Postprandial Increase in Blood Plasma Levels of Tissue Factor–Bearing (and Other) Microvesicles Measured by Flow Cytometry: Fact or Artifact?

Tissue factor (TF)–bearing microvesicles (MVs) and exosomes may play a role in hemostasis and thrombosis. MVs may be quantified by flow cytometry (FC)–based detection of phosphatidylserine (PS)-positive submicron particles carrying specific antigens, although interference from lipoproteins complicates this approach. In this study, we evaluated the effect of food intake on blood levels of TF-bearing particles measured by FC and small extracellular vesicles (EVs) measured by a protein microarray–based test termed EV Array. Platelet-free plasma (PFP) was obtained from 20 healthy persons in the fasting state and 75 minutes after consumption of a meal. Postprandial changes in the concentration of PS-positive particles, including subgroups binding labeled antibodies against TF, CD41, CD146, and CD62E, respectively (FC), small EVs (EV Array), and TF antigen and procoagulant phospholipids (PPLs) were measured. Furthermore, we tested the effect on FC results of in vitro addition of lipoproteins to fasting PFP. We found significantly increased plasma concentrations of PS-positive particles and all examined subgroups postprandially, while no changes in small EVs, PPL, or TF antigen levels were found. Levels of all types of particles measured by FC were also elevated by lipoprotein spiking. In conclusion, meal consumption as well as in vitro addition of lipoproteins to fasting plasma induces increased levels of PS-positive particles as measured by FC, including TF-positive subtypes and subtypes exposing other antigens. While the observed postprandial increase may to some extent reflect elevated MV levels, our results indicate a substantial interference from lipoproteins.


Introduction
Tissue factor (TF) has been established as a pivotal element in activation of the clotting cascade. 1 It is well described that TF expressed subendothelially in blood vessels comes into contact with blood upon tissue injury and, in combination with activated FVII, initiates the coagulation process. 2 However, TF is also produced in monocytes and may furthermore be expressed in platelets and activated endothelial cells. 1 In plasma, TF circulates both in its full-length form incorporated in the membranes of circulating cell-derived extracellular vesicles (EVs) and in an alternatively spliced soluble form. [3][4][5] The role of blood-borne TF in the coagulation process is not clear, 2,5 but TF-bearing EVs (TFþ EVs) have been suggested to play a role in hemostasis and thrombus formation, 6,7 and TF activity and TFþ EVs quantitated by flow cytometry (FC) have been shown to be increased in various thrombogenic conditions. 1 Although consensus on the nomenclature of different types of EVs has not been achieved, 8 smaller sized EVs, typically with a diameter of less than 150 nm, may be termed exosomes, whereas larger EVs are often referred to as microvesicles (MVs). 9 While these two types of EVs are formed and released from cells by different mechanisms, 10,11 indications exist that both types possess the ability to carry TF in their phospholipid bilayer membrane. 1,5 A significant postprandial increase in plasma concentration of particles within the size range of exosomes and MVs has been demonstrated using nanoparticle tracking analysis 12 and tunable resistive pulse sensing (TRPS). 12,13 However, these methods lack the ability to distinguish between EVs and other submicron particles, including phospholipid monolayer-bounded lipoproteins such as chylomicrons, low-density lipoproteins (LDLs), and very low-density lipoproteins (VLDLs), which are present in plasma and display an altered relative distribution postprandially. 13,14 FC enables some characterization of the particles by fluorescence-based detection of phosphatidylserine (PS) combined with celltype specific antigens and it has previously been reported in FC-based studies that blood concentrations of MVs are increased in the postprandial state. [15][16][17] However, recent investigations have indicated that MV measurement by FC on plasma may suffer from considerable interference from lipoproteins, which are not eliminated by state-of-theart MV isolation methods. 13 Although PS is a well-established marker of MVs 10,18,19 and has been used for MV labeling in FC-based investigations, 13,16 the sensitivity 18 as well as the specificity 20 of PS-positive (PSþ) particles as indicators of MVs can be questioned.
EV Array is a protein microarray-based technique designed for measurement of small EVs. 21 Biotinylated antibodies against CD9, CD63, and CD81, often regarded as exosome markers, but actually also present in MVs, 22 can be used in EV Array for sensitive detection of small EVs. 21 The aim of this study was to investigate the impact of food ingestion on plasma levels of PSþ particles (using FC) and small EVs (using EV Array) in general and their TF-bearing subtypes in particular as well as the effect of prandial state on levels of circulating TF antigen (Ag), TF activity, and procoagulant phospholipids (PPLs). Considering the recent questioning of the reliability of FC-based MV measurements, 13 we evaluated the impact of prandial state and, parallelly, the impact of in vitro addition of lipoproteins to fasting plasma on levels of particles positive and negative for PS and other markers indicating cell origin, discussing the potential influence of lipoproteins on postprandial changes of these analytes and the resulting limitations of FC in MV measurements on plasma.

