Tumor associated epilepsy is a common and debilitating co-morbidity of brain tumors, for which inadequate treatments are available. Additionally, animal models suggest a potential link between seizures and tumor progression. Our group has previously described a mouse model of diffusely infiltrating glioma and associated chronic epilepsy. G protein-gated inwardly rectifying potassium (GIRK) channels are important regulators of neuronal excitability, but their development as a target of antiseizure medications has been hampered by cross-reactivity with GIRK channels in the heart. Recently GiGA1, a novel GIRK agonist that is highly selective for brain tissue, was developed and shown to have antiseizure properties in an acute chemoconvulsant model. Here, we test GiGA1 ex vivo in our established mouse model of tumor associated epilepsy, demonstrating that a highly selective, small-molecule GIRK agonist can reduce seizure-like activity in the peritumoral region, where neurons and glioma cells interact and from which focal seizures arise.

Mouse Glioma Model

A retroviral vector containing PDGFA-IRES-Cre was injected into the subcortical white matter of transgenic C57BL/6 mice containing floxed p53, STOP-floxed RPL22^HA, and STOP-floxed mCherry-luciferase, resulting in conditional expression of these genes. Glioma cells were then isolated from these tumors and sustained in culture.

Wild-type C57BL/6 or transgenic Thy1-GCaMP6f mice, of both sexes, were used for experiments. In some experiments, mice aged 1-4 months were injected with 50,000 glioma cells in a volume of 1 μl at 0.2 μl/min. The cells were injected into right frontal cortex (coordinates relative to the bregma: 2 mm anterior, 2 mm right, and 2 mm deep). In vivo luciferase imaging was performed weekly to track tumor growth, and tumor-bearing mice were sacrificed 28-35 days post injection (dpi) to prepare acute slices for electrophysiology.

Electrophysiology

Mice were sacrificed by cervical dislocation and the whole brain was immediately placed in ice-cold Cutting Solution (sucrose 210 mM, potassium chloride 2.5 mM, sodium bicarbonate 26 mM, sodium phosphate monobasic 1.25 mM, dextrose 10 mM, calcium chloride 0.5 mM, magnesium chloride 7 mM). While submerged in ice-cold, aerated (95% O₂ and 5% CO₂) cutting solution, 400 µm coronal slices through frontal cortex were prepared using a Leica VT 1000S Vibratome (Nussloch, Germany). Slices were placed in aerated Recovery Solution (sodium chloride 125 mM, potassium chloride 2.5 mM, sodium bicarbonate 26 mM, sodium phosphate monobasic 1.25 mM, dextrose 10 mM, calcium chloride 2 mM, magnesium chloride 1.5 mM) at 35° C for 18 minutes, then at least 42 minutes (for a total of 1 hour) in the same aerated solution at room temperature.

Baseline recordings were performed in artificial cerebrospinal fluid (ACSF) comprising sodium chloride 125 mM, potassium chloride 5 mM, sodium bicarbonate 26 mM, sodium phosphate monobasic 1.25 mM, dextrose 10 mM, calcium chloride 2 mM, magnesium chloride 1.5 mM. To generate spontaneous seizure-like events while preserving inhibitory function, slices were recorded in Zero-Magnesium ACSF (sodium chloride 125 mM, potassium chloride 5 mM, sodium bicarbonate 26 mM, sodium phosphate monobasic 1.25 mM, dextrose 10 mM, calcium chloride 2 mM) containing either 0.1% DMSO (vehicle control), or 0.1% DMSO and 100 µM GiGA1. Recordings were performed at 32° C.

Each slice was placed on a 4 x 4 mm, orthogonal array of 96 penetrating microelectrodes (10 x 10 arrangement, with 4 corner reference electrodes), each 1 mm in length (Blackrock Neurotech Inc., Salt Lake City, UT). The slices were positioned such that electrodes covered cortex (including tumor, if present) as well as subcortical regions. Signals were continuously acquired on a CerePlex Direct acquisition system with digital preamplifier (Blackrock Neurotech Inc., Salt Lake City, UT) at 30 kHz, with 16-bit precision and a range of +/− 8 mV.

Calcium Imaging

Images were acquired using an upright Leica DM LFS microscope, with a 2.5x/0.07NA Leica objective and an I3 fluorescence filter cube (Leica Microsystems, Wetzlar, Germany). Tumor-bearing slices from Thy1-GCaMP6f transgenic mice were excited at 450-490 nm. Images were recorded using an Andor Zyla Plus sCMOS 4.2 MP camera (Oxford Instruments, Abingdon, UK), at 50 frames per second and a resolution of 1024 x 1024 pixels.

