Extracellular single-unit recordings from peripheral nerve axons in vitro by a novel multichannel microelectrode array

Abstract

The peripheral nervous system (PNS) is an attractive target for modulation of afferent input (e.g., nociceptive input signaling tissue damage) to the central nervous system. To advance mechanistic understanding of PNS neural encoding and modulation requires single-unit recordings from individual peripheral neurons or axons. This is challenged by multiple connective tissue layers surrounding peripheral nerve fibers that prevent electrical recordings by existing electrodes or electrode arrays. In this study, we developed a novel microelectrode array (MEA) via silicon-based microfabrication that consists of 5 parallel hydrophilic gold electrodes surrounded by silanized hydrophobic surfaces. This novel hydrophilic/hydrophobic surface pattern guides the peripheral nerve filaments to self-align towards the hydrophilic electrodes, which dramatically reduces the technical challenges in conducting single-unit recordings. We validated our MEA by recording simultaneous single-unit action potentials from individual axons in mouse sciatic nerves, including both myelinated A-fibers and unmyelinated C-fibers. We confirmed that our recordings were single units from individual axons by increasing nerve trunk electrical stimulus intensity, which did not alter the spike shape or amplitude. By reducing the technical challenges, our novel MEA will likely allow peripheral single-unit recordings to be adopted by a larger research community and thus expedite our mechanistic understanding of peripheral neural encoding and modulation.

1. Introduction

To interrogate the nervous system functions and mechanisms commonly requires high fidelity recordings of action potentials from individual neurons or axons, i.e., single-unit recordings. This is routinely conducted at the neural somata using a sharp liquid or metal electrode that penetrates the neural membrane to record intracellular membrane potentials (e.g., [1, 2]). Intracellular single-unit recording at the neural somata can also be conducted by a patch-clamp liquid electrode that forms a whole-cell giga-ohm seal with the neural membrane [3]. Intracellular single-unit recordings from neural axons are challenged by the much smaller axon geometry compared to that of the somata, and are conducted only in non-mammalian axons, e.g., the squid giant axons, whose axons are large enough to allow electrodes to penetrate into the axoplasm [4]. In contrast to intracellular recordings, to record action potentials extracellularly relies on the sensitive detection of ionic transmembrane currents that are usually estimated in the order of nano-Amperes in neural somata [5] and at individual Nodes of Ranvier in myelinated axons [6]. Since those tiny transmembrane currents can be easily dispersed into the surrounding electrolyte bath, extracellular single-unit recordings of action potentials require close proximity of the electrode to the neuronal tissue [7, 8].

Extracellular single-unit recordings are widely implemented in studying the neural circuitry in the central nervous system (CNS) by penetrating multi-shank electrode arrays [9], and recently by flexible neural probes or mesh-like electrodes (see [10] for a recent review). While protected by bony structures of skull and vertebrae, CNS tissues inside the dura and pia mater lack additional protective layers of connective tissue and allow electrodes to reach close proximity to neural somata and axons; the amplitude of recorded action potential spikes in the vicinity of electrode (<50 μm) is generally large enough (> 60 μV) for resolving action potentials from individual neurons via spike sorting algorithms [11]. In contrast to the CNS, neural axons in mammalian peripheral nervous system (PNS) are protected by multiple layers of connective tissues as illustrated in Fig. 1A, i.e., epineurium, perineurium and endoneurium, which are not only physical barriers that prevent the electrodes from reaching the axonal membrane but also insulating layers that severely attenuate the amplitude of extracellular recordings [12]. The impact of electrode location inside or outside connective tissues on the quality of extracellular recordings is comprehensively summarized by a recent report [13]. Consequently, most existing electrodes and electrode arrays are incapable of recording single-unit action potentials from mammalian PNS axons. For example, cuff electrodes non-invasively wrap around the nerve trunk outside the epineurium, providing long-term biocompatibility at the expense of low-resolution recordings of compound action potentials (CAPs) from a population of axons [14, 15]. Penetrating electrode arrays of various geometry were developed to be placed inside the epineurium, including LIFE (Longitudinal intra-fascicular electrode) [16, 17], TIME (Transverse intrafascicular multichannel electrode) [18, 19], USEA (Utah slant electrode array) [20], and flexible needle-structure electrodes [21–23]. Closer to nerve fascicles than the cuff electrodes, penetrating electrodes provide an enhanced signal-to-noise ratio and the opportunity to monitor single-unit action potentials [17, 20, 22]. Recently, a flexible microchannel electrode array was developed for single-unit recordings from teased spinal nerve or dorsal root filaments of about 100 μm thick [24, 25]. However, most recordings appear to be from large-diameter myelinated axons (A-fibers) (e.g., in [25]), while single-unit recordings from small-diameter unmyelinated axons (C-fibers) are rarely reported by those electrodes or electrode arrays.

