|Home | About | Journals | Submit | Contact Us | Français|
Recent studies in rodent and human cerebral cortex have shown that glutamate-glutamine neurotransmitter cycling is rapid and the major pathway of neuronal glutamate repletion. The rate of the cycle remains controversial in humans, because glutamine may come either from cycling or from anaplerosis via glial pyruvate carboxylase. Most studies have determined cycling from isotopic labeling of glutamine and glutamate using a [1-13C]glucose tracer, which provides label through neuronal and glial pyruvate dehydrogenase or via glial pyruvate carboxylase. To measure the anaplerotic contribution, we measured 13C incorporation into glutamate and glutamine in the occipital-parietal region of awake humans while infusing [2-13C]glucose, which labels the C2 and C3 positions of glutamine and glutamate exclusively via pyruvate carboxylase. Relative to [1-13C]glucose, [2-13C]glucose provided little label to C2 and C3 glutamine and glutamate. Metabolic modeling of the labeling data indicated that pyruvate carboxylase accounts for 6 ± 4% of the rate of glutamine synthesis, or 0.02 μmol/g/min. Comparison with estimates of human brain glutamine efflux suggests that the majority of the pyruvate carboxylase flux is used for replacing glutamate lost due to glial oxidation and therefore can be considered to support neurotransmitter trafficking. These results are consistent with observations made with arterial–venous differences and radiotracer methods.
Neuronal–glial glutamate trafficking is critical to maintain neuronal glutamate. Glutamate released by the neurons through neurotransmission is replenished by precursors released by surrounding astrocyte cells (Fig. 1). Because neurons lack pyruvate carboxylase (Shank et al. 1985), which is required for the net synthesis of glutamate from glucose, they are almost completely dependent upon the astrocytes for net glutamate synthesis (Patel 1974; Yu et al. 1983; Shank et al. 1993). The requirements for replenishing neuronal glutamate are large, due to fast neuronal–glial glutamate trafficking. Measurements using 13C magnetic resonance spectroscopy (MRS) in vivo have shown that the rate of this pathway is similar to the rate of neuronal-glucose oxidation (Mason et al. 1995; Gruetter et al. 1998, 2001; Sibson et al. 1998, 2001; Shen et al. 1999; Chhina et al. 2001; Lieth et al. 2001; Lebon et al. 2002; Oz et al. 2004), which is one of the largest metabolic fluxes in the brain. Glutamate in nerve terminals is released into the synaptic cleft and then transported into surrounding astrocytes (McLennan 1976). Once in the astrocyte, glutamate may undergo two fates. In the pathway known as the glutamate-glutamine cycle it is converted by glutamine synthetase into glutamine (Martinez-Hernandez et al. 1976). Glutamine is then released from the astrocytes, transported into the neurons, and converted to glutamate by phosphate-activated glutaminase (Hogstad et al. 1988; Erecinska and Silver 1990; Yudkoff et al. 1993). In an alternate pathway, referred to as glutamate oxidation, released neuronal glutamate is oxidized in the astrocyte and replaced by anaplerosis, starting with the pyruvate carboxylase reaction (Hertz and Schousboe 1986; Hertz et al. 1999; Lieth et al. 2001; Lebon et al. 2002). Glutamate formed by anaplerosis is then converted by glutamine synthetase to glutamine, released, and taken up by the neuron and converted back to glutamate as in the glutamate-glutamine cycle. The rate of glutamine synthesis devoted to repletion of neuronal glutamate is then
where VcycleTot is total neuronal glutamate repletion from glutamine, Vcycle is the rate of the glutamate/glutamine cycle, and VanaGlutOx is the fraction of neuronal glutamate repletion from glutamine derived from anaplerosis.
