Journal of Neural Engineering PAPER • OPEN ACCESS You may also like - Misaligned Circumstellar Disks and Orbital Characterization of optogenetically-induced Motion of the Young Binary XZ Tau Takanori Ichikawa, Miyu Kido, Daisuke cortical spreading depression in awake mice using Takaishi et al. - The subtraction of mutually displaced graphene micro-transistor arrays Gaussian Schell-model beams J Carlos G de Sande, Massimo Santarsiero, Gemma Piquero et al. To cite this article: Eduard Masvidal-Codina et al 2021 J. Neural Eng. 18 055002 - Orbital Evolution of Binary Black Holes in Active Galactic Nucleus Disks: A Disk Channel for Binary Black Hole Mergers? Ya-Ping Li, Adam M. Dempsey, Shengtai Li et al. View the article online for updates and enhancements. This content was downloaded from IP address 158.109.56.184 on 07/01/2022 at 18:19 J. Neural Eng. 18 (2021) 055002 https://doi.org/10.1088/1741-2552/abecf3 Journal of Neural Engineering PAPER Characterization of optogenetically-induced cortical spreading OPEN ACCESS depression in awake mice using graphene micro-transistor arrays RECEIVED 30 November 2020 Eduard Masvidal-Codina1,2,9, Trevor M Smith3,9, Daman Rathore3, Yunan Gao3, Xavi Illa1,2, REVISED Elisabet Prats-Alfonso1,2, Elena Del Corro5, Andrea Bonaccini Calia5, Gemma Rius1, 8 February 2021 Iñigo Martin-Fernandez1,7, Christoph Guger4, Patrick Reitner4, Rosa Villa1,2, Jose A Garrido5,6, ACCEPTED FOR PUBLICATION 9 March 2021 Anton Guimerà-Brunet1,2,∗ and Rob C Wykes3,8,∗ 1 PUBLISHED Institut de Microelectrònica de Barcelona, IMB-CNM (CSIC), Esfera UAB, Bellaterra 08193, Spain 2 6 April 2021 Centro de Investigación Biomédica en Red en Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN), Madrid 28029, Spain 3 Department of Clinical and Experimental Epilepsy, Queen Square Institute of Neurology, University College London, London WC1N Original content from 3BG, United Kingdom 4 this work may be used g.tec medical engineering GmbH, Guger Technologies OG, 8020 Graz, Austria under the terms of the 5 Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and The Barcelona Institute of Science and Technology (BIST), Creative Commons Campus UAB, Bellaterra, Barcelona 08193, Spain Attribution 4.0 licence. 6 ICREA, Barcelona 08010, Spain Any further distribution 7 Universitat Autònoma de Barcelona, Bellaterra 08193, Spain of this work must 8 maintain attribution to Nanomedicine Lab, Faculty of Biology Medicine and Geoffrey Jefferson Brain Research Centre, University of Manchester, Manchester the author(s) and the title M13 9PT, United Kingdom of the work, journal 9 These authors contributed equally to this work. citation and DOI. ∗ Authors to whom any correspondence should be addressed. E-mail: anton.guimera@imb-cnm.csic.es and r.wykes@ucl.ac.uk Keywords: graphene, electrophysiology, spreading depression, optogenetics Supplementary material for this article is available online Abstract Objective. The development of experimental methodology utilizing graphene micro-transistor arrays to facilitate and advance translational research into cortical spreading depression (CSD) in the awake brain. Approach. CSDs were reliably induced in awake nontransgenic mice using optogenetic methods. High-fidelity DC-coupled electrophysiological mapping of propagating CSDs was obtained using flexible arrays of graphene soultion-gated field-effect transistors (gSGFETs). Main results. Viral vectors targetted channelrhopsin expression in neurons of the motor cortex resulting in a transduction volume ⩾1 mm3 . 5–10 s of continous blue light stimulation induced CSD that propagated across the cortex at a velocity of 3.0 ± 0.1 mm min−1 . Graphene micro-transistor arrays enabled high-density mapping of infraslow activity correlated with neuronal activity suppression across multiple frequency bands during both CSD initiation and propagation. Localized differences in the CSD waveform could be detected and categorized into distinct clusters demonstrating the spatial resolution advantages of DC-coupled recordings. We exploited the reliable and repeatable induction of CSDs using this preparation to perform proof-of-principle pharmacological interrogation studies using NMDA antagonists. MK801 (3 mg kg−1 ) suppressed CSD induction and propagation, an effect mirrored, albeit transiently, by ketamine (15 mg kg−1 ), thus demonstrating this models’ applicability as a preclinical drug screening platform. Finally, we report that CSDs could be detected through the skull using graphene micro-transistors, highlighting additional advantages and future applications of this technology. Significance. CSD is thought to contribute to the pathophysiology of several neurological diseases. CSD research will benefit from technological advances that permit high density electrophysiological mapping of the CSD waveform and propagation across the cortex. We report an in vivo assay that permits minimally invasive optogenetic induction, combined with multichannel DC-coupled recordings enabled by gSGFETs in the awake brain. Adoption of this © 2021 The Author(s). Published by IOP Publishing Ltd J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al technological approach could facilitate and transform preclinical investigations of CSD in disease relevant models. 1. Introduction glass micropipettes with silver/silver-chloride elec- trodes (Masvidal-codina et al 2019). Spreading depolarization is the largest disruption of Existing experimental models to induce CSD brain homeostasis possible in living neural tissue. It often rely upon direct chemical (potassium chloride), is characterized by an abrupt, near-complete break- mechanical (pinprick), or electrical stimulation of the down of the transmembrane ion gradients and high- cortex. The CSD threshold is difficult to determine amplitude (tens of millivolts) negative extracellu- using mechanical stimulation and the above meth- lar potential shift that persists for tens of seconds ods can all result in tissue damage and injury (Ayata (Dreier 2011). Cortical spreading depolarization is 2013). Repeated trials may result in cumulative tis- usually accompanied by temporary suppression of sue damage affecting reproducibility and reliability, brain activity, although in some severe conditions, as well as altering thresholds required for CSD induc- where tissue is compromised, spreading depolariza- tion. Noninvasive optogenetic induction of CSD has tions can trigger epileptiform activity (Kramer et al recently been reported using transgenic expression of 2017). We use the term cortical spreading depression channelrhodopsin (Houben et al 2017, Chung et al (CSD) in this manuscript to describe a propagating 2019). We have adapted this approach using viral vec- wave of cortical depolarization that is accompanied tors to target opsin expression unilaterally in neur- by transient suppression of neuronal activity. CSDs ons of the motor cortex (MC), and establish that this contribute to the pathophysiology of several neuro- approach can be used to reliably and repeatedly trig- logical diseases, including stroke (Dreier et al 2018), ger CSD on demand in awake mice. epilepsy (Aiba and Noebels 2015) and migraine with Therapies targeting CSD may potentially treat aura (Ayata 2010, Charles and Baca 2013) where the a range of neurological conditions. However, many characteristics of the propagating wave, the duration drugs identified from preclinical studies result in of neuronal suppression and the degree of neural translational failure (Ikonomidou and Turski 2002). injury are determined by the local conditions of the In vivo drug discovery approaches have often been tissue (Pietrobon and Moskowitz 2014). performed under anesthesia. As general anesthesia Given the importance of CSDs to various neur- may impact on spreading depolarization frequency, ological diseases, the development of novel tech- propagation speed and sensitivity to pharmacolo- nology that permits the electrophysiological detec- gical suppression (Kudo et al 2008), an in vivo, tion of CSDs (both the spreading depolarization anesthesia-free assay, which permits repeated, repro- and the subsequent depression of neural activ- ducible rounds of CSD, will greatly improve the trans- ity), with high spatiotemporal resolution, and its latability of pre-clinical findings. Current evidence application to studying these events in the awake points to the essential role of NMDA receptors in CSD brain is warranted. Experimentally, CSD is tradition- generation and propagation (Kohr 2007). Therefore, ally electrographically recorded using solution-filled to determine the feasibility of our model as a drug glass micropipettes with Ag/AgCl wires, which lim- screening platform we investigated the efficacy and its the spatial resolution to just a few point meas- modulation of CSD by NMDA antagonists. urements. Microelectrode grids improve the spa- To avoid confounds associated with a craniotomy tial resolution but suffer from polarization-induced required for epidural array placement, we investigated drift and signal attenuation causing distortion of a minimally invasive approach to measure CSDs elec- the measured signal (Nelson et al 2008). As an trographically in awake rodents. We report the prac- alternative to the commonly used passive elec- ticability of recording CSD through a mouse skull trode technology, active transducers based on tran- using gSGFET transistor arrays. sistors can offer significant advantages in electro- Overall, this study establishes the feasibility and physiology (Khodagholy et al 2013). In particu- advantages of applying graphene micro-transistor lar, graphene solution-gated field-effect transistors technology to characterize CSD induced in the awake (gSGFETs) demonstrate a comparable signal-to-noise brain as well as the repetitiveness and reliability of ratio (SNR) to platinum black electrodes in typical optogenetic CSD induction. The application of this local field potential (LFP) frequencies (Blaschke et al in vivo approach in disease-relevant models, includ- 2017, Hebert et al 2018). Additionally, due to excel- ing familial hemiplegic migraine and ischemic stroke, lent stability in aqueous solutions, gSGFETs offer the will allow characterization of CSDs induced by the possibility of wide-bandwidth electrographic record- underlying pathology, the ability to trigger additional ings including DC-shifts and infraslow oscillations CSDs on demand, evaluate pharmacological inter- that are superior to those recorded by passive grids ventions, and record localized CSD waveforms and of electrodes and comparable to those obtained using propagation patterns in the awake brain. 