Assessment of metabolic fluxes in the mouse brain in vivo using ¹H-[¹³C] NMR spectroscopy at 14.1 Tesla

Abstract

¹³C magnetic resonance spectroscopy (MRS) combined with the administration of ¹³C labeled substrates uniquely allows to measure metabolic fluxes in vivo in the brain of humans and rats. The extension to mouse models may provide exclusive prospect for the investigation of models of human diseases. In the present study, the short-echo-time (TE) full-sensitivity ¹H-[¹³C] MRS sequence combined with high magnetic field (14.1 T) and infusion of [U-¹³C₆] glucose was used to enhance the experimental sensitivity in vivo in the mouse brain and the ¹³C turnover curves of glutamate C4, glutamine C4, glutamate+glutamine C3, aspartate C2, lactate C3, alanine C3, γ-aminobutyric acid C2, C3 and C4 were obtained. A one-compartment model was used to fit ¹³C turnover curves and resulted in values of metabolic fluxes including the tricarboxylic acid (TCA) cycle flux V_TCA (1.05±0.04 μmol/g per minute), the exchange flux between 2-oxoglutarate and glutamate V_x (0.48±0.02 μmol/g per minute), the glutamate-glutamine exchange rate V_gln (0.20±0.02 μmol/g per minute), the pyruvate dilution factor K_dil (0.82±0.01), and the ratio for the lactate conversion rate and the alanine conversion rate V_Lac/V_Ala (10±2). This study opens the prospect of studying transgenic mouse models of brain pathologies.

Introduction

In vivo ¹³C magnetic resonance spectroscopy (MRS) in conjunction with the administration of ¹³C labeled glucose provides a unique tool to assess cerebral metabolism noninvasively. For instance, following the infusion of [1-¹³C] or [1,6-¹³C₂] glucose, ¹³C MRS can dynamically monitor the ¹³C label incorporation into desired carbon positions of MR detectable metabolites, e.g., glutamate (Glu) and glutamine (Gln). Then, cerebral metabolic fluxes can be quantitatively derived from experimentally measured ¹³C-label time courses using a mathematical model.^{1, 2, 3}

Mice have been widely used for genetic modification to investigate the pathology of respective human diseases. The use of ¹³C MRS studies on transgenic mouse models may provide unique insights into the pathologies of various brain diseases. For instance, the pathogenesis of Huntington's disease may involve the impairment of glutamate-glutamine cycling⁴ assessed uniquely by ¹³C MRS. In vivo ¹³C MRS is commonly conducted in the human⁵ and rat brain.⁶ However, for the mouse brain, the intrinsically low sensitivity of ¹³C MRS and the small brain size render the measurement of ¹³C labeling into metabolites in vivo challenging. Hence, most ¹³C MRS studies in mouse brain were achieved by using brain extracts.^{7, 8} One in vivo ¹³C study in mouse brain, using direct ¹³C MRS by image selected in vivo spectroscopy localization with distortionless enhancement by polarization transfer, reported time courses of C4 and C3 of Glu and Gln.⁹ However, the cerebral metabolic fluxes remain to be determined.

In addition, a time-resolved isotopic enrichment of plasma glucose is typically required as an input function to derive metabolic fluxes from ¹³C labeled time courses using a given metabolic model in a ¹³C labeled glucose infusion study. However, frequent blood sampling in small animals such as mouse during ¹³C NMR experiment is hardly feasible due to the small blood volume of the mouse and rather big dead volume in the catheter from the center of the magnet to the entrance of the magnet bore.

Relative to direct ¹³C MRS techniques, indirect detection of ¹³C through ¹H combined with [1,6-¹³C₂] or [U-¹³C₆] glucose administration can improve the sensitivity and allow the direct measurement of fractional enrichments (FEs). However, the distinct separation of ¹³C labeling in Glu and Gln using indirect detection requires sufficiently high spectral resolution.¹⁰ Recently, a short-echo-time (TE) full-sensitivity ¹H-[¹³C] NMR sequence was proposed and validated at a high magnetic field strength of 14.1 T, which provides high spectral resolution and signal sensitivity.¹¹

Therefore, the aim of the present study was to use the short-TE full-sensitivity ¹H-[¹³C] MRS¹¹ combined with high magnetic field (i.e., 14.1 T) and infusion of [U-¹³C₆] glucose to measure the time-resolved ¹³C labeling of individual metabolites with a focus on glutamate and glutamine, to assess metabolic fluxes in vivo in the mouse brain.