In addition to glutamine synthesized for neuronal trafficking, a fraction of glutamine must be synthesized for ammonia detoxification. Glutamine synthesized by this pathway is released by the brain to maintain nitrogen balance. When this flow is included, the rate of glutamine synthesis is
Several studies using 13C MRS (Hassel et al. 1995; Mason et al. 1995; Gruetter et al. 1998, 2001; Sibson et al. 1998; Shen et al. 1999; Chhina et al. 2001; Sibson et al. 2001; Lebon et al. 2002; de Graaf et al. 2004; Oz et al. 2004; Patel et al. 2004) and 14C labeling methods (Dzubow and Garfinkel 1970; Lieth et al. 2001; Oz et al. 2004) have shown fast glutamate trafficking, ranging from 30% to 42% of the neuronal TCA cycle, which corresponds to ~60–84% of neuronal glucose consumption. However, a lack of consensus remains about the values of Vana,VanaNH4, and VanaGlutOx, particularly in humans. Human studies that used 13C MRS to follow labeling from [1-13C]glucose in the human brain at a magnetic field strength of 4T yielded values of 20% (Gruetter et al. 1998) and 35% (Gruetter et al. 2001) for Vana/Vgln. Based on the relative labeling of lactate and glutamate, Vana was assigned completely to ammonia detoxification (Gruetter et al. 1998), as opposed to glutamate oxidation and pyruvate recycling (Gruetter et al. 2001). In the same region of the brain, we reported a threefold lower value of VanaNH4 + VanaGlutOx (Shen et al. 1999), also using [1-13C]glucose, based primarily upon consideration of the initial labeling kinetics of glutamine C4. When [2-13C]acetate was supplied as an alternative precursor (Lebon et al. 2002), the rate of glutamate-glutamine cycling was in close agreement with that from [1-13C]glucose (Shen et al. 1999), but larger than the other human studies with [1-13C]glucose (Gruetter et al. 1998, 2001). Vgln, which is the sum of conversion from neurotransmitter glutamate and synthesis de novo via anaplerosis, was similar in all of the studies (Gruetter et al. 1998, 2001; Shen et al. 1999), but the reports of larger values of Vana (Gruetter et al. 1998, 2001) led to lower values of glutamate-glutamine neurotransmitter cycling.
A complication in the use of [1-13C]glucose to determine anaplerosis in the brain is that the label arrives at glutamate and glutamine through pyruvate dehydrogenase in both neurons and glia but also through pyruvate carboxylase in glia. Therefore, because the glial and neuronal pools are mixed by neurotransmitter cycling, the labeling of all positions of glutamate, glutamine, and other amino acids reflects both of these pathways. An alternate precursor for the rat brain has been [2-13C]glucose, to provide a more direct measurement of anaplerosis (Kanamatsu and Tsukada 1999; Sibson et al. 2001). The simplification is that [2-13C]glucose labels the amino acids glutamate and glutamine only in C2-C4 if label enters via pyruvate carboxylase, which is the first reaction in anaplerosis, and to a lesser extent by malic enzyme. Alternatively, if label enters through pyruvate dehydrogenase, it does not enter C2-C4.
In this paper we present results using [2-13C]glucose to calculate the fraction of total glutamine synthesis that is supplied by anaplerosis in the human brain. Results are compared with labeling observed with [1-13C]glucose in the same subjects to further assess the relative flux into the brain from pyruvate dehydrogenase and pyruvate carboxylase. Estimates of the fraction of anaplerosis devoted to ammonia detoxification relative to glutamate trafficking are made by comparison with previously published rates of glutamine release by the brain. It is concluded that anaplerosis accounts for approximately 6% of total glutamine synthesis, with a considerable fraction of this flux devoted to neurotransmitter repletion.
Six healthy subjects, four women and two men, were studied after obtaining informed consent in accordance with the procedures of the Yale Human Investigations Committee. Steady state measurements were made in two studies with [2-13C]glucose and [1-13C]glucose (99% atom percent enrichment; Cambridge Isotopes, Inc., Cambridge, MA, USA) on separate days in the same subjects. All NMR measurements were performed on a Bruker 2.1T whole-body magnet with an Avance spectrometer (Bruker Instruments, Billerica, MA, USA).
The overall experimental plan was to make three sets of measurements:
In one of the three sessions, graded image segmentation was performed using quantitative T1 images with B1 compensation to create images of percent gray matter, white matter, and CSF (Mason and Rothman 2002).
The subjects were admitted to the Yale University-New Haven Hospital General Clinical Research Center at 10 p.m. the night before the study and fasted from 10 p.m. onward. The following morning at 6–7 a.m., an i.v. catheter was placed in an antecubital vein in each arm, one for infusions and the other for measurements of plasma glucose concentration and [1-13C]glucose enrichment, and the subject rested quietly for 30 min to reduce potential effects of stress on glucose metabolism.
The glucose was infused during the study using a hyper-glycemic clamp, during which endogenous insulin secretion causes the body to absorb a large fraction of the infused glucose, which wastes the expensive labeled material. Therefore, to inhibit endogenous insulin secretion while maintaining basal insulin concentrations, somatostatin (Bachem, King of Prussia, PA, USA) was infused at a rate of 0.1 μg/kg/min, and insulin (U-100 Humulin, Eli Lily, Indianapolis, IN, USA) was infused at a rate of 24 pmol/m2/min to maintain basal insulin concentrations.
Solutions of 100 mm creatine, 25 mm glutamate, and 25 mM glutamine were measured using the same acquisition sequence to measure the relative detection efficiencies of each of the carbon resonances. The concentrations of natural abundance labeling of glutamate and glutamine were then calculated for the natural abundance measurements in humans by comparing the C2, C3, and C4 resonance areas to the natural abundance resonance area of creatine at 54.8 ppm; the total (labeled and unlabeled) concentrations of glutamate and glutamine were determined as the average 13C-labeled concentration divided by 0.011. Uncertainties in the natural abundance measurements of glutamate, glutamine, and creatine due to signal-to-noise were not determined independently but do contribute to the scatter seen in the measured concentrations of glutamate and glutamine.
