Ginseng-derived exosome-like nanovesicles protect against liver fibrosis by regulating TIMP2 pathways and gut dysbiosis

Abstract

Metabolic dysfunction-associated fatty liver disease (MASLD) and alcohol-associated liver disease (ALD) are prevalent chronic liver diseases that can progress to steatohepatitis, fibrosis, cirrhosis, and ultimately liver failure. Here, we demonstrated that oral administration of GNVs provided substantial protection against liver injury and fibrosis in MASLD and ALD mouse models. In a Western-style high-fat diet-induced MASLD model and a chronic binge alcohol-induced ALD model, GNVs treatment significantly reduced gut leakiness by restoring intestinal junctional complex proteins and rebalancing the gut microbiome. GNVs attenuated hepatic lipid accumulation, oxidative stress and fibrogenic markers. GNV treatment downregulated the fibrosis-associated tissue inhibitor of metalloproteinase-2 (TIMP2) pathway in hepatic stellate cells, which is linked to enhanced matrix degradation and reduced fibrogenesis. GNVs prevent MASLD- and ALD-associated gut barrier dysfunction and liver fibrosis through modulation of the gut–liver axis and the TIMP2 pathway. Edible GNVs represent a novel, multifaceted therapeutic strategy for managing chronic liver diseases.

Graphical abstract

1. Introduction

Metabolic dysfunction-associated fatty liver disease (MASLD) and alcohol-associated liver disease (ALD) are considerable global health concerns [1,2]. MASLD is characterized by excessive hepatic fat accumulation and components of metabolic syndrome including in the absence of alcohol consumption [3,4]. ALD results from chronic alcohol intake and shares similar pathological spectra with MASLD despite differing etiologies [5]. MASLD and ALD are prevalent chronic liver diseases that progress from the reversible stages of simple steatosis to inflammatory steatohepatitis, fibrosis, irreversible cirrhosis, and hepatocellular carcinoma if left untreated [1,[6], [7], [8], [9]]. A key feature of their pathogenesis is disruption of the intestinal barrier. This exposes the liver to gut-derived toxic agents, such as bacterial products, including endotoxins, and triggers inflammation. Liver dysfunction or disease also exacerbates intestinal permeability, creating a vicious cycle of gut–liver axis injury [10]. Therefore, MASLD and ALD are closely associated with gut–liver axis dysfunction, characterized by leaky gut, systemic endotoxemia-related inflammation, and progressive liver fibrosis/cirrhosis. Given the limited therapeutic options for these conditions, there is an urgent need for safe and effective interventions targeting gut–liver axis injury.

Ginseng (Panax ginseng C.A. Meyer) is a medicinal plant known for its diverse pharmacological benefits and has been a traditional medicine for thousands of years [11,12]. It has been shown to have adaptogenic, detoxifying, anti-cancer, anti-diabetic and anti-inflammatory properties in experimental models [13,14]. Ginseng extracts can suppress the proliferation of liver and breast cancer cells [15,16] while they enhance immune function [17]. In terms of their relevance to liver disease, specific ginsenosides, such as Rg1 and Rb1, have hepatoprotective effects [[18], [19], [20]]. These findings suggest that ginseng and its bioactive components hold promise as therapeutic agents for liver-associated diseases.

Plant-derived exosome-like nanovesicles (PENVs) are cell-derived vesicles 50–300 nm in diameter enclosed by a lipid bilayer that contain proteins, lipids, mRNAs and miRNAs from their source plant cells [21,22]. PENVs can also deliver their contents to other organisms, such as microbes and mammalian cells [23]. PENVs have several advantages as therapeutic delivery platforms, such as ease of large-scale preparation, high stability in physiological conditions, extremely low toxicity, and innate biocompatibility [[24], [25], [26], [27]]. Ginseng-derived exosome-like nanovesicles (GNVs), a subclass of PENVs enriched with ginsenosides, can cross biological barriers and exert potent bioactivities. GNVs were reported to cross the blood–brain barrier and modulate the tumor microenvironment. It has been reported that it induces apoptosis of 4T1 cells through PTEN/PI3K/Akt/mTOR pathways and caspase-dependent pathways to improve the effectiveness of tumor treatment [28,29], to alleviate inflammatory bowel disease (IBD) via toll-like receptor 4–mitogen-activated protein kinase (TLR4/MAPK) and p62/Nrf2/Keap1 signaling pathways [30], to inhibit osteoclast differentiation in bone disease [31], and to protect skin from ultraviolet (UV)-induced oxidative stress by modulating AP-1 signaling [32]. It also responds to the wound microenvironment of infectious wounds and controls them to have a wound healing effect [33]. GNVs are distinct from other PENVs due to their content of ginsenosides and other ginseng-specific phytochemicals, which may provide additional therapeutic effects. However, to date, the benefits of GNVs in the context of MASLD and ALD, nor their influence on intestinal injury and liver fibrosis through the gut–liver axis have been extensively examined.

Based on ginseng’s known hepatoprotective constituents and the multiple advantages of PENVs, we hypothesized that edible GNVs could ameliorate gut and liver pathology in chronic liver diseases. Therefore, in this study, we aimed to isolate and characterize nanovesicles from ginseng and evaluate their protective effects against Western-style high-fat diet-induced MASLD and ethanol-induced ALD in mice. This was conducted with a particular focus on improving leaky gut and liver fibrosis (Fig. 1). We examined the underlying mechanisms, including the regulation of the hepatic TIMP2 pathway and gut microbiome restoration, to establish GNVs as a potential therapeutic agent for effectively managing liver fibrosis.

Fig. 1.

Schematic diagram to describe isolation, administration and protective effects of GNVs against leaky gut and liver fibrosis in MASLD and ALD models. Up (red) and down (blue) arrows indicate increases and decreases in the indicated targets, respectively.

2. Materials and methods

2.1. Materials

Minimum essential medium alpha (MEM-α), fetal bovine serum (FBS), penicillin, streptomycin and trypsin were sourced from Gibco (Grand Island, NY, USA). Unless otherwise specified, most reagents were obtained from Sigma (St. Louis, MO, USA), following protocols outlined in previous studies [25,[34], [35], [36]]. C57BL/6 J mice (6 weeks old) were purchased from Orient Bio Inc. (Seongnam-si, Korea). All the animal experimental procedures were approved by the Committee for Animal Ethics and Experimental Safety at Andong National University (Approval No. 2022-4-0901-04). They were conducted in accordance with NIH guidelines for the care and use of small laboratory animals.

2.2. Isolation of GNVs by ultracentrifugation and tangential flow filtration

Medium-sized, 6-year-old ginseng plants was obtained from Punggi Ginseng Agricultural Cooperatives. Nanovesicles were isolated using a protocol established in our laboratory [25,[34], [35], [36]]. The ginseng was thoroughly washed and homogenized with 1 l cold phosphate buffer saline (PBS) using a blender for 5 min. To eliminate residual fibers, the homogenate underwent stepwise centrifugation processes: twice at 500 × g for 10 min, followed by 2,000 × g for 20 min, and 10,000 × g for 30 min. The final supernatant fraction was subjected to ultracentrifugation at 100,000 × g for 1 h, and the resulting pellet was resuspended in PBS. The protein contents in the purified GNVs were quantified using a bicinchoninic acid (BCA) protein assay (Pierce, Rockford, IL, USA).