Materials and methods

Animal Preparation

Male ICR-CD mice (33.4±3.7 g, n=5, Charles River, L'Arbresle Cedex, France) were housed in standard cages, fasted over 7 hours with free access to water and anesthetized using isoflurane (1% to 2%) mixed with O₂. One femoral vein was cannulated for the infusion of [U-¹³C₆] glucose (Sigma-Aldrich, St. Louis, MO, USA). After giving a bolus of a 20% (w/v) 99%-enriched [U-¹³C₆] glucose solution at an exponentially decaying rate over 5 minutes, a 67%-enriched [U-¹³C₆] glucose solution was infused continuously for up to 3 hours.^{12, 13} The respiration rate was continuously monitored (SA Instruments Inc., Stony Brook, NY, USA) and maintained in the range of 80 to 110 breath per minute by adjusting the amount of isoflurane in the O₂ to sustain mouse under physiological condition, i.e., pH=7.3 to 7.4, PaCO₂=30 to 42 mm Hg and PaO₂>100 mm Hg (data not shown), all of which allowed sustaining normal functional activity and normal cerebral blood flow.^{14, 15} The mouse was placed in a holder (RAPID Biomedical GmbH, Rimpar, Germany) and the head was stereotaxically fixed. Body temperature was measured by a rectal thermosensor and maintained at 36.0±0.5°C with the circulation of heated water. All animal preparation procedures were performed in accordance with local and federal guidelines (EXPANIM, Expérience sur animaux-SCAV, Service de la consommation et des affaires vétérinaires, Switzerland) and were approved by the Veterinary Office of Canton de Vaud.

In Vivo NMR Spectroscopy

Magnetic resonance spectroscopy experiments were performed in a 14.1 T magnet with a 26cm horizontal bore (Agilent Technologies, Palo Alto, CA, USA) using a homebuilt geometrically decoupled ¹H quadrature surface coil (13 mm diameter) and a linearly polarized ¹³C coil (10 mm diameter) as a transceiver. Images for voxel positioning were obtained using a multislice fast spin-echo sequence in sagittal and coronal planes (effective TE=54 ms, repetition time=4,000 ms, echo train length=8, average=4, slice thickness=0.6 mm, slices=25, field of view=20 × 20 mm², data matrix=256 × 256). ¹³C decoupled ¹H MR spectra were acquired using a previously proposed full signal intensity ¹H-[¹³C] NMR sequence ‘SPECIAL-BISEP' (Supplementary Figure S1), ¹¹ which was a hybrid sequence composed of SPECIAL (SPin ECho, full Intensity Acquired Localized spectroscopy) localization part¹⁶ and a preceding ¹³C editing block based on the inversion B₁-insensitive spectral editing pulse (BISEP), including a segmented 0^o BIR-4 pulse with two pulse interval delays (τ=1/2J) in the ¹H channel and an adiabatic full passage (AFP) in the ¹³C channel centered at the central segment of the 0^o BIR-4 pulse.^{17 13}C editing was achieved by turning on and off the AFP pulse in the ¹³C channel on alternate scans, and then the edited spectrum (i.e., ¹³C AFP off–¹³C AFP on) contained ¹³C coupled ¹H resonances. Outer volume suppression and water suppression with VAPOR (VAriable Pulse power and Optimized Relaxation delays)¹⁸ were applied before SPECIAL-BISEP. B₀ inhomogeneity was optimized using first- and second-order shimming with FAST(EST)MAP,^{19, 20} resulting in water linewidth of 24 to 28 Hz for a volume of 60 μL (5 × 4 × 3 mm³) containing primarily cerebral cortex and striatum (Figure 2). Adiabatic ¹³C decoupling¹⁰ was applied during 145 ms acquisition time. Spectra with ¹³C AFP off and on were acquired with 8-scan blocks in an interleaved mode (TE=2.8 ms, repetition time=4 seconds) during the whole infusion experiment.

Figure 2.