Study Population and Sample Handling
In this study, which was approved by the local research ethics committee, we analyzed platelet-free plasma (PFP) samples from 20 volunteers (13 females and 7 males) with a median age of 44 (interquartile range [IQR], 33-58) years who were considered healthy based on a health status questionnaire. Additional blood tests were performed to screen for ongoing disease and to measure total cholesterol and triglyceride (TG) concentrations in the fasting and postprandial samples. From each person, a fasting venous blood sample was collected in a 9-mL Vacuette 3.2% sodium citrate plastic tube (Greiner Bio-One, Kremsmünster, Austria) in the morning before consumption of a nonstandardized breakfast with a choice of bread, butter, cheese, jam, marmalade, cake, coffee, tea, milk, and sugar. Seventy-five minutes following commencement of breakfast, a postprandial sample was collected. The first 3.5 mL of blood from each sampling were discarded. Samples were centrifuged twice at 2,500 g for 15 minutes at room temperature, as recommended by Lacroix et al, 23 to obtain PFP, which was stored at À80°C until analysis.

Flow Cytometry
Plasma levels of MVs were evaluated by a modified FC approach based on a previously described method, 24 using a BD FACSAria III High Speed Cell Sorter (BD Biosciences, San Jose, California, United States). For each analysis, 50 μL of freshly thawed PFP were transferred to a TruCount tube (BD Biosciences) containing a lyophilized pellet, releasing a known number of fluorescent beads as an internal standard and enabling particle concentration determination according to the insert instruction of the manufacturer. Subsequently, 7 µL of fluorescein isothiocyanate (FITC)-conjugated lactadherin (Haematologic Technologies Inc., Essex Junction, Vermont, United States) were added to each sample, utilizing the binding affinity of lactadherin for PS, 25 to label particles exposing PS. All particles within the defined size gate (see below) were collectively designated as particles. Particles positively stained with FITCconjugated lactadherin (which binds to PS) were designated as PSþ particles. To identify PSþ particles originating from platelets, 26 particles were further labeled with 3 µL of allophycocyanin-conjugated antihuman CD41 IgG1, κ (BioLegend, San Diego, California, United States). We identified endotheliumderived particles 27 by labeling with 20 µL of phycoerythrin (PE)-conjugated antihuman CD146 IgG1, κ (BD Biosciences), and furthermore activated endothelium-derived particles 27  labeling with 20 µL of PE-conjugated antihuman CD62E IgG1, κ (BD Biosciences). TF-bearing (TFþ) particles were identified by labeling with 20 µL of PE-conjugated antihuman TF IgG1, κ (BioLegend). TFþ particles that were not positively stained with FITC-conjugated lactadherin were designated TFþ PSÀ particles. Isotype controls matching each antibody were used as negative controls. After 30 minutes of incubation at 4°C in the dark, 200 µL of Dulbecco's phosphate-buffered saline (PBS) buffer (Lonza, Basel, Switzerland) that had been filtered through a sterile 0.2-µm Q-Max syringe filter (Frisenette, Knebel, Denmark) were added to each labeled sample. A size gate was established by preliminary standardization experiments using a blend of size-calibrated fluorescent polystyrene beads with sizes ranging from 0.2 to 0.9 μm. The upper and outer limits of the gate were established just above the size distribution of the 0.9-µm beads and the lower limit was set to include the 0.2-µm beads to obtain a size gate of about 0.2-1.0 µm in a forward (FSC) and side scatter (SSC) setting. A discriminator was set to 200 at the SSC-H parameter (instead of FSC-H) to avoid exclusion of smaller sized particles. Event numbers were counted at a maximal flow rate of 20,000 events/second and stopped after 60 seconds or when the MV gate reached at least 500,000 events. Flow rates of postprandial samples with increased particle numbers were adjusted to comply with the maximal flow rate limit.
