Contrast Peter Reimer, Jack Wittenberg,
MD
Receptor of Liver
#{149} Ralph
MD
Weissleder,
J.
#{149} Thomas
Brady,
Imaging: Cancer’
A new
contrast agent for magnetic (MR) imaging, directed to asialoglycoprotein (ASG) receptors on hepatocytes, was used for detection of liver cancer in rats. Ultrasmall superparamagnetic (mean size, i2 nm) particles of iron oxide (USPIOs) were targeted to ASG receptors by coating particles with arabinogalactan (AG). Liver T2 relaxation times decreased more effectively after a single intravenous administration of AG-USPIO than after an equal dose of a conventional superparamagnetic liver MR contrast agent (AMI-25; mean size, 72 nm). Receptor affinity studies demonstrated that receptor-mediated attachment and subsequent cellular endocytosis do not occur in primary malignant (hepatocellular carcinoma) or metastatic (adenocarcinoma) tumors, because the surface ASG receptors are lost during malignant dedifferentiation. In vitro relaxation and in vivo MR imaging cxperiments of liver tumors show that targeting USPIO to hepatocytes rather than to the mononuclear phagocytic system allows a considerable dose reduction, increases tumom-liver contrast, and potentially allows distinction of ASG-positive (benign hepatocellular) and ASGnegative (malignant hepatocellular) tumors. resonance
Index terms: Contrast media, experimental studies #{149} Iron #{149} Liver neoplasms, diagnosis, 761.321, 761.332 #{149} Liver neoplasms, MR studies, 761.1214 #{149} Magnetic resonance (MR), contrast enhancement Radiology
1
From
Harvard
1990;
the
177:729-734
Department
Medical
School,
MD, PhD MD
of Radiology, 13th
sion requested June 26; revision tional Institutes of Health grant P.R. supported by the Deutsche R.W. C RSNA, 1990
St.
Bldg
MGH-NMR 149,
#{149} Albert
Application
C
S. Lee,
BA
.
to MR Imaging
tumors. We hypothesized that AGUSPIO would accumulate in normal hepatocytes but not in tumor cells, thus improving tumor-liver contrast. To prove this hypothesis, dose response studies, magnetic resonance (MR) imaging experiments with primary and secondary liver tumors, and cell-membrane receptor studies were performed. These results were compared with results obtained with a conventional superparamagnetic iron oxide (AMI-25; Advanced Magnetics, Cambridge, Mass), which has previously been used in clinical trials as a contrast agent for MR imaging of the liver.
superparamagnetic iron oxides with albumin (1), starch (2), and dextran (3,4) coating are rapidly phagocytosed by the mononuclear phagocytic system of liver and spleen. We have recently described a novel, hepatocyte-directed, ultrasmall supemparamagnetic iron oxide (AG-USPIO) preparation as a liver contrast agent (5). Iron oxide particles were targeted to asialoglycoprotein (ASG) receptors by coating particles with arabinogalactan (AG), a galactose-containing polysaccharide (6). ASG receptors have a high affinity for the terminal galactose groups (7), a concept previously exploited in the design of both diagnostic (8-9) and therapeutic drugs (10). The strategy of targeting superparamagnetic iron oxide to hepatocytes rather than to Kupffer cells allows a more homogenous tissue distribution, which potentially results in a greater tissue relaxivity. Previous biodistribution studies demonstrated selective accumulation of AC-USPIO in liver but not in othem organs of the mononuclear phagocytic system such as the spleen. Findings at electron microscopy confirmed preferential distribution of iron oxide to hepatocytes and showed selective binding of ACUSPIO to hepatocyte cell surface membranes. After internalization from surface receptors, AC-USPIO accumulates within intracellular hepatocyte lysosomes, where particles are degraded in a time-dependent fashion (5). Our study was designed to evaluate the potential of ASG receptom-directed AC-USPIO in detecting liver ONVENTIONAL
Center,
Charlestown,
received July 18; accepted no. CA 48279, PPG, and Forschungsgemeinschaft
Media
Massachusetts General Hospital and MA 02129. Received May 7, 1990; mcviJuly 23. Supported in part by the Naby the MGH-NMR development fund; (Re 758/1-1). Address reprint requests
MATERIALS Iron
AND
Oxide
METHODS
Preparations
Hepatocyte-directed
AC-USPIO
(Ad-
vanced Magnetics, Cambridge, Mass) was obtained by coating USPIO with AG (11), AG is known for its high specificity for the
asialoglycoprotein
tocytes.
