Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Dec 15.
Published in final edited form as: Biosens Bioelectron. 2020 Sep 22;170:112608. doi: 10.1016/j.bios.2020.112608

Intraneural ultramicroelectrode arrays for function-specific interfacing to the vagus nerve

Atefeh Ghazavi a,1, Maria A González-González b,1, Mario I Romero-Ortega b, Stuart F Cogan a,*,1
PMCID: PMC7654841  NIHMSID: NIHMS1635572  PMID: 33035896

Abstract

Selective interfacing to small multifunctional nerves such as the vagus nerve (VN) which is the main multimodal autonomic nerve that provides a major communication pathway from vital peripheral organs to the brain, can have significant potential in treating and diagnosing diseases as well as enhancing our understanding of peripheral nerve circuits. Here we describe the fabrication of a 16-channel intraneural electrode array with ultramicro-dimensioned electrodes to achieve improved functionally selective recording. We demonstrate that the amorphous silicon carbide ultramicroelectrode arrays (a-SiC UMEAs) provide selectivity in the detection of neural activity in the cVN related to changes in systemic oxygenation and blood pressure. We will also demonstrate spatially selective recording of micro-compound action potentials (μCAPs) by electrical stimulation of the subdiaphragmatic branches of the VN. Distinct neural activity was recorded on electrodes separated by less than about 100 μm. This is the first time that this level of spatially selectivity recording has been demonstrated in the cVN with an intraneural multielectrode array.

Keywords: Microelectrode array, Vagus nerve, Selective neural recording, Intraneural, Peripheral nerve interface, Ultramicroelectrode

1. Introduction

Multifunctional nerves in the peripheral nervous system sense and control the activity of different target organs, where subsets of axons modulate single organ function. The vagus nerve (VN) contains mostly parasympathetic afferents from liver, spleen, pancreas, kidney, intestine, heart, lung, stomach, and pharynx. Despite such multi-functionality, global stimulation of the cervical VN (cVN) with cuff electrodes has been used clinically, with varying degrees of success, for the treatment of epilepsy, heart failure, and depression, but suffers from off-target effects such as dysphonia, pain, and unwanted recruitment of the laryngeal muscles (Ben-Menachem, 2001; De Ferrari et al., 2017; Handforth et al., 1998). Clinically, nerve cuffs provide an extraneural interface to the cVN with electrodes that individually circumscribe a significant portion of the nerve diameter (Usami et al., 2013). Recent advances in extraneural electrode interfaces to improve selectivity within the cVN include multichannel cuff electrodes (Plachta et al., 2014), flexible cuff-like microelectrodes (Caravaca et al., 2017), twining nerve electrodes (Li et al., 2017; Zhang et al., 2019), flexible neural clips (Lee et al., 2017), wrappable microwire electrodes (Falcone et al., 2018), and multichannel nerve electrodes (Xue et al., 2018). Given that the extraneural implants are wrapped around the nerve, electrical stimulation is expected to have limited selectivity and affect multiple functions, although current steering with multichannel cuff electrodes has been shown to partially address this limitation (González-González et al., 2018). Likewise, extraneural recording electrodes have limited capability to selectively record from function–specific fibers within the nerve. Higher selectivity for both stimulation and recording is expected with intraneural electrodes that place active electrode sites within the nerve. Several intraneural electrode designs have been investigated for this purpose including the High Density Utah Slanted Electrode Array (HD-USEA)(Wark et al., 2013), the Transverse Intraneural Microelectrode Array (TIME)(Boretius et al., 2010), and the Longitudinal Intrafascicular Electrode (LIFE)(Navarro et al., 2007). However, their use in smaller nerves has been limited by their relatively large size (2.2 × 104 μm2−4 × 106 μm2 array size, 110 μm–280 μm shank width/diameter, and 200 μm–600 μm electrode pitch). Thus, functionally selective intraneural interfaces for small nerves, particularly autonomic nerves, need further development. Currently, the only functional intraneural electrode with dimensions suitable for interfacing with the cVN is the carbon nanotube (CNT) yarn electrode (McCallum et al., 2017). This CNT electrode elicited limited foreign body response due to the small fiber size (17 μm diameter), and reduced tissue encapsulation compared to LIFE and TIME arrays, four months post implantation (Stice et al., 2007). However, the intraneural CNT fiber electrodes have a relatively large geometric surface area (GSA) of 15700 μm2 (10 μm diameter, 500 μm length) (McCallum et al., 2017) and given the small diameter of the cVN fibers in our rat model, ranging from 0.6 μm to 2.7 μm (Pianca et al., 2015; Soltanpour and Santer, 1996), stimulation or recording with these electrodes will likely involve large axonal volumes (i.e., thousands of axons) and thus have limited functional selectivity.

In order to improve the selectivity of intraneural electrodes in small nerves, we have developed a 16-channel intraneural array based on amorphous silicon carbide (a-SiC UMEA). The a-SiC UMEA is designed to provide high spatial selectivity for stimulation and recording, while reducing insertion trauma and foreign body response. Our initial target for this device was the rat cVN, a multifunctional nerve of approximately 400 μm in diameter (Woodbury and Woodbury, 1991) composed of 188000–229000 axonal fibers with the majority being unmyelinated (~4:1) (Soltanpour and Santer, 1996). The a-SiC UMEA has sixteen electrodes, two each on eight shanks with each shank having cross-sectional dimensions of 23 μm by 10 μm. The shank lengths vary from 335 μm to 535 μm. The small cross-sectional shank dimensions provide mechanical flexibility and minimize the volume of tissue displaced by implantation and thus the invasiveness of the intraneural array. Each shank has two ultramicroelectrode contacts each with a GSA of 200 μm2 (8 × 25 μm2) (Fig. 1b–d).

Fig. 1. 16-channel amorphous silicon carbide ultramicroelectrode array.

Fig. 1.

Open in a new tab

a, Summary of the fabrication process and different material layers forming the array. b, Top optical view of a 10 μm thick array during the assembly process. The pins at the two ends of the connector are attached to the reference (left) and ground (right) wires. Scale bar, 2 mm c, Scanning electron micrograph (SEM) of the shanks. Scale bar, 100 μm c’, Magnified image of an electrode site. Scale bar, 20 μm d, Design and dimensions of the a-SiC UMEA (Magnifications are color-coded). e, SEM image of the Neurocase structure. Scale bar, 1 mm f, Sideview image of the array representing its flatness after annealing at 400 °C. Scale bar, 2 mm.

Here, we report on the electrochemical and mechanical properties of the a-SiC UMEAs and demonstrate their use as an intraneural interface for selective recording in the rat cVN. The unique spatial resolution of the a-SiC UMEA also has the potential for selective functional modulation or monitoring of single organ function from multi-functional nerves. This capability should allow selective intraneural targeting with vagus nerve stimulation (VNS) and the optimization of treatments for epilepsy, depression and obesity, while reducing the unwanted side effects associated with extraneural VNS. In addition, high special selectivity for recording and stimulation in peripheral nerves may be generally pertinent to a variety of applications that include limb pros-theses (Davis et al., 2016), electroceuticals (Horn et al., 2019), and functional nerve mapping (Zanos, 2019).

