The Jou nal o Clinical In es iga ion
RESEARCH ARTICLE
2118 jci.o g Volume 127 Numbe 6 June 2017
In oduc ion
Epilep ogenesis is a dynamic p ocess o molecula , cellula , and
unc ional eo ganiza ion ollowing p ecipi a ing e en s ha leads
o b ain pa hology capable o gene a ing spon aneous seizu es (1).
Cu en ly used an iepilep ic d ugs (AEDs) me ely p o ide symp-
oma ic con ol o seizu es, and a ound 30% o pa ien s ha e epi-
lepsy ha is e ac o y o AEDs (2). The de elopmen o e ec i e
he apies o ea o p e en epilep ogenesis and d ug esis ance
emains an u gen unme clinical need. Clinical ial designs o
no el he apeu ics agains epilep ogenesis a e likely o hinge on
disco e ing nonin asi e bioma ke s ha allow ea ly iden i ica ion
o pa ien s a high isk o de eloping he disease as well as pa ien s
who migh p e e en ially espond o no el ea men s.
Neu oin lamma ion in seizu e-p one b ain egions is a com-
mon ea u e o a ious o ms o d ug- esis an , ocal-onse symp-
oma ic epilepsies in humans and con ibu es o mechanisms o
seizu e gene a ion in animal models (3). S e ile neu oin lamma-
ion is a complex phenomenon in ol ing he ac i a ion o he
inna e immune sys em by damage-associa ed molecula pa e ns
(DAMPs). DAMPs a e eleased by cells unde going s ess ul o
deadly e en s o ale he mic oen i onmen o ac i a e homeo-
s a ic mechanisms o issue epai , bu i DAMPs pe sis in issue,
hey may induce cell dys unc ion o damage (4). High-mobili y
g oup box 1 (HMGB1), a p o o ypical membe o he DAMP am-
ily, is a known media o o s e ile neu oin lamma ion e oked by
epilep ogenic inju ies (4, 5). HMGB1 is a highly conse ed non-
his one nuclea p o ein exp essed by mos euka yo ic cells, whe e
i binds o ch oma in (6) and egula es gene ansc ip ion (7).
HMGB1 has se e al iso o ms, each o which has dis inc physi-
ological and pa hological unc ions. Nonace yla ed HMGB1 is
eleased passi ely om nec o ic cells (7), while ace yla ion o key
lysine esidues in HMGB1 indica es ac i e elease du ing in lam-
ma ion, a s ep ha equi es nucleus- o-cy oplasm ansloca ion
(6, 8). Fu he mo e, he edox s a e o he p o ein de e mines i s
ecep o in e ac ions: unde basal condi ions, HMGB1 is p edom-
inan ly ully educed inside he cell, bu can be oxidized by eac-
i e oxygen species upon ansloca ion o he cy oplasm o a e
i s ex acellula elease. Redox modi ica ion o 3 key cys eine es-
idues, C23, C45, and C106, de e mines he unc ional ac i i y o
HMGB1. In pa icula , disul ide HMGB1, con aining an in amo-
lecula disul ide bond be ween C23 and C45 and a educed C106
(9), speci ically binds and signals ia he TLR4/MD-2 complex
o induce cy okine elease in mac ophages (10), mic oglia, and
as ocy es (5, 11, 12). Reduced HMGB1, in which all 3 cys eines
App oxima ely 30% o epilepsy pa ien s do no espond o an iepilep ic d ugs, ep esen ing an unme medical need. The e
is e idence ha neu oin lamma ion plays a pa hogenic ole in d ug- esis an epilepsy. The high-mobili y g oup box 1
(HMGB1)/TLR4 axis is a key ini ia o o neu oin lamma ion ollowing epilep ogenic inju ies, and i s ac i a ion con ibu es
o seizu e gene a ion in animal models. Howe e , u he wo k is equi ed o unde s and he ole o HMGB1 and i s iso o ms
in epilep ogenesis and d ug esis ance. Using a combina ion o animal models and se a om clinically well-cha ac e ized
pa ien s, we ha e demons a ed ha he e a e dynamic changes in HMGB1 iso o ms in he b ain and blood o animals
unde going epilep ogenesis. The pa hologic disul ide HMGB1 iso o m p og essi ely inc eased in blood be o e epilepsy onse
and p ospec i ely iden i ied animals ha de eloped he disease. Consis en wi h animal da a, we obse ed ea ly exp ession
o disul ide HMGB1 in pa ien s wi h newly diagnosed epilepsy, and i s pe sis ence was associa ed wi h subsequen seizu es. In
con as wi h pa ien s wi h well-con olled epilepsy, pa ien s wi h ch onic, d ug- e ac o y epilepsy pe sis en ly exp essed he
ace yla ed, disul ide HMGB1 iso o ms. Mo eo e , ea men o animals wi h an iin lamma o y d ugs du ing epilep ogenesis
p e en ed bo h disease p og ession and blood inc ease in HMGB1 iso o ms. Ou da a sugges ha HMGB1 iso o ms a e
mechanis ic bioma ke s o epilep ogenesis and d ug- esis an epilepsy in humans, necessi a ing e alua ion in la ge -scale
p ospec i e s udies.
Molecula iso o ms o high-mobili y g oup box 1
a e mechanis ic bioma ke s o epilepsy
Lau en Elizabe h Walke ,1 Fede ica F ige io,2 Te esa Ra izza,2 Emanuele Ricci,1 Ka en Tse,1 Rosalind E. Jenkins,1
G aeme John Sills,1 And ea Jo gensen,1 Luca Po cu,3 Thimmase appa Thippeswamy,4 Tiina Alapi i,5 Jukka Pel ola,5
Ma in J. B odie,6 B ian Ke in Pa k,1 An hony Guy Ma son,1 Daniel James An oine,1 Annama ia Vezzani,2 and Muni Pi mohamed1
1Depa men o Molecula and Clinical Pha macology, Uni e si y o Li e pool, Li e pool, Uni ed Kingdom. 2Depa men o Neu oscience and 3Depa men o Oncology, IRCCS-Is i u o di Rice che
Fa macologiche Ma io Neg i, Milano, I aly. 4Depa men o Biomedical Sciences, College o Ve e ina y Medicine, Iowa S a e Uni e si y, Ames, Iowa, USA. 5Depa men o Neu ology and Rehabili a ion,
Tampe e Uni e si y Hospi al, Tampe e, Finland. 6Epilepsy Uni , Wes e n In i ma y, Glasgow, Uni ed Kingdom.
