Small Extracellular Vesicles Engineered Using Click Chemistry to Express Chimeric Antigen Receptors Show Enhanced Efficacy in Acute Liver Failure

ABSTRACT

Acetaminophen (APAP) overdose can cause severe liver injury and life‐threatening conditions that may lead to multiple organ failure without proper treatment. N‐acetylcysteine (NAC) is the accepted and prescribed treatment for detoxification in cases of APAP overdose. Nonetheless, in acute liver failure (ALF), particularly when the ingestion is substantial, NAC may not fully restore liver function. NAC administration in ALF has limitations and potential adverse effects, including nausea, vomiting, diarrhoea, flatus, gastroesophageal reflux, and anaphylactoid reactions. Mesenchymal stromal cell (MSC)‐based therapies using paracrine activity show promise for treating ALF, with preclinical studies demonstrating improvement. Recently, MSC‐derived extracellular vesicles (EVs) have emerged as a new therapeutic option for liver injury. MSC‐derived EVs can contain various therapeutic cargos depending on the cell of origin, participate in physiological processes, and respond to abnormalities. However, most therapeutic EVs lack a distinct orientation upon entering the body, resulting in a lack of targeting specificity. Therefore, enhancing the precision of natural EV delivery systems is urgently needed. Thus, we developed an advanced targeting technique to deliver modified EVs within the body. Our strategy aims to employ bioorthogonal click chemistry to attach a targeting molecule to the surface of small extracellular vesicles (sEVs), creating exogenous chimeric antigen receptor‐modified sEVs (CAR‐sEVs) for the treatment. First, we engineered azido‐modified sEVs (N₃‐sEVs) through metabolic glycoengineering by treating MSCs with the azide‐containing monosaccharide N‐azidoacetyl‐mannosamine (Ac4ManNAz). Next, we conjugated N₃‐sEVs with a dibenzocyclooctyne (DBCO)‐tagged single‐chain variable fragment (DBCO‐scFv) that targets the asialoglycoprotein receptor (ASGR1), thus producing CAR‐sEVs for precise liver targeting. The efficacy of CAR‐sEV therapy in ALF models by targeting ASGR1 was validated. MSC‐derived CAR‐sEVs reduced serum liver enzymes, mitigated liver damage, and promoted hepatocyte proliferation in APAP‐induced injury. Overall, CAR‐sEVs exhibited enhanced hepatocyte specificity and efficacy in ameliorating liver injury, highlighting the significant advancements achievable with cell‐free targeted therapy.

Our research confirmed the effectiveness of targeted small extracellular vesicles (sEVs) therapy in acute liver failure (ALF) models by targeting asialoglycoprotein receptor 1 (ASGR1), expressed on the surface of hepatocytes. Azido‐modified sEVs (N₃‐sEVs) were purified from the conditioned medium of pcMSCs incubated with Ac4ManNAz. ASGR1‐specific DBCO‐scFv was conjugated onto N₃‐sEVs to generate chimeric antigen receptor‐modified sEVs (CAR‐sEVs). These CAR‐sEVs exert hepatoprotective effects by delivering sEVs containing therapeutic cargos to the liver and enhancing hepatocyte repair in ALF through targeted delivery.

Abbreviations

Abbreviation

Definition

Ac4ManNAz

N‐azidoacetyl‐mannosamine

ALF

acute liver failure

APAP

acetaminophen

ASGR1

asialoglycoprotein receptor 1

CAR‐sEVs

chimeric antigen receptor‐modified small extracellular vesicles

DBCO

dibenzocyclooctyne

DBCO‐scFv

dibenzocyclooctyne‐tagged single‐chain variable fragment

EVs

extracellular vesicles

sEVs

small extracellular vesicles

MSCs

mesenchymal stromal cells

NAC

N‐acetylcysteine

NAPQI

N‐acetyl‐p‐benzoquinone imine

N₃‐sEVs

azido‐modified small extracellular vesicles

pcMSCs

placenta choriodecidua membrane‐derived mesenchymal stromal cells

pcMSC‐CM

conditioned medium derived from placenta choriodecidual membrane‐derived mesenchymal stromal cells

pcMSC‐sEVs

small extracellular vesicles derived from placenta choriodecidual membrane‐derived mesenchymal stromal cells

scFv

single‐chain variable fragment

SPAAC

strain‐promoted alkyne‐azide cycloaddition

1. Introduction

Acute liver failure (ALF) is a critical condition characterized by a rapid decline in liver function, often leading to multiple organ failure and posing a life‐threatening risk. In regions such as Europe, the United States, and Japan, the fatality rate for ALF is as high as 50% (Hosack et al. 2023; Stravitz and Lee 2019). Drug‐induced liver injury is frequently caused by lipophilic drugs such as acetaminophen (APAP), amiodarone, and statins, which undergo biomodification into reactive metabolites. This process leads to oxidative stress, mitochondrial dysfunction, and bile acid homeostasis disturbances (Hosack et al. 2023). APAP toxicity is a leading cause of drug poisoning worldwide, with excessive ingestion resulting in significant liver damage and ALF, particularly in Western countries (Ramachandran and Jaeschke 2019). While 90% of APAP is safely metabolized, 5%–9% is processed by CYP2E1, producing the highly reactive byproduct N‐acetyl‐p‐benzoquinone imine (NAPQI) (Ramachandran and Jaeschke 2021). The accumulation of NAPQI induces oxidative stress and cell death, and failure to manage APAP overdose may necessitate liver transplantation for survival (Lee 2017; Woolbright and Jaeschke 2017). N‐acetylcysteine (NAC) is administered as a clinical remedy for APAP intoxication to prevent and alleviate drug‐induced liver injury. While NAC has been shown to reduce APAP‐induced hepatotoxicity and mortality rates, there are limitations and potential adverse reactions associated with its use (Akakpo et al. 2022; Fisher and Curry 2019; Lewis et al. 2022). In cases of ALF, particularly following a significant overdose of APAP, NAC may not fully restore liver function; prolonged administration can hinder recovery and potentially lead to hepatocyte vacuolation and delayed liver regeneration (Wang et al. 2022; Akakpo et al. 2021). Additionally, intravenous NAC can cause rate‐related anaphylactoid reactions in up to 18% of patients (Kerr et al. 2005), and overdose has resulted in severe complications, including cerebral edema, seizures, uncle herniation, and permanent brain damage (Heard and Schaeffer 2011). Therefore, there is a continued need to develop an advanced treatment protocol for ALF.

Mesenchymal stromal cells (MSCs) have emerged as a promising therapeutic strategy for APAP‐induced liver failure (Hu et al. 2020; Hwang et al. 2019; Cai et al. 2020; Lin et al. 2022), demonstrating regenerative capabilities and extensive anti‐inflammatory and immune‐modulatory properties (Wang and Chen 2023; Zhou et al. 2021). The utilization of MSC‐based therapy has the potential to provide a therapeutic approach for treating drug‐induced liver diseases due to their ability to facilitate liver regeneration and suppress liver inflammation—primarily through paracrine effects (Harrell et al. 2019). This process may involve the release of extracellular vesicles (EVs) containing therapeutic substances such as miRNA, mRNA, DNA, proteins, and metabolites (Nikfarjam et al. 2020; Harrell et al. 2019; Liu et al. 2022). There has been a significant increase in studies investigating the function of EVs derived from MSCs, including small extracellular vesicles (sEVs), microvesicles, and apoptotic bodies (Lai et al. 2022; Wei et al. 2021). Among these, sEVs, ranging in size from 30 to 150 nm, play a crucial role in modulating cell interactions and their microenvironments (Han et al. 2023). They are involved in essential processes such as intercellular signaling, cell‐cell communication, immunomodulation, and the transfer of genetic information (Lotfy et al. 2023; Liu and Wang 2023). MSC‐derived sEVs can serve as therapeutic agents against various diseases due to their diverse cargo, including signalling proteins, genetic materials, and some small molecules (Bian et al. 2022; Veerman et al. 2021; Cao et al. 2021). They mimic the biological functions of MSCs, offering a viable substitute for whole‐cell therapy while reducing the risk associated with live‐cell administration.

However, sEVs lack inherent homing effects and are nondirectional, leading to the accumulation of specific body parts influenced by factors such as their source, size, and methods of injection (Sharma and Mukhopadhyay 2024). Therefore, enhancing the tissue penetration and organ specificity of sEVs to increase the local concentration of therapeutic substances at the target site is crucial for overcoming the limitations of EV therapy (Huang et al. 2023; Kooijmans et al. 2016). By achieving precise targeting and improving therapeutic effectiveness, cell‐free therapy has the potential to provide more treatment options for various diseases.

Recently, we have successfully established a serum‐free selection/culture method to generate human placenta choriodecidual membrane‐derived mesenchymal stromal cells (pcMSCs) (Su et al. 2017). The cells, pcMSCs, were produced in GTP Lab under clinical‐grade for the compassionate treatment of acute respiratory distress syndrome (ARDS) patients (Chen et al. 2022; Tsai 2024), and the Phase I clinical trial has been approved by the US Food and Drug Administration (Therapeutics 2023). The therapeutic potential of pcMSCs (Lin et al. 2019; Liou et al. 2019; Huang et al. 2020), along with their EVs and secretomes (Chang et al. 2021; Chiang et al. 2022; Le et al. 2024), in immune modulation strongly motivates us to advance our research into precision pcMSC‐derived chimeric antigen receptor‐modified sEVs (CAR‐sEVs) therapy for ALF.

In this study, we focused on the development of an advanced targeting technique to deliver EVs specifically to the liver. Our approach utilized click chemistry to conjugate a targeting material to the surface of sEVs derived from pcMSCs, resulting in CAR‐sEVs. Click chemistry, introduced in 2001 by Sharpless et al. (Kolb et al. 2001), offers a highly selective reaction under mild aqueous conditions, yielding favourable rates and yields compared to traditional methods. It provides versatile building blocks, such as azides and alkynes, suitable for both small‐ and large‐scale organic synthesis. A bioorthogonal click reaction‐based drug delivery system has recently entered Phase II clinical trials (Peplow 2023), highlighting the potential of click reactions for clinical applications (Yan et al. 2024). Our method involves three key steps: first, we introduced azido moieties onto the surface of sEVs (N₃‐sEVs) by treating pcMSCs in serum‐free culture with the azide‐containing monosaccharide, N‐azidoacetyl‐mannosamine (Ac4ManNAz), via metabolic glycoengineering. Second, we synthesized a single‐chain variable fragment (scFv) with a dibenzocyclooctyne (DBCO) motif (DBCO‐scFv) that targets asialoglycoprotein receptor 1 (ASGR1). The recombinant scFv is known for its high specificity and compact size (< 25 kDa), making it suitable for nanoparticle engineering and offering superior tissue penetration compared to conventional antibodies (De Aguiar et al. 2021; Chen et al. 2018). The chosen target, ASGR1, is predominantly located on the membrane of hepatocytes, constituting 80% of liver cells (Zhang et al. 2024; Witzigmann et al. 2016; D'Souza and Devarajan 2015). Finally, the ASGR1‐/specific DBCO‐scFv conjugates were click‐reacted with N₃‐sEVs via strain‐promoted alkyne‐azide cycloaddition (SPAAC) to produce CAR‐sEVs (Kim and Koo 2019). This advancement highlights the potential of CAR‐sEVs to selectively target specific cells, tissues, or molecules in clinical applications. Using this chemical conjugation strategy, we engineered CAR‐sEVs that improve therapeutic efficacy for ALF by specifically targeting the liver‐specific marker ASGR1.

