Hsa-miR-423-5p selectively loaded in hypoxic exosomes reduces the sensitivity of normoxic hepatocellular carcinoma to sorafenib via autophagy Authors: Bian Shu, Min Zou, Rui Liao, Diguang Wen, Xianzhi Gao, Jiao Lu, Hua Song, Ziyi Sheng, Zuojin Liu, Yu You Published: 2025 ABSTRACT AbstractBackgroundFor patients with advanced hepatocellular carcinoma, sorafenib is considered a highly effective targeted molecular drug; however, sorafenib resistance limits its therapeutic efficacy. Hepatocellular carcinoma, a type of solid tumour, contains hypoxic and oxygen-rich regions within its internal structure. This study investigated the mechanism by which hypoxic hepatocellular carcinoma cells influence the sorafenib sensitivity of normoxic hepatocellular carcinoma cells through the secretion of hypoxic exosomes containing miRNA signals.MethodsExperiments like CCK-8, IC50, and flow cytometry were used to determine how hypoxic exosomes affect the sorafenib sensitivity of normoxic hepatocellular carcinoma cells. A high-throughput sequencing approach was employed to identify the target of miR-423-5p selectively loaded in hypoxic exosomes. Subsequently, RNA pull-down, RIP, and other experiments were conducted to investigate the mechanism by which hsa-miR-423-5p is selectively loaded into hypoxic exosomes through the RNA-binding proteins (RBPs) acting upstream. Next, Western blot, quantitative real-time PCR, apoptosis flow cytometry, and dual-luciferase reporter assays were performed to investigate how hsa-miR-423-5p affects downstream direct targets and modulates the sorafenib sensitivity of normoxic hepatocellular carcinoma cells through an autophagy mechanism. Finally, the experimental results were validated in an orthotopic hepatocellular carcinoma mouse model by constructing lipid nanoparticles (LNPs) that carry hsa-miR-423-5p inhibitors.ResultsHypoxic hepatocellular carcinoma cells secreted hypoxic exosomes, which could be taken up by normoxic hepatocellular carcinoma cells. These exosomes subsequently reduced the sorafenib sensitivity of normoxic hepatocellular carcinoma cells through an autophagy mechanism. High-throughput sequencing revealed that hsa-miR-423-5p was selectively loaded into hypoxic exosomes via HNRNPA1. In normoxic hepatocellular carcinoma cells, hsa-miR-423-5p promoted autophagy by targeting the downstream target TAB2, thereby decreasing the sorafenib sensitivity of normoxic hepatocellular carcinoma cells to sorafenib.ConclusionsIn response to the upstream factor HNRNPA1, hsa-miR-423-5p is selectively loaded into hypoxic exosomes, which then target the downstream protein TAB2 to regulate the autophagy pathway and reduce the sorafenib sensitivity of normoxic hepatocellular carcinoma cells.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12935-025-04055-8.Keywords: Hepatocellular carcinoma (HCC), Sorafenib, Autophagy, Exosomes, MiRNA, Lipid nanoparticles Introduction Liver cancer represents the sixth most common cancer and the third leading cause of cancer-related mortality on a global scale. Hepatocellular carcinoma (HCC) is responsible for approximately 80% of primary liver cancer cases [1]. The prevalence of liver cancer in China is gradually increasing, as is the incidence of HCC [2]. The early diagnosis of liver cancer presents significant challenges, primarily due to the insidious onset of HCC and the lack of specific early markers. Consequently, the majority of patients are diagnosed at an advanced stage. For advanced HCC, treatment options are limited, and systemic therapy is a crucial component of the therapeutic approach [3]. The efficacy of sorafenib in inhibiting tumour cell proliferation and tumour angiogenesis has been substantiated. Sorafenib is a molecularly targeted drug that has demonstrated efficacy in patients with advanced HCC and is currently utilized as a first-line therapeutic option in clinical practice. It has been shown to markedly extend the survival of patients with advanced HCC [4, 5]. However, only a small group of patients might gain from systemic treatment plans that include sorafenib. However, this group of individuals developed resistance to sorafenib after a few months of treatment. Resistance to sorafenib in HCC, both primary and acquired, is associated with several mechanisms, including autophagy, the tumor microenvironment, cancer stem cells, epithelial-mesenchymal transition, and epigenetic regulation [6, 7]. HCC is considered a solid tumour. HCC cells that proliferate rapidly and have active metabolism are accompanied by an internal tumour vasculature system that is not well developed, resulting in uneven oxygen partial pressure within the tumour. This results in some HCC cells being situated in an oxygen-rich environment in close proximity to blood vessels, whereas other HCC cells are located in an anoxic environment (approximately 100 μm) away from these vessels [8, 9]. Hypoxia is a critical factor in tumour progression. Autophagy frequently occurs when cells are stimulated by environmental stresses, such as hypoxia. Autophagy plays a complex and multifaceted role in the tumorigenesis, cancer progression, and drug resistance of tumors. The critical regulatory function of autophagy in tumour drug resistance has been underscored by numerous researchers [10–12]. Some studies have indicated that tumour cells in solid tumours in a hypoxic environment facilitate the malignant progression of tumour cells in an oxygen-rich environment through exosomes [13, 14]. Exosomes are small extracellular vesicles characterized by a lipid bilayer structure and a diameter ranging from approximately 40 to 160 nm. They are secreted by cells into the surrounding environment and are widely distributed in body fluids. Exosomes contain various molecules, including DNA, RNA, lipids, and proteins, among which microRNAs (miRNAs) have attracted considerable attention as signalling molecules involved in intercellular communication [15]. MiRNAs are non-coding RNAs, about 22 nucleotides in length, that modulate cellular activities at the genetic level [16]. Increasing evidence has demonstrated that miRNAs play pivotal roles in diverse aspects of tumour biology, including the development of drug resistance, as well as the regulation of invasion and metastatic potential [17, 18]. Exosomes carry specific cytoplasmic cargoes during biogenesis, although the underlying control mechanisms involved remain unclear. Currently, the accepted mechanisms for miRNA loading into exosomes