Regenerative potential of epicardium-derived extracellular vesicles mediated by conserved miRNA transfer

Abstract Aims After a myocardial infarction, the adult human heart lacks sufficient regenerative capacity to restore lost tissue, leading to heart failure progression. Finding novel ways to reprogram adult cardiomyocytes into a regenerative state is a major therapeutic goal. The epicardium, the outermost layer of the heart, contributes cardiovascular cell types to the forming heart and is a source of trophic signals to promote heart muscle growth during embryonic development. The epicardium is also essential for heart regeneration in zebrafish and neonatal mice and can be reactivated after injury in adult hearts to improve outcome. A recently identified mechanism of cell–cell communication and signalling is that mediated by extracellular vesicles (EVs). Here, we aimed to investigate epicardial signalling via EV release in response to cardiac injury and as a means to optimize cardiac repair and regeneration. Methods and results We isolated epicardial EVs from mouse and human sources and targeted the cardiomyocyte population. Epicardial EVs enhanced proliferation in H9C2 cells and in primary neonatal murine cardiomyocytes in vitro and promoted cell cycle re-entry when injected into the injured area of infarcted neonatal hearts. These EVs also enhanced regeneration in cryoinjured engineered human myocardium (EHM) as a novel model of human myocardial injury. Deep RNA-sequencing of epicardial EV cargo revealed conserved microRNAs (miRs) between human and mouse epicardial-derived exosomes, and the effects on cell cycle re-entry were recapitulated by administration of cargo miR-30a, miR-100, miR-27a, and miR-30e to human stem cell-derived cardiomyocytes and cryoinjured EHM constructs. Conclusion Here, we describe the first characterization of epicardial EV secretion, which can signal to promote proliferation of cardiomyocytes in infarcted mouse hearts and in a human model of myocardial injury, resulting in enhanced contractile function. Analysis of exosome cargo in mouse and human identified conserved pro-regenerative miRs, which in combination recapitulated the therapeutic effects of promoting cardiomyocyte proliferation.


Aims
After a myocardial infarction, the adult human heart lacks sufficient regenerative capacity to restore lost tissue, leading to heart failure progression. Finding novel ways to reprogram adult cardiomyocytes into a regenerative state is a major therapeutic goal. The epicardium, the outermost layer of the heart, contributes cardiovascular cell types to the forming heart and is a source of trophic signals to promote heart muscle growth during embryonic development. The epicardium is also essential for heart regeneration in zebrafish and neonatal mice and can be reactivated after injury in adult hearts to improve outcome. A recently identified mechanism of cell-cell communication and signalling is that mediated by extracellular vesicles (EVs). Here, we aimed to investigate epicardial signalling via EV release in response to cardiac injury and as a means to optimize cardiac repair and regeneration.

Methods and results
We isolated epicardial EVs from mouse and human sources and targeted the cardiomyocyte population. Epicardial EVs enhanced proliferation in H9C2 cells and in primary neonatal murine cardiomyocytes in vitro and promoted cell cycle re-entry when injected into the injured area of infarcted neonatal hearts. These EVs also enhanced regeneration in cryoinjured engineered human myocardium (EHM) as a novel model of human myocardial injury. Deep RNA-sequencing of epicardial EV cargo revealed conserved microRNAs (miRs) between human and mouse epicardial-derived exosomes, and the effects on cell cycle re-entry were recapitulated by administration of cargo miR-30a, miR-100, miR-27a, and miR-30e to human stem cell-derived cardiomyocytes and cryoinjured EHM constructs.

. Conclusion
Here, we describe the first characterization of epicardial EV secretion, which can signal to promote proliferation of cardiomyocytes in infarcted mouse hearts and in a human model of myocardial injury, resulting in enhanced contractile function. Analysis of exosome cargo in mouse and human identified conserved pro-regenerative miRs, which in combination recapitulated the therapeutic effects of promoting cardiomyocyte proliferation.

