A Novel Approach to Study Coherent γ-Band Oscillations in Hippocampal Brain Sections

Abstract γ-Band oscillations (GBOs) are generated by fast-spiking interneurons (FSIs) and are critical for cognitive functions. Abnormalities in GBOs are frequently observed in schizophrenia and bipolar disorder and are strongly correlated with cognitive impairment. However, the underlying mechanisms are poorly understood. Studying GBOs in ex vivo preparations is challenging because of high energy demands and the need for continuous oxygen delivery to the tissue. As a result, GBOs are typically studied in brain tissue from very young animals or in experimental setups that maximize oxygen supply but compromise spatial resolution. Thus, there is a limited understanding of how GBOs interact within and between different brain structures and in brain tissue from mature animals. To address these limitations, we have developed a novel approach for studying GBOs in ex vivo hippocampal slices from mature animals, using 60-channel, perforated microelectrode arrays (pMEAs). pMEAs enhance oxygen delivery and increase spatial resolution in electrophysiological recordings, enabling comprehensive analyses of GBO synchronization within discrete brain structures. We found that transecting the Schaffer collaterals, a neural pathway within the hippocampus, impairs GBO coherence between CA1 and CA3 subfields. Furthermore, we validated our approach by studying GBO coherence in an Ank3 mutant mouse model exhibiting inhibitory synaptic dysfunction. We discovered that GBO coherence remains intact in the CA3 subfield of these mutant mice but is impaired within and between the CA1 subfield. Overall, our approach offers significant potential to characterize GBOs in ex vivo brain sections of animal models, enhancing our understanding of network dysfunction in psychiatric disorders.

GBOs can be investigated in ex vivo hippocampal sections and retain many characteristics of their in vivo counterparts (C.B. Lu et al., 2011). However, generating GBOs in ex vivo preparations is challenging because of the substantial increase in metabolic load and oxygen consumption during GBO generation (Kann et al., 2011), and sufficient oxygen must be delivered to meet these heightened energy demands Kann et al., 2016). Interface chambers have been used to overcome this obstacle by directly exposing the ex vivo tissue to highly oxygenated air Fisahn et al., 1998). Extensive research using interface chambers has explored GBOs in ex vivo hippocampal preparations Fisahn, 2005;Tsintsadze et al., 2015). However, solution exchange is slow in interface chambers, often requiring 1-2 h of drug delivery to pharmacologically induce GBOs (C. Lu et al., 2012;Pietersen et al., 2014;Lemercier et al., 2017;Xie et al., 2021). Submerged hippocampal sections allow faster pharmacological induction of GBOs (C. Lu et al., 2012;Cabungcal et al., 2013), but the limited availability of oxygen within the tissue cannot sustain continuous GBOs. Consequently, submerged preparations are primarily used to study transient GBOs triggered by brief electrical stimulation (Carmeli et al., 2013) or localized application of agonists (Gloveli et al., 2005;McNally et al., 2011).
To gain a better understanding of the generation and propagation of pharmacologically-induced GBOs throughout the complex structure of the hippocampus, we developed an approach with enhanced spatial resolution. Our approach combines the use of perforated microelectrode arrays (pMEAs) and fast superfusion, enabling the delivery of oxygenated artificial CSF (aCSF) and pharmacological agents through the interstitial space of brain sections. The electrode geometry on the MEA chips allowed for simultaneous recordings from multiple locations within the cornu ammonis (CA) CA1 and CA3 subfields of the hippocampus, which significantly improves spatial resolution. This enabled us to investigate the dynamic coupling of GBOs and confirm that GBO coherence between the CA1 and CA3 subfields, but not within each subfield, is modulated by severing the Schaffer collaterals. To the best of our knowledge, this report is the first to examine the temporal dynamics of GBO coherence within and between hippocampal subfields in ex vivo brain preparations from adult mice. Furthermore, we demonstrate the translational potential of our approach by characterizing abnormalities in GBOs evoked in hippocampal sections from a mouse model with Ank3-dependent inhibitory synaptic dysfunction.

