An Exploratory Study on the Pathogenic Role of Faecal Extracellular Vesicles in Metabolic Dysfunction‐Associated Steatotic Liver Disease Progression and in Drug‐Induced Liver Injury

ABSTRACT

The role of extracellular vesicles secreted by the gut microbiota present in faeces (fEVs) is not well known in metabolic dysfunction‐associated steatotic liver disease (MASLD) and idiosyncratic drug‐induced liver injury (DILI). We identify the microbiome profiles of fEVs in these liver diseases, and analyse the effects of fEVs from MASLD, without (F≤2) or with (F≥3) significant liver fibrosis, and DILI patients on inflammation, steatosis and mitochondrial function. DILI patients showed a consistent pattern in fEVs, characterised by a decrease in Paraprevotella and an increase in AAP99, Acinetobacter, Actinobacillus, Aerococcus and Anaeroglobus. A higher presence of 16S rDNA was observed in plasma EVs from MASLD and DILI patients. HepG2 cells treated with DILI and MASLD F≥3 fEVs increased TLR4, TLR5, IL6 and CASP3 expression, and accumulation of lipid droplets. DILI fEVs enhanced the hepatotoxic impact of diclofenac on the response to microbial components (TLR4, TLR5), inflammatory response (IL1B, IL6), accumulation of lipid droplets and mitochondrial dysfunction (OPA1, DNM1L). In conclusion, bacterial EVs enter the bloodstream and could modulate the immune response. DILI and MASLD F≥3 fEVs are drivers of the pro‐inflammatory response and hepatocyte steatosis. DILI fEVs have a distinct bacterial profile that enhances the hepatotoxic potential of diclofenac.

• Fecal extracellular vesicles (fEVs) from DILI patients have a distinct bacterial profile.

• This study provides evidence that bacterial EVs enter the bloodstream.

• DILI and MASLD F≥3 fEVs increase the pro‐inflammatory response and hepatocyte steatosis.

• FEVs from DILI patients enhance the hepatotoxic potential of diclofenac.

1. Introduction

Gram‐positive and Gram‐negative bacteria produce bacterial extracellular vesicles (BEVs) that are enclosed in a lipid bilayer membrane (Toyofuku et al. 2019). The specific content of BEVs may vary depending on bacterial species and the environmental conditions (Bitto et al. 2021). These BEVs are implicated in several biological processes, including intercellular communication, metabolism, horizontal gene transfer, immune response modulation, biofilm formation, nutrient acquisition by enzyme transport and release, quorum sensing, host pathogen interaction and stress response and survival (Ozkocak et al. 2022). BEVs have been studied in the context of various diseases, including cancer, inflammatory bowel disease, Alzheimer and liver disease, among others.

In liver diseases, such as metabolic dysfunction‐associated steatotic liver disease (MASLD, classically known as non‐alcoholic fatty liver disease (NAFLD)), alcoholic liver disease (ALD), idiosyncratic drug‐induced liver injury (DILI) and liver fibrosis, the gut‐liver axis and gut microbiota dysbiosis seem to play a crucial role. In a previous study, we demonstrated that the composition of gut microbiota differs according to the type of liver disease, with specific profiles in DILI and MASLD depending on the liver fibrosis status (Rodriguez‐Diaz et al. 2022).

Indeed, BEVs released by gut microbiota seem to be implicated in the pathogenesis and progression of these liver conditions. A previous study found that gut dysbiosis was associated to an increased intestinal permeability and an increased presence of BEVs in blood (Fizanne et al. 2023). They may interact with the immune system and trigger an inflammatory response (Fizanne et al. 2023; Villard et al. 2021). They have been also implicated in promoting liver fibrosis in NAFLD and ALD (Muñoz‐Hernández et al. 2022). Furthermore, BEVs may participate in several mechanisms involved in NAFLD onset and progression, such as intestinal barrier destabilisation and systematic inflammation (Villard et al. 2021). BEVs isolated from the faeces of patients with non‐alcoholic steatohepatitis (NASH) have been shown to increase intestinal permeability and inflammatory cytokines and chemokines, and activated profibrotic and proinflammatory proteins of hepatic stellate cells (Fizanne et al. 2023). Other studies have also described the beneficial effects of some bacteria and their extracellular vesicles (EVs) in the prevention and modulation of liver fibrosis. For example, a recent study in mice describes how EVs from Akkermansia muciniphila and its EVs improve intestinal permeability, modulate inflammatory responses and prevent liver injury (Keshavarz Azizi Raftar et al. 2021). However, to the best of our knowledge, there are no studies in human hepatocytes, the principal cell type of the liver, examining the potential impact of faecal EVs (fEVs) on steatosis and fibrogenesis, pivotal aspects in the pathogenesis of MASLD. This increased steatosis results in augmented lipotoxicity, a hallmark of NASH progression (Han and Kaufman 2016).

The underlying mechanism of DILI is the result of multiple interactions between drug properties, host and environmental factors (Chen et al. 2015), such as the gut microbiota. The metabolism of drugs in hepatocytes generates reactive oxidative species that can be potentially toxic to the cell, causing cellular stress and potentially leading to apoptosis (Andrade et al. 2019). If this response is inadequate, cell death occurs (Andrade et al. 2019). Indeed, hepatotoxicity can mimic any known liver disease (Andrade et al. 2019), including drug‐induced esteatosis. However, while there seems to be a correlation between faecal microbiota and liver diseases, there is a lack of research on the potential liver‐related effects of fEVs (Fizanne et al. 2023; Villard et al. 2021). Furthermore, to the best of our knowledge, no studies have been conducted with fEVs from DILI patients to demonstrate their involvement in liver processes related to the development of these liver diseases, such as the immune response, liver fibrosis, steatosis and apoptosis, as well as processes involving mitochondria.

The primary objective of this study was to characterise the BEVs derived from the faecal microbiome from patients with different degrees of liver fibrosis within MASLD, and from patients with DILI. This may enable us to identify possible microbial markers that may be associated with the onset and progression of these liver diseases. Secondly, the present study also aims to investigate the in vitro effects of fEVs derived from patients with MASLD and DILI on human hepatocytes in order to identify the molecular mechanisms underlying the pathophysiology of microbiota‐induced liver injury in these liver diseases. This will allow us to gain a deeper understanding of the effects of fEVs on human hepatocytes, which will in turn inform our approach to developing new treatments for these liver diseases.

2. Materials and Methods

2.1. Study Design

In this study, we analysed the gut microbiota and BEVs composition within a cohort study composed of nine patients with DILI, 17 patients with MASLD (seven patients with ‘No Significant Fibrosis’ (F≤2) and 10 patients with ‘Significant Fibrosis’ (F≥3)) and a control group of 10 healthy volunteers. Characteristics of these patients are shown in Table 1. All samples collected from patients were meticulously processed and rapidly frozen (–80°C) upon receipt at the Biobank of the Virgen de la Victoria University Hospital (Andalusian Public Health System Biobank). Other types of liver disorders, such as viral, autoimmune, genetic and alcoholic hepatitis were excluded. All patients included in the study were of Caucasian origin. Prior to participation, all patients gave written informed consent. The study protocol was executed in line with the ethical guidelines established by the World Medical Association Code of Ethics, specifically following the principles of the Declaration of Helsinki. In addition, the study was approved by the Provincial Research Ethics Committee of Málaga, Spain (Approval code: UMA18‐FEDERJA‐194).

TABLE 1.

Anthropometric and biochemical variables of the subjects included in the study.

	Control	MASLD F≤2	MASLD F≥3	DILI
N (men/women)	10 (4/6)	7 (1/6)	10 (3/7)	9 (3/6)
Age (years)	37 ± 8	49 ± 7	57 ± 7 a	47 ± 14
Weight (kg)	68 ± 9	81 ± 17	86 ± 7 a	75 ± 14
BMI (kg/m2)	24 ± 3	29 ± 5 a	30 ± 4 a	26 ± 3
Glucose (mg/dL)	87 ± 9	101 ± 5 a	112 ± 28 a	95 ± 9
Cholesterol (mg/dL)	175 ± 28	196 ± 35	186 ± 32	198 ± 39
Triglycerides (mg/dL)	76 ± 37	137 ± 63 a	149 ± 105 a	162 ± 68 a
Insulin (µU/mL)	5.9 ± 1.7	16.5 ± 6.6	17.6 ± 9.6 a	11.4 ± 9.8
HOMA‐IR	1.3 ± 0.4	4.1 ± 1.6 a	5.4 ± 3.9 a	2.7 ± 2.3
AST (UI/L)	21 ± 5	49 ± 23 a	45 ± 24 a	143 ± 124 a , c
ALT (UI/L)	22 ± 10	82 ± 58 a	81 ± 29 a	320 ± 225 a , b , c
GGT (UI/L)	27 ± 10	174 ± 200 a	142 ± 159 a	238 ± 260 a
ALP (UI/L)	44 ± 9	96 ± 51 a	90 ± 51 a	153 ± 63 a , c
IFG/T2DM (n (%))	0 (0%)	2 (28.6%)	6 a (60%)	3 (30%)
Hypercholesterolemia (n (%))	0 (0%)	3 (42.8%)	2 (20%)	1 (10%)
Hepatic parameters fibrosis
FLI	−24 ± 21	73 ± 26 a	81 ± 14 a	Nc
NAFLD FS	3.05 ± 0.50	−1.51 ± 1.77	−1.03 ± 1.04 a	Nc
FIB4	0.68 ± 0.18	1.78 ± 1.94	1.44 ± 0.66	Nc
APRI	0.26 ± 0.06	0.87 ± 0.77 a	0.63 ± 0.28 a	Nc

Note: Results are expressed as mean ± standard deviation for continuous variables or percentages for categorical variables.

Abbreviations: ALP, alkaline phosphatase; ALT, alanine aminotransferase; APRI, AST‐to‐platelet ratio index; AST, aspartate aminotransferase; BMI, body mass index; FIB4, fibrosis‐4; FLI, fatty liver index; GGT, gamma‐glutamyltransferase; HOMA‐IR, homeostasis model assessment of insulin resistance; IFG, impaired fasting glucose; NAFLD FS, NAFLD fibrosis score; Nc, not calculated; T2DM, type 2 diabetes mellitus.

^a p < 0.05: significant differences with regard to healthy control group.

^b p < 0.05: significant differences with regard to MASLD F≤2 group.

^c p < 0.05: significant differences with regard to MASLD F≥3 group.

2.2. Cohort of Patients With DILI, MASLD and Healthy Controls

Patients with DILI were enrolled from the prospective Spanish DILI Registry. In‐depth details of this registry have been published previously (Andrade et al. 2005). In summary, suspected DILI cases were evaluated based on (i) the alignment between the timing of drug intake and the onset of signs, symptoms or abnormal blood test results; (ii) comprehensive biochemical, histological and imaging data to rule out alternative liver diseases and (iii) the outcome of the liver injury. The RUCAM (Roussel Uclaf Causality Assessment Method) scale was then applied, and cases were reviewed by three DILI experts before being included in the registry. The biochemical criteria for DILI were established by an international expert group (Aithal et al. 2011). Fifty percent of the included DILI patients had jaundice, and 80% required hospitalisation. The type of liver injury (R = ALT/ULN/ALP/ULN) was hepatocellular (R≥5) in 70%, cholestatic (R≤2) in 20% and mixed (R>2 and R<5) in 10% of cases. DILI severity was mild in 50% of patients and 50% of patients showed a moderate injury. Culprit agents responsible for DILI were dietary supplements (n = 1), nanodrol (n = 1), clenbuterol/amoxicillin‐clavulanate (n = 1), terbinafine (n = 1), isoniazide (n = 1), levofloxacin (n = 1), amoxicillin‐clavulanate (n = 1), ampicillin (n = 1), clindamycin (n = 1) and trabectedin (n = 1).

Patients meeting both invasive and noninvasive criteria for MASLD diagnosis were prospectively recruited from the Digestive Diseases Unit at Virgen de la Victoria University Hospital (Malaga, Spain). Inclusion criteria required a histological diagnosis of MASH (presence of ≥5% steatosis in the liver, hepatocellular damage, hepatocyte ballooning or presence of fibrosis in a liver biopsy) or a noninvasive MASLD diagnosis (for those without a liver biopsy, diagnosis was confirmed by excluding other liver damage causes, performing noninvasive tests and identifying steatosis via abdominal ultrasound), plus at least 1 of 5: (1) BMI ≥ 25 kg/m², waist circumference >94 cm in men or >80 cm in women; (2) Fasting serum glucose ≥100 mg/dL or 2‐h post‐load glucose level ≥140 mg/dL or HbA1c ≥5.7% or on specific drug treatment; (3) Blood pressure ≥130/85 mmHg or specific drug treatment; (4) Plasma triglycerides ≥150 mg/dL or specific drug treatment; (5) Plasma HDL cholesterol <40 mg/dL for men and <50 mg/dL for women or specific drug treatment (Rinella et al. 2023). Exclusion criteria included (Chalasani et al. 2018): alcohol intake >20 g/day for men and >10 g/day for women, secondary causes of MASLD or other chronic liver diseases, use of drugs that could potentially induce MASLD (such as steroids, amiodarone, methotrexate, tamoxifen and sodium valproate), type 1 diabetes, severe psychiatric disorders and antibiotic use in the previous 3 months. MASLD patients were categorised into two groups based on liver fibrosis severity: no significant fibrosis (F≤2) (n = 7) and significant fibrosis (F≥3) (n = 10), as measured by FibroScan.

