Availability of neurotransmitter glutamate is diminished when β-hydroxybutyrate replaces glucose in cultured neurons


Address correspondence and reprint requests to Dr Trine Meldgaard Lund, Faculty of Pharmaceutical Sciences, University of Copenhagen, Universitetsparken 2, 2100 Copenhagen, Denmark.
E-mail: tml@farma.ku.dk


Ketone bodies serve as alternative energy substrates for the brain in cases of low glucose availability such as during starvation or in patients treated with a ketogenic diet. The ketone bodies are metabolized via a distinct pathway confined to the mitochondria. We have compared metabolism of [2,4-13C]β-hydroxybutyrate to that of [1,6-13C]glucose in cultured glutamatergic neurons and investigated the effect of neuronal activity focusing on the aspartate–glutamate homeostasis, an essential component of the excitatory activity in the brain. The amount of 13C incorporation and cellular content was lower for glutamate and higher for aspartate in the presence of [2,4-13C]β-hydroxybutyrate as opposed to [1,6-13C]glucose. Our results suggest that the change in aspartate–glutamate homeostasis is due to a decreased availability of NADH for cytosolic malate dehydrogenase and thus reduced malate–aspartate shuttle activity in neurons using β-hydroxybutyrate. In the presence of glucose, the glutamate content decreased significantly upon activation of neurotransmitter release, whereas in the presence of only β-hydroxybutyrate, no decrease in the glutamate content was observed. Thus, the fraction of the glutamate pool available for transmitter release was diminished when metabolizing β-hydroxybutyrate, which is in line with the hypothesis of formation of transmitter glutamate via an obligatory involvement of the malate–aspartate shuttle.

Abbreviations used



liquid chromatography mass spectrometry


malate–aspartate shuttle


molecular carbon labeling


monocarboxylate transporter

Glucose is the primary energy substrate for the adult mammalian brain, but lactate and ketone bodies can be used as alternative energy substrates (Owen et al. 1967; Schurr et al. 1988, 1997; Dalsgaard 2006). Ketone bodies include β-hydroxybutyrate, acetoacetate and acetone, of which the latter is generally considered of little metabolic significance. When glucose availability is too low to meet brain requirements, such as during starvation, the liver generates ketone bodies from the catabolism of fatty acids (Owen et al. 1967). The ketone bodies enter the brain via the monocarboxylate transporters (MCT) and during prolonged starvation it may even replace glucose as the primary fuel in adult brain (Owen et al. 1967).

It is well known that starvation or a ketogenic diet diminishes the number of seizures in some patients with epilepsy. Although its biochemical basis is not clarified, several hypotheses have been suggested (Morris 2005; Gasior et al. 2006; Bough and Rho 2007; Ma et al. 2007). One of these hypotheses involves an altered brain amino acid metabolism (Hawkins et al. 1971; Erecinska et al. 1996; Daikhin and Yudkoff 1998; Yudkoff et al. 2001; Melo et al. 2006). Another recently proposed hypothesis suggests an altered ATP distribution in β-hydroxybutyrate nourished cells affecting KATP channels (Ma et al. 2007). Glucose and β-hydroxybutyrate are both metabolized to acetyl-CoA, however, the initial catabolic steps take place in different compartments; i.e., glucose is mainly metabolized via glycolysis in the cytosol and subsequently pyruvate is metabolized in the mitochondria, whereas β-hydroxybutyrate directly enters the mitochondria. Thus, the required activity of the malate–aspartate shuttle (MAS) transferring reducing equivalents generated in glycolysis is negligible when β-hydroxybutyrate is metabolized and this could affect the homeostasis of essential neuroactive amino acids in the cell. In particular, the balance between aspartate and glutamate in glutamatergic neurons might be affected since synthesis of transmitter glutamate entails the operation of one of the carriers of MAS, i.e. the ketodicarboxylate carrier (Palaiologos et al. 1988).

Only a few studies have compared the metabolism of ketone bodies with that of glucose. A study using synaptosomes (McKenna et al. 1994) and one using cultured neurons (Lopes-Cardozo et al. 1986), both focus on the complete oxidation of the substrates, measuring the production of CO2, whereas a detailed metabolic mapping of the effects of these two substrates was not performed. Moreover, in a recent study by Scafidi et al. (2009)β-hydroxybutyrate metabolism was studied in rats in vivo using 13C-labeling. In analogy with this the present study was designed to compare metabolism of [1,6-13C]glucose and [2,4-13C]β-hydroxybutyrate to elucidate glutamate and aspartate homeostasis in resting as well as activated cultured cerebellar granule neurons, which are mainly glutamatergic (Drejer et al. 1982; Sonnewald et al. 2004). Neuronal activity was induced by repetitive exposure to NMDA selectively causing vesicular release of neurotransmitter glutamate (Bak et al. 2003). It has previously been shown that glucose can maintain neurotransmitter homeostasis as the only energy substrate, whereas this cannot be fulfilled by lactate, during synaptic activity induced in cultured cerebellar granule neurons (Bak et al. 2006). However, whether β-hydroxybutyrate can compensate for a lack of glucose and maintain neurotransmitter homeostasis is not known.

Materials and methods


NMRI mice were obtained from Taconic M&B (Ry, Denmark), plastic tissue culture flasks from NUNC A/S (Roskilde, Denmark) and fetal calf serum from Harlan Sera-Lab Ltd. (Sussex, UK). Culture medium, poly-d-lysine (MW > 300 000), l-glutamine, p-aminobenzoic acid (pABA), sodium (R)-(-)-3-hydroxybutyrate were from Sigma (St Louis, MO, USA) and penicillin from LEO (Ballerup, Denmark). o-Phthalaldehyde-reagent was obtained from Agilent Technologies (Nærum, Denmark). Sodium D-β-[2,4-13C]hydroxybutyrate (99% enriched) was from Cambridge Isotopes Laboratories (Woburn, MA, USA) and D-[1,6-13C]glucose (98% enriched) from Isotec (St Louis, MO, USA). For amino acid analysis by the LC-MS method, the EZ:faast LC-MS kit (KHO-7337) was used and for HPLC, a Gemini column was used, both from Phenomenex (Torrance, CA, USA). All other chemicals were of the purest grade available from regular commercial sources.

