Beta-caryophyllene is a dietary cannabinoid

Abstract

The psychoactive cannabinoids from Cannabis sativa L. and the arachidonic acid-derived endocannabinoids are nonselective natural ligands for cannabinoid receptor type 1 (CB1) and CB2 receptors. Although the CB1 receptor is responsible for the psychomodulatory effects, activation of the CB2 receptor is a potential therapeutic strategy for the treatment of inflammation, pain, atherosclerosis, and osteoporosis. Here, we report that the widespread plant volatile (E)-β-caryophyllene [(E)-BCP] selectively binds to the CB2 receptor (K i = 155 ± 4 nM) and that it is a functional CB2 agonist. Intriguingly, (E)-BCP is a common constituent of the essential oils of numerous spice and food plants and a major component in Cannabis. Molecular docking simulations have identified a putative binding site of (E)-BCP in the CB2 receptor, showing ligand π–π stacking interactions with residues F117 and W258. Upon binding to the CB2 receptor, (E)-BCP inhibits adenylate cylcase, leads to intracellular calcium transients and weakly activates the mitogen-activated kinases Erk1/2 and p38 in primary human monocytes. (E)-BCP (500 nM) inhibits lipopolysaccharide (LPS)-induced proinflammatory cytokine expression in peripheral blood and attenuates LPS-stimulated Erk1/2 and JNK1/2 phosphorylation in monocytes. Furthermore, peroral (E)-BCP at 5 mg/kg strongly reduces the carrageenan-induced inflammatory response in wild-type mice but not in mice lacking CB2 receptors, providing evidence that this natural product exerts cannabimimetic effects in vivo. These results identify (E)-BCP as a functional nonpsychoactive CB2 receptor ligand in foodstuff and as a macrocyclic antiinflammatory cannabinoid in Cannabis.

• Cannabis
• CB2 cannabinoid receptor
• foodstuff
• inflammation
• natural product

Full text

Beta-caryophyllene is a dietary cannabinoid
Ju
¨rg Gertsch*
†
, Marco Leonti
‡§
, Stefan Raduner*
§
, Ildiko Racz
¶
, Jian-Zhong Chen
储
, Xiang-Qun Xie
储
, Karl-Heinz Altmann*,
Meliha Karsak
¶
, and Andreas Zimmer
¶
*Institute of Pharmaceutical Sciences, Department of Chemistry and Applied Biosciences, Eidgeno¨ ssische Technische Hochschule (ETH) Zurich, 8092 Zu¨ rich,
Switzerland; ‡Dipartimento Farmaco Chimico Tecnologico, University of Cagliari, 01924 Cagliari, Italy; ¶Department of Molecular Psychiatry, University of
Bonn, 53115 Bonn Germany; and 储Department of Pharmaceutical Sciences, University of Pittsburgh, Pittsburgh, PA 15260
Edited by L. L. Iversen, University of Oxford, Oxford, United Kingdom, and approved May 6, 2008 (received for review April 14, 2008)
The psychoactive cannabinoids from Cannabis sativa L. and the
arachidonic acid-derived endocannabinoids are nonselective nat-
ural ligands for cannabinoid receptor type 1 (CB
1
) and CB
2
recep-
tors. Although the CB
1
receptor is responsible for the psychomodu-
latory effects, activation of the CB
2
receptor is a potential
therapeutic strategy for the treatment of inflammation, pain,
atherosclerosis, and osteoporosis. Here, we report that the wide-
spread plant volatile (E)-
␤
-caryophyllene [(E)-BCP] selectively binds
to the CB
2
receptor (K
i
ⴝ155 ⴞ4 nM) and that it is a functional CB
2
agonist. Intriguingly, (E)-BCP is a common constituent of the
essential oils of numerous spice and food plants and a major
component in Cannabis. Molecular docking simulations have iden-
tified a putative binding site of (E)-BCP in the CB
2
receptor,
showing ligand
␲
–
␲
stacking interactions with residues F117 and
W258. Upon binding to the CB
2
receptor, (E)-BCP inhibits adenylate
cylcase, leads to intracellular calcium transients and weakly acti-
vates the mitogen-activated kinases Erk1/2 and p38 in primary
human monocytes. (E)-BCP (500 nM) inhibits lipopolysaccharide
(LPS)-induced proinflammatory cytokine expression in peripheral
blood and attenuates LPS-stimulated Erk1/2 and JNK1/2 phosphor-
ylation in monocytes. Furthermore, peroral (E)-BCP at 5 mg/kg
strongly reduces the carrageenan-induced inflammatory response
in wild-type mice but not in mice lacking CB
2
receptors, providing
evidence that this natural product exerts cannabimimetic effects in
vivo. These results identify (E)-BCP as a functional nonpsychoactive
CB
2
receptor ligand in foodstuff and as a macrocyclic antiinflam-
matory cannabinoid in Cannabis.
Cannabis 兩CB2 cannabinoid receptor 兩foodstuff 兩inflammation 兩
natural product
Plant essential oils are typically composed of volatile aromatic
terpenes and phenylpropanoids. These lipophilic volatiles freely
cross cellular membranes and serve various ecological roles, like
plant-insect interactions (1, 2). The sesquiterpene (E)-
␤
-
caryophyllene [(E)-BCP] (Fig. 1) is a major plant volatile found in
large amounts in the essential oils of many different spice and food
plants, such as oregano (Origanum vulgare L.), cinnamon (Cinna-
momum spp.) and black pepper (Piper nigrum L.) (3–5). In nature,
(E)-BCP is usually found together with small quantities of its
isomers (Z)-
␤
-caryophyllene [(Z)-BCP or isocaryophyllene] and
␣
-humulene (formerly
␣
-caryophyllene) or in a mixture with its
oxidation product, BCP oxide (Fig. 1). Because of its weak aromatic
taste, (E)-BCP is commercially used as a food additive and in
cosmetics (6). (E)-BCP is also a major component (up to 35%) in
the essential oil of Cannabis sativa L (7). Although Cannabis
contains ⬎400 different secondary metabolites, including ⬎65
cannabinoid-like natural products, only ⌬
9
-tetrahydrocannabinol
(THC), ⌬
8
-tetrahydrocannabinol, and cannabinol have been re-
ported to activate cannabinoid receptor types 1 (CB
1
) and 2 (CB
2
)
(8). Here, we show that the essential oil component (E)-BCP
selectively binds to the CP55,940 binding site (i.e., THC binding
site) in the CB
2
receptor, leading to cellular activation and antiin-
flammatory effects.