2. Materials and methods

2.1. A-SiC UMEA fabrication and assembly

A-SiC UMEAs were fabricated using standard thin-film deposition methods and patterning by UV-photolithography. Briefly, the base structure of the device was an 8 μm thick layer of amorphous silicon carbide (a-SiC) deposited onto a silicon carrier wafer using plasma enhanced chemical vapor deposition (PECVD) (PlasmaTherm Unaxis 790 Series) at 350 °C, 270 W power, 1000 mtorr pressure, and 164 sccm Ar, 600 sccm SiH4/Ar, 36 sccm CH4 gas flow rates. A thin layer (~1 μm) of polyimide (PI) (HD Microsystems, PI2610) was used between the silicon wafer and a-SiC to facilitate the release of the UMEAs from the wafer after fabrication is complete. Array metallization compromised a tri-layer of Ti/Au/Ti (approximate thickness 40nm/300nm/40 nm) deposited by e-beam evaporation (CHA Mark 50E evaporator) and patterned by liftoff photolithography. The liftoff process used for patterning the metal employed a double layer resist: liftoff-resist (LOR5A, Kayaku Advanced Materials Inc. Marlborough, MA) and positive photoresist (Microposit S1805, Kayaku Advanced Materials Inc. Marlborough, MA). A second, 2 μm thick, layer of a-SiC was deposited over the metallization and base a-SiC layer, and vias exposing bond pads and electrode sites were opened in the second a-SiC layer by inductively coupled plasma-reactive ion etching (ICP-RIE) (Versaline ICP PSU, Plasma-Therm) with a SF6/O2 plasma.

Besides the advantages associated with a-SiC as the base and encapsulation material, PECVD a-SiC films have intrinsic residual stress which is usually compressive (Jean et al., 1992), and if not properly controlled can introduce curvature into the UMEA shanks, hindering insertion of the array into the nerve. To moderate the effects of the intrinsic compressive stress of a-SiC, the PI/a-SiC/Au/a-SiC structures were annealed at 400 °C, while still on the silicon wafer. Annealing reduces the overall stress of the multilayer structure by decreasing the hydrogen content of the a-SiC (Deku et al., 2019) and also acts to relieve stress in the gold metallization that is deposited with an intrinsic tensile stress in the e-beam evaporation process that we employed. The annealing process was effective in producing arrays with straight shanks having low or balanced tensile and compressive stresses (Fig. 1f), which enables insertion of the array into the nerve without bending or deflection of the shanks.

Following the annealing, a low impedance coating of sputtered iridium oxide (SIROF) was deposited onto the electrode sites in a DC sputtering system (AJA International, Scituate, MA), using the following process parameters: 30 mtorr pressure, 100 W power, 15 sccm Ar, 7.5 sccm O2, and 22.5 sccm water vapor, and patterned by lift off (Maeng et al., 2019). In order to singulate the devices, SF6/O2 and O2 plasmas in an ICP RIE system were used to etch a-SiC and PI films, respectively. Subsequently, the wafers were soaked in deionized water to facilitate the release of the fabricated devices from the silicon wafer.

After the devices were released, the assembly process of the device required connecting the UMEA structure to a 16-channel Omnetics connector (Omnetics Connector Corp., Minneapolis, MN), and attaching ground and reference wires to the array and encapsulating the connector (Fig. 1b). The 16-channel Omnetics connector was attached to the array by reflow soldering using indium alloy solder paste (Indalloy 1E 52In/48Sn, Indium Corporation, Clinton, NY). The flux residue was removed and the ground and reference wires were attached to the array with conductive silver epoxy (Epo-TEK H20E, Epoxy Technology Inc., Billerica, MA). In order to strengthen the connection and encapsulate the connector, a biocompatible epoxy with high moisture and heat resistance, and low curing temperature (Med-302, Epoxy Technology Inc., Billerica, MA) was employed. In order to increase the stiffness of the device and assure uniaxial insertion forces, a 127 μm thick laser-cut Kapton sheet (Cole-Parmer, IL, USA) was attached to the backside of the bond pad area of the device but not the shanks that insert into the nerve.

2.2. Characterization of a-SiC UMEAs

The a-SiC UMEAs were electrochemically characterized by cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS) and voltage transient (VT) measurements in air-equilibrated phosphate buffered saline electrolyte (PBS) at room temperature. The PBS had a composition of 126 mM NaCl, 22 mM NaH2PO4–7H2O, and 81 mM Na2HPO4–H2O at pH 7.2–7.4 (Cogan et al., 2007). Prior to characterization, to rejuvenate the SIROF (Cogan et al., 2014), the electrodes were subjected to 3 hours of continuous slow sweep rate (50 mV/s) CV cycles (200 cycles) in PBS between potential limits of −0.6 V and 0.8 V versus Ag|AgCl using a Reference 600+ potentiostat (Gamry Instruments, Warminster PA). The averaged voltammetric response of the SIROF electrodes (n = 16) changes notably before and after the 200 CV rejuvenation cycles (Supplementary data, Fig. S1). This cycling step was performed in order to electrochemically clean the electrodes of any contamination remaining from the fabrication process. Furthermore, as the SIROF electrodes are exposed to temperatures as high as 100–200 °C throughout the assembly process, it is necessary to rehydrate the film to recover the charge storage and charge-injection properties of SIROF for neural stimulation and recording.

To evaluate the charge storage capacity, CVs were performed between potential limits of −0.6 V and 0.8 V versus Ag|AgCl reference electrode with a scan rate of 50 mV/s. The cathodal charge storage capacities were calculated from the time integral of the negative current during a complete CV cycle. Furthermore, the total capacitive leakage current during CV measurements of 16-channel a-SiC UMEAs, between the electrode traces, was evaluated (Fig. S2). The EIS experiments were performed in PBS at room temperature using a sinusoidal voltage with 10 mV rms amplitude over a frequency range of 1 Hz-100 kHz.

Charge injection capacities were determined by voltage transient measurements using a Plexstim stimulator (Plexon Inc., Dallas, TX). Cathodic first biphasic current pulses were applied to the electrode at a rate of 50 pulse per second. The electrode was discharged during the interpulse period (counter and working electrodes shorted). The electrode potential 12 μs after the termination of the first phase of the pulse determined the electrode polarization. The maximum charge injection capacity was defined as the charge at which the electrode polarization was −0.6 V with respect to the platinum (Pt) counter electrode, normalized by the electrode surface area.

Mechanical analysis of the UMEA shanks was performed by measuring the critical buckling load of nonfunctional devices. Critical buckling loads were determined by measuring longitudinal force along electrode shanks as the electrode array was advanced, with a constant speed of 10 μm/s, towards a robust, high friction substrate (silicon wafer coated with a 5 μm-thick layer of silicone). The force was measured with a LSB200 load cell (Futek, Irvine, CA) and the array was advanced with hydraulically-controlled micropositioner (KOPF Instruments 2650, Tujinga, CA).

In order to predict the buckling force thresholds for different thicknesses of the same shank design, a 3D finite element model (FEM) of the longest shank was developed (COMSOL v5.2, COMSOL Inc., Burlington, MA). The shank was composed of a 1 μm-thick PI layer and an a-SiC layer with variable thicknesses (4 μm, 6 μm, 8 μm, 10 μm, 12 μm). The shank had a 536 μm length and 23 μm width at the base. The geometry was meshed with 671,563 tetrahedral elements and the COMSOL structural mechanics module was used to perform a linear buckling analysis. Elastic moduli of 8.5 GPa and 64 GPa and Poisson ratios of 0.34 and 0.19 (El Khakani et al., 1994) were assigned to the PI and a-SiC layers, respectively. The Young’s modulus (64 GPa) that provided the best fit between the measured and numerical buckling force thresholds for a 6 μm-thick a-SiC probe was used to numerically calculate the buckling force thresholds for the rest of the probes. The measured buckling force threshold for 8 μm-thick a-SiC probe matched well with the numerical results. One end of the shank was considered fixed and the other end free and a load was applied to the free end of the shank. A second FEM model was developed to evaluate the bending stiffness of the shanks. The same geometry, material properties, and mesh was used for the second model. One end of the shank was fixed, and a force perpendicular to the axis of the shank, was applied to the other end. The applied force was divided by the resulting displacement to calculate the bending stiffness of the shank.

2.3. Animal use

2.3.1. Ethics statement

Protocols and surgical procedures were designed to prevent animal discomfort at all times. The procedures were approved by the Institutional Animal Care and Use Committee protocol (IACUC, 14–09) at The University of Texas at Dallas, which also follow the National Institute of Health (NIH) guidelines.

