Tightly coupled brain activity and cerebral ATP
Fei Du†, Xiao-Hong Zhu†, Yi Zhang†, Michael Friedman‡, Nanyin Zhang†, Ka ˆmil Ug ˘urbil†, and Wei Chen†‡§
†Department of Radiology, Center for Magnetic Resonance Research and‡Department of Biomedical Engineering, University of Minnesota Medical School,
Minneapolis, MN 55455
Edited by Marcus E. Raichle, Washington University School of Medicine, St. Louis, MO, and approved March 18, 2008 (received for review
November 13, 2007)
A majority of ATP in the brain is formed in the mitochondria through
oxidative phosphorylation of ADP with the F1F0-ATP (ATPase) en-
zyme. This ATP production rate plays central roles in brain bioener-
getics, function and neurodegeneration. In vivo31P magnetic reso-
sole approach able to noninvasively determine this ATP metabolic
rate via measuring the forward ATPase reaction flux (Ff,ATPase). How-
ever, previous studies indicate lack of quantitative agreement be-
tween Ff,ATPase and oxidative metabolic rate in heart and liver. In
contrast, recent work has shown that Ff,ATPasemight reflect oxidative
animal study, using rats under varied brain activity levels from light
anesthesia to isoelectric state, to examine whether thein vivo31P MT
approach is suitable for measuring the oxidative phosphorylation
rate and its change associated with varied brain activity. Our results
conclude that the measured Ff,ATPasereflects the oxidative phosphor-
ylation rate in resting rat brains, that this flux is tightly correlated to
the change of energy demand under varied brain activity levels, and
that a significant amount of ATP energy is required for ‘‘housekeep-
ing’’ under the isoelectric state. These findings reveal distinguishable
characteristics of ATP metabolism between the brain and heart, and
they highlight the importance of in vivo31P MT approach to poten-
tially provide a unique and powerful neuroimaging modality for
noninvasively studying the cerebral ATP metabolic network and its
central role in bioenergetics associated with brain function, activa-
tion, and diseases.
brain metabolism ? In vivo31PMRS ? ATPase ? ATP synthesis ?
cells for supporting the energy needs of various cellular activities
and functions. In the brain, a majority of ATP is formed in the
mitochondria through oxidative phosphorylation of adenosine
diphosphate (ADP) catalyzed by the enzyme of ATP synthase
(ATPase) (1). A large portion of ATP energy is used in cytosol
to pump sodium and potassium across the cellular membrane for
maintaining transmembrane ion gradients and to support neu-
rotransmitters cycling and, thus, sustaining electrophysiological
activity and cell signaling in the brain. The ATP metabolism
regulating both ATP production and utilization plays a funda-
mental role in cerebral bioenergetics, brain function, and
neurodegenerative diseases (2–6).
The brain ATP metabolism is mainly controlled by ATPase
constitute a complex chemical exchange system involving ATP,
phosphocreatine (PCr), and intracellular inorganic phosphate
(Pi) (i.e., a PCr ^ ATP ^ Pi chemical exchange system) (7–10).
One vital function of this ATP metabolic network is to maintain
a stable cellular ATP concentration by adjusting the reaction
rates to ensure a continuous energy supply for sustaining elec-
trophysiological activity and maintaining normal function in the
brain. Logically, both the kinetics of ATP metabolism and the
associated chemical exchange rates should be more sensitive to
denosine triphosphate (ATP), a high-energy phosphate
(HEP) compound, is the universal energy currency in living
the brain activity and energy states than the steady-state ATP
concentration; they should provide an essential measure of
cerebral bioenergetics under different brain activity states. The
sole approach able to noninvasively and directly assess the
cerebral ATP metabolic rates is in vivo31P magnetic resonance
spectroscopy (MRS) combined with the magnetization transfer
(MT) method (9–16).
The in vivo31P MT approach has significantly advanced the
understanding of bioenergetics in the skeletal and cardiac mus-
cles and brain, yet it has also been a catalyst of controversies
regarding whether the measured ATP metabolic flux could truly
reflect the rate of cerebral oxidative phosphorylation, which is
expected to dominate the ATP production and bioenergetics.
The ATP synthesis (i.e., Pi 3 ATP reaction) rate measured by
the in vivo31P MT approach, in principle, can be mediated by
many cellular processes involving this conversion. Most pro-
cesses, however, are not fast enough relative to the intrinsic spin
lattice time (T1) of Pi to have a NMR detectable contribution.
Surprisingly, the reactions catalyzed by glycolytic enzymes have
been shown to be a major contributor to the Pi 3 ATP rate
measured by the in vivo31P MT approach in Escherichia coli (17),
yeast (18), liver (19), and myocardium (20). In the glucose
perfused rat heart, the exchange mediated by the two coupled
glycolytic enzymes of glyceraldehyde-3-phosphate dehydroge-
nase (GAPDH) and phosphoglycerate kinase (PGK) dominates
the unidirectional Pi 3 ATP rate¶measured by the in vivo31P
MT approach (20). This rate was shown to remain constant,
whereas the myocardial oxygen consumption rate was approxi-
mately doubled under high workload. Only after the GAPDH/
PGK contribution was eliminated, the Pi 3 ATP rate measured
by the in vivo31P MT approach in the myocardium was found to
match the rate of oxidative ATP synthesis rate (20).
