Wirelessly Powered 3D Printed Hierarchical Biohybrid Robots with Multiscale Mechanical Properties

Abstract

The integration of flexible and stretchable electronics into biohybrid soft robotics can spur the development of new approaches to fabricate biohybrid soft machines, thus enabling a wide variety of innovative applications. Inspired by flexible and stretchable wireless-based bioelectronic devices, we have developed untethered biohybrid soft robots that can execute swimming motions, which are remotely controllable by the wireless transmission of electrical power into a cell simulator. To this end, wirelessly-powered, stretchable, and lightweight cell stimulators were designed to be integrated into muscle bodies without impeding the robots’ underwater swimming abilities. The cell stimulators function by generating controlled monophasic pulses of up to ∼9 V in biological environments. By differentiating induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) directly on the cell stimulators using an accordion-inspired, three-dimensional (3D) printing construct, we have replicated the native myofiber architecture with comparable robustness and enhanced contractibility. Wirelessly modulated electrical frequencies enabled us to control the speed and direction of the biohybrid soft robots. A maximum locomotion speed of ∼580 μm/s was achieved in robots possessing a large body size by adjusting the pacing frequency. This innovative approach will provide a platform for building untethered and biohybrid systems for various biomedical applications.

Graphical Abstract

Wirelessly-powered, biohybrid soft robots that are inspired by stretchable bioelectronic devices and accordion-like scaffold designs are demonstrated. The integration of wireless cell stimulators into human cardiac-tissue-based muscle actuators realizes electrically controlled movement of robots without requiring batteries and wires. The long penetration depth of the stimulation signal and inherent system extensibility may advance the development of biohybrid robotic systems that are controllable in both in vivo and in vitro environments.

1. Introduction

Bioinspired soft robots embellished with adaptable, pliable, and smart materials, similar to living biological components, have become a promising research subject portending a modern robotic revolution ^[1–3]. Relative to conventional robots that are rigid, these devices offer superior flexibility and multiple degrees of freedom, enabling them to mimic the movement and response output of adaptive living organisms that are intrinsically intelligent against environmental cues. Such capabilities allow morphological adaptations in response to complex changes that are caused by environmental unpredictability. However, bioinspired soft robots still cannot fully replicate the peculiarity of our living systems, which provide autonomous energy generation, highly efficient actuation, high power-to-weight ratios, self-healing properties, behavioral flexibility, complex controllability, and advanced learning potential. Recently, attention has been drawn to the possibility of fusing living organisms and artificial systems to design more organic-like robotics that can mimic lifelike movements ^[4–10]. When incorporated with bio-hybridization strategies, such soft robots are called biohybrid soft robots; they are exclusively powered by the contractile forces of living muscle tissues and can be an alternative to compliant, material-based, bioinspired soft robots due to their inherent organic-originating advantages. Such innovative strategies have substantially contributed to the development of biohybrid soft robots, e.g., fish-like swimming robots ^{[4–6, 11]}, tendon-like interfaces ^[7], locomotive robots ^{[8, 12–14]}, transformable robots ^[15], and morpho butterfly wings ^[16–17], that incorporate muscle tissues-based actuation systems, such as cardiomyocytes, skeletal muscle cells, and insect muscular tissues. The maneuvering of these biohybrid soft robots is controlled using either a pulsed electrical signal transmitted via an electrode ^{[4, 7–8]} or an optical stimulation transmitted via optogenetically-modified cells, which can depolarize the muscle tissues ^{[5, 18–19]} and photosensitive hydrogels to deform the robots’ structures ^[15]. At a cellular level, relative to pulsed electrical signal transmission, optical stimulation provides a safe route for maneuvering, and it can be applied in such a manner that the remote triggering source, such as a laser or light from a light-emitting diode, is kept on the biohybrid soft robots. Although optical stimulation could successfully steer and control the motion of biohybrid soft robots, the penetration depth of light might constrain the usability of such light-controlled biohybrid systems in biomedical applications. The penetration of ultra-violet (UV) or visible light is limited to a depth of a few hundred micrometers, while that of infrared light is limited to a depth of a few millimeters. Limitations are the result of light scattering, which occurs in heterogeneous biological tissue ^[20–21], and thus would require the insertion of a light source, such as a fiber probe, into the body tissues for signal stimulation.

Therefore, as an alternative, advanced electrical signal transmission technology has recently been explored from the perspective of integrating flexible microelectrodes into soft and deformable engineered scaffolds to locally stimulate muscle actuators that can control biohybrid soft robots ^{[6, 22]}. Furthermore, the flexible microelectrodes could provide good cytocompatibility and high durability, while also being simply fabricated. However, the local electrical stimulation of biohybrid soft robots is still only achievable by the use of embedded microelectrodes that require tethered connections to external power supplies. To automatize the biohybrid soft robots for use in various practical applications, their driving components, i.e., power supply sources, must be fully integrated and embedded within their structure without rigid boundaries. Thus far, untethered systems typically rely on loading bulky batteries within the robots’ structures. However, the heaviness and rigidity of batteries render them difficult to integrate into relatively soft muscle actuators and can inhibit actuation performance; thus, they cannot be implemented into biohybrid systems. Therefore, a method to feed electricity to biohybrid soft robots needs to be developed in a wireless, lightweight, stretchable, and biocompatible manner to conserve structural softness.

Herein, we present an untethered biohybrid soft robot comprised of human cardiac tissue-based muscle actuators and wirelessly-powered bioelectronics that allows for electrically controlled movement without the need for batteries and wires (Figure 1A). Stretchable and flexible wirelessly-powered bioelectronics have been suggested as promising platforms for electronic skins ^[23–25], optogenetic neuromodulators ^[26–27], smart contact lenses ^[28], and electroactive cell stimulators ^[29–30]. The implantable device utility of such bioelectronics offers an innovative strategy for powering biohybrid soft robotic systems by averting the loading of bulky batteries and wires. Existing bioelectronic devices, however, have only been implemented in a wearable or implantable manner, onto biological tissues such as the skin, eyes, or organs. For example, Matsuhia et al. reported an on-skin wireless stretchable system capable of operating at a high frequency of 13.56 MHz in air to drive an electrochromic display pixel ^[23]. Park et al. demonstrated an on-eye wireless glucose sensor by incorporating wireless power transfer circuits and display pixels into a polymer lens, which caused it to operate in the presence of tear fluid ^[28]. Koo and Choi et als. also implanted wireless, bioresorbable electronic stimulators on sciatic nerves by encapsulating them with polymers (poly (lactic-co-glycolic acid) or polyurethane) that, driven in the presence of body fluid, could facilitate neuromuscular regeneration ^[29–30]. Based on the specific application, bioelectronics should be designed uniquely and optimized to generate desired functions at specific conditions; i.e., on the skin, on the eyes, on/under the tissues, or in water etc. Until now, there is no wirelessly-powered bioelectronic device that can be integrated directly on tissue engineered biohybrid actuators. The direct and seamless hybridization of bioelectronics with tissue engineered, muscle-based biohybrid actuators, through the strategy herein elucidated, will allow the full hybridization of bioelectronics with living tissues and enable the creation of flexible and stretchable wireless bioelectronics. These can further be enhanced for their functionality in numerous applications, including targeted sensing, delivery, and imaging under confined and enclosed spaces, such as inside the body, through an integrated wireless circuit design.

