Reliable measurements of extracellular vesicles by clinical flow cytometry

Abstract Extracellular vesicles (EVs) are cell‐derived particles with a phospholipid membrane present in all body fluids. Because EV properties change in health and disease, EVs have excellent potential to become biomarkers for diagnosis, prognosis, or monitoring of disease. The only technique capable of detecting, sizing, and phenotyping a million of EVs within minutes is (clinical) flow cytometry. A flow cytometer measures light scattering and fluorescence signals of single EVs. Although these signals contain valuable information about the presence and composition of EVs, the signals are expressed in arbitrary units, which make the comparison of measurement results impossible between instruments and laboratories. Additionally, unintended and undocumented variations in the source, preparation, and analysis of the sample lead to orders of magnitude variations in the measured EV concentrations. Here, we will explain the basics, challenges, and common misconceptions of EV flow cytometry. In addition, we provide an overview of recent standardization initiatives, which are a prerequisite for comparison of clinical data and thus for clinical biomarker exploration of EVs.


| FLOW C Y TOME TRY
Flow cytometers (FCMs) are commonly available in clinical laboratories and can provide information about the concentration, phenotype via fluorescent labeling, refractive index (RI), and size of single EVs. 8 To understand pitfalls and challenges that are specific to EV flow cytometry, we will first explain the basics of flow cytometry regarding fluidics, light scattering, and fluorescence, followed by a description of the standardization efforts ongoing in the EV field, and the prospects of EV research. This manuscript focuses on clinical flow cytometers and therefore does not cover cell sorters, imaging FCMs, or nanoparticle FCMs, which we define as FCMs specialized in the detection of EVs smaller than 100 nm. 9 In contrast to clinical flow cytometers, nanoparticle FCMs (a) have fluidics and optics unsuitable for cell characterization, (b) accomplish high-sensitivity at the expense of throughput and therefore have a maximum count rate <1000 events per second, and (c) are (still) rarely available.

| Fluidics
An FCM hydrodynamically focusses an EV sample into a stream of sheath fluid, as shown in Figure 1. Because the sheath fluid has a higher flow rate than the sample, the sample flow becomes relatively narrow, thereby centering the EVs in the flow cell. Increasing the sample flow rate increases the diameter of the sample flow and the number of EVs detected per unit time. [10][11][12] Sample flow rates used to detect cells typically range between 10-50 µL/min, 10 which results in a sample flow diameter in the size range of cells. Hence, single cells will pass through the center of the focused laser beam. However, because most EVs are smaller than 200 nm, EVs are considerably smaller than the sample flow diameter used for cell detection, which can affect EV measurements in two ways. First, EVs may pass the focused laser beam at different locations, thereby leading to variation in the detected signals. Second, a relatively large sample flow diameter increases the probability that multiple EVs and other particles are simultaneously illuminated, which may lead to an unwanted effect called swarm detection (see section Common misinterpretations). To prevent swarm detection and to ensure that EVs are centered in the laser beam, EVs are often detected with the lowest flow rates present on FCMs, typically in the range of 3-12 µL/min. 10 The choice of sample dilution buffer and sheath fluid depends on the application. 10 To prevent background scattering caused by RI differences between the interface of the sample and sheath fluid, the sample dilution buffer and sheath fluid are ideally the same. EVs need to be diluted in an isotonic buffer solution, such as phosphate-buffered saline (PBS), to prevent damage to the EVs by osmosis. To match the RI, PBS therefore seems an obvious choice for the sheath buffer. However, PBS may increase the formation of salt crystals in the fluidics, which causes artefacts such as additional background noise, clogging, or loss of laminar flow, making especially dim EV signals unreliable. 10,11 Due to these artifacts, several new flow cytometers dedicated to EV detection use deionized water as a sheath fluid. 13,14 The optimal sample dilution buffer and sheath fluid are subject for ongoing research.

| Light scattering
When an EV is illuminated by the laser beam, the EV scatters light into all directions. A typical FCM detects the scattered light in the forward scattering (FSC) and sideward scattering (SSC) directions, as shown in Figure 1. 12 In the FSC direction, the laser beam is blocked by an obscuration bar to prevent laser light impinging upon the detector. Although scattered light contains valuable information about the EVs, such as the diameter, getting access to this information is not straightforward for two reasons. First, the light scattering intensity is measured in arbitrary units, which differ between FCMs, making data comparison difficult. Secondly, the signal is difficult to interpret, because the dependence of light scattering on EV properties is complicated.

