Brain (2002), 125, 2446±2459
TNF-a reduces cerebral blood volume and disrupts
tissue homeostasis via an endothelin- and TNFR2dependent pathway
Nicola R. Sibson,1 Andrew M. Blamire,1 V. Hugh Perry,2 Jack Gauldie,4 Peter Styles1 and
Daniel C. Anthony3
1MRC
Biochemical and Clinical Magnetic Resonance Unit,
Department of Biochemistry, University of Oxford, Oxford,
2CNS In¯ammation Group, 3Molecular Neuropathology
Laboratory, School of Biological Sciences, University of
Southampton, Southampton, UK, 4Department of Pathology
and Molecular Medicine and Centre for Gene Therapeutics,
McMaster University, Hamilton, Ontario, Canada
Summary
TNF-a expression is elevated in a variety of neuropathologies, including multiple sclerosis, cerebral malaria and HIV encephalitis. However, the consequences
of such high cerebral TNF-a expression remain unresolved. Here, using MRI, we demonstrate that a focal
intrastriatal injection of TNF-a causes a signi®cant,
acute reduction (15±30%) in cerebral blood volume
(CBV), which is dependent on TNF-a-type 2 receptor
(TNFR2) activation, and can be ameliorated by pretreatment with a non-speci®c endothelin (ET) receptor
antagonist. An acute breakdown of the blood-cerebrospinal ¯uid barrier (B-CSF-B) and a delayed breakdown of the blood-brain barrier (BBB) were also
observed using contrast-enhanced MRI. Furthermore, a
Correspondence to: N. R. Sibson, MRC Biochemical and
Clinical Magnetic Resonance Unit, Department of
Biochemistry, University of Oxford, South Parks Road,
Oxford, OX1 3QU, UK
E-mail: [email protected]
signi®cant reduction in tissue water diffusion was
apparent 24 h after intrastriatal injection of TNF-a
injection, which may indicate compromise of tissue
energy metabolism. Prolonged expression of endogenous
TNF-a, achieved through the use of an adenoviral
vector expressing TNF-a cDNA (Ad5TNF-am), caused
a sustained depression in CBV in accordance with the
single TNF-a bolus data. These ®ndings identify vasoconstriction, disrupted tissue homeostasis and damage
to the BBBs as adverse effects of TNF-a within the
brain, and suggest that antagonists of the endothelin
and TNF-a type 2 receptors may be therapeutic in
TNF-a-associated neuropathologies.
Keywords: brain; cytokines; magnetic resonance; rats; vasoconstriction
Abbreviations: ADC = apparent diffusion coef®cient; BBB = blood brain barrier; B-CSF-B = blood cerebrospinal ¯uid
barrier; CBV = cerebral blood volume; ET = endothelin; Gd = gadolinium-based contrast agent; HRP = horseradish
peroxidase; rCBV = regional cerebral blood volume; rrTNF-a = recombinant rat TNF-a; rhuTNF-a = recombinant human
TNF-a; ROI = region of interest; TNFR1 = TNF-a type 1 receptor; TNFR2 = TNF-a type 2 receptor
Introduction
Expression of the proin¯ammatory cytokine tumour necrosis
factor alpha (TNF-a) is associated with the pathology of a
broad spectrum of CNS disease and injury. However, the
consequences of TNF-a expression, whether detrimental or
protective, remain a focus for debate.
TNF-a has been quanti®ed in post-mortem tissue from the
brains of patients with both acute and chronic neuropathologies, such as cerebral malaria (Brown et al., 1999) and
HIV-1 (Achim et al., 1993; Nuovo et al., 1994), indicating
local production of the cytokine. TNF-a expression has also
ã Guarantors of Brain 2002
been detected in post-mortem brain tissue from patients with
bacterial meningitis (Waage et al., 1989; Brown et al., 1999),
a condition in which intrathecal levels of TNF-a correlate
positively with the degree of blood-brain barrier (BBB)
breakdown, disease severity and indices of meningeal
in¯ammation (Sharief et al., 1992). Furthermore, TNF-a
expression is associated with demyelinating multiple sclerosis lesions (Woodroofe and Cuzner, 1993), and the presence
of TNF-a in cerebrospinal ¯uid from multiple sclerosis
patients correlates with disease activity (Hauser et al., 1990).
TNF-a alters CBV and tissue homeostasis
Thus, the accumulated evidence suggests a role for TNF-a in
the pathophysiology of a variety of CNS diseases, although
the mechanisms by which this cytokine contributes to disease
or injury severity remain unresolved.
Following both stroke and trauma, the in¯ammatory
response has been shown to contribute to secondary injury
and increased lesion volume (Clark et al., 1994; Matsuo et al.,
1994). However, although TNF-a is the archetypal proin¯ammatory cytokine, it can be both neurotoxic and
neuroprotective in models of cerebral ischaemia and head
injury (Bruce et al., 1996; Dawson et al., 1996; Shohami et al.,
1996, 1999; Barone et al., 1997; Nawashiro et al., 1997;
Lavine et al., 1998; Scherbel et al., 1999). It has been
suggested that in the early stages of injury, over-expression of
TNF-a is deleterious, while at later time points it may aid
recovery of injured tissue (Scherbel et al., 1999; Shohami
et al., 1999). Recently, Arnett and colleagues reported that
TNF-a activity contributes to remyelination in the cuprizone
model (Arnett et al., 2001). In addition, Gourin and
Shackford (1997) reported elevated TNF-a levels in cerebral
microvascular endothelium isolated from head-injured
patients, suggesting possible cerebrovascular effects of this
cytokine.
MRI is used clinically for the evaluation of many
neuropathologies in which in¯ammation is implicated.
Conventional MRI provides a sensitive measure of tissue
structure and water content and, together with intravenous
contrast agents, can measure BBB permeability and cerebral
perfusion. In addition, diffusion-weighted imaging has demonstrated sensitivity to reversible and irreversible alterations
in cellular homeostasis that are undetectable histologically,
notably in acute ischaemia and spreading depression (HoehnBerlage, 1995). Owing to the non-invasive nature of MRI,
these techniques are ideally suited to the temporal evaluation
of brain disease, and are particularly of value in understanding the contribution of cytokines to CNS injury and disease
in vivo (Blamire et al., 2000).
In this study, MRI techniques were used to investigate the
effects of a focal striatal injection of recombinant rat TNF-a
(rrTNF-a) on cerebral blood volume (CBV), on BBB and
blood-cerebrospinal ¯uid barrier (B-CSF-B) viability, and on
tissue water diffusion in the rat brain. The mechanism and
pathway underlying the observed changes in CBV were
investigated in two further experiments: (i) pre-treatment
with an endothelin (ET) receptor antagonist prior to
intrastriatal rrTNF-a; and (ii) intrastriatal injection of
recombinant human TNF-a (rhuTNF-a), which only binds
to the TNF-a type 1 receptor (TNFR1) in rat brain. In
addition, the effects of prolonged endogenous TNF-a
expression on CBV were determined by intrastriatal injection
of a replication-de®cient recombinant adenovirus expressing
membrane-bound TNF-a cDNA (Ad5TNF-am), which
causes local, sustained expression of TNF-a (Marr et al.,
1997). The results of these studies demonstrate diverse
actions of TNF-a in the brain and provide a mechanistic basis
by which this cytokine may contribute to the pathogenesis of
2447
diseases associated with TNF-a expression within the brain
parenchyma.
Materials and methods
Animal preparation
Adult male Wistar rats (Harlan-Olac, UK), weighing 218 6
20 g, were anaesthetized with fentanyl/¯uanisone and
midazolam (0.68 ml/kg of each). Using a <50 mm-tipped
glass pipette, 1 ml of TNF-a (NIBSC, Potters Bar, UK)
solution was injected stereotaxically over a 10-min period, 1
mm anterior and 3 mm lateral to bregma, at a depth of 4 mm
into the left striatum. Animals were injected with 0.3 mg/ml or
1.5 mg/ml of either recombinant rat TNF-a (rrTNF-a) or
recombinant human TNF-a (rhuTNF-a), each in 0.1% BSA
(bovine serum albumin) in low-endotoxin saline, or with
vehicle solution only. In addition, two further groups of
animals were injected stereotaxically in the same way with
107 p.f.u. of a replication-de®cient recombinant adenoviral
vector (Kolb et al., 2001) in 1 ml of saline: (i) adenovirus
expressing murine membrane-bound TNF-a cDNA
(Ad5TNF-am); and (ii) a control adenovirus with no insert
in the E1 region (AdDL70). Animals were positioned in the
MRI probe (3.4 cm internal diameter Alderman-Grant
resonator) using a bite-bar. During MRI, anaesthesia was
maintained with 0.8±1.2% halothane in 50% N2O/50% O2,
ECG was monitored non-invasively and body temperature
was maintained at ~37°C. All procedures were approved by
the United Kingdom Home Of®ce.
