Approaches to augment vascularisation and regeneration of the adult heart via the reactivated epicardium

Survival rates following myocardial infarction have increased in recent years but current treatments for post-infarction recovery are inadequate and cannot induce regeneration of damaged hearts. Regenerative medicine could provide disease-reversing treatments by harnessing modern concepts in cell and developmental biology. A recently-established paradigm in regenerative medicine is that regeneration of a tissue can be achieved by reactivation of the coordinated developmental processes that originally formed the tissue. In the heart, the epicardium has emerged as an important regulator of cardiac development and reactivation of epicardial developmental processes may provide a means to enable cardiac regeneration. Indeed, in adult mouse hearts, treatment with thymosin β4 and other drug-like molecules reactivates the epicardium and improves outcomes after myocardial infarction by inducing regenerative paracrine signalling, neovascularisation and de novo cardiomyocyte production. However, there are considerable limitations to current methods of epicardial reactivation that prevent direct translation into clinical practice. Here, we describe the rationale for targeting the epicardium and the successes and limitations of this approach. We consider how several recent advances in epicardial biology could be used to overcome these limitations. These advances include insight into epicardial signalling and heterogeneity, epicardial modulation of inflammation and epicardial remodelling of extracellular matrix.


INTRODUCTION
Coronary heart disease is a leading cause of worldwide morbidity and mortality 1 . Despite improved survival rates following myocardial infarction (MI), current treatments are unable to reverse loss of cardiac function after MI [1][2][3] . By harnessing modern concepts in cell and developmental biology, regenerative medicine could provide novel treatments to repair diseased hearts 4 .
The adult mammalian heart has long been regarded as non-proliferative and terminally differentiated but recent evidence has demonstrated that cardiomyocytes are capable of low level turnover, particularly after MI [5][6][7] . However, the limited proliferative capacity of the mammalian heart, beyond early neonatal stages 8 , is grossly insufficient to regenerate the ∼1 billion cardiomyocytes lost in a typical MI 9 . Moreover, regeneration of myocardial tissue is a complex process requiring appropriate integration of newly produced cardiomyocytes, production and integration of non-cardiomyocyte cell types, regenerative paracrine signalling, and moderation of inflammation that promotes excessive fibrosis. Therefore, even if cardiomyocytes could be easily replenished, such approaches, in isolation, are unlikely to induce sufficient regeneration.
Several different approaches to cardiac regeneration have been studied (Table 1), both in preclinical and clinical trials, described in 10 . These are principally based upon inducing cardiomyocyte cell cycle re-entry 11 , directly reprogramming cardiac-resident non-myocytes (fibroblasts) 12 or cell therapy-based strategies. The latter approach has applied a range of somatic 13 and haematopoietic 14 cell types, resident cardiac progenitor cells [15][16][17][18] , embryonic (ESCs) or induced pluripotent stem cells (iPSCs). Despite numerous multi-centre clinical trials of transplanted autologous bone marrow stem cells, transdifferentiation into cardiac muscle appears not to occur and any clinical improvement is modest 19 , attributable to the local release of paracrine factors that enhance repair, attenuate fibrosis and improve cardiac functional recovery. Resident cardiac progenitor cells were anticipated by many to confer superior benefit.
Whereas the findings from the SCIPIO trial appear promising 20 , concerns have been raised over some of the data 21 . Early evidence from the CADUCEUS trial 22 suggests that cardiosphere-derived autologous stem cells confer modest functional improvement. In contrast, recent studies in rodents, where lineage tracing technologies can be applied, revealed minimal contribution of c-kit+ cells to cardiac regeneration 23, 24 ; indeed, the value of c-kit as a marker of cardiac stem cell potential was questioned when c-kit positive cells were revealed to be an endothelial cell population with no cardiomyogenic contribution 24 . A recent proof-of-concept study in a non-human primate model highlighted the safety and potential for extensive remuscularisation using human ESCderived cardiomyocytes 25 . Given the ethical concerns that limit application of ESCs, the hopes of cardiac cell therapy may, therefore, depend upon improving functional maturation of iPSC-derived cardiomyocytes 26 .
