Monitoring Early Response of Experimental Brain Tumors to ... · extracellular volume, and membrane...
Transcript of Monitoring Early Response of Experimental Brain Tumors to ... · extracellular volume, and membrane...
Vol. 3, 1457-1466, September 1997 Clinical Cancer Research 1457
Advances in Brief
Monitoring Early Response of Experimental Brain Tumors to
Therapy Using Diffusion Magnetic Resonance Imaging1
Thomas L. Chenevert,2 Paul E. McKeever, and
Brian D. Ross
Departments of Radiology fT. L. C., B. D. R.] and Pathology[P. E. M.], University of Michigan, Ann Arbor, Michigan 48109
Abstract
Quantitative magnetic resonance imaging was per-
formed to evaluate water diffusion and relaxation times, Ti
and T2, as potential therapeutic response indicators forbrain tumors using the intracranial 9L brain tumor model.
Measurements were localized to a column that intersected
tumor and contralateral brain and were repeated at 2-day
intervals before and following a single injection of i,3-bis(2-
cffloroethyl)-i-thtrosourea (i3.3 mg/kg). Tumor growth was
measured using T2-weighted magnetic resonance imaging to
determine the volumetric tumor doubling time (Td) before
(Td = 64 ± i3h,mean ± SD,n = 16)andafter(Td = 75 ±
9 h, n = 4) treatment during exponential regrowth. Appar-ent diffusion coefficient of untreated tumors was independ-
ent of tumor volume or growth time, whereas relaxation
times increased during early tumor growth. Diffusion dis-played the strongest treatment effect and increased beforetumor regression by 55% 6-8 days following treatment.
Changes in relaxation times were also significant with in-creases of 16% for Ti and 27% for T2. Diffusion and
relaxation times returned to pretreatment levels by i2 daysafter treatment. Histological examination supports the
model that the observed increase in diffusion reflects an
increase of extracellular space following treatment. Further-
more, the subsequent apparent diffusion coefficient decrease
is a result of viable tumor cells that repopulate this space at
a rate dependent on the surviving tumor cell fraction andrecurrent tumor doubling time. Serial tumor volume meas-
urements allowed determination of log cell kill of i.O ± 0.3(n = 4). These results suggest that diffusion measurements
are sensitive to therapy-induced changes in cellular struc-ture and may provide an early noninvasive indicator oftreatment efficacy.
Received 3/3/97; revised 6/27/97; accepted 7/14/97.
The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely toindicate this fact.
I This research was supported in part by Grant BE-149 from the Amer-ican Cancer Society and NIH Grant R29 CA59009.
2 To whom requests for reprints should be addressed, at Department ofRadiology-MRI, University of Michigan Hospitals, 1500 East Medical
Center Drive, Ann Arbor, MI 48109-0030. Phone: (313) 936-8866; Fax:(313) 764-2412; E-mail: [email protected].
Introduction
The management of high-grade malignant tumors of thecentral nervous system is problematic, and despite the use of
multimodality therapy, malignant gliomas remain uniformly
fatal (1). Furthermore, quantitation of the response of brain
tumors to therapy is more difficult than in systemic cancers
because an improvement in patient function is multifactorial,
and changes in neurological deficits may be unrelated to
changes in tumor volume (2-5). MRI3 and X-ray CT scans ofmalignant human brain tumors do not readily allow quantitation
of the actual tumor volume. Administration of a contrast agentallows estimates of tumor areas from the largest cross-sectional
area of contrast enhancement indicating a compromised blood-brain barrier. The blood-brain barrier, however, may be altered
by chemotherapeutic agents (5) or corticosteroids; thus, meas-urements of tumor dimensions by the boundaries of contrast
enhancement is an indirect estimate of tumor size. Moreover,
not all tumors exhibit contrast enhancement (6).
Because of the difficulty in obtaining accurate measure-ments of intracranial tumor volumes, brain tumor therapy trials
usually report median survival time and median time to progres-sion as a quantitative measure of response (2-5). Improved
methods that would allow for earlier and accurate quantitationof the therapeutic response in individual patients are still
needed. Because diffusion MRI is sensitive to tissue structure at
the cellular level, we believe that this technique has the potentialto detect important quantitative information about the tumorcellular changes that would occur following successful thera-
peutic intervention.Cellularity and the integrity of cellular membranes that
impede water translational mobility can affect the diffusion of
water within tumor tissue. For example, water diffusion meas-
urements have been shown to be sensitive to tissue cellular size,
extracellular volume, and membrane permeability (7-9). In ad-
dition, it has been shown that structural anisotropy is manifest as
diffusion anisotropy (10-12). Diffusion MRI applications in-dude imaging of cerebral stroke (13-16) and as a probe in thestudy of central nervous system tumors in humans (17-19).
Diffusion studies on human (17-20) and animal (21-23) tumorshave demonstrated strong differences between solid tumors rel-
ative to high diffusion within necrosis and cysts. As both solid
tumor and peritumor edema can have a continuum of diffusion
values that are elevated relative to normal brain, the diffusion
coefficient alone does not readily distinguish tumor from edem-
atous tissue (19, 23). However, structurally anisotropic tissues,
such as white matter, can exhibit persistent diffusion anisotropy,
3 The abbreviations used are: MRI, magnetic resonance imaging; MR.
magnetic resonance; CT, computed tomography; ADC, apparent diffu-sion coefficient; NMR, nuclear magnetic resonance; BCNU, 1,3-bis(2-
chloroethyl)-l-nitrosourea; ROl, region(s) of interest; IR, inversionrecovery.
