Attenuation of Allergic Contact Dermatitis Through the Endocannabinoid System

Abstract

Allergic contact dermatitis affects about 5% of men and 11% of women in industrialized countries and is one of the leading
causes for occupational diseases. In an animal model for cutaneous contact hypersensitivity, we show that mice lacking both
known cannabinoid receptors display exacerbated allergic inflammation. In contrast, fatty acid amide hydrolase–deficient mice,
which have increased levels of the endocannabinoid anandamide, displayed reduced allergic responses in the skin. Cannabinoid
receptor antagonists exacerbated allergic inflammation, whereas receptor agonists attenuated inflammation. These results demonstrate
a protective role of the endocannabinoid system in contact allergy in the skin and suggest a target for therapeutic intervention.

Full text

DOI: 10.1126/science.1142265
, 1494 (2007); 316Science
et al.Meliha Karsak,
the Endocannabinoid System
Attenuation of Allergic Contact Dermatitis Through
www.sciencemag.org (this information is current as of June 8, 2007 ):
The following resources related to this article are available online at
http://www.sciencemag.org/cgi/content/full/316/5830/1494
version of this article at:
including high-resolution figures, can be found in the onlineUpdated information and services,
http://www.sciencemag.org/cgi/content/full/316/5830/1494/DC1
can be found at: Supporting Online Material
http://www.sciencemag.org/cgi/content/full/316/5830/1494#otherarticles
, 8 of which can be accessed for free: cites 22 articlesThis article
http://www.sciencemag.org/cgi/collection/immunology
Immunology
: subject collectionsThis article appears in the following
http://www.sciencemag.org/about/permissions.dtl
in whole or in part can be found at: this article
permission to reproduce of this article or about obtaining reprintsInformation about obtaining
registered trademark of AAAS.
c 2007 by the American Association for the Advancement of Science; all rights reserved. The title SCIENCE is a
CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
on June 8, 2007 www.sciencemag.orgDownloaded from

Attenuation of Allergic Contact
Dermatitis Through the
Endocannabinoid System
Meliha Karsak,
1
* Evelyn Gaffal,
2
* Rahul Date,
1
Lihua Wang-Eckhardt,
3
Jennifer Rehnelt,
1
Stefania Petrosino,
4
Katarzyna Starowicz,
4
Regina Steuder,
2
Eberhard Schlicker,
5
Benjamin Cravatt,
6
Raphael Mechoulam,
7
Reinhard Buettner,
8
Sabine Werner,
9
Vincenzo Di Marzo,
4
Thomas Tüting,
2
† Andreas Zimmer
1
†
Allergic contact dermatitis affects about 5% of men and 11% of women in industrialized countries
and is one of the leading causes for occupational diseases. In an animal model for cutaneous
contact hypersensitivity, we show that mice lacking both known cannabinoid receptors display
exacerbated allergic inflammation. In contrast, fatty acid amide hydrolase–deficient mice, which
have increased levels of the endocannabinoid anandamide, displayed reduced allergic responses in
the skin. Cannabinoid receptor antagonists exacerbated allergic inflammation, whereas receptor
agonists attenuated inflammation. These results demonstrate a protective role of the
endocannabinoid system in contact allergy in the skin and suggest a target for therapeutic
intervention.
A
llergic contact dermatitis is characterized
by a loss of immunological tolerance
toward small allergenic molecules (hap-
tens). The allergen or hapten first penetrates the
epidermis and is taken up by skin dendritic cells,
which migrate to the draining lymph nodes and
present haptenated peptide or major histocom-
patibility complexes (MHC) to naï ve an tigen -
specific T lymphocytes. Upon repeated allergen
contact, an inflammatory response is elicited by
the recruitment of antigen-specific effector Tcells
to the skin and the subsequent production of
inflammatory cytokines and chemokines (1).
Cannabinoid receptors CB1 and CB2 are hetero-
trimeric GTP-binding protein (G pr otein)–
coupled receptors, which are activated by the
cannabinoid D
9
-tetrahydrocannabinol (D
9
-THC),
as well as by endocannabinoids such as arachi-
donoylethanolamide (anandamide or AEA) or
2-arachidonoylglycerol (2-AG) (2). Both can-
nabinoid receptors are present on keratinocytes
in the epidermis, which also express enzymes
involved in the synthesis and degradation of
endocannabinoids (3, 4).