MRI studies
Magnetic resonance images were acquired using a 300 MHz
Varian Inova spectrometer (Varian, Palo Alto, CA).
Anatomical images were acquired using a T2-weighted
sequence [repetition time (TR) 3 s, echo time (TE) 80 ms].
Diffusion-weighted images were acquired using a pulsedgradient spin-echo sequence (TR 1.0 s, TE 40 ms), with
diffusion weighting values of 125, 750 and 1500 s/mm2, a
diffusion time of 20 ms and a diffusion gradient duration of
10 ms. Diffusion gradients were applied separately along
three orthogonal axes and apparent diffusion coef®cient
(ADC) `trace' maps were calculated (Basser et al., 1994).
Navigator echoes were used for motion correction (Ordidge
et al., 1994). Regional CBV (rCBV) maps were generated
from time-series images acquired during bolus injection of
contrast agent and tracer kinetic analysis (Rosen et al., 1990).
A series of 40 FLASH images (Frahm et al., 1986) were
acquired at a rate of 1 image per second, while 100 ml of a
gadolinium-based contrast agent (Gd) (Omniscan, Nycomed
Amersham, UK) was injected via a tail vein, over a 4 s period,
from image 8. Spin-echo T1-weighted images (TR 500 ms,
TE 20 ms) were acquired both pre-Gd injection and 10 min
post-Gd injection to look for image enhancement owing to
BBB/B-CSF-B permeability. Slice thickness was 1 mm for
2448
N. R. Sibson et al.
coronal images and 2 mm for horizontal images, except for
the rCBV data sets, which were all 1 mm. The matrix size and
®eld of view were 128 3 128 and 3 3 3 cm, respectively, for
all images, except for the rCBV data, which were acquired
with a 128 3 64 matrix and a 3 3 2 cm ®eld of view. All
regions of interest (ROI) and statistical analyses were
performed on images obtained in the horizontal plane (at
the level of the injection site) at the time-points stated,
although coronal plane data was also acquired at intermediate
times for qualitative purposes.
Experimental protocol
Six studies were performed to investigate different aspects of
the brain response to TNF-a.
Acute effects of a TNF-a single bolus injection
on rCBV and B-CSF-B/BBB viability
Three groups of animals were used and injected with: (i)
vehicle only (n = 4); (ii) 0.3 mg rrTNF-a (n = 6); or (iii) 1.5 mg
rrTNF-a (n = 5). Pre-Gd T1-weighted images, rCBV data and
post-Gd T1-weighted images were acquired at 1.5, 2.5, 3.5
and 5.5 h. One animal from each of the rrTNF-a-injected
groups was sacri®ced at 5.5 h for histological and immunocytochemical analysis. All other animals were recovered
from anaesthesia at the end of the MRI protocol, except for
two which died for reasons unrelated to the rrTNF-a (0.3 mg)
injection.
Acute effects of a TNF-a single bolus injection
on tissue water diffusion
Two groups of animals were used and were injected with: (i)
vehicle only (n = 4); or (ii) 0.3 mg rrTNF-a (n = 7). Diffusionand T2-weighted images were acquired each hour (1.5±6.5 h)
after TNF-a injection. All animals were recovered from
anaesthesia at the end of the MRI protocol, except for two,
which died for reasons unrelated to the rrTNF-a injection.
Chronic effects of a TNF-a single bolus injection
In studies (a) and (b), animals that were recovered from
anaesthesia after the ®nal acquisition were re-imaged using
all MRI protocols at either 24 h (control, n = 4; 0.3 mg
rrTNF-a, n = 8; 1.5 mg rrTNF-a, n = 4) or 72 h (control, n = 3;
0.3 mg rrTNF-a, n = 6) after stereotaxic injection.
The groups injected with 0.3 mg rrTNF-a include
additional animals that were imaged at either 24 h or 72 h
only, to replace animals that died or were used for histology at
earlier time points.
Effect of an ET receptor antagonist on acute
CBV changes
Two groups of animals were used, and were treated as
follows: (i) intravenous injection of the non-speci®c ET
receptor antagonist Ro 46-2005 (1 mg in 0.25 ml sterile
water) 10 min before injection of 1.5 mg rrTNF-a (n = 6); and
(ii) intravenous injection of sterile water (0.25 ml) 10 min
before injection of 1.5 mg rrTNF-a (n = 4). Pre-Gd T1weighted images, rCBV data and post-Gd T1-weighted
images were acquired at 1.5 h only. An additional group of
animals treated as for group (i) were imaged using all MRI
protocols at 24 h (n = 4). The dose of Ro 46-2005 used in this
study has been found to be an effective therapy in other
models of acute CNS injury and disease (Clozel et al., 1993).
Effects of TNF-a type I receptor activation alone
on acute CBV changes
Two groups of animals were used, and were injected with: (i)
0.3 mg rhuTNF-a (n = 5); or (ii) 1.5 mg rhuTNF-a (n = 5). PreGd T1-weighted images, rCBV data and post-Gd T1-weighted
images were acquired at 1.5 h only. All animals were
recovered from anaesthesia and subsequently re-imaged at
24 h with all MRI protocols.
Effects of prolonged endogenous TNF-a
expression on CBV
Animals were studied using MRI at either 24 h or 48 h after
adenovirus injection: (i) Ad5TNF-am (24 h, n = 4; 72 h, n =
3); or (ii) AdDL70 (n = 3 per time point). Pre-Gd T1-weighted
images, rCBV data and post-Gd T1-weighted images were
acquired for each animal.
Following end-point MRI, the brains of all animals were
perfusion-®xed for histological and immunocytochemical
analysis. To quantitate leukocyte recruitment and horseradish
peroxidase (HRP)-detectable changes in BBB/B-CSF-B permeability, additional animals were (i) injected with 0.3 mg
rrTNF-a and perfusion-®xed at 4 h (n = 3), or (ii) injected
with 1.5 mg rrTNF-a and perfusion-®xed at either 2.5 h (n =
2) or 5.5 h (n = 2). Additional animals were injected with
adenovirus (Ad5TNF-am, n = 6; or AdDL70, n = 6), and
perfusion-®xed at either 5 or 14 days to look for neuronal loss.
Histological analysis
At the end of each protocol, the rats were deeply
anaesthetized with sodium pentobarbitone, and transcardially
perfused with 200 ml 0.9% heparinized saline followed by
200 ml periodate lysine paraformaldehyde. After dissection,
the brains were post-®xed for several hours in the ®xative and
then immersed in 30% sucrose buffer for 24 h to cryoprotect
them. The tissue was then embedded in Tissue Tek (Miles
Inc, Elkhart, IN, USA) and frozen in isopentane at ±40°C.
Cresyl violet-stained sections (50 mm) were examined for
TNF-a alters CBV and tissue homeostasis
2449
Fig. 1 rCBV maps acquired after a focal striatal injection of TNF-a. (A) rCBV map calculated from perfusion time-series MRI data
acquired 1.5 h after injection of rrTNF-a (1.5 mg). The ROIs used for quantitative data analysis are indicated for each striatum. An area of
reduced CBV can be seen in the injected (left) striatum. (B) rCBV map calculated from perfusion time-series MRI data acquired 5.5 h
after injection of rrTNF-a (1.5 mg), demonstrating that CBV in the injected (left) striatum has mostly returned to normal by this time
point. (C) T2-weighted image acquired 1.5 h after injection of rrTNF-a (1.5 mg). The small area of increased signal intensity in the left
striatum indicates the site of intrastriatal injection. The ROIs used for quantitative data analysis are indicated for each striatum. All
non-CNS tissue has been excluded from the rCBV maps.
signs of neuronal damage and for the presence of leukocytes.
