Lethal pulmonary thromboembolism in mice induced by intravenous human umbilical cord mesenchymal stem cell-derived large extracellular vesicles in a dose- and tissue factor-dependent manner

Abstract

Mesenchymal stem cell-derived extracellular vesicles (MSC-EVs) have obvious advantages over MSC therapy. But the strong procoagulant properties of MSC-EVs pose a potential risk of thromboembolism, an issue that remains insufficiently explored. In this study, we systematically investigated the procoagulant activity of large EVs derived from human umbilical cord MSCs (UC-EVs) both in vitro and in vivo. UC-EVs were isolated from cell culture supernatants. Mice were injected with UC-EVs (0.125, 0.25, 0.5, 1, 2, 4 μg/g body weight) in 100 μL PBS via the tail vein. Behavior and mortality were monitored for 30 min after injection. We showed that these UC-EVs activated coagulation in a dose- and tissue factor-dependent manner. UC-EVs-induced coagulation in vitro could be inhibited by addition of tissue factor pathway inhibitor. Notably, intravenous administration of high doses of the UC-EVs (1 μg/g body weight or higher) led to rapid mortality due to multiple thrombus formations in lung tissue, platelets, and fibrinogen depletion, and prolonged prothrombin and activated partial thromboplastin times. Importantly, we demonstrated that pulmonary thromboembolism induced by the UC-EVs could be prevented by either reducing the infusion rate or by pre-injection of heparin, a known anticoagulant. In conclusion, this study elucidates the procoagulant characteristics and mechanisms of large UC-EVs, details the associated coagulation risk during intravenous delivery, sets a safe upper limit for intravenous dose, and offers effective strategies to prevent such mortal risks when high doses of large UC-EVs are needed for optimal therapeutic effects, with implications for the development and application of large UC-EV-based as well as other MSC-EV-based therapies.

Introduction

Extracellular vesicles (EVs) are nano-sized, membrane-bound particles that carry various bioactive molecules to facilitate intercellular communication and regulate biological processes [1]. EVs are commonly enriched through differential centrifugation, which utilizes sedimentation coefficients to pellet EVs of different sizes. According to the latest Minimal Information for Studies of Extracellular Vesicles (MISEV) guidelines [2, 3], large EVs are obtained at intermediate-speed centrifugation forces (10,000–20,000 × g), while small EVs require higher speeds (100,000–200,000 × g). Notably, there is a size overlap between large and small EVs [2, 4–7].

Mesenchymal stem cell-derived EVs (MSC-EVs) have received much attention because of their similar biological functions to MSCs [8–11], low immunogenicity, and ease of preparation, storage, and administration [12, 13]. These characteristics make MSC-EVs a promising and safer alternative to MSC therapy, including for conditions such as graft-versus-host disease, acute respiratory distress syndrome, and osteoarthritis [14–16]. While small EVs have been the most studied EV subtype for their critical roles in tissue regeneration, the diagnostic or therapeutic potential of large EVs is also an active area of research. The larger size of these EVs allows for increased disease antigen presentation due to their larger surface area [17–19], and they also transport a greater amount of biological cargoes [20]. Studies have revealed the therapeutic promise of large MSC-EVs in conditions such as systemic sclerosis [21] and tissue/organ damage [22–26]. Our group has been focusing on large EVs [27–29] and demonstrated that large EVs derived from human umbilical cord MSCs (UC-EVs) serve as novel senescence-associated secretory phenotype components of their parental cells [30], and these large UC-EVs (hereinafter, unless otherwise specified, “UC-EVs” refer to the large UC-EVs) exhibit a remarkable ability to rejuvenate aging bone marrow MSCs (BM-MSCs) and delay the degenerative functional decline in multiple tissues of aged mice, holding great clinical value [31].

Despite these encouraging results, recent studies have revealed that both small and large MSC-EVs have adverse effects on blood coagulation [32–36]. The procoagulant activity of MSC-EVs is believed to be associated with specific proteins related to coagulation, such as tissue factor (TF/CD142) and phosphatidylserine, on their surface [33, 37, 38]. TF/CD142 is a transmembrane glycoprotein that initiates the extrinsic pathway of coagulation by binding to factor VII/VIIa and activating factor X [39]. Circulating TF/CD142 positive EVs have been associated with an increased risk of coagulopathy and thrombus formation in patients with diseases such as cancer, sepsis, and various viral infections [40–43]. Notably, Skovronova et al. [4] reported that large MSC-EVs express more TF/CD142 than small MSC-EVs, suggesting potential increased risks for clinical applications. Although there is a growing body of preclinical and clinical studies supporting the safety of small MSC-EVs, clinical experience with large MSC-EVs is limited. Therefore, it is crucial to determine whether and how intravenous administration of large MSC-EVs poses the risk of therapy-induced thrombosis for clinical translational research. To our knowledge, no systematic data have yet been reported for the large MSC-EVs intravenous infusion induced thromboembolism and prevention strategies.

