Small extracellular vesicles as liquid biopsy targets and precision therapeutics for metabolic dysfunction-associated steatotic liver disease

Abstract

Metabolic dysfunction-associated steatotic liver disease (MASLD) affects over one-third of the global population, yet robust molecular biomarkers for early disease activity and treatment monitoring remain limited. Small extracellular vesicles (sEVs) carry nucleic acids, proteins, and lipids that reflect disease-associated intercellular signaling and are being investigated as liquid-biopsy analytes and therapeutic platforms. Candidate sEV miRNAs, proteins, and lipid signatures have shown potential for disease characterization, but evidence derives mainly from small, heterogeneous cohorts with limited external validation. No sEV-based assay is established for routine MASLD diagnosis, staging, or treatment monitoring. Mesenchymal stem cell-derived sEVs and engineered sEV platforms have shown therapeutic potential in preclinical MASLD-related models through modulation of inflammation, oxidative stress, lipid metabolism, and targeted cargo delivery, although some engineering strategies remain supported primarily by non-MASLD proof-of-concept evidence. Key barriers to translation include heterogeneity in extracellular vesicle isolation and characterization, lipoprotein co-isolation, prospective multicenter validation, and, for therapeutic products, manufacturing consistency, biodistribution, dosing, and safety. This narrative review integrates diagnostic and therapeutic evidence within an evidence-classified translational framework.

Key Words: Metabolic dysfunction-associated steatotic liver disease; Small extracellular vesicles; Extracellular vesicles; Liquid biopsy; Biomarkers; Noninvasive tests; Drug delivery systems; Therapeutic strategies

Core Tip: Metabolic dysfunction-associated steatotic liver disease (MASLD) lacks non-invasive biomarkers and targeted therapies. Small extracellular vesicles (sEVs) offer dual solutions: As diagnostic tools, sEV-derived miRNAs (e.g., miR-122), proteins (e.g., Kelch-like protein 41), and lipids enable accurate staging; as therapeutics, mesenchymal stem cell-derived sEVs reduce steatosis and fibrosis via autophagy and oxidative stress regulation, while engineered sEVs (e.g., cyclic arginine-glycine-aspartic acid-modified) enhance targeted delivery. sEVs thus shift from bystanders to key diagnostic and therapeutic players in MASLD.

INTRODUCTION

As the hepatic event of metabolic syndrome, nonalcoholic fatty liver disease (NAFLD) is associated with obesity, diabetes and features of the metabolic syndrome, such as hypertension, dyslipidemia and insulin resistance[1,2]. With the increasing prevalence of obesity and diabetes worldwide, the prevalence of NAFLD is estimated to approximately 25%, which has become a public health threat[3]. Given its strong association with metabolic dysfunction, the nomenclature of NAFLD has been updated to metabolic dysfunction-associated steatotic liver disease (MASLD). Correspondingly, its progressive subtype, non-alcoholic steatohepatitis (NASH), is now termed metabolic dysfunction-associated steatohepatitis (MASH)[4]. The transition from NAFLD/NASH to MASLD/MASH reflects an updated disease framework that more explicitly incorporates metabolic dysfunction into disease nomenclature and definition[5].

Current clinical assessment increasingly relies on sequential non-invasive testing (NIT) for fibrosis risk stratification rather than routine liver biopsy[6]. For instance, plasma aminotransferases, fibrosis-4 (FIB-4) score and imaging tests, including ultrasound, computed tomography or magnetic resonance imaging, are used in clinical practice to evaluate MASLD[5,7]. While NITs are central to routine risk stratification and can help identify patients at risk of advanced fibrosis, they cannot directly assess microscopic features such as hepatocellular ballooning or lobular inflammation. In most patients, liver biopsy is not required for routine MASLD management; however, it remains the reference standard for histological assessment of steatohepatitis and fibrosis in selected clinical contexts and may also help exclude alternative liver diseases[8,9]. Its invasiveness, sampling variability, and limited suitability for repeated monitoring have driven the search for complementary non-invasive biomarkers. While weight loss is currently recommended as the first-line therapy for MASLD, this approach is not universally suitable for all patients. In 2024, the United States Food and Drug Administration approved Resmetirom as the first pharmacological agent for the treatment of NASH with liver fibrosis[10]. Specifically, Resmetirom is indicated for adults with non-cirrhotic MASH accompanied by moderate to advanced liver fibrosis (stages F2 to F3), rather than for all stages of MASLD. Given the escalating global burden of MASLD and its progression to advanced liver disease, there is an urgent clinical need for robust non-invasive diagnostic tools and effective therapeutic strategies.

While recent advances in imaging and serum biomarkers have improved risk stratification, significant gaps remain in accurately detecting early disease activity and monitoring treatment response at the molecular level. In this context, small extracellular vesicles (sEVs) have emerged as a novel conduit of intercellular communication, reflecting the pathophysiological state of their parent hepatocytes and immune cells[11]. However, the rapidly evolving landscape of sEV research has revealed that these vesicles are far more than passive reflections of disease status. By transporting miRNAs, proteins, lipids and other bioactive cargos, sEVs actively mediate communication among hepatocytes, immune cells, hepatic stellate cells and extrahepatic metabolic tissues, thereby contributing to lipotoxicity, inflammation, fibrosis and metabolic remodeling in MASLD[12-14]. At the same time, their unique biological properties have highlighted their value as liquid biopsy targets and promising therapeutic platforms. Accordingly, this narrative review integrates sEVs as pathogenic mediators, candidate biomarkers, native therapeutic agents, and engineered delivery platforms within a MASLD-focused framework. Particular emphasis is placed on the directness and maturity of the supporting evidence, the incremental value of sEV biomarkers beyond established NITs, and the methodological and translational barriers that currently limit clinical implementation.

LITERATURE SEARCH STRATEGY

A structured literature search was conducted to inform this narrative review using PubMed, Web of Science, and Scopus to identify literature relevant to sEVs in MASLD. The search was performed on May 25, 2026, and included publications available up to that date. The search strategy combined terms related to extracellular vesicles (EVs), MASLD, and their diagnostic and therapeutic applications. The following search terms were used across the three databases, with database-specific syntax adapted accordingly: (“small extracellular vesicles” OR “extracellular vesicles” OR “exosomes”) AND (“MASLD” OR “MASH” OR “MAFLD” OR “NAFLD” OR “NASH” OR “metabolic dysfunction-associated steatotic liver disease”) AND (“biomarker” OR “diagnosis” OR “liquid biopsy” OR “therapy” OR “drug delivery” OR “engineered extracellular vesicles”). Only English-language publications were included. Studies were considered relevant if they investigated the biogenesis, pathological roles, diagnostic applications, therapeutic effects, or engineered delivery strategies of EVs in the context of MASLD and related metabolic liver diseases. Conference abstracts, editorials, case reports, and studies without sufficient relevance to liver or metabolic disease were not considered. Potentially relevant publications were identified by reviewing titles and abstracts and consulting the full text where necessary. This process was intended to support thematic coverage of the narrative review rather than systematic or PRISMA-compliant study selection. The literature was reviewed and synthesized according to the thematic scope of this narrative review, with findings from human studies, animal models, and in vitro experiments considered separately where appropriate. Historical disease terminology, including NAFLD/NASH and MAFLD, was retained when referring to the original studies, whereas MASLD/MASH was used for the current disease framework. EV terminology and the reported isolation/characterization methods were interpreted with reference to Minimal Information for Studies of Extracellular Vesicles (MISEV) principles and the information provided in the original studies. Additional relevant studies were identified through manual screening of reference lists and cross-referencing of related publications.

For evidence synthesis, studies were categorized according to their experimental setting and directness to MASLD as human clinical evidence, animal evidence from NAFLD/MASLD/MASH models, in vitro evidence from relevant cell-based models, or indirect non-MASLD evidence derived from liver fibrosis, other metabolic diseases, or broader EV-engineering studies. Studies involving more than one evidence type were classified accordingly, and indirect evidence was used only to illustrate potentially transferable biological or engineering principles rather than direct diagnostic or therapeutic efficacy in MASLD. Because this was a narrative rather than a systematic review, literature selection was guided by relevance to the review objectives and representation of the major pathogenic, diagnostic, therapeutic, and EV-engineering themes. Priority was given to original research, particularly human MASLD studies and MASLD-specific preclinical studies with clearly defined experimental models and EV characterization, whereas non-MASLD studies were included selectively as mechanistic or engineering proof of concept. Current clinical guidelines and consensus statements were used to contextualize disease terminology, diagnostic pathways, and established clinical management. Review articles were used mainly for background context and identification of additional primary studies, and preprints were not used as primary evidence for major diagnostic or therapeutic claims. The English-language restriction and narrative study-selection process may have introduced language and selective-citation bias; therefore, this review should not be interpreted as an exhaustive systematic evidence base.

TERMINOLOGY

In this review, the term “EVs” is used as a general term for EV populations, whereas sEVs refer to vesicles within the small-EV size range when supported by the original study. The term “exosomes” is retained when used by the cited studies to describe vesicles of presumed endosomal origin. Accordingly, EV-related terminology was not retrospectively converted when the original studies did not provide sufficient evidence for a specific vesicle subtype.