An emerging paradigm in regenerative medicine is that repair of tissue can be achieved by reactivation of the coordinated developmental processes that originally formed the tissue 27 . Of note, the epicardium has been described as an important regulator of myocardial development and experimental treatments that reactivate embryonic epicardial processes improve outcomes after an experimental model of MI in adult mice 28 . However, these treatments have major limitations and are unlikely to be clinically useful in their current form. We discuss the rationale for targeting the epicardium and explore the successes and limitations of pre-clinical studies that demonstrate proof-of-principle of epicardial reactivation for cardiac regeneration. Induce proliferation of mature CMs to replace CMs lost in heart disease.
• Autologous (eliminates risk of rejection and requirement for immunosuppression).
• No requirement for stimulating CM differentiation and maturation.
• New CMs are likely to have good mechanical, vascular and electrical integration.
• Potential off-target effects of treatment (particularly oncogenesis).
• Difficult to achieve the magnitude of required CM proliferation for clinically meaningful benefit.
• Inherent difficulties of gene therapy approaches -non-genetic approaches (eg. paracrine factors) more desirable.
• Progressive improvement in infarct size and cardiac functional parameters occurred in cyclin D2 transgenic mice 29 .
• Activation of the Hippo/YAP promitogenic signalling pathway improved heart function and survival after MI 11 .
• Inhibition of miR-15 induced cardiac proliferation and modestly improved cardiac function 31 .
• In a small clinical trial, infusion of human recombinant NRG-1 was welltolerated and improved cardiac function. NRG-1 has pro-proliferative effects via the ErbB2/ErbB4 receptor -oncogenic potential is an important concern 32 . • Deletion of Meis1, a cardiomyocyte cell cycle regulator, extends the postnatal window of proliferation and regeneration, although the ability to induce the effect in the adult heart was not reported 33 .
• Reconstituting the potent cardiogenic activity of FSTL1 in an epicardial patch promoted myocardial regeneration following MI, in mouse and pig 34 .
(continued on next page) • Successful long-term engraftment of substantial numbers of ESC-CMs has been achieved in animal models.
• Highly specific effects.
• Possibility of teratoma formation.
• Stem cell-derived CMs are immature -improper mechanical, electrical and vascular integration.
• Human ESC-CMs could be produced on a clinical scale. Delivery of human ESC-CMs to infarcted non-human primate hearts produced extensive remuscularisation 25 .
Isolate resident or non-resident cardiac progenitor cells, expand in vitro and deliver to the myocardium.
• Autologous CPCs can be obtained from biopsies collected during surgery.
• Clinical trials have demonstrated safety (but safety might be compromised at the higher engraftment rates that would be desirable for improved treatment).
• Limited efficacy and inconsistent results in clinical trials.
• Unclear mechanism of action -paracrine effects are likely to be important, in which case paracrine factor therapy may have advantages over cell-based therapy.
• CMs may not appropriately integrate.
• In a randomised, open-label phase 1 trial ('SCIPIO'), c-kit + CPC-treated patients had a small improvement in LVEF 20 . However, The Lancet has issued an ''expression of concern'' regarding data integrity. 21 . In independent murine studies, c-kit + CPCs made only minimal 23 , or no 24 , contribution to CMs.
• In the 'CADUCEUS' trial, intracoronary infusion of autologous CDCs after MI was safe and reduced scar mass and regional contractile dysfunction; however the patient cohort was small 22 .
(continued on next page) • Potential to regenerate a range of cell types in addition to CMs.
• Cell-free (less costly, fewer logistical difficulties and perhaps safer than cell-based therapies).
• Robust regenerative technique not yet described.
• ''Priming'' with Tβ4 activated the epicardium, resulting in de novo CMs and improved functional cardiac parameters 28 .
• Intra-myocardial injection of modRNA encoding VEGF-A enhanced mobilisation of epicardial progenitor cells and improved heart function and survival in mice 35 .
• In vivo injection of HGF and IGF-1 into murine hearts to mobilise and amplify resident c-kit + CPCs improved function and survival after MI 36 .

