ZFP36-ferroptosis axis as a key renal protective pathway in diabetic kidney disease

Abstract

Diabetic Kidney Disease (DKD) is strongly related to ferroptosis, an iron-dependent form of programmed cell death characterized by the accumulation of lipid peroxides. While ferroptosis is a pivotal mediator in DKD pathogenesis, its upstream regulatory mechanism remains poorly defined, thus impeding the development of targeted therapeutic strategies. In this study, by integrating multi-omics clinical data (GSE96804 and GSE104954) and combining machine learning algorithms, the RNA-binding protein ZFP36 was screened out as the core regulatory factor of ferroptosis in DKD. The results revealed that the expression level of ZFP36 in the DKD group was significantly lower than that in the normal group. Functional experiments demonstrated that overexpression of ZFP36 could significantly alleviate lipid peroxidation, iron ion accumulation, and cellular fibrosis, thereby inhibiting ferroptosis and alleviating kidney damage. Transcriptome analysis further revealed that ZFP36 regulated key genes related to oxidative stress and iron metabolism. Additionally, molecular docking simulations revealed strong binding affinity between ZFP36 and bioactive natural products such as berberine and astragalus, providing a potential mechanism for its renal protective effect. Overall, ZFP36 was hereby established as an important inhibitor of ferroptosis in DKD, highlighting its potential as both a biomarker and a therapeutic target for its precision diagnosis and intervention.

Graphical Abstract

1. Introduction

Diabetic Kidney Disease (DKD) represents a major global health concern closely associated with the increasing prevalence of diabetes mellitus [1, 2]. Long-term hyperglycemia and metabolic imbalance can trigger a series of pathophysiological changes in the kidneys [3-5]. Approximately 214 million adults worldwide are affected by DKD, accounting for 40% of the 537 million diabetes patients worldwide [6]. As the main cause of end-stage renal disease (ESRD), DKD accounts for over 50% of ESRD cases in developed countries [7]. China has an estimated 24 million DKD patients. Notably, since 2015, this population has suffered a 40% decline in quality of life, with their annual hospitalization rate rising by 12% year on year [7-10]. DKD is imposing a huge burden on both patients and the medical system.

Multiple factors contribute to the occurrence and development of DKD. Genetic susceptibility represents a key determinant, as specific gene polymorphisms can affect the susceptibility to DKD [11-13]. In addition, the courses of diabetes, blood sugar control levels, hypertension, and dyslipidemia are also known risk factors that accelerate kidney damage [14]. Their interaction forms a complex pathological network, posing challenges to the prevention and treatment of DKD [15]. Current treatment methods mainly focus on delaying the progression of the disease rather than targeting fundamental renal damage, highlighting an urgent need for novel treatment strategies [16].

Accumulating evidence shows that ferroptosis plays a core role in the DKD process [17-19]. Ferroptosis is a programmed cell death that relies on iron and is driven by lipid peroxidation. Its induction mechanisms include ROS generation and imbalance of the antioxidant system [20, 21]. Excess iron can catalyze the generation of ROS through the Fenton reaction, leading to lipid peroxidation of membrane polyunsaturated fatty acids [22]. In addition, inhibition of the cysteine-glutamic acid antitransporter (system xc-) can reduce the synthesis of glutathione (GSH), while glutathione peroxidase 4 (GPX4) inactivation can also promote the accumulation of lipid ROS, damage the integrity of the cell membrane, and cause cell death [23-29]. In DKD, hyperglycemia and metabolic abnormalities can induce excessive ROS generation beyond renal cells’ antioxidant capacity, inducing lipid peroxidation and ferroptosis that drive renal tubular epithelial cell damage, glomerular podocyte loss, and fibrosis [22, 30-32]. While the association between ferroptosis and DKD has been initially reported, the specific molecular mechanism remains unclear.

To fill this knowledge gap and identify potential therapeutic targets, this study integrated multi-omics data analysis with machine learning algorithms to screen out core genes closely related to ferroptosis in DKD. In vitro and in vivo experiments demonstrated ZFP36 as an important regulatory factor of ferroptosis in DKD, and its overexpression was observed to delay the process of ferroptosis. In addition, computer molecular docking studies suggested that natural compounds such as berberine and astragalus might exert renal protective effects by regulating ZFP36. Collectively, this study not only provides novel insights into the pathogenesis of DKD but also offers potential targets for precise targeted therapeutic interventions.