Au ho ship no e: L.E. Walke and F. F ige io a e co– i s au ho s.
Con lic o in e es : The au ho s ha e decla ed ha no con lic o in e es exis s.
License: This wo k is licensed unde he C ea i e Commons A ibu ion 4.0 In e na ional
License. To iew a copy o his license, isi h p://c ea i ecommons.o g/licenses/by/4.0/.
Submi ed: No embe 29, 2016; Accep ed: Ma ch 16, 2017.
Re e ence in o ma ion: J Clin In es . 2017;127(6):2118–2132.
h ps://doi.o g/10.1172/JCI92001.
Downloaded om h p://www.jci.o g on June 22, 2017. h ps://doi.o g/10.1172/JCI92001
The Jou nal o Clinical In es iga ion RESEARCH ARTICLE
2119
jci.o g Volume 127 Numbe 6 June 2017
cy oplasm be o e he onse o spon aneous seizu es, he e o e
demons a ing ha b ain changes o HMGB1 a e no solely a
consequence o ongoing seizu e ac i i y.
Liquid ch oma og aphy– andem mass spec ome y (LC/
MS-MS) analysis o HMGB1 iso o ms was subsequen ly pe -
o med in bo h hippocampus and blood in a di e en coho o
a s ha we e sac i iced a ep esen a i e ime poin s a e SE, as
in o med by he immunohis ochemical analysis. The p o eoly ic
diges ions o HMGB1 equi ed o LC/MS-MS analysis gene a es
pep ide agmen s ha ei he con ain he ace yla ion si es o he
edox sensi i e sequence (Supplemen al Figu e 3). HMGB1 iso-
o ms in se um ha e been de e mined by a pa en ed absolu e
quan i ica ion me hod (22, 23), which has been ex ensi ely ali-
da ed o bioanaly ical guidelines se ou by he bioma ke con-
so ium unded by he Sa e and Fas e E idence-Based T ansla-
ion Inno a i e Medicines Ini ia i e (SAFE-T IMI) (h p://www.
imi-sa e- .eu/). Since a me hod o absolu e quan i ica ion o
HMGB1 iso o ms has no been alida ed o b ain issue as ye ,
HMGB1 iso o ms in b ain a e exp essed as old change o espec-
i e basal alues. The esul s depic ace yla ed HMGB1 sepa a e-
ly om he educed and disul ide iso o ms. Since educed and
disul ide HMGB1 a e mu ually exclusi e iso o ms, hei amoun s
accoun o o al HMGB1 le els.
We de e mined whe he an associa ion exis s be ween b ain
and blood HMGB1 iso o ms (Figu e 1B and do plo s in Supple-
men al Figu e 4). In b ain lysa e om hippocampal issue, he
ace yla ed iso o m (indica o o ac i e elease) p og essi ely
inc eased be ween 3 hou s and 4 days a e SE, i.e., du ing a ime
ame p eceding he onse o spon aneous seizu es. The edox
s a e o HMGB1 also changed du ing epilep ogenesis: a 3 hou s
a e SE, only educed HMGB1 was o med (3- old inc ease o e
sham alue, no s a is ically signi ican ). Howe e , be ween 6
hou s and 4 days, disul ide HMGB1 was also gene a ed oge h-
e wi h educed HMGB1. In he blood o he same animals, he
ace yla ed and disul ide iso o ms inc eased wi h a delay o a leas
6 hou s when compa ed wi h b ain and be o e spon aneous seizu e
onse . The educed iso o m inc eased in blood concomi an ly wi h
b ain changes and, as in b ain, an icipa ed he gene a ion o disul-
ide HMGB1, which appea ed in blood a 4 days. Wi hin he i s 6
hou s om SE induc ion, ci cula ing HMGB1 was nonace yla ed,
e lec ing i s passi e elease om damaged cells, while a 4 days,
he majo i y o HMGB1 was ace yla ed, indica ing ac i e cellula
elease. I has p e iously been shown ha he ac i e cellula elease
o HMGB1 is h ough a Ca2+-dependen sec e o y lysosome mech-
anism ha , in u n, depends on HMGB1 ace yla ion (24, 25).
These blood changes may e lec ce eb ospinal luid– o–blood
(CSF- o-blood) anspo o HMGB1, which is ini ially (wi hin he
i s 3 hou s) passi ely eleased om damaged neu ons (5, 26) and
possibly also om BBB endo helium (27, 28). In pa icula , a 3
hou s a e SE, nonace yla ed and educed HMGB1 in blood o e -
sho he b ain changes by abou 2- old, sugges ing an addi ional
cellula sou ce o HMGB1 o he han he b ain pa enchyma, such
as he b ain endo helium. In acco dance, endo helial cells showed
inc eased exp ession o HMGB1 du ing epilep ogenesis (Figu e
1A). Fu he , i has p e iously been demons a ed ha endo helial
cells can be damaged by SE (29), sugges ing ha hese cells may
se e as a sou ce o HMGB1.
a e educed, o ms a he e ocomplex wi h he C-X-C mo i chemo-
kine 12 (CXCL12) and binds CXCR4 o ini ia e chemo axis (13).
HMGB1 ha is e minally oxidized o con ain sul onyl g oups on
all cys eines (sul onyl HMGB1) does no a ec cell mig a ion o
cy okine induc ion (14). The disul ide and educed iso o ms ha e
mu ually exclusi e unc ions. HMGB1 also ac i a es ecep o o
ad anced glyca ion end p oduc s (RAGE), bu he binding e i-
ciency o each iso o m o RAGE is s ill un esol ed.
