Neuroadhesive protein coating improves the chronic performance of neuroelectronics in mouse brain

Abstract

Intracortical microelectrodes are being developed to both record and stimulate neurons to understand brain circuitry or restore lost functions. However, the success of these probes is hampered partly due to the inflammatory host tissue responses to implants. To minimize the foreign body reactions, L1, a brain derived neuronal specific cell adhesion molecule, has been covalently bound to the neural electrode array surface. Here we evaluated the chronic recording performance of L1-coated silicon based laminar neural electrode arrays implanted into V1m cortex of mice. The L1 coating enhanced the overall visually evoked single-unit (SU) yield and SU amplitude, as well as signal-to-noise-ratio (SNR) in the mouse brain compared to the uncoated arrays across the 0–1500 μm depth. The improvement in recording is most dramatic in the hippocampus region, where the control group showed severe recording yield decrease after one week, while the L1 implants maintained a high SU yield throughout the 16 weeks. Immunohistological analysis revealed significant increases of axonal and neuronal density along with significantly lowered microglia activation around the L1 probe after 16 weeks. These results collectively confirm the effectiveness of L1 based biomimetic coating on minimizing inflammatory tissue response and improving neural recording quality and longevity. Improving chronic recording will benefit the brain-computer interface technologies and neuroscience studies involving chronic tracking of neural activities.

1. Introduction

Neural interfaces are used to record and/or stimulate from neural tissue to provide insight into processes such as movement control and to restore function for patient suffering from neurological disorders (Collinger et al. 2013; Donoghue 2008; Hochberg et al. 2006). However, the full potential of these devices remains unrealized due to inconsistent performance in chronic applications (Barrese et al. 2013; Chestek et al. 2011; Kozai et al. 2015a; Williams et al. 1999). There are different biotic and abiotic factors responsible for the poor recording performance (Prasad et al. 2011). It is believed that brain tissue reaction plays a central role (Barrese et al. 2013) which is triggered by insertion of the electrode device followed by a cascade of inflammatory immune responses, ultimately leading to device encapsulation (McConnell et al. 2009). More specifically, probe penetration damages blood-brain barrier (BBB) and results in activation of microglia and infiltration of macrophage (Kozai et al. 2015b; Ravikumar et al. 2014; Wellman and Kozai 2017). Further, BBB breakdown leads to an influx of plasma proteins onto the implant surface, which triggers the reactivity of inflammatory cells (Wellman and Kozai 2017). These cells produce reactive oxygen species (ROS) around the implants and release inflammatory cytokines which can further contribute to BBB disruption (Abdul-Muneer et al. 2015; Falcone et al. 2019; Kozai et al. 2015b; Potter-Baker et al. 2014). It is also revealed that probe insertion cause pathological calcium activation in the nearby neurons as well as cell body and process damages (Eles et al. 2018), while increased mechanical strain further compressed neurons in the vicinity of the implant site (Du et al. 2017). Over time, neuronal loss and neurodegeneration, demyelination, as well as astroglial encapsulation have been extensively observed, which could lead to decreased recording quality (Biran et al. 2005; Golabchi et al. 2018; McConnell et al. 2009).

To establish a chronically stable neural interface, several approaches have been investigated which include the design of subcellular size probes (Kozai et al. 2012a; Seymour and Kipke 2007), and softer electrode materials (Blau et al. 2011; Du et al. 2017; Kolarcik et al. 2015b; Lacour et al. 2010; Luan et al. 2017) to mimic the brain tissue mechanics. Recently, devices that are both small and flexible have shown excellent tissue integration and good chronic performance (Wei et al. 2018; Yang et al. 2019). Ultra-small and flexible devices face the challenges of increased electrode impedance, decreased mechanical and electrical durability, and increased handling difficulties in implantation and removal. Alternatively, surface coatings may be applied improve the integration of the more conventional stiffer and larger devices. The development of numerous coating strategies by the use of carbon nanotubes (CNTs) (Alba et al. 2015; Kolarcik et al. 2015a; Kozai et al. 2016a), hydrogels (Lu et al. 2009; Zhou et al. 2012), and conducting polymers (CPs) (Cui et al. 2003; Du et al. 2015; Kozai et al. 2016a; Ludwig et al. 2011) into the electrode surface to improve not only electrode material properties like charge transfer and impedance, but also enhancing tissue integration and minimizing mechanical mismatch (Wellman et al. 2017). Moreover, pursuing surface chemistry approaches through incorporating anti-inflammatory drugs (Luo et al. 2011; Wadhwa et al. 2006; Zhong and Bellamkonda 2007a, b), and immobilizing bioactive molecules (Azemi et al. 2011; Eles et al. 2017; He et al. 2006; Kolarcik et al. 2012; Kozai et al. 2012a) to modulate cell behavior are other promising methods to improve electrode-tissue interface (Woeppel et al. 2018). We have shown previously covalently attachment of cell adhesion molecule (L1) onto the silicon surfaces in vitro, increased neuronal attachment and neurite length and decreased astrocyte attachment (Azemi et al. 2008). Further in vivo evaluation of L1 coated probes implanted in the rat cortex showed improved neuronal and axonal density near the implants with decreased glial cell activation up to 8 weeks after implantation (Azemi et al. 2011). Moreover, a two-photon microscopy study demonstrated significantly reduced microglia process attachment and spreading on L1 coated implants than the uncoated control at the first hours post implantation suggesting an immediate interaction between the protein and inflammatory cells (Eles et al. 2017). L1 coated 4×4 Utah arrays have been implanted in rat visual cortex and an acute benefit of the coating on recording yield has been demonstrated (Cody et al. 2018). In the present study, we evaluated the chronic recording performance of L1 coated single shank silicon based microelectrode arrays implanted in the visual cortex for 16 weeks using a visually evoked recording model. Immunohistological evaluation of the tissue response associated with L1 coated probe compared to the control was quantified at 16 weeks after implantation. Compared to uncoated probes, there was significantly improved recording quality and longevity and significantly improved axonal regeneration and lower microglia activation around the L1 probes. This is the first direct demonstration of the chronic benefit of biomimetic protein coating on neural recording.

2. Material and Methods

This study evaluates the effect of L1 coating on chronic neural recording of intracortical microelectrode arrays in C57BJ/6 wild type (WT) mice. A visually evoked recording model was used (Brainard 1997; Cornelissen et al. 2002; Kleiner et al. 2007), and recording performance was investigated through single-unit (SU) yield, SU signal to noise ratio, SU and multi-unit (MU) amplitude, noise floor and impedance using previously published methods (Kozai et al. 2012a; Kozai et al. 2014b). After 16 weeks, postmortem immunohistochemistry was performed to demonstrate the molecular, cellular, and vascular responses around the implantation site. 12 male mice (C57BL/6J; 9 weeks; 20 to 25 g; The Jackson Laboratory, Bar Harbor, ME) were kept in the temperature-controlled animal facility center (DLAR) in a ventilated rack at the University of Pittsburgh with a12:12-h light: dark cycle and with ad libitum access to food and water. All procedures were aligned with the guideline of Institutional Animal Care and Use Committee of the University of Pittsburgh.