The observations documenting the dissociation between the
measured Pi 3 ATP rate and the oxidative ATP synthesis rate
have diminished the significance of the in vivo31P MT approach
for studying the cellular oxidative phosphorylation and its role in
these biological systems. In contrast, we have recently demon-
strated that the unidirectional Pi 3 ATP rate measured by the
consistent with the oxidative ATP synthesis rate (9). This finding
has led us to hypothesize that the in vivo31P MT approach could
provide a unique and completely noninvasive tool for quantita-
31P MT approach in the resting human brain was
W.C. performed research; M.F. and N.Z. contributed new reagents/analytic tools; F.D.,
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
§To whom correspondence should be addressed. E-mail: email@example.com.
¶In a reaction A % B, unidirectional rates are given as the rate of A 3 B or B 3 A, whereas
the net rate is defined as the difference between the two unidirectional rates. The terms
rate and flux are used interchangeably.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2008 by The National Academy of Sciences of the USA
April 29, 2008 ?
vol. 105 ?
no. 17 ?
tively determining the oxidative ATP synthesis rate in the brain,
which is vital for studying the cerebral bioenergetics and brain
function (9). In the present study, we test this hypothesis and
attempt to establish a noninvasive31P MRS approach for directly
and quantitatively measuring the rate of oxidative phosphory-
lation and its change in the brain.
In vivo31P MT measurements, using the rat model, were
performed at 9.4 tesla (T), concurrently with electroencephalo-
gram (EEG) measurements under different brain activity levels
attained by using varied anesthetics and/or anesthetic doses.
Three essential questions were addressed herein: (i) Does the
ATP synthesis rate measured by the in vivo31P MT approach
reflect the cerebral oxidative phosphorylation rate? (ii) If it
does, is it more sensitive to the change of brain activity than the
steady-state HEP concentrations? (iii) If yes, what is the quan-
titative relationship between the cerebral oxidative phosphory-
lation rate and brain activity changes? The answers to these
questions are important as a technical validation of in vivo31P
MT approach in determining the cerebral oxidative phosphor-
ylation rate and for understanding the neuro-metabolic
In this study, the rat brain activity was altered under four
conditions by using three types of anesthetic chemicals [2%
isoflurane (IsoF); ?-chloralose (?-Ch), and sodium pentobarbi-
tal (Pen)], and two Pen doses (low-Pen vs. high-Pen) were
applied, in which high-Pen led to complete suppression of
spontaneous EEG signal (i.e., isoelectric state) (21). The in vivo
31P MT approach (9–12) was applied to measure both the
forward chemical exchange fluxes (Ff) from PCr to ATP [i.e.,
forward CK reaction flux (Ff,CK) related to the PCr 3 ATP
reaction catalyzed by CK] and from Pi to ATP [i.e., forward
ATPase reaction flux (Ff,ATPase) related to the Pi 3 ATP
reaction catalyzed by ATPase] by selectively saturating the
?-ATP resonance peak under varied brain activity states in the
rat. The rat brain activities were determined by EEG measure-
ments and quantified by the spectral entropy index (SEI) based
on the Shannon spectral entropy method (22). The details
regarding the measurements of Ff,ATPase, Ff,CK, and SEI may be
found in Materials and Methods.
Animal Physiology. Significant efforts were made to maintain rats
at normal and stable physiologic conditions. The relevant phys-
iological parameters measured under four anesthesia conditions
are summarized in Table 1. When anesthesia depth achieved by
using ?-chloralose or sodium pentobarbital increased, the blood
glucose concentration (CGlu) gradually and significantly de-
creased compared to the initial IsoF condition, and it reached
of CGlubetween the ?-Ch and low-Pen conditions was statisti-
cally insignificant (P ? 0.05 with unpaired two-tailed t test).
From IsoF to high-Pen, a statistically significant but small
difference in blood pH was noted, whereas the brain tissue pH
remained constant. Other physiological parameters had not
changed significantly under four anesthesia conditions except
that arterial blood pressure and heart rate decreased 37 ? 9%
and 19 ? 4%, respectively, at the isoelectric state compared with
the IsoF condition.
Test of Independence of T1 on Physiology Change. Because of the
limited length of experimental time for performing multiple in
vivo31P MT measurements in the same rat, the conventional
saturation transfer (CST) approach with steady-state saturation
of ?-ATP resonance peak was used to measure the ATP
metabolic rates in the rat brains with three anesthesia conditions
(group III: IsoF, low-Pen, and high-Pen; see details in Materials
and Methods). By this approach, the unidirectional ATP reaction
fluxes, Ff,ATPaseand Ff,CK, can be simultaneously deduced from
two in vivo31P spectra acquired in the absence (control) and
presence of the steady-state ?-ATP saturation according to Eq.
