Nanotopography-enhanced biomimetic coating maintains bioactivity after weeks of dry storage and improves chronic neural recording

Abstract

We developed a nanoparticle base layer technology capable of maintaining the bioactivity of protein-based neural probe coating intended to improve neural recording quality. When covalently bound on thiolated nanoparticle (TNP) modified surfaces, neural adhesion molecule L1 maintained bioactivity throughout 8 weeks of dry storage at room temperature, while those bound to unmodified surfaces lost 66% bioactivity within 3 days. We tested the TNP+L1 coating in mouse brains on two different neural electrode arrays after two different dry storage durations (3 and 28 days). The results show that dry-stored coating is as good as the freshly prepared, and even after 28 days of storage, the number of single units per channel and signal-to-noise ratio of the TNP+L1 coated arrays were significantly higher (32% and 40% difference, respectively) than uncoated controls over 16 weeks. This nanoparticle base layer approach enables the dissemination of biomolecule-functionalized neural probes to users worldwide and may also benefit a broad range of applications that rely on surface-bound biomolecules.

1.0. Introduction

Proteins and other biomolecules manufactured by cells are used as key components of biosensors for chemical detection, [1–3] blood oxygenation, [4, 5] purification, [6, 7] biotechnology and pharmaceutical industry as well as surface modifications of bio-interface devices. [8–11] Despite their utility, the use of biomolecules is hindered by their short operational lifetime, resulting in losses of bioactivity under physiological conditions. For example, glucose detection with glucose oxidase decays over time, likely due to the degradation of the protein itself. [12–14] Carbonic anhydrase used for CO₂ capture rapidly loses activity when stored suspended in solution, [15] and a separate study observed that carbonic anhydrase lost 60% of its catalytic activity after 20 days in saline at 25°C. [16] The bioactivity of surface-immobilized biomolecules may further reduce upon drying, as such maintaining bioactivity during shipping and storage is a common challenge for immunoassays, protein arrays, enzyme-based biosensors, and peptide-functionalized medical implants. [17–21]

A particularly interesting use of biomolecule functionalized surfaces is for improving the biocompatibility of implantable neural electrodes. Chronic neural recording with microelectrode arrays (MEAs) is often limited by a gradual degradation in single unit yield, i.e. the % channels capable of detecting neuronal spikes. [22] This is in part due to the inflammatory host tissue reactions which occur after the MEA is implanted. [8, 22] These inflammatory reactions result in acute and long-term damage to the blood-brain barrier (BBB), [23] activation and migration of host glial cells to the site of the implant, and loss of viable neurons near the electrode sites. [8, 24] It is not surprising then that substantial research effort has been invested into controlling and minimizing this inflammation. One effective strategy is the use of biomimetic coatings, [8, 11, 25–28] composed of biological molecules covalently bound to the surface of the neural probes. Of particular interest are coatings derived from L1, a cell adhesion molecule found on neuronal axons and synapses. [29, 30] L1 elicits unique reactions from multiple CNS cell types, both in vitro and in vivo. Astrocytes and fibroblasts attachment is inhibited on L1-modified surfaces in vitro, [9, 31] and microglia cells that contact L1-modified probes do not proceed to encapsulate the device as would otherwise be observed. [32] Neurons can interact directly with L1, resulting in elevated neurite outgrowth on L1-modified substrate in vitro. [9, 11, 31] In vivo, it has been observed that the tissue surrounding L1-modified neural probes has elevated axonal density and greater counts of neuronal cell bodies. [10, 11, 25] These interactions between the host tissue and L1 create a seamless electrode-neuron interface and result in greatly improved chronic recording performance. [10].

However, these biomimetic coatings have some limitations which have stalled their widespread adoption in the neural interface field. First, the bioactivity of a coating is tied to both the number of bound biomolecules and their stability. Further, immobilized protein tends to denature in non-physiological conditions, necessitating the application of these coatings to occur on-site and immediately prior to implantation. These limitations hinder the dissemination of the biomimetic coating technology and thus motivated us towards developing a method to increase the stability of the bound protein.

We have previously investigated the use of nano-textured surfaces on the stability and bioactivity of surface-immobilized L1. [9, 11] Silica nanoparticles functionalized with thiol groups (TNP) on the surface were immobilized to silicon substrates prior to the deposition of the L1 coating. It should be noted that our TNP coating primarily targets the silicon dioxide surface, not the iridium electrode site. While some nanoparticles are bound to the metal sites, we did not observe increased impedance that would translate to compromised recording quality. We have previously applied TNP+L1 coating to neural probes with platinum electrode sites without observing compromised function. However, other electrode coatings may be affected more strongly by the TNP+L1 coating process, especially materials which are less inert or have microstructure/topography critical to its performance and such electrode materials should be carefully examined prior to use.

We discovered that the nanotextured TNP surface was able to bind more proteins and demonstrated higher bioactivity than smooth surfaces. [9, 11] Further, the protein coating on the TNP surface was stable for four weeks in buffered saline at 37°C, while the coating on the smooth surface lost over 60% of the bound protein and over 50% of bioactivity. The increased stability in wet conditions may be due to the higher number of bonds per protein on the nano-topographical surface compared to a smooth surface in addition to the higher amount of bound protein. These findings were promising but required coated devices to be stored in physiological saline which greatly complicates packaging and shipping. The aqueous condition also increases the chance of hydrolytic degradation of the protein molecules and their bonding with the underlying substrate. In order to enable the transportation of biomolecule-modified devices, dry storage is preferred over wet because it makes packaging and delivery of delicate microdevices easier. Attempts to stabilize surface immobilized biomolecules in dry conditions include freeze-drying, encasing the protein in polyethyleneglycol gel, [33] and using nanostructured minerals to stabilize the conformation of the protein. [34]

We speculate that the porous structure of the TNP surface may trap moisture even after drying, and as such the immobilized proteins retain their native state and function. Residue moisture has been proposed to be the mechanism by which microarrays maintain stability in dry storage [49]. Alternatively, the 3D topography may offer more anchor points to maintain the bound proteins in the correct folding, thereby maintaining their bioactivity. We have begun investigating how 3D topography features at different length scale contribute to protein binding in a different study [50]. In this work, we aim to characterize the stability of the TNP+L1 coating after drying with the goal of facilitating the dissemination of biomimetic coatings. In vitro, we examined the bioactivity of dried TNP+L1 surfaces for eight weeks with primary neuron cell cultures. We then implanted single shank 16 channel neural probes that have been coated with TNP+L1, then dried and stored for 3 days, and compared the recording performance to those that have been freshly coated with TNP+L1, to confirm that 3 day dried coating works just as well as the freshly prepared. 3 days will be enough time for the delivery of the coated probes to most laboratories worldwide. To further examine the shelf life of the TNP+L1 coating and demonstrate the versatility of the coating, we tested a second set of neural probes (32 channels) which were coated, dried, and stored for 28 days before implantation. 28 days would be sufficient for most users to receive and store the coated devices and implant them according to their own timeline. The electrophysiological performance of these devices was measured for 12-16 weeks, and subsequent postmortem histology and explant analysis were used to investigate the interface between the electrode device and the host tissues.