To study the effect of in vitro addition of lipoproteins to plasma, four fasting PFP samples were spiked with an LDL isolate and a VLDL isolate, respectively, prior to the same labeling and incubation steps and FC analysis as described above. However, it was performed at a later time point on different plasmas and different batches of antibodies. One part of human LDL 5 mg/mL (Kalen Biomedical, Montgomery Village, Maryland, United States) was added to 10 parts of PFP and, parallelly, 1 part of human VLDL 1 mg/mL (Kalen Biomedical) was added to another 10 parts of PFP, resulting in increases in TG concentration of 0.1 and 0.2 mmol/L, and increases in cholesterol concentration of 2.2 and 0.2 mmol/L, respectively. As controls, we tested the effect of addition of the LDL and VLDL isolates and performing the incubation steps on samples not containing PFP, which was, in these control samples, replaced by an equal amount of PBS.
Catching and visualization of the EVs were performed as described previously by Jørgensen et al 29 with the following modifications: After blocking, 10 µL of freshly thawed PFP diluted in wash buffer (0.05% Tween 20 [Sigma-Aldrich, St. Louis, Missouri, United States] in PBS) were incubated at room temperature for 2 hours followed by an overnight incubation at 4°C. Following a wash, the slides were incubated for 2 hours with detection antibodies. The detection was performed with an "exosome cocktail," which is a combination of fluorescently labeled versions of the exosome-related markers anti-CD9, anti-CD63, and anti-CD81 (Ancell Corporation, Stillwater, Minnesota, United States) diluted 1:1,500 in wash buffer. After a wash, a subsequent 30-minute incubation step with cyanine 5 (Cy5)-labeled streptavidin (1:1,500; Life Technologies, Carlsbad, California, United States) in wash buffer was performed for detection. Prior to scanning at 635 nm, the slides were washed once in wash buffer and once in MilliQ water and dried using a Microarray High-Speed Centrifuge (ArrayIt, Sunnyvale, California, United States). Scanning and spot detection was performed as previously described. 29 Additional Analyses PPL-dependent clotting time (PPL-CT) was measured with STA Procoag PPL (Diagnostica Stago, Asnières, France) on an STA Compact Coagulation Analyzer (Diagnostica Stago).
TF Ag was measured by Imubind antihuman Tissue Factor ELISA (Sekisui Diagnostics, Lexington, Massachusetts, United States) using a SpectraMax M2 Microplate Reader (Molecular Devices, Sunnyvale, California, United States) measuring the absorbance at a wavelength of 450 nm. In the following, we refer to this method as TF Ag ELISA .
A supplemental method for TF Ag measurement, represented by a modified version of EV Array with polyclonal anti-TF antibodies (R&D Systems Inc.) as capturing agents and biotin labeled polyclonal anti-TF antibodies (R&D Systems Inc.) as detecting agents, was applied and designated TF Ag EV Array .
TF activity was evaluated with Tissue Factor Human Chromogenic Activity Assay Kit (Abcam, Cambridge, UK), and the absorbance at a wavelength of 405 nm was measured with SpectraMax M2 Microplate Reader (Molecular Devices).
Lipid levels were measured as routine analyses at the Department of Clinical Biochemistry, Aalborg University Hospital, using a Cobas 8000 Modular Analyzer (Roche Applied Science, Penzberg, Germany).

Statistics
Statistical analyses were performed and graphs created using GraphPad Prism, version 6.01 (GraphPad Software, Inc., La Jolla, California, United States). Data, for which a normal distribution was not rejected when applying the Anderson-Darling test, were presented with mean value and standard deviation (SD), while p-value for pairwise comparison was calculated using paired t-test and degree of correlation indicated by Pearson's r. Data, for which a normal distribution was rejected, were presented with median value and IQR, while p-value for pairwise comparison was evaluated using Wilcoxon's matched pairs signed-rank test and degree of correlation indicated by Spearman's r s . This document was downloaded for personal use only. Unauthorized distribution is strictly prohibited.
Two-sided p-values were given, and values below 0.05 considered significant.
Pooled coefficient of variation (CV) indicating the degree of consistency of results was calculated as the square root of the mean of the squared CVs for measurements on each subject.

