Interneuron theta phase locking controls seizure susceptibility

Zoé Christenson Wick; Paul A Philipsberg; Cassidy Kohler; Sophia I Lamsifer; Elizabeth Katanov; Christopher D Adam; Kathryn E Gordon; Yu Feng; Lauren M Vetere; Genevra C Donnelly; Corin Humphrey; Denise J Cai; Tristan Shuman

doi:10.1101/2025.09.10.675457

This is a preprint.

It has not yet been peer reviewed by a journal.

The National Library of Medicine is running a pilot to include preprints that result from research funded by NIH in PMC and PubMed.

[Preprint]. 2025 Sep 11:2025.09.10.675457. [Version 1] doi: 10.1101/2025.09.10.675457

Interneuron theta phase locking controls seizure susceptibility

Zoé Christenson Wick ^1,^*,^†, Paul A Philipsberg ^1,^†, Cassidy Kohler ¹, Sophia I Lamsifer ¹, Elizabeth Katanov ¹, Christopher D Adam ¹, Kathryn E Gordon ¹, Yu Feng ¹, Lauren M Vetere ¹, Genevra C Donnelly ¹, Corin Humphrey ¹, Denise J Cai ¹, Tristan Shuman ^1,^*

PMCID: PMC12440021 PMID: 40964369

Abstract

The timing of neuronal activity is highly precise and often organized by brain-wide oscillations. Many neurons modulate their firing rates at specific phases of theta (known as theta phase locking), creating discrete windows for information processing. Disrupted theta phase locking has been found across several neurological and psychiatric disorders (e.g., epilepsy), but gaps in technology have prevented its causal influence from being tested. Here, we developed PhaSER, a closed-loop optogenetic system designed to control the phase locking of specific interneurons, and demonstrate a causal role for inhibitory phase locking in seizure susceptibility. We first found that parvalbumin (PV+) and somatostatin (SOM+) expressing interneurons in the dentate gyrus (DG) show distinct theta phase locking profiles and are differentially impacted in a mouse model of chronic temporal lobe epilepsy. In healthy mice, PV+ interneurons have extremely consistent phase-locked firing near the trough of CA1 theta, aligned with excitatory inputs to DG. However, in epileptic mice, PV+ interneuron activity is dispersed across the theta cycle, suggesting that altered inhibitory phase locking could be a causal mediator of seizure susceptibility in epilepsy. To test this hypothesis, we applied PhaSER to directly control the phase locking of DG interneurons during an acute test of seizure susceptibility. In epileptic mice, re-aligning DG PV+ interneuron theta phase locking reduced seizure susceptibility, while in healthy mice, disrupting normal phase locking of PV+ interneurons increased seizure susceptibility. Together, this provides the first causal evidence that inhibitory theta phase locking can directly control network function by shifting seizure susceptibility in the healthy and epileptic brain.

INTRODUCTION

The brain relies on precise temporal organization of neuronal spiking to control behavior. Across the brain, network-wide coordination of neural activity is often rhythmic, forming oscillations (e.g., theta) which organize the activity of neurons within and across brain regions^1,2. This rhythmic synchronization leads many neurons to be active at specific oscillatory phases (i.e., “theta phase-locked”) and creates temporal windows where co-activity can bind neural ensembles and drive distinct cognitive processes^3–8. Inhibitory neurons are particularly highly phase-locked to theta oscillations, with distinct interneuron subtypes active at different phases^9–14. This is thought to control the routing of inputs and outputs and maintain excitatory-inhibitory homeostasis. As such, disorganized inhibitory spike timing may impair cognitive computations and disrupt the balance of excitation and inhibition. Yet the only evidence for the importance of phase locking is correlational – extrapolated from situations in which it is disrupted. For instance, phase-locked spiking is altered in models of substance use, schizophrenia, autism, Alzheimer’s disease, and epilepsy^15–23, suggesting that it may be a common driver of disrupted processing across neurological diseases. Together, this strongly suggests that disruptions to phase locking may be sufficient to induce network dysfunction and alter behavior, but (due to technical limitations) this has never been causally tested.

Under healthy conditions, the dentate gyrus (DG) acts as a “gate” to the hippocampus, with powerful local inhibition selecting which excitatory signals get relayed to downstream regions^24–26. Waves of rhythmic excitatory inputs enter the DG, carrying spatial information and driving local theta^27,28. Meanwhile, phase-locked interneurons provide rhythmic inhibition to match these excitatory inputs^16,17. This puts DG interneurons in a critical position where they must prevent excess excitation from overrunning the network and generating hippocampal seizures^29,30 – a hallmark of temporal lobe epilepsy (TLE). TLE is the most common form of epilepsy in adults and is often resistant to anti-epileptic drugs. Since most anti-epileptic drugs boost inhibition indiscriminately, this pharmaco-resistance suggests that the mechanisms of seizures are more complex than a bulk imbalance of excitation and inhibition. In fact, using a mouse model of drug-resistant TLE (pilocarpine-induced status epilepticus), previous work has shown a breakdown in inhibitory theta phase locking in the DG of epileptic mice¹⁶, while major excitatory inputs to the DG maintain their rhythmicity¹⁷. This suggests that there is a mismatch between the precise timing of excitatory inputs to the DG and local interneuron spiking. We predicted that this breakdown in inhibitory theta phase locking that occurs in the epileptic DG may be a direct causal driver of seizure susceptibility in epilepsy.

To directly test the causal role of inhibitory theta phase locking, we developed and validated an open-source tool, PhaSER (Phase-locked Stimulation to Endogenous Rhythms), that can directly manipulate the phase locking of neurons during behavior. PhaSER performs low-latency theta phase detection and phase-locked optogenetic manipulations, while permitting simultaneous monitoring of spike timing. We used PhaSER to apply phase-specific optogenetic manipulations and directly test the hypothesis that inhibitory phase locking in the DG is a causal mediator of seizure susceptibility in both healthy and epileptic mice.

RESULTS

Dentate gyrus PV+ interneurons provide highly precise inhibition at the trough of theta

Very little is known about the in vivo firing patterns of distinct subtypes of interneurons in the DG of healthy or epileptic mice. Two prominent and non-overlapping inhibitory cell types in the DG are the parvalbumin (PV+) and somatostatin (SOM+) expressing cells, which have distinct spatial connectivity profiles and spike timing properties^31–35. While the in vivo firing patterns, including theta phase locking, of PV+ and SOM+ cells in the DG are not thoroughly characterized (in either healthy or epileptic animals), studies have shown their intrinsic properties and connectivity profiles are differentially impacted in epilepsy^16,36–39. These changes could drastically influence the role these cell populations play in their local network, including in their phase locking profiles. We therefore sought to characterize DG PV+ and SOM+ cell theta phase locking in both healthy and epileptic mice and determine how these cells’ phase-locked spiking may impact seizure susceptibility.

To determine the firing patterns of PV+ and SOM+ cells in the DG of healthy and epileptic mice, we conducted acute silicon probe recordings in awake, head-fixed male and female mice as they navigated a virtual linear environment for water rewards (Fig 1A). To isolate the activity of DG PV+ or SOM+ cells, we used a viral-transgenic optogenetic tagging strategy: a Cre-dependent virus expressing the excitatory opsin channelrhodopsin-2 (ChR2) was injected into the dorsal hippocampus of PV-Cre or SOM-Cre mice (Fig 1A). Single-units were identified as opsin-expressing PV+ or SOM+ neurons if they showed a low-latency (<4ms) increase in spiking in response to blue light pulses (Fig 1A). These units’ spike patterns during a baseline period, before any light stimulations were applied, were then analyzed post-hoc.

We first examined healthy animals and found that DG PV+ interneurons had extraordinarily tight phase locking to the trough of theta (Fig 1B), with all PV+ interneurons having a mean firing phase (mu) near the trough. This led to a very strong modulation of firing rate relative to the phase of theta, with ~5x higher firing rates near the trough compared to near the peak of theta (Fig 1C). DG SOM+ neurons also show phase-locked firing near the trough of theta, but the preferred theta phases were much more distributed (Fig 1B), and the population of SOM+ neurons showed notably less overall modulation of spiking across theta phases (Fig 1C). We found no differences in the strength of phase locking for individual neurons (r-values) across PV+ and SOM+ interneurons (Fig 1D), suggesting that the consistency of PV+ cell phase preferences (Fig 1B) is primarily responsible for the increased aggregate phase locking of this population (Fig 1C). Importantly, this precisely trough-aligned spiking matches the spike timing of entorhinal layer 2 neurons, a major excitatory afferent to the DG^17,28. Thus, in healthy animals, inhibition from PV+ and SOM+ cells is most robust at the trough of theta, when excitatory inputs are strongest. However, we found key differences in the overall phase locking of these populations, with PV+ cell spiking precisely targeted near the trough of theta and SOM+ cell firing more broadly distributed throughout the theta cycle.

PV+ interneuron theta phase locking is disrupted in epileptic mice

We next explored how the phase locking of PV+ and SOM+ neurons in the DG are impacted in epilepsy. First, we replicated prior work showing that putative inhibitory neurons (Fig S1A,D) in the DG of pilocarpine-treated epileptic mice have substantially altered theta phase locking^16,17. Specifically, in epileptic mice, inhibitory neurons in the DG had theta phase preferences that were more distributed throughout the theta oscillation (Fig 1E,F, Fig S1B,E), and their firing was less strongly modulated by theta phase (Fig 1G, Fig S1C,F). However, we saw no change in their overall firing rate (Fig S1G,I)¹⁷.

When looking at opto-tagged DG PV+ and SOM+ cells specifically, we found that DG PV+ cells in epileptic mice had significantly more distributed theta phase preferences than in healthy animals (Fig 1H,I), though there was no change in the strength of their theta phase modulation in individual neurons (r-value; Fig 1J). Thus, the distributed firing activity of the population of PV+ interneurons in epileptic mice (Fig 1H) was driven by the population having much more variable preferred phases of firing, rather than each neuron’s spiking being more distributed. We saw no overall change in PV+ cell firing rates (Fig S1H). Meanwhile, we were unable to identify any DG SOM+ cells in epileptic animals using optogenetic-tagging strategies. This is likely due to the substantial SOM+ cell loss that occurs in temporal lobe epilepsy^16,40–43 and the low throughput of opto-tagging. Altogether, while the majority of DG SOM+ cells are lost in TLE, the remaining interneurons, including PV+ cells, show more distributed theta phase locking.

Importantly, we confirmed that epileptogenesis was equally effective at producing spontaneous seizures in both PV-Cre and SOM-Cre mouse strains (Fig S1K), and saw virtually identical deficits in overall DG inhibitory theta phase locking in both strains as well (Fig S1B,C,E,F). However, there was a notable sex effect with the phase locking deficit primarily occurring in males (Fig S2A,B). Meanwhile, DG inhibitory neurons in epileptic females showed theta phase locking profiles that were more similar to healthy females (Fig S2A,B,D). Interestingly, this sex-specific change in theta phase locking occurred despite males and females having similar seizure frequencies (Fig S2C). Thus, there may be sex-specific mechanisms driving altered spike timing and seizures in TLE.

The more distributed inhibitory spike timing in epileptic mice (Fig 1F,H) may be a compensatory mechanism to tamp down hyperexcitability, but it also alters the relative distribution of inhibition across the theta cycle. This may disrupt homeostatic plasticity mechanisms^44,45 and leave the DG vulnerable to hyperexcitability when rhythmic excitatory inputs are strongest (i.e., near the theta trough). Given the theorized importance of inhibitory spike timing in shaping circuit-wide dynamics, and the changes we found in epileptic mice, we hypothesized that this disruption in inhibitory theta phase locking might causally contribute to the emergence of seizures in TLE. To test this hypothesis, we performed direct manipulations of PV+ and SOM+ cell phase locking while measuring seizure susceptibility.

PhaSER: a tool for real-time theta phase estimation and phase-locked manipulation

To experimentally manipulate phase locking, we developed PhaSER (Phase-locked Stimulation to Endogenous Rhythms), an open-source platform for delivering fast, flexible, and highly accurate closed-loop perturbations at any specified phase of an oscillation during behavior (Fig 2A; Fig S3). To estimate theta phase, PhaSER performs real-time zero-phase filtering for theta on a reference electrode (in our recordings, we referenced the deep pyramidal cell layer of CA1 and phase was estimated every 1ms). However, real-time filtering introduces distortions at the edge of the sampling window that must be trimmed off. An autoregressive forward prediction model⁴⁶ was therefore trained on each reference signal and then applied during the recording to extrapolate the trimmed/distorted signal. Then, the Hilbert transform was applied for phase estimation, triggering an LED for optogenetic manipulations at any specified phase (Figs S3, S4). We found that PhaSER estimates theta phase with outstanding accuracy, delivering light pulses to specified theta phases precisely and with low-latency (Fig 2B; Fig S4A–C) across a range of physiological theta powers in mice as they traverse a virtual linear track (Fig S4D).

