All the procedures in the present study were approved by the Committee for Animal Experimentation at the Institute of Life Sciences of the Hebrew University of Jerusalem. This study was conducted using female Sprague Dawley rats (215–340 g).

To address the technical features required to prepare thin sections of implanted gMμE-platforms along with the parenchyma around it, we fabricated non-functional implants constructed of a Perforated Polyimide (PI)-based MEA Platform (PPMP) that carries a dense array of gold mushroom shaped microelectrodes (gMμE, Figure 1). The 1.7 mm long, 280 μm wide, and 16 μm thick non-functional gMμE-PPMPs were divided to a 0.9 mm long solid proximal part and a 0.8 mm perforated distal part (Figure 1). The perforated segment tapered to form a sharp tip. The width of all the rectangular perforations was 7–8 μm and the lengths of the different perforations were 65, 47, and 44 μm (Figure 1). gMμEs were electroplated at a pitch of 8 μm in rows along the 15 gold conducting lines that run along the platform (Figure 1 and Supplementary Figure 1).

The gMμE-PPMPs were constructed using standard photolithography fabrication methods as follows (Supplementary Figure 1). First, an aluminum releasing layer was sputtered on a 3-inch silicon wafer (University Wafer, United States), followed by a spin-coated 15 μm thick polyimide layer (PI 2610, HD Microsystems, Germany) that served as an insulating layer and the main mechanical backbone of the platform. A triple metal layer of Cr/Au/Cr (20/120/20 nm) was then patterned and e-beam evaporated as interconnects, pads and scribe-lines. Next, a second 1 μm thick insulating layer of polyimide was spin-coated, followed by the deposition of a 1 μm SiO₂ with Plasma Enhanced Chemical Vapor Deposition (PECVD). A 1.5 μm photoresist layer was then patterned. Dry etching by RIE was used to define 1.5 μm molds for electroplating the gMμEs and pads through the SiO₂ and the one micrometer thick insulating PI layer. After removal of the top Cr layer by wet etch, gMμEs with cap diameters of ∼2 μm were electrodeposited at a pitch of 8 μm along the tip and a perforated section of the platform (Figure 1 and Supplementary Figure 1). An additional 300 nm of SiO₂ was deposited with PECVD and a photoresist layer were used to define the perforated pattern of the platform. The photoresist and SiO₂ layers were then removed and the platforms were then released from the wafer by anodic metal dissolution and thoroughly rinsed in distilled deionized water.

A 1–1.5 cm longitudinal cut of the skin on the head was made and the anterior, dorsal surfaces of the skull were exposed. Two craniotomies, one in the left and the other in the right frontal bones, were performed at the desired reference points (coordinates: AP: +3.5 mm; ML: ±2.5 mm from the Bregma) and the dura was gently resected (0.3–0.5 mm long incision). The 1.7 mm long platforms held by forceps mounted on a micromanipulator were slowly inserted into the motor cortex to a depth of 1.8 mm. The electrodes were gently released from the holder and the craniotomy was sealed with melted bone wax (W810, Ethicon, Belgium). The wound was treated in situ with antibiotic ointment (Synthomycine, chloramphenicol 5%) and sutured with nylon sutures. Then the rats received an intraperitoneal injection of the antibiotic Enrofloxacin 50 mg/ml (5% W/V) at a dose of 15 mg/kg diluted with saline to 1 ml (Baytril, Bayer Animal Health GmbH, Leverkusen, Germany). In line with standard protocols to prevent postoperative pain, the rats received for three consecutive days after gMμE-PPMP implantation non-steroidal anti-inflammatory/analgesic drugs. A subcutaneous injection of Carprofen 50 mg/ml (5% W/V) in a dose of about 12 mg/kg (Norocarp, Norbrook Laboratories Limited, Newry, Co. Down, Northern Ireland) during surgery. Then, to further reduce the stress and pain caused by injections and prevent mechanical stress to the skin around the implantation site, the rats were fed on days 2 and 3 post PPMP implantation by Meloxicam (Rheumocam, oral suspension 1.5 mg/ml, Chanelle pharma) dissolved in palatable Jelly. To that end, Meloxicam dissolved in agar (Meloxicam-jelly) prepared in a small Petri dish (diameter of 35 mm) was placed in the rat cages. The Petri dishes were removed at the end of days 2 and 3. Visual checks confirmed that the rats consumed the entire volume of the Meloxicam-jelly. After surgery, the animals were housed individually to prevent them from chewing the implants. Actually, this is relevant only for transcuteneous implants. In case of dummy probes it is important to keep rats separately only for the first 10–14 days after the surgery. I overlooked it because automatically, i keep all my transcutaneously implanted animals separately.