2.1. Neural probes and L1 protein immobilization

Functional 3 mm long and 15 μm thick single-shank planar silicon probes having 16-circular electrode site with 30 μm diameter and 100 μm site-spacing (NeuroNexus, Ann Arbor, MI, A1X16–3MM-100–703-CM16; Fig S. 1.) were used in this study. Each probe was submerged in 0.1 N NaOH for 1 h followed by generous rinsing with water followed by ethanol. The shank was then submerged in an ethanol solution containing 2.5% glycidyloxypropyl trimethoxysilane and 10 mM acetic acid for 1 hour. After which the electrode was washed with ethanol and sterile PBS. The shank was then submerged in L1 solution (100 μg ml⁻¹ in PBS) for 1 hour and rinsed just prior to implantation. To test the stability of L1 coating over time, L1 was immobilized onto glass substrates which contain the same surface chemistry as the insulation layer of the silicon probes. The L1 coated substrates were incubated at 37 °C for up to 28 days prior to plating primary rat cortical neurons harvested from E18 rat embryos. After 2 days of culture in neurobasal media (Neurobasal +2% B27 supplement +1% GlutaMAX + 1% Pen/Strep), cells were fixed in 4% PFA and stained for β(III)-Tubulin, a neuronal process marker. The outgrowth was calculated by summing the neurite lengths per image area, with N=27 for all groups, then normalizing to the 0-day control. It was determined that L1 maintained most of its activity for 7 days, during which the acute wound healing response takes place. A substantial portion (~25%) of the L1 activity was maintained out to 4 weeks. The data related to the stability of L1 over time is shown in Fig S. 2.

2.2. Surgical Implant Procedure

All animal surgeries were performed as previously described (Kozai et al. 2015a; Kozai et al. 2014b). Animals were allocated randomly into 2 groups; control (N=6, implanted with uncoated electrodes) and L1 (N=6, implanted with L1-coated protein). Mice were anesthetized by using an Isoflurane Vaporizer (Patterson Veterinary Inc.; isoflurane inflow was 2% for induction phase and maintained at 1.5% during surgery) that provides a mixture of isoflurane and oxygen. A warm heating pad (HTP 1500, Adroit Medical Systems, Loudon, TN) was set to 37 °C and placed underneath the anesthetized mouse to maintain body temperature. The mouse head was fixed in the stereotaxic frame (Kopf Instruments, Tujunga, CA) and the skull was exposed. Very small hole was drilled in the skull with a surgical drill (0.007 drill bit, Fine Science Tools, Inc., Foster City, CA) over the 1 mm anterior to Lambda and 1.5 mm lateral to midline (Fig. S. 3 A). Continuously, sterilized saline was applied for cutting heat dissipation in high-speed drilling of the bone. Three sterilized bone screws (Stainless Steel; shaft diameter: 1.17 mm, length: 4.7 mm; Fine Science Tools, Inc., Foster City, CA) were screwed bilaterally over the primary motor cortex (as the grounding electrode) and over the contralateral visual cortex (as the reference electrode) for anchoring dental cement to bone. A probe was implanted in the left visual cortex using a stereotaxic manipulator until the top edge of the last electrode site is below the brain surface. After filling the craniotomy with Kwik-Cast Sealant (World Precision Instruments, Sarasota, FL), the reference and grounding wires were connected to the bone screws and dental cement (Pentron Clinical, Orange CA) was cured with a dental curing light to make a head-cap (Fig. S. 3 B). After surgery, animals placed on an electric heating blanket under a warming light to wake up and received intraperitoneal (i.p.) injection of 5 mg/kg ketofen (100 mg/ml, Henry Schein) up to three days after.

2.3. Neurophysiological Recording

Recording of spontaneous and visually evoked SU, MU and local field potential (LFP) responses was performed each week as extensively described (Kozai et al. 2015a; Kozai et al. 2012a; Kozai et al. 2014b) under 1% isoflurane anesthesia. Briefly, electrophysiological recording was conducted from the animal inside of a Faraday cage in a dark condition. Visual stimuli were presented through the MATLAB-based Psychophysics toolbox (Brainard 1997; Kleiner et al. 2007; Pelli 1997) through a 24” LCD (V243H, Acer. Xizhi, New Taipei City, Taiwan) located outside of cage and placed 20 cm from the eye contralateral to the implant. Solid black and white bar gratings were presented drifting in a perpendicular direction and synchronized with the recording system (RX7, Tucker-Davis Technologies, Alachua FL) at 24.414 Hz. The raw data stream was filtered to produce spike (0.3 to 5 kHz) data streams. Further, the spike data stream was pre-processed through previously published methods (Ludwig et al. 2009; Wagenaar and Potter 2002) and individual single-units were initially identified by using a fixed negative threshold value of 3.5 standard deviation below the mean of the spike data and were centered in a 1.2 msec waveform snippet that was then removed from the spike voltage stream data to calculate mean peak-to-peak noise (Kozai et al. 2014b; Ludwig et al. 2009). Offline spike sorting was performed through a custom MATLAB script modified from previously mentioned methods (Kozai et al. 2015a). SU signal-to-noise ratio (SU SNR) was calculated from verified units as peak-to-peak amplitude of the mean waveform of the isolated cluster divided by two times the standard deviation of the voltage data after removing 1.2 msec threshold crossing waveforms. A channel without a confirmed SU is considered to have an SNAR and amplitude of 0 and 0 V, respectively. Only candidate units with detectable spikes (SNR >2) were analyzed. Channels with SNR between 2 and 3 were manually selected by examining the combination of waveform shape, auto-correlogram, peak threshold crossing offset, and peri-stimulus time histogram (PSTH) with 50 ms bins and candidate units with SNR greater than 3 were manually confirmed by looking at the waveform shape (Kozai et al. 2014b). A channel is considered to have evoked MU activity if MU spike count (from total threshold crossing) is significantly different in a 550 msec bin 50 msec after stimulus onset (during ‘ON’ state) from MU spike count in a 550 msec bin 50 msec before stimulus onset (during ‘OFF’ state) as determined using a paired t-test. Additionally, we quantified the MU Signal-to-Noise Firing Rate Ratio (SNFRR) as the average firing rate of the ‘ON’ state minus the average firing rate of the ‘OFF’ state to the visual trigger divided by the average standard deviation of the ‘ON’ and ‘OFF’ state both (Kozai et al. 2015a) to evaluate the activity and excitability of the neurons. Noise Amplitude is calculated as mean peak-to-peak amplitude of the spike filtered voltage signal following removal of threshold crossing 1.2 msec waveform snippets.

