B ec
P u
Na Borm
Ba trea
ph and a
Me ge

(ag . The
8.5 me-o
Re al an
con s incr
.00 (p

Co ng-t
no gula
the war
Ke
sul
A
com
mi
be
Bu
of
pa
sat
ge
dy
(2)
stu
sys
both food intake and rewardprocessing (3). The type1 cannabinoid
rec
sys
thr
ma
tor
pe
un
pre
foo
im
EC
of t
eat
lev
stra
of
on
hig
dro
wa
op
cha
is m
and
fol
Usi
ity
ma
in e
active, at least in AN, and a compensatory chronic upregulation of
Fro
Ad
Rec
00
doeptor (CB1R) is abundantly expressed in the central nervous
tem and induces mainly inhibition of neurotransmission
ough modulation of presynaptic neurotransmitter release, pri-
rily through retrograde signaling (4). Type 1 cannabinoid recep-
inverse agonists inhibit food intake through both central and
ripheral mechanisms (5), but their development is halted by
desired central side effects such as increased prevalence of de-
ssionandsuicidality. In contrast, cannabinoidagonists stimulate
d intake in humans and induce beneficial effects in acquired
mune deficiency syndrome related anorexia, suggesting altered
the CB1R would be the result.
Methods andMaterials
Participants and Procedure
Female AN and BN patients were recruited during hospitaliza-
tion in an inpatient university center for eating disorders. Seven
restricting and seven binging-purging AN patients were included,
as well as 16 purging BN patients, as diagnosed according to
DSM-IV criteria (12). Patientswere between17 and45 years old. The
control group (CON) consisted of 19 healthy age-matched women.
Informed consent was obtained from all participants before study
investigations, and the study was approved by the local ethics
committee and performed according to the World Medical Associ-
ation Declaration of Helsinki.
All participants were screened for absence of current medical
conditions or current psychosis and addiction and were free of any
psychoactive medication. Blood and urine testing as performed on
the day of PET scanning included screening for benzodiazepines,
neuroleptics, opiates (including synthetic), cocaine and metabo-
m theDivisionofNuclearMedicine (NG, KG, KVL), UniversityHospital and
Katholieke Universiteit Leuven, Leuven; University Psychiatric Centre
(GP), Katholieke Universiteit Leuven, Eating Disorder Clinic Kortenberg,
Kortenberg; and Laboratory for Radiopharmacy (GB), Katholieke Univer-
siteit Leuven, Leuven, Belgium.
dress correspondence to Koen Van Laere, M.D., Ph.D., D.Sc., University
Hospitals, Leuven, Division of Nuclear Medicine, Herestraat 49, B-3000
Leuven, Belgium; E-mail: koen.vanlaere@uzleuven.be.
eived Oct 26, 2010; revised Apr 19, 2011; accepted May 9, 2011.
BIOL PSYCHIATRY 2011;70:777–78406-3223/$36.00
i:10.1016/j.biopsych.2011.05.010 © 2011 Society of Biological Psychiatryrain Type 1 Cannabinoid R
atients with Anorexia and B
thalie Gérard, Guido Pieters, Karolien Goffin, Guy
ckground: The endocannabinoid system is a possible target in the
y to investigate the type 1 cannabinoid receptor (CB1R) in bulimic
thods: We investigated 16 female bulimia nervosa patients (BN) (a
e
20.5 3.6 years) using the selective CB1R ligand [18F]MK-9470
years). Statistical parametricmapping (pfamily-wise error .05) and volu
sults: Global CB1R availability was significantly increased in cortic
trol subjects (24.5%, p
.0003). Regionally, CB1R availability wa
04) and the inferior frontal and temporal cortex in AN patients only
nclusions: Global CB1R upregulation in AN patients is a possible lo
id system in anorectic conditions. There is a similarity in CB1R dysre
integration of interoceptive information, gustatory information, re
y Words: [18F]MK-9470, anorexia nervosa, bulimia nervosa, in-
a, positron emission tomography, type 1 cannabinoid receptor
norexia nervosa (AN) and bulimia nervosa (BN) are severe
psychiatric disorders with prevalence in Western countries
of .7% and up to 2%, respectively, and a poor clinical out-
e. Characteristic symptoms of AN are the refusal to maintain a
nimally normal bodyweight, an intense fear of gainingweight or
coming fat, and a disturbed perception of body shape and size.
limia nervosa is characterized by binge-eating episodes and loss
control over eating behavior. In addition, self-esteem in these
tients is largely determined by their body shape and weight (1).