We ha e shown ha o al HMGB1 and i s nucleus- o-
cy oplasm ansloca ion inc ease in neu ons and glia in human
d ug- esis an epilep ic oci (5, 12) and he co esponding animal
models (5) and ha disul ide HMGB1 is he iso o m ha p o-
mo es seizu es and cell loss (5, 15, 16). Fu he mo e, mice lacking
TLR4 o RAGE a e in insically less suscep ible o seizu es and
less p one o de eloping epilepsy (5, 15), sugges ing ha HMGB1
is implica ed in epilep ogenesis and ha he e o e, i s a ge ing
may ha e he apeu ic u ili y. Howe e , he ole o HMGB1 and
i s di e en iso o ms in bo h epilep ogenesis and d ug esis ance,
including whe he hey can be used as nonin asi e, mechanis ic
bioma ke s o epilep ogenesis and ea men esponse, is unclea .
In o de o add ess his, we ha e measu ed o al HMGB1 and i s
ace yla ed, educed, and disul ide iso o ms be o e epilepsy onse
and du ing disease de elopmen in animal models o acqui ed
epilepsy, co ela ing b ain changes o hose in blood, and ha e
unde aken b idging s udies in well-de ined pa ien s wi h d ug-
e ac o y epilepsy and in newly diagnosed pa ien s.
Resul s
B ain exp ession and blood le els o HMGB1 in oden models o
epilepsy. Ou i s objec i e was o es ablish whe he a ela ion-
ship exis s be ween b ain and blood le els o HMGB1 du ing he
acu e inju y and ea ly epilep ogenesis ollowing s a us epilep i-
cus (SE), a b ain insul leading o epilepsy in animal models and
humans (17, 18). We in es iga ed whe he HMGB1 ansloca es
om he nucleus o he cy oplasm du ing epilep ogenesis (Fig-
u e 1A), since his s ep is equi ed o i s cellula elease, he con-
sequen biological e ec s (19), and i s po en ial b ain- o-blood
ans e . Cellula localiza ion o HMGB1 was assessed immu-
nohis ochemically in he hippocampus o a s exposed o elec-
ically induced SE, a well-cha ac e ized model o epilepsy (20,
21), and hese hippocampi we e compa ed wi h sham con ols
(see Supplemen al Me hods; supplemen al ma e ial a ailable
online wi h his a icle; h ps://doi.o g/10.1172/JCI92001DS1).
In con ol issue, HMGB1 was exp essed in nuclei o neu ons and
glia (Figu e 1A and Supplemen al Figu e 1). F om 3 hou s un il
4 days a e SE, a ime window p eceding he onse o epilepsy
in his model, as shown by he lack o EEG and mo o seizu es
in a s unde con inuous ideo-EEG eco ding (spon aneous sei-
zu es a ise a 8.4 ± 1.8 days, see below), HMGB1 s aining showed
a p og essi e inc ease in GFAP-posi i e as ocy es (Figu e 1A
and Supplemen al Figu e 2). No ably, in all SE-exposed a s,
HMGB1 s aining was obse ed in he cy oplasm o glial cells,
including ac i a ed OX-42–posi i e mic oglia, and in endo heli-
al ba ie an igen–posi i e (EBA-posi i e) endo helial cells o he
blood b ain ba ie (BBB) (Figu e 1A) and was de ec able in he
cy oplasm o neu ons in 30% o animals (Supplemen al Figu e
1). These da a show ha HMGB1 ansloca es om nucleus o
Downloaded om h p://www.jci.o g on June 22, 2017. h ps://doi.o g/10.1172/JCI92001
The Jou nal o Clinical In es iga ion
RESEARCH ARTICLE
2120 jci.o g Volume 127 Numbe 6 June 2017
Downloaded om h p://www.jci.o g on June 22, 2017. h ps://doi.o g/10.1172/JCI92001
The Jou nal o Clinical In es iga ion RESEARCH ARTICLE
2121
jci.o g Volume 127 Numbe 6 June 2017
i enp odil (a blocke o he NR2B-con aining NMDA ecep o
[NMDAR], e . 34; 20 mg/kg, i.p.) ( ea men p o ocol in Me h-
ods). This ea men combina ion was designed o e icien ly
block he ac i a ion o he IL-1R1/TLR4 pa hway, a known con-
ibu o o seizu e mechanisms (4), as well as a key molecula a -
ge ( he NR2B-NMDAR) ha media es inc eased neu onal exci -
abili y induced by pa hway ac i a ion (16, 35). In ac , ou p e ious
e idence showed ha he single blockade o he IL-1R1 axis was
no su icien o p e en TLR4 signaling ac i a ion du ing epilep-
ogenesis (20) and ha TLR4 signaling i sel con ibu es o epi-
lep ogenesis (15). A g oup o ehicle-injec ed a s implan ed wi h
elec odes bu no exposed o SE was used as a sham con ol o
ime-ma ched blood HMGB1 measu emen s (n = 9).
We ound a p og essi e inc ease o bo h o al HMGB1 and i s
ace yla ed and disul ide iso o ms ( he la e being unde ec able in
sham a s) in blood du ing disease de elopmen in ehicle-injec -
ed a s; no ably, hese changes we e de ec ed be o e he onse o
he spon aneous seizu es (i.e., disease onse ), as assessed in a s
con inuously moni o ed by ideo-EEG, and pe sis ed un il he
ch onic phase o spon aneous seizu es (Figu e 2 and do plo s in
Supplemen al Figu e 7). Be o e disease onse , ace yla ed HMGB1
was he p edominan iso o m (7 days, mean 70% o o al HMGB1)
e lec ing ac i e elease. This iso o m pla eaued ea ly a e dis-
ease onse (14 days, mean 50%) and inc eased again in he ch onic
disease phase (4.5 mon hs, mean 80%). Reduced HMGB1 s ably
inc eased a ound 4- old h oughou he obse a ion pe iod (Fig-
u e 2; do plo s in Supplemen al Figu e 7). The inac i e sul onyl
HMGB1 was no de ec ed a any ime poin .
