a r t i c l e s
nature medicine VOLUME 16 | NUMBER 4 | APRIL 2010 4 1 3
Epilepsy is a disabling neurological disorder characterized by recur-
ring, unprovoked seizures. It affects about 1% of the population of
all ages and often requires lifelong medication
1
. In about 30% of
affected individuals, epilepsy is refractory to pharmacological treat-
ment
2
, and surgical removal of the epileptic focus is suitable only
for a minority of them. Understanding the molecular events
underlying the occurrence of seizures is necessary for devising
new therapeutic approaches.
Increasing evidence supports the involvement of inflammatory and
immune processes in the etiopathogenesis of seizures
3
. Inflammatory
responses induced by brain-damaging events such as neurotrauma,
stroke, infection, febrile seizures and status epilepticus are associated
with acute symptomatic seizures and a high risk of developing epi-
lepsy
4,5
. Pronounced inflammatory processes have been described in
epileptogenic brain tissue from drug-resistant patients with temporal
lobe epilepsy (TLE)
6–8
and epilepsies associated with developmental
malformations of the cortex
9,10
. Pharmacological and genetic studies
in animal models have shown that specific inflammatory mediators
such as cytokines, complement factors and prostaglandins substan-
tially contribute to seizures and that interfering with these molecules
or their receptors can reduce seizure frequency and severity
11
.
We previously showed in rats and mice that the proinflammatory
cytokine IL-1β is rapidly upregulated during seizures in microglia,
astrocytes and endothelial cells in the epileptic focus as well as in
forebrain regions recruited in epileptic activity
8,12–14
. IL-1β exerts
powerful proconvulsant actions via a signaling pathway in neurons
involving its receptor IL-1R1, the IL-1R accessory protein and myeloid
differentiation primary response protein (MyD88) complex and Src
family kinases, leading to NMDA receptor-2B (NR2B) phosphoryla-
tion and enhanced NMDA-dependent Ca
2+
influx
12,15,16
.
The pathway activated by IL-1β depends on MyD88, but other sur-
face receptors can recruit MyD88, notably TLRs, which have a key
role in pathogen recognition
17
. TLRs recognize various molecules
of microbial origin, called pathogen-associated molecular patterns
(PAMPs), and trigger inflammation by inducing the transcription
of genes encoding cytokines, including IL-1β. TLR4 in particular
detects lipopolysaccharide (LPS), a major outer membrane compo-
nent of Gram-negative bacteria. Given that TLR4 is expressed in the
brain
18
and LPS lowers the seizure threshold in rodents
19,20
, we sought
to investigate whether TLR4 has a role in the onset and recurrence
of seizures.
Increasing evidence indicates that, in the absence of pathogens,
TLR signaling can be activated by molecules released by injured tis-
sue
21
. These molecules, named damage-associated molecular pat-
terns (DAMPs), include HMGB1, a nearly ubiquitous chromatin
component that is passively released by necrotic cells, retained by
Toll-like receptor 4 and high-mobility group box-1
are involved in ictogenesis and can be targeted to
reduce seizures
Mattia Maroso
1
, Silvia Balosso
1
, Teresa Ravizza
1
, Jaron Liu
2
, Eleonora Aronica
3,4
, Anand M Iyer
3
, Carlo Rossetti
5,6
,
Monica Molteni
6
, Maura Casalgrandi
7
, Angelo A Manfredi
8
, Marco E Bianchi
2
& Annamaria Vezzani
1
Brain inflammation is a major factor in epilepsy, but the impact of specific inflammatory mediators on neuronal excitability
is incompletely understood. Using models of acute and chronic seizures in C57BL/6 mice, we discovered a proconvulsant
pathway involving high-mobility group box-1 (HMGB1) release from neurons and glia and its interaction with Toll-like receptor
4 (TLR4), a key receptor of innate immunity. Antagonists of HMGB1 and TLR4 retard seizure precipitation and decrease acute
and chronic seizure recurrence. TLR4-defective C3H/HeJ mice are resistant to kainate-induced seizures. The proconvulsant
effects of HMGB1, like those of interleukin-1b (IL-1b), are partly mediated by ifenprodil-sensitive N-methyl-d-aspartate (NMDA)
receptors. Increased expression of HMGB1 and TLR4 in human epileptogenic tissue, like that observed in the mouse model of
chronic seizures, suggests a role for the HMGB1-TLR4 axis in human epilepsy. Thus, HMGB1-TLR4 signaling may contribute
to generating and perpetuating seizures in humans and might be targeted to attain anticonvulsant effects in epilepsies that are
currently resistant to drugs.