In this study, we aim to confirm the procoagulant activity and mechanism of UC-EVs. Moreover, we investigated whether and how intravenous infusion of UC-EVs activates the coagulation system in vivo, potentially leading to thromboembolism or death in animals, and how to prevent it.

Materials and methods

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

UC-MSCs were isolated from the umbilical cords of full-term neonates delivered by cesarean section, as previously described [30, 31]. Briefly, umbilical cords were collected with the informed consent from the neonates’ parents and approval from the Ethics Committee of the Union Hospital affiliated with Tongji Medical College of Huazhong University of Science and Technology. After the arteries and veins were removed, the umbilical cords were cut into 1- to 2-mm pieces and laid flat in a culture dish. These pieces were then cultured in tissue culture dishes containing 10% fetal bovine serum (Gibco, CA, USA) and 1% penicillin/streptomycin (Solarbio, Beijing, China) until 80% to 90% confluence. In this study, all cultured UC-MSCs were routinely characterized following the MSC identification criteria established by the International Society for Cellular Therapy [44]. The isolated UC-MSCs adhered to culture surfaces and expressed MSC markers CD44, CD73, CD90, and CD105, while not expressing hematopoietic markers CD34 and CD45 (all from BD Biosciences, CA, USA), and demonstrated in vitro differentiation into osteoblasts and adipocytes. Passages 3 to 10 were used for experiments.

Extraction and characterization of UC-EVs

UC-EVs were isolated from cell culture supernatants through differential centrifugation, as previously described [30, 31]. The supernatants were collected from EVs-depleted complete culture medium (centrifuged at 20,000 × g for 1 h) and subjected to sequential centrifugation steps: 750 × g for 15 min to remove cells, 2000 × g for 20 min to remove apoptotic bodies and cell fragments, and 16,000 × g for 1 h to pellet UC-EVs. The UC-EVs were washed with phosphate buffered saline (PBS; Basalmedia, China), resuspended, and then stored at −80 °C until further analysis. The protein concentration of UC-EVs was measured by a BCA protein assay (Beyotime, China). The UC-EVs samples were characterized by transmission electron microscopy (TEM; HT7700, HITACHI, Japan), nanoparticle tracking analysis (NTA) on the Zetaview particle tracking analyzer (Particle Metrix, Germany), and Western blot [45].

Flow cytometry analysis of TF/CD142

The APC-labeled anti-TF/CD142 (BioLegend, CA, USA) was centrifuged at 20,000 × g for 1 h before staining to remove fluorescent particles, as described [46]. The UC-EVs were then stained with 5 μL of antibody for 30 min at 4 °C in the dark, according to the manufacturer’s instructions. An APC-labeled isotype antibody was used as a control. The samples were washed twice with PBS and analyzed by flow cytometry (ID7000, Sony, Japan). Fluorescent submicron beads of 0.2 μm, 0.5 μm, and 0.76 μm (Bangs Laboratories, IN, USA) were used to define an EV gate. TF/CD142 positive EV events were those with a diameter less than 0.76 μm and TF/CD142 positivity.

Western blot analysis of TF/CD142

Western blot was performed as previously reported [47]. The primary antibodies used in this study were rabbit anti-TSG101 (1:1000; Abcam, MA, USA), rabbit anti-CD81 (1:1000, Abcam, MA, USA), rabbit anti-CD9 (1:1000, Abcam, MA, USA), rabbit anti-TF/CD142 (1:1000, Proteintech, China), and rabbit anti-β-actin (1:1000, Abcam, MA, USA).

Collection of blood and preparation of EV-free plasma

Human blood from healthy volunteers (4 males and 3 females) was drawn from the cubital vein and collected into 3.8% sodium citrate to evaluate the procoagulant properties of UC-EVs. The blood was centrifuged at 3000 × g for 15 min to obtain platelet-poor plasma that was then centrifuged at 20,000 × g for 1 h to deplete EVs and the top two-thirds was collected. The EV-free plasma was aliquoted and stored at −80 °C and thawed in a water bath at 37 °C before measurements.

UC-EV-induced plasma clotting

UC-EVs at different protein concentrations were mixed with EV-free plasma and the coagulation was initiated by adding 100 μL of CaCl₂ (0.02 M, Solarbio, China). The plasma clotting time was immediately recorded on a coagulation analyzer using a magnetic bead mechanical method. To demonstrate the specificity of TF/CD142, different concentrations of the tissue factor pathway inhibitor (TFPI, 1, 2, 4, 8, and 16 μg/mL; MedChemExpress, China) were mixed with EV-free plasma before testing.

