Mesenchymal stem cell-derived extracellular vesicles attenuate liver transplantation-induced ischemia/reperfusion injury by suppressing hepatocellular complement C5 expression

Abstract

Background & aims

Hepatic ischemia–reperfusion injury (HIRI) is an unavoidable complication of liver transplantation (LT) that drives allograft dysfunction and mortality. Mesenchymal stem cell–derived extracellular vesicles (MSC-EVs) show therapeutic promise, yet their actions within the hepatic network remain incompletely defined. We investigated how MSC-EVs protect against LT-induced HIRI, focusing on intercellular communication and complement-mediated inflammation.

Methods

Single-cell RNA sequencing, single-nucleus ATAC-seq, and spatial transcriptomics were performed in rat LT and mouse HIRI livers with or without MSC-EV treatment. Cell migration assays, AAV-mediated overexpression, and PTPN13-knockdown MSC-EVs were used to prove cellular function and mechanisms.

Results

MSC-EVs attenuated HIRI, improving liver function and reducing histologic injury and hepatocyte apoptosis. Single-cell analyses showed selective curtailment of mononuclear phagocyte and neutrophil recruitment to injured liver. Mechanistically, HIRI induced hepatocyte C5 and C5a, which engaged C5aR1 on mononuclear phagocytes and neutrophils to drive chemotaxis. Consistent with this pathway, MSC-EVs delivered PTPN13 to hepatocytes, lowering C5 and promoting ERK1/2 dephosphorylation and enhanced C/EBPα activity. Elevated C/EBPα activity directly suppressed C5 transcription, whereas PTPN13-depleted MSC-EVs showed diminished efficacy.

Conclusions

We present a single-cell multi-omics atlas of MSC-EV therapy across rat LT and mouse HIRI, supported by clinical data. MSC-EVs mitigated HIRI by lowering hepatocyte C5 and reducing mononuclear phagocyte and neutrophil recruitment. MSC-EV delivery of PTPN13 dephosphorylated ERK1/2 and enhanced C/EBPα, repressing hepatocyte C5 and dampening C5a–C5aR1 signaling. These findings position MSC-EVs as a translational nanotherapy to improve graft outcomes after LT.

Graphical Abstract

Mesenchymal stem cell-derived extracellular vesicles attenuate liver transplantation-induced ischemia/reperfusion injury by suppressing hepatocellular complement C5 expression

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04388-0.

Introduction

Liver transplantation (LT) is the definitive, life-saving therapy for end-stage liver disease and selected hepatic malignancies. Despite substantial gains from advances in perioperative management, including surgical technique, care pathways, and immunosuppressive regimens, hepatic ischemia/reperfusion injury (HIRI) during LT remains unavoidable. Characterized by a sterile inflammatory cascade and hepatocellular apoptosis, HIRI predisposes recipients to early allograft dysfunction and contributes significantly to morbidity and mortality. Against the backdrop of a persistent global shortage of donor organs, there is growing emphasis on strategies that attenuate HIRI and preserve graft function.

Mesenchymal stem cells (MSCs) are a promising therapy for liver disease owing to their immunomodulatory and pro-regenerative properties. We previously demonstrated that MSC therapy effectively attenuates HIRI and offers a clinical translatable approach to mitigate severe post-transplant complications [1–5]. Increasing evidence indicates that extracellular vesicles (EVs; 30 to 300 nm in diameter) are principal mediators of MSC bioactivity by delivering functional cargo. MSC-derived EVs (MSC-EVs) recapitulate much of the theraepeutic efficacy of their parent cells while providing safety advantages, including reduced risks of tumorigenicity and embolism [6]. Our prior work showed that MSC-EVs alleviate HIRI through immune rebalancing and mitigation of oxidative stress [7–11]. They also show with additional potential to enhance regeneration and suppress ferroptosis in aged livers [12–14]. Nevertheless, most studies, including ours, have focused on a single cell type, whereas HIRI arises from coordinated interactions among multiple hepatic cell populations. This complexity obscures the mechanisms by which MSC-EVs confer protection and hinders rational integration with current clinical practice. Defining how MSC-EV therapy remodels the hepatic cellular network after LT-induced HIRI is therefore essential to identify dominant mechanisms, clarify how MSC-EVs mitigate injury, and enable clinical translation.

The advent of multi-omics has enabled comprehensive characterization of hepatic biology and intercellular interactions within the three-dimensional microenvironment. Building on our prior multi-omics studies, including single-cell RNA sequencing (scRNA-seq), spatial transcriptomics (ST), proteomics, and metabolomics, which systemically delineated temporal remodeling of the hepatic niche during aging [15–17], we here integrate, for the first time, scRNA-seq, ST, and snATAC-seq to construct the most comprehensive multi-omics atlas of MSC-EV therapy in LT-induced HIRI. This resource allowed us to identify cell type-specific responses to injury and treatment. Using this atlas, we demonstrate that HIRI triggers complement C5 overproduction, which promotes the recruitment of mononuclear phagocytes and neutrophils to the donor liver and exacerbates tissue damage. These pathological processes are effectively attenuated by MSC-EV treatment. Mechanistically, we found that Protein Tyrosine Phosphatase Non-Receptor Type 13 (PTPN13), enriched in MSC-EVs, suppresses ERK phosphorylation, reduces C/EBPα phosphorylation, enhances its transcriptional activity, and downregulates hepatic C5 expression. Together, our findings elucidate how MSC-EVs reprogram intercellular communication networks to alleviate LT-induced HIRI and highlight their potential to mitigate pathological injury during LT.

Methods and materials

Clinical specimen collection and data analysis

A total of 10 paired donor liver specimens (pre- and post-reperfusion) were included in this study. These specimens were collected from the Department of Hepatic Surgery and Liver Transplantation Center at the Third Affiliated Hospital of Sun Yat-sen University between March 2021 and March 2025. Pre-perfusion samples were obtained prior to cold perfusion, while post-reperfusion samples were collected within 2 hours following graft vascular reconstruction. Blood samples were collected from recipients after LT. The study protocol was approved by the Clinical Research Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University.

Animals

Wild-type C57BL/6J mice and Sprague-Dawley (SD) rats were obtained from Wukong Biotechnology (Jiangsu, China). The study included male 8-week-old mice and 3-month-old rats. All animals were housed in a specific pathogen-free (SPF) facility at the Animal Center of Guangzhou Gennio Biotechnology, under controlled conditions: temperature maintained at 20–25 °C, relative humidity around 50%, and a 12-hour light/dark cycle. All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Jennio Biotech (#JENNIO-IACUC-2025-A024), and were performed in accordance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals.

Establishment of HIRI mouse model

Following anesthesia induced with sodium pentobarbital (50 mg/kg, intraperitoneal), a hepatic ischemia-reperfusion injury (HIRI) model was established in mice by occluding the blood supply to the left and median liver lobes using a non-invasive vascular clamp for 90 min. Reperfusion was initiated by clamp removal. Immediately thereafter, mice in the MSC-EVs group received UC-MSC-EVs (100 µg in 100 µl PBS, containing approximately 1 × 10⁹ particles) via tail vein injection. Mice in the anti-C5a group were given 30 mg/kg C5a-neutralizing antibody during HIRI. All animals were euthanized 24 h after surgery, and blood and liver tissue samples were collected for subsequent analysis.

Establishment of orthotopic rat liver transplantation

The establishment of orthotopic rat liver transplantation were performed according to established protocols [18].

Donor Liver Procurement. Following intraperitoneal pentobarbital sodium anesthesia, a midline laparotomy was performed. The perihepatic ligaments were sequentially divided. The common bile duct (CBD) was isolated, stented, and secured. After further dissection of the portal vein (PV) and infrahepatic inferior vena cava (IHIVC), systemic heparinization was administered. The IHIVC and PV were transected sequentially, and the liver was promptly explanted.

Vascular Stent Placement: The explanted graft was immediately placed on ice. Polyethylene (PE) tubing was inserted and secured as a portal vein stent. Another PE tubing was fashioned as an infrahepatic IVC stent and firmly fixed in place with silk ties.

