Fabrication of a Multilayer Implantable Cortical Microelectrode Probe to Improve Recording Potential

Abstract

Intracortical neural probes are a key enabling technology for acquiring high fidelity neural signals within the cortex. They are viewed as a crucial component of brain–computer interfaces (BCIs) in order to record electrical activities from neurons within the brain. Smaller, more flexible, polymer-based probes have been investigated for their potential to limit the acute and chronic neural tissue response. Conventional methods of patterning electrodes and connecting traces on a single supporting layer can limit the number of recording sites which can be defined, particularly when designing narrower probes. We present a novel strategy of increasing the number of recording sites without proportionally increasing the size of the probe by using a multilayer fabrication process to vertically layer recording traces on multiple Parylene support layers, allowing more recording traces to be defined on a smaller probe width. Using this approach, we are able to define 16 electrodes on 4 supporting layers (4 electrodes per layer), each with a 30 μm diameter recording window and 5 μm wide connecting trace defined by conventional LWUV lithography, on an 80 μm wide by 9 μm thick microprobe. Prior to in vitro and in vivo validation, the multilayer probes are electrically characterized via impedance spectroscopy and evaluating crosstalk between adjacent layers. Demonstration of acute in vitro recordings in a cerebral organoid model and in vivo recordings in a murine model indicate the probe’s capability for single unit recordings. This work demonstrates the ability to fabricate smaller, more compliant neural probes without sacrificing electrode density.

I. Introduction

Intracortical neural probes can acquire high-fidelity neural recordings within the brain cortex and/or deliver therapeutic electrical current to a specific brain region. These probes enable the in vivo evaluation of brain functions and are viewed as promising therapeutic devices for neuro-rehabilitative applications and assistive technologies for patients with neuromuscular disorders (e.g., paralysis due to stroke or spinal cord injuries, amyotrophic lateral sclerosis (Lou Gehrig’s disease), etc.) including neurostimulation/ neuromodulation, brain-computer interfaces (BCIs), and neural prosthetics. Early intracortical probes consisted of miniaturized metal wires made from platinum, iridium, copper or stainless steel, insulated with a Teflon or polyimide coating except at the tip or housed in glass pipettes containing an ionic solution [1]–[4]. The structure of microwire probes limits the recording channel count to a single recording site per shank. As microelectromechanical systems (MEMS) fabrication techniques emerged, multiple recording electrodes could be patterned along the shank of silicon based probes, such as Michigan styled probes, which significantly increased the recording and stimulation proficiency [5].

Compared to other less invasive recording modalities like electroencephalography (EEG) and electrocorticography (ECoG) which have recording electrodes lying atop of the skull and cortex surface, respectively, intracortical neural probes have electrodes in close proximity to firing neurons. As a result, intracortical neural probes can capture single unit signals and transmit a sufficiently high bandwidth signal, so that they are capable of providing the highest-density information content for patient neural prosthetic systems [6]. In order for the use of intracortical neural probes to translate into treatments for clinically relevant neurological trauma and disease, they must function stably, ideally over a patient’s life-time [7]. However, due to their invasiveness, implanting intracortical probes leads to several safety concerns; the surgery itself involves general anesthesia, disruption of brain tissue and blood vessels as well as subsequent bleeding following probe insertion, and the possibility of infection and irritation induced by the chronic presence of the implanted probes [8]. Additionally, despite the feasibility of short-term recording potential from implanted intracortical probes, one persistent problem preventing their ability to provide a chronic, stable recording is the neural tissue response and subsequent gliosis that results in electrode encapsulation. Tissue inflammation around the implantation site induces an immune cascade and gliosis, thereby insulating and isolating the recording sites from neurons. This leads to a deterioration in recording signal quality and significantly limits their long-term recording capacity, preventing probes from providing chronically stable functionality for BCIs or other neural recording related applications [4], [9], [10].

One suspected cause of gliosis which promotes the subsequent failure of probes is the mechanical mismatch between a stiff probe, often fabricated from rigid materials such as silicon or metal, and the much softer brain tissue [11]. Micromotion and interfacial friction occurring between rigid implants and soft tissue due to respiration and body movements lead to a persistent inflammatory reaction, activation of microglia, astrocytes and oligodendrocytes, which eventually results in the formation of a glial sheath encapsulating the probes and isolating the electrodes from surrounding neurons [12]. Aside from the rigidity of the probe, other factors that contribute to the chronic instability of implanted probes in vivo are the size and shape of the implant, with larger probes expected to cause more tissue disruption and vascular damage upon surgical implantation, ultimately inducing a more severe chronic glial scar encapsulation [13]–[17]. Therefore, it has been recognized that smaller, more compliant intracortical probes produce less damage to brain tissue and elicit a weaker reactive tissue response during chronic implantation, which has propelled researchers to fabricate probes from softer polymer materials while also attempting to reduce the neural probes’ footprint [4]. However, a competing requirement is to maximize a probe’s recording potential by supporting a large number of recording sites on a single microprobe so that multiple neighboring neurons can be recorded simultaneously.

