Expression of Tissue Factor and Platelet/Leukocyte Markers on Extracellular Vesicles Reflect Platelet-Leukocyte Interaction in Severe COVID-19

1. Introduction

Thrombotic events and endothelial dysfunction are key pathomechanisms of coronavirus disease-19 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1,2]. In severe COVID-19, deregulated complement activation enhances the recruitment of neutrophils to the infected lungs and promotes tissue factor (TF) expression on neutrophils, monocytes, and endothelial cells, resulting in microvascular thrombosis and endothelial dysfunction [1,3,4]. These events are best described as immunothrombosis, a simultaneous overactivation of the innate immune response and the coagulation system [5]. Activated platelets are main effectors of immunothrombosis and interact with leukocytes to promote pro-inflammatory and pro-coagulant responses [4]. Platelets from COVID-19 patients show a hyperreactive phenotype characterized by the release of platelet factor 4 (PF4), platelet-derived growth factor (PDGF) and thromboxane A2, as well as increased P-selectin expression, which mediates platelet-leukocyte aggregate formation via P-selectin glycoprotein ligand (PSGL)-1 [6,7,8,9,10]. The enhanced aggregate formation of platelets with monocytes or neutrophils in COVID-19 has been correlated with poor prognosis [7,9,11].

The platelet-leukocyte crosstalk can boost the dysregulated cytokine and chemokine response seen in COVID-19 and triggers TF expression on monocytes, neutrophils, as well as endothelial cells [6]. Moreover, aggregate formation of platelets with monocytes drives TF expression on monocytes via P-selectin and integrin αIIb/β3-dependent signaling and is amplified by TNF-α and IL-1β [12]. Likewise, platelets can induce TF expression on neutrophils and promote the release of neutrophil extracellular traps (NETs), representing a scaffold of expelled nuclear DNA and histones that captures and kills pathogens by antimicrobial proteins and proteolytic enzymes [13]. NETs support the pro-thrombotic phenotype by their ability to initiate coagulation, either by presenting TF [14] or by triggering the contact-dependent pathway via factor XII [15]. NETs promote platelet binding and the formation of platelet-neutrophil aggregates, which fuel further NET release and thus exacerbate endothelial damage and thrombotic complications [6,16,17]. Upon activation, platelets as well as neutrophils and monocytes release extracellular vesicles (EVs) into the circulation, which are important mediators of thrombosis due to their exposure of phosphatidylserine (PS) [18,19,20] and TF [21,22,23,24,25]. Furthermore, EVs exert immunomodulatory functions by transferring regulatory and pro-inflammatory molecules, such as nucleic acids, C-reactive protein (CRP) or high mobility group box-1 protein (HMGB-1) [26,27,28].

Here, we characterized the cellular origin of PS-exposing EVs in plasma from COVID-19 patients, as well as their association with inflammation- and coagulation-related parameters, including CRP, HMGB-1, PF4, and TF. We identified TF-bearing EVs co-expressing platelet (CD41) and leukocyte (CD45) markers, reflecting the recently described platelet-leukocyte aggregate formation in severe COVID-19 on the EV level.

2. Results

2.1. Patient Characteristics

In total, 12 critically ill COVID-19 patients receiving mechanical ventilation were included in this study. Twenty-five healthy individuals served as controls. Baseline characteristics and comorbidities are summarized in Table 1. The medication regimens for individual patients are detailed in Supplementary Table S1. The mean age of COVID-19 patients was 79.5 [77.0–82.9] years and the majority of patients were males (83 %). Secondary bacterial (E. coli) and/or fungal (Aspergillus fumigatus) infections were confirmed in 58 % of all patients. The overall mortality was 58 %. The time points of sample collection for individual patients are shown in Figure 1.