Introduction
Cardiovascular diseases (CVD) are one of the leading causes of death worldwide. 1 Sublethal cardiac ischaemic events lead to the development of a non-contractile myocardium that evokes pathological compensatory mechanisms within the remaining viable tissue. If the damage cannot be compensated, progression towards heart failure ensues. Current stateof-the-art pharmacotherapies aim at protecting the heart from neurohumoral stressors and reducing the metabolic burden of the impaired myocardium. Replacement of the underlying loss of functional heart muscle via stimulating cardiomyocyte regeneration is not achieved with current treatments. Therefore, significant efforts have been made to develop alternate cell-based therapies, including those aimed at promoting cardiomyocyte proliferation to regenerate the injured heart and prevent the onset of heart failure.
Targeting the low proliferative capacity of adult human cardiomyocytes 2 to re-enter the cell cycle is a major therapeutic goal. Several cell types with an innate capacity to provide regenerative cues have been identified, such as the epicardium. [3][4][5] Epicardial cells form the outermost layer of the heart, and in adults, they are essentially quiescent with lowlevel turnover to maintain myocardial homeostasis. During cardiac development, the epicardium contributes key cell types, such as fibroblasts 6 and vascular smooth muscle cells 6 and endothelial cells to the developing coronaries. 7,8 The epicardium also promotes cardiomyocyte proliferation via paracrine signalling. 9,10 Furthermore, epicardial and epicardium-derived cells (EPDCs) reinvoke an embryonic gene  . . programme and developmental characteristics after injury in lower vertebrates 11,12 and regenerative murine models. 13 This in turn can be induced in adult mammalian hearts after injury by priming with factors, such as the small peptide Thymosin b4 to improve outcome in the nonregenerative setting. 4,14 As in development, a key aspect of the pro-regenerative function of the epicardium is via signalling and provision of mitogens and trophic growth factors. In zebrafish, retinoic acid production from the epicardium was shown to be essential for heart regeneration 15 and in mouse epicardium, retinoid X receptor alpha is known to promote foetal cardiomyocyte proliferation. 9 Whether epicardial signals can act to promote cardiomyocyte proliferation in adult mammalian hearts, either in steady state or following injury remains unknown.
Recently, there has been a significant focus around alternate mechanisms of intercellular signalling and novel modes of cell-cell communication mediated by extracellular vesicles (EVs) and the subclass known as exosomes. 16 EVs were first identified as a subpopulation of doublemembrane vesicles that are secreted by cells in culture. 17 They are endosomal in origin, express specific membrane markers 18,19 and have a defined size, ranging from 30 to 120 nm. EVs and exosomes have been described as potential biomarkers of cellular damage in plasma 20 and pericardial fluid, 21 and have also been proposed as a novel mode to deliver therapeutic factors. They are vehicles for mitogenic cargo, possess low immunogenicity, can be cryopreserved and are able to be isolated in large quantities. 22 EVs and exosomes derived from plasma, 23 mesenchymal stem cells, 24-26 embryonic stem cells, 27 cardiosphere-derived cells, 28 bone marrow, 29 and cardiac progenitor cells 30 have been reported to be cardioprotective and improve outcome following ischaemic heart injury.
In this study, we describe for the first time the secretion of mouse and human EVs by epicardial cells, and their ability to evoke a cardiomyocyte proliferative response in regenerative postnatal day 1 (P1) and nonregenerative P7 neonatal mice 13 after myocardial infarction (MI), and in human embryonic stem (hES) cell-derived cardiomyocytes in vitro. We selected neonatal mouse heart injury to model cardiovascular regeneration vs. repair whereby we can assess the documented shift from regeneration underpinned by cardiomyocyte proliferation to an adult default wound healing process of fibrosis. Injured P7 hearts respond by scarring and remodelling after injury and are unable to induce a proliferative response in the cardiomyocyte population to restore lost myocardial tissue; in this regard they mimic the non-regenerative response observed in adult hearts. 13,31,32 Employing the neonatal MI model enables us to determine effects of EVs on existing cardiomyocyte proliferative capacity at P1 vs. a direct comparison 7 days later when this proliferative capacity is negligible. We also utilized engineered human myocardium (EHM 33 ) to test the effect of acute EV delivery after cryoinjury. EHM is a human model, which parallel the neonatal mouse system via a more developmental and potentially regenerative response to injury, allowing for comparative analyses following EV treatment across both models. We describe here the first demonstration of (cryo) injury response in EHM, which mimics cell and tissue loss, as observed in AMI patients, and as such facilitates testing of the potential therapeutic effects of EVs. Contractile recovery of EHM was restored within 7 days post-injury following treatment with EVs, via increased cell cycle re-entry of the remaining cardiomyocyte population. Analysis of microRNA cargo of mouse and human epicardial EVs revealed conserved miR content across species. Adenoviral-mediated delivery of highly represented miRNAs promoted proliferation of human cardiomyocytes recapitulating the mitogenic effects of host EVs. These findings are of widespread interest in terms of understanding cell-cell communication in the context of ischaemic heart disease and pave the way to develop new approaches to promote myocardial tissue restoration based on paracrine signalling mediated by epicardial EV release.

Institutional ethics
All surgical and pharmacological procedures were performed in accordance with the Animals (Scientific Procedures) Act 1986 (ASPA) (Home Office, UK). All animal experiments were subject to local ethical committee approval (ACER) and all were undertaken in compliance with the standards set down in the ASPA, revised 2012, and with the requirements of the European Directive 63/2010 on the protection of animals used for scientific purposes. Home Office project licences (PPL 30/2987; PPDE89C84). The use of human embryonic stem cells (hESCs) for the tissue engineering work was approved according to the German Stem Cell Act by the Robert-Koch-Institute (permit #12; reference number: 1710-79-1-4-16).