1
Department of Neuroscience, Mario Negri Institute for Pharmacological Research, Milano, Italy.
2
Department of Genetics and Cell Biology, San Raffaele University
and San Raffaele Research Institute, Milano, Italy.
3
Department (Neuro) Pathology, Academisch Medisch Centrum, Amsterdam, The Netherlands.
4
The Netherlands
Foundation (Stichting Epilepsie Instellingen Nederland), Heemstede, The Netherlands.
5
Department of Biotechnology and Molecular Sciences, University of Insubria,
Varese, Italy.
6
Department of Environmental Health Sciences, Mario Negri Institute for Pharmacological Research, Milano, Italy.
7
HMGBiotech srl, Milano, Italy.
8
Department of Regenerative Medicine, San Raffaele University and San Raffaele Research Institute, Milano, Italy. Correspondence should be addressed to
A.V. (vezzani@marionegri.it).
Received 29 December 2009; accepted 25 February 2010; published online 28 March 2010; doi:10.1038/nm.2127
© 2010 Nature America, Inc. All rights reserved.
a r t i c l e s
4 1 4 VOLUME 16 | NUMBER 4 | APRIL 2010 nature medicine
cells undergoing programmed death (apoptosis) and actively secreted
by cells in profound distress
22,23
. HMGB1 secretion does not follow
the canonical endoplasmic reticulum–Golgi pathway and occurs via
relocation of the nuclear protein to the cytoplasm
24
. Extracellular
HMGB1 binds the receptor for advanced glycation end products
22
and several other receptors, including TLR4
25,26
.
We asked whether proconvulsive stimuli could lead to HMGB1
release and TLR4 activation, thereby contributing to seizures. Our
results indicate that HMGB1 and TLR4 contribute to the generation
and severity of seizures, which therefore can be reduced by TLR4 and
HMGB1 antagonists. Studies in spontaneously epileptic mice and
in human TLE tissue indicate that the HMGB1-TLR4 axis also
contributes to seizures in chronic epilepsy.
RESULTS
Expression of HMGB1 and TLR4 in mouse models of seizures
We first investigated whether seizure activity affects the expression of
HMGB1 and TLR4. We used two mouse models of focal-onset acute
seizures induced by unilateral intrahippocampal injection of kainic
acid (7 ng) or bicuculline (51 ng) (Supplementary Fig. 1). Kainic acid
is an agonist of one class of glutamate receptors, whereas bicuculline is
an antagonist of γ-aminobutyric acid A receptors; they trigger seizures
by increasing excitatory neurotransmission and decreasing inhibitory
neurotransmission, respectively. In these models, the seizure activity is
similar and occurs to the same extent in the injected and contralateral
side of the hippocampus. Kainic acid at low dose causes excitotoxic cell
damage only to pyramidal cells in the CA3 area of the injected hippo-
campus
16,27
, whereas bicuculline provokes seizures in the absence of
neurodegeneration
28
. We also used an established mouse model of
chronic epilepsy, where seizure activity develops within 1 week after
intrahippocampal application of 200 ng kainic acid
29,30
. This model
recapitulates the major neuropathological features of human TLE
(Supplementary Fig. 2) and, in particular, recurring spontaneous
seizures that do not respond to various anticonvulsant drugs
29
.
We investigated the distribution of the immunohistochemi-
cal signal of HMGB1 in mice injected with kainic acid (Fig. 1a–d).