include a passive mechanism that is driven by the intracellular concentration of RNA and a selective loading mechanism coordinated by RNA-binding proteins (RBPs) that facilitate the incorporation of miRNAs into exosomes [19]. RBPs can influence gene expression levels by binding to specific sequences in mRNAs [20]. For exosome-mediated loading of miRNAs, certain RBPs are also capable of recognizing specific sequences in miRNAs and promoting their selective loading into exosomes [21]. These miRNAs enter target cells via exosomes, where they exert their functions and regulate target cells. However, studies on the impact of exosome-mediated miRNA transfer from hypoxic cells on the drug resistance of normoxic tumour cells are scarce. This study delineates a mechanistic pathway whereby exosome-mediated communication between hypoxic and normoxic HCC cells plays a pivotal role. Specifically, exosomes released by hypoxic HCC cells selectively load specific miRNAs, which subsequently regulate autophagy pathways in recipient normoxic cells. This regulatory cascade ultimately modulates the sensitivity of normoxic HCC cells to sorafenib. This study provides experimental evidence for identifying new targets for overcoming sorafenib resistance in hepatocellular carcinoma. Methods Cell culture The culture gas conditions used were as follows: 37 °C, 95% air, and 5% carbon dioxide, which is considered a normoxic environment and was used to simulate the oxygen-rich environment in tumours; and 37 °C, 1% oxygen, 5% carbon dioxide, and 94% nitrogen, which is considered a hypoxic environment and was used to simulate the hypoxic environment in tumours. The human hepatoma cell line Huh7, human hepatoma cell line MHCC-97H (97H) and human renal epithelial cell line 293 T were purchased from Procell Life Science & Technology Co., Ltd. (Wuhan, China). The cells were cultured in Dulbecco's modified Eagle’s medium (DMEM; Gibco, USA) supplemented with 10% foetal bovine serum (PAN-Biotech GmbH, Germany), and the medium was replaced with fresh medium every 2 days. To assess the effects of hypoxia on HCC cells, Huh7 and 97H cells were grown under normoxic conditions to 70% confluence and then transferred to hypoxic conditions for 48 h. Exosome isolation, identification and tracing HCC cells were cultured in exosome-free medium under normoxic or hypoxic conditions for 48 h. The supernatants were collected, and exosomes were isolated using differential centrifugation. The morphology, particle size and surface markers (CD81 and TSG101) of the isolated exosomes were assessed via transmission electron microscopy (TEM), nanoparticle tracking analysis (NTA) and Western blotting. Following the manufacturer’s instructions for the PKH67 Green Fluorescent Cytomembrane Linker Kit (PKH67; Solarbio, Beijing, China), exosomes (5 μg/mL) were labeled with PKH67 dye and subsequently incubated with hepatocellular carcinoma cells for 24 h. The internalization of exosomes by recipient cells was visualized by laser confocal microscopy. Detailed methods and information on equipment suppliers are provided in Supplementary Material: Methods S1. Cell counting kit-8 (CCK-8) and IC50 assays After co-culturing the cells with exosomes for 48 h, cell viability was assessed using the Cell Counting Kit-8 reagent. Cells were treated with various concentrations of sorafenib (Huh7: 0, 1, 2, 3, 4, 5 μM; 97H: 0, 2, 4, 6, 8, 10 μM) for 24 h. The IC50 values were calculated using GraphPad Prism 8.0. Cells in the exosome group were treated with exosomes for 48 h. Cell viability at 12, 24, and 48 h was assessed using CCK-8. Absorbance was measured, and cell viability was calculated using GraphPad Prism 8.0. Detailed methods are provided in Supplementary Material: Methods S2. Mimics or inhibitor transfection and lentiviral transfection Mimics were produced by Sangon Biotech (Shanghai) Co., Ltd. (Shanghai, China). The 2ʹ-O-methyl (2ʹ-OMe)-modified inhibitor was produced by Shanghai GenePharma Co., Ltd., located in Shanghai, China. The lentivirus shRNA of HNRNPA1-knockdown (pCDH-U6-HNRNPA1-shRNA-EF1-copGFP-T2A-Puro) was procured from Hunan LeapWal Biotech Co., Ltd. (Hunan, China). The sequences employed were: mimics: 5ʹ-UGAGGGGCAGAGAGCGAGACUUU-3ʹ inhibitor: 5ʹ-AAAGUCUCGCUCUCUGCCCCUCA-3ʹ sh-HNRNPA1: 5ʹ-GAAGAGTTGTGGAACCAAAGA-3ʹ The cells were then transfected with mimics using Lipo8000™ Transfection Reagent (Beyotime Biotechnology, Shanghai, China). Expression levels of the mimics were assessed using RT-qPCR, and Western blot was utilized to assess the expression levels of the genes targeted by the mimics. In accordance with the manufacturer’s instructions, cells were transduced with lentivirus using Hitrans G A (Genechem, Shanghai, China), and medium containing puromycin (2 μg/ml; Beyotime Biotechnology, Shanghai, China) to select for cells with stable transduction. Western blot Cell lysis was performed using RIPA buffer (Beyotime Biotechnology, Shanghai, China), followed by centrifugation at 12,000 × g to obtain the total cellular proteins. Proteins were separated using SDS-PAGE in equal quantities and then transferred to a PVDF membrane (Millipore, USA). After blocking, membranes were incubated overnight at 4 °C with primary antibodies, and were then incubated with secondary antibodies for one hour. The proteins were then visualized using Clarity Western ECL Substrate (Bio-Rad, USA). The following primary antibodies were used: anti-β-actin (20536-1-AP, Proteintech, Wuhan, China), anti-TSG101 (ab125011, Abcam, Cambridge, UK), anti-CD81 (ab109201, Abcam, Cambridge, UK), anti-calnexin (ab22595, Abcam, Cambridge, UK), anti-LC3B (ET1701-65, Huabio, Hangzhou, China), anti-SQSTM1 (#8025, Cell Signaling Technology, USA), anti-HIF1α (#36169, Cell Signaling Technology, USA), anti-HNRNPA1 (ET1704-52, Huabio, Hangzhou, China), and anti-TAB2 (14410-1-AP, Proteintech, Wuhan, China). Quantitative real-time PCR (RT-qPCR) In accordance with the manufacturer’s instructions, the AFTSpin Tissue/Cell Fast RNA Extraction Kit for Animals (ABclonal, Wuhan, China) was used to extract total RNA from cells, while the BIOG Exosome RNA Easy Kit (BIOG, Guangzhou, China) was employed to isolate exosomal RNA. Subsequently, the miRNA 1 st Strand cDNA Synthesis Kit (Vazyme, Nanjing, China) was used to reverse-transcribe RNA into cDNA. The process of RT-qPCR employed SYBR Green Fast qPCR Mix (ABclonal, Wuhan, China), using U6 as the internal control. Tsingke Biotechnology Co., Ltd. (Beijing, China) synthesized