2.0. Methods

2.1. Chemicals and Materials

All reagents were purchased from Sigma Aldrich unless otherwise specified. Neural probes were purchased from NeuroNexus (Ann Arbor, MI). The first probe design (A1x16-3mm-100-703-CM16) consists of 16 electrode sites of 30μm diameter (area of 703 μm²), separated by 100μm from center to center. The second set of probe design (A1x32 poly2-3mm-100-703-CM32) consists of 32 electrode sites with 30μm diameter (area of 703 μm²), staggered into two lines down the length of a single shank. Electrodes were spaced 100μm from the center of each other. Both sets of electrodes are comparable in electrode site area and distance between sites, the 32-channel probes are slightly wider. Both electrode devices used iridium electrode sites. Glass coverslips were purchased from Electron Microscopy Sciences.

2.2. Nanoparticle Fabrication

Nanoparticles were fabricated as previously described [11] from tetraethyl orthosilicate (TEOS) and mercaptopropyl trimethoxysilane (MTS). In brief, a solution consisting of 5ml ethanol, 36ml water, and 5ml triethanolamine was stirred for 30 minutes at 60°C. 1.5ml of TEOS was added dropwise over the course of 5 minutes under vigorous stirring. The reaction commenced for 5 minutes at which point 250μl of MTS was added. A second bolus of 100μL MTS was added after 1 hour, after which the reaction continued for 1 more hour and was cooled to room temperature. Following formation, the nanoparticles were collected by centrifuge and washed with water and absolute ethanol, then stored dry. Before use, thiolated nanoparticles (TNP) were resuspended in an aqueous solution of tris-carboxyethylphosphine (TCEP, 1mg ml⁻¹) to reduce any disulfide bonds and then centrifuged and washed with water.

2.3. Sample Preparation

Modified substrates consisted of glass (coverslips) or silicon dioxide (probes). Samples were first cleaned with acetone and isopropyl alcohol then the surface was activated using O₂ plasma for 5 minutes, then submerged in 2.5% silane solution in absolute ethanol for one hour. The silane was either MTS for control samples or aminopropyltriethoxysilane (ATS) for nanoparticle immobilized samples. Samples destined to have immobilized nanoparticles were then submerged in an aqueous solution of the hetero-bifunctional crosslinker gamma-maleimidobutyryl-oxysuccinimide (GMBS, 2mg ml⁻¹) for 30 minutes. Following which the samples were washed with water and exposed to a suspension of TNP in water (10mg ml⁻¹) for one hour with gentle agitation of the solution performed every 15 minutes to prevent clumping of the particles.

2.4. Protein isolation and deposition

L1 was isolated from post-natal rat pups using an affinity column as previously described [29]. In brief, brains were extracted from euthanized rat pups, homogenized, and then spun in a centrifuge against a sucrose gradient. The layer containing the cell membranes and membrane proteins was then extracted and stored in cholamidopropyl dimethylammonio-1-propanesulfonate (CHAPS) buffer. The membrane in CHAPS was then passed over an antibody affinity column, which was then washed with buffered saline to remove unwanted compounds. The protein was then eluted from the column with saline adjusted to pH 11 with diethylamine. Prior to deposition, samples were treated with GMBS for 30 minutes then washed with water. Samples were then exposed to L1 solution (2μg ml⁻¹ in saline) for one hour followed by washing with saline. Selected samples were then dried by removing the saline with an aspirator until completely dry. Dried samples were stored in closed containers at room temperature (with around 45-50% environmental humidity) in the absence of light.

2.5. Neuron Culture

Primary neurons were isolated from E18 rat fetuses. The mother rat was euthanized with CO₂ and the embryos were removed and submerged in ice-cold Hank’s buffered salt solution (HBSS, Gibco). The pup brains were removed and the cortex was isolated and washed with HBSS. The cortex was then digested with trypsin solution (0.05wt%, Gibco) for 15 minutes at 37°C. The trypsin solution was then removed, and the cortex was broken up by gently trituration. Finally, cells were isolated, resuspended in neurobasal media (Gibco) supplemented with 2% B27, 1% GlutaMax, and 1% Penicillin/Streptomycin, and counted prior to plating at a density of 25,000 cells cm⁻². Cells were grown for 48 hours at 37°C in 5% CO₂ and then fixed with 4% paraformaldehyde. Cells were stained with β(III)-Tubulin and DAPI as previously described.[9] Images were taken on a Leica DMI 4000b and neurite outgrowth was quantified with the Neurite Tracer ImageJ plugin.[35]

2.6. Surgeries and Recording

All animal experiments were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Male mice (8-week-old, C57-BL6) were purchased from Jackson Labs. In the first set of experiments, 10 mice were implanted with 16-channel NeuroNexus electrode devices coated with either fresh TNP or fresh TNP+L1 (five mice per condition). Due to headcap failure, two mice from each group had to be sacrificed at different time points before 12 weeks. For the 3-day dried TNP+L1 group, five mice were implanted with 16-channel probes, and all survived the 12 weeks. Recording data from unmodified probes from 6 animals were obtained from a previous experiment[10] and plotted as a reference. A second set of experiments were performed with 32-channel NeuroNexus probes where 10 animals were divided into two groups of five, receiving either the control implants without modification or TNP+L1 modified implants which had been stored dry for 28 days prior to implantation. All animals in the second set were recorded for 16 weeks.