Fig. 2. — A) Schematic of PhaSER: a deep pyramidal layer CA1 reference channel is filtered for theta (4–9Hz) before theta phase detection is applied to independently trigger red and blue light delivery at specific theta phases with low latency. See Methods and Fig S3 for details on autoregression used for forward prediction of theta phase. B) Top: Circular distribution of red light pulses in a representative animal targeted to the theta trough (left, target 170°) or peak (right, target 0°) showing accurate LED triggers centered around the target phase (mean phase of red light delivery when target is 170° = 179.3°; when target is 0° mean phase = 2.4°) with high precision (trough r-value = 0.50, peak r-value = 0.56). Bottom: polar plot showing the distribution of the mean phase of light delivery across N=13 PV-Cre and N=12 SOM-Cre mice when targeting the trough (in PV-Cre mice: 177.0°, SOM-Cre: 182.8°) and peak (PV: −7.3°, SOM: 16.0°) of CA1 theta. The trigger precision (r-value) is shown on the radial axis (PV trough: 0.47±0.04; PV peak: 0.49±0.04; SOM trough: 0.57±0.05; SOM peak: 0.47±0.04). No difference in precision of trough or peak targeted stimulation in PV-Cre compared to SOM-Cre mice (two-sample t-tests, p>0.05). C) Representative traces showing excitatory red light targeted to the trough (left) or peak (right) of theta with inhibitory blue light targeted ± 90° around the opposing phase. Raw signal shown on top, with 5–12Hz theta filtered trace on bottom. D) Bidirectional manipulation of spiking in a somBiPOLES expressing (i.e., opto-tagged, or PV+) neuron in the dentate gyrus of a PV-Cre mouse. Left: average spike waveform of optotagged PV+ interneuron across neighboring electrode channels. Right: Red light increases spiking with low latency (<4ms) while blue light suppresses spiking in the same neuron. **E-F)** Mean firing profiles relative to CA1 theta of individual opto-tagged cells from a PV-Cre (E) and SOM-Cre (F) mouse during baseline (blue), trough (purple), and peak (pink) targeted stimulation periods. Peak Stim (red excitatory light targeted to peak, blue inhibitory light delivered around the trough) shifted preferred firing to the peak of theta and reduced firing at the trough, while Trough Stim (red excitatory light targeted to the trough, blue inhibitory light around the peak) enhanced preferred firing at the trough of theta.

Bidirectional manipulations shift the theta phase preference of PV+ and SOM+ cells

To validate PhaSER’s ability to control theta phase locking, we first tested if applying phase-specific excitatory blue light to ChR2-expressing PV+ or SOM+ cells could alter their theta phase locking profiles. We found that exciting PV+ or SOM+ cells at various theta phases produced additional spikes at the targeted phase but did not reduce firing at the initially preferred phase (Fig S5). We therefore tested whether a bidirectional manipulation, with excitation and inhibition locked to opposing phases of theta, could successfully shift the phase preference of opsin+ neurons. To do this, we expressed a Cre-dependent soma-targeted BiPOLES opsin⁴⁷ in the dorsal hippocampus of PV-Cre and SOM-Cre mice (Fig S6). BiPOLES permits bidirectional manipulation of spiking, with red light inducing spiking via Chrimson, a red-shifted channelrhodopsin variant, and blue light inhibiting spiking via GtACR2, a blue light-activated chloride channel (Fig 2D). To test if this bidirectional manipulation was capable of shifting the phase preference of BiPOLES-expressing PV+ and SOM+ cells, we triggered a 10ms excitatory red light pulse at the target phase (e.g., trough or peak) and triggered a 180° window of inhibitory blue light centered around the opposite phase of theta (e.g., peak or trough, respectively; Fig 2C). This pattern of theta-locked dual light manipulation was indeed capable of shaping the firing profiles of BiPOLES-expressing PV+ and SOM+ cells. In both PV+ (Fig 2E) and SOM+ (Fig 2F) interneurons, peak excitation with trough inhibition (Peak Stim) was able to shift their preferred firing phase to the peak of theta while reducing firing at the trough (Fig 2E–F). Meanwhile trough excitation with peak inhibition (Trough Stim) enhanced their trough preference (Fig 2E–F). Thus, this bidirectional approach was able to fully control the preferred phase of these subpopulations of inhibitory neurons in vivo during awake behavior, enabling us to directly test the causal role of theta phase locking in PV+ and SOM+ inhibitory neurons.

Oscillatory power and theta phase are not altered by phase-specific PV+ or SOM+ cell manipulations

Some inhibitory neurons in the dorsal hippocampus have been shown to contribute to the local generation of theta^48,49 (but see also^50,51) or reset the theta phase upon activation⁵². If our inhibitory phase locking manipulations were to alter the theta oscillation itself, this would not only impact PhaSER’s ability to perform accurate phase estimation but would also impact the interpretation of our experimental results. We therefore tested whether manipulating DG PV+ or SOM+ cell theta phase locking impacted the power or phase of the endogenous oscillatory activity broadly, and in the theta band in particular. To test this, we first compared oscillatory power during running bouts without stimulation (baseline) to running bouts where trough- or peak-locked bidirectional stimulations were applied (Fig 3A,E). Notably, we did not detect a change in local field potential (LFP) power (1–100Hz) in either the DG or CA1 during trough- or peak-locked manipulations of PV+ or SOM+ cell activity (Fig 3A,E; Fig S7). To further confirm that theta was not impacted by these phase-specific manipulations, we compared broadband theta (5–12Hz) in the referenced layer (CA1) during baseline, trough-, or peak-targeted stimulation periods and found no significant change in power when either PV+ cells (Fig 3B) or SOM+ cells (Fig 3F) were manipulated.

Fig. 3. — **A-H)** No change in oscillatory power or theta phase during DG PV+ cell (A-D) or SOM+ cell (E-H) trough- or peak-targeted manipulations compared to baseline. **A,E)** No changes in power spectral density locally in the DG granule cell layer with PV+ cell theta phase manipulations [A, left; two-way repeated measures ANOVA, no significant effect of manipulation (baseline vs trough- vs peak-stim, p=0.688) and no significant interaction between frequency and manipulation (p=0.279), N=10] or SOM+ cell theta phase manipulations [E, left; two-way repeated measures ANOVA, no significant effect of manipulation (baseline vs trough- vs peak-stim, p=0.609) and no significant interaction between frequency and manipulation (p=0.692), N=5]. No changes in power spectral density in the CA1 pyramidal cell layer with PV+ cell theta phase manipulations [A, right; two-way repeated measures ANOVA, no significant effect of manipulation (baseline vs trough- vs peak-stim, p=0.123) and no significant interaction between frequency and manipulation (p=0.138), N=10] or SOM+ cell manipulations [E, right; two-way repeated measures ANOVA, no significant effect of manipulation (baseline vs trough- vs peak-stim, p=0.123) and no significant interaction between frequency and manipulation (p=0.138), N=10]. **B,F)** No significant change in theta band power (5–12Hz) in the theta reference channel (i.e., CA1 deep pyramidal cell layer) between baseline, trough-, and peak-stimulation conditions when manipulating PV+ cells (B, one-way repeated measures ANOVA, p=0.956, N=10) or SOM+ cells (F, one-way repeated measures ANOVA, p=0.589, N=7). **C,G)** Filtered theta from the CA1 pyramidal cell reference layer from representative animals, aligned to 100 red light pulses (10ms duration) showing no theta phase reset when stimulating PV+ cells (C) or SOM+ cells (G). Randomly selected individual traces are colored in shades of blue, mean of all traces is in white. Note that this signal has been filtered for theta (5–12Hz) for visualization purposes, but analysis in D,H is on unfiltered data. **D,H)** The LFP envelope was calculated from down-sampled, unfiltered data aligned to light pulses, as in C and G, on a 50ms window starting 250ms before or 250ms after light delivery. No significant change in LFP envelope was noted before versus after stimulation of either PV+ cells (paired t-test, p=0.672, N=14) or SOM+ cells (paired t-test, p= 0.189, N=11).

Finally, we tested if stimulating PV+ or SOM+ cells in the DG would reset theta phase. To test for a theta phase reset, we aligned the reference channel’s LFP to excitatory red light pulses (as in opto-tagging protocol). We then compared the envelope of the averaged LFP signal before versus after light delivery. If stimulating PV+ or SOM+ neurons was resetting the theta phase, we would expect to see an increase in the mean LFP envelope lasting more than one theta cycle (i.e., >120ms). However, we did not see a change in the mean envelope when either PV+ cells (Fig 3C,D) or SOM+ cells (Fig 3G,H) were stimulated. Together, these results suggest that manipulating DG PV+ or SOM+ cell theta phase locking does not alter theta power or theta phase in the hippocampus.

PV+ cell theta phase locking bidirectionally controls seizure susceptibility

Hippocampal theta is generated, in large part, by rhythmic excitatory inputs from the medial entorhinal cortex, including direct inputs from layer 2 stellate cells to the DG²⁷. These excitatory inputs are maximally active near the trough of theta in healthy mice and their rhythmic firing is maintained in epileptic mice¹⁷. In the healthy hippocampus, inhibition is precisely matched to the trough of theta, but in epilepsy DG inhibition is much more dispersed (Fig 1, Figs S1, S2)^16,17. We predicted that this disruption in the precise timing of inhibition weakens the relative inhibitory strength at the theta trough, making the DG vulnerable to overexcitation and increased seizure susceptibility. We therefore set out to test whether re-aligning DG inhibition to the theta trough in epileptic male mice would reduce seizure susceptibility, and whether shifting DG inhibition away from the theta trough in healthy mice would increase seizure susceptibility (Fig 4A). Due to the relatively low rate of spontaneous seizures (Fig 1E, Figs S1K, S2C), and to allow investigation of seizure susceptibility in healthy mice, we measured seizure susceptibility using an acute intraperitoneal injection of the chemoconvulsant kainic acid (KA) and measured the latency to the first seizure (Fig 4B)^53,54. We titrated the dosage of KA differently in healthy and epileptic mice so that an acute seizure would be reliably induced within ~5 minutes in epileptic mice (20mg/kg KA), and within ~10 minutes in controls (30mg/kg KA). We were then able to determine if re-aligning DG inhibitory spiking to the trough would increase seizure latency in epileptic mice, and if mis-aligning DG inhibitory spiking to the peak might shorten seizure latency in otherwise healthy mice.

Fig. 4. — A) Schematic of hypotheses. Top: in epileptic mice, Trough Stim (trough excitation with peak inhibition) re-aligns DG inhibition to the trough of CA1 theta, when input excitation is strongest. Bottom: in control mice, Peak Stim (peak excitation with trough inhibition) mis-aligns DG inhibition, so inhibition is weakest when input excitation is strongest, creating seizure vulnerability points at the theta trough. B) Experimental timeline schematic. PV-Cre and SOM-Cre mice received pilocarpine status epilepticus (SE) or saline (Control), and four weeks later were injected with Cre-dependent somBiPOLES virus into dorsal DG, and a headbar and wireless EEG transmitter were implanted. Three weeks later, an acute silicon probe recording was performed with channels spanning the dorsal hippocampus. Following a baseline period, mice were injected intraperitoneally with kainic acid and either peak-targeted or trough-targeted stimulation was applied until seizure onset. C) Latency to seizure in epileptic PV-Cre mice (left, male mice) was significantly increased compared to opsin- when PV+ cells were re-aligned to the trough of CA1 theta (Kruskal-Wallis ANOVA p=0.040; with Dunn’s post hoc tests comparing Trough vs Opsin-, p=0.041; and Peak vs Opsin-, p>0.999). Latency to seizure in PV-Cre control mice (right, male and female mice) was significantly reduced compared to opsin- when PV+ cells were mis-aligned to the peak of CA1 theta (one-way ANOVA p=0.039 with Dunnett’s multiple comparison post hoc tests comparing Trough vs Opsin-, p>0.999; and Peak vs Opsin-, p=0.046). D) No significant effects of manipulating SOM+ cell phase locking on latency to seizure in epileptic (left, one-way ANOVA p=0.741) or control (right, one-way ANOVA p=0.273) mice.

In epileptic mice, we found that stimulating DG PV+ cells at the trough of theta (while inhibiting at the peak) significantly increased the latency to KA-induced seizure compared to opsin-negative (opsin-) animals, while stimulating PV+ cells at the peak of theta (and inhibiting at the trough) had no effect on seizure latency (Fig 4C, left). Notably, this effect was only seen when manipulating PV+ cell firing in epileptic male mice, and not in epileptic female mice (Fig S8A), potentially because epileptic females did not have severely disrupted theta phase locking (Fig S2A,B,D). Meanwhile, we found no change in seizure duration with peak or trough stimulation (Fig S8C–D), and neither circadian or multidien rhythms, nor time since epileptogenesis could explain the variability in seizure latency (Fig S8E). Thus, correcting DG PV+ interneuron theta phase locking is sufficient to reduce seizure susceptibility in the epileptic hippocampus.

We next tested if mis-aligning DG PV+ interneuron theta phase locking could exacerbate seizure susceptibility in otherwise healthy animals. Here, we combined sexes as we did not see differences in DG inhibitory theta phase locking between healthy male and female mice (Fig S2A,B; though a breakdown by sex can be found in Fig S8A,B). In healthy mice, we found that mis-aligning DG PV+ cell spiking to the peak of theta significantly reduced the latency to KA-induced seizure compared to opsin- animals, while stimulating PV+ cells at the trough had no effect on seizure latency (Fig 4C, right). Notably, this effect was specific to DG PV+ interneurons as manipulating theta phase locking of DG SOM+ neurons had no effect on seizure latency in healthy or epileptic mice (Fig 4D). While the lack of effect in the epileptic SOM-Cre animals is perhaps expected given the well-documented loss of DG SOM+ cells in TLE (Fig 1K–M)^40–42 (but see also⁵⁵), the lack of effect in healthy SOM-Cre animals suggests that SOM+ interneurons have a less powerful role in regulating DG excitability. Altogether, these results demonstrate that PV+ interneurons are a powerful regulator of hippocampal excitatory-inhibitory homeostasis, and PV+ interneuron theta phase locking is a causal mediator of seizure susceptibility.