UC-EV-induced Xa generation

Factor Xase buffer (TBS, 5% BSA) and stop buffer (TBS, 5% BSA, 16 mM ethylene diamine tetraacetic acid) were prepared as described [48]. In brief, UC-EVs were added to the reaction system containing factor VIIa (4 nM; Haematologic Technologies, VT, USA) and factor X (130 nM; Haematologic Technologies, VT, USA). The mixture was incubated at 37 °C for 20 min, and the reaction was then stopped with stop buffer. The luminescent substrate S-2765 (Chromogenix, Italy) was added, and the absorbance at 405 nm (OD value) was immediately measured via an enzyme-linked immunosorbent assay. The anti-Xa generation assay was performed following the same steps as before with the addition of TFPI (16 μg/mL).

UC-EV-induced thrombin generation

Thrombin generation was evaluated by adding 10 μL of UC-EVs to EV-free plasma. Thrombin generation was initiated via the addition of 50 μL of CaCl₂ (0.02 M) and monitored by the fluorogenic substrate Z-Gly-Gly-Arg-AMC acetate (MedChemExpress, China). Kinetic measurements were taken at 37 °C for 1 h with 3-min intervals using the SYNERGY H1 microplate reader (BioTek, WVPD, USA) at 380/440 nm. Each group had three replicate wells and the endogenous thrombin potential (ETP) was calculated according to Hemker [49]. The anti-thrombin generation assay was performed following the same steps as before with the addition of TFPI (16 μg/mL).

UC-EV-induced whole blood clotting

The effect of UC-EVs on whole blood clot formation was investigated using the Thrombelastograph Hemostasis Analyzer (Haemoscope, MA, USA). Thrombelastography (TEG) tests were performed within 2 h after blood collection at room temperature. Upon the addition of UC-EVs to 340 μL of whole blood, clotting was initiated by the addition of 20 μL of CaCl₂ (0.2 M). Parameters assessed were the clotting time (CT), clot formation time (CFT), maximum clot firmness (MCF), and α-angle. The anticoagulant assay was performed following the same steps as before with the addition of TFPI (20 μg/mL).

Animal experiments

All animal experiments and procedures were approved by the Institutional Animal Care and Use Committee of Tongji Medical College, Huazhong University of Science and Technology, and were conducted in strict accordance with the relevant animal welfare and protection standards. Male KM mice (6 to 8 weeks) were purchased from the Beijing Vital River Company and housed at the Pesticide Toxicology Research Center of Tongji Medical College. The mice were maintained under a 12-h light/dark cycle, a temperature of 24 °C, and 50%–55% relative humidity.

For intravenous bolus injection of UC-EVs, mice (n = 70) were randomly divided into seven groups (n = 10 each group) and injected with 100 μL PBS or UC-EVs (0.125, 0.25, 0.5, 1, 2, 4 μg/g body weight) in 100 μL PBS via the tail vein. Behavior and mortality were monitored for 30 min after injection and the mortality rate was calculated. The median lethal dose (LD₅₀) was analyzed by Probit analysis using SPSS 23.0 software. Following this, blood samples and organs (heart, liver, spleen, lung and kidney) were collected from the mice for further analysis.

For intravenous pumping of UC-EVs, mice were randomly divided into two groups and infused with 100 μL of UC-EVs (4 μg/g body weight) via the tail vein for either 0 and 1 h, respectively (n = 5). During the administration of UC-EVs via a micro-infusion pump, mice were anesthetized with 3.4% isoflurane and positioned on a tail vein injection instrument.

Routine blood tests and coagulation analysis

Whole blood was analyzed by an automated hematology analyzer (BC-2800vet, Mindray, China) to measure red blood cells (RBCs), white blood cells (WBCs), platelets, and hemoglobin. For coagulation analysis, plasma was separated from whole blood with 3.8% sodium citrate by centrifugation at 3500 × rpm for 20 min. An automated blood coagulation analyzer (RAC-1830, Rayto, China) was used to determine prothrombin time (PT), activated partial thromboplastin time (APTT), and fibrinogen (Fg). PT and APTT exceeding the detection range were recorded as “120 s” and fibrinogen below the detection range was recorded as “NC” indicating no clotting.

Histological analysis

The hearts, livers, spleens, lungs, and kidneys of mice were harvested and fixed in 4% paraformaldehyde for 24 h at room temperature. The organs were paraffin-embedded and sectioned at 5-μm thickness. The sections were dewaxed and stained with H&E. Thrombosis formation within the vasculature was observed by digital scanners (3DHISTECH, Hungary). The pulmonary embolism (PE) score for each mouse was calculated as the average of 4 to 5 lung lobes, following the scheme by Xu et al. [50]. A score of 0 indicates that no PE was detected microscopically. Thrombus formation observed in fewer than 2 sites, 3 sites, and at least 4 sites were assigned scores of 1, 2, and 3, respectively, to quantify the extent of embolism.

Fluorescence microscopy and immunofluorescence

UC-EVs were stained using a PKH26 Red Fluorescent Cell Linker Kit for General Cell Membrane Labeling (Sigma-Aldrich, MO, USA) following the manufacturer’s instructions. The PKH26-labeled UC-EVs were administered to mice via the tail vein. Lung and liver tissues were harvested, snap-frozen, and then sectioned at 5 μm. The vascular endothelium and the cell nuclei were stained with anti-CD31 and DAPI, respectively. Fluorescence microscopy (Olympus, Japan) was used to observe UC-EVs retained in the lung and liver tissues.