Orthotopic Liver Transplantation (Recipient Procedure): Recipient Procedure. Recipient rats were anesthetized intraperitoneally with pentobarbital sodium and positioned supine. The perihepatic ligaments, portal vein (PV), and the supra- and infrahepatic segments of the inferior vena cava (IVC) were mobilized; the hepatic artery was isolated, ligated, and transected. After clamping the recipient PV and cross-clamping the supra- and infrahepatic IVC, the native liver was rapidly removed. The donor liver was placed in the hepatic fossa, and the supra- and infrahepatic IVC anastomoses were completed. The graft PV stent was inserted into the recipient PV, secured with silk ties, and portal inflow was restored. The infrahepatic IVC stent was then inserted and secured in the same manner, re-establishing caval flow. Finally, the graft biliary stent was introduced into the recipient common bile duct and ligated to secure biliary drainage.

Immediately after reperfusion, MSC-derived extracellular vesicles (MSC-EVs) and a C5a-neutralizing antibody were administered.

Evaluation of liver injury

ALT and AST were measured using a Chemray 800 analyzer (Shanghai, China). Liver tissues were fixed in 4% paraformaldehyde, embedded in paraffin, sectioned at 8 μm, and stained with hematoxylin and eosin (H&E). Hepatic injury was evaluated using Suzuki’s scoring system, which grades three histological features—sinusoidal congestion, hepatocellular necrosis, and ballooning degeneration—on a 0–4 scale. For each specimen, five random fields were imaged, and the Suzuki score was calculated as the mean across these fields.

Isolation and characterization of human umbilical cord-derived MSCs (UC-MSCs)

Human umbilical cord mesenchymal stem cells (UC-MSCs) were isolated under sterile conditions according to established protocols and with approval from the Research Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University (RG2024-091-01) [3]. Following informed consent from the donors, fresh umbilical cords (UCs) were collected and stored in phosphate-buffered saline (PBS) at 4 °C. The cords were then rinsed with PBS to remove residual blood, minced into approximately 10 mm³ pieces, and digested for 4 h at 37 °C using type I collagenase supplemented with 0.1% hyaluronidase and 3 mM CaCl₂.

The digested tissue was cultured in low-glucose (1 g/L) Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum (FBS; Gibco) and maintained at 37 °C in a humidified 5% CO₂ atmosphere. The culture medium was replaced every three days to remove non-adherent cells. UC-MSCs were subsequently passaged to achieve purification.

For analysis of MSC surface markers, cells were harvested and washed twice with PBS containing 1% bovine serum albumin (BSA; Sigma-Aldrich, USA) to block non-specific binding. They were then stained in the dark at 4 °C for 30 min with the following fluorescently labeled monoclonal antibodies: FITC-CD90, PE-CD105, PE-CD73, FITC-CD45, FITC-HLA-DR, FITC-CD11b, FITC-CD19 and FITC-CD34. Stained cells were analyzed using a thirteen-color FACSCalibur flow cytometer (Beckman Coulter, USA), and data were processed with FlowJo software (TreeStar, USA).

To assess osteogenic and adipogenic differentiation potential, cells were induced using specific differentiation media (Gibco, USA) with medium changes every three days. After 21 days of induction, osteogenic differentiation was evaluated by Alizarin Red S staining, and adipogenic differentiation was assessed by Oil Red O staining.

Isolation and purification of UC-MSC-EVs

In this study, MSC-derived extracellular vesicles (MSC-EVs) were obtained from third-passage human umbilical cord mesenchymal stem cells (UC-MSCs) cultured in low-glucose (1 g/L) Dulbecco’s modified Eagle’s medium (DMEM; Gibco, Life Technologies, NY, USA; Product No. c11885500bt) supplemented with 10% fetal bovine serum (FBS; Gibco) under standard humidified conditions at 37 °C with 5% CO₂. To collect EVs, the cells were switched to low-glucose DMEM containing 10% exosome-depleted FBS and incubated for 48 h. The conditioned medium was then collected and subjected to sequential centrifugation: first at 3,200 × g for 10 min to remove cells and large debris, followed by a second centrifugation at 10,000 × g for 30 min to eliminate smaller particulate matter. Finally, UC-MSC-EVs were isolated via two rounds of ultracentrifugation at 100,000 × g for 2 h each. The purity of the isolated EVs was assessed according to a previously described method [7].

Quantification and identification of UC-MSC-EVs

1) Protein Quantification:

The total protein content of the purified UC-MSC-EVs was quantified using a bicinchoninic acid (BCA) assay kit (KeyGEN BioTECH, Jiangsu, China) in accordance with the manufacturer’s instructions.

2) Transmission Electron Microscopy (TEM):

The morphological features of UC-MSC-EVs were examined by transmission electron microscopy. Briefly, EVs were fixed in 4% paraformaldehyde (PFA) for 30 min at room temperature, washed three times with PBS, and then applied to formvar/carbon-coated grids for 20 min. Subsequently, the samples were fixed with 2% glutaraldehyde for 1 h and negatively stained with 3% phosphotungstic acid. After rapid dehydration in 50% ethanol and air-drying, the grids were imaged under a TEM.

3) Nanoparticle Tracking Analysis (NTA):

The particle size distribution and concentration of UC-MSC-EVs were determined using nanoparticle tracking analysis (NTA) on a NanoSight LM10 system (Malvern Instruments Ltd, UK) equipped with a sCMOS camera. Samples were diluted in distilled water to a concentration of 2 µl/µg and introduced into the instrument. System settings were calibrated according to the manufacturer’s software recommendations. Autofocus was applied to minimize blurred particles, and maximum particle capture was ensured during recording. Videos were acquired at 25 °C and analyzed using NTA software (version 3.1.54).

4) Western Blot Analysis:

Western blotting was performed to detect specific extracellular vesicle markers, including GRP94, TSG101, ALIX and CD63.

Enzyme-linked immunosorbent assay

C5a levels in both cell-culture supernatants and liver homogenates were measured using a commercial ELISA kit (MEIMIAN, Shanghai) according to the manufacturer’s instructions. Liver homogenates were prepared from frozen liver tissue samples. C5a concentrations were normalized to total protein content, which was determined using the BCA protein assay (Thermo Fisher Scientific). Results are expressed as pg/mg of protein.

Cell viability assay

Cell viability was quantified using the Cell Counting Kit-8 (CCK-8) according to the manufacturer’s protocol(Solarbio). MSCs were seeded into 96-well plates at a density of 2 × 10³ cells per well and treated with 10 µL of CCK-8 reagent. After incubation at 37 °C for 3 h, the optical density at 450 nm was recorded with a microplate reader to evaluate cellular metabolic activity.

Isolation and culture of mouse primary hepatocytes

Primary hepatocytes from C57BL/6 mice were isolated following established protocols [19]. Under sterile conditions, the liver was first flushed via portal vein perfusion with Hank’s Balanced Salt Solution (HBSS) to remove residual blood, followed by perfusion with pre-warmed collagenase IV (0.5 mg/mL) to digest the liver. After completion of digestion, the liver was excised, mechanically dissociated, and passed through a 70-µm cell strainer. Hepatocytes were enriched by differential centrifugation (50 ×g, 5 min, 4 °C; three cycles) to pellet parenchymal cells, which were then resuspended in 15 mL DMEM. A 90% Percoll density gradient was applied for further purification, followed by centrifugation at 200 ×g for 10 min at 4 °C. The purified hepatocyte pellet was collected for counting and subsequent culture.

Isolation and culture of rat primary hepatocytes

Under sterile conditions, the liver was first flushed via portal vein perfusion with Hank’s Balanced Salt Solution (HBSS) to remove residual blood, followed by perfusion with pre-warmed collagenase IV (0.5 mg/mL) to digest the liver. After completion of digestion, the liver was excised, mechanically dissociated, and passed through a 70-µm cell strainer. Hepatocytes were enriched by differential centrifugation (50 ×g, 5 min, 4 °C; three cycles) to pellet parenchymal cells, which were then resuspended in 15 mL DMEM. A 90% Percoll density gradient was applied for further purification, followed by centrifugation at 200 ×g for 10 min at 4 °C. The purified hepatocyte pellet was collected for counting and subsequent culture.

