A Bioreactor for Controlled Electrical and Mechanical Stimulation of Developing Scaffold-Free Constructs

In recent years, tissue engineers have sought insight into how micro-environmental stimuli impact tissue development and function [1,2]. One approach to probe this question is to deliver physiologic-like stimulation (e.g., mechanical and electrical) to engineered constructs using custom-built bioreactors [3–5]. Bioreactors offer great flexibility because the stimuli's key features (e.g., duty cycle, frequency, amplitude, duration, and waveform) can be controlled in an attempt to elicit the desired response. By mirroring complex stimulation patterns seen in vivo during development and healing, corresponding biophysical responses can be leveraged to accelerate construct maturation and tune the structure and biomechanical properties of engineered tissue constructs.

Mechanical stimulation can be applied using various methods, including spinner flasks (shear forces) [6], compressive pistons [6], or step-motors which allow for uniaxial strain loading, with applications in many different tissue engineering platforms (e.g., tendon, cartilage, and skeletal muscle) [7,8]. Additionally, cell culture systems, such as the Flexcell system, have been widely used to deliver dynamic mechanical stimulation to deformable substrates through the application of computer-controlled vacuum pressure [9]. Mechanical loading is traditionally applied to a scaffold-based structure in which cells are seeded or to engineered constructs after their cells have produced sufficient extracellular matrix to establish tissue structure. In tissue engineering, incorporation of mechanical stimulation has been shown to improve construct alignment, increase construct thickness and cell matrix production, and promote cell elongation and fusion [10–12].

Electrical stimulation has also been explored for engineering various musculoskeletal tissues (e.g., skeletal muscle, cardiac, and bone) [13–15]. Stimuli are applied via electric field or direct stimulation, and although the method of stimulation may vary, the primary goal is typically to mimic the neural stimulation cells experience in vivo. Electrical stimulation has been shown to increase construct alignment and cell proliferation, and promote differentiation and growth factor production [16,17]. In addition, electrical stimulation has been shown to activate different cell types in bone and generate active force production in cardiac and skeletal muscle tissues [13–16].

Due to the demonstrated individual benefits of electrical stimulation and mechanical strain, and considering that these stimuli often occur simultaneously in vivo, a number of recent studies have sought to explore potential synergistic benefits of concurrent mechanical and electrical stimulation [17–20]. In engineered skeletal muscle, combined stimulation increased the amount of fast myosin heavy chain proteins present [17] and increased construct thickness and force production, when compared to unstimulated controls [17–20]. These promising findings hint that such synergies exist and may factor prominently in future tissue engineering strategies. Hence, there is a need for further exploration into possible mechanical–electrical stimulation synergies in tissue-engineered construct development and subsequent biomechanical function [5,20,21].

Our laboratory established a scaffold-free, single-fiber tissue engineering platform, wherein cellular fibers form via directed self-assembly, with no provisional matrix present, and fiber growth channel assemblies are coupled to a modified Flexcell system to apply mechanical strain to the developing fibers to promote matrix deposition and functional maturation [22–24]. Using this to engineer single tendon fibers, we found that applied cyclic, low-amplitude uniaxial tensile strain during fiber development dramatically increased fiber longevity, tensile strength, Young's modulus, and toughness, compared to unstimulated controls [23]. We now seek to incorporate electrical stimulation to expand this platform for engineering active tissue (i.e., skeletal muscle). In this study, we integrate the ability to apply precise, custom electrical stimulation, during fiber development, into our scaffold-free single-fiber engineering platform, and perform robust calibration of mechanical and electrical stimuli, delivered independently and concurrently (Fig. 1). We then characterize the applied calibrated signals to ensure controlled, reproducible, and accurate cellular stimulation.

