Next generation of heart regenerative therapies: progress and promise of cardiac tissue engineering

The adult heart is a vital and highly specialized organ of the human body, with limited capability of self-repair and regeneration in case of injury or disease. Engineering biomimetic cardiac tissue to regenerate the heart has been an ambition in the field of tissue engineering, tracing back to the 1990s. Increased understanding of human stem cell biology and advances in process engineering have provided an unlimited source of cells, particularly cardiomyocytes, for the development of functional cardiac muscle, even though pluripotent stem cell-derived cardiomyocytes poorly resemble those of the adult heart. This review outlines key biology-inspired strategies reported to improve cardiomyocyte maturation features and current biofabrication approaches developed to engineer clinically relevant cardiac tissues. It also highlights the potential use of this technology in drug discovery science and disease modeling as well as the current efforts to translate it into effective therapies that improve heart function and promote regeneration.


INTRODUCTION
Cardiovascular diseases (CVDs) have been recognized as a major concern for global health among noncommunicable diseases 1 , and despite half-century of advances in cardiovascular science and preventive medicine, CVDs continue to yield high mortality and morbidity rates worldwide 2 . Revascularization interventions or pharmacological treatment allow in many cases to salvage the heart and prevent transplantation, even though such strategies are not recognized as a definitive cure 3 .
The adult human heart has a limited regenerative capacity 4 and so regenerative medicine therapies have shed new hopes to repair or replace damaged hearts. The marginal benefits reported by cellbased [5][6][7][8] and cell-free [9][10][11] approaches have grossly been associated with inappropriate delivery and retention, as well as ineffective therapeutic efficacy 12 . In the past decades, tremendous advances have been witnessed in the field of tissue engineering (TE), especially concerning stem cell engineering, the development of functional biomaterials and biomimetic scaffolds, and the implementation of biofabrication tools for the generation of complex biological structures with high resolution 13 . This progress has set a solid background to rethink current therapeutic approaches and design new bioengineering therapies 14 . Nonetheless, creating fully matured and functional cardiac tissue in vitro is still challenging. Despite this, it is becoming possible to robustly engineer miniaturized tissue versions able to compete, if not replace, commonly used platforms to assess drug safety and model human physiopathology to guide efficient drug discovery 15 .
This review highlights recent progress in cardiac stem cell bioengineering, with a particular focus on the identification of essential requirements for the design of clinically relevant human cardiac tissues: (i) the cell sources, (ii) bioinspired strategies to generate cardiomyocytes (CMs) with improved maturity, and (iii) biofabrication methodologies. The potential use of this technology in drug discovery science and disease modeling as well as the current efforts to translate it into effective heart regeneration therapies is also discussed.

CELL SOURCES TO ENGINEER CARDIAC TISSUE
Cardiac muscle is the functional unit of the heart, as it is involved in force generation and propagation of electrical signals through the CM-rich tissue to allow rhythmic pump contraction. Human cardiac muscle is composed of various cell types, the most abundant being CMs, pericytes, endothelial cells, and fibroblasts. The proportion of each cell population in the heart is still a matter of discussion, but there are evidences that CMs occupy the majority of the heart volume (roughly 75%) and account for 30-40% of all cells 16 . State-of-the-art analyses of single-cell/nuclei RNA-sequencing transcriptomic data of six heart regions across several unrelated individuals have clarified that CMs are the most prevalent cardiac cell population in the heart, while also highlighting a surprising cellular heterogeneity among CM, pericyte, endothelial, and fibroblast cell populations 17 . It appears that the cellular composition of cardiac muscle and its transcriptional signature varies across the heart in order to ensure that each anatomical region is highly specialized in a given function. Such complexity undeniably elevates the challenge of engineering artificial cardiac muscle.
Typically, engineering a contractile cardiac muscle requires foremost a source of functional CM. The first engineered cardiac tissue models relied on nonhuman cell cultures of different species, including cell lines of rat embryonic myoblasts (H9c2 18 ) and mouse CMs (HL-1 19 ) as well as primary CMs isolated from chicken embryos 20 and neonatal rats 21 . Despite limitations regarding human tissue availability, human adult CMs can be isolated from myocardial biopsies 22 ; however, soon after isolation, CMs undergo a profound structural and functional remodeling, leading to cell dedifferentiation and loss of viability 23 . Recently, applying near-physiologic electromechanical stimuli during 24 h proved capable of inactivating CM remodeling after isolation 24 . Long-term culture (up to 4 months) of adult human myocardial muscle strips subjected to biomimetic stimuli similarly prevented pronounced CM remodeling, even though changes at the structural, mechanical, and gene expression levels were detected 25 . Immortalized adult human ventricular CM cell lines have also been established by the fusion of primary cells with SV40-transformed human fibroblasts to create cardiac models 26 , but their poor recapitulation of human physiology discourages its use.
Due to their high self-renewal capacity and ability to generate any cell type of the human body, pluripotent stem cells (PSCs) are becoming an attractive alternative for cardiac TE. These PSCs can either be derived from the inner cell mass of the blastocyst, named embryonic stem cells (ESCs) 27 , or by forced expression of pluripotency genes through the delivery of the "reprogramming factors" Oct3/4, Sox2, Klf4, and c-Myc to somatic cells, termed induced PSCs (iPSCs) 28,29 . In particular, human iPSC (hiPSC) derivatives better capture patient-specific physiology 30,31 , besides surpassing the ethical concerns associated with human ESC (hESC). Acquired knowledge on stem cell biology together with the advances in PSC technology has allowed the establishment of robust differentiation protocols capable of obtaining every subtype of CM (ventricular, atrial, and pacemaker) with a high degree of purity 32 . Briefly, temporal modulation of the Wnt/ β-catenin signaling pathway is usually performed by first promoting activation for mesoderm formation and subsequent inhibition for cardiac specification 33,34 . The use of small molecules (e.g., CHIR99021, IWP2, IWR1) to control the Wnt/β-catenin signaling pathway is currently preferred to growth factors/ recombinant proteins due to their better diffusion properties and moderate cost 32,35,36 .
A lot of effort has been put to evaluate the maturity state of human PSC-derived CMs (hPSC-CMs) and it is unanimous that, after differentiation, these cells are more similar to fetal rather than adult CMs (Fig. 1). Adult ventricular CMs display a remarkably organized structure with sophisticated intracellular organelles optimized for contraction. Among these components are densely packed aligned sarcomeres 37,38 within a hypertrophic 39 (usually) binucleated 40 cell; the sarcolemma and a vast network of transverse tubules (T-tubules) 41,42 along the Z-lines to allow for fast depolarization; a functional longitudinal sarcoplasmic or endoplasmic reticulum (SR/ER) and terminal cisternae positioned near T-tubules adapted for quick calcium kinetics 43 ; a high density of mitochondria near the contractile apparatus to improve ATP consumption 44 ; and intercalated discs with well-developed cell junctions to promote electrical coupling 45,46 . hPSC-CMs have not shown this level of cellular complexity [47][48][49][50][51][52] , as they are less fit for contraction 53,54 and are incapable of matching the metabolic demands of adult CMs 55-57 needed for significant force generation [58][59][60][61] , which is apparent in their negative force-frequency relationship (FFR) 62,63 and minimal post-rest potentiation (PRP) 62 . Due to a dissimilar ion-channel density, hPSC-CMs exhibit a distinct action potential (AP) profile from adult CMs 64,65 , resulting in asynchronous and spontaneous beating rate 66 , less negative resting membrane potential (RMP) 67,68 , slower upstroke velocity 69 , and irregular AP duration 67,69,70 . In fact, the immature ultrastructure of hPSC-CMs not only affects calcium handling 71 due to an under-developed excitation-contraction coupling (ECC) machinery 50,51,62,72 but it also slows AP propagation 73,74 owing to the lack of organized electromechanical junctions compared to adult cardiac tissue 75,76 . Furthermore, these differences are seen at a transcriptomic level, with hPSC-CMs showing preferential expression of fetal sarcomeric gene isoforms (myosin heavy chain: MYH6 > MYH7; troponin I: TNNI1 > TNNI3; titin: TTN-N2BA > TTN-N2B) as well as lower expression of genes related with electrophysiological function (e.g., KCNJ2, KCND3, KCNJ8, KCNH7, SCN5A, HCN4, GJA1, JPH2) and ECC (e.g., CACNA1C, RYR2, ATP2A2, CASQ2, SLC8A1, PLN, ITPR3, CAV3, BIN1) 77 .
The maturation defects of hPSC-CMs have gained increased attention over the past decade since their immaturity greatly limits their application in several areas of precision medicine. A better understanding of cardiac development has allowed the identification of relevant players that partake in the transformation of the fetal heart into a competent and efficient organ. Complete maturation of hPSC-CMs has yet to be reported; however, nature-inspired biological strategies and/or engineering solutions proposed over the past few years have undoubtedly made significant contributions to improve the maturation of these cells.  Fig. 1 Differences between human pluripotent stem cell-derived cardiomyocytes (hPSC-CMs) and ventricular cardiomyocytes from adult cardiac tissue. hPSC-CMs obtained from the differentiation of pluripotent stem cells present fetal-like features in respect to structural and ultrastructural organization, contractile force, metabolism, and electrophysiological function. AP action potential, ECC excitation-contraction coupling, FFR force-frequency relationship, hPSCs human pluripotent stem cells, NCX sodium-calcium exchanger, PRP post-rest potentiation, RMP resting membrane potential, RYR2 ryanodine receptor type 2, SERCA2a sarcoplasmic/endoplasmic reticulum calcium-ATPase 2a, SR/ER sarcoplasmic/endoplasmic reticulum. cardiac development and corresponds to a series of orchestrated phenotypic events that prepare the fetal heart for efficient and competent life-long pumping function. Some of these events are initiated in utero, such as physiologic hypertrophic growth, sarcomeric isoform switch, cell cycle exit, and a glycolytic-tooxidative metabolic switch 79 . However, the highly specialized cellular machinery of adult CMs only fully develops years after birth 46 . Even though prolonging time in culture of hPSC-CMs (up to 1 year) improves structural and functional maturation 47,80,81 , these cells still show signs of a fetal phenotype. Early-stage hPSC-CMs transplanted in animal hearts were able to acquire adult-like features 82,83 , providing evidence of their plasticity and ability to be matured if the right environment is ensured. By attempting to mimic nature, various bioengineering approaches have been proposed to promote hPSC-CM maturation (Fig. 2).

Functional stimuli
In cardiac tissue, CMs are responsive to external electrical stimuli and mechanical forces to perform their function, and these are initiated early during cardiac embryogenesis.
Electrical excitability is a key feature of CMs, dictating electromechanical coupling. The effect of exogenous electrical stimuli on CM maturation was first tested using neonatal rat CMs seeded in collagen sponges and exposed to pulse waves at 1 Hz for up to 8 days, resulting in cell alignment and elongation along the electrical field as well as the formation of intercalated discs with gap junctions expressing connexin-43 84,85 . Both neonatal rat CMs 86,87 and hPSC-CMs 88 have also shown a more mature phenotype after being stimulated with pulse frequencies >1 Hz. Interestingly, applying a supraphysiological pacing regimen up to 1 week (increasing pulse frequency from 1 to 6 Hz) to hPSC-CMs seeded in a collagen matrix markedly contributed to cell alignment, establishment of an organized contractile ultrastructure, improved SR/ER function, and electrophysiological profile 89 .
In turn, mechanical stress induces CM rhythmic contraction, which has a direct impact on physiologic hypertrophic growth, cell elongation, tissue alignment, and force generation. By casting neonatal rat CMs in a ring-shaped mixture of collagen and Matrigel ® exposed to mechanical stretch, it was demonstrated for the first time the importance of applying stress to induce both ultrastructural and functional maturation of CM 90 . Since cylindrical casting molds provided little tissue deflection, mechanical stimulation via cyclic stretch is now achieved using fibrin-based hydrogels in flexible posts 91 or in thin elastomeric frames 92 , allowing for auxotonic muscle contraction. These technologies were first established using rodent cells, but were successfully translated to hPSC-CMs 62,93 . hPSC-CMs show alignment along the hydrogel and sarcomeric organization 62 , a metabolic shift towards fatty acid consumption 94 , improved calcium drug sensitivity 95 , and force development 62 , even though the force generated (up to~1 mN/mm 2 ) is still relatively inferior to those of adult cardiac muscle strips (~30-60 mN/mm 2 ) 58,59 . More recently, modulation of cyclic 96 and passive stretching 97 of hPSC-CMs revealed an impact on tissue functional performance, indicating that there might be ways to further improve tissue force generation. Of note, more complex systems simulating the hemodynamic loads of the cardiac cycle have also been used to improve hPSC-CM maturation 98 .
In addition, CMs can be electrically and mechanically stimulated simultaneously [99][100][101] , thereby enhancing maturation. In a recent work, hPSC-CMs subjected to combined cyclic stretch in elastomeric posts and a supraphysiological pacing regimen of 2 weeks (increasing pulse frequency from 2 to 6 Hz) resulted in these cells displaying adult-like features never reported before, such as a vast network of T-tubules, inotropic response to isoproterenol, a positive FFR, and noticeable PRP 101 . This represents the most advanced strategy reported for hPSC-CM maturation so far, although these more mature hPSC-CMs still presented inadequate contractile force generation ability (up tõ 4 mN/mm 2 ) when compared to adult counterparts.

Structural organization
Cardiac tissue function depends on a highly anisotropic architecture with orderly aligned myofibrils along the direction of muscle contraction, which is impossible to replicate when culturing CMs on standard and simple plastic two-dimensional (2D) surfaces. Promoting this level of cellular organization can be achieved using grooved patterned substrates where hPSC-CMs already showed elongation along the line patterns, electromechanical coupling with neighboring CMs, and electrophysiological improvements 68,102 . Of note, substrate stiffness, groove width, and cell density seem to have an impact on the structural maturation of hPSC-CMs 103 . In fact, the stiffness of a substrate can by itself modulate the mechanical load experienced by hPSC-CMs, affecting the cytoskeleton organization 104 and contractility 102 . In addition, various natural (collagen 20,89 , fibrin 91 , and Matrigel ®90,105 ) or synthetic (polyesters 106,107 , polyethylene glycol 108 , polylactones 109 , and elastomers 110 ) biomaterials can be used to guide cardiac tissue structural organization, even though their properties (mechanical, topographical, and biocompatibility) influence the degree of maturation obtained. Unfortunately, none of these biomaterials can fully recapitulate the architecture and functional composition of the cardiac ECM. Still, artificial ECM 111 or decellularized myocardial ECM [112][113][114] has been reported to promote hPSC-CM maturation without biomimetic regimens. Throughout cardiac development, CMs are in direct contact with other cell types. Even though non-myocytes occupy a relatively small volume in cardiac tissue (~25%), they are necessary for CM survival, contractile performance, cell-cell communication, vascular supply, and ECM deposition 115 . Coculture of hPSC-CMs with endothelial cells, either on Matrigel ® -coated surfaces 116,117 or Matrigel ® hydrogels 118 , resulted in higher expression of proteins related with CM maturity 116 , contributed for ultrastructural organization 116,117 , improved contractility 118 , and calcium handling 116,118 .
Assembling hPSC-CMs into three-dimensional (3D) tissues greatly improves cell-cell communication with more physiological relevance, promoting structural, metabolic, and functional maturation. Engineered tissues can assume various architectures, such as cell sheets 119 , spheroids 120,121 and organoids 122 , muscle strips in the form of rings 90,123 , ribbons 63,124 or cables 89,92 , patches 125,126 , and even ventricle analogs 127 . A recent study using a patch platform reported, without exogenous stimulation, advanced hPSC-CM mechanical maturity with a contractile force closer to adult ventricular tissue (~17-22 mN/mm 2 ) 126 , even though it had slightly negative FFR and an organized ECC apparatus without pronounced T-tubulation. Nonetheless, 3D approaches are extremely versatile and can be combined with other strategies to enhance maturation and tissue performance (e.g., intensive electrical training 101 , dynamic culture 92 , coculture with other cell types [128][129][130] ). For instance, culturing fibroblasts and hPSC-CMs in muscle strips with a ratio of 1:3 improved cell-cell communication and tissue remodeling 128 , in addition to also contributing to hPSC-CM maturation. Intriguingly, the electrical pacing of hPSC-CM muscle strips in flexible posts requires a supporting network of fibroblasts, otherwise the tissue loses its integrity when electrically stimulated 101,131 . Due to the increasing relevance of 3D maturation-based strategies to the field, we will further detail 3D tissue fabrication methodologies in this review (see "Fabrication of biomimetic cardiac tissues").

Biochemical factors
Several soluble and biological factors have been proposed to induce hPSC-CM maturation, such as the type of carbon substrate and hormones. In vivo, CM maturation is characterized by a metabolic shift from glycolysis to preferential fatty acid oxidation, a consequence of cellular hypertrophy and higher energetic demands 56 . hPSC-CMs display metabolic substrate plasticity that can be modulated to promote fatty acid β-oxidation 132,133 . Our group has demonstrated that hPSC-CM structural, functional, and metabolic maturation can be improved in fatty acid-containing medium with galactose to prevent lipotoxicity 134 , when compared with glucose-rich medium. Among hormones, triiodothyronine (T3) is essential during cardiac development 135 , and has revealed to play a key role in inducing hPSC-CM maturation 136 . Furthermore, by combining T3 with dexamethasone, a glucocorticoid hormone, noticeable improvements of SR/ER contribution in ECC and T-tubulation were detected, which otherwise was not seen with each hormone alone 137 .
Biological factors such as those secreted by neighboring cells within the cardiac niche can also positively impact CM maturation state. Recently, hPSC-CM and mesenchymal stem cell (MSC) coculture promoted cellular cross-talk via the release of cytokines and extracellular vesicles capable of modulating hPSC-CM sarcomeric organization, electrophysiology, and metabolism 138 . Another study used a spherical microtissue containing hPSC-CMs as well as cardiac-specific endothelial cells and fibroblasts, and revealed that this tri-cellular interaction upregulated the cAMP/ β-adrenergic pathway 130 . An increase in intracellular cAMP levels consequently affected the assembly of connexin-43 gap junctions, besides improving hPSC-CM structural, contractile, metabolic, and electrophysiologic features. Taken together, by identifying key biological factors capable of inducing CM maturation, these studies underline the importance of heterotypic cell interaction in the cardiac microenvironment.
A less conventional approach to induce CM maturation has been the use of β-adrenergic agonists, namely norepinephrine 139 , isoproterenol 140 , and phenylephrine 141,142 . These compounds are capable of causing structural changes and cellular hypertrophy by inducing intensive workload, even though these effects are likely dose-dependent to avoid pathological remodeling and cytotoxicity.

Genetic regulation
Adult CMs have defined gene regulation mechanisms, ultimately dictating cardiac tissue phenotype and maturation. Overexpressing KCNJ2 during differentiation resulted in hPSC-CMs with hyperpolarized RMP and eliminated the AP proarrhythmic trait during repolarization, even though calcium handling and structural features were not improved due to early loss of automaticity 143 . In turn, calcium kinetics were significantly improved in hPSC-CMs by forced expression of CSQ2 144 . Instead of manipulating the transcriptome, altering the epigenetic state of hPSC-CMs using histone H3 deacetylase inhibitors also showed to promote adult-like expression of calcium handling and conduction genes 145,146 , even though reports regarding electrophysiologic improvements are inconsistent. Priming cells with polyinosinicpolycytidylic acid at the cardiac progenitor stage of differentiation strikingly accelerated structural, electrical, and metabolic maturation of hPSC-CMs, and this acceleration of the developmental clock was associated with increased H3K9ac activation that promoted earlier transcription of cardiac myofilament genes and the Notch ligand JAG1 147 . In another study, the inhibition of aberrantly upregulated factors involved in aerobic glycolysis promoted an hPSC-CM metabolic shift towards oxidative phosphorylation, and consequently features of functional maturation 133 . Moreover, the comprehensive analyses of micro-RNA (miR) levels throughout hPSC-CMs' prolonged culture have identified those that become differentially expressed 148,149 , suggesting alternative routes to modulate gene expression to induce maturation. Let-7 family miRs 150 , miR-1 151,152 , miR-499 152 , and mir-200c 153 , and a combination of miRs (miR-125b, miR-199a, miR-221, and miR-222) 154 have proved useful to improve hPSC-CM maturation features.
Soon after birth, CMs become nonproliferative cells due to extensive hypertrophic growth and remarkable sarcomere reorganization hindering cytokinesis. Ultimately, these changes result in an increase in binucleation and polyploidy. A recent work revealed that knockout of the ECM protein agrin in neonatal mice resulted in the loss of CM proliferation alongside sarcomere assembly and maturation 155 . As expected, administration of agrin to hPSC-CMs avoided loss of proliferation, but negatively affected gene expression and conduction velocity 155 . By promoting cell cycle cessation, mitomycin C treatment similarly increased sarcomerogenesis in neonatal rat-derived PSC-CMs 156 . Further work will be needed to elucidate how cell cycle arrest can promote hPSC-CM maturation other than structural improvements.

FABRICATION OF BIOMIMETIC CARDIAC TISSUES
Inducing the maturation of hPSC-CMs is still a challenge in the field of cardiac TE; however, considerable progress has been achieved by seeking innovative approaches. Moving from simple beating cells to physiologically relevant tissues will require further efforts to engineer the complexity of the in vivo conditions. In the following section, we highlight the most promising techniques adopted to date to grow biomimetic cardiac tissue in the laboratory (Fig. 3). A more detailed overview of these current biofabrication methodologies, their advantages and disadvantages for translation to TE and regenerative medicine fields can be found in ref. 157 .

Hydrogel method
Hydrogels are the most widely used polymers for cardiac TE due to their biocompatible physical and chemical properties, including high water content, efficient gas and mass transfer (i.e., adequate exchange of oxygen and metabolites), ability to be molded into several geometries, easy to be modified/functionalized in order to include cell-binding sites (e.g., RGD integrin-binding domains and ECM proteins), and scalability 158,159 . Cardiac tissues can be Fig. 3 Overview of state-of-the-art biomimetic cardiac tissues and corresponding fabrication approaches. Engineered cardiac tissues range from hydrogels capable of encouraging muscle fiber organization, decellularized scaffolds for tissue/organ bioengineering, microfabricated systems for promoting tissue anisotropy and vascularization, electrospun fibers for close mimicry of tissue architecture, and even printable tissue/miniaturized organs. FRESH Freeform Reversible Embedding of Suspended Hydrogels, SWIFT Sacrificial Writing Into Functional Tissue.
fabricated by joint gelation of cells and a polymer 160 or by having a casting mold to confine the cells in a given polymeric organization 63,90 . Typically, natural hydrogels 89,91,101 are preferred for engineering cardiac tissues by promoting cell assembly and subsequent maturation, but functionalized synthetic polymers 106 or hybrid biomaterials 161,162 are also suitable options. Cells often self-organized within hydrogels in ways that are different from those found in native tissues, and so it can be advantageous to tailor the hydrogel's microarchitecture to encourage anisotropic rather than isotropic tissue growth, which can be done by relying on simple methods such as freeze-drying, solvent casting, gas foaming, and unidirectional freezing 163 . Hydrogels are generally soft, with inferior mechanical compliance in comparison to the native heart tissue. Anchoring supports are useful not only to retain hydrogel conformation but also to provide a more adequate mechanical load to CM 62,101 and facilitate force measurement 87,131,164 . Impressively, the resulting tissues can be several millimeters long (~7-9 mm 62,101 ).

Decellularized bioscaffolds
The main aim of decellularization methodologies is to generate scaffolds with native ECM ultrastructure and composition while removing all cells and genetic material present in the native tissue. Several decellularization protocols have been described for nearly all tissues and organs in the body, and include physical, chemical, and/or enzymatic approaches. Discussing the effects of these agents is beyond the scope of this review (reviewed in refs. 165,166 ); however, it is important to highlight that the choice of optimal protocol for tissue decellularization depends on many factors, including tissue source, cellularity, composition, density, thickness, and donor age, and these overall impact the efficiency of the decellularization process and the resultant ECM composition.
Cardiac tissue or whole heart decellularization is typically accomplished by using ionic and nonionic detergents. Decellularization of a whole rat heart through perfusion 167 and of thin slices of human myocardial tissue 168 using sodium dodecyl sulfate (SDS) resulted in lower DNA content and maintenance of native ECM structure, when compared with approaches using other chemicals. Typically, decellularizing human cardiac tissue 114 requires a higher SDS exposure time than animal-derived cardiac tissue 169,170 , and the reasons for this can be manyfold, such as higher collagen and lipidic content, a stiffer ECM, or donor age. Because of this, it is becoming clear that interspecies differences may hinder the use of decellularization protocols optimized in animals for human tissue. The most immediate application of decellularized myocardial sheets (~200-400 µm thickness 113,114 ), muscle strips (15 mm length, 2.5 mm diameter 114 ), or whole heart 114 is scaffold recellularization. Even though myocardial sheets and strips were successfully recellularized with CMs, recellularization of a human heart is an unmet goal and the best attempt so far has been añ 500 million CM intramyocardial injection to an~5 cm 3 volume of an acellular heart scaffold 114 , which is way below the billions of CMs necessary to repopulate this organ. Nevertheless, isolated ECM can be broken down, solubilized with pepsin, and then assembled into a nanofibrous hydrogel 171 , inevitably destroying the structural and mechanical architecture of the ECM scaffold. Interestingly, decellularized human cardiac ECM requires additional steps of lipid removal before hydrogel formation 172 . Soluble ECM in 3D settings promoted CM maturation 112 , but has a wider application for scaffold functionalization and has shown versatility to be implemented in advanced manufacturing techniques 169,173,174 . Despite its promising applications, decellularized human cardiac ECM is scarce due to donor organ shortage, and therefore research often relies on off-the-shelf animal-derived matrices.

Microfabrication
Advances in microfabrication have allowed detailed engineering of material features in order to resemble tissue in vivo conditions. In particular, photolithography enables the precise transfer of geometrical shapes from a mask onto a substrate to obtain defined topographies for anisotropic tissue organization, while with soft lithography, substrates can be designed to include microchannels and complex microfluidic networks that are useful to recreate tissue vasculature. However, manufacturing such platforms can be time consuming, requires specialized equipment and facilities, and production may be costly.
The simpler systems use microfabrication to pattern material substrates, so as to ease CM spreading and sarcomere axial alignment. For instance, one of the initial attempts used polydimethylsiloxane (PDMS) thin films that contained an array of alternating 20-µm-wide lines, and these grooves substantially improved CM alignment and showed compliance during the cardiac cycle 110 . Currently, more sophisticated patterned culture platforms have been established, which synergize well with biochemical cues. In fact, circular chambers (ranging from 200 to 600 µm diameter), obtained through plasma etching of polyethylene glycol substrates, allowed geometrical confinement of hiPSC colonies and mediated morphogenesis of cardiac microchambers during differentiation 122 . Patterned culture chambers can also be useful to support hydrogel-based cardiac tissue formation (~3-4 mm length) and its long-term maturation (e.g., Biowire TM II 175 ). Nonetheless, microfabrication methodologies can be harnessed to create complex tissues similar to the native myocardium. Pioneering work used laser microablation of poly(glycerol sebacate) membranes to produce an accordion-like honeycomb scaffold, which was designed by overlapping two 200 µm × 200 µm squares at a 45°angle 176 . The unconventional design was purposely conceived to closely resemble the heart anisotropy, and proved successful in guiding CM alignment and improving their mechano-microenvironment. In another study, a porous photomask (2.5 cm × 2.5 cm) was crafted based on cardiac fiber orientation of a specific epicardial plane retrieved from a 3D reconstruction of the human ventricle 125 . This mask was used to mold a PDMS substrate that allowed a patch-like cardiac tissue formation in a developmentally mimetic fashion. More contemporary improvements have advanced even further cardiac patch design by using conductive biomaterials with built-in nanoelectronics 177,178 (~20 mm × 5 mm), which are practical for online electrophysiological recording and electromechanical pacing.
Simple fabrication technologies, such as hydrogels, are unable to integrate organized vasculature within engineered tissue; however, microfabricated cardiac tissues can be assembled with perfusable and highly branched endothelialized channels 179,180 by relying on simple soft lithography approaches. In particular, the AngioChip device (5 mm × 3.1 mm × 150-300 µm), a biodegradable elastomer assembled with layer-by-layer 3D stamping, reported unprecedented functional vascularization of cardiac tissues, allowing perfusion through an open lumen, capillary outgrowth, enhanced permeability, structural integrity, and matrix remodeling 180 .
Despite the lack of standardized culture modes across these systems, remarkable tissue-level complexity can be achieved, as well as adequate tissue function for high-throughput analysis 175 .

Electrospinning
Electrospinning is a versatile method to fabricate ECM-like scaffolds as braided, woven, or knitted fibrous networks, from both naturally derived and synthetic materials. In solution electrospinning, the polymer is gravitationally forced to pass through an electrostatic field to generate a fibrous meshwork. Even though electrospun fibers frequently display inappropriate porosity compromising cell spreading and infiltration, certain parameters, like voltage and polymer flow rate, can be readily adjusted to fabricate the desired fibrous architecture. We have used solution electrospinning to produce poly(ε-caprolactone) (PCL) electrospun scaffolds (~2 cm × 2.5 cm × 229 nm), and the resulting fibers displayed an appropriate architecture that allowed CM alignment 181 . Furthermore, coating the scaffold with piezoelectric microfibers provided suitable electrical stimuli to the cells.
Over the past few years, there have been proposed other electrospinning-based techniques, which aimed at improving certain drawbacks of the traditional method, such as dependency on the polymer's solution conductivity, slow deposition rates, and imprecise fabrication of 3D structures. Producing nano-to submicrometer scale fibers with a fast rate can be achieved using a high-speed rotating reservoir that propels centrifugal polymer deposition in a cylindrical collector 182 . The resulting polyester scaffold (25 mm diameter) was shown to promote CM alignment. A more recent approach used an ellipsoidal mandrel to collect electrospun PCL gelatin-coated nanofibers 127 . The ventricleinspired structure (truncated ellipsoid; 4.5 mm diameter, 9 mm height) promoted CM laminar alignment, even though the ejection fractions and contractile work were~50-250 and~10 8 smaller than the physiologic values for a human ventricle, respectively. Another promising methodology is melt electrospinning (also known as melt electrowriting), which allows a more controlled scaffold fabrication by precise collector or nozzle positioning, similar to 3D printing. Despite being unsuitable for cell encapsulation purposes, a relatively recent study demonstrated that a foldable PCL melt electrospun hexagonal-shaped scaffold (8 mm diameter, 300 µm thick) contributed to CM alignment and maturation 183 . Future improvements to these techniques or integrating existing expertise in multistep biofabrication approaches could allow their widespread use for cardiac TE.

Bioprinting
Bioprinting is an emerging biofabrication method based on precise layer-by-layer deposition of biomaterials, biochemicals, and living cells, collectively named bioinks 184 . High-fidelity tissue and organ surrogates can be generated with the aid of medical imaging modalities, namely computed tomography and magnetic resonance imaging.
Synthetic materials were amongst the first to be printed, but now research focuses on using naturally derived materials, such as ECM hydrogels derived from decellularized tissue. In particular, cardiac-specific ECM-derived bioinks, when pressure extruded through a printing nozzle, have successfully generated grid-type structures in the form of rectangles 169 (~5 mm × 5 mm × 1 mm) or circles 173 (10 mm diameter, 0.6 mm thick), such that it resembles a patch. Extrusion bioprinting is one of the most popular methods to directly print cell-laden hydrogel-based cardiac tissue in an organized and reproducible fashion, besides allowing printing with high cell densities 184 , which is a necessary requirement to print more complex structures. Aside from the above-mentioned simple tissue constructs, extrusion bioprinting can be used to create more complex tissue models, but freeform printing of such structures requires a supporting sacrificial material. A novel method termed "FRESH" (i.e., Freeform Reversible Embedding of Suspended Hydrogels) achieved extrusion-based freeform printing by embedding the bioink in a thermally reversible supportive hydrogel during printing 185,186 , and this allowed fabrication of high-resolution 3D anisotropic structures, including a beating ventricle (truncated ellipsoid; 5.7 mm diameter, 8 mm height) and an acellular neonatal-sized heart analog (37 mm diameter, 55 mm height) 186 . Another study used a similar approach, but with an enzymatic or chemical reversible supportive hydrogel, and was able to print a heart-shaped model (14 mm diameter, 20 mm height) using an omental ECM biomaterial in combination with CMs and endothelial cells 174 . Further improvements are needed to create a functional tissue since the cells were reported to be arranged as nonaligned aggregates after 1 day in culture.
Another advantage of bioprinting is the possibility of creating tissues with inherent perfusable vasculature [187][188][189] . Direct extrusion-based bioprinting was used to manufacture one of the first vascularized cardiac tissues, in which the resulting endothelialized grid-type structure (~5.5 mm × 3.5 mm × 0.75 mm) was seeded with CMs and perfused with a microfluidic device 187 . Bioprinting a vascular network greatly improves the thickness these tissues can achieve without becoming necrotic (>1 cm 188 ), but they often lack physiologically relevant cell numbers (~1 × 10 8 cell/mL). A new innovative method termed "SWIFT" (i.e., Sacrificial Writing Into Functional Tissue) relies on assembling up to 400,000 spheroids/organoids as a tissue matrix and then extrudes a sacrificial ink throughout the compact granular tissue, which can later be evacuated and the resulting lumen perfused 190 , despite lacking robustly endothelialized channels. This clever technique allowed the assembly of a cardiac tissue with 240,000 million cells (6 mm top width, 4.2 mm bottom width, 12 mm height, 4.2 mm depth) that fuse and beat synchronously over 7 days in culture. Stereolithography, a higher resolution bioprinting methodology that uses a light projector to solidify layer-by-layer the bioink, has recently been used to create entangled vascular networks to perfuse an alveolar sac model 189 . Due to its exceptional level of functionality, translating this approach to cardiac TE could also be worthwhile.
It is important to take into account that additive printing with a high level of detail requires a slow printing velocity, affecting the scalability of current methods. To overcome this limitation, a recent study demonstrated that tissue can be printed within seconds with high fidelity by irradiating multiple light patterns representing projections of the object to be printed onto a photopolymer 191 . Despite being a promising technology, its potential use for cardiac TE still needs to be evaluated.

UNLOCKING CARDIAC TISSUE ENGINEERING FOR PRECISION MEDICINE
The advances in hPSC biology together with the development of novel materials and biofabrication methods have enabled the production of patient-specific cardiac tissues, which in turn opens up the possibility of fulfilling the promise of personalized medicine (Fig. 4). This section includes a brief overview of the current state-of-the-art applications of cardiac TE for drug discovery, disease modeling, and regenerative therapies.
Cardiac tissue engineering for drug discovery and disease modeling It is well known that cardiovascular safety liabilities remain a major cause of drug attrition, resulting in early termination of pipeline candidates or withdrawal of marketed drugs 192 . Commonly used platforms to assess safety include small animal models, primarily mice, and ion-channel overexpressing cell lines 193,194 , which fail to truly predict human drug response effects (proarrhythmic, cardiotonic, and cardiotoxic) due to differences in cardiovascular physiopathology. Thus, cardiac TE-based solutions, such as spheroids 120,195 , microtissues 63,124 , and "heart-on-a-chip"/microfluidic devices 175,196,197 , have been emerging as advanced culture systems, capable of improving the physiologic relevance over traditional 2D in vitro culture platforms. In these systems, hPSC-CMs have shown higher predictive insight regarding drug mechanistic (e.g., inotropic response 101,175,198 and ECC blockage 101,175 ) and toxic (e.g., loss of contractile function 198 and apoptosis 195 ) effects. Nonetheless, additional efforts are required to improve the reliability of these systems. For instance, a recent study demonstrated that linsitinib, a promising drug candidate to treat Ewing sarcoma, significantly reduced tumor viability and cardiac function when both of these engineered tissues were cultured separately under perfusion in an organ-on-a-chip system. However, when they were cultured in an integrated way in the same platform, linsitinib showed poor tumor response and less cardiotoxicity, which was in agreement with the unfortunate clinical trial outcome of this drug 199 . Testing the targeted or offtargeted effect of new cardiovascular drugs will require advancing the physiological relevance in which cardiac engineered tissues are cultured. In addition, the complexity of engineered cardiac tissues is still incompatible with a rapid screen of large libraries of compounds, despite their smaller size in relation to a whole tissue/ organ. Although some platforms have shown the ability to be scaled-down 124,200,201 , high-throughput culture modes and reproducible assessment of tissue functionality are still lacking 202 .
Expediting drug development will further require reliable models for a better understanding of disease mechanisms that guide target drug validation/discovery. In this context, hPSCs, as stated above, have emerged as attractive tools and a preferred source for obtaining cells for in vitro models, since these cells can be derived from patients using minimally invasive procedures, and preserve the genetic and phenotypic characteristics of the donors, including pathological traits responsible for disease onset and progression 203,204 . Several studies have been using 2D hiPSC-CM systems to investigate a plethora of genetic mutations associated with genetic heart diseases, including the main inherited cardiomyopathies (i.e., familial hypertrophic cardiomyopathy 205 , dilated cardiomyopathy 203,206 , arrhythmogenic cardiomyopathy 207 , and left ventricular noncompaction cardiomyopathy 208 ) and channelopathies (e.g., type-2 long-QT syndrome 209 and Timothy syndrome 210 ). Others are modeling diseases without a known disease-causing genetic variant, such as idiosyncratic druginduced cardiotoxicity 204,211,212 . In parallel with the use of patientderived hPSCs, recent advances in targeted gene editing using the CRISPR-Cas9 systems have been performed on hPSCs [213][214][215] . Employing this technology, any wild-type cell can be altered to harbor a specific mutation, and a disease-causing mutation can be corrected in patient-derived cells, generating an isogenic control cell line. To the best of our knowledge, there are 13 completed or active clinical studies (www.clinicaltrials.gov) with the goal of generating patient-specific CMs to study the molecular mechanisms involved in several cardiac diseases, such as genetic and hypertrophic cardiomyopathies, inherited arrhythmias and valvulopathies, and cardiac fibrosis, among others (Table 1).
Notwithstanding, the use of engineered cardiac tissues as disease models is a relatively new concept. hPSC-CM tissue platforms harboring genetic mutations characteristic of hypertrophic cardiomyopathy 216 , dilated cardiomyopathy 217 , and mitochondrial cardiomyopathy of Barth syndrome 218 have reported abnormalities in cardiac function congruent with a pathological phenotype. More recently, chamber-specific arrhythmia was demonstrated in a ringshaped cardiac tissue containing both ventricular and atrial hPSC-CMs 219 . Manipulating engineered tissue microenvironment to induce pathological changes, such as CM remodeling of myocardium failure 220 , excessive ECM deposition of cardiac fibrosis 221 , and thrombus formation 222 , could similarly provide invaluable insights into mechanistic studies or predict patient-specific disease progression. Even though promising, disease-carrying hPSC-CMs still present an immature phenotype, thus possibly confounding identified disease hallmarks and identified mechanisms. For   Fig. 4 Streamlining engineered cardiac tissues for precision medicine applications. Tissue engineering expertise enables the generation of physiologically relevant cardiac tissues grown from human pluripotent stem cell-derived cardiomyocytes (hPSC-CMs). Designing a proper maturation strategy will allow the maturation of hPSC-derived cardiac tissues to an adult-like phenotype, which can later on be used for patient-specific drug screening, disease modeling, and regenerative therapies.
instance, genetically engineered hPSC-CM microtissues carrying mutations that truncate the sarcomere protein titin, a known common cause of dilated cardiomyopathy, exhibited reduced contractility and decreased sarcomere length, among other abnormalities 217 , none of which were detected in adult samples 223 . As a matter of fact, cardiac tissue structural maturity imposed by intensive electrical training was necessary to model pathological ventricular hypertrophy 101,175 . Even with their current limitations, the identified pathomechanisms in diseased engineered cardiac tissues can, in some cases, be translated to patient care in a personalized way. Recent research found out diltiazem, an L-type calcium channel blocker, was effective to correct disease-specific cardiac abnormalities in patient-derived engineered muscle strips with a rare form of hypertrophic cardiomyopathy. Diltiazem was further translated into patient care and ameliorated the prolonged QTc interval after 1 month of treatment 224 . These encouraging results give hope that emerging TE platforms could be extremely useful to reverse the growing burden associated with cardiovascular disorders.
Cardiac tissue engineering for heart regenerative therapy Regenerative cardiac cell therapy broadly aims to achieve two complementary goals: (i) the direct cell replacement of the injured myocardium with contractile CMs and (ii) paracrine modulation of endogenous repair processes, such as angiogenesis, inflammation, apoptosis, and fibrosis, with both contractile and noncontractile cells 14,225,226 . To the best of our knowledge, there are over 40 completed clinical trials exploring the therapeutic potential of cellbased therapies for cardiac regeneration and repair (www. clinicaltrials.gov). Pioneer studies tested non-cardiomyogenic cell populations that can easily be prepared for clinical applications, namely bone marrow-derived mononuclear cells (BM-MNCs) or MSCs, which in spite of their limited regenerative potential, may stimulate endogenous regenerative responses 225,227 . Secondgeneration cell candidates require a more refined isolation process, and ex vivo amplification procedures, but have higher regenerative potential, such as cardiac-derived progenitor cells and hPSC-CMs 225,227 . However, the clinical outcome of most of these studies remains insufficient, with evidence suggesting the major common obstacles to be the low engraftment and survival rate of transplanted cells in the host tissue, and the scarcity of endogenous cells with repair capacity 225,[228][229][230] .
Cardiac TE has proven to be an effective method to improve the delivery, engraftment, and differentiation of stem cells in injured hearts, and therefore has gained much attention in recent years. Cardiac tissue surrogates developed to date are typically engineered cardiac muscle strips [231][232][233] , epicardial patches 126,234-236 , or cardiogenic scaffolds 237 , even though other strategies can be employed [238][239][240][241][242][243] , such as acellular hydrogels to improve infarct mechanical integrity [241][242][243] . A summary of relevant cardiac TE preclinical studies in animal models is highlighted in Table 2. When transplanted into infarcted or defected animal hearts, hPSC-CMbased tissues led to longer-term cell retentions 231 , progressively matured in vivo 231 , (imperfectly) electrically coupled with the host tissue 233 , reversed infarction-associated changes (e.g., ameliorated cardiac remodeling by decreasing the amount of fibrosis 241-243 , enhanced vascularization 126,233,242 ), and overall improved left ventricular function 231,240,243 . Recent advances have allowed rapid fabrication of clinically relevant sized human cardiac tissues able to maintain functionality preimplantation 126,234 , thus being suitable for large animal preclinical studies and future clinical applications.  Among relevant risks that have to be taken into account in these therapies are the development of arrhythmias 244,245 , uncontrolled proliferation, and unwanted differentiation of cells that may lead to tumor formation in the case of hPSC-based therapies [245][246][247][248] , or to calcification of the host myocardium by MSCs or BM-MNCs 247,249 . Another potential issue is the development of an adverse immune response against the cells and materials employed in the constructs 250 . Designing tissues with a more controlled cell composition, defined architectural structure and developing strategies favoring muscular maturation before implantation could further improve their regenerative capability.
Despite showing promising outcomes in preclinical settings, the potential use of cardiac TE-based therapies at a clinical level is still in its infancy. Preliminary data from first-in-man clinical trials are providing evidence that cardiac TE-based therapies are generally safe 251,252 , but do not provide relevant functional benefits. We are aware of five completed and two active phase 1/2 studies mainly aiming at evaluating the safety and feasibility of cardiac TE therapeutic interventions, but also collecting preliminary efficacy data ( Table 3).
The first studies being conducted are focused on the hypothesis that biomaterials could act as an internal wall support to increase infarct wall thickness, thereby improving cardiac function. Results from PRESERVATION I and AUGMENT-HF trials were, however, disappointing, with minor improvements in symptoms and clinical status for patients with advanced heart failure (HF), but no significant changes in left ventricle (LV) remodeling. Aiming to provide not only mechanical support for the prevention of infarct expansion but also to restore ECM homeostasis and architecture, other strategies relied on the epicardial or myocardial application of ECM patches or hydrogels. In particular, CorMatrix ® -ECM patches, derived from porcine small intestinal submucosa (SIS-ECM), were applied on the epicardium of patients undergoing coronary artery bypass grafting, after an acute ischemic event. Porcine SIS-ECM has been used for over 20 years in a range of cardiovascular applications, but of all the commercially available SIS-ECM products, CorMatrix ® is the only medical device with clearance, even though there are few reports regarding its use for myocardial regeneration after acute myocardial infarction in humans 253 . No results for this study have been published so far; however, based on preliminary data obtained in rodent and porcine models, the patches are expected to not only prevent infarct expansion and LV dilation but also restore ECM homeostasis and architecture 252,254 . Another recently completed study evaluated the effect of VentriGel, a myocardium-specific ECM hydrogel, on patients in a 60-to 3-year window since their first acute ST-elevation myocardial infarction. Using a minimally invasive procedure, the gel was injected using a NOGA-MyoStar catheter, allowing for an intramyocardial delivery of the therapeutic gel on mapped infarcted areas. Overall, VentriGel showed encouraging safety results, with no deaths or patients who discontinued from the study, nor adverse events reported as a direct consequence of the treatment 255 . Although efficacy was not a primary outcome, there were suggestions of improvements, including in the 6-min walk test distance and decreases in New York Heart Association functional classification, across the entire cohort of patients 255 . Improvements in LV remodeling data, viable mass, and ventricular natriuretic peptide levels were mainly observed in patients who had a myocardial infarction >12 months before treatment 255 .
A distinctive approach has been the delivery of clinical-grade hESC-derived cardiovascular progenitor cells in fibrin patches for the treatment of severe ischemic left ventricular dysfunction 251 . Compared with biomaterial-only methods, cell patches showed equal safety profiles, with no complications that could be specifically ascribed to the cellular graft, namely, no detection of teratomas that could be developed from residual pluripotent cells, no occurrence of arrhythmias, and no major alloimmunization events. Still, following a similar trend as previous strategies, no astonishing results were obtained in terms of efficacy, although all patients were symptomatically improved. A newly posted study aims at determining the safety of Wharton's jelly-derived MSCs from human umbilical cord seeded on ECM patches, and also test its efficacy to promote neomyogenesis in patients with previous myocardial infarction. Another active and recently recruiting promising clinical trial will evaluate the efficacy of an engineered heart muscle (EHM) to sustainably remuscularize and improve myocardial performance in patients with advanced HF with reduced ejection fraction (≤35%). The EHM, a construct generated from hiPSC-CMs and stromal cells in a bovine collagen type I hydrogel, will be the first of its kind to be tested in humans. Despite the marginally positive outcome of all these clinical trials, this does not necessarily imply a lack of therapeutic value and may result instead from an inadequate choice of biomaterial and/or cell type or inaccuracies in dosing and/or delivery. A better understanding underlying the mechanism of action of these advanced medicinal products may contribute to further improvements down the road.