SEVS: FORMATION, STRUCTURE AND PATHOLOGICAL ROLE IN MASLD

The formation and structure of sEVs

sEVs, typically approximately 100 nm in diameter[15], are secreted by virtually all mammalian cell types, including human umbilical vein endothelial cells, mesenchymal stem cells (MSCs)[16], T cells, B cells, macrophages, dendritic cells and natural killer cells[17-20]. A subset of sEVs can originate through the endosomal pathway, involving plasma membrane endocytosis, endosomal maturation, and the subsequent formation of multivesicular bodies (MVBs)[15]. As depicted in Figure 1, plasma membrane endocytosis initially leads to the formation of early endosomes, which subsequently mature into late endosomes and MVBs[21-23]. During endosomal maturation, inward budding of the endosomal limiting membrane generates intraluminal vesicles (ILVs) within MVBs. MVBs may then follow two major fates: Fusion with lysosomes results in degradation of their contents, whereas fusion with the plasma membrane releases ILVs into the extracellular space as sEVs of endosomal origin. In contrast, other EV populations, including microvesicles and apoptotic bodies, arise through distinct biogenesis pathways[15,24,25]. Alternatively, MVBs may fuse with lysosomes, resulting in their degradation. Structurally, sEVs are spherical lipid-bilayer vesicles that carry a wide array of bioactive molecules, including proteins, lipids, nucleic acids, cytokines, transcription factor receptors, and other signaling mediators[26,27]. Importantly, even when secreted from the same cell origin, sEVs display marked heterogeneity in both size and cargo composition[28].

Figure 1 Conceptual schematic of small extracellular vesicles biogenesis and structure. Endocytosis leads to the formation of early endosomes, followed by endosomal maturation and the generation of multivesicular bodies (MVBs). Intraluminal vesicles are formed by inward budding of the endosomal limiting membrane within MVBs. MVBs may either fuse with lysosomes for degradation or fuse with the plasma membrane, resulting in the extracellular release of intraluminal vesicles as small extracellular vesicles. Small extracellular vesicles contain diverse cargos, including nucleic acids, proteins, lipids, and enzymes, and may display extracellular vesicles (EV)-associated tetraspanins such as CD9, CD63, and CD81. These proteins are commonly used for EV characterization but are not universal or definitive markers of a specific EV subtype; multiparametric characterization is recommended. ILVs: Intraluminal vesicles; MVB: Multivesicular body; sEVs: Small extracellular vesicles; HSP: Heat shock protein.

Although sEVs contain diverse biomolecules, their cargo composition is not determined by random incorporation but is regulated by selective sorting mechanisms during vesicle biogenesis[15]. The endosomal sorting complex required for transport (ESCRT)-dependent pathway represents one of the major mechanisms involved in cargo selection and ILV formation. ESCRT components, including TSG101 and ALIX, together with associated adaptor proteins, facilitate the recruitment and clustering of specific cargos at endosomal membranes, followed by membrane budding and ILV formation within MVBs. In addition, ESCRT-independent mechanisms involving tetraspanin-enriched microdomains (such as CD9, CD63, and CD81), lipid domains enriched in ceramide or cholesterol, and RNA-binding proteins further contribute to selective cargo enrichment.

For RNA cargos, RNA-binding proteins act as molecular selectors by recognizing specific RNA motifs or structural features. For example, hnRNPA2B1 can recognize the GGAG motif within specific miRNAs and facilitate their enrichment into sEVs[29]. During MASLD progression, metabolic stress, lipotoxicity, and hepatocyte injury may alter these sorting processes, resulting in disease-associated changes in sEV cargo profiles. Such alterations may contribute to the enrichment of disease-associated miRNAs, including miR-122 and miR-192-5p, in circulating liver-derived EVs.

In terms of protein composition, sEVs contain proteins derived from their cells of origin as well as EV-associated proteins commonly detected across sEV preparations. For instance, SPHK2 (sphingosine kinase 2), which is present in sEVs from primary hepatocytes, is crucial for liver regeneration[30]. Likewise, tetraspanins (e.g., CD63, CD9, CD81, and CD82) are frequently detected in sEV preparations and are commonly used as EV-associated proteins for sEV characterization[31]. Some membrane-associated proteins, including integrins, may also contribute to EV targeting, adhesion, and uptake[28,32]. Apart from proteins, nucleic acids are also the normal content of sEVs, for instance mRNA and microRNA (miRNA), which were first reported in 2007 by Valadi et al[33]. Furthermore, sEVs have been reported to contain mitochondrial DNA, double-stranded DNA, single-stranded DNA, small nuclear RNA, small cytoplasmic RNA, and other non-coding RNAs[34,35]. It is noteworthy that the biological activity of sEVs is determined not only by cargos such as proteins and nucleic acids, but also by their lipid components, including sphingomyelin, cholesterol, glycosphingolipids, and phosphatidylserine[36]. Lipids also play important roles in sEV formation and secretion into the extracellular environment[37]. Thus, the structural complexity and diverse molecular composition of sEVs enable them to participate in a broad range of biological processes, including intercellular communication, immune regulation, cell proliferation, apoptosis, and tissue homeostasis[38].

The pathological role of sEVs in MASLD

The broad biological functions of sEVs are highly relevant to the pathogenesis of MASLD. Hepatocyte injury, a hallmark event during disease progression, is accompanied by increased release of sEVs that carry stress-related signals and contribute to communication between injured hepatocytes and surrounding cells[39]. Under lipotoxic conditions, excessive lipid accumulation alters both the quantity and cargo composition of hepatocyte-derived sEVs, thereby promoting metabolic disturbances and pathological signaling within the hepatic microenvironment[35]. Beyond their roles as mediators of intercellular communication, sEVs facilitate signal exchange among hepatocytes, immune cells, hepatic stellate cells and extrahepatic metabolic organs. For instance, sEVs actively participate in hepatic inflammatory responses by mediating crosstalk among hepatocytes, Kupffer cells and infiltrating immune cells, contributing to the progression from simple steatosis to MASH[39,40]. As disease advances, sEV-associated signaling also contributes to hepatic stellate cell activation and fibrogenic remodeling, which are central events in liver fibrosis development[41,42]. In addition, emerging evidence suggests that sEVs participate in metabolic crosstalk between the liver and extrahepatic organs, including adipose tissue, skeletal muscle and the gut, thereby influencing systemic metabolic homeostasis and disease progression. Through these interactions, sEVs coordinate intrahepatic and systemic responses that collectively contribute to disease progression[43]. The major pathological processes associated with sEV-mediated signaling in MASLD are illustrated in Figure 2. Therefore, although sEVs exert diverse biological functions across multiple physiological and pathological settings, their involvement in hepatocyte injury, lipotoxicity, inflammation, fibrosis and inter-organ metabolic communication highlights their central role in MASLD pathogenesis.

Figure 2 Conceptual schematic of pathological roles of small extracellular vesicles in metabolic dysfunction-associated steatotic liver disease progression. Small extracellular vesicle-mediated signaling contributes to major pathological processes in metabolic dysfunction-associated steatotic liver disease, including lipotoxicity and metabolic dysregulation, immune activation and inflammation, hepatocyte injury and apoptosis, fibrogenic remodeling, and inter-organ metabolic crosstalk. These processes are associated with abnormal lipid accumulation, activation of hepatic immune cells and hepatic stellate cells, and metabolic communication involving the gut, adipose tissue, and skeletal muscle, collectively contributing to metabolic dysfunction-associated steatohepatitis, liver fibrosis, metabolic remodeling, and disease progression. The depicted mechanisms are primarily supported by preclinical and experimental evidence. MASLD: Metabolic dysfunction-associated steatotic liver disease; MASH: Metabolic dysfunction-associated steatohepatitis; sEV: Small extracellular vesicle.

Although sEVs participate in a wide range of physiological and pathological processes across multiple organs and disease settings, many of these functions converge on mechanisms that are highly relevant to MASLD pathogenesis. In particular, the ability of sEVs to regulate lipid metabolism, inflammatory signaling, immune-cell activation and fibrogenic responses provide a mechanistic basis for their involvement in disease initiation and progression. Therefore, understanding these fundamental biological properties is essential for interpreting the emerging roles of sEVs as biomarkers and therapeutic targets in MASLD[44,45].

SEVS AS CANDIDATE BIOMARKERS IN MASLD

sEVs have emerged as candidate liquid-biopsy analytes[46]. These nanoscale, lipid-bilayer-enclosed particles are secreted by diverse cell types, including hepatocytes, and carry proteins, lipids, and nucleic acids that may reflect disease-associated molecular changes[47]. Recent studies have identified altered sEV cargo profiles in patients with MASLD, including specific protein signatures and miRNA panels (e.g., miR-574-3p, miR-542-3p, and miR-200a-3p)[48]. These findings support further investigation of sEVs as minimally invasive biomarker candidates for MASLD, as summarized in Table 1.

Table 1 Candidate small extracellular vesicle biomarkers investigated in metabolic dysfunction-associated steatotic liver disease.

