Quantitative and simultaneous measurement of oxygen consumption rates in rat brain and skeletal muscle using ¹⁷O MRS imaging at 16.4T

Abstract

Purpose:

Oxygen-17 MRS imaging, successfully used in the brain, is extended by imaging the oxygen metabolic rate in the resting skeletal muscle and to determine the total whole-body oxygen metabolic rate in the rat.

Methods:

During and after inhalations of ¹⁷O₂ gas, dynamic ¹⁷O MRSI was performed in rats (n=8) ventilated with N₂O or N₂ at 16.4T. Time courses of the H₂¹⁷O concentration from regions-of-interest located in brain and muscle tissue were examined and used to fit an animal-adapted three-phase metabolic model of oxygen consumption. Cerebral blood flow (CBF) was determined with an independent washout method. Finally, body oxygen metabolic rate was calculated using a global steady-state approach.

Results:

Cerebral metabolic rate of oxygen consumption (CMRO₂) was 1.97 ± 0.19 μmol/g/min on average. The resting metabolic rate of oxygen consumption in skeletal muscle (RMRO₂) was 0.32 ± 0.12 μmol/g/min, and > 6 times lower than CMRO₂. Global oxygen consumed by the body (VO₂) was 24.2 ± 3.6 ml O₂/kg body weight/min. CBF was estimated to be 0.28 ± 0.02 ml/g/min and 0.34 ± 0.06 ml/g/min for the N₂ and N₂O ventilation condition, respectively.

Conclusion:

We have evaluated the feasibility of ¹⁷O MRSI for imaging and quantifying the oxygen consumption rate in low metabolizing organs such as the skeletal muscle at rest. Additionally, we have shown that CBF is slightly increased in the case of ventilation with N₂O. We expect this study to be beneficial to the application of ¹⁷O MRSI to a wider range of organs, though further validation is advised.

Introduction

Noninvasively measuring cellular oxygen metabolism using ¹⁷O₂ tracer and in vivo ¹⁷O spectroscopic imaging (¹⁷O MRSI) at ultrahigh field (UHF) is a promising tool for studying cellular energy metabolism and physiology.¹ The ¹⁷O imaging approach allows to quantify the cerebral metabolic rate of oxygen (CMRO₂) in human² and animal brain.^3–5 Although imaging studies have been performed to differentiate the cerebral metabolic rates between gray and white matter with ¹⁷O,^2,6,7 few measurements were done outside the brain,^8–10 often focusing on aerobic organs with a high metabolic rate. Obviously lower metabolic rates, such as in resting muscle tissues, result in slower turnover rates from ¹⁷O₂ to H₂¹⁷O in the mitochondria. Thus, less labeled H₂¹⁷O signal in tissue could necessitate longer scans or require more application of ¹⁷O-isotope labeled O₂ to achieve an adequate signal-to-noise ratio (SNR) for imaging. The dynamics of H₂¹⁷O signal change as well as a uncertainty in modeling with resulting longer inhalation durations is affected by blood perfusion and recirculation. This unmet challenge motivates our investigation of the feasibility of imaging the low oxygen metabolic rate in resting skeletal muscle using ¹⁷O MRSI in simultaneous comparison to the brain oxygen metabolism rates in the same subject. By using significantly longer ¹⁷O₂ inhalation times, the amount of generated H₂¹⁷O in biological tissues and the resulting ¹⁷O MR signal is largely increased and even more so by multiple inhalations,^7,11 allowing to reliably observe metabolic and perfusion¹² parameters at increased sensitivity. However, in advantage to ¹⁵O PET no subtraction scans have to be performed for the subtraction of gaseous oxygen signal.^13–16

In this study, we simultaneously acquired dynamic time courses of H₂¹⁷O signals in brain and muscle tissue in rats during ¹⁷O₂ inhalations using three-dimensional (3D) ¹⁷O MRSI at an ultrahigh magnetic field of 16.4 Tesla (T). Repetitive and longer inhalations of ¹⁷O₂ gas than in previous measurements in rodents resulted in a large increase of the H₂¹⁷O concentration several times above natural abundance. The H₂¹⁷O dynamic signals were fitted with two commonly used models: a three-phase metabolic model using to the whole H₂¹⁷O dynamic time course, acquired before, during and after inhalations to determine the metabolic rates of oxygen consumption², and a washout model was applied to the post-inhalation brain data to estimate the cerebral blood flow (CBF).¹² The three-phase metabolic model, which was previously used in human brain gray and white matter, was modified to obtain the low resting-state metabolic rate of oxygen consumption in the rat skeletal muscle (Muscle RMRO₂) for exploring the feasibility using the ¹⁷O MRSI method in other tissues. The washout technique, allowing for estimation of CBF, as previously validated in rodent brain, was used to investigate two groups of rats ventilated with different blends of gases (oxygen with N₂ or N₂O). Finally, recirculation of H₂¹⁷O leading to a new equilibrium at the end of the post-inhalation period was observed and employed to estimate the organism’s global metabolism rate (i.e., total body oxygen expenditure VO₂), which was then compared to the regional metabolic rates of oxygen consumption.