For brain tissue fixation, individual rats were deeply anesthetized with isoflurane (Piramal, United States) followed by an IP overdose injection of pentobarbital (4.5 ml per 250 g rat, CTS Group, Israel). When breathing had stopped, the rats were transcardially perfused with phosphate buffer saline (PBS). This was followed by a 4% paraformaldehyde in PBS (PFA, Sigma-Aldrich) perfusion for immunohistology and 1–2.5% glutaraldehyde/2% paraformaldehyde (Agar Scientific) for transmission electron microscopy (TEM). In both cases, the perfusion rate was 10 ml/min and lasted for 40 min. Next, the skulls were removed and the implanted brains were post-fixed at 4°C for an additional 12–24 h. in either PFA (for immunohistology), or glutaraldehyde/paraformaldehyde (for TEM). Thereafter, the fixed and exposed brains destined for immunohistology were washed in PBS and incubated for 1–3 days in a 30% sucrose solution in PBS at 4°C.

To prepare the brain tissue for cryosectioning (immunohistology), cubic shaped portions of tissue (approximately 1 × 1 × 1 cm) with the PPMP in their center were isolated. The isolated piece was placed in a freezing medium (Tissue- Plus O.C.T. Compound, Scigen) and frozen at −80°C. The frozen tissues along with the implanted platform were then horizontally sectioned into 40 μm thick slices using a Leica CM1850 Cryostat. Individual slices were collected and placed in 24 well plates containing PBS. The tissue slices were then incubated in blocking solution (1xPBS, 1% heat-inactivated horse serum (Biological Industries), 0.1% Triton X-100 (Sigma Aldrich)) for 1 h at room temperature (RT) under gentle shaking. Next, the slices were incubated with a diluted primary antibody for 3 h at room temperature (RT) and washed three times with the blocking solution. This was followed by 1-h incubation at RT with the diluted secondary antibody after which the slices were washed with the blocking solution three times and stained with the nuclear marker DAPI (Sigma–Aldrich, 1 mg/ml 1:1,000) for 15 min at RT. After washing with the blocking solution and PBS, the slices were mounted on Superfrost Plus Slides (Thermo Fisher Scientific) and sealed by a Vectashield (VE-H-1000 -Vector Labs) mounting medium.

For TEM imaging, glutaraldehyde/paraformaldehyde fixed tissue along with the PI, MEA platform implants were washed with PBS and sliced by a Leica VT1000S Vibrotome using a ceramic blade (Campden Instruments Ltd.) into 200 μm thick horizontal sections. The slices were deposited in 24 well plates with PBS.

After eight washes with 0.1 M cacodylate buffer at pH 7.4 (SigmaAldrich) the tissue was post fixed by 1% osmium tetroxide (Electron Microscopy Sciences) and 0.6% K3Fe(CN)₆ in a 0.1 M cacodylate buffer for 1 hr. at room temperature. The slices were then washed again in a 0.1 M cacodylate buffer and dehydrated by a series of increasing concentrations of ethanol solutions of 10, 25, 50, 75, 90, 96, 100, and 100%. Finally, the slices were embedded in Agar 100 (Agar Scientific). The embedded preparation was then thin-sectioned and observed using a TEM Tecnai 12 microscope at 100 kV. The shown TEM images were taken from sections prepared across the perforated part of the implant (rows 2–6 as marked in Figure 1). Efforts were made to orientate the thin sections perpendicular to the long axis of the polyimide platform and along the gMμEs. As the diameter of the stalks and caps of the gMμEs are in the range of 1–3 μm, a slight deviation from a perfect sectioning angle, resulted in imperfect sections that did not pass through the entire length of the gMμE cap and stalk. Thus, in some cases, the sections went through the entire length of the mushrooms cap, stalks and the contact between the stalk and the conducting line (for example Figures 6, 7A). In others, the thin sections were slightly tilted in respect to the long axis of the gMμE stalks. In these cases the stalk of the mushroom appears to taper toward the polyimide platform. The observed TEM images and conclusions represent transmission electron microscope imaging of over 250 thin sections prepared from 25 different PPMPs implants.