receptor
on hepa-
The preparations-as
magnetophysical,
biologic,
cologic
properties-
USPIO
have
well and
and AGpreviously
of USPIO
been
as the
pharma-
described
(5,12). of 12.2 [SD]), sec, sec,
Briefly, AG-USPIO has a mean size nm ± 6.5 (standard deviation an Ri relaxivity of 23.3 mmol/L’an R2 relaxivity of 48.9 mmol/L’. and a blood half-life of 8 minutes T, 37#{176}C) (5). AG-USPIO was used at
(0.47 a concentration The
of 17 mg/mL.
efficacy
of AG-USPIO
in detection
of hepatic tumors was compared with that of a conventional, dextran-stabilized iron oxide (AMI-25) that has a mean particle size of 72 nm. Pharmacologic characteristics
have was mL.
to
(13) and
been used
in vivo
described
use (14) of this
previously.
at a concentration
agent
AMI-25
of 11.2 mg/
Abbreviations: AG arabinogalactan, ASF asialofetuin, ASG asialoglycoprotein, C/N contrast-to-noise ratio, CPMG Carr-PurcellMeiboorn-Gill, GRE gradient echo, SD standard deviation, SE spin echo, USPIO ultrasmall superparamagnetic iron oxide.
729
Animals A total relaxation
of 95 rats were used time measurements
MR imaging (n
(n
=
20),
4). Relaxation
or
time
Tumor
for either
(n
71),
receptor
Liver
studies
measurements
were used for dose response studies in male, retired breeder Sprague-Dawley rats (n = 41) (Charles River Breeding Laboratories, Wilmington, Mass, or Harlan Sprague Dawley, Indianapolis) and determination of relaxation times in tumor models in Fischer rats (n 15) and ACIAXC rats (n 15) (Harlan Sprague DawIcy). Different strains of rats were used because the models of liver metastasis (R3230AC tumor cells [Biomeasure, Hopkinton, Mass] in Fischer rats) and primary hepatocellular
ACI-AXC
carcinoma
rats) have
(H-4-II-E
been
well
Prf
Liver or Tumor
‘,
removed
from
was
rats, minced
performed
by
enzymatic
and
desegregation
in
Cells
+
cell fragments
Sonification,
in
Hanks buffered salt solution (Sigma Chemicals, St Louis), sieved through a 0.3 wire mesh, and resuspended in 20 mL of Hanks buffered salt solution. This solution contained tumor fragments and single cells. A total of 0.1-0.2 mL of this solution was then injected subcutaneously into host rats. In addition, intrahepatic inoculation
Mechanical
character-
ized. The properties of the H-4-II-E cell line, a poorly differentiated hepatoma (15-17), and the R3230AC cell line have been described in detail elsewhere (18). Tumor cell lines were propagated as follows: One-cubic-centimeter tumors
were
fragment
direct
centrifugation,
purification
Cell membranes
injec-
tion of the suspension into the liver with 25-gauge hypodermic needles (19). The rats were lightly anesthetized with ether. They were used for in vitro and in vivo experiments 7-10 days after tumor cell inoculation.
Receptor These
Affinity experiments
determine
whether
ceptors
would
bind
Experiments were tumor
performed
to
cell
re-
AG-USPIO
surface specifical-
ly or nonspecifically. Figure 1 summarizes the experimental procedures in a flow chart. Previous studies had shown that hepatic uptake of AG-USPIO is mcdiated specifically by ASG receptors on normal hepatocytes and can be blocked competitively with D(+)-galactose and asialoglycoproteins such as ASF (10,20-25). These initial observations were confirmed by results of a dose-dependent blockade
experiment
and
control
experi-
ments in which a “control blocker” (D(+)glucose and albumin) was used. Hepatocyte, H-4-II-E tumor cell, and R3230 tumor cell membrane suspensions were prepared with a modified technique that has been previously described (26,27). Normal livers from rats were perfused in vivo for 5 minutes by means of buffered salt solution (Hanks) with 0.05 mol/L of (N-[2-hydroxyethyl]piperazineN’-[2-ethanosulfonic acid]) (HEPES) and 0.5 mmol/L of ethylene glycol-bis($aminoethyl ether) N,N,N’,N’-tetraacetic acid (Sigma Chemicals, St Louis). After this initial perfusion to remove erythrocytes, the livers were perfused with colla-
730
Radiology
#{149}
AG-USPIO
usplo
I
AG-USPIO
I
I Incubation,
I
Relaxation
USPIO
time
I
washing
measurement
Figure 1. The flow chart schematically summarizes the steps for the preparation of hepatocyte and tumor cell membranes. Membrane suspensions were incubated with AG-USPIO or Usplo in the presence or absence of blockers (D[+]-galactose [D(+)-gal] and asialofetuin [ASFJ) of the ASG receptor system. After repeated washing cycles, relaxation times of membrane suspensions were determined.