Materials and Methods
Mice were housed in the University of Michigan's animal care facilities with controlled temperature and lighting conditions (12/12 h light/dark cycle). Mice had access to food and water ad libitum. All animal procedures were conducted in accordance with the guidelines of the University of Michigan's Institutional Animal Care and Use Committees (IACUC) and complied with NIH Guidelines for Animal Use. The mice used in the study were 3-7 weeks old. C57BL/6J mice (IMSR catalog #JAX:000664, RRID: IMSR_JAX:000664) were used for assay development. Ank3 p.W1989R knock in mice were used to model the impact of inhibitory synaptic dysfunction on GBO function.

Electrophysiological recordings
Electrophysiological recordings were captured on pMEAs comprised of 59 titanium nitride recording electrodes (30 mM in diameter) and one reference electrode arranged in a 6 Â 10 grid with a 100-mm interelectrode distance (Multichannel Systems Reutlingen, Germany). One electrode serves as an internal reference. Male mice were anesthetized with isofluorane and intracardially perfused with ice-cold (4°C) modified N-Methyl-D-glucamine (NMDG) HEPES artificial CSF (aCSF) consisting of in mM: 93 NMDG, 2.5 KCl, 0.5 CaCl 2 , 10 MgCl 2 , 1.2 H 2 PO 4 , 20 HEPES, 25 dextrose, 5 ascorbic acid, 2 thiourea, and 3 Napyruvate (Ting et al., 2018). pH was maintained at 7.4 by saturation with O 2 /CO 2 (95/5%, respectively). Mice were quickly decapitated, and the brain was dissected and transferred to a holding chamber with ice-cold NMDG HEPES aCSF. Horizontal hippocampal sections (300 mm thick) were prepared with a vibrating microtome (Leica VT1200). We bi-laterally hemisected brain sections and transferred them to an intermediate holding chamber filled with NMDG HEPES aCSF maintained at 33°C for 10-12 min. For experiments where the Schaffer collaterals were transected, we used a sterile scalpel to make a 3-to 5-mm incision along the dorsoventral axis beginning at the pial surface near CA1. Sections were transferred to an intermediate holding chamber filled with aCSF consisting of in mM: 126 NaCl, 3 KCl, 2 CaCl 2 , 1 MgCl 2 , 1.25 NaH 2 PO 4 , 25 NaHCO 3 , and 10 dextrose at 33°C for 35 min. Sections were transferred to ambient temperature for at least 15 min before recording. We mounted sections on pMEAs (Multichannel Systems) and secured them to the surface using a peristaltic perfusion system (PPS2, Multichannel Systems) to create a slight vacuum through the perforations. Sections were superfused in aCSF at a rate of 5-7 ml/min (29-31°C). Recordings were acquired at 20 kHz with a MEA2100 head stage (Multichannel Systems), and the digitized signals were recorded on the hard disk of a personal computer for offline analysis. Baseline recordings were obtained for 1 h. Chemically-induced oscillations were evoked by bath application of kainate (400 nM) for 1 h. All solutions were bubbled with with 95%O2/5% CO 2 to maintain pH 7.4.