Theory

Three-phase model adaptation

A previously published model for determining the human brain oxygen metabolism rate² fitting three phases of a H₂¹⁷O time course (Phase 1: before, Phase 2: during and Phase 3 after an ¹⁷O₂ inhalation) was adapted to the rat systemic characteristics. We chose this model for the study in rodents since it applies well to low metabolic rates involving a significant amount of recirculating H₂¹⁷O. In particular with longer inhalations, cardiopulmonary factors like the cardiac output are increasingly important.²

The time dependent brain tissue H₂¹⁷O concentration defined as molar volume $M_V^{H_2^{17}O}(t)$ in an imaging voxel can be described as (refer to Eqn. [2] in Reference 2 for more details):

$$\frac{{\mathit{dM}}_V^{H_2^{17}O}(t)}{\mathit{dt}}=2\cdot {\mathit{CMRO}}_2\cdot A^{{}^{17}O_2}(t)-K_L\cdot M_V^{H_2^{17}O}(t)+K_G\cdot B^{H_2^{17}O}(t).$$ [1]

The three terms on the right side of Eqn. [1] can be separated into: 1) the regional metabolic activity producing H₂¹⁷O (i.e., the cerebral metabolic rate of oxygen consumption: CMRO₂), depending on the arterial ¹⁷O-isotope enrichment [$A^{{}^{17}O_2}$] of oxygen gas delivered through hemoglobin; 2) the loss [K_L] of H₂¹⁷O mainly due to (cerebral) blood flow or perfusion washout into the draining venous vascularity; and 3) the gain [K_G] of H₂¹⁷O through inflow of blood [$B^{H_2^{17}O}$] containing H₂¹⁷O, recirculating from both local metabolizing tissue and whole body oxygen metabolism.

By integration over time, Eqn. [1] can be used to fit the time courses (see Eqn. [6] in Reference 2) of H₂¹⁷O signal for each imaging voxel to derive the oxygen metabolic rate (MRO₂) in brain or muscle tissue (as CMRO₂ or RMRO₂). We propose herein that, in principle, for any sufficiently perfused organ, oxygen consumption rates even below the systemic global aerobic rate (VO₂) can be measured. The quantification is simplified if the water content of the imaged tissue, which can be calibrated by the H₂¹⁷O natural abundance concentration and the ¹⁷O signal measured in Phase 1, is known.^3–5 The water content of muscle and brain can be approximated by assuming comparability to humans (i.e., mice¹⁷: 74.4% wt in muscle vs. human:¹⁸ 79.5% wt in striated muscle and 73.3% wt in brain). Furthermore, the tissue density for rodents (1.06 kg/liter for skeletal muscle¹⁹) was employed for unit conversion. Eqn. [1] can then be used to determine the oxygen metabolic rates of the rodent muscle and brain.

Systemic oxygen expenditure VO₂

The total body oxygen expenditure or metabolic rate (VO₂ in the unit of μmol/g body weight/min) can be defined as the cumulative amount of metabolic H₂¹⁷O added to the organism by inhalation and metabolism of ¹⁷O₂ tracer with a fixed enrichment within a given inhalation time. The average body oxygen metabolic rate (VO_2,average) per minute can then be determined using the equilibrium H₂¹⁷O signal in tissue measured during the late part of the post-inhalation period (Phase 3) assuming that a new equilibrium (or steady state of the tissue H₂¹⁷O signal) has been established:

$${\mathit{VO}}_2=\frac{M_V^{H_2^{17}O}(\text{equilibrium})}{t_{\text{inhalation}}\times f}={\mathit{VO}}_{2,\text{average}}∕f(\mathrm{t}>>{\mathrm{t}}_{\text{inhalation}})$$ [2]

t_inhalation is the inhalation duration; $M_V^{H_2^{17}O}(\text{equilibrium})$ is the average tissue H₂¹⁷O concentration ($M_V^{H_2^{17}O}$) at equilibrium (in this study, ~3 inhalation durations after ¹⁷O₂ inhalation) when its pre-inhalation level is set to zero; the conventional format of VO₂ in volume of oxygen gas is usually given in ml/kg body weight/min and requires a unit conversion f by division of 0.0446 μmol⁻¹ml.²⁰

Methods

Simulation of circulation impact and metabolic rate on the H₂¹⁷O time courses

Inhalations with a ¹⁷O₂ enrichment of 70% were simulated using the previously outlined three-phase model for two settings: 1) simulation with a fixed high metabolic rate (i.e. isometabolic CMRO₂=2 μmol/g/min) varying only the circulatory parameters (K_G and K_L) in ranges reported in the literature^2,6,9,21,7 (Phase 2 with 15.25 min inhalation duration); and 2) simulation of varying metabolic rate and corresponding changes in perfusion. In the second stage simulation, different levels of local oxygen metabolic rates were set (MRO₂=2; 1; 0.5; 0 μmol/g/min) with fixed parameters K_L=0.2 and K_G=0.3, unless otherwise noted, during an inhalation using Eqn. [1] to qualitatively assess the time dependence of the tissue H₂¹⁷O signal. Specifically this allowed investigation under idealized conditions of the transitions between phases of the model.²

Furthermore, we adapted and evaluated the rodent specific systemic parameter (ρ_rat) as detailed in the Supporting Information.

Animal preparation and physiology monitoring

All procedures and experiments were approved by the local authorities (Regierungspräsidium) and were in compliance with the guidelines of the European Community (EUVD 86/609/EEC) for the care and use of laboratory animals. A total of 8 male Wistar rats (Charles River Laboratories, Sulzfeld, Germany) were used in this study (Table 1, mean body weight 312 ± 93 g). Artificial ventilation and maintenance of physiological stability is also further described in detail in the Supporting Information.

Table 1.

Summary of performed inhalation numbers and inhalation durations and weight for each animal.