Experimental Drugs

GiGA1 was provided courtesy of the National Center for Advancing Translational Sciences (NCATS), National Institutes of Health (NIH). GiGA1 powder was reconstituted in DMSO at 100 mM, aliquoted, and stored at -20° C. Prior to each experiment an aliquot was diluted 1:1000 in ACSF, for a final concentration of 100 µM GiGA1 and 0.1% DMSO.

Data Analysis

All analyses were performed in MATLAB versions R2019a, R2022a or R2024a using published libraries including NSxFile (https://github.com/edmerix/NSxFile), NeuroClass (https://github.com/edmerix/NeuroClass); and custom scripts (https://github.com/rrifkin/RAR_MEA_GECI).

Local Field Potentials (LFP)

Raw multi-electrode array data was symmetrically band-pass filtered (2-50 Hz, 1024^th order, window-based FIR1 filter using MATLAB) and down-sampled to 2000 Hz. LFP data were visually analyzed to remove time periods and channels with excessive artifact. Channels for which the electrode was not contacting the slice were also removed. Additionally, for paired, ‘per-channel’ analyses, inactive channels (those with zero definite epileptiform discharges) were excluded. ‘Line-length’ was calculated from the 2-50 Hz LFP. ‘Peak Amplitude’ was determined by finding local maxima and minima with an absolute value ≥ 4 standard deviations from the mean. ‘Peritumoral’ electrodes were defined as those located within the tumor, or within a margin of two electrodes (i.e. 800 µm) of the tumor, based on light microscopy.    

Multi-Unit Activity (MUA)

Raw multi-electrode array data was symmetrically band-pass filtered (500-5000 Hz, 1024^th order, window-based FIR1 filter using MATLAB). Spikes within each channel were detected using a voltage threshold of 4.5 σ, where σ = median(|x|/0.6745) and x is the MUA of that channel. These putative spikes were subjected to several quality control measures. Spikes with no calculable full-width at half-maximum (FWHM), spikes with a Mahalanobis distance in principal component space greater than would be expected less than 1% of the time, and spikes with voltages not within 3 standard deviations of any unit in a library of 4,772 manually sorted single units, were deleted. For those remaining, the FWHM was compared to a Gaussian mixture model fitted to the FWHM of action potentials from human microelectrode array recordings. From this, each spike was assigned a probability of belonging to the fast-spiking interneuron or pyramidal cell populations, and these values were used to calculate a probabilistic instantaneous firing rate for each population.

Spike-triggered averaging was performed using spikes across all electrodes categorized as either fast-spiking interneuron (IN) or pyramidal cell (PC) by the above method. In addition to filtering for false-positives as described above, spikes occurring at the same time point were deleted, as they likely represented electrical noise. PC spikes occurring within +/- 20 ms of an IN spike were counted in 2 ms bins. PC spikes were normalized to mean PC firing rate and total number of IN spikes, to account for the effect of GiGA1 on overall firing as well as intrinsic differences between slices. Within each 2 ms bin, DMSO vs. GiGA1 were compared using Wilcoxon Signed-Rank test, and these p values were corrected using the Bonferroni-Holm method.

Calcium Imaging

Circular Regions of Interest (ROI) with a radius of 140 µm were created around each electrode, with the exception of electrodes not in contact with the slice, which were excluded. The average fluorescence within each of these ROIs was calculated for each frame of the recording. This fluorescence was divided by the mean of a sliding window of 125 frames before and after. ‘Peak Amplitude’ was determined by finding local maximal and minima with an absolute value ≥ 1% above the mean. To evaluate the hypersynchrony of adjacent regions of tissue, the normalized fluorescence data for each ROI was compared to each adjacent ROI via Pearson correlation, generating an r-value representing the degree of synchrony (duplicate comparisons were avoided). The means of these r-values were compared between groups.

Statistical Analysis

Non-paired data were analyzed using the Wilcoxon Rank-Sum (Mann-Whitney U) test. Paired data were analyzed using the Wilcoxon Signed-Rank test. For multiple comparisons, the Bonferroni-Holm correction method was used. For paired analyses in which the individual channels or ROIs were the data points, outliers ≥ 3 scaled median absolute deviations from the median were excluded (solely to allow every included data point to be plotted on a meaningful scale; inclusion or exclusion of outliers did not impact significance). Incalculable data points (‘NaN’) were excluded from analyses.

(Adapted from methods of the linked manuscript)