DISCUSSION

We found that dentate gyrus PV+ interneuron theta phase locking can bidirectionally control seizure susceptibility in both healthy and epileptic mice. DG inhibitory neurons, particularly PV+ and SOM+ cells, are critical mediators of hippocampal processing, selecting which incoming signals are passed on for further processing^24,32. In healthy mice, we found that PV+ cells provide highly precise inhibition at the trough of theta, while SOM+ neurons had more dispersed firing across theta phases. This suggests that PV+ interneurons are particularly important for balancing excitatory inputs from MEC that arrive near the trough, while the more dispersed firing of SOM+ interneurons suggests a broader role in dentate function^35,56,57. In epileptic mice, alterations in DG inhibitory connectivity and function^{16,36,37,40,42,55} are likely to have an avalanche of consequences. Indeed, in epileptic mice, DG inhibitory neurons had substantially altered theta phase locking, with much more dispersed firing compared to healthy mice, particularly in males. Furthermore, optogenetically identified PV+ interneurons were specifically altered, while SOM+ interneurons could not be recorded, presumably due to cell death of this population in the DG. We therefore predicted that altered PV+ interneuron phase locking may directly contribute to seizures in epilepsy.

To test this hypothesis, we developed, validated, and utilized a new tool, PhaSER, to control the phase locking of PV+ interneurons unilaterally during behavior and tested their causal role in seizure susceptibility. We found that re-aligning PV+ interneuron firing to the trough of theta was able to reduce seizure susceptibility in male epileptic mice, while disrupting normal PV+ interneuron phase locking in healthy mice increased seizure susceptibility. Together, these data provide the first causal evidence that DG PV+ cell theta phase locking directly and powerfully influences excitatory-inhibitory homeostasis in the hippocampal circuit and can bidirectionally control seizure susceptibility. This finding is especially compelling because we were manipulating a relatively small population of inhibitory neurons (~75% of PV+ cells in the dorsal DG expressed opsin, and manipulations were performed unilaterally), yet we saw robust changes in susceptibility to seize after systemic kainic acid.

Surprisingly, we found that altered inhibitory spike timing was much more dramatic in epileptic male mice than epileptic female mice, despite them having similar seizure frequencies. This suggests that the mechanisms of seizure susceptibility in male and female epileptic mice may be different. Here we found clear evidence that male epileptic mice have disrupted inhibitory phase locking in DG, and that restoring PV+ interneuron phase locking to the trough of theta reduces seizure susceptibility (Fig 4). However, in female epileptic mice, where inhibitory phase locking is less disrupted (Fig S2), we found that re-aligning PV+ interneuron activity had no effect (Fig S8A). This null result is similar to what we found in healthy animals that received trough stimulation (Fig 4C). In both cases, there was already intact inhibition near the trough of theta, and adding further inhibition at this phase had no effect on seizure susceptibility. Therefore, disrupted inhibitory phase locking is sufficient to induce seizure susceptibility, but is not always necessary for seizure development. Indeed, epilepsy is a complex and heterogeneous disorder, as illustrated by the many neurological and psychiatric disorders with comorbid seizures and the variability in treatment outcomes with anti-epileptic drugs. Thus, a personalized approach may be needed across patients to treat seizures.

Though theta phase locking has been widely observed and studied, interrogating its causal role in network function is not possible with traditional approaches. Technical advances in large-scale electrophysiological data collection and real-time signal processing have opened the door for closed-loop, real-time, phase-specific manipulations^3,58–63. Here, we describe and validate PhaSER, our open-source, user-friendly platform that allows fast, flexible, and highly accurate delivery of perturbations at any specified phase of an oscillation. The PhaSER set up is unique because it estimates theta phase continuously (i.e., it does not solely detect maxima or minima), it triggers manipulations with outstanding accuracy, and when combined with artifact-free silicon probes it permits continuous monitoring of single-unit activity during stimulations. Here we demonstrated that PhaSER can precisely alter the theta phase locking of DG interneurons with excitatory and inhibitory stimulation patterns applied to target the theta trough or theta peak. Notably, trough and peak stimulation paradigms are virtually identical, with the same amount of red and blue light delivered, except that each is shifted by half a theta cycle (only ~60ms). Despite this, these stimulation paradigms produced opposite effects on seizure susceptibility. This is powerful evidence that the timing of neural activity is a critical factor mediating behavior and that temporally agnostic stimulations may muddy experimental interpretability and produce mixed results in the clinic and at the bench. It is our hope that the open-source release of PhaSER will reduce hurdles to performing stimulations that are driven by the brain’s real-time endogenous activity.

Because seizures can serve as clear readouts of pathological network activity, the epileptic hippocampus is ideal for causal investigations of cell-circuit interactions such as theta phase locking. However, phase-locked spiking is also disrupted in models of substance use, schizophrenia, traumatic brain injury, autism, and Alzheimer’s disease^15–22. Notably, all of these disorders also have increased risk for seizures^64–69. This suggests that phase locking disruptions may be a relatively universal component of network dysfunction, including seizure susceptibility, across diseases. Therefore, our findings expand our fundamental understanding of inhibitory theta phase locking by directly demonstrating their causal role in seizure susceptibility, and inform the development and expansion of oscillation-driven neuromodulatory therapies aimed at restoring network function in epilepsy and beyond^70–75.

MATERIALS & METHODS

All experimental protocols were approved by the Icahn School of Medicine at Mount Sinai’s Institutional Animal Care and Use Committee, in accordance with the US National Institutes of Health guidelines.

Animals

The SOM-Cre and PV-Cre mice used for experiments were bred and maintained in-house. SOM-Cre mice (Jackson Laboratory strain #013044; Sst^{tm2.1(cre)Zjh}/J) were maintained homozygous⁷⁶. The initial PV-Cre mice used for opto-tagging experiments (Fig 1) were maintained homozygous (Jackson Laboratory strain #017320; B6.129P2-Pvalb^tm1(cre)Arbr/J)⁷⁷. Later (Figs 3–4) we switched to maintaining this strain heterozygous by crossing male PV-Cre mice with female C57BL/6 (Charles River) mice to improve pilocarpine-induced status epilepticus success rates. For confirmation of viral specificity in mice without Cre expression, C57BL/6 (Charles River) mice were purchased directly from the vendor and allowed to habituate for at least a week before saline or pilocarpine treatment. Male and female mice were used for all experiments. Mouse sex was identified based on their external genitalia at the time of weaning (postnatal day 21). Data from male and female mice are combined except where noted. All mice were housed in standard housing conditions (12-hour light, 12-hour dark) in the animal facility at the Icahn School of Medicine at Mount Sinai. Animals were group-housed with their littermates when possible, or with an ovariectomized female (Jackson Laboratory strain# 000691; 129X1/SvJ) when littermates were not available or compatible. Animals were allowed ad libitum access to food and water except when water-restricted for training on the virtual linear track.

Epileptogenesis with Pilocarpine

Chronic temporal lobe epilepsy was induced using the pilocarpine-status epilepticus (pilo) model as previously described^16,17, with modifications to account for mouse strain differences. At ~9 weeks of age, mice were injected with scopolamine methyl bromide (intraperitoneal, 1mg/kg, Sigma-Aldrich, Cat# S8502) to reduce the peripheral effects of pilocarpine. 30 minutes later, mice received intraperitoneal injections of pilocarpine hydrochloride (Sigma-Aldrich, Cat# P6503) to induce status epilepticus, or an equivalent volume of saline (Fresenius Medical Care, Cat# 060–10109) for control (i.e., healthy) mice. Initial pilocarpine dosage and booster protocols varied by strain and animal sex, each optimized to increase SE induction success rates. C57BL6 mice received initial pilocarpine dosages determined by their weight class (<20g mice received 285mg/kg, 20–25g mice received 275mg/kg, and >25g mice received 260mg/kg), with boosters of 50–100mg/kg delivered after 45 minutes if the animals had not seized. SOM-Cre mice received an initial pilocarpine dose of 365mg/kg for males and 345mg/kg for females, with boosters for both sexes (~33% volume of initial dose) delivered 30–45 minutes after the initial dose as needed until the mouse began having seizures. Additional boosters were given if the time between seizures was greater than 10 minutes. Homozygous PV-Cre mice received an initial pilocarpine dose of 400mg/kg, with boosters (~25% volume of initial dose) delivered 45–50 minutes after the initial dose as needed, with follow-up boosters every 20 minutes until SE was reached. Heterozygous PV-Cre mice received an initial pilocarpine dose of 285–315mg/kg, depending on experimenter, with boosters (~25% volume) delivered 45–50 minutes after the initial dose as needed. We found that SE-success rates in male heterozygous PV-Cre mice were particularly sensitive to experimenter sex, with male mice ~2.3x more likely to enter SE when a male experimenter was present (or when exposed to male experimenter scent via KimWipe or cage change the night before epileptogenesis).

SE was established when animals entered a continuous seizure and didn’t react to outside stimuli. Pilocarpine-treated mice were left in SE for 2 hours, after which they received diazepam (20mg/kg, i.p.; Dash Pharmaceuticals, NDC 69339-137-01) to terminate SE. Saline-treated animals were given equivalent injections of saline (rather than pilocarpine) and diazepam time-locked to the pilocarpine-treated mice. Pilocarpine-treated mice who didn’t reach SE were excluded from the study.

Stereotaxic Surgery

The intrahippocampal virus delivery, headbar implant, and (where applicable) the wireless EEG implantation was performed during the same surgery under 1–3% isoflurane anesthesia. First, the skull was exposed and a small burr hole was drilled above the viral delivery site: 2mm posterior to bregma and 1.35mm to the right of bregma. A Nanoject III (Drummond Scientific Company) with glass capillary loaded with virus was lowered into the burr hole until the tip was 2.1mm ventral to bregma to target the DG. Virus was injected at a rate of 2nL/s. Once the injection was complete, the syringe was left in place for a minimum of 3 minutes. Then, for experiments where animals received viral injections into CA1 as well, the syringe was raised to 1.5mm ventral from bregma where an additional bolus of virus was injected in the same manner.

To determine baseline firing profiles of optogenetically-identified PV+ or SOM+ cells in the DG of control and epileptic mice, and monitor responses to theta phase-locked manipulations, a Cre-dependent channelrhodopsin virus (ChR2; AAV1-EF1a-double floxed-hChR2(H134R)-EYFP-WPRE-HGHpA; Addgene cat #20298; 1.2×10¹³ GC/mL; gifted by Karl Deisseroth) was injected into the dorsal hippocampus (120nL per site, DG and CA1). To test how theta phase locking of DG PV+ or SOM+ cells impacts seizure susceptibility, a Cre-dependent soma-targeted BiPOLES virus (somBiPOLES; AAV1-hSyn-DIO-somBiPOLES-mCerulean, AAV construct produced by Virovek from Addgene plasmid #154951, 2.0×10¹³vg/mL, 200nL)⁴⁷ or Cre-dependent fluorophore-only control virus (GFP; AAV1-hSyn-DIO-eGFP, Addgene #50457, 2.2×10¹³ GC/mL, 200nL; gifted by Bryan Roth) was injected into the DG.

Following viral infusion, a stainless steel headbar was fixed onto the skull of the mouse. Lidocaine (2%, ~0.03mL) was first injected subcutaneously over the skull, and the epaxial muscles along the neck were cleared from the skull’s surface. The mouse skull was thoroughly scored and then was stereotactically aligned to the headbar before the headbar was fixed near the surface of the skull with cyanoacrylate glue and dental cement (Lang Dental). Dental cement was built up to create a well around the exposed skull which was then filled with Kwik-Sil (World Precision Instruments). Once dried, the Kwik-Sil was covered with a final layer of dental cement. Meloxicam (5mg/kg) or Carprofen (5mg/kg) was administered subcutaneously during and for 2 days following the surgery together with a 7-day course of ampicillin (20mg/kg).

Viral Validation

To confirm that the Cre-dependent viruses used in these studies would not express in the absence of Cre, control and epileptic C57BL6 mice were injected with Cre-dependent ChR2 (N=4 control, N=6 pilo), Cre-dependent somBiPOLES (N=3 control, N=3 pilo) or Cre-dependent fluorophore-only (N=3 control, N=3 pilo) viruses targeting dorsal DG and CA1. At least 20 days following viral injection, mice were deeply anesthetized with isoflurane (~5%, Isospire), quickly decapitated, and their brains were dissected and dropped in 4% paraformaldehyde (Wako) and kept at 4°C for ~24 hours. 50μm thick coronal sections spanning the entire hippocampus were prepared in 1x phosphate buffered saline (PBS; Fisher BP399) on a vibratome (Leica VT1000S). Every section containing hippocampus was mounted for fluorescent imaging with DAPI (SouthernBiotech DAPI Fluoromount-Gm, 0100–20) when compatible with viral expression (i.e., when visualizing the ChR-eYFP or eGFP signals), and mounted without DAPI (Invitrogen ProLong Gold P36930 or Vector Laboratories Vectashield H-1000) when the viral fluorophore was not compatible with DAPI (i.e., when visualizing the somBiPOLES-mCerulean virus). Every section was then viewed and imaged (Leica DM6B). No expression was noted in any of these animals, confirming that these Cre-dependent viruses do not express in the absence of Cre.

To validate Cre-dependent viral expression against immunohistochemistry, animals (N=3 PV-Cre control, N=3 SOM-Cre control, N=2 PV-Cre pilo, N=2 SOM-Cre pilo) were injected with AAV1-hSyn-DIO-somBiPOLES-mCerulean into the dorsal DG. 9–10 weeks after viral injection, they were deeply anesthetized with isoflurane before being perfused with ice-cold 1x PBS and 4% paraformaldehyde. Their brains were dissected and incubated in 4% paraformaldehyde at 4°C for <24 hours before being transferred to 30% sucrose in PBS. After brains were saturated with sucrose (~2 days), they were frozen at −80°C until sectioning at 40μm on a cryostat (Leica CM3050S).

