Abstract
A fundamental challenge for bioelectronics is to deliver power to miniature devices inside the body. Wires are common failure points and limit device placement. Wireless power by electromagnetic or ultrasound waves must overcome absorption by the body and impedance mismatches between air, bone, and tissue. Magnetic fields, on the other hand, suffer little absorption by the body or differences in impedance at interfaces between air, bone, and tissue. These advantages have led to magnetically-powered stimulators based on induction or magnetothermal effects. However, fundamental limitations in these power transfer technologies have prevented miniature magnetically-powered stimulators from applications in many therapies and disease models because they do not operate in clinical “high-frequency” ranges above 20 Hz. Here we show that magnetoelectric materials – applied for the first time in bioelectronics devices – enable miniature magnetically-powered neural stimulators that operate at clinically relevant high-frequencies. As an example, we show that ME neural stimulators can effectively treat the symptoms of a Parkinson’s disease model in a freely behaving rodent. We also show that ME-powered devices can be miniaturized to sizes smaller than a grain of rice while maintaining effective stimulation voltages. These results suggest that ME materials are an excellent candidate for wireless power delivery that will enable miniature neural stimulators in both clinical and research applications.
Wireless neural stimulators have the potential to provide less invasive, longer lasting interfaces to brain regions and peripheral nerves compared to battery-powered devices or wired stimulators. Indeed, wires are a common failure point for bioelectronic devices. Percutaneous wires present a pathway for infection1 and implanted wires can also limit the ability of the stimulators to move with the tissue, leading to a foreign body response or loss of contact with the target tissue2,3. Additionally, chronic stress and strain on wires, particularly for devices in the periphery, can lead to failure in the wire itself or its connection to the stimulator4. In small animals like rats and mice, wires used to power neural stimulators can interfere with natural behavior, particularly when studying social interaction between multiple animals5.
One of the primary challenges for wireless neural stimulators is to create efficient miniature devices (< 1 cm in length) that operate reliably beneath bone and tissue as an animal or human patient engages in normal activity. At lengths of less than 1 cm, devices could be fully implanted in the periphery and be light enough to allow for unrestricted animal behavior; however for devices this small, power delivery remains a challenge. Efficient power transfer with propagating electromagnetic waves requires antennas with feature sizes comparable to the electromagnetic wavelength. Thus, for sub-millimeter devices, such as the proposed RF powered “neurograins6,” effective power-transfer frequencies lie in the GHz range, where electromagnetic radiation is absorbed by the body7. Absorption of this radio-frequency electromagnetic energy limits the amount of power that can be safely delivered to implants deep inside tissue7. As a result, researchers typically turn to magnetic induction or batteries to power implanted devices; however, these techniques also limit the degree of miniaturization. Batteries increase the size of the device and add considerable weight. Additionally, batteries require replacement or charging, which can limit the potential uses. Inductively coupled coils, on the other hand, can be made smaller and lighter than batteries, however; the power a receiving coil can generate is directly related to the amount of flux captured by the area of the coils. Thus, when the receiver coils are miniaturized, the output power reduces and becomes more sensitive to perturbations in the distance or angle between the transmitter and receiver8. For example, Freeman et al. demonstrated that small inductive coils less than 1 mm in diameter can power stimulators for the sciatic nerve in anesthetized rats9; however, in its present form, this device would have difficulty providing stable stimulation in freely moving animals due to the reduced power coupling efficiency that accompanies changes in the angle and distance between the receiver and transmitter coils.
Additionally, for neural stimulators to treat a number of neurological disorders like Parkinson’s Disease (PD), obsessive-compulsive disorder, and epilepsy, they must operate safely and effectively in the high-frequency “therapeutic band” between 100 and 200 Hz10–12. This type of high-frequency neural stimulation is challenging because charge on the electrode must be dissipated between successive stimulation pulses to prevent electrolysis, tissue damage, and changes to the local pH13. Charge dissipation at high-frequencies is accomplished by using a biphasic stimulus waveform that actively or passively charges and discharges the electrode with each cycle. Indeed all clinically approved electrical neural stimulation therapies in this therapeutic band use various forms of “charge balanced” biphasic stimulation waveforms14.