Direct reprogramming of non-cardiomyocytes
Use cardiac reprogramming factors to induce conversion of non-cardiomyocytes (e.g. fibroblasts) into iCMs in vivo.
• Cell-free (less costly, fewer logistical difficulties and perhaps safer than cell-based therapies).
• Requires gene transfer by integrating viruses.
• Functional integration of iCMs is not established.
• Difficult to confirm fibroblast-to-iCM reprogramming in vivo in patients.
• Requires development of appropriate models for translation into humans.
• Retroviral gene transfer and expression of Gata4, Mef2c, and Tbx15 (GMT) 12 or additionally Hand2 (GHMT) 37 induced reprogramming of fibroblasts into CMs and improved cardiac function after MI in mice.
Moreover, we consider how recent evidence highlighting several novel concepts related to epicardial biology could help to overcome previous limitations. These avenues for future research may aid progress towards the ultimate aim of clinical induction of cardiac regeneration in the adult human.

WHY TARGET THE EPICARDIUM?
The epicardium is essential for mammalian cardiac development The epicardium develops principally from the mesodermal proepicardial organ soon after cardiac looping 40,41 . The epicardium then makes essential cellular and signalling contributions to cardiac development ( Figure 1).
The cellular contribution is mediated by a subpopulation of epicardial cells that undergoes epithelial-to-mesenchymal transition (EMT) to produce epicardium-derived cells (EPDCs). EPDCs then migrate into the myocardium and differentiate into several cardiac lineages 41 . In development, it is widely accepted that EPDCs make essential contributions to cardiac interstitial fibroblasts, along with adventitial fibroblasts, vascular smooth muscle cells and pericytes of the coronary vessels 41 . However, the contribution of EPDCs to cardiomyocyte and coronary endothelial lineages is more controversial. Modern fate map studies primarily use Cre-LoxP-based genetic lineage tracing, whereby Cre recombinase is expressed with a lineage-specific gene to induce genetic recombination and consequent lineage-specific expression of a reporter gene (eg. enhanced green fluorescent protein or β-galactosidase).
These genetic changes are inherited by daughter cells ('indelible labelling') and can be assessed by histological techniques. Using Tbox18 (Tbx18) and Wilms' tumour-1 (Wt1) as epicardial marker genes for such genetic lineage tracing, a subset of cardiomyocytes was found to be derived from Tbx18 + or Wt1 + cells 42,43 . However, an important consideration in genetic lineage tracing is the specificity and sensitivity of the marker gene. In this regard, Tbx18 is expressed in epicardial cells (high sensitivity) but is also expressed in the interventricular septum and left ventricular cardiomyocytes (low specificity) 44 . Similarly, a tamoxifen-inducible Wt1CreERT2 drives incomplete epicardial recombination (low sensitivity) and Wt1 is reported to be expressed in coronary endothelial cells (low specificity) 45,46 . Since Tbx18 and Wt1 expression is not strictly limited to the epicardium, cells labelled by such genetic lineage tracing could derive from sources other than epicardial cells, and therefore the epicardial contribution to the cardiomyocyte lineage has been disputed 44,45 .
Analogous inadequacy of marker genes may underlie the controversy surrounding the (pro)epicardial contribution to the coronary endothelium 47 . A substantial contribution to the coronary endothelium is made by recently-identified proepicardial sub-populations that express neither Wt1 nor Tbx18 47 , thus, previous studies using these marker genes did not include the relevant sub-population (low sensitivity) and therefore the previous conclusion of a lack of (pro)epicardial contribution to the coronary endothelium may be incorrect [42][43][44][47][48][49] . While alternative sources, notably the sinus venosus 48 and endocardium 50 , have also been described, those endothelial cells contributed from the proepicardium appear to derive from a distinct subpopulation, characterised by expression of Scleraxis (Scx ) and Semaphorin3D (Sema3D) 47 .
Conversely, the myocardium also signals to the epicardium to control epicardial activities, such as EMT and migration and differentiation of EPDCs 27,68 . This occurs by myocardial secretion of signalling factors, which include FGF family members 69 , transforming growth factor β (TGFβ) 70,71 , vascular endothelial growth factors (VEGFs) 72 , thymosin β4 (Tβ4) 73 and platelet-derived growth factor (PDGF) 61,74 . Thus, the epicardium and myocardium form a reciprocal signalling unit that is essential for cardiac development.