Biocompatible polymers, like polydimethysiloxane (PDMS) [18]–[20], polyimide [21]–[32], SU-8 [33]–[37], and Parylene C [7], [10], [7], [38]–[48] have all been investigated as probe substrate materials. PDMS has mainly been used as the substrate material for non-penetrating microelectrode arrays like ECoGs due to its low elastic modulus and high flexibility [20], but seldom for penetrating intracortical probes. Polyimide, due to its excellent insulating properties and ease of processing, was explored as a structural material in earlier probe fabrication [22]. Kloosterman and colleagues were able to use multiple flexible polyimide probes to record hippocampal signals in freely behaving animals for more than two months [26]. Pannu and colleagues explored the fabrication of a multilayer polyimide probe with up to 4 metal layers [49]. Xie and colleagues implemented a nanoelectronic thread to achieve a subcellular sized SU-8 probe [33], [34]. Lieber and colleagues described a SU-8 mesh-like probe which can promote close integration of the probe with the surrounding neural tissues [35], [36]. Poly(p-xylylene) (Parylene) has received a great deal of interest as a structural material for intracortical probes due to its ease of deposition in a specialized chemical vapor deposition (CVD) chamber, low permeability to water, pin-hole free deposition, low Young’s modulus, and biocompatibility [50], [51]. Parylene is designated as a USP Class VI material and has been used as a material in approved implant devices by the US Food and Drug Administration (FDA) [52]. Meng and colleagues reported various Parylene based probes to interrogate the central or peripheral nervous system. They include: (1) Parylene-based cuff electrodes with an interlocking design to ensure the electrodes to be securely wrapped around rat sciatic nerves and a microfluidic channel for local drug delivery [53]; (2) 3D sheath electrode arrays fabricated by utilizing the thermal plasticity properties of the Parylene [54], [55]; (3) long shank Parylene-based microelectrode arrays capable of detecting acute as well as chronic signals from pyramidal cells in the hippocampal layer [39], [56]; (4) Parylene probe with two layers of high density electrodes [57]. Tolosa and colleagues investigated integrating multi-functional modules into a multi-layer Parylene probe design [58]. Fedder, Cui and colleagues have also described a highly compliant Parylene probe due to its meandered arc shape [7], [59]. Xu and colleagues reported a partially flexible silicon-Parylene hybrid probe capable of penetrating the brain cortex without requiring any insertion aid while still preserving the flexibility of soft polymer probes to alleviate the mechanical mismatch between the probe and brain tissue [41]. Yoon and colleagues described an eight-shank Parylene-based probe array, on which each probe shank contained 8 micropatterned gold electrodes [43].

While polymeric probes are designed to be compliant in order to limit mechanical damage following implantation and to improve biocompatibility, they are prone to bending and buckling during implantation, especially when the size of the probes is reduced. This makes implanting compliant probes and penetrating the cortex without mechanical deformation and undue tissue damage more difficult [60]. Therefore, multiple strategies have been explored to aid effective implantation of flexible polymeric probes. One insertion approach is to use rigid insertion shuttles as backbones to support the flexible probe during insertion. The insertion shuttle can then be removed after the probes reach their target location in the cortex [27], [61]. Pannu and colleagues utilized molten polyethylene glycol (PEG) as an adhesive to attach polyimide probes to silicon insertion shuttles which can be retrieved after the PEG dissolves in the moisturized in vivo environment [27]. A recent alternative approach involves encapsulating the probes (or portions thereof) within biodegradable polymers which temporarily enhance the mechanical rigidity of the probes to allow effective insertion and subsequent degradation within a short period of time post-implantation, leaving behind the compliant probes at the target location. Common coating polymers include PEG [39], [40], carboxy-methyl-cellulose (CMC) [59], silk [62], [63], and saccharose [22]. Dip-coating is the most common coating method which allows for uniform probe coverage of the degradable polymer; however, the coating thickness and shape cannot be precisely controlled and the electrodes are covered by the coating which makes it difficult to record directly following implantation. As an alternative to dip-coating, micromolding in capillaries (MIMIC) is used to define coating dimensions within a microfluidic mold structure placed over the probe. Lecomte et al. used a PDMS mold to encapsulate the whole probe shank with PEG and silk [38]. Meng and colleagues reported using a PDMS mold to partially encapsulate a length of a Parylene probe within PEG [39]. This left the distal portion of the probe shank exposed which shortened the shank length so that it had sufficient rigidity to resist Euler buckling during insertion. Previously, our group adopted the MIMIC process to encapsulate Parylene probes within a degradable tyrosine-derived polycarbonate with controlled dimensions and geometry [64]. Using this method, Lo et al. [65] fabricated a series of Parylene probes of varying dimensions and coating sizes and evaluated the encapsulation response in a chronic animal study through immunohistochemical staining. It was found that by using an appropriately sized coating dimension and probe sizes smaller than 80 μm wide, the degree of gliosis could be greatly attenuated . This finding served as the design parameters for further developing the Parylene probes reported herein.

The small footprint of the compliant probe shank (80 μm wide), combined with the limited feature resolution offered by conventional long-wavelength ultraviolet (LWUV) lithography and polymer micromachining techniques significantly limits the number of recording electrodes which can be defined on each probe shank if they are defined as a single layer on a Parylene substrate as is conventionally done. When the number of recording sites is limited, such probes might lack the spatiotemporal resolution required for human neuroprosthetic use, such as driving robotic limbs with a high degree of freedom [43].

In order to overcome this limitation, we developed a novel compliant neural probe fabrication approach defining 16 individually patterned recording electrodes using conventional LWUV optical lithography on multiple supporting layers of Parylene to create an 80 μm wide by 9 μm thick microprobe. Each recording electrode consists of a 30 μm diameter recording window and 5 μm wide connecting traces. The fabrication of the multilayer probe involves repeated photolithographic definition of titanium/platinum (Ti/Pt) electrodes along with CVD of Parylene-C insulation as a layered laminate, with the final probe geometry defined via reactive ion etching (RIE) to remove the unmasked field Parylene in an oxygen plasma. The recording sites, along with their connection traces on each layer, are insulated from adjacent layers by an intervening Parylene C insulation layer which also serves as a support for the next layer of electrodes. Since the connection traces can overlap on different Parylene support layers, this approach allows us to increase the total number of recording channels without proportionally increasing the width of the probe shank which largely determines the invasiveness of the probe and resultant tissue immune response.