The an iin lamma o y d ug combina ion signi ican ly a enu-
a ed he inc ease in o al HMGB1 and i s iso o ms du ing he p o-
d omal phase and ab oga ed blood HMGB1 changes a e disease
onse (Figu e 2; do plo s in Supplemen al Figu e 7). No ably, he
d ug combina ion also p e en ed he inc ease in seizu e equency
de e mined by EEG analysis be ween 2.0 and 4.5 mon hs a e SE in
ehicle-injec ed a s (Figu e 3; Supplemen al Table 1), hus indica -
ing ha he an iin lamma o y d ugs p e en ed disease p og ession.
An addi ional salu a y e ec o ea men was p e en ion o body
weigh loss occu ing in SE-exposed a s (Supplemen al Figu e 5B).
The d ug combina ion did no modi y he du a ion o se e i y
o SE (Supplemen al Figu e 6, A and B) and he consequen neu o-
degene a ion (Supplemen al Figu e 8 and Supplemen al Me hods)
o he ime o spon aneous seizu e onse (SE+ ehicle, 8.4 ± 1.8
days; SE+d ug, 11.6 ± 1.7 days, n = 9 each g oup). Recei e ope a -
ing cha ac e is ic (ROC) analysis (Table 1) showed ha bo h o al
HMGB1 and i s iso o ms eliably di e en ia ed be ween sham
con ol and epilep ic a s (AUC = 1) in he ch onic phase o spon-
aneous seizu es, suppo ing hei alue as diagnos ic bioma ke s
The p og essi e ac i e elease o ace yla ed HMGB1 may o ig-
ina e mainly om as ocy es, as sugges ed by bo h he p ominen
as ocy ic HMGB1 exp ession and i s cy oplasmic ansloca ion by
immunohis ochemis y (Figu e 1A and Supplemen al Figu e 2).
No ably, he b ain changes in bo h ace yla ed and disul ide
iso o ms p eceded he espec i e blood inc eases, suppo ing a
CSF- o-blood anspo . We canno exclude, howe e , ha leuko-
cy es con ibu e o blood HMGB1 changes. Bo h he hypo halam-
ic-pi ui a y-ad enal (HPA) axis and he agal ne e can media e
signaling o he in lamma o y s a e in he b ain o he blood, and
his phenomenon could p omo e he elease o HMGB1 om ci -
cula ing leukocy es. I should be no ed, howe e , ha hese b ain-
o-blood signals usually esul in pe iphe al an iin lamma o y
( a he han p oin lamma o y) esponses (30).
To exclude model-speci ic changes, we also measu ed
HMGB1 in blood o adul male C57BL/6J mice exposed o SE
induced by sys emic kaina e (31). As in a s, we ound on a e age
a 3- old inc ease in educed HMGB1 be ween 3 hou s and 6 hou s
a e SE, while bo h he ace yla ed and disul ide iso o ms we e
unchanged (no shown).
HMGB1 as a bioma ke in animal models. To in es iga e he al-
ue o blood HMGB1 in p edic ing epilepsy de elopmen (Figu e
2 and Supplemen al Figu e 5A), we used he elec ically induced
SE model (also used in Figu e 1) in which all a s de elop epilepsy
o es whe he blood HMGB1 le els change be o e he onse o
he spon aneous seizu es independen ly o ongoing seizu e ac i -
i y. We measu ed blood HMGB1 longi udinally a 3 imes a e SE
ha a e ep esen a i e o disease de elopmen : 7 days (p eceding
spon aneous seizu e onse ), 14 days (ea ly a e disease onse ), and
4.5 mon hs (ch onic epilepsy phase). Using he same model, we
also es ed he u ili y o blood HMGB1 o p edic ing a he apeu ic
esponse o an iin lamma o y d ugs. Ra s exposed o elec ical SE
we e andomized in 1 ea men g oup (n = 9) injec ed wi h a com-
bina ion o an iin lamma o y d ugs du ing epilep ogenesis and 1
placebo g oup (n = 9), which ecei ed he co esponding ehicles.
D ugs o hei ehicles we e injec ed in o a s o 1 week s a ing
1 hou a e SE induc ion (Supplemen al Figu e 5A and Supple-
men al Figu e 6, A and B) o encompass he epilep ogenesis phase
p eceding disease onse ; hen ea men was s opped. Ra s we e
ideo-EEG moni o ed con inuously om he induc ion o SE un il
he i s spon aneous seizu es (i.e., epilepsy onse ). Spon aneous
seizu es we e coun ed o 2 weeks a 2.0 and 4.0 mon hs as he p i-
ma y ou come measu e. Based on ou backg ound e idence (5, 15,
20), we used a a ional combina ion o an iin lamma o y d ugs,
anakin a, ( he human ecombinan IL-1 ecep o an agonis , e .
32; 10 mg/200 μl/ a , i. .) and BoxA (a agmen o HMGB1 wi h
an agonis ac i i y, e . 33; 100 μg/200 μl/ a , i. .) oge he wi h
Figu e 1. B ain and blood HMGB1 measu emen s du ing epilep ogenesis e oked by elec ical SE in adul a s. (A) Rep esen a i e pho omic og aphs
o hippocampi om con ol a s (sham) o a s a 3 hou s, 6 hou s, and 4 days a e SE (n = 5 each g oup). Top ow shows HMGB1 immuno eac i i y in
cell nuclei (a ows) o in cy oplasm o glial cells (a owheads). Immuno luo escence panels show localiza ion o HMGB1 signal (g een) in OX-42–posi i e
mic oglia ( ed), GFAP-posi i e as ocy es ( ed), and EBA-posi i e endo helial cells ( ed); colocaliza ion signal is depic ed in yellow (me ge). Whi e a ows
depic cy oplasmic s aining. Hoechs -posi i e nuclei a e shown in blue. Rad, s a um adia um; LMol, s a um lacunosum molecula e. Scale ba s: 25 μm
( op ow); 20 μm (bo om ow; immuno luo escence panels). (B) Le els o HMGB1 iso o ms in b ain issue (hippocampus) and co esponding blood o a s
du ing epilep ogenesis. Da a a e shown as mean ± SEM (n = 5 each g oup). Do plo s a e shown in Supplemen al Figu e 4. *P < 0.05; **P < 0.01, K uskal-
Wallis es ( e e ed o bo h iso o ms in each ba ). Blood ace yla ed and disul ide HMGB1 le els a 4 days a e signi ican ly di e en om co esponding
3-hou and 6-hou le els (P < 0.01).