The epicardium is essential for non-mammalian cardiac regeneration
The epicardium is not only important in mammalian cardiac development but is also critical for cardiac regeneration in species such as zebrafish, which are able to fully recover after substantial cardiac injury. During zebrafish cardiac regeneration, the epicardium re-expresses embryonic genes Retinaldehyde dehydrogenase 2 (Raldh2), a regulator of retinoic acid synthesis, and Tbx18 resulting in epicardial EMT and EPDC migration to vascularise the myocardium 75 . By contrast, in adult mammals, epicardial proliferation and signalling occurs after injury but, without therapeutic intervention, this is unable to invoke substantial regeneration 76, 77 .

Rationale for targeting the epicardium
Epicardial programmes underpin cardiac development and re-activation of these programmes mediates regeneration in zebrafish. However, mammalian epicardial reactivation requires enhancement to be therapeutically useful. Importantly, the epicardium can uniquely contribute to coordinated repair by providing a range of cell types and signalling factors.

PRIOR SUCCESSES AND LIMITATIONS IN REACTIVATING THE QUIESCENT ADULT MAMMALIAN EPICARDIUM
Thymosin β4 treatment improves outcomes after myocardial infarction in mice Thymosin β4 (Tβ4) is a 43-amino acid actin monomer-binding peptide, which is important for both systemic and coronary vascular development 73,78 . In the adult mammalian heart, Tβ4 treatment stimulates reactivation of the epicardium after MI and results in regenerative neovascularisation 79 , recapitulating its embryonic role. Furthermore, 'priming' by treatment with Tβ4 before MI enables EPDCs to form de novo cardiomyocytes 28 , and induces a robust neovascularisation 79,80 . In the short time frame of 28 days after MI, Tβ4 treatment reduced infarct volume and improved ejection fraction 28 . New cardiomyocytes must integrate properly into the myocardium in order to maintain structural integrity and avoid arrhythmogenic electrophysiological heterogeneity. After Tβ4 stimulation, epicardium-derived cardiomyocytes formed adherens and gap junctions, suggesting structural integration 28 ; moreover, functional integration of the de novo cardiomyocytes was demonstrated by synchronous [Ca 2+ ] i transients between de novo and pre-exisiting cardiomyocytes 28 . However, given the complexities of coupling between cardiomyocytes, as well as other functional links such as proposed cardiomyocyte-fibroblast coupling 81 , more comprehensive assessment of de novo cardiomyocyte integration might be required in order to conclude complete integration.

Limitations of Thymosin β4 and other treatments
EPDCs from the adult mouse after Tβ4 treatment are not molecularly identical to embryonic EPDCs, despite functional similarities 82 . Reactivated adult EPDCs display a heterogeneous molecular profile, defined by both cardiac progenitor and mesenchymal stem cell markers, including Sca-1, CD29, CD90, PDGFRβ and CD44, thus are significantly different from their embryonic counterparts obtained at mouse embryonic day 12.5 (E12.5), despite the common expression of the early embryonic epicardial gene Wt1 82 . Consequences of the molecular differences are unclear but this finding suggests that Tβ4 priming does not achieve true recapitulation of embryonic processes, which perhaps limits maximal therapeutic benefit.
In contrast to Tβ4 treatment before MI, Tβ4 treatment after MI does not produce de novo cardiomyocytes, although it does confer modest benefits, probably via paracrine effects 83 . This is a considerable limitation for translation into clinical practice because it presents a requirement for prophylactic treatment in order to regenerate de novo cardiomyocytes after MI. Although such prophylactic treatment is possible, particularly if at-risk patients were targeted, widespread long-term Tβ4 treatment may give cause for concern over issues such as safety, cost, patient compliance, and feasibility of frequent administration by necessarily non-oral routes.
Perhaps the most important limitation of Tβ4 treatment is low efficiency of de novo cardiomyocyte production from EPDCs (∼0.59% of Wt1 + EPDCs) 28 . In addition to Tβ4, two other drug-like molecules, prokineticin-2 and VEGF-A modified RNA (modRNA), are also able to reactivate the epicardium but they too have important limitations 35,84,85 . On balance, these experimental treatments demonstrate the principle of epicardial reactivation for cardiac regeneration but several difficulties prevent direct translation into clinical practice.

