- Email: [email protected]
A rat model for assessing the long-term safety and performance of peripheral nerve electrode arrays
Journal of Neuroscience Methods 328 (2019) 108437
Contents lists available at ScienceDirect
Journal of Neuroscience Methods journal homepage: www.elsevier.com/locate/jneumeth
A rat model for assessing the long-term safety and performance of peripheral nerve electrode arrays Benjamin Shafera, Cristin Welleb, Srikanth Vasudevana,
T
⁎
a U. S. Food and Drug Administration, Center for Devices and Radiological Health (CDRH), Office of Science and Engineering Laboratory (OSEL), Division of Biomedical Physics (DBP), Silver Spring, MD, USA b University of Colorado, Anschutz Medical Campus, Departments of Neurosurgery and Bioengineering, Aurora, CO, 80045, USA
A R T I C LE I N FO
A B S T R A C T
Keywords: Peripheral nerve interface Microelectrode arrays Utah electrode Sciatic nerve
Background: High-resolution peripheral nerve interfaces (PNIs) can provide amputees with intuitive motor control and sensory feedback. Current PNIs are limited by early device failure and suboptimal long-term stability. The present study aims to incorporate functional assessment into an in vivo test platform to assess the longterm safety and performance of PNIs for recording and stimulation. New methods: Utah electrode arrays (EA) were implanted in the rat sciatic nerve along with electromyography wires in the gastrocnemius and tibialis anterior. Cranial EEG screws were implanted in the somatosensory cortex for 12 weeks. Spontaneous neural activity was recorded using the implanted EA and stimulation-induced activity was monitored throughout the experiment. The impedance of each electrode was measured, and nerve function tests were conducted throughout the EA lifetime. Post-hoc safety assessments included scanning electron microscopy (SEM) of the EA and nerve histomorphometric analysis. Results: EA recordings were stable, and stimulation with EA elicited somatosensory evoked potentials and muscle contractions. Motor and sensory function tests indicated minimal deficits. Histomorphometric analysis indicated changes in nerve microstructure. SEM indicated EA-tip fracture, while lead wire breakage primarily caused device failure. Comparison with existing methods: We improved our prior platform with the addition of functional assessments of sensory pathways, a robust EMG array design to increase device longevity, and quantitative analysis of nerve microstructure. Conclusion: We present a test platform for long-term assessment of peripheral nerve interfaces for stimulation and recording. Using this platform, we demonstrate recording and stimulation with minimal impact on nerve function, while EA lead wire breakage and tip fracture could limit long-term device use.
1. Introduction In 2008, one in 190 Americans were living with limb loss, and this number is projected to increase substantially by 2050 (Shafer et al., 2018; Ziegler-Graham et al., 2008). Amputations in the lower limbs are mostly carried out due to the comorbidities associated with diabetes mellitus (Moxey et al., 2011), while trauma, malignancy, and infections are associated with upper limb amputations (Fahrenkopf et al., 2018). Between 2001 and 2017, 1705 service members of the U.S. armed forces sustained deployment related lower and upper limb amputations (Farrokhi et al., 2018). For many, the loss of a limb means inability to return to duty and increased comorbidities, which can have a negative impact on quality of life. To alleviate the burden associated with limb amputation, there is an increased need for advancing intuitive ⁎
neuroprosthetic devices for restoring lost sensory and motor function. Recently, researchers have focused their efforts on peripheral nerve interfaces (PNI) as an alternative to interfacing with the brain, exploring possible benefits such as: ease of surgical access, bidirectional communication (both sensory and motor pathways) at one interface site, persistence of functional neural connections post-amputation, increased selectivity for stimulating sensory feedback, and access to nerves at the time of amputation, reducing the need for additional surgeries (Dhillon et al., 2004; Ghafoor et al., 2017; Hong et al., 2018; Navarro et al., 2005; Rossini et al., 2011; Vasudevan et al., 2017). There are two main classifications of PNIs, extraneural and intraneural, signifying the level of invasiveness and, hence, selectivity for both recording and stimulation (Navarro et al., 2005; Gunter et al., 2019). Extraneural PNIs do not penetrate the nerve and are relatively
Corresponding author. E-mail address: [email protected] (S. Vasudevan).
https://doi.org/10.1016/j.jneumeth.2019.108437 Received 26 April 2019; Received in revised form 5 August 2019; Accepted 13 September 2019 Available online 14 September 2019 0165-0270/ Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
Fig. 1. Implant design. (a) 16 channel Blackrock EA with 55 mm long lead wire bundle and reference wires. (b) Custom EMG wire electrode with 4 bipolar insulated stainless-steel wires. (c) Bottom part of the 3D printed connector mount made with Nylon. (d) Top part of the 3D printed connector mount made with bronze infused stainless-steel. (e) Fully assembled connector mount. Scale bar 10 mm.
these devices for long-term use. A beneficial assessment platform would allow longitudinal assessment of: PNI impact on nerve function (motor and sensory), tissue response to implantation, selectivity of stimulation, stability of recording, and identifying failure modes early in the development cycle. While studies have completed these comprehensive assessments for intraneural PNIs used for recording (Vasudevan et al., 2017), there persists a need to expand the assessment of long-term safety and performance to PNIs used for both recording and stimulation. In this work, we have appended our previously published test platform (Vasudevan et al., 2017) to include the comprehensive assessment of the long-term safety and performance of PNIs used for stimulation, as well as recording, using a rat sciatic nerve model. We used Blackrock electrode arrays (EA) as the PNI implanted into the sciatic nerve, a custom-designed EMG wire electrode for recording compound muscle action potentials (cMAP) from gastrocnemius and tibialis anterior muscles, and stainless-steel cranial screws for recording somatosensory evoked potentials (SSEP). We developed a transcutaneous connector mount for accessing the EA and EMG wire electrodes after implantation. After the animals recovered from surgical procedures, we measured electrode impedance, recorded neural activity from awake behaving animals, studied stimulation evoked responses, and assessed motor and sensory function over a period of 12 weeks. At week 13, we harvested the nerve for histomorphometric analysis and studied
less invasive than intraneural PNIs. Extraneural PNIs have been approved for clinical applications to treat epilepsy and depression (Howland, 2014; Milby et al., 2008), and are currently being tested for restoration of function and neuroprosthetic applications (Christie et al., 2017; Graczyk et al., 2016; Polasek et al., 2006). Due to the presence of epineurium between the electrode and nerve fascicles, a limitation of extraneural PNIs is the inadequate spatial resolution (Rijnbeek et al., 2018). Intraneural PNIs are designed to penetrate the nerve for accessing individual fascicles and nerve fibers. Penetration increases recording signal quality, bypassing the less conductive connective tissue layers and decreasing the distance between the electrode tips and target axons (Ghafoor et al., 2017; Navarro et al., 2005). Intraneural PNIs have been shown to elicit specific sensations (Davis et al., 2016; Dhillon and Horch, 2005; Oddo et al., 2016), as well as record neural activity for motor control (Rossini et al., 2010; Wendelken et al., 2017). However, intraneural PNIs struggle with long-term stability caused by lead wire breakage, electrode migration, and foreign body response (Vasudevan et al., 2017; Gunter et al., 2019; Straka et al., 2018). Nevertheless, the higher resolution bidirectional communication offered by intraneural PNIs is favorable for advancing neuroprosthetic devices. As the research and clinical need involved with PNI technologies continues to rise, it is increasingly important to develop and improve experimental test platforms to assess the safety and performance of 2
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
pneumatic impactor (Blackrock Microsystems, Inc., U.S.A.). EA insertion was achieved by applying ∼5-6 gentle taps using the impactor set at 10 psi (Fig. 2b). In some cases, a few electrode shanks were found to be outside of the nerve due to larger dimensions of the EA. After EA insertion, a few drops of fibrin sealant (TISSEEL, Baxter Healthcare Corporation, U.S.A.) was applied to the nerve and electrode construct for stability. The silicone block and Parafilm were removed once the fibrin sealant cured, and a silicone cuff (4 mm length and 3.2 mm inner diameter) was placed around the nerve and electrode construct. The silicone cuff lumen was filled with fibrin sealant as shown in Fig. 2c. After confirming that the fibrin sealant cured completely, the muscle incision over the nerve was closed and reference wires were delaminated ∼0.5 mm using a lighter and secured inside the muscle over the nerve. For implanting the EMG wire electrodes, gastrocnemius and tibial anterior muscles were exposed through a skin incision. EMG wires were then tunneled under the skin from the connector mount site to the implant site. Using a 25 G needle serving as a canula, two pairs of EMG wires were interested into each of the two muscles for redundancy and were secured in place with sutures (4-0 Prolene, Oasis, U.S.A.) as shown in Fig. 2d. Skin incisions over the nerve and muscle implant sites were closed with a combination of sutures and gluture as shown in Fig. 2e. To record somatosensory evoked potentials (SSEP), stainless-steel screws (0.2 mm × 3 mm, M0.8, US Micro Screw, U.S.A.) were used as electrodes. The skull was exposed, and burr holes were carefully drilled over the left and right somatosensory cortex. During the process, it was ensured that the burr holes did not pass through the skull. Screws were inserted into the burr holes (Fig. 2f) and secured using dental acrylic (Jet Denture Repair Powder, Lang Dental Manufacturing Company, Inc., U.S.A.). Finally, animals were administered Atipamezole (0.5 mg/ kg, intraperitoneal) for anesthesia reversal, Meloxicam (2 mg/kg, subcutaneous) for analgesia, and Gentamycin (8 mg/kg, subcutaneous) for antibiotic treatment. Topical antibiotic ointment was applied to all incisions after the surgery. Meloxicam (1 mg/kg, subcutaneous) was administered for another two days post-surgery for pain control.
mechanical damage to the EA using scanning electron microscopy (SEM). Our results showed that the current test platform was conducive for assessing the long-term safety and performance of PNIs used for recording and stimulation. 2. Methods 2.1. Animal protocol and experimental design This study was approved by the Institutional Animal Care and Use Committee (IACUC) at the Food and Drug Administration, White Oak campus. Female Lewis rats were purchased from Charles River Laboratories International Inc., for experiments described in this manuscript. Experiments were conducted on animals weighing between 200–280 g at the time of surgery. All animals were individually housed in plastic cages with 12 -h light and dark cycle before and after experiments. Animals were randomly assigned to one of two groups: control, n = 6, and EAs, n = 6. The control group animals were not subjected to surgical procedures. 2.2. Implant design Sixteen channel high-density Blackrock EAs (Blackrock Microsystems LLC, U.S.A.) with shanks (1 mm in length with a 0.4 mm pitch) arranged in a 4 × 4 configuration were used in this study. The EA and the connector (A73098-001, Omnetics Connector Corporation, U.S.A.) were linked via a 55 mm long gold lead wire (Fig. 1a). One of the reference wires served as the ground for recording and stimulation. Additionally, a custom-designed EMG wire electrode (Microprobes for Life Science, U.S.A.) with 8 insulated stainless steel wires of 110 mm length (0.1 mm diameter, arranged into 4 bipolar pairs) attached to a connector (A79038-001, Omnetics Connector Corporation, U.S.A.) was used to record muscle activity (Fig. 1b). The tip of the bipolar pair was de-insulated by the manufacturer, with greater insulation removed for one electrode per pair to allow it to serve as the reference. Two of six animals received connector mount designed using SolidWorks (Dassault Systèmes SolidWorks Corp., France) and 3D printed with Stainless Steel infused with bronze (Shapeways, Inc., U.S.A.). The other 4 animals received a two-part connector mount, top made with Stainless Steel infused with bronze and bottom part made with Nylon as shown in Fig. 1c and d (Straka et al., 2018). Connectors from the EA and EMG wire electrode were housed inside the connector mount and a Mersilene mesh (2” x 2”, Ethicon, Inc., U.S.A.) was attached to the bottom of the connector mount using epoxy (Loctite Hysol Epoxy, Henkel Corporation, U.S.A.) as shown in Fig. 1e. Kwik-cast (World Precision Instruments, Inc., U.S.A.) was applied to the bottom portion of the connector mount to provide softer tissue contacting surface.