Fig. 1.

(A) Schematic of a peripheral nerve and the three layers of connective tissues around and within the nerve: epineurium, perineurium and endoneurium. (B) Schematic of the novel microelectrode array (MEA) for in vitro simultaneous, single-unit recordings from multiple split nerve filaments.

For sensory afferent axons, unmyelinated C-fiber afferents out-number myelinated A-fiber afferents in almost all peripheral nerves [26, 27]. Compared with A-fibers, C-fibers usually have much slower conduction velocity, lower maximum firing rate, and broader width of action potentials [28]. These features collectively make C-fiber afferents less important than A-fiber afferents in encoding physiological stimuli that require high-fidelity encoding (large range of firing frequency) and rapid transmission (fast conduction velocity). However, in pathophysiological conditions, C-fibers play critical roles and are often responsible for the persistence of a diseased state [29]. For example, C-fiber nociceptors (afferents that encode tissue-injurious stimuli and commonly, initiate the sensation of pain) can sensitize, i.e., increasing their firing rate and decreasing response threshold [30], which likely drives the persistence of many chronic pain conditions [31, 32]. Targeting C-fiber afferents for treating diseases like chronic pain requires mechanistic understanding of afferent neural encoding functions, which can be revealed by single-unit recordings from individual C-fiber axons. However, the existing penetrating electrode arrays are incapable of recording single-unit action potentials from unmyelinated C-fibers inside the endoneurium, especially those C-fibers located centrally in the Remak bundle. Currently, single-unit recordings from mammalian peripheral C-fibers are typically conducted in vitro from manually split nerve filaments of 10 – 50 microns using a metal wire recording electrode, which is technically demanding. Recently, we expanded conventional single-wire electrode recordings into a five-channel multi-wire electrode array to enhance the single-unit recording efficiency [12] and successfully applied the multi-wire array to study the effect of ultrasonic neuromodulation on mouse sciatic nerves [33].

In this study, we aim to further reduce the technical challenge and enhance the efficiency of conducting single-unit recordings from peripheral nerve axons by developing a novel microelectrode array via microfabrication as illustrated in Fig. 1B. To eliminate the requirement of manually wrapping microns-thick nerve filaments onto wire electrodes as done previously [12], we developed a surface microelectrode array consisting of parallel channels with hydrophilic/hydrophobic patterns. Five recording electrodes are parallel hydrophilic “islands” surrounded by hydrophobic regions, which attract the split nerve filaments to attach onto hydrophilic electrodes in a hydrophobic mineral oil environment; nerve filaments are hydrophilic. We implemented our microfabricated electrode array in single-fiber recordings from mouse sciatic nerves in vitro, and achieved simultaneous single-unit recordings from multiple sciatic nerve axons of both myelinated A-fibers and unmyelinated C-fibers.

2. Materials and methods

2.1. Fabrication of the microelectrode array (MEA)

The MEA was fabricated on a standard 4-inch silicon wafer (UniversityWafer Inc., South Boston, MA) with procedures summarized in Fig. 2. All of the microfabrication procedures were performed in a clean room at Harvard University Center for Nanoscale Systems (Harvard CNS, Cambridge, MA).

Fig. 2.

Schematic of the silicon-based microfabrication processes for the MEA. SiO2: silicon dioxide; Ti: titanium; Au: gold; FDTS: 1H, 1H, 2H, 2H-Perfluorodecyltrichlorosilane. Refer to the texts for more details.

