doi:10.1093/brain/awp010
Brain 2009: 132; 955–964
| 955
BRAIN
A JOURNAL OF NEUROLOGY
Early brain temperature elevation and anaerobic
metabolism in human acute ischaemic stroke
Bartosz Karaszewski,1,2 Joanna M. Wardlaw,1 Ian Marshall,3 Vera Cvoro,1
Karolina Wartolowska,4 Kristin Haga,1 Paul A. Armitage,1 Mark E. Bastin3 and
Martin S. Dennis1
1 Division of Clinical Neurosciences, University of Edinburgh, Western General Hospital, Crewe Rd, Edinburgh, EH4 2XU, UK
2 Department of Neurology of Adults, Medical University of Gdansk, ul. Debinki 7, 80-211 Gdansk, Poland
3 Medical Physics, Division of Medical and Radiological Sciences, University of Edinburgh, Western General Hospital, Crewe Rd, Edinburgh,
EH4 2XU, UK
4 Oxford Centre For Functional Magnetic Resonance Imaging of the Brain, Department of Clinical Neurology, University of Oxford, Oxford, UK
Correspondence to: Prof. Joanna M. Wardlaw,
Division of Clinical Neurosciences,
Western General Hospital,
Crewe Rd, Edinburgh,
EH4 2XU, UK
E-mail: [email protected]
Early after acute ischaemic stroke, elevation of brain temperature might augment tissue metabolic rate and conversion of
ischaemic but viable tissue to infarction. This might explain the observed link between pyrexia, severe stroke and poor outcome.
We tested this hypothesis by measuring brain temperature and lactate concentration with multi-voxel magnetic resonance
spectroscopic imaging across the acute ischaemic stroke lesion and normal brain as determined on diffusion imaging. We
compared patterns of lactate concentration (reported in ‘institutional units’) and temperature elevation in diffusion lesion
core, potential penumbra, ipsilateral and contralateral normal brain and with stroke severity. Amongst 40 patients with moderate
to severe acute stroke imaged up to 26 h after onset, lactate concentration was highest in the ischaemic lesion core (42 versus
26 units in potential penumbra, P50.05), whereas temperature was highest in the potential penumbra (37.7 versus 37.3 C in
lesion core, P50.05). Neither sub-regional temperature nor lactate concentration correlated with stroke severity. With increasing
time after stroke, ipsilateral brain temperature did not change, but contralateral hemisphere temperature was higher in patients
scanned at later times; lactate remained elevated in the lesion core, but declined in potential penumbral and ipsilateral normal
tissue at later times. We conclude that early brain temperature elevation after stroke is not directly related to lactate concentration, therefore augmented metabolism is unlikely to explain the relationship between early pyrexia, severe stroke and
poor outcome. Early brain temperature elevation may result from different mechanisms to those which raise body temperature
after stroke. Further studies are required to determine why early brain temperature elevation is highest in potential penumbral
tissue.
Keywords: ischaemic stroke; lactate metabolism; brain temperature; MR spectroscopy; pyrexia
Abbreviations: FID = free induction decay; MRSI = magnetic resonance spectroscopic imaging; OCSP = Oxfordshire Community
Stroke Project; PRESS = Point Resolved Spectroscopy; T2W = T2-weighted; UCP-2 = uncoupling protein 2
Received July 12, 2008. Revised December 12, 2008. Accepted January 5, 2009
ß The Author (2009). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved.
For Permissions, please email: [email protected]
956
| Brain 2009: 132; 955–964
Introduction
Brain and body temperature rise after ischaemic stroke
(Reith et al., 1996; Schwab et al., 1997; Castillo et al., 1998;
Zaremba, 2004; Karaszewski et al., 2006). In ischaemic stroke
models, artificially raised brain and body temperature were associated with ischaemic lesion enlargement (Barber et al., 2004;
Krieger and Yenari, 2004; Ren et al., 2004). In stroke patients,
pyrexia is associated with severe stroke and poor functional outcome (Reith et al., 1996; Castillo et al., 1998). Hypothermia
reduces ischaemic lesion growth in animal models (Zhang et al.,
2001; Berger et al., 2004; Iwata et al., 2005) and cooling may
improve outcome in stroke patients (Schwab et al., 1998;
De Georgia et al., 2004).
One potential explanation for the association between pyrexia
and poor outcome is that pyrexia would increase brain temperature and, by augmenting the brain metabolic rate, could result in
more rapid exhaustion of limited energy and oxygen supplies, and
increased production of free radicals and other toxic substances
in ischaemic tissue, thereby increasing conversion of ischaemic
but viable tissue to infarction (LaManna et al., 1988; Ichord
et al., 1999; Asai et al., 2000). In keeping with this concept,
pyrexia after stroke was associated with worsening neurological
deficit, larger infarcts and higher cerebrospinal fluid (CSF) concentrations of glutamate and glycine (Castillo et al., 1999). Cooling,
started at a mean of 16 h after severe hemispheric ischaemic
stroke in 12 patients, reduced glutamate, lactate and pyruvate
concentrations in normal and peri-infarct brain but not in infarct
core, as assessed with brain microdialysis catheter measurements
(Berger et al., 2002).
In ischaemic tissue, elevated lactate concentration indicates that
energy production has switched mainly to anaerobic mechanisms
(Stillman and Latchaw, 1993; Schurr, 2002; Fillenz, 2005). As
some lactate is produced and used as an energy substrate in the
normal brain, the increased lactate in ischaemic brain reflects
the net effect of overproduction, impaired usage and decreased
washout. Most previous studies suggested that the ratio of lactate
synthesis to utilization in ischaemic brain was positive (Stillman
and Latchaw, 1993; Schurr, 2002; Fillenz, 2005) and correlated
with the level of oxygen deprivation and ischaemia (Frykholm
et al., 2005).
We previously found that brain temperature was elevated in
ischaemic brain soon after stroke, prior to any increase in body
temperature, and was significantly higher in probable penumbral
tissue than in infarct core or normal brain (Karaszewski et al.,
2006). The temperature in the contralateral normal cerebral
hemisphere (which may be more closely related to body temperature) only became elevated close to 24 h after stroke.