2.4. Electrical characteristics Electrochemical Impedance Spectroscopy (EIS) was used to measure broadband impedance (1 Hz-1 MHz, Gamry Instruments Inc., U.S.A.) of the individual channels pre- and post-implantation. Pre-implant measurements were obtained by immersing the electrodes in phosphate buffered solution (PBS) and measuring the impedance between one of the reference wires and the channel of interest. Measurements were obtained immediately after implantation, again at 2 weeks following implantation, and weekly thereafter until device failure. The channels with an impedance ranging between 100 kΩ and 800 kΩ is considered functional as per the manufacturer. Therefore, a functional impedance threshold of 1 MΩ (at 1 kHz) was decided to accommodate for variability in measurement technique. Channels with impedance > 1 MΩ were considered non-functional, and EA with all the channels > 1 MΩ were considered broken (Fig. 3b). Detailed analysis of EIS data was published previously (Straka et al., 2018) and is not discussed in this manuscript.
2.3. Surgical procedure All surgical procedures were performed under aseptic conditions. Intraperitoneal injection of cocktail made with ketamine (75 mg/kg) and dexmedetomidine (0.25 mg/kg) was used to deeply anesthetize the rats. Adequate anesthetization was confirmed by the loss of toe pinch reflex. The surgical site was shaved and sterilized using alcohol and iodine. Ocular ointment was applied to each eye before and during surgery to prevent drying. Sterile warm saline (3 ml) was subcutaneously injected to ensure adequate hydration. While the electrode implantation procedure described here can be implemented on either hind limb, we used the right hind limb in this study. After securing the connector mount to the lumbar fascia (Fig. 2a), the right sciatic nerve was exposed using blunt dissection technique as described previously (Vasudevan et al., 2017, 2013). A piece of silicone block (∼10 × 8 × 1 mm, Sylgard, Dow Corning, U.S.A.) and small sheet of Parafilm were placed under the nerve to prevent it from moving downward during EA implantation using the
2.5. Electrophysiological recording of spontaneous neural activity Starting at two weeks, weekly spontaneous neural activity recordings were acquired. Animals were briefly anesthetized with 3% isoflurane and then connected to the recording system (Neuralynx Inc., U.S.A.). Following recovery from anesthesia, spontaneous neural activity recorded by the EA was acquired for 15 min during free exploration of the home cage. Spike events from recordings were extracted and analyzed using Neuraview and SpikeSort 3D (Neuralynx, U.S.A.) to confirm the presence of neural activity in individual channels. Automatic spike sorting algorithms in SpikeSort 3D (KlustaKwik) 3
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
Fig. 2. Surgical procedure. (a) Connector mount implanted over the lumbar fascia. (b) Electrode array inserted into the sciatic nerve. (c) Silicone cuff applied around the nerve and electrode array, which is then filled with fibrin sealant (cloudy inside the tube). (d) EMG wire arrays implanted into the gastrocnemius and tibialis anterior muscles. (e) Skin incisions closed with sutures and gluture. (f) Stainless-steel screws implanted over the somatosensory cortex.
were first used, followed by manual refinement to identify neural activity as described elsewhere (Vasudevan et al., 2017). For each channel with neural activity, clusters with baseline noise and neural activity were identified as shown in Fig. 3a. The signal-to-noise ratio (SNR) was calculated by taking a ratio of the average signal peak-to-peak and average baseline noise peak-to-peak. For channels with multiple neural waveform clusters, the cluster with the highest average peak-to-peak was used for SNR calculations. Data analysis was limited to seven weeks to include neural activity from all six animals at each time point. Data between time points were compared using one-way analysis of variance (ANOVA, Prism 8, GraphPad Software, U.S.A.) (Fig. 3c).
only one channel that induced muscle activation and/or SSEP on the first day of stimulation, that channel along with a randomly selected channel was consistently tested for the entire course of the experiment. If an array did not have any channels that induced muscle activation and/or SSEP on the first day of stimulation, two channels were randomly selected and consistently tested for the entire course of the experiment. Twenty-five anodic-first charge balanced square wave pulses with 12.5 μA per phase, 100 μs per phase were applied at 3.1 Hz and the corresponding SSEP and cMAP signals were recorded for analysis. In total, 12 individual channels from all 6 EAs were used to analyze SSEP and cMAP recordings to calculate the percentage of active channels (number of channels with SSEP or cMAP/12 × 100%). SSEP signals were measured for 100 ms after each stimulation pulse. An average of 25 stimulation induced SSEPs was calculated from both the left and right-side cranial electrodes, and then re-referenced by subtracting the right from the left SSEP (Fig. 3d). The re-referenced signal was then zero-averaged and analyzed in two specific time regions associated with SSEPs in our study. cMAP signals were processed similar to SSEP and identified using an average of 25 signals 100 ms in duration starting 5 ms after the EA stimulation trigger.
2.6. Electrical stimulation to drive somatosensory evoked potentials and compound muscle action potentials Somatosensory evoked potentials (SSEP) were recorded from cranial screws over the left and right somatosensory cortex. Baseline (no stimulation) recordings and stimulation-evoked SSEP were acquired each week beginning at two weeks post-implantation (RZ5D processor, PZ2 preamplifier, ZC16 headstage, Tucker-Davis Technologies, U.S.A.). Recordings were acquired with a single-ended configuration and shared a common reference electrode (18 G subcutaneous needle temporarily inserted over the neck muscle) at 3 kHz sampling frequency (Fisher et al. (2016)). Compound muscle action potentials (cMAP) were recorded simultaneously with SSEP recordings using the same system. Recordings were acquired with a single-ended configuration from all bipolar EMG wires with respect to the shared common reference electrode at 3 kHz sampling frequency. Two channels for each EA was selected for applying electrical stimulation (one channel at a time) between the channel and the EA reference wire. On the first day of stimulation trial, all 16 channels were tested individually for muscle activation and/or SSEP. Two channels that induced muscle activation and/or SSEP were selected and consistently tested for the entire course of the experiment. If an array had
2.7. Functional assessment Walking Track analysis was used to measure motor function at baseline, 5 days post-implantation (1 W), 12 days post-implantation (2 W) and biweekly until week 12. Animals were confined to 8 × 40 cm path and foot print videos were recorded with a camera placed underneath the platform (FreeWalk, Cleversys Inc., U.S.A.). Five sets of footprints were extracted from the videos and analyzed using ImageJ to calculate print length, toe spread, and intermediary toe spread. The Sciatic Functional Index (SFI) (Fig. 4a) and Tibial Functional Index (TFI) (Fig. 4b) were calculated using established methods and the average was used for data analysis (Vasudevan et al., 2017; Bain et al., 1989; Vasudevan et al., 2019). 4
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
Fig. 3. Electrophysiology. (a) Isolated neural activity (green) and background noise (pink). (b) Number of functional devices and percentage of non-functional channels based on impedance threshold of 1MΩ. (c) Average SNR values of identified neural activity in all EAs up to 7 weeks. (d) Example of SSEP waveform with a negative peak followed by a positive peak, identified with a threshold ± 2 STD of the baseline noise, respectively. Note that the plot is inverted following the conventions of clinical neurophysiology field. (e) Average SSEP peak-to-peak amplitude and percentage of active channels eliciting SSEP during simulation. (f) SSEP latency measured as the time from stimulation onset to the time of each peak. (g) Example cMAP waveform with two peaks identified with a threshold ± 3 STD of the baseline noise. (h) Average gastrocnemius cMAP peak-to-peak amplitude and percentage of active channels eliciting cMAP in response to EA simulation. (i) Average tibialis anterior cMAP peak-to-peak amplitude and percentage of active channels eliciting cMAP in response to EA simulation.
followed by Tukey’s multiple comparison test and between group time points were compared using two-way ANOVA followed by Bonferroni’s multiple comparison test (Prism 8, GraphPad Software, U.S.A.).
Von Frey test was used to assess sensory function. The animal was placed in a cage (25.4 × 25.4 cm) with a perforated floor and allowed to acclimate for 5 min before commencing measurements. With the animals in stationary position, a rigid tip of the Von Frey anesthesiometer (IITC Life Science, U.S.A.) was pressed against the central region of the right plantar foot. Response to mechanical allodynia was recorded (in grams) for 5 consecutive trials, and the average was used for further analysis (Fig. 4c) (Vasudevan et al., 2017, 2019). For walking track analysis and Von Frey test, data within groups (control and electrode array) were compared using two-way ANOVA
2.8. Semiautomated quantitative histomorphometry All animals were sacrificed at 13 weeks post-implantation by first administering ketamine/dexmedetomidine anesthesia cocktail followed by pentobarbital (200 mg/kg, intraperitoneal) injection. Nerve segment ∼1 cm distal to the EA implant site was carefully dissected and
Fig. 4. Function. (a) Sciatic Functional Index (SFI) of control and EA implanted animals. (b) Tibial Functional Index (TFI) of control and EA implanted animals. (c) Von Frey test of control and EA implanted animals. Data shown as mean ± S.E. 5
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
Fig. 5. Histomorphometry. (a) Light microscopy image of nerve cross-section obtained from one animal in the EA group. (b) Nerve fiber density comparison between EA and control group. (c) Myelin area comparison of EA and control group. (d) Comparison of myelinated fiber distribution between EA and control group. Data shown as mean ± S.E. *p ≤ 0.05.
Fig. 6. Scanning electron microscopy. (a) SEM image of 16 channel EA post-harvest. (b) Fractured electrode tip. (c) Intact electrode tip.
immersion fixed in 3% glutaraldehyde at 4 °C. Nerve samples (n = 3 control, n = 6 treatment) were processed for epoxy embedding. Samples were post-fixed in 1% osmium tetroxide and subjected to serial dehydration and embedding in Araldite 502 (Nerves Incorporated, U.S.A.). The epoxy embedded samples were cut into semi-thin sections
and stained with 1% toluidine blue dye for histomorphometric analysis in the Laboratory of Dr. David Brogan, Department of Orthopedic Surgery, Washington University School of Medicine (Fig. 5a). Histomorphometry of each nerve section was performed using a digital image analysis system with linked morphometry software (Leco
6
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
throughout the study to avoid unintended electrode damage. An example of SSEP waveform is shown in Fig. 3d. The time region between 8–15 ms post-stimulation was observed with a negative peak and the region between 18–23 ms, was observed with a positive peak. Peaks within each region were noted if they exceeded +/- 2 standard deviations of the baseline re-referenced recording (Fig. 3d) and manually examined to confirm the presence of an SSEP signal. Peak-to-peak calculations were made based on the noted peaks (Fig. 3e). The SSEP latency was measured for the negative and positive peaks as the time from stimulation until each peak (Fig. 3f). In cases where there was only one peak during a specific recording, the average of the signal within the missing peak’s time region was taken and used for the peakto-peak calculation. The average SSEP peak-to-peak amplitude decreased over time after reaching a peak around 4–5 W post-implantation (Fig. 3e). A maximum of 50% channels elicited SSEP across all EAs. SSEP latency was analyzed by measuring the time to peak of the negative and positive deflections from stimulation onset. The latency of both peaks remained consistent through the 7 weeks of stimulation (Fig. 3f). cMAP recordings were analyzed to assess muscle activation in response to EA stimulation. Stimulation waveforms were designed to prevent EA damage, so they did not elicit maximal contraction of the muscles. An example cMAP waveform is shown in Fig. 3g. The stimulation artifact was apparent in most signals (Fig. 3g). After the stimulation artifact, the cMAP was characterized with two peaks. At least one needed to be greater than -/+ 3 STD of the baseline noise for analysis (Fig. 3g). Once identified, the amplitude difference between the two peaks were measured as the peak-to-peak amplitude of the cMAP for gastrocnemius (Fig. 3h) and tibialis anterior (Fig. 3i) muscles. If only one peak was present, the amplitude of that peak along with the average of the signal within missing peak’s time region was used as the peak-to-peak amplitude. For target muscles (gastrocnemius and tibialis anterior), cMAPs were observed in some of the channels in response to stimulation of the nerve (Fig. 3h and 3i). The highest gastrocnemius peak-to-peak amplitude occurred at 5 W, while the highest tibialis anterior peak-to-peak amplitude occurred at 7 W. Lowest peak-to-peak values for gastrocnemius was observed between 2 and 4 weeks. While the lowest peak-to-peak average amplitude for tibialis anterior was observed between 2 and 6 weeks.
