[CANCER RESEARCH 50. 2943-2948. May 15, 1990]
Preliminary Correlations of Clinical Outcome with in Vitro Chemosensitivity of
Second Passage Human Breast Cancer Cells1
Hélène
S. Smith,2 Wainer Zoli, Annalisa Volpi, Alan Hiller, Marc Lippman, Sandra Swain, Brian Mayall,
Charles Dollbaum, Adeline J. Hackett, and DiñoAmadori
GéraldineBrush Cancer Research Institute, Pacific Presbyterian Medical Center, San Francisco, California 94115 [H. S. S., A. H.J; Morgani-Pieranloni Hospital, Farli,
Italy [W. Z., A. V., D. A.]; Vincent T. Lombardi Cancer Research Center, Georgetown University Medical Center, Washington, DC 20007 [M. L., S. S.J; Department of
Laboratory Medicine, University of California, San Francisco, San Francisco, California 94143 {B. M.J; and Peralta Cancer Research Institute, Oakland, California
94609 [C. D., A. J. HJ
ABSTRACT
These studies describe the clinical correlations of 63 in vitro chemosensitivity assays on breast cancer cells after short-term monolayer
culture. Forty-five of the assays were single agent correlations. Based on
cut-off values determined empirically, the test accurately predicted re
sistance for 36 of 41 patients (88%) who did not respond to the drug. It
also predicted sensitivity with a high degree of accuracy: 21 of 22 patients
(95%) who responded to the drug tested had a sensitive assay. In five
cases, two biopsies were evaluated from the same patient. Whenever
assays were performed before and after treatment with a given drug,
tumor cells from the second biopsy were more resistant in vitro if the
patient failed on therapy. If the patient did not fail, but stopped therapy
for other reasons, or if there was no intervening therapy with the tested
drug, the two biopsies remained similar in drug sensitivity. These results
suggest that in vitro chemosensitivity assays which accurately predict
both sensitivity and resistance can be obtained with breast cancer cells
after short-term culture and that further prospective trials are warranted.
INTRODUCTION
A number of different in vitro chemosensitivity assays have
been described which have proven to be highly reliable in
predicting response of a patient's tumor to chemotherapy (for
reviews see Refs. 1-3). All of these assays have in common that
tumor cells taken directly from the patient are treated with
drugs and then cultured. Unfortunately, the small size of many
breast cancer biopsies makes these assays unsuitable for 70 to
80% of breast specimens (2). This problem is likely to become
even more acute because (a) breast cancer biopsies will be
smaller as early diagnostic modalities are more frequently used;
and (b) new laboratory tests for predicting prognosis will require
additional specimen material.
One solution to the problem of small breast cancer size is to
amplify the available tumor cells by in vitro culture prior to
testing for drug sensitivity. This report presents preliminary
studies evaluating one such system. To obtain sufficient prolif
eration at first passage in culture, a medium developed specifi
cally for human mammary epithelium (4) was utilized. In this
medium, the proliferating cultures are easily trypsinized to
single cells which clone with high efficiency when plated on
irradiated fibroblasts (5). Analysis of X-ray survival curves
excluded the possibility that clumps, rather than single cells,
were being plated (6).3 Furthermore, the number of colonies
scored was directly proportional to the number of cells plated
(5). With this culture system, the vast majority (>70% of
Received 6/23/89; revised 12/21/89.
The costs of publication of this article were defrayed in part by the payment
of page charges. This article must therefore be hereby marked advertisement in
accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
' Supported by the American Cancer Society (Grant PDT 282), the Abramson
Foundation, National Cancer Institute Grant POI CA44768. and the Instituto
Oncologico Romagnolo (Grant 87221.2), Forli, Italy.
2To whom requests for reprints should be addressed, at GéraldineBrush
Cancer Research Institute, Pacific Presbyterian Medical Center, 2330 Clay St.,
San Francisco, CA 94115.
3 H. S. Smith, unpublished observations.
primary breast cancers and hypodermal métastasesgrow well
and clone with high efficiency). In contrast, effusion métastases
grow rather slowly in primary culture and usually do not clone
(7). Other types of cancers, such as colon or non-small cell lung
cancers grow poorly and do not clone.3
The cells growing in this assay have been extensively char
acterized. All of the cultures were uniformly positive for kera
tins (8) and other epithelial markers (5), proving that epithelial,
rather than stromal cells, were being cultured. Other criteria
consistently distinguished cultures derived from malignant and
nonmalignant tissues, including amnion invasiveness (9), multinucleation in response to cytochalasin (10), and presence of a
tumor-associated glycoprotein (8). Most recently, the assay was
used to show that tumor-derived cultures were sensitive to
tumor necrosis factor while nonmalignant mammary epithe
lium-derived cultures were not (11).