Immunolabeling, imaging, image processing, and analysis were conducted as detailed in previous studies from our laboratory (Huang et al., 2020; Sharon et al., 2021). Briefly, neurons were concomitantly labeled with two antibodies: one for neurite labeling [mouse anti-Neurofilament 160/200 monoclonal antibody (Sigma-Aldrich N2912, 1:10,000–1:20,000)] and the other for neuronal nuclei [mouse anti-NeuN monoclonal antibody (Merck MAB377, 1:200)]. Astrocytes were labeled with chicken anti-glial fibrillary acidic protein (GFAP) polyclonal antibodies (Thermo Fisher PA1-10004, 1:500–1,000). Microglia were labeled using rabbit anti-Iba-1 monoclonal antibody (Abcam ab178846, 1:1,000). For the secondary antibodies we used goat anti-mouse Alexa 488, goat anti-chicken Alexa 647 (Thermo Fisher A-11001 and A21449, respectively, 1:100) and sheep anti-rabbit Cy3 (Sigma–Aldrich C2306, 1:100).

Confocal image stacks of the immunolabeled slices were acquired with an Olympus FLUOVIEW FV3000 confocal scan head coupled to an IX83 inverted microscope, using a 20X air objective (NA = 0.75). Typically, 15–30 confocal slices were acquired, with a vertical spacing of 1 μm. Image processing of the immunolabeled sections was conducted using the Fiji distribution of ImageJ (Schindelin et al., 2012; Schneider et al., 2012).

Two methods of analysis and representation of the cell densities in contact and around the PPMPs were used: (1) The densities of the astrocytes and neurons, including their cell bodies and neurites, were analyzed and displayed as the relative fluorescent intensities with respect to the normal background. These are referred to in Figures 2E2,E3 as the Normalized Fluorescent Intensity (NFI) values (Huang et al., 2020). (2) The density of the microglia and neuronal cell bodies per 100 μm², at a given shell around the implant, and within the pores were calculated by manual counting (Figures 2E1,E4). The counting of these cell bodies was done by merging Iba1 labeled microglia or NeuN labeled neurons with the nuclear marker DAPI.

Average fluorescent values and cell counting characterizing the FBR in space and time were measured and calculated from cortical brain slices prepared from sections across rows 4–8 of the perforated part of the implant (Figures 1B, 2). We used 2–10 hemispheres/experimental points in time (Supplementary Table 1). Each brain hemisphere was used to prepare 1–6 tissue slices. Each slice was used to prepare a single maximal projected image generated by 10 consecutive optical sections. For more data on the numerical values and statistical tests, see Supplementary Table 2.

To address the technical features required to prepare thin sections of implanted gMμE platforms along with the parenchyma around it, we fabricated non-functional implants constructed of a Perforated Polyimide (PI)-based MEA Platform (PPMP) that carries a dense array of gold mushroom shaped microelectrodes (gMμE-PPMP, Figure 1). PI was selected because it is a biocompatible polymer with a Young’s modulus of 2.5 GP. Importantly, based on studies demonstrating that PI implants can be thin-sectioned for histological examinations (Mercanzini et al., 2007, 2008; Richter et al., 2013; Xie et al., 2014; Boehler et al., 2017; Huang et al., 2020) our laboratory has developed procedures to section implanted gMμE-PPMPs along with the surrounding brain parenchyma for light and transmission electron-microscope (TEM) studies (Huang et al., 2020; Sharon et al., 2021). In the present study, we fabricated 1.7 mm long, 280 μm wide and 16 μm thick non-functional gMμE-PPMPs. The proximal 0.9 mm of the implant was constructed of solid PI, and the remaining distal part was perforated (Figure 1). The perforated segment tapered to form a sharp tip. The width of all the rectangular perforations was 7–8 μm and the lengths of the different perforations were 65, 47, and 44 μm (Figure 1). gMμEs were electroplated in rows along the conducing gold lines which run in between the perforations (Supplementary Figure 1). The high density of the gMμEs served to increase the probability of successfully preparing thin sections (80 nm) for TEM imaging through gMμEs and PI along with the interfacing brain parenchyma. The perforated microarchitecture of the platform reduced the projected solid surface area of the perforated part by 35% and allowed cells to extend branches or migrate through the perforations. Each pore in the PI platform approximately doubled the PI surface to which the cells could adhere.

To examine the interfaces formed between the brain parenchyma and implanted gMμE-PPMPs, cross-sections for transmission electron microscopy of gMμE-PPMPs along with the surrounding tissue were prepared. We selected to examine the ultrastructure of the inflammatory scar at 2, 4, and 8 weeks after electrode implantation, since our earlier immunohistological studies showed characteristic alterations in the distribution and densities of the microglia, astrocytes and neurons at these points in time (Figure 2 and Huang et al., 2020; Sharon et al., 2021). For the reader’s convenience, the overall spatiotemporal relationships between microglia, astrocytes, neurons, and PPMP implants is briefly presented in Figure 2 using conventional immunohistological imaging.