Figure 1.

Concept and system design of the wirelessly controllable biohybrid soft robots. (A) Schematic illustration of the wirelessly controllable biohybrid soft robot. (B) Design of the muscle fiber alignment, which yields a soft robot capable of upward contraction. (C) Schematic illustration of the stretchable wireless device. (D) Schematic of the wireless power transfer and control system with the excitation coil (L_EC) and receiver coils (L_RC) with associated series resistances (R_EC and R_RC), an induced voltage source (V_RC), tank capacitors (C_TANK), load resistor R_CELLs (representing cells resistance) and stimulation voltage V_CELLs, and radio frequency (RF) diode (D). (E) Working frequency of the wireless device. The plot shows the self-resonant frequency (SRF) of the receiving antenna coil. (F) Top view of the numerically calculated electric potential contour plot for the wireless device.

In this study, we have demonstrated the first seamless hybridization of stretchable wireless bioelectronics with a cell stimulation circuit into biohybrid soft robotics through an embodied device architecture that differentiates muscle fibers directly on the cell stimulator. In this robotic design, voltages were induced using a magnetic coupling between the wireless device and the excitation circuit to generate pulses that could electrically stimulate the muscle actuators directly created on the device. The transmission of voltage signal was not attenuated through the tissue and the penetration depth reached up to ∼130 mm. We designed the wireless bioelectronics to generate high output voltages (of ∼up to 9 V) and a pulse type of signal (from 0.5Hz to 3 Hz) in cell culture media. This was done using a full-wave electromagnetic simulator, considering the high permittivity environment of the surrounding aqueous media. To enable human cardiac stem cells to culture with high cell viability and sufficient differentiation without the elution of any cytotoxic compounds during electrical stimulation, cytocompatible electrode materials (silver/urethane and acrylate-based dielectric) were selected. Furthermore, to provide a strong contractile performance of muscle actuators, we designed scaffolds with multiscale mechanical properties by hierarchically aligning two mechanically different hydrogel micropatterns into an accordion-inspired structure. The 3D-printed, multilayered scaffolds enabled the replication of thick and dense native cardiac myofiber structures with strong contractibility on the wireless cell stimulators, which allowed for the efficient driving of soft robots. Controlling the speed and direction of the locomotive biohybrid soft robots has been attainable through the transmission of wirelessly modulated electrical frequencies into cell simulators. A comparison of our wireless technique with other remote-control methods that drive biohybrid or bioinspired soft robots is shown in Table S1 (Supporting Information). While the magnetic control system that is implemented in magnetically-driven bioinspired soft robots allows for the more precise maneuverability and transformability of robots ^[31–34], the inherent system extensibility through the integrated wireless circuit design of our wireless system, may advance the development of locomotive bioelectronic and biohybrid robotic systems that are controllable in both in vivo and in vitro environments. This is abetted by the long penetration depth of the stimulation signal through the tissue, which is over 10 cm, compared to optical control systems.

2. Results and Discussion

2.1. Design of wireless powered biohybrid robots with multiscale mechanical properties for achieving strong actuation behavior

The muscle actuator should provide enough contractile force and actuation performance to synchronously deform the wirelessly-powered bioelectronics. Engineered scaffolds were designed by drawing inspiration from accordion-like structures and our previous stingray-inspired biohybrid soft robots ^[6]. These are composed of parallel rigid hydrogels micropatterned on soft and flexible hydrogel substrates, which guide the strong unidirectional deflection while retaining the mechanical stability of the whole structure, as shown in Figure S1A (Supporting Information). Hierarchical structures with multiscale mechanical properties enhance the stretchability and foldability of nano or microstructures ^[35–38]. To evaluate this hypothesis, we fabricated actuators composed of a rigid poly(ethylene glycol) (PEG) hydrogel pattern and a soft gelatin methacryloyl (GelMA) hydrogel film, which showed a large unidirectional structural deformation of ∼12% (Figure S1D and Movie S1, Supporting Information). Maturation and organization of muscle tissue on the scaffolds is another major factor that enhances actuation performance. The neonatal rat cardiomyocyte-based biohybrid systems have higher contractile specific stresses (1–4 kPa) compared with skeletal muscle-based systems (0.1–1.0 kPa) ^{[8–9, 39–41]}. Although the contractile performance of iPSC-CMs is weaker than that of neonatal rat cardiomyocytes because of their immature nature ^[42], they can generate consistent contractile force for more than one month compared with neonatal rat cardiomyocytes and other cell sources ^[43]. Additionally, the use of iPSC-CMs can avert the rise of ethical issues related to animal sacrifice, as these cells provide an abundant source of cardiac muscle tissue for tissue engineering, drug screening, and regenerative medicine applications. Therefore, iPSC-CMs can be a better cell source for the scale-up and manufacturing of muscle-based biohybrid soft robots, and they are selected as a model muscle cell source for our biohybrid system. To facilitate better maturation and organization of the iPSC-CMs, we adopted a carbon nanotube (CNT)-incorporated GelMA (CNT/GelMA) hydrogel that has a modulus of ∼36 kPa and nanofibrous architecture which mimics the native extracellular matrix (ECM) of cardiac tissue, as demonstrated in our previous studies for engineering cardiac constructs ^{[6, 22, 44–47]}. The electrically conductive CNT networks embedded in the hydrogels might facilitate electric signal propagation under the cardiac tissue, initiated by a wireless device. As a rigid component, the PEG/GelMA hydrogel with a high modulus of ∼650 kPa was selected due to its rigidity. Furthermore, the PEG/GelMA hydrogel has higher stiffness compared to other photo-crosslinkable hydrogels ^[48–49]. Even at such a high modulus, the PEG/GelMA can still retain enough deformability under both cardiac tissue contraction and relaxation while retaining their mechanical and structural stability ^[6]. Thus, the untethered biohybrid soft robot system comprises four parts: a thin CNT/GelMA hydrogel layer to embody the wireless device; the stretchable wirelessly-powered cell stimulator to generate a uniform pulsed direct current (DC) voltage of 1–9 V; a multilayered scaffold of 3D-printed CNT/GelMA hydrogel patterns to induce the alignment of muscle cells and PEG/GelMA patterns as a rigid structural frame; and contractile iPSC-derived muscle tissue.