2 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al 2. Materials and methods support screw (00-96X3-32, Plastics One, USA). Using vetbond (1469SB, 3M, USA), the headplate 2.1. Animal preparation (Model 9, Neurotar, Finland) was firmly attached Animal experiments were conducted in accordance and strengthened using dental cement before Kwik- with the United Kingdom Animal (Scientific Proced- cast (W.P.I., UK) covered the exposed skull. Mice ures) Act 1986, and approved by the Home Office were checked daily to ensure recovery. Habituation (license PPL70-13691). C57Bl/mice (38 males and 6 was performed by placing the mouse in the Neur- females) were bred and the majority (n = 33) were otar mobile-home cage for increasing periods of time used at the ages of ∼3–6 months (140.6 ± 4.65 d), (15–60 min) over several days the week prior to crani- while a subset (n = 11) were used at an age of otomy surgery. ∼7–9 months (250.0 ± 7.259 d), due to the clos- ure of the lab resulting from the Coronavirus pan- 2.3. Surgeries for craniotomy demic (p = <0.001, ***; unpaired T-Test). Anim- Due to the lab closure as a result of Covid19 the als were housed on a 12 h/12 h dark/light cycle, and amount of time cortical neurons expressed ChR2 food and water were given ad libitum. Animals were prior to optogenetic experiments varied. Pre-Covid, group housed until their headbar attachment surgery the average length of time from viral injection to and were then subsequently individually housed. For experiment was 59.61 ± 3.142 d, n = 33 animals, all procedures, animals were weighed before being and post-Covid this was extended to 175.9 ± 2.959 d, anaesthetized in a chamber using isoflurane (100% n = 11 animals (p = <0.001, ***; unpaired T-Test). w/w; 988-3245, Henry Schein, USA) and placed There was no difference in the intensity of light in a stereotaxic frame (David Kopf Instruments required to induce a CSD (threshold) between the two Ltd, USA). Viscotears (Bausch + Lomb, USA) were groups, 34.26 ± 7.971 mJ, n = 33 animals pre-Covid applied to each eye and both pain relief, consisting verses 33.15 ± 15.76 mJ, n = 11 animals post-Covid, of Buprenorphine (0.5 mg kg−1 ; subcutaneous; Ceva, (p = 0.255; Mann–Whitney unpaired T-Test). France) and Metacam (15 mg kg−1 ; subcutaneous; On the day of recording, craniotomies were Boehringer Ingelheim, Germany) and ∼0.5 ml of performed. Three areas were exposed: (a) a large saline, and dexamethasone (1 mg kg−1 ; subcu- (2 × 2 mm) craniotomy over somatosensory and taneous; MSD Animal Health, USA) were given. All visual cortex on the right hand side, (b) the skull surgical procedures were performed using sterilized was thinned or removed over the MC ipsilateral to tools and aseptic techniques. At the end of the surgical the main craniotomy, and (c) a small drill hole over procedure, warmed saline (0.15 ml/10 g) was admin- the MC on the left-hand side (contralateral to main istered subcutaneously. Mice were checked daily and craniotomy) (figure 1(B)). After completion, exposed recovered for at least a week before experiments com- dura was covered with cortex buffered saline, steril- menced, with the exception of the craniotomy sur- ized sylgard 184 (∼200 µm thickness), and a kwik- gery, where recording occurred on the same day, cast layer. After ∼2–4 h post-recovery, the animal was ∼2–4 h post-surgery. moved to the Neurotar frame and the craniotomies were exposed by removal of the kwickcast and sylgard. 2.2. Surgeries for viral injection and headbar attachment The skin on top of the skull was cut away. The skull 2.4. Electrophysiological recordings was cleaned using a delicate bone scraper (10075- Electrophysiological recordings were performed 16; Fine Science Tools, Germany) and brief topical using flexible gSGFETs arrays. The transistor arrays application of hydrogen peroxide (6%). Small bur- were carefully connected to a PCB and lowered holes were drilled in the following two positions onto the dura or skull using a micromanipulator. from bregma: (a) antero-posterior: +2.2 mm; lat- Whereas most currently available electrodes are pass- eral: +1.0 mm and (b) antero-posterior: +0.75 mm; ive, gSGFETs are active devices that transduce local lateral: +1.6 mm. To express channelrodopsin 2 voltage changes to current and permit a wide record- (ChR2), high titer (1 × 1013 vg ml−1 ) AAV9-hSyn- ing bandwidth (Masvidal-codina et al 2019). A cus- ChR2-EYFP virus (AAV9_hSyn_hCHR2(H134R)- tom g.HIamp biosignal amplifier, (g.RAPHENE, g.tec EYFP; 26973, Addgene, USA) was diluted 1:1 in 0.9% medical engineering GmbH Austria) was used for sig- NaCl and delivered using a 5 µl syringe (Model 95 nal acquisition at 9600 kHz and 24 Bit. The system Hamilton Company, Switzerland) with a 35-gauge enables simultaneous recording in two frequency needle. At both sites, the needle tip was lowered from bands with different gains preventing amplifier sat- pia mater until 500 µm deep. Using a microinjec- uration. Prior to recording, I ds − V gs curves were tion pump (WPI Ltd, USA), 500 nl was injected at obtained at the start and end of each experiment 100 nl min−1 to achieve expression in both layer 5 to allow determination of the performance of the and layer 2/3 pyramidal neurons. To apply the head- gSGFET and custom code was used to calculate the bars for the Neurotar system a small hole in the skull optimal bias point (average across the arrays of the over the left visual cortex was drilled for a metal maximum of the mean absolute transconductance 3 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Figure 1. Description and characterization of optogenetic-induced CSD in awake head-fixed mice. (A) Timeline of the experiment, indicating the range of time for viral expression and the habituation trials (three) to the head-fixed apparatus prior to experimentation. (B) Vectorized image showing an awake mouse in the Neurotar mobile homecage head fixation system during a habituation training session. (C) Sagittal schematic of the experimental arrangement, indicating ChR2 expression in frontal motor cortex areas (M1 and M2) and the epidural positioning more posterior of a 16-channel transistor array over somatosensory cortex (SC). (Di and Dii) (Di) Top view schematic showing the ChR2 expression areas activated by optogenetic stimulation (Opto.), gSGFET placement, and the position of the reference electrode (Ref.) in the contralateral motor cortex. (Dii) Photograph of a 4 × 4, 16-channel gSGFET neural array placed epidurally over the somatosensory cortex for recordings. (E) Schematic of the recording and signal calibration process used. gSGFETs are biased in common gate mode at a V ds = 50 mV and an optimal V gs determined from an I ds − V gs curve measured once the array is in position during the acute in vivo experiment. Extracellular signals are recorded as current variations flowing through the graphene transistor. Recovery of the original extracellular voltage signals (black trace; full bandwidth data) is obtained by interpolation of the current signal (red) into the I ds − V gs curve of the graphene sensor element. (Fi–iii) Representative data from one transistor showing DC-coupled (top; black), high-pass filtered 1–40 Hz (middle; black) and z-scored power spectra of 1–300 Hz (bottom) before, during and after illumination with continuous blue light (10 s) (Fi), continuous green light (10 s) (Fii), or pulsed blue light (20 Hz) (Fiii). For the z-scored power spectra plots, z-score is calculated for each frequency with respect to the before (−100 s to −10 s) LED stimulation epoch. Color scale goes from +2 (red) to −2 (blue) standard deviations. (G) Quantification and statistical comparison of evoked CSDs success in response to continuous blue light illumination (n = 27 animals), continuous green light (n = 23 animals) and pulsed blue light (n = 9 animals). Fisher’s exact test was used to compare between groups. left from charge neutrality point). Subsequent inter- (section 2.9, figures 8 and 9) and larger craniotomies polation of acquired current signals into the transfer were performed as required for the protocol. curve results in DC-coupled voltage signals. A reference wire (Ag/AgCl2 ) was placed in the 2.5. Determination of light threshold for contralateral MC. An LED cannula (400 µm dia- optogenetically-induced CSD and array calibration meter, CFM14L20, Thor Labs, NJ, USA) was lowered Prior to experimentation the LED current-to-power via a manipulator over right-hand side MC. For some relationship was calibrated using a power meter experiments (n = 8), two transistor arrays were used (PM16-130, Thor Labs, NJ, USA), and mW values 4 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Figure 2. Representative example and quantification of AAV9-ChR2 transduction volume. Ai and Aii: (Ai) representative example of cross sections through a mouse brain expressing AAV9-hSyn-ChR2-EYFP with the distance (mm) from bregma. (Aii) Image of a mouse brain showing the approximate neuroanatomical location of the slices used for analysis in Ai. (B) Higher magnification fluorescent image showing expression of AAV9-hSyn-ChR2-EYFP in M2 and M1. (C) Quantification of area of viral transduction plotted as the average area calculated per 250 µm blocks of tissue from the front of the brain until expression decreases beyond bregma. Black line illustrates the transduction volume extracted from animals where optogenetically induced CSDs could be repeatedly and reliably evoked using low light power, n = 5 mice. The red line illustrates the transduction volume from mice where an optogenetically induced CSD was not possible n = 2, or not consistently evoked n = 1. (D) Plot between transduction volume (mm3 ) and energy (mJ) required (black), or tested (red) for optogenetic induction of CSDs, n = eight mice. The dashed line indicates 1 mm3 . The two