Glassy Carbon Neural Interface for Chronic Epidural Stimulation in Rats With Cervical Spinal Cord Injury

There is growing evidence on the efficacy of electrical stimulation delivered via spinal neural interfaces to improve functional recovery following spinal cord injury. For such interfaces, carbon-based neural arrays are fast becoming recognized as a compelling material and platform due to biocompatibility and long-term electrochemical stability. Here, we introduce the design, fabrication, and in vivo characterization of a novel cervical epidural implant with carbon-based surface electrodes. Through finite element analysis and mechanical load tests, we demonstrated that the array could safely withstand loads applied to it during implantation and natural movement of the subject with minimal stress levels. Furthermore, the long-term in vivo performance of this neural array consisting of glassy carbon surface electrodes was investigated through acute and chronic spinal motor evoked potential recordings and electrode impedance tests in rats. We demonstrated stable stimulation performance for at least four weeks in a rat model of spinal cord injury. Lastly, we found that impedance measurements on all carbon-based spinal arrays were generally stable over time up to four weeks after implantation, with a slight increase in impedance in subsequent weeks when tested in spinally injured rats. Taken together, this study demonstrated the potential for carbon-based electrodes as a spinal neural interface to accelerate both mechanistic research and functional restoration in animal models of spinal cord injury.


I. INTRODUCTION
U PPER extremity function is the highest priority for functional restoration in individuals with cervical spinal cord injury (SCI) [1], [2]. Limited interventions, however, are available to improve physical impairments following SCI. Building on early work [3], [4], there is emerging evidence on the efficacy of spinal cord stimulation for upper extremity function following SCI [5], [6], [7], [8], [9]. The effect of cervical spinal cord stimulation has been examined through neurophysiological and histological assessments in rats and monkeys [10], [11], [12], [13], [14], [15]. Furthermore, the long-term implantation of an epidural stimulation electrode over the rodent cervical spinal cord has been demonstrated [16], [17]. Still, chronic stimulation of the mobile cervical spinal cord remains challenging due to the heterogeneous mechanical stress with physiological movement [18] and foreign body tissue reactions [19]. The introduction of new materials may be a key to further improving spinal implant function and longevity.
Over the past several years, carbon has emerged as a compelling material of choice for applications that require microelectrodes to interface with neural tissue for long periods. This is due to carbon's chemical inertness, biocompatibility, good electrical conductivity, high charge injection capacity, and excellent long-term electrochemical stability [20], [21], [22], [23], 24]. Carbon-based neural arrays have been used for neural recording [20], [21], [22], [23], [24], stimulation [25], [26], [27], magnetic resonance compatibility [28], and neurotransmitter detection [29], [30]. Further, carbon's multimodal capabilities allow not only electrical stimulation and sensing but also electrochemical recording. This is enabled by the fast diffusion of analytes to the surface of carbon and carbon's chemically active functional groups at edge planes and dangling bonds [29], [30]. Integrating carbon-based neural microelectrodes into cervical spinal implants has, in turn, raised significant interest in the wider use of carbon, predominantly for long-term applications for preclinical studies and future spinal implants for use in a clinical setting to improve function for treatment of neurological conditions [21].
To examine the efficacy of carbon-based electrodes for preclinical tools and future clinical use, it is vital to demonstrate that the electrodes on the surface of the cervical spinal cord can effectively activate the spinal cord circuits innervating upper extremity muscles in injured animals. Additionally, it is important to examine the capability of long-term functional stimulation performance in an animal model of injury. Here, we tested the performance of glassy carbon, one type of carbon, for use as epidural stimulation electrodes when fabricated on flexible polymer and implanted on the cervical spinal cord in a rat model of SCI. Electrophysiological testing and longterm (six weeks) in vivo studies were conducted to explore the functional stability of carbon-based microelectrodes.