all the primers., and the universal reverse Q primer was obtained from Vazyme (Nanjing, China). The sequences of the primers employed were as follows: U6: 5ʹ-GCTTCGGCAGCACATATACTAAAAT-3ʹ Hsa-miR-423-5p: 5ʹ-TGAGGGGCAGAGAGCGAGA-3ʹ Hsa-miR-29a-3p: 5ʹ-CTAGCACCATCTGAAATCGGTTA-3ʹ Hsa-miR-1273f: 5ʹ-CGCGGGAGATGGAGGTT-3ʹ Cell apoptosis assay Cells were harvested by trypsinization using trypsin–EDTA (Beyotime, Shanghai, China), and centrifuged at 1000 × g for 5 min. In accordance with the manufacturer’s instructions, apoptosis was assessed by flow cytometry using an Annexin V-FITC kit (Beyotime, Shanghai, China). Dual luciferase reporter assay The plasmids TAB2-WT and TAB2-MUT were constructed by Genecreate (Wuhan, China). These plasmids contained the mutant or wild-type TAB2 3ʹ-UTR. 293 T cells were placed in a 96-well plate, and once the cells had adhered, mimics/mimics-NC and TAB2-WT/TAB2-MUT were transfected using Lipo8000™ Transfection Reagent (Beyotime Biotechnology, Shanghai, China). Following a period of 48 h of incubation, the luciferase luminescence intensity was quantified using a Dual-Lumi Luciferase Reporter Gene Assay Kit (Beyotime Biotechnology, Shanghai, China) following the manufacturer's instructions. RNA pull-down The 3ʹ-biotin-tagged hsa-miR-423-5p probe was constructed by Sangon Biotech Co., Ltd. The following probe sequence was used: 5ʹ-UGAGGGGCAGAGAGCGAGACUUU-3ʹ-biotin. Following the manufacturer’s instructions, an RNA pull-down kit (BersinBio, Guangzhou, China) was used to conduct the experiment. Briefly, the probe was incubated with streptavidin magnetic beads, and the probe–bead complex was incubated with total cell protein supernatant at room temperature for 2 h. After eluting the collected magnetic beads for 2 h, the supernatant was collected. The analysis was performed via Western blotting. RNA immunoprecipitation (RIP) assay The RIP experiments were performed using the RNA-Binding Protein Immunoprecipitation (RIP) Kit (Genecreate, Wuhan, China) following the manufacturer’s instructions. In summary, specific antibodies and magnetic beads were incubated, and the antibody–magnetic beads complexed with the cell lysate were incubated at 4 °C for a period of 12 h. Thereafter, the magnetic beads were collected and washed five times. The magnetic beads were subsequently resuspended, and RNA was isolated and purified with TRIzol. Finally, the expression levels of microRNAs were detected by RT-qPCR. LNPs production and characterization The 2ʹ-O-methyl (2'-OMe)-modified inhibitor/NC was dissolved in acetate buffer (25 mM, pH = 4.0) to prepare the aqueous phase. DLin-MC3-DMA (SunLipo NanoTech, Shanghai, China), DSPC (Macklin, Shanghai, China), cholesterol (Macklin, Shanghai, China) and DMG-PEG2000 (Macklin, Shanghai, China) were dissolved in ethanol at a molar ratio of 50/10/38.5/1.5 for the preparation of the lipid phase. The aqueous and lipid phases were mixed at a volume ratio of 3:1 using a microfluidic chip (Microfluidic ChipShop, Germany). The mixed solution was dialyzed overnight at 4 °C in PBS to obtain ethanol-stripped LNPs, and the pH was adjusted to 7.4. Therapeutic evaluation in an orthotopic hepatocellular carcinoma mouse model Approval for all animal testing procedures was obtained from the Animal Ethics Committee at Chongqing Medical University. This experiment used 4-week-old BALB/C-Nude mice. The luciferase virus (pCDH-CMV-MCS-EF1-luciferase-neo) was designed and produced by Guangzhou IGE Biotechnology Co., Ltd. As mentioned above, Huh7 cells were transduced with luciferase virus. On a sterile bench, a 20 μl cell suspension containing 1 × 106 Huh7 cells was inoculated into the left lobe of each mouse liver. The mice were randomly divided into four groups and injected with LNPs-NC, LNPs-inhibitor, sorafenib, or sorafenib + LNPs-inhibitor every three days. On Day 12, all the mice were injected with D-Luciferin potassium (Solarbio, Beijing, China) under anaesthesia, and the tumour size was detected using in vivo imaging technology. Following euthanasia, the mice's livers were collected for further study. Detailed methods are provided in Supplementary Material: Methods S3. H&E staining and TUNEL analysis The tumour tissue was processed into paraffin-embedded sections, which were subsequently subjected to dewaxing and hydration. The glass slides were stained using a Haematoxylin and Eosin Staining Kit (Beyotime Biotechnology, Shanghai, China) following the manufacturer’s instructions. The slides were then dehydrated, cleared, and mounted, followed by observation under a microscope. A Fluorescein (FITC) TUNEL Cell Apoptosis Detection Kit (Servicebio, Wuhan, China) was used to stain the sections for cell apoptosis following the manufacturer’s protocol. After the sections were permeabilized, labelling was performed, followed by DAPI nuclear staining. The sections were eventually mounted and observed under a fluorescence microscope. Immunofluorescence (IF) The sections were then dewaxed and hydrated. Antigen retrieval was performed under high temperature and pressure using citrate antigen retrieval solution (Beyotime Biotechnology, Shanghai, China). Permeabilization was carried out by incubating the sections with immunostaining permeabilization buffer containing Triton X-100 (Beyotime Biotechnology, Shanghai, China) for 10 min. The sections were treated with endogenous peroxidase blocking buffer (Beyotime Biotechnology, Shanghai, China) to inhibit endogenous peroxidase activity. Goat serum (AiFang Biological, Hunan, China) was used to block nonspecific binding for 30 min. Following incubation with the primary antibody overnight at 4 °C, the sections were treated with secondary antibodies for an hour the next day. The sections were eventually mounted with Antifade Mounting Medium with DAPI (Beyotime Biotechnology, Shanghai, China) and observed under a fluorescence microscope. Fluorescence in situ hybridization (FISH) FISH was performed on tumor tissue samples using custom-designed probes synthesized by Servicebio (Wuhan, China). FAM 488-labelled probes specific for nucleic acid hsa-miR-423-5p probes were used during hybridization. Finally, the sections were carefully mounted onto slides. They were then observed using a fluorescence microscope to examine the fluorescence signals. The targeted sequences of the probe was as follows: Hsa-miR-423-5p: 5ʹ-AAAGTCTCGCTCTCTGCCCCTCA-3ʹ. Ad-mCherry-GFP-LC3B transfection and autophagy induction Cells were seeded onto confocal culture dishes