Quantitative assessment of brain glucose metabolic rates using in vivo deuterium magnetic resonance spectroscopy

Abstract

Quantitative assessment of cerebral glucose consumption rate (CMR_glc) and tricarboxylic acid cycle flux (V_TCA) is crucial for understanding neuroenergetics under physiopathological conditions. In this study, we report a novel in vivo Deuterium (²H) MRS (DMRS) approach for simultaneously measuring and quantifying CMR_glc and V_TCA in rat brains at 16.4 Tesla. Following a brief infusion of deuterated glucose, dynamic changes of isotope-labeled glucose, glutamate/glutamine (Glx) and water contents in the brain can be robustly monitored from their well-resolved ²H resonances. Dynamic DMRS glucose and Glx data were employed to determine CMR_glc and V_TCA concurrently. To test the sensitivity of this method in response to altered glucose metabolism, two brain conditions with different anesthetics were investigated. Increased CMR_glc (0.46 vs. 0.28 µmol/g/min) and V_TCA (0.96 vs. 0.6 µmol/g/min) were found in rats under morphine as compared to deeper anesthesia using 2% isoflurane. This study demonstrates the feasibility and new utility of the in vivo DMRS approach to assess cerebral glucose metabolic rates at high/ultrahigh field. It provides an alternative MRS tool for in vivo study of metabolic coupling relationship between aerobic and anaerobic glucose metabolisms in brain under physiopathological states.

Introduction

Cerebral glucose metabolism is of importance for brain function since glucose is the major fuel for energy production in the form of ATP, which is essential to maintain electrophysiological activity for neuronal firing, signaling and cellular functionality. Quantitative assessment of cerebral glucose consumption rate (CMR_glc) and the associated major metabolic fluxes, such as the TCA cycle flux (V_TCA), α-ketoglutarate/glutamate exchange and oxygen consumption rates, is crucial for understanding neuroenergetics under various physiological and pathological conditions. However, simultaneous measurement of both CMR_glc and V_TCA has been challenging due to the complexity of brain glucose metabolisms and limitations of current technology.

For decades, in vivo ¹³C MRS in combination with ¹³C-labeled substrate infusion has been the unique tool to investigate brain metabolic fluxes (e.g. V_TCA) and neurotransmission cycling rates noninvasively by analyzing ¹³C-labeled glutamate time courses using ¹³C-labeled glucose or acetate infusion and complex metabolic modeling.^1–6 To date, despite of its extensive use, it remains a challenging technique owing to relatively low MR sensitivity and the requirement of ¹³C-¹H dual channels, ¹H NOE preparation and decoupling. In addition, except for a small number of studies using ¹³C-labeled glucose signal for modeling glucose transport,⁷ the majority of ¹³C MRS studies do not utilize the ¹³C-labeled glucose signal, the CMR_glc measurement is indirectly estimated based on an assumed coupling relationship between glycolysis and the measured V_TCA.^4,5 Nevertheless, such relationship could change substantially or become decoupled in diseased tissues, e.g. in brain tumor and stroke.

Another gold-standard imaging modality for regional CMR_glc assessment has been the ¹⁸F-fluorodeoxyglucose (¹⁸FDG)-PET imaging method established since 1980s.^8–11 It has been widely used not only in animal but also in human brains under normal or diseased conditions,^12–19 though it requires administration of radioactive agent. Since the ¹⁸FDG-PET can only measure the ¹⁸FDG uptake in the cytosol, it is incapable of providing mitochondrial oxidative metabolic information including V_TCA. Therefore, to obtain both CMR_glc and V_TCA for comprehensive evaluation of cerebral glucose metabolisms, researchers commonly perform ¹⁸FDG-PET and in vivo ¹³C MRS measurements separately using two imaging modalities. It is critical to exploit emerging metabolic imaging methods for undertaking the technical challenge.

Deuterium (²H) contains one proton and one neutron, and it is a stable and non-radioactive isotope of hydrogen with spin quantum number of 1 (thus, dominated by quadrupolar relaxation mechanism) and a low natural abundance of 0.0156% on earth. Having a similar range of chemical shift (in ppm) to proton NMR, ²H NMR spectra are popular in the solid state NMR research to study lipid membrane, protein and peptide structure or dynamics due to its relatively small quadrupole moment.²⁰ High-resolution ²H NMR has been applied in elucidating metabolic pathways in gluconeogenesis through plasma samples²¹ and in fermentation.^22,23 However, except the first preliminary attempt by Mateescu et al.,²⁴ there has been no other in vivo deuterium magnetic resonance (DMR) spectroscopy (DMRS) study reported that investigates glucose metabolism and mitochondrial respiration. The advantage of in vivo DMRS becomes obvious when combining with ²H-isotope labeled glucose infusion, which results from its excellent chemical specificity and informative dynamic spectra in that labeling of the substrate, i.e. glucose, and intermediate metabolites through glycolysis and oxidative phosphorylation pathways can be distinguished, especially at high/ultrahigh field. These merits allow for the possibility of isotope kinetic analysis and eventually for quantification of glucose metabolisms. Thus, in view of the benefits it can bring, a well-developed quantitative in vivo DMRS approach is highly desirable.