Immunohistochemistry was performed on every fourth section to stain for either somatostatin (SOM; rat anti-somatostatin, EMD Millipore MAB354, 1:300) or parvalbumin (PV; rabbit anti-parvalbumin, Swant PV27a, 1:1000) in red (goat anti-rat Alexa Fluor 555, Invitrogen A21434, 1:500 or goat anti-rabbit Alexa Fluor 568, Invitrogen A11036, 1:500, respectively). Sections were then mounted with Vectashield H-1000 and imaged on a confocal microscope (Zeiss LSM780). Colocalization of virus with immunofluorescence was determined by an experimenter blinded to genotype and experimental group. Manual counts of mCerulean-labeled cells were performed and comparing against PV or SOM immunofluorescence in three sections per mouse: anterior to the injection site (1.8mm posterior to bregma), at the injection site (2mm posterior to bregma), and posterior to the injection site (2.2mm posterior to bregma).

In addition, viral expression was verified alongside probe tract imaging for all recorded animals: virally expressing cell bodies were confirmed in the strata and cell morphologies we would expect (i.e., Cre-dependent viral expression in PV-Cre mice was primarily found in somata located on the hilar border of the granule cell layer, with dense axons throughout the granule cell layer, and expression in SOM-Cre mice was primarily found in somata located in the hilus, with axons densest in the molecular layer) and animals without viral expression were excluded from further analysis. No animals, control or epileptic, showed unexpected or inappropriate labeling.

Chronic Seizure Monitoring

To chronically monitor seizures, a wireless EEG transmitter (PhysioTel ETA-F10; Data Sciences International) was implanted under the skin, posterior to the left ribcage, but anterior to the left haunch. The EEG transmitter leads were fixed between the skull and dura in small burr holes. The signal wire was fixed with cyanoacrylate glue under the left frontal bone, and the ground wire was fixed under the right side of the interparietal bone. EEG signals were monitored via Ponemah and NeuroScore software platforms (Data Sciences International), and seizures were manually identified. Pilocarpine-SE epileptogenesis produces highly stereotyped behavioral seizures that can be easily identified in the EEG signal by eye, and we have previously performed blinded video monitoring to confirm that these EEG seizures correspond to behavioral seizures¹⁶. EEG data from N=10 control, saline-treated mice (combined genotypes) was assessed by a blinded experimenter, and no seizures were found. Therefore, to increase throughput in follow-up experiments (Fig 4, Fig S8), some control mice were implanted with a sham transmitter (ETA-F10 Training Module, Data Sciences International).

Virtual Reality (VR) Training

In the days following surgery, mice were gently handled for 5 minutes daily. Once recovered from the surgery and fully habituated to being handled (~3 days), mice were introduced to head-fixation (5 minutes per day for ~3 days). Mice were then water restricted and maintained at a body weight of ~82% of their initial weight while they continued the rest of their daily training (as in¹⁶). During water restriction, mice were weighed and monitored daily for signs of dehydration, fatigue, or infection and were immediately given free access to water if any signs of distress were observed. After water restriction began, mice were introduced to head-fixation while standing/walking on a spherical treadmill locked to rotate on only one axis. Once the mice had gained coordination and strength on the spherical treadmill (usually ~3 days of training for 10–20 minutes per day), mice were trained to lick from a water port that delivered a 4μL drop of water per lick. Once the mice had learned to reliably earn water from the water port, they were trained to run along virtual linear tracks of increasing length. For animals expressing ChR (included in Figs 1, S1, S2, S4, and S5), water rewards were delivered automatically at the end of the track, before the mice were teleported back to the beginning of the track. For animals expressing BiPOLES (included in Figs 2, 3, and S7), mice had to stop inside a reward zone to earn water⁷⁸. Note that animals used for testing acute seizure susceptibility (Figures 4, S8) were not trained in VR. The virtual linear tracks were created with ViRMEn⁷⁹ and were displayed across three flat monitors angled around the front of the spherical treadmill. Once the mice were consistently earning their daily water within 1 hour on the longest (~2 meter) virtual track, they were prepared for acute silicon probe recording. Virtual track training took ~5 days.

Craniotomy & Ground Implantation Surgery

The day prior to an acute silicon probe recording, a craniotomy and ground implantation was performed. First, the most superficial layer of dental cement was drilled off and the Kwik-Sil was removed, exposing the skull. Then a burr hole was drilled above the left hemisphere of the cerebellum, and an Ag/AgCl-coated silver reference wire (Warner Instruments) was slipped between the skull and dura and fixed with cyanoacrylate glue and dental cement. A 1.5mm diameter craniotomy was also drilled over the right hippocampus at this time (centered on the following coordinates: 2mm posterior, 1.45mm right from bregma). The craniotomy was covered with buffered artificial cerebrospinal fluid (ACSF; in mM: 135 NaCl, 5 KCl, 5 HEPES, 2.4 CaCl2, 2.1 MgCl2, pH 7.4)^16,80 and the exposed skull was again covered with Kwik-Sil. Mice were returned to their home cages overnight after recovering on a heating pad.

Acute Silicon Probe Recordings & Recording Hardware

The day following craniotomy, mice were set up for acute silicon probe recordings [15.5 ± 0.5 weeks of age (109 ± 4 days of age; 19 ± 1 day after initial surgery)]. First, a 64 channel Cambridge NeuroTech (ASSY-77-H3) silicon probe (single shank, 8mm length, sharpened tip) with attached lambda-b optic fiber 100μm core (tapered 1.2mm) was painted with dye (Invitrogen, Vybrant DiI, V22885) so the probe track could be visualized later. Mice were then head-fixed on the spherical treadmill and the Kwik-Sil covering the skull was replaced with a buffered ACSF solution. Head-flat skull positioning in the dorsal-ventral and medial-lateral directions was then established to within 50μm (bregma-to-lambda) before the silicon probe was lowered into the right dorsal hippocampus at the following approximate coordinates: 2mm posterior, 1.45mm lateral, 2.1mm ventral to bregma. After reaching the target depth and confirming probe location with electrophysiological signatures of the dorsal hippocampus, mineral oil was placed over the buffered ACSF solution and the probe was allowed to settle for 1 hour before starting the virtual linear track and recording^16,81.

Electrophysiological signals were sent to an Intan headstage (RHD 64-Channel Recording Headstage, Intan Technologies) via a Samtec to Omnetics adaptor (ADPT A64-Om32×2, Cambridge NeuroTech) for pre-amplification and digitization. An Intan recording controller (RHD2000 Intan 1024ch Recording Controller, Intan Technologies) collected and logged signals at a 25kHz sampling rate from each electrode channel as well as from analog inputs reflecting the speed of the spherical treadmill, position on the virtual linear track, time of water delivery, time of the mouse’s licking behavior, and LED trigger and delivery times. Because the Intan recording controller logs data to the PC over a USB connection, which is relatively slow, we selected a single electrode channel in the deep CA1 pyramidal cell layer (as identified by its location relative to a stereotypical theta phase shift and higher density of spikes) to route as an analog output going to a separate, low-latency, PCIe data acquisition device (NI-DAQ PCIe-6321, National Instruments).

For recordings on mice expressing ChR2 (Figs 1, S1, S2, S4, S5): mice ran down the virtual linear track at least 50 times to collect a baseline before any photo-manipulations were introduced. Immediately following this baseline period, single-units were photo-tagged with 100 1Hz square pulses (10ms duration) of blue light (LED, 455nm, Prizmatix, 1mW) sent via a Pulser Plus (Prizmatix) pulse train generator. Five minutes following the photo-tagging procedure, a baseline of at least 120s was acquired to train the auto-regressive model for phase estimation (described in detail below). Photo-stimulation locked near the trough (170°) or peak (0°) was then delivered for at least 25 trials during locomotion. No phase-specific photo-stimulations were delivered during rest. The order of trough- and peak-targeted stimulations was counterbalanced across animals with at least 5 minutes between the two manipulation periods.

For VR recordings on mice expressing BiPOLES (Figs 2, 3, S7): mice ran down the virtual linear track 25 times to collect an initial baseline before any photo-manipulations were introduced. Then, trough (170°) or peak (0°) targeted red light stimulations (10ms; LED, 625nm, Prizmatix, <0.72mW), with blue inhibitory light pulses (variable duration) targeted ±90° around the opposite phase (from 270° to 90°, or from 90° to 270°, respectively), were delivered in 25 trial intervals with baseline periods in between. Trough and peak targeted manipulations were counterbalanced across animals (e.g., 25 trials baseline, 25 trials trough/peak, 25 trials baseline, 25 trials peak/trough, 25 trials baseline). These manipulations were delivered regardless of locomotion status, but single-unit and LFP analyses were limited to periods when the mice were running.

Testing Seizure Susceptibility with Kainic Acid

To test whether PV+ or SOM+ cell theta phase locking impacted seizure susceptibility, control or epileptic PV-Cre or SOM-Cre animals were randomly assigned to receive an injection of either an opsin- (AAV1-hSyn-DIO-eGFP) or opsin+ virus (AAV1-hSyn-DIO-somBiPOLES-mCerulean) in the dorsal DG. At the same time, mice underwent surgery to implant a headbar and a wireless EEG telemetry device (or sham device for some control animals) to monitor spontaneously occurring seizures. Following recovery from surgery, mice were habituated to experimenters with gentle handling for 3 days (5 minutes per day). Once habituated to the experimenters, mice were then habituated to head-fixation for an additional 3 days. At this point, mice were introduced to the Styrofoam ball used for head-fixed recordings and trained to walk on the ball for increasing durations until comfortable (minimum of 3 days). Simultaneously, mice were habituated to receiving intraperitoneal injections of 0.2mL saline while head-fixed on the ball. Once fully habituated to the experimental set-up, mice underwent craniotomy and ground implantation surgery, and the seizure susceptibility test was performed the following day. Details of the craniotomy and recording hardware are described above. Note that in a few cases, the seizure susceptibility recording was delayed by a day or two if spontaneous seizures occurred within 4 hours of the planned silicon probe recording.

To test if PV+ or SOM+ cell phase locking impacts seizure susceptibility, an acute silicon probe was lowered into the right dorsal hippocampus, with channels spanning CA1 and the DG, of control or epileptic PV-Cre or SOM-Cre mice expressing either eGFP (opsin-) or BiPOLES (opsin+) virus. Mice were recorded for 30 minutes to establish a baseline and ~2 minutes of baseline data from a channel in CA1 deep pyramidal cell layer was used to train PhaSER’s auto-regressor (details below). Mice were then injected intraperitoneally with 2mg/mL kainic acid (KA; Tocris Bioscience, 7065) mixed in saline (30mg/kg KA for control animals; 20mg/kg for epileptic animals). Immediately following KA injection, phase-specific red/blue light delivery began with excitatory red light pulses (10ms) targeted at either the peak (0°) or trough (170°) of CA1 theta, and inhibitory blue light pulses (variable duration) targeted ±90° around the opposite phase (from 90° to 270°, or from 270° to 90°, respectively). Light delivery continued until the experimenter visually confirmed electrographic seizure onset. Shortly after seizure cessation, the recording was ended and the mice were swiftly euthanized and their brains dissected for histology.

Animals were excluded from analyses if the mean phase-angle of excitatory (red) light delivery was >45° from the target phase (N=10 of 121 animals), or if the excitatory red light deliveries were not phase-specific (N=1 animal, Rayleigh test for non-uniformity of circular data). Other reasons for exclusion included poor KA injection, no seizure within 70 minutes of KA injection, seizure occurrence in the 4 hours leading up to KA injection, lack of viral expression, mis-targeted silicon probe, severe hippocampal sclerosis that prevented CA1 theta referencing, or evidence of seizure-like activity occurring before the KA injection in a control mouse.

Real-time Signal Processing & Phase-Specific Manipulation with PhaSER

In order to apply light stimulations locked to the ongoing theta oscillation in real-time, we created a custom LabVIEW program, which can be found on our lab’s GitHub (https://github.com/ShumanLab) along with user-friendly documentation. The 25kHz recorded signal from the CA1 pyramidal layer reference channel was first down-sampled to 1kHz. Then, in order to do theta phase detection, we filtered the down-sampled signal for the theta band (4–9Hz) in real-time. Because digital filters introduce phase lags and distortions, a zero-phase filter (first-order Butterworth) was employed. Zero-phase filtering is a technique in which a digital filter (in this case, an infinite impulse response (IIR) filter), which introduces a phase lag, is applied first in the forward direction and then in the reverse direction to cancel out the phase lags in the output signal. To extract the theta phase from this filtered signal, we used the Hilbert transform. The Hilbert transform can be used to calculate the phase angle of a real signal at every point in time, however both digital filters and the Hilbert transform introduce edge effects – distortions at the edge of the analysis window. To circumvent these distortions, and therefore avoid incorrectly identifying the theta phase, we removed 150 samples from the beginning and end of the filtered signal within the 1000 sample sliding filter window and then used a 13th order auto-regressive model that has been described previously⁴⁶. The auto-regressive model was trained on ~2 minutes of baseline data using the Yule-Walker method. This model was then used to reconstruct the 150 most recent datapoints without edge effects and then forward predict an additional 150 samples (which equals approximately 1 theta cycle). This places the current time point at the center of the 300-sample prediction window (Fig S3). The Hilbert transform was then applied to the resulting signal and the current phase angle was calculated from the center of the phase-estimation window, which corresponds to the real-time signal without edge effects. A schematic of this process is shown in Fig S3. To limit photo-stimulation to periods when the animal was locomoting, a movement threshold was set based on the analog signal from the spherical treadmill. When this value passed the threshold indicating locomotion, and the Hilbert transform phase angle crossed the threshold of the light delivery target phase, a TTL trigger was generated by the NI-DAQ and passed to a pulse generator (Pulser Plus, Prizmatix) which was programmed to deliver 10ms of blue or orange-red light (LED, 455nm or 625nm, Prizmatix, ~1.7mW blue light power, ~0.72mW red light power) at the target phase. The theta trough stimulation target was 170° and the theta peak stimulation target was 0°. When bidirectional manipulations were applied, a TTL step was generated by the NI-DAQ to toggle the blue LED on 90° after the target phase and off 90° before the following target phase (i.e., inhibitory blue light was applied ±90° around the phase opposite the stimulation target phase). The latency of this real-time signal processing system from electrode to light trigger is approximately 3ms, with the majority of the latency produced when the data is read from the DAQ.