Data processing
Oscillations from ;25-150 Hz have been designated for the g band, and the "slow" g band comprises frequencies of ;25-60 Hz (Colgin et al., 2009). We focused our analysis on frequencies from 25-59 HZ, as oscillations in this range, can be elicited by kainate in ex vivo hippocampal sections (McNally et al., 2011), and to omit 60 Hz line noise. Power analysis was done by multitaper spectrum analysis using custom-written MATLAB scripts and the multi-tapered Fourier estimation (http://chronux.org/; Mitra and Bokil, 2008). The absolute power in the g band was calculated from the integral of the power spectrum between 25 and 59 Hz. Data were preprocessed on a workstation and analyzed using MATLAB (MathWorks) and exported to GraphPad Prism for statistical analysis and plotting, unless otherwise specified. Local field potentials were isolated from MEA recordings by applying a low pass IIR Butterworth filter at 100 Hz and downsampled to 1 kHz on each channel. Spectrograms were made by convolving the signals with a morelet wavelet as follows: Sections were arranged on the electrode array to maximize electrode placement in CA1 and CA3 subfields. C, Schematic of the pMEA chip. Independently-controlled perfusion systems deliver aCSF above and below the MEA chamber. Perfusate is removed from the bottom chamber at a faster rate than it is delivered. The difference in rate creates a slight vacuum in the bottom chamber, drawing perfusate from the top chamber through the tissue, the perforated membrane, and into the bottom chamber. This results in increased diffusion of oxygenated perfusate throughout the tissue.
a normalization factor equal toðs t ffiffiffiffi p p ÞÀ 1 2 . The width of the wavelet was set to 25, Interhemispheric, zero-phase-lag coherence of g oscillations was computed between the LFP on all electrodes on the MEA using the multitapered approach above (http:// chronux.org/; Mitra and Bokil, 2008). We used the final 5 min of drug application to calculate the fractional difference in oscillatory power using the equation Power Drug À Power Drug =Power Vehicle , where the drug was kainate or kainate and bicuculline. GBO coherence was obtained from data acquired during the last 5 min of kainate application. We used the Chronux toolbox and calculated the autocorrelation on each channel, Sx(f) and Sy(f), and the cross-spectrum between the channels Sxy(f), which is calculated as follows: Gregoriou et al., 2009). Calculations for the GBO power coherence were done on the Great Lakes High Performance Computing Cluster at the University of Michigan.
We characterized GBO onset by evaluating GBO 90 -the time required for GBO power to attain 90% of the maximum value. GBO 90 was obtained by fitting the power in the g band to the sigmoid function c 1 1 expðÀapx 1 apbÞ 1d.
Recordings from channels exhibiting transient excessive noise in the g band were cleaned using linear interpolation for the missing data points, or discarded if the sigmoidal fit could not be performed. The fit window and the starting parameters for estimating the sigmoid fit were obtained manually for each recording. GBO 90-10 was obtained from the falltime function in MATLAB and used to measure the interval between GBO power rising and falling above thresholds of 90% and 10% of maximum, respectively. GBO 90-10 and GBO 10 were derived from data acquired during the last 2 min of bath application of kainate and the final 10-min application of kainate 1 bicuculline. The Q factor of the oscillations (Johnson, 1947;Kneubühl, 1997;Shea, 1929) was calculated by the equation Q = f0/B, where f0 is the peak frequency, and B is the bandwidth at 50% of maximum peak power. Recordings obtained from electrodes with artifacts or excessive noise in the g band were excluded from the kinetic analysis.

Statistical tests and models
Wilcoxon signed-rank test was used for paired data. Kolmogorov-Smirnov or Kruskal-Wallis tests were used for unpaired data, followed by Dunn's multiple comparisons for grouped data. Linear mixed-effects models were used to calculate GBO coherence within and between CA1 and CA3 with the following datasets: assay development, transected Schaffer collaterals, and Ank3 mutants. All linear mixed-effects models were generated using MATLAB's fitlme function, and maximum likelihood estimation was used as the fitting method. The full covariance matrix was calculated using the

Facilitating perfusate delivery with perforated MEAs
The pMEA chips used in our study consist of upper and lower perfusion chambers separated by a thin, perforated membrane. We employed a dual-perfusion system to deliver oxygenated artificial CSF (aCSF) both above and below the tissue sections (Fig. 1A). The top chamber received warm aCSF, while an independent perfusion system controlled the delivery and removal of aCSF from the bottom chamber. The electrodes on the pMEA were arranged in a 6 Â 10 rectangular grid, allowing us to position ;25 electrodes in the CA3 and CA1 hippocampal subfields (Fig. 1B). By removing perfusate from the bottom chamber more rapidly than it was delivered, a slight vacuum was created, which drew aCSF from the top chamber through the brain section, securing it to the array (Fig. 1C).