NOVEL RESEARCH AVENUES THAT COULD IMPROVE EPICARDIAL REACTIVATION STRATEGIES
Understanding the regulatory pathways that restrict epicardial cell behaviour Extrapolating from our understanding of epicardial processes in development and from recent insights into their redeployment for regenerative benefit by zebrafish following injury, we can clearly appreciate the shortcomings of the cardiac healing response in mammals. Therapeutically instilling developmental or cardioregenerative mechanisms into the refractory mammalian heart may support a greater degree of muscle regeneration. As far as engaging epicardial involvement in regeneration is concerned, at least three key processes need to be targeted: (i) reactivation and restoration of pluripotency; (ii) EMT and inward migration; (iii) cell fate determination. While partial success in stimulating each of these steps has been achieved, notably with Tβ4, the extent, as previously stated, is inadequate, particularly of EMT and in directing cardiomyocyte cell fate. While screening for potent small molecules 86 may lead to a breakthrough, candidate approaches, based on understanding embryonic and zebrafish mechanisms, may also prove fruitful. Taking EMT as an example, the principal drivers of epicardial EMT in the embryo are FGFs, notably basic FGF and TGFβ 68 , as discussed above. Furthermore, the regenerative capacity of zebrafish is FGF-dependent; expression of a dominant-negative FGF receptor blocks epicardial EMT, neovascularisation and regeneration 75 . Identifying suitable targets to achieve enhanced activation of these pathways in epicardial cells may be beneficial. Understanding the pathways that regulate cardiac cell differentiation during development may also reveal novel targets to enhance the complement of pro-regenerative cell types, at the expense of the predominant fibroblast fate.

Understanding the intrinsic response of the epicardium to myocardial injury
Despite its recognised failure to regenerate, the adult mammalian heart displays an intrinsic response to injury which, to date, has only been superficially characterised. Whilst scarring, ventricular dilatation and hypertrophic responses are widely recognised 87 , other elements have been largely overlooked. One such endogenous response is the reactivation of the epicardium, in the form of re-expression of embryonic epicardial genes 28 and expansion via proliferation 28 and infiltration of haematopoietic cells 40 . These responses would logically appear to be beneficial, an attempt at self-repair, for example by secretion of pro-regenerative paracrine factors 77 and induction of new vessel growth 79 . An appreciation of what these pathways contribute, and how they may be enhanced, may facilitate a greater degree of regeneration. However, not all intrinsic responses constructively influence repair, fibrosis being a prime example; while moderate fibrosis is essential, at least initially, to prevent cardiac rupture, excessive fibrosis leads to permanent scarring at the expense of myocardial replenishment 87 . Although some embryonic epicardial processes are redeployed 27 , others are either ineffectively induced or even actively suppressed. A notable, recently-identified example of this is expression of the secreted protein, Follistatin-like 1 (FSTL1) 34 . FSTL1 is a potent cardiogenic factor that is actively expressed by embryonic and adult epicardial cells. Curiously, epicardial FSTL1 declines following myocardial infarction but application of human FSTL1 via an epicardial patch was able to stimulate cardiomyocyte proliferation, to improve cardiac function and survival, in mouse and swine MI models 34 . These findings suggest that the loss of epicardial FSTL1 is a maladaptive response to injury, and that its restoration can reverse myocardial death and remodelling following MI. Thus, understanding endogenous responses empowers regenerative strategies but requires additional insight into whether to therapeutically enhance or override intrinsic mechanisms.

