Cerebrospinal fluid-derived extracellular vesicles: a proteomic and transcriptomics comparative analysis of enrichment protocols

Proteomic and transcriptomic analyses of cerebrospinal fluid (CSF)-derived extracellular vesicles (EVs) offer unique insights into molecular changes associated with central nervous system (CNS) diseases and may result in biomarker identification. No gold standard method to enrich EVs from CSF is established and head-to-head comparisons of outputs of different protocols s' outputs are scarce. Using a large pool of CSF, we characterized the EV preparations resulting from four enrichment protocols and compared them in terms of yield and purity. We found that particles enriched by ultracentrifugation (UC) or a combination of ultrafiltration and size exclusion chromatography (UF-SEC) bared exhibited the typical morphological and biochemical characteristics of small EVs and were highly enriched in proteins and polyadenylated transcripts associated with EV-related biological processes. UF-SEC preparations had higher particle yields while more proteins were identified in UC preparations. Approximately 40% of the EV preparations' proteome was not identified in unenriched CSF, among which a core proteome of 45 proteins identified in 30 EV preparations from independent experiments which may serve as CSF-derived EV markers. Enrichment scores to protein contaminants albumin and apolipoprotein E were higher in UF-SEC preparations. In conclusion, all protocols analyzed here resulted in enrichment of particles with small EVs characteristics, with EV enrichments from UF-SEC resulting in highest yield and purity. Overall design: We used polyA RNA-seq which comprehensively quantifies messenger RNAs (mRNAs) and long non-coding RNAs (lncRNAs) to characterize and compare EV preparations' transcriptome. RNA gel analysis showed the presence of low molecular weight bands below 200 nucleotides, also in accordance with literature.41,62 Read counts were similar in all 3 protocols (3.3*106 for the UF-SEC35 EV preparation, 3.4*106 for the UF-SEC and UC EV preparations). Of these reads, more than 90% survived after adaptor trimming. For the UF-SEC35 and UF-SEC70 EV preparations, 59.2 % (1.9*106) and 59.7% (2.0*106) of the surviving reads were uniquely mapped to the human genome respectively, whereas this proportion was significantly higher in the UC EV preparation (86.9%, 2.9*106). Of the uniquely mapped reads, the number of genes detected with ? 10 counts were of 22'440 (UF-SEC35), 18'753 (UF-SEC70), and 30'171 (UC) respectively (Supplementary Table 5). Of these, 12999 genes overlapped between the 3 protocols (Figure 8A), 13092 genes (62.4%) between the UF-SEC35 and UF-SEC70 EV preparations (Figure 8B), 15070 (49.5%) between the UF-SEC35 and UC EV preparations (Figure 8C), and 17605 (56.2%) between the UF-SEC70 and UC EV preparations (Figure 8D). For all 3 protocols, the highest proportion of uniquely mapped genes with ? 10 counts were protein coding, the remainder consisting of lncRNAs. As expected, RNA species (ribosomal RNA, micro RNAs) were rare, given the type of sequencing protocol chosen (Supplementary Table 5). Cellular component GO enrichment analysis of the protein coding genes with >10 counts showed that these were enriched for genes associated with EV-related biological processes (Figure 9A-C) very similar to those identified on a proteins level (Figure 5).