Extracellular vesicles from plasma of patients with minimal hepatic encephalopathy induce neuroinflammation and cognitive impairment in rats, which are prevented by treating the patients with rifaximin Authors: Adria Lopez-Gramaje, Yaiza M. Arenas, Juan-José Gallego, Paula Izquierdo-Altarejos, Amparo Urios, Desamparados Escudero-García, Salvador Benlloch, Elena Kosenko, Vicente Felipo, Carmina Montoliu Published: 2026 ABSTRACT Minimal hepatic encephalopathy (MHE) appearance is associated with a pro-inflammatory shift in peripheral inflammation. Treatment of MHE patients with rifaximin reverses this shift in and improves MHE. How peripheral alterations are transmitted to brain to induce MHE remains unclear. In rats with MHE, plasma extracellular vesicles (EV) induce cognitive impairment. We hypothesized that in cirrhotic patients, the shift in peripheral inflammation is associated with alterations in plasma EV which contribute to induce MHE in cirrhotic patients. We also hypothesized that rifaximin treatment reverses MHE by reversing changes in EV. The aims were: 1) assess if injecting rats with plasma EV from MHE patients (MHE-EV) induce cognitive impairment, 2) identify the underlying mechanisms, 3) assess if treating MHE patients with rifaximin reverses the pathological effects of their EV and, 4) if this is associated with reversal of changes in EV protein cargo. We isolated plasma EV from cirrhotic patients without and with MHE, treated or not with rifaximin, and controls and injected them to rats. We analysed cognitive function, neuroinflammation and glutamate receptors membrane expression in hippocampus. MHE-EV, but not EV from patients without MHE show increased TNFα content and trigger a strong cognitive impairment in rats. This is mediated by altered membrane expression of AMPA and NMDA glutamate receptors in hippocampus, due to increased neuroinflammation, with glial activation and enhanced activation of the TNFα-TNFR1-S1PR2-IL-1β-IL-1R pathway. The EV from MHE patients treated with rifaximin did not show increased TNFα levels and did not induce the above pathological effects. Rifaximin treatment reverses the changes in the cargo of the EV from plasma of MHE patients, including the increase in TNFα and eliminates the pathological effects of the EV. The data show that in cirrhotic patients the presence of MHE is associated with changes in their blood EV which transmit pathological effects to the brain, inducing neuroinflammation and altered glutamatergic neurotransmission in hippocampus which leads to cognitive impairment. Rifaximin treatment reduces TNFα and other altered proteins in the EV and reverses their pathological effects. EV could be a therapeutic target to improve MHE by modifying EV content or blocking the TNFα effects.Graphical Abstract Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12974-026-03922-z. Background Cirrhotic patients may show minimal hepatic encephalopathy (MHE), a complex neuropsychiatric syndrome, with mild cognitive impairment, psychomotor slowing, attention deficits and motor incoordination that reduce their ability to perform daily life tasks, their quality of life and life span [6, 16, 23, 40, 46, 48]. Peripheral inflammation plays a key role in MHE induction both in cirrhotic patients and animal models [1, 5, 7, 8, 10, 12, 32, 33]. MHE appearance in cirrhotic patients is associated with a shift in peripheral inflammation to a more pro-inflammatory form that is transmitted to the brain to induce neurological alterations [35]. Treatment with rifaximin reverses this shift in peripheral inflammation and cognitive impairment in most MHE patients [36]. However, how changes in the peripheral immune system are transmitted to the brain to induce neuroinflammation and cognitive impairment in MHE remain unclear. A main mechanism by which peripheral inflammation may affect brain function in different pathological situations is the infiltration of extracellular vesicles (EV) from blood [21, 28]. We have shown that EV from plasma of rats with hyperammonemia and MHE induce neuroinflammation and cognitive and motor impairment when injected to normal rats and that this is associated with increased TNFα and TNFR1 content in the EV, which mediate their pathological effects [28–30]. It is therefore likely that also in cirrhotic patients EV from blood could play a role in the transmission of peripheral alterations to the brain to induce MHE. We have recently shown that plasma EV from MHE patients have altered miRNA and protein contents, and are enriched in inflammatory factors compared to controls and patients without MHE. EV from MHE patients modulate the expression of pro-inflammatory IL-17, IL-21, and TNF-α in cultured CD4+ T lymphocytes, and increase the activation of Th17 cells [18]. This indicates that plasma EV play an important role in the immune changes observed in MHE. We hypothesized that plasma EV play a role in the induction of MHE in cirrhotic patients. Moreover, we also hypothesized that a) reversal of the alterations in peripheral inflammation of MHE patients by rifaximin treatment would be associated with reversal of the pathological effects of plasma EV and b) this would be associated with reversal of the pathological changes in the plasma EV protein cargo. To assess these possibilities, the aims of this work were: 1) assess if injection of EV from plasma of patients with MHE induce cognitive impairment in normal rats, 2) identify the underlying mechanisms, 3) assess if rifaximin treatment reverses the pathological effects of EV from MHE patients and 4) if this is associated with reversal of the pathological changes in the plasma EV protein cargo. We first isolated EV from plasma of cirrhotic patients without and with MHE and from control subjects without liver disease and injected them to normal rats. We analysed if the EV injection alters cognitive function as assessed in the object location, Y-maze and radial maze tests. Cognitive impairment in rat models of MHE is mediated by glial activation, neuroinflammation, with increased pro-inflammatory factors (TNFα and IL-1β), which induce alterations in membrane expression of NMDA and AMPA receptors for glutamate in hippocampus [12, 24, 25, 41, 43]. We also analysed the effects of EV injection on these parameters. In a second study, we assessed if rifaximin treatment reverses the pathological effects of EV from plasma of MHE patients by analysing again all the above parameters. We also analysed the protein cargo of the EV. Experimental procedures Patients and controls Fifty-six patients with liver cirrhosis (19 without MHE, 37 with MHE) were included. They were recruited from the outpatient clinics at Hospital Clínico and Hospital Arnau de Vilanova of Valencia, Spain. Cirrhosis was diagnosed based on clinical, biochemical, and ultrasonographic data. The exclusion criteria were: overt hepatic encephalopathy, infection, recent (< 6 months) alcohol intake, recent (< 6 weeks) antibiotic use, gastrointestinal bleeding or use of drugs affecting cognitive function; hepatocellular carcinoma or neurological or psychiatric disorder. 