Cargo	Biomarker application	Study population/evidence	EV isolation/characterization	Key finding/diagnostic performance	Ref.
Nucleic acids
miR-122, miR-192, miR-128-3p	MASLD discrimination and progression	Human: Controls (n = 14), non-MASH MASLD (n = 8), biopsy-proven MASH (n = 6)	Plasma EVs; size-exclusion chromatography and ASGR1 immunocapture; TEM, NTA and EV protein markers	Increased with disease severity; study-cohort AUCs of 0.830, 0.895 and 0.803, respectively	[56]
miR-122-5p, miR-27a, miR-335-5p	Pediatric MASLD biomarker candidates	Human: MASLD (n = 5) vs simple obesity (n = 5)	Serum exosomes; TEM, NTA and western blotting	Upregulated; identified by high-throughput sequencing in an exploratory cohort	[63]
Proteins/EV features
PLIN3/CD63 ratio	Risk stratification, particularly lean MASLD	Human: Controls (n = 57), obese MASLD (n = 83), lean MASLD (n = 87)	Serum EVs; iodixanol density-gradient ultracentrifugation; NTA and western blotting	Decreased in lean MASLD; external validation is still required	[60]
CD63+ and ASGR1+ hepatocyte-derived EVs	MASH progression and liver-specific EV detection	Human: Patients with steatosis and MASH	Circulating EVs; NTA, TEM and flow cytometry; CD63 capture with ASGR1 detection	Increased in MASH and associated with disease severity	[61]
FZD-7	MASLD severity and potential treatment-response marker	Human: Healthy controls (n = 20) and MASLD (n = 95)	Plasma exosomes; differential ultracentrifugation; DLS, TEM, zeta potential and exosomal protein markers	Higher in MASLD and reduced after lifestyle intervention	[67]
Lipids
C16:0 ceramide and S1P	MASH-associated EV lipid candidates	Human: Obese controls (n = 11), simple steatosis (n = 16), early MASH (n = 16); parallel mouse experiments	Plasma EVs; sphingolipid profiling by tandem mass spectrometry	Enriched in MASH; supported by parallel experimental models	[71]
FFA (18:0), LPC (22:6/0:0), FFA (18:1), PI (16:0/18:1)	MASH discrimination (exploratory)	Human: Non-MASH MASLD (n = 43) and MASH (n = 40)	Urinary EVs; EXODUS isolation and LC-MS/MS lipidomics	Four-lipid panel distinguished MASH with a study-cohort AUC of 0.923 (exploratory)	[72]

Evidence categories are reported according to the study population and experimental setting, including human, animal, and combined human/experimental evidence where applicable. Reported area under the receiver operating characteristic curves represent performance estimates from the respective study cohorts unless otherwise stated and should not be interpreted as independently externally validated performance. Confidence intervals and comparisons with established non-invasive tests are reported only when available in the original studies. EV: Extracellular vesicle; MASLD: Metabolic dysfunction-associated steatotic liver disease; MASH: Metabolic dysfunction-associated steatohepatitis; ASGR1: Asialoglycoprotein receptor 1; TEM: Transmission electron microscopy; NTA: Nanoparticle tracking analysis; AUC: Area under the receiver operating characteristic curve; PLIN3: Perilipin 3; FZD-7: Frizzled class receptor 7; DLS: Dynamic light scattering; S1P: Sphingosine 1-phosphate; FFA: Free fatty acid; LPC: Lysophosphatidylcholine; PI: Phosphatidylinositol; EXODUS: Extracellular vesicle isolation by dodecyl sulfate-based enrichment; LC-MS/MS: Liquid chromatography-tandem mass spectrometry.

Nucleic acid carried by sEVs

The liver comprises hepatocytes and several non-parenchymal cell populations, including hepatic stellate cells, liver sinusoidal endothelial cells, Kupffer cells, and cholangiocytes, which collectively contribute to hepatic homeostasis and the regulation of hepatocyte function[49]. Studies have shown that circulating miRNAs, including miR-122, and miR-192, are upregulated in NAFLD/NASH and associated with disease severity[50-52]. Notably, circulating and hepatic expression patterns may differ, as hepatic miR-122 has been reported to decrease in NASH, whereas hepatic miR-34a-5p may increase[50,53]. In the subsequent analysis, Chen et al[54] found that adipocyte-derived exosomal miR-122 was associated with NAFLD progression. EV-related miR-122 and miR-192-5p have subsequently been proposed as candidate biomarkers of MASLD progression[55,56]. Serum miR-192-5p levels have also been shown to be positively correlated with hepatic inflammatory activity score and disease progression in MASH patients and rats fed with a high-fat and high-cholesterol diet, supporting its potential as a noninvasive biomarker for MASH[55,57]. The pathogenic signaling roles of adipocyte-derived miR-122 and hepatocyte-derived sEV miR-192-5p are discussed separately in the following sections. Besides, Kim et al[58] performed small RNA sequencing in serum samples from 12 patients with NAFL and 12 patients with MASH and identified increased levels of miR-21-5p, miR-151a-3p, miR-192-5p, and miR-4449 in MASH. Although these miRNAs showed discriminatory performance in this small exploratory cohort, larger independent studies are required to establish their diagnostic utility. Ezaz et al[59] also observed that miR-200a was associated with liver fibrosis in 132 patients with biopsy-confirmed MASLD, suggesting its potential as a candidate biomarker for liver fibrosis in MASLD. The research by Chen et al[60] also reported altered circulating EV profiles in patients with MASLD. Notably, lean patients exhibited a distinct decrease in the EV PLIN3/CD63 ratio, suggesting that this ratio may represent a candidate marker for disease risk stratification, although independent external validation is still required.

More recently, Masrour et al[61] reported increased circulating EV/particle (EVP) populations in patients with steatohepatitis, with CD63+/ASGR1+ EVPs being particularly increased. EVP abundance was associated with the severity of hepatic inflammation and steatosis. Functionally, exposure of macrophages to CD63+ EVPs from patients with MASLD induced transforming growth factor-β (TGF-β) and tissue inhibitor of metalloproteinase-1 secretion, linking disease-associated EVPs with potentially relevant fibrogenic signaling.

Likewise, Becker et al[62] reported that elevated serum miR-21 levels were significantly associated with MASH, supporting its potential as a candidate biomarker. In a pediatric cohort, Zhang and Pan[63] performed high-throughput sequencing of serum sEV-derived miRNAs and identified increased levels of miR-122-5p, miR-27a, and miR-335-5p in children with MASLD compared with those with simple obesity. Given the small exploratory cohort, these miRNAs should currently be regarded as candidate molecular markers that require validation in larger independent populations. In addition to miRNAs, mitochondrial DNA associated with hepatocyte-derived circulating EVs has also been reported to increase in both experimental models and patients with MASH[64], suggesting another potential disease-associated EV cargo. Collectively, these studies support the continued investigation of liver-derived EV nucleic acids as candidate biomarkers for MASLD; however, their diagnostic performance and clinical utility remain to be established in larger and independently validated cohorts.

Proteins carried by sEVs

In addition to nucleic acids, EV-associated proteins have also been investigated as candidate biomarkers in MASLD. Zhang et al[65] reported higher levels of hepatocyte-derived EV-associated glucose transporter 1 in patients with NASH than in those with NAFL. In biopsy-based fibrosis stratification, hepatogenic exosomal glucose transporter 1 was also increased in advanced NASH (F2-F4) compared with early NASH (F0-F1), with an area under the receiver operating characteristic of 0.85 [95% confidence interval (CI): 0.77-0.93] for fibrosis assessment in the study cohort. Vanin-1 has also been identified in hepatocyte-derived EVs and implicated in angiogenesis and liver injury during MASH[66], although its diagnostic performance requires further clinical validation. Scavo et al[67] observed higher exosomal frizzled-7 levels in patients with MASLD than in healthy controls and reported a reduction following lifestyle intervention, suggesting that EV-associated frizzled-7 may represent a candidate marker associated with disease severity and treatment response. In contrast, Yao et al[68] identified hepatocyte-derived EV-associated retinol-binding protein 4 (RBP4) as a disease-relevant pathogenic EV signal in high-fat diet (HFD)-fed mice. Because this study primarily provides mechanistic rather than diagnostic evidence, the therapeutic-target relevance of EV-associated RBP4 is discussed separately in the following sections.

In a murine MASLD/MASH model induced by Western diet with or without carbon tetrachloride, Wang et al[69] profiled serum-derived EV proteins at 14 weeks and 28 weeks and identified stage-associated proteomic signatures. Ctsb and Psap were enriched in early MASLD, whereas F13a1 and Pigr were associated with early MASH. Psma2, Psmb3, and Psmb5 were enriched in more advanced MASH stages. Because these findings were derived from a single murine experimental system, the identified EV protein signatures should be considered preclinical candidates that require validation in human MASLD cohorts. These findings indicate stage-associated changes in EV protein composition across the MASLD spectrum, although their reproducibility and diagnostic utility require validation in independent clinical cohorts.

Beyond individual protein markers, high-throughput proteomic profiling has enabled the identification of stage-associated EV protein panels. Using plasma EVs from biopsy-proven MASLD patients, Li et al[70] identified several proteins associated with different disease stages. Kelch-like protein 41 levels were negatively associated with hepatic steatosis severity, GP1BA enrichment was associated with inflammatory activity and MASH, and complement component 7 was increased in advanced fibrosis. In the study cohort, complement component 7 achieved an area under the curve (AUC) of 0.80 (95%CI: 0.69-0.91) for distinguishing < F2 from ≥ F2 fibrosis and an AUC of 0.83 (95%CI: 0.72-0.93) for distinguishing < F3 from ≥ F3 fibrosis. Its performance exceeded that of the non-invasive fibrosis scores evaluated in that study. These findings support further evaluation of multi-protein EV panels as candidate tools for fibrosis stratification; however, independent external validation and prospective comparison with established NITs are required before any incremental clinical value can be established.