Immunohistochemistry was used to con®rm the presence and
distribution of speci®c cell populations. Frozen, 10-mm thick
serial sections were cut from the ®xed tissue and mounted on
gelatin-coated glass slides. Antigens were detected using a
three-step indirect method (Hsu et al., 1981). Polymorphonuclear neutrophils were identi®ed using the antineutrophil serum HB199 (Anthony et al., 1998) and
macrophages were identi®ed using the monoclonal antibody
ED1 (Dijkstra et al., 1985), which stains most macrophage
populations including recruited monocytes, but not quiescent
microglia. In each group of animals, leukocyte in®ltration
was quanti®ed by counting the number of ED1-positive or
HB199-positive cells in 10-mm thick cresyl violet counterstained sections from regions immediately adjacent to the
injection site. Three non-overlapping ®elds containing the
highest density of recruited cells within the parenchyma were
chosen and the number of neutrophils or macrophages was
calculated as an average number per mm2 for each animal.
Neurones were identi®ed using anti-NeuN monoclonal antibody (Chemicon, Europe, Ltd, Chandlers Ford, Hampshire,
UK) and 25-mm thick frozen sections were processed for
cholinesterase histochemistry using acetylthiocholine as
substrate.
Some animals were injected intravenously with type-II
HRP, an indicator of increased BBB permeability (Sigma
Chemical Co., St Louis, MO, USA), 104 U/kg, 30 min before
perfusion±®xation. Coronal, free-¯oating sections were cut
for HRP localization by a modi®ed Hanker±Yates method
(Perry and Linden, 1982). HRP has been used extensively as a
tracer of altered vessel permeability, and does not increase
endothelial transport under normal conditions (Claudio et al.,
1990).
MRI data analysis
ROIs encompassing the striatum were de®ned on T2weighted images in each hemisphere (see Fig. 1C), and
applied to all images or calculated data maps for quantitation.
During the acute time courses, animals remained in the
magnet throughout, enabling the same ROI to be used for all
images acquired over this period. At the later time points,
both the imaging slice position and the ROIs were matched to
those used at earlier time points. For the rCBV maps and T2weighted images, the data are expressed as a ratio of injected/
non-injected striatal values. The signal intensity was calculated for the injected striatum in the pre-Gd and post-Gd
images, and the percentage increase in signal in the post-Gd
images calculated (normalized for the changes in signal
intensity of the contralateral striatum between pre- and postGd images). All values are mean 6 standard deviation (SD).
Since data were not acquired at every time point from all
animals over the acute time course (for technical reasons), a
mixed-effect model followed by pair-wise t-tests (Vonesh
and Chinchilli, 1997) was used to determine any statistical
differences between the rCBV time courses for each group.
Unpaired or paired t-tests were used to determine signi®cant
differences at 24 and 72 h for all MRI parameters.
Results
The minimally invasive technique used to focally microinject
TNF-a into the brain parenchyma resulted in no visible
damage per se, as demonstrated by vehicle-injected animals
in which no leukocyte recruitment or damage to the brain
parenchyma was found at either 24 h or 72 h after injection.
Previous studies have demonstrated that there is no leukocyte
recruitment at earlier times in vehicle-injected animals
2450
N. R. Sibson et al.
Fig. 2 Time course of injected/non-injected striatal rCBV ratios. Graph showing effect of a focal striatal
injection of either rrTNF-a or vehicle on rCBV. Values are expressed as ratios of rCBV obtained from
ROIs in the treated (left) striatum versus the untreated (right) striatum. Data are presented for three
groups of animals: (i) control, intrastriatal injection of vehicle only (black bars); (ii) intrastriatal injection
of 0.3 mg rrTNF-a (grey bars); and (iii) intrastriatal injection of 1.5 mg rrTNF-a (white bars). Values
close to 1.0 indicate no change in striatal rCBV, as seen in control animals. All values are represented as
mean 6 SD. Asterisks indicate a signi®cant difference between control and treated groups: *P < 0.05,
**P < 0.02, ***P < 0.005, ****P < 0.002, *****P < 0.001.
(Anthony et al., 1997). On T2-weighted scout images at 1 h,
the injection bolus was visible as a small hyperintense area in
the left striatum in all animals (Fig. 1C), and subsequent scans
were positioned directly through the injection site. Owing to
the focal nature of the cytokine injection into the left striatum,
inter-hemisphere comparisons were possible between injected
and non-injected striatal values.
Acute effects of a TNF-a single bolus injection
on CBV
An acute reduction in rCBV was observed in the injected
striatum at 1.5 h after rrTNF-a injection into the brain
(Fig. 1A), which gradually returned to normal by ~5.5 h
(Fig. 1B). No differences in rCBV between the two
hemispheres were visible in vehicle-injected animals at any
time point. Local changes in CBV were assessed by
calculating the ratio of rCBV within an ROI in the injected
striatum to that in a matched area in the non-injected striatum
of the same animal (Fig. 1). In animals injected with either 0.3
or 1.5 mg of rrTNF-a, the ratio of injected/non-injected
striatal rCBV was signi®cantly reduced compared with the
vehicle-injected group at 1.5 h (unpaired t-tests: P < 0.02, 0.3
mg rrTNF-a; P < 0.0002, 1.5 mg rrTNF-a) (Fig. 2). This
reduction in rCBV was dose-dependent, with a greater
reduction at the higher dose (~25%) than at the lower dose
(~14%). The rCBV changes occurred prior to leukocyte
recruitment to the brain parenchyma, which was ®rst evident
4 h after the injection of rrTNF-a. At 4±5.5 h after rrTNF-a
injection, a small number of recruited monocytes (ED1-
stained cells) were visible in cuffs around the vessels
penetrating the cortex (50.8 6 5.0 mm±2 with 0.3 mg
rrTNF-a, n = 3; 59.2 6 19.4 mm±2 with 1.5 mg rrTNF-a,
n = 3).
The reduction in rCBV at 1.5 h in the injected striatum was
eliminated by intravenous injection of the ET receptor
antagonist Ro 46-2005 (5 mg/kg) 10 min prior to intracerebral
rrTNF-a injection (Fig. 3). In control animals injected
intravenously with the vehicle 10 min prior to intracerebral
rrTNF-a (1.5 mg), the reduction in striatal rCBV was still
evident, and comparable to that in the initial group of animals
injected with 1.5 mg rrTNF-a (Fig. 3). The difference
between injected/non-injected striatal rCBV ratios for the
two groups receiving an intravenous injection (either Ro 462005 or vehicle) prior to intracerebral rrTNF-a injection was
highly signi®cant (unpaired t-test, P < 0.005).
Animals injected with either 0.3 or 1.5 mg rhuTNF-a
exhibited no reduction in rCBV at 1.5 h (Fig. 3). There was no
signi®cant difference in the injected/non-injected striatal
rCBV ratios between either of these two groups and the
vehicle-injected group, indicating that activation of the rat
TNFR1 alone by rhuTNF-a does not result in a reduction in
rCBV.
Acute effects of a TNF-a single bolus injection
on B-CSF-B and BBB integrity
From as early as 1.5 h after injection of 1.5 mg rrTNF-a (2±
3 h with 0.3 mg rrTNF-a), enhancement of the ipsilateral
meninges was evident on T1-weighted images acquired 10
TNF-a alters CBV and tissue homeostasis
2451
Fig. 3 Effects of an ET receptor antagonist and activation of TNFR1 alone on striatal rCBV ratios. Graph showing (i) the effect of an ET
receptor antagonist (Ro 46-2005) on the rrTNF-a-induced rCBV changes 1.5 h after intrastriatal injection, and (ii) the effect of activation
of TNFR1 alone, by intrastriatal injection of rhuTNF-a, on rCBV 1.5 h after intrastriatal injection. Values are expressed as ratios of rCBV
obtained from ROIs in the treated (left) striatum versus the untreated (right) striatum. Values close to 1.0 indicate no change in striatal
rCBV, and all values are represented as mean 6 SD. ***P < 0.005, unpaired t-test. No signi®cant differences were found between: (i) the
control and 1.5 mg rrTNF-a + Ro 46-2005 groups; (ii) the 1.5 mg rrTNF-a and 1.5 mg rrTNF-a + H2O groups; or (iii) the control, 0.3 mg
rhuTNF-a and 1.5 mg rhuTNF-a groups.
min after intravenous administration of a contrast agent.