Briefly, the silicon wafer (Fig. 2a) was first deposited with a 2 μm-thick layer of silicon dioxide (SiO₂) by low-pressure chemical vapor deposition (LPCVD) in tetraethoxysilane (Fig. 2b). We then performed photolithography to deposit a pattern of titanium (Ti) and gold (Au) onto the SiO₂ layer, consisting of five parallel electrodes, connecting interconnects, and bond pads. In photolithography, SiO₂ surface was first spun with a bilayer of photoresists (LOR3A/S1805, MicroChem Corp., Westborough, MA), patterned with a maskless aligner (MLA150, Heidelberg Instruments, Woburn, MA), and developed in MF CD-26 Developer (MicroChem Corp., Westborough, MA) (Fig. 2c). Then, we sequentially deposited Ti (10 nm thick) and Au (200 nm thick) on the patterned photoresist layers via e-beam evaporation (Denton Vacuum LLC., Moorestown, NJ) (Fig. 2d). We then conducted lift-off by soaking the wafer overnight in Remover PG (MicroChem Corp., Westborough, MA) (Fig. 2e). To electrically insulate the Ti/Au connecting wires, we deposited a layer of silicon dioxide (500 nm) using plasma-enhanced chemical vapor deposition (PECVD, SPTS Technologies, Newport, UK) (Fig. 2f). Subsequently, we conducted a second photolithography to expose the electrodes and bond pads. Similarly, the photoresist (LOR3A/S1813) pattern was created by a maskless aligner (MLA150) and developed in MF CD-26 Developer (MicroChem Corp., Westborough, MA) (Fig. 2g). The SiO₂ covering the electrodes and pads was removed by reactive ion etching (RIE) in oxygen plasma (SPTS Technologies, Newport, UK) (Fig. 2h). Residual photoresist was stripped with solvent Remover PG (MicroChem Corp., Westborough, MA). To facilitate the attachment of split nerve filaments to the hydrophilic electrode surface (Ti/Au), we modified the areas other than the electrode surface to be hydrophobic by conducting a third photolithography (Fig. 2i). Photoresist patterns were created by a maskless aligner to cover the area of electrodes and bond pads, and the entire wafer was deposited with a layer of hydrophobic silane, 1H,1H,2H,2H-Perfluorodecyltrichlorosilane (#L16584, FDTS, 96%, Alfa Aesar, Haverhill, MA) (Fig. 2j), followed by a lift-off (Fig. 2k), as detailed in Section 2.3 below. The hydrophobic surface surrounding the electrodes restricts the distribution of conductive electrolytes, which not only eliminates cross recordings between electrodes but also enhances the signal-to-noise ratio by increasing the shunting impedance of the electrode to the bath solution [34].

2.2. Metrology of microfabrication

The quality of microfabrication was controlled by quantifying key features in each step using ellipsometer, profilometer, and an optical microscope. The LSE-WS Scanning Ellipsometer (ES-2), which measured the surface refractive index (n) and absorption coefficient (K) (Gaertner Scientific Corp., Skokie, IL), was implemented to quantify the uniformity of thickness of oxide and nitride layers over large areas following deposition of photoresist or other thin film layers. In addition, visual observation was performed under an optical microscope (100×) to check the outcome after each lift-off and etching process. Etch depth following RIE was measured by a profilometer (PL-8 DektakXT, Bruker, Billerica, MA), which was a stylus-based profilometer for step-height measurement. We moved the stylus probe down to our device surface and scanned forward for a programmed distance to determine the film thickness and etch depth.

2.3. Hydrophobic surface by vapor-phase silanization

The surfaces of the MEA, except for the electrodes and bond pads, were covered by hydrophilic SiO₂, and were modified to be hydrophobic by salinizing the SiO2 with fluorocarbon chains. The silane reagent, 1H, 1H, 2H, 2H-Perfluorodecyltrichlorosilane (CF3(CF2)7(CH2)2SiCl3, FDTS) was chosen based upon a reported study, in which multiple silane reagents were compared; the FDTS-treated surface yielded the largest contact angle, indicative of the strongest hydrophobicity [35]. The FDTS is also biocompatible and allows cell growth on the silanized surface for up to 14 days according to a cell culture study [36]. It was also reported that vapor-phase silanization treatment yielded superior coating performance to water-phase [37]. Hence, we deposited FDTS onto the SiO₂ surface of our MEA in a custom-built vacuum chamber filled with vaporized FDTS for 20 min. Vaporization of FDTS was achieved by applying a negative pressure of 10⁻³ Torr.

