Intra and inter-organ communication through extracellular vesicles in obesity: functional role of obesesomes and steatosomes

Abstract

Background

Extracellular vesicles (EVs) represent a sophisticated mechanism of intercellular communication that is implicated in health and disease. Specifically, the role of EVs in metabolic regulation and their implications in metabolic pathologies, such as obesity and its comorbidities, remain unclear.

Methods

Extracellular vesicles (EVs) were isolated through serial ultracentrifugation from murine adipocytes treated with palmitate or oleic acid, whole visceral and subcutaneous adipose tissue (obesesomes) of bariatric surgery obese donors, and human hepatocytes under steatosis (steatosomes) for functional in vitro experiments. Functional effects on inflammation and glucose and lipid metabolism of target cells (human and murine macrophages and hepatocytes) were assessed using ELISA, RT-PCR, and immunodetection. Isolated EVs from human steatotic (steatosomes) and control hepatocytes (hepatosomes) were characterized for quantity, size, and tetraspanin profile by NTA and Single Particle Interferometric Reflectance Imaging Sensor (SP-IRIS), and their protein cargo analyzed by qualitative (DDA) and quantitative (DIA-SWATH) proteomics using LC–MS/MS. Proteins identified by proteomics were validated by capturing EVs on functionalized chips by SP-IRIS.

Results and Conclusions

In this study, we investigated the role of EVs in the local communication between obese adipocytes and immune cells within adipose tissue, and the interaction of steatotic and healthy hepatocytes in the context of fatty liver disease progression. Furthermore, we analyzed obese adipose tissue-to-liver interactions through EV-obesesomes to elucidate their role in obesity-associated hepatic metabolic dysregulation. Our findings reveal that obesesomes promote inflammation and the secretion of pro-inflammatory cytokines upon interaction with macrophages, exerting a significant impact on reducing insulin resistance and altering lipid and glucose metabolism upon interaction with hepatocytes; in both cases, EVs from palmitate-loaded adipocytes and obesesomes from human visceral adipose depots demonstrated the most deleterious effect. Additionally, EVs secreted by steatotic hepatocytes (steatosomes) induced insulin resistance and altered lipid and glucose metabolism in healthy hepatocytes, suggesting their involvement in MASLD development. Proteomic analysis of steatosomes revealed that these vesicles contain liver disease-associated proteins, rendering them significant repositories of real-time biomarkers for the early stages and progression of MASLD.

Graphical Abstract

Introduction

Adipose tissue accumulation and inflammation during the development of obesity play important roles, not only at the metabolic level, as they are closely linked to dyslipidemia, diabetes, and cardiovascular disease, but also contribute to and accelerate other associated pathologies [1, 2]. Thus, the list of obesity-linked disorders or conditions exacerbated by obesity is gradually increasing, including those that are not initially related, such as osteoarthritis, gallbladder and kidney diseases, and various types of cancer. Being overweight or obese is among the primary risk factor for metabolic dysfunction-associated steatotic liver disease (MASLD), the most prevalent chronic liver disease worldwide [3]. The underlying molecular events in obese adipose tissue are related to abnormal hyperplasia and hypertrophy of adipocytes, leading to lipotoxicity and alterations in adipocytokines secreted by this endocrine tissue. These hypertrophy-associated alterations in adipocytes and their secretion patterns are responsible for the changes in lipid and carbohydrate metabolism, macrophage infiltration, and low-grade inflammation [4]. It has been profoundly described those obese-related alterations on known adipokines such as hormones, cytokines, extracellular matrix proteins and growth and vasoactive factors, that have revealed many pathological mechanisms and potential targets to fight obesity. In this context, new players have recently been implicated in non-classical protein secretion [5], such as long non-coding microRNAs [6], and extracellular vesicles [7], which are shed by all adipose tissue cell components, reflecting their physiological and pathological status.

Extracellular vesicles (EVs) are small, round-shaped membrane spheres of various sizes (30–1000 nm) liberated by virtually all cell types and can be found in the extracellular environment and in all body fluids, including blood, urine, saliva, cerebrospinal fluid, and tears. Depending on their biogenesis and size, they can be classified as microvesicles (100 nm-1 µm) derived by blebbing of the plasma membrane, exosomes (30–100 nm) compiled in multivesicular endosomes (MVEs) that are secreted by exocytosis, and larger vesicles (50–5000 nm) that comprise apoptotic bodies released by cells prior to apoptosis [8, 9]. EVs contain membrane and cytosolic components such as proteins, lipids, and RNAs, and this composition is conditioned by the site of biogenesis. Therefore, they have emerged as a good source of disease biomarkers, with potential use in monitoring pathology, including treatment effectiveness. In addition, because EVs are essential transporters of cellular information, owing to their bioactive cargo, they can alter the activity or properties of specific target cells and tissues. The contact of EVs with target cells induces cell transduction signals and genetic or epigenetic changes that have been demonstrated to promote stimulatory or inhibitory functional outcomes such as cell proliferation, apoptosis, cytokine production, immune modulation, and metastasis. Adiposomes shed by adipocytes, mesenchymal stem cells, immune cells, and endothelial cells have been implicated in adipokine release, browning of white adipocytes, adipogenesis, insulin resistance, modulation of the immune microenvironment, and tissue remodeling [10,11,12,13,14,15]. Moreover, EVs secreted from adipose tissue may exert long-distance effects on peripheral organs such as the liver, pancreas, muscle, and even the central nevous system (CNS), and are implicated in physiological homeostasis as well as in pathogenesis. Thus, adipocyte-shed vesicles can alter insulin sensitivity in hepatocytes and muscle cells [16], cell death, and dysfunction of human β-pancreatic cells [17], crossing the blood–brain barrier and affecting body energy intake by regulating the expression of arcuate pro-opiomelanocortin (POMC) through hypothalamic mTOR signaling [18].

In this study, we aimed to further expand our understanding of extracellular vesicles, specifically their role as mediators in an evolving intercellular communication network within adipose tissue that connects with other cells and tissues associated with obesity. Consequently, we investigated on extracellular vesicles, particularly obesesomes, isolated from adipocytes treated with palmitic and oleic acids, as well as from whole human obese adipose tissues, with specific attention to the anatomical origin of obesesomes (visceral and subcutaneous) in patients undergoing bariatric surgery. We performed functional assays to analyze the role of obesesomes in local inflammation through their interaction with macrophages, and at a distant level, by examining white adipose tissue-to-liver interaction. Furthermore, we established a human cell model of hepatocyte steatosis to obtain information on EVs released in the context of MASLD, termed steatosomes. Functional and protein cargo analyses of steatosomes compared to those of healthy vesicles, hepatosomes, were conducted.

Methods

Primary cells cultures and cell lines

The murine mesenchymal stem cell line, C3H10T1/2, was kindly donated by Prof. Eduardo Domínguez Medina (University of Santiago de Compostela, Spain) and cultured at 37 °C under 5% CO₂ in DMEM (4.5 g/L glucose, Corning, USA) containing 10% fetal bovine serum (FBS; Sigma-Aldrich, MO, USA) and 1% penicillin and streptomycin (P/S, Biowest) until differentiation into adipocytes, as previously described [19]. In brief, differentiation was performed after cell confluence with induction medium [maintenance medium supplemented with 1 µM dexamethasone, 0.5 mM isobutylmethylxanthine, 1 µM rosiglitazone (Sigma-Aldrich, MO, USA), and 5 µg/mL insulin (Actrapid, NovoNordisk)]. Two, four, and 6 days after induction, the medium was replaced with maintenance medium containing 5 µg/mL insulin (Actrapid, NovoNordisk). The accumulation of cytoplasmic triglycerides in these cells was detected by staining with Oil Red O (Sigma-Aldrich, MO, USA).

A high-glucose and high-insulin (HG/HI) insulin resistance, and lipid hypertrophy/insulin resistance cell models were established as described previously [11, 20]. Briefly, cells were differentiated until day 6, washed three times with PBS, and incubated for 2 h in low-glucose (1 g/L) and serum-free cell culture medium. Cells were then cultured in high glucose (4.5 g/L) and high insulin (100 nM), or with palmitate (500 µM in 2% fatty acid-free BSA [Capricorn Scientific, USA]) or oleic acid [1 mM, conjugated in 0.5 mM fatty acid-free BSA [Capricorn Scientific, USA]) in serum-free medium for 24 h in oleic and HG/HI, and 18 h in palmitate. For vesicle isolation, the treatment was prolonged to 48 h. For pAkt/Akt pathway analysis, cells were washed three times in PBS and stimulated with insulin (100 nM) for 10 min.

The murine macrophages Raw 264.7 cells were cultured in DMEM (4.5 g/L glucose, Lonza) containing 10% fetal bovine serum (Sigma-Aldrich, MO, USA) and 1% P/S (Biowest) at 37 °C under 5% CO₂.

