Regional Metabolite Concentrations in Human Brain as
Determined
by
Quantitative Localized Proton MRS
Petra
J.
W.
Pouwels,
Tens
Frahm
The regional distribution of brain metabolites was studied in
several cortical white and gray matter areas, cerebellum, and
thalamus
of
young adults with use of quantitative single-voxel
proton
MRS
at
2.0
T.
Whereas the neuronal compound
N-
acetylaspartate is distributed homogeneously throughout the
brain,
N-acetylaspartylglutamate
increases caudally and ex-
hibits higher concentrations in white matter than in gray mat-
ter. Creatine, myo-inositol, glutamate, and glutamine are less
concentrated in cortical white matter than in gray matter. The
highest creatine levels are found in cerebellum, parallel to the
distribution of creatine kinase and energy-requiring pro-
cesses in the brain.
Also
myo-inositol has highest concentra-
tions in the cerebellum. Choline-containing compounds ex-
hibit a marked regional variability with again highest
concentrations in cerebellum and lowest levels and a strong
caudally decreasing gradient in gray matter. The present find-
ings neither support a metabolic gender difference (except for
a 1.3-fold higher myo-inositol level in parietal white matter of
female subjects) nor a metabolic hemispheric asymmetry.
Key words: proton spectroscopy; brain metabolism; cerebral
metabolite distribution.
INTRODUCTION
Quantification of absolute metabolite concentrations in
human brain is required not only for a better understand-
ing of
in
vivo
neurochemistry but also for an unambigu-
ous
assessment of pathologic alterations and pertinent
applications in clinical diagnosis and therapy monitor-
ing. Initial emphasis of localized proton MRS studies has
been placed on the evaluation of metabolite patterns
specific for tissue types such as gray and white matter
(1-4).
However, because both gray and white matter ex-
tend over large cortical and subcortical areas with dis-
tinct anatomic and functional properties, more detailed
differences are likely to be expected. This hypothesis has
been supported by a preliminary short-echo time proton
MRS study of cortical gray matter
(5).
Pertinent findings
are in line with earlier MRS reports on regional metabolic
heterogeneity although such studies were often compro-
mised by the use of mixed tissue volumes, long echo
times, and/or ratios of resonance intensities
or
concen-
trations (6-10).
The primary aim of the present work therefore was
a
reliable and detailed determination
of
absolute metabo-
MRM
3953-60
(1998)
From the Biomedizinische NMR Forschungs GmbH am Max-Planck-lnstitut
fur biophysikalische Chemie, Gbttingen, Germany.
Address correspondence to: Petra J. W. Pouwels, Ph.D., Biomedizinische
NMR Forschungs GmbH am Max-Planck-lnstitut fur biophysikalische
Chemie, D-37070, Gottingen, Germany.
Received March 24, 1997; revised June
2,
1997; accepted July
6,
1997.
This work was supported
by
a research grant from the European Union
(Program Biomed
II)
(to P.J.W.P.).
Copyright
0
1998 by Williams
&
Wilkins
All rights
of
reproduction in any form reserved.
0740-3194/98 $3.00
lite concentrations in multiple locations of cortical and
subcortical gray and white matter. In addition, the data
were analyzed for putative gender differences and hemi-
spheric asymmetry.
METHODS
Studies of 34 healthy volunteers
(17
female/l7 male, age
range 18-39 years, mean
27
5
4
years) were conducted at
2.0 T (Siemens Magnetom SP4000, Erlangen) using the
standard imaging headcoil. Informed written consent
was obtained before all examinations. A total of 41 stud-
ies were performed yielding about five spectra each.
Figure
1
shows representative volumes-of-interest
(VOI) selected for localized proton MRS. Cortical white
matter was measured in frontal, parietal, and occipital
brain (Fig. la) with small volumes of
4.1-6.4
ml to pre-
vent contamination with gray matter. Gray matter vol-
umes in frontal (8-12 ml), parietal (8-18 ml), and occip-
ital brain (8 ml) were located in paramedian positions
spanning
20
mm across the interhemispheric fissure (Fig.