Instruments, U.S.A.) as previously described by Hunter et al. (Hunter et al. (2007)). Multiple field images were acquired to calculate fiber density (Fig. 5b), myelin area (Fig. 5c) and myelinated fiber distribution (Fig. 5d). Myelin area and fiber distribution data were compared with unpaired t-test, and multiple t-tests were used to compare myelinated fiber distribution (Prism 8, GraphPad Software, U.S.A.). 2.9. Scanning electron microscopy The nerve along with the EA was immersion fixed in 4% paraformaldehyde solution at 4 °C overnight. EA was carefully extracted from the nerve using microdissection methods and immersed in a deionized water solution containing 1% Tergazyme (Alconox, USA), 0.5% Triton X-100 for 3 days to remove biological debris. The EA was then washed in de-ionized water and air dried for at least 24 h before subjecting to scanning electron microscopy (SEM) imaging. SEM images (Jeol 6390 L V, USA) were acquired under low vacuum mode (15 kV, 30 Pa) to assess the integrity of EA tips (Fig. 6) (Vasudevan et al., 2017). 3. Results 3.1. Implant design and surgical procedure The ability to perform long-term assessment of PNIs in animal models depends on stable implantation of both the implanted electrodes and the connectors (Vasudevan et al., 2017). Therefore, reliable connector mount design is critical for long-term recording or stimulation studies. The two-part connector mount setup was assembled as shown in Fig. 1e, with the tissue-contacting surface made suitable for reducing animal discomfort. The metal top prevents damage to the connectors inflicted by animal chewing or mechanical damage inside the home cage. The Mersilene mesh attached to the bottom of the connector mount as shown in Fig. 1e encourages tissue ingrowth, which is critical for long-term stability. The assembled connector mount with housed electrode connectors shown in Fig. 1e allowed for easy and reliable access to the implanted electrodes over long periods of time. 3.2. Impedance measurements Detailed analysis of EIS data has been previously reported (Straka et al., 2018). Here, we used the data to determine if the devices were functional. EAs were considered non-functional when all the channels had impedance measurements above 1 MΩ at 1 kHz. Device failure was mainly the result of lead wire breakage at the connector mount interface. All devices were functional until 7 W post-implantation (Fig. 3b). For functional devices, greater than 75% of channels were within functional range. Overall, 50% of the tested EAs had functional channels for the 12 W experimentation period and greater than half of the channels were below 1 MΩ at 12 W (Fig. 3b).
3.5. Functional assessment Walking track analysis was used to assess motor function in the sciatic nerve (Fig. 4a) and the tibial nerve (Fig. 4b) in response to surgery and experimental procedures. Both SFI and TFI showed a trend towards deficits at 1 W and recovered close to pre-implant levels around 2 W. None of the time points were significantly different, indicating that the surgical procedure or the presence of EA in the nerve did not alter the motor function significantly. Presence of the EA also did not impede with recovery of motor function. Von Frey test was used to assess alteration in sensory function over time. There was no significant difference between the control and EA group, indicating that neither the surgical procedure nor the presence of the EA led to significant loss in sensory function or increase in sensitivity (Fig. 4c).
3.3. Electrophysiological recording assessment Spontaneous action potentials from the sciatic nerve were recorded for 15 min each week starting at week 2, up to 12 weeks or device failure, while the animals moved freely inside the home cage. Offline analysis of the recordings was conducted to extract single units using SpikeSort3D. Data analysis was performed up to 7 weeks to include all 6 devices, after which device failure was observed (Fig. 3b). An example of a neural activity is shown in Fig. 3a. The average SNR of recorded units was ∼4 at 2 W and did not substantially change from this value for up to 7 weeks of analysis (Fig. 3c).
3.6. Semiautomated quantitative histomorphometry Histomorphometry was used to study the effects of EA implantation on myelinated fibers at 13 weeks post-implantation (Fig. 5a). The fiber density was comparable between control and EA group (Fig. 5b), while the myelin area was lower in the EA group compared to control (Fig. 5c). Distribution of the myelinated fibers show that there is a significant difference in distribution of 4–6 μm fibers between EA and control groups (Fig. 5d). Overall, histomorphometry points to changes in nerve microstructure because of EA implantation and experimental procedures.
3.4. Electrophysiological stimulation performance SSEP and cMAP recordings were used to assess stimulation performance of the implanted EA. The stimulation parameters were fixed 7
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
surgical procedure was used to verify if damage occurred anytime during the surgery. The impedance for most of the electrodes did not increase beyond the functional threshold of 1 MΩ (Straka et al., 2018), suggesting that the implantation procedure did not cause lead wire breakage. Impedance could be characterized for all electrode tips over the course of the experiment unless lead wire breakage occurred. Wire breakage was indicated by simultaneous impedance increase above 1 MΩ for all electrode channels. Beyond lead wire breakage, abiotic and biotic factors could affect intraneural PNI impedance in the long term (Vasudevan et al., 2017; Straka et al., 2018). Broadband EIS measurements can potentially reveal failure modes beyond simple mechanical failure (Straka et al., 2018). In combination with characterization of the abiotic and biotic factors influencing an implanted PNI, chronic EIS measurement through this platform gives valuable feedback for electrode material selection and fabrication. Electrophysiological recording in awake behaving rats was made possible with the stable connector interface that allowed experimenters to easily and safely connect to the electrodes, minimizing the duration of anesthetic use and mechanical damage at the connector mount interface. The ability to record from the EA in freely-moving animals is important to assess the reliability of recorded signals without confounds, such as the influence of anesthesia on nerve conduction. In addition, multiple individual neural units could be isolated from the EA channels demonstrating the high fidelity of intraneural PNIs for advancing motor control using neuroprosthetic devices signals (Davis et al., 2016; Clark et al., 2011). The average SNR for these detected action potentials was ∼4 up to 7 W, suggesting applicability to pattern recognition for neuroprosthetic control applications (Shafer et al., 2018). SSEP and cMAP evaluation in animal models can be used to characterize the potential of a PNI to be used for closed-loop neuroprosthetic applications. The implanted EA displayed the ability to produce cMAPs and SSEPs in the long-term, but this was not consistent between time points. This inconsistency could be a result of micromotion inside the nerve, as the electrode tip distance from axons within the nerve could change. Gastrocnemius cMAP peak-to-peak values peaked at 5 W and decreased with time (Fig. 3h) while Tibialis Anterior peak-to-peak values increased with time and peaked around 7 W (Fig. 3i). This activation pattern could be attributed to the change in relative distance between the electrode tips with respect the nerve fibers innervating either of the two muscles, along with biotic factors such as fibrotic encapsulation at the electrode tip. Similar to the gastrocnemius cMAP signals, the peak-to-peak amplitudes of SSEPs reached its highest value around 4 W and then trended downward (Fig. 3e). Micromotion of the EA inside the nerve could result in the stimulation of different fibers leading to inconsistent cMAP and SSEP stimulation or decreased amplitudes of stimulation if the electrodes were to move farther from a previously stimulated fiber over time (Davis et al., 2016). While micromotion is a noted limitation of microelectrode arrays (Ghafoor et al., 2017), the variety of stimulated fiber types throughout the duration of the experiment reveals the potential of intraneural PNIs to interact with various fibers if electrode tips were placed intentionally. Stimulation of sensory nerves in humans using intraneural PNIs elicit a variety of sensations and intensities noted consciously by users (Davis et al., 2016; Raspopovic et al., 2014). The foreign body response could also influence the stimulation and recording performance of intraneural PNIs (Christensen et al., 2014). Changes in the PNI materials may show promise for mitigating these effects but more development is needed (Wurth et al., 2017). The timing of the SSEP peaks did not change with time (Fig. 3f), indicating stable physiological response to stimulation. The potential for sensory feedback from neuroprostheses is apparent but further development of technologies are needed to address stability on the microscopic scale. Overall, micromotion and changes at the electrode interface due to biotic and abiotic factors may have a strong effect on the selectivity of neural stimulation over long-term for these EAs.
3.7. Scanning electron microscopy Mechanical failure of the EA tips after extraction was studied using SEM (Fig. 6a). Some of the electrode tips were observed to be fractured (Fig. 6b), which could be a result of implantation. A high magnification of intact electrode tip is shown in Fig. 6c. Four out of the six EAs were observed with a small number of fractured electrode tips. 4. Discussion The need to improve intuitive prosthesis control is leading to innovation in PNI design and implementation (Ghafoor et al., 2017; Jung et al., 2018). Development of chronic animal models to assess the safety and performance of PNIs is critical for determining failure modes at an early stage of device development and for successful clinical translation of these new technologies. We previously published a rodent model to assess the safety and performance of recording electrodes (Vasudevan et al., 2017), which laid the foundation for the current study. The rat model has been extensively studied in the field of peripheral nerve repair and has well characterized functional and histological outcome measures (Vasudevan et al., 2017, 2013; Hunter et al., 2007; Deuis et al., 2017; Sarikcioglu et al., 2009). These measures for non-clinical assessment can be used to evaluate longitudinal performance of PNIs, and to compare across PNIs of different form factors or modalities. This model also opens the door for understanding histological, electrophysiological and material failure modes, which can be rectified before finalizing device design for clinical translation. To demonstrate the ability to perform long-term assessment, we designed a robust connector mount system as shown in Fig. 1, that allowed for stable access to the implants during the study. This component is critical to maintain the integrity of the lead wires as they undergo repeated mechanical stress during chronic testing (Vasudevan et al., 2017; Straka et al., 2018). In four out of six animals, the bottom part of the connector mount, in contact with tissue, was 3D printed with Nylon, and further softened with Kwik-cast to mitigate potential discomfort to the animal. The top part of the connector mount was 3D printed with bronze infused stainless steel to prevent damage to the connectors inflicted by the animal, cage components or user handling. Finally, the use of Mersilene mesh to support tissue ingrowth after implantation contributed to wound healing and stability (Vasudevan et al., 2017; Straka et al., 2018). The surgical method described here was previously developed in our lab to study floating microelectrode arrays and accounted for device integrity from the transcutaneous mount down to the PNI (Vasudevan et al., 2017). The lumbar fascia is used secure the connector mount, while providing easy access to the implanted electrodes through connectors (Fig. 2a and e). Accessing the two implants (EA and EMG) over long-term via the connector mount did not produce infections at the surgical sites, which is a major concern involving transcutaneous implants (Chou et al., 2010). Critical to the success of this procedure is securing the electrodes in the nerve as seen in Fig. 2c using fibrin sealant. The fibrin sealant forms a stable construct immediately upon application, preventing the EA from dislodging. At later points in the healing process, electrode stabilization is augmented by fibrotic encapsulation, and presumably limits electrode dislodgment during locomotion (Vasudevan et al., 2017; Straka et al., 2018). The EMG wire electrode fixation technique (Fig. 2d) was reliable, as demonstrated by the stability in the location of the recording tips inside the muscles during the experiments. The most common failure observed with the EA was lead wire failure at the connector mount interface, which is a region exposed to high mechanical stress. Lead wire breakage continues to be a hurdle to chronic experimentation in both animal and humans (Vasudevan et al., 2017; Davis et al., 2016). Nevertheless, surgical procedures described here provides implant stability for long-term assessment of PNIs in the sciatic nerve using a rodent model. Impedance of the nerve electrodes at 1 kHz before and after the 8
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
their assistance with histomorphometric analysis.