CB
1
and CB
2
cannabinoid receptors are GTP-binding protein (G
protein) coupled receptors that were first cloned in the early 1990s
(9, 10). Although the CB
1
receptor is expressed in the central
nervous system and in the periphery, the CB
2
receptor is primarily
found in peripheral tissues (11). In vivo, CB receptors are activated
by arachidonic acid-derived endocannabinoids, such as 2-arachido-
noyl ethanolamine (anandamide or AEA) and 2-arachidonoylglyc-
erol (2-AG) (12, 13). In addition to a wide range of primarily CB
1
receptor-mediated physiological effects on the central nervous
system, different cannabinoid ligands have been reported to mod-
ulate immune responses (14). In particular, CB
2
receptor ligands
have been shown to inhibit inflammation and edema formation
(15), exhibit analgesic effects (16), and play a protective role in
hepatic ischemia-reperfusion injury (17). In the gastrointestinal
tract, CB
2
receptor agonists have been shown to prevent experi-
mental colitis by reducing inflammation (18). Moreover, the CB
2
receptor has been described as a potential target for the treatment
of atherosclerosis (19) and osteoporosis (20). Consequently, CB
2
receptor-selective agonists that are devoid of the psychoactive side
effects typically associated with CB
1
receptor activation are poten-
tial drug candidates for the treatment of a range of different
diseases.
Results
Identification of CB
2
Receptor-Selective Ligands in
Cannabis
Essential
Oil. In our ongoing search for new CB
2
cannabinoid receptor-
selective ligands from natural sources, we observed that C. sativa
essential oil (5
␮
g/ml) devoid of the classical cannabinoids strongly
displaced the high-affinity radioligand [
3
H]CP55,940 (21) from
hCB
2
but not hCB
1
receptors [supporting information (SI) Fig. S1].
The same observation was made for a number of other essential
oils, thus suggesting that Cannabis essential oil may contain CB
2
Author contributions: J.G. and M.K. designed research; J.G., M.L., S.R., I.R., and J.-Z.C.
performed research; A.Z. contributed new reagents/analytic tools; J.G., X.-Q.X., K.-H.A.,
M.K., and A.Z. analyzed data; and J.G. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
†To whom correspondence should be addressed. E-mail: juerg.gertsch@pharma.ethz.ch.
§M.L. and S.R. contributed equally to this work.
This article contains supporting information online at www.pnas.org/cgi/content/full/
0803601105/DCSupplemental.
© 2008 by The National Academy of Sciences of the USA
O
HH
HH H H
(E)-BCP
12
3
4
512
6
7
8
13 9
10 11
14
15
(Z)-BCP BCP oxideα-humulene
Fig. 1. Caryophyllane- and humulane-type sesquiterpenes found in C. sativa
and numerous other plants. Shown are the chemical structures of the bicyclic
sesquiterpenes (E)-
␤
-caryophyllene, (Z)-
␤
-caryophyllene, caryophyllene ox-
ide, and the ring-opened isomer
␣
-humulene (
␣
-caryophyllene).
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0803601105 PNAS
兩
July 1, 2008
兩
vol. 105
兩
no. 26
兩
9099–9104
PHARMACOLOGY

receptor active compounds other than classical cannabinoids and
that these ligands may also commonly occur in other plant species.
Fractionation of Cannabis essential oil by column chromatography
and screening of the isolated constituents yielded (E)-BCP as a CB
2
receptor binding compound (Fig. 1).
Quantitative radioligand binding experiments showed that (E)-
BCP and its isomer, (Z)-BCP (Fig. 1), dose-dependently displaced
[
3
H]CP55,940 from hCB
2
receptors expressed in HEK293 cells with
apparent K
i
values in the nM range (Fig. 2A). (E)-BCP, which is the
isomer predominantly found in plants, showed a slightly higher CB
2
receptor binding affinity (K
i
⫽155 ⫾4 nM) than (Z)-BCP (K
i
⫽
485 ⫾36 nM). Notably, BCP oxide (Fig. 1), which is the volatile
BCP oxidation product sensed by narcotic detection dogs (22), and
the ring-opened isomer
␣
-humulene did not displace [
3
H]CP55,940
from the hCB
2
receptor (K
i
⬎20
␮
M) (Fig. S2). In accordance with
the data obtained with Cannabis essential oils (Fig. S1), BCP
isomers did not show significant binding affinity to the hCB
1
receptor (Fig. S2). Moreover, none of the other major Cannabis
terpenes (10
␮
M) showed a significant displacement of
[
3
H]CP55,940 (⬎50%) in either hCB
2
or hCB
1
receptor radioli-
gand binding assays (Fig. S2).
(E)-BCP Competitively Binds to the
h
CB
2
Receptor THC Binding Site.
Receptor binding studies were hampered by the poor water solu-
bility of the apolar (E)-BCP (clogP ⫽6.7). (E)-BCP led to the
formation of oil droplets at concentrations ⬎1
␮
M, whereas
(Z)-BCP showed a somewhat better water solubility (data not
shown). Accordingly, the displacement curve for (E)-BCP showed
a biphasic trend and K
i
values for (E)-BCP and (Z)-BCP calculated
from the Hill plot were 780 ⫾12 nM and 974 ⫾65 nM, respectively
(Fig. 2B). Thus, the K
i
values for (E)-BCP obtained in the nonlinear
regression (Fig. 2A), and the Hill plot (Fig. 2B) showed a statisti-
cally significant 5-fold difference. Because the Hill plot for (E)-BCP
deviates from linearity, the possibility of a biphasic nature of the
displacement was explored by using the GraphPad Prism software.
The biphasic displacement curve exhibited a poor correlation
(R
2
⬍0.8), suggesting that solubility rather than the presence of
multiple binding sites was responsible for the discrepancy between
K
i
values from nonlinear regression analyses and Hill plots. This was
subsequently confirmed in a Dixon analysis in which competitive
binding of (E)-BCP to the CP55,940 binding site was shown (Fig.
2C). The inhibition constant estimated from the Dixon plot was
⬇500 nM, again confirming the nanomolar range of the (E)-BCP
binding affinity. Because THC and CP55,940 share the same
binding site in the hCB
2
receptor, it can be concluded that (E)-BCP
binds to the THC binding pocket or an overlapping site.