1b). In addition, gray matter was investigated in insular
areas
(10
ml)
and thalamus (6.4 ml) (Fig. 1c). The cere-
bellum
(8
ml) was examined both in a central position
covering the vermis (Fig. Ib) and in hemispheric loca-
tions containing mainly white matter (Fig. Id).
Fully relaxed, short-echo time proton MR spectra were
recorded with
use
of a single-voxel STEAM localization
sequence
(TR/TE/TM
=
6000/20/30 ms,
64
accumula-
tions) as described previously
(11).
Particular care was
taken to optimize magnetic field homogeneity and water
suppression by localized adjustments. Spectroscopic
time-domain data were corrected for residual eddy cur-
rent effects
(12)
and calibrated in proportion to the actual
coil loading by using the transmitter reference amplitude
(3). For
display purposes only, spectral postprocessing
involved zero-filling to
4K
complex data points (2048
ms), Gaussian filtering (half-width 317 ms), and manual
phase-correction. Such spectra were not used for quanti-
fication.
Absolute metabolite concentrations were estimated by
LCModel, a user-independent fitting routine that em-
ploys a library of concentration-calibrated model spectra
of
all individual metabolites
(3,
13).
The method exploits
the full spectroscopic information
of
each metabolite and
not just isolated intense resonances. It therefore allows a
distinction even between metabolites with overlapping
signals provided they exhibit additional proton reso-
nances at different frequencies. Because the separation of
two overlapping metabolites also depends on the spectral
resolution attainable
in
vivo,
LCModel not only yields
individual concentrations of such metabolites, but also
offers a treatment as the
sum
of both. For example, total
NAA (tNAA) is determined in addition to separate con-
53
54
Pouwels
and
Frahm
FIG.
1.
T,-weighted
MR
images
(RF
spoiled
3D FLASH,
TRITE
=
15/6
ms,
flip
angle
20’)
showing
representative locations selected for proton
MRS
of
white matter
in
frontal, parietal, and occipital
brain (a); gray matter
in
paramedian frontal, parietal, and occipital brain, as well as
in
the central
cerebellum
(b);
gray matter
in
the insular area and
in
thalamus (c); and white matter
in
the cerebellar
hemisphere
(d).
centrations of N-acetylaspartate (NAA) and N-acetylas-
partylglutamate (NAAG)
(14).
Similarly, the amount of
choline-containing compounds (Cho) is expressed as the
sum of phosphorylcholine (PCh) and glycerophosphoryl-
choline (GPC) as suggested by
in
vivo
phosphorus MRS
(15)
and
in
vitro
proton MRS (16,
17).
The need for
including GPC in the library of model spectra was evi-
denced by a discrepancy between experimental spectra
and the corresponding LCModel fits, i.e., the appearance
of a 3.68-ppm proton resonance in the residual “base-
line,” when PCh was the only Cho-compound. Most
likely, this resonance belongs to the methylene protons
of the glycerol backbone of GPC.
Macromolecules with short
T2
relaxation times also
contribute to short echo-time spectra
(18,
19). However,
the baseline estimated by LCModel includes resonances
with broad linewidths. Thus, the quantitation of small,
mobile metabolites is not affected by macromolecular
contributions. In this study
no
attempt has been made to
further characterize residual broad resonances.
Concentrations are expressed as mmol/liter VOI (i.e.,
as mM) and
are
not corrected for CSF contributions and
residual
T2
relaxation effects.
T,
saturation was ignored
as
all experiments were performed under fully relaxed
conditions using sufficiently
long repetition times
TI?.
Be-
cause the model spectra were
measured with similar param-
eters as the
in vivo
spectra,
only differences in
T,
relax-
ation between
in vivo
and
in
vitro
must be taken into ac-
count. For
Tz
relaxation times
between 130 and 370 ms
in
vivo
and approximately
800
ms
in vitro
correction factors
range from 1.03 to
1.14
at an
echo time of
TE
=
20
ms (3).
