Voltage-Dependent Block of N-Methyl-d-Aspartate Receptors by the Novel Anticonvulsant Dibenzylamine, a Bioactive Constituent of l-(+)-β-Hydroxybutyrate


Address correspondence and reprint requests to Dr. J. M. Rho at Department of Pediatrics, ZC 4482, UC Irvine Medical Center, Bldg. 2, 3rd Floor, 101 The City Drive S, Orange, CA 92868, U.S.A. E-mail: jmrho@uci.edu


Summary: Purpose: Previously we demonstrated that l-(+)-β-hydroxybutyrate (L-BHB), acetoacetate (ACA), acetone, and dibenzylamine (DBA) were anticonvulsant in an audiogenic seizure–susceptible model, and that DBA was a bioactive contaminant identified in commercial lots of L-BHB. In the present study, we asked whether these effects could be mediated by ionotropic glutamate or γ-aminobutyric acidA (GABAA) receptors.

Methods: We studied the effects of both stereoisomers of BHB (as well as the racemate), ACA, and DBA on N-methyl-d-aspartate (NMDA), α-amino-3-hydroxy-5methyl-4-isoxazole-proprionic acid (AMPA), and GABAA receptors in cultured rodent neocortical neurons by using whole-cell voltage-clamp recording techniques.

Results: Only L-BHB and DBA exerted a concentration- and voltage-dependent block of NMDA-evoked currents, whereas none of the tested substrates affected AMPA- or GABA-activated currents. The kinetics of whole-cell block by L-BHB and DBA were similar, providing additional evidence that DBA is responsible for the anticonvulsant activity of L-BHB.

Conclusions: BHB and ACA do not exert direct actions on GABAA or ionotropic glutamate receptors in cultured neocortical neurons. In addition, we provide additional evidence that DBA is responsible for the anticonvulsant activity of L-BHB, and that this action may be mediated in part by voltage-dependent blockade of NMDA receptors.

During a prolonged (i.e., >48 h) fast, hepatic fatty acid oxidation and ketone body production provide alternative energy sources, especially for the brain. Ketogenesis, resulting primarily in the formation of β-hydroxybutyrate (BHB; 3-hydroxybutyrate or 3-hydroxybutanoic acid), and acetoacetate (ACA), and to a minor degree acetone, also is observed during diabetic ketoacidosis and a high-fat, low-carbohydrate diet known as the ketogenic diet (KD). The KD is an effective nonpharmacologic treatment for patients with intractable epilepsy (1,2). Despite >80 years of clinical experience with the KD, the mechanisms underlying its anticonvulsant actions remain poorly understood.

Recently attention has again focused on the possibility that one or more ketone bodies may modulate neuronal excitability. Although it has been observed that ACA and acetone possess anticonvulsant properties in vivo (3–5), the mechanisms underlying these effects are unknown; whether BHB, the principal ketone body, exerts similar effects is less clear. The major unresolved question is whether ketone bodies can directly affect the excitability of neuronal membranes through actions on ion channels that are the targets of most anticonvulsant medications (AEDs). Thus far, BHB and ACA do not appear to affect synaptic activity directly, at least not in the hippocampus (6).

Previously we demonstrated that l-(+)-BHB exhibits anticonvulsant activity in audiogenic seizure–susceptible mice, but that this was likely due to dibenzylamine (DBA), a chemical contaminant contained within commercial lots of this stereoisomer (4). Further, we provided evidence that ACA and acetone also were anticonvulsant in this model. To determine whether this in vivo activity could be correlated with actions on postsynaptic ligand-gated ion channels, we studied the effects of d-(−)-BHB, l-(+)-BHB, DL-BHB, ACA, and DBA on N-methyl-d-aspartate (NMDA), α-amino-3-hydroxy-5-methyl-4-isoxazoleproprionic acid (AMPA), and γ-aminobutyric acidA (GABAA) receptors in cultured mouse and rat neocortical neurons by using whole-cell voltage-clamp recording techniques. We found that L-(+)-BHB exerted a concentration- and voltage-dependent block of NMDA-evoked currents, and provided additional evidence that the in vivo anticonvulsant activity of this isomer is due to DBA.