Lipids carried by sEVs

EV-associated lipid species have also been investigated as potential biomarkers in MASLD. Kakazu et al[71] reported enrichment of C16:0 ceramide in EVs released from palmitate-stimulated hepatocytes and detected related alterations in experimental models and patients with MASH. EV-associated sphingosine 1-phosphate, a ceramide metabolite involved in macrophage chemotaxis and hepatic inflammation, was also identified in this setting. These findings suggest that EV-associated ceramide and sphingosine 1-phosphate may reflect disease-related lipid remodeling, although their diagnostic performance in independent clinical cohorts remains to be determined. Nevertheless, Zhu et al[72] used urinary EV lipidomics to identify a four-lipid panel comprising free fatty acid (18:0), lysophosphatidylcholine (22:6/0:0), free fatty acid (18:1), and phosphatidylinositol (16:0/18:1). In the reported cohort of patients with non-MASH MASLD and MASH, this panel distinguished MASH with an AUC of 0.923. Although this exploratory performance is encouraging, independent prospective validation and direct comparison with established NITs are required before the panel can be considered for clinical diagnostic use. Likewise, Garcia-Martinez et al[73] further identified EVs as carriers of saturated fatty acids involved in lipotoxic inflammatory signaling. These observations provide additional evidence that EV lipid composition reflects disease-related metabolic changes, but they do not by themselves establish diagnostic performance in MASLD.

Overall, most sEV biomarker candidates remain at an exploratory or early-validation stage, and large prospective cohorts with standardized EV isolation and independent external validation are still required before routine clinical implementation. Reported AUCs should therefore be interpreted as study-specific performance estimates rather than evidence of established clinical diagnostic utility.

Potential incremental value of sEV biomarkers beyond established NITs

Established NITs, including FIB-4 and imaging-based approaches, are widely used for risk stratification in MASLD, particularly for assessing liver fibrosis[5]. However, these approaches provide limited information on dynamic molecular changes associated with hepatocyte injury, inflammation, and disease activity. In contrast, sEV cargos may provide complementary molecular information reflecting the pathophysiological state and cellular origin of circulating vesicles. Therefore, the potential clinical value of sEV biomarkers may lie in complementing, rather than replacing, established NITs, particularly for molecular disease characterization and treatment-response monitoring. Exploratory human studies have identified EV-associated protein and lipid signatures related to MASLD severity, fibrosis stage, and MASH discrimination[70,72]. Nevertheless, no sEV-based biomarker has yet demonstrated sufficient incremental clinical utility over established NITs in independently validated prospective cohorts to support routine clinical use.

Accordingly, the clinical utility of sEV biomarkers should be evaluated according to specific intended uses. Current evidence is insufficient to support population-level screening. Plausible near-term evaluation settings may include patients with indeterminate results from established NITs, patients with suspected at-risk MASH, treatment-response monitoring, and clinical-trial enrichment. Early evidence also suggests that circulating EV profiles may change with disease resolution and response to weight-loss surgery[46]. Plasma and serum are currently the most practical sample types, whereas urinary EV biomarkers remain exploratory[72]. Depending on the intended use, selective enrichment of liver-derived EV populations may improve biological specificity and biomarker performance[56]. Future prospective studies should directly compare or combine sEV biomarkers with established clinical comparators, including FIB-4, elastography, magnetic resonance imaging-based assessments, and currently available serum biomarkers used in MASLD assessment pathways[5], and determine whether sEV measurements provide reproducible incremental information that improves patient classification or clinical decision-making. Major implementation barriers include pre-analytical variability, lack of standardized EV enrichment and analytical workflows, limited external validation, assay reproducibility, and the cost and technical requirements of advanced EV analysis.

Collectively, unlike free circulating molecules, sEV cargos are protected by a lipid bilayer membrane, which enhances their stability against enzymatic degradation. More importantly, sEVs retain molecular signatures of their cells of origin. Hepatocyte-derived, macrophage-derived, and hepatic stellate cell-derived sEVs therefore provide structured and biologically contextualized information regarding disease activity. This unique feature enables sEVs to function not only as biomarkers of liver injury but also as indicators of ongoing pathogenic mechanisms within MASLD, thus holding the potential to function as a theoretical agent or target for MASLD treatment.

THE THERAPEUTIC VALUE OF SEVS IN MASLD

Emerging studies have explored the therapeutic potential of sEVs in MASLD, and selected sEV preparations have shown favorable stability, biocompatibility, and relatively low immunogenicity in preclinical settings[74]. However, current evidence for sEV-based therapy in MASLD remains predominantly preclinical, and no sEV-based therapeutic strategy has yet been established for routine clinical management. In the following sections, studies conducted directly in NAFLD/MASLD/MASH models are distinguished from pathogenic sEVs proposed as therapeutic targets and from proof-of-concept studies performed in liver fibrosis, other metabolic disorders, or broader EV-engineering settings. These indirect studies are included to illustrate potentially transferable therapeutic or engineering principles and should not be interpreted as direct evidence of therapeutic efficacy in MASLD.

The therapeutic potential of native sEVs may originate from their intrinsic role in physiological intercellular communication. Under normal conditions, sEVs participate in the exchange of proteins, lipids, and nucleic acids among different cell types, thereby maintaining tissue homeostasis. During MASLD progression, these communication networks become dysregulated, contributing to hepatocyte injury, inflammation, and fibrosis. Administration of therapeutic sEVs may therefore partially restore disrupted signaling pathways and re-establish metabolic and immunological balance. In this regard, EVs can act both as carriers of molecular cargo and as biological mediators whose targeting, uptake, and intracellular fate influence responses in recipient cells[75]. Currently, sEV-based therapies for MASLD are pursued via two distinct routes: One relying on direct injection of unmodified sEVs into the circulation, and the other leveraging sEVs as nanocarriers for exogenous drugs, with engineered variants further optimized to enhance target specificity and delivery efficiency[76].

Native therapeutic sEVs in MASLD-related preclinical models

In addition to their diagnostic value, sEVs derived from selected therapeutic cell sources, particularly MSCs and immune cells, have shown protective effects in preclinical NAFLD/MASLD/MASH models. The evidence discussed in this subsection therefore focuses on native sEV preparations evaluated for therapeutic effects in MASLD-related experimental settings. Researchers have found that liver cells in MASLD patients were severely damaged by toxic lipid accumulation, resulting in hepatocyte damage or death, further leading to inflammation, liver fibrosis, cirrhosis and hepatocarcinoma[77]. Therapeutic sEVs derived from specific cellular sources can partially restore damaged hepatic functions by reducing lipid accumulation, regulating metabolic homeostasis, alleviating oxidative stress and suppressing inflammatory responses[78].

The therapeutic potential of MSC-derived EVs in MASLD has been increasingly recognized[79], alongside broader MSC-based interventions[80]. A recent comprehensive review by Zou and Brigstock[79] consolidated the current evidence, highlighting three primary mechanisms: Regulation of lipid metabolism (via SREBP-1c and PPARα pathways), suppression of hepatic inflammation (through macrophage polarization modulation), and inhibition of fibrogenesis (by targeting hepatic stellate cell activation). Previous reviews have separately addressed EV-associated miRNAs across the MASLD spectrum[81], EV pathobiology and therapeutic potential in NAFLD/alcoholic liver disease[78], and the therapeutic application of MSC-derived EVs in MASLD[79]. Building on these perspectives, the present review integrates pathogenic sEV-mediated signaling, candidate liquid-biopsy biomarkers, native therapeutic sEVs, engineered MASLD-specific interventions, and transferable EV-engineering strategies within an evidence-classified framework. Particular emphasis is placed on evidence directness, incremental value beyond established NITs, and the methodological and translational barriers that currently limit clinical implementation.

Mitochondrial dysfunction is known to drive the accumulation of hepatotoxic lipids, and in this context, a recent study reported that MSC-derived exosomal miR-24-3p attenuates lipid deposition in hepatocytes via targeting Keap-1 (Kelch-like ECH-associated protein 1)[82]. Moreover, this miR-24-3p was shown to inhibit palmitate-induced fat accumulation, reactive oxygen species (ROS) production, and inflammatory responses in vitro. In vivo, administration of miR-24-3p-containing MSC-derived sEVs significantly ameliorated metabolic dysfunction, hepatic injury, and steatosis in a mouse model of NAFLD. Conversely, when the miR-24-3p was knocked down or antagonized in controlled experiments, the beneficial effects were abolished, confirming its specific role. Collectively, these findings support further preclinical evaluation of MSC-derived miR-24-3p as a candidate therapeutic mechanism in MASLD[82].

Other studies have shown that myeloid-specific interleukin-6 (IL-6) signaling can induce macrophages and neutrophils to release miR-223, which then transfers to liver cells to inhibit liver fibrosis[83]. In a liver fibrosis model rather than a MASLD-specific setting, human umbilical cord-MSC-derived exosomal miR-499a-5p was reported to modulate the ETS1/GPX4-associated ferroptosis pathway in hepatic stellate cells and attenuate fibrotic progression[84]. This finding provides indirect mechanistic evidence relevant to fibrosis but should not be interpreted as direct evidence of sEV therapy in MASLD.

