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

Severe COVID-19 is frequently associated with thromboembolic complications. Increased platelet activation and platelet–leukocyte aggregate formation can amplify thrombotic responses by inducing tissue factor (TF) expression on leukocytes. Here, we characterized TF-positive extracellular vesicles (EVs) and their cellular origin in 12 patients suffering from severe COVID-19 (time course, 134 samples overall) and 25 healthy controls. EVs exposing phosphatidylserine (PS) were characterized by flow cytometry. Their cellular origin was determined by staining with anti-CD41, anti-CD45, anti-CD235a, and anti-CD105 as platelet, leukocyte, red blood cell, and endothelial markers. We further investigated the association of EVs with TF, platelet factor 4 (PF4), C-reactive protein (CRP), and high mobility group box-1 protein (HMGB-1). COVID-19 patients showed higher levels of PS-exposing EVs compared to controls. The majority of these EVs originated from platelets. A higher amount of EVs in patient samples was associated with CRP, HMGB-1, PF4, and TF as compared to EVs from healthy donors. In COVID-19 samples, 16.5% of all CD41+ EVs displayed the leukocyte marker CD45, and 55.5% of all EV aggregates (CD41+CD45+) co-expressed TF, which reflects the interaction of platelets and leukocytes in COVID-19 on an EV level.

The platelet-leukocyte crosstalk can boost the dysregulated cytokine and chemokine response seen in COVID-19 and triggers TF expression on monocytes, neutrophils, and 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 via antimicrobial proteins and proteolytic enzymes [13].NETs support the pro-thrombotic phenotype via 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.

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.

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.

Time Course of Coagulation-and Inflammation-Related Mediators
Statistical modeling 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, and IP-10 in COVID-19 patients with a probability of >95% (Table 2; see also Supplementary Figure S3 for the time course of inflammatory mediators).IL-8, IL-1β, granulocyte colony-stimulating factor (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%.
[1335-3945] EVs/µL, respectively).In the COVID-19 group, 8.0 [6.0-11.0]% of all EVs (which were defined as Anx5 + events in the EV gate as described in Section 4.2) were associated with TF as compared to 4.0 [2.0-5.0]% in healthy donors (Figure 4, right panel).TF + EVs correlated with soluble PF4 (r = 0.5282, p < 0.001).No correlation of TF + EVs with D-dimer or nucleosomes was detected.The baseline levels of TF + EVs did not differ between survivors and non-survivors.However, the small number of patients did not allow for a correlation to clinical endpoints.

Time Course of Coagulation-and Inflammation-Related Mediators
Statistical modeling 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, and IP-10 in COVID-19 patients with a probability of >95% (Table 2; see also Supplementary Figure S3 for the time course of inflammatory mediators).IL-8, IL-1β, granulocyte colony-stimulating factor (G-CSF), nucleosomes, TNF-α, Ddimer, IL-10, and HMGB-1 levels increased over time with a probability in the range of 100.0-92.7%.

Discussion
Immunothrombosis, the joint overactivation 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, and 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 [27].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 [40].
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 [41,42].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 [21,24,42], 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 the different antibody clones used for the detection of TF or to the 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 at 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, 55.5% of all CD41 + CD45 + EVs expressed TF, likely of monocyte or neutrophil origin.As a limitation, our flow cytometric protocol based on the pan leukocyte marker CD45 did not allow for the discrimination between monocyte-derived and neutrophil-derived EVs, to further characterize the source of blood-borne TF.
It is well established that activated platelets express P-selectin and interact with monocytes through PSGL-1 [43], a mechanism that 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 [44].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].
To conclude, we report here for the first time that TF-expressing platelet-and leukocytederived EV aggregates are present in severely ill COVID-19 patients, and we propose that these aggregates may act as amplifiers of immunothrombosis.

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× g for 15 min at 22 • C and stored at −80 • C until further analysis.Routine laboratory measurements (C-reactive protein, CRP; procalcitonin, PCT; and 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.

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") and can be detected using Annexin V (Anx5).To avoid interference of lipoproteins, which can also expose PS, we combined Anx5 staining with additional membrane-bound markers (CD41, CD45, CD235a, CD105, and CD36) that are absent on PS-exposing lipoproteins.EVs were characterized by flow cytometry using a CytoFLEX LX device (Beckman Coulter, Brea, CA, USA) 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, and 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 [47,48].
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, USA), 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 S5.Prior to use, all fluorochrome conjugates were centrifuged at 18,600× g for 10 min at 4 • C to remove eventual precipitates.All fluorochrome conjugates and 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.

