The objective of this study was to establish a three-dimensional (3-D) in vitro model system of cardiac muscle for electrophysiological studies. Primary neonatal rat ventricular cells containing lower or higher fractions of cardiac myocytes were cultured on polymeric scaffolds in bioreactors to form regular or enriched cardiac muscle constructs, respectively. After 1 wk, all constructs contained a peripheral tissue-like region (50–70 μm thick) in which differentiated cardiac myocytes were organized in multiple layers in a 3-D configuration. Indexes of cell size (protein/DNA) and metabolic activity (tetrazolium conversion/DNA) were similar for constructs and neonatal rat ventricles. Electrophysiological studies conducted using a linear array of extracellular electrodes showed that the peripheral region of constructs exhibited relatively homogeneous electrical properties and sustained macroscopically continuous impulse propagation on a centimeter-size scale. Electrophysiological properties of enriched constructs were superior to those of regular constructs but inferior to those of native ventricles. These results demonstrate that 3-D cardiac muscle constructs can be engineered with cardiac-specific structural and electrophysiological properties and used for in vitro impulse propagation studies.
- impulse propagation
cultured cardiac myocytes offer many advantages for developmental, physiological, and pharmacological studies of cardiac tissue because they allow for direct cell manipulation and control of environmental parameters without interference from the compensatory feedback mechanisms that exist in vivo. Compared with monolayer cultures, it has been suggested that three-dimensional (3-D) multilayered cultures of cardiac myocytes more closely resemble intact cardiac tissue with respect to cellular differentiation (8) and electrical properties (38, 39). Three-dimensional cardiac myocyte cultures could thus be used for in vitro studies of cardiac tissue development and function and, if sufficiently large and functional, for in vivo cardiac repair.
Impulse propagation studies in cultures of cardiac myocytes can improve our understanding of the electrophysiological behavior of normal and pathological cardiac tissues. Such studies are currently performed in one-dimensional cardiac strands and two-dimensional (2-D) isotropic, anisotropic, and photolithographically patterned monolayers using optical mapping techniques (9, 10, 27). Impulse propagation studies cannot be performed in 3-D myocyte aggregates (17, 30) because of their small size (100–300 μm) and isopotential nature. Other 3-D cultures of cardiac myocytes grown on microcarrier beads (1, 31), collagen fibers (1), synthetic, biodegradable polymeric templates (3,12), or in collagen gels (8) have not yet been evaluated electrophysiologically.
The goal of the present work was to establish a 3-D in vitro model system for impulse propagation studies in cardiac muscle using tissue engineering principles. This approach relies on the use of primary cells in conjunction with biodegradable polymer scaffolds (13, 18) and tissue culture bioreactors (11, 12). The polymer scaffold provides a 3-D substrate for cell attachment and tissue formation, whereas the mixing of culture medium in the bioreactor promotes mass transfer of nutrients and gases to the forming tissue. Primary neonatal rat ventricular cells were cultured on polymer scaffolds in bioreactors to form tissue constructs, which were characterized histologically, biochemically, and electrophysiologically and compared with neonatal and adult rat ventricular tissues.
MATERIALS AND METHODS
All experiments involving animals were performed according to the Institutional Committee on Animal Care of the Massachusetts Institute of Technology, which follows federal and state guidelines.
Cardiac myocyte preparation.
Primary cultures of cardiac myocytes were prepared by enzymatic digestion of ventricles obtained from neonatal (2 day old) Sprague-Dawley rats (Taconic), as previously described (44). Briefly, ventricles (n = 50, 5 litters in 3 independent studies) were incubated with 0.1% trypsin overnight and dissociated in four to five sequential steps using 0.1% collagenase. Isolated cells were resuspended in culture medium [DMEM, supplemented with 10% fetal bovine serum (FBS), 50 U/ml penicillin and 10 mM HEPES, all obtained from GIBCO-BRL].