2.4. Electrochemical Impedance Spectroscopy (EIS)

EIS was used to characterize the electrode properties and glial scar encapsulation. Data was collected before implantation and after every recording session under anesthesia (1.5% isoflurane) through connecting implanted electrode to an Autolab potentiostat using a 16 channel multiplexer. A voltage of 10 mV RMS sine wave from 10 to 32,000 Hz was applied to measure EIS, using individual electrode sites as the working electrodes and stainless-steel screw (19010–00, Fine Science Tools, Inc., Foster City, CA) as the counter electrode.

2.5. Immunohistochemistry

According to University of Pittsburgh IACUC approved methods, mice were sacrificed at the end (16 weeks). Each animal was deeply anesthetized using 80–100 mg/kg ketamine, 5–10 mg/kg xylazine cocktail. Once mice were unresponsive to tail/toe pinches, animals were transcardially perfused using phosphate buffered saline (PBS) flush at <80 mmHg followed by 4% paraformaldehyde (PFA) at <80 mmHg. Mice were decapitated and the skulls were removed to post-fix the brain in a 4% PFA at 4°C for 4–6 h. After removing the probes, brains were soaked in a 15% sucrose (Sigma-Aldrich Corp., St. Louis, Missouri) bath at 4°C overnight followed by a 30% sucrose solution for 24 h. Brains were then carefully frozen in a 2:1 20% sucrose in PBS:optimal cutting temperature compound (Tissue—Plus O.C.T. Compound, Fisher HealthCare, Houston, TX) blocking media blend with dry ice. Frozen tissue was then horizontally sectioned into 25 μm thick sections normal to the tract of the probes using a cryostat (Leica CM1950, Buffalo Grove, IL).

Cortical sections of implanted and non-implanted hemisphere were mounted on the same slide for comparison and staining for each antibody combination was performed at the same time to minimize variability. Antibodies to visualize neurons (NeuN, 1:250, MAB377 Millipore), neurofilament (NF200, 1:250, MAB5256 Millipore), microglia (Iba-1, 1:500, NC9288364, Fisher), astrocytes (GFAP, 1:500, Z033401 Dako), blood vessels (tomato-plant lectin, 1:250, B1175 Vector Labs), and blood-brain barrier injury (immunoglobulin G (IgG), 1:16, Alexa Flour 647-conjugated AffiniPure Fab Fragment goat anti-mouse IgG 115–607-003 Jackson ImmunoResearch Laboratories, Inc.) were used. These antibodies were used to evaluate the L1 coating effect in glial activation, probe encapsulation, and the number of viable neurons.

Tissue sections were rehydrated in 1 × PBS for 2×5 min. The tissues were then incubated in 0.01 M sodium citrate buffer for 30 min at 60 ° C. Then, a peroxidase block (PBS with 10% v/v methanol and 3% v/v hydrogen peroxide) was performed for 20 min at room temperature (RT) on a table shaker. Next, tissue sections were incubated in carrier solution (1 × PBS, 5% normal goat serum, 0.1% Triton X-100) for 30 min at RT. Lastly, the tissue sections were blocked with Alexa Flour 647-conjugated AffiniPure Fab Fragment goat anti-mouse IgG (IgG, 1:16, 115–607-003 Jackson ImmunoResearch Laboratories, Inc.) or Fab fragment only (1:13, 115–007-003, JacksonImmunoResearch Laboratories, Inc.) for 2 hours then rinsed 6 times each 4 minutes. Following blocking, sections incubated in a primary antibody solution consisting of carrier solution and antibodies overnight (12–18 hours) at RT were then washed with 1 × PBS for 3 × 5 min and incubated in carrier solution and secondary antibodies (1:500, Alexa Flour 488 goat-anti mouse, Invitrogen, and 1:500 Alexa Flour 568 goat-anti rabbit, Invitrogen, 1:500, DyLight 649 Streptavidin, Vector Labs, 1:500 Alexa Flour 568 goat-anti mouse, Invitrogen, 1:500 Alexa Flour 488 goat-anti rabbit, Invitrogen, 1:500 Alexa Flour 633 goat-anti chicken, Invitrogen) for 2 hours at RT. Then sections were rinsed with PBS for 3 × 5 min and exposed to Hoechst (1:1000, 33342 Invitrogen) for 10 min and washed in PBS for 3 × 5 min before being cover slipped with Fluoromount-G (Southern Biotech, Associate Birmingham, AL). The reagents used in this experiment are shown in supplementary section, Table 1.

Table 1.

List of reagents used for this experiment.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-NeuN, Mouse	Millipore Sigma	MAB377
Anti-Neurofilament 200 kDa (NF- 200)	Millipore	MAB5256
Wako Chemicals USA Rabbit Ant- Iba1, Rabbit	Fisher	NC9288364
Polyclonal Rabbit Anti-Glial Fibrillary Acidic Protein (GFAP)	Dako	Z033401
DyLight 594 labeled Lycopersicon Esculentum (Tomato) Lectin	Vector Labs	B1175
IgG	Jackson ImmunoResearch Laboratories, Inc.	115–607-003
Fab Fragment Only	Jackson ImmunoResearch Laboratories, Inc.	115–007-003
Alexa Fluor® 488 goat anti-mouse IgG	Invitrogen	A-11001
Alexa Fluor 488 Goat Anti-Rabbit IgG	Invitrogen	A-11034
Alexa Fluor® 568 goat anti-rabbit IgG	Invitrogen	A-11032
Alexa Flour 568 goat-anti mouse	Invitrogen	A-11004
Alexa Fluor® 633 goat anti-chicken IgG	Invitrogen	A-21103
Hoechst 33342, Trihydrochloride, Trihydrate	Invitrogen	H3570
Fluoromount-G	Southern Biotech	0100–01
DyLight 649 Streptavidin	Vector	SA-5649
Chemicals
Isothesia (isoflurane) solution	Covetrus North America	029405
Experimental Models: Organism/Strain
Mouse: C57BL/6J, male	The Jackson Laboratory	000664
Other
Recording electrode (1 shank, 16 channels	NeuroNexus	A1×x16–3mm-100- 703-CM16
Data acquisition system	Tucker Davis Technologies	RX7
Isoflurane Vaporizer	Patterson Veterinary Inc.
PAD, heat therapy SM 13” × 18” for HTP-1500 soft-temp localized heat therapy system by Adroit Medical	Medline	ADRST017H
Stereotaxic Frame	Kopf Instruments
Burrs for Micro Drill	Fine Science Tools, Inc.	19008–07
Self-T apping Bone Screws - 4.8mm/1.19mm Shaft Diameter	Fine Science Tools, Inc.	327433–1.7G
Kwik-Cast Sealant	World Precision Instruments	KWIK-CAST
Triple Antibiotic Ointment	Henry Schein	061148
Absorption Triangles	Fine Science Tools, Inc	18105–03
Restoratives, Composite, Flowable - Natural Elegance Intro Kit	Henry Schein	037752
Ketofen Ketoprofen 100mg	Henry Schein	005487
Fisher Healthcare™ Tissue-Plus™	Fisher Scientific	23–730-571
O.C.T. Compound