Animal husbandry, MI, and EV delivery in postnatal mice
Mice were housed and maintained in a controlled environment by University of Oxford Biomedical Services. FUCCI R26Rp-mice were kept on a C57BL/6J inbred background and have been described previously. 34 Animals were euthanized by cervical dislocation.
Neonatal (P1/P7) mice used in MI experiments were anaesthetized using a combination of hypothermia and inhaled isoflurane 3% v/v in oxygen. Neonates were exposed to this for 15 s followed by ice for 30-45 s for a P1 pup or 60-90 s for a P7 pup. MI was performed by permanent ligation of the left anterior descending 35 coronary artery. For sham controls, a suture was passed under the LAD but not ligated. Mice treated with epicardial EVs received one intracardiac injection into the observable infarct border zone (EV titre of 10 9 , in a volume of 5 uL, amount of EV within the range previously used in the literature 28 ) or phosphatebuffered saline (PBS) (vehicle) at the time of the injury. In order to maximize the amount of particles injected, several tests were performed injecting larger volumes or administering more than one injection around the injured area in neonatal hearts. Unfortunately, both of these tests reduced the survival rate of pups after surgery. Therefore, a compromise was reached to inject the largest amount of EVs that enabled a good survival rate, and still fall within the reported dosing in the literature. 28 Whole hearts were then excised at 48 h, 7 and 21 days following MI, which correspond to Days 2, 7, and 21 post-injury. Hearts collected for immunostaining were washed in ice-cold PBS prior to overnight fixation in a 2% formaldehyde solution at 4 C.

Immunodetection methods
Cultured cells from neonatal mice or hES-CM were fixed at 4 C overnight in 2%PFA prior to staining. Cryosections were processed for indirect immunofluorescence using standard methods. Primary antibodies are listed in Supplementary Table S2. Alexa Fluor secondary antibodies (Invitrogen, 1:200) were used in all cases. Imaging was performed using an Olympus FV1000 confocal microscope (tissue sections/cell culture) and Leica SP8 Navigator. Images were digitally captured and processed using Fiji software.
Epicardial extracellular vesicles promote regeneration

Histology
On Day 21 after MI, hearts were harvested and processed for paraffin embedding, cut into Superfrost slides, and deparaffinized using standard methods. For Masson's trichrome staining (Abcam), sections were stained according to the manufacturer's instructions. For Picrosirius red staining, sections were stained using the Picro Sirius Red Stain Kit (Abcam) for 60 min according to the manufacturer's protocol and digitalized using a scanner. Picrosirius staining was also imaged under polarized light using a Nikon 3i 4Â and 20Â objectives. For FUCCI sections, tissue was co-immunostained with the cardiomyocyte marker a-actinin. For subsequent analyses, only cardiomyocytes expressing a-actinin were taken-into-account. Total of 10-15 sections per heart were quantified, at an average of 900 cells throughout each section.

Cell culture
H9C2 cells (ATCC) and mouse epicardial cells 36 according to standard procedure.
Epicardial cells were cultured maintaining epithelial morphology until 80% confluency in T25 (2.5 10 6 ) or T75 (8 10 6 ) flasks according to procedures described upon line generation. 36 Prior to EV isolation, cells were cultured in OPTIMEM without serum for 48 h. Treatment with epicardial EVs (both for H9C2 cell line and neonatal primary cardiomyocytes) was performed for 24 h (at a concentration of 1.5 Â 10 8 EVs, consistent with previously described doses 28,37 ). Where applicable, bromo deoxyuridine (BrdU) administration was also added at 50 ng/mL for 24 h. H9C2 cells were seeded (at an initial density of 0.4x10 6 ) in P12 plates on coverslips prior to treatment.

EV isolation and analysis
EVs were isolated from conditioned media (OPTIMEM without serum) by ultracentrifugation following standard procedures. 38 Briefly, cells were cultured in OPTIMEM for 48 h prior to collection of CM. EV isolation was performed following an initial centrifugation step of the conditioned media at 2000g to remove cellular debris followed by filtration through a 0.22 um filter. Resulting supernatant was subjected to ultracentrifugation of 1 h at 120 000g followed by a second ultracentrifugation of 1 h at 120 000g after washing the pellet with PBS. Analyses of vesicle size were performed by Nanoparticle-tracking assays (Nanosight). Labelling of EVs was performed with PKH26 (Sigma) according to the manufacturer's directions.

RNA isolation and gene expression profile by qRT-PCR
Total RNA from cultured cells and epicardial EVs prior to sequencing was isolated using Trizol reagent (Invitrogen) according to the manufacturer's instructions. Total RNA was reverse transcribed using oligo-dT primers and Superscript III RT (Invitrogen). qRT-PCR analysis was performed on a ViiA 7 Real-Time PCR System (Thermo Fisher Scientific) using Fast SYBR Green Master Mix (Thermo Fisher Scientific). Data were normalized to Gapdh housekeeping gene expression. Fold changes in gene expression were determined by the 2 -DDCT method. Statistical differences were detected using an unpaired, 2-tailed Student's t-test.

Tem processing and immunogold labelling
EVs were resuspended in TBS-3% PFA and adsorbed to the grids for 5 min. Primary antibody CD63 (Santa Cruz) was used at a 1/50 dilution and then the grid was placed for 60 min on a drop of secondary antibody coupled with 10 nm diameter gold particles diluted 1/40. The grids were negatively stained with 2% uranyl acetate for 10 s.