In control hippocampi, HMGB1 is present mostly in nuclei of the
pyramidal neurons (Fig. 1a) and granule cells of the dentate gyrus
(data not shown). In the strata radiatum, lacunosum-moleculare and
moleculare, we observed scattered cells with nuclear staining as well
as neurons with both nuclear and cytoplasmic staining; most cells
were HMGB1 negative (Fig. 1a). Between 1 and 3 h after the onset of
a
d
f
i
g h
e
b c
CA1
Rad
LMol
GFAP CD11b GFAP CD11b
HMGB1 Hoechst
CA1 CA3
Hilus
CA1 CA3
Hilus
HMGB1 Hoechst
GFAP CD11b
HMGB1 Hoechst
0
C 1 3
*
*
#$
*
#$
*
$
C 1 3
Nuclear staining
Extranuclear staining
C 1 3 h C
ND NDND NDND ND
1 3 C 1 3 C
C
1 3 h
3h
25
50
Percentage
HMGB1-positive
astrocytes
Number of HMGB1-positive
monocyte/microglia cells
HMGB1/GAPDH
(arbitrary units)
75
100
0
5
10
15
0
0.5
1.0
1.5
*
CA1
MoDG
3V
TLR4 CD11b
TLR4 NeuN
TLR4 NeuN
TLR4 CD11b
TLR4 CD11b
TLR4 GFAP
TLR4 GFAP
Figure 1 HMGB1 and TLR4 immunoreactivity
in the CA1 pyramidal layer of hippocampi
of kainic acid–injected C57BL/6 mice.
(a–c) Photomicrographs of hippocampi
injected with vehicle (a), or 1 h (b) and
3 h (c) after kainic acid–induced seizures.
Top two rows, HMGB1 immunoreactivity
in nuclei of pyramidal neurons and cells
(arrows) of the strata radiatum (Rad) and
lacunosum-molecolare (LMol); some cells
with neuronal morphology (green arrows) show
cytoplasmic immunoreactivity. Cells with
astrocytic morphology (b,c) show HMGB1
in the cytoplasm (arrowheads). Bottom row,
HMGB1 signal only (left) and colocalization of
HMGB1, DNA (Hoechst), GFAP for astrocytes
(middle) and CD11b for microglia-like cells
(right). GFAP-positive cells show HMGB1
in nuclei (a) and around nuclei (b,c).
(d) Quantification of HMGB1-positive cells
in control-injected hippocampi (C), 1 h and
3 h after seizures (means ± s.e.m., n = 4).
Nuclear staining: *P < 0.05 versus control;
#P < 0.05 versus 1 h; Extranuclear staining:
$P < 0.05 versus 1 h and control; one-way
analysis of variance (ANOVA) followed
by Tukey’s test. ND, not detectable.
(e) Quantification of western blots for HMGB1
in mouse hippocampal homogenate from
control mice (C) and mice 3 h after kainic
acid seizures. Error bars (means ± s.e.m.,
n = 5) represent the ratios of the optical
densities of the HMGB1 and glyceraldehyde-
3-phosphate dehydrogenase (GAPDH) bands;
*P < 0.05 by Mann-Whitney test. (f–i) TLR4
immunoreactivity in hippocampi injected with
vehicle (f) or 1 h (g) or 3 h (h) after kainate-
induced seizures. Arrows in g and h point
to neurons. Colocalization of TLR4 with
cell-specific markers (g,h, second row) in
NeuN-positive neurons, GFAP-positive astrocytes,
CD11b-positive microglia-like cells. (i) TLR4
immunoreactivity in CD11b-positive cells in mouse hippocampus after intracerebroventricular lipopolysaccharide injection. MoDG, molecular layer of
dentate gyrus; 3V, third ventricle. Scale bars: a–c, f–h (top row) 75 µm; a–c (middle row) 25 µm; a–c (bottom row), g,h (bottom row) and i, 15 µm.
© 2010 Nature America, Inc. All rights reserved.
a r t i c l e s
nature medicine VOLUME 16 | NUMBER 4 | APRIL 2010 4 1 5
seizures, we found a progressive increase in nuclear and perinuclear
HMGB1 staining in astrocytes (Fig. 1a–c) in all subfields of injected
(Fig. 1d) and contralateral hippocampi (Supplementary Fig. 3b).
