Development of17O NMR approach for fast imaging
of cerebral metabolic rate of oxygen in rat
brain at high field
Xiao-Hong Zhu, Yi Zhang, Run-Xia Tian, Hao Lei, Nanyin Zhang, Xiaoliang Zhang, Hellmut Merkle, Kamil Ugurbil,
and Wei Chen†
Center for Magnetic Resonance Research, Department of Radiology, University of Minnesota Medical School, 2021 6th Street SE, Minneapolis, MN 55455
Communicated by Robert G. Shulman, Yale University, New Haven, CT, August 6, 2002 (received for review June 11, 2002)
A comprehensive technique was developed for using three-
dimensional17O magnetic resonance spectroscopic imaging at
9.4T for rapidly imaging the cerebral metabolic rate of oxygen
consumption (CMRO2) in the rat brain during a two-min inha-
lation of17O2. The CMRO2value (2.19 ? 0.14 ?mol?g?min, n ?
7) was determined in the rat anesthetized with ?-chloralose by
independent and concurrent17O NMR measurements of cerebral
H217O content, arterial input function, and cerebral perfusion.
CMRO2 values obtained were consistent with the literature
results for similar conditions. Our results reveal that, because of
its superior sensitivity at ultra-high fields, the17O magnetic
resonance spectroscopic imaging approach is capable of detect-
ing small dynamic changes of metabolic H217O during a short
inhalation of17O2gas, and ultimately, for imaging CMRO2in the
small rat brain. This study provides a crucial step toward the goal
of developing a robust and noninvasive17O NMR approach for
imaging CMRO2in animal and human brains that can be used for
studying the central role of oxidative metabolism in brain
function under normal and diseased conditions, as well as for
understanding the mechanisms underlying functional MRI.
through oxygen consumption mediated by the mitochondrial
respiratory chain, coupled to oxidation of glucose as the main
carbon source (2). The oxidative energy need of the human brain
to sustain neuronal activity constitutes a large fraction of the
played by oxygen consumption is also evident in pathologies
associated with the brain; perturbations in brain oxidative me-
tabolism have been closely linked to many brain diseases such as
schizophrenia, Alzheimer’s disease, Huntington’s disease, Par-
kinson’s disease, and mitochondrial dysfunction, as well as aging
problems (3–7). One line of evidence linking these diseases and
cerebral oxidative metabolism is the histopathological finding
that the activity of cytochrome oxidase, the key mitochondrial
enzyme that catalyzes the reduction of oxygen to form water, is
impaired in patients with schizophrenia (3) and Alzheimer’s
disease (4, 5).
Basal cerebral oxygen consumption is not uniform across
different regions in the brain. It is recognized that capillary
density in the brain is inhomogeneous (ref. 8 and refs. therein),
suggesting that mitochondrial density, and hence, oxygen con-
sumption are also likely to be inhomogeneous because a good
correlation exists between capillarity and mitochondrial density
in tissues (9). Alterations in neuronal activity due to stimulation
or performance of a task increases this spatial nonuniformity by
inducing regional changes in oxygen utilization as well as in blood
flow and glucose consumption (10–12). In addition, the impair-
ment in cytochrome oxidase content in the diseased brain was
also found to affect different regions in the brain selectively
(3–5). These considerations imply that the ability to image the
cerebral metabolic rate of oxygen consumption (CMRO2) in vivo
n the brain, the majority of energy consumption occurs by
neuronal activity (1). This energy need is met predominantly
is essential for efforts aimed at investigating cerebral oxidative
metabolism under normal and pathological conditions.
The earliest approach of CMRO2measurement is based on the
Kety-Schmidt method for measuring cerebral blood flow (CBF)
together with arteriovenous differences of oxygen content (13).
This method, thought to be the most accurate one in the literature,
provides only global CMRO2information without spatial differen-
tiation within the brain. Positron emission tomography (PET) has
been widely used for imaging CMRO2 in humans (14–16). By
introducing oxygen gas enriched with the isotope15O into the
human body and monitoring the spatial distribution and accumu-
CMRO2and provide a CMRO2image. However, the15O-PET is
unable to distinguish the radioactive signals between the15O2
H215O. To overcome this drawback, a measurement of cerebral
blood volume using inhalation of C15O has to be performed in
addition to the CBF measurement based on i.v. injection of H215O.
These PET requirements make CMRO2measurements difficult to
execute experimentally and complicate the calculation of CMRO2
The13C NMR methods based on monitoring the cerebral
metabolism of intravenously infused [1-13C] glucose (e.g., (10,
11, 17, 18)) provide an alternative approach for CMRO2mea-
surements. The labeling kinetics for several intracellular metab-
olites, most importantly glutamate, can be unequivocally mea-
sured by these methods; extraction of CMRO2 from
experimental data relies on extensive modeling (10, 11, 18, 19).
In addition, the13C NMR methods require a long measurement
time (60–120 min) because of relatively slow turnover rates of
cerebral metabolites, and it is difficult, although not impossible,
to achieve three-dimensional (3D) CMRO2imaging because of
a relatively long repetition time required for signal acquisitions.
