Content uploaded by Jerry Nickles
Author content
All content in this area was uploaded by Jerry Nickles
Content may be subject to copyright.
High-affinity dopamine D
2
/D
3
PET radioligands
18
F-fallypride and
11
C-FLB457: A comparison of
kinetics in extrastriatal regions using a
multiple-injection protocol
Nicholas T Vandehey
1
, Jeffrey M Moirano
1
, Alexander K Converse
2
, James E Holden
1
,
Jogesh Mukherjee
3
, Dhanabalan Murali
2
, R Jerry Nickles
1
, Richard J Davidson
2
,
Mary L Schneider
4
and Bradley T Christian
1,2
1
Department of Medical Physics, University of Wisconsin—Madison, Madison, Wisconsin, USA;
2
Waisman
Laboratory for Brain Imaging and Behavior, University of Wisconsin—Madison, Madison, Wisconsin, USA;
3
Brain Imaging Center, University of California—Irvine, Irvine, California, USA;
4
Department of Psychology,
University of Wisconsin—Madison, Madison, Wisconsin, USA
18
F-Fallypride and
11
C-FLB457 are commonly used PET radioligands for imaging extrastriatal
dopamine D
2
/D
3
receptors, but differences in their in vivo kinetics may affect the sensitivity for
measuring subtle changes in receptor binding. Focusing on regions of low binding, a direct
comparison of the kinetics of
18
F-fallypride and
11
C-FLB457 was made using a MI protocol. Injection
protocols were designed to estimate K
1
,k
2
,f
ND
k
on
,B
max
, and k
off
in the midbrain and cortical regions
of the rhesus monkey.
11
C-FLB457 cleared from the arterial plasma faster and yielded a ND space
distribution volume (K
1
/k
2
) that is three times higher than
18
F-fallypride, primarily due to a slower k
2
(FAL:FLB; k
2
= 0.54 min
1
:0.18 min
1
). The dissociation rate constant, k
off
, was slower for
11
C-FLB457,
resulting in a lower K
Dapp
than
18
F-fallypride (FAL:FLB; 0.39 nM:0.13 nM). Specific D
2
/D
3
binding could
be detected in the cerebellum for
11
C-FLB457 but not
18
F-fallypride. Both radioligands can be used to
image extrastriatal D
2
/D
3
receptors, with
11
C-FLB457 providing greater sensitivity to subtle changes
in low-receptor-density cortical regions and
18
F-fallypride being more sensitive to endogenous
dopamine displacement in medium-to-high-receptor-density regions. In the presence of specific
D
2
/D
3
binding in the cerebellum, reference region analysis methods will give a greater bias in BP
ND
with
11
C-FLB457 than with
18
F-fallypride.
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994–1007; doi:10.1038/jcbfm.2009.270; published online
30 December 2009
Keywords: D
2
/D
3
; dopamine; fallypride; FLB457; multiple injection; PET
Introduction
Extrastriatal dopamine D
2
/D
3
receptors have a func-
tional role in normal cognitive processing, motivation
regulation, and reward processing; and disruption
of their circuitry may be implicated in the etiology
of several neuropsychiatric disorders. The positron
emission tomography (PET) radioligands
18
F-fally-
pride and
11
C-FLB457, both high-affinity D
2
/D
3
antagonists, provide the ability to image receptors in
low concentrations (< 1 nmol/L) and serve as valuable
biomarkers for studying the dopaminergic system. For
high-affinity radiotracers, longer scan times ( > 2h) are
required for accurate assay in high-density regions,
thus limiting the utility of
11
C-FLB457 for striatal
D
2
/D
3
measurement (Olsson and Farde, 2001). A
number of studies have been conducted to charac-
terize the in vivo kinetics of these radiotracers and to
evaluate experimental designs for measuring extra-
striatal D
2
/D
3
receptor binding. Of particular interest
is the determination of the sensitivity of these
radioligands to variations in endogenous dopamine
concentrations (Mukherjee et al, 2005; Narendran
et al, 2009; Okauchi et al, 2001; Riccardi et al,2008;
Slifstein et al, 2004b), determination of in vivo affinity
(Christian et al, 2004; Mukherjee et al, 1999; Olsson
Received 5 June 2009; revised 3 December 2009; accepted 4
December 2009; published online 30 December 2009
Correspondence: Dr NT Vandehey, Department of Medical Physics,
University of Wisconsin—Madison, Lawrence Berkeley National
Laboratory, 1 Cyclotron Road, MS55R0121, Berkeley, CA 94720,
USA.
E-mail: nickvandehey@gmail.com
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
&
2010 ISCBFM All rights reserved 0271-678X/10
$32.00
www.jcbfm.com
et al, 2004; Slifstein et al, 2004a), as well as the testing
the validity of the cerebellum as a reference region
(Asselin et al, 2007; Christian et al, 2004; Delforge
et al, 1999; Olsson et al, 1999; Olsson et al, 2004).
For assessment of endogenous dopamine competi-
tion, it has been suggested that the radiotracer
vascular transport rate constants, K
1
and k
2
, have a
significant effect on the sensitivity for detecting
endogenous dopamine release (Morris and Yoder,
2006), despite these rate constants not being directly
related to receptor binding. In a challenge-type
experiment (i.e., amphetamine challenge), competi-
tive dopamine release is evoked after administration
of radioligand; it was simulated that
11
C-raclopride
yielded the highest sensitivity for detecting dopa-
mine release in the striatum, as compared with other
radiotracers, including
18
F-fallypride,
11
C-FLB457,
and several other frequently used radioligands. The
increased detection sensitivity was not attributed
to the rate constants involved in specific binding of
the radioligand to the receptor site, but rather to
the tissue-to-plasma efflux rate constant, k
2
. A fast k
2
constant provides rapid clearance of the radioligand
from the free space after displacement by endogen-
ous neurotransmitter, thus enhancing the change
in the PET signal, which represents both bound
and free states. However, in the case of extrastriatal
D
2
/D
3
receptor binding,
11
C-raclopride does not have
sufficient target-to-background signal to detect subtle
changes in the specific binding. In these regions,
higher-affinity radiotracers are required to provide a
suitable signal from the specifically bound compo-
nent of the PET measurement.
High affinity by itself is not sufficient for extra-
striatal assay; a radiotracer must also have
low nonspecific uptake, which is dependent on
lipophilicity and nonspecific protein binding.
This issue was illustrated by a semi-quantitative
in vivo comparison of
11
C-fallypride,
11
C-FLB457
with
11
C-cyclopropyl-FLB457 (Airaksinen et al,
2006).
11
C-cyclopropyl-FLB457 has been investigated
as a candidate radioligand for extrastriatal D
2
/D
3
binding due to its high in vitro D
2
/D
3
affinity, which
is 10-fold greater than that of fallypride and FLB457
(Airaksinen et al, 2008). Semi-quantitative compa-
risons of these radiotracers showed similar target/
cerebellum ratios in the subcortical and cortical
regions, despite the significantly higher affinity of
11
C-cyclopropyl-FLB457 for the D
2
/D
3
receptors.
Designing and conducting PET experiments to
separate and uniquely identify the radioligand
delivery and binding components is challenging for
a single bolus injection PET study. Frequently there
is high covariance between the parameter estimates
of delivery (K
1
,k
2
) and binding (k
3
,k
4
), particularly
for high-affinity radiotracers with rapid specific
binding (k
3
). This high covariance limits the inter-
pretation of each parameter independently. To
effectively uncouple the in vivo parameters for the
characterization of
18
F-fallypride and
11
C-FLB457, it
is necessary to introduce several injections of ligand,
each time varying the concentrations of unlabeled
and radiolabeled ligand. These multiple-injection
(MI) PET experiments methodologically perturb and
observe the system to separate the high covariance
between parameters by introducing competition
between the labeled and unlabeled ligand for the
receptor site (Morris et al, 2004).
MI PET strategies have been used for both FLB457
and fallypride to provide in vivo estimation of radio-
ligand–receptor characteristics. Using the long-lived
76
Br (t
1/2
= 16.1 h) radiolabel, FLB457 has been eval-
uated in baboons for estimation of receptor density
(B
max
) (Delforge et al, 1999) and for assay of endo-
genous dopamine competition (Delforge et al, 2001).