Functional testing allowed the quantification of both sensory and motor function of the target nerve. We conducted walking track analysis to monitor motor function and Von Frey testing to monitor sensory sensation. Results show that motor and sensory function did not change significantly and returned to pre-implantation levels within 2–4 W postimplantation (Fig. 4a–c). These results are necessary to conclude if a device causes significant nerve damage making the PNI unsuitable for long-term use. It is also important to note that the surgical procedures did not cause irreversible target nerve damage. In addition, it seems that electrical stimulation did not impact function over time, while the parameters and duration of stimulation used in this study could contribute to this observation. Although this experiment was not specifically designed to assess functional effects of stimulation, future work using this platform could isolate the relationship between stimulation and nerve function. Histomorphometric analysis is commonly used to assess the health and distribution of myelinated nerve fibers (Hunter et al., 2007). Histomorphometric analysis of the distal nerve segment showed comparable fiber density, while there was significant difference in myelin area and distribution of fibers between 4–6 μm fibers (Fig. 5). This difference in myelin area and fiber distribution could be a result of either the surgical procedures and/or the presence of EA inside the nerve. While there may have been axon damage during implantation, histomorphometric analysis at 13 weeks post-implantation shows that the presence of EA did not prevent the recovery of fiber density. It is important for an EA to have minimal nerve injury during surgery, but it is even more important for it to not induce complications associated with obstruction of regenerating nerve fibers. SEM can be used to evaluate factors leading to electrode failure (Vasudevan et al., 2017). SEM of the explanted electrodes revealed mechanically damaged electrode tips (Fig. 6b), which could be attributed to the implantation procedures. Breakage at the tip can alter device recording and stimulation performance, rendering the use of a PNI to be suboptimal for long-term use. Changes in materials specifically designed for peripheral nerve applications and refining surgical procedures could potentially mitigate this issue. Overall, the comprehensive test platform described here has allowed for detailed characterization of electrode safety and performance, which can provide information to steer future improvements to PNI design and fabrication.
References Bain, J.R., Mackinnon, S.E., Hunter, D.A., 1989. Functional evaluation of complete sciatic, peroneal, and posterior tibial nerve lesions in the rat. Plast. Reconstr. Surg. 83, 129–138. Chou, T.G., Petti, C.A., Szakacs, J., Bloebaum, R.D., 2010. Evaluating antimicrobials and implant materials for infection prevention around transcutaneous osseointegrated implants in a rabbit model. J. Biomed. Mater. Res. A 92, 942–952. Christensen, M.B., Pearce, S.M., Ledbetter, N.M., Warren, D.J., Clark, G.A., Tresco, P.A., 2014. The foreign body response to the Utah Slant Electrode Array in the cat sciatic nerve. Acta Biomater. 10, 4650–4660. Christie, B.P., Freeberg, M., Memberg, W.D., Pinault, G.J.C., Hoyen, H.A., Tyler, D.J., Triolo, R.J., 2017. Long-term stability of stimulating spiral nerve cuff electrodes on human peripheral nerves. J. Neuroeng. Rehabil. 14, 70. Clark, G.A., Ledbetter, N.M., Warren, D.J., Harrison, R.R., 2011. Recording sensory and motor information from peripheral nerves with Utah slanted electrode arrays. Conf. Proc. IEEE Eng. Med. Biol. Soc. 2011, 4641–4644. Davis, T.S., Wark, H.A., Hutchinson, D.T., Warren, D.J., O’Neill, K., Scheinblum, T., Clark, G.A., Normann, R.A., Greger, B., 2016. Restoring motor control and sensory feedback in people with upper extremity amputations using arrays of 96 microelectrodes implanted in the median and ulnar nerves. J. Neural Eng. 13, 36001. Deuis, J.R., Dvorakova, L.S., Vetter, I., 2017. Methods used to evaluate pain behaviors in rodents. Front. Mol. Neurosci. 10, 284. Dhillon, G.S., Horch, K.W., 2005. Direct neural sensory feedback and control of a prosthetic arm. IEEE Trans. Neural Syst. Rehabil. Eng. 13, 468–472. Dhillon, G.S., Lawrence, S.M., Hutchinson, D.T., Horch, K.W., 2004. Residual function in peripheral nerve stumps of amputees: implications for neural control of artificial limbs. J. Hand Surg. 29, 605–615 discussion 16-8. Fahrenkopf, M.P., Adams, N.S., Kelpin, J.P., Do, V.H., 2018. Hand amputations. Eplasty 18, ic21. Farrokhi, S., Perez, K., Eskridge, S., Clouser, M., 2018. Major deployment-related amputations of lower and upper limbs, active and reserve components, U.S. Armed Forces, 2001-2017. MSMR 25, 10–16. Fisher, J.A., Huang, S., Ye, M., Nabili, M., Wilent, W.B., Krauthamer, V., Myers, M.R., Welle, C.G., 2016. Real-time detection and monitoring of acute brain injury utilizing evoked electroencephalographic potentials. IEEE Trans. Neural Syst. Rehabil. Eng. 24, 1003–1012. Ghafoor, U., Kim, S., Hong, K.S., 2017. Selectivity and longevity of peripheral-nerve and machine interfaces: a review. Front. Neurorobot. 11, 59. Graczyk, E.L., Schiefer, M.A., Saal, H.P., Delhaye, B.P., Bensmaia, S.J., Tyler, D.J., 2016. The neural basis of perceived intensity in natural and artificial touch. Sci. Transl. Med. 8, 362ra142. Gunter, C., Delbeke, J., Ortiz-Catalan, M., 2019. Safety of long-term electrical peripheral nerve stimulation: review of the state of the art. J. Neuroeng. Rehabil. 16, 13. Hong, K.S., Aziz, N., Ghafoor, U., 2018. Motor-commands decoding using peripheral nerve signals: a review. J. Neural Eng. 15, 31004. Howland, R.H., 2014. Vagus nerve stimulation. Curr. Behav. Neurosci. Rep. 1, 64–73. Hunter, D.A., Moradzadeh, A., Whitlock, E.L., Brenner, M.J., Myckatyn, T.M., Wei, C.H., Tung, T.H., Mackinnon, S.E., 2007. Binary imaging analysis for comprehensive quantitative histomorphometry of peripheral nerve. J. Neurosci. Methods 166, 116–124. Jung, R., Abbas, J.J., Kuntaegowdanahalli, S., Thota, A.K., 2018. Bionic intrafascicular interfaces for recording and stimulating peripheral nerve fibers. Bioelectron. Med. (Lond.) 1, 55–69. Milby, A.H., Halpern, C.H., Baltuch, G.H., 2008. Vagus nerve stimulation for epilepsy and depression. Neurotherapeutics 5, 75–85. Moxey, P.W., Gogalniceanu, P., Hinchliffe, R.J., Loftus, I.M., Jones, K.J., Thompson, M.M., Holt, P.J., 2011. Lower extremity amputations—a review of global variability in incidence. Diabet. Med. 28, 1144–1153. Navarro, X., Krueger, T.B., Lago, N., Micera, S., Stieglitz, T., Dario, P., 2005. A critical review of interfaces with the peripheral nervous system for the control of neuroprostheses and hybrid bionic systems. J. Peripher. Nerv. Syst. 10, 229–258. Oddo, C.M., et al., 2016. Intraneural stimulation elicits discrimination of textural features by artificial fingertip in intact and amputee humans. Elife 5, e09148. Polasek, K.H., Hoyen, H.A., Keith, M.W., Kirsch, R.F., Tyler, D.J., 2006. Spiral nerve cuff electrodes for an upper extremity neuroprosthesis. Conf. Proc. IEEE Eng. Med. Biol. Soc. 1, 3584–3587. Raspopovic, S., et al., 2014. Restoring natural sensory feedback in real-time bidirectional hand prostheses. Sci. Transl. Med. 6, 222ra19. Rijnbeek, E.H., Eleveld, N., Olthuis, W., 2018. Update on peripheral nerve electrodes for closed-loop neuroprosthetics. Front. Neurosci. 12, 350. Rossini, P.M., et al., 2010. Double nerve intraneural interface implant on a human amputee for robotic hand control. Clin. Neurophysiol. 121, 777–783. Rossini, P.M., Rigosa, J., Micera, S., Assenza, G., Rossini, L., Ferreri, F., 2011. Stump nerve signals during transcranial magnetic motor cortex stimulation recorded in an amputee via longitudinal intrafascicular electrodes. Exp. Brain Res. 210, 1–11. Sarikcioglu, L., Demirel, B.M., Utuk, A., 2009. Walking track analysis: an assessment method for functional recovery after sciatic nerve injury in the rat. Folia Morphol. (Warsz) 68, 1–7. Shafer, B., Yaghouby, F., Vasudevan, S., 2018. Short-time fourier transform based spike detection of spontaneous peripheral nerve activity. Conf. Proc. IEEE Eng. Med. Biol. Soc. 2018, 2418–2421. Straka, M.M., Shafer, B., Vasudevan, S., Welle, C., Rieth, L., 2018. Characterizing
5. Conclusion We have advanced a previously developed test platform to assess the safety and performance of intraneural PNIs using a rodent model. The current test platform allows for long-term assessment up to 12 weeks of implantation. The ability to conduct these assessments on PNI devices provides a comprehensive understanding of reliability and failure modes in the non-clinical setting, which is critical to improve upon device design for novel PNI technologies. Disclaimer The mention of commercial products, their sources, or their use in connection with material reported herein is not to be construed as either an actual or implied endorsement of such products by the Department of Health and Human Services. Acknowledgements This research project was supported by the Defense Advanced Research Projects Agency (DARPA), Biological Technologies Office (BTO), Hand Proprioception and Touch Interfaces (HAPTIX) Program, through an Interagency Agreement with the US Food and Drug Administration. The authors would like to acknowledge the Orthopedic Nerve Research Laboratory at Washington University and Jason Wever for 9
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
D.J., Hutchinson, D.T., Clark, G.A., 2017. Restoration of motor control and proprioceptive and cutaneous sensation in humans with prior upper-limb amputation via multiple Utah Slanted Electrode Arrays (USEAs) implanted in residual peripheral arm nerves. J. Neuroeng. Rehabil. 14, 121. Wurth, S., et al., 2017. Long-term usability and bio-integration of polyimide-based intraneural stimulating electrodes. Biomaterials 122, 114–129. Ziegler-Graham, K., MacKenzie, E.J., Ephraim, P.L., Travison, T.G., Brookmeyer, R., 2008. Estimating the prevalence of limb loss in the United States: 2005 to 2050. Arch. Phys. Med. Rehabil. 89, 422–429.
