The Glycolytic Metabolite, Fructose-1,6-bisphosphate, Blocks Epileptiform Bursts by Attenuating Voltage-Activated Calcium Currents in Hippocampal Slices

Manipulation of metabolic pathways (e.g., ketogenic diet (KD), glycolytic inhibition) alters neural excitability and represents a novel strategy for treatment of drug-refractory seizures. We have previously shown that inhibition of glycolysis suppresses epileptiform activity in hippocampal slices. In the present study, we aimed to examine the role of a “branching” metabolic pathway stemming off glycolysis (i.e., the pentose-phosphate pathway, PPP) in regulating seizure activity, by using a potent PPP stimulator and glycolytic intermediate, fructose-1,6-bisphosphate (F1,6BP). Employing electrophysiological approaches, we investigated the action of F1,6BP on epileptiform population bursts, intrinsic neuronal firing, glutamatergic and GABAergic synaptic transmission and voltage-activated calcium currents (ICa) in the CA3 area of hippocampal slices. Bath application of F1,6BP (2.5–5 mM) blocked epileptiform population bursts induced in Mg2+-free medium containing 4-aminopyridine, in ~2/3 of the slices. The blockade occurred relatively rapidly (~4 min), suggesting an extracellular mechanism. However, F1,6BP did not block spontaneous intrinsic firing of the CA3 neurons (when synaptic transmission was eliminated with DNQX, AP-5 and SR95531), nor did it significantly reduce AMPA or NMDA receptor-mediated excitatory postsynaptic currents (EPSCAMPA and EPSCNMDA). In contrast, F1,6BP caused moderate reduction (~50%) in GABAA receptor-mediated current, suggesting it affects excitatory and inhibitory synapses differently. Finally and unexpectedly, F1,6BP consistently attenuated ICa by ~40% without altering channel activation or inactivation kinetics, which may explain its anticonvulsant action, at least in this in vitro seizure model. Consistent with these results, epileptiform population bursts in CA3 were readily blocked by the nonspecific Ca2+ channel blocker, CdCl2 (20 μM), suggesting that these bursts are calcium dependent. Altogether, these data demonstrate that the glycolytic metabolite, F1,6BP, blocks epileptiform activity via a previously unrecognized extracellular effect on ICa, which provides new insight into the metabolic control of neural excitability.


INTRODUCTION
Accumulating evidence suggests that neuronal activity is tightly coupled to energy metabolism (Magistretti and Pellerin, 1996;Ames, 2000), which is not surprising given that the brain consumes ∼20% of the body's energy. Metabolic manipulation provides a unique strategy to regulate excessive neuronal activity (such as seizures), which is particularly important given that at least 1/3 of patients with epilepsy are resistant to drug treatment. One relevant example of seizure control through metabolism is the low-carbohydrate, adequate-protein, high-fat ketogenic diet (KD), which has proven to be highly effective in controlling many types of drug-refractory seizures (Freeman and Vining, 1998;Stafstrom and Rho, 2012;Thakur et al., 2014). More recently, a new approach has been to alter energy metabolism by inhibiting glycolysis with a non-metabolizable glucose analog, 2-deoxy-glucose (2-DG). Partial inhibition of glycolysis with 2-DG blocks seizure activity in brain slices and in several animal models of seizures and epileptogenesis (Garriga-Canut et al., 2006;Stafstrom et al., 2009;Forte et al., 2016;Shao and Stafstrom, 2017).
The pentose phosphate pathway (PPP) is a secondary metabolic pathway that comprises a ''branch'' off glycolysis. The PPP is stimulated by the glycolytic intermediate, fructose-1,6-bisphosphate (F1,6BP), diverting glycolysis into the PPP, which produces glutathione (GSH), an antioxidant and endogenous anticonvulsant (Abe et al., 2000; Figure 1). Therefore, stimulation of the PPP has the potential to control excessive neuronal activity and seizures. Interestingly, a previous study reported that administration (i.p.) of the PPP stimulator F1,6BP is associated with anticonvulsant effects in animal models (Lian et al., 2007), supporting the hypothesis that PPP may play an important role in regulating seizures. Yet, how PPP alters neural excitability and seizure activity is largely elusive.
In this study, we further examined the role of the PPP in regulating neuronal excitability and seizure activity, by stimulating the PPP with F1,6BP in hippocampal slices. We found that while F1,6BP consistently blocks epileptiform bursts in hippocampal slices, it does not significantly reduce intrinsic neuronal firing or excitatory synaptic transmission. Unexpectedly, F1,6BP directly blocks voltage-activated calcium current (I Ca ), an important contributor to epileptiform bursts. Thus, F1,6BP exerts its anti-epileptiform action partly through a previously unrecognized acute extracellular action to attenuate Ca 2+ influx (Figure 1), at least in this in vitro seizure model. This novel mechanism of an intracellular metabolite on an extracellular target provides new insight into metabolic control of neural excitability and seizures.