Induction of GBOs through kainate bath application
LFPs recorded using planar MEAs typically exhibit smaller amplitudes compared with those obtained with Figure 2. Kainate evokes oscillations in CA1 and CA3. A, Experimental timeline consisted of 1-h acclimation period followed by the perfusion of 400 nM kainate to evoke oscillations for 1 h followed by a 12-to 15-min period of pharmacological manipulation. B, Prototypical extracellular recordings during vehicle or kainate in CA1 and CA3 (low-pass filtered, 100 Hz). C, D, Prototypical periodograms generated from recordings acquired during bath application of vehicle (black) or kainate (blue) in (C) CA1 and (D) CA3.
Research Article: Methods/New Tools aCSF-filled glass micropipettes. In mouse hippocampal sections, micropipette-recorded GBOs can range from ;100 mV (Zemankovics et al., 2013) to over 4 mV in optogenetically-stimulated signals (Betterton et al., 2017). In contrast, LFPs captured by planar MEAs have amplitudes ranging from 25-50 mV in both hippocampal (Rolston et al., 2009) and cortical sections (Carmeli et al., 2013). Ex vivo hippocampal sections generally exhibit low levels of basal electrical activity (Herman and van Amsterdam, 2014). During vehicle application, we observed minimal spontaneous oscillatory activity across most frequencies (Fig. 2). However, on the application of kainate to the bath, we noticed a pronounced increase in broadband LFP power in the CA1 and CA3 subfields of the majority of sections (10 out of 11). This heightened activity stabilized into consistent oscillatory patterns (Figs. 2, 3A,B). By visually inspecting the periodograms derived from these recordings, we observed a significant enhancement in LFP power within the frequency range of 20-60 Hz, indicating the successful generation of GBOs in both the CA1 and CA3 subfields (Fig. 2C,D).

Continuous GBOs elicited by kainate in adolescent mouse hippocampal sections
To visualize and compare the kinetics of kainate-induced oscillations in the CA1 and CA3 subfields, we generated spectrograms of the LFP recordings ( Fig. 3A-D). The spectrograms revealed that kainate administration triggered persistent GBOs in both CA1 and CA3 throughout the duration of the experiment. Notably, stable GBOs emerged in CA1 earlier than in CA3 (Fig. 3G), and the peak frequency of GBOs in CA1 was ;50% higher ( Fig.  3E). We assessed the periodicity of these oscillations using the quality factor, also known as the "Q factor," a parameter that quantifies the ratio of the peak frequency to the half bandwidth and indicates periodicity in oscillators (Kneubühl, 1997). Previous findings demonstrated that bath application of kainate-induced oscillations in rat hippocampal LFP signals with Q factors .0.5 (Lemercier et al., 2017). We calculated the Q factor values from the last 10 min of kainate bath application using the LFP recordings. In CA1 and CA3, the LFPs exhibited Q factor values of 1.01 6 0.06 and 0.74 6 0.05, respectively (p , 0.0001; Fig. 3F). These results clearly indicate that bath-applied kainate can generate continuous GBOs in the hippocampus.

Rapid modulation of kainate-evoked GBOs through drug application
One notable advantage of using ex vivo tissue preparations is the ability to swiftly administer drugs for the pharmacological modulation of GBOs. In our experimental setup, we investigated the speed at which kainate-induced GBOs could be modulated by co-applying the GABA A receptor antagonist, bicuculline, to the perfusate ( Fig. 4). Bath application of bicuculline attenuated ;60% in GBO power generated in both CA1 and CA3 (Fig. 4C, H). By analyzing the kinetics of bicuculline action, we determined that 90% of its pharmacological effect was achieved in ,90 s ( Fig. 4D-H). These findings highlight the remarkable speed at which water-soluble compounds can be delivered within our experimental setup and further support the critical involvement of GABA A receptors in hippocampal GBOs Fisahn et al., 1998;C. Lu et al., 2012). Notably, Hájos and colleagues previously described a custom-made perfusion chamber that expedites the onset of pharmacologically-evoked GBOs .