Two experimental groups were established as follows (Fig.1 A):1) a regular group of ventricular cells isolated as described above and2) an enriched group with a higher fraction of cardiac myocytes, prepared from the regular group by centrifugation at 600 rpm for 5 min, followed by two preplatings, 75 min each (Fig. 1 A); cells that remained unattached after the second preplating were used. Cell yields were ∼6 × 106 and 5 × 106 cells/ventricle for the regular and enriched group, respectively. Cell viability was 91 ± 3%, as assessed by trypan blue exclusion.
Cells from the regular and enriched groups were cultured in monolayers at a cell density of 1.3 × 104cells/cm2 in 12-well dishes, T75 flasks, and on glass coverslips to assess spontaneous contractions and biochemical and immunohistochemical parameters, respectively. After 2 days of static culture, monolayers were placed on an orbital shaker set to 75 rpm. Medium was completely replaced on day 3 and by 50% on day 5. Spontaneous contractions were assessed by videomicroscopy, by manually counting the number of beats per minute using five randomly selected fields (0.3 × 0.4 mm2 each) per plate and six plates per experimental group, on days 3,5, and7. Cells in T75 flasks were removed after 7 days by a 5-min incubation with 0.05% trypsin-EDTA (GIBCO-BRL) and counted, and a suspension of 2 × 106 cells/ml was stored at −20°C for determination of DNA and protein contents and lactate dehydrogenase (LDH) activity per cell. Cells on glass coverslips were fixed with HistoCHOICE (Amresco) for immunohistochemical analysis.
3-D tissue culture studies.
Cells from the regular and enriched groups were cultured on polyglycolic acid (PGA) scaffolds, which are highly porous (97%) meshes of randomly entangled 13-μm fibers formed as 5 × 2-mm (diameter × thickness) disks (Fig.1 A; Ref. 13). Briefly, scaffolds were prewetted in culture medium, positioned on thin stainless steel wires using segments of silicone tubing, and fixed to a silicone stopper placed in the mouth of a spinner flask (8 scaffolds per flask) (12). Flasks were filled with 120 ml of culture medium, placed in a humidified 37°C, 5% CO2incubator with the side arm caps loosened to permit gas exchange, and mixed at 50 rpm using a magnetic stir bar. After 24 h, flasks were inoculated with cells (8 × 106 cells per scaffold). Culture medium was replaced by 100% on day 3and by 50% on day 5. Cell-polymer constructs (n = 22, from 3 independent studies) were harvested after 7 days for morphometric, histological, biochemical, and electrophysiological assessments.
To verify the analytical methods, evaluate the developmental state of cardiac myocytes in constructs, and establish baseline values for parameters studied in engineered constructs that were not readily found in the literature, two control groups were examined. Adult ventricles (n = 10) were obtained from 3- to 4-mo-old Sprague-Dawley rats following anesthesia by intramuscular injection of 65 mg/kg ketamine and 5 mg/kg xylazine (Sigma). Hearts were rapidly removed, and ventricular sections were excised from 1 mm below the atrioventricular groove to 1–2 mm above the apex. For electrophysiological studies, full-thickness pieces of the ventricular wall (∼9 × 7 mm2, 2–4 mm thick) were then prepared by making two longitudinal cuts parallel to the base-apex line. Neonatal ventricles (n = 10, from 3 litters) were obtained from 2-day-old rats following decapitation. For electrophysiological studies, full-thickness pieces of the ventricular wall (∼6 × 4 mm2, 1.5–2.5 mm thick) were prepared by bisecting the ventricle. Smaller pieces of the adult and neonatal ventricles (7–13 mg wet wt) were used for biochemical and histological assessments. The properties of neonatal and adult ventricles were compared with those of constructs without a priori assumption that the engineered tissue resembled either of the native ventricular tissues.
Histological and immunohistochemical assessments.
Cells on glass coverslips were incubated for 30 min with mouse antisarcomeric tropomyosin monoclonal antibody (clone CH1, Sigma) diluted 1:100 in PBS containing 0.5% Tween 20 and 1.5% horse serum and then for 30 min with a secondary antibody (Vectastain), diluted 1:200. Coverslips were then incubated with avidin-biotin complex reagent and 3,3′-diaminobenzidine (Sigma). Ten randomly selected fields (0.3 × 0.4 mm2 each) from six coverslips from each group were analyzed using videomicroscopy and NIH Image 1.60 software to estimate cardiac myocyte fraction as a percentage of cell area stained positively for tropomyosin.
