Pharmacological reduction of reverse-translated hippocampal hyperactivity in mouse: relevance for psychosis

Abstract

Hippocampal hyperactivity (HH) is a potential biomarker in schizophrenia psychosis, which also appears in several other brain disorders, compromising specificity. We hypothesized that the reversal of HH in an established, reverse-translational animal preparation, coupled with a behavioral marker of psychosis may be a predictor of antipsychotic efficacy of a medication. We used a chemogenetic reverse-translational mouse preparation relevant to schizophrenia psychosis which shows HH and aberrant psychosis-relevant behaviors, specifically disrupted social recognition memory (SRM). Mice with and without HH were treated with three drugs; two known antipsychotics and one HH-reducing anticonvulsant, to assess their effects on both HH and SRM performance. All animals received one of the four treatments: vehicle (N = 15–24), haloperidol (N = 8–15), xanomeline (N = 8–13) or levetiracetam (N = 6–15) and were subsequently tested for baseline c-Fos protein expression within the hippocampal subfields (CA3 and CA1) as a measure of neuronal activity, or tested with the SRM task as a measure of social memory. All three drugs acutely reduced baseline HH compared to vehicle treatment. Subacute administration of haloperidol or xanomeline, the two drugs known to have antipsychotic activity, but not levetiracetam, normalized the SRM behavior to control levels. These results suggest that the reversal of HH alone cannot be a predictor of antipsychotic efficacy of an experimental drug and HH as a biomarker could benefit from a more sensitive readout approach.

Introduction

A well-established need exists for novel and more effective treatments for psychosis. Current antipsychotic medications show insufficient clinical effectiveness and significant side effects [1–7]. Animal preparations which can, to an extent, mimic the human pathophysiology and display quantifiable biomarkers can facilitate the process of screening drug candidates for antipsychotic efficacy, especially those outside of dopamine-mediated outcomes. One of those biomarkers identified in recent years is hippocampal hyperactivity (HH) [8–16].

Hippocampal hyperactivity is an emerging human biomarker in psychosis, and the phenomenon also appears in several different brain diseases, such as epilepsy, Alzheimer’s disease and early dementia [8–16]. HH is defined here as persistent, elevated baseline neuronal activity in the hippocampus when compared to healthy controls. HH is often present before and in early psychosis, when measured with high resolution magnetic resonance imaging techniques [12, 17–21]; a subset of human psychotic individuals exhibit elevated hippocampal activity at rest, with lower activation during tasks, when compared to healthy controls [12]. Neuroimaging studies indicate an association between HH and progression to psychosis from a premorbid state [10, 18, 21, 22]. We presume that HH can be induced by different mechanisms in these different disorders, often resulting in distinctive molecular or behavioral outcomes [23–27].

Models of psychosis in small animals can be challenged by inadequate validity for human psychosis, because there is no reliable marker of pathology to define the model [28]. However, our laboratory completed human post-mortem tissue studies of hippocampal subfields from people with schizophrenia vs. healthy controls. These experiments demonstrated a decrease in NR1 protein in the dentate gyrus (DG) with concurrent increase in proteins related to synaptic transmission in CA3 and CA1. Such protein expression patterns suggest low glutamatergic transmission in DG and neural hyperactivity in downstream subfields [29–31]. These findings were also accompanied by synaptic reorganization, represented by an increased number of thorny excrescences (giant post-synaptic spine head clusters) on proximal apical dendrites of CA3 pyramidal cells [30]. Subsequent transcriptome analysis revealed upregulated excitatory genes in CA3 and altered excitatory-inhibitory balance genes in CA1 [32].