Kainate induces synchronization of GBOs within and between hippocampal subfields
GBOs can transiently emerge within local networks of neurons, but they can also interact and synchronize across distinct networks (Buzsáki and Wang, 2012). Leveraging the geometric spacing between the electrodes on our MEA chips, we simultaneously recorded signals from multiple locations within the CA1 and CA3 hippocampal subfields. By analyzing LFP signals from channels in CA1 and CA3, we characterized how kainateinduced GBOs synchronize within and between these hippocampal subfields. During basal conditions, the coherence of GBO power was generally low within and between CA1 and CA3 (Fig. 5A,C). However, on kainate   Table 1 for results from linear mixed effects model. application, we observed a rapid increase in GBO power coherence (Fig. 5B,D-F). Specifically, kainate enhanced the mean GBO power coherence within CA1 by ;32% and within CA3 by ;22% (Fig. 5E,F). Notably, kainate also increased the mean coherence of GBO power between CA1 and CA3 by ;15% (Fig. 5G). To visualize the variability of GBO power coherence within each brain section, we plotted pair-wise comparisons of GBO coherence per slice (Fig. 5G) and fitted the results to a linear mixed model (Table 1). Consistent with changes observed in mean GBO power coherence within each brain slice, the results from the linear mixed model indicated that kainate induced more robust increases in GBO coherence within the CA1 and CA3 subfields (p = 0.225) compared with between the subfields (Fixed effect coefficient À0.104; p , 0.0001; Fig. 5H). These data demonstrate that kainate application induces synchronization of GBOs within and between hippocampal subfields, with more robust coherence enhancements observed within the CA1 and CA3 subfields compared with between them.
Schaffer collaterals play a key role in GBO synchronization between CA3 and CA1 hippocampal subfields The Schaffer collaterals serve as pathways through which CA3 neurons project to CA1 neurons, connecting anatomically distant neuronal networks via a single excitatory synaptic connection (Amaral and Witter, 1989). This unilateral circuit enables CA3 excitatory projection neurons to modulate the activity of CA1 neurons, particularly within the frequencies of slow g (;25-50 Hz; Combe et al., 2018). However, while focal application of kainate can induce GBOs in CA1 (Chittajallu et al., 2013), it has been shown that CA1 GBOs can exist independently of CA3 GBOs (Craig and McBain, 2015).
To investigate the mechanism underlying slow GBO coherence between CA3 and CA1 g , we conducted experiments in which we transected the Schaffer collaterals. The power of kainate-evoked GBOs was similar in in hippocampal sections with transected Schaffer collaterals, but peak frequency was reduced (Extended Data Fig. 6-1). The mean GBO coherence detected between electrodes within the CA1 or CA3 subfields of the transected hippocampal sections increased by ;17% or 19%, respectively (Fig. 6B,D-F).
As expected, transecting the Schaffer collaterals prevented an increase in GBO coherence between electrodes in CA1 and CA3 (Fig. 6H). However, direct comparison of the mean GBO coherence between all possible pair-wise comparisons within CA1 and CA3 of control slices or transected slices did not yield statistical significance (Kruskal-Wallis p = 0.020; Dunn's multiple comparisons within CA1 p ! 0.999; within CA3 p . 0.999; between CA1 and CA3 p = 0.260). It is important to note that comparing the means of averages between control and experimental groups can oversimplify the data by disregarding variability. To better understand the differences within the data and preserve variability, we employed a linear mixed model (Table 2). When we incorporated the Schaffer collaterals as a predictor variable in the linear mixed model, the results indicated that severing the Schaffer collaterals was not a predictor of kainate-induced changes in overall GBO coherence. However, severing the Schaffer collaterals was a strong predictor for the decrease in GBO coherence between CA1 and CA3 (p , 0.0001), providing support for the crucial role of Schaffer collaterals in synchronizing GBOs between the CA3 and CA1 subregions.