2.3.2. Surgical procedures

A total of 8 Sprague-Dawley male rats (300 g–350 g; Charles River, Wilmington, MA) were used for the experiments. The animals were anesthetized with vaporized isoflurane (2%) in a constant oxygen flow (2 L/min) delivered by a calibrated vaporizer and maintained throughout the experiment. A warm pad was used to maintain body temperature and vital signs were monitored throughout the experiment.

To expose the cVN, a longitudinal medial incision was made in the anterior part of the neck at the cervical level, using the anterior sternum as reference. The sternomastoid muscle was separated in oblique orientation to the midline. The cVN was identified running along with the lateral portion of the carotid artery. To access the subdiaphragmatic vagus nerve (SDVN), a midline incision (2.5 cm) was made on the abdominal wall. The stomach was carefully manipulated to expose the esophagus, and the SDVN trunks were identified between the diaphragm and the gastric cardia.

For systemic drug delivery, the femoral vein was cannulated. A 1.0 cm–1.5 cm incision in the midline of the leg allowed the exposure of the femoral vein. A small incision was made to insert a cannula (0.6 mm outer diameter) previously filled with heparinized saline (20 IU/mL) and coupled to an infusion system in order to administrate phenylephrine (100 μg/mL). The same procedure was employed in the femoral artery to continuously monitor the blood pressure, with a canula filled with heparin. The arterial canula was connected to a calibrated pressure transducer (MLT1199, ADInstruments, Colorado Springs, CO), which was coupled to a bridge amplifier and power supply module (FE221 and ML826, respectively, ADInstruments, Colorado Springs, CO). PowerLab data acquisition system (ADInstruments, Colorado Springs, CO) and LabChart Pro software (ADInstruments, Colorado Springs, CO) were used for recording the blood pressure data.

The cVN is attached to the carotid artery and the pulsatile movements of the artery complicate the implantation of the electrode arrays. For this study, we designed a temporary support device, the “Neurocase”, that supports the VN ventrally, greatly reducing movement during UMEA implantation (Supplementary-Video S2). The Neurocase was fabricated in SU8 photoepoxy (Microchem Corp. Marlborough, MA) by multilayer photolithography and provides a channel for placement of the cVN. Scanning electron microscopy (SEM) images of the fabricated array and the Neurocase structure are shown in Fig. 1c and e, respectively.

2.4. Neural recording

Neural activity in the cVN was recorded using an Omniplex data acquisition system (Plexon, Dallas TX). Spike sorting was performed by projecting the detected spikes onto the 2D feature space (features being the first two principle components) and using the K-means algorithm for clustering (Offline Sorter software, Plexon Inc., Dallas, TX). Further analysis of the sorted units was performed using NeuroExplorer software (Nex Technologies, Colorado Springs, CO).

2.5. Histological analysis

The ultrastructure visualization of the cVN was assessed by transmission electronic microscopy (TEM, n = 2) (Fig. 3e). Tissue fixation was obtained by intracardiac perfusion of physiological saline (NaCl, 0.9%), followed by fixative solution that consisted of 4% formaldehyde, 0.5% glutaraldehyde, diluted in sodium arsenate (cacodylate) buffer (0.1 M, pH 7.4). The samples were post-fixated in 3.0% glutaraldehyde in cacodylate buffer (0.1 M pH 7.4), at 4 °C and processed for resin embedding and ultrathin sectioning. 60–70 nm ultrathin sections were obtained, mounted on copper grids, and contrasted with uranyl acetate and silver citrate. The sections were analyzed and imaged by TEM (JEOL-JEM-1010 model), at 80 kV. A subset of cVN tissue was collected after the first fixation to process for cryosection as described elsewhere (González-González et al., 2018) and 25 μm sections where obtained in a cryostat to analyze electrode traces and contrast to array dimensions (Fig. 3d).

Fig. 3. A-SiC UMEA implantation in the cVN.

Fig. 3.

Open in a new tab

Schematic representation of a, The A-SiC UMEA with electrode numbers on the array and b, Neurocase structure as a stabilizing tool for in vivo implantation. The cervical vagus nerve (cVN) is represented in pink and a side view is represented in b′, including the a-SiC UMEA. c, Photograph of intraneural implantation of the a-SiC UMEA in the cVN stabilized with the Neurocase. Scale bar, 1 mm. d, Overlay of the a-SiC UMEA on the cVN cross section to show the respective array and nerve dimensions. Scale bar, 100 μm. e, Schema of a single electrode site (200 μm2) (left) and magnified TEM image of ~44 μm2 of the cVN cross section containing 21 myelinated and 160 unmyelinated axons (yellow and red arrowheads, respectively). Scale bar, 1 μm. f, Neural activity in response to hypoxia and changes in blood pressure was recorded with an a-SiC UMEA implanted in the rat left cervical vagus nerve (recording site). Elevation of blood pressure is primarily sensed by the carotid sinus baroreceptors which communicate with centers in the medulla via the glossopharyngeal nerve (blue arrow). Blood pressure changes are also transmitted to the medulla via left vagus afferent signals from baroreceptors in the aortic arch (not shown), but at a much lower sensitivity than that from the carotid sinus. In response to an increase in blood pressure, parasympathetic efferents (blue arrows) from the medulla are transmitted along both branches of the vagus nerve to the heart, inhibiting cardiac output. Changes in the blood oxygen levels (PO2) is sensed by the carotid sinus chemoreceptors and transmitted to the medulla via glossopharyngeal nerve. Changes in respiration in response to hypoxia are transmitted as vagal afferents (blue arrows) from pulmonary stretch receptors to the medulla and apneustic center of the pons (not shown). Parasympathetic vagal efferents are also involved in modulating airway caliber although the sympathetic pathways are species dependent. Hypoxia also increases the cardiac vagal efferent activity (for more details see Canning and Fischer, 2001; Feldman et al., 2014; Fukuda et al., 1989; Olshansky et al., 2008) (blue arrows = afferent, red arrows = efferent).

3. Results

The structure of the 16-channel a-SiC UMEA is shown in Fig. 1. The a-SiC UMEAs have eight nerve-penetrating shanks varying in length from 335 μm to 535 μm. The shanks have two electrode sites arranged at different positions along the length of the shank. These shanks have 68 μm pitch, spanning a distance of 500 μm, which is the maximum longitudinal nerve segment that can be accessed with this particular design. Since the electrode sites are small (nominal GSA of 200 μm2), and such small electrode size is associated with high impedance and low charge injection capacity for stimulation, the electrode sites were coated with SIROF. Fig. 1a shows the different material layers that form the a-SiC UMEA devices.

The average impedance spectrum of 16, 200 μm2 SIROF ultra-microelectrodes over a frequency range of 1 Hz–100 kHz in phosphate buffered saline (PBS) at pH 7.2 is shown in Fig. 2a. At a frequency of 1 kHz, the electrodes have an average impedance magnitude of 22.8 ± 0.6 kΩ (mean ± std. dev., n = 16).

Fig. 2. Electrochemical characterization and buckling force measurements of a 16-channel a-SiC UMEA with SIROF-coated electrode sites.

Fig. 2.