Healthy controls were recruited from workers at the Virgen de la Victoria University Hospital and the University of Malaga. Inclusion criteria required the absence of any history of liver disease. Exclusion criteria included: a previous history of hepatotoxicity or any other chronic liver disease, abnormal liver profile at the time of inclusion, body mass index >25 kg/m², diabetes mellitus, dyslipidemia, metabolic syndrome or MASLD, and the administration of antibiotics in the previous 3 months.

2.3. Non‐Invasive Evaluation of Steatosis, MASH and Fibrosis

The fatty liver index (FLI) was used to assess hepatic steatosis (Bedogni et al. 2006). To evaluate liver fibrosis, the NAFLD fibrosis score (NAFLD FS) (Angulo et al. 2007), the fibrosis‐4 index (FIB4) (Sterling et al. 2006) and the AST‐to‐platelet ratio index (APRI) (Bourliere et al. 2006) were used. Patients were also examined in a fasting state by transient elastography (TE) using the ECHOSENS FibroScan 402 (Echosens, Paris, France). This device employed either an M or XL probe, targeting the right lobe of the liver. The liver stiffness measurement (LSM) was established based on the median value obtained from 10 successful measurements. To ensure the reliability of the results, a criterion was set: LSM measurements were considered reliable only if IQR/med was less than 30% and success rate was greater than 60%. To gather comprehensive data, all patients underwent 10 successful acquisitions. MASLD patients were categorised into two distinct groups according to the extent of liver fibrosis. The groups were classified as ‘No Significant Fibrosis’ (F≤2) and ‘Advanced Fibrosis’ (F≥3) based on the TE results. Specifically, for the M probe, liver stiffness measurements of ≥7, ≥8.7 and ≥10.3 kPa were indicative of fibrosis stages ≥F2, ≥F3 and F4, respectively. Similarly, for the XL probe, the corresponding values were ≥6.2, ≥7.2 and ≥7.9 kPa for fibrosis stages ≥F2, ≥F3 and F4, respectively (Kaswala et al. 2016).

2.4. Biochemical Analysis

Faeces and blood samples were collected and immediately stored at –80°C until analysis. Blood samples were collected in a fasting state. The serum was separated and immediately frozen at an ultra‐low temperature of –80°C. In DILI patients, blood samples were collected between Days 1–11 after DILI recognition. Serum biochemical variables were measured in duplicate in a modular analytics E170 analyser (Roche Diagnostics GmbH, Mannheim, Germany). HOMA‐IR was calculated with the following equation: HOMA‐IR = fasting insulin (μIU/mL) × fasting glucose (mmol/L)/22.5.

2.5. Clasification of Patients

Patients were categorised into two distinct groups according to the presence of an alteration of the glucose metabolism: presence of Hb1Ac ≥5.7%, impaired fasting glucose or type 2 diabetes mellitus (T2DM) (fasting glucose >100 mg/dL) or treatment for T2DM (American Diabetes Association Professional Practice Committee 2025). The presence of hypercholesterolemia was defined by the treatment with cholesterol lowering therapies (i.e., statins and/or ezetimibe).

2.6. Isolation of EVs From Human Faeces

Isolation and purification of fEVs was performed as previously described (Caballano‐Infantes et al. 2023). Briefly, a total of 10 g of faeces was inoculated in 40 mL of sterile phosphate‐buffered saline (PBS) and homogenised. The fEVs were isolated by centrifugation as previously described, with some minor modifications (Rodríguez‐Díaz et al. 2023). A first centrifugation of the homogenate was performed (40 min, 4000 × g, 4°C), and the supernatant was recovered and filtered using sterilised Nalgene Rapid‐Flow vacuum filtration units of 0.2 µm in cold ice (ThermoFisher Scientific, Waltham, MA, USA). The filtrate was transferred to 10 mL polycarbonate open top thick wall tubes and ultracentrifuged at 100,000 × g for 3 h at 4°C, with a fixed angle rotor Type 70.1 Ti in a Beckman Optima XL‐100K ultracentrifuge (Beckman Coulter Life Sciences, Lakeview Pkwy S Drive, Indianapolis, IN, USA). Pellets were resuspended in 200 µL of PBS and the fEVs were purified using qEVoriginal size exclusion columns (SEC) of 70 nm (Izon Science Europe Ltd., Oxford, UK) (Dauros Singorenko et al. 2017; Mehanny et al. 2020), following the manufacturer's recommendations. Fractions 6–9 were collected, mixed, concentrated with Vivaspin 20 100 kDa (Sartorius Stedim Biotech GmbH, Göttingen, Germany) centrifugal concentrators, aliquoted and frozen at −80°C until used for metagenomics and in vitro culture analysis. The protein concentration of fEVs was determined using the bicinchoninic acid (BCA) assay (Thermo Fisher Scientific, Waltham, MA, USA).

2.7. Electron Microscopy of fEVs

Isolated fEVs and plasma EVs were fixed in 2% paraformaldehyde–0.1 M PBS for 30 min. The glow discharge technique (60 s, 7.2 V, using a Bal‐Tec MED 020 Coating System) was applied over carbon‐coated copper grids, and immediately, these grids were placed on top of sample drops for 15 min. Then, the grids with adherent MVs were washed in a 0.1 M PBS drop and additional fixation in 1% glutaraldehyde was performed for 5 min. After washing properly in distilled water, the grids were contrasted with 1% uranyl acetate and embedded in methylcellulose. Excess fluid was removed and allowed to dry before examination with a transmission electron microscope FEI Tecnai G2 Spirit (ThermoFisher Scientific, Oregon, USA). All images were obtained using a Morada digital camera (Olympus Soft Image Solutions GmbH, Münster, Germany) (Caballano‐Infantes et al. 2023).

2.8. Nanoparticle Tracking Analysis (NTA)

The fEVs size was assessed using the NanoSight NS300 system (Malvern Panalytical, Malvern, UK). Particles were automatically tracked and sized‐based on Brownian motion and the diffusion coefficient. The EVs were resuspended and diluted with 0.22 µm‐filtered PBS at a concentration range of 10⁹ particles/mL, and 1 mL was used for NanoSight analysis. Five replicates of 30 s videos were captured to analyse concentration and size distribution of EVs at threshold detection of 5. Data analysis was performed using NanoSight analysis software (Rodríguez‐Díaz et al. 2023).

2.9. Western Blot of EVs

EVs from plasma or from faeces of a DILI patient were lysed with RIPA buffer (ThermoFisher Scientific, Oregon, USA) and supplemented with a protease inhibitor cocktail (Merck KGaA, Darmstadt, Germany). The protein lysate was incubated with Laemmli Buffer 5× (Bio‐Rad Laboratories, Inc., Hercules, CA, USA) and supplemented with 2‐mercaptoethanol (5%) at 95°C for 5 min. Precision Plus Protein Standards (Dual Colour) (Bio‐Rad Laboratories, Inc., Hercules, CA, USA) was used as the standard for molecular weight. The samples were subjected to 4%–20% SDS‐PAGE (NB12‐420) (NuSep, Inc., Germantown, MD, USA) and transferred onto polyvinylidene fluoride membranes (Trans‐Blot Turbo Midi 0.2 µm PVDF Transfer Packs) (Bio‐Rad Laboratories, Inc., Hercules, CA, USA) at 13 V and 1.1 A for 20 min. The membranes were subsequently blocked in PBS–bovine serum albumin (BSA) 5% for 1 h at room temperature. The membranes were then incubated for 24 h at 4°C with a mouse monoclonal anti‐bacterial peptidoglycan antibody, clone 3F6B3 (MAB995, Merck KGaA, Darmstadt, Germany) for fEVs, or for 48 h at 4°C with a recombinant monoclonal rabbit antibodyanti‐CD9 (ab92726, Abcam, Cambridge, UK) for plasma EVs. Peptidoglycan is a marker for BEVs from Gram‐positive and Gram‐negative bacteria (Toyofuku et al. 2023). The membranes were washed three times with PBS + 0.05% Tween‐20 washing buffer and incubated with a horseradish‐peroxidase‐conjugated secondary antibody (VeriBlot for IP Detection Reagent (HRP), ab131366) (Abcam, Cambridge, UK) for 3 h at room temperature. Finally, after another three washes, the membranes were revealed with Clarity Western ECL substrate (Bio‐Rad Laboratories, Inc., Hercules, CA, USA). The proteins were visualised in an ImageQuant LAS 4000 (GE Healthcare, Buckinghamshire, UK).

2.10. DNA Extraction and 16S rDNA Metagenomic Sequencing Library Preparation

Total DNA extraction was performed using a QIAamp Power Faecal Pro DNA (QIAGEN Science, Hilden, Germany) for faeces and DNeasy Blood and Tissue Kit (QIAGEN Science, Hilden, Germany) for fEVs following the manufacturer´s recommendations. The DNA obtained was eluted into DNase/RNase‐free water, and its concentration and purity were evaluated by absorbance measurement (NanoDrop ND‐1000 spectrophotometer, Isogen Lifescience B.V., Utrecht, The Netherlands). 16S rDNA metagenomic sequencing library was prepared as previously described (Rodriguez‐Diaz et al. 2022). Briefly, amplification of the 16S rRNA targeting the V3‐V4 hypervariable region was performed using the primers 16S‐V3–314 forward (5′TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG3′) and V4–805 reverse (5′GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTAC HVGGGTATCTAATCC3′) with added Illumina adapter overhang nucleotide sequences. PCR conditions used were 3 min at 95°C, followed by 25 cycles of 30 s at 95°C, 30 s at 55°C and 30 s at 72°C, with a final extension at 72°C for 5 min. Each reaction mixture (25 µL) contained 5 ng of genomic DNA, 0.5 µL of amplicon PCR forward primer (0.2 µM), 0.5 µL of amplicon PCR reverse primer (0.2 µM) and 12.5 µL of 2 × KAPA HiFi HotStart ReadyMix (Roche Molecular Systems, Inc., Pleasanton, CA, USA). According to the manufacturer's protocol, PCR products were purified with Agencourt AMPure XP beads (Beckman Coulter Genomics, Danvers, MA, USA) to remove excess primers and primer dimers. In a second index PCR, 5 µL of each amplicon was used as a template. Dual indices and Illumina sequencing adapters for each amplicon were attached in the second PCR using a Nextera XT Index Kit v2 (Illumina Inc. San Diego, CA, USA). In this case, amplification was carried out under the following conditions: 3 min at 95°C, followed by eight cycles of 30 s at 95°C, 30 s at 55°C and 30 s at 72°C, with a final extension at 72°C for 5 min. Constructed 16S rDNA metagenomic libraries were purified with Agencourt AMPure XP beads. Quantification of the library, quality control and average size distribution were determined with an Agilent Tapestation 4200 (Agilent Technologies, Inc., Santa Clara, CA, USA). Libraries were normalised and pooled to 40 nM based on quantified values. Pooled samples were denatured and diluted to a final concentration of 6 pM with a 30% PhiX (Illumina Inc. San Diego, CA, USA) control. Amplicons were subjected to sequencing using a MiSeq Reagent Kit V3 in the Illumina MiSeq System (Illumina Inc. San Diego, CA, USA).

2.11. Bioinformatic, Ordination and Statistical Analysis

Sequence reads were analysed using Mothur v1.44.3 and VSearch, facilitating alignment, clustering and chimera detection (Rognes et al. 2016; Schloss et al. 2009). Sequences were clustered into operational taxonomic units (OTUs) at 97% identity. The SILVA 138 database of full‐length 16S rDNA gene sequences was used for alignments of unique sequences and taxonomical assignations (Quast et al. 2013). All statistical analyses were performed at the genus level. Good's coverage index and ecological indicators, including the α‐diversity (inverse Simpson's index and Shannon index), bacterial richness (Chao1 index) and evenness (Simpson index‐based measure) were calculated with MOTHUR v1.43. Differences between groups were assessed using Kruskal–Wallis multiple testing with Benjamini–Hochberg FDR corrections (for BEVs samples comparison) (q threshold  A  0.05), and according Wilcoxon matched‐pairs signed rank test (for faeces and BEVs samples). Barplots were built using PRISM 9.0. The bacterial community was visualised using Bray–Curtis dissimilarity matrix using vegan and vegan3 packages in RStudio. Post hoc pairwise differences between groups were assessed with Deseq2 package in R using Benjamini–Hochberg FDR correction (Fernandes et al. 2014). This study has been registered in the Bioproject (National Center for Biotechnology Information (NCBI)) at https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA1349782, reference number [PRJNA1349782].

2.12. Isolation of Circulating EVs From Human Plasma

Blood samples were collected in EDTA K2 tubes. All steps are carried out under sterile conditions. The plasma was separated and immediately frozen at –80°C until analysis. Plasma was thawed and centrifuged at 21,000 g for 30 min. Supernatant was collected and ultracentrifuged at 100,000 g for 90 min at 4°C with a fixed angle rotor Type 70.1 Ti in a Beckman Optima XL‐100K ultracentrifuge (Beckman Coulter Life Sciences, Lakeview Pkwy S Drive, Indianapolis, IN, USA). The pellet was resuspended in 200 µL of PBS and purified using qEVoriginal size exclusion columns (SEC) of 70 nm (Izon Science Europe Ltd., Oxford, UK), following the manufacturer's recommendations. Fractions 7–10 were collected, mixed, filtered with a 0.22 µm filter and concentrated with Vivaspin 20 100 kDa (Sartorius Stedim Biotech GmbH, Göttingen, Germany) centrifugal concentrators. Additionally, we used a sample of distilled and filtered water as blank sample for the western blot analysis, which underwent the same procedures for isolating and purifying plasma EVs.