Neonatal cardiomyocyte isolation and culture
P1 pups were sacrificed by cervical dislocation followed by decapitation and hearts dissected for cardiomyocyte isolation using Pierce Primary cardiomyocyte isolation kit according to manufacturer's directions. Digested cardiomyocytes were seeded (at a density of 0.25 Â 10 6 ) in P12 well plates on coverslips. After 4 days in culture, neonatal cardiomyocytes were treated with epicardial EVs. Treatment with EVs was performed for 24 h (1.5 Â 10 8 particles). Where applicable, BrdU was added for 24 h at 50 ng/mL.

Epicardial isolation from human right atrial appendage
Right atrial appendage (RAA) from patients undergoing bypass surgeries were dissected for epicardial and EPDC culture according to References. 39 All patients provided written informed consent and the investigation conformed to the principles outlined in the Declaration of Helsinki.

Generation and cryoinjury of EHM 41
EHM were generated as previously described. 33 Briefly, hES2-LUC þ derived cardiomyocytes and human foreskin fibroblasts (70:30 ratio) were interspersed within a bovine collagen hydrogel and cast into circular silicone moulds. These circular EHM were transferred to flexible stretchers 3 days post-casting that permitted auxotonic spontaneous contractions. EHM were maintained in culture for 28 days and cryoinjuries performed thereafter. Cryoinjuries were performed utilizing a liquid nitrogencooled stainless-steel oral gavage attached to a plastic syringe barrel. Injuries to only one separated arm of each EHM were performed twice to ensure a under control conditions irreversible injury.

Assessment of EHM contractile function
At specific time points post-cryoinjury (Cry), EHM were mounted onto isometric force transducers (Föhr Medical Instruments, Germany), lowered into 37 C water-jacketed organ baths and bathed in Tyrode's solution (in mM: 119.8 NaCl, 5.4 KCl, 1.05 MgCl 2 , 0.42 NaH 2 PO 4 , 22.6 NaHCO 3 , 5 glucose, 0.28 L ascorbic acid). The EHM were gradually stretched according to the Frank-Starling mechanism in the presence of 1.8 mM Ca 2þ to determine the maximal force of contraction (F max ). All subsequent functional assessments were performed at F max .  Bioinformatic analysis was performed using Oasis. 42 Sequence reads were pre-processed to remove adapters and retain only 16-30 bp length reads. Reads were aligned to reference genome. The reads aligning to each known mature miRNA were counted using Bioconductor packages for next-generation sequencing data analysis based on miRNA definitions in miRBase database. The differential expression analysis between different sample types was performed using the negative binomial statistical model of read counts as implemented in the DESeqBioconductor package. The statistical significance of differential expression is established based on the false discovery rate (FDR)-adjusted P-values. The cluster analysis of all differentially expressed miRNA was performed using the Bayesian infinite mixture model. Pathway enrichment analysis was conducted with Gene Ontology. Enrichment P-values were corrected for FDR and were considered significant when adjusted P< _0. 05. GEO accession number: GSE161630 (https://www.ncbi.nlm.nih.gov/ geo/query/acc.cgi?acc=GSE161630).

AAV vectors
All of the AAV vectors used in this study were generated by the International Centre for Genetic Engineering and Biotechnology AAV Vector Unit (http://www.icgeb.org/avu-core-facility.html) using a dualtriple plasmid co-transfection procedure followed by polyethylene glycol precipitation and purification through CsCl2 gradient centrifugations as described previously. AAV particles were used to infect the cells at a MOI of 10 4 for 48 h before fixing for further analysis.

AAV treatment of EHM
EHMs were infected immediately after cryoinjury (as in Section 2.12) with AAV serotype 9 (AAV9)-miR100 or AAV9-miR30a with a MOI of

Western blotting
Primary neonatal cardiomyocytes were homogenized using RIPA buffer in presence of protease and phosphatase inhibitors [Inhibitor cocktail (11 697 498 001, Roche, Switzerland) and 1 lM sodium orthovanadate]. After clarification by centrifugation, protein concentration was measured using Pierce BA Protein Assay kit (23 227, Thermo Scientific; USA) following manufacturer instructions. Protein (30 lg) were denaturized at 95 C for 5 min, loaded on an 8% polyacrylamide SDS-PAGE gel and run at 120 V for 90 min. Subsequently, samples were transferred to a nitrocellulose membrane by wet transfer at 400 mA for 2 h. Membranes were blocked with 5% BSA for 1 h and incubated with primary antibodies O/N at 4 C. Next day, membranes were washed in TBS-T buffer and incubated with the corresponding HRP-conjugated secondary antibodies (Dako, Denmark) at 1:5000 dilution for 1 h at RT. After washing, signal was developed using ECL Primer Western Blotting Detection Reagent (Amersham; UK) and detected by a LAS-3000 imaging system (Fujifilm; USA). The primary antibodies used were diluted 1:1000.