| Parameters affecting light scattering detection
The detected light scattering signal depends on both the EVs and the (clinical) FCM. Properties of EVs affecting light scattering are the diameter, RI, and shape. The RI depends on the EV composition and is a property that is often overlooked. To emphasize the importance of the RI, Figure 2A shows the measured (A60-Micro, Apogee Flow Systems) scattering intensities of similar-sized polystyrene (PS) F I G U R E 1 A flow cytometer utilizes a sheath flow to hydrodynamically focus the sample flow. Extracellular vesicles (EVs) in the sample pass through a focused laser beam and scatter light in all directions. A fraction of the scattered light is collected by lenses, which are typically placed in the forward scattering (FSC) and side scattering (SSC) direction. The fluorescence of EVs labeled with fluorophores is collected by the SSC lens and spectrally filtered to differentiate fluorescent signals from SSC  Figure 2B shows SSC vs diameter for the aforementioned beads and EVs. Generally, light scattering increases with increasing diameter and RI. [15][16][17] Properties of the FCM that affect the detected light scattering signal involve the power and wavelength of the illumination and the collection angle of the lens. The scatter to diameter relation in Figure 2B therefore differs between FCMs. Because most EVs are spherical, light scattering detected by an FCM can be well-described by Mie theory. 5,17,18 Mie theory takes into account all the aforementioned parameters of both the FCM and EVs and is valid for all EV sizes.

| From arbitrary units to comparable units
To make results insightful and comparable, the arbitrary units of light scattering should be converted to SI units. Such a calibration is generally performed using well-characterized reference beads, such that the data are physically accurately expressed and can be converted to SI units. Figure 2B shows how Mie theory can be used to relate the measured scatter signals to the diameter of EVs, thereby assuming an effective RI of EVs based on the measured RI of lipid bilayers (shell) and the cytosol (core) of cells. 17 To implement sizing of EVs by flow cytometry in the workflow, different software are available to apply Mie theory, such as FCM pass and Rosetta Calibration. 19,20 Daily runs of beads without applying Mie theory, as is common practice in the field, do not convert arbitrary units to SI units and are therefore not a calibration. Nevertheless, daily runs of beads are useful to set up the FCM and to check whether the instrument measures consistent over time and are thus a quality control.

| Common misinterpretations
Despite the availability of free and commercial software to size EVs with light scattering, 17 the most common pitfall is the use of PS beads to define gates for EV detection. Figure 2B shows, for example, that SSC of a 200 nm PS bead corresponds to EVs larger than 1000 nm.
Despite their lower RI, also silica beads scatter considerably more light than similar-sized EVs. Moreover, the scatter to diameter relation depends on the collection angles, which differ between FCMs.
Thus, neither PS nor silica beads can be used to gate EVs because (a) F I G U R E 2 A, The scattering intensities of hollow organosilica beads (HOBs) with a diameter of 180 nm, silica beads with a diameter of 183 nm and polystyrene beads with a diameter of 203 nm show that beads with higher refractive indices have a higher light scattering intensity. Polystyrene and silica beads have a refractive index of 1.633 and 1.475, respectively. Because the core of HOBs contains water, HOBs scatter light less efficiently than both polystyrene and silica. B, Side scattering intensity versus diameter measured (symbols) by flow cytometry (A60-Micro, Apogee Flow Systems) and calculated (lines) with Mie theory for polystyrene beads (squares), silica beads (circles), HOBs (triangles), and EVs (shaded area between dotted lines). Please note that the vertical scale is logarithmic. Mie theory calculations are performed with Rosetta Calibration software (Exometry, The Netherlands) assuming a refractive index of 1.633 for polystyrene beads and 1.475 for silica beads. HOBs are modeled as core-shell particles, with a core refractive index of 1.343, a shell refractive index of 1.475, and a shell thickness of 10 nm. EVs were also modeled as core-shell particles, but with a core refractive index ranging from 1.343 to 1.36, a shell refractive index of 1.46, and a shell thickness of 5 nm. The side scattering intensity increases with increasing diameter, but also with increasing refractive index. Hence, for this flow cytometer, polystyrene beads of 200 nm scatter light more efficiently than 1000 nm EVs  Figure 2B, only HOBs mimic the physical light scattering properties of EVs. 16 Hence, only HOBs are suitable to set true EV size gates without correction with Mie theory. Table 1 summarizes whether the aforementioned reference beads and methods can be used as a calibration or quality control.
Swarm detection is a special case of coincidence detection, where instead of two or a few particles, multiple (tenths to hundreds) particles at or below the detection limit are simultaneously and continuously present in the laser beam of the flow cytometer and erroneously measured as single counts. 15 Swarm detection can be prevented by diluting the sample prior to analysis, or by lowering the sample flow rate, which will result in higher sensitivity. 15 Measured light scattering signals may originate from other sources than EVs, such as optical and electrical noise and particles in the sheath fluid or buffer. To confirm that background noise sources are negligible, it is important to perform a buffer-only control in every experiment. The buffer-only control involves measuring the buffer, which is supposed to be clean and therefore is expected to result in low counts compared to the EV sample.

