Simultaneous assessment of cerebral glucose and oxygen metabolism and perfusion in rats using interleaved deuterium (²H) and oxygen‐17 (¹⁷O) MRS

Abstract

Cerebral glucose and oxygen metabolism and blood perfusion play key roles in neuroenergetics and oxidative phosphorylation to produce adenosine triphosphate (ATP) energy molecules in supporting cellular activity and brain function. Their impairments have been linked to numerous brain disorders. This study aimed to develop an in vivo magnetic resonance spectroscopy (MRS) method capable of simultaneously assessing and quantifying the major cerebral metabolic rates of glucose (CMR_Glc) and oxygen (CMRO₂) consumption, lactate formation (CMR_Lac), and tricarboxylic acid (TCA) cycle (V _TCA); cerebral blood flow (CBF); and oxygen extraction fraction (OEF) via a single dynamic MRS measurement using an interleaved deuterium (²H) and oxygen‐17 (¹⁷O) MRS approach. We introduced a single‐loop multifrequency radio‐frequency (RF) surface coil that can be used to acquire proton (¹H) magnetic resonance imaging (MRI) or interleaved low‐γ X‐nuclei ²H and ¹⁷O MRS. By combining this RF coil with a modified MRS pulse sequence, ¹⁷O‐isotope‐labeled oxygen gas inhalation, and intravenous ²H‐isotope‐labeled glucose administration, we demonstrate for the first time the feasibility of simultaneously and quantitatively measuring six important physiological parameters, CMR_Glc, CMRO₂, CMR_Lac, V _TCA, CBF, and OEF, in rat brains at 16.4 T. The interleaved ²H–¹⁷O MRS technique should be readily adapted to image and study cerebral energy metabolism and perfusion in healthy and diseased brains.

Cerebral perfusion, cerebral energy metabolism, and metabolic rates along key metabolic pathways were quantitatively assessed using an interleaved deuterium (²H)–oxygen‐17 (¹⁷O) magnetic resonance spectroscopy (MRS) approach, which enables measurement of the three cerebral metabolic rates of glucose consumption (CMR_Glc), oxygen consumption (CMRO₂), lactate production (CMR_Lac); tricarboxylic acid (TCA) cycle rate (V _TCA); and cerebral blood flow (CBF).

Abbreviations

¹³C

carbon‐13

¹⁷O

oxygen‐17

¹H

proton

²H

deuterium

³¹P

phosphorus‐31

CBF

cerebral blood flow

CMR_Glc

cerebral metabolic rate of glucose consumption

CMR_Lac

cerebral metabolic rate of lactate formation

CMRO₂

cerebral metabolic rate of oxygen consumption

CSI

chemical shift imaging

GEMS

gradient‐echo multiple slice

HDO

deuterated water

OEF

oxygen extraction fraction

RF

radio frequency

SNR

signal‐to‐noise ratio

SPULS

single‐pulse acquire sequence

TCA

tricarboxylic acid

V _TCA

cerebral metabolic rate of the TCA cycle

1. INTRODUCTION

The brain is one of the most metabolically active organs in the body, using glucose as its primary energy source. ¹ It accounts for only 2% of an adult's body weight (BW) but consumes approximately 20% of the body's total glucose and oxygen consumption at rest. ² Glucose is delivered to the brain through circulating blood and crosses the blood–brain barrier via glucose transporters. Once inside the brain cells, glucose is metabolized through the glycolytic pathway to form two pyruvate molecules, which are an important intermediate in glucose metabolism, and this metabolic activity can be quantified by the cerebral metabolic rate of glucose consumption (CMR_Glc). Pyruvate can be converted into lactate via lactate dehydrogenase in the cytoplasm, a reaction that can occur in the presence of oxygen in a healthy brain and is therefore termed aerobic glycolysis. This conversion, measurable by the cerebral metabolic rate of lactate formation (CMR_Lac), generates only two energy molecules of adenosine triphosphate (ATP) per consumed glucose. Normally, the majority of pyruvate can be converted to acetyl‐CoA by pyruvate dehydrogenase and then enter the tricarboxylic acid (TCA) cycle occurring inside mitochondria, a biochemical process that can be quantified by the cerebral metabolic rate of the TCA cycle (V _TCA). The TCA cycle supports oxidative phosphorylation, which is closely related to oxygen usage (measurable by the cerebral metabolic rate of oxygen consumption [CMRO₂]), and produces over 30 ATP molecules per consumed glucose. ³ , ⁴ Figure 1A summarizes the main metabolic pathways and associated rates of brain glucose energy metabolism, including non‐oxidative (CMR_Glc and CMR_Lac) and oxidative (V _TCA and CMRO₂) metabolic processes and perfusion process (measurable by the cerebral blood flow [CBF]). There is an unmet need to develop a noninvasive method that enables quantitative assessment of these major metabolic rates, CBF, and oxygen extraction fraction (OEF) ideally from a single in vivo measurement.

FIGURE 1.

Quantitative assessment of cerebral perfusion, cerebral energy metabolism, and metabolic rates along key metabolic pathways using an interleaved ²H–¹⁷O magnetic resonance spectroscopy (MRS) approach. (A) Glucose is supplied and transported via the circulating blood flow to brain cells, where it is metabolized via cytosolic glycolysis and the mitochondrial tricarboxylic acid (TCA) cycle, coupled with oxygen metabolism and oxidative phosphorylation to generate adenosine triphosphate (ATP) molecules in supporting neuronal activity and brain function. (B) Dynamic changes in ²H‐labeled metabolite signals measured by in vivo ²H MRS enable the estimation of the three cerebral metabolic rates of glucose consumption (CMR_Glc), lactate formation (CMR_Lac), and TCA cycle activity (V _TCA). In parallel, dynamic changes in the ¹⁷O‐labeled metabolic water signal facilitate the estimation of the cerebral metabolic rate of oxygen consumption (CMRO₂), cerebral blood flow (CBF), and oxygen extraction fraction (OEF ∝ CMRO₂/CBF). ADP, adenosine diphosphate.