genase (100 U/mL) in Williams E medium (Sigma Chemicals) buffered with 0.05 mol/L of HEPES buffer. Livers and tumoms were then washed in complete medium and sieved through a 36-mesh stainless steel mesh. The resulting tissue fragments and cells were separately incubated in 40 mL of collagenase (at a concentration of 0.3 mg/mL) in complete medium before sonication was performed (10 W for three 20-second cycles on ice [Sonifier 450; Branson Ultrasonics, Dan-
bury, tant
Conn]). Subsequently, was washed twice by
tion and minutes.
centrifuged
Samples suspensions minutes tions
of the
block brane,
at 1,500 resulting
of D(+)-galactose
and complete
existing ASG or a “control
for
10
membrane
(0.1
ASF
supernaof dilu-
rpm
were then incubated at 37#{176}Cwith different
ma Chemicals) Chemicals)
the means
for 30 concentra-
mg/mL;
(2 mg/mL;
medium
receptors blocker”
(control)
Sig-
Sigma to
on the mem(D[+]-glu-
December
1990
Pr#{149}-Incubatlon
Post-Incubation Blockads
Post-IncubaUon Control
Hepatoma (H-uI-4-E)
Hepatocyte (SD)
Adenocarcinoma (R3230)
U
so
. 0 E
60
C,’
!
40
I-
! I
20
q .
“ !
q q
E
C 0
.$
#{149}
0
C
E 0
.
C.)
U
C.) D
-
D
b. 2. (a) T2 relaxation times (CPMG sequence) of concentrated hepatocyte cell membrane solutions (MS) at 37#{176}C and 0.47 T. AG-USPIO incubated with hepatocyte membrane receptor solution has a significantly lower T2 (third column) than solutions incubated after blockade of the receptor with D(+)-galactose (Gal) and ASF (second column). Competitive blocking with different concentrations demonstrates the dose-dependent relationship. Incubation with “control” blockers (D[+]-glucose [Gic]) and albumin [Mb]) does not affect the relaxation times of AG-USPIO (first and third columns). (b) T2 relaxation times (CPMG sequence) of aqueous hepatocyte or tumor cell membrane solutions at 37#{176}C and 0.47 T. AG-USPIO incubated with hepatocyte membrane receptor solution has a significantly lower T2 than solutions incubated after blockade of the receptor with D(+)-galactose and ASF (first column). No significant T2 changes are seen in the H-H-4-E tumor cell line (second column) after incubation with AG-USPIO. T2 of R3230AC (mammary adenocarcinoma) cell membranes are similar to those of H-I!4-E cells (second column). Note that there is no difference in T2 if tumor cell membrane solutions are incubated with or without blocker of the ASG receptor. Vertical bars in a and b represent the mean ± SD. Figure
cose and/or albumin, 0.1 mg/mL). Dccreasing concentrations of blockers (from 1 mg/mL to 0.0001 mg/mL of D[+J-galac-
weight.
tose and ASF) were
with
used
for the competi-
tive blocking experiment. After this incubation, either USPIO or AG-USPIO was added to the test tubes (10 Mmol iron per tube). After 30 minutes of incubation, samples were washed with buffer and ultracentrifuged twice to remove excessive unbound AG-USPIO and/or USPIO. Relaxation
were
times
then
of membrane
measured
solutions
as described
below.