Downloaded om h p://www.jci.o g on June 22, 2017. h ps://doi.o g/10.1172/JCI92001
The Jou nal o Clinical In es iga ion
RESEARCH ARTICLE
2122 jci.o g Volume 127 Numbe 6 June 2017
simila se e i y (Supplemen al Figu e 6C) and du a ion (26.3 ± 0.5
hou s). Blood was d awn longi udinally a e SE a 23 days (p o-
d omal phase), 73 days (encompassing he ime o disease onse
in 70% o a s), and 7.5 mon hs (ch onic epilepsy phase) (Sup-
plemen al Figu e 5C). In he ch onic epilepsy phase, we coun ed
spon aneous seizu es du ing 2 weeks o con inuous ideo-EEG
moni o ing: epilep ic a s expe ienced 3.7 ± 0.5 clinical seizu es,
while no seizu es we e obse ed in nonepilep ic a s. Epilep ic a s
also displayed a signi ican educ ion in he h eshold o a e dis-
cha ge (ADT) induc ion (116 ± 38 μA, n = 12; P < 0.05) as compa ed
wi h sham con ols (220 ± 40 μA, n = 7); nonepilep ic a s wi h-
ou spon aneous seizu es we e simila o con ols (190 ± 40 μA,
n = 5). Fo HMGB1 blood measu emen , we andomly selec ed
5 epilep ic a s (4.0 ± 1.3 seizu es/2 weeks) and compa ed hem
wi h 5 nonepilep ic a s and 7 ehicle- ea ed sham con ols.
Only a s who de eloped epilepsy had inc eased blood le els
o ac i e epilepsy. Addi ionally, o al HMGB1 and i s iso o ms also
p edic ed wi h high ideli y (AUC = 0.8–1 ange) he he apeu ic
esponse o an iin lamma o y d ugs be o e disease onse and a
epilepsy diagnosis.
The e was no co ela ion be ween he numbe o ch onic sei-
zu es in he 2-week eco ding pe iod p eceding blood wi hd awal
(4.5 mon hs a e SE) and he co esponding le el o o al blood
HMGB1 (Supplemen al Figu e 9A).
Since he changes in blood HMGB1 we e de ec ed in a s
be o e he onse o he disease, we con inued ou in es iga ions
o alida e he po en ial ole o HMGB1 o p edic ing epilepsy
de elopmen . In a subsequen s udy, we e oked SE by sys emic
li hium+piloca pine in p eadolescen (P21) a s (Supplemen al
Figu e 5C), since only a ound 70% o a s de elop epilepsy a e
an a e age ime o 70 days (36–40). In his model, 12 ou o 17
a s de eloped spon aneous seizu es a e being exposed o SE o
Figu e 2. Blood HMGB1 le els inc ease be o e epilepsy onse in elec ical SE–exposed a s and p edic he apeu ic esponse o an iin lamma o y d ugs.
Longi udinal analysis o o al HMGB1 and le els o ace yla ed, educed, and disul ide iso o ms in blood plasma a ep esen a i e ime poin s o disease de el-
opmen in SE-exposed a s ecei ing ea men (d ug) o co esponding ehicle (see key). T ea men included anakin a+BoxA+i enp odil (SE+d ug) (de ailed
p o ocol in Me hods; Supplemen al Figu e 5A). Da a a e shown as mean ± SEM (n = 9 a s each g oup; do plo s a e shown in Supplemen al Figu e 7). Ra s a e
he same as epo ed in Figu e 3. *P < 0.05; **P < 0.01, 1-way ANOVA ( e e ed o bo h iso o ms in each ba , excep o p od omal phase, whe e
P < 0.01 in SE+d ug e sus SE+ ehicle e e s only o educed HMGB1). The ace yla ed iso o m le el in he ch onic epilepsy phase (SE+ ehicle) was signi ican ly
di e en om co esponding le els a disease onse and p od omal phases (P < 0.01, epea ed measu es 1-way ANOVA). The disul ide iso o m (SE+ ehicle)
le el a disease onse and in he ch onic epilepsy phase was signi ican ly di e en om he co esponding le el in he p od omal phase (P < 0.05 and P < 0.01,
espec i ely); he disul ide iso o m le el in he ch onic phase was signi ican ly di e en om he co esponding le el a disease onse (P < 0.05, epea ed
measu es 1-way ANOVA).
Downloaded om h p://www.jci.o g on June 22, 2017. h ps://doi.o g/10.1172/JCI92001

The Jou nal o Clinical In es iga ion RESEARCH ARTICLE
2123
jci.o g Volume 127 Numbe 6 June 2017
o bo h o al HMGB1 and i s iso o ms; hese changes occu ed
be o e he onse o he disease and we e main ained du ing dis-
ease de elopmen (Figu e 4A; do plo s in Supplemen al Figu e
10). Nonepilep ic a s we e simila o con ols a all ime poin s
a e SE (Figu e 4A; do plo s in Supplemen al Figu e 10). The
inac i e sul onyl HMGB1 iso o m was no de ec ed unde any
expe imen al condi ions. Mo eo e , pos mo em b ain analy-
sis showed ha only epilep ic a s displayed inc eased HMGB1
exp ession in he hippocampus a 7.5 mon hs wi h concu en
nucleus- o-cy oplasm ansloca ion in glial cells, while none-
pilep ic a s we e simila o con ols (Figu e 4B). ROC analysis
showed pe ec disc imina ion be ween epilep ic and nonep-
ilep ic a s (AUC = 1; Table 2), wi h bo h o al HMGB1 and i s
iso o ms measu ed ei he be o e he onse o spon aneous sei-
zu es (Supplemen al Figu e 11), a he onse o disease, o in he
ch onic epilepsy phase (Table 2). The blood changes in HMGB1
did no me ely e lec eac i e gliosis, since he hippocampal
le el o myo-inosi ol, a me aboli e e lec ing as ocy e ac i a-
ion, was simila ly augmen ed (P < 0.05) in all a s ei he wi h o
wi hou HMGB1 blood inc ease, as assessed by p o on magne ic
esonance spec oscopy (1H-MRS) 23 days a e SE a he ime
o blood wi hd awal (Supplemen al Figu e 12A). Mo eo e , he
le el o gliosis assessed in hese a s 7.5 mon hs a e SE by pos -
mo em glial ib illa y acidic p o ein (GFAP) immunos aining
was simila in he hippocampus o a s wi h epilepsy (showing
ele a ed blood HMGB1) o wi hou epilepsy (wi h no ele a ion o
blood HMGB1) (Supplemen al Figu e 12B).