The possibility of using the17O NMR spectroscopy?imaging
techniques for monitoring labeled H217O as a tracer of oxygen
(20–30). These early studies can be divided into two distinct
groups. The first group involved the simple17O NMR approach
for detecting H217O directly (20, 21, 24–27). Nonlocalized17O
NMR was used for most studies to estimate CMRO2in the whole
brain of experimental animals as well as in the human occipital
lobe (31) after inhalation of oxygen gas enriched with the17O
image (?800 ?l voxel size) in the cat brain with a long mea-
surement time (25, 27). These previous efforts, all of which were
conducted at relatively low fields compared with what is avail-
able currently, demonstrated the dramatic limitations imposed
by the low inherent sensitivity of
17O NMR due to its low
Abbreviations: CMRO2, cerebral metabolic rate of oxygen; CBF, cerebral blood flow; PET,
positron emission tomography; 3D, three-dimensional; MRS, magnetic resonance spectro-
scopic; SNR, signal-to-noise ratio.
†To whom reprint requests should be addressed. E-mail: email@example.com.
October 1, 2002 ?
vol. 99 ?
gyromagnetic ratio. These sensitivity limitations have hindered
further developments of the17O NMR approach and lead to the
exploration of a second group of methods that attempt to
circumvent the sensitivity limitations of
indirect detection through17O-coupled protons.
The indirect17O detection approach relies on the effect of
water protons, which is considerably shortened by the presence of
the17O nucleus in the same water molecule (22). It has also been
shown that the water proton T2of H217O can be restored to that of
H216O when17O decoupling is applied; in this case, the difference
between proton signals obtained with and without17O decoupling
correlating the proton signal changes to the concentrations of
H217O because the water T2 is sensitive to many physiologic
parameters such as pH and temperature. In addition, the fractional
change in1H signal induced by variations in metabolic H217O
concentration is small at basal conditions. These indirect
detection approaches have so far not been successful in imaging
Recently, we have investigated the longitudinal and transverse
relaxation times [T1, T2, and apparent T2(T2*)] as well as the
signal-to-noise ratio (SNR) of17O spins in water by using the17O
magnetic resonance spectroscopic (MRS) imaging method at
4.7T and 9.4T field strengths (32). Our results showed that the
relaxation times of17O water are field-independent, and the17O
NMR sensitivity increased approximately fourfold at 9.4T com-
pared with a field strength of 4.7T. With this sensitivity gain, we
were able to obtain two-dimensional17O MRS images of natural
abundance H217O in the rat brain at 9.4T with a temporal
resolution of 12 s (32). These results demonstrated the potential
of17O MRS imaging at ultra-high field for mapping the spatial
distribution of metabolic H217O dynamically in the rat brain.
In this study, we investigated the feasibility for developing the
during a brief inhalation of
applied the 3D17O MRS imaging method to determine CBF
after a bolus injection of H217O through an internal carotid
artery and the accumulation rate of H217O generated by oxida-
tive metabolism in the rat brain during a two-min17O2inhalation
with superior spatial (27 ?l) and temporal (11 s per 3D image)
resolution. Furthermore, we performed experiments where such
images were obtained together with continuous experimental
monitoring of the arterial input function related to recirculated
H217O via an implanted carotid artery MR coil. These measure-
ments were used to calculate CMRO2values rigorously in image
voxels by using a complete modeling that accounts for all
contributions to the cerebral H217O content accumulated during
an inhalation of17O2and demonstrate the feasibility for imaging
CMRO2. Based on these results, we present high-resolution 3D
CMRO2images obtained with17O NMR under the basal con-
dition in the rat brain anesthetized with ?-chloralose.
17O NMR by using
17O2 gas. We have successfully
The dynamic change of cerebral H217O concentration accumu-
lated during an inhalation of
processes: oxygen consumption generating metabolic H217O in
the brain, perfusion resulting in H217O washout from the brain,
and flow recirculation bringing extra H217O into the brain. The
mass balance equation of labeled H217O concentrations during
an inhalation of17O2is given by
17O2 is determined by three
? 2?f1CMRO2? CBF?f2?Ca?t? ? Cv?t???,
where Ca(t), Cb(t), and Cv(t) are the time-dependent H217O
concentrations expressed as [H217O] in excess of the natural
abundance [H217O] level in the arterial blood, brain tissue, and
venous blood, respectively. The constant ? is the enrichment
fraction of17O-labeled O2 gas. Proper use of Eq. 1 requires
consistencies of units among all terms used in the equation. The
determinations of Cb, Ca, and Cvare based on the17O NMR
measurements and calibrated by using the natural abundance
H217O concentration (20.35 ?mol?(g of brain water) for brain
tissue, and ?mol?(g of blood water) for blood, calculated from
natural abundance H217O enrichment of 0.037% and the mo-
lecular weight of H217O ? 19.0. Therefore, we prefer to use
convenient units of ?mol?(g of brain water) for Cb(t) and
?mol?(g of blood water) for Ca(t) and Cv(t). CMRO2 is ex-
pressed in the conventional unit of ?mol?min?(g brain tissue)
and is converted from the unit of ?mol?min?(g brain water) by
using the conversion constant f1 given by f1 ? (g of brain
tissue)?(g of brain water) ? 1??brain? 1.266, where ?brainis 0.79
(14, 27). The constant f2is also a unit conversion factor and is
given by f2? ?blood??blood??brain? 1.05 ? 0.81?0.79 ? 1.077,
where ?blood? (g blood)?(ml blood) ? 1.05 (33), and ?blood?