Recently, several promising alternatives to magnetic induction and batteries have enabled miniature neural stimulators; however, these approaches have yet to demonstrate in vivo operation in the therapeutic high-frequency band in freely moving animals. Montgomery et al. and Ho et al. have shown that one can use the mouse body as an electromagnetic resonant cavity to effectively transfer energy to sub-wavelength scale devices implanted inside the animal15,16. This approach is particularly effective to drive tiny LEDs for optogenetic stimulation. However, because the electrical waveform is monophasic, electrical stimulation has been limited to < 20 Hz. Using superparamagnetic nanoparticles to absorb energy from high-frequency (500 kHz) magnetic fields17, one can heat specific regions of the brain 18,19 in freely moving animals19. This local heat can stimulate neural activity when the targeted brain region is genetically modified to respond to changes in temperature18,19. However this approach requires transgenesis, which adds regulatory complexity and has yet to show high-frequency operation due the requirement for the tissue to cool between stimulation intervals. Ultrasound provides a promising and efficient method to power bioelectronic implants because ultrasound wavelengths are 105 times smaller than electromagnetic waves at the same frequency allowing sub-millimeter-sized devices to have wavelength-scale piezoelectric antennas 20,21. However, implementation of these “neural dust” motes can be challenging in freely moving animals because the impedance mismatch between air, bone, and tissue requires contact between soft tissue and the ultrasound transducer for efficient power transfer. As a result, there has yet to be a demonstration of ultrasound-powered neural stimulators in freely moving animals22.
Here we show that magnetoelectric (ME) materials enable the first magnetically powered miniature neural stimulators that operate in the therapeutic high-frequency band. Similar to inductive coils, these materials transform a magnetic field to an electric field, but instead of using an implanted coil we use a material that generates a voltage via mechanical coupling between magnetostrictive and piezoelectric layers in a thin film. Namely, the magnetic field generates strain in the magnetostrictive layer as the magnetic dipoles align with the applied field. That strain exerts a force on the piezoelectric layer, which generates a voltage (Fig. 1). By exploiting this transduction mechanism, magnetoelectrics do not suffer from the same miniaturization constraints as coils and can be driven by weak magnetic fields on the order of a few millitesla. These properties have led researchers to propose magnetoelectrics as a promising material for bioelectronic implants23–27. Here we demonstrate the first proof-of-principle wireless neural stimulators based on ME materials in a freely behaving rodent model for Parkinson’s Disease (PD), and that these materials could power miniature devices deep within the human brain.
Fabrication and characterization of ME stimulators
We fabricated proof-of-principle ME stimulators by bonding a rectangular magnetostrictive layer (Metglas) to a platinum coated piezoelectric layer, polyvinlydine fluoride (PVDF). We then encapsulated the films in a protective parylene-C layer (8-10 μm thick) (Fig. 1a, see Methods). We used PVDF layers between 28 and 110 μm, which yielded total device thicknesses between 50-150 μm. When we measured the voltage across the film, we found a dramatic voltage increase when the applied magnetic field frequency, matches an acoustic resonant frequency (Fig. 1b). Because the resonant frequency is proportional to the inverse of the film length, we can design multiple films and selectively activate them by changing the stimulus frequency (Fig. S2b). Using this principle, we can use different magnetic field frequencies to activate separate devices that may be in different areas of the body, or create biphasic stimulators by interleaved resonant stimulation of two different films, with each film driving either the positive or negative phase of the neural stimulus.
We can further enhance the voltage generated by the ME films by applying a constant bias field with a permanent magnet or an electromagnet (Fig. 1c). Because the strain in the magnetostrictive material is a sigmoidal function of the magnetic field strength, the change in voltage produced by the alternating field is largest when the field oscillates about the midpoint of the sigmoid (Fig. S1)28,29. Thus, we use a bias field to offset the magnetic field near the center of the sigmoidal magnetostrictive response curve. This bias field allows us to generate therapeutic voltage levels while applying a small (few mT) alternating magnetic field using an electromagnetic coil and custom control circuitry that specifies the frequency and timing of the alternating magnetic field (Fig. S3).
To identify the operational frequencies for our ME stimulators we tested them in saline and found that with a biphasic stimulation waveform we could apply constant stimulation up to at least 800 Hz without significant hydrolysis. For this test we operated either one film for monophasic stimulation or two films for biphasic stimulation attached to a stereotrode (Microprobes) in saline (see Methods). We then measured the time at which we could see bubbles at the electrode tip resulting from hydrolysis. This hydrolysis event indicates conditions that would lesion the surrounding tissue. We found that with a monophasic stimulation waveform stimulation frequencies above 50 Hz produced hydrolysis while biphasic charge-balanced stimulation showed no hydrolysis up to the maximum tested frequency of 800 Hz. Compared to previously demonstrated miniature magnetic neural stimulators, the biphasic ME devices shown here are the first to access the high-frequency bands used for clinical applications like the treatment of Parkinson’s disease and obsessive-compulsive disorder (Fig 1d).