In addition, Cheng et al[85] reported that miR-627-5p was highly enriched in MSC-derived sEVs. This miRNA upregulated the expression of gluconeogenic and lipogenic genes, including G6Pc, PEPCK, FAS, and SREBP-1c, while downregulating PPARα expression, thereby improving glycolipid metabolism. Mechanistically, miR-627-5p was found to alleviate liver injury and attenuate the progression of MASLD by directly targeting and suppressing fat mass and obesity-associated protein expression. Meanwhile, sEVs derived from adipose-derived stem cells (ADSCs) have also been shown to play a critical role in regulating obesity-related inflammation and metabolic disorders. Beyond MASLD-specific models, ADSC-derived sEVs have also shown metabolic regulatory effects in obesity-related experimental settings. In obese mice, administration of ADSC-derived sEVs promoted M2 macrophage-associated metabolic adaptations and improved systemic metabolic homeostasis[86]. This finding provides indirect evidence for the broader metabolic effects of ADSC-derived sEVs rather than direct therapeutic evidence in MASLD.

Except for lipid metabolism, the elimination of inflammatory reactions and ROS also have been shown to relate with sEVs. For instance, Nrf2/NQO-1 pathway has also been shown to have a regulatory effect in a pathway of Antioxidant factors. It has been shown that human umbilical cord-MSC sEVs can reduce inflammatory response and ROS by regulating the expression of proteins associated with the Nrf2/NQO-1 pathway[87]. Moreover, bone marrow MSC-derived sEVs have also shown therapeutic potential in experimental MASLD-related models. For instance, El-Derany and AbdelHamid[88] evaluated bone marrow-MSCs and bone marrow-MSC-derived exosomes in an experimental MASH animal model and observed upregulation of fatty acid oxidation-related factors, including PPARα and CPT1, together with reduced steatosis and hepatocyte ballooning. These effects were associated with increased miR-96-5p expression and downregulation of its downstream target caspase-2, suggesting that the miR-96-5p/caspase-2 axis may represent a potential therapeutic mechanism in experimental MASH. In a fibrosis-focused experimental setting rather than a MASLD-specific model, miR-6766-3p-enriched sEVs derived from 3D-cultured human embryonic stem cells were reported to suppress hepatic stellate cell activation through the TGF-βRII/SMAD pathway[89]. This study therefore represents indirect antifibrotic evidence with potential relevance to advanced MASLD.

Apart from sEVs derived from MSCs, sEVs originated from immunocytes also display the therapeutic value for the treatment of MASLD. It has been demonstrated that exosomal miRNA-223 derived from macrophages has an antifibrotic effect, which has the potentially important therapeutic target to relieve liver fibrosis via myeloid-specific IL-6 signaling[83]. Likewise, Gao et al[90] recently found that Kupffer cells-derived miR-690 can directly inhibit fibrogenesis in hepatic stellate cells, inflammation in recruited hepatic macrophages, and de novo lipogenesis in hepatocytes via targeting NAD kinase (NADK) mRNA and suppression NADK, thus supporting its preclinical therapeutic potential for MASLD.

Collectively, these findings suggest that source-specific therapeutic sEVs may represent promising candidates for MASLD intervention, whereas pathogenic sEVs may provide additional opportunities for therapeutic targeting. However, the therapeutic effects of sEVs should be interpreted with consideration of the heterogeneity of experimental models used to mimic MASLD progression, as different models represent distinct pathological stages and characteristics of the disease. In vitro lipotoxicity models, such as palmitate-stimulated hepatocytes, mainly reflect early hepatocellular lipid accumulation and lipotoxic stress. HFD-induced animal models better represent metabolic dysregulation associated with steatosis and mild inflammatory responses, whereas fibrosis-related models involving hepatic stellate cell activation reflect more advanced fibrogenic remodeling. Therefore, therapeutic outcomes obtained from different experimental systems should be considered according to their specific pathological contexts rather than generalized across the entire spectrum of MASLD progression.

Pathogenic sEVs as therapeutic targets in MASLD

In addition to their potential use as therapeutic agents, disease-promoting sEVs may themselves represent targets for intervention in MASLD. Unlike native therapeutic sEVs, which are administered for their beneficial biological effects, pathogenic sEVs contribute to disease progression by transferring disease-promoting molecular signals between donor and recipient cells. These pathogenic sEVs should therefore be distinguished from therapeutically administered sEVs: They represent potential targets for inhibiting EV release, cargo transfer, or recipient-cell signaling rather than therapeutic agents themselves.

Several representative pathogenic EV-associated cargos illustrate this concept. Adipocyte-derived exosomal miR-122 has been associated with NAFLD progression and, in experimental models, was reported to target Sirt1 and promote disease progression[54]. Liu et al[57] reported that hepatocyte-derived sEVs enriched in miR-192-5p activate proinflammatory macrophages through Rictor/Akt/FoxO1 signaling, thereby contributing to hepatic inflammation. Similarly, Yao et al[68] demonstrated that hepatocyte-derived EV-associated RBP4 promotes Kupffer-cell M1 polarization, hepatic lipid accumulation, and inflammation in experimental NAFLD. These findings indicate that pathogenic EV cargos can participate in MASLD progression through distinct metabolic and inflammatory signaling pathways.

Collectively, pathogenic sEVs represent a therapeutic concept distinct from native or engineered therapeutic sEV platforms. Potential intervention strategies may therefore involve reducing pathogenic EV release, limiting harmful cargo transfer, or interfering with recipient-cell signaling. However, current evidence supporting these approaches remains predominantly mechanistic and preclinical, and their feasibility, specificity, and therapeutic efficacy require further validation. Representative evidence is summarized separately in Table 2.

Table 2 Evidence-classified therapeutic roles and strategies of small extracellular vesicles relevant to metabolic dysfunction-associated steatotic liver disease.

Source	Cargo	Experimental model/evidence directness	Mechanism/biological effect	Ref.
Native therapeutic sEVs in MASLD-related models
hUC-MSCs	miR-24-3p	In vitro and animal evidence: NAFLD models	Targets Keap-1 and reduces lipid accumulation, oxidative stress, and inflammation	[82]
hUC-MSCs	miR-627-5p	In vitro and animal evidence: NAFLD models	Represses FTO expression and improves hepatic glucose and lipid metabolism	[85]
hUC-MSCs	Not specifically identified	In vitro and animal evidence: Experimental NASH models	Regulates the Nrf2/NQO-1 pathway and reduces inflammation and oxidative stress	[87]
BM-MSCs	miR-96-5p	Animal evidence: Experimental NASH rat model	Downregulates caspase-2, enhances fatty acid oxidation, and reduces steatosis	[88]
Myeloid cells	miR-223	In vitro and animal mechanistic evidence: NAFLD-associated fibrosis	Myeloid IL-6 signaling promotes miR-223-enriched sEV release and attenuates liver fibrosis	[83]
Kupffer cells	miR-690	Cell-based and animal evidence: NASH models	Targets NADK and suppresses fibrogenesis, inflammation, and de novo lipogenesis	[90]
Engineered MASLD-specific interventions
GalNAc-modified RBC-EVs	PJ34; GalNAc surface modification	Animal evidence: NAFLD mouse model	Enhances hepatocyte-targeted delivery and reduces hepatic triglyceride accumulation through PARP-1 inhibition	[96]
RBC-sEVs	siNUP85	In vitro and animal evidence: MASLD models	Silences NUP85 and attenuates steatosis, inflammation, and oxidative stress, partly through SLC27A1/PPAR-γ signaling	[103]
cRGD-modified ADSC-EVs	Angiotensin-(1-7); cRGD targeting	In vitro and animal evidence: NAFLD-associated fibrosis	Activates Mas receptor signaling, enhances Akt/Foxo1-associated autophagy and metabolic reprogramming, and reduces lipid accumulation and fibrosis	[113]
Pathogenic sEVs as therapeutic targets
Adipocytes	miR-122	Animal and in vitro evidence: NAFLD models	Targets Sirt1 and promotes disease progression	[54]
Hepatocytes	miR-192-5p	Human, animal, and cell-based evidence: NAFLD/MASH	Activates proinflammatory macrophages through Rictor/Akt/FoxO1 signaling	[57]
Hepatocytes	RBP4	Human association, cell-based and animal evidence: NAFLD	Promotes Kupffer-cell M1 polarization, hepatic lipid accumulation, and inflammation	[68]
Transferable proof-of-concept approaches
hESC-derived sEVs	miR-6766-3p	Indirect non-MASLD evidence: In vitro and animal liver fibrosis models	Targets TGF-βRII/SMAD signaling and suppresses hepatic stellate-cell activation and fibrotic progression	[89]
5HT1D/CD47-engineered exosomes	miR-29b; 5HT1D antibody and CD47 surface modification	Indirect non-MASLD evidence: In vitro and animal liver fibrosis models	Enhances targeting of activated hepatic stellate cells and suppresses fibrotic signaling through the TGF-β/SMAD pathway	[114]