Attenuated GBO power in Ank3 mouse hippocampal sections
To showcase the utility of our approach, we investigated kainate-induced GBOs in hippocampal brain sections obtained from mice carrying a disease-associated variant in Ank3 (Nelson et al., 2020). Ankyrins, a family of scaffolding proteins widely expressed throughout the body, play critical roles in neuronal structure and function (Bennett and Lorenzo, 2013;Nelson and Jenkins, 2017). Genome-wide association studies (GWAS) have linked Ank3 with psychiatric disorders, including bipolar disorder (Stevens and Rasband, 2021). In a previous study, the disease-associated Ank3 variant, Ank3 p.W1989R, was shown to diminish GABAergic neurotransmission from fast-spiking interneurons (FSIs) and reduces the power of kainate-evoked GBOs in hippocampal sections from mice carrying this variant (Nelson et al., 2020). We hypothesized that impaired GABAergic inhibition resulting from the Ank3 mutations would lead to a decreased capacity of hippocampal sections to generate and sustain GBOs. Consistent with WT littermates, the bath application of kainate promptly increased broadband LFP power in hippocampal sections from Ank3 mutant mice, which subsequently stabilized into GBOs (Fig. 7A,B,D,E; Extended Data Fig. 7-2). However, we observed a reduction in the power of kainate-induced GBOs in Ank3 mice (Fig. 7C,F; Extended Data Fig. 7-1), consistent with a previous report (Nelson et al., 2020). Notably, there were no discernible differences in the temporal onset of GBOs between the CA1 and CA3 subfields of hippocampal sections from Ank3 mutant mice (Kruskal-Wallis statistic = 4.957, p = 0.175; Extended Data Fig. 7-2). Moreover, co-application of bicuculline rapidly abolished kainate-induced GBOs in hippocampal sections of Ank3 mutants, following a similar time course as observed in WT littermates (Kruskal-Wallis statistic = 1.953, p = 0.582; Extended Data Fig. 7-3). Thus, in Ank3 mutant mice, the power of kainate-induced GBOs in hippocampal sections was reduced, underscoring the impact of Ank3 mutations on GBO generation. A portion of the data presented in Figure 7 was previously published in Figure 2 and Supplementary Figure 8 of Nelson et al. (2020).
GBO coherence is impaired in CA1 but not CA3 of Ank3 mice GBO coherence in Ank3 mice showed comparable robustness to that observed in hippocampal sections from WT littermates (Fig. 7A-F; Extended Data Fig. 7-4). Under basal conditions, mean GBO coherence within and between all hippocampal subregions did not differ significantly between WT and Ank3 mice: within CA1  Fig. 7-4E). Spatial mapping of electrodes revealed that kainate application broadly increased GBO coherence across most electrodes localized in a hippocampal slice from an Ank3 mutant mouse (Fig. 7G,H). The mean GBO coherence within and between hippocampal subfields differed between sections taken from Ank3 MT mice and WT littermates (Kruskal-Wallis statistic = 20.53; p = 0.0010). However, multiple comparison tests demonstrated that mean GBO coherence within CA3 (p . 0.999), within CA1 (p = 0.334) and between CA3 and CA1 (p . 0.999) did not differ (Extended Data Fig. 7-4F). We incorporated mouse genotype as a predictor variable in a linear mixed effects model to account for variance within the coherence measurements (Fig. 7I). The model revealed that the Ank3 mutation itself was not a significant predictor (p = 0.434). However, the mutation exhibited interactions within CA1 region coherence (p , 0.0001) and between CA1 and CA3 coherence (p , 0.0001; Table 3). Taken together, continued p = 0.260). Data shown as mean 6 SEM, n = number of slices. I, Kainate-evoked -fold change in GBO coherence per electrode, grouped by hippocampal subfield. Each circle represents a paired comparison of GBO power between two electrodes in hippocampal sections with intact (black) or transected Schaffer collaterals (magenta). See Table 2 for results from linear mixed effects model Panels J and K from Extended Data Figure   these findings imply that the reduced GABAergic inhibition caused by the Ank3 mutant does not impair the capacity to generate GBOs, but does significantly reduce GBO power. We previously demonstrated that the dysfunction of parvalbumin-expressing interneurons contributed to the decreased inhibitory neurotransmission and altered hippocampal GBOs in Ank3 mutant mice (Nelson et al., 2020). Thus, although the overall power of GBOs is reduced, the coherence of GBOs remains intact within CA3 of Ank3 mice, but attenuated within CA1 and between CA3 and CA1. This suggests that deficits to parvalbumin-expressing interneurons caused by the Ank3 mutation may specifically impair mechanisms of GBO coherence within and between the CA1 subfield.