An additional challenge for any wirelessly powered neural stimulator is to maintain a well-regulated stimulation voltage. This challenge is especially prevalent as devices become small, which often reduces the power transfer efficiency resulting in a greater sensitivity to the alignment between the device and power transmitter. ME materials offer two main advantages that can enable stable and effective stimulation even as devices become small and move with respect to the driver coils:
First, ME devices generate voltages well in excess of the effective stimulation potential, allowing them to be effective even if the materials are misaligned with the driver coils. At resonance, we have measured ME voltages in excess of 30 V at a field strength of only 1 mT (Fig. S2c, d). Because effective stimulation voltages are usually between 1-5 V, we can cap the applied voltage to this effective stimulation range using an LED or Zener diode. As long as the voltage generated by the ME film is greater than or equal to the capping voltage, we will apply approximately the same stimulus voltage regardless of the angle or distance between the driver coil and the ME film. For a typical film we found that we could reorient the film by +/-80 degrees and maintain voltages in excess of 3 V (Fig. S1h). This large angular tolerance is aided by the large magnetic permeability of the Metglas layer, which helps to direct the magnetic field lines along the long axis of the film, where they are most effective at creating a magnetostrictive response.
Second, the voltage generated by a piezoelectric material depends on the thickness of the piezoelectric layer and not the area of the film30, allowing us to fabricate small magnetoelectric films that generate roughly the same stimulation voltage as larger devices. Figure S2 shows the peak voltage generated and quality factor for ME films of different areas. We found that, for a 52 μm thick PVDF layer, the voltage remains around 10 V even as the film length decreases. Variations of +/-40 % in peak voltage and quality factors are likely due to defects produced during film fabrication, which may be reduced with improved manufacturing. We also verified that the output voltage depends only on the piezoelectric film thickness by measuring the peak voltages from ME devices with three different thicknesses of PVDF: 28 μm, 52 μm, and 110 μm. As expected, we see that the peak voltage increases linearly with the PVDF thickness and is independent of the film length. We calculated (and experimentally confirmed) that the power generated by a ME device is proportional to the film width for a given thickness and a length-to-width ratio >3 (see Fig. S2f). Despite the decrease in power as films become smaller, we calculate that films less than 1 cm long can generate up to 4 mW, which is more than sufficient for many wireless applications including neural stimulation31.
Monophasic stimulation by ME films evoke action potentials in vitro
Using fluorescence microscopy to image voltage in cultured cells, we found that monophasic stimulation for 50 ms at 100 Hz by ME films reliably stimulated action potentials (APs). For these experiments we used “spiking” human embryonic kidney (HEK) cell lines that were modified to express sodium and potassium channels. These cells have spike-like electrical waveforms that are rectangular in shape and can last for a few seconds depending on the confluency of the culture32. To determine the relative timing between magnetic stimulation and action potential generation, we transfected these cells with ArcLight33-a genetically encoded voltage indicator that allows us to measure action potentials using fluorescence microscopy.
To image fluorescence while we applied magnetic fields, we developed an experimental setup that allows us to place cells and ME films beneath an objective lens at the center of a 10 cm long solenoid with a 3 cm gap in the center. This configuration allowed us to place ME films, cells, and the objective lens at the center of the applied magnetic field (Fig. 2a). Two slightly larger coils placed on either side of the gap provide the constant bias field.
We then approximated an implanted ME stimulator using two experimental configurations: 1) growing cells directly on the ME film (Fig. S4) and 2) laying a coverslip with adherent cells on top of the ME film (Fig. 2). To culture cells directly on the ME film, we coated the top parylene layer with poly-l-lysine. The healthy proliferation of HEKs on the ME device indicates that this encapsulation approach prevents the ME materials from limiting cell growth (Fig. S4b). However, in a typical use case, the target cells may not adhere to the ME stimulator, so we also tested the response of cells laid on top of the ME materials. In this configuration we first grew the cells on coverslips for 3-5 days before inverting the coverslips and laying them on the ME for testing (Fig. 2, see Methods).
To create fringing electric fields that interact with the cultured cells, we stamped holes in the ME film (Fig. 2b). The films were otherwise fabricated as described above (Fig. 1, Methods). In experiments using ME films and Pt electrodes we found that high-frequency biphasic stimulation at the ME resonance frequency (typically 20-150 kHz) was not effective to stimulate APs in cultured HEKs, as predicted by the low-pass filtering properties of the cell membrane9. To create an effective monophasic stimulus waveform, we used a Schottky diode to rectify the voltage to create entirely positive or negative voltage waveforms depending on the diode direction. This rectified waveform has a slowly varying monophasic envelope in the <500 Hz frequency band where cells are responsive (Fig. 2c,d).