Animals were anesthetized under 1.5% isoflurane in O₂ at a flow rate of 1L min⁻¹ and anesthesia was maintained with 1.3% isoflurane. The scalp hair was removed, and the scalp was sterilized with an iodine solution and 70% ethanol. The scalp was then removed, and a hole was drilled in the animal skull above the visual cortex. The hole was covered with a sterile gelatin sponge to prevent the brain from drying, and three bone screws (Fine Science Tools) were secured to the skull. The gelatin sponge was then removed, and the electrode device was inserted into the brain with a stereotactic manipulator. The implant location was 1 mm anterior to Lambda and 1.5 mm lateral to midline. The electrode device was implanted normal to the brain surface, to a depth of 1700um below the brain surface. A small amount of medical-grade silicone epoxy (Kwik-Sil) was then used to fill the hole around the probe. The reference and ground wires were wrapped around the bone screws, then the probe was secured to the skull with UV-curable dental acrylic. A small amount of triple antibiotic ointment was applied to the injury site following surgery to minimize the risk of infection. Following the operation, the animal received analgesic Ketofen injections for 3 days (5mg kg⁻¹).

Impedance measurements were taken with a potentiostat (Metrohm PGSAT128N) prior to the coating modification and after implantation on a weekly schedule. Impedance measurements lower than 20 kΩ and higher than 2000 kΩ were excluded. These out-of-range values are most likely due to a transient imperfect connection between the head-mounted Omnetic connector and the adapters that passively connect the channels to the potentiostat on the day of the measurement. These values did not always appear on the same electrode sites or demonstrate a time-dependent pattern. The number of channels excluded based on this criterion ranges from 2 to 5 out of 16 channels, and 8 to 12 out of 32 channels. Electrophysiology was performed under a visual stimulation paradigm. Animals were administered 2% isoflurane for induction, after which isoflurane was reduced to 1% to maintain a lightly anesthetized state. Anesthetized animals (at 1%) were then subjected to visual stimulation consisting of a series of moving bars on a monitor displayed to the eye contralateral to the implant. Visual stimuli were performed in a dark room to minimize outside interference.

Electrophysiology data were analyzed with a custom MATLAB script using data collected during visual stimulation. Visually evoked data was filtered between 300 and 3000Hz and a threshold of 3.5 standard deviations of the data stream was applied to isolate potential spiking activity. Next, the threshold crossing events were centered in a 1.2 ms waveform snippet which was subsequently removed from the spike voltage stream data, to calculate mean peak-to-peak noise. To further isolate single neuronal units, the waveform snippets were subjected to principal component analysis, and the first three principal components were used to separate the waveforms into individual clusters by K means. The single unit (SU) signal quality was defined as signal-to-noise ratio (SNR) and was calculated as the maximum peak-to-peak amplitude of the mean waveform of the cluster, divided by the standard deviation of the noise, which is defined by the data stream after subtracting the spiking events. Only candidate units with detectable spikes of SNR greater than 4 were analyzed. Only channels with an SNR above 4 were manually selected by two human operators, who independently inspected a combination of the waveform shape, the raw spike data stream, and the inter-spike interval with 50 ms bins. Units that surpassed these criteria were deemed single units. Several metrics were used to evaluate the recording quality. % Yield was determined as the percentage of channels that recorded at least one unit out of the total channel count. The number of units per channel was defined as the average number of units recorded per channel across all electrode sites. The average SNR was determined by averaging the SNR for each time point across all electrode sites, with channels without a sortable single unit assigned an SNR = 0, while using the maximum SNR for channels detecting more than one single unit.

Following the 12- or 16-week experiment, animals were anesthetized with Ketamine (90mg kg⁻¹) and Xylazine (9mg kg⁻¹) and perfused with 100ml of PBS followed by 100ml 4% paraformaldehyde (PFA) in PBS. The bottom of the animal skull was removed and post-fixed in 4% PFA for 2 hours. The brain was then removed from the skull, dehydrated in 15% and 30% sucrose, and frozen prior to sectioning. Sections were taken on a Leica CM 1950 cryostat with a thickness of 10μm. Sections were stored frozen until staining. Sections chosen for staining spanned a depth of approximately 750-1500 microns deep in the visual cortex.

Staining was performed by rehydrating the slices in citrate buffer and then blocking them with 10% goat serum, following which the brain slices were treated with 0.1% Triton-x for 45 minutes. Staining of the brains was performed in groups consisting of NeuN (Millipore mouse 1:250), NF200 (Abcam rabbit 1:500), Iba-1 (Millipore rabbit 1:500), GFAP (DAKO rabbit 1:500), and lectin (Vector Laboratories, tomato, 1:250). Images were taken on a confocal microscope (Olympus Fluoview 1000) and analyzed with a custom MATLAB script. 25 bins, each covering 10μm were created concentrically around the probe implant, and the intensity of the stain or number of cells labeled per bin was quantified [51]. To control for the variability, quantified intensities, and counts were scaled to control regions at the corners of the images. Background intensity was calculated from the corners of the image (20% of the total image area) by removing pixels greater than 1 standard deviation (STD) above the mean and calculating the mean and variance of the remaining pixels. All pixels greater than 1STD above the mean background intensity were used for analysis.

2.7. Scanning Electron Microscopy Imaging

Scanning electron microscopy (SEM) was used to visualize the surface of the explanted 32-channel probes after 16 weeks in vivo. The animal skull, with the probe still attached, was carefully trimmed away so that the only remaining portion was the top of the skull directly surrounding the implant and dental cement cap. The skulls were then washed with DI water and dehydrated with an ethanol gradient of 30%, 50%, 70%, 90%, and 100% ethanol, allowing for at least 24 hours at each step. The skulls were then desiccated over CaCl₂ for 1 week prior to sputter coating and imaging (Zeiss Sigma500 VP).

2.8. Statistics

All statistical analyses were performed in GraphPad PRISM 9.4.1. Cell culture experiments are the combined results of 3 separate culture preparations, each culture with 3 samples and 3 images per sample and normalized to the TNP group. Samples were compared across groups with ANOVA followed by Tukey’s post hoc. Unless stated otherwise, a mixed model for repeated measures (MMRM) statistical analysis was performed on the longitudinal electrophysiological data followed by Sidak’s multiple comparison posthoc test to compare metrics between conditions over time. Interaction effects between time and condition are reported for the significance values. Data is plotted as mean ± standard error unless stated otherwise.