Although nickel-containing ear tags are gen-
erally well tolerated by individuals in our mouse
colony (5), we observed that mice lacking both
cannabinoid CB1 and CB2 receptors [CB1
receptor–deficient (Cnr1
−/−
) and CB2 receptor–
deficient (Cnr2
−/−
) mice] (6, 7) frequently scratch
their ear tags, leading to severe ulcerations in
the head and neck regions (Fig. 1, A and B, and
fig. S1). Because localized necrotic lesions ini-
tially developed around the ear tag, we first
considered that impaired wound healing might be
the potential primary cause for the skin lesions.
However , experiments to test the wound healing
ability revealed no measurable difference be-
tween wild-type (WT) and Cnr1
−/−
/Cnr2
−/−
double-knockout (DKO) mice (5, 8) (fig. S2).
We also noted that skin ulcerations appeared to
be particularly prominent in mice with tags con-
taining high nickel content (65 to 70%). In these
cases, skin ulcerations were observed in 88 out of
304 Cnr1
−/−
/Cnr2
−/−
DKO mice (29%), but were
not observed in single Cnr1
−/−
or Cnr2
−/−
knock-
outs or in any other mouse strain. Furthermore, no
ulcerations were noted in mice with pure brass
tags (fig. S1). Subsequent histopathological analy-
ses of ulcerated ear tissue revealed intense
infiltration of mast cells in close proximity to the
ulcers (Fig. 1, C and D), suggesting that an allergic
reaction might be involved in the pathology seen
in Cnr1
−/−
/Cnr2
−/−
DKO mice. Consistently, the
regional preauricular and submental lymph nodes
(Fig. 1, E and F) were swollen, reaching a diam-
eter of 51 ± 2.4 mm in comparison with 20.5 ±
1.9mminhealthyCnr1
−/−
/Cnr2
−/−
mice, as a re-
sult of mixed follicular and interfollicular lym-
phatic hyperplasia.
We therefore considered the possibility that
an increase in allergic responses might exist in
knockout (KO) mice. To test this hypothesis, we
evaluated cutaneous contact hypersensitivity
(CHS) in Cnr1
−/−
/Cnr2
−/−
animals using the ob-
ligate contact allergen 2,4-dinitrofluorobenzene
(DNFB), which generates a specific cutaneous T
cell–mediated allergic response upon repeated
allergen contact (9). A marked increase in allergic
1
Department of Molecular Psychiatry, University of Bonn,
Germany.
2
Laboratory of Experimental Dermatology, Depart-
ment of Dermatology, University of Bonn, Germany.
3
Life &
Brain GmbH, Bonn, Germany.
4
Endocannabinoid Research
Group, Institute of Biomolecular Chemistry, Consiglio
Nazionale delle Ricerche, Pozzuoli (Napoli), Italy.
5
Institute
of Pharmacology and Toxicology, University of Bonn,
Germany.
6
The Skaggs Institute for Chemical Biology and
Departments of Cell Biology and Chemistry, The Scripps
Research Institute, La Jolla, CA, USA.
7
Department of
Medicinal Chemistry and Natural Products, The Hebrew
University of Jerusalem, Israel.
8
Department of Pathology,
University of Bonn, Germany.
9
Institute of Cell Biology,
Eidgenössische Technische Hochschule, Zurich, Switzerland.
*These authors contributed equally to this work.
†To whom correspond ence should be addressed. E-mail:
a.z imme r@uni-bonn.de (A.Z.); thomas.tueting@u kb.u ni-
bonn.de (T.T.)