Representative coronal and sagittal fast spin-echo images of the mouse brain with the volume of interest (VOI) for magnetic resonance spectroscopy (MRS) measurement. Averaged nonedited (A) and edited (B) ¹H magnetic resonance (MR) spectra acquired in the mouse brain during first hour of [U-¹³C₆] glucose infusion (VOI=60 μL, nt=960, no apodization was applied). Ala, alanine; Asp, aspartate; GABA, γ-aminobutyric acid; Glc, glucose; Gln, glutamine; Glu, glutamate; Glx, Glu+Gln; Lac, lactate; MM, macromolecules; NAA, N-acetyl aspartate; tCr, total creatine.

Data Analysis

After Fourier transformation and frequency correction, every four blocks (64 averages) of spectra were summed and quantified with LCModel (Stephen Provencher Inc., Oakville, ON, Canada).²¹ Nonedited ¹H spectra acquired with ¹³C AFP off, which contain ¹H resonances coupled to both ¹³C and ¹²C nuclei, were quantified with a standard ¹H MRS basis set containing a measured macromolecular baseline²² and simulated metabolite spectra.²³ Another basis set including simulated metabolite spectra of ¹H resonances coupled to N-acetyl aspartate (NAA) C6, Glu (C2, C3, and C4), Gln (C2, C3, and C4), γ-aminobutyric acid (GABA) (C2, C3, and C4), aspartate (Asp) (C2 and C3), glucose (C2-C6), total creatine (C2 and C3), lactate (Lac) C3, and alanine (Ala) C3 was used to quantify ¹³C edited MR spectra. All metabolites concentrations were calculated using total creatine as an internal reference by assuming its concentration of 8 μmol/g.²⁴

Metabolic Modeling

In the metabolic reaction chain of glucose utilization in brain tissue, pyruvate (Pyr) is the last intermediate of the glycolysis and is not only a precursor for oxidative metabolism in the tricarboxylic acid (TCA) cycle, but also of lactate and alanine. Given the strong activity of lactate dehydrogenase allowing fast exchange between pyruvate and lactate as compared with the TCA cycle rate,^{25, 26} the ¹³C turnover of LacC3 closely follows that of PyrC3 in terms of ¹³C FE. On the time scale of the ¹³C turnover of the amino acids labeled through the TCA cycle, the FE of LacC3 can therefore be used as an input function of the metabolic system, as a substitute to the nonmeasurable pool of pyruvate. The LacC3 turnover curve was thus smoothed by being fitted with an exponential function FE(t)=(a·t+b)(1−e^−c·t) and used as the input function for the metabolic model. Cerebral metabolic fluxes were explored by fitting a one-compartment metabolic model of mitochondrial metabolism (Figure 1A)^{2, 27} to the measured ¹³C labeling time courses of metabolites (i.e., GluC4, GlnC4, GluC3+GlnC3 (GlxC3) and AspC2). The curve fitting process was undertaken in MATLAB (Version 7.11, The MathWorks, Inc., Natick, MA, USA) using a standard built-in ordinary differential equation solver and a modified Levenberg-Marquardt nonlinear regression method. The fitting procedure was weighted with the square root of the inverse of the variance of the experimental noise, to compensate for the different precisions in the measurement of the turnover curves.

Figure 1.

(A) Scheme of the mathematical model used to describe ¹³C label incorporation into cerebral metabolites during ¹³C labeled glucose infusion. ¹³C label derived from ¹³C labeled glucose is incorporated into C3 of pyruvate. Through pyruvate dehydrogenase ¹³C label enters the tricarboxylic acid (TCA) cycle and labels C4 position of 2-oxoglutarate (OG) in the first TCA cycle turn. Via transmitochondrial transport OG exchanges with cytosolic glutamate (Glu) which gets labeled at position C4. In the second turn of TCA cycle, OG C3 receives the label from OG C4 and Glu C3 gets further labeled. Aspartate (Asp) gets labeled via the transmitochondrial exchange with TCA cycle intermediate oxaloacetate (OAA). C4 and C3 positions of glutamine (Gln) accumulate labels from Glu C4 and C3 through Glu-Gln cycle. V_TCA, the TCA cycle rate; V_x, exchange flux between mitochondrial TCA cycle intermediates and cytosolic glutamate; V_gln, glutamate-glutamine exchange rate; K_dil, the dilution factor of the lactate (Lac)/pyruvate pool induced by uptake of unlabeled plasma lactate. (B) Scheme of the metabolic model for the ¹³C labeling of AlaC3 and LacC3 from ¹³C labeled glucose. V_Lac, lactate conversion rate; V_Ala, alanine conversion rate; CMR_glc, cerebral metabolic rate of glucose consumption; V_dil, dilution flux in the lactate pool induced by uptake of unlabeled plasma lactate. All subscript numbers represent ¹³C labeled carbon position.

