Plant-Derived Interventions for MASLD/MASH: From Extracts and Bioactive Compounds to Exosome-Like Nanoparticles

1. Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD) affects an estimated 30%-38% of adults worldwide and is now the most common chronic liver disease(Younossi et al., 2023; Younossi et al., 2025). Its spectrum extends from isolated steatosis to metabolic dysfunction-associated steatohepatitis (MASH), progressive fibrosis, cirrhosis, and hepatocellular carcinoma(Kanwal et al., 2018; Younossi et al., 2023; Younossi et al., 2025). MASLD commonly coexists with obesity, type 2 diabetes, cardiovascular disease, and chronic kidney disease, and may overlap with other liver diseases, including chronic viral hepatitis(John et al., 2026; Liang et al., 2022b; Mantovani et al., 2022; Tao et al., 2024). These comorbidities and overlapping etiologies substantially increase its clinical burden. In 2023, a multisociety consensus replaced the terms non-alcoholic fatty liver disease and non-alcoholic steatohepatitis with MASLD and MASH, respectively(Rinella et al., 2023). The revised nomenclature places metabolic dysfunction at the center of diagnosis rather than defining disease mainly by the absence of alcohol exposure.

The sequential “two-hit” model no longer captures the biology of MASLD. Current evidence supports a multiple-hit process in which insulin resistance, excess fatty-acid delivery, de novo lipogenesis, lipotoxic lipid species, mitochondrial and endoplasmic reticulum stress, immune activation, intestinal barrier dysfunction, microbial dysbiosis, and fibrogenic cell–cell interactions develop in parallel(Buzzetti et al., 2016; Loomba et al., 2021; Targher et al., 2025; Wang et al., 2025; Wang et al., 2023a). Disease progression depends not simply on hepatic triglyceride content, but on lipid composition and handling, hepatocyte adaptation to metabolic stress, and the inflammatory and fibrotic responses of non-parenchymal cells. The approval of resmetirom and semaglutide for selected adults with noncirrhotic MASH and moderate-to-advanced fibrosis has expanded the therapeutic landscape (Harrison et al., 2024; Newsome and Loomba, 2025; Petta et al., 2024; Sanyal et al., 2025). Nevertheless, responses remain heterogeneous, and neither agent addresses every metabolic, intestinal, inflammatory, and fibrogenic component of the disease. This leaves room for complementary, multi-pathway pharmacological strategies.

Medicinal and edible plants provide extracts, defined fractions, and isolated compounds with potential relevance to MASLD. Reported actions include reduced de novo lipogenesis, increased fatty-acid oxidation, improved insulin sensitivity, preservation of mitochondrial function, suppression of inflammatory signaling, and modulation of bile-acid and gut-barrier homeostasis(Cai et al., 2024; Gong et al., 2024; Kaushal et al., 2020; Liu et al., 2024; Tung et al., 2021; Wan et al., 2021; Wang et al., 2023a; Wang et al., 2023b; Yang et al., 2024; Zhang et al., 2023a; Zhang et al., 2023b). Extracts may retain interactions among multiple constituents, but their composition varies with species, cultivation, harvest, extraction, and formulation, complicating dose definition, mechanism assignment, quality control, and clinical reproducibility. Isolated compounds permit more precise target and exposure–response studies, yet many have limited oral stability or bioavailability and may not reproduce the broader effects of the parent preparation.

Plant-derived exosome-like nanoparticles (PELNs) have emerged as another form of plant pharmacology. These nanoscale lipid-bilayer preparations may contain proteins, lipids, small RNAs, and low-molecular-weight metabolites from the source material(Huang et al., 2025; Kang et al., 2016; Kim et al., 2025c; Lu et al., 2025; Sanchez-Terron et al., 2024; Wu et al., 2022; Zhang et al., 2026; Zheng et al., 2025). We use PELNs as an operational term because plant-vesicle biogenesis and classification remain unsettled. Vesicular packaging may protect selected cargo during gastrointestinal transit and facilitate interactions with intestinal epithelial, immune, or hepatic cells. Some preparations have altered intestinal permeability, microbial composition, endotoxin exposure, hepatic lipid metabolism, inflammatory signaling, or extracellular-matrix turnover. These effects are source- and preparation-dependent, however, and evidence for preferential hepatic delivery, cross-kingdom RNA regulation, or cell-selective uptake remains incomplete. Direct MASLD/MASH evidence is still fragmented and entirely preclinical.

These three evidence streams are related but not interchangeable. Extracts support multi-component pharmacology, isolated compounds allow target-focused analysis, and PELNs raise additional questions about vesicle identity, cargo, oral fate, cellular uptake, intercellular communication, and engineering. Activity of a parent plant cannot be assumed to reside in its vesicular fraction, and effects of purified vesicles must be separated from those of co-isolated phytochemicals or other non-vesicular components.

This review focuses on plant sources for which PELNs have been tested in experimental steatotic liver disease and places those studies alongside evidence from the corresponding extracts, defined fractions, and isolated compounds. We first summarize the disease processes most relevant to plant-derived intervention, then examine the preparation, composition, oral behavior, cellular uptake, and pharmacological effects of PELNs. Clinical evidence for source-plant preparations is considered separately, followed by the methodological and translational requirements for PELN development.

2. Literature Search and Evidence Selection

A structured search of PubMed, Web of Science Core Collection, Scopus, and Embase was conducted from database inception to July 2026. Google Scholar was used for citation tracking. The search proceeded in two stages. Terms related to plant-derived vesicles, including “plant exosome,” “plant-derived extracellular vesicle,” “exosome-like nanoparticle,” “exosome-like nanovesicle,” and “nanovesicle”, were first combined with “MASLD,” “MAFLD,” “NAFLD,” “MASH,” and “NASH.” The plant sources identified in this search were then combined with “extract,” “fraction,” “phytochemical,” “bioactive compound,” and “natural product” to retrieve evidence on the corresponding extracts, defined fractions, isolated constituents, and clinical studies.