2. Methods

2.1 Data Acquisition

Ferroptosis-related genes (FRGs) associated with DKD were identified through integrative analysis of GEO datasets (GSE96804 and GSE104954). A systematic search revealed 7 candidate datasets, with GSE96804 (30 DKD/28 control) and GSE104954 (24 DKD/22 control) meeting the inclusion criteria (human renal tissues, DKD/control comparisons, and sample size ≥20 per group). We conducted differential expression analysis via limma (v3.56.0) and applied a linear mixed-effects model to adjust for disease status, age, sex, and batch effects.

2.2 Functional Annotation and Pathway Enrichment

For functional gene annotation, the clusterProfiler package was used for gene ontology (GO) analysis (covering biological processes, molecular functions, and cellular components). Through hypergeometric tests combined with Benjamini-Hochberg correction, significantly enriched items (FDR <0.05) were screened and visualized using bubble plots and circular phylogenetic trees.

To explore the biological pathways correlated with ZFP36 expression variation across samples, we performed Gene Set Enrichment Analysis (GSEA) and Gene Set Variation Analysis (GSVA). For GSEA, genes were ranked based on their expression correlation with ZFP36 across all samples. GSEA was then run using the KEGG pathway gene sets from the MSigDB collection to identify pathways enriched at either end of this correlation spectrum, with significance assessed using 1,000 permutations (nominal p-value <0.05, FDR <0.1). Meanwhile, for GSVA, we calculated single-sample pathway activity scores for each sample and subsequently analyzed the correlation between these scores and ZFP36 expression levels to identify pathways whose activity varied consistently with ZFP36 expression. The results of these analyses, highlighting pathways whose enrichment or activity shows a significant association with ZFP36 expression levels.

2.3 Machine Learning

Machine learning algorithms, including random forest (ensemble decision trees), SVM-RFE (feature selection via iterative elimination), and LASSO regression (variable selection with regularization), were employed to identify clinically relevant molecular signatures.

2.4 Animal Model

In the experiment, 8-week-old male C57BL/6 mice (GemPharmatech) were used. After 7 days of adaptive rearing under standard conditions, they were randomly divided into two groups: the control group was given a normal diet, and the intervention group was given a high-fat diet (Research Diets 12492, 60% fat calories). After 4 weeks of intervention, mice in the intervention group were intraperitoneally injected with streptozotocin (STZ, Sigma, 50 mg/kg/day, dissolved in citric acid buffer with pH 4.5) for 4 consecutive days; The control group was injected with an equal volume of normal saline (Randomization was performed using GraphPad Prism, with all mice housed in the same environment and fed the same diet).

Random blood glucose was measured using the OneTouch Verio^TM blood glucose meter on the 3rd, 7th, and 14th days after injection. If the blood glucose value was ≥16.7 mmol/L for three consecutive times, it was determined that the diabetes model was successfully established. All mice were sacrificed at 16 weeks of age, and kidney samples were taken for experimental analysis. All mice were anesthetized by intraperitoneal injection of pentobarbital sodium (50 mg/kg). After complete loss of consciousness and pain reflex cervical dislocation was performed by firmly securing the head and neck with one hand while swiftly and forcefully pulling the tail backward with the other. Death was confirmed by the absence of respiration heartbeat and corneal reflex. The experimental protocol was approved by the Animal Ethics Committee of Nanjing Medical University (IACUC-NJMU 2104014).

2.5 Immunohistochemical staining

After dewaxing with xylene and rehydration with gradient ethanol, the renal tissue sections were subjected to antigen repair using citric acid buffer (pH 6.0) by high-pressure thermal repair for 20 minutes. After cooling at room temperature, they were treated with 3% hydrogen peroxide for 10 minutes to block the endogenous peroxidase activity. Subsequently, 5% goat serum was used for blocking for 30 minutes and incubated overnight at 4°C with ZFP36 primary antibody (Proteintech #66737-1-Ig, 1:150) . After PBS washing, the sections were incubated with horseradish peroxidase-labelled secondary antibody for 60 minutes, and then stained with DAB, and finally rinsed with distilled water. Nuclear staining was performed with hematoxylin for 1-2 minutes, followed by bluing, gradient dehydration, xylene transparency and sealing with neutral resin.

