the small heat shock protein hsp27 is not an independent … · vol. 2, 1199-1206, july 1996...
TRANSCRIPT
Vol. 2, 1199-1206, July 1996 Clinical Cancer Research 1199
The Small Heat Shock Protein HSP27 Is Not an Independent
Prognostic Marker in Axillary Lymph Node-negative Breast
Cancer Patients’
Steffi Oesterreich, Susan G. Hilsenbeck,
Daniel R. Ciocca, D Craig Allred,
Gary M. Clark, Gary C. Chamness,
C. Kent Osborne, and Suzanne A. W. Fuqua2
Department of Medicine, Division of Medical Oncology [S. 0..
S. G. H.. G. M. C., G. C. C., C. K. 0., 5. A. W. F.], and Department
of Pathology ID. C. A.]. University of Texas Health Science Center at
San Antonio, San Antonio, Texas 78284-7884. and Laboratonio de
Reproduccion y Lactancia, LARLAC. CC. 855. 5500 Mendoza,
Argentina [D. R. C.]
ABSTRACT
Heat shock protein 27 (hsp27) belongs to the family of
heat shock proteins and is thought to be involved in then-
motolenance, cell proliferation, drug resistance, and chaper-
one processes. The aim of this study was to investigate
whether hsp27 levels are correlated with clinical outcome in
axillary lymph node-negative breast cancer patients. We
describe a Western blot study measuring hsp27 levels in 425
patients and an immunohistochemistry (IHC) study analyz-
ing 788 patients. Results obtained by both methods were
concordant. Univaniate survival analysis was performed
considering hsp27 either as an optimally dichotomized van-
able or as a continuous variable. Additional data include age
at biopsy, tumor size, estrogen receptor (ER) and progester-
one receptor status, tumor ploidy and percentage of cells in
S phase, and adjuvant therapy. hsp27 levels correlated pos-
itively with ER status (P = 0.0001 in Western blot and IHC
study), progesterone receptor status (P = 0.0001 in Western
blot and IHC study), and aneuploidy (Western blot study, P
= 0.0012; IHC study, P 0.0004) but not with tumor size
(Western blot study, P 0.69; IHC, P 0.53) or S phase
(Western blot study, P = 0.19; IHC study, P 0.38). Over-
all, there was no relationship between hsp27 expression and
disease-free survival (Western blot study, P 0.70/0.54;
IHC, P = 0.47/0.30) or overall survival (Western blot study,
P = 0.16/0.15; IHC, P 0.46/0.78). Exploratory subset
analyses defined by ER status and use of adjuvant treatment
indicated that in ER+/untreated patients, high hsp27 levels
correlated modestly with shorter disease-free survival
Received I 2/2 1/95; revised 4/5/96: accepted 4/1 1/96.
� Supported by Specialized Programs on Research Excellence Grants
P50 CA58183. P30 CA54174. and P01 CA30195 and in part by
CONICET/Governrnent of Mendoza (Argentina).2 To whom requests for reprints should be addressed. at Department of
Medicine. Division of Medical Oncology, University of Texas Health
Science Center at San Antonio. 7703 Floyd Curl Drive, San Antonio,
TX 78284-7884. Phone: (210) 567-4787: Fax: (210) 567-6687.
(Western blot, P 0.04/0.04; IHC, P 0.1 1/0.03). hsp27 is
not a useful prognostic marker for the clinic in axillary
lymph node-negative patients. However, the finding of mod-
est prognostic value of hsp27 in the subgroup of ER+/
untreated patients raises new questions about the biological
function of hsp27 in breast cancer.
INTRODUCTION
hsp273 belongs to the family of heat shock proteins (re-
viewed in Ref. I). A variety of different functions have been
described for hsp27, including a role in thermotolerance (2), cell
proliferation, drug resistance (3). adtin polymerization (4, 5),
and chaperone processes (6, 7). In breast cancer cells, hsp27 was
originally identified as an estrogen-responsive gene (8. 9). The
expression of hsp27 is very high in normal estrogen-sensitive
target organs, such as breast, uterus, placenta. fallopian tubes,
vagina, and skin ( 10), but varies widely in breast cancer ( 1 1).