17 MHE patients were treated with rifaximin for 12 months (1.2 g/day, three doses of 400 mg every 8 h) and considered a separate group. Regarding control group, 28 healthy volunteers were enrolled in the study once liver disease was discarded. Participants were included in the study after signing their written informed consent. The main demographic, clinical and neuropsychological characteristics are summarized in Supplementary Table 1. The study protocols were approved by the Scientific and Ethical Committees of both hospitals (2018/051 and 2023/130). The procedures followed were in accordance with the ethical guidelines of the Declaration of Helsinki. Diagnosis of Minimal Hepatic Encephalopathy (MHE) The gold standard for diagnosing MHE, Psychometric Hepatic Encephalopathy Score (PHES), was used to classify cirrhotic patients as showing MHE or not [17]. The scores were adjusted for age and education level using Spanish normality tables (www.redeh.org/TEST_phes.htm). Threshold for MHE diagnosis is −4. Control subjects also undertook PHES. Extracellular vesicles isolation EV were isolated from −80ºC stored plasma samples by size-exclusion chromatography using the qEV 2/70 nm columns from IZON (Izon, Lyon, France) as in [18]. Fractions from 6 to 10 were collected and concentrated to 100–200 uL. Protein was measured with the bicinchoninic acid (BCA) assay (Thermo Fisher Scientific Inc., Waltham, MA, USA) and the samples were stored at −80ºC. Nanoparticle Tracking Analysis (NTA) Distribution profile, size and quantity of vesicles were assessed by Nanoparticle Tracking Analysis with a NanoSight NS300 system (Malvern Panalytical, Malvern, UK) using a 1:1000 dilution in PBS. Transmission Electron Microscopy (TEM) To confirm de presence of EVs in sample TEM was performed. Briefly, isolated vesicles were loaded onto carbon-coated copper grids and contrasted with 1% uranyl acetate. Samples were examined with a transmission electron microscope and images were acquired using a digital camera. Analysis of extracellular vesicle markers by Western blot Samples were subjected to electrophoresis and immunoblotting as described as in [15]. Primary antibodies were used against Alix (1:500, PROTEINTECH), Flotillin-2 (1:500, INVITROGEN), CD9 (1:1000, SIGMA) and Calnexin (1:500, NOVUS). Secondary antibody (1:4000) against rabbit was IgG conjugated with alkaline phosphatase (Sigma). Alix was considered a positive control for protein loading. EV purity was determined as a comparison with a plasma sample. The images were captured using a Hewlett Packard ScanJet 5300 C and band intensities were quantified using AlphaImager 2200 software and expressed like percentage of controls. Fluorescent labelling of extracellular vesicles To determine whether the administered EVs were able to reach the brain, an additional experiment was conducted in a separate group of rats (n = 4). EVs were fluorescently labeled with DiI dye (Sigma) following the protocol described by [28]. Each animal received an intravenous injection of 100 μg of labeled vesicles and was euthanized 24 h later, followed by transcardial perfusion with PBS. Brains were subsequently collected and embedded in OCT before immediate freezing. Cryosections of 10 μm thickness were prepared and immunostained using an antibody against Iba1 (1:300, WAKO). Appropriate donkey anti-rabbit Alexa Fluor 488 secondary antibody (1:400, Invitrogen) was then applied, followed by DAPI nuclear staining. Fluorescence images were obtained using a Leica TCS SP8 inverted laser scanning confocal microscope equipped with 40 × or 63 × oil-immersion Plan-Apochromat Lambda Blue objectives (1.4 1.4). Multiplex profiling of EV protein cargo EV protein cargo was analyzed by multiplex proximity extension assay (PEA) using the Olink Target-96 Inflammation Panel (T96INF02). EV samples were diluted with RIPA buffer up to 0.5 μg protein/μL. The PEA of 92 proteins was performed by Cobiomic Bioscience S.L, Cordoba, Spain as in [4]. Results are reported as percentage of controls on an arbitrary Log2-based NPX (normalized protein expression) scale adjusted to 2^NPX as in [4]. Experimental design The experimental design of study 1 is summarized in Fig. 1A. Male Wistar rats (120–140 g) were distributed into five groups: C: control rats injected with PBS, C-EV: control rats injected with EV from control subjects, NMHE-EV: control rats injected with EV from cirrhotic patients without MHE, MHE-EV: control rats injected with EV from cirrhotic patients with MHE, and HA: hyperammonemic rats injected with PBS. HA rats were made hyperammonemic by feeding them a diet containing ammonium acetate as in [43]. The conditions were held the same for control and hyperammonemic rats. Two identical experiments were performed, one with 7 and another with 8 rats per group. EV injected to rats were a pool of EV from different subjects, as detailed in Supplementary Table 1. All subjects were males. Although the age of patients was around 60 years we included young subjects (around 32 years) as controls to assess if EV induce effects in rats due to their human origin. The results show that MHE-EV induce cognitive impairment while NMHE-EV do not and both groups have the same age, indicating that EV do not induce effects due to their human origin and that patient’s age does not influence the cognitive impairment induction.Fig. 1Experimental design. A. Experiment 1. Extracellular vesicles were isolated from plasma of control subjects (C-EV) and from cirrhotic patients without (NMHE-EV) or with MHE (MHE-EV). On weeks 2 and 3 of the ammonium-containing diet, control rats were injected in the tail vein either with EV or PBS as vehicle. Analysis of cognitive impairment in the Y-Maze, Novel Object Location, Novel Object Recognition and Radial maze was performed after three, four and five weeks of hyperammonemia. Rats were sacrificed at 6 weeks of hyperammonemia. B. Experiment 2. Was performed as experiment 1, using EV from MHE patients treated (RIFAX-EV) or not (MHE-EV) with rifaximin Experimental design. A. Experiment 1. Extracellular vesicles were isolated from plasma of control subjects (C-EV) and from cirrhotic patients without (NMHE-EV) or with MHE (MHE-EV). On weeks 2 and 3 of the ammonium-containing diet, control rats were injected in the tail vein either with EV or PBS as vehicle. Analysis of cognitive impairment in the Y-Maze, Novel Object Location, Novel Object