2.6 Cell Culture and Treatment

HK-2 cells were maintained in low-glucose DMEM (Gibco; 5 mM glucose) supplemented with 10% heat-inactivated FBS (Gibco) and 1% penicillin–streptomycin (Gibco) under standard conditions (37°C, 5% CO₂, humidified). Medium renewal occurred every 48 hours to sustain optimal nutrient levels. The intervention group (n = 3) received high-glucose/palmitate treatment in 48h comprising 200 μM palmitic acid and 30 mM glucose (Adhere to the principle of biological triplicate).

2.7 Cell Transfection

ZFP36 overexpression plasmids (OE-ZFP36) and corresponding negative control (NC) vectors were synthesized by Corues Biotech (Nanjing, China). For transfection, HK-2 cells seeded in 6-well plates were transfected with 4 μg of OE-ZFP36 plasmid or NC plasmid mixed with Vazyme® LipoFiter 3.0 transfection reagent (Vazyme Biotech, China). The transfection complexes were maintained for 6 hours in serum-free medium before they were replaced with complete growth medium. Posttransfection incubation was continued for 24 hours under standard culture conditions.

2.8 Western blotting

Protein extraction was carried out using RIPA lysis buffer (containing protease inhibitors). Protein quantification was accomplished by the BCA method, with denaturation at 95°C for 10 minutes. Proteins were electrophoresed on a 10% SDS-PAGE gel (80 V 30 min → 120 V 65 min), and then transferred to a PVDF membrane by the wet transfer method (300 mA, 90 min). The membrane was sealed with 5% skimmed milk powder. After washing with TBST, it was incubated overnight with the primary antibody (FN1, α-SMA, GPX4, ACLS4, 1:1,000) at 4°C, and then incubated with HRP secondary antibody (1:10,000) at room temperature for 60 minutes. Signal detection was carried out using ECL reagents, and images were obtained under the ChemiDoc MP imaging system (Bio-Rad).

2.9 Quantitative Polymerase Chain Reaction

Comprehensive RNA extraction from both murine renal cortical tissues and cultured HK-2 cell monolayers was systematically executed via TRIzol reagent (Invitrogen) according to the manufacturer’s specifications, ensuring the preservation of RNA integrity throughout the isolation workflow. The resulting high-purity RNA was subjected to critical reverse transcription into complementary DNA (cDNA) via PrimeScript^TM RT Master Mix (Vazyme Biotech, Nanjing) under controlled thermal cycling conditions. Subsequent quantitative polymerase chain reaction (qPCR) analyses were conducted in optical 96-well plates via SYBR Green-based fluorescence detection chemistry, with amplification monitoring performed on the QuantStudio^TM16 Real-Time PCR platform (Applied Biosystems). Each experimental sample was analyzed in triplicate, with cycle threshold (Ct) values subjected to rigorous normalization against the endogenous reference gene GAPDH via the comparative ΔΔCt quantification methodology. The comprehensive sequences of primers used to support these analyses are cataloged in Supplementary Table 1.

2.10 ROS Assay

To assess the level of intracellular oxidative stress, a DCFH-DA fluorescent probe (Beyotime, China) was used in our experiments. For the specific operational process, the cells were first incubated with 10 μM DCFH-DA in serum-free medium at 37°C for 20 min. After the incubation was complete, the cells were washed three times with PBS to remove any residual substances that might interfere with the subsequent imaging. Finally, the EVOS Imaging System was utilized to image the cells, thereby enabling us to obtain relevant data and visualize the intracellular oxidative stress status.

2.11 Glutathione (GSH), Malondialdehyde (MDA), and Iron Assays

For our experimental setup, commercial assay kits (Beyotime, China) were used for the determination of MDA and GSH levels, and an iron assay kit (Proteinbio, China) was used for the measurement of iron content at 520 nm. In addition, all relevant assays were conducted on RIPA-lysed samples. To ensure the accuracy and reliability of the experimental results, the absorbance values obtained were normalized to the total protein concentration, which was quantified via a BCA assay (Thermo Fisher Scientific, USA).

2.12 Transcriptomic Sample Preparation and Data Analysis

The cells were categorized into two groups: a high-glucose/high-palmitate (HG + PA) group (30 mM glucose + 200 μM palmitate) and a ZFP36-overexpressing HG + PA group (HG + PA + OE-ZFP36), with three biological replicates each. Total RNA was isolated via TRIzol reagent, and qualified samples were processed for paired-end sequencing (Illumina NovaSeq 6000) by Biotree Biotechnology Company.

