Precise tubular braid structures of ultrafine microwires as neural probes: significantly reduced chronic immune response and greater local neural survival in rat cortex

Abstract

Braided multi-electrode probes (BMEPs) for neural interfaces comprise ultrafine microwire bundles interwoven into tubular braids. BMEPs provide highly flexible probes and tethers, and an open lattice structure with up to 24 recording/stimulating channels in precise geometries, currently all within a 150 ~ 200 μm diameter footprint. This paper compares the long-term tissue effects of BMEPs (12 × 9.6 μm wires) versus single conventional 50 μm wires, by testing nearby chronic immune response and neural survival in rat cortex. Four different types of electrodes were implanted in cortex in each of 8 rats: 1. BMEP with tether, 2. tethered 50 μm wire, 3. BMEP without a tether and 4. untethered 50 μm wire. Quantitative immunohistological statistical comparisons after 8 weeks using GEAP, ED1, and NeuN staining clearly showed that both BMEP implants had significantly less tissue immune response and more neuronal survival than either of the 50 μm wires (p < 0.05) in each of the 8 rats. Data strongly indicate that BMEP tissue responses are superior, and that BMEP designs partly alleviate chronic tissue inflammatory responses and neural losses. The flexible body, tether and open braid lattice and liner wire diameters of BMEP designs may all contribute to reducing the biological long-term response.

I. Introduction

Implanted brain machine interfaces (BMIs) are an important aspect of neuroprosthetics [1] in animals [2, 3] and human subjects [4–6]. However, implanted BMIs need the conditions for longevity to be established and risk ameliorated. Available implanted microelectrodes are relatively unreliable for truly long-term applications. This severely limits chronic clinical use [8–11], and electrodes pose significant risk. Penetrating electrodes can damage brain tissue during insertion and later evoke chronic tissue responses due to their prolonged presence in the brain [12]. Commonly, the chronic responses show: (1) the loss of many recording sites in an array over time and (2) varying degrees of immunoreactive encapsulation or ‘glial scar’ wrapping the whole electrode body [12, 13]. In addition, there is neural loss nearby. The relationship between these outcomes and electrode design is still unclear. Various electrode and insulation materials [14–22], differently sized and shaped electrodes [13, 23, 24], and different anti-inflammatory agent coatings [25, 26] have been tested. One possible source of chronic responses is brain tissue motion and its interactions with stiff microelectrodes tethered to the skull [27–30]. The brain floats in cerebrospinal fluid (CFS) [31] and can temporarily deform within the confines of the skull [32]. Rhythmic mechanical waves are generated by respiratory or cardiac pulsations and transfer to brain, which thus frequently pulses over micrometer scales [33]. Voluntary activities such as running, jumping, or fast head turning may produce more intense brain motions [34]. Conventional microelectrodes are usually very much stiffer than brain tissue, and do not absorb or follow tissue micromotions and deformations to any appreciable extent. Significant mechanical stress to surrounding tissue may likely be sufficient for the tissue responses observed. Tissue stress depends both on material modulus and on geometric form and spatial scale of the implanted structures. Additional factors may be the fine spatial scale of elements comprising foreign bodies and their actions as diffusion barriers. We have focused on mechanical designs that can improve the overall mechanical compliance and diffusion. Our approach uses tubular braids with ultrafine wires, of 9.6 μm in diameter. These form an open lattice of fine members. This partly addresses compliance, spatial scale and diffusion barrier concerns. We term these "braided multi-electrode probes" (BMEPs) [7]. In the BMEP, each filament has a spring-like helical form (albeit with curved zigzag ’wrinkles’). There are no bonds at crossing points, and overall the BMEP has very high mechanical compliance both laterally and axially. Mechanical tests have shown that bodies of BMEPs with 12 × 9.6 μm Nichrome wires had 21 times higher lateral compliance and the tethers had 97 times higher lateral compliance when compared to a single 50 μm Nichrome wires. However, structurally, the BMEPs are 2 times larger in outer diameter [7]. BMEPs have diamond-shaped open diffusion windows among wires and multiple recording sites within the 150 μm diameter footprint.

In this paper we test if BMEPs can improve brain tissue responses, despite the larger outer diameter of the assembled structure. For comparison, we chose single 50 μm Nichrome wires as a conventional reference electrode. 50 μm wire designs are still widely used in microwire-based arrays [35] and in terms of size the single 50 μm wire could represent a single channel or a shank of other microelectrodes, e.g., Michigan probes [13, 36, 37] and Utah arrays [38, 39]. To perform comparisons, we implanted both types of electrodes (braid and 50 μm wires) in the brains of each adult rat, to control for individual rat variations in inflammatory state, etc., and after 8 weeks, we performed immunohistological analyses to assess the chronic changes in the brain that occurred near each probe.

II. Materials and Methods

All procedures were compliant with PHS and USDA recommendations, and had full approval of the Drexel Institutional Animal Care and Use Committee. Female adult rats (Sprague Dawley, 3 months old) were used exclusively here avoiding variations in size, motor behaviors, and inflammatory responses between sexes.