Open in a new tab

a, Average impedance spectra of 16 SIROF electrodes on an a-SiC UMEA in PBS. The average electrode impedance at 1 kHz is 22.8 ± 2.4% kΩ. b, Average cyclic voltammogram of 16 SIROF electrodes in PBS at a sweep rate of 50 mV/s with cathodal charge storage capacity of 68 ± 1.9% mC/cm2. The two redox couples correspond to the peaks observed in the CV graph: Ir3+/Ir4+ (Cogan et al., 2008) and Ir4+/Ir5+ (Steegstra et al., 2013). c, Average voltage transient of 16 SIROF electrodes of an a-SiC UMEA in PBS in response to a 200 μs cathodal first, symmetrical biphasic rectangular current pulse. The average charge injection capacity (Qinj) is 3.7 ± 2.5% mc/cm2 with Emc at the cathodal limit of −0.6 V and zero interpulse potential with respect to a platinum counter electrode. Data are plotted as mean ± standard deviation (SD) (n = 16) (where error bars are not shown, SD is smaller than the symbol size). d, Measured critical load versus time using a nonfunctional a-SiC UMEA (8-shank, 7 μm thick, and 23 μm shank-width). e, Single shank geometry meshed with tetrahedral elements (left) and the modeling results representing the critical load and displacement of a shank (normalized to the maximum displacement). The shank dimensions are: 536 μm length, 23 μm width, 9 μm thickness (8 μm a-SiC and 1 μm PI). f, Image of the buckled shanks corresponding to the graph in panel d (note that reflection of the array on the silicon substrate is also present at the bottom of the image). g, Modeling results of buckling force threshold (green circles) and stiffness (blue squares) of shanks made of 1 μm thick PI layer and different thicknesses of a-SiC layer. The black triangle represents the experimentally measured buckling force threshold value.

The cathodal charge storage capacity (CSCc) of the electrodes determined from CV measurements was 68 ± 1 mC/cm2 (mean ± std. dev., n = 16). As shown in the SIROF CV response (Fig. 2b), two redox couples, Ir+3/Ir4+ (Cogan et al., 2008) and Ir4+/Ir5+ (Steegstra et al., 2013), are observed with corresponding oxidation and reduction peaks at 0.21 V/0.11 V and 0.61 V/0.51 V, respectively.

To determine maximum cathodal charge injection capacities, the electrodes were subjected to symmetric biphasic cathodal-first rectangular current pulses using a Plexstim stimulator (Plexon Inc., Dallas TX). The stimulator was configured such that the electrodes were discharged by shorting to a large-area Pt return electrode during the interpulse period to prevent any potential charge accumulation due to incomplete discharge. The voltage transient response, averaged over 16 electrodes on one array, to rectangular current pulses with 200 μs pulsewidth, 100 μs interphase period, and 50 Hz frequency, is shown in Fig. 2c. The maximum charge injection capacity was defined as the charge which polarized the SIROF to a potential of −0.6 V with respect to the Pt return electrode. Since Pt adopts an equilibrium potential that is usually 0.2 V–0.3 V positive of Ag|AgCl at pH 7.2, our criterion for the maximum electrode polarization is 0.2 V–0.3 V more positive than the limit that corresponds to the water reduction potential on SIROF (−0.6 V vs. Ag|AgCl limit). The average charge injection capacity of the electrodes polarized to −0.6 V vs. Pt was 3.7 ± 0.1 mC/cm2(mean ± std. dev. n = 16).

As an important consideration for the implantation of the a-SiC UMEA is the need to avoid buckling of the shanks during insertion into the nerve, we were interested in insertion mechanics, particularly for acute studies. To this end, we evaluated the buckling force threshold of the shanks on the array using nonfunctional devices compromising a-SiC alone (no metalization layer), with thickness levels of 6 μm and 8 μm, and the PI release layer (1 μm). Buckling forces were measured by advancing an array onto a silicone-coated (~5 μm thick) silicon wafer at an anticipated surgical implantation speed of 10 μm/s and the force was measured as shown in Fig. 2d for a representative array. As each shank contacts the substrate there is an incremental increase in the force (0.2 mN–0.5 mN) that remains substantially constant until the next shank contacts. The average critical buckling load of the longest shank on an array with a-SiC shank dimensions of 536 μm length, 23 μm width, and 6 μm a-SiC thickness was 1.39 ± 0.05 mN (mean ± std. dev, n = 3). As demonstrated in Fig. 2e, five shanks which have a 20 μm incremental length difference from each other on the array, are buckled consecutively as the array proceeds towards the substrate (Supplementary-Video S3). Reflection of the shanks on the substrate (silicon wafer) can also be observed in Fig. 2e. As can be seen from the graph (Fig. 2d), each shank contributes about the same buckling force. The incremental increase in the buckling forces is due to the 20 μm reduction in length of each shank from the previous one. The constant-force plateau between successive contacts of the shanks is expected since once buckling of a slender column is initiated, very little additional force is necessary to sustain the deflection of the buckled shank as it is pressed onto the substrate. Of note is the lack of reduction in the force after each buckling event which indicates that the shanks do not slip on the substrate as they advance. The force pattern is reversed as the shanks are withdrawn from the substrate.

An estimate of the critical buckling load for shanks of different thickness a-SiC layer (4 μm, 6 μm, 8 μm, 10 μm, and 12 μm) was made using a 3D FEM of a single shank (536 μm long). The modeling result for an 8 μm-thick shank (3.32 mN buckling force) was validated by comparing it with the experimental buckling force value (3.34 mN) (Fig. 2g). Fig. 2e shows the shank geometry meshed with tetrahedral elements and the modeling results presenting the normalized displacement of the shank (8 μm a-SiC, 1 μm PI). To assess the stiffness corresponding to the shanks with different a-SiC thicknesses another FEM with different boundary conditions was developed. Fig. 2g presents the modeling results of stiffness and buckling force values of a single shank (23 μm wide, 536 μm long) for different thicknesses of the a-SiC layer.

The capability to selectively record neural activity from different sites within an autonomic nerve was assessed by implanting the device in the left cVN in anesthetized rats. After isolating the nerve and placing the nerve into the Neurocase (Fig. 3b), we gently opened a 0.5–1 mm window in the epineurium. The 16-channel a-SiC UMEA was then inserted in a diagonal orientation into the cVN (Fig. 3c), allowing the 16 electrodes (numbering based on Fig. 3a), to contact axonal populations at different depths and to transverse the width of the nerve as shown in Fig. 3d. A cross-sectional TEM from the cVN was used to calculate the number of axons expected to be in contact with each a-SiC UMEA electrode site (Fig. 3e). The noise level of the 16 recording channels (GSA = 200 μm2) in saline was around 35 μV peak-to-peak (Supplementary, Fig. S3).

To evaluate recording selectivity of the a-SiC UMEA in the cVN in response to physiological events, we manipulated the mean arterial pressure (MAP) pharmacologically. This resulted in an increase in MAP and cVN activity. The increase in neural activity as a result of phenylephrine injection, which results in hypertension, is expected based on previous reports (Kuo et al., 2005). A simplified overview of the general organization of the afferents and efferent neural pathways associated with the vagus nerve in response to autonomic modulation of cardiac homeostasis is shown in Fig. 3f. After phenylephrine injection, the elevation of blood pressure is primarily sensed by the carotid baroreceptors which communicate with centers in the medulla via the glossopharyngeal nerve. Blood pressure changes are also transmitted to the medulla via left vagus afferent signals from the aortic baroreceptors, but with a much lower sensitivity than that from the carotid bodies. In response to an increase in blood pressure, parasympathetic efferents from the medulla are transmitted along both branches of the vagus nerve to the heart, inhibiting cardiac output (Câmara and Griessenauer, 2015; Min et al., 2019; Olshansky et al., 2008; Wehrwein and Joyner, 2013).

We also manipulated the blood oxygen level by inducing oxygen restriction. A schema of the neural circuitry associated with vagus nerve involved in pulmonary hemostasis (regulating the blood oxygen level) is included in Fig. 3f. Changes in the blood oxygen levels (PO2) are sensed by the carotid sinus chemoreceptors and transmitted to the medulla via the glossopharyngeal nerve. Changes in respiration in response to hypoxia are transmitted as vagal afferents from pulmonary stretch receptors to the medulla and apneustic center of the pons. Parasympathetic vagal efferents are also involved in modulating airway caliber, although the sympathetic pathways are highly species dependent. Hypoxia also increases the cardiac vagal efferent activity (Canning and Fischer, 2001; Feher, 2017; Feldman et al., 2013; Fukuda et al., 1989).

