Introduction

The hippocampus is a brain region specialized for learning and memory functions, with a well-described anatomy, molecular functions, and circuit connectivity. Seminal observations regarding human memory have been generated through decades of research into the functions of the medial temporal lobe. These data have driven curiosity around clinical conditions associated with hippocampal dysfunction. Learning and memory disorders are associated with hippocampal pathology. Among these are age-related memory loss and dementias [1, 2], anxiety with PTSD [3, 4], depression [5], and, of particular emphasis here, psychotic disorders [6, 7].

Increased resting state neural activity in hippocampus, (i.e. hippocampal hyperactivity) has been consistently observed in young people with schizophrenia, ascertained in vivo with brain imaging techniques [8,9,10], and localized to CA1 using high resolution MR methods [10,11,12,13]. In human post-mortem tissue analyses, in schizophrenia vs healthy control, we sought potential mechanisms for this hyperactivity [14], and reported reduced expression of GluN1 in dentate gyrus (DG), accompanied by increased markers of excitatory activity in CA3 and CA1. Further, subfield-selective transcriptome analysis of this human hippocampal tissue showed an increase in transcription of excitatory genes in CA3 in schizophrenia, while the transcriptome in CA1 was characterized by a compensatory alteration in the expression of genes involved in the excitatory/inhibitory balance [15]. Recent DG transcriptome analysis in schizophrenia, specifically targeting granule cells, further confirms a decrease in genes related to neurotransmission in schizophrenia cases [16]. While there is some evidence that hippocampal hyperactivity in schizophrenia may be associated with antipsychotic treatment [17, 18], several studies show hippocampal hyperactivity and increased hippocampal glutamate in unmedicated people with schizophrenia [8, 19, 20], a measure which correlates with symptom severity [10], show that antipsychotics reduce hippocampal activity [8], and that this effect could predict good antipsychotic response [21]. Based on these published human studies, we have developed a model of hippocampal pathology in psychosis, hypothesizing that reduced DG excitatory activity in the mossy fiber pathway to CA3, resulting in elevated pyramidal cell activity in CA3 (an outcome predicted by in vivo hippocampal cell studies [22]), with that hyperactivity projected downstream to CA1, degrading learning and memory functions, accompanied by mistaken and abnormal, even psychotic, memories. This model has been further informed by the studies in this DG-inhibited animal model, which seek to assess the correlates of the emergence of hippocampal hyperactivity in the mouse, and the manifestation of neural activity in hippocampus in real time.

The availability of a DG-selective GluN1 knockout (KO) mouse [23] as a reverse-translated preparation of changes in human schizophrenia tissue introduced the studies described here. This KO mouse shows reductions in glutamatergic activity in the mossy fiber pathway from DG to CA3, accompanied by increased NMDAR- and AMPAR-mediated EPSCs in the apical dendrites of CA3 pyramidal cells, accompanied by increased pyramidal cell activity in CA3 and CA1 and behavioral changes associated with psychosis [24]. Moreover, we demonstrated that acute activation of CA3 with a designer receptor exclusively activated by designer drugs (DREADD) can induce a similar behavioral phenotype as KO mouse, suggesting the sufficiency of CA3 hyperactivity in the manifestation of these behavioral changes [24]. Because the reduction of DG GluN1 in the KO mouse develops over the course of 4 months, the time-course of the development of hippocampal hyperactivity in CA3/CA1 has been unclear. Moreover, since psychosis in humans characteristically emerges during adolescence, we are interested in assessing parallel sensitivity of CA3 to DG dysfunction specifically during mouse adolescence.

In this study, we test a DG-placed inhibitory DREADD in the C57BL/6J mouse, enabling temporally-specific analysis. We queried whether DG inhibition during adolescence vs adulthood would differentially affect hippocampal hyperactivity outcomes. We find and report here that DG inhibition during adolescence, but not adulthood, is associated with markers of hippocampal hyperactivity, along with a behavioral phenotype associated with disrupted hippocampal function. Moreover, we report that this elevation in hippocampal activity is driven by the emergence of synchronous bursts of cellular activity throughout the hippocampus, measured from the analysis of local field potentials, a phenomenon we refer to here as “hyper-synchronous events” (HSEs).

Materials and methods

All experiments were performed with experimenters unaware of the treatment of the cohorts tested. Behavioral analysis and terminal procedures were performed at approximately the same time of day, in accordance with institutional and national guidelines on the use of laboratory animals, and protocols approved by the UTSW Animal Resource Center.

Animals

Male and female C57BL/6J mice, aged 4 or 8 weeks old, were purchased from the UTSW Wakeland breeding facility or Jackson Laboratories (Bar Harbor, ME). Animals had food and water available at libitum and maintained at ambient temperature in a 12:12 h light dark cycle. For these studies, we define adolescence as the period from P35 to P70 [25, 26]. Surgery to infuse the AAV was performed at 5 or 9 weeks of age, and Compound 21 (C21) exposure commenced at 6 or 10 weeks, for adolescent and adult groups, respectively. Behavioral tasks were performed at 9 or 13 weeks, and mice were sacrificed for tissue analysis at 10 or 14 weeks.