In the hierarchical, multilayered scaffold, the arrangements of the CNT/GelMA and PEG/GelMA micropatterns that have different mechanical properties are key criteria to designing the aligned muscle fibers and determining the kinematic motions of the soft robots. Therefore, the 3D printing technique, which can precisely fabricate complex and multilayered scaffolds in desired locations with multiple materials, was used, instead of other microfabrication techniques. Employing 3D multi-materials printing technology is especially beneficial in this approach, compared to other anisotropic cell alignment and scaffold fabrication techniques (such as a microcontact printing ^[50–52], uniaxial stretching of a substrate ^[53–55], and photolithography ^{[6, 56]}), as it readily allows the optimization of the hydrogel micropattern geometry to fulfill the cellular alignment/environment requirements and to couple the scaffold with the wireless device. In order to create a finite-element simulation of deflections of soft robots (Figure S2, Supporting Information), soft CNT/GelMA hydrogels were printed with a pattern perpendicular to the orientation of the hard PEG/GelMA hydrogel pattern. The sarcomeric longitudinal muscle layout was determined to facilitate electrical conduction between neighboring muscle cells (Figure 1B). The muscle fibers of iPSC-CMs along the CNT/GelMA hydrogel lines were estimated to guide the actuation dynamics parallel to the CNT/GelMA hydrogel lines, thus yielding a soft robot capable of performing upward contraction (Figure S2 and Movie S2, Supporting Information). Furthermore, the multilayered scaffold that was fabricated by repeating the 3D printing of the hydrogel micropatterns could produce thick and dense 3D cardiac tissue constructs with strong contraction behavior.

2.2. Design of a stretchable and biocompatible wireless powering device

The wireless device comprises a polyurethane film (PU), a receiving antenna coil, a dielectric layer, top electrodes, and a diode (Figure 1C and Figure S3, Supporting Information), fabricated on a stretchable and biocompatible 37-μm-thick PU (Figure S4, Supporting Information) via screen printing with stretchable silver- and acrylate-based dielectric inks. The dielectric layer isolates the antenna coil from the electrodes, thereby preventing short circuits and avoiding direct contact with the cardiac tissues. The voltage rectifier circuit, composed of the diode with a capacitance of 0.35 pF and a tank capacitor (C_TANK), converts the radio frequency (RF) signal (V_RC), which is induced in the receiving coil through the mutual inductance (M), into a DC voltage (Figure 1D). This DC voltage is applied to cardiac tissues using top electrodes for stimulating the cardiomyocytes, represented by a resistor (R_CELLs) in the electrical model. To ensure that the device functions properly in an aqueous environment, the receiving antenna coil (L_RC) was designed and simulated using the full-wave electromagnetic simulator to define the self-resonant frequency (SRF) while considering the relative permittivity, ε_r = 80, of the cell culture medium (Figure 1E and Figures S5 and S6, Supporting Information). The SRF at 85 MHz was well above the excitation frequency at 13.56 MHz, and the value of the inductance was 495 nH at the excitation frequency. The voltage distribution across the excitation area is shown in Figure 1F. The center of the wireless device between the cathode and the anode has a denser electric potential. Thus, the cells between these electrodes can be activated simultaneously, and then the electrical signal can be propagated from cells close to electrodes to cells away from the electrodes, toward the outer part of the wireless device via pulse conduction.

We fabricated stretchable, thin, lightweight, and biocompatible wirelessly powering devices by a screen-printing technique (Figure 2A, Figures S7–S9 and Movie S3, Supporting Information). The devices have two different sizes with 20 × 20 mm (trace width: 200 μm, weight: ∼28 mg, thickness: ∼70 μm, density: ∼1,000 Kg/m³, Figure 2A) and 10 × 10 mm (trace width: 50 μm, weight: ∼7 mg, thickness: ∼70 μm, density: ∼1,000 Kg/m³, Fig. S7 A) dimensions. The dielectric device was double-layered, ensuring complete electrical isolation between the antenna coil and the cell stimulation electrodes (Figure S7 B, Supporting Information). These devices have Young’s moduli of ∼6.4 MPa for the 20 × 20 mm device and ∼4.5 MPa for the 10 × 10 mm (Figure S8 B, Supporting Information). The difference in the Young’s modulus between the devices might be caused by the difference in their electrode geometry (line width) ^[57]. Although the Young’s modulus of these devices is high compared with that of cardiac tissues (∼0.05–0.195 MPa) ^[58], it still allows embodied integration in the soft robots and generation of buoyant robots compared to thicker and heavier printed circuit board (PCB)-based devices (thickness: ∼1.5 mm, Young’s modulus: ∼10,000 MPa, density: ∼2,200 Kg/m³) ^[59]. The 10 × 10 mm devices exhibited higher resistances, which might influence their output voltage performances (Figure S8 A, Supporting Information). Therefore, we mainly focused on fabricating soft robots using 20 × 20 mm wireless devices. Figures 2B–D highlight the electrical performance characteristics of the 20 × 20 mm wireless devices after connecting the diode. An electrical signal was transmitted through the excitation circuit based on a current-mode class-D (CMCD) power amplifier (PA) (Figures S10–S13, Supporting Information). The PA injects a sine wave current (carrier) at 13.56 MHz into the excitation coil. This carrier is then modulated by a square wave at 0.5–3.0 Hz with a pulse width of 50 ms using transistor M₃ (Figure S11 A, Supporting Information). The voltage induced in the wireless device is proportional to the excitation current (Figure S13 C, Supporting Information). The voltage rectifier generates a DC voltage, which follows the envelope of the modulated signal. The excitation coil current can be controlled by adjusting the PA voltage (V_sup).