A. Experiment design

Most implantable microelectrodes require an electrical connector to make a connection to an external electronic device for recording neural signals or stimulating neural tissue. For this connection a tether system is commonly used between the microelectrode in the brain to a connector mounted on the skull, allowing them to move somewhat with the brain surface. This arrangement is here termed a ’tethered’ electrode. If the stiffness of a tether is high, the tether forces increase, it can effectively partly anchor the electrodes to the skull. ’Floating’ probes have no mechanical anchor to the skull. The stiff body of chronically implanted but floating microelectrodes may also act as a persistent mechanical stressor to the softer neural tissue during the frequent micromotions and macromotions of brain in CSF. Stiffer tethers further aggravate these conditions and tissue responses. BMEPs can mechanically support both compliant bodies and compliant tethers. We hypothesize these are both important factors. To test the effects of both the flexible body and the tether on brain neuroimmune responses and to separate these effects from one another, we used two implant configurations for each probe type in the study: (1) probe with tethers attached to the skull in order to assess body and tether effects (’tethered’), and (2) probe without skull tethers (’floating’) to assess only the floating body effect (see figure 1(c)). We selected 50 μm Nichrome wires as a reference electrode design to compare to BMEPs with 12 × 9.6 μm Nichrome wires. Others have also used microwires as a reference against which to compare novel probes (e.g. 125 μm diameter stainless steel microwires in [39] and 50 μm diameter tungsten microwires in [40]). These BMEPs have been characterized in their mechanical compliance and neural recordings in our previous studies [33]. Since ’floating’ probes were electrically isolated, this prevented recording, and we focus here only on chronic tissue responses. In total 4 electrodes were implanted in the cortex per rat: 2 different types of electrodes each in tethered and floating configurations. We used 8 rats for this experiment design, and found this number provided more than sufficient power to detect differences and effects, with clear statistical significance.

Fig. 1.

BMEPs with 12 × 9.6 μm polyimide insulated Nichrome wires (forming a braid structure with about 135 μm outer diameter) and the schematic of methods for implantation and electrode preparation, (a) BMEP with a removable tungsten core (about 90 μm diameter). The BMEP tightly holds the tungsten rod by mechanical friction, (b) BMEP after the removal of tungsten core. Total 12 microwires mechanically maintain the tubular braid structure without collapse though there is no use of glue on the whole braid structure. Dimond shape diffusion windows on the braid wall have about 65 μm side length, (c) Regular animal experiment setup with the BMEP. For BMEP insertion and the tungsten core removal, we use our custom 3 axes micromanipulator (WPI Kite) having another one axis mini micromanipulator (Newport) with a custom adaptor plate to adjust the position of piston inside the BMEP syringe set. The core rod is attached to the piston surface. MP: micromanipulator, (d) Schematic of prepared electrodes before insertion. All BMEPs have a tungsten core inside the tubular braid. The braid forms a sheath over the core. The core acts to enforce temporarily the stiffness of the BMEP for better penetration and is then smoothly removed after insertion is complete, see [7]. (e) A conceptual schematic of our methods to implant two different intracortical electrode designs for testing the effect of flexible tethers, (f) A conceptual schematic of our methods to implant two intracortical electrodes for tests of the effects of a more compliant body design, (g) Four implant positions per rat used in this study. (h) Four possible combinations of position rotations maintaining symmetrical implantations among electrode types. BT: Braid with Tether, B: Braid without tether, 50T: 50 μm wire with Tether, 50: 50 μm wire without tether.

B. Microelectrode preparation

The BMEPs with 12 × 9.6 μm polyimide insulated Nichrome wires over a sharp tip tungsten core (see figure 1(a) and (b)) were built on a syringe platform (see figure 1(c)) by our standard methods [7], but did not connect to an EIB (Electrode Interfacing Board), instead simply to a skull anchor point (see figure 1(e)). The other end of the tungsten rod was attached to the piston surface inside the syringe and so the core was removed from the BMEP by pulling the piston with a mini micromanipulator attached to a 3 axes micromanipulator (see figure 1(b) and (c)). The length of each BMEP body was 2mm and each was prepared in advance and inspected thoroughly before insertion. For standard microwires used as controls, a 50 μm polyimide insulated Nichrome wire was inserted into a 32G blunt syringe needle and fixed with super glue (see figure 1(d)), allowing insertion with the same micromanipulator and 1mm syringe system used for the BMEPs. Following insertion of a 2mm length into the brain at the same insertion speed as BMEPs, and after dental cement headpiece completion with about 5mm of wire embedded in the dental cement, the probe was cut free. For the BMEP without tether, to test floating body compliance effects, the tethers (of non-braided wires) from BMEP constructions were cut and trimmed in advance. The 50 μm wires were prepared in the exactly same way for the ('floating') microwire without tether.

C. Implantation procedure

Animals were anesthetized by intraperitoneal injection of ketamine-xylazine-acepromazine cocktail (KXA, targeted dose 0.1 ml / 100 g body weight). Aseptic procedure was followed. The head was shaved swabbed with Betadyne and rat placed securely in a stereotaxic holder and an incision made to expose the skull. Using a trephine, four holes of 2mm diameter were drilled through the bone, 2mm away from bregma and mirrored over the right and left hemisphere, in rostral and caudal direction each was 3mm away from bregma (see figure 1(g)). By using a 30G syringe needle as a microscalpel, the dura was carefully pierced and an area opened wide enough to insert the electrodes (see below).

