Advancing functional engineered cardiac tissues toward a preclinical model of human myocardium

Abstract

Cardiac experimental biology and translational research would benefit from an in vitro surrogate for human heart muscle. This study investigated structural and functional properties and interventional responses of human engineered cardiac tissues (hECTs) compared to human myocardium. Human embryonic stem cell-derived cardiomyocytes (hESC-CMs, >90% troponin-positive) were mixed with collagen and cultured on force-sensing elastomer devices. hECTs resembled trabecular muscle and beat spontaneously (1.18±0.48 Hz). Microstructural features and mRNA expression of cardiac-specific genes (α-MHC, SERCA2a, and ACTC1) were comparable to human myocardium. Optical mapping revealed cardiac refractoriness with loss of 1:1 capture above 3 Hz, and cycle length dependence of the action potential duration, recapitulating key features of cardiac electrophysiology. hECTs reconstituted the Frank-Starling mechanism, generating an average maximum twitch stress of 660 μN/mm² at L_max, approaching values in newborn human myocardium. Dose-response curves followed exponential pharmacodynamics models for calcium chloride (EC₅₀ 1.8 mM) and verapamil (IC₅₀ 0.61 μM); isoproterenol elicited a positive chronotropic but negligible inotropic response, suggesting sarcoplasmic reticulum immaturity. hECTs were amenable to gene transfer, demonstrated by successful transduction with Ad.GFP. Such 3-D hECTs recapitulate an early developmental stage of human myocardium and promise to offer an alternative preclinical model for cardiology research.—Turnbull, I. C., Karakikes, I., Serrao, G. W., Backeris, P., Lee, J.-J., Xie, C., Senyei, G., Gordon, R. E., Li, R. A., Akar, F. G., Hajjar, R. J., Hulot, J.-S., Costa, K. D. Advancing functional engineered cardiac tissues toward a preclinical model of human myocardium.

Human embryonic stem cell-derived cardiomyocytes (hESC-CMs) are an attractive cell source as an in vitro model system to study human cardiac physiology and development (1) and offer an important species-specific advantage over cultured animal cells for developing new therapeutics (2). Most hESC-CM-based experiments have relied on traditional 2-dimensional (2D) culture conditions, but the nonphysiological environment of rigid tissue culture plastic can adversely affect cardiomyocyte function and maturation (3, 4). Tissue engineering offers a more physiological 3-dimensional (3D) culture platform to examine cell-cell and cell-matrix interactions; it allows the direct measurement of electromechanical function, and it may promote transition of immature cells to a more adult-like phenotype (5–7). It is, therefore, compelling if hESC-CMs can be used to engineer surrogate human cardiac tissue.

Important advances have been made toward developing human engineered cardiac tissues (hECTs) as patches for myocardial repair, using biodegradable scaffolds (2, 8), fibrin-based hydrogels (9), or scaffold-free constructs (10, 11). However, such patches are better suited for electrical propagation studies than for quantifying contractile twitch function. The tissue engineering field has also recognized the value of developing in vitro models for basic science, as well as drug discovery and toxicology screening applications (12, 13). Indeed, high-throughput screening platforms have been designed using microelectromechanical system technologies, including muscular thin films (MTFs; ref. 14) and microfabricated tissue gauges (μTUGs; ref. 15), but these have yet to be applied for human cardiac tissue engineering.

Murry and colleagues (16) were the first to report beating hECTs created from pluripotent human stem cells, demonstrating the length-dependent Frank-Starling mechanism of heart muscle, and showing that mechanical loading during hECT culture induced features of early cardiomyocyte development. Eschenhagen and Zimmerman and colleagues, pioneers in the cardiac tissue engineering field (17), more recently published a study using hECTs created from human embryonic stem cells, demonstrating responses to proarrhythmic compounds similar to the industry standard hERG channel assay (18). They also found that the 3-D culture environment of hECT was more favorable than embryoid body culture for inducing cardiomyocyte maturation. Nevertheless, interpretation of hECT responses to such interventions remained complicated by the heterogeneous population of differentiated cells, nearly half of which comprised an unspecified noncardiomyoycyte lineage (16, 18). More recently, functional cardiac tissues have been created using purified human cardiomyocyte populations (9, 19), allowing the potential benefits of nonmyocytes to be systematically examined (9). However, hECT characterization has primarily focused on comparisons with alternative embryoid body or monolayer culture conditions rather than with the desired target of natural human cardiac muscle. Thus, for engineered tissue constructs to serve as reliable in vitro preclinical models of human myocardium, further development is required, and comprehensive testing and characterization remain a priority.

Therefore, the objectives of the present study were to help advance the field of human cardiac tissue engineering by examining hECTs created from enriched hESC-CM cell populations obtained using an efficient small-molecule-mediated directed differentiation, by expanding the characterization of hECTs using multiple complementary testing platforms, and by analyzing the dose response to known inotropic agents with comparisons to natural human cardiac muscle. The 3-D hECTs described herein demonstrate ultrastructural features and cardiac gene expression levels approaching native human myocardium, with electrical, mechanical, and pharmacologic responses showing functional properties that resemble aspects of immature human cardiac physiology. The efficacy of adenoviral-mediated gene transfer is also demonstrated, opening a new avenue for developing gene therapies using engineered human heart muscle.