Ethics Approval Statement
All procedures used in this study were approved by the Institutional Animal Care and Use Committee of Johns Hopkins University.
FIGURE 1 | The roles of fructose-1,6-bisphosphate (F1,6BP) in glycolysis and pentose phosphate pathway (PPP) and its proposed extracellular action. Diagram showing the main steps in glycolysis, its relationship to the PPP and the roles of F1,6BP. Glucose enters cytoplasm via glucose transporters (GLUT3 in neurons or GLUT1 in glia), where glycolysis takes place. Glucose is first phosphorylated to glucose-6-P (G6P), is further converted to F1,6BP, and is ultimately catalyzed to pyruvate, which then enters mitochondria to participate in the TCA cycle. G6P also enters the PPP, which generates reduced nicotinamide adenine dinucleotide phosphate (NADPH) and glutathione (GSH) to prevent the cell damage caused by reactive oxygen species (ROS). Accumulation of F1,6BP weakly inhibits F6P conversion to F1,6BP, and potently shunts G6P into PPP, leading to the production of GSH. In addition, exogenous F1,6BP may interfere with voltage-gated calcium channels (VGCC) and reduce Ca 2+ influx, based on evidence in this study.

Brain Slice Preparation
Hippocampal slices were prepared from Sprague-Dawley rats (Envigo International, Indianapolis, IN, USA) aged 8-16 days, when they are most susceptible to epileptiform bursting activity in slice electrophysiology experiments (Swann and Brady, 1984). Rat pups were deeply anesthetized by isoflurane inhalation and decapitated. Their brains were quickly removed and placed in pre-chilled and oxygenated low Ca 2+ /high Mg 2+ cutting solution containing (in mM): 125 NaCl, 3 KCl, 1.25 NaH 2 PO 4 , 25 NaHCO 3 , 0.25 CaCl 2 , 10 MgSO 4 and 11 glucose, for ∼1-2 min. One hemisphere was glued onto the platform of a specimen syringe and embedded with 1.6% low-melting point agarose (type I-B), and placed into a buffer tank filled with cutting solution. Coronal hippocampal slices (350 µm) were cut using a VF-300 compresstome (Precisionary Instruments Inc., Greenville, NC, USA), and transferred to a storage chamber filled with a holding solution (Feldmeyer et al., 2006) containing (in mM): 125 NaCl, 3 KCl, 1.25 NaH 2 PO 4 , 25 NaHCO 3 , 0.5 CaCl 2 , 5 MgSO 4 and 11 glucose, and continuously bubbled with 95% O 2 and 5% CO 2 . Slices were allowed to recover for ∼2 h, first at 34 • C for ∼45 min and thereafter at room temperature.