For sterilization all electrodes were bathed in 90% Ethanol, and then exposed to UV (Ultraviolet) light on a mirrored surface for at least 30 minutes in advance before insertion. After electrode insertion with our custom manual micromanipulator (see figure 1(c)) at a consistent slow insertion speed (~ 100 μm per minute), 1% agar gel droplets were applied to the four holes to make a small gel dome rising slightly above the skull surface. This protected the brain and dura from the applied dental cement and provided a space for tethers to move more freely. After confirming that the exposed skull was dried, dental cement was poured to cover the exposed skull and capture the protruding wires of the two tethered probes, and the implanted bone screws, and allowed to solidify for at least 15 minutes. After confirming the cement was completely solidified, the 50 μm wire tether and the tether microwires of the BMEP above the cement were cut, and finally the skin was closed over the head cap by wound clips.

We implanted 4 different types of electrodes per rat. To ensure that a specific neuroimmune response was not caused by a specific positioning in the brain, either chronically or during implantation procedures, we varied the implant positions used for each probe type among rats. We flipped the positions in y axis and x axis, and double flipped in x and y axes in order to keep the matching bilateral symmetrical positioning configuration between both tethered BMEPs and 50 μm wires, and between both floating BMEPs and 50 μm wires (see figure 1(h)).

D. Tissue preparation

Most glia scar formation is completed by 6 weeks after microelectrode insertion [40]. Immunoresponses against implanted microelectrodes after 4 weeks are usually similar over time regardless of different types of electrodes and defined as prolonged reactive response [13]. Thus, we selected 8 weeks as tissue harvesting time, by which point responses were representative of chronic effects. Animals were euthanized with Euthasol for immunohistochemistry at 8 weeks post implant. Right after the animal was injected, but before cardiac arrest, perfusion was performed with buffered saline, then followed by 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer solution (PBS), pH7.4. After one day stored at 4°C, the decapitated rat head was stored again in fresh 4% PFA solution at 4°C for 3 days. After 3 days, the skin was removed, wide holes were drilled near electrodes with tethers, the tethers inside the holes were carefully cut and then the whole cap of dental cement was removed. Finally, the whole brain was extracted from the cranium. The extracted brains with electrodes were stored in 30% sucrose in PBS including antifungal Thimerosal at 4°C for about 1 week until the brain sank down to the bottom of solution. Before freezing, only the 50 μm wires were removed carefully from the tissue and all BMEPs remained in the tissue for cutting, as the ultrafine wires did not impede sectioning. The brains with BMEPs were then frozen in the embedding medium (Sakura, Tissue-Tek OCT compound) at −80°C. The frozen brains were sectioned down to 2mm depth in 30 μm thickness with a cryostat in the horizontal plane. During sectioning, microwires forming BMEPs were cut by the cryostat blade and their fragments washed away with PBS. The sections were stored in PBS at 4°C before immunostaining.

E. Immunochemistry

The floating sections were blocked with 4% normal goat serum with 0.5% Triton X-100 (sigma) in PBS solution for one hour at room temperature. Section incubation was performed with primary anti-bodies at 4°C overnight. Sections were then incubated again with secondary antibodies at room temperamre for one hour. The primary antibodies were GFAP (1:2000, DAKO, Z0334), EDI (1:1000, Serotec, MCA341R) and NeuN (1:500, Millipore, MAB377). The secondary antibodies were goat anti-rabbit IgG Alexa Fluor 488 (Invitrogen, A11034) and goat anti-mouse IgG Alexa Fluor 594 (Invitrogen, A11032). Double staining (GFAP/ED1 and GFAP/NeuN) was performed with separate sections.

F. Quantitative immunofluorescence analysis

Fluorescence intensity images were captured by SlideBook 4.3 software using a monochrome camera (Qimaging Retiga-SRV) and Leica DM5500B microscope with fluorescent light sources. Images were captured at 1392 × 1040 pixels with 4096 gray scale levels (12bit ADC). The calibrated square pixel distance was 0.9146 μm per pixel. All images were captured with the same setting using a × 10 magnification objective lens. Since double staining was performed, we used the automatic double capturing function of SlideBook which captures independent images twice for GFAP and ED1 or GFAP and NeuN at a same position of a double stained section by automatically changing filters and fluorescent light sources. This function generates a pair of two images.

To quantify the fluorescence intensity, we developed a modified version of the custom method of Kim et al. [28]. In Matlab, users set points around the implanted electrode surface on an image using a mouse, the program then generated a smooth curved contour of the electrode surface with a total of 180 points interpolated by the MATLAB spline function, and calculated the center of the contour shape, and then generated a total of 180 line intensity profiles (`line profiles’) which radiated from this center in 2° increments from 0° to 358° (see Figure 2(a)). Because most line profiles lie between pixels, the intensity value at every 1.8292 μm distance (2 pixels) along each line profile was calculated with a 2 × 2 weight matrix used to interpolate intensity along each line. The 2 × 2 weight matrix is based on the closest nearby 4-pixel centers expressed as the floating point numbers capturing pixel x and y positions relative to the measurement points at every 1.8292 μm interval along each line (the pixel centers closer to the current point have proportionally more weight). Before normalizing, a black level subtraction was performed. This subtracted the lowest intensity in each section from all intensities in an image, setting the black level to zero. A background reference intensity was calculated by averaging all intensity values in a 100×100 pixels window far from the implant site in each section. We then normalized [26] by dividing the intensity of images along each line by the reference intensity. All intensity values along line profiles were thus converted to relative values in relation to a reference intensity. Reference windows were located at or near one of four corners on a section image at a diagonal distance of 650 – 800 pixels (845 – 1040 μm, diagonal pixel distance = 1.2934 μm) from the implanted site (usually located near the center of the image). Due to the irregular contour shape of electrode surfaces, electrode surface positions from the calculated center were different among line profiles. Thus, line profiles were aligned using the electrode surface positions as the zero starting point, a ’section profile’ at aligned distances was then obtained by averaging all 180 profiles per section.