A. Microfabrication of Spinal Array
We microfabricated a 10-channel spinal array specifically designed for epidural spinal cord stimulation. As shown in Figure 1, the array has a U-shaped geometry with maximum length of 65 mm with ten 250 μm diameter microelectrodes and 70 μm wide interconnects. We used a silicon on insulator wafer with 0.5 μm thick oxide layer on top of which ten μm thick SU8-10 negative photoresist (Microchem, MA, USA) was spin-coated at 2000 rpm. The SU8-10 layer was then patterned following a conventional negative lithography process of microelectrodes described elsewhere [25], [27].
As shown in Figure 1, the coated substrate is then softbaked at 65 • C for 10 min, 95 • C for 20 min, then UV exposed at ∼400 mJ cm −2 and post-baked at 65 • C for 1 min and 95 • C for 1 min. This was followed by the development of pyrolysis of the negative-tone resist at 1000 • C in an inert N 2 environment following protocols described elsewhere [25] resulting in glassy carbon microelectrodes with high graphitic content. Then a layer of photo-patternable polyimide HD 4100 (HD Microsystems, DE, USA) was spin-coated on top of glassy carbon microelectrodes at 2500 rpms for 45 seconds, soft-baked at 90 • C for 3 min and 120 • C for 3 min, then cooled to room temperature, and patterned through UV exposure at ∼400 mJ cm −2 . This was followed by partial curing at 300 • C for 60 min in a N 2 environment. Following this, metal traces were patterned using lift-off process with a sacrificial layer of NR9-1000PY photoresist (Futurrex, Inc, NJ, USA). A Ti adhesion layer of 20 nm and 200 nm Pt layer were deposited through sputtering. For electrical insulation, an additional 6 μm of polyimide HD4100 (300 rpms) was spun, patterned (400 mJ cm −2 ), and cured (350 • C for 90 min) in a N 2 environment. The device was then released from the wafer through selective etching of silicon dioxide layer with buffered hydrofluoric acid.
The U-shape geometry of the array and lead was selected to enable a longer arm that consists of a bump pad that allows placement over the head of the animal and connection to the recording unit through a printed circuit board (PCB). The U-shaped geometry was designed to be thin and flexible, to implant easily on the spinal cord under the vertebral lamina, and sufficiently robust to resist mechanical loads caused during implantation and animal movement [18].

B. Microfabrication of Spinal Array
The mechanical load carrying capacity of the arrays were tested through a tensile test using the INSTRON Universal system (INSTRON, MA, USA). A tensile load was applied until failure while measuring the array deformations. The experiment setup included small c-clamps that were attached between the jaws of upper and lower clamps of the Instron machine. The array was fixed to the clamps by applying polyimide tape to secure the array position and prevent the clamps from damaging the array surface. The linear displacement associated with the crosshead motion of the INSTRON system was used to get the displacement values. The crosshead displacement rate was set at 1 mm/min with a ramp load of 1 N/min.
We also built a finite element analysis (FEA) model to determine the stresses induced during surgical placement of the array and natural animal movement under chronic implantation. COMSOL Multiphysics FEM program (COMSOL AB, Sweden) was used for building a two-dimensional finite element model of the array consisting of 8,245 quadrilateral elements subjected to a 200 kdyne force (2 N) horizontal and vertical load. The load level is considered appropriate for in vivo surgical placement. The electrode component of the array was considered fixed, while other parts were considered free. The modulus of polyimide is 3.1 GPa.

C. In Vivo Validation of Neural Recording and Stimulation
The in vivo study investigated spinal cord stimulation performance and chronic electrode impedance of the glassy carbon electrodes through acute and chronic experiments in adult female Long-Evans rats (250g-300g; Charles River Laboratories, MA, USA). A total of nine adult female Long Evans rats were involved in this study. Five animals underwent acute surgery and spinal motor evoked potential (MEP) recordings (Group A, N = 5). One of five animals from Group A also had a chronic implantation during the same surgery and was monitored for electrode impedance for six weeks. Four additional animals received a contusion SCI and chronic implantation of carbon electrodes for longitudinal spinal MEP recordings and electrode impedance tests (Group B, N = 4). Thus, for the chronic monitoring of stimulation performance and electrode impedance, we had a total of five animals (N = 1 from Group A, N = 4 from Group B). All procedures were approved by the Institutional Animal Care and Use Committee of the University of Washington under protocol number 4265-01.

