Here, using remaining samples from the previous study, we evaluated whether adding nEVs and oEVs pS129-α-syn, total tau, tau phosphorylated at Thr 181 (pT181-tau), and/or serum/plasma neurofilament light chain (NfL) to the previously measured α-syn might improve the diagnostic power when added to the model. The methods used are provided as Supplemental Information. The clinical and epidemiologic data are in Supplemental Tables 1–3. Most results are presented as log-transformed values to improve normalization and facilitate statistical analysis. Non-transformed values are summarized in Supplemental Table 4.
The levels of pS129-α-syn in normal adult human brain are ~ 4% of total α-syn and may increase up to ~ 90% in LBs (6). pS129-α-syn also is highly enriched in GCIs in the MSA brain, though to a lesser degree than in LBs (7). These differences suggest that the pS129-α-syn concentrations in nEVs and/or oEVs may help distinguish among the groups.
Measurement of pS129-α-syn in 32 HC, 46 PD, and 30 MSA samples showed that the concentrations in many cases were a small fraction of the total α-syn (Supplementary Fig. S1a). In the nEVs, the pS129-α-syn concentrations trended toward a decrease in the order HC > PD > MSA but the differences were statistically insignificant (Fig. 1a). In contrast, pS129-α-syn concentrations in oEVs increased in the same order, HC < PD < MSA, and were significantly higher in both disease groups compared to the HC group (Fig. 1a). The data suggest that EV-mediated removal of pS129-α-syn from neurons is not affected by synucleinopathy, whereas in oligodendrocytes it is increased in MSA and in a subgroup of patients with PD compared to oligodendrocytes in normal, age-matched brains.
The oEV:nEV pS129-α-syn ratio showed a similar behavior, increasing in the order HC < PD < MSA (Supplementary Fig. S1b). In most patients with MSA this ratio was > 1, as reported previously for total α-syn (5). However, in the PD group roughly equal numbers of samples were < 1 or > 1. Consequently, unlike the oEV:nEV total α-syn ratio, which strongly discriminated between PD and MSA, the same ratio of pS129-α-syn concentration only moderately separated the groups.
Although tau and hyperphosphorylated tau oligomerization and aggregation typically are thought of in the context of Alzheimer’s disease and other tauopathies, genome-wide association and other studies investigating genetic risks for idiopathic PD revealed that polymorphisms in the MAPT gene are strongly linked to PD (8). Abnormal tau has been shown in postmortem studies to be abundant in the striatum of patients with PD and tau aggregates have been observed in ~ 50% of PD brains. In contrast, brain hyperphosphorylated tau pathology only rarely has been reported in patients with MSA and in those cases, it might have reflected a co-pathology of AD. Thr 181 is a common phosphorylation site in tau found in various pathological conditions and CSF pT181-tau has been used as a biomarker for PD and atypical parkinsonian disorders.
Total tau was measured in 54 HC, 51 PD, and 41 MSA samples. One sample was excluded from the oEV MSA group because it was 39 standard deviations above the mean. The tau concentrations were approximately an order of magnitude lower than the α-syn concentrations (Supplemental Table 4), below 1 pg/ml in many of the samples. In a few samples, the signal was below the LLoD and was imputed as the minimum/2 value (Fig. 1b). The average concentration in both nEVs and oEVs was significantly lower in the MSA group compared to the HC and PD groups, in agreement with the scarce observation of tau pathology in MSA brain. The HC and PD groups had similar tau concentrations in both nEVs and oEVs. The differences between nEVs and oEVs were statistically insignificant in all three groups (Fig. 1b), unlike the total α-syn (5) and pS129-α-syn results (Fig. 1a), suggesting that in contrast to α-syn, the oEV:nEV tau concentration was not useful for distinguishing MSA from HC or PD and that measurement of tau in one type of EV is sufficient.
In our first attempt to measure pT181-tau we isolated nEVs an oEVs from 38 samples (9 HC, 17 PD and 12 MSA). Unfortunately, pT181-tau levels were detectable only in 6 of the 76 EV lysates. One PD and one MSA samples showed detectable pT181-tau concentrations in both nEVs and oEVs, whereas another MSA sample and one HC sample had detectable pT181-tau concentrations in oEVs only. These findings suggest that pT18- tau levels in CNS-originating EVs isolated from the samples used in our study were only a small fraction of the total tau concentration and demonstrate that at the present assay sensitivity levels this analyte is not useful for distinguishing among HC, PD, and MSA. Alternatively, the low levels may be due to sample age.