2.2. Cellular Origin of Extracellular Vesicles

COVID-19 patients showed significantly increased levels of PS-exposing circulating EVs compared to healthy donors (204,000 [122,375–450,000] vs. 53,500 EVs/µL, Figure 2). These EVs were derived from platelets (CD41⁺; 59,465 [25,980–112,155] vs. 9000 [5850–10,913] EVs/µL), red blood cells (CD235a⁺; 12,385 [7266–57,816] vs. 5400 [4193–7985] EVs/µL), leukocytes (CD45⁺; 11,958 [7359–22,659] vs. 2450 [1830–3378] EVs/µL), and endothelial cells (CD105⁺; 9598 [5843–22,898] vs. 2400 [1788–3335] EVs/µL; see also Supplementary Figures S1 and S2 for the time course of EV concentrations). Regarding the relative abundance of PS-exposing EVs in COVID-19 patients, 26.0 [16.0–33.0] % originated from platelets, 7.0 [4.0–21.3] % from red blood cells, 6.0 [4.0–8.0] % from leukocytes, and 5.0 [4.0–5.0] % from endothelial cells. CD36, which has been linked to platelet hyperactivation [28,29], was expressed on 12 [9.0–14.0] % of all EVs.

2.3. Association of Extracellular Vesicles with CRP, PF4, and HMGB-1

Next to the characterization of their cellular origin, we assessed the association of EVs with CRP, PF4, and HMGB-1, which are mediators of immunothrombosis (Figure 3). COVID-19 patients showed significantly increased levels of CRP⁺ EVs, which represented 43.0 [19.0–57.0] % of the total count of PS-exposing EVs, vs. 4.0 [3.0–7.5] % in healthy controls.

Albeit only a small percentage of EVs was associated with PF4 (COVID-19, 3.0 [3.0–4.0] %; healthy controls, 2.0 [0.5–3.0] %), PF4⁺ EVs were significantly elevated in COVID-19 patients compared to healthy donors (7525 [3878–16,385] vs. 1050 [170–2280] EVs/µL).

HMGB-1 was also detected on EVs. Compared to healthy controls, COVID-19 patients had significantly higher concentrations of HMGB-1⁺ EVs (15,443 [8548–36,240] vs. 1980 [1500–2688] EVs/µL).

2.4. Tissue Factor Expression on Platelet- and Leukocyte-Derived Extracellular Vesicles

In COVID-19 patients, 16.5 [9.0–22.5] % of all PS-exposing platelet EVs were CD41⁺ and CD45⁺, providing evidence for aggregate formation between platelet- and leukocyte-derived EVs, whereas only 3.4 [1.6–8.5] % of CD41⁺CD45⁺ EVs were detected in healthy donors (Figure 4). Remarkably, 55.5 [42.9–69.2] % of all EV aggregates (CD41⁺CD45⁺) in COVID-19 patients co-expressed TF, while the percentage of TF⁺ EV aggregates remained below the detection limit in healthy donors. For individual patients, the percentages of CD41⁺CD45⁺TF⁺ EVs ranged from 25.0–76.5%. EVs from COVID-19 patients expressed TF to a significantly higher extent than EVs from healthy donors (18,925 [10,053,33,028] vs. 1880 [1335–3945] EVs/µL, respectively). In the COVID-19 group, 8.0 [6.0–11.0] % of all EVs were associated with TF as compared to 4.0 [2.0–5.0] % in healthy donors.

2.5. Time Course of Coagulation- and Inflammation-Related Mediators

Statistical modelling of the time effect on coagulation- and inflammation-related mediators revealed a decrease over time in leukocytes, neutrophils, platelets, PF4, EV counts, CRP⁺ EVs, TF⁺ EVs, as well as IP-10 in COVID-19 patients with a probability of >95 % (Table 2; see also Supplementary Figures S3 and S4 for the time course of individual patients). IL-8, IL-1β, G-CSF, nucleosomes, TNF-α, D-dimer, IL-10, and HMGB-1 levels increased over time with a probability in the range of 100.0–92.7 %.

3. Discussion

Immunothrombosis, the joint over-activation of the innate immune response and coagulation, is a central pathomechanism in sepsis and severe COVID-19. Complement activation and cytokine release, platelet hyperactivity, as well as coagulopathy play critical roles in this complex scenario [31,32]. In particular, the crosstalk of platelets and leukocytes has been shown to amplify inflammatory effector functions, as discussed in more detail below [9,17].