Electrophysiologic studies

Neocortical cells were cultured from either Swiss Webster mouse fetuses (15-day gestational age) or Sprague–Dawley rat fetuses (18 days old). Animal procedures were approved by the Institutional Animal Care and Use Committee at the University of Utah. All electrophysiologic recordings were conducted, 1–2 weeks after cell plating, on the stage of an inverted phase-contrast microscope at room temperature (23–25°C) according to previously described techniques (7). Before each recording session, cells were rinsed completely with a buffer solution containing in (mM): NaCl, 142; KCl, 1.5; CaCl2, 0.1; HEPES, 10; glucose, 10; and sucrose, 20 (320 mOsm; pH 7.4). The bathing solution also contained 500 nM tetrodotoxin (to block voltage-activated Na+ channels) and 1 μM strychnine (to block glycine-activated Cl currents). In all recordings of NMDA-activated currents, 1 μM glycine was present as the obligate coagonist. Voltage-clamp recordings were made with an Axopatch 200A amplifier (Axon Instruments, Burlingame, CA, U.S.A.) and digitized for off-line analysis. Currents were filtered at 1–2 KHz, digitally sampled at 1 KHz, and acquired on an IBM-compatible microcomputer by using Axotape or pClamp7 software (Axon Instruments) as well as on a chart recorder.

Patch pipettes (2–4 MΩ) were prepared from filament-containing thin-wall glass capillary tubes (1.5-mm outer diameter; World Precision Instruments, Sarasota, FL, U.S.A.) with a four-stage horizontal pipette puller (model P-97 Flaming Brown; Sutter Instruments, Novato, CA, U.S.A.). Pipettes were backfilled with a solution containing (in mM): CsCl, 153; EGTA, 10; HEPES, 10; MgCl2, 4 (290 mOsm; pH 7.4). Recordings were carried out at a holding potential of −60 mV, unless otherwise noted.

Drugs were dissolved in buffer on the day of use. Solutions containing test compounds were applied by using a rapid perfusion system that consisted of a gravity-fed multibarreled microperfusion pipette that was positioned 200–400 μm from the cell being studied (7; or Warner Instruments, Hamden, CT, U.S.A.) under pCLAMP (Axon Instruments) control.

Data analysis

Concentration–response curves were generated by using a nonlinear least-squares program (NFIT; Island Products, Galveston, TX, U.S.A.). Percentage control of whole-cell currents was calculated according to the formula


where B is the percentage block, Io is the steady-state control current evoked by agonist (NMDA), and ID is the steady-state current evoked by agonist in the presence of the test compound. Data points for concentration–response curves were fit to the logistic equation


where n indicates nH, a parameter indicating the steepness of fit, B is the percentage block, C is the concentration of the test compound, and IC50 is the concentration resulting in Bmax/2 block.


DBA was purchased through Aldrich (Milwaukee, WI, U.S.A.). All other chemicals were obtained from Sigma Chemical Co. (St. Louis, MO, U.S.A.).


The concentration range chosen to test ketone bodies was based on previous clinical (8,9) and experimental data (10) indicating that serum levels of BHB can be found in the low millimolar range in subjects treated successfully with a KD. Although the true brain concentrations of BHB and ACA in patients on a KD remain unknown, evidence suggests that acetone may be present in the brain at concentrations approaching 1 mM (11).

β-Hydroxybutyrate and acetoacetate

In the current study, neither stereoisomer of BHB (≤30 mM) nor ACA (≤10 mM) affected whole-cell currents activated by 3 μM GABA (n = 6; data not shown) or 100 μM kainate (n = 6; data not shown). However, at a supratherapeutic concentration (30 mM), L-BHB produced a 23% block of currents evoked by 100 μM kainate (n = 6; data not shown). Similarly, but in a more clinically relevant (9) concentration range (300 μM to 10 mM), we found that only the l-isomer produced a rapid and reversible concentration-dependent block of currents activated by 10 μM NMDA (in the presence of 1 μM glycine; Fig. 1). The median effective concentration (IC50) for the block by l-isomer was 6.9 mM, and the Hill coefficient was 0.87. The d-isomer of BHB was without effect on NMDA-activated currents at concentrations ≤30 mM.

Figure 1.

l-(+)-β-Hydroxybutyrate (L-BHB) and dibenzylamine (DBA) block of N-methyl-d-aspartate (NMDA)-activated currents. A: Representative whole-cell current traces demonstrating inhibition by L-BHB (1, 3, and 10 mM) of currents activated by 10 μM NMDA and 1 μM glycine (left), but no effect of 10 mM D-BHB (right). B: Whole-cell current traces showing that DBA (30 and 300 μM) also blocks NMDA-activated currents. C: Concentration–response curve for inhibition of NMDA-evoked currents by D-BHB (open triangle), L-BHB (solid circle) and DBA (solid triangle). Each point represents the mean ± SEM of data from three to eight cells; errors bars, when not visible, are smaller than the size of the symbols.