Recognition and Radial maze was performed after three, four and five weeks of hyperammonemia. Rats were sacrificed at 6 weeks of hyperammonemia. B. Experiment 2. Was performed as experiment 1, using EV from MHE patients treated (RIFAX-EV) or not (MHE-EV) with rifaximin On week 2 of the ammonium-containing diet, rats were injected in the tail vein either with 50 µg of EV in 300 μl or the same volume of PBS as vehicle. The injected groups were: C: control rats injected with PBS, C-EV: control rats injected with EV from control subjects, NMHE-EV: control rats injected with EV from cirrhotic patients without MHE, MHE-EV: control rats injected with EV from cirrhotic patients with MHE, and HA: hyperammonemic rats injected with PBS. One identical additional injection was given one week later. Analysis of cognitive impairment in the Y-Maze, Novel Object Location, Novel Object Recognition and Radial maze were performed at three, four and five weeks of hyperammonemia. Rats were sacrificed after 6 weeks of hyperammonemia. In a second study, we assessed if treatment with rifaximin of MHE patients reverses the pathological effects of their plasma EV. The experimental design of study 2 is summarized in Fig. 1B. The study was performed exactly as study 1, including the groups of rats: C (n = 8), HA (n = 8), and MHE-EV (n = 10), as in study 1 and, RIFAX-EV: rats injected with EV from plasma of rifaximin-treated MHE patients (n = 20). The conditions of this second experiment were the same than for the first experiment. The experiments were approved by the Comite Ético de Experimentación Animal (CEEA) of our center (2024–26) and by the Conselleria de Agricultura of Generalitat Valenciana, were performed in accordance with the guidelines of the Directive of the European Commission (2010/63/EU) for care and management of experimental animals, and comply with the ARRIVE guidelines. Analysis of protein content by Western blot Rats were sacrificed at 6 weeks of hyperammonemia. Hippocampus was dissected from 9 rats per group and homogenized in 50 mM TRIS–HCl pH 7.5, 50 mM NaCl, 10 mM EGTA, 5 mM EDTA and protease and phosphatase inhibitors. Protein content was analyzed by Western blot as in [15] using antibodies against TNF-α (1:250) and IL-1β (1:250) from R&D SYSTEMS, TNFR1 (1:1500) from Proteintech, and S1PR2 (1:1000) from ABCAM, and actin or GAPDH as a control for protein loading. Secondary antibodies (1:4000) against rabbit, mouse, or goat were IgGs conjugated with alkaline phosphatase (Sigma). The images were captured using a Hewlett Packard ScanJet 5300 C and band intensities were quantified using AlphaImager 2200 software. Analysis of membrane expression of receptors and transporters Membrane expression of proteins in hippocampal slices was analysed by cross-linking with BS3 (Pierce cat# 21,580, Rockford, IL) as described in [9] using antibodies against IL1R (1:1000), S1PR2 (1:1000) both from ABCAM, NR2B (1:1000), NR2A (1:1000), GluA1 (1:1000) all from Millipore, and GluA2 (1:2000) and TNFR1 (1:1500) from Proteintech. Samples treated with or without BS3 were analysed by Western blot using the above antibodies. The membrane expression of receptor subunits was calculated as the difference between the intensity of the bands without BS3 (total protein) and with BS3 (non-membrane protein) [9]. Immunohistochemistry in brain sections Six rats per group were anaesthetized with sodium pentobarbital and subjected to transcardial perfusion with 0.9% saline followed by 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). The brains were removed and post-fixed in the same fixative for 24 h at 4 °C. Paraffin-embedded sections. (5-μm thick) were cut and mounted on coated glass slides, processed with the Envision Flex + Kit (Dako) to block endogenous peroxidase activity for 5 min, and incubated with antibodies against Iba1 (Wako; 1:300), GFAP (SIGMA; 1:400), TNF-α (ABCAM; 1:300) or IL-1β (ABCAM; 1:300) overnight. Then, slides were incubated with Goat anti-mouse or anti-rabbit (biotinylated) secondary antibodies (Vector Laboratories) for 1 h and diaminobenzidine for 2 min. Sections were counterstained with Mayer’s hematoxylin (DAKO) for 1 min. Once the slides were dry, they were scanned using an Aperio Versa scanner (Leica Biosystems Nussloch GmbH, Germany). Scanned slides were analyzed using ImageScope64 software. Analysis of astrocytes and microglia activation and neuronal content of IL-1β and TNFα Analysis of Iba1 and GFAP staining was performed in the hippocampal slices using the Image J software as in [34]. Analysis of IL-1β and TNF-α staining was performed as in [30]. Hippocampal slices from six animals per group were used. Cognitive function assessment NOR and NOL memory tests were performed as in [44] in an open-field arena. Rats were habituated during 3 days, 10 min per day. NOL test was performed on day 4. NOR test was performed on day 5. The time exploring the familiar stimulus and the novel stimulus (object with different location for NOL and unexplored object for NOR) was counted. Discrimination ratio for each test was calculated as: ((Time exploring novel stimulus – Time exploring familiar stimulus)/Total exploration time). Evaluation of Spatial Learning in the 8-Arms Radial Maze was performed as in [25]. Rats were kept under a caloric restricted diet during the test. For the two days prior to the test, the rats' food intake was slightly reduced to maintain the motivation to seek for food. Caloric restriction was maintained throughout the test, and the animals' weight was monitored to ensure that weight loss did not exceed 10%. We have performed the radial maze test previously in many studies in control and hyperammonemic rats (e.g. [7, 25, 30]). In these previous studies we have set up the conditions to restrict slightly the diet in hyperammonemic rats to maintain the motivation to seek for food while maintaining high blood ammonia levels. Food restriction was similar in control and hyperammonemic rats. On the first day, rats were habituated to the maze. The test was performed in the following 4 days with 3 trials per day. The number of reference memory errors (unbaited arms visited) and working memory errors (entries to arms already visited in the same trial) were calculated. Learning index was defined as the difference between the number of right choices and reference errors as in [25]. Evaluation of Short-term Memory in the Y-Maze was performed as in [30]. Each rat was placed on the starting arm (vertical) and allowed to explore the maze with one of the arms closed for 2 min (training round) three times, with a 1-min interval between each round. The rat was then placed back on the vertical arm and allowed to explore all three arms of the maze for 2 min (test round). The time spent on each