For
18
F-fallypride, MI studies have been reported
using rhesus monkeys for in vivo characterization
(Christian et al, 2004) and using baboons for
measurement of in vivo affinity (Slifstein et al,
2004a). It must be stressed that MI studies are
uniquely designed to optimize the estimation of
a particular parameter of interest. For example, an
MI design occurring over several separate imaging
sessions can yield estimates of the apparent affinity
(through scatchard type of analysis) (Holden et al,
2002), but are often not suitable for uncoupling
the radioligand transport parameters (K
1
,k
2
), which
is possible with a single-session MI study. Thus,
attempting to compare the in vivo characteristics
of FLB457 and fallypride based on the literature
findings is difficult because the experiments were
not optimized for direct comparison.
The goal of this study was to perform a direct
comparison of
18
F-fallypride and
11
C-FLB457 using
the MI protocol in the rhesus monkey model. The
experiments were designed to obtain estimates of
both radioligand transport and binding parameters,
with particular interest in the tissue-to-plasma efflux
constant (k
2
) and the in vivo equilibrium dissociation
constant (K
D
). Knowledge of these radioligand
characteristics will aid in the design of future
experiments with the goal of maximizing sensitivity
to subtle differences in extrastriatal D
2
/D
3
receptor
binding and endogenous dopamine competition in
applications for studying diseases where disruptions
in the dopaminergic system are implicated.
Materials and methods
Chemical Synthesis of
18
F-Fallypride and
11
C-FLB457
The radionuclides were produced with an 11-MeV RDS
112 cyclotron (CTI, Knoxville, TN, USA). For obtaining
11
C-FLB457,
11
C was produced by static irradiation of 10%
H
2
/N
2
, producing
11
C-CH
4
in target and converted to
11
C-methyl-triflate using an automated radiochemical
system (Larsen et al, 1997). Labeling of FLB604 precursor
and subsequent high-performance liquid chromatography
(HPLC) separation were performed according to previously
described methods (Lundkvist et al, 1998). For producing
18
F-fallypride, a modified chemical-processing control unit
(CPCU) was used for labeling of the tosyl-fallypride
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
995
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
precursor, as previously described (Mukherjee et al,
1995). After evaporation of the HPLC mobile phase, both
18
F-fallypride and
11
C-FLB457 were dissolved in 0.9%
NaCl and passed through a 0.22-mm filter for injection. The
precursors and reference standards were purchased from
ABX (Radeberg, Germany). Specific activity of the radio-
tracers was determined using reference standards and
analytic HPLC analysis.
For each study, a stock solution of either unlabeled
FLB457 or unlabeled fallypride was prepared by dissolving
a reference standard in a sterile 10% ethanol/saline
solution. For the second and third injections of each
experiment, a given volume of the stock solution was
thoroughly mixed with the labeled radiotracer, lowering its
specific activity as needed per the scan injection protocol.
PET Scans
MI PET data were acquired from two male rhesus monkeys
(Macaca mulatta; M1: 7 kg, 6 years; M2: 8 kg, 4 years)
using the experimental designs described in Table 1.
The experimental procedures were approved by the
University of Wisconsin Institutional Animal Care and
Use Committee. For the scanning procedure, the monkeys
were initially anesthetized with ketamine (10 mg/kg,
intramuscular) and maintained with 0.75% to 1.5%
isoflurane for the entire duration of the scan. Atropine
was administered intramuscularly at 0.25 mg to minimize
secretions. Body temperature, breathing rate, heart rate,
and SpO
2
levels were monitored and logged during the
course of each scanning session. A catheter was placed in
the saphenous vein for the administration of ligand and
another was placed in the femoral artery for withdrawing
arterial plasma samples.
The PET scans were acquired using a Concorde micro-
PET P4 scanner (Tai et al, 2001), with the animal mounted
in a custom head holder in the prone position. After
positioning, attenuation scans were acquired for 518 secs
using a
57
Co transmission point source. Collection of
emission data was initiated with the first bolus injection
of radiotracer and continued throughout each of the
multiple injections. After the scan, the animals were
removed from the scanner bed and from anesthesia. On
recovery of the swallowing reflex, the animals were
returned to their home cage where they were monitored
until they were alert.
Input Function Determination
Arterial blood samples (B0.5 ml each) were drawn
throughout the course of each study. The sampling
frequency varied from 10 samples/minute after each
injection to 0.1 samples/minute after 30 mins. The whole
blood samples were mixed with heparinized saline and
assayed for radioactivity using a 300 NaI(Tl) well counter,
cross-calibrated to the PET scanner. Blood samples were
then centrifuged at 2200gfor 5 mins and 250 mL plasma was
extracted. The plasma samples were alkynized with 50 mL
sodium bicarbonate before performing two ethyl acetate
extractions (500 mL each) to extract the lipophilic species.
Both extractions were combined and assayed and con-
verted to radioactivity concentration units after applying a
correction for extraction efficiency (B90%). The ethyl
acetate was evaporated and thin-layer chromatography was
performed to assess the fraction of lipophilic plasma signal
arising from lipophilic metabolites.
Whole blood and parent ligand plasma time–activity
curves were then parameterized to an analytic function
with three exponents to describe the decline of the radio-
tracer, as given in Figure 1A. Curve stripping was used
to generate separate time courses corresponding to each
injection subject to a constraint that the slowest component
of decline was equal for all injections of each study. The
parent ligand plasma time–activity data were converted to
units of molar concentration (pmol/mL) through division
by the injected specific activity. A decay correction was
Table 1 Experimental design
Animal/experiment M1a M2a M1b M2b M1c
Tracer
18
F-fallypride
18
F-fallypride
11
C-FLB457
11
C-FLB457
11
C-FLB457
Monkey scan day
a
0 166 16 0 220
t
1
(mins) 0 0 0 0 0
a
1
(MBq) 130 120 120 120 60
m
1
(nmol) 2.0 0.7 1.8 0.7 1.0
SA
1
(GBq/mmol) 67 170 67 180 59
t
2
(mins) 100 100 30 40 100
a
2
(MBq) 120 110 130 130 130
m
2
(nmol) 13.1 10.6 11.6 12.8 11.4
SA
2
(GBq/mmol) 9.3 10.3 10.7 9.6 11.5
t
3
(mins) 150 — 60 81 150
a
3
(MBq) 47 — 12 19 117
m
3
(nmol) 100 — 109 102 111
SA
3
(GBq/mmol) 0.5 — 0.1 0.2 1.1
t
end
(mins) 180 150 120 121 182
Estimated maximum receptor
occupancy (%) 13%, 64%, 98% 12%, 77% 38%, 97%, 100% 19%, 94%, 100% 23%, 98%, 100%
a
Monkey scan day represents number of days after the monkey’s first MI study.
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
996
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
applied to the molar concentration to represent the ‘cold’
ligand, which does not undergo radioactive decay.
Image Processing
The PET list mode data were binned into sinograms with
durations of 30 seconds per frame, with corrections applied
for scanner deadtime and random coincidence events.
Emission sinograms were reconstructed with filtered
backprojection using a 0.5 cm
1
ramp filter, zoom of
1.5, and 128 128 63 voxel matrix size with a voxel size
of 1.26 mm 1.26 mm 1.21 mm, with corrections applied
for attenuation, scanner normalization, and scatter to
create images with quantitative units. The data were not
decay-corrected, as radioactive decay is accounted for in
the model.
The reconstructed time series for each monkey were
spatially transformed into a common space for generation
Figure 1 Model input function description and fitting results. (A) The model input function used in fitting the measured input
functions. (B) Fraction of the injected dose, time shifted to t= 0. Each curve represents the average of all injections of a particular
study. The dotted lines are for fallypride, solid lines are for FLB457. (C) The measured input function radioactivity from the M1a
fallypride study (circles) and the fit of measured data, as fit with three separate input functions (dotted lines) and the sum of the three
input functions (solid line). The horizontal bar indicates the time period where no blood samples were drawn due to an obstructed
arterial draw line. (D) Three input functions scaled by specific activity, shown in pmol/cc (M1a fallypride). (E,F) Same as in panels
Cand D, but for FLB457 (M1c).
