Cell characterisation

CMSCLCs were isolated from right atrial appendage (RAA) tissues obtained from patients undergoing coronary artery bypass grafting (CABG). Donor properties and cell uses are listed in Table 1:

Table 1 Properties of CMSCLC donors and use of cells

A schematic diagram of the overall experimental design is shown in Fig. 1A. Representative images showing the morphology of CMSCLCs at 24 h, 7 days, 12 days, and after the first passage (P1), are shown in Fig. 1B. At D7 colonies were visible and after D12 cells took on a spindle-shaped, fibroblastic morphology. At passage 1 an average of 2.20 × 106 cells were obtained per donor (Fig. 1C). The mean cell doubling time (Fig. 1D), was ~ 40 h until after the fourth passage. CMSCLCs grew poorly in DMEM and were dependent on FGF2 (Supplementary Fig. 1). Flow cytometry was used to immunophenotype three separate donor lines, as shown in Fig. 1E, F. BM-MSCs and peripheral blood mononuclear cells (PBMCs) were used as positive controls. CMSCLC isolates at passage two were positive (≥ 99%) for CD44, CD105 and CD166 and negative (< 0.2%) for CD19 and CD45, as were all BM-MSCs. PBMCs were > 95% positive for CD45 and a subpopulation of PBMCs (approximately 3%) were positive for CD19, as expected. Gene expression levels of additional MSC markers are shown in Fig. 1G. CMSCLCs and BM-MSCs both met positive and negative ISCT-defined minimum criteria for MSCs [40]. We next analysed the capacity of these cells to carry out trilineage differentiation (Supplementary Fig. 2). CMSCLCs showed intracellular Oil Red O-positive droplets and calcified extracellular matrix after adipogenic and osteogenic differentiation, but BM-MSCs displayed significantly more. CMSCLCs showed little capacity for chondrogenesis. These observations were also reflected in expression of genetic markers pre/post differentiation.

Fig. 1

Comparison of cardiac and bone marrow-derived cells. (A) Schematic diagram of experimental design. (B) Images of CMSCLC morphology at 24 h, 7 days and 12 days after isolation, and 24 h after the first passage (P1). Scale bar 100 μm. (C) Number of viable cells obtained at passage 1 for n = 6 donor lines used in this study. (D) Cell doubling time plotted against passage number (average of n = 6 donors). (E) Representative histogram plots of flow cytometric analysis of CMSCLCs (green), BM-MSCs (dark grey) and unstained CMSCLCs (white). Negative markers CD19 and CD45, and positive markers CD44, CD105 and CD166 are shown. PBMCs (light grey) were used as positive controls for CD19 and CD45. (F) Quantification of flow cytometric analysis. n = 3 donors were compared to BM-MSCs by one-way ANOVA. (G) BM-MSC and CMSCLC (n = 3 donors) gene expression levels (as mRNA/GAPDH ratio) of positive and negative MSC markers. A dotted line shows the cut-off for low-expressed genes, which were considered negative. BM-MSC and CMSCLC samples were compared by two-way ANOVA with Sidak’s multiple comparison test. (H) BM-MSC and CMSCLC (n = 3 donors) gene expression of common MSC-associated paracrine factors. A dotted line shows the cut-off for low-expressed markers, which were considered negative. BM-MSC and CMSCLC samples were compared by two-way ANOVA with Sidak’s multiple comparison test. ns = not significant, ** = P ≤ 0.01, **** = P ≤ 0.0001. Pairs without annotations are also not significant (P > 0.05)

Examination of CMSCLC paracrine factors

Comparing gene expression levels of several well-known cardioprotective factors (Fig. 1H) showed that CMSCLCs expressed high levels of VEGFA, ANGPT1, IGF1, FGF2, HGF and TGFB1; equal to BM-MSCs from young healthy donors. We have validated in-house that CMSCLCs were negative for Islet-1 and NKX2.5, are NANOG positive, and express low levels of PDGFR-alpha. CMSCLCs also have low levels of p16 and SA-B-GAL at passage five [41]. Taking together CMSCLC surface markers, gene expression, colony formation ability, FGF2 dependence, and differentiation capacity these cells can be appropriately described as mesenchymal stromal cells [40].