| Fluorescence intensity measurements
Besides light scattering, FCMs can detect EVs labeled with fluorophores, which are used to establish the cellular origin of EVs, or to study the presence of proteins or lipids. Fluorescence occurs after a fluorophore is illuminated by laser light. If the laser wavelength matches the excitation wavelengths of the fluorophore, the fluorophore will go to an excited energy state. After relaxation to its original energy state, the fluorophore emits light at a wavelength longer than the excitation wavelength. More fluorophores bound to an EV yield a higher fluorescence intensity signal. 1988 Gate based on light scattering of activated platelets EVs are smaller than platelets; thus, events with lower scattering than platelets are labeled "EVs" Note: For each approach, the table provides an example, the suitability to function as quality control and calibration, the suitability to determine the EV size, and a literature reference.
part of the labeled EVs will be below the detection limit. Choosing a brighter fluorophore therefore would result in the detection of more EVs. 22 As a rule of thumb, APC and PE are 5-to 10-fold brighter than FITC. 12 When combinations of fluorophores are simultaneously used to label EVs, these fluorophores should not overlap in their emission spectrum. Spectral overlap causes multiple fluorophores to be detected by the same detector, resulting in misidentification of EVs. 11,12 Although for cell analysis methods are available to compensate for spectral overlap, more research is required to develop and validate spectral compensation methods for the analyses of EVs. 10,12 EVs are typically labeled with fluorophores conjugated to antibodies, which bind to specific antigens exposed by the EVs. As both fluorophores and antibodies consist of mostly proteins, which have a higher RI than the medium, the scatter signal from a stained EVs will likely increase, but we are not aware of any experimental confirmation. The most widely studied body fluid in EV research is plasma. 8 Plasma contains EVs originating from blood cells, including platelets, erythrocytes, and leukocytes, and the endothelium. The cellular origin of single EVs can be determined by labeling with an antibody that specifically binds to an antigen that is exposed exclusively on EVs originating from a particular cell type. For example, platelets expose the fibrinogen receptor, the integrin α IIb β 3 . Antibodies directed against the α IIb and β 3 subunit of this complex are categorized as cluster of differentiation (CD) 41 and CD61, respectively. 23,24 Each numbered CD includes all the (clones of an) antibodies that can be bound to a particular surface protein at the surface of EVs, thereby revealing the cellular origin of the EV. Thus, EVs in plasma that bind CD41 and/or CD61 are considered to be "platelet-derived EVs." Similarly, antibodies directed against glycophorin A, categorized as CD235a, are considered to identify erythrocyte EVs. 5,6,25 In addition to antibodies, there are generic dyes of which most stain the phospholipid membrane or primary amines found on the surface of proteins. Thus far, none of these generic dyes label all and exclusively EVs. From five investigated generic dyes, the protein lactadherin, which binds to phosphatidylserine in the membrane of EVs, resulted in the highest sensitivity and specificity to stain EVs >200 nm in diameter. 26,27 Another protein which binds to phosphatidylserine is annexin V, but annexin V needs calcium ions to bind to phosphatidylserine. Because calcium ions are also a cofactor for coagulation, annexin V is often not recommended to label plasma EVs. 26,28

| From arbitrary units to comparable units of fluorescence
Similar to the scattering intensity, fluorescence intensity is reported in arbitrary units, which are difficult to compare and interpret. EVs, reagents can be added to buffer and measured as a control. 33 Based on the buffer with reagents control, we identified aggregates in the fluorophores added to the sample measured in Figure 3.
Furthermore, the specificity of the used antibodies has to be confirmed using isotype controls. Isotype controls will determine if antibodies are binding FC receptors on the membrane. 33 In some assays, it is required to isolate labeled EVs. For example, generic dyes may form micelles themselves, which may be erroneously detected as EVs and therefore must be removed prior to analysis. Because frequently used isolation methods may affect the concentration of labeled EVs, it is good practice to run the buffer-only, as well as the buffer with reagents by the isolation method, to confirm the absence of positive events resembling EVs. This control was recently introduced as "procedural control". 33 An additional pitfall is the limited number of antigens exposed on EVs. The small diameters of EVs result in low numbers of antigens present on their surface compared to cells. It is important to realize that in most assays, a fraction of labeled EVs remains below the detection limit of the flow cytometer. 10,34 Therefore, the brightness of the fluorophores is important, and bright dyes, such as APC and PE, are preferred.
Also for fluorescence, swarm detection may lead to misinterpretation. When multiple small particles, including aggregated or unbound fluorophores, are simultaneously illuminated, their total (auto)fluorescence might exceed the trigger threshold, thereby contributing to the overall measured fluorescence signal. 35 Solutions to avoid swarm detection are sample dilution and lowering the sample flow rate. 10,15 Serial dilution is the control to confirm absence of swarm detection (see also Light scattering). 33 To reach the eventual goal of reliable measurements of the same properties of EVs between FCMs and laboratories, Table 2 lists the most important pitfalls to evade and techniques to apply.

