Comparison of EV characterization by commercial high-sensitivity flow cytometers and a custom single-molecule flow cytometer

Abstract

High-sensitivity flow cytometers have been developed for multi-parameter characterization of single extracellular vesicles (EVs), but performance varies among instruments and calibration methods. Here we compare the characterization of identical (split) EV samples derived from human colorectal cancer (DiFi) cells by three high-sensitivity flow cytometers, two commercial instruments, CytoFLEX/CellStream, and a custom single-molecule flow cytometer (SMFC). DiFi EVs were stained with the membrane dye di-8-ANEPPS and with PE-conjugated anti-EGFR or anti-tetraspanin (CD9/CD63/CD81) antibodies for estimation of EV size and surface protein copy numbers. The limits of detection (LODs) for immunofluorescence and vesicle size based on calibration using cross-calibrated, hard-dyed beads were ∼10 PE/∼80 nm EV diameter for CytoFLEX and ∼10 PEs/∼67 nm for CellStream. For the SMFC, the LOD for immunofluorescence was 1 PE and ≤ 35 nm for size. The population of EVs detected by each system (di-8-ANEPPS⁺/PE⁺ particles) differed widely depending on the LOD of the system; for example, CellStream/CytoFLEX detected only 5.7% and 1.5% of the tetraspanin-labelled EVs detected by SMFC, respectively, and median EV diameter and antibody copy numbers were much larger for CellStream/CytoFLEX than for SMFC as measured and validated using super-resolution/single-molecule TIRF microscopy. To obtain a dataset representing a common EV population analysed by all three platforms, we filtered out SMFC and CellStream measurements for EVs below the CytoFLEX LODs as determined by bead calibration (10 PE/80 nm). The inter-platform agreement using this filtered dataset was significantly better than for the unfiltered dataset, but even better concordance between results was obtained by applying higher cutoffs (21 PE/120 nm) determined by threshold analysis using the SMFC data. The results demonstrate the impact of specifying LODs to define the EV population analysed on inter-instrument reproducibility in EV flow cytometry studies, and the utility of threshold analysis of SMFC data for providing semi-quantitative LOD values for other flow cytometers.

Keywords: CellStream; CytoFLEX; equivalent reference fluorophore calibration beads; extracellular vesicles; limit of detection; single‐molecule flow cytometry.

FIGURE 1

Calibration and determination of limits of detection for CytoFLEX and CellStream using Cellarcus vCal nanoRainbow beads. Cellarcus vCal nanoRainbow beads are a mixture of four populations of 500 nm diameter beads conjugated with incremental amounts of PE and di‐8‐ANEPPS dyes and assigned MESF and MESA values by Cellarcus. (a) Scatterplot of fluorescence intensities of the beads in the CytoFLEX vFRed and PE channels. The three bright bead populations (Beads 2–4) are used for calibration; the dim bead population (Bead 1) is used for determination of resolution by comparison with a blank. (b) Plot of the MFI measured in the CytoFLEX PE channel versus the MESF values assigned to each bead population. (c) Plot of MFI measured in the CytoFLEX vFRed channel versus MESA values assigned to each bead population. (d)–(f) Same as (a)–(c) but for CellStream instead of CytoFLEX. The LOD for PE copy number (number of PE dyes per EV) was obtained by determining the MFI (PE) value at two rSD above the MFI (PE) for a buffer‐only sample and using the calibration curves shown in Panels B and E (extrapolated to the y‐axis) to determine the corresponding MESF value. Similarly, the size LOD was obtained by determining the MFI in the vFRed channel at two rSD above the MFI for a buffer‐only sample and using the calibration curves shown in Panels C and F (extrapolated to the y‐axis) to determine the corresponding MESA value. LOD, limit of detection; MESA, molecules of equivalent surface area MESF, molecules of equivalent soluble fluorophore MFI, median fluorescence intensity; rSD, robust standard deviations.

FIGURE 2

Immunofluorescence LOD, copy number measurements, size calibration, and size LOD for the SMFC. (a) Limit of detection of PE. Top: Photon burst traces of a buffer solution and a 10 pM solution of PE using the PE channel (excitation at 561 nm, 585/30 nm emission filter). Bottom: distribution of burst intensities with lognormal fitting. ∼100% of the fitted distribution was above the detection threshold of SNR = 3 (blue dashed line), determined using the buffer‐only sample, indicating 100% single‐molecule detection efficiency. (b) EV data collection. Overlaid photon‐burst traces of EVs labelled with PE‐antibodies (red) and with di‐8‐ANEPPs (green). (c) Copy number measurements. PE‐channel fluorescence intensity distributions for EVs and for unbound PE dye‐antibody conjugates were deconvoluted to calculate antibody copy numbers. (d) EV size calibration and LOD. Top: Distribution of square‐root intensity of di‐8‐ANEPPS signals from EVs (green) and of EV size as measured by dSTORM (red). Bottom: Calibration curve. y‐axis values are based on STORM measurements. The SMFC was able to detect and size the smallest EVs observed by dSTORM (35 nm) (LOD ≤ 35 nm). dSTORM, direct stochastic optical reconstruction microscopy; LOD, limit of detection; SMFC, single‐molecule flow cytometer; SNR, signal‐to‐noise ratio.