hsp27 has been studied as a prognostic marker in several
tumor types including prostate and bladder cancer, malignant
fibrous histiocytoma, and neuroblastoma (12-14). Of seven
reported studies in breast cancer using different assay methods
and patient populations (Table 1 ), three reported that high ex-
pression of hsp27 is a marker for worse outcome (15-17), two
showed no correlation with clinical outcome (18, 19), and two
studies indicated an association with improved outcome (20,
21 ). In view of these partially contradictory results, we decided
to repeal our original pilot study (15) using two different nieth-
ods in a larger, independent group of patients. We focused on a
node-negative population, most of whom did not resume adju-
vant therapy. Since most axillary lymph node-positive patients
will receive adjuvant therapy independent of the status of other
tumor markers, the challenge is to identify a subset of axillary
lymph node-negative patients who can forego adjuvant therapy.
In this study, we describe a Western blot analysis meas-
uning hsp27 levels in 425 axillary lymph node-negative patients
and an IHC study analyzing 788 node-negative patients. Ele-
vated hsp27 levels correlate with a shorter DFS in ER+/un-
treated (surgical intervention, but no adjuvant treatment) pa-
tients but are not associated with DFS survival or overall
survival in the whole patient population. Therefore, hsp27 is not
a useful, independent prognostic marker in axillary lymph node-
negative patients.
3 The abbreviations used are: hsp27. heat shock protein 27: ER. estrogen
receptor: PgR. progesterone receptor: IHC. irnrnunohistochernistry:
DFS. disease-free survival; OS. overall survival.
Research. on October 29, 2020. © 1996 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1200 hsp27 Is Not an Independent Marker in Node-negative Breast Cancer
Table 1 Prognostic significance of hsp27 in breast cancer patients
Association with
Investigators (Ref.) ii Follow up (rno) Nodal status Assay DFS OS
Tandon et al. (15) 198
161
81
68
Neg.”
Pos.
Western blot
Western blot NS
NS
NSThor ci’ al. (16) 291
(total)
92 Neg. IHC NS NS
1-3 Pos. nodes�4 Pos. nodes
IHC
IHC
(a = 66)
NS
NS
NS
Love and King (17) 361 74 Pos. vs. neg. IHC tin node neg.
patients: ii =
NS
Darrnstrup et al.(18) 103 ? Mixed IHC NS NS
Tetu ci’ al. (19) 890 78 Pos. IHC NS NSHurlirnan et al. (20) 196 81 Mixed IHC I “
Seymour et al. (21) 51 ? Pos. IHC ? I
(1 Neg.. negative: Pos., positive: NS, nonsignificant; � , decreased: I � increased: ?. unknown.
I, The survival was measured at 5 years after diagnosis: borderline significance was reached in ER-negative tumors.
PATIENTS AND METHODS
Patient Population
The patients in this study were identified from a large
database of patients who had hormone receptor assays per-
formed in our laboratory and comprise a total of 903 axillary
node-negative breast cancer patients diagnosed and resected
between 1974 and 1988. Selection criteria included presentation
with primary breast cancer treated with mastectomy or lumpec-
tomy, with or without postoperative radiation therapy, and ab-
sence of involvement of axillary lymph nodes or any distant
metastasis. Available clinical data included age at biopsy (diag-
nosis), tumor size, ER and PgR status, and tumor ploidy and
percentage of cells in S phase, as determined by flow cytometry
(22, 23), and adjuvant therapy. Patients were followed for dis-
ease recurrence and survival as described previously (24). The
Western blot study was done first and includes 425 cases. Three
hundred ten of these samples were later analyzed by IHC;
however, for some cases, no paraffin sections were available.
The IHC study was extended by using an additional 478 cases,
analyzing a total of 788 cases by IHC. The median follow-up
from diagnosis for unrelapsed cases was 7 1 and 53 months for
the Western blot and IHC studies, respectively.
hsp27 Analyses
Western Blot. Breast tumor specimens were frozen in
liquid nitrogen immediately after excision and stored at -70#{176}C.
Breast tumor tissues were fractured in the frozen stale, and an
average of 50 rng of pulverized tissue was used from each case.