In this paper we describe the design, fabrication, and initial evaluation of our multilayer microelectrode probe focused toward addressing the aforementioned challenges of balancing polymer probe size with number of recording sites in order to improve recording potential while also enhancing chronic stability of the acquired signals.

II. Material and Methods

A. Probe Design and Fabrication

Fabrication starts with a 4-inch silicon wafer. A sacrificial layer consisting of 5 nm chromium, 50 nm gold, and 20 nm chromium was sputtered on top of the wafer in order to allow the final release of the fabricated probes [66]. Next, a 3 μm thick Parylene layer was deposited via CVD (SCS Labcoter^®2 Parylene Deposition System PDS 2010). Afterwards, the wafer was treated for 1 minute in an O₂ plasma (March PX-250) (100W, 80 mTorr) to roughen the Parylene layer. This is followed by the patterning of the first layer of Ti/Pt electrodes, traces, and connection pads via photolithographic definition of the electrode geometry, physical vapor deposition (PVD) of a 200 nm thick Ti/Pt layer (Kurt J. Lesker PVD75), and lift-off in acetone (Fig. 1a and Fig. 1b). Following that, A174 adhesion promoter was applied. The O₂ plasma and A174 adhesion promoter were adopted to prevent the multiple Parylene layers from delaminating. Then a 1 μm thick layer of Parylene was deposited, which serves as both an insulation layer as well as the support layer for patterning the second layer electrodes (Fig. 1c). The metal electrode definition, layer treatment, and Parylene deposition were repeated multiple times to achieve the desired layers of electrodes (Fig. 1d).

Fig. 1.

Fabrication procedure of multilayer probes containing 4 layers of recording sites. (a) A base sacrificial layer of Cr/Au/Cr and support layer of Parylene are deposited; (b) the first layer of electrodes and connection pads are patterned via lift off; (c) a second insulating layer of Parylene is deposited; (d) the steps highlighted in (b) and (c) are repeated to create the multilayer electrode set; (e) the probe geometry, recording windows, and connection pads are defined through a Cr mask; (f) the excess Parylene is etched away by reactive ion etching (RIE); (g) the probe is released by wet etching the top Cr mask and the base layer of sacrificial Cr/Au/Cr.

Following the deposition of the final 3 μm Parylene insulation layer, a 4000Å thick chromium mask was patterned to define the probe geometry, recording windows, and connection pads (Fig. 1e). The recording sites and connection pads were exposed by etching through the Parylene using RIE in an oxygen plasma (JLS Design Ltd Mini-Lab Plasma-Pod System) for 100 min (50W, 80 mTorr) where the etch would stop on exposed platinum films to open up the connection pads and recording windows on each layer. The probe geometry was defined simultaneously by RIE as well (Fig. 1f). Following RIE etching, the probe was released by wet etching the sacrificial layer in Cr etchant (Fig. 1g). The Au and Cr layers of the sandwich sacrificial layer form an anode and cathode of an internal battery, providing an increased corrosion current, allowing the probe to be released much faster [66].

Fig. 2a shows the design used to pattern 16 electrodes on four different layers, with 4 electrodes on each layer and a total probe width of 80 μm. Fig. 2b indicates the dimensions of the important design features of the shank: recording window diameter: 30 μm; recording electrode pad diameter: 40 μm; trace width: 5 μm; spacing: 5 μm; and margin on the edge: 5 μm. The recording electrode size, trace width and spacing were chosen as a balance between our lithographic resolution, sequential mask alignment tolerances, and overall fabrication yield while maintaining an 80 μm overall probe width and recording electrode impedances in the range capable of resolving single unit activities. Fig. 2c shows an exploded view of each individual layer and Fig. 2d shows a SEM image of a completed probe with 16 recording sites fabricated on 4 electrode layers.

Fig. 2.

A four layer electrode neural probe with its shank containing 16 electrodes. (a) Schematic of the mask design with a different color for each electrode layer illustrating how the 16 electrodes are patterned on four different layers, as well as a top mask to define the final probe geometry and open recording windows (scale bar: 80 μm). (b) A zoomed-in view of a single recording window depicted by the blue box in (a). Numbers represent the dimension of each feature. (c) Exploded view of the multi-layer probe. (d) SEM image showing the open 16 recording sites along the probe shank. Each layer of electrodes is delineated by the dashed line (scale bar: 100 μm). The insets show a zoomed-in image of a recording window from each layer. Scale bars in all four zoomed-in images represent 10 μm.

In order to assess potential delamination of the Parylene layers in an in vivo environment, we placed probes fabricated with an earlier 6 channel design but using the same fabrication protocol in a 1× PBS solution at 37°C for up to 14 days. We then inspected the probes via SEM for evidence of delamination as well as comparing the electrodes’ impedance before and after soaking.