In vivo X‐nuclear (e.g., phosphorus‐31 (³¹P), carbon‐13 (¹³C), deuterium (²H), and oxygen‐17 (¹⁷O)) magnetic resonance spectroscopy (MRS) or imaging techniques are promising for studying cerebral metabolism and bioenergetics in animal and human brains. ² , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ Their detection sensitivities and capabilities are significantly improved at high/ultrahigh fields. ¹⁰ , ¹¹ , ¹² However, these technologies are often used separately to gather limited information from the resonance signals of metabolites associated with the desired nucleus. Performing interleaved or simultaneous multi‐nuclei measurements is challenging, hindering the ability to capture comprehensive metabolic and hemodynamic information in the same brain and MRS scan session.

One essential component of X‐nuclear MRS measurement is the dual‐frequency radio‐frequency (RF) coil, which can be operated at the proton (¹H) Larmor frequency for acquiring brain anatomic images and performing magnetic field (B₀) shimming and at an X‐nuclear Larmor frequency for acquiring MRS. Including multiple RF surface coils ¹³ tuned to the desired multiple resonant frequencies in an RF coil probe creates technical challenges because it is difficult to decouple them electronically. This complexity in the coil structure may degrade coil performance or require manual tuning of a single coil to multiple frequencies at a given time, ² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ thus challenging to achieve rapid frequency switching between different MRS operation frequencies, making it difficult to capture time‐sensitive multi‐nuclei metabolic information and dynamics through interleaved or simultaneous scans.

In this study, we developed a single‐loop RF surface coil with active tuning/matching capability to facilitate ¹H magnetic resonance imaging (MRI) and interleaved ²H–¹⁷O MRS measurements of rat brains at 16.4 T. By integrating dynamic ²H–¹⁷O MRS acquisitions with ²H‐labeled glucose administration and ¹⁷O‐labeled oxygen gas inhalation, we demonstrated for the first time the feasibility of simultaneously measuring and quantifying six key neurophysiological parameters: CMR_Glc, CMR_Lac, and V _TCA from dynamic ²H MRS data and CMRO₂, CBF, and OEF from dynamic ¹⁷O MRS data, both obtained from the same in vivo measurement, as illustrated in Figure 1B.

2. METHODS

Firstly, we designed and constructed a prototype single‐loop multifrequency surface coil that can operate at the ¹H Larmor frequency for acquiring brain anatomic images and performing B₀ shimming and also allow rapid switch between ²H and ¹⁷O Larmor frequencies via a circuit controller. ²⁰ , ²¹ The coil performance was first tested on bench, and then a phantom test was conducted to validate the functionality of both the hardware (i.e., MRI scanner, RF coil, and circuit controller) and software (i.e., MRS pulse sequences) essential for the in vivo study using the rat brain model. Lastly, two sets of in vivo rat brain experiments were conducted by placing the RF coil on top of the animals' heads for the proof of concept of the interleaved ²H–¹⁷O MRS technique. The first in vivo experiment involved intravenous infusion of isotope‐enriched D₂O and H₂ ¹⁷O water as dual‐perfusion tracers ²² to evaluate the sensitivity, quality, and reliability of the acquired ²H and ¹⁷O MRS data. The second set of in vivo experiments used intravenous administration of deuterium‐labeled d‐glucose‐6,6‐d₂ (D66) in parallel with an inhalation of ¹⁷O‐labeled ¹⁷O₂ gas to explore the feasibility of simultaneously assessing cerebral glucose and oxygen metabolic rates, CBF, and OEF in the rat brain. All MRI/MRS experiments were conducted using a 16.4‐T/26‐cm‐bore‐size animal MRI scanner (Varian/VNMRJ, California) equipped with a newly designed and constructed multifrequency RF surface coil. All animal studies were conducted in strict accordance with the Guide of the National Research Council for the Care and Use of Laboratory Animals. The Institutional Animal Care and Use Committee of the University of Minnesota approved the protocol.

2.1. Coil design and bench test

To obtain a wide frequency tuning capability, a conventional single‐loop RF coil loop of approximately 2.5 cm in diameter was connected to a tuning‐matching circuit via two short copper wires with the desired length, used as an isolating inductor (L₁) to form two primary resonant circuits: One resonates at a very high ¹H Larmor frequency (698 MHz at 16.4 T), and the other resonates at a lower frequency range covering the ¹⁷O (94 MHz) and ²H (107 MHz) Larmor frequencies, as shown in Figure 2A,B (details about the coil design can be found in the literature ²⁰ , ²³ ). To enable interleaved ²H–¹⁷O MRS measurements, a positive‐intrinsic‐negative (PIN)‐diode circuit (controller) was added to actively switch the coil operation frequency between the ²H and ¹⁷O resonant frequencies. ²¹ By introducing additional direct current (DC) input combined with the RF power provided by the coaxial cable, biasing the diode became feasible. This design allows for the RF current to flow through two capacitors (C₁, C₂), thereby increasing the capacitance of the tuning capacitors (C_L plus C₂) and the matching capacitors (C_S plus C₁), enabling dropping the coil resonant frequency from 107 MHz (²H) to 94 MHz (¹⁷O), creating an active tuning/matching mechanism that rapidly switches (<100 μs) the RF coil operation mode between ²H and ¹⁷O resonant frequencies (Figure 2A). In this study, the ¹H and ²H resonant frequencies were initially tuned and matched manually under the loaded condition, whereas automatic tuning and matching of the ¹⁷O resonant frequency was achieved through an actively tuning/matching PIN‐diode circuit (Figure 2A). Using the single‐loop multifrequency RF coil design and the circuit controller, the coil can work for ¹H/²H/¹⁷O triple resonant frequencies. The performance of the RF coil circuit was simulated using Keysight's ADS RF software, employing carefully selected electronic component values designed to minimize reflection coefficients at the triple targeted resonant frequencies. Subsequently, the RF coil circuit was built to specifications with traps added to reduce coaxial cable interference. The ¹H/²H/¹⁷O multifrequency surface coil was validated on bench using a network analyzer by measuring the reflection coefficients.

FIGURE 2.