(g blood water)?(g blood) ? 0.81 (14). If water in brain tissue is
in equilibrium with water in venous blood (i.e., fast water
exchange across capillaries), f2Cv(t) ? Cb(t)?? where ? ? 0.90 is
the brain?blood partition coefficient with the unit of (ml
blood)?(g of brain tissue) (34). Substituting this relation and
introducing two new correction parameters (n and m) into Eq.
1 leads to
? 2?f1CMRO2? mCBF?f2Ca?t? ?nCb?t?
The solution of Eq. 2 is
mnCBFCMRO2?1 ? e
?t ? t??dt?,
where m is a correction factor accounting for the limited
the brain-blood barrier; the value of m is 0.84 for the CBF range
that we studied herein (35). The constant n is another correction
factor that accounts for the additional restriction on the perme-
ability of the H217O generated through oxidative metabolism in
the mitochondria (‘‘metabolic’’ H217O). This additional restric-
tion is included in this modeling because we have observed that
the washout rate of the metabolic H217O after the cessation of
17O2inhalation was significantly slower than the washout rate of
the H217O that permeates brain tissue subsequent to a bolus
injection of H217O through the internal carotid artery in the rat.‡
The cause of this difference could be permeability restrictions
imposed by the inner and outer mitochondrial membranes that
must be traversed by all water molecules generated by oxygen
consumption. The ratio of the washout rate of the metabolic
H217O after the cessation of17O2 inhalation vs. the cerebral
H217O after a bolus injection of H217O through the internal
carotid artery gives the value of n. Limited permeability for the
metabolic water has been reported in the human brain (36).
The CMRO2value can be precisely calculated by using Eq. 3
if all parameters of Cb(t), CBF, and Ca(t) can be experimentally
measured for the same animal brain. The Cb(t) and CBF values
can be determined by using17O NMR approaches. To measure
‡Zhu, X. H., Lei, H., Zhang, Y., Zhang, X. L., Zhang, N. Y., Ugurbil, K. & Chen, W. (2002) Proc.
Int. Soc. Mag. Res. Med. 10, 1094 (abstr.).
Zhu et al.PNAS ?
October 1, 2002 ?
vol. 99 ?
no. 20 ?
Ca(t), we designed an implanted17O rf coil that permits the
continuous measurement of Ca(t) in a rat carotid artery. This
measurement can be simultaneously performed with Cb(t) mea-
surement during an inhalation of
channels on the MR spectrometer equipped with multiple
receivers. In principle, the Ca(t) function is determined by the
metabolic H217O generated in the entire animal body and the
equilibration of this H217O with the blood compartment. Con-
sidering that the oxygen consumption rate is constant (i.e., time
independent) in all tissues during the short inhalation time, one
can approximate Ca(t) as a linear function of time [i.e., Ca(t) ?
At, where A is a constant]. This approximation was supported by
our experimental results (see Fig. 5 presented later) and the
literature. Then, the solution of Eq. 3 for CMRO2is
17O2 by using separate rf
1 ? e
Therefore, according to Eq. 4, the CMRO2value can be calcu-
lated by experimental data on CBF, A, n, Cb(t), and other known
constants (f1, f2, m, ?, and ?) for each data point measured at
different inhalation times for each voxel. All CMRO2 values
reported (except for those in Fig. 5) are averages of the values
calculated as a function of time, excluding the first two time
points because of their relatively large measurement errors and
the fact that they may represent transient values during the
approach to a steady-state rate of H217O production by oxidative
phosphorylation. Such transient values are expected during the
build-up of17O2in the brain at values that exceed the Kmof
cytochrome oxidase for oxygen. Ability to obtain measurements
as a function of time and the use of only the time independent
values avoids this complication.
CBF values can be determined from the ‘‘washout’’ rate of the
tracer H217O in the brain tissue following a rapid bolus injection
of H217O through one carotid artery (32) according to Eq. 5:
Cb?t? ? Cb?0?exp? ? t?mCBF????,
where Cb(0) is the cerebral H217O concentration after the arrival
of all labeled water after the bolus injection. Fitting the expo-
nential decay of the H217O washout curve according to Eq. 5
gives the CBF value.
Materials and Methods
NMR Methods. All NMR experiments were performed on a
9.4T?31-cm bore magnet (Magnex Scientific, Abingdon, U.K.)
interfaced to a Unity INOVA console (Varian). A multinuclear
coil (? 1 cm ? 2 cm; 54.25 MHz) and a large butterfly-shaped
1H coil (400 MHz) was used.
A TURBOFLASH (fast low-angle shot) image sequence was used
to acquire scout images [TR?TE ? 8 ms?4 ms, field of view
(FOV) ? 3 cm ? 3 cm, and image matrix size ? 128 ? 128]. The
spatial localization of17O MRS was achieved by using the 3D
Fourier Series Window MRS imaging technique (32, 37). In this
method, the k-space sampling is weighted according to the
Fourier coefficients of a predetermined voxel shape to achieve
an optimal filter applied at the acquisition stage. A cylindrical
voxel shape (circular shape on the coronal orientation) was used
in this study (37). Short T1value (4.5 ms) of cerebral H217O in
the rat allows rapid signal acquisition, and ultimately, more
signal averages within the same sampling time (32). Therefore,
a short TR of 12 ms (?3T1) was used for gaining SNR. The
acquisition time for each 3D17O-MRS image was 11 s (total scan
number ? 948; rf pulse width ? 50 ?s, spectral width ? 30 kHz;
FOV ? 18 ? 18 ? 15 mm3, 9 ? 9 ? 5 phase encodes).
