Glassy Carbon Fiber-like Multielectrode Arrays for Neurochemical Sensing and Electrophysiology Recording

Abstract

Real-time measurement of neurochemical and electrophysiological activities in the living brain is vital for advancing neuroscience and understanding neurological and psychiatric disorders. Carbon fiber microelectrodes (CFEs) have been widely utilized for in vivo electrochemical detection due to their subcellular size, biocompatibility, and desirable electrochemical properties, and arrays of CFEs have demonstrated excellent multi-site chronic sensing of dopamine (DA) and neural activity recording capabilities. However, manual fabrication of CFE arrays lacks reproducibility and batch production capacity. Alternatively, photolithography enables batch fabrication of glassy carbon multielectrode arrays (GC-MEAs) with high resolution and reproducibility, but the insertion of planar flexible MEAs poses challenges. In this study, we present glassy carbon fiber-like (GCF) MEAs fabricated using photolithography. The GCF MEAs feature fiber-like GC electrodes with small cross-section, facilitating self-insertion into brain tissue without additional aids. Batch fabrication of GCF MEAs allows for customizable designs with no or minimal manual intervention. Comparative analysis with CFEs demonstrates improved electrochemical properties of GCF. In vivo experiments in mouse brains confirm the GCF MEAs’ ability to measure neurotransmitter concentrations and record neural activities. This novel fabrication approach is promising for multimodal neural activity and neurotransmitter concentration measurements with minimal footprint, offering significant advancements in neural interface technology for neuroscience research and clinical applications.

Graphical Abstract

A photolithographic approach for batch-fabrication of glassy carbon fiber-like multielectrode arrays (GCF MEAs) is presented. The subcellular-sized GCF MEAs implant deep into the mouse brain without the need for insertion aids, effectively measure neurotransmitter concentrations and record neural activities, while creating minimal damage. This method represents a significant advancement in the development of multimodal and chronically stable neural interface devices.

1. Introduction

The real-time measurement of neurochemical and electrophysiological activities in the living brain is crucial for advancing neuroscience and enhancing the understanding and treatment of neurological and psychiatric disorders [1–3]. Neurons communicate by firing action potentials and releasing neurotransmitters, and measuring both neural electrical activity and neurotransmitter levels is essential for comprehensively understanding synaptic connections [4, 5]. Therefore, the advancement of implantable multi-modal probes capable of sensing chemicals and recording electrical signals represents a pivotal step forward in elucidating the intricate relationship between electrical and electrochemical signaling in the normal and diseased brain [2, 3, 5, 6]. Carbon fiber microelectrodes (CFEs) have been the gold standard for in vivo electrochemical detection of rapid changes in neurotransmitter concentrations, leading to valuable insight into normal and pathological neurochemical dynamics [7–15]. They are widely-used implantable sensors because they are small in size (5 – 10 µm), biocompatible, flexible, and have desirable electrochemical properties.[16–18] Target neurochemicals, including dopamine (DA), [7, 19] serotonin (5-HT), [20–22] adenosine (AD), [23–25] or melatonin (MT), [26–28] can adsorb onto the CFE surface, and the concentration is calculated by the measurement of oxidation and reduction current [7, 15, 29]. Using electrochemical methods such as fast scan cyclic voltammetry (FSCV), CFEs can provide microscale spatial resolution and millisecond temporal resolution to investigate rapid, localized neurotransmitter changes in both animal models as well as humans [30]. More recently, CFEs have been also introduced as subcellular-scale probes for electrophysiological recording [1, 31–33]. To record a single neuron, carbon fibers are usually insulated with pulled glass or parylene, trimmed to expose a small area [34–36], and coated with poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS) [34]. Thus, CFEs can combine neural recordings with neurotransmitter measurements, enabling the exploration of a multimodal chemical and electrical neural interface [1, 31–33].

CFEs have also been arranged into arrays to allow measurement of multiple channels simultaneously, for both neurotransmitter sensing and neural activity recording [18, 33, 37–39]. For example, the Chestek group developed CFE arrays capable of both chronic electrophysiology and chemical sensing, and the probe can also be precisely localized to identify the recording location [40–43]. Schwerdt et al. demonstrated a carbon fiber splaying array for navigating to multiple subcortical sites to record neurochemical signaling dynamics [37]. However, the fabrication of CFE arrays requires extensive manual work, including mounting carbon fibers on the substrate, connecting carbon fiber to the circuit, providing an adequate insulation, and exposing the tips with trimming, burning or etching [44].

Conversely, photolithography can be utilized to batch fabricate carbon-based multielectrode arrays (MEAs) with high resolution, yield, and reproducibility [45–48]. Our group has previously fabricated flexible glassy carbon (GC) MEAs for the electrochemical detection of DA and 5-HT [49, 50]. The GC microelectrodes were produced by controlled patterning of SU-8 followed by pyrolysis, and then transferred into thin film polymer devices with metal traces. However, these flexible MEA probes require a metal wire shuttle to penetrate into the brain and have a larger footprint than a single CFE [49, 50]. Therefore, the MEA design still needs to be optimized to enable self-insertion, while also allowing combined measurements of neurotransmitter sensing and neurophysiological recording.