(A) Schematic circuit diagram of the multifrequency coil design. For ¹H (with positive‐intrinsic‐negative [PIN] diodes off), adjust C_S and C_L to tune and match the coil resonant frequency to 698 MHz at 16.4 T; for ²H (with PIN diodes off), tune and match C_S and C_L to the ²H operation frequency at 107 MHz; and for ¹⁷O, ensure that the setup for ²H is established, then apply a direct current (DC) and activate the PIN diodes, and use additional C₁ and C₂ to tune and match to the ¹⁷O operation frequency at 94 MHz. (B) Photo of the prototype coil. (C) S11 parameters measured using the network analyzer with the coil tuned to the three targeted resonant frequencies: 698 MHz (¹H), 107 MHz (²H), and 94 MHz (¹⁷O).

2.2. Phantom test

The ¹H/²H/¹⁷O multifrequency surface coil underwent MRS testing at 16.4 T using a spherical phantom (2‐cm diameter) filled with deionized water. ¹H MRI was performed using a multi‐slice two‐dimensional (2D) gradient‐echo multiple slice (GEMS) imaging sequence with the following acquisition parameters: repetition time (TR)/echo time (TE) = 100 ms/4 ms, nominal RF pulse flip angle (FA) = 20°, field of view (FOV) = 30 mm × 30 mm, slice thickness = 1 mm, and matrix = 128 × 128. RF pulse power calibration and B₀ shimming were performed before the ¹H MRI and ²H–¹⁷O MRS acquisitions.

We conducted interleaved ²H–¹⁷O MRS measurements using a modified single‐pulse acquire sequence (SPULS; Figure 3A), with each data block repeated 20 times for signal averaging. Each repetition consists of six segments (each lasting 50 ms): Five ¹⁷O free‐induction‐decay (FID) acquisitions were captured when the RF coil and RF pulse carry frequency were switched to the ¹⁷O Larmor frequency when the PIN didoes were short, and then one ²H FID was collected when the RF coil and RF pulse carry frequency were switched back to the ²H Larmor frequency when the PIN didoes were open. The apparent TRs were 300 ms for ²H and 50 ms for ¹⁷O. Key MRS acquisition parameters include a spectral width (SW) of 4006 Hz, an FID acquisition time (AT) of 44 ms, an RF hard pulse width of 200 μs, and an average (nt) of 20 FIDs for ²H and 100 FIDs for ¹⁷O across 20 repetitions (Figure 3A). The longitudinal relaxation time (T ₁) of deuterium water (HDO) in the rat brain at 16.4 T is approximately 360 ms, ⁶ whereas the T ₁ of H₂ ¹⁷O is less than 10 ms.² Therefore, the RF pulse FA was optimized to a 64° Ernst angle for acquisition of ²H FIDs under partially saturated conditions and a nominal 90° FA for acquisition of ¹⁷O FIDs under fully relaxed conditions.

FIGURE 3.

(A) Schematic representations of the modified single‐pulse acquire sequence and parameter configurations for the interleaved ¹⁷O–²H magnetic resonance spectroscopy (MRS) acquisitions using the novel radio‐frequency (RF) coil (Figure 2A,B), which can be operated at ¹⁷O/²H/¹H multifrequencies. (B) Sequential ²H (nt = 1 × 20, top) and ¹⁷O (nt = 5 × 20, bottom) spectra obtained from a natural abundance water phantom scanned using the interleaved ¹⁷O–²H MRS technique. Phantom test results of (C) ¹H magnetic resonance imaging (MRI) acquired with a two‐dimensional (2D) gradient‐echo multiple slice (GEMS) imaging sequence and (D) averaged ²H (top panel, nt = 20) and ¹⁷O (bottom panel, nt = 100) signals from control scans (in black) and interleaved ²H–¹⁷O scans (in red) of natural abundance water. FIDs, free induction decays; PIN, a positive‐intrinsic‐negative; TR, repetition time.

To evaluate the performance of the interleaved ²H–¹⁷O MRS measurements, we collected control ²H and ¹⁷O MRS data from the phantom separately using the original SPULS sequence and the same acquisition parameters used in the interleaved ²H–¹⁷O MRS measurements. All obtained FIDs, whether using original or modified SPULS sequences, underwent the same signal processing. Additionally, interleaved ²H–¹⁷O MRS scans were averaged to ensure the same signal average as the control ²H or ¹⁷O MRS acquisition for performance comparison.

2.3. Rat brain experiment with bolus administration of D₂O and H₂ ¹⁷O as perfusion tracers

A Sprague Dawley rat was anesthetized with 2% isoflurane in a mixture of N₂O/O₂ (about 3:1 volume ratio) and mechanically ventilated to maintain stable physiology during MRI/MRS measurements. The rat received 0.5 mL of deuterium‐labeled water (D₂O, 99.8% enrichment) mixed with 3 mL of ¹⁷O‐isotope‐labeled water (H₂ ¹⁷O, approximately 2% enrichment) via femoral vein infusion. The interleaved ²H–¹⁷O MRS pulse sequence (Figure 3A) was used to acquire dynamic ²H and ¹⁷O MRS FIDs with the following acquisition parameters: for ²H, TR = 300 ms, nt = 1, SW = 4006 Hz, AT = 44 ms, and FA = 64° (Ernst angle); for ¹⁷O, TR = 50 ms, nt = 5, SW = 4006 Hz, AT = 44 ms, and FA = 90°. Dynamic ²H–¹⁷O MRS data acquisition lasted 30 min and included a 1‐min baseline, 6‐min dual‐perfusion tracer infusion, and a 23‐min post‐infusion period. Prior to Fourier transformation, all ²H and ¹⁷O FIDs were subjected to 15‐Hz line broadening to enhance the spectral signal‐to‐noise ratio (SNR). The decay rate constants (k) of the HDO and H₂ ¹⁷O water resonance intensities measured from the post‐infusion spectra were determined by fitting the water resonance signals to an exponential function. These k values were utilized to approximately represent the apparent CBF values.

