A 96-Well Ultrafiltration Approach for the High-Throughput Proteome Analysis of Extracellular Vesicles Isolated From Conditioned Medium

Abstract

Extracellular vesicles (EVs), nanoscale vesicles that are secreted by cells, are critical mediators of intercellular communication and play a crucial role in diverse pathologies such as cancer development. Therefore, EVs are regarded as having high potential in the clinic, both for diagnostic and therapeutic applications. Unfortunately, EVs reside in complex biofluids and their consistent preparation at sufficient purity for mass spectrometry-based proteomics has proven to be challenging, especially when increased high-throughput is required. Here, we describe the incorporation of our previously reported filter-aided EV enrichment (FAEVEr) strategy for the separation of EVs from conditioned medium, from harvest to proteomic analysis completely to a streamlined 96-well format. We compared our approach with ultracentrifugation, the most widely used method for EV enrichment, in terms of protein identifications, consistency, reproducibility and overall performance, including the invested time, resources and required expertise. In addition, our results show that including relative high percentages of Tween-20, a mild detergent, markedly improves the final purity of the EV proteome by removing the bulk of non-EV proteins (e.g., serum proteins) and significantly increases the number of identified transmembrane proteins. Moreover, our FAEVEr 96-well strategy improves the overall reproducibility with a consistent number of protein identifications and decreased number of missing values across replicates. This promotes the validity and comparability between results, which is essential in both a clinical and research setting, where consistency is paramount.

Keywords: 300 kDa MWCO 96‐well ultrafiltration; Tween‐20; conditioned medium; proteomics.

FIGURE 1

Description of the FAEVEr 96‐well workflow. Conditioned medium containing 10% EV‐depleted FBS is isolated and pre‐cleared by centrifugation and filtration (0.2 µm filter) in a 96‐well format. The filtrate is collected and either frozen (−80°C) or processed immediately. After loading the pre‐cleared conditioned medium on a 300 kDa MWCO 96‐well filtration plate, the plate is centrifuged. The retained particles are purified by three consecutive wash steps. For LC‐MS/MS analysis, the purified EVs are lysed on the filter (yellow) followed by the collection of the proteome by centrifugation and processing by S‐Trap 96‐well. For characterization, the particles are recuperated from the filter (cyan) after the wash steps and analysed by NTA and TEM.

FIGURE 2

Recombinant EVs are successfully enriched and purified by FAEVEr using a 300 kDa MWCO 96‐well filter plate. rEV particles were enriched and recuperated from the filter for characterization by (a) TEM and (b) NTA. (c) Western blot analysis of the individual fractions shows that no luminal overexpressed Gag‐eGFP is observed in the initial flow‐through or wash, but is highly abundant in the enriched rEVs. (d) Eight rEV protein markers were searched across the different fractions after LC‐MS/MS analysis, indicating a high abundance (Log2(LFQ) intensities) in the lysate fraction. (e) Multiple–sample testing between the individual fractions revealed two major clusters of proteins that were more abundant in the lysate fraction compared to the other fractions. (f) Absolute and relative precursor intensities originate largely (> 96%) from bovine precursors in the input, load and wash fractions, but are greatly reduced in the lysate fraction (43%).

FIGURE 3

Comparison of the effect of the Tween‐20 concentration on the rEV purity. Retained rEVs were characterized by NTA (a) and TEM (b). (c) Number of protein identifications (spheres) with the number of proteins identified across all replicates (n = 3) indicated by the dash. The corresponding coefficient of variance (%CV) is indicated by ‘x’ on the right y‐axis. (d) Bovine precursor abundance decreases exponentially proportional to the concentration of Tween‐20; from nearly 70% at 0% Tween‐20 to 10% using 5% Tween‐20 in the washing buffer. (e) High Pearson correlation between the different replicates and clustered grouping of enrichment strategies (f) suggests high reproducibility. (g) Hierarchical clustering of significantly (BH corrected q value < 0.001) differentially abundant proteins reveals two major clusters. (h) Eight protein markers specific for rEVs were found across all the experiments with the transmembrane (CD63, CD9 and CD81), peripheral (FLOT1) and membrane‐associated (SDCBP) proteins being relatively higher abundant in the FAEVEr with Tween‐20 compared to the luminal markers (PDCD6IP, TSG101 and Gag‐eGFP) in ultracentrifugation. FAEVEr 96‐well without Tween‐20 has the overall lowest relative abundance of rEV markers.

FIGURE 4

Quantitative comparison of the identified proteins between FAEVEr 96‐well in combination with 5% Tween‐20 versus EVs isolated by ultracentrifugation. Volcano plot of the log₁₀(p values) and corresponding log₂ differences indicate that in FAEVEr 96‐well with 5% Tween‐20, the transmembrane (TM) proteins were relatively more abundant whereas bovine and human secreted proteins were significantly decreased, compared to the proteomes of EVs isolated by ultracentrifugation.