Human macrophage THP-1 cells were cultured in RPMI (without L-Glutamine, Lonza) containing 10% FBS (Sigma-Aldrich, MO, USA), 1% P/S (Biowest) and 0.05 mM β-mercaptoethanol at 37 °C under 5% CO₂. THP-1 monocytes were differentiated into macrophages by adding 5 µg/mL Phorbol Myristate Acetate (PMA) for 24 h. For RNA extraction, the treatment was performed for 24 h, whereas for the isolation of vesicles in the cell secretome, the treatment was prolonged to 48 h.

The human hepatocytes HepaRG cells were cultured in William’s E medium supplemented with GlutaMAX (Gibco) containing 10% fetal bovine serum (FBS; Sigma-Aldrich, MO, USA) and 1% P/S (Biowest) at 37 °C under 5% CO₂. However, before differentiation, HepaRG cells were cultured in proliferation medium [maintenance medium with 5 µg/mL insulin (Actrapid, NovoNordisk) and 0.5 µM hydrocortisone hemisuccinate sodium salt (Sigma-Aldrich, MO, USA) for 6 days, renewing the medium every two days. HepaRG cells were differentiated using differentiation medium [maintenance medium with 5 µg/mL insulin (Actrapid, NovoNordisk), 50 µM hydrocortisone hemisuccinate sodium salt, and 2% dimethyl sulfoxide (DMSO, Thermo Fisher Scientific, Massachusetts, USA)] for 15 days, renewing the medium for the first three days and then every three days. Once differentiated, HepaRG cells were washed twice with PBS and serodeprived for 2 h. The cells were then cultured with high glucose (4.5 g/L) and high insulin (100 nM), with palmitate [500 µM in 2% fatty acid-free BSA (Capricorn Scientific, USA)], oleic acid [1 mM, conjugated in 0.5 mM fatty acid-free BSA (Capricorn Scientific, USA)], or with combination treatment [glucose (4.5 g/L) and insulin (100 nM), 250 µM palmitate (Sigma-Aldrich, MO, USA) in 2% fatty acid-free BSA (Capricorn Scientific, USA) and 0.5 mM oleic acid (Sigma-Aldrich, MO, USA), conjugated in 0.5 mM fatty acid-free BSA (Capricorn Scientific)]. For RNA or protein extraction, the treatment was performed for 24 h, whereas for the isolation of vesicles from the cell secretome, the treatment was prolonged to 48 h.

Murine hepatocyte primary cell culture was obtained by hepatic perfusion of 14-week-old male C57BL/6 mice weighing 20 g following previously described protocols [21, 22]. Hepatocytes from the perfusion were resuspended in 100 mL of complete medium [DMEM (Gibco) with 10% FBS (Lonza) and 1% P/S (Gibco)], divided into 2 × 50 mL tubes, and centrifuged at 400 rpm (SorvallTM ST16, Fisher Scientific) for 4 min. The pellet was resuspended in 30 mL complete medium and 15 mL 90% Percoll (GE Healthcare) and centrifuged at 500 rpm for 10 min. Subsequently, the cell pellet from both tubes was transferred to a single tube, resuspended in complete medium, and centrifuged three times at 500 rpm for 5 min 3 times. Finally, the pellet of hepatocytes was resuspended in 50 mL of complete medium to analyze cell viability by Trypan Blue staining on slides (Cell Counting Chamber Slides, Invitrogen) for cell counting using the Countess® equipment (Invitrogen). Once the viability of the hepatocytes was obtained, 250.000 cells were seeded in 9.2 cm² plates (Corning, Thermo Fisher Scientific, Massachusetts, USA) previously treated with collagen (Sigma-Aldrich, MO, USA).

Human adipose tissue explants and secretome collection for EVs isolation

Human visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) were obtained from 15 consecutive obese and morbidly obese patients (clinical data in Supplementary Table 1) at the Obesity Treatment Unit of the Hospital Clínico Universitario de Santiago, Spain, who underwent laparoscopic gastrectomy (body mass index [BMI] > 35; average body mass index [BMI]: 51) after obtaining written informed consent, in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki). All procedures were approved by the Clinical Ethical Committee of Galicia (CEIC), Spain, under code number 2020/497. Visceral fat (VAT) was excised from the hypogastric region around the internal organs and subcutaneous fat (SAT) was excised from the mesogastric region. The tissues were transported from the operating room to the laboratory in sterile PBS containing penicillin (100 U/ml) and streptomycin (100 μg/ml) for further processing. Secretome collection was performed for adipose tissue explants, as previously described [11, 13, 14]. Briefly, tissue explants were processed to remove any contaminants by performing an intensive wash in PBS, which was repeated several times. Tissue pieces were transferred to a tube containing 25 mL PBS and centrifuged for 5 min at 1800 rpm at room temperature to eliminate red blood cells and debris. One gram (n = 2 replicates) of each tissue type (VAT or SAT) from independent patients was incubated at 37 °C and 5% CO₂ in 5 mL/tissue piece/well in serum-free DMEM without phenol red (Sigma-Aldrich), supplemented with 1% (v/v) penicillin–streptomycin. The medium was refreshed after 1.5 and 24 h. Following the last wash (24 h), all dishes received fresh DMEM (3 mL/1 g tissue) and were cultured for an additional 48 h to allow vesicle secretion. The media were then collected, centrifuged for 5 min at 1800 rpm, and stored at -80 °C until further analysis.

Extracellular vesicles isolation

Secretomes of human adipose tissue explants, cultured adipocytes, and hepatocytes were obtained, as previously described [11]. Briefly, cells were treated with the corresponding control or treatment medium, and the resulting secretome was collected and centrifuged at 1800 rpm for 5 min and then filtered through a 0.22 µm filter to remove contaminating cell debris. These secretomes were stored at − 80 °C until ultracentrifugation.

EVs isolation was performed according to our own standardized protocols following previous publications [11, 13, 14]. The secretomes were centrifuged in a Beckman Coulter Optima L-100XP at 10,000 g at 4 °C for 20 min, followed by ultracentrifugation at 100,000 × g at 4 °C for 90 min with a Type SW 41Ti rotor and 13.2 mL tubes (Thinwall Polypropylene Tubes, 14 × 89 mm; Beckman Coulter), with acceleration and deceleration brake profile 9, to pellet vesicles. The supernatant was carefully removed, and the vesicle-containing pellets were resuspended in ice-cold PBS. A second round of ultracentrifugation (100,000 g at 4 °C for 90 min) was performed. The final pellet containing the isolated vesicles was resuspended in ice-cold PBS or RIPA sample buffer [200 mM Tris/HCl (pH 7.4), 130 mM NaCl, 10% (v/v) glycerol, 0.1% (v/v) SDS, 1% (v/v) Triton X-100, and 10 mM MgCl₂] according to the analysis.

Nanoparticle tracking analysis (NTA)

Vesicle size and concentration distribution were determined by Nanoparticle Tracking Analysis (NTA) using a Malvern NanoSight NS300 equipment (v3.3 Malvern Panalytical, Ltd, UK) according to the manufacturer’s instructions. Briefly, samples were vortexed and diluted to a final dilution of 1:100 in Milli-Q H₂O. Blank filtered H₂O was used as a negative control. Each sample analysis was conducted for 60 s and measured five times using Nanosight automatic analysis settings. The detection threshold was set to level 4 and the camera level was set to 14.

SP-IRIS analysis

Hepatosomes and steatosomes were characterized using the SP-IRIS ExoView R200 platform (now Leprechaun, Unchained Labs, USA) [23] and human tetraspanin Flex kits (NAV251-1047, Unchained Labs, USA). The chips were prescanned using a protocol to identify any previously adhered particles during manufacturing. For incubation, the chips were placed in specific plates to avoid contact between the chip corners and sides of the well. The secretomes were diluted 1:2 in the incubation solution and 50 μL of the diluted sample was incubated overnight without agitation. Several washes were performed on the day following the manufacturer´s protocol, using a chip washer (Unchained Labs, USA). The chips were then incubated with the antibodies. A first analysis was started to see the tetraspanin profile of the different samples by adding an antibody cocktail to each chip using 300 μL of blocking buffer per chip and 0.6 μL of antibodies fluorescently labelled for 1 hour with gentle agitation: anti-CD9 (CF 488A), anti-CD81 (CF 555A) and anti-CD63 (CF 647A) provided by the kit. Subsequently, analysis was performed to detect biomarkers using 300 μL of blocking buffer per chip and 0.6 µL of antibodies (in case of surface proteins) or 1.5 μL of antibodies (in case of cargo proteins) for one hour with gentle agitation. The biomarker antibodies used were anti-clusterin purchased from Santa Cruz Biotechnology (CA, USA) and fluorescently labelled with DyLight 755 Conjugation Kit; anti-HMG-1 and anti-osteopontin from Novus Biologicals (Novus Bio, CO, USA) and fluorescently labelled with Alexa Fluor 488 Conjugation Kit and Alexa Fluor 647 Conjugation Kit, respectively, and anti-vimentin from Invitrogen (Thermo Fisher Scientific, Massachusetts, USA) fluorescently labelled with Alexa Fluor 647 Conjugation Kit. For the detection of intravesicular proteins, captured EVs were treated with cargo solution kit (Unchained Laboratories) according to the manufacturer’s instructions. The chips were washed and dried before being transferred to the scanner for interferometric and fluorescent imaging (Exoview R200 scanner; scanner software 3.2.2). For data analysis, fluorescent cut-offs were modified from 200 a.u. [for red, far-red and green channels] and from 400 a.u. [for blue channel] to limit the number of detected particles on MIgG to 50 events [for red, far-red, and green] or 100 events [for blue] following the technical instructions of the manufacturer using the ExoView Analyzer v3.2 (Unchained Laboratories).