Flexcell^® Tissue Train^® plates (Flexcell International Corp., Burlington, NC) were modified with vertical loading pins in each of the plate's six wells, as previously described [23]. Briefly, 17-gauge stainless steel nails (Hillman, d = 1 mm), cut to ∼7-mm height, were attached to the nylon loading tabs of the plates using cyanoacrylate glue (Krazy Glue Maximum Bond Gel). To incorporate electrical stimulation, 30-gauge stainless steel wire leads were soldered to the loading pins, such that the pins that enable mechanical loading also function as stimulating electrodes. Leads were passed out through the side of modified plates and secured in place with some slack in the wire. To further stabilize the stimulating electrodes, rectangular reinforcement tabs, fashioned out of recycled silicone Flexcell membranes and biopsy punched with 1-mm through holes, were placed over stimulating electrodes and secured with cyanoacrylate glue. After the glue cured, the wells of newly modified “dual stimulation” plates were rinsed with three 5-min washes of 70% ethanol before use in experiments.

Agarose single-fiber growth channel assemblies were fabricated using previously established methods [23]. Briefly, a 2-wt % agarose solution—UltraPure™ Agarose (Invitrogen™, Thermo Fischer Scientific, Waltham, MA) dissolved in Dulbecco's Modified Eagle Medium (DMEM, VWR, Bridgeport, NJ)—is poured under a custom-fabricated aluminum 6061-T6 micromold (Precision MicroFab, Curtis Bay, MD), inverted on 2-mm risers, so that once the agarose solidifies (∼10 min), the mold can be removed, leaving a precise channel (150-mm wide, 300-mm deep, and 17.5-mm long) in the agarose. Agarose growth channels were then UV-crosslinked (Spectroline, Westbury, NY) at 1 J/cm² and stored at 4 °C, for use within 1 month of fabrication.

Sterile collagen sponge disks (4-mm diameter; Bovine Type I; Royal DSM, Exton, PA) wet with 10 μl cell culture-grade water (Mediatech, Manassas, VA) are then inserted into 4-mm holes punched at both ends of the growth channel to provide anchor points for the developing fiber, and 1-mm through-holes in each disk allow the growth channel assembly to be coupled to the dual stimulation plates via the vertical loading pins [23].

Although fibers were not created as part of this study, to engineer fibers, human plasma-derived fibronectin (Corning, MA), diluted to 0.375 μg/ml, is wicked through the collagen disks into the growth channels (20 μl/channel) and allowed to dry, making them differentially adherent. Channels are then seeded with 10 μl of a high-density cell suspension (e.g., ∼5 × 10⁶ myoblasts/ml [25] or 15 × 10⁶ fibroblasts/ml [23]) pipetted through each collagen disk, and 10 μl directly over the channel in each direction, then allowed 10 min for cellular attachment before immersion in growth medium and culture in standard conditions [23]. Electrical/mechanical stimulation can be applied as early as 3 h (myoblasts) or 18 h (fibroblasts) postseeding [25].

A two-camera digital image correlation (DIC) system (Correlated Solutions, Inc., Columbia, SC) was used to dynamically measure strain within the growth channels, as previously described (Figs. 2(a)–2(e)) [26]. To prepare for DIC, the top surface of each growth channel assembly was blotted dry and speckle-coated with an anisotropic, high-contrast pattern of black spray paint (Rust-Oleum, Troy, NY) [26]. The 2 ml of DMEM were then added to cover the bottom of each well, mimicking typical culture conditions. Dual stimulation plates with coupled growth channel assemblies were then mounted in a pneumatic Flexcell FX-4000 system equipped with 24-mm Arctangle loading posts (Flexcell International Corp., Burlington, NC) to enable application of controlled uniaxial strain via applied vacuum pressure (Fig. 1).

Mechanical stimulation was calibrated over a range of physiologically relevant strains [23]. Two cycles at each of five different strain amplitudes (0.50%, 0.70%, 0.75%, 1.0%, and 1.5% strain) were applied to six independent growth channel assemblies, at three different frequencies (0.1, 0.5, and 1 Hz). DIC cameras, rigidly mounted above the Flexcell baseplate, imaged each growth channel assembly's top surface, including the entire channel length, at 50 frames per second during cyclic mechanical strain. Noncontacting strain analysis was performed using vic-3d 2009 DIC software (Correlated Solutions, Inc.) to create an optical strain map of the growth channel assembly, and a virtual extensometer created along the length of the growth channel calculated the tensile strain experienced within the growth channel. These measured tensile strains (output strains), averaged over the six independent growth channel assemblies, were plotted against strains programed in Flexcell (input strains), and fit with a line to establish calibration curves.