Subjects lay supine on the scanner bed, with the occipital region of the head lying on padding against the 13C and 1H transceivers. Sagittal and occipital gradient-echo scout images were acquired for subject positioning, and shimming was performed using FASTERMAP (Shen et al. 1997). 13C detection was performed in a 144-cm3 voxel in the occipital-parietal lobe (6 cm × 6cm coronally oriented box 4 cm thick) using a polarization transfer sequence described previously (Shen et al. 1999), modified to use adiabatic surface suppression pulses for suppression of the signal from surface lipids. For 13C time course measurements, the free induction decays (FIDs) were accumulated in 5-min averages of 128 acquisitions each, using a sweep width of 4496 Hz and 1024 complex points. The average line width of the glutamate C4 single peak was 2.5 ± 1.1 Hz (mean ± SD). Because studies required separate experimental sessions, the position of the bed and subject head holder were noted and used for approximate subject positioning on separate days. In one experimental session, 13C MR spectroscopic data were accumulated in three 20-min accumulations to measure the 1.1% natural abundance signal. In another session, [2-13C]glucose was infused during the 13C acquisition, and in further session, [1-13C]glucose was infused during the 13C acquisition.
The data acquisition continued for 120 min, and the concentrations of plasma glucose were measured every 5 min. Blood samples for determination of plasma glucose 13C atom percent excess (APE) were drawn at 0, 5, 10, and 30 min and thereafter every 30 min for the duration of the study. The plasma glucose APE were measured by gas chromatography-mass spectrometry (Cline et al. 1999).
The 13C-labeled concentrations of glutamate C2, C3, and C4 and glutamine C2, C3, and C4 that resulted from the infusion of [2-13C]glucose were also calculated by comparison to the natural abundance resonance of creatine and adjusted according to the relative detection efficiencies of the glutamate and glutamine carbon positions (Blüml 1999). The percent enrichment of each carbon position was calculated by dividing the 13C-labeled concentration at each carbon position, obtained during the infusion, by the total concentration of each metabolite. Then the atom percent excess (APE) was determined by subtracting the 1.1% natural abundance labeling.
The 13C-labeled concentrations of glutamate C2, C3, and C4 and glutamine C2, C3, and C4 that resulted from the natural abundance measurement and infusions of [2-13C] and [1-13C]glucose were calculated by comparison with the natural abundance resonance of creatine, which was assumed to have a concentration of 9 mm (Petroff et al. 1989; Kreis et al. 1993; Michaelis et al. 1993; Hetherington et al. 1996) and adjusted according to the relative detection efficiencies of the glutamate and glutamine carbon positions as measured in phantoms of 100 mm creatine, 25 mm glutamine, and 25 mm glutamate. The cortical concentration of creatine was used instead of a level reflecting mixed white and gray matter, because the fivefold greater metabolic rate of gray matter (Lebrun-Grandié et al. 1983; Heiss et al. 1984; Wang et al. 1994; Mason et al. 1999; de Graaf et al. 2004) means that the measurements here reflect primarily cortical metabolism. The percent enrichment of each carbon position was calculated by dividing the 13C-labeled concentration at each carbon position, obtained during the infusion, by the total concentration of each metabolite, as measured from the natural abundance acquisition. Then the percent enrichment (PE) was determined by subtracting the 1.1% natural abundance labeling.
The spectral data were prepared for analysis with – 1 Hz/4 Hz Lorentzian-to-Gaussian conversion and 16-fold zero-filling followed by Fourier Transformation. Automated, least-squares curve fitting was used to fit the peak areas of glutamate C2, C3, and C4, glutamine C2, C3, and C4, and the creatine resonance at 54 ppm. Briefly, the fitting routine used model spectra of C4-labeled 13C glutamate and 13C-labeled glutamine to fit the data, using a spline baseline fit and treating any lipid contamination as three Gaussian peaks. The peak areas and line widths of the fitted model spectra were recorded.
A model of the flow from [2-13C]glucose to glutamate and glutamine C4, C3, and C2 was implemented in CWave software (Mason et al. 2003) (see Fig. 1a) using the equations in Table 1. The modeling of the pathway was extended from our previous results (Sibson et al. 2001) by the addition of the exchange between α-ketoglutarate and glutamate. Assumptions in the model include that anaplerosis occurs by the conversion of CO2 and pyruvate to oxaloacetate, only in the astrocyte (Patel 1974; Yu et al. 1983; Shank et al. 1993).