Functional EVs assays

Inflammation assays

For inflammation experiments, murine macrophages (Raw 264.7) were cultured for 24 h in the presence of EVs isolated from differentiated control C3H10T1/2 adipocytes and HG/HI, palmitate, and oleic cell models (obesesomes) at physiological concentrations of approximately 3E+09 to 3E+10 particles/mL. Thus, following the Minimal Information for Studies of Extracellular Vesicles MISEV23 recommendations by the International Society for Extracellular Vesicles (ISEV) [24], we established a 1:1 ratio; vesicles from a determined number of cells in culture were added to the same number of target cells. Subsequently, the cell culture was collected and assayed using a mouse inflammation antibody membrane array with 40 targets (ab133999, Abcam), following the manufacturer’s instructions. The negative control was PBS vehicle, and treatment with 1 µg/mL LPS (lipopolysaccharide from Escherichia coli O111:B4; Sigma-Aldrich) was used as a positive control for inflammation. For human THP-1 macrophages, each 9.5 cm² well of differentiated cells was treated with EVs secreted from 125 mg of human visceral or subcutaneous adipose tissue explants. After 24 h, the RNA was extracted for real-time PCR analysis.

Liver crosstalk

On the one side, EVs from steatotic model of HepaRG cells treated with combination of HG/HI, palmitate, and oleic acids (steatosomes) were used to treat healthy HepaRG cells and on the other side, steatotic HepaRG cells were treated with EVs from healthy HepaRG cells (hepatosomes) following the 1:1 ratio described above. In both cases, 5 µL of EVs (physiological concentration of approximately 3E+09 to 3E+10 particles/mL) was used to treat 3.8 cm² wells, and 5 μL of PBS was used as a vehicle in the control wells. After 24 h, proteins were extracted for immunoblotting assays, or RNA was extracted for real-time PCR analysis.

White adipose tissue to liver crosstalk

5 μL of EVs (physiological concentration of approximately 3E+09 to 3E+10 particles/mL) from adipocyte cell models (obesesomes) described above were used to treat murine primary hepatocytes (9.5 cm² wells), as well as 5 μL of PBS as a vehicle in control wells. After 24 h, RNA was extracted for real-time PCR analysis, and insulin resistance was assayed by insulin stimulation (100 mM, 5 min) prior to 2 h of serodeprivation, before protein extraction. For experiments with human hepatocytes, 5 μL of obesesomes (physiological concentration of approximately 3E+09 to 3E+10 particles/mL) was used to treat healthy and steatotic human hepatocytes (HepaRG), and 5 μL of PBS was used as the vehicle in the corresponding wells. After 24 h of incubation, the RNA was extracted for real-time PCR analysis.

Immunobloting

Protein extracts were prepared as previously described [11]. Briefly, protein samples were extracted by homogenization using TissueLyser II (QIAGEN, Tokyo, Japan) in cold RIPA buffer containing anti-proteases and anti-phosphatases (Sigma-Aldrich, St Louis, MO, USA). Thirty micrograms of cell lysates from at least three independent experiments were separated on 10% SDS-PAGE gels and electroblotted onto nitrocellulose membranes, as previously described [13, 21]. Primary antibodies against pAkt, anti-Akt, anti-pS6, anti-S6 and anti-pACC antibodies were purchased from Cell Signaling Technology (Danvers, MA, USA). Data are expressed as percentages corrected for anti-GAPDH (Invitrogen, Thermo Fisher Scientific, Massachusetts, USA) (arbitrary units) in western blots with mean ± SEM.

Quantitative real-time PCR

RNA from macrophages and the human hepatocyte cell line was isolated using TRIzol (TRI Reagent, Sigma-Aldrich, MO, USA), in which chloroform (Sigma-Aldrich, MO, USA) was used as the denaturalization agent for the cells, and isopropanol (Sigma-Aldrich, MO, USA) was used as the RNA precipitating agent. RNA was extracted from murine primary hepatocytes using the RNeasy mini-50-column kit (Qiagen, Germany) according to the manufacturer's recommendations. The precipitate was diluted in DEPC water (Ambion) containing 1% SUPERase RNase Inhibitor (Thermo Fisher Scientific, Waltham, Massachusetts, USA). One microgram of RNA was treated with DNase I (Invitrogen, Thermo Fisher Scientific, Massachusetts, USA) prior to reverse transcription using qScript Super cDNA SuperMix (QuantaBio, USA). The cDNA obtained was used for gene expression analysis, which was performed using SYBER Select Master Mix (ROX) (Applied Biosystems, Massachusetts, USA) in QuantStudio 3 (Applied Biosystems). The PCR amplification conditions were as follows: denaturation step with 2 min at 95  ºC; 40 cycles of 15 s at 95 °C, 1 min at 60 °C, followed by a cooling step. Each sample was analyzed in triplicate for each experiment. Primers used for murine samples:

PRIMER	SEQUENCE
FAS	F: 5′—GCCACCTCAGTCCTGCTATC-3′ R: 5′—GGTATAGACGACGGGCACAG-3′
PPARα	F: 5′—GCAGTGCCCTGAACATCGAG-3′ R: 5′—GCCCTTACAGCCTTCACATGC-3′
PPARγ	F: 5′—ATCATCTACACGATGCTGGCC-3′ R: 5′—CTCCCTGGTCATGAATCCTTG-3′
Perilipin-3	F: 5′—CGCCTATGAACACTCCCTCG-3′ R: 5′—CGAGCACACTTGTTAAGCTGC-3′
PDK1	F: 5′—CCACTGAGGAAGATCGACAGAC-3′ R: 5′—AGAGGCGTGATATGGGCAATCC-3′
GLUT1	F: 5′—GTCCGATTGCCAGCTAATGC-3′ R: 5′—CACAGGCAGAAATCGCCAAC-3′
IL-6	F: 5′—CCCAACAGACCTGTCTACCA-3′ R: 5′—CAGAATTGCCATTGCACAAC-3′
TNFα	F: 5′—CCACCACGCTCTTCTGTCTA-3′ R: 5′—AGGGTCTGGCCATAGAACT-3′
HPRT	F: 5′—CAATGCAAACTTTGCTTTCCC-3′ R: 5′—TCCTTTTCACCAGCAAGCTTG-3′

The following primers were used for the human samples:

PRIMER	SEQUENCE
FAS	F: 5′—CCAGGACAGCCTGCTAGGTA-3′ R: 5′—GAGTGGATGAGCAGCGTCT-3′
PPARα	F:5′—TGAAGGCTGCAAGGGCTTCTT-3′ R: 5′—CAGATCTTGGCATTCGTCCAAA-3′
PPARγ	F: 5′—AAAGAAGCCAACACTAAACC-3′ R: 5′—TGGTCATTTCGTTAAAGGC-3′
PLIN3	F: 5′—GATCAGAAGCTGGTGGAAGG-3′ R: 5′—ACCTGGTCCTTCACATTGGT-3′
GLUT1	F: 5′—CTTGGCTCCCTGCAGTTTGG-3′ R: 5′—AACGAAAAGGCCCACAGAGA-3′
HPRT	F: 5′—AGAAGTTTTGTTCTGTCCTGGAA-3′ R: 5′—GGGAACTGCTGACAAAGATTCAC-3′

Proteomic analysis by tripleTOF 6600 LC–MS/MS system

To perform global protein identification and quantification, an equal number of cells and their corresponding secretome was used to isolate EVs from control and steatotic hepatocytes (n = 4 independent EVs isolation experiments: 3 × 6 multiwell plates = 36 mL/sample) and processed as previously described [11, 13]. The digested peptides were separated using reverse-phase chromatography. The gradient was created using a micro liquid chromatography system (Eksigent Technologies nanoLC 400, SCIEX, Foster City, CA) coupled with high-speed Triple TOF 6600 mass spectrometers (SCIEX) with a micro flow source, as described in the supplementary methods, using a data-dependent acquisition method (DDA). For this analysis, only proteins with FDR < 1% (99% protein confidence) were selected. For relative quantification using the SWATH data-independent acquisition (DIA) method, we first built a spectral library grouping each cell type in a pool (control and steatotic hepatocyte lysates). Each sample and its technical replicates (four samples from each vesicle type/run per triplicate) were run in TripleTOF 6600 using the SWATH-MS acquisition method, as detailed in Supplementary Methods.