Furthermore, in contrast to the known association between
pyrexia and poor outcome, we found no clear association between
elevated brain lesion temperature within the first 24 h and clinical
outcome (Karaszewski et al., 2006), suggesting that early perilesional brain temperature elevation might precede pyrexia and
might reflect a different mechanism to that which causes pyrexia
after stroke.
We hypothesized that if the early rise in brain temperature
increased the metabolic rate in ischaemic tissue, and was thus
B. Karaszewski et al.
the link between elevated body temperature, infarct growth,
severe stroke and poor outcome, then lactate concentration
should be highest in the same areas of the brain as those with
the highest temperature. Therefore, using magnetic resonance
spectroscopic imaging (MRSI), we determined the distribution of
lactate concentration (reported in MRSI-relevant ‘institutional
units’) and temperature across normal and ischaemic brain, in
acute stroke patients and tested for any association with stroke
severity.
Methods
Patient recruitment
We prospectively recruited patients with symptoms of moderate to
severe acute cortical ischaemic stroke without contraindications to
MRI. A stroke physician conducted a detailed medical history
and examination, measured baseline stroke severity using the
National Institutes of Health Stroke Scale (NIHSS) and determined
the Oxfordshire Community Stroke Project (OCSP) classification of
the stroke subtype (Bamford et al., 1991). The time of onset was
taken as the time when the symptoms were first observed or, if the
patient awoke with a stroke, then the time last known to be well.
Patients underwent MRI as soon as possible after stroke, but within
a maximum of 26 h. The study was approved by the local Research
Ethics Committee, and informed consent was obtained from the
patient or assent from their relative.
MRI and MRSI technique
All scans were carried out using a GE Signa Echospeed LX 1.5T
(General Electric, Milwaukee, WI, USA) MR scanner with self-shielding
gradients (22 mT/m maximum) and ‘birdcage’ quadrature head coil.
We performed axial T2-weighted (T2W) fast spin echo imaging, axial
diffusion tensor imaging (DTI, with field-of-view (FOV) 240 240 mm,
15 axial slices of thickness 5 mm, slice gap 1 mm, acquisition matrix
128 128, echo time 97.4 ms, repetition time 10 s and diffusion
sensitizing gradients with scalar b-values of 1000 s/mm2 applied
in six non-collinear directions), and multi-voxel Point Resolved
Spectroscopy (PRESS)-localized proton MRSI with the voxel grid
centered on the slice showing the maximum ischaemic lesion extent
on diffusion imaging. The MRSI voxel grid was carefully placed within
brain to include as much of the visible ischaemic lesion, ipsilateral and
contralateral normal brain as possible and to avoid contamination of
the spectra with lipid signal from bone marrow or subcutaneous fat
(Fig. 1). The imaging parameters for MRSI were: FOV 320 320 mm,
slice thickness 10 mm, acquisition matrix 24 24, echo time 145 ms
and repetition time 1000 ms. We used the echo time of 145 ms as
previous work showed it to be more reproducible than MRS detection
of metabolites at short echo times (Marshall et al., 2002). We used the
scanner’s standard three-pulse chemical shift selective (CHESS) water
suppression and shimming, optimized on the slice of interest. There is
sufficient residual water signal in each voxel for frequency (and hence
temperature) estimation. Spectroscopic data were Fourier transformed
for display and visual quality control purposes, but were modelled
in the time domain by five Gaussian components (corresponding to
choline, creatine, N-acetyl aspartate containing compounds and lactate) using the Advanced Method for Accurate Robust and Efficient
Spectral Fitting (AMARES) algorithm within the Magnetic Resonance
User Interface (MRUI) package (http://www.mrui.uab.es/mrui).
Lactate and temperature in human stroke
Brain 2009: 132; 955–964
| 957
Figure 1 Method of categorizing spectroscopy voxels according to the appearance of the brain on diffusion-weighted imaging.
The chemical shifts (i.e. frequency) of the fitted metabolite peaks were
reported to a precision of 0.001 ppm. Spectra were automatically
discarded if fitted line widths were less than 1 Hz or greater than
10 Hz, if the metabolite peaks were more than 0.1 ppm offset
from their expected values, if the voxels lay on the edges of the
PRESS excitation region or came from voxels containing CSF. In addition, all spectra were inspected visually, and discarded if judged to be
of poor quality, e.g. having a badly elevated baseline or containing
spurious peaks. For each phase encoding, 512 complex data points
were acquired with a sampling interval of 1 ms. Body temperature
was not monitored during scanning as all sequences operated well
within the specific absorption rate limits. The image acquisition took
20 min, not including patient settling time. Regular quality assurance
checks were performed (including a weekly spectroscopy check) with
appropriate phantoms to ensure stability of the scanner.
Image processing
Bulk patient motion and eddy current-induced artifacts were removed
from the DTI data using a three dimensional (3D) computational
image alignment program to register the component echo-planar
imaging volumes to the T2W volumes acquired with the DTI protocol.
Maps of the average DTI signal were obtained from the six DTI images
acquired for each slice. Spectroscopic images were interpolated to
a 32 32 matrix, yielding 1000 mm3 voxels and all processing was
carried out on a voxel-by-voxel basis (Figs. 1 and 2). Subsequent
processing consisted of zero-order phase correction and frequency
reference using the residual water signal (effectively bringing water
to a chemical shift of 4.70 ppm) followed by removal of the residual
signal using the Hanckel-Lanczos Singular Value Decomposition
method (van den Boogaart et al., 1994). Each MRSI data set took
9 minutes to acquire, and the data are effectively ‘averaged’ over
this period. Since frequency analysis was carried out on each individual
spectrum in parts-per-million, the absolute scanner frequency (i.e. Bo)
was not relevant.
