Fig 2: Optimised acquisition by Imaging flow cytometry and florescence membrane labelling can distinguish large extracellular vesicles from background and enable multiplex labelling. (A) Hierarchical gating can confidently refine the EV population for characterisation. Samples were acquired for 5 min using the ISX INSPIRE software and all events visualised using bivariate plots for fluorescence intensity in channel 03 (CMO) and channel 06 (side scatter). Initial gates for speed beads and CMO- events were set using the Acquisition buffer only and unstained EV samples included in each run (Gate i). These gates were applied to the labelled samples. Saturated events were excluded from the analysis by gating on the histogram plot: Raw Max Pixel for Channel 06 (Gate ii) thus removing the speed beads and very high side scatter events. An initial CMO + gate was placed (Gate iii). A lower fluorescence threshold of MESF > 50 in channel 3 (CMO) was set in each experiment using the regression analysis of Quantibrite PE Beads as described. The CMO + gate was anchored using co-ordinates for the equivalent fluorescence intensity and a further CMO + MESF > 50 gate applied to the histogram (Gate iv). Finally, a refined gate was used to eliminate outliers (Gate v). The analysis template was set for each experiment and applied to all samples. (B) Percentage positive CMO+ events (MESF > 50). The percentage of CMO+ events (MESF > 50) is represented by an orange dot for 10 replicate experiments. Black bar represents mean 18.73% (± 1.2 SEM). (C) A compensation matrix was applied to all samples. The compensation wizard was used to create a matrix which was applied to all samples. (D) Single, double, and triple labelling protocols were used for compensation and accurate gating. Dual labelled EVs provided Fluorescence minus one (FMO) controls for (i) AF647; EVs labelled with CMO and CD63 BV421) or (ii) BV421; (EVs labelled with CMO and LAMP1 AF647) and used to set positive gates. (E) UW228-2 cells were used to optimise multiplex labelling. Cell mask orange (CMO) provides a general membrane label (i) and could be used to co-label with EV markers CD63 BV421 and CD9 AF647 (ii) or CD63 BV421 and LAMP1 AF647 (iii). Representative gallery images from IDEAS software are shown.

Fig 3: Influenza mediates C3 release from platelets through TLR7. a C3 in plasma from healthy donors (n = 14) and influenza-infected patients (n = 18). The graphs represent the average ± SD; significance was assessed by Mann–Whitney U test, *p < 0.0001. b C3 release from human platelets that express TLR7 (assessed by qPCR) had been isolated from healthy donors and mixed with influenza strain WSN/33 at a proportion of 1 pfu to 100 platelets (n = 4, 3F, 1M); p < 0.0001, F = 27.49, df = 3. c C3 release from human platelets that do not express TLR7 (assessed by qPCR) treated as in (b). Graph is representative of n = 3 different blood draws; p = 0.2813, F = 1.833, df = 3. d C3 release from human platelets from the same donors as in (b) treated with the TLR7 agonist loxoribine 1 mM in the presence or absence of TLR7 inhibitor IRS661; p = 0.0086, F = 7.989, df = 3. b–d Data in graphs are represented as average ± SD; significance was assessed using ANOVA followed by Bonferroni multiple comparison test and star symbol (*) indicates p < 0.05. Source data for (a–d) are provided as a Source Data file. e Confocal images of permeabilized isolated platelets from influenza-infected patient stained for flu-FITC; TLR7-APC; lysosomal marker CD63-BV421. f Confocal images of isolated healthy human platelets incubated with WSN/33 for 30 min (1 pfu to 10 platelets) and stained as in (e). e, f Representative images of platelets from n = 4 (2M, 2F) different donors are shown