Based on these human tissue findings, we attempted to replicate the schizophrenia pathology in mice (reverse-translated) first using a DG-specific NR1 knockout mouse, as a model analogous to both the protein expression and activity imbalance found in human tissue [33, 34]. Subsequently, to ensure temporal specificity, we used Designer Receptors Exclusively Activated by Designer Drugs (DREADDs) in adolescent mice, as previously described [35]. DREADDs started an era of “chemogenetic” experiments in animals, bringing valuable spatiotemporal specificity to subchronic and chronic experiments [36]. We have previously shown in this preparation that subchronic DG granule cell inhibition in adolescence leads to lasting hippocampal hyperactivity (as measured by increased basal cFos+ cell density within the subfields, see Fig. 1) as well as psychosis-like behaviors, including disrupted social recognition memory (SRM, see Fig. 2) [34, 35]. Deficits in social cognition are commonly seen in schizophrenia [37–41] and social memory performance relies on the hippocampus and has clinical relevance [38, 41–45]. These phenotypes were not observed in the DREADD-expressing mice with DG inhibition occurring in adulthood, suggesting developmental sensitivity. Local field potentials studies in the DREADD-expressing mice (recorded from DG and CA1) do not show continuous, sustained increase in hippocampal pyramidal cell activity, but rather irregular episodes of high spiking activity [35].

Fig. 1. Subchronic DG inhibition results in hippocampal hyperactivity.

A Example of dorsal CA3 c-Fos protein expression in a control animal. B Example of dorsal CA3 c-Fos protein expression in a Compound 21-treated animal. C Example of DREADD expression in the DG and mossy fibers. D Timeline of animal preparation from surgery to testing. E Compound 21 (C21) exposure leads to a lasting increase of c-Fos protein expression within the CA3 and CA1, but not within DG (measured a week after C21 withdrawal). 2-way ANOVA (F(3,81) = 6.248, p = 0.0007) with Fisher’s LSD test, N = 15–24. Panel (A): p = 0.0795, 95% CI[–19.09, 1.085]. Panel (B): p = 0.0003, 95% CI[–28.44, –8.893]. Panel (C): p = 0.0066, 95% CI[–31.99, –5.337].

Fig. 2. Hippocampal hyperactivity affects social memory performance.

A Social Recognition Memory is significantly disrupted in Compound 21 treated animals, one week after cessation of Compound 21 administration (Dunnett’s T3 p = 0.0364, 95% CI[–0.1943, –0.004239], N = 19–21, Brown-Forsythe ANOVA (F(7, 77.85) = 2.562, p = 0.0198)). B Effects of drug treatment on behavior in control animals. Only haloperidol administration caused significant disruption of SRM (Dunnett’s T3 p = 0.0258, 95% CI[–0.2347, –0.01057], N = 11–21, Brown-Forsythe ANOVA (F(7, 77.85) = 2.562, p = 0.0198)).

In this study, we aim to assess whether treatment with antipsychotic medications reliably antagonizes HH, and whether treatment with drugs which can affect HH show antipsychotic potential. We selected three distinct and informative medications.

Haloperidol is an established first-generation antipsychotic, a potent D2 receptor antagonist [46–50]. Motor side effects, reflecting its potent antidopaminergic activity, are common and debilitating [51, 52]. This drug was chosen due to its predominant antidopaminergic activity in vivo and its antipsychotic efficacy in the clinic.

Xanomeline is a M4/M1 muscarinic agonist, originally developed for cognitive impairment associated with Alzheimer’s disease [53–55]. Recently, xanomeline has been paired with a peripheral muscarinic antagonist, trospium chloride [56], to mitigate its peripheral side effects. This novel formulation was approved for schizophrenia by the FDA in September 2024 and is sold as “Cobenfy” [57]. The major effects of xanomeline on hippocampal and striatal signaling have been described [58, 59] and the clinical antipsychotic efficacy of its combination with trospium chloride is documented [60–64]. Cobenfy (xanomeline-trospium) is the first antipsychotic drug without primary action at the dopamine receptor [65].

Levetiracetam is a commonly used atypical anticonvulsant and has no clinical use in psychiatry [66]. The mechanism of action is multifaceted and involves the ubiquitous presynaptic SV2A protein, voltage-gated calcium channels and Kv3.1 potassium channels [67–73]. Interestingly, it is a drug with clinically demonstrated pro-cognitive and HH-normalizing property in humans [8, 74–76], with some reports of psychiatric side effects in individuals with epilepsy [77–80].

Here, we test the hypothesis that a compound with antipsychotic action in humans can be distinguished in the laboratory using HH and relevant animal behavioral outcomes.