Although we did not detect HMGB1 in cells with microglial mor-
phology in control hippocampi, a short-lived wave of cytoplasmic
expression in microglia appeared 1 h after seizure onset and was again
absent after 3 h (Fig. 1a–d). In agreement with immunohistochemical
results, we measured, by western blotting, an average 27% increase in
HMGB1 abundance in kainic acid–injected hippocampi 3 h after the
onset of seizures (Fig. 1e). This increase in the whole tissue represents
a substantial increase in the fraction of cells affected.
The induction of HMGB1 in the hippocampus after bicuculline-
induced seizures was bilateral and similar to that described for kainic
acid–induced seizures (Supplementary Figs. 3 and 4), although the
changes were less prominent. In chronic epileptic mice, the cell pattern
and extent of HMGB1 expression in hippocampi (Supplementary
Figs. 3 and 5) were similar to those described after acute kainic acid–
induced seizures.
We did not find a parallel change in the number of neurons showing
nuclear staining, cytoplasmic staining or both in any seizure model
(data not shown). All changes observed in the CA1 region were similar
in the CA3 region and in dentate gyrus (data not shown).
We next investigated TLR4 expression (Fig. 1f–h and Supplementary
Figs. 4 and 5). Whereas we found no signal in control slices (Fig. 1f),
we observed substantial TLR4 expression 1 h (Fig. 1g) and 3 h (Fig. 1h)
after seizure onset in neurons within the pyramidal layers; glial fibril-
lary acidic protein (GFAP)-positive astrocytes also expressed TLR4 in
the various hippocampal subfields including CA1 (Fig. 1g,h), CA3 and
hilus (data not shown). TLR4 was not expressed in CD11b-positive
cells of microglial morphology (Fig. 1g,h). As a positive control, we did
detect TLR4 in CD11b-positive cells in the hippocampus after intra-
ventricular LPS injection (Fig. 1i). We observed a similar pattern of
TLR4 induction in both neurons and astrocytes, but not in microglia, in
chronic epileptic mice (Supplementary Fig. 5). In bicuculline-injected
mice, TLR4 staining increased bilaterally in neurons, but not in astro-
cytes, 1.5 h after the onset of seizures, when epileptic activity had just
abated (Supplementary Fig. 4). These differences from the kainic acid
model may be due to the lack of cell loss, the shorter duration of sei-
zures (90 min bicuculline versus 120 min kainic acid) or both.
Expression of HMGB1 and TLR4 in human epileptogenic tissue
Having characterized a specific pattern of HMGB1 and TRL4 expression
in the hippocampi of chronic epileptic mice, we looked for a similar
pattern in hippocampal specimens obtained at surgery in subjects with
drug-resistant TLE with hippocampal sclerosis (TLE-HS) (Fig. 2).
In autoptic control tissue, we detected HMGB1 signal in nuclei of cells
identified a posteriori as neurons and astrocytes, as well as in neuronal
cytoplasm (Fig. 2a). In TLE-HS, cytoplasmic HMGB1 staining was sub-
stantially increased in GFAP-positive astrocytes as well as in human
leukocyte antigen DR (HLA-DR)-positive processes (Fig. 2a–c).
TLR4 was undetectable in autoptic control hippocampi, whereas
it was clearly detectable in pyramidal neurons and GFAP-positive
astrocytes in TLE-HS specimens (Fig. 2d). HLA-DR–positive micro-
glia-like cells did not show TLR4 staining (Fig. 2e), whereas HLA-
DR–positive cells in multiple sclerosis specimens did stain for TLR4
(Fig. 2f). We found similar patterns in the CA3 and hilus of the den-
tate gyrus (data not shown). HMGB1 and TLR4 staining in surgical
hippocampal specimens from subjects with focal epileptogenic lesions
not involving the hippocampus proper was similar to that in autoptic
control tissue (data not shown).