Next, we tested whether EVs or other freely-secreted compounds were the most cardioprotective components of the CMSCLC secretome using hypoxic rat cardiomyoblast cells as a screening tool. The results (Supplementary Fig. 3A) showed that CMSCLC-conditioned medium was protective compared to basal medium (68% viability vs. 51% viability, P = 0.038). After ultracentrifugation, conditioned medium particle count was reduced by 96.3% (Supplementary Fig. 3B) and protein content was reduced by 29.9%, indicating successful EV depletion. This EV-depleted conditioned medium lacked any significant protective effects (58% viability, P = 0.423), whereas basal medium supplemented with isolated CMSCLC EVs (equalised by protein concentration) significantly protected cell viability (85.9%, P = 0.002). This demonstrates that EVs are the main protective component of the EV secretome. To measure other factors, we used an antibody array to detect 80 secreted cytokines and growth factors. Results from three CMSCLC donors are shown in Supplementary Fig. 3C-D. Proteins detected in conditioned medium included IL-8, IL-6, MCP-1/CCL2, TIMP-1 and − 2, osteoprotegerin and GRO-alpha. Eotaxin, angiogenin, IL-10 and VEGF were present at moderate concentrations and GM-CSF, CCL8, IGF-1 and FGF9 were detected in lower amounts.

Fig. 2

Extracellular vesicle isolation and characterisation. (A) Representative nanoparticle tracking analysis (NTA) size distribution plots for CMSCLC EVs (green) and BM-MSC EVs (grey). (B) Mean diameter and particle counts of n = 4 separate EV isolations and comparison by unpaired t-test. (C) Representative cryoEM images of isolated EVs. Scale bar 100 nm. A crop showing the lipid bilayer is also shown (inset). (D) Antibody-based membrane array showing human-specific EV surface markers and cargo markers. Two positive controls, a blank, and GM130 (cis-golgi marker) are also included. 50 µg total protein was added per membrane. (E) EV protein concentration (n = 4 per group) (F) Particle to protein ratio. Samples were compared by unpaired t-test. ns = not significant

Extracellular vesicle isolation and characterisation

Next, we compared isolated CMSCLC EVs (C_EVs) and BM-MSC EVs (B_EVs). Nanoparticle tracking analysis (NTA) showed single peaks for both types of EV (Fig. 2A), as typical from ultracentrifugation. Over four separate batches, the mean particle size measured by NTA was 104.7 ± 11.2 nm and mean particle count was 9.19 × 1011 ± 5.80 × 1010 particles/ml for C_EVs, which was very similar to B_EVs (Fig. 2B). Conventional TEM (Supplementary Fig. 4A) showed “cup shaped” particles of approximately 100 nm diameter for both EV isolations. CryoEM (Fig. 2C) confirmed the presence of abundant spherical 50–200 nm diameter vesicles with lipid bilayer membranes in both isolates. Vesicle diameters were measured with an average diameter of 108.1 ± 4.2 nm for C_EVs and 107.3 ± 7.1 nm for B_EVs (Supplementary Fig. 4B), in close agreement with the NTA results (Fig. 2B). We then detected EV protein markers, as shown in Fig. 2D. Tetraspanin EV surface markers CD63 and CD81 were present in both populations, as were cargo markers ALIX, Flotillin 1, ICAM, TSG101 and ANXA5. GM130, a cis-golgi marker protein, showed only a faint signal, indicating low levels of contamination with non-EV cellular components. The EV protein concentration did not differ between four separate batches of C_EVs and B_EVs (Fig. 2E). Neither did the particle/protein ratio (Fig. 2F), which was in the range of 2–8 × 1011 particles/mg protein, indicating a high purity of EVs. Since we controlled cell density and standardised medium collection and EV isolation procedures, these results indicate that CMSCLCs and BM-MSCs produce EVs at a similar rate [22]. Lastly, we used Western blot to confirm additional EV markers for C_EVs (Supplementary Fig. 4C). HSP70 was detected in EVs and whole CMSCLC lysates (WCL), GAPDH was weaker in EVs compared to WCL while CD9 was enriched in EVs compared to WCL. These data show that CMSCLC- and BM-MSC-derived EVs were successfully isolated at high purity, suitable for further experimentation.