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
997
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
of regional time–activity curves. A rigid-body registration
was performed to the integrated images of
11
C-FLB457 and
18
F-fallypride using the FSL flirt software, and the affine
transformation was applied to the entire dynamic PET
series (Smith et al, 2004). Regions of interest (ROIs) were
centrally placed within the boundaries of each region on
the integrated PET image and applied to the dynamic time
series for each study. Time–activity curves were obtained
for the general regions of the thalamus, substantia nigra
(SN), temporal cortex, prefrontal cortex, caudate, and
cerebellum. The ROI placement for each region is shown
in Figure 2. The thalamus region (0.13 cm
3
) was placed
predominantly over the medial dorsal portion of the
thalamus; the SN (0.14 cm
3
) over the pars compacta and
other SN sub-regions unidentifiable in PET images; the
temporal cortex region (TmpCtx) (0.87 cm
3
) represented the
superior temporal sulcus; and the frontal cortex region
(1.26 cm
3
) encompassed the dorsal prefrontal cortex (PFC).
The caudate time–activity data were also incorporated
into the parameter estimation process (although not for
estimation of receptor density). The caudate (caud) region
was defined over the central area of the caudate head using
a volume of 0.16 cm
3
. No distinction was made between
left and right regions. The cerebellum ROI (cbm) (0.63 cm
3
)
was drawn over the cortex of the cerebellar lobes and
positioned to avoid signal from the white matter, vermis,
and surrounding regions.
Optimization of Experimental Design
A single-session, MI design using the labeled–unlabeled
model describes the in vivo kinetics of the ligands (Delforge
et al, 1990). The MI experiments were designed with the
goal of yielding optimal parameter identifiability for the
PET model parameters describing plasma–tissue transport
[K
1
,k
2
] and ligand–receptor binding [f
ND
k
on
,k
off,
B
max
]
(Innis et al, 2007), with special focus on regions of low
binding, such as the cortex. The protocol design involved a
multiple-step optimization procedure, first using sensitiv-
ity analysis and then Monte Carlo (MC) methods. Initial
designs investigated protocols with three serial injections
of radiolabeled and unlabeled ligand. The entire experi-
mental design was constrained to be less than 3 hours.
Sensitivity analysis was performed to determine the
injection protocol that maximizes the determinant of the
reduced hessian matrix (H
R
) based on the preliminary
parameter estimates reported for
18
F-fallypride (Christian
et al, 2004) and
76
Br-FLB457 (Delforge et al, 1999), using
arterial plasma input functions obtained from preliminary
studies. Only binding parameters (f
ND
k
on
,B
max
,k
off
) were
included in H
R
. The parameters for the optimization
algorithms included injection times for the second and
third injections as well as the radioactivity and mass
injected for each of the three injections. The optimization
was performed to select the protocols that yielded the
highest parameter precision in the H
R
.
MC simulations were then performed to assess the effects
of noise on parameter precision and to examine the
sensitivity for estimation of K
1
and k
2
, which were not
included in H
R
during the initial sensitivity analysis. These
simulations were performed by simulating 100 instances
of MI PET curves, adding noise to the simulated data,
ROI
model
, using a noise model similar to a previously
published method (Logan et al, 2001):
ROIðtÞ¼ROImodelðGð0;1ÞÞc1ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ROImodel
pÞþGð0;1ÞÞc2
Subscripts ‘*’ and ‘model’ represent simulated noise-added
and pre-noise (simulated) data, and G(0,1) represents a
normally distributed random number with mean 0 and
standard deviation 1. Constants c
1
(0.5 to 2.5) and c
2
(0 to
75) represent noise levels, each adjusted to mimic the noise
level of the ROI being simulated. The noisy simulated data
(ROI
*
(t)) were fit using methods as described below. The
protocol with fits giving the most accurate and precise
estimation of the input parameters was used for the PET
scanning sessions.
Both sensitivity analysis and MC simulations were
repeated before each PET scanning session, using refined
starting estimates based on measured data from earlier
studies. After the initial
18
F-fallypride study (M1a), the
optimization protocol was revised, leading to the second
18
F-fallypride study (M2a) using a protocol with only
two injections. After the first
11
C-FLB457 study (M1b),
the transport (K
1
,k
2
) and k
off
estimates were updated
for the sensitivity analysis procedures. The sensitivity
analysis results suggested multiple protocols that were
further examined using the MC methods, leading to
selection of a new protocol for M2 (M2b) with an increase
in the time between injections. After acquisition of the
Figure 2 Representative locations of ROIs over the six regions used for analysis. ROIs are shown on a single slice (
18
F-fallypride, 5 to
20 mins), but were drawn on multiple adjacent slices.
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
998
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
second
11
C-FLB457 study, the protocol was revised in
the same manner leading to another protocol for a third
11
C-FLB457 experiment (M1c), which was slightly adjusted
to match the timing in M1a. The experimental protocols
used for the multiple-injection (MI) studies are given in
Table 1.
Multiple Injection (MI) Model
Model Description: For MI experiments with three injec-
tions (i= 1, 2, 3), there are a total of three differential
equations describing the rate of change of the free ligand
concentration (dF/dt) and another three differential equa-
tions describing bound (dB/dt) ligand concentration,
which are as below:
dFi
dt¼K1Cpik2FifNDkon ðBmax X
i
BiÞFiþkoff Bi
dBi
dt¼fNDkon ðBmax X
i
BiÞFiþkoff Bi
These six differential equations share the same values
for the plasma-to-tissue rate constant (K
1
), tissue-to-plasma
efflux constant (k
2
), the ligand–receptor association
(f
ND
k
on
), dissociation (k
off
) rate constants, and receptor
density (B
max
). The molar concentration of the ligand in the
arterial plasma (C
pi
) for each injection serves as the input
function to the model. The output signal measured by the
PET scanner (Model
j
) is related to the state variables by
the equation:
Modelj¼1
tj
end tj
begin Z
tj
end
tj
begin X
i
½SAiðtÞð1FVÞðFiðtÞþBiðtÞÞ þ FVCWBðtÞ
()
dt
with jdynamic PET frames and specific activity injection
of SA
i
for each injection, and t
begin
and t
end
the beginning
and ending times of each frame, respectively. The blood-
borne component in each region is modeled as the whole
blood time course (C
WB
(t)) multiplied by the fractional
blood volume (F
V
), which was set at F
V
= 0.04.
Parameter Estimation: The following assumptions were
used in an effort to increase parameter identifiability in
these regions of the brain with varying levels of dopamine
D
2
/D
3
receptors:
(i) Both
18
F-fallypride and
11
C-FLB457 were assumed to
bind to the same set of D
2
/D
3
receptors in the brain.
Thus, receptor density B
max
was constrained to a single
value for both radiotracers.
(ii) The in vivo association (k
on
) and dissociation (k
off
) rate
constants, and the non-displaceable (ND) free fraction
(f
ND
) are uniform across the examined brain regions for
both radiotracers.
Model configuration and parameter estimation were per-
formed using the COMKAT software algorithms (Muzic
and Cornelius, 2001). A constrained non-linear search
using the Levenberg–Marquardt optimization algorithm
was performed for estimation of the kinetic parameters.
The parameter estimates were made both on an individual
ROI basis and simultaneously in multiple regions using an
objective function that minimized the total sum of squares
residual, RSS, across NROIs consisting of Jframes each,
as given below.
RSSðpÞ¼X
N
n¼1X
J
j¼1
wn;jðPETn;jModeln;jÞ2
Depending on what parameters we were trying to esti-
mate, a subset of the pwas used, where p¼½fND kon;koff ;
½K1n;K2n;Bmaxnn. Uniform weighting (w
n,j
= 1) was used for
all of the frames (which were of equal duration) (Muzic and
Christian, 2006). All rate constants (K
1
,k
2
,f
ND
k
on
,k
off
) were
constrained to fall within the bounds of [0,1] (min
1
,
ml cm
3
min
1
). B
max
was constrained to [0,100] (nmol L
1
).