The quality of the raw sequencing reads was checked with FastQC. Differential expression analysis was conducted via DESeq2 (thresholds: |log2FC| >1, adjusted p < 0.05). Functional enrichment analysis (GO pathway enrichment) was performed via clusterProfiler (FDR < 0.05). Visualization included volcano plots (ggplot2), heatmaps, pathway enrichment diagrams (enrichplot), and Reactome plots.

2.13 Molecular Docking

The canonical SMILES structures of five bioactive phytoconstituents—calycosin (PubChem CID: 124044), berberine (CID: 2353), umbelliferone (CID: 5281426), glycyrrhizic acid (CID: 14982), and platycodin D (CID: 129716308)—were retrieved from PubChem. The predicted tertiary structure of the human ZFP36 protein (UniProt ID: Q07352) was obtained from the AlphaFold protein structure database (AF-Q07352-F1), as not experimentally determined PDB entry currently exists for this protein. The active sites were predicted via AutoDockTools, which defines a 25 ų binding region. Ligands underwent topology optimization (Gasteiger charges, rotatable bonds) prior to high-throughput docking with AutoDock Vina (exhaustiveness = 32). Binding poses were analyzed in PyMOL to map hydrogen bonds (≤3.5 Å) and hydrophobic interactions.

2.14 Statistical analysis

Statistical analyses were performed via GraphPad Prism. The data are presented as the means ± SEMs from ≥3 biological replicates, 95% confidence interval, with significance defined as p < 0.05 (two-tailed tests).

3. Results

3.1 Integrated Transcriptomics Identifies ZFP36 as a Key Ferroptosis-Related Gene in Human DKD

The study integrated and analyzed the GSE96804 and GSE104954 datasets, including a total of 89 patients, such as 38 controls and 51 cases of DKD. Batch effect correction was performed after data merging. Differentially expressed genes (DEGs) analysis showed that 72 genes were upregulated and 32 genes were downregulated in the DKD group (Fig. 1A–D). This unbiased transcriptome analysis clearly presents the abnormal gene expression of DKD, laying the foundation for targeted research on ferroptosis-related pathways.

Fig. 1. Differentially expressed genes.

A, B. Data distribution analysis pre- and post-batch effect correction. C. Volcano plot showing differentially expressed genes (DEGs) in DKD versus controls. D. A heatmap displaying DEG expression patterns across cohorts.

3.2 Ferroptosis-Associated DEGs Are Enriched in Lipid Metabolism and Oxidative Stress Pathways

The study combined the verified ferroptosis regulatory factors and DKD differentially expressed genes in the FerrDb database and we screened out 9 ferroptosis-related DEGs (p < 0.05, log2FC >1) (Fig. 2A). Box plot analysis showed that ACO1, EGR1, EMP1, GDF15, LOX, PDK4, PTGS2, S100A8 and ZFP36 were significantly downregulated in the DKD group, while the expressions of APOC1 and EMP1 were increased (Fig. 2B, C). GO functional enrichment analysis revealed that DEGs were involved in lipid and ketone body metabolism (biological processes), extracellular matrix of collagen content (cellular components), and fatty acid/carboxylic acid binding (molecular functions) (Fig. 2D), suggesting that iron metabolism and REDOX homeostasis imbalance are the core pathological features of DKD.

Fig. 2. Enrichment analysis of ferroptosis related genes.

A. Venn diagram of overlapping differentially expressed genes (DEGs) and ferroptosis-related genes (9 shared). B. The changes of the differential genes in DKD condition. C. Network analysis of enriched ferroptosis-associated pathways, with node size indicating gene count and color intensity reflecting enrichment significance (–log10(adjusted p-value)). D. GO enrichment analysis (biological processes): bar length represents enriched gene numbers; dot color denotes –log10(*p*-value).

3.3 Machine Learning Algorithms Converge on ZFP36 as a Central Hub Gene in DKD-Associated Ferroptosis

The study adopted multiple algorithms such as LASSO-Cox, SVM-RFE and Random Forest (RF) for comprehensive analysis. LASSO-Cox screened EGR1, EMP1, S100A8, LOX and ZFP36 (Fig. 3A), SVM-RFE obtained the extended feature set (Fig. 3B), and RF identified PDK4, ZFP36, EMP1 and GDF15 (Fig. 3C). ROC curve analysis showed that the AUC of ZFP36 was 0.949, with the best predictive performance (Fig. 3D). All three algorithms positioned ZFP36 as the core hub of the ferroptosis network (Fig. 3E), indicating its significant biological role in DKD.