2.4. Contact angle measurement

To quantify the hydrophobicity of the silanized surface, we measured the contact angle using a light microscope connected with a live camera. A drop of water of approximately 5 microliters was placed on the surface, the outline of the droplet was captured by the camera, and the image was analyzed post-hoc to measure the contact angle (ImageJ, National Institutes of Health).

2.5. Harvest of mouse sciatic nerves

All experimental procedures were approved by the University of Connecticut Institutional Animal Care and Use Committee. C57BL/6 mice of both sexes (6–8 weeks of age, 20–30 g body weight, Taconic, Germantown, NJ) were anesthetized by isoflurane inhalation, followed by transcardiac perfusion from left ventricle to right atrium with oxygenated (95% O2, 5% CO2) Krebs solution (in mM: 117.9 NaCl, 4.7 KCl, 25 NaHCO3, 1.3 NaH2PO4, 1.2 MgSO4, 2.5 CaCl2, 11.1 D-glucose). The carcass was transferred to a tissue dissection chamber filled with cold (~8°C) oxygenated Krebs solution for dissection of bilateral sciatic nerves from proximal branches (L3, L4, and L5 spinal nerves) to distal branches (sural, tibial and common peroneal nerves). One dissected sciatic nerve with proximal and distal branches (~3 cm in total length) was transferred to a custom-built recording chamber consisting of two compartments, one perfused with oxygenated Krebs solution at 32°C and the other covered with mineral oil (Fisher Scientific, Hampton, NH). The proximal branches of the sciatic nerve were pinned in the Krebs compartment while the distal branches were pulled into the adjacent mineral oil compartment for extracellular single-unit recordings.

2.6. Single-unit recordings from mouse sciatic nerve axons

To validate our microfabricated MEA, we conducted single-unit recording from distal sciatic nerve branches in vitro in the two-compartment recording chamber described above. Action potentials were evoked by electrical stimulation of the proximal end of either L3 or L4 spinal nerves (5 – 15 volts, 0.5 Hz, monopolar anodic) in the Krebs compartment using an extracellular suction electrode (WPI, Sarasota, FL) and a stimulus isolator (A365, WPI, Sarasota, FL). Single-unit recordings were obtained from the distal branches in the adjacent mineral oil compartment. The epineurium of a distal branch was carefully removed and the nerve trunk was split into fine filaments of 10 – 50 μm thick, a process that usually removed the perineurium and severely disrupted the endoneurium [12]. When brought in proximity to the hydrophilic electrodes of the MEA in the hydrophobic mineral oil compartment, the split hydrophilic nerve filaments self-attached to the hydrophilic “islands” of the five parallel electrode lines surrounded by hydrophobic regions. To conduct simultaneous multichannel single-unit recordings, we connected the MEA via bond pads to a TDT system (Tucker-Davis Technologies Inc., Alachua, FL), which consisted of an RZ5D BioAmp processor, a PZ5 preamplifier, a high-impedance ZC32 head stage, and a ZCA-DIP16 adaptor. The single-unit recording signals were amplified at 104 dB, filtered between 300 and 3000 Hz, and digitized at 25 kHz. To reduce electrical noise, we enclosed the entire recording setup in a custom-built Faraday cage, and used DC battery packs to power the PZ5 preamplifier and A365 stimulus isolator.

2.7. Histological assessment of split nerve filaments

We qualitatively estimated the axon numbers in each split nerve filaments by imaging the cross-section using an electron microscopy protocol we reported previously [12, 33]. Briefly, the split nerve filaments were submerged in a mixed fixative solution containing 0.12M phosphate buffer solution (PB, pH 7.2), 2.5% glutaraldehyde, 2% paraformaldehyde, and 3 mM MgCl2 at 4°C for 60 min. The tissue was then rinsed twice in PB and fixed with 1% Osmium tetroxide in 0.12 M PB for 2 h in a dark environment at room temperature. The tissue was then dehydrated by ascending series of 30, 50, 70, 95 and 100% ethyl ethanol (10 min each), each was followed by two exposures to 100% propylene oxide for 10 min. After embedding in epoxy resin at 60°C for 48 h, the tissue was sectioned transversely on an ultramicrotome to reveal the cross-sections of the nerve filaments (Leica, Bannockburn, IL). The tissue sections were collected on grids and stained in 2% uranyl acetate and 2.5% Sato’s lead citrate. The cross-sectional images of split nerve filaments were captured by a transmission electron microscope (FEI Tecnai T12, Thermo Fisher Scientific, Waltham, MA) coupled with an AMT 2 K XR40 CCD camera (4 megapixel) at an accelerating voltage of 80 kV.

