HK-2 cells (Cell Bank of China Academy of Sciences) were cultured in DMEM/F12 (Gibco) supplemented with 10% FBS (Gibco) and 1% penicillin/streptomycin (MilliporeSigma). MPC-5 cells (Procell) were maintained in RPMI 1640 (Procell) containing 10% FBS (Gibco) and 1% penicillin/streptomycin (MilliporeSigma). Both cell lines were incubated at 37 °C under 5% CO₂. Upon reaching 50% confluency, cells were treated with 30 mM high-glucose (HG; MilliporeSigma) for 72 h. Mannitol (MilliporeSigma) served as an osmotic control. LPS (Solarbio) was dissolved in sterile PBS to prepare a 10-mg/mL stock solution. HK-2 cells (10 μg/mL) and MPC5 cells (45 μg/mL) were treated with these LPS solutions for 24 h to model cell injury. Cells were subsequently harvested for Western blot analysis of target protein expression.

We employed four parallel wells for each group of HK-2 cells: HK-2 cells transfected with the control lentivirus were cultured in a high-glucose medium containing 30 mM glucose (HG-NC) for 72 h, while a separate group of control lentivirus-transfected HK-2 cells was cultured in a medium containing 30 mM mannitol to serve as an osmotic control (OS-NC). Similarly, HK-2 cells transfected to overexpress the hnRNP-F lentivirus (Gene ID:98758, Lentiviral expression vector LV5) were maintained in a 30 mM glucose medium (HG-OE) and in a mannitol medium (OS-OE) for 72 h, respectively. MPC5 cells were cultured under HG conditions (30 mM, 72 h), mannitol treatment (30 mM, 72 h), or LPS stimulation (10 μg/mL, 24 h) and then transfected with an hnRNP-F overexpressing plasmid packaged in a lentiviral vector (Gene ID: 98758, vector name: HBLV-ZsGreen-PURO).

Total RNA was isolated from the renal cortex and cells separately using the TRIzol method, and 1 μL of total RNA was used as the template, reverse-transcribed to cDNA, and continued to be amplified by using cDNA as the template, sequentially, at 95 °C for 3 min, 1 cycle, 95 °C for 10 s, and 62 °C for 40 s, for a total of 40 cycles. The mRNA levels of HG-NC and HG-OE were determined using the 2^−ΔΔCT method, with β-actin and GAPDH serving as internal references. Similarly, mRNA levels in renal tissues of db/db and db/m mice were calculated. The specific primer sequences are shown in Table 1.

After extracting the proteins from the cells OE-hnRNP-F and the NC HK-2 cells, the lysates were pre-cleared with rabbit IgG (3 μg/mg protein) and protein A/G magnetic beads. Then, they were incubated overnight at 4 °C with anti-hnRNP-F or control IgG (3 μg/mg protein). The complexes were captured with fresh magnetic beads (20 μL/500 μL lysate, room temperature for 2 h), washed three times with lysis buffer, and eluted in 1× Laemmli buffer (95 °C, 5 min) for immunoblotting.

Total protein lysates (30 µg/lane) from cells or renal tissues were separated by 10% SDS-PAGE and transferred to PVDF membranes (Millipore). After blocking with 5% non-fat milk/TBST for 1 hour, the membranes were incubated overnight at 4°C with the following primary antibodies:hnRNP-F, β-actin, GAPDH (1:5000; Proteintech, 67701-1-Ig, 20536-1-AP, 60004-1-Ig), p-p65 (1:500; Invitrogen, MA5-15160), p65 (1:5000; Abclonal, A19653) and TNF-α (1:1000; Abcam, ab183218). Following TBST washes (3 × 10 min), membranes were incubated 1 h with HRP-conjugated secondary antibodies: Goat anti-mouse, Goat anti-rabbit (1:1000; Proteintech, SA00001-1, SA00001-2). Signals were detected by ECL (Proteintech, P0018S) and quantified using ImageJ (NIH, v1.53e), normalized to β-actin/GAPDH.

All RNA was processed with RQ1 DNase (Promega) to remove DNA. The quality and quantity of the purified RNA were determined by measuring the absorbance at 260 nm/280 nm (A260/A280) utilizing SmartSpec Plus (BioRad). RNA integrity was further verified by 1.5% agarose gel electrophoresis.