D. Surgical Procedure for Acute Implantation
Five glassy carbon electrode arrays (Devices 1-5) were evaluated by acute MEP recordings in five animals (Group A). The animals were anesthetized using 2-3% isoflurane mixed with oxygen. Body temperature was maintained on a heating pad set at 37 • C during and after the surgical procedure. The skin on the neck to upper back was shaved and cleaned three times with alcohol and betadine. The skin incision was made between C2 to T2 spinous processes, and the soft tissue was removed to expose the C3 to T1 vertebral bones (Figure 2a). Following a T1 unilateral laminectomy, the array that underwent hydrogen peroxide gas plasma sterilization was placed in the epidural space underneath the C3 to C7 lamina ( Figure 2b). The T1 laminectomy created access to the epidural space at C7. We delivered the probe at the caudal side of C7 via the T1 laminectomy site. The device was stiff enough to be slid into the epidural space at C7 to C3 lamina levels without any additional delivery device. The electrodes contacted approximately the C3 to C7 spinal segments. Subsequently, the caudal side of the array was sutured to the dura over the dorsal aspect of the right T1 segment (7-0 Polypro, Surgipro II, Covidien, Ireland).
The U-shaped array traces were placed near the spinal cord to provide strain relief. The array traces were anchored to the T2 spinous process with a 5-0 nylon suture (918B, Surgical Specialties Corporation, MA, USA) for increased stability. A common ground wire (AS631, Cooner wire, CA, USA) was placed in the paraspinal muscles at the C6 vertebral level. The array and common ground were soldered to a PCB for joining the wires to a connector before the surgery. As the final step, the head was fixed within a stereotaxic frame, and three screws were placed in the skull after a skin incision and soft tissue removal. The PCB was placed on the skull and stabilized by acrylic dental cement with cement glue preparation (C&B Metabond, Parkell Products Inc., NY, USA). Following this, we conducted the acute MEP recordings.

E. Surgical Procedure for Cervical Contusion Injury and Chronic Implantation
Four arrays (Device 6-9) were evaluated by chronic stimulation and evoked-response recordings in four animals (Group B). One animal with Device 2 from Group A also underwent chronic implantation with contusion injury surgery after the acute MEP recording. The animal with Device 2 received the contusion injury and chronic implantation immediately following the acute recording, so that the acute recording was conducted in the intact spinal cord, as was the case with all other animals in Group A. The Group B animals were anesthetized using 2-3% isoflurane mixed with oxygen. Body temperature was maintained on a heating pad set at 37 • C during and after the surgical procedure. The skin was prepared for incision as described above.
Following dissection of muscle layers and a C4 unilateral laminectomy, the animals received a right lateralized C4 contusion injury by applying a 200 kdyne force (2 N) using the Infinite Horizon Impactor (Precision Systems and Instrumentation, LLC., VA, USA). This hemi-contusive model of the cervical spinal cord provides consistent severe motor impairments in the upper limb [31], [32]. Subsequently, we performed the chronic implantation of the array in the same surgery, following the procedure for implantation as described above. The array and common ground wire were implanted at the C3 to C7 vertebral levels and in paraspinal muscles next to the C6 vertebral bone, respectively. The PCB connected to the array and common ground wire was placed on the skull. Acrylic cement and cement glue were used to fix the PCB to the skull, all incisions were closed for the muscle and connective tissue layers (4-0 PGA Vicryl, Ethicon, NJ, USA) and for the skin (4.0 Ethilon Nylon Suture, Ethicon, NJ, USA). Postoperatively, Buprenorphine (0.05 mg/kg) was administered twice per day for three days to manage possible pain.

F. Electrode Impedance and Spinal Motor Evoked Potential Recordings
All spinal MEP recordings and electrode impedance measurements were conducted under the same anesthesia, 2-3% isoflurane mixed with oxygen. Body temperature was maintained on a heating pad during recordings. The impedance of each electrode was measured during the implantation surgery, and during weeks 2, 4 and 6 post-surgeries. Impedance was measured at a frequency of 1 kHz (IMP-2A Single Channel, MicroProbes, MD, USA) in five animals (N = 1 from Group A, N = 4 from Group B).
For the electromyography (EMG) recordings, two wires (0.27 mm diameter, 2 mm de-insulated tip, AS631, Cooner wire, CA, USA) were subcutaneously implanted 1 mm apart in the right biceps and triceps muscles. The EMG wires were then connected to a multichannel data acquisition system (Tucker-Davis Technologies, FL, USA) through a active, unity-gain head-stage. The PCB with the implanted array was connected to an analog stimulus isolator (Model 4100 Isolated Pulse Stimulator, A-M System, WA, USA). Monopolar epidural stimulation between one selected epidural electrode and the common ground was delivered to evoke spinal MEPs. In each recording session, 15-20 single, bi-phasic, square-wave stimulation pulses were delivered at 500 μs cathodic charge followed by a 500 μs anodic charge using pulse amplitudes 0.125 mA − 2 mA with pulses delivered at 2 Hz. Stimulation was applied to the carbon epidural electrodes, and evoked responses from 2 ms to 30 ms after stimulation onset were recorded in the muscles.
The signals were amplified (1000×) using the Tucker Davis Technologies RZ5 system and digitized at 24.4 kHz and stored on a PC for offline analysis. Subsequently, raw signals were bandpass filtered (30-1000 Hz, 4th order Butterworth, zero-phase filter) using MATLAB R2021a (MathWorks, MA, USA). Baseline activity corresponding to 150 ms to 100 ms before the stimulation onset was taken as an estimate for the background noise. MEPs were analyzed offline using custom MATLAB scripts. A single trial of MEP was defined as evoked responses during the 30 ms time window after each stimulation event. The rectified evoked responses between 2 ms to 5 ms and responses between 5 ms to 8 ms after stimulation onset were considered an estimate for the direct motor response (early-latency response) and monosynaptic response (middlelatency response), respectively ( Figure 3) [33]. To compare the response patterns between the stimulation electrodes at C4 spinal segment and C7 spinal segment, the raw value of each area under the curve was normalized by the maximum mean area under the curve during the entire recording session (C3-C7 stimulation) in each muscle from each animal. This normalization was conducted for early-latency responses and middle-latency responses separately. We then tested the spatial selectivity using the normalized early-latency responses and middle-latency responses. Additionally, we assessed the normalized MEPs among animals without injury and animals with SCI at weeks 2 and 4 for the MEP changes in the postinjury condition.