For both cells grown directly on the ME films and those placed in contact we found that five stimulation pulses with an envelope frequency of 100 Hz consistently stimulated APs in the spiking HEK cells (Fig. 2e, S4c, Supplementary Video 1). Critically, this 10-500 Hz stimulation frequency is spans the therapeutic window for many deep brain stimulation treatments34, and difficult to achieve with other wireless stimulators that compensate for low-efficacy energy harvesting by charging on-board capacitors35. For our experiments, the carrier frequency of the magnetic field was at the resonant frequency of the device, which varied between 20-40 KHz depending on device length. To test stimulation reliability, we repeated the 5-pulse stimulus three times over a period of 30 seconds. We observed APs for each stimulation pulse in n = 43 cells on coverslips and n = 144 cells grown on films. We confirmed that the APs stimulated by the ME film were in fact the result of resonant excitation of the film and not an artifact of the applied magnetic fields by imaging voltage-sensitive fluorescence when the magnetic field was tuned off of the resonant frequency. For non-resonant excitation we observed no correlation between the applied field and fluorescently detected APs in the spiking HEKs (Fig. 2f, S4d), supporting the conclusion that APs were stimulated by the ME film at resonance. We also confirmed that the fluorescent signal recorded indeed represents the voltage-dependent ArcLight response by imaging cells transfected with voltage-independent cytoplasmic GFP. These cells showed no change in fluorescence when the films were driven at the resonant frequency (Fig. 2g).
Biphasic stimulation by ME films evoke action potentials in brain slices
As described above, biphasic stimulation is preferred for most applications due to the desire to create a charge-balanced stimulus that reduces charge buildup and undesired electrochemical reactions at the electrode surface13. While the voltage waveform produced by ME films at resonance is biphasic, these resonant frequencies (typically 20 – 150 kHz) are too high to produce reliable cell stimulation, as described above. To create an effective biphasic stimulus in the therapeutic window (100 – 200 Hz), we use two films with distinct resonant frequencies connected to the same stimulating electrodes (Fig. 3a). The first film is attached to a full wave rectifier, which is oriented to generate a positive pulse, while the second film is attached to a full wave rectifier that generates a negative pulse.
The transistors block currents generated by one film from propagating through the circuitry attached to the other film, ensuring that only one half of the circuit is active at a time. By switching the magnetic field frequency between the two ME resonance frequencies, we can alternate positive and negative phase stimulation to create a biphasic neural stimulator (Fig. 3b-d). In this case the residual charge of -2.3 nC, which discharges in <2 ms, implies that this stimulator can safely operate at frequencies up to >500 Hz without accumulating charge.
We found that our biphasic ME stimulator is capable of repeatable neural stimulation using neocortical brain slices derived from mice that express the genetically encoded calcium indicator GCaMP3 in GABAergic neurons. To image neural activity following ME stimulation we inserted a stereotrode attached biphasic ME stimulator described above while we imaged GCaMP activity using fluorescence microscopy (Fig. 3e-g, Methods). We chose neural stimulation parameters similar to those commonly used for deep brain stimulation34: 100 biphasic pulses at 200 Hz with each phase lasting 175μs. When the magnetic field was on we observed a corresponding increase in fluorescence in n=23 recordings in neocortical layer 5 consistent with activity-mediated calcium increases (Supplementary Video 2). Following bath application of tetrodotoxin (TTX, 500 nM) fluorescence increases were completely blocked in n=9 recordings confirming that ME stimulation reliably evoked sodium-channel dependent action potentials in nearby neurons.
ME Neural Stimulation in Freely Moving Rats Shows Behavioral Efficacy
A major advantage of our ME stimulators is the fact that remote activation enables experiments with freely behaving animals. As a proof-of-principle we adapted our biphasic stimulator for deep brain stimulation (DBS) in freely moving rats (Fig. 4). To test ME stimulator efficacy, we used a previously reported protocol to test DBS in hemi-parkinsonian rats36. In these experiments rats are injected with 6-OHDA in the left medial forebrain bundle (MFB) to create a unilateral lesion of the substantia nigra pars compacta (SNc). The animals are then placed in a 30 cm diameter circular enclosure. Following a dose of methamphetamine, the hemi-parkinsonian rats have been shown to rotate ipsilateral to the injection (e.g. left for injection into the left MFB). During these rotations, the rat primarily moves using its contralateral (right) forepaw, rarely placing the ipsilateral (left) forepaw onto the ground. When a biphasic stimulus is applied at 200 Hz in the sub-thalamic nucleus (STN) using a tethered electrode array stimulator, rats typically stop turning to the left and exhibit more normal behavior such as moving with both forepaws, maintaining a steady orientation, or turning to the contralateral side34.