Animal adeno-associated virus-8 (AAV8) injection and processing

Hepatocyte-specific overexpression of Cebpa in mice and rats was achieved via tail-vein delivery of adeno-associated virus serotype 8 (AAV8) carrying a thyroxine-binding globulin (TBG) promoter–driven CEBPA cassette (AAV8-TBG-CEBPA). For mice, the vector encoded Cebpa (NCBI RefSeq NM_007678.4) under the TBG promoter; for rats, the vector encoded Cebpa (NCBI RefSeq NM_012524.3) likewise under the TBG promoter. Corresponding control vectors lacked CEBPA but with an mCherry tag. All recombinant AAV8 vectors were designed and purchased from PUZONGene Technology (Guangzhou, China). Four weeks after administration, the HIRI model was established in the respective AAV8-treated mice or rats. Because the AAV8 control vector encodes the red fluorescent protein mCherry, its in vivo biodistribution was assessed four weeks after administration using a Bruker small-animal optical imaging system, and successful hepatocellular overexpression of CEBPA was confirmed by Western blot analysis.

Western blotting assays

Western blotting was performed in accordance with published procedures [19]. Briefly, proteins were separated on 12% SDS–polyacrylamide gels (SDS-PAGE) and transferred to 0.45-µm PVDF membranes (Merck Millipore, Billerica, MA, USA). Membranes were blocked with 5% non-fat milk for 1 h at room temperature, incubated with primary antibodies overnight at 4 °C, washed three times with Tris-buffered saline containing Tween-20 (TBST), and then incubated with horseradish peroxidase (HRP)–conjugated secondary antibodies for 1 h at room temperature. After additional TBST washes, signals were developed using an enhanced chemiluminescence (ECL) substrate and captured with a FluorChem imaging system (ProteinSimple, California, USA). The following primary antibodies were used: Phospho-C/EBPalpha (Ser21) (2841T, CST), CEBPA (MA1-825, Thermo Fisher Scientific), PTPN13 (AF3577, R&D Systems), TSG101 (ab125011, abcam), CD63(ab271282, abcam), ALIX(YM8063, Immunoway), GRP94(#20292S, CST), phospho-ERK (#4370, CST), ERK (4695#, CST), and β-actin (RM2001, RuiKang Biotechnology). Detailed antibody information is provided in Supplementary Table 1.1.

RNA extraction and qRT-PCR

Total RNA was extracted with TRIzol (Invitrogen) according to the manufacturer’s instructions, quantified, and reverse-transcribed into cDNA using the Transcriptor First Strand cDNA Synthesis Kit (Accurate Biotechnology, Cat. No. AG11728), with 25 ng cDNA used as template. Quantitative real-time PCR was performed with the SYBR Green Pro Taq HS Premix qPCR Kit (Accurate Biotechnology, Cat. No. AG11701) in a 10-µL reaction volume; β-actin served as the internal control, and primer sequences are listed in Supplementary Table 1.3.

Plasmids, transfection, infection

The pCMV6-FLAG-MCS-Neo and pCMV-FLAG-CEBPA plasmids were obtained from Miaoling Bio (Wuhan, China). The pGL3-C5-Promoter-WT plasmids were synthesized by Sangon Biotech Co., Ltd. (Shanghai, China). siRNAs targeting CEBPA and PTPN13 were designed and synthesized by GenePharma (Shanghai, China). The transfection of plasmids and siRNAs were performed using jetPRIME (Polyplus-transfection) according to the manufacturer’s procedure. Supplementary Table 1.3 contains a list of all siRNA sequences.

Immunofluorescence staining

Paraffin-embedded liver sections were deparaffinized in xylene and rehydrated through a graded ethanol series (95%, 85%, 75%). Antigen retrieval was performed using buffers appropriate for each primary antibody. Sections were then blocked with 5% normal goat serum for 30 min at room temperature and incubated overnight at 4 °C with primary antibodies (C5a, ab11877, Abcam; CK18, ab181597, Abcam). Sections were incubated at room temperature for 1 h with Cy3- and FITC-conjugated secondary antibodies, counterstained and mounted using a DAPI-containing mounting medium, and subsequently examined and imaged using a fluorescence microscope(Zeiss, Germany).

Dual luciferase assay

To investigate the interaction between CEBPA and C5, we first constructed the pGL3-C5-Promoter-WT plasmid. 293T cells were co-transfected with pGL3-C5-Promoter-WT, pRL-TK, and either pCMV6-FLAG-MCS-Neo or pCMV6-FLAG-CEBPA using jetPRIME (Polyplus-transfection, France). Luciferase activities were then quantified with the Dual-Luciferase Reporter Assay System (Promega, USA), with Renilla luciferase serving as the internal control for normalization; both firefly and Renilla luminescence were measured strictly according to the manufacturer’s protocol.

Chromatin immunoprecipitation (ChIP) qPCR

Chromatin immunoprecipitation (ChIP) was performed using the ChIP Assay Kit (Beyotime, China) in accordance with the manufacturer’s instructions. In brief, more than 1 × 10⁷ primary hepatocytes were cross-linked with 1% formaldehyde for 10 min at room temperature, the reaction was quenched with glycine for 5 min at room temperature, and cells were washed twice with ice-cold PBS. Cells were lysed on ice for 10 min in SDS lysis buffer containing 1 mM PMSF, chromatin was sheared by sonication, and the lysate was clarified by centrifugation at 14,000 g for 5 min at 4 °C. The supernatant was diluted as recommended and 10% was reserved as input, while the remainder was incubated overnight at 4 °C with an anti-CEBPA antibody (2 µg, MA1-82, Thermo Fisher Scientific) or control IgG. After the recommended washes and elution, cross-links were reversed, DNA was purified, and precipitated fragments were analyzed by quantitative PCR. ChIP–PCR primers were designed with mouse GAPDH as a negative control; primer sequences are provided in the Supplementary Table 1.3.

In vitro hypoxia reoxygenation (H/R) model

To model hepatic ischemia–reperfusion injury (HIRI) in vitro, primary hepatocytes and macrophages or neutrophils underwent a hypoxia–reoxygenation (H/R) protocol adapted from established procedures [19]. In brief, cells were transferred to a humidified, temperature-controlled hypoxia chamber pre-equilibrated to 1% O₂, 5% CO₂, and 94% N₂ for 6 h, after which they were returned to normoxic culture conditions for a 4-hour reoxygenation phase; fresh, pre-warmed medium was exchanged at the onset of reoxygenation to minimize confounding by acidification and metabolites, and culture conditions were maintained at 37 °C throughout to ensure physiological relevance.

Immunohistochemistry (IHC) staining assays

Paraffin-embedded liver sections were deparaffinized in xylene and rehydrated through a graded ethanol series (95%, 85%, 75%). Antigen retrieval was performed using buffers appropriate for each primary antibody. Sections were then blocked with 5% normal goat serum for 30 min at room temperature and incubated overnight at 4 °C with primary antibodies (F4/80, ab300421, Abcam; CD68, ab283654, Abcam; Ly6G, YM837, Immunoway; Ly6G, GTX40912, GeneTex; MPO, YT5351). The following day, sections were incubated with the corresponding secondary antibodies for 1 h at 37 °C, immunoreactivity was visualized with DAB, and nuclei were counterstained with hematoxylin before dehydration and mounting.

Tracking MSC-EVs in the mouse and rats HIRI model

To track the biodistribution of MSC-EVs in rat and mouse HIRI models, MSC-EVs were labeled with the DiR membrane fluorescent probe (MB12482, Meilunbio) and administered via tail-vein injection to the mouse and rats HIRI model. At 24 h post-injection, in vivo whole-body fluorescence imaging was performed in mice using a Bruker small-animal optical imaging system (In-Vivo Xtreme II, Billerica, MA, USA) to assess biodistribution, after which hearts, lungs, livers, spleens, and kidneys were harvested from both species for ex vivo imaging, and DiR signal intensity in each organ was quantified on the same platform.