2. Materials and Methods

Alexa Fluor 488‐conjugated anti‐His tag antibody (Cat No. 652509) was purchased from BioLegend (San Diego, CA, USA). Recombinant anti‐ASGR1/HL‐1 antibody (Cat No. ab256352), anti‐6X His tag antibody (Cat No. ab18184), HRP anti‐6X His tag antibody (Cat No. ab1187), and a GSH assay kit (Colorimetric) (Cat No. ab239727) were purchased from Abcam (Cambridge, UK). ASGR1 recombinant human protein (Cat No. NM_001671) was purchased from OriGene (Rockville, MD, USA). APAP (Cat No. A7085), the PKH26 Red Fluorescent Cell Linker Kit (Cat No. PKH26GL), DiR', DiIC18(7) (1,1'‐dioctadecyl‐3,3,3',3'‐tetramethylindotricarbocyanine iodide) (Cat No. D12731), and D‐(+)‐galactose (Cat No. G0750) were purchased from Sigma‒Aldrich (St. Louis, MO, USA). Ac4ManNAz‐tetraacylated (Cat No. 1084), AZDye 488 DBCO (Cat No. 1278), DBCO Amine (Cat No. A103), and Biotin‐PEG3‐Azide (Cat No. AZ104) were purchased from Click Chemistry Tools (Scottsdale, AZ, USA). The CellTiter‐Glo Luminescent Cell Viability Assay Kit (Cat No. G7570) was purchased from Promega (Madison, WI, USA). The Pierce BCA Protein Assay Kit (Cat No. 23225) was purchased from Thermo Scientific (Rockford, IL, USA). The human hepatoblastoma cell line C3A (HepG2/C3A) was obtained from the Bioresource Collection and Research Center (BCRC), Taiwan. A 3 kDa MWCO Amicon centrifugal filter (Cat No. UFC5003) was purchased from Merck KGaA (Darmstadt, Germany). Paraformaldehyde (8%, Cat No. 157–8) was purchased from Electron Microscopy Sciences (Hatfield, PA, USA). Alexa Fluor 488‐conjugated AffiniPure Donkey Anti‐Rabbit IgG (Cat No. 711‐545‐152) was purchased from Jackson ImmunoResearch Laboratories (Pennsylvania, USA). FITC Annexin V (Cat No. 560931) and Annexin V binding buffer (Cat No. BDB556454) were purchased from Biosciences (BD Pharmingen, USA). DuoSet ELISA Ancillary Reagent Kit 2 (Cat No. DY008) was purchased from R&D Systems (Minneapolis, MN, USA).

2.1. Generation and Characterization of Human Placenta pcMSCs

In this study, human placentas were obtained from women undergoing caesarean sections at Taipei Medical University Hospital. The collection and use of placental tissues were conducted following the protocols approved by the Institutional Review Board of Taipei Medical University. Informed written consent was provided by all donors, and all experiments were carried out following the relevant ethical guidelines and regulations.

pcMSCs were generated from the human placenta choriodecidual membrane using a serum‐free culture system as described previously (Su et al. 2017). In brief, the tissues of the choriodecidual membrane were digested using a combination of SMEM supplemented with 1 mg/mL DNase I, 0.5 mg/mL collagenase B, and 0.5 mg/mL protease. This digestion process was carried out overnight at 4°C, and the resulting mixture was filtered through a 100‐µm nylon membrane. After centrifugation, the cells were collected and resuspended in a culture medium consisting of MCDB201 supplemented with 1% penicillin/streptomycin, 10 ng/mL epidermal growth factor, and 1% insulin transferrin selenium. Subsequently, the cells were plated on culture dishes coated with human collagen type IV. Adherent cells were maintained in the culture medium, with regular changes every 3 or 4 days to remove nonadherent cells. The cultured cells that demonstrated characteristics of pcMSCs were further cultured in a serum‐free medium and exhibited a fibroblast‐like morphology after attachment. These cells showed positive expression for CD44, CD73, CD90, and CD29 and negative expression for CD45, CD34, CD14, and HLA‐DR (Lin et al. 2019).

2.2. Cell Culture

HepG2/C3A, A549, and C2C12 cells were maintained in a 37°C, 5% CO₂, water‐saturated atmosphere and grown in Dulbecco's modified Eagle medium (DMEM; Gibco BRL, Grand Island, NY, US) (Cat No. 31600034) supplemented with 1% (v/v) penicillin and streptomycin, 10% (v/v) FBS, 2 mM L‐glutamine, 0.1 mM nonessential amino acid, and sodium bicarbonate (3.7 g/L). For the cellular uptake experiments, the medium was changed to serum‐free DMEM before sEV treatment.

2.3. In Vitro APAP Treatment and pcMSC‐Based Therapy

HepG2/C3A cells were treated with APAP in serum‐free DMEM for 30 min. Following this incubation, cells were treated with either pcMSCs, pcMSC‐CM, unlabelled sEVs, or CAR‐sEVs for 24 h. Subsequent experiments and analyses were performed after the treatment period.

2.4. Animals

Male C57BL/6J mice, aged 8–12 weeks, were obtained from the National Laboratory Animal Center (NLAC) in Taipei, Taiwan. The mice were housed in a specific pathogen‐free (SPF) environment at the Laboratory Animal Center (LAC) of the National Taiwan University College of Medicine. The mice were given unrestricted access to food and water. The experimental protocol was approved by the Institutional Animal Care and Use Committee (IACUC) in Taiwan, ensuring compliance with ethical guidelines for animal research.

2.5. ALF Mouse Model

Male C57BL/6J mice were fasted for 12–16 h, after which they were intraperitoneally injected with APAP (12 mg/mL) dissolved in a 0.9% NaCl solution. To optimize the APAP dosage, the fasted mice were given doses of 0, 50, 100, 150, 200, 250, and 300 mg/kg body weight APAP. In the cell treatment group, a total of 1 × 10⁷ pcMSCs were suspended in 1 mL of PBS and transplanted through intraperitoneal injection 30 min after the administration of 250 mg/kg body weight APAP. In the mesenchymal stromal cell‐conditioned medium (pcMSC‐CM) treatment group, 1 mL of culture supernatant from the pcMSCs was collected, and a single intraperitoneal injection was performed 30 min after the administration of 250 mg/kg body weight APAP. For the sham and control groups, 1 mL of PBS was administered intraperitoneally 30 min after the administration of 0.9% NaCl solution and APAP, respectively. For the sEV and CAR‐sEV groups, 10⁹ or 10¹⁰ particles were administered 30 min after APAP injection. At 24 h following APAP administration, the mice were sacrificed, and serum was harvested for analysis.

2.6. Conditioned Medium Collection, sEV Isolation, and Characterization

Conditioned medium was collected from pcMSC serum‐free culture medium, and sEVs were obtained and purified via ultracentrifugation. In summary, the conditioned medium from pcMSCs incubated for 4 days was collected. The medium was then centrifuged at 300 × g for 10 min to eliminate dead cells. The resulting supernatant was subjected to sequential centrifugation at 2000 × g for 10 min to eliminate apoptotic bodies and 10,000 × g for 30 min to remove microvesicles. Afterward, the sEVs were purified through ultracentrifugation at a speed of 100,000 × g for 90 min. To minimize the presence of soluble factors, the sEVs were washed once with PBS. Following the removal of the supernatant, sEV pellets were resuspended in PBS. A BCA protein assay kit was used to measure the protein content of the sEVs. Nanoparticle tracking analysis (NTA, NanoSight NS300) and a Litesizer DLS 500 (Anton Paar, Austria) were used to determine the particle size, zeta potential (ZP), and particle concentration. Transmission electron microscopy (TEM) was used to observe and analyse the structure and shape of the sEVs. NanoFCM (NanoFCM, China), a dedicated flow cytometer built specifically for analysing nanosized particles, was utilized to identify sEV‐associated markers, including CD9, CD63, and CD81.

2.7. Super‐Resolution Microscopy

Super‐resolution images of sEVs were captured using a temperature‐controlled Nanoimager S microscope from ONI (Oxford Nanoimaging, Oxford, UK), equipped with a 100×, 1.4 NA oil immersion objective, an XYZ closed‐loop piezo stage, and a single‐molecule localization microscopy (SMLM) system featuring four lasers and two emission channels, split at 640 nm. sEVs from the N₃‐sEVs group were conjugated with DBCO‐Cy5, while those from the CAR‐sEVs group were labeled with DBCO‐488‐scFv via click chemistry. Sample preparation followed the manufacturer's protocol. Briefly, the surface of the EV Profiler 2 assay chip was coated with biotinylated capture molecules conjugated to polyclonal antibodies against CD9, CD63, and CD81. sEV concentrations of 10⁹–10¹¹ particles/mL were applied to the chip and immobilized. The surfaces were then stained with detection antibodies—mouse anti‐human CD9‐CF488, CD63‐CF568, and CD81‐CF647. Following staining, post‐fixation was performed, and the surface was incubated with a dSTORM imaging buffer. Images were acquired in dSTORM mode using total internal reflection fluorescence (TIRF) for sequential data capture. Single‐molecule data were processed and filtered using NimOS software (v1.18.3, ONI). Further analysis, including drift correction, was performed on ONI's Collaborative Discovery (CODI) platform (version 0.2.3).

2.8. ASGR1‐Specific scFv Preparation

The DNA fragment encoding the ASGR1‐specific scFv sequence, consisting of 261 amino acid residues (Trahtenherts and Benhar 2009) with a C‐terminal His‐tag, was chemically synthesized and inserted into the pET‐21a(+) vector (Novagen, Germany) for scFv production in E. coli. Additionally, using a quick‐change site‐directed mutagenesis kit (Stratagene, USA), the wild‐type scFv was modified with a Ser19 to Cys mutation (S19C) for conjugation purposes. Subsequently, the plasmid was transformed into the E. coli BL21(DE3) strain, and the cells were cultured in Power Broth medium (Molecular Dimensions, UK) supplemented with 0.1 mg/L ampicillin at 37°C until the OD600 reached 0.8. The cell culture was then cooled to 16°C, and protein expression was induced by adding 1 mM isopropyl β‐D‐1‐thiogalactopyranoside (IPTG). Following overnight incubation, the cells were harvested by centrifugation, washed, and suspended in a buffer containing 500 mM NaCl and 50 mM Tris·HCl at pH 8.0. The cells were disrupted using a cell disruptor (Constant Systems, UK), and the cell debris was removed by centrifugation. The supernatant was loaded onto a HisTrap Excel Ni‐NTA column (GE Healthcare, USA) and washed with buffer, and the proteins were stepwise eluted with 80, 250, and 500 mM imidazole. The 250 mM fraction contained the majority of the scFv protein and was concentrated to 15 mg/mL while simultaneously exchanging the buffer with 100 mM NaCl and 50 mM Tris·HCl at pH 8.0. The mutant scFv was expressed and purified using the same methods as those used for the wild‐type scFv.