MATERIALS AND METHODS

Generation of hECT constructs

All experiments used cells derived from the HES2 hESC line (Wicell, Madison, WI, USA) propagated in feeder-free culture, as described previously (20). For cardiac differentiation, we used a new small-molecule directed differentiation technique (21); small clusters (50–100 cells) were cultured in suspension on ultra-low-attachment cell culture dishes (Corning, Lowell, MA, USA) with differentiation medium (StemPro34, 50 μg/ml ascorbic acid and 2 mM GlutaMAx-I; Invitrogen Carlsbad, CA, USA) supplemented with recombinant human (rh) cytokines and small molecules at the following concentrations for the indicated time periods; d 0–1: rh bone morphogenic protein 4 (rhBMP4; 10 ng/ml) and blebbistatin (5 μM); d 1–4.5: rhBMP4 (10 ng/ml) and rhActivin-A (25 ng/ml); and d 4.5–8: Wnt antagonist 1 (IWR-1; 2.5 μM). The differentiated cardiomyocytes were maintained in differentiation medium without supplements after d 8 until used for ECT creation. All cytokines were purchased from R&D Systems (Minneapolis, MN, USA), and the small molecules were purchased from Sigma-Aldrich (St. Louis, MO, USA). All differentiation cultures were incubated in a humidified, 37°C, 5% CO₂ environment. On d 13 of differentiation, cell clusters were enzymatically dissociated into single cells with trypsin/EDTA solution (0.04% trypsin, 0.03% EDTA; Promocell, Heidelberg, Germany) for 15 min and then resuspended in differentiation medium at a concentration of 0.2 × 10⁶ cells/ml and cultured in suspension.

After 48 h, the cells were collected by centrifugation (800 rpm, 5 min), resuspended in culture medium at a concentration of 0.75 × 10⁸ to 1 × 10⁸ cells/ml, and combined with a mixture of ice-cold bovine collagen type I (Sigma-Aldrich) and Matrigel (BD Biosciences, San Jose, CA, USA), at a ratio of 1:8:1 (v/v/v), as described previously (22). The cells/collagen/Matrigel mixture was then pipetted into a custom mold made of polydimethylsiloxane (PDMS) elastomer (Fig. 1), filling the rectangular well with 100 μl (10⁶ cells/well), and then incubated at 37°C, 5% CO₂ for 2 h to allow the collagen to polymerize. They were then bathed with differentiation medium (StemPro34, ascorbic acid, and GlutaMAX-I), and maintained in culture with daily half-medium exchanges. Inserts in the casting mold (Fig. 1A) were removed at 48 h, yielding a self-assembled ECT held between 0.5-mm-diameter PDMS posts at each end. At 5 d after ECT creation, culture medium was changed to high-glucose Dulbecco's modified Eagle medium (DMEM) supplemented with 10% newborn bovine serum (Atlanta Biologicals, Lawrenceville, GA, USA), 1% penicillin/streptomycin (Gibco, Carlsbad, CA, USA), and 0.2% amphotericin B (Sigma-Aldrich).

Figure 1.

Creation of hECT. A) Elastomer mold with integrated endposts and removable inserts. B) Side view of mold with attached hECT 23 d after creation. C–E) Inverted light microscope images at one end of an hECT at 0, 4, and 10 d after creation, respectively, showing tissue compaction with time in culture. Scale bars = 2 mm. C) Initially, hESC-CMs appeared as dispersed clusters across the thickness of the hECT, with isolated spontaneous contractions within the 3D construct. D) By d 4, cell alignment became more evident as the tissue compacted with a dense central region, and rhythmic contractions were visible within the tissue. E) By d 7–10, the hECT was highly compacted with well-defined edges.

Characterization of ECT contractile function and electrophysiology

Characterization of contractile function of the hECTs was performed using two methods. For the first, the integrated flexible PDMS posts were used as force sensors, with the post deflection captured in real time with a high-speed camera (100 frames/s) and LabView software (National Instruments, Austin, TX, USA), applying a beam-bending equation from elasticity theory to calculate twitch force, as described previously (22). These measurements were obtained with the hECTs maintained in the original molds where they were created, and all experiments were carried out under sterile conditions inside a laminar flow hood, with the hECTs bathed in culture medium at 37°C, with and without electrical field stimulation.

For the second method of testing, the hECTs were transferred to a physiological muscle bath (Scientific Instruments, Heidelberg, Germany) and held between a force transducer and a stepper motor to control stretch and perform measurements under isometric conditions. The tissues were superfused with gassed (95% O₂ and 5% CO₂) Tyrode's solution (in mM: 2.2 MgCl₂, 2.6 KCl, 136.8 NaCl, 0.4 NaH₂PO₄, 5.5 glucose, 11.9 NaHCO₃, and 2.3 CaCl₂; all reagents from Sigma-Aldrich) at 37°C under electrical field stimulation, using a 5-ms suprathreshold pulse, with a stepper motor for length control and high-sensitivity transducer (Scientific Instruments) for measuring isometric twitch force, controlled by a custom LabView program. After establishing force-length and force-frequency relationships, the inotropic dose responses to step changes in calcium chloride (CaCl₂; Sigma-Aldrich), verapamil (Sigma-Aldrich), and isoproterenol (Hospira, Lake Forest, IL, USA) concentrations were evaluated.

Optical action potential mapping

Action potential recordings were carried out using an optical mapping system. hECT preparations were stained with the voltage-sensitive dye di-4-ANEPPS (Invitrogen) and perfused with Tyrode's solution at 36°C in a custom-designed imaging chamber equipped to record action potentials with high spatial and temporal resolution (23). To suppress motion artifacts, hECT preparations were stabilized by rigid pins and treated with 10 mM 2,3-butanedione monoxime (BDM; Sigma-Aldrich). To examine the rate dependence of the action potential duration (APD) and assess the refractoriness and loss of 1:1 pacing, hECT preparations underwent a stimulation protocol with pacing cycle length (PCL) decreasing from 1000 to 200 ms, using a 1.5× diastolic threshold amplitude with 2-ms pulse width.