2. However, the intrinsic spin lattice relaxation times of Pi (T1,Pi)
and PCr (T1,PCr) in the absence of exchange with ATP have to
be determined at a given magnetic field strength and are, in
general, independent of the physiologic condition changes (9,
10). In this study, we have conducted a progressive saturation
transfer experiments (see Materials and Methods and Eq. 1) in
the rat brain at 9.4T under the IsoF and ?-Ch anesthesia
conditions (i.e., groups I and II). Intrinsic T1,PCr(i.e., the T1of
PCr in the absence of the PCr % ATP exchanges) was deter-
mined to be 3.68 ? 0.50 s for the IsoF condition and 3.97 ? 0.63 s
for the ?-Ch condition without a statistically significant differ-
ence (P ? 0.22). Similarly, intrinsic T1,Piwas 4.17 ? 0.58 s and
3.89 ? 0.69 s for the IsoF and ?-Ch condition (P ? 0.16),
respectively. These results indicate that both T1,PCrand T1,Piare
insensitive to the changes of animal physiology, and they can be
treated as constant (9, 13, 23). Thus, the averaged T1,PCr(3.83 ?
0.57 s) and T1,Pi(4.03 ? 0.64 s) were used in the present study
steady-state saturation measurements according to Eq. 2.
Coupling Between Varied Brain Activity and ATP Metabolic Rates.Fig.
1 shows typical in vivo31P MT spectra from a representative rat
brain acquired in the absence (Left) and presence (Center) of the
steady-state ?-ATP saturation under three anesthesia conditions
(Fig. 1A, IsoF; Fig. 1B, low-Pen; and Fig. 1C, high-Pen). The
corresponding EEG time courses measured in another rat are
also illustrated in Fig. 1 (Right), showing a gradual decrease of
brain EEG activity from the IsoF condition with burst EEG
pattern to the isoelectric state with silent EEG activity.
The in vivo31P MT spectra shown in Fig. 1 clearly demonstrate
and Piresonances due to the steady-state ?-ATP saturation and
the magnetization transfer effect through the exchange reaction
of PCr 3 ATP and Pi 3 ATP, respectively. The relative
Table 1. Physiologic parameters of rats measured by blood gas analysis and in vivo31P MRS under four anesthesia conditions
pCO2, mm Hg
pO2, mm HgCGlc, mg/dlABP, mm Hg
per min BloodTissue
33.4 ? 2.9
31.0 ? 3.0
30.6 ? 2.8
29.6 ? 4.8
141.5 ? 36.3
155.2 ? 20.2
131.6 ? 28.3
156.0 ? 34.9
162.3 ? 14.4
112.0 ? 12.4*
95.5 ? 15.3*
53.0 ? 6.0*
103 ? 7
108 ? 10
104 ? 13
64 ? 6*
370 ? 25
350 ? 33
394 ? 16
300 ? 14*
7.49 ? 0.03
7.43 ? 0.06
7.48 ? 0.05
7.36 ? 0.03*
7.15 ? 0.01
7.15 ? 0.02
7.16 ? 0.01
7.15 ? 0.04
The results were averaged from two blood samples collected before and after in vivo31P MT or EEG experiments under each physiological condition. The pH
values of brain tissue were measured according to pH ? 6.75 ? log[(? ? 3.26)/(5.70 ? ?)], where ? is the chemical shift difference between the Piand PCr
resonances. ABP, arterial blood pressure; HR, heart beating rate; CGlc, blood glucose concentration.
*The measured parameter was statistically significant different (P ? 0.05 with unpaired two-tailed t test) from that measured under the IsoF condition.
www.pnas.org?cgi?doi?10.1073?pnas.0710766105Du et al.
magnetization reduction is proportional to the forward reaction
rate constant according to Eq. 2. It is evident that the unidirec-
tional forward PCr 3 ATP and Pi 3 ATP rate constants (kf)
gradually decreased with increasing anesthesia depth. These
constants and the Pi and PCr concentrations were used to
calculate the forward unidirectional ATP reaction fluxes of
Ff,ATPaseand Ff,CKunder the four anesthesia conditions accord-
ing to Eq. 2. These data are summarized in Table 2.
were unaltered under IsoF, ?-Ch, and low-Pen anesthesia condi-
tions. Under the isoelectric state (i.e., high-Pen), however, we
observed a decrease of 8 ? 2% in [PCr] and an increase of 42 ?
6% in [Pi] compared to the IsoF condition, whereas [ATP] still
remained constant. The unidirectional reaction fluxes for PCr 3
ATP and Pi 3 ATP tended to decrease from the lightest (IsoF) to
deepest (high-Pen) anesthesia conditions. A similar trend was
consistently observed in the measured brain EEG activities (Table
2). The time domain EEG signal patterns and its entropy quanti-
fications at four anesthesia conditions with various brain activity
levels were similar to those reported in refs. 22 and 24. The general
tendency of deepening anesthesia is to suppress both cerebral ATP
the EEG signal to a lower-frequency domain and reduces the
spectral entropy index.
Fig. 2 shows the relationship between the rat brain activity
quantified by SEI versus the normalized ATP metabolic fluxes
through the CK (Ff,CK) and ATPase (Ff,ATPase) reactions, re-
spectively. The ?-Ch and low-Pen anesthesia conditions did not
induce a statistically significant difference (P ? 0.05) in the
values of Ff,CK(59.5 ? 9.3 versus 58.4 ? 10.5), Ff,ATPase(9.7 ?
1.2 versus 8.6 ? 1.0), and SEI (0.67 ? 0.04 versus 0.61 ? 0.03),
indicating a similar anesthesia depth (or anesthetic effect)
between these two conditions. In contrast, both the ATP met-
abolic rates and brain activities were significantly suppressed
under the isoelectric state related to the IsoF condition, resulting
in kfand Ffdecreases of 20 ? 8% and 28 ? 13% for the CK (i.e.,
PCr 3 ATP) reaction and 59 ? 17% and 48 ? 11% for the
ATPase (Pi 3 ATP) reaction, respectively, and reducing the
EEG spectral entropy index from 0.74 ? 0.06 to 0.44 ? 0.08.