Discussion
In this study, we have successfully developed a cuttingedge approach for investigating kainate-evoked g -band oscillations (GBOs) in ex vivo hippocampal slices. Our method showcases the ability to rapidly generate, sustain, and pharmacologically modulate GBOs in mature mouse hippocampal sections for extended durations of over 90 min. Leveraging multilocation LFP recordings within hippocampal subregions, we harnessed computational algorithms to precisely quantify GBO coherence within and between the CA1 and CA3 hippocampal subfields. To validate the broad applicability of our approach, we employed a transgenic mouse model with inhibitory synaptic dysfunction, effectively demonstrating how our methodology can uncover comprehensive insights into generalized deficits in oscillatory activity. Notably, our findings not only substantiate previous observations of reduced GBO power in Ank3 mutant mice but also unveil nuanced impairments in the synchronization of GBOs within and between the hippocampal subfields of Ank3 mice.

GBOs elicited by kainate unveil glutamatergic activation pathway
In our investigation, we induced g -band oscillations (GBOs) by employing a bath application of kainate, a potent glutamatergic agonist. It is important to note that GBOs can also be elicited in ex vivo hippocampal sections through various pharmacological manipulations, including the use of cholinergic agonists . The specific activation of GBOs by kainate is mediated by the engagement of GluK5 and GluK6 receptors (Fisahn et al., 2004), which are believed to be predominantly expressed by GABAergic interneurons within the hippocampus (Paternain et al., 2000).

Kainate and cholinergic agonists elicitation GBOs by distinct mechanisms
In contrast, the activation of GBOs by cholinergic agonists occurs through the stimulation of M1 receptors , which are equally expressed by both pyramidal neurons and GABAergic interneurons in the hippocampus (Cea-del Rio et al., 2010). Notably, kainate is known to elicit more robust GBOs (Cunningham et al., 2003), whereas cholinergic agonists tend to induce GBOs with higher peak frequencies . These findings suggest that kainate and carbachol evoke GBOs through partially distinct mechanisms. However, it is important to acknowledge that direct comparisons of results obtained from different studies can be challenging because of variations in experimental setups, such as differences in apparatuses, perfusion kinetics, and local oxygen partial pressures. To gain further insights and distinguish the underlying mechanisms, it would be beneficial to characterize the temporal dynamics of GBO coherence specifically between kainate-induced and carbachol-induced GBOs under consistent experimental conditions, similar to those employed in our study. The significant progress achieved in this study was largely driven by two crucial technological advancements.
pMEAs facilitate the robust generation and modulation of GBOs in ex vivo tissue The utilization of pMEA technology is the first key technological advancement in our approach. pMEAs played a pivotal role in reliably evoking GBOs in ex vivo brain sections by addressing the energetic demands of these oscillations that require near-maximal mitochondrial oxidative capacity (Kann et al., 2011). By enhancing the delivery of oxygenated aCSF to the hippocampal sections and increasing the local O 2 partial pressure within the interstitial space (Egert et al., 2005), pMEAs ensured consistent generation of GBOs and facilitated their pharmacological modulation (Figs. 3, 4). Additionally, the density and geometry of the electrode arrays employed in our study enabled simultaneous recordings from .10 electrodes in each hippocampal subfield, significantly enhancing spatial resolution. In contrast, a previous study on kainate-induced GBOs only recorded LFPs from a single electrode placed in each CA1 and CA3 subfield (C. Lu et al., 2012). This improved spatial resolution enabled us to investigate the temporal development of GBO coherence within and between CA1 and CA3. By selectively severing the Schaffer collaterals, the axonal projections of CA3 pyramidal cells that innervate the apical dendrites of CA1 pyramidal cells or CA1 parvalbumin-expressing FSIs, we discovered that kainate independently induces GBOs within CA1 and CA3 subfields while eliminating power-power coherence between them. Our interpretation of these findings suggests that CA1 and CA3 possess distinct mechanisms for generating kainate-induced GBOs, with the synchronization of GBOs between subfields dependent on intact Schaffer collaterals. Notably, empirical studies (Colgin et al., 2009) and computational models (Mysin et al., 2019) support these interpretations, reinforcing the robustness and validity of our conclusions.