The fact that the cells cultured from tumors differ from those
of nonmalignant specimens by so many criteria suggests that at
least some bona fide tumor cells are in fact being cultured.
However, it is likely that only some, less aggressive, tumorderived subpopulations are selectively grown in the system.
Although cells from hypodermal métastaseswere more abnor
mal than those from primary breast cancers, cells from both
types of lesions showed only minimal karyotypic changes after
culture rather than the gross aneuploidies usually associated
with carcinomas. In contrast, effusion métastases,a more ag
gressive manifestation of the disease which usually grew poorly,
were uniformly aneuploid (9, 12, 13).
We have undertaken studies to test the hypothesis that ac
curate predictive chemosensitivity assays could be obtained with
this assay even though (a) the cell cycle kinetics differ from the
in vivo state since all of the cells are cycling after culture; and
(b) there is less heterogeneity since only selected subpopulations
grow in this culture system. In a previous study (14), response
to doxorubicin in this system was examined. A broad range of
sensitivities in specimens from previously untreated patients
was shown, indicating inherent differences in drug sensitivity
even in proliferating cell cultures. That these differences might
be clinically meaningful was suggested by the fact that speci
mens from donors who had failed doxorubicin-containing
chemotherapy regimens had dose-response curves that clustered
at the more resistant end of the spectrum.
In this report, we present our continuing studies to evaluate
the clinical relevance of the assay by extending our observations
to other drugs besides doxorubicin and to prospective predic
tions of sensitivity. We show that the assay predicts both
sensitivity and resistance with a high degree of accuracy.
MATERIALS
AND METHODS
Specimen Material. Specimens were obtained from discarded surgical
material or from patients who had excisional biopsies of skin métastases
for the purpose of in vitro culture after obtaining appropriate informed
2943
Downloaded from cancerres.aacrjournals.org on July 12, 2017. © 1990 American Association for Cancer Research.
IN VITRO CHEMOSENSITIVITY
OF HUMAN
consent. In vitro assays were performed at Oakland, CA, and Forli,
Italy. The assay was developed by one of the authors (H. S. S.) in
Oakland, CA and another of the authors (W. Z.) was trained in its use
both at that site and in Forli, Italy. For each drug tested, preliminary
studies found that both laboratories observed similar cloning efficien
cies and colony morphologies. Furthermore, both laboratories observed
the same range of sensitivities among carcinomas from donors with no
prior therapy (data not shown).
Tissue Collection and Preparation. The processing of tissue for the
separation of carcinoma from stromal components was as described
previously (5). Briefly, tissue was carefully minced, and a portion of the
minced tissue was fixed in formalin and processed for histológica!
examination. The remainder of the mince was treated with collagenase
and hyaluronidase to digest stroma and basement membranes. The
digestion was terminated when observation of a sample showed small
clumps and single cells. The clumps were either separated from the
dissociated stromal cells by filtration with polyester screen filters or
plated directly into T-25 flasks (Corning Glass Works, Corning, NY)
in an enriched medium (MM4) (4) and which consisted of the following:
30% Dulbecco's modified Eagle medium, 30% Ham's F-12 medium,
15% conditioned medium from human fetal intestine epithelial cells,
74 Int; 15% conditioned medium from human bladder epithelial cell
line, 767B1; 9.5% conditioned medium from human myoepithelial cell
line 578 Bst; 0.5% fetal calf serum; 10 ¿ig
insulin/ml (Sigma Chemical
Co., St. Louis, MO), 5 ng/ml epidermal growth factor (Collaborative
Research, Waltham, MA), 10~8 M triiodothyronine (Sigma), 0.1 ^g
hydrocortisone/ml (Sigma), 10~9M estradici (Sigma), and 1 ng cholera
toxin/ml (Sigma). The cells lines used for producing conditioned media
can be obtained from the American Type Culture Collection (Rockville,
MD). Alternately, endothelial cells from human umbilical vein (15) or
bovine aorta3 can replace the cell lines as a source of conditioned media.