In vivo imaging

100 μg EVs were stained with DiR fluorescent dye (Rengen Biosciences, China) for imaging following the manufacturer’s instruction, and then injected into mice via the tail vein. The hearts, livers, spleens, lungs, and kidneys were harvested and imaged by the Image System Lago/Lago X imaging system (Spectral Instruments Imaging, AZ, USA) to assess the UC-EVs biodistribution.

Statistical analysis

Continuous data (mean ± SD) were analyzed by Student’s t-test for two groups and parametric or non-parametric ANOVA for more than two groups. The Kruskal-Wallis test was used for multiple comparisons of non-normal or heteroscedastic data. Data analyses were performed by SPSS 23.0 or GraphPad Prism8.0. P < 0.05 were considered statistically significant.

Results

Validation and characterization of UC-EVs

The UC-MSCs obtained were fibroblast-like adherent cells with the ability to differentiate into osteogenic and adipogenic lineages (Fig. 1a). They expressed CD44, CD73, CD90, and CD105 but not the hematopoietic cell markers CD34 and CD45 (Fig. 1b). TEM revealed that UC-EVs were spherical in shape with a diameter of less than 1.0 μm and possessed characteristic intact membrane structures (Fig. 1c). NTA showed a multi-peaked and highly heterogeneous particle size distribution, ranging from 0 to 1000 nm. The size range was consistent with the results of TEM, with the majority of particles were concentrated between 0 and 300 nm (Fig. 1d). Western blot showed that UC-EVs expressed the representative EV markers TSG101, CD81, and CD9 (Fig. 1e). Flow cytometry detected TF/CD142-positive events in UC-EVs extracted from UC-MSCs culture medium. As shown in Fig. 1f, approximately 4.5% of all particles <0.76 μm was TF/CD142-positive. Furthermore, Western blot also confirmed the presence of the TF/CD142 protein in UC-EVs (Fig. 1g).

Fig. 1. Validation of UC-EVs and detection of their surface TF/CD142 antigens.

a The isolated UC-MSCs were fibroblast-like adherent cells capable of differentiating into osteogenic and adipogenic lineages, respectively. Scale bar: 50 μm. b Flow cytometry analysis of MSC surface markers. c TEM images of UC-EVs. Scale bar: 200 nm. d The size distribution of UC-EVs measured by NTA. e EV marker proteins (TSG101, CD81, CD9, and β-actin) in both UC-MSCs and UC-EVs were analyzed by Western blotting. f The expression of TF/CD142 in UC-EVs was assessed by flow cytometry. Standard microspheres of 0.2 μm, 0.5 μm, and 0.76 μm were used to define events with a diameter less than 0.76 μm as EVs, with APC-labeled IgG antibody as a control. g TF/CD142 was identified on Western blots of UC-EVs samples.

Procoagulant properties of UC-EVs

To evaluate the procoagulant properties of UC-EVs, we performed a series of assays, including the plasma clotting assay, Xa generation assay, thrombin generation assay, and TEG assay. For the plasma clotting assay, EV-free plasma did not clot when recalcified for over 120 s, whereas UC-EVs shortened plasma clotting time, which decreased with increasing doses of UC-EVs (Fig. 2a). In the Xa generation assay, a positive correlation between Xa production and UC-EV levels was observed, as indicated by the increased optical density OD value of Xa products with higher UC-EV concentrations (Fig. 2b). For the thrombin generation assay, thrombin generation increased in a time-dependent manner in the representative thrombin generation curves (Fig. 2c). Only a small amount of thrombin was detected in the control group. In contrast, the addition of UC-EVs at different levels accelerated the thrombin generation process and increased ETP (Fig. 2d).

Fig. 2. Procoagulant properties of UC-EVs.

a The effect of different protein levels in UC-EVs on the plasma clotting time (n = 7). b OD value of exogenous Xa production activated by different protein levels in UC-EVs (n = 3). Representative curves of thrombin production (c) and ETP (d) after adding UC-EVs (n = 3). e–j Data from the TEG test (n = 3). e Representative diagram of the TEG showing the dose-dependent effects of UC-EVs on blood coagulation. Blood was mixed with UC-EVs at concentrations of 1 mg, 5 mg, and 25 mg per mL of blood. f The effect of different protein levels in UC-EVs on the CT of whole blood. g CT, (h) CFT, (i) alpha-angle, and (j) MCF measured by the TEG (n = 3). Data are shown as the mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns: not significant.