Cell-migration assay

Primary mouse hepatocytes, liver macrophages, and neutrophils were isolated and used for Transwell migration assays. A total of 1.5 × 10⁵ liver macrophages or 4 × 10⁵ neutrophils were seeded in the upper chamber, while 1 × 10⁵ primary hepatocytes were plated in the lower chamber and co-cultured. Two sets of experimental conditions were established: hypoxia–reoxygenation (H/R) control, H/R, and H/R with anti-C5a; and vector control versus CEBPA overexpression. At the predetermined time point, the insert was removed, cells were fixed and incubated with primary antibodies against F4/80 and Ly6G to label liver macrophages and neutrophils, respectively, and the cells were then counted under a fluorescence microscope.

10x Genomics snATAC-seq library construction, sequencing and analysis

Single-nucleus ATAC-seq was performed on the 10x Chromium platform using the Chromium Next GEM Single Cell ATAC Kit v2 (CG000496, 10x Genomics) following the manufacturer’s instructions. Fresh or frozen tissue was processed to isolate nuclei according to CG000505. Approximately 20,000 nuclei were selected for tagmentation, then loaded onto the Chromium Controller for GEM generation, linear amplification, and DNA cleanup per protocol. Libraries were constructed as described by the manufacturer, quantified by qPCR, and assessed with an Agilent Bioanalyzer and Qubit. Sequencing was carried out on an Illumina NovaSeq 6000 using 2 × 150 bp reads, targeting at least 25,000 fragments per nucleus. Cell Ranger ATAC v2.1.0 (10x Genomics)‘cellranger-atac count’ command with ‘force-cells’ settings was used for Read alignment, barcode counting, and peak calling with mm10 reference genome that generated fragments, barcodes metadata, and count matrix files. Downstream analyses were conducted in R withSignac v1.15.0 package in R [20].

For quality control, we computed standard metrics for each nucleus, including total fragments in peaks (peak_region_fragments), fraction of fragments in peaks (pct_reads_in_peaks), nucleosome banding pattern (nucleosome_signal), and TSS enrichment (TSS.enrichment). Nuclei were retained if peak_region_fragments were between 1,500 and 40,000, pct_reads_in_peaks exceeded 40%, the blacklist ratio was below 0.01, nucleosome_signal was below 2, and TSS.enrichment exceeded 4.

10x Genomics Visium HD library construction, sequencing and analysis

FFPE tissues were profiled using the Visium CytAssist Spatial Gene Expression platform for FFPE (Visium HD, 10x Genomics). Five 5 μm sections were cut, and RNA quality was assessed on an Agilent 2100 Bioanalyzer; sections with DV200 ≥ 30% were accepted. Sections mounted on plain glass slides underwent deparaffinization, H&E staining, and imaging per the Visium HD FFPE Tissue Preparation Handbook (CG000684). Probe hybridization and ligation, slide preparation, probe release and extension, library construction, and sequencing followed the Visium HD Spatial Gene Expression User Guide (CG000685). Libraries were sequenced on an Illumina NovaSeq 6000 with paired-end reads configured as 43 cycles for Read 1, 10 cycles for i7, 10 cycles for i5, and 50 cycles for Read 2. Space Ranger v3.0 mapped FASTQs to the reference, aligned sequencing data to microscope and CytAssist images, and generated gene–barcode matrices. These matrices were converted into Seurat objects using the Seurat v5.1.0 package in R, with 8-µm binned data employed for analysis following official Visium HD tutorials [21].

10x Genomics Chromium library construction, sequencing and analysis

Single-cell RNA-seq libraries were generated with the Chromium Next GEM Single Cell 3′ Reagent Kits v3 (10x Genomics). Cell suspensions were adjusted to 1,000 cells per µL and loaded onto the Chromium Controller to form gel bead–in–emulsions (GEMs) that isolate single cells with barcoded gel beads bearing 10x cell barcodes, unique molecular identifiers, and poly(dT) primers. Reverse transcription was performed in GEMs on a Veriti 96-well thermal cycler. cDNA was then amplified and libraries were prepared by fragmentation, end repair with A-tailing, adaptor ligation, index PCR, and SPRIselect purification. Final libraries were sequenced on an Illumina NovaSeq 6000 with 2 × 150 bp reads to approximately 500 million reads per library.

Raw scRNA-seq data were processed using 10× Genomics Cell Ranger (v7.1.0) with default parameters for demultiplexing, alignment to rat and human reference genomes, and UMI counting (by using the cellranger count function). The UMI count matrixes were transformed into Seurat objects using the R package Seurat v4.4.0 [21]. Cells expressing fewer than 200 genes or with mitochondrial gene content exceeding 10% of total UMIs were excluded. Doublets were identified and removed using DoubletFinder (v2.0.3) with the 92.5th percentile as the cutoff [22]. Additionally, cells with UMI counts over 40,000 or detecting more than 6,000 genes were filtered out. Filtered matrices were then merged for downstream analysis.

Each dataset was normalized using the SCTransform function in Seurat v4.4.0, and the heterogeneity associated with mitochondrial contamination was regressed out [21]. Features and anchors for downstream integration were selected according to the SCTransform pipeline. Datasets were integrated using canonical correlation analysis with FindIntegrationAnchors and IntegrateData functions. Principal component analysis was performed with RunPCA function, and appropriate principal components were selected for subsequent analysis. Dimensionality reduction was achieved using RunTSNE function, and a shared nearest neighbor graph was constructed with FindNeighbors function. Clustering was conducted using FindClusters function. Clusters with high mitochondrial gene expression, and putative cross-lineage multiplets were excluded. Prediction the transcriptional factor of C5 was conducted using TF-Target Finder (TFTF), and the binding sites were obtained by JASPAR database [23, 24]. Visualization was carried out using SCP (v0.5.1) (https://github.com/zhanghao-njmu/SCP). To identify significant signaling molecules among liver microenvironment populations, we utilized CellChat (v2.1.2) [25]. The normalized counts of merged liver samples were loaded into CellChat, and the data were pre-processed using the identifyOverExpressedGenes, indentifyOverExpressedInteractions functions based on the “CellChatDB.mouse” database. The computeCommunProb, computeCommunProbPathway and aggregeNet functions were applied in a sequential manner to infer significant ligand-receptor pairs across cell types. Additionally, to predict changes in potential ligand–target interactions, we employed the R package NicheNet (v2.1.0) with default parameters [26]. The top 20 predicted ligands, as well as their corresponding receptor genes were obtained.

Statistical analysis

Statistical analyses and graphing were conducted using GraphPad Prism 10 (GraphPad Software, USA). Group comparisons employed unpaired two-tailed t-tests (for two groups) or one-way ANOVA (for multiple groups), while human data were analyzed using the Mann-Whitney U test. Data are presented as individual values or mean ± SD. Correlation was assessed using Pearson’ s coefficient. Statistical significance was defined as p < 0.05, and exact p-values are marked above error bars.

Results

Cross-species multi-omics links MSC-EVs hepatoprotection to reduced mononuclear phagocytes and neutrophils recruitment

To evaluate the therapeutic efficacy of MSC-EVs against LT-induced HIRI, we employed a rat LT model and a mouse HIRI model and generated multimodal datasets, including single-cell RNA-seq (scRNA-seq; N = 18), single-nucleus ATAC-seq (snATAC-seq; N = 6), and 10x Genomics Visium HD spatial transcriptomics (N = 2) (Fig. 1A). Firstly, we isolated MSCs from umbilical cords, which displayed a fibroblast-like morphology and retained adipogenic and osteogenic differentiation potential (Fig. S1A). Flow cytometry confirmed an MSC immunophenotype—CD73⁺, CD90⁺, and CD105⁺ with concomitant absence of CD11b, CD19, CD45, CD34, and HLA-DR (Fig. S1B). MSC-CM was collected and subjected to sequential ultracentrifugation to isolate MSC-EVs. Characterization by transmission electron microscopy (TEM) revealed that MSC-EVs were round vesicles with a characteristic bilayer membrane, while nanoparticle tracking analysis (NTA) indicated a particle size distribution ranging from 30 to 300 nm (Fig. 1B-C). Western blotting corroborated successful EV isolation by detecting canonical exosomal markers (CD63, ALIX, TSG101) and the absence of the endoplasmic reticulum protein GRP94 (Fig. 1D).