Ginseng homogenates underwent sequential centrifugation steps identical to those described above. The supernatant was then processed through a tangential flow filtration (TFF) system equipped with 50 nm hollow fiber filter pores (Hansa BioMed, Tallinn, Estonia). The sample was passed through the filter pores by alternately pushing syringes in opposite directions, after which the concentrate was washed with PBS. The purified GNVs were collected from the retentate, resuspended in PBS, and used for further analyses.

2.3. Size, morphology and surface proteins of GNVs

The size and concentration of GNVs were determined by nanoparticle tracking analysis (NTA) using a Nanosight NS300 system (NanoSight, Amesbury, UK). For transmission electron microscopy (TEM) analysis, GNV samples were loaded onto glow-discharge copper grids coated with a continuous carbon film and negatively stained with 0.75% uranyl formate [37].

GNV surface proteins were observed and quantified using the EV Profiler Kit (ONi, Oxford, United Kingdom) and direct stochastic optical reconstruction microscopy (dSTORM). GNV samples were affinity captured on microfluidic chips using CD9/CD81/CD63 antibodies provided in the kit. HCT116 EVs were used as controls. Immobilized GNVs were fixed with F1 solution provided in the kit for 10 min before being labeled with CD9‐CF488 [excitation (ex)/emission (em): 490 nm/515 nm] and CD63‐CF568 (ex/em: 562 nm/583 nm) antibodies provided. Labeled GNVs were again fixed with F1 for 10 min. GNV samples were imaged on the NanoImager S Mark II microscope (Oxford Nanoimaging, ONi) with a 100× oil‐immersion objective. Labeled proteins were imaged sequentially at 65%, 52% and 76% power for the 647, 561 and 488 nm lasers, respectively, at 1,000 frames per channel with the angle of illumination set to 54.2° The dSTORM‐imaging buffer was freshly prepared and added immediately before image acquisition. The system was calibrated using the bead slide manual assembly before use. Data was processed on NimOS software (version 1.19.7, ONi). Subpopulation analyses of EVs that express one or two markers were analyzed using ONi's online platform CODI (https://alto.codi.bio). We used a density‐based clustering analysis with drift correction and filtering to evaluate each vesicle.

2.4. Lipidomic, proteomic and microRNA analyses of GNVs

Purified GNVs were analyzed for their lipid and protein compositions at the Korea Ginseng Corporation. Lipid extraction and separation were performed using ultra-high performance liquid chromatography (UHPLC), followed by mass spectrometry analysis (Basil Biotech, Incheon, South Korea). Data processing and analysis were conducted using Lipid Search software. For proteomics analysis, proteins were extracted from GNVs, enzymatically digested, and analyzed using liquid chromatography-tandem mass spectrometry (LC–MS/MS). Identified proteins were annotated using Mascot software, with pathway and gene ontology (GO) analyses conducted through the Kyoto Encyclopedia of Genes and Genomes (KEGG) and DAVID databases. Total RNA, including small microRNAs, was extracted from GNVs using a miRNA-specific isolation kit following the manufacturer’s instructions (Qiagen, Hilden, Germany). The quality and quantity of isolated RNA were assessed using a Bioanalyzer. MicroRNA expression profiles were analyzed using next-generation sequencing (NGS) or a microRNA microarray platform. Sequence reads were mapped using the bowtie2 software tool to obtain the bam file. Mature miRNA sequence is used as a reference for mapping. Read counts mapped on mature miRNA sequence were extracted from the alignment file using bedtools v2.25.0.

2.5. Animal disease models and therapeutic experimental design

Female C57BL/6 J mice (6 weeks old) housed under a 12 h light–dark cycle with unrestricted access to food and water. MASLD was induced by feeding a Western-style high-fat, fructose-, and cholesterol-rich diets (FFC) (DooYeol Biotech, Seoul, Republic of Korea) for 8 weeks. Mice were divided into four groups (n = 5–10 mice per group): control (standard chow), FFC diet alone, FFC diet + 0.5 mg/kg/d GNVs for 5 weeks, and FFC diet + 1 mg/kg/d GNVs for 5 weeks. ALD was induced by orally administering binge ethanol (EtOH, 6 g/kg/dose) or dextrose (as control) three consecutive times at 12 h intervals [35,[38], [39], [40]]. Mice were divided into three groups: control group (dextrose), EtOH (6 g/kg/dose) administration group, and EtOH with GNVs administration group (1 mg/kg/dose) (n ≥ 6 per group). The GNVs-treated group received 1 mg/kg GNVs orally daily for 14 d before ethanol exposure.

2.6. Histological and IHC analyses

Liver and intestinal tissues were fixed in neutral formalin, embedded in paraffin, and sectioned for hematoxylin & eosin (H&E) or Sirius Red staining at Kyungpook National University core lab [39]. Histological changes were evaluated under a light microscope [39]. For immunohistochemistry (IHC), tissue sections were stained with the specific antibody against F4/80 (macrophage marker) or Ly6G (neutrophil marker) to assess immune cell population and infiltration.

2.7. Immunoblot analysis

Tissues and cells were lysed using 1× RIPA buffer and homogenized with a plastic/glass homogenizer. Equal amounts of protein lysates were separated by SDS/PAGE and transferred onto nitrocellulose membranes. The membranes were incubated with the respective, specific antibody listed in Table S1. This was followed by stepwise washing steps and then HRP-conjugated secondary antibodies. Protein detection was performed using enhanced chemiluminescence (ECL) solutions (Thermo Fisher Scientific). Band intensities were quantified relative to GAPDH as a loading control using the FUSION SOLO S chemiluminescence imaging system (Vilber, Collégien, France).

2.8. Microbiome analysis, including microbial 16S sequencing and bioinformatics

Cecal stool samples were collected under sterile conditions and immediately stored at −80 °C. Genomic DNA was extracted using the Mag-Bind Universal Pathogen DNA Kit (CJ Bioscience, Seoul, South Korea) according to the manufacturer’s guidelines. DNA sequencing and bioinformatic analyses for bacterial 16S ribosomal RNA were performed at CJ Bioscience (https://www.cjbioscience.com/ngs). To analyze the function of the microbial community, PICRUSt2-based KEGG pathway analysis was conducted. Gut microbiota composition and diversity in MASLD and ALD mice without or with GNV treatment were analyzed according to the levels of Chao1, which is an indicator of microbial species alpha diversity. The distribution pattern was analyzed including beta-diversity.