Individual animal	gas mixture	Body weight [g]	Inh. number	Inh. Duration [min]
rat A	N2	300	#1	15.3
	N2		#2	15.3
rat B	N2	232	#1	15.3
rat C	N2	275	#1	15.1
	N2		#2	15.1
rat D	N2	233	#1	15.1
	N2		#2	12.9
rat E	N2O	250	#1	15.4
	N2O		#2	15.1
	N2O		#3	15.1
rat F	N2O	510	#1	15.1
rat G*	N2O	332	#1	15
rat H*	N2O	367	#1	15
Pop. Mean ± SD:		312 ± 93 g		15 ± 0.6 min

Each row represents one resting ¹⁷O₂ inhalation measurement, which for rats A, C, D and E was repeated multiple times within the same experimental session per animal.

Ventilation mixtures with enriched ¹⁷O₂ gas (Oxygen gas fraction ~25-35% with 70% enriched ¹⁷O₂ Nukem GmbH, Germany) were prepared in non-diffusive gas bags (Hans Rudolph, Inc., Shawnee KS, USA). Oxygen was mixed with N₂ in one group (Table 1, animals A-D, n=4) and with N₂O in a second group of animals (Table 1, animals E-H, n=4). At the end of each experiment, the animals were euthanized followed by post-mortem imaging as previously reported.²²

MRI instrumentation and data acquisition

Magnetic resonance imaging was performed on a BioSpec Avance III system (Bruker Biospin MRI GmbH, Ettlingen, Germany) using a 26 cm bore 16.4 Tesla magnet and gradients with 12 cm inner diameter, 1 T m⁻¹ maximum strength and 212 μs ramp time (Resonance Research Inc., Billerica, MA, USA). Custom-built quadrature surface coils (elliptical loops each ~1.5×1.2 cm) were tuned to the ¹⁷O Larmor-frequency (94.6 MHz) for ¹⁷O imaging and a separate ¹H butterfly RF coil passively decoupled from the ¹⁷O coils was used. Anatomical ¹H MRI FLASH images with TR=2 s, TE=10 ms (nt=4 averages), 59×59 μm² in-plane resolution and 29 axial slices (thickness=1 mm) were acquired within 25 min 36 s.

A k-space acquisition-weighted 3D CSI pulse sequence was used for all ¹⁷O MRSI acquisitions. Two types of time-series were acquired for each rat: natural abundance tissue H₂¹⁷O signal before any ¹⁷O₂ gas inhalation for calibration of H₂¹⁷O concentration in each CSI voxel, and during and after a single or repeated inhalation for metabolic rate and CBF analysis. In all in vivo ¹⁷O MRSI acquisitions, we used a field of view (FOV) of 27.5×12.5×18 mm³, spectroscopic sampling points 375, and acquisition duration of 3.75 ms with a delay of 0.538 ms from an excitation RF pulse. TR was 4.92 ms, optimized for tissue T₂*,²³ and RF-excitation was performed with a 68° hard pulse of 200 μs duration.

In the majority of animals (Table 1, animals A-F, referred to as “high-resolution protocol”) the FOV was scanned by an acquisition matrix of 15×7×7, resulting in a voxel volume of 43.1 μl as defined by the width of the spatial response function (SRF).^24–26 Each 3D ¹⁷O CSI volume was acquired within 30.2 s, with a maximum number of averages nt _max=74 at the k-space center (a total of 6144 FIDs or 735 k-space points per CSI volume). Fifty natural abundance H₂¹⁷O CSI were acquired within ~25 min at baseline, and a total of 109 volumes per full inhalation, started shortly before it, were collected within 54 min 57 s (see Table 1 for individual inhalation durations) including a ~38 min long post-inhalation acquisition (i.e., the H₂¹⁷O washout period).

In a subgroup of 2 animals (Table 1, animals G & H, referred to as “low-resolution protocol”), the same FOV was scanned with an acquired matrix of 9×7×7, leading to a voxel size of 77.3 μl by SRF adjustment with nt_max=45 averages at the k-space center (a total of 2048 FIDs or 441 k-space points per CSI volume) and 10.1 s acquisition per 3D CSI volume. Natural abundance H₂¹⁷O CSI volumes (n=50) were acquired within ~8 mins 24 s and the same acquisition duration of 54 min 57 s was used to acquire 327 volumes of inhalation data. Other acquisition parameters remained the same.

Post-mortem CSI-acquisitions were performed without k-space weighting (12 ms TR and 70° flip angle) and with a pulse length of 400 μs. A FOV of 27.5×12.5×25 mm³ was sampled with a matrix of 41×19×25 voxels (nominal voxel size 0.44 μl). Approximately a total of 2.5 million FIDs with 1000 points each and a spectral bandwidth of 100 kHz were acquired in 8 h 18 min.

Brain co-registration and tissue selection

The 3D ¹⁷O-CSI data were co-registered with ¹H anatomic images and high resolution (post-mortem) H₂¹⁷O images with the same FOV as illustrated in Figures 1A–C. Equally-sized regions of interest (ROI) were selected (−3 mm Bregma)²⁷ for brain and in lateral muscle compartments in the same coronal slices. The topography of the temporalis muscle was verified anatomically^28–30 and left and right lateral ROIs (42.4 μl, n=40 voxels after zero-filling in animals A-F, 49.4 μl, n=28 voxels after zero-filling in animals G-H) were chosen as a subset of the temporalis volume (0.422 ml)³¹ carefully avoiding partial volume contamination from adjacent brain tissue.

FIGURE 1.