Lactate concentration estimation
Lactate was identified by its characteristic appearance at echo
time = 145 ms (namely, an inverted doublet with peak separation of
7 Hz corresponding to the J-coupling). Metabolite quantification
took into account coil loading (using the scanner’s radio-frequency
958
| Brain 2009: 132; 955–964
B. Karaszewski et al.
Figure 2 Example of spectral data as acquired (left) and the corresponding fitted model (right). Voxel 433 of this patient was within
the PAL+ region.
transmitter gain) and receiver gain thus enabling inter-subject comparison of individual metabolite levels. Furthermore, all patients were
positioned in nearly the same place in the head coil; it was found
previously that coil uniformity is very good across an axial slice near
the centre (data not published). Our metabolite quantification was in
‘institutional units’ (Fig. 2). It is worth noting that using relative
concentrations of lactate to choline, creatine or N-acetyl aspartate
would not be appropriate as an alternative method for comparisons
between patients as all three of these metabolites change in stroke
(Muñoz Maniega et al., 2008). We did not correct for T1 and T2
effects since measuring T1 and T2 would make the scan sessions
unacceptably long for severely ill patients.
Brain temperature estimation
Cerebral temperature for each remaining voxel was calculated from
the relative chemical shifts of water and N-acetyl aspartate (Fig. 2).
Temperature-dependent changes in hydrogen bonding cause the
water chemical shift to vary linearly with temperature at 0.01 ppm
per C (Cady et al., 1995; Germain et al., 2001), whilst the chemical
shift of N-acetyl aspartate is independent of temperature. Both
chemical shifts are essentially independent of pH (Martin et al.,
1993). The use of MRUI for spectral quantification is advantageous
because the frequency resolution of MRUI is very high as quantification is carried out in the time domain, effectively looking for
small phase changes in the signals over the entire duration of the
free induction decay (FID). In other words, frequency resolution is
determined by the Cramer-Rao bounds (Marshall et al., 2006) and
not by the apparent digital resolution of the spectra, giving better
determination of any chemical shifts.
Validity of MRS temperature
measurement
This technique has been validated previously (Corbett et al., 1995;
Corbett et al., 1997; Karaszewski et al., 2006; Marshall et al., 2006;
Harris et al., 2008). Brain MRS thermometry readings correlated highly
(r = 0.93) with temperatures measured by the fluoroptic method in pigs
during body cooling and heating (Corbett et al., 1997). We previously
validated the MRSI thermometry technique with phantoms of known
temperature in normal volunteers (to test within and between session
repeatability) (Marshall et al., 2006) and during head cooling in
normal subjects (Harris et al., 2008). We found very close correlation
between temperature measured spectroscopically and by thermometer
in a phantom (Marshall et al., 2006). We scanned four healthy volunteers, four consecutive times within one MRI session: the mean brain
temperature across all scan times was 36.5 C; within each volunteer
and each scan time, and there was no statistically significant difference
between the hemispheres (difference 0.14 C, P = 0.302). Moreover,
across the four sequential MRS scan acquisitions within each
volunteer, we detected a statistically significant reduction in brain
temperature of 0.09 per scan (P = 0.0001), possibly due to the cooling
effect of the air current in the bore of the magnet or relative cerebral
inactivity during the scan (Marshall et al., 2006). Head cooling in
healthy volunteers caused a statistically significant drop in brain
temperature of 0.45 (P = 0.01) as measured by MR spectroscopy
and was 0.3 lower in voxels located near the brain surface compared
with those located centrally (Harris et al., 2008) but this difference
was not statistically significant (P = 0.15). This contrasts with numerical
modelling which suggests that the brain surface temperature relative
to the brain core may only differ in the outer 1 cm and by a maximum
of 1 (Nelson and Nunneley, 1998). This was confirmed using direct
Lactate and temperature in human stroke
Brain 2009: 132; 955–964
| 959
using Pearson correlation co-efficients, and between temperatures or
lactate concentrations and NIHSS scores using Spearman correlation
co-efficients. We used t-tests to compare lactate concentrations/
temperature readings in different brain sub-regions in patients scanned
before versus after 12 h of stroke. We performed all comparisons
‘within patient’ or ‘within region and within patient’ thereby avoiding
potential confounding by fluctuations in scanner performance or other
factors that might influence apparent between-patient and betweentissue region differences. For maximum transparency, the statistically
insignificant results are shown in the tables in addition to the significant ones. The P-values were not adjusted to allow for multiple testing
since the statistical methods for their correction may be inappropriate
(Perneger, 1998).
temperature probes inserted during neurosurgery where lower
temperatures were only found in the outer 1 cm of brain tissue in
normothermic patients, but this could reflect direct surface cooling
by exposure during surgery (Stone et al., 1997) and would explain
the difference to our results in volunteers through the intact cranium
(Harris et al., 2008).
Tissue regional classification
MRSI and DTI data were co-registered using an affine transformation.
The MRSI voxel grid was superimposed on the DTI data, and the
voxels were coded according to the DTI visual appearance, blind to
all other information. We used an operational tissue classification
based on the DTI visual appearance (Fig. 1), as described previously
(Karaszewski et al., 2006). This distinguished definitely abnormal DTI
voxels (DAL, probable core); possibly abnormal DTI voxels (PAL); a rim
of normal appearing tissue one voxel thick immediately outside the
possibly abnormal tissue (PAL+); and ipsi- and contra-lateral normal
voxels (INL and CNL respectively). In this operational classification,
‘possibly abnormal’ and ‘possibly abnormal plus’ tissues were
considered to correspond to ‘potential penumbra’ (Karaszewski
et al., 2006).
Results
We recruited 40 patients aged 58–95, with NIHSS of 1–29 (for full
demographic details see Karaszewski et al., 2006). Brain lactate
concentrations were not obtained for one patient due to technical
reasons. Twenty-six patients were scanned within the first 12 h of
stroke and fourteen patients between 12 and 26 h after stroke.