Estimation of the individual rate constants was per-
formed in a stepwise manner, incorporating the assump-
tions outlined above, following these three steps:
Step 1. Determination of k
off
:The identifiability of k
off
is
greatest in the regions of the brain with high specific recep-
tor binding, particularly during the period of radioligand
displacement. The PET time series from the caudate,
thalamus, and SN were used to obtain the estimate of k
off
,
using the parameter set p¼½fNDkon ;koff;½K1n;K2n;Bmaxnn
and n = 1,2,3 for these regions. It should be noted that in
the caudate, only k
off
could be sufficiently uncoupled from
the other parameters and this region was not used for
estimation of the other parameters. These methods pro-
vided a single k
off
value for each individual study (and
radiotracer).
Step 2. Determination of regional B
max
:The PET time
series from the thalamus, SN, PFC, and TmpCtx were then
used for estimation of the parameter set p¼½fNDkon ;
½K1n;K2n;Bmaxnn, using a fixed value for k
off
as determined
from Step 1. Thus f
ND
k
on
was constrained to be uniform
across all brain regions, whereas regional differences were
accounted for by differing receptor density and plasma
transport rates. A mean B
max
was then obtained for each
brain region based on an average of the data obtained from
the
18
F-fallypride and the
11
C-FLB457 studies.
Step 3. Determination of f
ND
k
on
and regional K
1
,k
2
:Using
fixed values for k
off
and B
max
, the PET time series was then
used for the final estimation of
p¼½fNDkon ½K1n;K2nn
The apparent equilibrium dissociation constant was then
calculated for each radiotracer in each monkey as follows:
K
Dapp
=k
off
/f
ND
k
on
.
Fitting Cerebellum Data: In the cerebellum, the rate
constants were determined using both a one-compartment
model (1CM) and a two-compartment model (2CM). For
1CM fitting, we used p=[K
1
,k
2
] and for 2CM fitting we
used p=[K
1
,k
2
,B
max
], with f
ND
k
on
and k
off
fixed to values
determined using the methods described above from the
other brain regions. The Akaike Information Criterion
(Akaike, 1974) was calculated for comparing the models
to examine whether the presence of the additional term for
receptor density (B
max
) in the 2CM model was justified.
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
999
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
Estimation of Uncertainty
Uncertainty in parameter estimates was performed using
MC methods similar to those described by Salinas et al
(2007). Noise-free data were first simulated on the basis of
the implemented experimental protocols and final para-
meter estimates. Noise was then added in a manner
identical as described in the optimization section. A total
of 65 trials were run for each radiotracer (five ROIs each).
Multi-step fitting procedures were used as described
above. The standard deviation (s.d.) and mean of the
parameter estimates across trials were calculated to give a
coefficient of variation (COV = s.d./mean).
Results
Input Function Determination
The results of the input function fitting procedure
for the M1
18
F-fallypride study are shown in Figures
1C–1F. Figure 1B shows a comparison of the arterial
plasma time–activity curves of parent radioligand for
11
C-FLB457 and
18
F-fallypride. The data are normali-
zed to the injected dose and shown for the first
40 mins after injection and averaged over the three
injections. These plots show that native
11
C-FLB457
was cleared from the arterial plasma faster than native
18
F-fallypride. The fraction of parent compound was
2 to 4 times higher for
18
F-fallypride than for
11
C-
FLB457 at approximately 5 mins after injection. The
faster rate of clearance of
11
C-FLB457 continued
throughout the course of the study, with the slowest
exponential component of 0.033 min
1
for
11
C-FLB457
and 0.017 min
1
for
18
F-fallypride, on average.
Optimization of Experimental Design
Using MC methods, it was found that the identifia-
bility of K
1
and k
2
was not affected by the experi-
mental design within the range of schemes required
for f
ND
k
on
,k
off
, and B
max
identifiability. For
18
F-
fallypride, the parameter identifiability could be
achieved with only two injections, thus the saturat-
ing dose (injection-3) was eliminated from the fall-
ypride protocol for M2. For
11
C-FLB457, the initial
parameter estimates used for optimization led to a
protocol with injection times at 0, 30, and 60 mins
(see Table 1). This preliminary study did not ade-
quately identify B
max
and f
ND
k
on
, yielding a correla-
tion between parameter estimates greater than 0.9. To
better identify f
ND
k
on
and B
max
, this protocol was then
refined by increasing the time interval between the
injections for the second and third
11
C-FLB457
experiments.
Parameter Estimates
Of the specific binding parameters f
ND
k
on
,B
max
, and
k
off
; identifiability was greatest for k
off
, showing
a small covariance with the other parameters. k
off
was found to be slower for
11
C-FLB457 than for
18
F-fallypride, with an average value of 0.016 min
1
and 0.022 min
1
, respectively (Table 2). As indicated
in the methods, the estimate of k
off
was best for M1b
due to the long period of scanning after the third
injection, but other parameters were most identi-
fiable from the M1c study. Accordingly, the M1
11
C-
FLB457 results presented in Table 2 represent an
average of data from M1b and M1c.
All of the experiments for
11
C-FLB457 and
18
F-
fallypride showed high identifiably for both trans-
port parameters and ligand–receptor interaction
parameters, with the exception of f
ND
k
on
for
11
C-
FLB457 for M2. For this study, a range of estimates
for f
ND
k
on
and k
off
are reported that provided similarly
acceptable fits to the data, with the average values
reported in Table 2. Within this range estimates
for f
ND
k
on
and k
off
,K
1
and k
2
were largely unaffected,
due to the small covariance between the parameters
of transport and binding. When averaged across both
monkeys, K
Dapp
for
18
F-fallypride is found to be
higher than that of
11
C-FLB457 (0.39 nmol/L versus
0.13 nmol/L). Figure 3A shows an example of the
measured PET data and the model fit to the data.
The precision of the parameter estimates is given
as COV, as derived by MC-based methods. For
11
C-
FLB457 fits, the most precise COVs were for K
1
and
k
2
estimates (2%), followed by B
max
(6%), k
off
(9%),
and f
ND
k
on
(12%). The error associated with
18
F-
fallypride estimates gave COVs of 4% for K
1
and
k
2
, 5% for B
max
, and 3% for f
ND
k
on
and k
off
.
Cerebellar Kinetics
The parameter estimates for cerebellar data are
shown in the bottom of Table 2. An example of the
data fit for M1 is shown in Figure 3B, comparing both
2CM and 1CM models. For both
11
C-FLB457 studies,
the 2CM model provided the most appropriate fit
using the Akaike information criterion, which is
visually evident in Figure 3B. For
18
F-fallypride, the
1CM model was adequate for describing the data. It
was also found that
11
C-FLB457 showed consider-
ably higher ND volume of distribution, V
ND
(=K
1
/k
2
),
than
18
F-fallypride; primarily attributable to the
lower k
2
of
11
C-FLB457. This difference in V
ND
holds
for both 1CM and 2CM models.
Discussion
The motivation for this study was to perform a direct
comparison of the kinetics of two commonly used
PET radioligands for assaying extrastriatal D
2
/D
3
binding,
18
F-fallypride and
11
C-FLB457. Although
a single-bolus-injection experimental design is most
feasible for studying changes in receptor–ligand
binding (through DVR or BP
ND
) in humans, such a
design cannot uncouple the individual transport and
binding processes of the radiotracer. We have chosen
to implement this experimentally complex protocol
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
1000
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
to directly compare and characterize the in vivo rate
constants of
11
C-FLB457 and
18
F-fallypride to guide
experimental design for future studies and to
evaluate the strengths and weaknesses of each
radiotracer for extrastriatal D
2
/D
3
assessment.
Plasma Analysis
Measurements show that
11
C-FLB457 is removed from
the blood much more quickly than
18
F-fallypride.
Rapid metabolism and clearance of radioligand from
the plasma could be advantageous as it permits a
shorter scanning duration to achieve a stable measure
of receptor binding. However, the accuracy of the
measured arterial concentration is greatly diminished
due to low counting statistics due to the short half-life
of the
11
C radiolabel.