Co-registered rat head images with 5mm scale bar: (A), ¹H FLASH structural image of the coronal rat head anatomy. Regions of interest (ROI) for brain tissue (green) are highlighted for each hemisphere and laterally left and right for muscle tissue (i.e. predominantly temporalis lateralis; red); (B), in vivo H₂¹⁷O image acquired within 30 s shortly before the onset of a ¹⁷O₂ inhalation, with intensity highlighting (2× brighter) of the ROIs for better visualization; and (C), very high resolution H₂¹⁷O enriched post-mortem image acquired after repeated inhalations with ROIs marked as in (A). All slices cover the same FOV at the Z-position of the Bregma. Visualization of brain and muscle (dotted line) based on proton images (A). ¹⁷O images in (B,C) are zero-filled (×2) in the spatial dimensions.

Post processing, in vivo T₂* estimation and metabolic fitting

Acquired CSI datasets were Fourier-transformed and the peak of the magnitude spectrum of H₂¹⁷O after apodization (T₂* = 1.8 ms in time domain) was normalized to the natural abundance H₂¹⁷O concentration of 16.3 μmol/g wet tissue of both muscle and brain, assuming equal H₂¹⁷O concentrations (i.e., water content) in muscle and brain tissue^18,32–34. Calibration was performed through normalization from the previously defined natural abundance acquisitions of each rat, last 20 CSI volumes for the high resolution protocol, 40 volumes for low resolution, and pre-inhalation time points (Phase 1) were 12 CSI volumes and 23 volumes, respectively. Signals were smoothed by a nearest neighbor moving average (three adjacent CSI time points).⁷

Separately, for each rat, the data of two rats in the low and high resolution protocols were phased and the localized semilogarithmic FIDs were fitted against time for in vivo T₂* relaxation measurement as described in detail in Ref. 22.²²

The metabolic model was fitted according to Eqn. [1] using a non-linear least-squares algorithm (Curve fitting toolbox, Matlab) to the H₂¹⁷O signal time courses of tissue signal (inhalation time (t) as independent variable; CMRO₂ for brain and RMRO₂ for muscle, K_G, K_L as dependent variables).

Estimation of CBF

CBF was estimated from the same brain ROIs, only based on the H₂¹⁷O signal after the end of the inhalation (i.e., ~38 min of washout). The previously validated washout model¹² is based on the return of local H₂¹⁷O overproduction to a new systemic equilibrium in relation to the rest of the body (VO₂). The concentration of brain tissue H₂¹⁷O using mono-exponential fitting against time courses can be described by the following equation:^12,35

$$\mathit{Cb}(t)=k3\times \mathit{exp}\frac{\mathit{CBF}\times t}{k_1}+k_4.$$ [3]

The primary decay constant, proportional to CBF/k1, can be converted by multiplication with 1.86 to absolute CBF units of ml/g/min (whereas k3 and k4 are scale factors).¹² Then, the two groups with different ventilation mixtures were compared (N₂ vs. N₂O).

All results are reported in mean ± standard deviation (SD).

Results

Proton structural images showed a clear anatomical contrast between brain and muscle tissue (Figure 1A). Coregistered geometry of ¹⁷O contrast in both in vivo (Figure 1B) and ex vivo ¹⁷O high-resolution images (Figure 1C) matched the anticipated intensity distribution of the ¹⁷O surface coil, i.e., stronger ¹⁷O water signal at the surface and in the quadrature B₁ field overlap region in the brain. Figure 2 illustrates representative natural abundance H₂¹⁷O spectra summed over the ROIs before inhalation from low-resolution (Figure 2A, 10 s acquisition averaging) and high-resolution (Figure 2B, 30 s acquisition averaging) ¹⁷O MRSI, indicating a high SNR offered at 16.4T, in particular, in the brain.

FIGURE 2.

Representative natural abundance H₂¹⁷O spectra summed over the brain and muscle ROIs in one CSI volume with: (A), lower spatial resolution (10 s acquisition averaging) and (B), higher spatial resolution (30 s acquisition averaging) protocol. Stacked displays of phase corrected spectra are shifted to the vertically and horizontally for visualization (Note, the H₂¹⁷O signal was heterogeneously affected in SNR because of the B₁ sensitivity profile of the two ¹⁷O surface coils being used).

Simulation of parameterized H₂¹⁷O dynamics

The simulation results shown in Figure 3A demonstrate the sensitivity of the parameters of the three-phase model, in particular the K_L or K_G values on the H₂¹⁷O dynamics, which represents the strong influence of perfusion. Time courses of the simulated ¹⁷O signal with varying metabolic rate are shown in Figure 3B for four different metabolic rates, exemplifying representative values for the brain and the muscle.

FIGURE 3.

(A) A simple iso-metabolic comparison of varying degrees of perfusion and circulatory parameters (i.e., the K_G and K_L parameter values) using the three-phase model (Eqn.[1]) highlighting phases in high metabolizing tissue: Phase 1 - pre-inhalation natural abundance H₂¹⁷O; Phase 2 - during inhalation with initial increases dominated by locally produced H₂¹⁷O (Phase 2A as the first ~2-4 min) and subsequently varied slopes (especially in later Phase 2B) from the different perfusion and circulation parameters; Phase 3 – post-inhalation H₂¹⁷O with varying washout rates to reach equilibrium levels that were linearly affected by the K_G / K_L ratio at the same local metabolic rate.