In this work, we used photolithography to develop glassy carbon fiber-like (GCF) MEAs for both neurochemical sensing and electrophysiology recording. In this novel design, we created fiber-like GC on thin film substrates; the long, relatively stiff GCF electrodes with an extremely small cross-section allow the device to penetrate the brain tissue without aid while exhibiting desirable electrochemical properties. The GCF MEAs were fabricated with GC-fiber-like backbones, which were not connected to the metal patterns but only used to stiffen the MEAs for insertion. Using photolithography, we can batch fabricate more than 50 probes of a variety of designs in one 4-inch wafer with little manual work. We compared the surface and electrochemical properties of the GCF MEAs with 7 µm CFEs and found that the GCFs has improved electrochemical properties compare to CFEs.

With the GC fiber-like structure, the GCF MEA self-penetrated into the mouse brain without requiring the use of potentially tissue-damaging shuttles or temporary stiffening agents. The GCF MEAs successfully measured tonic DA and 5-HT concentrations in vivo using specifically optimized SWV waveforms, as well as stimulation-evoked phasic DA with FSCV. They also successfully recorded neural activities in the striatum of mouse brains. This novel method enabled the fully customizable batch fabrication of GCF MEAs in a precisely controlled and highly reproducible manner. The GCF MEAs show promising potentials for multimodal measurements of neural activities and neurotransmitter concentrations with an extremely minimal footprint.

2. Methods

2.1. Glassy carbon fiber-like batch fabrication

The GCF batch-fabrication process is adapted from our previous works.[49, 50] We used 4-inch Si wafers with a 1 μm thick SiO₂ layer (University Wafer Inc., USA) to fabricate the microelectrode arrays. The wafer was first cleaned with acetone, isopropanol, and deionized (DI) water, sequentially. Then, the wafer was dried and heated on a hot plate at 150 °C for 5 min and treated with O₂ plasma using a reactive ion etcher (RIE, Trion Phantom III LT, Trion Technology, USA) for 2 min at 200 mTorr pressure and 150 W power. The cleaned wafer was spin-coated with SU-8 3050 (MicroChemicals GmbH, Germany) at 3000 rpm for 1 min and soft baked at 65 °C for 3 min and 95 °C for 10 min. The wafer was then exposed to 365 nm ultraviolet (UV) light using a direct writing mask-less aligner (MLA, MLA100 Heidelberg Instruments, Switzerland) with a dose of 800 mJ/cm². After UV exposure, the wafer was first post baked at 65 °C for 3 min and 95 °C for 5 min, and then developed with SU-8 developer (MicroChemicals GmbH, Germany) for 4 min. Following isopropanol and DI water cleaning, the patterned SU-8 was subsequently hard baked at 200 °C , 180 °C , and 150 °C for 5 min each and allowed to cool below 95 °C.

The first pattern of SU8–3050 was pyrolyzed at 900 °C in an inert N₂ environment for 60 min at 0.8 Torr. This step turns all the polymerized photoresist on the wafer into carbon, while retaining the position with a shrinkage in size. After the pyrolysis, the wafer was again cleaned with acetone, iso-propanol, and DI water sequentially, and treated with O₂ plasma with RIE for 90 s at a pressure of 200 mTorr and 150 W power. The cleaned wafer was then spin-coated with SU-8 2010 (MicroChemicals GmbH, Germany) at 2000 rpm for 1 min, and subsequently soft baked at 65 °C for 3 min and 95 °C for 5 min. The first SU-8 layer was then patterned using the MLA with a dose of 700 mJ/cm2 to define the bottom insulation layer and to open a connection between GC electrodes and the metal traces. After a post-bake at 65 °C for 3 min and 95 °C for 5 min, the wafer was developed using SU-8 developer. Finally, the patterned wafer was cleaned with isopropanol and DI water and hard baked at 200 °C, 180 °C, and 150 °C for 5 min each, and allowed to cool below 95 °C.

Next step is to pattern the metal connections. The wafer was spin-coated with AZ 4210 photoresist (MicroChemicals GmbH, Germany) at 3000 rpm for 1 min and baked at 105 °C for 3 min. After baking, the wafer was exposed using MLA with a dose of 400 mJ/cm², then developed using AZ400k 1:4 developer (MicroChemicals GmbH, Germany) and rinsed with. A 10 nm Ti adhesion layer and 100 nm Au layer were evaporated on the wafer using an electron beam evaporator (Plassys MEB550S, Angstrom Engineering, USA) and then the metal was lifted-off in acetone to define the metal traces and connection pads.

The top insulation layer of SU-8 2010 was spin-coated at 4000 rpm for 1 min, soft baked at 65 °C for 3 min and 95 °C for 5 min, and photopatterned using the MLA with a dose of 700 mJ/cm² to expose the connection pads and to define the top insulation layer. After post baking and development with the SU-8 developer, the wafer was cleaned with isopropanol and DI water, and hard baked at 200 °C, 180 °C, and 150 °C for 5 min each and allowed to cool below 95 °C.

The GCF MEAs were lifted off from the wafer using buffered oxide etchant (BOE 7:1, Transene Company, Inc., Danvers, MA, USA) in an acid hood for 4 to 6 h.