tumor-bearing
Relaxation
time rats were
administration
of AG-USPIO
a concentration
mol/kg. guination. diately,
ments hour
Time
after
trometer (PC-20 Conn) operating
times were meawith an MR spec-
Minispec; IBM, Danbury, at 0.47 T at 37#{176}C. Prior to
each measurement the spectrometer was tuned and calibrated. Ti was measured from eight data points generated by an inversion-recovery pulse sequence. T2 was measured from 10 data points by means of a Carr-Purcell-Meiboom-Gill
(CPMG) pulse sents
pulse
sequence
with
an inter-
time of 1 msec. Each value the mean of three separate
repremeasure-
ments. In Vitro
Relaxation
MR relaxation
times
Studies of liver
and tumor
tissue were measured to determine the magnetic effects of AG-USPIO and AMI25 in these tissues. Liver relaxation times in normal Sprague-Dawley rats were measured
as part
of the
dose
response
study by use of either AG-USPIO 25 with a concentration of 0, 2.5, or 20.0 mol of iron per kilogram
Volume
177
Number
#{149}
3
or AM!5.0, 10.0, of body
dow”
5 or 10
venous
MR
were
hematoxylin-eosin
Imaging was performed
with
superconducting
with
a 0.6-
a section
Muimages
thickness
of
4 mm and a 128 X 256 matrix. A head coil with a field of view of 16 cm was used;
stain
study
or absence
dimensions were 0.6 X 1.2 X 4 mm. Spin-echo (SE) images were obtained with 250/20, 500/30, 1,500/40, and 1,500/ 80 (repetition time [TRJ msec/echo time [TEl
msec)
pulse
sequences;
averaged and two.
the
number
was, respectively, Gradient-echo
six, (GRE)
images were obtained with a 120/18 sequence, a flip angle of 60#{176}, and eight signals averaged. Homogeneous regions of interest (more than 100 pixels) within individual livers and tumors were used to calculate tumor-liver contrast-to-noise tios (C/Ns) (mean signal intensity mor minus mean signal intensity divided
by
the
noise). All in vivo
were
standard
MR imaging
performed
raof tuof liver
deviation
of
of hepatic
the pres-
tumors.
Analysis
Differences C/N among
in relaxation different study
in rats anesthetized injection
(35 mg/kg). All and 60 minutes half-lives) after
(more than three the administration
with
of pentobarbital
imaged
before blood of ci-
times and groups were
evaluated statistically with the nonparametric Wilcoxon signed rank test (28). RESULTS Receptor
Affinity
Studies
Membrane receptor studies performed to study whether cell membranes or hepatocyte bind
AG-USPIO
by
a receptor
were tumor would system
(Fig i). The T2 relaxation time of medium containing hepatocyte ASG membrane receptors was significantly (P < .05) shorter (13.4 msec ± 1.7) after incubation with AC-USPIO than before incubation (i22.7 msec ± i5.7)
experiments
intraperitoneal
rats were
intra-
correlative
for
to determine
Statistical
pixel
of signals four, two,
after
magnet
GE Medical Systems, Transverse multisection
obtained
is 1-6 hours
administration of either agent. After the MR imaging experiment, the liver of each animal was removed, fixed in buffered 10% formaldehyde for 24 hours, and individually processed with
ence
MR imaging
or AMI-25. Previous studthat the optimum “win-
for imaging
histologic
waukee). Ti and T2 relaxation sured in milliseconds
ther AG-USPIO ies have shown
or AMI-25
of either
death.
(Technicare,
Measurements
in after
All rats were killed with exsanOrgans were removed immeand relaxation time measurewere performed at 37#{176}C within 1
T (25.i-MHz)
Relaxation
measurements performed
(Fig
2b).
This
is best
explained
by the retention of AC-USPIO on membrane receptors despite repeated washing cycles. When ASC membrane receptors on hepatocytes were blocked with i mg/mL of D(+)-galactose and ASF prior to incubation Radiology
731
#{149}
Table 1 Relaxation AMI-25
Times
of Different
Tissues
before
and after
Intravenous After
Before Administration (n3)
Tissue
Liver
Doses
of Different
Administration, 5ornol
of AG-USPIO
After
and
Administration,
10imol
AG-USPIO (n3)
AMI-25 (n3)
AG-USPIO (n3)
AMI-25 (n=3)
(CDF)
Ti T2
307 39.7
Liver
±27.0 ± 1.2
252 23.4
± ±
4.9 0.6*
298 33.2
±13.2 ± 0.9
0.7
250±7.4 30.1 ±
230 ± 18.6 ±
6.1* 0.6*
266 28.2
±33.4 ± 0.9
223±4.9* 19.4 ±
1.7*
234±5.9 25.6 ±
(AC!)