Electrophysiology
For recording, hippocampal slices were transferred to a submerged chamber and perfused with oxygenated artificial cerebrospinal fluid (aCSF) containing (in mM): 125 NaCl, 3 KCl, 1.25 NaH 2 PO 4 , 25 NaHCO 3 , 1.3 CaCl 2 , 1.3 MgSO 4 and 11 glucose, at 2 ml/min. For experiments eliciting epileptiform bursts and NMDA currents (see below), Mg 2+ was omitted from aCSF and Ca 2+ was elevated to 2 mM. All recordings were conducted at 32-33 • C through a digital temperature control system (Scientifica, East Sussex, UK). To visualize and approach individual neurons, a PatchPro 6000 patch-clamp recording system (Scientifica Instruments, East Sussex, UK) was used. This system is comprised of a custom designed motorized microscope with infrared differential interference contrast (IR-DIC) components, light-emitting diode (LED) IR and fluorescence light sources, 40× water immersion objective and CCD camera, and two motorized micromanipulators sitting on top of a motorized movable base plate. The entire visualized patch-clamp procedure was monitored on a computer screen and operated remotely through a control panel and LinLab software (Scientifica Instruments, East Sussex, UK). Electrophysiological signals were acquired by a MultiClamp 700B amplifier, a Digidata-1550 digitizer, and Clampex 10.4 software (Molecular Devices, Sunnyvale, CA, USA). Single or dual whole-cell recordings, extracellular field-potential recording, or simultaneous field potential-whole cell recordings were conducted in hippocampal CA3. In some field-potential experiments, two slices were recorded simultaneously. Synaptic stimulation was delivered by a concentric bipolar electrode (FHC, Bowdoin, ME, USA) placed in the CA3 stratum radiatum 100-200 µm lateral to the recording electrodes, through a digital stimulus generator (STG 4002, Multichannel Systems, Reutlingen, Germany).
Population epileptiform bursts in CA3 were induced by Mg 2+ -free medium containing 50 µM 4-aminopyridine (4-AP), and recorded extracellularly in non-clamp mode (I = 0) with a gain of 100-200. Intrinsic firing of CA3 neurons was recorded in the whole-cell current-clamp mode and in the presence of AMPA-, NMDA-and GABA A -receptor antagonists DNQX (10 µM), AP-5 (50 µM) and SR95531 (10 µM), to eliminate synaptic influences. Glutamatergic synaptic currents were evoked by electrical stimulation and recorded in whole-cell voltageclamp mode holding at −65 mV. Stimuli were delivered at 0.05 Hz at an intensity of 150-300 µA, 100 µs, which was determined as 2-3 times threshold stimulation, defined as the minimal current stimulation (∼50-75 µA) to evoke an EPSC. AMPAR-mediated excitatory postsynaptic current (EPSC AMPA ) was recorded in the presence of SR95531 (10 µM) and AP-5 (50 µM), while NMDAR-mediated excitatory postsynaptic current (EPSC NMDA ) was evoked in Mg 2+ -free medium and in the presence of SR95531 (10 µM) and DNQX (10 µM). GABA A R-mediated current was evoked by local stimulation in the CA3 pyramidal cell layer to directly activated GABAergic neurons, and recorded in the presence of DNQX (10 µM) and AP-5 (50 µM) and held at 0 mV. Voltage-activated calcium current (I Ca ) was recorded in whole-cell voltage-clamp mode in the presence of SR95531 (10 µM), DNQX (10 µM) and AP-5 (50 µM). A ramp voltage command (−70 to +20 mV, 200-500 ms) and a series of stepwise voltage commands (−70 to +20 mV, 200-500 ms, step: 10-mV) were used to characterize voltage-dependent activation kinetics. For steadystate inactivation kinetics, the membrane potential was first clamped at different levels (−70 to +20 mV) for 1 s (i.e., prepulses) and then stepped to −10 mV (200 ms, test pulses). High voltage-activated I Ca was evoked by a single voltage step from −70 mV to −10 mV (200-500 ms), repeated at 0.1 Hz. The series resistance (Rs) was compensated by 55%-75% and I Ca was leak subtracted online using a P4 subtraction paradigm in Clampex 10 (Molecular Devices, Sunnyvale, CA, USA). In a subset of experiments, neurons were loaded with F1,6BP (5 mM) intracellularly via recording pipettes by adding F1,6BP to the internal solution. All signals were acquired at 10 kHz and low-pass filtered at 2 kHz. For analysis, field-potential population bursts were high-pass filtered at 1 Hz offline.