The rapid metabolism for both tracers resulted in
hydrophilic species, which did not cross the blood–
brain barrier. At later time points the lipophilic
metabolites were present in the ethyl acetate extrac-
tion along with the parent compound as assayed
by thin-layer chromatography. The fraction of
non-parent lipophilic species did not exceed 20%
for either radiotracer at the time prior to subsequent
injection. However, the uncertainty in this measure-
ment was high, particularly for
11
C-FLB457, and
corrections for the presence of these radiolabeled
species were not applied to the input functions.
Previous analysis of MI experiments has shown that
with MI protocols involving multiple injections of
radioligand, the presence of the lipophilic fraction
had a negligible effect on the parameter estimates
because the relative proportion remained small with
the addition of parent compound at each injec-
tion (Christian et al, 2004). These effects of lipophilic
metabolites on the parameter estimates were further
examined for both
18
F-fallypride and
11
C-FLB457
through simulations of various levels of lipophilic
metabolites. As a greater fraction of lipophilic plasma
concentration was attributed to lipophilic metabolites
Table 2 Results of fitting procedures
M1 M2
18
F-fallypride
11
C-FLB457
18
F-fallypride
11
C-FLB457
M1a M1b+M1c M2a M2b
f
ND
k
on
nmol/L/min 0.046 0.13 0.080 0.15
a
k
off
min
1
0.024 0.014 0.021 0.018
a
k
Dappb
nmol/L 0.52 0.11 0.25 0.15
a
PFC K
1
ml/cm
3
/min 0.22 0.49 0.55 0.53
k
2
min
1
0.37 0.13 0.46 0.20
B
max
nmol/L 0.6 0.4
TMP K
1
ml/cm
3
/min 0.43 0.51 0.58 0.53
k
2
min
1
0.53 0.14 0.52 0.21
B
max
nmol/L
1
0.6 0.4
Thalamus K
1
ml/cm
3
/min 0.42 0.66 0.76 0.66
k
2
min
1
0.49 0.18 0.68 0.25
B
max
nmol/L 1.5 0.8
SN K
1
ml/cm
3
/min 0.35 0.36 0.41 0.40
k
2
min
1
0.54 0.10 0.53 0.19
B
max
nmol/L 3.0 2.3
Cbm (2CM) K
1
ml/cm
3
/min 0.57 0.65 0.51 0.53
k
2
min
1
0.71 0.24 0.56 0.21
B
max
nmol/L 0.0 0.4 0.0 0.1
V
NDc
ml/m
3
0.8 2.7 0.9 2.5
Cbm (1CM) K
1
ml/cm
3
/min 0.57 0.64 0.51 0.51
k
2
min
1
0.71 0.22 0.54 0.19
V
Td
ml/cm
3
0.8 2.9 0.9 2.7
Abbreviations: Cbm, cerebellum; 1CM, one-compartment model; 2CM, two-compartment model; PFC prefrontal cortex; SN, substantia nigra; TMP, temporal
cortex.
The values in bold indicate values the most appropriate model, based on the Akaike Information Criterion.
a
The values reported for M2 FLB are averages of estimates in a range k
on
= 0.1 to 0.2, k
off
= 0.010 to 0.025, and K
Dapp
=0.1 to 0.2.
b
K
Dapp
=k
off
/f
ND
k
on
. This is a calculated value based on f
ND
k
on
and k
off
, and was not a parameter included in the fit to the data.
c
V
ND
=K
1
/k
2
for 2CM. This is a calculated value based on K
1
and k
2
, and was not a parameter included in the fit to the data.
d
V
T
=V
ND
=K
1
/k
2
for 1CM. This is a calculated value based on K
1
and k
2
, and was not a parameter included in the fit to the data.
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
1001
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
(up to the extreme case of a 50% lipophilic metabo-
lites at 100 mins), the associated change in kinetic
parameter estimates was only significant in f
ND
k
on
and
resulted in an approximately 10% reduction in this
parameter.
Optimization of Experimental Design
MI experiments must be properly designed for the
injection timing and the proportion of radiolabeled
and unlabeled compound to yield estimated para-
meters that are uniquely identified. However, MI
optimization procedures present a challenge because
the procedures used require aprioriinformation
about the transport and binding characteristics of a
tracer when the goal of the experiments is to deter-
mine these very parameters. On the basis of our
previous experience with
18
F-fallypride (Christian
et al, 2004), it was possible to achieve excellent
identification of all the parameters with
18
F-fallypride
in both animals. The results of the first
18
F-fallypride
study (M1a) led to elimination of the third injection
for the second study (M2a) on the basis of a robust
design-optimization strategy described by Salinas
et al (2007). Finding a protocol giving good para-
meter estimation for
11
C-FLB457 was not so straight-
forward. We acquired two scans (M1b, M2b) with
poor identifiability for f
ND
k
on
and B
max
before finding
a protocol that was able to sufficiently identify these
parameters of binding (M1c). This was achieved
by increasing the interval between the injections to
provide additional data, which were needed to
adequately uncouple the delivery and rapidly bind-
ing processes.
Model Assumptions
Several constraints were enforced to increase the
identifiability of the estimated in vivo rate cons-
tants (by minimizing the total number of simulta-
neously estimated parameters). For
11
C-FLB457 and
18
F-fallypride, the ligand–receptor association (k
on
)
Figure 3 Representative time–activity curves and model estimations. (A) A summary of model estimation and measured data for the
thalamus, SN, prefrontal cortex, and temporal cortex (in M1). (B) A summary of cerebellar data fits with both 1CM and 2CM models.
The 1CM is represented by a dotted line and the 2CM is shown with a solid line. The measured data have been re-binned from a
30-secs frame duration into longer frame durations for illustrative purposes only.
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
1002
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
and dissociation (k
off
), and the ligand-free fraction
(f
ND
) were assumed to be uniform across all regions of
the brain for each individual. This assumption of a
uniform apparent K
D
across brain regions also
implies that synaptic dopamine concentration is
uniform across regions. The estimation of k
off
can
be made with high precision (B3%) in the brain
regions where the PET signal is dominated by the
specifically bound compartment, which is visually
evidenced by displacement of radioligand after
saturating doses of ligand. This was the motivation
for including the region of the caudate for the
estimation of k
off
only. To further validate the use of
a uniform k
off
, each of the regions (caudate, thalamus,
SN) were first fit independently for k
off
and the
resulting estimates were all within 30% of the shared
value. Consistent estimates of k
off
across medium-
and high-density D
2
/D
3
regions have also been
previously reported for these radiotracers (Delforge
et al, 1999, Christian et al, 2004). In the low-receptor-
density regions, the use of a fixed k
off
serves to mini-
mize the coupling with the transport and forward
binding constants. The assumption of uniform f
ND
across regions is commonly assumed in all PET
methods of analysis using a reference region to
account for the ND compartment (e.g., BP
P
and BP
ND
)
(Innis et al, 2007). As a result, the constraints of
f
ND
k
on
and k
off
across all regions of the brain will
yield a uniform apparent K
D
for each radioligand.
The concentration of receptor sites available for
binding (B
max
) was also constrained to be equal for
11
C-FLB457 and
18
F-fallypride. To accommodate this
assumption, both radiotracers must not only have
high selectivity for the D
2
/D
3
receptors, but also
possess a similar ratio of D
2
and D
3
affinities. In vitro
comparisons of D2
long
and D3 affinities have been
reported for fallypride (Ki
D2long
:Ki
D3
= 2.2 nmol/L:
1.6 nmol/L) and FLB457 (Ki
D2long
:Ki
D3
= 0.65 nmol/L:
0.42 nmol/L) using
3
H-spiperone, thus showing a
consistent relation for both ligands (Stark et al, 2007).
Some discrepancy was reported when compa-
ring the D2
short
and D2
long
isoform affinities between
fallypride (Ki
D2long
:Ki
D2short
= 2.2 nmol/L:2.1 nmol/L) and
FLB457 (Ki
D2long
:Ki
D2short
= 0.65 nmol/L:1.6 nmol/L). In
our data, there was no apparent bias for the estimation
in the region of the SN, believed to consist predomi-
nantly of D2
short
receptors (Kahn et al, 1998), by
assuming an identical B
max
for
11
C-FLB457 and
18
F-
fallypride binding, and the constraint was applied to be
consistent with the other brain regions. Accordingly,
the constrained values for B
max
were similar to the
values as derived independently (i.e. M1 SN B
max
:
fallypride = 2.8 nmol/L, FLB457 = 3.3 nmol/L, average =
3.0 nmol/L).