DREADD injection

Mice were anesthetized with isoflurane and placed into a stereotaxic apparatus (David Kopf Instruments, Tujunga, CA). Openings were drilled bilaterally into the skull, and infusions of an inhibitory DREADD (pAAV-CaMKIIa-hM4D(Gi)-mCherry, Addgene plasmid # 50477) or a control vector (pAAV-CaMKIIa-EGFP, Addgene plasmid # 50469), assigned randomly to mice in each cohort, were given at 0.1 μl per infusion via a Hamilton syringe into the DG, and the syringe was left in place for 5 min. Four infusions/hemisphere were given to cover the entire anterior/posterior extent of the DG, a requirement for inclusion in analysis. The syringe was lowered into the skull at an angle of 10°. Targeted regions are shown in Fig. 1. Adjusted coordinates relative to bregma, taking the 10° angle into account used were: AP −1.70 mm, ML ±1.07 mm, DV −2.13 mm; AP −2.30 mm, ML ±2.15 mm, DV −2.13 mm; AP −2.92 mm, ML ±2.10 mm, DV −2.03 mm; AP −3.52 mm, ML ±2.84 mm, DV −4.82 mm. Placement of the virus was confirmed through the detection of mCherry in the DG.

Fig. 1: Verification of experimental procedures.
figure 1

Localization of the DREADD as visualized by mCherry expression in adolescent (a) and adult (b) mice. Red dot represents targeted location for each of four infusions per hemisphere along the longitudinal axis of the hippocampus, accompanied by representative images showing mCherry fluorescence. Specific coordinates used in the study were adjusted to optimize DREADD expression. Reference images from Allen Mouse Brain Atlas, mouse.brain-map.org and atlas.brain-map.org. Additional images are shown in Supplementary Fig. S1. c Expression of mCherry in granule cell axons in CA3 (left) and limited expression of mCherry in hilar mossy cells (right). GCL granule cell layer. Additional images are shown in Supplementary Fig. S1. d Reduction in cFos+ cell density in DREADD-expressing and control mice in the dentate gyrus following 3, 6, 15, and 21 days of C21 exposure in adolescent and adult mice. Two-way ANOVA revealed a main effect of group (F(3.51) = 27.03, p < 0.0001), but no effect of day (F(3,51) = 0.817, p = 0.490) or interaction (F(9,51) = 0.504, p = 0.865), suggesting a similar, stable degree of inhibition in adolescent and adult DREADD-expressing animals. Adolescent control: N = 4 (3 days), 5 (6 days), 4 (15 days), 6 (21 days). Adolescent DREADD N = 4 (3 days), 6 (6 days), 4 (15 days), 6 (21 days). Adult control: N = 4 (3 days), 3 (6 days), 4 (15 days), 3 (21 days). Adult DREADD N = 4 (3 days), 3 (6 days), 4 (15 days), 3 (21 days). e Compound 21 (C21) consumption by cage, recorded every 3 days for 21 days. Two-way ANOVA, with day as a repeated measure, revealed main effects of age (F(1,77) = 8.123, p = 0.0056) and day (F(4.532,335.1) = 21.69, p < 0.0001), but no interaction (F(6,462) = 2.017, p = 0.062). Adolescent: N = 39; adult: N = 40. f Plasma and g brain concentration of C21 after 3, 6, and 21 days of exposure, and after 24 h of withdrawal following 21 days of exposure. N = 4/group. Data are presented as mean ± SEM.

Full size image

Compound 21 (C21) exposure

Following recovery from surgery, mice were housed in groups of 2 or 3, and provided with a solution containing C21 (50 μg/ml, HelloBio, Bristol, UK) in a 1% w/v sodium saccharin solution as their sole drinking source. When an indwelling probe was placed in the brain, mice were single-housed for the duration of the study. Every 3 days, mice were weighed, the amount of C21 consumed recorded, and fresh solution provided. Following 21 days of exposure, C21 was replaced with tap water.

Separate cohorts of mice were exposed to C21 for analysis of DREADD function or C21 tissue analysis. To assess DREADD function, male mice expressing an inhibitory DREADD in DG were exposed to C21 for 3, 6, 15, or 21 days prior to sacrifice. DREADD function was determined by a reduction in basal levels of cFos+ cell density within DG. For pharmacokinetic analyses of C21, naïve male mice, aged 6 or 10 weeks, were exposed to oral C21 for 3, 6, or 21 days prior to sacrifice. One additional cohort of mice was exposed to C21 for 21 days, then allowed 24 h withdrawal from C21 prior to sacrifice. Mice were then decapitated, trunk blood collected to isolate plasma, and brains immediately frozen in dry ice. C21 levels in the brain and plasma were analyzed by the UTSW preclinical pharmacology core facility.