G. Statistics
All in vivo impedance and MEP data are reported as mean ± standard error of mean. All in vivo data were assessed for normality with the Kolmogorov-Smirnov test. None of the data in this study were normally distributed. For the acute and chronic MEP recording analyses, differences between C4 stimulation and C7 stimulation, as well as between intact and post-injury conditions, were evaluated by Kruskal-Wallis test and Wilcoxon Rank-Sum test with Bonferroni correction. For the chronic impedance analysis, differences recorded on the same device over time were evaluated by Friedman's test and Wilcoxon Singed Rank test with Bonferroni correction.  All differences were evaluated using the adjusted p-value based on the Bonferroni method, such that p < 0.05 was divided by the number of comparisons before being considered significant. All analyses were performed with MATLAB R2021a.

III. RESULTS
Below we report the outcomes of the glassy carbon spinal stimulation arrays following mechanical tests, finite element modelling, electrical stimulation to evoke muscle contractions, and long-term stability of the carbon epidural electrodes.

A. Spinal Array Mechanical Characterization and FEA Model
The ultimate tensile strength of the array consisting of carbon-based microelectrodes and the supporting polyimide substrate was 129 MN/m 2 . The failure of the arrays occurred at a tensile load of 5.5 N at 7.7 mm extension. The elastic and plastic regions were well defined in the load-extension curve as shown in Figure 4. Failures were observed to always occur at locations where the long arms of the probes bend towards the short arms causing stress concentration.
The FEA simulation demonstrated that the maximum stresses under a vertical load of 2 N were less than the yield stress of 90 MPa for polyimide ( Figure 5). The array is notably stiff in the axial direction, as expected. However, the array is compliant in the horizontal direction due to low flexural rigidity and, also confirmed by the simulation results. In reality, there is significant lateral resistance from the tissues near the array and its true lateral stresses are much less than those shown in Figure 5, indicating that the array is mechanically sound for use in chronic applications. The model also confirmed that stress concentrations occur at the 'U-shaped' bent sections of the probe.

B. Acute in Vivo Spinal Motor Evoked Potential Recordings in Intact Animals
To test the stimulation performance of the carbon-based electrodes and spatial selectivity of stimulation at the different spinal segments, we conducted the in vivo spinal MEP recordings during acute surgeries. We removed electrode contacts with impedance value > 600 k at 1kHz frequency from all analyses due to likely mechanical or connection failure based on a previous study [34]. First, we measured spinal MEPs in five arrays (Device 1-5) in animals without SCI. The stimulation train through each available electrode elicited spinal MEPs recorded in the biceps and triceps muscles.
We analyzed both early-latency responses and middlelatency responses to verify the direct activation of motoneurons and the activation of spinal circuits via afferent fibers, respectively. As expected, spinal MEPs of biceps and triceps muscles exhibited different recruitment patterns with electrodes at different spinal segments before normalization as shown in Figure 6.
To investigate the spatial characteristics of rostral and caudal electrodes on the cervical spinal cord, we compared the normalized peak responses of spinal MEPs between C4 stimulation and C7 stimulation for all animals. Wilcoxon rank-sum test revealed a significant difference only in peak middle-latency responses of biceps muscles between stimulation at C4 spinal segment and at C7 spinal segment ( * p = 0.0036; Figure 7b Biceps). However, we could not detect other significant spatial selectivity between C4 spinal segment and C7 spinal segment when normalizing data and comparing across all animals.