filled red dots represent animals with a transduction volume calculated to be less than 1 mm3 and it was not possible to induce a light-induced CSD. The open red circle represents data from an animal where an optogentically induced CSD could be evoked but not consistently, even at maximum light power. were multiplied by the duration (seconds) of stim- was obtained from four mice due to hardware or soft- ulation to obtain a measurement of light power in ware problems during experimentation. mJ. After 10 min of baseline recording, the light- threshold required to induce a CSD was determined 2.7. Comparison of CSD induction methods using continuous 5 or 10 s stimulations of blue light CSD was induced in the MC by the following three (470 nm; M470F3, Thors Labs, NJ, USA). Controls methods: (a) optogenetic (b) pinprick of a 35 gauge were performed using the same power intensity and Nanofil injection needle (NV33BV-2, World Preci- duration with a green LED (595 nm; M595F2, Thors sion Instruments, USA) attached to a Nanofil Injec- Labs, NJ, USA) or by using pulse blue light (20 Hz, tion system (UMP3T-1, SMARTtouch Microinjec- 50% duty cycle) with a combined light-on duration tion System, World Precision Instruments, USA) or identical to the square pulsed light duration (5 or (c) injection of 1000 nL, at a rate of 100 nl s−1 , of 1M 10 s), (figure 1(G)). KCl into the superficial layers of the cortex (∼500 µm deep from pia; using aforementioned injection sys- tem). For this protocol, the order above was used each 2.6. Inclusion criteria time, due to the trauma caused from the pinprick, 63 mice were prepared for this study. 44 mice (38 which was required for the local injection of KCL. males, 6 females) showed reproducible induction of CSD from light stimulation and were included in this 2.8. Pharmacological screen assay manuscript. The remaining 19 mice were excluded The lowest LED current found to induce a CSD for the following reasons: eight animals because the twice was multiplied by 1.5 to ensure that optogen- initial light-evoked CSD was followed by at least etic induction was supra-threshold throughout the one spontaneous CSD. Post-mortem analysis of these experiment. The inter-induction interval was set to animals revealed lesions through subcortical (motor) 10 min to allow full recovery of neuronal activity white matter tracts, indicative of accidental damage (figure 3(D)). Two baseline light-induced CSDs were during surgery (data not shown). Four animals were followed by a randomized I.P. administration of a excluded as it was not possible to evoke a light induced substance, either: vehicle (Saline; 0.5 ml), MK801 CSD and three animals were excluded due to an inab- (3 mg kg−1 ; 10009019, Cayman Chemical, Michigan, ility to trigger a repeatable CSD every trial. Post hoc USA), or ketamine (15 mg kg−1 , Dechra, UK). A analysis of these brains revealed a marked reduction minimum of six post-injection light stimuli were in transduction volume ⩽1 mm3 (figure 2). No data delivered. Analysis of neural activity was conducted 5 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Figure 3. Waveform properties and frequency band analysis of optogenetically-induced spreading depression in awake mice. (A) Example of a CSD waveform recorded by an individual graphene transistor, highlighting the parameters extracted including duration, amplitude, latency from delivery of LED light, area under the curve (AUC), prehyperpolarization (PreHyper) and recovery hyperpolarization (RecHyper). (Bi–iv) Histograms, with a line showing the population estimate, of CSD parameters for amplitude (Bi), Area Under the Curve (Bii), duration (Biii) and propagation velocity (Biv). Bi–iii includes a total of 633 CSD waveforms recorded from 54 CSDs events (two CSD events induced in n = 27 mice), while Biv is derived from the same 54 CSD events. (C) Extracellular voltage depolarization (black line) and filtered voltage of different neural frequency bands (Delta 1–4 Hz, theta 4–8 Hz, alpha 8–12 Hz, beta 12–30 Hz, gamma 30–100 Hz) during a CSD event recorded by a gSGFET. Vertical grey lines indicate middle point of the epochs: before light stimulus (bf), immediately after (aft), at onset (ons), at maximum depolarization (dep), and every minute after the end of the depolarization (r1 − r7). The vertical scale bar y-axis is 3 mV for the DC coupled trace (black); for the other frequency bands is as follows: delta: 0.6 mV, theta: 0.4 mV, alpha: 0.2 mV; beta: 0.3 mV; gamma: 0.15 mV. (D) Average change in the frequency band root mean square (rms) voltage normalized with respect to the bf. timepoint. There is an immediate and progressive silencing for delta-theta band and an immediate increase in gamma band rms after light stimulus. Data from the same CSD events shown in panel B, grouped by mice (n = 27). post injection of NMDA antagonists. One animal with subdural recordings (n = 5; figure 9). Here, the injected with ketamine showed no change post- craniotomy for on dura placement was narrower and injection in power (theta), similar to saline injec- over the precise neuroanatomical area of interest. ted animals, and in contrast to all other animals injected with either MK801 (n = 5) and ketamine 2.10. Transduction volume calculations (n = 3), and was excluded from the study due to The experiment was terminated using a sodium concerns that the i.p injection had failed, and the pentobarbital overdose (i.p. injection). Brains were drug did not reach the brain. To check for a differ- removed and placed into 4% PFA (J19943.K2; ence in the stimulation thresholds between treatment Alfa Aesar, UK) chilled at 4◦ overnight. Coronal groups we compared the OptoThreshold and this was sections of 50 µm thickness were cut using a vibrat- not significantly different between treatment groups ing microtome (Leica, Wetzlar, Germany). Sections (Vehicle, 13.18 ± 4.99 mJ; MK-801, 18.00 ± 4.99 mJ; were mounted using Vectashied (2B Scientific Ltd). ketamine, 4.55 ± 2.17 mJ) (p = 0.0591, Kruskal– Tiling images were acquired using a Zeiss Axio Wallis ANOVA). Observer inverted widefield microscope and Zen microscope software (ZEISS), with 10× objective 2.9. Assays using dual gSGFETs at 1024 × 1024 pixels resolution. All images were To explore the neurophysiology at the site of induc- acquired using the same microscope parameters for tion of the CSD, experiments were performed with analysis purposes. Cortical fluorescence distribu- arrays over both the MC and somatosensory/visual tion in coronal slices was quantified using ImageJ. cortex (n = 3; figure 8). Here, a larger craniotomy The cortex was manually selected as the region of was completed over the MC to facilitate the place- interest and an intensity threshold, consistent across ment of the gSGFET. Some experiments were con- animals applied. Pixels above this value were coun- ducted to compare recording CSD through the skull ted with total area in mm2 calculated based on 6 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al pixel dimensions and magnification. Using the Allen defined as the minimum value of the wave and repor- Mouse Brain Atlas, each slice was binned based on ted as an absolute value. CSD duration was defined distance from bregma. The transduced area within as the time that the extracellular voltage remained each slice was taken as a representation of the sur- below −4 mV; it was determined using below and rounding posterior 250 µm of tissue. The total trans- above thresholds. AUC was defined as the area below duction volume (mm3 ) was calculated as the summa- zero voltage during the duration of the depolariza- tion of each 250 µm volume. tion determined. CSD propagation speed was estim- ated as the distance between the LED induction site 2.11. gSGFET arrays and the middle of the transistor array (4 mm) divided Flexible epidural neural probes (10 µm in thickness) by the mean latency CSD onset. Code available upon containing an array of 16 graphene micro-transistors request. (4 × 4 array, 400 µm separation) with active areas of either 150 × 100 µm2, 100 × 100 µm2 or 50 × 50 µm2 2.14. CSD wave propagation direction were used in this study. gSGFET arrays were fab- To determine the direction of propagation, a vector ricated at the clean room facilities of IMB-CNM as pointing from the lowest to highest latency values reported in (Masvidal-codina et al 2019). In brief, in the transistor array matrix was used. A threshold single-layer graphene was grown by chemical vapor of 2.5 standard deviations from the mean of the deposition and transferred to a silicon wafer previ- latencies was set to remove outliers that could dis- ously coated with a polyimide layer and patterned tort the direction of propagation. The angles between metal traces. After defining the graphene channels the propagation vector of the first induced CSD and and before evaporating a second metal layer, UVO the subsequent ones was determined by ∠ (x, y) = treatment was applied to improve the graphene-metal arccos xx·,yy ., where x and y are the two vectors. interface and reduce its contact resistance (Schaefer et al 2020). Finally, SU-8 was used as passivation 2.15. CSD band analysis layer and the polyimide layer was etched to define the For each CSD extracellular voltage shift, the calib- geometry of the neural probes. Devices were gently rated raw signals were either not filtered for full-band peeled off from the wafer and inserted to zero inser- or bandpass filtered for the different bands (delta: tion force connectors for electronic interfacing. 