(B) Zoomed (~2x) in time three-phase model plots at different metabolic rates: during the first few minutes, the H₂¹⁷O water content in low metabolizing tissue increases much slower than that in the higher ones (Phase 2A), signals approach a similar slope during Phase 2B and at the end of the inhalation, the H₂¹⁷O in low metabolizing tissue continues to rise with gradually decreased slope; both high and low metabolizing tissue approach the same equilibrium due to recirculation of body water at a new steady-state level determined by the global metabolic rate (VO₂, according to Eqn. [2]). Unless otherwise stated in the legend, all time courses in (B) had K_G=0.3 and K_L=0.2.

The dashed and dotted lines indicate the beginning and end of the Phase 2, respectively.

The simulated metabolic rates at different levels showed a qualitatively distinct shape of the H₂¹⁷O signal dynamics at low metabolism (i.e., sigmoidal). Despite significant differences in the early Phase 2 (Phase 2A), the slopes converge in a non-linear way during the late Phase 2 (Phase 2B) as shown in Figure 3B. The simulation results indicate that the early dynamic change of the tissue H₂¹⁷O signal after inhalation of ¹⁷O₂ gas is more sensitive to the local metabolic rate than that of late Phase 2.

A novel observation from this simulation was that the same K_G/K_L ratio leads to the same equilibrium level of H₂¹⁷O signal at the end of Phase 3 (Figure 3A for brain and 3B for muscle at K_G/K_L ratio = 1.5). This suggests that even if the oxygen metabolic rates vary greatly in different tissues (e.g., brain vs. muscle), the relative contributions of the H₂¹⁷O signal gain and signal loss due to recirculation and perfusion in different voxels remain the same. Thus, the voxels containing different tissue types will eventually reach the same H₂¹⁷O concentration level.

Metabolic rate estimates for brain and resting skeletal muscle tissue

As shown in Figure 4, the H₂¹⁷O signal intensity in muscle ROIs grew in a slower fashion, then accelerated during the late Phase 2 (i.e. Phase 2B) before approaching a saturation after the inhalation ended. An absent H₂¹⁷O signal decrease in muscle tissue during the post inhalation phase due to competing processes between H₂¹⁷O recirculation and washout was in stark contrast to the obvious H₂¹⁷O signal decay observed in the brain ROIs (Figure 4B). Reproducible time courses were observed during three repeated inhalation measurements in the same animal and MR imaging session (Fig. 4C).

FIGURE 4.

(A) Time courses of raw H₂¹⁷O signal intensity in one representative rat during and after a 15-minute inhalation of ¹⁷O₂ gas in ROIs of brain (green) and muscle (red) tissues; and (B) post-processed molar H₂¹⁷O concentration time courses of the same data. The dashed and dotted vertical lines highlight the beginning and end of the ¹⁷O₂ inhalation. Non-linear least-square fits of the three-phase metabolic model (continuous lines) yielded a brain CMRO₂ of 1.94 μmol/g tissue/min and muscle RMRO₂ of 0.32 μmol/g tissue/min. An equilibrium value of around 17 μmol/g tissue over the pre-inhalation H₂¹⁷O concentration in both tissue types was confirmed by matching 99% prediction bounds of the fit (dashed surrounding lines). The baseline H₂¹⁷O concentration was set at zero, which was above natural abundance level due to a prior ¹⁷O₂ inhalation. Finally, a comparable fit quality was shown by the very similar spread of the confidence intervals despite their significantly different metabolic estimates. (C) Normalized display of multiple inhalations showing the reproducibility of the technique in the same animal. Note the reproducible convergence at the end of each inhalation time course despite the differences in metabolic rate. With longer experimental duration, maintaining stable anesthetic conditions becomes more challenging as also reflected in higher signal fluctuations.

Fitting the metabolic rates of brain ROIs, an overall average of CMRO₂ = 1.97 ± 0.19 μmol/g/min (n=26 ROIs from all 8 rats) was determined. For the two subgroups consisting of 4 rats each a CMRO₂ of 2.07 ± 0.15 (n=14 ROIs) and slightly lower 1.84 ± 0.14 μmol/g/min (n=12 ROIs) were estimated with N₂ and N₂O, respectively (Table 2), and no significant differences between the two hemispheres were detected. In muscle ROIs, an average RMRO₂ of 0.32 ± 0.12 μmol/g/min (n=21 ROIs) was determined with some notable intra-subject left and right lateral differences. The estimated muscle oxygen metabolic rates were only a sixth of that of the brain. The perfusion and diffusion related parameter K_G was higher than the parameter K_L for both tissue types (for brain: averaged K_G=0.34 ± 0.05, K_L =0.22 ± 0.03, n=26; and for muscle: K_G=0.63 ± 0.33, K_L=0.40 ± 0.17, n=21). Group averages for brain tissue were K_G=0.34 ± 0.04 (n=14) for N₂ and K_G=0.34 ± 0.07 (n=12) for N₂O without a statistically significant difference. In contrast, K_L=0.20 ± 0.02 (n=14) for brain within the N₂ group was increased by +22% to K_L=0.24 ± 0.02 (n=12) in the N₂O group with statistical significance (two-sided unpaired t-test at p<0.005).

Table 2.

