Study Cohort
We enrolled a total of 20 HC, 16 RRMS patients with active disease not receiving DMT, 14 RRMS patients with stable disease not receiving DMT, 17 RRMS patients with stable disease receiving natalizumab, and 14 RRMS patients with stable disease receiving ocrelizumab (Table 1). Active disease was determined by recent physician-confirmed relapse or detection of a CEL via MRI. Of the 16 RRMS patients with active disease, 14 had confirmed CEL (87.5%). Donors with significant comorbidities known to alter circulating EV concentrations including cancer, other autoimmune diseases, morbid obesity (body mass index of 40 or more), diabetes mellitus type 2, or recent cardiovascular or cerebrovascular events were excluded(68–73). Subject groups were comparable in age however the sex ratio varied. For the RRMS groups, EDSS scores were comparable between groups however disease duration (DD) was varied.
Table 1
Demographics and clinical characteristics of study participants
| HC | Active | Stable | NTZ | OCZ |
| n = 20 | n = 16 | n = 14 | n = 17 | n = 14 |
Age, mean, SD, y | 37(13)a | 36(10) | 43(12) | 39(10) | 41(11) |
% Female (n) | 75%(15)* | 69%(11) | 93%(13)* | 94%(16)* | 64%(9) |
Clinical State | | RRMS | RRMS | RRMS | RRMS |
Disease State | | Activeb | Stable | Stable | Stable |
DD, median (IQR), mo | | 22.5(2.5–56)# | 142.5(62.5–183) | 87(47.5-157.5) | 159.5(33-239.3)# |
EDSS, median, (IQR) | | 1.5(0.3–2.4) | 1.5(0.0-4.1) | 1(0.0–3.0) | 3.3(1.5–4.5) |
a one undetermined value, not included in correlation analysis |
b 14 out of 16 (87.5%) patients have confirmed contrast enhancing lesion |
*p < 0.002 HC vs. stable, HC vs. NTX, determined by Fishers exact test between pairs |
#p = 0.007 Active vs. OCZ, determined by ANOVA with Tukey's multiple comparisons test |
Abbreviations: DD = disease duration, EDSS = Expanded Disability Status Scale, IQR = interquartile range, m = month, SD = standard deviation, y = year |
Determination of flow cytometry size resolution
To determine the optimal collection procedure for isolation of EV via SEC for flow cytometry analysis, twelve, 0.5ml fractions were collected via SEC after addition of 1ml PFP (Fig. 1A). To detect submicron events via flow cytometry, we used 1um beads (polystyrene microspheres) diluted in PBS to select appropriate flow cytometry settings (Fig. 1B). A size gate was designated based on 1um beads (polystyrene microspheres) and excluded mechanical noise observed in PBS alone. Analysis revealed that SEC fractions seven, eight and nine contained the highest number of events within our designated size gate. Based on these results, fractions seven, eight and nine were pooled together for all subsequent EV analysis. TEM analysis of the pooled aliquots from HC and RRMS patients revealed morphologies indicative of EV (Fig. 1D). Although there was a wide range of EV diameters observed via TEM, there was no significant difference in the mean size between EV isolated from HC or RRMS patients (Fig. 1E). These results were confirmed via NTA (Fig. 1F). Finally, western blot analysis of pelleted EV revealed the presence of EV related proteins including flotillin, CD9, CD63, CD86 (Fig. 1G).
To determine the size resolution of our flow cytometry size gate, we completed a series of experiments using high speed centrifugation (Fig. 2A-D). It is well accepted that centrifugation between 10–20,000g causes sedimentation of larger EV (~ 150nm and larger) while centrifugation at 100,000g causes sedimentation of smaller EV (~ 30-150nm and larger)(74–76). To determine if our flow cytometry settings detected small or large EV, isolated EV were centrifuged at 18,000g for one hour to pellet large EV and 100,000g for one hour to pellet small EV (Fig. 2B and C). Supernatants were harvested and analyzed via flow cytometry and concentrations compared to no-spin controls. EV pellets were resuspended in the same volume of PBS as the starting material prior to analysis. Results revealed that the 18,000g and 100,000g supernatants contained 5% and 3% of events compared to no-spin controls, respectively (Fig. 2C), indicating that flow cytometry analysis is detecting EV larger than ~ 150nm. These results were confirmed when examining the number of events detected in the resuspended pellets. The 18,000g and 100,000g resuspended pellets contained 73% or 44% of events compared to no-spin controls, respectively. The reduction in events seen in the resuspended pellets is most likely due to EV aggregation, as observed via TEM (Fig. 2D). This aggregation was not observed in the no-spin controls. Taken together, this data indicates that our size gate is detecting EV larger than 150nm.