The epicardium is a complex heterogeneous structure
The epicardium was, for a long time, regarded as a simple mesothelial layer with little heterogeneity, but recent work indicates hereto unappreciated complexities 41,47 . Epicardial sub-populations have been classified according to factors such as activation by signalling molecules (e.g. Notch or Tβ4) or expression of molecular markers (e.g. Wt1, Sca-1 or c-kit) 27,28,82,88 . However, one limitation of the literature is that different studies often assess different activating factors, molecular markers or developmental time points. The extent of overlap between different populations reported in separate studies is, therefore, not immediately apparent. For example, there is considerable molecular heterogeneity among the EPDCs that are reactivated by Tβ4 82 and it is therefore likely that some of the Tβ4-stimulated epicardial cells could also be classified into subpopulations described elsewhere on the basis of other distinguishing factors. A detailed understanding of the different cell types that populate the epicardium might allow for more precise targeting of sub-populations relevant to regeneration. The first study to systematically characterise epicardial heterogeneity, using a single cell transcriptomic approach, confirmed that at least three distinct sub-populations of tcf21+ epicardial cells exist in zebrafish 89 . The specific functional roles of these sub-populations in cardiac development, homeostasis and regeneration remains to be explored and, crucially, the extent to which this heterogeneity is conserved, or potentially more complex, in mammals remains to be determined.
Assessment of murine epicardial structure has recently revealed additional heterogeneity in the form of clusters of CD45 + Wt1 − haematopoietic cells, encased in ECM, which resemble stem cell niches 40 . After MI, the CD45 + cells proliferated, encapsulating ECM was degraded by matrix metalloproteinases, and proliferative CD45 + cells were released into the underlying myocardium. Although the precise extent of this movement into the myocardium was not assessed, these results hint at a possible role for the CD45 + cells that deserves further investigation 40 , specifically to investigate whether the CD45 + cells contribute adversely or beneficially to cardiac regeneration. This study both challenges the prevailing dogma that the epicardium derives solely from the proepicardial organ and identifies a novel sub-population of epicardial cells that might be a useful therapeutic target.

Controlling the inflammatory response to cardiac injury
In zebrafish and newts, regeneration occurs with only short-lived fibrosis. By contrast, MI in adult mammalian hearts results in inflammation, fibrosis and permanent scarring 90 . Regeneration of the adult mammalian heart may, therefore, require moderating the response away from excessive inflammation and fibrosis, which preclude regeneration, and towards cellular regeneration and integration.
Although use of Tβ4, prokineticin-2 and VEGF-A modRNA to reactivate epicardial developmental programmes is beneficial, it does not necessarily follow that epicardial activity, in the absence of developmental reactivation, is beneficial. Indeed, it was reported that epicardial activity may underlie, at least in part, the inflammatory fibrosis of the endogenous response to MI 91 . After identifying the CCAAT/enhancer binding protein (C/EBP) transcription factor family as important in epicardial activation, it was found that viral gene transfer of dominant-negative mutant C/EBP improved ejection fraction and fibrosis after MI and reduced local neutrophil count; however, the causal link between lowering the inflammatory infiltrate and improving outcome was not directly demonstrated 91 .
Thymosin β4 may also be useful for controlling inflammation. One study found that treatment of mice after MI with a biologically-occurring form of Tβ4 (Tβ4-sulfoxide) increased the local macrophage count at day 2 post-MI but reduced the macrophage count at day 7 compared to control 92 . In combination with the finding of reduced infarct volume after Tβ4 treatment, these results were interpreted to mean that Tβ4-sulfoxide hastens early phagocytic debris removal after MI and enhances subsequent clearance of immune cells. Although the epicardium may not have a direct role in these anti-inflammatory effects, these findings, along with its proven role in cardiomyocyte protection 93 , suggest that Tβ4 might be clinically useful for prevention of cardiac scarring, even if its role in stimulation of production of de novo cardiomyocytes by EPDCs is limited 80,92,94 .
The concept of beneficial effects of early macrophages in cardiac regeneration is supported by recent evidence showing that macrophage depletion prevents the regeneration that normally occurs in neonatal mice after MI 95 . The pro-regenerative macrophages had pro-angiogenic effects, although the underlying mechanisms remain unclear. Molecular profiling of the early macrophages revealed no clear bias towards M1-type or M2-type macrophages, instead pointing to a transient, pro-regenerative phenotype and secretion of distinct soluble factors that may facilitate myocardial regeneration. Although these findings identify a potentially important new macrophage subset, the non-canonical characteristics of the pro-regenerative neonatal macrophages might limit translation into clinical therapy 96 .
Thus, inflammation, fibrosis and scarring remain as major obstacles to cardiac regeneration, and the underlying mechanisms are complex. Improved understanding of the inflammatory response to cardiac injury, including the role played by the epicardium, might enable targeted modification of adverse fibrosis while preserving beneficial debris removal by macrophages 97,98 .