To create a wireless, biphasic ME stimulator for freely moving animals we added a small permanent magnet to the ME stimulator to generate a bias field, and wrapped the behavioral chamber with 18 AWG copper wire to create a solenoid (Fig. 4a, S5). By integrating the small permanent magnet (< 0.25g) into the ME stimulator, we could ensure that the bias field was constantly aligned with ME films as the animal moved within the enclosure. We could also ensure that the positive and negative stimuli had equal amplitudes by independently adjusting the distance between each film and the permanent magnet. This ME stimulator was then connected to a commercial electrode array (Microprobes) implanted in the STN (Fig 4b, see Methods). We ensured that the stimulation voltage and current were within the safe and therapeutic range by measuring the output of the ME stimulator connected to an equivalent circuit model of the brain (Fig. 4c, see methods). Specifically, we observed peak voltages of approximately +/-1.5 V and peak currents of approximately +/-100 μA for 400 us at approximately a 50% duty cycle (200 μs of overall current per phase), which is within the effective stimulation range reported for conventional wired stimulators36. When we tune the magnetic field frequency off resonance we observe almost no generated voltage or current (Fig. 4c).
We then tested the wireless version of our biphasic ME stimulator mounted to the head of a freely behaving rat and found that ME stimulation showed efficacy comparable to previously reported wired DBS stimulators (Fig. 4). With a magnetic field applied at resonance, we found that one-minute periods of 200 Hz biphasic pulses resulted in a significant decrease in the animal’s rotation rate (Fig. 4d green intervals). This decreased rotation was not observed when the magnetic stimulus frequency was tuned off resonance (Fig. 4d blue intervals). Plots of the head trajectories show that the pathological rotations observed during off-resonant magnetic field stimulation are not present when the ME stimulator is active during resonant magnetic field stimulation (Fig. 4e, Methods). When averaged over all trials, average rotation rate during the first half of stimulation fell to a statistically significant 1.6 rotations per minute (rpm), compared to 9.3 rpm in the absence of stimulation, or 9.4 rpm during off-resonant stimulation (paired t-test, Fig. 4f). We further demonstrated the repeatability of this stimulator by repeating this stimulation protocol on a second rat and found similar results (Fig S5b).
With a weight of 0.67 g, the ME stimulators described here are the first reported miniature, magnetic, high frequency stimulator. Furthermore, by changing the frequency and timing of the external drive coils, we can generate a variety of stimulation patterns throughout the therapeutic window of 100-200 Hz with applications to other disease models. Additionally, calculations of the magnetic field strengths suggest that we can reconfigure the drive coils for a number of behavioral experiments by placing coils beneath the floor of an animal enclosure. Finite element simulations and measurements show that even at distance 4-5 cm above a drive coil, ME films generate sufficient voltage for stimulation (Fig. S6). This distance could be further improved by optimizing the geometry of the coils or increasing the power of the magnetic field.
Demonstration of multichannel deep brain stimulation in skull phantom using rice-sized ME films
In addition to supporting experiments in freely moving rodents, ME materials could enable miniaturized wireless stimulators that operate deep in the brain of large animals or human patients and are individually activated with an external electromagnet. To our knowledge, this is the first technology that enables independent external wireless control of multiple miniature stimulators deep beneath a human skull phantom. Figure 5a shows the predicted depth that various miniature antennas could be safely implanted under the skull and generate 1 mW of power, which is in the approximate maximum power required for high-frequency continuous neural stimulation31. As mentioned above, radio-frequency (RF) powered antennas that operate at frequencies above ~1 MHz have limitations in the amount of power that can safely be delivered to an implanted device without causing potentially harmful tissue heating. Simulations show that when operating with the safe power limits, RF-antennas must be placed on the surface of the brain or in very shallow regions to harvest 1 mW of power. “Mid-field” techniques37, improve the RF coupling efficiency enabling deep operation, but because this approach operates at a fixed frequency there have yet to be demonstrations of individually addressable motes or biphasic stimulation. Other techniques for wireless power delivery discussed previously, such magnetic induction, also cannot achieve deep multichannel stimulation. For example, even using a higher operating frequency of 1 MHz an inductive coil with the same orientation and cross-sectional area as the ME films shown here would require a minimum of 500 turns of wire to generate 2 V using the same 0.5 mT field used here (assuming a typical Q-factor of 10). Thus, devices based on magnetic inductors cannot be miniaturized without sacrificing available power as described previously9.