2.6. Quantitative tissue analysis

Sections were imaged using confocal fluorescent microscopy to evaluate the cellular reactions associated with the implanted electrodes. Images were acquired using an Olympus Fluoview FV-1000 Confocal Microscope (Olympus America, Center Valley, PA) at the Center for Biologic Imaging at the University of Pittsburgh. For each antibody, images were taken using the same laser power, exposure time, and detector settings to decrease variability. Images were centered on the implant site and multi-channel images were acquired simultaneously. For Iba-1, GFAP, NF200, and IgG images were analyzed using a pixel-based radial image intensity analysis, as previously described (Kozai et al. 2014a). Sections in the two ranges of 300–800 μm, and 1000–1500 μm depth from the brain surface were compared with non-implanted sections. For each image, the center of insertion site was chosen and by using the MATLAB script, masks of concentric rings every 20 μm for 240 μm were generated. Then, MATLAB scrip calculated and normalized the average gray scale intensity for all pixels above the threshold of the background noise intensity in each 20 μm bin. Finally, intensities were averaged for each group, and then bar graphs for intensity-based radial analysis of fluorescent markers were plotted by mean ± standard error as a function of distance.

For NeuN, cells were manually counted in each bin and the cell density (cell count per tissue area) was calculated. As previously described, data were averaged for each group, and the bar graphs for density-based radial analysis of cells (the mean and standard error) were plotted as a function of distance to the implantation site.

2.7. Statistics

General linear model repeated measure ANOVA and Tukey’s post hoc tests were conducted in SPSS to compare the effect of time and condition on recording or tissue characteristics among L1 group compared to the control at the same time points. P<0.05 is deemed statistically significant.

3. Results

3.1. Effects of L1 coating on impedance spectroscopy

Fig. S. 4 shows the impedance of electrodes as a function of frequency before and after coating with L1 in PBS. The control group impedance at 1 kHz (318 kΩ ± 40 kΩ) was not different from L1-coated (369 kΩ ± 154 kΩ; before modification) groups. The magnitude and the phase of Bode plots did not change significantly after coating with L1 in all frequencies (N=5; due to the Autolab technical issue, the pre-coating impedance data from one of the probes was not recorded correctly and was omitted from the averaging). The in vivo impedance performance is displayed in Fig. 1 with the average 1 kHz impedance magnitude of L1 coated electrodes compared to the control group over 16 weeks of implantation (N=6 for all impedance measures from week 0 (the day of surgery) to week 16). As shown in Fig. 1 A the control group impedance was not different from L1-coated groups in the brain on the day of surgery (control, 909 kΩ ± 90 kΩ; L1-coated, 1041 kΩ ± 119 kΩ) and both groups increased in impedance and reached a plateau (control, 929 kΩ ± 72 kΩ; and L1-coated, 435 kΩ ± 19 kΩ) at 4 weeks and stayed indistinguishable for the remainder of the study. Besides, the noise amplitude was almost the same during the whole study in both groups (Fig. 1 B).

Fig. 1. L1 coating improves chronic recording performance.

Evoked recording performance of L1 coated electrode (red square) compared to the control (black circle) over 16 weeks of implantation. A) Average 1 kHz impedance. B) Voltage amplitude of the noise floor. C) The percentage of recording sites detected SU. D) Average SNR value for all recording sites. E) Average SNR of recording sites able to detect SU (active channels). F) Evoked MU SNFRR. G) Signal strength of MU. H) Average peak-to-peak voltage amplitude of SU for all recording sites. I) Average peak-to-peak voltage amplitude of active channels. All data presented as mean ± SE.* indicates p<0.05, and **** presents p<0.0001 for all. (N=6 for both groups).

3.2. Neurophysiological recording performance

3.2.1. Depth independent chronic recording performance

The performance of electrodes was investigated over time by averaging metrics of total number of recording sites without depth consideration. Visually evoked results are shown in Fig. 1 C-I. The percentage of recording sites detecting SU signal are plotted as SU yield (Fig. 1 C). Yield for the control started at 46% ± 8%, decreased to 23% ± 5% (control) on the following week of surgery, and then increased back to 43 ± 5% between week 1 and 3, and then continued to decline until stabilizing at 30% after week 5. L1 coated group started with a yield of 55% ± 5% (no statistical difference to control), decreased slightly to 48% ± 5% after 1 week, increased to 62% ± 5% by week 3, and maintained a higher yield for most of the weeks between week 5 and 16 than control. In general, results directly illustrated the elevated recording performance of L1-treated implants compared to the uncoated group from week 3 until the rest of study and the significant difference was observed over 16 weeks of study (p<0.05). Moreover, the quality of SU signal is reported as SNR (Fig. 1 D&E). As shown in Fig. 1 D and G, SNR and signal amplitude (channels with no detectable SUs were considered with 0 value) of the L1 group was higher than control group at all time points, while the active SNR (channels with no detectable SU were excluded from final analysis) of L1 group was higher on the day of surgery and week 1 (Fig. 1 E). Although, evoked MU SNFRR, MU amplitude, and active signal amplitude were almost the same in both groups (Fig. 1 F, G, and I), the significant difference was observed in SNR, signal amplitude (Fig. 1 H), and active SNR between two groups over 16 weeks of study (p<0.05).

3.2.2. Depth dependent chronic recording performance analysis

All implants were aligned to the average layer IV depth of the corresponding animal between week 2 and week 16. Layer IV was determined using current source density (CSD) after visual stimulation (Kozai et al. 2015a). The SU yield of L1 and control group as a function of depth and time is shown in Fig. 2 A. An immediate observation is that layer II/III, IV, and V in both groups have higher yield than the other layers, with L1 group having the higher yield (above 80%) throughout the 16 week of study (Fig. 2 A), than control (60%) (Fig. 2 B). Besides, at the day of surgery and 1 week after, almost at all depth, recording sites were able to pick up single unit activity in both groups, however, after 1 week, the SU yield of control group at CA1 area dropped (Fig. 2 B), while L1 coated group presented sustained high SU yield in this region over 16 weeks (Fig. 2 A). In addition, the average SU yield at shallow depth (300–800 μm) and deep depth (1000–1500 μm) between L1 and control group are shown in Fig. 2 C&D. In both figures, L1 group showed significantly higher SU yield over 16 weeks. The amplitude of L1 and control group as a function of time and depth showed the same behavior as SU yields as is shown in supporting data (Fig. S. 5). Furthermore, the SNR of L1 group around Layer IV was higher than control acutely (Fig. S. 6).

Fig. 2. Evoked SU yield as a function of depth and time.

