Analysis of extracellular vesicle miRNA profiles in heart failure

Abstract Extracellular vesicles (EVs) have recently emerged as an important carrier for various genetic materials including microRNAs (miRs). Growing evidences suggested that several miRs transported by EVs were particularly involved in modulating cardiac function. However, it has remained unclear what miRs are enriched in EVs and play an important role in the pathological condition. Therefore, we established the miR expression profiles in EVs from murine normal and failing hearts and consecutively identified substantially altered miRs. In addition, we have performed bioinformatics approach to predict potential cardiac outcomes through the identification of miR targets. Conclusively, we observed approximately 63% of predicted targets were validated with previous reports. Notably, the predicted targets by this approach were often involved in both beneficial and malicious signalling pathways, which may reflect heterogeneous cellular origins of EVs in tissues. Lastly, there has been an active debate on U6 whether it is a proper control. Through further analysis of EV miR profiles, miR‐676 was identified as a superior reference control due to its consistent and abundant expressions. In summary, our results contribute to identifying specific EV miRs for the potential therapeutic targets in heart failure and suggest that miR‐676 as a new reference control for the EV miR studies.

microRNAs (miRs) and proteins are transported by EVs. Therefore, EVs are being re-evaluated as a critical modifier of cellular functions. 5 During last decade, tremendous efforts have been placed to understand the mechanism and the function of EVs in diverse fields of studies including cardiovascular researches. For example, EVs derived from stem cells or cardiac progenitor cells were generally beneficial, [6][7][8][9] whereas EVs from fibroblasts or endothelial cells exhibited deleterious effects on cardiac functions. [10][11][12] In these previous studies, the origin of EVs was confined to in vitro cultured cells or blood plasma, but the role of EVs in cardiac tissues has been poorly investigated. Therefore, it is important to characterize in tissue EVs under pathological conditions. Recently, Loyer X. et al characterized the functions of EVs isolated from mouse hearts subjected to myocardial infarction. 13 In this study, we further analysed the expression profiles and functions of miRNAs in EVs isolated from pressure overload-induced failing hearts compared to normal hearts.
Apart from the characterization study, we also sought to identify an alternative internal reference for miR studies. In general, U6 is the most commonly used control for the normalization of miR expression. However, it is yet controversial whether U6 is a suitable control because variable expression patterns of U6 were observed in a number of studies. [14][15][16][17][18][19] Analysing the abundancy and stability of miR expression, U6 was confirmed as an appropriate control being consistently expressed throughout our samples. In addition, we identified miR-676 as an alternative candidate as a control and observed that it exhibited a significant correlation with U6. Taken together, although a further validation is necessary in other tissues or organs, miR-676 may serve as an alternative control for the EV miR studies in the heart.
In summary, this study analysed miR profiles in the cardiac EVs isolated from murine normal and failing hearts. Our results demonstrated that the cardiac EVs contain diverse miRs with contradictory functions. The results may reflect various cellular origins of EVs and dynamic alternations of miR expressions in these EVs during heart failure. Nevertheless, our study may contribute to understand intercellular communication between cardiac cells under the pathological condition and suggests EV miRs as potential biomarkers and therapeutic targets for the intervention of heart failure. Studies were conducted in male C57BL/6 mice aged 8 ~ 10 weeks (weight, 25 ~ 30 g) obtained from Jackson Laboratories. Transverse aortic constriction (TAC) was conducted as previously described with minor modifications. 10 Briefly, mice were anesthetized with a solution mixture of 95 mg/kg ketamine and 5 mg/kg xylazine administered via intraperitoneal injection. The transverse aortic arch was ligated between the innominate and left common carotid arteries with an overlaid 27-gauge needle. The needle was then immediately removed, leaving a discrete region of constriction. Two month after TAC operation, the mice showed less than 40% of fractional shortening in average which is typically considered as failing heart and included in this study ( Figure S3). Sham-operated mice underwent the same surgical procedures, except that the ligature was not tied.