Extracts from pulverized frozen tumors were prepared, and 100
�ig of protein were subjected to Western blot analysis as de-
scnibed previously (25, 26). The primary antibody was a mono-
clonal antibody directed against human hsp27, which was de-
veloped in our laboratory (27). It was used at a dilution of
1 : 1000 in blocking buffer. After washing, the blots were incu-
baled for 2 h with ‘251-labeled anti-mouse antibody (100,000
cpm/ml; Amersham Corp. Arlington Heights, IL). Each poly-
acrylamide gel contained an internal protein standard from
human MCF-7 breast tumor cells. Evidence of hsp27 protein
degradation, such as multiple faster migrating bands, could not
be detected (data not shown). The hsp27 bands were quantitated
by densilomelnic scanning, as described elsewhere (15). The
levels of hsp27 were calculated in arbitrary units per 100 �ig of
protein by the ratio of the integrated densitometry signal in the
tumor samples relative to the MCF-7 standard on each gel.
IHC. IHC was performed on frozen, partially pulverized,
small breast carcinoma samples prepared for paraffin sections as
reported elsewhere (28). Briefly, frozen pulverized tissue was
rehydraled at room temperature in phosphate buffer and fixed
for 4 h in 10% formalin; then the particles were concentrated in
warm agar. After embedding in paraffin wax, tissue sections (5
p.m thick) were used to evaluate hsp27 expression by IHC (29),
using a commercial avidin-biotin-peroxidase method (Vector
Laboratories, Burlingame, CA). A quality control examination
was performed on each sample (stained with H&E) to exclude
those samples with an inadequate number of malignant cells,
unusual cellular damage, or necrosis. Diaminobenzidine was
used for developing color; methyl green was used for counter-
staining. Each run included positive control sections from
MCF-7 cells, and negative controls were run by omitting the
primary antibody. Immunostained slides were examined by light
microscopy, and a positive signal was defined as the presence of
distinct staining in the tumor cell. The signal was semiquanti-
latively scored estimating the proportion of positively staining
tumor cells on the entire slide (0, none; 1 , less than one-
hundredth; 2, one-hundredth to one-tenth; 3, one-tenth to one-
third; 4, one-third to two-thirds; 5, greater than Iwo-thirds) and
the average intensity of staining (0, no staining; I , weak slain-
ing; 2, moderate staining; 3, strong staining). Intensity and
proportion scores were added to obtain the final total score. This
scoring system has been successfully applied to evaluate steroid
receptors and other prognostic factors in previous studies (30,
31).
Statistical Analyses
Statistical analyses focused on Iwo issues: (a) the associ-
ation between hsp27 measurements and other clinical variables;
and (b) the prognostic value of hsp27. Associations between
hsp27 values and other clinical variables were first examined
Research. on October 29, 2020. © 1996 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Clinical Cancer Research 1201
Table 2 Summary of patie nt characte nistics and associations with hsp27 level s measured by West em blot or IHC
Western blot study IHC study
hsp27 association hsp27 association
Groups F,” % -Ysp/, P ii” % -y,� P
All Cases 424 100 788 100
Age �50 yrs
>50 yrs
1 19
305
28.1 0.08 0.09
71.9
191
597
24.2
75.2
0.09 0.60
ER Negative
Positive
99
325
23.3 0.34 0.0001
76.7
159
627
20.2
79.8
0.18 0.0001
PgR NegativePositive
186219
45.9 0.19 0.000154.1
310
457
40.4
59.6
0.15 0.0001
Tumor size �2 cm
>2 cm
156
268
36.8 0.02 0.69
63.2
410
355
53.6
46.6
0.02 0.53
Ploidy Diploid
Aneuploid
141
268
34.5 0.0012’
65.5
307
474
39.3
60.7
0.0004’
% S Phase <6.7%�6.7%
201
170
54.2 0.07 0.19
45.8
404
282
58.9
41.1
0.03 0.38
Endocrine therapy No
Yes
317
92
77.5 0.94’
22.5
653
89
88
12
0.63’
Chemotherapy No
Yes
357
52
87.3 0.38’
12.7
677
68
90.9
9.1
0.02’
“ Not all variables were measurable in all cases.