In Silico
Docking Analysis of the (E)-BCP CB
2
Receptor Binding Inter-
action. Computational docking analyses, using Surflex-Dock cal-
culations (see SI Materials and Methods) with an established CB
2
receptor homology model (23), suggest that (E)-BCP binds into the
hydrophobic region of the water-accessible cavity. The (E)-BCP-
CB
2
complex energy was minimized by MD/MM simulations. The
putative binding site of CB
2
receptor ligands is located adjacent to
helices III, V, VI, and VII at the near extracellular site of the 7TM
bundles (Fig. 3). In this model, the lipophilic (E)-BCP docks into the
hydrophobic cavity of the amphiphatic binding pocket and the
binding mode of (E)-BCP appears to be facilitated by
␲
–
␲
stacking
interactions with residues F117 (4.0 Å) and W258 (4.6 Å). More-
over, (E)-BCP closely interacts with the hydrophobic residues I198,
V113, and M265 (Fig. 3). Only four different geometries are
allowed in (E)-BCP, owing to the constraints imposed by the
nine-membered ring. These conformations are designated
␣␣
,
␣␤
,
␤␣
, and
␤␤
, showing the relative orientations of the C8-C13
exocyclic double bond and of the C4–C5 (–C12) vinylic moiety, in
which
␣
or
␤
denotes the position of the C8–C13 double bond and
the C12 methyl below or above the molecular plane (24). Based on
the data obtained by molecular modeling, the bioactive conforma-
tion of (E)-BCP in the CB
2
receptor complex resembles the
␤␤
low
1.5 2.0 2.5 3.0 3.5 4.0
-1
0
1
2
3
(Z)-BCP
(E)-BCP
Log nM
Log (BL/(BLO-BL))
-8 -7 -6 -5
25
50
75
100
(E)-BCP
(Z)-BCP
Anandamide
Log [M]
[
3
H]CP-55,940 (% bound to CB
2
receptor s)
K
i
= 155 ±4 nM
K
i
= 485 ±36 nM
A
0
2
4
6
8
10
12
14
16
-4000 -1500 1000 3500 6000 8500 11000
(E)-BCP (nM)
1 / Bound (nM )
84 pM
168 pM
252 pM
B
C
K
i
= 974 ±65 nM
K
i
= 780 ±12 nM
K
i
= 296 ±15 nM
Fig. 2. (E)-BCP and (Z)-BCP displace [3H]CP-55,940 from hCB2receptors
expressed in HEK293 cells. (A) Sigmoidal displacement curves (R2⫽0.93 for
(E)-BCP and 0.98 for (Z)-BCP) show overall displacement (71% and 84%,
respectively) with Kivalues of 155 and 485 nM, respectively. Data show mean
values of nine measurements ⫾SEM. (B) Hill plot showing linearized data and
the corresponding Kivalues. (C) Dixon plot of the competitive binding inter-
action of (E)-BCP with the CP-55,940 receptor binding site in the CB2receptor.
Radioligand assays were performed by using 84 pM, 168 pM, and 252 pM
[3H]CP-55,940.
Y190
F281
M265
V113
F117
W258
I198
4.0Å
4.6Å
(E)-BCP
Fig. 3. Model of the putative interaction of (E)-BCP with the CB2receptor
determined by Surflex–Dock and MD/MM calculations. (E)-BCP is located in
the hydrophobic region of the amphipathic CB2receptor binding pocket
where it closely interacts with hydrophobic residues F117, I198, W258, V113,
and M265. In this model, significant
␲
–
␲
stacking interactions between the
(E)-BCP double bonds and F117 and W258, respectively, facilitate binding.
9100
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0803601105 Gertsch et al.

energy conformer (24). Because the Surflex-Dock scores are
expressed in ⫺log(K
d
) units, the modeling results also indicate that
(E)-BCP (⫺log(K
d
)⫽7.86) and (Z)-BCP (7.96) exhibit ⬎1,000-
fold higher CB
2
receptor binding affinities than
␣
-humulene (4.17),
which is in agreement with the experimental data shown in Fig. S2.
(
E
)-BCP is a Full CB
2
Receptor Agonist Leading to G
i
and G
o
Signals. We
next determined whether the interaction of (E)-BCP with the CB
2
receptor leads to G protein mobilization. Like the potent high-
affinity cannabinoid ligand WIN55,212–2 (K
i
hCB
2
⫽1.2 nM) (21),
(E)-BCP inhibited forskolin-stimulated cyclic adenosine mono-
phosphate (cAMP) production in hCB
2
receptor transfected
CHO-K1 cells (Fig. 4A). The difference in potency between
WIN55,212-2 (EC
50
⫽38 ⫾5.7 nM) and (E)-BCP (EC
50
⫽1.9 ⫾
0.3
␮
M) (Fig. 4A) reflects the ⬇150-fold difference in CB
2
receptor
binding affinity between these ligands. To assess whether (E)-BCP
is a fully functional agonist at the CB
2
receptor, we measured
CB
2
-mediated intracellular calcium transients ([Ca
2⫹
]
i
) in promy-
elotic HL60 cells. Like the endocannabinoid 2-arachidonoylglycerol
(2-AG), which is a potent activator of the G
o
pathway that leads to
[Ca
2⫹
]
i
release (25), also (E)-BCP concentration-dependently trig-
gered the release of [Ca
2⫹
]
i
(Fig. 4B). The maximum increase in
[Ca
2⫹
]
i
of (E)-BCP (E
max
⫽312 ⫾41%) was somewhat weaker than
the maximum effect achieved with 2-AG (E
max
⫽409 ⫾17%) (Fig.