Cell cultures

Cerebellar neurons were isolated from 7-day-old mice and cultured as described by Schousboe et al. (1989). In brief, after dissection, the tissue was trypsinized (0.25 mg/mL trypsin, 15 min, 37°C) followed by trituration in a Dnase solution (75 IU/mL) containing soybean trypsin inhibitor (0.52 mg/mL). The cells were suspended in modified Dulbecco’s minimum essential medium (Schousboe et al. 1989) containing 24.5 mM KCl, 12 mM glucose, 7 μM pABA, 50 μM kainic acid [to inhibit GABAergic cell function (Drejer and Schousboe 1989)], 0.05 I.U. insulin, 50 000 I.U. penicillin and 10% fetal calf serum and seeded in poly-d-lysine (50 mg/L) coated 25 cm2 or 80 cm2 culture flasks at a density of 2.75 × 106 or 3.50 × 106 cells/mL, respectively. The medium was supplemented twice (day 2 and 5) with an aliquot of glucose (1.2 M) to maintain a minimum concentration of 12 mM. Cytosine arabinoside was added to a final concentration of 20 μM after 2 days in culture to prevent astrocytic proliferation. The cells were cultured for 7–8 days at which point the cultures consist of 80–90% cells with glutamatergic characteristics (Drejer and Schousboe 1989).

Superfusion experiments

The culture medium was replaced with a Mg2 + -free Tris-buffered saline solution (50 mM Tris, 135 mM NaCl, 5 mM KCl, 1 mM CaCl2, pH 7.4, 37°C) containing either 1 mM [1,6-13C]glucose or 1 mM [2,4-13C]β-hydroxybutyrate or combinations of labeled and unlabeled substrates. (It should be noted that Tris-buffer does not enter the cells and thus does not contribute to intracellular buffering). Immediately thereafter, the cultures were placed in a heated (37°C) superfusion system in a sloping position so that the media could flow in from one end of the flask and be sucked out from the other. In this system the superfusion medium could be repetitively alternated as described in detail by Drejer et al. (1987). The cells were covered by a nylon mesh (80 μm) and superfused with the Tris-buffered saline solution containing the labeled substrates (37°C, 4 mL/min) for 74 min (control cultures). In some of the cultures the medium was repetitively changed every third minute for 30 s to a medium known to induce neuronal activity, i.e. neurotransmitter release (Bak et al. 2006). This depolarizing medium contained 300 μM NMDA, 10 μM glycine and an elevated potassium concentration (15 mM KCl) with an equimolar reduction in sodium (to 125 mM NaCl). Using this superfusion system the effect of repetitive depolarization on intermediary metabolism in neuronal cultures could be investigated in the presence of either [1,6-13C]glucose or [2,4-13C]β-hydroxybutyrate. The cell cultures were subsequently washed twice in ice-cold phophate-buffered saline (137 mM NaCl, 2.7 mM KCl, 7.3 mM Na2HPO4, 1.5 mM KH2PO4, 0.9 mM CaCl2, 0.5 mM MgCl2, pH 7.4) and metabolites extracted using 70% v/v ethanol. The cell residues were scraped off the dish and collected. The cell extract was separated from insoluble proteins after centrifugation at 20 000 g for 20 min at 4°C. The cell extracts were lyophilized and re-dissolved in water for analysis by LC-MS and 13C NMR. Protein contents in the pellets were determined according to Lowry et al. (1951) using bovine serum albumin as standard.


Aspartate and glutamate were quantified by HPLC using precolumn, on-line o-phthalaldehyde-reagent-derivatization and fluorescence detection on a Shimadzu 10A HPLC system equipped with a Shimadzu RF-10AXL fluorescence detector (Geddes and Wood 1984). The chromatography was performed on a Gemini C18 column (150 × 4.6 mm, particle size 5 μm, pore size 110 Å).


Lyophilized cell extracts were dissolved in 200 μL D2O containing 0.1% ethylene glycol. Proton decoupled 150.92 MHz 13C NMR spectra were obtained on a Bruker 600 MHz spectrometer (Bruker Analytik GmbH, Rheinstetten, Germany). Spectra were accumulated using a 30° pulse angle, an acquisition time of 1.08, and a 0.5 s relaxation delay. The number of scans was typically 50 000. Relevant peaks were integrated and quantified using ethylene glycol as quantification standard. Peak areas were corrected for nuclear Overhauser and relaxation effects by applying correction factors.


Aspartate and glutamate in the cell extracts were derivatized using Phenomenex EZ:faast amino acid kit (KHO-7337) and analyzed on a Shimadzu LCMS-2010 mass spectrometer with electrospray ionization coupled to a Shimadzu 10A VP HPLC system. Percentage incorporation of 13C in each isotopomer of glutamate and aspartate was determined after correction for natural abundance as described by Biemann (1962). Mono labeling was designated (M + 1), double labeling (M + 2), triple labeling (M + 3) etc. To obtain a measure of the total 13C incorporation in each amino acid, an average percent of labeling, called percent molecular carbon labeling (MCL) for each metabolite was calculated as described by Bak et al. (2006). As an example, glutamate can have from one to five 13C atoms (designated M + 1 to M + 5). Percentage of each isotopomer (M + 1, M + 2, etc.) is multiplied by the ratio of labeled carbon atoms (1/5 for M + 1, 2/5 for M + 2 etc.); these values are summed and expressed as the average percentage of labeling for each metabolite. MCL reflects total labeling and these values are dependent upon both the acetyl CoA enrichment and the TCA cycling.