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
despite a diffusion coefficient elevated by edema. This could every 2 days. A rigid bite-bar was used to hold the anesthetized
potentially provide contrast between edema and an isotropic rat within a 35-mm diameter transmit-receive slotted-resonator
solid tumor ( 18-2 1 ). radiofrequency coil (28). Gradient-recalled-echo MRI inter-
The consistent observation of high diffusion in necrotic leaved in orthogonal planes was used to position the animal
tissue relative to solid tumor provides a rationale for the use of rapidly and reproducibly. Quantitative in vivo measurements
diffusion to monitor cellular changes following anticancer ther- included tumor volume, localized water diffusion, and localized
apies. Preliminary studies in gliomas revealed that the evolution Tl and T2 relaxation times. The 9L glioma appears hyperintense
toward necrosis that occurs following successful therapy could on T2-weighted MRI and has relatively distinct tumor margins
be detected as an increase in ADC relative to pretreatment levels with only moderate pentumoral edema. Therefore, tumor vol-
in the intracranial 9L model after administration of a chemo- ume was assessed using multislice T2-weighted MRI acquired
therapeutic agent (24, 25). The observed increase in water with repetition time TR = 2850 ms, echo time TE = 80 ms, 32
mobility was anticipated to be the result of increased interstitial coronal 0.8-mm slices, 128 X 128 matrix, two-signal average,
water volume and cellular permeability as the tumor cells ne- and a 30-mm field of view. These images were acquired as two
crosed, thus indicating the potential of diffusion MRI for pre- interleaved 16-slice sets collected near the beginning and at the
dicting therapeutic efficacy. More recently, others have made end of the NMR session. This indicated whether the animal had
similar observations in a s.c. tumor model using diffusion- moved due to failed sedation, in which case only the initial
weighted spectroscopy (26). 16-slice set was used for tumor volume estimation. Tumor
In this study, we sought to determine if quantitative MR volume was quantified as the product of slice-to-slice separation
diffusion measurements could be used as a predictor of treat- and the sum of areas from manually drawn tumor ROI on
ment outcome through early detection of cellular changes before images.
tumor regression in an intracerebral tumor model. In addition, Volumetric tumor doubling time, Td, and constant, K, re-
NMR relaxation times, Tl and T2, have been shown to be lated to initial viable tumor volume, were calculated using a
sensitive to extracellular volume (23, 27) and thus may also two-parameter, least-squares-fit of measured tumor volume ver-
demonstrate changes in response to treatment. In this regard, sus time by:
diffusion and Tl and T2 relaxation times were obtained from rat
intracerebral 9L gliosarcomas before and following treatmentTumor volume (t) = K � 2:/Td
with BCNU. We found that of the MR parameters evaluated,
diffusion MRI was the most sensitive to changes in membrane
integrity affecting intra- and extracellular water volumes fol-
lowing treatment. Furthermore, MRI diffusion values varied
among x, y, and z directions, indicating the presence of diffusion
Data during exponential pretreatment and tumor regrowth were
fit independently to yield Tdpre and � respectively. Pre-
treatment tumor volume, Vpre, and the effective volume of tumor
that survives treatment, � were derived by extrapolation of
fitted pretreatment and regrowth curves to the time of treatment.anisotropy near tumor boundaries that correlated with histolog- j� the cellular density of the tumor at time of treatment andically distinct features. These results should help motivate future during exponential regrowth are approximately the same, thenexperiments correlating diffusion changes in human brain tu-
mors with clinical outcome following therapeutic intervention.the following provides an estimate of log cell kill:
Materials and Methods
F � 1Log cell kill = log10l j�- I
L postJ
Glioma Model. Rat 9L tumor cells were grown as mono- T2-weighted images were also used to prescribe subse-
layers in sterile plastic flasks in modified Eagle’s MEM with quent water diffusion and relaxation time measurements. The
10% fetal bovine serum. Cells were cultured in an incubator at diffusion pulse sequence used orthogonal 90#{176}and 1 80#{176}slice-
37#{176}C95% air, 5% CO2 atmosphere until confluent. Cells were selective pulses that defined a right-to-left oriented, 2 X 2-mm
then harvested by trypsinization, counted, and resuspended in column through the most homogeneous tumor region and con-
serum-free medium for intracerebral injection. tralateral brain ( 1 1 , 2 1). Diffusion measurement is susceptible to
Adult male Fischer 344 rats weighing approximately I 50 g errors due to tissue motion. These artifacts were lessened by
were anesthetized via i.p. injection of a ketamine (87 mg/kg using a frequency encode gradient along the column axis for
body weight) and xlazine (13 mg/kg) cocktail. A small incision spatial encoding and magnitude processing of Fourier trans-
over the right hemisphere was made, and a high-speed drill was formed echoes. As a result, phase cancellation effects com-
used to create a I -mm diameter burr hole through the skull. The monly associated with tissue motion in the presence of diffusion
rat head was affixed in a stereotactic holder for inoculation of gradient pulses were greatly reduced (1 1). Spatial resolution
l0� 9L tumor cells in 5 p.! in the right forebrain at a depth of 3 along the one-dimensional image of the column was 0.23 mm.
mm via a 25-�il Hamilton syringe. The surgical region was Potential directionality of water mobility is probed by the con-
irrigated with 70% ethanol, and the burr hole was filled with trolling diffusion gradient direction. To account for potential
bone wax to reduce extracerebral extension of the tumor and diffusion anisotropy, paired diffusion sensitization gradient
loss of cerebrospinal fluid. The skin incision was sutured closed, pulses were trapezoidal, 15 ms each, and applied independently
and the rats were allowed to recover. Tumors were produced in on x, y, and z axes. Diffusion gradient direction and amplitude
20 rats for this experiment. were interleaved during the scan and stepped from 14 to 60
MRI Measurements. In vivo MRI was initiated 10-12 milliTesla/meter for a total of 42 “b-factors” (17) on each axis,
days after tumor inoculation and performed serially on average ranging from 87 to 1 669 s/mm2. ADCs were calculated for all
1458 Therapy Assessment via Quantitative Diffusion MRI
(A)
(B)
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Clinical Cancer Research 1459
Fig. 1 Serial coronal T2-weighted MR images of a ratbrain with time after tumor im-plantation. Images were ac-quired immediately before treat-ment with BCNU (a) and at167 h (b), 319 h (c), and 442 h
(d) following BCNU treatment.The full 12 time point series onthis animal spanned 2 weeks.The column used for obtaininglocalized diffusion, TI and T2data, and ROl of tumor andbrain are overlaid on each im-age.