2.8. Data analysis

The single-unit recordings were processed post-hoc in MATLAB v2018 (Mathworks Inc., Natick, MA). The single-unit action potential spikes were detected by setting a negative threshold of 8 (for A-fibers) or 1.5 (for C-fibers) times the root mean square of the background noise. Conduction delay of each axon was extracted by measuring the time delay between the stimulus artifact and the onset of action potential spike, which was used to calculate the afferent conduction velocity. Spike waveforms also underwent principal component analysis (PCA) and were classified based upon the first three dominant components. Data are presented as means ± SE. One-way ANOVA, two-way ANOVA or Student’s t-tests were performed as appropriate using SigmaPlot v9.0 (Systat Software, San Jose, CA). Differences were considered significant when p < 0.05.

3. Results and discussion

3.1. Successful microfabrication of the MEA

As shown in Fig. 3A, the microfabricated MEA consists of five parallel Ti/Au recording electrodes, each 50 μm wide and 4000 μm long, and with a 300 μm center-to-center distance between electrodes. The Ti/Au electrodes, interconnects and bond pads were protected by a layer of SiO₂ (500 nm). The recording sites of the electrodes (50×50 μm) and bond pads (1×0.5 mm) were exposed by removing the SiO₂ with RIE as indicated by light microscopy in Fig. 3B. The thickness of the deposited Ti/Au layer measured by the profilometer showed consistent thickness of ~210 nm across all five electrodes as indicated in Fig. 3C. An automatic dicing saw (Disco DAD321) was utilized to define the shape of the device. One 4 inch wafer was cut into 40 devices (7×19 mm). The MEA was connected to an 18-pin DIP socket at the bond pads via insulated Nichrome wires (0.0026”, A-M systems, Inc.), which were pasted to the bond pads by Silver Conductive Epoxy Adhesive (MG Chemicals, British Columbia, CA). We then protected the bonding junction at the bond pads with a layer of silicone (Sylgard 184, Dow Corp., Garrison, NY).

Fig. 3.

Validation of MEA fabrication by light microscopy and profilometry. (A) The pattern of deposited gold electrodes and bonding pads sandwiched and insulated by two layers of silicon dioxide. (B) Opening of the electrode/electrolyte interface (50×50 μm) by removing the silicon dioxide layer on the gold electrode by reactive ion etching. (C) Thickness of the metal deposition measured by stylus surface profilometry. The inset shows the stylus tip.

3.2. Hydrophobic/hydrophilic surface patterning of the MEA

The five electrodes are hydrophilic. To modify the surrounding SiO2 surfaces to be hydrophobic, we conducted vapor-phased silanization as illustrated in Fig. 4A. Tricholoro-silane groups (Si-Cl) of FDTS are converted into silanol groups (Si-OH) which covalently attach to the hydroxyl groups of the SiO₂ surface via Si-O-Si bonds to form a self-assembled monolayer (SAM). Unreacted terminal chlorine groups are replaced with hydroxyl groups which produce a byproduct of HCl. These hydroxyl groups then condense and cross-link with silanols on other precursor molecules to generate a siloxane network. After silanization, the heavy fluorinated tail of FDTS increases the surface hydrophobicity as quantified by increase in contact angle. As shown in Fig. 4B, the contact angle of the MEA surface interfaced with a water droplet was significantly increased following silanization treatment (52° ± 3.1° vs 92.6° ± 2.9°, p < 0.001). Further, to assess the stability of silanized surface, we submerged the silanized surfaces in mineral oil overnight, removed the mineral oil with ethanol and measured the contact angle again; there was no significant change from before the mineral oil treatment (92° ± 3.1° vs. 92.6° ± 2.9°, p > 0.9). This strongly indicates that silanization with vaporized FDTS results in stable hydrophobic surfaces of the MEA suitable for prolonged in vitro single-unit recordings in the mineral oil compartment of our recording chamber.

Fig. 4.