Characterization of hECT structure

Hematoxilyn and eosin (H&E) staining was performed on 4-μm-thick sections from hECTs fixed in 4% paraformaldehyde and embedded in paraffin. For immunofluorescence, fixed hECTs were frozen and embedded in Tissue-Tek optimal cutting temperature (OCT) compound (Sakura, Torrance, CA, USA). Cryosections (8 μm) were stained for sarcomeric α-actinin (ab72592, 1:400; Abcam, Cambridge, MA, USA) and membrane-bound connexin 43 (MAB3068, 1:100; Millipore, Billerica, MA, USA), followed by Alexa Fluor 488 and Alexa Fluor 555 secondary antibodies (A21206 and A31570, 1:500; Invitrogen). Samples were also stained with fluorescein isothiocyanate (FITC)-conjugated anti-α-smooth muscle actin (α-SMA) antibody (F3777, Sigma-Aldrich). Sarcoplasmic reticulum (SR) was stained using anti-sarcoendoplasmic reticulum Ca²⁺-ATPase (SERCA2a) antibody (1:100; 21st Century Biochemicals, Marlborough, MA, USA) and Alexa Fluor 555 (A31572, 1:500; Invitrogen). All sections were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and mounted in Vectashield (Vector Laboratories, Burlingame, CA, USA). Images were obtained using an epifluorescent microscope (IX71; Olympus, Center Valley, PA, USA) or a laser-scanning confocal microscope (Leica TCS SP5 DMI; Leica Microsystems, Buffalo Grove, IL, USA). Paraffin-embedded sections were used for troponin T staining using anti-troponin T, cardiac isoform AB-1, clone:13-11 (1:100; Thermo Scientific, Pittsburgh, PA, USA) and secondary antibody Alexa Fluor 568 (A11004; Invitrogen). For transmission electron microscopy (TEM), hECTs were fixed in 3% glutaraldehyde, and then processed at the Electron Microscopy Core Facility of the Icahn School of Medicine at Mount Sinai; thin sections (50–60 nm) stained with uranyl acetate solution and Reynold's lead citrate solution were examined in a transmission electron microscope (H-7650; Hitachi High Technologies, Pleasanton, CA, USA). Qualitative assessment was made on 4 different hECTs ranging from 18 to 23 d old.

Characterization of hECT genetic profile

Relative gene expression was determined using 2-step real-time quantitative polymerase chain reaction (qPCR). Total RNA was extracted from hECTs using the RNeasy isolation kit with on-column DNase I treatment (Qiagen, Valencia, CA, USA) to eliminate contaminating genomic DNA, according to the manufacturer's protocol, and RNA was reverse transcribed using the high-capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA, USA). Real-time qPCR was performed using iTaq SYBR Green Mastermix with Rox (Bio-Rad, Hercules, CA, USA) using the 7500 real-time PCR system (Applied Biosystems). The PCR protocol consisted of 1 cycle at 95°C (3 min) followed by 40 cycles of 95°C (15 s) and 60°C (1 min), followed by a dissociation step. The primers used in qPCR were α-cardiac muscle actin (ACTC1), α-myosin heavy chain (α-MHC), β-myosin heavy chain (β-MHC), atrial natriuretic factor (ANF), and SERCA2a. Fold changes in gene expression were determined using the ΔΔC_t method relative to the normal adult human left ventricular (LV) myocardium, with normalization to β-2-microgobulin (B2M) endogenous control.

Gene transfer with recombinant adenovirus

Recombinant adenoviruses expressing humanized Renilla green fluorescent protein (hrGFP) under the control of the cytomegalovirus (CMV) promoter were generated using the AdEasy adenoviral vector system (Agilent Technologies, Santa Clara, CA, USA), as described previously (24). Briefly, the linearized pshuttle-IRES-hrGFP vector was introduced into the BJ5183 bacterial cells that harbor the AdEasy-1 backbone vector, mediating homologous recombination between the shuttle vector and the adenoviral backbone vector. The resulting recombinant adenoviral DNA was PacI digested and transfected into AD-293 cells (Agilent Technologies). The viruses were harvested 14–20 d later and further amplified in AD-293 cells. The viral titers were determined by plaque-forming assays, and their replication deficiency was confirmed by the absence of the wild-type E1 region. HES2-derived cardiomyocytes were infected 24 h prior to creating hECTs, or after hECT creation on culture d 2 at multiplicity of infection (MOI) of 10 and 100, respectively. The expression of GFP was monitored by fluorescence microscopy and confirmed by PCR.

Statistical analysis

Descriptive statistics are reported as means ± sd. Repeated-measures analysis of variance (ANOVA) with a single within-subjects factor, with post hoc pairwise contrast analysis of means, was used to examine effects of percentage muscle length for maximum active tension (L_max), [CaCl₂], [verapamil], and [isoproterenol] on developed stress, and for the effect of PCL on APD. Statistical significance was accepted at the P < 0.05 level, using Greenhouse-Geisser (G-G) sphericity correction. Linear regression was used to test significance of the force-frequency relationship. Student's t test was used for comparison of individual gene expression levels at 2- and 3-wk culture times.