Eligible primary studies included cellular, animal, and human investigations reporting outcomes relevant to steatosis, glucose or lipid metabolism, oxidative stress, mitochondrial function, inflammation, intestinal barrier integrity, gut microbiota, bile-acid signaling, cell death, autophagy, hepatic stellate-cell activation, or fibrosis. The core PELN dataset was restricted to studies that directly evaluated isolated, enriched, or engineered plant-derived vesicles in MASLD-, MASH-, NAFLD-, or NASH-related models. Reviews were used for background and citation tracing only. Duplicate reports, conference abstracts lacking sufficient data, unrelated studies, and complex multi-herbal formulations were excluded from the mechanistic evidence set. Source-plant extracts and isolated compounds were assessed separately from the corresponding PELNs.

The core PELN dataset comprised 12 primary studies covering 11 plant sources. Most sources were represented by a single report, and the studies differed in isolation method, vesicle characterization, disease model, dose, route, and outcome assessment. The evidence was therefore synthesized narratively rather than quantitatively.

3. Pathological Modules Relevant to Plant-Derived Interventions

MASLD develops through interacting metabolic, cellular, and tissue-level disturbances rather than a single sequential pathogenic event. The major processes relevant to plant-derived intervention include genetic and metabolic susceptibility, lipotoxicity and insulin resistance, mitochondrial and endoplasmic reticulum stress, inflammatory and immune activation, gut–liver axis dysfunction, regulated cell death, impaired autophagy, and fibrogenic tissue remodeling (Figure 1).

Host genetics modifies both susceptibility to hepatic steatosis and the likelihood of disease progression. The PNPLA3 I148M variant remains the most consistently associated genetic risk factor, whereas variants in TM6SF2, MBOAT7, GCKR, HSD17B13, and MTARC1 influence hepatic lipid handling, inflammation, fibrosis, and liver-related outcomes to different degrees(Chen et al., 2025; Pei and Goh, 2025; Romeo and Valenti, 2025; Sveinbjornsson et al., 2022). These variants do not act independently of metabolic stress. Their effects are amplified by obesity, insulin resistance, adipose-tissue lipolysis, dietary substrate excess, and impaired hepatic lipid disposal.

Lipotoxicity, rather than triglyceride accumulation alone, is a central driver of hepatocellular injury. Triglyceride storage may initially buffer excess fatty acids, whereas diacylglycerols, ceramides, free cholesterol, and oxidized phospholipids disrupt membrane integrity, insulin signaling, calcium homeostasis, and organelle function(Dua et al., 2025; Iturbe-Rey et al., 2025; Li et al., 2025c; Nakagawa, 2025; Venkatesan et al., 2023; Yang et al., 2025b; Zhong et al., 2024). Increased de novo lipogenesis through SREBP-1c and carbohydrate-responsive element-binding protein, excessive fatty-acid delivery from adipose tissue, and insufficient mitochondrial or peroxisomal oxidation increase the intracellular burden of toxic lipids. Early mitochondrial adaptation may compensate for substrate overload, but sustained excess eventually impairs respiration, raises reactive oxygen species, and damages mitochondrial DNA. Activation of the PERK, IRE1, and ATF6 branches of the unfolded-protein response further links endoplasmic reticulum stress to insulin resistance, inflammation, and hepatocyte death(Dua et al., 2025; Iturbe-Rey et al., 2025; Koliaki et al., 2015; Li et al., 2025c; Nakagawa, 2025; Steinberg et al., 2025; Venkatesan et al., 2023).

Injured hepatocytes sustain inflammation by releasing damage-associated molecular patterns, oxidized lipids, mitochondrial DNA, and extracellular vesicles. These signals activate Kupffer cells, recruited macrophages, and other immune populations through NLRP3, cGAS–STING, TLR4, NF-κB, and related pathways(Henao-Mejia et al., 2012; Li et al., 2024; Mridha et al., 2017; Yu et al., 2019; Zhu et al., 2025). Immune activation is accompanied by metabolic reprogramming: macrophages and lymphocyte subsets alter glucose and lipid use, which shapes cytokine output and hepatocyte–immune-cell crosstalk. Apoptosis, pyroptosis, necroptosis, and ferroptosis further amplify injury by releasing intracellular contents and lipid-peroxidation products(Peleman et al., 2024; Zhao et al., 2024).

The gut–liver axis supplies an additional source of metabolic and inflammatory signals. Dysbiosis and impaired epithelial barrier function increase portal exposure to lipopolysaccharide and other microbial products, while altered short-chain fatty acids and bile acids affect host metabolism and immunity(Lau et al., 2025; Yu et al., 2024). The LPS–TLR4–NF-κB pathway and FXR–FGF19 signaling connect intestinal dysfunction with hepatic inflammation, insulin resistance, and lipid accumulation. These pathways are especially relevant to oral plant preparations and PELNs, which may act within the intestine without high systemic exposure

Fibrosis arises from sustained communication among injured hepatocytes, inflammatory macrophages, and hepatic stellate cells. Single-cell and spatial studies show that fibrogenesis is organized within discrete tissue niches rather than uniformly across the liver(Kim et al., 2025a; Li et al., 2025a; Li et al., 2025d; Martin et al., 2025; Matchett et al., 2024; Schwabe et al., 2025; Xu et al., 2025). Lipid-associated macrophages, activated stellate-cell subsets, damaged hepatocytes, and matrix-producing cells form localized networks of persistent inflammation, extracellular-matrix deposition, and impaired regeneration. This spatial organization helps explain why reducing steatosis alone may not reverse established MASH fibrosis.