Statistics
Statistical analyses were performed using GraphPad Prism 7 software. The statistical significance between two groups was determined using a non-parametric Mann-Whitney test or an unpaired two-tailed Student's t-test; these included an F test to confirm the two groups had equal variances, and the data were reported as mean ± SEM. Among three or more groups 1-way ANOVA followed by Tukey's multiple comparisons test was used. P < _ 0.05 was considered statistically significant.

Mouse epicardial cells release EVs, which are taken up by cardiomyocytes and promote cell cycle activity
There is extensive evidence that epicardial cells play a major role in cardiomyocyte proliferation during development and regeneration through paracrine signalling (reviewed in Reference 10 ). This led us to determine Epicardial extracellular vesicles promote regeneration whether the epicardium communicates with cardiomyocytes via EV shedding. To begin to address this, we cultured immortalized mouse embryonic epicardial cells from a previously established cell line 36 ( Figure 1A) and isolated EVs from conditioned medium utilizing a previously described ultracentrifugation protocol. 38 We identified EVs in the epicardium conditioned medium that displayed typical exosomal characteristics, such as a modal particle size of 98 nm (analysed by Nanosight and TEM) ( Figure 1B and C and Supplementary Figure S1A-C), canonical morphology and expression of the exosomal marker CD63 ( Figure 1D).
To assess the potential function of the epicardial EVs in vitro, we first screened for proliferation-inducing effects in the well characterized H9C2 rat myocyte cell model. 43 Exposure of H9c2 cells to isolated mouse epicardial EVs (mepic EV) resulted in significantly increased level of cell cycle activity, by more than two-fold as reported by phosphohistone H3 staining (PH3) PBS 1.

Epicardial EVs promote cell cycle activity in mouse cardiomyocytes ex vivo
To further interrogate the effect on cardiomyocyte cell cycle activity, we treated primary mouse cardiomyocytes isolated at P1 with mouse epicardial-derived EVs. We detected PH3 þ cardiomyocytes (co-stained with a-actinin) (Figure 2A We further analysed cell cycle progression and activity by taking advantage of the FUCCI reporter mouse line. 34,44 In the FUCCI reporter, cells progressing through the cell cycle are labelled with fluorescent Cherry when in G1 (Cherry reports ubiquitination domains of Cdt1) and with the Venus fluorescence when transitioning through S phase and mitosis (Venus reports ubiquitination domains of Geminin). 34 Treatment of primary neonatal cardiomyocytes isolated from FUCCI hearts revealed an increase in cardiomyocytes that were undergoing S phase and mitosis when treated with mepic exo (PBS 1.53±0.69; EV 3.98±0. 16, P< _0.05) ( Figure 2J-L). We then sought to analyse downstream pathways that were mediating the increase in cell cycle activity in murine neonatal cardiomyocytes. We investigated activation of ERK, AKT, and Hippo pathways, all of which have been previously implicated in cell cycle activity and proliferation in cardiomyocytes. [45][46][47][48] Following treatment of cardiomyocytes with epicardial EVs, we detected a significant increase in phosphorylation of Akt and a trend in the reduction in (inactive) phosphorylated Yap, favouring the non-phosphorylated (activated) Yap isoform, albeit this was not significant. We did not detect a significant increase in the phosphorylation of ERK, but observed a trend towards activation of the ERK pathway followed epicardial EV treatment (Akt: PBS: 0. 63±0 Figure S3). Taken together, these data revealed that epicardial-derived EVs significantly induced a pro-proliferative response in murine neonatal cardiomyocytes ex vivo and that this response is mediated by activation of the Akt pathway and partial activation of Hippo and ERK signalling pathways.

Epicardial EVs induce a proliferative response in vivo during and beyond the murine neonatal regenerative window
The regenerative capacity of the mouse neonatal heart has been described in injury models of apical resection, 13 cryoinjury, 49,50 and left anterior descending 35 artery ligation. 51 The extent of regeneration is linked with the proliferation of pre-existing cardiomyocytes 13 and angiogenesis, 52 similar to that described in lower vertebrates. 11 However, this regenerative potential is lost within the first week of life by P7 correlating with cardiomyocytes exiting the cell cycle. 13 Injured P7 hearts, therefore, behave comparably to adult mammalian hearts, regarding their response to injury; after MI a scar is formed impeding regeneration and cardiomyocytes are unable to proliferate and replace the loss myocardium. 13,31,32 To test the regenerative capacity of epicardial EVs in murine hearts, we induced a MI by LAD ligation in regenerative P1 and nonregenerative P7 FUCCI R26Rp mouse pups (in which FUCCI expression is directed by the Rosa26 promoter) 34 followed by intracardiac injection of epicardial EVs into the muscle wall of the left ventricle, around the injury area. Expression of the FUCCI reporter in cardiomyocytes was assessed by co-detection of a-actinin throughout orthogonal views (zaxis) per section. Only nuclei clearly contained within an a-actininþ cardiomyocyte were considered for quantification ( Figure 3A).
Analysis of FUCCI cell cycle progression in P1 mouse hearts 1 week post-MI showed an increase in proliferation, relative to sham-operated control hearts without any treatment ( Figure 3B, C and F and Supplementary Figure S4A Figure S5B). Cumulatively, these results suggest that the innate cardiomyocyte proliferative response in P1 mouse hearts after acute injury can be further enhanced by exposure to epicardialderived EVs and significantly, that the relatively quiescent, more adultlike cardiomyocytes at P7, can be induced to re-enter the cell cycle by EV treatment.