For each sample, 1 μg of total RNA was used for RNA-seq library preparation. mRNAs were captured by VAHTS mRNA capture beads (Vazyme, N401). The purified RNA was treated with RQ1 DNase (Promega) to remove DNA before being used for directional VAHTS with a Universal V8 RNA-seq Library Prep Kit for Illumina (NR605). Polyadenylated mRNAs were purified and fragmented. Fragmented mRNAs were converted into double-stranded cDNA. Following end repair and A tailing, the DNAs were ligated to Adapter (N323). After purification of the ligation product and size fractioning to 300–500 bps, the ligated products were amplified and purified, then quantified and stored at −80 °C before sequencing. The strand marked with dUTP (the second cDNA strand) is not amplified, allowing strand-specific sequencing.

For high-throughput sequencing, the libraries were prepared following the manufacturer’s instructions and applied to an Illumina NovaSeq 6000 system for 150-nt paired-end sequencing.

First, raw reads containing more than 2-N bases were discarded. Then, adapters and low-quality bases were trimmed from raw sequencing reads using FASTX-Toolkit (Version 0.0.13). The short reads less than 16 nt were dropped as well. Afterward, clean reads were aligned to the GRCh38 genome by HISAT2 (Kim et al., 2015), allowing four mismatches. Uniquely mapped reads were used for gene read number counting and FPKM calculation (fragments per kilobase of transcript per million fragments mapped) (Trapnell et al., 2010).

The R Bioconductor package DESeq2 was applied to screen out the differentially expressed genes (DEGs) (Love et al., 2014). The P-value for correction <0.05 and fold change ≥2 or ≤0.5 were set as the cut-off criteria for identifying DEGs.

To minimize potential batch effects and technical variability in RNA-seq data, we applied ComBat_seq, an empirical Bayes method implemented in the “suva” R package, to adjust for known batch information across samples while preserving biological variance. Prior to batch correction, principal component analysis (PCA) was performed to visualize sample clustering and assess batch-related variation. After correction, PCA and hierarchical clustering confirmed improved consistency within experimental groups.

The AS events and regulated alternative splicing events (RAS) between OE-hnRNP-F and NC samples were defined and quantified by using the splice sites usage variation analysis (SUVA) pipeline as described previously. Differential splicing of each pair of cells was analyzed. The frequency and reads proportion of the SUVA AS event (pSAR) of each AS event were calculated. For alternative splicing validation, we performed RT-qPCR on independent samples (n = 3 biological replicates) to confirm SUVA predictions, reporting both p-values and AS ratios in supplementary GraphPad data.

In order to sort out functional categories of DEGs, Gene Ontology (GO) terms and KEGG pathways were identified using the KOBAS 2.0 server (Xie et al., 2011). The hypergeometric test and the Benjamini–Hochberg FDR controlling procedure were used to define the enrichment of each term.

GSEA is an analytical method for genome-wide expression profile microarray data. By comparing genes with predefined gene sets, it can identify functional enrichment. A gene set means a group of genes sharing localization, pathways, functions, or other features. GSEA was conducted using the clusterProfiler package (version 4.6.2). The fold change of gene expression between the Mets group and the Primary group was calculated, and the gene list was generated in accordance with the change of |log2FC|. Afterward, we utilized GSEA-based enriched HALLMARK gene sets of the Molecular Signature Database.

Public sequence data files of CLIP-seq data of hnRNP-F in human 293T cells from GSE34993 were downloaded from the Sequence Read Archive (SRA). After reads were aligned onto the genome, only uniquely mapped reads were used for the following analysis. The “ABLIRC” strategy was used to identify the binding regions of RBP on the genome (Xia et al., 2017). Reads with at least 1-bp overlap were clustered as peaks. For each gene, computational simulation was used to randomly generate reads with the same number and lengths as reads in peaks. The output reads were further mapped to the same genes to generate random max peak heights from overlapping reads. The whole process was repeated 500 times. All the observed peaks with heights higher than those of random max peaks (*P < 0.05) were selected. The target genes of hnRNP-F were finally determined by the peaks, and the binding motifs were called by HOMER software (Heinz et al., 2010).