2.9. RNA sequencing analysis and bioinformatics

LX-2 hepatic stellate cells (HSCs) were pretreated with GNVs or vehicle (PBS) for 24 h. Total RNA was isolated using TRIzol reagent, and RNA integrity was verified using an Agilent 2100 Bioanalyzer (Agilent Technologies, USA) at Ebiogen (https://www.e-biogen.com/product1.php). A paired-end mRNA library was prepared using the NovaSeq Reagent Kits (Illumina, USA), and sequencing was conducted using the NovaSeq 6000 system (Illumina). The KEGG pathway database (http://www.genome.jp/keg/) was used to identify enriched pathways. Functional gene enrichment was determined using Fisher’s exact test, with pathways containing at least two differentially expressed genes (DEGs) and a P-value < 0.05 considered significant.

2.10. Quantitative real-time PCR analysis

Gene expression levels related to oxidative stress, inflammation, liver fibrosis, and intestinal function were measured using quantitative real-time PCR (qRT-PCR). Total RNA was isolated with TRIzol reagent (Invitrogen) and subjected to overnight precipitation to enhance yield. cDNA was synthesized from 600 ng DNase I-treated RNA using a cDNA reverse transcription kit (Applied Biosystems). qRT-PCR was performed using the QuantStudio 1 Real-Time PCR system (Applied Biosystems) using SYBR green PCR Master Mix (Applied Biosystems) with GAPDH as a housekeeping gene. The primer sequences are provided in Table S2.

2.11. TIMP2 knockdown using small RNA interference

Small interfering RNA (siRNA) was used to suppress TIMP2 expression. The siRNA sequences are listed in Table S3. A scrambled siRNA was used as a negative control. TIMP2-specific siRNA was reverse-transfected into LX-2 stellate cells using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer’s protocol. Briefly, 10 pmol of TIMP2 siRNA was mixed with 1 µl Lipofectamine RNAiMAX in 100 µl Opti-MEM (Invitrogen) and incubated for 20 min before being added to 1 × 10⁵ LX-2 cells cultured in 500 µl medium. After 24 h of incubation, gene expression was analyzed [41].

2.12. Immunofluorescence staining of α-SMA and TIMP2

LX-2 HPSs were cultured on chamber slides for immunofluorescence staining of α-smooth muscle actin (α-SMA) or TIMP2. Cells were fixed with 4% paraformaldehyde (PFA) and incubated with a blocking solution. After overnight incubation with α-SMA primary antibody at 4 °C, cells were treated with an Alexa Fluor 488-conjugated secondary antibody (Thermo Fisher Scientific). Nuclei were counter-stained using DAPI in a mounting solution, and fluorescent images were captured using a fluorescence microscope (Logos Biosystems, Inc.).

2.13. In vivo biodistribution assay of GNVs

The biodistribution of GNVs in mice was assessed using an in vivo imaging system (IVIS). GNVs (1 mg/kg) were labeled with DiR dye (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide) by incubating with 10 µl DiR for 30 min, followed by ultracentrifugation at 100,000 × g for 1 h. The pellet was resuspended in PBS and orally administered to mice (n = 3–4 per group). After 24 h, the mice were sacrificed, and organs were excised to assess the degradation and absorption patterns of DiR-labeled GNVs. For intestinal or liver tissue labeling, tissue sections were fixed with 4% PFA for 10 min, blocked with 3% bovine serum albumin (BSA), and immunostained with the primary antibody targeting ZO-1 (SC-33725, Santa Cruz) or mouse anti-albumin (sc-2701605, Santa Cruz). AlexaFluor 488-conjugated secondary antibodies (1:100, Thermo Fisher Scientific) were used for visualization [42], and fluorescence images were acquired using a fluorescence microscope (Logos Biosystems, Inc.).

2.14. In vitro cell uptake assay of GNVs

To track cellular uptake, GNV membrane lipids, proteins, and RNAs were labeled with PKH67 Green fluorescent dye (Sigma), ExoGlow Protein EV Labeling Kit (System Biosciences, CA, USA), and ExoGlow RNA EV Labeling Kit (System Biosciences), respectively, following the manufacturer’s protocols. Labeled GNVs were incubated with the LX-2 human HSC line, AML12 murine hepatocytes, and T-84 human colon carcinoma cells for 24 h. Cells were washed with 1× PBS and analyzed via fluorescence microscopy (Logos Biosystems, Inc., Anyang-si, South Korea).

2.15. In vivo toxicity assessment

C57BL/6 J 6-week-old male mice (n = 5 per group) were randomly assigned to receive either vehicle (PBS) or GNVs (1 mg/kg/dose in PBS) via oral gavage once daily for seven consecutive days. In a separate experiment, a single oral dose of 1 mg/kg/d of GNVs was administered to the mice. Blood samples were collected via cardiac puncture and centrifuged at 1,200 × g for 15 min (Eppendorf Model 5413). Serum samples were immediately frozen at −80 °C. Serum alanine aminotransferase (ALT) levels were assessed to evaluate liver function, while blood urea nitrogen (BUN) and creatinine levels were determined for kidney function. Serum ALT, aspartate aminotransferase (AST) (BioVision, Milpitas, CA, USA), hepatic triglyceride (TG) (Asan Co., Ltd., Gimpo, Korea, respectively), and endotoxin (endpoint LAL Chromogenic Endotoxin Quantitation Kit; Thermo Fisher Scientific, Waltham, MA, USA) levels were quantified using commercial assay kits [7,38], followed by measurements using the DRI-CHEM NX500 analyzer (Fujifilm, Japan).

2.16. Gelatinase/Collagenase assay

The levels of gelatinase/collagenase activity in the mouse liver extracts from the different groups were analyzed using the EnzChek™ Gelatinase/Collagenase assay kit (Thermo Fisher Scientific) following the manufacturer’s protocol.

2.17. Statistical analysis

All in vitro and in vivo data are expressed as mean ± standard deviation and mean ± standard error of mean, respectively. Statistical analyses were conducted using SPSS/Windows 27.0 (SPSS Inc., Chicago, IL, USA). Differences were considered statistically significant at the P < 0.05 level and were evaluated using one-way or two-way ANOVA followed by Duncan’s multiple range test post hoc analysis [7,38,43,44].

3. Results and discussion

3.1. Isolation and characterization of GNVs

GNVs were isolated using two methods, including differential ultracentrifugation (dUC) and TFF, to determine the optimal preparation protocol (Fig. 2A and S1A) [45]. TEM analysis confirmed that the isolated GNVs possess the expected round, nano-sized vesicular morphology with a characteristic lipid bilayer (Fig. 2B). The dUC method yielded a higher number of small-sized vesicles compared to TFF-related isolation. NTA showed that dUC-isolated GNVs had an average diameter of 168.7 ± 7.4 nm with a particle concentration of 1.4 × 10⁹ particles/ml. Meanwhile, TFF-derived GNVs were larger (228.7 ± 4.5 nm on average) and fewer in number (3.8 × 10⁶ particles/ml) (Fig. 2C and S1B). Approximately 77.9% of GNVs from the dUC method were under 200 nm, compared to only 34.9% of those from TFF. Given prior observations that smaller PENVs (<200 nm) have superior cellular uptake and bioactivity [39], the dUC method was better than the TFF isolation. Therefore, it was used for all subsequent experiments. Super-resolution fluorescence imaging further showed that GNVs carry typical extracellular vesicle markers. GNVs prepared by dUC showed a clear presence of CD63 and CD9, whereas TFF-prepared GNVs showed only CD9 (Fig. 2D and S1C). dUC-derived GNVs exhibited a strong expression of ubiquitin and PEN1, a feature often noted in PENVs (Fig. 2E and S1D) [46,47]. However, ubiquitin was weakly expressed in TFF-prepared GNVs. These results indicate that the dUC isolation enriches small, exosome-like ginseng nanovesicles with canonical surface protein markers.