Regional Binding and Transport
Examination of the results show that the main
differences between radiotracers lie in their affinity
for D
2
/D
3
receptors and the rate at which they clear
out of tissue from the free space. Using the averages
of both studies, we report an apparent K
D
(in vivo,
K
Dapp
=k
off
/f
ND
k
on
) of 0.39 nmol/L for
18
F-fallypride
and 0.13 nmol/L for
11
C-FLB457. This three-fold
difference can be attributed to the combination of
faster binding and slower dissociation of
11
C-FLB457
as compared with that of
18
F-fallypride. This differ-
ence is relatively consistent, with a twofold differ-
ence in humans (Narendran et al, 2009) and in vitro
measures of K
D
(0.030 nmol/L versus 0.018 nmol/L)
(Halldin et al, 1995; Mukherjee et al, 1995) and K
i
(2.2 nmol/L versus 0.65 nmol/L, using
3
H-spiperone)
for fallypride and FLB457, respectively. For fally-
pride, the measured apparent dissociation constant
is in close agreement with 0.38 nmol/L from previous
studies using the rhesus monkey (Christian et al,
2004), but lower than 1.3 nmol/L as reported for
the baboon (Slifstein et al, 2004a). For FLB457, the
literature values have been reported as 0.21 nmol/L
for the baboon (Delforge et al, 1999) and 0.35 nmol/L
and 0.9 nmol/L for humans (Olsson et al, 2004;
Suhara et al, 1999).
The delivery of the parent compound from the
plasma to the free space of the tissue (K
1
) was similar
for both radiotracers, yielding a mean of K
1
= 0.48±
0.15 mL/cm
3
/min for
18
F-fallypride and K
1
= 0.54±
0.10 mL/cm
3
/min for
11
C-FLB457 when averaged
across all brain regions. However, there was a large
difference in the tissue-to-plasma efflux constant,
k
2
, with values of 0.54±0.10 min
1
for
18
F-fallypride
and 0.18±0.05 min
1
for
11
C-FLB457. On the basis of
these findings, it is seen that the ND distribution
volume (V
ND
=K
1
/k
2
) is more than three-fold larger for
11
C-FLB457 than
18
F-fallypride: 3.0 mL/cm
3
as com-
pared with 0.9 mL/cm
3
. This measure of V
ND
for
18
F-
fallypride is consistent with our previous findings
using the rhesus monkey through the MI approach
(Christian et al, 2004). For
76
Br-FLB457, previous MI
experiments reported a V
ND
of 0.35 mL/cm
3
for the
baboon (Delforge et al, 1999). Recent studies of
humans found a V
ND
of 2.9 mL/cm
3
for
11
C-FLB457
(Asselin et al, 2007), which is in closer agreement
with the value reported here. This large discrepancy
in V
ND
for FLB457 may be species or methodology
dependent; however, it does emphasize the impor-
tance of performing radioligand comparisons on the
same subjects.
The significant difference in V
ND
between
11
C-
FLB457 and
18
F-fallypride may, in part, be due to
the higher lipophilicity of fallypride. Generally,
higher lipophilicity produces a reduction in the
free compound in both the plasma (f
p
) and brain
tissue (f
ND
), and V
ND
depends on the ratio of these
fractions at equilibrium. On the basis of literature
values using HPLC measurement,
18
F-fallypride has a
log k
w
= 2.43 (Mukherjee et al, 1995) and
11
C-FLB457
has a log k
w
= 1.89 (Schmidt et al, 1994). In silico
measurements support the higher lipophilicity of
fallypride, with an average log P-value of 3.3 for
fallypride and 2.9 for FLB457, as calculated using
ALOGPS (Tetko and Yu, 2005).
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
1003
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
Effect of k
2
Differences
One of the primary differences in the in vivo
behavior of
11
C-FLB457 and
18
F-fallypride is the
tissue-to-plasma efflux constant (k
2
). For
11
C-FLB457,
the approximately threefold reduction in k
2
results in
a significantly higher signal in the ND compartment,
frequently measured in the cerebellum. Such an
increase in this component of the PET signal is
advantageous for radiotracers that require an ex-
tended PET scanning duration to achieve BP
ND
stability by improving the statistics in the PET signal.
On the basis of the examination of M1 in this study, it
was found that at 80 mins after injection, the decay-
corrected
11
C-FLB457 study yielded a cerebellar
concentration that was 70% greater than that in the
18
F-fallypride study, despite having a plasma con-
centration that was 85% less than that of
18
F-
fallypride. This increased signal in the ND compart-
ment also improves visualization of the cortical
regions, which can aid in the definition of ROIs or
in spatial co-registration or normalization. However,
as the signal-to-noise ratio is dependent on measured
counts (and not decay corrected counts), this
advantage of increased V
ND
is lost for
11
C-FLB457 as
compared with
18
F-fallypride, due to its shorter half-
life of the radiolabel.
The k
2
parameter also plays a role in the sensitivity
of a radioligand for measuring changes in endogen-
ous dopamine competition. Through use of computer
simulations, Morris and co-workers performed a
detailed comparison of prospective radioligands
available for measuring dopamine transmission in
the brain (Morris and Yoder, 2006). In their survey of
these radiotracers, it was found that fallypride was
almost three-fold more sensitive than FLB457 to
dopamine displacement. In the case of dopamine
displacement, the radioligand is administered prior
to a manipulation (i.e. administration of amphet-
amine), which induces the release of endogenous
dopamine, and in turn displaces the radiotracer from
the specifically bound receptors. For displacement
experimental designs, the PET detection sensitivity of
a radioligand will be enhanced by rapid dissociation
(k
off
) and tissue clearance (k
2
), thus enhancing the
contrast in the PET signal before and after dopamine
release. For dopamine displacement experiments,
the faster kinetics of fallypride will improve detection
sensitivity, but only in regions where sufficiently high
PET signal is present. A recent study by Narendran
et al (2009) directly compared
11
C-fallypride- and
11
C-
FLB457-measured BP
ND
changes before and after endo-
genous dopamine competition in humans using
an amphetamine challenge. Their results found that
11
C-FLB457 showed 30% to 70% higher sensitivity
(through DBP
ND
) to dopamine competition than
11
C-fallypride in the cortical regions of the brain. In
contrast to the ‘displacement’-type study, this ‘block-
ing’-type study is designed to have the compe-
ting dopamine present prior to the administration of
the radioligand. This difference in study design favors
11
C-FLB457 for the blocking-type experiment, because
this design is less sensitive to the combination of slow
binding dissociation, increased PET signal in both the
specifically bound and ND states, and the higher
proportion in the specifically bound state relative to
the ND state.
Cerebellar Kinetics
The radiotracer kinetics in the cerebellum are of great
importance for PET D
2
/D
3
receptor studies because it
is commonly used as a reference region in BP
ND
deter-
mination. In this study, it was found that significant
specific binding could be measured in the cerebellum
for
11
C-FLB457 but not for
18
F-fallypride.
The presence of specific D
2
/D
3
binding in cerebel-
lum has been frequently reported in the literature for
11
C-FLB457 (Asselin et al, 2007; Delforge et al, 1999;
Olsson et al, 2004), leaving one to conclude that
similar issues with specific cerebellar binding would
be present with
18
F-fallypride. The inability of
fallypride to detect specific cerebellar binding in
this study can be explained, in part, by its lower in
vivo affinity when compared with FLB457. A B3
higher apparent K
D
of fallypride will result in a 3
lower bound-to-ND fraction at equilibrium. However,
the insensitivity to cerebellar binding is also attrib-
uted to several other characteristics of fallypride,
including the reduced radiotracer clearance from the
blood, faster k
2
, and slower rate of receptor–ligand
association (f
ND
k
on
) as compared with FLB457.