Ventricles and 7-day constructs were fixed in 2% glutaraldehyde for 10 min, rinsed in PBS, and immersed in 10% neutral buffered Formalin (Sigma). Samples were embedded in paraffin, sectioned at 5 μm, and stained with hematoxylin and eosin (H + E) for general evaluation and Masson’s trichrome stain for collagen assessment. Immunohistochemical staining for tropomyosin was used to assess the fraction of cardiac myocytes in constructs. Sections were incubated with 1 mg/ml trypsin (Sigma) at 37°C for 15 min and 0.3% hydrogen peroxide for 30 min, blocked with horse serum for 30 min, and incubated with antisarcomeric tropomyosin as described above. A humidified chamber was used for all incubation steps. Sections were counterstained with Mayer’s hematoxylin (Sigma) and coverslipped using glycerol mounting media (Sigma). Specificity of staining for tropomyosin was confirmed by staining for sarcomeric α-actin, another myocyte-specific protein, using otherwise identical methodology. Construct macroscopic architecture was assessed from stained tissue sections using videomicroscopy and NIH Image 1.60 software.
Transmission electron microscopy.
Samples were fixed in Karnovsky’s reagent (0.1 M sodium cacodylate with 2% paraformaldehyde and 2.5% glutaraldehyde, pH = 7.4), postfixed in 2% osmium tetroxide, dehydrated in ethanol in propylene oxide, and embedded in Poly/Bed812 (Polysciences). Sections were cut at 60 nm, stained with lead citrate and uranyl acetate, and examined using a transmission electron microscope (JEOL-100CX, JEOL).
Physiological ranges of (115–130 mmHg), (48–55 mmHg), and pH (7.21–7.33) were maintained for the duration of cultivation, as measured by a blood gas analyzer (IL 1610, Instrumentation Laboratory). Glucose and lactate concentrations were measured using a glucose/lactate analyzer (2300 StatPlus, YSI). The activity of LDH in the culture media was monitored using a LDH-L reagent kit (Chiron Diagnostics). Media samples were sonicated using a Sonic Dismembrator (Vibra-Cell, Sonics and Materials), and absorbance was measured at 340 nm (Spectronic 1001+, Milton Roy) against cell-free medium. An LDH activity of 1 U/l corresponded to 3,600 cells in monolayers.
DNA and protein assays.
DNA and protein assays were performed on engineered constructs and native ventricles using modifications of previously described methods (7). Samples were homogenized in buffer (1 N ammonium hydroxide/2% Triton X-100, 0.04 ml/mg wet wt) for 1 min. For the DNA assay, homogenates were incubated at 37°C for 10 min, diluted with assay buffer (100 mM NaCl, 1 mM EDTA, 10 mM Tris, pH 7.00), and centrifuged. DNA contents of supernatants were determined using a spectrofluorometer (PTI) and calf thymus DNA as a standard (7). DNA contents measured for regular and enriched monolayers were comparable (7.1 ± 0.2 pg/cell) and consistent with published values (7).
For protein assays, the viscosity of homogenates was reduced by several passages through a 26-gauge needle. After centrifugation, protein concentration was measured in the supernatant using a Bio-Rad DC protein assay kit and a microplate spectrophotometer (MR5000, Dynatech). Regular and enriched monolayers had comparable protein contents (290 pg/cell), resulting in protein-to-DNA ratios of 41 mg/mg that were consistent with published values (29).
Metabolic activity assays.
Metabolic activities of cells within constructs and ventricular tissues were assessed by the uptake and enzymatic reduction of the tetrazolium dye 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; Sigma). Samples (2–15 mg wet wt) were rinsed with PBS and incubated with MEM (GIBCO-BRL) without phenol red and 0.5 mg/ml MTT for 4 h on an orbital shaker at 37°C and 60 rpm. Medium was replaced with an equal volume of 0.1 N HCl in absolute isopropanol and pipetted directly through the constructs to solubilize the resulting formazan crystals. After 10 min of incubation at 37°C, the absorbance was read at 570 nm, using a microplate spectrophotometer.