Metabolic pathways and labeling patterns

The position of 13C-labeling in [1,6-13C]glucose and [2,4-13C]β-hydroxybutyrate was chosen to make a comparison possible, since they are both metabolized to [2-13C]acetyl-CoA (Fig. 1a). [1,6-13C]Glucose can be converted to [3-13C]pyruvate via glycolysis and subsequently to [2-13C]acetyl-CoA. On the other hand, [2,4-13C]β-hydroxybutyrate is metabolized to [2-13C]acetyl-CoA via [2,4-13C]acetoacetate and [2,4-13C]acetoacetyl-CoA. [2-13C]Acetyl-CoA generated from either [1,6-13C]glucose or [2,4-13C]β-hydroxybutyrate may condense with unlabeled oxaloacetate producing [2-13C]citrate initiating the first turn of label incorporation into TCA-cycle intermediates and related amino acids, i.e. glutamate and aspartate (Fig. 1a). The subsequent steps yield [4-13C]α-ketoglutarate, which may be transaminated to [4-13C]glutamate giving rise to a singlet in the 13C NMR spectrum and mono labeling (M + 1) in the mass spectrum. If [4-13C]α-ketoglutarate is decarboxylated to succinyl-CoA for further TCA-cycle metabolism, labeling of malate and oxaloacetate will occur in either the C2 or C3 position (due to the symmetry of succinate). [2/3-13C]Oxaloacetate may be transaminated to [2/3-13C]aspartate (singlet, M + 1) catalyzed by aspartate aminotransferase. Alternatively, [2/3-13C]oxaloacetate may be condensed with either labeled or unlabeled acetyl-CoA, giving rise to double labeling (M + 2), i.e. [3,4-13C] or [2,4-13C]glutamate (Fig. 1b) or mono labeling (M + 1), i.e. [2-13C] or [3-13C]glutamate, respectively. Triple labeling (M + 3) may arise from three consecutive condensations of [2-13C]acetyl-CoA. Incorporation of 13C, labeling patterns plus integrals for glutamate and aspartate were analyzed by LC-MS and 13C NMR in cell extracts. The cellular content of glutamate and aspartate was determined by HPLC.

Figure 1.

 Metabolic scheme showing the metabolism of [2,4-13C]β-hydroxybutyrate and [1,6-13C]glucose and the position of 13C-labeling in aspartate and glutamate subsequent to condensation of unlabeled oxaloacetate and [2-13C]acetyl-CoA, i.e. the first turn of the TCA cycle (a). A simplified scheme showing labeling patterns (as black circles) in glutamate and aspartate subsequent to condensation of unlabeled oxaloacetate and [2-13C]acetyl-CoA and the following turns in which labeled oxaloacetate condenses with [2-13C]acetyl-CoA (b).

Data analysis

Glutamate is labeled from the TCA-cycle intermediate α-ketoglutarate by transamination, primarily by aspartate aminotransferase having a very high activity (Drejer et al. 1985; Yudkoff et al. 1994), thus, the labeling pattern of glutamate reflects that of the TCA-cycle intermediate α-ketoglutarate (Bak et al. 2006). From 13C NMR data a TCA-cycling ratio was calculated as the sum of labeling in glutamate C2 and C3, generated in the second and subsequent turns of the TCA cycle, divided by labeling in C4 formed whenever [2-13C]acetyl-CoA condenses with oxaloacetate. Thus, as this is a relative value, where the labeling in the numerator as well as the labeling in the denominator are proportional to enrichment in the acetyl-CoA pool, the enrichment in the acetyl-CoA pool is eliminated and the TCA-cycling ratio reflects activity in the TCA cycle independently of enrichment in the acetyl-CoA pool.


Results are presented as means ± standard error of the mean (SEM). Statistically significant differences were determined using Student’s t-test where indicated and otherwise one or two way anova and pair wise multiple comparison by the Holm-Sidak method. The level for statistically significant differences was < 0.05.


Cellular content of aspartate and glutamate

The cellular content of the two neuroactive amino acids glutamate and aspartate was determined in cell extracts of cultured cerebellar granule neurons superfused with glucose or β-hydroxybutyrate or with a combination of the two substrates (Table 1). Neurons receiving only glucose or glucose in combination with β-hydroxybutyrate had a decreased content of glutamate upon depolarization, but neurons receiving only β-hydroxybutyrate exhibited no significant changes upon depolarization. The content of aspartate was not affected by depolarization under any of the three substrate conditions.

Table 1.   Intracellular content of glutamate, aspartate and the sum of the two amino acids in cerebellar neurons after superfusion with either 1 mM glucose, 1 mM β-hydroxybutyrate or the combination of the two substrates
SubstrateConditionAmino acid content (nmol/mg protein)
GlutamateAspartateGlutamate + aspartate
  1. Cultured cerebellar neurons were superfused (see Materials and methods) with either 1 mM glucose, 1 mM β-hydroxybutyrate (β-HOB) or the combination of the two and repetitively depolarized with 300 μM NMDA, 10 μM glycine and 15 mM K+. The cell extracts were subsequently analyzed with HPLC and cellular contents of glutamate and aspartate determined. Results are mean ± SEM, with = 4 cultures for each condition with all cultures from the same batch. All conditions were repeated in other batches with similar results. Statistically significant differences between control and repetitively depolarized cells are indicated with an asterisk and differences between substrates are indicated as follows; #: between glucose and β-hydroxybutyrate, ¤: between β-hydroxybutyrate and the combination of the two substrates, all determined by two way anova and pair wise comparison by the Holm-Sidak method (< 0.05).

GlucoseControl41.2 ± 1.98.0 ± 0.749.3 ± 1.8
Depolarized29.9 ± 1.8*9.0 ± 0.138.9 ± 1.7*
β-HOBControl31.1 ± 1.7#18.1 ± 2.6#49.1 ± 3.1
Depolarized26.8 ± 1.813.8 ± 3.140.6 ± 3.8
Glucose + β-HOBControl44.9 ± 3.0¤4.6 ± 0.5¤49.6 ± 3.3
Depolarized29.9 ± 1.1*6.3 ± 0.3¤36.3 ± 1.1*

Under resting conditions neurons exposed to only β-hydroxybutyrate exhibited a cellular content of aspartate which was 10 nmol/mg protein higher than that observed in neurons exposed to only glucose, and the content of glutamate was decreased correspondingly to a value significantly lower than that found in neurons provided with glucose. When applying a combination of the two substrates, the amount of glutamate again increased and the amount of aspartate was reduced to the level observed in neurons receiving only glucose. The differences observed, between the three different substrate combinations under control conditions, were almost eliminated by repetitive depolarization, in which case no difference was found in the glutamate content. Furthermore, no difference was found in the aspartate content when comparing glucose and β-hydroxybutyrate, however, a lower content was found using a combination of the two substrates compared with β-hydroxybutyrate alone.