2.9. scFv and DBCO‐PEG4‐Maleimide Conjugation by a Site‐Specific Cysteine‐Cyclooctyne Reaction

DBCO‐scFv was produced by incubating scFv with a 4‐fold molar excess of DBCO‐PEG4‐maleimide in 50 mM HEPES (200 mM NaCl, pH 7.0) at 30°C for 2 h. The DBCO‐scFv conjugates were then verified by reacting them with an equivalent number of moles of Calfluor 488 azide at 25°C for 1 h. The final product was analysed by SDS‐PAGE under nonreducing conditions.

2.10. Cell‐Surface Labelling of Azido Glycans with AZDye 488 DBCO on pcMSCs and pcMSC‐sEVs by Metabolic Glycoengineering and Click Chemistry

For the Ac4ManNAz concentration optimization experiment, pcMSCs were seeded in 96‐well plates at a density of 3.4 × 10³ cells per well. The cells were subsequently supplemented with different concentrations of Ac4ManNAz (0, 1, 5, 10, 20, 50, and 100 µM) and incubated for 4 days. After removing the supernatant, the cells were washed twice with PBS and treated with AZDye 488 DBCO (20 µM) for 1 h at 37°C. After five washes with PBS, a microplate reader (Synergy H1 Hybrid Multi‐Mode Reader, Biotek Inc., Winooski, VT, USA; model number: 3013208037–58) was used to measure the fluorescence intensity (λ_ex: 485 nm and λ_em: 528 nm). Similarly, the incubation time of Ac4ManNAz on pcMSCs was optimized by incubating the cells with 20 µM Ac4ManNAz for 1 to 6 days. AZDye 488 DBCO (20 µM) was then added for 1 h, and each sample was quantified using a microplate fluorescence reader (Biotek, USA). To evaluate the toxicity of Ac4ManNAz on pcMSCs, different concentrations of Ac4ManNAz (0, 1, 5, 10, 20, 50, and 100 µM) were added to the cells, followed by incubation for 4 days. The viability of cells in each well was measured using a CellTiter‐Glo luminescent cell viability assay kit (G7570, Promega, USA), and the luminescence was detected using a microplate reader (Biotek, USA). For the pcMSC‐sEVs labelling experiment, pcMSCs were incubated with 20 µM Ac4ManNAz for 4 days, and N₃‐sEVs were purified via ultracentrifugation as previously described. AZDye 488 DBCO (0.005, 0.05, 0.5, 5, 10, 20, 40, and 80 µM) was then added to N₃‐sEVs, followed by incubation at 25°C for 30 min. The final product was washed twice with PBS through ultracentrifugation, and the fluorescence intensity of the pcMSC‐sEVs‐AZDye 488 DBCO (λ_ex: 485 nm and λ_em: 528 nm) was analysed using a NanoAnalyzer (NanoFCM, China) and NF Professional 1.0 software. The experimental procedure was conducted by Reliance Biosciences, Taiwan.

2.11. Production of CAR‐sEVs

The production of CAR‐sEVs involved conjugating DBCO‐scFv with azido glycans on pcMSC‐sEVs. To this end, 10 µM DBCO‐scFv was prepared and incubated with N₃‐sEVs (4.5 × 10⁹ particles) at 25°C for 30 min. Following incubation, the CAR‐sEVs were washed with PBS using ultracentrifugation. The harvested CAR‐sEVs pellet was then subjected to further analysis. The procedure used for sEVs production was identical to that used for CAR‐sEVs production but with PBS substituted for scFv.

2.12. Flow Cytometry Analysis

For single‐EV analysis, sEVs were isolated from conditioned media (CM) of control and N₃‐modified pcMSC cultures via ultracentrifugation. The isolated N₃‐sEVs were incubated with 10 µM DBCO‐AF488‐scFv for 30 min at room temperature for the click chemistry reaction to conjugate the fluorescently labeled scFv onto the surface of sEVs. Subsequently, both sEVs and N₃‐sEVs were incubated with 5 µL APC‐conjugated CD9 antibody (312107, BioLegend) per 10⁶ cells in 100 µL staining buffer on ice for 1 h. sEVs were washed via ultracentrifugation to remove unbound reagents. Labeled sEVs were analysed using a CytoFlex V5‐B5‐R3 flow cytometer (Beckman Coulter). CD9‐positive sEVs were identified using the APC‐CD9 antibody (detected in the APC channel). DBCO‐AF488‐scFv labelling efficiency was assessed by measuring fluorescence at 488 nm/525 nm. The percentage of CD9+/AF488+ sEVs was calculated to quantify CAR‐sEVs. Control sEVs served as a baseline for comparison. For cell uptake analysis, HepG2/C3A cells were cultured and treated with either N₃‐sEVs or CAR‐sEVs. sEVs were added to the culture medium at specified time points (0.5, 1, 2, 4, and 6 h) to evaluate cellular uptake. After each incubation period, cells were fixed with 100% ethanol. Cellular uptake was quantified by measuring the mean fluorescence intensity (MFI) using a BD LSRFortessa flow cytometer. Data was analysed to compare the uptake efficiency of N₃‐sEVs and CAR‐sEVs in HepG2/C3A cells over the specified time course.

2.13. sEV miRNA Extraction, Library Preparation, and Sequencing

Total RNA was extracted using TRIzol LS reagent (10296010, Invitrogen, CA, USA) following the manufacturer's protocol. The RNA concentration was measured with a NanoDrop spectrophotometer (ND‐1000, NanoDrop Technology, USA) at 260 nm. To assess RNA quality, we used the LabChip RNA 6000 kit (Agilent Technologies, USA) and the Bioanalyzer 2100 (Agilent Technologies). Sample libraries were created using the QIAseq miRNA Library Kit (Qiagen, Germany), following the manufacturer's instructions, and sequenced on an Illumina platform (75‐cycle single‐end reads, 75SE). The miRNA data generated from this study are available in the Gene Expression Omnibus database under accession number GSE247568.

Sequencing data were processed using the Illumina BCL2FASTQ v2.20 software (Illumina, USA) for demultiplexing. Trimmomatic was used to filter out reads shorter than 18 nucleotides, ensuring high‐quality sequences (Bolger et al. 2014). The data were analysed using miRDeep2 (Friedländer et al. 2012), aligning the reads to the GRCh38 reference genome from the University of California, Santa Cruz (Langmead et al. 2009). To identify miRNAs, we considered only reads that aligned to the genome at up to five locations, as human miRNAs have limited genomic alignment sites (Rastegari et al. 2020). Normalized miRNA expression was calculated using reads per million mapped reads (RPM), by dividing the signal of each miRNA by the total number of mapped reads. miRNA target prediction was performed using the MirTarget V4 tool (Liu and Wang 2019) of miRDB version 6 (Chen and Wang 2020), with only functional human miRNAs from the FuncMir collection (http://www.mirdb.org/FuncMir.html) included. We excluded gene targets with fewer than 60 target prediction scores and miRNAs with over 2000 predicted targets. TargetScan (v8.0; targetscan.org (Agarwal et al. 2015; Mcgeary et al. 2019)) was utilized to identify miRNAs targeting specific genes, with a context++ score ≤ −0.1. For differential miRNA expression analysis, we used the read counts of mature miRNAs obtained from miRDeep2's Quantifier module, normalizing each miRNA's read count to RPM for sample comparison.

2.14. Immunocytochemical Staining and Confocal Imaging

HepG2/C3A, A549, and C2C12 cells were seeded on 12‐mm glass microscope coverslips in 24‐well plates at a concentration of 1.25 × 10⁵ cells per well. The cells were then fixed using a 4% PFA aqueous solution and blocked with 3% BSA in PBS. For the scFv internalization experiment, recombinant anti‐ASGR1 antibody (10 µg/mL) and scFv (10 µg/mL) were prepared using 1% BSA in PBS. In the scFv/galactose competition assay, scFv (10 µg/mL) was added after blocking with D‐(+)‐galactose (100 mg/mL) for 1 h. The secondary antibodies used for the control and scFv groups were Alexa Fluor 488‐conjugated anti‐His tag antibodies, which were diluted in blocking buffer at a ratio of 1:15000. For the recombinant anti‐ASGR1 antibody group, Alexa Fluor 488‐conjugated AffiniPure Donkey Anti‐Rabbit IgG diluted in blocking buffer at a ratio of 1:200 was used. The cells were imaged using a Zeiss LSM780 confocal microscope system, and the obtained images were analysed using ZEN5 microscopy software.

2.15. Membrane Labelling of HepG2/C3A Cells with PKH26 Dye

To label HepG2/C3A cells with PKH26 dye, 2 × 10⁷ cells were washed in a serum‐free medium to remove proteins and lipids. After centrifugation at 300 × g for 5 min, the supernatant was discarded, and the cells were suspended in Diluent C to create a 2× cell suspension. Subsequently, the cell suspension was mixed with 1 mL of 2× dye solution and incubated for 5 min. Staining was stopped by adding an equal volume (2 mL) of serum. Following centrifugation at 400 × g for 10 min at 20°C–25°C, the supernatant was removed, and the cells were resuspended in 10 mL of complete medium. The washing process was repeated twice using 10 mL of complete medium each time to eliminate unbound dye. Finally, after the last wash, the cell pellet was resuspended in 10 mL of complete medium to assess cell recovery, viability, and fluorescence intensity.

2.16. In Vitro Analysis of APAP‑Treated HepG2/C3A Cells Treated with pcMSC‐Based Therapies

HepG2/C3A cells were seeded in 96‐well plates at a density of 1.5 × 10⁴ cells/well. For the viability assay, the cells were treated with different concentrations of APAP (0, 2, 5, 10, 20, and 40 mM) and incubated for 24 h. After incubation, cell viability was assessed using a CellTiter‐Glo luminescent cell viability assay kit, and luminescence was measured using a microplate reader (BioTek). To evaluate the therapeutic effects of pcMSC‐CM and pcMSC‐sEVs, an equal volume of pcMSC‐CM and various doses of sEVs and CAR‐sEVs (10⁷, 10⁸, 10⁹ particles/well) were added to each well after 30 min of incubation with a 10 mM APAP. The viability of the treated cells was measured after 48 h of incubation with pcMSC‐CM or pcMSC‐sEVs.

2.17. Glutathione (GSH) Assay

GSH levels in C3A cells were measured using a Glutathione Assay Kit (Abcam, Cat. No. ab239727) according to the manufacturer's protocol. C3A cells were seeded in a 6‐well plate and treated with 10–40 mM APAP as indicated. After an initial 30‐min incubation with APAP, cells were further treated with either a pcMSC‐conditioned medium or a medium control. After 48 h, cells were harvested, and GSH was extracted by adding a 5% sulfosalicylic acid (SSA) solution to the cell pellets. The lysates were vortexed vigorously and kept on ice for 10 min, followed by centrifugation at 12,000 × g for 20 min at 4°C to remove cell debris. The supernatant was collected, and GSH levels were quantified by mixing the samples with assay buffer and detection reagent in a 96‐well plate. Absorbance was measured at 405 nm using a microplate reader. A standard curve was generated using known concentrations of GSH, which was used to calculate the GSH concentration in each sample. The remaining lysate underwent BCA protein quantification, allowing GSH levels to be normalized to protein content. Results were expressed as nmol of GSH per mg of protein. Each experiment was performed in triplicate, and data are presented as the mean ± SEM.