However, because
T,
relax-
ation times of metabolites
show a regional dependence
(1,
4,
6,
7),
such small correc-
tions were ignored as they
would be at the expense of
time-consuming determina-
tions of metabolite
T,
values
in each VOI that are not toler-
able for related MRS studies of
patients.
For
similar reasons,
data were not corrected for po-
tential contributions from
CSF. Pertinent correction fac-
tors amount to about
1.05-1.07
for gray matter of young
healthy adults
(3,
4),
and even
less for other regions. In pa-
tient studies the observation of
reduced metabolite concentra-
tions may reflect brain atrophy
due to larger CSF contribu-
tions. In that case a compari-
son with healthy controls ei-
ther may be based on concentration ratios,
or
requires an
estimation of the CSF contribution to the VOI, eg., by
using image segmentation.
Statistical comparisons between regions were per-
formed with two-sided unpaired
t
tests assuming un-
equal variances. Gender differences were assessed by
applying these tests
to
metabolite concentrations of fe-
male and male subjects
for
each region separately. Data
from hemispheric regions (cortical white matter, insular
gray matter, cerebellar hemispheres, and thalamus) were
further examined for right-left asymmetry, both intrain-
dividually and interindividually. The intraindividual
asymmetry parameter (individual laterality) was defined as
the concentration ratio of a single metabolite
in
right versus
left
hemispheres for each subject with bilateral spectral
acquisitions. The interindividual asymmetry (species later-
ality) was assessed by comparing concentrations from right
and
left
hemisphere
of
all subjects with a two-sided
t
test.
RESULTS
White Matter
Representative proton MR spectra of frontal, parietal, and
occipital white matter are shown in Fig.
2.
The regional
Regional Concentrations
of
Human Brain Metabolites
55
WM
frontal
1
4.0
3.5 3.0
2.5 2.0
1.5
1.0
NAA
WM
parietal
~
4.0
3.5
3.0
2.5
2.0
1.5
1.0
WM
occipital
4.0
3.5 3.0 2.5
2.0
1.5
1.0
Chemical Shift
/
ppm
FIG.
2.
Localized proton
MR
spectra
(STEAM,
TR/TE/TM
=
6000/
20/30
ms,
64
accumulations) of a 22-year-old subject
in
frontal
(5.1
ml),
parietal (6.4
ml),
and occipital white matter
(5.1
ml).
Major
metabolites refer to N-acetylaspartate
(“I),
N-acetylaspartylglu-
tamate
(NAAG),
glutamate
(Glu),
glutamine
(Gln),
total creatine
(Cr),
choline-containing compounds (Cho), and myo-inositol
(Ins).
similarity is confirmed by an analysis
of
mean metabolite
concentrations
in
white matter of all subjects summa-
rized in Table
1.
Creatine and phosphocreatine (Cr) as
well as Cho are distributed uniformly throughout cortical
white matter. In contrast, the concentration of tNAA is
significantly lower in frontal white matter than in pari-
etal
or
occipital white matter. Separation of NAA and
NAAG shows that this distribution of tNAA is almost
exclusively due
to
NAAG, whereas no significant re-
gional differences are observed for NAA
(14).
Thus, the
contribution of NAAG to tNAA varies from
15%
in fron-
tal white matter to
25%
in parietal and occipital white
matter.
A
regional dependence is also observed for myo-inosi-
to1 (Ins). While this metabolite is equally distributed in
periventricular white matter frontally and occipitally, its
concentration in parietal white matter is significantly
lower. Finally, no regional variations are revealed for
glutamate (Glu) and glutamine (Gln).
Gray
Matter
Gray matter spectra of frontal, parietal, and occipital
brain, as well as of the insular area are shown in Fig.
3.
The most striking observation is a gradient in Cho yield-
ing an intensity decrease from frontal to occipital cortex.