4B). However, the E
max
and EC
50
values obtained in nonlinear
regression analyses were not statistically different for 2-AG
(EC
50
⫽13.8 ⫾0.9
␮
M) and E-BCP (EC
50
⫽11.5 ⫾2.8
␮
M). In
HL60 cells devoid of CB
2
receptor surface expression E-BCP did
not trigger [Ca
2⫹
]
i
(Fig. 4C). Moreover, the [Ca
2⫹
]
i
transients
induced by (E)-BCP and 2-AG were fully blocked by the CB
2
receptor antagonist SR144528, which supports a strictly CB
2
re-
ceptor-dependent G
o
mechanism of [Ca
2⫹
]
i
release by (E)-BCP
(Fig. 4D). BCP oxide induced significant [Ca
2⫹
]
i
transients in CB
2
positive, but induced them even more strongly in CB
2
deficient
HL60 cells (Fig. S3A), suggesting a CB
2
receptor-independent
mechanism of action. As expected, this effect could not be inhibited
by SR144528 (Fig. S3B). Because CB
2
receptor-selective agonists
have been shown to increase phosphorylation of mitogen-activated
protein (MAP) kinases p38 (26) and extracellular receptor kinases
1/2 (Erk1/2) (27), we incubated the CB
2
receptor-selective agonist
JWH133 (K
i
⫽3.4 nM) and (E)-BCP with HL60 cells and human
primary CD14
⫹
monocytes to determine the phosphorylation of
these kinases upon CB
2
receptor activation. (E)-BCP and JWH133
(1
␮
M) led to rapid phosphorylation of Erk1/2 in both HL60 and
primary CD14
⫹
monocytes, and this effect could be blocked by
prior incubation with the CB
2
receptor antagonist SR144528 (Fig.
S4). Although (E)-BCP weakly increased phosphorylated p38 in
both HL60 cells and primary CD14
⫹
monocytes, JWH133 did not
trigger p38 phosphorylation in primary CD14
⫹
monocytes.
(
E
)-BCP Inhibits Lipopolysaccharide (LPS)-Stimulated TNF-
␣
and IL-1
␤
Expression in Peripheral Blood. CB
2
receptor agonists have repeat-
edly been shown to inhibit the release of cytokines from LPS-
stimulated monocytes, such as tumor necrosis factor-
␣
(TNF-
␣
)
(14, 28, 29). To explore whether (E)-BCP triggers CB
2
receptor-
dependent effects on cytokine expression in vitro, we measured
cytokine levels in LPS-stimulated human whole blood after 18 h in
the presence and absence of (E)-BCP. At 500 nM, (E)-BCP
significantly inhibited LPS-stimulated IL-1
␤
and TNF-
␣
expression
(Fig. 5A). This inhibition was clearly reversed by the CB
2
receptor-
selective antagonist AM630 (5
␮
M), thus indicating a functional
CB
2
receptor-dependent mechanism. AM630 was used because
SR144528 potently inhibits cytokine expression in whole blood, as
reported in ref. 29. As shown in Fig. 5A, the antiinflammator y effect
90
120
150
180
210
240
(E)-BCP(E)-BCP +
SR144528
SR1
44528
[Ca
2+
]
i
(% of c ontrol)
-7 -6 -5 -4
100
200
300
400
500 2-AG
(E)-BCP
Log [M]
[Ca
2+
]i (% of vehicle contr ol)
**
-11 -10 -9 -8 -7 -6 -5 -4
40
50
60
70
80
90
100
110
WIN 55,212-2
(E)-BCP
Log [M]
cAMP (% of vehicle cont rol)
C
-7 -6 -5 -4
100
200
300
400
500 2-AG
(E)-BCP
Log [M]
[Ca
2+
]i (% of vehicle contr ol)
A
D
B
M1
M1
CB2
CB2
Fig. 4. G protein triggered effects of (E)-BCP upon
CB2receptor binding. (A) The high-affinity CB ligand
WIN55,212–2 and (E)-BCP dose-dependently inhibit
forskolin-stimulated cAMP production in CB2recep-
tor-transfected CHO-K1 cells. Data are mean values of
three independent experiments measured in tripli-
cates ⫾SEM in a nonlinear dose-response curve (R2⬎
0.97). (B) Like 2-AG, which was used as positive control,
(E)-BCP dose-dependently triggers [Ca2⫹]itransients in
CB2expressing HL60 cells. The FACS histogram shows
CB2immunofluorescence of HL60 cells. Data are mean
values of three independent experiments ⫾SEM
shown in a nonlinear dose-response curve (R2⫽0.99).
(C)No[Ca
2⫹]itransients were induced in HL60 cells
devoid of CB2surface expression. The FACS histogram
shows the lack of CB2Ab immunofluorescence in these
HL60 cells. (D) Addition of the CB2-selective antagonist
SR144528 (1
␮
M) inhibited the [Ca2⫹]irelease trig-
gered by (E)-BCP (10
␮
M) in CB2-positive HL60 cells.
Data are mean values from three independent exper-
iments ⫾SEM (paired ttest **,P⬍0.01).
Gertsch et al. PNAS
兩
July 1, 2008
兩
vol. 105
兩
no. 26
兩
9101
PHARMACOLOGY

was less pronounced with increasing concentrations of (E)-BCP,
probably reflecting the formation of (E)-BCP aggregates or a weak
simultaneous cellular stimulation of cytokines via a low-affinity
target at
␮
M concentrations. Below 500 nM, cytokine inhibition
was concentration-proportional, and (E)-BCP produced less inhi-
bition (data not shown). The LPS-stimulated expression levels of
IL-6, -8, and -10 were not significantly influenced by (E)-BCP after
18h(Fig. S5).
(
E
)-BCP Inhibits Lipopolysaccharide-Stimulated Erk1/2 and JNK1/2
Activation in Primary Monocytes. Based on the observation that
(E)-BCP weakly induced p38 and Erk1/2 phosphorylation in pri-
mary CD14
⫹
monocytes and, at the same time, inhibits LPS-
stimulated TNF-
␣
and IL1
␤
protein expression in whole blood, we
investigated whether LPS-triggered p38, Erk1/2, and JNK1/2 acti-
vation was modulated by (E)-BCP. LPS stimulation of primary
human monocytes led to a rapid and strong phosphorylation of p38
and JNK1/2, whereas a weaker activation of Erk1/2 was detected.
Incubation of cells with (E)-BCP (500 nM) for 1 h before LPS
stimulation led to a significant reduction of Erk1/2 and JNK1/2
activation (phosphorylation) as determined by Western blot (Fig.
S6A) and cytometic bead array (CBA) analyses (Fig. S6B), whereas
p38 activation was not influenced by (E)-BCP (data not shown).
Oral (
E
)-BCP Inhibits Carrageenan-Induced Edema in Wild Type Mice.