Phase-encoding gradient was 0.35 ms long and had a half-sine
waveform, which allowed for a short gradient echo time (TE ?
0.4 ms). This short TE is critical for minimizing signal loss
because of the relatively fast transverse relaxation of the17O
magnetization (T2? 3.0 ms at 9.4T; ref. 32). The peak strength
of the phase-encoding gradient increment was 0.09–0.11 G?cm.
The image voxel size was 27 ?l. A 17 ? 17 ? 9 matrix of free
5 phase encode data for each 3D17O image. The FIDs were
zero-filled, and a 100-Hz line broadening (LB) was used before
fast Fourier transformation for SNR enhancement. The17O
NMR signals from each voxel were quantified by measuring the
H217O resonance peak intensities.
General Animal Preparation. Male Sprague–Dawley rats (250–300
g body weight) were anesthetized with ?2% (vol?vol) isoflurane
in a mixture of O2and N2O gases (2:3) during surgery. After oral
intubations, femoral artery and vein were catheterized for
physiological monitoring, blood sampling, and chemical admin-
istration. After surgery, anesthesia was switched to ?-chloralose
by using continuous infusion of 0.4 ml?hr of 15 mg?ml ?-chlo-
ralose solution. Blood gases were sampled for monitoring phys-
iological conditions [pH ? 7.35 ? 0.03, pCO2 ? 42.1 ? 3.0
mmHg, pO2 ? 124 ? 16 mmHg (1 mmHg ? 133 Pa)]. The
arterial blood pressure (80–110 mmHg) and EtCO2level (?3%)
were monitored throughout the experiment. The rectal temper-
ature of the rats was maintained at 37 ? 1°C by using a heated
water blanket. All animal surgical procedures and experimental
protocols were conducted under the guidelines of the National
Institutes of Health and the Institutional Animal Care and Use
Committee of the University of Minnesota.
CBF Measurements. For CBF measurements, one external carotid
artery was catheterized with a PE-50 tubing for gaining access to
the internal carotid artery without interrupting blood circulation
into the brain (32). A 0.05–0.1 ml 50%-enriched H217O was
rapidly injected into the brain through the internal carotid artery
for CBF measurements using17O MRS imaging (32).
Ca(t) Measurements. An implanted vascular rf coil was designed
and constructed for continuously detecting the17O NMR signal
changes of H217O in the rat carotid artery at 9.4T. This coil was
based on a modified solenoid coil design combined with an rf
shielding.§The rf shielding ensured that the NMR signal de-
tected by the implanted coil was only attributed by the artery
blood (?7 ?l) without contaminations from surrounding tissues.
Therefore, additional spatial localization was not necessary for
determining Ca(t). In addition, the rf shielding minimized the
electromagnetic coupling between the implanted17O coil and
the head17O surface coil tuned at the same operating frequency
(two-coil arrangement), which allowed simultaneous measure-
the same temporal resolution (11 s).
17O2Inhalation Experiments. For17O2inhalation studies, the17O-
labeled O2gas with 58.2–72.1% enrichment (Isotec) was mixed
with N2O gas (?2:3) and then stored in a cylindrical gas
reservoir. The 3D17O MRS imaging acquisitions were started
§Zhang, X. L., Tian, R. X., Zhu, X. H, Zhang, Y., Merkle, H. & Chen, W. (2002) Proc. Int. Soc.
Mag. Res. Med. 10, 889 (abstr.).
www.pnas.org?cgi?doi?10.1073?pnas.202471399 Zhu et al.
with the normal gas mixture first for 2 min; these natural
abundance 3D17O MRS images were used as references for
quantifications of metabolic H217O concentrations in the rat
brain during and after the17O2 inhalation. Subsequently, the
respiration gas was quickly switched to the17O2-labeled gas
mixture. After 2 min of17O2inhalation, the gas line was switched
back to the normal gas mixture, and the 3D17O MRS imaging
acquisitions were continued throughout the procedure; a total of
100 images were collected in about 19 min.
Determination of Constant n. The ratio between the washout rate
of the metabolic H217O after the cessation of17O2inhalation and
the washout rate of the H217O trace from the H217O bolus
injected through the rat internal carotid artery gave the value of
constant n for each voxel.
All results are presented as mean ? SD.
The multinuclear surface-coil probe was used to collect a series
of global17O and1H spectra from the natural abundance water
in the rat brain to estimate the relative signal fluctuations
between17O and1H NMR. Two stacked plots of 2517O and1H
spectra acquired consecutively by using the17O coil and the
much larger1H coil, respectively, are shown in Fig. 1. Although
the number of acquisitions (NT) for each spectrum was different
for the17O and1H nuclei, the repetition time (TR) was ?3T1in
both cases; the total acquisition time was 6.2 s for all spectra.
Under this condition, we found that the signal fluctuations in the
consecutively acquired spectra for the two nuclei were compa-
rable (SD ? 0.20% for1H spectra; SD ? 0.35% for17O spectra).
This result demonstrates that with respect to the capability for
detecting fractional changes in NMR signal intensity, the17O
method is comparable to the1H method.
