Immune stimuli shape the small non-coding transcriptome of extracellular vesicles released by dendritic cells
- 2k Downloads
The release and uptake of nano-sized extracellular vesicles (EV) is a highly conserved means of intercellular communication. The molecular composition of EV, and thereby their signaling function to target cells, is regulated by cellular activation and differentiation stimuli. EV are regarded as snapshots of cells and are, therefore, in the limelight as biomarkers for disease. Although research on EV-associated RNA has predominantly focused on microRNAs, the transcriptome of EV consists of multiple classes of small non-coding RNAs with potential gene-regulatory functions. It is not known whether environmental cues imposed on cells induce specific changes in a broad range of EV-associated RNA classes. Here, we investigated whether immune-activating or -suppressing stimuli imposed on primary dendritic cells affected the release of various small non-coding RNAs via EV. The small RNA transcriptomes of highly pure EV populations free from ribonucleoprotein particles were analyzed by RNA sequencing and RT-qPCR. Immune stimulus-specific changes were found in the miRNA, snoRNA, and Y-RNA content of EV from dendritic cells, whereas tRNA and snRNA levels were much less affected. Only part of the changes in EV-RNA content reflected changes in cellular RNA, which urges caution in interpreting EV as snapshots of cells. By comprehensive analysis of RNA obtained from highly purified EV, we demonstrate that multiple RNA classes contribute to genetic messages conveyed via EV. The identification of multiple RNA classes that display cell stimulation-dependent association with EV is the prelude to unraveling the function and biomarker potential of these EV-RNAs.
KeywordsImmune activation Immune suppression Small RNA sequencing Biomarker Extracellular RNA
Bovine serum albumin
Small nuclear RNA
Small nucleolar RNA
Extracellular vesicles (EV) released by cells are considered as important mediators of intercellular communication [1, 2]. These 50–200 nm-sized vesicles are released by a broad range of cells, and have been detected in a wide range of body fluids, such as milk, plasma, urine, and semen [1, 3]. The collective term ‘EV’ refers to a heterogeneous population of secreted vesicles that are formed via different pathways. Exosomes are formed as intraluminal vesicles (ILVs) inside multi-vesicular endosomes (MVE) and are released upon fusion of the MVE with the plasma membrane, whereas microvesicles directly bud off from the plasma membrane . These vesicle populations overlap in size and molecular composition, which currently hampers their discrimination based on biophysical or biochemical parameters [4, 5, 6].
EV are multi-component entities that transfer proteins, lipids, and RNA between cells. The RNA associated with EV mainly consists of small non-coding RNA species (ncRNA), which are thought to be of major importance in altering molecular processes in the recipient cell [3, 4, 7, 8]. In addition, the RNA composition of EV can give information about the (patho)physiological status of the EV-producing cell. Since EV can be easily obtained from body fluids, such EV-associated RNAs may be used as non-invasive biomarkers for the early detection of disease [9, 10]. The molecular composition of EV is often assumed to be a ‘snapshot’ of the producing cell, from which the tissue type or mutational status of the parental cell can be deduced [9, 11, 12].
Although the EV-RNA field receives major attention, with quickly rising numbers of publications on this topic, the technical difficulties to obtain pure EV remain underexposed. It is important to note that a substantial amount of extracellular RNA is not contained in EV, but is associated with ribonucleoprotein particles (RNPs), including argonaute 2 (AGO2), or lipoprotein complexes [13, 14, 15]. These contaminating structures have been shown to co-isolate together with EV via commonly used isolation procedures, and may, therefore, greatly affect the outcomes of experiments [4, 5, 16]. The currently most reliable method to obtain pure EV is by high-speed ultracentrifugation followed by density centrifugation in which EV are separated from contaminating structures based on differences in buoyant density . Subsequently, analyzing the presence of common EV markers (such as CD9, CD63, and CD81) and absence of contaminant proteins should be performed to assess the purity of EV preparations used in the study [17, 18].
MiRNAs are the most intensely studied type of small RNA in EV, likely due to the interest in gene-regulatory functions of these RNAs [19, 20, 21]. It has been shown that changes in the activation status of the EV-producing cell lead to changes in the EV-miRNA composition [22, 23, 24]. Moreover, miRNAs in EV were shown to be functionally transferred between cells, leading to repression of genes in target cells [25, 26]. It is, therefore, thought that differences in the miRNA content of EV underlie distinct functional effects of EV released by differentially stimulated cells [22, 23, 24]. In addition, changes in EV-miRNA levels in plasma or serum have been associated with diseases such as cancer, rheumatoid arthritis, and Alzheimer’s disease [27, 28, 29].
Despite the interest in miRNAs, a larger part of EV-associated RNA consists of other ncRNA classes. We and others previously showed that EV released by cultured cells and EV present in a variety of body fluids contain many other small non-coding RNA species (20–300 nt), such as tRNA, snoRNA, snRNA, Y-RNA, Vault RNA, and SRP-RNA (also named 7SL) [23, 30, 31, 32, 33, 34, 35]. Some of these non-coding RNAs are relatively enriched in EV, suggesting that these RNAs are specifically shuttled into EV for release into the extracellular milieu. Inside cells, the above-mentioned non-coding RNAs are known to play a role in basic cellular processes such as translation, RNA splicing, and RNA quality surveillance. Recent studies showed that full-length and fragmented forms of these non-coding RNAs can additionally be involved in other processes including immunological signaling, gene regulation, and guiding of other RNAs and nucleases [36, 37, 38, 39, 40, 41]. However, it is not known whether activation stimuli imposed on the EV-producing cell affect the levels of these ncRNA classes in EV, similar to EV-associated miRNAs. Evaluation of cell stimulation-dependent changes in EV-associated ncRNA classes is a first step in uncovering their function in EV-mediated signaling processes and their potential as biomarkers indicative of the activation status of cells.