FIGURE 3

EV concentration, size distribution and PE‐anti‐tetraspanin antibody copy number distribution for all tetraspanin‐labelled EVs detected by each platform. EVs were labelled with a mixture of PE‐conjugated anti‐CD9, ‐CD63 and ‐CD81 antibodies ('TS mix') and with the membrane dye di‐8‐ANEPPS. Different EV populations were detected and analysed by each system due to their different limits of detection. (a) Concentration of EVs detected by each system (mean ± SD; n = 5−7 replicate runs using the same processed sample on the same day). (b) EV size distribution measured by the three systems using membrane dye fluorescence. The size LODs for CellStream and CytoFLEX were determined using calibration beads; the LOD for SMFC was determined using calibration with EV measurements from super‐resolution microscopy (dSTORM). The SMFC could detect the smallest EVs observed by dSTORM (35 nm). (c) Total PE‐anti‐tetraspanin (CD9, CD63 and CD81) antibody copy number distribution for each system, based on signal from a mixture of PE‐conjugated antibodies against all three tetraspanins. The LODs for CellStream and CytoFLEX were based on calibration beads; the LOD of SMFC was determined by plotting the photon burst intensity distribution for single PE dyes conjugated to antibodies; all photon burst intensities were well above a signal‐to‐noise ratio of 3 (shown in Figure 2a), indicating that the SMFC detected even weakly emitting single PE fluorophores. The median TS antibody copy number determined by single‐molecule TIRF microscopy was 5.7. dSTORM, direct stochastic optical reconstruction microscopy; SMFC, single‐molecule flow cytometer; TIRF, total internal reflection fluorescence.

FIGURE 4

EV concentration, size distribution, and PE‐anti‐EGFR antibody copy number distribution for all EGFR‐labelled EVs detected by each platform. EVs were labelled with PE‐conjugated anti‐EGFR antibody and with the membrane dye di‐8‐ANEPPS. (a) Concentration of EVs detected by each system (mean ± SD; n = 5−7 replicate runs using the same processed sample on the same day). (b) EV size distribution measured by the three systems using membrane dye fluorescence. See Figures 1 and  2 caption and main text for methods used to determine the LODs for each system. (c) EGFR antibody copy number distribution for each system, based on signal from PE‐conjugated anti‐EGFR antibodies. The median EGFR antibody copy number determined by single‐molecule TIRF microscopy was 2.4. EGFR, epidermal growth factor receptor; LOD, limit of detection.

FIGURE 5

EV populations detected and analysed by the three platforms. Plot of the distribution of EV size and PE copy numbers for all EVs detected by all three systems (overlaid). Dashed lines indicate immunofluorescence and size LOD values for CellStream and CytoFLEX determined using bead calibration or threshold analysis. The common EV population analysed by all three platforms were EVs above the CytoFLEX LODs (black dashed line according to bead calibration [10 PEs, 80 nm], green dashed line according to threshold analysis [21 PEs, 120 nm]). LOD, limit of detection.

FIGURE 6

Filtering the dataset using CytoFLEX's LODs based on bead calibration (10 PEs, 80 nm) to obtain a common EV population analysed by all three platforms. 'TS' = EVs labelled with anti‐tetraspanin antibodies; 'EGFR' = EVs labelled with anti‐EGFR antibodies. Applying the 10 PE, 80 nm filter based on LOD values from bead calibration resulted in better agreement in mean EV concentrations (log scale) (top row), median EV sizes (middle row), and median PE copy numbers (bottom row) (which correspond to TS and EGFR antibody copy numbers for TS‐ and EGFR‐labelled EVs, respectively). However, there was still a discrepancy in the TS and EGFR antibody copy numbers between CytoFLEX and the other two platforms. CF, CytoFLEX; CS, CellStream; SMFC, single‐molecule flow cytometer.

FIGURE 7

Filtering the dataset using CytoFLEX's LODs based on threshold analysis (21 PEs, 120 nm) yielded better agreement in TS and EGFR antibody copy numbers. 'TS' = EVs labelled with anti‐tetraspanin antibodies; 'EGFR' = EVs labelled with anti‐EGFR antibodies. Applying the 21 PE, 120 nm filter based on LOD values from threshold analysis resulted in better agreement in median PE copy numbers than the 10 PE, 80 nm filter based on bead calibration. CF, CytoFLEX; CS, CellStream; SMFC, single‐molecule flow cytometer.