RESULTS

Three-dimensional hECTs developed from hESC-CMs

A mixed population of cardiomyocytes and nonmyocytes can improve viability and contractile performance of 2D and 3D cardiac cell cultures (6, 9, 11, 25–27) but may also have disadvantages for some applications (6, 9, 28). Interpretation of cell-cell interactions is further complicated when a “cardiomyocyte-only” population includes substantial fractions of nonmyocytes. For this study, hESCs were driven toward the cardiac lineage by small-molecule-mediated directed differentiation that yields ∼3 hESC-CMs/original stem cell in enriched populations that were >90% positive for expression of cardiomyocyte-specific markers cardiac troponin-T type 2 (TNNT2) and α-sarcomeric actinin (ACTN2), with <10% positive for smooth muscle myosin heavy chain (MYH11) and <5% positive for the endothelial cell markers CD31 and CD34 (Supplemental Fig. S1). Without requiring any further cell purification, the resulting hESC-CMs (10⁶ cells/hECT) were mixed with type I collagen-Matrigel scaffold solution and cultured in a custom PDMS mold (Fig. 1), yielding 2 to 4 hECTs from 10⁶ starting hESCs/preparation. Over several days, the embedded cells self-assembled into a thin, spontaneously beating hECT held between the integrated flexible endposts of the PDMS mold, which deflected inward with each contraction (Supplemental Video S1).

hECT histology reveals features of cardiac microstructure

Histological analysis revealed highly cellular tissues with anisotropic, heterogeneous microstructural organization (Fig. 2A), including a central core region comprising a dense concentration of aligned, TNNT2-, and ACTN2-expressing cardiomyocytes exhibiting organized sarcomeres (Fig. 2B). The membrane-bound gap junction protein connexin 43 exhibited a punctate distribution along cell boundaries rather than being restricted to intercalated discs (Fig. 2C), similar to neonatal human heart tissue (29). Cells lining the edges of the hECT stained positive for α-SMA (Fig. 2D), with a cell-depleted collagenous transition zone. On the basis of the expression of SERCA2a, which localizes to the SR membrane, hESC-CMs within hECTs contained SR structures (Fig. 2E), but the appearance lacked the organized longitudinal patterns typically seen in normal adult cardiomyocytes. TEM revealed typical ultrastructural features of cardiomyocytes (Fig. 2F), including parallel myofilaments within sarcomeres with well-formed Z discs, intercellular junctions at lateral cell boundaries, and a sarcolemma with caveolae and well-formed basement membrane. Evidence of heterochromatic nuclei and small clusters of mitochondria with intact cristae was apparent throughout the cell-dense core region, and distance from the tissue periphery did not appear to factor against cell viability.

Figure 2.

Structural and molecular analysis of hECTs. A) H&E stain of hECT sections shows highly cellular core (thick arrow) with compact borders (thin arrows) separated by a dense collagen zone (brackets). B) Confocal microscopy of hECT labeled with cardiac troponin T (red) and DAPI (blue) reveals cardiomyocytes with aligned myofibrils and organized sarcomeres. Magnified insets (A, B) correspond to boxed regions. C) α-Sarcomeric actin (green), membrane-bound connexin 43 (red) and DAPI (blue) show punctate gap junction protein along intermyocyte boundaries. D) Epifluorescence for α-SMA (green) and DAPI (blue) shows preferential assembly of stromal cells at the hECT border (arrows). E) Confocal microscopy of SERCA2a (red) and DAPI (blue) shows SR structures that lack patterns of longitudinal organization. F) TEM at ×12,000 shows organized sarcomeres with Z discs (Z) and myofilaments (brackets); nascent intercellular junctions (J); caveolae (C); mitochondria (M) with intact cristae; and cytoplasmic glycogen (gly).

hECT cardiac gene expression compared to adult human myocardium

While gene expression of hECTs has been evaluated relative to age-matched embryoid bodies (18) and monolayer cultures (9), comparison to adult human myocardium has been limited (30). Real-time qPCR analysis of mRNA extracted from hECTs after 2 or 3 wk in culture revealed mean expression levels of key cardiac genes (α-MHC, SERCA2a, and ACTC1) that approached (0.4× to 1.0×) levels in adult human LV myocardium, whereas β-MHC expression was <0.1× and ANF expression was ∼10× (Fig. 3). Interestingly, a 10× increase in ANF mRNA expression was also reported for human fetal heart compared to nonfailing adult heart mRNA (31). Several genes showed trends toward adult LV mRNA expression levels from wk 2 to 3 (increasing β-MHC, SERCA2a, ACTC1, and decreasing ANF), suggesting a possible phenotypic maturation of hESC-CM during hECT culture.

Figure 3.

Real-time qPCR analysis of hECTs after culture for 14.8 ± 1.5 d (solid bars, n=4) or 22.3 ± 1.3 d (shaded bars, n=4), relative to mRNA from adult human LV myocardium. Pairwise comparisons were not statistically significant, but showed trends toward adult expression levels (unity line) from 2 to 3 wk. Primers: α-MHC, β-MHC, SERCA2a, ACTC1, and ANF.