Summary of the oxygen metabolic rate (μmol/g tissue/min) results measured in brain and muscle based on region of interest (ROI) analysis

Individual animal	RMRO2 Left muscle ROI	RMRO2 Right muscle ROI	CMRO2 Left S1 ROI	CMRO2 Right S1 ROI
rat A	0.15	0.11	2.11	2.08
	0.35	0.20	2.14	2.01
rat B	0.28	0.43	2.03	1.90
rat C	0.27	0.30	2.02	2.08
	0.33	0.31	2.06	2.01
rat D	0.32	*	2.30	2.45
	0.37	*	1.85	1.96
Mean ± SD	0.30 ± 0.07	0.27 ± 0.12	2.07 ± 0.14	2.07 ± 0.18
rat E	0.45	0.34	1.82	1.77
	0.45	0.16	1.82	1.83
	0.46	0.42	1.68	1.74
rat F	0.13	*	2.14	2.08
rat G**	0.44	0.57	1.86	1.89
rat H**	0.34	0.20	1.80	1.69
Mean ± SD	0.38 ± 0.11	0.34 ± 0.14	1.85 ± 0.15	1.83 ± 0.14
Mean values of right and left ROIs:	RMRO2=0.32 ± 0.12***		CMRO2=1.97 ± 0.19***

^*No convergence of the fitting procedure.

^**This subgroup of 2 animals was acquired at a higher temporal resolution (10s per 3D CSI volume) with the lower spatial resolution protocol.

^***p < 0.01 significant tissue-type difference between muscle and brain (paired t-test).

The overall ratio of K_G/K_L determined within sessions was 1.51 ± 0.23 (n=21 ROIs) for muscle and 1.58 ± 0.23 (n=26 ROIs) for brain tissue, respectively; no statistically significant difference between the two tissue types was observed. The same K_G/K_L ratios between the brain and muscle converged to the same level of equilibrium H₂¹⁷O signal at the later Phase 3 (Figure 4B) despite > 6 times of difference in the metabolic rate between the two tissues. This finding is in agreement with the prediction from the simulations shown in Figure 3.

Cerebral blood flow and VO₂ in N₂O vs. N₂ ventilated animals

The estimated average oxygen metabolic rate of the entire body per gram tissue VO_2,average according to Eqn. [2] was 1.08 ± 0.20 μmol/g/min (n=13) and 1.08 ± 0.16 μmol/g/min (n=13) as inferred from right and left averaged muscle and brain ROI time courses, respectively (Table 3). Consistent with the conventional unit commonly used in the literature, VO₂ was converted to 24.2 ml/kg body weight/min, derived from the steady-state H₂¹⁷O signals from both tissue types. Cerebral blood flow in the N₂ ventilated animal brain from average k1 = 0.15 ± 0.01 (n=14) resulted in CBF 0.28 ± 0.02 ml/g/min and significantly elevated CBF (+21%, p<0.005 with two-sided unpaired t-test) was observed in the N₂O ventilated group with a mean of 0.34 ± 0.06 ml/g/min (n=12, Table 3).

Table 3.

Summary of VO_2,average (μmol/g tissue/min) and cerebral blood flow (ml/g tissue/min) results based on washout in brain.

Individual animal	VO2,average Left & Right Muscle ROI	VO2,average Left & Right Brain ROI	CBF Left Brain ROI	CBF Right Brain ROI
rat A	1.17	1.19	0.27	0.27
	1.19	1.17	0.28	0.26
rat B	1.21	1.04	0.28	0.25
rat C	1.19	1.16	0.26	0.28
	1.06	1.09	0.31	0.28
rat D	1.39	1.37	0.30	0.30
	1.15	1.14	0.28	0.26
Mean ± SD	1.19 ± 0.10**	1.17 ± 0.10**	0.28 ± 0.02**	0.27 ± 0.02**
rat E	1.14	1.11	0.41	0.33
	1.03	1.05	0.34	0.25
	1.02	0.95	0.29	0.29
rat F	0.64	0.91	0.32	0.33
rat G*	1.17	1.13	0.32	0.32
rat H*	0.74	0.73	0.46	0.38
Mean ± SD	0.96 ± 0.22**	0.98 ± 0.15**	0.36 ± 0.06**	0.32 ± 0.05**
Overall average of both N2 & N2O	VO2,average (n=13) 1.08 ± 0.20 μmol/g body/min	VO2,average (n=13) 1.08 ± 0.16 μmol/g body/min	CBF (n=26) 0.30 ± 0.05 ml/g tissue/min

^*This subgroup of 2 animals was acquired at a higher temporal resolution (10s per 3D CSI volume) and with lower spatial resolution protocol.

^**p<0.05 significant population difference between N₂ (rats A-D) and N₂O (rats E-H) groups (unpaired t-test).

In vivo T₂* in muscle and brain tissue

Figure 5 shows the lower T₂* of H₂¹⁷O in muscle tissue and ~40% higher T₂* in brain tissue, that are correlated against the independent metabolic rates in the two types of tissues. In the same rats a more than 5-fold difference in metabolic rate between muscle and brain is apparent.

FIGURE 5.

Scatter plot of in-vivo tissue T₂* vs. metabolic rate showing high correlation in both tissue types of brain and muscle from clearly separated clusters of the independent properties of metabolic rate and relaxometric behavior inside ROIs. The distinction benefits from the fact of a stark difference in T₂* as also previously reported for both in vivo and post-mortem tissues. Each cluster is based on the pooled tissue type of two representative rats (rat A and rat G), of each high resolution and low resolution protocols, respectively. Note: this is only an observation to confirm the accurate selection and size (i.e. partial volume contamination) of the ROIs and does not imply a causal relation between relaxation rate and metabolic rate in either direction.

Discussion

This study demonstrates three perspectives about the utility of the noninvasive and quantitative ¹⁷O MRSI or MRI method with inhalation of ¹⁷O₂ gas determining the oxygen consumption rates in organs with higher and/or lower metabolic activity, measuring the systemic global oxygen consumption rate using a steady-state model, and characterizing local cerebral blood flow under two experimental conditions.