Probe implantation

A separate cohort of male C57BL/6J mice were implanted with a 32-channel silicon probe (A1x32-6mm-50-177-H32_21mm, Neuronexus, Ann Arbor, MI) 3 days after receiving DREADD or control vector injections. Mice were implanted unilaterally in the right dorsal hippocampus at AP −1.80 mm, ML +1.30 mm, DV −1.90 mm to record dorsal CA1 and DG simultaneously. The implant was secured onto the skull by filling the gap between anchor screws (USMICROSCREW: M08-20-M-SS-P, Seattle, WA) and bottom of the implant using dental acrylic resin. To protect the silicon probe from damage, each probe was mounted on a custom-designed miniature microdrive (3 g weight) [27]. All mice were allowed to recover for 5 days before subsequent experiments. At the end of the experimental window, 7 weeks after the 1st day of recording, mice were euthanized, and the electrode track and tip location were identified with DiI fluorescence on the silicon probes. For these chronic LFP experiments, ten adolescent mice (five DREADD-expressing and five control vector-expressing) were assessed in pairs. To contrast the findings in adolescent mice, an additional three adult DREADD-expressing mice were included.

cFos immunohistochemistry (IHC)

Mice were sacrificed breathing CO2, and transcardially perfused with physiological saline, followed by 4% paraformaldehyde. Brains were passed through a sucrose gradient, then coronally sliced at 50 μm. Every sixth section was quenched free of endogenous peroxidases in 0.3% H2O2, and blocked in a buffer containing 2% normal donkey serum, prior to a 72 h incubation in rabbit anti-cFos polyclonal antibody (1:2000; Synaptic Systems 226-008, Goettingen, Germany). The primary antibody was detected by sequential incubation with biotinylated donkey anti-rabbit IgG (1:1000, Jackson ImmunoResearch, West Grove, PA) and avidin–biotin complex (Vector Laboratories, Burlingame, CA). Diaminobenzidine chromogen was used to detect the immunoperoxidase signal. Images were captured from mounted sections using a light microscope, and cFos+ cell nuclei were counted using ImageJ. Total number of cFos+ cell nuclei divided by the area of the subfields was used to determine cFos+ cell density. Due to differences in staining efficiency between antibody lots, data were normalized to the average of the control animals in each cohort.

Behavior

All behavioral experiments were performed at least 24 h following the cessation of C21, when C21 levels are undetectable in the brain to ensure that the behaviors assessed are not affected by active inhibition of DG. We assessed social memory and spatial working memory, two behaviors reliant on hippocampal function [28,29,30] and impaired in people with psychosis [31,32,33,34]. Sample sizes of >12 mice/group were used, with at least three replications of each task were performed.

Social memory

Mice were habituated to an empty cage for 15 min, after which a 4-week-old C57BL/6J mouse of the same sex as the resident mouse was placed in the cage for 2 min; the time the resident mouse spent in contact/sniffing, following, nosing/grooming, or pawing/general inspection was quantified. The procedure was repeated 24 h later, introducing the same juvenile mouse to the same resident.

Spontaneous alternation

For this task, we used a Y-maze, with arms offset by 120°. Arms were 36 cm long, with walls 27 cm high. For each trial, mice were confined in the home arm for 5 s, then allowed to explore the apparatus for 2 min. Upon the mouse entering either goal arm, the mouse was confined to that arm for 30 s, ending the trial. The animal was then returned to the home arm to begin the next trial. If neither of the goal arms was entered after 2 min, the animal was returned to the home arm for 30 s, and no choice was recorded. Successful alternation is defined as the animal entering the opposite arm as the last trial in which the animal made a choice. Each animal was run through eleven trials, providing ten opportunities for alternation.

In vivo electrophysiology

To detect abnormalities in the freely behaving mouse, we used silicone linear arrays to record LFP profiles along the somatodendritic axis of dorsal CA1 and DG identical to our previously published methods [35]. Mouse pairs were recorded in parallel in a treatment-blind manner using Neuralynx recording system (Digital Lynx SX 256 ch, Bozeman, MT) in their home cage for 1 h daily (between 2 p.m.–6 p.m., 7 p.m.–7 a.m.: dark cycle), starting 1 or 2 days before the C21 administration period (baseline) through the duration of C21 administration for 3 weeks, and for 4 weeks after the cessation of C21 administration totaling 7 weeks of recording. The local field potentials (LFPs) and multi-unit spiking activities (MUAs) were sampled at 32 kHz. The head position of the mouse in the cage was extracted from the light emitting diodes (LEDs) mounted on the headstage. The overhead camera monitored the animal’s movement, and the recording system (Digital Lynx SX) tracked the animal’s head position at 30 Hz.