2.18. DCFDA Assay for ROS Measurement

Reactive oxygen species (ROS) levels in C3A cells were measured using a DCFDA/H₂DCFDA—Cellular ROS Assay Kit (Thermo Fisher Scientific, Cat. No. MP36103), following the manufacturer's protocol. C3A cells were seeded at a density of 1 × 10⁵ cells per well in a 12‐well plate and incubated overnight to allow cell attachment. Following incubation, cells were treated with 40 mM APAP for 30 min to induce oxidative stress, followed by treatment with PBS, sEVs, or CAR‐sEVs for 24 h.

After the treatment period, cells were harvested, washed, and resuspended in PBS. A final working concentration of 5 µM DCFDA was added to the cell suspension, and the cells were incubated with DCFDA for 30 min in the dark at 37°C. Following incubation, fluorescence signals were immediately measured using flow cytometry (BD LSRFortessa). ROS levels were quantified based on the fluorescence intensity of DCFDA, which was analysed in triplicate. Data are presented as the MFI ± SEM.

2.19. Histology and Immunohistochemical Staining

Slides were incubated in a heating chamber at 60°C for 30 min, then immersed in xylene and rehydrated through a graded alcohol series. The deparaffinized sections were subjected to antigen retrieval using 10 mM citrate buffer (pH 6.0) at 98°C for 20 min, followed by washing with TBS and permeabilization in TBST (0.2% Tween 20 in TBS). Next, the sections were blocked with 5% (v/v) normal goat serum (Vector Laboratories, S‐1000, USA) for 1 h. After blocking, the slides were incubated overnight with the following primary antibodies: anti‐ASGR1 (Invitrogen, PA5‐81318), anti‐CD9 (Abcam, ab2215), anti‐CD63 (Abcam, ab59479), anti‐cleaved caspase‐3 (Proteintech, 68773‐1‐Ig), anti‐PCNA (Chemicon, CBL407), and the TUNEL assay using the In situ Cell Death Detection Kit, POD (Roche, Cat. No. 11684817910, Switzerland). The following day, after washing the sections with TBST, they were incubated for 2 h at room temperature with secondary antibodies, including goat anti‐rabbit IgG conjugated to Alexa Fluor 594 and goat anti‐mouse IgG conjugated to Alexa Fluor 488, Alexa Fluor 594, or Alexa Fluor 633. Afterward, the sections were counterstained with DAPI (4’,6‐diamidino‐2‐phenylindole). Fluorescence signals were captured using a Stellaris 8 confocal microscope (Leica, Germany). The specific experimental conditions for each antibody are detailed in Supplementary Tables S2 and S3. In addition, haematoxylin and eosin (H&E) staining was performed to evaluate the histological structure of the tissues. Masson trichrome staining was carried out to assess collagen fibre deposition. The stained sections were then examined under a light microscope to analyse the distribution and morphology of fibrotic tissue.

2.20. Statistical Analysis

The mean values of all experimental data were analysed using GraphPad Prism 8.40 software. Comparisons between two groups were performed using a two‐tailed unpaired t‐test. One‐way ANOVA was conducted to compare multiple groups, and two‐way ANOVA was used to analyse the effects of multiple variables and their interaction. Data are presented as the mean ± standard error of the mean (SEM). Statistical significance was established at p values < 0.05. All graphs depict mean values accompanied by SEMs.

3. Results

3.1. Optimizing the Surface Labelling of Azido Glycans on pcMSCs and pcMSC‐sEVs Through Metabolic Glycoengineering and Click Chemistry

To label sEVs effectively, a strategy involving the introduction of azido moieties onto the sEV surface through metabolic glycan synthesis in living cells was used (Figure 1A). Based on the mechanisms of metabolic glycoengineering, the cells metabolize the azide‐containing monosaccharide Ac4ManNAz, which results in the presentation of azido groups on cell membrane glycans for glycoprotein modification (Li et al. 2024; Ying et al. 2022). The azido‐conjugated building blocks were introduced on the surface of pcMSCs. The presence of azido groups on the membrane glycans of pcMSCs was confirmed by performing SPAAC with AZDye 488 DBCO, which comprises a DBCO motif linked to the green fluorophore Alexa Fluor‐488. The DBCO‐containing fluorescent dye is utilized for labelling azido‐modified biomolecules in living cells. The number of azido groups on the surface of the pcMSCs increased in a dose‐dependent manner with increasing concentrations of Ac4ManNAz after 3 days of incubation (Figure 1B). We selected 20 µM Ac4ManNAz for further time‐dependent optimization. The highest level of azido group expression on the surface of the pcMSCs was observed after 4 days of incubation with 20 µM Ac4ManNAz (Figure 1C). We assessed the viability of pcMSCs treated with Ac4ManNAz and found that cell survival remained unaffected at concentrations less than 50 µM (Figure 1D). To determine whether azido‐containing monosaccharides are necessary for membrane labelling via click chemistry, we utilized AZDye 488 DBCO. The results showed that pcMSCs incubated with Ac4ManNAz exhibited a strong AZDye 488 DBCO fluorescent signal in contrast to that of the control group (Figure 1E,F). This indicates that the presence of azido groups is essential for membrane‐based SPAAC. Having confirmed the level of azido‐glycan expression on the surface of the pcMSCs, we focused on the azido‐glycans expressed on the pcMSC‐sEVs.

FIGURE 1.

Cell‐surface labelling of azido glycans on pcMSCs and pcMSC‐sEVs via metabolic glycoengineering and SPAAC. (A) Scheme for the CAR‐sEVs manufacturing strategy. Following the treatment of pcMSCs with Ac4ManNAz, N₃‐sEVs were subsequently purified from the conditioned medium. These N₃‐sEVs were then subjected to DBCO‐scFv conjugation via click chemistry, resulting in the formation of CAR‐sEVs. (B) Azido groups dose‐dependently expressed on the surface of pcMSCs via metabolic glycoengineering when the cells were treated with increasing concentrations of Ac4ManNAz. Three replicates per sample were performed, and the experiment was repeated three times. The data are presented as the mean ± SEM. (C) The highest expression level of azido groups on the pcMSC surface was observed after 4 days of incubation with Ac4ManNAz (20 µM). Three replicates per sample were performed, and the experiment was repeated three times. The data are presented as the mean ± SEM. (D) The viability of Ac4ManNAz‐treated pcMSCs was analysed by a luminescent cell viability assay. Three replicates per sample were performed, and the experiment was repeated three times. The data are presented as the mean ± SEM. (E) Compared with that of the group without Ac4ManNAz pretreatment, the labelling yield (%) of the group treated with AZDye 488 DBCO (green) to pcMSCs incubated with 20 µM Ac4ManNAz for 4 days significantly increased. Nuclei were stained with DAPI (blue). Scale bar = 50 µm. (F) The mean fluorescence intensity of azido groups on pcMSCs conjugated with AZDye 488 DBCO per field was calculated across a total of six fields. Significant differences (p < 0.001) in mean fluorescence intensity were observed between two groups of pcMSCs, one that was pretreated with Ac4ManNAz and the other that was not. The fluorescence intensities are presented as arbitrary units with SEMs. (G) Illustration of sEVs surface labelling with a fluorescent probe through click chemistry. The labelling strategy involved the reaction of AZDye 488 DBCO with N₃‐sEVs via SPAAC, resulting in the formation of AF488‐sEVs that exhibit a fluorescence signal on their surface. (H) (I) The number of azido groups on the surface of pcMSC‐sEVs gradually increased with increasing doses of AZDye 488 DBCO (0.005, 0.05, 0.5, 5, 10, 20, 40, and 80 µM), reaching a plateau at 10 µM. The labelling yield (%) of AZDye 488 DBCO‐conjugated sEVs was determined by NanoFCM. (J) (K) Compared with that of the group without Ac4ManNAz pretreatment, the labelling yield (%) of AZDye 488 DBCO to N₃‐sEVs purified from the conditioned medium of pcMSCs after 4 days of 20 µM Ac4ManNAz pretreatment significantly increased (p < 0.001). The labelling yield (%) of AZDye 488 DBCO‐conjugated sEVs was determined by NanoFCM. All data are presented as the mean ± SEM. ***p < 0.001, ****p < 0.0001.

Recent studies have shown that metabolic glycoengineering allows for the modification of cell‐derived EVs with chemical tags (Song et al. 2020; Lee et al. 2018). EVs secreted by azido‐labelled cells similarly display azido groups on their surfaces as a result of the exosome biogenesis pathway (Bhatta et al. 2023). In our studies, we quantified the distribution of azido groups on the surface of pcMSC‐sEVs by performing click reactions with increasing concentrations of AZDye 488 DBCO (Figure 1G). Our findings indicated the successful labelling of azido group‐presenting membranes with pcMSC‐sEVs and AZDye 488 DBCO, and the labelling efficiency increased in a dose‐dependent manner with increasing AZDye 488 DBCO concentration (Figure 1H,I). Furthermore, compared with those from pcMSCs without Ac4ManNAz, pcMSC‐sEVs purified from the conditioned medium of pcMSCs treated with Ac4ManNAz exhibited a significant increase in the intensity of the AZDye 488 DBCO fluorescent signal (Figure 1J,K).

These findings validate the efficacy of our membrane labelling strategy for modifying sEVs. We preliminarily succeeded in introducing azido moieties onto the surface of pcMSC‐sEVs via metabolic glycoengineering. These modified sEVs were used for subsequent conjugation with targeting materials to synthesize CAR‐sEVs.

3.2. DBCO‐scFv Synthesis and ASGR1 Targeting Validation

To achieve targeted hepatic delivery, a scFv was specifically developed to selectively bind to ASGR1 (Figure S1); the sequence used was described by Trahtenherts (2009) (Trahtenherts and Benhar 2009). The recombinant scFv consisted of the variable regions of both the heavy and the light chains of an antibody, which were conjoined by a flexible (SGGG)₄ peptide linker. To assess the targeting efficacy of the scFv, a sandwich ELISA was performed, which confirmed that the scFv specifically targeted the ASGR1 antigen in a dose‐dependent manner (Figure 2A). In addition, the synthesized scFv was bound to the human hepatoma cell line HepG2/C3A within 30 min of incubation and was gradually internalized into the cytoplasm with prolonged incubation time (Figure 2B,C). Meanwhile, it showed no targeting specificity toward non‐hepatic cells, including A549 and C2C12 (Figure S2). The results confirmed that the synthesized scFv can specifically bind to the ASGR1 antigen and can be internalized by HepG2/C3A cells.

FIGURE 2.

