Rejuvenating silicon probes for acute electrophysiology

The work was supported by grants from the NIH (R01-EB028156, R37-NS128416). 14 15 Author contributions: A.M.S., M.A.F, and R.S. conceived and designed research; A.M.S, N.P., M.A.F, P.H., 16 and S.P.O. performed experiments; A.M.S., N.P., M.A.F., and P.H. analyzed data; A.M.S. and R.S. 17 interpreted results of experiments; A.M.S. and R.S. prepared figures; A.M.S and R.S. drafted manuscript; 18 A.M.S. and R.S. edited and revised manuscript; R.S. approved final version of manuscript. 19 20 Competing interests: None. 21 22 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 2 23

24 Electrophysiological recording with a new probe often yields better signal quality than with a used 25 probe. Why does the signal quality degrade after only a few experiments? Here, we considered silicon 26 probes in which the contacts are densely packed, and each is coated with a conductive polymer that 27 increases its surface area. We tested 12 Cambridge Neurotech silicon probes during 61 recording 28 sessions from the brain of 3 marmosets. Out of the box, each probe arrived with an electrodeposited 29 polymer coating on 64 gold contacts, and an impedance of around 50k Ohms. With repeated use, the 30 impedance increased and there was a corresponding decrease in the number of well-isolated neurons. 31 Imaging of the probes suggested that the reduction in signal quality was due to a gradual loss of the 32 polymer coating. To rejuvenate the probes, we first stripped the contacts, completely removing their 33 polymer coating, and then recoated them in a solution of 10 mM EDOT monomer with 32 uM PSS using 34 a current density of about 3mA/cm2 for 30 seconds. This recoating process not only returned probe 35 impedance to around 50k Ohms, it also yielded significantly improved signal quality during 36 neurophysiological recordings. Thus, insertion into the brain promoted loss of the polymer that coated 37 the contacts of the silicon probes. This led to degradation of signal quality, but recoating rejuvenated 38 the probes. 39 40 41 42 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 3

43 Our current ability to simultaneously record extracellular activities of neurons in the brain is due to the 44 development of high-density, multi-contact silicon probes. The large number of contacts makes it 45 possible to record extracellular voltages from not only multiple neurons, but also from multiple regions 46 of a single neuron, thus providing a pseudo image of the geometry of each neuron as well as the spike 47 interactions between neurons. 48 However, the high density spatial configuration dictates a small contact size (hundreds of 49 micrometers square area) (Ludwig et al., 2006), which results in high impedance. In theory, high 50 impedance can make it more difficult to record the microvolt-range activities of neurons (Loeb et al., 51 1995; Ludwig et al., 2006; Baranauskas et al., 2011), though the relationship between impedance and 52 recording quality is still unclear (Alba et al., 2015; Neto et al., 2018; Jones et al., 2020). Some probe 53 manufacturers have addressed the impedance issue by using electrodeposition to adhere a conductive 54 polymer on the surface of each contact, increasing the surface area and reducing the impedance (Cui et 55 al., 2001; Ferguson et al., 2009; Niederhoffer et al., 2023). Indeed, contacts with a polymer coating may 56 produce better electrophysiological recordings, but this approach introduces another problem: in acute 57 recordings the probe must be inserted and then retracted, which can remove the polymer coating, thus 58 producing a probe with contacts that have high impedance. The net effect is that a probe with a polymer 59 coating is often useful for only a handful of recordings before it is discarded. Given the high cost of these 60 probes, is there a way to rejuvenate them? 61 Here, we used 64-contact silicon probes (Cambridge Neurotech) and made impedance and 62 electrophysiological measurements during repeated in vivo recordings from the marmoset brain. We 63 found that with repeated use, the probe’s impedance increased, the spike magnitudes decreased, and 64 the number of contacts which failed to record any signals increased. Imaging of the probes suggested 65 that insertion and retraction promoted the removal of the polymer coating. This raised the idea that 66 recoating the contacts via electrodeposition might not only reduce the probe’s impedance, but also 67 restore its signal quality. 68 To test this idea, we first used an enzyme cleaner to strip the coating from all contacts, and then 69 recoated the contacts using electrodeposition. The results were rejuvenated probes that retained their 70 low impedance and produced high quality recordings that rivaled new probes. 71 72