arm was quantified, and the discrimination ratio was calculated as: (time on the new arm – time on the familiar arm)/total time on both arms. Statistical analysis Results are expressed as mean ± standard error. All statistical analyses were performed using the software program GraphPad Prism v.9.0. Normality was assessed using the D’Agostino and Pearson Omnibus test and the ShapiroWilk normality tests. Differences in variances of normally distributed data were assessed using Bartlett’s test. Data with the same variance across groups were analyzed by a parametric one-way analysis of variance (ANOVA) followed by Fisher’s LSD multiple comparisons test or two‐way ANOVA when appropriate. Data with different variance across groups were analyzed using the non-parametric Kruskal–Wallis test followed by Dunnett’s post hoc test. A confidence level of 95% was accepted as significant. The number of rats used for each parameter and the statistical procedure used in each case is indicated in the corresponding Figure legend. Results Characterization of EV from plasma We first characterized the EVs isolated from plasma of control subjects and of cirrhotic patients without MHE and with MHE, treated or not with rifaximin. The EVs exhibited properties compatible with exosomes regarding to morphology, size and surface markers. Electron microscopy images of the isolated EVs are shown in Fig. 2A. The content of markers typical of exosomes was analyzed by Western blot. The EVs were positive for Alix, Flotillin-2, and CD9 (Fig. 2B) and negative for calnexin (Fig. 2C), indicating that they are purified EV and have no contaminants such as calnexin, a marker for purity of EV [13]. No significant differences were observed in the content of Alix (Fig. 2D), Flotillin-2 (Fig. 2E) or CD9 (Fig. 2F) between the different groups.Fig. 2Characterization of EV isolated from plasma of control subjects, cirrhotic patients without MHE and with MHE treated or not with rifaximin. A Representative images obtained by transmission electron miscroscopy after negative staining. Content of markers Alix, (B, D) Flotillin2 (B, E), and CD9 (B, F) and of calnexin (C) analyzed by Western blot. Calnexin was analysed as an indicator of EV purity compared to plasma sample (C). Representative images are shown in B and C. Particle size mode (nm, H), total particles per mililiter (I) measured by nanoparticle tracking analysis (NTA), and total protein (µg/sample, J) in EVs. The total particles and total protein of EVs were referred to the initial volume of plasma used to isolate EVs (5 mL). Examples of the particle size distribution profiles analized by NTA in the four groups of study are show in G. Values are the mean ± SEM of 6–8 samples per group for Western blot and 10–25 for NTA. Values significantly different from controls are indicated by asterisk (*p < 0.05, ***p < 0.001, ****p < 0.0001), and from MHE-EV by “a” (aa: p < 0.01, aaa: p < 0.001, aaaa: p < 0.0001) Characterization of EV isolated from plasma of control subjects, cirrhotic patients without MHE and with MHE treated or not with rifaximin. A Representative images obtained by transmission electron miscroscopy after negative staining. Content of markers Alix, (B, D) Flotillin2 (B, E), and CD9 (B, F) and of calnexin (C) analyzed by Western blot. Calnexin was analysed as an indicator of EV purity compared to plasma sample (C). Representative images are shown in B and C. Particle size mode (nm, H), total particles per mililiter (I) measured by nanoparticle tracking analysis (NTA), and total protein (µg/sample, J) in EVs. The total particles and total protein of EVs were referred to the initial volume of plasma used to isolate EVs (5 mL). Examples of the particle size distribution profiles analized by NTA in the four groups of study are show in G. Values are the mean ± SEM of 6–8 samples per group for Western blot and 10–25 for NTA. Values significantly different from controls are indicated by asterisk (*p < 0.05, ***p < 0.001, ****p < 0.0001), and from MHE-EV by “a” (aa: p < 0.01, aaa: p < 0.001, aaaa: p < 0.0001) Size distribution profile was analyzed by Nanoparticle Tracking Analysis. Representative profiles are shown in Fig. 2G. The mean diameter of EVs was 102 ± 5 nm for control subjects and was increased in cirrhotic patients without (128 ± 3 nm) and with (123 ± 5 nm) MHE. Rifaximin treatment reversed this increase, returning the diameter to 87 ± 2 nm (Fig. 2H). The amount of EVs in plasma was 4.5 × 1011 ± 0.7 × 1011 particles/ml for control subjects and was reduced in cirrhotic patients without (2.7 × 1011 ± 0.3 × 1011 particles/ml) and with (2 × 1011 ± 0.2 × 1011 particles/ml) MHE. Rifaximin treatment reversed this reduction, returning the concentration to 5.9 × 1011 ± 0.9 × 1011 particles/ml (Fig. 2I). The total protein content in the EV was increased in MHE patients (269 ± 28 μg/sample) compared to control subjects (167 ± 11 μg/sample) and this increase was reversed by rifaximin, returning the protein content to 182 ± 13 μg/sample (Fig. 2J). We confirmed that injected EVs reached the hippocampus (Fig. 3). DIL-labeled EVs (red signal) co-localized with CA1 neurons (Fig. 3A) and with microglia cells stained with Iba1 (Fig. 3B).Fig. 3The plasma EV injected to rats reach the hippocampus. Intravenously injected Dil-labeled extracellular vesicles (red) reach the hippocampus of control rats after 24 h and are detected in the neurons of the CA1 region (A). Co-localization with microglia (B), indicated by arrows. Scale bar = 50 µm (white); 15 µm (yellow) The plasma EV injected to rats reach the hippocampus. Intravenously injected Dil-labeled extracellular vesicles (red) reach the hippocampus of control rats after 24 h and are detected in the neurons of the CA1 region (A). Co-localization with microglia (B), indicated by arrows. Scale bar = 50 µm (white); 15 µm (yellow) Injection of EV from MHE patients impairs cognitive function in normal rats. Rifaximin treatment reverses this effect Injection of MHE-EV induced cognitive impairment in the novel object location (NOL) test, reducing the discrimination index to −0.03 ± 0.04 compared to 0.15 ± 0.08 in control rats. NOL was also impaired in hyperammonemic rats but was not affected by injection of EV from cirrhotic patients without MHE (NMHE-EV) or EV from control subjects without liver disease (C-EV). (Fig. 4A). EV isolated from plasma of MHE patients treated with rifaximin (RIFAX-EV) did not impair NOL (Fig. 4B).Fig. 4Injection of EV from MHE patients impairs cognitive function in control rats. Cognitive function was assessed in the Novel Object Location (NOL) test (A, B); Novel Object Recognition (NOR) test (C, D) and in the Y maze (E, F). Values are the mean ± SEM of 13–15 rats per group. One-way ANOVA followed by Fisher’s