Here, we studied how changes in the activation/differentiation status of the parental cell affect the small non-coding RNA content of EV by employing a primary immune cell model, two different types of cell stimulation, and methods yielding highly pure EV populations. For our analysis, we used primary dendritic cells (DC), which are master regulators of immune responses and have been shown to communicate with various other immune cells via EV [2, 8, 42]. In addition, the use of DC-derived EV has been proposed as a strategy for cancer immune(chemo)therapy [43, 44, 45, 46]. Depending on external stimuli, DC can be differentiated into functionally different phenotypes that either activate or down-regulate immune responses . These DC, therefore, present a suitable model to investigate how different external stimuli affect incorporation of a broad range of RNA classes into EV. We compared cellular and EV-associated RNA levels in control, immune-stimulating, or immune-suppressing DC conditions using RNA deep sequencing and RT-qPCR. Our data indicate that differentiation signals imposed on dendritic cells affect the EV-associated levels of particular RNA classes, such as miRNA, snoRNA, and Y-RNA, while other non-coding RNA types remain largely unchanged. Moreover, only a minor part of the stimulation-induced changes in EV-RNA content reflects changes in cellular RNA levels. This study exemplifies how comprehensive analysis of RNA obtained from highly purified EV yields candidate small ncRNAs beyond miRNAs for further exploration as biomarkers or functional entities within EV.
Materials and methods
Complete culture medium was prepared by supplementing Iscove’s Modified Dulbecco’s Medium (IMDM, Lonza, Verviers, Belgium) with 2 mM Ultraglutamine (Lonza, Verviers, Belgium), 10% fetal calf serum (FCS, GE Healthcare Bio-Sciences, Pasching, Austria), 100 IU/ml penicillin and 100 μg/ml streptomycin (Gibco, Paisley, United Kingdom), and 50 µM β-mercaptoethanol. The GM-CSF producing NIH 3T3 cell line (R1) was grown in complete medium. Primary bone marrow cells were grown in complete medium supplemented with 30% conditioned medium from R1 cells. To prepare EV-depleted medium, a mixture of 150 ml conditioned R1 culture supernatant and 42.5 ml FCS (end concentration 30% FCS) was depleted of EV by overnight centrifugation at 100,000g in an SW28 rotor (k-factor 334.2) (Beckman Coulter, Brea, CA). The EV-depleted supernatant was carefully pipetted from the tubes, leaving 5 ml in the tubes to prevent disturbance of the pellet. This supernatant was filtered through a 0.22 μm bottle top filter (Millipore, Billerica, MA) after which additional IMDM, ultraglutamine, antibiotics, and β-mercaptoethanol were added to prepare complete medium as indicated above. The efficiency of EV depletion from culture medium, as regularly assessed by high-resolution flow cytometric analysis, is on average ~ 90%. Primary bone marrow cells were flushed from the femur and tibia of 8–12-week-old C57bl/6 mice and differentiated to dendritic cells according to Lutz et al. . To generate tolerogenic DC, cells were treated from day 2 onward with 10 nM 1α,25-dihydroxyvitaminD3 (Sigma, St Louis, MO). To generate immunogenic DC, 1 µg/ml lipopolysaccharide (LPS, cat. L-2630, Sigma, St. Louis, MO) was added on day 12. On day 13, non-adherent and semi-adherent cells were recovered and cultured at 3 × 106 cells/dish for 20 h in EV-depleted culture medium. Cell viability, as determined by Trypan blue exclusion, did not differ between treatments and was above 90% for all cultures. All cells were maintained at 37 °C and 5% CO2 in a humidified incubator. Experiments were approved by the institutional ethical animal committee at Utrecht University (Utrecht, The Netherlands).
Day 12 bone marrow DC (BMDC) were collected and labeled for 30 min in PBS + 1% BSA (Bovine Serum Albumin, cat. K45-001, GE Healthcare Bio-Sciences, Pasching, Austria) containing anti-CD11c-APC (eBioscience, clone N418, 1:400), anti-MHCII-FITC (eBioscience, clone M5/114.15.2, 1:1500), and anti-CD40-PE (BD Biosciences, clone 3/23, 1:200), or anti-CD86-PE (eBioscience, clone GL1, 1:200) or anti-CD274-PE (PD-L1, eBioscience, clone M1H5, 1:400). As control, corresponding isotype control antibodies (eBioscience, San Diego, CA) were used. Surface labeling was analyzed using a FACS Calibur Flow Cytometer (BD Biosciences, San Jose, CA). Data analysis was performed using FCS Express V3 (DeNovo Software, Glendale, CA).