73 PEDOT coating degraded with each recording session, increasing impedance and decreasing signal 74 quality 75 We used 64-contact silicon probes (M1 checkerboard and M2 linear Cambridge Neurotech) to acquire 76 neurophysiological data from three marmosets. Each session began with an impedance measurement of 77 the probe in saline (see Methods). Next, we inserted the probe through a craniotomy into the brain, 78 traversing the artificial dura (Duragel, Cambridge Neurotech), the intact dura, the visual cortex, and the 79 tentorium, to finally arrive in lobule VI and VII of the cerebellar vermis or the fastigial nucleus (Sedaghat-80 Nejad et al., 2019). The probe was inserted using a piezoelectric, high precision microdrive (0.5 μm 81 resolution) with an integrated absolute encoder (M3-LA-3.4– 15 Linear smart stage, New Scale 82 Technologies), using step size of 5 um, at a speed of 2.5 um/sec. Following 1-3 hours of 83 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 4 neurophysiological recordings, the probe was retracted at the same step size and speed. The typical 84 insertion depth was 4-9 mm depending on whether the target was the cerebellar cortex or nucleus. 85 Despite regular cleaning between recording sessions, there was a gradual increase in both the 86 magnitude and phase of the probe’s impedance (Fig. 1A, 1st and 2nd row, linear mixed effects model, 87 main effect of number of days of recording: 22% increase in impedance magnitude per day of recording, 88 t(2110) = 36.2, p = 1.46 e-223, 3.06-degree increase in impedance phase per day of recording, t(2110) = 89 35.1, p = 7.4 e-213). Each of the probes had 64 contacts. We characterized a contact as ‘bad’ when it had 90 an impedance magnitude of over 150k Ohms. The number of bad contacts increased with recording 91 sessions (Fig. 1A, third row, linear mixed effects model, main effect of number of days of recording: 4.6-92 contact increase in bad contacts per day of recording, t(31) = 2.89, p = 0.0069). 93 Impedance magnitude represents the resistive properties of the contact, where lower 94 magnitude indicates that charge can pass more easily. Impedance phase measures the capacitive 95 properties, where a perfect capacitor has a -90o phase (Macdonald, 1987). The increase in magnitude 96 and phase that we found with repeated use suggested that the charge-carrying capacity of the contacts 97 was decreasing. To understand the cause of these changes, we imaged the probes in their new state, 98 then again after each recording. Fig. 1C illustrates a new and a used probe, along with the impedance 99 magnitude of various contacts. On the new probe, the contacts were dark blue and uniform in color. 100 This indicated an intact PEDOT coating, matching the measured average impedance of 39k Ohms with a 101 standard deviation of .003k Ohms. The used probe had some contacts that were still well-coated (dark 102 blue, 37k Ohms and 57k Ohms), but also many contacts that were almost bare (164k Ohms, 306k Ohms, 103 1113k Ohms), exhibiting a light-blue or gold color. This suggested that some of the coating was lost 104 during the recording sessions. 105 This loss of the coating polymer appeared to affect both the quality and quantity of units that 106 we were able to record. Typically, the spike waveform of the sortable units on the first day of a probe 107 was much larger than the waveform on later recordings. For example, on probe 10460, the best unit on 108 Day 2 was 2-3 times larger than the best unit for the same probe on Day 7 (Fig. 1B). We examined the 109 data over days of recording by plotting spike magnitude as a function of impedance magnitude (Fig. 1D). 110 As impedance increased, spike magnitude decreased, and the percentage of contacts that had no units 111 increased (binned by magnitude of impedance into 4 groups. ANOVA on median spike magnitude per 112 contact for 976 contacts per group, f(3) = 7.46, p = 5.63e-5. Significance test for slope of percentage of 113 contacts with no units across 4 groups (top plot, red line), data assigned randomly to each of 4 groups to 114 create null distribution, repeating 400 times. Slope significantly different from null distribution with p < 115 1e-5, z-score = 9.4939. Thus, with repeated use, the probes lost some of their polymer coating, and this 116 coincided with increased impedance and a reduction in signal quality. 117 118 Each recording session deposited tissue on the probe, requiring cleaning that removed the tissue but not 119 the polymer coating 120 An additional challenge that we faced was that with each recording, there was an accumulation of 121 debris on the probe. Under the microscope the debris appeared as a fibrous tissue containing proteins, 122 dura-gel, and dust from the surrounding air (Fig. 2A, left image). Thus, we needed to clean the probes, 123 but there was a risk that the cleaning method would remove both the tissue and the polymer coating. 124 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 5 We tested two conventional methods (enzyme cleaner and Trypsin), as well as an 125 unconventional method of cleaning (hard-boiled egg). Surprisingly, cleaning with a hard-boiled egg 126 followed by Trypsin proved to be the most reliable method. We began by repeatedly inserting the probe 127 into a peeled, hard-boiled egg. This removed most of the tissue that had adhered to the probe (Fig 2A, 128 middle image). Next, we removed the remaining debris, and potentially any debris resulting from the 129 egg cleaning, using an overnight soak in Trypsin (Fig 2A, right image). Trypsin is a solution commonly 130 used to dissociate cells from each other during cell culture but is also used to clean biological matter 131 from probes (Tokiwa et al., 1979; Cui et al., 2001; Cui and Martin, 2003; Alba et al., 2015; Neto et al., 132 2016, 2018). We quantified the effectiveness of this method by measuring the magnitude and phase of 133 impedance, as well as the number of bad contacts (Fig. 2B), and observed that egg cleaning followed by 134 Trypsin reduced impedance magnitude and phase, as well as the number of bad contacts (linear mixed 135 effects model, main effect of egg cleaning then trypsin: 88% reduction in impedance magnitude, t(638) = 136 -21.1, p = 1.74 e-75, 8.2 degree reduction in impedance phase, t(638) = -29.8, p = 6.10 e-123, 33.4-contact 137 reduction in bad contacts, t(8) = -2.7, p = 0.027 ). 