Methods

Animals

4 weeks old male C57BL/6 J mice (Jackson Labs, Bar Harbor, ME) were purchased from Jackson Laboratories and housed at UTSW vivarium in temperature-controlled rooms on a constant 12:12 light dark cycle with ad libitum food and water. Experiments were conducted during the light cycle at similar time each day. See Supplementary Materials and Methods.

DREADD placement

5 weeks old mice were anesthetized with isoflurane (1–1.5% with 1 l/min flow) and placed within stereotaxic apparatus. Skull burr holes were prepared after an aseptic scalp incision. A Hamilton microliter syringe was lowered into the skull at a 10° angle. Inhibitory DREADD (pAAV-CaMKIIa-hM4D(Gi)-mCherry, plasmid #50477, Addgene, Watertown, MA) was infused at 0.1 μl per infusion. Infusions allowed the delivery of ~10¹¹ TU/infusion, with 4 infusions per hemisphere to ensure coverage of the longitudinal axis of DG. See Fig. S2.

Target coordinates for the AAV infusion: (1) AP -1.70 mm, ML ± 0.70 mm, DV –2.04 mm (2) AP –2.30 mm, ML ± 1.75 mm, DV –2.04 mm (3) AP –2.92 mm, ML ± 1.85 mm, DV –1.95 mm AP (4) –3.52 mm, ML ± 2.25 mm, DV –4.50 mm.

Animal eyes were protected with Paralube ointment (Dechra, USA) during the surgery, and each animal received carprofen (10 mg/kg, sc) at the end of surgery. Following the surgery, the wound was closed with non-absorbable sutures and lidocaine gel was topically administered. Subsequently, each animal was transferred to a heated recovery cage, and once awake and mobile, transferred to its home cage. Carprofen injections continued once per 24 h, for a total of 72 h.

Compound 21 (C21) exposure

Following one week recovery from surgery, mice were group housed (2-3/cage) and provided with ligand for the DREADD [81], Compound 21 (C21; 50 mg/L, dissolved in 1% saccharin sodium solution) or vehicle ad libitum as their sole source of drinking water. Bottles were covered with aluminum foil to protect from light. Mice and bottles were weighed every three days to determine the average consumption per cage. Subsequently, a fresh solution was provided, in order to avoid C21 breakdown. The duration of C21 or vehicle exposure amounted to 21 days. Following this period, C21/vehicle was removed and animals were provided with tap water. The average C21 consumption remained stable throughout the treatment.

All experiments followed institutional guidelines, approved by The Institutional Animal Care and Use Committee at UT Southwestern.

Drug preparation

Acute treatment

Haloperidol (Sigma Aldrich, Cat. No. H1512, St. Louis, MO) was dissolved in a minimal amount of glacial acetic acid. The solution was subsequently diluted to the desired concentration using sterile 0.9% sodium chloride. The resulting haloperidol solution was balanced to pH=6 and filtered with a 0.22μM syringe filter, prior to administration. Haloperidol was administered at 0.3 mg/kg intraperitoneally.

Levetiracetam (TCI, Cat. No. L0234, Portland, OR) was diluted to the desired concentration with sterile 0.9% sodium chloride. The resulting levetiracetam solution was filtered with a 0.22μM syringe filter prior to administration. Levetiracetam was administered at 5 mg/kg intraperitoneally.

Xanomeline tartrate (MCE, Cat. No. HY-105182A, Monmouth Junction, NJ) was dissolved in 5% DMSO, 15% Solutol and 80% water. The target dosage was 3.2 mg/kg of free base.

All drugs were administered 90 min before sacrifice, to capture the peak of c-Fos protein expression [82].

Subacute treatment

Haloperidol was diluted into a 5 mg/L stock solution with distilled water. The solution to be administered orally (as the sole source of drinking water) was prepared by dilution of the stock with 1% saccharin sodium in water. The target dosage was 1.75 mg/kg/day.

Levetiracetam was dissolved in 1% saccharin sodium solution. The solution was administered orally as the sole source of drinking water. The target dosage was 100 mg/kg/day.

Xanomeline solution was prepared by dissolving xanomeline tartrate in 5% DMSO, 15% Solutol and 80% water. The solution was loaded into osmotic minipumps (Alzet, Cat. No. 1007D, Cupertino, CA) and implanted subcutaneously between scapulae under isoflurane anesthesia 72 h before the last day of testing. The target dosage was 8 mg/kg/day of free base.