2.13. DNA Extraction and 16S rDNA Detection in Circulating EVs

Total DNA extraction was performed from circulating EVs using a DNeasy Blood and Tissue Kit (QIAGEN Science, Hilden, Germany) following the manufacturer´s recommendations. The DNA obtained was eluted into DNase/RNase‐free water, and its concentration and purity were evaluated by absorbance measurement (NanoDrop ND‐1000 spectrophotometer, Isogen Lifescience B.V., Utrecht, The Netherlands). The presence of 16S rDNA in samples was determined by PCR using the following primers, RIBOFOR 5′‐GTAGCGGTGAAATGCGTAGA‐3′ and RIBOREV 5′‐CTTTCGTACCTCAGCGTCAG‐3′, generating an approximately 85 bp 16S rDNA product. Migration of amplicons was visualised using a 2.5% agarose gel.

2.14. Cell Culture Conditions

The HepG2 cell line was obtained from human hepatocarcinoma (ATCCHB‐8065, 85011430–1VL, Merck LifeScience S.L.U., Darmstadt, Germany) and cultured in low‐glucose Dulbecco's modified Eagle's medium (DMEM) (Gibco Carlsbad, CA, USA) supplemented with 10% inactivated foetal bovine serum (FBS) (Gibco Carlsbad, CA, USA), 1% penicillin/streptomicine (Merck LifeScience S.L.U., Darmstadt, Germany), 2 mM glutamine (Gibco, Carlsbad, CA, USA) and 1% MEM non‐essential amino acids solution (Merck LifeScience S.L.U., Darmstadt, Germany). Cells were cultured in a humidified atmosphere at 37°C, 21% O₂ and 5% CO₂. Cells were sub‐cultured by dispersion using a trypsin/EDTA solution (Gibco, Carlsbad, CA, USA). Once HepG2 cells had reached 80%–90% confluence, they were either subjected to a 24‐h pre‐treatment to induce steatosis and hepatotoxicity, or maintained without any pre‐treatment. To induce steatosis, HepG2 cells were pre‐treated for 24 h as follows: HepG2 cells + vehicle (0.5% bovine serum albumin—fatty acid free (BSA‐FAF)) and HepG2 cells + palmitic:oleic acids (PA:OA) (0.5% BSA‐FAF + palmitic acid 250 µM and oleic acid 500 µM) (Rodriguez‐Pacheco et al. 2014; Longhitano et al. 2024). In addition, we used diclofenac (DIC) 500 µM as a model drug with a known risk hepatotoxicity in vivo (Knöspel et al. 2016). Additionally, we utilised control samples, designated as ‘Control’, which were obtained from distilled and filtered water and underwent the same procedures for isolating and purifying fEVs as the faecal samples. After these pre‐treatments, HepG2 cells were exposed to fEVs (10 µg of protein/mL) (Patten et al. 2017) for 24 h, after which different experiments were performed, as described below. This fEVs concentration is in line with other studies conducted with fEVs (Zhu et al. 2024; Diaz‐Garrido et al. 2022), and with a pilot study with fEVs from DILI patients in HepG2 cells (the disease that is the main focus of this study) (Figure S1).

2.15. Oil Red O Staining

To determine the accumulation of lipid droplets into HepG2 with the different treatments, HepG2 cells were washed with PBS and then incubated with oil red‐O staining solution for 30 min. Images were taken using an DM IL LED inverted microscope (Leica Microsystems, Wetzlar, Germany) at a magnification of 40×, and stained lipid droplets were quantified as the absorbance at 520 nm after solubilisation with 96% ethanol (Rodríguez‐Pacheco et al. 2017).

2.16. Soluble Collagen Analysis

To determine the secretion of soluble collagen by HepG2 cells, the culture medium, obtained following a 24‐h period of exposure to the various treatments, was isolated and centrifuged at 10,000 × g for 15 min at 4°C to pellet any debris. The resulting clarified supernatant was then used for the analysis for soluble collagen with the Soluble Collagen Assay Kit (ab241015) (Abcam, Cambridge, UK).

2.17. Mitochondrial Membrane Potential

For immunofluorescence analysis of mitochondrial membrane potential in the HepG2 cell line, cells were seeded on an 18‐well μ‐slide (Cat.No: 81816, IBIDI GmbH, Gräfelfing, Germany). When cells were semi‐confluent (80%–90%), they were treated with the same conditions as previously described (Section 2.14). At 48 h, media was removed, cells were washed with pre‐warned non‐supplemented DMEM, and stained with DAPI (1 µg/mL, ref. 62247, Thermo Fisher Scientific Inc. Waltham, MA, USA), MitoTracker Red CMXRos (350 nM final concentrations) (ref. M7512, Thermo Fisher Scientific Inc. Waltham, MA, USA) and CellMask Deep Red Plasma Membrane Stain (1:5000 dilution) (ref. C10046, Thermo Fisher Scientific Inc. Waltham, MA, USA) for 15 min at 37°C. Cells were washed twice with PBS 1× and fixed with 10% formalin (15 min at 37°C). Cells were washed twice with PBS 1×, and covered with ProLong Diamond Antifade Mountant (ref. P36961, Thermo Fisher Scientific Inc. Waltham, MA, USA). Cells were visualised on a SP5 confocal microscope (Leica Microsystems, Wetzlar, Germany). A total of three experiments were run, and five pictures per condition were analysed. Fluorescence was measured as integrated density, and values from mitotracker were normalised to fluorescence from DAPI.

2.18. RNA Isolation and Quantitative Real‐Time PCR

Total RNA from HepG2 cells was extracted with an RNeasy mini kit (QIAGEN Science, Hilden, Germany). First strand cDNA was synthesised by retrotranscription using M‐MLV retrotranscriptase (Promega, Madison, WI, USA). Gene expression levels were analysed in triplicate by real‐time PCR using a SensiFAST SYBR Green Kit (Bioline, London, UK) in a 7500 Fast (Thermo Fisher Scientific Inc. Waltham, MA, USA). Primers for the PCR reaction were designed using the primer design tool from IDT (Integrated DNA Technologies, Coralville, IA, USA): collagen type I alpha 1 chain (COL1A1) (forward: GCTATGATGAGAAATCAACCG, reverse: TCATCTCCATTCTTTCCAGG), fatty acid synthase (FASN) (forward: CAATACAGATGGCTTCAAGG, reverse: GATGTATTCAAATGACTCAGGG), glutathione peroxidase 1 (GPX1) (forward: CTACTTATCGAGAATGTGGC, reverse: CAGAATCTCTTCGTTCTTGG), carnitine palmitoyltransferase 1A (CPT1A) (forward: AAGTTTTATCTGAGCCTTGG, reverse: AGAACTTGGAAGAAATGTGG), interleukin 1 beta (IL1B) (forward: GCAACAAGTGGTGTTCTC, reverse: CAGATTCTTTTCCTTGAGGC), interleukin 6 (IL6) (forward: GCAGAAAAAGGCAAAGAATC, reverse: CTACATTTGCCGAAGAGC), interleukin 10 (IL10) (forward: GCCTTTAATAAGCTCCAAGAG, reverse: ATCTTCATTGTCATGTAGGC), OPA1 mitochondrial dynamin like GTPase (OPA1) (forward: CAGAAGACCTTGTAAAGTTAGC, reverse: TAACCAATTTGTGACCTGAG), dynamin 1 like (DNM1L) (forward: GAGAGGAATGCTGAAAACTTC. Reverse: GAGTCGTTCAATAACCTCAC), TIMP metallopeptidase inhibitor 1 (TIMP1) (forward: CACCTTATACCAGCGTTATG, reverse: TTTCCAGCAATGAGAAACTC), matrix metallopeptidase 9 (MMP9) (forward: AAGGATGGGAAGTACTGG, reverse: GCCCAGAGAAGAAGAAAAG), transforming growth factor beta 1 (TGFB1) (forward: TGTACCAGAAATACAGCAAC, reverse: CAAAAGAAACCACTCTGGC), toll like receptor 4 (TLR4) (forward: GATTTATCCAGGTGTGAAATCC, reverse: TATTAAGGTAGAGAGGTGGC), toll like receptor 5 (TLR5) (forward: ATCTTTCACATGGGTTTGTC, reverse: TTCCCCCAGAAGGTTATATG), glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH) (forward: TCGGAGTCAACGGATTG, reverse: CAACAATATCCACTTTACCAGAG), diacylglycerol O‐acyltransferase 1 (DGAT1) (forward: ATCTTCTTCTACTGGCTCTTC, reverse: CAGAAGTAGGTGACAGACTC), sterol regulatory element binding transcription factor 1 (SREBF1) (forward: TGCATTTTCTGACACGCTTC, reverse: CCAAGCTGTACAGGCTCTCC), sterol regulatory element binding transcription factor 2 (SREBF2) (forward: CAGCAGGTCAATCATAAACTG, reverse: GGACATTCTGATTAAAGTCCTC), CD36 (forward: CATGTCTTGCTGTTGATTTG, reverse: AGCCCATTTTTCTTGTTCAG), patatin like phospholipase domain containing 3 (PNPLA3) (forward: CAGCACTGAGTGAAGAAATG, reverse: GACATTATCCTAATGGGTAGC), caspase 3 (CASP3) (forward: AAAGCACTGGAATGACATC, reverse: CGCATCAATTCCACAATTTC), perilipin 5 (PLIN5) (forward: GATCAGAGGAGACAGCAG, reverse: GTGGTCTATCAGCTCCAG). The relative expression levels of all genes were normalised to the expression of the house‐keeping gene (GAPDH) and data were analysed with the ∆∆CT method.

2.19. Statistical Analysis

The statistical analysis for the metagenomic study has been described above. For the other statistical analyses performed in the study, data were analysed with GraphPad Software (Prism 9.3.1) (GraphPad Software, San Diego, CA, USA). Differences between groups were compared using Kruskal–Wallis tests followed by post hoc analyses using the Dunn's test. Chi‐square tests were performed to compare proportions between groups. Values were considered statistically significant when p < 0.05.

3. Results

3.1. Characterisation of EVs

We characterised the fEVs after isolating using ultracentrifugation and qEV IZON columns. First, we evaluated the purity of the fEVs, verifying that most of the soluble proteins had been removed (Figure 1A). Additionally, Figure 1B shows an image of the fEVs before and after purification using size exclusion columns. The absence of background noise in the sample after passing through these columns is evident. Analysis of the size distribution by NTA revealed a mean value of 177.1 ± 5.6 nm for fEVs (Figure 1C). This size distribution is consistent with BEVs ranging from 20 to 400 nm (Díaz‐Garrido et al. 2021; Brown et al. 2015). The Figure 1D shows a Western blot for peptidoglycan, a marker for BEVs from Gram‐positive and Gram‐negative bacteria, which is a variable‐sized polymer that forms the bacterial cell wall (Miyakawa et al. 2024). Electron microscopy revealed a variety of spherical shapes in samples from each patient group, consistent with their different bacterial origins (Figure 1E).

FIGURE 1.

Characterisation of faecal extracellular vesicles (fEVS) after ultracentrifugation and size exclusion chromatography (SEC). (A) Protein concentration in the different fractions obtained after passing the filtered and concentrated faecal supernatant by ultracentrifugation through SEC (image from a DILI patient). (B) Transmission Electron Microscopy (TEM) images of fEVs before and after purification using SEC (image from a DILI patient). (C) Nanoparticle Tracking Analysis (NTA) (image from a DILI patient). (D) Western blot of peptidoglycan in a sample of fEVs from a DILI patient and in a blank sample. Std.: Molecular weight standard. (E) TEM images of fEVs from a patient of each group.

3.2. Comparative Analysis of Microbiome Profiles in Faecal Samples and BEVs From Patients With MASLD and DILI

The 16S amplicon sequencing yielded 9000 identifications per sample. Bacterial populations with a median value of zero across the entire group of patients were dismissed. The mean alpha‐diversity (inverse Simpson index), evenness (derived from the Simpson index) and bacterial richness (Chao1 richness index) exhibited significant differences when comparing faecal samples and BEVs across the different patient groups. Notably, the mean alpha‐diversity and evenness (Figure 2) showed significant differences between faeces and BEVs in MASLD F≤2 patients (p value = 0.015 and p value = 0.031, respectively). Additionally, significant variations in bacterial richness (Figure 2) were observed between faeces and BEVs in patients with DILI (p value = 0.039) and MASLD F≥3 patients (p value = 0.002).

FIGURE 2.

Bacterial diversity (Inverse Simpson diversity index), bacterial evenness (Simpson evenness index) and bacterial richness (Chao1 richness index), expressed as the mean value with standard deviation, for control (n = 10), MASLD F≤2 (n = 7), MASLD F≥3 (n = 10) and DILI patients (n = 9). Statistical differences are calculated according Wilcoxon matched‐pairs signed rank test (for faeces and BEVs samples). Barplots were built using Prism 9. M F≤2: MASLD F≤2 patients; M F≥3: MASLD F≥3 patients.