To obtain in vivo evidence of the antiinf lammatory effects induced
by (E)-BCP in vitro, we examined the effectiveness of orally
administered (E)-BCP in wild-type mice (Cnr2
⫹/⫹
) and CB
2
re-
ceptor-deficient (Cnr2
⫺/⫺
) mice. (E)-BCP (5 and 10 mg/kg) dosed
orally significantly inhibited carrageenan-induced paw edema in
wild-type mice by ⬇70% and 50%, respectively (Fig. 5B). Some-
what unexpectedly, the lowest dose of (E)-BCP was most effective
in this experiment (Fig. 5B). Because no antiinflammatory effect
could be observed with the two lower doses of (E)-BCP in Cnr2
⫺/⫺
mice, the antiinflammatory effects of (E)-BCP in wild-type mice
show that this natural product exerts CB
2
receptor-dependent
cannabimimetic effects in vivo. Moreover, 10 mg/kg (E)-BCP
showed an interesting biphasic effect, such that edema formation
was augmented after 30 min relative to vehicle control but clearly
inhibited after 120 min. This was not obser ved in Cnr2
⫺/⫺
mice, thus
demonstrating that the CB
2
receptor is involved in both effects.
Because 50 mg/kg (E)-BCP showed only a weak antiinflammatory
0
0.05
0.1
0.15
0.2
0.25
03060120240
t (min)
change in paw volume (ml)
vehicle
5 mg/Kg
10 mg/Kg
50 mg/Kg
**
**
Cnr2-/-
*
0
0.05
0.1
0.15
0.2
0.25
0 30 60 120 240
t (min)
change in paw volume (ml)
veh ic le
5 mg/Kg
10 mg/Kg
50 mg/Kg
*
**
***
Cnr2+/+
*
Cnr2-/-
B
20
40
60
80
100
120
% protein expression of stimulated
control
**
*** ***
**
0.5 µM1µM5µM
(E)-BCP
AM630 -+- +- +
IL-1β
20
40
60
80
100
120
% protein expres sion of stimulated
control
***
***
**
*
0.5 µM1µM5µM
(E)-BCP
AM630 -+- - +
**
TNF-α
A
+
020406080100
1
paw swelling (% of contr ol)
(E)-BCP 5 mg/Kg
(E)-BCP 10 mg/Kg
JWH133 10 mg/Kg
JWH133 20 mg/Kg
***
C
**
**
*
**
Fig. 5. Anti-inflammatory effects of (E)-BCP in vitro
and in vivo.(A)(E)-BCP (1 h incubation before stimu-
lation) inhibits LPS-stimulated (313 ng/ml) IL-1
␤
and
TNF-
␣
protein expression in human peripheral whole
blood (determined after 18 h). Prior incubation with
the CB2receptor antagonist AM630 (5
␮
Mfor1h)
blocked the effect of (E)-BCP. Data show values from
three independent experiments ⫾SEM (paired ttest *,
P⬍0.05; **,P⬍0.01; ***,P⬍0.001). (B) Intraplantar
carrageenan (30
␮
l)-induced edema formation is in-
hibited by orally administered (E)-BCP (5 and 10 mg/
kg) in C57BL/6J wild-type (Cnr2⫹/⫹) mice but not in CB2
(Cnr2⫺/⫺) knockout mice. Shown is the mean increase
in paw volume over time of at least nine mice per
group ⫾SEM (ANOVA, *,P⬍0.05; **,P⬍0.01; ***,
P⬍0.001). (C) Comparison of the effects of oral
JWH133 (CB2selective agonist, Ki⫽3.4 nM) and (E)-
BCP on carrageenan (30
␮
l)-induced edema formation
in C57BL/6J wild-type (Cnr2⫹/⫹) mice after 240 min.
Shown is the percentage of increase in paw volume
(relative to control) of at least nine mice per group ⫾
SEM (paired ttest, *,P⬍0.05; **,P⬍0.01; ***,
P⬍0.001).
9102
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0803601105 Gertsch et al.

activity in both wild-type and Cnr2
⫺/⫺
mice (Fig. 5B), higher doses
of (E)-BCP may lead to off-target effects. In comparison, the CB
2
receptor-selective agonist JWH133 (K
i
⫽3.4 nM) was significantly
less effective in inhibiting carrageenan-induced edema formation
than (E)-BCP after oral administration (Fig. 5C) despite its 45-fold
higher CB
2
receptor affinity.
(
E
)-BCP Content in
Cannabis
Essential Oil Correlates with CB
2
Receptor
Activation. (E)-BCP contents are known to vary in Cannabis strains
and preparations (7), and we therefore analyzed five commercial
Cannabis essential oils lacking classical cannabinoids by quantita-
tive gas chromatography (GC). The percentage of (E)-BCP in the
essential oils varied between 12.5 and 35% (Fig. S7). The (E)-BCP
content in Cannabis essential oils positively correlated with the
displacement of [
3
H]CP55,940 and the amount of [Ca
2⫹
]
i
triggered
via CB
2
receptors in HL60 cells (Fig. S7), thus confirming that this
natural product is the only cannabimimetic in Cannabis essential oil.
Discussion
The abundant plant sesquiterpene (E)-BCP is shown to competi-
tively interact with the CP55,940 binding site (i.e., THC binding
site) of the peripheral hCB
2
cannabinoid receptor with a K
i
value
in the nanomolar range (Fig. 2). Results from molecular modeling
studies suggest that (E)-BCP binds to a previously described
putative ligand binding pocket in the hCB
2
receptor (Fig. 3) (29).
Although other ligands like N-alkylamides may span the entire
amphiphatic binding pocket or bind adjacent to the water accessible
cavity (e.g., close to Y190) (29), (E)-BCP binds to the hydrophobic
cavity, where it interacts with hydrophobic residues F117, I198,
W258, V113, and M265 (Fig. 3). The C4–C5 double bond in
(E)-BCP exhibits a
␲
–
␲
stacking interaction with F117 (4.0 Å), and
thus its geometry likely plays an important role for hCB
2
receptor
binding. This hypothesis is in line with the experimental difference
in CB
2
receptor binding between (E)-BCP and (Z)-BCP (Fig. 2)
and the loss of binding affinity of BCP oxide (Fig. S2). (E)-BCP
activation of the hCB
2
receptor triggers a full stimulation program,
involving inhibition of cAMP, stimulation of [Ca
2⫹
]
i
transients, and
weak activation of the MAP kinases p38 and Erk1/2 (Fig. 4 and Fig.