Increased concentrations of NfL in the CSF have emerged in recent years as a measure of neuroaxonal injury and neurodegeneration not specific to a particular disease. Several studies have shown that plasma NfL concentrations correlate well with those in the CSF. Previously, Hansson et al. reported significantly higher plasma NfL concentrations in patients with MSA compared to HC and patients with PD (9). Therefore, we tested whether similar differences could be detected in our cohort. An important difference between our measurement and that of Hansson et al. was that they used plasma whereas the majority of our samples were serum. NfL concentrations in serum and plasma have been reported to be strongly correlated in patients with multiple sclerosis, yet we are not aware of similar side-by-side comparisons in patients with synucleinopathies.
Because the volume needed for the direct measurement of NfL in each sample, 20 µL, is ~ 10% of the volume needed for EV isolation, we could measure NfL in a larger number of samples – 88 HC, 79 PD, and 74 MSA. Unlike the other biomarkers, which were measured using ECLIA, we used Simoa to quantify NfL to match most of the current literature. Our analysis showed a significantly lower average NfL concentration in the PD group compared to both the HC and the MSA groups (Fig. 1c). The lower concentration of NfL in PD compared to HC samples was in agreement with a recent paper by Chen et al. who reported similar findings in an Asian cohort (10). Moreover, Chen et al. reported that the lower in NfL plasma concentration was statistically significant only in females. Therefore, we tested if the same phenomenon could be found in our cohort. Indeed, we found lower NfL concentrations in females but not in males, in the PD group, corroborating the findings of the previous study (Supplementary Figure S2). In contrast to the report by Hansson et al., we did not find significant differences between HC and MSA when both sexes were analyzed together (Fig. 1c). When males and females were analyzed separately, the median NfL concentration in female patients with MSA (270 pg/mL) was significantly lower than in the healthy females (326 pg/mL, Supplementary Figure S2).
Previously, a multinomial logistic regression (MLR) model with LASSO variable selection performed the best out of four models examined and selected nEV α-syn concentration, the oEV:nEV α-syn concentration ratio, and total EV concentration to create a receiver operating characteristic (ROC) model that separated best among the HC, PD, and MSA groups (5). Here, we used the same approach to test whether pS129-α-syn, tau, or NfL might improve the separation when added separately to the previously selected parameters. The number of samples in which both pS129-α-syn and tau could be measured was too small in the current study to allow meaningful evaluation of both together. NfL was measured in most of the samples and therefore we tested it in combination with pS129-α-syn or tau.
For the subset of samples with pS129-α-syn measurements, as might be expected based on the data presented in Fig. 1a, the MLR model selected to add oEV pS129-α-syn concentrations to the previous parameters in the ROC analysis. The addition of pS129-α-syn improved the model predictions compared to the model that did not include oEV pS129-α-syn (Akaike information criterion (AIC) = 136.1 vs 138.7, respectively; AIC change = − 2.6). The improved model separated the PD and HC groups with AUC = 0.874, MSA and HC group with AUC = 0.993, and the PD and MSA groups with AUC = 0.936 (Fig. 1d, Supplemental Table 5). Addition of NfL did not improve the separation in this subset of samples.
For the subset of samples with total tau measurements, the model added oEV tau to the previously selected parameters. However, the addition of this parameter did not result in a significant improvement in the model predictions compared to the model that did not include oEV tau (Fig. 1e, Supplemental Table 5). When NfL was added to oEV tau, there was a significant improvement in model predictions (AIC = 260.8 vs. 263.1; AIC change = − 2.2) and the separation between the HC and PD groups in ROC analysis increased compared to tau alone from AUC = 0.633 to 0.720 (p = 0.055).
α-Syn concentration in both nEVs and oEVs in all the groups combined together correlated positively with oEV pS129-α-syn (rnEV = 0.38, roEV = 0.35; p < 0.0001) but not with nEV pS129-α-syn. To assess the contribution of the different groups to these correlations we tested them in each group separately. In the control group, no significant correlation was found between nEVs or oEVs total α-syn and nEVs or oEVs pS129-α-syn indicating that this group did not contribute to the observed correlations. In the PD group, both nEVs and oEVs α-syn correlated positively with oEVs pS129-α-syn (rnEV = 0.37, p = 0.010; roEV = 0.33, p = 0.025). Similarly, positive correlations were found also for the MSA group (rnEV = 0.38, p = 0.037; roEV = 0.38, p = 0.036). None of the biomarkers correlated with disease duration or disease progression as assessed by Unified Parkinson’s Disease Rating Scale-III (UPDRS-III), Unified Multiple System Atrophy Rating Scale (UMSARS), or Hoehn and Yahr scale (H&Y).
Our study is the first to measure pS129-α-syn in nEVs and oEVs and the first to assess tau concentrations in these types of EVs in patients with MSA. Though the results need validation in larger cohorts, the data support measurement of these and additional biomarkers in CNS-originating EVs toward constructing a panel for improved diagnosis of parkinsonian disorders.