Fig. 2.

Isolation and properties of purified GNVs. (A) Flowchart of dUC procedures to isolate GNVs; (B) TEM image of GNVs; (C) NTA of the size and concentration of the isolated GNVs by dUC; (D) ONi super-resolution nanoimaging of GNVs using the dUC method; (E) Detection of plant protein markers (Ubiquitin and PEN1) in the dUC-GNVs; (F) Pattern of SDS-PAGE of GNV proteins. The proteins extracted from GNVs were separated using SDS-PAGE and stained with Coomassie blue; (G) Major cellular pathways after proteomic profiling of GNVs; (H) Analysis of ginsenosides of GNVs; (I) Lipid profiles of GNVs. Abbreviations for (I): wax esters (WE), sterols (St), sphingomyelin (SM), phosphatidylserine (PS), phosphatidylgylcerols (PG), polyethylene terephthalate (PEt), phosphatidylcholine (PC), monogalactosylacylglycerols (MGDG), mono-ether phosphatidylcholine (MePC), lysophosphatidylethanolamine (LPE), lysophosphatidic acid (LPA), trihexosylceramide (Hex3Cer), monohexosylceramide (Hex1Cer), diacylglycerol (DG), diacylglycerol-triacylglycerol (D5-TG), cholesteryl ester (ChE), ceramide (Cer), bis(monoacylglycero)phosphate (BisMePA), N-acetylhexosaminyl sulfate esters (AcHexStE), N-acetylhexosaminyl sphingosine (AcHexSi), triglycerides (TG), sphingosine (SPH), sphingomyelin equivalent (SiE), phosphatidylinositol (PI), perfluoroalkyl acids (PFAA), phosphatidylethanolamine (PE), phosphatidic acid (PA), monoacylglycerol (MG), lysophosphatidylglycerol (LPG), lysophophatidylcholine (LPC), lysobisphosphatidic acid (LBPA), hexosylceramide (Hex2Cer), digalactosyldiacylglycerol (DGDG), dioleoylphosphatidic acid (DAP), coenzyme (Co), cholesterol (Ch), bis(monoacylglycero)phosphate ethanolamine (BisMePE), bis(monoacylglycero)phosphate lysophosphatidic acid (BisMeLPA), N-acetylhexosaminyl sphingosine ether (AcHexSiE), N-acetylhexosaminyl ceramide ether (AcHexCmE).

We analyzed the molecular composition of GNVs. SDS-PAGE analysis showed that GNVs contain abundant proteins in the 20–30 kDa range (Fig. 2F). The average total protein yield of GNVs from three different isolation batches was 0.4 mg proteins per gram of ginseng (4–5 µg/ml proteins in 9 ml ginseng juice). The concentration of protein per unit of GNVs was 2.86 µg/ml (Fig. S2), which is similar to that reported for other ginseng-derived nanovesicles at 0.2–0.3 mg proteins per gram of ginseng [30]. Proteomic profiling showed that GNVs are enriched in proteins related to stress responses, ion & protein transports, and glycolysis (Fig. 2G and Table S4). In addition to proteins, GNVs were found to carry ginsenosides (Fig. 2H), including Rg1, Re, Rf, Rb1, Rc and R2, and a variety of other lipid compounds (Fig. 2I), including phospholipids and sphingolipids. Our RNA sequencing results showed that GNVs contained various microRNAs, such as pgi-miR6135e.2-5p, pgi-miR6135j, pgi-miR6136a.1, pgi-miR6136a.2, pgi-miR6138 and pgi-miR6139 (Table S5). These molecular characterizations showed that GNVs are naturally loaded with multiple bioactive constituents, that is, proteins, lipids, and ginsenosides, and closely resemble mammalian extracellular vesicles in structure, size, and composition [39]. The successful isolation of bioactive, nanoscale vesicles from ginseng provided the foundation to investigate their therapeutic potential against chronic liver disease models.

3.2. Therapeutic efficacy of GNVs in a MASLD mouse model

We first evaluated the effects of oral GNVs administration against a mouse MASLD model induced by a Western-style high-fat, fructose and cholesterol (FFC) diet. Chronic FFC feeding led to intestinal barrier disruption (“leaky gut”) and dysbiosis. These are known to drive liver inflammation and fibrosis in MASLD via the gut–liver axis [48]. GNV treatment by oral gavage at 0.5 or 1 mg/kg, thrice weekly (Fig. 3A) significantly improved gut integrity in MASLD mice. Histological examination of the small intestine showed that the abnormal gut villus structure caused by the FFC diet was restored toward normal morphology in GNV-treated groups (Fig. 3B). The elevated serum endotoxin (LPS) level observed in FFC-fed mice was significantly reduced by GNV supplementation (Fig. 3C), indicating mitigation of gut leakage. F4/80 and Ly6G staining in the small intestine showed decreased levels of macrophages and neutrophil infiltration, respectively, in the GNV-treated group. This indicated that there was protection against immune cell activation by GNVs (Fig. 3D and S3A). At the molecular level, GNVs preserved and restored the expression of intestinal tight junction (TJ) proteins such as ZO-1, Claudin-4 and Occludin, and adherens junction (AJ) proteins, such as β-catenin, γ-catenin and α-tubulin. All of these were decreased by the FFC diet in MASLD mice (Fig. 3E&3F). In parallel, GNVs had a corrective effect on the gut microbiome. Sequencing of 16S rDNA of fecal microbiota showed that FFC feeding shifted the microbial community toward a pro-inflammatory profile with increased abundance of Proteobacteria and depletion of Bacteroidetes (Fig. 3G). This is a pattern often observed in patients with hepatic cirrhosis or advanced liver disease [[49], [50], [51]]. GNV treatment partially reversed these trends by decreasing the relative abundance of Proteobacteria while increasing Bacteroidetes compared to the untreated FFC controls (Fig. 3G). Treatment with GNVs also inhibited FFC diet-induced alterations in gut microbial composition and abundance at the genus levels (Fig. S4). In contrast, the abundance of beneficial genera, such as Lactobacillus (Fig. 3H) and Alistipes (Fig. 3I) [[52], [53], [54]], which were severely diminished in MASLD mice, was significantly increased after GNVs administration. GNV treatment significantly attenuated the FFC-mediated reduction in microbial species diversity in MASLD mice (Fig. 3J&3K). Although the microbiota of GNVs-treated mice did not fully return to those of healthy controls, the changes indicate a shift toward a healthier, less inflammatory gut ecosystem. These data demonstrate that GNVs effectively prevent FFC-induced gut barrier failure and microbial dysbiosis in MASLD mice, reducing systemic endotoxemia and its downstream inflammatory burden in the liver.