Figure 2.

Fabrication and electrical characterization of the wireless device. (A) Photo images from the screen-printed wireless devices on a polyurethane film with a thickness of ∼40 μm and a dimension of 20 × 20 mm. (B) Comparison of the output voltage (peak-to-peak voltage) at different amplifier voltages (V_sup) in air and medium using a single-turn excitation coil (n = 3 devices). Error bars represent the standard error of the mean. (C) Representative output signals from the wireless device at different modulation frequencies in air and medium (V_sup: 3.2 V, pulse width: 50 ms, single-turn excitation coil). (D) Contour plots of output voltage (peak-to-peak voltage) distributions on the single-turn excitation coil when the wireless device is placed at Z_sd (stand-off distance from the excitation coil surface) = 1 and 5 mm above the excitation coil surface in medium (V_sup: 3.2 V, pulse width: 50 ms, 2 Hz). (E) Oxidation/reduction process for methyl viologen (MV) with a redox potential of −0.704 V (versus saturated calomel reference electrode (SCE)), and color change in an MV 50 mM solution on the wireless device under wireless electrical stimulation (V_sup: 3.2 V, pulse width: 50 ms, 2 Hz, single-turn excitation coil). (F) Temporal trace of output voltage (peak-to-peak voltage) from the wireless device with a CNT/GelMA layer and C2C12 cells in medium (V_sup: 3.2 V, pulse width: 50 ms, 2 Hz, single-turn excitation coil). The microscopic images show live/dead staining images of C2C12 cells on the device after 14 days of the electrical test.

We designed two planar resonant antenna (excitation) coils: one using a single-turn coil (coil radius: ∼85 mm, Figure 2B) and the other using a multi-turn coil (coil internal radius: ∼55 mm, Figure S14 A, Supporting Information). The 20 × 20 mm wireless devices generated an output voltage (peak-to-peak voltage) of up to ∼5 V by the single-turn coil and ∼9 V by the multi-turn coil in cell culture media (frequency: 2 Hz, pulse width: 50 ms) (Figure 2B and Figures S14 A, D, E, Supporting Information). For both coils, the output voltage depends on the V_sup and the standoff distance (Z_sd) between the excitation coil and the wireless device. The multi-turn excitation coil markedly enhanced the output voltage. Figure 2C illustrates the monophasic output voltage obtained from the wireless device at different frequencies in air and medium (V_sup: 3.2 V, pulse width: 50 ms, single-turn coil); this output voltage induces the electrical impulses that stimulate the target cardiac tissues. Our system can generate a monophasic output voltage at up to 15 Hz with a pulse width of 50 ms (Figure S14 F, Supporting Information). Even 30% uniaxial strain did not alter the output voltage performance from the wireless device (Figure S8C, Supporting Information). The in-plane distribution of output voltage above the excitation coil is shown in Figure 2D. The output voltage (peak-to-peak voltage) decreases toward the center of the excitation coil because of the increased magnetic field deviation; however, even at a Z_sd of 5 mm, the minimum output peak voltages were ∼2 V, which are higher than the threshold for cardiac cell stimulation through an embedded microelectrode ^[6], and can be generated in culture media. The PA voltage should be adjusted to avoid overstimulation of cells at the near edge of the excitation coil. Future work should aim at optimizing the excitation coil geometry to generate a more uniform magnetic field distribution ^[60–63]. Other fundamental limitations that relate to the miniaturization of the wireless receiver coils and the dimensions of the transmitter coil for wireless systems can also be overcome by utilizing an electromagnetic, mid-field wireless powering technology that enables power transfers to deep-tissue implants using a mm-size receiver coil ^[64], and by optimizing the geometry of transmitter coils ^[65], respectively. Notably, the transmission of voltage signal was not attenuated through tissue, and the penetration depth reached up to ∼130 mm (Figure S15, Supporting Information). Therefore, our wireless systems could be operated inside the body tissues for use in various biomedical applications. The 10 × 10 mm devices also exhibited similar electrical performances (Figure S16, Supporting Information).

The successful electrochemical reduction of N,N’-dimethyl-4,4’-bipyridinium (MV²⁺) ^[66] further manifested a potential in the wireless cell stimulator. When the MV²⁺ solution (50 mM) was dropped onto the wireless device, the color above the anode immediately turned deep violet because of the transmitted voltage (V_sup: 3.2 V, frequency: 2 Hz, pulse width: 50 ms, single-turn excitation coil) (Figure 2E and Movie S4, Supporting Information). To evaluate the durability and cytotoxicity of the wireless device, C2C12 myoblast cells were seeded on the device with a 100-μm-thick CNT/GelMA layer. Furthermore, the temporal output (V_sup: 3.2 V, frequency: 2 Hz, pulse width: 50 ms, single-turn excitation coil) and cell viability were monitored (Figure 2F, Supporting Information). The output voltage remained constant for over 14 days, and the C2C12 cells showed good viability (98.0% ± 0.5%), even after 14-day-long voltage pulse trains were used for electrical stimulation. Although we confirmed the cytocompatibility of the wireless device with C2C12 cells, there was no observed decreased viability of iPSC-CMs while they were cultured on the device or during the stimulation period. Consequently, during long-term electrical stimulation and under the stimulation conditions, the wireless device barely experienced any wear, and no cytotoxic compounds were released from the stretchable silver electrode or its dielectric layers.

2.3. Integration of wireless device into the biohybrid soft robots

To fabricate a biohybrid soft robot with an embedded 20 × 20 mm wireless device (body size: ∼35 × 33 mm, total thickness: ∼150 μm (wireless device-embedded CNT/GelMA layer: ∼100 μm + muscle and 3D printed 6-layered hydrogel scaffold: ∼50 μm) (Figure 3A and Figure S17, Supporting Information), and total weight: ∼160 mg (after wiping the surface dry to remove unbound water)), we first photo-crosslinked the CNT/GelMA hydrogel that was directly dropped onto the wireless device using a shadow mask to encapsulate the device with the hydrogel layer (Figures S18 and S19, Supporting Information). An RF diode, which is part of the voltage rectifier, was covered by polydimethylsiloxane (PDMS, density: ∼947 Kg/m³) to isolate it from the media and generate buoyancy of the soft robot. It was then connected between the antenna coil and the electrodes before embedding. The thickness of the CNT/GelMA hydrogel layer was decided to be ∼100 μm, considering the finite-element simulation results and wireless device thickness of ∼70 μm. In the finite-element simulations, the bending angle was markedly decreased as the thickness of the CNT/GelMA hydrogel layer increased to 100 μm; the deflection for the constructs with and without the wireless device was reduced to ∼47% (Figure 3B and Figure S2, Supporting Information). However, the CNT/GelMA layer with the PU substrate, which was 100-μm thick, was estimated to be still bendable with a maximum deflection of ∼2.1 mm.