Within 24-48 h of plating, tumor cells migrated from the clumps
and began to proliferate extensively. By 6-8 days after plating, the
combined effect of cell migration and extensive proliferation was a wide
area of epithelial cells with extensive mitotic activity surrounding each
clump. Very brief treatment with trypsin at this stage resulted in
dissociation of contaminating fibroblasts (which also grew poorly in
MM) while leaving the tumor epithelial cells attached to the flask. One
to 2 days later, a somewhat more vigorous trypsinization removed many
of the cells which had migrated from the central clump. This process
was termed "partial trypsinization." The dish was then refed and the
cell-harvesting process was repeated one or two more times. The
suspended cells were utilized directly in the drug assays.
Clonal Assay for Drug Sensitivity. Breast cancer cells under condi
tions of clonal growth yielded plating efficiencies ranging from 1 to
10%. As described previously (5, 14), UV-irradiated human fibroblasts
were seeded at 1.5 x IO4 cells onto 35-mm tissue culture dishes to
provide a feeder layer and all human fibroblast strains tested were
equally effective. The next morning, tumor cells harvested from primarycultures were dispersed to a single-cell suspension and plated. Sixteen
to 19 h after seeding of tumor cells, the medium was removed and MM
with the desired drug concentration was added to each dish. For each
drug dose tested, cells are seeded at 350, 1,000, and 35,000 cells/25mm-diameter culture dish. All dishes with a countable number of
surviving colonies are scored. Three drug concentrations were used to
establish a dose-response relationship. The drug-containing and control
dishes were incubated at 37°C.The medium was removed 4 h later and
the dishes were washed once with basal salts. Each dish was then refed,
without drug, with MM containing penicillin and streptomycin (100
units/ml each) and IO4 freshly prepared irradiated fibroblasts. The
dishes were refed twice weekly until readily visible colonies were present
(usually within 5-10 days). The dishes were rinsed with phosphatebuffered saline (80 mg NaCl/ml, 0.2 mg KCl/ml, 0.2 mg KH2PO4/ml,
0.1 mg CaCl2/ml, and 0.1 mg MgCl2-6H2O/ml), fixed with methanol,
and stained with May-Grunwald-Giemsa.
The colonies in each dish were counted, and the number of colonies
per 100 plated cells was calculated. The value from the drug-treated
cultures was divided by the mean number of colonies per 100 cells
BREAST
plated on
generated
countable
response
CANCER CELLS
control dishes (without drug). The dose-response curves were
by calculation of the mean and standard deviation for all
dishes for a given point. There were no differences in dose
at different seeding levels.' The relative resistance of the
cultures was expressed as the LD90 values. The staff responsible for
performing the in vitro assays had no knowledge of the patients' clinical
response until after assay results were reported to the clinicians.
Drug Preparation. Stock solutions of each drug at 1 mg/ml were
divided into aliquots and frozen at -70°C. For each experiment, a new
ampoule was used. Drugs and stock solutions were as follows: doxorubicin (Adria Laboratories, Columbus. OH) diluted with sterile phys
iological saline; cisplatinum (Bristol, Syracuse, NY) diluted with sterile
water, vinblastine sulfate (Eli Lilly, Indianapolis, IN) diluted with sterile
physiological saline; 4-epidoxorubicin and idarubicinol (a generous gift
of Dr. S. Ganzina, Farmitalia, Milan, Italy) were diluted with sterile
physiological saline. Idarubicinol is the active metabolite of 4-demethoxydaunorubicin, a new derivative of daunorubicin synthesized by
Farmitalia Carlo Erba (termed idarubicin, or MI30, IMI30, NSC
256439). After i.v. and p.o. administration in humans, 4-demethoxydaunorubicin is rapidly metabolized to its 13-dihydroderivative, idarub
icinol. which has been shown to be an active metabolite in experimental
models.
Patient Correlations. To obtain clinical correlations, all patients were
clinically evaluated by two different clinicians who were not aware of
the in vitro results. The following criteria were used. Complete response,
the disappearance of all known disease determined by two observations
separated by at least 4 weeks. None of the cases in this study had a
complete response. Partial response, at least a 50% reduction in the
size of all measurable tumor areas as measured by the sum of the
products of the greatest length and maximum width. No lesions may
progress and no new lesions may appear. These parameters must have
been present for at least 2 measurement periods separated by at least 4
weeks. Stable disease, when a patient's status fails to qualify for either
a response or progressive disease. Progressive disease, an increase of
>25% and at least 2 cm2 of original lesions and of the development of
any new lesion.