1–4 Hz, theta: 4–8 Hz, alpha: 8–12 Hz, beta: 12–30 Hz and gamma: 30–100 Hz). Then, the root mean 2.12. Data analysis squared (RMS) was calculated using a sliding window Recorded signals were analyzed using Python 3.7 (window_size: 0.5 s, steptime = window_size/10) and packages (Matplotlib 3.2.0, Numpy 1.17.4, Pan- resulting traces were passed through a median filter das 0.25.3, Seaborn 0.9.0, Neo 0.8.0) and the cus- (scipy.signal.medfilt) to remove artefacts. RMS val- tom library PhyREC (PhyREC4, https://github.com/ ues for each band were averaged at given epochs with aguimera/PhyREC). Transistor recordings were calib- respect to the LED stimulation (referred to as 0 s), rated by interpolating the current signals into the cor- the depolarization onset (t_onset, defined as the time responding branch of the in vivo measured transfer extracellular voltage goes below −4 mV of the raw curve as reported by Masvidal-codina et al (2019). For signal) and the depolarization end (t_end, defined two experiments the signal interpolation to the trans- as the time the extracellular voltage crossed again fer curve was not possible due to insufficient meas- the −4 mV). The epochs are: before the LED stim- ured voltage range or presence of artifacts, for these ulus (bf.: −100 s to −10 s), a few seconds after (aft.: experiments the gain at the V gs bias point of the trans- 0 s to t_onset − 10 s), depolarization onset (onset: fer curve for each transistor was used. No high pass t_onset − 10 s to t_onset + 10 s), depolarization filtering is applied to the recorded extracellular poten- (dep.: t_onset + 10 s to t_end) and recovery 1–7 tials for the analysis of CSD waveforms, extraction of being every minute after the end of the depolariza- CSD parameters or plotting to avoid any signal distor- tion (r1: t_end to t_end + 60 s; r2: t_end + 60 s to tion. Error bars in all plots represent standard error. t_end + 120 s, etc). To evaluate the neural effects of CSD, RMS power values where normalized with 2.13. CSD parameter extraction respect to the bf. timepoint. Values for each transistor For each CSD extracellular voltage shift the following in the array over multiple CSD events were averaged parameters were extracted: latency, peak amplitude, to yield a single value for each experiment. duration and area under the curve (AUC). Raw sig- nals were down sampled to 3 Hz and zero voltage was 2.16. CSD shape clustering set to the average value of the 5 s before the light stim- To classify the different shapes of the depolarization ulus. We defined the onset of the CSD as the onset of waves, raw signals were resampled at 10 Hz. Signals the negative shift with a threshold at −4 mV. Latency from −30 to 400 s around each light-stimulus were was calculated as the difference between the starting extracted and zero voltage was set to the average value stimulation time and CSD onset. Peak amplitude was of the 30 s before the light stimulus. A threshold of 7 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Figure 4. Electrophysiological comparison of gender and method of induction for CSD in awake mice. (A)–(C) Schematics (i) and raw traces (ii) showing CSDs induced using different modalities: optogenetic (A), pinprick (B), or potassium chloride (1 µl at 1 M) injection (C). The traces come from the same animal. (D) Comparison of the amplitude (Di), AUC (Dii), and propagation velocity (Diii) of CSD from induction by opto (blue), pinprick (red) and KCl (grey) and, also, gender with females (closed squares) and males (open squares). No statistical difference was found between genders, or between waveform characteristics of CSDs induced by different methods. (E) Normalized, to the before CSD timepoint, rms band changes during CSD for the delta (Ei) and theta (Eii) frequency bands displayed. n = five male and five females for opto and KCl, and four male and four females for pinprick. F or E, statistics used a mixed-model effects analysis with Sidak’s post-hoc tests. Asterisk are color-coded depending on the CSD-induction type that is compared against KCl. −4 mV was used for CSD detection. Then, all detec- DP (PreHyper <1); SPh (RecHyper >1) and others ted waves were time aligned to the minimum of the were classified as SP, LV if amplitude was lower than (first) depolarization peak (0 s) and epochs from −50 −6 mV or ND if a depolarization was not detected at to 140 s extracted. Depolarization voltage is defined that channel. as the mean voltage value between −1 and 1 s. Then, three features where selected to perform the cluster- 2.17. Statistical analysis and experimental design ing: integral of the prehyperpolarization (area >0 mV In these studies, the variable of interest was isol- in the −15 s to 0 s), integral of recovery hyperpolariz- ated and statistically examined. In general, for com- ation (area >0 mV in the 0s to 75 s) and the second to parisons between groups of two, unpaired or paired first peak amplitude ratio. The second peak was detec- T-tests were used, while for three or more groups, ted during the 100 s after the first peak by finding the ANOVA with post hoc corrections were utilized. To valley (with the minimum of the first derivative) and examine the frequency of evoked CSDs for different then searching from the minimum of voltage after optogenetic stimulations (figure 1), since the same the valley. Based on these three features waveforms animals were used in a non-randomized fashion for were clustered by defining a threshold for each group: each modality, Fisher’s exact test was used to com- first they were classified as DP (second to first peak pare between groups. For comparing the paramet- amplitude ratio >0) or SP. Then DPh (PreHyper >1); ers of CSDs to different modalities (figure 4), each 8 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Figure 5. Classification of distinct CSD waveforms recorded using 16-channel gSGFETs as CSD propagates through cortex. (A) 14-channels from a gSGFET array recording an optogenetically induced CSD. Different waveforms can be observed across the grid. These can be categorized into four clusters (panel B) and color-coded. (B) CSD waveforms for each of the four clusters DP (red), DPh (blue), SPh (green), SP (purple). DP = double-peak; SP: single-peak, h = hyperpolarization. Bold lines in black highlight an example case for each group. (C) 3D scatterplot of the three features used for clustering (F1: prehyperpolarization, F2: recovery posthyperpolarization and F3: second to first peak ratio). Dots are colored to indicate the clusters, as described in B, with orange showing low voltage (<−6 mV). (D) Pie chart showing the frequency of the four different clusters, with colors the same as in B and C. Orange indicates the percentage of nondetectable waveforms. (Ei and Eii) Density plots of the distribution of duration (i) and amplitude (ii) extracted from the waveforms for each cluster. (F) and (G) Rms changes, normalized to baseline, during CSD in the delta (F) and theta (G) frequency bands. All four CSD types are graphed, with grey lines showing combined single and black line combined double peaks. ANOVA with Sidak’s post hoc tests were performed. animal, either male or female, was given the same measures ANOVA with Sidak’s post-hoc test to exam- sequential treatment (due to the nature of induction ine the differences between responses over the 10 s methods), with at least a single CSD being evoked of optogenetic stimulation and up until recovery of to each modality. With this data collected, a one-way the voltage. To examine electrophysiological record- ANOVA was run to compare between (a) genders, and ings between epidural and skull positioning of the (b) groups. As genders showed no differences between gSGFET, the two arrays were placed and optogen- induction using either optogenetics, pin-prick from a etic stimulation was applied in the MC (figure 9). As needle, or injection of KCl, the data from each gender the CSD propagated across the brain, it was recorded was combined to allow for a more informed compar- at both sites. SNR was calculated as the ratio of the ison between modality types. Examining the change peak CSD amplitude to the standard deviation of the in normalized frequency bands between the different 30 s prior to CSD induction. To compare the SNR, an CSD shapes (figure 5) utilized the two-way repeated artificial SNR level was set to 3, with slight variation measures ANOVA with Sidak’s post-hoc comparis- around this, and then a one-way ANOVA with Tukey’s ons for both (a) sub-types and (b) combined sub- post-hoc test was used to compare this regularly used types. For the pharmacological study (figure 7) para- artificial SNR level to the SNR from epidural and skull meters (Amplitude), a two-way repeated measures gGFET recordings. Next, for other parameters such ANOVA was used with Sidak post-hoc comparisons. as amplitude, AUC, and propagation velocity (pro. Examining the time response profile for recording at vel.), Welch’s, unpaired, T tests were used to exam- the site of induction also utilized a two-way repeated ine whether there were differences between recording 9 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al sites. Finally, two-way ANOVA was used to compare recorded extracellular voltage lasting tens of seconds normalized frequency band activity at the two record- (46.0 ± 1.2 s) in duration were recorded. Data cor- ing sites. Throughout the manuscript standard sig- responding to 633 CSD waveforms recorded from nificance classifications were used (* = p < 0.05; a total of 54 CSDs (2 CSDs evoked in each of ** = p < 0.01; *** = p < 0.001). 