In a second step, the ratio of lactate conversion rate V_Lac and alanine conversion rate V_Ala was obtained by fitting time courses of AlaC3 and LacC3 to a compartmental model (Figure 1B) of the chemical reactions involving pyruvate in brain glucose metabolism. In this model, the FE of glucose is assumed to follow a step function at a level of 67% after the start of tracer infusion, as previously shown using this type of infusion protocol.^{12, 13} The cerebral metabolic rate of glucose consumption (CMR_glc) was fixed to 0.5 μmol/g per minute, which is a representative value for brain glucose metabolism in anesthetized mice, as found in this study and in the literature.⁷ In the case of the considered model, with a bidirectional flux with the same pyruvate pool, the FE of Lac is expected to be very close to the FE of Ala, since the dilution flux in lactate also dilutes the labeling in Ala through the exchange with Pyr. Note that the calculated FE of AlaC3 was slightly higher than that of LacC3, which may be ascribed to a potential underestimation of the small Ala concentration given its overlap with macromolecule resonances at 1.4 ppm. Therefore, the FE turnover curve of AlaC3 at steady state was scaled with the FE of LacC3, based on the ratio between FELacC3/FEAlaC3 at steady state found in brain extracts in Duarte et al.¹³ The errors of all adjusted metabolic fluxes were evaluated by Monte–Carlo simulation.

The equations describing the mathematical modeling approaches used in this study are reported in appendix, as Supplementary information.

Results

Nonedited (Figure 2A) and edited ¹H MR spectra (Figure 2B) acquired from the volume of interest of 60 μL in the mouse brain using SPECIAL-BISEP show good spectral quality at 14.1 T in this study, such as high spectral resolution, e.g., the separation of GluC4 and GlnC4, and the absence of contamination signal from extraneous lipids, which allows the measurement of lactate and alanine. In the ¹³C edited spectra, ¹³C labeled ¹H resonances in the mouse brain can be observed in LacC3, AlaC3, GluC4, GlnC4, GlxC3, GlxC2(GluC2+GlnC2), AspC3, AspC2, GABAC3, GABAC2, and Glc(C2-C6) (Figure 2B).

To show the dynamic incorporation of ¹³C label into individual carbon positions of metabolites, edited spectra containing ¹³C-coupled ¹H resonances were summed over 5 minutes for 90 minutes after the start of [U-¹³C₆] glucose infusion (Figure 3). Note that ¹³C labeling into Glc and Lac can already be observed during the second 5-minute acquisition after the start of infusion. Subsequently, carbon positions of Ala, Glu, Gln and GABA were labeled, respectively (Figure 3).

Figure 3.

Stack plot of ¹³C labeling incorporation into cerebral metabolites during the first 90 minutes of [U-¹³C₆] glucose infusion with time resolution of 5 minutes. No baseline correction and water signal removal were applied.

To determine the total concentration of metabolites (¹³C+¹²C) and the concentrations of ¹³C labeled metabolites, nonedited and edited ¹H spectra were analyzed using LCModel. The pool sizes of glutamate (9.1±0.5 μmol/g, mean±s.e.m.), glutamine (3.4±0.1 μmol/g), and aspartate (2.1±0.2 μmol/g) were measured simultaneously from the nonedited ¹H spectra. The average time courses of ¹³C labeled GluC4, GlnC4, GlxC3, AspC2 and GABA(C2, C3, and C4) concentration were obtained from the edited spectra with a temporal resolution of 4.3 minutes during 180-minute infusion of [U-¹³C₆] glucose (Figure 4).

Figure 4.