In this study, a novel in vivo DMRS approach that enables simultaneous measurement of CMR_glc and V_TCA is proposed and examined in rat brain at 16.4 Tesla (T). Briefly, following a brief intravenous infusion of deuterated glucose, the dynamic labeling of glucose, glutamate/glutamine (Glx) and water in the brain tissue can be monitored using their well-resolved resonance signals in dynamic ²H spectra with excellent sensitivity and temporal resolution, which can be attributed to the very short longitudinal relaxation time (T₁) of ²H-labeled molecules as compared to that of ¹H or ¹³C spins, thus, allowing more signal averaging within the same sampling time. A new kinetic model incorporating glycolysis, TCA cycle and α-ketoglutarate/Glx exchange is then developed. By least-square fittings of the model with the dynamic DMRS glucose and Glx data, CMR_glc and V_TCA can be concurrently determined. The sensitivity and reliability of this approach in assessing cerebral glucose metabolism and its change are also evaluated in this work by using two different brain conditions: 2% isoflurane (deep anesthesia) vs. morphine infusion (analgesia).

Materials and methods

Animal preparation and experimental protocol

Eight male Sprague Dawley rats (body weight: 350 ± 47 g) were anesthetized with 2% isoflurane in a mixture of O₂ and N₂O gases (∼2:3) and prepared for in vivo DMRS scans. The rat femoral arteries and veins were catheterized for blood sampling, physiological monitoring, deuterated glucose and/or morphine infusion. Rectal temperature was maintained at 37 ± 1℃ using heated circulating water. Blood gases were sampled for monitoring physiological conditions. The arterial blood pressure and heart rate were monitored during the entire experiment. The end-tidal CO₂ was monitored (Capnomac Ultima; Finland) and maintained at normal levels (3–4%). The animals were maintained under physiological conditions throughout the measurements. The animal procedures and experiments were conducted in accordance with the National Research Council’ s Guide for the Care and Use of Laboratory Animals and under the protocols approved by the Institutional Animal Care and Use Committee of University of Minnesota, and were compliance with the ARRIVE guidelines (Animal Research: Reporting in Vivo Experiments).

For each rat, after surgical preparation and a waiting period (∼1 h) for stabilizing animal physiological condition inside the 16.4 T magnet, 10-min baseline ²H spectra (described in the following section) were acquired that followed by 2 min intravenous infusion of D-Glucose-6,6-d₂ (d66, Sigma-Aldrich; 1.3 g/kg body weight and dissolved in 2.5 mL saline) and 120 min continuous DMRS acquisitions (see the experimental protocol as shown in Figure S1). Multiple arterial blood samples were drawn during the DMRS experiment for assessing the dynamics of blood (²H-labeled and non-labeled) glucose levels.

To test the sensitivity of in vivo DMRS method in response to altered metabolic activity, four of the eight rats were switched from deep anesthesia condition using 2% isoflurane to analgesia condition with constant infusion of morphine sulfate (West-Ward, NJ; 25 mg/kg/h), which was started before the onset of baseline DMRS acquisitions. Figure S1 displays the schematic diagram of the experimental protocol used in this study.

Deuterium MR spectroscopy and metabolites quantification

All in vivo MR experiments were conducted at 16.4 T/26 cm horizontal scanner (Varian/VNMRJ, CA) using a passively decoupled radio frequency (RF) dual-coil probe, which includes an oval-shaped single-loop ²H surface coil (10 mm × 20 mm) for DMRS acquisition and a butterfly-shaped ¹H surface coil for shimming and anatomical images. The ²H RF probe was placed over the rat head between the eyes and ears and tuned to 107 MHz for acquiring the ²H MRS data. The ²H RF coil size and RF pulse flip angle were optimized to maximize the ²H signals from the rat brain; nevertheless, the partial signal contribution from the surrounding muscle was still inevitable. A single-pulse-acquire sequence was applied to obtain dynamic DMR spectra from the rat brain with the following parameters: 3 kHz spectral width, 512 points for each free induction decay (FID), 0.3 s repetition time with 50 averages (15 s per spectrum) and a total of 520 spectra (130 min in total). Raw FID signals were converted into frequency domain spectra by Fourier transformation using an exponential filter with 13 Hz to enhance the signal-to-noise ratio (SNR).

A spherical phantom containing 5 mM of d66 dissolved in saline solution was also prepared for identifying and assigning the chemical shifts of water (set at 4.8 ppm as a reference) and glucose resonance peaks in DMR spectrum and validating quantification accuracy of metabolites.

All ²H resonance signal integrals (except for the weak lactate resonance signal observed in the healthy rat brain) were fitted using a MATLAB-based program for quantification. The concentrations of deuterated glucose (d66) and labeled Glx in the brain tissue or phantom solution were quantified by normalizing their fitted resonance integrals to that of the baseline water signal, which has the natural abundance ²H concentration of 17.2 mM in water phantom (=55 × 2 × 0.0156% in considering 55 M water concentration and 0.0156% deuterium natural abundance) and 13.7 mM for brain tissue water in considering the water content fraction of 0.8 in the brain tissue.^25,26

The T₁ values of deuterated water, Glx and d66 were measured in rat brains at 16.4 T using the inversion-recovery method under fully relaxed condition with a long pre-inversion delay of 2 s and 16 signal averages. To compensate the RF magnetic field (B₁) inhomogeneity of the ²H surface coil and its implication on the T₁ measurement, a B₁-insensitive hyperbolic secant inversion pulse²⁷ was used followed by magnetization dephasing gradients. Each T₁ value was determined by a three-parameter least-square fitting of a single exponential function to the resonance signals with a total of 10 different inversion-recovery times (0.012–12 s). For comparison, T₁ values of deuterated water and d66 in the phantom solution were also measured. Blood glucose concentration (labeled plus non-labeled ones) was measured with glucose meter (ACCU-CHEK, Roche). Plasma-labeled glucose level was quantified by analyzing high-resolution ²H NMR spectrum obtained at 11.7 T (Varian Inova-500 with 5 mm Broadband probe tuned to ²H frequency) using the standard d66 (20 mM) solution as an internal reference. The results were used to determine the glucose fraction of ²H enrichment (FE) and its dynamic change. Figure S2 demonstrates the representative results from one rat study.

