In vitro models of the cardiac microenvironment to study myocyte and non‐myocyte crosstalk: bioinspired approaches beyond the polystyrene dish

Abstract

The heart is a complex pluricellular organ composed of cardiomyocytes and non‐myocytes including fibroblasts, endothelial cells and immune cells. Myocytes are responsible for electrical conduction and contractile force generation, while the other cell types are responsible for matrix deposition, vascularization, and injury response. Myocytes and non‐myocytes are known to communicate and exert mutual regulatory effects. In concert, they determine the structural, electrical and mechanical characteristics in the healthy and remodelled myocardium. Dynamic crosstalk between myocytes and non‐myocytes plays a crucial role in stress/injury‐induced hypertrophy and fibrosis development that can ultimately lead to heart failure and arrhythmias. Investigations of heterocellular communication in the myocardium are hampered by the intricate interspersion of the different cell types and the complexity of the tissue architecture. In vitro models have facilitated investigations of cardiac cells in a direct and controllable manner and have provided important functional and mechanistic insights. However, these cultures often lack regulatory input from the other cell types as well as additional topographical, electrical, mechanical and biochemical cues from the cardiac microenvironment that all contribute to modulating cell differentiation, maturation, alignment, function and survival. Advancements in the development of more complex pluricellular physiological platforms that incorporate diverse cues from the myocardial microenvironment are expected to lead to more physiologically relevant cardiac tissue‐like in vitro models for mechanistic biological research, disease modelling, therapeutic target identification, drug testing and regeneration.

Abbreviations

αSMA

α smooth muscle actin

CM

cardiomyocyte

Cx

connexin

EC

endothelial cell

ECM

extracellular matrix

EHT

engineered heart tissue

Fb

fibroblast

hESC

human embryonic stem cell

hiPSC

human induced pluripotent stem cell

HUVEC

human umbilical vein endothelial cell

MEF

mouse embryonic fibroblast

MyoFb

myofibroblast

non‐CM

non‐cardiomyocyte

SR

sarcoplasmic reticulum

Introduction

The heart is a pluricellular organ composed of cardiomyocytes (CMs) and non‐myocytes (non‐CMs) such as fibroblasts (Fbs), endothelial cells (ECs), smooth muscle cells, neurons and immune cells. Each cell type has distinct features and functions. While CMs are responsible for electrical conduction and contractile force generation, other cells are involved in matrix deposition, vascularization, inflammation and autonomic regulation. Non‐CMs process and transmit biochemical, mechanical and electrical cues, and are therefore essential components of the myocardial microenvironment.

Crosstalk between CMs and non‐CMs is critical for cardiac development and the regulation of postnatal morphology and function in health and disease. Both Fbs and ECs communicate with CMs and exert regulatory effects on their size and electro‐mechanical function (reviewed by Zhang et al. 2012 and Brutsaert, 2003, respectively). Mouse models with cell type‐selective gene targeting have provided in vivo evidence for crosstalk between CMs and Fbs (Nishida et al. 2008; Takeda et al. 2010; Yoon et al. 2010) and CMs and ECs (Aird et al. 1997). Importantly, CM/non‐CM crosstalk is bidirectional, and miscommunication has been implicated in the pathophysiology of heart failure and arrhythmias (Nguyen et al. 2014b; Lim et al. 2015). In this review, we focus on Fbs and ECs as the major non‐CMs. Much less is known about communication of CMs with resident immune cells and those recruited under pathological conditions (Kamo et al. 2015), and postganglionic efferent neurons for the modulation of contractile force and frequency (Franzoso et al. 2016).

In response to stress or injury, the myocardium undergoes structural remodelling, which entails CM growth or death, vascular rarefaction, fibrosis and inflammation, with specific changes depending on the insult (Burchfield et al. 2013). While initially adaptive, cardiac remodelling can ultimately lead to contractile dysfunction and irregular rhythm. Alterations in cellular composition and distribution, myocardial stiffness, mechanical contraction and beating rate represent important changes in the cardiac microenvironment of the diseased heart.

Investigation of the mechanisms and consequences of cellular crosstalk requires suitable experimental models. Suitability depends on the questions asked, the outcomes desired, and the tools available. In vivo models have the distinct advantage that cells are in their native microenvironment, but they are complex, difficult to control, low‐throughput and expensive. The ability to target genes in a cell type‐specific manner in the heart in vivo is hampered by cellular heterogeneity. In vitro studies offer more efficient and precise control of variables and analyses, but cells are uncoupled from their native environment. Researchers have begun to recombine key features of the in vivo tissue environment in microphysiological in vitro platforms that range from single‐cell functional assays to engineered tissues (Kurokawa & George, 2016; Tzatzalos et al. 2016).

This review focuses on cardiac in vitro platforms for the investigation of ventricular CM/non‐CM crosstalk. First, we highlight the complex, pluricellular composition of the myocardium, and discuss important considerations regarding the source and phenotype of CMs, Fbs and ECs for co‐culture studies. We then turn our attention to the diverse cues that cardiac cells are exposed to in the tissue context and summarize progress and challenges in incorporating them into in vitro platforms. In closing, we offer our thoughts on the remaining challenges, key questions, and new opportunities.