longitudinal changes in the impedance spectra of in-vivo peripheral nerve electrodes. Micromachines (Basel) 9. Vasudevan, S., Yan, J.G., Zhang, L.L., Matloub, H.S., Cheng, J.J., 2013. A rat model for long-gap peripheral nerve reconstruction. Plast. Reconstr. Surg. 132, 871–876. Vasudevan, S., Patel, K., Welle, C., 2017. Rodent model for assessing the long term safety and performance of peripheral nerve recording electrodes. J. Neural Eng. 14, 16008. Vasudevan, S., Vo, J., Shafer, B., Nam, A.S., Vakoc, B.J., Hammer, D.X., 2019. Toward optical coherence tomography angiography-based biomarkers to assess the safety of peripheral nerve electrostimulation. J. Neural Eng. Wendelken, S., Page, D.M., Davis, T., Wark, H.A.C., Kluger, D.T., Duncan, C., Warren,
10
Contents lists available at ScienceDirect
Journal of Neuroscience Methods journal homepage: www.elsevier.com/locate/jneumeth
A rat model for assessing the long-term safety and performance of peripheral nerve electrode arrays Benjamin Shafera, Cristin Welleb, Srikanth Vasudevana,
T
⁎
a U. S. Food and Drug Administration, Center for Devices and Radiological Health (CDRH), Office of Science and Engineering Laboratory (OSEL), Division of Biomedical Physics (DBP), Silver Spring, MD, USA b University of Colorado, Anschutz Medical Campus, Departments of Neurosurgery and Bioengineering, Aurora, CO, 80045, USA
A R T I C LE I N FO
A B S T R A C T
Keywords: Peripheral nerve interface Microelectrode arrays Utah electrode Sciatic nerve
Background: High-resolution peripheral nerve interfaces (PNIs) can provide amputees with intuitive motor control and sensory feedback. Current PNIs are limited by early device failure and suboptimal long-term stability. The present study aims to incorporate functional assessment into an in vivo test platform to assess the longterm safety and performance of PNIs for recording and stimulation. New methods: Utah electrode arrays (EA) were implanted in the rat sciatic nerve along with electromyography wires in the gastrocnemius and tibialis anterior. Cranial EEG screws were implanted in the somatosensory cortex for 12 weeks. Spontaneous neural activity was recorded using the implanted EA and stimulation-induced activity was monitored throughout the experiment. The impedance of each electrode was measured, and nerve function tests were conducted throughout the EA lifetime. Post-hoc safety assessments included scanning electron microscopy (SEM) of the EA and nerve histomorphometric analysis. Results: EA recordings were stable, and stimulation with EA elicited somatosensory evoked potentials and muscle contractions. Motor and sensory function tests indicated minimal deficits. Histomorphometric analysis indicated changes in nerve microstructure. SEM indicated EA-tip fracture, while lead wire breakage primarily caused device failure. Comparison with existing methods: We improved our prior platform with the addition of functional assessments of sensory pathways, a robust EMG array design to increase device longevity, and quantitative analysis of nerve microstructure. Conclusion: We present a test platform for long-term assessment of peripheral nerve interfaces for stimulation and recording. Using this platform, we demonstrate recording and stimulation with minimal impact on nerve function, while EA lead wire breakage and tip fracture could limit long-term device use.
1. Introduction In 2008, one in 190 Americans were living with limb loss, and this number is projected to increase substantially by 2050 (Shafer et al., 2018; Ziegler-Graham et al., 2008). Amputations in the lower limbs are mostly carried out due to the comorbidities associated with diabetes mellitus (Moxey et al., 2011), while trauma, malignancy, and infections are associated with upper limb amputations (Fahrenkopf et al., 2018). Between 2001 and 2017, 1705 service members of the U.S. armed forces sustained deployment related lower and upper limb amputations (Farrokhi et al., 2018). For many, the loss of a limb means inability to return to duty and increased comorbidities, which can have a negative impact on quality of life. To alleviate the burden associated with limb amputation, there is an increased need for advancing intuitive ⁎
neuroprosthetic devices for restoring lost sensory and motor function. Recently, researchers have focused their efforts on peripheral nerve interfaces (PNI) as an alternative to interfacing with the brain, exploring possible benefits such as: ease of surgical access, bidirectional communication (both sensory and motor pathways) at one interface site, persistence of functional neural connections post-amputation, increased selectivity for stimulating sensory feedback, and access to nerves at the time of amputation, reducing the need for additional surgeries (Dhillon et al., 2004; Ghafoor et al., 2017; Hong et al., 2018; Navarro et al., 2005; Rossini et al., 2011; Vasudevan et al., 2017). There are two main classifications of PNIs, extraneural and intraneural, signifying the level of invasiveness and, hence, selectivity for both recording and stimulation (Navarro et al., 2005; Gunter et al., 2019). Extraneural PNIs do not penetrate the nerve and are relatively
Corresponding author. E-mail address: [email protected] (S. Vasudevan).
https://doi.org/10.1016/j.jneumeth.2019.108437 Received 26 April 2019; Received in revised form 5 August 2019; Accepted 13 September 2019 Available online 14 September 2019 0165-0270/ Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
Fig. 1. Implant design. (a) 16 channel Blackrock EA with 55 mm long lead wire bundle and reference wires. (b) Custom EMG wire electrode with 4 bipolar insulated stainless-steel wires. (c) Bottom part of the 3D printed connector mount made with Nylon. (d) Top part of the 3D printed connector mount made with bronze infused stainless-steel. (e) Fully assembled connector mount. Scale bar 10 mm.
these devices for long-term use. A beneficial assessment platform would allow longitudinal assessment of: PNI impact on nerve function (motor and sensory), tissue response to implantation, selectivity of stimulation, stability of recording, and identifying failure modes early in the development cycle. While studies have completed these comprehensive assessments for intraneural PNIs used for recording (Vasudevan et al., 2017), there persists a need to expand the assessment of long-term safety and performance to PNIs used for both recording and stimulation. In this work, we have appended our previously published test platform (Vasudevan et al., 2017) to include the comprehensive assessment of the long-term safety and performance of PNIs used for stimulation, as well as recording, using a rat sciatic nerve model. We used Blackrock electrode arrays (EA) as the PNI implanted into the sciatic nerve, a custom-designed EMG wire electrode for recording compound muscle action potentials (cMAP) from gastrocnemius and tibialis anterior muscles, and stainless-steel cranial screws for recording somatosensory evoked potentials (SSEP). We developed a transcutaneous connector mount for accessing the EA and EMG wire electrodes after implantation. After the animals recovered from surgical procedures, we measured electrode impedance, recorded neural activity from awake behaving animals, studied stimulation evoked responses, and assessed motor and sensory function over a period of 12 weeks. At week 13, we harvested the nerve for histomorphometric analysis and studied
less invasive than intraneural PNIs. Extraneural PNIs have been approved for clinical applications to treat epilepsy and depression (Howland, 2014; Milby et al., 2008), and are currently being tested for restoration of function and neuroprosthetic applications (Christie et al., 2017; Graczyk et al., 2016; Polasek et al., 2006). Due to the presence of epineurium between the electrode and nerve fascicles, a limitation of extraneural PNIs is the inadequate spatial resolution (Rijnbeek et al., 2018). Intraneural PNIs are designed to penetrate the nerve for accessing individual fascicles and nerve fibers. Penetration increases recording signal quality, bypassing the less conductive connective tissue layers and decreasing the distance between the electrode tips and target axons (Ghafoor et al., 2017; Navarro et al., 2005). Intraneural PNIs have been shown to elicit specific sensations (Davis et al., 2016; Dhillon and Horch, 2005; Oddo et al., 2016), as well as record neural activity for motor control (Rossini et al., 2010; Wendelken et al., 2017). However, intraneural PNIs struggle with long-term stability caused by lead wire breakage, electrode migration, and foreign body response (Vasudevan et al., 2017; Gunter et al., 2019; Straka et al., 2018). Nevertheless, the higher resolution bidirectional communication offered by intraneural PNIs is favorable for advancing neuroprosthetic devices. As the research and clinical need involved with PNI technologies continues to rise, it is increasingly important to develop and improve experimental test platforms to assess the safety and performance of 2
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
pneumatic impactor (Blackrock Microsystems, Inc., U.S.A.). EA insertion was achieved by applying ∼5-6 gentle taps using the impactor set at 10 psi (Fig. 2b). In some cases, a few electrode shanks were found to be outside of the nerve due to larger dimensions of the EA. After EA insertion, a few drops of fibrin sealant (TISSEEL, Baxter Healthcare Corporation, U.S.A.) was applied to the nerve and electrode construct for stability. The silicone block and Parafilm were removed once the fibrin sealant cured, and a silicone cuff (4 mm length and 3.2 mm inner diameter) was placed around the nerve and electrode construct. The silicone cuff lumen was filled with fibrin sealant as shown in Fig. 2c. After confirming that the fibrin sealant cured completely, the muscle incision over the nerve was closed and reference wires were delaminated ∼0.5 mm using a lighter and secured inside the muscle over the nerve. For implanting the EMG wire electrodes, gastrocnemius and tibial anterior muscles were exposed through a skin incision. EMG wires were then tunneled under the skin from the connector mount site to the implant site. Using a 25 G needle serving as a canula, two pairs of EMG wires were interested into each of the two muscles for redundancy and were secured in place with sutures (4-0 Prolene, Oasis, U.S.A.) as shown in Fig. 2d. Skin incisions over the nerve and muscle implant sites were closed with a combination of sutures and gluture as shown in Fig. 2e. To record somatosensory evoked potentials (SSEP), stainless-steel screws (0.2 mm × 3 mm, M0.8, US Micro Screw, U.S.A.) were used as electrodes. The skull was exposed, and burr holes were carefully drilled over the left and right somatosensory cortex. During the process, it was ensured that the burr holes did not pass through the skull. Screws were inserted into the burr holes (Fig. 2f) and secured using dental acrylic (Jet Denture Repair Powder, Lang Dental Manufacturing Company, Inc., U.S.A.). Finally, animals were administered Atipamezole (0.5 mg/ kg, intraperitoneal) for anesthesia reversal, Meloxicam (2 mg/kg, subcutaneous) for analgesia, and Gentamycin (8 mg/kg, subcutaneous) for antibiotic treatment. Topical antibiotic ointment was applied to all incisions after the surgery. Meloxicam (1 mg/kg, subcutaneous) was administered for another two days post-surgery for pain control.
mechanical damage to the EA using scanning electron microscopy (SEM). Our results showed that the current test platform was conducive for assessing the long-term safety and performance of PNIs used for recording and stimulation. 2. Methods 2.1. Animal protocol and experimental design This study was approved by the Institutional Animal Care and Use Committee (IACUC) at the Food and Drug Administration, White Oak campus. Female Lewis rats were purchased from Charles River Laboratories International Inc., for experiments described in this manuscript. Experiments were conducted on animals weighing between 200–280 g at the time of surgery. All animals were individually housed in plastic cages with 12 -h light and dark cycle before and after experiments. Animals were randomly assigned to one of two groups: control, n = 6, and EAs, n = 6. The control group animals were not subjected to surgical procedures. 2.2. Implant design Sixteen channel high-density Blackrock EAs (Blackrock Microsystems LLC, U.S.A.) with shanks (1 mm in length with a 0.4 mm pitch) arranged in a 4 × 4 configuration were used in this study. The EA and the connector (A73098-001, Omnetics Connector Corporation, U.S.A.) were linked via a 55 mm long gold lead wire (Fig. 1a). One of the reference wires served as the ground for recording and stimulation. Additionally, a custom-designed EMG wire electrode (Microprobes for Life Science, U.S.A.) with 8 insulated stainless steel wires of 110 mm length (0.1 mm diameter, arranged into 4 bipolar pairs) attached to a connector (A79038-001, Omnetics Connector Corporation, U.S.A.) was used to record muscle activity (Fig. 1b). The tip of the bipolar pair was de-insulated by the manufacturer, with greater insulation removed for one electrode per pair to allow it to serve as the reference. Two of six animals received connector mount designed using SolidWorks (Dassault Systèmes SolidWorks Corp., France) and 3D printed with Stainless Steel infused with bronze (Shapeways, Inc., U.S.A.). The other 4 animals received a two-part connector mount, top made with Stainless Steel infused with bronze and bottom part made with Nylon as shown in Fig. 1c and d (Straka et al., 2018). Connectors from the EA and EMG wire electrode were housed inside the connector mount and a Mersilene mesh (2” x 2”, Ethicon, Inc., U.S.A.) was attached to the bottom of the connector mount using epoxy (Loctite Hysol Epoxy, Henkel Corporation, U.S.A.) as shown in Fig. 1e. Kwik-cast (World Precision Instruments, Inc., U.S.A.) was applied to the bottom portion of the connector mount to provide softer tissue contacting surface.