Analyzing high dimensional LFP data with machine learning GAM methods

Generalized additive models (GAMs) are statistical methods used to model independent non-linear relationships allowing for exploration of the underlying patterns in the data, especially complex neural time series data. In this study, we employed GAMs featuring basic spline relationships—a non-parametric approach—to quantify the connection between the target and factors such as amplitude of LFP and spatial position of the animal. We used a matrix of smoothing functions (n = 20) and we selected a logistic GAM for the purpose of classification of the fast spike events into HSEs and artifacts.

The GAM structure is:

$$g\left(E\left(Y\right)\right)={f}_{1}\left({x}_{1}\right)+\ldots +{f}_{p}\left({x}_{p}\right)\ldots \ldots$$
(1)

where \(Y\) is the dependent variable, E(Y) is the expected value, and g(Y) is the logistic link function that connects the expected value to the predictor variables. The terms f1(x1), …, fp(xp) denote smooth, non-parametric functions.

The initial predictor in our analysis was the averaged amplitude of the time series raw data (LFP) across 32 channels. Additionally, animal position data recorded from the head was also used as a predictor. This positional data specifically focused on quiet and stable periods serving as a distinctive factor to discern from artifacts. Within GAMs, it is possible to adjust the influence of each predictor as needed; however, for our analysis, we presumed equal weight for both LFP amplitude and position. We randomly chose about 100 min of data (time series of varying duration 1–60 min) with fast spikes where we marked HSEs and artifacts in LFP recorded from CA1 and the corresponding position. We used this labeled data from two DG (−) Adolescent mice across 6 days as the training dataset for the GAM (L1). While this GAM layer cleaned up the data, we included another layer to further differentiate between HSEs and high amplitude control events. We randomly chose two single channel time series data from each of the 13 animals, specifically from CA1 and DG consisting of amplitude and normalized amplitude but excluding position data. This exclusion was possible because the HSE labels were already preserved from the first layer (L1) (Supplementary Fig. S3a, b). The model consisted of a two-layer processing approach: initially to eliminate artifacts, followed by determining the accurate count of HSEs/h.

Results

Dentate gyrus inhibition

To assess the effectiveness of DREADD-mediated DG inhibition, we first verified the regional specificity of DREADD expression through localization of the fluorescent marker mCherry. DG was effectively targeted in both adolescent and adult animals (Fig. 1a, b). There was a slight degree of fluorescence detected outside the target region. Visual inspection of the hippocampus showed fluorescence in CA3, but only in axonal fibers in this region (Fig. 1c). We did detect limited expression of the DREADD in cell bodies of the hilus, between the blades of the DG granule cell layer (Fig. 1c). There was no signal detected in the pyramidal cell layer of CA1 or CA3, and while there was a degree of fluorescence outside the pyramidal cell layer in CA1 and CA3, there were no labeled cell bodies in these regions, indicating selective expression in DG granule cells. Additional images are shown in Supplementary Fig. S1. In order to consistently inhibit DREADD-expressing granule cells, we provided the ligand for the DREADD, C21, as the sole source of drinking water for all mice. As the mice remained group-housed throughout the entirety on the experiments, C21 consumption was calculated as total mg/kg for each cage. Chronic exposure to oral C21 decreased total basal activity in DG granule cells, as detected by reduced cFos+ cell density in adolescent and adult mice. Both adolescent and adult mice showed a stable degree of inhibition for the period of C21 administration, suggesting little desensitization of the receptors. This indicates that oral exposure of the DREADD ligand is sufficient to induce long-term inhibition of DREADD-expressing granule cells [36] (Fig. 1d). When assessing C21 consumption, there were significant effects of exposure day and age, suggesting an increased intake in adolescents compared to adult mice over the course of the study (Fig. 1e). However, analysis of plasma and whole brain tissue using LC/MS/MS showed indistinguishable levels of C21 between age groups (Fig. 1f, g), with almost complete elimination after 24 h of withdrawal from C21.

Regional cFos activity analysis

We examined the effect of DREADD-mediated DG inhibition on hippocampal subfield activity by quantifying cFos+ cell density 1 week following the removal of C21 (Fig. 2a), normalized to the controls within each cohort. As statistical analyses revealed no main effect or interaction with sex, male and female mice were pooled for analysis (adolescent: sex: F(1,18) = 1.215, p = 0.285, sex × AAV: F(1,18) = 1.215, p = 0.285; adult: sex: F(1,31) = 0.281, p = 0.600, sex × AAV: F(1,31) = 0.281, p = 0.600). Results from individual subfields were compared using unpaired t-tests, corrected for multiple comparisons. In adolescent mice, C21-induced activation of the DG inhibitory DREADD resulted in an increase in the density of cFos+ pyramidal neurons in CA3, a hyperactivity signal which propagated downstream to CA1 (Fig. 2b–e). However, in adult-prepared mice, the same process resulted in decreased cFos+ cell density in DG, despite the subfield no longer being actively inhibited, and this hypofunction is further apparent in CA3 (Fig. 2f–i). DG inhibition had no effect on activity in CA1 in adult mice.