Brain ATP Energy Budget and ‘‘House-Keeping’’ Energy. In general,
?90% of cerebral ATP production occurs in the mitochondria
through oxidative phosphorylation (25). ATP utilization mainly
occurs in the cytosol, providing chemical energy for supporting
various cellular functions, including phospholipid metabolism,
proteins synthesis, neurotransmitter cycling, and transportation
of ions across cellular membranes. A significant amount of ATP
energy budget in the brain is spent for maintaining and restoring
the transmembrane Na?/K?ion gradients that are diminished by
neuronal firing associated with spontaneous brain activity (4,
25). At the isoelectric state, we observed that ?50% of the
oxidative ATP synthesis capacity relative to the lightly anesthetic
state (i.e., IsoF condition) remained when EEG became silent.
This result indicates that brain ‘‘house-keeping’’ activities and
the spontaneous EEG activity under the IsoF condition each use
approximately one-half of the total brain ATP energy as mea-
of ‘‘house-keeping’’ ATP for maintaining cellular integrity in the
Cerebral metabolic activity can be assessed experimentally by
measuring the cerebral metabolic rates of oxygen (CMRO2)
and/or glucose (CMRglu) consumption. These metabolic pro-
cesses and their rates have been intensively studied by a variety
of techniques, and their relationships to varied brain activity
were investigated by using different anesthetics and doses (24,
26). The degree of CMRglcsuppression by deep pentobarbital
anesthesia was reported to be 40–60% compared with nitrous
-16 -11-6 -149
-16 -11 -6-149
Chemical Shift (ppm)
Chemical Shift (ppm)
under IsoF (SEI ? 0.75) (A), low-Pen (SEI ? 0.65) (B), and high-Pen (isoelectric;
acquired in the absence (control) (Left) and presence (Center) of ?-ATP satu-
ration. The magnetization ratio quantified by the NMR signals obtained at
steady-state saturation versus control was 56%, 59%, and 63% for PCr and
59%, 67%, and 78% for Pi at IsoF, low-Pen, and high-Pen anesthesia condi-
tions, respectively. The ratio changes indicate reduction in the measured ATP
metabolic rates with increased anesthesia depth. (Right) EEG time courses
recorded at three anesthesia conditions.
Table 2. Measurements of ATP energy metabolism and brain activity under four varied anesthesia conditions
Condition PCr, mM ATP, mMPi, mM
5.03.01.30.24 ? 0.02
0.21 ? 0.03*
0.21 ? 0.02*
0.19 ? 0.03*
0.17 ? 0.03
0.13 ? 0.02*
0.12 ? 0.02*
0.07 ? 0.03*
65.4 ? 7.3
59.5 ? 9.3
58.4 ? 10.5
46.8 ? 10.0*
12.1 ? 1.9
9.7 ? 1.2*
8.6 ? 1.0*
6.4 ? 1.9*
0.74 ? 0.06
0.67 ? 0.04*
0.61 ? 0.03*
0.44 ? 0.08*
4.99 ? 0.21
4.95 ? 0.14
4.60 ? 0.29*
3.03 ? 0.21
3.05 ? 0.15
2.98 ? 0.10
1.28 ? 0.13
1.26 ? 0.17
1.85 ? 0.19*
*The measured parameter was statistically significant different (P ? 0.05 with unpaired two-tailed t test) from that measured under the IsoF condition.
Spectral Entropy Index of EEG (SEI)
Normalized ATP Metabolic Flux
and full diamonds) and ATPase (Ff,ATPaseand open circles) reactions versus the
averaged spectral entropy index of EEG measured under varied brain activity
Ff,ATPasewere normalized to the values measured under the IsoF condition.
Correlations between the averaged ATP metabolic rates of CK (Ff,CK
Du et al.
April 29, 2008 ?
vol. 105 ?
no. 17 ?
oxide analgesia, light ?-chloralose anesthesia, or the conscious
animal (24, 27, 28). It was also reported that administration of
sodium pentobarbital reduced CBF and CMRO2by 66% and
61%, respectively, relative to a light anesthetic condition with
morphine sulfate (21). Other reports showed that thiopental-
induced isoelectric condition resulted in a reduction of global
CMRO2 and CMRglc by 47% and 61%, respectively, in the
non-human primate brain relative to the awake condition (29,
30). The range of 40–61% reduction in the brain metabolic rates
as found in these reports is in line with the reduction of ?50%
in the forward unidirectional ATPase reaction flux measured in
our study, indicating a significant ATP consumption for ‘‘house-
keeping’’ under the isoelectric condition. However, caution
a percentage of the total brain energy budget, because the latter
depends on the selection of the reference brain state (or baseline
level). For instance, ?35% reduction in CMRO2was reported in
the dog brain under anesthesia with 2% end-tidal isoflurane
concentration compared to an awake state (31). Thus, the
‘‘house-keeping’’ ATP requirement at the isoelectric state mea-
sured in our study would become ?33% in comparison with that
in the awake brain, consistent with the concept that the majority
of ATP energy is used to support brain activity and neuronal
signaling in the resting and conscious brain (4–6).