Pharmacotherapeutic interventions for AN and BN remain un-
isfactory, as there is still a lack of understanding of their patho-
nesis. Functional neuroimaging has shown cortical metabolic
sfunction and changes in monoamine neurotransmitter systems
. While the serotonergic and dopaminergic systems have been
died most intensively, the endocannabinoid neurotransmission
tem (ECS)has recentlybeen recognizedas an important target ineptor Availability in
limia Nervosa
ans, and Koen Van Laere
tment of eating disorders. We used positron emission tomogra-
norectic patients.
23.8 7.1 years) and 14 female anorexia nervosa patients (AN)
control group consisted of 19 age-matchedwomen (age
25.2
f-interest analyses of CB1R availability were performed.
d subcortical brain areas in AN patients compared with healthy
eased in the insula in both AN and BN patients (p
.01 and p

.02).
erm compensatory mechanism to an underactive endocannabi-
tion both in AN and BN in the insular cortex, which is involved in
d, and emotion processing.
S neurotransmission in anorectic conditions (6). Different alleles
he CB1R gene have been associated with restricting and binge-
ing/purging AN subtypes (7). Moreover, an increase in plasma
els of the endocannabinoid anandamide has been demon-
ted in AN but not in BN (8). Also in animals, an orexigenic effect
CB1R agonists is present, even in the satiated condition (9). In
e clinical trial in primary anorexia nervosa, patients received
h doses of cannabis’ psychoactive component 9-tetrahy-
cannabinol, but no improvement in food intake or weight
s observed (10).
Recently, theCB1R-specific radioligand [18F]MK-9470wasdevel-
ed for CB1R positron emission tomography (PET) imaging and
racterized preclinically and clinically (11). Its affinity (.7 nmol/L)
uch higher than the endogenous cannabinoids anandamide
2-arachidonoylglycerol (2-AG), and its selectivity is several ten-
d higher for the CB1R over the type 2 cannabinoid receptor (11).
ng this radioligand, we have investigated in vivo CB1R availabil-
in both female AN and BN patients, in comparison with age-
tched healthy volunteers. Based on current literature on the ECS
ating disorders, we hypothesized that the ECS would be under-

lites, amphetamines, and cannabinoids. Magnetic resonance imag-
ing
an
gra
ord
(ED
dis
Ra
We
an
sis
qu
pre
rity
m
Im
me
be
[18
tio
sam
an
12
a s
sho
as
blo
rem
an
Us
im
filt
Im
to
ag
all
pa
fra
pro
cal
tis
co
of
(m
mi
bo
jec
pa
fie
the
sho
pre
bia
fic
ses
ize
wis
per
com
Da
ize
wit
cat
tio
by
ass
set
(pc
tai
.00
old
sm
de
Zu
Mo
ma
cle
po
sub
can
ana
ho
Ok
sco
Fig
sta
to V
ver
dat
sta
typ
val
778 BIOL PSYCHIATRY 2011;70:777–784 N. Gérard et al.
ww(MRI) was performed to exclude structural brain abnormalities
d for voxel-based morphometry to assess possible changes in
y matter concentration (13). Subjects completed the Eating Dis-
er Evaluation Scale (EDES) and the Eating Disorder Inventory
I) to assess psychological and clinical aspects related to eating
orders (14,15).
diotracer Characteristics and Preparation
The radiotracer [18F]MK-9470 (Merck Research Laboratories,
st Point, Pennsylvania) is an inverse agonist with a high affinity
d specificity for the humanCB1R. The precursor for tracer synthe-
was obtained from Merck Research Laboratories and subse-
ent labelingwas performedonsitewith 2-fluoroethyl bromide as
viously described (11). The radioligand had a radiochemical pu-
of 95% and a mean specific radioactivity of 261  144 GBq/
ol.