S4). Like other CB
2
receptor-selective agonists, (E)-BCP leads to
antiinflammatory effects in vitro and in vivo (Fig. 5). An inhibition
of LPS-stimulated TNF-
␣
and IL-1
␤
expression (Fig. 5A) is typically
exerted by CB receptor ligands, such as CB
2
receptor agonists (14,
28, 29). However, the underlying molecular mechanism of this
effect remains to be elucidated. Our data suggest that a CB
2
receptor-mediated suppression of Erk1/2 and JNK1/2 signaling is
involved (Fig. S6). Because Erk1/2 and JNK1/2 signaling pathways
are critical for LPS-stimulated expression of IL-1 and TNF-
␣
(30),
CB
2
receptor ligands capable of inhibiting the activation of these
kinases may down-regulate IL-1
␤
and TNF-
␣
expression. Data have
shown (15) that different CB
2
receptor-selective ligands, including
CB
2
receptor agonists, are able to inhibit carrageenan-stimulated
edema formation in mice. Our results on the (E)-BCP-mediated
inhibition of carrageenan-stimulated edema in wild-type but not
(Cnr2
⫺/⫺
) mice (Fig. 5B) are in agreement with these data. Some-
what paradoxically, activation of the CB
2
receptor has also been
shown to exert proinflammatory effects in the periphery, such as in
the skin (31, 32). Thus, different cell types and the context of
stimulation appear to mediate distinctly different CB
2
receptor-
dependent immunomodulatory effects. This is also shown by our
data, where (E)-BCP in monocytes activates constitutive Erk1/2
(Fig. S4) but inhibits LPS-stimulated Erk1/2 (Fig. S6). Whether the
inverse dose-dependency of (E)-BCP and the initial increase in
edema formation in wild-type mice at 10 mg/kg reflect the action
of differentially activated G protein subsets needs to be elucidated.
However, in the carrageenan model of acute inflammation, low
doses of (E)-BCP are clearly antiinflammator y (Fig. 5B). Moreover,
in this model, (E)-BCP is more potent than the high-affinity CB
2
receptor-selective agonist JWH133 (Fig. 5C). (E)-BCP has been
reported to exert diverse antiinflammatory effects in vivo (33, 34),
including gastric antiinflammatory and cytoprotective effects.
These findings are in agreement with the pharmacology of CB
2
receptor agonists (35–37). In a recent report (38), (E)-BCP and
␣
-humulene were shown to be orally available inhibitors of hista-
mine-triggered edema and inflammation in mice with equipotent
effects as dexamethasone. Although the molecular basis for these
findings was not elucidated in this previous work, it is clear from our
own studies that
␣
-humulene does not interact with the CB
2
receptor (Fig. S2). Thus, it is possible that the in vivo effects of
macrocyclic sesquiterpenes may not be exclusively mediated
through the CB
2
receptor, which also appears to be supported by
the carrageenan experiments with 50 mg/kg oral (E)-BCP (Fig. 5B).
However, although other yet unknown targets may be involved, the
antiinflammatory effects of the low oral doses of (E)-BCP are
directly related to CB
2
receptor activation, because (E)-BCP is
ineffective in (Cnr2
⫺/⫺
) mice (Fig. 5B).
(E)-BCP is the first Cannabis-derived functional CB receptor
ligand with a fundamentally different structure from the classical
cannabinoids. Thus, it represents a new type of CB
2
receptor-
selective agonist that is based on an unusual cyclobutane-
containing scaffold. Recently (39), a Cannabis extract (Sativex) was
approved in Canada for the treatment of neuropathic pain in
multiple sclerosis. Because (E)-BCP is a major constituent in
Cannabis essential oil (Fig. S7) and shows significant cannabimi-
metic effects, it may also contribute to the overall effect of Cannabis
preparations, including Sativex. Moreover, (E)-BCP is commonly
ingested with vegetable food, and an estimated daily intake of
10–200 mg of this lipophilic sesquiterpene could be a dietary factor
that potentially modulates inflammatory and other pathophysio-
logical processes via the endocannabinoid system. Consequently,
the pharmacokinetics of (E)-BCP in humans and its potential
impact on health should be addressed in future studies.
Materials and Methods
Drugs and Antibodies. See SI Materials and Methods.
Data Analysis. Results are expressed as mean values ⫾SD or ⫾SEM for each
examined group. Statistical significance of differences between groups was
determined by the Student’s ttest (paired ttest) with GraphPad Prism4 software.
Outliners in a series of identical experiments were determined by Grubb’s test
(ESD method) with alpha set to 0.05. Statistical differences between treated and
vehicle control groups were determined by Student’s ttest for dependent sam-
ples. For animal experiments, statistical differences between treated and vehicle
control groups were determined by repeated measurements (ANOVA) and post
hoc least square difference tests. Differences between the analyzed samples were
considered as significant at Pⱕ0.05. Nonlinear regression analysis (curve fitting)
was performed with GraphPad Prism4 software.
Cell Cultures. See SI Materials and Methods.
FACS Analysis of CB2Expression. Cellular surface expression of the CB2receptor
was quantified by immunofluorescence as described in ref. 29 (for additional
details, see SI Materials and Methods).
Radioligand Displacement Assays on CB1and CB2Receptors. [3H]CP-55,940
binding and displacement experiments were performed as described in ref. 29.
(For additional details, see SI Materials and Methods.) Data were fitted in a
sigmoidal curve and graphically linearized by projecting Hill plots, which for both
cases allowed the calculation of IC50 values. Derived from the dissociation con-
stant (KD)of[
3H]CP-55,940 (0.39 nM) and the concentration-dependent displace-
ment (IC50 value), inhibition constants (Ki) of competitor compounds were calcu-
lated by using the Cheng–Prusoff equation [Ki⫽IC50/(1 ⫹L/KD)] (40).
Molecular Modeling. The CB2receptor homology model used for molecular
modeling was described in ref. 23. Docking of (E)-BCP, (Z)-BCP and
␣
-humulene
and CB2protein-ligand complex MD/MM studies were performed on the basis of
published docking protocols, using Tripos molecular modeling packages
Sybyl7.3.3 and Tripos force field (23, 29). (For additional details, see SI Materials
and Methods.)
