Sleeve Gastrectomy Alters Exosomal miR-497-5p Cargo to Ameliorate Metabolic Dysfunction–Associated Steatotic Liver by Targeting GABARAPL1

Introduction

Obesity, recognized by the World Health Organization as a global chronic disease, results from complex genetic-environmental interactions and has emerged as a major public health issue.¹ It is linked to over 200 chronic conditions—including type 2 diabetes, metabolic dysfunction-associated steatotic liver disease (MASLD), cardiovascular and cerebrovascular diseases, and certain cancers—which collectively increase mortality risk.^2,3 In individuals with severe obesity, the incidence of complications, disability, and mortality can be several to dozens of times higher than in non-obese populations.⁴ Among the complications of obesity, MASLD is the most prevalent hepatic manifestation, closely associated with caloric excess, insulin resistance, and systemic metabolic dysregulation.^5,6 It encompasses a spectrum from simple steatosis to metabolic dysfunction-associated steatohepatitis (MASH), which can progress to fibrosis, cirrhosis, and hepatocellular carcinoma. Approximately one-quarter of the global population is affected, and its prevalence continues to rise with the obesity epidemic.^7–9 Notably, MASLD exhibits a pronounced characteristic of multi-system involvement, which not only leads to adverse hepatic outcomes but also significantly elevates the risk of extrahepatic complications such as cardiometabolic diseases.^10–12 Furthermore, MASLD independently increases the risk of fatal and non-fatal cardiovascular events by approximately 1.5-fold, regardless of traditional risk factors. This risk is further amplified (to approximately 2.5-fold) with disease progression, particularly among individuals with advanced fibrosis.¹³ Conventional lifestyle and pharmacological treatments are often ineffective in severe obesity. Laparoscopic sleeve gastrectomy (LSG) has emerged as an effective intervention for obesity with metabolic comorbidities and is recommended by authoritative bodies, such as the International Diabetes Federation, due to its sustained weight loss and systemic metabolic benefits.^14,15 In addition to mechanical restriction, growing evidence suggests that exosomes and their bioactive cargos—particularly microRNAs (miRNAs)—play a role in the metabolic improvements observed following bariatric surgery.¹⁶

Exosomes are 40–100 nm vesicles that transfer proteins, metabolites, and nucleic acids to recipient cells, facilitating intercellular communication.^17–20 They are involved in obesity pathogenesis by shuttling metabolites and miRNAs across pancreatic β-cells, adipose tissue, skeletal muscle, and the liver.^21–23 Notably, bariatric surgery-induced remodeling of circulating exosomes has been linked to metabolic improvements. For instance, Bae et al demonstrated that LSG significantly altered 72 serum exosomal miRNAs (including miR-122-5p, miR-193b-5p, and miR-26b-3p), suggesting that exosomal miRNAs mediate surgery-induced systemic effects.²⁴ In a MASLD rat model, targeting adipose-derived exosomal miR-122 reversed hepatic dysregulation of glucose and lipid metabolism, apoptosis, inflammation, and fibrosis.²⁵ Similarly, the delivery of miR-122 mimics via mesenchymal stem cell-derived exosomes alleviated liver fibrosis by activating the IGF1R/CCNG1/P4HA1 pathway in hepatic stellate cells.²⁶

Despite these advances, significant gaps in knowledge remain. First, while the burden of MASLD is well-recognized, the precise mechanisms underlying its remission after metabolic surgery are not fully elucidated. Second, most existing studies rely on animal models or cross-sectional human analyses, lacking longitudinal data that capture the dynamic changes in plasma exosomal components following LSG in humans. Third, the functional role of these dynamically altered exosomal cargos, especially miRNAs, in driving MASLD remission is poorly understood. Consequently, there is a critical need for longitudinal human studies to comprehensively characterize how LSG remodels the plasma exosomal landscape and to functionally link these changes to MASLD improvement.

This study profiled plasma exosomal miRNAs before and after LSG using high-throughput sequencing, identified key miRNAs associated with MASLD improvement through integrative analyses, and validated their roles in hepatocellular lipid metabolism using in vitro models. Our findings contribute to a model wherein the exosomal miR-497-5p-GABARAPL1 axis may participate in the systemic metabolic benefits of LSG, highlighting its potential as both a biomarker and a therapeutic avenue for MASLD.