HMGB1 promotes seizures in a TLR4-dependent way
HMGB1 and TLR4 form a ligand-receptor pair. To examine whether
TLR4 activation by HMGB1 contributes to ictogenesis, we used
a
b
c
d
e f
Control
Control
CA1
HS
HS HS
HMGB1 GFAP HMGB1 GFAP
HMGB1 HLA-DR
CA1 CA3 Hilus CA1 CA3 Hilus
0
C HS C HS C HS
C HS C HS
C
HS
ND ND ND
25
50
Percentage HMGB1-positive
astrocytes
Percentage HMGB1-positive
monocyte/microglia cells
75
100
0
25
50
75
100
*
*
Nuclear staining
Control
HS
Extranuclear staining Nuclear + extranuclear staining
TLR4 NeuN TLR4 GFAP TLR4 HLA-DR TLR4 HLA-DR
*
Figure 2 HMGB1 and TLR4 immunoreactivity in the hippocampi of
control subjects and subjects with TLE-HS. (a) Immunohistochemical
staining for HMGB1 in the CA1 region of control individuals and
individuals with TLE. Arrows indicate pyramidal neurons. Double
arrows point to neurons with prominent cytoplasmic immunoreactivity;
arrowheads point to cells with glial morphology with nuclear staining and
double arrowheads point to cells with glial morphology with cytoplasmic
staining. (b) Immunofluorescence of HMGB1, GFAP and HLA-DR in
hippocampi from control individuals and individuals with TLE-HS.
(c) Quantification of HMGB1-positive cells in control subjects (C) and subjects with TLE-HS (means ± s.e.m., n = 6). Extranuclear staining: *P < 0.05
versus control, one-way ANOVA followed by Tukey’s test. (d) TLR4 immunostaining in the CA1 region; arrows point to neurons and arrowheads to
reactive glial cells. (e) Confocal images showing colocalization of TLR4 with NeuN in neuronal cells, or with GFAP in reactive astrocytes, and lack of
colocalization of TLR4 with HLA-DR in cells of the microglia/macrophage lineage. (f) TLR4 staining in HLA-DR–positive cells in a brain tissue sample of
a subject with multiple sclerosis. Sections in a and d are counterstained with hematoxylin. Scale bars: a,d, 50 µm; b,e,f, 20 µm.
© 2010 Nature America, Inc. All rights reserved.
a r t i c l e s
4 1 6 VOLUME 16 | NUMBER 4 | APRIL 2010 nature medicine
pharmacological and genetic approaches. First, we increased extracel-
lular HMGB1 amounts in C57BL/6 mice by intrahippocampal injec-
tion of recombinant HMGB1 15 min before kainic acid application
(Fig. 3a). HMGB1 addition (5.5 and 9.5 µg per mouse) increased the
frequency of kainic acid–induced seizures by about 2.5-fold (repre-
sentative electroencephalogram (EEG) tracings in Supplementary
Fig. 1a–d) and the total time spent in seizures. The time of onset of
the first seizure was markedly reduced by 9.5 µg HMGB1. An HMGB1
dose of 3.2 µg per mouse was ineffective.
We next investigated whether HMGB1 signals through TLR4 by using
C3H/HeJ mice, which have a spontaneous mutation in the Toll/IL-1
receptor domain of TLR4 (Tlr4
Lps-d
), leading to functional inactivation
of the receptor
31
. HMGB1 (5.5 µg per mouse) had no effect on kainic
acid–induced seizures when injected into C3H/HeJ mice but was
proconvulsant in their wild-type C3H/HeouJ counterparts, increasing
the time spent in seizures by about twofold (Fig. 3b and Supplementary
Fig. 1f–i). C3H/HeJ mice also had a reduced intrinsic susceptibility to
seizures, as shown by a delay in kainic acid–induced seizure onset and a
substantial reduction in seizure frequency and total duration (Fig. 3b).
Overstimulated neurons release HMGB1
A high local concentration of glutamate, resulting from hyperexcita-
tion of the neuronal network, is thought to have a key role in the initia-
tion and spread of seizures. We thus hypothesized that overstimulation
of glutamate receptors might induce neurons to release HMGB1.