Flow Cytometry
FC results on the fasting and postprandial samples are shown in ►Fig. 1 (concentrations of total and PSþ particles for each subject) and ►Fig. 2 (levels of antibody labeled subtypes). The concentration of total particles increased substantially in all subjects after food intake, the median rise being 14-fold (IQR, 5-30-fold). A less extensive, yet clearly statistically significant, postprandial increase in the concentration of PSþ particles and all studied subtypes of these, including TFþ PSþ particles, was observed. The relative increases from fasting to postprandial state are shown in ►Table 1. Total TFþ particles (i.e., TFþ PSþ and TFþ PSÀ particles) were also detected and TFþ PSÀ particles constituted a considerable part of all TFþ particles and increased even more postprandially. Representative scatter plots of TFþ PSþ events in a fasting and a postprandial sample are shown in ►Fig. 3A, B.
Also shown in ►Table 1 are the relative increases after addition of lipoproteins in vitro. The absolute numbers are shown in ►Figs. 1 and 2. This investigation was performed later than the previous investigation on the fasting/postprandial samples, and generally the numbers of particles were lower, but the data indicate that spiking PFP with either LDL or VLDL results in variably increased levels of measur-able total particles, PSþ particles, and subgroups staining positive for TF, CD41, CD146, and CD62E, respectively. Since the measurements on samples with added VLDL or LDL showed an increased level of the particles measured in each subgroup after addition, it is highly significant for most of the subgroups: for each subgroup, p < 0.005, n ¼ 8 (►Table 1). Representative scatter plots of TFþ PSþ events in a fasting sample before and after addition of VLDL are shown in ►Fig. 3C, D. In control samples, where PBS was spiked with LDL and VLDL, none of the antibody-labeled subtypes of PSþ particles appeared.

EV Array
Fasting and postprandial fluorescence intensities measured with EV Array are shown in ►Fig. 4. No significant differences between the fasting and postprandial samples were found. CD62Eþ EVs were undetectable in the majority of the study participants, while the other investigated subgroups, including TFþ EVs and PSþ EVs (lactadherin capture), were detected in most subjects. We observed a substantial variation and markedly lower fluorescence intensities, including several below the detection limit, using anti-CD63 antibodies as compared with anti-CD9 and anti-CD81 antibodies. Results on the two last-mentioned exhibited consistency between fasting and postprandial results, showing a very low difference between the two measurements.

PPL, TF Antigen, and TF Activity Analysis
Results on biological variation in PPL-CT and TF Ag levels as measured by TF Ag ELISA and TF Ag EV Array are given in ►Table 2. No significant difference between fasting and postprandial    and 2 (Q2), respectively. Increased particle numbers lead to a higher coincident rate. Flow rates were, therefore, adjusted in postprandial samples with many events to maintain event rates below the maximal flow rate limit (<20,000 events/second) and thereby reduce coincident detection. Note: A lower flow rate reduces both the numbers of TruCount beads (the framed dots in quadrant 2) and PSþ events acquired in the postprandial scatter plot (B).

Lipid Level Determination
Fasting and postprandial total cholesterol and TG concentrations for each subject are given in ►Fig. 5D, E. TG concentrations were significantly higher in the postprandial as compared with the fasting samples (p < 0.001) with a mean increase of 0.2 mmol/L or 22%. No significant change in total cholesterol concentration was observed. A positive correlation was found between TG concentrations and total particle levels as measured by FC in postprandial (►Fig. 6B) but not fasting samples.