Complexity of the pluricellular myocardial microenvironment

Although CMs occupy 70–80% of the volume of the adult mammalian heart, they are much smaller in number than the non‐CM population. Fbs were long viewed as the largest non‐myocyte cell population in the heart (e.g. Banerjee et al. 2007). Recent analysis of human hearts maintain this dominance (Bergmann et al. 2015), but ECs were recently shown to outnumber Fbs in the mouse heart by a factor of 4 (Pinto et al. 2016). Study‐specific differences in relative cell numbers are in part attributable to methodologies (stereology in tissue sections, flow cytometry of dissociated cells) and molecular reagents used for cell type identification (antibodies for cell markers, genetic labels) (reviewed by Zhou & Pu, 2016). The relative proportion of cardiac cell types changes during development and can differ between species (Banerjee et al. 2007), while being overall comparable in different ventricular regions (Banerjee et al. 2007; Pinto et al. 2016).

The proximity between different cell types facilitates heterocellular communication. CMs and Fbs are spatially intermingled, with virtually every CM bordering one or more Fbs in the healthy myocardium (Kohl et al. 2005). The endothelium communicates with subjacent CMs at the endocardium and in the capillaries (Brutsaert, 2003). The network of capillaries in the myocardium is intricate, with at least one capillary next to every CM. In the diseased heart, the relative proportion of cardiac cells changes (e.g. CMs and ECs die, Fbs proliferate, and immune cells invade after an ischaemic insult), and the distance between them can be extended by fibrosis, which also increases myocardial stiffness. Excess extracellular matrix (ECM) deposition can be interstitial and perivascular (pressure overload) or compact (infarct). Cell ratios and their spatial distribution are important considerations for the design of cardiac in vitro models, as is the stiffness.

Heterocellular communication is mediated through the exchange of biochemical, electrical and mechanical messages via the release of paracrine factors, direct cell–cell interactions and/or ECM‐mediated cellular communication (Zhang et al. 2012). Paracrine factors are released from both CMs and non‐CMs. In addition to many well‐characterized paracrine factors (growth factors, cytokines, chemokines), CM‐Fb communication can also be mediated by microRNAs (miRNAs; Bang et al. 2014) and metabolites (Sassi et al. 2014). EC‐derived paracrine factors that regulate CMs include neuregulin, nitric oxide and endothelin‐1 (Lim et al. 2015). Conditioned medium and transwell co‐cultures have long been used to investigate paracrine communication in the absence of direct cell–cell contact. Microfluidic platforms for the investigation of paracrine, juxtacrine and endocrine cell–cell communication have been developed (reviewed by Nahavandi et al. 2014), but to our knowledge they have not yet been applied to cardiac cells. Direct cell–cell contact between CMs and Fbs can be mediated via gap junctions (reviewed by Ongstad & Kohl, 2016), adherens junctions (Thompson et al. 2011) and nanotubes (He et al. 2011; Quinn et al. 2016). Cell contact‐mediated effects have also been suggested for EC and CM interactions (Narmoneva et al. 2004).

The intricate relationship among cellular and acellular tissue components in the ECM meshwork is another important and dynamic feature of the myocardial microenvironment. ECM‐mediated communication is involved in the regulation of cellular growth, survival, proliferation, differentiation and migration (reviewed by Valiente‐Alandi et al. 2016). The highly organized interstitium is critical for muscle alignment and force transmission and provides micro‐ and nanoscale topographical and molecular cues (Eckhouse & Spinale, 2012). Choices of ECM proteins for coatings and matrices should be carefully considered for crosstalk platforms, and the reader is referred to an excellent review on how ECM structure, chemistry and mechanics can inform cardiac tissue engineering (Capulli et al. 2016).

Cellular considerations for cardiac in vitro models

CM and non‐CM phenotypes vary depending on their origin, developmental stage and activation state, which must be considered in cardiac in vitro model design.

The differentiation state is a major determinant of CM properties and function (reviewed in Peter et al. 2016). Native adult ventricular CMs are terminally differentiated, rod‐shaped cells that are aligned with the direction of contraction and have highly organized sarcomeres. Isolated adult CMs from most species can maintain their rod‐shaped morphology and phenotype but only in short‐term culture without serum. They resemble CMs in the intact heart tissue but usually require lower density culture conditions and can therefore not be easily incorporated into cardiac tissue models. Neonatal CMs and those derived from human embryonic stems cells (hESC‐CMs) or induced pluripotent stem cells (hiPSC‐CMs) are suitable for tissue engineering but much less differentiated than adult CMs, which remains a major caveat. The phenotype and function of hiPSC‐CMs has been reviewed by Karakikes et al. (2015). The developmental stage of CMs affects electromechanical function, as shown by reduced conduction velocity and contractile function in a direct comparison of engineered tissue generated from ESC‐derived vs. neonatal CMs (both from mice) (Feinberg et al. 2013). CM maturation is subject to regulation by diverse cues, including topographical, electrical, mechanical, biochemical and cellular interaction cues. Some of them can be leveraged to enhance directed differentiation and advance maturation of stem‐cell derived and neonatal CMs (see below). Importantly, in diseased hearts, CMs change in size (with pathological trigger‐dependent increases in their length/width ratio) and in cellular and subcellular structural organization. CMs from surgically induced or genetically modified animal models of cardiac diseases (Houser et al. 2012) as well as hiPSC‐CMs from patients with hereditary sarcomeric cardiomyopathies (Eschenhagen et al. 2015) and arrhythmic disorders (Sallam et al. 2015) are recognized for their potential for disease modelling. Although many caveats remain and it will only be possible to recapitulate a subset of the in vivo phenotype in vitro, microtissues engineered from patient‐derived hiPSC‐CMs were recently shown to be a powerful platform for evaluating the pathogenicity of titin gene variants (Hinson et al. 2015).