When comparing the BEVs among the four groups (control, MASLD F≤2, MASLD F≥3 and DILI) (Figure 3A), no significant differences in bacterial alpha‐diversity were detected (Simpson and Shanon indices). There was significant differences in bacterial evenness between the BEVs from control and MASLD F≤2 groups (p value = 0.033). However, bacterial richness showed a significant difference (p value = 0.007), with significant differences between the control group and the other groups (MASLD F≤2: p value = 0.017; MASLD F≥3: p value = 0.009 and DILI: p value = 0.005). The exclusive comparison of faecal samples between these groups was not performed in this study, as this has been previously published (Rodriguez‐Diaz et al. 2022). The Bray–Curtis dissimilarity matrix is shown in Figure 3B.

FIGURE 3.

(A) Bacterial diversity (Inverse Simpson diversity index), bacterial evenness (Simpson evenness index), bacterial richness (Chao1 richness index) and Shannon index, expressed as the mean value with standard deviation, for BEVs samples comparison. Control (n = 10), MASLD F≤2 (n = 7), MASLD F≥3 (n = 10) and DILI patients (n = 9). The results are presented in a unified graph and in separate graphs against the control group to clarify the differences. Statistical differences are calculated according to Kruskal–Wallis multiple testing with Benjamini–Hochberg FDR corrections Wilcoxon. Barplots were built using Prism 9. (B) Sample distribution in function of the sample group using three‐dimensional dynamic ordination. Beta‐diversity analysis using a Bray–Curtis‐based NMDS model of beta‐diversity. M F≤2: MASLD F≤2 patients; M F≥3: MASLD F≥3 patients.

The identity and relative abundance of the microbial populations in faeces and BEVs across the four patient groups were initially analysed in order to facilitate a global comparison. Overall, the five most abundant taxa, considering both faecal and BEVs samples from all four groups, included Bacteroides, Prevotella, Ruminococcus, Faecalibacterium and Roseburia (Figure 4A). All groups were statistically compared using DESeq2 analysis with Benjamini–Hochberg FDR corrections. A total of 33 significant bacterial differences were identified across conditions, as detailed in Table 2. In the control group, Pseudomonas was the only bacterial genus that showed a significant difference between BEVs and faecal samples, with a significant increase observed in BEVs compared to faeces (p value = 0.047). In the DILI group, the key findings included increases in faeces and decreases in BEVs for 4 and 7 bacterial taxa, respectively (Table 2). An increase in faeces was detected for Phascolarctobacterium (p value = 0.004), Paraprevotella (p value = 0.007), Akkermansia (p value = 0.007) and Gemminger (p value < 0.001). Conversely, a significant decrease in faeces and an increase in BEVs were noted for Abiotrophia (p value = 0.009), Adlercreutzia (p value < 0.001) and Acinetobacter (p value = 0.007), among others.

FIGURE 4.

(A) Faecal microbiota and BEVs composition for control (n = 10), MASLD F≤2 (n = 7), MASLD F≥3 (n = 10) and DILI patients (n = 9). Composition at the genus level shown in a bar chart detailing the mean cumulative abundance (%) of the 16 core genera common to the four groups of patients. Only genera with relative abundance >1% were plotted. (B) Intra‐group analysis of BEVs‐associated microbiota in faecal samples for control, MASLD F≤2 and MASLD F≥3 and DILI groups. The bar charts display the dominant proportions of BEVs within each individual study group, focusing on the mean cumulative relative population abundance (%) at genus level. The proportions were calculated independently for each group, showcasing the distinct microbial profiles within each condition. (C) BEVs composition in faecal samples for control, MASLD F≤2, MASLD F≥3 and DILI patients. Composition at the genus level shown in a bar chart detailing the mean cumulative abundance (%) of the 16 core genera common to the four groups of patients. Only genera with relative abundance >1% were plotted. M F≤2: MASLD F≤2 patients; M F≥3: MASLD F≥3 patients. (D) Correlations between the 16 core genera common to the four groups of patients in faeces and BEVs with host clinical variables. The correlation analysis was conducted exclusively on genera exhibiting relative abundance greater than 1%. *p < 0.05.

TABLE 2.

Differential abundance analysis of microbiome composition in faeces versus BEVs of control, MASLD F≤2, MASLD F≥3 and DILI patients using DESeq2 analysis and Benjamini–Hochnberg FDR corrections.

Group	Comparison	Bacteria	p value
Control	Decrease in faeces and increase in BEVs	Pseudomonas	0.047
DILI	Increase in faeces and decrease in BEVs	Phascolarctobacterium	0.004
		Paraprevotella	0.007
		Akkermansia	0.007
		Gemminger	<0.001
	Decrease in faeces and increase in BEVs	Abiotrophia	0.009
		AAP99*	<0.001
		Acinetobacter	0.007
		Actinomyces	0.044
		Aerococcus	0.007
		Actinobacillus	0.044
		Adlercreutzia	<0.001
MASLD F≤2	Increase in faeces and decrease in BEVs	Phascolarctobacterium	0.007
		Dorea	0.008
MASLD F≥3	Increase in faeces and decrease in BEVs	Blautia	<0.001
		Dorea	<0.001
		Erysipelothrix	0.017
		Escherichia	0.001
		Eubacterium	0.002
		Lactobacillus	0.003
		Pseudobutyrivibrio	0.006
		Roseburia	0.001
		Shuttleworthia	0.001
		Streptococcus	<0.001
		Slackia	0.026
		Methanobrevibacter	0.04
		Veillonella	0.049
		Bifidobacterium	<0.001
	Decrease in faeces and increase in BEVs	Alistipes	<0.001
		Anaerotroncus	0.026
		Barnesiella	<0.001
		Butyricimonas	0.008
		Clostridia_UCG_014	<0.001

*beta proteobacterium AAP99.

In the context of MASLD, distinct microbial patterns were identified. For MASLD F≤2, there were significant increases in faeces and decreases in BEVs for only two genera: Phascolarctobacterium (p value = 0.007) and Dorea (p value = 0.008) (Table 2). In MASLD F≥3, numerous taxa exhibited significant changes, with increases in faeces and decreases in BEVs for Blautia (p value < 0.001), Dorea (p value < 0.001) and Bifidobacterium (p value < 0.001), among others (Table 2). Additionally, decreases in faeces and increases in BEVs were observed for five taxa, including Alistipes (p value < 0.001), Anaerotroncus (p value = 0.026), Barnesiella (p value < 0.001), Butyricimonas (p value < 0.001) and Clostridia_UCG_014 (p value < 0.001).

3.3. Dominant Proportions of BEVs in DILI and MASLD Patients

Figure 4B illustrates the dominant proportions of BEVs in each study group, focusing specifically on the mean cumulative relative population abundance within the different categories of DILI, MASLD F≥3, MASLD F≤2, and controls. It should be noted that the graphs do not provide comparative data between groups, but rather show the relative abundance of BEVs exclusively for each individual group. The proportions shown reflect the distinct microbial profiles of each category, while comparisons between the groups will be addressed in subsequent analyses. This approach allows for a clearer understanding of the dominant BEVs populations within each condition before exploring potential variations between them. The microbial composition of BEVs varies significantly across the conditions studied, with each group exhibiting distinct microbial signatures. While genera such as Prevotella, Faecalibacterium and Bacteroides are present in all groups, their relative abundance varies. This variation in bacterial proportions suggests that different pathological conditions, such as DILI and MASLD, may lead to specific changes in the BEVs‐associated microbiota.

The bacterial community in the control BEVs group exhibited a distinct profile, with Bacteroides and Prevotella being the most dominant genera, together comprising a significant proportion of the microbial population. Other significant contributors include Phascolarctobacterium and Barnesiella, reflecting a more stable and balanced microbial ecosystem (Figure 4B).

The microbiota associated with BEVs in patients with DILI displays a distinctive profile that was not observed in the other groups studied. Specifically, we found a marked predominance of bacterial genera such as Abiotrophia, AAP99, Oscillospiraceae_UCG_002 and Acidaminococcus in DILI patients. This specific microbial composition, together with the presence of other genera in lower proportions, such as Lactobacillus and Streptococcus, highlights the unique heterogeneity of the DILI‐associated microbiota compared to the other conditions analysed.

The microbiota associated with MASLD F≤2 BEVs was characterised by a predominance of Prevotella, Ruminococcus and Faecalibacterium. Other genera of interest included Bifidobacterium, Bacteroides and Anaerofilum, which were present in varying proportions. In MASLD F≥3 BEVs, Prevotella, Ruminococcus and Faecalibacterium continued to constitute the majority of the bacterial community, but with the presence of other notable taxa including Subdoligranulum (Figure 4B).

3.4. Comparative Analysis of Microbiome Profiles in BEVs Isolated From Faecal Samples of Patients With MASLD and DILI

In the previous section, microbiome profiles were analysed independently for each group: patients with DILI, MASLD F≤2, MASLD F≥3 and healthy controls. In contrast, the present analysis integrates these groups, highlighting significant differences in bacterial composition across all four cohorts. The stacked bars in the Figure 4C represent the relative distribution of bacterial genera in each group. These proportions were further compared using DEseq2 analysis, revealing a total of 73 statistically significant differences, which are detailed in Table 3. Of the total differences identified, 58 were related to DILI BEVs when compared with the other groups.

TABLE 3.

Differential abundance analysis of microbiome composition in BEVs samples from control, MASLD F≤2, MASLD F≥3 and DILI patients using DESeq2 analysis.

Group	Comparison	Bacteria	p value
Control vs. DILI	Increase in Control and decrease in DILI	Prevotellaceae_NK3B31_group	<0.001
		Paraprevotella	<0.001
		Enterobacteriaceae	<0.001
		Turicibacter	0.003
	Decrease in Control and increase in DILI	Abiotrophia	<0.001
		AAP99*	<0.001
		Acidaminococcus	<0.001
		Oscillospiraceae_UCG_002	<0.001
		Anaerofilum	<0.001
		Acinetobacter	<0.001
		Adlercreutzia	<0.001
		Neisseriaceae	<0.001
		Anaeroglobus	<0.001
		Aerococcus	0.002
		Actinomyces	0.007
		Actinobacillus	0.014
		Anaerotruncus	0.018
Control vs. MASLD F≤2	Increase in Control and decrease in MASLD F≤2	Dialister	0.012
		Alistipes	0.003
		Prevotellaceae_NK3B31_group	<0.001
		Parabacteroides	0.037
		Barnesiella	<0.001
		Bacteroidales	0.004
		Muribaculaceae	<0.001
		Enterobacteriaceae	0.004
	Decrease in Control and increase in MASLD F≤2	Anaerotruncus	0.005
		Anaerovibrio	<0.001
Control vs. MASLD F≥3	Increase in Control and decrease in MASLD F≥3	Muribaculaceae	<0.001
		Enterobacteriaceae	0.002
Bacteroidales	0.002
	Decrease in Control and increase in MASLD F≥3	Subdoligranulum	<0.001
		Anaerotruncus	0.002
DILI vs. MASLD F≤2	Increase in DILI and decrease in MASLD F≤2	AAP99*	<0.001
		Acinetobacter	0.028
		Aerococcus	0.014
		Actinobacillus	0.043
		Adlercreutzia	<0.001
		Anaeroglobus	<0.001
	Decrease in DILI and increase in MASLD F≤2	Ruminococcus	0.036
		Clostridium	<0.001
		Paraprevotella	0.021
		Anaerobranca	0.028
		Alkaliphilus	<0.001
		Flavobacterium	0.028
		Caloramator	0.015
		Natronincola	0.003
		Desulfotomaculum	0.028
		Caldicellulosiruptor	0.02
		Anaerovibrio	<0.001
DILI vs. MASLD F≥3	Increase in DILI and decrease in MASLD F≥3	AAP99*	<0.001
		Acinetobacter	0.001
		Aerococcus	0.004
		Actinobacillus	0.016
		Anaeroglobus	<0.001
		Abiotrophia	<0.001
	Decrease in DILI and increase in MASLD F≥3	Ruminococcus	0.022
		Clostridium	<0.001
		Paraprevotella	0.013
		Anaerobranca	0.011
		Alkaliphilus	0.004
		Flavobacterium	0.005
		Caloramator	0.038
		Natronincola	0.004
		Desulfotomaculum	0.005
		Caldicellulosiruptor	0.024
		Faecalibacterium	0.005
		Oscillospira	0.017
		Coprococcus	0.036
		Catenibacterium	0.013
		Turicibacter	0.009
		Lachnospira	0.038
MASLD F≤2 vs. MASLD F≥3	Decrease in MASLD F≤2 and increase in MASLD F≥3	Barnesiella	<0.001
		Clostridia_UCG_014	<0.001

*beta proteobacterium AAP99.

3.4.1. Comparative Analysis of BEVs Microbiome Profiles in DILI in Relation to Other Patient Groups

Tables 3 (for p values) and 4 (for a more schematic representation) organise these findings in relation to DILI. A consistent pattern emerged, showing that the bacterial genera AAP99, Acinetobacter, Actinobacillus, Aerococcus and Anaeroglobus were consistently elevated in BEVs from DILI patients, regardless of the comparison group (Table 4) (p values shown in Table 3). In contrast, Paraprevotella consistently showed a significant decrease in DILI BEVs (Table 4) (p values shown in Table 3). When comparing DILI patients with the control group, additional genera were notably increased in DILI BEV, while the Prevotellaceae NK3B31 group (p value < 0.001), Enterobacteriaceae (p value < 0.001) and Turicibacter (p value = 0.003) showed reductions in DILI BEVs (Tables 3 and 4). Similarly, when compared to MASLD F≤2 and MASLD F≥3, DILI BEVs exhibited elevated levels for some of the same genera found in the control group, but a reduction of nine other genera besides Paraprevotella.

TABLE 4.