138 A critical question was whether this cleaning method harmed the polymer coating. Trypsin has 139 been documented as a safe cleaning method for silicon probes with PEDOT ((Scott et al., 2012; Alba et 140 al., 2015; Neto et al., 2016, 2018)), but what about the egg cleaning? In order to test if egg cleaning was 141 safe for the PEDOT coating, probes with a fresh PEDOT coating were subjected to 5 egg cleanings. Each 142 egg cleaning consisted of inserting the probe into the egg and moving up and down the shank with a 143 scrubbing motion. The scrubbing motion was repeated for approximately the same amount of time it 144 took to clean a dirty probe, an average of 15 times. This was repeated five times for each probe, 145 representing five days of cleaning between recording sessions. We found little evidence that egg 146 cleaning affected the PEDOT coating (Fig 2C, signed rank test, median increase of 5kOhm over 5 147 simulated egg cleanings, median absolute deviation of 3kOhm, p = 2.97 e-37). The slight increase in 148 impedance may have been due to deposition of egg particles on the probe, which was removed by 149 soaking the probe in Trypsin for an hour. Thus, the combination of hard-boiled egg cleaning and 150 subsequent soaking in Trypsin appeared to be an effective way to clean the probes. 151 The conventional method of cleaning probes, including tungsten and non-coated silicon probes, 152 is soaking in a solution of enzyme cleaner (Tergazyme, Alconox Inc.). This cleaner can remove proteins 153 and tissue without harming the base material ((Chen et al., 2023)). However, we observed adverse 154 effects using this method, finding that impedance magnitude increased greatly, with many contacts in 155 the mega Ohms range (Figure 3B), an indication of reduced quality of PEDOT coating. For example, on a 156 typical probe there was visible debris and fibers after removal from the brain (Fig. 3A). After an 157 overnight soak in 2% Tergazyme solution with stirring, the probe appeared to still have some debris. This 158 debris was removed with egg cleaning, which revealed that the contacts were gold color, indicating a 159 loss of the PEDOT coating. Therefore, Tergazyme enzyme cleaner removed the PEDOT coating. 160 In summary, enzyme cleaner was an inappropriate method for cleaning these probes because it 161 stripped the polymer coating, dramatically increasing the probe’s impedance. An effective method was 162 hard-boiled egg cleaning followed by Trypsin. 163 164 Recoating via electrodeposition 165 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 6 Despite regular cleaning, insertion into the brain led to gradual loss of the factory-deposited PEDOT 166 coating, resulting in reduced signal quality (Fig 1A). In order to restore the probes, the PEDOT coating 167 needed to be re-deposited. We approached this with electrodeposition, creating a mixture of PEDOT 168 and PSS and using the NanoZ Electroplate mode to plate each contact. However, prior to recoating, we 169 found it essential to remove all PEDOT coating. We did this using Tergazyme enzyme cleaner, and then 170 deposited new PEDOT onto the cleaned gold contacts. This method proved to rejuvenate the probes to 171 like-new condition. 172 Fig. 4A shows the probe following enzyme stripping, with clean gold contacts (left image). After 173 the probe was recoated with fresh PEDOT, the contacts took on a dark blue-black color, indicating an 174 even coating. The quality of recoating was tested with impedance measurements. Fig. 4B shows the 175 impedances of four probes across the rejuvenation process. These probes started with a median 176 impedance of about 150-200k Ohms, with large variance in impedance across the probes, indicating 177 uneven wear of the PEDOT coating. We next stripped the PEDOT coating with Tergazyme, leaving the 178 contacts clean of any debris and mostly clear of the original factory PEDOT. This greatly increased probe 179 impedance (linear mixed effects model, main effect of enzyme stripping: 483% increase in impedance 180 magnitude, t(510) = 24.0, p = 9.44 e-86). The probes were then recoated, resulting in a median 181 impedance at approximately 50k Ohms, which is as good or better than the impedance measurements 182 tested in the new probes. The phase component was restored as well, indicating that the recoated 183 probe had similar resistive and capacitive properties as a new probe. 