Lactate concentration and temperature
in brain sub-regions
Statistical analysis
We calculated mean temperatures and lactate concentrations in
definitely and possibly abnormal, possibly abnormal+ and normal
voxels. We examined lactate concentrations, tissue temperatures and
National Institutes of Health Stroke Scale (NIHSS) for associations
within the whole group and in those imaged within 12 h and after
12 h. Some data were not normally distributed (W Shapiro-Wilk test)
so we used non-parametric tests where appropriate. We tested
for correlations between lactate and temperature, and between temperatures or lactate concentrations and delay from stroke to imaging
Mean lactate concentration (i.u.) was highest in the definitely
abnormal tissue on DTI and lowest in contralateral normal
voxels
(DAL = 42.1,
SD = 24.3;
PAL = 26.0,
SD = 16.3;
PAL+ = 14.5, SD = 10.6; INL = 10.4, SD = 8.8; CNL = 8.7,
SD = 4.6; P50.05; Fig. 3). Temperature was highest in potential penumbral tissues (37.66 C), then in definitely abnormal
tissue (37.3 C), and then contralateral (37.25 C) and ipsilateral
Lactate [i.u.]
Temperature [°C]
45.0
40.0
42.1
37.7
37.66
37.63
35.0
37.6
30.0
37.5
26.0
25.0
20.0
37.4
37.3
37.30
37.25
37.2
15.0
14.5
10.0
37.16
10.4
37.1
8.7
37.0
5.0
0
36.9
DAL
PAL
PAL+
INL
CNL
Figure 3 Mean lactate concentration (i.u.; grey bars; left-hand axis) and temperature ( C; blue bars; right-hand axis) in sub-regions of
ischaemic and normal brain at baseline. DAL = definitely abnormal tissue; PAL = possible abnormal tissue; PAL+ = tissue one voxel thick
immediately outside the definitely or possible abnormal tissue; INL = ipsilateral normal brain; CNL = contralateral normal brain;
i.u. = standardized institutional units.
960
| Brain 2009: 132; 955–964
B. Karaszewski et al.
(37.16 C) normal voxels respectively (Fig. 3). Thus the highest
lactate levels and highest temperatures occurred in different
regions of the ischaemic brain. There was no association
between lactate and temperature in any of the individual
sub-regions.
Discussion
Lactate elevation and temperature elevation occur in different
regions of the acute ischaemic stroke lesion. Lactate was highest
in the lesion core and temperature was highest in potential
penumbral regions. There was no association of either lactate
or temperature with initial stroke severity as measured by
NIHSS. We did not find any correlations between brain temperature in individual sub-regions and functional outcome in contrast
to the well established relationship between pyrexia, severe stroke
and clinical outcome (Reith et al., 1996; Castillo et al., 1998).
Lactate concentrations and tissue temperature elevation also
showed different temporal profiles. In patients scanned within
the first 12 h, lactate concentration was significantly higher in
potential penumbral tissue and in ipsilateral normal tissue than in
patients scanned after 12 h, suggesting that patients being
scanned later had less potential penumbral tissue. We previously
showed that patients scanned within the first few hours of stroke
had significantly higher temperatures in ischaemic lesion core than
in contralateral brain, whereas in those scanned later there was no
difference in temperature between these brain regions because the
temperature in contralateral normal voxels rose with time
(Karaszewski et al., 2006). Temperature in the normal brain contralateral to the ischaemic lesion is most likely to reflect body core
temperature. Contralateral brain temperature elevation 12–26 h
after stroke is temporally consistent with patterns of pyrexia
observed in other studies. Considering the ischaemic hemisphere,
these observations suggest that increased lactate concentration is
unlikely to be consequent upon rises in local brain temperature, at
least within the first 26 h. This, in turn, suggests that the early rise
in brain temperature is not simply an early systemic inflammatory
response to ischaemia—if it were then the most ischaemic areas
i.e. those where lactate concentration was highest—reflecting a
positive lactate production/utilization ratio in the ischaemic brain
(Stillman and Latchaw, 1993; Schurr, 2002; Fillenz, 2005), should
have shown the highest temperatures.
The study has limitations including the relatively small sample
size, and the difficulty in drawing conclusions about temporal
changes from individual patients imaged at different times. It is
Brain lactate, temperature and stroke severity
There was no association between lactate or temperature and
NIHSS in any individual regions (Table 1).
Brain lactate and temperature and time from
stroke onset
In patients scanned 412 h after stroke, lactate concentration was
lower in possibly abnormal+ and ipsilateral normal tissue than in
patients scanned within 12 h indicating that lactate concentrations
in these regions fell with increasing time after stroke (Table 2).
However there was no change over time in possibly or definitely
abnormal or contralateral normal voxels (Table 2), i.e. lactate
stayed elevated in possibly and definitely abnormal tissue and
remained low in contralateral normal tissue. Brain temperature
did not change with time, except that temperature of contralateral
tissue was higher in those scanned at later than at earlier times
(Table 2).
Table 1 Correlations between lactate concentration and
temperatures in individual brain regions and baseline
stroke severity (NIHSS)
Brain sub region
DAL
PAL
PAL+
INL
Lactate correlation with stroke severity by sub-region
Spearman r-value
0.02
0.3
0.05
0.11
P-value
0.93
0.07
0.77
0.54
CNL
0.25
0.12
Temperature correlation with stroke severity by sub-region
Spearman r-value
0.13
0.22
0.09
0.18
0.17
P-value
0.49
0.18
0.62
0.29
0.29
DAL = definitely abnormal tissue; PAL = possible abnormal tissue; PAL+ = tissue
one voxel thick immediately outside the definitely or possible abnormal tissue;
INL = ipsilateral normal brain; CNL = contralateral normal brain.
Table 2 Association between time to scanning, mean lactate concentrations and temperatures in brain sub-regions
DAL
PAL
Lactate concentration
Pearson r-value (correlation)
P-value (correlation)
Mean512 h of stroke
Mean412 h of stroke
P-value (t-test)
0.17
0.41
43.35
40.43
0.75
0.35
0.076
29.09
20.01
0.10
Brain Temperature
Pearson r-value (correlation)
P-value (correlation)
Mean512 h of stroke
Mean412 h of stroke
P-value (t-test)
0.07
0.74
37.47
37.42
0.91
0.03
0.88
37.67
37.63
0.9
Significant values are indicated in boldface.