Principal component analysis (PCA) was performed by the R package factoextra (https://cloud.r-project.org/package=factoextra) to show the clustering of samples with the first two components. After controlling the reads by tags per million TPM) of each gene in samples, an in-house script (sogen) was used for visualization of next-generation sequence data and genomic annotations. The pheatmap package (https://cran.r-project.org/web/packages/pheatmap/index.html) in R was used to perform the clustering based on Euclidean distance.

Seven-week-old male db/db mice (C57BLKS/J background, 12 weeks old, mean body weight: 45.2 ± 3.1 g) and their db/m littermates were purchased from GemPharmatech Co., Ltd. (Chengdu, China) and maintained in the specific pathogen-free (SPF) animal facility at Hubei University of Chinese Medicine (Wuhan, China). All experimental procedures involving animals were performed in strict compliance with the institutional guidelines and approved by the Animal Ethics Committee of Hubei University of Chinese Medicine (Approval No. HUCMS00303837). DKD modeling success was defined by (1) fasting blood glucose ≥16.7 mmol/L for three consecutive tests, (2) urine output >150% of controls, and (3) persistent proteinuria (Wang et al., 2021).

All results are presented as the average value plus or minus the standard deviation (SD). Statistical analyses were performed using GraphPad Prism 10.1.2 software (GraphPad, San Diego, CA). Differences between experimental groups were evaluated using either a paired two-tailed Student’s t-test or one-way ANOVA followed by Bonferroni’s post hoc test for multiple comparisons. A P-value of ≤0.05 was considered statistically significant.

In cells treated under normal glucose and hypertonic conditions, there was no significant difference in hnRNP-F protein levels between the two groups (P > 0.05). However, in HK-2 cells cultured with HG concentrations, hnRNP-F levels were significantly reduced (**P < 0.01) (Figures 1A–C). Western blot analysis demonstrated that high-glucose downregulated hnRNP-F expression. Mannitol, used as an osmotic control, exhibited no significant effect on hnRNP-F gene expression. Consistent with protein-level observations, RT-qPCR confirmed significant upregulation of hnRNP-F mRNA in high-glucose-treated HK-2 cells (*P < 0.05 vs. control) (Figures 1D,E), while treatment with equiosmolar mannitol showed no comparable effect (P > 0.05).

Overexpression of hnRNP-F was constructed in HK2 cells cultured in a HG environment. qPCR results showed that hnRNP-F gene expression levels were upregulated in HK-2 cells after infection with overexpression of the hnRNP-F lentivirus (Figure 2A). To comprehensively investigate the hnRNP-F-mediated transcriptional regulation in high-glucose (HG) conditions, we constructed cDNA libraries prepared from control and hnRNP-F-overexpression cells (three biological replicates), which were incubated in high-glucose and mannitol. After removing adapters and contaminating sequences, we obtained a total of 742.7 million high-quality reads from each sample (Supplementary Table S2). Approximately 91.2%–96.39% paired-end reads per sample were then aligned to the human GRCH38 genome. RNA-seq yielded robust expression for 18,051 genes (Supplementary Table S3). PCA was carried out and revealed excellent clustering of expression gene changes between the OE-hnRNP-F HK-2 cells and control samples for different treatment conditions (Figure 2B). To evaluate the dynamics of gene expression between OE-hnRNP-F vs. Ctrl cells, we compared expression among all pairwise combinations of the samples using DESeq2 (Love et al., 2014).

OE-hnRNP-F affects many gene expressions under HG conditions. A total of 890 DEGs were obtained (P-value of <0.05, fold change ≥2, or ≤0.5, FDR ≤0.05), of which 568 genes were upregulated and 322 genes were downregulated (Supplementary Table S4). Mannitol treatment was found to affect the expression of some genes (Supplementary Table S5), while the number of DEGs was larger in HK-2 cells treated with high-glucose than in HK-2 cells treated with mannitol. A Venn diagram illustrating the profiles of DEGs reveals an overlap between glucose and mannitol treatment (Figure 2C). This analysis displays unique and overlapping sets of DEGs in hnRNP-F overexpressing cells under HG treatment. Venn diagram analysis revealed an intersection of 118 genes between the high-glucose-treated and mannitol-treated OE-hnRNP-F HK-2 cells. Mannitol did not drastically affect overall gene expression when used as an osmolar control treatment.