aging Procedure
All PET acquisitions were performed on an HR camera (Sie-
ns, Ehrlangen, Germany). Subjects fasted for at least 4 hours
fore the PET scan and received on average 306  55 MBq of
F]MK-9470 in a slow bolus injection, under standardized injec-
n circumstances. Full dynamicmeasurementswith arterial blood
pling were conducted for the first 10 subjects (5 AN and 5 BN)
d for the control subjects, between 0 and 75 minutes and 90 and
0minutes post injection. Thesedatawere used to assesswhether
implified noninvasive quantification procedure (see further), as
wn for healthy volunteers, was justified in these patient groups
well (16). Arterial sampling for input curve determination and
od metabolites was done as described previously (16). For the
aining patients, a simplified dynamic acquisition between 90
d 120 minutes post injection was used, without blood sampling.
ing a transmission scan (68Ge source) for attenuation correction,
ages with 4 mm resolution were reconstructed by a standard
ered backprojection algorithm.
age Processing
Each subject’s dynamic [18F]MK-9470 scan was co-registered
the corresponding volumetric magnetic resonance (MR) im-
e. Regional time-activity curves were calculated from a manu-
y defined volume-of-interest (VOI) template to assess kinetic
rameters (16). For the subjects with arterial sampling, the
ctional uptake value (fractional uptake rate [FUR]), which is
portional to the CB1R total volume of distribution (VT), was
culated as the ratio of total radioactivity concentration in
sue at the end of the scan and the integral of metabolite-
rrected plasma radioactivity from time of injection to the end
the scan (16). Also, parametricmodified standard uptake value
SUV) images were generated by summation of the 90 to 120
nutes image data, corrected for injected dose and subject’s
dy weight: mSUV
{[activity concentration (kBq/mL) (sub-
t’s weight [kg]  70)/2]/injected dose (MBq)} (17). With com-
rable peripheral tracer metabolization in subjects, this simpli-
d method can be used as indicator of CB1R availability,
reby not requiring invasive blood sampling (16). Figure 1
ws that no difference in FUR versus mSUV relationship was
sent between both groups, indicating absence of systematic
s or peripheral metabolization effects because of this simpli-
ation.
Optimized voxel-basedmorphometry (VBM)was performed to as-
s structural gray matter differences between groups (18). Normal-
d MR images were segmented into gray matter maps and a voxel-w.sobp.org/journale comparison of the local gray matter concentration was
formedwith Statistical ParametricMapping version 2 (SPM2;Well-
e Trust Centre for Neuroimaging, London, United Kingdom).
ta Analysis
For Statistical Parametric Mapping analysis, spatially normal-
d gray matter maps and mSUV PET images were smoothed
h an isotropic 10 mm Gaussian kernel and analyzed using a
egorical design. Both absolute and regional relative distribu-
ns of CB1R availability were investigated. The latter was done
normalizing activity to total cerebral values. For statistical
essment, a relative gray matter analysis threshold of 80% was
to exclude extracerebral activity, and only significant clusters
luster  .05, corrected for multiple comparisons) were re-
ned, in combination with sufficient localizing power (pheight
1 uncorrected for multiple comparisons). The extent thresh-
kext was set at 50 voxels (.4 cm
3) to minimize false-positive
all clusters.
Additionally, a VOI analysis was performed by loading a pre-
fined VOI map on all parametric PET images (PMOD; PMOD Inc.,
rich, Switzerland). This VOImapwasdrawnonanMRI template in
ntreal Neurological Institute space, representing cortical Brod-
nn areas and subcortical gray matter structures (caudate nu-
us, putamen, nucleus accumbens, pallidum, thalamus, and hy-
thalamus). Individual adjustments were performed for the
cortical brain areas based on the individual MR images. Type 1
nabinoid receptor availability values were compared using
lysis of variance and Tukey honestly significant difference post
c tests (p  .05), by use of Statistica v. 9.0 (StatSoft, Inc., Tulsa,
lahoma).
With both analysis methods, correlations with EDES and EDI
res were assessed.
ure 1. Relationship between simplified quantification using modified
ndard uptake values and fractional uptake, which is directly proportional
T (total receptor distribution volume) (16), for control subjects (n
19)
sus a subgroup of patients (n
10). Full lines and triangles
patient
a, dotted lines and circles
control subject data. Error bars indicate the
ndard error of themean. AN, anorexia nervosa; BN, bulimia nervosa; CB1,
e 1 cannabinoid; CON, control group; mSUV, modified standard uptake
ues.