B. Electrical Package

External electrical connections with the multilayer probe were established via zero-insertion-force (ZIF) connectors (Hirose Electric. Part#: FH12-16S-0.5SH(55)). A poly(ether ketone) (PEEK) polymer sheet was attached to the connection pad region of the probe with EPO-TEK 301 adhesive (Epoxy Technology) to stiffen the Parylene connection pad region and match the required cable thickness for the ZIF interface. The thickened connection pad region was inserted into the ZIF connector, the pins on ZIF connector and connection pads were aligned, and the connection pad was lock flipped into place. Fig. 3a shows a completed probe held by tweezers, with its connection pad region attached to PEEK to facilitate its electrical connection to the ZIF connector. The connection between the ZIF connector and the external recording or testing equipment was achieved through a customized PCB and a 16-channel Omnetics connector (A79038-001). A fully packaged probe is shown in Fig. 3b.

Fig. 3.

(a) Photograph of a multi-layer probe with its connection pad region thickened by PEEK (scale bar: 2mm). (b) Photograph of a fully packaged device. The probe shank is attached to a SU-8 insertion aid (scale bar: 2mm). (c) Optical microscope image of a probe attached to a SU-8 insertion aid. The droplets on the SU-8 insertion aid are from condensation of the EPO-TEK 301 epoxy used to affix the probe to the SU-8 (scale bar: 200 μm).

Probes to be implanted in vivo were also affixed to a SU-8 insertion shuttle to provide sufficient rigidity during insertion. Briefly, a 35 μm thick SU-8 2075 layer was spin cast onto a silicon wafer which was precoated with a 40 nm thick Ominicoat (Microchem) sacrificial layer. Then the SU-8 was lithographically patterned in the shape of the probe with the support having a 200 μm wide and 3 mm long shank. The SU-8 was developed and removed from the wafer. After drying, the SU-8 shuttle was dipped onto a thin spincast layer of EPO-TEK 301 adhesive and affixed to a Parylene probe, taking care not to get epoxy on the front of the probe. The probe tip is roughly 0.8 mm - 1 mm above the tip of the insertion aid (Fig. 3c). This probe/shuttle structure was then inserted into the electrical package for cleaning and characterization prior to animal implantation.

C. Electrical Characterization

Following fabrication and packaging, probes were cleaned by cyclic voltammetry (CV) in diluted H₂SO₄ (−0.2V–1.2V, scan rate 0.1V/s for 30 cycles) to remove any remaining photoresist and other impurities on the electrode surfaces, especially due to the oxygen plasma etch which defined the probe shape and opened the recording windows. A Platinum mesh electrode was used as a counter electrode and a Ag/AgCl electrode as a reference electrode [56].

The impedance of each electrode on the multilayer probe were characterized through frequency sweeping performed in a beaker of 1× phosphate buffered saline (PBS) solution (VWR Life Science) with a 50 mV_rms excitation voltage over a frequency range of 100 Hz to 10 kHz.

D. Crosstalk Evaluation

One concern which arose from minimizing our probes’ footprint is that the small distances between electrical traces may increase the coupling capacitance between adjacent layers because the neighboring layers are only insulated by a 1 μm thick Parylene-C layer and a conductive path through the fluid environment. This may lead to signals carried on one trace to be detected by traces from neighboring layers, resulting in crosstalk and signal contamination [67].

Assuming the shunting capacitance and recording instrument input impedance are zero, then the equivalent circuit model for two overlapping recording lines from neighboring layers is depicted in Fig. 4a and the crosstalk can be expressed as follows:

$$\mathit{\text{ Crosstalk }}=\frac{V_{out2}}{V_{in1}}\mathit{\text{ when }}{V}_{in2}=0$$ (1)

Fig. 4.

Crosstalk evaluation of the electrodes between neighboring layers. (a) Equivalent circuit model of two adjacent recording channels vertically overlapping with each other. Z_electrode represents the electrode electrolyte impedance, and Z_p is the parasitic coupling impedance composed of parasitic capacitance and resistance of the PBS solution. (b) Schematic illustration of the setup to evaluate the crosstalk. All 16 electrodes are submerged in a beaker filled with 1× PBS solution. V_in1 (50 mV_rms, 100 – 10kHz) was applied on one channel and the output V_out2 from its vertically overlapping channel was monitored.

The crosstalk of the multilayer probe was evaluated by submerging the probe in 1× PBS and stimulating one layer of electrodes with a 50 mV_rms AC voltage scanning from 100 Hz to 10 kHz while recording the output voltage from the neighboring layers in order to determine the magnitude of crosstalk using the setup shown in Fig. 4b.

E. In Vitro Cerebral Organoid Recording

In order to verify the probe’s capability of recording neural signals, electrophysiological measurements were first performed on cerebral organoids generated from human induced pluripotent stem (hiPS) cells. Cerebral organoids were generated using modified published protocols [68]–[70]. Briefly, hiPS cell embryoid bodies were induced by dual-SMAD inhibition towards a neuroectodermal fate for 1 week in neural induction medium in low-adherence 6-well plates. Organoids were then moved to an orbital shaker (110 RPM) in neural maturation medium until recording at day 63 or 81 with a diameter of approximately 2–3 mm.

In order to hold the organoid in place during probe implantation and recording, a dual-chamber PDMS device consisting of two interconnected mini-wells was fabricated using conventional soft lithography techniques and bonded onto a glass coverslip (Fig. 5a). Prior to recording, the PDMS wells were filled with an artificial cerebral spinal fluid (aCSF) solution (125 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 25 mM NaHCO₃, 2.5 mM CaCl₂, 1.2 mM MgCl₂, pH 7.4 bubbled with 5% CO₂, 95% O₂). The organoid was then transferred from incubator into the recording bath. A grounded reference screw was placed into the adjacent mini-well and the packaged probe was carefully controlled by a micromanipulator to penetrate the organoid (Fig. 5b). All electrophysiological recordings took place inside of a grounded Faraday cage.