Histology Following Recordings

Following recording, the silicon probe was removed from the brain and the mice were deeply anesthetized with isoflurane (5%; Isospire) before being decapitated. The brains were quickly dissected and dropped in 4% paraformaldehyde (Wako) where they stayed for ~24 hours. 50μm thick coronal sections were prepared in 1x phosphate buffered saline (PBS; Fisher BP399) using a vibratome (Leica VT1000S). All sections containing hippocampus were then mounted for fluorescent imaging with DAPI (SouthernBiotech DAPI Fluoromount-Gm, 0100–20) when compatible with viral expression (i.e., when visualizing the ChR-eYFP signal), and mounted without DAPI (Invitrogen ProLong Gold P36930; or Vector Laboratories VECTASHIELD H-1000) when the viral fluorophore was not compatible with DAPI (i.e., when visualizing the somBiPOLES-mCerulean virus). Sections were later imaged to confirm viral expression (visualized via eYFP or mCerulean) and to determine probe location (DiI). Images were taken using a Leica DM6B fluorescence microscope equipped with a Lumencor Light Engine and Leica DFC9000 GT camera. Hippocampal anatomical locations of the probe tracks and stereotaxic coordinates were compared to their electrophysiological signals and a mouse brain atlas⁸². Viral expression was confirmed and compared to expected expression patterns (e.g., expected expression of Cre-dependent virus in SOM-Cre mice was in CA1 stratum oriens and DG hilus while expected expression in PV-Cre mice was in CA1 pyramidal cell layer and on the border of DG hilus/granule cell layer) in all recorded mice.

Post-Processing & Analysis of Single-Unit and Local Field Potential Data

All data analysis was performed with custom scripts using MATLAB 2017a. Data files were first concatenated into continuous signals for each channel. To extract single-units, all channels were high-pass filtered and background subtracted for clustering into putative single-units using Kilosort2.5^83,84 and Phy2⁸⁴. This automated spike sorting pipeline reliably isolates single-units using a template-matching approach that is iteratively updated and includes drift correction, a critical step for stable cell isolation across the entire recording. Once single-units were isolated using Kilosort2.5 and manually confirmed in Phy2, we calculated the percentage of spikes in each putative cell cluster that violated a 2ms refractory period, and only continued analysis with clusters where 3% or fewer spikes violated this refractory window. We then assigned cell clusters as “inhibitory,” “excitatory,” or “unclassified” based on their firing rate, distribution of inter-spike intervals, and complex spike index (Fig S1A,D). We then tested each single-unit for its response to our photo-tagging stimulation protocol. To identify ChR2-expressing cells, we used a 1Hz 10ms blue light tagging protocol and built a spike probability distribution based on the 500ms prior to each of 100 light stimulations. If a single-unit’s spike frequency during the first 4ms of light delivery was significantly above the distribution mean (where α = 0.001), then the cell was classified as ChR2+. To identify BiPOLES-expressing cells, we used a similar 1Hz photo-tagging protocol, but tested for bidirectional responsiveness to 100 excitatory 10ms red light pulses and 100 inhibitory 20ms blue light pulses. To be identified as BiPOLES+, cells had to be both excited by red light and inhibited by blue light. A cell was classified as ‘excited by red light’ if the spike probability was significantly higher during the 10ms red light delivery than during the 500ms leading up to each light stimulus (α = 0.05), and was classified as ‘inhibited by blue light’ if the spike probability during the 20ms of blue light delivery was significantly lower than the 500ms prior (α = 0.05), or if there were fewer spikes in the 20ms blue light stimulation period than in any of the equivalently sized bins in the 500ms prior. Due to our histological confirmation that neurons labeled with our Cre-dependent viruses in healthy and epileptic PV-Cre mice co-localized with PV immunofluorescence, and those labeled in healthy and epileptic SOM-Cre mice co-localized with SOM immunofluorescence (and did not co-localize with PV immunofluorescence), we will refer to opto-tagged cells in the DG of PV-Cre mice as PV+ cells and opto-tagged cells in the DG of SOM-Cre mice as SOM+ cells. Putative inhibitory cells and opsin-expressing PV+ and SOM+ cells in the DG were then assessed for their phase preference during the baseline period as well as their response to the trough and peak phase-specific manipulations. Phase locking was quantified by an r-value (magnitude of phase preference) and a mu-value (circular mean phase of firing) relative to the deepest pyramidal layer channel (i.e., the pyramidal cell layer channel closest to the stratum oriens). Note that mu-values are only reported for cells that were significantly phase-locked during baseline (Rayleigh’s test for non-uniformity), whereas r-values are reported for cells regardless of the strength of their phase preference.

For local field potential (LFP) analyses, signals were down-sampled to 1kHz, bandpass filtered (theta: 5–12Hz) using zero-phase digital filtering and LFP power was compared using the Chronux library^85,86. Power spectral density (PSD) was calculated from the unfiltered, down-sampled LFP at each electrode position during running bouts that were at least 3 seconds in duration. A spectrogram was constructed by taking the mean PSD across running bouts within each frequency bin (1Hz bin size) and electrode channel. Sublayers of the hippocampus were identified from histology showing location of the dye-coated probe and electrophysiological markers, including peak theta and ripple power, theta phase shifts, gamma coherence, and density of action potentials^87–90. Channels from each layer were aligned and spectrograms for each stimulation condition were averaged across animals (Fig S7).

To determine if stimulating SOM+ or PV+ neurons reset the theta phase, we aligned the down-sampled LFP from the deepest CA1 pyramidal cell layer channel to the onset of red light pulses (10ms duration) delivered at 1Hz. The LFP was averaged across 100 light deliveries and the mean envelope was calculated during a 50ms wide bin starting 250ms before and after the aligned light deliveries. With the exception of the opto-tagging identification and theta phase reset analyses, during which the mice were spontaneously running and resting, all other single-unit and LFP power analyses were limited to periods when the animals were running.

Automated Detection of KA Injection, Seizure Onset, and Seizure Duration

Seizure latency was quantified as the amount of time between KA injection and seizure start time using data from one channel at the DG hilus/upper granule cell layer border as a reference. The time of injection, seizure start, and seizure end were identified with custom MATLAB scripts, which were compared against recording notes and confirmed via manual inspection by a blinded experimenter.

To identify the time of KA injection, we detected the end of the signal artifact that was introduced by touching the mouse. To do this, we filtered the DG reference channel for 57–63 Hz and then calculated line length values over a 1 second sliding window for the recording duration. The line length values from the baseline period (first 30 minutes) established a threshold using a method similar to prior work⁹¹. Briefly, the threshold was initially defined as 2 standard deviations above the median baseline line length. Each baseline line length value was then compared to this threshold, and a “hit” was counted if the value met or exceeded the threshold. If any hits were recorded, the number of standard deviations (SDfactor) was increased by 0.5 and hit testing performed again. This process was repeated until there were no hits in the baseline period, and the injection detection threshold was subsequently set to that SDfactor. To detect the injection, we compared the line length values in the post-baseline injection time window to this threshold. If no threshold crossings were detected, the SDfactor was decreased by 0.5 and the test was repeated again. The last threshold crossing was considered the injection time, corresponding to the last time in the injection window when 60Hz noise was higher than baseline.

The time of seizure onset (used to determine seizure latency from KA injection) was detected using a similar hit testing strategy, however line lengths were calculated from unfiltered LFP data from the DG reference channel and a moving average filter was applied to prevent short, non-seizure events (e.g., interictal spikes) from triggering seizure detection. For the majority of recordings (89 of 118), the baseline period was used to set the crossing threshold. However, in about 25% of cases (29 of 118), LFP dynamics after the injection were considerably altered compared to baseline activity, causing false seizure onset detection times. These changes in the LFP can be largely attributed to changes in movement (i.e., some mice changed their locomotion status in response to the injection). In these cases, a variable duration “post-injection/pre-seizure” period (10 seconds after the detected injection end time to 60 seconds before the manually recorded seizure start time) was used to set the detection threshold instead. Seizure onset was then identified as the first upwards threshold crossing following the injection end time. No significant differences in seizure latency were found between opsin-animals of the same genotype and experimental condition that received either peak- or trough-targeted stimulation (unpaired t-tests, p>0.05), and therefore these stimulation paradigms were collapsed into their respective PV-Cre Control/PV-Cre Epileptic/SOM-Cre Control/Som-Cre Epileptic Opsin- groups for follow-up analyses.

The end of the seizure (used to calculate seizure duration), was identified as the first time following seizure onset when the signal line length dropped and stayed below 66% of threshold for a minimum of 5 seconds. In six cases, manual inspection by a blinded experimenter did not agree with the detected seizure end. In five of these six cases, the percent reduction was set manually at either 33%, 50%, or 90% threshold. In the remaining case, no seizure duration was calculated as the signal suggested the mouse continuously seized until the end of the recording.

When latency to seizure or seizure duration was normally distributed (Shapiro-Wilk test, p>0.05) and standard deviations were not significantly different between trough/peak/opsin- groups (Brown-Forsythe test, p>0.05), ordinary one-way ANOVAs were used to establish whether there was a significant difference between the stimulation paradigm (trough/peak/opsin-) means, with post hoc Dunnett’s multiple comparisons tests where appropriate. However, if latency to seizure or seizure duration was not normally distributed, a Kruskal-Wallis ANOVA was performed, with post hoc Dunn’s multiple comparisons tests where appropriate.

Statistics

All statistics were performed in MATLAB 2017a or GraphPad Prism. For circular data, all phases were converted to radians and were analyzed using the Circular Statistics Toolbox in MATLAB⁹². Linear data is presented as mean ± standard error of the mean while circular data is presented as circular mean ± standard deviation. For comparisons of mean preferred phase preference (mu), Watson-Williams and circular k-tests were performed post-hoc to test for a change in mean phase preferences or concentration of phase preferences only if the Kuiper test showed a significant difference. Furthermore, only neurons that showed significant phase modulation (Rayleigh’s test for non-uniformity) were included in mu comparisons. Threshold for significance (α) was set to 0.05 unless otherwise noted. n=cells, N=mice. In figures * indicates p<0.05, ** indicates p<0.01, and *** indicates p<0.001 unless otherwise noted.

Supplementary Material

NIHPP2025.09.10.675457V1-supplement-1.pdf^{(2.2MB, pdf)}

ACKNOWLEDGEMENTS

We would like to wholeheartedly thank the Shuman and Cai lab groups for providing their support and thoughtful feedback throughout the course of this project. In particular, we would like to thank Alia Abdelhameed and Helen Liu for assistance with animal training and tissue processing, Lucia Page-Harley for methodological set-up and support, and Lingxuan Chen, Nadia Khan, William Mau, Zachary Pennington, Justin Lines, Yosif Zaki, and Brian Sweis for helpful conversations that shaped this project. Corin Humphrey passed away in October 2020 after making substantive methodological contributions to this project. She agreed to be an author on this project before her death.

Funding:

National Institutes of Health grant F32NS116416 (ZCW)

American Epilepsy Society Postdoctoral Research Fellowship (ZCW)

Simons Collaboration on Plasticity and the Aging Brain Transition to Independence Award (ZCW)

Friedman Brain Institute Postdoctoral Innovator Award (ZCW)

National Institutes of Health grant F31NS134301 (PAP)

American Epilepsy Society Predoctoral Fellowship (YF)

National Institutes of Health grant F31AG069496 (LV)

National Institutes of Health grant R01MH120162 (DJC)

National Institutes of Health grant DP2MH122399 (DJC)

Brain Research Foundation Award (DJC)

Klingenstein-Simons Fellowship (DJC)

NARSAD Young Investigator Award (DJC)

McKnight Memory and Cognitive Disorder Award (DJC)

One Mind-Otsuka Rising Star Research Award (DJC)

Hirschl/Weill-Caulier Award (DJC)

Mount Sinai Distinguished Scholar Award (DJC)

CURE Taking Flight Award (TS)

American Epilepsy Society Junior Investigator Award (TS)

National Institutes of Health grant R01NS136590 (TS)

National Institutes of Health grant R01NS116357 (TS)

National Institutes of Health grant RF1AG072497 (TS)

Funding Statement

National Institutes of Health grant F32NS116416 (ZCW)

American Epilepsy Society Postdoctoral Research Fellowship (ZCW)

Simons Collaboration on Plasticity and the Aging Brain Transition to Independence Award (ZCW)

Friedman Brain Institute Postdoctoral Innovator Award (ZCW)

National Institutes of Health grant F31NS134301 (PAP)

American Epilepsy Society Predoctoral Fellowship (YF)

National Institutes of Health grant F31AG069496 (LV)

National Institutes of Health grant R01MH120162 (DJC)

National Institutes of Health grant DP2MH122399 (DJC)

Brain Research Foundation Award (DJC)

Klingenstein-Simons Fellowship (DJC)

NARSAD Young Investigator Award (DJC)

McKnight Memory and Cognitive Disorder Award (DJC)

One Mind-Otsuka Rising Star Research Award (DJC)

Hirschl/Weill-Caulier Award (DJC)

Mount Sinai Distinguished Scholar Award (DJC)

CURE Taking Flight Award (TS)

American Epilepsy Society Junior Investigator Award (TS)

National Institutes of Health grant R01NS136590 (TS)

National Institutes of Health grant R01NS116357 (TS)

National Institutes of Health grant RF1AG072497 (TS)

Footnotes

Competing interests:

Authors declare that they have no competing interests.