The SU yield depth dependent of L1 group (A) and the control (B) over 16 weeks of recording are plotted. Note: there are greater yields around layer IV and CA1 of hippocampus in L1 group, while, it is only observed around layer 4 in control group. L1 group showed higher SU yield in shallow depth (C; 300–800 μm) and deeper region (D; 1000–1500μm) than control. All data presented as mean ± SE. * indicates p<0.05, and **** p<0.0001 for all.

3.3. Histology

3.3.1. Effect of L1 coating on neuron viability and axonal immunoreactivity

Neuronal density and axonal damage were visualized using NeuN+ (neuronal nuclei) (Fig. 3) and NF-200 (axonal neurofilament) (Fig. 4), respectively. Representative images of viable neurons around the implanted electrode in control and L1-coated group at 300–800 μm (Fig. 3 A) and 1000–1500 μm (Fig. 3 B) depth from the brain surface at week 16 are shown in Fig. 3. The density of neurons was quantified using a MATLAB script (Kozai et al. 2014a) and were plotted as a function of distance to the implantation site at 300–800 μm (Fig. 3 C) and 1000–1500 μm (Fig. 3 D) depth from the brain surface. Although, the number of neurons immediately around the implant site in control and L1 treated group had no significant difference (P>0.05), there was significant higher density of neuronal cells in L1 group compared to the control across all distance from the implantation site at both depth (Fig. 3 C&D; P<0.0001), which is consistent with stronger electrophysiological outcome. Moreover, NF-200 is a marker for mature axons. Representative images for axonal staining at the 16-week time point for Control and L1 probes at the depth of at 300–800 μm (Fig. 4 A) and 1000–1500 μm (Fig. 4 B) depth from the brain surface are shown in Fig. 4. At both depths, the NF-200+ axons were increased around L1 coated probe compared to the control. Then, the average pixel intensity of NF-200 stain was calculated using MATLAB script and normalized to the intensity of the background and plotted as a distance from electrode-tissue interface at two different depths (Fig. 4. C&D). At 300–800 μm, the first bin of 0–20 μm NF-200 staining intensity increase as much as 117% was observed around L1 modified probes compared to the control ones (Fig. 4. C) and the intensity associate with further distance (40 and 60 μm), was increased to almost 49% when compared to that around the Control implants (Fig. 4. C). The differences in NF-200 intensity between the two groups were significant at all three bins (p<0.05). Moreover, at deeper tissue section, L1 probe yields 45% increase in NF-200 staining intensity at the first 20 μm bin than control which was statistically significant (p<0.05) (Fig. 4. D). In general, the L1 implants showed a significant increase in NF-200 intensity compared to the control across the distances from the electrode-tissue interface.

Fig. 3. Immunohistochemically evaluation of neuronal viability at 16 weeks.

Representative images of neuronal density from tissue acquired at 300–800 μm (A) and 1000–1500 μm (B) depth from the brain surface around implanted electrode (dashed squares) in the control and L1 coated group. Neuronal cell counts were measured by automated analysis of NeuN staining with respect to the distance from the implantation hole at 300–800 μm (C) and 1000–1500 μm (D) depth. (Control N=6 animal, N=8 sample (300–800 μm) and N=10 sample (1000–1500 μm); L1-coated N=6 animal, N=14 samples (300–800 μm) and N=16 sample (1000–1500 μm) staining. Results were binned 20 μm from the electrode-tissue interface until 240mm away, with 20 μm bin size. All data presented as mean ± SE. **** presents p<0.0001 for all.

Fig. 4. Immunohistochemically evaluation of axons at 16 weeks.

Representative images of axonal damage from tissue acquired at 300–800 μm (A) and 1000–1500 μm (B) depth from the brain surface around implanted electrode (dashed squares) in the control and L1 coated group. Normalized intensity value for NF-200 expression with respect to the distance from the implantation hole at 300–800 μm (C) and 1000–1500 μm (D) depth. (Control N=6 animal, N=8 sample (300–800 μm) and N=13 sample (1000–1500 μm); L1-coated N=6 animal, N=8 samples (300–800 μm) and N=10 sample (1000–1500 μm) staining. Results were binned 20 μm from the electrode-tissue interface until 240mm away, with 20 μm bin size. All data presented as mean ± SE. * indicates p<0.05, *** shows p<0.0005 and **** presents p<0.0001 for all.

3.3.2. Effect of L1 coating on BBB permeability

The integrity of the BBB is examined through Immunoglobulin G (IgG) staining. Representative images of BBB leakage at the 16-week time point for Control and L1 probes at 300–800 μm (Fig. 5 A) and 1000–1500 μm (Fig. 5 B) depth from the brain surface are shown in Fig. 5. All experimental groups showed the same amount of BBB leakage at both depths (Fig. 5 A&B). Moreover, the average normalized intensity of IgG stain as a function of distance from the electrode-tissue interface from the tissues that acquired at 300–800 μm (Fig. 5 C), and 1000–1500 μm (Fig. 5 D) depth was not different between L1 and control probes.

Fig. 5. Immunohistochemically evaluation of blood brain permeability at 16 weeks.

Representative images of blood brain permeability (IgG expression) from tissue acquired at 300–800 μm (A) and 1000–1500 μm (B) depth from the brain surface around implanted electrode (dashed squares) in the control and L1 coated group. Normalized intensity value for IgG expression with respect to the distance from the implantation hole at 300–800 μm (C) and 1000–1500 μm (D) depth. (Control N=6 animal, N=8 sample (300–800 μm) and N=7 sample (1000–1500 μm); L1-coated N=6 animal, N=12 samples (300–800 μm) and N=8 sample (1000–1500 μm) staining. Results were binned 20 μm from the electrode-tissue interface until 240 mm away, with 20 μm bin size. All data presented as mean ± SE.

3.3.3. Effect of L1 coating on Gliosis

The activity of microglia was studies in Fig. 6. The intensity of ionized calcium binding adaptor molecule (Iba-1), a selective stain for both resting and activated microglia and macrophages (Golabchi et al. 2018), was evaluated surrounding the 16-week implant at both depth of 300–800 μm and 1000–1500 μm. Iba-1 within the 60 μm from the interface in L1 modified probe showed lower intensity compared to the control (Fig. 6 A&B). Then, quantification of Iba-1 expression at the interface was normalized to background expression levels of Iba-1. Significantly less Iba-1 interfacial expression was noted within 60 μm surrounding the L1 implant than control at 300–800 μm depth. This intensity reduction in L1 probe compared to control was 51%, 89%, and 88% at 20, 40, and 60 μm away from the interface, respectively (Fig. 6. C; p<0.05). Furthermore, the expression of Iba-1 around L1 implant was significantly lower than control (by 81% and 75%) at 20 and 40 μm away from the interface in the deeper tissue sections collected, respectively (Fig. 6. D; p<0.05). Also, the expression of glial fibrillary acidic protein (GFAP), a marker for astrocytes with higher expression in activated astrocytes, was evaluated between L1 and control conditions. There was no difference at the interface for both conditions at the shallow depths (Fig. 7. A and C), but L1 coated group showed statistically significant lower intensity compared to control across 240 μm distances from implantation site at the deeper depth (Fig. 7. B and D; p<0.0001).