Plasma levels quantification

Dosage of haloperidol, levetiracetam and xanomeline (N = 4–5 each) was adjusted, in order to establish intraperitoneal, oral or subcutaneous dose resulting in an equal to human effective plasma concentration range [74, 83]. Previous research literature was used as a reference for the dosage of xanomeline [58]. See Supplementary Materials and Methods for a detailed protocol.

Behavioral testing

Social recognition memory

As previously published [84, 85], experimental mice were placed into an empty cage and allowed to habituate for 15 min. Following this, a juvenile (3–4 weeks old) male C57BL/6 J mouse was placed into the cage for 2 min, and the amount of time the resident spent engaged in nosing, grooming, following, or active social interaction was recorded. 24 h later, the procedure was repeated with the same resident/juvenile pairing. Tested mice which interacted less than 20% of total interaction time (i.e. <24 s) on the first day were excluded from analysis.

Calculation of SRM index

Established formula for Social Recognition Memory index was used to quantify the interaction between tested adult mice and paired juveniles [44, 86–88]. Due to the use of only one intertrial interval in this study, the equation was simplified. Otherwise, Day 2 signifies the testing phase, while Day 1 signifies the training phase.

$$SRM=\frac{{Interactiontime}_{Day2}-{Interactiontime}_{Day1}}{{Interactiontime}_{Day1}+{Interactiontime}_{Day2}}$$

c-Fos labeled tissue imaging and analysis

Stained tissue was imaged using a Leica light microscope and 10x magnification (see Fig. S3). Scale was determined using a dedicated microscale (Imaging Research Inc., USA). A subset of the tissue was imaged by UTSW Whole Brain Microscopy Facility (RRID:SCR_017949) using a NanoZoomer (Hamamatsu Photonics, Japan) and 20x magnification. See Supplementary Materials and Methods.

Statistical methods

Statistical analyses were performed using GraphPad Prism (Boston, MA). ANOVA tests were used for initial analyses. c-Fos multiple comparisons were performed using uncorrected Fisher’s LSD, due to small number of a priori comparisons with small effects and the risk of introducing Type II error. Dunnett’s test was used for multiple comparisons of behavior, where indicated (assumed heteroscedasticity). Effect size was calculated using Cohen’s d. See Table S2.

Results

Hippocampal hyperactivity can be caused by subchronic inhibition of dentate gyrus in adolescence and alters hippocampus-dependent behaviors

c-Fos protein expression

We measured c-Fos protein expression within the hippocampal subfields in group-housed mice in their home cages, at rest, as a proxy for baseline hippocampal activity.

7 days after C21 withdrawal, there was no change in DG (Fig. 1E) between C21-treated animals and vehicle-treated animals. We found elevated c-Fos in saline injected C21-treated animals (test group) compared to saline injected vehicle-treated animals (control group) in CA3 by 72.57% (Fisher’s LSD p = 0.0003, d = –1.187, N = 15–24, 2-way ANOVA (F(3,81) = 6.248, p = 0.0007)) and CA1 by 68.71% (Fisher’s LSD p = 0.0066, d = –0.828, N = 15–24, 2-way ANOVA (F(3,79) = 2.945, p = 0.0380)) (Fig. 1E).

Behavioral testing

SRM test measured the difference in interaction time between the test mouse and juvenile conspecific across two trials, 24 h apart. Calculated SRM indices demonstrate that control vehicle-treated mice show a significant decrement in interaction time on the second trial relative to the first, suggesting intact social memory, while control C21-treated mice interact approximately the same amount of time on both trials. The difference between the SRM indices (Fig. 2A) identifies an aberrant behavior in control C21-treated animals, compared to vehicle-treated animals (Dunnett’s T3 p = 0.0364, d = –0.931, N = 19–21, Brown-Forsythe 2-way ANOVA (F(7,77.85) = 2.562, p = 0.0198)).