Kinetic model and analysis

Figure 1 shows the ²H-isotope labeling scheme detectable by dynamic DMRS in the brain tissue after d66 infusion and the associated metabolic pathways. Following the infusion, d66 glucose is transported together with non-labeled glucose into the brain and metabolized through glycolysis and oxidative phosphorylation pathways. Dynamics of the cerebral d66 and incorporation of the label into the pools of lactate, glutamate and water could be detected by sequential DMRS acquisitions with adequate SNR in each spectrum. Similar to ¹³C MRS, the ²H NMR signal of glutamate is the result from its relatively high labeling concentration of intracellular glutamate that is in constant exchange with the TCA cycle via α-ketoglutarate (Figure 1). Since it is difficult to resolve the resonances of glutamate and glutamine (labeled through glutamate/glutamine cycle) in the DMR spectrum, Glx served as a combined pool of them in this study. In contrast to ¹³C MRS, in vivo DMRS can also provide new information for assessing ²H labeling dynamics of brain tissue water (Figure 1).

Figure 1.

The ²H-labeling scheme in the brain tissue for dynamic DMRS application using D-Glucose-6,6-d₂ (d66) as an isotopic tracer. Labeling firstly incorporates into pyruvate pool to form [3,3-d₂] Pyruvate through glycolysis, which is then converted to [3,3-d₂] lactate catalyzed by lactate dehydrogenase (LDH). [3,3-d₂] Pyruvate can also be transported into mitochondria and transformed into [2,2-d₂] Acetyl-CoA via pyruvate decarboxylation by pyruvate dehydrogenase (PDH). By entering the TCA cycle, intermediates of [4-d] or [4,4-d₂] Citrate and [4-d] or [4,4-d₂] α-ketoglutarate will be produced, which could exchange with glutamate to generate [4-d] or [4,4-d₂] glutamate. During the following steps of TCA cycle, ²H-labeling may depart from the cycle and exchange with the proton(s) in water molecule to become deuterated water. ‘*’: Pools labeled with ²H; black square boxes: pools to be detected by in vivo DMRS.

Analysis of metabolic activity and kinetics is based on the pre-steady state DMR spectra of glucose and Glx reflecting their ²H labeling dynamics. A new and simplified kinetic model incorporating glycolysis, TCA cycle and α-ketoglutarate/Glx exchange was developed. As shown in Figure 2, five cerebral metabolites plus the blood glucose were involved in this lumped model.

Figure 2.

Simplified kinetic modeling. Symbols: Glc: glucose; Gly: glycogen; L: combined pool for Pyr (pyruvate) and Lac (lactate); K: α-KG, α-ketoglutarate; Glx: combined pool for Glu (glutamate) and Gln (glutamine). V_x stands for the α-KG/Glx exchange rate. V_y is the glycogen synthetic rate. V_out represents an efflux of lactate. ‘*’: ²H-labeled metabolites.

Both of the changes in blood glucose level and plasma ²H-labeled glucose concentration during the post-deuterated glucose infusion period (note that the time point of 5 min after infusion ending was set to zero for performing the kinetic analysis and model regression) were fitted based on exponential decay functions, which were then served as model inputs for further kinetic analysis and quantification.

Ten differential equations (see below) from all five metabolite pools in the brain tissue (two equations for each metabolite: one for total, i.e., non-labeled plus labeled; the other one for labeled only) constituted the mathematical model. By least-square fittings of the model outputs with the time course of ²H labeled brain glucose and Glx concentrations, major metabolic fluxes of CMR_glc and V_TCA and the glycogen synthetic rate (V_y) were determined.

Mass and isotope balance equations describing the metabolic model

(a) Brain glucose [Glc]_brain (labeled plus non-labeled; unit: mM, converted to µmol/g brain tissue by assuming the brain tissue density of 1.1 g/ml)

$$\frac{\text{d}[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}{\mathit{\text{dt}}}=\frac{{\mathit{\text{T}}}_{max}·[\mathit{\text{Glc}}{]}_{\mathit{\text{blood}}}}{{\mathit{\text{K}}}_{\mathit{\text{T}}}+[\mathit{\text{Glc}}{]}_{\mathit{\text{blood}}}+[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}-\frac{{\mathit{\text{T}}}_{max}·[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}{{\mathit{\text{K}}}_{\mathit{\text{T}}}+[\mathit{\text{Glc}}{]}_{\mathit{\text{blood}}}+[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}-{\mathit{\text{CMR}}}_{\mathit{\text{glc}}}$$ (1)

where K_T: half-saturation constant (mM); T_max: maximum transport rate (mM/min); [Glc]_blood: blood glucose level (labeled plus non-labeled, mM); CMR_glc: cerebral metabolic rate of glucose (mM/min, converted to µmol/g brain tissue/min by assuming the brain tissue density of 1.1 g/ml).