To investigate the sensitivity of both tracers to a
small level of cerebellar binding, we compared
simulations of the cerebellum-to-plasma ratios for
both radiotracers assuming a receptor density of
0.1 nmol/L, as shown in Figure 4. In the absence of
specific binding, this ratio should plateau at V
ND
.In
the presence of specific binding, the ratio will plateau
at V
T
(=V
ND
+V
S
) (as simulated in Figure 4). The
distinctive upward bend in the
11
C-FLB457 curve is
due to the fast f
ND
k
on
combined with the rapidly
declining plasma concentration. It can be seen that
this shape in the curve is enhanced by the slower k
2
and k
off
of
11
C-FLB457. Thus, the slow clearance of
11
C-FLB457 from the tissue acts as a leaky integrator of
the radiotracer, enhancing the opportunity for the
ligand to specifically bind to the receptor sites. This
shape is not seen with
18
F-fallypride because of the
combination of slower plasma clearance of the parent
ligand, a slower rate of specific binding, and faster
dissociation from specific binding sites.
When using a reference region method of analysis,
such as the Logan DVR graphical method, specific
cerebellar binding will result in an underestimation
of BP
ND
, with the following relationship between
measured BP
ND
and true BP
ND
being derived:
BPNDmeasured
BPNDtrue
¼1Bmaxref ð1þBPNDtissue Þ
Bmaxtissue ð1þBPNDref Þ
"#
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
1004
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
The effect of specific binding in the cerebellum on
the measurement of BP
ND
is shown in Figure 5,
where the higher degree of underestimation for
FLB457 can be attributed to its higher affinity. Of
primary importance is the non-linearity in bias in the
low-density regions where the receptor density is
only slightly greater than that found in the cerebellar
lobes. Across a cohort of animals with variable
levels of cerebellar binding, this will result in
high variability of BP
ND
estimation in low-receptor-
density regions such as the cortex. We have recently
reported high variability in cortical
18
F-fallypride
binding (in BP
ND
) in a large cohort of rhesus monkeys
(Christian et al, 2009), which may be a direct
consequence of cohort variation in cerebellar D
2
/D
3
receptor concentration. The extent and variability
of D
2
/D
3
density in the cerebellum (lobes) in the
population is not known, and thus it is not possible
to assess the variability in BP
ND
due to cerebellar
specific binding. Without knowledge of specific
cerebellar binding, caution must be used in inter-
preting binding in low-density regions when refer-
ence region methods are employed.
Conclusions
There is a high degree of similarity in the visual
appearance of the PET images of
18
F-fallypride and
11
C-FLB457, with both radiotracers yielding similar
target-to-background ratios throughout the brain
regions, with differing D
2
/D
3
receptor densities.
However, assessing the performance of a radiotracer
simply on the basis of target-to-background ratios can
lead to a misguided understanding of a radiotracer’s
ability to provide an accurate measure of receptor–
ligand binding (Eckelman et al, 2009). Compared
with fallypride, FLB457 shows faster ligand–receptor
binding and slower dissociation from the receptor,
translating into an equilibrium dissociation constant
that is approximately three times lower. Fallypride
clears from the ND space faster and remains in the
plasma longer than FLB457. The higher affinity and
ND component of FLB457 will provide higher tracer
uptake in the cortical and cerebellar regions; how-
ever, this will be offset by lower counting statistics
when using a
11
C radiolabel. These properties suggest
that
18
F-fallypride is better suited for endogenous
displacement-type experiments. For reference region
methods of analysis, both radiotracers will suffer
from potential BP
ND
bias in the low binding regions.
Clom/Plasma
Clom/Plasma
true FLB457
FLB457 kinetics
FLB457 plasma input
fallypride kinetics
fallypride plasma input
FLB547 kinetics
true fallypride
fallypride kinetics
150% kon
100% kon
50% kon
Figure 4 Cerebellum-to-parent plasma ratios assuming a cerebellar D
2
/D
3
receptor concentration of 0.1 nmol/L. (A) Simulated
cerebellum/plasma curves using FLB457 measured plasma input function; (B) curves generated using measured
18
F-fallypride
plasma concentration. Both fallypride-like kinetics and FLB457-like kinetics are represented using each plasma input function.
Variations in k
on
from the tracer’s true k
on
value (±50%) are shown by dotted and dashed lines. The thick solid lines indicate the
kinetics parameters of FLB457, with a FLB457 input function (A) and fallypride kinetic parameters and fallypride input (B).
Figure 5 Theoretical underestimation of BP
ND
using reference
region methods. The series of plots are the theoretical regions of
the cortex (*,B
max
= 0.5 nmol/L), thalamus (J,B
max
= 1.5 n-
mol/L), and SN ( ,B
max
= 3.0 nmol/L). Fallypride curves are
black; FLB457 curves are gray.
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
1005
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
The presence of cerebellar binding may preclude the
use of reference region methods for cortical D
2
/D
3
assay.
Acknowledgements
We acknowledge the help provided by Dr Todd
Barnhart and Jon Engle in radiotracer preparation;
Liz Ahlers in data acquisition; and Leslie Resch,
Allison Theide, and Julie Larson in animal handling.
This work was supported by funding through NCI
T32 CA009206-30, NIH T90 DK070079, NIH/NIBIB
EB006110, and NIAAA RO1 AA12277.
Disclosure/conflict of interest
The authors declare no conflict of interest.
References
Airaksinen AJ, Finnema SJ, Nag S, Mukherjee J, Gulyas B,
Halldin C (2006) A new PET radioligand 11C cyclo-
propyl-FLB 457 for imaging extrastriatal dopamine D2
receptors: evaluation in monkey and comparison to 11C
FLB 457 and 11C fallypride. Neuroimage 31:T15-T
Airaksinen AJ, Nag S, Finnema SJ, Mukherjee J, Chatto-
padhyay S, Gulya
´s B, Farde L, Halldin C (2008)
[11C]Cyclopropyl-FLB 457: a PET radioligand for low
densities of dopamine D2 receptors. Bioorg Med Chem
16:6467–73
Akaike H (1974) A new look at the statistical model identi-
fication. Automatic Control, IEEE Transact Biomed Eng
19:716–23
Asselin MC, Montgomery AJ, Grasby PM, Hume SP (2007)
Quantification of PET studies with the very high-affinity
dopamine D-2/D-3 receptor ligand [C-11]FLB 457:
re-evaluation of the validity of using a cerebellar
reference region. J Cereb Blood Flow Metab 27:378–92
Christian BT, Narayanan T, Shi B, Morris ED, Mantil J,
Mukherjee J (2004) Measuring the in vivo binding
parameters of [18F]-fallypride in monkeys using a PET
multiple-injection protocol. J Cereb Blood Flow Metab
24:309–22
Christian BT, Vandehey NT, Fox AS, Murali D, Oakes TR,
Converse AK, Nickles RJ, Shelton SE, Davidson RJ,
Kalin NH (2009) The distribution of D2/D3 receptor
binding in the adolescent rhesus monkey using small
animal PET imaging. Neuroimage 44:1334–44
Delforge J, Syrota A, Mazoyer BM (1990) Identifiability
analysis and parameter-identification of an in vivo
ligand–receptor model from PET data. IEEE Transact
Biomed Eng 37:653–61
Delforge J, Bottlaender M, Loc’h C, Guenther I, Fuseau C,
Bendriem B, Syrota A, Mazie
`re B (1999) Quantitation of
extrastriatal D2 receptors using a very high-affinity
ligand (FLB 457) and the multi-injection approach.