points along the column and each direction via least-squares-fit
to the model:
S(b1) = S0e��”�; i = x, y, z
Disparate ADC versus gradient direction is an indication of
diffusion anisotropy ( 10, 1 1 ). Although a 3 X 3 tensor is
required to fully characterize diffusion in an anisotropic system
such as tissue (29), the quantity ADCo = [ADCX + ADC� +
ADC�]/3 is useful because it represents a scalar invariant of the
diffusion tensor. That is, ADCo is more closely related to a
tissue-inherent property rather than one dependent on the rela-
tive orientation of gradient axes and the rat brain.
An IR experiment was used for Tl measurements (T1IR).
Orthogonal slice-selective 90#{176}and 180#{176}pulses following non-
selective inversion and variable inversion recovery delay were
used for rapid T1IR data collection along the same frequency
encoded column probed for diffusion. Data at 16 inversion times
between 10 and 5000 ms were acquired within 4 mm. Tl
relaxation times were estimated from a three-parameter T1IR
model fit for each 2 X 2 X 0.23-mm volume element along the
column (30).
Unlike diffusion and Ti measurements, additional echoes
can be collected without a time penalty; therefore, localized T2
measurements were derived from full two-dimensional multie-
cho images of a 2-mm coronal slice acquired with TR = 2000
ms, TE = n X 20 ms (n = 1, 2 8), 128 X 128 matrix, and
30-mm field of view. Pixels were averaged within a 2-mm
vertical range to synthesize a one-dimensional column at each
echo time equivalent to that probed for TI and diffusion meas-
urements. A least-squares-minimization fit to a monoexponen-
tial decay was used to estimate T2 for each point along the
column. This approach is consistent with the observations of
others that noted monoexponential T2 behavior in experimental(C)
brain tumors (23).
All NMR data were acquired on a two Tesla, 20-cm bore
system equipped with actively shielded 1 8-cm gradient coils
(Omega console; GE NMR Instruments). Tumor volume, diffu-
sion, and Tl and T2 data reduction were performed on a
SPARC1O workstation (SUN Microsystems, Mountainview,
CA) using custom software developed within the AVS software
environment (Advanced Visual Systems, Inc., Waltham, MA).
As indicated above, diffusion and relaxation time data were
reduced to one-dimensional plots of each NMR observable
versus right-to-left position through the tumor and brain. Points
within a linear segment through a relatively homogeneous re-
gion of the solid tumor were averaged for each one-dimensional
parametric image. That is, the right-to-left extent of an ROI was
selected using the corresponding T2-weighted image to include
apparent tumor while excluding grossly necrotic or cystic areas
that exhibit high ADC and T2. A separate linear ROI was
defined for contralateral brain.
Treatment Protocol. Seven of the 20 rats did not receive
chemotherapy and served as controls; the other 13 rats received
a single i.p. injection of 13.3 mg/kg of BCNU given between 14
and 19 days after tumor cell inoculation. BCNU was adminis-
tered without anesthesia at least 1.5 days following the last
anesthetic injection and at least 1 h before the next anesthetic
injection to minimize induction of liver enzymes that could
accelerate the clearance of BCNU as described previously (31).
The average tumor volume at the time of treatment was 101
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Diffusion Changes with BCNU Treatment
I #{149} ireateal� - - -x- - Control
BCNU Tx
16 20 24
a
Ti Changes with BCNU Treatment
. Treated
--.)(-- Control
1400
I .4
4. 1.2
Ea
C
U
.� 0.8
0.6-8
2000
1800
i.14,
a.� 1600
I-
1200 -
140
120
�1a,
.! 100
(.4
i-a
80
60 -
24
Time (days)
C
Fig. 2 Average values (means; bars, SE) for water diffusion (a) and Ti(b) and T2 (c) relaxation times for 9L tumors over time for pretreatment(n 15) and post-BCNU treatment (n = 8, day 6; n 4, day 16; n =
1, days 18 -22) are shown. Also plotted are the average values of the MRparameters for evolution of untreated (control) 9L tumors (n 7, day 4;
n 3, day 6).
-8 -4
- - - �BCNUTx
Time (days)
12 16 20 24
b
T2 Changes with BCNU Treatment
. Treated- .-x-. Control
-8 -4 tBCNU Tx � 12 16 20
1460 Therapy Assessment via Quantitative Diffusion MRI
mm3. Four of the treated rats were sacrificed at selected time
intervals following treatment for histological analysis.
Histology. Rat brains were removed and immersed for 3
days in 10% neutral buffered formalin solution. Several coronal
sections corresponding to those studied by MRI were cut from
each brain and embedded in paraffin. The 6-iim sections were
cut, stained with H&E, and evaluated microscopically.
Results
Dependence of MR Parameters on Untreated TumorVolume and Treatment Efficacy. Intracranial 9L tumor dou-bling times (Td) were determined using Eq. A and three or four
pretreatment MRI tumor volume measurements in BCNU-
treated animals and all available volumetric points for untreated
animals. The average doubling time of untreated 9L tumor was
Td = 64 ± 13 h (n = 16 ± SD), which is in close agreement
with earlier studies using this model (31).