To manipulate the mean arterial pressure phenylephrine (10 μg/kg) was intravascularly injected, while simultaneously monitoring the nerve activity in all 16 channels. A schematic diagram of the electrodes that recorded evoked activity, and the experimental approach are shown in Fig. 4a–b. Channels 2, 5, 8, and 13 (Fig. 4a) recorded changes in neural activity from baseline, which coincided with an increase in MAP. Specifically, an increase in MAP from 90 mmHg to 170 mmHg resulted in an increase in the frequency and amplitude of compound action potentials (CAPs) which were identified by PCA (Fig. 4c–f). Following the notation employed by Zanos et al. (2018), we describe these neural signals as micro-compound action potentials (μCAPs) to describe the neural signal, which because of our electrode size, arise from a comparatively small number of fibers compared with that recorded with extraneural surface electrodes. Although we cannot specifically identify the fiber type from these studies, it seems likely that the increase in neural activity arises from parasympathetic efferents that act to inhibit cardiac output in response to the increase in blood pressure as shown in Fig. 3f. Besides channel 1, which had a high baseline noise level, the rate of neural activity did not change on the remaining channels, indicating selective detection of axon clusters responsive to BP changes. Fig. S4 (Supplementary) presents the electroneurograms (ENGs) of the four channels that recorded an increase in neural activity and four channels as representative examples of channels with no change in activity.

Fig. 4. Selective recording of clustered neural activity in the cVN evoked by an increase in BP.

Fig. 4.

Open in a new tab

a, Schematic representation of the a-SiC UMEA. Channels 2, 5, 8, 13 (inscribed by ellipses) detected the increase in MAP. b, Diagram of the experimental set up. 10 μg/kg phenylephrine was injected into the left femoral vein and blood pressure measured from the left femoral artery while recording neural activity from the left cVN. c, Electroneurogram of channel 13 representing the changes in neural activity upon phenylephrine administration (bottom) and the corresponding BP change (top). d, μCAP waveform recorded on channel 13 in 2D feature space. The features are the first two principal components of spike waveforms. e, The corresponding average waveform of the cluster in panel d. The arrow points to the depolarization peak. The dotted line denotes ±1 standard deviation. f, Histogram of the average firing rate of the μCAP detected on channel 13 (4 ms bin). g, Autocorrelogram of the μCAP on channel 13 (2 ms bin size).

We also examined recordings from individual channels in more detail. Spike sorting of channel 13, as an example, resulted in identification of 139 waveforms. The projection of the evoked waveforms in 2D feature space (the first two principal components, PC1 and PC2) and the corresponding average waveform of the detected spikes, putatively μCAPs, are presented in Fig. 4d–e. The increase in MAP upon phenylephrine injection coincided with an increase in the neural activity (at t = 660 s) as can be observed in the rate histogram in Fig. 4f. A minimum refractory period of 4 ms was determined for the unit on channel 13 based on the autocorrelogram (Fig. 4g), which evaluates the temporal relation between the detected waveforms. At the end of the procedure, the nerve activity was blocked by adding lidocaine to the cVN topically (~100 μL, 2% concentration) which substantially abolished neural activity, confirming the neural origin of the recordings.

Since cVN fibers from the carotid sinus respond to changes in systemic oxygenation (Fukuda et al., 1989), we induced oxygen restriction to further evaluate the capability of a-SiC UMEAs to differentially detect the activity of subgroups of vagal fibers. After 2 min of oxygen deprivation, an increase in the activity was observed on channel 4 (Fig. 5a), but not on channels 9 and 12. Conversely, an increase in BP induced by phenylephrine evoked an increase in the neural activity detected on channels 9 and 12 (Fig. 5a), but not on channel 4, demonstrating the selective recording of different physiological functions by tightly packed axons in the cVN (ENGs of channels 4, 9, and 12 are presented in Supplementary, Fig. S5).

Fig. 5. Selective recording of distinct functional fibers in cVN evoked by oxygen restriction and hypertension.

Fig. 5.

Open in a new tab

a, Schematic representation of the a-SiC UMEA showing channels 4, 9, and 12 on the array (inscribed by rectangles and ellipses, respectively) that detected differential evoked neural activity induced by oxygen restriction (Channel 4) and hypertension (Channels 9 and 12). The waveform recorded on channel 4 presented an increase in activity upon oxygen restriction but no change upon phenylephrine administration. However, channels 9 and 12 responded to drug administration but not to oxygen restriction. The rest of the channels either did not record a change in neural activity or the biological noise level was higher than neural signal. b, Units identified on channel 4 and c, Channel 9 in 2D feature space. d, Corresponding average waveform of unit a detected on channel 4 (The dotted line denotes ±1 standard deviation) and histogram of the average firing rate of the μCAP (bin size = 4 ms) before and after inducing oxygen restriction and applying lidocaine to the nerve. e, Average waveform and rate histogram of unit a and f, Unit b on channel 9 before and after heparin (anticoagulant) and phenylephrine administration. The arrows on the waveforms point to the depolarization peaks.

The increase in neuronal activity on channel 4 occurred approximately 2 minutes after initiating oxygen deprivation (this delay is expected based on the dimensions of the gas delivery tube and the oxygen flow rate) as shown in the rate histogram in Fig. 5d. At t = 420 s the recording was paused in order to apply lidocaine to the nerve which resulted in a noticeable decrease in the firing activity.

The statistical analysis of the μCAPs on channels 4 and 9 is also presented in Fig. 5. One μCAP was identified on channel 4 (Fig. 5b) and two on channel 9 (Fig. 5c). The projection of 166 detected waveforms on channel 4, and 95 and 120 detected waveforms on channel 9, onto 2D PC feature space, along with their corresponding waveforms are shown in Fig. 5d and 5e–f, respectively. The two μCAPs detected on channel 9 responded differently to the increase in BP. An increase in neuronal activity of unit a and a decrease in the activity of unit b can be observed (Fig. 5e–f).

To further investigate whether selective recordings from small unmyelinated fibers can be detected by a-SiC UMEAs, we evoked activity in the ventral trunk of the SDVN, lateral to the esophagus and proximal to the gastric branching, which is virtually devoid of myelinated axons (11275 unmyelinated vs. 52 myelinated) (Prechtl and Powley, 1990). Using a bipolar Pt cuff implanted on the SDVN, biphasic cathodal first symmetric rectangular current pulses with 200 μs pulse width, 100 μs interphase delay, 20 Hz frequency, and 50 μA and 75 μA amplitudes were applied for 1 s using a Plexstim stimulator. The current amplitude was 50 μA for the first four stimulations (initiated at t = 33 s, 90 s, 167 s, and 241 s) and 75 μA for the rest (initiated at t = 373 s, 440 s, 507 s). The cVN activity was recorded using an Omniplex data acquisition system.

Fig. 6a shows the sixteen channels on the array which can be categorized into three different groups based on the recorded neural activity on them; channels 1, 3, 4, 6, 7, 12, 13, 14, 16 showed an increase in activity at different time points during the 10-min recording session after multiple SDVN stimulations; channels 9 and 11 showed an increase in activity after the first 50 μA stimulation, and the activity on the other sites (channels 2, 5, 8, 10, 15) did not noticeably change (Supplementary, Fig. S6).

Fig. 6. Selective spatial recording of neural activity in the cVN, evoked by stimulating the subdiaphragmatic vagus nerve (SDVN).

Fig. 6.

Open in a new tab

a, Schematic representation of the array. Channels 1, 3, 4, 6, 7, 12, 13, 14, and 16 (black ellipses) responded to SDVN stimulation by an increase in neuronal activity. Channels 9 and 11 detected an increase in spontaneous activity after the first SDVN stimulation (green rectangles). Channels 2, 5, 8, 10, and 15 (blue hexagons) did not record changes in neuronal activity. b, An increase in neural activity was recorded on channel 11 after first 50 μA SDVN stimulation. c, Channels 3 and 16 showed increased activity at different time points after repetitive SDVN stimulation likely due to secondary events. d, Average waveforms of unit b recorded on channel 3 (The dotted line denotes ±1 standard deviation) (left), corresponding histogram of the average firing rate of the unit (4 ms bin size) (middle), corresponding autocorrelograms of the unit (2 ms bin size) (right). The arrows point to depolarization peaks.