When the differential equations of Table 1 were solved to determine the value of Vana/Vgln, the resulting expression has the form
The term fPyr3 was assumed to be represented by the measured enrichment at glutamate C4, because [3-13C]pyruvate yields C4-labeled glutamate through pyruvate dehydrogenase. The equation for the fraction of glutamine synthesis that contributes directly to glutamate-glutamine neurotransmitter cycling is Vcycle/Vgln = 1 – Rana. An additional assumption that was made to arrive at eqn. 3 was that the concentration of astrocytic glutamate was much less than the concentration of neuronal glutamate (Van den Berg and Garfinkel 1971; Ottersen et al. 1992; Lebon et al. 2002), so that the fractional enrichment of the neuronal glutamate pool could be approximated as the total glutamate pool measured by MRS. To evaluate this assumption and the general validity of eqn. 3, the time-dependent calculation in CWave was run out to the isotopic steady-state to verify the correctness of the complicated steady-state expression. Eqn. 3 was solved for Rana under a variety of assumed values for RtcaA and RxA, beginning with the values deemed most likely and moving to upper and lower extremes.
Because eqn. 3 is complicated, a simpler approximation was derived that is valid if
The behavior of eqn. 4 was compared with the results obtained using the more the exact solution of eqn. 3. If one further assumes that RtcaA*[(fGln2 + fGln3)–2fGln3] < < 1 and < < (fGlu2 + fGlu3), with a small correction for pentose phosphate cycling, that fGln2 + fGln3 ≈ fGluA2 + fGluA3, and that fGlu2 + fGlu3 ≈ fGluN2 + fGluN3, then for Vcycle/Vgln one obtains the equation from Sibson et al. (2001):
The time to reach a steady-state level of isotopic labeling of glutamate and glutamine was determined primarily by the rates of the neuronal TCA cycle and the glutamate-glutamine neurotransmitter cycling, which is approximately 45 min with this protocol (Mason et al. 1995). The justification and testing of this assumption are provided in the Discussion.
The kinetic data from the infusions of [1-13C]glucose for each individual were fitted with the model shown in Fig. 1b using the time-dependent equations in Table 1, with Vana/Vgln set to the value determined from the steady-state portion of that subject's [2-13C]glucose infusion.
For a semiquantitative examination of the data, the labeling from [1-13C] and [2-13C]glucose was compared for each subject. To make a meaningful comparison, it was necessary to account for differences in the percent labeling of the glucose substrate in the blood. Therefore, the percent labeling above natural abundance for each metabolite was normalized by the percent labeling of plasma glucose.
The concentrations of glutamate and glutamine were 9.8 ± 2.4 and 4.2 ± 1.2 μmol/g (mean ± SD, n = 6), respectively, similar to values reported previously by 13C NMR in a similar volume of cortex (Gruetter et al. 1994), as well as in measurements published from autopsy samples (Perry et al. 1977, 1985). The tissue composition was 54 ± 3% gray matter (mean ± SD, n = 6).
Figure 2 shows a comparison of spectra obtained from the same subject during the isotopic steady state period of labeling during an infusion of [1-13C]glucose and [2-13C] glucose. Also shown is the spectrum obtained with 1 h of acquisition of the 1.1% 13C natural abundance measurement. As reported previously (Gruetter et al. 1994; Shen et al. 1999; Blüml et al. 2001) there is substantial enrichment from [1-13C]glucose in the C4, C3, and C2 positions of glutamine and glutamate. In contrast, when [2-13C]glucose was the precursor there was substantially less labeling in all positions of glutamate, consistent with slow anaplerosis relative to other metabolic flows in the brain. Table 2 shows the values of the APE for glutamate and glutamine from the [2-13C]glucose studies. The most intensely labeled position was the glutamine C3, as anticipated based upon the [2-13C]glucose label entering via pyruvate carboxylase in the astrocyte. The glutamate C3 position was labeled at approximately half the percent enrichment of glutamine C3, reflecting a substantial flow of glutamine synthesized in the astrocyte into the neuron (Dzubow and Garfinkel 1970; Erecinska and Silver 1990; Gruetter et al. 1998, 2001; Sibson et al. 2001). The labeling in glutamate C4 and in glutamine C4 was low but significant. The labeling of GABA and of aspartate is not reported because the relatively low concentrations combined with the very low percentage labeling of these metabolites made the measurement of their enrichments too imprecise to be useful.