Increased levels of circulating EVs released from activated blood cells are well documented in sepsis and severe COVID-19 [33]. EVs released from the plasma membrane expose PS and support coagulation by catalyzing the formation of the tenase and prothrombinase complexes of the coagulation cascade [19]. Therefore, we focused on the characterization of this EV population in our study. Our data confirm the presence of increased levels of circulating PS-exposing EVs in patients with severe COVID-19. In line with previous reports [10,24,34,35], the majority of these EVs originated from platelets.

While flow cytometry is versatile and well established to characterize EVs in plasma samples, the limitations of this approach have to be taken into account. There is evidence that Anx5, which we used to detect PS-exposing EVs in this study, also labels apolipoprotein B-containing lipoproteins, such as low-density lipoprotein [36], challenging the use of Anx5 to uniquely identify EVs in lipoprotein-containing samples. This is why we combined the PS-based approach for EV detection with additional membrane-bound markers (CD41, CD45, CD235a, CD105), which are not exposed on lipoproteins. Moreover, the fact that we compared TF expression on EVs in patient samples to samples from healthy donors greatly limits a potential bias in our results, since lipoproteins would have been present in the control samples, as well.

We found that EVs from COVID-19 patients carry CRP, PF4, and HMGB-1 on their surface, which may further fuel immunothrombosis. Elevated levels of CRP⁺ EVs have been described in sepsis, but also in myocardial infarction, and several studies have highlighted the pro-inflammatory characteristics of these CRP-bearing EVs [26,37,38]. Sustained platelet activation in COVID-19 boosts the release of PF4 and HMGB-1, inducing neutrophil activation and the release of NETs, which strongly promote coagulation [4,10,17,39]. Platelet-derived HMGB-1⁺ EVs have been reported as markers of platelet activation and are associated with a poor prognosis in COVID-19 patients [40]. Furthermore, we found increased levels of CD36⁺ EVs in COVID-19 patients. Platelet glycoprotein CD36 acts as a receptor for membrane-derived EVs, as it binds to PS on their surface [30], resulting in platelet activation, aggregation, and thrombus formation [29]. Accordingly, increased CD36 expression indicates a higher risk of venous and arterial thromboembolism [41].

Next to characterizing PS-exposing EVs, which propagate coagulation as discussed above, we focused on the expression of TF on EVs, which is the main initiator of coagulation. Monocytes represent the predominant source of blood-borne TF, which can be passed on to EVs originating from these cells [42,43]. While we were not able to obtain samples for all time points from each COVID-19 patient due to extubation or death, our data show that each patient had increased levels of TF⁺ EVs at all available time points in comparison to healthy controls. While several previous studies have also reported increased levels of TF-expressing EVs in COVID-19 patients, which correlated with disease severity [21,24,43], others failed to detect differences in TF⁺ EVs between COVID-19 patients and healthy controls [35]. These conflicting results may at least in part be attributed to different antibody clones used for the detection of TF or to different fluorochrome-to-protein ratios of antibody-fluorochrome conjugates used for the flow cytometric detection of TF⁺ EVs.

Enhanced platelet-leukocyte aggregate formation is known to occur in various pathological conditions, including sepsis and COVID-19, where increased platelet-monocyte interaction has been linked to disease severity [39]. Our data suggest that the enhanced interaction of platelets and leukocytes in COVID-19 is also reflected on the EV level. About 17 % of all platelet-derived EVs displayed the leukocyte marker CD45, suggesting aggregate formation between platelet-derived and leukocyte-derived EVs. Notably, 56 % of all CD41⁺CD45⁺ EVs expressed TF, likely of monocyte or neutrophil origin. It is well established that activated platelets express P-selectin and interact with monocytes through PSGL-1 [44], a mechanism which may also mediate aggregate formation between platelet- and leukocyte-derived EVs. Platelets induce TF expression on monocytes through P-selectin and integrin αIIb/β3 signaling [39] and stimulate the release of TF-conveying NETs by neutrophils [45]. In turn, increased TF expression has been associated with an upregulation of CD16 on monocytes, inducing a shift from CD16^– classical monocytes, which are mainly phagocytic, towards inflammatory CD16⁺ intermediate and non-classical monocytes [12].