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 statistically significant.
To assess the effect of time on different parameters, we used the Bayesian methods due to advantages over the frequentist framework [49,50].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 in the parameters was associated with survival, we employed the change in the measure from the first observation as the predictor variable.A logistic model with survival as the outcome variable was created.Again, the patient was used as the grouping variable.Priors for the slope were set to normal, with a mean = 0 and a standard deviation = 3.We used the logit of the slope estimate (transformed to percentages) for the effect of changes in the parameters on survival.The analysis was carried out using R (version 4.2.2).The tidyverse [51] package was used for data wrangling and brms [52] to create the statistical models.

Figure 1 .
Figure 1.Timeline of sampling.Samples from 12 COVID-19 patients were analyzed in this study (134 samples in total).Sampling was initiated on the day of intubation.Blue frames indicate sampling, (♦) marks extubation, and () death of patients.

Figure 1 .
Figure 1.Timeline of sampling.Samples from 12 COVID-19 patients were analyzed in this study (134 samples in total).Sampling was initiated on the day of intubation.Blue frames indicate sampling, ( ) marks extubation, and ( †) death of patients.

Figure 2 .
Figure 2. Cellular origin of extracellular vesicles in COVID-19 patients.Extracellular vesicles in COVID-19 patients (n = 134; black) and healthy controls (n = 25; white) were characterized by flow cytometry and stained by a combination of anti-CD41 as platelet marker, anti-CD235a as red blood cell marker, anti-CD45 as leukocyte marker, anti-CD105 as endothelial marker, and anti-CD36 to detect platelet glycoprotein IV as described in Section 4. Annexin V (Anx5) was used as marker for EVs exposing phosphatidylserine.Data are given as median [IQR; interquartile range] and were compared using the Mann-Whitney test (** p < 0.01, *** p < 0.001).2.3.Association of Extracellular Vesicles with CRP, PF4, and HMGB-1Next to the characterization of their cellular origin, we assessed the association of EVs with CRP, PF4, and HMGB-1, which are mediators of immunothrombosis (Figure3).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

Figure 2 .
Figure 2. Cellular origin of extracellular vesicles in COVID-19 patients.Extracellular vesicles in COVID-19 patients (n = 134; black) and healthy controls (n = 25; white) were characterized by flow cytometry and stained by a combination of anti-CD41 as platelet marker, anti-CD235a as red blood cell marker, anti-CD45 as leukocyte marker, anti-CD105 as endothelial marker, and anti-CD36 to detect platelet glycoprotein IV as described in Section 4. Annexin V (Anx5) was used as marker for EVs exposing phosphatidylserine.Data are given as median [IQR; interquartile range] and were compared using the Mann-Whitney test (** p < 0.01, *** p < 0.001).

Figure 4 .
Figure 4. Aggregate formation between platelet-derived and leukocyte-derived EVs and TF expression on EVs in COVID-19 patients.Extracellular vesicles in COVID-19 patients (n = 134; black) and healthy controls (n = 25; white) were characterized by flow cytometry.Aggregates of plateletand leukocyte-derived EVs were identified as CD41 + CD45 + events in the EV gate as described in Section 4 (left: EV counts/µL and the percentage of platelet EVs carrying CD45 as leukocyte marker are shown).A total of 55.5% of CD41 + CD45 + EVs in COVID-19 patients expressed TF (middle).Overall, TF expression on EVs was significantly enhanced in COVID-19 patients (right).Data are given as median [IQR] and were compared using the Mann-Whitney test (*** p < 0.001); b.d., below detection.

Figure 4 .
Figure 4. Aggregate formation between platelet-derived and leukocyte-derived EVs and TF expression on EVs in COVID-19 patients.Extracellular vesicles in COVID-19 patients (n = 134; black) and healthy controls (n = 25; white) were characterized by flow cytometry.Aggregates of platelet-and leukocyte-derived EVs were identified as CD41 + CD45 + events in the EV gate as described in Section 4 (left: EV counts/µL and the percentage of platelet EVs carrying CD45 as leukocyte marker are shown).A total of 55.5% of CD41 + CD45 + EVs in COVID-19 patients expressed TF (middle).Overall, TF expression on EVs was significantly enhanced in COVID-19 patients (right).Data are given as median [IQR] and were compared using the Mann-Whitney test (*** p < 0.001); b.d., below detection.

Table 1 .
Patient characteristics at baseline.

Table 1 .
Patient characteristics at baseline.
Data are represented as median [interquartile range] and counts (percentage); n.a., not applicable.

Table 2 .
Effect of time on inflammation-and coagulation-related parameters.Probability (P) increase and decrease indicate the probability that a certain parameter will increase or decrease, respectively, depending on time, according to the statistical model.The Estimate refers to the effect size, as an increase of time by one will change the parameter by the Estimate value.The lower and upper 95% credibility intervals (CI) are denoted as CI lower and CI upper .As a measure for the model fit, Rhat as well as the Bulk and Tail effective sample size (ESS) are given, confirming acceptable model fit.

Table 3 .
Antibodies and fluorochrome conjugates used for flow cytometry.