Discussion
In this study, we aimed to evaluate the effect of a shift from fasting to postprandial state on blood plasma levels of particles measured with FC and EVs measured with EV Array, especially of TF-bearing subtypes. Using FC, we observed significantly increased postprandial plasma levels of total particles, TF-bearing particles (PSÀ as well as PSþ particles), total PSþ particles, and subtypes exposing antigens indicating platelet and endothelial origin in the postprandial as compared with the fasting state. In vitro addition of lipoproteins to PFP also induced increased levels of all types of particles measured by FC in this study. No changes in total levels of EVs measured with EV Array, PPL-CT, or circulating TF Ag were observed postprandially and in no samples TF activity was above the limit of detection. Therefore, the seeming postprandial increases in MV concentrations (relying on the assumption that PSþ events represent MVs) measured with FC may, at least partially, be artifacts.
The postprandial increase in total particle concentration was expected and the magnitude of the increase agreed well with observations on the same study population counting particles by use of TRPS showing a median postprandial increase in particle concentrations by close to 1,400%. 12 Intestinal lipid absorption is followed by release of chylomicrons from enterocytes. 30 We expect these TG-rich lipoproteins, which have diameters between 75 and 1,200 nm, 30 to represent the main part of the increased number of particles measured after food intake. The positive correlation between the concentration of TG and total number of particles in the postprandial state supports this. The concentration of VLDLs, another type of TG-rich lipoprotein particles, is also markedly increased postprandially 31,32 but we would not expect single VLDL particles to be detected by FC as their diameter is appreciably below 100 nm except from a smaller fraction, named large VLDLs, with diameters up to 200 nm. 14 Interestingly, the amount of PSþ particles also increased considerably postprandially, but it is questionable to which degree the significant increases in PSþ particles and the measured subtypes really reflect changes in the MV population.
The nomenclature in the EV field is not consistent. 8,13,33 A distinction between MVs and exosomes, relying on their exposure of PS and specific tetraspanins, respectively, has been suggested but it is not possible in practice to differentiate between these two types of EVs. 18 Exosomes are actually not necessarily PS-free. 11 Although Lacroix et al advocated that identifying PS on particles using lactadherin seems preferable to distinguish true MV events from other particles in the same size range, 34 it has also been reported that a substantial part of circulating MVs do not expose measurable amounts of PS. 18,24,35 Moreover, Shi et al Table 2 Biological variation in procoagulant phospholipiddependent clotting time (PPL-CT) and tissue factor antigen (TF Ag) levels in platelet-free plasma (PFP)