When the BBB and B-CSF-B are intact, the intravascular
contrast agent cannot cross these barriers, and, consequently,
no signal changes are seen on post-contrast T1-weighted
images. Thus, enhancement in post-contrast T1-weighted
images, as observed here, indicates breakdown of the B-CSFB, and was ®rst seen in the meninges overlying the parietal
cortex. The breakdown of the B-CSF-B was not detected by
histochemical localization of the tracer HRP at this time
point, and preceded recruitment of any leukocytes to the
meninges. Pre-treatment with Ro 46-2005 did not signi®cantly alter the effect of rrTNF-a on the B-CSF-B at 1.5 h.
Intrastriatal injection of rhuTNF-a yielded a similar degree of
meningeal enhancement at 1.5 h to that seen with rrTNF-a.
Over subsequent hours, the B-CSF-B breakdown spread to
encompass meningeal layers surrounding the frontal cortex
(Fig. 4B and C). Marked mononuclear cell recruitment to the
meninges occurred from ~4 h, and by 5.5 h the B-CSF-B
breakdown was just visible histologically using HRP
(Fig. 4D). In some cases, the MRI signal enhancement
appeared to have spread into the outermost cortical layers by
5.5 h, suggesting compromise of the pial and cortexpenetrating vessels. In the coronal plane, we observed
meningeal enhancement around the entire injected hemisphere (Fig. 4C), and this was often particularly clear around
the piriform cortex where upon histological examination we
found large numbers of monocytes (1390 6 444 mm±2 with
1.5 mg rrTNF-a, n = 3) (Fig. 4E). No breakdown of the
B-CSF-B was observed in vehicle-injected animals at any
time point.
Acute effects of a TNF-a single bolus injection
on tissue water diffusion
From 1 to 4 h, we observed small increases in tissue water
diffusion at the injection site, which corresponded spatially to
regions of T2 hyperintensity, in all animals imaged for T2 and
diffusion over the acute time course. This acute increase in
both T2 signal intensity (5±13%) and diffusion (6±8%)
re¯ects the small increase in extracellular water arising from
the injection bolus, and resolved as the ¯uid was cleared.
Chronic effects of a TNF-a single bolus
injection on tissue water diffusion, B-CSF-B/
BBB integrity and CBV
Although ELISA (enzyme-linked immunosorbent assay)
measurements show that all TNF-a has been cleared
from the brain parenchyma by 24 h after a bolus injection
(Anthony et al., 2001), the ADC of tissue water in the injected
striatum (ROI delineated as for rCBV) was signi®cantly
reduced compared with the non-injected striatum in both of
the rrTNF-a-injected groups (0.3 and 1.5 mg) at 24 h (paired
t-tests, P < 0.02) (Table 1). Despite the focal nature of the
2452
N. R. Sibson et al.
Fig. 4 Post-contrast images and histology demonstrating B-CSF-B breakdown and macrophage
recruitment. (A) Pre-contrast T1-weighted image acquired at 1.5 h after injection of rrTNF-a. (B) Post-Gd
T1-weighted image (horizontal plane) acquired 5.5 h after injection of rrTNF-a indicating a region of BCSF-B breakdown (hyperintensity) in the meninges overlying the frontoparietal cortex of the injected
hemisphere (arrow). (C) Post-contrast agent T1-weighted image (coronal plane) acquired 5.5 h after
injection of TNF-a, indicating a region of B-CSF-B breakdown around the entire ipsilateral meningeal
layer (arrows). (D) Horizontal brain section illustrating the changes in vessel permeability to HRP
following injections of 1.5 mg rrTNF-a at 5.5 h. Dark staining in the meninges indicates regions of
increased vessel permeability. HRP-positive macrophages can also be seen within the brain parenchyma.
(E) ED-1 stained macrophages (brown-stained cells) in the meninges around the piriform cortex at 5.5 h
after the injection of 1.5 mg rrTNF-a. Cresyl violet counterstained (scale bar = 20 mm). Note the
conspicuous macrophage recruitment to the meninges, which is less evident in the brain parenchyma.
cytokine injection, the reduction in ADC was not restricted to
the striatum and also encompassed surrounding cortical
regions (Fig. 5A). The ADC reduction observed in the rrTNFa-injected animals did not appear to be dose-dependent, with
similar reductions (7±13%) in both groups. There were no
signi®cant differences between the injected and non-injected
hemispheres in the control animals. The reduction in ADC
was not affected by pre-treatment with the ET-receptor
antagonist Ro 46-2005, with a signi®cant difference (paired ttest, P < 0.03) between the injected and non-injected striatal
values being evident (Table 1). Similarly, there was a
signi®cant difference between the injected and non-injected
striatal ADC values in the animals injected with 1.5 mg
rhuTNF-a (paired t-test, P < 0.02) (Table 1). However,
although a reduction in ADC was apparent in the injected
hemisphere in three out of ®ve animals injected with the
lower dose (0.3 mg) of rhuTNF-a, this did not reach
signi®cance (paired t-test, P = 0.136).
Breakdown of the B-CSF-B in all animals injected with
rrTNF-a and rhuTNF-a persisted to 24 h, when large
numbers of monocytes were present in the meninges. Lowlevel breakdown of the BBB in the brain parenchyma was
Maps of the ADC of tissue water were calculated from the diffusion-weighted images, and the ADC was measured within a ROI in each striatum (Left = treated, Right = control).
Values are represented as mean 6 SD for n = 4 (control groups), n = 8 (0.3 mg rrTNF-a, 24 h), n = 6 (0.3 mg rrTNF-a, 72 h), n = 4 (1.5 mg rrTNF-a, 24 h), n = 4 (Ro 46-2005 +
1.5 mg rrTNF-a, 24 h), n = 5 (0.3 mg rhuTNF-a, 24 h) and n = 5 (1.5 mg rhuTNF-a, 24 h). Signi®cant differences from control (right) striatum were determined by paired t-tests:
aP < 0.02, bP < 0.05.
6.78 6 0.35 6.79 6 0.35 6.28a 6 0.38 6.99 6 0.46 6.30a 6 0.52 6.90 6 0.35 6.72b 6 0.25 7.23 6 0.17 7.15 6 0.28 7.33 6 0.26 6.76a 6 0.20 7.14 6 0.31
6.59 6 0.25 6.50 6 0.22 6.82 6 0.42 6.55 6 0.36
24 h
72 h
Right
Left
Right
Left
Right
Left
Right
Left
Right
Left
Right
Left
Ro 46-2005
+1.5 mg rrTNF-a
1.5 mg rrTNF-a
0.3 mg rrTNF-a
Time Vehicle
Table 1 Apparent diffusion coef®cients (310±4 mm2/s) of tissue water in each striatum
0.3 mg rhuTNF-a
1.5 mg rhuTNF-a
TNF-a alters CBV and tissue homeostasis
2453
Fig. 5 MRI data acquired 24 h after focal striatal injection of
TNF-a. (A) Map of tissue ADC calculated from diffusionweighted images, showing a widespread area of reduced diffusion
in the injected striatum and surrounding cortical regions (left side
of image) 24 h after focal striatal injection of rrTNF-a (1.5 mg).
(B) T1-weighted image acquired 10 min after administration of an
intravenous contrast agent in the same animal 24 h after focal
striatal injection of rrTNF-a (1.5 mg). The region of hyperintensity
indicates BBB breakdown in regions corresponding to those
showing reduced tissue water diffusion. Breakdown of the B-CSFB is also still evident in the ipsilateral meninges.
also observed 24 h after rrTNF-a injection on post-contrast
T1-weighted images. This breakdown was more evident with
the higher dose (1.5 mg) of rrTNF-a (increase in signal
intensity of injected striatum post-Gd versus pre-Gd of 8.1 6
2.1%), but less apparent than that observed previously
following intrastriatal IL-1b injection (Blamire et al.,
2000). The pattern of BBB breakdown was similar to the
tissue water diffusion changes at 24 h, encompassing both
striatal and cortical regions (Fig. 5B). At this time point,
recruitment of monocytes into the brain parenchyma was
marked (189 6 7 per mm2 with 0.3 mg rrTNF-a, n = 3). Pretreatment with the ET-receptor antagonist Ro 46-2005 did not
signi®cantly affect the level of BBB breakdown observed in
animals injected with 1.5 mg TNF-a (increase in signal
intensity of injected striatum of 10.3 6 2.7%). However, in
animals injected with 1.5 mg rhuTNF-a, the degree of BBB
breakdown appeared to be substantially reduced (increase in
signal intensity of injected striatum = 5.2 6 1.6%), which
may be related to a lower level of monocyte recruitment (95
6 33 mm±2, n = 3) compared with that induced by rrTNF-a at
this time point.