Fig. 1. Histological views of C nr1
−/−
/
Cnr2
−/−
mouse ears. Clips with high nickel
content cause chronic ear ulceration in
Cnr1
−/−
/Cnr2
−/−
mice(fig.S1).(A)Hematox-
ylin and eosin (H&E)–stained sections of ear
tissue from healthy Cnr1
−/−
/Cnr2
−/−
mice
with normal size and histology. Ct, cartilage;
D, dermis; E, epidermis; HF, hair follicle. (B)
H&E–stained sections of ulcerated ear tissue
from Cnr1
−/−
/Cnr2
−/−
mice carrying nickel
ear tags. Arrowheads point to keratinocytes
on the re-epithelized body of the wound. The
eschar (Es), the strong hyperproliferative
epithelium (H), and the granulation tissue
(G) are indicated. (C) Toluidine blue–stained
sections of ear tissue from healthy Cnr1
−/−
/
Cnr2
−/−
mice showing few metachromatic
mast cells (arrowheads). (D) Toluidine blue–
stained sections of ulcerated ear tissue from
Cnr1
−/−
/Cnr2
−/−
mice carrying nickel tags
with intense infiltration of mast cells (arrow-
heads). (E)H&E–stained sections of lymph
nodes from healthy Cnr1
−/−
/Cnr2
−/−
mice. ln,
lymph node; sg, salivary gland. (F)H&E–
stained sections of lymph nodes from Cnr1
−/−
/
Cnr2
−/−
mice carrying nickel ear tags with
reactive enlargement and lymphadenitis of
preauricular lymph nodes. The preauricular
salivary glands located next to the lymph
nodesarethesamesizeinCnr1
−/−
/Cnr2
−/−
mice without (E) and with (F) ear ulceration and serve as an internal standard.
8 JUNE 2007 VOL 316 SCIENCE www.sciencemag.org1494
REPORTS
on June 8, 2007 www.sciencemag.orgDownloaded from

responses was apparent in Cnr1
−/−
/Cnr2
−/−
mice
as compared with WT controls (Fig. 2A). The
difference was particularly prominent 48 hours
after the second and the third DNFB challenge.
Infiltration of the skin with Gr-1
+
granulocytes
was increased in the ears of DNFB-exposed
Cnr1
−/−
/Cnr2
−/−
mice, as compared with WT ani-
mals (fig. S3), and correlated with a higher myelo-
peroxidase activity, which is indicative of enhanced
neutrophil recruitment (10, 11 ). Furthermore, we
found an increased number of MHC II
+
antigen-
presenting cells in Cnr1
−/−
/Cnr2
−/−
mice. Upon
examination of mice with a sin gle deletion of either
CB1 or CB2 receptors, we observed a similarly
pronounced increase of allergic ear swelling (Fig.
2B), suggesting a nonredundant role for each
cannabinoid receptor in the allergic response to
DNFB. This finding would also suggest that the
response to the milder stimulus originating from
the nickel ear tag was only detectable in mice
lacking both receptors.
In order to obtain independent evidence for a
role of the CB1 and CB2 receptors in the reg-
ulation of CHS, we examined WT mice after
administration of the CB1 receptor antagonist
SR141716A (Rimonabant) and the CB2 receptor
antagonist SR144528, respectively. After the in-
duction of CHS, animals received three injections
of the corresponding antagonist 30 min before
and after each challenge. In these cases, ear swell-
ing was significantly increased in treated mice as
compared with control mice (Fig. 2C), further
supporting the role of both receptors in CHS. In
contrast, however, acute oral or topical adminis-
tration of the CB2 receptor antagonists appeared to
ameliorate inflammatory skin responses (12, 13),
thus suggesting that CB2 antagonism may be
initially beneficial but detrimental upon chronic
blockade.
To explore this issue more carefully , we inves-
tigated the synthesis of endogenous cannabinoids
and the expression of cannabinoid receptors
during experimental contact allergy studies. After
DNFB treatment, the levels of 2-AG increased in
the DNFB-treated ears of WT controls and to an
even larger extent in the ears of Cnr1
−/−
/Cnr2
−/−
mice (Fig. 3A), whereas AEA production was
strongly induced in both genotypes. Endocanna-
binoids may have been overproduced in KO
animals as a result of the lack of communication
between receptors and ligands (14) or as a further
consequence of the exacerbated allergic response.
Keratinocytes are known to express CB1 recep-
tors (4), although the expression of CB2 on these
Fig. 2. Contact allergy in canna-
binoid receptor–deficient mice.