(b) Brain labeled glucose ^*[Glc]_brain (unit: mM, converted to µmol/g brain tissue by assuming the brain tissue density of 1.1 g/ml)

$$\frac{\mathit{\text{d}}*[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}{\mathit{\text{dt}}}=\frac{{\mathit{\text{T}}}_{max}·*[\mathit{\text{Glc}}{]}_{\mathit{\text{blood}}}}{{\mathit{\text{K}}}_{\mathit{\text{T}}}+[\mathit{\text{Glc}}{]}_{\mathit{\text{blood}}}+[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}-\frac{{\mathit{\text{T}}}_{max}·*[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}{{\mathit{\text{K}}}_{\mathit{\text{T}}}+[\mathit{\text{Glc}}{]}_{\mathit{\text{blood}}}+[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}-{\mathit{\text{CMR}}}_{\mathit{\text{glc}}}·\frac{*[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}{[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}+{\mathit{\text{V}}}_{\mathit{\text{y}}}·\frac{*[\mathit{\text{Gly}}]}{[\mathit{\text{Gly}}]}-{\mathit{\text{V}}}_{\mathit{\text{y}}}·\frac{*[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}{[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}$$ (2)

where ^*[Glc]_blood: blood labeled glucose level (mM); V_y: glycogen synthetic rate; [Gly] and ^*[Gly]: brain glycogen.

(c) Combined pool for brain pyruvate and lactate, [L] and ^*[L]

$$\frac{\mathit{\text{d}}[\mathit{\text{L}}]}{\mathit{\text{dt}}}=2{\mathit{\text{CMR}}}_{\mathit{\text{glc}}}-[{\mathit{\text{V}}}_{\mathit{\text{out}}}+(2{\mathit{\text{CMR}}}_{\mathit{\text{glc}}}-{\mathit{\text{V}}}_{\mathit{\text{out}}})]=0$$ (3)

$$\frac{\mathit{\text{d}}*[\mathit{\text{L}}]}{\mathit{\text{dt}}}={\mathit{\text{CMR}}}_{\mathit{\text{glc}}}·\frac{*[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}{[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}-(2{\mathit{\text{CMR}}}_{\mathit{\text{glc}}}+{\mathit{\text{V}}}_{\mathit{\text{out}}})·\frac{*[\mathit{\text{L}}]}{[\mathit{\text{L}}]}$$ (4)

where V_out: an efflux of lactate.

(d) Brain α-ketoglutarate (α-KG), [K] and ^*[K]

$$\frac{\mathit{\text{d}}[\mathit{\text{K}}]}{\mathit{\text{dt}}}=(2{\mathit{\text{CMR}}}_{\mathit{\text{glc}}}-{\mathit{\text{V}}}_{\mathit{\text{out}}})+{\mathit{\text{V}}}_{\mathit{\text{x}}}-{\mathit{\text{V}}}_{\mathit{\text{TCA}}}-{\mathit{\text{V}}}_{\mathit{\text{x}}}=0$$ (5)

$$\frac{\mathit{\text{d}}*[\mathit{\text{K}}]}{\mathit{\text{dt}}}=(2{\mathit{\text{CMR}}}_{\mathit{\text{glc}}}-{\mathit{\text{V}}}_{\mathit{\text{out}}})·\frac{*[\mathit{\text{L}}]}{[\mathit{\text{L}}]}-({\mathit{\text{V}}}_{\mathit{\text{x}}}+{\mathit{\text{V}}}_{\mathit{\text{TCA}}})·\frac{*[\mathit{\text{K}}]}{[\mathit{\text{K}}]}+{\mathit{\text{V}}}_{\mathit{\text{x}}}·\frac{*[\mathit{\text{Glx}}]}{[\mathit{\text{Glx}}]}$$ (6)

where V_TCA and V_x stand for the TCA cycle flux and α-KG/Glx exchange rate; [Glx] and ^*[Glx]: combined pool for glutamate and glutamine.

(e) Combined pool for brain glutamate and glutamine, [Glx] and ^*[Glx]

$$\frac{\text{d}[\mathit{\text{Glx}}]}{\mathit{\text{dt}}}=0$$ (7)

$$\frac{\mathit{\text{d}}*[\mathit{\text{Glx}}]}{\mathit{\text{dt}}}={\mathit{\text{V}}}_{\mathit{\text{x}}}·\frac{*[\mathit{\text{K}}]}{[\mathit{\text{K}}]}-{\mathit{\text{V}}}_{\mathit{\text{x}}}·\frac{*[\mathit{\text{Glx}}]}{[\mathit{\text{Glx}}]}$$ (8)

(f) Brain glycogen, [Gly] and ^*[Gly]

$$\frac{\text{d}[\mathit{\text{Gly}}]}{\mathit{\text{dt}}}=0$$ (9)

$$\frac{\mathit{\text{d}}*[\mathit{\text{Gly}}]}{\mathit{\text{dt}}}={\mathit{\text{V}}}_{\mathit{\text{y}}}·\frac{*[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}{[\mathit{\text{Glc}}{]}_{\mathit{\text{brain}}}}-{\mathit{\text{V}}}_{\mathit{\text{y}}}·\frac{*[\mathit{\text{Gly}}]}{[\mathit{\text{Gly}}]}$$ (10)

The following parameters of pool sizes and rates were taken from different reports in the literature: [Glc]_brain = 1.6 mM²⁸ (Baseline, Pre-infusion); [L] =2.2 mM;²⁹ [K] = 0.12 mM;²⁹ [Glx] = 20 mM;^4,5,29 [Gly] = 5.3 mM;^30,31 V_out = 0.13 mM/min.⁴

Other parameters of T_max (=3.1 ± 0.2 µmol/g/min), K_T (=6.4 ± 4.4 µmol/g) and V_x (=4.2 ± 3.3 µmol/g/min) were determined by kinetic analysis in the present study. By using variance and the Jacobian matrix, the covariance of CMR_glc and V_TCA was calculated (=0.013). The impact of V_out variation on the modeling was examined. Within a reasonable range of the change in V_out (0.13 ± 10%, e.g. +8% reported in the literature³²), the maximal estimated changes in CMR_glc and V_TCA were within 4% under the anesthesia condition as employed in the present study. Therefore, small variation in V_out should not impact the CMR_glc and V_TCA measurements significantly. In addition, our simulation indicated that the influence of the ²H-labeled glucose in blood on the outcome of regressed brain metabolic rates was negligible owing to a small fraction of blood volume in brain tissue and rapid decay of the labeled glucose concentration in the blood during the post-infusion period.