Fig. 5.

Organoid recording setup. (a) Schematic illustration of the setup for organoid recording. The PDMS microdevice is composed of two interconnected mini-wells on a glass slide. One well is used to hold the organoid in place while the other is for placement of the reference screw to ground the solution surrounding the organoid. (b) Top-down view of the probe penetrating the organoid held inside of the PDMS well.

After placing the 16 microelectrodes inside of an organoid, four sessions of baseline recording were taken using a portable ME32-System amplifier (Multichannel Systems, Germany) with each session lasting approximately 1 minute. Then, a KCl solution was applied to the organoid mini-well to a final concentration of 60 mM to depolarize neurons within the organoid. After the KCl addition, three sessions of recording were taken with each session lasting approximately 5 minutes. Finally, sodium channel blocker tetrodotoxin (TTX) was added at a 1 μM concentration to terminate neural activity by inhibiting action potential generation and propagation. Following TTX addition, three sessions of recording were taken with each session lasting approximately 5 minutes. The recording was performed in two different organoids at day 63 and day 81. The sampling rate was 25 kHz, and a 500 Hz low pass and high-pass filter was applied to extract local field potentials (LFPs) and single-unit recordings, respectively.

After recording experiments, organoids were fixed, sectioned, and processed for immunofluorescence staining using previously reported protocols [71]–[73]. Briefly, organoids were fixed in 4% paraformaldehyde, cyroprotected in 15% and 30% sucrose, embedded in OCT freezing medium, snap-frozen, and sectioned at 20 μm using a cryostat. The primary antibodies used for immunohistochemistry include TBR1 (rabbit), CTIP2 (rat), MAP2 (chicken), C-FOS (rabbit). Alexa Fluor secondary antibodies were used at a 1:500 dilution. Slides were mounted using Fluoromount-G medium containing DAPI (Southern Biotechnology). Images were obtained using a Zeiss 710 confocal microscope. Fluorescent micrographs were processed using Fiji software (NIH ImageJ).

F. Surgical Implantation and in Vivo Acute Recording

Wild-type male mice housed at the Nelson Animal Resources Center at Rutgers University (in a 12-hour light/dark cycle at 22°C, with food and water ad libitum) were used for the multilayer probe implantation. Prior to surgery, mice were placed in an induction box and exposed to isoflurane (4%) vaporized in oxygen. Once general anesthesia was reached, the animals were transferred to the surgical table and isoflurane (2%) was delivered via nosecone. Body temperature was maintained with a feedback controlled heating blanket. After the skull was exposed, a craniotomy, 1 mm in diameter, was made in the bone of the skull above the motor cortex region, specifically under the SI cortex coordinates of 1.4 mm anterior, 1.4 mm lateral from bregma and the dura was removed. The probe with the SU-8 insertion aid attached was inserted with the electrode holder of the stereotaxic equipment at a speed of approximately 1 mm/min, as shown in Fig. 9a. When the electrodes on the probe reached the target depth of 1 mm beneath the cortex surface, four sessions of acute recording were taken with a portable ME32-System amplifier (Multichannel Systems, Germany) with each session lasting 3 minutes. The sampling rate was 25 kHz, and a 500 Hz high-pass filter was applied for single-unit recording. The high-pass filtered signal was then post-processed offline using a sorting software (Plexon Inc.,Dallas, TX). Briefly, spikes above a set threshold are first detected. Time-coincident spikes present in more than 4 channels are categorized as an artifact caused by motion or interference from the environment and are invalidated. Then, the remaining spikes are sorted with automatic valley seeking method. All surgical procedures and recordings were performed as in previous work under a protocol approved by Rutgers’ Institutional Animal Care and Use Committee (IACUC) [74].

Fig. 9.

Surgical implantation and acute in vivo recording. (a) Photograph of a mouse ready for probe implantation. (b) Typical spontaneous unit activities recorded by 16 electrodes on an implanted multi-layer probe from motor cortex in an anesthetized mouse. A high-pass filter (500 Hz) was applied. The numbers on the left of the plot represent the channel number and correspond to the microelectrodes located along the shank in Fig. 8f (vertical scale bar: 100 μV; horizontal scale bar: 0.1 s). (c) (d) Raster plots showing spikes in each channel before and after cross-channel artifacts were removed. Time-coincident spikes present in at least 4 channels were categorized as artifacts (scale bar: 1s). (e) Single units isolated from each channel (vertical scale bar: 200 μV; horizontal scale bar: 1.28 ms).

III. Results

A. Electrical Characterization

Figure 6 shows the results of electrochemical cleaning (Fig. 6a) and impedance characterization (Fig. 6b) of a completed probe. Following electrochemical cleaning of the recording sites in H₂SO₄, the 30th cycle CV curve encloses a larger region compared to the 2nd cycle CV curve indicating a larger electroactive surface area of the electrode, resulting from the electrochemical cleaning of the recording windows.

Fig. 6.

(a) CV curve of all 16 electrodes cleaned in 0.5M H₂SO₄. (b) Impedance magnitude vs. frequency within PBS solution. Electrodes were stimulated at 50 mV and frequency swept from 100 Hz to 10 kHz while the impedance was monitored.

Following cleaning, impedance spectroscopy of the multilayer probe confirmed that all microelectrodes along the probe shank had impedance values capable of acquiring resolvable neural spikes. Frequency sweeping of all 16 electrodes revealed that the impedance values of all 16 electrodes fall within the range of 0.3 MΩ-0.6 MΩ at 1 kHz, which is within the optimal range reported for single unit recording (Fig. 6b) [39].