184 Although impedance is an indicator of the PEDOT coating quality, we sought to test the 185 rejuvenation method via signal quality of electrophysiological recording. Despite confounding 186 differences such as location of recording in the brain and other factors that affect quality of recording, 187 we found that recoating indeed improved signal quality. Fig. 4C demonstrates data for a single probe on 188 a recording day immediately before and after the recoating procedure. The best unit from the recording 189 day before recoating, when the probe had been used for 9 recording sessions, had spikes that were 2-3 190 times smaller than the best unit from the first day after recoating. In four probes that we examined 191 before and after recoating, there was a 57.4 uV increase in average spike magnitude of isolated units 192 with recoating (Fig 4D left panel, linear mixed effects model, main effect of recoating: t(510) = 6.12, p = 193 1.88 e-9). In addition, the number of contacts with no isolated units decreased by 27% (Fig 4D, right 194 panel, ANOVA f(1) = 6.65, p = .0419). 195 In summary, recoating of the conductive polymer returned impedance magnitudes to around 196 50k Ohms and significantly improved signal quality. 197 198

that did not yield positive results 199 For cleaning, we first tried soaking the probe in Trypsin-EDTA overnight. This did not remove the bulk of 200 debris from the probe or greatly improve the impedance on the debris-obscured contacts. We also tried 201 changing the order of cleaning to first trypsin, and then hard-boiled egg cleaning. This method was less 202 effective than egg followed by Trypsin, potentially due to the mechanical agitation from the egg 203 loosening all of the debris to a state where any remaining debris could be removed chemically. We also 204 tried cleaning the probe mechanically, with similar technique to the hard-boiled egg method, in a cup of 205 either 1:1 ratio cured dura-gel (Cambridge Neurotech) or prepared 1.5%, 2%, 3%, or 4% laboratory-206 grade Agar (Innovating Science, Amazon). The dura-gel was sticky but not strong enough, and instead 207 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 7 deposited itself onto the probe in small chunks. The agar was not sticky enough, despite being almost 208 too hard to penetrate safely with the probe. 209 We also tried other methods for recoating the probes. Originally, we attempted plating over the 210 existing PEDOT coating in order to thicken it, only cleaning the probe beforehand with egg or trypsin, or 211 both. However, those methods resulted in uneven coating, with most contacts still in the high-212 impedance range. This suggests that the existing PEDOT either prevents new PEDOT from adhering to 213 the gold contact, or the existing PEDOT coating is dirty or damaged in a way that prevents appropriate 214 electroplating conditions at the site. 215 216

217 To better understand why the signal quality of some silicon probes degrades with use, we employed 218 Cambridge probes to record from the marmoset brain. On the one hand, insertion into the brain caused 219 a loss of the probe’s PEDOT coating, and on the other hand it deposited tissue on the probe, both of 220 which promoted increased impedance and reduced signal quality. Inserting the probe into a hard-boiled 221 egg followed by soaking in Trypsin was effective in removing the tissue. To replace the lost coating, we 222 first stripped the contacts using an enzyme cleaner, then recoated the probes via electroplating. This 223 cleaning, stripping, and recoating procedure rejuvenated the probes, returning their impedance to 224 around 50k Ohms. Critically, recoating dramatically improved the capacity of the used probes to record 225 electrophysiological signals from the brain. 226 227 Impedance is correlated to spike magnitude and the ability to isolate single units 228 While low impedance is generally a desirable property in recording electrodes, it is not clear that lower 229 impedance corresponds to improved neuronal signal quality. For example, (Neto et al., 2018) compared 230 bare iridium and PEDOT-PSS coated probes in an anesthetized preparation and found no significant 231 differences in acutely recorded spike magnitudes. This was despite significant differences in impedance 232 and noise level in saline. In contrast, (Baião, 2014) compared uncoated gold and PEDOT-coated gold 233 contacts in acute anesthetized recordings and found that the coated probes not only exhibited reduced 234 impedance, they also had better signal quality. (Cui and Martin, 2003) recorded acutely in anesthetized 235 animals and found larger signal amplitudes in the PEDOT coated contacts as compared to the uncoated 236 contacts, which shared the same significant difference in impedance between coating conditions. In 237 contrast, (Scott et al., 2012) did not find a signal quality difference in lower-impedance gold-coated 238 iridium probes, but did observe a noise reduction. Others have found that in chronic recordings, coated 239 probes produced a greater number of well-isolated units, but the relationship to impedance was unclear 240 because those measurements were taken while the probe was in the brain (Ludwig et al., 2006; 241 Venkatraman et al., 2009; Alba et al., 2015). Taken together, previous work had suggested that 242 uncoated electrodes had a higher innate impedance and lower ability to record units than PEDOT-243 coated electrodes. 