RESULTS
Three anthracycline derivatives were evaluated: doxorubicin,
4'-epidoxorubicin, and idarubicinol (the active metabolite of 4demethoxydaunorubicin).
Dose-response curves representing
sensitive and resistant assays for each analogue are presented
in Fig. 1. All of the anthracycline data accumulated for speci
mens from patients with no prior therapy are illustrated in Fig.
2.
To obtain clinical correlations, specimens were divided into
two categories, those clinically sensitive and those clinically
resistant to chemotherapy. Patients were considered clinically
resistant with either "progressive" or "stable" disease, and
100.0
0.2
0.4
0.6
0
0.4 0.6
Drug Concentration
0
0.2 0.4 0.6
(/¿g/ml)
Fig. 1. Anthracycline dose-response curves for human breast cancer. A, dox
orubicin; B, 4' epidoxorubicin; C, idarubicinol. D. typical sensitive specimen; •
typical resistant specimen; bars. SD.
4 The abbreviations used are: MM, enriched medium; LD«,,drug concentration
killing 90% of colony-forming cells.
2944
Downloaded from cancerres.aacrjournals.org on July 12, 2017. © 1990 American Association for Cancer Research.
l\
VITRO CHEMOSENSITIVITY
OF HUMAN
BREAST CANCER CELLS
legend, the different symbols indicate whether the correlations
were prospective or retrospective and whether the patients were
treated with single or multidrug regimens.
For each of the analogues, there was a broad range of LD90
10.0values for specimens from patients with no prior therapy. Most
of these were primary breast cancers; however, samples of
hypodermal métastasesfrom patients with no prior therapy,
:=
although fewer in number, showed the same broad range of
LD90 values. Those samples classified as clinically resistant in
vivo tended to cluster with LD90 values at the resistant end of
the spectrum when tested in vitro. In contrast, tumors from
0 0.2 0.4 0.6 0 0.2 0.4 0.6 0 0.2 0.4 0.6
patients who responded with either complete or partial re
Drug Concentration
(¿¿g/ml)
sponses clustered with more sensitive LD90 values.
The accumulated correlations for Vinca alkaloids are illus
Fig. 2. Composite of anthracycline dose-response curves for human breast
cancers from patients without prior therapy. A, doxorubicin; B. 4'-epidoxorubicin;
trated in Fig. 4. Preliminary studies in culture indicated that all
C, idarubicinol.
specimens gave similar LD90 values when assayed with vincristine or vinblastine (data not shown). All of the in vitro assays
sensitive with either "partial" or "complete" remissions. Sam
were done with one compound, vinblastine. Most of the patients
ples from clinical sensitive patients were obtained by biopsy were also treated with vinblastine; however, two received vincristine not vinblastine. Both of these patients were nonreimmediately prior to a chemotherapy regimen where sensitivity
is defined as a partial or complete response to that therapy. For sponders; assays of their tumor yielded LD90 values of 0.043
fig/ml and >0.075 ng/m\. Similar to the anthracyclines, there
in vitro predictions of sensitivity, only single agent therapies
were included in the study since one cannot ascertain which was a broad range of LD90 values for both primary breast
drug or drugs in a combination regimen are responsible for a cancers and hypodermal métastasesfrom previously untreated
patient's response to multiagent combinations. Samples from patients. All of the cases where patients responded to therapy
with Vinca alkaloids clustered at the sensitive end of the LD90
clinically resistant patients were obtained in 2 ways: (a) those
value spectrum. Most of the specimens from nonresponders
biopsied immediately prior to treatment with chemotherapy
had LD90 values more resistant than those of the responders.
when the patient subsequently showed either stable or progres
In three cases, the cells were sensitive to vinblastine, but the
sive disease on therapy (prospective correlation); or (b) those
patients did not show either an objective partial or complete
biopsied at the time of therapeutic failure following treatment
remission. These three cases are considered negative correla
(retrospective correlation). We evaluated specimens from clin
tions for the assay. Two of these patients progressed on therapy
ically resistant patients treated with either single or multidrug
with vinblastine; however, the third had stable disease for 5
regimens since patients failing a given combination regimen
months. When this third patient relapsed, a second biopsy was
would be resistant to all drugs in that regimen.