Average time courses of (A) GlnC4 (□) and GluC4 (○), (B) GlxC3 (○) and AspC2 (□), (C) LacC3, and (D) GABA C2(*), C3(□), and C4(○) during 180-minute [U-¹³C₆] glucose infusion (n=5). LacC3 time course was fitted with an exponential function. The fitting curves of GluC4, GlnC4, GlxC3, and AspC2 were the best fit obtained from one-compartment metabolic modeling. Spectra were average over 17 minutes for time courses of GABA C2, C3, and C4. Polynomial function was used to fit GABA C2, C3, and C4 for visualization purpose. Asp, aspartate; Gln, glutamine; Glu, glutamate; Lac, lactate.

To determine the TCA cycle flux (V_TCA=1.05±0.04 μmol/g per minute, mean±s.d.), the exchange flux between 2-oxoglutarate (mitochondrial TCA cycle intermediate) and glutamate (cytosolic amino acid) (V_x=0.48±0.02 μmol/g per minute), the Glu-Gln exchange rate (V_gln=0.20±0.02 μmol/g per minute) and the dilution factor (K_dil=0.82±0.01), the average time courses of GluC4, GlnC4, GlxC3, and AspC2 were fitted by a one-compartment model of mitochondrial metabolism using fitted LacC3 time course as the input function (Figures 4A–4C). The comparison of metabolic fluxes obtained in mouse (ex vivo), rat, and human (in vivo) brains is summarized in Table 1.

Table 1. Metabolic fluxes (μmol/g per minute, mean±s.d.) obtained in rat, human, and mouse brain using ¹³C MRS.

	VTCAn	VTCA	Vx	VNT	Vg	VPC	Anesthetic	VOI location	Reference
Rata	0.44±0.01	0.74±0.02	0.76±0.07b	0.12±0.01	0.23±0.02	0.07±0.01	α-Chloralose	Whole brain	Duarte et al13
Ratc	0.37±0.06	0.73±0.06	0.46±0.05b	0.15±0.01	0.27±0.02	0.09±0.01	α-Chloralose	Cortex	Lanz et al37
Ratd	−	0.71±0.02	0.88±0.08	−	−	−	α-Chloralose	Whole brain	Henry et al12
Rata	0.79±0.15	0.93±0.15	9.6±6.9	0.31±0.07	0.04e	0.096e	Halothane	Gray matter	de Graaf et al6
Rata	0.20±0.11	0.31±0.11	6.4±3.8	0.02±0.04	0.01e	0.096e	Halothane	White matter	de Graaf et al6
Rata	0.42±0.09	0.54±0.09	6.7±4.1	0.18±0.12	0.02e	0.096e	Halothane	Subcortex	de Graaf et al6
Humand	−	0.80±0.10	−	−	−	−	−	Gray matter	Mason et al39
Humand	−	0.17±0.01	−	−	−	−	−	White matter	Mason et al39
Humand	−	0.73±0.19	57±26	0.47±0.70	−	−	−	Occipital-parietal	Mason et al40
Humand	0.57±0.06	0.72±0.07	0.57±0.19	0.17±0.05	0.06±0.04	0.09±0.02	−	Visual cortex	Gruetter et al38
Mousef	0.91±0.05	1.06±0.07	99.96	0.36±0.02	0.13±0.04	0.02±0.03	Halothane	Cortex extract	Tiwari et al7
Mousef	0.75±0.05	0.83±0.07	99.96	0.18±0.01	0.02±0.04	0.06±0.03	Halothane	Striatum extract	Tiwari et al7
Mouseg	−	1.05±0.04	0.48±0.02	0.20±0.02h	−	−	Isoflurane	Cortex and striatum	Current study

Abbreviations: MRS, magnetic resonance spectroscopy; TCA, tricarboxylic acid; V_TCAⁿ, rate of neuronal TCA cycle; TCA cycle flux V_TCA=V_TCAⁿ+V_g+V_PC; V_g, rate of glial TCA cycle; V_PC, rate of pyruvate carboxylase; V_NT, apparent rate of glutamate neurotransmission; VOI, volume of interest.

^a[1,6-¹³C₂] glucose infusion.

^bExchange rate between 2-oxoglutarate and glutamate in neurons.