These modules provide the framework for the evidence reviewed below. Extracts and isolated compounds have mainly been studied for effects on lipid metabolism, oxidative stress, inflammation, gut microbiota, autophagy, and stellate-cell activation. PELN studies address many of the same processes but add questions about oral stability, cargo, cellular uptake, intercellular signaling, and intestinal or hepatic exposure.

4. Pharmacology of Source-Plant Extracts and Bioactive Compounds

Extracts, defined fractions, and isolated compounds from the plant sources represented in the PELN literature have been investigated more extensively than their vesicular preparations. Most studies have focused on hepatic lipid handling, mitochondrial stress, inflammatory signaling, gut–liver communication, regulated cell death, autophagy, and fibrogenesis (Figure 2 and Table 1). Detailed study-level information on experimental models, outcomes, proposed mechanisms, and PubMed identifiers is provided in Supplementary Table S1.

Most evidence remains preclinical. Many reports describe pathway-associated proteins or circulating biomarkers, whereas relatively few use gene deletion, pharmacological blockade, metabolic-flux analysis, or cell-specific models to establish causality. The sections below therefore distinguish repeated pathway associations from direct mechanistic validation.

4.1. Regulation of Hepatic Lipid Metabolism and Insulin Sensitivity

Hepatic steatosis reflects an imbalance among fatty-acid uptake, de novo lipogenesis, fatty-acid oxidation, lipid export, and storage. Plant extracts and isolated compounds have been reported to influence each of these processes, most commonly through AMPK, SIRT1, PPARα, SREBP-1c, and bile-acid receptor signaling(Chiang and Ferrell, 2020; Hardie and Pan, 2002; Kersten and Stienstra, 2017; Nagaya et al., 2010; Wan et al., 2011).

AMPK activation is one of the most frequently reported responses. Extracts or defined fractions from honeysuckle, tea, pomegranate, Artemisia capillaris, blueberry, ginger, garlic, and ginseng have been associated with lower expression of SREBP-1c, ACC, fatty-acid synthase, and stearoyl-CoA desaturase 1, together with higher expression of PPARα, carnitine palmitoyltransferase 1A, or acyl-CoA oxidase 1(Chun et al., 2018; Cui et al., 2023; Han et al., 2023; Huang et al., 2017; Kim et al., 2018; Kim et al., 2013; Lee et al., 2016; Li et al., 2020; Liang et al., 2022a; Liu et al., 2018; Liu et al., 2023b; Mohammed, 2022; Park et al., 2019; Raffaele et al., 2020; Ren et al., 2014; Ren et al., 2019; Wu et al., 2025; Xia et al., 2019; Xiao et al., 2013; Yan et al., 2017; Yang et al., 2023b; Yuan et al., 2025). This convergence suggests a common metabolic response, but it does not establish AMPK as the direct target of each preparation.

Ginseng illustrates the greater mechanistic resolution possible with isolated compounds. Compound K promotes fatty-acid oxidation through AMPK/ACC and PPARα-associated signaling, whereas ginsenoside Rb2 restores autophagy under lipotoxic conditions through AMPK and SIRT1(Huang et al., 2017; Kim et al., 2013). Ginsenoside Rk1 reduces CD36 expression and mTORC1–SREBP-1c signaling, affecting both fatty-acid uptake and synthesis(Wu et al., 2025). Ginsenosides Rd and Rc have also been reported to activate SIRT6 and enhance PPARα-dependent oxidation(Cui et al., 2023; Yang et al., 2023b). Most of these studies, however, remain confined to hepatocyte cultures or rodent models.

Ginger-derived constituents show a similar pattern. Zingerone and 6-gingerol activate AMPK-related pathways and reduce SREBP-dependent lipogenesis in high-fat- or fructose-based models(Liu et al., 2023b; Mohammed, 2022). Garlic organosulfur compounds, including S-allyl cysteine and S-allylmercaptocysteine, suppress lipogenesis and restore autophagy through AMPK-related pathways(Hwang et al., 2013; Xiao et al., 2013). Blueberry leaf polyphenols and blueberry preparations combined with probiotics have been linked to SIRT1/PGC-1α/SIRT3 signaling, improved mitochondrial lipid utilization, and reduced hepatic lipid accumulation(Li et al., 2020; Ren et al., 2014; Ren et al., 2017; Ren et al., 2019).

Some interventions act beyond the AMPK–SREBP axis. Broccoli preparations regulate hepatic lipid uptake and export through CD36, microsomal triglyceride transfer protein, and FXR/LXR-associated pathways(Chen et al., 2016; Lu et al., 2025). Yellow tea polysaccharides alter bile-acid composition and the FXR–FGF15–CYP7A1 axis, whereas fermented dark tea has been associated with improved PI3K/AKT signaling and reduced gluconeogenic gene expression(Huang et al., 2024; Yuan et al., 2025). A. capillaris extracts have been linked to both SIRT1-dependent hepatic lipid metabolism and microbiota-associated production of short-chain fatty acids(Liang et al., 2022a; Liang et al., 2025).

Across these studies, the most consistent findings are reduced de novo lipogenesis and fatty-acid uptake, increased fatty-acid oxidation, and changes in bile-acid or insulin signaling. Lower hepatic triglycerides or serum lipids should not be taken as evidence of broad metabolic correction unless accompanied by measurements of insulin sensitivity, lipid flux, mitochondrial function, or histological disease activity.