Although the majority of ATP is produced in neurons via the
oxidative phosphorylation of ADP, there is considerable oxida-
tive ATP synthesis in the glia cells based on in vivo13C MRS
studies (6, 27, 28). The fraction of brain energy consumption
between neuron and glia cells will change with varied brain
activity; and substantial energy consumption in the glia cells has
been reported under deep pentobarbital anesthesia or isoelectric
conditions (6, 27, 28). The agreement between the ATPase
reaction flux measurement in the present study and other
measurements of total oxygen or oxidative glucose consumption
as a function of brain activity suggests that the forward ATPase
reaction flux measured by the in vivo31P MT approach might be
contributed from both neuron and glia cells; although the glia
contribution is relatively small, in most cases, it could become
significant under the isoelectric state.
Coupling Among Forward ATPase Reaction Flux, Oxidative Phosphor-
ylation, and Brain Activity. Oxygen and glucose are the two most
important exogenous substrates for supporting brain energy and
function. The majority of brain glucose is oxidatively metabo-
lized. Under physiological conditions, oxidative metabolism is
tightly coupled with the oxidative phosphorylation process that
produces ATP molecules from Pi and ADP through the mito-
chondrial ATPase. This metabolic pathway can produce 15 times
more ATP molecules than the glycolysis pathway. Thus, the
oxidative glucose metabolism associated with the oxygen utili-
zation dominates the ATP production in the brain. Moreover,
the high energy demand in the brain requires correspondingly
high rates of ATP production and utilization, resulting in fast
chemical cycling among ADP and Pi versus ATP. The net rate
of oxidative ATP synthesis in this cycling is given as the product
of 2 ? (P:O ratio) ? CMRO2. The unidirectional rate of ATP
synthesis measured by the31P MT approach can be equal to or
greater than this rate. In all organs examined except in the brain,
unidirectional rate of ATP synthesis measured by31P MT (i.e.,
Ff,ATPase) exceeded net rate of oxidative ATP synthesis substan-
tially. The notable exception has been the brain as founded by
previous measurements in the resting human brain from our
group and previous rat brain studies in refs. 9, 10, and 13. This
surprising finding was further examined in the present study
under the conditions in which we have performed rigorous
CMRO2measurements, using high-field17O MRS, ultimately, to
test whether Ff,ATPasedetermined by31P MT changed as brain
energy demand was altered.
measured in this study and the net oxidative phosphorylation
rate calculated from the P:O ratio and the CMRO2 value,
and oxygen utilization (3–6, 9, 10). We had determined the
CMRO2value to be 2.2 ?mol?g?1?min?1in the rat brain (32)
under the same ?-chloralose anesthesia condition (i.e., ?-Ch
condition) used in the current study. Multiplication of this
CMRO2value with 2? P:O ratio with a P:O ratio of 2.4 (20, 33)
yields a rate of 10.6 ?mol?g?1?min?1for net ATP synthesis by
oxidative phosphorylation, in excellent agreement with Ff,ATPase
values of 9.7 ? 1.2 ?mol?g?1?min?1measured by31P MT under
the same animal condition in this study. This result further
validates the conclusion reached in the awake human brain (9,
10) and in the rat brain with sodium pentobarbital anesthesia
(13). The significance of the current data are particularly im-
portant, because the CMRO2and Ff,ATPasewere examined under
identical experimental conditions, albeit in different animals,
thus removing potential sources of error in the comparison. In
addition, the 50% reduction in Ff,ATPasemeasured by31P MT
under the isoelectric state compared to light anesthetic state in
this study was in excellent agreement with the 40–61% reduction
of CMRO2reported. This indicates that Ff,ATPaseremains ap-
proximately proportional to the CMRO2 value across a wide
range of brain activity. Consistent with this conclusion, our
results also demonstrate that with increasing anesthesia depth,
Ff,ATPasedecreased progressively (Table 1 and Fig. 1), and its
reduction was parallel to the decreased EEG activity as quan-
tified by SEI, revealing a tight neurometabolic correlation
between the oxidative ATP production rate measured by the in
vivo31P MT approach and the brain activity level assessed by
This study clearly demonstrates that a close relationship
between Ff,ATPaseand the level of brain activity can be extended
from the resting brain to a wide range of brain activity levels
from awaked human brain to the rat brain with complete
suppression of spontaneous EEG. In contrast, such a relation-
ship is lacking in the heart, because the measured Ff,ATPasein
myocardium is dominated by the glycolytic enzyme activities
(20). The cellular mechanism for understanding this discrepancy
between the brain and heart is still unclear (9, 10). However, the
measurement of Ff,ATPase relies on the Pi signal reduction
induced by the saturation of ?-ATP resonance peak and the
magnetization transfer effect between Pi and ?-ATP; it is
possible that the Pi signal detected by in vivo31P MRS in the
brain may mainly anticipate the Pi 3 ATP reaction through the
oxidative, rather than glycolytic, pathway; thereby, the measured
Ff,ATPaselargely reflects the majority of ATP production through
oxidative phosphorylation in the brain mitochondria (9, 10).
However, the exact mechanism requires further investigation,
using cell cultures of neuron and astrocyte.