As a proof-of-concept demonstration we show that two rice-sized ME films can be individually addressed at the center of a human skull phantom using an external electromagnet. These two-films with lengths of 8 mm and 10 mm have acoustic resonant frequencies of 180 and 200 kHz, which are determined by the film length. When these films are attached to an orange LED, their output voltage is capped at approximately 1.8 V, which helps to regulate the stimulation voltage and allows us to visualize film activation. ME films of this size are smaller than current DBS leads and could potentially be implanted into deep brain areas as shown in Fig 5c. Additionally, the magnetic stimulation coil is small enough to be incorporated into a stylish hat or visor that could be worn comfortably by a patient. When we placed the two ME films at the center of a skull phantom we found that we could individually illuminate the LEDs on each film when we applied a magnetic field at the resonant frequency of the selected film (Fig 5d-g). For this experiment we used a 400 W power supply, which produced a field of approximately 0.5 mT at the center of the skull phantom. The top of the skull phantom was removed for visualization, but had no affect on our ability to drive the LED indicators. The number of stimulation channels could be increased with the addition of ME films with different resonant frequencies.
Outlook
To our knowledge, this is the first demonstration of a miniature, magnetic neural stimulator that 1) operates in the therapeutic band (100-200 Hz) in freely moving animals and 2) enables individually addressable miniature stimulators deep within a human skull phantom; however, the advantages of ME materials extend beyond these proof-of-principle demonstrations.
ME stimulators such as the one described in the in vivo rat experiment could have an immediate impact on the study of DBS therapies using rodent disease models. Because the ME stimulator is compatible with commercial implanted electrodes, and the magnetic stimulators can be adapted to a number of standard behavioral experiments or animal enclosures, our ME stimulators could readily replace the wired DBS stimulators currently in use. As a result, new experiments can be developed to probe the effects of chronic and continuous DBS or DBS in social contexts where wired DBS stimulators would be impracticable.
Additionally, ME materials have the potential to enable miniature neural stimulators that can be implanted deep in the brain of large animals or humans and addressed externally with a small electromagnet. As shown here, rice-sized films can be selectively activated based on unique resonant frequencies. Additional miniaturization is not expected to reduce the voltage produced by these films since the voltage depends on the thickness of the piezoelectric field and not the film length (Fig. S2 c), suggesting that even smaller films could serve as effective stimulators.
We also foresee applications for ME materials as a wireless power technology for more complex implantable bioelectronic devices. For example, the demonstrated ability of ME films to power LEDs implies that ME materials could power implantable optogenetic stimulators, or small integrated circuits for physiological monitoring.
To realize these fully implantable bioelectronic devices, work is needed to improve ME materials and fabrication processes to reliably produce high-quality miniature ME films, and encapsulate them for chronic use. For wearable technologies, it is also necessary to further miniaturize magnetic field generators so that they can be battery powered and comfortably worn. These advances must also be accompanied by in vivo testing to show safety and efficiency for chronic use.
Overall, ME materials have the potential to fill a key need for wireless power delivery to miniature biphasic neural stimulators and other bioelectronic devices where the major challenge is transferring energy over distances of several centimeters without heating the tissue or suffering loss at interfaces between tissue, bone, and air.
Methods
General Statistical Methods
Error bars in Figure S2fg denote +/-one standard deviation for n=~50 data points. We furthermore performed a Tukey’s Honest Significant Difference test on the data in Figure S2g, which indicated that the voltage produced at each different PVDF thickness is significantly different. Paired t-tests were used for the rotation tests in figure 4e.
Film Fabrication
To fabricate ME films, we used Metglas SA1 alloy (Metglas Inc) for the magnetostrictive layer and polyvinylidenefluoride “PVDF” (precision acoustics) for the piezoelectric layer. The PVDF films used for these experiments were pre-stretched and poled by the manufacturer. The two layers were bonded together using an epoxy capable of transferring the mechanical stress between the two layers (Hardman double bubble red epoxy). Prior to bonding the two layers together, we sputtered a thin layer of platinum (<100 nm) as a top electrode on the PVDF. Both the Metglas and PVDF were plasma cleaned using O2 plasma for five minutes before epoxying. After the epoxy set, the films were cut into the desired rectangular shape using scissors, taking care to cut the long axis of the film along the stretching direction of the PVDF. We then attached wires using conductive epoxy to either side of the films in order to measure the electrical capabilities of the film. We found that attaching wires in the center dramatically increased the resonant voltage. However for convenience, the wires were attached near the ends of the films during the in vitro experiments. In many cases we also attached additional electronic components such as diodes or LEDs to the wires attached to the films as noted in the appropriate sections in the main text. Finally the devices were coated in 5-10 μm of parylene-C (Labcoater 2). Initially this coating was used to electrically insulate and protect the devices during in vitro experiments, but we also found that the encapsulation increases the resonant voltage, which could be due to increased mechanical coupling from the encapsulation.