27 animals). However, as the histograms indicate (figure 3(B)), there was a large degree of variability in 3. Results these parameters, the 25%–75% range for amplitude and duration were 9.8–16.4 mV, and 28.9–50.2 s 3.1. Optogenetic CSD model in awake head-fixed respectively. This variability reflects the heterogen- mice eity in CSD waveforms recorded (see figure 5). In The experimental paradigm for triggering light- contrast, propagation speed was more consistent, induced CSDs in awake mice is illustrated in 3.0 ± 0.1 mm min−1 (2.8–3.3 mm min−1 , 25%–75% figure 1(A). Surgery was performed to inject viral vec- range). Light-evoked CSDs in the MC also induced tor to transduce neurons in the MC and express light- transient changes in LFP activity. For these exper- sensitive channelrhodopsin (ChR2) (see section 2, iments we recorded CSDs in areas of propagation, figure 2). Head bars were attached and in the week somatosensory and visual cortex (figures 1(C) and prior to experimentation mice were habituated (D)). Within seconds of light stimulation, a gradual to the head-fixation apparatus over several days decrease in 1–30 Hz RMS power could be detected, (figure 1(B)). To detect CSD, a 16-channel gSGFET minutes before the actual CSD invaded the area of the was placed on the dura over the somatosensory and brain underneath the transistors. In contrast, gamma visual cortex posterior to the ChR2-injected area RMS power (30–100 Hz) increased just prior to CSD (illustrated in figures 1(C) and (D)). Before record- invasion, indicative of intense neuronal firing pre- ing, a transfer curve is obtained that is used to determ- ceding the massive depolarization and subsequent ine the optimal recording settings and also for post- neuronal silencing (figures 3(C) and (D)) similar to recording current-to-voltage conversion (figure 1(E) an increase in multi-unit activity previously repor- and section 2). ted (Houben et al 2017). The rate of recovery of Continuous 488 nm blue light illumination neuronal activity post CSD were similar for frequen- (5–10 s) reliably induced CSD (figure 1(Fi)). Note cies 1–30 Hz with full recovery evident 4 min after the correlated activity suppression shown in the the end of the DC-shift. In some recordings there high-pass filtered signal and in the corresponding was a prominent increase in gamma RMS power z-scored spectrogram. To examine the specificity of 30–100 Hz upon recovery, indicative of rebound light wavelength and stimulation parameters we used excitation (figures 3(D) and 5(D)). a green LED 594 nm (figure 1(Fii)) with the same light duration and power intensity that successfully 3.2. Comparison with more invasive CSD evoked a CSD with blue light, or used pulsed blue induction methodologies light (20 Hz) with a total light-on duration identical The electrophysiological features of optogenetically- to the square pulsed light duration (figure 1(Fiii)). induced CSDs in virally-injected mice were com- Whereas constant blue light illumination always pared to standard induction methodologies including evoked a CSD, pulsed blue light or continuous green direct application of KCl to the brain and pinprick light rarely evoked a CSD (figure 1(G)). This indicates in the same animal (figures 4(A)–(C)). CSD wave- that a sustained neuronal depolarization, caused by form parameters (i.e. peak amplitude, duration and a single, seconds-long square pulse stimulation with AUC values were not statistically different between blue light in ChR2-expressing areas is an effective and CSD induction types, nor between male and female robust mechanism of CSD induction. mice. These values (figure 4(D)) are within previ- Examination of viral transduction volume post ously reported ranges for KCl-induced CSDs in non- hoc revealed that expression of cortical tissue, greater virally-injected, wild-type mice (Eikermann-Haerter than 1 mm3 , was required to express ChR2 to reli- et al 2015). However, although there is no difference ably induce CSD optogenetically (figure 2). Viral in the duration or AUC of CSDs between groups, we injections were variable in the total volume of tissue did observe a difference in the time taken for activ- transduced (5.46 ± 1.7 mm3 n = five animals) but ity suppression in frequencies above 1 Hz to fully consistently transduced the area of MC from +0.5 recover (figure 4(E)). Compared to optogenetic- and to +2.5 mm (anteroposterior coordinates relative to pinprick-induced CSDs, recovery from KCl-mediated bregma) where the optic fiber light was positioned CSD was significantly slower (figure 4(E)). during experiments. The light intensity required to trigger a CSD ranged 5.89–147.6 mJ, with mean of 3.3. CSD waveform analysis 33.98 ± 7.065 mJ (n = 44 animals). Due to the mapping capabilities of the gSGFET CSD parameters were extracted from indi- arrays, we observed variability in CSD amplitudes vidual micro-transistors (figure 3(A)). High amp- and duration (figure 3(B)), resulting from distinct litude (14 ± 0.3 mV) negative shifts in the waveform patterns. Figure 5(A) displays the CSD 10 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Figure 6. Reproducible and stable CSDs recorded using 16-channel gSGFETs in response to repetitive optogenetic induction. (A) DC-coupled recordings of 14 individual transistors in an array showing reliable, repetitive induction of CSDs triggered by blue light illumination over a ∼100 min recording session, with eight CSD inductions, indicated by the light blue bars. (B) Left: schematic indicating the array and fiber optic canula location. CSDs propagated from the motor cortex towards the visual cortex. Color-coded spatial map of the 16-channel gSGFET array based on CSD waveform at each transistor location. Arrow indicates CSD propagation direction. The angular variability of propagation varied little between repetitive induction of CSD. (C) Polar histogram of CSD wave propagation direction, n = 48 CSDs, from six mice. (Di) Mean frequency of CSD shapes recorded by individual channels. Although there was an increase in SP and a decrease in SPh (purple and green lines respectively), the ratio of single to double peak CSDs varied little during eight repetitions. (Dii) Combining the single (grey) and double (black) peaks. White area indicates the percentage of channels that did not record CSD waveforms. waveform captured by 14 individual transistors as each). DP and DPh were less common (10%), 5% a CSD propagated over the array. Some wave- each. Double peak waveforms (DP and DPh) had forms have a single peak, whereas others present a higher duration and amplitude than single peak a double peak. An additional feature is the pres- waveforms (SP, SPh) (figure 5(E)) and interestingly, ence of either a pre- or post-CSD hyperpolariza- also a faster rate of depolarization (data not shown). tion. We obtained a database of 2323 CSDs wave- DPh was the cluster with the highest values in all para- forms recorded from 224 optogenetically-induced meters tested. We performed frequency band ana- CSD events from 26 mice. We analyzed three wave- lysis on the waveform types and observed that double form features pre-hyperpolarization (F1), recovery peak waveforms caused a higher percentage of activ- post-hyperpolarization (F2) and second-to-first peak ity suppression compared to single peak waveforms amplitude ratio (F3) to classify the recorded wave- (figures 5(F)–(G)). The arrays were predominately forms into 4 different clusters: DP (double-peak), placed in areas of propagation. In these locations no DPh (double-peak hyperpolarization), SPh (single- spatial distributional trends across the grid for spe- peak), SPh (single-peak hyperpolarization) and a 5th cific waveforms were identified. category where the CSD was of low amplitude (<6 mV We examined the reproducibility of repeated peak), (figures 5(B)–(E) and section 2). The high optogenetic stimulation (figure 6(A)). Separation channel number and low-invasive induction meth- between light stimulations was set to 10 min to allow odology of our preparation allowed us to study the recovery from transient silencing of neural activ- occurrence and spatial distribution of each waveform. ity (figure 3(D)). Figure 6(A) displays 14 transistor Under our recording conditions, using ‘normal’, non- recordings from one animal in response to eight disease model mice, single peak accounted for 84% of optogenetic stimulations. In contrast to AC-coupled, waveforms split similarly between SP and SPh (42% or DC-coupled passive electrodes which encounter 11 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al baseline drift and require either high-pass filter- 3.5. Investigation of CSD induction ing or electronic off-set readjustments (Nelson et al To examine whether MK-801 and ketamine preven- 2008, Nasretdinov et al 2017), gSGFETs are able to ted induction, as well as propagation of CSDs, we record repeated rounds of CSD over several hours modified the experimental protocol and used two with minimal baseline drift (0.16 ± 0.9 mV/10 min, epidural arrays; one at the site of LED illumination n = six animals), highlighting the superiority of this (gFET#1), motor cortex, and the other, (gFET#2), methodology for long (hours) continuous record- over somatosensory/visual cortex (figure 8(A)). LED ings (figure 6(A)). We analyzed the propagation pat- illumination of the ChR2 expressing area resulted in terns of CSD across the transistor array (figure 6(B)), an immediate depolarization, most prominently on and categorized the CSD waveform recorded from the top-left corner of the transistor array (gFET#1). each transistor for repeated trials. Overall CSD wave- This immediate depolarization is not seen in gFET#2, form remained stable with roughly the same per- ∼3 mm away from the illumination site (figures 8(B) centage of events classified as single peak and double and (C)). CSDs induced in the MC propagated and peak, although SPh waveforms decreased and SP were detected later by the second transistor array waveforms increased during repeated stimulations gFET#2. After administration of MK-801 (3 mg kg−1 , (figure 6(D)), and there was a trend towards more n = 2), or ketamine (15 mg kg−1 , n = 1), no CSD double peaks by stimulation 7 and 8. Figure 6(C) was detected at gFET#2 in accordance with the single- shows an angular histogram of the difference between transistor array experiments (figure 7); while at the the propagation direction of any CSD and the direc- induction site (MC) a neuronal depolarization can tion of the first induced CSD of that experiment. Res- still be observed (figure 8(D)). Light stimulation res- ults indicate very small changes in the propagation ults in a time-locked depolarization of ∼5 mV, which angle over repeated optogenetic stimulations. then transitions into a larger amplitude depolariza- tion and full CSD (figure 8(D)). In contrast, after 3.4. Pharmacological suppression of CSD in awake administration of MK-801 or ketamine the initial mice light-induced depolarization returns quickly post- To test the effect of drugs capable of suppressing or light stimulus to baseline values without induction of modulating CSDs we increased the light stimulus to a CSD (figure 8(E)). 