2.2. Carbon fiber microelectrode fabrication.

CFEs were fabricated as previously described in [28] [27, 51, 52]. Briefly, borosilicate capillaries (0.4 mm ID, 0.6 mm OD; A‐M systems Inc., Sequim, WA, USA), each containing a single carbon fiber (7μm diameter, T650; Cytec Carbon Fibers LLC., Piedmont, SC, USA), were pulled to a fine tip using a vertical puller (Narishige, Los Angeles, CA, USA). The tip was sealed with epoxy (Spurr Epoxy; Polysciences Inc., Warrington, PA, USA) and the exposed fiber was cut 400 μm from the glass seal using a scalpel under an optical microscope (Szx12, Olympus). A mercury drop was placed in the barrel for electrical contact to a hookup wire (Nichrome; Goodfellow, Oakdale, PA, USA).

2.3. Raman Spectroscopy

Raman spectroscopy measurements were performed using the Renishaw inVia Micro-Raman Microscope system (Renishaw, Hoffman Estates, IL, USA); an 1800 lines/mm diffraction grating was used. A 633 nm laser was focused to a spot size on the sample through a 100x objective. Laser intensity was 100% and the scan range was 100 cm⁻¹ to 3200 cm⁻¹. Before each measurement, the system was calibrated with a Si (100) Reference Sample. All data points are an average of 5 measurements.

2.4. Electrochemical characterization

EIS and CV were first performed in 1x phosphate buffered saline (PBS, composition: 11.9 mM Na2HPO4 and KH₂PO₄, 137mM NaCl, 2.7 mM KCl, pH 7.4) applying a sine wave (10 mV RMS amplitude) onto the open circuit potential while varying the frequency from 1 to 10⁵ Hz. EIS and CV were carried out using a potentiostat/galvanostat (Autolab, Metrohm, USA) connected to a three-electrode electrochemical cell with a platinum counter electrode and an Ag/AgCl reference electrode. During the CV tests, the working electrode potential was swept between 1.2 and −0.6 V (vs Ag/AgCl) with a scan rate of 100 mV/s. The charge storage capacity (CSC, mC/cm²) was calculated as CSC=(∫didt)/(geometric area) in an entire CV cycle. CV was also performed in 10 mM Ru(NH₃)₆ ^3+/2+, and 10 mM DA (100 mV/s scan rate, 0.0244 interval time, 0.00244 V steps), to investigate the redox activity of CFEs and GCFs.

2.5. In vitro fast scan cyclic voltammetry measurements

FSCV was performed with an EI 400 (Ensman Instruments; Bloomington, IN, USA), controlled by the CV Tar Heels LabVIEW program (CV Tar Heels v4.3, University of North Carolina, Chapel Hill, NC, USA), and a 4-channel WaveNeuro FSCV Potentiostat System (Pine Research, Durham, NC 27705 USA), controlled by HDCV software (UNC at Chapel Hill, North Carolina, USA). Data were analyzed using HDCV software (UNC Chapel Hill, North Carolina, USA). The electrode was scanned using a triangular waveform with a negative holding potential of −0.4 V, a 1.3 V switching potential, and applied using a 400 V/s scan rate at 10 Hz. DA and 5-HT were identified by inspection of the background-subtracted cyclic voltammograms. Electrodes were calibrated using 1–10 µM DA/5-HT concentrations dissolved in 1x phosphate buffered saline (PBS, composition: 11.9 mM Na₂HPO₄ and KH₂P O₄, 137mM NaCl, 2.7 mM KCl, pH 7.4). The different concentrations were diluted starting from freshly prepared 1 mM DA/ 5-HT solutions.

2.6. In vitro square wave voltammetry measurements

Electrochemical detection of DA and 5-HT was performed using two different SWV waveforms specifically optimized for DA [50, 51] and 5-HT [53]. SWV experiments were carried out using a potentiostat/galvanostat (Autolab PGSTAT128N, Metrohm, USA) connected to a three-electrode electrochemical cell, where the MEAs were the working electrodes, a platinum wire was the counter electrode, and an Ag/AgCl was the reference electrode. The SWV waveform for dopamine was repeatedly applied from −0.2 V to 0.3 V with a 40 Hz step frequency, a 50-mV pulse amplitude, and a 5-mV step height every 15 seconds. The potential was held at 0 V between scans [51]. The SWV waveform for 5-HT was repeatedly applied from 0.15 V to 0.5 V with a 40 Hz step frequency, a 50-mV pulse amplitude, and a 5-mV step height every 15 seconds. The potential was held at 0.15 V between scans [53]. In vitro DA calibrations were performed using freshly prepared DA solutions dissolved in 1xPBS in a 50 nM–1 µM concentration range. Electrode sensitivity was determined by the slope of the linear range of the calibration plot relating the DA peak current at 0.15 V to the DA concentration. For each electrode used in vivo, a preimplant calibration (pre-calibration) was performed using freshly prepared DA and 5-HT solutions as previously described. DA and 5-HT peaks were isolated from the non-faradaic background current for each SWV scan by subtracting a modeled polynomial baseline using a previously described methodology [51, 53].