Surface modification by silanization to convert hydrophilic silicon dioxide surface to be hydrophobic. (A) Silanization reactions for generating a hydrophobic FDTS monolayer on the silicon dioxide surface via covalent bonding. (B) Surface hydrophobicity quantified by contact angle measurement of the electrode array interfaced with a water droplet before and after silanization.

3.3. Electrical impedance of the MEA

The electrical impedance of the MEA affects the quality of extracellular recordings. Generally, reduced contact area of the electrode/electrolyte interface will restrict the ‘listening’ area to facilitate recordings from individual neurons or axons, i.e., single-unit recordings [8]. However, this is at the expense of increased electrode impedance that leads to increased thermal noise in the recordings attenuating the signal-to-noise ratio [38]. Based on our extensive experience with metal wire recordings, relatively large electrode size permits single-unit recordings from peripheral nerve axons, likely due to the significantly lower density of axons/neurons in the PNS as compared to in the CNS [39, 40]. Hence, we designed our electrode/electrolyte interface to be 50 by 50 μm, a surface area comparable to that of conventional wire electrodes. We determined the impedance of our MEA in phosphate-buffered saline using a potentiostat instrument (Metrohm Autolab, Utrecht, Netherlands). Each electrode of the MEA was tested individually along with a large platinum counter electrode (PINE Research, NC, USA) and a silver/silver chloride reference electrode (Cole-Parmer, IL, USA). The electrode impedances were measured from 0.1 Hz to 100 kHz to generate an impedance spectroscopy with a representative Bode plot shown in Fig. 5. The average impedance of our MEA was 23.6 kΩ ± 8.8 kΩ (N = 11) at 1 kHz, which falls in the anticipated range for metal electrodes with an exposed tip dimension of 2500 μm² in area.

Fig. 5.

Electrical impedance spectroscopy of the MEAs. The top and bottom panels are impedance magnitude and phase angle, respectively. The inset shows the magnitude of impedance around 1 kHz.

3.4. Simultaneous recording from multiple sciatic nerve axons in vitro

Following the procedures reported in Sections 2.5 and 2.6, we conducted single-unit recordings from mouse sciatic nerve axons using our microfabricated MEA as shown in Fig. 6A. In the Krebs compartment, action potentials were evoked by a suction electrode at 0.5 Hz (5 – 15 volts, monopolar anodic 0.2 ms duration). In the mineral oil compartment, five split filaments were attached to the 5 hydrophilic electrodes in the MEA. The reference electrode was placed perpendicular to the 5 recording electrodes to permit bipolar recordings of action potentials. The ground electrode was placed in the bath solution in the Krebs compartment.

Fig. 6.

Validation of the MEA by in vitro single-unit recordings from mouse sciatic nerve axons. (A) A photograph of the single-unit recording setup in the two-compartment recording chamber. The branches of L3-L5 spinal nerves were placed in the tissue compartment and the distal branches of sciatic nerve were pulled into the adjacent mineral oil compartment. Action potentials were evoked by electrically stimulating spinal nerves with a suction electrode. The magnified view shows the interface between five split nerve filaments and the five channels of electrodes in the MEA. (B) The cross-section of a split nerve filament revealed by electron microscopy. (C) Simultaneous single-unit recordings from five electrode channels. The vertical dotted line indicates an approximate conduction velocity of 1 m/s (30mm axon length), i.e., the criterion for distinguishing myelinated A-fibers from unmyelinated C-fibers. Single-unit spikes are marked by unfilled arrows. The stimulus artifact is marked by a filled arrow.

For multichannel recordings, electrical insulation between electrodes is crucial to avoid action potentials from the same axon being recorded in multiple channels [12]. As shown in Fig. 6A, the large hydrophobic surface that surrounds individual electrodes prevent conductive electrolyte solution from accumulating between electrodes, thus reducing chances for cross-electrode recordings. To achieve distinct recordings from individual axons, the sciatic nerve was split into fine filaments of 10 – 50 μm thick (Fig. 6B), which as indicated by the electron microscopy image in Fig. 6B loosened up the individual axons to facilitate closer contact to the electrodes. The small number of axons in each filament (<100) also enhances the chances of recording temporarily distinct action potential spikes, i.e., single-units. As shown in Fig. 6C, single-unit recordings using the MEA showed clearly different spike patterns, in both amplitude and conduction delay, between adjacent MEA electrodes. The MEA was capable of recording from both fast-conducting A-fibers (conduction velocity, CV, > 1m/s) as well as slow-conducting C-fibers (CV < 1m/s).