Fluorescent labeling and purification of EV
Conditioned cell supernatants were pooled to volumes of 100–140 ml per condition and were subjected to differential centrifugation as described previously . In brief: supernatant was sequentially centrifuged 2 × 200g for 10 min, 2 × 500g for 10 min, and 1 × 10,000g for 30 min. Next, EV were pelleted by ultracentrifugation at 100,000g for 65 min using an SW28 rotor (k-factor 334.2) (Beckman Coulter, Brea, CA). For EV quantification by high-resolution flow cytometry, pellets were resuspended in 20 µl PBS + 0.2% BSA (cleared from aggregates by overnight ultracentrifugation at 100,000g) and labeled with 1.5 µl PKH67 (Sigma, St. Louis, MO) in 100 µl Diluent C per pellet. For EV-RNA isolation, 100,000 g pellets were resuspended in 50 µl PBS + 0.2% BSA. PKH67-labeled samples or samples for RNA isolation were mixed with 1.5 ml 2.5 M sucrose, and overlaid with a linear sucrose gradient (2.0–0.4 M sucrose in PBS). Gradients were centrifuged 15–18 h at 192,000g in a SW40 rotor (k-factor 144.5) (Beckman Coulter, Brea, CA). PKH67-labeled EV were used for high-resolution flow cytometric analysis (see below). Alternatively, the fractions with densities of 1.12–1.18 g/ml (as measured by refractometry) were pooled, diluted six times in PBS + 0.2% EV-depleted BSA, and centrifuged again at 192,000g for 65 min in a SW40 rotor (k-factor 144.5) prior to EV-RNA isolation.
Small RNA was isolated from EV pellets and from 1 × 106 cells using the miRNeasy Micro kit according to the small RNA enrichment protocol provided by the manufacturer (Qiagen, Hilden, Germany). RNA yield and size profile were assessed using the Agilent 2100 Bioanalyzer with Pico 6000 RNA chips (Agilent Technologies, Waldbronn, Germany).
Preparation of small RNA sequencing libraries
50 ng cellular small RNA and 2 ng EV-derived small RNA was treated with DNase [Turbo DNA-free kit (Life Technologies, Carlsbad, CA)] according to the manufacturer’s instructions. The DNase-treated RNA was pelleted using Pellet Paint (Merck, Darmstadt, Germany) according to the manufacturer’s instructions, reconstituted in milliQ (MQ) and subsequently subjected to ribosomal RNA depletion using RiboZero Gold kit (Human/Mouse/Rat) (Illumina, San Diego, CA) according to the manufacturer’s instructions. This is required to deplete residual rRNA from cellular RNA samples and, for comparability, we subjected EV-RNA samples to the same procedure. Hereafter, RNA was pelleted with Pellet Paint and reconstituted into 6 µl MQ. cDNA libraries were prepared using the NebNext smallRNA library prep kit for Illumina (New England Biolabs, Ipswich, MA), according to the manufacturer’s instructions but with the following adaptations: 3′ adapter-ligation was carried out overnight at 16 °C; PCR amplification was done using Kapa HiFi Readymix 2× PCR mastermix (Kapa Biosystems, Wilmington, MA) using barcoded primers and the following PCR programme: 2 min at 95 °C, 15 cycles of 20 s at 98 °C, 30 s at 62 °C, 15 s at 70 °C, and a final elongation step of 5 min at 70 °C. cDNA was purified using magnetic AMPure XP beads (Beckman Coulter, Brea, CA) and quantified using Agilent 2100 Bioanalyzer and DNA HiSensitivity chips (Agilent Technologies, Waldbronn, Germany). Adapter dimers (126 nt in size) were removed by running the cDNA on 6% TBE gels (Life Technologies, Carlsbad, CA) for 60 min at 145 V, after which cDNA products of 15–300 nt (+ 126 nt adapters) were cut from the gel and purified. Subsequently, all libraries were pooled at equimolar ratios and run on a 4–12% TBE gel (Life Technologies, Carlsbad, CA). Size fractions of 15–25 nt (includes miRNAs), 25–60 nt, 60–80 nt (includes tRNAs), and 80–275 nt (each + 126 nt adapters) were cut from the gel, and purified and pooled in a volumetric ratio of 2:2:3:3. Sequencing was done using 150 bp paired-end reads on an Illumina HiSeq 4000 machine (Illumina, San Diego, CA) at ServiceXS (Leiden, The Netherlands).
RNAseq data analysis
All data was processed as paired-end reads. Data quality was checked with FastQC and reads were processed with cutadapt (version 1.8) to remove low-quality reads, clip the reads to 100 bp, and remove adapter sequences. Sequences with a minimal length of 17 bp after adapter trimming were retained to ensure high quality reads (cutadapt -q 30 -u -51 -U -51 -n1 --minimum-length 17 -a AGATCGGAAGAGCACACGTCTGAAC -A GATCGTCGGACTGTAGAACTCTGAA). Reads were mapped to the mouse genome mm10 using gsnap (v2014-12-23; gsnap --novelsplicing 1 --npaths 3 --format sam). Count tables were prepared using Htseq (v0.6.1p1; htseq-count --minaqual 0 --format bam --order pos --stranded yes). For miRNA count tables, miRbase v20 annotation was used. Gene annotation for snoRNA, snRNA, and miscRNA was retrieved from Biomart (Ensembl 84). tRNA and rRNA chromosomal positions were retrieved from the UCSC mm10 repeat mask track. Reproducibility of triplicate experiments (performed on different days) was assessed by evaluating Pearson correlation values (after logarithmic transformation of the small RNA counts).