To quantify anaplerosis, the enrichments of glutamate and glutamine C2, C3, and C4 were applied to eqns 3 and 4, using first the values of VtcaA = 0.14 μmol/g/min (Lebon et al. 2002) and VxA = 100 × VtcaA. Table 2 shows the results for this first set of assumed values, with a value of 0.059 ± 0.045 (mean ± SD). for Vana/Vgln. When the ratios of Vana/Vgln are combined with the kinetic studies with [1-13C]glucose, we obtain a value of 0.022 ± 0.020 μmol/g/min (Table 3)., The low rate of anaplerosis is consistent with earlier work showing that pyruvate carboxylase is on the order of 3–10% of the glucose utilization in the brain (Berl et al. 1962; Cheng 1971; Van den Berg and Garfinkel 1971; Siesjo 1978; Cooper and Plum 1987; Kanamatsu and Tsukada 1999).
Because few data exist about the values of VtcaA and VxA, upper and lower estimates for RtcaA and RxA (i.e. VtcaA/Vgln and VxA/Vgln) were also tested for their effects on the determination of Rana using eqn. 3. The solution was determined at the minimum value of RtcaA, for which the condition RtcaA = Rana was used, corresponding to the astrocytic TCA cycle functioning as a generator of intermediates to be shunted to glutamine and out of the system. The maximum value of VtcaA was 0.20 μmol/g/min, which is the value of 0.14 μmol/g/min estimated previously for VtcaA in humans, plus one standard deviation (Lebon et al. 2002). The minimum value of VxA was VtcaA, and the maximum value was 100, which is in practice equivalent to infinity. A term ‘s’ was used in eqn. 3 to address the effects of Vsc, the rate of scrambling of label between the C2 and C3 of oxaloacetate, with s 0 corresponding to Vsc = 0 and s = 0.5 for Vsc >> Vana. Table 3 shows the values of Vana/Vgln obtained under each of the conditions tested. Within all the extremes tested, Vana/Vgln lay between 0.033 and 0.107.
To evaluate the impact of assuming a negligibly small pool of astrocytic glutamate, the differential equations of Table 1 were integrated out to isotopic steady state and compared with the results seen for eqn. 3. For the current assumed pool size of 0.7 μmol/g (Lebon et al. 2002), differences in Vana/Vgln of less than 0.001 were seen.
To test the potential impact of an incorrect assumption about the time to isotopic steady state being only 45 min after the start of the infusion of [2-13C]glucose, the data were reanalyzed with the exact steady-state solution (eqn. 3) using only the data between 90 and 120 min after the start of the infusion. The mean value of Vana/Vgln differed insignificantly from the ratios obtained with longer averaging, having a mean value of 0.043 ± 0.031.
Figure 3 shows fits of the time courses of glutamate and glutamine C4 labeling from infusions of [1-13C]glucose in two of the subjects. Table 4 shows the absolute rate determinations provided by the kinetic data during the infusion of [1-13C]glucose for the same subjects, fixing the value of Vana/Vgln to the results obtained for each individual's own [2-13C]glucose. Because the [1-13C]glucose data lacked the sensitivity to determine the value of VtcaA, a previously published value of 0.14 μmol/g/min (Lebon et al. 2002) was used, along with the upper and lower limits of VtcaA = 0.20 μmol/g/min and VtcaA = Vana, respectively. From the expected value to the lower=limit, the absolute value of Vana changed insignificantly (0.021 vs. 0.022 μmol/g/min), whereas Vgln and VtcaTotal changed by – 7% and + 9%, respectively. From the expected value to the upper limit, the absolute value of Vana changed almost imperceptibly (0.22 μmol/g/min in both cases), and Vgln and VtcaTotal changed by only + 2.7% and −2.6%, respectively.
The rate of anaplerosis was measured in the awake human brain and found to constitute 6% of the rate of glutamine synthesis, in agreement with a previous finding that used [1-13C]glucose (Shen et al. 1999), as well as earlier arterial–venous differences and other measurements of brain CO2 incorporation vs. total glucose oxidation (Waelsch et al. 1964; Cheng et al. 1967; Cheng 1971; Siesjo 1978; Cooper and Plum 1987; Aureli et al. 1997). The uncertainty comprises a large fraction of the rate, with SD = 4% or SEM = 0.6%, because the rate is so low and the amount of labeling is therefore very small. We discuss here the comparison of our findings with previous results, implications for the role of anaplerosis in supporting neuronal glial glutamate trafficking and rebuilding glutamate lost to oxidation, and the validity of the steady-state measurement.