Evidence categories are assigned according to the experimental setting and disease model reported in the original studies. Studies conducted in liver fibrosis or other non-metabolic dysfunction-associated steatotic liver disease settings are explicitly classified as indirect non-metabolic dysfunction-associated steatotic liver disease evidence. Table 2 is intended to classify therapeutic evidence according to experimental setting and disease-model relevance; administration route, dose, and extracellular vesicle isolation protocols vary across studies and are not used here as evidence-ranking criteria. sEVs: Small extracellular vesicles; MASLD: Metabolic dysfunction-associated steatotic liver disease; hUC-MSCs: Human umbilical cord mesenchymal stem cells; Keap-1: Kelch-like ECH-associated protein 1; NAFLD: Nonalcoholic fatty liver disease; FTO: Fat mass and obesity-associated protein; NASH: Non-alcoholic steatohepatitis; NQO-1: NAD(P)H quinone oxidoreductase 1; Nrf2: Nuclear factor erythroid 2-related factor 2; IL-6: Interleukin-6; GalNAc: N-acetylgalactosamine; NADK: NAD kinase; RBC-EVs: Red blood cell-derived extracellular vesicles; PARP-1: Poly(ADP-ribose) polymerase 1; cRGD: Cyclic arginine-glycine-aspartic acid; ADSC-EVs: Adipose-derived mesenchymal stem cell-derived extracellular vesicles; RBP4: Retinol-binding protein 4; hESCs: Human embryonic stem cells; BM-MSCs: Bone marrow mesenchymal stem cells; TGF‑βRII: Transforming growth factor-beta receptor II; SMAD: Mothers against decapentaplegic homolog; 5HT1D: 5-hydroxytryptamine receptor 1D; CD47: Cluster of differentiation 47; TGF-β: Transforming growth factor-β.

sEVs as drug delivery vehicles for the treatment of MASLD

Various physical and chemical approaches have been developed for loading therapeutic cargos into sEVs. Representative methods illustrated in Figure 3 include electroporation, sonication, passive incubation, extrusion, and freeze-thaw cycles. Other approaches, including saponin-assisted loading and pH-gradient-based loading, have also been reported for EV cargo incorporation[91-93].

Figure 3 Conceptual schematic of therapeutic roles and engineering strategies of small extracellular vesicles relevant to metabolic dysfunction-associated steatotic liver disease. The upper panels summarize metabolic dysfunction-associated steatotic liver disease (MASLD)-related evidence for native therapeutic small extracellular vesicles and pathogenic small extracellular vesicle-associated targets. The lower panels summarize general or transferable extracellular vesicle-engineering approaches, including cargo loading and surface engineering. Electroporation, sonication, freeze-thaw cycling, passive incubation, and extrusion represent general cargo-loading methods and should not be interpreted as established MASLD-specific therapeutic strategies unless validated directly in nonalcoholic fatty liver disease/MASLD/metabolic dysfunction-associated steatohepatitis models. Evidence categories and disease-model relevance for representative therapeutic studies are detailed in Table 2. sEV: Small extracellular vesicle; MASLD: Metabolic dysfunction-associated steatotic liver disease.

However, these loading approaches differ substantially in their loading efficiency, cargo compatibility, preservation of cargo integrity, and effects on the structural and biological properties of sEVs. Therefore, selecting an appropriate loading strategy requires comprehensive consideration of the characteristics of therapeutic cargos and the preservation of native sEV functions. The major advantages and limitations of commonly used sEV cargo loading approaches are summarized in Table 3[91,94,95]. Importantly, the examples discussed below do not represent equivalent levels of evidence. Studies evaluated directly in NAFLD/MASLD/MASH models are considered MASLD-related preclinical evidence, whereas studies derived from liver fibrosis, other metabolic disorders, or broader EV-delivery research are explicitly treated as indirect proof-of-concept evidence. The latter are included only to illustrate potentially transferable delivery or engineering principles and are not considered evidence of established therapeutic efficacy in MASLD.

Table 3 Comparison of commonly used small extracellular vesicles cargo loading and surface engineering strategies for therapeutic delivery.

Strategy	Principle	Suitable cargos/ modification	Effect on vesicles	Key advantages and limitations
Passive incubation	Passive diffusion or interaction across the vesicle membrane	Small molecules; hydrophobic drugs	Minimal disruption; cargo integrity generally preserved	Simple and mild, but loading efficiency is low to moderate and nucleic acid loading is limited
Electroporation	Electrical pulses generate transient membrane pores	miRNA; siRNA; mRNA	It may cause membrane disruption, aggregation, or RNA precipitation	Useful for nucleic acids with moderate to high loading, but reproducibility is variable and optimization is required
Sonication	Mechanical energy transiently increases membrane permeability	Drugs; proteins; nucleic acids	It may alter vesicle morphology and membrane components	Generally improves loading efficiency, but excessive treatment can damage cargo and compromise native vesicle properties
Extrusion	Mechanical deformation promotes cargo incorporation	Drugs; nucleic acids	It may alter vesicle size distribution and biological activity	It can yield relatively uniform vesicles, but membrane remodeling may affect native functions
Freeze-thaw cycles	Repeated membrane disruption and resealing	Proteins; small molecules	Possible structural alteration and membrane fusion	Easy to perform, but loading efficiency is usually low to moderate and cargo preservation can vary
Saponin-assisted loading	Mild detergent transiently permeabilizes the vesicle membrane	Small molecules; hydrophilic drugs	It may increase membrane permeability and damage membranes at excessive concentrations	It can improve loading efficiency, but saponin concentration and removal of residual detergent require careful control
Endogenous loading (donor-cell engineering)	Donor-cell modification promotes incorporation of selected cargos during vesicle biogenesis	miRNAs; proteins; therapeutic molecules	Usually preserves physiological vesicle structure and biological activity	Good biological compatibility and cargo enrichment, but manufacturing and quality control are more complex
Chemical surface conjugation	Targeting ligands are covalently attached to molecules on the vesicle membrane	Peptides; antibodies; glycans; other targeting ligands	It may alter membrane proteins or surface properties under harsh reaction conditions	It improves targeting specificity, but reaction chemistry and ligand density must be optimized to preserve vesicle integrity

Chemical surface conjugation represents a surface engineering strategy rather than internal cargo loading. miRNA: MicroRNA; siRNA: Small interfering RNA; mRNA: Messenger RNA.

Among them, incubation is the simplest way of drug delivery, which achieves drug diffusion through different drug concentration gradients inside and outside the membrane. This method takes a long time and is inefficient. In contrast, ultrasonic, extrusion and electroporation are much more efficient, but sEVs membranes will be severely damaged by these relatively violent methods, resulting in a decline in membrane integrity. In addition, repeated freeze-thaw cycles can also induce the formation of pores and membrane fusion. The use of mild detergents such as saponins can also form drug loading holes. Adding curcumin during the drug loading process can enhance its hydrophobic performance and drug loading efficiency, and improve the anti-inflammatory effect[94].

New research shows that N-acetylgalactosamine (GalNAc)-modified red blood cell-derived EVs can play a protective role against liver diseases. Huang et al[96] developed GalNAc-modified red blood cell-derived EVs (GalNAc-RBC-EVs) for hepatocyte-targeted delivery and evaluated GalNAc-RBC-EVs loaded with PJ34 in a mouse model of NAFLD. GalNAc-RBC-EVs/PJ34 reduced hepatic triglyceride levels (P < 0.0001), with the therapeutic effect associated with poly(ADP-ribose) polymerase 1 inhibition. No significant toxicity was detected in the biochemical and histological assessments reported in this preclinical study. These findings support further investigation of GalNAc-modified RBC-EVs as liver-targeted drug-delivery vehicles.

sEVs-based chemical drug transport: The pharmacokinetic behavior and in vivo fate of sEV-based drug delivery systems are important considerations for therapeutic application. Exosomes and exosome-mimetic platforms have been explored as carriers for biological therapeutics because of their capacity to encapsulate and transport bioactive cargos[97]. However, these carrier properties do not necessarily imply prolonged systemic persistence, and biodistribution and clearance remain important determinants of delivery efficiency. Among drugs, the therapeutic mechanism on MASLD is different, including eliminating ROS and inflammation and inhibiting nucleic acid synthesis, etc. In a liver fibrosis setting rather than a MASLD-specific model, Ji et al[98] developed an sEV-liposome hybrid delivery system loaded with the antifibrotic agent clodronate and nintedanib. This strategy enhanced Kupffer-cell inhibition by clodronate while facilitating nintedanib delivery to fibrogenic liver cells, thereby improving antifibrotic efficacy.

sEVs-based bioactive molecules transport: CRISPR-Cas9-based gene editing requires safe and effective in vivo delivery systems[99]. In the context of hepatic fibrosis rather than MASLD specifically, AML12 cell-derived sEVs have been used to deliver CRISPR/dCas9-VP64 for hepatic stellate cell reprogramming, demonstrating the feasibility of sEV-mediated gene delivery in fibrotic liver disease[100]. Similarly, in a familial hypercholesterolemia model rather than MASLD, Li et al[101] demonstrated that engineered sEVs loaded with Ldlr mRNA restored hepatic Ldlr expression in Ldlr-deficient mice and alleviated hypercholesterolemia. This study provides a transferable example of sEV-mediated gene delivery for metabolic disorders. In parallel, although not developed specifically for MASLD, anchor-peptide strategies have been used to capture, target, and load exosomes of diverse origins[102], providing a potentially transferable approach for future sEV-mediated delivery of bioactive proteins or peptides in MASLD.