Fig. 3

Protection of hypoxic human cardiomyocytes using CMSCLC and BM-MSC EVs (A) Experimental design showing hiPSC-CM seeding and hypoxia treatment. (B) Example images of hiPSC-CMs following 48 h normoxia or hypoxia + vehicle (Veh), or hypoxia with 67 ng/µl CMSCLC EVs (C_EVs) or BM-MSC EVs (B_EVs). Scale bar 100 μm. (C) Culture medium LDH levels after 48 h of hiPSC-CM exposure to each treatment group. Blank samples (without hiPSC-CMs) are also included. Hypoxic hiPSC-CM groups were compared by one-way ANOVA with Tukey’s multiple comparison test. *** = P ≤ 0.001, **** = P ≤ 0.0001 (D) Apoptosis protein arrays from each group (n = 2 per group). Examples of significant differences between samples are highlighted with red boxes. Positive controls are shown as blue boxes in the upper left and lower right corners. (E) Heatmap showing quantification of integrated density of high concentration apoptosis-related proteins (n = 2 per group). All groups were compared using two-way ANOVA with Tukey’s post-test. The table above the heat map describes statistical significance; 1 = P < 0.05, 2 = P ≤ 0.01, 3 = P ≤ 0.001, 4 = P ≤ 0.0001, ns = not significant (P > 0.05)

Protection of hypoxic human cardiomyocytes using CMSCLC and BM-MSC EVs

Next, we compared the ability of C_EVs and B_EVs to protect hypoxic hiPSC-CMs. LDH release was used as a sensitive metric to measure hiPSC-CM damage at two time points [36]. The experimental design is shown in Fig. 3A. 48 h hypoxia was used as an injury model based on previous studies, resulting in 30–40% cardiomyocyte death [20, 32]. A time course of hypoxic injury is shown in Supplemental Fig. 5A. Hypoxic cells showed noticeable vacuolisation, with fragmentation and plentiful debris (Fig. 3B). B_EV-treated hiPSC-CMs had improved morphology, but C_EV-treated cells appeared more like normoxic cells. As expected, hypoxia resulted in significant (P ≤ 0.0001) LDH release compared to normoxia (Fig. 3C). Based on lysing hiPSC-CMs and measuring total LDH release, this corresponded to ~ 35% cell death. LDH release was completely prevented by C_EVs (1.08-fold, P = 0.99). hiPSC-CMs treated with B_EVs at the same dose had significantly lower LDH than the vehicle control (P = 0.0003), but they were not as effective as C_EVs. To detect hiPSC-CM apoptosis we used an antibody array to measure multiple apoptosis-related proteins (Fig. 3D, E). Hypoxia significantly increased pro-apoptotic protein expression, which were significantly lowered by C_EVs, including activated caspase 3 and 8, cytochrome C and p53. hiPSC-CMs treated with B_EVs had higher expression of most pro-apoptotic markers than C_EV-treated cells. Some additional experiments were conducted using C_EVs. Seven days after restoring hiPSC-CMs to normoxia (in fresh culture medium, without EVs) the control group showed further elevation of LDH release due to re-oxygenation injury (Supplementary Fig. 5B) [36]. However, hiPSC-CMs which were previously incubated with C_EVs showed significantly less LDH release (P = 0.03), eight days after the EV treatment ended, demonstrating that C_EVs had a long-lasting protective effect. A higher dose of C_EVs (167 ng/µl, ~ 5,000 EVs per cell) was also tested, which did not offer any benefit over 67 ng/µl (P = 0.68). Conventional viability assays (WST/CCK-8) found that hiPSC-CMs had very low baseline dehydrogenase activity under normoxic conditions, which increased during hypoxia (Supplementary Fig. 5C, D). H9C2 cells and AC16 cells both showed a large decrease in CCK-8 activity after hypoxia, while CMSCLCs were less affected. C_EVs increased CCK-8 conversion in a dose-dependent manner, implying that they may affect CM metabolism. Since cell-secreted EVs can contain cytoplasmic components such as LDH or dehydrogenases we confirmed (Fig. 3C and Supplementary Fig. 5C) that neither C_EVs nor B_EVs had any effect on the assays. Together, results show that C_EVs reduced membrane damage, apoptosis and cell death of hypoxic hiPSC-CMs more effectively than B_EVs.