Bench Top Electrolysis Tests
The stimulator shown in Fig 4a was wired to a stereotrode immersed in saline under a microscope in order to observe the formation of bubbles from electrolysis at the tips. During monophasic stimulation we used only one resonant frequency and during biphasic stimulation we used two frequencies as demonstrated above. In each case the pulse time was a 400 μs/phase. We determined the limit of stimulation time as when the first bubble began to appear at the tips of the electrode and repeated each data point 4 times.
Magnetic Field Generation (Fig. S3)
Each magnetic field generator consists of two major components, 1) Magnetic coils used for the alternating magnetic field (described in the main text and optimized for each experiment) and 2) Electronic drivers to control voltage and timing of the alternating current in the coils (the same for all experiments).
To maintain simplicity, efficiency, and low cost the coils were driven with full H-Bridge style switching circuits. The drivers are designed to deliver high currents to the drive coils in the form of bi-phasic pulse trains. This reduces the cost and complexity of the driver itself, as well as the power supply and control circuitry when compared to arbitrary function generators. The design also has potential for improved operational efficiency through impedance matching with the drive coils. Furthermore, it is also possible to regulate power delivered to the drive coils on the fly by adjusting the duty cycle of the current pulses, allowing power being delivered to the ME film to be easily controlled digitally while maintaining the resonant carrier frequency. The output carrier and pulse frequencies of the magnetic field are generated using a TeensyLC board and custom Arduino code to generate the specific pulse timings to deliver controlled ME stimulation (Fig. S3c,d).
These coils and drivers can be combined in different ways to generate the appropriate field for a given experiment. For example, the setup used to generate the alternating field in the in vivo rotation experiments consisted of four sets of coils each with five turns powered by one driver with all four drivers synced to the same output signal. In this way we can generate sufficient power to generate a mT-scale magnetic fields over the whole behavioral area (Fig. S5a).
Cell Culture
For experiments performed on coverslips, HEK cells expressing sodium channel Na1.3 and potassium channel K2.1 were grown on 12 mm poly-l-lysine coated coverslips to approximately 30% confluency. The cells were then transfected with the genetically encoded voltage indicator ArcLight using Lipofectamine (Invitrogen) following manufacturer’s recommendations. Two to three days after transfection the coverslips were inverted onto ME films for testing. Preparation of GFP controls followed the same procedure with the exception of replacing the ArcLight vector (AddGene) with a GFP expression vector (AddGene). For experiments performed with cells grown on the films, HEK cells transfected with ArcLight were placed onto parylene coated poly-l-lysine treated films. The films were placed in cellular media overnight and tested the following day.
ArcLight and GFP were excited at 460 nm with an LED light source. Fluorescence images were collected at 33 fps using a CCD camera. Images were analyzed using Matlab to quantify fluorescence changes in individual cells. In vitro testing was performed in extracellular buffer (ECB, in mM: NaCl 119, KCl 5, Hepes 10, CaCl2 2, MgCl2 1; pH 7.2; 320mOsm)
Figure S4b was obtained by growing unmodified HEK cells on a film submerged in cellular media for five days. The cells were then stained with Hoechst and Calcein-AM to label the nucleus and membrane respectively in living cells. The cells were then fixed and imaged using a confocal microscope.
Mouse Brain Slice Procedures
We used 40 day old GAD2-GCaMP3 mice, generated by crossing GAD2-Cre (JAX # 10802) with flox-GCaMP3 (JAX # 14538) animals. Preparation of brain slices followed procedures described by Ting et al.45 and was carried out in accordance with National Institutes of Health guidelines and approved by the UTHealth animal welfare committee. Mice were deeply anesthetized with Isoflurane and perfused with ice cold NMDG-based solution consisting of (in mM): 92 NMDG, 2.5 KCl, 1.25 NaH2PO4, 10 MgSO4, 0.5 CaCl2, 30 NaHCO3, 20 glucose, 20 HEPES, 2 thiouera, 5 Na-Ascorbate, 3 Na-pyruvate, saturated with 95% O2 and 5% CO2., at a rate of ~6 ml/min. Coronal brain slices (300 μm) were cut using a vibratome (Leica VT1200S), incubated for 15 min at 35 °C in NMDG solution, and then transferred to a chamber held at room temperature containing (in mM): 92 NaCl, 2.5 KCl, 1.25 NaH2PO4, 2 MgSO4, 2 CaCl2, 30 NaHCO3, 25 glucose, 20 HEPES, 2 thiouera, 5 Na-Ascorbate, 3 Na-pyruvate, saturated with 95% O2 and 5% CO2. For experiments, slices were placed into a recording chamber perfused with ACSF containing (in mM): 126 NaCl, 2.5 KCl, 1.25 NaH2PO4, 2 MgCl2, 2 CaCl2, 26 NaHCO3, 10 glucose), held at 32-34 °C using an inline heater. NBQX (10 μM) was included in the bath solution to block AMPA receptor-mediated synaptic transmission. The stereotrode was placed in layer 5 of somatosensory (barrel) cortex.