All results were presented as mean ± standard deviation (SD). Comparisons were performed using Student’s t-test with the differences in mean values considered to be statistically significant at a probability level of p < 0.05.

Results

Figure 3 demonstrates excellent spectral quality and good fittings of the original spectra not only for the ²H spectrum of phantom solution but also for the in vivo DMRS data. Two groups of resonances appeared in the DMR phantom spectrum, including a large water peak (natural abundance, set at 4.8 ppm) and a small glucose peak (d66, 3.8 ppm) (Figure 3(a), upper panel). Calculation based on the fitted spectrum (Figure 3(a)lower panel) predicted a deuterated glucose (d66) concentration of 5.2 mM in the phantom solution (close to the known value of 5 mM), which indicates the accuracy and reliability of the quantification method. As shown in Figure 3(b) to (f), besides the large water peak (4.8 ppm), three additional well-resolved resonance signals (glucose at 3.8 ppm, Glx at 2.4 ppm and lactate at 1.4 ppm) were detected in the rat brain following the brief (2 min) infusion of d66. Excellent SNR and resonance separation in the in vivo DMR spectra acquired within 15 s per spectrum made it possible to monitor the dynamic changes of water, glucose and Glx robustly. Figure S3 illustrated the traces of ²H signal changes in cerebral glucose, Glx and water signals from the stack-plot of original spectra obtained under isoflurane or morphine condition, respectively.

Figure 3.

Representative original (black trace in upper row and grey trace in bottom row) and fitted (red trace in bottom row) DMR spectra obtained from deuterated glucose (d66) phantom solution (a) and a rat brain under constant morphine sulfate infusion pre- (b), 5 min (c), 30 min (d), 60 min (e) and 120 min (f) post-deuterated glucose (d66) infusion. Each in vivo spectrum displayed in this figure was summed from 1 min of data acquisitions (four spectra). ²H resonance assignments: (1) water (4.8 ppm, set as a chemical shift reference); (2) glucose (3.8 ppm); (3) Glx (2.4 ppm); and (4) lactate (1.4 ppm).

Figure 4 shows the time courses of deuterated glucose and labeled Glx concentrations measured from a representative rat brain under each condition. The rates of cerebral glucose consumption and labeled Glx accumulation were found significantly increased in the morphine group when compared with the isoflurane control. Delayed Glx labeling appearance and the onset of cerebral glucose decay were also observed in the isoflurane group. All of these observations indicate an accelerated glucose metabolism with increased CMR_glc and V_TCA under morphine condition. Increased heart rate (416 ± 22 vs. 340 ± 20 bpm using isoflurane, p < 0.05) and mean blood pressure (141 ± 12 vs. 111 ± 12 mmHg using isoflurane, p < 0.05) were also found in the morphine group. Nevertheless, there were no significant difference in pH = 7.37 ± 0.05, pCO₂ = 42.5 ± 6 mmHg and pO₂ = 151 ± 38 mmHg between the two groups.

Figure 4.

Dynamic changes and time courses of deuterated brain glucose (d66) and labeled Glx concentrations during sequential DMRS acquisitions (15 s temporal resolution) in two representative rat brains under 2% isoflurane anesthesia vs. constant morphine sulfate infusion, respectively. Time = 0, starting time point for model fitting.

The T₁ values of deuterated water, glucose (d66) and Glx measured in the rat brain in vivo were 0.36 ± 0.01 s (n = 15), 0.05 ± 0.02 s (n = 4) and 0.20 ± 0.05 s (n = 3) (obtained in separate experiments), respectively. As a comparison, T₁ values of water and d66 in the phantom solution (pH = 7.0, room temperature) were 0.45 s and 0.06 s, respectively, indicating similar T₁ values between phantom solution and in vivo brain.

Figure 5 demonstrates excellent fittings of the kinetic model with the dynamic changes of labeled glucose and Glx concentrations in representative rat brains under isoflurane (Figure 5(a)) or morphine condition (Figure 5(b)). As a result of kinetic analysis, increased CMR_glc (0.46 ± 0.06 vs. 0.28 ± 0.13 µmol/g/min for isoflurane, p < 0.05) and V_TCA (0.96 ± 0.4 vs. 0.6 ± 0.2 µmol/g/min for isoflurane, p = 0.1) were determined in the morphine treated brains (n = 4 for each group). Interestingly, we also found a significant increase in the glycogen synthetic rate under the morphine condition (V_y: 0.08 ± 0.02 µmol/g/min vs. 0.002 ± 0.001 µmol/g/min for isoflurane, p < 0.01), indicating an elevated glycogen synthetic activity.

Figure 5.

Fitting of both glucose (d66, blue circles) and Glx (green circles) labeling curves with kinetic modeling for representative rats under 2% isoflurane anesthesia (a) vs. constant morphine sulfate infusion (b). Solid lines are the model fittings of labeled glucose (red) and Glx (black) changes. Starting time point (0 min) for model fitting corresponds to 5 min after ending of the d66 infusion.