A series of coronal T2-weighted images from a single
animal harboring an intracerebral 9L tumor before and follow-
ing BCNU treatment is shown in Fig. 1. Each T2-weighted
image was selected to intersect the largest cross-sectional area
of the tumor and is the section used to prescribe quantitative
acquisitions. The tumor is clearly evident as a hypenntense
lesion in the right hemisphere. At 167 h after treatment, an area
of peritumoral edema is observed followed by tumor shrinkage
and regrowth at 319 and 442 h, respectively. Overlaid on each
image are the locations of the column and ROI representative of
where diffusion and Ti and T2 measurements for tumor and
brain were acquired.
Fig. 2 displays the mean values for the MR-observable
parameters for untreated and BCNU-treated 9L tumors. Data
from treated rats acquired before or within 1 day of treatment
were combined with control rats to improve statistical estimates
of untreated tumor. Data are grouped into 2-day bins. For
control rats, “day 0” corresponds to when its tumor volume was
closest to that of treated rats on the day of treatment (mean
volume, 101 mm3). As shown in Fig. 2a, there is a clear increase
from baseline diffusion values 2-4 days after treatment, which
peaked at 6 days after treatment, then a return to pretreatment
values. The therapeutic effect is accentuated by the fact that
untreated tumor diffusion values remained essentially un-
changed during the entire growth period of the 9L tumors. Of
measured parameters, diffusion exhibited the greatest independ-
ence of untreated tumor volume or growth time, whereas Ti,
and to a lesser extent T2, tended to increase with untreated
tumor growth. Despite this evolution unrelated to treatment,
significant therapeutic changes in tumor Ti and T2 values were
still apparent (Fig. 2, b and c). Tumor NMR quantities were
higher in treated rats on day 6 by 53% for diffusion, 10% for Ti,
and 47% for T2 relative to the three control rats that survived to
day 6.
Summarized in Table 1 are the average values for tumor
volume, ADC, Ti, and T2 for 9L tumor and contralateral brain.
Baseline values represent control rats (when tumor volume was
closest to iOi mm3) combined with treated rats using the last
pretreatment time point or within 24 h of treatment. Only data
collected at the point of peak change in diffusion for the 8 of 13
treated animals that survived 6-8 days after BCNU are repre-
-4 0 4 6 12
TIme (days)
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Td pr.�T. - S5hrs
aa
aCC
C
aC
I-a
100
10
�‘BCNU Tx
---Tumor
- -x- -Contrslstsral� Brain
-1 .4 0 4
Time (days)
12 16 20
X._X�)(��X -X�.-X-X-��
4I�J:rx . . . . . .
a bFig. 3 a, plot of intracerebral 9L tumor volume obtained from MR images versus time for one animal. The initial four-tumor volume measurements
determined the pretreatment tumor doubling time (Td, 55 h) before BCNU (13.3 mg/kg) administration on day 0. Immediately after treatment, thetumor growth rate began to deviate from initial exponential growth, followed by a period of shrinkage. Apparent exponential regrowth occurredbetween 12 and 15 days for a posureatment Td of 86 h. Extrapolation of the regrowth line back to the time of treatment provides an estimate oftreatment log cell kill in this animal. b, plot of diffusion for the 9L tumor and contralateral brain tissue for the animal shown in a. The animal diedduring the last NMR session, accounting for the final drop in diffusion values.
Clinical Cancer Research 1461
Tabl e 1 Peak change in NMR parameters 6-8 days follow ing BCNU treatment”
Pretreatment baseline (n = 15) Posttreatm ent (n = 8)
Solid tumor Contralateral brain Solid tumor Contralateral brain
ADCo (l0� X mm2/sec) 0.91 ± 0.05 0.66 ± 0.02 1.41 ± 0.12p<o.000l”
0.64 ± 0.03NSC
11 (ms) 1665 ± 109 1049 ± 35 1935 ± 66
P<0.000l’1073 ± 50
NSC
T2(ms)
Tumor volum e (mm3)
98±8 59±2
98 ± 45
124±17P < 0005b
182 ± 57P< 0005b
58±4
NSC
a Data are presented as means ± SD.b p comparing baseline and posttreatment groups using two-sample t test.C Not significant to P = 0.05 level.
1000
sented in Table 1 . The observed treatment change in tumor
diffusion was highly significant. Treatment effects on relaxation
times were also significant, whereas no significant changes were
observed in contralateral brain values. The relative increase in
treated tumor values from baseline (i.e., pretreatment) were 55%
for diffusion, 16% for Tl, and 27% for T2.