Despite both diffusion and BBB changes in the TNF-ainjected groups, there were no signi®cant differences in either
rCBV or T2 signal intensity between the injected and noninjected hemispheres at 24 h in any animals.
Seventy-two hours after rrTNF-a injection, both the BBB
and B-CSF-B were intact, and no signi®cant differences were
found in the ADC of tissue water, T2 intensity or rCBV
between the hemispheres in any animals. However, the
number of ED-1-positive macrophages present within the
brain parenchyma was elevated further (361 6 79 mm±2 with
0.3 mg rrTNF-a, n = 3) at this time. There was no apparent
neuronal cell death at any time point following the single
2454
N. R. Sibson et al.
Fig. 6 Effect of sustained endogenous TNF-a expression on striatal rCBV ratios. Graph showing changes
in rCBV following injection of either AdDL70 (black bars) or Ad5TNF-am (grey bars). Values are
expressed as ratios of rCBV obtained from ROIs in the treated (left) striatum versus the untreated (right)
striatum. All values are represented as mean 6 SD. ***P < 0.005, unpaired t-tests.
bolus injections of rrTNF-a or rhuTNF-a, as evidenced by
cresyl violet staining.
Effects of prolonged endogenous TNF-a
expression on CBV
A progressive decrease in rCBV was observed in the injected
striatum of animals injected with Ad5TNF-am. The reduction
in rCBV reached ~20% by 48 h (Fig. 6). This decrease in
rCBV was highly signi®cant compared with animals injected
with a control adenovirus (AdDL70) at both 24 and 48 h
(unpaired t-tests, P < 0.005). At both 5 and 14 days after the
injection of Ad5TNF-am, unlike the single bolus injection,
there was evidence of neuronal cell death in the lesion as
indicated by a loss of NeuN antigen expression and a
reduction in acetylcholinesterase activity (Fig. 7). There was
no reduction in NeuN staining or loss of acetylcholinesterase
activity following injection of AdDL70.
Discussion
In this study we have shown that a focal, intrastriatal injection
of TNF-a in the rat brain results in: (i) an acute, dosedependent reduction in CBV that is mediated by endothelin,
and coupled to activation of the TNF-a type 2 receptor
(TNFR2) pathway; (ii) early breakdown of the blood-CSF
barrier and delayed breakdown of the blood-brain barrier;
and (iii) a delayed reduction in tissue water diffusion. At all
times leukocyte recruitment to the brain (parenchyma and
meninges) was restricted solely to mononuclear cells, as
reported previously (Anthony et al., 1997; Glabinski et al.,
1998). In addition, experiments using a replication-de®cient
recombinant adenovirus expressing membrane-bound TNF-a
cDNA con®rmed that prolonged expression of endogenous
TNF-a causes a sustained decrease in rCBV. These results
are in contrast to our previous ®ndings following intrastriatal
injection of IL-1b, which induced an increase, rather than a
decrease, in CBV and recruited only neutrophils to the brain
parenchyma (Blamire et al., 2000). In peripheral tissues, IL1b and TNF-a have been reported to have similar effects, and
it is surprising, therefore, that these cytokines have different
effects within the CNS. Despite these differences, both
cytokines result in a decrease in tissue water diffusion,
although this is delayed in TNF-a-injected animals compared
with IL-1b-injected animals. The implications of the current
®ndings are discussed below.
Effects of TNF-a on CBV
Our data demonstrate that there is a profound, acute reduction
in striatal rCBV as a direct consequence of focal rrTNF-a
injection. Few investigations of the effects of TNF-a on
cerebral perfusion have been reported previously, and where
data is available the results are somewhat contradictory.
Megyeri and colleagues demonstrated vasoconstriction in pial
arterioles following intracisternal injection of rhuTNF-a into
newborn piglets (Megyeri et al., 1992). In contrast, Brian and
Faraci (1998) demonstrated dilation of pial arterioles following superfusion of the rat cortex with TNF-a. Both of these
studies report the effects of TNF-a on the super®cial pial
arterioles of the brain, rather than the intraparenchymal
microvasculature. Similarly, large intracisternal injections of
TNF-a have been shown to decrease whole brain CBF in
rabbits (Tureen, 1995), but to increase cortical blood ¯ow in
rats (Angstwurm et al., 1998). Again, it is likely that in both
studies the effects of TNF-a were exerted on the super®cial,
rather than intraparenchymal vessels, and that the differences
re¯ect either species or dose differences. In rat models of
TNF-a alters CBV and tissue homeostasis
2455
Fig. 7 Immunohistochemistry at 5 or 14 days after adenovirus injection. (A) Neurones identi®ed by NeuN
immunohistochemistry (brown) in the right (non-injected) striatum 5 days after the injection of the
adenoviral vector Ad5TNF-am. (B) Absence of NeuN staining in the left (injected) striatum is visible 5
days after injection of the adenoviral vector Ad5TNF-am. Sections were cresyl violet counterstained
(scale bar = 20 mm). (C) Cresyl violet-stained section obtained 14 days after the injection of Ad5TNFam, demonstrating the hypercellular lesion. (D) Acetylcholinesterase activity in a serial section to C
reveals an area of decreased staining at the lesion site.
cerebral ischaemia, inhibition of endogenous TNF-a has been
shown to improve microvascular perfusion (Dawson et al.,
1996) and enhance cerebral blood ¯ow during reperfusion
(Lavine et al., 1998). On this basis, it has been suggested that
expression of TNF-a following focal cerebral ischaemia may
contribute to impairment of microvascular perfusion, either
as a consequence of recruited leukocytes obstructing cerebral
vessels or via a direct vasoconstrictor effect of the cytokine
itself (Dawson et al., 1996). Our data demonstrate clearly that
an intracerebral injection of rrTNF-a causes acute, temporary
vasoconstriction of local parenchymal vessels that is independent of recruited leukocytes.
Since the reduction in rCBV precedes monocyte recruitment, we hypothesized that this might occur via TNF-ainduced expression of endothelin peptides (ET-1 and ET-3),
which are known vasoconstrictors. Many pathologies associated with increased cytokine production also exhibit
elevated levels of circulating ET-1, and peripheral injection
of TNF-a into rats signi®cantly increases plasma ET-1
concentrations within 15 min (Klemm et al., 1995). Our data
demonstrate that the vasoconstrictor effects of rrTNF-a
within the brain parenchyma in vivo can be completely
eliminated by prior administration of an ET receptor antagonist which blocks both ETA and ETB receptors (Clozel et al.,
1993). We suggest, therefore, that the observed effects of
rrTNF-a on rCBV are mediated via the action of ET on its
receptors. ET-1 production by both bovine and human
cerebral endothelial cells in culture is increased by TNF-a
(Durieu-Trautmann et al., 1993; Skopal et al., 1998), and
vascular smooth muscle cells have also been shown to secrete
ET-1 in in¯ammatory lesions (Kedzierski and Yanagisawa,
2001). It is likely, therefore, that the observed reduction in
rCBV is caused by the action of TNF-a on the brain
microvessel endothelial and smooth muscle cells to provoke
the release of ET, which subsequently causes vasoconstriction, primarily through its action on the smooth muscle cell
ETA receptors (Kedzierski and Yanagisawa, 2001).
TNF-a binds to two transmembrane receptors of approximately 55 (p55, TNFR1) and 75 kDa (p75, TNFR2)
(Aggarwal and Natarajan, 1996). While the TNFR1 is
ubiquitously expressed, the TNFR2 is predominantly expressed by haematopoietic and endothelial cells, and they are
thought to activate distinct signalling pathways and mediate
distinct cellular processes. The rrTNF-a used in these studies
binds non-speci®cally to both TNF-a receptor subtypes,
whilst rhuTNF-a will bind only to TNFR1 in rat brain (Lewis
et al., 1991; Stefferl et al., 1996). In contrast to rrTNF-a
injection, intrastriatal injection of rhuTNF-a caused no
reduction in rCBV. These data indicate that activation of
the TNFR2, either alone or in combination with the TNFR1,
is essential for the TNF-a-induced reduction in rCBV, and
that activation of this pathway leads to ET release, probably
2456
N. R. Sibson et al.
via activation of the microvascular endothelial cells. These
data may explain the discrepancies in TNF-a-mediated
effects on cerebral perfusion reported previously, where
rhuTNF-a was used in rodent studies before rrTNF-a became
widely available.