Figure 3.

Fabrication of the biohybrid soft robots with the embedded wireless device and wireless electrical pacing of the muscle actuator. (A) Three-dimensional (3D) printing design of the CNT/GelMA and PEG/GelMA hydrogel patterns on the soft robot with the embedded wireless device and schematic illustration of the soft robot with a scaffold composed of six-layered hydrogel pattern. (B) Global deflection for the soft robot with the wireless device (CNT/GelMA layer thickness: 100 μm). (C) A fluorescent image under an ultraviolet (UV) light from the 3D printed construct, which was printed using hydrogel inks containing fluorescent materials. The CNT/GelMA and PEG/GelMA patterns glow red and yellow, respectively. (D) A photo image of the soft robot with a printed pattern. (E) Microscopic image from soft robot of the photo of (D). (F), (G) Confocal images from the immunostained human-induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) on the soft robot of (D). (H) Schematic illustration of the soft robot with interconnected muscle fiber alignment and images from z-stack projections of (F). (I) Confocal images from the iPSC-CMs on the soft robot with a two-layered hydrogel pattern. The inset shows a magnified image of sarcomere structures. All images were taken after 6 days of culture. (J) Traces of calcium transients in iPSC-CMs on the wireless device on day 6. I: Microscopic image at an area highlighted using broken white lines, which correspond to the measurement area for calcium imaging. Activation maps of wirelessly paced activation at 0.5 Hz (II), 1.0 Hz (III), 2.0 Hz (IV), and 3.0 Hz (V) from the muscle actuator (V_sup: 3.2 V, pulse width: 50 ms, single-turn excitation coil). (K) Excitation threshold PA voltage at 0.5, 1.0, and 2.0 Hz (pulse width: 50 ms, single-turn excitation coil) (n = 3 devices, *p < 0.05, Tukey’s test). Error bars represent the standard error of the mean. (L) Representative trace demonstrating the wireless electrical pacing effect on the contractile activity (beating rate) of the iPSC-CMs on the wireless device (V_sup: 3.2 V, pulse width: 50 ms, single-turn excitation coil).

A 3D printing approach was implemented to construct the multilayered hydrogel scaffold with multiscale mechanical properties onto the wireless powering device. To develop the 3D printing of the CNT/GelMA and PEG/GelMA hydrogel with precise narrow lines and good printability, we optimized the printing parameters, such as speed and line width, to provide good printability and improve cell–cell electrical coupling and maturation of muscle tissue, based on our previous reports ^[6] (Figure S20, Supporting Information). Under optimized conditions, the fine CNT/GelMA and PEG/GelMA patterns, with a width of ∼60–100 μm and height of ∼30–50 μm, were printable (Figure 3C and Figure S21, Supporting Information) and acted as the ECM-like scaffolds for alignment, maturation, and long-term cultivation of iPSC-CMs. In the latter, the alignment of sarcomeric structures, expression of key mature sarcolemma (connexin 43) and myofilament (Troponin I) markers for cardiomyocytes ^[67–69], and stable spontaneous beating rate and contraction amplitude for over one month (Figure S22 and Movie S5, Supporting Information) occurred. Therefore, the printed hydrogel pattern does not inhibit the cardiomyocyte differentiation. The protein structure of GelMA, which is a chemically modified gelatin consisting of denatured collagen, in the CNT/GelMA and PEG/GelMA hydrogels could be enzymatically degraded by collagenase type I and type II ^[70–71]. However, the incorporated CNTs and PEG can serve as a water insoluble backbone and nondegradable component hindering the permeation of collagenase into the hydrogels, hence slowing their biodegradation rate ^{[45, 72]}. Furthermore, without collagenase type I and II, the degradation rate of GelMA hydrogels in biological media; i.e., PBS and cell culture media, is significantly slow ^[73–74]. Therefore, structural integrity of the GelMA-based, 3D-printed hydrogel patterns was minimally affected by the material’s degradation, their structures were preserved even after 17 days of culture (Figure S22A, Supporting Information). The cardiomyocyte alignment caused an upward contraction of the posterior region of the soft robot, as revealed in the results of optical flow analysis and actuation dynamics of the construct from 20 to 240 ms (Figure S22 H and Movie S6, Supporting Information).

After optimizing the printability of CNT/GelMA and PEG/GelMA inks, we constructed a multilayered hydrogel scaffold with both two and six layers of printed hydrogel patterns onto the wireless powering device to achieve a dense, 3D cardiac tissue construct that produces strong contractile behavior (Figure S23, Supporting Information). We obtained the CNT/GelMA and PEG/GelMA micropatterns with a line width as narrow as ∼70 μm onto the wireless device-embedded CNT/GelMA hydrogel layer to construct the muscle actuator (Figures 3D and 3E, Supporting Information). After seeding iPSC-CMs on the printed construct, the muscle actuator with the embedded wireless device exhibited well-organized sarcomeric α-actinin structures and the homogeneous distribution of connexin 43 (Figures 3 F, G, and I, Supporting Information) with interconnected networks beyond the patterned lines in the outermost layers (Figure 3H, Supporting Information); additionally, no cytotoxicity of the device was observed. Such a narrow line width allows the cardiomyocytes to bridge the adjacent CNT/GelMA hydrogel patterns, and thus creates interconnected muscle networks aligned toward the CNT/GelMA hydrogel patterns. The 3D cardiac tissue formed by a scaffold with a multilayered hydrogel pattern was thicker and denser than that formed by a scaffold with a lower number of hydrogel layers; this was confirmed through the z-stack projection of the constructs (Figure S24, Supporting Information). The interconnected outermost cardiac tissue allows the whole construct to synchronize its contraction. The soft robot showed strong synchronous and spontaneous beating behavior after 6 days of culture (Movie S7, Supporting Information).