The 3D17O MRS imaging technique was applied to measure
the natural abundance H217O distribution in the rat brain with
high temporal resolution of 11 s and spatial resolution (27 ?l
voxel size). Fig. 2A shows representative data obtained from
three adjacent coronal images in a rat brain showing anatomical
images acquired by the1H coil and the17O MRS images of
natural abundance H217O acquired by the17O coil. Fig. 2B
displays one17O chemical shift image from the middle image
slice as shown in Fig. 2A. The spatial distributions of17O NMR
signal intensity are not uniform because of the inhomogeneous
B1 of the17O surface coil. Nevertheless, excellent SNR was
achieved with a relatively short acquisition time at 9.4T, espe-
cially for the central voxels where SNR was optimized by the B1
profile of the17O coil (SNR ? 40:1).
The large sensitivity gained at ultra-high field revealed the
potential of17O MRS imaging for detecting small dynamic changes
of cerebral H217O in small animal brains during a short inhalation
H217O from one representative voxel before (natural abundance),
during, and after a 2-min inhalation of17O2(72.1%17O enrich-
ment). It is clearly evident that the17O signal intensity of cerebral
H217O in the entire measurement is characterized by three distinct
phases: (i) constant before the17O2inhalation; (ii) approximately
linearly increasing during the17O2inhalation; and (iii) decreasing
after the cessation of17O2inhalation, approximately exponentially,
and reaching a new steady state within a short time (?10 min).
state after the17O2inhalation experiment, the CBF measure-
ment was performed by a rapid bolus injection of H217O into one
internal carotid artery. The dynamic distribution of cerebral
H217O signal during the washout period was monitored by using
3D17O MRS imaging. The washout curves excluding the first
to Eq. 5, and these fits yielded CBF values. Fig. 4 demonstrates
one example of CBF measurement obtained from the same rat
and image voxel used in Fig. 3 showing the stacked plot of17O
spectra and the exponential fit. Because the H217O bolus was
injected into one of the internal carotid arteries, and the H217O
tracer was mainly distributed into the ipsilateral hemisphere,
CBF and, subsequently, CMRO2 quantifications were per-
formed only for the17O image voxels in the ipsilateral hemi-
sphere of the rat brain.
Fig. 5A illustrates the natural abundance H217O spectrum
from the blood in the rat carotid artery. Fig. 5B plots the results
of simultaneous measurements of Ca(t) change in one carotid
artery and Cb(t) change in a voxel in the ipsilateral hemisphere
adjacent slices (Left, color images) and corresponding anatomical images
(Right, gray images) in the coronal orientation from a representative rat.
(A) 3D17O brain images of natural abundance H217O from three
voxel (27 ?l voxel size) as indicated in the anatomical image (Left Inset)
acquired before (natural abundance), during (as indicated by the dark bar
under the stacked plots) and after a 2-min17O2inhalation.
Stacked plots of the cerebral H217O spectra from one representative
the rat brain acquired and processed with the following parameters: NT ? 1,
TR ? 6.2s, LB ? 25Hz (? half linewidth). The SD of the1H water peak height
in these spectra is 0.20%. (B) Stacked plot of 2517O NMR spectra of natural
abundance water in the rat brain acquired and processed with the following
the SD of the17O water peak height in these spectra is 0.35%. The total
acquisition times for each1H and17O spectra were the same (6.2 s).
(A) Stacked plot of 251H NMR spectra of natural abundance water in
Zhu et al.PNAS ?
October 1, 2002 ?
vol. 99 ?
no. 20 ?
brain during an inhalation of17O2. The Ca(t) data were fitted by
a linear function for calculating the value of constant A. Finally,
the values of Cb(t), CBF, and n from each voxel and the value of
A measured from each17O inhalation measurement were used
to calculate CMRO2according to Eq. 4. Fig. 5C demonstrates
one example of CMRO2calculations as a function of time from
a single voxel. It is evident in Fig. 5C that the CMRO2value is
independent of the inhalation time if the initial two points are
excluded. The averaged CMRO2 values (excluding two initial
points) and CBF values from seven measurements are 2.19 ?
0.14 ?mol?g?min and 0.53 ? 0.07 ml?g?min, respectively, in the
rat brains anesthetized with ?-chloralose.
The mean CBF and n values obtained from the entire hemi-
sphere ipsilateral to the internal carotid artery where a H217O
bolus was injected were subsequently used to calculate CMRO2
values for both ipsilateral and contralateral hemispheres approx-
imately and to obtain 3D CMRO2images. This approximation
was made based on the fact that the CBF distribution in the
hemisphere where it was measured was relatively uniform and
that two hemispheres should be equivalent with respect to CBF.
Fig. 6 illustrates three adjacent CMRO2images in the coronal
orientation from a representative rat brain. The inhomogeneous
signal intensity caused by the surface coil sensitivity profile (Fig.
2) is not reflected in the CMRO2maps, as expected.
The17O T1value in H217O is less than 5 ms in the rat brain at
9.4T (32). This short T1permits much faster data acquisitions,
more signal averages, and, consequently, a high SNR per unit of
time. This advantage is critical for the success of CMRO2
measurement using direct17O detection. The comparable frac-
tional signal fluctuations between consecutively acquired1H and
17O NMR spectra (Fig. 1) indicated that direct17O detection of
changes in metabolic H217O signal is likely to be more sensitive
than the indirect17O detection through1H NMR. This notion is
based on the fact that the fractional change of water proton
signal caused by a 2-min17O2inhalation at basal condition is
estimated to be 0.15–0.30% (29), which is difficult to detect
reliably because this change is comparable to the signal fluctu-
ations in consecutively acquired1H spectra or images. These
signal fluctuations are not dominated by the inherent SNR of a
single spectrum or image but are mainly determined by the
physiological processes of breathing, heart pulsation, and vaso-
motion, as well as NMR scanner instability. In contrast,17O
NMR has a much larger dynamic range for detecting the changes
of metabolic H217O (20–40%) during the same17O2inhalation.