PAL+
INL
CNL
0.52
0.006
17.21
9.13
0.029
0.45
0.02
12.66
5.95
0.029
0.24
0.23
7.9
10.18
0.15
0.002
0.99
37.46
37.68
0.56
0.18
0.37
37.25
37.09
0.54
0.39
0.046
37.07
37.5
0.03
Lactate and temperature in human stroke
possible that patients with larger strokes presented earlier than
those with small lesions, were scanned earlier and could influence
the analysis of the time-dependent changes. Alternatively, MRI
timing may have been delayed in the most severely ill patients
due to the longer time needed for stabilization of their clinical
condition, but there was no evidence of this. Serial imaging of
the same patients every few hours soon after acute stroke
would interfere with their acute stroke care. We limited the potential bias due to differences in stroke severity between patients
arriving early and late by focusing on within patient analyses.
However there are likely to have been differences between
patients scanned early and late, emphasizing the need for larger
studies. We performed quality controls for MRS by regular checking against phantoms of known metabolite concentration. Our
metabolite measurements were in arbitrary units, but corrected
for scanner gain settings to enable comparisons between patients,
and therefore referred to as ‘institutional units’. Because of the
unknown relaxation times (T1 and T2) in pathological tissues,
we were unable to ‘normalize’ the measurements and interpret
them in terms of absolute millimolar metabolite concentrations.
Unfortunately, detailed measurement of T1 and T2 would considerably prolong imaging times and not be possible in acutely
ill stroke patients. A limitation of the spectroscopic imaging
technique is the difficulty of achieving a satisfactory shim over
the whole slice, especially on scanners that are not fitted with
high-order shimming coils. Several spectra from frontal regions
had to be discarded because of susceptibility effects from the
nearby sinuses. The radio-frequency excitation is non-uniform
near the edges of the PRESS localized MRS volume, leading to
several poor quality spectra from cortical regions of interest. We
discarded all poor quality spectra as part of a rigorous quality
control process. In in vitro studies, we have found that modest
variations in shimming (linewidths ranging from 1.4 Hz at the
centre of a phantom to 1.7 Hz at the edges) and water suppression factor (range from 85 to 105) do not lead to corresponding
variations in estimated temperatures (data submitted for publication). Although the situation is more complex in vivo, MRSI
appears to be capable of regional temperature estimation. Future
studies should examine the degree to which metabolite concentrations can be evaluated precisely in the situation where limited
signal-to-noise ratio is limited and to quantify the effect of other
variables that could affect temperature estimation. Inevitably,
patients with small lesions will have contributed more normal
brain voxels than patients with large lesions, the influence of
which we tried to minimize using within patient analyses. The
variable lesions with different numbers of voxels contributing to
each tissue type are an inevitable factor in acute stroke studies
and emphasize the need for larger confirmatory studies.
In the meantime, MRS appears to be a valid method for
measuring brain temperature in patients non-invasively using the
principle that water frequency shift relative to N-acetyl aspartate is
temperature-dependent (Cady et al., 1995; Corbett et al., 1997;
Marshall et al., 2006). Experimental studies in test phantoms
(Corbett et al., 1995; Ishihara et al., 1995) and experimental
models (Corbett et al., 1995; Ishihara et al., 1995; Corbett
et al., 1999; Kuroda et al., 2003; Trubel et al., 2003;
McDannold et al., 2004) show very close correlation between
Brain 2009: 132; 955–964
| 961
temperature measured by MRS and implanted probes (r = 0.93 in
Corbett et al., 1997). In 4 normal volunteers, we showed that
the technique is capable of detecting small differences in brain
temperature as small as 0.09 C (Karaszewski et al., 2006;
Marshall et al., 2006). Each volunteer was scanned four times
and within each volunteer and each scan time there was no statistically significant difference in MR measured brain temperatures
between the hemispheres (difference 0.14 C, P = 0.302) and the
mean temperature across all scan times was 36.5 C. However,
across the four sequential MRS scan acquisitions within each
volunteer, we were able to detect a statistically significant reduction in temperature of 0.09 per scan (P = 0.0001), presumably
due to cooling by the air current in the bore of the magnet (or
relative cerebral inactivity during the scan). Convective device
head cooling in the scanner in healthy volunteers caused a statistically significant drop in brain temperature of 0.45 (P = 0.01)
as measured by MRS, of similar magnitude within the ‘core’,
‘intermediate’, and ‘surface’ located voxels (Harris et al., 2008).
These data did not support the hypothesis that there is a major
surface-to-core temperature gradient in the brain (Nelson and
Nunneley, 1998) in the intact skull.
Previous studies of brain temperature and its relation to different metabolites were limited by reliance on invasive measurement
techniques. Brain tissue temperatures were measured using
temperature probes implanted either directly in brain tissue
(Stone et al. 1997; LaManna et al., 1998; Nemoto et al., 2005)
or suspended in CSF (e.g. Meylaerts et al., 2000). Although both
methods might be useful in some circumstances, they can only
sample brain temperatures from discrete regions. Inserting any
probe into the brain exposes the brain surface to cooler room
air, is likely to change any physical heat exchange conditions
and thus alter temperature readings. Invasive probes may cause
microlesions and a local inflammatory response, thus possibly
changing the tissue temperature. The MRS thermometry is not
confounded by any of these artifacts. Although further standardization, including in in vivo models is essential to quantify the
effect of different known and unknown factors, the available
data suggest that the technique provides reliable and valid
temperature measurements within the brain. Other potential
methods for non-invasive thermometry, such as infrared thermal
imaging, are useless for deep brain thermometry due to their
limited depth penetration.
Previous studies indicate that the highest lactate concentrations
should be in the most ischaemic tissues (Stillman and Latchaw,
1993; Schurr, 2002; Frykholm et al., 2005; Fillenz, 2005).
Lactate is synthesized constantly at low levels in healthy brain
and is the principal product mainly of anaerobic and, as suggested
recently, possibly also aerobic neuronal metabolism (Schurr and
Payne, 2007). It might be utilized by mitochondria as a substrate
for oxidative energetic pathways followed by oxidation to pyruvate (Bergersen, 2007; Schurr and Payne, 2007). Additionally,
some CSF lactate is likely to derive from brain cells although
there are no actual studies to support this thesis. Deprivation of
tissue oxygen causes increased lactate production, but simultaneously might impair its rate of cellular utilization (Frykholm
et al., 2005; Hillered et al., 2005; Macri et al., 2006; Bergersen,
2007). Nevertheless, the lactate synthesis/utilization ratio is
962
| Brain 2009: 132; 955–964
thought to be positive in ischaemic nervous tissue and therefore
useful to estimate the level of ischaemia in acute ischaemic
stroke. Our data concur with this interpretation as the lactate
concentration was the highest in the core of the ischaemic
lesion and the lowest in the brain remote from the ischaemic
area. A previous study (Parsons et al., 2000) suggested that
ischaemic core lactate/choline ratio measured by single voxel
MRS was associated with poor functional outcome. In our
sample, core lactate concentration was not associated with
stroke severity or clinical outcome. This difference might be
because we did not use the ratio of lactate to choline because
choline (or other potential reference-metabolites such as creatine
or N-acetyl aspartate) is not normal and therefore an unreliable
denominator in ischaemic tissue (Muñoz Maniega et al., 2008), or
our core lactate concentration was taken as a mean of all ischaemic core voxels rather than from a single large lesion-centered
voxel (Parsons et al., 2000), or both factors combined.
How do we interpret our findings? We hypothesized that elevated brain temperature might promote conversion of ischaemic
but viable tissue to infarction by augmenting metabolic reactions
and promoting rapid exhaustion of residual energy. However, the
tissue with the highest temperature elevation was not the most
ischaemic as assessed by lactate concentration. This questions
which other factors might promote the ischaemic brain temperature increase. An important one is local tissue blood flow (ParryJones et al., 2008), discussed in our previous work (Karaszewski
et al., 2006). Reduced brain perfusion could act in several ways.
Reduced blood flow could impair heat exchange resulting in
higher ischaemic lesion temperature where cells were still actively
metabolizing. Alternatively, the same scenario might result in
lower ischaemic than normal brain temperature if body and
brain temperature were elevated due to pyrexia due to failure to
transfer heat into the ischaemic lesion, whether or not the ischaemic lesion cells were still actively metabolizing. Blood flow is rarely
completely absent from the ischaemic lesion so the relative contribution of blood flow to temperature distribution probably lies
between the two scenarios. We previously found that patients
with higher blood flow levels in the ischaemic tissues had lower
tissue temperatures (Karaszewski et al., 2006) consistent with
recent experimental data showing marginally lower ischaemic
lesion temperatures following reperfusion (Parry-Jones et al.,
2008). However, this may not account for the difference in
temperatures observed between definitely abnormal and probable
penumbral tissues which would have fallen within the whole area
of reduced perfusion (by mean transit time). Similarly, if we
assume that local inflammatory response, which is likely to be
more intense in the more ischaemic tissues, is linked to the
tissue temperature elevation, than the temperature should
be the highest within the lesion core rather than in the healthylooking tissue surrounding the core, but it is not.
The early rise in ischaemic lesion temperature may result from
different processes to those which lead to pyrexia following stroke.
One explanation for the observed brain temperature distribution,
and the lack of association between brain temperature and stroke
severity or clinical outcome, might be the existence of a local
process that is at first neuroprotective and secondly, causes
regional heat generation. An example is up-regulation of the
B. Karaszewski et al.
gene for uncoupling protein 2 (UCP-2), an inner mitochondrial
membrane molecule, which regulates ATP synthesis by uncoupling
oxidation from phosphorylation, thus dissipating energy as local
heat (Horvath et al., 2003). UCP-2 is up-regulated in human
ischaemic brain (Nakase et al., 2007) and following transient
ischaemic attack in animal models, increasing resistance to neuronal damage with subsequent repeated ischaemia (Mattiasson
et al., 2003). Its overexpression is associated with decreased
brain damage following experimental acute stroke and better
neurological outcome (Mattiasson et al., 2003; Kim-Han and
Dugan, 2005). UCP-2 up-regulation may therefore constitute a
natural neuroprotective mechanism, possibly part of the ischaemic
preconditioning process, and simultaneously be responsible for the
early rise in lesion brain temperature after stroke.
Many other factors could influence local heat production in the
brain. These might be related to some common stroke-coexisting
pathologies such as impaired insulin action affecting many cellular
metabolic pathways in diabetes and metabolic syndrome, or to
some individual reactions e.g. the strength of immune or stress
response to ischaemia, in addition to variation in time of sampling
after stroke. We were unable to explore all these factors in the
present analysis. Our data suggest that within the first 26 h there
is unlikely to be a simple relationship between higher temperature,
increased energy metabolism, worse stroke and hence worse
outcome. Instead, early brain temperature elevation after stroke
and pyrexia probably represent responses to different stimuli
occurring after stroke. The pathophysiological links between
metabolic reactions taking place in ‘tissue at risk’, their thermal
effects, disease severity and their ability or otherwise to limit
deterioration, need further investigation with detailed measurements of brain and body temperature to profile the sequence of
temperature changes with other factors like regulation of selected
genes, infection, systemic inflammatory response, and local blood
flow patterns in and around the infarct.
Acknowledgements
This study was conducted in the SFC Brain Imaging Research
Center (www.sbirc.ed.ac.uk), The University of Edinburgh.
Funding
United Kingdom Stroke Association (TSA 02/01); European Union
Marie Curie PhD program/studentship (to B.K.); The Foundation
for Polish Science (to B.K.); The International Brain Research
Organization (to B.K.); Dr James & Bozena Bain Memorial Trust
Fund (to B.K.); The Dlugolecka-Graham Trust Fund (to K.W.);
The Row Fogo Charitable Research Foundation and SFC eDIKT
project (to P.A.); The Scottish Chief Scientist Office.
References
Asai S, Zhao H, Kohno T, Takahashi Y, Nagata T, Ishikawa K.
Quantitative evaluation of extracellular glutamate concentration in
post-ischemic glutamate re-uptake, dependent on brain temperature,
Lactate and temperature in human stroke
in the rat following severe global brain ischemia. Brain Res 2000; 864:
60–8.