Hippocampal cell loss in CA3 py amidal cell laye and hilus
was de ec ed in epilep ic a s (n = 5) e sus sham con ols (no
exposed o SE, n = 7): numbe o cells, CA3, sham, 546 ± 54; epi-
lep ic, 454.5 ± 13 (P < 0.05); hilus, sham, 93.9 ± 7.1; epilep ic, 38.9
± 2.9 (P < 0.01). Nonepilep ic a s had a pa e n and ex en o cell
loss simila o hose in epilep ic a s (no shown).
HMGB1 allows s a i ica ion o pa ien s wi h pha maco esis an
epilepsy. Because HMGB1 was inc eased in bo h b ain and blood
in a s wi h ac i e epilepsy and i s le els we e modula ed by he -
apeu ic in e en ions ha educed seizu es, we hypo hesized ha
o al HMGB1 and i s iso o ms may be used as bioma ke s o d ug
esis ance in human epilepsy. We examined HMGB1 le els in se a
om 65 pa ien s wi h d ug- esis an epilepsy unde going con-
inuous inpa ien ideo-EEG moni o ing, 20 pa ien s wi h well-
con olled epilepsy (diagnosed mo e han 12 mon hs p e iously,
seizu e- ee o mo e han 6 mon hs on AED he apy) includ-
ed as d ug- esponsi e epilepsy con ols, and 74 heal hy con ols
(Supplemen al Tables 2–4). Pa ien s wi h d ug- esis an epilepsy
exp essed signi ican ly highe le els o o al HMGB1 (8.70 ± 0.47
ng/ml) han bo h heal hy con ols (1.11 ± 0.07 ng/ml, P < 0.01) and
hose wi h well-con olled epilepsy (1.25 ± 0.15 ng/ml, P < 0.01;
Figu e 5A). Pa ien s wi h abno mal b ain MRI showed signi ican -
ly highe HMGB1 le els (mean ± SEM, 9.8 ± 0.7 ng/ml, n = 35)
Table 1. Recei e ope a ing cha ac e is ic analysis ela ed o Figu e 2
Vehicle e sus ea men
P od omal phase
Vehicle e sus ea men
Disease onse
Sham e sus epilep ic
Ch onic phase
HMGB1 AUC P alue Sensibili y
a 95%
speci ici y
Cu -o
(ng/ml)
AUC P alue Sensibili y
a 95%
speci ici y
Cu -o
(ng/ml)
AUC P alue Sensibili y
a 95%
speci ici y
Cu -o
(ng/ml)
To al 0.9 0.0052 0.5 (0.2–0.8) 1.1 1 0.0005 1 (0.6–1) 10.5 1 0.0002 1 (0.7–1) 3
Ace yla ed 0.9 0.0028 0.9 (0.5–1) 3.9 1 0.0005 1 (0.6–1) 5.6 1 0.0002 1 (0.7–1) 1.5
Reduced 0.9 0.0028 0.9 (0.5–1) 3.9 1 0.0005 1 (0.6–1) 3.7 1 0.0002 1 (0.7–1) 2.2
Disul ide 0.8 0.0833 0.5 (0.2–0.8) 1.1 1 0.0005 1 (0.6–1) 3.7 1 0.0002 1 (0.7–1) 3
Blood HMGB1 le el disc imina es be ween con ol and epilep ic a s in he ch onic epilepsy phase and be ween ehicle- and d ug- ea ed a s du ing
epilep ogenesis (p od omal phase) and a disease onse . The cu -o alue o o al HMGB1 and co esponding iso o m concen a ion (sensi i i y a 95%
speci ici y) is epo ed o each expe imen al condi ion.
Table 2. Recei e ope a ing cha ac e is ic analysis ela ed o Figu e 4A
Epilep ic e sus nonepilep ic
P od omal phase
Epilep ic e sus nonepilep ic
Disease onse
Epilep ic e sus nonepilep ic
Ch onic phase
HMGB1 AUC P alue Sensibili y
a 95%
speci ici y
Cu -o
(ng/ml)
AUC P alue Sensibili y
a 95%
speci ici y
Cu -o
(ng/ml)
AUC P alue Sensibili y
a 95%
speci ici y
Cu -o
(ng/ml)
To al 1 0.0143 1 (0.4–1) 4.5 1 0.009 1 (0.5–1) 5.8 1 0.009 1 (0.5–1) 9.4
Ace yla ed 1 0.0143 1 (0.4–1) 0.8 1 0.009 1 (0.5–1) 1.4 1 0.009 1 (0.5–1) 6.2
Reduced 1 0.0143 1 (0.4–1) 3.1 1 0.009 1 (0.5–1) 3.4 1 0.009 1 (0.5–1) 3.5
Disul ide 1 0.0143 1 (0.4–1) 0.7 1 0.009 1 (0.5–1) 2.2 1 0.009 1 (0.5–1) 4.8
Blood HMGB1 le el disc imina es be ween epilep ic and nonepilep ic a s du ing disease de elopmen . The cu -o alue o o al HMGB1 and co esponding
iso o m concen a ion (sensi i i y a 95% speci ici y) is epo ed o each expe imen al condi ion.