We exposed mixed primary cocultures of neurons and glia to
0.25 mM glutamate for 2 h then fixed and immunostained the cells
(Supplementary Fig. 6). HMGB1 was located only in the nuclei
of unchallenged neurons and glial cells; after glutamate exposure,
HMGB1 signal was still present only in the nuclei of glial cells, but in
neurons it was located in both the nucleus and the cytoplasm (neuN-
positive cells, Supplementary Fig. 6a). We confirmed these results
in primary cortical neurons infected with a lentivirus expressing
an HMGB1-GFP fusion protein. Unchallenged neurons contained
HMGB1-GFP only in the nucleus, whereas neurons exposed to
0.25 mM glutamate had a large amount of HMGB1-GFP throughout
the cell bodies and processes (Supplementary Fig. 6b).
Neurons overstimulated by glutamate eventually undergo excitotoxic
death. Cortical neurons exposed in vitro to 0.25 mM glutamic acid
started to die after about 2 h and were almost completely dead after
24 h, as judged morphologically (Supplementary Fig. 6c) and by the
release of lactate dehydrogenase (Supplementary Fig. 6d). Notably,
HMGB1 was also released into the medium (Supplementary Fig. 6d).
We showed previously that HMGB1 retention upon apoptosis depends
on caspase-3 activation
32
; caspase-3 was not activated by glutamate
exposure, nor was DNA fragmented (ref. 33 and data not shown).
HMGB1 and TLR4 antagonists reduce acute and chronic seizures
Overall, our findings predict that pharmacological blockade of
the HMGB1-TLR4 axis should reduce the frequency and duration
of seizures. We then interfered with the activity of endogenous
HMGB1 by injecting BoxA, a fragment of HMGB1 with antago-
nistic activity
34–36
. BoxA injection (7.5 µg per mouse) delayed the
onset of seizures (Fig. 4a), and 2.5 and 7.5 µg of BoxA significantly
a b
0
–
3.2
5.5
9.5
*
**
**
**
**
**
**
**
**
**
**
–
3.2
5.5
9.5
–
–
–
3.2
5.5
Wild typeWild type Tlr4
Lps-d
Tlr4
Lps-d
Wild type Tlr4
Lps-d
–
5.5
–
5.5
–
5.5
–
5.5
0
10
20
30
0
10
20
30
5.5
Time in seizures (min)
5.5
9.5
(µg)
(µg)
5
10
Onset (min)
Onset (min)
15
0
10
20
Number of seizures
Number of seizures
Time in seizures (min)
30
0
0
10
40
80
5
10
15
20
20
Vehicle + KA
HMGB1 + KA
**
Figure 3 Dose-dependent proconvulsant effect of HMGB1 in wild-type mice and absence of HMGB1 effect in Tlr4
Lps-d
mice. (a) Seizure parameters (onset,
number and duration) in C57BL/6 mice (n = 8 or 9) injected in the hippocampus with HMGB1 at the indicated doses, 15 min before kainic acid (KA) injection.
Vehicle + KA group represents mice receiving the corresponding volume of PBS before kainic acid. Error bars are means ± s.e.m.; *P < 0.05, **P < 0.01 versus
vehicle + KA by one-way ANOVA followed by Tukey’s test. (b) Seizure parameters (onset, number and duration) in mice defective in TLR4 signaling (Tlr4
Lps-d
)
and the corresponding wild-type mice (n = 6 each group) after injection in the hippocampus with 5.5 µg HMGB1 or the corresponding volume of vehicle
followed 15 min later by kainic acid injection. Error bars are means ± s.e.m., **P < 0.01 versus wild type by two-way ANOVA followed by Tukey’s test.