3.5. Validation of single-unit recordings with ascending stimulus intensities

To confirm that recorded spikes are from individual axons, we conducted additional experiments by gradually increasing the electrical stimulus intensity by 0.5 or 1 volts. There are two criteria to isolate single-unit action potentials: consistent shape and unchanged amplitude [41]. For a typical 5-channel single-unit recording shown in Fig. 6C, action potential spikes were marked by unfilled arrows (from potentially 10 different axons), extracted from recordings conducted at different electrical stimulus intensities, and overlaid with one another in Fig. 7A. We then conducted a principal component analysis on all spikes recorded from multiple stimulus intensities and used the first three dominant components to separate the spikes into ten clusters (N = 54, 81, 29, 301, 134, 356, 465, 203, 114, 185) as shown in Fig. 7B, which confirmed the unique identity of spike shapes from different peripheral axons. In addition, the single-unit spike amplitudes recorded at different stimulus intensities were plotted in 7C, which showed no significant difference in amplitude. The fact that individual spikes in our recordings do not change in shape or amplitude confirms that they are single-unit recordings from 10 different, individual axons.

Fig. 7.

Validation of single-unit recordings by increasing the electrical stimulus intensity and post-hoc spike analysis. (A) The waveforms of the 10 action potentials spikes indicated in Fig. 6C by unfilled arrows were extracted and overlaid after application of six different electrical stimulus intensities. Spike waveforms are unchanged by varying stimulus intensity. (B) Principal component analysis of spike waveforms to cluster the same spikes in (A) into 10 groups based upon the first three principal components in the Cartesian coordinate. (C) Amplitudes of the 10 labeled spikes in Fig. 6C evoked by varying stimulus intensities from 5 to 15 volts. (D) Spike amplitude is proportional to conduction velocity (R² = 0.49). (E) Electrical stimulus threshold appears to be inversely proportional to conduction

As shown in Fig. 7D, the peak-to-peak amplitude of single-unit action potentials is proportional to their conduction velocities. This is not unexpected given that fast-conducting A-fibers are stronger electrical current sources than slowly-conducting C-fibers, because of the significantly larger axon diameter and the presence of Nodes of Ranvier with concentrated ion channels. As shown in Fig. 7E, the stimulus thresholds that activate individual axons appear to be inversely proportional to their conduction velocities, consistent with other findings that A-fiber axons have lower stimulus activation thresholds than C-fiber axons [42, 43].

3.6. Distribution of conduction velocity from different types of axons

The conduction velocities (CV) of recorded action potentials reflect their myelination and physical diameters: axons with CV greater than 1 m/s in mice are generally myelinated A-fibers with large diameters (5 – 20 microns) whereas axons with CV less than 1 m/s are unmyelinated C-fibers with small diameters (~1 microns) [12]. To validate that our MEA is capable to record action potentials from different types of axons with different conduction velocities, we conducted recordings on 49 split nerve filaments from 18 sciatic nerves, from which 147 single-unit spikes were recorded. A typical recording shown in Fig. 8A allows the calculation of CV from conduction delays, and the CV of the 147 axons were plotted in a histogram in Fig. 8B, including 43 A-fibers (CV > 1 m/s) and 104 C-fibers (CV < 1m/s).

Fig. 8.

Ability to record from both A-type and C-type axons by the MEA. (A) Typical recordings from both myelinated A-fibers (unfilled arrows) and unmyelinated C-fibers (filled arrows). (B) The histogram of conduction velocities (m/s) from 43 A-fibers (orange) and 104 C-fibers (magenta).