Validation of the production and targeting of DBCO‐scFv in the HepG2/C3A cell model. (A) Sandwich ELISAs were performed to detect serial dilutions of scFv that bound to ASGR1. scFv was then detected using an HRP‐conjugated anti‐6X His tag antibody, and the absorbance at 450 nm was quantified using an absorbance microplate reader. Three replicates per sample were performed, and the experiment was repeated three times. The data are presented as the mean ± SEM. The figure to the right presents a schematic representation of the sandwich ELISAs for assessing the binding affinity of scFv to ASGR1. (B) Confocal microscopy images (630×) showed that after 30 min, scFv, which was stained with an Alexa Fluor 488‐conjugated anti‐His tag antibody, specifically targeted ASGR1 expressed in the membrane of HepG2/C3A cells. ASGR1 was gradually internalized by the cells as the incubation time increased (0.5, 1, 3, and 6 h). Images of the nuclei (DAPI, blue), cell membrane (PKH26, red), and scFv (green) were merged. Scale bar = 50 µm. (C) Quantification of the mean fluorescence intensity of the scFv targeting ASGR1 in HepG2/C3A cells incubated for different durations (0.5, 1, 3, and 6 h). The mean fluorescence intensity per field was calculated across a total of six fields. The fluorescence intensities are presented as arbitrary units with SEMs. (D) Different molar ratios of galactose to scFv were mixed and added to ELISA wells coated with the ASGR1 antigen. The scFv signal was gradually inhibited concomitant with increasing proportions of galactose. Three replicates per sample were performed, and the experiment was repeated three times. The data are presented as the mean ± SEM. (E) Confocal microscopy images (×400) showing that scFv (1 µM) stained with an Alexa Fluor 488‐conjugated anti‐His tag antibody specifically targeted ASGR1 expressed on the membrane of HepG2/C3A cells and that the targeting was inhibited by preincubation with galactose (500 mM) for 1 h. Images of the nuclei (DAPI, blue), cell membrane (PKH26, red), and scFv (green) were merged. Scale bar = 50 µm. (F) The quantification results revealed that the mean fluorescence intensity of the scFv (1 µM) targeting ASGR1 in HepG2/C3A cells was inhibited after preincubation with galactose (500 mM) for 1 h. The mean fluorescence intensity per field was calculated across a total of six fields. The fluorescence intensities are presented as arbitrary units with SEMs. (G) Schematic representation of DBCO‐scFv production through the conjugation of scFv (S19C) and DBCO‐PEG4‐maleimide by a site‐specific cysteine‐cyclooctyne reaction. (H) Sandwich ELISAs were performed to detect serial dilutions of DBCO‐scFv bound to ASGR1. DBCO‐scFv was detected with biotin‐PEG3‐azide (10 µM) via a click reaction and subsequently interacted with streptavidin HRP and TMB. The absorbance was measured at 450 nm by an absorbance microplate reader. Three replicates per sample were performed. The data are presented as the mean ± SEM. The figure to the right presents a schematic representation of the sandwich ELISAs for assessing the binding affinity of DBCO‐scFv for ASGR1. (I) SDS‐PAGE of DBCO‒scFv conjugates was performed under nonreducing conditions. Coomassie blue staining (left) and fluorescence image (right) of the SDS‐PAGE gel of the DBCO‒scFv conjugates. M: marker ladder. Lane 1: scFv protein control group. Lane 2: Calfluor 488 azide control group. Lane 3: scFv + Calfluor 488 azide control group. Lane 4: scFv:DBCO‐PEG4‐maleimide:Calfluor 488 azide = 1:1:1. Lane 5: scFv:DBCO‐PEG4‐maleimide:Calfluor 488 azide = 1:2:2. Lane 6: scFv:DBCO‐PEG4‐maleimide:Calfluor 488 azide = 1:4:4. Lane 7: scFv:DBCO‐PEG4‐maleimide:Calfluor 488 azide = 2:4:4. Lane 8: scFv:DBCO‐PEG4‐maleimide:Calfluor 488 azide = 4:4:4. (J) Confocal microscopy images (×400) showing that after 30 min, DBCO‐scFv, detected by click reaction with Calfluor 647 Azide (10 µM), specifically targeted ASGR1 expressed on the membrane of HepG2/C3A cells. DBCO‐scFv was gradually internalized by the cells after a prolonged incubation time (6 h). Images of nuclei (DAPI, blue), cell membranes (FM1‐43, green), and DBCO‐scFv (red) were merged. Scale bar = 50 µm. (K) The mean fluorescence intensity was significantly higher (p < 0.01) in the DBCO‐scFv‐treated HepG2/C3A cells than in the control cells without scFv treatment. The mean fluorescence intensity per field was calculated across a total of six fields. The fluorescence intensities are presented as arbitrary units with SEMs. (L) Confocal microscopy images (×630) showing that after 3 h of incubation, DBCO‐scFv, detected by click reaction with Calfluor 647 Azide (10 µM), specifically targeted ASGR1 and was internalized into the cytoplasm by HepG2/C3A cells. Images of nuclei (DAPI, blue), cell membranes (FM1‐43, green), and DBCO‐scFv (red) were merged. Scale bar = 50 µm. All data are presented as the mean ± SEM. *p < 0.05, **p < 0.01.

To confirm that the scFv specifically binds to ASGR1 in cells, a competition assay was conducted. Galactose, a ligand that is specifically recognized by ASGR1, was utilized as a competitor to scFv (Mamidyala et al. 2012; Hoober 2020). Mixtures containing various molar ratios of galactose to scFv were added to ELISA wells precoated with the ASGR1 antigen. The results revealed that an increase in the galactose concentration corresponded to a decrease in the scFv signal (Figure 2D). Moreover, scFv displayed specificity for ASGR1 in the membrane of HepG2/C3A cells, with this specific binding being inhibited when the cells were preincubated with galactose for one hour (Figure 2E,F). Furthermore, we conjugated the scFv to DBCO via a site‐specific cysteine‐cyclooctyne reaction, in which the engineered cysteine residue (S19C) in the scFv reacted with the maleimide group of DBCO‐PEG4‐maleimide, forming a thioether bond (Figure 2G). This reaction ultimately yielded the DBCO‐scFv conjugate. Subsequent experimental results confirmed that the binding affinity of the scFv was retained after DBCO modification, with DBCO‐scFv maintaining dose‐dependent binding efficiency to ASGR1 (Figures 2H and S3). To confirm the successful synthesis of the DBCO‐scFv conjugate, SDS‐PAGE analysis was performed, and the results demonstrated that scFv (25 kDa) was successfully conjugated to DBCO‐PEG4‐maleimide in a dose‐dependent manner (Lanes 6, 7, and 8), as detected by the click reaction with CalFluor 488 azide (Figure 2I). Finally, we confirmed that the synthesized DBCO‐scFv exhibited a binding pattern to HepG2/C3A cells similar to that of scFv, with significant targeting within 30 min of incubation and progressive internalization into the cytoplasm with extended incubation (Figure 2J,L).

Collectively, these results suggest that the synthesized DBCO‐scFv conjugate exhibited enhanced targeting of ASGR1 on human hepatic cell lines. The conjugate was subsequently incorporated into N₃‐sEVs for CAR‐sEVs construction.

3.3. Characterization of sEVs and CAR‐sEVs

To generate CAR‐sEVs, DBCO‐scFv was incubated with N₃‐sEVs, and subsequent characterizations of both CAR‐sEVs and unmodified pcMSCs‐derived sEVs were performed (Figure 1A). TEM analysis revealed that both the unmodified pcMSCs‐derived sEVs and CAR‐sEVs exhibited round or cup‐shaped morphologies, with diameters ranging from 30 to 150 nm (Figure 3A). Flow cytometry analysis was conducted to examine the membrane CD markers of the unmodified pcMSC‐sEVs. The findings indicated that pcMSC‐sEVs exhibited significant expression levels of the tetraspanins CD9, CD63, and CD81, with low expression of calnexin, a negative marker of sEVs (Figure 3B and S4). Nanoparticle tracking analysis (NTA) was conducted to determine the size and concentration of both types of sEVs (Veerman et al. 2021; Bachurski et al. 2019). The results indicated that the average size of sEVs in both groups ranged from 80 to 120 nm (Figure 3C,D). To further elucidate the properties of the sEVs, a dynamic light scattering (DLS) analyser was used to measure the ZP of each sEV group. ZP is a quantitative measurement of the electrical charge and stability of particles suspended in a fluid, offering insights into particle‐particle interactions and the propensity for particle aggregation in the surrounding medium (Rasmussen et al. 2020). The analysis revealed that the surfaces of the sEVs, N₃‐sEVs, and CAR‐sEVs displayed a similar negative charge, enhancing the stability of the EVs by preventing their aggregation and clumping through electrostatic repulsion (Figure 3E,F). This confirmed that the azido moiety and scFv modifications did not affect the surface charge of the pcMSC‐sEVs. As demonstrated, the sEVs and CAR‐sEVs were comprehensively characterized using various techniques, and the results suggested that the CAR‐sEVs exhibited characteristics similar to those of the sEVs. Moreover, the expression patterns of the tetraspanins CD9, CD63, and CD81 on a single sEV were observed on pcMSC‐sEVs using ONI Nanoimager (Figure 3G). Super‐resolution microscopy images of native sEVs and modified N₃‐sEVs show the expression of highly concentrated tetraspanins CD9, CD63, and CD81 on the sEVs surface. We also used DBCO‐Cy5 to validate the presence of azido groups on the surface of N₃‐sEVs, confirming that DBCO successfully conjugates with the N₃‐groups on the sEVs surface (Figure 3G.i). The quantitative analysis further demonstrates the expression percentage of tetraspanins and azido groups on the N₃‐sEV surface, confirming the high percentage presence of azido groups (Figure 3G.ii). Clustering analysis and spatial distribution patterns highlight differences in marker organization, with CD63/CD81 clusters in native sEVs and DBCO‐Cy5/CD63 clusters in modified N₃‐sEVs, supporting the successful incorporation of azido groups onto N₃‐sEVs (Figure 3G.iii).

FIGURE 3.

Characterization of sEVs and CAR‐sEVs. (A) Transmission electron microscopy (TEM): The morphology and structure of both the sEVs and CAR‐sEVs were characterized as either spherical or cup‐shaped, with diameters ranging from 30 to 150 nm. Scale bar = 200 nm (left). The scale bar for enlarged images is 50 nm (right). (B) Flow cytometry was used to analyse EV surface markers, including CD9, CD63, and CD81. The gating strategy for each marker is shown, with the respective fluorescence intensities on the y‐axis and side scatter on the x‐axis. (C) (D) Nanoparticle tracking analysis (NTA): Quantification and size of sEVs and CAR‐sEVs isolated from conditioned medium of pcMSCs ranging from 80 to 120 nm. The data are presented as the mean ± SEM; n = 6 per group. p value was measured by an unpaired t‐test; ns, non‐significant. (E) (F) The zeta potential (mV) profile and quantification of the sEVs, N₃‐sEVs, and CAR‐sEVs isolated from the conditioned media of the pcMSCs were measured with a Litesizer DLS 500. The surface potential of the vesicles was negatively charged in both groups. The data are presented as the mean ± SEM; n = 6 per group. ns, non‐significant. (G)(i) Representative super‐resolution microscopy images of single pcMSC‐derived native sEVs and modified N₃‐sEVs labelled with surface markers CD9 (green), CD63 (blue), or CD81 (red). The modified N₃‐sEVs were conjugated with DBCO‐Cy5 (red) via click chemistry. Merged images demonstrate the colocalization of these markers on both native and modified sEVs. Scale bars represent 50 nm for individual marker images and 100 nm for merged images. (ii) Quantitative analysis of the percentage of single‐, double‐, and triple‐positive sEVs expressing CD9, CD63, CD81, or DBCO‐Cy5 was conducted using CODI software. The bar charts represent the distribution of sEVs populations in both the native sEV and N₃‐sEV groups. (iii) Clustering analysis strategy for native sEVs and N₃‐sEVs, showing a large field of view with multiple clusters (left panel), a zoomed‐in view of a selected EV cluster (middle panel), and a corresponding graph (right panel) depicting the cluster distribution of CD81/CD63 for native sEVs or DBCO‐Cy5/CD63 for modified N₃‐sEVs. (H) The diagram represents the conjugation of N₃‐sEVs with DBCO‐AF488‐scFv (green) to produce CAR‐sEVs through click chemistry. (I) Representative super‐resolution microscopy image of a single CAR‐sEV particle, labelled with EV surface markers CD9 (red), CD63 (blue), and DBCO‐AF488‐scFv (green). Merged images demonstrate the colocalization of these markers on CAR‐sEVs. (J) Flow cytometry histogram showing the fluorescence intensity of CD9‐positive N₃‐sEVs labelled with DBCO‐AF488‐scFv. The percentage of CAR‐sEVs (CAR+) is calculated to be 45.3% under a 10 µM DBCO‐AF488‐scFv, compared to the control sEVs.