(A)Cnr1
−/−
/Cnr2
−/−
and Cnr1
+/+
/
Cnr2
+/+
mice were sensitized (indi-
cated by “s”)withDNFBonthe
shaved abdomen. On day 5, mice
were challenged (indicated by “c” )
with DNFB on the right ear. A second
and third challenge was performed
on day 13 and 21. Shown is the
mean ear swelling over time ± SEM
of a representative experiment with nine mice per group. Statistical significance
was evaluated with the Wilcoxon-Mann-Whitney two-sample test (**P <0.01).This
experiment was independently repeated four times with similar results. (B)
Cnr1
−/−
/Cnr2
+/+
,Cnr1
+/+
/Cnr2
−/−
,andCnr1
+/+
/Cnr2
+/+
mice were sensitized as
described in (A). Ear swelling 48 hours after the third challenge of a rep-
resentative experiment with eight mice per group is shown. Similar results were
obtained in four independent experiments with a total of 23 mice (**P <0.01).
(C) Contact allergic response in C57BL/6J mice after treatment with the
indicated CB receptor antagonists. Ear swelling 48 hours after the second
challenge of a representative experiment with 10 mice per group is shown.
Similar results were obtained in four independent experiments with a total of 25
mice (***P < 0.001). Error bars in (B) and (C) indicate SEM.
0.4
0.3
30
0.2
20 25
0.1
10 15
0.0
0 5
0.5
ss c
Cnr1
-/-
/Cnr2
-/-
Cnr1
+/+
/Cnr2
+/+
cc
**
**
**
**
**
**
**
**
**
**
**
0.4
0.3
0.2
0.1
0.0
0.5
Cnr1 +/+ -/- +/+
+/+ +/+ -/-Cnr2
BA
C
mean ear swelling (mm)
mean ear swelling (mm)
mean ear swelling (mm)
+
+
-
+
+
-
-
-
SR141716A
SR144528
0.0
0.1
0.2
0.3
***
***
***
**
A
C
D
E
B
0 2 4 6 8 10 12 14 16 18 20 22
0.0
0.1
0.2
0.3
**
**
**
***
***
*
*
day
vehicle
HU-308
∆
9
-THC
)mm( gnillews rae naem
gm/lomp
g/lomp
0 2 4 6 8 10 12 14 16 18 20 22
0.0
0.1
0.2
0.3
vehicle
HU-308
∆
9
-THC
day
*
*
*
*
*
**
**
***
)mm( gnillews rae naem
2-AG
0
20
40
60
80
DCDC
Cnr1
-/-
Cnr2
-/-
Cnr1
+/+
Cnr2
+/+
AEA
0
100
200
300
400
DCDC
Cnr1
-/-
Cnr2
-/-
Cnr1
+/+
Cnr2
+/+
FAAH
-/-
FAAH
+/+
0 5 10 15 2520
0.0
0.1
0.2
day
)mm( gnillews rae naem
**
**
*
**
*
**
*
stinu evitaler [01 x
4-
]
CB1
0
4
8
12
16
**
CB2
0
40
80
120
***
DCDC
Fig. 3. The endogenous cannabinoid system in CHS. (A) Endocannabinoid levels in
ear samples of Cnr1
+/+
/Cnr2
+/+
and Cnr1
−/−
/Cnr2
−/−
mice treated with DNFB after
the second challenge (indicated by “D”), in comparison with the control sides of the
same animals (indicated by “C”), were measured by means of liquid chroma tography–
mass spectrometry. Data represent mean values ± SEM. (B) CB1 and CB2 mRNA
expression levels in ear samples of Cnr1
+/+
/Cnr2
+/+
mice treated with DNFB after the
second challenge (“D”), in comparison with the control sides of the same animals
(“C”). Relative expression units were determined by quantitative real-time PCR. Data
represent mean values ± SEM. ( C and D) Contact allergic responses in C57BL/6J mice
after subcutaneous (C) or topical (D) treatment with the agonis ts D
9
-THC and HU-308
or with vehicle alone. Error bars indicate SEM. (E) Contact allergic responses in
FAAH
−/−
and FAAH
+/+
mice. Shown is the mean ear swelling over time ± SEM of
representative experiments with 10 mice per group (*P <0.05,**P <0.01,***P <
0.001).