Methods

Study Population and Ethical Approval

Study Population and Ethics

This prospective study enrolled patients undergoing LSG at Beijing Shijitan Hospital between September 2022 and September 2024. All procedures were performed by a single specialized surgical team. The study protocol was approved by the Institutional Ethics Committee of Beijing Shijitan Hospital (sjtkyll-lx-2022(076)) and adhered to the Declaration of Helsinki. Written informed consent was obtained from all participants.

Inclusion criteria were as follows: (1) BMI meeting the 2024 Chinese Guidelines for metabolic surgery; (2) voluntary participation with consent; (3) no previous abdominal surgery; (4) normal cardiopulmonary and renal function without significant organic disease; (5) total weight loss (TWL%) ≥ 20% at 1-year follow-up; (6) male participants aged 18–45 years. Exclusion criteria were as follows: (1) incomplete preoperative or follow-up data; (2) perioperative complications requiring reoperation (eg., fistula, hemorrhage); (3) gastrointestinal surgery during follow-up; (4) substance abuse, alcohol dependence, uncontrolled psychiatric disorders, or poor protocol compliance.

Healthy controls were male (18–45 years) with normal BMI (18.5–23.9 kg/m²), no history of MASLD, normal laboratory results and physical examination, no medications affecting hepatic or metabolic function, and provided written consent. Exclusion criteria for controls included overweight/obesity, MASLD, liver or metabolic disease history, medications affecting hepatic or metabolic parameters, severe cardiovascular, renal, or systemic diseases, or smoking/alcohol abuse.

Fasting venous blood was collected under calm conditions. Plasma was separated by centrifugation at 3,000 rpm for 15 minutes at 4°C (Beckman Coulter) and stored at −80°C (Thermo Fisher Scientific) until analysis.

Exosome Isolation and Characterization

Plasma exosomes were isolated by differential ultracentrifugation (Hitachi). Briefly, samples were centrifuged at 2,000 × g for 30 minutes and 10,000 × g for 45 minutes, followed by filtration through a 0.45-μm membrane (JET BIOFIL) and ultracentrifugation at 100,000 × g for 70 minutes. Pellets were washed with PBS, ultracentrifuged again at 100,000 × g, and then resuspended and stored at −80°C.

For transmission electron microscopy (TEM), 10 μL of exosome suspension was placed on Formvar/carbon-coated copper grids, incubated for 1 minute, negatively stained with 2% uranyl acetate, air-dried, and imaged at 80 kV (Hitachi, Japan). Protein content was quantified using a BCA assay (Beyotime). Western blot analysis was performed to assess exosomal markers CD63 (Abclonal) and TSG101 (Abcam), with Calnexin (SAB Biotherapeutics) used as a negative control marker.

Western Blot

Proteins (20 μg) were mixed with 5 × SDS–PAGE loading buffer, denatured at 100°C for 5 minutes, and separated on 12% resolving/5% stacking gels. Electrophoresis was conducted at 80 V until samples entered the resolving gel, followed by 100 V until the dye front reached the end of the gel. Proteins were transferred to methanol-activated PVDF membranes (Merck; activation for 20 seconds) at 300 mA for 30 minutes in ice-cold transfer buffer. Membranes were blocked with 5% (w/v) non-fat milk in TBST for 1 hour at room temperature, then incubated overnight at 4°C with primary antibodies (1:1000 in TBST). After washing three times for 10 minutes, membranes were incubated with HRP-conjugated secondary antibodies (Invitrogen; 1:5000 in TBST) for 1 hour at room temperature, followed by washing and development using ECL (Shanghai Qinxin) on a chemiluminescence imager (Shanghai Qinxin). Brightness/contrast adjustments were applied uniformly across the entire image.