1.5 times the threshold, OptoThreshold, to ensure supra-threshold stimulation across the experiment 3.6. Through-skull detection of CSD duration. The protocol consists of two CSD induc- Finally, we investigated the possibility of recording tions prior to intraperitoneal systemic injection of CSDs electrographically through an intact skull. To vehicle or drug followed by an additional six supra- do this and to quantify our results with epidural threshold stimulations post-drug (figure 7(A)). In placement of arrays, we utilized two gSGFET arrays this study, we evaluated the efficacy of two NMDA in a single experiment, one placed over the motor/ receptor antagonists. A representative trace of the somatosensory cortex, and the other over recorded signal corresponding to one transistor of the ipsilateral somatosensory/visual cortex and the 16-channel array is shown for an animal in the optogenetically-induced CSDs (figure 9(A)). One saline, MK-801 and ketamine groups (figure 7(A)). of the arrays was placed epidurally and the other To validate the methodology for testing drug effects on top of the skull. There is a reduction in spati- we used MK-801, a well reported CSD-inhibitor otemporal properties of recorded CSDs across the (Chung et al 2019). Intraperitoneal (i.p.) injection array (figure 9(Eii)). The resolution in propagation of 3 mg kg−1 MK-801 produced rapid (<10 min) and CSD waveform across and between individual suppression of CSDs that did not appear in any of transistors achieved by subdural placement of the the subsequent light stimulations (n = 5). MK-801 gSGFET is lost. There is also a reduction in amp- cannot be used therapeutically due to a number of litude of the detected CSDs when recording through serious adverse side-effects, so we next turned our bone (epidural was 17.14 ± 2.66 mV, through skull attention to a potentially more clinically relevant was 9.75 ± 1.57 mV; n = six animals; p = 0.0436, NMDA antagonist, ketamine. Rodent studies indic- Welch’s unpaired T-Test) (figure 9(Cii)). However, ate that ketamine can reduce the amplitude of CSDs CSDs can be detected through the skull travelling at (Krüger et al 1999), or block CSD induction com- a propagation velocity of 3.198 ± 0.265 mm min−1 pletely (Hernándéz-cáceres et al 1987, Rashidy-pour (figure 9(Civ)). Additionally, we could detect sup- et al 1995). The effects are however transient, last- pression of higher frequency (1–30 Hz) rms amp- ing 20–45 min after a single injection. An effect litude correlated with the CSD, although this suppres- we also observed in our studies (figure 7); com- sion recovered faster compared to epidural recordings plete inhibition only lasted 15–25 min post-injection, (figure 9(D)). In order to exclude potential elec- although the CSD amplitude and AUC was smaller trical paths through burr holes and craniotomies, we (figures 7(B) and (D)) and propagation was slower performed non-invasive optogenetic induction and (figure 7(E)) for the remaining trials compared to recording of CSD (supplementary figure 1 (available pre-drug values. online at stacks.iop.org/JNE/18/055002/mmedia)). 12 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Figure 7. Pharmacological sensitivity of optogenetically induced CSD to NMDA antagonists. (A) Representative DC-coupled recordings from a single transistor from an animal injected with either saline (blue), MK-801 (red) or ketamine (brown). The blue and magenta bars indicate the time of light stimulus and i.p. injection, respectively. (B) Peak amplitude of CSDs from repetitive induction of CSDs in the presence of vehicle, MK-801 and ketamine. n (mice) = six vehicle, five MK-801, and three ketamine. Absence of CSD was considered as 0 mV amplitude. (C) Binary visualization of the presence or absence of CSDs at each experimental timepoint showing the variability in the time and duration of the effects. Each line indicates an experiment and is color-coded by treatment. (D) and (E) (i) Using only datapoints and treatment groups where a CSD was present, a graphical representation of area under the curve (D) and propagation velocity (E) for vehicle and ketamine. For ketamine, stimulations 3, 4 and 5 did not induce CSDs in all animals in this group and so the n numbers for successful CSD induction are indicated underneath the data, in brown. (ii) Datapoints for AUC and propagation velocity when CSD could be detected in all animals (timepoints 6,7 and 8) were analyzed by normalization to the average of pre-drug tests (stimulations 1 and 2) to illustrate that although CSDs had returned, they propagated slower and did not recover pre-injection parameters within the time course of this experiment (55 min post drug injection). Statistics. Two-way repeated measures ANOVA with Sidak’s post hoc tests to examine differences between specific stimulations. These experiments demonstrate that CSDs can be invasive manner. Although cerebral blood flow tech- electrographically recorded through a completely niques can be a useful complementary method cap- intact mouse skull using gSGFETs. able of recording the downstream effects of CSD on blood flow; the gold standard for CSD detection 4. Discussion requires electrophysiological methods (Ayata 2013), usually recorded using solution-filled glass pipettes A difficulty with studying CSDs, either independ- with Ag/AgCl2 wires. These are not practical for ently or concurrently with higher frequency activ- chronic recordings, nor for high-density mapping of ity, is the lack of appropriate tools to record them CSDs. Using passive electrodes, even with the use in vivo with high spatiotemporal fidelity, and to of DC-coupled amplifiers is sub-optimal (Fabricius reliably, on-demand, induce CSDs in a minimally et al 2008). In contrast, gSGFETs are active devices 13 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Figure 8. NMDA receptor antagonism blocks CSD induction and propagation. (A) Cross sectional schematic showing the positioning of both epidural gSGFETs, over either the motor or somatosensory cortex, and the LED fiber optic for optogenetic activation of ChR2-expressing neurons (green). (B) and (C) Waveforms recorded by the two epidural arrays for an example CSD event with electrophysiological traces both before (black) and after (red) i.p. injection with MK-801 (3 mg kg−1 ; NMDA receptor antagonist). In this example, the LED stimulus was located top-left of motor cortex array. (D) The top-left channel of the gSGFET over the motor cortex, showing the light-induced neuronal depolarization before and after administration of MK-801 with dashed lines indicating the timepoints (0, 1, 3, 10, 15, 20, 25 s from LED onset) used for analysis. (E) Absolute amplitudes taken from the listed timepoints from the onset of LED stimulation (0, 1, 3, 10, 15, 20, 25 s) and grouped by before and after drug. n numbers consisted of combined data from NMDA receptor antagonists that inhibited CSDs in the somatosensory cortex (MK-801 at 3 mg kg−1 (n = two mice) and ketamine at 15 mg kg−1 (n = one mouse)). Statistics. two-way repeated measures ANOVA with Sidak’s post hoc tests to examine differences at specific time points. that dynamically transduce LFP changes into cur- CSD resulting in tissue damage, problems with repro- rent (Blaschke et al 2017, Hebert et al 2018) allowing ducibility, and therefore limited translational value. multisite DC-coupled electrophysiological recordings Recently transgenic mice expressing ChR2 under the with negligible baseline drift (Masvidal-codina et al Thy-1 promoter have been used to non-invasively 2019), see figure 6(A). Experimental models of CSD induce CSD optogenetically via depolarization of pre- often employ highly invasive mechanisms to induce a dominately layer 5 excitatory neurons in the cortex 14 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Figure 9. Comparing recordings of Optogenetically induced CSD through the dura and the skull. (A) and (B) cross-sectional schematics showing the positioning of two gSGFETs, either over the dura and skull (Ai) or skull and then dura (Bi). Beneath each are representative recordings over the dura and skull (Aii) or skull and dura (Bii), with colours to indicate recording position: skull (purple) and epidural (black). (C) Quantification of SNR (Ci), amplitude (Cii), AUC (Ciii) and propagation velocity (Civ) for data recorded through skull and dura. For SNR (i), the artificial level was set to 3, with s.d. of ±0.1871, and then statistics (one-way ANOVA, with Tukey’s post-hoc tests). For amplitude, AUC, and propagation velocity (ii–iv), statistics were performed using unpaired T-Tests. n (mice) = 6. (D) Quantification of normalized rms of delta (Di) and theta (Dii) frequency bands for both skull (purple) and dura (black) recordings. (E) Select channels, positioned furthest apart on the gSGFET array showing electrophysiological full bandwidth traces (i) for skull (purple) and dura (black). Note the loss of spatiotemporal resolution in skull recordings compared to epidural recordings depicted as the histogram and population estimate (line) of the latency dispersion around the mean latency in an array (ii), data from six optogenetic CSDs (one CSD from six different mice). (Houben et al 2017, Chung et al 2019, Takizawa synapsin (hSyn) promotor to drive expression in et al 2020), and have been used to evaluate the excitatory and inhibitory neurons in all cortical lay- CSD threshold in ischemic stroke (Sugimoto et al ers (Nathanson et al 2009), using a modified version 2020). We have built on this approach by using a of channelrhodopsin with a point mutation (H134R) viral vector-mediated method. We used the human which generates larger photocurrents (Nagel et al 15 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al 2005) to maximize neuronal depolarization. By com- used in collaboration with modeling approaches to bining these methodologies, we have developed an investigate predominate ionic contributions to CSD awake head-fixed mouse preparation that facilitates waveform under certain disease states. This techno- preclinical CSD research. The reproducibility of opto- logy could also allow real-time evaluation of under- genetic induction and the enhanced fidelity of micro- lying brain health. Spreading depolarisation prop- transistor recordings permits high density mapping erties, i.e. waveform shape and duration, could be of CSD induction and propagation, examination and used to infer the boundaries between metabolically categorization of CSD waveform, and the ability to challenged and/or dying tissue from uncomprom- monitor neuronal activity responses across physiolo- ised