2.4. Rat brain experiments with D66 administration and ¹⁷O₂ inhalation for studying cerebral glucose metabolism and perfusion

Two Sprague Dawley rats were mechanically ventilated under 2% isoflurane anesthesia in a mixture of N₂O/O₂ to maintain physiological conditions during MRI/MRS measurements. D66 (Cambridge Isotope Lab; 99% enrichment, 1 g/kg dissolved in 1.8 mL saline) was intravenously administered within 2 min after 5‐min baseline acquisition of interleaved ²H–¹⁷O spectra. Each rat underwent two consecutive inhalations of ¹⁷O₂ gas (74.8% enrichment, Sigma‐Aldrich, mixed with N₂O and isoflurane), each lasting approximately 2 min, occurring during and 30 min after the D66 glucose administration.

After B₀ shimming and ¹H MRI acquisition, dynamic ¹⁷O–²H MRS data from the rat brains were acquired for 80 min using the same interleaved ²H–¹⁷O MRS pulse‐acquire sequence (Figure 3A) and parameters as described above. The ²H–¹⁷O FID data underwent 15‐Hz line broadening before Fourier transformation to enhance the SNR. A custom Matlab‐based program was employed to fit all resonance signals, and the resulting spectral integrals were used to quantify the metabolite concentrations after correcting the saturation effects (for ²H spectra). The natural abundance ¹⁷O or ²H water signal acquired in the brain at the baseline condition served as an internal reference for quantification of metabolite concentration. ⁶ , ⁸

2.5. Kinetic models and quantification

For an aerobic organ with a higher oxygen metabolic rate and blood perfusion (such as the brain or heart), the detected ¹⁷O MRS signal of brain tissue H₂ ¹⁷O and its time‐dependent change can be characterized by three phases as shown in Figure S1, and the modeling and quantification can be significantly simplified for rodents with a relatively small body size and a short ¹⁷O₂ inhalation as previously described ⁸ , ²⁴ , ²⁵ :

1. Phase I (pre‐inhalation): stable natural abundance ¹⁷O MRS signal of brain tissue H₂ ¹⁷O, which can be used as an internal reference for quantifying tissue H₂ ¹⁷O concentration and its change measured in later phases

2. Phase II (¹⁷O₂ inhalation): ¹⁷O MRS signal during a brief ¹⁷O₂ inhalation (usually lasting a few minutes), showing a rapid increase in brain tissue H₂ ¹⁷O content through the metabolism of inhaled ¹⁷O₂ in brain cells; the dynamic H₂ ¹⁷O signal change can be used to quantify CMRO₂

3. Phase III (post‐inhalation): ¹⁷O MRS signal after the ¹⁷O₂ inhalation, where the labeled H₂ ¹⁷O molecules produced by the metabolism of inhaled ¹⁷O₂ in the brain tissue can be washed out via blood perfusion; the H₂ ¹⁷O signal can be used as an endogenous perfusion tracer to measure and quantify CBF

The mass balance relationship for the changes in the labeled H₂ ¹⁷O concentration in different tissue compartments during an ¹⁷O₂ gas inhalation is given by Equation (1) ⁸ , ²⁴ :

$$\frac{d{C}_{\mathrm{b}}\left(t\right)}{\mathit{dt}}=2·\alpha ·f_1·\mathrm{CMR}{\mathrm{O}}_2+f_2·\mathrm{CBF}\left[C_{\mathrm{a}}\left(t\right)-C_{\mathrm{v}}\left(t\right)\right],$$ (1)

where C _a, C _b, and C _v are the H₂ ¹⁷O concentrations expressed in excess of the natural abundance H₂ ¹⁷O level in arterial blood, brain tissue, and venous blood, respectively, as a function of the ¹⁷O₂ inhalation time t (unit = minute). One labeled ¹⁷O₂ molecule converts to two H₂ ¹⁷O molecules through oxygen metabolism; thus, a constant of 2 is used in Equation (1). The parameter α in Equation (1) is the ¹⁷O enrichment fraction in the inhaled ¹⁷O₂ gas. The natural abundance H₂ ¹⁷O concentration (equal to 20.4 μmol per gram of tissue water or per gram of water in blood; see references 8 and 24) can be used to calibrate the absolute concentrations of C _b(t), C _a(t), and C _v(t). For convenience, the unit for C _b is chosen as micromoles per gram of brain tissue water and the units for C _a and C _v are micromoles per gram of blood water, and the parameters f ₁ (=1.266) and f ₂ (=1.077) are conversion factors used for consistency of units applied in Equation (1). ⁸ , ²⁴

For in vivo ¹⁷O MRS brain application in rodents, which have smaller bodies and rapid pulmonary gas exchange, the contributions of blood recirculation (i.e., the C _a term in Equation (1)) and blood washout via perfusion (i.e., the C _v term in Equation (1)) during a short ¹⁷O₂ gas inhalation (e.g., 2–3 min) are similar, so the second term on the right‐hand side of Equation (1) becomes negligible, and this approximation leads to a linear differential equation ⁸ , ²⁴ :

$${\text{CMRO}}_2\approx \frac{d{C}_{\mathrm{b}}\left(t\right)}{\mathit{dt}}·\frac{1}{2·\alpha ·f_1},$$ (2)

where the dC _b(t)/dt term represents the slope of the brain tissue H₂ ¹⁷O concentration (i.e., [H₂ ¹⁷O]) change in phase II, and it can be determined by linear regression.