After MS/MS analysis (MS2 data), data files were processed using ProteinPilotTM 5.0.1 software from Sciex, which uses the algorithm ParagonTM for database search and ProgroupTM for data grouping. Data were searched using a human-specific Uniprot database specifying iodoacetamide at cysteine alkylation as a variable modification and methionine oxidation as a fixed modification. The false discovery rate was determined using a nonlinear fitting method, displaying only the results that reported a 1% or better global false discovery rate [25].

Protein functional analysis

Functional analysis was performed using FunRich open access software (Functional Enrichment analysis tool version 3.1.3) for functional enrichment and interaction network analysis [http://www.funrich.org/] [26], Reactome [https://reactome.org/] [27], and Metascape [http://metascape.org] [28].

Statistical analysis

Statistical significance among multiple groups was analyzed using GraphPad Prism 10.0.3 software by two-way ANOVA followed by Dunnett's multiple comparisons test, parametric one-way ANOVA followed by Dunnett's multiple comparisons test, or non-parametric one-way Anova-Kruskal Wallis followed by Dunn’s multiple comparison test. The Mann–Whitney U test was used to compare results between the two groups. Normal distribution of data was assayed using the Shapiro–Wilk test. Statistical significance was set at P ≤ 0.05.

Results

Extracellular vesicles from obese adipose tissue exert inflammatory paracrine functions on macrophages

To investigate the role of extracellular vesicles in local communication between obese adipocytes and immune cells within adipose tissue, we exposed murine macrophages to previously well-characterized EVs [11] from control (adiposomes) and pathological (obesesomes) adipocytes. Although we previously demonstrated that obesesomes from lipid-hypertrophied adipocytes significantly increased IL-6 and TNF-α expression in macrophages [11], the capacity of these vesicles to induce inflammatory cytokine secretion remains unexplored. Therefore, EVs isolated from well-established cell models of control differentiated murine, lipid-treated/insulin-resistant [palmitate and (500 μM/L)/oleic acid (1 mM)] adipocytes were used to treat macrophages at physiological concentrations for 24 h (Fig. 1A). The resulting macrophage-conditioned cell culture medium was collected to assay for 40 cytokines using commercial antibody arrays. Of these, 16 cytokines showed significant changes in secretion (Fig. 1B). As anticipated, following our previous mARN expression assays, TNF-α and IL-6 secretion demonstrated a significant increase in the cell culture medium of macrophages treated with EVs from palmitate- and oleic acid-hypertrophied adipocytes (Fig. 1B). Furthermore, the levels of other cytokines were significantly elevated in the cell culture medium of macrophages stimulated with pathological vesicles, particularly in macrophages treated with obesesomes isolated from palmitate-hypertrophied adipocytes, including CXXL13, CCL11, GCS-F, IFN, IL-6, IL-10, CXCL11, lymphotactin, MIF, TCA-3, TECK, CCL24, SDF-1, and TIMP1/2 (Fig. 1B).

Fig. 1

Obesesomes induce macrophage inflammation. EVs isolated from control and pathological adipocytes [high glucose/high insulin (HG/HI), palmitate (PALM) and oleic acid (OLEIC)] were used to treat Raw 264.7 murine macrophages during 24 h (A); cell conditioned medium was collected for assaying 40 cytokine array. Significant changes in fold change towards macrophages treated with control vesicles are represented [densitometry arbitrary units of n = 3 independent experiments] (B). Human obesesomes isolated from obese subcutaneous (SAT) and visceral (VAT) fat explants were used to treat THP-1 human macrophages for 24 h (C); BMI [kg/m²], metabolic syndrome index [TG/HDL-C], type 2 diabetes [HbA1c], and inflammation [CRP] status of adipose tissue donors used for functional assays is shown (D); and relative TNF-α expression in THP-1 represented (E). Differences were analyzed using the Two-Way Anova test, followed by Dunnett's multiple comparisons test (P ≤ 0.05, considered statistically significant: * P < 0.05, ** P < 0.01, *** P < 0.001, and ****P < 0.0001); detailed clinical data of donors are shown in Supplementary Fig. 1, and graphs row data in Supplementary file 9. Figure created using Biorender (https://biorender.com/)

Following the demonstration of EV-mediated macrophage inflammation in murine cell models in vitro  which corroborated previous evidence, we extended the analysis to unexplored vesicles secreted by human obese adipose tissue. Subsequently, obesesomes from obese human visceral and subcutaneous whole-tissue explants of independent obese and morbidly obese patients, selected for bariatric surgery, were isolated and used to treat human THP-1 macrophages (Fig. 1C). The analysis of clinical data from donor patients revealed an average body mass index of 51 kg/m² (Fig. 1D), with nearly 50% exhibiting an elevated metabolic syndrome index [TG/HDL-C Index > 2.2] (Fig. 1E). Furthermore, 60% of the patients had type 2 diabetes, characterized by elevated levels of HbA1c > 5.6 (Fig. 1F), and systemic inflammation, indicated by abnormal CRP ultra values > 0.5 mg/dL (Fig. 1G). Comprehensive clinical data of the donor patients are presented in Supplementary Table 1. The functional interaction assay demonstrated that EVs secreted by both obese fat depots induced TNF-α expression in macrophages, with a particularly significant effect observed when treated with obesesomes of visceral origin (Fig. 1H). Additionally, a moderate correlation between TNF-α expression and the systemic inflammation level [CRP ultra] of obese AT donors was observed (r = 0.42 for obese SAT-EVs; r = 0.30 for obese VAT EVs; Supplementary Fig. 1). Moreover, a moderate positive correlation was identified between TNF-α expression levels in THP-1 cells and BMI [ranging from 33 to 64 kg/m²; Suppl. Table 1] for obesesome donors [r = 0.41 for EVs from visceral adipose tissue], as illustrated in Supplementary Fig. 1A.

Consequently, the secretion of obesesomes by adipose tissue in obese individuals, particularly from visceral harmful deposits that contribute to obesity-related complications, may exacerbate the inflammatory state locally, and probably at systemic level.

Steatotic hepatocytes secrete extracellular vesicles, steatosomes, that are enriched with liver disease-associated proteins that may contribute to the progression of fatty liver disease

To elucidate the role of extracellular vesicles (EVs) in the communication between steatotic hepatocytes and adjacent healthy hepatocytes within the context of fatty liver disease progression, we isolated EV-hepatosomes from healthy human HepaRG hepatocytes and pathological steatosomes from an in vitro lipid-hypertrophied human hepatocyte cell model (Fig. 2A–D). Initially, we optimized the steatosis hepatocyte cell model by subjecting cells to high glucose/high insulin (HG/HI), oleic acid, palmitate, or a combination of these three treatments. The combination of lipids with high glucose and insulin was found to exert an optimal effect on hepatocytes, resulting in lipid accumulation (Fig. 2A, B), and insulin resistance, as evidenced by a significant reduction in pACC (Ser79) and pAkt levels, respectively (Fig. 2C, D). EVs from these cells (steatosomes) and healthy control cells (hepatosomes) were isolated for characterization and functional assays (Fig. 2E–K; Fig. 3). Nanotracking analysis was conducted on vesicles isolated from hepatocytes treated with HG/HI, palmitate, oleic acid, or a combination of the three (all data are presented in Supplementary Data 9 and Supplementary Fig. 2). Notably, steatotic hepatocytes exhibited a significantly higher concentration of secreted vesicles (steatosomes) than healthy hepatocytes, with no observed variation in particle size (Fig. 2F-H). Interferometry and tetraspanin profiling of isolated vesicles using SP-IRIS were used to analyze hepatosomes and steatosomes, revealing the predominant capture of CD63-positive EVs, followed by CD9 and CD81 EVs (Fig. 2I). The detection of fluorescently labeled anti-CD81 [green], CD63 [red], and CD9 [blue] after immunocapture facilitated the identification of combinations of the three tetraspanin colocalizations (Fig. 2J, K). Steatosomes, in comparison with hepatosomes, were characterized by a strong signal for CD63/CD9-positive EVs in CD81-captured vesicles (Fig. 2 K). Vesicles captured by CD63 exhibited all combinations of colocalization, particularly in steatosomes. Conversely, in CD9-captured vesicles, a greater number of double CD63/CD9 labeling was observed in steatosomes than in hepatosomes and in all captured vesicles in CD81 and CD63 wells. Reduced detection of CD63 was observed in steatosomes compared to hepatosomes using the three captured antibodies: CD81, CD63, and CD9 (Fig. 2K).