Moreover, TEG was used to assess the UC-EVs-induced whole blood clotting. We observed that recalcified whole blood of healthy volunteers clotted after approximately 10.50 min. However, the CT was significantly reduced to 1.77 ± 0.12 min, 1.23 ± 0.06 min, and 0.83 ± 0.06 min, respectively, by the addition of 1 μg/mL, 5 μg/mL, and 25 μg/mL UC-EVs (Fig. 2e, f), suggesting that UC-EVs could rapidly activate coagulation compared to the control group. Specifically, 1 μg/mL UC-EVs decreased CT and CFT by 5.9 times and 2.3 times (Fig. 2g, h), respectively, and increased the α-angle from 48.77 ± 7.62° to 71.37 ± 3.11° (Fig. 2i). Meanwhile, another TEG parameter, MCF, did not exhibit obvious changes (Fig. 2j). These results suggest that UC-EVs possess procoagulant properties and can influence blood coagulation in a dose-dependent manner.

TF/CD142-dependent procoagulant activity of UC-EVs

Based on these findings, we further investigated whether the TF/CD142 on the surface of UC-EVs was responsible for inducing coagulation. Upon adding various concentrations of TFPI to EV-free plasma, we observed a gradual slowing of plasma clotting to over 120 s as the TFPI concentration increased (Fig. 3a). Moreover, the generation of Xa (Fig. 3b) and ETP was also inhibited by TFPI (Fig. 3c, d). In the TEG test, TFPI prolonged CT by 70% (Fig. 3e, f). These results indicate that TF/CD142 is the main contributor to the procoagulant activity of UC-EVs, and TFPI can effectively inhibit this activity.

Fig. 3. TF/CD142-dependent procoagulant activity of UC-EVs.

a TFPI prolonged the plasma clotting time of UC-EVs (n = 7). b TFPI suppresses the production of Xa (n = 3). c, d TFPI suppresses thrombin production. Representative curves of thrombin production (c) and ETP (d) after adding TFPI and UC-EVs (n = 3). e Representative TEG diagram showing the inhibitory effect of TFPI on the procoagulant activity of UC-EVs. f CT measured by the TEG (n = 3). Data are shown as the mean ± SD. **P < 0.01, ***P < 0.001, ****P < 0.0001, ns: not significant.

Risk of thrombosis after intravenous infusion of UC-EVs

Next, mice were intravenously injected with UC-EVs at a base dose of 0.125 μg/g body weight, gradually increased to 4 μg/g body weight (Supplementary Fig. S2a). Consistent with most animal experiments [51], no obvious acute adverse reactions were observed when the injection dose of UC-EVs was less than 0.25 μg/g (Table 1). However, higher doses of UC-EVs induced a series of adverse symptoms, which were consistent with the phenomenon we observed in our previous study [31]. At moderate doses (0.5 μg/g and 1 μg/g), some mice exhibited respiratory distress, cyanosis, convulsions, hemiplegia, or quadriplegia following the intravenous injection, indicating respiratory and circulatory failure. Some of the mice gradually recovered to normal, while others died. The mortality rate was 10% in the 0.5 μg/g group and 50% in the 1 μg/g group (Table 1). As the dosage of UC-EVs increased, both the incidence and mortality of adverse reactions in mice correspondingly increased. When the injection doses were increased to 2 μg/g and 4 μg/g, the incidence of adverse events increased to 100%, and all mice died within 30 min (Table 1). Based on the mortality rate of mice, probit regression was used to calculate the LD₅₀ of UC-EVs, showing that the administration dose had a significant effect on the mortality rate (P < 0.001). The probit regression equation was obtained as Probit (P) = 0.209 + 5.692 × dose (where P is the dose-response probability); that is, the LD₅₀ of UC-EVs for mice was 0.919 μg/g, with the 95% confidence limit of 0.786–1.082 μg/g.

Table 1.

Incidence and mortality of acute adverse reactions in mice.

Dose	Total number	Wheezing	Cyanosis	Hemiplegia or quadriplegia	Deaths	Mortality rate
(μg/g)	(n)	(n)	(n)	(n)	(n)	(%)
0	10	0	0	0	0	0
0.125	10	0	0	0	0	0
0.25	10	0	0	0	0	0
0.5	10	4	0	1	1	10
1	10	9	5	9	5	50
2	10	10	10	10	10	100
4	10	10	10	10	10	100

To determine the cause of death in mice, we immediately dissected the dead mice and observed several pathological manifestations: the lungs were pink with hemorrhagic spots, the hearts were black, red, and stiff, and the livers were dark red and slightly enlarged. In contrast, the lungs of normal mice were homogeneously pink, the hearts were still beating, and the livers and kidneys were both meaty red with a lighter color (Fig. 4a). We then examined microscopic pathological changes in these organs. As shown in Fig. 4b, there was obvious thrombus formation in the blood vessels around the bronchioles of the lungs, accompanied by fibrin exudation in the interstitial space around the blood vessels. In the left ventricular myocardium of the heart, focal myocardial rupture with hemorrhage was evident (Fig. 4c). Solid organs such as the liver, spleen, and kidney mainly showed congestion (Fig. 4d–f).