During phase III, the dynamic [H₂ ¹⁷O] change reflects the perfusion process in the brain tissue according to an exponential decay (measurable by the decay rate constant k) and can be used to estimate the CBF value according to Equation (3) ⁵ , ²⁶ :

$$\mathrm{CBF}\approx 1.86\times k,$$ (3)

where the constant k is determined by fitting the post‐inhalation $C_{\mathrm{b}}\left(t\right)$ time course to an exponential function. ⁵ , ²⁶ , ²⁷ The OEF value can be determined from the corresponding CMRO₂ and CBF values according to Equation (4) ²⁷ :

$$\mathrm{OEF}=\frac{{\text{CMRO}}_2}{C_{\mathrm{a},{\mathrm{O}}_2}\times \mathrm{CBF}},$$ (4)

where $C_{\mathrm{a},{\mathrm{O}}_2}$ is the arterial oxygen concentration, assumed constant at 18 mL O₂/dL blood (equivalent to 18/2.24 μmol/mL blood). Note that the simplified models as described by Equations (2) and (3) provide good approximations for rodent brain application. However, they are not applicable to the human brain owing to a much larger body size and much slower pulmonary gas exchange process; thus, both oxygen metabolism and perfusion processes can contribute to the dynamic [H₂ ¹⁷O] changes in phases II and III, and a sophisticated quantification model is needed. ²⁸ , ²⁹ , ³⁰

For in vivo ²H MRS application in rat brains, the original kinetic model does not include the glycolysis pathway and CMR_Lac owing to a low level of ²H‐labeled lactate in healthy rat brains. ⁶ In this study, we modified the previous kinetic model ⁶ by replacing the glycogen pathway with the lactate formation pathway as illustrated in Figure S2. Using the modified kinetic model and the arterial blood input functions reported in our early rat brain work using the same D66 infusion protocol and anesthesia condition (2% isoflurane), ⁶ we estimated three key cerebral metabolic rates: CMR_Glc, CMR_Lac, and V _TCA. This involved fitting the concentration time courses of deuterated metabolites (glucose, lactate, and the combined glutamate/glutamine: Glx) to the modified kinetic model (Figure S2).

Ultimately, through a single dynamic interleaved ²H–¹⁷O MRS measurement, six key physiological parameters, CMRO₂, CBF, OEF, CMR_Glc, V _TCA, and CMR_Lac, were simultaneously measured and determined.

3. RESULTS

3.1. RF coil design, prototype, and bench test

Figure 2A illustrates the circuit layout of the newly designed RF surface coil that can be operated at ¹H/²H/¹⁷O triple resonant frequencies. Figure 2B displays the prototype of this coil. Figure 2C presents the S11 measurement outcomes for the prototype coil when it was tested under the heavy loading condition using a large water phantom. The RF coil quality (Q) factors, derived from the S11 parameters under the loaded condition, were calculated to be 23.3 for the ¹H resonant frequency, 20.6 for the ²H resonant frequency, and 20.6 for ¹⁷O resonant frequency.

3.2. Design of the interleaved ²H–¹⁷O MRS pulse sequence and phantom validation

Figure 3A depicts a schematic of a modified single‐pulse acquisition (SPULS) sequence used to collect five ¹⁷O FIDs followed by one ²H FID in an interleaved way for each element and then repeated 20 times for signal averaging. In this configuration, the TR for acquiring ¹⁷O signals was set to 50 ms; however, as the ¹⁷O MRS acquisition was repeated five times, the effective TR of ²H was 300 ms (=6 × 50 ms). The total acquisition count for acquiring the interleaved ²H–¹⁷O MRS was set to 20, yielding a temporal resolution of 6 s for the phantom test. This count can be adjusted to accommodate different experimental designs or SNR requirements. ²H FIDs were acquired when the PIN diode was reverse‐biased in an off state and the RF hard pulse was set to the ²H Larmor frequency (107 MHz at 16.4 T) with an Ernst FA, whereas ¹⁷O FIDs were collected when the PIN diode was forward‐biased in an on state and the RF hard pulse was set to the ¹⁷O Larmor frequency (94 MHz at 16.4 T) with a nominal 90° FA. Figure 3B illustrates a series of water spectra acquired from a phantom using the interleaved ²H–¹⁷O MRS pulse sequence (Figure 3A), with the upper panel (black lines) showing the ²H spectra of naturally abundant HDO and the lower panel (red lines) showing the ¹⁷O spectra of naturally abundant H₂ ¹⁷O from the water phantom. The ²H signal exhibited greater fluctuation compared with the ¹⁷O signal, which was acquired under full relaxation conditions and had a signal average five times greater than that of the ²H signal, thus a better SNR and a smaller fluctuation. Figure 3C displays the ¹H MRI of a spherical water phantom in three orientations. Figure 3D compares the averaged ²H and ¹⁷O water resonance spectra acquired in an interleaved manner with those acquired individually in the conventional manner (i.e., the control). The well‐matched control and interleaved spectra of ²H and ¹⁷O confirm the excellent performance of the multifrequency surface coil and the modified pulse sequence. Additionally, to prevent mixing up of the ²H and ¹⁷O water signals in the interleaved scans, we intentionally shifted the ¹⁷O water signal in the receiver frequency by approximately 400 Hz.

3.3. Rat brain experiment with bolus injection of D₂O and H₂ ¹⁷O

Figure 4A,B shows the results of the dynamic ²H–¹⁷O MRS measurements before, during, and after the concurrent intravenous injection of D₂O and H₂ ¹⁷O tracers. The stack plots depict the HDO and H₂ ¹⁷O spectra, presenting averaged labeled water signals with a temporal resolution of 1 min per spectrum. Exponential fitting of the post‐injection water signals yields decay rate constant (k) values of 0.37 and 0.38 min⁻¹ for HDO and H₂ ¹⁷O measurements, respectively, reflecting the perfusion process in the rat brain. Although the HDO signal intensity is significantly higher than that of H₂ ¹⁷O due to the higher deuterium enrichment (or dose) of the D₂O tracer, the measured decay rate constants of the HDO and H₂ ¹⁷O signals are almost identical as anticipated. Note that the accurate quantification of the actual CBF values ideally requires the blood input functions after the infusion of tracers, which were not measured in this study.

FIGURE 4.

Dynamic ²H and ¹⁷O MRS results obtained from a rat brain during a 6‐min intravenous infusion of D₂O and H₂ ¹⁷O water, with the onset of infusion indicated by the blue arrows. The dynamic spectra of averaged HDO (A) and H₂ ¹⁷O (B) resonance signals are presented at a temporal resolution of 1 min per spectrum. The decay rate constants (k) for the post‐infusion HDO and H₂ ¹⁷O signals were determined as 0.37 and 0.38 min⁻¹, respectively. Stack plots of the dynamic brain ²H spectra (C, D) and ¹⁷O spectra (E, F) acquired continuously over 80 min in two rats (C, E for rat A and D, F for rat B), undergoing simultaneous intravenous administration of D66 glucose and inhalation of ¹⁷O₂ gas at 5–7 min, followed by another ¹⁷O₂ gas inhalation at 35–37 min.