Differential abundance was determined separately for each ncRNA class, and separately for EV and cells, using the edgeR package in R . Data were normalized using the TMM method (weighted trimmed mean of M values). Estimating the common, trended (overexpression values), and tagwise dispersion, a generalized linear model, containing the effect of cell stimulation and the effect of the day-to-day variation between triplicate experiments, was fit. The log-likelihood ratio test was used to evaluate differential expression between treatments relative to control. p values were adjusted for multiple testing using Benjamini and Hochberg’s false discovery rate (FDR). Average fold-change over three independent experiments and standard deviation were plotted. Analysis of RNA fragments was done using the UCSC genome browser and Integrated Genome Viewer .
Quantitative real-time PCR
cDNA was generated from cellular or EV-derived small RNA using the miScript RT2 kit (Qiagen, Hilden, Germany). An equivalent of 20 pg RNA was used per qPCR reaction and mixed with 100 nM primers (Isogen Life Sciences, De Meern, The Netherlands) and 4 µl SYBR Green Sensimix (Bioline Reagents Ltd., United Kingdom) in an 8 µl reaction. No-RT-controls confirmed the absence of genomic DNA and non-specific amplification.
Cycling conditions were 95 °C for 10 min followed by 50 cycles of 95 °C for 10 s, 57 °C for 30 s, and 72 °C for 20 s. All PCR reactions were performed on the Bio-Rad iQ5 Multicolor Real-Time PCR Detection System (Bio-Rad, Hercules, CA). Quantification cycle (Cq) values were determined using Bio-Rad CFX software using automatic baseline settings. Thresholds were set in the linear phase of the amplification curve.
High-resolution flow cytometric analysis of EV
High-resolution flow cytometric analysis of PKH67-labeled EV was performed using a BD Influx flow cytometer (BD Biosciences, San Jose, CA) with an optimized configuration, as previously described [49, 52]. In brief, we applied threshold triggering on fluorescence derived from PKH67-labeled EV passing the first laser. Forward scatter (FSC) was detected with a collection angle of 15°–25° (reduced wide-angle FSC). Fluorescent 100- and 200-nm polystyrene beads (FluoSpheres, Invitrogen, Carlsbad, CA) were used to calibrate the fluorescence and rw-FSC settings. Sucrose gradient fractions containing PKH67-labeled EV were diluted 25× in PBS and vortexed just before measurement. This dilution factor was sufficient to avoid ‘coincidence’ (multiple EV arriving at the measuring spot at the same time), thereby allowing accurate quantitative comparison of EV numbers in different conditions. Moreover, samples were measured at maximally 10,000 events per second, which is far below the limit in the electronic pulse processing speed of the BD Influx .
Cell pellets were lyzed in PBS + 1% Nonidet-P40 with protein inhibitor cocktail (Roche, Basel, Switzerland) for 15 min on ice. Nuclei were spun down at 16,000 g for 15 min at 4 °C, supernatant was used for Western blotting. Cell lysates and EV were denatured in SDS-sample buffer at 100 °C for 3 min, and separated using 12% SDS-PAGE gels, after which proteins were transferred onto Immobilon-P 0.45 μm PVDF membranes (Millipore, Cork, Ireland). After blocking for 1–2 h in blocking buffer (0.5% Cold Fish Skin Gelatin (Sigma-Aldrich, St. Louis, CA) in PBS + 0.05% Tween-20), blots were incubated overnight at 4 °C with primary antibodies [anti-mouse-CD9 (eBioscience, clone KMC8, 1: 1000), anti-mouse-CD63 (MBL, clone D263-3, 1:1000), anti-mouse–galectin-3 (eBioscience, clone M3/38, 1:500), anti-MHCII-p55 (GenScript, Piscataway, NJ, custom Ab raised against MHCII bèta chain peptide sequence RSQKGPRGPPPAGLLQC, 1:5000), or anti-mouse-beta-actin (ThermoScientific, polyclonal PA1-16889, 1:5000)] in blocking buffer, and washed and incubated for 1–2 h with HRP-coupled secondary antibodies (Dako, cat P0450 and P0448, 1:5000). ECL solution (ThermoScientific, SuperSignal West Dura Extended Duration Substrate, cat. 34075) was used for detection on a Chemidoc imager (Bio-Rad, Hercules, CA). Images were analyzed by the Image Lab software (Bio-Rad, Hercules, CA).
Nanoparticle tracking analysis (NTA)
EV purified by density gradient ultracentrifugation as described above were resuspended in 50 µl PBS + 0.4% EV-depleted BSA. Samples were diluted 20 times in PBS (confirmed to be particle-free when analyzed with the same settings as used for the EV samples) before measurement on a Nanosight NS500 instrument (Malvern, Worchestershire, UK). Data acquisition and processing were performed using NTA software 3.1. Each sample was recorded for 5 times 30 s, 25 frames per second, camera level 14, detection threshold 5.
EV and RNP were separated by density gradient centrifugation as described above, and resuspended in 15 µl PBS. 3 µl aliquots of either EV or RNP were added on top of carbon-coated 300 mesh, copper grids, and incubated for 2 min at room temperature. The grids were then washed two times in 50 µl PBS and stained with 2% uranyl acetate. Imaging was performed on a T20 electron microscope (FEI) operated at 200 keV. Images were recorded on a CCD Eagle camera (FEI).