Fig. 4. Lethal PE induced by intravenous infusion of UC-EVs in mice.

a Gross examination of the lung, heart, liver, kidney, and spleen of normal and deceased mice. Representative H&E-stained images of the lung (b), heart (c), liver (d), spleen (e), and kidney (f) of normal mice and dead mice. Scale bar: 100 μm. g Fluorescence images of dissected organs from dead mice injected with 4 μg/g DiR-labeled UC-EVs. h Typical PE in pulmonary microvessels (upper) and the presence of UC-EVs at the site of fibrin-like thrombi in blood vessels (lower). PKH26-labeled UC-EVs (bright red), CD31-marked vascular endothelium (green), and DAPI-stained vascular nuclei (blue). Scale bar: 100 μm.

To provide direct evidence for PE occurrence after UC-EVs injection, we used DIR-labeled UC-EVs to detect their organ distribution. The images of excised organs from the dead mice revealed accumulation of UC-EVs in the lungs, while no fluorescence was detected in other organs (Fig. 4g). We also examined the distribution of PKH26-labeled UC-EVs in tissues through a fluorescence microscope (Supplementary Fig. S2b). Immunofluorescence at corresponding H&E fibrin-like thrombus sites showed that red PKH26-positive UC-EVs co-localized with red blood cells in blood vessels around the bronchioles (Fig. 4h). We further quantified the number of pulmonary thrombi in each group of mice and performed PE scoring (Supplementary Table S1), revealing that multiple thrombi in the lungs of mice injected with UC-EVs at a dosage of ≥1 μg/g body weight (Fig. 5a–h).

Fig. 5. UC-EVs infusion induced coagulation reaction in vivo.

a–g Representative H&E-stained lung sections from mice in different UC-EV dose groups. The black-boxed area in the upper image is enlarged and shown in the lower image. A black arrow indicates a thrombus. Scale bar: 100 μm. h PE scores of mice in different UC-EV dose groups (n = 6). i Platelet, Hg (j), WBC (k), and RBC (l) counts were measured in blood samples from mice that received different doses of UC-EVs. m–o Coagulative indices of plasma samples from these mice (n = 5). Data are presented as the means ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 versus the 0 μg/g UC-EV group (control). # indicates that the value was below the measurement range of the analyzer.

The results of blood routine and coagulation function of mice in each group showed a gradual dose-dependent decrease in platelet count with increasing UC-EVs dosage. Notably, platelet counts in groups with doses above 0.25 μg/g were significantly lower than those of the control group (P < 0.05). Infusion of a lethal dose of UC-EVs (≥1 μg/g) resulted in a reduction of platelet levels by more than 75% (Fig. 5i), while levels of WBCs, RBCs, and hemoglobin remained unaffected (Fig. 5j–l). Furthermore, Fg showed a downward trend with increasing doses of UC-EVs (Fig. 5m). The injection of UC-EVs (≥1 μg/g) led to a significant reduction in Fg and a significant prolongation of PT and APTT (P < 0.0001, Fig. 5m–o), exceeding the measurement capabilities of the coagulation analyzer, indicating a severe depletion of coagulation factors.

Strategies to prevent the adverse effect of UC-EVs infusion

Exactly as we hypothesized, intravenous infusion of UC-EVs induced coagulation reaction, with high doses of UC-EVs leading to PE and death in mice. This significant adverse effect poses a challenge to the clinical application of UC-EVs. To prevent this risk, two strategies were proposed in this study.

The first strategy involves controlling the rate of UC-EVs entering into the blood per unit time (Fig. 6a), as the degree of coagulation activation correlated positively with UC-EVs dosage. Our results showed that all mice died after an intravenous bolus injection of a high dose of UC-EVs (4 μg/g). In contrast, no acute symptoms, PE, or deaths (Fig. 6b, Table 2, and Supplementary Table S2) occurred within 24 h in the 1-h intravenous infusion groups with the same dose. Furthermore, routine blood and coagulation analysis showed normal PT, APTT, Fg, and platelet levels in the pump injection group (Fig. 6c–f).

Fig. 6. Prevention of PE by slower infusion of UC-EVs or pre-injection of heparin.

a Schematic representation of continuous intravenous pumping injection of UC-EVs to mice, with UC-EVs being continuously infused intravenously for 1 h. b Representative H&E-stained lung sections and PE scores from mice in different infusion mode groups (n = 5). Scale bar, 100 μm. c–f Coagulative indices and platelet counts of blood samples from mice with different infusion modes (n = 5). # indicates that the value was below the measurement range of the analyzer. g After pre-injection of PBS or heparin (400 U/kg), mice were injected with 4 μg/g UC-EVs via the tail vein. Mice in the PBS group died while those in the heparin group survived. h H&E staining of lungs and PE score of mice (n = 5). Scale bar, 100 μm. j–m Coagulative indices and platelet counts of blood samples from mice with pre-injection of PBS or heparin (n = 5). # indicates that the value was below the measurement range of the analyzer. The organ distribution of UC-EVs was observed through fluorescence imaging (i) or under a microscope (n, o). Data are presented as the means ± SD. **P < 0.01, ***P < 0.001, ****P < 0.0001; ns: not significant; AFI: average fluorescence intensity.