Modeling dynamics of H₂¹⁷O signal in muscle and brain

We have simulated the H₂¹⁷O signal dynamics using an animal-adapted three-phase metabolic model as described by Eqn. [1] using different parameter settings to mimic the experimentally measured H₂¹⁷O time courses (see examples in Figure 3). The simulation data indicate that the initial change of the tissue H₂¹⁷O signal during the early inhalation period (Phase 2A) is dominated by the metabolically produced H₂¹⁷O and the contribution from recirculating H₂¹⁷O is small. Therefore, the initial slope of the H₂¹⁷O concentration in Phase 2A is sensitive to the oxygen metabolic rate of the tissue,³⁶ which is much slower in muscle as compared to the steeper increase in brain tissue. Despite the expected differences in the local metabolic rate, the time course of the H₂¹⁷O signal in the late inhalation phase (Phase 2B) converged to a relatively similar slope for all tissues (see Figures 3, 4 and also Supporting Information Figure S1B). The contribution of recirculating water increased with inhalation time and gradually dominated the H₂¹⁷O signal in the later phase of the inhalation, resulting in converging slopes between high and low activity tissues as observed in experimental data.

Determining the oxygen metabolic rates in muscle and brain

Despite the limited spatial specificity, arterio-venous difference measurements can still be regarded as the gold standard for oxygen consumption measurements. However, due to their invasiveness they are less convenient and the variability of draining vascular territory effects on reproducibility motivates the use of non-invasive alternatives like ¹⁷O MRSI/MRI with ¹⁷O₂ tracer inhalation, as in parallel has been attempted through ¹⁵O PET.^37,38 By fitting the H₂¹⁷O signal dynamics of the rat muscle ROIs to the adapted three-phase metabolic model, the resting-state metabolic rate of oxygen consumption in skeletal muscle (RMRO₂) was 0.32 ± 0.12 μmol/g/min. Comparing to the literature reports of oxygen metabolic rates in skeletal muscle from the earliest in vitro estimates³⁹ to more recent studies ⁴⁰ in Wistar rats, the results of the present study show a good agreement with the literature values (Table 4). Perfused rat hindquarter muscle metabolic rate was reported similar (e.g., 0.37 μmol O₂/g/min),⁴¹ depending on modality.⁴² Other differences could be inherent to the heterogeneity of muscle fibers,^43–46 which in the case of the temporalis muscle is low^30,47 compared to other muscles (e.g., soleus or gastrocnemius) and in other species.^29,47,48 To the best of our knowledge, this study is the first to report measurements of oxygen metabolic rates using ¹⁷O MR imaging for resting skeletal muscle, although working cardiac muscle with a high oxygen consumption rate has been shown before in isolated heart⁸ as well as in vivo rat heart¹⁰. In muscle, alternative pathways (i.e., fatty acids) are possible in contrast to the glucose-based metabolism of the brain^30,31. However, both are based on of oxygen as the substrate in the predominant mitochondrial electron transfer chain as origin of metabolic H₂¹⁷O. Therefore, this study is in agreement with previous measurements in the cardiac muscle both perfused⁸ and in vivo¹⁰, but it extends to a much lower regime of metabolic rates in the immobilized, resting skeletal muscle with very distinguishable characteristics.²

Table 4.

Muscle oxygen metabolic rate literature values based on various techniques in rats and humans.

Species / Technique	Muscle MRO2 μmol/g/min	Skeletal muscle type	Literature Reference
Rat – perfused	0.65	Global muscle estimate	Field et. al. 1939 39
Rat – perfused	0.37	Hindquarter	Hood et. al. 1986 41
Rat - perfused	0.23	Hindquarter	Rolfe & Brand 1996 40
Rat – perfused	0.75	Spinotrapezius	Behnke et. al. 2002 42
Rat – this study	0.32	Temporalis muscle	-
Human – invasive ΔA/V	0.13	Whole Leg	Oikonen et. al. 1998 60
Human - 15O2 inhalation PET	0.11	Whole Leg	Oikonen et. al. 1998 60
Human - 15O2 inhalation PET	0.10	Whole Leg	Nuutila et. al. 2000 61
Human - 15O2 inhalation PET	0.05	Whole Leg	Heinonen et. al. 2011 37

Comparison between selected muscle metabolism estimates using different techniques in rodents and humans, with the latter being more similar to the ¹⁷O₂ inhalation technique used in this study.

The averaged CMRO₂ value (= 1.97 ± 0.19 μmol/g/min) as determined in this study is in agreement with the value (= 2.19 ± 0.14 μmol/g/min) from a literature report in the rat brain under relatively lower dose α-chloralose anesthesia obtained with a different modeling and experimental protocol.³ These comparisons provide strong evidence to support the validity and reliability of the quantitative ¹⁷O MRS imaging method as described in this work for noninvasively imaging oxygen metabolic rates in the brain and resting muscle with a very low metabolic activity. Thus, we conclude that the same imaging approach should be applicable for most organs across a wide range of metabolic rates.