| EV isolation and purification
To isolate in tissue cardiac EVs from the heart tissue, Langendorffbased isolation was applied as utilizing previous protocol to isolate in tissue EV from liver. 20 Hearts were extracted from sham or TACoperated animals, and the aorta was retrogradely perfused at 37°C After 20 minutes of digestion, the heart tissue was removed, cut into several chunks and gently pipetted for 2 minutes in Tyrode buffer with 5% BSA. The mixtures of cell, extracellular matrix and EVs were then separated by several centrifugations according to a previously described protocol. 21 EVs were pelleted by ultracentrifugation at 100 000 × g at 4°C for 90 minutes and resuspended in 200 μL of PBS solution.

| Transmission Electron microscopy for EVs with negative staining
EVs from fresh ventricles were collected as described above and pre-fixed with 2% paraformaldehyde (PFA) for 30 minutes. Deposit 5 μL resuspended pellets on Formvar-carbon-coated EM grid. Add 10 μL of 2% uranyl acetate on the grid and incubate for another 10 minutes. After thorough washes, store grid in the grid box at room temperature and dark until imaging. Samples were viewed under a transmission electron microscope (HITACHI H-7650, Japan) operated at 80 kV. Images were taken at 10K, 20K and 30K-folds magnification. 50 μL of EV samples from Sham and TAC EV was directly mixed with 5x SDS sample buffer, boiled at 95°C for 3 minutes and then separated on a SDS-PAGE gel. The total loaded protein amount was calculated by the intensity of Coomassie blue stain. The membrane was blocked in 5% skim milk solution and incubated overnight with an antibody directed against hnRNA A2B1 (Ab6102, Abcam, 1:5000), CD63 (Ab1318, Abcam, 1:1000), Flotillin 1 (Ab133497, Abcam, 1:10 000), TSG101 (Ab83, Abcam, 1:1000), H3 (Ab1791, Abcam, 1:1000) and SERCA2a (custom antibody from 21st Century Biochemicals, 1:3000). The membrane was then incubated with a horseradish peroxidase-conjugated secondary antibody (Sigma) and developed with Western Lighting chemiluminescence reagent (PerkinElmer).

| Nanoparticle tracking analysis
Samples were loaded into the assembled sample chamber of a NanoSight LM10. 60-second video images were acquired by a Hamamatsu C11440 ORCA-Flash 2.8 digital camera and analysed by NanoSight NTA 2.3 software.

| Extracellular vesicle microRNA isolation and microRNA array profiling
100 μL of EV samples from Sham and TAC EV was used to extract RNA for microRNA array profiling and validation using qRT-PCR. Total EV RNA was isolated with mirVana miRNA Isolation Kit (Ambion). The RNA samples were sent, and the microarray procedure was carried out at Exiqon Vedbaek, Denmark, as followed the previous description. 22 Briefly, the samples were labelled using the miRCURY LNA™ microRNA Hi-Power Labeling Kit, Hy3™/Hy5™ and hybridized on the miRCURY LNA™ microRNA Array (7th Gen).
Bioinformatics analyses (Gene Ontology) were performed using the UniProt database and Human Protein Reference Database.23 Protein-protein interactions analyses by STRING 11.0 database with medium confidence STRING annotated interactions (STRING score > 0.4) for the bait proteins. 24 Cytoscape software (Ver 3.6.1; http://www.cytos cape.org/) was used to visualize the network.

| Statistical analysis
Where appropriate, the data are expressed as means ± s.e.m.
Comparisons of the group means were made by using Student's t test or one-way ANOVA with a Bonferroni post-test analysis. Correlation test was performed with the Pearson's correlation coefficient measure. A P-value of <.05 was considered to be statistically significant.