Composition analysis of BEVs in DILI in relation to other patient groups.

INCREASE IN DILI	DECREASE IN DILI
In comparation with all Groups
AAP99*, Acinetobacter, Actinobacillus, Aerococcus, Anaeroglobus	Paraprevotella
In comparation with the Control Group
AAP99*, Acinetobacter, Actinobacillus, Aerococcus, Anaeroglobus, Acidaminococcus, Actinomyces, Anaerofilum, Anaerotruncus, Neisseriaceae, Oscillospiraceae_UCG_002, Abiotrophia, Adlercreutzia	Paraprevotella, Prevotellaceae_NK3B31_group, Enterobacteriaceae, Turicibacter
In comparation with MASLD F≤2 and MASLD F≥3
AAP99*, Acinetobacter, Actinobacillus, Aerococcus, Anaeroglobus, Abiotrophia, Adlercreutzia	Paraprevotella, Caldicellulosiruptor, Desulfotomaculum, Natronincola, Caloramator, Alkaliphilus, Anaerobranca, Clostridium, Flavobacterium, Ruminococcus
In comparation with MASLD F≤2
AAP99*, Acinetobacter, Actinobacillus, Aerococcus, Anaeroglobus, Adlercreutzia	Paraprevotella, Caldicellulosiruptor, Desulfotomaculum, Natronincola, Caloramator, Alkaliphilus, Anaerobranca, Clostridium, Flavobacterium, Ruminococcus, Anaerovibrio
In comparation with MASLD F≥3
AAP99*, Acinetobacter, Actinobacillus, Aerococcus, Anaeroglobus, Abiotrophia	Paraprevotella, Caldicellulosiruptor, Desulfotomaculum, Natronincola, Caloramator, Alkaliphilus, Anaerobranca, Clostridium, Flavobacterium, Ruminococcus, Faecalibacterium, Oscillospira, Coprococcus, Catenibacterium, Turicobacter, Lachnospira

*beta proteobacterium AAP99.

3.4.2. Composition Analysis of BEVs Microbiome Profiles in MASLD in Relation to Other Patient Groups

The analysis of the microbial composition of BEVs in patients with MASLD revealed significant alterations in comparison to the control group and among different degrees of MASLD. Tables 3 (for p values) and 5 (for a more schematic representation) organise these findings. However, these changes are notably smaller than those observed in DILI. In the MASLD F≤2 BEVs, increases were noted in Anaerotruncus and Anaerovibrio, while decreases were observed in 8 bacterial groups, including Muribaculaceae, Bacteroidales, Enterobacteriaceae, Barnesiella, Dialister, Alistipes, Prevotellaceae_NK3B31 and Parabacteroides (p values shown in Table 3) compared to controls. In comparison with the MASLD F≥3 BEVs, MASLD F≤2 BEVs exhibited decrease in Barnesiella and Clostridia_UCG_014 (p values shown in Table 3). In the MASLD F≥3 BEVs, increases in Anaerotruncus and Subdoligranulum were observed compared to controls, accompanied by decreases in Muribaculaceae, Bacteroidales and Enterobacteriaceae (p values shown in Table 3). It is noteworthy that a comparison of the MASLD F≥3 to MASLD F≤2 BEVs revealed that only Barnesiella and Clostridia_UCG_014 exhibited increases in MASLD F≥3, with no significant changes in other genera (p values shown in Table 3).

TABLE 5.

Composition analysis of BEVs in MASLD in relation to other patient groups.

INCREASE IN MASLD F≤2	DECREASE IN MASLD F≤2
In comparation with the Control Group
Anaerotruncus, Anaerovibrio	Muribaculaceae, Bacteroidales, Enterobacteriaceae, Barnesiella, Dialister, Alistipes, Prevotellaceae_NK3B31_group, Parabacteroides
In comparation with MASLD F≥3
—	Barnesiella, Clostridia_UCG_014
INCREASE IN MASLD F≥3	DECREASE IN MASLD F≥3
In comparation with the Control Group
Anaerotroncus, Subdoligranulum	Muribaculaceae, Bacteroidales, Enterobacteriaceae,
In comparation with MASLD F≤2
Barnesiella, Clostridia_UCG_014	—

3.4.3. Correlations Between the 16 Core Genera Common to the Four Groups of Patients in Faeces and BEVs With Host Clinical Variables

Figure 4D illustrates the correlations between the 16 core genera common to the four groups of patients in faeces and BEVs with host clinical variables. The correlation analysis was conducted exclusively on genera exhibiting relative abundance greater than 1%. The findings indicate that the composition of BEVs exhibited a stronger correlation with clinical variables compared to microbiota composition. A higher presence of Bacteroides, Alistipes, Prevotellaceae_NK3B31_group and Parabacteroides has been associated with a better metabolic profile, as indicated by lower body weight, body mass index, glucose, insulin and HOMA‐IR, as well as a better hepatic profile, as indicated by higher levels of AST, ALT, GGT and ALP. However, a higher presence of Faecalibacterium and Clostridium was associated with a worse metabolic profile, as indicated by weight, body mass index, glucose, insulin and HOMA‐IR.

3.5. Presence of Prokaryotic EVs in Circulating EVs

The circulating EVs were isolated from plasma samples. Figure 5A shows an image of EVs isolated from a DILI patient as described in Section 2. In addition, a Western blot (Figure 5B) reveal the presence of CD9, a specific marker of EVs. Once the EVs have been isolated, the presence of 16S rDNA was observed in the circulating EVs from all patients, but significantly increased in samples from with MASLD and DILI patients (Figure 5C).

FIGURE 5.

Presence of 16S rDNA in circulating EVs (cEVs) from control, MASLD F≤2, MASLD F≥3 and DILI patients. (A) Transmission Electron Microscopy image of EVs after ultracentrifugation (UC) and purification using size exclusion chromatography (SEC) (image from a DILI patient). (B) Western blot of CD9 in a blank sample and in a cEVs sample from a DILI patient (in duplicate) after UC and SEC. Std.: Molecular weight standard. (C) Relative intensity (% relative to control cEVs band) of 16S rDNA PCR bands obtained from cEVs samples from control, MASLD F≤2, MASLD F≥3 and DILI patients and visualised by agarose gel electrophoresis (n = 3/group). Data are expressed as means ± SEM. Groups with different letters statistically differ (p < 0.05). NC: Negative control. M F≤2: MASLD F≤2 patients; M F≥3: MASLD F≥3 patients.

3.6. Effects of Faecal EVs From Patients With MASLD

3.6.1. Effects on TLRs and Immune Response

Following an in‐depth examination of the microbial composition of BEVs, we wanted to test whether these fEVs might affect liver cells differently and whether their effects may be related to pathways associated with MASLD. In this context, hepatic steatosis was induced in vitro by treatment with PA/OA. First, we analysed the mRNA expression of TLR4 and TLR5, receptors involved in the interaction between BEVs and eukaryotic cells triggering the activation of signalling pathways leading to inflammation and fibrosis development (Zheng et al. 2023). We found that MASLD F≥3 fEVs increased the mRNA expression of TLR4 and TLR5 compared to control fEVs (p value = 0.006 and p value = 0.048, respectively) and MASLD F≤2 fEVs (p value = 0.040 and p value = 0.014, respectively), regardless of PA/OA treatment (Figure 6A).

FIGURE 6.

Effects of fEVs from MASLD patients on the mRNA expression of different genes in HepG2 cells (n = 4). (A) Receptors involved in bacterial recognition (TLR4 and TLR5). (B) Genes related to the immune response (IL1B, IL6 and IL10). (C) Secretion of soluble collagen into the culture medium and mRNA expression of genes involved in fibrogenesis (COL1A1 and TGFB1). (D) Genes related to matrix degradation (MMP9 and TIMP1). (E) Genes involved in lipid synthesis (SREBF1, SREBF2, FASN, CD36, DGAT1, PLIN5 and PNPLA3). (F) Genes involved in fatty acid β‐oxidation (CPT1A) and oxidative stress (GPX1). (G) Analysis of the mitochondrial membrane potential: image from HepG2 cells incubated with a control fEVs sample and results from all conditions tested. (H) Receptors involved in mitochondrial dysfunction (OPA1 and DNM1L). (I) Gene involved in apoptosis (CASP3). PA/OA: Palmitic acid/oleic acid; M F≤2: MASLD F≤2 patients; M F≥3: MASLD F≥3 patients. Data are expressed as means ± SEM. Groups with different letters statistically differ (p < 0.05).

After, the impact of fEVs from patients with MASLD on the mRNA expression of specific cytokines was investigated. No significant differences were observed in the mRNA expression of IL1B and IL10 with MASLD fEVs (Figure 6B). In addition, our findings indicate that MASLD F≥3 fEVs increased the mRNA expression of IL6 in comparison to control fEVs (p value = 0.035), regardless of PA/OA treatment (Figure 6B).

3.6.2. Effects on Fibrogenesis and Extracellular Matrix Degradation

The impact of fEVs derived from patients with MASLD on the mRNA expression of markers associated with fibrogenesis and extracellular matrix (ECM) remodelling was also investigated. No significant differences were observed in the mRNA expression of COL1A1, the major component of type I collagen. However, an increase in the TGFB1 mRNA expression, another fibrosis marker, was only observed with MASLD F≤2 fEVs compared to MASLD F≥3 fEVs (p value = 0.009), which was even higher when the cells were simultaneously treated with PA/OA (p value = 0.001) (Figure 6C). Nevertheless, our findings revealed that the secretion of soluble collagen was elevated in the presence of MASLD F≤2 fEVs compared to control (p value = 0.008) and MASLD F≥3 fEVs (p value = 0.003) (Figure 6C).

In the process of ECM degradation and remodelling, metalloproteinases (MMPs) and their inhibitors, known as tissue inhibitors metalloproteinases (TIMPs), play a central role. It was observed that MASLD F≥3 fEVs produced a significant increase of MMP9 mRNA expression compared to control (p value = 0.007) and MASLD F≤2 fEVs (p value = 0.001) (Figure 6D). However, TIMP1 mRNA expression was similar in HepG2 cells incubated with control and MASLD fEVs (Figure 6D).

3.6.3. Effects on Fat Accumulation

To investigate whether fEVs from patients with MASLD induce an increase in liver steatosis, HepG2 cells were treated with the aforementioned fEVs. As expected, PA/OA induced an increase in the number of intracellular lipid droplets in HepG2 cells (Figure 7A,B). Notably, treatment with fEVs from the MASLD F≥3 group resulted in a higher number of lipid droplets in HepG2 cells compared to those treated with control fEVs (p value = 0.029) (Figure 7A,B). This increase was not significant in HepG2 cells treated with MASLD F≤2 fEVs.

FIGURE 7.

(A) Representative images of HepG2 cells treated with fEVs from control, MASLD F≤2, MASLD F≥3 and DILI patients on the accumulation of lipid droplet stained with oil red. The scale is represented by a black bar, denoting a measurement of 50 µm. (B) Optical density (OD) of accumulated red coloration in HepG2 cells treated with fEVs from control, MASLD F≤2 and MASLD F≥3 patients. (C) Optical density (OD) of accumulated red coloration in HepG2 cells treated with fEVs from control and DILI patients. (D) Comparison of the optical density (OD) of the accumulated red coloration in HepG2 cells treated with fEVs from MASLD and DILI patients. PA/OA: Palmitic acid/oleic acid; M F≤2: MASLD F≤2 patients; M F≥3: MASLD F≥3 patients; DIC: Diclofenac; DILI: DILI patients. Data are expressed as means ± SEM (n = 4). Groups with different letters statistically differ (p < 0.05).

In light of these findings, we further investigated the effect of fEVs from patients with MASLD on the mRNA expression of genes associated with adipogenesis, lipid metabolism and oxidative stress. Regarding the expression of genes involved in lipid metabolism, control fEVs upregulated the mRNA expression of FASN compared to MASLD F≤2 (p value < 0.001) and MASLD F≥3 fEVs (p value = 0.001) (Figure 6E). However, the treatment with the fEVs from MASLD F≤2 and MASLD F≥3 resulted in decreased mRNA expression of SREBF1 compared to control fEVs (p value = 0.042 and p value = 0.029, respectively) (Figure 6E). PLIN5 and PNPLA3 were increased with MASLD F≤2 fEVs compared to control fEVs (p value = 0.004 and p value = 0.050, respectively). DGAT1 was increased with MASLD F≥3 fEVs compared to control (p value = 0.035) and MASLD F≤2 (p value < 0.001) (Figure 6E). MASLD F≤2 and MASLD F≥3 fEVs induced an increase in the mRNA expression of SREBF1 (p value = 0.006 and p value = 0.065, respectively), SREBF2 (p value = 0.163 and p value = 0.050, respectively), FAS (p value = 0.309 and p value = 0.024, respectively) in the presence of PA/OA pretreatment compared to its absence (Figure 6E).

3.6.4. Effects on Fatty Acid β‐Oxidation, Mitochondrial Dysfunction and Apoptosis

Regarding the mRNA expression of genes involved in mitochondrial fatty acid uptake and oxidative stress, a decrease in CPT1A mRNA expression was observed in HepG2 cells incubated with MASLD F≥3 fEVs compared to MASLD F≤2 fEVs (p value = 0.038) (Figure 6F). GPX1 mRNA expression was decreased with MASLD F≤2 fEVs compared to control fEVs (p value = 0.003) and MASLD F≥3 fEVs (p value = 0.022). Moreover, the mitochondrial membrane potential was increased in HepG2 cells incubated with MASLD F≥3 fEVs in the presence of PA/OA compared to MASLD F≤2 fEVs (p value = 0.004) (Figure 6G).

There is evidence that gut dysbiosis may disrupt mitochondrial homeostasis (Galloway et al. 2014), including changes in DNM1L‐dependent mitochondrial fragmentation and OPA1. In this study, we investigate whether fEVs from MASLD patients may affect the mRNA expression of genes associated with mitochondrial fusion and fission. With respect to the mitochondrial fusion gene OPA1, no significant effects were observed in HepG2 cells treated with control and MASLD fEVs (Figure 6H). Importantly, MASLD F≤2 fEVs in the presence of PA/OA increased the mRNA expression of DNM1L compared to control fEVs (p value = 0.024) and MASLD F≥3 fEVs (p value = 0.029) (Figure 6H).

The mRNA expression of CASP3, an apoptosis marker, was increased in the presence of MASLD F≤2 (p value = 0.001) and MASLD F≥3 fEVs (p value = 0.003) compared to control fEVs, with a slight increase, although not significant, in the presence of PA/OA (Figure 6I).

3.7. Effects of Faecal EVs From Patients With DILI

3.7.1. Effects on TLRs and Immune Response

As well with the MASLD fEVs, we wanted to test whether the fEVs from DILI patients could affect liver cells. Drug‐induced liver injury was induced in vitro by treatment with diclofenac. Incubation of HepG2 with DIC resulted in upregulation of TLR4 (p value = 0.010) and TLR5 mRNA expression (p value = 0.004) (Figure 8A). Incubation of HepG2 cells with DILI fEVs resulted in an increase in TLR5 mRNA expression compared to control fEVs (p value = 0.008) (Figure 8A). Furthermore, TLR4 and TLR5 mRNA levels were higher when DILI fEVs were simultaneously co‐cultured with DIC compared to the DIC condition (p value = 0.010 and p value = 0.046, respectively), thereby indicating an enhanced response (Figure 8A). These results suggest that fEVs from DILI patients induce TLR4 and TLR5 mRNA expression.

FIGURE 8.

Effects of fEVs from DILI patients on the mRNA expression of different genes in HepG2 cells (n = 4). (A) Receptors involved in bacterial recognition (TLR4 and TLR5). (B) Genes related to the immune response (IL1B, IL6 and IL10). (C) Secretion of soluble collagen into the culture medium and mRNA expression of genes involved in fibrogenesis (COL1A1 and TGFB1). (D) Genes related to matrix degradation (MMP9 and TIMP1). (E) Genes involved in lipid synthesis (SREBF1, SREBF2, FASN, CD36, DGAT1, PLIN5 and PNPLA3). (F) Genes involved in fatty acid β‐oxidation (CPT1A) and oxidative stress (GPX1). (G) Analysis of the mitochondrial membrane potential: image from HepG2 cells incubated with a DILI fEVs sample and results from all conditions tested. (H) Receptors involved in mitochondrial dysfunction (OPA1 and DNM1L). (I) Gene involved in apoptosis (CASP3). PA/OA: Palmitic acid/oleic acid; DILI: DILI patients. Data are expressed as means ± SEM. Groups with different letters statistically differ (p < 0.05).

The effect of fEVs from DILI patients on the mRNA expression of specific cytokines in HepG2 cells was then investigated. Incubation of HepG2 cells with DILI fEVs resulted in increased mRNA expression of IL1B (p value = 0.007) and IL6 (p value = 0.007) compared to control fEVs, independent of DIC treatment (p value = 0.001 and p value = 0.001, respectively) (Figure 8B). In contrast, IL10 mRNA expression was increased only in the presence of DIC (p value = 0.003), but not with DILI fEVs (Figure 8B). Thus, the results suggest that fEVs from patients with DILI induce an enhanced immune response.

3.7.2. Effects on Fibrogenesis and Extracellular Matrix Degradation

The effect of DILI fEVs on the mRNA expression of markers associated with fibrogenesis and extracellular matrix (ECM) remodelling was also investigated. The mRNA expression of COL1A1 (p value < 0.001) and MMP9 (p value < 0.001) was increased with the co‐incubation of DILI fEVs + DIC compared to control fEVs (Figure 8C,D). Conversely, only DIC was found to significantly increase TGFB1 (p value = 0.034) and TIMP1 expression (p value = 0.001) and the secretion of soluble collagen (p value = 0.046) (Figure 8C,D). These results suggest that DIC induces an enhanced mRNA expression of fibrogenic markers.

3.7.3. Effects on Fat Accumulation

Interestingly, DIC increased the accumulation of lipid droplets in HepG2 cells (p value = 0.003) (Figure 7C). Furthermore, the incubation with DILI fEVs also resulted in a greater number of lipid droplets within HepG2 cells (p value < 0.001) compared to those treated with control fEVs (Figure 7D). We further investigated the effect of DILI fEVs on the mRNA expression of genes associated with adipogenesis, lipid metabolism and oxidative stress. DILI fEVs increased the mRNA expression of DGAT1 (p value = 0.050) and PNPLA3 (p value = 0.001) compared to cells treated with control fEVs (Figure 8E). However, DILI fEVs decreased the mRNA expression of FASN (p value = 0.031) (Figure 8E). DIC administration increased the mRNA expression of SREBF2 (p value = 0.009), FASN (p value = 0.003) and PNPLA3 mRNA expression (p value = 0.002), indicating an activation of lipid synthesis (Figure 8E). Furthermore, co‐incubation of DILI fEVs with DIC resulted in a greater increase in the SREBF1 (p value = 0.002) and DGAT1 mRNA expression (p value = 0.020) (Figure 8E).

3.7.4. Effects on Fatty Acid β‐Oxidation, Mitochondrial Dysfunction and Apoptosis

Regarding the mRNA expression of genes involved in mitochondrial fatty acid uptake and oxidative stress, only DIC increased the mRNA expression of CPT1A (p value = 0.003) (Figure 8F). In addition, only DIC increased the mitochondrial membrane potential compared to the other conditions (p value < 0.05) (Figure 8G).

Regarding the mitochondrial fusion gene OPA1 and DNM1L, no significant effects were observed in HepG2 cells treated with DILI fEVs compared to control fEVs (Figure 8H). However, DIC increased the mRNA expression of OPA1 (p value = 0.009), which was enhanced by the co‐incubation with DILI fEVs (p value < 0.001) (Figure 8H). Importantly, the combination of DIC with DILI fEVs results in a significant upregulation of DNM1L mRNA expression (p value = 0.002) (Figure 8H).

The mRNA expression of CASP3, an apoptosis marker, was increased in the presence of DILI fEVs compared to control fEVs (p value = 0.042) and DIC (p value = 0.012) (Figure 8I).

3.8. Comparison of the Effects of Faecal EVs From Patients With MASLD and DILI

3.8.1. Effects on TLRs and Immune Response

After a thorough analysis of the effects of fEVs from patients with MASLD and DILI, the goal was to compare their effects to determine if there was any similarity between the molecular pathways activated by the fEVs from these liver diseases. The findings showed that only MASLD F≥3 fEVs produced an increase of TLR4 mRNA expression compared to MASLD F≤2 and DILI fEVs (p value = 0.008 and p value = 0.039, respectively) (Figure 9A). FEVs from DILI and MASLD F≥3 increased TLR5 mRNA expression (p value = 0.039 and p value = 0.008, respectively) compared to MASLD F≤2 fEVs (Figure 8A). However, only DILI fEVs increased IL1B (p value = 0.006) compared to the effect of MASLD F≤2 fEVs (Figure 9B).

FIGURE 9.

Comparison between the effects of fEVs from patients with MASLD and DILI on the mRNA expression of different genes in HepG2 cells (n = 4). (A) Receptors involved in bacterial recognition (TLR4 and TLR5). (B) Genes related to the immune response (IL1B, IL6 and IL10). (C) Secretion of soluble collagen into the culture medium and mRNA expression of genes involved in fibrogenesis (COL1A1 and TGFB1). (D) Genes related to matrix degradation (MMP9 and TIMP1). (E) Genes involved in lipid synthesis (SREBF1, SREBF2, FASN, CD36, DGAT1, PLIN5 and PNPLA3). (F) Genes involved in fatty acid β‐oxidation (CPT1A) and oxidative stress (GPX1). (G) Analysis of the mitochondrial membrane potential. (H) Receptors involved in mitochondrial dysfunction (OPA1 and DNM1L). (I) Gene involved in apoptosis (CASP3). M F≤2: MASLD F≤2 patients; M F≥3: MASLD F≥3 patients; DILI: DILI patients. Data are expressed as means ± SEM. Groups with different letters statistically differ (p < 0.05).

3.8.2. Effects on Fibrogenesis and Extracellular Matrix Degradation

The results showed that fEVs from MASLD F≥3 and DILI patients decreased the TGFB1 mRNA expression (p value = 0.014 and p value = 0.024, respectively) and soluble collagen secretion (p value = 0.014 and p value = 0.024, respectively) compared to the effects of MASLD F≤2 fEVs, and increased the MMP9 mRNA expression (p value = 0.006 and p value = 0.050, respectively) (Figure 9C,D).

However, the mRNA expression of COL1A1 was significantly increased in response to DILI fEVs compared to MASLD F≤2 fEVs (p value = 0.014) and MASLD F≥3 fEVs (p value = 0.050).

3.8.3. Effects on Fat Accumulation

Treatment of HepG2 cells with fEVs from MASLD F≥3 patients resulted in a higher accumulation of lipid droplets than that induced by MASLD F≤2 fEVs (p value = 0.002), with DILI fEVs having an intermediate effect (Figure 7D).

The effects of MASLD F≥3 and DILI fEVs on FASN (p value = 0.036 and p value = 0.002, respectively) and DGAT1 mRNA expression (p value = 0.002 and p value = 0.047, respectively) were similar and significantly higher than those induced by MASLD F≤2 fEVs (Figure 9E). In addition, DILI fEVs increased SREBF1 (p value = 0.039 and p value = 0.008, respectively) and PNPLA3 mRNA expression (p value = 0.050 and p value = 0.006, respectively) compared to the effects induced by MASLD F≤2 and MASLD F≥3 fEVs (Figure 9E).

3.8.4. Effects on Fatty Acid β‐Oxidation, Mitochondrial Dysfunction and Apoptosis

The results showed an increased mRNA expression of CPT1A (p value = 0.008) (Figure 9F) and CASP3 (p value = 0.002) (Figure 9I) with DILI fEVs compared to MASLD F≥3 fEVs, but similar to MASLD F≤2 fEVs. Conversely, a substantial increase in GPX1 mRNA expression was observed with DILI fEVs (p value = 0.002) compared to MASLD F≤2 fEVs (Figure 9F). In addition, DILI fEVs increased the mitochondrial membrane potential compared to MASLD F≤2 fEVs (p value = 0.009) (Figure 9G).

4. Discussion

The results of this study indicate a notable decline in the diversity index of microbial populations in BEVs compared to faecal samples. A comparison of BEVs microbiome profiles in patients with DILI, MASLD and healthy controls revealed a complex interplay between BEVs microbiome and liver disease, with higher changes in the acute liver disease (DILI) compared to chronic liver disease (MASLD). This discrepancy in microbiome composition may be at the root of the different effects observed with these faecal EVs. The results showed an increase in the response to bacterial products, the pro‐inflammatory response and the accumulation of lipid droplets in hepatic cells in response to fEVs derived from DILI patients and from MASLD patients with significant liver fibrosis (MASLD F≥3). Moreover, fEVs from the latter group of patients decreased the entry of fatty acids into the mitochondria for β‐oxidation (CPT1A). Furthermore, there was a trend towards increased mitochondrial fission (DNM1L) with fEVs from MASLD and DILI patients when co‐incubated with PA/OA or diclofenac (DIC), respectively. In addition, our results suggest that collagen secretion was enhanced only by fEVs from MASLD F≤2 patients, but not by those from MASLD F≥3 or DILI patients. This collagen secretion is consistent with the TGFB1 expression. Conversely, our results suggest that DILI fEVs exert a potentiating effect, significantly increasing the hepatotoxic impact of DIC. This is evidenced by an increased response to microbial components, inflammatory response, lipid droplet accumulation and mitochondrial dysfunction.

Our analysis revealed significant differences in the microbial diversity index between faecal samples and the BEVs. Notably, in patients with MASLD F≤2, both alpha‐diversity and evenness were significantly lower in BEVs compared to faecal samples. In addition, bacterial richness showed significant variation between faeces and BEVs in patients with DILI and MASLD F≥3, and between BEVs from the four patient groups. These findings may suggest a differential distribution of bacterial communities within BEVs, possibly reflecting different pathophysiological mechanisms specific to each patient group, as previously described for faecal bacterial communities (Rodriguez‐Diaz et al. 2022). In this context, we identified significant bacterial differences between patient groups when comparing faeces composition and BEVs, which may reflect the influence of disease‐specific factors on the distribution of microbial communities in BEVs (Rodriguez‐Diaz et al. 2022). While there was only one genus with significant differences between BEVs and faecal samples in the control group and two in the MASLD F≤2 group, a wider range of taxa showed significant differences in the MASLD F≥3 and DILI groups. The divergence observed in alpha diversity indices between faecal samples and BEVs underscores the biologically selective nature of vesicle biogenesis within the gut ecosystem. BEVs do not constitute a direct reflection on the entire bacterial community; rather, they derive from specific taxa actively engaged in vesicle secretion under distinct physiological or pathological contexts. This selective release process may preferentially capture microbial populations with heightened metabolic activity or stress responsiveness, thereby shaping the distinct microbial profiles identified in BEVs samples. In patient groups such as MASLD F≤2, the marked reduction in BEVs associated with diversity compared to stool suggests that vesicle production may be driven by a restricted set of bacterial taxa. This pattern could indicate that inflammatory or metabolically altered intestinal conditions favour vesicle release from specific microbial populations with adaptive or stress‐related bacterial groups. In contrast, in the DILI cohort, although the gut microbiota exhibits clear signs of dysbiosis, the absence of significant differences in alpha diversity between stool and BEV samples may reflect that such alterations equally affect both the total bacterial community and the vesicle‐producing bacteria. Consequently, the relative diversity between these two fractions remains comparable despite underlying compositional disruption. Differences in bacterial load, vesicle isolation efficiency and host‐related variables, including mucosal inflammation, intestinal permeability or bile acid metabolism, may further contribute to the observed diversity patterns. This suggests that these bacteria may actively modulate the immune or inflammatory response through the molecular signalling of these BEVs (Bhardwaj et al. 2018). However, the specific role of BEVs in these liver diseases remains poorly understood.

Most studies have focused on the association between gut microbiota and various diseases. However, there are few studies analysing the association with the composition of BEVs. A detailed analysis of the most prevalent taxa in faeces and fEVs in our study revealed a stronger correlation between the clinical characteristics of patients and the composition of fEVs than with the composition of gut microbiota. However, nothing has been described in this regard. Our findings partially corroborate the associations previously described in gut microbiota, which are sometimes contradictory. For instance, the associations of Ruminococcus, Bacteroides, Faecalibacterium, Parabacteroides and Clostridium with obesity and metabolic disorders have been inconsistent in previous studies (Hoseini Tavassol et al. 2023). Regarding the significant correlations found with liver enzymes, a low abundance of Prevotellaceae_NK3B31_group, Bacteroides and Parabacteroides has been demonstrated to be strongly associated with a worse liver profile. Our results obtained in fEVs corroborate those obtained directly in the gut microbiota, which demonstrated that the abundance of Bacteroides and Prevotellaceae_NK3B31_group was diminished in subjects with high liver fat compared to subjects with low liver fat (Driuchina et al. 2023). This study even proposed that low faecal Prevotellaceae_NK3B31_group and Bacteroides abundance may serve as biomarkers of high liver fat. In a separate study, a species of the genus Parabacteroides (Parabacteroides distasonis) has been demonstrated to play a potential role in the protection of hepatic diseases (Duan et al. 2024).

A comparison of the microbiome profile of BEVs from patients with DILI, MASLD and healthy controls revealed a complex interplay between BEVs microbiome and liver disease. BEVs from DILI patients showed a specific profile compared to controls and MASLD patients, with a significant increase in AAP99, Acinetobacter, Actinobacillus, Aerococcus and Anaeroglobus, and a marked decrease in Paraprevotella. However, no effects of their BEVs on the liver have been described. In addition, the relationship between these bacteria and liver disease is complex, and there is little information on the existence of any relationship. A previous study identified Paraprevotella BEVs as a dominant taxon in meconium, suggesting they are shed during gut colonisation (Turunen et al. 2023). An increase in Acinetobacter has been found in the gut microbiota of mice after paracetamol overdose (Ren et al. 2024). It is known that Acinetobacter produces BEVs (Moon et al. 2012), which can contain multiple virulence factors (Kwon et al. 2009), and can induce early onset apoptosis in dendritic cells (Mehanny et al. 2020). In this regard, our results also showed an increase in apoptosis (CASP3) after the incubation of HepG2 cells with DILI fEVs. In addition, a study in nonalcoholic hepatic steatosis showed that the Acinetobacter abundance was positively associated with an increase in kynurenine (Sui et al. 2021). The increase of this metabolite could indicate a common pathway affected by this bacterium and DILI. A previous study in a mouse model of doxorubicin‐induced acute liver injury found that this drug activates the L‐tryptophan/L‐kynurenine metabolic pathway, increasing kynurenine levels (Tang et al. 2025). Moreover, our results are also in line with those showing an increase in hepatic IL1B and IL6 after the infection of pigs with Acinetobacter pleuropneumoniae (Skovgaard et al. 2010). On the other hand, in the context of iDILI, there is an association between cytochrome P450 metabolism and the most commonly involved drugs (Teschke and Uetrecht 2021). In this regard, the suppression of oxidative hepatic drug metabolism and changes in the hepatic cytochrome P4503A mRNA levels have been shown in pigs infected with Actinobacillus pleuropneumoniae (Teschke and Uetrecht 2021). Regarding Aerococcus, other drugs such as methotrexate also increased its abundance (Wang et al. 2022). However, there is no information on the possible relationship between Anaeroglobus and liver metabolism. Only an increase in Anaeroglobus geminatus and Actinobacillus pleuropneumoniae was found in patients with primary biliary cirrosis (Lv et al. 2016). All of these results suggest that there may be a relationship between the gut microbiota, and its BEVs, and certain drugs that can cause hepatotoxicity, although further studies are needed.

The analysis of BEVs in patients with MASLD reveals alterations in their composition compared to the control group and, to a lesser degree, between different stages of the disease. Although evident, they appear to be less pronounced than those observed in DILI, indicating that alterations of BEV composition in MASLD may be more gradual or nuanced compared to more acute liver diseases (DILI). Nevertheless, the modification of the BEVs composition in the MASLD F≥3 group in comparison to the control group may be associated with the advanced age of the former group or the presence of a greater number of patients with altered glucose metabolism. That is to say, this group of patients exhibits a worse metabolic profile. It has been established that advanced age and the presence of T2DM are associated with alterations in the composition of the gut microbiota (Chong et al. 2025). This may be the underlying cause of the alterations observed in the composition of BEVs. However, given the close interrelatedness of age, T2DM, and MASLD, it is difficult to separate their possible contributions. The presence of either age or T2DM could serve as a confounding factor in the observed outcomes. Regarding the BEVs composition, in both MASLD F≤2 and MASLD F≥3 groups, the observed increase in Anaerotruncus suggests that BEVs from this genus may contribute to the onset of early metabolic disturbances in the liver and promote MASLD progression. An increase in Anaerotruncus has been reported in mice with NAFLD compared to controls (Quesada‐Vázquez et al. 2022), and negatively correlated with the levels of indolepropionic acid (IPA) (Zhang et al. 2022), a metabolite with known antioxidant and anti‐inflammatory properties. The analysis of BEVs in patients with MASLD also reveals significant decreases in Muribaculaceaee, Bacteroidales, Enterobacteriaceae compared to the control group, which is in agreement with the findings of various studies indicating that changes in the abundance of bacteria from Bacteroidales and Enterobacteriaceae were associated with NAFLD (Li et al. 2021; Hiippala et al. 2020). Furthermore, a study conducted on C57BL/6J mice fed a high‐fat diet suggested a relationship between a reduced abundance of Muribaculaceae and the development of NAFLD (Zhu et al. 2023). It is noteworthy that the present study has barely identified any differences between the composition of fEVs of MASLD F≤2 and MASLD F≥3 patients. This may reinforce the significance of microbial diversity, rather than the prevalence of specific species, in MASLD progression. This observation underscores the need for further investigation into the role of BEVs and the microbial signatures they convey in liver disease progression, representing a promising avenue of research to comprehend microbial dynamics at various stages of disease.

As previously demonstrated (Rodriguez‐Diaz et al. 2022), there are differences in the gut microbiota of these patient groups. Consequently, the composition of fEVs may undergo alterations, as evidenced by the findings of this study. We found a different distribution of bacterial communities within the fEVs, according to the nature of the underlying liver disease. This may imply different functional capabilities within the host organism, and may also be related to the different impact observed on HepG2. These fEVs contain lipopolysaccharides, pathogen‐associated molecular patterns, peptidoglycans, lipoteichoic acid and other molecules as an integral part of BEVs membranes. Consequently, the presence of these molecules would be responsible, to a greater or lesser extent, for the effects of fEVs found in the in vitro studies (De Langhe et al. 2024; Toyofuku et al. 2023). We have shown that MASLD and DILI patients have a higher presence of circulating BEVs than healthy controls. This could be due to a higher intestinal permeability in these patients, as other studies have shown (Benedé‐Ubieto et al. 2024). Their presence in circulating blood could have different effects on the liver. However, the possible effect of certain BEVs may be enhanced or counteracted by the effect of other BEVs. For this reason, we wanted to analyse the effects of all the EVs present in faeces, regardless of their bacterial origin. These BEVs can be recognised by the target cells via pattern recognition receptors (PRRs), such as TLRs, and initiate signalling cascades. Our results showed an increase in TLR4 and TLR5 mRNA expression in response to fEVs derived from patients with MASLD F≥3. Our results are also consistent with a previous study showing that fEVs from MASLD patients with advanced fibrosis (MASH) induced inflammation through TLR4 activation (Fizanne et al. 2023). However, unlike the previous study in stellate cells, our investigation was carried out in HepG2 cells, a cell line derived from human hepatocytes, which are the predominant cell type in the liver and a suitable model for studying lipids in human hepatocytes in vitro. The activation of these TLRs is closely linked to the stimulation of the immune system, which is associated with the development of MASLD. Our results showed an increase in the pro‐inflammatory response only with the fEVs from MASLD patients with higher liver fibrosis (IL6). This result was independent of the previous presence of hepatic steatosis (treatment with PA/OA). This is in line with another study comparing the effect of fEVs from healthy individuals and patients with different degrees of NAFLD on the inflammatory profile of the LX2 hepatic stellate cell line (Fizanne et al. 2023). However, that study does not analyse the effects on hepatocytes, nor does it analyse subjects with acute liver damage such as DILI. Nevertheless, the anti‐inflammatory response (IL10) seems to be triggered exclusively by MASLD F≥3 fEVs and in conjunction with the presence of steatosis (treatment with PA/OA), that is, in a state where the hepatic cells are already altered.

Metagenomic analysis of fEVs revealed changes in the profile depending on the degree of liver fibrosis in MASLD patients. This phenomenon is a hallmark of MASH patients. However, there are few studies on the potential effects of these fEVs on steatosis and fibrogenesis. Previous studies in Kupffer and stellate cells have demonstrated that LPS has the ability to induce liver fibrosis by activating the TLR4/TRIF/GBPs signalling pathway (Villard et al. 2021; Jiang et al. 2023) and to increase collagen deposition by increasing the mRNA expression of genes involved in the collagen synthesis, such as COL1A1 and TGFβ1 (Affò et al. 2014). In another study, NASH fEVs were observed to increase the expression of COL1A1 and TGFB1 in LX2 human hepatic stellate cells (Fizanne et al. 2023). Nevertheless, our findings in parenchymal liver cells do not fully confirm these results. Our results indicate that collagen secretion was increased with fEVs from MASLD F≤2 patients, but not with fEVs from MASLD F≥3 patients. This collagen secretion is more consistent with TGFB1 mRNA expression, which is increased only in the fEVs of MASLD F≤2 patients. TGFB1 is considered to be a central player in liver fibrosis, contributing to collagen production and deposition (Longhitano et al. 2024; Choi et al. 2020). However, it is important to note that HepG2 cells, unlike stellate cells, are not the primary source of collagen in the liver. Perhaps, the secretion of TGFB1 and pro‐inflammatory cytokines by HepG2 cells may increase the secretion and deposition of collagen by other cell types, such as stellate cells. However, this needs to be confirmed in further studies.

The main feature of MASLD is the accumulation of lipids in liver cells (Keshavarz Azizi Raftar et al. 2021). Our study showed an increase in lipid droplets in hepatic cells treated with fEVs from MASLD F≥3 patients compared to those treated with control fEVs. However, our results suggest that there is no increase in de novo lipogenesis, as SREBF1 and FASN mRNA expression were decreased with MASLD fEVs. In contrast, the mRNA expression of other proteins involved in the regulation of lipid accumulation, such as PLIN5 and PNPLA3, were increased with MASLD fEVs (Ma et al. 2019; Dorairaj et al. 2021). In addition, DGAT1 mRNA expression, the other enzyme analysed related to lipogenesis, was significantly increased with MASLD F≥3 fEVs, which may contribute to the higher accumulation of lipid droplets. This increase in DGAT1 with MASLD F≥3 fEVs is also associated with a decrease in CPT1A mRNA expression, the rate‐limiting enzyme of β‐oxidation in the liver, suggesting that a higher triglyceride synthesis would be associated with the decreased entry of fatty acids into the mitochondria for β‐oxidation. This mitochondrial process leads to energy production under normal conditions. However, when this process becomes excessive, as it does in MASLD, an increase in ROS production would be observed, and consequently, an increased antioxidant response. However, our results show a downregulation of GPX1 mRNA expression. This would be consistent with a decrease in the antioxidant capacity of liver cells in MASLD, a characteristic of MASH progression (Han and Kaufman 2016). In this regard, we found that MASLD F≥3 fEVs in a state of steatosis (with PA/OA pretreatment) were the only ones capable of increasing mitochondrial membrane potential. This increase in mitochondrial membrane potential would increase GPX1 mRNA expression as a cellular defence mechanism (Lubos et al. 2011), as we have verified. However, the increase observed in GPX1 mRNA expression with control fEVs without an increase in mitochondrial membrane potential could reflect an additional defence mechanism activated by another pathway, which is not activated by fEVs from patients with MASLD. Taken together, our findings suggest a possible involvement of fEVs from MASLD F≥3 patients in the development of hepatocyte steatosis.