2.7. In vivo Testing:

2.7.1. Fast scan cyclic voltammetry

All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh. A female Sprague-Dawley rat (250–350g, Taconic, Rensselaer, NY, USA) was anesthetized with isoflurane (5% for induction, 2.5% for maintenance), placed on a thermal pad, and the head was fixed into a stereotaxic frame (Kopf Instruments, Tujunga, CA, USA). The skin and connective tissue were removed, and holes were drilled in the skull over the dorsal striatum (DS, AP +1.0 mm, ML 3.8 mm from bregma) and the medial forebrain bundle (MFB, AP −4.3 mm, ML 1.2 mm from bregma). An additional hole was drilled for the insertion of an Ag/AgCl reference electrode (0.005” diameter wire, A-M Systems, coated with AgCl by amperometric deposition in 3M KCl at 4V for 3 minutes). A bipolar stimulating electrode (MS303S/1, Plastics One, Roanoke, VA) was inserted in the MFB. A standard CFE electrode, prepared as previously described, was inserted into the striatum.[15, 52] The FSCV response at the CFE was used to determine the final dorsoventral position of the stimulated electrode to produce a maximal evoked DA concentration. Then, the CFE was removed, and a GCF electrode was inserted into the striatum. The MFB was stimulated with 600 biphasic 2 ms constant current 250 µA pulses and the consequent evoked DA was detected at the GCF electrode.

2.7.2. Square Wave Voltammetry

Two male C57BL/6J mice (C57BL/6J, 22–35 g; Jackson Laboratory, Bar Harbor, ME, USA) were anesthetized with isoflurane (2% for induction, 1.25% for maintenance), placed on a thermal pad, and the head was fixed into a stereotaxic frame (Kopf Instruments, Tujunga, CA, USA). After the animal head was fixed in a stereotaxic frame (Narishige International USA, Inc. Amityville, NY, USA), the skin and connective tissue on the surface of the skull were removed. For DA SWV measurements, a small pinhole craniotomy was made over the dorsal striatum (DS, 1 mm anterior to bregma, and 1.5 mm lateral from midline) with a high-speed dental drill (0.007 drill bit, Fine Science Tools, Inc., Foster City, CA, USA), and bone fragments were carefully removed with forceps and saline. Saline was applied continuously onto the skull to dissipate heat from the high-speed drill. The GCF was lowered 3.0 mm below the cortical surface into the DS using a micromanipulator. Two additional small pinhole craniotomies were performed for the introduction of the Ag/AgCl reference electrode contralaterally to the GCF and a bone screw counter electrode caudally to the reference. EIS was measured immediately after the GCF implantation to confirm functionality. Then, the tonic DA response was measured using the DA SWV waveform over a 20 min period.

For 5-HT SWV measurements, a second mouse underwent an identical procedure, except the GCF electrode was implanted into the CA2 region of the hippocampus (AP: 2.9, ML: +3.3, DV: −2.5 mm relative to bregma).

After the GCF was lowered 2.5 mm below the cortical surface into the CA2, EIS was measured to confirm functionality. Then, the tonic 5-HT response was measured using the 5-HT SWV waveform over a 20 min period. Upon reaching the predetermined experimental endpoint, the GCFs were explanted, and the animals were humanely sacrificed using approved procedures.

2.8. Electrophysiology Recordings

A male C57BL/6J mouse (C57BL/6J, 22–35 g; Jackson Laboratory, Bar Harbor, ME, USA) was anesthetized under isoflurane (2.5%) and head-fixed in a stereotaxic frame (David Kopf Instruments, Tujunga, CA, USA). Animal body temperature was maintained at 37 °C using an isothermal pad connected to a SomnoSuite system (Kent Scientific Corporation, Torrington, CT, USA). Heart rates were monitored using the SomnoSuite system as well. A holder was used to secure the custom-designed PCBs. One skull screw was carefully positioned above the left visual cortex of the mice. A 1.4mm diameter window above the ventral striatum of the right hemisphere was opened using a motorized drill. The coordinates for the center of the window were 1 mm posterior to Bregma and 1.1 mm lateral to the midline. A drop of saline was used to hold all 5 shanks together and all the shanks were manually implanted at the same time. TDT recording system (RX5, 16-channel Medusa amplifier, Tucker Davis Technologies (TDT), Alachua, FL) was used to record electrophysiology. Neural signals were imported to MATLAB with custom scripts to remove stimulation artifacts. Plexon offline sorter (Plexon Inc Dallas, TX, USA) was used to identify single units. Raw data was filtered between 300 Hz and 10 k Hz. Threshold crossing events were identified by using a fixed negative threshold value of 3 standard deviations. A 3D PCA feature space was used to identify waveform features, and the K-means clustering method was used to identify individual units. K-means was set up using an adaptive standard EM between 2 to 5.