Fig. 1.

Cross-species multi-omics links MSC-EVs hepatoprotection to reduced mononuclear phagocytes and neutrophils recruitment. (A) Schematic of the experimental workflow integrating single-cell RNA sequencing, single-nucleus ATAC-seq, and spatial transcriptomics to profile intrahepatic cells from a rat liver transplantation model and a mouse hepatic ischemia–reperfusion injury model. (B-C) Transmission electron microscopy (B) and nanoparticle tracking analysis (C) of umbilical cord mesenchymal stem cell–derived extracellular vesicles (UC-MSC-EVs) showing typical vesicular morphology and size distribution (Scale bar=1 μm and 0.6 μm). (D) Representative Western blot of the specific vesicle-related markers (CD63, ALIX, TSG101 and GRP94). (E) Representative IVIS images of heart, lung, liver, spleen, and kidneys harvested from rats after liver transplantation or mice after HIRI following injection with DiR-treated supernatant control (left) or DiR-labeled MSC-EVs (right). (F-G) Serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) in rats after liver transplantation (F) and in mice after HIRI (G) (n = 5 animals per group). (H–I) Representative H&E staining and TUNEL staining with quantification in rat liver transplantation (H) and mouse HIRI cohorts (I) (n = 5 animals per group). (J) t-SNE map of 14 hepatic cell clusters derived from rat samples spanning normal livers, liver transplantation, and MSC-EV–treated liver transplantation. (K) Corresponding distributions of cell types across the rat groups. (L) Immunohistochemistry for CD68 and Ly6G across rat groups. (M) t-SNE map of 14 hepatic cell clusters derived from mouse samples spanning normal livers, HIRI, and MSC-EV–treated HIRI. (N) Corresponding distributions of cell types across the mouse groups. (O) Immunohistochemistry for F4/80 and Ly6G across mouse groups ( n = 5 animals per group). Scale bar: 50 μm

Subsequently, the biodistribution of MSC-EVs in vivo was investigated. MSC-EVs were labeled with the near-infrared fluorophore DiR, and at 24 h post-administration fluorescence was predominantly localized to the liver in both models (Fig. 1E). Consistent with this biodistribution, serum biochemistry showed that MSC-EVs treatment attenuated hepatic injury in rats after LT and in mice after HIRI, with lower ALT and AST levels (Fig. 1F–G). Histopathology supported these findings in both species, as evidenced by reduced parenchymal damage on H&E staining and fewer TUNEL-positive hepatocytes (Fig. 1H–I). At the molecular level, spatial transcriptomics of injured mouse livers demonstrated suppression of canonical pro-inflammatory programs after MSC-EV therapy compared to the control group, including reduced expression of Casp1, Il1b, Nlrp3, and Relb (Fig. S2A). These results collectively demonstrate that MSC-EVs treatment effectively mitigates HIRI.

Given that HIRI involves a complex interplay of multiple cell types, single-cell RNA sequencing data was used to investigate whether MSC-EVs exert their effect by reshaping the hepatic microenvironment. At single-cell resolution, the rat liver scRNA-seq atlas identified 14 major populations, including hepatocytes, cholangiocytes, endothelial cells (ECs), hepatic stellate cells (HSCs), mononuclear phagocytes (MPs), conventional dendritic cells (cDCs), plasmacytoid dendritic cells (pDCs), neutrophils, T cells, natural killer (NK) cells, natural killer T (NKT) cells, B cells, plasma cells, and dividing cells (Fig. 1J; S3A). Prior studies have established that MPs and neutrophils play critical roles in exacerbating HIRI [27]. Accordingly, our scRNA-seq data revealed that MPs and neutrophils were modest at baseline yet expanded most after LT and became the dominant post-injury immune subsets. Both populations contracted with MSC-EV treatment, and IHC corroborated their increase after LT and reduction with MSC-EV therapy (Fig. 1K–L). Cross-species profiling in mouse livers showed an analogous 14-cluster architecture with conservation of the principal lineages (Fig. 1M; S3B). Immune shifts paralleled the rat data. MPs and neutrophils expanded and constituted the predominant immune subsets after injury. These populations contracted with MSC-EV therapy, which IHC likewise validated (Fig. 1N–O).

Together, these findings show that MSC-EVs mitigate LT-induced HIRI and that efficacy aligns with reduced recruitment of MPs and neutrophils.

MSC-EVs limit MPs and neutrophils recruitment by suppressing hepatocellular C5/C5a level

To delineate how MSC-EVs reduce recruitment of MPs and neutrophils during HIRI, we mapped cell–cell communication. Global interaction strength increased after HIRI and decreased with MSC-EV treatment (Fig. 2A). In parallel, DiR-labeled MSC-EVs localized predominantly to hepatocytes, indicating that MSC-EVs primarily act on hepatocytes in vivo and suggesting a hepatocyte-centered mechanism for dampening intercellular signaling and myeloid recruitment (Fig. 2B). Consistent with this rationale, communication analysis showed that hepatocytes provided the dominant post-injury inputs to MPs and to neutrophils, with signal strength exceeding that from other immune compartments in HIRI (Fig. 2C). With MSC-EV therapy, hepatocyte-to-MP and hepatocyte-to-neutrophil signaling declined markedly, supporting a model in which MSC-EVs blunt hepatocyte-derived cues that attract these cells (Fig. 2C).

Fig. 2.

MSC-EVs limit MPs and neutrophils recruitment by suppressing hepatocellular C5/C5a level. (A) Global intercellular communication strength within the hepatic microenvironment across mouse groups, inferred by ligand–receptor analysis. (B) Immunofluorescence showing the localization of DiR-labeled MSC-EVs in mouse liver sections. Magenta indicates DiR-labeled MSC-EVs, green indicates CK18, and blue indicates DAPI. (C) Comparative interaction networks in Sham, HIRI, and MSC-EV–treated HIRI mice. In the left network, red lines denote signaling interactions upregulated in HIRI versus Sham and blue lines denote downregulated interactions. In the right network, red lines denote interactions upregulated after MSC-EV treatment versus HIRI and blue lines denote downregulated interactions. Line thickness reflects the magnitude of change in interaction strength. (D) Top-ranked hepatocyte-derived ligands predicted to drive gene-expression changes in mononuclear phagocytes (left) and neutrophils (right) during HIRI. (E) Primary hepatocyte C5 mRNA levels by qPCR across mouse groups (left) and hepatic C5a content by ELISA across the same groups (right) ( n = 5 animals per group). (F) Primary murine hepatocytes subjected to in vitro hypoxia–reoxygenation with or without subsequent MSC-EVs treatment, showing C5 mRNA abundance (left) and C5a release to the culture medium (right). Three independent cell experiments. (G) Single-cell RNA-seq features showing, within mononuclear phagocytes and neutrophils across the three groups, the fraction of C5ar1-positive cells and corresponding expression levels. (H) Spatial transcriptomics showing the proportions of C1qa-positive mononuclear phagocytes, S100a8-positive neutrophils, and C5ar1-positive cells in HIRI and MSC-EV–treated HIRI livers. Scale bar: 50 μm

We next identified the hepatocyte ligands underlying these interactions. Among predicted hepatocyte ligands to MPs, C5 ranked first and was also prominent among ligands to neutrophils (Fig. 2D). Because C5 is cleaved to the chemoattractant C5a, we examined regulation of this axis [28]. In vivo, MSC-EV treatment reduced hepatocyte C5 mRNA and lowered hepatic C5a levels after HIRI (Fig. 2E). In vitro, primary mouse hepatocytes subjected to hypoxia/reoxygenation (H/R) showed increased C5 transcripts and C5a release, both of which were suppressed by treatment with MSC-EVs (Fig. 2F). Ligand–receptor pairing implicated hepatocyte C5 engaging C5aR1 on MPs and neutrophils (Fig. S4A). Concordantly, scRNA-seq showed that C5aR1⁺ MPs and neutrophils increased after HIRI and declined with MSC-EV therapy (Fig. 2G). Spatial transcriptomics further confirmed reduced accumulation of MPs and neutrophils after MSC-EVs treatment and a marked decrease in the fraction of C5aR1⁺ cells within myeloid aggregates (Fig. 2H).