The finding of 6 ± 4% for the value of Vana/Vgln, or 0.022 ± 0.020 μmol/g/min, is similar to the value of 0.04 ± 0.02 μmol/g/min (mean ± SD) reported using [1-13C]glucose as a precursor (Shen et al. 1999), but lower than reports of 20% (i.e. 0.08 μmol/g/min) in two subjects (Gruetter et al. 1998) and 35%, or 0.09 ± 0.02 μmol/g/min (Gruetter et al. 2001). In those studies, which used [1-13C]glucose as a precursor, the flow through pyruvate carboxylase was estimated from the labeling time courses of the C2, C3, and C4 positions of glutamine and glutamate. Because of extensive flow of label into glutamate and glutamine C2 and C3 from [1-13C]glucose via pyruvate dehydrogenase, the analysis places high demands on sensitivity and spectral resolution to discriminate the much smaller label flow from pyruvate carboxylase. In contrast, the [2-13C]glucose method is less demanding methodologically, because it labels the glutamate and glutamine C2 and C3 almost exclusively by anaplerosis. The low labeling from anaplerosis relative to pyruvate dehydrogenase is evident when data obtained with [1-13C]glucose and [2-13C]glucose are compared for the same subject (Fig. 2). The labeling from [2-13C]glucose is at most 10% of the labeling from [1-13C]glucose. Note that the absolute amount of amino acid labeling from anaplerosis from [2-13C]glucose is the same as for [1-13C]glucose. The low labeling from [2-13C]glucose is consistent with studies of the cerebral cortex in the rat (Kanamatsu and Tsukada 1999; Sibson et al. 2001). To evaluate whether sensitivity limitations might explain the disagreement with the larger reported values reported recently (Gruetter et al. 1998, 2001), CWave was used to solve the differential equations in Table 3 to calculate values for the APE of Gln and Glu C2 and C3 given different values of Vana/Vgln. MR spectra were then simulated for the cases of Vana/Vgln = 0.20 and 0.35 (Fig. 4), which are the largest values reported using [1-13C]glucose as a tracer (Gruetter et al. 1998; Gruetter et al. 2001). The spectra with higher values of anaplerosis differ markedly from the observed data, and are well within the sensitivity of the measurement.
An alternate explanation for the differences obtained between the two labels is in limitations of the present metabolic modeling. Based on our sensitivity analyses the greatest effect will be due to the astroglial TCA cycle and the exchange rate between astroglial cytosolic glutamate and mitochondrial α-ketoglutarate. In the future, dynamic time courses using C2 glucose infusion at magnetic field strengths of 4T or higher could potentially resolve this issue. However, we should point out that for all human studies calculating Vana, the rate is a significant fraction of total glutamine synthesis and greater than can be explained by the need to synthesize glutamine for ammonia detoxification (see below).
Given a minimum estimate of CMRgl 0.5×VtcaTotal, or 0.39 μmol/g/min, net anaplerotic flow =must account for between 1% and 10% of glucose oxidation, in agreement with most previous studies using arterial–venous difference and other methods (Waelsch et al. 1964; Cheng et al. 1967; Cheng 1971; Siesjo 1978; Cooper and Plum 1987; Aureli et al. 1997).
The ratio of Vana/Vgln reported here is less than the 0.18–0.30 found in rats with 13C NMR or equivalent 14C methods (Patel 1974; Shank et al. 1993; Lapidot and Gopher 1994; Aureli et al. 1997; Sibson et al. 2001; Merle et al. 2002; Oz et al. 2004; Xu et al. 2004). With [2-13C]glucose in rats, the metabolic modeling and other methodology were similar to those used in the present study, so the higher ratio (Sibson et al. 2001) reflects either species differences or effects of anesthesia. Anesthesia reduces glutamate-glutamine trafficking (Sibson et al. 1998) and could therefore increase the ratio of Vana/Vgln. Consistent with this latter possibility, a recent report of gas chromatographic-mass spectrometric measurements of neurotransmitter glutamate enrichments in extra-cellular fluid yielded Vana/Vgln ≈ 10% (Kanamori et al. 2002).
In the other cited studies on rodents, quantitative comparisons with the present results are not possible because the glutamate-glutamine cycle was not modeled to evaluate the data, and in all but one study (Lapidot and Gopher 1994) an isotopic steady state was not achieved. If an isotopic steady state is not achieved with [1-13C]glucose as the labeled precursor, then the effect of the kinetic portion of the labeling influences the finding more heavily. Some of the label enters the glutamine pool through glial pyruvate carboxylase and goes almost directly to glutamine. However, most of the label flow into glutamine is derived from neuronal glutamate at the rate Vcycle, and before significant amounts of label can flow to glutamine from that source, the large neuronal glutamate pool must be enriched, which introduces a lag for Vcycle as a label source for glutamine. Therefore, the value of Vana/Vgln will be overestimated, because labeling from pyruvate carboxylase enters the glutamine directly in the glia, while label entry from neuronal pyruvate dehydrogenase via the glutamate-glutamine cycle is delayed by the time required to label the large neuronal glutamate pool.