2.9. Immunohistochemistry

The removed brain tissue was cryoprotected using optimal cutting temperature compound (OCT, Tissue-Tek, Torrance CA), frozen, and sectioned. Tissue sections were hydrated in PBS and blocked with 10% normal goat serum. Following 45 min in 0.5% Triton X-100 in PBS, the sections were incubated overnight with primary antibodies rabbit polyclonal anti-Iba1 (Wako Chemicals, 01919741, 1:500), chicken GFAP (Sigma, AB5541, 1:500) and DAPI. Sections were then rinsed three times for five minutes each with PBS and incubated for two hours with AlexaFluor488 (goat anti rabbit, 1:500), AlexaFluor568 (goat anti mouse 1:500), and AlexaFluor633 (goat anti chicken, 1:500), from ThermoFisher Scientific (Waltham, Massachusetts, USA). Images were acquired with an Olympus FluoView 1000 confocal fluorescent microscope (Olympus, Inc., Tokyo, Japan) at the Center for Biologic Imaging at the University of Pittsburgh.

3. Results and Discussions

Figure 1A shows an illustration of the shank design and the procedure to fabricate the GCFs. As before described, SU-8 3050 was spin-coat and patterned in fiber-like structures and long backbone stiffeners, and subsequently pyrolyzed at 900 °C in a tube furnace under controlled inert atmosphere for GC synthesis [49, 50]. The long GC backbones aim to maintain the shank stiffness to assist with the penetration into the animal brain. After pyrolysis, these long GC backbones are fully insulated by SU-8 2010 patterning. Instead, the fiber-like structures (3 for each shank) are the electrode sites and are thus connected to the metal traces through a small opening on the first SU-8 insulation layer. Finally, a second SU-8 2010 insulation layer was patterned to insulate the metal traces while leaving an opening on the GC electrodes and connection pads. In this way, the GC fiber-like structures are open on both sides after releasing from the wafer. For each probe, we can fully customize the shank layout, the number of shanks, and the electrode site numbers per shank. Figures 1B and 1C shows example pictures of a design having five shanks and three GC fiber-like electrode on each site, where each shank has a length of 4.5 mm and a width of 30 µm. The displacement between two shanks is 200 µm. On one 4-inch wafer, we can batch-produce more than 50 probes with high reproducibility and yield.

Figure 1.

The design and fabrication of glass carbon fiber-like microelectrode arrays (GCF MEAs). A) The fabrication process of the GCF. SU-8 is spin-coated and patterned on a wafer, followed by pyrolysis to produce carbon electrode sites and the stiffening structure. SU-8 is spin-coated and patterned to insulate the stiffening structure but exposing the electrode sites. Metal is patterned using lift-off procedures. A top layer of SU-8 is patterned to insulate the metal patterns. B) The layout of the GCF. C) Optical microscopy images of the GCF.

To test the properties of the pyrolyzed GCFs, we performed Raman spectroscopy and electrochemical characterization and compared them to traditionally constructed CFEs.

Figure 2A presents the Raman spectra of the carbon in GCF and CFE. The Raman spectra show the two primary peaks typical for carbon materials. The D band at 1323 cm⁻¹ is the breathing mode of six-atom rings, indicating defects from the boundaries like edge-plane or doping that interrupt the graphene structure. The G band at 1604 cm⁻¹ is the E_2g phonon mode, indicating sp² graphitic carbon structures. The Raman spectra also show some secondary peaks.[16, 54] The 2D peak, which is the overtone of the D peak, is at 2642 cm⁻¹, and the D + D’ peak at 2916 cm⁻¹.[54] The defect level of the carbon materials can be evaluated by calculating the D/G peak height ratio, and the GCF and CFE have similar values, indicating they have similar defect sites which are beneficial for adsorption and electron transfer [55]. The D/G ratio value was 1.17 for the pyrolyzed GC from SU-8 3050, which is comparable with previously reported pyrolyzed GC from SU-8 100 [48, 53], pointing to the presence of a substantial graphitization degree and a high sp2 content in the GC material. Additionally, The GCFs have more distinct D and G peaks, indicating that the pyrolyzed GC has higher ordering and graphitization than CFEs.[56] The second-order Raman shift for the samples is related to the crystallinity of carbon materials. The CFE has only a broad peak, indicating the low crystallinity of the material. The increased resolution of the 2D and the D + D’ peaks for GCF confirms that GCF has higher crystallinity.[56] This is consistent with data reported in literature for GC obtained with pyrolysis temperature of 900 °C [57] and with our previous data obtained with similarly prepared GC [48, 53].Overall, this result indicates that the GC created from SU-8 pyrolysis present high level of graphitization and is rich in defect sites, beneficial for adsorption and electron transfer [48, 55].

Figure 2.

Characterizations of glassy carbon electrodes (GCF) and carbon fiber electrodes (CFE). Currents are normalized by geometric area. A) Raman spectra of GCF and CFE. B) CV of background current in PBS buffer, C) CV in 10 mM Ru(NH₃)₆ ^3+/2+ and D) in 10 mM dopamine (100 mV/s scan rate, 0.0244 interval time, 0.00244 V steps).

To understand the electrochemical properties of the GCF and CFE electrodes, we performed cyclic voltammetry (CV) in a 1x PBS buffer background and different redox couples. The CV current in the PBS buffer results from the double-layer capacitance, and the current is proportional to the surface area. Figure 2B shows GCF and CFE have nearly identical charging current densities around the electrochemical window (- 0.2 V to 1.0 V), indicating the two materials have similar surface roughness. Ru(NH₃)₆ ^3+/2+ is an outer-sphere redox couple insensitive to the surface, so the cyclic voltammograms of the two electrodes are similar (Figure 2C). DA is a surface-sensitive redox couple, which is known to adsorb onto the carbon surface. GCF electrodes are more sensitive to DA compare to CFEs, most likely due to an higher level of graphitization as revealed by the CV plots in Figure 2D [54].