Species‐specific and regional differences in the electrical and Ca²⁺‐handling properties of CMs are additional key considerations. A prominent feature of ventricular action potentials from mice and rats is the triangular shape and lack of a distinct plateau phase compared to larger species, which is primarily due to differences in the expression of repolarizing K⁺ channels, but species‐specific variations in Na⁺ and Ca²⁺ channels also exist (reviewed by Bartos et al. 2015). Action potential waveforms also differ between CMs from different cardiac regions (i.e. atria vs. ventricles, left vs. right ventricle) and even within the ventricular wall (endo/mid/epicardium), owing to heterogeneity in the expression and/or properties of the underlying ion channels (reviewed by Nerbonne & Kass, 2005; Antzelevitch & Dumaine, 2011). Cardiac excitation–contraction coupling shows marked species‐specific and regional differences as well. Species markedly differ with regard to the source of Ca²⁺ in activating contraction (i.e. extracellular via voltage‐activated Ca²⁺ channel influx vs. release from the sarcoplasmic reticulum (SR) via ryanodine receptors), and the fraction of Ca²⁺ removal via plasmalemmal extrusion vs. SR Ca²⁺ reuptake (reviewed by Bers, 2002). Small rodents rely more heavily on SR Ca²⁺ release and reuptake in ventricular CMs than larger animals and humans. Shorter action potentials and larger SR dependence in adult mice and rats support rapid contraction at higher heart rates. The transverse (T)‐tubule network and junctions with the SR, which enable synchronized Ca²⁺ release during membrane excitation and coordinated contraction within ventricular CMs, are absent in atrial CMs (reviewed by Bootman et al. 2006). In ventricular CMs, T‐tubules vary in density between species (e.g. higher in rodents) and evolve during maturation (reviewed by Guo et al. 2013), which is associated with maturation of Ca²⁺ release units at the SR and developmental changes in SR Ca²⁺ fluxes (reviewed by Louch et al. 2015). The immaturity of the T‐tubular network of neonatal and stem cell‐derived CMs is one of the driving forces behind on‐going efforts to mimic their native microenvironment in tissue engineering to enhance their maturation. CM maturation is further gauged by additional structural parameters (sarcomere assembly, myofibre orientation, mitochondrial distribution) as well as functional (electrophysiology, calcium handling, force of contraction) parameters, molecular (shift from fetal to adult gene expression), and metabolic (shift from glycolysis to fatty acid oxidation) parameters (Liaw & Zimmermann, 2016).

The source, maturation and activation states are also important considerations for non‐CMs. Ideally, Fbs and ECs of cardiac origin should be used for crosstalk studies because of anatomical site‐specific differences in gene expression (Chang et al. 2002; Chi et al. 2003). EC structure and function also varies in time and space, in part because of adaptation to many different microenvironments to serve diverse site‐specific functions (reviewed in Aird, 2007). ECs and Fbs from all developmental stages can be cultured in vitro but maturation can affect function. For example, embryonic Fbs promote hyperplasia in embryonic CMs, whereas adult Fbs promote hypertrophy (Ieda et al. 2009). Activated Fbs change their morphology and functional properties when they convert into contractile and hypersecretory myofibroblasts (MyoFbs) in response to stress or injury (Weber et al. 2013). Fb activation can be experimentally induced in culture, but is critically dependent on culture substrate stiffness (see below). Alternatively, activated Fbs can be obtained from fibrotic hearts for disease modelling. Like ECs (Duong et al. 2011), Fbs can maintain their phenotype in culture in early passages (Squires et al. 2005), but phenotypic drift should be carefully monitored.

Taken together, species, regional origin, developmental stage, maturation and activation states are important determinants for the phenotypic properties of CMs and non‐CMs. They must be taken into account for the design and interpretation of cardiac in vitro models. For the reasons described above, stem cell‐derived or neonatal CMs are most commonly used for cardiac tissue engineering. Depending on the differentiation, maturation and culture conditions, stem cell‐derived CMs show heterogeneity in their phenotypes, such as ventricular‐, atrial‐ and nodal‐like electrophysiological properties (reviewed by Blazeski et al. 2012). CMs, Fbs and ECs that are isolated from neonatal rodents often originate from both ventricles, although whole hearts are also used for cell isolations. Non‐cardiac sources for Fbs and ECs that have been incorporated into cardiac in vitro models include Fb cell lines (3T3), mouse embryonic Fbs (MEFs), human umbilical vein ECs (HUVECs) and stem cell‐derived ECs.