An electrophysiological system was custom-designed to enable stimulation and recording of unipolar extracellular potentials in constructs and ventricular tissues under controlled environmental conditions using a linear array of microelectrodes (Fig.1 B). A cylindrical Plexiglas chamber was tightly fitted inside an electrically grounded brass casing placed on a 37°C heater (VWR). The brass case distributed the heat evenly through the chamber and served as an electrostatic shield. The chamber was gassed with a prewarmed mixture of 5% CO2 in air and filled with 50 ml of culture medium (DMEM with 15 mM HEPES, 4.5 g/l glucose), which was recirculated (at 60 ml/min for constructs and 120 ml/min for ventricular tissues) using a pulseless gear pump (Cole-Parmer). Temperature and pH were maintained at 37 ± 0.1°C and 7.32 ± 0.02, respectively.
A photomicrograph of the microelectrode array is shown in Fig.1 B. All microelectrodes were made of insulated tungsten wire and had uninsulated tips with diameters of 50 ± 6 μm (Microprobe). Two electrodes for bipolar stimulation were positioned 200 μm apart and connected to a programmable cardiac stimulator (SEC-3102, Nihon Kohden). Eight recording electrodes were positioned 500 μm apart in a linear array, 1.5 to 5 mm from the stimulating site. Exact distances between electrodes were measured using a microscope and NIH 1.60 image analysis software. Shielded cables connected recording electrodes to bioelectric amplifiers (AB.601G, Nihon Kohden). A reference Ag-AgCl electrode (WPI) was placed in the medium 3.5 cm away from the microelectrode array.
Samples were placed in a tissue holder 2–3 mm under the surface of the culture medium, secured using Teflon screws, and left to equilibrate for 15 min. An XYZ mechanical micropositioner (Taurusr, WPI) was used to gradually advance the microelectrode array toward either the top surface of the construct or the epicardial surface of the ventricle, and pacing impulses were simultaneously applied (3–5 V, 1-ms pulses at a rate of 60 beats/min). The position of the array was fixed at the point where the amplitudes of the recorded responses appeared maximal, and a recording protocol was performed as follows.
Spontaneous beating, if present, was recorded for 3–5 min. After 15 min, monophasic pacing pulses (1-ms duration) were applied at a rate of 60 beats/min, starting at a pacing voltage of 0.1 V, which was then increased in 0.1-V increments until the sample was captured (i.e., until each pacing impulse was followed by a recorded tissue response). The corresponding pacing voltage, defined as the excitation threshold, represented the lowest stimulus that produced a stable propagation (for at least 1 min at a rate of 60 beats/min) over the length of the recording array. For the next 20–30 min, the sample was continuously paced at 60 beats/min using pacing amplitudes 1.5 times higher than the excitation threshold, and responses were recorded every 4–5 min for a period of 1 min. The pacing rate was then increased every 5 min by 30 beats/min, and responses at each rate were recorded for the last 40 s, similar to the protocol in Ref. 37. The maximum pacing frequency at which the sample could be captured for at least 5 min was defined as the maximum capture rate. After reaching the maximum capture rate, stimulation was stopped for 10 min and then reapplied at 30 and 60 beats/min for 5 min each to check for reproducibility of the recorded waveforms. At the end of the experiment, double and triple extrastimuli and rapid stimulation at frequencies above the maximum capture rate were applied in an attempt to induce arrhythmia.
All recorded signals were amplified and band-pass filtered between 0.3 and 1,000 Hz. The unfiltered noise level was 35 μV, peak to peak, with virtually no 60-Hz component. Analog recordings were digitized at a sampling rate of 3 kHz using a 16-bit analog-to-digital board (AT-MIO-16X, National Instruments), real-time displayed using LabView data acquisition software, and stored and analyzed using MATLAB (The Mathworks).