To test our ability to deliver precise tensile strains, we used the aforementioned calibration curves to determine the input magnitudes that should produce strains at each of the five desired strain magnitudes. The resulting strains in six independent growth channel assemblies were measured dynamically using a DIC-based virtual extensometer, as described above.

Fabricated growth channel assemblies were coupled to dual stimulation plates, as done for mechanical characterization, and individual wells were filled with DMEM (high conductivity, ∼1.5 S/m) to cover the growth channel. Wire leads from the dual stimulation plates were connected to a 16-bit, 250 kS/s M Series multifunction bus-powered data acquisition module (NI USB-6211, National Instruments, Austin, TX), and a LabVIEW Virtual Instrument was used to deliver a biphasic square waveform of desired magnitude, frequency, and duty cycle. For electrical stimulation characterization, biphasic pulses with amplitudes of 1, 5, or 8 V were applied to six independent growth channel assemblies at four different stimulation frequencies (0.2, 1.0, 2.0, and 10.0 Hz). The resulting voltages were measured via recording electrodes held at three distinct locations along the length of the growth channel—0, 5, and 13 mm from the stimulating electrode—to assess the delivered electrical stimulation throughout the channel (Figs. 2(f) and 2(g)). Average measured voltages at each frequency were plotted against voltages input in LabVIEW, and lines of best fit were applied to generate calibration curves.

To test our ability to apply desired electrical stimulation, we applied voltages calculated using calibration curves to six independent growth channel assemblies at the four frequencies of interest, and recording electrodes placed along the length of the growth channel were used to measure resulting electrical stimulation voltages.

Having calibrated mechanical and electrical stimulations independently, we assessed the potential effects of combined mechanical and electrical stimulation on independent calibrations. Six independent growth channels were sequentially subjected to: (i) independent mechanical stimulation, (ii) independent electrical stimulation, and (iii) concurrent electrical and mechanical stimulation using specific waveforms of interest for skeletal muscle tissue engineering [23]. Growth channel assemblies were prepared for DIC strain assessment, as described above, with an additional 1 ml of DMEM pipetted over the growth channel to allow for recording electrode placement.

First, independent mechanical stimulation was characterized, in six independent growth channel assemblies, via DIC using two cycles each of five calibrated strain values (0.50%, 0.70%, 0.75%, 1.00%, and 1.50%) applied at 0.5 Hz. Next, we characterized independent electrical stimulation in the same six growth channel assemblies, using the established calibration curves to deliver a biphasic square pulse (1.5 V, 10% duty cycle) at 1.0 Hz, and measured the resulting voltages using a single recording electrode at the growth channel's center. Finally, we applied concurrent mechanical and electrical stimuli to the same six growth channels, and compared mechanical strains and stimulating voltages measured within the growth channel during concurrent stimulation to their respective independently delivered values, to determine whether combined stimulation affected independent calibrations.

A single-factor analysis of variance (ANOVA) test was used to assess differences between individual growth channel assemblies used for mechanical and electrical characterization, and two-tailed student's t-tests examined differences between averaged output strains or voltages and the desired target values. Two-factor ANOVAs were employed to examine (i) whether mechanical or electrical stimulation differed when delivered independently or concurrently, and (ii) whether there were any channel-to-channel differences. For mechanical loading, six two-factor ANOVAs were performed (one at each strain magnitude), and a single two-factor ANOVA was used for the electrical stimulation. All data are reported as mean ± SD, with a significance level of α = 0.05.