www.sciencemag.org SCIENCE VOL 316 8 JUNE 2007
1495
REPORTS
on June 8, 2007 www.sciencemag.orgDownloaded from

cells has been debated (3, 12, 15). We could
identify both receptor mRNAs in human HaCat
keratinocytes (fig. S4). CB1 receptor mRNA
was down-regulated in HaCat cells that had
been stimulated for 24 hours with polyinosinic-
polycytidylic acid and also in DNFB-exposed
ears, whereas CB2 mRNA and protein expres-
sion was up-regulated in both cases (Fig. 3B and
fig. S4). Taken together, these findings indicate
that DNFB-induced CHS caused an activation of
the endocannabinoid system in the skin and im-
balanced expression of both receptors.
Because our data indicate that activation of
the endocannabinoid system may function to
dampen the CHS response, we considered the
possibility that administration of cannabinoids
such as D
9
-THC might attenuate CHS in WT
animals. Therefore, after CHS was induced, we
injected mice with D
9
-THC (5 mg per kilogram
of body weight, administered subcutaneously)
30 min before, as well as 24 and 48 hours after,
DNFB challenge. Indeed, D
9
-THC significantly
decreased ear swelling (Fig. 3C) and reduced the
recruitment of Gr-1
+
granulocytes in comparison
with swelling and granulocyte recruitment in
untreated mice [vehicle: 483 ± 62 Gr-1
+
cells per
high power field (HPF) versus 200 ± 50 Gr-1
+
cells per HPF in subcutaneous THC treatment].
A similar therapeutic effect of D
9
-THC was
observed after topical application of 30 mgof
D
9
-THC 30 min before, as well as 24 and 48
hours after, DNFB administration (Fig. 3D). In
contrast, the CB2-specific agonist HU-308 either
showed no efficacy or even increased the allergic
response after subcutaneous or topical applica-
tion (Fig. 3, C and D) (13). It has recently been
shown that leukocyte chemotaxis is inhibited by
the CB2 receptor inverse agonist N-[1(S)-[4-[[4-
methoxy-2-[(4-methoxyphenyl)sulfonyl]phenyl]-
sulfonyl]phenyl]ethyl]methanesulfonamide
(Sch. 336) and, thus, HU-308 administration to
the inflamed ear may have further increased
leukocyte infiltration (16).
Because endocannabinoid levels were reg-
ulated in DNFB-treated ears and D
9
-THC had a
beneficial effect on CHS, we asked whether mice
having higher endocannabinoid levels would
show altered allergic responses after DNFB treat-
ment. Therefore, we analyzed the allergic reac-
tion in mice lacking the enzyme fatty acid amide
hydrolase (FAAH), which have augmented anan-
damide levels (17).AsshowninFig.3E,we
found a significantly decreased allergic response
in FAAH KO mice after the second and third
DNFB challenge when compared with the re-
sponse in WT mice. Similar to our results, mice
with only peripherally elevated fatty acid amides
showed a reduced inflammation in the carra-
geenan foot-pad model (18). All data indicate that
the anti-inflammatory responses were periph-
eral, rather than central, and that FAAH inhib-
itors are a potential therapeutic approach for
inflammation (and a potentially useful alterna-
tive to direct CB1/CB2 agonists).
In order to gain insight into the molecular
mechanism that may contribute to the increased
magnitude of CHS in cannabinoid receptor–
deficient mice, we performed a series of microarray
experiments with RNA isolated from DNFB-
treated ears of Cnr1
−/−
/Cnr2
−/−
and Cnr1
+/+
/Cnr2
+/+
mice, as well as from the untreated ears of control
mice. In all, 830 probe sets were differentially
expressed between control and DNFB-treated
ears, of which 674 were up-regulated and only
156 were down-regulated. We grouped the probe
sets by using gene ontology (GO) annotation
(19, 20) (Fig. 4A). Immune-related probe sets
represented the largest group of differentially
expressed genes (40 down-regulated and 117
up-regulated probe sets). Notably, large num-
bers of chemokine C-C and C-X-C motif ligands
(21) and receptors were found in the up-regulated
group (table S1) and represented 22% of the
immune-related probe sets (table S2). Fifty-four
genes were differentially expressed in DNFB-
treated whole-ear specimens of Cnr1
+/+
/Cnr2
+/+
mice versus those of Cnr1
−/−
/Cnr2
−/−
mice (fig.