Ti T2
263±7.1 41.6 ±
Tumor
241±3.9* 26.1 ±
1.0
1.9
0.8
(R3230AC)
Ti T2 Tumor Ti
777 75.1
±41.3 ± 8.7
803 75.7
±49.5 ± 8.1
796 75.4
±43.2 ± 5.8
777 76.2
±17.2 ± 6.0
782 77.6
± 17.4 ± 6.7
655 63.0
±i3.0 ± 1.4
657 ±li.2 64.2 ± 3.7
657 63.1
±12.9 ± 7.5
657 65.6
±14.1 ± 4.4
651 63.6
± 16.5 ± 3.9
(H-4-!!-E)
T2 Note-Relaxation .
Administration
are expressed
times
lower
(P
C/N
of Different
Significantly
Table 2 Tumor-Liver
in milliseconds (mean and SD). the relaxation time obtained after administration
< .05) than
Doses,
Contrast
Agents,
and Pulse After
of AMI-25.
Sequences
Administration, 5Mmol
After
Administration, 10imol
Before
Administration PulseSequences SE 250/20
120/18
7.9 19.2 35.0 39.7
18.6 ±
3.3
Note.-Six signals were averaged in the SE 250/20 sequence, sequences. Data are expressed as mean plus or minus SD. * Significantly (P < .05)lower than AMI-25. t Significantly (P < .005) lower than AMI-25.
AG-USPIO,
nificantly i22.7 msec
± ± ± ±
AG-USPIO (n5)
(n5)
1.2 4.6 5.Ot 2.5*
2.3 8.7 16.8 20.1
± ± ± ±
5.3 1.7 4.3 2.2
75.3 ± 4.2*
40.4
± 4.8
T2 values
were
higher (69.9 msec ± 15.7, P < .05).
sig-
± 3.4 vs This ef-
four
USPIO
signals
because
in the
SE 500/30
of incomplete
sequence,
incubation D(+)-glucose in control
Dose USPIO function
response.-The effect of AGand AMI-25 on T2 of liver as a of administered dose is
shown
in Figure
medium and solutions
was
(Fig 2b). When albumin were used no receptor
seen
(Fig
of primary tumor cell suspensions
to those
of hepatocyte
(Fig
The
2b).
membranes
overall
of the
centration.
T2 values
membrane
After
of the
differ from in Figure
lower
AG-USPIO, the mary hepatocellular msec ± 9.2) nor
2a). (H-4-II-E) and (R3230AC) were similar
membrane solutions those of the solutions
cause
2a be-
con-
incubation
with
T2 of neither
the
carcinoma
the
pri(479.0
metastatic
adeno-
Relaxation
3. The
of the
those
dicating
732
test
of the contamination
Radiology
#{149}
± 12.8*
58.7
± 5.5
samples
control
in relaxation compared
samples, with
free
inAG-
and
two
signals
SE 1,500/40
and
in the
demon-
1.1 1.4 2.7 1.6
1,500/80
20
of AGof liver
Dosi
Figure
fore and different 41).
for clinical The liver
times
imaging T2 values
(14,29) obtained
(Fig 3). in
USPIO Table crease
and
after
administration
sizes. tumors.times
be-
AG-
or AMI-25 are summarized 1. There is a significant dein liver Ti and T2 relaxation
3.