Axelrod YK, Diringer MN. Temperature management in acute neurologic
disorders. Crit Care Clin 2006; 22: 767–85.
Bamford J, Sandercock P, Dennis M, Burn J, Warlow C. Classification and
natural history of clinically identifiable subtypes of cerebral infarction.
Lancet 1991; 337: 1521–6.
Barber PA, Hoyte L, Colbourne F, Buchan AM. Temperature-regulated
model of focal ischemia in the mouse: a study with histopathological
and behavioral outcomes. Stroke 2004; 35: 1720–5.
Berger C, Schabitz WR, Georgiades D, Steiner T, Aschoff A, Schwab S.
Effects of hypothermia on excitatory amino acids and metabolism in
stroke patients. A microdialysis study. Stroke 2002; 33: 519–24.
Berger C, Schabitz WR, Wolf M, Mueller H, Sommer C, Schwab S.
Hypothermia and brain-derived neurotrophic factor reduce glutamate
synergistically in acute stroke. Exp Neurol 2004; 185: 305–12.
Bergersen LH. Is lactate food for neurons? Comparison of monocarboxylate transporter subtypes in brain and muscle. Neuroscience 2007;
145: 11–9.
Cady EB, D’Souza PC, Penrice J, Lorek A. The estimation of local brain
temperature by in vivo 1H magnetic resonance spectroscopy. Magn
Reson Med 1995; 33: 862–7.
Castillo J, Davalos A, Marrugat J, Noya M. Timing for fever-related brain
damage in acute ischemic stroke. Stroke 1998; 29: 2455–60.
Castillo J, DAvalos A, Noya M. Aggravation of acute ischaemic stroke by
hyperthermia is related to an excitotoxic mechanism. Cerebrovasc Dis
1999; 9: 22–7.
Corbett R, Laptook A, Weatherall P. Noninvasive measurements
of human brain temperature using volume-localized proton
magnetic resonance spectroscopy. J Cereb Blood Flow Metab 1997;
17: 363–9.
Corbett RJ, Laptook AR, Tollefsbol G, Kim B. Validation of a noninvasive
method to measure brain temperature in vivo using 1H NMR
spectroscopy. J Neurochem 1995; 64: 1224–30.
Corbett RJ, Purdy PD, Laptook AR, Chaney C, Garcia D. Noninvasive
measurement of brain temperature after stroke. Am J Neuroradiol
1999; 20: 1851–7.
De Georgia MA, Krieger DW, Abou-Chebl A, Devlin TG, Jauss M,
Davis SM, et al. Cooling for Acute Ischemic Brain Damage (COOL
AID): a feasibility trial of endovascular cooling. Neurology 2004; 63:
312–7.
Emsley HC, Tyrrell PJ. Inflammation and infection in clinical stroke.
J Cereb Blood Flow Metab 2002; 22: 1399–419.
Fillenz M. The role of lactate in brain metabolism. Neurochem Int 2005;
47: 413–7.
Frykholm P, Hillered L, Långström B, Persson L, Valtysson J, Enblad P.
Relationship between cerebral blood flow and oxygen metabolism, and
extracellular glucose and lactate concentrations during middle cerebral
artery occlusion and reperfusion: a microdialysis and positron emission
tomography study in nonhuman primates. J Neurosurg 2005; 102:
1076–84.
Germain D, Chevallier P, Laurent A, Saint-Jalmes H. MR monitoring
of tumour thermal therapy. Magn Reson Materials Phys Biol Med
(MAGMA) 2001; 13: 47–59.
Harris BA, Andrews PJD, Marshall I, Robinson TM, Murray GD. Forced
convective head cooling device reduces human cross-sectional brain
temperature measured by magnetic resonance: a non-randomized
healthy volunteer pilot study. Br J Anaesth 2008; 100: 365–72.
Hillered L, Vespa PM, Hovda DA. Translational neurochemical research in
acute human brain injury: the current status and potential future for
cerebral microdialysis. J Neurotrauma 2005; 22: 3–41.
Horvath TL, Diano S, Barnstable C. Mitochondrial uncoupling protein 2
in the central nervous system: neuromodulator and neuroprotector.
Biochem Pharmacol 2003; 65: 1917–21.
Ichord RN, Northington FJ, van Wylen D, Johnston MV, Kwon C,
Traystman RJ. Brain O2 consumption and glutamate release during
hypoglycemic coma in piglets are temperature sensitive. Am J
Physiol 1999; 276: H2053–62.
Brain 2009: 132; 955–964
| 963
Ishihara Y, Calderon A, Watanabe H, Okamoto K, Suzuki Y, Kuroda K,
et al. A precise and fast temperature mapping using water proton
chemical shift. Magn Reson Med 1995; 34: 814–23.
Iwata O, Thornton JS, Sellwood MW, Iwata S, Sakata Y, Noone MA,
et al. Depth of delayed cooling alters neuroprotection pattern after
hypoxia-ischemia. Ann Neurol 2005; 58: 75–87.
Karaszewski B, Wardlaw JM, Marshall I, Cvoro V, Wartolowska K,
Haga K, et al. Measurement of brain temperature with magnetic
resonance spectroscopy in acute ischemic stroke. Ann Neurol 2006;
60: 438–46.
Kim-Han JS, Dugan LL. Mitochondrial uncoupling proteins in the central
nervous system. Antioxid Redox Signal 2005; 7: 1173–81.
Krieger DW, Yenari MA. Therapeutic hypothermia for acute ischemic
stroke: what do laboratory studies teach us? Stroke 2004; 35: 1482–9.
Kuroda K, Takei N, Mulkern RV, Oshio K, Nakai T, Okada T, et al.
Feasibility of internally referenced brain temperature imaging with a
metabolite signal. Magn Reson Med Sci 2003; 2: 17–22.
LaManna JC, McCracken KA, Patil M, Prohaska O. Brain tissue temperature: activation-induced changes determined with a new multisensor
probe. Adv Exp Med Biol 1988; 222: 383–9.