When assessing changes in probe electrode impedance following soaking in 1× PBS for 14 days, we did not see a consistent decrease in electrode impedance due to delamination. We observed that some of the electrodes showed an impedance increase following soaking which was attributed to corrosion occurring at the electrode surface or other factors contributing to this impedance increase.

When evaluating crosstalk between electrodes, one layer was stimulated with 50 mV excitation, while voltage signals sensed by all the individual electrodes from their neighboring layers are monitored. The crosstalk between these two layers is calculated as the average of the sensed signals divided by the 50 mV input. Fig. 7 shows that at 1 kHz, the crosstalk between all vertically neighboring layers is below −36.5 dB, which corresponds to less than 1.5% signal transmission between layers due to crosstalk.

Fig. 7.

Crosstalk versus frequency between different neighboring layer pairs. Each curve corresponds to the crosstalk between three pairs of vertically adjacent layers. The dashed horizontal line delineates a crosstalk of 1.5% (−36.5dB).

B. In Vitro Cerebral Organoid Recording

When recording from the day 63 cerebral organoid, minimal spontaneous activity was observed prior to KCl addition (Fig. 8a). After KCl was added to induce depolarization, neural activity was recorded in all 16 channels (Fig. 8b). Neurons on the outer organoid surface are expected to be at a higher KCl concentration than the more central neurons due to the proximity of the tissue to the outer solution and the need for KCl to diffuse through the organoid. Spiking events are indicated by arrows in Fig. 8c with the largest number of spikes occurring in channels 13 and 15, which correspond to the two topmost microelectrodes along the probe shank closest to the edge of the organoid where the KCl has penetrated the organoid (Fig. 8f). In contrast to the relatively sparse spiking activity, consistent low-frequency local field potentials (LFPs) were detected across all channels following KCl addition (Fig. 8d). A conspicuous drop in neural activity was observed following TTX addition, as expected (Fig. 8e).

Fig. 8.

In vitro recording traces. (a) Recording traces before KCl was applied. Scale bar: 100 μV (vertical) and 5 s (horizontal). (b) Recording traces after KCl was applied. Scale bar: 100 μV (vertical) and 5 s (horizontal). (c) High pass (>500 Hz) filtered signal of recording traces in (b). Spike activities are indicated by arrows. Scale bar: 30 μV (vertical) and 5 s (horizontal). (d) Low pass (<500 Hz) filtered signal of recording traces in (b). The initial spikes which are the same across all 16 channels represents the artifacts created by KCl addition. Scale bar: 30 μV (vertical) and 5 s (horizontal). (e) Recording traces after TTX was applied. Scale bar: 100 μV (vertical) and 5 s (horizontal). Recording traces after TTX was applied. Scale bar: 100 μV (vertical) and 5 s (horizontal). (f) Schematic of the microelectrode number map. In (a)-(e), the numbers listed on the left of each plot represent the channel number and correspond to the microelectrodes located along the shank in (f).

Immunofluorescence staining in Fig. S1 revealed high expression of neuron-specific markers TBR1, CTIP2, and MAP2 in cerebral organoids, confirming their neuronal identity. Robust expression of neuronal cytoskeletal protein MAP2 is demonstrated in dendrites within the organoid (Fig. S1a). C-FOS is an Immediate Early Gene (IEG) and indirect marker of neuronal activity that is upregulated by action potential firing [75]. Widespread C-FOS expression was observed in a day 81 organoid exposed to 60 mM KCl for 1 hour (Fig. S1c).

C. In Vivo Recording

The multilayer probe was also used to record acute spontaneous neuronal signals from the mouse motor cortex. Neural action potentials were detected in most of the channels, as shown in the high pass filtered signal data indicated in Fig. 9b. Following the spike detection, spikes in each channel were sorted to individual units shown in Fig.9e.

Portions of the recordings displayed synchronized spikes across multiple channels within the high pass filtered signal shown in Fig. 9b. These are artifacts resulting from environmental noise, since the recording was taken right on the surgical table. These artifacts were removed in the subsequent signal processing and spike sorting procedure. They are displayed as raster plots showing when events occur on a given probe electrode before (Fig.9c) and after artifact removal (Fig.9d).

In some cases, more than one single unit was detected on a single electrode (Fig.9e).

IV. Discussion

The purpose of this study is to introduce a multilayer approach to increase the number of recording sites vertically, rather than laterally. Thus, the main determinant of the probe invasiveness, the probe width, remained constant. Although more advanced lithography techniques, like E-beam or EUV lithography can achieve submicron features, which make it possible to pattern more recording electrodes on a fixed probe width [36], patterning narrower metal interconnects with smaller spacing also increases the likelihood of crosstalk between adjacent channels. Using a multilayer approach can keep both the lateral and vertical distance between electrode traces constant when additional recording sites are added on the probe, minimizing the risk of crosstalk. Additionally, these higher resolution patterning techniques are inherently more expensive and complicated to process compared to LWUV lithography.