In this study we have used a single bolus injection of TNFa into the striatum. Previously, we have demonstrated by
ELISA that following a bolus injection of TNF-a, the level of
immunoreactive TNF-a in the brain parenchyma has fallen to
50% of maximum after 4 h and is no longer quanti®able by 24
h (Anthony et al., 2001). As a result, effects of TNF-a are
likely to be short-lived under this experimental protocol in
comparison to neurological conditions, in which TNF-a may
be expressed chronically. Therefore, we used an adenovirusexpressing membrane-bound TNF-a cDNA to investigate the
effects of prolonged TNF-a expression within the brain on
rCBV. These experiments con®rmed the ®ndings of the single
bolus study and demonstrated that longer durations of
endogenous TNF-a expression cause a sustained reduction
in rCBV. In experimental stroke models, it is well documented that a reduction of 80±90% in cerebral perfusion of
short duration invariably results in energy failure and
neuronal death (Tamura et al., 1981; Bolander et al., 1989).
However, much less severe reductions in cerebral perfusion, if
prolonged, can also lead to neuronal death (Jacewicz et al.,
1992). Therefore, in neurological conditions where TNF-a
expression is prolonged, this may cause a long-term perfusion
de®cit that is detrimental to neuronal viability. Following the
injection of adenovirus, it is clear that long-term expression of
membrane-bound TNF-a can give rise to neuronal degeneration. However, at the present time it is unclear whether this is
due to the reduction in rCBV, a direct action of TNF-a on the
neurones, effects of molecules secreted from activated
in¯ammatory cells, or a combination of all of these factors.
Both cerebral malaria (Kwiatkowski et al., 1990; Brown
et al., 1999) and the Plasmodium berghei ANKA model of
cerebral malaria (Neill and Hunt, 1995) are associated with
high levels of cerebral TNF-a expression. Furthermore, a
signi®cant increase in the expression of TNFR2, but not
TNFR1, has been found on brain microvessels during
cerebral malaria in susceptible mice, and mice de®cient in
TNFR2 (but not those de®cient in TNFR1) are signi®cantly
protected from experimental cerebral malaria (Lucas et al.,
1997). Thus, the effects of TNF-a on rCBV, mediated via the
TNFR2 pathway and ET production, may be a contributing
factor to neuronal dysfunction or degeneration in cerebral
malaria, in which the causes of neuronal damage, and
ultimately patient death, are still unknown.
Effects of TNF-a on B-CSF-B and BBB
integrity
TNF-a is thought to play a role in BBB disruption associated
with brain injury (Feuerstein et al., 1994; Yang et al., 1999)
and bacterial meningitis (Sharief et al., 1992), and in vitro has
been shown to decrease the trans-endothelial resistance in
cerebrovascular-derived endothelial cells (de Vries et al.,
1996). However, few studies have considered variations in
BBB compromise between the different CNS compartments.
In the current study, the early unilateral increase in B-CSF-B
permeability (as distinct from BBB permeability) preceded
leukocyte recruitment to the brain. This effect on the B-CSFB may re¯ect direct actions of TNF-a on the vasculature, as
studies with tracers (N.R.S and D.C.A., unpublished data)
indicate that a bolus of ¯uid (as injected in this study) will
diffuse fairly rapidly out of the striatum and alongside the
major cortical vessels, to reach the meninges within 1.5±2 h.
Furthermore, the data indicate that the B-CSF-B breakdown
is leukocyte-independent, since it preceded macrophage
recruitment to the meninges and was no longer apparent at
72 h when recruited macrophages were numerous. Although
monocytes can cross an intact BBB and B-CSF-B, breakdown
of these barriers may facilitate presentation of chemokines
and thus recruitment to the meninges.
In contrast, breakdown of the BBB within the brain
parenchyma at 24 h was coincident with signi®cant macrophage recruitment to the parenchyma. This ®nding differs
from our previous studies of BBB viability using HRP, in
which only very minimal leakage of tracer, localized
speci®cally to the larger parenchymal vessels, was observed
24 h after a single bolus intraparenchymal injection of TNF-a
(Anthony et al., 1997). In addition, the B-CSF-B breakdown
observed at the early time points was visible by contrastenhanced MRI before it became detectable using HRP. The
current data suggest, therefore, that contrast-enhanced MRI
measurements offer a more sensitive method of detecting
BBB/B-CSF-B permeability than the HRP method, probably
owing to the considerably smaller molecular weight of the
gadolinium-based agent (0.57 kDa) compared with HRP (40
kDa).
Pre-treatment with the non-speci®c ET receptor antagonist
Ro 46-2005 had no effect on the changes in BBB and B-CSFB permeability, suggesting that these events are not mediated
by the TNF-a-induced ET pathway responsible for the rCBV
reduction. However, the ET system is widespread in the brain,
with ETA, ETB, ET-1 and ET-3 being expressed by vascular,
neuronal and glial cells (Kedzierski and Yanagisawa, 2001).
Given that the Ro 46-2005 injection was administered
intravenously, it is possible that it may not antagonize nonvascular effects of ET occurring deep within the brain
parenchyma, such as the observed BBB breakdown and ADC
reduction. Interestingly, there appeared to be a reduction in
the degree of BBB permeability at 24 h following rhuTNF-a
injection compared with the rrTNF-a-injected animals. These
data suggest that both receptor pathways contribute to the
processes underlying the BBB breakdown. Rosenberg and
colleagues have previously demonstrated a dose-dependent
parallel increase in capillary permeability and expression of
proteolytic enzymes 24 h after intracerebral infusion of TNFa, which could be blocked by an inhibitor of matrix
metalloproteinases (Rosenberg et al., 1995). On this basis,
TNF-a alters CBV and tissue homeostasis
they suggest that TNF-a may modulate delayed capillary
permeability via the matrix metalloproteinase gelatinase B.
Effects of TNF-a on tissue water diffusion
The areas of reduced tissue water ADC observed at 24 h
corresponded to the regions of BBB breakdown, and indicate
a relatively widespread effect of the focal cytokine injection.
Again, this is likely to result from spread of the injected bolus
to neighbouring cortical regions. Reduced tissue water
diffusion has been extensively documented in acute brain
ischaemia (Hoehn-Berlage, 1995), although the exact mechanisms responsible for these changes remain unclear. In
ischaemia, the temporal evolution of reduced diffusion
appears to follow the loss of high-energy metabolites (van
der Toorn et al., 1996) and is thought to re¯ect compromise of
tissue energy metabolism. Observations of a transient reduction in ADC during spreading depression (Latour et al., 1994)
indicate that changes in tissue water diffusion are linked to
disruption of tissue energy homeostasis, rather than ischaemia
per se. This hypothesis is supported by our previous ®nding
that IL-1b causes a reduction in ADC that is accompanied by
an increase in rCBV and no indicators of ischaemia (Blamire
et al., 2000). In the current study, rCBV was found to be
normal within the areas of reduced ADC at 24 h, again
suggesting that ischaemia is unlikely to be the cause of these
changes.
As with the BBB permeability changes, pre-treatment with
the non-speci®c ET receptor antagonist Ro 46-2005 had no
effect on the observed ADC reduction, although as discussed
above this does not necessarily preclude the ET system from
playing a role in these changes. However, it does indicate that
the reduction in ADC is independent of the acute reduction in
rCBV. In animals injected intrastriatally with rhuTNF-a,
there appeared to be a dose-dependent effect on tissue ADC.
This ®nding suggests that, as for the BBB permeability
changes, the pathways induced by both TNF-a receptors may
be involved in the processes underlying the ADC changes.
Moreover, the reduction in ADC may be a consequence of
BBB breakdown, which spatially corresponded closely to the
region of decreased ADC. Alternatively, the ADC changes
may be induced by direct effects of the cytokine on cellular
metabolism and function. It has been shown that TNF-a is not
directly toxic to neurones (Piani et al., 1992; Barone et al.,
1997). However, recent studies have shown that TNF-a
markedly inhibits glutamate uptake in both human and rat
astrocytes in culture (Fine et al., 1996; Ye and Sontheimer,
1996; Hu et al., 2000). Thus, glutamate-induced toxicity and
resultant energetic compromise of neurones may contribute to
the observed reduction in tissue water diffusion at 24 h. It has
also been suggested that TNF-a may impair the ability of
astrocytes to provide adequate energy substrates to neurones
for oxidation (Yu et al., 1995), which also could result in
neuronal dysfunction.