Fig 4: The ability of COVID19 plasma to activate platelets is increased for patients with evidence of organ damage, high circulating levels of ferritin, and at timepoints preceding a thrombotic event. (A) Plasma from COVID-19 inpatient (n=63), COVID-19 convalescent (n=20) and healthy donors (n=9) was incubated with platelets isolated from healthy volunteers. The gMFI levels of CD63, CD62P, CD32 and C3aR are shown in these three groups. Kruskal-Wallis non-parametric testing was used to compare all three groups and pair-wise comparisons were also performed; p-values are depicted. (B) Spearman correlation of the gMFI surface levels of the markers used in (A) and selected clinical parameters of the COVID-19 inpatient group (n=63). ? correlation coefficient shown on the key (bottom). Asterisks *, ** and *** denote p values less than 0.05, 0.01 and 0.001 respectively. Highlighted squares denote FDR values less than 0.05. (C) CD63 gMFI surface levels of control platelets induced by plasma derived from COVID-19 patients with thrombosis drawn at different timepoints relative to the thrombotic event. (D) CD63 gMFI surface levels of control platelets induced by plasma derived from COVID-19 patients prior to thrombosis and after thrombosis, COVID-19 patients without thrombosis and COVID-19 patients in convalescence. Kruskal-Wallis non-parametric testing was used to compare all four groups and pair-wise comparisons were also performed using Wilcoxon testing; p-values are depicted. (E) Representative scatter plots for creatinine vs CD63, platelet count vs CD62P and APACHE III score vs CD63. Samples are colored based on the absence (grey color) or presence of thrombosis relative to the time of draw (red before thrombosis and blue after thrombosis). The Spearman correlation co-efficient and the corresponding p-value are depicted.

Fig 5: Isolated large EVs can be triple-labelled and individual events can be scrutinised post-acquisition. (A) The proportion of CMO+ events expressing a single EV marker varied across replicates. The percentage positive events within the CMO+ MESF > 50 gate which expressed either no EV marker (CMO only), CD63, CD9, or LAMP1 varied across replicates. Black bar represents the mean across 10 (CMO or CD63) or 5 (CD9 or LAMP1) replicates. CMO only 63.3% (± 4.6 SEM), CD63 only 19% (± 2.2 SEM), CD9 (± 0.5 SEM) only 10.3% and LAMP1 only 47.6% (± 3.8 SEM). (B) UW228-2 cell derived EVs differentially express EV markers CD63 and CD9, or both. EVs were labelled with CMO, CD63 BV421 and CD9 AF647. CMO + events (MESF > 50) were analysed for co-expression with CD63, CD9 or both. Quadrant gating of bivariate plots for fluorescence intensity in channel 07 (BV421) and channel 11 (AF647) demonstrated single labelled (CMO+ only), dual and triple labelled EV populations (i). Percentage positive in each quadrant is shown. Representative experiment. Gallery images display examples of individual events from each of the 4 quadrants (ii). Bright Field (BF) and side scatter (SSC) channels are shown alongside CMO, CD63 BV421 and CD9 AF647. Of the CMO+ MESF >50: 75.1% (± 3.0 SEM) were CMO + only; 13.3% (± 2.4 SEM) expressed CD63 BV421 and 3.5% (± 0.8 SEM) expressed CD9 AF647. 7.1% (± 1.1 SEM) of the CMO+ events expressed both CD63 and CD9. (C) UW228-2 cell derived EVs differentially express EV markers CD63 and LAMP1, or both. EVs were labelled with CMO, CD63 BV421 and LAMP1 AF647. CMO+ events were analysed for co-expression with CD63, LAMP1 or both. Quadrant gating of bivariate plots for fluorescence intensity in channel 07 (BV421) and channel 11 (AF647) demonstrated single labelled (CMO+ only), dual and triple labelled EV populations (i). Percentage positive in each quadrant is shown. Representative experiment. Gallery images display examples of individual events from each of the 4 quadrants (ii). Bright Field (BF) and side scatter (SSC) channels are shown alongside CMO, CD63 BV421 and LAMP1 AF647. Of the CMO+ MESF >50: 54.1% (± 4.3 SEM) were CMO + only; 2.8% (± 0.4 SEM) expressed CD63 BV421 and 25.6% (± 2.7 SEM) expressed LAMP1 AF647 whilst 18.0% (± 2.8 SEM) of the CMO + events expressed both CD63 and LAMP1. (* p = 0.0159 or 0.0317; ** p = 0.0079). Mann Whitney U test.