This difference is the major merit that makes17O NMR more
suitable for CMRO2imaging.
The17O approach has several major advantages for CMRO2
imaging in comparison with other existing methods. First,17O
signals from the17O2bound to hemoglobin or17O2dissolved in
tissue space. When bound to hemoglobin, the17O2resonance is
broadened beyond detection because of the slow rotational motion
for the large molecular weight oxyhemoglobin complex.17O2as an
unbound molecule in gas phase or dissolved in water is strongly
paramagnetic because of its unpaired electrons and, hence, unde-
tectable because of the dipolar coupling between the electrons and
the nucleus. As a result,17O NMR avoids the complication in the
calculation of CMRO2encountered in the PET methodology. In
addition, unlike PET, the17O isotope is stable and nonradioactive,
and the experimental procedure of17O2inhalation does not intro-
These merits can significantly simplify the modeling of CMRO2
calculation as well as provide an approach for imaging CMRO2in
living organs with minimal biological risk. Second, the natural
abundance signal of H217O, which can be accurately imaged,
provides an internal reference for rigorously calibrating and calcu-
lating the absolute H217O concentration changes for both Cb(t) and
Ca(t) during17O2inhalation; such a quantification is crucial for
determining absolute CMRO2values and is independent of the
inhomogeneities of17O coil sensitivity. Third, there is only a single
analysis simple. Fourth, the17O resonance peak is relatively broad
and has a low resonance frequency (only 14% of the1H resonance
frequency); hence, it is significantly less sensitive to static magnetic
field inhomogeneities. Fifth, after the cessation of17O2inhalation,
the cerebral H217O concentration reaches a new steady state within
a short time (6–10 min); this fast recovery should allow repeated
session. This capability is essential for studying the oxidative
metabolism changes related to perturbations in physiology and
function, where at least two measurements are required under
control and perturbation conditions. Finally, at ultra-high fields the
images, even relative to the small size of the rat brain, with an 11-s
acquisition time; this sensitivity makes the17O approach possible
for imaging CMRO2within a short measurement duration (2 min).
Because of the technical challenges faced for most existing
CMRO2imaging methods, there is virtually no literature report-
ing direct measurements of CMRO2 in the rat brain. For
instance, the image voxel size of most conventional PET scan-
ners is on the order of several milliliters (e.g., ref. 38); this size
is already larger than the size of the entire rat brain. Neverthe-
less, we compared our result of CMRO2measurement by using
in Fig. 3) after a fast bolus injection of H217O into the rat brain through an
internal carotid artery. (B) Exponential decay fitting of the H217O washout
curve for calculating CBF (0.56 ml?g?min for this voxel).
(A) Stacked plots of the cerebral H217O spectra from one voxel (used
blood (7 ?l) obtained before inhalation of17O2. (B) Time course of17O MR
signals of Ca(t) in one carotid artery (F) and Cb(t) from a representative voxel
3D coronal CMRO2images of rat brain measured during a 2-min17O2
www.pnas.org?cgi?doi?10.1073?pnas.202471399Zhu et al.
the17O NMR approach with the literature results obtained by
indirect methods under similar condition. Based on an autora-
diographic approach (39), cerebral metabolic rate of glucose
(CMRglc) was recently determined to be ?0.37 ?mol?g?min in
the somatosensory and motor cortices of rats anesthetized with
?-chloralose. By using the CMRO2?CMRglcratio of 5.5 (2), a
CMRO2 value of 2.04 ?mol?g?min is calculated from this
CMRglcmeasurement, which is in excellent agreement with our
17O-based CMRO2result (2.19 ? 0.14 ?mol?g?min). The CBF
value of 0.58–0.68 ml?g?min reported in the same autoradio-
graphic study (39) is also close to the mean CBF value of 0.53 ?
0.07 ml?g?min measured in our study. The heteronuclear editing
MRS and1H MRI methods covering a large range of cortical
activity at different anesthesia conditions had been used to
measure CMRO2and CBF values in the sensory motor cortex of
mature rats, and a linear correlation between CMRO2and CBF
was observed (CMRO2? 3.76?CBF ? 0.18, R2? 0.99; ref. 40).
The estimated CMRO2from this relation is 2.16 ?mol?g?min if
the CBF value observed in our study (0.53 ml?g?min) is applied.
This CMRO2 value is again in excellent agreement with our
result. These comparisons provide mutual support between our
direct CMRO2measurements and other indirect measurements
reported in the literature and support the validity of the 3D
CMRO2imaging using17O NMR.
invasive procedures: one for determining Ca(t) by means of the
implanted17O rf coil, and the second for measuring CBF by an
intra-arterial catheter and bolus injection of H217O. The CBF
measurement can be performed noninvasively by using the arterial
spin tagging MR approaches (41). The implanted rf coil, on the
other hand, is not practical for routine measurements of CMRO2
is it suitable for human applications. Therefore, it would be
important to explore further the feasibility of the
Ca(t) measurements. A successful outcome from such a study will
establish a completely noninvasive
In comparison with most existing approaches, the high-field17O
NMR method requires a significantly short measurement time for
imaging CMRO2 and can be performed within a 2-min
inhalation period. However, this method assumes that CMRO2is
the CMRO2 measurement time can further shorten at 9.4T or
method, this time-scale is faster than any other technique currently
available for assessing CMRO2, especially at the spatial resolution
available from the17O NMR approach.