S5). The only chemokine gene transcript in this
group was monocyte chemotactic protein 2
(MCP-2)/chemokine (C-C motif) l igand 8
(CCL8). Using real-time reverse transcription
polymerase chain reaction (R T-PCR) and in situ
hybridization, we confirmed the differential
expression of MCP-2/CCL8 mRNA in keratino-
cytes (Fig. 4, B and C). Further in vitro
experiments showed dynamic regulation of
MCP-2/CCL8 production in activated keratino-
cytes through cannabinoid agonists (fig. S6).
MCP-2/CCL8 is a member of the MCP family ,
whichhasbeenshowntoplayamajorroleinthe
recruitment of macrophages, activated lympho-
cytes, and mast cells into inflammatory sites
(22, 23). Taken together, our results suggest that
the endocannabinoid system regulates allergic
inflammation through a modulation of the chemo-
kine system.
The observation that the endocannabinoid
system is activated in a model of CHS, together
with the fact that genetic deletion or pharmaco-
logic blockade of CB receptors enhanced con-
tact allergic inflammation, whereas stimulation
of CB receptors reduced such inflammation,
strongly suggests that the endocannabinoid sys-
tem serves to attenuate the inflammatory re-
50
0
50
100
150
200
250
erutcurts llec
srehto
sisotpopa
msilobatem
noisehda llec
nwonknu
sisotim / elcyc llec
noitacilper / riaper AND
noitalsnart / noitpircsnart
tropsnart
gnillangis
esnopser esnefed
enummi
stpircsnart fo rebmuN
detaluger-pu-nwod
A
realtime PCRAffymetrix
0
2500
5000
7500
10000
level evitaler
***
***
##
B
***
***
***
##
**
CD
Cnr1
Cnr2
stinu evitaler
0.00
0.02
0.04
0.06
0.08
CDCD C D CD
C
D
+/+
+/+
Cnr1
Cnr2
+/+
+/+
+/+
-/-
-/-
-/-
-/-
-/-
-/-
+/+
Cnr1
+/+
/Cnr2
+/+
Cnr1
-/-
/Cnr2
-/-
sense control
C
Dermis
Epidermis
CDCD
0.1 mm
Fig. 4. Gene expression analysis. (A) Microarray analysis of ear samples of
Cnr1
+/+
/Cnr2
+/+
and Cnr1
−/−
/Cnr2
−/−
mice treated with DNFB 48 hours
after the second challenge (“D”) in comparison with the control sides of
thesameanimals(“C”) revealed 674 up-regulated and 156 down-
regulated probe sets in response to the treatment. Grouping was performed
by GO annotation. Chemokine C-C motif and C-X-C motif ligands and
receptors in the up-regulated group are presented in light gray in the
column of immune-related genes. (B) Transcripts for the chemokine MCP-
2/CCL8 in DNFB-treated ears of Cnr1
+/+
/Cnr2
+/+
and Cnr1
–/–
/Cnr2
–/–
mice
resulting from microarray expression study (left panel) and real-time RT-
PCR quantification (right panel). Shown are representative results of three
independent experiments (**P <0.01,##P <0.01,***P < 0.001). Err or
bars indicate SEM. (C) In situ hybridization with the use of a
digoxigenin-labeled probe for MCP-2/CCL8 in ear tissue of DNFB-treated
and control ears from Cnr1
+/+
/Cnr2
+/+
and Cnr1
−/−
/Cnr2
−/−
mice. The sense
probe showed no staining.
8 JUNE 2007 VOL 316 SCIENCE www.sciencemag.org
1496
REPORTS
on June 8, 2007 www.sciencemag.orgDownloaded from

sponse in cutaneous CHS. The finding that
topical application of D
9
-THC reduced allergic
inflammation points to the promising potential of
developing pharmacological treatments (24)with
the use of selective CB receptor agonists or
FAAH inhibitors.
References and Notes
1. S. Grabbe, T. Schwarz, Immunol. Today 19, 37 (1998).
2. T. Bisogno, A. Ligresti, V. Di Marzo, Pharmacol. Biochem.
Behav. 81, 224 (2005).