Gertsch et al. PNAS
兩
July 1, 2008
兩
vol. 105
兩
no. 26
兩
9103
PHARMACOLOGY

cAMP Assay. Human CB2-receptorexpressing CHO-K1 cells were plated in 96-well
plates at a density of 3 ⫻105cells per ml and incubated overnight. After aspirating
the media, the cells were chilled for 10 min at room temperature in RPMI medium
1640 (w/o supplements) containing 500
␮
M 3-isobutyl-1-methylxanthine. Cells
were then treated with different concentrations of test-compounds and incu-
bated for 30 min at 37°C in a total volume of 100
␮
l. After another 30 min of
incubation at 37°C with 20
␮
M forskolin, intracellular cAMP levels were detected
by HitHunter for adherent cells EFC chemiluminescent detection assay (Amer-
sham; catalog no. 90000302) according to the manufacturer’s instructions and
measured on a Microlumat Plus Microplate Luminometer LB 96V (EG&G
Berthold). The high-affinity CB receptor ligand WIN55,212–2 was used as positive
control.
Measurement of [Ca2ⴙ]i.Intracellular Ca2⫹was quantified in HL60 cell lines by
FACS measurements as described in ref. 29. (For additional details, see SI Materials
and Methods.)
CBA Quantification of Cytokines in Human Blood Plasma. Human peripheral
whole blood cultures were obtained as described in ref. 29. Cytokine production
in human peripheral whole blood was analyzed in blood plasma of whole blood
cultured for 18 h at 37°C, 5% CO2, using Cytometric Bead Arrays (BD Biosciences;
human inflammation CBA kit 551811) as described in ref. 29. (For additional
details, see SI Materials and Methods.)
Determination of p38, Erk1/2, and JNK1/2 Activation. Phsophorylation of p38 and
Erk1/2 was analyzed in HL60 CB2-positive cells and CD14⫹peripheral blood
mononuclear cells (PBMCs). PBMCs were isolated from human buffy coats by
density gradient centrifugation as reported in ref. 29. Phosphoproteins were
quantified with CBA Phospho p38 MAPK Flex Set 560010 (T180/T182), Phospho
Erk1/2 Flex Set 560012 (T202/Y204), and the Phospho JNK1/2 Flex Set (T183/Y185)
from BD Biosciences according to the manufacturer’s instructions.
Western Blot Analysis. Western blots were carried out by standard methods (see
SI Materials and Methods).
Animals. Male CB2knockout mice (Cnr2⫺/⫺) on a C57BL6/J congenic background
(32) and their C57BL6/J (Cnr2⫹/⫹) wild-type controls ⬇3 months of age were used.
Animals were housed in groups of 3–5 and had access to water and food ad
libitum. The housing conditions were maintained at 21 ⫾1°C and 55 ⫾10%
relative humidity in a controlled light– dark cycle (light on between 7:00 a.m. and
7:00 p.m.). All experimental procedures and animal husbandry were conducted
according to standard ethical guidelines.
Animal Treatment. (E)-BCP was dissolved in olive oil (Fluka) and gavage-fed to the
animals (with the help of feeding needle) 30 min (50 mg/kg) and 60 min (5 and 10
mg/kg) before carrageenan treatment. Olive oil without (E)-BCP was gavage-fed
to the animals as vehicle control.
Carrageenan–Paw Edema. The experiments were performed as described in ref.
41. Briefly, Carrageenan (2%, 20 mg/ml suspended in saline; Sigma) was injected
intraplantar in a volume of 30
␮
l into the hind right paw, using a 27-gauge needle.
The left paw received the same amount of saline and it was used as control.
Edema was measured by using a Volume meter (TSE) at a several time points after
carrageenan injection. Edema was expressed in milliliters as the difference be-
tween the right and left paw.
GC Measurements. Gas chromatography measurements were carried out as
described in the European Pharmacopoeia 5.5 (Pinus silvestris oil) with a Thermo
Electron Focus GC instrument (Thermo Fisher Scientific) fitted with a BGB wax
column (60 m, 0.25-mm diameter, 0.25-
␮
m film; serial no. 13651937). (For addi-
tional details, see SI Materials and Methods).
ACKNOWLEDGMENTS. We thank Dr. Irmgard Werner and Alex Hermann for
their help and technical assistance for the GC measurements, Andreas Nievergelt-
Meier for his help with the CBA analyses, and Dr. Michael Detheux (Euroscreen
S.A., Brussels, Belgium) for the CB2-transfected CHO-K1 cell line. This work was
supported by the Deutsche Forschungsgemeinschaft Grants FOR926 and GRK804.
1. Pichersky E, Noel JP, Dudareva N (2006) Biosynthesis of plant volatiles: Nature’s
diversity and ingenuity. Science 311:808– 811.
2. Rasmann S, et al. (2005) Recruitment of entomopathogenic nematodes by insect-
damaged maize roots. Nature 434:732–737.
3. Orav A, Stulova I, Kailas T, Mu¨u¨ risepp M (2004) Effect of storage on the essential oil
composition of Piper nigrum L. fruits of different ripening states. J Agric Food Chem
52:2582–2586.
4. Jayaprakasha GK, Jagan Mohan Rao L, Sakariah KK (2003) Volatile constituents from
Cinnamomum zeylanicum fruit stalks and their antioxidant activities. J Agric Food
Chem 51:4344– 4348.
5. Mockute D, Bernotiene G, Judzentiene A (2001) The essential oil of Origanum vulgare
L. ssp. vulgare growing wild in vilnius district (Lithuania). Phytochemistry 57:65–69.
6. Sko¨ ld M, Karlberg AT, Matura M, Bo¨ rje A (2006) The fragrance chemical beta-
caryophyllene-air oxidation and skin sensitization. Food Chem Toxicol 44:538–545.
7. Hendriks H, Malingre T, Battermann S, Boss R (1975) Mono- and sesquiterpene hydro-
carbons of the essential oil of Cannabis sativa. Phytochemistry 14:814– 815.
8. Pertwee RG (2007) The diverse CB(1) and CB(2) receptor pharmacology of three plant
cannabinoids: Delta(9)-tetrahydrocannabinol, cannabidiol and Delta(9)-tetrahydro-
cannabivarin. Br J Pharmacol 53:199–215.
9. Matsuda LA, Lolait SJ, Brownstein MJ, Young AC, Bonner TI (1990) Structure of a
cannabinoid receptor and functional expression of the cloned cDNA. Nature 346:561–
564.
10. Munro S, Thomas KL, Abu-Shaar M (1993) Molecular characterization of a peripheral
receptor for cannabinoids. Nature 365:61–65.