17O approach for CMRO2
We have successfully developed a comprehensive technique by
using17O MRS imaging at ultra-high field for fast imaging of
CMRO2 in the rat brain and validated it with complementary
measurements. The overall results from our study provide an
essential step toward the goal for developing a robust, fast, reliable,
oxidative metabolism for both animal and human brains and,
potentially, for other living organs. Such an approach would be
useful for studying the central role of oxidative metabolism in brain
function under normal and diseased conditions, as well as for
understanding the mechanisms underlying functional MRI (42).
17O NMR approach for imaging the rate of
We thank Dr. Jae-Hwan Kwag and John Strupp for their technical
assistance and Dr. Haiying Liu for scientific discussions. This work was
supported by National Institutes of Health Grants NS41262, NS38070,
NS39043, and EB00329; National Research Resource Grant P41
RR08079 (from the National Institutes of Health); the MIND Institute;
and the W. M. Keck Foundation.
1. Attwell, D. & Laughlin, S. B. (2001) J. Cereb. Blood Flow Metab. 21, 1133–1145.
2. Siesjo, B. K. (1978) Brain Energy Metabolism (Wiley, New York).
3. Maurer, I., Zierz, S. & Moller, H. (2001) Schizophr. Res. 48, 125–136.
Z., Rachel, R. & Jones, J. (1997) Vision Res. 37, 3593–3608.
5. Maurer, I., Zierz, S. & Moller, H. J. (2000) Neurobiol. Aging 21, 455–462.
6. Frackowiak, R. S., Herold, S., Petty, R. K. & Morgan-Hughes, J. A. (1988) Brain
7. Beal, M. F. (1992) Ann. Neurol. 31, 119–130.
8. Harrison, R. V., Harel, N., Panesar, J. & Mount, R. J. (2002) Cereb. Cortex 12,
9. Hoppeler, H. & Kayar, S. R. (1988) News Physiol. Sci. 3, 113–116.
10. Hyder, F., Chase, J. R., Behar, K. L., Mason, G. F., Siddeek, M., Rothman, D. L.
& Shulman, R. G. (1996) Proc. Natl. Acad. Sci. USA 93, 7612–7617.
11. Chen, W., Zhu, X. H., Gruetter, R., Seaquist, E. R. & Ugurbil, K. (2001) Magn.
Reson. Med. 45, 349–355.
12. Raichle, M. E. (1987) in Handbook of Physiology–The Nervous System, eds.
Mountcastle, V. B., Plum, F. & Geiger, S. R. (Am. Physiol. Soc., Washington, DC),
13. Kety, S. S. & Schmidt, C. F. (1948) J. Clin. Invest. 27, 476–483.
14. Ter-Pogossian, M. M., Eichling, J. O., Davis, D. O. & Welch, M. J. (1970) J. Clin.
Invest. 49, 381–391.
15. Mintun, M. A., Raichle, M. E., Martin, W. R. & Herscovitch, P. (1984) J. Nucl.
Med. 25, 177–187.
16. Lenzi, G. L., Frackowiak, R. S., Jones, T., Heather, J. D., Lammertsma, A. A.,
Rhodes, C. G. & Pozzilli, C. (1981) Eur. Neurol. 20, 285–290.
17. Rothman, D. L., Novotny, E. J., Shulman, G. I., Howseman, A. M., Petroff,
O. A., Mason, G., Nixon, T., Hanstock, C. C., Prichard, J. W. & Shulman, R. G.
(1992) Proc. Natl. Acad. Sci. USA 89, 9603–9606.
18. Gruetter, R., Seaquist, E. R. & Ugurbil, K. (2001) Am. J. Physiol. 281,
19. Mason, G. F., Rothman, D. L., Behar, K. L. & Shulman, R. G. (1992) J. Cereb.
Blood Flow Metab. 12, 434–447.
20. Mateescu, G. D., Yvars, G. M. & Dular, T. (1988) in Water and Ions in Biological
Systems, eds. Lauger, P., Packer, L. & Vasilescu, V. (Birkhauser, Basel), pp.
21. Mateescu, G. D., Yvars, G., Pazara, D. I., Alldridge, N. A., LaManna, J. C., Lust,
D. W., Mattingly, M. & Kuhn, W. (1989) in Synthesis and Application of
Isotopically Labeled Compounds, eds. Baillie, T. A. & Jones, J. R. (Elsevier,
Amsterdam), pp. 499–508.
22. Hopkins, A. L. & Barr, R. G. (1987) Magn. Reson. Med. 4, 399–403.
23. Kwong, K. K., Hopkins, A. L., Belliveau, J. W., Chesler, D. A., Porkka, L. M.,
B. R. (1991) Magn. Reson. Med. 22, 154–158.
24. Arai, T., Nakao, S., Mori, K., Ishimori, K., Morishima, I., Miyazawa, T. &
Fritz-Zieroth, B. (1990) Biochem. Biophys. Res. Commun. 169, 153–158.