Wireless local electrical stimulation of the muscle actuator initiated controlled activation waves that propagated outward from the center of the wireless device, as revealed via calcium imaging (Figure 3J, Figures S25 and S26, and Movie S8, Supporting Information). To control the activation rates of the muscle actuator at a specific target frequency (0.5–2.0 Hz), a monophasic pulse voltage (V_sup: 1.6–3.2 V, pulse width: 50 ms, single-turn excitation coil) was wirelessly applied to the wireless device. Calcium transient traces, during wireless pacing at different frequencies, revealed that the muscle actuator operates over a wide frequency range, from 0.5 to 2.0 Hz (Figure S26, Supporting Information). Our muscle actuator did not reach the stimulation frequency of 3.0 Hz under the applied voltage range, which corresponds to the results that paced the soft robots up to 2.0 Hz by external carbon rod electrode stimulation (Movie S9, Supporting Information). This is presumably due to the immature nature of iPSC-CMs and/or delayed relaxation/contraction behavior of muscle tissue induced by hysteresis during the actuation response of the hydrogel scaffold ^[6], especially for the large-sized tissue constructs compared with small-scale ones. It has been confirmed that the iPSC-CMs exhibit reliable pacing up to 2.0 Hz at an electrical field potential of 5 V/cm under the monophasic voltage, with a 50 ms pulse width (Figure S27, Supporting Information). The activation wave had a pacing frequency-dependency on the propagation velocity. Higher frequency pacing brought about slower propagation velocities (propagation velocity: ∼52.4 mm/s at 0.5 Hz, ∼37.7 mm/s at 1.0 Hz, ∼31.4 mm/s at 2.0 Hz, and ∼28.6 mm/s at 3.0 Hz), which may be attributed to gradually increased action potential durations and increased post-repolarization refractoriness that impair the sodium (Na) ions current recovery (as a result of the increased frequencies) ^[75–76]. The excitation threshold voltage increased with increasing frequency (Figure 3K, Supporting Information). Full synchronization was obtained with a V_sup voltage of more than 3.1 V, which corresponds to the maximum output peak-to-peak voltage of ∼3.4 V from the electrodes. Although we could not estimate the actual voltage that was applied to the cardiac layer because the voltage signal was attenuated by the CNT/GelMA layer, this threshold voltage was slightly lower than that for external carbon rod electrode stimulation (5.0 V for full synchronization at 0.5–2.0 Hz when the carbon rod electrodes with a radius of 5 mm are placed at an interspaced distance of 5 cm) (Movie S9, Supporting Information). The wireless electrical stimulation produced a forward change of ∼0.5 s and a reverse change of ∼3.5s in the beating rate of iPSC-CMs on the wireless device (Figure 3L and Movie S10, Supporting Information).

2.4. Wireless pacing of the biohybrid soft robots

When the soft robots were immersed in cell culture media (∼25 mL, ∼5 mm above bottom level) inside a 10-cm-radius Petri dish that was placed on the single-turn excitation coil, they floated in the medium and were propelled by the motion of the robot’s lateral part, initiated through muscle contractions (Movie S11, Supporting Information), as observed from finite-element deflection simulations. The anterior-most parts of the soft robots floated on the media surfaces due to the lower density of the PDMS-covered RF diodes (∼950 kg/m³), but their bodies were tilted to the bottom toward their posterior. The density of the remaining part of the muscle actuator (∼1,100 kg/m³) is high compared with that of the medium (∼1,000 kg/m³) (Figure S17, Supporting Information). Figure 4A and Movie S12 (Supporting Information) reveal the wave-like deformation dynamics of the soft robot’s lateral part, resulting from the CNT/GelMA hydrogel’s soft properties. Furthermore, they show the actuation behavior from a side-angle view recorded during a single stroke. The anterior region, which is placed at a long distance from the electrode boundary, is largely deformed in a wavy manner after the relaxation of muscle, which may cause a slight delay in the robot’s deflection to the posterior region. Optical flow and video analyses revealed that during a single stroke, the posterior region of the lateral part was lifted upward ∼70 ms earlier, followed by the subsequent upward bending of the anterior region. Both regions reached their maximum displacement of ∼3.5 mm at around ∼0.5 s. Afterward, it gradually relaxed to its original state at ∼1.6 s (Figures 4B and 4C). The deviation in the simulated maximum deflection of ∼2.1 mm from the finite-element simulation may have arisen from the applied passive tension force of ∼2 kPa; the passive tension force of 3 kPa gives a close match (∼3.2 mm) between the simulation and the actual measurement (Figure S2E, Supporting Information). The relatively slow actuation dynamics allowed for the wireless control of the beating rate of the soft robot up to 1.0 Hz (Movie S13, Supporting Information).

Figure 4.

Actuation behaviors of biohybrid soft robots. (A) Side-angle views during the single stroke of a soft robot. (B) Optical flow analyses highlighting the production of alternating upward and downward vectors by the soft robot’s lateral region. (C) Deflection at different locations highlighted using red x-marks in (A). (D) Locomotion of the robot illustrated by comparing snapshots during the single stroke of a 1 s recording (no wireless pacing, day 7). (E) Comparing the moving distance during the single stroke at different wireless pacing (V_sup: 3.2 V, pulse width: 50 ms, single-turn excitation coil) and spontaneous beating (1 s recording, day 7). (F) The temporal change in the spontaneous beating rate of the soft robot after wireless electrical stimulation (n = 3 independent measurements) and the locomotion speed (n = 6 strokes, V_sup: 3.2 V, pacing rate: 0.7 Hz, pulse width: 50 ms, single-turn excitation coil). The inset shows the average reduction in speed under 0.7 Hz pacing and spontaneous beating rate after three weeks from robot release (n = 3 robots). Error bars represent the standard error of the mean. (G) Comparison of temporal traces in moving position under 0.7-Hz wireless pacing and spontaneous beating (day 8, V_sup: 3.2 V, pulse width: 50 ms, single-turn excitation coil). The inset shows the average speed under 0.7 Hz pacing (n = 3 robots). Error bars represent the standard error of the mean. (H) Comparison of the moving trajectory under 0.7-Hz wireless pacing and spontaneous beating (day 8, V_sup: 3.2 V, pulse width: 50 ms, single-turn excitation coil).