| Cardiac EVs from normal and failing hearts have distinctive physical and molecular signatures
Pressure overload-induced heart failure was generated in murine hearts by applying transverse aortic constriction (TAC). The cardiac EVs were isolated from normal (Sham EVs) and failing (TAC EVs) hearts using the Langendorff-based isolation followed by purification through multiple rounds of ultracentrifugations ( Figure 1A).
The isolated EVs were negatively stained and visualized by electron microscopy with 30 000-fold magnification. Morphological analysis revealed that EVs showed a typical cup-shaped structure with a diameter of 70 ~ 160 nm and that TAC EVs tended to be larger than Sham EVs ( Figure 1B). Nanoparticle tracking analysis (NTA) was performed to precisely measure the concentration and sizes of EVs. Failing hearts secrete ~5 × 10 9 EVs per heart, whereas normal hearts release ~2 × 10 9 EVs per heart ( Figure 1C). Average size of TAC EVs was significantly larger than that of Sham EVs, which is consistent with the electron microscopic data ( Figure 1D  Furthermore, Sham and TAC EVs exhibited a significant difference in their protein contents. Same volumes of EV samples (50 μL, a quarter of the total EVs from a heart) were loaded in a SDS-PAGE gel and quantified by Coomassie blue staining. In line with NTA results, 50% more proteins were detected in TAC EVs ( Figure 1E and F). As shown in Figure 1F, several EV markers were analysed by Western bolt analysis. The Flotillin level was not altered between Sham and TAC EVs, but hnRNP A2B1 expression was significantly increased in TAC EVs. On the other hand, TSG101 and CD63 expressions were substantially reduced in TAC EVs ( Figure 1F). Taken together, the cardiac EVs from normal and failing hearts showed distinctive features in their physical and molecular properties.

| microRNA profiles of cardiac EVs and validation
To further characterize cardiac EVs, microRNA (miR) expression profiles were explored. Cardiac EVs were isolated from three normal and three failing hearts. All EV preparations were immediately subjected to quality control and miRNA array analyses that were conducted using the 7th generation of miRCURY™ LNA array system F I G U R E 1 Characterization of in tissue cardiac EVs. A, Schematic procedure for the EV isolation. B, Representative images of electron microscopy. The bar indicates 100 nm. Arrows indicate EVs. C, Average densities and sizes of Sham and TAC EV as analysed by NTA. n = 6-7. D, Ensemble averages from several identical size distributions of Sham (n = 7) and TAC EV (n = 6). Upper: Absolute particle numbers per 1 mL, Lower: Relative concentration normalized to total nanoparticle concentrations. Arrow heads and arrows indicate the primary peak and the second or third peak, respectively. E, Representative images of Western blot and (F) quantified data. Sham EV, TAC EV and NTA indicate EVs from normal hearts, EVs from failing hearts and nanoparticle tracking analysis, respectively. *P < .05, **P < .01, ***P < .001 versus Sham EV, as determined by Student's t test. Data are presented as mean ± s.e.m F I G U R E 2 microRNA array results of tissue cardiac EVs. A, A scheme of entire procedure to identify differentially expressed microRNAs. B, Heat map and hierarchical clustering. The clustering was performed on the top 50 miRs with highest standard deviation. C, Volcano plot depicting the fold changes (X-axis) and P-value (Y-axis) in miR expression levels between Sham and TAC EV. Coloured points refer significantly up-regulated (Red), down-regulated (Blue) or unchanged (Green) microRNAs F I G U R E 3 Validation of microRNA expression profiles by qRT-PCR. The levels of the up-regulated (A) or downregulated (B) miRs were quantified by qRT-PCR. The relative levels of each microRNA were normalized to the U6 snRNA. n = 4. *P < .05, **P < .01, ***P < .001 versus Sham EV, as determined by one-way ANOVA. Data are presented as mean ± s.e.m (Exiqon). Of the 1,195 miRs displayed on the array, 731 miRs were eliminated due to a lack of significant signal intensity. The remaining 464 miRs were then sorted out by criteria indicated in Figure 2A ( Table S1). The heat map diagram demonstrated a clear separation of miR expressions in Sham and TAC EVs ( Figure 2B). Compared to Sham EVs, 107 miRs were up-regulated (23%) and 47 miRs were down-regulated (10%) in TAC EVs ( Figure 2C). miRs with p-values higher than 0.05 were considered as 'not altered' (67%). The 154 miRs differentially expressed on the array were further validated for their abundancy using qRT-PCR. Consequently, we identified seven up-regulated (miR-378a, miR-665, miR-139, miR-345, miR-328, miR-767 and miR-92b) and three down-regulated (miR-99b, miR-124 and miR-411) miRs ( Figure 3A and B).