We next explored the possibility that L-BHB might interact directly with either the NMDA- or glycine-binding sites. The degree of whole-cell block produced by 10 mM L-BHB was similar at low (1–3 μM) or high (100 μM) concentrations of either of the two coagonists, NMDA or glycine (data not shown), indicating that the block could not be overcome by increasing the concentration of either coagonist. This result suggested that L-BHB antagonism of NMDA receptors was not due to competitive binding at either the glycine or NMDA recognition sites. Further experiments were conducted to determine whether the block produced by L-BHB was voltage-dependent. We found that 10 mM L-BHB blocked NMDA-activated currents by 6.0 ± 1.4% (n = 6) and 44 ± 1.4% (n = 8) at holding potentials of +60 mV and −60 mV, respectively.


Interpretation of the voltage-dependence data for L-BHB, however, became problematic. Such findings would not be expected for this compound, a monocarboxylic acid that is negatively charged at physiologic pH. Additionally, given the low affinity of this compound against NMDA receptors (IC50, 6.9 mM), we would have expected faster kinetics of whole-cell block, something we did not observe. More important, when the dl-racemate (which contains an equal mixture of both d- and l-isomers) of BHB was used in a similar concentration range, blockade of NMDA-evoked currents could not be confirmed. These data were consistent with a substance other than L-BHB (i.e., DBA) being responsible for our previous observations.

As expected, we found that DBA blocked NMDA-evoked currents in a concentration-dependent manner (Fig. 1B). The IC50 of block by DBA was 92.1 μM and the Hill coefficient was 0.84, the latter being nearly identical to that calculated from the concentration–response curve with L-BHB (i.e., 0.87). Given the estimated 0.4% concentration of DBA in the L-BHB (4), the revised IC50 for NMDA-receptor block by L-BHB was ∼28 μM, roughly comparable to that obtained with DBA alone.

Further evidence that the block of NMDA-evoked currents initially seen with L-BHB was due to DBA was provided by kinetic data. The on-rates of whole-cell block by 10 mM L-BHB and 100 μM DBA were determined with the best single-exponential fits to current traces. The time constants (τ) for the on-rates of L-BHB and DBA block were 1,040 ± 120 ms (n = 6) and 1,030 ± 122 ms (n = 4), respectively. The voltage dependence of DBA block also was confirmed. Figure 2 demonstrates that the degree of whole-cell block induced by DBA was greater at negative potentials than at positive ones; NMDA-activated currents in the presence of DBA were 19 ± 2.9% (n = 4) and 85 ± 7% (n = 4) of control at −80 mV and +40 mV, respectively. This voltage dependence was similar to that observed previously for L-BHB.

Figure 2.

Voltage dependence of dibenzylamine (DBA) block of N-methyl-d-aspartate (NMDA)-activated currents. Left: Representative current traces demonstrating 300 μM DBA inhibition of whole-cell currents evoked by 10 μM NMDA and 1 μM glycine at holding potentials +40 mV and −80 mV. Right: Relative block expressed as fraction of control at various holding potentials. Each point represents the mean ± SEM of data from four cells.


By using whole-cell voltage-clamp recording techniques and a rapid drug-perfusion system, we tested the actions of both stereoisomers and racemate of BHB, ACA, and DBA on NMDA, AMPA, and GABAA receptors in rodent cultured neocortical neurons. We found that both L-BHB and DBA (a contaminant in commercial lots of L-BHB) blocked NMDA-activated currents in a concentration- and voltage-dependent manner. In extending our earlier observations (4), we herein provide additional evidence that the in vivo anticonvulsant activity of L-BHB is due to DBA alone.

Ketone bodies

Given the strong (but not universal) correlation between blood ketone levels and seizure control, it is important to determine whether ketones can directly modulate neuronal excitability and/or synchronization. We previously demonstrated that BHB, the major ketone moiety, is not directly anticonvulsant, at least in an audiogenic seizure–susceptible model (4). Furthermore, we found that, in cultured neocortical neurons, neither BHB nor ACA directly interacts with either GABAA or ionotropic glutamate receptors, the principal molecular targets of many AEDs (12–14).

Our data are consistent with the study by Thio et al. (6), who applied standard cellular electrophysiologic techniques to evaluate the direct effects of ketone bodies in hippocampal synaptic transmission. In their hands, short-term application of BHB and ACA did not affect (a) excitatory postsynaptic potentials (EPSPs) and population spikes in CA1 pyramidal neurons after Schaffer collateral stimulation; (b) spontaneous epileptiform activity in the hippocampal–entorhinal cortex slice seizure model; and (c) whole-cell currents evoked by glutamate, kainate, and GABA in cultured hippocampal neurons.