Fig. 3. Machine learning-based screening of ferroptosis-related genes.

A. LASSO regression for feature selection. B. SVM-RFE ranking key genes. C. Random Forest importance scoring. D. Analysis of ROC curves for differentially expressed genes. E. Venn diagram showing the intersected gene ZFP36 between LASSO, SVM-RFE and RF algorithms.

3.4 ZFP36 Expression Correlates with Metabolic and Fibrotic Pathway Dysregulation in DKD

To investigate the global pathway landscape associated with ZFP36 expression variation in DKD, we performed Gene Set Enrichment Analysis (GSEA) using ZFP36 expression as a continuous ranking variable.

Pathways exhibiting negative enrichment were significantly correlated with lower ZFP36 expression. These included core metabolic and antioxidant processes such as “Glutathione metabolism,” “Citrate cycle (TCA cycle),” “Oxidative phosphorylation,” and multiple amino acid degradation pathways (Fig. 4A). Conversely, pathways showing positive enrichment were associated with higher ZFP36 expression and involved “Focal adhesion,” “ECM-receptor interaction,” and “TGF-beta signaling” (Fig. 4B). Gene Ontology analysis yielded consistent results, with terms related to fatty acid oxidation and oxidoreductase activity being inversely correlated with ZFP36 expression.

Fig. 4. Pathway analyses associate ZFP36 expression with a pro-ferroptotic molecular landscape in DKD.

A. Representative pathways that are enriched in samples with lower ZFP36 expression. B. Representative pathways that are enriched in samples with higher ZFP36 expression. C. Bar plot summarizing the top KEGG pathways whose enrichment shows a significant correlation with ZFP36 expression levels. Pathways on the left (green, negative NES) are enriched in contexts of lower ZFP36 expression, while pathways on the right (blue, positive NES) are enriched with higher ZFP36 expression. D. Bar plot of the top GO terms correlated with ZFP36 expression. Reduced ZFP36 expression is linked to downregulated metabolic/antioxidant pathways and upregulated proliferative/fibrotic pathways, fostering a cellular environment prone to ferroptosis.

Gene Set Variation Analysis (GSVA) further supported these findings, revealing a significant association between lower ZFP36 expression and decreased activity of antioxidant/metabolic pathways, alongside increased activity of fibro-proliferative pathways (Fig. 4C, D).

Collectively, these analyses indicate that variation in ZFP36 expression, particularly its reduction, is associated with a molecular signature marked by impaired antioxidant and metabolic capacity coupled with enhanced fibrotic signaling, a cellular state conducive to ferroptosis.

3.5 ZFP36 Expression Is Consistently Downregulated Across in vitro and Clinical DKD Models

Clinical data analysis (GSE96804/GSE104954) revealed a significant downregulation of ZFP36 in the renal tissues of DKD patients (Fig. 5A). In HK-2 cells treated with high glucose/palmitic acid, the expression of ZFP36 was also significantly decreased (Fig. 5B). The STZ-induced DKD mouse model showed that ZFP36 was significantly downregulated at the mRNA and protein levels, and immunohistochemical staining also indicated a decrease in the intensity of ZFP36 in renal tissue (Fig. 5C, D), indicating that the downregulation of ZFP36 in DKD is a common pathological feature.

Fig. 5. ZFP36 expression is downregulated in DKD.

A. ZFP36 was downregulated in DKD patients. B. Western blot analysis of ZFP36 in renal tissues from DKD model mice. C. Western blot analysis of ZFP36 in HG + PA-induced HK-2 cells. D. Representative immunohistochemical staining showing ZFP36 in renal tissues from DKD model mice. *p < 0.05, **p < 0.01, ***p < 0.001 compared with NC group.