Rebuilding native tissue‐level (3‐D) structure

Compared to conventional 2‐D monolayers, 3‐D cultures enhance cell–cell and cell–ECM contact, establish molecular concentration gradients, and can eliminate the need for unnaturally stiff and adherent substrates (Baker & Chen, 2012). Developed 3‐D platforms include cells on thin films, cells in gels/scaffolds, cells in micro‐fluidic devices, and cells aggregated into microtissues (reviewed by Mathur et al. 2016; Zuppinger, 2016). Much of the research in this area is geared toward cardiac patch development. Scaffold‐based approaches provide synthetic and/or natural matrices as a structural template for cells to attach to, whereas scaffold‐free approaches promote and depend on cellular self‐assembly and organization. In scaffold‐based models, cell‐to‐scaffold interactions predominate. The addition of neonatal rat ventricular Fbs to CMs in scaffolds enhances CM elongation, alignment and viability (Nichol et al. 2008), and pre‐treatment of scaffolds (conditioning) with these Fbs can improve electrical excitability and contraction (Radisic et al. 2008). Co‐culture of hESC‐derived CMs with hESC‐derived ECs (or HUVECs) was shown to enhance CM proliferation without affecting elongation, orientation or alignment (Caspi et al. 2007). It can also inhibit CM apoptosis and necrosis, which was enhanced when neonatal mouse CMs were seeded on preformed EC networks compared to co‐culture (Narmoneva et al. 2004). In pluricellular models with all three cell types, crosstalk between embryonic Fbs and ECs was indicated by decreased EC death and increased proliferation in the presence of Fbs (Caspi et al. 2007). A synergistic effect on endothelial structure formation was reported when mouse embryonic Fbs or human marrow stromal cells and HUVECs were co‐cultured with hESC‐CMs (Tulloch et al. 2011).

While there is variation amongst engineered heart tissues (EHTs, Eschenhagen et al. 1997) and other engineered cardiac (micro)tissues and wires (Boudou et al. 2011; Thavandiran et al. 2013; Turnbull et al. 2014), generally these platforms are 3‐D hydrogel‐based muscle constructs, reconstituted in non‐adhesive moulds with attachment or restraint points. The resulting spontaneously formed engineered tissues show contractile and electrophysiological properties of the working myocardium (Zimmermann et al. 2002) and have higher levels of binucleation, improved sarcomere assembly, and a more physiological response to hypertrophy than neonatal cells in standard culture conditions (Tiburcy et al. 2011). These tissues can be downscaled from centimetres to millimetres (Hansen et al. 2010; Boudou et al. 2011), but the adoption of flexible posts to produce smaller scale tissues and reduce tissue handling may not maximize force generation (Zimmermann et al. 2006; Turnbull et al. 2014). Forces generated by these tissues are an order of magnitude less than forces generated in vivo (Eder et al. 2016). Extracellular material properties have been shown to affect these organized 3‐D cultures as more force is generated from tissues on more rigid cantilevers, and stiffer matrix also results in higher cell tension (Boudou et al. 2011). Bracing of EHTs with non‐compliant materials can induce afterload and model pathological hypertrophy in vitro (Hirt et al. 2012).

Spinner flasks, rotation systems, hanging drops, microfluidic systems and non‐adhesive substrates have all been used to self‐assemble cardiac cells into spheroidal 3‐D tissues in vitro (reviewed in Fennema et al. 2013). The advantages of self‐assembly are organotypic density and architecture, maximal cell‐to‐cell communication and movement among different cell types, self‐produced ECM and proper adrenergic responses. Interconnectivity is essential to determine the potential for arrhythmogenesis and can be affected by the presence, expression pattern and functionality of gap junction proteins of the connexin (Cx) family. Self‐interspersion of CMs and Fbs (Desroches et al. 2012) and self‐networking of ECs in CM sheets (Sekine et al. 2008) have both been observed in vitro with neonatal rat heart cells, mimicking pluricellular organization in vivo. Importantly, stem‐cell derived CMs show enhanced maturation and CM enrichment when cultured in aggregates or cardiospheres compared to 2‐D cultures (Nguyen et al. 2014a; Beauchamp et al. 2015). In addition they self‐organize with Fbs and ECs to form tissue patches with vascular networks without the need for exogenous materials (Stevens et al. 2009). Inclusion of Fbs in self‐assembled tissues prolongs action potential duration (Desroches et al. 2012) and improves sarcomere organization and Cx43 expression (Ou et al. 2011).

Additive manufacturing or 3‐D printing also holds promise as an in vitro tool for mechanistic experimentation. The incorporation of living cells and biological factors was limited by exposure to high temperatures and laser energy, and complex microenvironments must first be deconstructed into a series of 2‐D layers. More recent developments allow for bioprinting precise architecture, pluricellularity and physiological microenvironments (reviewed in Duan, 2017). For example, laser‐induced‐forward‐transfer (LIFT) printing of mesenchymal cells and ECs (HUVECs) was used to generate defined patterns of EC lumens surrounded by support cells, increasing vessel formation and showing the utility of 3‐D printing (Gaebel et al. 2011).