Interestingly, the increase in neural activity on channels 1, 3, 4, 6, 7, 12, 13, 14, 16 happened after multiple SDVN stimulations and this increase occurred at different time points on each of the channels (ENGs of all 16 channels are provided in the Supplementary-Fig. S6). Channels 1, 4, 7, 13, and 14 recorded a gradual increase in activity, whereas channels 2 and 16 recorded a sudden increase in activity after the second 50 μA stimulation, channel 6 recorded the increase after the third 50 μA stimulation, and channel 3 after the first 75 μA stimulation. As might be expected, higher stimulation currents will recruit a larger number of fiber and consequently neural activity may be recorded on additional channels in the cVNs. Overall, this result, highlights the ability of the a-SiC UMEA to record temporal changes in the cVN at different locations, and revealed that highly dynamic electrophysiological events are spatially organized in the cVN.

The ENG of channel 11 is a representative of the channels that recorded an immediate increase in their spontaneous activity after SDVN stimulation (Fig. 6b). The ENG of channels 6 and 16 are representative examples of channels that showed an increase in activity with some delay after multiple SDVN stimulations (Fig. 6c). The results of spike sorting on channel 3, are shown in Fig. 6d. The increase in activity on channel 3 occurred at around t = 400 s following SDVN stimulation with a 75 μA current stimulus (1 s pulsing at 20 Hz) at t = 373 s and this increase continued till the end of the recording session (Fig. 6c). Unit b, one of the waveforms identified on channel 3, showed low levels of activity initially and a sudden increase in the activity at t = 400 s after multiple stimulations of the SDVN. Autocorrelation analysis determined a refractory period of t ≥ 2 ms for unit b (Fig. 6d).

4. Discussion

Enhancing the spatial resolution of neural recording from peripheral nerves is expected to allow higher levels of control in neuroprosthetic devices and facilitate the development of focused bioelectronic medicine as it allows targeting small subsets of axons within complex nerves such as the vagus. The existing electrode interfaces for cVN do not provide a high degree of functional selectivity due to the placement and size of the electrodes, as extraneural electrodes do not have selective access to the fibers deep inside the nerve. Recently, an intraneural electrode made of CNT yarn (McCallum et al., 2017) was successfully used to interface the rat cVN. That study demonstrated the possibility of chronically interfacing the cVN with indwelling electrodes. However, the GSA of the fiber electrodes is comparatively large (~80 times larger than the GSA of the a-SiC UMEA’s electrode), which clearly reduces the spatial selectivity of neural recordings, at least from more heterogeneous multimodal nerves such as the vagus nerve, composed of Aα, Aβ, Aγ, Aδ, B, and C fibers (Helmers et al., 2012). The thin-film processes employed in fabricating the a-SiC UMEAs allow great latitude in the design of the intraneural component of the arrays, including selection of electrode surface area, precise control of the two-dimensional spatial arrangement of electrode sites and the placement of multiple electrode sites along a single intraneural shank. Although the chronic tissue response to the a-SiC UMEAs remains to be established, the small shank dimensions achievable with the a-SiC devices results in a comparatively small displacement of the nerve tissue per electrode site and this is expected to minimize any adverse foreign body response. The reduced tissue displacement, compared for example with microwire electrodes, suggests the possibility of having a large number of recording or stimulation sites within the nerve that are well-tolerated.

In this study we demonstrated the functional selectivity in recording from the cVN using a 16-channel a-SiC-based array. Amorphous-SiC has previously been used in the fabrication of UMEAs for intracortical applications (Deku et al., 2018). The a-SiC has several desirable properties as the base and encapsulation material of the UMEA including chemical inertness (Iliescu et al., 2008), compatibility with high temperature thin-film fabrication processes (Sarro, 2000), high electronic resistivity (~3 × 1013 Ωcm) (Cogan et al., 2003) and high elastic modulus (Cros et al., 1997; El Khakani et al., 1994), enabling implantation without the need for mechanical support to suppress buckling of the UMEA shafts, at least for the cross-sectional dimensions employed in the current study (Supplementary-Video S1). Biocompatibility of a-SiC has been demonstrated in cell culture (Iliescu et al., 2008) and a-SiC coatings have been employed in chronic animal studies (Black et al., 2018; Joshi-Imre et al., 2019). Amorphous-SiC has also been evaluated as a hemocompatibile coating in coronary stents (Dinne et al., 2002).

The selectivity of the a-SiC UMEAs was evaluated by evoking physiological responses by phenylephrine administration, oxygen restriction, and electrical stimulation of the SDVN. We observed that different channels of the array at different intraneural locations, differentially recorded neural activity evoked by two distinct physiological events: The vasodepressor phenylephrine induced the expected increase in blood pressure that temporally coincided with an increase in neural depolarization detected on some channels (channels 9 and 12, Fig. 5). The activity is likely from arterial baroreceptors (Weston et al., 2003) and excitation of vagal afferents which reflexly activate the cardiac efferent fibers (Kunze, 1972). Conversely, a functionally distinct intraneural activity was evoked on channel 4 (Fig. 5) by oxygen deprivation and hypertension likely due to the activation of pulmonary stretch receptors (Canning and Fischer, 2001; Feher, 2017; Feldman et al., 2013). Based on the waveforms of PCA-identified μCAPs, phenylephrine induced vasoconstriction stimulated what are likely B fibers almost immediately and possibly C fibers after a 400 s delay (Channel 9, Fig. 5e–f), whereas oxygen deprivation increased the activity of C fibers after a 120 s delay (Channel 4, Fig. 5d).

Further, we provide evidence that stimulation of the SDVN evoked an increase in spontaneous activity that was identified selectively on channels 9 and 11 (Fig. 6b), likely representing the activation of gastrointestinal Aδ sensory afferents with a conduction velocity~8.5 m/s and C fiber nociceptors (Chen et al., 2008). Repetitive stimulation of the SDVN at 50 μA and 75 μA, also induced a delayed increase in the cVN activity which initiated at different times at specific recording sites. This likely represents the activation of different subgroups of unmyelinated afferent axonal fibers (Andrews et al., 1980). Even though adjacent recording sites on the UMEAs are only separated by ~80 μm, different ENGs were recorded on the adjacent electrodes (Supplementary, Fig. S6).

A challenge with these measurements is the small amplitude of the neural signals. We expect this is due, at least in-part, to the small number of fibers that we are recording from with our ultramicroelectrodes. Increasing the size of the electrodes will result in an increase in signal amplitude but at the expense of selectivity. There is therefore a compromise between signal amplitude and selectivity for the μCAPs. Besides increasing the electrode size, it is also possible that low impedance coatings such as poly(3,4-ethylenedioxythiophene) (PEDOT) might also increase signal amplitude.

While this work did not identify the specific nature of the neuron population being recorded, it is known that these physiological stimuli activate multiple types of neurons including Aβ-Aδ sensory afferents and C fiber nociceptors (Chen et al., 2008; Mellema, 2008). Additional experiments are needed to further resolve the functional origin of the recorded activity from these intravagus fibers evoked by SDVN stimulation.