RNA Extraction and Quality Control

Total RNA from exosomes was extracted using the miRNeasy Mini Kit (Qiagen). RNA concentration was determined with a Quantus Fluorometer (Promega), and integrity was assessed using the Qsep100 (Bioptic). Libraries were prepared with the QIAseq miRNA Library Kit (Qiagen), following adapter ligation, cDNA synthesis, and PCR amplification. Libraries were purified with magnetic beads and quality-checked with Qubit 4.0 and the Agilent 2100 Bioanalyzer. Sequencing was performed on an Illumina platform. Quality control and filtering were performed with FastQC and fastp. Clean reads were quantified against miRBase using miRDeep2. Differential expression analysis was conducted using DESeq2 (|log₂FC| ≥ 1; FDR-adjusted P < 0.05). Weighted Gene Co-expression Network Analysis (WGCNA) and random forest models in R (“WGCNA”, “randomForest”) were used to identify key miRNAs. Target genes were predicted using miRDB,²⁷ miRTarBase,²⁸ TargetScan,²⁹ miRanda,³⁰ and DIANA-microT-CDS.³¹ Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed with clusterProfiler. Two liver biopsy transcriptome datasets (GSE106737³² and GSE83452³³) before and after bariatric surgery were retrieved from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). Raw data underwent background correction, normalization, and probe summarization using RMA. Batch effects were assessed by boxplots and principal component analysis (PCA). A standardized matrix of 49 paired samples was obtained. Differential expression analysis (limma; |log₂FC| > 0.585; FDR-adjusted P < 0.05) identified differentially expressed genes (DEGs), which were then compared with miRNA-predicted targets.

Dual-Luciferase Reporter Assay for miRNA–Target Validation

The pmirGLO vector was used to clone wild-type (WT) 3′UTR sequences containing the predicted binding site (5′-…GCTCCCAG…-3′) and mutant (MUT) sequences with seed-region substitutions (5′-…CGAGGGTC…-3′) using SacI and SbfI. Inserts were confirmed by Sanger sequencing. After digestion, purification, PCR amplification, and ligation, constructs were transformed into TOP10 competent cells, and positive colonies were verified by sequencing. HEK293T cells were cultured in high-glucose DMEM, seeded 24 hours before transfection, and co-transfected with plasmids and miRNA mimics using Lipofectamine 3000. The medium was refreshed after 6 hours. At 24–48 hours post-transfection, Firefly and Renilla luciferase activities were measured, and the Firefly/Renilla ratio was used to assess miRNA-target interactions. Each condition was tested in triplicate with appropriate controls.

Cell Culture and Transfection

HepG2 cells (Wuhan E-Bio) were cultured in MEM (89%) supplemented with FBS (10%; Thermo Fisher) and penicillin–streptomycin (1%; Wuhan Saiweier). Transfections were performed with hsa-miR-497-5p mimics or inhibitors (MedChemExpress) using HY-K2017 (MedChemExpress). After 6 hours, the medium was replaced, and cells were harvested 48 hours post-transfection.

For MASLD modeling, a 100 mM free fatty acid (FFA) stock was prepared by dissolving oleic acid (28.2 mg) and palmitic acid (25.6 mg) in 1 mL DMSO (OA:PA = 2:1), sterile-filtered (0.22 μm), diluted into complete medium, and used immediately. Cells were treated with 0.5 mM FFA for 24 hours. Experimental groups included: HepG2-FFA (FFA only), HepG2-FFA-Inhibitor (FFA + miR-497-5p inhibitor), and HepG2-FFA-Mimic (FFA + miR-497-5p mimic).

Biochemical Analysis

Cells were washed twice with PBS and lysed in isopropanol (5 × 10⁷ cells/mL). After ultrasonication and centrifugation (12,000 × g for 5 minutes at 4°C), supernatants were collected. Intracellular triglyceride (TG) and total cholesterol (TC) levels were measured using commercial kits (Shanghai Biotium), with standard curves (TG: 0–10 mM; TC: 0–500 μM). Samples and standards were assayed in duplicate.

Oil Red O Staining

For histological analysis, cells were fixed in 4% paraformaldehyde for 20 minutes, stained with Oil Red O (15 minutes), washed, and imaged at 20 × magnification. Five random fields per sample were quantified for integrated optical density (IOD) using Image-Pro Plus v6.

Statistical Analysis

Continuous variables are presented as mean ± SD or median [IQR], as appropriate, and categorical variables as n (%). Two-group comparisons were made using two-tailed t-tests or Mann–Whitney U-tests. For comparisons among ≥3 biological replicates, one-way ANOVA with Tukey’s post-hoc test or Kruskal–Wallis with Dunn’s correction was applied. Sequencing analyses controlled for the false discovery rate using the Benjamini–Hochberg method. Analyses were performed using R (v4.3.1) or GraphPad Prism (v9.5.1). A P-value of <0.05 was considered statistically significant.

Results

Isolation and Characterization of Plasma Exosomes

Plasma exosomes isolated by ultracentrifugation displayed typical round or oval morphology with clear bilayer membranes on TEM (Figure 1A–C). Western blot analysis confirmed the presence of exosomal markers CD63 and TSG101, while the endoplasmic reticulum marker Calnexin was absent (Figure 1D), indicating high purity of the exosome preparation.