Together, these data identify hepatocytes as the source of a C5a–C5aR1 axis that drives recruitment of MPs and neutrophils after HIRI and suggest that MSC-EVs mitigate this axis by lowering hepatocyte C5 expression.

Hepatocyte C5a correlates with graft injury, and C5a blockade reduces macrophage and neutrophil infiltration

To assess the clinical relevance of hepatocyte C5a to donor liver injury, we performed immunofluorescence on pre- and post-reperfusion samples during LT. The proportion of C5a⁺ hepatocytes was markedly higher after reperfusion (Fig. 3A). Consistent with this increase, cases with a higher C5a⁺ hepatocyte fraction had worse postoperative liver function. Serum ALT and AST were higher in the C5a-high group compared to the C5a-low (Fig. 3B). Extending to tissue pathology, the same cases showed greater histologic injury and more hepatocyte death, accompanied by increased infiltration of macrophages and neutrophils (Fig. 3C–D).

Fig. 3.

Hepatocyte C5a correlates with graft injury, and C5a blockade reduces macrophage and neutrophil infiltration. (A) Immunofluorescence of human liver samples before and after reperfusion. C5a in green, CK18 in red, and nuclei in blue with DAPI. (B) Trajectories of liver function over the first 7 days after transplantation in donor livers stratified by baseline C5a level, comparing C5a-high and C5a-low grafts. (C) H&E (up) and TUNEL (low) staining for C5a-low versus C5a-high donor livers, with quantitative analyses. (D) Immunohistochemistry for CD68 and MPO in C5a-low versus C5a-high donor livers, with quantitative analyses. (E) Serum ALT and AST in mice across three groups: Sham, HIRI, and HIRI with C5a neutralization. (F) Serum ALT and AST in rats across three groups: Sham, liver transplantation, and liver transplantation with C5a neutralization. (G-H) Mouse (G) and rat (H) liver sections stained with H&E (up) and TUNEL (low) across different groups, with quantitative analyses. (I-J) IHC for macrophages (up) and neutrophils (low) in mouse (I) and rat (J) livers across different groups, with quantitative analyses. Normal group, n = 10; Post-LT group, n = 10 (A–D). n = 5 animals per group (E–J). Scale bar: 50 μm

We then tested whether C5a derived from hepatocytes mediates the recruitment of macrophages and neutrophils. In transwell assays using sorted primary mouse hepatocytes, macrophages, and neutrophils, H/R of hepatocytes enhanced chemotaxis of both cell types, and this effect was markedly reduced by an anti-C5a neutralizing antibody (Fig. S5A). We next applied anti-C5a neutralizing antibody in vivo in mouse HIRI and rat LT models. C5a blockade decreased liver injury, reflected by lower ALT and AST, reduced histologic damage, and lowered hepatocyte apoptosis (Fig. 3E–H). Importantly, hepatic infiltration of macrophages and neutrophils was diminished in both models (Fig. 3I–J).

From a clinical perspective, these findings indicate that hepatocyte-derived C5a is associated with increased recruitment of macrophages and neutrophils and with worse graft injury after LT. Conversely, C5a blockade mitigates LT-induced HIRI via reducing intrahepatic macrophage and neutrophil accumulation.

MSC-EVs increase hepatocyte C/EBPα transcriptional activity to suppress C5 expression

To elucidate the mechanism by which MSC-EVs reduce C5 expression in hepatocytes during LT-induced HIRI, we identified candidate transcription factors that directly regulate C5 based on bioinformatic prediction and prioritized those enriched in hepatocytes. Only C/EBPα met these criteria (Fig. 4A). We then assessed whether MSC-EVs modulate C/EBPα activity in vivo. SnATAC-seq profiling of mouse livers after HIRI with or without MSC-EVs treatment revealed increased C/EBPα regulatory activity in the hepatocytes from treated group (Fig. S6A; 4B).

Fig. 4.

MSC-EVs increase hepatocyte C/EBPα transcriptional activity to suppress C5 expression. (A) Venn diagram showing the intersection of transcription factors that are relatively enriched in hepatocytes and predicted to regulate C5, identified via FIMO_JASPAR, ChIP-Atlas, and GTRD databases. (B) C/EBPα binding sequence logo (left) and snATAC-seq analysis comparing C/EBPα regulatory activity across HIRI, and MSC-EVs treated groups. (C) 293T cells were co-transfected with pGL3-C5 promotor and pRL-TK plasmids, plus pcDNA3.1 or Cebpa expression vector. Luciferase activity was measured 48 h post-transfection. (D) C/EBPα binding to the C5 promoter was analyzed by ChIP-qPCR in primary hepatocytes derived from mice. (E) Multi-level validation of Cebpa and C5 expression in mouse primary hepatocytes treated with a C/EBPα-overexpression plasmid. mRNA levels were quantified by RT-qPCR (β-actin as control), protein levels by western blotting, and secreted C5a levels in culture supernatants by ELISA. (F) Multi-level validation of Cebpa and C5 expression in mouse primary hepatocytes treated with Cebpa-targeting siRNA. mRNA levels were quantified by RT-qPCR (β-actin as control), protein levels by western blotting, and secreted C5a levels in culture supernatants by ELISA. (G) Macrophages and neutrophils migrating from the upper to the lower chamber were analyzed using fluorescence microscopy. Three independent experiments (C-G). Scale bar: 100 μm

To determine whether C/EBPα directly regulates the C5, we performed a dual-luciferase assay and found that its overexpression significantly repressed C5 promoter-driven activity, indicating that C/EBPα may function as a transcriptional repressor of C5 (Fig. 4C). On this basis, we mapped C/EBPα motifs within the C5 promoter to identify candidate binding sites. Using JASPAR, we predicted five putative C/EBPα binding sites within the C5 promoter (Fig. S6B). ChIP–qPCR in primary mouse hepatocytes confirmed C/EBPα occupancy at site 1 of C5 promoter (Fig. 4D). Having confirmed that C/EBPα directly binds the C5 promoter, we assessed its functional impact. Following H/R treatment, C/EBPα overexpression (confirmed by qPCR and immunoblotting) reduced both C5 mRNA levels and C5a release, indicative of its role as a transcriptional repressor (Fig. 4E). Conversely, C/EBPα knockdown elevated C5 mRNA and C5a release under the same conditions (Fig. 4F). In addition, C/EBPα overexpression in hepatocytes diminished macrophage and neutrophil chemotaxis in vitro (Fig. 4G).

Collectively, these data show that MSC-EVs enhance hepatocyte C/EBPα activity, which could directly repress C5 transcription and C5a production, thereby limiting macrophages and neutrophils chemoattraction.

Overexpression of hepatocyte-specific C/EBPα attenuated LT-induced HIRI by suppressing targeted chemotaxis of macrophages and neutrophils

To further investigate the role of C/EBPα in HIRI, we delivered an adeno-associated virus 8 (AAV8) vector designed for hepatocyte-specific C/EBPα overexpression in mice (Fig. 5A). One month after administration, whole-organ imaging demonstrated preferential hepatic accumulation of AAV8, and immunoblotting of isolated primary hepatocytes confirmed robust C/EBPα upregulation (Fig. S7A-B). We then established HIRI model. C/EBPα overexpression attenuated injury, as shown by improved liver function, reduced histological damage, and fewer apoptotic hepatocytes (Fig. 5B-C). Consistent with the proposed mechanism, the C5/C5a level in livers expression was suppressed and infiltration of macrophages and neutrophils decreased (Fig. 5D–E). To test generalizability, we administered AAV8–C/EBPα in rats (Fig. 5F). The AAV8 again accumulated in liver and increased hepatocellular C/EBPα expression (Fig. S7C-D). In the rat LT model, C/EBPα overexpression mitigated liver injury, reflected by lower serum enzyme levels and improved histopathology, accompanied by reduced hepatic C5/C5a and diminished macrophage and neutrophil infiltration (Fig. 5G–J).