Control animals: the influence of pharmacological treatment on hippocampal subfields and behavior

c-Fos protein expression

Next, we measured basal c-Fos protein expression within the hippocampal subfields in mice which did not undergo DG inhibition at all, but each animal received either vehicle or selected drug before sacrifice. The drug dosage was established to mimic effective human plasma levels (see Supplementary Materials and Methods).

Haloperidol (0.3 mg/kg ip) did not affect c-Fos protein expression in DG, CA3 or CA1 in control animals (Fisher’s LSD p = 0.8717, d = −0.090; p = 0.4429, d = −0.413; p = 0.8019, d = 0.205; respectively; N = 11–15).

Xanomeline (3.2 mg/kg sc) increased c-Fos protein expression in CA1 in vehicle-treated animals by 113.8% (Fisher’s LSD p = 0.0002, d = –1.154, N = 9–24, 2-way ANOVA (F(3,79) = 2.945, p = 0.0380), Fig. 3B).

Fig. 3. Effects of drug treatment in control animals.

A None of the medications affected CA3 c-Fos protein expression in control animals. 2-way ANOVA (F(3,81) = 6.248, p = 0.0007) with Fisher’s LSD test, Panel (a): p = 0.4429, 95% CI[-16.82, 7.423], N = 8–24. Panel (b): p = 0.1206, 95% CI[–20.76 to 2.454], N = 9–24. Panel (c): p = 0.1783, 95% CI[–22.80, 4.303], N = 6–24. B Only xanomeline treatment produced an increase in CA1 c-Fos protein expression in control animals. 2-way ANOVA (F(3,79) = 2.945, p = 0.0380) with Fisher’s LSD test, Panel (d): p = 0.8019, 95% CI[–14.44, 18.62], N = 8–24. Panel (e): p = 0.0002, 95% CI[–47.41, –15.76], N = 9–24. Panel (f): p = 0.9315, 95% CI[–17.68, 19.28], N = 6–24.

Levetiracetam (5 mg/kg ip) increased c-Fos protein expression in DG in vehicle-treated animals by 79.25% (Fisher’s LSD p = 0.0045, d = 0.084, N = 6–24, 2-way ANOVA (F(3,79) = 4.469, p = 0.0060), Fig. S4).

Behavioral testing

Haloperidol showed disrupted SRM in vehicle-treated control animals (Fig. 2B, Dunnett’s T3 p = 0.0258, d = −1.063, N = 15–21, Brown-Forsythe ANOVA (F(7,77.85) = 2.562, p = 0.0198)).

We found no effect of levetiracetam (100 mg/kg/day po) or xanomeline (8 mg/kg/day sc) on SRM in vehicle-treated control animals (Dunnett’s T3 p = 0.9946, d = 0.075; p = 0.995, d = 0.085; respectively, N = 11–21).

Hippocampal hyperactivity: the influence of pharmacological treatment by subfield

c-Fos protein expression

Haloperidol and levetiracetam had a normalizing effect on hippocampal c-Fos protein expression, decreasing it to levels seen in control animals. Xanomeline normalized c-Fos protein expression in CA3, but not in CA1 (see Discussion).

Haloperidol (0.3 mg/kg ip) reduced the baseline c-Fos protein expression by 53.7% in CA3 (Fisher’s LSD p = 0.0059, d = 0.849, N = 11–15, 2-way ANOVA (F(3,81) = 6.248, p = 0.0007), Fig. 4A) and by 48% in CA1 (Fisher’s LSD p = 0.0327, d = 0.762, N = 9–15, 2-way ANOVA (F(3,79) = 2.945, p = 0.038), Fig. 4B).

Fig. 4. Effects of drug treatment on hippocampal hyperactivity.

A All medications significantly decrease c-Fos protein expression in CA3. 2-way ANOVA (F(3,81) = 6.248, p = 0.0007),with Fisher’s LSD test, Panel (a): p = 0.0059, 95% CI[4.967, 28.54], N = 11–15. Panel (b): p = 0.0011, 95% CI[9.124, 35.12], N = 8–15. Panel (c): p = 0.0019, 95% CI[7.927, 33.92], N = 8–15. B Haloperidol and levetiracetam influence c-Fos protein expression in CA1. 2-way ANOVA (F(3,79) = 2.945, p = 0.038) with Fisher’s LSD test, Panel (d): p = 0.0327, 95% CI[1.574, 35.71], N = 9–15. Panel (e): p = 0.7877, 95% CI[–15.32, 20.13], N = 8–15. Panel (f): p = 0.0207, 95% CI[3.297, 38.75], N = 8–15.