^c[2-¹³C] acetate infusion.

^d[1-¹³C] glucose infusion.

^eAssumed values.

^fEx vivo, [1,6-¹³C₂] glucose and [2-¹³C] acetate infusion.

^g[U-¹³C₆] glucose infusion.

^hV_NT was set to V_gln in the one-compartment model.

To determine the ratio of V_Lac/V_Ala (10±2), the turnover curves of LacC3 and AlaC3 (Figure 5A) were fitted, assuming a glucose utilization CMR_glc of 0.5 μmol/g per minute.⁷ To investigate the effect of the assumed CMR_glc value on the assessment of V_Lac/V_Ala ratio, V_Lac/V_Ala ratios were calculated by varying CMR_glc values (Figure 5B). For CMR_glc values lower than 0.45 μmol/g per minute, the simulated curves failed to describe the initial rising portion of the measured LacC3 enrichment curve. This value was therefore used as a minimum in the analysis. Error bars represents the standard deviation determined by Monte–Carlo simulations. The V_Lac/V_Ala ratio was found to be between 6.6 and 11.6 for the considered CMR_glc range. For an assumed value of CMR_glc of 0.5 μmol/g per minute,⁷ the ratio was 10±2. The pool size of Ala and Lac may vary due to physiological conditions, such as the higher plasma glucose level during [U-¹³C₆] glucose infusion. From the ¹H MRS data, we observed a net production of lactate (0.01 μmol/g per minute) and in a smaller extent of alanine (0.002 μmol/g per minute). When this metabolic nonsteady-state condition was added to the metabolic model, i.e., considering linearly increasing pool sizes, no statistically significant difference was found for the V_Lac/V_Ala ratio (data not shown).

Figure 5.

(A) Average time courses of AlaC3 and LacC3 (n=5) were shown and fitted with a mathematical model to calculate V_Lac/V_Ala. (B) The effect of the assumed cerebral metabolic rate of glucose consumption (CMR_glc) value on the ratio of V_Lac/V_Ala. Ala, alanine; FE, fractional enrichment; Lac, lactate.

Discussion

The present study measures for the first time in vivo in the mouse brain the ¹³C label incorporation into brain metabolites using ¹H-[¹³C] MRS with the concomitant determination of the TCA cycle flux V_TCA, the exchange flux between 2-oxoglutarate and glutamate V_x, the Glu-Gln exchange rate V_gln, the pyruvate dilution factor K_dil and the ratio V_Lac/V_Ala.

¹H-[¹³C] Magnetic Resonance Spectroscopy in the Mouse Brain

Due to the small size of the mouse brain and the intrinsic low sensitivity of ¹³C MRS, several strategies were used in this study to enhance sensitivity of the measurement. The use of [U-¹³C₆] glucose infusion increased the sensitivity of the measurement by two-fold relative to [1-¹³C] glucose by labeling both molecules of [3-¹³C] pyruvate from [U-¹³C₆] glucose through glycolysis. In addition, as an alternative to direct-detected ¹³C MRS,⁹ detection of ¹³C labeling through coupled ¹H nuclei intrinsically increased the experimental sensitivity. The application of SPECIAL-BISEP¹¹ at ultra-high magnetic field of 14.1 T further increased the quality of in vivo spectra from the mouse brain in terms of signal-to-noise ratio, spectral resolution, and localization performance. Moreover, the nonedited ¹H spectra allow the straightforward measurement of metabolites pool size in vivo, which is especially practical for recording the pool size of those metabolites that may vary during the infusion experiment.^{13, 28} Although the application of ¹H-[¹³C] MRS at 14.1 T improved the spectral dispersion of GluC4 and GlnC4 (Figure 2B), the separation of labeling in C2 and C3 of Glu and Gln was not complete.