Mechanical strains measured within the growth channels were smaller than their programed input strains, and plotting measured output strains against their corresponding input strains (Fig. 3(a)) revealed that mechanical calibration was frequency-specific, with slightly different calibration curves for each cyclic strain frequency. Data for the five strain amplitudes (0.50%, 0.70%, 0.75%, 1.0%, or 1.5%) were well described by a linear calibration curve at each frequency—R² = 0.845, 0.945, and 0.941—for oscillation frequencies of 0.1, 0.5, and 1 Hz, respectively (Fig. 3(a)).

Utilizing the calibration curves, we determined the input strain values necessary for desired output strains (e.g., to mechanically strain a fiber 0.5% at 0.5 Hz, 1.11% strain is programed in the Flexcell computer). Indeed, calibration-determined input strains produced average output strains, termed “calibration strains,” that were not statistically different from their intended target strains (p > 0.05, Fig. 3(b)).

Output voltages measured along the length of the growth channels were smaller than their programed input voltages. Voltages recorded at the three different locations were not statistically different within each channel; therefore, their average was used as the channel's representative voltage measure. Measured output voltages showed a strong linear correlation (⁠$R2$ = 0.99) with input voltages (1, 3, or 8 V) and revealed that electrical calibration was not frequency-dependent for stimulation voltages of 0.2, 1.0, 2.0, and 10.0 Hz (Fig. 3(c)).

Similar to our approach for mechanical calibration, we used the calibration curve to calculate input voltages needed to produce desired output voltages (e.g., to electrically stimulate a fiber 1.5 V at any frequency, 1.9 V is programed into LabVIEW). For the two voltages of interest (1.5 V and 3.5 V), the calibration-determined input voltages produced output voltages statistically similar (p = 0.032 for 1.5 V and p = 0.022 for 3.5 V) to the desired voltages at each frequency (Fig. 3(d)).

The addition of electrical stimulation during mechanical loading had no significant effect on the independent mechanical calibration (<2% difference), at each of the six strain magnitudes (p > 0.05; Fig. 4(a)). Likewise, the electrical stimulation calibration was unaffected by the addition of cyclic mechanical strain (<1% difference; p = 0.79), and there were no observed differences between growth channels (p = 0.15; Fig. 4(b)). Additionally, calibrated strains and voltages measured for concurrent mechanical and electrical stimulation were not statistically different from their desired target values (p > 0.05) (Fig. 4). Taken together, these results demonstrate our bioreactor's ability to deliver precise mechanical and electrical stimulation, alone or in combination, to developing scaffold-free tissue-engineered fibers.

It was critical to calibrate mechanical and electrical input signals to their output delivered within growth channels, to know the exact stimuli that the cells would receive. Our modifications to Flexcell plates altered their compliance and added mass, thereby affecting their factory calibrations. Prior to our calibration, the input mechanical strain magnitudes were consistently higher than the strains delivered to the growth channels, indicating that our modifications for both mechanical and electrical stimulation affect how the applied Flexcell vacuum pressure translates to tensile strains within the growth channel. Because the mechanical strain is delivered by coupling growth channel assemblies to modified dual stimulation plates, any slight fabrication differences, in either the growth channels or the modified plates, would affect how the input strain is transferred into the growth channels, and their variances would be additive. One-way ANOVAs revealed no well-to-well difference within each plate (p > 0.05), but a slight difference between plates (p = 0.04), suggesting that difference between the individual Flexcell plates may contribute to variations in the output mechanical strain. Despite these variations, applied input strain magnitudes, determined using calibration curves, were not statistically different from their target desired strain values, illustrating the ability to precisely deliver prescribed cyclic mechanical strain. We also observed that the mechanical calibration was frequency-dependent, with increased resistance to mechanical stimulation seen at higher frequencies. Yet, despite differing absolute values at the various frequencies (0.1, 0.5, and 1.0 Hz), all calibrations were linear, reflecting the linear relationship between input strain and strain magnitude within the growth channel.