Macrı̀ MA, D’Alessandro N, Di Giulio C, Di Iorio P, Di Luzio S, Giuliani P,
et al. Regional changes in the metabolite profile after long-term
hypoxia-ischemia in brains of young and aged rats: a quantitative
proton MRS study. Neurobiol Aging 2006; 27: 98–104.
Marshall I, Wardlaw JM, Graham C, Murray L, Blane A. Repeatability of
long and short echo-time in vivo Proton Chemical Shift Imaging.
Neuroradiology 2002; 44: 973–80.
Marshall I, Karaszewski B, Wardlaw JM, Cvoro V, Wartolowska K,
Armitage PA, et al. Measurement of regional brain temperature
using proton spectroscopic imaging: validation and application to
acute ischemic stroke. Magn Reson Imaging 2006; 24: 699–706.
Martin M, Labouesse J, Canioni P, Merle M. N-acetyl-L-aspartate and
acetate 1H NMR signal overlapping under mild acidic pH conditions.
Magn Reson Med 1993; 29: 692–4.
Mattiasson G, Shamloo M, Gido G, Mathi K, Tomasevic G, Yi S, et al.
Uncoupling protein-2 prevents neuronal death and diminishes brain
dysfunction after stroke and brain trauma. Nat Med 2003; 9: 1062–8.
McDannold N, King RL, Hynynen K. MRI monitoring of heating
produced by ultrasound absorption in the skull: in vivo study in pigs.
Magn Reson Med 2004; 51: 1061–5.
Meylaerts SA, Kalkman CJ, de Haan P, Porsius M, Jacobs MJ. Epidural
versus subdural spinal cord cooling: cerebrospinal fluid temperature
and pressure changes. Ann Thorac Surg 2000; 70: 222–8.
Muñoz Maniega S, Cvoro V, Armitage PA, Marshall I, Bastin ME,
Wardlaw JM. Choline and creatine are not reliable denominators for
calculating metabolite ratios in acute ischaemic stroke. Stroke 2008;
39: 2467–9.
Nakase T, Yoshida Y, Nagata K. Amplified expression of uncoupling
proteins in human brain ischemic lesions. Neuropathology 2007; 27:
442–7.
Nelson DA, Nunneley SA. Brain temperature and limits on transcranial
cooling in humans: quantitative modelling results. Eur J Appl Physiol
1998; 78: 353–9.
Nemoto EM, Jungreis C, Larnard D, Kuwabara H, Horowitz M,
Kassam A. Hyperthermia and hypermetabolism in focal cerebral
ischemia. Adv Exp Med Biol 2005; 566: 83–9.
Parry-Jones AR, Liimatainen T, Kauppinen RA, Gröhn OH, Rothwell NJ.
Interleukin-1 exacerbates focal cerebral ischemia and reduces ischemic
brain temperature in the rat. Magn Reson Med 2008; 59: 1239–49.
Parsons MW, Li T, Barber PA, Yang Q, Darby DG, Desmond PM, et al.
Combined (1)H MR spectroscopy and diffusion-weighted MRI improves
the prediction of stroke outcome. Neurology 2000; 55: 498–505.
Perneger TV. What’s wrong with Bonferroni adjustments. BMJ 1998;
316: 1236–8.
Reith J, Jorgensen HS, Pedersen PM, Nakayama H, Raaschou HO,
Jeppesen LL, et al. Body temperature in acute stroke: relation to
stroke severity, infarct size, mortality, and outcome. Lancet 1996;
347: 422–5.
964
| Brain 2009: 132; 955–964
Ren Y, Hashimoto M, Pulsinelli WA, Nowak TS Jr. Hypothermic
protection in rat focal ischemia models: strain differences and
relevance to ‘reperfusion injury’. J Cereb Blood Flow Metab 2004;
24: 42–53.
Schurr A. Lactate, glucose and energy metabolism in the ischemic brain
(Review). Int J Mol Med 2002; 10: 131–6.
Schurr A, Payne RS. Lactate, not pyruvate, is neuronal aerobic glycolysis
end product: an in vitro electrophysiological study. Neuroscience 2007;
147: 613–9.
Schwab S, Spranger M, Aschoff A, Steiner T, Hacke W. Brain
temperature monitoring and modulation in patients with severe
MCA infarction. Neurology 1997; 48: 762–7.
Schwab S, Schwarz S, Aschoff A, Keller E, Hacke W. Moderate hypothermia and brain temperature in patients with severe middle cerebral
artery infarction. Acta Neurochir Suppl 1998; 71: 131–4.
Stillman AE, Latchaw RE. On cerebral lactate production and blood flow
in acute stroke. J Magn Reson Imaging 1993; 3: 682–3.
Stone JG, Goodman RR, Baker KZ, Baker CJ, Solomon RA. Direct
intraoperative measurement of human brain temperature.
Neurosurgery 1997; 41: 20–4.
B. Karaszewski et al.
Trubel HK, Maciejewski PK, Farber JH, Hyder F. Brain temperature
measured by 1H-NMR in conjunction with a lanthanide complex.
J Appl Physiol 2003; 94: 1641–9.
van den Boogaart A, van Ormondt D, Pijnappel WWF, de Beer R, AlaKorpela M. Removal of the water resonance from 1H magnetic
resonance spectra. In: McWhirter JG, editor. Mathematics in signal
processing. 3rd edn., Oxford: Clarendon Press; 1994. p. 175–95.
Wardlaw JM, Marshall I, Wild J, Dennis MS, Cannon J, Lewis SC. Studies
of acute ischemic stroke with proton magnetic resonance
spectroscopy. Relation between time from onset, neurological deficit,
metabolite abnormalities in the infarct, blood flow, and clinical
outcome. Stroke 1998; 29: 1618–24.
Yenari MA, Han HS. Influence of hypothermia on post-ischemic
inflammation: role of nuclear factor kappa B (NFkappaB).
Neurochem Int 2006; 49: 164–9.
Zaremba J. Hyperthermia in ischemic stroke. Med Sci Monit 2004; 10:
RA148–53.
Zhang Y, Wong KC, Zhang Z. The effect of intraischemic mild hypothermia on focal cerebral ischemia/reperfusion injury. Acta Anaesthesiol
Sin 2001; 39: 65–9.