Coupling Between CK Reaction Flux and Brain Activity. The high
energy demand in the brain results in fast chemical cycling
among cellular ATP, ADP, and Pi, which requires rapid energy
transportation between the cytosol and mitochondria. This can
can facilitate the effective transport of ATP from mitochondria
to sites of energy utilization and vice versa for ADP (8, 23). Thus,
a close coupling between the CK reaction flux and the brain
activity level could exist in a wide range of brain activity levels.
Such a neuro-metabolic coupling between Ff,CK and SEI was
examined in this study and shown in Fig. 2. In comparison with
Ff,ATPase, the measured unidirectional CK reaction flux was
relatively less sensitive to the different brain activity levels. This
is not unexpected, because Ff,CK exceeds Ff,ATPase by fivefold
(Table 2) under all conditions in this study, consistent with
reports in refs. 9, 10, and 13. These results reveal that the PCr
www.pnas.org?cgi?doi?10.1073?pnas.0710766105 Du et al.
and CK reaction have a supportive role in the cerebral ATP
HEP and Pi Concentrations vs. Brain Energy and Activity Changes.
Although the steady-state concentrations of cellular HEP and Pi
are closely linked to the cerebral ATP metabolism and are
vulnerable to brain pathology, they are relatively stable and
insensitive to brain energy demand and its change under phys-
iological conditions. This notion is clearly evident in this study,
showing a very stable brain ATP concentration under four
anesthesia conditions (Table 2) despite an ?50% reduction in
the oxidative ATP production rate in the rat brain at the
isoelectric state related to the light isoflurane anesthesia con-
dition. Thus, the cerebral ATP metabolic rate should logically
provide a more sensitive and accurate measure for quantifying
brain energetics and its change under different brain activity
states compared to the steady-state ATP concentration.
The ADP concentration and/or the ratio of [Pi]/[PCr] are
commonly used to reflect the oxidative metabolic activity and
bioenergetic level in the tissues under steady-state conditions.
They were found to increase significantly in the human skeletal
muscle during exercise (34, 35) and in the rabbit brain under a
prolonged epileptic condition (36) owing to higher energy
demand. One of the surprising findings from our study is the
observation of increases in [Pi]/[PCr] and [ADP] under the
isoelectric condition with high dose of pentobarbital anesthesia,
which presents a minimal level of metabolic and brain activity.
We found that [PCr] decreased 8 ? 2% and [Pi] increased 42 ?
6% at the isoelectric state compared with the light isoflurane
anesthesia condition without detectable variation in pH and
[ATP] (Tables 1 and 2). These changes result in an increase of
55% in [Pi]/[PCr] and 17% in [ADP]; the latter was calculated
by assuming that the creatine kinase reaction remained at
equilibrium during these two conditions and that the [PCr] is
approximately one-half of the total creatine concentration in the
brain. It had been shown that severe brain stress, such as hypoxia
[Pi]/[PCr] (36). To exclude this possibility, we have conducted a
localized in vivo1H MRS experiment to examine whether the
brain cellular lactate concentration, a common indicator of
hypoxia, increased under the isoelectric condition. Supporting
information (SI) Fig. S1 demonstrates the spectra acquired in a
representative rat brain under three anesthesia conditions (IsoF,
low-Pen, and high-Pen), showing no sign of significant lactate
increase under the isoelectric condition. This result excludes the
possible contribution of hypoxia to the increased [ADP] and
[Pi]/[PCr] under the isoelectric state as observed in our study. It
is, thus, possible that, under the isoelectric condition, Vmaxfor
oxidative ATP synthesis decreases and that corresponding in-
creases in ADP and Pi contents are required to maintain an
equilibrium balance between ATP production and utilization.
Summary. The following conclusions may be drawn from this
study: (i) the forward unidirectional ATPase reaction flux
measured by the in vivo
oxidative ATP synthesis rate in the brain; (ii) this flux is highly
correlated to brain energy demands at different brain activity
levels; and (iii) the cerebral ATP metabolic rate, which can be
explicitly and noninvasively determined by the in vivo31P MT
approach, would provide a more sensitive and quantitative
measure of brain bioenergetics and its change associated with
brain activity change under physiological condition compared
with the steady-state HEP and Pi concentrations. These com-
pelling findings reveal distinguishable characteristics of ATP
metabolism between the brain and other organs. They also
highlight the importance of the in vivo31P MT approach in
studying the central roles of oxidative ATP metabolism in
31P MT approach reflects the net
cerebral bioenergetics, brain function, and neurodegenerative
Materials and Methods
anesthetized via inhalation of 2% (vol/vol) isoflurane in the nitrous oxide/
oxygen (3:2) mixture. In group II, rats were anesthetized by ?-chloralose with
MT measurements, using the progressive saturation method (9–12), were
performed for both groups I and II to measure (i) the intrinsic spin lattice
the Pi3 ATP and PCr 3 ATP reaction, respectively.
In group III, rats were anesthetized under three conditions: (i) For IsoF, by
a 30 mg/kg bolus followed by continuous infusion (30 mg/kg/h); and (iii) for
high-Pen, by increasing the sodium pentobarbital infusion rate to 70
mg?kg?1?h?1to achieve an isoelectric state ?40 min after the high-dose
infusion. In vivo31P MT measurements, using the steady-state saturation
method (9–12), were performed for group III to measure the forward rate
constants and fluxes for the Pi3 ATP and PCr 3 ATP reactions, respectively,
under the three anesthesia conditions.