(tmoI
Flag)
T2 relaxation times of liver beafter intravenous administration of doses of AG-USPIO or AMI-25 (n
after
administration of 5 .&mol or 10 imol (P < .05) of AGUSPIO, and of liver T2 after administration of iO mol of AMI-25 (P < .05). However, the decrease in liver T2 after administration of AG-USPIO is significantly greater than that after administration of AMI-25 at both doses investigated (P < .05). Relaxation times of either primary hepatocellular carcinoma or metastatic adenocarcinoma did not significantly de-
(P < .05)
Spmague-Dawley rats (Fig 3) differ only minimally from those in Fischer or AC! rats (Table 1), probably be-
fore
with
97.0
T2 (from 35.4 msec ± 2.4 to 24.5 msec ± 0.7) similar to the decrease caused by a dose of 20 mol/kg of AMI-25 (from 35.4 msec ± 2.4 to 22.4 msec ± 0.4 msec), the dose originally used
was
decrease
data
strate that a dose of 5 mol/kg USPIO results in a decrease
cause of different sample Relaxation times of liver Liver and tumor relaxation
times
± ± ± ±
Studies
carcinoma (479.0 msec ± 11.5) decreased to the extent that the T2 of the hepatocyte suspension did. Theme
a smaller
9.7 13.1 25.9 26.7
wash-
In Vitro
T2 values secondary membrane
16.8 ± 4.4 25.3 ± 4.4t 50.7 ± 5.i 52.9 ± i.8
ing.
fect was clearly dependent on the total amount of blocking agent in the
blockade
AMI-25 (n=5)
with
60#{176} flip angle
with
AMI-25
(iz=5)
-14.2 ± 10.5 -1.5 ± 1.1 11.7 ± 10.7 15.3 ± 3.2
SE 500/30 SE 1,500/40 SE 1,500/80 GRE
AG-USPIO
(n20)
in
December
1990
Figure imaged
4. (a) Tumor (R3230AC)-liver with SE 250/20 sequence
signals
averaged
quence
with
(top
four
row),
signals
model six
with
SE 500/30
se-
averaged
(middle
row), and SE 1,500/40 sequence with two signals averaged (bottom row). These sequences were performed before administration (first column), after administration of 5 omol (middle column), and after administration of 10 mol (right column) of AG-
USPIO.
: -
‘l;_
Note
the excellent
tumor
(f)-liver
(1)
contrast at low doses (5 tmol) in mixed Tiand T2-weighted pulse sequences (SE 500/ 30 with four signals averaged). (b) Tumorliver model imaged with an identical protocol as in a but after administration of AM!25 instead of AG-USPIO. Note that the liver appears less hypointense for any given
pulse
sequence
than
liver contrast sequence than USPIO.
in a. As a result,
is lower for any after administration
tumor-
given
pulse of AG-
,
:‘
mor-liver
.
‘:
C/N
higher dose preparation.
was
ministration
a.
of 5 mol
USPIO, tumor-liver ally higher than of AMI-25.
4
higher
at the
of either iron Interestingly,
oxide after
Fe/kg
ad-
of AG-
C/N was genertwice the dose
with
.1#{149}
DISCUSSION 4
the feasibilto ASG receptors on hepatocytes (5). Hepatocyte ASG receptors (400,000-500,000 receptors per cell) are responsible for the plasma clearance of desialylated glycoprotein by the liver. The directability of iron oxide was achieved by coupling galactose terminals in the form of AG Previously
ity
A
to
of
we
USPIOs.
AG-USPIO
predominantly cells
of
reported
USPIO
targeting
accumulates
in hepatic
liver
but
not
in
parenchymal other
organs
or
the the
mononuclear phagocytic system at low doses used for MR imaging (5). Our present and previous studies (5) confirm that AG-USPIO is selectively taken up by hepatocytes via the ASG receptor
system.
take
of iron
reduction
the
in of
by
the
AG-USPIO
administration
of either
duction in liver (Fig 4b), resulting
conspicuity MR
Imaging
Experiments
In vivo MR imaging experiments were performed in rats with implanted tumors by use of both SE and GRE sequences (Fig 4). Visual observation demonstrates that, at equal dose, AGUSPIO
Volume
(Fig
177
4a)
provides
#{149} Number
greater 3
re-
signal than AMI-25 in improved tumor on both SE and GRE im-
ages. Quantitative tumor-liver C/N measurements are summarized in Table 2. Maximum tumor-liver contrast was achieved with the GRE pulse Sequence (5 mol/kg AG-USPIO: 75.3 msec ± 4.2; 10 mol/kg AG-USPIO: 97.0 msec ± 12.8). As expected, tu-
USPIO AMI-25
USPIO
oxide
T2 caused
in
larger
3) despite
(Fig
The
increased
is thought
even
spatial
at
the
tocytes
the are
more
because targeted.
cells so
that
mass) (