Study
Mean AE SD  mentioned the possibility of lactadherin binding to lipoproteins that may themselves display PS. 20 With regard to exosomes, while tetraspanins considered to be exosome markers, i.e., CD9, CD63, and CD81, 36 are particularly abundant in the membranes of exosomes, they are also found in MVs. 22 Hence, we have designated particles detected by EV Array "small EVs" and particles detected with FC "PSþ particles" or "PSÀ particles." The combination of lactadherin for labeling particles exposing PS and an antibody for labeling cell type markers should theoretically help prevent lipoproteins from being mistaken for MVs. However, the fact that all the subtypes of PSþ particles that we measured in this study were more abundant in the postprandial state may indicate that FC counts of PSþ particles including any subgroup existing in PFP are in general increased as a result of interference from lipoproteins. Indeed, when we spiked plasma samples with VLDL or LDL, the results clearly indicated that the in vitro admixture of either type of lipoprotein prompted elevations of all examined subtypes of particles, although the magnitude of the rise varied substantially among the different subtypes, with the increase in CD41þ particles being least pronounced. Sódar et al recently demonstrated that LDLs do in fact adhere to MVs as well as to exosomes when mixed with them in vitro. 13 It could be speculated that LDLs adhering to vesicles smaller than those normally detectable by FC cause formation of fluorescently labeled particles large enough for light scatter-based detection. However, even without the occurrence of adhesion, lipoproteins and EVs simultaneously passing the optical system of the flow cytometer may conceivably be mistaken for a fluorescent particle within the size range detectable by FC, thus causing a variant of the phenomenon termed swarm detection. 37 Such mechanisms could explain the apparent increase in the antibody-labeled subtypes of PSþ particles postprandially. The TFþ PSÀ particles observed in our study may represent TFþ MVs that did not expose measurable amounts of PS or possibly TFþ exosomes attached to or "swarm detected" along with lipoproteins.
Our spiking study has some limitations. We did not add chylomicrons, which appear in high amounts in the true postprandial state. The sizes of LDLs and VLDLs in the added suspensions probably differ from their in vivo counterparts since previous studies have shown that the isolates contain particles with larger diameters than normally described for LDLs and VLDLs, probably due to aggregates of the lipoprotein particles. 38 FC of the samples spiked with VLDL or LDL was performed later and on other plasmas and other batches of antibodies, and consequently the numbers of particles were somewhat different from the first experiments. Nevertheless, the relative increases (experiments performed at the same time with the same type of antibodies) were clearly significant (►Table 1), and the results indicate that lipoproteins do interfere with the FC measurements. Also, we considered the possibility that the commercially purchased LDL and VLDL isolates may themselves contain some PSþ EV residuals, since co-purification of EVs and lipoproteins is a known phenomenon. 13 However, addition of LDLs and VLDLs in a solution containing PBS instead of PFP did not produce antibody-labeled events, indicating that this should not be a major source of error.
Although an artifactual effect on FC results of lipoproteins appearing postprandially is likely and although plasma content of smaller EVs measurable by EV Array did not change upon food intake, it is possible that part of the postprandial increase in antibody-labeled PSþ particles may actually be due to cellular MV release. Notably, among the PSþ particles, the CD62Eþ fraction increased markedly more than the others, indicating a distinct effect on this subtype, which may reflect an increase in MVs from activated endothelial cells. Mechanistically, a rise in the number of CD62Eþ PSþ MVs is plausible as increased plasma levels of TG have been shown to cause endothelial activation 39,40 and MVs can be released in response to cell activation. 41 Ferreira et al reported that postprandial hypertriglyceridemia induced elevated levels of endothelium-derived MVs both in the early (1 hour) postprandial phase and 3 hours after food intake. That study also relied on FC for MV detection, 42 thus being subject to the same type of possible interference from lipoproteins as the present study. Platelets can also be activated in the postprandial state. 43,44 Interestingly, Michelsen et al described elevated postprandial concentrations of platelet-derived MVs after a fat-rich meal using a non-FC-based immunometric method reportedly displaying a linear relation to FC results on plateletderived MVs. 45 Our FC results on subtypes of CD41 þ PSþ particles are consistent with these findings. Further elucidation of the potential interactions between MVs and lipoproteins and the relative contribution of either to the increase in cell type marker-and PS-bearing events measured by FC may be possible by addition of lipoprotein markers (e.g., antiapolipoprotein B) or by detergent-based lysis of EVs as applied by Sódar et al. 13 The existence of circulating TF-bearing EVs in blood from healthy subjects has been described, 6,46 although it has also been stated that cells do not shed TF-bearing EVs under nonpathological conditions. 47 Since our various measurements of TF did not correlate very well, we clearly measured different populations of TF. However, consistency between fasting and postprandial EV Array results indicated that small TFþ EVs are not significantly influenced by food ingestion, whereas the reason for the postprandial increase of TFþ particles (PSþ and PSÀ) remains debatable. It would have been advantageous to measure TF activity but the results demonstrating levels below the limit of detection in all plasma samples are in accordance with the findings by others on healthy individuals. 48,49 Measurement of TF Ag may not solely measure active TF but the unaltered postprandial level indicates no major change of TF concentration postprandially. This is in accordance with the suggestion that the increase in TFþ particles measured with FC primarily reflects a higher number of small EVs adhering to or "swarming" alongside lipoproteins in formations that are large enough to be detected by FC rather than de novo release of TFþ MVs.
The finding that TF Ag levels are unaltered by food intake contrasts observations described by Motton et al. 48 However, postprandial blood samples in that study were drawn 3.5 and 6 hours following a standardized 40% fat meal that induced an increase in mean TG concentration by over 100% (3.5 hours postprandial), whereas the mean postprandial increase in TG concentration in our study confined itself to 22%. The postprandial samples in our study were drawn at a feasible time for blood sampling of patients in the morning after a normal breakfast. Sódar et al showed that the concentration of particles was almost maximal after 90 minutes. 13 Regarding PPL levels, our data demonstrated that although the amount of PSþ particles detectable by FC increased significantly in the postprandial state, this was not accompanied by a change in the coagulation state that can be monitored by PPL-CT. Tushuizen et al found that the total amount of PS in MV fractions was unaffected by food intake in healthy subjects, although PSþ MV concentration increased. 16 This corresponds well with the apprehension that a considerable part of the apparent postprandial increase in the concentration of PSþ particles may be an artifact resulting from increased lipoprotein interference.

Conclusion
Food ingestion resulted in elevated plasma levels of PSþ particles, including subgroups exposing TF and other antigen markers measured by FC. While the observed postprandial increase may to some extent reflect elevated MV levels, the finding that in vitro addition of lipoproteins to fasting plasma had a similar effect suggests that it may at least partly be due to postprandially appearing lipoproteins. Thus, PSþ events that are also positive for TF or cell-specific antigen markers should not uncritically be interpreted as MVs, when performing FCbased investigations on MVs in blood.

Funding
The study was supported by research grants from the Danish Council for Strategic Research.