Table 2.

Incidence and mortality of acute adverse reactions in mice with different infusion modes.

Injection mode	Total number	Wheezing	Cyanosis	Hemiplegia or quadriplegia	Deaths	Mortality rate
	(n)	(n)	(n)	(n)	(n)	(%)
Bolus injection of UC-EVs	5	5	5	5	5	100
Pump injection of UC-EVs	5	0	0	0	0	0

In animal models, anticoagulant therapy has proven highly effective in mitigating the procoagulant activity of MSCs, which is also related to TF/CD142 [33]. Therefore, another strategy involves the use of anticoagulant treatment with heparin to prevent the coagulation reaction of UC-EVs (Fig. 6g). We found that a pre-injection of heparin (400 U/kg) completely abolished all acute adverse reactions in mice that received high doses of UC-EVs (4 μg/g), with no deaths or PE (Fig. 6h, Table 3 and Supplementary Table S3). This was accompanied by a significant prolongation of PT and APTT, indicating the presence of heparin in the blood (Fig. 6j, k), and normal Fg and platelet levels (Fig. 6l, m). Moreover, heparin anticoagulation reduced the accumulation of UC-EVs in the lungs and increased their presence in the liver (Fig. 6n, o). These findings suggest that heparin successfully inhibited the coagulation response triggered by UC-EVs.

Table 3.

Incidence and mortality of acute adverse reactions in mice after pre-injection of heparin.

Group	Total number	Wheezing	Cyanosis	Hemiplegia or quadriplegia	Deaths	Mortality rate
	(n)	(n)	(n)	(n)	(n)	(%)
Pre-injection of PBS	5	5	5	5	5	100
pre-injection of heparin	5	0	0	0	0	0

Discussion

In this study, we demonstrated that large UC-EVs activated the coagulation cascade in a dose- and TF/CD142-dependent manner. Intravenous infusion of a high dose of these UC-EVs resulted in intravascular coagulation, PE, and acute death in mice. This is the first thorough report of vascular occlusion and animal death caused by large MSC-EVs infusion, highlighting the potential risks associated with activated coagulation. We further found that modifying the infusion protocol (by slowing down the infusion rate and using heparin as an anticoagulant) could effectively prevent the adverse effects of UC-EVs infusion, as illustrated in Fig. 7.

Fig. 7. Schematic illustrations for the risk and mechanism of PE-induced acute death with the preventive strategies.

When UC-EVs enter the blood, the tissue factor (TF/CD142) on the surface of UC-EVs activates factor VII and forms a complex with it, which in turn activates factor X and generates sufficient thrombin, leading to blood coagulation and thrombus formation, thereby increasing the risk of pulmonary embolism. Conversely, slowing down the infusion rate of UC-EVs or pre-injecting heparin anticoagulant are efficient strategies to block coagulation activation and avoid PE.

Previous studies have reported the procoagulant activity of both small and large MSC-EVs from various sources, and various coagulation-related proteins expressed on the surface of MSC-EVs are thought to be the main triggers of the coagulation cascade [33, 37, 38]. Nonetheless, the procoagulant effect and the underlying mechanism of the large UC-EVs remain unclear. Here, we demonstrated the presence of TF/CD142 protein on the surface of UC-EVs by flow cytometry and Western blot. TF/CD142 triggers the extrinsic coagulation pathway by binding and activating coagulation Factor VII, forming the Factor VIIa/TF complex. This complex subsequently activates Factor X to Xa, which further generates thrombin [32]. Notably, flow cytometry analysis showed that only 4.5% of UC-EVs were TF/CD142 positive, lower than that reported by Skovronova et al. [4]. This might be due to the low sensitivity of the immunological method we used [40, 52]. We further applied more sensitive functional assays to verify the procoagulant activity of TF/CD142 on the UC-EVs. We found that TFPI, an inhibitor of factor Xa and factor VIIa/TF complex [53], could effectively inhibit the procoagulant activity of UC-EVs in vitro, playing a highly effective role against the generation of coagulation factor Xa and thrombin. Therefore, we concluded that the UC-EV-induced coagulation was mainly mediated by TF/CD142. Our results are similar to the previous reports of TF/CD142 involvement in the procoagulant effect caused by large EVs derived from adipose MSC and BM-MSCs [32, 37]. In addition, the role of other procoagulant proteins in the procoagulant activity of UC-EVs remains unknown and deserves further investigation.