3.6 HG + PA Treatment Induces Ferroptosis In Vivo/In Vitro

To directly confirm that our HG + PA modeling strategy successfully induces ferroptosis, we evaluated a comprehensive panel of ferroptosis-specific hallmarks, with FER-1, a well-recognized gold-standard ferroptosis inhibitor, used to validate the specificity of the cell death process. Western blot analysis demonstrated that compared with the NC group, the HG + PA model exhibited a significant downregulation of the key ferroptosis suppressor GPX4 and a robust upregulation of the ferroptosis driver ACSL4 (Supplementary Fig. 1A). These alterations are canonical signatures of ferroptotic commitment. Consistently, HG + PA treatment triggered marked oxidative stress, as evidenced by the dramatic accumulation of intracellular ROS (Supplementary Fig. 1B). Further biochemical quantification revealed that HG + PA induced iron overload, enhanced lipid peroxidation (MDA level), and exhausted the key antioxidant GSH—all core mechanistic features of ferroptosis (Supplementary Fig. 1D). Notably, co-treatment with FER-1 completely reversed all the above HG + PA-induced changes: it restored GPX4 expression, suppressed ACSL4 levels, abrogated ROS accumulation, normalized iron content, reduced MDA levels, and replenished GSH (Supplementary Fig. 1A–D). Collectively, these findings provide direct, multi-dimensional evidence that our HG + PA modeling strategy effectively induces ferroptosis, and the reversal of this phenotype by the gold-standard inhibitor FER-1 confirms the bona fide induction of ferroptotic cell death.

3.7 ZFP36 Overexpression Confers Protection Against Fibrosis and Ferroptosis In Vitro

After overexpression of ZFP36 (OE-ZFP36) in HK-2 cells, cells treated with high glucose/palmitic acid showed decreased fibrosis markers (FN1, α-SMA) and lipid peroxidation mediator ACSL4, and increased antioxidant enzyme GPX4 (Fig. 6A). ROS generation (Fig. 6B), iron load (Fig. 6C), and lipid peroxidation (MDA, Fig. 6D) were significantly reduced, and the level of glutathione (GSH) recovered (Fig. 6E). These experiments verified that ZFP36 plays a causal role in inhibiting ferroptosis and renal tubular fibrosis.

Fig. 6. ZFP36 overexpression mitigates fibrosis and ferroptosis in HK-2 cells.

A. Western blot analysis of fibrosis markers (FN1, α-SMA) and ferroptosis-related proteins (GPX4, ACSL4) in NC, HG + PA + OE-ZFP36, and HG + PA + OE-CON groups (normalized to HSP90). B–E. Quantification of ROS (B), iron content (C), MDA (D), and GSH (E) levels. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 vs. NC.

3.8 Transcriptomic Profiling Reveals ZFP36 Modulation of Oxidative Stress and Iron Homeostasis Networks

To better understand how ZFP36 affects ferroptosis in DKD, we performed a transcriptomic analysis of control (CON) and ZFP36-overexpressing (OE-ZFP36) HK-2 cells. Among the 1,685 DEGs identified, 791 genes were significantly increased, and 894 genes were markedly decreased (Fig. 7A, B). GO analysis revealed significant enrichment in “cellular response to oxidative stress” (GO:0034599) and “iron ion binding” (GO:0005506), while Reactome pathway analysis indicated alterations in “fatty acyl-CoA biosynthesis” and “iron uptake and transport” (Fig. 7C, D). We further screened for genes involved in ferroptosis and identified 19 ferroptosis-related differentially expressed genes, which could be functionally categorized into oxidative stress regulators (TXNIP, ACO1) [33, 34], lipid metabolism mediators (SQLE, HMGCR) [35, 36], iron transport modulators (SLC3A2, SLC4A4) [37], and cell cycle controllers (CDKN1A, CDKN2AIP) [38, 39]. Experimental validation via qPCR in high-glucose- and palmitate-treated cells revealed significant EPAS1 downregulation, accompanied by elevated SLC3A2, ACO1 and TXNIP expression, and these changes were alleviated to some extent by the overexpression of ZFP36, whereas SLC4A4 tended to be upregulated in DKD conditions and was alleviated after the overexpression of ZFP36, although the difference was not statistically significant. (Fig. 7E–I). The broad regulatory impact of ZFP36 on genes involved in iron handling, lipid metabolism, and oxidative stress response suggests it functions as a master coordinator of cellular homeostasis pathways disrupted in DKD.

Fig. 7. Transcriptome analysis of ZFP36-regulated genes.

A. A heatmap of differentially expressed genes in CON vs. OE-ZFP36 cells. B. Volcano plot showing DEGs in CON vs. OE-ZFP36 cells. C. GO enrichment (biological process). D. Enriched GO terms. E–I. qPCR analysis of the mRNA levels of ferroptosis-related genes in HK2 cells. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 vs. NC.