Discussion and conclusion

This study demonstrates that dynamic DMRS is well suited for quantitative assessment of cerebral glucose metabolisms in vivo at high/ultrahigh field. Our results suggest that this approach is robust and reliable for detecting dynamic changes in labeled glucose, Glx and water concentrations in the rat brain with excellent spectral quality and sensitivity. When combined with metabolic modeling, fluxes through important pathways such as glycolysis and TCA cycle could be determined via kinetic analysis from the same in vivo measurement, thus, allows simultaneous quantification of CMR_glc and V_TCA in the brains under physiological or pathological conditions. It also provides an opportunity for in vivo study of coupling relationship between aerobic and anaerobic glucose metabolisms in the healthy brain and potential de-coupling relationship in neurological disorders, which can be examined via the ratio of the simultaneously measured CMR_glc and V_TCA values.

Evaluation of the in vivo DMR spectroscopy

As shown in Figure 3(b), before the infusion of deuterated glucose (d66), a large single peak at 4.8 ppm representing natural abundance signal of deuterium water was detected reliably in the rat brain. Following the d66 infusion, well-resolved resonances of glucose (3.8 ppm), Glx (2.4 ppm) and lactate (1.4 ppm) became visible, and their dynamic changes (Figure 3(c) to (f)) reflected the glucose transportation and metabolism. It is interesting to notice that the chemical shift of lactate or Glx in the in vivo DMR spectrum is identical to that in the ¹H-MRS (e.g. δ_[3,3-D2] = δ_[3,3,3-H3] for lactate, and δ_[4,4-D2] = δ_[4,4-H2] for glutamate). These findings collectively reveal that in vivo brain DMRS shares a similar chemical shift range to that of proton MRS, and this knowledge simplifies the chemical shift assignment and identification of brain metabolites detected by in vivo DMRS.

Considering the measured in vivo ²H T₁ values (0.36 s for water, 0.05 s for glucose and 0.2 s for Glx) and the 0.3 s repetition time used in this study, there were partial saturation effects on the water and Glx signals, which had been corrected in the quantification. On the other hand, the glucose (d66) with a very short T₁ value was acquired under fully relaxed condition even with a very short repetition time.

Importantly, the excellent sensitivity and temporal resolution of in vivo DMRS ensures the reliable detection of dynamic change of labeled metabolites, which revealed a significantly delayed Glx labeling (for few minutes) under isoflurane condition (see Figure 4). This observation may provide possible explanation on the challenges of hyperpolarized ¹³C technique^33,34 in detecting the labeling of glutamate and/or glutamine in animal brain under deep anesthesia condition. A recent study has demonstrated the feasibility for detecting the ¹³C signal of 2-oxoglutarate, a TCA cycle intermediate, after infusion of hyperpolarized [1,2-¹³C]acetate in the rat brain anesthetized with isoflurane; however, the study indicated an absence of detectable ¹³C Glx intermediates in the brain.³⁵ The slow turnover of carbon isotope label into Glx (>2 min under 2% isoflurane, as demonstrated in Figure 4) makes it difficult to detect the usable hyperpolarized ¹³C Glx signals within a critical time window (usually <2 min) that is limited by the T₁ relaxation time of labeled Glx in vivo. Similar technical challenge could be anticipated for the hyperpolarized ¹³C application in the resting human brain since the reported CMR_glc (∼0.3 µmol/g/min) in the human brain³⁶ is almost identical to that of rat brain under 2% isoflurane anesthesia condition (0.28 µmol/g/min) as measured in the present study. Nevertheless, it is interesting to see if the hyperpolarized ¹³C MRS technique enables detection of the hyperpolarized ¹³C Glx signal change, thus, assessment of TCA cycle metabolic activity in the human visual cortex during visual stimulation, in which an approximately 50% increase of CMR_glc can be expected.^35,36

Validity of dynamic DMRS for assessing cerebral glucose metabolisms

A simplified model consisting of mass and isotope balance equations (see details in Materials and methods and Figure 2) written for individual metabolites were developed in this study. Several model assumptions were made for performing the kinetic analysis: (i) glucose and its isotope (d66) shared same mechanism and property for transportation, which was described by the symmetric Michaelis–Menten model; (ii) small pools of glycolytic and TCA cycle intermediates were assumed to have negligible concentrations;⁴ (iii) metabolites in rapid chemical and isotopic exchange were assumed to be in equilibrium and represented by a single kinetic pool with the size equals the sum of its constituents (e.g. pyruvate and lactate);⁴ (iv) glucose (d66) was the only ²H-labeling source for TCA cycle and gluconeogenesis was negligible; (v) glutamate and glutamine were treated as a combined pool, since their resonance signals could not be differentiated robustly in the ²H spectra; (vi) exchanges of α-ketoglutarate and glutamate in their mitochondrial and cytosolic pools were represented as a single exchange reaction.³⁷

When comparing with the brains under deeper anesthesia condition (2% isoflurane), kinetic analysis predicted significant increases of CMR_glc (0.46 vs. 0.28 µmol/g/min) and V_TCA (0.96 vs. 0.6 µmol/g/min) for the morphine-treated brains (also see an example in Figure 5). These increases in metabolic fluxes reflect elevated glucose metabolism at a corresponding higher level of neuronal activity under the constant morphine infusion. The values of CMR_glc and V_TCA reported herein are in agreement with the previous results obtained from in vivo ¹³C or ¹H MRS studies under similar (∼2%) isoflurane anesthetic condition;^5,28 however, the measured CMR_glc value in the present study is relatively lower than of ∼0.5 µmol/g/min reported from other studies under lighter anesthetic condition using ≤1.5% isoflurane,^38–40 suggesting that CMR_glc is sensitive to the isoflurane dose or brain anesthetic level. However, the partial volume effect and very low glucose metabolism activity in the muscle surrounding the rat brain could also attribute a lower CMR_glc value, since a RF surface coil was applied to collect the DMRS data without using other localization technique in the present study.