Since the two amino acids glutamate and aspartate are interconverted very rapidly by transamination, the sum of the two was calculated to test whether the total pool stayed constant under the different conditions. Indeed it did, no difference was found for either the control or the depolarized condition among the three different substrates; glucose, β-hydroxybutyrate or the combination (Table 1). However, there was a significant decrease from control to depolarized for both glucose and the combination of the two substrates, whereas when employing β-hydroxybutyrate as substrate no significant decrease was found in the total pool of glutamate plus aspartate upon depolarization.

Incorporation of 13C into glutamate and aspartate determined by 13C NMR

Typical 13C NMR spectra obtained from cell extracts of cerebellar neurons superfused with either [1,6-13C]glucose or [2,4-13C]β-hydroxybutyrate are presented in Fig. 2. [For explanation of 13C-labeling and splitting patterns, see Materials and methods (Fig. 1) and Waagepetersen et al. (1998)]. The spectra were integrated and the intracellular 13C contents of glutamate (C4, C2 and C3), aspartate (C2 and C3) as well as β-hydroxybutyrate (C2 and C4) are shown in Table 2. Upon depolarization using glucose as substrate, significantly lower 13C contents were found in both the singlet and the doublet of glutamate C4 and C2 as well as in the doublet and doublet of doublets of glutamate C3. In contrast, the singlets of aspartate C2 and C3 were not significantly changed upon depolarization. These singlets were the only quantifiable peaks of aspartate in the spectra, compatible with a low cellular content of aspartate in cultures superfused in buffer containing only [1,6-13C]glucose. Depolarization resulted in similar changes for glutamate when utilizing [2,4-13C]β-hydroxybutyrate, in which case the C4 singlet and doublet, the C2 doublet as well as the C3 doublet and doublet of doublets decreased. However, the decrease was not as pronounced as that observed in neurons receiving [1,6-13C]glucose. When utilizing [2,4-13C]β-hydroxybutyrate, a complex coupling pattern was detectable for aspartate C2 and C3, i.e. singlet, doublet and doublet of doublets, during the resting condition. Upon depolarization, the splitting pattern, i.e. doublet of doublets, which is obtained during extensive TCA-cycle metabolism incorporating [2-13C]acetyl-CoA, disappeared leaving only a singlet and a doublet for aspartate C2 and C3. Depolarization resulted in a slightly although not significantly higher content of [2,4-13C]β-hydroxybutyrate.

Figure 2.

13C NMR spectra of cell extracts from cultured cerebellar neurons (for details, see Materials and methods) superfused with either [2,4-13C]β-hydroxybutyrate or [1,6-13C]glucose for 74 min. Glu, glutamate; Asp, aspartate; Lac, lactate; β-HOB, β-hydroxybutyrate.

Table 2.   Intracellular content 13C in glutamate, aspartate and β-hydroxybutyrate from cultured cerebellar neurons after superfusion with either [1,6-13C]glucose or [2,4-13C]β-hydroxybutyrate during repeated depolarization and control conditions
SubstrateConditionContent of 13C-labeled isotopomer (nmol/mg protein)
Glu C4Glu C2Glu C3
Doublet of doublets
[2,3-13C]; [3,4-13C]
Doublet of doublets
  1. Cultured cerebellar neurons were superfused (see Materials and methods) in the presence of either 1 mM [1,6-13C]glucose or 1 mM [2,4-13C]β-hydroxybutyrate. The effect of repetitive neuronal depolarization using pulses (30 s) of 300 μM NMDA and 10 μM glycine in the presence of 15 mM K+ was investigated using NMR spectroscopy on extracted cellular metabolites. Results are means ± SEM, = 4 for each group; control or depolarized with either [1,6-13C]glucose or [2,4-13C]β-hydroxybutyrate. Statistically significant differences (< 0.05) between groups were determined by two way anova and pair wise comparison by the Holm-Sidak method. Differences between control and repetitively depolarized cells are indicated by * and differences between cells superfused with either 1 mM [1,6-13C]glucose or 1 mM [2,4-13C]β-hydroxybutyrate by †. Glu, glutamate; Asp, aspartate; β-HOB, β-hydroxybutyrate. ND, not detectable.

[1,6-13C]glucoseControl16.3 ± 1.012.8 ± 0.59.5 ± 0.47.2 ± 0.2ND1.9 ± 0.26.9 ± 0.26.7 ± 0.6
Depolarized3.9 ± 0.4*4.0 ± 0.3*3.7 ± 0.4*3.5 ± 0.4*ND2.1 ± 0.42.6 ± 0.4*2.5 ± 0.2*
[2,4-13C]β-HOBControl7.7 ± 1.0†9.5 ± 1.83.6 ± 0.4†4.3 ± 0.5†1.9 ± 0.41.1 ± 0.1†4.7 ± 0.8†5.1 ± 1.0
Depolarized4.8 ± 0.4*3.6 ± 0.3*2.5 ± 0.32.1 ± 0.2*†1.2 ± 0.11.3 ± 0.12.6 ± 0.3*1.9 ± 0.2*
  Asp C3Asp C2β-HOB C2β-HOB C4
Doublet of doublets
Doublet of doublets
[1,6-13C]glucoseControl1.1 ± 0.1NDND1.2 ± 0.1NDNDNDND
Depolarized0.9 ± 0.1NDND1.0 ± 0.1NDNDNDND
[2,4-13C]β-HOBControl2.1 ± 0.3†2.8 ± 0.41.4 ± 0.32.7 ± 0.5†4.2 ± 0.71.9 ± 0.313.6 ± 1.913.8 ± 2.0
Depolarized1.7 ± 0.2†1.3 ± 0.2ND1.6 ± 0.22.7 ± 0.9ND19.9 ± 1.920.4 ± 1.6

When utilizing [2,4-13C]β-hydroxybutyrate compared with [1,6-13C]glucose, a significantly lower content of 13C was found in the C2, C3 and C4 singlets as well as the C2 and C3 doublets of glutamate in the control situation and in the C2 doublet of glutamate in the depolarizing condition. However, significantly higher contents of 13C were found in aspartate C2 and C3 singlets under control as well as in the C3 singlet under depolarizing conditions using [2,4-13C]β-hydroxybutyrate compared with [1,6-13C]glucose.