11. Mackie K (2006) Cannabinoid receptors as therapeutic targets. Annu Rev Pharmacol
Toxicol 46:101–122.
12. Devane WA, et al. (1992) Isolation and structure of a brain constituent that binds to the
cannabinoid receptor. Science 258; 1946–1949.
13. Sugiura T, Waku K (2000) 2-Arachidonoylglycerol and the cannabinoid receptors.
Chem Phys Lipids 108:89–106.
14. Klein TW (2005) Cannabinoid-based drugs as anti-inflammatory therapeutics. Nat Rev
Immunol 5:400– 411.
15. Iwamura H, Suzuki H, Kaya T, Inaba T (2001) In vitro and in vivo pharmacological
characterization of JTE-907, a novel selective ligand for cannabinoid CB2 receptor.
J Pharmacol Exp Ther 296:420– 425.
16. Ibrahim MM, et al. (2005) CB2 cannabinoid receptor activation produces antinocicep-
tion by stimulating peripheral release of endogenous opioids. Proc Natl Acad Sci USA
102:3093–3098.
17. Batkai S, et al. (2007) Cannabinoid-2 receptor mediates protection against hepatic
ischemia/reperfusion injury. FASEB J 21:1788–1800.
18. Kimball ES, Schneider CR, Wallace NH, Hornby PJ (2006) Agonists of cannabinoid
receptor 1 and 2 inhibit experimental colitis induced by oil of mustard and by dextran
sulfate sodium. Am J Physiol Gastrointest Liver Physiol 291:G364–371.
19. Steffens S, et al. (2005) Low dose oral cannabinoid therapy reduces progression of
atherosclerosis in mice. Nature 434:782–786.
20. Ofek O, et al. (2006) Peripheral cannabinoid receptor, CB2, regulates bone mass. Proc
Natl Acad Sci USA 103:696–701.
21. Mukherjee S, et al. (2004) Species comparison and pharmacological characterization of
rat and human CB2 cannabinoid receptors. Eur J Pharmacol 505:1–9.
22. Stahl E, Kunde P (1973) Die Leitsubstanzen der Haschisch-Suchhunde [Leading sub-
stances for hashish narcotic dogs]. Kriminalistik, 9:385–388.
23. Chen JZ, Wang J, Xie X-Q (2007) GPCR structure-based virtual screening approach for
CB2 antagonist search. J Chem Inf Model 47:1626–1637.
24. Clericuzio M, Alagona G, Ghio C, Toma L (2000) Ab initio and density functional
evaluations of the molecular conformations of beta-caryophyllene and 6 hydroxy-
caryophyllene. J Org Chem 65:6910– 6916.
25. Shoemaker JL, Buckle MB, Mayeux PR, Prather PL (2005) Agonist-directed trafficking of
response by endocannabinoids acting at CB2 receptors. J Pharmacol Exp Ther 315:828–
838.
26. Herrera B, Carracedo A, Diez-Zaera M, Guzma´ n M, Velasco G (2005) p38 MAPK is
involved in CB2 receptor-induced apoptosis of human leukaemia cells. FEBS Lett
579:5084–5088.
27. Correa F, Mestre L, Docagne F, Guaza C (2005) Activation of cannabinoid CB2 receptor
negatively regulates IL-12p40 production in murine macrophages: Role of IL-10 and
ERK1/2 kinase signaling. Br J Pharmacol 145:441–448.
28. Klegeris A, Bissonnette CJ, McGeer PL (2003) Reduction of human monocytic cell
neurotoxicity and cytokine secretion by ligands of the cannabinoid-type CB2 receptor.
Br J Pharmacol 139:775–786.
29. Raduner S, et al. (2006) Alkylamides from Echinacea are a new class of cannabinomi-
metics. Cannabinoid type 2 receptor-dependent and -independent immunomodula-
tory effects. J Biol Chem 281:14192–14206.
30. Guha M, Mackman N (2001) LPS induction of gene expression in human monocytes. Cell
Signal 13:85–94.
31. Oka S, et al. (2006) Involvement of the cannabinoid CB2 receptor and its endogenous
ligand 2 arachidonoylglycerol in oxazolone-induced contact dermatitis in mice. J Im-
munol 177:8796– 8805.
32. Karsak M, et al. (2007) Attenuation of allergic contact dermatitis through the endo-
cannabinoid system. Science 316:1494–1497.
33. Martin S, et al. (1993) Anti-inflammatory activity of the essential oil of Bupleurum
fruticescens. Planta Med 59:533–536.
34. Cho JY, et al. (2007) Amelioration of dextran sulfate sodium-induced colitis in mice by
oral administration of beta-caryophyllene, a sesquiterpene. Life Sci 80:932–939.
35. Tambe Y, Tsujiuchi H, Honda G, Ikeshiro Y, Tanaka S (1996) Gastric cytoprotection of
the non-steroidal anti-inflammatory sesquiterpene, beta-caryophyllene. Planta Med
62:469– 470.
36. Thuru X, et al. (2007) Cannabinoid receptor 2 is required for homeostatic control of
intestinal inflammation. 17th Annual Symposium on the Cannabinoids (International
Cannabinoid Research Society, Burlington, VT), p 19. Available at www.cannabinoid-
society.org/symposium.2007/2007.icrs.program.and.abstracts.pdf.
37. Di Marzo V, Izzo A (2006) Endocannabinoid overactivity and intestinal inflammation.
Gut 55:1373–1376.
38. Fernandes ES, et al. (2007) Anti-inflammatory effects of compounds alpha-humulene
and (⫺)-trans-caryophyllene isolated from the essential oil of Cordia verbenacea. Eur
J Pharmacol 569:228–236.
39. Russo EB, Guy GW, Robson PJ (2007) Cannabis, pain, and sleep: Lessons from thera-
peutic clinical trials of Sativex, a cannabis-based medicine. Chem Biodivers 4:1729–
1743.
40. Cheng YC, Prusoff WH (1973) Relationship between the inhibition constant (K1) and
the concentration of inhibitor which causes 50 per cent inhibition (I50) of an enzymatic
reaction. Biochem Pharmacol 22:3099–3108.
41. Mogil JS, et al. (1999) Heritability of nociception I: Responses of 11 inbred mouse strains
on 12 measures of nociception. Pain 80:67–82.
9104
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0803601105 Gertsch et al.