3.4. Rat brain experiments with D66 glucose administration and ¹⁷O₂ gas inhalation

Two rats used in this experiment exhibited some differences in their BW, end tidal CO₂ (EtCO₂), and body temperature (T): BW = 335 g, EtCO₂ ≈ 2.7%–3.0%, and T ≈ 38.0–38.2°C for rat A and BW = 254 g, EtCO₂ ≈ 3.5%–3.8%, and T ≈ 38.6–38.8°C for rat B. The stack plots of the dynamic brain ²H and ¹⁷O spectra obtained during the 80‐min interleaved ²H–¹⁷O MRS data acquisition in two rats are shown in Figure 4C–F. Both rats underwent simultaneous D66 glucose intravenous administration and ¹⁷O₂ gas inhalation after 5‐min baseline acquisition and a subsequent ¹⁷O₂ gas inhalation 30 min after the D66 glucose injection. The quality of the ²H–¹⁷O spectra and the dynamic changes in the ¹⁷O‐labeled water (H₂ ¹⁷O) signal and ²H‐labeled water (HDO), glucose (Glc), mixed glutamate and glutamine (Glx), and lactate (Lac) signals are clearly captured and demonstrated herein. Subsequent data analysis yielded concentration (presented in millimolar unit) time courses of H₂ ¹⁷O with ¹⁷O₂ inhalations as measured by ¹⁷O MRS (Figure 5), as well as Glc, Glx, and Lac after an intravenous infusion of deuterium‐labeled glucose as measured by ²H MRS (Figure 6), and they were utilized in model fitting to estimate the relevant physiology parameters of interest.

FIGURE 5.

Time courses and model regressions of brain tissue H₂ ¹⁷O concentration and the corresponding cerebral blood flow (CBF) and cerebral metabolic rate of oxygen consumption (CMRO₂) values for two rats ((A) rat A and (B) rat B) during two repeated ¹⁷O₂ inhalations. The shaded areas indicate the durations of ¹⁷O₂ inhalation, and the blue and red lines depict the linear and exponential fittings of the experimental data (star symbols) used to calculate the CMRO₂ and CBF values according to Equations (2) and (3), respectively. D66, d‐glucose‐6,6‐d₂.

FIGURE 6.

Time courses and model fittings of brain deuterated glucose (Glc), glutamate/glutamine (Glx), and lactate (lac) concentrations in two rats ((A) rat A and (B) rat B) during and after d‐glucose‐6,6‐d₂ (D66) administration (indicated by shaded areas) using the modified kinetic model (Figure S2). The derived cerebral metabolic rate of glucose consumption (CMR_Glc), cerebral metabolic rate of the tricarboxylic acid cycle (V _TCA), and cerebral metabolic rate of lactate formation (CMR_Lac) values from the model fitting were 0.60, 1.16, and 0.05 μmol/g/min for rat A and 0.67, 1.17, and 0.17 μmol/g/min for rat B, respectively.

CMRO₂ and CBF values were calculated by regression of the brain [H₂ ¹⁷O] time courses according to Equations (2) and (3), respectively, as summarized in Figure 5, showing moderately higher CBF values in rat A (0.83 and 0.54 mL/g/min for the first and second measurements) than in rat B (0.66 and 0.37 mL/g/min) and slightly lower CMRO₂ values in rat A (1.26 and 1.19 μmol/g/min) than in rat B (1.36 and 1.24 μmol/g/min). Interestingly, both rats exhibited a decrease in CBF (34.9% in rat A and 43.9% in rat B) and a smaller reduction in CMRO₂ (5.6% in rat A and 8.8% in rat B) about 30 min after D66 administration, suggesting a potential hyperglycemia effect of glucose administration or a potential decline in animal physiology condition. Nevertheless, further investigation is necessary to confirm this observation. The averaged OEF was 0.28 in rat A, significantly lower than the OEF of 0.40 in rat B, which may be due to their different physiological conditions. The relatively low (global) CMRO₂ values observed in this study can be attributed to the partial volume effect caused by the large RF coil size relative to the rat brain, resulting in significant contamination from surrounding tissues such as muscle with several times low metabolic activity. ³⁰

Figure 6 displays the concentration dynamics of deuterated metabolites fitted to the modified kinetic model (Figure S2), demonstrating excellent fitting quality. The derived values of CMR_Glc and V _TCA were similar between rat A (0.60 and 1.16 μmol/g/min, respectively) and rat B (0.67 and 1.17 μmol/g/min, respectively), although rat B showed a higher CMR_Lac value compared with rat A (0.17 μmol/g/min vs. 0.05 μmol/g/min for rat A) and a higher Glx turnover.

Table 1 summarizes all quantified physiological parameters from two rats with mean values of CMRO₂ = 1.26 μmol/g/min, CBF = 0.60 mL/g/min, OEF = 0.34, CMR_Glc = 0.64 μmol/g/min, V _TCA = 1.17 μmol/g/min, and CMR_Lac = 0.11 μmol/g/min. These values are in general consistent with the literature reports (e.g., in the reference ³¹ ).

TABLE 1.

Summary of measured rat brain metabolic rates, cerebral blood flow (CBF), and oxygen extraction fraction (OEF) results.