To quantify the moving velocity, we recorded the locomotion of the soft robot and tracked its position via video analysis and modeling (Figures 4D and 4E), wherein the wireless device was marked by a black bar to track its movement trajectory. The locomotion speed was calculated by adjusting the x and y axes of the robot in the same direction as the direction of locomotion for each stroke. The upward deflection of the posterior region of the robot initially generated forward locomotion. However, with the increase in upward deflection amplitude, the increased viscous drag from a larger effective frontal area compensated and suppressed the ability to displace larger masses of the fluid, thereby generating backwards locomotion. Then, the downward deflection caused forward locomotion again. Therefore, forward locomotion was counterbalanced by the upward and downward deflection amplitude of the lateral part. Wireless electrical pacing resulted in a large locomotion amplitude; however, pacing at a frequency exceeding 1.0 Hz impaired forward locomotion due to the large fluctuation of the cell culture medium at the air–liquid interface, which was caused by the increased locomotion amplitude (Figure 4E). A maximum locomotion speed of ∼580 μm/s was demonstrated by adjusting the pacing rate to 0.7 Hz (Figure 4G and Movie S14, Supporting Information). The calculated relative speed was ∼0.0166 body lengths/s and is in a similar order range with that of other biohybrid soft robots (Table S2, Supporting Information). Further optimization of the design should consider increasing the locomotion speed. The process should involve the fabrication of hydrodynamically favorable soft robots with neutral buoyancy to avoid the strong surface tension force from the water–air interface, which can be approached by carefully optimizing the components of the muscle actuators and optimally aligning the muscle fibers, hence enabling the larger deflection of robots. Possible design options of 3D-printed hydrogel patterns for increasing deflection through the optimization of muscle fibers alignment are demonstrated in the Figure S28 (Supporting Information). The robots with the hydrogel pattern geometries whose CNT/GelMA lines are aligned at an angle of 120° or 150° to PEG/GelMA lines may have an enhanced deflection magnitude compared to robots with the orthogonalized geometry of hydrogel patterns. The hydrogel pattern geometry in which CNT/GelMA lines are aligned at an angle of 120° to PEG/GelMA lines will allow for the largest deflection. At the pacing rate, the soft robots exhibited side locomotion (Figure 4H). Additionally, we checked the locomotion of the 10 × 10 mm device-based soft robot (body size: ∼17 × 16 mm, weight: ∼45 mg, density: 990 Kg/m³, Figure S17, Supporting Information), which fully floated near the water–air interface. The robots exhibited similar side locomotion at a speed of ∼115 μm/s (Movie S15, Supporting Information).

Finally, we traced the temporal changes in spontaneous beating rates and locomotion speeds to evaluate the lifetime of the biohybrid robot’s actuation (Figure 4F). After releasing the robots into medium, on day 7, during a week of wireless electrical testing, major degradation in the beating rate and locomotion speed did not occur, and the robots retained 36% ± 6% and 83% ± 4% of their initial speed and beating rate, respectively, even after three-week periods. These results imply that the integrity of the cell and its contractile performance is preserved over a long period through wireless device integration and electrical stimulation. These results strongly support our concept of a biohybrid soft robot with an embedded soft driving system powered and controlled by a wireless power transfer system. Given that our system has the capability to not only attain the required compliance between the structurally flexible biohybrid robots and the soft driving platform, but also to integrate high-performance electronic functionality, our developed systems can be an attractive tool for applications in universal untethered soft robotic activation and in various biomedical devices, such as in wireless cardiac/skeletal muscles and in brain pacemakers. Possible concrete, short-term applications of our wireless bioelectronics will be muscle training and stimulation through wireless electrical input to improve and rebuild muscle function and muscle weight to reverse aging and to treat injury. These muscle therapeutic devices will be wearable, wireless, implantable, and miniaturized. In addition, one of the most exciting subjects that is being examined in the long-term for biohybrid soft robots is their complexity, which enables them to multitask between actions, such as interactive sensing, targeting, and manipulation. For example, the further integration of biological neural networks into the wireless bioelectronics with a neuromorphic integrated circuit will allow interactive actuating, sensing, and information processing for use in neuroprostheses and in human-machine interfaces. This work will provide useful knowledge for developing such multimodal biohybrid robots with interactive capabilities that can be achieved using integrated wireless circuits.

3. Conclusion

We developed stretchable, thin, lightweight, and biocompatible wireless cell stimulators that can be integrated into self-actuating soft muscle bodies. Furthermore, we adopted an in tandem architecture of a receiving coil and cell stimulation electrodes, thereby enabling embodied integration in untethered biohybrid soft robots that can swim. Additionally, we optimized their electrode geometry and wireless-powering system to transfer and receive power efficiently in a liquid biological environment. The cell stimulators generated controlled monophasic pulses of up to ∼9 V in media. By differentiating iPSC-CMs directly on the cell stimulator using hydrogel-based 3D multi-materials printing, we replicated the native myofiber architecture together with its function in terms of robustness and enhanced contractibility. This embodied soft robotic design allowed us to control the swimming motion of the robots by wirelessly transmitting electrical power into the cell simulator. The concept of our developed bioelectronic devices is applicable to the development of untethered and biohybrid systems with a multimodal function, such as interactive sensing, targeting, and manipulation through an integrated circuit design for various future biomedical and soft robotic applications.

4. Experimental Section

Fabrication of the stretchable wireless device:

The stretchable wireless devices were fabricated on a ∼37-μm-thick polyurethane film (STRC-40, Nikkan Industries Co., Ltd.) through a screen-printing method, based on the design shown in Figures S3 and S7 (Supporting Information). Silver ink (ECM CI-1036, Engineered Materials Systems, Inc.) and dielectric ink (ECM DI-7542, Engineered Materials Systems, Inc.) were used to print the electrodes and dielectric layer, respectively. The screen-printing of the wireless devices was performed by Nissha Co., Ltd, using a custom-made printer. During the printing, a stainless-steel screen with 500 mesh and a polyester screen with 200 mesh were used for silver ink deposition and dielectric ink deposition, respectively. Each layer of the printed wireless device was annealed at either 120˚C for 30 min (silver ink) or UV cured under a light intensity of 2,400 mJ/cm² (dielectric ink) before the other layers were deposited. The detailed fabrication process is shown in Figure S3 (Supporting Information).