To correlate the hnRNP-F-regulated gene expression and biological functions under high-glucose, we subjected all 890 DEGs to GO annotation (Supplementary Tables S6, S7). In the biological processes (BPs) of GO analysis, the upregulated genes in the OE-hnRNP-F samples were highly enriched in the “extracellular matrix organization” and “cell adhesion” processes (Figure 2D). The downregulated genes were mainly enriched in the inflammatory response and other related biological processes, as well as “regulation of insulin secretion,” “response to ischemia,” “regulation of insulin secretion,” positive regulation of angiogenesis,” and “response to hypoxia,” which are closely related to the pathogenesis of DKD (Figure 2E; Supplementary Table S7). Among these, the decreased expression of inflammation-related genes was of particular interest. Representative genes from inflammatory-related genes (CXCL8, IL6, GDF15, PTX3, and TFPI2) were selected for RT-qPCR validation of their mRNA levels. CXCL8 and IL6 were found to be enriched in the tumor necrosis factor pathway (Supplementary Table S6). KEGG pathway enrichment analysis was also performed (Supplementary Figures S1A,B; Supplementary Table S8). The downregulated genes were also enriched in the TNF signaling pathway (Supplementary Figure S1B). The GO enrichment pathway of differentially expressed genes following mannitol treatment in HK-2 cells with hnRNP-F overexpression differs from that observed under HG conditions (Supplementary Figures S1C,D). Compared with HK-2 cells treated with HG and transfected with the empty vector (HG-NC group), overexpression of hnRNP-F could significantly downregulate the expression of CXCL8, IL6, GDF15, PTX3, and TFPI2 (****P < 0.0001). The qPCR results were consistent with RNA sequencing data (Figure 2F). The primers are listed in Supplementary Table S1. In the hyperosmotic mannitol control condition, the overexpression of hnRNP-F resulted in the significant downregulation of only two genes, IL6 and GDF15. This suggests that the upregulation of hnRNP-F under normoglycemic conditions did not exert any significant effect. However, under hyperglycemic conditions, it led to a marked reduction in the expression of genes associated with the inflammatory response, particularly those involved in the TNF signaling pathway and the pathogenesis of DKD.

We performed GSEA of genes differentially expressed upon hnRNP-F overexpression under high-glucose. The overexpression of hnRNP-F resulted in a significant inhibition of the TNFα/NFκB signaling pathway (Figure 3A). As demonstrated in Figure 3B, the upregulation of hnRNP-F resulted in a notable decrease in the expression of PTX3 and IL6, both of which are genes associated with the TNFα/NFκB signaling pathway. Even in the hypertonic control of mannitol, overexpression of hnRNP-F inhibited the expression of these genes. The findings indicated that the upregulation of hnRNP-F suppressed the transcription of genes associated with the TNF signaling pathway, such as CXCL8, IL6, and PTX3.

Given the multi-functional nature of hnRNP-F as an RNA-binding protein, our analysis also encompasses the impact of hnRNP-F on the regulation of AS events. We obtained a total of 125 million uniquely mapped reads from each sample, in which 33.15%–44.04% were junction reads (Supplementary Table S2). We analyzed the RNA-seq data using the software SUVA (9). Our analysis revealed the presence of distinct alternative splicing events (ASEs) and regulated alternative splicing events (RASEs) among the OE-hnRNP-F and control cells in the HG and mannitol-treated groups. Specifically, alt3p and alt5p were the main ASEs and RASEs between OE-hnRNP-F and control cells (Figure 4A; Supplementary Figure S3A; Supplementary Table S9. The SUVA-identified ASEs corresponded to the classical ASEs. hnRNP-F overexpression processing in HG cultured HK-2 cells resulted in a large number of differential variable splicing events, with a total of 1,158 significant RASEs detected. The main variable splicing event types included 65 3pMXE, 66 5pMXE, 230 A3SS, 27 A3SS&ES, 177 A5SS, 52 A5SS&ES, 261 ES, 4 IntronR, 67 MXE, and 209 cassette exons (Figure 4B; Supplementary Table S9). Mannitol treatment mainly affects the AS events of A3SS and ES (Supplementary Figure S3B; Supplementary Table S9). Figure 3C illustrates the presence of novel splicing events among the RASEs under HG conditions, a finding that aligns with the results observed following mannitol treatment (Supplementary Figure S3C). These findings suggest that the overexpression of hnRNP-F can modulate intracellular alternative splicing in response to hypertonic treatment conditions.