and transfected with adenovirus expressing mCherry-GFP-LC3B fusion protein (Beyotime Biotechnology, Shanghai, China) according to the manufacturer’s instructions. Following transfection for 24 h at 37 °C, the medium was replaced with fresh DMEM, and cells were further cultured for an additional 24 h. Subsequently, equal amounts of exosomes (5 μg/mL) were added to each group for 48 h. Fluorescence signals were then examined using a laser-scanning confocal microscope. Statistical analysis Quantitative data were analysed using Student’s t-test or one-way analysis of variance (ANOVA) with GraphPad Prism 8.0 software to assess the significance of differences between groups. Statistical significance was defined as p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), or p ≥ 0.05 (N.S.). Results Characterization of exosomes and their uptake by HCC cells HCC cells were cultured in either a normoxic environment (designated normoxic HCC cells) or a hypoxic environment (designated hypoxic HCC cells). The exosomes secreted by each group of cells were extracted by differential ultracentrifugation. Exosomes secreted by HCC cells in a normoxic environment are designated normoxic exosomes (NExs), whereas those secreted by HCC cells in a hypoxic environment are referred to as hypoxic exosomes (HExs). Transmission electron microscopy (TEM) was used to observe vesicular structures (Fig. 1A), and the particle diameter was determined to be approximately 150 nm by nanoparticle tracking analysis (NTA) (Fig. 1B). Western blotting confirmed the presence of exosomal markers (TSG101 and CD81) and the absence of calnexin, indicating exosome purity (Fig. 1C). The results showed that we successfully extracted NExs and HExs secreted by Huh7 and 97H cells. PKH67 (green fluorescence) was used to label the NExs and HExs secreted by Huh7 and 97H cells, which were then incubated with Huh7 and 97H cells under normoxic conditions. The distribution of green fluorescence from PKH67 was observed via laser confocal microscopy to surround the nuclei of Huh7 and 97H cells (Fig. 1 D&E), confirming that the obtained exosomes could be taken up by normoxic HCC cells. Characterization of exosomes and their uptake by HCC cells. A. Morphology of the exosomes identified by transmission electron microscopy (TEM) (30,000 ×). B. Particle size distribution and concentration of exosomes identified by nanoparticle tracking analysis (NTA) after a sample dilution of 1000 ×. C. The expression of exosome surface markers was detected via Western blotting. D&E. The phagocytosis of exosomes by normoxic HCC cells was observed by laser confocal microscopy (200 ×) HExs reduced the sorafenib sensitivity of normoxic HCC cells via autophagy Huh7 and 97H cells were preincubated with or without exosomes (NExs or HExs, 5 μg/mL) and subsequently treated with sorafenib. The occurrence of apoptosis was determined through Annexin V-FITC/PI flow cytometry, which revealed that HExs pretreatment markedly diminished the extent of sorafenib-induced apoptosis (Fig. 2A&B). The IC50 of sorafenib was subsequently evaluated in Huh7 and 97H cells before and after treatment with exosomes under normoxic conditions. Notably, HExs pretreatment attenuated the sorafenib sensitivity of HCC cells (Fig. 2C). These results suggested that HExs modulated the therapeutic response to sorafenib more effectively than NExs. CCK-8 assay results further demonstrated that HExs significantly enhanced the viability of normoxic HCC cells following sorafenib treatment (Fig. 2D). To clarify the function of autophagy, we artificially modulated the autophagic activity of Huh7 and 97H cells using either rapamycin (an autophagy activator) or chloroquine (an autophagy inhibitor), followed by treatment with sorafenib. Annexin V-FITC/PI flow cytometry revealed that enhancing autophagy via rapamycin treatment decreased the sorafenib sensitivity of Huh7 and 97H cells, while suppressing autophagy with chloroquine increased their susceptibility to the drug (Fig. 2E&F). In addition, Western blot analysis was then performed to assess the expression levels of autophagy-related markers. The results demonstrated that HExs enhanced autophagic activity in normoxic HCC cells (Fig. 2G&H). To further confirm this finding, we conducted an autophagic flux assay using Huh7 and 97H cells transduced with adenovirus expressing mCherry-GFP-LC3B fusion protein. Consistent with the Western blot results, HExs co-culture markedly increased LC3B fluorescence signals compared with NExs co-culture, indicating enhanced autophagic flux in normoxic HCC cells (Supplementary Materials: Figure S1A). The data imply that HExs enhance autophagy in normoxic HCC cells while concurrently impairing their responsiveness to sorafenib treatment. HExs reduced the sensitivity of normoxic HCC cells to sorafenib via autophagy. A&B. Following pretreatment of the cells with exosomes, Annexin V-FITC/PI flow cytometry was used to determine the extent of sorafenib-induced apoptosis. C. The IC50 of sorafenib in Huh7 and 97H cells was detected using a CCK-8 assay. D. Huh7 and 97H cells were preincubated with or without HExs and subsequently treated with sorafenib (12, 24, and 48 h). The final assessment of cell viability was performed using a CCK-8 assay. E&F. After autophagy, sorafenib-induced apoptosis was detected via Annexin V-FITC/PI flow cytometry. G&H. Western blot analysis was used to detect the expression levels of autophagy markers in Huh7 and 97H cells that had been preincubated with or without hypoxic exosomes and subsequently treated with sorafenib (* p < 0.05, ** p < 0.01) Hsa-miR-423-5p was selectively loaded by HExs Exosomes are extracellular vesicles that are secreted by cells and contain a variety of biological molecules, including miRNAs, which play important roles in cell-to-cell communication. High-throughput sequencing of miRNAs secreted by Huh7 and 97H cells in NExs and HExs was employed to analyse miRNAs that exhibited increased expression in HExs relative to NExs (screening criteria: fold change > 2, p value < 0.05). Analysis shows that HExs derived from Huh7 and 97H cells exhibit upregulated target miRNAs, namely hsa-miR-29a-3p and hsa-miR-1273f (Fig. 3A-C). RT-qPCR revealed that the target microRNA was highly expressed in HExs but expressed at low levels in hypoxic HCC cells (Fig. 3D). Exosomes secreted by HCC cells are capable of selectively loading target miRNAs. To explore the impact of selective loading on miRNA expression, we conducted