As noted earlier, red blood cell-derived sEVs have shown promise for targeted delivery. In a recent study, red blood cell-derived sEVs loaded with siNUP85 were designed to specifically target hepatocytes and effectively knockdown the overexpression of NUP85 in a mouse model of MASLD. This therapeutic approach attenuated hepatic steatosis, inflammation, and oxidative stress, at least in part through the SLC27A1/PPAR-γ signaling pathway, supporting the preclinical potential of sEV-mediated nucleic acid delivery in MASLD[103].

EV-mediated gene delivery has been increasingly explored as a strategy for transporting therapeutic nucleic acids[104,105]. Beyond cargos already investigated in EV-based systems, other disease-relevant nucleic acids may provide candidates for future delivery studies. For example, miR-376b-3p has been shown to ameliorate hepatic steatosis by targeting fibroblast growth factor receptor 1 and regulating lipid oxidation[106]; however, that study did not investigate EV-associated miR-376b-3p or sEV-mediated delivery. Therefore, this finding provides mechanistic rationale for future cargo-selection studies rather than direct evidence for sEV-mediated therapy in MASLD. Representative native therapeutic sEVs, engineered MASLD-specific interventions, pathogenic sEV targets, and selected transferable proof-of-concept strategies are summarized in Table 2 according to their evidence category and disease-model relevance.

Engineered sEV strategies relevant to MASLD

Engineered sEVs are EVs modified to incorporate defined therapeutic cargos or alter surface properties for specific delivery applications[107]. These approaches provide a flexible platform for investigating cargo delivery and targeting; however, most engineered sEV strategies relevant to MASLD remain at the preclinical or proof-of-concept stage.

Compared with conventional nanoparticles, engineered sEVs provide a biologically derived platform with substantial flexibility for therapeutic design[108]. Favorable biocompatibility and relatively low immunogenicity have been reported for selected EV preparations; however, these properties cannot be generalized across all sEV products and may vary with vesicle source, manufacturing process, engineering strategy, dose, and route of administration[109]. Surface or cargo engineering may improve delivery specificity in selected systems, but targeting efficiency, biodistribution, stability, and safety require product-specific evaluation[95]. Importantly, engineered sEVs are biologically active delivery platforms that may retain intrinsic intercellular communication properties in addition to carrying exogenous therapeutic cargos. Collectively, these features support continued preclinical investigation of engineered sEVs as potential delivery platforms relevant to MASLD.

As shown in Figure 4, engineered sEVs can be generated either by modifying parental cells before vesicle release or by modifying isolated sEVs after purification[110]. Parental-cell engineering, including genetic modification or chemical metabolic labeling, can alter vesicle cargo or surface properties. For example, azide-containing precursors can be introduced to label surface glycoproteins, followed by click-chemistry-based functionalization[111]. After isolation and purification, sEV membranes can be transiently permeabilized to facilitate the entry of exogenous molecules into the vesicle lumen for cargo loading. Besides, covalent conjugation of functional molecules to proteins or lipids on the sEV membrane can be used to modify targeting properties, although reaction chemistry and ligand density should be optimized to preserve vesicle integrity and delivery performance[112]. However, the examples discussed below include both MASLD-specific applications and engineering strategies investigated in related liver diseases. Non-MASLD examples are presented as proof-of-concept approaches with potential relevance to MASLD rather than as established MASLD-specific therapies. In a recent preclinical study of targeted EV therapy for NAFLD-associated fibrosis, Niu et al[113] developed cyclic arginine-glycine-aspartic acid-modified adipose-derived MSC-derived EVs loaded with Angiotensin-(1-7). The system was evaluated in LX-2 cells and an HFD-induced NAFLD mouse model. Angiotensin-(1-7) delivery activated Mas receptor signaling and enhanced Akt/Foxo1-associated autophagy and fatty acid metabolic reprogramming, accompanied by reductions in hepatic lipid accumulation and fibrosis. These findings provide preclinical evidence for targeted EV-mediated peptide delivery in NAFLD-associated fibrotic disease.

Figure 4 Conceptual schematic of major engineering strategies for small extracellular vesicles. This conceptual schematic illustrates two major approaches for engineering small extracellular vesicles. A: Parental cells can be engineered through approaches such as genetic modification or plasmid transfection, enabling the resulting small extracellular vesicles to acquire desired therapeutic cargos or functions; B: After isolation, small extracellular vesicles can be further engineered through cargo loading and membrane surface modification. Therapeutic agents may be introduced by transiently altering membrane permeability, while targeting ligands can be conjugated to the vesicle membrane to enhance delivery specificity. sEV: Small extracellular vesicle.

As an example derived from hepatic fibrosis rather than a MASLD-specific model, Zhang et al[114] developed a nanosecond pulsed microfluidic system (T-nsPMs) for the production of exosomes co-modified with 5-hydroxytryptamine receptor 1D antibody and CD47 protein. These engineered exosomes effectively penetrated the extracellular matrix, evaded macrophage phagocytosis, and precisely delivered miR-29b to activated hepatic stellate cells, resulting in significant downregulation of fibrotic markers including α-smooth muscle actin, collagen type I alpha 1 chain, and tissue inhibitor of metalloproteinase-1 through suppression of the TGF-β/SMAD signaling pathway. Regarding MASLD, studies have shown that the excessive opening of mitochondrial permeability pores in hepatocytes leads to mitochondrial dysfunction, which plays a core role in the occurrence and development of MASLD.

Comparative analysis of therapeutic sEV strategies

Importantly, selection of therapeutic sEV strategies should be determined by therapeutic objectives, disease characteristics, and the complexity of pathological mechanisms. Although sEVs can function either as intrinsic therapeutic agents or as delivery vehicles for exogenous therapeutic cargos, these approaches represent distinct but complementary therapeutic paradigms. Native therapeutic sEVs, such as MSC-derived sEVs, exert biological effects primarily through their endogenous cargos, including miRNAs, proteins, and lipids, which can simultaneously regulate multiple pathological processes, such as inflammation, oxidative stress, and metabolic imbalance. Therefore, native sEV-based therapies may be particularly suitable for complex metabolic disorders such as MASLD, where multiple interconnected pathways contribute to disease progression. In addition, their physiological origin and endogenous biological properties may contribute to favorable biocompatibility and immunomodulatory effects. However, the therapeutic efficacy of native sEVs may be influenced by donor heterogeneity, cellular status, and production conditions, which can complicate potency standardization and quality control[109].

In contrast, sEVs used as drug delivery vehicles provide a more controllable strategy by enabling the incorporation of defined therapeutic cargos, including chemical drugs, nucleic acids, proteins, or peptides, to modulate specific disease-associated pathways[115]. This approach may be preferable when precise regulation of well-characterized molecular targets is required[116]. In MASLD, fibrosis-related signaling, lipid metabolism, and mitochondrial dysfunction represent potential disease-relevant targets, although the suitability of EV-mediated delivery for each requires direct validation. For example, engineered sEV delivery systems may be particularly relevant when pathway-specific intervention is required, whereas native therapeutic sEVs may be more suitable for restoring broad metabolic and immunological homeostasis. Nevertheless, engineered sEVs are not necessarily superior to native sEVs, as additional cargo loading, surface modification, and manufacturing procedures may influence vesicle integrity, biological activity, and long-term safety. Therefore, native therapeutic sEVs and engineered sEV delivery platforms should not be considered competing approaches; rather, they represent complementary strategies that should be selected according to disease context, therapeutic requirements, and the desired balance between biological regulation and functional controllability.

CRITICAL DISCUSSION AND METHODOLOGICAL CAVEATS

Across current sEV biomarker studies, cross-study consistency remains difficult to assess because study populations, disease definitions, biospecimens, EV isolation and enrichment methods, analytical platforms, and clinical endpoints differ substantially. Reported biomarker performance should therefore not be interpreted as directly comparable across studies or as evidence of a single reproducible MASLD-associated sEV signature. These methodological and analytical differences remain major barriers to cross-cohort reproducibility and biomarker standardization.

Although increasing evidence supports the therapeutic potential of sEV-associated miRNAs in MASLD, the causal relationships between specific miRNA cargos and therapeutic outcomes require further clarification[117]. Current investigations often rely on miRNA mimic or inhibitor approaches, bioinformatic target prediction, and luciferase reporter assays to identify miRNA-target interactions. However, these approaches mainly demonstrate regulatory associations between miRNAs and their predicted targets and may not fully establish whether the therapeutic effects of sEV-derived miRNAs are directly mediated through the proposed molecular pathways in vivo. More rigorous validation strategies, including rescue experiments by restoring target gene expression without the corresponding 3’ untranslated region, loss-of-function approaches, and in vivo genetic validation, are required to confirm the causal contribution of specific sEV cargos to MASLD improvement. Therefore, future studies should integrate mechanistic validation with functional evaluation to strengthen the evidence linking sEV-derived miRNAs with therapeutic effects.