4.2. Mitochondrial Homeostasis and Oxidative Stress

Mitochondrial dysfunction contributes to MASLD progression by impairing oxidative metabolism and increasing the production of reactive oxygen species. Prolonged lipid overload can also disrupt mitochondrial membrane potential, respiratory-chain activity, ATP production, and mitochondrial quality control. Plant-derived interventions have been studied mainly through the Nrf2 antioxidant pathway, SIRT1/PGC-1α-dependent mitochondrial biogenesis, and autophagy or mitophagy.

Blueberry polyphenols and anthocyanin-rich preparations reduce lipid peroxidation and increase glutathione or antioxidant-enzyme activity in experimental models(Li et al., 2020; Ren et al., 2014; Ren et al., 2019). Blueberry juice combined with probiotics has also been reported to improve mitochondrial morphology and respiratory function through SIRT1/PGC-1α-associated signaling(Ren et al., 2019). These findings suggest an effect beyond nonspecific radical scavenging, although the contribution of the probiotic component cannot always be separated from that of blueberry constituents.

Pomegranate preparations provide another well-studied example. Pomegranate seed oil and punicalagin have been associated with enhanced antioxidant responses, improved mitochondrial function, and reduced hepatic injury in diet-induced steatotic liver disease (Raffaele et al., 2020; Zou et al., 2014). Punicalagin has additionally been linked to p62/Nrf2 signaling, enhanced autophagy, and suppression of the TXNIP–NLRP3 axis(Ma et al., 2025). Pomegranate seed oil improves antioxidant capacity and mitochondrial markers, although its lipid composition and metabolic effects differ substantially from those of polyphenol-rich pomegranate extracts.

Ginseng preparations act through several mitochondrial pathways. Wild ginseng cambial meristematic cells increase PGC-1α, NRF1, and TFAM expression, consistent with enhanced mitochondrial biogenesis(Lee et al., 2016). Ginsenoside Rg1 maintains FOXO1-associated antioxidant responses and reduces lipid peroxidation in an aging-related fatty-liver model(Qi et al., 2020). Fermented ginseng preparations have also been linked to restoration of mitophagy and suppression of excessive mTORC1 activity(Choi et al., 2019).

Mitochondrial homeostasis is therefore a plausible target of several plant-derived interventions, but the depth of evidence varies. Many studies rely on reactive oxygen species, malondialdehyde, antioxidant enzymes, or expression of mitochondrial regulators; far fewer measure oxygen consumption, respiratory-chain function, mitochondrial turnover, or substrate oxidation. Antioxidant biomarkers alone should not be described as restoration of mitochondrial function.

4.3. Inflammatory Signaling and Immune-Metabolic Regulation

Inflammatory pathways help drive the transition from steatosis to MASH and fibrosis. Plant-derived preparations have been studied for effects on pattern-recognition receptors, NF-κB signaling, inflammasome activation, cytokine production, and macrophage function.

Black ginseng reduces TLR4 expression, NF-κB p65 phosphorylation, and inflammatory or fibrosis-associated genes in a Western-diet and carbon tetrachloride model(Wei et al., 2022). Dried tangerine peel polysaccharide binds MD-2 and attenuates TLR4-associated inflammatory and endoplasmic reticulum stress responses(Wang et al., 2024). Ginger essential oil suppresses the microbiota–LPS–TLR4 pathway and downstream NLRP3 inflammasome activation in experimental NASH(Panyod et al., 2024). These studies indicate that plant preparations may intervene at several points upstream of cytokine production rather than simply lowering circulating TNF-α or IL-6.

Several ginseng-derived compounds have been investigated more directly. The ginsenoside metabolite 20(S)-protopanaxatriol inhibits NLRP3 inflammasome activation and reduces inflammation and fibrosis in a methionine- and choline-deficient model(Lu et al., 2023). Panaxydol has been reported to interact with the ATP-binding region of NLRP3, suppressing caspase-1 activation, cytokine maturation, and pyroptosis(Kim et al., 2024b). Ginsenoside Re affects both PI3K/AKT and TLR4/NF-κB signaling, whereas fermented red ginseng preparations have been associated with altered macrophage inflammatory programs through PPARα and mTORC1(Choi et al., 2019; Zhang et al., 2024).

Blueberry preparations combined with probiotics increased interleukin-22 and activated Janus kinase 1/signal transducer and activator of transcription 3 signaling in experimental fatty liver disease(Zhu et al., 2018). Pomegranate, green tea, honeysuckle, ginger, and cannabis-derived preparations have also reduced inflammatory markers in animal models(Carmona-Hidalgo et al., 2021; Gong et al., 2023; Han et al., 2023; Kim et al., 2018; Liu et al., 2018; Ma et al., 2025; Noori et al., 2017; Panyod et al., 2024; Park et al., 2019; Tzeng et al., 2015; Xia et al., 2019; Zou et al., 2014). In most cases, however, immune effects were inferred from liver cytokines, histology, or bulk-tissue protein expression. Immune-cell depletion, conditional gene deletion, single-cell profiling, and adoptive-transfer experiments remain uncommon.

Several plant-derived interventions therefore show anti-inflammatory activity, but evidence for cell-specific immune reprogramming is much weaker. Terms such as “macrophage polarization” or “immune-metabolic remodeling” should be reserved for studies with functional or cell-resolved data.

4.4. Modulation of the Gut–Liver Axis

The gut–liver axis is especially relevant to orally administered plant preparations. Changes in microbial composition, epithelial permeability, microbial metabolites, and bile-acid signaling can alter the inflammatory and metabolic signals reaching the liver.