Epicardial EVs fail to promote a longerterm regenerative response in P7 neonatal hearts post-MI
In order to assess the longer-term effects of epicardial EV injection on non-regenerative P7 mouse hearts, we analysed scar formation 21 days after P7 injury and EV treatment. Histological examination showed no    Figure S6D and E).
We then sought to further characterize the longer-term effects of EVs and analysed whether they elicited a pro-angiogenic response in P7 infarcted hearts. We did not observe significant differences in the vascular density after injury or EV treatment (PBS injected 9.25±1.14; EV injected 8. 76±0.77) (Supplementary Figure S6F and G). Moreover, analysis of cardiomyocyte sarcomeric structure by a-actinin staining showed no obvious changes in extent of myocardial disarray. Taken together, these results show that whilst we observed early effects (within the first 7 days) on cardiomyocyte proliferation ( Figure 3) (Supplementary Figures S4 and  S5), there was no long-term beneficial effect of epicardial EV injection in response to MI in the non-regenerative, adult-like P7 setting; suggesting that increased cardiomyocyte proliferation alone is insufficient to promote optimal repair, longer-term regeneration or improved outcome.

Human primary and embryonic stem cell-derived epicardial cells secrete EVs that enhance proliferation in mouse and human cardiomyocytes
To determine whether our findings in mice might be conserved in humans, we sought to determine whether EVs could be isolated from epicardial cells from RAA biopsies from patients undergoing coronary artery bypass graft surgery 39 and differentiated H9 human embryonic stem cells. 40 EPDCs from RAA biopsies were expanded with a cobblestone (inactive) epithelial morphology ( Figure 4A) and underwent spontaneous, or evoked, activation to a more spindle-like, mesenchymal form that recapitulates epithelial-mesenchymal transition 39 (EMT) ( Figure 4B). EMT has also been observed to inherently occur with epicardial activation in injured adult mouse hearts. 53 Human primary EPDCs secreted exosomal-like EVs that were within the expected size range ( Figure 4C) and displayed a characteristic cupped-shaped morphology ( Figure 4D). Similarly, hESC-derived epicardial cells also displayed a cobble-stone-like morphology (Supplementary Figure S7A) and strongly expressed epicardial markers Wt1, Tbx18, and Tcf21 (Supplementary Figure S7A-D 40 ). Analysis of their cell-conditioned media revealed particles that were within the size range and morphological appearance for EVs (Supplementary Figure S7E).

Human primary EPDC-derived EVs induce proliferation of cardiomyocytes in mouse and human 2D cardiomyocyte cultures and in EHM after cryoinjury with improved functional outcome
To explore the functional potential of human epicardial EVs, we initially tested the effect of the more activated, mesenchymal (spindle) EPDCderived EVs on primary mouse cardiomyocytes expressing the FUCCI cell cycle reporter. Spindle EPDC-derived EVs significantly increased the number of cardiomyocytes undergoing cell cycle activity (PBS 1.  Figure S8D-I).
Next, we employed the EHM model 41 with properties of postnatal human hearts 33 to assess structural and functional consequences of epicardial EV exposure. To simulate myocardial damage, we established a cryoinjury EHM model (Supplementary Video S1 and Figure S9A and B) and treated with human EPDC-derived EVs to determine effects on functional outcome (force generation). Cryoinjury itself induced %40%-60% decrease in force generation, that was sustained until D7 postinjury [Day 3 (D3) Cry 46%±23% of uninjured control (C), D7 Cry 57%±24%, P< _0.05] ( Figure 5A) concomitant with the loss of cardiomyocytes ( Figure 5B, right panels). This impaired function was detected in all cryoinjured groups irrespective of EV treatment until D3 post-injury ( Figure 5A). At D7 Cry, functional restoration to levels comparable in uninjured controls was detected in cryoinjured EHMs acutely treated with human EPDC-derived EVs. EV (þE) treatment in uninjured EHM did not significantly alter functionality (D7 Cry 57%±24%; D7 CryþE 123%±50%, P< _0.05) ( Figure 5A).   Interestingly, EVs alone evoked a strong increase in a-smooth muscle actin (a-SMA) expression in the fibroblast population at D3 after cryoinjury relative to vehicle -treated controls, indicating an activation of myofibroblasts 54 in the absence of an external inflammatory stimulus (C 2.74±1.05; Cry 1.74±0.69; E 5.25±0.08; CryþE 2.13±0.65 P< _0.01 for EV treatment vs. control, P< _0.001 for EV treatment vs. cryoinjury and Cryoinjury plus EVs vs. EVs alone) ( Figure 5D and E). Following injury there was a delayed increase in myofibroblast a-SMA expression, however, this was not dependent upon EV treatment (C 3.07±0. 68 Figure S9E and F).
These data collectively reveal that human epicardial EVs can stimulate an initial increase in cardiomyocyte proliferation in a model of injured human myocardium, which manifests longer term as improved contractile function.