Epicardium-controlled remodelling of the extracellular matrix
The ECM is a dynamic network of fibrous and non-fibrous proteins that can control cell function through several signalling mechanisms 65,99 . In development, the ECM has important effects on the epicardium and elsewhere in the heart 99 . For example, binding of epicardial α4-integrin to ECM ligands in the sub-epicardium inhibits EMT and migration, and influences epicardial cell differentiation 66 . Moreover, remodelling of the ECM recapitulates embryonic programmes to facilitate regeneration of limbs, tails and fins in fish and amphibians [100][101][102][103] .
Recently it was suggested that dynamic spatiotemporal ECM remodelling by the epicardium plays an important role in cardiac regeneration in zebrafish and newts 104,105 .
In situ hybridisation and transgenic reporter analyses revealed that deposition of fibronectin-1 and fibronectin-1b by epicardial cells in zebrafish was dynamically upregulated during heart regeneration 104 . Furthermore, concomitant expression of integrin β3 and αV also occurred in cardiomyocytes, perhaps facilitating the fibronectin signalling to enhance mobilisation and integration of cardiomyocytes. Importantly, fibronectin-1-defective zebrafish displayed impaired myocardial regeneration and increased fibrosis 104 .
These findings suggest that the ECM acts as an important intermediate messenger for regenerative epicardium-to-myocardium signalling. Similarly, DNA microarray profiling and subsequent gene ontology analysis has identified an enrichment of genes associated with ECM in the regenerating hearts of newts and zebrafish 105 . Of particular interest were the ECM components hyaluronic acid, tenascin C and fibronectin, which were upregulated in the epicardium of the regenerating newt heart. Tenascin C induced cell cycle re-entry of primary cultured newt cardiomyocytes and may facilitate migration with 'counter-adhesive domains'. It was speculated that the 'regeneration-specific matrix' may also facilitate migration by altering tissue stiffness through hyaluronic aciddependent ECM hydration 105 .
However, while these studies show that ECM remodelling is important in newt and zebrafish cardiac regeneration, the implications for adult mammalian non-regenerative hearts are unclear. Zebrafish heart regeneration occurs by cardiomyocyte dedifferentiation and proliferation, both around the injury site and throughout the whole organ, as well as by cardiomyocyte migration [106][107][108][109] . This differs mechanistically from the adult mammalian response to cardiac injury, even if partial regeneration is induced. It is therefore difficult to extrapolate between species. On the other hand, comparative gene ontology suggested that species differences in ECM remodelling may be partly responsible for the different regenerative capacities 105 . Modification of injury-induced epicardial ECM remodelling might therefore present a method of converting the characteristics of the non-regenerative response towards those seen in regenerative hearts. Tissue engineering of pro-regenerative biomimetic matrices might be one approach to exploiting these novel findings 105 .

The role of pericardial fluid
The pericardial sac enhances epicardial activation by constraining heart-derived signalling factors in close proximity to the epicardium 88 . Myocardially secreted FGF-1 110 and FGF-2 levels 111 in the pericardial fluid of MI patients were found to correlate with the severity of ischaemia and a possible role in mediating collateral growth, although a direct involvement of the epicardium in this process has not been explored.
A recent study in patients profiled the miRNA content of pericardial fluid and identified a number of previously implicated heart failure markers 112 ; this was interpreted to be an active and selective paracrine mechanism, involving non-coding RNAs as well as growth factors, to mediate cross-talk between cardiac cell types, which likely includes epicardial cells. Administration of pericardial fluid into the pericardial cavity of non-infarcted mouse was sufficient to induce epicardial proliferation and re-expression of embryonic genes 88 .
Insulin-like growth factor 1 (IGF1), hepatocyte growth factor (HGF) and High mobility group box 1 protein (HMGB1), factors known to induce resident cardiac progenitors, were significantly elevated in pericardial fluid from MI patients 88 . Thus, pericardial fluid contains trophic factors which might be harnessed to invoke epicardial activation and repair. In a recent study, clusterin, was found to be secreted in exosomes of MI patients; addition of clusterin to the pericardial sac of mice post-MI enhanced epicardial EMT and arteriogenesis and led to improved cardiac function 113 . These studies demonstrate that intra-pericardial injection may be an effective delivery method for any future treatments to augment epicardial contribution to myocardial regeneration 114 .

Translation into humans
The preclinical experiments described above have necessarily used non-human animal models, with the non-regenerative adult mouse heart implicitly assumed to be similar to adult human hearts. Nevertheless, experiments on human epicardial cells are required for translation of therapies into clinical practice. Several studies have used cultured primary epicardial cells, taken from right atrial appendages of human patients during right coronary artery bypass 115, 116 ; in vitro characterisation of the cells demonstrated basic similarities between mouse and human. Some species similarities were also noted in an in situ analysis of primary embryonic and foetal tissue samples 117 , in terms of marker expression and, broadly speaking, in its formation and apparent role during development. However, consistent differences were also reported between species both in fetal and adult epicardium; the embryonic human epicardium is not a simple squamous epithelium, as previously reported 118 . Instead, the external layer of flat mesothelial cells overlies a thin basal lamina with an underlying layer of connective tissue, the subepicardial space, containing elastic fibres as well as large vessels. In the adult human myocardium, the subepicardial space consists mainly of adipose tissue, which surrounds coronary vessels. By contrast, murine epicardium does not contain adipose tissue and comprises only a monolayer of mesothelial cells on a thin layer of connective tissue formed from elastic fibres 118 . These species differences suggest that experiments on human tissue may be important for translation from animal models to humans.
Furthermore, at least in the fetal human epicardium, atrial-ventricular differences in cellular behaviour were reported between epicardial cells. The ventricular, but not atrial, epicardium exhibited greater cell alignment and spindle-like morphology and ex vivo ventricular cells spontaneously differentiate and lose epicardial identity, whereas atrialderived cells remained more 'epithelial-like' 117 . The utility of cultured human atrial EPDCs as a model may therefore be limited not only by low availability but also by dissimilarity to the epicardial cells that may be stimulated in vivo. If atrial epicardial cells differ from ventricular epicardial cells, cells from atrial appendages may not necessarily represent the population of cells that partakes in the majority of reactivation after MI, albeit the extent of contribution in vivo from the individual chambers has not been addressed.
In an attempt to overcome these difficulties, epicardial-like cells have been generated from human pluripotent stem cells (hPSCs) 119 . One approach used stage-specific activation of bone morphogenetic protein (BMP) and WNT signalling pathways to generate cells with epicardial morphology expressing the epicardial markers WT1, TBX18 and RALDH2. These epicardial-like cells could be induced to undergo EMT to produce populations displaying characteristics of fibroblast and vascular smooth muscle cell lineages 116 . An independent approach used a two-stage process to generate similar epicardial-like cells. The hPSCs were first converted to lateral plate mesoderm and then differentiated into epicardial-like cells by stimulation of BMP, WNT and retinoic acid signalling pathways 120 . These approaches can provide unlimited sources of cultured human epicardial-like cells which could be used to optimise potential treatments or could be combined with chemical genetic screening techniques to identify novel regenerative compounds and provide insight into poorly-understood signalling pathways 86,121 .

CONCLUSION
High rates of cardiovascular disease mean that safe, effective and feasible cardiac regeneration could transform clinical practice. Reactivation of epicardial developmental processes is an attractive approach to coordinated regeneration of the heart and, if refined based on emerging concepts in epicardial biology, could provide novel strategies for regenerative epicardial reactivation.