To further define our size resolution via flow cytometry, we performed an experiment using an array of differently sized fluorescent-green submicron beads (polystyrene microspheres) (Fig. 2E and F). 1um, 0.5 and 0.2um green-fluorescent beads were analyzed via flow cytometry using the settings described above. Unlabeled 1um beads were used as a negative control. By selecting for events with positive green fluorescence compared to unlabeled 1um beads, we determined that our size gate detects 99% of 1um green-fluorescent beads, 63% of 0.5um green-fluorescent beads, and 15% green-fluorescent beads (Fig. 2F). This data indicates that our size gate detects events ranging in size from 0.2 to 1um, with the majority of events ranging in size from 0.5 to 1um. This size gate was used for all subsequent analysis of SEC isolated EV.
Characterization of EV isolated from patient plasma
To further characterize the events detected within our size gate, SEC isolated EV were analyzed for EV specific markers CD9, CD63, and CD81 via flow cytometry (Fig. 3A-F). SEC isolated EV were stained with fluorescently conjugated anti-CD antibodies and appropriate isotype controls (Fig. 3A and B). When staining for CD9, a significantly higher percent of EV stained positive for CD9 versus the isotype control, 60.7% versus 1.5%, respectively. When staining for CD63, a significantly higher percent of EV stained positive for CD63 versus the isotype control, 9.7% versus 1.3%, respectively. In comparison, when staining for CD81, a significantly lower percent of EV stained positive for CD81 versus the isotype control, 0.1% versus 1.23%, respectively. Taken together this data indicates EV within our size gate are positive for CD9 and CD63, but not CD81. Analysis of EV isolated from three HC and three RRMS patients revealed that the majority of events within the size gate expressed CD9 (14.8–78.7%) while a smaller portion expressed CD63 (1.4–9.7%) (Fig. 3C). When comparing the percentage of size events positive for CD9, CD63, or CD81 between HC and RRMS patients, no difference was observed. When examining size events for dual expression of CD9 and CD63, the majority of events either expressed CD9 only (CD9 + CD63-, 14.0-71.7%) or none (CD9- CD63-, 21.2–84.9%) (Fig. 3D and E). A small percentage of events were positive for both CD9 and CD63 (CD9 + CD63+, 0.8–7.6%) and very few events were positive for CD63 only (CD9- CD63+, 0.2–0.4%). No significant difference was observed between HC or RRMS patients. Taken together, this data indicates that our flow cytometry methods are successfully detecting events consistent with EV.
EV isolated from patient plasma express the Myelin and Lymphocyte protein MAL
We have previously demonstrated that CNS endothelial cells specifically express MAL compared to endothelial cells from other organ systems(62). To identify EV originating from CNS endothelial cells, we have proposed to detect a combination of pan-endothelial markers (CD31, CD105, or CD144) and MAL on individual EV. However, we first wanted to determine if MAL expression was detectable on EV by binding of the MAL-specific ligand, Clostridium perfringens epsilon protoxin (pETX)(77, 78) via flow cytometry. To detect MAL expression, SEC isolated EV were probed with Alexaflour-647 conjugated pETX (pETX-647). As a negative control, EV were also probed with pETX-647 pretreated with an antibody that prevents binding pETX binding to MAL(67) (Fig. 4A). Unstained EV were used as an additional control. Staining with pETX-647 demonstrated a significant increase in fluorescence compared to EV stained with the antibody treated pETX-647 (Fig. 4B). This data indicates that MAL expression on EV can be detected by pETX-647 binding.