Fig. 2.

Schematic to visualize the two image analyses for immunoresponses and neuron counts, (a) The relative fluorescent intensity analysis with line profile techniques and GFAP/ED1 staining, (a1) The green color is the pseudo color representing GFAP. White 180 lines (0° ~358°) are the individual line profile radiating from the electrode center. The white curved line enclosure is the contour of the implanted electrode surface generated from several surface points with a spline function. Blue dots are the actual pixel positions representing the white contour. Red dots are the closest pixel positions to the blue dots on the line profiles. The positions of red dots were used as points of electrode surface for this analysis. (a2) Line average of the line profiles from one image to represent one section. (a3) Section average of line averages to represent one animal. (a4) Animal average of section averages to represent data for one electrode combined among 8 rats. (b) The image was magnified to show the detail of line profiles in (a1). (c) Neuron counts in the vicinity of implanted 50 μm wire. The area around the probe was divided into 12 annular subareas. Red and green colors are pseudo colors representing neurons (NeuN stained) and GFAP staining respectively. All cross hairs on this image are the positions of neurons automatically and then manually corrected and selected by our custom program. Once the contour of electrode surface is drawn, the program keeps expanding the electrode contour by radial distances of 25 μm and calculating the number of neurons within the annular subarea up to 300 μm from the contour. Densities are then assessed. Each colored line circle is the outer boundary of each annular subarea and the same colored cross hairs are neuron positions in the subarea.

Sections were collected from 60 μm to 1680 μm depth. We examined section profiles for systematic depth variations in glial reactions among sections but found none. Accordingly we selected 8 sections for each electrode type in each rat for statistical analysis. Next, one ’animal profile’ per rat was obtained by averaging 8 such section profiles for the specific electrode probe design in each rat. Finally all 8 animal average line profiles were averaged to obtain one ‘grand profile’ per electrode design and configuration (see figure 2(a)). Since 4 different electrodes were used for this experiment, 4 grand profiles were for plotting relative fluorescent intensity as a function of distance from the implanted electrode surface, for each stain type used. Taken together, the individual normalized fluorescence profiles for each electrode type formed a N = 180 × 8 × 8 data set (representing lines × sections × rats) of samples for normalized fluorescence for each electrode type at a given distance from the probe. These data supporting either a single factor repeated measures ANOVA, with line profiles, sections and rats as repeated measures, or a 180 × 8 multifactor repeated measures ANOVA including rat and electrode types as factors (see below) and line profiles as repeated measures, or multifactor ANOVA of grand average data.

G. Statistical analysis

All intensity data are presented as the average value and the standard error of the mean (S.E.M). A two-way ANOVA (rat and electrode type as factors) was used in the analysis of results here to compare the grand average data, and then in following comparisons of electrodes and rats we used post hoc t-tests with the Bonferroni correction for specific comparisons. These methods were applied for both fluorescent intensity and cell density analyses. The ANOVA calculation was performed by the anovan.m function included in the Matlab statistical toolbox and the output p-values were also verified with the IBM SPSS package. The p-value threshold to indicate statistical significance was set at 0.05. The factors for two-way ANOVA were electrode types (denoted "Electrodes" in tables) and animals (denoted "Rats") as noted. To compare statistical differences among electrodes in the fluorescent intensity analysis, we selected three variables from line profiles: intensity peaks, intensity peak distances from the surface and total intensity of segmented areas in 25 μm thickness increments up to 125 μm. Total fluorescent intensity of areas was obtained by adding all intensity values within 25 μm distance from a reference point along a profile. In cell density neuron counts, normalized values obtained by measurements in 12 annular subareas were used as the outcome variables.

To compare the viability or neurons in the vicinity of different implants, we performed cell counts with images from GFAP/NeuN stained sections. To count neurons and calculate the cell density, we developed and used a custom Matlab GUI code set which combined automatic blobs detection techniques to find cell bodies, with manual correction cell adding/removing techniques to correct any errors introduced by the automation process. The region in close vicinity of an implant was divided into 12 annular subareas by radial 25 μm increments from the contour surface of each electrode up to 300 μm (see figure 2(c)). The cell density per 100 × 100 pixels (8365.6072 μm²) was calculated in each subarea, based on annulus area, and each density value was then normalized to the average cell density per 100 × 100 pixels of the whole image, which was estimated using the cell density in a much larger area not including the electrode (see figure 2(c)). Because annuli areas closer are smaller, and the annuli areas further larger, the statistical confidence in accuracy of density estimates collected in this way is the greatest furthest from the probe. However, results close to the probes also showed significant statistical differences among probes (see Results). The animals’ cell density per rat was calculated by averaging the average cell density values of 4 sections in each rat, and the grand cell density per electrode was obtained by averaging animal cell density values from 8 rats. Repeating the analyses after removing the most extreme responding rat’s data (i.e., rat in figure 3 (d), (e) and (f)) did not alter the significant results of the analysis. We here report all rats’ data analyses.