3.0. Results

The covalent immobilization of nanoparticles onto silicon and glass substrates (Figure S2 and S3) has previously demonstrated an increase in protein immobilization and bioactivity.[9, 11] We do observe some TNP coverage on the iridium electrodes sites (Figure S3), but most of the TNP coverage is on the silicon shanks since the covalent linking chemistry is primarily targeting SiO₂ substrate. Additionally, the proteins bound to these substrates were stable for at least four weeks under aqueous conditions at 37°C. However, transportation of probes in liquid could prove difficult and it is desirable to develop a coating that is stable under ambient conditions after drying. To test the bioactivity of the TNP+L1 coating after drying and storage, samples were prepared up to eight weeks prior to culture and dried thoroughly with an aspirator vacuum, then stored at room temperature away from light for 1, 2, 4, 6, and 8 weeks. For comparison, a second set of samples was prepared by coating L1 directly on a smooth substrate without the nanoparticle base layer (referred to as smooth L1).

3.1. L1 bound to TNP retains its bioactivity when stored dry and at room temperature

Primary neurons were plated over the dried TNP+L1 samples along with the controls for 48 hours, followed by fixation and staining with beta(III)-tubulin to visualize the neurite outgrowth. Neurites were visible on all TNP-modified substrates but were notably longer and denser on TNP+L1-modified samples (Figure 1A–D). The lowest outgrowth was observed on the smooth sample without L1, while freshly prepared smooth L1 samples performed far better (p<0.0001; ANOVA with Tukey’s post hoc). However, after only three days we observed a significant reduction in neurite outgrowth on smooth L1 samples (p<0.0001; ANOVA with Tukey’s post hoc). In contrast, all TNP+L1 modified samples performed better than the TNP modified sample (Figure 1E, p<0.001; ANOVA with Tukey’s post hoc) and more importantly, dry storage for 1 to 8 weeks did not affect the neurite outgrowth, suggesting maintained bioactivity for at least 8 weeks in dry air. Even though there is a slight upward trend in the neurite length for the dried TNP+L1 groups from 1-week to 8-week of drying, this change is not statistically significant. While the neuron cell culture was performed on the same day, the samples had to be prepared on different days for the different duration of aging, which could contribute to the non-significant variability in the neurite length between groups.

Figure 1. L1 bound to TNP retains its bioactivity when stored dry and at room temperature.

(A-D) Representative images of neurons grown on TNP and TNP+L1 substrates after aging. (E) Quantified neurite outgrowth represented as % normalized to TNP group on glass coverslips. Groups included only silane modification (smooth), L1 without topographical modification prepared the day of culture and 3 days prior to culture (Fresh and 3-day Smooth L1), and TNP or TNP+L1 modifications, prepared freshly or dried for 1, 2,4,6 and 8 weeks . ****p<0.0001, ANOVA with Tukey’s post hoc. n=9 for smooth, fresh smooth L1, and 3-day smooth L1, n=27 for the TNP modified samples.

3.2. TNP+L1 modified Probes Enhanced Neural Recording Even After Three Days of Dry Storage

After confirming that the bioactivity of the TNP+L1 surface, as assessed by L1’s neurite extension promoting property, was not lost after drying and storage, we examined how the dried coatings would perform in vivo. In the first step of the experiment, mice were implanted with 16-channel NeuroNexus electrode arrays modified with one of three conditions: TNP only, TNP+L1 prepared the day of surgery, and TNP+L1 dried and stored at ambient temperature for three days prior to implantation. The recording and impedance performance were compared to control data previously published by our lab. [36] Following implantation, neural activity, and impedance were recorded every week for 12 weeks, followed by sacrificing the animal and performing post-mortem histology. We compared the impedance and electrophysiologic recording performance over the course of the implantation period, in addition to performing endpoint analysis at the final week of recording.

Initial impedances were measured directly prior to implantation, and account for potential changes in the electrode impedance between the pristine and modified probes (Figure 2A). Despite the surface modification, the pre-implant impedance for the TNP and TNP+L1 groups did not drastically increase as compared to the control. On average, after the pristine electrodes were modified with TNP+L1, impedance did increase from 181 kΩ to 257 kΩ (data not shown) but it was not a significant change (p= 0.49; Welch’s t-test). The slight impedance increase is likely due to binding of the nanoparticles around the electrode site (Figure S3B). This binding can alter the interactions with the electrolyte and decrease the exposed surface area of the electrodes. However, once implanted, impedance measurements taken just prior to electrophysiology show a characteristic increase over the first weeks of implantation but level off after approximately 2-4 weeks. Impedances of the TNP-only group were significantly higher than other conditions between weeks 5-12 (p<0.0001; mixed-effects model for repeated measures (MMRM)). While both fresh and 3-day dried TNP+L1 showed lower and more stable impedances than the TNP modification alone, 3-day dried TNP+L1 exhibited significantly lower impedance than the fresh TNP+L1 (p<0.0001; MMRM). Control electrode impedances showed a less stable trajectory than the two TNP+L1 groups, with significantly lower impedances at the 12-week timepoint.

Figure 2. TNP+L1-modified probes enhanced neural recording even after three days of dry storage.

(A) Changes in impedance magnitude at 1kHz immediately prior to implantation, and weekly prior to each recording session from per-modification values. Occasionally some channels have abnormally high (>2000kΩ) or low impedance (<20 kΩ) due to poor connection and these were removed from the average (B) Single unit yield (number of channels with single unit per probe/16 channels) was evaluated for each probe. One-way ANOVA was performed with Tukey’s multiple comparison test. (C) the noise floor and (D) Signal to noise ratio (SNR) for probes throughout the 12-week recording period. Electrophysiological recordings for the control group are based on a prior published study [10] and were not included in the statistical comparison. 5 animals were implanted for each group (TNP, TNP+L1, 3D TNP+L1, total of 15 mice). Due to headcap failure, only 3 mice survived for each of the TNP and TNP+L1 groups, and all 5 survived in the 3D TNP+L1 for the entire 12 weeks. “ns”=not significant, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 mixed-model for repeated measures with Sidak’s multiple comparison test. Data presented as mean ± standard error.