PDMS culture device allows noninvasive assessment of hECT contractile properties

We (22) and others (15, 32) have developed PDMS culture devices based on a simple beam-bending principle for optically monitoring forces generated by engineered tissues in a sterile, temperature-controlled environment. Our hECTs beat spontaneously with an average rate of 1.18 ± 0.48 Hz [∼70 beats per minute (bpm); n=13], and responded to pacing by electrical field stimulation with an excitation threshold of 128 ± 90 mV/mm required to capture the tissue using a 2-Hz, 5-ms pulse train. The resulting twitch force stabilized between d 5 and 10 of culture, and the hECTs remained viable and responsive to electrical stimulation for ≥4 wk in culture. We used a suprathreshold stimulus, and hECTs were reliably paced to at least 3 Hz (180 bpm), with loss of 1:1 capture at higher frequencies. An expanded tracing of hECT force during 2-Hz electrical stimulation is provided in Fig. 4A, also showing the filtered data used to identify the maximum and minimum force for each twitch; the difference yielded the developed force (DF) used for characterizing contractile function, and other twitch parameters, including time to peak force (TPF), time from peak to 90% relaxation (RT₉₀), maximum rate of force increase (+dF/dt), and maximum rate of force decrease (−dF/dt), are also readily computed. Treatment with isoproterenol, a β-adrenergic agonist, caused a significant dose-dependent increase in spontaneous pulse rate (Fig. 4B), indicating a positive chronotropic response, as expected for human heart muscle.

Figure 4.

Contractile function of hECTs on PDMS device. A) Representative twitch tracing during pacing at 2 Hz shows raw force vs. time and illustrates parameters used for twitch analysis, including developed force (DF), time to peak force (TPF), time to 90% relaxation (RT₉₀), maximum rate of force increase (+dF/dt), and maximum rate of force decrease (−dF/dt). Squares and triangles show detected maximum and minimum force per twitch based on the smoothed data (trace). Inset: top view of hECT on PDMS posts during testing. B) Spontaneous beating rate (relative to pretreatment) dose response to increasing concentrations of isoproterenol, demonstrating a positive chronotropic effect (means±sd; n=3 hECTs). *P < 0.03 vs. pretreatment.

Isometric twitch properties of hECTs in a physiological muscle bath demonstrate characteristics of human cardiac muscle

The physiological muscle bath is widely used for testing contractile properties of cardiac tissue samples (33, 34) and provides an established basis for comparison of hECT twitch properties with a wealth of published data on human myocardium. For this study, hECTs were initially cultured on the PDMS mold for 18–24 d and then mounted in the muscle bath. The force-length relationship was investigated with incremental stretches of 0.025 mm under continuous 2-Hz pacing (Fig. 5A). Twitch stress was obtained from force normalized by cross-sectional area, which was 0.96 ± 0.19 mm² for these hECTs (n=5). Maximum and minimum twitch stresses increased with axial stretch (Fig. 5B), as did the resulting developed stress (DS = max − min), which reached a plateau at a tissue length defined as L_max (Fig. 5B, inset). The increase in DS, from 0.33 ± 0.13 mN/mm² at 90% L_max to 0.66 ± 0.19 mN/mm² at 100% L_max (Fig. 5C), was statistically significant by repeated-measures ANOVA (P<0.005), indicating a functional Frank-Starling mechanism as found in natural human myocardium (35).

Figure 5.

Length and frequency dependence of hECT contractions in muscle bath system. A) Raw isometric twitch tracing during 0.025-mm step changes in hECT length while pacing at 2 Hz. B) Representative maximum and minimum twitch stress (force/area) vs. axial stretch (percentage of original length), reflecting systolic and diastolic properties of hECTs. Inset: corresponding developed (max-min) stress increases with tissue stretch, reaching a well-defined plateau at a stretched tissue length that defines L_max. C) Developed stress at 90, 95, and 100% L_max (means±sd; n=5), indicating a functional Frank-Starling mechanism. *P < 0.03 vs. 100% L_max. D) Stress-frequency relationship showed a statistically significant downward slope by linear regression (means±sd; n=5; P<0.0001). Inset: representative twitches at 1 and 3 Hz, illustrating a lower systolic force and slightly elevated diastolic force at the higher frequency.

The force-frequency relation was examined by increasing stimulation rate from 1.0 to 3.0 Hz with the hECT held at 95% L_max. Mean developed stress was relatively uniform at 0.58 mN/mm² below 1.5 Hz, but decreased at higher frequencies (Fig. 5D), whereas healthy adult human heart muscle contracts more strongly at higher stimulation rates. Table 1 summarizes the isometric twitch parameters at baseline conditions (2-Hz stimulation, 95% L_max, 2.3 mM CaCl₂) and at 1 Hz. More rapid pacing is accompanied by shorter twitch duration parameters (TPF and RT₉₀) as expected, and although the reduced DS and slower rise and decline rates (+dF/dt and −dF/dt) are generally associated with failing rather than healthy adult myocardium (33, 35), these trends are also observed in newborn heart muscle (34), suggesting that hECTs may represent an immature human cardiac phenotype.

Table 1.

Isometric twitch parameters before and after intervention in hECTs, with trends compared to published studies on human cardiac cells and tissues