| Selecting a flow cytometer for EV research
To evaluate the suitability of an FCM to study EVs, we recommend calibrating the scatter and relevant fluorescence detectors. By expressing the sensitivities of detectors in comparable units, sensitivities can be compared to FCMs in the field (eg, Figures 2B and 3B) and to the requirements for future research. Please note that FCM manufacturers typically specify the scatter sensitivities in terms of the smallest detectable diameter of polystyrene beads, which is misleading, because polystyrene beads scatter light substantially more efficient than EVs, as shown in Figure 2B. Furthermore, the flow rate should be stable and measurable. Only after measuring an EV sample on a calibrated FCM, one can decide if the performance of the FCM satisfies the preferred requirements for EV flow cytometry. Although it is beyond the scope of this manuscript to discuss how to set up an FCM, we recommend reading MIFlowCyt-EV to get an impression of the relevant parameters and assay controls. 33 The sensitivity of FCMs does not only depend on the brand and type, but also on the state of maintenance and instrument setup.

| G E T TING E X TR ACELLUL AR VE S ICLE S INTO THE CLINI C
Thus far, most if not all studies in which concentration of cell type-

| Standardization
There are three sources of variation that affect the outcome of EV measurements. The first source is the sample itself, for example, a body fluid such as blood and urine. The body fluid composition is donor-dependent and depends for example on age, diurnal rhythm, fasting state, and medication. 33,36,37 The second source of variation is the pre-analytical phase, in which the samples are collected, handled, and stored until analysis of EVs. The third source of variation is the analytics, the hardware, and software used to detect EVs and analyze the obtained results. 38 Because the analytics have been incompletely understood for a long time, monitoring and optimizing the pre-analytics was difficult. At present, this is rapidly changing, due to international standardization initiatives.
Standardization of the sample itself is possible only to a limited extent. As mentioned before, the composition of a body fluid, including endogenous EVs, will differ per person. 33  The PS beads were used to set so-called "microparticle gates" by FSC with the aim to select platelet-derived EVs. These studies had two shortcomings. Firstly, the difference in RI between the PS beads and EVs was not taken into account. Secondly, the differences in light collection angles of various FCMs were not taken into account. Day-to-day variation Apply daily cleaning and regular maintenance and monitor day-to-day variation daily using appropriate quality controls for scatter, fluorescence and flow rate.
Dim fluorescence signals Select the brightest fluorophores and incubate longer during staining.
Event signals originate from noise or other particles than the envisioned EVs Apply appropriate controls (buffer-only control, buffer with reagents control, unstained control, isotype control, FMO and single-stained control, procedural control, serial dilution, detergent treatment) to ensure that events are actual EVs.
FCMs cannot measure all sizes of EVs Currently, no FCM exist with a dynamic range suitable for measuring all EV sizes. Calibrate scatter and fluorescence of the FCM to know the measurement ranges of your FCM.
Incomparable data between FCMs Calibrate scatter, fluorescence and flow rate of the FCM daily to generate data in comparable standardized units. Wrong estimation of EV size Take the optical configuration of the FCM as well as the difference in RI between EVs and the reference beads into account using Mie theory.

Pre-analytical variation
Abbreviations: EVs, extracellular vesicles; FCM, flow cytometer; FMO, fluorescence minus one; RI, refractive index. well as differences in the optical configuration of the FCMs were taken into acount. 20  sizes, but both sizes fall in the RI range of EVs. 16 The lack of suitable reference materials and procedures for standardizing EV flow cytometry measurements has led to the foundation of the METVES II project (metves.eu), which started in June 2019 (see section Outlook).

| Applications
The infrastructure that is being developed will allow the measurements of the concentrations of cell type-specific EVs in all body fluids. For example, in plasma also EVs originating from a tumour or the placenta may be present. 1

| Outlook
The sights. If EV researchers will start using the MIFlowCyt-EV framework and calibrations now, we are confident that, upon the arrival of faster and more sensitive FCMs together with improved reference materials and procedures, both comparable FCM measurements and EV-based biomarkers will become reality.

ACK N OWLED G M ENTS
This work is part of the project "METVES II" and received funding from the EMPIR program co-financed by the Participating States and from the European Union's Horizon 2020 research and innovation program. E. van der Pol acknowledges funding from the Netherlands

Organisation for Scientific Research-Domain Applied and Engineering
Sciences (NWO-TTW), research program VENI 15924. We thank Aleksandra Gąsecka for providing the data underlying Figure 3.

CO N FLI C T O F I NTE R E S T
E. van der Pol is co-founder and shareholder of Exometry BV.