Ginseng extracts and polysaccharides from Panax japonicus have been reported to reduce intestinal permeability, modify microbial composition, and alter short-chain fatty-acid or lipid-associated metabolic pathways in high-fat-diet models(Wu et al., 2024). Pomegranate peel extract, pomegranate-derived supplements, punicalagin, and punicic acid influence microbial communities and metabolites alongside improvements in hepatic inflammation and lipid metabolism(Chen et al., 2024; Liu et al., 2021; Sanchez-Terron et al., 2024). Broccoli preparations, yellow tea polysaccharides, and fermented dark tea have likewise been associated with increased abundance of selected beneficial taxa, altered bile-acid metabolism, and improved insulin sensitivity(Huang et al., 2024; Lu et al., 2025; Yuan et al., 2025; Zandani et al., 2021).

A. capillaris water extract altered microbial composition and increased short-chain-fatty-acid production while activating hepatic PPARα and suppressing NF-κB signaling(Liang et al., 2022a; Liang et al., 2025). Two cannabis extracts also changed microbial diversity in a high-fat-diet model, but produced modest metabolic effects and did not establish a clear host target(Zandani et al., 2022). A shift in microbial composition alone does not demonstrate a therapeutically meaningful effect on liver disease.

Combination studies require separate interpretation. Honeysuckle combined with metformin improved steatosis, insulin resistance, endotoxin exposure, and microbial composition in OLETF rats(Shin et al., 2017), but the contribution of honeysuckle cannot be separated from that of metformin.

Most microbiome studies remain associative. Changes in individual bacterial genera are often interpreted as beneficial or harmful without direct testing of their functional contribution. Antibiotic depletion, fecal microbiota transplantation, gnotobiotic models, microbial-metabolite tracing, and barrier-specific experiments are still uncommon. The strongest conclusions are therefore limited to coordinated changes in microbiota, barrier markers, metabolites, and hepatic outcomes rather than proof of a microbiota-dependent mechanism.

4.5. Autophagy, Regulated Cell Death, and Fibrogenesis

Progression from steatosis to MASH involves impaired autophagy, hepatocyte death, inflammatory amplification, hepatic stellate-cell activation, and extracellular-matrix accumulation. Several plant-derived compounds influence these processes, although their effects often overlap with metabolic and antioxidant pathways.

S-allylmercaptocysteine reduced hepatocyte apoptosis and enhanced autophagy through LKB1–AMPK and PI3K–AKT–mTOR signaling(Xiao et al., 2013). Ginsenoside Rg1 increased BCL-2, reduced BAX, and preserved viability in lipid-challenged hepatocytes(Qi et al., 2020). The blueberry-derived isoflavone tectorigenin was linked to tRF-47-dependent control of autophagy and pyroptosis in experimental NASH(Zhu et al., 2022). Fermented red ginseng restored mitophagy while reducing mTORC1 activity and inflammatory signaling(Choi et al., 2019).

Inflammasome-dependent pyroptosis provides a mechanistic link between hepatocyte injury and fibrogenesis. Ginseng-derived protopanaxatriol and panaxydol suppress NLRP3–caspase-1 signaling and reduce inflammatory liver injury(Kim et al., 2024b; Lu et al., 2023). These compound-level studies are more informative than reports based solely on changes in IL-1β or caspase-1 expression because they include evidence of direct or proximal pathway regulation.

Evidence for anti-fibrotic activity comes mainly from prevention models. Black ginseng reduces collagen deposition and fibrosis-associated gene expression in a Western-diet plus carbon tetrachloride model(Wei et al., 2022). Honeysuckle ethanol extract attenuates steatohepatitis and fibrotic changes in a methionine- and choline-deficient model(Tzeng et al., 2015). Cannabis-derived compounds have also been associated with reduced inflammatory and fibrotic responses(Carmona-Hidalgo et al., 2021). However, these models do not necessarily demonstrate regression of established fibrosis. Few studies begin treatment after advanced fibrosis is present, and even fewer assess matrix turnover, stellate-cell heterogeneity, portal pressure, or durable fibrosis reversal.

Source-plant extracts and isolated compounds act across metabolic, mitochondrial, inflammatory, intestinal, and fibrogenic pathways relevant to MASLD/MASH. Extracts offer broader but less standardized pharmacology; isolated compounds permit clearer attribution but may have limited oral exposure. These data provide context for PELNs, but they do not show that vesicular fractions contain the same active constituents or reproduce the effects of the parent preparation.

5. PELNs in MASLD/MASH: Current Preclinical Evidence

The direct PELN evidence base comprises 12 preclinical studies covering 11 plant sources, including native preparations and engineered vesicle-based systems. Reported outcomes include reduced steatosis, oxidative stress, inflammation, intestinal barrier dysfunction, endotoxemia, or fibrosis. Most sources are represented by a single study, and differences in preparation, disease model, dose, and outcome assessment limit cross-study comparison.

PELNs are commonly described as natural carriers of plant lipids, proteins, RNAs, and small molecules. Their activity may instead reflect several components: the vesicle membrane, encapsulated cargo, membrane-associated molecules, or non-vesicular compounds carried through purification. Because these components are rarely separated experimentally, PELNs should be treated as complex biological preparations rather than assumed to be purified analogues of mammalian exosomes.

5.1. Terminology, Isolation, and Physicochemical Characterization

Terminology, isolation procedures, and physicochemical characterization remain inconsistent across the PELN literature. Published studies use differential ultracentrifugation, polyethylene glycol precipitation, tangential-flow filtration, size-exclusion chromatography, or combinations of these methods, each of which differs in recovery, purity, scalability, and risk of co-isolating soluble plant components (Figure 3 and Table 2).