While it is important to evaluate the performance of the new GCFs in relation to the gold standard CFEs, the primary focus of this work is not to directly compare the sensing properties of GCFs versus CFEs. The main objective is to introduce a method for the batch fabrication of miniaturized GC MEAs that require minimal manual intervention while preserving the electrochemical advantages of CFEs. One of the most serious drawbacks of CFE electrodes is the fabrication process; even state-of-the-art multichannel CFE arrays involve painstaking manual steps [18, 37, 43], leading to electrodes with slightly varying quality. In contrast, the GCFs present here are precisely designed using photolithography followed by pyrolysis, resulting in consistent surface and electrochemical properties across batches.

We then evaluated the sensing capabilities of the GCFs in vitro compared to CFEs. FSCV is the predominant electrochemical technique used to measure real-time neurotransmitter changes in the brain. [7, 19, 58] Figures 3A and 3B show the background subtracted FSCV plot corresponding to the detection of DA and 5-HT in vitro using GCF. The corresponding plots obtained using CFE are reported in Supplementary Figure S1. Representative color plots are reported in Supplementary Figure S2.

Figure 3.

FSCV characterization of glassy carbon fiber-like electrodes in comparison with CFEs. A) FSCV background subtracted plot corresponding to the detection of DA at different concentrations (1–10 µM). B) FSCV background subtracted plot corresponding to the detection of 5-HT at different concentrations (1–10 µM). C) Calibration curves of n=6 GCF electrodes in DA and 5-HT. D) Comparison of CFE (green) and CGF (red) normalized FSCV background subtracted plot corresponding 1µM DA detection, and E) Comparison of CFE (green) and CGF (red) normalized FSCV background subtracted plot corresponding 1µM 5-HT detection F) Calibration curves of n=6 CFE electrodes in DA and 5-HT (1–10 µM).

Both using GCFs and CFEs the background subtracted FSCVs exhibit characteristic anodic DA oxidation and cathodic dopamine-o-quinone (DAoQ) reduction peaks, confirming the detection of DA. However, we can observe a larger cathodic DAoQ reduction peak when using GCFs in Figure 3D, reporting the comparison of CFE (green) and CGF (red) normalized FSCV background subtracted plot of 1µM DA detection. The ratio of the peak oxidation current (anodic, ip,a) to the peak reduction current (cathodic, ip,c) provides information about the equilibrium of absorption/desorption properties of DA and DAoQ and the reversibility of the reaction [7, 15]. The increase in the cathodic peak of DAoQ, the oxidation product of DA, indicates a tighter adsorption of the oxidized DAoQ towards GCF than CFE surfaces, corresponding to a better reversibility of the reaction [59]. This is similar, even if less dramatic, to what was previously observed using 3D fuzzy graphene (3DFG) [60], carbon nanotube (CNT) treated CFEs [61], and CNT fiber microelectrodes [62]. A possible explanation can be that at GCF surface, due to the higher content of defect, DAoQ binding is enhanced due to the DAoQ’s affinity to partially charged carbon atoms near the edge [63]. Indeed, the electron transfer kinetics for DA oxidation has been demonstrated to be catalyzed by the adsorption of quinone containing species, including DAoQ, onto the carbon electrode surface [64].

From the FSCV background subtracted plot of 5-HT detection in Figure 3B we can observe that 5-HT shows a secondary peak in addition to the primary peak, and the secondary peak is more obvious for GCFs than CFEs (direct comparison presented in Fig. 3 E). Two separate 5-HT oxidation peaks have been previously observed for GC [48] and 3DFG [60] microelectrodes. Their capability to distinguish the peak around 0.5 V is likely assigned to the intermediate step of the 5-HT oxidation reaction [65], and can be attributable to their higher sensitivity afforded by high-density defect sites [48, 60].

Figure 3C and F shows the calibration curve of DA and 5-HT for GCFs and CFEs respectively. Electrode sensitivity was determined by the slope of the linear range of the calibration plot relating the oxidation peak currents to DA and 5-HT concentrations. The sensitivity of DA is 2.0 ± 0.2 pA µM⁻¹ µm⁻² for GCFs and 1.5 ± 0.1 pA µM⁻¹ µm⁻² for CFEs. The sensitivity of 5-HT is 4.0 ± 0.2 pA µM⁻¹ µm⁻² for GCFs, and 1.9 ± 0.3 pA µM⁻¹ µm⁻² for CFEs. GCF electrodes have significantly enhanced sensitivity (Bonferroni post-test p < 0.001, n = 6) for both DA and 5-HT. The theoretical lower detection limit (LOD), defined as 3 times the standard deviation of the noise [23, 48, 66, 67], was estimated to be 1.18 and 0.89 nM for DA and 5-HT respectively, when using FSCV. These results are comparable to what was previously observed using planar GC microelectrodes [48].