Downloaded om h p://www.jci.o g on June 22, 2017. h ps://doi.o g/10.1172/JCI92001
The Jou nal o Clinical In es iga ion
RESEARCH ARTICLE
2124 jci.o g Volume 127 Numbe 6 June 2017
han hose wi hou imaging abno mali ies (7.4 ± 0.5 ng/ml, n = 30,
P < 0.01; Supplemen al Figu e 13). No ably, ace yla ed HMGB1
was obse ed in d ug- esis an epilepsy pa ien s, bu no in con-
ols o d ug- esponsi e pa ien s. Pa ien s wi h d ug- esis an epi-
lepsy could be u he s a i ied on he basis o he p esence o he
disul ide iso o m (Figu e 5A). The e was no signi ican associa ion
be ween he p esence o disul ide HMGB1 and any o he clinical
a iables (Supplemen al Table 2). Fu he mo e, he e was no co -
ela ion be ween he numbe o seizu es in he p eceding mon h
and he le el o o al HMGB1 in blood (Supplemen al Figu e 9B).
Fo he compa ison be ween con ol alues and hose ob ained
om pa ien s ha we e d ug esis an , a o al ROC-AUC o 0.99
(P = 0.0001) was eco ded (Figu e 5B). The sensi i i y a 95%
speci ici y ga e an ROC-AUC o 0.94 (95% CI 0.85–0.98) wi h
a cu -o HMGB1 concen a ion o 2.3 ng/ml. Fo he compa i-
son be ween alues om pa ien s wi h well-con olled epilepsy
and hose ob ained om pa ien s ha a e d ug esis an , a o al
ROC-AUC o 0.99 (P = 0.0001) was eco ded (Figu e 5B). The
sensi i i y a 95% speci ici y ga e an ROC-AUC o 0.93 (95% CI
0.85–0.99), which also had a cu -o o HMGB1 concen a ion o
2.3 ng/ml ( o alues used o de e mine ROC).
HMGB1 in newly diagnosed epilepsy. In a pilo s udy, we exam-
ined HMGB1 iso o ms in he pe iphe al blood o 27 pa ien s wi h
newly diagnosed epilepsy. The ace yla ed and disul ide iso o ms
o HMGB1 (DS-A) we e de ec able a diagnosis (i.e., baseline isi )
in 25 o 27 pa ien s (median, 2.00 ng/ml; ange, 0.0-8.10), wi h 2
pa ien s alling below he quan i iable limi . The e was no signi -
ican co ela ion be ween he o al p e ea men seizu e coun a
diagnosis and he co esponding amoun o HMGB1 p esen in
blood (Supplemen al Figu e 14). No was he e any signi ican ela-
ionship be ween he numbe o days since he las seizu e and he
o al amoun o blood HMGB1 p esen (Supplemen al Figu e 15).
A he 6-week ollow-up isi , 13 o 27 pa ien s we e seizu e
ee (median DS-A, 0.8 ng/ml; ange, 0.0–7.40). The emaining
14 pa ien s who had expe ienced subsequen seizu es had sig-
ni ican ly highe DS-A le els (median DS-A, 3.25 ng/ml; ange,
0.0–10.60, P = 0.04), wi h no pa ien s ha ing unde ec able le els.
Fi e o he seizu e- ee pa ien s had unde ec able DS-A le els a
6 weeks; 4 o hese pa ien s emained seizu e ee a 6 mon hs,
while a u he pa ien was also seizu e ee, bu no sample was
a ailable o analysis.
By 6 mon hs, 5 pa ien s we e los o ollow-up. Ele en pa ien s
we e seizu e ee (median DS-A, 0.0 ng/ml; ange, 0.0–8.70); in 7
o he pa ien s, DS-A le els we e unde ec able. The emaining 11
pa ien s had subsequen seizu es, and hey had signi ican ly highe
DS-A le els (median DS-A, 8.30 ng/ml; ange, 0.0–8.90, P = 0.02).
An addi ional 2 pa ien s ec ui ed o he s udy had hei epilep-
sy diagnosis e ised o nonepilep ic a ack diso de (NEAD). DS-A
HMGB1 iso o m was absen in hese 2 pa ien s. I was no de ec -
able in he se a o heal hy human subjec s (41) o pa ien s wi h
well-con olled ch onic epilepsy who we e seizu e ee (Figu e 5A).
Discussion
Ou comp ehensi e assessmen shows ha ci cula ing iso o ms
o HMGB1 may se e as mechanis ic bioma ke s o epilep ogene-
sis in animal models as well as d ug- esis an epilepsy in humans.
Elec ically induced SE in adul oden s inc eased nonace yla ed
and educed HMGB1 iso o ms in blood wi hin hou s. This is con-
sis en wi h passi e elease o HMGB1 om inju ed neu onal and
endo helial cells (5, 9), which is known o occu in he hippocam-
pus a e SE (26, 27). Following a di e en piloca pine-induced
epilep ogenic insul , we also iden i ied an ea ly inc ease in nonac-
e yla ed and educed HMGB1 in he blood o a s ha de eloped
epilepsy, bu no in hose who did no , despi e simila neu onal
cell loss. Thus, his e idence sugges s ha he passi e elease o
HMGB1 ollowing SE no only e lec s cell dea h, bu may also be
a ma ke o he o e all s a e o cell s ess in neu ons, glia, and BBB
endo helium induced by he epilep ogenic insul .
Figu e 3. An iin lamma o y d ugs adminis e ed du ing epilep ogenesis p e en disease p og ession in elec ical SE–exposed a s. (A) P og ession index
( he a io be ween he numbe o seizu es a 4.5 mon hs and a 2.5 mon hs a e SE) in ehicle- and d ug- ea ed a s. Value equi alen o 1 de ines lack
o p og ession. Da a a e shown as mean ± SEM (n = 9 a s each g oup). *P < 0.05 e sus ehicle, Mann-Whi ney U es . Raw da a a e shown in Supplemen-
al Table 1. (B) Odds a io (0.24; P = 0.008) indica es ha d ug- ea ed a s e alua ed a 2.5 mon hs ha e a 76% educed isk o de eloping seizu es a 4.5
mon hs e sus ehicle- ea ed a s. T ea men included anakin a+BoxA+i enp odil (SE+d ug) (see Me hods; Supplemen al Figu e 5A). OR, odds a io.