2.4. Electrical characteristics Electrochemical Impedance Spectroscopy (EIS) was used to measure broadband impedance (1 Hz-1 MHz, Gamry Instruments Inc., U.S.A.) of the individual channels pre- and post-implantation. Pre-implant measurements were obtained by immersing the electrodes in phosphate buffered solution (PBS) and measuring the impedance between one of the reference wires and the channel of interest. Measurements were obtained immediately after implantation, again at 2 weeks following implantation, and weekly thereafter until device failure. The channels with an impedance ranging between 100 kΩ and 800 kΩ is considered functional as per the manufacturer. Therefore, a functional impedance threshold of 1 MΩ (at 1 kHz) was decided to accommodate for variability in measurement technique. Channels with impedance > 1 MΩ were considered non-functional, and EA with all the channels > 1 MΩ were considered broken (Fig. 3b). Detailed analysis of EIS data was published previously (Straka et al., 2018) and is not discussed in this manuscript.
2.3. Surgical procedure All surgical procedures were performed under aseptic conditions. Intraperitoneal injection of cocktail made with ketamine (75 mg/kg) and dexmedetomidine (0.25 mg/kg) was used to deeply anesthetize the rats. Adequate anesthetization was confirmed by the loss of toe pinch reflex. The surgical site was shaved and sterilized using alcohol and iodine. Ocular ointment was applied to each eye before and during surgery to prevent drying. Sterile warm saline (3 ml) was subcutaneously injected to ensure adequate hydration. While the electrode implantation procedure described here can be implemented on either hind limb, we used the right hind limb in this study. After securing the connector mount to the lumbar fascia (Fig. 2a), the right sciatic nerve was exposed using blunt dissection technique as described previously (Vasudevan et al., 2017, 2013). A piece of silicone block (∼10 × 8 × 1 mm, Sylgard, Dow Corning, U.S.A.) and small sheet of Parafilm were placed under the nerve to prevent it from moving downward during EA implantation using the
2.5. Electrophysiological recording of spontaneous neural activity Starting at two weeks, weekly spontaneous neural activity recordings were acquired. Animals were briefly anesthetized with 3% isoflurane and then connected to the recording system (Neuralynx Inc., U.S.A.). Following recovery from anesthesia, spontaneous neural activity recorded by the EA was acquired for 15 min during free exploration of the home cage. Spike events from recordings were extracted and analyzed using Neuraview and SpikeSort 3D (Neuralynx, U.S.A.) to confirm the presence of neural activity in individual channels. Automatic spike sorting algorithms in SpikeSort 3D (KlustaKwik) 3
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
Fig. 2. Surgical procedure. (a) Connector mount implanted over the lumbar fascia. (b) Electrode array inserted into the sciatic nerve. (c) Silicone cuff applied around the nerve and electrode array, which is then filled with fibrin sealant (cloudy inside the tube). (d) EMG wire arrays implanted into the gastrocnemius and tibialis anterior muscles. (e) Skin incisions closed with sutures and gluture. (f) Stainless-steel screws implanted over the somatosensory cortex.
were first used, followed by manual refinement to identify neural activity as described elsewhere (Vasudevan et al., 2017). For each channel with neural activity, clusters with baseline noise and neural activity were identified as shown in Fig. 3a. The signal-to-noise ratio (SNR) was calculated by taking a ratio of the average signal peak-to-peak and average baseline noise peak-to-peak. For channels with multiple neural waveform clusters, the cluster with the highest average peak-to-peak was used for SNR calculations. Data analysis was limited to seven weeks to include neural activity from all six animals at each time point. Data between time points were compared using one-way analysis of variance (ANOVA, Prism 8, GraphPad Software, U.S.A.) (Fig. 3c).
only one channel that induced muscle activation and/or SSEP on the first day of stimulation, that channel along with a randomly selected channel was consistently tested for the entire course of the experiment. If an array did not have any channels that induced muscle activation and/or SSEP on the first day of stimulation, two channels were randomly selected and consistently tested for the entire course of the experiment. Twenty-five anodic-first charge balanced square wave pulses with 12.5 μA per phase, 100 μs per phase were applied at 3.1 Hz and the corresponding SSEP and cMAP signals were recorded for analysis. In total, 12 individual channels from all 6 EAs were used to analyze SSEP and cMAP recordings to calculate the percentage of active channels (number of channels with SSEP or cMAP/12 × 100%). SSEP signals were measured for 100 ms after each stimulation pulse. An average of 25 stimulation induced SSEPs was calculated from both the left and right-side cranial electrodes, and then re-referenced by subtracting the right from the left SSEP (Fig. 3d). The re-referenced signal was then zero-averaged and analyzed in two specific time regions associated with SSEPs in our study. cMAP signals were processed similar to SSEP and identified using an average of 25 signals 100 ms in duration starting 5 ms after the EA stimulation trigger.
2.6. Electrical stimulation to drive somatosensory evoked potentials and compound muscle action potentials Somatosensory evoked potentials (SSEP) were recorded from cranial screws over the left and right somatosensory cortex. Baseline (no stimulation) recordings and stimulation-evoked SSEP were acquired each week beginning at two weeks post-implantation (RZ5D processor, PZ2 preamplifier, ZC16 headstage, Tucker-Davis Technologies, U.S.A.). Recordings were acquired with a single-ended configuration and shared a common reference electrode (18 G subcutaneous needle temporarily inserted over the neck muscle) at 3 kHz sampling frequency (Fisher et al. (2016)). Compound muscle action potentials (cMAP) were recorded simultaneously with SSEP recordings using the same system. Recordings were acquired with a single-ended configuration from all bipolar EMG wires with respect to the shared common reference electrode at 3 kHz sampling frequency. Two channels for each EA was selected for applying electrical stimulation (one channel at a time) between the channel and the EA reference wire. On the first day of stimulation trial, all 16 channels were tested individually for muscle activation and/or SSEP. Two channels that induced muscle activation and/or SSEP were selected and consistently tested for the entire course of the experiment. If an array had
2.7. Functional assessment Walking Track analysis was used to measure motor function at baseline, 5 days post-implantation (1 W), 12 days post-implantation (2 W) and biweekly until week 12. Animals were confined to 8 × 40 cm path and foot print videos were recorded with a camera placed underneath the platform (FreeWalk, Cleversys Inc., U.S.A.). Five sets of footprints were extracted from the videos and analyzed using ImageJ to calculate print length, toe spread, and intermediary toe spread. The Sciatic Functional Index (SFI) (Fig. 4a) and Tibial Functional Index (TFI) (Fig. 4b) were calculated using established methods and the average was used for data analysis (Vasudevan et al., 2017; Bain et al., 1989; Vasudevan et al., 2019). 4
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
Fig. 3. Electrophysiology. (a) Isolated neural activity (green) and background noise (pink). (b) Number of functional devices and percentage of non-functional channels based on impedance threshold of 1MΩ. (c) Average SNR values of identified neural activity in all EAs up to 7 weeks. (d) Example of SSEP waveform with a negative peak followed by a positive peak, identified with a threshold ± 2 STD of the baseline noise, respectively. Note that the plot is inverted following the conventions of clinical neurophysiology field. (e) Average SSEP peak-to-peak amplitude and percentage of active channels eliciting SSEP during simulation. (f) SSEP latency measured as the time from stimulation onset to the time of each peak. (g) Example cMAP waveform with two peaks identified with a threshold ± 3 STD of the baseline noise. (h) Average gastrocnemius cMAP peak-to-peak amplitude and percentage of active channels eliciting cMAP in response to EA simulation. (i) Average tibialis anterior cMAP peak-to-peak amplitude and percentage of active channels eliciting cMAP in response to EA simulation.
followed by Tukey’s multiple comparison test and between group time points were compared using two-way ANOVA followed by Bonferroni’s multiple comparison test (Prism 8, GraphPad Software, U.S.A.).
Von Frey test was used to assess sensory function. The animal was placed in a cage (25.4 × 25.4 cm) with a perforated floor and allowed to acclimate for 5 min before commencing measurements. With the animals in stationary position, a rigid tip of the Von Frey anesthesiometer (IITC Life Science, U.S.A.) was pressed against the central region of the right plantar foot. Response to mechanical allodynia was recorded (in grams) for 5 consecutive trials, and the average was used for further analysis (Fig. 4c) (Vasudevan et al., 2017, 2019). For walking track analysis and Von Frey test, data within groups (control and electrode array) were compared using two-way ANOVA
2.8. Semiautomated quantitative histomorphometry All animals were sacrificed at 13 weeks post-implantation by first administering ketamine/dexmedetomidine anesthesia cocktail followed by pentobarbital (200 mg/kg, intraperitoneal) injection. Nerve segment ∼1 cm distal to the EA implant site was carefully dissected and
Fig. 4. Function. (a) Sciatic Functional Index (SFI) of control and EA implanted animals. (b) Tibial Functional Index (TFI) of control and EA implanted animals. (c) Von Frey test of control and EA implanted animals. Data shown as mean ± S.E. 5
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
Fig. 5. Histomorphometry. (a) Light microscopy image of nerve cross-section obtained from one animal in the EA group. (b) Nerve fiber density comparison between EA and control group. (c) Myelin area comparison of EA and control group. (d) Comparison of myelinated fiber distribution between EA and control group. Data shown as mean ± S.E. *p ≤ 0.05.
Fig. 6. Scanning electron microscopy. (a) SEM image of 16 channel EA post-harvest. (b) Fractured electrode tip. (c) Intact electrode tip.
immersion fixed in 3% glutaraldehyde at 4 °C. Nerve samples (n = 3 control, n = 6 treatment) were processed for epoxy embedding. Samples were post-fixed in 1% osmium tetroxide and subjected to serial dehydration and embedding in Araldite 502 (Nerves Incorporated, U.S.A.). The epoxy embedded samples were cut into semi-thin sections
and stained with 1% toluidine blue dye for histomorphometric analysis in the Laboratory of Dr. David Brogan, Department of Orthopedic Surgery, Washington University School of Medicine (Fig. 5a). Histomorphometry of each nerve section was performed using a digital image analysis system with linked morphometry software (Leco
6
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
throughout the study to avoid unintended electrode damage. An example of SSEP waveform is shown in Fig. 3d. The time region between 8–15 ms post-stimulation was observed with a negative peak and the region between 18–23 ms, was observed with a positive peak. Peaks within each region were noted if they exceeded +/- 2 standard deviations of the baseline re-referenced recording (Fig. 3d) and manually examined to confirm the presence of an SSEP signal. Peak-to-peak calculations were made based on the noted peaks (Fig. 3e). The SSEP latency was measured for the negative and positive peaks as the time from stimulation until each peak (Fig. 3f). In cases where there was only one peak during a specific recording, the average of the signal within the missing peak’s time region was taken and used for the peakto-peak calculation. The average SSEP peak-to-peak amplitude decreased over time after reaching a peak around 4–5 W post-implantation (Fig. 3e). A maximum of 50% channels elicited SSEP across all EAs. SSEP latency was analyzed by measuring the time to peak of the negative and positive deflections from stimulation onset. The latency of both peaks remained consistent through the 7 weeks of stimulation (Fig. 3f). cMAP recordings were analyzed to assess muscle activation in response to EA stimulation. Stimulation waveforms were designed to prevent EA damage, so they did not elicit maximal contraction of the muscles. An example cMAP waveform is shown in Fig. 3g. The stimulation artifact was apparent in most signals (Fig. 3g). After the stimulation artifact, the cMAP was characterized with two peaks. At least one needed to be greater than -/+ 3 STD of the baseline noise for analysis (Fig. 3g). Once identified, the amplitude difference between the two peaks were measured as the peak-to-peak amplitude of the cMAP for gastrocnemius (Fig. 3h) and tibialis anterior (Fig. 3i) muscles. If only one peak was present, the amplitude of that peak along with the average of the signal within missing peak’s time region was used as the peak-to-peak amplitude. For target muscles (gastrocnemius and tibialis anterior), cMAPs were observed in some of the channels in response to stimulation of the nerve (Fig. 3h and 3i). The highest gastrocnemius peak-to-peak amplitude occurred at 5 W, while the highest tibialis anterior peak-to-peak amplitude occurred at 7 W. Lowest peak-to-peak values for gastrocnemius was observed between 2 and 4 weeks. While the lowest peak-to-peak average amplitude for tibialis anterior was observed between 2 and 6 weeks.