TCA-cycle metabolism

For evaluation of TCA-cycle metabolism, a cycling ratio was calculated from the labeling patterns in glutamate as determined by 13C NMR (see Materials and methods for calculations). The values for the two substrates during resting and depolarizing conditions are presented in Table 3. Depolarization induced an approximately 60% increase in TCA-cycling ratio using [1,6-13C]glucose, whereas the TCA-cycling ratio was unaffected when [2,4-13C]β-hydroxybutyrate was the substrate. Furthermore, the cycling ratio in the depolarized condition employing [2,4-13C]β-hydroxybutyrate was significantly lower than when using [1,6-13C]glucose.

Table 3.   TCA-cycling ratio determined from the 13C-labeling in glutamate
SubstrateConditionTCA-cycling ratio inline image
  1. Cultured cerebellar neurons were superfused (see Materials and methods) with either 1 mM [1,6-13C]glucose or 1 mM [2,4-13C]β-hydroxybutyrate (β-HOB) and repetitively depolarized with 300 μM NMDA, 10 μM glycine and 15 mM K+. The cell extracts were subsequently analyzed with NMR and the cycling ratio was calculated from the glutamate peaks in the spectrum. The content of 13C labeling in glutamate from the second turn and onwards was divided by labeling in C4 originating from the first turn. Results are means of cycling ratio ± SEM, = 4 for each condition. Statistically significant difference (< 0.05) between control and repetitively depolarized cells is indicated with an asterisk and difference between [1,6-13C]glucose and [2,4-13C]β-hydroxybutyrate is indicated by †, as determined by two way anova and pair wise comparison by the Holm-Sidak method.

[1,6-13C]glucoseControl1.12 ± 0.03
Depolarized1.81 ± 0.12*
[2,4-13C]β-HOBControl1.21 ± 0.04
Depolarized1.37 ± 0.02†

Molecular 13C-labeling in glutamate and aspartate determined by LC-MS

To obtain more detailed information about the relative importance of glucose and β-hydroxybutyrate for maintaining aspartate–glutamate homeostasis, a series of experiments were performed in which either [1,6-13C]glucose or [2,4-13C]β-hydroxybutyrate was used alone or in combination with its unlabeled counterpart. Using mass spectrometry, the MCL was determined for glutamate and aspartate in cell extracts of cerebellar neurons during resting or depolarizing conditions. As shown in Fig. 3a labeling in glutamate decreased significantly, whereas no significant change was found in aspartate upon depolarization in the presence of [1,6-13C]glucose only. However, when only [2,4-13C]β-hydroxybutyrate was present a significantly lower incorporation into aspartate was found upon depolarization (Fig. 3b). Comparing the MCL for aspartate under resting conditions for the two different substrates it was significantly higher (= 0.01), i.e. 44%, using [2,4-13C]β-hydroxybutyrate compared with 28% using [1,6-13C]glucose. When using 1 mM [1,6-13C]glucose and 1 mM unlabeled β-hydroxybutyrate no significant change was found for either glutamate or aspartate upon depolarization (Fig. 3c), whereas a decrease in labeling was found in both amino acids when β-hydroxybutyrate was the labeled compound in combination with unlabeled glucose (Fig. 3d). A dilution of the labeled acetyl-CoA pool and thus a lower MCL would be expected using a combination of unlabeled and labeled substrate as observed comparing results obtained by utilization of labeled glucose with and without unlabeled β-hydroxybutyrate (Fig. 3a and c). Surprisingly, no reduction was found in the MCL for glutamate under neither resting nor depolarizing conditions when comparing the situation with [2,4-13C]β-hydroxybutyrate alone (Fig. 3b) with that where labeled β-hydroxybutyrate was used together with unlabeled glucose (Fig. 3d). However, the additional presence of glucose caused a shift in the labeling from aspartate to glutamate, as aspartate had a significantly higher labeling in the control situation when β-hydroxybutyrate was used alone (aspartate 44.0 ± 2.8%; glutamate 28.3 ± 3.5%, = 0.01), whereas glutamate had the highest labeling when a combination of the two substrates was used (aspartate 28.3 ± 1.9%; glutamate 33.4 ± 1.0%, = 0.04) (Fig. 3b and d). In the two experiments with either 1 mM labeled glucose and 1 mM unlabeled β-hydroxybutyrate (Fig. 3c) or vice versa (Fig. 3d), the cultures had been exposed to exactly the same conditions, only the position of 13C label being changed from glucose to β-hydroxybutyrate. 13C from [2,4-13C]β-hydroxybutyrate (Fig. 3d) was more readily incorporated into glutamate and aspartate under both control and depolarizing conditions than 13C-labeling from [1,6-13C]glucose (Fig. 3c) when both substrates were present concurrently (< 0.05 for all four comparisons). However, depolarization caused a significantly lower incorporation of 13C from [2,4-13C]β-hydroxybutyrate into both glutamate and aspartate, whereas incorporation of 13C coming from [1,6-13C]glucose seemed to increase, although this was not significant. Altogether this gives a relative shift in the preference of substrate towards glucose upon depolarization. Still having in mind that the cultures were exposed to the same substrates in the two experiments (Fig. 3c and d), this shift can be further emphasized by calculating the relative contribution of 13C in glutamate or aspartate (% MCL) coming from [1,6-13C]glucose compared with the sum of 13C incorporation coming from both [1,6-13C]glucose and [2,4-13C]β-hydroxybutyrate, e.g. (7.2/(7.2 + 33.42)) × 100% = 17.8% for glutamate in the control situation. Thus, upon depolarization a shift was seen in the preference of substrate towards glucose as the relative contribution of 13C-labeling from [1,6-13C]glucose in glutamate increased from 17.8 ± 0.9 to 31.8 ± 1.4% (Student’s t-test, < 0.001) and in aspartate from 17.6 ± 1.0 to 36.6 ± 4.4% (Student’s t-test, = 0.002).