Determination of treatment log cell kill in individual ani-
mals was accomplished using serial tumor volume data col-
lected over a period long enough to exhibit exponential re-
growth. Extrapolation of the flued exponential regrowth line
back to the time of treatment provides an estimate oflog cell kill
by Eq. B. This method is illustrated for one animal in Fig. 3a,
with the corresponding diffusion time course in Fig. 3b. A total
of four treated rats survived long enough to exhibit exponential
regrowth (Td 75 ± 9 h, n = 4) for an estimate of efficacy: log
cell kill = 1.0 ± 0.3 (mean ± SD, n = 4). These data suggest
that a single LD10 dose (13.3 mg/kg) of BCNU kills approxi-
mately 90% of the 9L tumor cells. At 6-8 days following
I .80
I .30
.� 1.10aa �
0.70C
U
� 0.50
0.30
-8 -4 0 4 8 12 16 20
Time (days)
BCNU treatment, tumor diffusion values increased markedly by
over 50% and returned to pretreatment values during 8-14 days
after treatment. A direct comparison of tumor volume (Fig. 3a)
and diffusion over time following treatment (Fig. 3b) from one
animal reveals that as the tumor growth begins to slow, diffusion
is increased and peaks just before regression of the tumor mass
occurs. Furthermore, the tumor volume exhibits exponential
regrowth at the same time when the diffusion value drops to
pretreatment levels. The relatively high diffusion environment
suggests a significant increase in extracellular water or an in-
crease in permeability between intra- and extracellular water
spaces. For reference, unrestricted water at 40#{176}Chas a diffusion
coefficient of m3.O X l0� mm2/s. Observed peak tumor dif-
fusion values are well below that of pure water, indicating that
although 90% of the tumor cells present at time of treatment
ultimately die, the fractional volume of extracellular space does
not approach 90% at any one time. Upon mass shrinkage and
then regrowth, there is a radical shift back toward a low diffu-
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
- � 44� -�-�-�.:-- � � � ‘?�‘ �
1462 Therapy Assessment via Quantitative Diffusion MRI
Fig. 4 H&E-stained coronal histological sections of 9L tumors from an untreated rat (a), a rat at 6 days (b), 9 days (c), and 12 days (6) followingBCNU treatment. At 12 days after treatment, areas of tumor regrowth (d, left) are histologically similar to untreated tumor (a), although coagulativenecrosis is also apparent (d, right). All photomicrographs are at the same magnification.
sion; this is consistent with a more restricted diffusion or cel-
lular environment. Tumor diffusion values at or below pretreat-
ment levels suggest a relatively high cellular density with a
moderate extracellular volume fraction as exhibited by untreated
9L tumors.
Histopathology of treated animals sacrificed at selected
intervals following BCNU treatment support the model of a
therapy-induced increase of extracellular space. Relative to a
control tumor (Fig. 4�i), tumors 6 days following BCNU treat-
ment revealed a marked increase in tumor extracellular space,
which was filled with mucosubstance (Fig. 4b). The lympho-
cyte-predominant mixed inflammatory response was most in-
tense 6 days after treatment. There was also a noticeable in-
crease in pleomorphism, large cells, and in cells with structural
features of apoptosis. Peritumoral edema was also most exten-
sive 6 days after treatment. Nine days after BCNU treatment, the
extracellular space diminished (Fig. 4c). At 12 days, the number
of inflammatory cells decreased, and regions of reduced diffu-
sion were associated with dense regrowth of tumor cells (Fig.
4d, left) that was similar in appearance to untreated tumors (Fig.
4a). Geographic regions of coagulative necrosis appeared in the
same specimen as shown in Fig. 4d, right. Fig. 5a illustrates the
T2-weighted MRI of the animal sacrificed for the histology 12
days after treatment. Associated diffusion plots (Fig. Sb) clearly
demonstrate tumor heterogeneity, with cellular dense and ne-
crosed regions as low and high diffusion environments, respec-
tively. These findings underline the importance of spatially
resolved diffusion measurements because even relatively homo-
geneous tumors, such as the 9L gliosarcoma, may become
heterogeneous following therapy.
Diffusion Anisotropy. Diffusion anisotropy is known to
exist in highly ordered structures such as myelinated white
mauer fiber tracts (10-12). It was interesting to note that sig-
nificant diffusion anisotropy also exists within and near the
edges of the 9L tumor mass. The most conspicuous and con-
sistent anisotropy pattern was that of low diffusion perpendic-
ular to the mass surface (ADCX in these experiments) and
relatively high diffusion in orthogonal directions. Although it
was anticipated that tumors would have a highly disorganized
cellular pattern (i.e., isotropic), histological inspection revealed
a significant cellular directionality at the tumor boundary as
shown in Fig. 6. Note there is an apparent lamination of tumor
and adjacent brain cells that are radially compressed by mass
effect. As a result, diffusion perpendicular to the tumor surface
is impeded relative to diffusion with the cellular “grain,” i.e.,
parallel to the tumor surface. This pattern of strong diffusion
anisotropy at the tumor edges was observed in all rats studied
and is exemplified in Fig. 7a. Amsotropic diffusion in the
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Post-Tx 281 Hr
Clinical Cancer Research 1463
12 Days Post-BCNU1 .8
1 .6
1 .4
I .2
0.8
0.6
0.4
0.2
.6mm
a,”
�CSF
�.. Coag. �N ecro. a -12 -6 .4 0 4 5
RightLocstion(mm) Lsft
bFig. 5 a, coronal T2-weighted image of a rat with a 9L tumor that was treated 12 days earlier with BCNU. Tissue regions corresponding to recurrent
tumor, coagulative necrosis, and cerebrospinal fluid in the ventricle are shown, which were positively identified by histology (shown in Fig. 4d). Thelocation of the column used for obtaining diffusion data is enscribed on the image. b, plot of water diffusion as a function of position along the column
shown in a. Note the high contrast in diffusion between solid recurrent tumor and the necrotic region.
.. L #{149}�., � �
,. � 0’ #{149}
� (a
� �‘ ‘:‘
� � i�’
U
a
a
C
U
Fig. 6 H&E-stained coronal histological section of the 9L tumor/braininterface from an untreated rat. Note the anisotropic pattern of cellular
shape and distribution at the brain/tumor interface.
central tumor was apparent in some rats, but this pattern was not
observed consistently. The degree of anisotropy varied with
time toward more isotropic high diffusion 6-8 days after treat-
ment and was isotropic in areas of coagulative necrosis. Pre-
sumably, this is a result of less directional cellular order because
as cells necrose, the mass “loosens” with increasing extracellu-
lar volume leading to a more isotropic environment. This is
illustrated in Fig. 7a at 30 and I 5 1 h after treatment (Fig. 7b).
Discussion
Human brain tumors are diverse in response to treatment.