Fig. 3.

GNVs prevented FFC-induced leaky gut and restored microbiome composition and diversity in MASLD mice. (A) Experimental design to study the effect of GNVs against leaky gut in MASLD mice; (B) H&E-stained images of small intestines; (C) Level of serum endotoxin; (D) Immunohistology images of F4/80-stained macrophages in the small intestines; (E–F) Levels of gut TJ proteins (ZO-1, Claudin-4, Occludin) and AJ junction proteins (β-catenin, γ-catenin, α-tubulin) in different groups. Data represent as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 between CON and FFC groups; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 between FFC and FFC + 1 mg/kg GNVs groups. Significance of mean values for each group was determined using one-way ANOVA with Dunnett’s multiple comparisons test; (G) Relative composition and abundance of various bacterial phyla in the overall gut microbiome, and (H–I) the relative abundances of genus Lactobacillus or genus Alistipes in the indicated groups. (J) Chao1 and (K) Shannon indices representing alpha diversity within the microbial community. Data are presented as mean ± SEM. Significance of the mean values for each group was determined using one-way ANOVA and Tukey’s Honestly Significant Difference (HSD) test.

GNV’s intervention also protected the liver from FFC-induced injury. MASLD mice fed the FFC diet developed hallmark features of non-alcoholic steatohepatitis (NASH), including hepatic steatosis, inflammation and fibrosis with elevated serum ALT and AST. GNV-treated mice showed significant improvements in all these aspects. H&E-stained liver histology of FFC-fed mice exhibited increased fat droplets and inflammatory foci compared to those of GNV-treated mice, indicating GNV-mediated attenuation of steatohepatitis (Fig. 4A). F4/80 and Ly6G staining of the liver sections showed decreased levels of macrophages and neutrophil infiltration, respectively, in the GNV-treated group. This indicated the prevention from immune cell activation by GNVs (Fig. 4B and S3B). Serum ALT and AST levels were significantly elevated in untreated MASLD mice. However, GNV treatment at low (0.5 mg/kg/dose) and high (1 mg/kg/dose) doses lowered serum ALT/AST levels nearly to baseline (Fig. 4C&4D). Hepatic TG content was substantially increased in the FFC-fed mice compared to controls. However, it was significantly reduced after GNV administration, and a greater effect was observed at the higher dose (Fig. 4E). These improvements in liver pathology were associated with modulation of key metabolic and inflammatory pathways. GNV treatment significantly suppressed the expression of fatty acid synthase (FAS), sterol regulatory element-binding protein-1 (SREBP-1), and peroxisome proliferator-activated receptor-γ (PPARγ) that were upregulated in the FFC-fed mice (Fig. 4F). Likewise, the amounts of arginase isoforms (Arg1 and Arg2), which were elevated in fatty liver in the FFC-fed MASLD mice [55,56], were brought down to basal levels by GNVs (Fig. 4G). GNVs exhibited strong antioxidant effects in vivo. The FFC diet increased the hepatic levels of oxidative stress markers, including cytochrome P450–2E1 (CYP2E1), inducible nitric oxide synthase (iNOS), and nitrated proteins assessed by anti-3-nitrotyrosine (3-NT) antibodies. However, GNV treatment sharply reduced the expression of all these oxidative stress markers (Fig. 4H). Oral administration of GNVs mitigated hepatic fat accumulation and oxidative injury in MASLD mice.

Fig. 4.

GNVs treatment attenuated FFC-induced fatty liver and inflammatory injury in MASLD mice. (A) H&E-stained liver histology; (B) Images of F4/80-stained hepatic macrophages; (C–E) Levels of serum ALT, AST and hepatic TG in different groups. (F–H) Immunoblot results for (F) the liver proteins involved in the lipid metabolism (FAS, SREBP-1, PPARγ), (G) arginase (Arg1, Arg2), and (H) oxidative stress markers (CYP2E1, iNOS, 3-NT) in the indicated groups. Data represent as mean ± SEM, *P < 0.05, **P < 0.01, ***P < 0.001 between CON and FFC groups; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 between FFC and FFC + 0.5 mg/kg GNVs; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 between FFC and FFC + 1 mg/kg GNVs group. Significance of mean values for each group was determined using one-way ANOVA with Dunnett’s multiple comparison test.

We further studied the effects of GNVs on FFC-mediated liver fibrosis. GNVs significantly inhibited fibrogenesis in the FFC-induced MASLD mice. Sirius Red staining for collagen fiber showed extensive fibrosis in the livers of FFC diet-fed mice. Meanwhile, GNVs-treated mice had significantly decreased collagen-rich areas (Fig. 5A). Collagenase activity was also activated in GNVs-treated livers compared to the FFC-induced MASLD group, supporting the beneficial effects of GNVs against liver fibrogenesis through ECM remodeling (Fig. 5B). GNV administration led to dose-dependent reductions in multiple fibrosis markers that were elevated by the FFC diet. Immunoblotting confirmed that pro-collagen-1, collagen-1, TGF-β, MMP-2, MMP-9 and α-SMA were all upregulated in untreated MASLD mouse livers but were downregulated in GNVs-treated groups (Fig. 5C&5D). These data suggest that GNVs prevent new collagen deposition, but may also suppress HSC activation, as indicated by decreased α-SMA.

Fig. 5.

GNV treatment attenuated FFC-induced liver fibrosis in MASLD mice. (A) Sirius Red-stained liver images; (B) Gelatinase/collagenase activity; (C–D) Immunoblot results for the various liver fibrosis markers, including (C) pro-collagen-1, collagen-1 and TGF-β, and (D) MMP-2, MMP-9 and α-SMA in the indicated groups. Data represent as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 between CON and FFC groups; #P < 0.05, ^##P < 0.01, ^###P < 0.001 between FFC and FFC + 0.5 mg GNVs; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 between FFC and FFC + 1 mg/kg GNVs groups. Significance of mean values for each group was determined using one-way ANOVA with Dunnett’s multiple comparison test for B-D.

GNV therapy alleviates FFC diet-induced steatohepatitis and prevents the progression to liver fibrosis, likely through indirect effects of restoring gut barrier function and reducing endotoxin-driven inflammation and direct effects on liver fibrosis by suppressing HSCs activation and the pro-fibrotic signaling.