Differential centrifugation followed by ultracentrifugation is the most common isolation route in MASLD-related studies and has been used for pomegranate, hemp sprout, tangerine peel, broccoli, and A. capillaris PELNs(Kim et al., 2024a; Kim et al., 2025c; Xu et al., 2024; Zhang et al., 2026; Zou et al., 2024). Honeysuckle-derived PELNs were isolated from dried honeysuckle as a vesicle-enriched preparation and had a reported mean diameter of approximately 152.7 nm(Li et al., 2025b). Other reports used polyethylene glycol precipitation, filtration, tangential-flow filtration, size-exclusion chromatography, or combined methods(Guo et al., 2025; Kim et al., 2025b; Lei et al., 2025; Liu et al., 2023a; Ma et al., 2026; Zhao et al., 2022). Ultracentrifugation may promote aggregation and co-isolation; polymer precipitation increases recovery at the expense of purity; tangential-flow filtration is more scalable but sensitive to membrane and process conditions; and size-exclusion chromatography improves separation from soluble proteins but yields diluted fractions.

Reported particle diameters generally fall within the nanoscale range, although size distributions vary across plant sources and preparation methods. Engineered hybrid systems may be larger than native vesicular preparations. Particle size alone is insufficient to establish vesicle identity or purity. At minimum, PELN characterization should combine particle-size analysis, particle concentration, morphology, membrane-lipid assessment, total protein and RNA measurements, and evaluation of non-vesicular contaminants. Few studies report all of these parameters, and plant-specific reference markers remain unavailable.

This methodological heterogeneity has direct pharmacological consequences. Changes in isolation conditions can alter vesicle yield, membrane composition, cargo recovery, colloidal stability, and cellular uptake. Comparisons between studies should therefore account for preparation method rather than treating all PELNs from the same plant as equivalent products.

5.2. Endogenous Cargo and Evidence for Cargo-Dependent Activity

PELN preparations contain varying combinations of membrane lipids, proteins, small RNAs, phytochemicals, and other metabolites (Figure 4). The reported functional effects and strength of mechanistic evidence for individual PELN preparations are summarized in Table 3.

Garlic-derived PELNs provide one of the clearest examples of cargo-specific activity. Vesicular miR-396e reduced macrophage glycolysis by targeting PFKFB3, lowered inflammatory mediator production, and indirectly improved hepatocyte lipid metabolism through macrophage–hepatocyte communication(Liu et al., 2023a). This study moved beyond cargo profiling by testing a defined vesicular RNA.

Small RNAs in A. capillaris-derived PELNs have also been implicated in hepatic lipid metabolism and inflammation, but the contribution of individual RNAs to the overall preparation remains uncertain. Tea-derived PELNs altered host miRNAs and metabolic regulators in oleic-acid-treated HepG2 cells(Lei et al., 2025). Because the latter evidence is largely in vitro, its relevance to oral delivery and whole-body pharmacology remains unclear.

Ginseng-derived vesicles contain ginsenosides, lipids, proteins, and RNAs(Kim et al., 2025b). Their antifibrotic effect has been linked to reduced TIMP2 expression in hepatic stellate cells, but it is not yet clear whether this response is mediated by a specific encapsulated ginsenoside, an RNA cargo, membrane lipids, or the combined preparation. Broccoli-derived vesicles contain multiple small RNAs associated with metabolic and immune pathways, but most of these associations arise from sequencing and pathway prediction rather than direct cargo-level validation(Zhang et al., 2026).

Selected PELNs clearly contain biologically active material, but cargo inventories are not mechanisms. Intact vesicles should be compared with disrupted vesicles, cargo-depleted preparations, free-cargo controls, and reconstituted vesicles to distinguish vesicle-dependent delivery from the effects of co-isolated or soluble plant constituents.

5.3. Oral Fate, Intestinal Effects, and Gut–Liver Communication

Oral administration is frequently considered a potential advantage of PELNs. Available studies suggest that selected preparations may interact with the intestinal epithelium, microbiota, and mucosal immune system, thereby influencing barrier integrity, microbial metabolites, and endotoxin exposure (Figure 5A and B).

Several PELNs appear to act mainly through the intestinal compartment. Broccoli-derived PELNs reduced steatosis and inflammation while improving barrier markers, microbial composition, and bile-acid-related pathways(Zhang et al., 2026). Pomegranate-derived PELNs reduced gut leakage and liver injury(Kim et al., 2025c), and hemp sprout-derived PELNs attenuated fibrosis together with lower endotoxemia and oxidative stress(Kim et al., 2024a). Honeysuckle-derived PELNs improved barrier function and reduced steatosis, inflammation, and collagen deposition in high-fat-diet-fed mice; these effects disappeared after antibiotic depletion of the microbiota and were transferred by fecal microbiota transplantation, supporting a microbiota-dependent mechanism(Li et al., 2025b). Tangerine peel-derived PELNs altered gut microbiota and hepatic lipid metabolism in diabetes-associated steatosis(Zou et al., 2024).

These findings are consistent with a gut–liver axis-associated effect, but they do not demonstrate preferential hepatic targeting. Improvement in liver outcomes after oral administration may result from local intestinal activity, reduced endotoxin translocation, altered microbial metabolism, systemic absorption of released cargos, or direct uptake of intact vesicles. These routes have not been distinguished in most studies.

Claims of cross-kingdom RNA regulation also require caution. Plant-derived RNAs can be detected in vesicular preparations, and selected studies support functional effects in mammalian cells. Nevertheless, RNA abundance, sequence stability, cellular exposure, and target specificity are not consistently reported. Cross-kingdom regulation should therefore be discussed as a source-dependent mechanism supported in selected experimental systems rather than a general property of all PELNs.