Xanomeline (3.2 mg/kg sc) reduced the baseline c-Fos protein expression by 46.9% in CA3 (Fisher’s LSD p = 0.0011, d = 1.366, N = 8–15, Fig. 4A), but not in CA1 (Fisher’s LSD p = 0.7877, d = 0.101, N = 8–15, Fig. 4B).

Levetiracetam (5 mg/kg ip) reduced the baseline c-Fos protein expression by 46.48% in CA3 (Fisher’s LSD p = 0.0019, d = 1.252, N = 8–15, Fig. 4A) and by 43.5% in CA1 (Fisher’s LSD p = 0.0207, d = 0.919, N = 8–15, Fig. 4B).

Both haloperidol and xanomeline, but not levetiracetam, reduced DG c-Fos protein expression in C21-treated animals (Fisher’s LSD p = 0.0312, d = 0.730; p = 0.0029, d = 1.149; p = 0.0629, d = 0.691; respectively. 2-way ANOVA (F(3,79) = 4.469, p = 0.006), Fig. S5).

Behavioral testing

Haloperidol administered for 72 h (1.75 mg/kg/day po) rescued the SRM phenotype in C21-treated animals and normalized the SRM index, as evidenced by comparison with control C21-treated animals (Dunnett’s T3 p = 0.0457, d = 1.023, Brown-Forsythe ANOVA (F(7,77.85) = 2.562, p = 0.0198), N = 11–21, Fig. 5).

Fig. 5. Effects of drug treatment on social recognition memory in animals with HH.

Haloperidol and xanomeline lead to a decrement in SRM index and normalization of SRM performance. Brown-Forsythe ANOVA (F(7,77.85) = 2.562, p = 0.0198) with Dunnett’s T3 MC test. Panel (A): adjusted p = 0.0457, 95% CI[0.001517, 0.2272], N = 15–19. Panel (B): adjusted p = 0.0429, 95% CI[002577, 0.2257], N = 13–19. Panel (C): adjusted p = 0.8200, 95% CI[–0.9249, 0.2220], N = 15–19.

Xanomeline (8 mg/kg/day sc) showed an effect similar to haloperidol, normalizing the SRM to control levels (Dunnett’s T3 p = 0.0429, d = 1.084, Fig. 5).

Levetiracetam (100 mg/kg/day po) did not affect SRM when compared to controls (Dunnett’s T3 p = 0.82, d = 0.435, Fig. 5).

Discussion

We postulated that antipsychotic drug administration would normalize HH in the recently established chemogenetic reverse-translational animal preparation with the psychosis-relevant biomarker (HH) as well as a behavioral phenotype (aberrant social recognition memory; SRM). We aimed to test whether the reversal of HH alone can be used as an indication of antipsychotic medication efficacy.

We have replicated previous findings showing that subchronic adolescent DG granule cell inhibition induces adult lasting hyperactivity of CA3 and CA1 at rest, as measured by c-Fos+ cell density [34, 35]. We report that lasting CA3 and CA1 hyperactivity can be reversed by acute systemic injection of haloperidol, xanomeline or levetiracetam (see Table S1). Moreover, both haloperidol and xanomeline, but not levetiracetam, provoked a reversal of the aberrant psychosis-relevant behavior, SRM, once delivered subacutely (po/sc).

Here, we were able to generate a reverse-translational preparation of HH induced by subchronic DG inhibition with the use of DREADDs in adolescent mice. Following 3 weeks of DG inhibition, we looked at the basal c-Fos protein expression pattern throughout the hippocampal subfields as a measure of gross neuronal activity. When investigated 1 week after the withdrawal of DREADD ligand (C21), c-Fos protein expression in DG does not differ from control animals which did not receive the ligand (see Study limitations). Nonetheless, CA3 and CA1 subfields show increased basal c-Fos expression (see Fig. 1A, E), implying persistent HH. Our recently published paper reveals that this hyperactivity presents itself in the form of hypersynchronous events, i.e. periodic bouts of high neuronal firing [35].

We hypothesized that a normalization of HH in the reverse-translational mouse preparation could be a predictor of antipsychotic efficacy of a drug, therefore we tested a selection of promising medications, checking the c-Fos expression pattern and quantifying their impact on behavior. To establish the influence of these drugs on c-Fos and SRM in normal animals, we have treated the control animals with haloperidol, xanomeline or levetiracetam. There was no effect of drug treatment on control animals in terms of c-Fos+ density, with the exception of xanomeline, which increased c-Fos in CA1 (Fig. 3e). A similar outcome was described previously in slice electrophysiology by Thorn [58] and was assumed to be mediated by activation of local M1 receptors, which leads to an increase in spontaneous neuronal firing in the area.

Simultaneously, DG-inhibited animals were treated with haloperidol, xanomeline or levetiracetam. In our experiments, at the time of testing the DG has “normalized” its activity, while CA3 and CA1 still show a pattern indicating hyperactivity. We were testing whether haloperidol would produce a robust reversal of CA3/CA1 hyperactivity. Here, haloperidol effectively reversed the HH to the control group levels.

Xanomeline’s antipsychotic effect could be mediated via M4 receptors within the striatum and M1 receptors in the cortex, affecting dopaminergic transmission, therefore acting indirectly upon the same neurotransmitter as haloperidol [89–93]. Based on this reasoning, we predicted that this novel antipsychotic would show an effect similar to haloperidol. In our study, xanomeline treatment reversed the HH in CA3, but it did not normalize CA1 activity. We consider this finding consistent with the aforementioned report of activation of CA1 neurons (Fig. S6) and perhaps emphasizing the complexity of hippocampal signaling driving behavioral symptoms. Levetiracetam does not show any antipsychotic activity in humans, apart from a pro-cognitive effect [94, 95]. We used this drug based on its HH-normalizing property in humans and hypothesized a similar effect, although possibly through a different mechanism of action [8, 72, 74, 96–101]. In our study, levetiracetam treatment reversed the HH in both CA3 and CA1, as observed with haloperidol.

All medications, despite their different mechanisms of action, seemingly normalized the CA3 and CA1 c-Fos protein expression back to levels equal to control animals, with the exception of xanomeline’s partial effect. Next, we administered selected drugs for 72 h to another cohort of animals and measured social recognition memory, as a learning and memory paradigm relevant for psychosis. The effect of drug treatment on behavior of control animals appeared different than their effect on c-Fos. There was a significant negative effect of haloperidol, which may be attributed to its influence on learning and memory [102–104] (Fig. 2B). Xanomeline or levetiracetam did not influence the behavior.

Concurrently, we treated experimental mice (DG inhibited in adolescence) with haloperidol, xanomeline or levetiracetam. We hypothesized that haloperidol treatment would rescue the aberrant behavioral phenotype, primarily due to its antipsychotic efficacy demonstrated in the clinic, and to a lesser degree its influence on glutamatergic transmission [51, 52, 105–110]. This indeed happened; the animals treated with haloperidol showed SRM index similar to control animals. With xanomeline, we also observed a robust normalization of the SRM index. It is known that the activation of M1 receptors rescues social interaction deficits in other models of psychosis and M4 receptors may also be involved [111, 112]. Both receptors are expressed throughout the brain, including the hippocampus, and from our c-Fos data as well as prior studies [58, 113], we know that xanomeline may act directly on the hippocampus. The effect on SRM can be compared to haloperidol, but perhaps the aforementioned difference in their influence on control animals could be meaningful in the context of side effects of both drugs. Lastly, we administered levetiracetam, but it did not normalize the SRM index. Levetiracetam is not antipsychotic per se and may induce clinical psychiatric symptoms [114], despite prior promising preclinical data; this commands careful dosage [72, 74, 80, 95, 97, 115, 116] and as such, could influence the result. We suspect three possible reasons for levetiracetam’s ineffectiveness on behavior, despite HH rescue. First, the trend in our data suggests that dosage or length of treatment could be adjusted to yield a larger effect. The effect on cognition is sensitive to dosage [96]. We used subacute treatment rather than chronic; some acute studies produced an effect on behavior [117], but others suggest that acute treatment may not be effective; even though our dose corresponded to human plasma levels [8, 74, 96, 97].