Cellular or EV-derived small RNA (between 100 ng and 1 ng per lane, as indicated) was denatured in gel loading buffer II (Invitrogen) for 2 min at 70 °C and snap-cooled on ice before loading on a denaturing 15% PAGE gel (National Diagnostics, Nottingham, UK) as described previously . RNA was transferred onto a Hybond-N Nylon membrane (Amersham Pharmacia Biotech, GE Healthcare Bio-Sciences, Pasching, Austria) and chemically crosslinked with 1-ethyl-3-(3-dimethylaminopropyl) carbodimide (Sigma) at 55 °C for 2 h . 32P-labeled DNA oligo probes perfectly complementary to each RNA species were hybridized overnight at 42 °C in PerfectHyb (Sigma) solution. Blots were analyzed by phosphorimaging using a Typhoon Scanner (GE Healthcare).
List of DNA oligo probes
TGA GAA CTG AAT TCC ATG GGT
GGG TTA ATG CTA ATT GTG ATA G
GGG TCT TTG GTT ATC TAG C
TAC CCT GTA GAT CCG AAT TTG TG
TTC ACA GTG GCT AAG TTC TGC
ACT GGA CTT GGA GTC AGA AG
GTT ATC TCA ATT GAT TGT TCA CAG TC
GGC TGG TCC GAG TGC AGT GG
TTG AGA TAA CTC ACT ACC TTC GGA CCA GCC
GTC AAG TGC AGT AGT GAG AAG
CCG AAT CCT AAC CAC TAG ACC ACC AG
GCA TTG GTG GTT CAG TGG TAG AAT TCT CGC C
Differences between the numbers of EV released by LPS- and VitD3-stimulated DC versus control DC were analyzed by one-way ANOVA with Dunnett’s two-sided post hoc test. Similar statistical testing was performed on differences in RNA yield. Differences in RNA biotypes in EV and cells were analyzed by two-tailed t test. Differences in fold-changes of individual RNAs as measured in RT-qPCR were analyzed by one-way ANOVA. Significance was defined as p < 0.05. Statistical tests on data not derived from RNA sequencing were done in SPSS (v24, IBM).
Availability of data and materials
Raw data and count tables are deposited in the GEO database under accession number GSE105151. Written details on experimental procedures have been submitted to the EV-TRACK knowledgebase (EV-TRACK ID: EV170030) .
In contrast to tRNAs and snRNAs, the EV-associated levels of three other RNA types tested here, i.e., miRNAs, snoRNAs, and miscRNAs, changed in response to the different stimulate imposed on the DC. We aimed to identify the RNAs differentially present in EV released by LPS- versus VitD3-treated DC, since such RNAs may underlie distinct effects of EV on recipient cells.
MicroRNAs enriched in LPS- or VitD3-EV with known functions in DC
Function in DC
Master regulator in DC maturation
Downregulated in mature/activated DC
Inhibits DC activation and Th1/Th17 cell immune responses
Down regulates IL-12p70, IL-6, and TNF-α production by DC
Regulatory circuitry controlling monocyte activation by LPS
Repression of pro-inflammatory cytokine release by DC
Upregulated in VitD3-treated DC
Upregulated in tolerogenic DC
Upregulated in tolerogenic DC
Suppression of inflammatory cytokine production via NF-κB
Upregulated in LPS/IFNg stimulated DC
Reduces the responsiveness of DCs to TLR7/9 ligands
Suppresses NF-κB signaling
Inhibition of CD40 and MHCII expression
Suppression of inflammatory cytokine production
MiRNAs known to be involved in the pro-inflammatory function of DC were enriched in LPS-induced EV, whereas EV from VitD3-treated DC were enriched in miRNAs that dampen immune-stimulatory signaling cascades in DC. We validated three LPS-enriched and three VitD3-enriched RNAs by RT-qPCR. The use of stable reference transcripts in RT-qPCR analysis is important to ensure reliable normalization between different samples. However, the identity and general applicability of reference transcripts stably present in EV are understudied. Here, we used our sequencing data set to select four non-coding RNAs that were present in EV at comparable levels across all conditions, and confirmed their stability by RT-qPCR (Supplementary Fig. 7A–B). We normalized our qPCR data to the geometric mean of these four genes to minimize errors caused by technical and/or biological variation . Using this normalization strategy, RT-qPCR analysis confirmed the sequencing data for the tested miRNAs, although miR146-5p did not reach significance (Fig. 5b). Together, these data indicate that opposing stimuli imposed on the same type of parental cell induce the release of functionally different sets of miRNAs via EV.