high-throughput sequencing of miRNAs in normoxic and hypoxic Huh7 and 97H cells. The significantly upregulated miRNAs in HExs (fold change > 2, p value < 0.05) were filtered, as were the significantly downregulated miRNAs in hypoxic HCC cells (fold change < 0.5, p value < 0.05). Notably, the difference in the miRNA expression profiles between different cell lines may have resulted in the failure to screen for the required miRNAs. Consequently, the screening criteria were adjusted to identify upregulated miRNAs in HExs (fold change > 1, p value < 0.05) and miRNAs with insignificant or downregulated changes in hypoxic HCC cells (fold change < 2, p value < 0.05). The intersection of the screening results was then taken (Fig. 3E-G). In conclusion, only one target miRNA, hsa-miR-423-5p, was screened. RT‒qPCR verification demonstrated that hsa-miR-423-5p was selectively loaded into HExs (Fig. 3H). These results indicated that HExs have a selective loading capacity for hsa-miR-423-5p. Hsa-miR-423-5p was selectively loaded by HExs. A&B. Heatmaps and volcano plots of differentially expressed miRNAs between NExs and HExs secreted by Huh7 or 97H cells. C. Venn diagram of upregulated miRNAs in HExs secreted by Huh7 and 97H cells. D. The expression of the target miRNAs (hsa-miR-29a-3p and hsa-miR-1273f) in exosomes and cells was verified by RT‒qPCR. E&F. Heatmaps and volcano plots of differentially expressed miRNAs in Huh7 and 97H cells (hypoxic vs. normoxic) and exosomes (HExs vs. NExs). G. Venn diagram of miRNAs whose expression levels were not significantly changed in hypoxic HCC cells, downregulated in hypoxic HCC cells, or upregulated in HExs. H. The expression levels of hsa-miR-423-5p in exosomes and cells were verified by RT‒qPCR (N.S. p ≥ 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001) HNRNPA1 mediated the selective loading of hsa-miR-423-5p in HExs To gain more insight into how hsa-miR-423-5p is selectively loaded by HExs, catRAPID, a tool designed for the prediction and calculation of the characterization of RNA‒protein interactions [22], was employed. We initially used the catRAPID database to predict potential RNA-binding proteins (RBPs) (Fig. 4A). HNRNPA1 was the most significant RNA-binding protein among the top ten proteins predicted (Table 1). Western blot analysis revealed that HNRNPA1 is highly expressed in hypoxic Huh7 and 97H cells (Fig. 4B). An RNA probe for biotin-labelled hsa-miR-423-5p (miR-423-5p probe) was subsequently designed for RNA pull-down and successfully pulled down HNRNPA1 (Fig. 4C). Furthermore, the RIP assays demonstrated that hsa-miR-423-5p was enriched in the RNA precipitated with the anti-HNRNPA1 antibody in both Huh7 and 97H cells (Fig. 4D). To determine whether hsa-miR-423-5p is loaded into exosomes via HNRNPA1, we infected Huh7 and 97H cells with a lentiviral shRNA vector specific for HNRNPA1 (sh-HNRNPA1), which successfully knocked down the expression of the HNRNPA1 gene (Fig. 4E). Exosomes were obtained from the conditioned media of Huh7 and 97H cells infected with sh-HNRNPA1 or sh-NC. RT-qPCR analysis demonstrated that the knockdown of HNRNPA1 resulted in notable reductions in the levels of hsa-miR-423-5p in exosomes secreted by Huh7 and 97H cells. No significant alterations in the expression of hsa-miR-423-5p were observed in Huh7 and 97H cells (Fig. 4F). These findings indicate that HNRNPA1 plays a role in the loading of hsa-miR-423-5p into exosomes secreted by Huh7 and 97H cells. HNRNPA1 mediated the selective loading of hsa-miR-423-5p in HExs. A. The catRAPID website predicted RNA-binding proteins that could interact with hsa-miR-423-5p. B. HNRNPA1 expression was detected by Western blotting in Huh7 and 97H cells cultured under normoxic or hypoxic conditions. C. RNA pull-down experiments were performed using the miR-423-5p probe to isolate RNA-binding proteins, and the expression level of HNRNPA1 was subsequently assessed by Western blot analysis. D. RIP experiments were performed using an HNRNPA1-specific antibody to pull down miRNAs, and the expression level of hsa-miR-423-5p was subsequently measured by RT‒qPCR. E. Huh7 and 97H cells were infected with a specific knockdown lentivirus (sh-HNRNPA1) or a control lentivirus (sh-NC). The knockdown efficiency of sh-HNRNPA1 on HNRNPA1 was assessed by Western blotting. F. Cells were collected after infection with a lentivirus, along with their secreted exosomes. The expression levels of hsa-miR-423-5p in both cells and exosomes were measured by qRT-PCR (* p < 0.05, ** p < 0.01, *** p < 0.001) Table 1The catRAPID database was used to predict the top ten potential RNA-binding proteins (ten proteins in the dataset)ProteinTop_RNA_fragmentRNA_binding_domains_IDsRankingHNRNPA126–77PF00076, PF116270.820917PTBP126–77PF13893, PF000760.820333RBM15B44–95PF07744, PF000760.669RBM1544–95PF07744, PF000760.662167HNRNPH226–77PF00076, PF080800.655708ELAVL226–77PF00076, PF138930.653875HNRNPH319–70PF000760.650208A1CF19–70PF00076, PF147090.649292PTBP31–52PF13893, PF000760.647625SEC6344–95–0.509833The potential RNA-binding proteins that could interact with hsa-miR-423-5p were predicted using the online database catRAPID (http://service.tartaglialab.com/page/catrapid_omics2_group), and the top ten candidates were identified based on their predicted rankings from the website. The full results are available in the supplementary file The catRAPID database was used to predict the top ten potential RNA-binding proteins (ten proteins in the dataset) The potential RNA-binding proteins that could interact with hsa-miR-423-5p were predicted using the online database catRAPID (http://service.tartaglialab.com/page/catrapid_omics2_group), and the top ten candidates were identified based on their predicted rankings from the website. The full results are available in the supplementary file Hsa-miR-423-5p promoted autophagy in normoxic HCC cells by targeting TAB2, thereby reducing the sorafenib sensitivity of normoxic HCC cells A mimic of hsa-miR-423-5p (miR-423-5p mimics) was constructed and verified by RT‒qPCR to demonstrate that the transfection of miR-423-5p mimics can result in the overexpression of hsa-miR-423-5p in Huh7 and 97H cells (Fig. 5A). Next, via Western blotting, we confirmed that the miR-423-5p mimic promoted autophagy in Huh7 and 97H cells (Fig. 5B). Following the transfection of Huh7 and 97H cells with miR-423-5p mimics and subsequent treatment with sorafenib, Annexin V-FITC/PI flow cytometry demonstrated that enhancing