Fig. 3.

GFAP/ED1 fluorescent images, which were captured at 4 different positions in the same tissue section which included each type of electrode. Shown with GFAP/ED1 double staining. All colors are pseudo color: green for GFAP and red for ED1. (a), (b), and (c) have exact same 4 image sets with separate staining showing relatively weak tissue responses against 4 implants at 990 μm depth, (d), (e), and (f) have another same 4 image sets showing relatively strong tissue responses against 4 implants at 660 μm depth in a different rat. (a) and (d) show only GFAP. (b) and (e) show only ED1. (c) and (f) show both GFAP and ED1 overlapped. The abbrev. ‘50’ stands for the 50 μm wire, the letter ‘B’ stands for the BMEP and the letter ‘T’ after 50 or B means “with tether”. The white bar represents 50 μm length.

III. Results

A. GFAP Staining

Figure 3 shows reactive GFAP staining used to identify reactive astrocytes near the implanted electrodes. The figure shows the significant differences among electrodes. In most cases a complete ring or GFAP layer encapsulated the untethered 50 μm wire (figure 3(a) and (d), 50), but the shapes of astrogliosis surrounding the BMEPs were more irregular (Figure 3(a) and (d), 50T, BT and B). We obtained plots of the relative fluorescent intensity of the average GFAP response to the 4 different types of implanted electrodes (figure 4(a)). The untethered 50 μm wire had the highest level of GFAP within 25 μm of the electrode surface. The more complete and uniform circular shape of the GFAP layer around the 50 μm wire sometimes contributed to a higher intensity average peak as compared to more irregular shapes or broken circles of astrogliosis. However, the intensity of astrogliosis in an irregular shape was clearly sometimes much stronger (compare figure 3(a) and (d), for the 50T electrode). The order of decreasing peak GFAP responses, in figure 4(a) was as follows: the untethered 50 μm wire highest because of the high radial coherence, the tethered 50 μm wire was in the 2^nd place, the tethered BMEP in the 3^rd place and the untethered BMEP last.

Fig. 4.

Comparisons of grand averages of relative fluorescent intensity of GFAP and ED1 obtained from averages with 180 lines in each of 8 sections from 8 rats. Comparisons among the 4 different electrode types, (a) GFAP results, (b) ED1 results. The cross marks represent the positions of intensity peaks and the asterisk marks are the positions of the half max drops from peaks. The palest colored cloud/ribbon plots with dashed outlines represent the standard deviation and the darker narrower pale colored cloud plots represent the standard errors. The abbrev. ‘50’ means the 50 μm wire, the letter ‘B’ stands for the BMEP, and the letter ‘T’ after 50 or B means “with tether”. The GFAP and ED1 stains show least response to B and then BT. Greatest response is to 50T and then 50.

Results of 2 way ANOVA are shown in Table I. In terms of the GFAP intensity peaks there was a significant difference among both electrodes and rats as determined by the 2 way ANOVA. Post hoc tests with the Bonferroni correction showed the peaks were not significantly different among the 50T, B, and BT. Only the 50 peak was significantly different from the others (p = 0.039, 0.028 and 0.024) as shown figure 4(a), presumably due to the higher radial coherence of glial encapsulation. Rat to rat variations in GFAP response were small but not insignificant in post hoc tests. In terms of peak distances from the electrode surface, there was a significant difference among electrodes and rats in the ANOVA. The post hoc tests revealed that there were no significant differences between the same type of electrodes (i.e., between the two 50 μm wires, or between the two BMEPs) regardless of the tethered or untethered states of each, but there were significant differences between the two different types of electrodes (50 μm vs BMEP) also regardless of their tethering. This peak distance difference among probes is in part consistent with the differing regularity of the two types of electrodes.

TABLE I.

Two Way Anova Results For GFAP And ED1

Factor	Peak	Peak	Area 1	Area 2	Area 3	Area 4	Area 5	Area 6	Area 7	Area 8
		Distance	0–25	26–50	51–75	76–100	101–125	126–150	151–175	176–200
GFAP
Electrodes	0.009*	0.000*	0.000*	0.000*	0.000*	0.000*	0.000*	0.000*	0.000*	0.000*
Rats	0.021*	0.001*	0.010*	0.011*	0.038*	0.044*	0.063	0.050	0.043*	0.078
EDI
Electrodes	0.001*	0.007*	0.000*	0.000*	0.000*	0.000*	0.000*	0.000*	0.000*	0.000*
Rats	0.130	0.880	0.029*	0.016*	0.016*	0.018*	0.020*	0.018*	0.010*	0.004*

The number means the position of segmented by radial 25 pm increment from the electrode surface to 200 pm.

^*significant and below p = 0.05 (Bonferroni corrected). MANOVA Electrodes: GFAP Peaks and distance F (6,38) = 7.028, p = 0.000; GFAP F (24,38) = 13.150, p = 0.000; ED1 Peaks and distance F (6,38) = 5.887, p = 0.000; ED1 F (24,38) = 15.882, p = 0.000 (All p from Hotelling’s trace).