Besides changes in miRNA levels, the levels of several EV-associated snoRNAs were found to differ between LPS and VitD3 conditions (Fig. 5c), with three H/ACA box snoRNAs (snora2b, snora32, and snora55) displaying FDR levels < 0.05. The overall low abundance of this RNA class in EV (Fig. 2c) hampered reliable quantification of snoRNAs in EV by RT-qPCR (data not shown). Nevertheless, our data are a first indication that EV-associated snoRNA levels change with the activation status of parental cells. Although none of the RNAs in the miscRNA group reached FDR levels < 0.05, analysis by RT-qPCR indicated that the levels of Y3-RNA (Rny3) and an additional member of the Y-RNA family, Y1-RNA (Rny1), significantly differed between LPS and VitD3 conditions (Fig. 5d, e). In cells, the conserved family of Y-RNAs is known to bind to several different proteins that play a role in RNA quality control [75, 76]. Members of the small non-coding Y-RNA family have frequently been detected in EV from multiple different cell types and different body fluids [30, 31, 32, 34, 35, 77, 78, 79, 80, 81]. We here provide evidence that Y-RNA levels in EV can be regulated by stimuli imposed on the EV-producing cell. The reduced Y-RNA levels in EV from LPS-stimulated DC compared to EV from VitD3-stimulated DC (Fig. 5e) indicate that the presence of Y-RNAs in EV may be indicative for the immune function of the parental cells. Overall, our data indicate that immune cell stimuli cause changes in the EV-associated levels of specific miRNAs, snoRNAs and Y-RNAs.
Overall, the results obtained for our primary DC cultures show that the molecular messages enclosed in EV are composed of multiple RNA classes that are specifically shuttled into EV dependent on the activation status of the cell. Moreover, the stimulus-induced changes in the small RNA content of cells and EV only partly overlap.
The data presented here broaden our view on the plasticity of the small RNA content of EV in relation to changes in the activation status of the EV-producing cell. Whereas most of the previous studies focused on the miRNA content of EV, we show that cell status-dependent changes in the RNA composition of EV also extend to other small RNA classes, which may, therefore, also contribute to the specific genetic messages conveyed by EV to recipient cells. The immune cell model which we employed, i.e., DC that are well known for their strong and diverse functional responsiveness to external stimuli, served two purposes. First, the cell system illustrates that diverse stimuli can induce the release of EV with variable levels of multiple RNA types, and that only some of the stimulation-induced changes in EV-associated RNAs reflect changes in the cellular RNA. Second, the data extend our knowledge on the central role of DC in raising and regulating immune responses. The identification of multiple EV-RNA types that are specifically associated with either the immune-activating or immune-suppressing status of DC is the prelude to unraveling the function of these RNAs in EV.
So far, RNA sequencing studies in which multiple RNA classes were identified in EV employed EV isolation methods known to co-isolate extracellular protein–RNA complexes (RNP) that overlap in size with EV, e.g., ultracentrifugation and precipitation-based methods [4, 24, 30, 32, 34, 80, 85, 86]. The results of our present study indicate that up to 55% of total extracellular RNA present in ultracentrifugation pellets of cell culture supernatant can be associated with RNPs. Buoyant density-based separation of EV from RNP, as employed in this study, allowed us to specifically identify RNA species that cells differentially sort into EV upon exogenous stimulation. In RNA sequencing libraries prepared from these highly pure EV, we observed that the percentage of alignment to the mouse genome was lower than in cellular RNA libraries. This is likely caused by the (ultra)low-input quantity of RNA in our EV-sequencing libraries, which amplifies the contribution of lab-derived contaminant RNAs, e.g., from commercial nucleic acid extraction kits or sample cross-contamination [87, 88]. The percentage of alignment may also be dependent on the size range of RNAs selected for sequencing. Whereas, in most published EV-RNA sequencing studies, RNAs in the 20–40 nt size range were sequenced, thereby strongly selecting for microRNAs, we here sequenced RNAs of a much larger size range (20–300 nt). We previously showed that several of these larger and highly modified/structured RNAs, for which the sequencing efficiency may be compromised, are enriched in EV compared to cellular RNA , which may have caused the percentages of alignment to be lower in EV compared to cells. For our studies, we selected a commercial adapter-ligation-based method that has been frequently used in EV-RNA sequencing studies [23, 24, 31, 32, 61]. While optimized for miRNA analysis, it should be noted that these library preparation methods show bias in the efficiency with which sequences with base modifications, terminal triphosphate groups, and strong secondary structures are ligated and sequenced . As an example, we here provide Northern blot-based evidence that the frequently reported predominance of Y-RNA fragments over full-length Y-RNA in EV [30, 32, 35, 90] represents a sequencing artefact. We observed a similar discrepancy between sequencing read coverage and Northern blot analysis for tRNAs (Supplementary Fig. 8). While the triphosphate group on the 5′-end  or their strong secondary structure may hamper efficient sequencing of full-length Y-RNA, difficulties with sequencing of full-length tRNAs are likely due to the inability of the reverse transcriptase enzyme used during sequencing library preparation to read through the highly modified tRNA structure . This is corroborated by recent sequencing studies deploying reverse transcriptases that are insensitive to secondary RNA structures [93, 94]. Together, these findings urge caution in the interpretation of RNA fragmentation based on RNA sequencing data alone. Moreover, we reduced the effect of sequencing biases on the assessment of quantitative differences in EV-RNA content by determining the fold changes in identical transcripts between different conditions.