Another barrier to comparison with the present results is that except for some localized measurements (Sibson et al. 2001; Kanamori et al. 2002; Kondrat et al. 2002), 13C labeling was obtained from the whole brain. The rate of neuronal-glial glutamate cycling in cerebral cortex should be considerably higher than in white matter and other regions with fewer glutamatergic synapses, resulting in a smaller Vana/Vgln in the cortex compared to whole brain. Future studies in which the [2-13C]glucose method is applied quantitatively to look at anaplerosis in the cerebral cortex of the rat brain will resolve these issues.
The ratio of Vana/Vgln requires an isotopic steady state in glutamate and glutamine. It may seem counterintuitive that an isotopic steady state could be reached if Vana supplies label slowly, but the time to steady state is dominated by the other, rapid flows. However, in case even VtcaN and Vgln fail to generate an isotopic steady state, the potential effects of incomplete turnover were assessed by two procedures.
First, we measured labeling in glutamate and glutamine C2, C3, and C4 at 45–70 min, 70–95 min, and 95–120 min. Because the 13C labeling was low in all of the studies, we averaged the data across all subjects for each time segment. The Vana/Vgln for each time segment differed insignificantly from one another and from the enrichments obtained with 45–90 min averaged together. As a further evaluation of the steady-state assumption, we simulated the labeling from [2-13C]glucose using rates of VtcaN and Vgln from previous studies (Mason et al. 1995; Shen et al. 1999; Gruetter et al. 2001), for values of Vana/Vgln from 0 to 40%. The value of Vana/Vgln was calculated from the simulations averaged over 45–120 min, 90–120 min, or infinity. In all cases the calculated ratios of Vana/Vgln from the intervals using eqn. 3 were close to the corresponding simulated values (Fig. 5), with underestimates of 1–4% (for Vana/Vgln = 0.06–0.4) resulting from missing steady state and from approximations that were made to derive eqn. 3. We also performed the same calculation with the data obtained in vivo from 90 to 120 min, finding that Vana/Vgln = 0.05 instead of the value of 0.06 that was calculated using the data from 45 to 120 min. The value of 39% for the ratio VtcaA/Vgln in eqn. 3 was taken from a previous study in which it was measured directly with [2-13C]acetate (Lebon et al. 2002). This ratio is greater than the value of 56% reported by Gruetter et al. (2001), but even a value of 56% changed the values of Vana/Vgln and Vcycle/Vgln by less than 0.5%.
In the glutamate oxidation pathway, referred to as pyruvate recycling (Cerdan et al. 1990; Künnecke et al. 1993; Hassel and Sonnewald 1995; Sonnewald et al. 1996; Aureli et al. 1997), glutamate is converted through glial glutamate dehydrogenase or transamination to α-ketoglutarate, oxidized to malate, converted to pyruvate by malic enzyme, and oxidized by pyruvate dehydrogenase (Hertz et al. 1999; Lieth et al. 2001). Glutamate thus oxidized is replaced by anaplerosis by conversion of pyruvate to oxaloacetate (OAA). Glial glutamate formed in this pathway cycles back to the neuron either as glutamine or as a TCA cycle intermediate such as α-ketoglutarate, although recent human cortical findings support glutamine as the major trafficking substrate (Lebon et al. 2002). As shown in eqn. 2, the fraction of anaplerosis associated with glutamate oxidation and neuronal-glial trafficking may be obtained by comparing Vana from the present study with previous measurements of human brain glutamine efflux. arterial–venous difference measurements yielded rates of 0.010–0.019 μmol/g/min (Lying-Tunell et al. 1981; Grill et al. 1992; Strauss et al. 2001), similar to the lower bound observed for Vana in the current study. The net efflux of all amino acids is 0.04 μmol/g/min (Strauss et al. 2001), slightly above the upper estimate of Vana in the current study. Even 0.04 μmol/g/min is considerably lower than the value of 0.09 μmol/g/min that was reported for VanaNH4 previously (Gruetter et al. 1998, 2001). Comparing the arterial–venous difference values with the present Vana of 0.01–0.04 μmol/g/min, eqn. 2 indicates that up to 75% of anaplerosis may be used to replenish oxidized neurotransmitter glutamate, with the remainder for repletion of amino acids and ammonia-related glutamine loss from the brain.
Also consistent with glutamate oxidation is the observed labeling of glutamate C4, which results from the conversion of glutamate to acetyl-CoA during pyruvate recycling. Under physiological conditions, pyruvate recycling through malic enzyme or PEPCK is considered to occur in the astroglial compartment only (Haberg et al. 1998; Lieth et al. 2001; Sibson et al. 2001), although there may be neuronal glutamate repletion through malic enzyme (Hassel et al. 2002). Pyruvate from this pathway may be converted to lactate and transported to neurons for conversion oxidation by pyruvate dehydrogenase (Bouzier et al. 2000). In a [2-13C]glucose experiment, this pathway may be identified by C4 labeling in glutamine and then neuronal glutamate. Small but significant glutamate and glutamine labeling in the C4 position was observed in the present study, consistent with this possibility.