Although the multilayer fabrication approach could be adapted to produce electrode arrays with a varying number of layers and electrode densities, the number of layers cannot be increased endlessly due to several fabrication limitations. This includes the accumulation of film stress with increasing number of layers, accumulation of lithographic alignment errors, decreased polymer etching rates through thicker films, and an increased distance of the deeper electrodes from the tissue interface which could reduce recording sensitivity. Defining each electrode layer requires depositing a supporting layer of Parylene followed by metal deposition and patterning; the associated film stress builds up during the deposition and processing steps. Increasing the number of layers increases the propensity of the probe to curl upon release from the substrate and/or cause electrode failure. Additionally, as the number of layers is increased, the bottommost layers of electrodes will be further away from the probe surface. Thus, opening the recording site window of the bottommost electrode layers via RIE etch will be more difficult because the etch rate could be substantially decreased when etching deep trenches due to the microloading effect. Increasing the number of layers also increases the thickness of the probe such that the bottommost layer electrodes are at a greater distance from the tissue interface, which may increase the recording impedance and/or ability to record single unit activity. Finally, increasing the number of layers also increases the allowable alignment tolerance. Due to the complexities of probe fabrication, our current design allows for a broad +/−5 μm misalignment between layers, which is sufficient for up to four layers. Increasing the number of layers would require a larger misalignment allowance which may result in a wider probe shank, thus offsetting the benefits of the multi-layer fabrication approach. Other issues that arise with adding more electrode layers include a higher possibility of layer delamination and an overall longer, more complicated fabrication protocol. Thus, at some point it becomes both cost and technically prohibitive to design and fabricate probes with a number of layers above a certain threshold.

When the recording windows were opened by RIE, the time to expose the electrodes varied across different layers. The electrodes from the topmost layer will be exposed first, and those from the base layer will take longer time to expose. However, we did not observe any electrode window size variation or damage of the electrode from the topmost layer, which can be seen from the SEM images in Fig.1d, Fig S2 and Fig S3. The zoomed-in image of the electrodes from different layers show that all recording windows are 30 μm in diameter, which is the same as the diameter of the windows defined on the Cr etching mask. This indicates good anisotropy of the RIE process.

However, when considering the recording windows in the multilayer design probe, the depth from the probe surface to the electrode surface on different recording layers are different, which might cause electrode impedance variation across different layers. However, how the depth of electrode affects its impedance cannot be generalized here. As depicted in Fig. 6(b), the electrodes from the same layer might have higher or lower impedances than the impedance of electrodes from another specific layer. In earlier designs we tested with larger electrodes (80 to 100 μm in diameter), which had lower overall impedances, the effect of electrode depth on their impedance was more pronounced. Deeper electrodes generally had larger impedance compared with their shallower counterparts. One possible reason is that as the electrode is further away from the probe surface, there is a longer conductive pathway which results in an increase of the overall electrode impedance. This effect was only be observed for the larger low impedance electrode. As the electrode size is decreased and electrode impedance increases, this trend is masked by other factors.

Critically, electrodes patterned at different depths could affect signal acquisition during recording, since electrodes from the bottommost layer are at a greater distance from firing neurons, compared to other shallower electrodes. Future work will explore planarizing the electrode surfaces by electrodepositing conductive layers into the trenches to ensure all recording interfaces are at a constant depth. This can be accomplished by electroplating gold or conductive polymers such as poly (3,4-ethylenedioxythiophene) (PEDOT) layers of varying thickness onto the electrode surfaces from different layers. The use of PEDOT may not only compensate for the distance differences of electrodes to firing neurons, but also provide a more intimate contact between the electrode surface and the soft brain tissue [76], since these polymers have substantially lower Young’s modulus than metal.

The crosstalk between all vertically neighboring layers was found to be lower than 1.5% at 1 kHz. The crosstalk between the 2nd layer and 3rd layer is around 0.5%, much lower than the other two layer pairs (Top layer versus 2nd layer: 1.4%; 3rd layer versus Base layer: 1.3%). The relatively high crosstalk between the top and 2nd layer is a bit unexpected, since overlapping traces formed by these two layers are the shortest compared with other two layer pairs, thus they have the smallest overlapping area and would be expected to generate the lowest parasitic coupling capacitance and have the lowest crosstalk rate. This inconsistency indicates that other factors aside from the parasitic coupling capacitance may also impact the crosstalk rate. In Gustavo’s thesis [77], he investigated the influence of the electrode impedance on crosstalk rate and concluded that as the electrode impedance drops, the crosstalk rate will also decrease, if other parameters remain constant. Thus, in our future design, two different strategies may be explored to further decrease the crosstalk: (1) decrease the electrode impedance, which can be achieved by electroplating PEDOT on the electrode surface as mentioned above; (2) decrease the parasitic coupling capacitance by slightly increasing the thickness of the laminate Parylene layers.

When considering in vitro recordings from cerebral organoids, they do not contain blood vessels and other connective tissue found in brain tissues. Thus, the organoids have a much lower stiffness which allowed penetration of our flexible Parylene-based probe without the need for an insertion aid. When probes were retracted from the organoid at the end of the recording, no discernable implantation damage is observed at the implantation site, possibly due to the small size and relatively sparse existence of cells compared to brain tissue. Additionally, at 63 days, the cerebral organoid is considered relatively immature which likely contributed to the observed lack of spontaneous electrical activity. Recent studies have demonstrated the emergence of network spiking activity and low-frequency oscillatory dynamics in cerebral organoids over prolonged culturing periods [78], [79]. In future studies, recordings in more mature organoids may present more elevated firing activity and synchronous network events, as previously reported. Thus, the organoid recording setup described herein can be used as an in vitro model to validate the probe’s electrophysiological recording capability, as well as offer a quantitative measure of the development of organoid electrophysiological dynamics under different conditions.