Figure 1 Morphological and molecular characterization of plasma exosomes. (A–C) TEM images of plasma-derived exosomes from healthy controls, pre-LSG, and post-LSG patients (scale bar, 100 nm). (D) Western blot detecting exosomal markers CD63 and TSG101, and the absence of the endoplasmic reticulum marker Calnexin.

Quality Assessment of Exosomal RNA

Exosomal RNA from healthy controls (H1–H6), pre-LSG (Pre-LSG 1–6), and post-LSG (Post-LSG 1–6) samples exhibited a predominant enrichment of small RNAs in the 18–25 nt range, as shown in Qsep100 electropherograms, confirming a miRNA-rich profile suitable for downstream analyses (Supplementary Figure 1).

Differential Expression of Exosomal miRNAs

Differential expression analysis of plasma exosomal miRNAs associated with obesity and LSG revealed 298 differentially expressed miRNAs (DEmiRNAs) when comparing Pre-LSG to healthy controls (291 upregulated, 7 downregulated). Comparison of Post-LSG to Pre-LSG identified 333 DEmiRNAs (4 upregulated, 329 downregulated) (Figure 2A and B). Overall, exosomal miRNA levels were significantly reduced post-LSG, indicating extensive metabolic remodeling.

Figure 2 Differential expression of exosomal miRNAs and WGCNA modules. (A) Volcano plot of differentially expressed miRNAs (DEmiRNAs) in pre-LSG versus healthy controls. (B) Volcano plot of DEmiRNAs in post-LSG versus pre-LSG. Gray indicates not significant; red, upregulated; blue, downregulated (thresholds as defined in Methods). (C) Scale-free topology fit index (R2) and mean connectivity across soft-threshold powers (β). (D) Hierarchical clustering dendrogram with module assignment. (E) Module–trait correlation heatmap (Pearson r, two-sided P-values). (F) Correlation between gene significance and module membership within the blue module.

Co‑expression Network Analysis

WGCNA identified seven co-expression modules at a soft-threshold power of 9 (Figure 2C and D). The blue module was positively correlated with the Post-LSG group (cor = 0.42), while the grey module showed a negative correlation (cor = −0.42) (Figure 2E). Within the blue module, module membership correlated with gene significance (cor = 0.18; P = 6.5×10⁻³), supporting the biological relevance of this module to LSG-associated processes (Figure 2F).

Key Exosomal miRNAs Prioritized by Machine Learning

Sixty-five miRNAs were found to overlap among the Pre-LSG vs. healthy controls, Post-LSG vs. Pre-LSG, and the blue module (Figure 3A). Random-forest modeling prioritized six key miRNAs with a Mean Decrease Gini > 0.4: miR-186-5p, miR-760, miR-320d, miR-154-5p, miR-4508, and miR-497-5p (Figure 3B and C). Target prediction across five databases identified 44 consensus targets for miR-497-5p (Figure 3D) and one for miR-154-5p (Figure 3E), while no overlapping targets were found for the other miRNAs (Figure 3F–H). Gene ontology (GO) and KEGG enrichment analyses implicated pathways related to immune regulation, inflammatory response, autophagy, and metabolism, including RIG-I signaling, type I interferon production, lysosomal targeting, autophagosome assembly, mitophagy, p53, TGF-β, and AMPK pathways (Figure 3I).

Figure 3 Prioritization of key exosomal miRNAs. (A) Venn diagram showing overlapping DEmiRNAs among comparisons and the blue module. (B) Out-of-bag error rate of the random-forest model across increasing numbers of trees. (C) Top-30 miRNAs ranked by importance (Mean Decrease Gini). (D–H) Consensus predicted target genes across five databases (DIANA-microT, miRanda, miRTarBase, TargetScan, miRDB) for miR-497-5p (D), miR-154-5p (E), miR-186-5p (F), miR-760 (G), and miR-4508 (H). (I) GO and KEGG enrichment of predicted targets; bars indicate enrichment in biological processes, molecular functions, and KEGG pathways; x-axis, −log10(FDR).