The concentration of total EV (Fig. 4C) and the concentration of pETX/MAL + EVs (Fig. 4D) were then determined for our different patient groups. Enumeration of total EV concentrations revealed a significant difference in concentrations between RRMS patients with active disease and RRMS patients receiving natalizumab, but no other significant differences were observed (Fig. 4C). In comparison, when examining concentrations of pETX/MAL + EV, active RRMS patients had significantly higher concentrations than all other patient groups (Fig. 4D). This data indicates pETX/MAL + EV concentrations are increased in RRMS patients with active disease.
CNS-EEV concentrations are increased in active MS
To identify EV originating from CNS endothelial cells (CNS-EEV), we developed a phenotyping strategy to identify EV that express both pan-endothelial markers CD31, CD105, or CD144 and MAL (Fig. 5A and 5B). Representative scatter plots of anti-CD-stained EV and corresponding isotype controls are depicted in Supplemental Fig. 1A-D. Because platelets also express CD31, we excluded platelet derived EV by eliminating events positive for CD41. Because lymphocytes also express MAL, we also excluded any lymphocyte derived EV by eliminating events positive for CD3. EV negative for CD3 and CD41 (CD3/CD41-) were further analyzed for the presence of CD31, CD105, and CD144 to determine if they were endothelial derived EV (EEV). CD3/CD41- EV positive for CD31, CD105, or CD144 were referred to as EEV31, EEV105, and EEV144 respectively. To determine if EEV originated from CNS endothelial cells, MAL expression on EEV31, EEV105, and EEV144 was evaluated by pETX-647 binding. EEV31, EEV105, and EEV144 positive for pETX/MAL are referred to as CNS-EEV31, CNS-EEV105, and CNS-EEV144, respectively. Enumeration of CNS-EEV31, CNS-EEV105, and CNS-EEV144 from our different patient groups revealed significant increases in the concentration of CNS-EEV in active RRMS patients compared to all other patient groups (Fig. 5C). When examining CNS-EEV31 concentrations, active RRMS patients had significantly higher concentrations than stable RRMS patients not receiving DMT, stable RRMS patients receiving natalizumab, and a trend towards significance compared to HC. When examining CNS-EEV105, active RRMS patients had significantly higher concentrations than all other patient groups. When examining CNS-EV144, active RRMS patients had significantly higher concentrations compared to HC and RRMS patients receiving natalizumab or ocrelizumab. Expression of CD31, CD105, and CD144 on EV was confirmed via western blot analysis of EV lysate from three separate donors (Fig. 5D). This indicates that active RRMS patients have increased levels of CNS-EEV31, CNS-EEV105, and CNS-EEV compared to HC and stable RRMS patients.
Finally, we wished to determine if individual CNS-EEV populations or individual EEV populations were more sensitive markers of active disease in RRMS patients. When examining the separate EEV populations, we observed an increase in EEV31, EEV105, and EEV144 concentrations in active RRMS patients compared to some of the other patient groups (Supplementary Fig. 2F-G). However, the statistical difference observed between levels of CNS-EEV31, CNS-EEV105, and CNS-EEV144 were more common and more significant than their EEV counterparts. In addition, we saw no significant difference in CD3/CD41 + EV concentrations (Supplementary Fig. 2E). Taken together, this data indicates that measurement of individual CNS-EEV concentrations is a more specific marker of disease activity in RRMS patients compared to total EV concentrations, individual EEV populations, or CD3/CD41 + EV populations.