Global systemic metabolic rates

It should be reasonable to assume that the metabolite pools are in equilibrium upon a stable physiological condition of the animal.⁴⁹ As observed in both simulation and experimental data, the post-inhalation H₂¹⁷O concentrations of different ROIs containing brain or muscle tissue eventually converged to the same steady-state level, which represented the new equilibrium H₂¹⁷O concentration after the ¹⁷O₂ inhalation. Based on that information and Eqn. [2], we were able to derive the global systemic metabolic rate. Metabolic inter- or intra-subject fluctuations are likely caused by variations in the physiological animal condition (i.e., ventilation parameters, anesthesia status and body weight). Thus, in contrast to other studies,⁷ our estimates of the average global oxygen metabolic rate (Table 3) were robust and consistent, independent of whether they were inferred from brain or muscle ROI time courses. Previous studies have used ¹⁷O to assess the total metabolic rate of oxygen so far in dogs⁵⁰ and mealworms^51,52. Very early studies on Wistar rats³⁹ measured oxygen consumption in muscle in vitro, with more recent reports estimating VO₂ for muscle of 18.7 ml O₂/kg/min in anesthetized rats⁵³ and 24.5 ± 8.5 ml O₂/kg/min in awake rats of the same strain, remarkably close to our results (24.2 ml O₂/kg/min).⁵⁴ The variations in literature values also highlight possible inter-subject variations and different approaches used for these studies.^55–57

Increased washout of locally produced H₂¹⁷O during N₂O ventilation

The washout of H₂¹⁷O in brain tissue during the post-inhalation period (i.e., related to perfusion or CBF) has been established previously.¹² It reflects the dynamics of perfusion washout of the metabolically produced H₂¹⁷O in brain tissue and an inflow of global recirculating H₂¹⁷O. However, there is no observable “washout” in the lower metabolic muscle tissue (RMRO₂ = 0.36 μmol/g/min) below the average body oxygen metabolic rate (VO_2,average ~1.1 μmol/g/min), presumably due to a substantial inflow effect from recirculating H₂¹⁷O and low metabolic activity. Thus, in contrast to brain tissue, a significant extent of “wash in” from systemic recirculation after the ¹⁷O₂ inhalation was observed in muscle (Figure 4, from t=15 min onwards).

An increase in cerebral blood flow through vasodilation has been observed and reported before with high percentage N₂O administration.⁵⁸ Thus, the anesthetic properties and vasodilatory effects of N₂O may reduce the global metabolism and possibly uncouple it partially from the narrowly regulated cerebral local oxygen metabolism.⁵⁹

Validation of the three-phase model in future research

Although the influence of recirculating metabolic water is substantial, depending on the regional and global organism rates, the three-phase model accounts accurately for the metabolic rate differences between tissue types. Our measurements used long inhalation times of over 15 min, thus, requiring a non-linear metabolic model.² It can also be concluded that the longer duration of the inhalation phase does not linearly increase the CMRO₂ measurement sensitivity: it is limited by the accumulation of recirculating total body H₂¹⁷O.

An internal ROI validation confirmed whether the voxels selected truly reflected the chosen tissue type by assessment of T₂* against metabolic rate in brain and muscle. Figure 5 shows a plot of the independent properties of tissue T₂* and metabolic rate values for the ROIs taken from muscle and brain under the two different ¹⁷O MRS imaging protocols (low versus high spatial resolution protocol). Two well-separated clusters associated with the two types of tissues because of stark difference in transverse relaxation between the tissues (a much longer T₂* in brain than that of muscle)²² confirm the placement, especially the muscle ROI covered sufficiently accurate the temporalis muscle. It also shows the fact that the strong divergence in metabolic rate is reflecting an underlying tissue difference. However, this approximate separation is only possible because of the significantly shorter T₂* value of H₂¹⁷O in muscle than that of brain tissue.²²

It also has to be noted that certain metabolic rate variability stems from tissue heterogeneity within ROIs. For example, in the case of brain tissue estimates, despite low intra-session variance (e.g., see rat A) a hemispheric difference was likely induced through ROI choice near the boundary between brain and muscle tissues leading to partial volume effects. Another technical limitation is the relatively low SNR of ¹⁷O signal detected in the muscle due to short T₂*,²² and lateral differences in B₁ resulting in ~half SNR than that of brain tissue (see the ¹⁷O spectra in Figures 2A and 2B). Therefore, the fidelity in imaging muscle could be improved, for instance, by using a coil array covering both brain and muscle with optimal detection sensitivity.

Finally, we would anticipate smaller variations of the ¹⁷O MRSI approach when potentially activating the muscle by stimulation, as was done in a different paradigm during varying workload for instance, in cardiac muscle,^8,10 resulting in an elevated oxygen metabolic rate. In previous brain experiments, with an implantable ¹⁷O RF coil, the measurement of an arterial input function and the measurement of blood flow through H₂¹⁷O bolus measurements was used for a detailed investigation, which also allowed the calculation of oxygen extraction fraction (OEF).³ Thus, in future studies in other rat muscles (e.g., in the leg, by implantation of an arterial ¹⁷O RF coil on the femoral artery or separately on the tail artery) the metabolic rate could be validated after electrical stimulation over a wide range of metabolic rates and perfusion. Dynamically measuring the increased metabolic rate during ¹⁷O₂ inhalations, could give new insights to different muscle fiber types. Furthermore, we would expect a simultaneous measurement to be robust in consideration of systemic changes in animal physiology.

Conclusion

In this study, we have extended the applicability of in vivo ¹⁷O MR imaging to measure and image the resting skeletal muscle with a very low oxygen metabolic rate (~16% of the brain tissue). We have also confirmed the consistency of the CMRO₂ results measured during prolonged and repeated inhalations of ¹⁷O₂ gas in this study with previous findings. Since the brain has a very high metabolic rate of oxygen consumption, in contrast to the very low rate in the resting muscle, we anticipate that the same ¹⁷O MR imaging approach and modeling will be useful for other organs such as liver and heart. Therefore, we expect a broad impact of using the ¹⁷O MR imaging technology for metabolic rate measurements in normal and diseased organs beyond the brain.