3.3. Therapeutic efficacy of GNVs in ALD mouse model

Having established the protective effects of GNVs in MASLD mice, we also studied their efficacy in an ALD mouse model to determine whether GNVs exert their benefits through a similar gut–liver axis mechanism. For the ALD mouse model, we used a binge ethanol exposure regimen (Fig. 6A). This induces acute intestinal injury, microbial dysbiosis and hepatic inflammation/fibrosis akin to alcoholic hepatitis. GNV treatment produced substantial benefits in ALD mice that closely paralleled those observed in MASLD mice. In the small intestine, binge alcohol caused shortening and blunting of villi with compromised mucosal structure (Fig. 6B). However, mice that received oral GNVs alongside ethanol maintained normal villus structure, indicating prevention of alcohol-induced gut damage (Fig. 6B). Consistent with improved gut structural integrity, GNVs significantly lowered the high serum endotoxin (lipopolysaccharide, LPS) levels observed in ethanol-fed mice (Fig. 6C). F4/80 and Ly6G staining in the small intestine showed decreased levels of macrophages and neutrophil infiltration in the GNV-treated group, indicating the prevention of immune cell activation by GNVs (Fig. 6D and S5A). Fecal microbiome analysis showed alcohol-driven shifts in microbial composition and abundance. The ethanol-only group had an elevated abundance of Proteobacteria and depletion of beneficial taxa, such as Alistipes. This is similar to the dysbiosis in MASLD mice. GNV co-administration blunted these microbial changes, partially restoring a healthier balance of major phyla with increased abundance of Alistipes relative to ethanol controls Fig. 6E&6F and S6). The functional profiles of the microbial community evaluated by PICRUSt2-based KEGG pathway analysis showed ABC transformers and Butanoate metabolism (Fig. S7). Analysis of gut microbiome indicated that alpha- and beta-diversity were decreased in ALD mice but significantly improved by GNV treatment and the distribution pattern by beta-diversity (Fig. 6G and S7). Therefore, our in-depth analysis of gut microbiomes in different groups confirmed that GNVs attenuated alcohol-mediated changes in gut microbiota composition and diversity. GNVs counteracted alcohol-induced increments of oxidative stress markers, that is iNOS, CYP2E1 and nitrated proteins, and disruption of gut TJ/AJ proteins (ZO-1, Claudin-4, Occludin, E-cadherin, β-catenin and α-tubulin). GNV treatment normalized the levels of these proteins at both protein and mRNA levels (Fig. 6J–6N), by reducing mucosal oxidative injury and restoring the expression of junctional complex proteins. Therefore, GNVs preserved gut barrier integrity, prevented endotoxemia, and maintained gut microbiota homeostasis in ALD mice similarly as in MASLD mice.

Fig. 6.

GNVs prevented binge alcohol-induced leaky gut in ALD mice. (A) Schematic experimental design to study the beneficial effects of GNVs in ALD mice. (B) H&E-stained small intestine images; (C) Level of serum endotoxin; (D)F4/80-stained images of the small intestine; (E–F) Relative composition and abundance of various bacterial phyla and the relative abundance of genus Alistipes in different groups. (G) Alpha diversity indicator Chao1. Significance of the mean values for each group was determined using one-way ANOVA and Tukey’s post hoc analysis. (H–J) Immunoblot results for the gut oxidative stress (H) marker proteins (iNOS, CYP2E1, 3-NT) and junctional complex proteins, including (I) ZO-1, Claudin-4, and Occludin, and (J) E-cadherin, β-catenin, γ-catenin, α-tubulin, and (K–L) qRT-PCR results for the mRNA transcripts of gut junction-related genes, including (K) ZO-1, Claudin-1, and Occludin, and (L) E-cadherin, β-catenin, γ-catenin 1, γ-catenin 2 in the indicated groups. All data represent as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 between CON and EtOH groups; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 between EtOH and EtOH + 1 mg/kg GNVs groups. Differences among group means were assessed using a one-way ANOVA followed by Dunnett’s multiple-comparison test.

Protective effects of GNVs were also evident in the liver injury of ALD mice. Histopathology analyses (H&E staining) showed that livers of ethanol-exposed mice had increased fat vacuoles, apoptosis and inflammatory loci. Meanwhile, GNV-treated mouse livers had substantially less fat deposition and inflammatory cells (Fig. 7A). F4/80 and Ly6G staining in the liver showed decreased levels in macrophage and neutrophil infiltration, respectively, in the GNV-treated group. This indicated that GNVs offer protection against immune cell activation in ALD mice (Fig. 7B and S5B). Elevated serum ALT and AST in alcohol-exposed mice were significantly reduced with GNVs co-treatment (Fig. 7C&7D). GNVs also prevented alcohol-induced hepatic TG content (Fig. 7E), suggesting an attenuation of ethanol-mediated steatosis. At the molecular level, GNV administration mitigated the ethanol-triggered surge in oxidative stress markers, fibrogenic and apoptotic pathways in the liver. Protein levels of CYP2E1 and 3-NT-related nitrated proteins were substantially elevated in ethanol-exposed mice, reflecting oxidative stress. However, these marker proteins were substantially decreased in the GNVs-exposed group (Fig. 7F). The pro-fibrotic proteins MMP-9 and α-SMA, which were increased in the ethanol group, were significantly reduced by GNV co-treatment (Fig. 7G). Likewise, markers of hepatocellular apoptosis, including phosphorylated p-JNK, cleaved caspase-3, and Bax, were all upregulated by ethanol exposure. All these proteins were markedly reduced by GNVs treatment (Fig. 7H). The activity measurements for the different liver extracts confirmed that hepatic collagenase/gelatinase activity was significantly increased in the GNV-treated group compared to that of the alcohol-exposed group (Fig. 7I). In liver tissue, GNVs also suppressed the ethanol-induced upregulation of genes associated with inflammation and oxidative stress. mRNA levels of iNOS, CYP2E1, Cox-2, TLR4 and NF-κb were significantly lower in GNVs co-treated mice than those in ethanol-exposed mice (Fig. 7J). GNVs mitigate alcohol-induced liver injury through multiple mechanisms, via inhibiting oxidative stress, hepatocyte death and fibrogenesis. The pattern of protection in ALD mice closely mirrored that in MASLD mice, highlighting the broad therapeutic potentials of GNVs against liver injury of different etiologies.

Fig. 7.

GNV treatment attenuated binge alcohol-induced fatty liver and inflammation injury in ALD mice. (A) H&E-stained and (B) F4/80-stained images of the liver sections; (C–E) Levels of serum ALT, AST and hepatic TG in the indicated groups; (F–H) Immunoblot results for the markers of (F) oxidative stress (CYP2E1, 3-NT), (G) fibrosis (MMP-9, α-SMA), and (H) apoptosis (p-JNK, cleaved caspase 3, Bax) in the ALD mouse livers in different groups; (I) Hepatic gelatinase/collagenase activity in the indicated groups; (J) RT-PCR analysis results for the hepatic mRNA transcripts of the genes associated with oxidative stress (left panel) and inflammation (right panel) in the different groups. All data present as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 between CON vs EtOH groups; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 between EtOH vs EtOH + 1 mg/kg GNVs groups. Significance of mean values for each group was determined using one-way ANOVA with Dunnett’s multiple comparisons test.