Although impedance measurements can allow for investigating the nature of the electrode-tissue interface, it is often poorly correlated with recording performance. To examine the effect of the TNP+L1 coatings on electrophysiology, we performed weekly recordings for 12 weeks. Recording yield was defined as the number of electrode sites or channels that recorded single units, divided by the total number of channels for each array (Figure 2B). Recording yield was highest for the fresh and 3-day dried TNP+L1 groups, maintained between 60-80% for the entire 12-week experiment, which was significantly higher than the TNP-modified probes when compared across the entire recording period (p<0.01 for TNP compared to TNP + L1, p <0.001 for TNP compared to 3-day dried TNP + L1; one-way ANOVA). The TNP-modified probes also outperformed the control, producing higher recording yields than the control probes when examined across all weeks.

We then examined the noise and signal-to-noise ratio (SNR) (Figure 2C and 2D, respectively) of the probes. The noise floor was significantly lower for the fresh and 3-day dried TNP+L1 groups compared to TNP and control throughout the recording period (p<0.0001 for TNP compared to TNP+L1, p<0.0001 for TNP compared to 3-day dried; MMRM). The noise floor reported in the 3-day dried TNP+L1 was significantly lower than both fresh TNP alone and fresh TNP+L1 (p<0.0001 for 3-day dried compared to TNP+L1; MMRM). Lastly, it was observed that the SNR was highest for both TNP+L1 groups (Figure 2D), both performing significantly higher than the TNP alone condition (p<0.0001 for TNP compared to 3-day dried and p<0.0001 for TNP compared to TNP+L1; MMRM).

3.3. Elevated Neuronal and Axonal Density and Decreased Gliosis Were Observed in Tissues Surrounding both fresh and 3 d dried TNP+L1 than TNP coated probes

Postmortem histology was performed to examine the extent and the nature of inflammation and gauge neuronal tissue health around the probe following implantation. We first visualized the neuronal soma with NeuN while co-labeling with Iba-1, which is a marker for the microglia (Figure 3A). In all conditions, there was a decrease in the relative NeuN density directly adjacent to the probe, but there was a quicker return to baseline in the fresh and 3-day dried TNP+L1 samples (Figure 3B). We quantified the number of neurons present in the first 50μm from the implant and observed that there was a significant elevation in the number of neuron cell bodies adjacent to the freshly prepared and 3-day dried TNP+L1 modified probes compared to the TNP-modified probes (Figure 3C).

Figure 3. Elevated neuron density and decrease microglial encapsulation were observed in tissues surrounding the TNP+L1-modified probes.

Endpoint histology showing NeuN (neuron soma), Iba-1 (microglia), and Dapi (cell nuclei) staining of cortical tissues after 12 weeks of recording. (A) Representative images of cortical tissues stained with NeuN (green), Iba-1 (red), and Dapi (blue) for tissues implanted with TNP (top), TNP+L1 (middle), and TNP+L1 which was dried for three days prior to implantation (bottom). (B) Relative NeuN counts as a function of distance from the probe surface. (C) Neurons were quantified over the first 50μm from the probe surface, and significant differences were observed between the number of cells adjacent to the protein-modified samples compared to the TNP alone (ANOVA with Tukey’s Post Hoc). (D) Relative Iba-1 intensity as a function of distance from the probe surface. Significant differences were observed between the TNP group and the other two groups over the first four 10μm bins. *p<0.05 two-way ANOVA with Tukey’s Post Hoc. Scale bars are 100μm.

Microglia are often the first responders to implantation injury and are expected to have the greatest intensity in their staining closest to the implant surface. We observed that there was the greatest expression of the Iba-1 staining in the closest binned distances for all conditions, with a significant increase in the Iba-1 staining intensity around the TNP-modified group compared to the TNP+L1 groups over the first 40μm from the probe surface (Figure 3D). Following the closest binned distances, tissue around all the probes exhibited a rapid decline in Iba-1 intensity, approaching the baseline values by approximately 100μm from the probe surface.

Reactive astrocytes were visualized with Glial Fibrillary Acidic Protein (GFAP) while neuronal axons were stained with NF200 (Figure 4A). GFAP quantification revealed elevated astrocytic activity around the TNP-modifted probes relative to the TNP+L1 probes (Figure 4B). Significant differences in GFAP staining were observed between the TNP+L1 and TNP groups between 20 and 90μm from the electrode, and an overall statistical significance was found between groups across all time points (p<0.01; two-way ANOVA). Additionally, there was an overall significant decrease in GFAP expression in 3-day dried TNP+L1 electrodes compared to the TNP probes across all binned distances (p<0.05; ANOVA). Axons were minimally affected by the probe insertion in all experimental groups (Figure 4C). There appears to be an elevated expression of NF200 directly adjacent to the TNP+L1 modified probes compared to TNP modified probes, but this only resulted in statistically significant differences for the first 10μm bin (p<0.05; two-way ANOVA).

Figure 4. Elevated axon density and decreased astrocytic encapsulation were observed in tissues surrounding the TNP+L1 modified probe, GFAP (activated astrocytes) and NF200 (neuronal axons) staining of cortical tissues following 12 weeks of recording.

(A) Representative images of NF200 (green) and GFAP (red) for tissues implanted with TNP (top), TNP+L1 (middle), and TNP+L1 which were dried for three days prior to implantation (bottom). (B) GFAP intensity vs distance from the probe surface. GFAP intensity is highest nearer to the probe and drops towards the baseline in all cases. (C) NF200 intensity vs distance from the electrode surface. *p<0.05 **p<0.01, green asterisk signifies TNP+L1 significantly differs from other groups, scale bars are 100μm.

3.4. TNP+L1 Modified-Probes Stored Dry for 28 Days Outperformed Uncoated Control Electrodes in Neural Recording Performance

Next, we expanded the study by 1) extending the dry storage time prior to implantation to 28 days, and 2) increasing the electrode site density and channel counts. 32-channel NeuroNexus neural probes were coated with TNP+L1 and stored for 28 days (28D TNP+L1) before implantation into the mouse visual cortex. Electrochemical impedance characterization and electrophysiological recordings were performed weekly for 16 weeks and compared to unmodified probes.