Using an electron micrograph from the cVN covering ~44 μm2 area and given the area of a single recording site (200 μm2) on our a-SiC UMEAs, we estimated that a single electrode likely records from approximately 730 unmyelinated and 95 myelinated axons in close proximity (Fig. 3e). Given the fact that extraneural stimulation of SDVN ventral trunk activates an expected 11,327 axons (0.76 μm–2.15 μm in diameter) (Prechtl and Powley, 1990), the recorded activity by single electrodes on the UMEA represent μCAPs from bundled axons in proximity to a particular electrode contact. The recorded μCAPs in this study, provide insight into the highly dynamic and asynchronous activity evoked by physiological events or electrical stimulation, in which some bundled axons inside the nerve are activated by immediate specific stimuli, while others are evoked by indirect or secondary events. This capability enabled by the a-SiC UMEA, opens the possibility for future studies directed towards mapping intraneural activity with precise temporal and spatial correlation of the μCAPs to physiological and clinical therapeutic effects.

Currently, VNS is a therapeutic option for epilepsy and depression and is under investigation for the treatment of heart failure, sepsis, pain, obesity, lung injury, diabetes, rheumatoid arthritis, traumatic brain injury, post-stroke motor recovery, tinnitus, and peripheral nerve injury (Johnson and Wilson, 2018; Meyers et al., 2019). Although the number of potential therapeutic benefits of VNS is growing, the precise neuronal type and pathways involved are mostly unknown. In addition, a major limitation of cervical VNS is the adverse side effects commonly observed including neck pain, coughing, dyspnea, voice alteration or hoarseness, which is a result of multiple fiber-type recruitment during stimulation (Handforth et al., 1998). Selective recording from the cVN may enables us to identify the location of intraneural axonal bundles involved in different therapeutic effects associated with VNS. It also suggests that selective stimulation can be achieved at the level of small axonal bundles while avoiding fibers associated with side effects, which might significantly enhance the level of control for modulating different autonomous activities.

5. Conclusion

The selectivity of a 16-channel a-SiC UMEA with 200 μm2 intraneural electrodes in recording neural activity was assessed in the rat cVN. Distinct neural activity was recorded on different channels upon inducing hypertension and oxygen restriction. The electrodes also recorded dissimilar spatial and temporal neural activity consequent to repetitive electrical stimulation of the SDVN. The type of fiber activated during electrical stimulation of SDVN was not resolved in this study. As the vagus nerve is one of the main autonomous nerves, the level of anesthesia is expected to affect also its functionality. In this study we did not evaluate the effect of the anesthesia level on the recorded nerve activity.

To further improve selectivity smaller electrode sizes can be incorporated in the array design. The feasibility of using electrodes as small as 20 μm2 for neural stimulation and recording based on benchtop measurements in an inorganic model of interstitial fluid (model-ISF) is presented elsewhere (Ghazavi et al., 2020). Studies involving intravagal microstimulation with a-SiC UMEAs to explore the possibility of therapeutic applications would also be of great interest. Similar to cortical implants (Stice et al., 2007), we expect that a-SiC UMEAs with small shank cross-sectional areas will evoke low levels of FBR and thus have higher stability for in vivo applications. From a device perspective, based on the properties of amorphous silicon carbide encapsulation (Cogan et al., 2003) (Cogan, 2017) and SIROF electrode coatings (Maeng et al., 2019), good long-term stability of the UMEA structures is expected. The extension of these studies to chronic preparations is warranted in order to confirm this possibility and to determine the stability of the electrode-tissue interface.

Supplementary Material

mmc1
mmc2

Acknowledments

SFC acknowledges financial support from the United States, National Institutes of Health under grant R01 NS104344-02.

Footnotes

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data related to this article can be found at https://doi.org/10.1016/j.bios.2020.112608.