A quantitative analysis of all gray matter spectra in Table
2
confirms highly significant differences in Cho concen-
tration between frontal and parietal, as well as between
parietal and occipital gray matter. When gray and white
matter regions are compared, even the highest levels of
Cho in frontal gray matter are lower than those found in
white matter.
The regional variability of tNAA, which has a higher
concentration in occipital than in parietal
or
frontal gray
matter, is due to enhanced levels of both NAA and NAAG
in occipital cortex. Whereas the concentration of NAA in
frontal and parietal gray matter is similar to that in white
matter, it is significantly higher in occipital cortex. Be-
tween and within white and gray matter, the largest
variability
is
observed for NAAG, which has lower con-
centrations in gray matter, but shows a caudal increase in
both tissues (14). Cr, Ins, Glu, and Gln are distributed
homogeneously throughout paramedian gray matter but
at higher levels than in white matter.
For
all metabolites,
concentrations in the insular area are higher than in
parietal gray matter. Similar concentration ratios with
respect to
Cr
suggest that this may partly be explained by
reduced CSF contributions in insular areas versus para-
median locations.
Concentrations of other brain metabolites like
scyllo-
inositol, glucose, lactate, taurine, y-amino butyric acid,
and aspartate are also determined by LCModel. However,
unfavorable conditions such as considerable spectral
overlap, strong coupling, and rather low concentrations
usually prevent a reliable determination in individual
subjects. On the other hand, group averages result in
reasonable estimates of pertinent concentrations in pari-
etal gray matter (n
=
19) of
0.14
t
0.09
mM for
scyllo-
inositol, 0.5
5
0.6
mM for glucose,
0.6
5
0.4 mM for
lactate, 1.0
2
0.4
mM for taurine,
1.4
t
0.8
mM for
y-amino butyric acid, and
1.4
t
0.8
mM for aspartate.
Table
1
Metabolite Concentrations
in
White Matter
(mM
f
SD)a
Frontal Parietal Occipital
n
=
22
n
=
20
n
=
27
Cr
5.7
f
0.5
5.7
f
0.6
5.5
IT
0.8
tNAA
9.6
2
1.1**
10.6
t
0.8
10.4
f
0.9
NAA
8.1
f
0.9
8.0
t
1.0
7.8
i
0.9
NAAG
1.5
5
0.9**‘
2.7
f
1.2
2.6
IT
1.0
Cho 1.78
f
0.41
1.68
5
0.27 1.64
t
0.21
Ins
3.8
f
0.V
3.1
rt
0.6
4.1
f
0.8***
Glu
7.0
IT
2.6
6.7
r
1.8
6.0
f
1.2
Gln
1.8
f
1.6
1.5
t
1.3
2.2
2
1.6
FWHMb
3.51
i
0.39
3.45
2
0.31
3.31
5
0.34
a
rnM
refers to rnrnol/liter
VOI.
FWHM:
full
linewidth at half maximum
(Hz)
of the Cr
signal
at
3.03
pprn.
***
p
<
0.001,
*’
p
<
0.01,
*
p
<
0.05
for comparisons to parietal white
matter.
56
Pouwels
and
Frahm
GM
frontal
4.0
3.5 3.0 2.5 2.0
1.5
1.0
GM
parietal
4.0
3.5 3.0 2.5 2.0
1.5
1.0
I
GM
occipital
4.0
3.5
3.0 2.5 2.0 1.5
1.0
GM
insular
4.0
3.5
3.0
2.5 2.0
1.5
1.0
Chemical
Shift
/
ppm
FIG.
3.
Localized proton
MR
spectra
of
a
21
-year-old subject
in
frontal
(1
2
ml),
parietal
(1
8
ml),
and occipital gray matter
(8
ml),
as
well as
of
a 25-year-old subject
in
the insular area
(10
ml).
Other
parameters as
in
Fig.
2.
Cerebellum and Thalamus
Spectra from the cerebellar hemisphere and vermis are
shown in the upper traces of Fig.
4.
Although the
VOI
in
the cerebellar hemisphere mainly contains white matter,
its
metabolite composition clearly differs from that of
cortical white matter (e.g., compare Fig.
2).
This is even
better reflected in the absolute concentrations given
in
Table
3.
Concentrations of Cr, Cho, and Ins are similar in
both cerebellar areas but significantly higher than in any
neocortical location. A prominent difference between the
two cerebellar regions
is
the concentration of tNAA,
which is significantly higher in the cerebellar hemi-
spheres than in the central part. This is due to much
higher amounts
of
NAAG in the hemispheres, whereas
the NAA concentration in both areas is constant and
slightly lower than in most cortical regions. Within the
cerebellum no regional differences are observed for Glu
and Gln. Whereas Glu levels are comparable to neocorti-
cal regions, Gln concentrations resemble those in gray
matter and are higher than in white matter.
A representative spectrum of thalamus is shown in the
bottom trace of Fig.
4.
As a result of large variations in
magnetic susceptibility that are mainly caused by a high
iron load in adult tissues, spectra typically suffer from
broad lines (i.e., a mean linewidth of
5.5
Hz as opposed
to 3.5
Hz
for most other brain regions at
2.0
T). Still,
concentrations
of
major metabolites can be determined
very reliably as demonstrated by standard deviations that
are comparable to those in other brain areas. The main
differences between gray matter in cortex and thalamus
are the relatively high Cho and tNAA concentrations in
the latter, whereas similar levels are found for
Cr
and Ins.
A separate analysis of NAA and NAAG indicates the
presence
of
high concentrations
of
NAAG in thalamus.
Gender Effects and Hemispheric Asymmetry
The intraindividual asymmetry parameters for
Cr,
tNAA,
Cho, and myo-Ins are summarized in Table
4.
For these
metabolites no asymmetry
is
observed in any of the re-
gions investigated bilaterally.
In
addition,
110
statistically
significant interindividual asymmetry was detected by a
t
test of group averages of all right- and left-hemispheric
spectra.
The data base was further evaluated for gender differ-
ences. Because these assessments were performed in ret-
rospect, the measured spectra were not equally distrib-
uted between the two sexes for all regions. However,
using the data available, no statistically significant gen-
der differences were obtained. The only exception was
for
Ins in parietal white matter where the ratio of female-
to-male concentrations reached 1.3
(p
<
0.01), i.e.,
3.4
i-
0.5 mA4for female
(n
=
14)
and
2.6
i
0.4
mMfor male
subjects
(n
=
6).
DISCUSSION
Apart from ignoring residual
T,
and CSF corrections, the
accuracy of an average concentration value as expressed
by its standard deviation in Tables 1-3, reflects the mea-
surement precision
of
individual spectra (depending
on
SNR
and
VOI
size), the reliability of the evaluation pro-
cess (characterized by LCModel estimates of individual
concentration values), and true intersubject variability.
For example, for relatively large gray matter volumes
with high spectral SNR, evaluation by LCModel results
in individual metabolite concentrations with experimen-
tal uncertainties of only 3-7% for tNAA and Cr, and
7-12% for Cho and Ins. The corresponding mean con-
centrations averaged over all subjects exhibit larger stan-
dard deviations of 6-12% for tNAA and Cr, 11-15% for
Cho, and 15-21% for Ins (Table
2).
This suggests that the
Regional Concentrations
of
Human Brain Metabolites
57
Table
2
Metabolite Concentrations
in
Gray Matter
(mM
f
SD)a
Cr
tNAA
NAA
NAAG
Cho
Ins
Glu
Gln
FWHM~
Frontal
n
=
12
6.4
f
0.7
8.4
i
1.0
7.7
f-
1.0
0.7
f
0.3
4.3
i
0.9
8.5
i
1.0
4.4
i
1.4
3.51
i
0.34
1.38
f
0.1
7***
Parietal
n
=
19
6.5
i
0.6
8.7
2
0.8
8.2
t
0.8
0.5
?