Interestingly, both for CFE and CFE electrodes, 5-HT sensitivity is higher than DA sensitivity, in this concentration range. This can be possibly attributed to 5-HT stronger adsorption properties compared to DA at GC microelectrodes, as we previously observed for GC electrodes conducting multi waveform FSCV experiments [48]. Similar observation has been reported for CFEs [68].

Due to the tendency of 5-HT to foul carbon surfaces [48, 60], we performed an electrochemical fouling test. We observed that the detection sensitivity did not significantly change before and after the continuous FSCV cycling in the presence of 5-HT over an 8-h period (Supplementary Information Figure S3), which aligns with our previous findings that GC electrodes can resist chemical fouling [69]-[70]. This resistance has been attributed to presence of dense edge planes rich in functional groups, that have also shown to increase hydrophilicity and reduce fouling [8, 66, 70].

To evaluate the effect of prolonged FSCV cycling on the electrochemical stability of these miniaturized GCF electrodes. We applied the FSCV waveform (−0.4 to 1.3 V vs. Ag/AgCl, scan rate: 400 V/s) at 10 Hz on the microelectrodes in 1x PBS for over 40 hours, and we monitored the impedance and CV at different time points (starting point and after electrode activation, 20 and 40 hours). We observed a reduction of the impedance, as shown in Supplementary Figure S4, particularly at the low-frequency range indicating an increase in surface area and capacitance. Concurrently, the charge storage capacity (CSC), calculated as the time-integral of an entire CV cycle between the water oxidation and reduction potential limits (−0.6 to 1.2 V vs Ag/AgCl, 100mV/s scan rate), increased from 41.2 mC/ cm² after activation to 70.4 mC/cm² after 40 h of FSCV cycling. This increase in capacitance is predominantly due to continued surface activation that is very common in carbon microelectrodes. The use of a 1.3 V switching potential is known to regenerate CFE surface by continuous etching of the carbon electrode surface [71, 72]. Similar electrochemical results have been observed using bigger GC planar electrode under application of biphasic electrical stimulation [73].

Square wave voltammetry (SWV) is a pulsed voltammetry technique where a symmetrical square-wave pulse is superimposed onto a staircase wave. SWV offers excellent sensitivity, and it can minimize the non-Faradaic capacitive charging current, so SWV is ideal for measuring basal-level neurotransmitter concentration.[28, 50, 51, 53] Figures 4A and 4B show the SWV detection of DA and 5-HT at 0 to 1000 nM concentrations for GCF electrodes. The sensitivity is defined as the slope of the calibration curve per surface area over the concentration range of 0 to 200 nM, where it is linear. We observe that at higher concentrations the analyte saturates the adsorption sites of the carbon surface, which results in the nonlinearity of peak current over concentration. SWV is used to detect tonic concentrations of DA and 5-HT in the brain, expected to be in the range of 50–100 nM [50, 51, 53], so we determined (and plotted) the electrode sensitivity as the slope of the linear range of the calibration plot at the lower concentrations up to 200 nM.

Figure 4.

SWV characterization of glassy carbon fiber electrodes and carbon fiber electrodes. A) The SWV detection of DA at different concentrations. B) The SWV detection of 5-HT at different concentrations (0, 50, 100, 200, 300, 500, 700, 1000 nM). C) and D) The calibration curves of DA and 5-HT using SWV for GCFs (C) and CFEs (D), respectively.

The average SWV sensitivity of GCF for DA (0.45 µA cm⁻² nM⁻¹) is determined to be 36% higher than the sensitivity of CFEs (0.33 µA cm⁻² nM⁻¹). For 5-HT, the sensitivity of GCFs (1.22 µA cm⁻² nM⁻¹) is 22% higher than the sensitivity of CFEs (1.00 µA cm⁻² nM⁻¹). These results further demonstrate that GCFs outperform CFEs in the electrochemical detection of neurotransmitters.

As a proof of concept of the in vivo performance of GCFs for the detection of tonic DA and 5-HT, using two different SWV waveforms specifically optimized for DA [50, 51] and 5-HT [53]. Acute experiments were performed in the dorsal striatum (DS) (richly innervated by mesencephalic dopaminergic neurons [74, 75]), and in the CA2 region of the hippocampus (where 5-HT chemistry is extensively studied [22, 76]) of isoflurane-anesthetized mice for the validation of DA and 5-HT respectively. The CGFs successfully self-penetrate the mouse brain without the use of a metal shuttle, indicating the carbon fiber-like structure is stiff enough to insert (Movie S1). Immediately after the insertion in the DS, the tonic DA response was measured using the DA SWV waveform over a 20 min period. Figure 5A shows a clear peak at 0.12 V which is characteristic of DA [50, 51], demonstrating that GCF MEAs can detect DA in the mouse DS by SWV. Using a similar method, tonic 5-HT detection was validated in the CA2 region of the hippocampus using the 5-HT SWV waveform over 20 min period. Figure 5B reveals a clear 5-HT peak at 0.29 V, similar to what we previously reported using PEDOT/CNT-coated GC electrodes [53].

Figure 5.

In vivo detection of dopamine and serotonin with GCF MEAs. A) SWV detection of tonic DA. B) SWV detection of tonic 5-HT. C) FSCV detection of stimulated DA.