CONCLUSIONS
To date, inducing maturation of hPSC-CMs is a challenge in the field, limiting their use in clinical settings. Considerable progress has been achieved by continuously seeking innovative ways to improve hPSC-CM maturation state, and, currently, it is already possible to obtain cardiac tissue with adult-like gene signatures, remarkable ultrastructural organization, functional contractile machinery, and improved electrophysiology. Even though in vitro maturation may not follow the in vivo paradigm, identifying the spatio-temporal mechanisms dictating tissue development and designing a combinatorial strategy capable of addressing such complexity might be required to engineer fully matured biomimetic cardiac tissues. Given the interdisciplinarity of TE, from stem cell biology to stem cell bioprocessing, material science, and advanced biofabrication, taking advantage of this knowledge would be beneficial to grow physiologically relevant cardiac tissues in vitro. Recreating cardiac tissue complexity is still challenging, limiting the size of the engineered tissue possible to be obtained, and thus redirecting their application towards other areas rather than the clinical transplantation arena. Benchmarked "on-a-chip" engineered cardiac tissues, although promising, still need to show standardization and high-throughput compatibility to contribute to more reliable results than commonly used platforms, as well as reduction of costs to expedite drug development. Successful attempts at scaling up engineered cardiac tissues have made it possible to investigate their regenerative efficacy in large animal studies, and the results of pioneering TE-based clinical trials are eagerly awaited. While there are still few clinical trials, these have been an important step in the ability to engineer artificial constructs that mimic some of the important compositional, architectural, and functional properties of human cardiac tissues. Even if it is not unrealistic to foresee a translation of engineered cardiac tissues for in vitro clinical studies, their potential to remuscularize failing hearts and integrate the host tissue without adverse effects, despite a better retention, is still uncertain, and the causes associated with mild improvements unknown. Nonetheless, advances in the field of cardiac TE have undoubtedly imposed a new paradigm in cardiovascular science, paving the way one day for restoring or even replacing damaged hearts.