Hence improved methods for reliable quantitation of treatment
response in individual patients would provide additional time
for applying alternative treatments to tumors unresponsive to the
initial therapeutic regimen. A common approach used in esti-
mating tumor “size” is by measuring two major axes of the
tumor from clinical CT or MR images. A positive treatment
response is typically defined as a substantial decrease in cross-
sectional tumor area from the pretreatment size. This approach
requires substantial shrinkage of the tumor mass that may pro-
ceed relatively slowly due to the time necessary for absorption
of cellular debris. Further complications in the estimation of
treatment efficacy and tumor size often arise from malignant
brain tumors that have ill-defined margins between viable tumor
and brain tissue due to peritumoral edema, as observed by
T2-weighted MR images. This edema may be secondary to
tumor involvement or represent regional alterations in noncan-
cerous brain parenchyma in response to treatment. Contrast-
enhanced CT and MR imaging offers additional sensitivity, but
enhancement characteristics may not always be directly related
to the presence or absence of tumor cells. Consequently, al-
though morphological imaging is relied upon as the principle
clinical tool, it often leads to only crude estimates of tumor size,
and pronouncement of treatment response can be made only
relatively late following treatment initiation through evidence of
substantial tumor regression or progression of growth.
The main objective of this preclinical study was to evaluate
quantitative diffusion and relaxation time MRI measurements as
early therapeutic response indicators using the 9L brain tumor
model. This widely used model has been shown previously to be
responsive to BCNU treatment (3 1 , 32). Serial tumor volume
measurements provided a classic indicator of treatment response
for comparison with quantitative NMR parameters. As is com-
monly observed with some tumor models and human gliomas,
volume determinations are sensitive to error from ill-defined
tumor boundaries, particularly when the ROI are manually de-
fined. The 9L model was chosen for these studies because it is
reported to have well-demarcated borders and minimal peritu-
moral edema, features that allow for noninvasive quantitation of
tumor volumes and cell kill in individual animals using MRI
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1 .8
1 .6
1.25 Days Post-B CNU
ADCx (R/L)
-----ADCy
ADCz
0.4
0.2
.9 -6 -3 0 3 6
Right Location (mm) Left
a
9?
Days Post-B CNU
AL)cX (FIlL)
ADCz
I .8
1 .6
I .4
� 1.2
E
0
� 0.8
C
c) 0.6
.� 0.4
Fig. 7 Overlaid plots of ADCX, ADC�, and ADC2 1.25 days (a) after BCNU treatment and 6.3 days (b) after treatment in the same animal when
diffusion changes are maximal. Note that the high diffusion anisotropy decreases with time following treatment, particularly at the tumor edges.
.9
Right
.3 0
Location (mm)
bLeft
1464 Therapy Assessment via Quantitative Diffusion MRI
__I .4C.,a,IC
E
‘-4
0 0.8
Q
.� 0.6
(3 1 ). Even in the 9L model, however, tumor boundaries become
indistinct during the period following treatment but prior to
substantial regrowth. Indeed, it is during this time that cellular
density changes are greatest as indicated by diffusion. If, how-
ever, survival is long enough for most of the cellular debris to be
removed, then tumor regrowth will appear exponential, thus
allowing an estimate of doubling time. Note knowledge of
pretreatment tumor doubling time is not essential if volume is
directly measured at the time of treatment. If one assumes the
tumor cellular density during regrowth is the same as prior to
treatment, then Eq. B provides an estimate of log cell kill. In this
regard, histology from control 9L and posttreatment tumor dur-
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Clinical Cancer Research 1465
ing regrowth appear qualitatively similar in terms of cellular
density. A quantitative measure of cell density from histology
was not performed because the fixation process is known to alter
extra- and intracellular spaces. We can consider the theoretical
impact of reduced cellular density (i.e., number of cells/unit
volume) during regrowth on calculated posttreatment Td and
cell kill. If upon regrowth the cellular density is constant but
lower than before treatment and tumor cell division rate is
unchanged, then the apparent posttreatment tumor doubling time
will be reduced. This is an unlikely scenario given that, empir-
ically, Td during regrowth is not reduced. Alternatively, if there
is a transient elevation in extracellular volume fraction due to an
incomplete removal of necrotic debris, then there will be an
apparent increase in Td, leading to an underestimation of cell
kill. Although the potential of this error cannot be excluded, the
estimation of cell kill using only animals that exhibited signif-
icant regrowth minimized it. In these animals, the final tumor
volume was greater than 4-fold the volume at the time of
treatment such that the necrotic volume was a relatively small
fraction of the mass.
Diffusion provides further evidence that cellular density
during late regrowth is similar to untreated tumor. Using a
mathematical model of the dependence of diffusion on cellular
properties, others have argued that the extracellular fluid volume
fraction is the main determinant of measured diffusion (9). The
observation of diffusion returning to pretreatment levels during
tumor regrowth suggests that the extracellular volume fraction,
and thus cellular density, is similar to untreated tumor. There-
fore, the effects of treatment on MR-observable parameters can
be quantitated and compared with treatment efficacy in the same
animal. This provides a powerful approach for evaluating the
dynamic changes in MR parameters of interest following treat-
ment for correlation with the effects on tumor growth and
histopathology.