C. Chronic in Vivo Spinal Motor Evoked Potential Recordings in Animals With SCI
We next evaluated the stimulation performance of electrodes at C4 and C7 spinal segments in the chronically implanted arrays in animals with cervical SCI. We compared the normalized peak responses of the spinal MEPs in intact animals to the spinal MEPs at two-and four-week post-SCI in five injured animals (Device 2 and Devices 6-9). All available electrodes of the chronically implanted devices produced spinal MEPs at least four weeks post implant.
In early-latency responses, there was a significant difference in the peak responses to C4 epidural stimulation in the triceps muscles between the intact animals and the animals with SCI (χ 2 (2) = 6.02, p = 0.0493, Kruskal-Wallis test). Wilcoxon rank-sum test revealed a significant reduction in peak early-latency responses of triceps muscles to C4 epidural stimulation in the injured animal at two-week post-SCI compared to in the intact animals ( * p = 0.0148; Figure 8a Right).
There was also a significant difference in the peak middlelatency response to C4 epidural stimulation (χ 2 (2) = 9.94, p = 0.0069, Kruskal-Wallis test) and C7 epidural stimulation (χ 2 (2) = 6.44, p = 0.0399, Kruskal-Wallis test) that appeared only in biceps muscles between the intact and injured animals. Wilcoxon rank-sum test only revealed a significant reduction in peak middle-latency responses of biceps muscle to C4 epidural stimulation in the injured animal at two-week postinjury compared to in the intact animals ( * * p = 0.0011; Figure 8b Left), while no significant difference was detected in response to C7 epidural stimulation by the post-hoc test.