Our experimental approach allowed comprehensive analysis of changes in EV-RNA classes induced by diverse immune-relevant stimuli. The most prominent stimulation-induced changes in EV-associated RNA levels were found for miRNA, Y-RNA, and snoRNAs (Fig. 3). Although detected in EV from multiple sources [23, 31, 33, 34, 35], the presence of snoRNAs in EV has been largely understudied. Here, we provide evidence that the levels of EV-associated snoRNAs can be regulated by stimuli imposed on the EV-producing cell. Since changes in the cellular levels of snoRNA have been associated with diseases [36, 95, 96], our present data point to EV-associated snoRNAs as potential functional entities in EV and interesting candidate biomarkers indicative of the cellular activation status.
Numerous studies already demonstrated that miRNA levels in EV change upon disease induction or cell stimulation [22, 23, 24]. Unique aspects of our present RNAseq study are the use of two different stimuli for comparative analysis of RNA levels in highly purified EV and the parallel assessment of cellular and EV-associated RNA. The enrichment of endotoxin-responsive miRNAs, such as miR-155 and miR-146a [20, 97, 98], that we observed in EV from LPS-stimulated DC concurs with previously published semi-quantitative microarray-based data on the miRNA content of EV from stimulated DC [22, 26]. Both miR-155 and miR-146a are known to be endotoxin-responsive, but miR-155 has an immune-activating role, while miR-146a is involved in dampening of immune responses and could act as a molecular brake on inflammation . It is thought that the coordinated action of these two is important in regulation of immune response . Differences in the miR-155/miR-146a ratios that we observed in LPS-EV and VitD3 EV (Fig. 5b) may, therefore, lead to different effects of these EV on immune activation. In addition, we here demonstrated that a different (i.e., immune downregulatory) type of stimulus imposed on DC led to both a reduction of EV-associated levels of these immune-activating miRNAs and an enrichment in miRNAs implicated in the suppression of immune responses via modulation of multiple pathways. Examples include dampening of TLR signaling by miR-27a and miR-126a [70, 73], and downregulation of (regulators of) PI3K/Akt and NF-κB by miR-378 , miR-708  and miR-27b-3p , processes that have been implicated in the suppression of pro-inflammatory cytokine release. Such differences in the EV-miRNA content may therefore contribute to immune-activating or immune-suppressive functions of differentially stimulated DC.
Not only immune-suppressive miRNAs but also two types of Y-RNA were found to be specifically excluded from EV released by LPS-stimulated DC (Fig. 5e). This raises the question whether Y-RNAs contribute to immune-related functions of EV. Until now, different types of Y-RNA (mY1 and mY3 in mice and hY1, hY3, hY4, and hY5 in humans) have been detected in EV from a multitude of different cell types and in different body fluids [30, 31, 32, 34, 35, 77, 78, 79, 80, 81]. The extracellular presence of Y4- and Y5-RNA and fragments thereof has mostly been associated to cancer [32, 81, 84, 102] and coronary artery disease [78, 103]. Immune-related functions of Y-RNAs include their capacity to trigger TLR signaling. Interestingly, Y-RNAs may vary in their specificity for different TLRs, with Y3-RNA triggering predominantly TLR3, while Y1-, Y3-, and Y4-RNA trigger TLR7 . On the contrary, the reduction of Y1- and Y3-RNA levels in EV released by immune-stimulatory DC, shown in this study, suggests an immune downregulatory role for these EV-associated Y-RNAs. In line with this, high levels of Y-RNAs have been reported in immune-suppressive EV released by the parasite Heligmosomoides polygyrus and in seminal fluid EV for which immunosuppressive effects have been described [35, 105, 106, 107]. Overall, the observed variable presence of Y-RNA in EV from immune cells has raised further interest in unraveling the immune-related function and biomarker potential of these RNAs in EV.
With regard to the biomarker potential of EV-associated RNAs, our data also urge caution in interpreting EV as snapshots of the cell from which they arise. We found that only a subset of changes in EV-RNA reflected changes that occurred at the cellular level. The highest proportion of reflective RNAs was observed within the category of miRNAs and included prominent immune-related RNAs such as miR-155, miR-9, and miR-10-5p (Fig. 7h). For a substantial number of EV-associated miRNAs and both Y-RNA types, however, stimulation-induced changes were observed in EV, but not in cells. This could suggest that EV-levels of reflective RNAs are mainly regulated by transcription, whereas levels of EV-responsive RNAs are mainly determined by shuttling rate. For RNAs known to primarily reside in the nucleus (such as snRNAs and snoRNAs), subtle changes in the cytoplasmic pool of these RNAs may be overshadowed by the much larger nuclear pool. Comparison of EV-associated levels with cytoplasmic instead of total cellular RNA levels may, therefore, provide a better insight in shuttling of these nuclear RNAs into EV. Interestingly, we also observed that many of the stimulation-induced changes in cellular RNA levels were not reflected in the RNA levels of EV released by these cells. This strengthens the hypothesis that cells release EV with selective sets of RNAs into the extracellular milieu. From a biomarker perspective, these data implicate that, although the presence of specific EV-RNAs may correlate with disease, the EV transcriptome is not necessarily predictive for RNA levels in the EV-producing cell.