areas. A change from single to double peak CSD gically relevant frequencies. waveform with a concurrent increased duration of We used a fiber optic cannula to deliver dif- neuronal depression, or indeed a switch to post-SD fused blue (488 nm) light and determined that a hyperexcitability, could be used to tract infarct or volume of tissue expressing ChR2 greater than 1 mm3 lesion progression. (figure 2) was required to reliably induce a CSD. In contrast to results reported in transgenic A single continuous (5–10 s) light stimulation was Thy-1-ChR2 mice (Houben et al 2017), we did not more effective than trains of brief depolarizations observe a large change in the threshold required to with matched total light duration (figure 1), pre- illicit CSDs during repetitive rounds of induction. sumably as a result of efficient inter-pulse extracel- Although there could be heterogeneity in waveform lular potassium buffering mechanisms and relaxa- shape across the array, CSD waveform parameters tion of neuronal depolarization block. As modeling were stable, and changed little during repeated CSD studies indicate that a rise of extracellular potassium inductions (figure 6). In contrast to a metabolic- above 10–12 mM within a restricted area of cortex ally compromised brain where SDs are believed to (>1 mm3 ) in rodents is required for CSD induction spiral around the edge of an ischemic core (Nakamura (Matsuura and Bures 1971), we hypothesize that a et al 2010, von Bornstadt et al 2015), CSDs in non- single continuous (5–10 s) light stimulation increases ischemic tissue are thought to propagate radially extracellular potassium sufficiently to trigger a CSD. across the brain (Shibata and Bures 1972). Our results Future experiments using potassium sensitive elec- confirm that in normoxic awake mice, optogenetic- trodes can be used to investigate this hypothesis. induced CSDs behave as non-recurrent propagating Spreading depolarization-induced spreading waves with a preferred direction that is preserved for depression is mechanistically explained by the fact repeated stimulations (figure 6(E)). These results are that the sustained depolarization exceeds the inac- in accordance with the small CSD propagation vari- tivation threshold for the action potential generat- ability measured by imaging methods in anesthetized ing channels (Kager et al 2002). The depression of mice (Donmez-Demir et al 2020). activity typically outlasts the neuronal depolarization The majority of published pharmacological stud- suggesting that it is maintained by other mechan- ies have investigated whether a compound reduces isms that affect synaptic function. We have quantified CSD frequency, typically in response to topical that depression outlasts depolarization by around application of KCl. Few studies report an effect on 4 min in epidural recordings in normoxic awake brain CSD threshold, amplitude or waveform characterist- (figure 3(D)). We recorded >2000 CSD waveforms ics as these can only reliably be investigated in vivo and categorized these into distinct clusters based on when CSD is induced electrically or optogenetic- certain parameters including single or double peak ally and detected using electrophysiological meth- and the presence or absence of a pre-event hyper- ods. The stability, reproducibility and consistent CSD polarization. Despite individual transistors being threshold during repeated trials in a single record- close to each other (400 µm), we were able to detect ing session, and the ability to conduct experiments local changes in CSD waveform across the cortex. in awake rodents, extenuating any confounders due These waveforms were stable during repeated rounds to anesthesia (Kitahara et al 2001) make optogenetic- of induction (figure 6). The ability to record spe- induction coupled with gSGFET array recording cific CSD waveforms from localized areas of cortex an advantageous model to conduct pharmacological highlights the advantage of spatial mapping using perturbation studies. We performed experiments in DC-coupled micro-transistors. The shape of a CSD awake brain and focused on NMDA receptor antag- is thought to reflect a number of ionic processes onists as a proof-of-principle to evaluate the in vivo including activation of persistent sodium channels, drug-screening platform. Our results confirm previ- NMDA receptors and extracellular potassium diffu- ous reports that MK-801 and ketamine can reliably sion (Kager et al 2002, Tuttle et al 2019). Double inhibit the induction of CSD (Klass et al 2018). Extra- peaks are thought to represent enhanced NMDA polation of the pharmacological results obtained in receptor activation (Krüger et al 1999). In our study, healthy animals has to be taken cautiously, particu- using naïve mice, double peaks were rare, (∼10%) larly in regard to severe diseases such as traumatic (figure 5). The ability to record multiple CSD wave- brain injury or stroke where energy failure plays forms with high-fidelity using gSGFETs could be an important role in CSD pathogenesis. Although 16 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al ketamine can be used to inhibit spreading depolar- depolarisations in subcortical areas in awake izations in severe brain injury (Carlson et al 2018), rodents. studies indicate that NMDA receptor antagonists Although we take care to keep the dura intact and effective under normoxic brain conditions are con- record epidurally, therefore minimizing tissue dam- siderably less effective in tissue modeled to reflect age and swelling, the ability to record electrographic- brain trauma by artificially raising baseline potassium ally through an intact skull will be a major advance. concentration (Petzold et al 2005). The precise rela- To explore the possibility of through-skull detection tionship between CSD and the extent of the patho- of CSD electrographically we placed two gSGFET logy in ischemic stroke, migraine, and epilepsy, is still arrays, one epidurally and the other on an intact skull, poorly understood. Future studies will benefit from and compared the recorded signals in response to the versatility that this in vivo assay offers, including an optogenetically induced CSD. Although some of viral delivery of opsins to discreet brain regions and the spatiotemporal properties are lost with subdermal awake recordings. Implementation of this methodo- recordings, including waveform shape due to volume logical approach in disease relevant preclinical mod- conduction through the bone, 5–10 mV amplitude els, including transgenic migraine mice (Loonen et al waves of depolarization travelling across the cortex at 2019) or ischemic stroke (von Bornstadt et al 2015) ∼3 mm min−1 could clearly be detected (figure 9). will permit selection of agents with enhanced transla- Our data demonstrates that subdermal recordings tional value. While patients with severe brain trauma detect both the depolarization wave and neuronal require sedation, the vast majority of CSDs will occur activity suppression in mice. Through-skull graphene in awake, unanaesthetized brains, and result in lower micro-transistor recordings, coupled with noninvas- quality of life for patients suffering migraines and epi- ive optogenetic induction of CSDs will facilitate lon- lepsy. Therefore results obtained in awake brain will gitudinal studies examining inflammatory responses have enhanced mechanistic and clinical relevance. post-CSD where invasive methodology complicates Indeed, studies inducing ischemic stroke in awake the interpretation of results (Takizawa et al 2020). brain demonstrate how anesthesia reduces not only Additionally, some relevant rodent models, such as the extent of ischemic damage after stroke, but also the mouse model of familial hemiplegic migraine masks the protective effects of a stroke therapy (Seto type 1 expressing the human mutation S218L in the et al 2014). CaV 2.1 (P/Q-type) calcium channel subunit, usually In this study we demonstrate the usefulness of our die within a few hours/days following invasive brain methodological approach to investigate mechanisms procedures as a result of uncontrolled brain swelling of CSD induction by directly confirming the import- and edema (Loonen et al 2019). Recording noninvas- ance of NMDA receptor activation for CSD initiation ively from these mice will allow the recording of spon- (figure 8). Similar approaches could be employed to taneous CSDs associated with migraine, without per- investigate the roles of persistent sodium channels turbation of the brain, and for extended periods of and extracellular potassium to CSD initiation in the time. normoxic brain, as well as in specific disease states. Although we focused in this study on short (5–10 s) 5. Conclusion continuous light stimulation to evoke CSDs, as this was most efficient (figure 1), our approach could be We demonstrate the advantages of combining used to investigate how more physiological/patho- gSGFET arrays with optogenetic CSD induction to physiological firing patterns in distinct sets of neur- map electrographically with high spatiotemporal ons could result in CSD induction. The flexibility of fidelity both infraslow activity concurrently with cor- a viral vector approach permits cell-type specific pro- related activity suppression and recovery in specified motors to be used in future studies, as well as the abil- frequency bands. The high-density recordings allow ity to optogenetically induce CSD in mouse models of detailed analysis of CSD waveform characteristics and neurological disease while mitigating expenses result- propagation. This in vivo platform is stable, allowing ing from maintaining and/or crossing multiple trans- reliable, repetitive rounds of CSDs to be induced genic lines, and thus reducing overall animal numbers on demand. This experimental methodology offers used in research (Prescott and Lidster 2017). Applic- many advantages over traditional approaches to study ation of optogenetic approaches could be used to CSD in vivo in the awake brain, and can be utilized in gain a better understanding of how spreading depol- future studies to gain mechanistic insight into disease arisations may promote epileptiform field poten- pathology, and as a pharmacological screening plat- tials rather than neuronal depression in damaged form for evaluating potential CSD modulating drugs brain tissue. Minimally invasive optogenetic stimula- in several distinct neurological disorders. tion was restricted to cortical brain regions, however recent advances in opsin engineering now permit Data availability statement transcranial activation of any region of the mouse brain expressing ultrasensitive opsins (Gong et al The data that support the findings of this study are 2020) opening the possibility of inducing spreading available upon reasonable request from the authors. 