Future studies will also focus on evaluating chronic recording capabilities of the multilayer probe. Therefore, rather than using a rigid SU-8 insertion aid as we adopted for acute recording evaluation in this study, probes can be coated with degradable polymers to aid insertion [64]. The smaller, compliant intracortical probe is expected to limit the reactive tissue response during chronic implantation while still allowing multichannel recording capabilities.

Delamination between different Parylene layers has also been reported and might prevent probes from functioning in a chronic recording setting [80]–[82]. The application of A174 adhesion promoter and roughening the previous Parylene layer via O₂ plasma was adopted to prevent Parylene delamination in our fabrication procedure. After soaking our probes in 1× PBS at 37°C for up to two weeks we saw no evidence of delamination through impedance changes or SEM evaluation. Although no delamination was observed in our short term test (Fig.S2, Fig. S3), longer term implantation in a living environment may give rise to delamination. High temperature annealing to strengthen the adhesion between Parylene layers has also been proposed as a method to prevent delamination [83]–[85] and we could investigate this approach if needed.

V. Conclusion

A multilayer fabrication approach was adopted to produce compliant intracortical Parylene neural probes, allowing for increasing the total number of recording channels without proportionally increasing the size and invasiveness of the final probe. Electrical characterization of the 16 recording electrodes confirms good functionality of the probe with a low degree of crosstalk between adjacent layers. In vitro and in vivo acute recording test demonstrated the probes’ capability for single unit recording.

Biography

Xin Liu received the B.E. degree in control technology and instrument from Wuhan University in 2003, and the M.S. degree in marine molecular biology from the Chinese Academy of Sciences in 2007. She is currently pursuing the Ph.D. degree in biomedical engineering with the BioMEMS and Microfluidics Laboratory, Rutgers University, under the supervision of Dr. J. D. Zahn. Then, she entered medical equipment industry. Her current research interests include flexible neural probe fabrication and neurosciences.

Yelena Bibineyshvili received the M.S. degree in 2010, and the Ph.D. degree in biophysics from Moscow State University, Russia, in 2015, under the supervision of Prof. G. Maksimov. From 2016 to 2019, she was a Post-Doctoral Associate with the Department of Cell Biology and Neuroscience, Rutgers University, NJ, USA, with Prof. D. Margolis. She currently works as a Post-Doctoral Associate with the Department of Anesthesiology, Weill Cornell University, NY, with Prof. D. Calderon. Her research is focused on the cortical activity in alive mice experiencing brain injury, coma, and other brain disorders.

Denise A. Robles received the B.E. degree in biomedical engineering from the Macaulay Honors College, The City College of New York in 2017. She is currently pursuing the Ph.D. degree in biomedical engineering with the BioMEMS and Microfluidics Laboratory, Rutgers University, under the supervision of Dr. J. D. Zahn. Her research focus is in the development of miniaturized technologies that closely model neurological systems and disorders.

Andrew J. Boreland received the B.A. degree in genetics from Rutgers University in 2016, where he is currently pursuing the Ph.D. degree in cell and developmental biology, under the supervision of Dr. Zhiping Pang and Dr. Peng Jiang. His research focuses on human stem-cell derived neural systems in both 2-D culture and 3-D cerebral organoids for modeling neuropathology. He has been supported by two NIH-sponsored training programs in Biotechnology (T32) and Clinical and Translational Science (TL1).

David J. Margolis received the Sc.B. degree in neuroscience from Brown University in 1999, and the Ph.D. degree in neurobiology and behavior from the University of Washington, Seattle, in 2007. He did post-doctoral work at the University of Zurich, Switzerland, before starting as an Assistant Professor at Rutgers, the State University of New Jersey in 2013. He is currently an Associate Professor of cell biology and neuroscience, where his laboratory studies the neural circuitry underlying sensory processing, plasticity, and learning in mouse models. His research has been funded by the National Institutes of Health, the National Science Foundation, the Brain and Behavior Research Foundation, the Whitehall Foundation, and among others. He serves as an Associate Editor for the journal Science Advances.

David I. Shreiber received the B.S. degree in mechanical and aerospace engineering from Cornell University in 1991, and the M.Eng. and Ph.D. degrees in bioengineering from the University of Pennsylvania in 1993 and 1998, respectively. After post-doctoral studies at the University of Minnesota in chemical engineering and materials science, he joined the Faculty at Rutgers in 2002, where is he currently a Professor and the Chair of biomedical engineering. His research interests span biomechanics and biomaterials with an emphasis on regenerative medicine. His research has been supported by the NSF, NIH, CDC, and other state commissions and private foundations. He was a recipient of the Whitaker Foundation Young Investigator Award, the NSF CAREER Award, and the Johnson & Johnson Discovery Award.

Jeffrey D. Zahn received the B.S. degree in chemical engineering with a minor in biology from the Massachusetts Institute of Technology (MIT), Cambridge, MA, USA, in 1995, and the Ph.D. degree in bioengineering from the Graduate Group in Bioengineering, University of California at San Francisco, CA, USA, and University of California at Berkeley, CA, in 2001.

He completed post-doctoral studies with the Department of Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh, PA, USA, in 2001. From 2001 to 2006, he was an Assistant Professor of bioengineering with Pennsylvania State University, State College, PA. Since 2007, he has been with Rutgers, the State University of New Jersey, Piscataway, NJ, USA, where he is currently a Professor of biomedical engineering. His research interests include biomedical microdevices, electroporation, bioMEMS, and microfluidics. His research has been supported by NSF, NIH, and the New Jersey Commission on Spinal Cord Research. He was a recipient of the ADA Young Investigator Award and the Wallace H. Coulter Foundation Early Career Translational Research Award.