For the ultrastructural analysis, gMμE-PPMP implanted brains were chemically fixed by standard transcardial perfusion of glutaraldehyde/paraformaldehyde fixative. Since MEA platform implantation unavoidably damages blood capillaries along the insertion path, concern was raised whether the quality of tissue fixation around the implant will suffice to preserve the tissue ultrastructure. In retrospect, based on the preservation qualities of the cell membranes and subcellular organelles including the mitochondria, the smooth and rough endoplasmic reticulum, synaptic vesicles, post-synaptic densities and myelin, we concluded that the perfusion of the fixative was not impaired in the surroundings of the implant. It is important to note, however, that as in other ultrastructural studies of the CNS, the extracellular spaces between the various cell types is reduced by approximately 20% (Korogod et al., 2015; Hrabetova et al., 2018; Soria et al., 2020). Since the volume of the implanted PPMPs is not altered by the fixatives, transcardial fixation led to the generation of mechanical tension around the implant. This often tears the tissue around the implant. Importantly, tissue growing into the PPMP pores and adhering to the platform surfaces remained tightly attached to the platforms, and the break in the tissue took place between cells or even across cells a few micrometers away from the implant surface.

Since the relative positioning of the PPMP and the cells around it are not altered by the classical method of transcardial fixative perfusion, the TEM analysis presented here suffices to provide the essential and missing information on implant brain-tissue interfaces. In addition, because the range of the shrinkage factor is known, the genuine extracellular clefts can be estimated. It is important to note that TEM examination of hundreds of thin sections representing over 25 gMμE-PPMP implants revealed that the gMμEs maintain stable contact with the conducting lines used for their electroplating. That is, the gMμEs are not striped off during the platforms insertion or during the thin sectioning of the tissue along with the implant for TEM imaging.

Typically, at 2 weeks post gMμE-PPMPs implantation, tightly adhering dark cytoplasmic microglia processes (dark as compared to other cell profiles in their surroundings) encapsulated individual PI “ribs” (Figures 3, 4). The cell bodies from which the dark cytoplasm emanated contained characteristic microglia nuclei with clumps of heterochromatin beneath the nuclear envelope and throughout their nucleoplasm (Figure 3 and see Tremblay et al., 2012; Garcia-Cabezas et al., 2016; Savage et al., 2018; Nahirney and Tremblay, 2021). The electron-dense cytoplasm of these microglia was bordered by a clear plasma membrane and contained rough endoplasmic reticulum, mitochondria and dark inclusions which may plausibly be lysosomes and lipofuscin granules (Figures 3, 4).

In line with Figure 2 and our earlier immunohistological studies (Huang et al., 2020; Sharon et al., 2021), 2 weeks post-PPMP implantation, TEM images revealed the presence of astrocyte cell bodies and branches in close proximity to the platform surface and within the platform pores (Figures 3B, 4A). Astrocyte cell bodies could be identified by their pale nuclei that had a thin rim of heterochromatin and pale cytoplasm (Figures 3, 4A). Typically, the cytoplasm of astrocyte branches are characterized by sparse electron- dense material containing intermediate filaments (glial fibrillary acidic protein, GFAP, Garcia-Cabezas et al., 2016; Nahirney and Tremblay, 2021).

Two weeks after implantation, the dark microglia cytoplasm that adheres tightly to the gMμE-PPMPs surfaces are often interposed between the PPMP surfaces and the astrocytic branches, thus mechanically preventing direct contact between the neurite and astrocytic branches and the platform surfaces (Figure 3B).

Immunohistological imaging of the neurite revealed that at 2 weeks post-implantation, neurites extended into the PI platform pores (Figures 2C,E3). Based on the presence of microtubules in axons and GFAP in astrocytes (Figure 3B) it was possible to differentiate between the branches of the astrocytes and the unmyelinated neurites (axons and dendrites). At 2 weeks post-PPMP implantation, no myelinated axonal profiles were observed in the immediate vicinity (<10 μm) to the gMμE-PPMPs. Further away from the implants (>10 μm) myelinated axons were observed (Figure 3).

Neuronal cell bodies characterized by typical round euchromatic nuclei, the presence of electron dense nucleoli and nuclear membrane invaginations were observed as close as ∼20 μm from the PI platform and onwards (see also Figure 2E4). Chemical presynaptic terminals identified by the presence of profiles containing clusters of synaptic vesicles or chemical synapses identified by presynaptic fibers in association with typical post-synaptic densities were imaged at a distances of approximately 10 μm from the platform.