Inducing native anisotropy and patterning

Cellular orientation is of critical importance for biological and mechanical tissue function, especially in the heart given its laminar structure and well‐organized arrangement of CMs (and other cell types) in the connective tissue architecture (LeGrice et al. 1995). Loss of cellular organization is often associated with tissue malfunctioning, and cell shape changes are known to impact cell viability, growth and signalling (Chen et al. 1997; Huang & Ingber, 1999). Loss of CM organization can impact excitability as the loss of T‐tubules affects excitation–contraction coupling (reviewed by Guo et al. 2013). A major goal for the engineering of cardiac in vitro models has been to incorporate cues that can modulate cell shape, orientation and alignment of CMs and non‐CMs to mimic the myocardial tissue in vivo.

CMs in the healthy adult heart are elongated and rod‐shaped and changes in their shape are concurrent with whole heart changes in diseased states such as cardiomyopathy and heart failure (Gerdes & Capasso, 1995; Kanzaki et al. 2012). Yet CMs that have lost their native shape are a central tenant of mechanistic research in vitro. Induction of uniaxial stretch was a pioneering technique to stimulate anisotropy in cardiac cells (e.g. Vandenburgh, 1992), which will be discussed as a component of the in vitro mechanical environment below. Micropatterned islands of adhesive proteins such as fibronectin can effectively influence morphogenesis of cardiac cells (Parker et al. 2008), and can be designed to maintain native length/width ratio of CMs (Bray et al. 2008). These substrates have been used to demonstrate changes in sarcomere organization (Bray et al. 2008) and nuclear morphology (Bray et al. 2010) in neonatal rat ventricular CMs. In addition to cellular anisotropy, CMs also align to each other at the tissue level to form compact, parallel myofibres, which is essential for spatially and temporally coordinated conduction and contraction. To mimic this organization, CMs have been cultured on 2‐D and 3‐D features such as microfabricated surface topography, patterned ECM/adhesive/non‐adhesive proteins (Karp et al. 2006), wrinkles (Luna et al. 2011), fibres (Kenar et al. 2011), biowires (Nunes et al. 2013) and decellularized ECM (Ott et al. 2008). For an in‐depth review of engineering cellular alignment in vitro, the reader is referred to Li et al. (2014). Most of the elongated patterns and structures used to generate CM anisotropy consist of straight lines, but other effective patterns include ‘accordion‐like honeycombs’ (Engelmayr et al. 2008) and traces of cardiac fibres from magnetic resonance imaging data (Badie & Bursac, 2009). Anisotropic arrangements of neonatal rodent and hESC‐CMs have been achieved with features from the nanoscale (D.‐H. Kim et al. 2010) or multiple scales of geometric cues (Luna et al. 2011), mimicking the ultrastructure of aligned ECM between aligned myofibres. Similar to single CMs on anisotropic patterns, CMs in anisotropic tissue cultures have increased cellular aspect ratios. Additionally, CMs in aligned clusters are often smaller and multinucleated (Pong et al. 2011), functionally connect via Cx43 and N‐cadherin (McDevitt et al. 2002; Motlagh et al. 2003), and have improved action potentials, conduction velocities (Thomas et al. 2000) and calcium handling (Pong et al. 2011) over isotropic cultures. These studies were performed with neonatal ventricular CMs from rat (McDevitt et al. 2002; Motlagh et al. 2003; Pong et al. 2011) or mouse (Thomas et al. 2000). Similarly, alignment of stem cell‐derived CMs induced greater molecular, morphological and functional maturation (Rao et al. 2013; Ribeiro et al. 2015).

Cellular shape and alignment are also important considerations for non‐CMs. Fbs and ECs were used in key studies to reveal details of the relationship between cellular function and morphology (e.g. Chen et al. 1997). 2‐D and 3‐D features can also influence both Fb and EC alignment, though Fbs appear to require shallower features than ECs (Biela et al. 2009). The effect of substrate anisotropy (and cyclic mechanical stretch, see below) on the orientation of adhesive cells is reviewed elsewhere (Tamiello et al. 2016).

Beyond controlling cell shape, orientation and alignment, micropatterns can be utilized to control cellular spatial location to generate controlled heterotypic cell–cell interactions (reviewed by Kaji et al. 2011). Studies in which CMs paired with Fbs, skeletal myoblasts, mesenchymal stem cells, engineered cell lines, or other CMs have shown that unexcitable cells modulate CM action potential shape and pacemaking activity, and that non‐CMs can couple to CMs via gap and adherens junctions and nanotubes (Pedrotty et al. 2008; McSpadden et al. 2012; Ma et al. 2013; Zhang et al. 2014). Until recently, random attachment of two cell types resulted in heterotypic pairs on less than 5% of patterns. More precision with laser‐patterned biochips (Ma et al. 2013) will improve the throughput of this approach. Isotropic co‐cultures fail to recapitulate the nature of crosstalk interactions, but there has been limited application of tissue‐level anisotropy to crosstalk models. Importantly, co‐culture improves CM alignment as pre‐seeding aligned scaffolds with Fbs improved sarcomere organization (Parrag et al. 2012), and aligning CMs and Fbs co‐cultures may change the makeup of heterocellular junctions (Thompson et al. 2011). These platforms point toward the temporal importance of pluricellular interactions as patterned templates cultured with CMs, Fbs and ECs simultaneously were not contractile, but pre‐culturing these patterns with Fbs or ECs produced contractile tissues with elongated, more mature CMs (Iyer et al. 2009). Alignment and position can also be considered in crosstalk models as patterned CMs with precisely controlled intervening areas of Fbs have been utilized to gain insights into electrical signal propagation by Fbs (Gaudesius et al. 2003).