Our results revealed a substantial increase in tumor water
diffusion that occurred early following administration of BCNU
and before tumor regression. Although relaxation times were
also affected by the treatment, diffusion values were found to be
a more robust predictor of treatment-induced changes to cellular
tissue integrity. In fact, histological comparisons of 9L tumors
treated with BCNU suggest that treatment-induced changes in
tumor diffusion values are reflective of the changes in micro-
scopic water environment that occurred following treatment and
during repopulation of the tumor from the surviving clonogenic
tumor cells. Based on this data, we believe the increased diffu-
sion is primarily a result of a proportional increase in extracel-
lular water because a large fraction of tumor cells necrose with
dissolution of cell membranes and secondarily from vasogenic
edema and an inflammatory response. An alternative scenario is
that observed diffusion changes are primarily due to vasogenic
edema and an inflammatory response to treatment, with a coin-
cident arrest of tumor growth (in contrast to tumor cell death),
followed by resolution of edema and regrowth. This scenario,
however, is not supported by colony-forming assays that yield
yet greater cell kill in the 9L model at the same BCNU dose
(32). This suggests the our log kill of 1 is not an overestimate.
Consistent with studies of untreated tumor (17-23), we
have observed that necrosis is clearly evident by high diffusion.
Furthermore, recurrent tumor could be distinguished from ther-
apy-induced regional necrosis by localized MR diffusion meas-
urement, thereby providing important diagnostic information
not easily obtainable by other nonlocalized or low-resolution
spectroscopic methods. The ability to detect diffusion anisot-
ropy within tumors prior to treatment and progression to isotro-
pic diffusion following treatment reveals the sensitivity of this
approach to obtain potentially important information related to
patterns of organized cellular distribution.
In conclusion, in this present study we extended our pre-
vious work by quantitation of cell kill following treatment and
identified treatment-induced histological changes that may be
important factors in affecting the observed changes in water
mobility. Furthermore, diffusion anisotropy was observed near
the boundaries of intracranial 9L tumors, which reflected direc-
tionality in water mobility that decreased upon successful treat-
ment. Although growth retardation of treated 9L tumors was
observable by serial MR volume measurements, precise tumor
volume measurements are not readily obtainable from human
brain tumor patients for reasons discussed above. Therefore, the
results of this study suggest that quantitative water diffusion
offers significant potential in the early assessment of antineo-
plastic treatment response and should provide the motivation for
validating this quantitative noninvasive approach for monitoring
tumor treatment in the clinical setting.
References
I. Walker, M. D.. Green, S. B., Byar, D. P., Alexander, E. Jr., Batzdorf,
U., Brooks, W. H., Hunt, W. E., MacCarty, C. S., Mahaley, M. S., Jr.,
Mealey. J., Jr., Owens, G.. Ransohoff, J., Robertson, J. T., Shapiro,W. R., Smith, K. R., Jr., Wilson, C. B., and Strike, T. A. Randomizedcomparisons of radiotherapy and nitrosoureas for the treatment of ma-
lignarn glioma after surgery. N. Engl. J. Med., 303: 1323-1329, 1986.
2. Grossman, S. A., and Burch, P. A. Quantitation of tumor response to
antineoplastic therapy. Semin. Oncol., 15: 441-454, 1988.
3. Levin, V. A., Crafts, D. C.. Norman, D. M., Hoffer. P. B.. Spire, J. P.,and Wilson, C. B. Criteria for evaluating patients undergoing chemo-
therapy for malignant brain tumors. J. Neurosurg., 47: 329-335, 1977.
4. Wilson, C. B., Crafts, D., and Levin, V. A. Brain tumors: criteria of
response and definition of recurrence. Natl. Cancer Inst. Monogr., 46:
197-203, 1977.
5. Grossman, S. A. Chemotherapy of brain tumors. In: M. Salcman(ed), Concepts in Neurosurgery: 4, pp. 321-340. Baltimore: Williams
and Wilkins, 1991.
6. Bradley, W. G., Yuh, W. T. C., and Bydder, G. M. Use of MRimaging contrast agents in the brain. J. Magn. Reson. Imaging, 3:199-218, 1993.
7. Stejskal, E. 0. Use of spin echoes in a pulsed magnetic-field gradient
to study anisotropic, restricted diffusion and flow. J. Chem. Phys., 43:
3597-3603, 1965.
8. Tanner, J. E. Transient diffusion in a system partitioned by perme-
able barriers. Application to NMR measurements with a pulsed field
gradient. J. Chem. Phys., 69: 1748-1754, 1978.
9. Latour, L. L., Svoboda, K., Mitra, P. P., and Sotak, C. H. Time-dependent diffusion of water in a biological model system. Proc. NatI.
Acad. Sci. USA, 91: 1229-1233, 1994.
10. Moseley, M. E., Cohen, Y., Kucharczyk, J., Mintorovitch, J., As-
garitt, S., Wendland, M. F., Tsuruda, J., and Norman, D. Diffusion-
weighted MR imaging of anisotropic water diffusion in cat central
nervous system. Radiology, 176: 439-445, 1990.
I 1. Chenevert, T. L., Brunberg. J. A., and Pipe, J. G. Anisotropic
diffusion in human white matter: demonstration with MR techniques invivo. Radiology, /77: 401-405, 1990.
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1466 Therapy Assessment via Quantitative Diffusion MRI
12. Doran, M., Hajnal. J. V., Van Bruggen, N., King, M. D., Young.
I. R., and Bydder, 0. M. Normal and abnormal white matter tractsshown by MR imaging using directional diffusion weighted sequences.
J. Comput. Assisted Tomogr., 14: 865-873. 1990.
13. Mintorovitch, J., Moseley, M. E., Chileuitt, L., Shimizu, H., Cohen,
Y., and Weinstein, P. R. Comparison of diffusion- and T2-weightedMRI for the early detection of cerebral ischemia and reperfusion in rats.Magn. Reson. Med., 18: 39-50, 1991.
14. Moseley, M. E., Kucharczyk, J., Mintorovitch, J., Cohen, Y.,
Kurhanewicz, J., Derugin, N., Asgari, H., and Norman, D. Diffusion-
weighted MR imaging of acute stroke: correlation with T2-weighted and
magnetic susceptibility-enhanced MR imaging in cats. Am. J. Neurora-
diol., Ii: 423-429, 1990.