We observed a temporary increase in the device impedances following insertion (Figure 5A), as is expected after implanting into the tissue. Initially, impedances were lower for the protein-modified device, but a gradual decrease in the impedance measured on the control device resulted in the impedance magnitudes nearing each other after eight weeks. Overall, as compared to the control devices, the coated group exhibited stable impedance measurements over the course of 16 weeks (p<0.0001; MMRM).

Figure 5. TNP+L1-modified probes stored dry for 28 days outperformed control probes.

(A) Impedances measured on the 28D TNP+L1 probes were overall significantly lower than the control probes (p<0.0001; MMRM). Channels with abnormally low (<20kΩ) or high impedance (>2000kΩ) were removed. (B) The number of average single units recorded per channel was significantly higher for the 28D TNP+L1 probes compared to the control for the overall duration of the implant (p<0.001) (C) The noise, defined as the data stream after spiking events are subtracted, followed similar trends for both the 28D TNP+L1 probes and the control probes with an overall significant difference for the 16-week duration of the implant (p<0.0001). (D) The overall signal-to-noise ratio (SNR) was overall significantly elevated for the 28D TNP+L1 probes compared to the control probes (p<0.0001). Overall SNR includes channels with units assigned an SNR of 0 based on manual waveform sorting. N = 5 male mice were implanted for each condition (N = 10 total). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, mixed-model for repeated measures with Sidak’s multiple comparison tests was used for (A – D). Data presented as mean ± standard error.

In terms of neural recording, the unmodified probes exhibit a decrease in the average number of single units recorded per channel following the first week of recording and a more gradual decline thereafter (Figure 5B). This trend was not observed on the 28D TNP+L1 probes, which were able to maintain an average number of 0.7 units per channel for the duration of the implant. 28D TNP+L1 probes recorded a significantly higher number of units per channel than the control over the course of the experiment (p<0.001; MMRM). The average unit number per channel across time was 0.65 for TNP+L1, and 0.47 for control, resulting in a 32% increase by the coating. The observed trend of unit yield increase (non-significant) during the last 4-5 weeks of implantation is due to some electrode sites having imperfect connection during a single recording session but recovering for future sessions. While both probes exhibited similar trends for the recording noise floor (Figure 5C), stabilizing at approximately 7μV after five weeks post-implant, the overall difference in noise floor levels between control and coated conditions was significant for the implant duration (p<0.0001; MMRM). The strength of the recorded units, the background noise, and the number of sites actively recording units all contribute to the calculated SNR values. We observed that the SNR was consistently higher for the TNP+L1 probes over the course of the experiment with an overall significant difference between groups over the course of the experiment (Figure 5D, p<0.0001; MMRM). The average SNR across time was 4.65 for TNP+L1, and 3.1 for control, resulting in a 40% increase by the coating. At the 16^th week time-point, the TNP+L1 group not only had more channels recording at least one sortable unit but the sorted units also had higher peak-to-peak amplitude, as compared to the control group (Figure S5A and S5B). Despite having similar levels of noise floor to the control group, the higher amplitude waveforms recorded from TNP+L1 coated probes led to higher overall SNR values, as reported in Figure 5D. Previous work demonstrated an increase in neuronal cell bodies around L1 coated implants and direct attachment of neuronal processes. [9,11] This increase in the number and proximity of viable neurons around the TNP+L1 probe likely contributes to higher number and larger amplitude of units recorded as compared to the uncoated controls.

3.5. Tissue Surrounding 28D TNP+L1 Probes had Elevated Neuron Density and Decreased Gliosis than Uncoated Controls

We then performed immunohistochemical studies to compare the tissue responses between the 28D TNP+L1 coated and uncoated 32-channel implants. The NeuN and Iba-1 staining are shown in Figure 6A and 6C. The number of neuron cell bodies is expected to correlate to increased recording yield, and we observed a significant increase in the number of cell bodies adjacent to the probe at the 25μm binned distance (p<0.05) while also demonstrating an overall significant trend (p<0.001; two-way ANOVA, Figure 6B)). Notably, the observed trend in neuron cell body counts is consistent between the 32Ch and 16Ch probes. There was a significant decrease in Iba-1 staining from the experimental group in comparison to the uncoated control groups both at the 20μm bin (p<0.05) and overall (p<0.001; two-way ANOVA) (Figure 6D).

Figure 6. The tissue surrounding 28D TNP+L1 probes had elevated neuron density and decreased microglial encapsulation.

(A) Representative images of the tissues adjacent to control and 28D TNP+L1 probes stained for NeuN (green) for Control (left) and 28D TNP+L1 (right). Blue is a nuclear stain. The number of neuronal cell bodies was quantified in 25μm bins from the probe surface (B), and significant increases in the number of neurons adjacent to the probes were observed at the 25μm bin and overall. (C) Representative images of the tissues adjacent to control and 28D TNP+L1 probes stained for Iba-1 (white) for Control (left) and 28D TNP+L1 (right). (D) The Iba-1 staining intensity was elevated overall for the control probes compared to the 28D TNP+L1 probes, as well as at the 20μm binned distance from the probe surface. *p<0.05, ***p<0.001, two-way ANOVA with Tukey’s post hoc.

GFAP and NF200 staining of the 32 Ch devices are shown in Figure 7A and C). As with the 16Ch probes, we observed that NF200 staining was diminished adjacent to the control 32Ch probes but was maintained around the 28D TNP+L1 modified 32Ch probes (Figure 7B). Although there were no significant differences in the axon density around either probes at any specific distance from the probe, there was an overall significant decrease in axon density around the control probes that was not present on the 28D TNP+L1 modified probes (p<0.001, two-way ANOVA). GFAP was significantly elevated around the control compared to the 28D TNP+L1 probes at binned distances between 10 and 60μm from the implant surface (Figure 7D).

Figure 7. Tissue surrounding 28D TNP+L1 probes had elevated axon density and decreased astrocytic encapsulation.