A more informative assessment of oral fate will require labeled-vesicle tracking, measurement of intact particle recovery, distinction between membrane and cargo distribution, and quantitative analysis of intestinal, portal, hepatic, and systemic exposure. Without these data, oral efficacy cannot be equated with oral bioavailability of intact PELNs.

5.4. Cellular Uptake and Intercellular Effects

Uptake of selected PELNs by hepatocytes, macrophages, and hepatic stellate cells has been reported in specific experimental systems (Figure 5C). These findings indicate cell-type-associated effects but do not establish universal cell selectivity or preferential hepatic targeting.

A. capillaris-derived PELNs entered AML12 hepatocytes through macropinocytosis and caveolin-associated endocytosis and reduced lipid accumulation and inflammatory signaling(Xu et al., 2024). Garlic-derived PELNs acted mainly through macrophages: suppression of macrophage PFKFB3 and glycolytic activity lowered inflammatory output and indirectly improved hepatocyte metabolism(Liu et al., 2023a). These studies illustrate both direct hepatocyte uptake and intercellular mechanisms.

Ginseng-derived vesicles were taken up by hepatic stellate cells and reduced TIMP2 expression, favoring extracellular-matrix degradation(Kim et al., 2025b). This observation is relevant to fibrogenesis, but the study does not establish whether the vesicles selectively target stellate cells in vivo. Similarly, intestinal epithelial uptake has been reported or inferred for several orally administered preparations, but cell specificity has rarely been quantified against other intestinal and hepatic cell populations.

Current data show uptake by hepatocytes, macrophages, or hepatic stellate cells in specific models, but not universal cell selectivity. Whether these patterns reflect active targeting, tissue exposure, membrane composition, or experimental design remains unresolved.

Future studies should use quantitative cell-resolved biodistribution, receptor-blocking experiments, and competitive uptake models. These approaches would help determine whether uptake is mediated by specific membrane ligands or represents nonspecific internalization of nanoscale lipid particles.

5.5. Native and Engineered PELN Systems

Most studies have examined native or minimally processed preparations, which retain endogenous plant cargo but offer limited control over composition, dose, and distribution. Engineering approaches seek greater control through drug loading, membrane fusion, ligand modification, or hybridization with synthetic nanocarriers.

A ginger-derived vesicle–liposome hybrid loaded with berberine and modified with ursodeoxycholic acid illustrates the engineering potential of PELN-based systems (Figure 5D) (Ma et al., 2026). Because its activity reflects the combined contributions of plant-derived vesicles, synthetic lipids, berberine, and a targeting ligand, it should be distinguished from native ginger PELNs.

The ginger hybrid demonstrates the engineering potential of plant vesicles, but it does not show that native ginger PELNs are intrinsically liver-targeted. Native preparations and engineered hybrids require separate evaluation because their manufacturing controls, pharmacokinetics, safety, and regulatory pathways differ.

Processing may also modify native vesicle composition. Fermentation of tea-derived vesicles altered particle characteristics and metabolic profiles and was associated with enhanced lipid-lowering and anti-inflammatory activity in an experimental NAFLD model(Guo et al., 2025). This finding suggests that plant processing can influence PELN composition, but it also raises questions about batch reproducibility and whether the final product should still be considered equivalent to vesicles from untreated plant material.

Engineering may broaden the use of plant vesicles, but once synthetic lipids, drugs, or targeting ligands are added, the product is a composite nanomedicine rather than a native PELN.

5.6. Strength and Limitations of the Current Evidence

The studies provide proof of principle that selected PELNs can affect metabolic, inflammatory, intestinal, or fibrotic processes relevant to MASLD/MASH. Mechanistic confidence is highest when a defined cargo or causal intermediary is tested, as in the garlic miR-396e–PFKFB3 pathway and the honeysuckle study using microbiota depletion and fecal transfer(Li et al., 2025b; Liu et al., 2023a). Most other mechanisms rest on pathway-associated proteins, microbiota profiles, or predicted RNA targets.

Several limitations recur across the literature. First, most plant sources are represented by a single study, with little independent replication. Second, isolation and characterization procedures vary widely, making it difficult to determine whether studies examined comparable vesicle populations. Third, few reports distinguish intact vesicle activity from the effects of co-isolated polyphenols, lipids, proteins, or soluble RNAs. Fourth, dose is reported using inconsistent units, including particle number, protein content, vesicle mass, or source-plant mass. Fifth, quantitative pharmacokinetic and biodistribution data are scarce. Finally, most studies use short-term steatosis or early inflammatory models rather than models with established MASH fibrosis.

Direct comparisons with source-plant extracts and isolated compounds are also uncommon. Such comparisons are needed to determine whether vesicular delivery improves potency, stability, intestinal retention, cellular uptake, or therapeutic index. Without head-to-head experiments, the proposed advantages of PELNs remain largely inferential.

The present evidence supports PELNs as an emerging preclinical extension of plant-derived pharmacology but does not yet establish them as a reproducible therapeutic class. Progress will require standardized preparation, rigorous cargo validation, quantitative exposure studies, independent replication, and testing in models that reflect established MASH and fibrosis.

6. Clinical Evidence for Source-Plant Interventions and the Translational Gap of PELNs

Clinical studies have evaluated Korean red ginseng, ginger, garlic, pomegranate, green tea, and selected herbal formulations in patients classified under earlier NAFLD/NASH criteria(Baghdadi et al., 2025; Ekhlasi et al., 2015; Goodarzi et al., 2021; Hong et al., 2016; Jafarirad et al., 2023; Rafie et al., 2020; Rahimlou et al., 2016; Sakata et al., 2013; Soleimani et al., 2016; Soleimani et al., 2020; Soleimani et al., 2021; Xu et al., 2026). Some trials reported improvements in liver enzymes, insulin resistance, lipids, body composition, inflammation, or imaging-based steatosis, but formulations, sample sizes, treatment duration, and outcome measures varied substantially (Table 4).