Data and materials availability:

All software and hardware requirements to implement real-time phase manipulations during behavior are available online (https://github.com/ShumanLab/PhaSER). All LFP and spike data will be made publicly available upon publication. Any additional information required to re-analyze the data reported in this paper is available from the lead contact upon request.

REFERENCES

1.Buzsáki G., S. L. L.-W. & Vanderwolf C. H. Cellular bases of hippocampal EEG in the behaving rat. Brain Res. Rev. 6, 139–171 (1983). [Google Scholar]
2.Colgin L. L. Oscillations and hippocampal-prefrontal synchrony. Current opinion in neurobiology 21, 467–74 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Siegle J. H. & Wilson M. A. Enhancement of encoding and retrieval functions through theta phase-specific manipulation of hippocampus. Elife 3, e03061 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Kay K. et al. Constant Sub-second Cycling between Representations of Possible Futures in the Hippocampus. Cell 180, 552–567.e25 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Santos-Pata D., Barry C. & Ólafsdóttir H. F. Theta-band phase locking during encoding leads to coordinated entorhinal-hippocampal replay. Curr. Biol. 33, 4570–4581.e5 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Ku S.-P. et al. Phase locking of hippocampal CA3 neurons to distal CA1 theta oscillations selectively predicts memory performance. Cell Rep. 43, 114276 (2024). [DOI] [PubMed] [Google Scholar]
7.Hasselmo M. E., Bodeln C. & Wyble B. P. A Proposed Function for Hippocampal Theta Rhythm: Separate Phases of Encoding and Retrieval Enhance Reversal of Prior Learning. Neural Comput 14, 793–817 (2002). [DOI] [PubMed] [Google Scholar]
8.Mehta M. R., Lee A. K. & Wilson M. A. Role of experience and oscillations in transforming a rate code into a temporal code. Nature 417, 741–746 (2002). [DOI] [PubMed] [Google Scholar]
9.Klausberger T. et al. Brain-state- and cell-type-specific firing of hippocampal interneurons in vivo. Nature 421, 844–8 (2003). [DOI] [PubMed] [Google Scholar]
10.Klausberger T. & Somogyi P. Neuronal Diversity and Temporal Dynamics: The Unity of Hippocampal Circuit Operations. Science 321, 53–57 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Somogyi P., Katona L., Klausberger T., Lasztóczi B. & Viney T. J. Temporal redistribution of inhibition over neuronal subcellular domains underlies state-dependent rhythmic change of excitability in the hippocampus. Philosophical Transactions Royal Soc Lond Ser B Biological Sci 369, 20120518 (2013). [Google Scholar]
12.Csicsvari J., Hirase H., Czurkó A., Mamiya A. & Buzsáki G. Oscillatory Coupling of Hippocampal Pyramidal Cells and Interneurons in the Behaving Rat. J Neurosci 19, 274–287 (1999). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Varga C. et al. Functional fission of parvalbumin interneuron classes during fast network events. eLife 3, (2014). [Google Scholar]
14.Royer S. et al. Control of timing, rate and bursts of hippocampal place cells by dendritic and somatic inhibition. Nature neuroscience 15, 769–75 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Adam C. D. et al. Disrupted Hippocampal Theta-Gamma Coupling and Spike-Field Coherence Following Experimental Traumatic Brain Injury. eLife 13:RP100642, (2024). [Google Scholar]
16.Shuman T. et al. Breakdown of spatial coding and interneuron synchronization in epileptic mice. Nat Neurosci 23, 229–238 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Feng Y. et al. Distinct changes to hippocampal and medial entorhinal circuits emerge across the progression of cognitive deficits in epilepsy. Cell Rep. 44, 115131 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Vetere L. M. et al. Medial entorhinal-hippocampal desynchronization parallels the emergence of memory impairment in a mouse model of Alzheimer’s disease pathology. bioRxiv 2025.01.15.633171 (2025) doi: 10.1101/2025.01.15.633171. [DOI] [Google Scholar]
19.Guillem K. & Ahmed S. H. Reorganization of theta phase-locking in the orbitofrontal cortex drives cocaine choice under the influence. Sci. Rep. 10, 8041 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Lazaro M. T. et al. Reduced Prefrontal Synaptic Connectivity and Disturbed Oscillatory Population Dynamics in the CNTNAP2 Model of Autism. Cell Reports 27, 2567–2578.e6 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Sigurdsson T., Stark K. L., Karayiorgou M., Gogos J. A. & Gordon J. A. Impaired hippocampal–prefrontal synchrony in a genetic mouse model of schizophrenia. Nature 464, 763–767 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Dickerson D. D., Wolff A. R. & Bilkey D. K. Abnormal Long-Range Neural Synchrony in a Maternal Immune Activation Animal Model of Schizophrenia. J. Neurosci. 30, 12424–12431 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Lopez-Pigozzi D. et al. Altered Oscillatory Dynamics of CA1 Parvalbumin Basket Cells during Theta-Gamma Rhythmopathies of Temporal Lobe Epilepsy. eNeuro 3, (2016). [Google Scholar]
24.Braganza O., Mueller-Komorowska D., Kelly T. & Beck H. Quantitative properties of a feedback circuit predict frequency-dependent pattern separation. eLife 9, e53148 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Cutsuridis V. & Poirazi P. A computational study on how theta modulated inhibition can account for the long temporal windows in the entorhinal-hippocampal loop. Neurobiol Learn Mem 120, 69–83 (2015). [DOI] [PubMed] [Google Scholar]
26.Cutsuridis V. & Hasselmo M. GABAergic contributions to gating, timing, and phase precession of hippocampal neuronal activity during theta oscillations. Hippocampus 22, 1597–1621 (2012). [DOI] [PubMed] [Google Scholar]
27.Zutshi I., Valero M., Fernández-Ruiz A. & Buzsáki G. Extrinsic control and intrinsic computation in the hippocampal CA1 circuit. Neuron (2021) doi: 10.1016/j.neuron.2021.11.015. [DOI] [Google Scholar]
28.Mizuseki K., Sirota A., Pastalkova E. & Buzsáki G. Theta oscillations provide temporal windows for local circuit computation in the entorhinal-hippocampal loop. Neuron 64, 267–80 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Krook-Magnuson E. et al. In vivo evaluation of the dentate gate theory in epilepsy. The Journal of physiology 593, 2379–88 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Heinemann U. et al. The dentate gyrus as a regulated gate for the propagation of epileptiform activity. Epilepsy research. Supplement 7, 273–80 (1992). [PubMed] [Google Scholar]
31.Hainmueller T., Cazala A., Huang L.-W. & Bartos M. Subfield-specific interneuron circuits govern the hippocampal response to novelty in male mice. Nat. Commun. 15, 714 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Hainmueller T., Heynold E. A., Paleologos N. & Buzsáki G. Dentate Gyrus Interneurons Modulate Winner-Take-All Network Dynamics in Freely Behaving Mice. (2025) doi: 10.2139/ssrn.5282602. [DOI] [Google Scholar]
33.Freund T. F. & Buzsáki G. Interneurons of the hippocampus. Hippocampus 6, 347–470 (1996). [DOI] [PubMed] [Google Scholar]
34.Espinoza C., Guzman S. J., Zhang X. & Jonas P. Parvalbumin+ interneurons obey unique connectivity rules and establish a powerful lateral-inhibition microcircuit in dentate gyrus. Nat. Commun. 9, 4605 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Yuan M. et al. Somatostatin-positive interneurons in the dentate gyrus of mice provide local- and long-range septal synaptic inhibition. Elife 6, e21105 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Christenson Wick Z., Leintz C. H., Xamonthiene C., Huang B. H. & Krook-Magnuson E. Axonal sprouting in commissurally projecting parvalbumin-expressing interneurons. Journal of neuroscience research 95, 2336–2344 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Peng Z. et al. A Reorganized GABAergic Circuit in a Model of Epilepsy: Evidence from Optogenetic Labeling and Stimulation of Somatostatin Interneurons. The Journal of Neuroscience 33, 14392–405 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Lee S.-H., Kang Y.-J. & Smith B. N. Activation of hypoactive parvalbumin-positive fast-spiking interneurons restores dentate inhibition to reduce electrographic seizures in the mouse intrahippocampal kainate model of temporal lobe epilepsy. Neurobiol. Dis. 203, 106737 (2024). [DOI] [PubMed] [Google Scholar]
39.Halabisky B., Parada I., Buckmaster P. S. & Prince D. A. Excitatory Input Onto Hilar Somatostatin Interneurons Is Increased in a Chronic Model of Epilepsy. J. Neurophysiol. 104, 2214–2223 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Buckmaster P. & Dudek F. Neuron loss, granule cell axon reorganization, and functional changes in the dentate gyrus of epileptic kainate-treated rats. The Journal of comparative neurology 385, 385–404 (1997). [PubMed] [Google Scholar]
41.Lanerolle N. C. de, Kim J. H., Robbins R. J. & Spencer D. D. Hippocampal interneuron loss and plasticity in human temporal lobe epilepsy. Brain Res. 495, 387–395 (1989). [DOI] [PubMed] [Google Scholar]
42.Sloviter R. S. Decreased Hippocampal Inhibition and a Selective Loss of Interneurons in Experimental Epilepsy. Science 235, 73–76 (1987). [DOI] [PubMed] [Google Scholar]
43.Cossart R. et al. Dendritic but not somatic GABAergic inhibition is decreased in experimental epilepsy. Nat. Neurosci. 4, 52–62 (2001). [DOI] [PubMed] [Google Scholar]
44.Pavlides C., Greenstein Y. J., Grudman M. & Winson J. Long-term potentiation in the dentate gyrus is induced preferentially on the positive phase of θ-rhythm. Brain Res 439, 383–387 (1988). [DOI] [PubMed] [Google Scholar]
45.Hyman J. M., Wyble B. P., Goyal V., Rossi C. A. & Hasselmo M. E. Stimulation in Hippocampal Region CA1 in Behaving Rats Yields Long-Term Potentiation when Delivered to the Peak of Theta and Long-Term Depression when Delivered to the Trough. J Neurosci 23, 11725–11731 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Chen L. L., Madhavan R., Rapoport B. I. & Anderson W. S. Real-Time Brain Oscillation Detection and Phase-Locked Stimulation Using Autoregressive Spectral Estimation and Time-Series Forward Prediction. Ieee T Bio-med Eng 60, 753–762 (2013). [Google Scholar]
47.Vierock J. et al. BiPOLES is an optogenetic tool developed for bidirectional dual-color control of neurons. Nat Commun 12, 4527 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Amilhon B. et al. Parvalbumin Interneurons of Hippocampus Tune Population Activity at Theta Frequency. Neuron 86, 1277–1289 (2015). [DOI] [PubMed] [Google Scholar]
49.Bezaire M. J., Raikov I., Burk K., Vyas D. & Soltesz I. Interneuronal mechanisms of hippocampal theta oscillations in a full-scale model of the rodent CA1 circuit. eLife 5, (2016). [Google Scholar]
50.Strüber M., Sauer J.-F. & Bartos M. Parvalbumin expressing interneurons control spike-phase coupling of hippocampal cells to theta oscillations. Sci Rep-uk 12, 1362 (2022). [Google Scholar]
51.Müller-Komorowska D., Fujishige T. & Fukai T. Recurrent Interneuron Connectivity does not Support Synchrony in a Biophysical Dentate Gyrus Model. eNeuro ENEURO.0097–25.2025 (2025) doi: 10.1523/eneuro.0097-25.2025. [DOI] [Google Scholar]
52.Christenson Wick Z., Tetzlaff M. R. & Krook-Magnuson E. Novel long-range inhibitory nNOS-expressing hippocampal cells. Elife 8, e46816 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Skrabak D. et al. Slack K+ channels limit kainic acid-induced seizure severity in mice by modulating neuronal excitability and firing. Commun. Biol. 6, 1029 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Gross C. et al. MicroRNA-Mediated Downregulation of the Potassium Channel Kv4.2 Contributes to Seizure Onset. Cell Rep. 17, 37–45 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Zhang W. et al. Surviving hilar somatostatin interneurons enlarge, sprout axons, and form new synapses with granule cells in a mouse model of temporal lobe epilepsy. The Journal of neuroscience : the official journal of the Society for Neuroscience 29, 14247–56 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Raven F., Vankampen A. A., He A. & Aton S. J. Dentate gyrus network regulation by somatostatin- and parvalbumin-expressing interneurons differentially impacts hippocampal spatial memory processing. bioRxiv 2025.07.23.666335 (2025) doi: 10.1101/2025.07.23.666335. [DOI] [Google Scholar]
57.Yuan M. et al. Predictive goal coding by dentate gyrus somatostatin-expressing interneurons in male mice. Nat. Commun. 16, 5382 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Zelmann R. et al. CLoSES: A platform for closed-loop intracranial stimulation in humans. Neuroimage 223, 117314–117314 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Schatza M. J., Blackwood E. B., Nagrale S. S. & Widge A. S. Toolkit for Oscillatory Real-time Tracking and Estimation (TORTE). J Neurosci Meth 366, 109409 (2022). [Google Scholar]
60.Desideri D., Zrenner C., Gordon P. C., Ziemann U. & Belardinelli P. Nil effects of μ-rhythm phase-dependent burst-rTMS on cortical excitability in humans: A resting-state EEG and TMS-EEG study. Plos One 13, e0208747 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
61.Gordon P. C., Belardinelli P., Stenroos M., Ziemann U. & Zrenner C. Prefrontal theta phase-dependent rTMS-induced plasticity of cortical and behavioral responses in human cortex. Brain Stimul 15, 391–402 (2022). [DOI] [PubMed] [Google Scholar]
62.Rahsepar B. et al. Theta-phase-specific modulation of dentate gyrus memory neurons. eLife 12, (2023). [Google Scholar]
63.Grennan I. et al. Phase-dependent closed-loop deep brain stimulation of the fornix provides bidirectional manipulation of hippocampal theta oscillations. Brain Stimul. 18, 993–1003 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Wolfe C. E. et al. Seizures as a complication of recreational drug use: Analysis of the Euro-DEN Plus data-set. NeuroToxicology 73, 183–187 (2019). [DOI] [PubMed] [Google Scholar]
65.Mattoo S. K. et al. Prevalence and correlates of epileptic seizure in substance-abusing subjects. Psychiatry Clin. Neurosci. 63, 580–582 (2009). [DOI] [PubMed] [Google Scholar]
66.Hyde T. M. & Weinberger D. R. Seizures and schizophrenia. Schizophr. Bull. 23, 611–622 (1997). [DOI] [PubMed] [Google Scholar]
67.Fordington S. & Manford M. A review of seizures and epilepsy following traumatic brain injury. J. Neurol. 267, 3105–3111 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Capal J. K. & Jeste S. S. Autism and Epilepsy. Pediatr. Clin. North Am. 71, 241–252 (2024). [DOI] [PubMed] [Google Scholar]
69.Vossel K. A., Tartaglia M. C., Nygaard H. B., Zeman A. Z. & Miller B. L. Epileptic activity in Alzheimer’s disease: causes and clinical relevance. Lancet Neurol. 16, 311–322 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
70.Philipsberg P. A. & Shuman T. The Best Things Happen When You Least Expect Them: Responsive Neurostimulation During Low-Risk Brain States is Associated with Improved Long-Term Seizure Suppression. Epilepsy Curr. 24, 350–352 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Anderson D. N. et al. Closed-loop stimulation in periods with less epileptiform activity drives improved epilepsy outcomes. Brain 147, 521–531 (2023). [Google Scholar]
72.Khambhati A. N., Shafi A., Rao V. R. & Chang E. F. Long-term brain network reorganization predicts responsive neurostimulation outcomes for focal epilepsy. Sci Transl Med 13, eabf6588 (2021). [Google Scholar]
73.Földi T., Lőrincz M. L. & Berényi A. Temporally Targeted Interactions With Pathologic Oscillations as Therapeutical Targets in Epilepsy and Beyond. Front Neural Circuit 15, 784085 (2021). [Google Scholar]
74.Ferrero J. J. et al. Closed-loop electrical stimulation prevents focal epilepsy progression and long-term memory impairment. Nat. Neurosci. 28, 1753–1762 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
75.Figee M. & Mayberg H. The future of personalized brain stimulation. Nat. Med. 27, 196–197 (2021). [DOI] [PubMed] [Google Scholar]
76.Taniguchi H. et al. A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron 71, 995–1013 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
77.Hippenmeyer S. et al. A developmental switch in the response of DRG neurons to ETS transcription factor signaling. PLoS biology 3, e159 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]
78.Tennant S. A. et al. Stellate Cells in the Medial Entorhinal Cortex Are Required for Spatial Learning. Cell Reports 22, 1313–1324 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
79.Aronov D. & Tank D. W. Engagement of Neural Circuits Underlying 2D Spatial Navigation in a Rodent Virtual Reality System. Neuron 84, 442–456 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
80.Cai D. J. et al. A shared neural ensemble links distinct contextual memories encoded close in time. Nature 534, 115 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
81.Shobe J. L., Claar L. D., Parhami S., Bakhurin K. I. & Masmanidis S. C. Brain activity mapping at multiple scales with silicon microprobes containing 1,024 electrodes. J Neurophysiol 114, 2043–52 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
82.Franklin K. B. J. & Paxinos G. The mouse brain: in stereotaxic coordinates. Academic Press, Elsevier; (2007). [Google Scholar]
83.Stringer C. et al. Spontaneous behaviors drive multidimensional, brainwide activity. Science 364, 255 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
84.Steinmetz N. A. et al. Neuropixels 2.0: A miniaturized high-density probe for stable, long-term brain recordings. Science 372, (2021). [Google Scholar]
85.Mitra P. & Bokil H. Observed Brain Dynamics. Oxford University Press; (2008). [Google Scholar]
86.Miri M. L., Vinck M., Pant R. & Cardin J. A. Altered hippocampal interneuron activity precedes ictal onset. eLife 7, (2018). [Google Scholar]
87.Karlsson K. AE. & Blumberg M. S. Temperature-Induced Reciprocal Activation of Hippocampal Field Activity. J Neurophysiol 91, 583–588 (2004). [DOI] [PubMed] [Google Scholar]
88.Lubenov E. V. & Siapas A. G. Hippocampal theta oscillations are travelling waves. Nature 459, 534–539 (2009). [DOI] [PubMed] [Google Scholar]
89.Schomburg E. W. et al. Theta Phase Segregation of Input-Specific Gamma Patterns in Entorhinal-Hippocampal Networks. Neuron 84, 470–485 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
90.Senzai Y. & Buzsáki G. Physiological Properties and Behavioral Correlates of Hippocampal Granule Cells and Mossy Cells. Neuron 93, 691–704.e5 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
91.Bergstrom R. A. et al. Automated identification of multiple seizure-related and interictal epileptiform event types in the EEG of mice. Sci. Rep. 3, 1483 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
92.Berens P. CircStat: A MATLAB Toolbox for Circular Statistics. Journal of Statistical Software (2009) doi: 10.18637/jss.v031.i10. [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