Fig. 6. Immunohistochemically evaluation of microglia activation at 16 weeks.

Representative images of Iba-1 around implanted electrode (dashed squares) at 300–800 μm (A) and 1000–1500 μm (B) depth from the brain surface in the control and L1 coated group. Normalized intensity value for Iba-1 expression with respect to the distance from the implantation site at 300–800 μm (C) and 1000–1500 μm (D) depth. (Control N=6 animal, N=8 sample (300–800 μm) and N=7 sample (1000–1500 μm); L1-coated N=6 animal, N=12 samples (300–800 μm) and N=8 sample (1000–1500 μm) staining. Results were binned 20 μm from the electrode-tissue interface until 240mm away, with 20 μm bin size. All data presented as mean ± SE. * indicates p<0.05, and **** presents p<0.0001 for all.

Fig. 7. Immunohistochemical evaluation of astrocyte activation at 16 weeks.

Representative images of GFAP expression around implanted electrode (dashed squares) from tissue acquired at 300–800 μm (A) and 1000–1500 μm (B) depth from the brain surface in the control and L1 coated group. Normalized intensity value for GFAP expression with respect to the distance from the implantation site at 300–800 μm (C) and 1000–1500 μm (D) depth. (Control N=6 animal, N=9 sample (300–800 μm) and N=13 sample (1000–1500 μm); L1-coated N=8 animal, N=10 samples (300–800 μm) and N=8 sample (1000–1500 μm) staining. Results were binned 20 μm from the electrode-tissue interface until 240mm away, with 20 μm bin size. All data presented as mean ± SE. **** presents p<0.0001 for all.

4. Discussion

Despite the substantial progress that has been reported using intracortical neural implants for neuro-prosthetic applications, many studies have demonstrated instability in chronic recordings over time due to the corrosion of the recording sites, degradation of insulating coatings, and the brain neuro-inflammatory responses that cause persistent microglia activation, glial encapsulation, and neuronal loss (Golabchi et al. 2018; Jorfi et al. 2015; Kozai et al. 2015b; Wellman et al. 2017). Although, the precise mechanism of device failure is a considerable interplay among different modes, the neuro-inflammatory responses to the implant are considered a key obstacle in realizing seamless neural electrode-tissue interface. In general, recording yield is not optimum at the beginning and the signal quality degrades over weeks to months (Kozai et al. 2015a). A growing number of research studies have investigated biological intervention strategies aimed at improving electrode performance. Covalently linking neuronal cell adhesion molecule L1 to the silicon surface is a promising approach to selectively enhance neuronal attachment and growth while reducing astrocyte and fibroblast attachment (Azemi et al. 2008; Collazos-Castro et al. 2013; Webb et al. 2001). Our previous research (Azemi et al. 2011; Eles et al. 2017; Kolarcik et al. 2012) showed that the L1 coated probe implanted in the cortex, spinal cord, and dorsal root ganglia (DRG) reduced glial activation and maintained higher neuronal density and axonal regeneration around the implants, and even neuronal attachment to the probe’s surface from acute to a chronic period (Azemi et al. 2011; Kolarcik et al. 2012; Kolarcik et al. 2015b). Furthermore, we demonstrated that L1 coated Utah arrays yield significant recording benefits at acute time points (Cody et al. 2018). Present study is the first to evaluate the long-term electrophysiology recording performance of L1 coated single shank Michigan arrays along.

4.1. Depth independent chronic neurophysiological recording

We first examined the overall recording performance across all depths. Evoked SU and MU activity are reported since it is more reliable than spontaneous recording for comparing day to day performance (Kozai et al. 2014b). To remove further motion artifact associated with awake recording, data are obtained from anesthetized animals (Michelson et al. 2017).

3.1.1. Surgery day

Previously, Two-photon studies revealed that upon electrode insertion, microglia send their process toward the implant surface causing encapsulation of the devices by microglia processes hours after insertion (Kozai et al. 2016b; Kozai et al. 2016c; Kozai et al. 2012b), while L1 coating reduced the encapsulation for at least 6 hours after implantation (Eles et al. 2017). Microglia might function as a barrier between the electrode and nearby neuron, or affect the neuronal health and activity by releasing inflammatory cytokines and reactive oxygen species (Bechade et al. 2013). On the day of surgery, we did not observe any difference in 1 kHz impedance between the L1 and control group (Fig. 1. A), this suggests that the microglial process encapsulation does not affect 1 kHz impedance to a significant degree. Despite the similar impedances, significantly higher SU SNR and active SU SNR in L1 group compared to the control were observe, which might be the result of closer distance between neurons and electrodes due to diminished microglial barrier in between (Fig. 1 D&E). Although, it is not clear how L1 is affecting the particular functions of microglia that mentioned above, it is possible the sialic acid residue on L1 can suppress microglia phagocytosis through the SICLEC-E receptor, similar to NCAM (Brown and Neher 2014; Eles et al. 2017).

4.1.2. Acute to early chronic (1–4 weeks post-implantation)

After implantation, impedance in both groups gradually increased and reached the peak after 3~4 weeks of residing in the brain. As expected, the noise floor also increased over this period (Fig. 1. B). The increase in impedance may be the result of gliosis reaction peaking at 4 weeks or the result of wound healing during which the insertion created liquid path from the electrodes to the sub-meningeal space is sealed. In the past, we have found that gliosis is reduced at 4 weeks of implantation for the L1-coated non-functional dummy probe (Azemi et al. 2011), which would have resulted in a decrease in impedance. Instead, no differences in 1 kHz impedance between the two groups were observed, suggesting that the liquid path mechanism is more dominant. However, even though there was no difference in 1 kHz impedance, changes in the characteristics of the full spectrum are observed between L1 and uncoated electrodes (not shown). In the follow up work, equivalent circuit modelling will be performed to tease out the contribution of each potential circuit component (Alba et al. 2015; Cody et al. 2018; Williams et al. 2007). During this phase, the SU yield increased up to week 3, consistent with the wound healing pattern speculated above. Although 1 kHz impedance and noise floor showed no significant difference, greater SU yield, SNR, and signal amplitude in L1 group compared to the control was observed, which may be due to the higher neuronal density and health around the L1 coated implants compared to the controls.

4.1.3. Early chronic (4–8 weeks post-implantation)

During this period, the impedance is stabilized early at week 4 in the L1 groups, but continues to increase in the control until week 7. There was a drop of recording yield in both but stabilized for the L1 group by week 5 while control continued to decline till week 8. The increase in impedance may be dominated by gliosis at this phase, which is reduced by the L1 coating.