3. M. M. Ibrahim et al., Proc. Natl. Acad. Sci. U.S.A. 102,
3093 (2005).
4. M. Maccarrone et al., J. Biol. Chem. 278, 33896 (2003).
5. Materials and methods are available as supporting
material on Science Online.
6. A. Zimmer, A. M. Zimmer, A. G. Hohmann, M. Herkenham,
T. I. Bonner, Proc. Natl. Acad. Sci. U.S.A. 96, 5780 (1999).
7. N. E. Buckley et al., Eur. J. Pharmacol. 396, 141 (2000).
8. S. Werner et al., Science 266, 819 (1994).
9. J. Knop, R. Stremmer, C. Neumann, E. De Maeyer,
E. Macher, Nature 296, 757 (1982).
10. R. I. Lehrer, J. Hanifin, M. J. Cline, Nature 223, 78 (1969).
11. C. Nathan, Nat. Rev. Immunol. 6, 173 (2006).
12. S. Oka et al., J. Immunol. 177, 8796 (2006).
13. Y. Ueda, N. Miyagawa, T. Matsui, T. Kaya, H. Iwamura,
Eur. J. Pharmacol. 520, 164 (2005).
14. V. Di Marzo et al., J. Neurochem. 75, 2434 (2000).
15. S. Stander, M. Schmelz, D. Metze, T. Luger, R. Rukwied,
J. Dermatol. Sci. 38, 177 (2005).
16. C. A. Lunn et al., J. Pharmacol. Exp. Ther. 316, 780
(2006).
17. B. F. Cravatt et al., Proc. Natl. Acad. Sci. U.S.A. 98, 9371
(2001).
18. B. F. Cravatt et al. , Proc. Nat l. Acad. Sci. U.S.A. 101,
10821 (2004).
19. J. S. Lee, G. Katari, R. Sachidanandam, BMC Bioinformat.
6, 189 (2005).
20. M. Ashburner et al., Nat. Genet. 25, 25 (2000).
21. M. F. Bachmann, M. Kopf, B. J. Marsland, Nat. Rev.
Immunol. 6, 159 (2006).
22. A. de Paulis et al., Int. Arch. Allergy Immunol. 124, 146
(2001).
23. D. D. Taub et al., J. Clin. Invest. 95, 1370 (1995).
24. T. W. Klein, Nat. Rev. Immunol. 5, 400 (2005).
25. This work was supported by grants from the Deutsche
Forschungsgemeinschaft [SFB645 and GRK804 (to M.K.
and A.Z.) and Tu90/5-1 (to T.T.)], by a Bonfor stipend
to E.G., and a grant from Epitech S.r.l. to V.D.M. We
thank M. Krampert for her help in wound healing
experiments, L. Cristino for her help with
immunohistochemistry, and J. Essig, A. Zimmer,
E. Erxlebe, and I. Heim for technical assistance.
Supporting Online Material
www.sciencemag.org/cgi/content/full/316/5830/1494/DC1
Materials and Methods
Figs. S1 to S6
Tables S1 and S2
References
8 March 2007; accepted 4 May 2007
10.1126/science.1142265
Genome-Wide Mapping of in Vivo
Protein-DNA Interactions
David S. Johnson,
1
* Ali Mortazavi,
2
* Richard M. Myers,
1
† Barbara Wold
2,3
†
In vivo protein-DNA interactions connect each transcription factor with its direct targets to form a
gene network scaffold. To map these protein-DNA interactions comprehensively across entire
mammalian genomes, we developed a large-scale chromatin immunoprecipitation assay (ChIPSeq)
based on direct ultrahigh-throughput DNA sequencing. This sequence census method was then
used to map in vivo binding of the neuron-restrictive silencer factor (NRSF; also known as REST, for
repressor element–1 silencing transcription factor) to 1946 locations in the human genome. The
data display sharp resolution of binding position [±50 base pairs (bp)], which facilitated our
finding motifs and allowed us to identify noncanonical NRSF-binding motifs. These ChIPSeq
data also have high sensitivity and specificity [ROC (receiver operator characteristic)
area ≥ 0.96] and statistical confidence (P <10
−4
), properties that were important for inferring new
candidate interactions. These include key transcription factors in the gene network that regulates
pancreatic islet cell development.