Fig. 5.

Overexpression of hepatocyte-specific C/EBPα attenuated LT-induced HIRI by suppressing targeted chemotaxis of macrophages and neutrophils. (A) Schematic of the vector induction experimental workflow. Cebpa OE-AAV8 or control-AAV8 was injected 30 days prior to HIRI model induction in mice. (B) Serum levels of ALT and AST in each group. (C) Histology (H&E staining, up, left) and TUNEL staining (low, left) in liver tissues in each mouse group. Statistical analyses of Suzuki injury score and the percentage of TUNEL-positive cells in each field (right). (D) C5 transcript levels in primary mouse hepatocytes quantified by RT-qPCR (left), and C5a concentrations in liver tissue homogenates quantified by ELISA (right). (E) Immunohistochemistry for F4/80 and Ly6G in each group, with quantitative analyses. (F) Schematic of the vector induction experimental workflow. Cebpa OE-AAV8 or control-AAV8 was injected 30 days prior to liver transplantation model induction in rats. (G) Serum levels of ALT and AST in each group. (H) Histology (H&E staining, up, left) and TUNEL staining (low, left) in liver tissues in each rat group. Statistical analyses of Suzuki injury score and the percentage of TUNEL-positive cells in each field (right). (I) C5 transcript levels in primary rat hepatocytes quantified by RT-qPCR (left), and C5a concentrations in liver tissue homogenates quantified by ELISA (right). (J) Immunohistochemistry for CD68 and Ly6G in each group, with quantitative analyses. n = 5 animals per group (B–E, H–I)

In summary, these data demonstrate that hepatocyte-specific C/EBPα overexpression significantly attenuates LT-induced HIRI by modulating C5/C5a levels and the downstream chemotaxis of macrophages and neutrophils.

PTPN13 enriched within MSC-EVs downregulated C5/C5a level by affecting the ERK/ C/EBPα signaling pathway activation

To investigate how MSC-EVs enhance the transcriptional activity of C/EBPα in hepatocytes, we first analyzed chromatin accessibility at the Cebpa locus using snATAC-seq. The results revealed comparable promoter accessibility among the HIRI and MSC-EV groups (Fig. 6A). Consistent with this, immunoblotting of primary hepatocytes isolated from sham, HIRI, and MSC-EV–treated mice showed no significant change in C/EBPα protein abundance (Fig. 6B). These data indicate that MSC-EVs potentiate C/EBPα activity through post-translational mechanisms, such as protein modification or altered cofactor recruitment, rather than by modulating its expression. Since extracellular signal–regulated kinases 1/2 (ERK1/2) are known critical regulators of C/EBPα, and their dephosphorylation promotes C/EBPα transcriptional activity, we next evaluated the phosphorylation status of both ERK and C/EBPα [29–31]. We found that phosphorylation levels of both proteins were significantly increased following HIRI, which was reversed by MSC-EVs treatment (Fig. 6B). Furthermore, administration of SCH772984, an ERK inhibitor, not only suppressed C/EBPα phosphorylation but also downregulated hepatocellular C5/C5a levels (Fig. 6C).

Fig. 6.

PTPN13 enriched within MSC-EVs downregulated C5/C5a level by affecting the ERK/ C/EBPα signaling pathway activation. (A) snATAC-seq comparison of chromatin accessibility at the Cebpa locus across experimental groups (left) and across individual samples (right). (B) Western blot analysis of indicated target proteins in primary mouse hepatocytes isolated from the specified groups (n = 5 animals per group). (C) Primary mouse hepatocytes subjected to the indicated in vitro treatments, with intracellular target proteins (left), C5 mRNA levels by RT-qPCR (middle), and C5a concentrations in culture supernatants by ELISA (right). Three independent experiments. (D) Proteomic intersection of three independent MSC-EV preparations, showing shared proteins and the ten most abundant shared proteins. (E) Western blot of indicated target proteins in primary mouse hepatocytes after the in vitro treatments, with quantitative analyses. Three independent experiments. (F) Primary mouse hepatocytes after the in vitro treatment, with C5 mRNA levels by RT-qPCR (left) and C5a concentrations in culture supernatants by ELISA (right). Three independent experiments. (G) Macrophages and neutrophils migrating from the upper to the lower chamber were analyzed using fluorescence microscopy. Three independent experiments. Scale bar: 100 μm

EVs serve as critical mediators of intercellular communication by encapsulating diverse functional components. Our prior studies demonstrated that MSC-EVs mitigate HIRI through delivery of MnSOD2, CCT2, and functional mitochondria, while also enhancing proliferative capacity in senescent hepatocytes via DDX5 transfer [13]. Furthermore, we recently established that MSC-EVs ameliorate HIRI in aged models by delivering miR-1275 to suppress ferroptosis [12]. In this study, we performed high-solution LC-MS/MS analysis to investigate the proteome in MSC-EVs. As shown in Figs. 6D and 3911 proteins were detecting enriched in MSC-EVs. According to expression levels, protein tyrosine phosphatase non-receptor type 13 (PTPN13) was of particular interest among the top 10 proteins. PTPN13 has been demonstrated to exert phosphatase activity to dephosphorylate target proteins [32, 33]. Therefore, we speculated that MSC-EVs modulate C/EBPα activation, at least partially, through PTPN13-mediated dephosphorylation of ERK1/2. Western blot analysis confirmed the enrichment of PTPN13 expression in both MSCs and MSC-EVs (Fig. S8A). Furthermore, a lentivirus carrying PTPN13-siRNA was constructed to downregulate PTPN13 expression in MSCs, generating siPTPN13-transducted MSCs (Fig. S8B). Results confirmed that there were no significant differences in viability or MSC-related features between these siPTPN13-transducted MSCs and control MSCs (Fig. S8C). Subsequently, EVs were purified from the siPTPN13-transducted MSCs, which retained characteristic EV properties (Fig. S8D-F). When applied to primary mouse hepatocytes, MSC-EVs^siPTPN13 increased phosphorylation of C/EBPα and ERK1/2 and elevated C5 mRNA and C5a release after H/R compared with control MSC-EVs (Fig. 6E–F). In parallel, MSC-EVs^siPTPN13 enhanced hepatocyte-driven chemotaxis of macrophages and neutrophils in vitro (Fig. 6G).

Taken together, these data support a mechanism whereby PTPN13 in MSC-EVs promotes ERK1/2 dephosphorylation, preserves C/EBPα activity, and in turn suppresses C5/C5a induction in H/R-treated hepatocytes.

MSC-EV-derived PTPN13 mitigates LT-induced HIRI

Finally, to determine whether the protection conferred by MSC-EVs against LT-induced HIRI requires PTPN13 and the downstream ERK1/2–C/EBPα–C5 axis, we used MSC-EVs produced from PTPN13-silenced MSCs (MSC-EVs^siPTPN13). In the mouse HIRI model, PTPN13 knockdown partially abrogated MSC-EV hepatoprotection, as indicated by higher serum hepatic enzyme levels, aggravated histopathology, and increased hepatocyte apoptosis (Fig. S9A and Fig. 7A). In line with an effect on inflammatory recruitment, immunofluorescence showed that the MSC-EV–mediated reduction in hepatic macrophage and neutrophil infiltration was lost with PTPN13 silencing (Fig. 7B). Consistent with the proposed mechanism, livers treated with MSC-EVs^siPTPN13 exhibited increased ERK1/2 and C/EBPα phosphorylation together with higher C5 and C5a levels (Fig. 7C–D). These findings were corroborated in a rat LT model, where PTPN13 silencing similarly attenuated the protective efficacy of MSC-EVs (Fig. 7E–G and Fig. S9B). Collectively, these data demonstrate that PTPN13 delivered by MSC-EVs is essential for attenuating HIRI and for limiting macrophage and neutrophil chemotaxis to donor livers during LT.

Fig. 7.