Fig. 2: Analysis of hippocampal cFos expression.
figure 2

a Experimental schedule indicating timing of C21 exposure and sacrifice for analysis of hippocampal cFos+ cell density. Red X indicates time of sacrifice. b Representative images of hippocampal cFos expression in adolescent mice following DG inhibition. Quantification of cFos+ cell density is shown for c DG, d CA3, and e CA1. f Representative images of hippocampal cFos expression in adult mice following DG inhibition. Quantification of cFos+ cell density is shown for g DG, h CA3, and i CA1. Arrowheads indicate cFos+ cell nuclei. cFos+ cell density in adolescent mice is elevated relative to control mice after DG inhibition in CA3 and CA1. In adult mice, cFos+ cell density remains depressed in DG relative to controls despite the lack of active DREADD-mediated inhibition. Data are analyzed within each subfield using unpaired t-tests, corrected for multiple comparisons. Adolescent: N = 11/group; DG: t(20) = 0.769, p = 0.451; CA3: t(20) = 4.178, p = 0.0009; CA1: t(20) = 5.608, p < 0.0001. Adult: Control N = 18; DREADD N = 17; DG: t(33) = 3.809, p = 0.0017; CA3: t(33) = 2.722, p = 0.020; CA1: t(33) = 0.537, p = 0.595. *, ***, **** represent p < 0.05, 0.001, 0.0001, respectively. Data are presented as mean ± SEM.

Full size image

Behavioral analysis

To test the functional significance of DG inhibition and the resulting changes in hippocampal activity in adolescent vs adult mice, we assessed animal behavior on tests of social memory/cognition and spatial working memory, behaviors which reflect hippocampal function [28,29,30] and are disrupted in people with psychosis [31,32,33,34] following the removal of C21 (Fig. 3a).

Fig. 3: Behavioral analyses.
figure 3

a Experimental timeline indicating timing of C21 exposure and behavioral analyses. Gray box indicates timing of behavioral analyses. b Description of social memory analysis. Mice are allowed to interact with a novel juvenile mouse for 2 min, then exposed to the same mouse for 2 min 24 h later. Time spent in non-aggressive interaction in recorded, and a decrease in interaction time between the first and second tests indicate social memory. Social memory in adolescent (c) and adult (d) mice following DG inhibition. Bars represent time of active social interaction with a juvenile mouse. Data are analyzed using a two-way ANOVA with factors of AAV and test day as a repeated measure, and individual differences were determined using Sidak’s post-hoc comparisons. Adolescent: AAV: F(1,43) = 0.651, p = 0.424; day: F(1,43) = 31.41, p < 0.0001; AAV × day: F(1,43) = 49.01, p < 0.0001. Adult: AAV: F(1,44) = 0.035, p = 0.853; day: F(1,44) = 210.3, p < 0.0001; AAV × day: F(1,44) = 2.767, p = 0.103. Adolescent: Control N = 21; DREADD N = 24. Adult: Control N = 25; DREADD N = 21. e Description of spontaneous alternation. Alternation is defined as the animal making the opposite choice as was made on the previous trial. Spontaneous alternation in adolescent (f) and adult (g) mice following DG inhibition. Groups are compared using unpaired t-tests. Adolescent: t(26) = 3.366, p = 0.002. N = 14/group. Adult: t(25) = 0.759, p = 0.455. Control N = 14; DREADD N = 13. **, **** represent p < 0.01, 0.0001, respectively. Data are presented as mean ± SEM.

Full size image

Social memory is intact if a mouse spends significantly less time interacting with a juvenile partner on their second interaction relative to the first (Fig. 3b). As statistical analyses revealed no main effect or interaction with sex (adolescent: sex: F(1,41) = 0.263, p = 0.611; sex × test day: F(1,41) = 0.350, p = 0.557; sex × AAV: F(1,41) = 0.573, p = 0.453; sex × test day × AAV: (F1,41) = 0.009, p = 0.926; adult: sex: F(1,42) = 2.33, p = 0.134; sex × test day: F(1,42) = 1.012, p = 0.320; sex × AAV: F(1,42) = 0.013, p = 0.909; sex × test day × AAV: (F1,42) = 1.000, p = 0.323), male and female mice were pooled for analysis. Data were then analyzed using a two-way ANOVA, with factors of AAV and test day as a repeated measure. Social memory was compromised in DREADD-modified adolescent relative to control mice (Fig. 3c). Control mice showed a significant decrease in interaction time 24 h following the initial introduction, reflective of intact social cognition. While DREADD-expressing mice showed a lower level of social interaction on the 1st day of testing, after C21-mediated DG inhibition, mice explored the familiar mouse at comparable levels as a novel partner, suggesting an impairment in social recognition. Consistent with the hyperactivity analysis, social memory was normal in both the DREADD- and control vector-expressing adult mice (Fig. 3d).