A
lthough much is known about transcrip-
tion factor binding and action at specific
genes, far less is known about the com-
position and function of entire factor-DNA
interactomes, especially for organisms with large
genomes. Now that human, mouse, and other
large genomes have been seq uenced, it is
possible, in principle, to measure how any
transcription factor is deployed across the entire
genome for a given cell type and physiological
condition. Such measurements are important for
systems-level studies because they provide a
global map of candidate gene network input
connections. These direct physical interactions
between transcription factors or cofactors and the
chromosome can be detected by chromatin
immunoprecipitation (ChIP) (1). In ChIP ex-
periments, an immune reagent specific for a
DNA binding factor is used to enrich target DNA
sites to which the factor was bound in the living
cell. The enriched DNA sites are then identified
and quantified.
For the gigabase-size genomes of vertebrates,
it has been difficult to make ChIP measurements
that combine high accuracy , whole-genome com-
pleteness, and high binding-site resolution. These
data-quality and depth issues dictate whether pri-
mary gene network structure can be inferred with
reasonable certainty and comprehensiveness, and
how effectively the data can be used to discover
binding-site motifs by computational methods.
For these purposes, statistical robustness, sam-
pling depth across the genome, absolute signal
and signal-to-noise ratio must be good enough
to detect nearly all in vivo binding locations for
a regulator with minimal inclusion of false-
positives. A further challenge in genomes large
or small is to map factor-binding sites with high
positional resolution. In addition to making com-
putational discovery of binding motifs feasible,
this dictates the quality of regulatory site anno-
tation relative to other gene anatomy landmarks,
such as transcription start sites, enhancers, introns
and exons, and conserved noncoding features
(2). Finally, if high-quality protein-DNA inter-
actome measurements can be performed rou-
tinely and at reasonable cost, it will open the
way to detailed studies of interactome dynam-
ics in response to specific signaling stimuli or
genetic mutations. To address these issues, we
turned to ultrahigh-throughput DNA sequenc-
ing to gain sampling power and applied size
selection on immuno-enriched DNA to enhance
positional resolution.
The ChIPSeq assay shown here differs
from other large-scale ChIP methods such as
Ch I PAr r a y, also called ChIPchip (1); ChIPSAGE
(SACO) (3); or ChIPPet (4) in design, data
produced, and cost. The design is simple (Fig.
1A) and, unlike SACO or ChIPPet, it involves no
plasmid library construction. Unlike microarray
assays, the vast majority of single-copy sites in
the genome is accessible for ChIPSeq assay (5),
rather than a subset selected to be array features.
For example, to sample with similar complete-
ness by an Affymetrix-style microarray design, a
nucleotide-by-nucleotide sliding window design
of roughly 1 billion features per array would be
needed for the nonrepeat portion of the human
genome. In addition, ChIPSeq counts sequences
and so avoids constraints imposed by array
hybridization chemistry , such as base composition
constraints related to T
m
, the temperature at which
50% of double-stranded DNA or DNA-RNA
hybrids is denatured; cross-hybridization; and
secondary structure interference. Finally , ChIPSeq
is feasible for any sequenced genome, rather than
being restricted to species for which whole-
genome tiling arrays have been produced.
ChIPSeq illustrates the power of new se-
quencing platforms, such as those from Solexa/
Illumina and 454, to perform sequence census
counting assays. The generic task in these appli-
cations is to identify and quantify the molecular
1
Department of Genetics, Stanford University School of
Medicine, Stanford, CA, 94305–5120, USA.
2
Biology
Division, California Institute of Technology, Pasadena, CA
91125, USA.
3
California Institute of Technology Beckman
Institute, Pasadena, CA 91125, USA.
*These authors contributed equally to this work.
†To whom correspondence should be addressed. E-mail:
woldb@its.caltech.edu (B.W.); myers@shgc.stanford.edu
(R.M.M.)
www.sciencemag.org SCIENCE VOL 316 8 JUNE 2007 1497
REPORTS
on June 8, 2007 www.sciencemag.orgDownloaded from