0.4
1.10
i
0.14
4.3
i
0.7
8.2
i
1.1
3.8
i
1.4
3.50
f-
0.33
Occipital
n
=
14
6.9
t
0.7
10.6
f
0.8***
9.2
i
0.9**
1.4
f
0.8**
0.88
f
0.10**'
4.1
f
0.6
8.6
f-
1.1
3.9
f-
1.1
4.06
t
0.31
Insular
n
=
16
7.0
5
0.6'
9.6
t
0.6**
8.7
f
0.8
0.8
f
0.5*
1.30
t
0.19**
4.7
i
0.6
8.8
i
0.8
4.9
t
1.6'
4.06
i
0.43
a
mM
refers to mmol/liter
VOI.
FWHM:
full linewidth at half maximum
(Hz)
of
the Cr signal at
3.03
ppm.
***
p
<
0.001,
**
p
<
0.01,
*
p
<
0.05
for comparisons to parietal gray matter
interindividual variability due to small differences in
VOI
location and/or true physiologic differences is larger
than the inaccuracy of the
MRS
quantification method.
Similar arguments hold true for white matter, although
smaller volumes and correspondingly lower SNR yield
Cerebellum hemisphere
4.0
3.5
3.0
2.5
2.0
1.5
1.0
Cerebellum central
4.0
3.5
3.0
2.5
2.0
1.5
1.0
Thalamus
4.0
3.5
3.0
2.5
2.0
1.5
1.0
Chemical
Shift
/
ppm
FIG.
4.
Localized proton
MR
spectra
of
a 25-year-old subject
in
the cerebellar hemisphere
(8
mi),
of
a 25-year-old subject
in
the
central cerebellum (8
ml),
and
of
an 18-year-old subject
in
thala-
mus
(6.4
ml).
Other parameters as
in
Fig.
2.
slightly larger experimental uncertainties for individual
concentrations.
It is also possible that the presence of intersubject
variability precludes the detection of small differences
between group averages of right and left hemispheres
or
between females and males. However, with the present
accuracy (Table
4),
our
data do not support concepts of
metabolic laterality previously hypothesized not only for
the temporal lobe (20,
21),
but also for parietal and oc-
cipital lobes
(21).
Pertinent reports are based on ratios of
resonance intensities
from
long-echo time recordings
without absolute quantification. Conversely, a long-echo
time spectroscopic imaging study is in line with
our
finding
of
metabolic hemispheric symmetry
(8).
In general, the present work confirms the absence of a
gender difference. This particularly applies to major me-
tabolites such as tNAA and
Cr
with the best experimental
accuracy. After ruling
out
that artifacts contribute to the
exceptional result of a 1.3-fold higher Ins concentration
in parietal white matter of female subjects, we fail to
report an explanation.
NAA
and
NAAG
The finding of a heterogeneous distribution of tNAA in
human brain
is
supported by studies in which no dis-
tinction between NAA and NAAG has been made. The
frontal to parieto-occipital increase of tNAA has been
described
(6,
71,
as well as the generally higher tNAA
concentration in white matter as compared to gray matter
(3,
4,
8,
9).
A relatively higher tNAA level in the cerebel-
lar hemisphere as compared to the vermis has also been
detected by spectroscopic imaging
(22).
Extending these observations by separating NAA and
NAAG demonstrates that the significant differences in
tNAA within and between different brain tissues are
mainly due to a corresponding regional variation
of
NAAG. The distribution of NAA
is
rather homogeneous,
with the exception of a slightly, but significantly higher
concentration in occipital gray matter, i.e., in the func-
tional area of the visual cortex. These findings parallel
neuroanatomic reports
of
a constant neuronal cell den-
sity in the neocortex except for a higher density exclu-
sively in primary visual cortex
(23,
24).
It further
sup-
ports the use of NAA as a neuronal marker but also