Extracellular vesicle miRNA cargo analysis

We next compared C_EV and B_EV miRNA cargo using three donors per cell type. Out of 752 probed miRNAs, 450 C_EV and 334 B_EV miRNAs were detected with cycle threshold (CT) values below 36 (Fig. 4A). All samples showed equal efficiency of miRNA isolation, reverse transcription and amplification (Supplementary Fig. 6A). Plotting normalised C_EV/ B_EV miRNA expression levels (Fig. 4B) showed a high correlation (R2 = 0.697). miR-21-5p was the highest detected miRNA in both EV types, and miR-1260a, miR-27a and miR-23a were highly-detected in both B_EVs and C_EVs. Comparing the most abundant miRNAs (Fig. 4C) showed 70–77% overlap between C_EV/B_EV cargo. Interestingly, some miRNAs (miR-202-5p (~ 11.6% of C_EV cargo), miR-451a (~ 5.1%) and miR-142-3p (~ 1.0%)) were found in high abundance in three separate C_EV donors but none of the B_EV samples. B_EVs contained hsa-miR-138-5p (1.0%) and hsa-miR-10b-5p (0.26%), which were not detected in C_EVs. Included among the most abundant miRNAs in both populations were miR-21-5p and miR-125b; both of which are stem cell-associated miRNAs [27].

Fig. 4

Extracellular vesicle miRNA cargo analysis. (A) Percentage of miRNAs detected in CMSCLC EVs (C_EV) and BM-MSC EVs (B_EVs) for three separate donor samples per group. Those with cycle threshold (CT) values of < 36.0 (green bar) were included in subsequent analyses. (B) Scatter plot of C_EV (Y axis) versus B_EV (X axis) mean miRNA expression levels normalised to reference miRNA (GeNorm) levels. The R-squared correlation is shown in the upper left. (C) Venn diagrams showing degree of overlap between the top 10, 20, 50 and 100 highest expressed C_EV miRNAs compared to B_EV miRNAs. (D) Gene ontology (GO) predictions for biological process (BP) for top 50 expressed C_EV miRNAs. Bars show the % of miRNAs belonging to each GO (lower X axis) and the green line shows the adjusted Fisher P value (upper X axis)

Target prediction of abundant CMSCLC and BM-MSC EV miRNAs

Since EV miRNAs act in combination to exert their effects, target pathway prediction was performed for the top 50 expressed C_EV and B_EV miRNAs. Categorisation by cellular component (CC) (Supplementary Fig. 6B) unsurprisingly showed high enrichment of exosome-related pathways. Categorisation by biological process (BP) (Fig. 4D) predicted GO:0010667 (negative regulation of cardiac muscle cell apoptotic process, (13 miRNAs, modified Fisher P-value = 1.6 × 10− 8)), angiogenesis (GO:1903589, GO:0016525), inflammation (GO:0050728), and cardiac muscle cell development (GO:0061049). The same analyses for B_EV miRNAs are shown in Supplementary Fig. 7. Due to the overlap between B_EV and C_EV miRNA cargo, target prediction results were overall similar.