4.1.4. Chronic (8–16 weeks post-implantation)

At this time, recording properties (SU yield, SNR, and amplitude) in L1 implants increased initially and stabilized thereafter, while no increases were observed in control animals. This was followed by continuous drop in impedance in control starting week 10 post-implant, which may be due to the degradation of insulating layer or the iridium recording sites. Interestingly, such decrease was delayed in L1 group till week 14. One assumption is that the lower inflammatory condition around L1 implant results in lower number of free radicals that affects the electrode materials stability (Park et al. 2018; Prasad et al. 2012). Further study is necessary to focus on exploring this assumption.

4.1.5. Summary

In summary, although, the overall 1 kHz impedance modulus was not significantly different between the coated and control groups, there are interesting temporal patterns that revealed some difference between the two groups. Nevertheless, the recording performance of L1 coated array over the 16 weeks indwelling period is significantly improved from control in SU yield, SU SNR, and SU amplitude (Fig. 1 C, D, G; p<0.0001), and different dynamic phase of recording was observed. Different studies in the field define acute, sub-chronic and chronic time points differently (McConnell et al. 2009; Potter et al. 2012; Prodanov and Delbeke 2016). This study showed similar trend to our previous chronic recording work (Kozai et al. 2014b) with slightly different timings. In the previous study, electrophysiological recordings were divided into acute to early chronic (0–1 week), early chronic (1–4 weeks), early chronic stabilization (4–7 weeks), early chronic to chronic transition (8–12 weeks), and chronic (after 12 weeks) periods (Kozai et al. 2014b). In our case, we considered the surgical stage as a separate time point, since statistically significant higher SNR in L1 implants than the control was observed, which may be explained by the immediate effect of L1 coating on decreasing the microglia coverage towards the implant (Eles et al. 2017). We also combined the last two periods defined in the previous study into a single chronic phase due to the temporal patterns that span both phases in our study. Recently, Capadona’s group divided the neural recording performance into two separate phases, a dynamic phase (first 12 weeks) and a persistent chronic or late-stage modified state (CMS), after 12 weeks (Bedell et al. 2018) and they observed a significant decline between the initial dynamic phase and the CMS in electrophysiological output. In our case, after averaging the data across recording sites and dividing them to the dynamic and CMS phases, we did not observe a significant difference between the two phases in both groups. In fact, the first 8 weeks seems more dynamic than the last 8 weeks as far as the channel yield is concerned.

In our previous study, L1 coated Utah arrays only demonstrated improvement in the acute time points in rats (Cody et al. 2018). In that study, the gross array ejection was observed for all implants regardless of coated or not. Such massive host tissue response is likely to be a result of the large durotomy, 16 penetrating shanks and the floating design of the array top. As a result, the potential effects of the L1 coating were overwhelmed at later time points. In this study, the single shank probe requires a very small craniotomy, with a single insertion injury, and the top of the probe is anchored on the skull preventing any gross movement. Such configuration allows us to better evaluate the coating effect. Furthermore, we compared the evoked MU signal to noise firing rate ratio between L1 and control group and no significant difference across any time points were observed (Fig. 1 F). This may indicate that the L1 coating does not modify the local neural circuitry.

4.2. Depth dependent chronic neurophysiological recording

Since neuronal distribution varies significantly across cortical layers and brain regions, it is important to examine the recording results from different depths and compare the control vs. coated electrode performance from similar depth. One of the important aspects of this study is that we were able to record from different layers of the brain, thanks to the configuration of the electrodes. The layer dependent recording metrics in both groups were reported (Fig. 2 and supplementary Fig. S. 5 & Fig. S. 6) to present the precise change at different depth (every 100 μm) over time. In addition, we group the neural recording performance into the two depths, 300–800 μm, which spans cortical layers III, IV, V, and VI with expected higher neural activity, and 1000–1500 μm, which include the axonal rich callosum with no expected single unit yield and hippocampus.

4.2.1. 300–800 μm depth

After examining the data at the shallower depths that contain higher density of active neurons (Keller et al. 2018), we observed that from the day of surgery, control group were capable of picking up SUs activity until the end time point of study. However, higher numbers of recording sites in L1 group were able to detect single units, indicating L1 coating’s benefit (Fig. 2 A, B, and C).

4.2.2. 1000–1500 μm depth of the brain surface

By looking at SU yield at the depth of 1000–1500 μm (CA1 of the hippocampus), at the day of surgery, both groups have equal number of channels recording SU signals (Fig. 2 A, B, and D). However, the yield in control group dropped significantly to about 10% by week 4, and never recovered. More severe recording yield degradation in the deeper depth compared to the shallow depth, may be the result of several possible mechanisms. First, pyramidal neurons are more robust than the neurons in CA1 sublayers (Mizuseki et al. 2011). Secondly, there is a higher degree of mechanical damage in deeper region compared to the shallow depth due to micromotion of the tip. The single shank probe acts as a cantilever with one end fixed on the skull but the other end hanging in the tissue experiencing more micromotion. In contrast, the L1 treated group decreased in yield in the first two week, then recovered by week 3, maintaining a high SU yield throughout the 16 weeks, demonstrating L1’s potent ability to enhance neuron survival and recovery post injury.

4.3. Chronic neuro-inflammatory responses

Chronic foreign body response is a complex phenomenon emerging from multiple, interconnected parallel events. It is reported that the population of neuronal cells around the implanted electrode correlates with the quality of recording (Jorfi et al. 2015). We observed a significant difference in neuronal population between the groups overall at both depths (Fig. 3 C & D). It is worth mentioning that the control groups have much higher variation in neuronal counts than the L1 groups across all distances and at both depths. Implantation of electrodes is known to cause a large degree of inflammatory tissue response based on the non-uniform vascular distribution from animal to animal, and across the depths of the same animals (Kozai et al. 2010). These contribute to the large variation in neuronal distribution and consequently recording performance. L1 coating reduces the variability by rescuing neurons and maintaining a high number of healthy neurons near the interface. Many studies have reported a significant decrease of axonal density at the vicinity of the neural implants indicated by NF-200 staining reduction, compared to the distant background (Azemi et al. 2011; Biran et al. 2007). Here, significant enhancement of axonal density at the vicinity of the L1 implant was observed compared to the background tissue after 16 weeks (Fig. 4 C&D). Similar result was reported in our previous study using non-functional probes at week 1, 4, and 8 (Azemi et al. 2011). It was suggested that L1 on the probe surface has interacted with L1 on the axons through hemophilic bindings (Azemi et al. 2011). Upon binding, L1 is known to be potent at promoting neurite extension in vitro and axonal regeneration in vivo (Hulley et al. 1998; Zhang et al. 2005). Therefore, the overall improved axonal density around L1 probes suggests that L1 may promote the regeneration of injured axons close to the probe insertion.