Figure 3.

 Molecular carbon labeling (% MCL) of glutamate and aspartate in cell extracts of cerebellar granule cells (for details, see Materials and methods) superfused with 1 mM [1,6-13C]glucose (a), 1 mM [2,4-13C]β-hydroxybutyrate (b), 1 mM [1,6-13C]glucose and 1 mM β-hydroxybutyrate (c) and 1 mM [2,4-13C]β-hydroxybutyrate and 1 mM glucose (d) under resting (bars in grey) and depolarizing conditions (bars in black). Bars indicate mean ± SEM and the number of individual cultures was 6–8 originating from two to three different experiments. * Significantly different from the corresponding resting condition, § significantly different from the same condition without unlabeled counterpart (i.e. c versus a; d versus b) (< 0.05) as determined by two way anova and pair wise multiple comparison by the Holm-Sidak method.

Molecular distribution of 13C in glutamate and aspartate from [2,4-13C]β-hydroxybutyrate

To further elucidate β-hydroxybutyrate metabolism and its impact on aspartate–glutamate homeostasis a detailed mapping of the labeling patterns for glutamate and aspartate was performed (Fig. 4). Employing [2,4-13C]β-hydroxybutyrate as substrate a significantly higher labeling was found in M + 1 for both glutamate and aspartate, whereas M + 2 and M + 3 decreased for aspartate upon depolarization (Fig. 4a and 4b). When [2,4-13C]β-hydroxybutyrate was combined with unlabeled glucose, M + 1 increased for glutamate upon depolarization, whereas M + 2, M + 3 and M + 4 decreased (Fig. 4c), as also observed for M + 2 and M + 3 of aspartate (Fig. 4d). An augmented labeling in M + 1 and a lower labeling in M + 2, M + 3 and M + 4 is compatible with a dilution of labeling in the acetyl-CoA pool. Labeling in glutamate increased by adding unlabeled glucose for M + 2 and M + 3 in the control situation as well as for M + 2 in the depolarized situation (Figs 4a and 4c). As opposed to this, labeling in aspartate was reduced for M + 3 in the control and for all isotopomers in the depolarized condition when unlabeled glucose was added. (Fig. 4d compared with 4b). Thus, the impact of the presence of glucose in combination with [2,4-13C]β-hydroxybutyrate was a higher labeling in glutamate and a reduced labeling in aspartate.

Figure 4.

 Percentage 13C-labeling in each isotopomer, monolabeling (M + 1), double-labeling (M + 2) etc. in glutamate and aspartate in cell extracts of cerebellar granule cells (for details, see Materials and methods) superfused with either 1 mM [2,4-13C]β-hydroxybutyrate (a and b) or 1 mM [2,4-13C]β-hydroxybutyrate plus 1 mM unlabeled glucose (c and d) under resting (bars in grey) and depolarizing conditions (bars in black). Bars indicate mean ± SEM and the number of individual cultures was 6–8 originating from 2–3 different experiments. *Significantly different from the corresponding resting condition, § significantly different from the same isotopomer without glucose (i.e. a versus c; b versus d) (< 0.05) as determined by two way anova and pair wise multiple comparison by the Holm-Sidak method.


Glucose versus β-hydroxybutyrate as energy substrates

Glucose is the primary energy substrate for the brain; but in the brain of suckling rats and neonate humans, ketone bodies are used as precursors for synthesis of amino acids and lipids (Morris 2005). However, most patients with inborn errors of ketogenesis develop normally, suggesting that the role for ketone bodies is as an alternative energy substrate to glucose during development and also during diseases or prolonged fasting (Morris 2005). Thus, a ketogenic diet can be used to treat patients with GLUT1 deficiency syndrome, which causes an impaired glucose transport into the brain. In these patients, ketone bodies compensate for the low brain glucose concentration and are used as an alternative fuel, which effectively restores brain energy metabolism (Klepper et al. 2005). A ketogenic diet has also shown remarkable effects in the treatment of drug-resistant childhood epilepsy (Gasior et al. 2006; Bough and Rho 2007; Yellen 2008) and in an animal model of traumatic brain injury infusion of β-hydroxybutyrate 1 h after injury led to a significant decrease in edema (Scafidi et al. 2009). However, underlying mechanisms need still to be further explored (Yudkoff et al. 2007).

In the present study, we tried to elucidate a part of this mechanism by studying the metabolic changes at the cellular level when changing energy substrate from glucose to β-hydroxybutyrate or by using combinations of the two substrates. We used cultured glutamatergic neurons and investigated the effect of inducing neuronal activity, which is known to stimulate neuronal oxidative metabolism (Bak et al. 2006).

Effect of depolarization on oxidative metabolism

During neuronal activity, sodium ions enter and potassium ions leave the cell, changing the membrane potential. As the sodium–potassium-ATPase restores the potential at the expense of energy, increased metabolic activity is expected upon re-polarization (Clarke and Sokoloff 1999). In the present study, a higher TCA-cycling ratio was found upon depolarization monitored by the use of [1,6-13C]glucose (Table 3), indicating increased oxidative metabolism. This is supported by an increased CO2 production during depolarization (Peng and Hertz 1993).