Platelet-monocyte aggregates also promote the secretion of inflammatory mediators. Incubation of platelets from COVID-19 patients with monocytes from healthy donors triggers the release of IL-1β, IL-6, IL-8, MIP-1α, MCP-1, TNF-α, PF4, and PDGF [12,39], and elevated levels of these factors were observed in clinical samples from COVID-19 patients [46]. Likewise, our study cohort presented increased levels of inflammatory mediators that dynamically changed during the course of mechanical ventilation, with an increase of IL-8, IL-1β, G-CSF, nucleosomes, TNF-α, D-dimer, IL-10, and HMGB-1 over time. A similar time-dependent increase has been reported for IL-6, IL-8, IL-1β, and TNF-α in COVID-19 [47]. We further observed decreasing platelet, leukocyte, and neutrophil counts in mechanically ventilated COVID-19 patients over time.

To conclude, we report here for the first time that TF-expressing platelet- and leukocyte-derived EV aggregates are present in severely ill COVID-19 patients, and we propose that these aggregates may act as amplifiers of immunothrombosis.

4. Materials and Methods

4.1. Patients and Sample Collection

Twelve patients with PCR-confirmed or suspected SARS-CoV-2 infection, requiring mechanical ventilation were included in this study at the Department of Internal Medicine, Hospital St. Vinzenz, Zams, Austria between November 2020 and January 2021 [32]. Sample collection was approved by the ethics committee of the Medical University of Innsbruck (1144/2020). The study was conducted in accordance with the declaration of Helsinki. Whole blood samples anticoagulated with EDTA (S-Monovette® K3 EDTA, Sarstedt, Nümbrecht, Germany) were obtained during routine blood collection every 24 h. Overall 134 samples were collected. The control group (n=25) consisted of healthy individuals from whom EDTA-anticoagulated blood was obtained after written consent. Platelet-poor plasma was obtained by centrifugation of whole blood at 2000 x g for 15 min at 22°C and stored at –80°C until further analysis. Routine laboratory measurements (C-reactive protein, CRP; procalcitonin, PCT; D-dimer) and blood cell counts were obtained as part of standard medical care. Non-survival of patients was defined as death during mechanical ventilation or within 14 days following extubation.

4.2. Flow Cytometric Characterization of Phosphatidylserine-Exposing Extracellular Vesicles

Due to their pro-coagulant properties, we focused on PS-exposing EVs, which represent a subpopulation of larger EVs derived from the cell membrane (“microvesicles”). These EVs were characterized by flow cytometry using a CytoFLEX LX device (Beckman Coulter, Brea, CA) equipped with 405 nm, 488 nm, 561 nm, and 631 nm lasers. Calibration of the flow cytometer was performed with fluorescent silica beads (1 μm, 0.5 μm, 0.1 μm; excitation/emission 485/510 nm; Kisker Biotech, Steinfurt, Germany). The triggering signal for EVs was set to the violet side scatter (405 nm), and the EV gate was set below the 1 µm bead cloud as previously described [48,49].

For staining, plasma samples were diluted 1:100 in 0.1 µm sterile-filtered Annexin V (Anx5) binding buffer (BD Biosciences, San Jose, CA). The cellular origin of EVs was assessed by staining of the diluted samples (100 μL each; 30 min; room temperature; in the dark) with a combination of 2 µL PC7-conjugated anti-CD41 (Beckman Coulter) as platelet marker, 2 µL FITC-conjugated anti-CD235a (eBioscience, San Diego, CA) as red blood cell marker, 5 µL PE-conjugated anti-CD105 (Becton Dickinson, Franklin Lakes, NJ) as endothelial marker, 2 µL PB-conjugated anti-CD45 (Beckman Coulter) as leukocyte marker, and 2.5 µL AF700-conjugated anti-hCD36 (BioLegend) to detect platelet glycoprotein IV. The association of EVs with coagulation- and inflammation-related mediators was assessed by staining aliquots of the diluted samples (100 µL each; 30 min; room temperature; in the dark) with 1 µL FITC-conjugated anti-CRP (Abcam) to detect CRP, 5 µL AF700-conjugated anti-hHMGB-1 (R&D Systems) to detect HMGB-1, and 2.5 µL PE-conjugated anti-hCXCL4 (R&D Systems) to detect PF4. Calibration and all controls for the flow cytometric characterization of EVs are shown in Supplementary Figure S5.