Further, tissue destruction during implanting the electrode and the presence of foreign material in the brain initiates the cytokines release which later triggers an inflammatory process, and neo-vascularization (Prodanov and Delbeke 2016). Initially, this reaction causes BBB breakdown and infiltrating blood components including the plasma proteins and inflammatory cells cause persistent neuro-inflammation by activating microglia and macrophages around the implantation site (Jorfi et al. 2015), (Selvakumaran et al. 2008). In the present study, the IgG staining, used as a marker for BBB leakage, in L1 probe was not different from control (Fig. 5), which is in contrast with some other studies that reported a negative correlation between BBB leakage and recording performance (Saxena et al. 2013). This suggests that IgG staining may not be a good marker to predict electrode performance (Golabchi et al. 2018).

Microglia activation is a hallmark of inflammatory tissue response to the brain injury and they can cause secondary neuronal damage by releasing cytotoxic molecules and cytokines (Donat et al. 2017). L1 coated probe significantly decreased microglia activation (Fig. 6 C&D). This observation was consistent with our previous studies (Azemi et al. 2011; Kolarcik et al. 2012). Besides, in response to injury and signals from the activated microglia/macrophage, astrocytes are activated and increase in GFAP expression, changes appearance and forms a glial sheath in an attempt to wall of the injury and promote healing (Sofroniew 2014; Wellman and Kozai 2017), but could also isolate the recording electrodes from the neurons. L1 coating showed significantly less overall GFAP staining than control at the deep region (Fig. 7 D, but not the shallow depth (Fig. 7 C), which is correlated the more dramatic recording improvement by the L1 coating from the deeper depth.

In summary, by specifically looking at inflammatory tissue responses in the top and deep region of the brain, we observed statistically significant higher NF-200 (in 60 μm implant zone), higher neuronal count (overall), and lower Iba-1 (within 60 μm of implant) activity in the L1 group compared to the control in 300–800 μm depth, which corresponds well to the improved SU yield in the heatmap. Moreover, at the deeper region (1000–1500 μm), there was significantly higher NF-200 (in 20 μm away from implant), and NeuN (overall), and lower Iba-1 (40 μm away from implant), and astrocyte (overall) intensity around the L1 than control, which clearly correlate with higher SU yield that shown in Fig. 2 A vs B.

This study shows that the L1 coating is effective in improving the chronic recording quality of intracortical implants by promoting neuronal survival, axonal regeneration and reducing microglia activation. This is the first demonstration of a biomimetic and protein-based coating promoting neural recording quality and longevity over a long period of time. Improved recording quality and longevity will benefit the brain-computer interface studies and many other scientific research and clinical applications that rely on chronic neural recording. More dramatic improvement in recording yield is found in the hippocampus, which is an area of the brain involved in memory and learning. Being able to stably record from this area chronically is indispensable in the effort of understanding the neural basis of memory and learning.

Questions remain as to how long a protein coating lasts in vivo and what are the exact mechanisms of action. With the current analytical tools, we are not able to identify the L1 proteins applied in the coating from the endogenous proteins that adsorb on the surface. L1 is covalently bound to the surface which in theory would be more stable than previously reported protein-based coatings that were physically adsorbed (He et al. 2006; Oakes et al. 2018). However, it would be difficult for a foreign protein to survive the various enzymes and oxidative stress available in the host tissue. Regardless, we observed a long-lasting benefit of the coating up to 16 weeks, which could be a result of long lasting coating or effective modulation of the initial host response that leads to better chronic outcome. Additional studies will be performed to more closely monitor the cellular response to the L1 coated implant with 2 photon microscopy with pre-labeled proteins to track the fate of L1 coating. To improve the stability of the protein coating, different material science strategies may be applied to preserve the protein during storage as well as in use (Dankwardt et al. 1998; Grasso et al. 2007; Nath et al. 2008; Nosworthy et al. 2009; Seetharam et al. 2006; Stabenau and Winter 2007). Finally, since the protein is derived from animal tissue, care needs to be taken to rule out immunogenicity or other unknown negative biological effect that the protein and protein degradation products can cause. Fortunately, our coating is nm-thick bound on microscale devices; the risk is minimized due to the very low dose.

5. Conclusion

The current study showed that covalent attachment of the neuronal specific cell adhesion protein, L1, on the silicon-based intracortical implants significantly improved recording performance in the terms of SU yield, SU SNR, and SU amplitude over 16 weeks of implantation. Quantitative immunohistochemitry showed significantly reduced microglia and improved neuronal density and axonal regeneration around the L1 coated implants in both the shallow (300–800 μm) and deep (1000–1500 μm) depths. A dramatic improvement of recording yield is observed at the hippocampus region which is at the tip of implant with a reduced gliosis by L1. These observations emphasize the potent effects of L1 coating in maintaining the high quality electrode-tissue interface, and with further optimization, this technology has the potential to improve the performance of many implantable neural interface devices.

Supplementary Material

1

Fig. S. 1. Schematic presentation of implanted electrode. Single shank Michigan style electrode with 16 circular recording sites (diameter = 30 μm) was ordered from NeuroNexus Technologies in Ann Arbor, MI, USA.

Fig. S. 2. Stability of L1 over time. The relative neurite outgrowth of primary cortical neurons on incubated L1 coated substrates is shown. All data presented as mean ± SE (N=27). * indicates p< 0.05.

Fig. S. 3. Surgery setup. A) Small hole was drilled at a location corresponding to the left visual cortex on the mice skull. The green dashed square shows the location of hole. Three anchoring screws were used to stabilize the dental acrylic head cap and attach the reference and ground wires of electrode. B) The image shows after the electrode was inserted into the visual cortex and sealed with the dental acrylic.

Fig. S. 4. Impedance characteristics. Bode plot of a bare IrOx and L1 coated electrodes in PBS (N=5). The black/red circles show the impedance of uncoated electrodes compared to L1-coated electrodes (gray/orange squares). All data presented as mean ± SE.

Fig. S. 5. Evoked signal amplitude as a function of depth and time. The signal amplitude depth dependent of L1 group (A) and the control (B) over 16 weeks of recording are plotted. Note: there are greater signal amplitude around layer IV and CA1 of hippocampus in L1 group, while, it is only observed around layer 4 in control group. L1 group showed higher signal amplitude in shallow depth (C; 300–800 ¼m) and deeper region (D; 1000–1500 μm) than control. All data presented as mean ± SE. * indicates p<0.05, and **** presents p<0.0001 for all.

Fig. S. 6. Evoked SNR as a function of depth and time. The SNR depth dependent of L1 group (A) and the control (B) over 16 weeks of recording are plotted. Note: there are greater SNR around layer IV acutely in L1 group, while, it is not observed in control. L1 group showed higher SNR in shallow depth (C; 300–800 ¼m) and deeper region (D; 1000–1500μm) than control. All data presented as mean ± SE. * indicates p<0.05, and **** presents p<0.0001 for all.

Highlights.

Neuroadhesive protein coating:

• Improves chronic recording performance in mouse brain

• Increases neuronal density and axonal intensity close to the implant

• Reduces inflammation and glial scar around the implant

These findings demonstrate Neuroadhesive protein coating as a promising candidate for improving the longevity of neural implants.