J Cereb Blood Flow Metab 19:533–46
Delforge J, Bottlaender M, Pappata S, Loc0h C, Syrota A
(2001) Absolute quantification by positron emission
tomography of the endogenous ligand. J Cereb Blood
Flow Metab 21:613–30
Eckelman WC, Kilbourn MR, Mathis CA (2009) Specific to
nonspecific binding in radiopharmaceutical studies: it’s
not so simple as it seems!. Nucl Med Biol 36:235–7
Halldin C, Farde L, Ho
¨gberg T, Mohell N, Hall H, Suhara T,
Karlsson P, Nakashima Y, Swahn CG (1995) Carbon-11-
FLB 457: a radioligand for extrastriatal D2 dopamine
receptors. J Nucl Med 36:1275–81
Holden JE, Jivan S, Ruth TJ, Doudet DJ (2002) In vivo
receptor assay with multiple ligand concentrations: an
equilibrium approach. J Cereb Blood Flow Metab 22:
1132–41
Innis RB, Cunningham VJ, Delforge J, Fujita M, Gjedde A,
Gunn RN, Holden J, Houle S, Huang S-C, Ichise M, Iida
H, Ito H, Kimura Y, Koeppe RA, Knudsen GM, Knuuti J,
Lammertsma AA, Laruelle M, Logan J, Maguire RP,
Mintun MA, Morris ED, Parsey R, Price JC, Slifstein M,
Sossi V, Suhara T, Votaw JR, Wong DF, Carson RE (2007)
Consensus nomenclature for in vivo imaging of rever-
sibly binding radioligands. J Cereb Blood Flow Metab
27:1533–9
Kahn ZU, Mrzljak L, Gutierrez A, de la Calle A, Goldman-
Rakic PS (1998) Prominence of the dopamine D2 short
isoform in dopaminergic pathways. Proc Natl Acad Sci
USA 95:7731–6
Larsen P, Ulin J, Dahlstrom K, Jensen M (1997) Synthesis of
[11C]iodomethane by iodination of [11C]methane. Appl
Radiat Isot 48:153–7
Logan J, Fowler JS, Volkow ND, Ding YS, Wang G-J, Alexoff
DL (2001) A strategy for removing the bias in the
graphical analysis method. J Cereb Blood Flow Metab
21:307–20
Lundkvist C, Sandell J, Nagren K, Pike VW, Halldin C
(1998) Improved syntheses of the PET radioligands,
[C-11]FLB457, [C-11]MDL100907 and [C-11]beta-CIT-
FE, by the use of [C-11]methyl triflate. J Labelled
Compounds Radiopharm 41:545–56
Morris E, Yoder K (2006) Positron emission tomography
displacement sensitivity: predicting binding potential
change for positron emission tomography tracers based
on their kinetic characteristics. J Cereb Blood Flow
Metab 27:606–17
Morris ED, Christian BT, Yoder KK, Muzic RF (2004)
Estimation of local receptor density, B’max, and other
parameters via multiple-injection positron emission
tomography experiments. Methods Enzymol 385:
184–213
Mukherjee J, Yang Z-Y, Das M, Brown T (1995) Fluorinated
benzamide neuroleptics—III Development of (S)-N-
[(1-allyl-2-pyrrolidinyl)methyl]-5 (3-[18F]fluoropropyl)
2,3-dimethoxybenzamide as an improved dopamine
D-2 receptor tracer. Nucl Med Biol 22:283–96
Mukherjee J, Yang Z-Y, Brown T, Lew R, Wernick M,
Ouyang X, Yasillo N, Chen C-T, Mintzer R, Cooper M
(1999) Preliminary assessment of extrastriatal dopamine
d-2 receptor binding in the rodent and nonhuman
primate brains using the high affinity radioligand,
18F-fallypride. Nucl Med Biol 26:519–27
Mukherjee J, Christian BT, Narayanan TK, Shi B, Collins D
(2005) Measurement of d-amphetamine-induced effects
on the binding of dopamine D-2/D-3 receptor radio-
ligand, 18F-fallypride in extrastriatal brain regions in
non-human primates using PET. Brain Res 1032:77–84
Muzic RF, Cornelius S (2001) COMKAT: compartment
model kinetic analysis tool. J Nucl Med 42:636–45
Muzic RF, Christian BC (2006) Evaluation of objective
functions for estimation of kinetic parameters. Med Phys
33:342–53
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
1006
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
Narendran R, Frankle WG, Mason NS, Rabiner EA,
Gunn RN, Searle GE, Vora S, Litschge M, Kendro S,
Cooper TB, Mathis CA, Laruelle M (2009) Positron
emission tomography imaging of amphetamine-induced
dopamine release in the human cortex: a comparative
evaluation of the high affinity dopamine D2/3 radio-
tracers [11C]FLB 457 and [11C]fallypride. Synapse
63:447–61
Okauchi T, Suhara T, Maeda J, Kawabe K, Obayashi S,
Suzuki K (2001) Effect of endogenous dopamine on
endogenous dopamine on extrastriated [(11)C]FLB 457
binding measured by PET. Synapse 41:87–95
Olsson H, Halldin C, Swahn CG, Farde L (1999) Quanti-
fication of [11C]FLB 457 binding to extrastriatal dopa-
mine receptors in the human brain. J Cereb Blood Flow
Metab 19:1164–73
Olsson H, Farde L (2001) Potentials and pitfalls using high
affinity radioligands in PET and SPET determinations
on regional drug induced D2 receptor occupancy—
a simulation study based on experimental data. Neuro-
image 14:936–45
Olsson H, Halldin C, Farde L (2004) Differentiation of
extrastriatal dopamine D2 receptor density and affinity
in the human brain using PET. Neuroimage 22:794–803
Riccardi P, Baldwin R, Salomon R, Anderson S, Ansari MS,
Li R, Dawant B, Bauernfeind A, Schmidt D, Kessler R
(2008) Estimation of baseline dopamine D2 receptor
occupancy in striatum and extrastriatal regions in
humans with positron emission tomography with
[18F] fallypride. Biol Psychiatry 63:241–4
Salinas C, Muzic JRF, Ernsberger P, Saidel GM (2007)
Robust experiment design for estimating myocardial
beta adrenergic receptor concentration using PET. Med
Phys 34:151–65
Schmidt DE, Votaw JR, Kessler RM, Depaulis T (1994)
Aromatic and amine substituent effects on the apparent
lipophilicities of N-(2-pyrrolidinyl)methyl-substituted
benzamides. J Pharm Sci 83:305–15
Slifstein M, Hwang D-R, Huang Y, Guo N, Sudo Y,
Narendran R, Talbot P, Laruelle M (2004a) In vivo
affinity of [18F]fallypride for striatal and extrastriatal
dopamine D 2 receptors in nonhuman primates.
Psychopharmacology 175:274–86
Slifstein M, Narendran R, Hwang DR, Sudo Y, Talbot PS,
Huang Y, Laruelle M (2004b) Effect of amphetamine on
[(18)F]fallypride in vivo binding to D(2) receptors in
striatal and extrastriatal regions of the primate brain:
single bolus and bolus plus constant infusion studies.
Synapse 54:46–63
Smith SM, Jenkinson M, Woolrich MW, Beckmann CF,
Behrens TE, Johansen-Berg H, Bannister PR, De Luca M,
Drobnjak I, Flitney DE, Niazy RK, Saunders J, Vickers J,
Zhang Y, De Stefano N, Brady JM, Matthews PM (2004)
Advances in functional and structural MR image
analysis and implementation as FSL. Neuroimage
23(Suppl 1):S208–19
Suhara T, Sudo Y, Okauchi T, Maeda J, Kawabe K, Suzuki
K, Okubo Y, Nakashima Y, Ito H, Tanada S, Halldin C,
Farde L (1999) Extrastriatal dopamine D2 receptor
density and affinity in the human brain measured by
3D PET. Int J Neuropsychopharmacol 2:73–82
Stark D, Piel M, Hubner H, Gmeiner P, Grunder G, Rosch F
(2007) In vitro affinities of various halogenated benza-
mide derivatives as potential radioligands for non-
invasive quantification of D-2-like dopamine receptors.
Bioorg Med Chem 15:6819–29
Tai YC, Chatziioannou A, Siegel S, Young J, Newport D,
Goble RN, Nutt RE, Cherry SR (2001) Performance
evaluation of the microPET P4: a PET system dedicated
to animal imaging. Phys Med Biol 46:1845–62
Tetko IV, Tanchuk VY (2005) ALOGPS (http://www.
vcclab.org) is a free online program to predict lipophi-
licity and aqueous solubility of chemical compounds.
Proceedings of the 229th National Meeting of
the American Chemical Society, San Diego, CA, 13–17
March 2005
Comparison of
11
C-FLB457 and
18
F-fallypride
NT Vandehey et al
1007
Journal of Cerebral Blood Flow & Metabolism (2010) 30, 994 –1007