Rebuilding native passive mechanical properties

The myocardium must be able to withstand mechanical loads and comply with pumping forces. Depending on the species, the heart experiences cyclic contraction and relaxation at the organ, tissue and cell level in the order of 1 Hz (human) to 10 Hz (mouse) over the entire lifespan. The mechanical properties of the cellular microenvironment play a key role in muscle functionality and differentiation, and mechanical forces can be transduced into biochemical signals capable of altering protein synthesis and gene transcription that induce changes in proliferation, hypertrophy, apoptosis, metabolic rates and gene expression (reviewed in Stoppel et al. 2016).

One characteristic of heart tissue that affects the forces experienced by cells is the material stiffness. Normal cardiac muscle has a Young's modulus of 10–15 kPa, whereas fibrotic tissue is typically 2–10 times higher (reviewed by van Putten et al. 2016). There is variability with age and species as neonatal rat heart has a modulus of 4–11.4 kPa, adult rat 11.9–70 kPa, and adult human 20–500 kPa (reviewed in Tallawi et al. 2014). While CMs can deform themselves and their surrounding microenvironment in vivo, they can only deform themselves on hard substrates like traditional cell culture materials (i.e. glass and polystyrene (GPa) or polydimethylsiloxane (MPa)). Softer 2‐D substrates can be fabricated without sacrificing compatibility and transparency. Decreasing substrate stiffness can improve rhythmic beating (as shown for embryonic quail and chick CMs; Engler et al. 2008), and increase force generation and calcium handling of neonatal CMs (Jacot et al. 2008), but the effect of substrate stiffness on myofibrils has been inconsistent across species (Engler et al. 2008; Galie et al. 2013; Yahalom‐Ronen et al. 2015). It appears that optimal differentiation and maturation occur on substrates with elasticity similar to that of the adult heart of the appropriate species (Jacot et al. 2008; Galie et al. 2013; Hazeltine et al. 2014). However, dynamic changes in substrate stiffness to mimic the increase in cardiac stiffness during development may have even greater effects, as shown for embryonic chick CMs (Young & Engler, 2011). Compliant substrates have also been shown to enhance stem cell differentiation toward CMs (Engler et al. 2006; Feaster et al. 2015), and material stiffness can also be applied as a 3‐D tool to control CM afterload (Jian et al. 2014).

Mechanical stress is also an important cue for non‐CMs. Fbs must be able to sense increased mechanical loads to respond with appropriate ECM deposition (reviewed in van Putten et al. 2016). The unnatural stiffness of classic cell culture substrates has led to speculation that many in vitro studies of Fbs actually characterize activated MyoFbs, since Fbs cultured on more compliant substrates do not express α smooth muscle actin (αSMA), whereas Fbs on substrates with stiffness >20 kPa have αSMA‐positive stress fibres (Yeung et al. 2005). ECs are also sensitive to substrate stiffness (e.g. cell spreading; Pompe et al. 2009). Non‐CMs also contribute to the material properties of the microenvironment. ECs are softer than CMs (Mathur et al. 2001), but the stiffness of pluricellular cardiac patches formed from hESC‐CMs, MEFs and HUVECs is 4‐fold higher than that of CM‐only patches (Stevens et al. 2009). In an effort to incorporate multiple cues from the cardiac microenvironment, stiffness and anisotropy have been combined for CMs (Wang et al. 2011; Agarwal et al. 2013; Annabi et al. 2013; Ribeiro et al. 2015).

Incorporating static and cyclic stretch

Matching in vitro substrate stiffness to the stiffness of heart tissue can improve the utility of cell culture platforms, but does not reproduce the directionality of static tension in the myocardium or induce the cyclic stress of contraction. Substrate deformation was initially utilized to produce uniaxial tension on cultured cells: mechanical loading was shown to induce neonatal rat CM hypertrophy and non‐CM hyperplasia, and gene expression changes were characterized to begin to explore potential mechanisms of cellular response to strain (Komuro et al. 1991; Sadoshima et al. 1992). Importantly, aligned CMs under perpendicular stretch accumulate branched myofibres and changes in actin organization (Simpson et al. 1999).