I’ � Spearman’s rank correlation, using actual values.C Groups compared by two-sample Wilcoxon rank sum test.
using Spearman rank correlations for continuous clinical van-
ables (tumor size, ER, PgR, and percentage in S phase) and then
using Wilcoxon rank sum statistics to compare distributions of
hsp27 for categorical variables (ploidy, endocrine therapy, and
chemotherapy). Two prognostic outcomes were examined: DFS,
defined as the time from diagnosis to first recurrence, or last
contact; and OS, defined as the time from diagnosis to death
(from any cause), or last contact. Survival times (DFS or OS)
longer than 10 years were truncated at 1 0 years and censored to
reduce the influence of a small number of cases with very long
follow-up. In prognostic analyses, clinical variables were di-
chotomized as follows: age at diagnosis, �50 versus age >50
years; tumor size, �2 cm versus >2 cm; ER, <3 versus
fmol/rng protein; PgR, <5 versus �5 frnollmg protein; diploidy
(DNA index, 1 .0) versus aneuploidy; percentage of S-phase
fraction, s6.7 versus >6.7% (24); no adjuvant endocrine ther-
apy versus endocrine therapy; and no adjuvant chemotherapy
versus chemotherapy. Univaniale survival curves were estimated
using the method of Kaplan and Meier and compared using the
log-rank test (32). Culpoinl analysis with P adjustment (33) was
used to dichotomize hsp27. In cutpoint analysis and related
methods that use repealed testing of the data set to select an
optimal cutoff value, type I errors (false positives) increase
with the number of culpoints examined (34), and P adjustment
(Padjusted) is necessary to correct for this. Cox proportional
hazards regression (32) was used to examine the prognostic
value of hsp27 as either a dichotomized or continuous
log-transformed variable.
RESULTS
Association of hsp27 Expression with Other Prognostic
Factors. General patient and tumor characteristics are shown
in Table 2. Four hundred twenty-five samples were examined by
Western blot analysis, and 788 samples were analyzed by IHC
for hsp27 expression; representative experiments of the Western
Blot analysis, as well as the IHC assay, are shown in Fig. 1 . The
Western Blot shows different hsp27 levels in 10 breast tumors
and the internal protein standard from MCF-7 cells. The IHC
reveals a strong cytoplasmic signal in breast cancer cells.
Complete data on all the measured prognostic variables
were available on 346 and 613 cases, respectively. The majority
of these node-negative patients did not receive endocrine ther-
apy (77.5 and 88% in the Western blot and IHC studies, respec-
tively) or chemotherapy (87.3 and 90.9%, respectively). In the
Western blot study, hsp27 expression correlated with the ex-
pression of the hormone receptors ER (P = 0.0001) and PgR
(P = 0.0001) but not with tumor size (P 0.69) or percentage
in S phase (P = 0. 19). Aneuploid tumors tended to have higher
hsp27 values (P = 0.012). Similar results were obtained for the
tumor specimens analyzed by IHC. In general, results obtained
by both methods, Western blot and IHC, were concordant (�y�
= 0.505; P = 0.0001). Again, hsp27 expression as obtained
with IHC correlated with ER (P = 0.0001), PgR (P = 0.0001),
and aneuploidy (P = 0.0004) but not with tumor size (P = 0.53)
or S phase (0.38).
Clinical Outcome and Expression of hsp27. Univariate
survival analyses are summarized in Table 3. In the Western blot
datasel, small tumor size and a low percentage in S phase were
significantly related to longer DFS and OS, whereas age over 50
was related to longer OS. In the larger IHC dataset, which had
shorter follow up (53 months versus 71 months) and smaller
tumors, we found that age over 50 years, small tumor size, low
percentage in S phase, and diploidy were all associated with a
longer DFS. Only tumor size and S phase were related to OS.
In the Western blot dataset, there was no detectable rela-
tionship between hsp27 expression and either DFS or OS (Table
3), regardless of whether hsp27 was considered as an optimally
dichotomized variable (P�,djusted 0.70 or 0. 16 for DFS and OS,
Research. on October 29, 2020. © 1996 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1202 hsp27 Is Not an Independent Marker in Node-negative Breast Cancer
Tumors MCF-7
1 2 3 4 5 6 7 8 910 �
� :lj�� ,�* ;�;:�#{248}�
Fig. 1 Western Blot (left) and IHC (right) using hsp27-specific antibodies. The Western Blot shows the hsp27 level in 10 breast tumors and the
internal protein standard (100, 50, and 25 �g) from MCF-7 cells.