Transcriptomic analysis of CMSCLC and BM-MSC EV activity

Next, we used RNA-seq to examine how B_EVs or C_EVs affected the hypoxic hiPSC-CM transcriptome. 97.63 ± 0.27% of transcripts were successfully mapped. Normoxic hiPSC-CMs expressed TNNT2 (7,397 transcripts per million, TPM), as well as CM markers TBX5, HEY2, MYL2, ACTN1, IRX4, GJA1 (connexin-43) and ATPA2 (SERCA2) and had low CDK1 (8.6 TPM) and FGF8 (0.2 TPM), indicating a ventricular CM phenotype [42, 43]. Principal components analysis (PCA) (Fig. 5A) showed distinct profiles for each experimental group, with high consistencies of samples within each group.

Plotting TPM distribution (Fig. 5B) of all genes (n = 60,671 total) showed that hypoxia increased overall expression levels (P = 1.8 × 10− 24 vs. normoxia, measured by Kolmogorov-Smirnov (KS) Test). Interestingly, C_EVs further increased total expression levels (P = 6.9 × 10− 6 vs. hypoxia) whereas B_EVs had no effect on overall gene expression levels (P = 0.99). The same finding was observed for protein-coding genes (Fig. 5C). Individual comparisons are shown in Supplementary Fig. 8A, B. Comparing hypoxic to normoxic hiPSC-CMs (Supplementary Fig. 8C-F) revealed 3,125 and 5,105 significantly down and up-regulated genes respectively. Unsurprisingly, hypoxic cells showed enriched pathways related to cellular stress, apoptosis, oxidoreductase activity, dehydrogenase activity, electron transport chain and muscle contraction. Hypoxia-related genes such as VEGFA and ENO2 were upregulated up to 40-fold in all three hypoxia groups and were not affected by either of the EV treatments. These findings are very similar to previously published microarray and RNA-seq of hypoxic human cardiomyocytes [32, 42]. This demonstrates that the utilised hypoxia model induced the relevant and appropriate responsive pathways in iPSC-CMs.

Venn diagrams comparing overlapping up- and down-regulated genes between normoxia/hypoxia/C_EV treatment groups are shown in Fig. 5D. C_EVs reversed the direction of many gene expression changes which were induced by hypoxia, mostly by increasing their expression. Comparing hypoxic hiPSC-CMs + C_EVs against EV vehicle (Fig. 5E, F) showed that C_EVs significantly up-regulated 1,507 genes and significantly down-regulated 541 genes. Categorising differentially expressed genes by KEGG (Fig. 5G) revealed significant up-regulation of Pi3k-akt signalling, ECM-receptor interaction, cell adhesion and calcium signalling pathways, all of which are important modulators of CM survival [36]. Sorting by molecular function (Fig. 5H), the most significant changes related to up-regulation of heparin and glycosaminoglycan (GAG) binding, ECM structural constituents, and metal ion transporter activity.

Next, we looked at the most differentially expressed genes in the C_EV-treated hiPSC-CMs by both fold-change and statistical significance (Fig. 5I, J). Of the most significantly upregulated genes, many are known to be cardioprotective, including A2M, NPPA, SELENON and THBS4. The most upregulated gene, A2M (alpha-2-macroglobulin), is a powerful anti-inflammatory protein which inhibits multiple cytokines and cellular proteases, and activates cardioprotective ERK1/2, Akt and PI3-kinase pathways. It was up-regulated 24-fold by C_EVs but unchanged by B_EVs. SELENON, coding for selenoprotein N, protects cells from oxidative stress and maintains calcium homeostasis and contractile function during stress. It was down-regulated by hypoxia, increased 4-fold by C_EVs (P = 1.12 × 10− 70) but was unchanged by B_EVs (P = 0.42). THBS4 has been previously shown as cardioprotective and was increased 6.8-fold by C_EVs and unaffected by B_EVs [44]. Together, these data indicate that C_EVs induced multiple protective responses in hypoxic cardiomyocytes.