a b
c d
e f
Vehicle + KA
Vehicle + bicuculline
BoxA + bicuculline
Lps-Rs + bicuculline
BoxA + KA
Lps-Rs + KA
Cyp + KA
0
0
0
100
200
Onset
(percentage of vehicle + KA)
Number of seizures
(percentage of vehicle + KA)
Ictal activity
(percentage of vehicle + KA)
Ictal activity
(percentage of vehicle + bicuculline)
Onset
(percentage of vehicle + bicuculline)
300
50
100
150
0
–
–
0.8
2.5
7.5
7.5
0.5
2.0
5.0
5.0
5.0
(µg)
(µg)
–
–
0.8
2.5
7.5
7.5
0.5
2.0
5.0
5.0
5.0
(µg)
(µg)
–
–
0.8
2.5
7.5
7.5
0.5
2.0
5.0
5.0
5.0
(µg)
(µg)
50
100
150
0
50
100
150
50
100
150
*
**
**
**
**
**
**
**
**
**
**
*
**
**
0
100
300
500
700
*
**
**
**
**
Number of seizures
(percentage of vehicle + bicuculline)
Figure 4 Anticonvulsant effects of BoxA, Lps-Rs and Cyp in acute seizure models. (a–f) Seizure parameters (onset, a,d; number, b,e; duration, c,f) in
mice injected in the hippocampus with vehicle or the various drugs at the doses indicated, followed 15 min later by kainic acid (a–c) or bicuculline
injection (d–f). Error bars (means ± s.e.m., n = 8–11 each group) represent percentage changes in inhibitor-treated mice versus mice injected with
vehicle + KA (onset 8.5 ± 1.4 min, number of seizures 10.0 ± 1.0, ictal activity 6.3 ± 0.5 min) or vehicle + bicuculline (onset 1.9 ± 0.3 min, number
of seizures 13.0 ± 1.0, ictal activity 9.1 ± 0.7 min). *P < 0.05, **P < 0.01 versus vehicle by one-way ANOVA followed by Tukey’s test.
© 2010 Nature America, Inc. All rights reserved.
a r t i c l e s
nature medicine VOLUME 16 | NUMBER 4 | APRIL 2010 4 1 7
reduced the frequency and duration of kainic acid–induced
seizures (Fig. 4b,c); a dose of 0.8 µg was ineffective.
We then tested two TLR4 antagonists, Rhodobacter sphaeroides LPS
(Lps-Rs)
37
and cyanobacterial LPS (Cyp)
38
. Lps-Rs had anticonvul-
sant effects already at 2 µg per mouse (Fig. 4a–c). Both antagonists
at 5 µg per mouse substantially ameliorated all parameters of kainic
acid–induced seizures (Fig. 4a–c).
BoxA and Lps-Rs were also highly effective in the bicuculline
model; mice experienced substantially fewer seizures, spent less time
in seizures and seizure onset was delayed (Fig. 4d–f).
Finally, we targeted the HMGB1-TLR4 axis in the model of
chronic spontaneous seizures (Fig. 5). We recorded EEG activity
in C57BL/6 mice with a stable baseline of spontaneous seizures
before and after injection of BoxA (7.5 µg per mouse) or Lps-Rs
(5 µg per mouse). Each antagonist lowered the number and
frequency of spontaneous seizures up to ~75% for about 2 h after
administration (Fig. 5c).
a
b
c
Start
Pre-injection baseline
Post-injection baseline
0 h
RHP
RHP
RHP
BoxA
RHP
LHP
Lps-Rs
LHP
LHP
LHP
200 µV
10 s
2 h 4 h 6 h 8 h
Drug
Checkpoint 1
Checkpoint 2
End
Baseline
0
0–2
Baseline
Drug
2–4
Time (h) Time (h)
4–6 0–2 2–4 4-6
Baseline
Drug
25
50
**
**
**
**
Baseline
BoxA
Lps-Rs
75
Number of seizures
(percentage of baseline)
lctal activity
(percentage of baseline)
100
125
0
25
50
75
100
125
Figure 5 Anticonvulsant effects of BoxA and Lps-Rs in the chronic seizure
model. (a) Experimental protocol. EEG seizures were recorded in eight
C57BL/6 mice with a stable baseline of spontaneous seizures (Online
Methods and Supplementary Fig. 2 contain details). The first EEG recording
period (0–2 h) was used to assess the baseline of spontaneous seizure
activity before drug injection (BoxA, 7.5 µg per mouse or Lps-Rs, 5 µg per
mouse). EEG activity was then measured continuously after drug injection,
and data were binned in 2-h periods. BoxA and Lps-Rs were injected into
the same mice, 5 d apart; mice were also injected with PBS 24 h after the
last drug administration. (b) Representative EEG tracings from mice with
spontaneous seizures recorded during the pre-injection baseline, during the
first 2 h after Lps-Rs or BoxA injection, and during the 4-h period before
ending the experiment (post-injection baseline). Ictal activity is delimited by
arrowheads. RHP and LHP, right and left hippocampus. (c) Quantification
of BoxA and Lps-Rs effects on chronic spontaneous seizures. Error bars
represent the means ± s.e.m. (n = 8) of ictal activity in each mouse,
expressed as percentage of the corresponding pre-injection baseline (raw
data in Supplementary Fig. 2). **P < 0.01 versus baseline by repeated
measures ANOVA followed by Dunnett’s test.