Data Analysis
Qualitative and quantitative data analyses were conducted using Clampfit 10.4 (Molecular Devices, Sunnyvale, CA, USA), SigmaPlot 11 (SPSS Inc., Chicago, IL, USA), and Excel (Microsoft Corp., WA, USA) programs. The mean amplitude of epileptiform population bursts, EPSC AMPA and EPSC NMDA , GABA A current and I Ca and the mean frequency of neuronal firing were calculated from a 5-min recording before and after application of F1,6BP, and averaged across the recorded cells/slices in their respective groups. One-way analyses of variance (ANOVA) was used for statistical analysis across multiple groups, followed by the Holm-Sidak test for comparison between each two groups. Student's t-test was used for comparisons between two groups. Data are presented as means ± SEM with statistical significance set at p < 0.05.

Pharmacological Agents and Chemicals
All chemicals used in this study were purchased from Sigma-Aldrich (St. Louis, MO, USA), except DNQX (di-sodium salt), which was obtained from Abcam (Cambridge, MA, USA).

F1,6BP Rapidly Blocks CA3 Epileptiform Population Bursts
F1,6BP has previously been shown to exhibit a prominent anticonvulsant effect in several animal models of acute seizures (Lian et al., 2007). However, no study to date has studied F1,6BP effects on epileptiform activity in vitro. We first tested the hypothesis that F1,6BP blocks epileptiform activity in the CA3 area of hippocampal slices. Epileptiform population bursts were consistently induced in 0 Mg 2+ aCSF with 50 µM 4-AP. Similar to our previous study (Shao and Stafstrom, 2017), these bursts were typically ''interictal-like'' discharges consisting of 1-4 short bursts, lasting for 100-500 ms and occurring at variable frequencies (Figures 2A,B). Bath application of F1,6BP (2.5-5 mM) markedly suppressed epileptiform bursts in majority (13 of 19, 68%) of the slices (Figures 2A-D). This dosage of F1,6BP is similar to previous in vitro studies (3.5 mM, Vexler et al., 2003); 10 mM, (Izumi et al., 2003) and comparable to (but slightly lower than) in vivo studies (0.25-1 g/kg, (Lian et al., 2007); 1 g/kg, (Rogido et al., 2003)). More specifically, F1,6BP caused two types suppression: partial and complete. In partial suppression, epileptiform population bursts were minimized in amplitude by F1,6BP but did not disappear, and recovered quickly when F1,6BP was washed out from bathing medium (Figure 2A). In complete suppression, epileptiform bursts were fully abolished by F1,6BP and recovered much more slowly (up to 35 min; Figure 2B). Interestingly, single-unit activity in the slices was not blocked by F1,6BP ( Figure 2B, middle inset). Overall, F1,6BP effectively inhibited epileptiform bursts activity in brain slices, corroborating its anticonvulsant effects in vivo (Lian et al., 2007). However, the rapid time to blockade with F1,6BP (4.2 ± 0.7 min, Figures 2A,B) was surprising, because it occurred faster than expected for an action mediated via an intracellular metabolic pathway (for comparison, glycolytic inhibition with 2-DG blocks similar epileptiform activity in CA3 in 14.2 ± 1.6 min; Shao and Stafstrom, 2017). Thus, we asked whether the quick F1,6BP effect might be mediated by an acute extracellular mechanism.
(6/9) fired action potentials spontaneously at their RMP; the other neurons (3/9) were manually depolarized to ∼60 mV via current injection to fire action potentials. Under these experimental conditions, the neurons fired action potentials continuously at an average frequency of ∼2-3 Hz, mostly in bursts (Figures 3A-D), persisting for prolonged periods (up to >60 min). Bath application of F1,6BP did not stop neuronal firing (Figures 3A-C); in fact, it tended to increase the firing frequency in some neurons (Figures 3B,C). Overall, F1,6BP did not significantly change the frequency of intrinsic firing of CA3 neurons (Figures 3B,D; 118 ± 28 spikes/min vs. 161 ± 33 spikes/min, p > 0.05, n = 9), suggesting that F1,6BP does not block epileptiform population bursts through inhibition of intrinsic firing mechanisms.