Integration with Postoperative Liver Transcriptomes Identifies GABARAPL1

Integration of 49 paired pre- and post-LSG liver transcriptomes (GSE106737, GSE83452), after RMA normalization and ComBat batch correction, revealed comparable medians and PCA distributions (Supplementary Figure 2). Differential expression analysis (|log₂FC| > 0.585; FDR < 0.05) identified 110 DEGs, with 23 upregulated and 87 downregulated (Figure 4A and B). Intersection of these DEGs with miRNA-predicted targets highlighted GABARAPL1 as a predicted target of miR-497-5p, which was significantly upregulated post-LSG (Figure 4C and D). Dual-luciferase assays confirmed specific binding of miR-497-5p to the GABARAPL1 3′UTR, leading to repression of reporter activity (P < 0.01), which was abolished by seed-site mutation (Figure 4E).

Figure 4 Integration with paired liver transcriptomes and validation of the miR-497-5p–GABARAPL1 axis. (A) Volcano plot of differentially expressed genes (DEGs) in paired pre-/post-LSG liver transcriptomes. (B) Heatmap of DEG expression across 49 paired samples. (C) Intersection of DEGs with miRNA-predicted targets highlighting GABARAPL1. (D) Schematic of the miR-497-5p seed-match site in the GABARAPL1 3′UTR. (E) Dual-luciferase reporter assay in HEK293T cells showing repression of the wild-type (WT) reporter by miR-497-5p and loss of repression with seed-site mutation (MUT); Firefly/Renilla normalized; Data are presented as mean ± SD, n = 3; two-tailed t-test; significance as indicated. ** P < 0.01, ns indicates no statistically significant difference.

miR‑497‑5p Downregulation Alleviates FFA‑induced Lipid Accumulation via GABARAPL1

In FFA-treated HepG2 cells, Oil Red O staining revealed prominent lipid droplet accumulation, confirming successful induction of steatosis (Figure 5A). Inhibition of miR-497-5p reduced the number and intensity of lipid droplets compared to FFA alone (Figure 5B), while overexpression of miR-497-5p exacerbated lipid deposition (Figure 5C). Quantification showed a significant increase in lipid area with FFA treatment compared to controls (P < 0.05), a reduction with the inhibitor (P < 0.05), and an augmentation with the mimic (P < 0.05) (Figure 5D). Intracellular TC and TG levels increased with FFA treatment, while miR-497-5p inhibition decreased both (P < 0.0001), and mimic treatment increased both (P < 0.05) (Figure 5E and F). qRT-PCR and Western blot analysis revealed elevated GABARAPL1 mRNA and protein levels in FFA-treated cells compared to controls (P < 0.05), with further increases following miR-497-5p inhibition (P < 0.05), consistent with the negative regulation of GABARAPL1 by miR-497-5p (Figure 5G and H).

Figure 5 miR-497-5p modulates FFA-induced lipid accumulation via GABARAPL1. (A–C) Oil Red O staining of HepG2 cells treated with FFA alone, FFA + miR-497-5p inhibitor, or FFA + miR-497-5p mimic (20 ×). (D) Quantification of lipid droplet area percentage. (E) Intracellular total cholesterol (TC). (F) Intracellular triglycerides (TG). (G) Relative GABARAPL1 mRNA expression levels determined by qRT-PCR. (H) Protein expression of GABARAPL1 by Western blot with densitometric analysis. (I) Statistical analysis was performed by one-way ANOVA with Tukey’s post-hoc test (n = 3). **** P < 0.0001, *** P < 0.001, ** P < 0.01, * P < 0.05.

Discussion

This study identified a circulating exosomal pathway through which LSG may improve MASLD. By integrating plasma exosomal miRNA sequencing from a prospective cohort with network analysis and machine learning, and cross-referencing the results with paired human liver transcriptomes, miR-497-5p was identified as a key hub miRNA. Its postoperative decline corresponds with derepression of the autophagy adaptor gene GABARAPL1. Direct miR-497-5p–GABARAPL1 binding was validated, and suppression of miR-497-5p was shown to reduce lipid accumulation in a hepatocellular steatosis model, while increasing GABARAPL1 expression at both mRNA and protein levels. These findings support a model in which postoperative remodeling of exosomal cargo contributes to the restoration of hepatocellular lipophagy and metabolic homeostasis following LSG.