Our results indicate that modulating the gut–liver axis is an effective strategy to combat MASLD and ALD. Common pathological features of these diseases is a “leaky gut” driven by dysbiosis and oxidative damage, leading to translocation of endotoxin (LPS) and other pro-inflammatory agents to the liver [7,38,43,44,57]. GNVs can effectively halt the progression of both metabolic dysfunction- and alcohol-associated liver diseases in experimental models, highlighting their efficacy in alleviating liver injury and fibrosis through preservation of the gut–liver axis and direct hepatic effects.

3.4. Mechanistic insights: TIMP2-Dependent anti-fibrotic pathways

To further examine the molecular mechanisms by which GNVs attenuate liver fibrosis, we performed transcriptomic and genetic analyses by focusing on hepatic stellate cell activation, a central driver of fibrosis. RNA sequencing (RNA-seq) was conducted on LX-2 human HSCs treated with or without GNVs to identify GNVs-responsive genes and pathways. Transcriptomic profiling showed that GNV exposure led to broad changes in gene expression, with numerous mRNAs significantly up- or downregulated relative to untreated cells (Fig. 8A). Gene ontology and KEGG pathway enrichment analysis of the downregulated transcripts showed that GNVs suppressed multiple pathways associated with general metabolic fibrosis and a few degenerative disease pathways, including neurodegeneration, Alzheimer’s disease, amyotrophic lateral sclerosis, and Huntington’s disease (Fig. 8B). Many genes within these pathways are associated with inflammation and fibrosis in metabolic disease contexts. This suggests that GNVs may exert antifibrotic effects in part by attenuating the pathways of cellular stress and injury. Network analysis of fibrosis-related genes highlighted distinct gene interaction clusters in GNVs-treated cells compared to controls, indicating that GNVs cause a concerted shift in the fibrotic gene network (Fig. 8C). TIMP2 has been reported to promote fibrosis. TIMP2 regulates ECM degradation and fibrotic tissue scaling in HSCs through inhibition of metalloproteinase activity [58,59]. However, deficiency of TIMP2 in heart and liver tissues promotes fibrosis [60,61]. These conflicting reports suggest that TIMP2 may have a complex effect on the fibrotic progression of hepatic tissues, depending on the cellular contexts in different fibrosis models. However, most studies showed that TIMP2 generally promotes fibrosis. Therefore, we hypothesized that TIMP2 downregulation is a crucial mechanism in GNVs-mediated anti-fibrotic activity. Confocal imaging showed that untreated LX-2 cells displayed intense intracellular staining for TIMP2. Meanwhile, GNVs-treated LX-2 cells and mouse primary HSC showed a dose-dependent reduction of these fibrosis markers (Fig. 8D&8E). To determine whether the therapeutic effect of GNVs is caused by proteins, small microRNAs, ginsenosides or other physiologically active molecules, LX-2 cells were exposed to GNVs that were pre-treated with RNase or proteinase (Fig. S8). Fibrosis was suppressed in both groups pre-treated with RNase and proteinase. However, substantially greater reductions were observed in the RNase pre-treated group (Fig. S8A&S8B). Therefore, the anti-fibrosis effects of GNVs are due to the complex interplay of proteins, RNAs, and other bioactive substances. However, the major beneficial effects appear to result from their RNA-related components, including microRNAs. GNV exposure significantly reduced the protein levels of TIMP2, α-SMA and MMP-2 (Fig. 8F), further confirming the anti-fibrotic effects of GNVs. RNA-seq analysis showed that the most strongly down-regulated genes by GNVs were key pro-fibrogenic factors such as TIMP2, collagen I (CoL1A1) and transforming growth factor-beta (TGF-β) (Fig. 8G). Quantitative RT-PCR confirmed that GNVs-treated LX-2 cells had significantly lower mRNA levels of TIMP2, COL1A1 and TGF-β than untreated cells (Fig. 8H).

Fig. 8.

GNVs attenuated hepatic fibrosis in a TIMP2-dependent mechanism. (A) Scatter plot of RNA-seq data to depict the values of log2 (normalized data) for all gene transcripts in the CON group (X-axis) vs the 10 µg/ml GNVs-treated group (Y-axis). The red and green dots represent the genes with increased and decreased expression, respectively; (B) Horizontal bar graph for the top five upregulated gene expression pathways with different functions obtained through the KEGG pathway database (I: Metabolic in fibrosis pathways, II: Pathways of neurodegeneration, III: Alzheimer disease, IV: Amyotrophic lateral sclerosis, and V: Huntington disease); (C) RNA interaction network related to the liver fibrosis pathway in LX-2 cells; (D–E) Confocal microscopy images of TIMP2 expression in LX-2 stellate cells and mouse primary HSC. (F) Levels of fibrosis marker proteins (TIMP2, α-SMA, MMP-2) in LX-2 stellate cells in the indicated conditions. Data represents mean ± SD, *P < 0.05, **P < 0.01, ***P < 0.001 between CON and FFC groups; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 between FFC and FFC+0.5 mg GNVs; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 between FFC and FFC + 1 mg/kg GNVs groups. Significance of mean values for each group was determined using one-way ANOVA with Dunnett’s multiple comparison test for (F); (G–H) Gene expression levels analyzed with RNA-seq and qRT-PCR. Significance of mean values for each group was determined using one-way ANOVA with Dunnett’s multiple comparison test. *P < 0.05, **P < 0.01, ***P < 0.001. (I–J) Levels of fibrosis marker proteins and their mRNA transcripts in TIMP2-knockdown LX-2 stellate cells without or with GNVs. *P < 0.05, **P < 0.01, ***P < 0.001 between siCON and siTIMP2 groups; siCON + GNVs group; siTIMP2 + GNVs group. Significance of mean values for each group was determined using one-way ANOVA with Dunnett’s multiple comparison test. All data represent as mean ± SEM.

To further evaluate the functional importance of TIMP2, we conducted TIMP2 knockdown experiments in LX-2 stellate cells. Cells were initially transfected with TIMP2-specific siRNA (siRNA-TIMP2) or a scrambled control siRNA, then treated with GNVs. Silencing TIMP2 alone led to decreased TIMP2 expression and a concomitant reduction in other fibrosis markers, supporting TIMP2’s role in regulating HSC activation (Fig. 8I). GNV treatment in addition to TIMP2 knockdown produced an even greater suppression of fibrogenic proteins (Fig. 8I) and mRNA transcripts (Fig. 8J). We observed comparable results of the anti-fibrotic effects, assessed by α-SMA levels, of GNVs in LX2 cells and mouse primary HSC (Fig. S9A&S9B). TIMP2 is a key regulator of the fibrosis program in stellate cells and that GNVs effectively exert anti-fibrotic effects through a TIMP2-dependent mechanism. By downregulating TIMP2, GNVs likely modulate MMPs, blocking fibrosis progression. This TIMP2/MMP2-mediated modulation of the fibrotic signaling is a crucial, direct hepatoprotective mechanism of GNVs.