For the anthracycline derivatives, Fig. 3 compares LD90 evaluated and found to have a much higher LD90 value (>0.075
values for samples from previously untreated patients with LD90 Mg/ml) than the first sample (0.013 /¿g/ml).
Although c/s-platinum is not usually used to treat breast
values for specimens from patients for whom clinical correla
cancers, response to this drug was also evaluated (Fig. 5). Again,
tions are available. In Fig. 3, the closed symbols indicated
primary breast cancers while the open symbols represent métas there was the characteristic range of sensitivities among speci
tases. One metastatic specimen was to lung (indicated in Fig. 3 mens from previously untreated donors. One patient whose
by **); all others were hypodermal lesions. As defined in Fig. 3 tumor showed a marked sensitivity to c/'s-platinum was treated
A=n•
Fig. 3. Response to anthracyclines of cul
tured human breast cancers. A, doxorubicin;
B, 4' epidoxorubicin; C. idarubicinol. Closed
symbols, primary breast cancers; open symbols,
hypodermal métastases,unless otherwise in
dicated. •.O. no prior therapy; D. retrospec
tive correlation, patient received combination
drug regimen that included drug assayed in
vitro; A, retrospective correlation, patient re
ceived single agent assayed in vitro; *. O. pro
spective correlation, patient received single
agent assayed in vitro; *. **, patient continued
to respond to doxorubicin, but did not reach
partial remission status when drug therapy was
stopped because of cumulative cardiac toxicity;
for **, specimen was a lung biopsy, not hypodermal metastasis.
0.06 - -
AB"t
O.
tj]
B*
°<E
*•*o•0
lOSl
o
±0iOo*!opatients
tt78•
>0.07 :
0.04 - *#*
01
o*A*OO
04
0.02-
«»o
0•
*patients
O1patients
0.00
respondingwith
non-
respondingB°=>
patientsprior
no responding
patientstherapy
patients
patients
with no
prior
therapy
non-
respondingCs
responding
patients
patients
nonwith no
prior
therapy
responding
2945
Downloaded from cancerres.aacrjournals.org on July 12, 2017. © 1990 American Association for Cancer Research.
/.V VITRO CHEMOSENSmVITY
OF HUMAN BREAST CANCER CELLS
ranges were picked empirically in order to distinguish optimally
between clinical responders and nonresponders. As additional
data accumulate, it will be possible to define these values more
precisely. Using these values. Table 2 summarizes the data.
Among 41 patients who did not respond to therapy, 36 had
assays in the resistant range. In contrast, among 22 patients
who responded with at least a partial remission, only 1 patient
had a resistant in vitro assay. These differences were highlysignificant (P< 0.01) by the x2 test. Included in Table 2 are the
>0.07
0.06-
0.04
0.02
results for anthracyclines and vinblastine analyzed separately.
In each case, x2 analysis indicated significance. However, the
< 0.01
0.00
responding
patients
Fig. 4. Response to vinblastine of cultured human breast cancers. All of the
symbols present in this figure are explained in legend to Fig. 3.
DISCUSSION
>5
1 :.
Fig. 5. Response to c/i-platinum of cultured human breast cancer. All of the
symbols present in this figure are explained in legend to Fig. 3.
Table 1 Sensitivity assays: two biopsies from the same individual
No. of cases
Second biopsy more resistant than first
4
Second biopsy similar to first
1
fact that patients who were resistant to multidrug regimens
were compared to patients sensitive to single agents might bias
the results by artificially increasing the differences between
sensitive and resistant. Therefore, Table 2 also summarizes the
data comparing only cases treated with the single agent tested
in vivo. Although the number of cases is smaller, the conclusions
remain the same.
Comments
Taken after treatment failure
Therapy stopped for cardiac
toxicity
No intervening therapy with
tested drug
with this drug and showed a partial response which lasted for
2 months, at which time therapy was changed to carboplatin, a
platinum analogue, for 3 cycles with no change for 4 months.
This was done because of neurotoxicity from a's-platinum. The
biopsy of a second patient who progressed on therapy with cisplatinum and VP 16 was considerably more resistant to cisplatinum in vitro.