Since early in development when the heart starts to contract, cardiac growth and reorganization occurs in the presence of cyclic mechanical stress, which is altered in the diseased heart upon remodelling. Cyclic stretch has been widely investigated in 2‐D cardiac cultures and 3‐D cardiac tissues. Cyclic stretch on 2‐D substrates have shown alterations in fibril organization, hypertrophy, gap junctions, and a number of signalling pathways (reviewed in Stoppel et al. 2016). Cyclic stretch of EHTs induces hypertrophic growth, increases metabolism and mitochondrial density, and improved contractile function (Fink et al. 2000). Static and cyclic stress conditioning both promote cell and matrix alignment, but although cyclic stress conditioning promoted more CM hypertrophy within the construct, it conferred no additional benefit to CM alignment over static stress conditioning (Tulloch et al. 2011). Cyclic mechanical stimulation also seems to enhance stem cell differentiation, though a standardized and optimized protocol has not yet been reached (reviewed in Stoppel et al. 2016). Some studies have also looked at localized cyclic mechanical stimulation on CMs in vitro. Stimulation of constructs with an atomic force microscopy cantilever can enhance contractility (Boudou et al. 2011; Galie et al. 2013). Recent work has shown that isolated neonatal rat CMs can be trained to beat at a given frequency by mechanically stimulating the underlying substrate. A mechanical probe induced deformations aimed to mimic the deformations generated by neighbouring cells in vivo. CMs maintained this paced beating more than an hour after stimulation was ended (Nitsan et al. 2016). This work has not yet been extended to adult or stem cell‐derived CMs.

The effect of cyclic strain on Fbs depends on the maturation state of the Fbs, when the strain is applied, the strain rate and the strain duration (reviewed in van Putten et al. 2016), but overall, increased strain on Fbs induces activation to MyoFbs by changing αSMA expression and recruiting stress fibres. Stretch has also been shown to shift the resting membrane potential in Fbs (reviewed in Abramochkin et al. 2014), which could impact excitation–contraction coupling if coupled to CMs. When cultured without other cell types in 3‐D tissue gauges, Fbs have significant influence on collagen organization (Legant et al. 2009). The inclusion of Fbs or stromal cells in CM tissues improved their formation and compaction (Thavandiran et al. 2013). Biomechanical stimulation of ECs is a critical determinant for EC gene expression, morphology and function, with opposing effects on EC alignment (along the direction of laminar shear stress and away from the direction of cyclic stretch). This has been studied most for vascular ECs (reviewed by Chien, 2007), but Terracio et al. (1988) showed a long time ago that neonatal rat heart ECs also become elongated and oriented perpendicular to the direction of cyclic linear stretch, akin to CMs and Fbs studied in parallel. Mechanical forces with a clear direction cause a transient upregulation of proinflammatory and proliferative pathways, where those without a clear direction cause sustained effects (Chien, 2007).

Inducing native electrical cues

In lieu of action potentials inducing rhythmic cardiac contractions in vivo, electrical stimulation is used to control and maintain the functionality of CMs in vitro. Several systems are commercially available (Tandon et al. 2009), and optimal protocols are suggested for rat neonatal CMs (Tandon et al. 2011). Electrical stimulation affects cell alignment, differentiation, metabolic activity, protein synthesis, Cx43 expression, and conduction velocity, calcium handling and force generation (e.g. Radisic et al. 2004; reviewed in Stoppel et al. 2016). Effective stimulation protocols mimic in vivo electrical properties in shape and order of magnitude (i.e. rectangular, 2 ms, 5 V cm⁻¹, 1 Hz), but some studies show improved electrical properties for hiPSC‐CMs with stimulation protocols at higher frequencies than native to the originating species (Nunes et al. 2013). Electrical conditioning of neonatal rat cardiac microtissues only modestly improved their maturation when also mechanically stimulated (Boudou et al. 2011), so these cues may not have synergistic effects.

Non‐CMs may also be responsive to electrical stimulation though this has not been studied extensively. Fbs cultivated in DC electric field oriented perpendicular to the electric field lines migrated towards the cathodal end of the field (Erickson & Nuccitelli, 1984). Biphasic field stimulation can improve electrical excitability of cardiac microtissues pre‐cultured with Fbs and ECs prior to CM seeding by improving the 3‐D organization of the cells, increasing cellular elongation and enhancing Cx43 expression (Chiu et al. 2011). When alignment is combined with electrical stimulation, the response of CMs and Fbs was different: the topographic cue influenced the orientation of CM more than electrical stimulation, while Fbs seemed to integrate both cues (Au et al. 2007). Because micropatterning leads to predictable geometries, it enables electrophysiological studies (e.g. microelectrode arrays positioned under patterned CMs). Electrodes have been incorporated into engineered heart tissues (Boudou et al. 2011), and nanowires have been introduced into 3‐D spheroids (Richards et al. 2016).