244 A limitation of many of these studies is that they only compared coated and uncoated 245 electrodes, which usually differ in impedance by one or two orders of magnitude. There is no study 246 currently that directly correlates impedance on a spectrum from well-coated electrodes to a degraded 247 or absent coating. Here, we measured impedance and signal quality on each of 64 contacts in 12 probes 248 across 61 acute recording sessions and found that spike magnitude was inversely varied with impedance 249 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 8 (Fig 1D). As impedance increased, our ability to isolate single units greatly diminished. Moreover, 250 because recoating reduced impedance and improved signal quality, the loss of the coating and its 251 associated increase in impedance was likely a causal factor in the reduced signal quality. 252 253

254 Although we used the same charge density for recoating each electrode, and generally recoated the 255 probes to an impedance magnitude of about 50k Ohms, we were unable to directly measure coating 256 thickness. Coating thickness is one of the more important characteristics of conductive polymer coating, 257 and is directly correlated with impedance (Cui et al., 2001; Dijk et al., 2020a, 2020b; Niederhoffer et al., 258 2023). However, because coating thickness measurements are generally destructive, we did not employ 259 them in our experiments. 260 Our methods and conclusions are only directly applicable to PEDOT-PSS coating on passive gold 261 contacts for acute recordings in the brain. There is a vast literature on the use of other dopants besides 262 PSS, or other coatings entirely for various applications (Cui et al., 2001; Cui and Martin, 2003; Ferguson 263 et al., 2009; Ludwig et al., 2011; Scott et al., 2012; Baião, 2014; Alba et al., 2015; Carli et al., 2019; Wang 264 et al., 2021; Niederhoffer et al., 2023). However, PEDOT-PSS seems to be an effective and popular 265 conductive polymer coating (Cui and Martin, 2003; Venkatraman et al., 2009; Park et al., 2013; 266 Koutsouras et al., 2017; Neto et al., 2018; Pranti et al., 2018; Boehler et al., 2019; Dijk et al., 2020a, 267 2020b; Jones et al., 2020). 268 Our conclusions regarding signal quality were based on acute recordings, which necessarily 269 varied in trajectory and individual experimenter electrophysiology technique among other factors that 270 may influence recording quality or wear on the probe. Despite these confounding factors, within-probe 271 comparisons suggested that use promoted loss of the coating, and recoating improved signal quality. 272 Perhaps our most surprising result was the observation that inserting the probe into a hard-273 boiled egg was an effective way to remove some of the tissues that are deposited on the probes. We 274 conjecture that the efficacy is due to the springy but sticky nature of the protein matrix in the egg white, 275 and the way in which it gently but firmly glides along the probe shank. This method does have a 276 drawback in that it is mechanical, not chemical, and thus care must be taken to not break the probe. 277 278 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 9

279 All data were acquired using 64-contact large animal silicon probes (M1 checkerboard and M2 linear 280 Cambridge Neurotech). Neurophysiological data were collected from three marmosets (Callithrix 281 Jacchus, 2 male and 1 female, 350–370 g, between 6 and 8 yrs old, subjects Mirza, Ramon, and Charlie), 282 using methods described earlier (Sedaghat-Nejad et al., 2019). The marmosets were born and raised in a 283 colony that Prof. Xiaoqin Wang has maintained at the Johns Hopkins School of Medicine since 1996. The 284 procedures on the marmosets were approved by the Johns Hopkins University Animal Care and Use 285 Committee in compliance with the guidelines of the United States National Institutes of Health. 286 287 Impedance measurements 288 At the onset of each electrophysiological recording session, we measured the impedance of each 289 contact by inserting the probe in sterile saline at 1k Hz via an Intan RHD2000-series Data Acquisition 290 (DAQ) System using the OpenEphys (Siegle et al., 2017) Impedance Measurement utility. The electrode 291 was connected to a 64-channel head stage amplifier and digitizer (Intan Technologies). Additional 292 impedance measurements were acquired using a NanoZ impedance device (White Matter LLC.) at a 293 range of frequencies in 0.01M phosphate buffered saline solution (PBS, Millipore Sigma) with a Ag/AgCl 294 (3M KCl) reference electrode (BASi Research Products). We used a 3D-printed holder to maintain a 295 consistent distance between the probe and the reference electrode. If the test signal was clipped at any 296 point during the measurement, the measurement was repeated. Each measurement was an average of 297 40 sine wave cycles at various frequencies. 