NIHPP2025.09.10.675457V1-supplement-1.pdf^{(2.2MB, pdf)}

Data Availability Statement

[R1] 1.Buzsáki G., S. L. L.-W. & Vanderwolf C. H. Cellular bases of hippocampal EEG in the behaving rat. Brain Res. Rev. 6, 139–171 (1983). [Google Scholar]

[R2] 2.Colgin L. L. Oscillations and hippocampal-prefrontal synchrony. Current opinion in neurobiology 21, 467–74 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R3] 3.Siegle J. H. & Wilson M. A. Enhancement of encoding and retrieval functions through theta phase-specific manipulation of hippocampus. Elife 3, e03061 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R4] 4.Kay K. et al. Constant Sub-second Cycling between Representations of Possible Futures in the Hippocampus. Cell 180, 552–567.e25 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R5] 5.Santos-Pata D., Barry C. & Ólafsdóttir H. F. Theta-band phase locking during encoding leads to coordinated entorhinal-hippocampal replay. Curr. Biol. 33, 4570–4581.e5 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R6] 6.Ku S.-P. et al. Phase locking of hippocampal CA3 neurons to distal CA1 theta oscillations selectively predicts memory performance. Cell Rep. 43, 114276 (2024). [DOI] [PubMed] [Google Scholar]

[R7] 7.Hasselmo M. E., Bodeln C. & Wyble B. P. A Proposed Function for Hippocampal Theta Rhythm: Separate Phases of Encoding and Retrieval Enhance Reversal of Prior Learning. Neural Comput 14, 793–817 (2002). [DOI] [PubMed] [Google Scholar]

[R8] 8.Mehta M. R., Lee A. K. & Wilson M. A. Role of experience and oscillations in transforming a rate code into a temporal code. Nature 417, 741–746 (2002). [DOI] [PubMed] [Google Scholar]

[R9] 9.Klausberger T. et al. Brain-state- and cell-type-specific firing of hippocampal interneurons in vivo. Nature 421, 844–8 (2003). [DOI] [PubMed] [Google Scholar]

[R10] 10.Klausberger T. & Somogyi P. Neuronal Diversity and Temporal Dynamics: The Unity of Hippocampal Circuit Operations. Science 321, 53–57 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R11] 11.Somogyi P., Katona L., Klausberger T., Lasztóczi B. & Viney T. J. Temporal redistribution of inhibition over neuronal subcellular domains underlies state-dependent rhythmic change of excitability in the hippocampus. Philosophical Transactions Royal Soc Lond Ser B Biological Sci 369, 20120518 (2013). [Google Scholar]

[R12] 12.Csicsvari J., Hirase H., Czurkó A., Mamiya A. & Buzsáki G. Oscillatory Coupling of Hippocampal Pyramidal Cells and Interneurons in the Behaving Rat. J Neurosci 19, 274–287 (1999). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R13] 13.Varga C. et al. Functional fission of parvalbumin interneuron classes during fast network events. eLife 3, (2014). [Google Scholar]

[R14] 14.Royer S. et al. Control of timing, rate and bursts of hippocampal place cells by dendritic and somatic inhibition. Nature neuroscience 15, 769–75 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R15] 15.Adam C. D. et al. Disrupted Hippocampal Theta-Gamma Coupling and Spike-Field Coherence Following Experimental Traumatic Brain Injury. eLife 13:RP100642, (2024). [Google Scholar]

[R16] 16.Shuman T. et al. Breakdown of spatial coding and interneuron synchronization in epileptic mice. Nat Neurosci 23, 229–238 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R17] 17.Feng Y. et al. Distinct changes to hippocampal and medial entorhinal circuits emerge across the progression of cognitive deficits in epilepsy. Cell Rep. 44, 115131 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R18] 18.Vetere L. M. et al. Medial entorhinal-hippocampal desynchronization parallels the emergence of memory impairment in a mouse model of Alzheimer’s disease pathology. bioRxiv 2025.01.15.633171 (2025) doi: 10.1101/2025.01.15.633171. [DOI] [Google Scholar]

[R19] 19.Guillem K. & Ahmed S. H. Reorganization of theta phase-locking in the orbitofrontal cortex drives cocaine choice under the influence. Sci. Rep. 10, 8041 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R20] 20.Lazaro M. T. et al. Reduced Prefrontal Synaptic Connectivity and Disturbed Oscillatory Population Dynamics in the CNTNAP2 Model of Autism. Cell Reports 27, 2567–2578.e6 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R21] 21.Sigurdsson T., Stark K. L., Karayiorgou M., Gogos J. A. & Gordon J. A. Impaired hippocampal–prefrontal synchrony in a genetic mouse model of schizophrenia. Nature 464, 763–767 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R22] 22.Dickerson D. D., Wolff A. R. & Bilkey D. K. Abnormal Long-Range Neural Synchrony in a Maternal Immune Activation Animal Model of Schizophrenia. J. Neurosci. 30, 12424–12431 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R23] 23.Lopez-Pigozzi D. et al. Altered Oscillatory Dynamics of CA1 Parvalbumin Basket Cells during Theta-Gamma Rhythmopathies of Temporal Lobe Epilepsy. eNeuro 3, (2016). [Google Scholar]

[R24] 24.Braganza O., Mueller-Komorowska D., Kelly T. & Beck H. Quantitative properties of a feedback circuit predict frequency-dependent pattern separation. eLife 9, e53148 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R25] 25.Cutsuridis V. & Poirazi P. A computational study on how theta modulated inhibition can account for the long temporal windows in the entorhinal-hippocampal loop. Neurobiol Learn Mem 120, 69–83 (2015). [DOI] [PubMed] [Google Scholar]

[R26] 26.Cutsuridis V. & Hasselmo M. GABAergic contributions to gating, timing, and phase precession of hippocampal neuronal activity during theta oscillations. Hippocampus 22, 1597–1621 (2012). [DOI] [PubMed] [Google Scholar]

[R27] 27.Zutshi I., Valero M., Fernández-Ruiz A. & Buzsáki G. Extrinsic control and intrinsic computation in the hippocampal CA1 circuit. Neuron (2021) doi: 10.1016/j.neuron.2021.11.015. [DOI] [Google Scholar]

[R28] 28.Mizuseki K., Sirota A., Pastalkova E. & Buzsáki G. Theta oscillations provide temporal windows for local circuit computation in the entorhinal-hippocampal loop. Neuron 64, 267–80 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R29] 29.Krook-Magnuson E. et al. In vivo evaluation of the dentate gate theory in epilepsy. The Journal of physiology 593, 2379–88 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R30] 30.Heinemann U. et al. The dentate gyrus as a regulated gate for the propagation of epileptiform activity. Epilepsy research. Supplement 7, 273–80 (1992). [PubMed] [Google Scholar]

[R31] 31.Hainmueller T., Cazala A., Huang L.-W. & Bartos M. Subfield-specific interneuron circuits govern the hippocampal response to novelty in male mice. Nat. Commun. 15, 714 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R32] 32.Hainmueller T., Heynold E. A., Paleologos N. & Buzsáki G. Dentate Gyrus Interneurons Modulate Winner-Take-All Network Dynamics in Freely Behaving Mice. (2025) doi: 10.2139/ssrn.5282602. [DOI] [Google Scholar]

[R33] 33.Freund T. F. & Buzsáki G. Interneurons of the hippocampus. Hippocampus 6, 347–470 (1996). [DOI] [PubMed] [Google Scholar]

[R34] 34.Espinoza C., Guzman S. J., Zhang X. & Jonas P. Parvalbumin+ interneurons obey unique connectivity rules and establish a powerful lateral-inhibition microcircuit in dentate gyrus. Nat. Commun. 9, 4605 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R35] 35.Yuan M. et al. Somatostatin-positive interneurons in the dentate gyrus of mice provide local- and long-range septal synaptic inhibition. Elife 6, e21105 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R36] 36.Christenson Wick Z., Leintz C. H., Xamonthiene C., Huang B. H. & Krook-Magnuson E. Axonal sprouting in commissurally projecting parvalbumin-expressing interneurons. Journal of neuroscience research 95, 2336–2344 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R37] 37.Peng Z. et al. A Reorganized GABAergic Circuit in a Model of Epilepsy: Evidence from Optogenetic Labeling and Stimulation of Somatostatin Interneurons. The Journal of Neuroscience 33, 14392–405 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R38] 38.Lee S.-H., Kang Y.-J. & Smith B. N. Activation of hypoactive parvalbumin-positive fast-spiking interneurons restores dentate inhibition to reduce electrographic seizures in the mouse intrahippocampal kainate model of temporal lobe epilepsy. Neurobiol. Dis. 203, 106737 (2024). [DOI] [PubMed] [Google Scholar]