Due to a splicing event involving two transcripts, which may account for a very small proportion of the entire gene expression, our study focused on identifying the more dominant transcripts in splicing events. We specifically quantified the number of splicing events with varying proportions of RASEs in the region covered by all reads. We also excluded splicing events with a low proportion (pSAR<50%). A total of 687 events with pSAR>50% were detected in overexpression hnRNP-F cells treated with high-glucose (Figure 4D). A total of 611 events were detected in the cells treated with mannitol (Supplementary Figure S3D). While a diversity of AS events was observed in cells treated with mannitol, the functional disparities in these splicing events between the two groups were pronounced when compared to the HG group. Notably, only 36 AS events were common to both treatment conditions (Figure 4E). This result indicated that overexpressed hnRNP-F HK-2 cells have a distinct AS profile in response to HG exposure. Gene Ontology biological process (GO-BP) enrichment analysis revealed that the HG group exhibited significant alterations in alternative splicing, predominantly enriched in pathways related to “regulation of RNA splicing” and “RNA splicing” (Figure 4F). The AS events of the mannitol group were mainly enriched in the “microvillus assembly” and “epithelial tube formation” pathway (Supplementary Figure S3E). To validate the accuracy of the predicted hnRNP-F-regulated ASEs selected from the RNA-seq data under the HG condition, two RASEs were selected for verification. The ratio of variable splicing events occurring in the gene OSMR (alt3p) decreased in the OE-HNRNP-F group (Figure 4H), and increased in the gene TRIP6 (alt5p) (Figure 4G), as expected. We present the designed PCR primer pairs in Supplementary Table S1. TRIP6 mediates inflammatory response and renal fibrosis in diabetic nephropathy (Lin et al., 2021).

The hnRNP-F CLIP-seq data in human 293T cells were obtained from the SRA database accession number GSE34993. These data were utilized to identify transcripts that interact with hnRNP-F in cells. Only reads that mapped uniquely were included in the subsequent analysis. Comparisons between the control group and the IP groups revealed that the reads in the latter were predominantly enriched in noncoding exons, introns, and the 3′UTR region (Figure 5A). RNA-binding proteins that bind to the 3′UTR region often have an impact on RNA stability, suggesting that hnRNP-F may influence RNA stability.

Hypergeometric Optimization of Motif EnRichment (HOMER v4.11, http://homer.ucsd.edu/homer/) was employed for motif analysis of the specific binding peaks identified in the experimental samples. The motif enrichment analysis of the immunoprecipitation (IP) groups revealed enrichment of the UA-rich motif 5′-UUA-3′ in the hnRNP-F-bound motif (Figure 5B). Subsequently, the gene sequences corresponding to the bound peak clusters were aligned with the GO database for annotation, which indicated enrichment ion the RNA splicing process. These primarily included the heterogeneous ribonucleoprotein (HNRNP) family as well as SRSF splicing factors, including HNRNPA2B1, HNRNPH, HNRNPU, SRSF1, SRSF5, and SRSF11 (Figure 5C). Studies have shown that HNRNPA2B1-binding motifs were UA rich (Wu et al., 2018).