Instruments, U.S.A.) as previously described by Hunter et al. (Hunter et al. (2007)). Multiple field images were acquired to calculate fiber density (Fig. 5b), myelin area (Fig. 5c) and myelinated fiber distribution (Fig. 5d). Myelin area and fiber distribution data were compared with unpaired t-test, and multiple t-tests were used to compare myelinated fiber distribution (Prism 8, GraphPad Software, U.S.A.). 2.9. Scanning electron microscopy The nerve along with the EA was immersion fixed in 4% paraformaldehyde solution at 4 °C overnight. EA was carefully extracted from the nerve using microdissection methods and immersed in a deionized water solution containing 1% Tergazyme (Alconox, USA), 0.5% Triton X-100 for 3 days to remove biological debris. The EA was then washed in de-ionized water and air dried for at least 24 h before subjecting to scanning electron microscopy (SEM) imaging. SEM images (Jeol 6390 L V, USA) were acquired under low vacuum mode (15 kV, 30 Pa) to assess the integrity of EA tips (Fig. 6) (Vasudevan et al., 2017). 3. Results 3.1. Implant design and surgical procedure The ability to perform long-term assessment of PNIs in animal models depends on stable implantation of both the implanted electrodes and the connectors (Vasudevan et al., 2017). Therefore, reliable connector mount design is critical for long-term recording or stimulation studies. The two-part connector mount setup was assembled as shown in Fig. 1e, with the tissue-contacting surface made suitable for reducing animal discomfort. The metal top prevents damage to the connectors inflicted by animal chewing or mechanical damage inside the home cage. The Mersilene mesh attached to the bottom of the connector mount as shown in Fig. 1e encourages tissue ingrowth, which is critical for long-term stability. The assembled connector mount with housed electrode connectors shown in Fig. 1e allowed for easy and reliable access to the implanted electrodes over long periods of time. 3.2. Impedance measurements Detailed analysis of EIS data has been previously reported (Straka et al., 2018). Here, we used the data to determine if the devices were functional. EAs were considered non-functional when all the channels had impedance measurements above 1 MΩ at 1 kHz. Device failure was mainly the result of lead wire breakage at the connector mount interface. All devices were functional until 7 W post-implantation (Fig. 3b). For functional devices, greater than 75% of channels were within functional range. Overall, 50% of the tested EAs had functional channels for the 12 W experimentation period and greater than half of the channels were below 1 MΩ at 12 W (Fig. 3b).
3.5. Functional assessment Walking track analysis was used to assess motor function in the sciatic nerve (Fig. 4a) and the tibial nerve (Fig. 4b) in response to surgery and experimental procedures. Both SFI and TFI showed a trend towards deficits at 1 W and recovered close to pre-implant levels around 2 W. None of the time points were significantly different, indicating that the surgical procedure or the presence of EA in the nerve did not alter the motor function significantly. Presence of the EA also did not impede with recovery of motor function. Von Frey test was used to assess alteration in sensory function over time. There was no significant difference between the control and EA group, indicating that neither the surgical procedure nor the presence of the EA led to significant loss in sensory function or increase in sensitivity (Fig. 4c).
3.3. Electrophysiological recording assessment Spontaneous action potentials from the sciatic nerve were recorded for 15 min each week starting at week 2, up to 12 weeks or device failure, while the animals moved freely inside the home cage. Offline analysis of the recordings was conducted to extract single units using SpikeSort3D. Data analysis was performed up to 7 weeks to include all 6 devices, after which device failure was observed (Fig. 3b). An example of a neural activity is shown in Fig. 3a. The average SNR of recorded units was ∼4 at 2 W and did not substantially change from this value for up to 7 weeks of analysis (Fig. 3c).
3.6. Semiautomated quantitative histomorphometry Histomorphometry was used to study the effects of EA implantation on myelinated fibers at 13 weeks post-implantation (Fig. 5a). The fiber density was comparable between control and EA group (Fig. 5b), while the myelin area was lower in the EA group compared to control (Fig. 5c). Distribution of the myelinated fibers show that there is a significant difference in distribution of 4–6 μm fibers between EA and control groups (Fig. 5d). Overall, histomorphometry points to changes in nerve microstructure because of EA implantation and experimental procedures.
3.4. Electrophysiological stimulation performance SSEP and cMAP recordings were used to assess stimulation performance of the implanted EA. The stimulation parameters were fixed 7
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
surgical procedure was used to verify if damage occurred anytime during the surgery. The impedance for most of the electrodes did not increase beyond the functional threshold of 1 MΩ (Straka et al., 2018), suggesting that the implantation procedure did not cause lead wire breakage. Impedance could be characterized for all electrode tips over the course of the experiment unless lead wire breakage occurred. Wire breakage was indicated by simultaneous impedance increase above 1 MΩ for all electrode channels. Beyond lead wire breakage, abiotic and biotic factors could affect intraneural PNI impedance in the long term (Vasudevan et al., 2017; Straka et al., 2018). Broadband EIS measurements can potentially reveal failure modes beyond simple mechanical failure (Straka et al., 2018). In combination with characterization of the abiotic and biotic factors influencing an implanted PNI, chronic EIS measurement through this platform gives valuable feedback for electrode material selection and fabrication. Electrophysiological recording in awake behaving rats was made possible with the stable connector interface that allowed experimenters to easily and safely connect to the electrodes, minimizing the duration of anesthetic use and mechanical damage at the connector mount interface. The ability to record from the EA in freely-moving animals is important to assess the reliability of recorded signals without confounds, such as the influence of anesthesia on nerve conduction. In addition, multiple individual neural units could be isolated from the EA channels demonstrating the high fidelity of intraneural PNIs for advancing motor control using neuroprosthetic devices signals (Davis et al., 2016; Clark et al., 2011). The average SNR for these detected action potentials was ∼4 up to 7 W, suggesting applicability to pattern recognition for neuroprosthetic control applications (Shafer et al., 2018). SSEP and cMAP evaluation in animal models can be used to characterize the potential of a PNI to be used for closed-loop neuroprosthetic applications. The implanted EA displayed the ability to produce cMAPs and SSEPs in the long-term, but this was not consistent between time points. This inconsistency could be a result of micromotion inside the nerve, as the electrode tip distance from axons within the nerve could change. Gastrocnemius cMAP peak-to-peak values peaked at 5 W and decreased with time (Fig. 3h) while Tibialis Anterior peak-to-peak values increased with time and peaked around 7 W (Fig. 3i). This activation pattern could be attributed to the change in relative distance between the electrode tips with respect the nerve fibers innervating either of the two muscles, along with biotic factors such as fibrotic encapsulation at the electrode tip. Similar to the gastrocnemius cMAP signals, the peak-to-peak amplitudes of SSEPs reached its highest value around 4 W and then trended downward (Fig. 3e). Micromotion of the EA inside the nerve could result in the stimulation of different fibers leading to inconsistent cMAP and SSEP stimulation or decreased amplitudes of stimulation if the electrodes were to move farther from a previously stimulated fiber over time (Davis et al., 2016). While micromotion is a noted limitation of microelectrode arrays (Ghafoor et al., 2017), the variety of stimulated fiber types throughout the duration of the experiment reveals the potential of intraneural PNIs to interact with various fibers if electrode tips were placed intentionally. Stimulation of sensory nerves in humans using intraneural PNIs elicit a variety of sensations and intensities noted consciously by users (Davis et al., 2016; Raspopovic et al., 2014). The foreign body response could also influence the stimulation and recording performance of intraneural PNIs (Christensen et al., 2014). Changes in the PNI materials may show promise for mitigating these effects but more development is needed (Wurth et al., 2017). The timing of the SSEP peaks did not change with time (Fig. 3f), indicating stable physiological response to stimulation. The potential for sensory feedback from neuroprostheses is apparent but further development of technologies are needed to address stability on the microscopic scale. Overall, micromotion and changes at the electrode interface due to biotic and abiotic factors may have a strong effect on the selectivity of neural stimulation over long-term for these EAs.
3.7. Scanning electron microscopy Mechanical failure of the EA tips after extraction was studied using SEM (Fig. 6a). Some of the electrode tips were observed to be fractured (Fig. 6b), which could be a result of implantation. A high magnification of intact electrode tip is shown in Fig. 6c. Four out of the six EAs were observed with a small number of fractured electrode tips. 4. Discussion The need to improve intuitive prosthesis control is leading to innovation in PNI design and implementation (Ghafoor et al., 2017; Jung et al., 2018). Development of chronic animal models to assess the safety and performance of PNIs is critical for determining failure modes at an early stage of device development and for successful clinical translation of these new technologies. We previously published a rodent model to assess the safety and performance of recording electrodes (Vasudevan et al., 2017), which laid the foundation for the current study. The rat model has been extensively studied in the field of peripheral nerve repair and has well characterized functional and histological outcome measures (Vasudevan et al., 2017, 2013; Hunter et al., 2007; Deuis et al., 2017; Sarikcioglu et al., 2009). These measures for non-clinical assessment can be used to evaluate longitudinal performance of PNIs, and to compare across PNIs of different form factors or modalities. This model also opens the door for understanding histological, electrophysiological and material failure modes, which can be rectified before finalizing device design for clinical translation. To demonstrate the ability to perform long-term assessment, we designed a robust connector mount system as shown in Fig. 1, that allowed for stable access to the implants during the study. This component is critical to maintain the integrity of the lead wires as they undergo repeated mechanical stress during chronic testing (Vasudevan et al., 2017; Straka et al., 2018). In four out of six animals, the bottom part of the connector mount, in contact with tissue, was 3D printed with Nylon, and further softened with Kwik-cast to mitigate potential discomfort to the animal. The top part of the connector mount was 3D printed with bronze infused stainless steel to prevent damage to the connectors inflicted by the animal, cage components or user handling. Finally, the use of Mersilene mesh to support tissue ingrowth after implantation contributed to wound healing and stability (Vasudevan et al., 2017; Straka et al., 2018). The surgical method described here was previously developed in our lab to study floating microelectrode arrays and accounted for device integrity from the transcutaneous mount down to the PNI (Vasudevan et al., 2017). The lumbar fascia is used secure the connector mount, while providing easy access to the implanted electrodes through connectors (Fig. 2a and e). Accessing the two implants (EA and EMG) over long-term via the connector mount did not produce infections at the surgical sites, which is a major concern involving transcutaneous implants (Chou et al., 2010). Critical to the success of this procedure is securing the electrodes in the nerve as seen in Fig. 2c using fibrin sealant. The fibrin sealant forms a stable construct immediately upon application, preventing the EA from dislodging. At later points in the healing process, electrode stabilization is augmented by fibrotic encapsulation, and presumably limits electrode dislodgment during locomotion (Vasudevan et al., 2017; Straka et al., 2018). The EMG wire electrode fixation technique (Fig. 2d) was reliable, as demonstrated by the stability in the location of the recording tips inside the muscles during the experiments. The most common failure observed with the EA was lead wire failure at the connector mount interface, which is a region exposed to high mechanical stress. Lead wire breakage continues to be a hurdle to chronic experimentation in both animal and humans (Vasudevan et al., 2017; Davis et al., 2016). Nevertheless, surgical procedures described here provides implant stability for long-term assessment of PNIs in the sciatic nerve using a rodent model. Impedance of the nerve electrodes at 1 kHz before and after the 8
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
their assistance with histomorphometric analysis.