GCaMP3 was excited at 460 nm with an LED light source. Fluorescence images were collected at 9.8 fps using a CCD camera attached to an Olympus BX51WI microscope. Images were analyzed using Matlab to quantify fluorescence changes in 600 x 600 μm regions around the stereotrode tips.
Implant Design and Rat Surgical Procedures
Two male Long-Evans rats (n ≈ 1, 400 g) were anesthetized with isoflurane gas. Five percent isoflurane was used to induce anesthesia and two percent was used to maintain anesthetic depth. Buprenorphine (0.04 mg/kg) was administered 30 minutes prior to ear bars for analgesia. 5 - 7 skull screws were placed to anchor the electrode array. Skull screws were bound to skull with Metabond dental acrylic. A craniotomy was made to accommodate the micro electrode array and expose an injection site for neurotoxin. A 30 gauge needle bent at the tip cut and pulled away the dura mater covering of the brain. Desipramine (DMI) reconstituted in saline at a concentration of 15 mg/mL was injected IP to protect noradrenergnic neurons. The dose of DMI was approximately 15 mg/kg and injected approximately 30 minutes prior to administration of neurotoxin. To induce a hemiparkinsonian lesion, 8 ug of 6-hydroxydopamine (OHDA) at 2 ug/uL in saline was injected at 0.2 uL/min into the mid forebrain bundle (MFB -1.2 ML, -4 AP, and -8.1 DV). STN stimulation was delivered via a 2×2 platinum iridium microelectrode array (Microprobes) with 600 x 600 μm spacing of 75 μm electrodes. Each electrode had a nominal 10 kOhm impedance. The electrode array was lowered to -2.6 ML, -3.6 AP, and approximately -8.2 DV from bregma. The array was fixed to the skull with dental acrylic. All experiments were approved by the Institutional Animal Care and Use Committee of Rice University.
Prior to stimulating each rat with the magnetoelectric stimulator, the stimulator power was estimated via a benchtop approximation of the rodent electrode impedance. Constant current stimulation of the rodent brain with an A-M Systems 4100 stimulator produced characteristic voltage waveforms that approximated a simplified parallel RC circuit. A 56 kOhm resistor, and 440 pF capacitor in parallel closely approximated the impedance characteristics of the rat brain across the stimulating electrodes. Using this circuit model, we estimated the field strengths and pulse durations necessary to produce the desired stimulation effects and confirm that the stimulation was charge balanced prior to rodent experimentation.
Rotation Test Experiments
Prior to performing the rotation tests the rat was briefly anesthetized with 5% isoflurane gas and injected intraperitoneally (IP) with methamphetamine (0.31 ml 1.25 mg/kg) and the wireless biphasic stimulator was plugged into the implanted electrode array. After the anesthesia had worn off (about 5-10 min) the rat was placed in the cylindrical behavioral chamber. The magnetic field was applied over the whole behavioral area to the films on the device (Fig S5a).
The magnetic field was applied on resonance and off resonance for one minute at various times during the 40-minute trial. The resonant frequencies were 73 kHz and 77 kHz and the off resonant frequencies were 63 kHz and 87 kHz.
Rodent Tracking
Head position on the rotation task was generated using a slightly modified version of DeepLabCut46 to track ears, snout, and implant. A dataset totaling 286 frames from both the on and off resonance rotation tasks was hand labeled and trained for approximately 140,000 iterations.
Skull Phantom Demonstration
At the magnetic field frequencies used for this experiment bone and tissue are effectively transparent47, so we selected a life sized skull with the size of an average human adult head as a phantom (Orient Infinity Limited). It was wrapped with 18 AWG magnet wire as shown in Fig 5. The coil consisted of four coils in parallel each wired to an individual magnetic field driver. All drivers were wired to the same input frequency signal and powered from the same power supply. The films were suspended at the center of the skull phantom. Orange LEDs (Chanzon) with a diode antiparallel were attached to the films for wireless verification of the voltage generated by the films. For visualization purposes the skull top was removed to better photograph the LED.