LSD post-hoc test was performed to compare groups. Values significantly different from control rats are indicated by asterisks, where *: p < 0.05, **: p < 0.01, ***: p < 0.001, and ****: p < 0.0001; and from MHE-EV by “a”: a: p < 0.05 Injection of EV from MHE patients impairs cognitive function in control rats. Cognitive function was assessed in the Novel Object Location (NOL) test (A, B); Novel Object Recognition (NOR) test (C, D) and in the Y maze (E, F). Values are the mean ± SEM of 13–15 rats per group. One-way ANOVA followed by Fisher’s LSD post-hoc test was performed to compare groups. Values significantly different from control rats are indicated by asterisks, where *: p < 0.05, **: p < 0.01, ***: p < 0.001, and ****: p < 0.0001; and from MHE-EV by “a”: a: p < 0.05 MHE-EV impaired performance in the novel object recognition (NOR) test, reducing the discrimination index compared to control rats. NOR was also impaired in hyperammonemic rats but was not affected by injection of NMHE-EV or of C-EV (Fig. 4C). RIFAX-EV did not impair NOL (Fig. 4D). MHE-EV impaired short-term memory in the Y maze compared to control rats. Y maze was also impaired in hyperammonemic rats but was not affected by injection of NMHE-EV or of C-EV (Fig. 4E). RIFAX-EV did not impair short-term memory (Fig. 4F). Cognitive function was also evaluated in the radial maze. Reference memory errors on day 4 of the test were increased by MHE-EV and in hyperammonemic rats compared to control rats and were not affected by NMHE-EV or C-EV. (Fig. 5A). RIFAX-EV did not impair reference memory (Fig. 5B).Fig. 5Injection of EV from MHE patients impairs spatial learning in control rats. In the radial maze reference errors (A-B), working errors (C-D) and learning index (E–F) were analyzed. Values are the mean ± SEM of 13–15 rats per group. One-way ANOVA followed by Fisher’s LSD post-hoc test was performed to compare groups. Values significantly different from control rats are indicated by asterisks, where **: p < 0.01, and ***: p < 0.001; and from MHE-EV by “a”: a: p < 0.05, aa: p < 0.01 Injection of EV from MHE patients impairs spatial learning in control rats. In the radial maze reference errors (A-B), working errors (C-D) and learning index (E–F) were analyzed. Values are the mean ± SEM of 13–15 rats per group. One-way ANOVA followed by Fisher’s LSD post-hoc test was performed to compare groups. Values significantly different from control rats are indicated by asterisks, where **: p < 0.01, and ***: p < 0.001; and from MHE-EV by “a”: a: p < 0.05, aa: p < 0.01 Working memory errors on day 4 of the test were increased by MHE-EV and in hyperammonemic rats compared to control rats and were not affected by injection of NMHE-EV or C-EV (Fig. 5C). RIFAX-EV did not impair working memory (Fig. 5D). Learning index on day 4 of the test was impaired by MHE-EV and NMHE-EV and in hyperammonemic rats and was not affected by C-EV (Fig. 5E). RIFAX-EV did not impair learning index (Fig. 5F). Supplementary Fig. 1 shows the progression of reference errors, working errors and learning index. Injection of EV from MHE patients alters membrane expression of NMDA and AMPA receptors in hippocampus in normal rats. Rifaximin treatment reverses this effect In hyperammonemic rats cognitive impairment is due to altered membrane expression of NMDA and AMPA receptors in hippocampus [9, 12, 34, 44]. We therefore analyzed the effects of injecting EV to normal rats on membrane expression of the NMDA receptor subunits NR2A and NR2B and of the AMPA receptor subunits GluA1 and GluA2. The membrane expression of NR2B was increased by MHE-EV, NMHE-EV and by hyperammonemia but not by C-EV (Fig. 6A) or by RIFAX-EV (Fig. 6B).Fig. 6In vivo treatment with MHE-EV alters membrane expression of NMDA and AMPA receptors in hippocampus of control rats. Membrane expression of NR2B (A, B), NR2A (C, D), GluA1 (E, F) and GluA2 (G, H) in hippocampus was analyzed by Western blot the cross-linker BS3. For each protein, samples without BS3 represent the total amount, while samples with BS3 represent the non-membrane fraction. Representative images of the blots are shown. Values are the mean ± SEM of 8–9 rats per group, expressed as percentage of controls. One-way ANOVA followed by Fisher’s LSD multiple comparisons test (A-F) or non-parametric Kruskal–Wallis test followed by Dunnett's post hoc test (G-H) were performed to compare groups with same or different variance of normality, respectively. Values significantly different from control rats are indicated by asterisks, where *: p < 0.05, **: p < 0.01, ***: p < 0.001 and ****: p < 0.0001; and from MHE-EV by “a”: a: p < 0.05, aaaa: p < 0.0001 In vivo treatment with MHE-EV alters membrane expression of NMDA and AMPA receptors in hippocampus of control rats. Membrane expression of NR2B (A, B), NR2A (C, D), GluA1 (E, F) and GluA2 (G, H) in hippocampus was analyzed by Western blot the cross-linker BS3. For each protein, samples without BS3 represent the total amount, while samples with BS3 represent the non-membrane fraction. Representative images of the blots are shown. Values are the mean ± SEM of 8–9 rats per group, expressed as percentage of controls. One-way ANOVA followed by Fisher’s LSD multiple comparisons test (A-F) or non-parametric Kruskal–Wallis test followed by Dunnett's post hoc test (G-H) were performed to compare groups with same or different variance of normality, respectively. Values significantly different from control rats are indicated by asterisks, where *: p < 0.05, **: p < 0.01, ***: p < 0.001 and ****: p < 0.0001; and from MHE-EV by “a”: a: p < 0.05, aaaa: p < 0.0001 The membrane expression of NR2A was reduced by MHE-EV, NMHE-EV and by hyperammonemia but not by C-EV (Fig. 6C) or RIFAX-EV (Fig. 6D). The membrane expression of GluA1 was reduced by MHE-EV, NMHE-EV and by hyperammonemia and, at a lesser extent by C-EV (Fig. 6E), but was not affected by RIFAX-EV (Fig. 6F). The membrane expression of GluA2 was increased by MHE-EV and by hyperammonemia but not by NMHE-EV or C-EV (Fig. 6G) nor by RIFAX-EV (Fig. 6H). Injection of EV from MHE patients enhances the activation of the TNFα-TNFR1-S1PR2-IL-1β-IL-1R pathway in hippocampus of normal rats. Rifaximin treatment reverses this effect Altered membrane expression of glutamate receptors in hyperammonemia is a consequence of neuroinflammation. Increased levels of TNFα and membrane expression of its receptor TNFR1 increase membrane expression and activation of the S1PR2, which increases IL-1β and membrane expression and activation of its receptor IL-1R, leading to increased membrane expression of NR2B which alters membrane expression of GluA1 and GluA2 [41, 43]. Injection of MHE-EV to normal rats induced similar effects to hyperammonemia, increasing TNFα levels (Fig. 7A), the content (Fig. 7C) and membrane expression of TNFR1 (Fig. 7E), content (Fig. 7G) and membrane expression of S1PR2 (Fig. 7I), IL-1β levels (Fig. 7K) and membrane expression of the IL-1R (Fig. 7M).Fig. 7Injection of EV from MHE patients enhances the activation of the TNFα-TNFR1-S1PR2-IL-1β-IL-1R pathway in hippocampus of normal rats. Protein content of TNFα (A, B), TNFR1 (C, D), S1PR2 (G, H) and IL-1β (K, L) was analyzed by Western blot. Membrane expression of TNFR1 (E, F), S1PR2 (I, J) and IL-1R (M, N) was analyzed using BS3 cross-linker. Representative images of the blots are shown. Values are expressed as percentage of controls and are the mean ± SEM of 8–9 rats per group. One-way ANOVA followed by Fisher’s LSD multiple comparisons test (A-D, G-L) or non-parametric Kruskal–Wallis test followed by Dunnett's post hoc test (E–F, M–N) were performed to compare groups with same or different variance of normality, respectively. Values significantly different from controls are indicated by asterisk (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Values significantly different from MHE-EV are indicated by “a”: a: p < 0.05, aa: p < 0.01, aaa: p < 0.001, aaaa: p < 0.0001 Injection of EV from MHE patients enhances the activation of the TNFα-TNFR1-S1PR2-IL-1β-IL-1R pathway in hippocampus of normal rats. Protein content of TNFα (A, B), TNFR1 (C, D), S1PR2 (G, H) and IL-1β (K, L) was analyzed by Western blot. Membrane expression of TNFR1 (E, F), S1PR2 (I, J) and IL-1R (M, N) was analyzed using BS3 cross-linker. Representative images of the blots are shown. Values are expressed as percentage of controls and are the mean ± SEM of 8–9 rats per group. One-way ANOVA followed by Fisher’s LSD multiple comparisons test (A-D, G-L) or non-parametric Kruskal–Wallis test followed by Dunnett's post hoc test (E–F, M–N) were performed to compare groups with same or different variance of normality, respectively. Values significantly different from controls are indicated by asterisk (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Values significantly different from MHE-EV are indicated by “a”: a: p < 0.05, aa: p < 0.01, aaa: p < 0.001, aaaa: p < 0.0001 Injection of NMHE-EV increased membrane expression of TNFR1 (Fig. 7E) and of IL-1R (Fig. 7M), but did not affect the remaining steps of the pathway (Fig. 7). Injection of C-EV or RIFAX-EV did not affect any step of the TNFα-TNFR1-S1PR2-IL-1β-IL-1R pathway (Fig. 7). Injection of EV from MHE patients increases activation of microglia and astrocytes and neuronal TNFα and IL-1β in hippocampus of normal rats. Rifaximin treatment reverses this effect The increase in TNFα and IL-1β in hyperammonemic rats is a consequence of glial activation and increased TNFα and IL-1β production by CA1 neurons [25, 41, 43]. Injection of MHE-EV to normal rats induced similar effects to hyperammonemia, increasing microglia activation in hippocampus (Fig. 8A, E and F) as reflected by reduced area (Fig. 8E) and perimeter (Fig. 8F). MHE-EV also increased activation of astrocytes, as reflected by increased area covered by GFAP (Fig. 8B and 8G). MHE-EV also increased TNFα (Fig. 8C and H) and IL-1β in CA1 neurons (Fig. 8D and I).Fig. 8In vivo treatment with EV from MHE patients increases activation of microglia and astrocytes and neuronal production of TNFα and IL-1β in CA1 hippocampal neurons. Representative images of immunohistochemistry against IBA1 (A), GFAP (B), TNFα (C) and IL-1β (D). Arrows indicate the staining of TNFα (C) and IL-1β (D) in neurons. The area and perimeter of microglia (IBA1 + cells) are expressed in µm2 and µm respectively (E and F), GFAP stained area as percentage of total area (G). Content of TNFα and IL-1β is expressed as intensity of staining and values are percentage of controls (H and I). Values are the mean ± SEM of 4–6 rats per group. One-way ANOVA followed by Fisher’s LSD post-hoc test was performed to compare all groups. Values significantly different from controls are indicated by asterisk (**p < 0.01, ***p < 0.001, ****p < 0.0001), and from MHE-EV by “a” (a: p < 0.05, aa: p < 0.01, aaa: p < 0.001) In vivo treatment with EV from MHE patients increases activation of microglia and astrocytes and neuronal production of TNFα and IL-1β in CA1 hippocampal neurons. Representative images of immunohistochemistry against IBA1 (A), GFAP (B), TNFα (C) and IL-1β (D). Arrows indicate the staining of TNFα (C) and IL-1β (D) in neurons. The area and perimeter of microglia (IBA1 + cells) are expressed in µm2 and µm respectively (E and F), GFAP stained area as percentage of total area (G). Content of TNFα and IL-1β is expressed as intensity of staining and values are percentage of controls (H and I). Values are the mean ± SEM of 4–6 rats per group. One-way ANOVA followed by Fisher’s LSD post-hoc test was performed to compare all groups. Values significantly different from controls are indicated by asterisk (**p < 0.01, ***p < 0.001, ****p < 0.0001), and from MHE-EV by “a” (a: p < 0.05, aa: p < 0.01, aaa: p < 0.001) Injection of NMHE-EV increased astrocytes activation (Fig. 8B and G) and slightly IL-1β in CA1 neurons (Fig. 8D and I) but did not induce microglia activation (Fig. 8A, E and F) nor increased TNFα in CA1 neurons (Fig. 8C and H). Injection of C-EV (Fig. 8) or of RIFAX-EV (Supplementary Fig. 3) did not activate microglia or astrocytes nor induced TNFα or IL-1β increase in CA1 neurons. The content of several pro-inflammatory proteins is selectively increased in MHE-EV and this increase is reversed by rifaximin treatment The contents of TNFα, IL-8, CDCP1, oncostatin M, CD6, PD-L1, HGF and, especially of CCL11 were increased in MHE-EV but not in NMHE-EV. Moreover, treating MHE patients with rifaximin reversed the increase of all these proteins (Fig. 9). The changes in other proteins are summarized in Supplementary Fig. 4 and Supplementary Tables 2, 3 and 4.Fig. 9The content of eight proteins is increased selectively in EV from MHE patients and is reversed by rifaximin treatment. The protein cargo of EV was analyzed by proximity extension assay. Data are the mean ± SEM and are reported as percentage of controls on an arbitrary Log2-based NPX (normalized protein expression) and adjusted to 2^NPX. One-way ANOVA followed by Fisher’s LSD post-hoc test was performed to compare all groups. Values significantly different from controls are indicated by asterisk (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001), from MHE-EV by “a”: a: p < 0.05, aa: p < 0.01, aaa: p < 0.001, aaaa: p < 0.0001, and from NMHE-EV by “b”: b: p < 0.05 The content of eight proteins is increased selectively in EV from MHE patients and is reversed by rifaximin treatment. The protein cargo of EV was analyzed by proximity extension assay. Data are the mean ± SEM and