Functional testing allowed the quantification of both sensory and motor function of the target nerve. We conducted walking track analysis to monitor motor function and Von Frey testing to monitor sensory sensation. Results show that motor and sensory function did not change significantly and returned to pre-implantation levels within 2–4 W postimplantation (Fig. 4a–c). These results are necessary to conclude if a device causes significant nerve damage making the PNI unsuitable for long-term use. It is also important to note that the surgical procedures did not cause irreversible target nerve damage. In addition, it seems that electrical stimulation did not impact function over time, while the parameters and duration of stimulation used in this study could contribute to this observation. Although this experiment was not specifically designed to assess functional effects of stimulation, future work using this platform could isolate the relationship between stimulation and nerve function. Histomorphometric analysis is commonly used to assess the health and distribution of myelinated nerve fibers (Hunter et al., 2007). Histomorphometric analysis of the distal nerve segment showed comparable fiber density, while there was significant difference in myelin area and distribution of fibers between 4–6 μm fibers (Fig. 5). This difference in myelin area and fiber distribution could be a result of either the surgical procedures and/or the presence of EA inside the nerve. While there may have been axon damage during implantation, histomorphometric analysis at 13 weeks post-implantation shows that the presence of EA did not prevent the recovery of fiber density. It is important for an EA to have minimal nerve injury during surgery, but it is even more important for it to not induce complications associated with obstruction of regenerating nerve fibers. SEM can be used to evaluate factors leading to electrode failure (Vasudevan et al., 2017). SEM of the explanted electrodes revealed mechanically damaged electrode tips (Fig. 6b), which could be attributed to the implantation procedures. Breakage at the tip can alter device recording and stimulation performance, rendering the use of a PNI to be suboptimal for long-term use. Changes in materials specifically designed for peripheral nerve applications and refining surgical procedures could potentially mitigate this issue. Overall, the comprehensive test platform described here has allowed for detailed characterization of electrode safety and performance, which can provide information to steer future improvements to PNI design and fabrication.
References Bain, J.R., Mackinnon, S.E., Hunter, D.A., 1989. Functional evaluation of complete sciatic, peroneal, and posterior tibial nerve lesions in the rat. Plast. Reconstr. Surg. 83, 129–138. Chou, T.G., Petti, C.A., Szakacs, J., Bloebaum, R.D., 2010. Evaluating antimicrobials and implant materials for infection prevention around transcutaneous osseointegrated implants in a rabbit model. J. Biomed. Mater. Res. A 92, 942–952. Christensen, M.B., Pearce, S.M., Ledbetter, N.M., Warren, D.J., Clark, G.A., Tresco, P.A., 2014. The foreign body response to the Utah Slant Electrode Array in the cat sciatic nerve. Acta Biomater. 10, 4650–4660. Christie, B.P., Freeberg, M., Memberg, W.D., Pinault, G.J.C., Hoyen, H.A., Tyler, D.J., Triolo, R.J., 2017. Long-term stability of stimulating spiral nerve cuff electrodes on human peripheral nerves. J. Neuroeng. Rehabil. 14, 70. Clark, G.A., Ledbetter, N.M., Warren, D.J., Harrison, R.R., 2011. Recording sensory and motor information from peripheral nerves with Utah slanted electrode arrays. Conf. Proc. IEEE Eng. Med. Biol. Soc. 2011, 4641–4644. Davis, T.S., Wark, H.A., Hutchinson, D.T., Warren, D.J., O’Neill, K., Scheinblum, T., Clark, G.A., Normann, R.A., Greger, B., 2016. Restoring motor control and sensory feedback in people with upper extremity amputations using arrays of 96 microelectrodes implanted in the median and ulnar nerves. J. Neural Eng. 13, 36001. Deuis, J.R., Dvorakova, L.S., Vetter, I., 2017. Methods used to evaluate pain behaviors in rodents. Front. Mol. Neurosci. 10, 284. Dhillon, G.S., Horch, K.W., 2005. Direct neural sensory feedback and control of a prosthetic arm. IEEE Trans. Neural Syst. Rehabil. Eng. 13, 468–472. Dhillon, G.S., Lawrence, S.M., Hutchinson, D.T., Horch, K.W., 2004. Residual function in peripheral nerve stumps of amputees: implications for neural control of artificial limbs. J. Hand Surg. 29, 605–615 discussion 16-8. Fahrenkopf, M.P., Adams, N.S., Kelpin, J.P., Do, V.H., 2018. Hand amputations. Eplasty 18, ic21. Farrokhi, S., Perez, K., Eskridge, S., Clouser, M., 2018. Major deployment-related amputations of lower and upper limbs, active and reserve components, U.S. Armed Forces, 2001-2017. MSMR 25, 10–16. Fisher, J.A., Huang, S., Ye, M., Nabili, M., Wilent, W.B., Krauthamer, V., Myers, M.R., Welle, C.G., 2016. Real-time detection and monitoring of acute brain injury utilizing evoked electroencephalographic potentials. IEEE Trans. Neural Syst. Rehabil. Eng. 24, 1003–1012. Ghafoor, U., Kim, S., Hong, K.S., 2017. Selectivity and longevity of peripheral-nerve and machine interfaces: a review. Front. Neurorobot. 11, 59. Graczyk, E.L., Schiefer, M.A., Saal, H.P., Delhaye, B.P., Bensmaia, S.J., Tyler, D.J., 2016. The neural basis of perceived intensity in natural and artificial touch. Sci. Transl. Med. 8, 362ra142. Gunter, C., Delbeke, J., Ortiz-Catalan, M., 2019. Safety of long-term electrical peripheral nerve stimulation: review of the state of the art. J. Neuroeng. Rehabil. 16, 13. Hong, K.S., Aziz, N., Ghafoor, U., 2018. Motor-commands decoding using peripheral nerve signals: a review. J. Neural Eng. 15, 31004. Howland, R.H., 2014. Vagus nerve stimulation. Curr. Behav. Neurosci. Rep. 1, 64–73. Hunter, D.A., Moradzadeh, A., Whitlock, E.L., Brenner, M.J., Myckatyn, T.M., Wei, C.H., Tung, T.H., Mackinnon, S.E., 2007. Binary imaging analysis for comprehensive quantitative histomorphometry of peripheral nerve. J. Neurosci. Methods 166, 116–124. Jung, R., Abbas, J.J., Kuntaegowdanahalli, S., Thota, A.K., 2018. Bionic intrafascicular interfaces for recording and stimulating peripheral nerve fibers. Bioelectron. Med. (Lond.) 1, 55–69. Milby, A.H., Halpern, C.H., Baltuch, G.H., 2008. Vagus nerve stimulation for epilepsy and depression. Neurotherapeutics 5, 75–85. Moxey, P.W., Gogalniceanu, P., Hinchliffe, R.J., Loftus, I.M., Jones, K.J., Thompson, M.M., Holt, P.J., 2011. Lower extremity amputations—a review of global variability in incidence. Diabet. Med. 28, 1144–1153. Navarro, X., Krueger, T.B., Lago, N., Micera, S., Stieglitz, T., Dario, P., 2005. A critical review of interfaces with the peripheral nervous system for the control of neuroprostheses and hybrid bionic systems. J. Peripher. Nerv. Syst. 10, 229–258. Oddo, C.M., et al., 2016. Intraneural stimulation elicits discrimination of textural features by artificial fingertip in intact and amputee humans. Elife 5, e09148. Polasek, K.H., Hoyen, H.A., Keith, M.W., Kirsch, R.F., Tyler, D.J., 2006. Spiral nerve cuff electrodes for an upper extremity neuroprosthesis. Conf. Proc. IEEE Eng. Med. Biol. Soc. 1, 3584–3587. Raspopovic, S., et al., 2014. Restoring natural sensory feedback in real-time bidirectional hand prostheses. Sci. Transl. Med. 6, 222ra19. Rijnbeek, E.H., Eleveld, N., Olthuis, W., 2018. Update on peripheral nerve electrodes for closed-loop neuroprosthetics. Front. Neurosci. 12, 350. Rossini, P.M., et al., 2010. Double nerve intraneural interface implant on a human amputee for robotic hand control. Clin. Neurophysiol. 121, 777–783. Rossini, P.M., Rigosa, J., Micera, S., Assenza, G., Rossini, L., Ferreri, F., 2011. Stump nerve signals during transcranial magnetic motor cortex stimulation recorded in an amputee via longitudinal intrafascicular electrodes. Exp. Brain Res. 210, 1–11. Sarikcioglu, L., Demirel, B.M., Utuk, A., 2009. Walking track analysis: an assessment method for functional recovery after sciatic nerve injury in the rat. Folia Morphol. (Warsz) 68, 1–7. Shafer, B., Yaghouby, F., Vasudevan, S., 2018. Short-time fourier transform based spike detection of spontaneous peripheral nerve activity. Conf. Proc. IEEE Eng. Med. Biol. Soc. 2018, 2418–2421. Straka, M.M., Shafer, B., Vasudevan, S., Welle, C., Rieth, L., 2018. Characterizing
5. Conclusion We have advanced a previously developed test platform to assess the safety and performance of intraneural PNIs using a rodent model. The current test platform allows for long-term assessment up to 12 weeks of implantation. The ability to conduct these assessments on PNI devices provides a comprehensive understanding of reliability and failure modes in the non-clinical setting, which is critical to improve upon device design for novel PNI technologies. Disclaimer The mention of commercial products, their sources, or their use in connection with material reported herein is not to be construed as either an actual or implied endorsement of such products by the Department of Health and Human Services. Acknowledgements This research project was supported by the Defense Advanced Research Projects Agency (DARPA), Biological Technologies Office (BTO), Hand Proprioception and Touch Interfaces (HAPTIX) Program, through an Interagency Agreement with the US Food and Drug Administration. The authors would like to acknowledge the Orthopedic Nerve Research Laboratory at Washington University and Jason Wever for 9
Journal of Neuroscience Methods 328 (2019) 108437
B. Shafer, et al.
D.J., Hutchinson, D.T., Clark, G.A., 2017. Restoration of motor control and proprioceptive and cutaneous sensation in humans with prior upper-limb amputation via multiple Utah Slanted Electrode Arrays (USEAs) implanted in residual peripheral arm nerves. J. Neuroeng. Rehabil. 14, 121. Wurth, S., et al., 2017. Long-term usability and bio-integration of polyimide-based intraneural stimulating electrodes. Biomaterials 122, 114–129. Ziegler-Graham, K., MacKenzie, E.J., Ephraim, P.L., Travison, T.G., Brookmeyer, R., 2008. Estimating the prevalence of limb loss in the United States: 2005 to 2050. Arch. Phys. Med. Rehabil. 89, 422–429.
longitudinal changes in the impedance spectra of in-vivo peripheral nerve electrodes. Micromachines (Basel) 9. Vasudevan, S., Yan, J.G., Zhang, L.L., Matloub, H.S., Cheng, J.J., 2013. A rat model for long-gap peripheral nerve reconstruction. Plast. Reconstr. Surg. 132, 871–876. Vasudevan, S., Patel, K., Welle, C., 2017. Rodent model for assessing the long term safety and performance of peripheral nerve recording electrodes. J. Neural Eng. 14, 16008. Vasudevan, S., Vo, J., Shafer, B., Nam, A.S., Vakoc, B.J., Hammer, D.X., 2019. Toward optical coherence tomography angiography-based biomarkers to assess the safety of peripheral nerve electrostimulation. J. Neural Eng. Wendelken, S., Page, D.M., Davis, T., Wark, H.A.C., Kluger, D.T., Duncan, C., Warren,
10