15. Warach. S., Chien, D., Li, W., Ronthal, M., and Edelman, R. R. Fast
magnetic resonance diffusion-weighted imaging of acute human stroke.
Neurology, 42: 1717-1723, 1992.
16. Sorensen, A. 0., Buonanno, F. S., Gonzalez, R. G., Schwamm,
L. H., Lev, M. H., Huang-Hellinger, F. R., Reese, T. 0., Weisskoff,R. M.. Davis, T. L., Suwanwela, N., Can, U., Moreira, J. A., Copen,w. A., Look, R. B., Finklestein, S. P., Rosen, B. R., and Koroshetz,W. J. Hyperacute stroke: evaluation with combined multisection diffu-sion-weighted and hemodynamically weighted echo-planar MR imag-
ing. Radiology, 199: 391-401, 1996.
17. Le Bihan, D., Breton, E., Lallemand, D., Grenier, P., Cabanis, E.,
and Laval-Jeantet, M. MR imaging of intravoxel incoherent motions:
application to diffusion and perfusion in neurologic disorders. Radiol-
ogy, 161: 401-407, 1986.
18. Hajnal, J. V.. Doran, M., Hall, A. S., Collins, A. 0., Oatridge, A.,
Pennock, J. M., Young, I. R., and Bydder, G. M. MR imaging ofanisotropically restricted diffusion of water in the nervous system:
technical, anatomic, and pathologic considerations. J. Comput. Assisted
Tomogr., 15: 1-18, 1991.
19. Brunberg, J. A., Chenevert, T. L., McKeever, P. E., Ross, D. A.,Junck, L. R., Muraszko, K. M., Dauser, R., Pipe, J. 0., and Betley, A. T.
In vivo MR determination of water diffusion coefficients and diffusionanisotropy: correlation with structural alteration in gliomas of the cer-
ebral hemispheres. Am. J. Neuroradiol., 16: 361-371, 1995.
20. Tsuruda, J. S., Chew. W. M., Moseley, M. E., and Norman, D.Diffusion-weighted MR imaging of the brain: value of differentiating
between extraaxial cysts and epidermoid tumors. Am. J. Neuroradiol.,
II: 925-931, 1990.
21. Chenevert, T. L., Ross, B. D., Pipe, J. G., and Simerville, S. J.
Quantitative diffusion and anisotropy in rat gliomas. Proc. Soc. Magn.
Reson. Med., 2: 787, 1991.
22. Maier, C. F., Paran, Y., Bendel, P., Rutt, B. K., and Degani, H.
Quantitative diffusion imaging in implanted human breast tumors.
Magn. Reson. Med., 37: 576-581, 1995.
23. Eis, M., Els, T., and Hoehn-Berlage, M. High resolution quantita-
tive relaxation and diffusion MRI of three different experimental brain
tumors in rat. Magn. Reson. Med., 34: 835-844, 1995.
24. Ross, B. D., Chenevert, T. L., Kim, B., and Ben-Yoseph, 0.
Magnetic resonance imaging and spectroscopy: application to experi-
mental neuro-oncology. Quart. Magn. Res. Biol. Med., 1: 89-106,1994.
25. Chenevert, T. L., and Ross, B. D. Diffusion, Tl, and T2 as chemo-therapeutic response predictors of experimental brain tumors. Proc. Soc.
Magn. Reson. Med., 2: 924, 1995.
26. Zhao, M., Pipe, J. G., Bonnett, J., and Evelhoch, J. L. Early
detection of treatment response by diffusion-weighted 1H-NMR spec-
troscopy in a murine tumour in vivo. Br. J. Cancer, 73: 61-64, 1996.
27. Jakobsen, I., Lyng, H., Kaalhus, 0., and Rofstad, E. K. MRI of
human tumor xenographs in vivo: proton relaxation times and extracel-
lular tumor volume. Magn. Reson. Imaging, 13: 693-700, 1995.
28. Nijhof, E. J. Slotted resonator: principles and applications for high-
frequency imaging and spectroscopy on electrically conducting samples.
Magn. Reson. Imaging, 8: 345-349, 1990.
29. Basser, P. J., Matiello, J., and Le Bihan, D. MR diffusion tensor
spectroscopy and imaging. Biophys. J., 66: 259-267, 1994.
30. Levy, 0. C., and Peat, I. R. The experimental approach to accurate
carbon-l3 Tl measurements. J. Magn. Reson., 18: 500-521, 1975.
31. Kim, B., Chenevert, T. L., and Ross, B. D. Growth kinetics and
treatment response of the intracerebral rat 9L brain tumor model: a
quantitative in vivo study using magnetic resonance imaging. Clin.
Cancer Res., I: 643-650, 1995.
32. Rosenblum, M. L., Knebel, K. D., Vasquez, D. A., and Wilson,
C. B. Brain-tumor therapy: quantitative analysis using a model system.J. Neurosurg., 46: 145-154, 1977.
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1997;3:1457-1466. Clin Cancer Res T L Chenevert, P E McKeever and B D Ross therapy using diffusion magnetic resonance imaging.Monitoring early response of experimental brain tumors to
Updated version
http://clincancerres.aacrjournals.org/content/3/9/1457
Access the most recent version of this article at:
E-mail alerts related to this article or journal.Sign up to receive free email-alerts
Subscriptions
Reprints and
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Permissions
Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)
.http://clincancerres.aacrjournals.org/content/3/9/1457To request permission to re-use all or part of this article, use this link
Research. on June 30, 2020. © 1997 American Association for Cancerclincancerres.aacrjournals.org Downloaded from