Rat	CMRO2 (μmol/g/min)	CBF (mL/g/min)	OEF	CMRGlc (μmol/g/min)	V TCA (μmol/g/min)	CMRLac (μmol/g/min)
A	1.26	0.83	0.28	0.60	1.16	0.05
	1.19	0.54
B	1.36	0.66	0.40	0.67	1.17	0.17
	1.24	0.37
Average	1.26	0.60	0.34	0.64	1.17	0.11

4. DISCUSSION

In a normal brain, most glucoses are metabolized through mitochondrial oxidative pathways to meet the high ATP energy demand required for various cellular activities and brain functions while maintaining ATP homeostasis. However, altered brain states may trigger metabolic reprogramming, leading to significant changes in the relative contribution of glycolysis to overall energy production. ³² , ³³ , ³⁴ , ³⁵ , ³⁶ It is widely recognized that brain cancer cells exhibit a significant increase in lactate production under aerobic conditions, a process known as the “Warburg effect.” ²⁷ , ³² , ³⁷ , ³⁸ , ³⁹ Furthermore, changes in aerobic glycolysis are implicated in various brain disorders, including Alzheimer's disease and other neurodegenerative conditions. ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ Interestingly, even in the healthy human brain, aerobic glycolysis is also enhanced in the human visual cortex during visual stimulation with a sufficient cellular oxygen level, resulting in an elevation in lactate and much large increases in CMR_Glc and CBF than CMRO₂. ³⁴ , ⁴⁷ , ⁴⁸ , ⁴⁹ Such a metabolic shift or enhanced aerobic glycolysis plays a crucial role in synaptic plasticity and adaptation learning ³⁵ , ³⁶ as well as brain development, ⁵⁰ and conversely, the contribution of aerobic glycolysis decreases with age at varied paces in different brain regions. ⁵⁰ , ⁵¹ The ample evidence highlights the fundamental role of metabolic reprogramming in healthy and diseased brains. ⁴⁷ , ⁵² Therefore, it is imperative to develop in vivo MRS and imaging techniques that can simultaneously measure and quantify the cerebral glucose and oxygen metabolic rates as well as blood flow.

In this work, we developed a novel RF surface coil approach based on a dual‐frequency coil design ²⁰ and a circuit controller ²¹ that enables ¹H MRI and interleaved ²H–¹⁷O MRS measurements. This design creates two primary resonant circuits in the same loop coil that cover a wide tuning range from the ¹H frequency of 698 MHz to the ¹⁷O or ²H frequency of 94 or 107 MHz at 16.4 T. This approach offers a simple and cost‐effective RF coil solution that overcomes prevalent challenges in multi‐nuclei MRI/MRS imaging and RF coil design, ensuring adequate detection sensitivity for ¹H MRI and achieving excellent performance for low‐γ X‐nuclear MRS imaging (e.g., ¹⁷O and ²H) as the single‐frequency surface coil with the same coil size and geometry.²⁰ Essentially, this coil design allows the measurement of multiple low‐γ X‐nuclei, such as ²H and ¹⁷O, within the tuning range. However, in practical applications, the transition between different nuclear spins or operation frequencies requires manual adjustment. To address this issue and enable rapid and active tuning and retuning, a PIN‐diode circuit was integrated into the dual‐frequency coil design to facilitate the active switching between the ²H and ¹⁷O operation frequencies (Figure 2). This integration makes interleaved in vivo ²H and ¹⁷O measurements possible. The PIN diode acts as a nonlinear, current‐controlled resistor at the desired RFs, providing a resistance of 0.1–0.5 Ω in forward bias, effectively creating a short circuit for the RF current and increasing to several kiloohms in reverse bias, resulting in an open circuit. ⁵³ This property makes PIN diodes important components in transmit–receive (T/R) coils and switches and enables rapid switching between different resonant frequencies in multifrequency coils. ⁵⁴ , ⁵⁵ , ⁵⁶ It is particularly useful when switching between nuclei with closely spaced resonant frequencies by adding extra parallel tuning and matching capacitance (controlled by PIN didoes), as illustrated in Figure 2A. ⁵⁷ Our current coil design is specifically optimized for active frequency switching between ²H and ¹⁷O at 16.4 T, given the proximity of their resonant frequencies (13 MHz apart) and their alignment within the tuning range suited for the low‐γ resonant frequencies.

The performance of the novel RF surface coil was evaluated through phantom testing to detect natural abundance HDO and H₂ ¹⁷O water signals, as well as through a rat brain perfusion experiment. The latter utilized dynamic ²H–¹⁷O MRS measurements with simultaneous administration of D₂O and H₂ ¹⁷O dual tracers. The analysis revealed that the decay rate constants regressed from the interleaved ²H and ¹⁷O data sets are nearly identical, although that the intensity of the HDO signal was significantly higher than that of H₂ ¹⁷O due to the higher dose of deuterium label introduced into the same brain (see Figure 4A,B). This is expected because the decay rate reflects brain tissue perfusion and should be independent of the dose of tracer used.

CMRO₂, CBF, and OEF are three important parameters for evaluating cerebral oxygen metabolism and perfusion and the coupling between them, which can be determined by ¹⁷O MRS or imaging. ⁸ , ²⁶ , ⁵⁸ Likewise, CMR_Glc, CMR_Lac, and V _TCA are critical for the assessment of cerebral glucose metabolism along major metabolic pathways and can be measured by the recently developed in vivo ²H MRS or imaging technique. ⁶ The method developed in this work enables simultaneous determination of those six parameters, making it possible for the first time to perform a comprehensive and quantitative assessment of cerebral glucose and oxygen metabolism and perfusion from a single dynamic measurement, as demonstrated in Figures 4–6.