We next tested spatial working memory by assessing spontaneous alternation in a Y-maze (Fig. 3e). As statistical analyses revealed no main effect or interaction with sex when tested with a two-way ANOVA (adolescent: sex: F(1,24) = 2.057, p = 0.1644; sex × AAV: F(1,24) = 0.129, p = 0.703; adult: sex: F(1,23) = 2.880, p = 0.103; sex × AAV: F(1,23) = 0.208, p = 0.653), male and female mice were pooled for analysis and data were analyzed using an unpaired t-test. We detected a deficit in spatial working memory when DG inhibition occurred during adolescence, evidenced by significantly less alternation between trials relative to control (Fig. 3f). However, when DG was inhibited during adulthood, there was no difference between DREADD-expressing and control mice, suggesting no impairment in spatial working memory (Fig. 3g).

Chronic local field potential monitoring

To further investigate potential hippocampal hyperexcitability temporally and qualitatively, we performed long-term chronic in vivo electrophysiological recording in these mice. We examined five control and five DREADD-expressing adolescent mice, and three DREADD-expressing adult mice chronically implanted with single shank silicone probes to record CA1 and DG, simultaneously (Fig. 4a). The mice were recorded 1 h/day for over 7 weeks (3 weeks of C21 administration followed by 4 weeks of tap water administration) and were allowed to freely move, rest and sleep during the recording session (Fig. 4b). In DG-inhibited mice, we identified abnormal periods in which LFP showed sharp large deflection followed by slow wave component that lasted about a second (Fig. 4c, red arrowhead). The abnormal events, termed HSEs, were different from movement artifacts and control activity or known oscillatory activities, such as sharp-wave ripples (SWRs) or epileptic discharges (Fig. 5a and Supplementary Figs. S2, S3). The spectrogram of HSE shows the spike and wave component, while the wave is absent in the artifact spectrogram in contrast to no spikes in the baseline spectrogram (Fig. 5b).

Fig. 4: Chronic local field potential monitoring of hippocampal hyperactivity.
figure 4

a Representative images of a coronal histology and a mouse post-surgery, illustrating the chronic implantation of a 32-channel silicon linear probe in the right dorsal hippocampus, spanning both the DG and CA1 regions. The microdrive for the probe is notably lightweight, at ~3 g for mice weighing 25–30 g. The head stage is counter-balanced for stress-free behaviors. b Timeline detailing the sequence of probe implantation, initial baseline recording, and the subsequent continuous daily recording schedule, encompassing both adolescence and adulthood in mice. This schedule includes 3 weeks in the C21 phase and 4 weeks in the water treatment phase. c Illustration of an abnormal hyper-synchronous event (HSE), marked by red arrowheads, and its progression over time. These events mostly occur when the subjects are sitting quietly or sleeping. The top panel shows wideband local field potential (LFP) traces from dCA1 (green) and DG (blue), while the bottom panel displays multi-unit spiking activity (MUA) from both regions, along with a summed spatial representation of MUAs. Different colors of spiking events denote distinct spiking activities.

Full size image
Fig. 5: Machine learning based approach to identify HSEs.
figure 5

a Representative waveforms of the hyper-synchronous events HSEs (red) followed by a wave, movement artifacts (black), and high amplitude control activity (green). b Representative power spectrograms of the different waveforms: HSE showing the high magnitude sharp spike followed by a low magnitude wave (top panel), multiple artifacts without the wave component (middle panel), and baseline spectrogram showing multiple frequency bands without any sharp spikes (bottom panel).

Full size image

To precisely and effectively identify the sporadic HSE events from the large-scale LFP data set, we employed a machine learning GAM algorithm. The initial layer (L1) effectively classified artifacts and HSEs throughout the 1-h recording period across 13 animals for 50 days, amounting to about 35,000 min of data. We validated the model on 60 min of data from two randomly selected DG (−) Adolescent mice, achieving 99.6% accuracy in distinguishing HSEs from artifacts compared to manual classification. This GAM L1 successfully differentiated HSEs from artifacts to interpret the data we recorded from DG-DREADD and control mice. The second layer GAM L2 successfully differentiated HSEs from the control group events from the first layer. The results of the two-layered GAM indicated that DG-inhibited adolescent mice developed increased hippocampal CA1 activity (Fig. 6a and Supplementary Fig. S4). The mean number of HSEs/h in DG-inhibited adolescent mice was 23.7 ± 13.1 events/h, which was significantly higher than the mean of the control group at 1.2 ± 1.2 events/h. When the GAM analysis was extended to the adult DG-inhibited group the mean HSEs/h was 1.5 ± 2.6 events/h, there was a significant difference between the DG-inhibited adolescent and DG-inhibited adult groups (Fig. 6a, right panel). When the individual mice data across the entire recording period were analyzed, the HSEs/h exhibited a notable increase in DG-inhibited adolescent mice around day 17 (Fig. 6a, left panel), reaching a maximum of ~80 events/h (Fig. 6a). In contrast, the control and DG-inhibited adult groups consistently maintained rates below 5 events/h throughout the same period (Fig. 6a). This showed increased activity in CA1 of the hippocampus when DG was inhibited only during adolescence. Additionally, the observed HSE activity not only exhibited a significant increase during the DREADD manipulation period with C21, but it also sustained a notably high frequency of HSEs even after reverting to the regular water period, during which DG function was presumed to return to its normal state (Fig. 6a). We also analyzed the individual characteristics of the HSEs such as their peak amplitude, time duration, cross-correlation (CC) between CA1 and DG. DG (−) Adolescent group HSE events exhibited high correlation across the 50-day period, the normalized CC factor seldom dropping below 0.5 (Fig. 6b), the DG (−) Adolescent group averaging at 0.88 ± 0.04 significantly differing from the average CC of 0.55 ± 0.1 in the Adolescent control group and 0.58 ± 0.04 in the DG (−) Adult group (Fig. 6b, right panel). The amplitude of HSEs fluctuated across days, with the average over 50 days totaling 1.9 ± 1.5 mV. However, the time duration of HSEs exhibited a trend similar to the number of HSEs/h over the 50-day period, with an overall average of 41.5 ± 32.4 ms (Fig. 6c). Overall, the machine learning based GAM analysis revealed that DG-inhibited mice showed a significant increase in the likelihood of HSE presentation beginning in the 1st week of C21 administration, and continued until sacrifice, 4 weeks from the end of C21 administration.