Fig. 2

Isolation and characterization of hepatosomes and steatosomes. Human hepatocytes were treated with [high glucose/high insulin (HG/HI), palmitate (PALM), oleic acid (OLEIC), or a combination of the three (COMBI)] to establish a steatosis cell model. Representative images of Oil Red O staining used to assess the triglyceride content of cultured hepatocytes (A) and their quantification (B) are shown (n ≥ 4 independent experiments). Western blotting and densitometry (n = 4 independent experiments) of pACC protein levels (C) and insulin sensitivity by immunodetection of p-Akt/Akt after insulin stimulation [100 mM, 10 min; n = 6] are shown (D). NTA quantification [particles/mL] (F) and size distribution [nm] (G–H) of isolated extracellular vesicles (n = 3 independent experiments) secreted from heathy (hepatosomes) and from steatotic hepatocytes (COMBI-steatosomes), and tetraspanin profile characterization by SP-IRIS [ExoView R200] is represented indicating interferometric measurement of particles/mL of immuno-captured particles by CD81, CD63 and CD9 (I), representative fluorescent mode images of captured particles for each tetraspanin (red: CD81, green: CD63, and blue: CD9) and colocalization data for each sample is represented in pies (K). Differences were analyzed using One-way Anova Kruskal–Wallis test, followed by Dunn´s multiple comparisons test, One way-Anova followed by Dunnett´s multiple comparisons test, and comparisons between two groups were performed using the Mann–Whitney U test (P ≤ 0.05, considered statistically significant: * P < 0.05, ** P < 0.01, *** P < 0.001, and ****P < 0.0001). Detailed raw data and complete immunoblot images are provided in Supplementary File 9. Figure created using Biorender (https://biorender.com/)

Fig. 3

Steatosomes alter glucose and lipid metabolism in healthy hepatocytes. Functional assays were performed by incubating steatotic hepatocytes with healthy hepatosomes and healthy hepatocytes with pathological steatosomes [secreted by cells treated with COMBI] for 24 h were performed (A). The effect of EVs was tested using an insulin sensitivity assay by quantifying pAkt/Akt; representative images of n = 6 independent experiments are shown (B, C). Glucose (Glut1) and lipid (FAS, PPARα, PPARγ, and PLIN3) metabolism-regulating gene expression compared to non-treated cells by real-time PCR of n = 6 independent experiments are shown (D–H). Differences were analyzed using the one-way ANOVA Kruskal–Wallis test, followed by Dunn´s multiple comparisons test, or by One way-Anova followed by Dunnett´s multiple comparisons test (P ≤ 0.05, considered statistically significant: * P < 0.05, ** P < 0.01, and *** P < 0.001). Detailed raw data and complete immunoblot images are provided in Supplementary File 9. Figure created using Biorender (https://biorender.com/)

Characterized hepatosomes and steatosomes were utilized to analyze local intercellular crosstalk, as illustrated in Fig. 3A. Healthy hepatosomes were used to treat steatotic hepatocytes, whereas steatosomes were used to treat healthy cells, and their effects on insulin sensitivity were subsequently evaluated (Fig. 3B-G). The results demonstrated that no beneficial effect was observed on the insulin sensitivity of steatotic hepatocytes treated with hepatosomes (Fig. 3B); however, steatosomes significantly increased insulin resistance in healthy hepatocytes (Fig. 3C). Considering the effect of steatosomes on decreasing insulin sensitivity of target healthy hepatocytes, both healthy and steatotic hepatocytes treated with hepatosomes and steatosomes were examined for the expression of genes implicated in glucose (Fig. 3D), and lipid metabolism (Fig. 3E-H). Notably, hepatosomes significantly induced the expression of Glut 1 in steatotic hepatocytes (Fig. 3D). Regarding lipid metabolism, a pronounced effect of steatosomes was observed when treating healthy hepatocytes as they significantly reduced perilipin 3 expression in these cells. Moreover, healthy hepatosomes appeared to counteract slightly (not significantly) the diminution of perilipin 3 in steatotic hepatocytes (Fig. 3E). Additionally, a significant decrease in FAS expression was detected in steatotic hepatocytes, but no effect was observed in healthy cells treated with pathological vesicles (steatosomes), nor was there any beneficial effect of hepatosomes on steatotic hepatocytes (Fig. 3F). In the case of PPARα, a decrease similar to that observed in steatotic cells was observed in healthy hepatocytes treated with steatosomes (Fig. 3G). The same effect was observed when PPARγ expression was analyzed (Fig. 3H). In both cases, healthy hepatosomes did not counteract the deleterious effects of lipid hypertrophy (Fig. 3G, H).

Based on their former functional role, proteomic analysis of these hepatocyte-shed vesicles was conducted to elucidate their protein content and potential role in propagating pathology, including the identification of potential early biomarkers of MASLD. Proteomic DDA and quantitative DIA-SWATH analyses of hepatosomes and steatosomes secreted by cells treated with a combination of HG/HI, palmitate, and oleic acid were performed (Fig. 4A–F). For a more comprehensive analysis, vesicles isolated from hepatocytes treated with HG/HI, oleic acid, or palmitate were independently examined (Supplementary Tables 2–7). Qualitative DDA revealed that hepatocyte steatosis was reflected in the secreted EVs protein cargo. Consequently, only 50% of the proteome was common to steatosomes and hepatosomes (Fig. 4A). Functional analysis of the exclusive proteins in steatosomes revealed an increase in proteins associated with inflammatory responses and their regulation (Fig. 4A). Furthermore, the enrichment analysis of DDA data from each vesicle type represented as a heatmap (Fig. 4B) revealed alterations in pathways related to the mediation of cellular adaptation to redox stress and amino acid metabolism (Fig. 4B).

Fig. 4

Proteomic analysis of steatosomes reveals potential MASLD biomarkers. Hepatosomes and steatosomes (n = 4 independent isolation experiments) were analyzed by mass spectrometry using qualitative DDA and quantitative DIA-SWATH analyses. Descriptive and comparative Venn diagrams showing the total number of proteins identified in DDA with FDR < 1% (99% protein confidence), and functional enrichment analysis [FunRich: Biological process] of specific proteins for hepatosomes and steatosomes are shown A. Enrichment analysis [Metascape gene annotation analysis] of the DDA data from the protein content in each vesicle type is represented as a heat map (B). PCA analysis of transformed SWATH areas for the quantitative comparison of hepatosome and steatosome samples (C), and quantitative DIA analysis of proteins identified in steatosomes compared to those in hepatosomes is represented as a volcano plot, x-axis = log2 (fold change), y-axis = log (p-value), indicating significance by dotted line (D). The horizontal dotted line shows the chosen p-value for selecting regulated proteins. Enrichment analysis [Metascape gene annotation analysis] of DIA data from the protein content in each vesicle type represented as a heat map (E), and biomarkers of interest are shown in schematic drawings indicating the elevated proteins in orange and those diminished in gray (F). Validation of selected biomarkers by SP-IRIS is shown as fluorescent mode colocalization (G) and particle counts for osteopontin (H), HMG-1 (I), vimentin (J), and clusterin (K). Differences were analyzed using the Mann–Whitney U test (P ≤ 0.05, considered statistically significant: *** P < 0.001, and ****P < 0.0001). DDA Data-Dependent Acquisition, SWATH Sequential Window Acquisition of All Theoretical Mass Spectra, DIA Data-Independent Acquisition, SP-IRIS Single Particle Interferometric Reflectance Imaging Sensor, HMG High Mobility Group, ICAM Intercellular adhesion molecule, UBA Ubiquitin-Associated, CAND Cullin-associated and neddylation-dissociated. figure created using biorender (https://biorender.com/)