IV. DISCUSSION
Here we presented the detailed fabrication and design of carbon-based arrays for cervical epidural stimulation in rats with spinal cord injury. The electrophysiological testing demonstrated the robust stimulation performance via spinal MEPs. The long-term implantation of the carbon-based arrays revealed the stability of the electrodes on the injured cervical spinal cord for at least four weeks.
The implanted carbon-based surface electrodes could activate the targeted upper limb muscles in rats. The acute spinal MEPs showed the trend of activating specific motor pools in the cervical spinal cord (Figure 6). In the middle-latency responses, we observed decreased magnitude of MEPs in higher intensity. As prior studies show, it is possible that the higher current intensity increases the direct activation of motoneurons [10], [33]. The increased direct motoneuron activation may lead to the decrease of capacity for  monosynaptic responses. Polysynaptic reflex via afferent fibers potentially also potentially recruits inhibitory interneurons [35] to suppress the monosynaptic responses. C4 epidural stimulation via the carbon-based surface electrodes evoked greater middle-latency responses of biceps muscles than C7 epidural stimulation [36]. However, we observed no significant differences in triceps middle-latency responses between C4 and C7 epidural stimulation in this study. Nonetheless, carbonbased electrodes could activate targeted cervical segments similar to a prior study [17]. Further work is needed to reveal the spatial resolution of cervical epidural stimulation.
Our implanted carbon-based arrays could maintain their stimulation performance for at least four weeks in rats with cervical SCI. The chronic spinal MEPs presented revealed consistent middle-latency responses in biceps and triceps muscles for four weeks after implantation. We observed decreased middle-latency responses to C4 stimulation in biceps muscles at Week 2, while there was no difference detected by C7 stimulation. Considering the C4 stimulation was applied directly over the contusion injury, the acute recovery processes like inflammation likely impact the spinal circuits around the lesion, as observed for tests of intraspinal stimulation following injury [37]. Similarly, another study found C6-C8 epidural stimulation below the C4 lesion could evoke similar or greater spinal MEPs in upper limb muscles compared to the pre-injury condition both one and ten weeks after C4 SCI in rats [12]. This is consistent with the spinal MEPs of C7 stimulation in the present study. These results suggest equivalent stimulation performance of the chronically implanted carbon-based arrays in a rat model of cervical SCI.
We observed stability of carbon-based arrays on the injured spinal cord for at least four weeks in a rat model of cervical SCI. The mechanical characterization of the arrays demonstrated the robustness of the devices as the force was applied and the array elongated 7.7 mm to reach its ultimate tensile strength of 129 ± 9 MPa, with a measured Young's modulus of 3.47 ± 0.35 GPa (n = 3). Together with the FEA simulation results, the device could withstand the forces applied during surgical and subsequent animal movement after implantation.
Previously, we have shown the stability of carbon-based arrays for long-term stimulation and recording of the brain [25], [27]. In this study, the electrode impedances were stable for at least four weeks after implantation on the injured spinal cord and showed a slight increase of electrode impedance by six weeks. This is similar to other multielectrode spinal implants in rats with cervical SCI, which demonstrate stable impedance for 4-6 weeks [38], [39]. The prior studies implanted electrodes below the injury on the lumbosacral spinal cord. Our implants were on and below the cervical spinal cord lesion. The direct contact on the injured C4 spinal cord may affect impedance due to mechanical compression from cord swelling or later biological responses such glial scar formation if not contained within the dura [19]. Nonetheless, carbon-based electrodes have shown high biocompatibility [25]. Other possible explanations for increased impedance include the malfunction of polyimide surface rather than the electrodes themselves, which occupy only about 10% of the implant. This is consistent with the results obtained in the mechanical characterizations as well as FEA modeling where mechanical failures were observed to always occur in the polyimide surface at locations where the long arms of the probes bend towards the short arms causing stress concentration. These locations are far from the electrodes; but high stress in polyimide surface could directly cause cracks in the metal traces connected to the electrodes themselves, thereby contributing to increase in impedances.
This study has several limitations that need to be addressed by further investigation. First, this report on the first carbonbased epidural spinal implant used a limited number of animals (N = 9). Second, we used a small electrode size (250 μm diameter) which could increase the impedance modulus at 1kHz. Other prior studies used larger diameter electrodes. Future work should evaluate larger diameter electrodes with effective charge delivery for stimulation. Furthermore, we did not evaluate the long-term biological interaction via histology or the post-explanation mechanical fatigue in the mobile cervical spinal cord inducing more applied strain compared to the brain and caudal spinal cord. As a prior study indicated [16], in vivo biological and mechanical changes can degrade the electrochemical properties of the electrode system.
Taken together, the stimulation performance and stability of glassy carbon epidural electrodes and array design may allow us to investigate underlying mechanisms of spinal cord stimulation for SCI through chronic implantation in preclinical models of SCI. Furthermore, glassy carbon -an important type of carbon -can provide other potential advantages for spinal neural interfaces. Based on studies in rats and monkeys [16], [17], it has been proposed that spinal implants need to be tailored based on 3D modeling from individual imaging for anatomical accuracy using materials such as platinum-silicone (Pt-PDMS) composite. Glassy carbon may expand the tailored design of implants with adaptable size, shape and number of electrodes [26]. Moreover, glassy carbon electrodes can offer quality recording and electrochemical detection functions on the spinal cord, potentially generating other applications for functional recovery following SCI [21].

V. CONCLUSION
In summary, this study provides the first evidence that glassy carbon epidural implants can reliably stimulate the cervical spinal cord during long-term implantations in rats with cervical SCI. This builds on previous results reported on the excellent electrochemical response, high signal-to-noise recordings, strong corrosion resistance, strong mechanical robustness at the interface of microelectrodes and the supporting substrates, stability under more than 3.5 billion cycles of stimulation, and multi-modality of glassy carbon electrodes [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30]. Therefore, a carbon-based platform for spinal cord stimulation may advance future development of preclinical experiments in rats with SCI and lead to more robust and versatile spinal implants for clinical use in the future.

AUTHORS CONTRIBUTIONS
Soshi Samejima designed and performed acute and chronic in vivo neural recording as well as stimulation experiments, designed and analyzed electrophysiological results, and contributed to writing all sections Brinda K. Cariappa, Rhea Montgomery-Walsh, and Rene Arvizu designed the architecture, fabricated the devices, implemented the microfabrication process, performed suite of electrical and electrochemical characterizations. Surabhi Nimbalkar did Raman and FTIR spectroscopy experiments, SEM imaging, coordinated for AFM and STEM imaging, and analyzed the results and contributed to writing the paper. Chet T. Moritz helped design and supervised in vivo tests and revised the paper. Abed Khorasani and Richard Henderson supported the electrophysiological experiments and revised the paper; Sam Kassegne formulated the array concept, supervised the project, structured the outline of the paper and contributed to discussion section of the paper.