Next, we asked whether differences in hnRNP-F binding genes were associated with different gene expressions. We performed gene-based differential binding analyses. We separately analyzed the transcriptome data of hnRNP-F in human 293T cells (GSE34995) and the transcriptome data of hnRNP-F in HK-2 cells that we independently measured. The results showed that hnRNP-F-binding genes overlap with differently expressed genes. Of particular interest was the observation that the lncRNA SNHG1, when bound by hnRNP-F in 293T cells, exhibited differential expression in HK-2 cells overexpressing hnRNP-F (OE-hnRNP-F). Otherwise, in 293T cells, the lncRNA SNHG1 underwent alternative splicing (Figures 5D,E). We also performed an association analysis utilizing CLIP-seq and AS methodologies, which revealed that AS events occurred in four gene regions where hnRNP-F binds in HK-2 and in 16 gene regions in 293T cells (Figure 5E). The AS events of gene RBM41 were detected in both cells (Figure 5E). The analysis of distribution maps indicated that the overexpression of hnRNP-F in HK-2 cells led to a diverse intron retention (ir) AS event of RBM41. Similarly, in 293T cells with hnRNP-F knockdown, an ir AS event was also observed in RBM41. There is an hnRNP-F bound site near the splicing site (Figure 5F). The findings suggest that hnRNP-F can interact with RBM41. The interaction between hnRNP-F and RBM41 results in the production of a truncated transcript of RBM41. Our hypothesis posits that the truncated transcript generated by RBM41 could potentially influence the AS events. Nonetheless, given that the experiment conducted in 293T cells involved the knockdown of hnRNP-F and was characterized by a relatively low sequencing depth, this AS event warrants further experimental investigation.

Initially, we observed a significant reduction in the levels of hnRNP-F protein in the kidney of db/db mice compared to db/m controls (***P < 0.001) (Figures 6A–C). Subsequently, we validated the differentially expressed genes identified through RNA-seq (DEGs: CXCL8, GDF15, IL6, PTX3, and TFPI2) by RT-qPCR. Notably, CXCL8 is a chemokine specific to humans and lacks a direct ortholog in mice, which precludes its validation in murine models. Compared to db/m controls, db/db mice demonstrated significantly elevated renal expression of genes associated with inflammation (GDF15, IL6, PTX3, and TFPI2, ****P < 0.0001) (Figures 6D–G), suggesting their critical roles in the progression of DKD. The downregulation of hnRNP-F protein may have facilitated the upregulation of these genes in the db/db mouse model.

Under HG conditions, MPC-5 cells demonstrated a significant reduction in hnRNP-F protein levels (**P < 0.01), with minimal correlation to hyperosmolarity induced by mannitol (Figures 7A–E). In cells stably transfected to overexpress hnRNP-F (validated by immunoblotting) (Figures 7F,G), TNF-α expression was significantly attenuated in HG conditions (*P < 0.05) (Figures 7H,K), and NF-κB p-p65 phosphorylation was notably suppressed (**P < 0.01) (Figures 7H–J).

Under LPS-induced inflammatory conditions, hnRNP-F protein levels were significantly decreased in both HK-2 cells and MPC-5 podocytes (*P < 0.001; P < 0.01; Figures 8A–F). In MPC-5 cells stably transfected to overexpress hnRNP-F, TNF-α expression was markedly attenuated (*P < 0.05) (Figure 8I), and NF-κB p-p65/p65 was significantly suppressed (**P < 0.01) (Figures 8G,H) under LPS exposure.

To elucidate the mechanistic role of hnRNP-F in transcriptional repression, we conducted Co-IP experiments to examine the hnRNP-F interactome in vivo. In these experiments, HK-2 cells were engineered to stably overexpress hnRNP-F. Total protein lysates were subjected to immunoprecipitation using antibodies specific to hnRNP-F, followed by WB with antibodies targeting ZFP36, HNRNPH, and FOXP3 (Figure 9). The Co-IP analysis using hnRNP-Fantibodies, followed by WB with ZFP36 antibodies, demonstrated a physical association between hnRNP-F and ZFP36. The ZFP36 gene, also known as tristetraprolin (TTP), is a crucial RNA-binding protein that plays a vital role in various biological processes. ZFP36 modulates mRNA stability through its interaction with AU-rich elements (AREs) within mRNA, consequently affecting gene expression and cellular function (Makita et al., 2021). Furthermore, empirical evidence suggests that ZFP36 plays a substantial role in the regulation of alternative splicing (Tu et al., 2019; Chan et al., 2025). We speculate that hnRNP-F and ZFP36 form a complex that regulates gene expression and alternative splicing.