The V_TCA values measured in healthy brains under two anesthesia conditions in this study were approximately two times of the CMR_glc values, which are consistent with the stoichiometric relationship that one glucose is converted to two pyruvates before entering the TCA cycle. Moreover, model simulations demonstrated that major metabolic fluxes of CMR_glc and V_TCA are not sensitive to other parameters, fluxes and pool sizes of the intermediates (see Table S1). These results confirm the validity and robustness of the in vivo DMRS approach for assessing and quantifying cerebral glucose metabolisms.

Methodology comparison of DMRS and ¹³C MRS for studying brain glucose metabolisms

In terms of the methodology, there is similarity between in vivo DMRS and ¹³C MRS approaches for monitoring and quantifying the dynamics of isotope labeling on intermediates following the introduction of isotope-labeled glucose. There are two ways in detecting ¹³C-labeled intermediates:^1,6 one based on direct ¹³C detection incorporated with ¹H NOE enhancement and decoupling^2–5 or with the DEPT (distortionless enhancement by polarization transfer) technique,^41,42 and another based on the indirect ¹H detection using the proton-observer-carbon-editing (POCE) technique to specifically measure the protons attached to the ¹³C-labeled carbon.^43–45 We discuss and compare the similarity and differences between in vivo DMRS and ¹³C MRS methods.

T₁ relaxation and sensitivity

The dipole–dipole relaxation of ¹³C spins detected by ¹³C MRS or ¹H spins detected by POCE leads to long T₁ values usually in a range of >1.4 s.^46,47 In contrast, the much shorter ²H T₁ relaxation times of deuterated metabolites (e.g. 0.05 ± 0.02 s for d66) dominated by quadrupolar relaxation should provide a substantial sensitivity gain (e.g. >5 folds for d66, since SNR is inversely proportional to the square root of T₁) for the in vivo DMRS detection via more signal averaging within the same sampling time. Practically, the DMRS SNR can be further improved by using a shorter repetition time and Ernst flip angle, if the correction of saturation for water and Glx is performed based on their known T₁ values for quantification. Nevertheless, it is interesting and critical to quantitatively compare the sensitivity of DMRS versus direct ¹³C and indirect POCE approaches for in vivo assessment of cerebral glucose metabolisms in future using the similar SNR quantification methods as reported in the literature.^48–50

Similarity and differences in spectral features and MRS technology

Comparing to the ¹³C MRS or ¹H POCE, in vivo DMRS has the advantage of a narrower chemical shift range (in Hz unit) and a “clean” spectral background. The ¹³C resonances of biological interest cover a very wide chemical shift range up to 200 ppm,⁶ which is approximately 10 times (in Hz unit) of that ¹H MRS (∼5 ppm covering from 1 ppm to 6 ppm) considering 4 times lower ¹³C gyromagnetic ratio (γ) than ¹H. This poses two challenges for in vivo ¹³C MRS applications: a large chemical shift displacement error to localize ¹³C NMR signals, especially at high and ultrahigh fields; and difficulty to achieve effective bandwidth of RF pulse excitation or refocusing covering the entire ¹³C chemical shift range. In contrast, DMRS has a similar chemical shift range of interest (1–5 ppm) as ¹H MRS, however, is almost seven times narrower (in Hz) than that of ¹H-based POCE method or ¹³C DEPT method based on ¹H localization because of a very low ²H γ, or 70 times narrower than that of ¹³C. This advantage results in minimal chemical shift displacement for spectral localization using the slice-selection gradients and optimal performance of RF pulses.

One technical hurdle commonly faced by ¹³C MRS or ¹H-based POCE is the contamination of intense macromolecule or lipid signals from subcutaneous fat, as well as the tissue water signal in ¹H-based POCE detection, thus, it requires advanced water and out-volume suppression techniques adding more RF pulses and gradients for obtaining high quality spectra. It becomes a daunting challenge to achieve three-dimensional ¹³C MRS imaging covering the entire brain with completely suppressed lipid or water signals. In contrast, the very low natural abundance of ²H (73 times lower than that of 1.1% for ¹³C) could limit the background ²H signals from macromolecule contamination down to the noise level as demonstrated by global brain ²H spectra in Figure 3, resulting in “clean” spectral background. Moreover, the sole resonance detected in the baseline ²H spectra before d66 infusion is from tissue water signal, which can serve as an internal reference for calibrating and quantifying the concentrations of ²H-labeled glucose and intermediates and their dynamic changes (see Figures 4 and 5), and the metabolic rates.

¹³C MRS, ¹³C-based DEPT or ¹H-based POCE requires dual ¹H-¹³C RF channels for performing NOE and/or decoupling (in either ¹H or ¹³C frequency) using a large RF power during FID acquisition, potentially leading to high specific absorption rate (SAR) of RF power: a safety concern. In addition, multiple bandpass filters are needed to isolate the cross interaction of RF power and noise between the ¹H and ¹³C channels. In contrast, DMRS relies on a single channel of RF system without the need of water suppression or out volume suppression RF pulses, thus, minimizing the RF pulse number, RF power usage and SAR. These advantages make DMRS simple, robust and reliable for implementation, quantification and spatial localization based on single voxel or whole brain chemical shift imaging.