Mouse and human epicardial EVs carry a conserved miRNA cargo
EVs from a wide range of sources have been described to carry a myriad of cargo, including microRNAs that have previously been implicated as potential mediators of regenerative signals in the context of CVD. 55 We performed microRNA sequencing of mouse and human primary cobble (epithelial) and spindle (mesenchymal)-derived EVs (data in Supplementary Table S1). Enriched microRNAs in epicardial EVs from different sources were consistent, suggesting conservation in microRNA cargo between mouse and human epicardial EVs ( Figure 6). Interestingly, three microRNAs: miR-99a-5p, miR-146a-5p, and miR-30e-3p, were present in mouse and human spindle (activated) epicardium, but absent from human cobble (inactive) epicardial EVs ( Figure 6).

miRNAs present in epicardial EVs
recapitulate the proliferative response exerted in cardiomyocytes by EV treatment and improve the functional outcome of cryoinjured EHM We initially tested the effects of individual identified miRNAs on the proliferation of hESC-derived cardiomyocytes. For this purpose, we generated AAV9 vectors expressing the pre-miRNAs for Homo sapiens (hsa)-miR-30a and hsa-miR-30e and for a series of other miRNAs derived from the cross-analysis of the epicardial EV miRNA cargo list with a database of human miRNAs, previously identified for their ability to induce proliferation in rat and mouse primary cardiomyocytes, assessed by EdU incorporation. 56 This published study validates the use of AAV-miRs for testing the ability to induce proliferative responses in hESCcardiomyocytes, as an initial step to triaging those miRs identified as EV cargoes herein for further in vivo testing in future studies.