An alternate repletion source for glutamate oxidation has been proposed in which neurons use malic enzyme to replenish glutamate directly (Hassel 2000). This pathway would result in the present experiment in labeling of the C3 position of glutamate. [2-13C]glucose and 14CO2 would both be incorporated by neuronal malic enzyme, and recent studies of the rat brain using these substrates in vivo indicate that the rate of this pathway in vivo is small (Lieth et al. 2001; Sibson et al. 2001). Both studies found labeling consistent with glial anaplerosis, the glutamate-glutamine cycle being the major glutamate repletion pathway. In the present study the glutamate C3 labeling was consistent with the majority of neurotransmitter cycling occurring via the glutamate-glutamine cycle.
As described previously (Sibson et al. 2001; Petroff et al. 2002), the relative steady state labeling of glutamate and glutamine C3 may be used to calculate the ratio between the rates of total neuronal glial glutamate cycling (VcycleTot = Vcycle + VanaGlutOx) and the neuronal TCA cycle VtcaN. The basis of this approach is that the majority of glutamate is neuronal (Van den Berg and Garfinkel 1971; Gundersen et al. 2001). As discussed earlier, neuronal malic enzyme activity is low, so the glutamate C3 labeling will come primarily from glutamine C3 supplied by glutamate from the astrocyte and secondarily from label entering neuronal pyruvate dehydrogenase due to pyruvate recycling and the pentose shunt. Table 2 shows that the fractional enrichment of glutamate C3 is several times higher than glutamate C4, indicating a substantial flow of label from glutamate neurotransmitter cycling. Quantitatively this can be expressed as the ratio
The models shows only the glial uptake of glutamine, which has been reported to be responsible for the vast majority of glutamate uptake (Danbolt et al. 1992; Tanaka et al. 1997; Maragakis and Rothstein 2004). Neuronal uptake of glutamate does occur, however, and any such transport into neurons is invisible to this model and is missing from the value of Vgln. Therefore, it is possible that the value of Vgln is somewhat larger than that reported here, meaning that the value of Vana/Vgln may be an overestimate of Vana relative to total glutamate release.
The application of a quantitative relationship to calculate the fractional rate of cycling is complicated by the experimental measurement interval of 90–120 min after the start of the infusion being insufficient for the glutamate C3 resonance to reach steady state labeling, resulting in a potential underestimation of VcycleTot/VtcaN. However, even with no compensation for incomplete labeling, a ratio of 0.73 was calculated for the group, providing additional evidence for a substantial cycling flux, and consistent with large values reported using [1-13C]glucose and [2-13C]acetate as tracers (Mason et al. 1995; Gruetter et al. 1998; Shen et al. 1999; Lebon et al. 2002).
A strategy of labeling from [2-13C]glucose simplifies the measurement of the rate of pyruvate carboxylase in the brain. With this approach, we have found that the rate of anaplerosis in awake human occipital parietal cortex is 6 ± 4% (mean ± SD) of glutamine synthesis, or 0.022 ± 0.020 μmol/g/min. Analysis of assumptions made in the modeling showed that the maximum and minimum values possible ranged from 2% to 11% of the total rate of glutamine synthesis. Based on comparisons with published values for the rate of efflux of glutamine from the human brain, as well as label scrambling into the glutamate and glutamine C4 positions, the majority of this flux is being used to replace neurotransmitter glutamate that is oxidized in the astrocyte, and hence may be considered part of neuronal glutamate repletion as opposed to ammonia detoxification. The labeling of glutamate and glutamine C3 is consistent with the finding in previous human studies that the rate of neuronal-glial neurotransmitter cycling is high in the awake, resting human cerebral cortex. Derangements in anaplerosis and glutamate oxidation may play a role in the pathogenesis of a variety of neurological disorders such as epilepsy, hepatic encephalopathy, and neurodegenerative disorders (During and Spencer 1993; Csernansky et al. 1996; Kanamatsu and Tsukada 1999; Blüml et al. 2001; Petroff et al. 2002). The combination of [2-13C]glucose labeling with non-invasive 13C MRS yields the potential to study the involvement of these pathways in human disease in a research setting.
We thank Ms Yanna Kossover, Mr. Mikhail Smolgovsky, Mr. Anthony Romanelli, and the staff of the Yale – New Haven Hospital Adult Clinical Research Center (GCRC) for technical assistance with the studies, Gary Cline for gas chromatography/mass spectrometry. This project was supported by grants from the National Institutes of Health: K02 AA-13430 (GFM), R01 DK-49230, P30 DK-45735, R01 NS-0375279 (DLR), K23 DK-02347. GFM was supported in part by a Young Investigator Award from the National Alliance for Research on Schizophrenia and Depression (NARSAD).