25. Pekar, J., Ligeti, L., Ruttner, Z., Lyon, R. C., Sinnwell, T. M., van Gelderen, P.,
Fiat, D., Moonen, C. T. & McLaughlin, A. C. (1991) Magn. Reson. Med. 21,
26. Fiat, D., Ligeti, L., Lyon, R. C., Ruttner, Z., Pekar, J., Moonen, C. T. &
McLaughlin, A. C. (1992) Magn. Reson. Med. 24, 370–374.
27. Pekar, J., Sinnwell, T., Ligeti, L., Chesnick, A. S., Frank, J. A. & McLaughlin,
A. C. (1995) J. Cereb. Blood Flow Metab. 15, 312–320.
28. Ronen, I. & Navon, G. (1994) Magn. Reson. Med. 32, 789–793.
30. Reddy, R., Stolpen, A. H. & Leigh, J. S. (1995) J. Magn. Reson. 108, 276–279.
31. Fiat, D., Dolinsek, J., Hankiewicz, J., Dujovny, M. & Ausman, J. (1993) Neurol.
Res. 15, 237–248.
Med. 45, 543–549.
33. Altman, P. L. (1961) in Biological Handbook Series (Fed. Am. Soc. Exper. Biol.,
Washington, DC), p. 14.
34. Herscovitch, A. & Raichle, M. E. (1985) J. Cereb. Blood Flow Metab. 5, 65–69.
35. Herscovitch, P., Raichle, M. E., Kilbourn, M. R. & Welch, M. J. (1987) J. Cereb.
Blood Flow Metab. 7, 527–542.
36. Ohta, S., Meyer, E., Thompson, C. J. & Gjedde, A. (1992) J. Cereb. Blood Flow
Metab. 12, 179–192.
37. Hendrich, K., Hu, X., Menon, R., Merkle, H., Camarata, P., Heros, R. &
Ugurbil, K. (1994) J. Magn. Reson. 105, 225–232.
38. Fox, P. T., Raichle, M. E., Mintun, M. A. & Dence, C. (1988) Science 241,
39. Nakao, Y., Itoh, Y., Kuang, T. Y., Cook, M., Jehle, J. & Sokoloff, L. (2001) Proc.
Natl. Acad. Sci. USA 98, 7593–7598.
40. Hyder, F., Kennan, R. P., Kida, I., Mason, G. F., Behar, K. L. & Rothman, D.
(2000) J. Cereb. Blood Flow Metab. 20, 485–498.
41. Detre, J. A., Zhang, W., Roberts, D. A., Silva, A. C., Williams, D. S., Grandis,
D. J., Koretsky, A. P. & Leigh, J. S. (1994) NMR Biomed. 7, 75–82.
42. Barinaga, M. (1997) Science 276, 196–198.
Zhu et al.PNAS ?
October 1, 2002 ?
vol. 99 ?
no. 20 ?
NEUROSCIENCE. For the article ‘‘Development of17O NMR ap-
proach for fast imaging of cerebral metabolic rate of oxygen in
rat brain at high field,’’ by Xiao-Hong Zhu, Yi Zhang, Run-Xia
Tian, Hao Lei, Nanyin Zhang, Xiaoliang Zhang, Hellmut
Merkle, Kamil Ugurbil, and Wei Chen, which appeared in
number 20, October 1, 2002, of Proc. Natl. Acad. Sci. USA (99,
13194–13199; First Published September 19, 2002; 10.1073?
pnas.202471399), the authors note the following correction.
error in the setup of parameters used for acquiring the 3D17O
magnetic resonance spectroscopic (MRS) images reported in
this article. This error led to overestimations of the spatial
resolution of 3D17O MRS images. The claimed voxel sizes of the
3D17O MRS images (Figs. 2 and 3) and the cerebral metabolic
rate of oxygen (CMRO2) image (Fig. 6) are 57% smaller than the
actual voxel size in each spatial dimension. Therefore, the
correct voxel size was 0.10 ml, and the correct field-of-view
(FOV) used in the 3D17O MRS images was 28 ? 28 ? 24 mm3.
This correction should not significantly affect the major conclu-
sions and methodology presented in this article. However, the
correction could reveal that the current sensitivity of17O NMR
and, alternatively, the spatial resolution of the17O MRS image
achieved at 9.4 tesla may be potentially limited for determining
and imaging CMRO2in small brain structures such as the white
matter in the rat brain.
COLLOQUIUM. For the colloquium paper ‘‘Unified scaling law for
earthquakes,’’ by Kim Christensen, Leon Danon, Tim Scanlon,
and Per Bak, which appeared in Suppl. 1, February 19, 2002, of
Proc. Natl. Acad. Sci. USA (99, 2509–2513), and for the Physical
Review Letters paper ‘‘Unified Scaling Law for Earthquakes,’’
by Per Bak, Kim Christensen, Leon Danon, and Tim Scanlon,
which appeared April 29, 2002, in Phys. Rev. Lett. (88, 178501),
Christensen, Danon, and Scanlon note that they were unaware
that publication of papers titled ‘‘Unified Scaling Law for
Earthquakes’’ as an Arthur M. Sackler Colloquium paper in
transfer and double publication policies of the journals. These
authors acknowledge significant overlap in data and figures
between the two papers and apologize for their oversight.
April 1, 2003 ?
vol. 100 ?
no. 7 www.pnas.org