Intervention	DS (mN/mm2)	TPF (ms)	+dF/dt (mN/s/mm2)	−dF/dt (mN/s/mm2)	RT90 (ms)
Baseline: 2.3 mM CaCl2, 2 Hz; n = 5	0.51 ± 0.07	84.9 ± 5.1	7.22 ± 0.64	5.04 ± 0.66	115.3 ± 9.1
Frequency: 1 Hz; n = 5	0.57 ± 0.07***	90.1 ± 5.4**	7.63 ± 0.51*	5.49 ± 0.46*	118.9 ± 8.3
Compares to	M (33), N (34)	H (33), M (33), N (34)	M (33)	M (33)	H (33), M (33), N (34)
CaCl2: 1 mM; n = 4	0.26 ± 0.14**	70.3 ± 11.9*	4.01 ± 1.35**	3.35 ± 1.25*	82.3 ± 16.3*
Compares to	H (33, 36), I (37), M (33, 36)	H (38)	H (36), M (36), I (37)	H (36), M (36)	H (38), M (38)
Verapamil: 1 μM; n = 5	0.10 ± 0.09***	51.6 ± 3.8***	2.90 ± 1.44***	1.95 ± 0.91***	70.2 ± 9.3***
Compares to	H (39), M (40), I (37)		H (39)	H (39)
Isoproterenol: 1 μM; n = 4	0.41 ± 0.24	74.6 ± 2.3**	6.41 ± 3.73	4.54 ± 2.85	105.0 ± 15.1
Compares to	I (37)	H (41), M (42)	I (37)	I (37)	I (37)

All measurements obtained at 95% L_max, 2 Hz, 2.3 mM CaCl₂ unless otherwise specified. Values represent means ± sd; numbers in parentheses indicate references. “Compares to” subentries indicate the direction of change from baseline was consistent with cited reports for healthy (H) or myopathic (M) adult human myocardium, newborn (N) human myocardium, or immature (I) human embryonic stem cell-derived cardiomyocytes.

^*P < 0.05

^**P < 0.005

^***P < 0.0005 vs. baseline using Student's t test (unpaired, except for frequency).

hECT responses to pharmacological interventions reveal similarities and differences with natural myocardium

Here we examine the pharmacodynamics of several well-known agents in terms of their dose-dependent effects on twitch force. A positive inotropic dose response was observed for hECT (95% L_max, 2-Hz pacing) exposed to incremental concentrations of extracellular CaCl₂ from 0.5 to 2.5 mM, showing an increase in systolic force and decrease in resting tension for each step change (Fig. 6A), as observed with natural myocardium (43). The average relative DS dose-response curve followed a 1-phase exponential growth model (Fig. 6B) yielding a half-maximal effective concentration (EC₅₀) of 1.8 mM. Lower than the EC₅₀ of 3.0 mM reported for adult human atrial trabeculae (44), the discrepancy may partly reflect known age-dependent differences in cardiac calcium sensitivity (45).

Figure 6.

Inotropic responses of hECTs in muscle bath system (95% L_max, 2 Hz) for calcium (A, B), verapamil (C, D), and isoproterenol (E, F). A) Example raw maximum and minimum twitch forces vs. time for 0.5 mM increments in [CaCl₂] from 0.5 to 2.5 mM. B) Developed stress (means±sd; n=4) relative response to external [Ca²⁺] (normalized to baseline at 2.3 mM) with fitted exponential growth model. *P < 0.01 vs. 2.5 mM. C) Example twitch forces vs. time for 10-fold increments in [verapamil] from 10⁻⁹ to 10⁻⁵ M. D) Relative response to verapamil (normalized to pretreatment) with fitted exponential decay model (means±sd; n=5). * P < 0.02 vs. pretreatment. E) Example twitch forces vs. time for increments in [isoproterenol] from 10⁻⁹ to 10⁻⁵ M. F) Relative response to isoproterenol (steady state normalized to pretreatment), with fitted exponential growth model (means±sd; n=4). Overall isoproterenol effect not significant by repeated-measures ANOVA (P=0.12).

hECTs exhibited a negative inotropic response to treatment with the L-type calcium channel blocker, verapamil, reflecting a decrease in systolic force with sequential increments in concentration from 1 nM to 10 μM (Fig. 6C). A semilog plot of the relative dose response (Fig. 6D) exhibited a sigmoidal decrease in DS, which followed a 1-phase exponential decay. The resulting half-maximal inhibitory concentration (IC₅₀) of 0.61 μM was comparable to 0.63 μM determined from the inotropic response of human papillary muscles (40).

Treatment with incremental doses of isoproterenol from 1 nM to 10 μM, while pacing at 2 Hz, elicited transient increases in systolic twitch force, but these rapidly returned to near-pretreatment levels (Fig. 6E) so that, in contrast to natural myocardium, the sustained inotropic effect was not statistically significant. The fitted relative dose-response curve (Fig. 6F) yielded an EC₅₀ value of 750 nM, which is substantially higher than the range of 15 to 50 nM reported for young nonfailing and old failing human myocardium (46). One study reported a dramatic age-related decrease in β-adrenergic responsiveness in nonfailing myocardium from older patients, with an EC₅₀ value of 460 nM (47), implicating possible down-regulation and decreased sensitivity of β₁ and β₂ receptors in hECTs. Alternatively, the combination of a strong chronotropic response (see Fig. 4B) with a negligible inotropic response to β-adrenergic stimulation may reflect functional immaturity of the SR (37), consistent with the structurally immature SR revealed by SERCA2a staining (Fig. 2E).

As summarized in Table 1, when hECTs were exposed to interventions that have a negative inotropic effect (reduced [Ca²⁺]_o or 1 μM verapamil), TPF, +dF/dt, −dF/dt, and RT₉₀ all decreased significantly along with the diminished DS. Several of these trends are also observed in natural human myocardium (see Table 1). In the presence of 1 μM isoproterenol, the isometric twitch parameters remained similar to the baseline conditions specified above, with only the decrease in TPF achieving statistical significance; correlations with reports from natural myocardium were inconsistent.