MSC-EV-derived PTPN13 mitigates LT-induced HIRI. (A) Histology (H&E staining, up) and TUNEL staining (low) in liver tissues in each mouse group. Statistical analyses of Suzuki injury score and the percentage of TUNEL-positive cells in each field. (B) Immunohistochemistry for F4/80 and Ly6G in each group, with quantitative analyses. (C) Western blot analysis of indicated target proteins in primary mouse hepatocytes isolated from the specified groups. (D) C5 transcript abundance in primary mouse hepatocytes from the indicated groups quantified by RT-qPCR (left) and C5a concentrations in liver tissue homogenates quantified by ELISA (right). (E) Histology (H&E staining, up) and TUNEL staining (low) in liver tissues in each rat group. Statistical analyses of Suzuki injury score and the percentage of TUNEL-positive cells in each field. (F) Immunohistochemistry for CD68 and Ly6G in each group, with quantitative analyses. (G) C5 transcript abundance in primary rat hepatocytes from the indicated groups quantified by RT-qPCR (left) and C5a concentrations in liver tissue homogenates quantified by ELISA (right). n = 5 animals per group

Discussion

HIRI is an unavoidable event during LT and a major driver of early graft dysfunction and rejection. However, therapeutic options remain limited owing to incomplete mechanistic insight. This study demonstrated for the first time that MSC-EVs have the potential to mitigate HIRI during LT. Integrating scRNA-seq, snATAC-seq and ST, we show that MSC-EVs mitigate HIRI in part by lowering hepatocellular C5/C5a output, a potent chemoattractant that fuels macrophage and neutrophil influx and amplifies tissue injury. We further demonstrate that PTPN13, enriched in MSC-EVs, attenuates ERK1/2 activity, restores C/EBPα transactivation, and represses C5 transcription. These data position MSC-EVs as a nanotherapeutic approach to protect donor livers during LT.

Macrophages and neutrophils are established drivers of LT-induced HIRI. Previous studies show that macrophage-mediated sterile inflammation during the post-transplant period is pivotal [34, 35]. In addition, our work and others’ demonstrate that neutrophil activation aggravates injury through extracellular trap formation and inflammatory cytokines [15, 36, 37]. What remained unclear was the upstream cue coordinating their recruitment. By systematically comparing donor liver transcriptomes before and after transplantation, we define a hepatocyte-centric signal that strengthens hepatocyte–myeloid communication post-transplant. Hepatocytes upregulate and release C5/C5a, which engages C5aR1 on macrophages and neutrophils. Although C5/C5a has been linked to hepatocyte apoptosis, innate immune activation, prothrombotic coagulation, and donor-specific antibody production in LT, our data show that C5/C5a also directs macrophage and neutrophil chemotaxis toward injured hepatocytes. This hepatocyte-initiated axis provides a mechanistic bridge from tissue damage to inflammatory amplification in HIRI. Consistent with this model, hepatic C5 levels correlate with post-transplant injury severity, and blockade of the C5a–C5aR1 pathway reduces myeloid recruitment and mitigates LT-induced HIRI.

As key secretory products of MSCs, MSC-EVs mitigate HIRI through multiple pathways. Here we show that administered MSC-EVs are predominantly taken up by hepatocytes and, consequentially, lower hepatocellular C5/C5a levels. This suppression attenuates the chemotaxis of macrophages and neutrophils toward injury sites, thereby curbing the initiation of the inflammatory cascade. Mechanistically, we observed dynamic shifts in C/EBPα activity from model induction through MSC-EV treatment. C/EBPα is a key transcription factor regulating essential cellular processes, including differentiation, proliferation, and tumorigenesis [38]. In hepatology, C/EBPα has ben reported to control hepatocyte proliferation and differentiation by modulating gluconeogenesis and DNA-binding activity [39, 40]. Moreover, Ren et al. and Liu et al. suggested its potential role in aging-related exacerbation of liver injury [41, 42]. In this study, we define C/EBPα as a transcriptional regulator of C5. During LT-induced HIRI, hepatocellular C/EBPα activity is suppressed, driving up C5/C5a, whereas MSC-EV treatment restores C/EBPα activity and lowers C5. Hepatocyte-specific C/EBPα overexpression corroborates this regulatory link.

Building on our prior work showing that MSC-EVs attenuate HIRI by delivering MnSOD, CCT2, and functional mitochondria, we recently found that EV-mediated transfer of DDX5 restores senescent hepatocyte function to promote regeneration, while delivery of miR-1275 suppresses ferroptosis in aged HIRI [12, 13]. To explain how MSC-EVs regulate C/EBPα and downstream C5/C5a, we performed quantitative proteomics of EV cargo and identified PTPN13 as a leading candidate. PTPN13 is a protein tyrosine phosphatase that dephosphorylates phosphotyrosine residues, and acts as a tumor suppressor that restrains aggressiveness and overcomes resistance to tyrosine kinase inhibitors [32, 43–45]. It also confers Fas resistance in lung fibroblasts and promotes pulmonary fibrosis, and siPTPN13-loaded polymeric micelles show efficacy in idiopathic pulmonary fibrosis [46, 47]. In this study, EV-derived PTPN13 translocates into hepatocytes and lowers C5 production after H/R. Mechanistically, it dephosphorylates ERK1/2, a key regulator of C/EBPα phosphorylation, thereby enhancing C/EBPα transactivation and repressing C5 transcription [48, 49]. Together, these findings identify PTPN13 as a central EV effector and a tractable target to protect against LT-induced HIRI by reducing hepatocellular C5/C5a and limiting macrophage and neutrophil recruitment.

In summary, this study integrates multi-omics analyses to systematically investigate the protective mechanism of MSC-EVs in attenuating LT-induced HIRI. We identify that the protective effects of MSC-EVs are primarily mediated through PTPN13 delivery. The transferred PTPN13 suppresses hepatocyte C5/C5a levels by dephosphorylating ERK1/2, thereby modulating C/EBPα transcriptional activity, which ultimately reduces macrophage and neutrophil chemotactic infiltration. By constraining the complement cascade, MSC-EVs blunt inflammatory amplification and mitigate HIRI pathogenesis. These findings support MSC-EVs as a translational candidate to improve donor liver quality during transplantation.

Conclusions

In conclusion, integrated single-cell multi-omics across rat LT and mouse HIRI models demonstrates that MSC-EVs mitigate LT-induced HIRI by modulating hepatocyte-myeloid communication. HIRI increased hepatocellular C5 expression, promoting C5aR1-dependent chemotaxis of mononuclear phagocytes and neutrophils and exacerbating inflammatory injury. MSC-EV treatment suppressed this complement-associated recruitment program, resulting in reduced myeloid infiltration and improved graft injury indices. These findings support MSC-EVs as a mechanism-defined translational nanotherapy to improve graft outcomes after liver transplantation.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

YL, JZhe and RL initiated the project; FZ, JX, ZW and TW designed experiments; FZ, JX, ZW, TW, QY, HC and JZha carried out experiments and analyzed the experimental data; YL analyzed the single-cell sequencing data; RL and JZhe verified statistics; YL, RL, FZ and JX wrote the paper. SY, YY and JZhe reviewed the manuscript and participated in the interpretation of data.

Funding

This work was supported by the grants from the National Key Research and Development Program of China (2024YFA1107200); Guangzhou Clinical High and New Technology, Key and Specialty Projects (2026P-GX007); the National Natural Science Foundation of China (U24A20655, 82470666, 82400770, 82400771, 82300747); Guangzhou Basic and Applied Basic Research Project Co-funded by Municipal Schools (institutes) (2024A03J0273); Natural Science Foundation of Guangdong Province (2024A1515011662, 2025A1515012524); Youth Talent Cultivation Program of 3rd Affiliated Hospital of SYSU (2024GZRPYMS07).

Data availability

The datasets analyzed during the current study are available from the corresponding author on reasonable request. This study did not use any unique codes, and all analyses were performed in R and Python using standard protocols from previously published packages. All primers and other nucleic acid sequences are provided in the Supplement Table 1.

Declarations

Competing interests

The authors declare no competing interests.