The co-expression of TF on EVs of platelet and leukocyte origin was assessed by staining with 2 µL FITC-conjugated anti-hTF (Biomedica Diagnostics, Stamford, CT), 2 µL PC7-conjugated anti-CD41, and 2 µL PB-conjugated anti-CD45 for 30 min at 4°C in the dark. APC-conjugated Anx5 (2 µL, BD Biosciences) was used as marker for EVs exposing PS. The gating strategy to define TF⁺ EVs is shown in Supplementary Figure S6. Prior to use, all fluorochrome conjugates were centrifuged at 18,600 x g for 10 min at 4°C to remove eventual precipitates. All fluorochrome conjugates as well as the respective antibody clones are listed in Table 3.

Prior to analysis, stained samples were diluted 1:5 in 0.1 µm sterile-filtered Anx5 binding buffer. Acquisition was performed for 2 min at a flow rate of 10 µL/min and Anx5 positive events in the EV gate were quantified. Data were analyzed using the Kaluza Software 2.1 (Beckman Coulter). Further details on the flow cytometric characterization of EVs are reported according to the MIFlowCyt-EV framework in Supplementary Table S2.

4.3. Quantification of Cytokines, Chemokines, and Growth Factors

A total of 27 cytokines, chemokines, and growth factors were analyzed using a magnetic bead array assay (Bio-Plex Pro^TM human cytokine 27-plex; Bio-Rad, Vienna, Austria), including interleukin (IL)-1β, IL-1 receptor antagonist (IL-1ra), IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-12p70, IL-13, IL-15, IL-17A, interferon-gamma (IFN-γ), tumor necrosis factor-alpha (TNF-α), monocyte chemotactic protein-1 (MCP-1), macrophage inflammatory protein-1 alpha and beta (MIP-1α, MIP-1β), regulated on activation, normal T-cell expressed and secreted (RANTES), eosinophil chemotactic protein (eotaxin), interferon-inducible protein 10 (IP-10), granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), basic fibroblast growth factor (bFGF), platelet derived growth factor (PDGF), as well as vascular endothelial growth factor (VEGF). Plasma samples were diluted 1:4 with sample diluent and analyzed according to the manufacturer instructions.

4.4. Quantification of PF4, HMGB-1, and Nucleosomes

Plasma concentrations of PF4 and HMGB-1 were quantified by ELISA (R&D Systems, Minneapolis, MN, and IBL International, Hamburg, Germany). The cell death detection ELISA (Roche, Mannheim, Germany) was used to determine nucleosome levels in plasma samples.

4.5. Statistical Analysis

Statistical analysis was carried out using GraphPad Prism version 9.5.1 (La Jolla, CA, USA). Data are represented as median [IQR; interquartile range]. Groups were compared using the Mann-Whitney test. A value of p < 0.05 was considered as statistically significant.

To assess the effect of time on different parameters, we used the Bayesian methods due to advantages over the frequentist framework [50,51]. Hierarchical models were created taking the repeated measures for the individual patients into account. To model the effect of time, we used the individual parameters as outcome variables, as well as time and patient as predictors. To estimate the probability of time being positively or negatively associated with the parameters (indicating an increase or decrease over time, respectively), we used the posterior percentages with positive vs. negative slope values for the predictor time. We used the default priors for this analysis. To estimate whether the change of the parameters was associated with survival, we employed the change of the measure from the first observation as predictor variable. A logistic model with survival as outcome variable was created. Again, the patient was used as grouping variable. Prior for the slope were set to normal with mean = 0 and standard deviation = 3. We used the logit of the slope estimate (transformed to percentages) for the effect of change of the parameters on survival. Analysis was carried out using R (version 4.2.2). The tidyverse [52] package was used for data wrangling and brms [53] to create the statistical models.