To validate the labelling efficiency of CAR‐sEVs, we conjugated N₃‐sEVs with AF488‐labeled DBCO‐scFv via click reaction (Figure 3H). CytoFlex analysis showed that the labelling efficiency of CAR‐sEVs increases with the dose of DBCO‐AF488‐scFv, reaching 92% when treated with 200 µM DBCO‐AF488‐scFv (Figure S5). This confirms that the click reaction efficiently labels N₃‐sEVs with DBCO‐scFv to generate CAR‐sEVs. The presence of DBCO‐scFv on the surface of a single CAR‐sEV particle was observed (Figure 3I). For further therapeutic assessment of CAR‐sEVs, we selected a dose of 10 µM DBCO‐AF488‐scFv to achieve approximately 45% labelling of DBCO‐scFv‐conjugated CAR‐sEVs, ensuring high specificity while preserving the intrinsic biofunction of CAR‐sEVs (Figure 3J). Based on these results, we confirmed that CAR‐sEVs can be efficiently constructed through the conjugation of DBCO‐scFv with N₃‐sEVs via the SPAAC click reaction. Further validation was conducted to assess the targeting efficiency of CAR‐sEVs.

3.4. CAR‐sEVs Markedly Improved Targeting Efficacy in the HepG2/C3A Cell Model

The ability of CAR‐sEVs to target human hepatic cells in vitro was investigated using a HepG2/C3A cell model that expresses ASGR1. To validate the targeting ability of CAR‐sEVs, we utilized a dual tagging approach, simultaneously tracking N₃‐sEVs with DBCO‐Cy5 and labelling the targeting materials on the surface of N₃‐sEVs with DBCO‐AF488‐scFv (Figure 4A). The uptake of CAR‐sEVs by HepG2/C3A cells exhibited a significant dose‐responsive interaction compared to sEVs, as evidenced by the increased fluorescence signals in the cells using flow cytometry analysis (Figure 4B,C). This cellular uptake was inhibited when ASGR1 was blocked through galactose pretreatment (Figure S6). Additionally, CAR‐sEVs significantly targeted HepG2/C3A cells as early as 30 min and continued to accumulate within the cells with prolonged incubation, while the N₃‐sEVs‐treated group exhibited minimal uptake even at 6 h of incubation (Figure 4D). In the flow cytometry analysis, HepG2/C3A cells treated with CAR‐sEVs showed a significant increase in fluorescence signal over incubation times ranging from 30 min to 6 h, compared to the N₃‐sEVs‐treated group (Figure 4E). This thorough analysis validated the superior targeting and uptake of CAR‐sEVs by HepG2/C3A cells, underscoring the potential of CAR‐sEVs as an efficient delivery system for human hepatic cell applications. These findings collectively highlight the promise of CAR‐sEVs as a precise delivery strategy for medical treatments targeting the liver.

FIGURE 4.

CAR‐sEVs significantly enhanced targeting efficacy in the HepG2/C3A cell model. (A) Schematic representation of the conjugation of N₃‐sEVs with DBCO‐Cy5 (red) and DBCO‐AF488‐scFv (green) using click chemistry to generate CAR‐sEVs. (B) Flow cytometry histograms showing the fluorescence intensity of HepG2/C3A cells treated with increasing doses (10⁷, 10⁸, or 10⁹ particles) of CAR‐sEVs or N₃‐sEVs for 18 h. The fluorescence was detected in the Alexa Fluor 488 and PE‐Cy5 channels. Sham treatment (no sEVs) was included as a control. (C) Quantification of targeting efficiency (%) of N₃‐sEVs and CAR‐sEVs at different doses (10⁷, 10⁸, or 10⁹ particles). (D) Confocal microscopy images (630×) showing the cellular uptake of N₃‐sEVs and CAR‐sEVs in HepG2/C3A cells at various time points (0.5, 1, 2, 4, and 6 h). The images illustrate the nuclei (DAPI, blue), sEV particles with azido groups (DBCO‐Cy5, red), and DBCO‐scFv (labelled with Alexa Fluor‐488, green). The merged DIC (differential interference contrast) images to fluorescent images. Scale bar = 10 µm. (E) Flow cytometry was used to measure the mean fluorescence intensity in HepG2/C3A cells treated with N₃‐sEVs (grey) or CAR‐sEVs (red) over 6 h. All data are presented as the mean ± SEM. ***p < 0.001.

3.5. CAR‐sEVs Alleviated APAP‐Induced Injury in HepG2/C3A Cells

To assess the protective effects of pcMSC‐based therapies on APAP‐induced injury in HepG2/C3A cells, in vitro studies were conducted. Exposure to APAP for 24 h dose‐dependently reduced the viability of HepG2/C3A cells (Figure 5A.i). Interestingly, incubation with pcMSC‐CM or pcMSC‐sEVs significantly increased cell viability in the presence of 10 mM APAP (Figure 5A.ii and iii). Additionally, incubation with APAP led to a dose‐dependent decrease in the expression of GSH (Figure S7A), a critical antioxidant that plays a key role in protecting against oxidative stress and actively participates in the detoxification of xenobiotics (Li et al. 2023; Lu 2020). Interestingly, GSH levels notably increased after 48 h of incubation with pcMSC‐CM in the presence of 10 mM APAP (Figure S7B). Moreover, both pcMSC‐CM (Figure 5A.iv) and pcMSC‐sEVs (Figure 5A.v) enhanced the viability of HepG2/C3A cells exposed to various APAP concentrations. Given the observed therapeutic potential of pcMSC‐based therapies, we proceeded to investigate the therapeutic effects of CAR‐sEVs. Surprisingly, CAR‐sEVs significantly enhanced the survival of HepG2/C3A cells in a dose‐dependent manner when the cells were incubated with 10 mM APAP (Figure 5A.vi). From the DCFDA analysis, CAR‐sEV treatment reduced ROS levels in APAP‐injured HepG2/C3A cells compared to the sEV treatment group (Figure S8).

FIGURE 5.

Compared to unmodified sEVs, CAR‐sEVs enhanced the therapeutic effects against the APAP challenge in C3A cells. (A) (i) APAP dose‐dependently reduced the viability of HepG2/C3A cells, with an IC50 of 10 mM APAP. (ii) The viability of HepG2/C3A cells significantly increased after 48 h of incubation with pcMSC‐CM in the presence of APAP (10 mM), as determined by a luminescent cell viability assay. (iii) The viability of HepG2/C3A cells significantly increased after 48 h of incubation with sEVs from pcMSCs in the presence of APAP (10 mM), as determined by a luminescent cell viability assay. (iv) The viability of HepG2/C3A cells significantly increased after 48 h of incubation with pcMSC‐CM in the presence of increasing doses of APAP (0, 10, 20, 40 mM). Three replicates per sample were performed, and the experiment was repeated three times. (v) The viability of HepG2/C3A cells significantly increased after 48 h of incubation with sEVs from pcMSCs in the presence of increasing doses of APAP (0, 10, 20, 40 mM). Three replicates per sample were performed, and the experiment was repeated three times. (vi) Compared with corresponding doses of sEVs, CAR‐sEVs significantly increased the viability of HepG2/C3A cells in a dose‐dependent manner after 48 h of incubation in the presence of APAP (10 mM). Three replicates per sample were performed, and the experiment was repeated three times. The data are presented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (B) The heatmap displays the predicted interactions between pcMSC‐sEVs miRNAs and genes of interest categorized into four major functional groups: anti‐inflammation (TGF‐β1, TGF‐β2, TGF‐β3, IL‐6, IL‐1β, TNF‐α, and NKB1), liver regeneration promotion (PTEN, MKK4, MFAP4), anti‐apoptosis (caspase‐3, caspase‐7, caspase‐8, caspase‐9, Bax and Bim), and anti‐fibrosis (COL1A1 and COL3A1). TargetScanHuman 8.0 was used to calculate the targetability score, with only the top miRNAs having a TargetScore context++ score of ≤ −0.1 selected. The miRNAs were selected using TargetScanHuman 8.0, with a TargetScore context++ score ≤ −0.1. The miRNA expression is shown as log2 RPM (reads per million), with higher values indicating greater miRNA expression. (C) (i) The line chart showing the highly expressed miRNAs significantly involved in specific biological processes of negative regulation of hepatocyte proliferation (GO:2000346) and positive regulation of the apoptotic process (GO:0043065). (ii) The Venn diagram illustrates the overlap of miRNAs involved in two biological processes. The overlapping area represents 15 miRNAs predicted to target genes involved in both processes. (iii) The top miRNAs from this overlap are listed in the table below.

Considering the significant therapeutic effect of pcMSC‐sEVs on ALF, we conducted a comparative analysis of pcMSC‐sEV miRNA profiles. Our analysis revealed that specific miRNAs within the pcMSC‐sEVs target several key regulatory genes, notably TGFB1, IL‐6, and TNFA (involved in inflammation), PTEN and MKK4 (implicated in liver regeneration), caspase‐3, caspase‐7, caspase‐8, and caspase‐9, along with Bax and Bim (all related to apoptosis), as well as COL1A1 and COL3A1 (associated with fibrosis) (Figure 5B and Table S1). These results suggest that the miRNA cargo of pcMSC‐sEVs has significant potential to influence key biological processes, particularly those involved in liver regeneration, apoptosis, and fibrosis. Furthermore, the enrichment analysis of pcMSC‐sEV miRNAs indicated a significant association with two Gene Ontology (GO) biological processes: positive regulation of the apoptotic process (GO:0043065) and negative regulation of hepatocyte proliferation (GO:2000346) (Figure 5C.i and .ii). Notably, several highly expressed miRNAs, including miR‐125b‐5p, miR‐199a‐3p, and miR‐16‐5p, were identified as overlapping between the two GO pathways (Figure 5C.iii). Their ability to influence multiple genes involved in these processes highlights the versatility and therapeutic potential of sEV‐based treatments for liver injury. These findings underscore the potential of targeted EV‐based therapies to protect HepG2/C3A cells from APAP‐induced injury in an in vitro setting. The results suggested that CAR‐sEVs may exert beneficial effects by specifically and effectively delivering therapeutic sEVs to injured cells, with potential clinical applications in the context of APAP‐induced liver injury.