| Potential target prediction for the identified miRs
To investigate the roles of the identified miRs under pathological conditions, bioinformatics analyses were performed as shown in Figure 4A ( Tables 1 and 2). First, putative targets listed in miRTar-Base (http://mirta rbase.mbc.nctu.edu.tw) as validated by both reporter assay and Western blot analysis were selected. Second, the selected targets were further sorted out according to their tissue distribution as reported in the proteomics database (www.prote omics db.org). Through these two subsequent analyses, 29 putative targets associated with the seven up-regulated miRs and 49 putative targets associated with the three down-regulated miRs were identified. Lastly, these putative targets were categorized based on their subcellular localization, biological process and protein-protein interactions (Tables 3 and 4). The analysis of subcellular localization revealed that miRs of TAC EVs appeared to target proteins predominantly located in the plasma membrane, nucleus and cytoskeleton ( Figure 4B). Functional analysis suggested that target proteins for the up-regulated miRs were involved in cell cycle (19%), signal transduction (15%) and apoptosis (15%). On the other hand, intracellular targets for the down-regulated miRs were predicted to contribute to transcription/translation (22%), apoptosis (18%) and cell cycle (14%) (Figure 4 and D). The protein-protein interaction map generated by the STRING database (https://strin g-db.org/) is shown in Figure 5. The most frequently matched targets controlled by the up-regulated miRs in TAC EVs were NOTCH, CD44 and PTEN. Meanwhile, targets modulated by the down-regulated miRs were CDH2, GRB2 and ITGB1. Taken together, these results suggest that miRs in TAC EVs are indeed closely related to the essential biological processes.

F I G U R E 4
Proteomics analysis of the putative microRNA target genes. A, Analyses were conducted based on miRNA database (miRTarBase; http://mirta rbase.mbc.nctu.edu.tw) and protein database (ProteomicsDB; www.prote omics db.org). Target proteins were displayed by their subcellular localization (B) and Gene Ontology (GO) categories (C and D) using the STRING database (https://strin g-db. org/)

| Identification of an alternative control for miR expression
There is no clear consensus on suitable controls for relative miR quantification in circulating miRs, although U6 has been considered as an endogenous normalization control. [14][15][16][17][18][19] In this study, we tested whether U6 is an adequate control for the miR expression.
U6 expression was shown to be abundant, stable and not significantly different across all 12 EV samples ( Figure 6A). Therefore, U6 expression was used to normalize the qRT-PCR data presented in Figure 3. To further support and strengthen our results, we searched out potential candidates for more reliable controls that showed minimal differences in expressions between groups yet with strong signal intensity. Initial screening led us to identify five miRs, miR-676, miR-300, miR-204, miR-453 and miR-3085.
Those were thoroughly tested and validated by qRT-PCR with multiple primer sets, and finally, miR-676 was selected as a new potential candidate for a EV miR expression control. As shown in Figure 6A, miR-676 indeed exhibited the most invariable expressions throughout all samples, which is comparable to U6. ( Figure   B and C). In fact, in regard to Ct values obtained by the Mann-Whitney test between Sham and TAC EVs, miR-676 expression (P = .52) was more consistent than U6 expression (P = .11). These data suggest that miR-676 can be used as an internal control for EV miRs with high fidelity.