In the broader area analyses, integrating fluorescence over distance increments (Table I), area 1 (0 ~ 25 μm), 50 and 50T were each significantly different from all other electrodes but the B and BT were not significantly different from each other. Encapsulation was greater in the 50 and 50T probes. In areas 2 ~ 6, only 50T was significantly different among all electrodes. In areas 7 and 8, 50T was significantly different from the other electrodes and 50 and B also were significantly different from each other. In summary, for glial encapsulation we found the floating 50 μm wire, and the floating and tethered BMEP were all largely equivalent in post hoc tests. The BMEPs were thus as low in gliosis response as a single floating 50 μm wire.

B. ED1 Staining

ED1 staining identified reactive microglia/macrophages in the vicinity of the probes. The images in figure 3(b) and (e) show ED1 stained cells in the vicinity of implanted electrodes and are paired with images in figure 3(a) and (d) respectively. Plots of relative fluorescent intensity of average ED1 staining (figure 4(b)) were again obtained as described in Methods. In many cases complete rings of ED1 staining were observed around both the untethered and tethered 50 μm wires, unlike the GFAP results. However, ED1 rings around the tethered 50 μm wires were thicker and sometimes the outer ring layers had oval or water droplet type shapes. The graph of tethered 50 μm wire ED1 response in figure 4(b) consequently has a peak level similar to the untethered 50 μm wire but also a broader peak. For both types of BMEPs, there was only partial ED1 fluorescence (figure 3(c) and (f), BT and B) across the rats. Accordingly, graphs of BMEPs response had lower mean levels. In figure 4(b), the order of decreasing EDI fluorescence intensity among electrodes was identical to the order found for the analysis for the GFAP fluorescence; tethered 50 μm wire had strongest and widest responses, untethered 50 μm wire was 2^nd, the tethered BMEP was 3^rd and the untethered BMEP showed the weakest ED1 fluorescence.

Two way ANOVA results are shown in Table I. The ANOVA results confirmed all electrodes were significantly different for peaks, peak distances and all areas values. The post hoc tests showed that the intensity peak of the untethered 50 μm wire was not significantly different from the tethered 50 μm wire, but in contrast it was significantly different from both the tethered and untethered BMEPs (p = 0.010 and 0.003). This matches the ordering in figure 4(b), with small differences between BMEPs and the tethered microwire. For peak distance, the tethered 50 μm wire differed significantly from both untethered and tethered BMEPs (p = 0.013 and 0.033), but not from the floating 50 μm wire. ED1 response was more localized and intense in the floating microwire, but the tethered microwire response was the broadest.

In area 1 (0 – 25 μm), closest to the electrodes, post hoc tests showed that the total intensity of the tethered 50 μm wire was significantly different from the others (50, p = 0,008; BT, p = 0.000; B, p = 0.000) and the untethered 50 μm wire differed significantly from both tethered 50 μm wire and untethered BMEP (B, p = 0.020). The tethered BMEP did not differ in ED1 intensity from the floating 50 μm wire in area 1. Beyond this, only the total intensity of tethered 50 μm was significantly different from the others. Thus, tethered BMEPs were superior to tethered 50 μm wires in ED1 response throughout, but equivalent to the floating 50 μm wire. The exception was the higher focused peak ED1 response of the floating 50 μm wire.

The animal variations in ED1 intensity were tested as a factor in a 2 way ANOVA, with significant differences for all areas. Post hoc tests revealed that compared with the GFAP between rat differences, ED1 differences between rats were more diverse and less specific to particular rats. Despite the animal variations, there were differences between electrodes. In ED1 reactivity BMEPs had better responses.

C. NeuN Staining

NeuN staining was used to identify surviving neurons. GFAP/NeuN double staining was performed. Sample images from each electrode are shown in figure 5(a). Neurons were not observed in the area heavily occupied by astrogliosis (e.g., figure 5(a), 50T). The grand averages of the normalized cell densities in 12 annular areas across 5 rats are plotted in figure 5(b).

Fig. 5.

GFAP/NeuN double staining images and normalized cell density graph from neuron counts, (a) Images captured in 4 different positions including each implant from the same tissue section with GFAP/Neun double staining. All colors are pseudo color: green for GFAP and red for neurons. Cell densities of images even from the same section can appear different due laminar variations between implant location differences, (b) Graph shows normalized cell densities in radial 12 annular sub-area averages. The red dotted horizontal line is the total average of the average cell densities in a reference area in the image far from electrodes defined as “reference”. Cell density was normalized by this reference value. The numbers at the top left corner in the graph are average numbers of neurons calculated within 100 × 100 pixels or 8365.6072 μm². The palest colored ribbon plots with dashed outlines represent the standard deviation and the darker narrower pale colored ribbon plots represent the standard errors. The abbrev. ‘50’ means the 50 μm wire, the letter ‘B’ stands for the BMEP, and the letter ‘T’ after 50 or B means “with tether”. Neural survival is highest in B (green) and then BT (magenta). The white bar in A represents 50 μm length.