Activation times at each recording electrode were determined as the minima of five-point derivatives (2) of the low-pass filtered signals. The stimulus-activation time intervals at each electrode (conduction times) were plotted against the corresponding distances and fitted by linear regression. The conduction velocity of a propagated beat was calculated as the inverse slope of the best linear fit (16). The peak-to-peak (p-p) amplitudes of the responses were determined from linearly detrended signals around the activation times. Recording sites with very low or fractionated (polyphasic) activity were ignored.
For each tissue sample, p-p amplitudes at each electrode and conduction velocities were averaged from recordings made during the initial 20 min of pacing at 60 beats/min (i.e., over at least 200 beats). Conduction velocity, maximum amplitude, and average amplitude were calculated, respectively, as the averages of conduction velocities, maximum p-p amplitudes, and all p-p amplitudes from all samples within a group. The maximum and average amplitudes, respectively, represented local and spatially averaged properties of constructs or ventricles.
Data were calculated as means ± SE and analyzed using either a paired t-test or one-way ANOVA followed by Fisher’s protected least significant difference post hoc test. To determine time-dependence trends for beating rates in monolayer cultures, a univariate repeated-measures ANOVA was used. Differences were considered statistically significant whenP < 0.05. All calculations were performed using SuperANOVA III for Macintosh.
After 24 h of culture, cardiac myocytes from both the regular and enriched groups started to contract spontaneously and byday 3–4 formed synchronously contracting networks. Rates of contraction decreased significantly between culture days 3 and7 (P< 0.05) in monolayers from both groups. At day 7, enriched monolayers had significantly higher cardiac myocyte fractions and contraction rates than regular monolayers (60.5 ± 1.5 vs. 43.8 ± 0.5% of the culture area,P < 0.04, and 169 ± 8 vs. 132 ± 10 beats/min, P < 0.01), which is consistent with previous reports (22).
After 7 days of culture, cell-polymer constructs appeared discoid [∼5 × 1.3 mm (diameter × thickness); Table1]. The peripheral zone was 50–70 μm thick (Fig. 2 A) and consisted of more cell layers in the enriched than in the regular group (7 ± 1 vs. 5 ± 1 layers, respectively). Cells in this outermost zone formed a continuous, 3-D tissuelike structure by attaching to other cells, spreading along the randomly oriented PGA fibers, and forming bridges between the fibers (Fig. 2, A and C). Distinct cardiac bundles, spatially oriented groups of cells (>100 μm in size), and interstitial collagen septa were not observed. Randomly oriented cells in the peripheral zone exhibited a variety of shapes, from elongated cells spread on the polymer fibers to round unattached cells, as assessed histologically. The majority of the cells expressed the muscle-specific proteins sarcomeric tropomyosin (Fig. 2,C and D) and sarcomeric α-actin (data not shown). Immediately below the peripheral zone was a 60- to 70-μm-thick region consisting mainly of cells that did not express tropomyosin. At the construct center, cells were sparsely distributed and either elongated, expressing tropomyosin, or round, with pyknotic nuclei and acidophilic cytoplasm (Fig. 2 B).
Cross striations were present in cells in the peripheral zone of the constructs as well as in neonatal and adult ventricles, as assessed immunohistochemically (Fig. 2, D–F). The presence of subcellular elements characteristic of cardiac myocytes, including myofilaments with well-defined sarcomeres, z-lines, glycogen granules (Fig.3 A), and mitochondria (Fig. 3 B) in the outermost layer of constructs, was demonstrated by transmission electron microscopy (TEM). Cell-to-cell connections were demonstrated by the presence of desmosomes (Fig.3 C) and intercalated disks (Fig.3 D).