Fig. 5

RNA sequencing of hypoxic EV-treated human cardiomyocytes. (A) Principal component analysis (PCA) for normoxia, hypoxia + vehicle (Hyp), hypoxia + CMSCLC EV (H + C_EV) and BM-MSC EV groups (H + B_EV). (B) TPM distribution of all gene transcripts or (C) protein-coding gene transcripts for the four experimental groups. Sample distributions were compared by Kolmogorov-Smirnov (KS) test, and the direction of change and P values are shown for each comparison. (D) Venn diagrams showing the number of overlapping genes between the stated comparisons. (E) Volcano plot of hypoxia + vehicle against hypoxia + C_EVs. The Y axis show statistical significance, with the solid line showing P = 0.05. The X axis shows log2 fold change with the red and green lines showing two-fold down- and up-regulation respectively. (F) Scatter plot of hypoxia + vehicle vs. hypoxia + C_EV. Each point represents one gene. Green points indicate P ≤ 0.05 and the box indicates the genes with ≥ 0.3 TPM which were included in subsequent analyses. (G) Pyramid plot of most significantly enriched pathways by KEGG for hypoxia + C_EV vs. hypoxia + vehicle. The X axis shows the number of upregulated and downregulated genes in each group and the bar colours indicate statistical significance. (H) Pyramid plot of molecular function (MF). (I) Scatter plot showing the 10 most differentially-expressed and most statistically significant genes (J) between hypoxia + vehicle vs. hypoxia + C_EV groups

Looking at the genes most significantly reduced by C_EVs, CKM (creatine kinase, M-type) was reduced by C_EVs but was increased by B_EVs. HMOX1 (heme oxygenase-1) was expressed at very low levels (2.2 TPM) in normoxia and increased to 774.1 TPM in hypoxia, as expected. Both C_EVs (108.2 TPM, P = 7.3 × 10− 22) and B_EVs (93.0 TPM, P = 4.81 × 10− 41) lowered HMOX1 expression. Interestingly, both CKM and HMOX1 are hypoxia-inducible and have cardioprotective functions; HMOX1 by anti-oxidant activity and CKM by preserving cardiomyocyte ATP production [45, 46]. Taken together, these data indicate that C_EVs significantly aided in up-regulating multiple cardioprotective genes in response to hypoxic stress, and many of these changes were not found after treatment with B_EVs. However, the genes which were strongly downregulated by C_EVs were mostly also downregulated by B_EVs.

The hypoxic hiPSC-CM response to B_EVs is shown in Supplementary Fig. 9. Here, there was more total gene downregulation (1,719) than upregulation (1,407) and the most significantly enriched biological pathways related to ion channels, calcium signalling and cAMP signalling. B_EVs also affected many of the same pathways as C_EVs, including GO: 0008201 heparin binding (P = 0.02), and GO: 0005539 glycosaminoglycan binding (P = 0.011). Direct comparison of C_EVs and B_EVs is shown in Supplementary Fig. 10 and a summary of strongly differentially regulated genes (based on Fig. 6I-J) is shown in Supplementary Fig. 11A and CM apoptosis-related genes in Supplementary Fig. 11B.