17 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Acknowledgments References This work has been funded by the European Union’s Aiba I and Noebels J L 2015 Spreading depolarization in the brainstem mediates sudden cardiorespiratory Horizon 2020 research and innovation programme arrest in mouse SUDEP models Sci. Trans. Med. under Grant Agreement Nos. 785219 and 881603 7 282ra46 (Graphene Flagship) and co-funded by the European Ayata C 2010 Cortical spreading depression triggers migraine Regional Development Funds (ERDF) allocated to attack: pro Headache 50 725–30 Ayata C 2013 Pearls and pitfalls in experimental models of the Programa operatiu FEDER de Catalunya 2014– spreading depression Cephalalgia 33 604–13 2020, with the support of the Secretaria d’Universitats Blaschke B M et al 2017 Mapping brain activity with flexible i Recerca of the Departament d’Empresa i Coneix- graphene micro-transistors 2D Mater. 4 9 ement of the Generalitat de Catalunya for emer- Carlson A P, Abbas M, Alunday R L, Qeadan F and Shuttleworth C W 2018 Spreading depolarization in acute ging technology clusters devoted to the valor- brain injury inhibited by ketamine: a prospective, ization and transfer of research results (Graph- randomized, multiple crossover trial J. Neurosurg. CAT 001-P-001702). RW is funded by a Senior 130 1513–9 Research Fellowship awarded by the Worshipful Charles A C and Baca S M 2013 Cortical spreading depression and migraine Nat. Rev. Neurol. 9 637–44 Company of Pewterers. EMC was awarded an EMBO Chung D Y et al 2019 Determinants of optogenetic cortical Short-Term Fellowship ASTF 8157 to spend time spreading depolarizations Cereb. Cortex 29 1150–61 in RW’s lab. DR is a Biotechnology and Biolo- Donmez-Demir B, Erdener Ş E, Karatas H, Kaya Z, Ulusoy I and gical Sciences Research Council (BBSRC) LIDo Dalkara T 2020 KCl-induced cortical spreading depression waves more heterogeneously propagate than sponsored PhD student. JAG and EDC acknowledge optogenetically-induced waves in lissencephalic brain: an the Ministerio de Ciencia, Innovación y Univer- analysis with optical flow tools Sci. Rep. 10 12793 sidades, la Agencia Estatal de Investigación (AEI) Dreier J P 2011 The role of spreading depression, spreading y el Fondo Europeo de Desarrollo Regional (FED- depolarization and spreading ischemia in neurological disease Nat. Med. 17 439–47 ER/UE) for the FIS2017-85787-R research project. Dreier J P, Lemale C L, Kola V, Friedman A and Schoknecht K This work has made use of the Spanish ICTS Net- 2018 Spreading depolarization is not an epiphenomenon work MICRONANOFABS partially supported by but the principal mechanism of the cytotoxic edema in MICINN and the ICTS ‘NANBIOSIS’, more specific- various gray matter structures of the brain during stroke Neuropharmacology 134 189–207 ally by the Micro-NanoTechnology Unit of the CIBER Eikermann-Haerter K et al 2015 Migraine prophylaxis, ischemic in Bioengineering, Biomaterials and Nanomedi- depolarizations, and stroke outcomes in mice Stroke cine (CIBER-BBN) at the IMB-CNM. Equipment 46 229–36 used was partially funded by Fondo Europeo de Fabricius M, Fuhr S, Willumsen L, Dreier J P, Bhatia R, Boutelle M G, Hartings J A, Bullock R, Strong A J and Desarrollo Regional (FEDER/UE) FICTS14/20-2- Lauritzen M 2008 Association of seizures with cortical 23. IMF is funded by the European Union’s Hori- spreading depression and peri-infarct depolarisations in the zon 2020 research and innovation programme acutely injured human brain Clin. Neurophysiol. under the Marie Skłodowska-Curie grant agreement 119 1973–84 Gong X et al 2020 An ultra-sensitive step-function opsin for No. 665919. minimally invasive optogenetic stimulation in mice and macaques Neuron 107 38–51.e8 Conflict of interest Hebert C et al 2018 Flexible graphene solution-gated field-effect transistors: efficient transducers for micro-electrocorti- cography Adv. Funct. Mater. 28 1703976 CG is the owner of g.tec medical engineering GmbH Hernándéz-cáceres J, Macias-gonzález R, Brožek G and Bureš J and Guger Technologies OG. PR is an employee of 1987 Systemic ketamine blocks cortical spreading depression g.tec medical engineering GmbH. but does not delay the onset of terminal anoxic depolarization in rats Brain Res. 437 360–4 Houben T et al 2017 Optogenetic induction of cortical spreading ORCID iDs depression in anesthetized and freely behaving mice J. Cereb. Blood Flow Metab. 37 1641–55 Eduard Masvidal-Codina  https://orcid.org/0000- Ikonomidou C and Turski L 2002 Why did NMDA receptor 0002-9579-8603 antagonists fail clinical trials for stroke and traumatic brain injury? Lancet Neurol. 1 383–6 Trevor M Smith  https://orcid.org/0000-0002- Kager H, Wadman W J and Somjen G G 2002 Conditions for the 5495-8906 triggering of spreading depression studied with computer Yunan Gao  https://orcid.org/0000-0003-1228- simulations J. Neurophysiol. 88 2700–12 1252 Khodagholy D et al 2013 In vivo recordings of brain activity using organic transistors Nat. Commun. 4 1575 Xavi Illa  https://orcid.org/0000-0002-3212-1128 Kitahara Y, Taga K, Abe H and Shimoji K 2001 The effects of Iñigo Martin-Fernandez  https://orcid.org/0000- anesthetics on cortical spreading depression elicitation 0001-5388-8611 and c-fos expression in rats J. Neurosurg. Anesthesiol. Anton Guimerà-Brunet  https://orcid.org/0000- 13 26–32 Klass A, Sanchez-porras R and Santos E 2018 Systematic review of 0003-1768-3293 the pharmacological agents that have been tested against Rob C Wykes  https://orcid.org/0000-0002-6141- spreading depolarizations J. Cereb. Blood Flow Metab. 6822 38 1149–79 18 J. Neural Eng. 18 (2021) 055002 E Masvidal-Codina et al Kohr G 2007 NMDA receptor antagonists: tools in neuroscience recording circuits and filters J. Neurosci. Methods with promise for treating CNS pathologies J. Physiol. 169 141–57 581 1–2 Petzold G C et al 2005 Increased extracellular K+ concentration Kramer D R, Fujii T, Ohiorhenuan I and Liu C Y 2017 Interplay reduces the efficacy of N-methyl-D-aspartate receptor between cortical spreading depolarization and seizures antagonists to block spreading depression-like Stereotact. Funct. Neurosurg. 95 1–5 depolarizations and spreading ischemia Stroke 36 1270–7 Krüger H, Heinemann U and Luhmann H J 1999 Effects of Pietrobon D and Moskowitz M A 2014 Chaos and commotion in ionotropic glutamate receptor blockade and 5-HT1A the wake of cortical spreading depression and spreading receptor activation on spreading depression in rat depolarizations Nat. Rev. Neurosci. 15 379–93 neocortical slices Neuroreport 10 2651–6 Prescott M J and Lidster K 2017 Improving quality of science Kudo C, Nozari A, Moskowitz M A and Ayata C 2008 The impact through better animal welfare: the NC3Rs strategy Lab. of anesthetics and hyperoxia on cortical spreading Anim. 46 152–6 depression Exp. Neurol. 212 201–6 Rashidy-pour A, Motaghed-larijani Z and Bures J 1995 Tolerance Loonen I C M et al 2019 Brainstem spreading depolarization and to ketamine-induced blockade of cortical spreading cortical dynamics during fatal seizures in Cacna1a S218L depression transfers to MK-801 but not to AP5 in rats Brain mice Brain 142 412–25 Res. 693 64–69 Masvidal-codina E et al 2019 High-resolution mapping of Schaefer N et al 2020 Improved metal-graphene contacts for infraslow cortical brain activity enabled by graphene low-noise, high-density microtransistor arrays for neural microtransistors Nat. Mater. 18 280–8 sensing Carbon 161 647–55 Matsuura T and Bures J 1971 The minimum volume of Seto A, Taylor S, Trudeau D, Swan I, Leung J, Reeson P, depolarized neural tissue required for triggering cortical Delaney K R and Brown C E 2014 Induction of ischemic spreading depression in rat Exp. Brain Res. 12 238–49 stroke in awake freely moving mice reveals that isoflurane Nagel G, Brauner M, Liewald J F, Adeishvili N, Bamberg E and anesthesia can mask the benefits of a neuroprotection Gottschalk A 2005 Light activation of channelrhodopsin-2 therapy Front. Neuroenerg. 6 1 in excitable cells of Caenorhabditis elegans triggers rapid Shibata M and Bures J 1972 Reverberation of cortical spreading behavioral responses Curr. Biol. 15 2279–84 depression along closed-loop pathways in rat cerebral cortex Nakamura H et al 2010 Spreading depolarizations cycle around J. Neurophysiol. 35 381–8 and enlarge focal ischaemic brain lesions Brain Sugimoto K et al 2020 Peri-infarct hot-zones have higher 133 1994–2006 susceptibility to optogenetic functional activation-induced Nasretdinov A, Lotfullina N, Vinokurova D, Lebedeva J, spreading depolarizations Stroke 51 2526–35 Burkhanova G, Chernova K, Zakharov A and Khazipov R Takizawa T et al 2020 Non-invasively triggered spreading 2017 Direct current coupled recordings of cortical depolarizations induce a rapid pro-inflammatory response spreading depression using silicone probes Front. Cell in cerebral cortex J. Cereb. Blood Flow Metab. 40 1117–31 Neurosci. 11 408 Tuttle A, Riera Diaz J and Mori Y 2019 A computational study on Nathanson J L, Yanagawa Y, Obata K and Callaway E M 2009 the role of glutamate and NMDA receptors on cortical Preferential labeling of inhibitory and excitatory cortical spreading depression using a multidomain electrodiffusion neurons by endogenous tropism of adeno-associated virus model PLoS Comput. Biol. 15 e1007455 and lentivirus vectors Neuroscience 161 441–50 von Bornstadt D et al 2015 Supply-demand mismatch transients Nelson M J, Pouget P, Nilsen E A, Patten C D and Schall J D 2008 in susceptible peri-infarct hot zones explain the origins of Review of signal distortion through metal microelectrode spreading injury depolarizations Neuron 85 1117–31 19