Differential quantitative analysis by DIA-Sequential Window Acquisition of all Theoretical fragment-ion spectra (SWATH) label-free proteomics was conducted to identify candidate biomarker proteins present in hepatic pathological EVs. Principal component analysis (PCA) of the differentially expressed proteins between hepatosomes and steatosomes demonstrated a distinct separation of these vesicles based on their cell of origin, indicating that EV protein cargo reflects hepatocyte metabolic alterations (Fig. 4C). Differential expression analysis of proteins with a fold change ≥ 1.5, and a p-value ≥ 0.05, facilitated the identification of proteins that were exclusively (8 proteins), elevated (176 proteins), or diminished (82 proteins) in steatosomes compared to hepatosomes (Supplementary Table 7). Additionally, differences between vesicles from hepatocytes subjected to individual treatments (palmitate and oleic acid) and control hepatosomes were analyzed (Supplementary Table 7). Quantitative analysis of the proteins identified in steatosomes compared with those in hepatosomes is represented as a volcano plot, indicating significant findings and highlighting the proteins of interest (Fig. 4D). Cluster plot analysis of proteins with a fold change ≥ 1.5, and a p-value ≤ 0.05, elucidated the presence of proteins implicated in the cellular response to stress, cell adhesion, extracellular matrix, and immune signaling, among others, in steatosomes (Fig. 4E). Considering the proteins elevated in steatosomes or vesicles secreted by hepatocytes hypertrophied with palmitate or oleic acid, we selected five upregulated proteins common to all pathological vesicles: Golgi reassembly stacking protein 2 (GORS2) associated with endoplasmic reticulum stress, hydroxyacyl-coenzyme A dehydrogenase (HCDH) and sphingosine-1-phosphatase 1 (SPP1) associated with β-oxidation of fatty acids (FunRich: Biological pathway) and hepatic necrosis and steatosis (Funrich: clinical phenotype), protein synthesis-associated eukaryotic translation initiation factor 2 (EIF2S3), and small ribosomal subunit protein S8 (RPS8) (Supplementary Data 7). In addition, 66 upregulated proteins were identified in at least two types of pathological vesicles, including ceruloplasmin and osteopontin (Supplementary Fig. 3; Supplementary data 7). Conversely, six downregulated proteins were common to the three pathological EVs, and 36 downregulated proteins were common to at least two types of pathological EVs, including those implicated in the c-MYC pathway, signaling events mediated by Akt, EGF receptor, and PDGFR-beta signaling, and the insulin pathway (Supplementary Fig. 3; Supplementary data 7). In the absence of comparable studies, all proteins identified by DDA and DIA were compared with previous proteomic data on EVs from primary rat hepatocytes (healthy lean and obese) [22, 29, 30], whole NAFLD plasma proteome databases [31], and proposed EV surface protein biomarkers for NAFLD based on the integration of various sources of information [32] (Supplementary Fig. 4).

Spectral count analysis of hepatosomes and steatosomes was conducted, corroborating several differences identified in SWATH, including vimentin, S100A8, and S100A9 (Supplementary Fig. 5; Supplementary data 7). Based on quantitative analysis of pathological EVs, we selected 16 proteins previously associated with liver diseases, including fibrosis, steatosis, steatohepatitis, MASLD, and hepatocellular carcinoma (Table 1): high-mobility group protein (HMG-1), alkaline phosphatase (PPB1), osteopontin (OSTP), 4F2 cell-surface antigen heavy chain (4F2), nucleophosmin (NPM), protein S100-A8 (S10A8), S100-A9 (S10A9), S100-A11(S10AB), vimentin (VIME), intercellular adhesion molecule 1 (ICAM1), annexin A5 (ANXA5), A2 (ANXA2), reticulon-4 (RTN4), nicotamide phosphoribosyltransferase (NAMPT), serpin B3 (SPB3), and cathepsin D (CATD) (Table 1, Fig. 4F). Furthermore, we identified a group of proteins that were downregulated in steatosomes that are considered protective against MASLD, including clusterin and ubiquitin-like modifier-activating enzyme 1 (Table 1; Fig. 4F). A selection of proteins of interest was validated using SP-IRIS, incorporating colocalization analysis (Fig. 4G), which corroborated the proteomic findings and suggested the potential for assessing liver damage via circulating EVs (Fig. 4H–K). Consequently, significant upregulation of osteopontin (Fig. 4H), HMG-1 (Fig. 4I), and vimentin (Fig. 4J), as well as downregulation of clusterin (Fig. 4K), were observed in steatosomes compared to hepatosomes.

Table 1 Selection of biomarkers identified in steatosomes

Obese adipose tissue-to-liver crosstalk through obesesomes alters glucose and lipid metabolism in hepatocytes

To elucidate novel factors contributing to the deleterious effects of pathological obese adipose tissue on hepatic metabolic dysregulation associated with obesity, we investigated the functional role of obesesomes through their interaction with hepatocytes. Consequently, healthy adipose EVs, adiposomes, and pathological obesesomes from murine adipocytes (insulin-resistant/palmitic/oleic lipid hypertrophied) were isolated to evaluate their interaction with primary hepatocytes culture established from C57BL/6 mice. Pathological obesesomes were used to treat hepatocytes for 24 h (Fig. 5A), followed by the stimulation of the insulin pathway [100 nM insulin/5 min]. This revealed a significant impairment in insulin sensitivity, as evidenced by the substantial reduction in the pAkt and pS6 levels (Fig. 5B–D). This effect was not observed when hepatocytes were treated with vesicles derived from insulin-resistant adipocytes (HG/HI) or control adipocytes. Furthermore, obesesomes from lipid-hypertrophied adipocytes significantly decreased the expression of genes associated with de novo FA synthesis (FAS) (Fig. 5E), FA oxidation (PPARα) (Fig. 5F), FA uptake and transport (PPARγ) (Fig. 5G), lipid metabolism (PLIN3) (Fig. 5H), and glucose homeostasis and uptake (PDK1 and Glut1) (Fig. 5I, J) in hepatocytes. However, they elicited a moderate increase in inflammatory IL-6 expression (non-significant) and no alteration in TNF-α expression (Fig. 5K–L). These findings suggest that EVs secreted by lipid-hypertrophied adipocytes, rather than those from insulin-resistant adipocyte cell models, modify the lipid and glucose metabolism upon reaching healthy hepatocytes.

Fig. 5

Adipose tissue-to-liver crosstalk: Murine obesesomes induce insulin resistance and alter lipid and glucose metabolism in healthy hepatocytes. The effects of obesesomes from lipid-hypertrophied (palmitate/oleic acid) and insulin-resistant (HG/HI) adipocytes on murine primary hepatocytes were assayed by focusing on the insulin pathway (A). Representative images and densitometry of bands expressed towards insulin (10 mM, 5 min stimuli) of pAkt/Akt and pS6/S6 immunoblots [n = 6 independent lysates] are shown (B–D). CONT, primary hepatocytes without insulin stimulation; INS, primary hepatocytes stimulated with insulin; PALM/OLEIC, palmitic/oleic-hypertrophied hepatocytes stimulated with insulin; + ADIPO EVs, primary hepatocytes treated with vesicles isolated from control adipocytes; + HGHI EVs, primary hepatocytes treated with obesesomes from insulin-resistant adipocytes (HG/HI); + PALM EVs, primary hepatocytes treated with obesesomes shed by palmitate-hypertrophied adipocytes; + OLEIC EVs, primary hepatocytes treated with obesesomes shed by oleic-hypertrophied adipocytes. Relative expression of lipid [Fatty Acid Synthetase-FAS, Peroxisome Proliferator-Activated Receptor α/γ, perilipin 3] (E–H), glucose [PDK1, Glut1] (I, J) metabolism, and inflammation [IL-6, TNF-α] (K, L) genes in primary hepatocytes treated with the above obesesomes (n = 6 independent studies) is represented by histograms. Differences were analyzed using the Kruskal–Wallis test, followed by Dunn´s multiple comparison test (P ≤ 0.05, considered statistically significant; * P < 0.05, ** P < 0.01, and *** P < 0.001). Detailed raw data and complete immunoblot images are provided in Supplementary File 9. Figure created using Biorender (https://biorender.com/)

To obtain a more comprehensive understanding of adipose tissue-liver crosstalk, we established a parallel experiment by utilizing human obesesomes isolated from visceral (VAT) and subcutaneous (SAT) adipose tissue explants of independent patients with obesity or morbid obesity who underwent bariatric surgery (Fig. 6A, B). These VAT and SAT obesesomes were employed to treat control and steatotic human hepatocytes in culture (from Fig. 2A: combination treatment with HG/HI, palmitate and oleic acid). Notably, obesesomes shed by obese human adipose tissues of certain patients, particularly those from visceral depots, decreased insulin sensitivity in healthy hepatocytes. However, EVs patient donor variability precluded statistical significance (Fig. 6D). Correlation analysis of insulin sensitivity of treated HepaRG hepatocytes with clinical data from obesesomes donors revealed an inverse and moderate correlation [r = − 0.41] between SAT-obesesomes induced phosphorylation of Akt [arbitrary units] and the levels of circulating glucose [mg/dL]. Further, vesicles secreted by obese VAT demonstrated a moderate negative correlation between phosphorylation of Akt and BMI [r = − 057] and TG/HDL-C index [r = − 0.43] (Supplementary Fig. 1B). Expression analysis of hepatocytes treated with VAT and SAT obesesomes indicated no significant alterations in PPARγ expression, although obesesomes from both fat depots appeared to reduce PPARγ expression in both healthy and steatotic hepatocytes (Fig. 6G). However, consistent with the findings in murine primary hepatocytes, a significant reduction in PPARα expression was observed in healthy hepatocytes treated with SAT obesesomes and in steatotic hepatocytes treated with VAT obesesomes (Fig. 6H).