In summary, these studies demonstrate that a single bolus
injection of TNF-a into the brain parenchyma gives rise to an
2457
acute reduction in rCBV, followed by breakdown of the
BBBs and a reduction in tissue water diffusion. Furthermore,
chronic expression of TNF-a in the brain leads to a prolonged
CBV de®cit. These observations identify detrimental consequences of TNF-a in the brain which may contribute to
neuronal dysfunction and degeneration in CNS injury and
disease. Our data suggest that both endothelin receptors and
the TNFR2 pathway are potential targets for therapeutic
intervention in neuropathologies, such as multiple sclerosis,
cerebral malaria and HIV encephalitis that are associated with
high cerebral TNF-a expression.
Acknowledgements
This work was supported by the Medical Research Council
under the MRC-funded co-operative entitled The Host
Response to Acute Brain Injury. The authors thank Ms Sara
Fearn for technical assistance, and Dr Mario Cortina Borja for
advice and assistance with the statistics.
References
Achim CL, Heyes MP, Wiley CA. Quantitation of human immunode®ciency virus, immune activation factors, and quinolinic acid in
AIDS brains. J Clin Invest 1993; 91: 2769±75.
Aggarwal BB, Natarajan K. Tumor necrosis factors: developments
during the last decade. [Review]. Eur Cytokine Netw 1996; 7: 93±
124.
Angstwurm K, Freyer D, Dirnagl U, Hanisch UK, Schumann RR,
Einhaupl KM, et al. Tumour necrosis factor alpha induces only
minor in¯ammatory changes in the central nervous system, but
augments experimental meningitis. Neuroscience 1998; 86: 627±34.
Anthony DC, Bolton SJ, Fearn S, Perry VH. Age-related effects of
interleukin-1b on polymorphonuclear neutrophil-dependent
increases in blood-brain barrier permeability in rats. Brain 1997;
120: 435±44.
Anthony DC, Miller KM, Fearn S, Townsend MJ, Opdenakker G,
Wells GM, et al. Matrix metalloproteinase expression in an
experimentally-induced DTH model of multiple sclerosis in the
rat CNS. J Neuroimmunol 1998; 87: 62±72.
Anthony DC, Blond D, Dempster R, Perry VH. Chemokine targets
in acute brain injury and disease. [Review]. Prog Brain Res 2001;
132: 507±24.
Arnett HA, Mason J, Marino M, Suzuki K, Matsushima GK, Ting
JP. TNF alpha promotes proliferation of oligodendrocyte
progenitors and remyelination. Nat Neurosci 2001; 4: 1116±22.
Barone FC, Arvin B, White RF, Miller A, Webb CL, Willette RN,
et al. Tumor necrosis factor-alpha. A mediator of focal ischemic
brain injury. Stroke 1997; 28: 1233±44.
Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor
spectroscopy and imaging. Biophys J 1994; 66: 259±67.
Blamire AM, Anthony DC, Rajagopalan B, Sibson NR, Perry VH,
Styles P. Interleukin-1b-induced changes in blood-brain barrier
permeability, apparent diffusion coef®cient, and cerebral blood
2458
N. R. Sibson et al.
volume in the rat brain: a magnetic resonance study. J Neurosci
2000; 20: 8153±9.
Bolander HG, Persson L, Hillered L, d'Argy R, Ponten U, Olsson
Y. Regional cerebral blood ¯ow and histopathologic changes after
middle cerebral artery occlusion in rats. Stroke 1989; 20: 930±7.
Brian JE Jr, Faraci FM. Tumor necrosis factor-alpha-induced
dilatation of cerebral arterioles. Stroke 1998; 29: 509±15.
Brown H, Turner G, Rogerson S, Tembo M, Mwenechanya J,
Molyneux M, et al. Cytokine expression in the brain in human
cerebral malaria. J Infect Dis 1999; 180: 1742±6.
Bruce AJ, Boling W, Kindy MS, Peschon J, Kraemer PJ, Carpenter
MK, et al. Altered neuronal and microglial responses to excitotoxic
and ischemic brain injury in mice lacking TNF receptors. Nat
Medicine 1996; 2: 788±94.
Clark RS, Schiding JK, Kaczorowski SL, Marion DW, Kochanek
PM. Neutrophil accumulation after traumatic brain injury in rats:
comparison of weight drop and controlled cortical impact models.
J Neurotrauma 1994; 11: 499±506.
Claudio L, Kress Y, Factor J, Brosnan CF. Mechanisms of edema
formation in experimental autoimmune encephalomyelitis. The
contribution of in¯ammatory cells. Am J Pathol 1990; 137: 1033±45.
Clozel M, Breu V, Burri K, Cassal JM, Fischli W, Gray GA,
et al. Pathophysiological role of endothelin revealed by the ®rst
orally active endothelin receptor antagonist. Nature 1993; 365:
759±61.
Dawson DA, Martin D, Hallenbeck JM. Inhibition of tumor necrosis
factor-a reduces focal cerebral ischemic injury in the spontaneously
hypertensive rat. Neurosci Lett 1996; 218: 41±4.
de Vries HE, Blom-Roosemalen MC, van Oosten M, de Boer AG,
van Berkel TJ, Breimer DD, et al. The in¯uence of cytokines on the
integrity of the blood-brain barrier in vitro. J Neuroimmunol 1996;
64: 37±43.
Dijkstra CD, Dopp EA, Joling P, Kraal G. The heterogeneity of
mononuclear phagocytes in lymphoid organs: distinct macrophage
subpopulations in the rat recognized by monoclonal antibodies
ED1, ED2 and ED3. Immunology 1985; 54: 589±99.
Durieu-Trautmann O, Federici C, Creminon C, Foignant-Chaverot
N, Roux F, Claire M, et al. Nitric oxide and endothelin secretion by
brain microvessel endothelial cells: regulation by cyclic
nucleotides. J Cell Physiol 1993; 155: 104±11.
Gourin CG, Shackford SR. Production of tumor necrosis factor-a
and interleukin-1b by human cerebral microvascular endothelium
after percussive trauma. J Trauma 1997; 42: 1101±7.
Hauser SL, Doolittle TH, Lincoln R, Brown RH, Dinarello CA.
Cytokine accumulations in CSF of multiple sclerosis patients:
frequent detection of interleukin-1 and tumor necrosis factor but not
interleukin-6. Neurology 1990; 40: 1735±9.
Hoehn-Berlage M. Diffusion-weighted NMR imaging: application
to experimental focal cerebral ischemia. [Review]. NMR Biomed
1995; 8: 345±58.
Hsu SM, Raine L, Fanger H. The use of antiavidin antibody and
avidin±biotin±peroxidase complex in immunoperoxidase technics.
Am J Clin Pathol 1981; 75: 816±21.
Hu S, Sheng WS, Ehrlich LC, Peterson PK, Chao CC. Cytokine
effects on glutamate uptake by human astrocytes. Neuroimmunomodulation 2000; 7: 153±9.
Jacewicz M, Tanabe J, Pulsinelli WA. The CBF threshold and
dynamics for focal cerebral infarction in spontaneously
hypertensive rats. J Cereb Blood Flow Metab 1992; 12: 359±70.
Kedzierski RM, Yanagisawa M. Endothelin system: the doubleedged sword in health and disease. [Review]. Annu Rev Pharmacol
Toxicol 2001; 41: 851±76.
Klemm P, Warner TD, Hohlfeld T, Corder R, Vane JR. Endothelin
1 mediates ex vivo coronary vasoconstriction caused by exogenous
and endogenous cytokines. Proc Natl Acad Sci USA 1995; 92:
2691±5.
Kolb M, Margetts PJ, Anthony DC, Pitossi F, Gauldie J. Transient
expression of IL-1b induces acute lung injury and chronic repair
leading to pulmonary ®brosis. J Clin Invest 2001; 107: 1529±36.
Kwiatkowski D, Hill AV, Sambou I, Twumasi P, Castracane J,
Manogue KR, et al. TNF concentration in fatal cerebral, non-fatal
cerebral, and uncomplicated Plasmodium falciparum malaria.