The major ketone BHB is structurally related to GABA; thus it has been speculated that BHB (and possibly ACA) may exert direct modulatory effects on GABA receptors. The evidence thus far fails to support this notion. This is not surprising, given that BHB does not contain important amine moieties critical for GABA-receptor binding (15). Nevertheless, Niesen et al. (16) presented preliminary data supporting the direct anticonvulsant actions of BHB and ACA. In their hands, BHB (150 mg/kg), administered in the short term, was effective in blocking pentylenetetrazol (PTZ)-induced seizures in adolescent male Wistar rats (16). In acute and cultured hippocampal slices, 1–3 mM BHB or ACA decreased the amplitude and number of multiple CA1 population spikes induced by four different proconvulsant conditions: 8 mM extracellular KCl, 100 μM 4-aminopyridine, 50 μM bicuculline, and a “rapid kindling” paradigm (17). Finally, in intracellular recordings of immature CA1 neurons from cultured rat hippocampal slices, BHB (0.03–3 mM) potentiated evoked early inhibitory postsynaptic potentials (IPSPs), suggesting an action on postsynaptic GABAA receptors (18).

In the current study, we found that neither isomer of BHB (through a concentration range of 30 μM to 30 mM) nor ACA changed the amplitude of whole-cell currents evoked by 3 μM GABA, consistent with the observations of Thio et al. (6). It is possible that BHB may preferentially modulate specific molecular isoforms of GABAA receptors found in CA1 hippocampal neurons but not in neocortex, but this is highly unlikely because pyramidal cells in both neocortex and hippocampus exhibit a similar spectrum of GABAA-receptor subtypes (19,20). In the studies reported by Niesen's group, it is unclear which stereoisomers of BHB were used. It is possible that the effects of BHB seen by these investigators may be due to block of presynaptic K+channels by DBA, as suggested by Doepner et al. (21,22), resulting in increased presynaptic release of neurotransmitter (i.e., GABA). This possibility, however, has yet to be confirmed. Doepner et al. (23) recently demonstrated that DBA blocked voltage-dependent K+ currents in mouse cardiac myocytes, specifically inhibiting the slow component of the recovery from inactivation.

For the current investigations, we chose not to study the effects of acetone on GABAA and ionotropic glutamate receptors because its extreme volatility renders its use in electrophysiologic experiments challenging at best. This is despite reports of solvents such as dimethylsulfoxide (DMSO) affecting GABA- and glutamate-activated currents in cultured neurons (24,25). Although we cannot exclude the possibility of direct effects of acetone on membrane-bound ion channels, no data support this hypothesis. The anticonvulsant efficacy of acetone has been recently confirmed in animal models (4,5), but its underlying mechanisms have yet to be established.


DBA is a benzaldehyde derivative whose clinical import stems from its use as a pharmacologic vehicle for certain antibiotics (26) and as a biologically active contaminant identified in commercial preparations of l-(+)-β-hydroxybutyrate (4,22). Although it is a novel compound with promising anticonvulsant properties, it is labeled as a toxic, potentially carcinogenic substance in the Environmental Protection Agency (EPA) inventory under the Toxic Substances Control Act (TSCA), and thus is a poor candidate for further preclinical and clinical development. Nevertheless, our finding that DBA is a biologically active contaminant in L-BHB is not the first (nor likely the last) instance of an unexpected compound possessing anticonvulsant properties. It is well known that the anticonvulsant activity of valproic acid (VPA) was serendipitously discovered after it was used as a vehicle to dissolve investigational compounds (27).

In summary, we have shown that DBA, a contaminant in L-BHB, is a voltage-dependent blocker of NMDA receptors. Consistent with what has been previously reported in hippocampus (6), we also demonstrated that BHB and ACA do not directly modulate GABAA or ionotropic glutamate receptors in cultured neocortical neurons. In extending our earlier work (4), we provided additional evidence that the in vivo anticonvulsant activity of L-BHB is due to DBA alone. The role of ketone bodies as anticonvulsant effectors and/or mediators of the KD remain to be clarified. Given the prominent role of ketone bodies in intermediary metabolism, there are likely other novel mechanisms through which they may exert an anticonvulsant, and potentially neuroprotective, effect (28–30).


Acknowledgment:  This work was supported by NIH grant K08 NS01974 (J.M.R.) and the Anticonvulsant Drug Development Program, The University of Utah (H.S.W. and S.D.D.).