There is evidence that MASLD is closely linked to mitochondrial dysfunction, which is regulated by mitochondrial fusion (OPA1) and fission (DNM1L) events (Hernández‐Alvarez and Zorzano 2021). However, there is a tendency towards increased mitochondrial fission (Radosavljevic et al. 2024), as our results show. There is an increase in DNM1L mRNA expression with fEVs from MASLD patients, with a more significant effect when these fEVs were incubated in steatotic HepG2 cells (cells treated with PA/OA). However, the fusion process (OPA1 mRNA expression) was not significantly affected by fEVs. It has been proposed that DNM1L‐mediated mitochondrial fragmentation observed in NAFLD exacerbates hepatic steatohepatitis and liver injury (Zheng et al. 2023). Furthermore, in a murine model of MASLD, reduced mitochondrial fission was associated with the amelioration of hepatic steatosis (Galloway et al. 2014). In this regard, previous studies have investigated the effect of BEVs on mitochondrial integrity. As previously reported by our group, treatment of HepG2 cells with Clostridioides difficile‐derived EVs resulted in the upregulation of the mitochondrial fission genes DNM1L and FIS (Caballano‐Infantes et al. 2023). This suggests a potential role for fEVs in the regulation of mitochondrial fusion and fission events and consequently, in the development of the MASLD spectrum, a hypothesis supported by previous studies (Legaki et al. 2022; Ramachandran et al. 2021).

There is a lack of studies on DILI. Previous studies have shown an association between DILI an altered gut microbiota composition and immune response (Rodriguez‐Diaz et al. 2022; Cueto‐Sanchez et al. 2021). Some authors have reported that DIC, a pharmaceutical agent commonly used to study DILI in vitro (Segovia‐Zafra et al. 2024), is able to modulate the inflammatory response by interacting with the TLR4/NF‐κB pathway (Barcelos et al. 2016). In this regard, we found that DIC activated the mRNA expression of TLR4, TLR5 and both a pro‐inflammatory (IL1B and IL6) and an anti‐inflammatory (IL10) response in HepG2 cells, in agreement with other studies (Fredriksson et al. 2014). In relation to the fEVs, the present study has demonstrated an increase in TLR5 and the pro‐inflammatory response (IL6 and IL1B) in response to fEVs derived from DILI patients. In addition, DILI fEVs significantly enhanced the DIC‐induced inflammatory response and the mRNA expression of receptors involved in the interaction between BEVs and eukaryotic cells. This finding suggests that the hepatotoxicity of certain drugs may be enhanced by the action of bacterial products, such as fEVs.

Although liver fibrosis is not one of the most important events in DILI, we have found that DIC increased the collagen secretion, which is also consistent with the increase in TGFB1 mRNA expression. Moreover, the combination of DIC and DILI fEVs increased the mRNA expression of COL1A1 and MMP9, suggesting that fEVs seem to be involved in the regulation of liver fibrosis, especially when there is concomitant treatment with certain hepatotoxic drugs. This may be related to the increase in TLR4 mRNA expression, as LPS appears to induce liver fibrosis by activating the TLR4/TRIF/GBPs signalling pathway (Villard et al. 2021; Jiang et al. 2023) and increasing collagen deposition by increasing gene expression of COL1A1 and TGFβ1 (Affò et al. 2014).

Our study also shows a significant increase in lipid droplet accumulation in hepatic cells treated with fEVs from DILI patients or with DIC, with increased mRNA expression of DGAT1 and PNPLA3. This was particularly evident when HepG2 cells were cocultured with DILI fEVs+DIC, which also increased the mRNA expression of CD36 and PLIN5, proteins involved in fatty acid uptake by liver cells and lipid droplet stabilisation, respectively. However, these DILI fEVs also increased the mRNA expression of CPT1A, suggesting that, in addition to increasing lipid synthesis, DILI fEVs appear to increase β‐oxidation. This process takes place within the mitochondria, whose fusion and fission processes are minimally affected by DILI fEVs. In addition, as seen also with MASLD fEVs, apoptosis appears to be increased with both DILI fEVs and DIC. In addition to the effects of DILI fEVs, we have shown how DIC can activate fat deposition in hepatic cells and lead to mitochondrial dysfunction by increasing the mitochondrial membrane potential, mitochondrial fusion (OPA1) and fission (DNM1L), as observed with other drugs (Bessone et al. 2018; Satapathy et al. 2015). However, studies on this topic are lacking.

A relevant finding of this study is the existence of a certain similarity between the effects of fEVs from MASLD F≥3 and DILI patients on the expression of different genes, which are mainly involved in bacterial recognition, immune response, fibrosis and lipid synthesis. Although conclusions must be drawn with some caution, these results may support the hypothesis of a certain relationship, at least at the molecular level, between MASLD and DILI, which has been shown in certain studies showing that up to 26% of DILI cases had some degree of steatosis (Kleiner et al. 2014).

These findings could have different clinical implications. The gut microbiota has gained increasing interest as a potential source of diagnostic and prognostic biomarkers for various liver diseases. Clinical studies have demonstrated that particular microbial signatures are associated with the presence and severity of liver disease, with specific microbial products differentiating simple steatosis from progressive forms that result in fibrosis (Rodriguez‐Diaz et al. 2022; Satthawiwat et al. 2024). In this regard, microbiota‐derived products, such as BEVs, are being explored as non‐invasive markers of liver injury and metabolic dysfunction, serving as non‐invasive alternatives to liver biopsy (Fizanne et al. 2023). The present study provides a distinctive BEVs‐based profile of DILI and MASLD patients according to fibrosis levels. These findings suggest that these biomarkers could be utilised as potential indicators of liver disease. The incorporation of these findings into clinical practice may have the potential to enhance diagnostic accuracy, enable the identification of patients at risk of progression, and facilitate personalised monitoring and treatment strategies. Furthermore, the functional capacities of these BEVs are associated with cellular and molecular alterations that have been described in the context of DILI and MASDL diseases. These results could provide multiple therapeutic opportunities for modulating the effects of microbiota to prevent or treat these liver diseases. In addition to targeted microbiota interventions, including probiotics, prebiotics, synbiotics and postbiotics, the modulation of BEVs could be readily engineered to be utilised for clinical applications (Peregrino et al. 2024). For instance, the utilisation of these BEVs could encompass applications such as drug delivery, vaccine therapy, the modulation of immune responses or the mitigation of drug resistance. However, as with other novel therapeutic approaches, further investigation and standardisation are necessary to ensure its efficacy and safety. This approach has the potential to serve as a personalised therapeutic intervention. However, the necessity for additional randomised controlled trials of a larger scale is evident.

A limitation of this study is the relatively small sample size of the groups. DILI is a rare event, which inherently limit the possibility of enrolling large patient cohorts. For this reason, we used very selective exclusion criteria to avoid likely confounding by liver diseases. Therefore, although we know the limited sample size in some patient groups, the statistical robustness of the majority of our findings supports their reliability. Although it would have been desirable to measure protein levels to reinforce the results, we believe these results offer valuable preliminary insights that can inform and guide future investigations. On the other hand, the effect of different drugs, mainly antibiotics, on the gut microbiota could influence the pathogenesis of these chronic liver diseases (Liu et al. 2022). In this regard, the use of antibiotics was documented for all patients included in the study. However, given the heterogeneity of DILI, which is caused by diverse medications and the limited sample size of the study, further explorations of how different drugs influence the composition and function of fEVs would be beneficial. In addition, extending the experiments to include primary hepatocytes, hepatic stellate cells and immune cells would allow a more comprehensive assessment of the effects of fEVs and provide a more holistic view of disease mechanisms. It is important to note that all statistical analyses were performed at the genus level, as identification at the species level based on 16S rDNA sequencing targeting the V3‐V4 hypervariable region in human faeces should be considered only with caution. Further studies investigating bacterial metabolic functions in more detail will elucidate their future applications in the detection, prevention and treatment of these diseases. Another important point to bear in mind is that primary hepatocytes are the most suitable cell model for establishing liver cell culture models. However, they are difficult to obtain and are very invasive. There are also certain disadvantages and differences between donors (Sison‐Young et al. 2015; den Braver‐Sewradj et al. 2016). Moreover, and unfortunately, in vivo animal models are not yet widespread for studying iDILI (Ballet 2015). On the other hand, HepG2 cells alone cannot reproduce the entire hepatic environment. A large number of cell types, such as liver sinusoidal endothelial cells, hepatic stellate cells, cholangiocytes, Kupffer cells, B cells, natural killer T cells, mucosa‐associated invariant T cells, myeloid dendritic cells and plasmacytoid dendritic cells, can affect the hepatic environment. For this reason, it is highly complex to establish an adequate model that combines a wide variety of cell types when analysing different aspects of metabolism, especially from an immune perspective. However, our focus was solely on the possible effect that BEVs may have on hepatocytes, an aspect not studied to date.

In conclusion, our results indicate a differential distribution of bacterial communities within BEVs according to the type of liver disease. The results suggest that the composition of BEVs may undergo more gradual or subtle changes in MASLD compared to more acute liver diseases (DILI). This observation highlights the need for further investigation into the role of BEVs and the microbial signatures they carry in the progression of liver disease. The results of our study suggest that fEVs have the ability to enter the bloodstream, and may contribute to the pathological mechanisms involved in the progression of the MASDL spectrum, the development of hepatocyte steatosis and DILI. Although they are different liver diseases, they appear to share common molecular mechanisms, particularly between MASLD patients with higher levels of fibrosis and DILI patients. Furthermore, in the latter group of patients, the fEVs appear to enhance the hepatotoxic injury caused by certain drugs involved in DILI, such as diclofenac. In light of the potential relationship identified in this study between the composition of fEVs and alterations in liver metabolism, our findings could facilitate a novel approach to modify the intestinal microbiota with the aim of mitigating the development of these liver pathologies. Further studies are needed to gain a deeper understanding of the gut‐liver relationship, particularly the role of the gut microbiome modulating the immune‐inflammatory response.

Author Contributions

Antonio J. Ruiz‐Malagón: investigation, formal analysis, writing – original draft. Marina Herraiz‐Vilela: investigation. Jose Pinazo‐Bandera: investigation. Juan Pedro Toro‐Ortiz: investigation. Carlos López‐Gómez: investigation. Ailec Ho‐Plagaro: investigation. Lourdes Garrido‐Sánchez: investigation. Mercedes Robles‐Díaz: investigation. Bernard Taminiau: investigation, formal analysis. Georges Daube: investigation, formal analysis. Judith Sanabria‐Cabrera: investigation. Gonzalo Matilla‐Cabello: investigation. M. Isabel Lucena: writing – review and editing, visualization, funding acquisition. Raúl J. Andrade: funding acquisition, visualization, writing – review and editing. Eduardo García‐Fuentes: conceptualization, methodology, formal analysis, resources, writing – original draft, writing – review and editing, funding acquisition, visualization. Miren García‐Cortes: methodology, formal analysis, resources, writing – review and editing, funding acquisition, visualization. Cristina Rodriguez‐Diaz: conceptualization, methodology, formal analysis, resources, visualization, writing – original draft.

Funding

This work was supported by the Instituto de Salud Carlos III (Spain) through the proyects PI18/01804, PI19/00883, PI21/01248 and ‘FORT23/00013’ Programa Fortalece of the Ministry of Science, Innovation and Universities, from the Consejería de Economía, Conocimiento, Empresas y Universidad (Junta de Andalucía, Spain) (PI18‐RT‐3364, UMA18‐FEDERJA‐194), and from the Consejería de Salud (Junta de Andalucía, Spain) (PI‐0285‐2016). This project has received funding from the European Horizon´s Research and Innovation Program HORIZON‐HLTH‐2022‐STAYHLTH‐02 under agreement No 101095679. This study has been co‐funded by European Union. A.J.R.‐M. is supported by the Sara Borrell program from the ISCIII (Spain) (CD23/00117). J.P.T.‐O. is supported by the Rio Hortega program from the ISCIII (Spain) (CM23/00126). J.P.‐B is supported by the Juan Rodes program from the ISCIII (Spain) (JR25/00027). C.L.‐G. is supported by the Miguel Servet program from the ISCIII (Spain), and co‐funded by the European Union (CP22/00050). A.H.‐P. is supported by the Sara Borrell program from the ISCIII (Spain), and co‐funded by the European Union (CD23/00089). L.G.‐S. is supported by the Nicolas Monardes program from the Consejería de Salud de Andalucía (Spain) (C‐0028‐2018). G.M.‐C. is supported by a FPU PhD fellowship from the Spanish Ministry of Science, Innovation and Universities (FPU22/03868). E.G.‐F. is supported by the Nicolas Monardes program from the Consejería de Salud de Andalucía (Spain) (RC1‐0006‐2025). C.R.‐D. is supported by the Miguel Servet program from the ISCIII (Spain) (CP23/00088).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Figure: jev270240‐sup‐0001‐FigureS1.tiff

Ruiz‐Malagón, A. J. , Herraiz‐Vilela M., Pinazo‐Bandera J., et al. 2026. “An Exploratory Study on the Pathogenic Role of Faecal Extracellular Vesicles in Metabolic Dysfunction‐Associated Steatotic Liver Disease Progression and in Drug‐Induced Liver Injury.” Journal of Extracellular Vesicles 15, no. 3: e70240. 10.1002/jev2.70240

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.