The potent procoagulant activity of these UC-EVs has prompted us to explore their safety profile when administered intravenously. Here, for the administration of UC-EVs to mice, we followed the established dosage regimen for large MSC-EVs in previous studies. Clinical trials have typically utilized a low dose below 0.1 µg/g (NCT02138331). In the animal experiments, assuming 1 µg in protein is equivalent to approximately 2 × 10⁹ MSC-EVs [54], the typical dose of large MSC-EVs ranges from 0 to 1 µg/g for an average mouse with a body weight of 25 g [55–57]. Considering the dose-dependent effects of UC-EVs on blood clotting, we set the dosage gradient of UC-EVs from 0 to a high dose of 4 µg/g, to observe the potential adverse effects. Consistent with previous findings, administration of UC-EVs at doses below 1 µg/g resulted in few adverse effects, while a high mortality rate among the mice, accompanied by severe PE, was observed when the dose exceeded 1 µg/g. Pathological examinations confirmed the involvement of UC-EVs in thrombus formation, leading to the obstruction of blood vessels around the bronchioles of the lungs. This explained the immediate occurrence of hypoxemia and respiratory failure following intravenous injection. Furthermore, PE inevitably led to pulmonary hypertension and right heart failure, with evident congestion in solid organs. Our research reveals the potential thromboembolism risk of the large UC-EVs infusion via blood vessels, and suggests similar vigilance is required when using other MSC-EVs (both small and large) products via blood vessels, especially when high doses are used for much stronger therapeutic effects. Nonetheless, animal results could not be directly applied to humans, and further studies are needed to determine whether and how MSC-EVs pose a thromboembolism risk for patients.

It should be noted that there are distinct procoagulant properties between large and small EVs. TF/CD142-presenting large tumor-derived EVs have been identified as the major contributing factors for forming complexes with coagulation tenase activity [58]. Small MSC-EVs have been extensively studied in both preclinical and clinical settings, with intravenous doses ranging from 0.001 to 100 µg/g in animal models and typically below 0.1 µg/g in human trials [59]. To date, no thromboembolic events have been reported, which is likely due to the inherently lower procoagulant activity of small MSC-EVs. Skovronova et al. [4] previously proposed that large MSC-EVs may possess greater procoagulant activity based on their findings of higher TF/CD142 expression in large MSC-EVs; however, the comparative procoagulant activity of large and small MSC-EVs has not been extensively explored due to the limited data for large MSC-EVs. In the present study, the addition of large UC-EVs resulted in a significantly reduced clotting time compared to small UC-EVs (Supplementary Fig. S1d). Additionally, no signs of PE were detected in mice following intravenous injection of 4 µg/g body weight of small UC-EVs (Supplementary Fig. S1e). These results further support the difference in procoagulant activity between the small and large MSC-EVs. To be noted, the procoagulant activity of MSC-EVs could be affected by various confounding factors, such as cell type, culture condition (including media and serum components), expansion number, and passage number [33, 35, 37], and even donor source, as well as EV separation methods. A thorough assessment of procoagulant activity is necessary for in vivo administration.

We also proposed and verified two effective strategies (slowing down the infusion rate or anticoagulation treatment by heparin) to prevent the adverse effect induced by UC-EVs. Slow intravenous injection is a common method to avoid side effects in patients, for example, potassium chloride can be administered via slow intravenous infusion to prevent treatment-induced death in cases of severe hypokalemia [60]. We successfully employed this method in our study for the infusion of UC-EVs. In addition, we demonstrated that heparin, an anticoagulant, efficiently prevented the acute adverse effect of infusing a high dose of UC-EVs. This strategy has previously been explored by various groups to reduce the procoagulant potency of MSCs, which also rely on TF/CD142 to activate the coagulation system and cause thromboembolism [61–63]. A notable example is a report by Liao et al. [64], which indicated that heparin treatment improved the safety and therapeutic effect of large-dose BM-MSCs. Other strategies, such as using different types of anticoagulants (e.g., fondaparinux, rivaroxaban, hirudin, bivalirudin), specific antibodies to directly block TF/CD142, or reducing the expression of TF/CD142 on the surface of parental cells through gene editing techniques, may be more effective in controlling the procoagulant activity of UC-EVs. However, these strategies require careful evaluation for clinical application.

In summary, we demonstrated the dose- and TF/CD142-dependent procoagulant characteristics and risk of intravenously injected UC-EVs in mice in detail, which can be successfully prevented by either slowing down the infusion rate or by heparin injection. Despite numerous explorations of MSC-EVs for various indications, no MSC-EV-based drugs have been approved for marketing. Given the need to increase the dosage of MSC-EVs for better therapeutic effects, the potential risk of thromboembolism should be carefully evaluated. Our research not only provides a basis for developing UC-EV-based therapies but also serves as a safety warning for other types of MSC-EVs, thus facilitating their drug development and clinical application.

Author contributions

ZCC and QBL conceived the idea and directed the entire study. BLY and YYL performed most of the experiments. BLY and YYL drafted the paper, and QBL revised it. QL, FG, WXR, YLC, DW, LYX, JQ, HL, YLY, AYZ, SW, and HXW contributed to conducting research, refining ideas, and finalizing the manuscript. All authors have approved the submitted version of the manuscript.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41401-024-01327-3.