We next wanted to determine if CNS-EEV31, CNS-EEV105, and CNS-EEV144 are unique EV populations or express multiple pan-endothelial markers. To achieve this, CD3/CD41- EV were analyzed for the presence of multiple endothelial markers by comparing expression of CD31 versus CD105, CD31 versus CD144, and CD105 versus CD144 (Fig. 6A-B). Examination of these events from seven HC and seven active RRMS patients revealed that the majority of EV only expressed a single endothelial marker (Fig. 4C). When comparing CD31 versus CD105 expression, only 0.3–0.6% percent of CD3/CD41- events were positive for both CD31 and CD105. When comparing CD31 versus CD144 expression, only 0.3–0.5% percent of CD3/CD41- events were positive for both CD31 and CD144. Finally, when comparing CD105 versus CD144, only 0.5–0.8% of CD3/CD41- events express both CD105 and CD144. This suggests that EEV31, EEV105, and EEV144 are unique populations and therefor CNS-EEV31, CNS-EEV105, and CNS-EEV144 are unique populations as well. We therefore reasoned that the total concentration of CNS-EEV in a single patient sample could be calculated by taking the sum of the individual CNS-EEV31, CNSEEV105, and CNS-EEV144 concentrations (Fig. 4D). This analysis revealed a significantly higher level of total CNS-EEV in active RRMS patients compared to all other patient groups. The same methodology was applied to determine the total EEV concentration (Supplemental Fig. 3). Active RRMS patients had significantly higher concentrations of total EEV compared to all other patient groups. However, the statistical difference observed between total CNS-EEV concentration was more significant than the differences seen between total EEV concentrations. For example, when examining total CNS-EEV concentrations, active RRMS were significantly higher versus HC (p = 0.001), stable RRMS receiving no DMT (p = 0.001), stable RRMS receiving natalizumab (p = 0.001), and stable RRMS patients receiving ocrelizumab (p = 0.001). In comparison, when examining total EEV concentrations, active RRMS were significantly higher versus HC (p = 0.02), stable RRMS receiving no DMT (p = 0.003), stable RRMS receiving natalizumab (p = 0.003), and stable RRMS patients receiving ocrelizumab (p = 0.001). Taken together, this data indicates that CNS-EEV concentrations are a more sensitive measurement of disease activity in MS patients compared to total EEV concentrations.
Finally, we wished to determine if total CNS-EEV concentrations were a more sensitive measurement of disease activity compared to the percentage of EV determined to be CNS-EEV (Supplemental Fig. 3). First, we examined the percent of total EV determined to be CNS-EEV (Supplemental Fig. 3A). To calculate this, we divided the total CNS-EEV concentration by the total EV concentration for each individual donor. No significant difference was observed between patient groups. Next, we examined the percent of total EEV that were determined to be CNS-EEV (Supplemental Fig. 3B). To calculate this, we divided the total CNS-EEV concentration by the total EEV concentration for each individual donor. No significant difference was observed between the different patient groups. Taken together, this data indicates that measurements of total CNS-EEV concentrations is a better measurement of disease activity in MS patients versus percentage calculations.
Analysis of Active RRMS Group
Interestingly, when examining the total CNS-EEV concentrations in the active RRMS patient group, we observed what appeared to be a bimodal distribution (Fig. 6D) with 11 of the 16 (68.75%) patients having EV concentrations higher than 100 EV per ul of plasma, and 5 out of the 16 patients (31.25%) below 100 EV per ul of plasma. Although these active RRMS patients were not receiving DMT at the time of blood draw, a portion of them had received steroids within 35 days to help manage symptoms (range one to 34 days). To determine if steroid use may influence the total CNS-EEV concentrations in these patients, we compared total CNS-EEV concentrations between active RRMS patients who had received steroid treatment within 35 days prior to their blood draw (+ steroids) and those who had not (- steroids). Steroid use information was available for 14 out of our 16 active RRMS patients. Active RRMS patients who had received steroid treatment (n = 6) demonstrated a trend towards significantly lower total CNS-EEV concentrations than those without (n = 8) (p = 0.16) (Fig. 7A). In addition, no correlation was observed between total CNS-EEV concentrations and individual patient EDSS (Fig. 7B) nor was a correlation observed between total CNS-EEV concentration and the time elapsed from blood draw to MRI (Fig. 7C). MRI dates were available for 11 patients. Taken together this indicates that steroid use may influence total CNS-EEV concentrations, however our small sample size lacks sufficient power for a conclusive analysis and more testing is required.