It could be speculated that the increased oxidation may involve glutamate as anaplerotic substrate, as we found a decrease in the glutamate content (Table 1). However, addition of unlabeled substrate, such as glutamate, into the pool of TCA-cycle intermediates via α-ketoglutarate has previously been shown to reduce the TCA-cycling ratio (Waagepetersen et al. 1998). In contrast, the TCA-cycling ratio calculated on the basis of 13C incorporation from β-hydroxybutyrate was unaltered upon depolarization. This finding is in line with a study in conscious rats by Cruz et al. (2005) showing that tracer amounts of [14C]hydroxybutyrate do not register an increase in energy metabolism during acoustic activation, as opposed to increases in glucose and acetate utilization. Moreover, in the present study, the MCL, i.e. total labeling, of aspartate was markedly decreased indicating a reduced metabolism of β-hydroxybutyrate in neurons exposed to depolarizing conditions (Fig. 3b). This was further emphasized by the detailed labeling patterns of glutamate and aspartate obtained by MS, indicating a decreased enrichment in the acetyl-CoA pool upon depolarization (Fig. 4a and b). The intracellular concentration of calcium increases in depolarized neurons, which in turn activates the mitochondrial Ca2+-uniporter (Kirichok et al. 2004). An augmented intra-mitochondrial Ca2+ concentration activates the pyruvate, isocitrate, and α-ketoglutarate dehydrogenases, potentially giving rise to an elevated mitochondrial NADH/NAD+ ratio (Nichols and Denton 1995; Pardo et al. 2006). The reduced catabolism of β-hydroxybutyrate may be caused by an inhibition of β-hydroxybutyrate dehydrogenase by the increased level of NADH and it may be compatible with the higher cellular content of β-hydroxybutyrate found in depolarized neurons, although this was not statistically significant. However, it seems in contrast to the finding by Maalouf et al. (2007) that ketone bodies increase NADH oxidation in isolated mitochondria. This apparent discrepancy may be explained by Maalouf et al. (2007) using a combination of β-hydroxybutyrate and acetoacetate clearly eliminating any involvement of the β-hydroxybutyrate dehydrogenase. Moreover, the effect observed in the present study is likely induced by increased mitochondrial [Ca2+], a condition which was not tested by Maalouf et al. (2007).

Although 13C from β-hydroxybutyrate was more readily incorporated into glutamate and aspartate than from 13C glucose when a combination of the two substrates were used, we found that depolarization shifted the preference slightly towards glucose, as an increased glucose and a decreased β-hydroxybutyrate metabolism was found during depolarizing conditions. This may support the conclusion by Bak et al. (2006) using an analogous preparation of glutamatergic neurons that glucose is necessary to maintain neurotransmitter homeostasis during synaptic activity.

Interaction between glucose and β-hydroxybutyrate metabolism

The present study showed an interesting interaction between glucose and β-hydroxybutyrate metabolism when using combinations of the two substrates, where one of them was 13C-labeled and the other unlabeled. The MCL in glutamate and aspartate, originating from [1,6-13C]glucose, was considerably reduced when also unlabeled β-hydroxybutyrate was present (Fig. 3a and c). This could be a consequence of dilution caused by an increase in the pool of unlabeled acetyl-CoA. However, using [2,4-13C]β-hydroxybutyrate, only a slight reduction was observed in the labeling of aspartate and that of glutamate was unaffected by addition of unlabeled glucose (Fig. 3b and d); the latter finding argues against a simple dilution effect in the acetyl-CoA pool. Similar results were obtained in synaptic terminals from weanling rats, where unlabeled glucose did not decrease the 14CO2 production from 14C-labeled β-hydroxybutyrate, whereas 14CO2 production from 14C-labeled glucose was decreased by β-hydroxybutyrate (McKenna et al. 1994). This observation may be explained by an inhibition of enzymes in glucose metabolism prior to acetyl-CoA formation or a competition between pyruvate and β-hydroxybutyrate for entry into the mitochondria. In favor of the first suggestion, β-hydroxybutyrate has been shown to inhibit the oxidative decarboxylation of pyruvate in rat brain mitochondria (Booth and Clark 1981), and in vivo studies have shown that i.v. infusion of β-hydroxybutyrate caused a markedly increased lactate production in the brain, indicating inhibition of pyruvate oxidation (Ruderman et al. 1974; Nehlig 2004). A likely mechanism involves allosteric feedback inhibition by acetyl-CoA of the pyruvate dehydrogenase complex as suggested by Nehlig (2004). Whether a competition between β-hydroxybutyrate and pyruvate for entry into mitochondria via MCT takes place remains uncertain, as little is known about the distribution and characteristics of MCT in the inner mitochondrial membrane.

The aspartate–glutamate homeostasis

The brain has a well-regulated interplay between excitation and inhibition of neurons. Excessive excitatory activity and/or low inhibitory activity is thought to lead to seizures, which is consistent with the repeated observation that glutamate receptor agonists or GABA receptor antagonists generate seizures in animal models (Croucher and Bradford 1989; Kondziella et al. 2002). As opposed to pharmacological correction of epileptic pathology, nutrition-based generation of ketone bodies - as in the ketogenic diet – may seek to establish excitatory-inhibitory balance via an alteration of neuronal intermediary metabolism (Yudkoff et al. 2007). Relevant to this, evidence is presented that the aspartate–glutamate homeostasis in glutamatergic neurons was affected differently by β-hydroxybutyrate or glucose.