Fig. 6

Human obesesomes induce insulin resistance in healthy hepatocytes. Obesesomes isolated from human SAT and VAT explants of independent patients who underwent bariatric surgery (n = 8) were used to treat the human HepaRG control and steatotic hepatocytes (A and B, respectively). Representative images and densitometry of pAkt/Akt to assay insulin resistance upon treatment by stimulating cells with 100 mM insulin for 5 min in healthy adipocytes (C) or steatotic adipocytes (D) are shown. Each patient is represented by a different color, as shown in Table B. Real-time PCR expression analysis of PPARα/γ in treated and non-treated hepatocytes is shown (G–H). Differences were analyzed using One-way Anova Kruskal–Wallis test, followed by Dunn´s multiple comparisons test, or One way-Anova followed by Dunnett´s multiple comparisons test (P ≤ 0.05, considered statistically significant: * P < 0.05, ** P < 0.01). SAT, subcutaneous adipose tissue; VAT, visceral adipose tissue. Detailed raw data and complete immunoblot images are shown in Supplementary File 9. Figure created using Biorender (https://biorender.com/)

To elucidate potential contributors of the observed effects of obesesomes in hepatocytes, we reexamined our previous comprehensive proteome characterization of visceral obesesomes [11,12,13]. This analysis revealed the absence of beneficial adiponectin and omentin, as well as the presence of deleterious molecules associated with liver inflammation, and fibrosis in MASLD such as leptin, IL-6, IL-8, IL-11 and osteopontin. These findings suggest that these adipokines and cytokines, secreted from obese visceral adipose tissue within EVs, may contribute to the development of fatty liver disease.

Discussion

Extracellular vesicles are becoming increasingly significant as a complex and sophisticated mechanism of cellular and organ communication under physiological and pathological conditions; therefore, there is substantial interest in acquiring more comprehensive knowledge about the molecular events mediated by these vesicles, as they have emerged as potential therapeutic targets and as a source of real-time biomarkers [55]. These vesicles are loaded with bioactive molecules, including surface antigens, which may facilitate the specific binding of target cells or tissues. They can reach both local and systemic levels because EVs can circulate freely through circulation, are present in various biological fluids, and can even penetrate the central nervous system, as small EVs cross through the blood–brain barrier [56].

In this study, we examined the role of extracellular vesicles in the local communication between obese adipocytes and immune cells within adipose tissue, as well as the communication of steatotic hepatocytes with neighboring healthy hepatocytes in the context of fatty liver disease progression. Furthermore, we analyzed the communication between obese adipose tissue and the liver through secreted EV-obesesomes to elucidate their role in obesity-associated hepatic metabolic dysregulation. Specifically, we describe the following: (a) the paracrine communication between obesesomes and macrophages, which was found to promote inflammation and the secretion of pro-inflammatory cytokines; (b) the inflammatory effect of obesesomes related to BMI and systemic inflammation status; c) the endocrine interaction between obesesomes and hepatocytes, which was shown to have a significant impact on reducing insulin resistance and altering the expression of genes that regulate lipid and glucose metabolism; (c) in both cases, upon interaction with macrophages and hepatocytes, palmitate-loaded adipocytes shed EVs and obesesomes from human visceral adipose depots were responsible for a more harmful effect; (d) isolated EVs secreted by lipid steatosic hepatocytes (steatosomes) may exacerbate the initial stages of MASLD development by exerting local paracrine interactions with healthy hepatocytes, reducing insulin sensitivity, and dysregulating genes controlling lipid and glucose metabolism; (e) proteomic analysis of steatosomes secreted during fatty liver disease progression may serve as real-time indicators of MASLD’s early stages and their progression. This study provides substantive evidence that corroborates previous findings and, more significantly, elucidates novel effects of EVs on adipose crosstalk and their role in obesity. Functional assays were conducted not only with adipocyte cell cultures, but also with substantial whole visceral and subcutaneous adipose tissue samples from a significant cohort of patients with obesity undergoing bariatric surgery. Furthermore, to the best of our knowledge, this is the first study to characterize human steatosomes, including the identification and validation of biomarkers of liver disease involved in fibrosis, steatosis, steatohepatitis, metabolic dysfunction-associated steatotic liver disease (MASLD), and hepatocellular carcinoma.

Previous studies conducted in rodents demonstrated that obesity-associated insulin resistance is initiated in adipose tissue by the secretion of interleukin-6 (IL-6), tumor necrosis factor-alpha (TNF-α), and macrophage inflammatory protein-1 alpha (MIP-1α) by macrophages. [57]. Evidence exists, including our own, indicating intercellular communication between adipocytes and adipose tissue-resident macrophages (ATM) via EVs. This suggests that hypertrophied adipocytes may activate ATM, potentially contributing to obesity-associated inflammation and insulin resistance. [11, 58, 59]. In this study, we augmented existing evidence by demonstrating that obesesomes secreted by insulin-resistant lipid-loaded murine adipocytes can induce the secretion of proinflammatory cytokines by macrophages, particularly those secreted from palmitate-hypertrophied adipocytes. Consequently, macrophages cultured in the presence of obesesomes from lipid-loaded adipocytes exhibited a significant increase in the secretion of cytokines implicated in the recruitment of immune cells and regulation of inflammation in metabolic diseases, as well as cytokines associated with insulin resistance, such as CXCL11 and 13 [60], IFNγ [61], eotaxin 1 (CCL11) [62], interleukins 6 and 10 [63], MIF [64], TIMP 1/2 [65], and TNF-α [66]. Notably, we observed a comparable pro-inflammatory effect when utilizing obesesomes secreted by human whole adipose tissue explants of visceral and subcutaneous origins excised from obese patients on the day of bariatric surgery. To the best of our knowledge, this is the first study in which EVs of visceral and subcutaneous origin from human obese and morbidly obese adipose tissues were evaluated using inflammatory functional assays. These findings align with previous data describing the role of EVs from human non-obese AT explants in stimulating the differentiation of monocytes into adipose tissue macrophages (ATM) and inducing the secretion of factors that can inhibit insulin signaling in adipocytes [67]. Mechanistically, as previously described, this phenomenon may be attributed to EV-microRNA-34a [68]. Consequently, we posit that the outcomes of our investigation are significant, as they facilitate a physiological and realistic approach to elucidate intercellular communication via EVs in obese adipose tissue. Obesesomes released by visceral and subcutaneous tissue explants from independent obese subjects significantly increased the expression of inflammatory TNF-α in human macrophages, particularly in those treated with vesicles released by visceral fat depots. This observation aligns with previous findings regarding the association between visceral fat adipokine secretion and systemic inflammation. [69, 70]. In support of this observation, the TNF-α expression in human macrophages, induced by obesesomes from both depots, was moderately correlated with the systemic inflammation status of the donor patient.

Furthermore, we isolated extracellular vesicles from hepatocytes (hepatosomes), and steatosomes as vesicles secreted from hepatocytes during metabolic steatosis, simulating the prevalent obesity-related comorbidity MAFLD SLD [71]. To the best of our knowledge, this study represents the first proteomic analysis of extracellular vesicles (EVs) secreted by a fatty liver cell model of human hepatocytes, facilitating the identification of potential therapeutic targets as repositories of liver disease biomarkers, including fibrosis, steatosis, steatohepatitis, metabolic dysfunction-associated steatotic liver disease (MASLD), and hepatocellular carcinoma. Previous investigations have conducted proteomic analyses of circulating extracellular vesicles in non-alcoholic fatty liver disease [72,73,74,75,76], which are not comparable to the results of our study on EVs secreted directly by hepatocytes. Conde-Vacells et al. conducted a comprehensive proteomic analysis of exosomes released by murine primary liver cells in culture, establishing a foundation for elucidating the hepatic function under physiological conditions [29]. Moreover, the aforementioned research group analyzed EVs secreted by primary hepatocytes from Zucker animals [30], which was complemented by our study in human cells.