Fig. 6: Dynamics and characterization of HSEs.
figure 6

a A summary overview showing the count of HSE events detected using the GAM machine learning algorithm. The DREADD-expressing (DG (−)) Adolescent trend of HSEs/hour (red) increased around day 17 to an average of about 80 across 50 days while the control (CTRL, green) and DG (−) Adult (blue) group means stayed consistently low. Individual DG (−) Adolescent mice HSEs/hour across 50 days increased around day 17 and increased after day 30 as well. Individual CTRL adolescent mice HSEs/hour did not significantly increase around day 17 different from the DG (−) Adolescent groups. Individual DG (−) Adult mice HSEs/hour did not significantly increase around day 17 (DG inhibition) different from the DG (−) Adolescent groups (left panel). Statistical analysis revealing a significant increase in HSE events exclusively in the adolescence DG-DREADD inhibited group, in contrast to the Adolescent control or Adult DG-DREADD groups. The right panel graph shows the average number of HSEs per hour in the DG silenced Adolescent group (red, N = 5 animals) at 23.7 ± 13.1 which is significantly higher than the average at 1.2 ± 1.2 in Adolescent control (green) and the average of 1.5 ± 2.6 in the Adult control groups (blue) (*p = 0.008, z = −2.619 and *p = 0.036, z = −2.249, respectively, Mann–Whitney U two-tailed, N1 = 5 animals, and adult N2 = 3 animals). b DG (−) Adolescent group HSE events were highly correlated across the 50 days rarely dropping below 0.5 (normalized factor, in red), while the control group activity showed correlation between CA1 and DG, it is not consistently high in both Adolescent and Adult groups (left panel). The right panel bar graph shows the average of the CC factor, the average CC of the DG (−) Adolescent group is 0.88 ± 0.04 (red, N = 5 animals) which is significantly different from the average CC of 0.55 ± 0.1 Adolescent control group (green) and 0.58 ± 0.04 in Adult control group (blue) (*p = 0.008, z = −2.611, and *p = 0.036, z = −2.236, respectively, Mann–Whitney U two-tailed, control N1 = 5 animals, and adult N2 = 3 animals). c Characteristics of HSEs: amplitude of the HSEs across days is not consistently high while the average of all 50 days is 1.9 ± 1.5 mV (top panel), time duration of HSEs across days has a similar pattern to the number of HSEs/h across 50 days and the total average is 41.5 ± 32.4 ms calculated using the sharp component (bottom panel). *represents p < 0.05.

Full size image

Discussion

DG inhibition with C21 exposure in DREADD-expressing mice during adolescence results in hippocampal hyperactivity in CA3 and CA1 associated with the emergence of HSEs and the presentation of psychosis-like hippocampal behaviors. In adult mice, DG inhibition depresses hippocampal activity, reveals infrequent HSEs, and fails to induce a behavioral phenotype. The data reported here demonstrate a period of susceptibility during adolescence, where DG inhibition generates subsequent hippocampal hyperactivity.

To quantify hippocampal hyperactivity, we assessed basal expression of the immediate early gene cFos as a proxy for neuronal activity. Although quantification of cFos expression is typically used to assess the consequences of acute external events [37], cFos+ neurons are present at a high level in the hippocampus of mice at rest [38], and cFos+ cell density can be used to determine the activity of populations of neurons even in the absence of salient external stimuli [39]. cFos+ neuronal density was examined within the granule cell layer of DG, and the pyramidal cell layers of CA3 and CA1. We demonstrate that when DG is inhibited during adolescence, hippocampal activity is increased, specifically in CA3 and CA1. This pattern of activity, and the scale of change, is similar to what is seen in the DG-GluN1 KO mouse [24], suggesting that distinct sources of DG inhibition can similarly affect downstream hippocampal activity, depending on the developmental timing. We contrast this to mice with DG inhibited during adulthood. Even a week following the cessation of active DG inhibition, DG activity remains depressed, and this hypoactivity is reflected in depressed activity in CA3. This effect during “recovery” on DG is particularly noteworthy. The adolescent DG appears to be resilient and recovers its function 1 week following the removal of C21, while the adult DG appears to suffer longer-term inhibition after C21. The mechanisms underlying this persistent decrease in DG activity remain unclear and are worthy of additional investigation.