Conclusions and future perspective

Crosstalk between CMs and non‐CMs is believed to be of critical importance for cardiac function in health and disease but the exact mechanisms are still poorly understood, because the complexity of the myocardium imposes substantial experimental challenges both in vitro and in vivo. Traditional 2‐D cell culture models fail to recapitulate the rich and complex contributions of the cardiac microenvironment, in which multiple cell types are embedded in the interstitium with a 3‐D anisotropic organization and continuously exposed to chemical, mechanical, and electrical stimuli. CMs, Fbs, ECs and other cell types communicate with each other via paracrine signalling, electrical and mechanical coupling, and interactions with ECM, but on tissue culture plastic these interactions are limited to a few planar cell–cell interactions on a non‐compliant surface, usually with non‐physiological coatings of ECM or synthetic molecules. Significant progress has been made in enhancing the maturation of stem cell‐derived and neonatal CMs and resulting engineered cardiac tissues, and contributions from non‐CMs have been reported (C. Kim et al. 2010). Cell shape, cell organization, electrical stimulation and mechanical cues are often considered in CM in vitro models, and a few studies have characterized the behaviour of non‐CMs in more organized, electrically active, or mechanical environments. Electro‐mechanical cues and cellular organization have yet to be widely applied to crosstalk studies, but given the phenotypic changes that occur individually in CMs and non‐CMs, these considerations are imperative. Advancements in the development of more complex pluricellular physiological platforms that incorporate diverse cues from the myocardial microenvironment hold promise for ultimate utility as physiologically relevant cardiac tissue‐like in vitro models with heterocellular communication for mechanistic biological research, disease modelling, therapeutic target identification, drug testing and regeneration. Importantly, it is not necessary (or possible) to mimic all of the cues in the cardiac microenvironment in vivo and to fully recapitulate the complexity of the myocardium. The gap between the structural, electrophysiological and contractile properties of engineered in vitro models to native cardiac tissue is still wide (reviewed by Eschenhagen et al. 2012; Yang et al. 2014; Eder et al. 2016; Tzatzalos et al. 2016), but a fully mature phenotype is not necessarily required depending on the application. For example, a less than mature phenotype is needed for regenerative and developmental studies, whereas disease modelling and pharmacological studies with relevance to the adult heart will benefit from a more adult‐like phenotype (Yang et al. 2014). The challenge will continue to be to determine for each model and specific application what the minimal requirements are for the microenvironmental cues to be mimicked and phenotype to be achieved.

Current limitations of cardiac in vitro models include reproducibility, accessibility, readouts, throughput and relevance. Few of the platforms reviewed here have standard cell sources, materials and protocols, which may lead to inconsistency in data collection and analysis. Protocol advancements have increased the efficiency of directed CM differentiation from human stem cells and are geared to foster maturation and reduce the heterogeneity of the CM population, but the relative immaturity remains a major obstacle. A better understanding of the disease contribution of Fbs of different origin (Ali et al. 2014) will aid in the refinement of Fb protocols and also enhance reproducibility. Many of the techniques and materials are sophisticated and not available to standard biological labs though the commercialization of platforms for mechanical and electrical stimulation, and 3‐D cell culture may improve access. Increasing platform complexity can complicate the techniques to gain structural, functional and molecular readouts. For example, imaging of 3‐D microtissues requires the adaptation of new tools for optimization (Kabadi et al. 2015). Techniques have been adapted to allow for multiple readouts from engineering tissues, such as assessment of molecular parameters, action potentials, calcium handling and contraction, and histological analysis. Techniques like laser capture microdissections (Datta et al. 2015) could aid in teasing out the contributions of individual cell types in pluricellular environments.

While it is well appreciated that the inclusion of non‐CMs to CM‐containing 3‐D tissues is beneficial, crosstalk studies are still limited in number and scope, with a focus on the communication between Fbs and CMs or ECs and CMs. Studies addressing crosstalk between Fbs and ECs, as well as communication with neuronal and immune cells, are also needed, and integration of more than two cell types is another important goal. Although much groundwork still needs to be laid for the development of pluricellular co‐culture models for cardiac crosstalk investigations, they can eventually be extended to disease studies by incorporating cells from remodelled hearts and/or the use of hiPSC‐CMs from patients with inherited mutations that affect cardiac structure or function. 3‐D platforms could also be developed to model tissue heterogeneity (e.g. endo/mid/epicardium) and fibrotic border zones. Validation of in vitro observations in the native tissue context are of critical importance and will be aided by the advancements with cardiac tissue slices (de Boer et al. 2009; Kang et al. 2016) and mouse models with cell type‐selective gene targeting models (reviewed in Minami & Aird, 2005; Doetschman & Azhar, 2012; Swonger et al. 2016).

Additional information

Competing interests

None declared.

Funding

Funding was provided by NIH/NHLBI 1F32HL126311(C.M.K.) and by NIH/NHLBI grant 1R01HL‐114784 (U.M.).

Biography

Celinda Kofron and Ulrike Mende are researchers at the Cardiovascular Research Center at Rhode Island Hospital and the Alpert Medical School at Brown University in Providence, Rhode Island. Dr Mende is a Professor of Medicine trained in pharmacology and molecular cardiology with expertise in G proteins and their regulators in cardiac cells. She received medical and postdoctoral training at the University of Hamburg (Germany). Her work as a principal investigator began at Brigham and Women's Hospital/Harvard Medical School and continued in establishing the Cardiovascular Research Center in 2005. Dr Kofron is a biomedical engineer, earning her doctorate at Brown University with a focus on models of neurite guidance. As a postdoctoral fellow, she applies her knowledge of in vitro cues and cellular responses to cardiac models. The multidisciplinary Mende laboratory targets fibroblasts as key regulators of myocytes and contributors to heart failure and arrhythmias using in vivo and in vitro platforms.