In group IV, rats were anesthetized under the four aforementioned anes-
thesia conditions: IsoF, ?-Ch, low-Pen, and high-Pen. This group was used for
EEG measurements with the same protocol used for the in vivo31P MT
All in vivo31P MT and EEG measurements were performed after animal
physiology approached a stable condition. The residue concentration of
isoflurane was controlled to remain ?0.1% (vol/vol) for the measurements
under ?-chloralose or sodium pentobarbital anesthesia conditions. Usually it
took ?90 min to switch anesthetics from isoflurane to ?-chloralose or sodium
pentobarbital. The femoral artery and vein were catheterized for blood
sampling, physiology monitoring, and infusing ?-chloralose or sodium pen-
throughout the experiments. All surgical procedures and experimental pro-
tocols were according to the guidelines of the National Institutes of Health
and approved by the Institutional Animal Care and Use Committee of the
University of Minnesota.
In Vivo31P MT Measurements. All In vivo31P MT experiments were conducted
INOVA console. A multinuclear radiofrequency (RF) surface-coil probe con-
sisting of a butterfly-shape1H surface coil and an elliptical-shape31P surface
coil with axes of 12 mm and 10 mm were used. The1H surface coil was used to
acquire brain anatomy imaging and for shimming magnetic field homogene-
ity, using the FASTMAP autoshimming algorithm (37). The31P detection
sensitivity was crucial for reliably measuring the small Pi signal and its change
caused by the MT effect to determine the oxidative ATP production rate (9,
10). To achieve maximal31P NMR sensitivity, the spatial localization of in vivo
31P MRS for acquiring the rat brain signal was achieved by the local detection
sensitivity profile defined by the31P RF surface coil without the use of other
to ensure that the Pi contribution from the muscle surrounding the brain was
negligible (Fig. S2), suggesting that the measured Pi signal and Ff,ATPasewere
mainly attributed to the rat brain tissue. The cortical gray matter likely
very small volume ratio of white mater to gray matter of ?0.14 in the rat (38)
(see SI Materials and Methods).
There are two saturation methods in performing the MT experiments. One
time (t) at near fully relaxed condition and then performing a curve fitting
according to Eq. 1 (9, 10):
Ma?t? ? Ma
where a stands for PCr or Pi resonance peak; Ma(t) and Ma
kfis the pseudo-first-order forward rate constant; and T1ais the intrinsic spin
lattice relaxation time of Pi (T1,Pi) or PCr (T1,PCr). Both kfand T1avalues can be
determined by the regression analysis of Eq. 1. The second saturation method
is the use of a sufficiently long saturation of ?-ATP resonance for reaching a
0are the resonance
Du et al.
April 29, 2008 ?
vol. 105 ?
no. 17 ?
steady-state of PCr or Pi magnetization (M* Download full-text
unidirectional ATP reaction fluxes can be determined by Eq. 2:
a). In this case, kfand the forward
;Ff,ATPase? kf,ATPase?Pi?;Ff,CK? kf,CK?PCr?.
The pulse sequence used for the in vivo31P MT experiments in this study is
resonance was selectively saturated with varied saturation time. For the steady-
steady-state magnetizations for both PCr and Pi. The31P spectra were acquired
under approximately fully relaxed condition (repetition time, 9 s) with other
acquisition parameters of 512 data points, a 5,000-Hz spectral width, and 128
signal averages. The quantification of the HEP metabolites was based on the
other anesthesia conditions were calculated from their PCr, ?-ATP, or Pisignal
integrals relative to that measured under the IsoF condition.
EEG Measurements. Two electrodes were used to record EEG signals from the
rat brain (group IV). One was put on the rat nose to served as a reference, and
the tip of the other was inserted into the rat cortex (2 mm deep from the
surface of skull, 3 mm posterior to bregma, and 3 mm from the brain midline)
through a small hole in the skull. The filtered EEG signal (0.0–30 Hz) was
sampled at a rate of 1,000 Hz, using commercially available EEG equipment
(Grass Instruments). Home-made software based on the Shannon spectral
entropy method (22) was used to quantify EEG data and to calculate the
spectral entropy index (SEI ? 1), where a high SEI value indicates a high brain
activity level. The EEG signal was divided into epochs of 10-s duration, and
then the spectral entropy index was calculated for each epoch in the 0.3- to
30-Hz frequency range. The reported EEG spectral entropy index value was
obtained by averaging 20 min data acquired under each physiological
Statistical Analysis. Unpaired two-tailed t tests were applied to statistical
analysis of the experimental data. The results are presented as mean
of Health Grants NS39043, NS41262, P41 RR08079, and P30 NS057091 and the
W. M. Keck Foundation.
1. Boyer PD (1999) What makes ATP synthase spin? Nature 402:247–249.
3. Clarke DD, Sokoloff L (1999) in Basic Neruochemistry: Molecular, Cellular and Medical
Aspects, eds Siegel GJ, et al. (Lippincott-Raven, Philadephia), pp 633–669.
brain. J Cereb Blood Flow Metab 21:1133–1145.