Further evidence supporting the clinical relevance of the assay
is summarized in Table 1 which describes cases where two
biopsies were evaluated from the same patient. In all 7 cases
examined, the results support the conclusion that the in vitro
assay accurately reflects the in vivo response. Whenever assays
were performed before and after treatment with a given drug,
the LD90 values always increased if the patient failed on the
therapy. If the patient did not fail, but stopped therapy for
other reasons, or if there was no intervening therapy with the
given drug, the LD90 values remained similar.
In Figs. 3 to 5, a threshold is drawn to separate LD90 values
that correspond to in vitro sensitivity and resistance. These
These studies describe in vitro chemosensitivity assays on
breast cancer cells after short-term culture. Using empirically
defined values, the test accurately predicted resistance for 36 of
41 patients whose tumors did not respond in vivo. The test also
predicted sensitivity with a high degree of accuracy: 21 of 22
patients, 95% of whom responded to the drug treated, had a
sensitive assay. These results suggest that in vitro chemosensi
tivity assays which accurately predict both sensitivity and re
sistance can be obtained with breast cancer cells after shortterm culture even though (a) the cell cycle kinetics differ from
the in vivo state, since all of the cells are cycling after culture;
and (b) there is less heterogeneity because only selected subpopulations grow in this culture system.
The negative correlations may be the result of an inherent
limitation of in vitro drug sensitivity assays, since there may be
pharmacological reasons for a patient's drug response that
cannot be measured in vitro. For example, sufficient drug con
centrations may not reach the tumor cells in vivo because the
tumor has a poor blood supply or the patient may metabolize
and excrete the drug at a rapid rate. If the tumor cells themselves
are very resistant, even good drug delivery will not help; there
fore, in vitro assays may be best for predicting resistance.
Predicting sensitivity may be more difficult because such phar
macological issues may prevent response to the drug in some
patients with very sensitive tumor cells.
Other studies have defined these cut-off values as the drug
levels achievable in vivo (16) rather than empirically defining
the cut-off values for sensitivity and resistance. The latter
approach does not consider the fact that effective drug concen
trations may vary, depending on factors such as drug bound to
culture fluid proteins or plastic substrates, or to drug uptake
being different in cultured cells compared to cells in vivo.
Therefore, we have relied upon the fact that whatever the effects
of culture may be, they will be constant; hence, the relative
LD90 values, regardless of the absolute values, will reflect the
in vivo situation.
The simplest way to obtain clinical correlations is to treat
patients with the single agent assayed in vitro. However, very
few breast cancer patients are treated with single agents. There
fore, patients treated with multidrug regimens were included in
clinical correlations, but only if they did not respond or if they
responded but subsequently failed prior to biopsy. Unlike other
studies (17), patients who responded to combination regimens
2946
Downloaded from cancerres.aacrjournals.org on July 12, 2017. © 1990 American Association for Cancer Research.
/¡VVITRO CHEMOSENSmVITY
OF HUMAN BREAST CANCER CELLS
SummaryNo. 2
DrugAll
correlations'1
clinical correlationsNo.
assaysNo.
of resistant
of patients1/15(7)
sensitive
assaysNo.
of resistant
of patients19/21
resistant
(90)*
Anthracyclines
0/7 (<14)
Vinblastine
17/20(85)
15.9
Total
36/41
(88)5/7
1/22(5)1/15(7)
40.910.2
Single-agent correlations
Anthracyclines
(71)
0/7 (<14)
Vinblastine
13/16(81)
13.4
TotalTable
18/23 (78)of
1/22(5)x224.7
25.1
" Includes specimens from clinically resistant patients treated with either single or multidrug regimens since patients failing a given combination regimen would be
resistant to all drugs in that regimen. All sensitive patients were treated with the single agent assayed in vitro.
* Numbers in parentheses, percentage.
were not included. Our rationale is that patients would fail a
multidrug regimen only if their tumors were resistant to every
drug in the combination. Hence, a resistant assay in vitro would
be a positive correlation while a sensitive assay would be a
negative correlation. If a patient responded to combination
chemotherapy, a resistant assay in vitro for one of the drugs
could not be scored as a negative correlation because the patient
might be responding to other drugs in the combination. There
fore, considering a sensitive in vitro assay for one of the drugs
as a positive correlation overestimates the accuracy of the assay.
Furthermore, there is no proof that the patient actually re
sponded to the drug tested.
For all drugs tested, there was a broad range of sensitivities
among specimens from previously untreated patients, suggest
ing that some cancers are inherently resistant to a given drug.