[R39] 39.Halabisky B., Parada I., Buckmaster P. S. & Prince D. A. Excitatory Input Onto Hilar Somatostatin Interneurons Is Increased in a Chronic Model of Epilepsy. J. Neurophysiol. 104, 2214–2223 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R40] 40.Buckmaster P. & Dudek F. Neuron loss, granule cell axon reorganization, and functional changes in the dentate gyrus of epileptic kainate-treated rats. The Journal of comparative neurology 385, 385–404 (1997). [PubMed] [Google Scholar]

[R41] 41.Lanerolle N. C. de, Kim J. H., Robbins R. J. & Spencer D. D. Hippocampal interneuron loss and plasticity in human temporal lobe epilepsy. Brain Res. 495, 387–395 (1989). [DOI] [PubMed] [Google Scholar]

[R42] 42.Sloviter R. S. Decreased Hippocampal Inhibition and a Selective Loss of Interneurons in Experimental Epilepsy. Science 235, 73–76 (1987). [DOI] [PubMed] [Google Scholar]

[R43] 43.Cossart R. et al. Dendritic but not somatic GABAergic inhibition is decreased in experimental epilepsy. Nat. Neurosci. 4, 52–62 (2001). [DOI] [PubMed] [Google Scholar]

[R44] 44.Pavlides C., Greenstein Y. J., Grudman M. & Winson J. Long-term potentiation in the dentate gyrus is induced preferentially on the positive phase of θ-rhythm. Brain Res 439, 383–387 (1988). [DOI] [PubMed] [Google Scholar]

[R45] 45.Hyman J. M., Wyble B. P., Goyal V., Rossi C. A. & Hasselmo M. E. Stimulation in Hippocampal Region CA1 in Behaving Rats Yields Long-Term Potentiation when Delivered to the Peak of Theta and Long-Term Depression when Delivered to the Trough. J Neurosci 23, 11725–11731 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R46] 46.Chen L. L., Madhavan R., Rapoport B. I. & Anderson W. S. Real-Time Brain Oscillation Detection and Phase-Locked Stimulation Using Autoregressive Spectral Estimation and Time-Series Forward Prediction. Ieee T Bio-med Eng 60, 753–762 (2013). [Google Scholar]

[R47] 47.Vierock J. et al. BiPOLES is an optogenetic tool developed for bidirectional dual-color control of neurons. Nat Commun 12, 4527 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R48] 48.Amilhon B. et al. Parvalbumin Interneurons of Hippocampus Tune Population Activity at Theta Frequency. Neuron 86, 1277–1289 (2015). [DOI] [PubMed] [Google Scholar]

[R49] 49.Bezaire M. J., Raikov I., Burk K., Vyas D. & Soltesz I. Interneuronal mechanisms of hippocampal theta oscillations in a full-scale model of the rodent CA1 circuit. eLife 5, (2016). [Google Scholar]

[R50] 50.Strüber M., Sauer J.-F. & Bartos M. Parvalbumin expressing interneurons control spike-phase coupling of hippocampal cells to theta oscillations. Sci Rep-uk 12, 1362 (2022). [Google Scholar]

[R51] 51.Müller-Komorowska D., Fujishige T. & Fukai T. Recurrent Interneuron Connectivity does not Support Synchrony in a Biophysical Dentate Gyrus Model. eNeuro ENEURO.0097–25.2025 (2025) doi: 10.1523/eneuro.0097-25.2025. [DOI] [Google Scholar]

[R52] 52.Christenson Wick Z., Tetzlaff M. R. & Krook-Magnuson E. Novel long-range inhibitory nNOS-expressing hippocampal cells. Elife 8, e46816 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R53] 53.Skrabak D. et al. Slack K+ channels limit kainic acid-induced seizure severity in mice by modulating neuronal excitability and firing. Commun. Biol. 6, 1029 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R54] 54.Gross C. et al. MicroRNA-Mediated Downregulation of the Potassium Channel Kv4.2 Contributes to Seizure Onset. Cell Rep. 17, 37–45 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R55] 55.Zhang W. et al. Surviving hilar somatostatin interneurons enlarge, sprout axons, and form new synapses with granule cells in a mouse model of temporal lobe epilepsy. The Journal of neuroscience : the official journal of the Society for Neuroscience 29, 14247–56 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R56] 56.Raven F., Vankampen A. A., He A. & Aton S. J. Dentate gyrus network regulation by somatostatin- and parvalbumin-expressing interneurons differentially impacts hippocampal spatial memory processing. bioRxiv 2025.07.23.666335 (2025) doi: 10.1101/2025.07.23.666335. [DOI] [Google Scholar]

[R57] 57.Yuan M. et al. Predictive goal coding by dentate gyrus somatostatin-expressing interneurons in male mice. Nat. Commun. 16, 5382 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R58] 58.Zelmann R. et al. CLoSES: A platform for closed-loop intracranial stimulation in humans. Neuroimage 223, 117314–117314 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R59] 59.Schatza M. J., Blackwood E. B., Nagrale S. S. & Widge A. S. Toolkit for Oscillatory Real-time Tracking and Estimation (TORTE). J Neurosci Meth 366, 109409 (2022). [Google Scholar]

[R60] 60.Desideri D., Zrenner C., Gordon P. C., Ziemann U. & Belardinelli P. Nil effects of μ-rhythm phase-dependent burst-rTMS on cortical excitability in humans: A resting-state EEG and TMS-EEG study. Plos One 13, e0208747 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R61] 61.Gordon P. C., Belardinelli P., Stenroos M., Ziemann U. & Zrenner C. Prefrontal theta phase-dependent rTMS-induced plasticity of cortical and behavioral responses in human cortex. Brain Stimul 15, 391–402 (2022). [DOI] [PubMed] [Google Scholar]

[R62] 62.Rahsepar B. et al. Theta-phase-specific modulation of dentate gyrus memory neurons. eLife 12, (2023). [Google Scholar]

[R63] 63.Grennan I. et al. Phase-dependent closed-loop deep brain stimulation of the fornix provides bidirectional manipulation of hippocampal theta oscillations. Brain Stimul. 18, 993–1003 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R64] 64.Wolfe C. E. et al. Seizures as a complication of recreational drug use: Analysis of the Euro-DEN Plus data-set. NeuroToxicology 73, 183–187 (2019). [DOI] [PubMed] [Google Scholar]

[R65] 65.Mattoo S. K. et al. Prevalence and correlates of epileptic seizure in substance-abusing subjects. Psychiatry Clin. Neurosci. 63, 580–582 (2009). [DOI] [PubMed] [Google Scholar]

[R66] 66.Hyde T. M. & Weinberger D. R. Seizures and schizophrenia. Schizophr. Bull. 23, 611–622 (1997). [DOI] [PubMed] [Google Scholar]

[R67] 67.Fordington S. & Manford M. A review of seizures and epilepsy following traumatic brain injury. J. Neurol. 267, 3105–3111 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R68] 68.Capal J. K. & Jeste S. S. Autism and Epilepsy. Pediatr. Clin. North Am. 71, 241–252 (2024). [DOI] [PubMed] [Google Scholar]

[R69] 69.Vossel K. A., Tartaglia M. C., Nygaard H. B., Zeman A. Z. & Miller B. L. Epileptic activity in Alzheimer’s disease: causes and clinical relevance. Lancet Neurol. 16, 311–322 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R70] 70.Philipsberg P. A. & Shuman T. The Best Things Happen When You Least Expect Them: Responsive Neurostimulation During Low-Risk Brain States is Associated with Improved Long-Term Seizure Suppression. Epilepsy Curr. 24, 350–352 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R71] 71.Anderson D. N. et al. Closed-loop stimulation in periods with less epileptiform activity drives improved epilepsy outcomes. Brain 147, 521–531 (2023). [Google Scholar]

[R72] 72.Khambhati A. N., Shafi A., Rao V. R. & Chang E. F. Long-term brain network reorganization predicts responsive neurostimulation outcomes for focal epilepsy. Sci Transl Med 13, eabf6588 (2021). [Google Scholar]

[R73] 73.Földi T., Lőrincz M. L. & Berényi A. Temporally Targeted Interactions With Pathologic Oscillations as Therapeutical Targets in Epilepsy and Beyond. Front Neural Circuit 15, 784085 (2021). [Google Scholar]

[R74] 74.Ferrero J. J. et al. Closed-loop electrical stimulation prevents focal epilepsy progression and long-term memory impairment. Nat. Neurosci. 28, 1753–1762 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R75] 75.Figee M. & Mayberg H. The future of personalized brain stimulation. Nat. Med. 27, 196–197 (2021). [DOI] [PubMed] [Google Scholar]

[R76] 76.Taniguchi H. et al. A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron 71, 995–1013 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R77] 77.Hippenmeyer S. et al. A developmental switch in the response of DRG neurons to ETS transcription factor signaling. PLoS biology 3, e159 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R78] 78.Tennant S. A. et al. Stellate Cells in the Medial Entorhinal Cortex Are Required for Spatial Learning. Cell Reports 22, 1313–1324 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R79] 79.Aronov D. & Tank D. W. Engagement of Neural Circuits Underlying 2D Spatial Navigation in a Rodent Virtual Reality System. Neuron 84, 442–456 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R80] 80.Cai D. J. et al. A shared neural ensemble links distinct contextual memories encoded close in time. Nature 534, 115 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R81] 81.Shobe J. L., Claar L. D., Parhami S., Bakhurin K. I. & Masmanidis S. C. Brain activity mapping at multiple scales with silicon microprobes containing 1,024 electrodes. J Neurophysiol 114, 2043–52 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R82] 82.Franklin K. B. J. & Paxinos G. The mouse brain: in stereotaxic coordinates. Academic Press, Elsevier; (2007). [Google Scholar]

[R83] 83.Stringer C. et al. Spontaneous behaviors drive multidimensional, brainwide activity. Science 364, 255 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R84] 84.Steinmetz N. A. et al. Neuropixels 2.0: A miniaturized high-density probe for stable, long-term brain recordings. Science 372, (2021). [Google Scholar]

[R85] 85.Mitra P. & Bokil H. Observed Brain Dynamics. Oxford University Press; (2008). [Google Scholar]

[R86] 86.Miri M. L., Vinck M., Pant R. & Cardin J. A. Altered hippocampal interneuron activity precedes ictal onset. eLife 7, (2018). [Google Scholar]

[R87] 87.Karlsson K. AE. & Blumberg M. S. Temperature-Induced Reciprocal Activation of Hippocampal Field Activity. J Neurophysiol 91, 583–588 (2004). [DOI] [PubMed] [Google Scholar]

[R88] 88.Lubenov E. V. & Siapas A. G. Hippocampal theta oscillations are travelling waves. Nature 459, 534–539 (2009). [DOI] [PubMed] [Google Scholar]

[R89] 89.Schomburg E. W. et al. Theta Phase Segregation of Input-Specific Gamma Patterns in Entorhinal-Hippocampal Networks. Neuron 84, 470–485 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R90] 90.Senzai Y. & Buzsáki G. Physiological Properties and Behavioral Correlates of Hippocampal Granule Cells and Mossy Cells. Neuron 93, 691–704.e5 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R91] 91.Bergstrom R. A. et al. Automated identification of multiple seizure-related and interictal epileptiform event types in the EEG of mice. Sci. Rep. 3, 1483 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R92] 92.Berens P. CircStat: A MATLAB Toolbox for Circular Statistics. Journal of Statistical Software (2009) doi: 10.18637/jss.v031.i10. [DOI] [Google Scholar]

PERMALINK

This is a preprint.

Interneuron theta phase locking controls seizure susceptibility

Zoé Christenson Wick

Paul A Philipsberg

Cassidy Kohler

Sophia I Lamsifer

Elizabeth Katanov

Christopher D Adam

Kathryn E Gordon

Yu Feng

Lauren M Vetere

Genevra C Donnelly

Corin Humphrey

Denise J Cai

Tristan Shuman

Roles

Abstract

INTRODUCTION

RESULTS

Dentate gyrus PV+ interneurons provide highly precise inhibition at the trough of theta

Fig. 1. DG PV+ cell firing is tightly locked to the theta trough in healthy mice, and is disrupted in epileptic mice.

PV+ interneuron theta phase locking is disrupted in epileptic mice

PhaSER: a tool for real-time theta phase estimation and phase-locked manipulation

Fig. 2. Closed-loop manipulation of inhibitory theta phase locking with PhaSER.

Bidirectional manipulations shift the theta phase preference of PV+ and SOM+ cells

Oscillatory power and theta phase are not altered by phase-specific PV+ or SOM+ cell manipulations

Fig. 3. Manipulating dentate PV+ or SOM+ cell theta phase preference does not impact the referenced LFP.

PV+ cell theta phase locking bidirectionally controls seizure susceptibility

Fig. 4. PV+ interneuron theta phase locking impacts seizure susceptibility.

DISCUSSION

MATERIALS & METHODS

Animals

Epileptogenesis with Pilocarpine

Stereotaxic Surgery

Viral Validation

Chronic Seizure Monitoring

Virtual Reality (VR) Training

Craniotomy & Ground Implantation Surgery

Acute Silicon Probe Recordings & Recording Hardware

Testing Seizure Susceptibility with Kainic Acid

Real-time Signal Processing & Phase-Specific Manipulation with PhaSER

Histology Following Recordings

Post-Processing & Analysis of Single-Unit and Local Field Potential Data

Automated Detection of KA Injection, Seizure Onset, and Seizure Duration

Statistics

Supplementary Material

ACKNOWLEDGEMENTS

Funding:

Funding Statement

Footnotes

Data and materials availability:

REFERENCES

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases