Hydroxycarboxylic Acid Receptor 2 Mediates β‐hydroxybutyrate's Antiseizure Effect in Mice

Abstract

Objective

The ketogenic diet, a high‐fat, low‐carbohydrate regimen, is often used to treat drug‐resistant seizures and is being studied for Alzheimer's disease and other neuropsychiatric disorders. However, its mechanism of action remains unclear. β‐hydroxybutyrate, a primary circulating ketone body produced by the ketogenic diet, may mediate its effects on seizures by binding to a recently identified Gi‐coupled receptor: hydrocarboxylic acid receptor 2 (HCAR2).

Methods

RNAscope in situ hybridization assay and real‐time quantitative polymerase chain reaction were used to assess HCAR2 expression in the mouse brain. We generated HCAR2⁻/⁻ using the CRISPR‐Cas technique on an S129 mouse background. Whole‐cell current‐clamp was performed to measure the passive and active membrane properties of hippocampal dentate granule cells. The voltage‐clamp was performed to record synaptic currents. Two complementary in vivo mouse models—continuous hippocampal stimulation to induce status epilepticus (SE) and kindling—were used to induce seizures.

Results

HCAR2 was localized in dentate granule cells and microglia. In mice with HCAR2, β‐hydroxybutyrate reduced neuronal excitability by hyperpolarizing the resting membrane potential, raising the action potential threshold, and reducing the firing frequency of dentate granule cells. β‐hydroxybutyrate suppressed excitatory synaptic transmission. These effects were nullified in HCAR2⁻/⁻ mice. HCAR2⁻/⁻ mice showed no cognitive impairment. Moreover, β‐hydroxybutyrate did not affect seizures in HCAR2⁻/⁻ mice. However, it diminished both the duration and severity of seizures in HCAR2⁺/⁺ mice.

Interpretation

These findings demonstrate that HCAR2 mediates β‐hydroxybutyrate's antiseizure effects by regulating neuronal excitability and synaptic transmission. These studies propose a new mechanism for the antiseizure action of the ketogenic diet. ANN NEUROL 2026;99:809–824

A[Color figure can be viewed at www.annalsofneurology.org]

The ketogenic diet (KD)—a high‐fat, low‐carbohydrate, moderate‐protein regimen—has re‐emerged since the 1990s as a clinically effective treatment of medically refractory epilepsy. ¹ Clinical trials showed KD's effectiveness in children with drug‐resistant epilepsy (DRE). ² , ³ Preclinical studies also demonstrated the antiseizure action of KD across animal seizure models. ⁴ , ⁵ Moreover, KD may have beneficial effects in other neurological disorders, such as Alzheimer's disease (AD), Parkinson's disease, autism, and multiple sclerosis. ⁶ The mechanism of antiseizure action remains under active investigation, with studies suggesting a role for lactate dehydrogenase, ⁷ ATP‐sensitive potassium (KATP) channels, ⁸ and adenosine A1 receptors. ⁵

The liver generates ketone bodies when on KD, which include most commonly β‐hydroxybutyrate (β‐HB), acetoacetate, and acetone. Elevated plasma β‐HB levels correlate with lower seizure frequency. ⁹ Early animal model studies did not demonstrate an antiseizure action of β‐HB. However, more recent studies suggested efficacy in flurothyl ¹⁰ and N‐methyl‐D‐aspartate (NMDA)‐induced seizures, ¹¹ as well as in a chronic epilepsy model. ¹² β‐HB increased inhibitory synaptic transmission ¹³ and reduced excitatory synaptic transmission. ¹⁴ Additionally, β‐HB reduced neuronal firing by increasing the open probability of KATP channels. ⁸ It is unclear whether β‐HB directly mediates such diverse actions or if it acts via a second messenger system.

Hydroxycarboxylic acid receptor 2 (HCAR2, HM74a, or GPR109A) was cloned in 2003, primarily known as a specific receptor that mediates the anti‐lipolytic effects of nicotinic acid. ¹⁵ HCAR2 is a 7‐transmembrane Gi‐protein‐coupled receptor, along with HCAR1 (GPR81) and HCAR3 (GPR109B), belonging to the hydroxycarboxylic acid receptor (HCA) family of metabolite‐sensing receptors. ¹⁶ β‐HB and butyrate are 2 endogenous ligands for HCAR2. ¹⁶ HCAR2 is expressed not only on adipocytes and immune cells, ¹⁷ , ¹⁸ but also in the brain. ¹⁹ , ²⁰ Some studies suggested selective expression in microglia. ¹⁷ , ²⁰ HCAR2 is also expressed in the human brain. ¹⁹

We describe the distribution of HCAR2 in hippocampal neurons and glia. Next, we assessed whether β‐HB reduces neuronal excitability and excitatory synaptic transmission in the hippocampal circuit by HCAR2. We tested whether abrogation of HCAR2 abrogates these effects. We also tested the impact of HCAR2 deletion on locomotor behavioral function. Finally, we tested whether β‐HB rapidly suppresses seizures in mice expressing the HCAR2 receptor and whether the genetic absence of HCAR2 affects the antiseizure effects of β‐HB.

Methods

Animals

All experiments were conducted according to protocols approved by the University of Virginia (UVA) Institutional Animal Care and Use Committee (IACUC). Animals were housed in standard cages (groups of 5) with ad libitum access to water and food in a facility maintained under a 12‐hour light–dark cycle, with a controlled temperature of 21 ± 1.5°C and humidity levels of 50 ± 10%. Both male and female mice were used. The results from both sexes were similar, therefore, the data were pooled. This study used 4 to 8‐week‐old C57BL/6 mice (Charles River), HCAR2 knockout (KO) (HCAR2⁻/⁻) and wild‐type (WT) (HCAR2⁺/⁺). The investigators performing the experiments and analyzing data were blind to the genotype. Littermate WT and KO mice were used, wherever possible.

RNAscope In Situ Hybridization

To quantify HCAR2 mRNA expression, RNAscope in situ hybridization and immunohistochemistry were performed based on the protocol. ²¹ The fixation and permeabilization process was compatible with RNAscope and immunohistochemistry, which preserve RNA and protein structure. Brain tissue was embedded in Leica tissue‐freezing medium and sliced into 15μm sections using a cryostat microtome. The sections are mounted on the charged slides and run on the same slide to maintain consistent conditions and solutions during the experiment. Sections were rinsed with sterile water and incubated with “pretreat 4” from the RNAscope Multiplex Fluorescent Assay kit for 30 minutes at 40°C. The Advanced Cell Diagnostics (ACD) RNAscope Multiplex Fluorescent Detection Kit v2 (Cat. 323110, ACD, Newark, CA, USA) was used, and the tissue was treated according to the manufacturer's protocol. Sections were incubated with RNAscope Mm‐HCAR2 Probe (Gene Alias: Gpr109a, Cat. 451591, ACD, Newark, CA, USA), Negative Control Probe‐DapB (Cat. 310043, ACD, Newark, CA, USA), and Positive Control Probe Mm‐Polr2a (Cat. 312471, ACD, Newark, CA, USA). To perform immunohistochemistry staining with RNAscope, following the target retrieval steps, primary antibodies (Mouse anti‐Neun, Millipore Sigma, #MAB377 (Temecula, CA, USA), Rabbit anti‐Iba 1, Biocare Medical, #CP290A (Pacheco, CA, USA)) were applied to the sections and left overnight at 4°C. Goat anti‐rabbit IgG (H + L) cross‐adsorbed secondary antibody conjugated with Alexa Fluor 488 (ThermoFisher, Cat.A‐11008) was used for the rabbit primary antibodies. Because we used 2 fluorophores to visualize RNA and protein signals, we scanned each fluorophore separately to prevent bleed‐through. Imaging was conducted using the confocal microscope. Analysis of RNAscope was performed with Imaris (version) using the Spot Tool. A quadrangular region of interest (ROI) was established, encompassing the cell soma, and the count of individual puncta was performed within the cell body of microglia located in the hilus, subgranular zone, granular zone, and molecular layer of the dentate granule cells (DGCs). Microglia were identified by IbA 1, and neuron was confirmed with NeuN.

Quantitative Real‐Time Polymerase Chain Reaction

The HCAR2 mRNA level was measured using a real‐time (RT) polymerase chain reaction (PCR) assay with a standard protocol based on SYBR Green dye and the 2^−ΔΔCT method. ²² The expressions of β‐actin, GAPDH, and HPRT1 served as internal controls (all primers were used as mentioned in Table 1). The results were expressed as fold changes calculated using the 2^−ΔΔCT formula.

TABLE 1.

Primers Used in the qRT‐PCR Assay

Gene	Primer sequence
HCAR2	F: TCC AAG TCT CGA AAG GTG GT R: TGT TTC TCT CCA GCA CTG AGTT
β‐Actin	F: CAT TGC TGA CAG GAT GCA GAA GG R: TGC TGG AAG GTG GAC AGT GAG G
GAPDH	F: ACA GTC CAT GCC ATC ACT GCC R: GCC TGC TTC ACC ACC TTC TTG
HPRT	F: ACA GGC CAG ACT TTG TTG GA R: ACT TGC GCT CAT CTT AGG CT

HCAR2 = hydrocarboxylic acid receptor 2; qRT‐PCR = quantitative real‐time reverse transcription polymerase chain reaction.

Slice Electrophysiology

The brain slices containing the hippocampus were prepared as previously described. ²³ In brief, animals underwent deep anesthesia with isoflurane, followed by quick decapitation and dissection to isolate the brain into ice‐cold slicing buffer saturated with 95% O₂ and 5% CO₂. Subsequently, slices were placed into a holding chamber containing 34°C oxygenated artificial cerebrospinal fluid (aCSF) and then maintained at 22 to 25°C for recording (TC‐324C, Warner Instruments). Electrophysiological recordings were performed as previously described. ²³ Patch pipettes were filled with a potassium gluconate‐based internal solution containing (in mM): 120 potassium gluconate, 10 NaCl, 2 MgCl₂, 0.5 EGTA, 10 HEPES, 4 NaATP, and 0.3 NaGTP (pH 7.2, 280–295 mOsm). Briefly, brain slices were submerged in aCSF flowing at a rate of 2 to 3mL/minute. We applied hyperpolarizing and depolarizing step currents ranging from −90 to +190pA in 20pA increments for 300ms. Once the patch electrodes broke the cell membrane, the resting membrane potential (RMP) was measured in the current‐equal‐to‐0 mode. Membrane passive and active properties were calculated based on our previous paper published. ²³ The voltage‐clamp technique used to record spontaneous excitatory synaptic currents (sEPSC) was performed based on our lab's previous publication, ²⁴ in DGCs by holding a potential of −65mV. Recording pipettes displayed open‐tip resistances ranging from 6 to 8MΩ and were filled with the internal solution containing (in mM): 110 D‐gluconic acid, 110 CsOH, 10 CsCl, 1 EGTA, 1 CaCl2, 10 HEPES, 5 Mg‐ATP, and 5 lidocaine (pH 7.3, 290–300 mOsm). To block NMDA and GABAA receptors, 50μM D‐2‐amino‐5‐phosphonopentanoic acid (APV; Tocris) and 50μM picrotoxin (Sigma‐Aldrich) were added to aCSF. The access resistance was monitored continuously, and if the series resistance increased by 20% at any point, the recording was stopped. The amplitude and frequency of sEPSC were measured. DL‐β‐HB (3mM, Sigma‐Aldrich) was applied to slices.

Behavioral Tests

To assess baseline cognitive and motor function, mice underwent open‐field, object placement, and fear conditioning tests, as described in previous studies. ²⁵ Data were analyzed to determine the freezing time indicative of recall. ²⁵

Seizure Induction Protocols

Status epilepticus was induced using previously published methods. ²⁶ Briefly, mice (6–7 weeks old) were anaesthetized with isoflurane. Bipolar‐insulated stainless‐steel electrodes were stereotaxically implanted bilaterally into the ventral hippocampus, and a cerebellar reference electrode was placed posterior to lambda. After‐discharge threshold (ADT) was established using a 10‐second or 2‐second pulse train with an initial current of 20μA. The hippocampus was stimulated for 60 minutes. Status epilepticus (SE) was defined as continuous epileptiform discharges above 1Hz frequency lasting more than 5 minutes, after the end of stimulation. Mice were injected intraperitoneally with saline or DL‐β‐HB (1 g/kg, Sigma‐Aldrich) 15 minutes after the end of stimulation. The duration of SE and the behavioral seizure score (BSS) were monitored and evaluated. BSS was assessed using a modified Racine scale. ²⁶

In the kindling experiment, the ADT was measured under electrical stimulation consisting of 2‐second biphasic trains with a pulse width of 1ms at 50 Hz delivered. Stimulus was set at 1.5 times ADT at 60‐minute intervals, 4 times per day. Mice were considered fully kindled when they experienced 5 consecutive seizures with a behavioral score of at least 5. On the day of the experiment, ADT was remeasured. Sixty minutes later, mice received 1 baseline stimulation. Thirty minutes after the first stimulation, mice received an intraperitoneal injection of either β‐HB (1g/kg), niacin (300mg/kg, Sigma‐Aldrich). After an additional 30 minutes, mice received a second stimulation and 3 more stimulations at 60‐minute intervals to assess the effect of β‐HB. After‐discharge duration (ADD) and BSS were measured for each stimulation.

Statistical Analysis

Data in the graphs are presented as mean ± standard error of the mean (SEM) unless the figure specifies otherwise. Symbols denote individual data points. Figure legends provide crucial information, including numerical values for mean, SEM, biological or technical replicates, statistical tests, and corrections. Data were tested with normal distribution, and appropriate parametric and non‐parametric tests were used. We used 2‐way analysis of variance (ANOVA) followed by Tukey's post hoc multiple comparison tests for experimental designs involving more than 2 variables. Statistical differences between groups are represented in graphs as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. All data were included in the analysis. Data were collected and analyzed in a blinded manner, and the study was performed using GraphPad Prism 10.

Results

HCAR2 Expression and Colocalization with Neurons and Microglia

HCAR2 is widely expressed in the central nervous system. ¹⁹ , ²⁰ , ²⁷ HCAR2 protein is expressed in microglia located in the cortex, ¹⁷ rostral ventrolateral medulla, ²⁷ and subiculum. ²⁰ However, the cellular distribution of HCAR2 in the hippocampus remains poorly established. Hippocampus is a common site of seizure origin in patients with temporal lobe epilepsy (TLE), ²⁸ therefore, we visualized HCAR2 mRNA expression using in situ hybridization within the hippocampus. We further characterized the association between HCAR2 mRNA expression in neurons, microglia using immunohistochemistry.

HCAR2 mRNA (red puncta) was widely distributed throughout the hippocampus, with a strong signal in the DGCs and the pyramidal cell layers in CA1 and CA3 regions (Fig 1A‐J). Higher magnification images of the dentate gyrus region revealed distinct HCAR2 transcripts in NeuN‐positive neurons. Further magnification of this area provided detailed visualization of NeuN immunostaining, individual HCAR2 mRNA puncta, and their colocalization. Landreth's group found that HCAR2 modulates the microglial response in Alzheimer's disease and reported HCAR2 expression in microglia located in the subiculum. ²⁰ Next, we assessed HCAR2 expression in the microglia. In low magnification images, HCAR2 transcripts (red) were detected inside Iba1‐positive microglial cells (green) throughout the hippocampus. Enlargement of the DGC and further magnification revealed detailed labeling of Iba1, individual HCAR2 puncta, and the merged image indicating colocalization. A representative microglial cell, highlighted by a dashed circle, displayed an overlapping expression of Iba1 and HCAR2, indicating that HCAR2 is also expressed in a subset of microglia within the dentate gyrus. These results demonstrate that HCAR2 mRNA is expressed in both neurons and microglia in the hippocampal dentate gyrus, suggesting a potential role for HCAR2 in modulating functions in both cell types.

FIGURE 1.

Hydrocarboxylic acid receptor 2 (HCAR2) distribution and colocalization with neurons and microglia in the hippocampus. (A–E) Representative images illustrating HCAR2 mRNA expression (red puncta) detected by RNAscope and NeuN protein (green) identified through immunohistochemistry in the mouse hippocampus. (A) A low magnification image of the entire hippocampus showcasing widespread HCAR2 expression and NeuN labelling. (B) A higher magnification of the boxed area in A, emphasizing the dentate granule cell layer DG. (C, D) Further magnification of the boxed region in B reveals detailed labelling of (C) NeuN; (D), HCAR2 mRNA; and (E), the merged NeuN with HCAR2. Dashed circles highlight 2 neurons, each colocalizing with approximately 5 HCAR2 mRNA puncta. (F–J) Images depict HCAR2 mRNA (red) and the Iba1 protein marker for microglia (green) in the DG of the hippocampus. (F) A low magnification image demonstrating the overall expression of HCAR2 and Iba1 throughout the hippocampus. (G) A higher magnification of the boxed area in F, focusing on the DG containing microglia and HCAR2. (H–J) Further magnification of the boxed area in G. (H) Depicts Iba1 labelling microglia, (I) HCAR2 mRNA puncta, and (J) their colocalization. A dashed circle in J indicates a representative microglial cell exhibiting colocalization of HCAR2 mRNA with Iba1. Scale bars: 100μm in A and F; 10μm in B, C, G, and H. [Color figure can be viewed at www.annalsofneurology.org]

Generation and Characterization of HCAR2 KO

To investigate the role of HCAR2 in the β‐HB antiseizure effects, we conducted experiments using both HCAR2 KO (HCAR2⁻/⁻) and WT (HCAR2⁺/⁺) mice. To this end, we generated HCAR2⁻/⁻ mice via CRISPR‐Cas9 genome editing techniques with the S129 mouse genetic background (see Materials and Methods). This method uses Cas9 nuclease along with single‐guide RNAs to create a precise double‐strand break at the HCAR2 gene location (Fig 2B). The deletion of 503 nucleotides within a single exon of HCAR2 resulted in a significant reduction in HCAR2 mRNA expression (see Fig 2C,D and Supporing data 1). PCR genotyping differentiated HCAR2⁻/⁻ (205bp) from HCAR2⁺/⁺ (708bp) alleles, confirming the successful deletion (see Fig 2E). Quantitative RT‐PCR analyses confirmed the HCAR2 mRNA expression in the brain tissue samples from 6 to 8‐weeks‐old HCAR2⁺/⁺ mice. The gene expression levels of HCAR2 transcripts exhibited a significant reduction in HCAR2 mRNA expression in the brains of HCAR2 KO mice compared to WT mice (fold change in gene expression, HCAR2⁺/⁺, 0.9577 ± 0.1115; HCAR₂ ⁻/⁻, 0.1901 ± 0.05605) (see Fig 2F).

FIGURE 2.

Generation and characterization of hydrocarboxylic acid receptor 2 (HCAR2)‐deficient mice. (A) Schematic diagram showing gene‐targeting to generate knock‐out and wild‐type HCAR2 mice. (B) The exon, 5′ upstream region, and 3′ region of the HCAR2 gene are depicted. (C) The on‐target guide RNA binding sites and the corresponding guide RNA sequences are shown. The red nucleotides indicate the guide RNA sequence. The genotyping polymerase chain reaction (PCR) strategy used for line 1 is outlined. (D) The sequence of a PCR fragment shows the deletion site marked by the star. Note that the reverse strand was sequenced. This line exhibited a 503bp deletion. (E) The gel illustrates the amplification products from the genotyping PCR. (F) Real‐time quantitative PCR analysis of HCAR2 mRNA expression in the brain tissue of 4–6‐week‐old HCAR2⁺/⁺ and HCAR2⁻/⁻ mice (***p < 0.001, paired t test, n = 7 per genotype). [Color figure can be viewed at www.annalsofneurology.org]

HCAR2 Mediates the Effects of β‐HB on Neuronal Excitability

An aberrant increase in neuronal firing and excitatory synaptic transmission occurs during seizures. ²⁹ , ³⁰ To evaluate whether HCAR2 mediates the effect of β‐HB (3mM) on neuronal excitability and synaptic transmission, we conducted whole‐cell patch‐clamp recording in DGCs from HCAR2⁺/⁺ and HCAR2⁻/⁻ mice. In humans, plasma β‐HB levels are 2 to 5mM following a KD or exogenous ketone supplementation, therefore, we applied 3mM β‐HB to slices. ³¹ Preclinical studies use this range to replicate the therapeutic ketosis observed in patients. ⁸ , ³² We focused on hippocampal DGCs, which serve as gatekeepers for cortical input to the hippocampus and whose breakdown precipitates seizures. ³³ We evaluated the effect of β‐HB (3mM) on the resting membrane properties and excitability of DGCs before and after using current‐clamp recordings in acute hippocampal slices obtained from HCAR2⁺/⁺ and HCAR2⁻/⁻ mice.

Applying β‐HB to hippocampal slices did not affect neuronal excitability in HCAR2⁻/⁻ mice, but reduced it in slices from HCAR2⁺/⁺ mice. β‐HB hyperpolarized the RMP of DGCs from HCAR2⁺/⁺ mice (aCSF: −60.99 ± 1.885mV; β‐HB: −64.04 ± 1.865mV), but produced no change in HCAR2⁻/⁻ mice (aCSF: −62.44 ± 1.533mV; β‐HB: −62.26 ± 1.585mV) (Fig 3B). Input resistance remained unchanged after the application of β‐HB in both HCAR2⁺/⁺ and HCAR2⁻/⁻ mice (Fig 3C). The DGCs' firing frequency in response to step current injections showed no change in the current‐frequency (IF‐curve)—the relationship between a neuron's firing and the input current—after β‐HB application in HCAR2⁻/⁻ mice. In contrast, β‐HB caused a downward shift in the current intensity versus firing frequency curve in HCAR2⁺/⁺ mice. Analysis of action potential (AP) features indicated that β‐HB did not alter AP properties in HCAR2⁻/⁻ mice, whereas it did in HCAR2⁺/⁺ mice. β‐HB did not change the AP threshold in HCAR2⁻/⁻ mice (aCSF: −37.42 ± 2.236mV; β‐HB: −37.93 ± 2.332mV), but it increased the AP threshold in HCAR2^+/+ mice (aCSF: −38.85 ± 2.338mV; β‐HB: −30.45 ± 3.134mV). The AP shape‐related parameters—AP amplitude and half‐width—remained unchanged in both HCAR2⁺/⁺ and HCAR2⁻/⁻ mice after β‐HB application. There were no changes in depolarization step‐induced spike count after β‐HB in HCAR2⁻/⁻ mice (aCSF: 65.70 ± 8.588; β‐HB: 67.10 ± 8.786), whereas it was reduced in HCAR2⁺/⁺ mice (aCSF: 68.80 ± 8.094; β‐HB: 58.22 ± 6.080) (Fig 3J).

FIGURE 3.

β‐hydroxybutyrate (β‐HB) modulates dentate granule cells (DGCs)' excitability through hydrocarboxylic acid receptor 2 (HCAR2). (A) Schematic representing the preparation of brain slices and the DGCs' whole‐cell patch clamp recording. (B) β‐HB (3mM) perfusion for 10 minutes did not change the resting membrane potential (RMP) of DGCs in HCAR2⁻/⁻ mice (p = 0.883, paired t test, n = 15 cells), whereas it hyperpolarized the RMP of DGCs from HCAR2⁺/⁺ mice (*p < 0.05, paired t test, n = 12 cells). (C) R_in did not change after applying β‐HB in HCAR2⁻/⁻ mice (p = 0.205, paired t test, n = 14 cells) and HCAR2⁺/⁺ mice (p = 0.3844, paired t test, n = 12 cells). (D) Representative traces of AP firing in DGCs under control conditions (artificial cerebrospinal fluid [aCSF]) and after 10 minutes of β‐HB perfusion, recorded in current‐clamp mode. β‐HB did not significantly alter the DGCs' firing frequency in HCAR2⁻/⁻ mice (p = 0.6835, 2‐way analysis of variance [ANOVA], Tukey's multiple comparison test, n = 12 cells). (E) Representative traces of AP firing in DGCs from HCAR2⁺/⁺ mice, recorded in aCSF and after 10 minutes of β‐HB perfusion. In HCAR2⁺/⁺ mice, β‐HB significantly reduced the firing frequency (*P < 0.05, 2‐way ANOVA, Tukey's multiple comparison test, n = 12 cells). (F) The action potential diagram represents AP features that were measured under current‐clamp recording. (G) AP threshold in HCAR2⁻/⁻ mice remained unchanged (p = 0.747, paired t test, n = 15 cells), whereas it significantly increased after β‐HB application (**p < 0.01, paired t test, n = 12 cells). (H) AP shape‐related remained unchanged, such as AP amplitude in both HCAR2⁻/⁻ mice (p = 0.309, paired t test, n = 15 cells) and HCAR2⁺/⁺ mice (p = 0.091, paired t test, n = 12 cells). (I) AP half‐width showed no significant differences after β‐HB perfusion in both HCAR2⁻/⁻ mice (p = 0.377, paired t test, n = 12 cells) and (p = 0.48, paired t test, n = 12 cells) remained unchanged. (J) The total number of APs remained unchanged after β‐HB application in HCAR2⁻/⁻ mice (p = 0.534, paired t test, n = 12 cells), whereas it was significantly reduced following β‐HB perfusion in HCAR2⁺/⁺ mice (***p < 0.001, paired t test, n = 12 cells). [Color figure can be viewed at www.annalsofneurology.org]

β‐HB Reduces sEPSC in DGCs through HCAR2

β‐HB and acetoacetate are structurally similar, and their conversion is catalyzed by β‐HB hydrogenase, an enzyme located on the inner mitochondrial membrane. ³⁴ Acetoacetate reduced the quantal glutamate release, which was measured through miniature excitatory postsynaptic current (mEPSC) in cultured CA1 pyramidal neurons. ¹⁴ Conversely, another study showed that 10mM acetoacetate has shown no effect on sEPSC amplitude under normal conditions, whereas it inhibited sEPSC in epileptiform activity, which was mediated by voltage‐dependent Ca²⁺ channels (VDCCs). ³⁵ β‐HB has shown little effect on sEPSC amplitude. ³⁵ To examine whether β‐HB influences excitatory synaptic transmission, we recorded sEPSC from DGCs using voltage‐clamp techniques. The slices were perfused with oxygenated aCSF containing D‐2‐amino‐5‐phosphonopentanoic acid (DL‐AP5, 50μM) and picrotoxin (50μM) to block the NMDA and GABA‐A receptors, respectively. ²²

We found that β‐HB reduced the frequency of sEPSC, but not amplitude in hippocampal DGCs from C57BL/6 mice in preliminary studies. To assess HCAR2's role in mediating β‐HB's effect on sEPSC, we used the voltage‐clamp technique to record sEPSC from DGCs in HCAR2⁻/⁻ and HCAR2⁺/⁺ mice (Fig 4A, B). The application of β‐HB (3mM) caused no changes in sEPSC frequency or amplitude in HCAR2⁻/⁻ mice (Fig 4C). In contrast, β‐HB reduced sEPSC frequency in HCAR2⁺/⁺ mice under the same condition without altering the amplitude, indicating a presynaptic mechanism of action (P < 0.05, Kolmogorov–Smirnov test, n = 8 cells) (Fig 4D). Together, these findings demonstrate that β‐HB's effects on both excitability and excitatory transmission in DGCs require HCAR2. This provides compelling evidence that HCAR2 mediates the acute neuronal effects of β‐HB in reducing neuronal network activity, a mechanism relevant to seizure suppression.

FIGURE 4.

β‐hydroxybutyrate (β‐HB) modulates dentate granule cells (DGCs)' excitatory synaptic transmission through hydrocarboxylic acid receptor 2 (HCAR2). (A) The schematic presents the spontaneous excitatory synaptic currents (sEPSC) recording of hippocampal DGCs using whole‐cell patch clamp recording. (B) Representative trace of sEPSC recorded from DGCs from HCAR2⁻/⁻ and HCAR2⁺/⁺ mice. β‐HB (3mM) perfusion for 10 minutes did not change the properties of sEPSC of DGCs in HCAR2⁻/⁻ mice, whereas it reduced the frequency of events in HCAR2. (C) Cumulative distribution analysis of sEPSC events demonstrated no significant change in frequency (n = 8 cells, p = 0.26, Kolmogorov–Smirnov test) following β‐HB application. (D) The cumulative distribution of sEPSC events indicated sEPSC frequency was significantly reduced (n = 8 cells, *p < 0.05, Kolmogorov–Smirnov test). Data are presented as mean ± standard error of the mean; *p < 0.05 was considered statistically significant. [Color figure can be viewed at www.annalsofneurology.org]

HCAR2 Gene Deletion Has No Impact on Cognitive and Locomotor Activity Functions

After establishing the colony, we used 3 behavioral tests: open field, object placement preference, and fear conditioning. The open field test showed no differences in locomotor activity, thigmotactic time, and index between HCAR2⁺/⁺ and HCAR2⁻/⁻ mice (Fig 5A–D). We then tested the animals on memory tasks. The object placement test, which included mobile object preference, total investigation time, and discrimination index, revealed no differences between the 2 genotypes (Fig 5E–H). Moreover, results from the fear conditioning test, which measured the percentage of time spent freezing in response to 5‐foot shocks during the conditioning day and the subsequent 2 days of context and cued recall at 24 and 48 hours after initial conditioning, indicated no differences between HCAR2⁺/⁺ and HCAR2⁻/⁻ mice (Fig 5I–K). These findings were consistent across all tests and suggest that HCAR2 gene deletion does not impact baseline cognition, anxiety‐related behavior, or movement.

FIGURE 5.

Hydrocarboxylic acid receptor 2 (HCAR2) gene knockout does not affect the behavioral tests. (A) Representative cartoon, a heat map illustrating the time spent in different zones of the open field arena during the testing phase, conducted 8 hours after the familiarization period for HCAR2⁺/⁺ and HCAR2⁻/⁻. Representative movement traces depict exploratory behavior in both genotypes in the open field arena. (B–D) Quantification of thigmotaxis time (row factor p = 0.0001, column factor p = 0.5021, row factor × column factor p = 0.0377, 2‐way analysis of variance [ANOVA]), thigmotaxis index (row factor p = <0.0001, column factor p = 0.5520, row factor × column factor p = 0.3846, 2‐way ANOVA), and total locomotor activity (row factor p = <0.0001, column factor p = 0.7562, row factor × column factor p = 0.8135, 2‐way ANOVA, HCAR2⁺/⁺ n = 11 and HCAR2⁻/⁻ n = 10), demonstrating no significant differences between HCAR2⁺/⁺ and HCAR2⁻/⁻ mice. (E) Cartoon, heat map, and movement traces illustrate the time spent on the mobile object placement task during the testing phase, which was performed 1 hour after familiarization. (F–H) Quantification of mobile object preference (p = 0.4312, unpaired t test), total investigation time (p = 0.8071, unpaired t test), and discrimination index (p = 0.4657, unpaired t test, n = 9 each), in HCAR2⁺/⁺ and HCAR2⁻/⁻ mice. (I–K) Cued and contextual fear conditioning assessments. (I) Percentage of freezing time in response to 5‐foot shocks during conditioning (row factor p = <0.0001, column factor p = 0.3078, row factor × column factor p = 0.4139, 2‐way ANOVA). (J) Freezing behavior during context recall was assessed 24 hours after conditioning (p = 0.1950, unpaired t test). (K) Freezing behavior during cued recall, performed 48 hours after initial conditioning (row factor p = <0.0001, column factor p = 0.8747, row factor × column factor p = 0.4446, 2‐way ANOVA, HCAR2⁺/⁺, n = 11 and HCAR2⁻/⁻, n = 10). Data are presented as mean ± standard error of the mean. [Color figure can be viewed at www.annalsofneurology.org]

HCAR2 Is Required for the Antiseizure Effects of β‐HB and Niacin in the Hippocampal Kindling

To further explore the role of HCAR2 in the antiseizure effects of β‐HB, we performed in vivo experiments in HCAR2⁺/⁺ and HCAR2⁻/⁻ mice. We aim to assess the acute antiseizure effect of β‐HB on seizures induced by the kindling paradigm, which mimics TLE by gradual sensitization to seizures. ³⁶ TLE is the most common type of focal onset epilepsy. First, we examined the HCAR2's role in susceptibility to hippocampal kindling. One week after hippocampal electrode implantation, 4 stimulation trains were administered daily at 1‐hour intervals in HCAR2⁺/⁺ and HCAR2⁻/⁻ mice until they reached the fully kindled stage of 5 seizures (Fig 6A, B). Notably, HCAR2⁻/⁻ mice reached the fully kindled state more quickly than HCAR2⁺/⁺ mice, and they required fewer stimulations to achieve grade 5 seizures (HCAR2⁻/⁻: 15.87; HCAR2⁺/⁺: 22.07) (see Fig 6C).

FIGURE 6.

Hydrocarboxylic acid receptor 2 (HCAR2) is essential for the antiseizure effect of β‐hydroxybutyrate (β‐HB) and niacin in a hippocampal kindling paradigm. (A) An HCAR2⁺/⁺ and HCAR2⁻/⁻ mouse underwent daily hippocampal kindling stimulation until they were fully kindled (5 consecutive stage 5 seizures). On the test day, β‐HB or niacin was administered intraperitoneally 30 minutes after the first stimulation train, which is shown with the green arrows. (B) Representative electroencephalography (EEG) trace and corresponding spectrogram depict the kindling stimulus train (50Hz, 1ms biphasic square wave pulses, delivered at 1.5× the after‐discharge threshold [ADT] for 2 seconds). (C) Behavioral seizure score (BSS) across the kindling acquisition demonstrates that HCAR2⁻/⁻ mice reached full kindling approximately 50% faster than HCAR2⁺/⁺, using a nonlinear fit. (D) On the test day, BSS recorded during 5 successive stimulation trials showed that β‐HB treatment in HCAR2⁺/⁺ mice reduced seizure severity from stage 5 to stage 4 during the third and fourth stimulations, with a return to stage 5 during the fifth stimulation (*p < 0.05, Mann–Whitney test, HCAR2⁺/⁺, n = 7; HCAR2⁻/⁻, n = 7). However, the β‐HB injection did not affect the HCAR2⁻/⁻ seizure severity. (E) β‐HB reduced the after‐discharge duration (ADD) measured from the EEG during the third and fourth stimulations in HCAR2⁺/⁺ mice, whereas no effect was observed in HCAR2⁻/⁻ mice (*p < 0.05, 2‐way repeated measures analysis of variance [ANOVA] followed by a post hoc Tukey's, HCAR2⁺/⁺, n = 7; HCAR2⁻/⁻, n = 7). (F) Administration of niacin (300mg/kg, i.p.) in HCAR2⁺/⁺ mice also decreased BSS, reducing it from stage 5 to stage 4 during the fourth stimulation, followed by a return to stage 5 during the fifth stimulation, however, it had no effect in HCAR2⁻/⁻ mice (*p < 0.05, Mann–Whitney test, HCAR2⁺/⁺, n = 4; HCAR2⁻/⁻, n = 6). (G) Niacin also reduced ADD during the third stimulation and returned to stage 5 during the fifth stimulus, thereby confirming a transient antiseizure effect mediated by HCAR2 activation, with no effect observed in HCAR2⁻/⁻ mice (*p < 0.05, 2‐way repeated measures ANOVA followed by a post hoc Tukey's, HCAR2⁺/⁺, n = 4; HCAR2⁻/⁻, n = 6). All data are presented as mean ± standard error of the mean; p < 0.05 was considered statistically significant. [Color figure can be viewed at www.annalsofneurology.org]

Five stimulation trains were administered at 1‐hour intervals on the testing day to investigate whether β‐HB suppresses seizures via HCAR2 in this chronic model, and β‐HB (1g/kg, i.p.) was given 30 minutes after the first train. Exogenous administration of β‐HB was tested in different doses, ranging from low to high doses (0.3–12g/kg), demonstrating its stability and effectiveness. ³⁷ , ³⁸ It should be noted that even at the high doses of 12g/kg of β‐HB, there was no adverse effect or acute toxicity. ³⁷ Initially, in the pilot study, we tested different doses of 0.3, 1, and 3g/kg of β‐HB on seizure parameters. Based on our pilot results, the dose of 1g/kg was efficient on seizure parameters such as duration and severity, so we selected this dose for the subsequent in vivo experiments. In HCAR2⁺/⁺ mice, β‐HB transiently reduced seizure severity, as indicated by a notable reduction in the BSS in HCAR2⁺/⁺ mice following the third and fourth stimuli (from stage 5 to stage 4), with a subsequent return to stage 5 behavioral seizure in response to the fifth stimulus train. This attenuation in BSS was associated with a shorter ADD in response to the third and fourth stimuli in HCAR2⁺/⁺ mice. In contrast, β‐HB did not change either BSS or ADD in HCAR2⁻/⁻ mice (see Fig 6D,E).

Next, to verify that HCAR2 mediates β‐HB's effect, we used a pharmacological probe, niacin, as a potent orthosteric agonist of HCAR2, ³⁹ and it has been used as a treatment for dyslipidemia. ¹⁵ Consistent with a receptor‐mediated mechanism, niacin (300mg/kg, i.p.) temporarily decreased both seizure severity and ADD in HCAR2⁺/⁺ mice (Fig 6F,G), mimicking the effect of β‐HB. In contrast, in HCAR2⁻/⁻ mice, niacin did not affect seizure severity or ADD, indicating that the antiseizure effects of both compounds depend on HCAR2 signaling.

HCAR2 Mediates β‐HB's Antiseizure Effects on the SE

Clinical evidence demonstrated the effectiveness of KD in treating individuals with super‐refractory SE when conventional treatments fail. ⁴⁰ , ⁴¹ We tested whether a single systemic dose of β‐HB (1g/kg) rapidly suppresses SE in mice. We first evaluated the duration and severity of SE induced by the CHS paradigm in HCAR2⁺/⁺ and HCAR2⁻/⁻ mice SE duration was similar in HCAR2⁺/⁺ and HCAR2⁻/⁻ mice (HCAR2⁺/⁺: 152.1 ± 22.42 minutes; HCAR2⁻/⁻: 132.4 ± 29.96 minutes) (Fig 7B), and the behavioral seizure intensity showed no differences between HCAR2⁺/⁺ and HCAR2⁻/⁻ mice (see Fig 7C).

FIGURE 7.

The antiseizure effects of β‐hydroxybutyrate (β‐HB) were attenuated in animals lacking hydrocarboxylic acid receptor 2 (HCAR2). (A) The schematic overview of the experimental paradigm: bilateral stimulating and recording electrodes were implanted in the ventral CA1 hippocampi, along with a reference electrode placed in the cerebellum. Seizures were induced using CHS, and their monitoring was conducted via video‐electroencephalography (EEG). (B) In response to CHS, mice lacking HCAR2 (HCAR2⁻/⁻) and littermate wild‐type controls (HCAR2⁺/⁺) had similar duration status epilepticus (SE) (respectively, p = 0.92, Kaplan–Meier survival analysis, Mantel‐Cox log‐rank test, HCAR2⁻/⁻ n = 11 and HCAR2⁺/⁺ n = 9). (C) Behavioral seizure scores (BSS) were comparable between HCAR2⁻/⁻ and HCAR2⁺/⁺ mice, indicating no genotype‐dependent differences in seizure severity. (D) In HCAR2⁻/⁻ mice, β‐HB administration (n = 14) did not significantly alter SE duration or BSS compared to saline treatment (p = 0.1447, Kaplan–Meier survival analysis, Mantel‐Cox log‐rank test, HCAR2⁻/⁻–β‐HB treated n = 14, HCAR2⁻/⁻‐saline treated n = 8). (E) BSS analysis indicated no difference after β‐HB administration (p = 0.1805, Mann–Whitney test, HCAR2⁻/⁻–β‐HB treated n = 9, HCAR2⁻/⁻‐saline treated n = 6). (F) Heat maps depicting seizure severity in randomly selected HCAR₂ ⁻/⁻ mice following β‐HB (n = 5) or saline (n = 6) injection. (G) Cortical EEG power spectrograms showing SE duration and frequency in HCAR2⁻/⁻ mice treated with β‐HB or saline. Green arrows indicate the time of injection (15 minutes post‐CHS), while white arrows denote the end of SE. (H) β‐HB treatment significantly reduced SE duration in HCAR2⁺/⁺ mice compared to saline controls (*p < 0.05, Kaplan–Meier survival analysis, Mantel‐Cox log‐rank test, HCAR2⁺/⁺–β‐HB treated n = 13, HCAR2⁺/⁺‐saline treated n = 5). (I) Analysis of the BSS demonstrated significantly reduced seizure severity following β‐HB administration compared with the saline group (**p < 0.01, Mann–Whitney test, HCAR2⁺/⁺–β‐HB treated n = 8, HCAR2⁺/⁺‐saline treated, n = 6). (J) Heat maps illustrate seizure severity in randomly selected HCAR2⁺/⁺ mice treated with β‐HB (n = 6) or saline (n = 5). (K) Cortical EEG power spectrograms during SE in a saline‐treated (left) and β‐HB‐treated (right) HCAR2⁺/⁺ mouse. Green arrows indicate saline or β‐HB administration (15 minutes post‐CHS), whereas white arrows denote the SE termination. [Color figure can be viewed at www.annalsofneurology.org]

Next, we determine whether HCAR2 mediates β‐HB's antiseizure effects. In HCAR2⁻/⁻ mice, β‐HB (1g/kg, i.p.) did not shorten SE or reduce its severity compared to saline‐treated animals (see Fig 7D–F). The electroencephalography (EEG) power spectrogram analysis did not reveal any differences between HCAR2⁻/⁻ mice treated with β‐HB and those treated with saline (see Fig 7G). These findings were contrasted with HCAR2⁺/⁺ mice. β‐HB (1g/kg, i.p.) decreased the SE duration by 67.38 ± 30.99 minutes compared to saline‐treated HCAR2⁺/⁺ mice (see Fig 7H). β‐HB attenuated seizure severity to grades 3 to 4, compared to saline‐treated mice that had grades 4 to 5. A heatmap representative of BSS illustrated that β‐HB reduced seizure severity compared to saline‐treated mice. Furthermore, EEG power spectrogram analysis revealed that β‐HB administration decreased power. These results demonstrated that HCAR2 mediates β‐HB's effect on SE.

Discussion

KD has long been recognized as an effective treatment for individuals with epilepsy who have not responded to conventional antiseizure medications. Numerous clinical and preclinical studies have demonstrated the efficacy of KD. ⁴² , ⁴³ The current study demonstrated that β‐HB, a key product of the KD, suppresses seizures via neuronal HCAR2. HCAR2 is expressed in hippocampal neurons and microglia. β‐HB modulates DGC's excitability and excitatory synaptic transmission by acting on HCAR2.

Our study suggests that HCAR2 agonists, such as β‐HB and niacin, can achieve at least some of the therapeutic benefits of the KD. Despite its established efficacy, adherence to use KD among PWE remains low because of its intolerability and undesirable side effects, such as gastrointestinal discomfort, dyslipidemia, and micronutrient imbalances, and there are no antiseizure drugs that mimic these dietary treatments. Several reviews have emphasized the urgent need to clarify the cellular targets and signaling pathways modulated by ketone bodies and other metabolic intermediates of the KD. ⁴⁴ Consistent with our findings, a recent study showed that an HCAR2 agonist, an endogenous KD metabolite, or GSK256073 (60mg/oral) suppresses seizures in the mouse 6Hz and rat PTZ animal models of epilepsy and reduces epileptiform discharge in both rat and human brains. ⁴⁵

Nicotinic acid can suppress seizures. Several HCAR2 agonists are available, including acipimox, acifran, monomethyl fumarate (MMF), and MK‐6892, whose anti‐seizure properties require testing. We used niacin as a potent and orthosteric agonist of HCAR2 to confirm its activation‐mediated antiseizure effects. However, niacin has multiple effects independent of HCAR2 activation. Niacin is a precursor for nicotinamide adenine dinucleotide (NAD⁺) and has direct metabolic and cellular actions. Niacin‐induced SIRT1 activation, a nuclear enzyme called a sirtuin, results in an anti‐inflammatory response, reduces tumor necrosis factor (TNF‐α), and mitigates oxidative stress. ⁴⁶ Niacin inhibits diacylglycerol acyltransferase 2 (DGAT2), lowering hepatic triglyceride synthesis and very‐low‐density lipoprotein secretion, and increases high‐density lipoprotein cholesterol by reducing apoA‐I catabolism. ⁴⁷ Additionally, it protects endothelial cells via antioxidant and anti‐inflammatory pathways and can suppress dendritic cell migration independently of HCAR2. ⁴⁸ Given that niacin has many effects beyond HCAR2, using HCAR2 KO mice confirmed that HCAR2 specifically mediates niacin's antiseizure effects.

KD effectively treated individuals with super‐refractory SE when conventional treatments failed. ⁴⁰ , ⁴¹ Two clinical observations raise questions about ketone bodies as mediators of the antiseizure effects of KD. One clinical observation is that in some studies, blood ketone (ie, β‐HB) levels do not correlate well with seizure control. ⁴⁹ , ⁵⁰ Second, the low glycemic index treatment (LGIT) does not induce systemic ketosis and also suppresses seizures. ⁵¹ Our studies show that β‐HB can shorten SE duration, suggesting a mechanism by which KD can shorten SE duration. We also found that β‐HB suppressed kindled seizures along with nicotinic acid. A previous study also reported suppression of kindled seizures by KD, however, the effect was transient. We did not assess the effect of repeated administration of β‐HB.

There are additional antiseizure mechanisms of β‐HB. It can rapidly suppress seizures through its metabolic actions and modulation of ion channels. It also offers long‐term benefits by inducing neurotrophic factors epigenetically and reducing inflammation. ⁵² β‐HB alters neuronal bioenergetics and protects mitochondria. ⁵³ Neurons rely on glycolysis to produce Acetyl‐CoA, which then enters the tricarboxylic acid cycle. β‐HB provides an alternative source of Acetyl‐CoA, generating more ATP and fewer reactive oxygen species. ⁵³ By lowering reactive oxygen species, β‐HB shields neurons from oxidative damage. ⁵³ Exercise increases β‐HB levels and specifically targets HDAC2 and HDAC3, which influence brain‐derived neurotrophic factor promoters. ⁵⁴ β‐HB also increases BDNF expression under normal glucose conditions and enhances the production of ciliary neurotrophic factor. ⁵⁵ Kim et al ¹² reported that β‐HB restored impaired hippocampal long‐term potentiation and spatial learning and memory deficits in Kcna1‐null mice. It raises the threshold for calcium‐induced mitochondrial permeability transition in Kcna1‐null animals. ¹²

Other studies in transfected Xenopus oocytes showed that micromolar concentrations of β‐HB activate the KCNQ 2/3‐encoded M‐Channel. ⁵⁶ Additional Xenopus oocyte studies report that β‐HB inhibits the effects of agonists for GABA‐A, glycine, and NMDA receptors at concentrations achieved in vivo. ⁵⁷ In Drosophila, β‐HB modulates KATP channels and GABA‐B receptors. ⁵⁸ These findings contrast with an early study in cortical cultures and hippocampal slices that found no effect on the primary voltage‐ and ligand‐gated ion channels mediating excitatory or inhibitory neurotransmission in the hippocampus. ⁵⁹

β‐HB and niacin dampen the inflammatory response of microglia, which are the brain's resident immune cells. ²⁰ , ⁶⁰ Moreover, clinical and experimental evidence showed that β‐HB has an immunomodulatory role by suppressing the NLR‐family pyrin domain‐containing 3 (NLRP3) inflammasome activation, proinflammatory cytokines. ⁶¹ , ⁶² The anti‐inflammatory effect of β‐HB depends on HCAR2, which is expressed on macrophages. ⁶³ , ⁶⁴ This is associated with elevated interleukin [IL]‐10 anti‐inflammatory cytokine levels and an enhanced anti‐inflammatory M2‐like phenotype of macrophages. ⁶³ In the brain, β‐HB's anti‐inflammatory action is mediated by HCAR2 through immune‐cell infiltration, macrophage anti‐inflammatory phenotype, and induces the production of prostaglandin D2 via cyclooxygenase‐1. ¹⁷

A long‐standing question in the studies of the KD mechanism is to identify molecular targets of ketones that can modify neuronal excitability. Previous studies have suggested a role for lactate dehydrogenase, ⁷ KATP channels, ⁸ and adenosine A1 receptors, ⁵ in mediating anti‐seizure actions of KD. Ketones like β‐HB are proposed to reduce neuronal excitability by opening KATP channels, but these channels open in response to low intracellular ATP. The KD raises intracellular ATP, which should shut down KATP channels. However, KATP channels may be activated by lowered cAMP levels. EPAC, a cAMP‐sensing protein, reduces KATP channel openings in pancreatic islet cells. ⁶⁵ Neurons express Epac, therefore, HCAR2 activation, which is a Gi‐coupled receptor and lowers cAMP levels, and may reduce Epac activity in neurons, which increase KATP channel openings. Alternatively, β‐HB‐HCAR2 signaling may leverage these intracellular signaling cascades to activate GIRK channels, allowing potassium ions to flow out of the neurons, resulting in cell membrane hyperpolarization. Future work aimed at elucidating these pathways could pave the way for more targeted and tolerable interventions, not only for epilepsy, in which hyperexcitability and circuit dysfunction converge.

β‐HB reduced sEPSC frequency on DGCs through HCAR2 receptors, but it had no impact on EPSCs' amplitude, suggesting a presynaptic action. Previous studies indicated that ketone bodies like acetoacetate inhibit vesicular glutamate transporters and N‐type Ca²⁺ channel currents. ¹⁴ , ⁶⁶ Additionally, decanoic acid, a medium‐chain saturated fatty acid produced by KD, suppressed epileptiform activity in hippocampal slices of rats by directly inhibiting AMPA receptor‐mediated current. ⁶⁷

β‐HB reduced DGCs' excitability, strengthening the “dentate gate”, a mechanism to suppress seizures. Dentate gyrus is characterized by a combination of 2 specific features, including DGCs' intrinsic features (eg, hyperpolarizing RMP, low input resistance, and higher threshold), and dense GABAergic transmission (feedforward and feedback inhibitory circuits). ³³ By hyperpolarizing the RMP and elevating the AP threshold, β‐HB dampened DGC responsiveness to depolarizing inputs. At the synaptic level, β‐HB reduced the frequency of sEPSC, suggesting a presynaptic modulatory role.

This study has limitations. Many clinical studies demonstrating the efficacy of KD are conducted in patients with generalized seizures, which involve the neocortex, and fewer with focal seizures. ⁶⁸ , ⁶⁹ Some recent studies suggest that HCAR2 is expressed in the mouse and human neocortex. ²⁰ , ⁷⁰ HCAR2's role in suppressing generalized neocortical seizures needs further investigation. Another limitation of our study is that we used the racemic mixture DL‐β‐HB. Because only the D‐enantiomer is biologically active, the effective dose of the active compound was approximately half of the nominal dose administered.

In conclusion, this study uncovers an uncharacterized cellular mechanism by which KD exerts its neuroprotective effects, through β‐HB acting on HCAR2 to modulate hippocampal neuronal activity. β‐HB‐HCAR2 signaling stabilizes DGCs' excitability and circuit dynamics by dampening hippocampal network activity, which is crucial for brain disorders characterized by hippocampal hyperexcitability, such as epilepsy and Alzheimer's disease. Our work addresses a long‐standing gap by revealing a direct receptor‐mediated pathway through which β‐HB modulates neuronal function. This potentially explains the broader clinical efficacy of the KD in various brain disorders.

Author Contributions

J.K., S.N., S.J., and J.W. contributed to the conception and design of the study. S.N., J.W., H.S., S.J., R.S., S.Z., S.S., C.S., and AB contributed to the acquisition and analysis of data. S.N., J.W., S.J., R.S., C.S., S.S., I.Z., and J.K. contributed to drafting the text or preparing the figures. [Correction added on 25 February 2026, after first online publication: Author contribution text has been revised in this version.]

Potential Conflicts of Interest

Nothing to report.

Supporting information

Data S1 Supporting Information.

Data Availability

The data that support the findings of this study are available from the corresponding author on reasonable request.