In addition to establishing causal relationships between sEV cargos and therapeutic outcomes, the intracellular trafficking and release of delivered cargos represent another critical determinant of therapeutic efficacy. Although several studies have demonstrated that sEV-associated miRNAs regulate specific targets in MASLD models, direct evidence confirming that these cargos successfully escape endosomal or lysosomal degradation and reach their functional sites remains limited. Recent studies in the EV field have highlighted that intracellular trafficking and endosomal escape are essential determinants of functional cargo delivery, as internalized vesicles may undergo endosomal retention or lysosomal degradation rather than productive cytosolic release[118]. Advanced approaches, including LysoTracker-based colocalization analysis and endosomal escape reporter systems, have been developed to evaluate the intracellular fate and functional availability of delivered cargos[119]. Therefore, future MASLD-related sEV studies should incorporate intracellular trafficking analyses to determine whether therapeutic cargos, such as miR-24-3p targeting Keap-1 and miR-690 targeting NADK, achieve effective intracellular delivery rather than being degraded after cellular uptake.

Systemic therapeutic use of sEVs also requires product-specific safety assessment beyond demonstration of efficacy. Route-dependent biodistribution and uptake by macrophages in the liver and spleen may reduce target exposure and contribute to off-target effects. In addition to immunogenicity, future development should evaluate procoagulant potential, unintended bioactive cargos, tumor-related safety signals, donor-cell identity and safety, batch-to-batch variability, and long-term toxicity. Together with dose standardization, validated potency assays, scalable manufacturing, and regulatory requirements, these unresolved issues help explain why promising preclinical sEV strategies have not yet translated into clinical trials for MASLD.

FUTURE CHALLENGES AND PROSPECTS

From conventional imaging and liver biopsy to sEVs as diagnostic markers for diagnosing MASLD, scientists’ research has also encountered a series of difficulties and adjustments. Firstly, the standardization and nomenclature of sEVs remain the first barrier for experiments to move towards clinical translation, and it is necessary to more strictly adhere to the MISEV guidelines of the International Society for EVs. Unifying terminology, transparent reporting processes and setting up negative/positive controls are the prerequisites for achieving comparability and reproducibility in the future[120].

A second important challenge for the clinical translation of sEV-based therapies is dose standardization. Currently, many preclinical studies define sEV doses based on total protein content rather than particle number, which may limit the comparability of therapeutic outcomes among different studies. Since different isolation and purification strategies can generate substantial variations in particle-to-protein ratios, protein-based dosing alone may not accurately represent the actual number of administered vesicles. Therefore, future studies should incorporate particle-based quantification methods, such as nanoparticle tracking analysis or tunable resistive pulse sensing, and report dosing information together with relevant quality control parameters to improve reproducibility and facilitate clinical translation[121].

A third major challenge for clinical translation is the lack of standardized and scalable sEV isolation strategies. Currently, ultracentrifugation, size exclusion chromatography, and immunoaffinity-based approaches are among the most commonly used methods for EV isolation, but each method has distinct advantages and limitations. Ultracentrifugation enables relatively high recovery and has been widely adopted in preclinical studies; however, it requires specialized equipment, may induce vesicle aggregation, and often shows limited reproducibility between laboratories. Size exclusion chromatography provides relatively high purity and preserves vesicle integrity by efficiently removing soluble proteins, but its recovery efficiency and scalability remain challenging for large-volume clinical applications. Immunoaffinity-based isolation achieves high specificity by targeting EV surface markers, but it may selectively capture only specific EV subpopulations and is limited by marker heterogeneity and large-scale production requirements. Therefore, future studies should optimize isolation workflows according to specific diagnostic or therapeutic purposes by balancing yield, purity, reproducibility, and scalability[120,121].

A fourth critical challenge for plasma/serum EV biomarkers in MASLD is lipoprotein co-separation[122]. Due to the substantial alterations in circulating lipid profiles during MASLD progression, lipoprotein particles, including low-density lipoproteins, very-low-density lipoproteins, and high-density lipoproteins, may be co-isolated with EVs because of their overlapping physicochemical properties. Such contamination may introduce non-EV-derived miRNAs and proteins, thereby leading to inaccurate interpretation of EV cargo profiling and functional studies. Therefore, quantitative quality control strategies should be incorporated into future MASLD EV studies. Assessment of residual lipoprotein-associated markers, such as ApoB and ApoA-I, may serve as important indicators for evaluating purification efficiency. Moreover, delipidation procedures, including optimized density-based separation or other lipoprotein depletion strategies, should be considered as routine steps in standardized workflows to improve the reliability and clinical applicability of EV-based biomarker studies.

Pre-analytical variability and clinical confounding represent additional challenges for sEV biomarker studies. EV abundance and cargo profiles may be influenced by specimen type (plasma, serum, or urine), fasting status, sample collection and processing procedures, storage conditions, and repeated freeze-thaw cycles. In addition, common metabolic and clinical factors associated with MASLD, including obesity, diabetes, dyslipidemia, renal dysfunction, medication use, and concurrent liver diseases, may independently alter circulating or urinary EV profiles and therefore complicate attribution of EV-associated changes specifically to MASLD. Future clinical studies should prospectively record and, where possible, standardize these variables and incorporate appropriate matching or statistical adjustment to reduce confounding and improve cross-cohort comparability.

Furthermore, the specific targeting and delivery efficiency of sEVs as drugs also urgently need to be addressed. GalNAc-decorated EVs have been explored for liver-directed delivery in non-MASLD settings[123], providing a transferable targeting concept rather than direct evidence for MASLD therapy. Whether similar strategies can improve nucleic-acid delivery in MASLD requires disease-specific evaluation of biodistribution, cellular uptake, efficacy, and safety. Subsequently, it is necessary to systematically compare the effects of different ligands (GalNAc, liver receptor peptide segments, ApoA1-mimic peptides, etc.) on biodistribution, cellular uptake and immunological safety. Taken together, the clinical translation of sEV-based diagnostics and therapeutics requires a stepwise framework encompassing standardized pre-analytical sample handling, MISEV-aligned EV isolation and characterization, reproducible manufacturing and batch quality control, and systematic evaluation of biodistribution and safety. For therapeutic applications, scalable production and regulatory requirements should also be considered before clinical implementation.

As mentioned above, although there are a series of challenges in using sEVs as markers and drug carriers for the diagnosis and treatment of MASLD, with the increasing incidence of MASLD year by year, human health urgently needs precise diagnosis and treatment at the microscopic level to achieve early detection, early intervention and early treatment[124]. Future research should determine whether disease-associated EV/miRNA signatures can be reproducibly validated and whether mechanistically supported miRNA targets can be translated into therapeutic strategies[125]. Meanwhile, the balance among precise targeting, uptake efficiency, and delivery stability of engineered sEVs requires further investigation. During systemic delivery, sEVs may also undergo rapid clearance by macrophages in organs such as the liver and spleen[126].

Selective enrichment of liver-derived EVs may become increasingly important for improving the specificity of liquid-biopsy approaches in MASLD. Selective enrichment of liver-derived EVs may improve biological specificity and biomarker performance in exploratory studies[56]. However, whether such approaches provide clinically useful information for MASLD staging beyond established NITs remains to be determined in prospective, externally validated cohorts. Potential combination strategies between EV-based interventions and approved therapies warrant preclinical investigation. In March 2024, the United States Food and Drug Administration accelerated the approval of resmetirom (trade name Rezdiffra) as the first drug for non-cirrhotic MASH complicated with F2-F3 liver fibrosis, for treatment in combination with diet and exercise[10]. For example, combination strategies involving EV-based interventions and resmetirom could be explored in future preclinical studies, but their efficacy and safety require direct evaluation.

Meanwhile, studies have shown that the Lactobacillus salivarius SNK-6 derived from probiotics also has certain therapeutic potential for MASLD. Oral administration of Lactobacillus salivarius SNK-6 to mouse models can effectively reduce the levels of various indicators of fatty liver, increase the secretion of anti-inflammatory factor IL-4, and simultaneously inhibit the secretion of pro-inflammatory cytokines in intestinal cells induced by lipopolysaccharide. This finding provides a new future approach for the treatment of liver-related diseases using probiotic-derived sEVs[127].

Furthermore, future translation of sEV-based diagnostics may benefit from integrating EV-miRNA profiles, proteomics, and lipidomics with established clinical NITs, such as FibroScan and serum-based scores, using computational and machine-learning approaches[70,72]. However, the feasibility of such strategies may differ substantially across regional and healthcare-resource settings. Advanced EV isolation platforms, high-throughput proteomic and lipidomic analyses, and machine-learning-assisted multi-omics workflows often require specialized equipment, technical expertise, and substantial financial resources, which may limit their accessibility in resource-constrained settings. Therefore, future translational studies should consider not only analytical performance but also cost, workflow complexity, turnaround time, and local laboratory capacity. Simplified and standardized EV assays that can be integrated with routinely available clinical tests may ultimately be more suitable for broad clinical implementation.

CONCLUSION

sEVs offer promising avenues for molecular characterization and therapeutic development in MASLD. Human studies have provided emerging evidence for the diagnostic potential of sEVs; however, no sEV-based biomarker or therapeutic product has yet been validated for routine clinical management of MASLD. Biomarker evidence remains insufficient for routine diagnosis, disease staging, or treatment-response monitoring, while evidence supporting therapeutic applications remains predominantly preclinical and is based largely on in vitro and animal studies. Clinical translation therefore remains limited by challenges in isolation, standardization, reproducibility, and large-scale validation. Future efforts should focus on multicenter clinical validation, methodological harmonization, liver-specific EV enrichment, and rational combination strategies with approved therapies. With prospective validation and methodological standardization, integration of sEV profiling with multi-omics and computational approaches may support the future development of molecular characterization and therapeutic strategies in MASLD.