F1,6BP Does Not Significantly Reduce Glutamatergic Synaptic Transmission
Next, we asked whether F1,6BP blocked glutamatergic transmission, and tested its effect on EPSC AMPA and EPSC NMDA separately. EPSC AMPA was evoked in the presence of AP-5 (50 µM) and SR95531 (10 µM) to block NMDA and GABA A receptors. The EPSC AMPA was typically fast (∼20 ms) with an amplitude of ∼200 pA ( Figure 4A). Addition of F1,6BP (5 mM) seemed to slightly reduce the amplitude of the EPSC AMPA but did not cause overall significant reduction (Figures 4A-C,G; 206 ± 23 pA vs. 173 ± 31 pA, p > 0.05, n = 9). EPSC NMDA was evoked in Mg 2+ -free solution in the presence of DNQX (10 µM) and SR95531 (10 µM) to block AMPA and GABA A receptors, and held at −65 mV. The extracellular Ca 2+ concentration was modestly elevated to 2 mM (from 1.3 mM) to partially compensate for membrane surface charge of Mg 2+ (Hille et al., 1975) and reduce multisynaptic responses. Compared with EPSC AMPA , the EPSC NMDA was usually larger and much slower (>100 ms; Figure 4D). Application of F1,6BP tended to decrease the amplitude of EPSC NMDA but the reduction was not significant overall (Figures 4E-G, 248 ± 45 pA vs. 186 ± 24 pA, p > 0.05, n = 8). These data suggest that it is unlikely that F1,6BP blocks epileptiform bursts through reduction of glutamatergic transmission.

F1,6BP Consistently Inhibits Voltage-Activated Calcium Currents (I Ca )
Given that two of the main contributors to epileptiform bursts, intrinsic neuronal firing and synaptic transmission, were not significantly altered by F1,6BP, it was reasonable to speculate that F1,6BP somehow acts through I Ca , which is known to be essential for the generation of epileptiform bursts (Schwartzkroin and Wyler, 1980;Kohling et al., 2000;Siwek et al., 2012). To test this hypothesis, we evoked I Ca in CA3 neurons using depolarizing voltage protocols (Figures 6A,E). I Ca began to be activated at ∼−40 mV and peaked at ∼−10 mV (Figures 6E,F). I Ca appeared to comprise a prominent inactivating component followed by a smaller steady-state component, and was completely abolished by the non-specific Ca 2+ channel blocker, CdCl 2 (100 µM; Figures 6A,E). These data are consistent with a previous study that showed that immature CA3 neurons have significantly more prominent inactivating I Ca than adult neurons (Thompson and Wong, 1991). The peak amplitude of I Ca ranged from 400 pA to 1400 pA (on average: 996 pA, n = 9). Application of F1,6BP (5 mM) quickly (2-3 min) and consistently attenuated I Ca by ∼40% (Figures 6A-D, 996 ± 97 pA vs. 600 ± 62 pA, p < 0.001, n = 9). To test the whether F1,6BP can block I Ca intracellularly, a subset of CA3 neurons were loaded with F1,6BP (5 mM) intracellularly through recording pipettes. In these neurons, the amplitude of I Ca activated by the same protocol was similar to that in non-loaded neurons (Figure 6D, 938 ± 91 pA, n = 6; vs. 996 ± 97 pA, n = 9; p > 0.05). Moreover, further extracellular application of F1,6BP caused similar reduction of I Ca in the F1, 6BP-loaded neuron (∼40%, Figure 6D, 938 ± 91 vs. 598 ± 64, p < 0.001, n = 6). These data suggest that F1,6BP most likely blocks I Ca extracellularly rather than intracellularly. F1,6BP did not seem to alter the voltage dependence of I Ca as it was activated at the same voltage range before and during F1,6BP application ( Figure 6F). Steadystate inactivation, assessed by the amplitude of I Ca after 1 s-long pre-pulses holding at different voltage levels (Figure 6G), was also not altered by F1,6BP. In fact, the inactivation kinetics of I Ca before and during F1,6BP were almost identical ( Figure 6H). Altogether, these data suggest that F1,6BP blocks a significant portion of I Ca , which may account for its anticonvulsant action on epileptiform bursts.