Optical mapping of hECT electrophysiology

Electrical functionality of the hECTs was tested using optical action potential (AP) mapping (23). Using the voltage-sensitive fluorescent dye di-4-ANEPPS, 3 preparations (of 15 attempted) exhibited motion-suppressed AP morphologies and stable signal-to-noise characteristics that allowed accurate steady-state APD measurement. Fig. 7 shows representative AP traces recorded from an hECT preparation during stimulation at a wide range of PCLs (PCL=frequency⁻¹), exhibiting a relatively fast upstroke velocity followed by a more gradual repolarizing phase (Fig. 7B). The APD shortened significantly at fast capture rates, but at slower rates (longer PCL), APD values were stable at ∼245 ms (Fig. 7C), comparable to measurements in human myocardium (48). As noted above, rapid stimulation of hECT preparations consistently resulted in the eventual loss of 1:1 capture (Fig. 7B), reflecting the underlying refractoriness of cardiac tissue.

Figure 7.

Optical mapping of hECT electrophysiology. A) Fluorescence image of one end of hECT mounted for optical mapping. B) Raw fluorescence intensity data (arbitrary units) from hECT AP recordings at several PCLs. C) APD as a function of PCL (measurement protocol started at 1000 ms PCL; means±sd n=3 hECTs). *P < 0.01 vs. 1000 ms PCL.

hECTs support gene transfer with recombinant adenovirus

We examined, for the first time, the ability of recombinant adenovirus to mediate gene transfer into hECTs. The hESC-CMs were infected either 24 h prior to creation of hECTs, or on d 2 of culture after hECT creation, with a recombinant adenovirus GFP (Ad.GFP) vector containing the hrGFP gene under control of the CMV promoter. The hECTs were successfully transduced, regardless of the timing of adenoviral delivery, as monitored by fluorescence imaging and confirmed by PCR (Fig. 8). However, the viral load required for transduction was higher after ECT generation, whereby a 10-fold higher MOI (MOI 100) was required to transduce the hECTs at 2 d after creation to achieve GFP expression levels comparable to preinfected hESC-CMs (MOI 10). Infection of compacted hECTs on d 7 of culture (MOI 100) was equally effective (not shown). Notably, similar to noninfected control hECTs, the Ad.GFP-infected hECTs were highly compacted with well-defined edges by 7 d, with spontaneous beating and post deflection observed, suggesting that the structure and function of the hECTs were not adversely affected by adenoviral infection. Furthermore, GFP expression was stable in hECTs throughout 3 wk of postinfection monitoring (not shown). These results demonstrate that adenoviral-mediated gene delivery in hECTs is feasible, supporting their use for basic research and possible therapeutic screening applications in genetically modified hECTs.

Figure 8.

Successful transduction of hECTs with Ad.GFP. A) Fluorescent images of representative hECTs from 3 conditions on d 9 of: hECT created with hESC-CMs infected with Ad.GFP (MOI 10) 1 d prior to hECT creation (a), hECT infected with Ad.GFP (MOI 100) on d 2 after creation (b), and noninfected hECT as negative control (right panel). Scale bars = 2 mm. B) Qualitative PCR of hECTs for the 3 conditions in A on d 15 of culture.

DISCUSSION

We examined the structure and function of hECTs created using hESC-CMs as an in vitro model of human myocardium. The principal findings were as follows: successful creation of functional hECT using an hESC-derived cell population of >90% CMs; ultrastructure and gene expression approaching that of native human myocardium; dose-response curves identifying similarities, as well as specific inconsistencies, with expected cardiac pharmacodynamics for adult myocardium; characteristic electrical functionality measured using optical mapping; contractile function measurements using two alternative systems for direct comparison; and feasibility of adenoviral-mediated gene delivery to functioning hECT. These findings support the potential of such hECTs to be developed for preclinical therapeutic screening, and help to advance the field of cardiac tissue engineering toward the promise of a reliable in vitro surrogate for human myocardium.

The understanding of cardiac muscle physiology over the past 50 yr has derived largely from muscle bath experiments using thin trabeculae and papillary muscles (43). Such studies on human tissue samples provide a valuable source of data under controlled conditions for comparison with the isometric twitch experiments performed on hECTs. As summarized in Table 2, hECTs generated isometric twitch DS on the order of 0.5 mN/mm², well below the typical range of 15 to 30 mN/mm² reported in nonfailing and failing adult myocardium (33, 35). On the other hand, they do approach the range of 1.0 to 1.5 mN/mm² reported for pediatric myocardium (34). Similarly, the TPF and RT₉₀ twitch duration times in baseline hECTs (Table 1) are about half as long as those in pediatric tissue paced at 2 Hz (34). Moreover, whereas a strong negative force-frequency relationship is a hallmark of failing adult myocardium (33), the hECT frequency dependence more closely parallels the flat or slightly decreasing trend observed in newborn myocardium (34), suggesting hECTs may represent a relatively healthy but immature human cardiac phenotype. This interpretation is further corroborated by the disparate chronotropic and inotropic responses to β-adrenergic stimulation and the lack of a distinct plateau phase in the optical action potential recordings, which both reflect cardiac immaturity. hECTs exhibit rate dependence of APD and loss of 1:1 capture at fast stimulation frequencies, lending further credence to an in vitro model system that recapitulates key electrophysiological features of cardiac tissue (48).

Table 2.