(A) Representative images of the tissues adjacent to control and 28D TNP+L1 probes stained for NF200 for Control (left) and 28D TNP+L1 (right). (B) Quantification of the NF200 staining as a function of distance from the probe surface. Significant differences were observed between the control and 28D TNP+L1 groups when compared across all distances (p<0.001 two-way ANOVA). (C) Representative images of the tissues adjacent to control and 28D TNP+L1 probes stained for GFAP for Control (left) and 28D TNP+L1 (right). (D) Quantification of the GFAP staining as a function of distance from the probe surface. Significant differences were observed between the control and 28D TNP+L1 groups when compared across all distances in addition to significant differences at 10μm-60μm from the implant (two way ANOVA). **p<0.01 ***p<0.001 ****p<0.0001, two-way ANOVA with Tukey’s post hoc.

3.6. Nanoparticle Topography and Neurites Were Observed on Explanted 28D TNP+L1 Probes After 16 Weeks In Vivo

Following the endpoint of recording and removal of the brains for postmortem histology, the underside of the skulls which still contained the attached 32Ch probe was prepared for scanning electron microscopy (SEM) (Figure 8 A–C). SEM images of the control probe showed a generally smooth surface morphology after 16 weeks of implantation (Figure 8B). On the other hand, the high magnification images of the 28D TNP+L1 surface revealed a nanotopographic structure, likely due to the immobilized TNP (Figure 8D). This structure was visible only on the 28D TNP+L1 probes and was notably absent on the control probes (more representative images are shown in Figure S1). There was also a substantial degree of biological material present on the 28D TNP+L1 device (Figure 8G, H), with what appears to be neurite projections originating from a cellular mass located directly on top of an electrode site. Increasing the magnification on the highlighted region, we can examine the morphology of these projections on the surface of the device which are straight in shape and 10s of μm long. Control probes lacked these long process-like biological features on their surfaces and are instead covered with thicker and more homogeneous matters that resemble inflammatory cell encapsulation and extracellular matrices (Figure 8E, F).

Figure 8. Nanoparticle topography and neurites were observed on explanted 28D TNP+L1 probes after 16 weeks In Vivo.

SEM images of the Control (A,B) and 28D TNP+L1 probe (C,D) in the skull (A,C) and zoomed in to show the morphology of the surface (B,D) are shown. Images of the control probe (E) with a higher magnification of the red highlighted section in (F) are shown. Control probes also showed biological encapsulation on the probes after explantation (E,F), however, the long neurite-like features were not prominent. Higher magnification images of the 28D TNP+L1 probes show apparent neurites extending from potential cell bodies (G,H).

4.0. Discussion

Biomimetic coatings have demonstrated a great potential of controlling undesired tissue response, promoting neuron health and density, and improving recording quality and longevity. Despite their promise, challenges in maintaining coating stability in use and storage have prevented their widespread adoption. In this work, we examined the effects of TNP pre-modification on stabilizing the L1 protein in dry storage. We further investigated the electrophysiological performance of the probes with TNP+L1 coating that was dried in air for 3 days to verify that the coating maintained its function. Finally, we expanded our scope to include storage of dried TNP+L1 probes for 28 days prior to use and a different probe design, allowing for even greater flexibility for the distribution and utilization of these modified devices.

The L1 coating was first subjected to dry-storage conditions to examine the long-term stability of the protein-modified surface. Our results indicated that there was rapid loss of L1’s neurite-promoting properties when the protein is bound to an unmodified smooth surface, while immobilization of L1 to the nano-textured surface maintained the stability of the coating for 8 weeks in ambient conditions.[34] Other studies have observed similar increases in protein stability when bound to nanotopographical surfaces. [15, 37] The elevated stability of L1 bound to the TNP surface is hypothesized to be due to multiple factors. Firstly, the nanotopographical coating may be able to support a controlled microenvironment within the porous structure, preventing the protein from being completely dehydrated. Secondly, there may be a scaffolding effect in play, where the protein is bound to multiple locations within the porous structure. This scaffolding effect could increase the strength of the protein-surface interactions, thereby protecting the protein from being cleaved while maintaining its folding conformation for optimal bioactivity.

Based on the findings in this study and our previous work, the L1 coatings encourage neuronal attachment to the surface, promote neuronal density, and inhibit inflammatory glial encapsulation. The effect of the coating not only benefits neural recording but is also expected to translate to improved stimulation performance. With a healthier and more seamless neural tissue and implant integration, the threshold stimulation current is expected to be lower and steadier than implants with a more severe foreign body response (FBR). Thus, a more precise and stable stimulation paradigm can be utilized with lower power consumption and higher safety [48]. The effects of L1 coating on neural stimulation may be examined in depth in future studies. In addition, we plan to modify flexible electrode devices with the L1 coating, combining the benefits of a soft flexible device with a biologically active coating. Immobilization of L1 coatings on TNP-modified surfaces and their collective sustained stability is particularly interesting in its potential applications outside of neural engineering. Many devices that employ biologically derived modifications such as artificial lungs[4, 5] and biosensors[3, 38, 39] may benefit from enhanced stability and bioactivity.

Following in vitro experiments (Figure 1), we transitioned into a mouse in vivo model to study how the enhanced coating can affect electrophysiological performance over 12 weeks. Prior to implantation, TNP-modified probe showed the lowest post-modification impedance, even lower than the unmodified control. This observation is likely due to the additional oxygen plasma treatment, which is the first step for the surface modification and can clean the surface. The impedance increase following L1 immobilization may be due to the adsorption of non-conductive L1 protein on the surface of the electrode sites. The 3-day dried L1 group has higher impedance than the fresh TNP+L1 group likely because the impedance was measured before water fully wet the nanopores at the surface of the 3-day dried surface. The difference in the in vitro impedance between groups does not necessarily indicate the expected trend once implanted. The TNP-modified probe, without L1, performed remarkably well considering there was no bioactive modification to the surface other than textural. We did observe significantly increased in vivo impedance values measured from the TNP-modified probes as compared to the control or L1-modified devices. This impedance increase is likely from heightened glial encapsulation for the TNP-only group, as supported by immunohistology (Figure 3). However, increased impedance does not compromise the recording performance of the TNP-coated devices.

The weekly single-unit yield from TNP-modified probes was on par or even greater than unmodified control probes, indicating that the modification did not induce any adverse reactions in the surrounding tissue that will affect neuronal health and function. These findings are further corroborated by the postmortem histology results. Although we observed many characteristic signs of inflammation, including elevated astrocytic encapsulation and decreased neuronal density adjacent to the probes, these responses were not out of line with previously reported literature examining unmodified implants.[10] Changes to the topography of substrates have previously been shown to change cellular responses and differentiation in vitro.[40–45] Moreover, topographical modifications have been utilized to provide modest reductions in inflammation following the implantation of probes in vivo[46]. This is believed to be partly due to the natural topography of biological environments,[47] where cells are interacting with a web of extracellular matrices, surrounding cells, and vessels with varying degrees of roughness.