Second, the unique mechanism of action of levetiracetam [72], which includes Kv3.1 potassium channels (implicated in schizophrenia [118], important for gamma oscillations [118, 119], involved in social behaviors [120–123]) [99], may produce different effects in extrahippocampal areas, depending on the animal preparation used [100, 101, 117, 124]. Hence, the drug may be ineffective in our animal, which had subchronically inhibited DG granule cells before measurements, most likely producing lasting downstream changes, which are not fully researched.

Third, levetiracetam’s effect on DG (involved in pattern separation [125–130]) differs from haloperidol and xanomeline (see Figs. S4, 5), which could influence the behavior [33, 126]. The effect on DG was demonstrated in the clinic, but is not replicated here [8, 74] (see Study limitations). Although preclinical data suggests dose-dependent effects on cognition [67, 97, 100, 131, 132], the evidence from clinical research is mixed [8, 74, 133–135].

The differential drug effects on hippocampal subfields and behavior underscore the very different molecular mechanisms that each drug acts through. These results imply the role of extrahippocampal brain areas in the behavioral response and show the limitations of c-Fos as a proxy for activity. Consequently, our current results need to be interpreted considering the study limitations.

Study limitations

Our animal preparation may not perfectly model the human hippocampal pathology seen in schizophrenia. It is based on human imaging and post-mortem evidence of HH, thus along with associated behaviors, shows a good degree of face (biomarker and ethological) validity [136]. Our results indicate satisfactory predictive (remission) validity as well. A reliable construct validity is a challenge to schizophrenia researchers, mainly due to the elusive disease etiology, and this preparation is not an exception – it is based on quantifiable disease biomarkers rather than an underlying known pathophysiology, which remains unknown [137–140].

Furthermore, HH itself may be elicited via multiple mechanisms, such as interneuron loss as seen in MAM-E17, hAPP-J20, CD2-KO models [23, 25, 141]; GABA deficiency, glutamate dehydrogenase deficiency or oxytocin dysregulation [24, 26, 142]. This heterogenous etiology implies lack of specificity of the biomarker itself and thus the possibility of variable treatment response. While the hyperactivity readout via c-Fos+ cell density showed a reversal of HH, it may not constitute a sufficient proxy of neuronal activity, since c-Fos provides a snapshot of overall activity of the past few hours [82], and may mask different patterns of activity, e.g. hypersynchronous events [35], which still occur in DG weeks after C21 withdrawal. Thus, a more direct activity measurement technique (e.g. in vivo electrophysiology) may be required to capture more granular activity data.

Finally, our behavior focused on cognitive and learning aspects of psychotic disorders, as those are often undertreated [3, 4]. SRM is a highly reproduced phenotype in this animal, but we recognize the need to include more psychosis-relevant behaviors in future studies.

Future directions

Based on the translational validity of HH in mice, it could be worthwhile to study differentially activated brain regions downstream of the hippocampus; at baseline and in the context of chronic drug treatment, especially considering studies of HH induced by other mechanisms [11, 143]. Such research could bring us closer towards biomarker-based treatment decisions and precision psychiatry, enabling safer and more effective therapies.

In conclusion, our study demonstrates that (i) HH as measured by c-Fos protein expression is not a specific biomarker, and (ii) c-Fos protein expression readout lacks the temporal specificity to capture the full complexity of HH as a biomarker.

Author contributions

FD, DS and CT contributed to experimental design. FD designed and performed the surgeries and all experiments, acquired and analyzed the data. FD and CT wrote the manuscript. DS contributed to the internal review of the manuscript.

Funding

This work was supported by NIMH R01 grant (5R01MH123479).

Data availability

Data presented here is available upon reasonable request.

Competing interests

Currently, CAT is a founding member of Neuventis, Stockholm, Sweden; she held Karuna stock during the development of Cobenfy; is a member of DSBM at Merck; and is a part-time advisor of Kynexis. DSS and FPD declare no potential competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41386-025-02077-4.