Quantitative comparison of isometric DS vs. frequency from hECT and published studies on human ventricular myocardial tissue

Study	Source	n	Age	DS (mN/mm2)	Frequency (Hz)
Current (see Table 1)	hECT	5		0.57 ± 0.07	1.0
		5		0.51 ± 0.07	2.0
Wiegerinck et al. (34)	Newborn	7	<2 wk	1.4 ± 0.3	1.0
		7		1.1 ± 0.3	2.0
	Infant	7	3–14 mo	1.2 ± 0.6	1.0
		7		1.7 ± 0.9	2.0
Rossman et al. (33)	Nonfailing adult	12	58 ± 5 yr	16.7 ± 1.6	0.5
		7		30.3 ± 4.6	2.5
	Failing adult	28	58 ± 3 yr	31.0 ± 2.8	0.5
		8		13.5 ± 1.8	2.5

Because differentiation age of hESC-CMs can affect the response to pharmacologic agents (49), strategies to enhance or accelerate maturation of hECTs will be of interest for drug discovery and toxicology applications. The demonstrated efficacy of adenoviral-mediated gene transfer at various stages of hECT culture may be particularly relevant to such efforts, since temporally regulated gene expression can be used to facilitate maturation of hESC-CMs (50). Meanwhile, as an in vitro model of immature human myocardium, hECTs may offer advantages for studying congenital heart diseases or for detecting acute cardiotoxicity of therapies for pediatric patients (51).

The PDMS culture-testing system in Fig. 1 was developed specifically for engineered cardiac tissue applications (22) to overcome key limitations of using the physiological muscle bath, namely, avoiding tissue handling and damage to maintain sterility and viability for longitudinal studies. However, the PDMS system yielded a developed stress of 0.13 ± 0.10 mN/mm² at 1 Hz, which was less than one-fourth the isometric twitch DS generated by the same 5 hECTs tested at 95% L_max using the muscle bath (0.57±0.07 mN/mm²). The lower stress observed on PDMS likely reflects a combination of mechanical unloading as the endposts bend during the nonisometric twitch, and the unknown and uncontrolled length of the resting hECT relative to L_max, both of which could introduce confounding effects when developing and screening new compounds.

Interestingly, Schaaf et al. (18) reported a DS of 0.12 mN/mm² for spontaneously beating hECT using a similar PDMS device, which is comparable to our average nonisometric DS of 0.13 mN/mm², despite differences in scaffold composition (fibrin/Matrigel vs. collagen/Matrigel), cell purity (<50 vs. >90% cardiomyocytes), and other culture conditions. Since nonmyocytes can substantially affect cardiac cells (6, 16, 25, 26, 28), generation of human ECTs using well-defined cell populations is critical for studying the effects of heterocellular interactions on hECT viability, maturation, and function under controlled conditions. Indeed, Kensah et al. (19) used an isometric bioreactor system to create bioartificial cardiac tissues from 90% highly purified ESC-derived CMs plus 10% human dermal fibroblasts in a collagen-Matrigel scaffold, and their maximum active force under control conditions was 0.78 mN, similar to our isometric DF of 0.63 mN at L_max. Optimized conditioning with ascorbic acid treatment and progressive static stretch during culture increased the generated force to 1.37 mN, equivalent to DS of 4.4 mN/mm² in their system (19). Alternatively, Zhang et al. (9) created a net-like cardiac patch with elliptical pores designed to enhance cell alignment and nutrient transport and found active force and stress amplitudes of 3 mN and 12 mN/mm², the strongest reported to date. Interestingly, both of these studies also reported a positive inotropic response to β-adrenergic stimulation with isoproterenol (9, 19), which may be related to a difference in maturity of the β-signaling pathway and its downstream effectors (involving β-adrenergic receptors, G protein, L-type calcium channel, SR, and myofilaments). Understanding the factors that affect hECT maturation and functional enhancement requires further investigation. In addition, while nonmyocytes are considered essential for the self-assembly of engineered cardiac constructs, Zhang et al. (9) showed that as hESC-CM purity increased from 48 to 90%, conduction velocity increased, while the contractile force per myocyte decreased, and other factors such as APD and maximum active force were not significantly affected by hESC-CM purity. Thus, future hECT studies may require application-specific optimization of cellular composition, i.e., the ideal cell ratio for in vitro assessment of arrhythmogenicity may differ from that for screening inotropic agents or for in vivo transplantation applications.

Although more physiological than standard cell culture conditions, and longer lasting than ex vivo myocardial tissue, hECTs remain a simplified model that lacks many components of a working heart, such as humoral, metabolic, and neurogenic factors that affect cardiac function and interventional responses. Even isolated natural heart preparations lack sensitivity to some known arrhythmogenic compounds (52), underscoring the challenges involved in designing model systems that can accurately identify promising new therapeutic agents. For some toxicity and disease-modeling applications, the long-term stability and maximum lifetime of hECTs will need to be established. Despite these challenges, a growing understanding of the similarities between engineered and natural human myocardium remains encouraging, and the species dependence of cardiac drug responses (53, 54) argues strongly in favor of the continued development of model systems based on human cell sources, including hECTs created from patient-specific induced pluripotent stem cells (iPSCs; refs. 16, 19), that promise to revolutionize the field of personalized in vitro drug screening.

In summary, hECTs allow physiological muscle function testing, and exhibit electrical and mechanical contractile properties, pharmacological responses, and structural and molecular characteristics mimicking key aspects of the newborn human heart. Such 3D engineered cardiac tissue constructs provide a more natural in vitro cell culture environment than the standard Petri dish, and promise to help fill a long-standing void in available preclinical models as a surrogate for human myocardium that could enhance the study of cardiac experimental biology, and improve the challenging process of translating cardiology research discoveries from the laboratory into effective therapies to benefit patients in the clinic.