Due to some tissue and section losses, data from 5 rats out of 8 were deemed sufficiently complete to be used for this analysis. From the graphs in figure 5(b), the order of decreasing cell density among electrodes was: (1) the untethered BMEP had highest cell density, (2) the tethered BMEP had second highest cell density, (3) the untethered 50 μm wire was in third place and (4) the tethered 50 μm wire had the lowest cell density. This order is consistent with the inverse of results for levels of GFAP and ED1 staining. Results from a 2 way ANOVA are shown in Table II. In subareas 1, 2, 3, 4 and 6, 8, 9, all electrodes were significantly different. The key issue for neural recording is neural density within 100 μm or less of the probe. The post hoc tests with Bonferroni correction showed that in Areal (0 – 25 μm), the tethered 50 μm wire was poorest and significantly different from both BMEPs (both p = 0.000) in normalized neuron density. In Area2 (26 – 50 μm), the tethered 50 μm wire remained poorer, but significantly different from only the untethered BMEP (p = 0.008) in normalized neuron density. However, in Area3 (51 – 75 μm), the tethered 50 μm wire was again significantly different from all the others (50, p = 0.025; B, p = 0.000; BT, p = 0.004). In addition, we found the untethered 50 μm wire was also poorer and significantly different from the untethered BMEP (p = 0.018) in normalized neuron density. In Area4 (76 – 100 μm), the tethered 50 μm wire remained significantly different from both BMEPs (B, p = 0.005; BT, p = 0.041). The loss of neurons around the probes was lowest in the untethered BMEP with ~ 50% loss within the first 25 μm, declining to ~ 25% within the next 25 μm of distance, and plateauing close to ~ 5 – 10% loss by 100 μm from the probe. The tethered BMEP showed corresponding losses of ~ 55%, reducing to ~ 25%, to a near plateau of ~ 10% by 100 μm. Other electrodes showed worse neural loss as note above.

TABLE II.

Two Way Anova Results For Neuron Counts

Factor	Area 1	Area 2	Area 3	Area 4	Area 5	Area 6	Area 7	Area 8	Area 9	Area 10	Area 11	Area 12
	0–25	26–50	51–75	76–100	101–125	126–150	151–175	176–200	201–225	226–250	251–275	276–300
Electrodes	0.000*	0.017*	0.000*	0.002*	0.053	0.010*	0.080	0.005*	0.009*	0.088	0.211	0.157
Rats	0.022*	0.006*	0.204	0.119	0.865	0.515	0.959	0.234	0.355	0.112	0.969	0.347

The area number means the position of segmented area by radial 25 pm increment from the electrode surface to 300 pm.

^*means the difference is significant and below p=0.05 after Bonferroni correction for multiple comparisons. Hotelling's trace N/A.

The results are also shown in Table II. The order of response of overall neural survival reflected in normalized neural density persisted as B>BT>50>50T within 100 μm of the probe.

IV. DISCUSSION

Our study demonstrates that BMEPs generate better immune responses and neural survival than control 50 μm diameter Nichrome wires, whether tethered or untethered when all are implanted in the same rats. The data may reflect both geometric strucmre and overall compliance effects in these differences. The material in every probe was identical (Nichrome with polyimide insulation), with similar modulus, but the member elements scales, geometry and overall compliance in BMEPs differed significantly from the 50 μm microwires. Size of the component braid elements, overall braided device compliance and the outer braided probe diameter all differed versus the control microwires. BMEPs had larger overall probe diameters since they were inserted over a stiff probe core which was then removed, but despite this showed significantly better tissue responses in the same animal in terms of gliosis, microglial effects, and neural survival close to the probe.

In terms of astrogliosis, 50 μm wires induced denser GFAP layers closer to the electrode than BMEPs, but BMEPs seemed to have slightly broader GFAP layers near the electrode surface. Sections suggested that astrogliosis didn’t wrap the contour surface of BMEPs completely (figure 5(a), BT) while it always did wrap 50 μm wires as a continuous ring. We think that the diamond-shaped open windows on the braid body may partly alleviate the astrogliosis. This idea is supported by two other studies in which microelectrode structures having lattices or open windows also showed less astrogliosis [24, 41]. A further difference between 50 μm wires and BMEPs was that astrogliosis for 50 μm wires could be much larger and extending beyond the ring of GFAP layers on the electrode surface. It could extend in one or two directions in a wing-like pattern (figure 5(a), 50T and 50). A wing-like long astrogliosis around 50um wires chronically implanted in rabbit cortex was also observed by others [42]. The anisotropic astrogliosis was most severe and thick with the tethered 50 μm wires. These anisotropies might be caused by interaction between the stiff body / tether and frequent tissue micromotions. The highly flexible tether and body of BMEPs likely exerted less tissue stressing effects of this type. In areas of heavy and thick astrogliosis, in some instances, we did not observe any neurons; any electrode that caused more astrogliosis likely also had less neurons in the same unit volume near the implant. This also matches our NeuN data. The recent report that the upper stiffer portion of Michigan like SMP (shape memory polymer) probes showed more astrogliosis and less neural density than the lower less stiff portion of the probe also corroborates our claim [43]. We cannot fully discount that there were variations in micromotion among rats in our study, although headpiece integrity generally seemed good at the time of rat retirement. Nonetheless, whatever the source of gliosis, the BMEPs performed better, and the body structure and tether compliance likely contributed. We observed all microgliosis was enclosed by astrogliosis, regardless of whether a partial or complete encapsulation occurred. Microgliosis always encapsulated the contour of untethered 50 μm wires. We speculate that there were two complete encapsulating layers around the 50 μm wires: an inner microgliosis encapsulation and an outer astrogliosis encapsulation. However, full inner microgliosis encapsulation was here never observed with either untethered or tethered BMEPs. Again, we believe that the flexible body and open structure of BMEPs may cause this difference. The deepest sections used in our analysis were obtained at about 1680 μm depth from the brain surface. The shallowest sections used here were at 60 μm. We could not find any trends of changes in immunoreaction along electrode length or with depth.