4. Conclusions

In summary, we have successfully developed a multichannel microelectrode array via silicon-based microfabrication that allows convenient single-unit, simultaneous recordings from mammalian peripheral nerve axons in vitro. The MEA consists of planar titanium/gold electrodes sandwiched between insulating silicon dioxide layers. The exposed electrode/electrolyte surface is 50 by 50 microns, yielding a relatively low electrode impedance around 24kΩ at 1 kHz to lessen recording noise, and may further be improved with use of iridium oxide or PEDOT. The innovative MEA surface design of parallel hydrophilic electrode channels surrounded by hydrophobic areas facilitates the alignment of split nerve filaments with electrode channels, reducing the technical challenge of traditional single-unit recordings from peripheral mammalian axons. The hydrophobic surface pattern was generated by silanization of a silicon dioxide surface by vaporized FDTS. The novel MEA achieved simultaneous five-channel recordings from 5 different peripheral nerve filaments, a high throughput approach compared with conventional recordings with wire electrodes. In addition, both myelinated A-fibers and unmyelinated C-fibers were recorded with the MEA. Increased electrical stimulus intensity did not alter the shape or amplitude of recorded action potential spikes, which strongly indicated that those spikes were single units from different, individual axons. We anticipate that with further development of this MEA, single-unit recordings from peripheral nerve axons will be implemented by a broader research community that will work collaboratively to advance our mechanistic understanding of peripheral neural encoding, transmission and modulation for treating diseases.

Research highlights.

• We have developed a novel microelectrode array (MEA) for in vitro extracellular recordings of action potentials (i.e., single units) from individual mammalian peripheral nerve axons.

• The MEA utilizes a hydrophilic/hydrophobic surface pattern to self-align the peripheral nerve filaments towards the hydrophilic electrodes, which dramatically reduces the technical challenges in conducting single-unit recordings.

• By reducing the technical challenges, our novel MEA will likely allow peripheral single-unit recordings to be adopted by a larger research community and thus expedite our mechanistic understanding of peripheral neural encoding and modulation.

Biographies

Tiantian Guo is a Ph.D. candidate in the Department of Biomedical Engineering at the University of Connecticut. Her research work includes electrode-nerve interface, microfabrication and bio-signal processing.

Longtu Chen is a Ph.D. candidate in the Department of Biomedical Engineering at the University of Connecticut. His research work includes peripheral neuromodulation, visceral pain neurophysiology, and bio-signal processing.

Khanh Tran is a Ph.D. candidate of biomedical engineering at the University of Connecticut. Her research is focused on microsystems for drug or vaccine delivery and microfabrication.

Pejman Ghelich is a Ph.D. candidate in the Department of Biomedical Engineering at the University of Connecticut. His research is focused on the microfabrication and optimization of chronically implanted electrode arrays.

Yi-Syuan Guo is a Ph.D. candidate in the Department of Biomedical Engineering at the University of Connecticut. Her research is focused on microfluidics and microfabrication.

Nicholas Nolta received his PhD from The University of Utah. He is currently a postdoctoral associate at the University of Connecticut, mainly working on the fabrication of a 3D multi-channel electrodes and evaluating the biocompatibility of implantable electrodes via immunochemistry.

Sharareh Emadi received her PhD from Pasteur Institute, France. Afterwards, she conducted postdoctoral research at INSERM/Hospital Paul-Brousse in France and Arizona State University in the USA, respectively. She is currently an Assistant Professor-in-Residence in the Department of Biomedical Engineering at the University of Connecticut. Dr. Emadi has extensive expertise and experience in molecular, biochemical, and cell biology and tissue histology.

Martin Han received a B.S. degree in Electrical Engineering from the Univ. of Hawaii, Honolulu, HI, and a M.S. degree in Electrical Engineering and a Ph.D. degree in Biomedical Engineering from the Univ. of Southern California, Los Angeles, CA. He is an associate professor of Biomedical Engineering at Univ. of Connecticut. From 2003–2016, he was Staff Scientist/Principal Investigator in the Neural Engineering Program at the Huntington Medical Research Institutes, Pasadena, CA. His main research interests include the development of silicon-based implantable microelectrode arrays using MEMS and hybrid microfabrication technologies for better understanding of the central nervous system and for treating neurological disorders such as profound hearing loss, Parkinson’s disease, and spinal cord injury.

Bin Feng received his B.S. degree in Precision Instruments at Tsinghua University, M.S. degree in Mechanical Engineering at the University of Oklahoma and Ph.D. degree in Biomedical Engineering at Purdue University, West Lafayette. He is currently an assistant professor of Biomedical Engineering at the University of Connecticut. His research interests include mechanisms and management of pain, neuromodulation, and sensory encoding and processing of the peripheral nervous system.