are reported as percentage of controls on an arbitrary Log2-based NPX (normalized protein expression) and adjusted to 2^NPX. One-way ANOVA followed by Fisher’s LSD post-hoc test was performed to compare all groups. Values significantly different from controls are indicated by asterisk (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001), from MHE-EV by “a”: a: p < 0.05, aa: p < 0.01, aaa: p < 0.001, aaaa: p < 0.0001, and from NMHE-EV by “b”: b: p < 0.05 Discussion There is increasing evidence that sustained peripheral inflammation results in neuroinflammation and cognitive impairment in many pathological situations, including multiple sclerosis, Alzheimer's, and Parkinson's diseases, rheumatoid arthritis or diabetes [11, 14, 38, 39]. We have previously shown that appearance of MHE in cirrhotic patients is associated with a shift in peripheral inflammation leading to increased activation of Th17 CD4 lymphocytes and increased plasma levels of pro-inflammatory cytokines [35]. Reversal of this shift in peripheral inflammation by treatment with rifaximin reverses MHE in cirrhotic patients [36]. This clearly support a key role for peripheral inflammation in triggering MHE. This idea is also supported by reports from other groups in cirrhotic patients and by studies in animal models of MHE [1, 5, 7, 8, 10, 12, 32, 33]. However, the mechanisms by which these alterations in the peripheral immune system are transmitted to brain to induce neuroinflammation and cognitive impairment remain unclear. Recent reports support that pathological EV from blood are transported to the brain to induce neuroinflammation, alterations in neurotransmission and cognitive impairment [21, 28–31]. EV are released by most cell types and play a key role in intercellular communication, especially in the immune system. EV contribute to the pathogenesis of many diseases, especially those associated to sustained inflammation, cancer metastasis, liver diseases and of neurodegenerative diseases [18, 20, 22, 31, 37]. It has been shown that EV from plasma may cross the blood–brain barrier and reach the microglia and neurons in hippocampus of injected rats [30]. The induction of pathological effects in brain and of cognitive and motor impairment by pathologic EV from blood seems a general process involved in many neurological diseases. We found that patients with liver cirrhosis show a lower amount of EV in plasma than control subjects, but the EV are larger in cirrhotic patients and contain more total protein in MHE patients than in controls. Similar results have been reported in cirrhotic patients by different groups using different methodologies [18, 47, 49]. A similar reduction in the concentration of EV in plasma [42] and an increased size [27], has been reported in other pathologies associated with sustained inflammation such as HIV. Weil et al [47] found that the decrease in the EV amount in blood of cirrhotic patients is due to the selective loss of a specific subset of small EV released by platelets, which produce most of the blood EV [3]. The loss of small EV results in an increased mean size of the remaining EV, explaining why cirrhotic patients show larger EV. Wei et al. [47] propose that, because low production of EV by platelets may affect all subtypes of EV, the low levels of small EV may be the consequence of an overconsumption of small platelet-derived EV. Small platelet-derived EV have a very strong pro-coagulant activity and may incorporate themselves either directly into microthrombi, or into circulating cells, resulting in decreased EV plasma levels. Cirrhotic patients show enhanced pro-coagulant environment [45] which could enhance small EV clearance. The same mechanism proposed by Weil et al [47] would explain the reduced amount and larger size of plasma EV found here in cirrhotic patients. We now also show that reduction in EV amount and the increase in EV size are completely reverted by long-term treatment with rifaximin. MHE patients treated with rifaximin show a similar concentration of EV in plasma than control subjects and also the size and total protein content are normalized by rifaximin, returning to values similar to control subjects. Rifaximin improves inflammation and complications of cirrhosis, including portal vein thrombosis [26, 36]. These beneficial effects of rifaximin reduce the pro-coagulant environment, which would normalize EV clearance, thus restoring the normal amount of small and total plasma EV. The restoration of the amount of small EV would be responsible for the normalization of the mean size of the EV in MHE patients treated with rifaximin. The EV from cirrhotic patients show a similar pattern of exosome markers (alix, flotillin2, CD9) than control subjects. However, they show a remarkable alteration in the protein cargo as analysed by proximity extension assay. The EV from patients with liver cirrhosis without or with MHE show altered content of 18 proteins. Moreover, the content of 8 proteins (TNFα, IL-8, CDCP1, oncostatin M, CD6, PD-L1, HGF and, especially of CCL11) was increased in MHE-EV but not in NMHE-EV and this increase was reverted by rifaximin treatment, suggesting that some of these proteins would play a key role in the pathologic effect of the EV from MHE patients. The results reported show that EV from patients with MHE induce a strong activation of both microglia and astrocytes, enhances the activation of all steps of the TNFα-TNFR1-S1PR2-IL1β-IL1 receptor pathway and alters the membrane expression of all subunits analysed of NMDA and AMPA receptors, leading to cognitive impairment in all parameters analyzed in the object location and object recognition tests, the Y-maze and in the radial maze. EV from cirrhotic patients without MHE induce a few of the above effects: induces activation of astrocytes but not of microglia; increases membrane expression of TNFR1 and IL1 receptor but not the other steps of the TNFα-TNFR1-S1PR2-IL1β-IL1 receptor pathway and alters membrane expression of NR2B and GluA1 but not of NR2A and GluA2 subunits of NMDA and AMPA receptor. EV from NMHE patients did not affect object location and object recognition and Y-maze tests nor reference or working memory in the radial maze but impaired learning index in the radial maze, inducing therefore only a very mild effect on cognitive impairment. EV from control subjects did not affect any of these parameters. These results, summarized in the graphical abstract, indicate that MHE in cirrhotic patients is associated with a pathological change in plasma EV and that EV from patients with MHE (MHE-EV) are enough to trigger cognitive impairment. This impairment is not induced by EV from control subjects or from cirrhotic patients without MHE, except for a mild impairment of learning