After 3 days, the respective numbers of viable cells present in enriched and regular constructs were 66 and 57% of those seeded attime 0, as calculated from medium LDH levels. LDH release between culture days 3 and 7 was one-third of that between days 0 and3, indicating that the cell death rate decreased with cultivation time. At 7 days, cell numbers in enriched and regular constructs were 38 and 47% of the respective numbers seeded at time 0, as determined by the DNA content of constructs. For comparison, 7-day cell monolayers from both groups contained 61 ± 6% of the initially plated cells. The number of cells seeded at time 0 (8 million per PGA disk) could be accounted for by summing cell numbers in constructs atday 7 (determined from DNA content) and in the medium over 7 days (calculated from cumulative LDH activity/construct) (Table 1), implying that no significant cell proliferation occurred during the cultivation period. Glucose consumption and lactate production rates were higher in enriched than in regular constructs (P < 0.005, Table 1), whereas the lactate-to-glucose molar ratios were similar for both construct groups (1.00 ± 0.20 and 1.30 ± 0.11, respectively).
Ventricular tissues from neonatal and adult rats had respectively six- and threefold higher DNA contents per unit wet weight (an index of cellularity) than engineered constructs from either group (P < 0.01, Fig.4 A), which is consistent with the relatively acellular appearance of the construct centers (Fig. 2 B). Relative cell size, assessed from the ratio of total protein to DNA, was comparable for cells in constructs, neonatal ventricles, and monolayers and lower than for cells in adult ventricles (P < 0.01) (Fig.4 B). This finding was consistent with the relative cross-sectional areas of cells in constructs and neonatal and adult ventricles observed histologically (Fig. 2,D–F, respectively). The MTT conversion per unit DNA (an index of metabolic activity) was similar for constructs and neonatal ventricles and was slightly higher in adult ventricles (Fig.4 C).
Spontaneous, macroscopic contractions of engineered constructs were visually observed in flasks between days 2 and 4 of cultivation, which indicated the presence of intercellular communication. At day 7 the majority of constructs and native ventricles exhibited transient spontaneous beating lasting for 1–10 contractions (Fig.5 A), which may have resulted from reentrant or triggered activity (4). Electrical stimulation resulted in impulse propagation in the peripheral cardiac tissue-like zone of the constructs. In contrast, impulses failed to propagate when the electrodes were advanced toward the central acellular region of the constructs. All 7-day constructs were electrically excitable and could be captured over a wide range of pacing frequencies (up to 270 beats/min, Fig. 5,B-D). Step increases in construct pacing frequency resulted in transient decreases in conduction velocity to steady-state values (data not shown). Rapid stimulation induced short tachyarrhythmias with rates close to the maximum capture rates in 3 of 6 enriched constructs, 2 of 6 regular constructs (Fig. 5 E), 1 of 10 adult ventricles, and 0 of 10 neonatal ventricles. A separate experiment showed that constructs remained electrically excitable for up to 4 wk of culture (data not shown).
Representative examples of impulse propagation in an enriched construct, a neonatal ventricle, and an adult ventricle are shown in Fig. 6,A–C, respectively. Propagating extracellular waveforms in constructs and native tissues showed fairly smooth, biphasic shapes with distinct downward deflections that enabled confident determination of activation times and implied nondecremental, macroscopically continuous propagation without wave collisions (34). Notches in extracellular waveforms occasionally observed in constructs (see E3 in Fig. 6 A) may have reflected asynchronous excitation in adjacent groups of cells caused by the presence of empty space, polymer fibers, and/or necrotic tissue (36). Conduction times in ventricles and constructs increased linearly with distance over 5 mm (Fig. 6 D, regression coefficients > 0.98), implying similar conduction velocities between adjacent electrodes and thus relatively homogeneous electrical properties throughout the cardiac tissue-like zone.
Conduction velocities descended in the following order: adult ventricles, neonatal ventricles, enriched constructs, and regular constructs (Table 2, Fig.6 D) (P < 0.001). Lower conduction velocities measured in neonatal than in adult ventricles were consistent with previously published values (40, 41). Conduction velocities measured in enriched constructs were ∼30 and 50% as high as those in adult and neonatal ventricles, respectively (P < 0.001) and were 27% higher than in regular constructs (P < 0.02). Excitation thresholds were higher in engineered constructs than in native ventricles (P < 0.001, Table 2) and were lower in neonatal than in adult ventricles (P < 0.01). Excitation thresholds of constructs and ventricles thus varied inversely with cellularity indexes (Table 2, Fig. 4 A).