Fig. 6

Determining hypoxia protection by abundant EV miRNAs. (A) CCK-8 activity of hypoxic AC16 cardiomyocytes incubated with miRNA negative control (miR-NEG), single miRNA mimics or combinations of mimics to a total of 15 nM. Percentages are relative to normoxia. (B) LDH secretion shown as change in absorbance relative to normoxia. Groups were compared to miR-NEG by one-way ANOVA with Dunnett’s multiple comparison test. * = P < 0.05, **** = P ≤ 0.0001. N = 24 samples per group. (C) Representative images from selected conditions. (D) Gene expression of predicted miRNA targets after normoxia, hypoxia + miR-NEG, miR-1260a or miR-142/202/451 at 15 nM. The Y axis shows log scale of gene expression normalised to GAPDH. N = 4 independent samples per bar. Statistical annotations show comparisons against hypoxia + miR-NEG by one-way ANOVA with Dunnett’s multiple comparison test. * = P < 0.05, ** = P ≤ 0.01

Integrating CMSCLC and BM-MSC EV miRNA cargo and down-regulated genes

Next, we integrated RNA-seq data with target prediction for the top 50 most abundant C_EV and B_EV miRNAs. We filtered for genes which were ≥ 2.0-fold up-regulated during hypoxia and ≥ 2.0-fold down-regulated with addition of C_EVs with p.adj < 0.05; nine genes met these criteria for C_EVs (Supplemental Fig. 12A). Of these 9 predicted genes, all except EGR1 were also significantly reduced by B_EVs. EGR1 was strongly up-regulated during hypoxia (6.1-fold), downregulated by C_EVs (4.1-fold, P = 0.001755), but not affected by B_EVs (P = 0.10). The same analysis was performed for B_EVs, as shown in Supplemental Fig. 6B.

To validate some of these findings we treated AC16 cardiomyocytes with miR-21-5p and miR-1260a, to represent the most abundant C_EV and B_EV miRNAs, as well as miR-202-5p, miR-451a and miR-142-3p to represent abundant C_EV-exclusive miRNAs. At a 5nM concentration (Supplemental Fig. 13a) miR-21-5p transfection reduced hypoxic CM viability compared to the negative control mimic, miR-1260a had no significant effect, and 5 nM of combined miR-142-3p/202-5p/451a increased viability. At 25 nM (Supplemental Fig. 13b), all miRNAs greatly and equally reduced cell viability, indicating saturation of the RISC system. At 15 nM (Fig. 6A) miR-21-5p again reduced viability, and miR-1260a increased viability compared to negative control transfection. Interestingly, single miR-142-3p, 202-5p or miR-451a at 15 nM had no significant effects on viability, but when combined (5 nM each), they significantly improved viability. A combination of the top 5 C_EV miRNAs (including miR-1260a and miR-21-5p) again had no significant effect, possibly from re-introduction of detrimental miR-21-5p. Measuring LDH release (Fig. 6B) showed the same findings of miR-1260a and miR-142/202/451 protecting hypoxic CMs, and miR-21 worsening CM injury. Representative images (Fig. 6C) also showed the detrimental effects of miR-21 and beneficial effects of miR-1260a and the C_EV-exclusive miR-142/202/451a combination. Gene expression analysis of the two cardioprotective miRNA treatments showed that some of the changes found in iPSC-CMs were also reproduced in CMs treated with miRNA mimics (Fig. 6D). For example, A2M increased with hypoxia, and was further increased by the combination of miR-142/202/451 (15.2-fold vs. miR-NEG). miR-1260a also reduced expression of several target genes including KITLG, JAK2 and PTEN. However, some of the C_EV modified targets such as HMOX1 and EGR1 did not change in response to these miRNA mimics, indicating that they may be targeted by other components of the EV cargo. Gene expression was also tested with mimics provided at 5 nM and 25 nM concentrations (Supplemental Fig. 13, d) which found similar trends. Notably, 5 nM of miR-142/202/451 was sufficient to increase A2M expression. Taken together, we summarise that EVs isolated from human RAA-derived cardiac stromal cells can robustly protect cardiomyocytes from hypoxic injury which is in part due to miRNA cargo and upregulating cardioprotective pathways.