3.5. Biodistribution and safety of GNVs

In addition to demonstrating the beneficial effects in MASLD and ALD mouse models, we assessed the biodistribution and safety profiles of orally administered GNVs to further evaluate their translational potentials. Ex vivo imaging showed that orally administered GNVs preferentially target the gut and liver (Fig. S10). Mice were orally administered fluorescently labeled GNVs (DiR dye) and examined at 24 h. IVIS imaging of excised organs showed strong fluorescence in the small intestine and colon, and detectable signals in the liver and, to a lesser extent, the brain of GNVs-treated mice (Fig. 9A). In contrast, systemic administration via intraperitoneal or intravenous injection of labeled GNVs led to a more widespread distribution across many indicated organs (Fig. 9A and S10). This indicates that the oral route naturally concentrates GNVs in the gut–liver axis. In situ confocal microscopy further confirmed that DiR-labeled GNVs were co-localized with the TJ protein ZO-1 in intestinal villi (Fig. 9B). This showed that orally administered GNVs reach and interact with the gut epithelium during transit. Similarly, in the liver, GNVs co-localized with hepatocyte marker protein albumin, which confirmed that orally delivered GNVs are absorbed from the gut and subsequently taken up by liver parenchymal cells (Fig. 9C). We also assessed the stability of GNVs under gastrointestinal (GI) conditions. GNVs after incubation with a simulated gastric fluid (pH 2.0) or GI fluid (pH 6.5) remained at 147.2 ± 2.6 nm in diameter and 3.28 × 10⁸ particles/ml for the pH 2.0 group while their sizes similar at 145.0 ± 1.7 nm with slightly less 1.37 × 10⁹ particles/ml for the pH 6.5 group (Fig. 11A&11B). GNVs were readily absorbed into LX-2 cells and showed anti-fibrotic activity, which indicated that GNVs are relatively stable to withstand harsh pH changes in the GI tract and enzymatic conditions (Fig. S11C&S11D). In vitro uptake assays showed efficient internalization of GNVs by relevant target cells. GNVs were labeled separately for their membrane lipids, proteins and RNAs using PKH67 lipophilic membrane dye, the ExoGlow-Protein EV Labeling Kit, and the ExoGlow RNA EV Labeling Kit, respectively. They were then incubated with various cell lines (Fig. S12). Confocal imaging showed that mouse hepatocytes (AML12), human colon epithelial cells (T84), and human HSC (LX-2) all readily took up GNVs, with the vesicles visible inside the cells (Fig. 9D–9F), suggesting a broad cellular uptake capability. These findings highlight that orally delivered GNVs can reach target organs and cells in the gut-liver axis, supporting their feasibility as an oral, nanotherapeutic platform. To find out the absorption mechanism of GNVs, we used GNVs after pre-treatment of MβCD to destroy the caveolae structure [62]. The absorption rates of GNVs in the MβCD pre-treated group decreased in T84 and LX-2 cells (Fig. S13). These results confirm that GNVs were absorbed via the caveolae-dependent endocytosis mechanism.

Fig. 9.

In vivo distribution of GNVs and in vitro cellular uptake of GNVs. (A) IVIS ex vivo images of various tissues, including small intestine, large intestine, brain, lung, heart, liver, spleen, kidney and lymph nodes, evaluated at 24 h after oral, IP or IV administration of DiR-labeled GNVs; (B) Small intestine sections immunofluorescent labeled with ZO-1 (Green) after treatment of the DiR-labeled GNVs (red, arrows for enlarged areas). Scale bar: 100 µm; (C) Liver sections immunofluorescent labeled with a hepatic protein albumin (green) after treatment of the DiR-labeled GNVs (red, arrows for enlarged areas). Scale bar: 200 µm; (D–F) mouse hepatocytes (AML12), human colon epithelial cells (T84), and human HSC (LX-2) were used to study the uptake and distribution of PKH67-labeled membrane lipids, proteins and RNAs of GNVs. Scale bars: 200 µm. Efficient uptake of GNVs by the indicated cells including LX-2, AML12 and T-84 cells.

GNVs exhibited an excellent safety profile in vivo. We conducted a sub-acute toxicity study in healthy C57BL/6 J mice by administering a high dose of GNVs (1–50 mg/kg/d by oral gavage) for seven consecutive days (Fig. 10A). Throughout the experimental period, all the mice remained healthy and did not show abnormal clinical signs, changes in body weight, or alterations in diet intake (data not shown). Throughout the treatment period, mice showed no signs of distress or illness. Their body weights and food intake remained normal with no significant changes compared to water-treated vehicle controls. Histological examination of major organs (liver, kidney, brain, heart, spleen and lung) on Day 8 showed no evidence of tissue damage or abnormality in GNVs-treated animals up to 50 mg/kg/d (Fig. 10B). The liver histology of GNVs-only mice showed normal histology without signs of inflammation or necrosis despite the liver being a primary accumulation site for GNVs. Consistent with the histology, blood biochemical analyses indicated no hepatic or renal toxicity from GNVs treatment. Serum ALT and AST, as well as hepatic TG and high-density lipoprotein levels, were unchanged by GNV administrations relative to controls (Fig. 10C). Kidney function markers (BUN and serum creatinine) and blood glucose were also comparable between GNVs-treated and control groups (Fig. 10C). All the measured parameters fell within normal physiological ranges, and no statistical differences were noted (one-way ANOVA, n.s. for all comparisons). These results confirm that orally administered GNVs are well-tolerated at doses equal to or exceeding those used for therapeutic efficacy, causing no detectable systemic toxicity or organ damage. The biodistribution and safety profiling data demonstrate that GNVs selectively accumulate in the gut and liver upon oral administration without inducing adverse effects. This reinforces their promise as a safe, oral therapeutic for effectively managing liver disease.

Fig. 10.

In vivo toxicity evaluation of orally administered GNVs. (A) Schematic experimental procedure to evaluate potential cytotoxicity of orally administered GNVs; (B) Images of H&E-stained liver, kidney, brain, heart, spleen and lung sections from control and GNVs-treated male mice. Scale bars are depicted in each image; (C) Levels of liver function-related parameters (serum ALT, AST, hepatic TG and HDL-C), and kidney function-related markers (BUN, serum creatinine and glucose) from control and GNVs-treated male mice. Data represents as mean ± SEM. Significance of mean values for each group was determined using one-way ANOVA with Dunnett’s multiple comparison test. NS: not significantly different. (HDL-C: high-density lipoprotein cholesterol concentration; BUN: blood urea nitrogen).

4. Conclusions

We have demonstrated that nanovesicles derived from edible ginseng can alleviate liver injury in MASLD and ALD mouse models. GNVs exerted these effects by suppressing oxidative stress, nitration, restoring intestinal integrity, reshaping the gut microbiome, reducing hepatic lipid accumulation, and attenuating fibrogenesis through TIMP2-dependent pathways. Given their natural origin, biodistribution with excellent safety profile, and oral efficacy, GNVs represent a promising therapeutic and drug delivery platform for effectively managing chronic liver diseases. The findings provide a solid foundation for developing PENVs-based interventions targeting the gut–liver axis and opens new avenues to innovative nanomedicine strategies for treating liver fibrosis and related pathologies.

Conflicts of interest

The authors declare that there is no conflicts of interest.