3.6. CAR‐sEVs Demonstrated Superior Therapeutic Effects to Those of sEVs in APAP‐Induced ALF Mice

The efficacy of CAR‐sEVs in alleviating APAP‐induced ALF was evaluated using a mouse model (Figure S9A). Male C57BL/6 mice were fasted for 12–16 h before receiving an intraperitoneal injection of 250 mg/kg body weight APAP. Thirty minutes post‐APAP injection, the mice were divided into various groups and subjected to different therapies, including intraperitoneal injections of pcMSCs (1×10⁷ cells), pcMSC‐CM (1 mL), sEVs (10⁹ particles), and CAR‐sEVs (10⁹ particles). After 24 h, the mice were euthanized, and blood and liver samples were collected for analysis (Figures S9B and S9C). Before administering pcMSC derivatives, we tested various doses of pcMSCs via IP injection in ALF mice, including 10⁵, 10⁶ and 10⁷ cells, to identify the optimal dose. Significant recovery was observed in the group treated with 10⁷ pcMSCs following a high dose of 250 mg/kg APAP (Figure S9D). In pcMSC‐based therapy, the liver morphology of mice in the sham group was smooth, dense, and brown, whereas the livers of mice in the pathological control group (APAP/PBS) exhibited severe haemorrhaging. Compared to pcMSC transplantation, pcMSC‐CM, and sEV treatments, CAR‐sEV treatment resulted in the least haemorrhaging and a liver morphology closest to that of the sham group (Figure 6A). Serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were measured to assess liver function. Compared to the pathological control (APAP/PBS), the transplantation of pcMSCs and administration of pcMSC‐CM, sEVs, or CAR‐sEVs significantly reduced ALT and AST levels, with CAR‐sEV treatments showing the lowest ALT and AST levels in the serum (Figure 6B,C), indicating the alleviation of liver damage. Notably, the serum levels of the proinflammatory cytokines interleukin‐1 beta (IL‐1β), interleukin‐6 (IL‐6), tumour necrosis factor‐alpha (TNF‐α), and monocyte chemoattractant protein‐1 (MCP‐1) decreased following pcMSC‐based treatments (Figure 6D.i‐iv), with the lowest concentrations observed in the CAR‐sEV treatment group, indicating its strong anti‐inflammatory effects. Additionally, histological analysis using haematoxylin and eosin (H&E) staining revealed that, compared to the pathological control group (APAP/PBS), treatments with pcMSCs transplantation, pcMSC‐CM, sEVs, and CAR‐sEVs effectively reduced the necrotic area of liver tissue, with CAR‐sEV treatment exhibiting the lowest percentage of necrosis (Figure 6E). Masson's trichrome staining (MT) revealed a reduction in fibrosis following pcMSC‐based therapy (Figure S10). Additionally, pcMSC‐based treatment increased the expression of proliferating cell nuclear antigen (PCNA) (Figure 6F), with the CAR‐sEV treatment group showing the highest level of hepatocyte proliferation. Tissue sections stained for apoptosis markers using TUNEL (green) and cleaved caspase‐3 (red) revealed that CAR‐sEV treatment achieved the most substantial reduction in apoptotic cells (Figure 6G). These findings demonstrate that pcMSC‐based therapies, including pcMSC transplantation and the administration of pcMSC‐CM, sEVs, or CAR‐sEVs, effectively improved liver morphology, reduced necrosis, decreased inflammation, and promoted liver cell proliferation and anti‐apoptotic activity in the APAP‐induced ALF model. Notably, CAR‐sEVs showed the highest efficacy among all groups, suggesting that our targeted therapy approach has the potential to greatly improve current stem cell treatments and overcome the limitations of traditional stem cell therapy. To verify the homing effects of CAR‐sEVs, we stained human CD9 and CD63 EV protein markers in liver tissue sections from APAP‐injured ALF mice treated with pcMSC‐based therapy. The results demonstrated that CAR‐sEVs specifically target and accumulate significantly in hepatocytes compared to the sEV‐treated group (Figure S11). Moreover, we evaluated ASGR1 expression in APAP‐injured ALF mouse liver tissue following pcMSC‐based treatment. The results demonstrated that pcMSC‐based therapy significantly restores ASGR1 expression compared to the pathological control group (APAP/PBS) (Figure S12). The notable improvement observed in the APAP‐induced ALF mouse model suggested that CAR‐sEVs effectively delivered therapeutic sEVs to the injured liver, indicating promise for clinical application for the treatment of APAP‐induced ALF.

FIGURE 6.

Compared with unmodified sEVs, CAR‐sEVs exhibited superior therapeutic efficacy in attenuating APAP‐induced liver injury in an ALF mouse model. (A) Representative liver morphology from different groups at 24 h post‐APAP injection: sham group (NaCl 0.9%), pathological control group (APAP/PBS), and pcMSC, pcMSC‐CM, sEVs (10⁹ particles), and CAR‐sEVs (10⁹ particles) treatment groups (n = 3 per group). The enlarged images are presented at 5× magnification with a scale bar of 5 mm. (B) (C) Serum levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) in the sham group (n = 9), pathological control group (APAP/PBS) (n = 8), and pcMSCs (n = 5), pcMSC‐CM (n = 8), sEVs (n = 5–8), and CAR‐sEVs (n = 5–8) treatment groups were measured 24 h after saline or APAP injection. (D) (i–iv) Serum levels of inflammatory cytokines (IL‐1β, IL‐6, TNF‐α, and MCP‐1) measured after the administration of pcMSCs, pcMSC‐CM, sEVs (10⁹ particles), or CAR‐sEVs (10⁹ particles) (n = 5–6 per group). (E–G) Representative images of liver sections stained with haematoxylin and eosin (H&E) (E), PCNA staining (F), and TUNEL assay with cleaved caspase‐3 staining (G) from the sham group, pathological control group (APAP/PBS), and groups treated with pcMSCs, pcMSC‐CM, sEVs (10⁹ particles), and CAR‐sEVs (10⁹ particles) 24 h after saline or APAP injection. Images are presented with a scale bar of 50 µm. Quantification was performed by analysing four to five fields per group, including necrotic area (%), PCNA+ cells (%), TUNEL+ cells (%), and cleaved caspase‐3+ cells (%). All the data are presented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns indicates non‐significance.

4. Discussion

In this study, we developed a method for delivering EVs directly to specific body sites, enhancing targeting effects and maximizing therapeutic benefits. Through this targeted approach, we enhanced the overall therapeutic efficiency and improved the availability of EVs for clinical use. Using advanced technologies, we engineered sEVs with customized targeting materials to precisely guide them to a desired location within the body, that is, the liver. We anticipate that this strategy will advance the clinical application of EVs and lead to more effective treatment outcomes.

APAP toxicity now ranks as the second leading cause of liver transplants worldwide, and patients usually experience severe ALF before a fatal event occurs (Caparrotta et al. 2018). Despite the efficacy of NAC administration, the gold standard treatment to mitigate APAP overdose, fatalities from APAP toxicity still occur in the United States and Europe due to severe APAP ingestion and delayed treatment (Lee 2017). While studies have shown that NAC effectively restores liver function soon after APAP intake, it fails to protect against APAP‐induced injury during later stages (Hu et al. 2020). We adopted stem cell therapy as a potential solution to overcome the clinical challenges associated with treating severe ALF. An ALF mouse model was established using a high dose of APAP. Subsequently, various stem cell therapies, including pcMSCs, pcMSC‐CM, and pcMSC‐sEVs, were administered to ALF model mice. Based on liver function tests (AST/ALT levels), we selected 10⁷ pcMSCs for IP injection in ALF mice, as this dose showed significant recovery after 250 mg/kg APAP (Figure S9D), establishing it as the baseline for pcMSC‐based therapy (Figure 6B,C). For pcMSC‐CM, we administered 1 mL derived from 10⁷ pcMSCs, containing approximately 10⁹ sEVs (NTA assay). In sEV treatments, we used 10⁹–10¹⁰ particles per animal to evaluate dose‐dependent therapeutic effects. The outcomes revealed the significant alleviation of pathological conditions and the restoration of liver function in C57BL/6 mice treated with pcMSC‐based therapies. The liver morphology normalized, and the serum levels of ALT, AST, IL‐1β, IL‐6, TNF‐α, and MCP‐1 decreased. Additionally, H&E staining revealed apparent liver tissue repair. These findings confirmed the significant enhancement of liver function and reduction in inflammation in APAP‐induced ALF mice treated with pcMSC‐based therapies. In our studies, we applied the IP injection route rather than the IV injection because the IV injection of EVs often leads to higher liver accumulation. However, we focus on clarifying CAR's targeting ability. The biodistribution of EVs via IV injection makes it difficult to distinguish CAR‐specific targeting, so we used IP injection to better assess targeting effects in vivo. In addition, IP injection is commonly used for abdominal drug delivery and has shown promising results in targeting organs like the liver, diaphragm, and spleen (Chaudhary et al. 2010; Chiang et al. 2022; Chang et al. 2021). Based on these findings, we chose IP injection for our CAR‐sEVs study to enhance biodistribution accuracy.

Furthermore, we showed that pcMSC‐based therapies can protect liver cell lines from APAP‐induced injury in vitro, as indicated by increased viability following treatment with pcMSC‐CM and pcMSC‐sEVs. Combining MSCs with NAC has proven more effective than either therapy alone, showing a synergistic therapeutic effect on other inflammatory diseases (Shin et al. 2019). Recent studies underscore the importance of MSC‐derived EVs in mediating these beneficial effects. Clinical studies and ongoing trials have evaluated MSC‐derived EVs as therapeutic agents for various diseases, including ARDS, kidney diseases, graft‐versus‐host disease, osteoarthritis, stroke, Alzheimer's disease, and type 1 diabetes (Lotfy et al. 2023). However, there are currently only limited trials underway for liver diseases (Lotfy et al. 2023). Therefore, the development of MSC‐derived EV treatments for ALF patients is urgently needed. Building on this evidence, we focused on pcMSC‐sEVs to expand their therapeutic scope. Through precise characterization, we identified pcMSC‐sEVs as well‐distributed, stable, nanoscale membrane‐bound vesicles that met the MISEV2023 guidelines (Théry et al. 2018). We sought to ensure that the isolated pcMSC‐sEVs complied with the current validation standards for therapeutic agents. Based on these results, we aimed to enhance the efficacy of these agents through targeted cell‐based therapy. We employed bioorthogonal click chemistry, successfully labeled the surface of pcMSC‐sEVs with azido groups, and facilitated subsequent conjugation with DBCO‐containing chemicals via click reaction (Kolb et al. 2001).