Electrical stimulation appeared to be frequency-independent, with a single line of best fit accurately representing all calibration frequencies (R² = 0.99). Additionally, electrical measurements were consistent, with no observed differences along the channel length (p = 0.94), or between channels (p =0.15). Variance in electrical stimulation was far less than that of mechanical stimulation. Therefore, we believe that many of the variations that can affect mechanical strain (e.g., fabrication of plate, growth channel, and plate/channel interface) have negligible influence on the electrical input signal delivery within the growth channel. Also, the input voltages applied using the calibration curve were statistically similar to the desired target values (1, 5, or 8 V), highlighting the bioreactor's capacity to provide controlled electrical stimulation.

Concurrent stimulation had a negligible influence on independent calibrations; causing a <2% difference in the independent mechanical calibration and <1% difference in the independent electrical calibration. Thus, the mechanical and electrical calibrations remain valid, whether applied alone or in combination, thereby allowing concurrent stimulation and other more complex electrical/mechanical loading patterns to be explored.

An important novelty of this bioreactor is its capacity to apply prescribed stimuli to constructs as they develop. There is growing interest in the mechanobiology of tissue genesis and maturation, and scaffold-free tissue engineering approaches have emerged as attractive models to study tissue development. However, scaffold-free techniques have faced technical challenges because the developing constructs need sufficient structural integrity to withstand applied mechanical stimulation. To overcome this, scaffold-free techniques typically allow the construct to develop significant structure before introducing mechanical loading. Applying stimulation after the construct has significantly formed promotes matrix remodeling, rather than tissue formation, and precludes investigation of how these stimuli influence tissue construct development [6,21,27]. Our bioreactor can apply mechanical stimulation to scaffold-free tissue constructs as they develop, promoting matrix (re)generation [22,23,27].

Similarly, there is great interest in the impact of electrical stimulation during engineered tissue development, particularly in electrically active tissues, such as muscles. Electrical stimulation is suggested to occur throughout skeletal muscle development [28,29], and it has been exogenously applied to upregulate muscle specific genes, like pax-7 [28], and increase muscle volume in aneurogenic muscles of chick embryos [29]. Since electrical stimulation contributes to muscle development and biomechanical function, its use in in vitro tissue engineering could be beneficial and ultimately translate to more functionally mimetic skeletal muscle constructs [13]. Indeed, some studies have utilized electrical stimulation in skeletal muscle tissue engineering and shown synergistic effects when delivered in combination with mechanical stimulation, as evidenced by increased matrix alignment, myotube formation, and force production [5,20,30]. However, because those studies were conducted using scaffold-based methods, or by applying the stimuli to scaffold-free constructs after they had fully formed, they could not explore the effect of concurrent stimulation on construct development. Thus, this bioreactor's ability to apply stimulation to scaffold-free tissue constructs as they develop can provide additional insight on the developmental and reparative roles of these stimuli, and their possible synergies, and may further enhance tissue engineering approaches [6,13,21,27].

The bioreactor framework allows it to be adapted to other platforms, including larger structural scales (e.g., fascicles and whole tissue), and scaffold-based constructs. Although our immediate interest is in tendon and skeletal muscle fiber engineering, this bioreactor allows exploration of mechanical and electrical stimulations in the development and maturation of other tissues, such as nerve, skin, ligament [7], smooth muscle [31], and cardiac muscle [14]. This can be achieved by changing the cell source and adjusting the stimulation patterns to better mimic those relevant to each tissue's development. Additionally, this bioreactor has potential for straightforward incorporation of chemical/environmental stimuli (e.g., growth factor addition, macromolecular crowding, and hypoxia) to allow for controlled multimodal stimulation, alone or in combination.

This relatively simple bioreactor enables precise, dynamic mechanical and electrical stimuli to be delivered, either alone or in combination, to engineered scaffold-free fibers as they develop, to provide unique insight into the influence of biophysical stimuli on the development and functional maturation of engineered musculoskeletal tissues.

Special thanks to Dr. Monica Agarwal for input on experimental design.

• This work was supported, in part, by the National Science Foundation under Grant No. 0954990 (NSF CBET-0954990) (DTC) (Funder ID: 10.13039/100000001).

No competing financial interests exist.