A central observation is the global reduction of circulating exosomal miRNA levels after surgery, with the post-LSG miRNA profile approaching that of healthy controls. This pattern aligns with previous reports indicating that bariatric surgery reprograms circulating extracellular vesicles (EVs) and their cargo.^24,34–36 One plausible mechanism is that improved insulin sensitivity and reduced adipose tissue hypoxia post-LSG dampen EV release and/or miRNA loading from adipose tissue and other metabolic organs.^37–39 Additionally, bile acid signaling and gut-derived hormones (eg., GLP-1) that surge after LSG may modulate miRNA processing and EV biogenesis pathways in a tissue-specific manner, reshaping the circulating “miRNome”.⁴⁰ Although correlative, these findings provide a coherent framework where exosomal downregulation relieves repression of metabolic effectors in target tissues.

Within this global shift, multistep prioritization identified six miRNAs, with miR-497-5p emerging as the most robust target, including GABARAPL1. The identification of GABARAPL1 is biologically relevant. As a member of the ATG8/LC3 family, GABARAPL1 facilitates autophagosome formation, cargo recognition, and membrane dynamics, playing a pivotal role in lipophagy—the selective autophagic degradation of lipid droplets.^38–44 Lipophagy mobilizes triglycerides for β-oxidation, alleviating steatotic burden and lipotoxicity. Disruptions in autophagic flux and lipophagy are increasingly implicated in MASLD pathogenesis, with context-dependent activation in early stages, followed by impairment under sustained lipid overload.^{39,40,42,43,45} Our integration of 49 paired human liver transcriptomes shows that GABARAPL1 is upregulated after bariatric surgery, consistent with the restoration of autophagic competence in the postoperative liver. The convergence of this tissue-level signal with the circulating exosomal decline of miR-497-5p, coupled with direct binding confirmed by luciferase assays, reinforces the existence of a causal regulatory axis.

Functional studies in FFA-challenged HepG2 cells further support this regulatory axis. Inhibition of miR-497-5p reduces intracellular triglyceride and cholesterol levels, decreases Oil Red O-positive lipid droplets, and increases GABARAPL1 expression. Conversely, miR-497-5p mimic exacerbates steatosis. In this context, miR-497-5p functions as a brake on GABARAPL1 expression: when miR-497-5p levels decrease—such as in the postoperative state—GABARAPL1 is derepressed, promoting lipophagic clearance and improving hepatocellular lipid handling. The source of circulating miR-497-5p post-LSG remains undefined. Adipose tissue is a likely candidate due to its reduction in mass, inflammatory remodeling, and known EV output, but contributions from skeletal muscle, the intestine, and immune cells are also possible. Investigating tissue-specific origins using cell-type-specific EV markers, RNA barcoding, or isotope/label tracking of patient-derived EVs could clarify this axis and identify additional therapeutic targets.

These findings have translational relevance. First, circulating exosomal miR-497-5p could serve as a minimally invasive biomarker for monitoring hepatic recovery after LSG. The extent and trajectory of its decline may predict the restoration of lipophagy and remission of MASLD, complementing imaging and serum aminotransferase assessments. Second, therapeutically, antagonizing hepatic miR-497-5p could partially replicate the metabolic benefits of surgery by relieving repression of GABARAPL1 and potentially other autophagy-related targets. In addition to miR-497-5p, other prioritized miRNAs (miR-186-5p, miR-760, miR-320d, miR-154-5p, miR-4508) may also contribute to postoperative metabolic remodeling. While their downstream targets necessitate further experimental validation and the consensus target sets exhibit limited consistency across databases, the possibility of additive or context-dependent effects remains plausible.

Several limitations should be considered. First, the association between circulating EVs and hepatic GABARAPL1 upregulation is based on correlative evidence, and future experimental validation through EV-cell interaction studies would strengthen the causal interpretation. Second, our in vitro model relies on HepG2 cells exposed to acute FFAs, which, while capturing steatotic features, do not fully replicate the complexity of MASLD, including lipotoxicity-induced apoptosis, stellate cell activation, and immune infiltration. Third, the use of an all-male cohort, while mitigating potential confounding from hormonal cyclical variation, inherently limits the generalizability of our conclusions to females. This underscores the need for validation in sex-balanced populations. Additionally, liver biopsies were not obtained from participants, and instead, public postoperative liver datasets were used.

Conclusion

In conclusion, LSG remodels circulating exosomal miRNAs. Postoperative downregulation of exosomal miR-497-5p derepresses hepatic GABARAPL1, enhancing lipophagy and reducing lipid accumulation. miR-497-5p is thus a potential biomarker and therapeutic target.