Evaluation of serum extracellular vesicle isolation methods for profiling miRNAs by next-generation sequencing

Abstract

Extracellular vesicles (EVs) are intercellular communicators with key functions in physiological and pathological processes and have recently garnered interest because of their diagnostic and therapeutic potential. The past decade has brought about the development and commercialization of a wide array of methods to isolate EVs from serum. Which subpopulations of EVs are captured strongly depends on the isolation method, which in turn determines how suitable resulting samples are for various downstream applications. To help clinicians and scientists choose the most appropriate approach for their experiments, isolation methods need to be comparatively characterized. Few attempts have been made to comprehensively analyse vesicular microRNAs (miRNAs) in patient biofluids for biomarker studies. To address this discrepancy, we set out to benchmark the performance of several isolation principles for serum EVs in healthy individuals and critically ill patients. Here, we compared five different methods of EV isolation in combination with two RNA extraction methods regarding their suitability for biomarker discovery-focused miRNA sequencing as well as biological characteristics of captured vesicles. Our findings reveal striking method-specific differences in both the properties of isolated vesicles and the ability of associated miRNAs to serve in biomarker research. While isolation by precipitation and membrane affinity was highly suitable for miRNA-based biomarker discovery, methods based on size-exclusion chromatography failed to separate patients from healthy volunteers. Isolated vesicles differed in size, quantity, purity and composition, indicating that each method captured distinctive populations of EVs as well as additional contaminants. Even though the focus of this work was on transcriptomic profiling of EV-miRNAs, our insights also apply to additional areas of research. We provide guidance for navigating the multitude of EV isolation methods available today and help researchers and clinicians make an informed choice about which strategy to use for experiments involving critically ill patients.

Keywords: Extracellular vesicle; biomarker; exosome isolation; miRNA; next-generation sequencing; precipitation; sepsis; small RNA sequencing; ultracentrifugation.

Figure 1.

Schematic summary of EV isolation, RNA extraction and downstream analyses. EVs were isolated from human serum using five (healthy donors) or four (sepsis patients) different methods. After extracting total RNA from EV isolates, small RNA species were profiled by NGS. Differential expression of miRNAs between volunteers and patients was assessed to identify potential biomarker candidates. Sera from a subset of volunteers and patients were used to additionally characterize isolates from each method by Western blot (WB), nanoparticle tracking analysis (NTA) and transmission electron microscopy (TEM).

Figure 2.

Mean library size and mapped miRNAs for EVs isolated from healthy volunteers (a) and sepsis patients (b). miRNA mapping frequencies (red diamonds) are expressed as percentages of total library size and plotted against the right x-axes. Enrichment of miRNA reads was highest for miRCURY (35.08% and 27.56% for volunteers and patients, respectively) and lowest for qEV (0.79% for volunteers and 0.57% for patients). All data are mean ± SD for 10 volunteers and 9 sepsis patients.

Figure 3.

Mapping statistics for various classes of small non-coding RNA. Highest frequencies of miRNA mapping were observed in isolates from precipitation, sedimentation and membrane affinity. Both SEC-based methods were prone to capture short sequences, while libraries from membrane affinity-derived samples contained an increased share of tRNA fragments. Short: sequence is shorter than 15 nt; unmapped: sequence did not align to human rRNA, snRNA, snoRNA, tRNA or miRNA. Data are expressed as mean mapping percentages for 10 volunteers (V) and 9 sepsis patients (S).

Figure 4.

Differential expression of miRNAs in EVs isolated by commercial methods. Precipitation and membrane affinity yielded high numbers of differentially regulated miRNAs (miRCURY: 90; exoRNeasy: 60). Far fewer regulated miRNAs were detected in SEC-derived samples (Exo-spin: 14; qEV: 6). Two differentially regulated miRNAs were detected in EVs isolated by all methods. Data are filtered for baseMean ≥50, log₂ fold change ≥|1| and adjusted p-value ≤0.05.

Figure 5.

Hierarchical clustering analysis of miRNAs in EVs isolated by commercial methods. Samples split up into two clusters, separating precipitation and membrane affinity from both SEC-based methods. miRCURY (blue) and exoRNeasy (red) accurately distinguished between healthy volunteers (darker shades, V) and sepsis patients (lighter shades, S). miRNAs isolated from SEC-EVs (Exo-spin, qEV) showed noticeable heterogeneity and were less capable of separating volunteers and patients.

Figure 6.

Analysis of EVs by NTA demonstrates differences in size distribution (a). Whiskers indicate 1st and 99th percentiles; line: mean diameter; dot: modal diameter; V: volunteer; S: sepsis patient. Precipitation- and membrane affinity-based methods isolated the smallest and largest EVs, respectively. Concentration and purity of isolated EVs differed depending on isolation strategies (b). Black bars indicate the absolute number of vesicles isolated from 1 ml serum; red diamonds plotted against the right x-axis represent vesicle purity defined as the particle to protein ratio. While precipitation most efficiently isolated EVs from serum, SEC-based isolation yielded fewer but highly pure vesicles. Asterisks indicate significant differences in particle numbers compared to miRCURY. *p < 0.05; **p < 0.01; NS: not significant. All data are mean ± SD for five volunteers and five sepsis patients.

Figure 7.

Morphology of serum EVs by transmission electron microscopy. Images are representative for three separate biological replicates for both volunteers (top panel) and sepsis patients (bottom panel). Scale bars are 500 nm (top row) and 100 nm (bottom row).

Figure 8.

Analysis of marker proteins in EVs from volunteers (left) and sepsis patients (right). EV markers CD63 and syntenin were detected in vesicles isolated by membrane affinity (exoRNeasy) and SEC (qEV, Exo-spin), but not precipitation (miRCURY) and UC. All EV isolates were negative for TSG101, CD81 and calnexin. Significant albumin contamination of EVs was found for non-SEC isolation methods. Results are representative for three separate biological replicates for both volunteers and sepsis patients.

Figure 9.

Analysis of EV markers and human serum albumin in EVs isolated by precipitation and sedimentation and further purified by iodixanol density gradient centrifugation. CD63 and syntenin were detected in a density fraction of 1.18 g/ml, while the majority of albumin floated in fractions of 1.02–1.05 g/ml. Results are representative for two separate biological replicates for both volunteers (top panel) and patients (bottom panel).