The TNP+L1 modification has previously been examined for up to four weeks of recording. [11] The probes modified with TNP+L1 had lower impedances, higher recording yield, greater neuronal density, and lower expression of inflammatory cells than the unmodified or TNP-modified probes. In this longer-term study, we again observe decreased impedances for the TNP+L1 probes relative to the TNP probes, potentially indicating decreased scarring and improved device-tissue integration. The recording channel yield and SNR of the TNP+L1 modified probes appear higher than controls in our previous report, [10] and outperformed the TNP-modified probes with statistical significance. Drying and storing the probe for 3 days prior to implantation did not negatively affect the recording performance of the device, and the 3-day dried probes produced the effect in reducing impedances and elevating single unit yields and SNR, as observed on the fresh TNP+L1 devices.

Postmortem histology further elaborated on the tissue reactions to the implant. Specifically, certain histological makers were more affected by the TNP+ L1 coatings. The degree of astrocytic encapsulation was reduced in tissue adjacent to the TNP+L1 probes, as compared to the TNP probes. L1 is known to interact directly with neurons. These interactions likely encourage neurons and their axons to bind directly to the probe and enhance the number of cells proximal to the surface. A second potential interaction may be a survival cue for these cells, preventing the cells from undergoing apoptosis following probe implantation or during the chronic inflammation phase.

Although 3 days of dry storage is expected to be sufficient for off-site production and shipping of the device, this creates tight windows between device arrival and use. We examined the shelf-life of L1-coated substrates stored in −20°C for up to 10 weeks and observed no loss of bioactive effect on in vitro neuronal cell cultures (Figure S4). These observations confirm the expected effect of sub-zero temperatures maintaining the stability of biological molecules. In addition, to prevent heat-induced damage, coated devices would be packaged on ice for overnight shipment. In an effort to enable flexibility in the schedule of the supplier and end user, we further extended our examination of the electrode performance after drying to 28 days. The electrophysiology performance of the TNP+L1 probes after drying and storing for 28 days was compared to control probes. Despite being prepared an entire month prior to use and dry-stored at room temperature, the TNP+L1 probes consistently outperformed the control and produced a higher average number of single units recorded per channel (32% increase) and signal-to-noise ratio (40% increase). In addition, we observed lower and more steady impedance values for the 28D TNP+L1 probes compared to the control probes, similar to the fresh and 3-day dried probes. The postmortem histological analysis indicated that the tissues surrounding the 28D TNP+L1 probes were still healthier than the tissue surrounding the control. Most notably, there was an increased number of neuronal axons and cell bodies and decreased astrocytic encapsulation. We have previously reported that the TNP+L1 modification was capable of increasing neurite extension and outgrowth while limiting astrocytic spreading in vitro[9] and in vivo after one and four weeks.[11] The findings reported here serve to demonstrate that the interactions between the TNP+L1 surface and both neurite and astrocytes are maintained for at least 16 weeks in vivo and even after the modified surface has been stored for 28 days prior to implantation. Future work may also explore how the extent of drying the TNP+L1 coatings can affect the results and investigate ways to control this effect.

Following the endpoint of recording, the probes were explanted to examine the underlying topography. Previously, we have confirmed the presence of the TNP coating on explanted probes after four weeks of implantation indicating strong adhesion of the TNP to the surface [11]. In this study, the explant analysis of the 28D TNP+L1 probes provides evidence for the adhesion of the TNP coating even after 16 weeks in vivo. In addition, there appear to be neuron-like structures intimately interacting with the TNP+L1 surface. These projections are long, straight, and thin, and highly resemble neurites that appear on L1-modified substrates in vitro. These features were not observed on the control probes, potentially due to the preferential adherence of different cell types to the electrode surface. Further analysis will be required to confirm the nature of these biological features, but the results suggested a continued intimate interaction between the electrode and the host neurons throughout the duration of the study.

Another important consideration prior to the dissemination of TNP+L1-modified neural probes is the sterilization process. One of the major challenges with utilizing biological coatings for in vivo implantation is their sensitivity to most common sterilization procedures. For instance, gamma irradiation, steam sterilization, or ethylene oxide sterilization can all deactivate the surface-bound protein through various mechanisms. Hence, the best approach is to sterilize the uncoated electrode device first and subsequently coat the device under sterile conditions. Once coated, the device can be packaged in sterile packaging and prepared for shipping to the end-user.

While this coating strategy was mainly tested for animal research models, with additional validations and regulatory approvals, the TNP+L1 coating could be used for clinical recording electrodes in the future. Taking clinical brain-computer interfaces (BCI) as an example, utilizing coated electrodes with a 32% increase in unit yield and 40% increase in SNR would enable recording of higher quantity and quality neuronal spiking information which can then be used for better decoding and improved BCI performance (more degree of freedom control, accuracy, and speed etc.). This is especially useful, since signal quality deterioration is a common phenomenon observed in chronic microelectrode implants in for BCI applications [52].

5.0. Conclusions

The use of L1 as a surface modification for neural electrode devices can greatly increase the short and long-term recording performance of the devices while maintaining the health of the neural tissues adjacent to the implant. However, protein-based coatings are notoriously fragile which limits the widespread use of the technology. We have investigated an enhancement to the conventional L1 coating which employs a nanotopographical base layer derived from the immobilization of silica nanoparticles to the electrode surface. The resulting TNP+L1 surface was superior to smooth L1 substrates in vitro and was even able to stabilize the protein such that it could be dried and stored prior to use in vivo. We have demonstrated that the TNP+L1 surface can be prepared and stored under ambient conditions for 28 days without losing functionality in improving neural recording quality and tissue health. Such results provide the feasibility to deliver the biomimetic coating technology to researchers world-wide. The nanoparticle base coating technology can also be broadly applicable to applications where stability of surface bound biomolecules is of utmost importance, including biosensors, tissue engineering, biotechnology, and medical devices.

Data Availability

Data will be made available on request.