Discussion
This study provides the first account of EVs produced by quiescent versus activated epicardium and an analysis of their functional roles in murine and human models of cardiomyocyte proliferation and heart injury. We describe a pro-proliferative effect of epicardial EVs on cardiomyocytes that implicates signalling from the injury-activated epicardium in regeneration. Other cell types have been shown to secrete EVs that promote modest cardiomyocyte cell cycle activity, although the majority are reported to exert cardioprotective effects. 24,25,28,57 Many of these prior studies, however, have focused on progenitor or stem cells that are controversial in terms of their origins and precise mechanisms underlying their transient beneficial effects. 58 The epicardium, due to its known trophic role in the developing heart, anatomical location and well characterized paracrine activity 9,59 represents a prime candidate celltype to reinvoke a proliferative response in cardiomyocytes.
We report efficient uptake of epicardial EVs by cardiomyocytes that promotes an increase in cell cycle activity (Figure 2 and Supplementary Figure S2), mediated via activation of the Akt pathway and partial activation of Hippo and ERK pathways (Supplementary Figure S3); all previously implicated in mediating cell cycle activity and proliferation in cardiomyocytes, [45][46][47][48] leading to improvement after MI 60 and regenerative responses. 46 Moreover, we show that this effect is also present in vivo after injury (Figure 3 and Supplementary Figures S4 and S5). In the P1 mouse heart injury model, an increase in proliferation of pre-existing cardiomyocytes has been previously reported as a driving mechanism for regeneration. 13,52 Our results suggest that epicardial EVs are able to augment this response in an organ-wide manner (Supplementary Figures   S4 and S5). Most strikingly, cardiomyocyte proliferation was also increased in EV-treated P7 non-regenerative, infarcted hearts 7 days after injury (Figure 3 and Supplementary Figure S5). This is significant, given the regenerative response has been reported to be lost in mice 1 week after birth (at P7), coinciding with cardiomyocyte cell cycle withdrawal. 13 Instead, injured P7 hearts adopt the default adult wound healing process of permanent scar formation. 13 However, despite evoking an early increase in cardiomyocyte proliferation in the EV-treated P7 infarcted hearts, we were unable to detect evidence of longer-term benefit; with no effect on scarring, vascularization or myocardial disarray to indicate more optimal repair or reduced pathological remodelling (Supplementary Figure S6). Regeneration is a complex process, in which different cell types contribute to tissue restoration at the expense of scarring and remodelling via simultaneous and synergistic roles. Thus, it is possible that even though epicardial EVs promote an increase in cardiomyocyte cell cycle activity in non-regenerative postnatal hearts, it is not enough to promote regeneration. This contrasts with the view that places cardiomyocyte proliferation as the major or exclusive driver of regeneration in both neonatal mice and adult zebrafish, 13,15,31,51,61,62 but reflects the fact that other factors, such as collateral blood vessel formation, 63 macrophages, 52 and extracellular matrix, 32,64 are also important contributors to heart regeneration. One major limitation of our study is the lack of functional in vivo data to assess whether the increase in cardiomyocyte cell cycle activity, mediated by epicardial EV treatment after injury, in a non-regenerative model is able to improve heart function regardless of a lack of full regeneration. This limitation can be addressed by adaptation of functional imaging to the neonatal model in future studies.
We also describe for the first time the release of EVs by human epicardial cells, derived from RAA patient biopsies and hESC-derived epicardial cells (Figure 4 and Supplementary Figure S7). Reports of EVs isolated from the pericardial fluid of patients undergoing aortic valve surgery 21 suggested putative EV secretion by epicardial cells; however, the cellular source was not identified. In this study, human epicardial-derived EVs induced a significant proliferative response in cardiomyocytes that mimicked that of mouse EVs. To analyse human EV effects related to myocardial injury, we established cryoinjury of three-dimensional EHM 41 constructs. 65,66 We ensured injury to a localized site by repeat application of the cryoprobe, which resulted in irreversible structural and functional damage at days 1-3 post-injury. Epicardial EV treatment of injured EHMs invoked significant cardiomyocyte proliferation, which resulted in functional recovery 7 days post-injury ( Figure 5 and Supplementary Figure  S9). The transfer of EV cargo is predicted to be a rapid event and, therefore, the delay in improved outcome, as observed at 7 days, likely reflects the need to invoke sufficient cardiomyocyte proliferation, subsequent maturation and functional integration within the injured tissue prior to impacting positively on contractile function.
EVs mediate their biological effects via specific cargoes, and have been reported to transport proteins, nucleic acids and lipids. 67,68 Their ability to transport microRNAs in particular has been a major focus regarding mechanisms of cell-cell signalling 28,57,69 We identified miRNA cargo in epicardial EVs by deep RNA-sequencing and tested whether miRNAs residing in mouse and human EVs might be responsible for inducing cardiomyocyte proliferation and the pro-regenerative response observed in the neonatal mouse and human EHM models. Interestingly, we identified a series of miRNAs ( Figure 6) (miR-99a-5p, miR-30e-3p, miR-30a-5p, miR-21-5p, miR-23b-3p, miR-181a-5p, miR-27a-3p, miR-100-5p, and miR-146a-5p) common to both epicardial sources, suggesting conserved function. Moreover, we observed a shift in miRNA content when comparing EVs from activated (spindle; mesenchymal) and inactive (cobblelike; epithelial) epicardial cells that induced different levels of responses on target cardiomyocytes. Specific miRNAs were also unique to spindle epicardial cell EVs (including miR-30e), which suggests a specific role in promoting a more potent proliferative response once the epicardial cells undergo EMT after activation. 5 From our data, no single miRNA exerted a proliferative response to the same extent as EV treatment; albeit individually they were able to increase proliferation in human ES-derived cardiomyocytes (Figure 7), suggesting a combinatorial and synergistic role for the cohort of identified cargo miRNAs. In addition, other putative unexplored epicardial EV cargoes, such as alternate nucleic acids (mRNA and lncRNA) and protein might also contribute to the overall EV-enhanced proliferative effect on cardiomyocytes. Interestingly, two of the miRNAs (miR-30 and miR-100) were also able to elicit a functional improvement in injured EHMs, comparable to EV treatment ( Figure 7G) suggesting more potent effects within human myocardial tissue.
Some of the most abundant miRNA species in epicardial EVs have previously been implicated therapeutically in the context of CVD. miR-21 secreted by bone marrow-derived cells has been reported to enhance tissue contractility in mature ischaemic adult cardiomyocytes. 70 Moreover, miR-21 has been shown to promote a cardioprotective effect via regulation of cell death and survival pathways, contributing to reduced myocardial damage. 71 Other highly represented miRNA species in epicardial EVs have not previously implicated in CVD, but share common downstream targets that could mediate cardiomyocyte proliferation: such as Tor (miR-30a, miR-100, and miR-30e) 72 and Notch (miR-21 and miR-30a 73 ). They have also been implicated in metabolic regulation and autophagic responses (miR-30a 74,75 and miR-27a 72 ), which can lead to increased proliferation and increased protection against cellular damage. 76 In conclusion, this study is the first to implicate EV release as a mediator of epicardium-derived paracrine signalling. Epicardial EVs from both murine and human sources have the potential to invoke cardiomyocyte cell cycle re-entry, both in vitro and in vivo, mediated by conserved miRNA cargo, which results in improved outcome in human heart injury models. Moreover, activated epicardial cells having undergone EMT present with a different EV signature as compared to when residing within the epicardium proper, with altered miRNA cargo that may increase myocardial proliferation. The focus herein has been on invoking cardiomyocyte proliferation to potentially extend the regenerative window in the neonatal mouse model, and in a more developmental model of EHM following injury; further studies are required to extrapolate the findings to the adult setting post-MI. Thereafter, understanding and targeting EV release from activated epicardium may represent a therapeutic opportunity for treating patients following acute MI.

Supplementary material
Supplementary material is available at Cardiovascular Research online.
OxStem Cardio. All other authors have no conflict of interest to declare.

Data availability
The data underlying this article will be shared on reasonable request to the corresponding author. The data underlying this article corresponding to miRNA sequencing are available in Gene Expression Omnibus.