A total transcriptome profiling method for plasma-derived extracellular vesicles: applications for liquid biopsies

Abstract

Extracellular vesicles (EVs) are key mediators of intercellular communication. Part of their biological effects can be attributed to the transfer of cargos of diverse types of RNAs, which are promising diagnostic and prognostic biomarkers. EVs found in human biofluids are a valuable source for the development of minimally invasive assays. However, the total transcriptional landscape of EVs is still largely unknown. Here we develop a new method for total transcriptome profiling of plasma-derived EVs by next generation sequencing (NGS) from limited quantities of patient-derived clinical samples, which enables the unbiased characterization of the complete RNA cargo, including both small- and long-RNAs, in a single library preparation step. This approach was applied to RNA extracted from EVs isolated by ultracentrifugation from the plasma of five healthy volunteers. Among the most abundant RNAs identified we found small RNAs such as tRNAs, miRNAs and miscellaneous RNAs, which have largely unknown functions. We also identified protein-coding and long noncoding transcripts, as well as circular RNA species that were also experimentally validated. This method enables, for the first time, the full spectrum of transcriptome data to be obtained from minute patient-derived samples, and will therefore potentially allow the identification of cell-to-cell communication mechanisms and biomarkers.

Figure 1

Scheme of the bioinformatics analysis pipeline.

Figure 2

Characterization of EVs isolated from plasma by ultracentrifugation. (A) Whole-mount transmission electron micrograph of EVs displaying the characteristic “cup-shaped” morphology. (B) NanoSight quantification plot of EVs concentration in function of size shows that the majority of EVs are less than 250 nm in the healthy control samples (n = 5). (C) Western blot of the vesicle-associated markers CD63, FLOTILLIN, HSP70 and RAB27B from a pool of plasma samples (n = 4) and corresponding Ponceau staining of membrane before blocking.

Figure 3

Bioanalyzer electropherogram analysis of fragmentation time-points of leukocyte RNA. Arbitrary fluorescence units (FU) are plotted as a function of RNA size in nucleotides (nt). (A) Analysis of fragmentation time-points 15–180 min with Agilent RNA 6000 Pico Kit shows the expected ribosomal RNA peaks in non-fragmented sample (in red), whereas after all fragmentation time-points the majority of RNAs are below 200 nt in size. (B) Analysis of fragmentation time-points from 45 to 420 min with Agilent Small RNA Kit shows that the majority of RNAs are smaller than 40 nt in size for all time-points, and that a plateau is reached after approximately 180 min, as no further reduction in size is observed with longer fragmentation times. For the longer fragmentation periods, reduced amounts of RNA (marked with a star) were used to better simulate the enzymatic kinetics in the presence of less (but still detectable) RNA, and the plateau region is still the same.

Figure 4

Distribution of the RNA classes identified in the sequencing analysis of the total transcriptome of EVs isolated from the plasma of healthy controls (average values for five samples). Pie charts show the distribution of mapped reads according to gene biotype, as defined by Ensembl, and sequencing coverage. Distribution of reads is shown for the four major Ensembl biotypes (A,D); the short noncoding biotype (B,E) and the misc_RNA biotype (C,F). Genes considered were covered by at least 2 reads (A–C), or by a minimum of 10 reads (D–F).

Figure 5

Quantitative RT-PCR analysis of two miRNAs (miR-223-3p and let-7g-5p) identified by sequencing. The treatment of intact EVs with RNase before RNA extraction minimally altered Ct values of both miRNAs, strongly suggesting that these miRNAs were derived from the EVs-cargo and thereby protected from digestion by the membrane bilayer, whereas there was a large increase in Ct values when lysed EVs were similarly treated. Undetermined values (no detection by qRT-PCR) were assigned as Ct = 35. (A,B) Bar-graphs for miR-223-3p and let-7g-5p, showing Ct values for each treatment per sample; (C,D) Box-plots for miR-223-3p and let-7g-5p, showing Ct values per treatment (dots are results from each one of the five samples).