Isotope labeling pathway, metabolic modeling and quantification

One major merit of ¹³C MRS is the well-resolved ¹³C resonances including homonuclear J-coupling peaks and it provides rich information for performing kinetic analysis and model fitting. In contrast, DMRS has relatively poor spectral resolution and is not robust to resolve the Glu and Gln resonances; thus, it is difficult to determine the Glu-Gln transmission cycling rate.

Although the glucose metabolism pathways are identical between the ²H-labeled glucose (e.g. d66 herein) and ¹³C-labeled glucose, the fate of isotope labeling and detectable intermediates are similar but not identical in these two approaches. For instance, [1-¹³C]glucose is commonly employed for in vivo ¹³C studies in the literature; however, either α- or β-D-[1-¹³C]glucose resonance peak detected by POCE overlaps with intense water resonance or its residual, making them difficult to be reliably detected or quantified. It is equally difficult for ¹³C MRS detection due to a large chemical shift difference between [1-¹³C]glucose and ¹³C-labeled Glx resonances. In contrast, the ²H resonances of d66 glucose (at around 3.8 ppm) separate from the water resonance by nearly 1 ppm, thus, can be easily resolved as shown in Figure 3.

Considering the glycolysis and TCA cycle pathways, the metabolism of ¹³C-labeled glucose will label Glx C4 first, then label other carbons after multiple turns of TCA cycling, and thus sophisticated modeling approaches are critical to account the complex labeling and dynamics for determining the metabolic rates of interest.^32,51–53 On the other hand, when the ²H-labeled glucose (d66) enters the TCA cycle, it will replace the protons attached on Glx C4 via the α-ketoglutarate/Glx exchange; however, other ²H labels will depart from the TCA cycle intermediates and become water (see Figure 1), and thus will not enter next-turn of TCA cycles. Therefore, DMRS with d66 infusion only labels Glx on the C4 position, and makes the modeling relatively straightforward and simpler than the ¹³C modeling.

Accumulation of labeled water was clearly demonstrated in Figures 3 and S3 under both isoflurane and morphine conditions. This nascent mitochondrial water formation is partially associated with TCA cycle and oxygen consumption. Unfortunately, due to the complexity of the fate in deuterium label departing from the TCA cycle intermediates (or labeling dilution), we were unable to use the dynamic information of ²H water signal, though with excellent SNR, for quantitative determination of cerebral metabolic rate of oxygen consumption without the knowledge of labeling dilution. This requires further investigation.

To simplify measurement and quantification, traditional ¹³C studies commonly use a clamp protocol for introducing ¹³C-labeled glucose into venous blood stream, thus, keeping a constant level of ¹³C-labeled glucose in plasma during hours of ¹³C MRS measurement. In this study, we employed a brief (few minutes) d66 infusion protocol,²⁸ resulting in large dynamics of d66 glucose signal change in brain (Figures 3 to 5), which is critical for modeling and determination of CMR_glc, V_TAC and V_y simultaneously. Such a protocol should be suitable for translational application in patients.

Potential, limitation and future direction of dynamic DMRS

Besides the dynamic changes of the labeled glucose and Glx, valuable information provided by the in vivo DMR spectra also involves the labeling dynamics of lactate (Figure 3). In this study, the resonance signal of lactate remained as a small peak in the entire duration of the DMRS experiment (Figure 3(c) to (f)). Due to its relatively poor SNR resulting from the low in vivo concentration of ²H labeling and low activity of anaerobic glucose metabolism in the healthy brain, the quantification of cellular lactate labeling in this study was not robust for in vivo ²H MRS with a temporal resolution of 15 s. However, if adequate SNR becomes available (such as in the case of brain ischemia or tumor with an elevated lactate level) and the lactate pool is considered in the quantification model, simultaneous quantification of lactate production, CMR_glc and V_TCA would be possible.

Since the in vivo dynamic DMRS was unable to separate the labeling signals between glutamate and glutamine, a combined pool was applied in the kinetic modeling. Also for simplification, the model used in this work was a lumped one. Therefore, unlike ¹³C MRS, the rate of glutamate-glutamine cycle could not be determined.

Although the present study was performed at ultrahigh field of 16.4 T, the applications of dynamic DMRS could be extended to lower field strength in animal and human brains or other tissue/organs. Furthermore, the excellent spectral quality and sensitivity of the dynamic ²H spectra should become feasible for in vivo application using localized DMRS or imaging methods, and thus, has the potential to assess regional CMR_glc and V_TCA in the brain.

One unanswered question is: if the in vivo DMRS approach described herein could offer significantly higher sensitivity or better SNR than in vivo ¹³C MRS methods based on either direct or indirect detection, and it deserves rigorous investigation.

In conclusion, the results of this work indicate that in vivo DMRS approach is feasible, robust and reliable for noninvasively assessing the cerebral glucose metabolisms. This novel approach provides an alternative to the established ¹³C MRS methods for in vivo study of metabolic coupling relationship between aerobic and anaerobic glucose metabolisms and neuro-metabolic coupling in the animal and human brains under physiopathological condition.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was partly supported by the National Institute of Health, grant numbers R01 NS057560, NS070839, MH111447 and MH111413; R24 MH106049 and MH106049 S1, P41 EB015894, P30 NS076408, and S10 RR025031; and the W.M. Keck Foundation.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

WC, GM, ML and XHZ devised the study. ML, XHZ, YZ and WC performed the study. YZ performed the animal surgical procedure and maintenance of animal physiological condition. ML performed the MRS scans, kinetic and spectral analysis. ML, WC and XHZ discussed and interpreted the data and finding, and prepared the manuscript.

Supplementary material

Supplementary material for this paper can be found at the journal website: http://journals.sagepub.com/home/jcb