Included among these specimens were both primary breast
cancers and hypodermal métastases. Since skin métastases
showed the same range of sensitivity as primary breast cancers,
it may be possible to use data obtained with primary breast
cancers to indicate sensitivity of subsequent lesions if there is
no intervening therapy with the same drug. Further studies with
primary lesions and metalases from the same patient are needed
to test this hypothesis.
One advantage of the assay for use in adjuvant trials is that
the success rate for culturing primary breast cancers is greater
than 70% (5). A similar high success rate has been found for
hypodermal métastases.In contrast, although the enriched me
dium grows effusions and some lymph node métastasesin
primary culture, we have found that these cultures tend to grow
more slowly than the primary tumors and usually do not clone
at second passage (7). Hence, further work to improve the
culture system is needed if these advanced tumors also are to
be utilized for predictive assays. It is noteworthly that the agar
clonogenic assay is most successful with effusion métastases
and least effective for primary breast cancers and hypodermal
métastases(18). Hence, the two assays may complement each
other.
The fact that the assay predicts both sensitivity and resistance
for breast cancers is an important first step in using the assay
clinically. The correlations in the study were found in patients
with metastatic breast cancer and the responses seen were all
partial responses. Further studies will be required to determine
whether utilization of this assay can modify morbidity and
survival of patients with metastatic breast cancer and whether
the cost versus benefit ratio is sufficiently high to warrant
routine clinical application.
Even if it proves useful to predict which drugs will be most
effective against metastatic breast cancer, the ability of results
from in vitro assays to change the natural history of advanced
breast cancer, using currently available drugs, will be limited.
Adjuvant therapy has, however, altered the natural history of
breast cancer. Unfortunately, only about 20% of the women
treated with adjuvant therapy show a prolongation of overall
survival (19). It is highly likely that many of those women who
recur after adjuvant therapy do so because of the inherent
resistance of their tumors to the drugs used. Thus, a long-range
goal of these studies is to develop an in vitro assay which can
be used to select the optimal drug regimen when other prog
nostic tests indicate that adjuvant therapy is warranted. The
success of the assay in predicting sensitivity suggests that an
appropriate study question to consider would be to use this
assay to choose between the standard of care alternatives for
adjuvant therapy (i.e., anthracycline versus nonanthracyclinebased regimens).
ACKNOWLEDGMENTS
We thank Dr. S. Ganzina (Farmitalia, Milan, Italy) for the experi
mental drug.
REFERENCES
1. Von Hoff. D. D. In Vitro predictive testing: the sulfonamide era. Int. J. Cell
Cloning, 5: 179-190, 1987.
2. Von Hoff, D. D. Human tumor cloning assays: applications in clinical
oncology and new antineoplastic agent development. Cancer Metastasis Rev.,
7:357-371. 1988.
3. Weisenthal, L. M., and Lippman, M. E. Clonogenic and nonclonogenic in
vitro chemosensitivity assays. Cancer Treatment Rep., 69: 615-633, 1985.
4. Stampfer, M. R., Hallowes, R. C., and Hackett, A. J. Growth of normal
human mammary cells in culture. In Vitro Cell. Dev. Biol., 16: 415-425,
1980.
5. Smith, H. S.. Lan, S.. Ceriani, R., Hackett. A. J., and Stampfer, M. R. donai
proliferation of cultured nonmalignant and malignant human breast epithelia.
Cancer Res., 41: 4637-4643, 1981.
6. Yang, T. C., Stampfer, M. R., and Smith, H. S. Response of cultured normal
human mammary epithelial cells to x-rays. Radiât.Res., 96: 476-485, 1983.
7. Hackett, A. J., and Smith, H. S. Human malignant mammary' epithelial cells
in culture. In: M. M. Webber and L. I. Sekely (eds.). In Vitro Models for
Cancer Research, Vol. 3, pp. 31-49. Boca Raton, FL: CRC Press. 1986.
8. Stampfer, M. R., Hackett, A. J., Hancock, M. C., Leung, J. P., Edgington.
T. S., and Smith, H. S. Growth of human mammary epithelium in culture
and expression of tumor specific properties. Cold Spring Harbor Conf. Cell.
Proliferation, 9: 819-829, 1982.
9. Smith, H. S., Liotta, L. A., Hancock, M. C., and Wolman, S. R. Invasiveness
and ploidy of human mammary carcinomas in short-term culture. Proc. Nati.