Maximum and average amplitudes were significantly lower in constructs than in native ventricles (P < 0.0001, Table 2). Maximum amplitudes were 1.7-fold higher in enriched than in regular constructs (P < 0.002), whereas average amplitudes did not differ significantly. The range of recorded amplitudes was ∼3-fold higher in constructs than in native ventricles (data not shown). Maximum capture rates differed significantly (P < 0.001) among groups and descended in the following order: neonatal ventricles, adult ventricles, enriched constructs, and regular constructs (Table 2). The higher maximum capture rates in neonatal ventricles than in adult ventricles were consistent with the higher resting heart rates and higher tolerance to ischemia previously reported for neonatal ventricles (47).
The present study demonstrates that 3-D cardiac muscle constructs with cardiac-specific structural and electrophysiological properties can be engineered in vitro using isolated cells and biodegradable polymer scaffolds. In particular, constructs contained a peripheral cardiac tissue-like zone in which differentiated cardiac myocytes were organized in multiple layers and attached to other cells and/or polymer fibers in a 3-D configuration. Impulse propagation studies carried out using an array of extracellular microelectrodes demonstrated that the peripheral cardiac tissue-like zone of constructs sustained macroscopically continuous impulse propagation (Fig.6 A) that depended on the fraction of seeded cardiac myocytes (Table 2). Functional constructs may thus enable in vitro electrophysiological studies that may complement those currently carried out using thin ventricular slices (5, 14, 35) and monolayers of cardiac myocytes (9).
Structurally, constructs were 5 × 1.3-mm (diameter × thickness) disks and contained a 50- to 70-μm-thick outer cardiac tissue-like zone composed of cells that expressed sarcomeric tropomyosin (Fig. 2 C) and contained myofilaments, desmosomes, and intercalated disks (Fig.3 D). For comparison, a recently reported (8) heart muscle model system based on cardiac myocyte-populated 3-D collagen gels (15 mm long × 8 mm wide × 180 μm thick) contained several layers of differentiated cells at the edges and less concentrated cells centrally. The small thickness of the cardiac tissue-like zone in constructs (Fig.2 A) and collagen gels (8) can be attributed to the low survival rate of metabolically demanding cardiac myocytes located more than 50 μm from a source of gas exchange (15).
The molar ratios of lactate to glucose of 1.0–1.3 indicated aerobic cell metabolism in the constructs (21). Compared with the regular group, enriched constructs had higher glucose consumption rates (Table 1), probably due to the relatively higher fraction of myocytes. The absence of cell proliferation in constructs (Table 1) was consistent with the previous findings that neonatal ventricular cardiac myocytes lose their ability to proliferate after 2–3 days in vitro (45), whereas fibroblasts proliferate slowly in 3-D cultures (20).
Electrophysiologically, impulse propagation in constructs was studied on a macroscopic level using a linear array of extracellular electrodes (Fig. 1 B). Interelectrode distances of 500 μm were selected on the basis of previously reported in vivo and ex vivo epicardial mapping studies (6, 46, 48). Bipolar point stimulation and unipolar recording (16, 25) in the custom-designed test chamber (Fig. 1 B) did not adversely affect samples with respect to their electrical properties (waveform shapes were stable) or structure (no apparent tissue damage was observed histologically). Automated data analysis was facilitated by the high average signal-to-noise ratios (of ∼10 and 470 for constructs and native ventricles, respectively). Whereas 1- to 5-V amplitude, 1-ms duration electrical pulses were sufficient to induce impulse propagation in slices of ventricles and in the peripheral zone of 7-day constructs, it was difficult to overdrive 7-day confluent monolayers of neonatal cardiac myocytes even when using stimuli of twice this amplitude and duration. In addition, impulse propagation in monolayers could not be assessed using extracellular electrodes because of fractionation and low amplitudes of recorded waveforms. These findings may be due to 3-D electrotonic interactions between cells (9) and relatively high cell density around the stimulating and recording electrodes in 3-D constructs compared with 2-D monolayers.