index in the radial maze. It should be noted that patients are classified as with or without MHE according to the results on the PHES battery, recognized as the gold standard to diagnose MHE in the 11th World Congress. of Gastroenterology in 1998 [17]. However, we have shown that the PHES battery does not detect all cirrhotic patients with early neurological deficits, which are different amongst patients [19]. This suggests that the effects induced by NMHE-EV on astrocytes activation and on learning index would be due to a milder alteration in these EV associated to a mild cognitive impairment in the NMHE patients that is not being unveiled by the PHES battery [19]. It is noteworthy that treatment of MHE patients with rifaximin reverses the pathological effects of plasma EV, indicating that rifaximin reverses the mechanisms leading to the generation of pathological EV. We have already shown that MHE alters protein and miRNAs cargo of EV in plasma [18]. EV from plasma of MHE patients show an increase in pro-inflammatory molecules such as TNFα and a reduction in anti-inflammatory molecules such as TGFβ compared to control subjects or NMHE patients [18]. We show here that MHE-EV exhibit a selective increase in eight proteins, including TNFα, which are not increased in NMHE-EV and are normalized in RIFAX-EV from MHE patients treated with rifaximin. These proteins are therefore specifically associated with the pathological effects of MHE-EV. It is noteworthy that the results reported show that injection of MHE-EV to normal rats induces the same effects than hyperammonemia per se. We have already demonstrated in previous studies that increased TNFα levels and enhanced activation of the TNFα-TNFR1-S1PR2-IL-1β-IL1 receptor pathway is responsible for the altered membrane expression of NMDA and AMPA receptors in hippocampus and for cognitive impairment in hyperammonemic rats. Hyperammonemia increases TNFα levels and activation of this pathway both in hippocampus and cerebellum. We have shown that the steps of the pathway follow the indicated order because inhibiting each step with specific inhibitors or antagonists (anti-TNFα for TNFα, R7050 for TNFR1 signaling; JTE-013 for S1PR2; the IL1-receptor antagonist IL1-Ra for IL1 receptor) reverses the increase in the subsequent steps but not in previous steps of the pathway [7, 41, 43] and [2, 8, 30] and [28]. We have also shown that blocking TNFα with infliximab [7, 8, 41] or the IL1 receptor with IL1-Ra [44] in vivo reverses the alterations in NMDA and AMPA receptors and cognitive impairment in hyperammonemic rats. These studies clearly demonstrate that enhanced activation of the TNFα-TNFR1-S1PR2-IL-1β-IL1 receptor pathway is responsible for the altered membrane expression of NMDA and AMPA receptors in hippocampus and for cognitive impairment in hyperammonemic rats. We have also shown that injecting EV from plasma of hyperammonemic rats to normal rats also induces glial activation, neuroinflammation, glutamatergic neurotransmission and cognitive impairment by increasing TNFα and the TNFα-TNFR1-S1PR2-IL-1β-IL1 receptor pathway and altering membrane expression of NMDA and AMPA receptors. These effects are not induced if EV from hyperammonemic rats are pre-treated with anti-TNFα to block the action of TNFα [30] and [28]. This demonstrates that the TNFα in the pathologic EV is responsible for triggering the activation of the TNFα-TNFR1-S1PR2-IL-1β-IL1 receptor pathway and cognitive impairment in normal rats. We show here that EV from MHE patients also increases TNFα and the TNFα-TNFR1-S1PR2-IL-1β-IL1 receptor pathway, alter membrane expression of NMDA and AMPA receptors and induces cognitive impairment. Moreover, EV from MHE patients show increases levels of TNFα which are reversed by treating the patients with rifaximin. In parallel with the loss of TNFα, the EV-RIFAX also loss its pathogenic capacity and did not affect TNFα, the above pathway, glutamate receptors or cognitive function when injected to normal rats. Altogether our previous studies and the current work support that the increased levels of TNFα in MHE-EV would contribute to their deleterious effects, including the induction of cognitive impairment. Reversal of the increase in TNFα by rifaximin would explain the reversal of the pathologic effects in RIFAX-EVand points to a new key target to improve MHE in cirrhotic patients. Conclusions The results reported show that cirrhotic patients with MHE have pathological EV in blood, with altered protein content, which contribute to transmit to the brain the peripheral alterations in the immune system. EV from MHE patients show increased content of 8 proteins, including TNFα, which are not altered in EV from patients without MHE. MHE-EV, but not EV from patients without MHE trigger a strong cognitive impairment in rats. This is mediated by altered membrane expression of AMPA and NMDA glutamate receptors in hippocampus, due to increased neuroinflammation, with glial activation and enhanced activation of the TNFα-TNFR1-S1PR2-IL-1β-IL-1R pathway in hippocampus. The EV from MHE patients treated with rifaximin did not show increased TNFα levels and did not induce the above pathological effects. Rifaximin treatment reverses the changes in the cargo of the EV from plasma of MHE patients, including the increase in TNFα and eliminates the pathological effects of the EV. The data show that in cirrhotic patients the presence of MHE is associated with changes in their blood EV which transmit pathological effects to the brain. These pathological EV could be a new therapeutic target to improve MHE by modifying the EV content using rifaximin or by blocking the effects of TNFα. This report provides two relevant advances in the understanding of the pathological mechanisms and treatment of MHE. First: we have previously shown that appearance of MHE in cirrhotic patients is associated with a shift in peripheral inflammation and immunophenotype. However, how peripheral alterations are transmitted to brain to induce MHE, including cognitive impairment was not known. This work shows, for the first time, that EV from plasma of MHE patients are enough to induce cognitive impairment. This unveils a main new mechanism involved in the pathogenesis of MHE in cirrhotic patients. Second: we also show that treating MHE patients with rifaximin reverses the alterations in the cargo of plasma EV and the capacity of MHE-EVs to induce cognitive impairment. This provides a relevant advance on the mechanisms by which rifaximin induces beneficial effects in MHE patients and opens new possibilities to investigate other therapeutic treatments directed to manipulate the EV cargo to improve MHE. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 1. Supplementary Material 2.