| D ISCUSS I ON
As EVs are highlighted as an important mediator of cell-to-cell communications during heart failure, it becomes essential to analyse the contents of cardiac EVs. In the present study, we isolated and characterized cardiac EVs from normal and failing mouse hearts. In the physical and molecular analyses, TAC EVs showed distinctive features in terms of density, size, protein expression patterns and miR contents as compared to those from Sham EVs (Figure 1). EVs are bound to cell surface proteins, extracellular matrix (ECM) molecules and components of the blood plasma through its surface molecules, and located in the niche of ECM or plasma. 25  information was highly limited since most of the studies were conducted using blood plasma. 26,27 Recently, Loyer X. et al investigated cardiac EVs in the model of myocardial infarction. 13 In their report, they minced ventricles to obtain cardiac EVs, which might damage cardiac cells and increase potential contaminants from other intracellular components. In our study, on a contrary, Langendorff-based EV isolation was employed to minimize damaging cells ( Figure 1A). In addition, we have also compared our data with a previous study about the circulating miRs in a canine heart failure model. 27 However, only thirteen miRs were confirmed to be identical out of fifty seven miRs reported in this paper (Table S1). This weak correlation suggests distinctive miR profiles of EVs in the ECM and plasma, which is probably due to different cellular origins of EVs. Plasma EVs could be released from the other organ or tissues, but EVs used for our study is strictly from cardiac tissues. In a similar context, we found the different size distribution and protein profiles between Sham and TAC EVs ( Figure 1D-F size of exosomes were changed in remote ischaemic preconditioning-mediated cardioprotection, 28 which is correlated with our morphological observations. Taken together, these differences are also presumably related to the dynamic changes in the cellular composition of the heart under pathological conditions. In addition, we also observed the differences in the expression of exosomal markers. For example, CD63 and TSG101 were decreased in TAC EVs and these markers were previously reported to be involved in cardiac fibrosis and hypertrophy, respectively. 29,30 On the other hand, hnRNP A2B1 expression was increased and this was known to control the sorting of miRs into exosomes through specific binding motif. 31 Although we particularly focused on miR profiles in EVs, this result also suggests a possibility that proteins in EVs could actively regulate cardiac intercellular communication.
From EV miR profiles, we have identified seven up-regulated and three down-regulated miRs in failing hearts compared to con-

F I G U R E 5
In silico protein-protein interaction map for putative microRNA targets. The protein-protein maps were constructed using STRING 11.0 database and Cytoscape. The putative molecular targets by up-regulated microRNAs (A) and down-regulated microRNAs (B). The iBAQ intensity represents the expression level of proteins in the heart of down-regulated miRs, miR-124, is also known to induce cardiac hypertrophy, 38 attenuate cardiac function and stimulate angiogenesis. 39 Therefore, EV miRs released from failing hearts may play important roles in cardioprotection (Table 5). On the other hand, a negative-inotropic miR, miR-328, was found to be increased and an anti-fibrotic miR, miR-411, was decreased. In addition, we evaluated EV miR-146a, a negative-inotropic miR, that we previously reported, 10 as a positive control and it was also significantly increased ( Figure S1). Therefore, even though anti-hypertrophic and anti-fi- Furthermore, bioinformatics was applied as an alternative approach to predict the impact of EV miRs on cardiac function.
First, we predicted the possible cardiac outcomes with previously reported targets for miRs identified in this study. Next, we performed literature studies and compared with previous results.
In brief, molecular targets experimentally validated by both a reporter assay and Western blot analysis were initially selected based on the miR database (miRTarBase) and 52 targets that are not expressed in the heart were discarded. The remaining 77 putative targets of EV miRs are summarized in Table 3 (targets for up-regulated miRs) and Initially, U6 was confirmed to be abundant and stable in our samples and used as a reference control in this study. To avoid potential critics for the usage of U6, we decided to identify alternative reference miR. Through the miRNA profile analysis, miR-676 was appeared as a promising candidate due to its superior abundance and stability.
Upon this result, additional normalization was performed using miR-676 ( Figure S2) and a significant correlation between U6 and miR-676 was confirmed. Although miR-676 is reported to be elevated in the liver in the response to a high-fat diet, 58

TA B L E 5 Previously described functions of differentially expressed miRs in TAC EV
In summary, we isolated and characterized cardiac EVs in the context of miR contents using normal and failing hearts. EVs from both hearts exhibited substantial contrasts in physical and functional features. We also found that EVs from failing hearts contains both beneficial and detrimental miRs presumably because current EV miR profiles that we reported here were generated from mixed cellular origins in failing hearts, which is a limitation of this study.
Taken together, our study shows dynamic intercellular communications via EV miRs during heart failure and it will be critical to identify the origin of beneficial EVs for a therapeutic application to treat heart failure.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
JGO, PL and DJ designed the research, performed the experiments, analysed the data and wrote the paper. REG contributed in performing EM microscopy. SS and CK served as scientific advisor. DJ revised a manuscript and data.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analysed during this study are included in this published article (and its supplementary information files).
Furthermore, the raw data sets generated during the current study are also available from the corresponding author on reasonable request.