298 299 Electrophysiological recordings 300 Neurophysiological data was recorded acutely in awake, behaving marmosets as they performed a 301 targeted saccade task (Sedaghat-Nejad et al., 2019, 2022). The recordings were from lobules VI and VII 302 of the cerebellar cortex, or from the fastigial nucleus of the cerebellum. Sessions lasted between 1 and 3 303 hours of recording on any given day, not including penetration and retraction time which varied from 1 304 to 3 hours total. Thus, the total experiment time was 5-6 hours. Endpoint depth varied between 4 and 9 305 mm from the skull surface. The probes were inserted through a craniotomy that was covered with a 306 layer of Dura-Gel (Cambridge Neurotech) which protected the exposed dura and supported the probe 307 during penetration. We used a Narishige SM-11 stereotaxic micromanipulator to align and advance into 308 the brain until we reached the cerebellum, at which point we switched to a piezoelectric, high precision 309 microdrive (0.5 μm resolution) with an integrated absolute encoder (M3-LA-3.4-15 Linear smart stage, 310 New Scale Technologies) to advance the electrode to the final recording position. 311 We connected each probe to a 64-channel head stage amplifier and digitizer (Intan 312 Technologies), and then connected the head stage to the same Intan DAQ that was used for the 313 impedance measurement. Data were sampled at 30 kHz. We used OpenEphys for interfacing with the 314 RHD2000 system and recording of signals. The neurophysiological data were sorted either via P-sort 315 (Sedaghat-Nejad et al., 2021) or via Kilosort 2.0 (Pachitariu et al., 2016) and then curated using the 316 open-source toolbox Phy. 317 318 Cleaning method 319 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 10 During acute electrophysiology, we inserted the probe into the brain, recorded from neurons, and then 320 removed it. This led to deposition of tissue on the probes. To remove this tissue, we tried three 321

inserting the probe into a hard-boiled egg, washing it in a Trypsin solution, and washing it with 322 an enzyme cleaner. For egg cleaning, the probe was mounted onto a micromanipulator and repeatedly 323 inserted into a peeled hard-boiled egg. This process generally consisted of 5-20 cycles of insertion-324 removal, using a gentle ‘scrubbing’ up-down motion with the manipulator. This process was done 325 carefully to reduce bending of the shank. 70% alcohol was sprayed on the shank between insertions to 326 aid in removal of debris. After each insertion the probe was moved slightly to a new insertion point. 327 Cleanliness was checked between insertions using a microscope, and the process concluded when the 328 probe was visibly free of debris. 329 For Trypsin cleaning, the probe shank was submerged in Trypsin-EDTA 0.25% solution (Millipore 330 Sigma) overnight. In some cases, the solution was stirred during submersion, but that did not make an 331 appreciable difference as compared to non-stirred submersion. 332 For enzyme cleaning, the probe shank was submerged in freshly made 2% Tergazyme (Alconox 333 Inc.) dissolved in deionized (DI) water. The solution was stirred continuously over a 12 hour period. If 334 there was visible debris left on the probe, the probe was egg cleaned before taking any measurements. 335 This procedure was repeated once or twice, or until the magnitude of impedance on all contacts was 336 over 300k Ohm and the contacts appeared gold-colored (without PEDOT coating) under the microscope. 337 338 Recoating method 339 The probes were recoated using a degassed solution of 10 mM EDOT (3,4-Ethylenedioxythiophene, 340 Millipore Sigma) with 32 uM PSS (Poly(sodium 4-styrenesulfonate), Millipore Sigma) in DI water. The 341 NanoZ provided a current source, and utilizing the NanoZ software’s DC electroplating mode, we 342 recoated each contact with a current density of about 3mA/cm2 for 30 seconds. The deposition process 343 was repeated on each contact until the impedance dropped to approximately 50k Ohms, although a 344 majority of the contacts only needed one cycle of the process. 345 346 Data Analysis: spike waveform 347 The magnitude of a neuron’s spike waveform was determined by obtaining an average unit waveform 348 using Phy’s built-in function _get_mean_waveforms, which averages 100 random spike samples per unit 349 over a roughly 3 ms window. We then computed the peak-to-peak voltage in microvolts from the 350 maximum and minimum values of the average waveform. The value was taken from the contact where 351 the spike waveform was largest. 352 The best unit for a given day was determined as the largest average waveform size out of all 353 units except complex spikes. Complex spike units were excluded because they are unique to the 354 molecular and purkinje layers of the cerebellar cortex and have a characteristically large peak-to-peak 355 voltage when recording somatically that could bias the data towards having a higher value when 356 recording from specific depths or trajectories. During a given day of data collection, neurophysiology 357 was logged in ~30 minute durations to reduce file size, so the best unit was taken only from units 358 isolated in the final recording of the session, when the cells were most stable. 