5. Raichle ME, Mintun MA (2006) Brain work and brain imaging. Annu Rev Neurosci
6. Hyder F, et al. (2006) Neuronal-glial glucose oxidation and glutamatergic-GABAergic
function. J Cereb Blood Flow Metab 26:865–877.
7. Slater EC, Holton FA (1953) Oxidative phosphorylation coupled with the oxidation of
ylation reaction and adenine nucleotide specificity. Biochem J 55:530–544.
8. Kemp GJ (2000) Non-invasive methods for studying brain energy metabolism: What
they show and what it means. Dev Neurosci 22:418–428.
Sci USA 100:14409–14414.
10. Du F, Zhu XH, Qiao H, Zhang X, Chen W (2007) Efficient in vivo31P magnetization
transfer approach for noninvasively determining multiple kinetic parameters and
11. Forsen S (1963) Study of moderately rapid chemical exchange reactions by means of
nuclear magnetic double resonance. J Chem Phys 39:2892–2901.
12. Brown TR, Ugurbil K, Shulman RG (1977)31P nuclear magnetic resonance measure-
ments of ATPase kinetics in aerobic Escherichia coli cells. Proc Natl Acad Sci USA
13. Shoubridge EA, Briggs RW, Radda GK (1982)31P NMR saturation transfer measure-
ments of the steady state rates of creatine kinase and ATP synthetase in the rat brain.
FEBS Lett 140:289–292.
14. Ugurbil K (1985) Magnetization-transfer measurements of individual constants in the
presence of multiple reactions. J Magn Reson 64:207–219.
15. Degani H, Alger JR, Shulman RG, Petroff OA, Prichard JW (1987)31P magnetization
transfer studies of creatine kinase kinetics in living rabbit brain. Magn Reson Med
16. Spencer RG, Balschi JA, Leigh JS, Jr, Ingwall JS (1988) ATP synthesis and degradation
rates in the perfused rat heart.31P-nuclear magnetic resonance double saturation
transfer measurements. Biophys J 54:921–929.
17. Mitsumori F, Rees D, Brindle KM, Radda GK, Campbell ID (1988)31P-NMR saturation
in Saccharomyces cerevisiae: Characterization of phosphate exchange reactions by
iodoacetate and antimycin A inhibition. Biochemistry 26:7483–7492.
perfused rat liver. Biochimica et Biophysica Acta 893:225–231.
20. Kingsley-Hickman PB, et al. (1987)31P NMR studies of ATP synthesis and hydrolysis
kinetics in the intact myocardium. Biochemistry 26:7501–7510.
21. Hyder F, et al. (2000) Dependence of oxygen delivery on blood flow in rat brain:
A 7 tesla nuclear magnetic resonance study. J Cereb Blood Flow Metab 20:485–
to the measurement of the electroencephalographic effects of desflurane. Anesthe-
23. Sauter A, Rudin M (1993) Determination of creatine kinase kinetic parameters in rat
brain by NMR magnetization transfer. Correlation with brain function. J Biol Chem
24. Koga K, Miura I (1988) A measurement of cerebral glucose uptake rate by31P MRS.
Biochem Biophys Res Commun 157:1258–1263.
metabolic rate in mammals. Physiol Rev 77:731–758.
26. Sokoloff L (1981) Relationships among local functional activity, energy metabolism,
and blood flow in the central nervous system. Fed Proc 40:2311–2316.
27. Choi IY, Lei H, Gruetter R (2002) Effect of deep pentobarbital anesthesia on neuro-
transmitter metabolism in vivo: On the correlation of total glucose consumption with
glutamatergic action. J Cereb Blood Flow Metab 22:1343–1351.
with brain activity. J Neurosci 24:11273–11279.
29. Nemoto EM, Klementavicius R, Melick JA, Yonas H (1996) Suppression of cerebral
metabolic rate for oxygen (CMRO2) by mild hypothermia compared with thiopental.
J Neurosurg Anesthesiol 8:52–59.
flow and oxygen and glucose metabolism in monkeys. J Neurosurg Anesthesiol
cerebral metabolism to low concentrations of halothane, enflurane, isoflurane, and
thiopental. Anesthesiology 46:28–34.
32. Zhu XH, et al. (2002) Development of17O NMR approach for fast imaging of cerebral
metabolic rate of oxygen in rat brain at high field. Proc Natl Acad Sci USA 99:13194–
33. Hinkle PC (2005) P/O ratios of mitochondrial oxidative phosphorylation. Biochim
Biophys Acta 1706:1–11.
34. Chance B, et al. (1985) Control of oxidative metabolism and oxygen delivery in human
Natl Acad Sci USA 82:8384–8388.
35. Chance B, et al. (1986) Multiple controls of oxidative metabolism in living tissues as
studied by phosphorus magnetic resonance. Proc Natl Acad Sci USA 83:9458–9462.
36. Prichard JW, Alger JR, Behar KL, Petroff OA, Shulman RG (1983) Cerebral metabolic
studies in vivo by31P NMR. Proc Natl Acad Sci USA 80:2748–2751.
shim coils. Magn Reson Med 29:804–811.
38. Zhang K, Sejnowski TJ (2000) A universal scaling law between gray matter and white
matter of cerebral cortex. Proc Natl Acad Sci USA 97:5621–5626.
www.pnas.org?cgi?doi?10.1073?pnas.0710766105 Du et al.