Lancet 1990; 336: 1201±4.
Latour LL, Hasegawa Y, Formato JE, Fisher M, Sotak CH.
Spreading waves of decreased diffusion coef®cient after cortical
stimulation in the rat brain. Magn Reson Med 1994; 32: 189±98.
Lavine SD, Hofman FM, Zlokovic BV. Circulating antibody against
tumor necrosis factor-alpha protects rat brain from reperfusion
injury. J Cereb Blood Flow Metab 1998; 18: 52±8.
Feuerstein GZ, Liu T, Barone FC. Cytokines, in¯ammation, and
brain injury: role of tumor necrosis factor-a. [Review]. Cerebrovasc
Brain Metab Rev 1994; 6: 341±60.
Lewis M, Tartaglia LA, Lee A, Bennett GL, Rice GC, Wong GH,
et al. Cloning and expression of cDNAs for two distinct murine
tumor necrosis factor receptors demonstrate one receptor is species
speci®c. Proc Natl Acad Sci USA 1991; 88: 2830±4.
Fine SM, Angel RA, Perry SW, Epstein LG, Rothstein JD,
Dewhurst S, et al. Tumor necrosis factor alpha inhibits glutamate
uptake by primary human astrocytes. Implications for pathogenesis
of HIV-1 dementia. J Biol Chem 1996; 271: 15303±6.
Lucas R, Juillard P, Decoster E, Redard M, Burger D, Donati Y,
et al. Crucial role of tumor necrosis factor (TNF) receptor 2 and
membrane-bound TNF in experimental cerebral malaria. Eur J
Immunol 1997; 27: 1719±25.
Frahm J, Haase A, Matthaei D. Rapid NMR imaging of dynamic
processes using the FLASH technique. Magn Reson Med 1986; 3:
321±7.
Marr RA, Addison CL, Snider D, Muller WJ, Gauldie J, Graham
FL. Tumour immunotherapy using an adenoviral vector expressing
a membrane-bound mutant of murine TNF alpha. Gene Ther 1997;
4: 1181±8.
Glabinski A, Krajewski S, Rafalowska J. Tumor necrosis factor-a
induced pathology in the rat brain: characterization of stereotaxic
injection model. Folia Neuropathol 1998; 36: 52±62.
Matsuo Y, Onodera H, Shiga Y, Nakamura M, Ninomiya M, Kihara
T, et al. Correlation between myeloperoxidase-quanti®ed neutrophil
TNF-a alters CBV and tissue homeostasis
2459
accumulation and ischemic brain injury in the rat. Effects of
neutrophil depletion. Stroke 1994; 25: 1469±75.
factor alpha in brain injury. Cytokine Growth Factor Rev 1999; 10:
119±30.
Megyeri P, Abraham CS, Temesvari P, Kovacs J, Vas T, Speer CP.
Recombinant human tumor necrosis factor alpha constricts pial
arterioles and increases blood-brain barrier permeability in newborn
piglets. Neurosci Lett 1992; 148: 137±40.
Skopal J, Turbucz P, Vastag M, Bori Z, Pek M, deChatel R, et al.
Regulation of endothelin release from human brain microvessel
endothelial cells. J Cardiovasc Pharmacol 1998; 31 Suppl 1: S370±2.
Nawashiro H, Martin D, Hallenbeck JM. Inhibition of tumor
necrosis factor and amelioration of brain infarction in mice. J Cereb
Blood Flow Metab 1997; 17: 229±32.
Neill AL, Hunt NH. Effects of endotoxin and dexamethasone on
cerebral malaria in mice. Parasitology 1995; 111: 443±54.
Nuovo GJ, Gallery F, MacConnell P, Braun A. In situ detection of
polymerase chain reaction-ampli®ed HIV-1 nucleic acids and tumor
necrosis factor-alpha RNA in the central nervous system. Am J
Pathol 1994; 144: 659±66.
Ordidge RJ, Helpern JA, Qing ZX, Knight RA, Nagesh V.
Correction of motional artifacts in diffusion-weighted MR images
using navigator echoes. Magn Reson Imaging 1994; 12: 455±60.
Perry VH, Linden R. Evidence for dendritic competition in the
developing retina. Nature 1982; 297: 683±5.
Piani D, Spranger M, Frei K, Schaffner A, Fontana A. Macrophageinduced cytotoxicity of N-methyl-D-aspartate receptor positive
neurons involves excitatory amino acids rather than reactive oxygen
intermediates and cytokines. Eur J Immunol 1992; 22: 2429±36.
Rosen BR, Belliveau JW, Vevea JM, Brady TJ. Perfusion imaging
with NMR contrast agents. Magn Reson Med 1990; 14: 249±65.
Rosenberg GA, Estrada EY, Dencoff JE, Stetler-Stevenson WG.
Tumor necrosis factor-a-induced gelatinase B causes delayed
opening of the blood-brain barrier: an expanded therapeutic
window. Brain Res 1995; 703: 151±5.
Stefferl A, Hopkins SJ, Rothwell NJ, Luheshi GN. The role of TNFalpha in fever: opposing actions of human and murine TNF-a and
interactions with IL-b in the rat. Br J Pharmacol 1996; 118: 1919±24.
Tamura A, Graham DI, McCulloch J, Teasdale GM. Focal cerebral
ischaemia in the rat: 2. Regional cerebral blood ¯ow determined by
[14C]iodoantipyrine autoradiography following middle cerebral
artery occlusion. J Cereb Blood Flow Metab 1981; 1: 61±9.
Tureen J. Effect of recombinant human tumor necrosis factor-a on
cerebral oxygen uptake, cerebrospinal ¯uid lactate, and cerebral blood
¯ow in the rabbit: role of nitric oxide. J Clin Invest 1995; 95: 1086±
91.
van der Toorn A, Sykova E, Dijkhuizen RM, Vorisek I, Vargova L,
Skobisova E, et al. Dynamic changes in water ADC, energy
metabolism, extracellular space volume, and tortuosity in neonatal
rat brain during global ischemia. Magn Reson Med 1996; 36: 52±
60.
Vonesh EF, Chinchilli VM. Linear and nonlinear models for the
analysis of repeated measurements. New York: Marcel Dekker; 1997.
Waage A, Halstensen A, Shalaby R, Brandtzaeg P, Kierulf P, Espevik
T. Local production of tumor necrosis factor alpha, interleukin 1, and
interleukin 6 in meningococcal meningitis. Relation to the
in¯ammatory response. J Exp Med 1989; 170: 1859±67.
Woodroofe MN, Cuzner ML. Cytokine mRNA expression in
in¯ammatory multiple sclerosis lesions: detection by nonradioactive in situ hybridization. Cytokine 1993; 5: 583±8.
Scherbel U, Raghupathi R, Nakamura M, Saatman KE, Trojanowski
JQ, Neugebauer E, et al. Differential acute and chronic responses of
tumor necrosis factor-de®cient mice to experimental brain injury.
Proc Natl Acad Sci USA 1999; 96: 8721±6.
Yang GY, Gong C, Qin Z, Liu XH, Lorris Betz A. Tumor necrosis
factor alpha expression produces increased blood-brain barrier
permeability following temporary focal cerebral ischemia in mice.
Brain Res Mol Brain Res 1999; 69: 135±43.
Sharief MK, Ciardi M, Thompson EJ. Blood-brain barrier damage
in patients with bacterial meningitis: association with tumor
necrosis factor-alpha but not interleukin-1b. J Infect Dis 1992;
166: 350±8.
Ye ZC, Sontheimer H. Cytokine modulation of glial glutamate uptake:
a possible involvement of nitric oxide. Neuroreport 1996; 7: 2181±5.
Shohami E, Bass R, Wallach D, Yamin A, Gallily R. Inhibition of
tumor necrosis factor alpha (TNFa) activity in rat brain is
associated with cerebroprotection after closed head injury.
[Review]. J Cereb Blood Flow Metab 1996; 16: 378±84.
Shohami E, Ginis I, Hallenbeck JM. Dual role of tumor necrosis
Yu N, Maciejewski-Lenoir D, Bloom FE, Magistretti PJ. Tumor
necrosis factor-a and interleukin-1 alpha enhance glucose
utilization by astrocytes: involvement of phospholipase A2. Mol
Pharmacol 1995; 48: 550±8.
Received March 28, 2002. Revised May 3, 2002.
Accepted May 16, 2002