Next, we wanted to determine if there was a consistent relationship between total CNS-EEV concentrations as well as individual CNS-EEV31, CNS-EEV105, CNS-EEV144 concentration in active RRMS patients. In other words, we wanted to know if CNS-EEV31 concentration were low for an individual patient, would CNS-EEV105 and CNS-EEV105 concentrations also be low? To test for this, we performed a correlation analysis and paired t tests. We observed significantly strong, positive correlations between total CNS-EEV concentrations and CNS-EEV31 concentrations (r = 0.75, r2 = 0.56, and p = 0.0008), CNS-EEV105 concentrations (r = 0.92, r2 = 0.85, and p < 0.0001), ad CNS-EEV1144 concentrations (r = 0.82, r2 = 0.67, and p = 0.0001) (Fig. 7D). Paired T test analysis also revealed a significant or close to significant relationship between individual CNS-EEV population concentrations including CNS-EEV31 and CNS-EEV105, CNS-EEV31 (p = 0.09) and CNS EEV144(p < 0.05), and CNS-EEV105 and CNS-EEV144 (p < 0.003) (Fig. 7E-G). Taken together, this indicates that CNS-EEV populations have a consistent relationship with one another in a single patient. In other words, if CNS-EEV31 concentrations are low for a single active RRMS patient, CNS-EEV105 and CNS-EV144 concentrations are typically low too and vice.
Total CNS-EEV concentrations correlate with total EEV concentrations in RRMS patients
To determine if the total CNS-EEV concentrations correlated with the total EV concentrations, we performed a correlation analysis examining all donors as a single group (Fig. 8A). When examining all donors as a single group, total CNS-EEV concentrations demonstrated a moderate, but significant positive correlation with total EV concentrations (r = 0.40, r2 = 0.16, and p = 0.0002). However, when donors were separated into their individual patient groups, only active RRMS patients demonstrated a positive correlation between total CNS-EEV and total EV concentrations (r = 0.55, r2 = 0.30, and p = 0.03) (Fig. 8B). This indicates that an increase in total CNS-EEV concentrations correlates with an increase in total EV concentrations in active RRMS.
To determine if the total CNS-EEV concentrations correlated with total EEV concentrations, we performed a correlation analysis examining all donors as a single group (Fig. 8C). When examining all donors as a single group, total CNS-EEV concentrations demonstrated a strong and significant positive correlation with total EEV concentrations (r = 0.74, r2 = 0.54, and p < 0.0001). When donors were separated into their individual patient groups, a similar trend was observed for all RRMS groups, but not HC (Fig. 8B). Active RRMS demonstrated a significantly strong correlation (r = 0.74, r2 = 0.55, and p = 0.001) while stable RRMS patients on natalizumab had a moderate but significant correlation (r = 0.62, r2 = 0.39, and p = 0.007). Stable RRMS patients not receiving DMT (r = 0.46, r2 = 0.21, and p = 0.09) and stable RRMS patients receiving ocrelizumab (r = 0.50, r2 = 0.25, p = 0.07) both demonstrated a moderate correlation approaching significance. Taken together, this data indicates that CNS-EEV concentrations positively correlate with total EEV concentrations in RRMS patients, but not HC.
Total CNS-EEV concentrations do not depend on patient age but may be influenced by gender
To determine if patient age influences total CNS-EEV concentrations, a correlation analysis was performed examining at all donors as a single group or when donors were separated into their individual patient groups (Fig. 9A and B). No significant correlation was observed. This indicates that age does not influence total CNS-EEV concentrations.
To determine if gender influences total CNS-EEV concentrations, total CNS-EEV concentrations were compared between female and male donors in the individual patient groups (Fig. 9C). This analysis revealed no significant difference between males and females. However, this may be due to the low number of men enrolled in this study. To try and overcome this, we compared total CNS-EEV concentrations when all RRMS patients were grouped together or when all donors including HC were grouped together. This analysis revealed that male donors had significantly higher levels of total CNS-EEV compared to female donors when all RRMS donors were grouped together and when all donors including HC were grouped together. This indicates that men with RRMS may have higher levels of total CNS-EEV compared to females with RRMS. However, these results are only preliminary because of the small sample size.