Cell preparation:

Single-cell suspensions of iPSC-CMs and their plating and maintenance medium (iCell Cardiomyocytes² Kit) were purchased from FUJIFILM Cellular Dynamics, Inc. The culture medium for the C2C12 and CF cells contained Dulbecco’s modified Eagle medium (DMEM, Thermo Fisher Scientific Inc.) with 10% or 1% fetal bovine serum (FBS, Thermo Fisher Scientific Inc.) and 1% antibiotic–antimycotic solution (Sigma–Aldrich). The cytotoxicity was tested using a live/dead viability/cytotoxicity kit (Thermo Fisher Scientific Inc.). In the cytotoxicity assessments of the device, the C2C12 cells were seeded onto the devices at a low density of <1 × 10³ cells per device to avoid fast confluency.

Preparation of TMSPMA-coated glass:

First, glass slides (1-mm-thick and 5.08 × 7.62 cm in size) were washed with ethanol. Then, they were placed in a 250 mL beaker, and 5 mL of TMSPMA was poured onto them using a syringe. Subsequently, they were incubated in an oven at 80°C overnight, and, then, they were again washed with ethanol and dried. Finally, the TMSPMA-coated glass slides were wrapped in aluminum foil and left at room temperature (RT).

Fabrication of biohybrid soft robots without the wireless device:

Figure S18 (Supporting Information) presents our strategy for fabricating the biohybrid soft robots without the wireless-coil device, beginning with photo-crosslinking of the CNT/GelMA hydrogel. First, we placed 100-μm spacers made by stacking two layers of commercial invisible tape (50-μm-thick, 3M) on a TMSPMA-coated glass. Then, 300 μL of CNT/GelMA hydrogel solution was poured on top of the TMSPMA-coated glass with and without a wireless-coil device, and an uncoated glass slide was placed over the CNT/GelMA solution. Then, to photo-crosslink the CNT/GelMA hydrogel, a UV light (200 W mercury vapor short arc lamp with 320–390 nm filter, OmniCure S2000) was exposed through a photomask placed on the top of the uncoated glass slide at 120 mW/cm² for 40 s. The slides were carefully rinsed in Dulbecco’s phosphate-buffered saline (DPBS, Thermo Fisher Scientific Inc.) to detach the patterned CNT/GelMA hydrogel layer from the uncoated glass substrate. Subsequently, CNT/GelMA patterns and then PEG/GelMA patterns were 3D printed (INKREDIBLE 3D printer, CELLINK) at a speed of 1,600 mm/m, a spacing of 300 or 1,000 μm, a pressure of ∼15 kPa, and a conical nozzle of 30 G at RT. After photo-crosslinking the 120 mW/cm² construct for 120 s, we applied a sterilization protocol on the constructs, and then the iPSC-CMs were cultivated on them. Finally, the constructs were spontaneously detached from the TMSPMA-coated glass on day 4.

Fabrication of biohybrid soft robots with the wireless device:

A diode with a size of 1.2 × 0.8 × 0.6 mm and weight of <0.01 mg (BAT62, Infineon Technologies AG) was connected to the wireless-coil device using a silver paste (8330S, MG Chemicals) at 80°C for 30 min and subsequently covered with PDMS (Sylgard 527, Dow Inc.) cured at 80°C for 120 min (Figure S17, Supporting Information). The wireless-coil device was placed on the TMSPMA-coated glass with two spacers. Subsequently, we applied the same preparation protocol with the biohybrid soft robots without the wireless device. For the long-time operation and locomotion characterization of soft robots in Figure 4, we partially covered the cathode with a conductive carbon paste (16050, Ted Pella, Inc.) through a mask to avoid corrosion of the cathode.

Sterilization protocol of biohybrid soft robots:

Following 3D printing, the constructs were washed with 10% antibiotic–antimycotic solution (Thermo Fisher Scientific Inc.) for 5 min. Then, we drained the 10% antibiotic–antimycotic solution and incubated the constructs in a 1% antibiotic–antimycotic solution at 4°C overnight. Next, we drained the 1% antibiotic–antimycotic solution and washed the constructs with media once before seeding the iPSC-CMs. For the constructs with the wireless-coil device, in addition to the above process, after the incubation of constructs in a 1% antibiotic–antimycotic solution at 4°C overnight, the constructs were washed with medium five times at 37°C for 1 h to decrease the toxicity of the device. Then, the constructs were again incubated in the medium at 37°C overnight.

Culture of the iPSC-CMs on the wireless device:

After sterilizing the constructs, we thawed the iPSC-CMs and seeded them on the constructs at a density of ∼5 × 10⁵ cells/cm². After post-plating the iPSC-CMs on the constructs in a plating medium at 37°C for 4 h, the medium was changed to a maintenance medium. The maintenance medium was replaced every other day. Any electrical signals were not applied during the iPSC-CMs culture period to make cardiac tissue on the wireless device. The output voltage was only applied after creating cardiac tissue for controlling the biohybrid robots for a short time (typically, 5–15 minutes per experiment).

Actuation assessment for the wirelessly powered biohybrid soft robots:

The soft robots were released in the maintenance medium or DMEM (no phenol red, Thermo Fisher Scientific Inc.) to record their locomotion abilities. Videos of the locomotion of the soft robots were recorded using a digital camera with a macro lens (α6400, Sony). The deflections of the soft robots caused by muscle contractions and position during locomotion were measured from the video recordings using a video analysis and modeling software, Tracker (http://physlets.org/tracker).

Statistical analysis:

Measurements of the output voltages from the wireless devices, shown in Figure 2B, were performed for n = 3 devices, and the error bars represent the standard error of the mean. The excitation threshold PA voltages in Figure 3K were performed for n = 3 devices, and the error bars represent the standard error of the mean. Tukey’s multiple comparisons were used to compare the data groups in Figure 3K. The temporal traces of beating rate and speed of the soft robot in Figure 4F were performed for n = 3 independent measurement repeats and 6 strokes, respectively, and the error bars represent the standard error of the mean. The average reduction in speed and beating rate, shown in the inset of Figure 4F, were assessed for n = 3 robots, and the error bars represent the standard error of the mean. The average speed in Figure 4G was evaluated for n = 3 robots, and the error bars represent the standard error of the mean. The p-values and number of samples (n) are provided in the figure captions of the corresponding graphs.

Data availability Statement

All data needed to evaluate the conclusions in the paper are presented in the main manuscript and/or the Supplementary Information. Additional data are available from the authors upon request.