Enhanced viability and age flexibility in generating GBOs through NMDG recovery
Another significant advancement in our approach was the incorporation of the NMDG recovery method, which expanded the age range of mice from which GBOs could be continuously evoked. Typically, GBOs are studied using ex vivo brain sections from young animals (Carmeli et al., 2013;Tsintsadze et al., 2015). Although there are reports of GBOs evoked in hippocampal sections from older mice, the ability to evoke GBOs is often transient (Gloveli et al., 2005), making ex vivo tissue from neonatal animals more viable for these experiments.
Generating GBOs ex vivo requires the presence of healthy fast-spiking interneurons (FSIs) and maintaining tissue viability throughout the experiment is crucial. By incorporating the NMDG recovery method into our approach, we significantly enhanced the viability of our brain sections (Ting et al., 2014). Importantly, the combination of pMEAs and NMDG recovery allowed us to evoke GBOs from ex vivo hippocampal sections across a broader range of developmental ages when FSIs are physiologically mature (Goldberg et al., 2011). This expanded age flexibility has valuable implications, particularly for studying GBOs in animal models of diseases that manifest later in life, such as schizophrenia or dementia. The utilization of the NMDG recovery method opens up new avenues for investigating oscillatory dynamics in more relevant disease models and deepening our understanding of the underlying mechanisms.
Our approach proved instrumental in uncovering previously unrecognized deficits in hippocampal GBOs by investigating a mouse model with impaired GABAergic neurotransmission. Previous research demonstrated reduced power in kainate-induced GBOs in Ank3 mice (Nelson et al., 2020). In this study, we not only confirmed and expanded on those findings but also revealed that while kainate-induced GBO power is diminished in both the CA1 and CA3 regions of Ank3 mice, GBO coherence is specifically attenuated in CA1. Although the mechanistic basis for the divergent GBO responses between CA1 and CA3 in Ank3 mice requires further elucidation, it is plausible that the deleterious impact of the Ank3 mutation primarily affects the mechanics of GBO coherence within CA1, while sparing CA3, given the partially distinct mechanisms underlying GBO generation in these regions (Combe et al., 2018).
In summary, we have presented a novel assay that successfully generates and sustains GBOs in ex vivo tissue from juvenile mice. This approach holds promise for characterizing a wide range of oscillatory frequencies in ex vivo sections from various excitable tissues and animal models. Moreover, we propose that the utility of this approach can be significantly expanded by incorporating focalized methods such as electrical, optogenetic, or chemogenetic stimulation to precisely generate oscillations. By leveraging these advancements, we can further unravel the complexities of oscillatory dynamics and deepen our understanding of their functional implications in health and disease.