Immunohistological examination of the changes in cell composition and distribution within and around the implanted PPMPs 4 weeks after PPMP implantation (Figure 2) suggested that the parenchyma around the implant had undergone regenerative processes. These included: (a) a reduction in the average microglia density in the first shells around the implant, but not within the PPMP pores (Figure 2E1 and Huang et al., 2020); (b) a significant increase in the average density of the neuronal cell bodies in the first shell around the implant, preceded by the extension of neurites toward the implant and into the PPMP pores (Figures 2E3,E4 and Huang et al., 2020). In contrast to these regenerative processes, the astrocyte branches and cell bodies continued to increase during the fourth week post- implantation both within the PPMPs and in the first shell around it (Figure 2E2).

The overall regenerative processes observed at the confocal microscope resolution were reflected and more finely delineated at the ultrastructural levels. TEM images revealed myelinated axons extending toward and in the vicinity of the implant surface (Figure 5). A considerable increase in the density of structurally mature chemical synapses was seen. At 4 weeks post-implantation, the PPMP’s “ribs” were no longer enwrapped by dark protoplasmic protrusions emanating from microglia cell bodies. Rather, relatively thin layers of electron opaque cytoplasm adhered to the surface of the PPMPs and the gMμEs (Figure 5). Relevant to the electrophysiological recording functions of the implant (see discussion), it is noted that the extracellular cleft formed between the microglia membranes that enwrapped the gMμE remained in the range of 10–20 nm (Figure 5B). The microglia clearly interposed between the myelinated axons extending in the electrodes’ vicinity and between pale cytoplasmic profiles of astrocytic branches and unmyelinated axons. Astrocyte branches and cell bodies, microglia and unmyelinated axonal profiles were seen occupy the pores between the PI “ribs” and directly adhere to the PI surfaces.

In addition, large profiles of dark cytoplasm containing phagocytosed materials were occasionally observed to reside within the pores (Supplementary Figure 5). In a few cases, the section went through the nucleus of these large cells. Based on the heterochromatin distribution of the nucleus, these cells were likely to be microglia. These cell types were never observed outside the PI implant pores.

In summary, whereas clear regeneration of the neuron cell body densities, axons, dendrites and synapses took place within the first shell around the implant, dark microglia branches adhering to the gMμEs and PPMPs were still present. These adhering microglia can be assumed to electrically insulate the electrodes from the surrounding neurons (see section “Discussion”).

Confocal microscope imaging of the cortical parenchyma interfaced 8 weeks after PPMP implantation revealed that the overall regenerative processes that were observed 2–4 weeks after implantation persisted. (a) The microglia density within the implant was further reduced to half of its peak value and to a third in the first shell around the implant (Figure 2E1 and Huang et al., 2020). (b) Whereas, the neurite density (NFI values, Figure 2E3) did not change, the average neuronal cell body density in the first shell around the implant further increased to 86% with respect to the control level (Figure 2E4). (c) On the other hand, the astrocyte (branches and cell bodies together) continued to increase mainly in the first shell around the implant (Figure 2E2). These regenerative trends were reflected at the TEM level, in particular in that neuronal cell bodies were imaged to reside as close as ∼2 μm from the gMμE caps (Figures 6, 7A). The narrow space between the cell bodies membrane and the gMμE caps were occupied by small profiles (with a diameter in the range of <1 μm) of either astrocytes or neurites. Parts of the gMμE stalk and caps were enwrapped by narrow (∼100 nm) dark cytoplasmic protrusions probably corresponding to microglia branches, while other parts appeared to be free of microglia. Myelinated axonal profiles were observed to form a direct contact with gMμE caps (Figure 7B). Chemical synaptic profiles were observed as close as ∼0.5 μm to the PI platform and the gMμEs, and within the parenchyma surrounding the implant (Figures 6, 7). The pores within the PI platform were mainly occupied by astrocytic branches and unmyelinated axonal profiles. No synaptic structures were observed within the pores.

TEM observations conducted 8 weeks after the PPMP implantation, occasionally revealed gMμE that were not insulated by microglia. Under these conditions, the gMμEs formed a direct contact with many small (∼100 nm) axonal or astrocytic profiles (Figure 7D). It is conceivable that the small surface area of these tentatively identified unmyelinated axonal profiles were too small to generate sufficient current to be measured by the gMμE system.