150
Onset
(percentage of vehicle
+ KA)
Number of seizures
(percentage of
baseline)
Number of
seizures (percentage
of baseline)
lctal activity
(percentage of
baseline)
lctal activity
(percentage of
baseline)
lctal activity
(percentage of
baseline)
lctal activity
(percentage of
baseline)
lctal activity
(percentage of
baseline)
lctal activity
(percentage of
baseline)
Number of
seizures (percentage
of baseline)
Number of
seizures (percentage
of baseline)
0
–
2
2
–
4
Number of
seizures (percentage
of baseline)
Number of
seizures (percentage
of baseline)
Number of seizures
(percentage of vehicle
+ KA)
lctal activity
(percentage of vehicle
+ KA)
250
200
150
50
0
125
100
75
50
25
0
100
250
200
150
50
0
100
100
50
0
a
b
c d
e f
g h
Pre-injection baseline
Ifenprodil
Ifenprodil
RHP
LHP
Post-injection baseline
125
100
75
50
25
0
125
100
75
50
25
0
125
100
75
50
25
0
125
100
75
50
25
0
125
100
75
50
25
0
125
100
75
50
25
0
ND
ND
ND
ND
125
100
75
50
25
0
125
100
75
50
Time (h)
25
0
125
100
75
50
25
0
125
100
75
50
25
0
125
100
75
50
25
0
RHP
LHP
LHP
LHP
RHP
RHP
Vehicle + KA
Ifenprodil + KA Ifenprodil + HMGB1 + KA
Baseline
200 µV
10 s
Ifenprodil
HMGB1 + KA
4
–
6
6
–
8
0
–
2
2
–
4
Time (h)
4
–
6
6
–
8
0
–
2
2
–
4
Time (h)
4
–
6
6
–
8
0
–
2
2
–
4
Time (h)
4
–
6
6
–
8
**
**
**
Figure 6 Effect of ifenprodil on acute and chronic seizures. (a) Acute seizure parameters (onset, number and duration) in C57BL/6 mice injected
intrahippocampally with HMGB1 (9.5 µg per mouse, n = 8 per group) or vehicle (n = 10 per group) and intraperitoneally with ifenprodil (1 mg per kg body
weight) where indicated. HMGB1 was injected 15 min, and ifenprodil 20 min, before intrahippocampal kainic acid injection. Error bars represent means ±
s.e.m. normalized to the vehicle + KA group (onset 8.7 ± 0.9 min; number of seizures 10.0 ± 1.0; ictal activity 6.6 ± 1.0 min). **P < 0.01 versus vehicle +
KA by two-way ANOVA followed by Tukey’s test. (b) Representative EEG tracings recorded in mice with chronic spontaneous seizures during the pre-injection
baseline, during the first 2 h period after ifenprodil i.p. injection (40 mg per kg body weight) and the following 4-h period (post-injection baseline). Ictal
activity is delimited by arrowheads. (c–h) Ifenprodil effects on chronic spontaneous seizures. Bars represent seizure parameters in each of six chronic epileptic
mice, expressed as percentage of the corresponding pre-injection baseline in the same mouse (raw data in Supplementary Fig. 2). ND, not detectable.
© 2010 Nature America, Inc. All rights reserved.