Epileptiform Bursts in CA3 Are Ca 2+ Dependent
If it were true that F1,6BP blocks epileptiform bursts by reducing I Ca , it would be expected that Ca 2+ channel antagonists would also block epileptiform bursts as F1,6BP did. To confirm this, we applied CdCl 2 , a non-selective Ca 2+ channel antagonist. As expected, a low concentration of CdCl 2 (20 µM) completely and reversibly blocked epileptiform bursts (Figures 7A-C). Similar to F1,6BP, CdCl 2 did not block the single-unit activities in the slices (Figures 2B, 7A, middle insets). Taken together, these data confirm that the epileptiform bursts in CA3 are Ca 2+ -dependent and can be blocked by Ca 2+ channel blockers or modulators (such as F1,6BP).

DISCUSSION
The main findings of the present study are: (1) the glucose metabolite and PPP stimulator, F1,6BP, effectively blocks epileptiform activity in hippocampal slices; (2) F1,6BP does not stop intrinsic neuronal firing or significantly reduce AMPAR or NMDAR-mediated synaptic currents; (3) F1,6BP causes a reduction of GABA A R-mediated synaptic currents; (4) F1,6BP consistently attenuates I Ca ; and (5) epileptiform bursts in this model are Ca 2+ dependent. To our knowledge, this is the first examination of the role of PPP in seizure reduction and the anticonvulsant action of F1,6BP in brain slices, focusing on underlying cellular mechanisms.