hsa-miR-423-5p expression markedly diminished the degree of apoptosis induced by sorafenib (Fig. 5C). To further elucidate the mechanism by which hsa-miR-423-5p regulates autophagy, we used the online database miRTarBase 8.0 to predict the target genes of hsa-miR-423-5p (Table 2), which has accumulated more than three hundred sixty thousand miRNA‒target interactions [23]. We also obtained the autophagy-related protein data from the GSEA database (Table 3). The two datasets were then compared and intersected to obtain potential downstream autophagy-related targets of hsa-miR-423-5p (Table 4). It has been demonstrated that TAB2 operates as a tonic inhibitor of autophagy [24]. Therefore, TAB2 is a potential direct target of hsa-miR-423-5p in the autophagy-related downstream pathway. The 3ʹ-UTR of TAB2 contains a putative hsa-miR-423-5p binding site. To confirm the direct binding between hsa-miR-423-5p and the 3ʹ-UTR of TAB2, we constructed wild-type (TAB2-WT) or mutant (TAB2-MUT) plasmids in the pmirGLO dual-luciferase reporter vector. A dual-luciferase reporter gene assay demonstrated that hsa-miR-423-5p could interact with TAB2-WT but not TAB2-MUT (Fig. 5D). Furthermore, Western blot analysis showed that elevating hsa-miR-423-5p levels resulted in a reduction in TAB2 expression following the transfection of hsa-miR-423-5p mimics into Huh7 and 97H cells (Fig. 5E). These findings suggest that hsa-miR-423-5p derived from hypoxic exosomes targeted and inhibited TAB2, thereby activating autophagy in normoxic HCC cells and reducing their sensitivity to sorafenib. Hsa-miR-423-5p promoted autophagy in normoxic HCC cells by targeting TAB2, thereby reducing the sensitivity of normoxic HCC cells to sorafenib. A. Huh7 and 97H cells were transfected with miR-423-5p mimics or mimics-NC, and the expression level of hsa-miR-423-5p in the cells was detected by RT‒qPCR. B. Expression levels of the autophagy markers LC3B and SQSTM1 in cells were detected by Western blotting. C. Huh7 and 97H cells were transfected with or without the miR-423-5p mimic, followed by sorafenib treatment, and apoptosis was detected via Annexin V-FITC/PI flow cytometry. D. Wild-type and mutated binding sites between hsa-miR-423-5p and TAB2. Relative luciferase intensity of cells cotransfected with miR-423-5p mimics or mimics-NC and TAB2-WT or TAB2-MUT. E. Expression level of TAB2 in Huh7 and 97H cells transfected with miR-423-5p mimics or mimics-NC (N.S. p ≥ 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001) Table 2The online database miRTarBase 8.0 was used to predict that hsa-miR-423-5p targets genes (ten proteins in the dataset)miRNASpecies (miRNA)TargetSpecies (Target)hsa-miR-423-5phsaCAPN15hsahsa-miR-423-5phsaABL1hsahsa-miR-423-5phsaACTG1hsahsa-miR-423-5phsaADCY3hsahsa-miR-423-5phsaAK2hsahsa-miR-423-5phsaALDOChsahsa-miR-423-5phsaARHGDIAhsahsa-miR-423-5phsaARRB2hsahsa-miR-423-5phsaASCL2hsahsa-miR-423-5phsaASPHhsaThe potential direct downstream targets of hsa-miR-423-5p were predicted using the miRTarBase 8.0 online database (https://mirtarbase.cuhk.edu.cn/~miRTarBase/miRTarBase_2019/php/index.php). Some of the results are shown in Table 2, and the full results are available in the supplementary file The online database miRTarBase 8.0 was used to predict that hsa-miR-423-5p targets genes (ten proteins in the dataset) The potential direct downstream targets of hsa-miR-423-5p were predicted using the miRTarBase 8.0 online database (https://mirtarbase.cuhk.edu.cn/~miRTarBase/miRTarBase_2019/php/index.php). Some of the results are shown in Table 2, and the full results are available in the supplementary file Table 3The GSEA database was used to obtain data on autophagy-related proteins (fifty proteins in the dataset)ProteinsABL1ATG12ATP6V0BATP6V1C2BECN1ABL2ATG13ATP6V0CATP6V1DBMFACER2ATG14ATP6V0D1ATP6V1E1BNIP3ADCY10ATG16L1ATP6V0D2ATP6V1E2BNIP3LADRA1AATG2AATP6V0E1ATP6V1G1BOKADRB2ATG5ATP6V0E2ATP6V1G2C9orf72AKT1ATMATP6V1AATP6V1HCALCOCO2AMBRA1ATP13A2ATP6V1B1BADCAMKK2ATF6ATP6V0A1ATP6V1B2BAG3CAPN1ATG101ATP6V0A2ATP6V1C1BCL2CAPNS1The dataset of human autophagy-related proteins was obtained through the GSEA online database (https://www.gsea-msigdb.org/gsea/msigdb/human/geneset/GOBP_REGULATION_OF_AUTOPHAGY). Some of the results are shown in Table 3, and the full results are available in the supplementary file The GSEA database was used to obtain data on autophagy-related proteins (fifty proteins in the dataset) The dataset of human autophagy-related proteins was obtained through the GSEA online database (https://www.gsea-msigdb.org/gsea/msigdb/human/geneset/GOBP_REGULATION_OF_AUTOPHAGY). Some of the results are shown in Table 3, and the full results are available in the supplementary file Table 4Potential downstream autophagy-related targets of hsa-miR-423-5pTargetSpecies (Target)ABL1hsaATP6V1E1hsaGAPDHhsaMTMR3hsaSQSTM1hsaTAB2hsaULK1hsaThe two datasets were then compared and intersected to obtain potential downstream autophagy-related targets of hsa-miR-423-5p Potential downstream autophagy-related targets of hsa-miR-423-5p The two datasets were then compared and intersected to obtain potential downstream autophagy-related targets of hsa-miR-423-5p Lipid nanoparticles loaded with hsa-miR-423-5p inhibitors enhanced the sorafenib sensitivity of HCC cells in vivo LNPs are recognized as an effective means of delivering RNA to target organs in vivo, with the potential for significant implications in the treatment of tumours [25–28]. To elucidate the role of hsa-miR-423-5p on the sorafenib sensitivity of hepatocellular carcinoma (HCC) in vivo, a 2ʹ-O-methylated hsa-miR-423-5 inhibitor (inhibitor) was constructed with the objective of inhibiting the function of hsa-miR-423-5p in vivo. The constructed inhibitor was transfected into Huh7 and 97H cells in vitro, and RT-qPCR was used to confirm that the inhibitor effectively knocked down the expression of hsa-miR-423-5p in the cells (Fig. 6A). Lipid nanoparticles loaded with the hsa-miR-423-5p inhibitor downregulated the expression level of hsa-miR-423-5p in vivo, thereby increasing the sensitivity of HCC to sorafenib. A. RT‒qPCR was used to detect the knockdown efficiency of the inhibitor on hsa-miR-423-5p. B. Morphology of LNPs identified by transmission electron microscopy (TEM) (30,000 ×). C. Particle size analysis was performed to determine the size of the LNPs. D. Flow cytometry was used to assess the phagocytosis and uptake of LNPs by cells at different time points (0, 4, 8, and 12 h). E. The