Glucose and β-hydroxybutyrate are metabolized in different pathways and cellular compartments which may affect aspartate–glutamate homeostasis. In the presence of β-hydroxybutyrate, the aspartate content was higher and that of glutamate lower, than in the presence of glucose. During glycolysis, MAS is active, transporting reducing equivalents to mitochondria (Fig. 5a), whereas β-hydroxybutyrate is metabolized to acetyl-CoA directly in the mitochondria (Fig. 5b). In the latter case, the activity of cytosolic malate dehydrogenase reducing oxaloacetate to malate may be diminished, and consequently the exchange of malate and α-ketoglutarate via the ketodicarboxylate carrier of the MAS may be decreased (Fig. 5b). Such scenario may potentially lead to more of the mitochondrial α-ketoglutarate entering the TCA cycle, and thus enhancing the formation of oxaloacetate via TCA-cycle reactions and subsequently aspartate via aspartate aminotransferase; a pathway described by Hertz et al. (1992) as the truncated TCA cycle. The fact that β-hydroxybutyrate metabolism consumes succinyl-CoA in the step from acetoacetate to acetoacetyl-CoA may further accelerate the truncated TCA cycle. Depending upon the continued uptake of glutamate into the mitochondria either via the glutamate-hydroxyl-carrier or the aspartate–glutamate exchanger, the high activity of the truncated TCA cycle may lead to a limited accessibility of α-ketoglutarate in the cytosol for synthesis of glutamate. Interestingly, the biosynthesis of particularly the transmitter pool of glutamate has been shown to be dependent upon the operation of the ketodicarboxylate carrier (Palaiologos et al. 1988). Moreover, an enlarged pool of aspartate was found in cerebellar granule cells exposed to phenylsuccinate, an inhibitor of the ketodicarboxylate carrier (Passarella et al. 1987; Palaiologos et al. 1988). The MCL of aspartate was considerably higher under resting conditions, using [2,4-13C]β-hydroxybutyrate compared with using [1,6-13C]glucose. This supports the notion that the increased cellular aspartate content observed in neurons receiving only [2,4-13C]β-hydroxybutyrate is generated via the truncated TCA cycle. A similar shift in the content of glutamate and aspartate has been found both in rat brain synaptosomes and in cerebellar granule cells metabolizing glutamine when changing from a condition with glucose to a condition without glucose in which the production of cytosolic NADH and MAS activity is limited (Yudkoff et al. 1994; Peng et al. 2007). However, in these studies the lack of glucose is compensated for by glutamine as an anaplerotic substrate and probably pyruvate recycling to form acetyl-CoA (Olstad et al. 2007), whereas in our studies using β-hydroxybutyrate acetyl-CoA is not a limiting factor.

Figure 5.

 The activity of the individual processes involved in the malate–aspartate shuttle is affected under conditions when glucose is present (a) and when only β-hydroxybutyrate is present (b). In the latter case, no reducing equivalents are formed in glycolysis and the activity of the cytosolic malate dehydrogenase and subsequently the ketodicarboxylate carrier may decrease. This may result in an increased flow of intramitochondrial α-ketoglutarate into the TCA cycle and thus enhancing the formation of oxaloacetate, the latter being transaminated to aspartate. Such scenario might explain an increased intracellular content of aspartate in neurons receiving β-hydroxybutyrate only. The shift from glutamate to aspartate via the truncated TCA cycle is indicated with bold arrows. AcAc, acetoacetate; AcAc-CoA, acetoacetyl-CoA; Ac-CoA, acetyl-CoA; Asp, aspartate; cAAT, cytosolic aspartate aminotransferease; Cit, citrate; ETC, electron transport chain; Glu, glutamate; β-HOB, β-hydroxybutyrate; α-KG, α-ketoglutarate; mATT, mitochondrial aspartate aminotransferease; OAA, oxaloacetate; Suc, succinate; Suc-CoA, succinyl-CoA. Note that the equations are not stoichiometric as acetoacetyl-CoA reacts with CoA and yields two acetyl-CoA.

In line with the hypothesis presented above re-establishment of the activities of cytosolic malate dehydrogenase and the ketodicarboxylate carrier was expected when cerebellar neurons were superfused in the combined presence of glucose and β-hydroxybutyrate and, indeed, the aspartate–glutamate homeostasis was restored (Table 1). The additional presence of glucose caused a shift in the labeling from aspartate to glutamate, as aspartate had a significantly higher labeling when β-hydroxybutyrate was used alone, whereas glutamate had the highest labeling when a combination of the two substrates was used (comparing Fig. 3b with 3d and emphasized in detail in Fig. 4). Thus, as discussed earlier, the addition of glucose activated the MAS, reconstituting transport of α-ketoglutarate out of the mitochondria, which supports subsequent formation of transmitter glutamate via transamination at the expense of aspartate formation. In agreement with this, the energy substrate available, i.e. glucose, β-hydroxybutyrate or both, was found to have an impact on how cultured granule neurons responded to the repetitive exposure to NMDA leading to vesicular release of the neurotransmitter pool. In the presence of glucose, either alone or in combination with β-hydroxybutyrate the glutamate content as well as the total pool of glutamate plus aspartate decreased significantly upon activation of neurotransmitter release (Table 1). In contrast to this no significant decrease was observed in the presence of only β-hydroxybutyrate, neither in the glutamate content nor in the total pool of glutamate plus aspartate (Table 1). As mentioned above the ketodicarboxylate carrier, an obligatory entity of the MAS, has been shown to be functionally involved in the synthesis of transmitter glutamate (Palaiologos et al. 1988), which might explain a diminished pool of transmitter glutamate and a subsequent limited vesicular release in neurons receiving β-hydroxybutyrate as the only substrate.


Cultured glutamatergic neurons metabolized β-hydroxybutyrate more readily than glucose, when both substrates were available. This was evident during resting conditions as well as during induced neuronal activity, although glucose metabolism increased while that of β-hydroxybutyrate decreased due to depolarization. Interestingly, neurons metabolizing β-hydroxybutyrate as energy substrate, had an altered aspartate–glutamate homeostasis, with an increased aspartate and a correspondingly decreased glutamate content compared with those metabolizing glucose. We suggest that these changes are caused by a decreased activity of the MAS in neurons metabolizing β-hydroxybutyrate; a malfunctioning MAS has previously been shown to limit the amount of transmitter glutamate available for vesicular release (Palaiologos et al. 1988). Thus, our observation supports the hypothesis of reduced excitatory transmission playing a role in the anti-convulsive effect of the ketogenic diet (Patel et al. 1988). In patients treated with the ketogenic diet, the brain has access to both glucose and β-hydroxybutyrate, a condition in which the current experiments were unable to show any significant reduction in neuronal glutamate content. However, a shift towards ketone based energy metabolism may have a capacity to tune down the overall excitatory capacity of glutamatergic neurons potentially leading to an anti-convulsive effect.


Ms. Lene Vigh, Mette Clausen and Johanne Kroon Hansen are cordially acknowledged for expert technical support. This study was supported by grants from The Danish State Medical Research Council (271-07-0262), the Horslev, Alfred Benzon and Lundbeck Foundations.