3.9 Molecular Docking Predicts High-Affinity Binding Between ZFP36 and Nephroprotective Compounds

Molecular docking showed that ZFP36 could highly affinity bind with berberine, astragaline, glycyrrhizic acid, umbelliferone and platycoside D (ΔG < –5 kcal/mol), among which platycoside D had the highest binding stability (ΔG = –8.2 kcal/mol) (Supplementary Fig. 2B). The combination mode shows hydrogen bonds, π-π stacking and hydrophobic interaction, suggesting that natural compounds may regulate ferroptosis through ZFP36 to achieve renal protection (Supplementary Fig. 2A–C).

4. Discussion

DKD constitutes one of the main causes of end-stage renal disease worldwide. Its occurrence and development involve a variety of complex mechanisms, including metabolic disorders, inflammatory responses, oxidative stress, fibrosis, etc. [40]. Despite substantial progress achieved in blood glucose and blood pressure control, the current clinical diagnostic and therapeutic approaches remain highly dependent on late-stage indicators such as decreased eGFR and elevated urinary albumin, lacking precise capture of early molecular pathological changes [41]. This typically delays the diagnosis of DKD until the disease has progressed markedly, leading to the missed opportunity of the optimal early intervention window. Concurrently, existing treatment methods are mostly centered on delaying disease advancement, with little capacity to reverse the kidney damage that has already occurred. Furthermore, the commonly used animal models are still difficult to fully reproduce the chronic and multifactorial characteristics of human DKD, further complicating the early diagnosis and targeted intervention [42]. Therefore, seeking new molecular targets and mechanism bases holds considerable significance for improving the early diagnosis rate of DKD and developing novel therapeutic strategies.

In recent years, ferroptosis (an iron- and lipid peroxidation-dependent programmed cell death pathway) has been increasingly identified as a critical mediator in multiple metabolic diseases. Studies have also uncovered that ferroptosis can directly lead to dysfunction and death of renal tubular cells by promoting ROS accumulation and membrane lipid damage, thereby promoting the occurrence and development of DKD [32, 43, 44]. However, the upstream regulatory mechanism of ferroptosis in DKD is not yet fully understood, and the dominating role of RNA-binding proteins in this process still lacks systematic research. This study, through multi-omics analysis and in vitro and in vivo experimental verification, for the first time revealed the mechanism of action of ZFP36 as a core regulatory factor of ferroptosis in DKD, enriching the molecular map of ferroptosis in kidney diseases.

ZFP36 belongs to the TIS11 family of RNA-binding proteins and is known to be involved in the regulation of inflammatory response, cell cycle, and stress response by binding to mRNA degradation mediated by AU-rich element (ARE) [45, 46]. Previous studies have reported that ZFP36 can exert anti-inflammatory effects by down-regulating the stability of inflammatory factors such as TNF-α and IL-6, while simultaneously exhibiting dual regulatory functions in tumorigenesis [46]. The innovation of this study lies in identifying ZFP36 as a key regulator of ferroptosis in the context of DKD. The present results demonstrated that ZFP36 was significantly downregulated in both DKD renal tissues and cell models induced by high glucose/lipid toxicity. Its absence aggravated iron homeostasis disorder and lipid peroxidation, while ZFP36 overexpression could effectively alleviate ROS accumulation and ferroptosis phenotypes. This not only suggested that ZFP36 might be an upstream regulatory factor of ferroptosis but also provided innovative clues for understanding the early molecular pathological changes of DKD.

Further bioinformatics analysis and GSEA/GSVA results indicated that the gene set regulated by ZFP36 was enriched in pathways related to REDOX homeostasis, iron ion metabolism, and cellular stress response. Downstream target genes EPAS1, SLC3A2, ACO1, and TXNIP were all observed to exert significant regulatory effects on both ferroptosis and energy metabolism processes. Previous studies confirmed ACO1 (also known as IRP1) as a core regulatory factor in iron metabolism, and its inactivation was found to lead to excessive iron accumulation and ROS generation. TXNIP was also closely related to oxidative stress and mitochondrial dysfunction, generally upregulated in diabetic complications [47-49]. Therefore, ZFP36 might maintain the iron homeostasis and antioxidant capacity of renal tubular cells at the global level by coordinating the expression of multiple downstream targets, thereby inhibiting ferroptosis occurrence. This post-transcriptional regulation, iron metabolic homeostasis, and the proposal of the cell death chain could jointly provide a new mechanism perspective for explaining the complex pathological network of DKD (Graphical Abstract).