Novel Mechanism of F1,6BP's Anti-Seizure Activity
Our data in hippocampal slices corroborates the pioneering work showing that F1,6BP inhibits seizures in animal models (Lian et al., 2007). As in that study (Lian et al., 2007), we initially hypothesized that the anti-seizure activity of F1,6BP was mediated by an intracellular metabolic effect of shifting glycolysis to PPP (Figure 1), thus promoting the production of the endogenous antioxidant and anticonvulsant, GSH (Lian et al., 2007). Yet, we noticed that the time to block epileptiform bursts by F1,6BP (4.2 ± 0.7 min) was much quicker than epileptiform bursts blocked by the glycolytic inhibitor 2-DG in our previous study (14.2 ± 1.6 min; Shao and Stafstrom, 2017).
The short time to blockade suggested that F1,6BP might have a direct extracellular effect independent of its assumed intracellular effect and prompted us to examine its effect on extracellular mechanisms that contribute to neural excitability. However, F1,6BP did not stop neuronal firing nor did it significantly reduce excitatory synaptic transmission, two of the major contributors to epileptiform bursting (Traub and Wong, 1983). Finally and unexpectedly, we discovered that F1,6BP consistently blocks a significant portion of voltage-activated I Ca extracellularly. Since epileptiform bursts in CA3 are Ca 2+ -dependent and readily blocked by Ca 2+ channel blockade (CdCl 2 ), these data strongly suggest that F1,6BP achieves its anticonvulsant activity in part by attenuating Ca 2+ influx, at least in our in vitro seizure model.
Epileptiform bursts may be mediated by both presynaptic and postsynaptic Ca 2+ -dependent mechanisms. Presynaptic Ca 2+ influx is important for transmitter release. Postsynaptic (somatic) I Ca is known to be essential for the genesis and maintenance the plateau potentials of epileptiform bursts (Schwartzkroin and Wyler, 1980;Kohling et al., 2000;Siwek et al., 2012). Our data that F1,6BP did not cause significant reduction in EPSCs suggest that F1,6BP blocks epileptiform bursts mainly by reducing postsynaptic (somatic) rather than presynaptic Ca 2+ influx; and that a moderate reduction (40%) of postsynaptic Ca 2+ influx is sufficient to block seizures . The 40% reduction of Ca 2+ influx is derived from somatic high-voltage activated (HVA) Ca 2+ currents. Whether presynaptic Ca 2+ entry is similarly reduced by F1,6BP is unclear. Possibly, Ca 2+ channels at presynaptic terminals of glutamatergic synapses may be less sensitive to F1,6BP or F1,6BP might block certain types Ca 2+ channels that are not coupled to glutamate release (Turner et al., 1992;Dunlap et al., 1995). In contrast, Ca 2+ influx at the GABAergic presynaptic terminals is likely affected by F1,6BP as indicated by the reduction of GABA A current. These data suggest that F1,6BP differentially affects glutamatergic and GABAergic transmission. At the developmental period in this study (P8-16), the action of GABA on CA3 neurons is believed to have switched from excitatory (P0-P6) to inhibitory (Ben-Ari et al., 1990). However, the roles of GABAergic inhibition in seizures are complex. Recent studies have demonstrated that selective activation of perisomatic targeting interneurons (such as parvalbumin-expressing neurons) leads to initiation of seizures (Toyoda et al., 2015;Yekhlef et al., 2015;Khoshkhoo et al., 2017), while activation of axonic targeting interneurons (or axon-axonic synapses) inhibits seizures (Wang et al., 2014). In this study, we stimulated the CA3 pyramidal layer, where the axons of perisomatic interneurons are located, thus the resulting IPSCs are mostly derived from perisomatic synapses. A reduction in the strength of this type of inhibition by F1,6BP might therefore contribute to the reduction of epileptiform activity. Whether F1,6BP exerts additional effects on intracellular metabolic pathways in our model remains to be determined. Apparently, the mechanisms underlying the anticonvulsant action of F1,6BP in animal models (Lian et al., 2007) are likely to be more complex than in brain slices. The extent to which this extracellular mechanism contributes to F1,6BP's in vivo anticonvulsant action is uncertain. Arguably, administration of antioxidants does not consistently block seizures in the same animal models (Xu and Stringer, 2008), which suggests that F1,6BP may act on other targets in addition to PPP activation and GSH production. Nonetheless, the present study revealed a novel extracellular mechanism of F1,6BP that may explain its anticonvulsant activity, at least in this in vitro model of seizures.
CA3 neurons in the developing hippocampus possess both HVA and low-voltage activated (LVA) Ca 2+ channels (Thompson and Wong, 1991). It is generally recognized that Ca 2+ influx through HVA channels (particularly L-and R-types) are critically involved in the genesis of the plateau potential of epileptiform bursts Siwek et al., 2012), while Ca 2+ influx through LVA channels (i.e., T-type) contributes to rebound bursting and pacemaking activity (Siwek et al., 2012). The major component of I Ca in the present study resembles the HVA inactivating currents (R, N-, P/Q) with an activation threshold of −40 mV and fast decay kinetics (Yamakage and Namiki, 2002;Lacinová, 2005;Catterall, 2011), mixed with a smaller component of non-inactivating current (L-type). This is consistent with a previous study showing that inactivating I Ca contributed to a significantly larger percentage of the total I Ca in immature than adult CA3 neurons (Thompson and Wong, 1991). F1,6BP reduces the amplitude of I Ca without altering channel activation and inactivation kinetics, suggesting that F1,6BP does not bind to HVA Ca 2+ channels and works in a way that is different from conventional Ca 2+ channel blockers. We did not further differentiate which specific subtypes of HVA Ca 2+ channel are blocked by F1,6BP nor did we evaluate the effect of F1,6BP on LVA (T-type) I Ca , which is beyond the main focus of the present study and may be addressed in future studies. Alternatively, F1,6BP may reduce Ca 2+ influx by chelating Ca 2+ ions because of its negative charges (Hassinen et al., 1991). A reduction of Ca 2+ influx may also affect neuronal firing (by reducing Ca 2+ -activated K + currents). Our data show that F1,6BP tends to increase neuronal firing but this effect did not reach statistical significance. F1,6BP has previously been shown to be neuroprotective against hypoxic-ischemic injury (Farias et al., 1990;Kelleher et al., 1995;Sola et al., 1996) and excitotoxic injury (Rogido et al., 2003) by stabilizing intracellular Ca 2+ mediated by activation of phospholipase C (PLC; Donohoe et al., 2001;Fahlman et al., 2002;Rogido et al., 2003). Whether this mechanism occurs in seizures is not known. Also, we do not completely rule out the possibility that F1,6BP acts through an intracellular metabolic pathway (i.e., PPP), in addition to its action on Ca 2+ channels. Theoretically, with its negative charge and lack of known transporter, it may be difficult for exogenous F1,6BP to enter cells, but several studies have shown a specific membrane permeability for F1,6BP (Hardin and Roberts, 1994;Wheeler et al., 2004;Wheeler and Chien, 2012).