In vivo DA and 5-HT concentrations were determined for all in vivo experiments by converting the SWV peak current to the DA concentration using the pre-calibration electrode sensitivity, as previously reported by our group [28, 50, 51, 53]. The DA tonic level in the mouse DS was estimated to be 40.11 ± 3.35 nM. This DA concentration is comparable to previously measured concentrations using PEDOT/CNT-coated GC MEAs in mouse DS (56.2 ± 12.3 nM) [50] and slightly lower, but still comparable, to previous measurements of basal DA levels obtained using the same SWV techniques with PEDOT/CNT-coated CFEs in the rat DS (82 ± 6 nM) [51]. These values are also consistent with those obtained using other electrochemical techniques, such as fast-scan controlled-adsorption voltammetry (FSCAV) in the mouse nucleus accumbens (90 ± 9 nM) [77] and convolution-based FSCV in the rat nucleus accumbens (41 ± 13 nM) [78]. The 5-HT tonic level in the mouse CA2 was estimated to be 68.75 ± 29.75 nM, in line with what we previously observed using SWV at PEDOT/CNT coated GC MEAs (77 ± 12 nM) [53] . This agrees with values previously detected in the CA2 region of the mouse hippocampus (64.9 ± 2.3 nM) using FSCAV [22].

We have also conducted in vivo FSCV detection of DA. A stimulating electrode was lowered into the MFB of an isoflurane-anesthetized rat and was used to stimulate nigrostriatal dopaminergic neurons. A GCF electrode was inserted into the rat DS and used to detect the consequent DA overflow. The onset of the 60 Hz, 250 µA stimulation (Figure 5C, red square) resulted in an increase in the detected background-subtracted signal. Upon the cessation of stimulation (red triangle), DA concentrations began to return to baseline over several seconds. This response is in line with previously observed stimulation-evoked DA concentration changes in our lab.[15, 52]

Finally, to test the neural recording performance of the GCF electrodes, we implanted the flexible 5 shank and 15 channel GCF MEAs into the striatum of the mouse brain for electrophysiology recordings (Figure 6). Using a drop of saline, the 5 shanks were bundled together due to the electrostatic force and were inserted together into the striatum without aid (Figure 6C). Representative waveforms of recorded single units are shown in Figure 6D. The waveforms are arranged spatially to correspond with the channel locations on the array. 14 out of 15 channels recorded high-quality single units with large peak-to-peak amplitudes and well-defined single-unit waveforms. These results demonstrate the GCF MEAs allow multi-channel electrophysiological recording.

Figure 6.

In vivo electrophysiology. A) The scheme of a 5-shank and 15-channel MEA, B) The scheme showing the location of the GCF electrode sites C) Bundled insertion of the 5-shank GCF MEA. The probe was inserted until the tip rested 4.16 mm below the dura with the electrode sites spanning the stratum. D) the representative single unit waveforms of the electrophysiology recordings from each channel, arranged according to the spatial location shown in (B) (channel 13 did not detect any single unit).

In our design, the GCF was patterned to resemble the very small cross-section of a CFE array. Our GCF electrodes can self-penetrate the mouse brain without the use of potentially damage-inducing insertion shuttles or stiffening agents. Like CFEs [79], the ultrasmall cross-section of the GCF also resulted in minimal tissue damage, as we observed in immunohistology images in Figure 7. After 1 week of implantation, the electrode track is barely visible with minimum GFAP and Iba-1 expressions indicating a very low inflammatory response [80].

Figure 7:

Histological analysis of the tissue surrounding the GCF electrodes after 1 week. Images of GCF electrode tracks in the brain (three different depts (Z) from three different implants (GCF) are reported in the three different rows) after 1 week of implantation stained for (A) DAPI (blue) (B) Iba-1 (red), (C) GFAP (white) and (D) all merged. Z1 is −750 µm, Z2 is −1.2 mm and Z3 is −1.3 mm. Scale bare is 100 µm.

To our knowledge, this is the smallest batch-fabricated GC electrode array used for in vivo sensing, and it is comparable in size to the smallest recording electrode arrays [81, 82]. Additionally, GCFs demonstrated promising capabilities in detecting both basal and phasic neurotransmitter concentrations, and they are also capable of performing electrophysiological recordings.

4. Conclusions

In this study, we utilized photolithography to engineer GCF arrays capable of neurotransmitter sensing and electrophysiology recording. Our GCF array showcased herein embodies the key characteristics of CFEs – such as small size, biocompatibility, and favorable electrochemical properties – combined with the reproducibility, adaptability and batch-production enabled by photolithography. This approach allows for the rapid and easy fabrication of GCF MEAs with varying electrode lengths, site numbers, shank configurations, and dimensions on a single wafer, offering unparalleled versatility. Notably, our GCF design enabled aid-free implantation, a significant advantage over conventional GC MEAs. Upon implantation, GCF MEAs allowed for real-time detection of tonic levels of DA and 5-HT using optimized SWV waveforms, alongside electrically evoked release of DA using FSCV in vivo. Furthermore, the GCFs demonstrated exceptional performance in recording single-unit activity with a high channel yield. This technology represents a paradigm shift in the creation of multimodal carbon-based neural probes and, with extensive optimization and customization, can be tailored to suit a wide range of neuroscience research and clinical applications.