Graphical Abstract.

Notably, this study revealed the role of ZFP36 in ferroptosis in DKD and also predicted potential natural compound intervention pathways through molecular docking. Results showed that components such as berberine, astragaloin, and glycyrrhizic acid could form stable hydrogen bonds and hydrophobic interactions with ZFP36, confirming the renal protective effects of these active ingredients of traditional Chinese medicine by regulating ZFP36 activity. This discovery provides both experimental evidence for traditional Chinese medicine molecular target research and new perspectives for small-molecule drug development. In the context of the current lack of specific ferroptosis inhibitors in clinical practice, the ZFP36-targeted intervention strategy based on natural products is feasible and holds substantial application potential.

From a clinical perspective, the results of this study carry dual implications: On the one hand, the level of ZFP36 is expected to serve as an early biomarker for risk stratification and efficacy prediction. Given that its expression is downregulated in the early stage of DKD, it is more sensitive and specific compared to traditional indicators. On the other hand, restoring or simulating the function of ZFP36 may become a new therapeutic strategy to slow the progression of DKD by inhibiting ferroptosis. This offers the possibility for future precision medical management, such as monitoring ZFP36 levels in high-risk populations and intervening with ferroptosis inhibitors, which is anticipated to slow down disease progression and reduce end-stage renal disease incidence.

Undoubtedly, this study still has several limitations. Firstly, the clinical sample size is limited, and the results should still be verified on a larger scale and in multiple centers to ensure universality. Secondly, the present research mainly relies on transcriptome analysis and in vitro models, lacking direct verification of human renal tissue function. Thirdly, the specific molecular targets and action networks of ZFP36 in regulating ferroptosis still warrant further analyses through techniques such as RIP-seq or CLIP-seq. Finally, despite the molecular docking data of natural compounds pointing to potential binding patterns, no systematic in vivo pharmacodynamic or toxicological assessments have been conducted to date, which calls for additional verification in animal models and preclinical investigations.

Future research should prioritize three directions: Firstly, explore the sensitivity and specificity of ZFP36 in the early diagnosis of DKD, and evaluate its detection value in non-invasive samples such as blood or urine. Secondly, further verify the causal role of ZFP36 in ferroptosis using gene editing technologies such as CRISPR/Cas9 and animal models, and clarify its contribution to DKD progression. Thirdly, based on the molecular docking results, small molecule compounds should be systematically screened and optimized to search for efficient and specific ZFP36 agonists or mimics, providing candidate drugs for clinical transformation. Additionally, integrating multi-omics data such as proteomics, metabolomics, and single-cell transcriptomics could facilitate the elucidation of the specific functions of ZFP36 in different types of kidney cells, further promoting the application of precision medicine.

In summary, through multi-omics data integration, in vivo and in vitro functional experiments, and computational prediction, this study for the first time revealed the key role of ZFP36 in the regulation of ferroptosis in DKD. Down-regulation of ZFP36 could lead to iron homeostasis imbalance and ROS accumulation, thereby aggravating renal tubular injury, while its overexpression could significantly alleviate ferroptosis and renal pathological changes. Molecular docking analysis suggested that natural compounds such as berberine and astragalus could bind to ZFP36 with high affinity, providing a structural basis for drug development based on ZFP36. Taken together, these findings not only expand the understanding of the ferroptosis mechanism in DKD but also suggest the dual potential of ZFP36 as both a diagnostic biomarker and a therapeutic target, providing new insights and forging experimental foundations for the future precise medical management of DKD.

Ethical approval and consent to participate

Approval under NJMU-IACUC #2104014. All procedures followed the ARRIVE guidelines, with maximal efforts to minimize animal suffering.

Consent for publication

All the authors have read and approved the final version of this manuscript. We confirm that this work is original and has not been published elsewhere.

Data availability

This study analyzed the following public datasets: phytochemical structures (PubChem CIDs: 124044,2353,5281426,14982,129716308), ZFP36 structures (AlphaFold DB: AF-Q07352-F1), and the transcriptomic profiles GSE96804 and GSE104954 from the GEO database. The experimental data are available upon request to the corresponding author.

Disclosure

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Supplementary Material

Supplementary Table 1

Primer sequence.

Supplementary Figs.