Downloaded om h p://www.jci.o g on June 22, 2017. h ps://doi.o g/10.1172/JCI92001
The Jou nal o Clinical In es iga ion RESEARCH ARTICLE
2125
jci.o g Volume 127 Numbe 6 June 2017
blood leakage due o BBB damage (27) o CSF- o-blood di usion
(42), al hough we canno exclude a con ibu ion o pe iphe al leu-
kocy es o HMGB1 blood changes (43). To al HMGB1 and all i s
iso o ms inc eased in blood be o e he onse o spon aneous sei-
zu es, and hese changes pe sis ed only in animals wi h ac i e epi-
lepsy. Impo an ly, blood HMGB1 p edic ed which animal would
de elop he disease wi h high ideli y (AUC = 1). On ha basis, i is
A la e ime poin s o epilep ogenesis, no ably be o e he onse
o spon aneous seizu es and du ing disease de elopmen in he 2
animal models, bo h ace yla ed and disul ide iso o ms p og es-
si ely inc eased and accoun ed o he majo i y o o al HMGB1
blood le els. This phenomenon is consis en wi h ac i e elease
o HMGB1 om ac i a ed glial cells, as suppo ed by immunohis-
ochemis y, and i may he e o e e lec ei he HMGB1 b ain- o-
Figu e 4. Ea ly p edic ion o epilepsy de elopmen by moni o ing blood HMGB1 le el in li hium+piloca pine SE a s. (A) Longi udinal analysis o o al
HMGB1 and ace yla ed, educed, and disul ide iso o m le els in blood a ep esen a i e ime poin s o disease de elopmen (see key in A). Blood was d awn
a 23 days (epilep ogenic phase p od omal o epilepsy onse ), 73 days (encompassing he ime o disease onse in 70% o a s), and 7.5 mon hs (ch onic
epilepsy). Da a ep esen mean ± SEM; n = 7 sham; n = 5 epilep ic (Epi); n = 5 nonepilep ic (Nonepi) a s. Do plo s a e shown in Supplemen al Figu e 10.
**P < 0.01 ( e e s o bo h iso o ms in each ba ); P < 0.05, ace yla ed HMGB1 in epilep ic e sus sham, K uskal-Wallis es . To al HMGB1, ace yla ed and disul-
ide iso o m le els in he ch onic phase (Epi) a e signi ican ly di e en e sus co esponding le els in p od omal phase (P < 0.01, epea ed measu es 1-way
ANOVA). (B) Rep esen a i e immunohis ochemical pic u es o CA1 s a um adia um depic ing HMGB1 s aining in con ol (sham, n = 7) and in SE a s wi h
(Epi, n = 5 ou o 12) o wi hou (Nonepi, n = 5) spon aneous seizu es. Ra s a e he same as epo ed in A. B ains we e ha es ed 7.5 mon hs a e SE. A ows,
nuclea s aining o HMGB1 in sham and nonepilep ic a s; a owheads, cy oplasmic HMGB1 s aining in glia in epilep ic a s. Scale ba : 15 μm.
Downloaded om h p://www.jci.o g on June 22, 2017. h ps://doi.o g/10.1172/JCI92001
The Jou nal o Clinical In es iga ion
RESEARCH ARTICLE
2126 jci.o g Volume 127 Numbe 6 June 2017
ma o y cy okine elease om immune cells by ac i a ing TLR4
(6, 9) and con ibu es o hype exci abili y and seizu es in animal
models by modula ing Ca2+ pe meabili y o NR2B-NMDAR (5, 15,
16). Addi ionally, TLR4 has been implica ed in epilep ogenesis
(15). The e o e, he elease o disul ide HMGB1 in he b ain ol-
lowing SE may con ibu e o epilep ogenesis and he consequen
onse o spon aneous seizu es (pe sonal communica ions, Vezza-
ni). Acco dingly, disul ide HMGB1 was gene a ed in he a b ain
possible ha blood HMGB1 migh iden i y pa ien s wi h he high-
es isk o de eloping spon aneous seizu es ea ly a e a po en ial-
ly epilep ogenic inju y. This hypo hesis can be es ed in p ospec-
i e human clinical s udies, such as in pos s oke o pos auma ic
epilepsy, by examining he p ognos ic alue o o al HMGB1 and
i s iso o ms be o e he i s p esen a ion o seizu es.
While educed HMGB1 has chemoa ac an p ope ies o
mac ophage/mic oglia (6, 9), disul ide HMGB1 induces p oin lam-
Figu e 5. HMGB1 le els in blood o d ug- esis an and d ug-sensi i e epilepsy pa ien s and heal hy con ols. (A) To al HMGB1 le els in heal hy con ols
(1.1 ± 0.07 ng/ml, n = 74) and pa ien s wi h well-con olled epilepsy (1.2 ± 0.15 ng/ml, n = 20) o d ug- esis an epilepsy (8.7 ± 0.47 ng/ml, n = 65; P < 0.01).
Ace yla ed HMGB1 was de ec able only in pa ien s wi h d ug- esis an epilepsy, and hese indi iduals could be u he s a i ied on he basis o he p es-
ence o disul ide HMGB1 (d ug- esis an , educed HMGB1 7.1 ± 0.4 ng/ml, n = 41; educed+disul ide, 11.0 ± 0.9 ng/ml, n = 24; P < 0.01). Da a a e p esen ed in
each g oup as indi idual alues as well as he mean ± SEM. **P < 0.01, 1-way ANOVA. (B) ROC analysis shows ha o al HMGB1 disc imina es (AUC = 0.99)
be ween pa ien s wi h d ug- esis an seizu es and bo h heal hy con ols and hose wi h well-con olled seizu es.
Downloaded om h p://www.jci.o g on June 22, 2017. h ps://doi.o g/10.1172/JCI92001