Input Function for Metabolic Modeling

As a major energy source of the brain, glucose is first transported into brain across the blood–brain barrier and then degraded into pyruvate by the glycolytic pathway. Pyruvate can then be transported into mitochondria and oxidized in the TCA cycle. It can also be converted to lactate through lactate dehydrogenase, which has a higher activity than that of the pyruvate dehydrogenase.²⁵

Therefore, following the glucose metabolic pathway, several candidates, e.g., plasma glucose, brain glucose, lactate, and pyruvate could serve as an input function in the mathematic modeling to derive metabolic fluxes. However, due to small pool size of pyruvate, it can only be detected with sensitivity enhancement using dynamic nuclear polarization.²⁹

To date, when using direct ¹³C MRS, it has been hardly feasible to measure the time-resolved FE of metabolic pools such as glucose or lactate in vivo. Therefore, a time-resolved FE of plasma glucose was commonly obtained from blood sampling during the MRS experiment and then used as an input function to derive metabolic fluxes using a given metabolic model in a ¹³C labeled glucose infusion study.^{13, 30} However, frequent blood sampling during ¹³C NMR experiments remains challenging in the mouse due to its small blood volume. Recently, one study showed the blood sampling during ¹³C MR experiments in young rat (~100 g) by using a specifically designed animal holder.³¹

Alternatively, the measure of FE of brain glucose using ¹H-[¹³C] MRS has been reported in the rat brain;¹⁰ however, this remains challenging as it heavily relies on excellent water suppression and shimming performance. The optimization of field inhomogeneity in mouse brain is highly demanding on shimming capacity because of an increased susceptibility gradient comparing with rat brain²⁴ and becomes even more challenging at 14.1 T compared with 9.4 T.³² Therefore, in the current study, the reliable measurement of C1 glucose at 5.2 ppm was not applicable in mouse brain. Other ¹H resonances coupled to C2-C6 of glucose present a number of peaks (3.3 to 3.8 ppm) overlapping with macromolecule baseline and many metabolites. This results in the underestimation of total glucose concentration in the nonedited spectra and overestimation of FE of brain glucose. Therefore, brain glucose might not be best candidate as the input function in this study.

However, the mouse brain contains higher level of lactate (3 to 5 mmol/L)^{24, 33} across four brain regions relative to rats (1 to 2 mmol/L)³⁴ and human (~1 mmol/L),³⁵ leading to direct time course measurements of lactate FE using ¹H-[¹³C] MRS (Figure 4C). Compared with plasma or brain glucose, which is several biochemical steps further from the TCA cycle, the lactate pool is likely to represent closely the FE of pyruvate. In addition, lactate can also be transported into the brain via monocarboxylate transporters and utilized as brain fuel.³⁶ The use of lactate as an input function could take into account the potential ¹³C labeling from the blood. Therefore, the brain lactate signal is the most favourable and straightforward candidate for input function in the mouse brain.

Metabolic Fluxes

The value of TCA cycle flux (V_TCA=1.05±0.04 μmol/g per minute) representing essentially glutamatergic neuron activity is in agreement with the results obtained from mouse brain extracts (Table 1).⁷ As shown in Table 1, mouse brains showed a higher TCA cycle flux compared with human and rat brain, suggesting distinct kinetics for the mouse. The Glu-Gln exchange rate of 0.20±0.02 μmol/g per minute for cerebral cortex and striatum falls in the range of reported ex vivo values of 0.18±0.01 and 0.36±0.02 μmol/g per minute in striatum and cortex.⁷ The transmitochondrial flux V_x (0.48±0.02 μmol/g per minute) summarizing the exchange between 2-oxoglutarate and cytosolic glutamate was measured for the first time in the in vivo mouse brain and was found on the same order of magnitude as the TCA cycle rate, similarly to previous studies in rats^{13, 37} and humans.³⁸

In this study, we applied a one-compartment model analysis of brain energy metabolism based on the turnover curves GluC4, GlnC4, GlxC3, and AspC2. Since brain tissue is composed of many different cell types with distinct roles in energy metabolism, it would be desirable to develop the model into a more complex scheme closer to the tissue biochemical complexity. A next step would be to consider a two-compartment approach^{1, 13, 37} to distinguish between neuronal and astrocytic metabolism in the ¹³C labeling of glutamate, glutamine and aspartate. However, to characterize properly the activity of astrocytic and neuronal TCA cycles separately, the separate acquisition of the positions C4 and C3 of glutamate and glutamine is at least necessary,¹ when infusing [1,6-¹³C]glucose. The main reason is the different labeling pattern of the position C3 induced in the astrocytic compartment by the astrocytic-specific pyruvate carboxylase. This cell-specific labeling pattern enables to decouple the effects of neuronal and astrocytic TCA activity on the observed turnover curves by a glial-specific dilution of glutamine C3. In the present study using [U-¹³C]glucose and limited to the measurement of GlxC3 due to spectral overlap, this cell-specific labeling pattern was not observed and the use of two-compartment modeling approach leads to strong correlations between the determined metabolic rates in the neuronal and astrocytic compartments (data not shown). This underdetermination can either be solved by fixing some metabolic rates using a priori knowledge or decreasing the level of complexity of the model, i.e., reducing it to a one-compartment approach. Future studies enabling the separate measurement of the C3 and C2 positions of glutamate and glutamine are expected to give a cell-specific insight in cerebral oxidative metabolism in the mouse.