In this study, we identified proteins that are differentially present in vesicles secreted by healthy, steatotic, and insulin-resistant hepatocytes. Notably, comparison of steatosomes protein cargo from hepatocytes treated individually with palmitate, oleic acid, or a combination of the two together with high glucose and high insulin treatment, revealed three proteins that were common and upregulated, from which Golgi protein-73 (GP73) and sphingosine-1-phosphate phosphatase 1 (S1P) were previously identified. GP73 is a transmembrane glycoprotein present in the Golgi apparatus that is significantly increased in diseased livers, regardless of etiology, and serves as a diagnostic serological marker for liver diseases including liver fibrosis, cirrhosis, and hepatocellular carcinoma [76,77,78]. S1P has been implicated in ceramide synthesis and hepatic pro-fibrotic inflammatory processes through the modulation of vascular permeability, leukocyte infiltration, cellular survival, migration, proliferation, and differentiation of HSCs into myofibroblasts [79]. Evidence indicates that ceramide production in NAFLD is predominantly attributed to the activation of the de novo synthesis pathway of ceramides in hepatocytes, and its presence in steatosomes may reflect this metabolic alteration in the cell of origin [80]. Furthermore, among the proteins found to be upregulated in steatosomes compared to healthy hepatosomes are alkaline phosphatase (PPB1) and albumin, which are standard parameters utilized to evaluate liver function in MASLD and hepatocellular carcinoma in the clinical practice [81]. Furthermore, several biomarkers previously associated with liver pathology, such as high mobility group protein (HMG-1), were found to be upregulated in steatosomes. HMG-1, in conjunction with osteopontin (OSTP), another upregulated protein in these vesicles, responds to liver damage, thereby promoting fibrosis [33]. We also identified alkaline phosphatase (PPB1) as an independent liver fibrosis biomarker in the serum of obese patients undergoing surgery [34], intercellular adhesion molecule 1 (ICAM1) [42], and reticulon-4 [46]. As components of steatosomes protein cargo, we also identified upregulated proteins such as 4F2 cell-surface antigen (CD98) and cathepsin D, which are implicated in hepatic steatosis [38, 49, 82], nucleophosmin associated with hepatic insulin resistance [39]; S100A8 and A9 linked to fatty liver, steatohepatitis, MASLD, and HCC [40]; vimentin associated with steatohepatitis [41]; annexins A2 and A5 related to HCC [44]; nicotinamine phospohribosyltransferase implicated in the novo lipogenesis in MASLD; and serpin B3, which participates in the progression of MASLD and HCC [48]. In contrast, proteins such as clusterin were downregulated in steatosomes, as reported by Mlecko et al., who observed its downregulation in extracellular vesicles from primary hepatocytes of Zucker rats [30]. This is consistent with the results showing that elevated clusterin levels protect against Western diet-induced obesity and NAFLD [83]. Similarly, we identified ubiquitin-like modifier-activating enzyme 1, whose deficiency promotes pathological liver damage [52], septin-9 as a tumor suppressor whose methylation is associated with HCC [53], and cullin-associated NEDD8-dissociated protein 1, which mitigates NAFLD [54]. Additionally, our study identified several potential NAFLD EV biomarkers that were previously proposed in a multistep literature review [32]^. This finding suggests that EVs secreted during liver disease progression may become real-time subrogates for early stages of MASLD and indicators of disease progression.

Pathological proteins identified in steatosomes, in conjunction with specific miRNAs, may explain the pronounced deleterious effects on healthy hepatocytes observed in our study. An increasing body of literature has documented the role of EVs in non-alcoholic fatty liver disease [84, 85], and, conversely, the beneficial effects of EVs, such as whole-body glycemic control [86]. However, the observation that pathological steatosomes can significantly reduce insulin sensitivity in healthy hepatocytes has not been reported. Furthermore, our findings indicate that, although healthy hepatosomes did not significantly restore insulin sensitivity in steatotic hepatocytes, they exhibited a beneficial effect by increasing Glut1 expression and potentially improving glucose metabolism in these cells, aligning with recent results showing that liver EVs are responsible for blood glucose-lowering effects by acting on peripheral tissues [86]. Conversely, we demonstrated that pathological steatosomes were detrimental to healthy hepatocytes by downregulating the expression of lipid metabolism-regulating genes PLIN3, PPARα, and PPARγ, which is consistent with established knowledge regarding PPARα/γ downregulation promoting NAFLD and liver inflammation [87, 88]. Although our study was limited to hepatocyte-to-hepatocyte interactions without communication with other important cells (hepatic stellate, Kupffer, and liver sinusoidal endothelial cells), our data substantiated the hypothesis that excessive fatty acid accumulation is toxic to hepatocytes, and that this lipotoxicity can induce the release of EVs (hepatocyte-EVs), which can facilitate the progression of fibrosis via the activation of neighboring macrophages and hepatic stellate cells [89].

Finally, to discern the inter-organ crosstalk through EVs in obesity, we investigated white adipose-to-liver communication by using obesesomes to treat healthy hepatocytes. Remarkably, we observed that, consistent with our findings in healthy adipocytes, obesesomes derived from lipid-hypertrophied adipocytes, upon interaction with control hepatocytes, were capable of inducing insulin resistance after just 24 h of incubation. This detrimental effect of obesesomes on murine primary hepatocytes was accompanied by the reduced expression of genes implicated in de novo FA synthesis, FA oxidation, FA uptake and transport, and lipid and glucose metabolism, a phenomenon previously described to be associated with liver steatosis [90]. Interestingly, when assaying obesesomes isolated from whole visceral and subcutaneous adipose tissues of obese patients who underwent bariatric surgery, we observed a moderate positive correlation between obesesomes of visceral origin, reduction in insulin sensitivity, circulating levels of insulin-resistant HBA1c, and CRP ultra inflammation parameters in tissue donors. This observation suggests that the effects of obese adipose tissue-EVs at local and distant sites may correlate with systemic inflammation and insulin resistance. Our findings align with those of previous reports by Kranendonk et al., who conducted a similar experiment by treating hepatocytes with vesicles secreted by subcutaneous and omental adipose tissue explants from normal to overweight individuals (mean BMI: 25.8) undergoing surgery for aneurysmatic aortic disease. Although they described great variations, some patients showed a reduction in pAkt levels after insulin stimulation [16]. The fact that the obesesomes used in our experiments were obtained from obese and morbidly obese patients may explain the more pronounced effect observed on the insulin sensitivity of hepatocytes in our study. Upon reexamination of our previous findings analyzing the qualitative and quantitative protein cargo of obesesomes from lipid-hypertrophied adipocytes and human subcutaneous and visceral obese depots [11, 13], we identified adipokines that are implicated in obesity-related liver disease. Thus, we showed that obesesomes of visceral fat origin do not carry adiponectin or omentin, as expected; but indeed, they transport leptin. Leptin is profibrotic and enhances the release of TNF-α by Kupffer cells in pathogenic liver disease [91]. Additionally, we identified IL-6 pro-inflammatory cytokine in human obese visceral obesesomes, whose mRNA expression in omental adipose tissue was positively correlated with hepatic steatosis and insulin resistance in humans [92]. Furthermore, we observed the presence of osteopontin and IL-8 in obesesomes, both of which are associated with hepatic fibrosis in patients with NAFLD [93, 94]. Finally, IL-11, which is associated with liver fibrosis, inflammation, and steatosis in NAFLD, [95], was also detected in the obesesomes.

The limitations of this study include the inherent challenge of obtaining EVs from the adipose tissue of lean individuals, which would enhance the characterization of their functional roles. Furthermore, it is imperative to acknowledge that the findings of this investigation require additional validation through in vivo experiments.

Conclusions

In summary, this study contributes valuable and novel data regarding EV crosstalk in metabolic diseases, building upon previous research, including our own. First, we demonstrated that obese adipocytes and human obese adipose tissue not only induce inflammation in macrophages through the secretion of obesesomes, but also stimulate the secretion of pro-inflammatory cytokines that may exacerbate the cycle of inflammation and insulin resistance at both local and distant levels. Furthermore, we demonstrated the direct interaction of obesesomes with hepatocytes, which resulted in a deleterious reduction in insulin sensitivity and alteration of lipid and glucose metabolism pathways associated with liver steatosis. We have shown for the first time that both adverse effects (local inflammation of macrophages and distant metabolic alteration of hepatocytes) are more pronounced for vesicles released by visceral obese adipose tissue compared to subcutaneous tissue, and they appear to correlate with the systemic glucose and inflammation status of the individual. Moreover, we advanced the field by characterizing, for the first time, the protein cargo of EVs secreted by hepatocytes under steatosis conditions, compared to the vesicles of normal healthy hepatocytes, which demonstrated the significance of steatosomes as reservoirs of liver disease biomarkers. Additionally, we showed that steatosomes can independently induce insulin resistance, potentially exacerbating fatty liver disease once established. Importantly, we propose the potential therapeutic role of vesicles secreted by healthy hepatocytes in improving glucose metabolism at both local and systemic levels.