Randomized studies of ginger and garlic suggested modest metabolic or hepatic benefit, although most were small, short, and based on ultrasonography or biochemical markers rather than quantitative imaging or histology(Rafie et al., 2020; Rahimlou et al., 2016; Soleimani et al., 2016; Soleimani et al., 2020; Soleimani et al., 2021). Pomegranate juice, extract, and peel preparations affected metabolic indices, liver enzymes, inflammatory markers, and steatosis; one trial also measured liver stiffness and the hepatorenal ultrasound index(Baghdadi et al., 2025; Ekhlasi et al., 2015; Goodarzi et al., 2021; Jafarirad et al., 2023). Catechin-rich green tea improved biochemical measures and the liver-to-spleen computed-tomography attenuation ratio, while a trial of Yinchenhao decoction used magnetic resonance imaging proton-density fat fraction as its primary endpoint(Sakata et al., 2013; Xu et al., 2026). The decoction is a multi-herbal formulation, so its effect cannot be assigned specifically to A. capillaris. Observational associations involving cannabis remain vulnerable to confounding and do not support therapeutic use(Adejumo et al., 2017; Vazquez-Bourgon et al., 2019).

Most trials measured aminotransferases, triglycerides, cholesterol, body weight, insulin resistance, inflammatory markers, or semiquantitative ultrasonographic steatosis. These outcomes may reflect metabolic improvement but do not establish resolution of steatohepatitis or reversal of fibrosis. Quantitative liver-fat imaging, elastography, validated fibrosis biomarkers, and histology were uncommon, and no source-plant trial demonstrated accepted histological endpoints for MASH resolution or fibrosis improvement. Safety and herb–drug interactions were also reported inconsistently.

Clinical activity of a juice, powder, extract, or herbal formula does not validate the corresponding PELN. No purified or enriched PELN formulation has entered clinical testing for MASLD/MASH, and vesicular fractions differ from parent preparations in composition, dose, biodistribution, and pharmacokinetics.

Before human testing, candidate PELNs need reproducible manufacturing, defined quality attributes, standardized particle and cargo measurements, stability data, biodistribution, and repeated-dose safety assessment. Early studies should establish tolerability, exposure, and pharmacodynamic activity before testing efficacy with quantitative liver-fat and fibrosis endpoints.

7. Translational Challenges and Future Directions

PELNs remain at an early preclinical stage. Most plant sources are represented by one study, and differences in isolation, characterization, disease model, dose, and outcome assessment limit reproducibility(Guo et al., 2025; Kim et al., 2025b; Kim et al., 2024a; Kim et al., 2025c; Lei et al., 2025; Liu et al., 2023a; Ma et al., 2026; Xu et al., 2024; Zhang et al., 2026; Zhao et al., 2022; Zou et al., 2024). The immediate task is to define comparable products and reproduce key findings, not simply to add more plant sources.

Standardized preparation and quality control are essential. Studies should report source material, tissue, harvest and processing conditions, particle concentration and size distribution, morphology, membrane composition, cargo content, stability, contaminants, and batch consistency. Isolation methods differ substantially in recovery and purity(Guo et al., 2025; Kim et al., 2025b; Kim et al., 2024a; Lei et al., 2025; Liu et al., 2023a; Ma et al., 2026; Nemati et al., 2022; Ning et al., 2025; Xu et al., 2024; Yang et al., 2025a; Zhao et al., 2022), so method-specific effects must be considered when comparing studies.

Mechanistic work must separate vesicle-mediated effects from co-isolated phytochemicals, proteins, lipids, and free RNAs. Garlic miR-396e is a useful model of cargo-dependent activity(Liu et al., 2023a), but most proposed mechanisms are based on sequencing or downstream markers (Kim et al., 2025b; Lei et al., 2025; Xu et al., 2024; Zhang et al., 2026). Intact and disrupted vesicles, vesicle-free fractions, matched free cargo, and source extracts should be compared directly.

Oral fate and biodistribution remain poorly resolved. Benefit after oral dosing may arise from intestinal activity, microbiota modulation, reduced endotoxin translocation, released cargo, or uptake of intact particles(Kim et al., 2024a; Kim et al., 2025c; Zhang et al., 2026; Zou et al., 2024). Quantitative studies should distinguish these routes and measure intestinal, portal, hepatic, and cell-specific exposure.

Therapeutic claims also require models with established MASH and fibrosis rather than prevention of early steatosis alone. Candidate PELNs should be compared with the corresponding source extract, major free compounds, and appropriate pharmacological controls. Native PELNs and engineered hybrids must be evaluated separately because their composition, pharmacology, safety, and regulatory requirements differ(Ma et al., 2026) .

No purified PELN has entered clinical testing for MASLD/MASH. Translation will require scalable manufacturing, validated potency assays, quantitative biodistribution, repeated-dose toxicology, and liver-specific pharmacodynamic endpoints. Until these requirements are met, PELNs should be described as an emerging preclinical platform rather than an established therapeutic class.

8. Conclusions

Plant extracts and isolated compounds act across several processes relevant to MASLD/MASH, but their evidence base is uneven and clinical studies have largely measured metabolic indices, liver enzymes, or hepatic fat rather than histological disease modification.

PELNs add vesicular packaging and delivery to this pharmacology, yet direct evidence remains limited to 12 heterogeneous preclinical studies. The central questions are now product identity, causal cargo, oral fate, comparative advantage over source extracts or free compounds, and safety. Resolving these issues, not broader claims of natural targeting or biocompatibility, will determine whether PELNs can progress toward clinical evaluation.