phagocytosis and uptake of LNPs by cells were observed under a confocal microscope (the inhibitor was labelled with Cy5 fluorescent dye) (400 ×). F. In vivo distribution of LNPs in mice was observed using an in vivo imaging system. G. The distribution of LNPs in various organs was monitored using an in vivo imaging system. H. The tumour fluorescence intensity in the mice was assessed using an in vivo imaging system. I. Tumour volumes in the liver under different treatments were measured. J. HCC orthotopic tumour formation was observed through HE staining (40 ×). K. Tumour cell apoptosis in the liver after different treatments was detected by TUNEL analysis (200 ×) (*** p < 0.001) LNPs-inhibitor modified with 2ʹ-O-methyl RNA was constructed (the inhibitor was labelled with Cy5 fluorescent dye). The characteristics of the particles were verified by TEM and particle size analysis (Fig. 6B&C). We subsequently coincubated LNPs with cells and confirmed the phagocytosis of LNPs by the cells by flow cytometry and laser confocal microscopy (Fig. 6D&E). Next, we constructed an orthotopic HCC mouse model by implanting Huh7 cells, injecting LNPs into the mice through the tail vein, placing them under an in vivo imaging instrument for observation, and then dissected the mice to remove organs such as the liver and kidneys and placed them under an in vivo imaging instrument for observation. These results confirmed that LNPs can accumulate in the liver and thus exert their RNA delivery function (Fig. 6F&G). To further investigate the impact of the LNPs-encapsulated miR-423-5p inhibitor on intratumoral hsa-miR-423-5p expression in the orthotopic HCC mouse model, the orthotopic HCC mouse model was treated according to the established groupings (LNPs-NC, LNPs-inhibitor). After 12 days of treatment, all the mice were euthanized, and the orthotopic tumour tissues were harvested. PCR and fluorescence in situ hybridization (FISH) analyses revealed that the levels of hsa-miR-423-5p were lower in the LNPs-inhibitor group compared to the LNPs-NC group (Supplementary Materials: Figure S2A&B). Furthermore, immunofluorescence (IF) analysis revealed that the inhibitor reduced the inhibitory effect of hsa-miR-423-5p on TAB2 in vivo (Supplementary Materials: Figure S2C). The orthotopic HCC mouse model was treated according to the established groupings (LNPs-NC, LNPs-inhibitor, sorafenib, or sorafenib + LNPs-inhibitor). Subsequent in vivo imaging revealed a notable reduction in tumour size among those subjects treated with sorafenib + LNPs-inhibitor compared with the other groups (Fig. 6H). After the mice from each group were euthanized, the livers were harvested, and the tumour volumes were measured. The results aligned with previous findings, as the tumour size in the sorafenib + LNPs-inhibitor treatment group was significantly smaller compared to the other groups (Fig. 6I). H&E staining of the mouse livers confirmed the occurrence of tumours (Fig. 6J). Moreover, apoptosis staining of tumour sections revealed that the apoptosis rate in the sorafenib + LNPs-inhibitor treatment group was greater compared to the other groups (Fig. 6K). In summary, in vivo delivery of the hsa-miR-423-5p inhibitor via LNPs increased the sorafenib sensitivity of HCC by downregulating hsa-miR-423-5p expression. Discussion HCC is a type of solid tumour characterized by a poorly developed intratumoral vascular system. The oxygen partial pressure gradually decreases in tumour areas (> 100 μm) that are distant from blood vessels, thereby forming an area of hypoxia. Hypoxia is a salient feature of the internal microenvironment of solid tumours. Hypoxia exerts a regulatory influence on tumour cell metabolism, proliferation, and survival by activating transcription factors such as hypoxia-inducible factors (HIFs). Additionally, hypoxia impacts the autophagy process in tumour cells [11, 29]. Autophagy is a critical intracellular degradation mechanism that plays a pivotal role in the elimination of damaged organelles and proteins, thereby maintaining homeostasis within the cell [30]. There is an emerging consensus that autophagy is a multifaceted process in tumour treatment. Autophagy activation has been shown to facilitate the adaptation of tumour cells to extreme environmental stresses, such as hypoxia and drug treatments [31]. The majority of research on hypoxia has focused on the impact of this condition on tumour cells within a hypoxic environment. However, few studies have examined the effects of hypoxic tumour cells on normoxic tumour cells. Nevertheless, within the tumour microenvironment, many cell-secreted signalling molecules facilitate intercellular communication [32]. Exosomes are of interest to researchers as signalling molecules for cell-to-cell communication. This study is the first to demonstrate that exosomes secreted by hypoxic HCC cells affect the sorafenib sensitivity of normoxic tumour cells via autophagy. Exosomes are defined as extracellular vesicles secreted by cells. Hypoxic conditions have been reported to enhance exosome secretion from tumor cells, consequently influencing tumor angiogenesis, proliferation, invasion, and migration [33, 34]. Exosomes are nanoscale extracellular vesicles that carry diverse signaling molecules, including microRNAs (miRNAs). These molecules have garnered significant attention because of their role in facilitating cell-to-cell communication. For example, hypoxic exosomes secreted by hypoxic glioma cells deliver miR-1246 and miR-10b-5p to normoxic glioma cells, thereby enhancing their migratory and invasive capacities [13]. Previous studies by our group have confirmed that hypoxic exosomes secreted by HCC cells can deliver miR-1273f to normoxic HCC cells, thereby promoting their proliferation and metastasis [35]. As previously reported, certain miRNAs can enter exosome-like structures through a selective loading mechanism [36, 37]. Interestingly, during the course of our experiments, we observed that miRNAs that were highly enriched in HExs were not necessarily upregulated in hypoxic cells. Accordingly, we believed that the selective loading mechanism was not restricted to miRNAs that were concurrently upregulated in both HExs and hypoxic cells. Through specific sequence recognition by RBPs, target miRNAs that were not highly expressed in hypoxic cells were selectively enriched into HExs via a selective loading mechanism. Specifically, miRNAs exhibiting minimal changes or even low expression levels in