To examine functional behavioral outcomes, we carried out analyses of hippocampally mediated paradigms which show psychosis-relevance, and have been demonstrated with validated animal models of schizophrenia [40,41,42,43,44]. Unlike some paradigms used to model psychosis in animals, the behaviors assessed here focus on cognition-driven outcomes [45]. Social cognition assesses the ability of the animal to recognize a previously encountered social partner; this behavior is deficient in people with schizophrenia [46], and plausibly contributes to impaired social functioning in people with the disorder. While similar to general cognition, social cognition appears to be independently regulated, and therefore is a distinct construct [47]. Spontaneous alternation assesses spatial working memory, a behavior impaired in humans with psychosis [31, 32], through the innate preference of an animal to explore a previously unexplored area in a Y-maze, even without the delivery of a reward [48]. Our results show deficient social cognition and spatial working memory with adolescent, but not adult DG inhibition. This is consistent with the deficit seen in the reverse-translated DG-GluN1 KO mice, and seen with direct excitation of CA3 [24], suggesting this is a direct effect of the induced hippocampal hyperactivity in the adolescent DG-inhibited mice.

The chronic in vivo electrophysiology demonstrates that in adolescent mice, HSEs emerge in the 2nd week of DG inhibition, and the frequency of HSEs increase over the remaining course of DG inhibition. This suggests a minimum duration of DG dysfunction necessary to induce a pathological state in the hippocampus. That the alterations in hippocampal activity were not due to a generalized increase in total power in the hippocampus was a surprise. The HSEs described here are fundamentally different from other known types of electrophysiological network activity, such as sharp-wave ripples (SWRs) or any kind of epileptic discharges (Figs. 4c, 5a, b and S2). The SWRs are more concentrated in the cell layer of the CA1, CA2, and CA3 areas, not spanning the DG to CA1. Although HSEs might bear resemblance to interictal spikes reported in epilepsy, we have not consistently observed the high-frequency burst-like activity characteristic of seizures in epileptic waveforms. Instead, these changes were associated with spontaneous and discrete occurrences of HSEs as episodic events. Moreover, despite the brief (3 weeks) period of DG inhibition, the presentation of HSEs continues for weeks following the cessation of inhibition, suggesting a persistent, or even permanent effect resulting from DG inhibition during adolescence.

Technically, our study has demonstrated the effectiveness of interpretable machine learning in extracting crucial features from neural data. The analysis of this study contributes to a deeper understanding of the electrical activity associated with psychosis-related disorders. Generalized linear models have been used in analyzing intracortical electrophysiological data to extract properties from neural activity such as place fields from CA1 regions [49, 50] however this is the first time modified generalized additive models (GAM) has been used for multimodal analysis for processing neural data, differentiating HSEs from artifacts in our study, without extensive data cleaning before using machine learning algorithms.

The overall implications of this study show that in this mouse reverse-translated schizophrenia preparation based on the pathology observed in human psychosis, DG inhibition causes downstream hippocampal hyperactivity when the inhibition is delivered during an adolescent (but not adult) period. With temporal control of DG activity in this animal preparation, we show that adolescence is a particularly sensitive time for the generation of hippocampal hyperactivity relative to adulthood. This could reflect a developmental match between the sensitivity to DG inhibition in mice, and the common emergence of psychotic illness in humans during adolescence, both associated with hippocampal hyperactivity. Of particular interest is that this hippocampal hyperactivity is associated with the presentation of a novel pattern of synchronous electrical activity in the mouse hippocampus. That hippocampal hyperactivity in response to DG inhibition, linked to adolescence, could translate to functional alterations in human psychosis, especially as adolescence is known to be a period of psychosis sensitivity in humans [51].

These results (i) confirm hippocampal hyperactivity as a consequence of DG inhibition; (ii) demonstrate a sensitive period during which inhibition of excitatory DG signaling can induce long-lasting hippocampal hyperactivity; (iii) show that this hippocampal hyperactivity is associated with psychosis-relevant behaviors; and (iv) reveal that the pattern of this hyperactivity may be reflected by patterned excitatory HSEs. This provides support for a vulnerability model where DG inhibition could impact overall hippocampal activity only when this functional lesion is initiated during adolescence. These data support the growing realization of the critical nature of the adolescent period and hippocampus for adult-onset psychotic disorders as is supported here [52,53,54].