References

  1. Andrews P, Grundy D, Scratcherd T, 1980. J. Physiol 298, 513–524. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Ben-Menachem E, 2001. J. Clin. Neurophysiol 18, 415–418. [DOI] [PubMed] [Google Scholar]
  3. Black BJ, Kanneganti A, Joshi-Imre A, Rihani R, Chakraborty B, Abbott J, Pancrazio JJ, Cogan SF, 2018. Chronic recording and electrochemical performance of Utah microelectrode arrays implanted in rat motor cortex. J. Neurophysiol 120 (4), 2083–2090. [DOI] [PubMed] [Google Scholar]
  4. Boretius T, Badia J, Pascual-Font A, Schuettler M, Navarro X, Yoshida K, Stieglitz T, 2010. Biosens. Bioelectron 26, 62–69. [DOI] [PubMed] [Google Scholar]
  5. Câmara R, Griessenauer CJ, 2015. Anatomy of the vagus nerve In: Tubbs RS, Rizk E, Shoja MM, Loukas M, Barbaro N, Spinner RJ (Eds.), Nerves and Nerve Injuries: Vol 2: Pain, Treatment, Injury, Disease and Future Directions. Academic Press, London, p. 385, 297. [Google Scholar]
  6. Canning BJ, Fischer A, 2001. Respir. Physiol 125, 113–127. [DOI] [PubMed] [Google Scholar]
  7. Caravaca AS, Tsaava T, Goldman L, Silverman H, Riggott G, Chavan SS, Bouton C, Tracey KJ, Desimone R, Boyden ES, Sohal HS, Olofsson PS, 2017. J. Neural. Eng 14, 0066005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Chen SL, Wu XY, Cao ZJ, Fan J, Wang M, Owyang C, Li Y, 2008. Am. J. Physiol. Gastrointest. Liver Physiol 294, 1441–1449. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cogan SF, 2017. Conductive and insulative materials In: Horch K, Kipke D (Eds.), Neuroprosthetics: Theory and Practice. World Scientific, Singapore, pp. 240–258. [Google Scholar]
  10. Cogan SF, Edell DJ, Guzelian AA, Ping Liu Y, Edell R, 2003. J. Biomed. Mater. Res 67A, 856. [DOI] [PubMed] [Google Scholar]
  11. Cogan SF, Ehrlich J, Plante TD, 2014. 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE, Chicago, pp. 6850–6853. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Cogan SF, Ehrlich J, Plante TD, Smirnov A, Shire DB, Gingerich M, Rizzo JF, 2008. J. Biomed. Mater. Res. B Appl. Biomater 89B, 353–361. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Cogan SF, Troyk PR, Ehrlich J, Gasbarro CM, Plante TD, 2007. J. Neural. Eng 4, 79–86. [DOI] [PubMed] [Google Scholar]
  14. Cros B, Gat E, Saurel JM, 1997. Characterization of the elastic properties of amorphous silicon carbide thin films by acoustic microscopy. J. Non-Cryst. Solids 209 (3), 273–282. [Google Scholar]
  15. Davis TS, Wark HAC, Hutchinson DT, Warren DJ, O’Neill K, Scheinblum T, Clark GA, Normann RA, Greger B, 2016. J. Neural. Eng 13, 36001. [DOI] [PubMed] [Google Scholar]
  16. De Ferrari GM, Stolen C, Tuinenburg AE, Wright DJ, Brugada J, Butter C, Klein H, Neuzil P, Botman C, Castel MA, D’Onofrio A, 2017. Int. J. Cardiol 244, 229–234. [DOI] [PubMed] [Google Scholar]
  17. Deku F, Mohammed S, Joshi-Imre A, Maeng J, Danda V, Gardner TJ, Cogan SF, 2019. J. Biomed. Mater. Res. B Appl. Biomater 107, 1654–1661. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Dinne I, Kalnins U, Kumsars I, Erglis A, Jegere S, 2002. Clinical outcomes of silicon carbide coated stents in patients with coronary artery disease. Med. Sci. Monit 8 (2), PI16–PI20. [PubMed] [Google Scholar]
  19. El Khakani MA, Chaker M, Jean A, Boily S, Kieffer JC, O’Hern ME, Ravet MF, Rousseaux F, 1994. J. Mater. Res 9, 96–103. [Google Scholar]
  20. Falcone JD, Liu T, Goldman L, Pogue DD, Straka M, Reith LW, Bouton CE, Sohal HS, 2018. bioRxiv 402925. [Google Scholar]
  21. Feher J, 2017. Control of ventilation In: Feher J (Ed.), Quantitative Human Physiology: an Introduction. Academic press, London, pp. 672–681. [Google Scholar]
  22. Feldman JL, McCrimmon DR, Morrison SF, 2013. Neural control of respiratory and cardiovascular functions In: Squire LR, Berg D, Bloom F, du Lac S, Ghosh A, Spitzer NC (Eds.), Fundamental Neuroscience. Academic Press, pp. 749–766. [Google Scholar]
  23. Fukuda Y, Sato A, Suzuki A, Trzebski A, 1989. J. Auton. Nerv. Syst 28, 61–74. [DOI] [PubMed] [Google Scholar]
  24. Ghazavi A, Maeng J, Black M, Salvi S, Cogan SF, 2020. J. Neural. Eng 17, 016022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. González-González MA, Kanneganti A, Josh A, Hernandez-r G, Bendale G, Modi R, Ecker M, Khurram A, Cogan SF, Voit WE, Romero-Ortega MI, 2018. Sci. Rep 8, 1–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Handforth A, DeGiorgio CM, Schachter SC, Uthman BM, Naritoku DK, Tecoma ES, Henry TR, Collins SD, Vaughn BV, Gilmartin RC, Labar D, 1998. Neurology 51, 48–55. [DOI] [PubMed] [Google Scholar]
  27. Helmers SL, Begnaud J, Cowley A, Corwin HM, Edwards JC, Holder DL, Kostov H, Larsson PG, Levisohn PM, De Menezes MS, Stefan H, Labiner DM, 2012. Acta Neurol. Scand 126, 336–343. [DOI] [PubMed] [Google Scholar]
  28. Horn CC, Ardell JL, Fisher LE, 2019. Physiology 34, 150–162. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Iliescu C, Chen B, Poenar DP, Lee YY, 2008. Sens. Actuators, B Chem 129, 404–411. [Google Scholar]
  30. Jean A, Chaker M, Diawara Y, Leung PK, Gat E, Mercier PP, Pépin H, Gujrathi S, Ross GG, Kieffer JC, 1992. J. Appl. Phys 72, 3110–3115. [Google Scholar]
  31. Johnson RL, Wilson CG, 2018. J. Inflamm. Res 11, 203–213. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Joshi-Imre A, Black BJ, Abbott J, Kanneganti A, Rihani R, Chakraborty B, Danda VR, Maeng J, Sharma R, Rieth L, Negi S, Pancrazio JJ, Cogan SF, 2019. J. Neural Eng 16, 046006. [DOI] [PubMed] [Google Scholar]
  33. Kunze DL, 1972. J. Physiol 222, 1–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Kuo TB, Lai CJ, Huang YT, Yang CC, 2005. J. Cardiovasc. Electrophysiol 16, 864–869. [DOI] [PubMed] [Google Scholar]
  35. Lee S, Peh WYX, Wang J, Yang F, Ho JS, Thakor NV, Yen SC, Lee C, 2017. Adv. Sci 4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Li C, Sohal HS, Li F, Zanos T, Goldman L, Narayan RK, Bouton CE, 2017. 19th Int. Conf. Solid-State Sensors, Actuators Microsystems 51–54. [Google Scholar]
  37. Maeng J, Chakraborty B, Geramifard N, Kang T, Rihani RT, Joshi-Imre A, Cogan SF, 2019. J. Biomed. Mater. Res. B Appl. Biomater [DOI] [PubMed] [Google Scholar]
  38. McCallum GA, Sui X, Qiu C, Marmerstein J, Zheng Y, Eggers TE, Hu C, Dai L, Durand DM, 2017. Sci. Rep 1–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Mellema MS, 2008. J. Vet. Emerg. Crit. Care 18, 561–571. [Google Scholar]
  40. Meyers EC, Kasliwal N, Solorzano BR, Lai E, Bendale G, Berry A, Ganzer PD, Romero-Ortega M, Rennaker RL, Kilgard MP, Hays SA, 2019. Nat. Commun 10, 1–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Min S, Chang RB, Prescott SL, Beeler B, Joshi NR, Strochlic DE, Liberles SD, 2019. Cell Rep. 29, 2192–2201. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Navarro X, Lago N, Vivó M, Yoshida K, Koch KP, Poppendieck W, Micera S, 2007. IEEE 10th Int. Conf. Rehabil. Robot. ICORR’07 00 643–649. [Google Scholar]
  43. Olshansky B, Sabbah HN, Hauptman PJ, Colucci WS, 2008. Circulation 118, 863–871. [DOI] [PubMed] [Google Scholar]
  44. Pianca E, Krause Neto W, Pithon-Curi TC, Gama EF, Sabbag A, Souza R.R. de, 2015. Rev. Bras. Med. do Esporte 21, 403–406. [Google Scholar]
  45. Plachta DTT, Gierthmuehlen M, Cota O, Espinosa N, Boeser F, Herrera TC, Stieglitz T, Zentner J, 2014. J. Neural. Eng 11, 036011. [DOI] [PubMed] [Google Scholar]
  46. Prechtl JC, Powley TL, 1990. Anat. Embryol 181, 101–115. [DOI] [PubMed] [Google Scholar]
  47. Sarro PM, 2000. Sens. Actuators, A 82, 210–218. [Google Scholar]
  48. Soltanpour N, Santer RM, 1996. J. Auton. Nerv. Syst 60, 93–101. [DOI] [PubMed] [Google Scholar]
  49. Steegstra P, Busch M, Panas I, Ahlberg E, 2013. J. Phys. Chem 117, 20975–20981. [Google Scholar]
  50. Stice P, Gilletti A, Panitch A, Muthuswamy J, 2007. J. Neural. Eng 4, 42–53. [DOI] [PubMed] [Google Scholar]
  51. Usami K, Kawai K, Sonoo M, Saito N, 2013. Brain Stimul 6, 615–623. [DOI] [PubMed] [Google Scholar]
  52. Wark H, Sharma R, Mathews KS, Fernandez E, Yoo J, Christensen B, Tresco P, Rieth L, Solzbacher F, Normann RA, Tathireddy P, 2013. J. Neural. Eng 10, 045003. [DOI] [PubMed] [Google Scholar]
  53. Wehrwein EA, Joyner MJ, 2013. Regulation of blood pressure by the arterial baroreflex and autonomic nervous system In: Buijs RM, Swaab DF (Eds.), Handbook of Clinical Neurology,, 3rd Seriesvol. 117 Elsevier; BV, Amsterdam, pp. 89–102. [DOI] [PubMed] [Google Scholar]
  54. Weston M, Wang H, Stornetta RL, Sevigny CP, Guyenet PG, 2003. J. Comp. Neurol 460, 525–541. [DOI] [PubMed] [Google Scholar]
  55. Woodbury JW, Woodbury DM, 1991. Pacing Clin. Electrophysiol 14, 94–107. [DOI] [PubMed] [Google Scholar]
  56. Xue N, Martinez ID, Sun J, Cheng Y, Liu C, 2018. Biosens. Bioelectron 112, 114–119. [DOI] [PubMed] [Google Scholar]
  57. Zanos TP, 2019. Cold Spring Harb. Perspect. Med 9, a034157. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Zanos TP, Silverman HA, Levy T, Tsaava T, Battinelli E, Lorraine PW, Ashe JM, Chavan SS, Tracey KJ, Bouton CE, 2018. Proc. Natl. Acad. Sci. U.S.A 115, E4843–E4852. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Zhang Y, Zheng N, Cao Y, Wang F, Wang P, Ma Y, Lu B, Hou G, Fang Z, Liang Z, Yue M, Li Y, Chen Y, Fu J, Wu J, Xie T, Feng X, 2019. Sci. Adv 5. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

mmc1
NIHMS1635572-supplement-mmc1.docx (4.5MB, docx)
mmc2
NIHMS1635572-supplement-mmc2.docx (38.8KB, docx)

RESOURCES