We observed two classes of RNA, i.e., snRNA and tRNA, of which the EV-associated levels remained stable after differential stimulation of DC. These RNA classes have been commonly detected in small RNAseq studies of EV from multiple cell types and body fluids, but factors impacting their incorporation into EV had until now not been investigated. Although these data need confirmation in different cell types with various external stimuli, our data may be a first indication that tRNAs and snRNAs are constitutive components of EV. Speculatively, these RNAs may be required for basal scaffolding functions in EV biogenesis, or may aid the functional transfer of gene-regulatory RNAs. Based on their stable association with EV, these types of RNAs also have potential to be used as EV-reference RNAs for RT-qPCR normalization, since comparing the EV transcriptome of differentially stimulated cells is complicated by potential differences in the number of EV, RNA yield, and RNA composition. Importantly, the use of well-known stably expressed cellular RNAs is not necessarily valid for normalization of EV qPCR data due to potential differences in RNA shuttling between tested conditions. Therefore, normalization of RT-qPCR data to reference genes which are stable in EV is a reliable way to quantitate differences in individual RNAs. Here, we show how RNA sequencing and RT-qPCR validation can be used to identify candidate EV-reference genes which are constitutively expressed between differently stimulated cells.
In conclusion, this study indicates that external stimuli imposed on cells can lead to changes in the levels of some EV-associated RNA classes, while leaving others unchanged. In our DC-model, two different immune-related cell stimuli led to disparate changes in the type and quantity of miRNAs, snoRNAs, and Y-RNAs in EV. Only part of the changes in RNA contents of EV reflected changes in the cellular RNA, which urges caution in interpreting EV as snapshots of cells. These findings show that differences in multiple structurally distinct RNA classes shape the EV transcriptome. These RNA types may contribute to the functional properties of EV in cellular communication, and may be further explored as candidate EV-RNA-based indicators for the immune status of the EV-producing cell.
The authors would like to thank Dr. G. J. A. Arkesteijn for the assistance with high-resolution flow cytometric analysis of EV, Dr. Z. Zeev Ben Mordehai for the assistance with transmission electron microscopy, and Drs. S. L. Wolin, S. Sim, and D. Staněk for the valuable discussions. This work was supported by European Research Council under the European Union’s Seventh Framework Programme [FP/2007-2013]/ERC Grant Agreement number  to [ENMNtH]; COST Action BM1202 Microvesicles and Exosomes in Health and Disease Short-Term Scientific Mission Grant [191116-083388] to [TAPD]; and the Human Frontiers Science Program [RGY0069] to [AHB].
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest.
- 1.Colombo M, Raposo G, Théry C (2014) Biogenesis, secretion, and intercellular interactions of exosomes and other extracellular vesicles. Annu Rev Cell Dev Biol 30:255–289. https://doi.org/10.1146/annurev-cellbio-101512-122326 CrossRefGoogle Scholar
- 11.Kahlert C, Melo SA, Protopopov A et al (2014) Identification of doublestranded genomic dna spanning all chromosomes with mutated KRAS and P53 DNA in the serum exosomes of patients with pancreatic cancer. J Biol Chem 289:3869–3875. https://doi.org/10.1074/jbc.C113.532267 CrossRefPubMedPubMedCentralGoogle Scholar
- 17.Lötvall J (2014) Minimal experimental requirements for definition of extracellular vesicles and their functions: a position statement from the. Int Soc Extracell Vesicles. 1:1–6Google Scholar
- 30.Nolte-’t Hoen ENM, Buermans HPJ, Waasdorp M et al (2012) Deep sequencing of RNA from immune cell-derived vesicles uncovers the selective incorporation of small non-coding RNA biotypes with potential regulatory functions. Nucleic Acids Res 40:9272–9285. https://doi.org/10.1093/nar/gks658 CrossRefPubMedPubMedCentralGoogle Scholar
- 31.van Balkom BWM, Eisele AS, Pegtel DM et al (2015) Quantitative and qualitative analysis of small RNAs in human endothelial cells and exosomes provides insights into localized RNA processing, degradation and sorting. J Extracell Vesicles 4:26760. https://doi.org/10.3402/jev.v4.26760 CrossRefPubMedGoogle Scholar
- 77.Dhahbi JM, Spindler SR, Atamna H et al (2013) 5′-YRNA fragments derived by processing of transcripts from specific YRNA genes and pseudogenes are abundant in human serum and plasma. Physiol Genom 45:990–998. https://doi.org/10.1152/physiolgenomics.00129.2013 CrossRefGoogle Scholar
- 91.Hendrick JP, Wolin SL, Rinke J et al (1981) Ro small cytoplasmic ribonucleoproteins are a subclass of La ribonucleoproteins: further characterization of the Ro and La small ribonucleoproteins from uninfected mammalian cells. Mol Cell Biol 1:1138–1149. https://doi.org/10.1128/MCB.1.12.1138.Updated CrossRefPubMedPubMedCentralGoogle Scholar
- 100.Ruckerl D, Jenkins SJ, Laqtom NN et al (2012) Induction of IL-4R alpha-dependent microRNAs identifies PI3K/Akt signaling as essential for IL-4-driven murine macrophage proliferation in vivo. Blood 120:2307–2317. https://doi.org/10.1182/blood-2012-02-408252.The CrossRefPubMedPubMedCentralGoogle Scholar
- 101.Lee J-J, Drakaki A, Illiopoulos D, Struhl K (2012) MiR-27b targets PPARy to inhibit growth, tumor progression, and the inflammatory response in neuroblastoma cells. Oncogene 31:3818–3825. https://doi.org/10.1146/annurev-immunol-032713-120240.Microglia CrossRefPubMedGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.