359 For a given day of acute neurophysiology, units were sorted for each ~30 minute recording. The 360 average unit size was calculated for units found on each recording. Then, the spike magnitude for each 361 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 11 contact was calculated as the median size of units found on the given contact for the whole day. If no 362 units were found on a given contact for any recording all day, the magnitude value for that contact is 363 zero. 364 365 366 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 12 367 368 Figure 1. The factory-deposited PEDOT coating degraded across recording sessions, leading to 369 increased impedance and decreased recording quality. A. Magnitude, phase, and number of bad 370 contacts (impedance magnitude > 150k Ohms) as a function of recording sessions. B. Representative raw 371 data recordings from the best neuron, Day 2 and Day 7. Spikes from the best unit are 2-3 times larger on 372 Day 2 as compared to Day 7. C. Microscope images from two probes. Top: new probe 11870, which had 373 an average impedance magnitude of 39k Ohm. The dark blue contacts have good PEDOT coating, 374 characteristic of a new probe. Bottom: used probe 10460, which had contacts of highly variable 375 impedance magnitude. The gold contacts have poor PEDOT coating with a corresponding high 376 magnitude of impedance. Many contacts are on a spectrum between these two extremes due to uneven 377 wear. D. Top: the spike magnitude of the signal recorded on a contact vs. the impedance of that contact. 378 The red trace is the percentage of contacts per bin that had no well-isolated units. Error bars indicate 379 variance, and are computed from bootstrapped data (n = 1000 simulated probes). Bottom: spike 380 magnitude as a function of impedance. Error bars are SEM. 381 382 383 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 13 384 Figure 2. Method for removing tissue that electrophysiological recordings deposited on the probes. 385 The objective was to find a method that removed tissue without also removing the PEDOT coating. A. 386 Image of a probe immediately after completion of recording, after inserting into a hardboiled egg, and 387 after soaking in Trypsin. B. The effects of egg cleaning and Trypsin on the impedance properties of the 388 various probes. C. Effects of egg cleaning on the PEDOT. Error bars are box and whisker plots, indicating 389 the median, upper and lower quartile, and upper and lower extreme. 390 391 392 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 14 393 394 Figure 3. Enzyme cleaning stripped the PEDOT coating. A. Image of a probe immediately after 395 completion of recording, after enzyme cleaning, and after egg cleaning. B. Effect of enzyme cleaning on 396 probe impedance. Error bars are box and whisker plots, indicating the median, upper and lower quartile, 397 and upper and lower extreme. 398 399 400 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 15 401 402 Figure 4. Recoating rejuvenated the probes. A. Image of a probe after enzyme cleaning, which stripped 403 the PEDOT coating, and after recoating. B. Impedance properties of the probes following recording, 404 enzyme cleaning, and recoating. C. Representative raw data from the best unit that was recorded on 405 Day 9, and the best unit that was recorded after the probe was recoated. D. Spike magnitude and 406 number of contacts with no units before and after recoating. Error bars are box and whisker plots, 407 indicating the median, upper and lower quartile, and upper and lower extreme. 408 409 410 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 16 Alba NA, Du ZJ, Catt KA, Kozai TDY, Cui XT (2015) In Vivo Electrochemical Analysis of a PEDOT/MWCNT 411 Neural Electrode Coating. Biosensors 5:618–646. 412 Baião P de AA (2014) Nanostructuring silicon probes via electrodeposition: characterization of electrode 413 coatings for acute in vivo neural recordings. Available at: 414 https://run.unl.pt/handle/10362/14666 [Accessed March 21, 2023]. 415 Baranauskas G, Maggiolini E, Castagnola E, Ansaldo A, Mazzoni A, Angotzi GN, Vato A, Ricci D, Panzeri S, 416 Fadiga L (2011) Carbon nanotube composite coating of neural microelectrodes preferentially 417 improves the multiunit signal-to-noise ratio. J Neural Eng 8:066013. 418 Boehler C, Aqrawe Z, Asplund M (2019) Applications of PEDOT in bioelectronic medicine. Bioelectron 419 Med 2:89–99. 420 Carli S, Bianchi M, Zucchini E, Di Lauro M, Prato M, Murgia M, Fadiga L, Biscarini F (2019) 421 Electrodeposited PEDOT:Nafion Composite for Neural Recording and Stimulation. Adv Healthc 422 Mater 8:1900765. 423 Chen FD, Chameh HM, Movahed M, Wahn H, Mu X, Ding P, Xue T, Straguzzi JN, Roszko DA, Sharma A, 424 Govdeli A, Jung Y, Chua H, Luo X, Lo PGQ, Valiante TA, Sacher WD, Poon JKS (2023) Implantable 425 nanophotonic neural probes for integrated patterned photostimulation and electrophysiology 426 recording. :2023.11.14.567101 Available at: 427 https://www.biorxiv.org/content/10.1101/2023.11.14.567101v1 [Accessed February 17, 2024]. 428 Cui X, Lee VA, Raphael Y, Wiler JA, Hetke JF, Anderson DJ, Martin DC (2001) Surface modification of 429 neural recording electrodes with conducting polymer/biomolecule blends. J Biomed Mater Res 430 56:261–272. 431 Cui X, Martin DC (2003) Electrochemical deposition and characterization of poly(3,4-432 ethylenedioxythiophene) on neural microelectrode arrays. Sens Actuators B Chem 89:92–102. 