To further increase the efficiency of our targeting system, we synthesized scFv and validated its strong affinity for the ASGR1 antigen. The chosen target, ASGR1, is a promising option for selective drug delivery due to its high expression in hepatocytes and transformed liver cells (Witzigmann et al. 2016). Using sandwich ELISAs and galactose competition assays, we demonstrated the high specificity of scFv for ASGR1; this was evidenced by its binding and blockade in the presence of a 10⁷‐fold higher concentration of galactose. Notably, our internalization experiments showed that scFv was not only bound to ASGR1 on HepG2/C3A cells but was also internalized into the cytoplasm, confirming the recognition and binding specificity of scFv. Upon ligand binding, ASGR1 undergoes conformational changes, initiating receptor‐mediated endocytosis. This internalization process occurs via clathrin‐coated pits, leading to receptor sorting and processing within the cell. After internalization, receptors move to endosomes for ligand‐receptor separation. Ligands are degraded, while receptors are recycled to the cell surface for future ligand binding and uptake. This recycling is crucial for subsequent ligand binding and uptake (D'Souza and Devarajan 2015). Understanding the endocytosis mechanism of ASGR1 has significant implications for drug development, particularly for targeted therapies. The affinity of ASGR1 for carbohydrates can aid in designing agents that are selectively internalized by ASGR1‐expressing hepatocytes, offering promise for treating liver diseases such as hepatocellular carcinoma, viral hepatitis, and liver fibrosis. This approach has the potential to advance precision medicine and improve patient outcomes (Ahn et al. 2021). Compared with galactose, a ligand typically used for liver targeting (D'Souza and Devarajan 2015), the designed scFv demonstrated superior binding affinity. Various strategies, such as regioselective isomers and synthetic modifications, enhance the targeting efficiency of hepatocytes, indicating strong potential for liver‐specific delivery. N‐acetylgalactosamine (GalNAc), which has a greater affinity than galactose, is a key ligand for ASGR1‐mediated endocytosis, but challenges arise with its delivery via liposomes, especially those with sizes exceeding 70 nm. However, GalNAc targeting is currently being extensively researched, including in clinical trials (Böttger et al. 2020). Lactoferrin, another potential ligand, shows promise for hepatocyte‐targeted delivery; however, more in vivo research is needed to fully understand its efficacy (Wei et al. 2015). Overall, these findings underscore the ongoing efforts to optimize ligand‐mediated delivery systems for the effective treatment of liver diseases.

To verify the functionality of our system, the ASGR1‐specific scFv was linked with DBCO‐PEG4‐maleimide to generate DBCO‐scFv via a site‐specific cysteine‐cyclooctyne reaction. The scFv sequence was specifically designed with the S19C mutation, enabling the thiol (‐SH) group to react with maleimide. The specifically introduced cysteine residue ensured that maleimide was conjugated to a specific site, thereby not impeding the binding affinity of the targeting construct. Subsequently, DBCO‐scFv was incorporated into our system to achieve precise targeting and enhance therapeutic efficacy. We meticulously engineered DBCO‐scFv to ensure its stability and functionality, confirming its ability to selectively bind azido‐containing substances via click reaction.

In this study, a crucial step was the surface modification of N₃‐sEVs with DBCO‐scFv. We validated that DBCO‐scFv modification of N₃‐sEVs did not alter the ZP, as azides are neutral and the anti‐ASGR1 scFv has a pI of 7.0, ensuring the physical properties of sEVs were preserved, preventing aggregation and maintaining bioavailability and targeting efficacy. By attaching scFv to pcMSC‐sEVs surface, we aimed to confer targeting capabilities, enabling specific interactions with desired targets. Indeed, CAR‐sEVs showed therapeutic effects in the APAP‐induced ALF model, normalizing liver morphology, reducing necrosis, and decreasing injury markers and inflammatory cytokine levels. The significant enhancement in the ALF model suggested that CAR‐sEVs can effectively deliver therapeutic agents, such as miRNAs, to the injured liver, indicating their promise for clinical use in APAP‐induced liver injury. The therapeutic potential of pcMSCs and their derivatives, pcMSC‐CM, was initially tested on in vitro ALF cell models, showing superior effectiveness over sEVs alone (Figure 5.A.ii–v). However, in vivo studies revealed limited efficacy of pcMSC‐CM in severe liver necrosis, likely due to systemic biodistribution and a lack of targeted specificity. To address this, we created CAR‐sEVs, which demonstrated significantly enhanced therapeutic efficacy compared to both unmodified sEVs and conventional pcMSC‐CM (Figure 6). These findings indicate that CAR‐sEVs have immense potential for various clinical applications, offering enhanced treatment efficacy while reducing potential side effects.

In summary, our research aimed to advance the development of DBCO‐scFv as a targeting agent, facilitating the generation of CAR‐sEVs. We sought to establish a robust and adaptable therapeutic platform with precise targeting capabilities for future clinical use. Our findings demonstrate that CAR‐sEVs display promising therapeutic effects and exceptional specificity for hepatocytes. This innovative strategy holds great potential for mitigating liver damage in APAP‐induced ALF, providing a targeted delivery system to increase the local concentration of therapeutic agents.

In this study, we discovered that CAR‐sEVs could substantially improve existing stem cell therapies and overcome their limitations, including low retention rates and a lack of targeting specificity. Furthermore, we envision that this targeted EVs delivery system has the potential to treat diverse diseases. We propose that by identifying specific markers for organs or tissues, site‐specific scFvs can be customized using the bioorthogonal‐based click reaction method with cysteine introduction. This approach can be applied not only to therapeutic EVs but also to whole cell‐based therapeutic strategies. We envision extending this concept to immune‐based systems, such as simplifying the production of CAR‐T cells or CAR‐NK cells and enhancing the effectiveness of targeted therapy. Nonviral endogenous nanoparticles offer a safe and effective method that permits repeated administration. However, the feasibility of applying this concept to other cell‐based targeted therapies still needs to be validated. For cells intolerant to plasmid transfection, this method offers a convenient and efficient alternative that does not disrupt cellular homeostasis. We anticipate that this targeted delivery system will have broad application prospects across various therapeutic fields, providing precise and personalized medicine for individuals. Regarding the indication for CAR‐sEVs, we anticipate that intrinsic drug‐induced acute liver injury caused by lipophilic drugs could be treated with CAR‐sEVs because intrinsic drug‐induced acute liver injury is typically dose‐related, with onset occurring within a short period (hours to days) and may lead to lethal conditions if not treated promptly (European Association for the Study of the Liver, Clinical Practice Guideline Panel: Chair, Panel members, and EASL Governing Board representative 2019). Future applications could involve combining this EV delivery system with gene therapy for rare genetic diseases (Cheng et al. 2020), a targeted delivery system for anticancer drugs (Xu et al. 2021), or a brain‐targeted drug delivery system to bypass the blood‐brain barrier for the treatment of neurological diseases and stroke with the aim of brain tissue repair (Naqvi et al. 2020). We will continue to explore the potential applications of CAR‐sEVs to maximize their benefits. Future studies will expand on various aspects of our findings.

Despite the promising results of our study, several critical questions remain unanswered and require further investigation. First, we aimed to elucidate the mechanism of action responsible for the therapeutic effects of CAR‐sEVs. To gain mechanistic insights, we conducted a comparative analysis of pcMSC‐sEVs miRNA profiles to investigate the potential mechanism of action (MOA). EVs facilitate intercellular communication by transporting miRNAs, which are sorted in a regulated manner. Once delivered, these miRNAs influence key cellular processes like apoptosis, migration, and proliferation by modulating gene expression that affects recipient cells. In this study, we analysed the miRNAs present in the sEVs of pcMSCs. Using the RPM method for normalizing miRNA sequencing results, we achieved a more accurate comparison of miRNA expression levels, allowing us to assess the relative abundance of specific miRNAs in pcMSC‐sEVs (Figure 5B,C). These miRNAs were categorized into groups associated with anti‐apoptosis, anti‐inflammation, anti‐fibrosis, and liver regeneration. Among the top eight highly expressed miRNAs (log2 (RPM) >15), the anti‐apoptosis group was predominant. Notably, miRNAs linked to Caspase‐9 included miRNA‐125b‐5p, miRNA‐199a‐3p, and miRNA‐199b‐3p. For Caspase‐3, the associated miRNAs were let‐7a‐5p, miRNA‐29a‐3p, let‐7b‐5p, and let‐7f‐5p. Additionally, miRNA‐29a‐3p was linked to both Caspase‐7 and Caspase‐8. These seven sEV miRNAs regulate both endogenous and exogenous apoptotic pathways. We identified anti‐inflammatory miRNAs targeting IL‐6, including let‐7a‐5p, let‐7b‐5p, and let‐7f‐5p. MiRNAs associated with TGF‐β family members were also identified, highlighting their roles in fibrosis regulation. Second, a comprehensive understanding of the pharmacokinetics (PK) and pharmacodynamics (PD) of CAR‐sEVs is still lacking. Clarifying the PK and PD profiles is crucial for advancing the formulation of therapeutic sEVs for clinical application. Finally, challenges persist regarding the expansion and stability of CAR‐sEVs for clinical use. These challenges include the need to scale up production while maintaining reaction efficiency, as well as addressing the inherent instability and heterogeneity of sEVs. Strategies aimed at enhancing the stability of isolated sEVs include optimizing isolation methods, establishing standardized storage conditions, and exploring techniques to bolster EV stability during therapeutic applications. These efforts are vital for successfully translating targeted EV therapy into clinical practice. In conclusion, our exploration of stem cell therapy, coupled with the advancement of targeted EV therapy using pcMSCs, represents a significant advance. The anticipated therapeutic outcomes and the potential of CAR‐sEVs hold immense promise for the future treatment of specific diseases. By surmounting clinical hurdles, we aim to redefine stem cell therapy, enhancing its efficacy through targeted delivery via a streamlined approach. A major obstacle in treating complex diseases is the inefficiency of drug action at specific sites. Addressing this challenge is crucial to developing a more effective and promising treatment approach for such conditions. We envision targeted EV therapy as a groundbreaking solution within the realm of cell therapy, charting a path toward a brighter future for human health and well‐being.

Author Contributions

Yen‐Ting Lu: conceptualization (equal), data curation (equal), formal analysis (equal), investigation (equal), methodology (equal), validation (equal), visualization (equal), writing–original draft (equal). Tzu‐Yu Chen: data curation (equal), formal analysis (equal), visualization (equal). Hsin‐Hung Lin: formal analysis (equal), methodology (equal), project administration (equal), resources (equal), software (equal), writing–original draft (equal). Ya‐Wen Chen: formal analysis (equal), validation (equal). Yu‐Xiu Lin: data curation (equal), formal analysis (equal), validation (equal), writing–review and editing (equal). Duy‑Cuong Le: data curation (equal), formal analysis (equal), validation (equal). Yen‐Hua Huang: data curation (supporting), formal analysis (equal), funding acquisition (equal), supervision (equal), validation (equal), writing–review and editing (equal). Andrew H.‐J. Wang: resources (equal), supervision (equal), writing–review and editing (equal). Cheng‐Chung Lee: conceptualization (equal), data curation (equal), funding acquisition (equal), investigation (equal), methodology (equal), project administration (lead), resources (equal), supervision (equal), validation (equal), writing–original draft (equal), writing–review and editing (equal). Thai‐Yen Ling: conceptualization (equal), data curation (equal), funding acquisition (equal), investigation (equal), methodology (equal), project administration (lead), resources (equal), supervision (equal), validation (equal), writing–review and editing (equal).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting information

Lu, Y.‐T. , Chen, T.‐Y. , Lin, H.‐H. , Chen, Y.‐W. , Lin, Y.‐X. , Le, D.‐C. , Huang, Y.‐H. , Wang, A. H.‐J. , Lee, C.‐C. , & Ling, T.‐Y. (2025). Small Extracellular Vesicles Engineered Using Click Chemistry to Express Chimeric Antigen Receptors Show Enhanced Efficacy in Acute Liver Failure. Journal of Extracellular Vesicles, 14, e70044. 10.1002/jev2.70044

Data Availability Statement

Data openly available in a public repository that issues datasets with DOIs