hypoxic cells could be recognised by RBPs and subsequently enriched into HExs through this selective loading process. Taken together, these findings provide an alternative perspective to the traditional "concentration-driven" paradigm, expands the mechanistic understanding, and provides novel perspectives with experimental evidence for investigating these sorting mechanisms. Therefore, high-throughput miRNA sequencing was employed on cell and exosome samples. The target microRNA (hsa-miR-423-5p) that was selectively loaded was identified by comparing the exosomal high-throughput sequencing data. The expression level of this target microRNA was then experimentally verified. The aim of our study was to elucidate the precise mechanism of the selective loading of hsa-miR-423-5p and its regulation of autophagy. Further elucidation was achieved through a combination of experimental validation and database analysis, leading to the identification of HNRNPA1 as a key player in the packaging of hsa-miR-423-5p into hypoxic exosomes. Moreover, it was shown that hsa-miR-423-5p affects the TAB2 target gene, thereby modulating autophagy levels within cells. HNRNPA1 is known to participate in the selective loading of miRNAs into exosomes. For example, HNRNPA1 mediates the packaging of miR-522 into exosomes derived from cancer-associated fibroblasts [38]. In this study, we predicted the interaction between HNRNPA1 and hsa-miR-423-5p using an online database and experimentally validated this interaction. Additionally, we constructed an HNRNPA1-knockdown lentiviral cell line and observed a significant reduction in the expression of hsa-miR-423-5p in exosomes. These results suggest that HNRNPA1 is involved in the selective loading of hsa-miR-423-5p into hypoxic exosomes derived from HCC cells. TAB2 has been identified as a strong inhibitor of autophagy. The overexpression of TAB2 inhibits autophagy, whereas its depletion triggers autophagy [24, 39, 40]. In this study, we transfected HCC cells with a mimic of hsa-miR-423-5p and assessed the levels of autophagy and TAB2 expression. Consistent with previous studies, we observed that hsa-miR-423-5p downregulated TAB2 and activated autophagy in HCC cells. Our study suggested that hsa-miR-423-5p affects the sorafenib sensitivity of HCC cells in normoxic regions by reducing the expression of its target gene TAB2. The depletion of TAB2 triggered autophagy in HCC cells, helping the cells maintain intracellular homeostasis and reducing their sensitivity to sorafenib. Exosomes, as messengers of intercellular communication, carry hsa-miR-423-5p into exosomes under the influence of HNRNPA1 under hypoxic conditions. This facilitates the transmission of sorafenib resistance between hypoxic and normoxic HCC cells. Upon in vivo administration of a LNPs-inhibitor to inhibit the expression of hsa-miR-423-5p, we effectively increased the sorafenib sensitivity of HCC cells. In conclusion, hypoxic HCC cells are capable of transmitting hypoxic signals and sorafenib resistance to surrounding HCC cells via exosomes. Hsa-miR-423-5p, the focus of our study, offers a novel potential combination therapeutic strategy to overcome sorafenib resistance in HCC. Conclusions In our study, HExs reduced the sorafenib sensitivity of normoxic HCC cells by promoting autophagy. These findings suggest that HExs carry ‘autophagy-activating signals’, which are delivered to normoxic HCC cells, thereby decreasing their sensitivity to the drug. Upon further investigation, we identified the target miRNA (hsa-miR-423-5p) that showed a greater degree of change in HExs than in NExs. Furthermore, we demonstrated that the RNA-binding protein HNRNPA1 specifically bound to hsa-miR-423-5p, facilitating its selective loading into HExs. Based on the experimental results, normoxic HCC cells engulf HExs, resulting in an increased level of hsa-miR-423-5p within the exosomes. This miRNA targets its downstream protein (TAB2), promoting autophagy and reducing the sorafenib sensitivity of normoxic HCC cells. Finally, we confirmed this conclusion in an animal model. These findings indicate that hsa-miR-423-5p could serve as a novel target for combating sorafenib resistance. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Acknowledgements The authors thank all participants in this study for their enthusiastic cooperation. Abbreviations Hepatocellular carcinoma Normoxic exosomes Hypoxic exosomes Transmission electron microscope Nanoparticle tracking analysis Tumour susceptibility 101 CD81 molecule Hypoxia inducible factor 1 subunit alpha Heterogeneous nuclear ribonucleoprotein A1 Sequestosome 1 Microtubule associated protein 1 light chain 3 beta TGF-beta activated kinase 1 (MAP3K7) binding protein 2 Lipid nanoparticles Author contributions BS conducted the experiments, curated the data, and wrote the main manuscript text. MZ performed the formal analysis and wrote the main manuscript text. RL visualized the data and wrote the main manuscript text. DGW conceptualized the study. XZG and JL curated the data. HS prepared Figs. 1, 2, 3. ZYS prepared Figs. 4, 5, 6. All the authors have approved the final version of the manuscript and have agreed to its publication. Funding This work was supported by the National Natural Science Foundation of China (No. 82370672), the Chongqing Medical University Program for Youth Innovation in Future Medicine (No. W0084), the Kuanren Talents Program of the Second Affiliated Hospital of Chongqing Medical University (No. kryc-yq-2208), the Chongqing Natural Science Foundation (CSTB2023NSCQ-MSX0150), and the Chongqing Municipal Science and Health Joint Medical Research Project (2024QNXM003). Data availability The raw data supporting the findings of this study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate All animal experiments were conducted in accordance with the guidelines and regulations of Institutional Animal Care and Use of Chongqing Medical University (Approval Number: IACUC-CQMU-2025-0058). Consent for publication The manuscript has been approved for publication by all the authors. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Bian Shu, Min Zou and Rui Liao contributed equally to this work. Contributor Information Zuojin Liu, Email: 300376@hospital.cqmu.edu.cn. Yu You, Email: youyu@cqmu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Data Availability Statement The raw data supporting the findings of this study are available from the corresponding author upon reasonable request.