Metabolic Control of Neural Excitability and Metabolic Substrates as Novel Antiepileptic Drugs
Manipulation of neurometabolism exerts broad action and mechanisms influencing neural excitability and seizures. For instance, the anti-seizure effects of the KD may involve activation of adenosine receptors (Kawamura et al., 2010(Kawamura et al., , 2016 and K ATP (Ma et al., 2007;Lutas and Yellen, 2013), production of ketone bodies that inhibit transmitter release (Juge et al., 2010) or alteration of mitochondrial activity (Kim et al., 2015). Glycolytic inhibition with 2-DG not only blocks acute epileptiform activity in brain slices and animal models (Stafstrom et al., 2009;Forte et al., 2016;Shao and Stafstrom, 2017), but also interrupts the chronic process of epileptogenesis (Garriga-Canut et al., 2006;Stafstrom et al., 2009). Other metabolic interventions include disruption of the glial-neuronal energy coupling by targeting the so-called ''astrocyte-neuron lactate shuttle'' (Belanger et al., 2011) which also effectively blocks seizure activity (Sada et al., 2015). The efficacy of anticonvulsant activity of F1,6BP first reported by Stringer and colleagues and Lian et al. (2007) was quite remarkable in that it not only blocked seizures in all three animal models tested (KA, pilocarpine, PTZ), but also exceeded the treatment efficacy of 2-DG, valproic acid and KD. Therefore, F1,6BP appears to be a promising antiepileptic agent. The present study reveals an unexpected extracellular effect of F1,6BP, effectively blocking seizures by attenuating I Ca without disrupting normal physiological function (neuronal firing and excitatory synaptic transmission), which is desirable for clinical anti-seizure treatment. Moreover, its neuroprotective effect against hypoxic-ischemic injury would be additionally beneficial to patients, as seizures are often accompanied by or caused by hypoxic-ischemic injury, particularly in neonates (Kang and Kadam, 2015). Also, it is important to note that as a natural intracellular glucose metabolite, F1,6BP exerts very little toxicity even at a dose eight times larger than its effective dose (Vexler et al., 1999). All these features of F1,6BP makes it an attractive candidate for an effective and safe antiepileptic agent.
In summary, metabolic interventions represent a novel strategy for treatment of drug-refractory seizures. We have identified a novel mechanism of the promising metabolic anticonvulsant, F1,6BP, which may provide new insight into its anti-seizure and/or neuroprotective activity.

AUTHOR CONTRIBUTIONS
L-RS and CS conceptualized the study and wrote the manuscript; L-RS designed the research and conducted the experiments, and analyzed data; and GW contributed to experiments.

FUNDING
This work was supported in part by generous gifts from the Mathias Koch Memorial Fund of the Community Foundation of Southern Wisconsin and the Sandra and Malcolm Berman Foundation. We thank Kennedy Krieger Institute for providing lab space and other facilities, and we are grateful to Dr. Michael Johnston for his enthusiastic support for our research and the Johnston lab for sharing lab supplies and providing administrative assistance.