Although the novel RF coil and associated MRS acquisition and quantification methods described herein allow simultaneous and quantitative assessment of rat brain glucose and oxygen metabolism and perfusion at 16.4 T, the current study is limited by using only surface coil localization. The in vivo ²H and ¹⁷O MRS signals obtained using the RF coil (2.5‐cm diameter, Figure 2B) and the modified SPULS sequence (Figure 3A) should originate mainly from the rat brain; however, there is significant contamination from the surrounding muscle with lower metabolic activity, so the values reported in Table 1 may be affected by the partial volume effect. In particular, contamination may lead to an underestimation of CMRO₂ due to the significant effect of the recirculating H₂ ¹⁷O into the brain and surrounding tissues. Nevertheless, this may be less critical for measuring other metabolic rates based on the labeled metabolites (e.g., Glx), which remains primarily within the brain cells and absent in the surrounding muscle. Technical limitations of this study precluded imaging of the six parameters in different brain regions. Nonetheless, the interleaved ²H–¹⁷O single‐pulse acquisition sequence can be extended to a three‐dimensional (3D) interleaved ²H–¹⁷O chemical shift imaging (CSI) theme by adding phase encoding and dephasing gradients, for instance, as illustrated in Figure S3. Advanced post‐processing and denoising techniques can also be applied to the 3D CSI data to further improve SNR and achieve high‐resolution metabolic mapping of rodent brains. ⁵⁹ , ⁶⁰ Additionally, the interleaved ²H–¹⁷O MRS or imaging method could be adapted for human brain applications, which requires new hardware such as RF coils or coil arrays customized for the human brain imaging, or could be employed for a relatively lower field strength potentially for translational or clinical applications (e.g., at 3 T) with a decent imaging resolution. ⁶¹ , ⁶²

Although the measured metabolic rates and perfusion parameters in this study are within the ranges as reported in the literature, there are some limitations beside surface coil localization as discussed above. First, the dual‐perfusion tracers (HDO and ¹⁷H₂O) were injected into the body via the femoral vein rather than the internal carotid artery as reported in our early studies, ⁸ , ¹⁶ invariably involving tracer dilation in the heart and lung. Therefore, the measured decay rate constants as shown in Figure 4A,B are not accurate for directly determining the true CBF values, and additional information about blood input function is needed. Second, the blood input functions related to the total and deuterium‐labeled glucoses in blood were not measured during the dynamic measurements of interleaved ²H–¹⁷O MRS in this study; therefore, the reported CMR_Glc, CMR_Lac, and V _TCA values present approximated estimations. Finally, the number of animals used for this study was small and it limits the accuracy of the measurement and derived values. More studies are needed.

5. CONCLUSIONS

To address the technical hurdles of using ²H and ¹⁷O MRS technologies to simultaneously study cerebral glucose and oxygen metabolism and perfusion, we developed a novel surface coil paired with a modified pulse sequence and a circuit controller to achieve active switching between the ²H and ¹⁷O Larmor frequencies. With this capability, our study marks the first time that interleaved ²H–¹⁷O MRS technology combined with concurrent ²H‐labeled glucose administration and ¹⁷O‐labeled oxygen gas inhalation can simultaneously measure and quantify six key metabolic and perfusion parameters (CMR_Glc, CMR_Lac, CMRO₂, V _TCA, CBF, and OEF) in a single dynamic scan. This method not only saves scanning time but also improves measurement accuracy by minimizing variations associated with changes in the animal's condition commonly across multiple measurements. Looking forward, further developments, such as incorporating spatial localization or imaging capabilities, could greatly expand the utility of this technology in a wide range of biomedical applications, especially in exploring brain energy metabolism, metabolic reprograming, and their interplay with cerebral perfusion in healthy brains and aging‐related changes and more importantly in many disordered brains, including tumor, stroke, and neurovegetative diseases. Finally, the same interleaved ²H–¹⁷O MRS or imaging technology can be also applied to other organs or different field strengths.

AUTHOR CONTRIBUTIONS

Wei Chen and Xiao‐Hong Zhu designed the study; Guangle Zhang, Wei Zhu, and Xiao‐Hong Zhu conducted the study; Parker Jenkins built the RF coil and circuit; Guangle Zhang analyzed the data; and Guangle Zhang, Wei Chen, and Xiao‐Hong Zhu wrote the manuscript.

CONFLICT OF INTEREST STATEMENT

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

SUPPORTING INFORMATION

The supporting information for this article is available online.

Abbreviations: CMR_Glc, cerebral metabolic rate of glucose consumption; CMR_Lac, cerebral metabolic rate of lactate formation; CMRO₂, cerebral metabolic rate of oxygen consumption; V _TCA, cerebral metabolic rate of the tricarboxylic acid cycle.

Supporting information

Figure S1. The dynamics of brain H₂ ¹⁷O signal measured by in vivo ¹⁷O MRS can be characterized by three phases: (Phase I) steady‐state natural abundance ¹⁷O‐water signal prior ¹⁷O₂ inhalation; (Phase II) a brief period of ¹⁷O₂ inhalation (indicated by a grey bar, usually a few minutes), resulting in a rapid increase in the brain tissue H₂ ¹⁷O signal; (Phase III) post ¹⁷O₂ inhalation period, the metabolized H₂ ¹⁷O in the brain tissue begins to decay and can serve as an endogenous perfusion tracer.

Figure S2. The kinetic model modified from our previous publication (Lu et al. JCBFM, 2017; 37:3518–3530) for quantifying three cerebral metabolic rates of glucose consumption (CMR_Glc), lactate production (CMR_Lac) and TCA cycle (V_TCA) by analyzing the dynamics of Glc, Glx and Lac metabolites measured by in vivo ²H MRS. Glc_blood: blood glucose concentration; Glc_brian: brain glucose concentration; Pyr: pyruvate; Lac: lactate; Glx: mixed glutamate and glutamine; α‐KG: α‐ketoglutarate; V_out: lactate efflux rate; T_max and K_T are two glucose transportation parameters.

Figure S3. (A) The interleaved ²H/¹⁷O MRS (or FID) based on the modified single pulse acquire (SPULS) sequence as presented in this study. (B) The interleaved ²H/¹⁷O chemical shift imaging (CSI) pulse sequence that is adapt from the interleaved ²H/¹⁷O MRS sequence in (A) by inserting 3D (x, y and z) phase encoding (G_pe) and dephasing (G_dp) gradients. G_pe is for spatial encoding or imaging, and G_dp is for canceling the residues of the transversal magnetization of metabolites after the FID acquisition.

Zhang G, Jenkins P, Zhu W, Chen W, Zhu X‐H. Simultaneous assessment of cerebral glucose and oxygen metabolism and perfusion in rats using interleaved deuterium (²H) and oxygen‐17 (¹⁷O) MRS. NMR in Biomedicine. 2025;38(1):e5284. doi: 10.1002/nbm.5284

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.