Acad. Sci. USA. 82: 1805-1809, 1985.
10. Asaga, T., Suzuki, K.. Takemiya, S., Okamoto, T., Tamura, N., and Umeda,
M. Difference in polynucleation of cultured cells from human mammary
tumors and normal mammary glands on treatment with cytochalasin B.
Gann. 74: 95-99, 1983.
11. Dollbaum, C., Creasey. A., Dairkee, S., Hiller, A. J.. Rudolph, A. R., Lin,
L.. Vitt, C., and Smith, H. S. Specificity of TNF toxicity for human mammary
carcinomas relative to normal mammary epithelium and correlation with
response to doxorubicin. Proc. Nati. Acad. Sci. USA, 85: 4740-4744, 1988.
12. Wolman, S. R.. Smith, H. S., Stampfer, M., and Hackett. A. J. Growth of
diploid cells from breast cancers. Cancer Genet. Cytogen., 16:49-64, 1985.
13. Wolman, S. R. Chromosomes in breast cancer. In: D. Medina, W. Kidwell,
G. Heppner, and E. Anderson (eds.). Cellular and Molecular Biology of
Mammary Cancer, pp. 47-66. New York: Plenum Press, 1987.
14. Smith, H. S., Lippman, M. E., Hiller, A. J., Stampfer, M. R., and Hackett,
A. J. Response to doxorubicin of cultured normal and malignant mammary
epithelial cells. J. Nati. Cancer Inst., 74: 341-348, 1985.
2947
Downloaded from cancerres.aacrjournals.org on July 12, 2017. © 1990 American Association for Cancer Research.
IN VITRO CHEMOSENSITIVITY
OF HI MAN BREAST CANCER CELLS
15. Moore, C. J.. Eldridge. S. R.. Tricomi, W. A., and Gould. M. N. Quanlilation
of benzo(a)pyrene and 7,12-dimethylbenz(a)amhracene
binding to nuclear
macromolecules in human and rat mammary epithelial cells. Cancer Res.,
¿7:2609-2613. 1987.
16. Salmon, S. E., Hamburger. A. W.. Soehnlen, B., Durie. B. G. M., Alberto,
D. S., and Moon. T. E. Quantitation of differential sensitivity of humantumor stem cells to anticancer drugs. N. Engl. J. Med., 298: 1321-1327,
1978.
17. Ajani. J. A., Baker, F. L., Spitzer, G.. Kelly. A.. Brock. W., Tomasovic, B..
Singletary, S. E.. McMurtrey, and Plager, C. Comparison between clinical
response and in vitro drug sensitivity of primary human tumors in the
adhesive tumor cell culture system. J. Clin. Oncol.. 5: 1912-1921. 1987.
18. Dittrich. C.. Sattelhack, E., Jakesz, R., Kolb, R„Holzer, H.. Havelec, L.,
Lenzhofer, R., Steininger, R., Vetterlein, Z. M., Moser. K., and Spitzy, K.
H. Testing of mammary cancer in the human tumor stem cell assay. In: S.
E. Salmon and J. M. Trent (eds.). Human Tumor Cloning, pp. 551-555.
New York: Gruñe& Stratton, 1984.
19. Harris. J. R., Hellman, S., Canellos. G. P., and Fisher, B. Cancer of the
breast. In: V. T. DeVita, S. Hellman, and S. A. Rosenberg (eds.). Cancer
Principles and Practice of Oncology, Ed. 2, pp. 1119-1178. Philadelphia: J.
B. Lippincott Co., 1985.
2948
Downloaded from cancerres.aacrjournals.org on July 12, 2017. © 1990 American Association for Cancer Research.
Preliminary Correlations of Clinical Outcome with in Vitro
Chemosensitivity of Second Passage Human Breast Cancer
Cells
Helene S. Smith, Wainer Zoli, Annalisa Volpi, et al.
Cancer Res 1990;50:2943-2948.
Updated version
E-mail alerts
Reprints and
Subscriptions
Permissions
Access the most recent version of this article at:
http://cancerres.aacrjournals.org/content/50/10/2943
Sign up to receive free email-alerts related to this article or journal.
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Department at [email protected].
To request permission to re-use all or part of this article, contact the AACR Publications
Department at [email protected].
Downloaded from cancerres.aacrjournals.org on July 12, 2017. © 1990 American Association for Cancer Research.