433 Dijk G, Ruigrok HJ, O’Connor RP (2020a) Influence of PEDOT:PSS Coating Thickness on the Performance 434 of Stimulation Electrodes. Adv Mater Interfaces 7:2000675. 435 Dijk G, Rutz AL, Malliaras GG (2020b) Stability of PEDOT:PSS-Coated Gold Electrodes in Cell Culture 436 Conditions. Adv Mater Technol 5:1900662. 437 Ferguson JE, Boldt C, Redish AD (2009) Creating low-impedance tetrodes by electroplating with 438 additives. Sens Actuators Phys 156:388–393. 439 Jones PD, Moskalyuk A, Barthold C, Gutöhrlein K, Heusel G, Schröppel B, Samba R, Giugliano M (2020) 440 Low-Impedance 3D PEDOT:PSS Ultramicroelectrodes. Front Neurosci 14 Available at: 441 https://www.frontiersin.org/articles/10.3389/fnins.2020.00405 [Accessed March 25, 2023]. 442 Koutsouras DA, Gkoupidenis P, Stolz C, Subramanian V, Malliaras GG, Martin DC (2017) Impedance 443 Spectroscopy of Spin-Cast and Electrochemically Deposited PEDOT:PSS Films on Microfabricated 444 Electrodes with Various Areas. ChemElectroChem 4:2321–2327. 445 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 17 Loeb GE, Peck RA, Martyniuk J (1995) Toward the ultimate metal microelectrode. J Neurosci Methods 446 63:175–183. 447 Ludwig KA, Langhals NB, Joseph MD, Richardson-Burns SM, Hendricks JL, Kipke DR (2011) Poly(3,4-448 ethylenedioxythiophene) (PEDOT) polymer coatings facilitate smaller neural recording 449 electrodes. J Neural Eng 8:014001. 450 Ludwig KA, Uram JD, Yang J, Martin DC, Kipke DR (2006) Chronic neural recordings using silicon 451 microelectrode arrays electrochemically deposited with a poly(3,4-ethylenedioxythiophene) 452 (PEDOT) film*. J Neural Eng 3:59. 453 Macdonald JR (1987) Impedance spectroscopy: emphasizing solid materials and systems. New York: 454 John Wiley & Sons. Available at: http://catdir.loc.gov/catdir/toc/onix01/86032582.html 455 [Accessed January 5, 2024]. 456 Neto JP, Baião P, Lopes G, Frazão J, Nogueira J, Fortunato E, Barquinha P, Kampff AR (2018) Does 457 Impedance Matter When Recording Spikes With Polytrodes? Front Neurosci 12 Available at: 458 https://www.frontiersin.org/articles/10.3389/fnins.2018.00715 [Accessed January 15, 2024]. 459 Neto JP, Lopes G, Frazão J, Nogueira J, Lacerda P, Baião P, Aarts A, Andrei A, Musa S, Fortunato E, 460 Barquinha P, Kampff AR (2016) Validating silicon polytrodes with paired juxtacellular recordings: 461

and dataset. J Neurophysiol 116:892–903. 462 Niederhoffer T, Vanhoestenberghe A, Lancashire HT (2023) Methods of poly(3,4)-463 ethylenedioxithiophene (PEDOT) electrodeposition on metal electrodes for neural stimulation 464 and recording. J Neural Eng 20. 465 Pachitariu M, Steinmetz NA, Kadir SN, Carandini M, Harris KD (2016) Fast and accurate spike sorting of 466 high-channel count probes with KiloSort. Adv Neural Inf Process Syst:4448–4456. 467 Park H-S, Ko S-J, Park J-S, Kim J, Song H-K (2013) Redox-active charge carriers of conducting polymers as 468 a tuner of conductivity and its potential window. Sci Rep 3:2454. 469 Pranti AS, Schander A, Bödecker A, Lang W (2018) PEDOT: PSS coating on gold microelectrodes with 470 excellent stability and high charge injection capacity for chronic neural interfaces. Sens 471 Actuators B Chem 275:382–393. 472 Scott KM, Du J, Lester HA, Masmanidis SC (2012) Variability of acute extracellular action potential 473 measurements with multisite silicon probes. J Neurosci Methods 211:22–30. 474 Sedaghat-Nejad E, Fakharian MA, Pi J, Hage P, Kojima Y, Soetedjo R, Ohmae S, Medina JF, Shadmehr R 475 (2021) P-sort: an open-source software for cerebellar neurophysiology. J Neurophysiol 476 126:1055–1075 Available at: https://journals.physiology.org/doi/full/10.1152/jn.00172.2021 477 [Accessed September 30, 2021]. 478 Sedaghat-Nejad E, Herzfeld DJ, Hage P, Karbasi K, Palin T, Wang X, Shadmehr R (2019) Behavioral 479 training of marmosets and electrophysiological recording from the cerebellum. JNeurophysiol 480 122:1502–1517. 481 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint 18 Sedaghat-Nejad E, Pi JS, Hage P, Fakharian MA, Shadmehr R (2022) Synchronous spiking of cerebellar 482 Purkinje cells during control of movements. Proc Natl Acad Sci 119:e2118954119 Available at: 483 https://www.pnas.org/doi/10.1073/pnas.2118954119 [Accessed March 30, 2022]. 484 Siegle JH, Lopez AC, Patel YA, Abramov K, Ohayon S, Voigts J (2017) Open Ephys: an open-source, plugin-485 based platform for multichannel electrophysiology. JNeural Eng 14:045003. 486 Tokiwa T, Hoshika T, Shiraishi M, Sato J (1979) Mechanism of cell dissociation with trypsin and EDTA. 487 Acta Med Okayama 33:1–4. 488 Venkatraman S, Hendricks J, Richardson-Burns S, Jan E, Martin D, Carmena JM (2009) PEDOT coated 489 microelectrode arrays for chronic neural recording and stimulation. In: 2009 4th International 490 IEEE/EMBS Conference on Neural Engineering, pp 383–386. 491 Wang A, Jung D, Lee D, Wang H (2021) Impedance Characterization and Modeling of Subcellular to 492 Micro-sized Electrodes with Varying Materials and PEDOT:PSS Coating for Bioelectrical 493 Interfaces. ACS Appl Electron Mater 3:5226–5239. 494 495 496 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 23, 2024. ; https://doi.org/10.1101/2024.02.20.581222doi: bioRxiv preprint