In our study we compared electrodes within individual rats. Although rats were a significant factor in the 2 way ANOVA for both GFAP and ED1, the post hoc test with Bonferroni correction revealed that the results from only 1 or 2 rats were consistently and significantly different among the 8 rats. Removal of extreme outlier rats identified in this way from the analysis were tested and did not alter any of the main results.

Our method for neural density per 100 × 100 pixels calculated the relative neuron density to the average cell density per 100 × 100 pixels of a whole section image which was estimated using the neuron number counted in a large area covering 30% ~ 40% of the whole image, but not including the implant site. These average cell density values as reference values were in the expected ballpark range when we compared them with the neural density calculated via published estimation techniques. We calculated an estimate of rat’s cortex neuron density in a unit volume using data in the literature from two reference methods: measuring brain tissue density using pycnometry and using the isotropic fractionator to estimate the neuron numbers per milligram with a chunk of brain tissue block [44, 45]. From the first method, we calculated that the rat brain tissue in 1mm³ volume could be roughly converted to 1.05 mg. From the isotropic fractionator table, we estimate that the average number of neurons in the rat cortex is 41.09 × 10³ / mg and the standard deviation is 8 × 10³ / mg. The estimated unit volume of our reference window is 91.46 μm × 91.46 μm × 30 μm = 2.509 × 10⁻⁴mm³ (100 × 100 pixels, 0.9146 μm per pixel, 30 μm tissue thickness) and converts into 2.634 × 10⁻⁴mg. Finally from these calculations we obtained an expected average 11 ± 2(11 ± 18%) neurons per unit volume. The average number of neurons in the equivalent reference window measured directly from this study was 16 neurons per unit volume. Our results showed 5 more neurons per unit volume (~ 20 – 45%) compared to the mean estimated value from the literature, but the data is within the ballpark of the estimated neuron number in rat cortex calculated from the isotropic fractionator tables. Further, we used proportional losses in our statistics, which were thus unlikely to be biased by our specific counting methods.

Gliosis and NeuN responses showed inverse patterns. Neural losses correlated to gliosis levels in our analyses. The most extensive anisotropic glial barrier surrounding an implanted tethered 50 μm wires that we observed extended almost 450 μm. The available recording distance, of 100 μm [46, 47] and even the theoretically available recording distance, of 130 μm [48][48][48] showed electrode relevant effects within these ranges. Stiff electrodes anchored to the skull may induce more inflammation and generate more gliosis due to frequent tissue micro- and macro-motions. In this study, the length of the tether between the anchor point and the tissue surface was about 1mm regardless of electrode types due to the applied 1 % agar gel dome. Generally increases in tether length would help both types of electrodes to follow tissue motions more freely, but according to our previous mechanical bending data, die shortest tether length (3 mm) had the maximum difference (97 times) of bending stiffness between BMEPs and the 50 μm wire and the longer tether (5 mm) had relatively less difference (44 times) between them [7]. The short tether length is good for the purpose of tissue response comparison between BMEPs and the 50 μm wire, not for better floating effects. Floating electrodes clearly partly overcome this tissue motion problem. However, we also found better responses of BMEPs than 50 μm wires even when both were without tethers. We do not here know the electrophysiological viability of neurons identified as surviving near the probes, but the BMEP designs clearly had significantly greater neural survival (up to 20%) within the critical area within 100 μm range from the probes. This suggests the body geometry and compliance in our BMEPs design structures further reduced biological responses and neural losses beyond just the reduced tether forces in BMEP probes. This happened despite the overall larger outer diameter of our BMEP structures and replicates the works of Kim et al. in non-porous implants [28] and Thelin et al. in 200 μm diameter stainless steel probe implants [30].

In this research, we did not explant BMEPs from brains before sectioning. We hoped to be able to observe any chronic tissue responses inside BMEPs lumen, an empty space with lots of open diffusion windows on the tubular braid wall as shown in Figure 1. Unfortunately, we could not see any effects inside BMEPs because the lumen content was not preserved after sectioning (see Figure 3 and 5, - the insides of all BMEPs were empty, much like all implant sites from explanted 50 μm wires).

In summary, results here showed that all BMEPs consisting of 12 × 9.6 μm Nichrome wires caused less tissue response than single 50 μm Nichrome wires, regardless of a tethered or untethered state. The untethered BMEPs also showed better immune response than the tethered BMEPs and the untethered 50 μm wires showed less tissue response than the tethered 50 μm wires. Significantly more neurons survive in the vicinity of implanted BMEPs than in the vicinity of implanted 50 μm wires.

V. Conclusion

These data strongly suggest that despite a strong material modulus mismatch between neural tissue and electrode build materials, an improved design geometry supporting higher probe mechanical compliance, better diffusion, and comprising many small spatial scale elements can result in significantly less immunoreaction and neural loss than occurs with a single and slightly larger spatial scale element that forms a stiffer overall electrode, even if both designs are untethered and floating in the tissue. The use of ultrafine wire structures arranged in precise geometries may support mechanically more robust (but high compliance) devices for insertion and handling that nonetheless minimize tissue inflammatory response and neural losses when in situ in the brain. Moreover, the gains in tissue response reduction and neural survival seen here can likely be improved further in future, by employing still smaller diameter component elements, or by using materials in component elements with tunable material modulus, such as shape memory polymers.