The inferior electrophysiological properties of constructs compared with native ventricles (Table 2) can be attributed to differences in their macroscopic tissue architecture. In particular, the relatively high excitation thresholds (24) and low response amplitudes were associated with low construct cellularity (Fig.4 A). Low maximum capture rates and conduction velocities in constructs probably resulted from decreased cell coupling, the presence of intercellular clefts, and geometric current-to-load mismatches (due to tissue discontinuities) (9, 26). Other mechanisms that could contribute to inferior construct electrophysiological properties include cell depolarization, reduced excitability, and slower repolarization resulting from injury during isolation and/or cultivation (32, 39). Intracellular recordings would be necessary to test the proposed mechanisms.
Compared with enriched constructs, lower conduction velocities, maximum capture rates, and amplitudes in regular constructs probably resulted from 1) the higher fraction of noncardiomyocytic cells, which would be expected to form high-resistance junctions with cardiac myocytes (28) and act as passive current sinks (9), and 2) the thinner cardiac tissue-like zone (Table 1). Lower maximum capture rates in the regular than enriched constructs could also be due to the relatively longer duration of cellular action potentials (as previously observed in fibrotic compared with normal cardiac tissue; Ref. 42).
Neonatal and adult ventricular tissues did not exhibit spontaneous beating ex vivo in a previous (39) or the present study. In contrast, enzymatically isolated ventricular cardiac myocytes cultured in monolayers are known to revert to a less differentiated phenotype, depolarize, and regain spontaneous contractile activity for as yet unknown reasons (39). In the present study, visible spontaneous contractions in constructs ceased after 4 days of cultivation. This finding might be attributed to gradual depolarization and decoupling of cardiac myocytes due to injury during cultivation. However, it is more likely that the cultivation of cardiac myocytes on 3-D biomaterial scaffolds in tissue culture bioreactors (Fig.1 A) promoted differentiated cellular phenotype and function. In support of this hypothesis, Sperelakis (38) showed that 3-D aggregates composed of electrically differentiated cardiac myocytes did not contract spontaneously but responded to electrical stimulation.
The aim of the present study was to demonstrate basic cardiac-specific features in constructs and to evaluate construct structure and electrophysiological properties on a macroscopic (tissue) level, rather than on a cellular level. In ongoing work, we are expanding our electrophysiological studies to include whole cell clamp and sharp microelectrode intracellular recordings and assessment of the spatial distribution of the gap junctional protein connexin 43 (23). We are also attempting to culture constructs with a thicker cardiac tissue-like zone by direct perfusion of constructs during cultivation (to improve mass transfer) and by coculturing cardiac myocytes with microvascular endothelial cells (as a first step toward inducing vascularization).
In conclusion, cardiac-specific features of engineered cardiac muscle constructs were demonstrated structurally and electrophysiologically and were related to the cellular composition of constructs. The 3-D multilayer structure in conjunction with macroscopic impulse propagation in engineered constructs can offer advantages for in vitro studies of cardiac muscle. In addition, structurally and functionally improved 3-D engineered cardiac muscle constructs could be eventually applied in vivo. To date, attempts to regenerate cardiac tissue have involved the injection of different muscle cell types (33, 43) or small tissue fragments (19) into the heart. Implantation of cardiac muscle constructs with a defined shape instead of isolated cells could potentially improve the efficiency and localization of tissue repair.
N. Bursac and M. Papadaki contributed equally to this study.
Address for reprint requests and other correspondence: L. E. Freed, Massachusetts Institute of Technology, Div. of Health Science and Technology, MIT, Bldg. E25–342, Cambridge, MA 02139 (E-mail:).
We thank R. Langer for advice, R. Padera for help with animal surgery, H. Shing for carrying out the transmission electron microscopy, Y. Lee for help establishing the electrophysiological recording system, and J. Merok, H. Cho, and P. Gupta for help with biochemical assays.
This work was supported by National Aeronautics and Space Administration Grant NAG9-836.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
- Copyright © 1999 the American Physiological Society