Metabolic concentrations, required for the determination of the metabolic fluxes, are commonly varying strongly with cerebral regions as shown in the human³⁵, rat,³⁴ and mouse²⁴ brains. Therefore, it is crucial to determine the metabolic pool size in the studied volume of interest (VOI). One of the advantages of ¹H-[¹³C] MRS is the simultaneous measurement of metabolic pool sizes from nonedited spectra. The metabolite values measured in the VOI containing mainly cerebral cortex and striatum in this study are in the range of those reported in vivo in the mouse striatum and cortex.²⁴

Metabolic Flux Ratio of V_Lac/V_Ala

In the present study, we report for the first time the in vivo measurement of ¹³C label incorporation into Lac and Ala in the mouse brain. A separate metabolic model of the biochemical reactions involving the production of alanine and lactate from pyruvate was therefore developed to enable the characterization of the dynamics of the labeling of alanine and lactate (Figure 1B). The conversion rate between pyruvate and alanine through alanine aminotransferase was expected to be on the same order of magnitude as the conversion between pyruvate and lactate. Therefore, the strategy of substituting the unknown FE time course of PyrC3 by the measured FE turnover curve of LacC3 as used for the modeling of the TCA cycle activity was not applicable. An alternative approach would be to work with another upstream substrate, such as the plasma glucose FE as an input function.

With the applied glucose infusion protocol, the FE of glucose in blood has been shown to reach steady state within 5 minutes.¹³ Blood glucose FE was therefore modeled as a step function. However, the biochemical pathways and intermediates from plasma glucose to cerebral pyruvate are numerous (i.e., glucose transport, phosphorylation, and glycolysis). All these intermediate labeling pools potentially delay the turnover of pyruvate as compared with blood glucose and therefore may affect the kinetics of ¹³C uptake in alanine and lactate. Since the glycolytic process is expected to be relatively slow compared with the conversion between pyruvate and lactate or alanine, this effect may not be negligible. Therefore, the separate evaluation of CMR_glc and the labeling fluxes V_Ala and V_Lac is not possible without measuring the FE of PyrC3. However, since this labeling delay due to glycolysis affects pyruvate labeling, which is the last precursor for lactate and alanine, the dynamics of AlaC3 and PyrC3 turnover curves carries information on the relative conversion rates V_Ala and V_Lac, which could be determined in this study, as confirmed by Monte–Carlo simulation (V_Lac/V_Ala=10±2).

In this study, a representative CMR_glc of 0.5 μmol/g per minute⁷ was assumed to avoid potential mathematical underdetermination of the metabolic model. Further analysis (Figure 5B) showed that assumed CMR_glc in a range of 0.45 to 2 μmol/g per minute resulted in a ratio V_Lac/V_Ala from 6.6 to 11.6. Note that the gradual increase in lactate and alanine concentrations measured by ¹H MRS did not affect the determination of the ratio V_Lac/V_Ala.

We conclude that high quality ¹H-[¹³C] NMR spectra can be acquired from a volume as small as 60 μL in mouse brain in vivo at 14.1 T. This allows to obtain ¹³C labeling turnover curves of GluC4, GlnC4, GlxC3, and for the first time, AspC2, AlaC3, GABA (C2, C3, and C4) as well as LacC3 that offers precursor information to determine a number of metabolic fluxes without additional blood sampling. This study opens the prospect of studying transgenic mouse models of brain pathologies.

The authors declare no conflict of interest.