Table 3 Univari ale su rvival analysis of all p atients an d all variables
DFS#{176} 0 5�
Western Western
blot study IHC study blot study IHC study
HRb p HRb p HRb p HR” p
Age 1.12 0.6 1.33 0.12 0.61 0.04 0.83 0.29ER 0.71 0.11 0.82 0.27 0.89 0.61 0.85 0.37PgR 0.92 0.67 0.75 0.06 0.79 0.26 0.79 0.14
Tumor size 1.95 0.003 2.56 0.0001 1.53 0.05 1.83 0.0004% S Phase 1 .55 0.03 1 .55 0.007 1 .56 0.04 1 .46 0.03Ploidy 1.24 0.31 1.78 0.0006 1.31 0.23 1.27 0.17hsp27c 0.70 0.47 0.16 0.46hsp27” 1.07 0.56 1.25 0.29 1.18 0.16 1.07 0.77
a Log-rank results.
b HR. hazard ratios from Cox regression with a single covariate. For each variable, the reference group is the first group listed for that variable
in Table 2.C p� for hsp27 are from cutpoint analyses and are corrected for multiple comparisons. Hazard ratios could not be determined.
d Ps for hsp27 are from Cox regression where log(hsp27 + 1) was the only explanatory variable. Hazard ratios represent the effect of a 10-fold
increase in hsp27 score.
respectively) or as a log-transformed continuous variable (P =
0.54 and 0. 15, respectively). Given the sample size of this
dataset, the rate of censoring, and the modest correlation of
hsp27 with other variables of known prognostic value, we
estimate that a moderate independent prognostic value (a hazard
ratio of 2.0 or more) for hsp27 would have been detectable with
more than 85% power at the 5% level of significance. Likewise,
in the IHC dataset, there was no detectable relationship between
hsp27 and either DFS or OS, regardless of whether hsp27 was
considered as an optimally dichotomized variable (�adjusted
0.47 and 0.46, respectively) or as a continuous log-transformed
variable (P = 0.30 and 0.78, respectively). Again, given the
sample size, rate of censoring, and the correlation of hsp27 with
other variables, a moderate independent prognostic value for
hsp27 would have been detectable with more than 95% power at
the 5% level of significance. Survival curves (shown in Fig. 2)
clearly confirm these findings. The curves for low versus high
hsp27 expression are nearly overlapping for DFS as well as OS.
Umvariate Survival Analysis in ER+IUntreated Pa-tients. Since our datasets are fairly large, we decided to per-
form subgroup analysis of the tumors. Subgroups were defined
by ER status and use of adjuvant treatment (ER+/treated,
ER+/untreated, ER-/treated, and ER-/untreated). These
groups are of interest for several reasons: (a) it was recently
shown that hsp27 can be induced by different cylotoxic drugs
commonly used in the treatment of cancer patients, including
daunomycin, doxorubicin, 3 ‘ -fluorodeoxythymidine, and
vincnisline (35). Furthermore, hsp27 itself may be involved in
drug resistance, e.g. , overexpression of hsp27 results in de-
creased sensitivity to doxorubicin in breast cancer cells (3).
(Unfortunately, direct influence of hsp27 expression on the
effect of doxorubicin treatment in patients could not be
evaluated in this study, since the few treated patients received
various nonstandardized treatments); and (b) hsp27 is an
estrogen-regulated gene, and its expression correlates with
ER expression in a range of estrogen-sensitive cells, includ-
ing primary breast cancer (36). We repeated the univaniale
survival analyses in each of the four subgroups (Table 4). In
ER+/untreated patients, but not in the other subsets, a weak
correlation between hsp27 clinical outcome was found (Table
4). In the Western blot datasel, hsp27 expressed levels once
considered either as an optimally dichotomized variable
Research. on October 29, 2020. © 1996 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
I� hsp27 (n=�3)
p=O.58 High h8p27 (n=211)
A0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
B0.9
0.8
0.7
0.6
0.5
High hsp27 (n’=310)
Low hsp27 (n=478)
C1.o0.9
0.8
0.7
0.6r)� 0.5
0.4
0.3
0.2
0.1
0.0
D1.o0.9
0.8
0.7
0.6
C,,0 0.5
0.4
0.3
02
0.1
0 24 48 72 96 120
Months
� � hsp27 (n=�3)
p=O.22 High hsp27 (n=211)
0 24 48 72 96 120
Months
� L� hsp27 (n=478)
High hsp27 (n=310)
p=0.13U)0
_________________________________ 0.0
0 24 48 72 96 120
Months
Fig. 2 Survival curves (OS and DFS) of all patients as a function of hsp27 expression measured by Western blot (A and B) or IHC (C and D). Thegraphs are based on dichotomization of hsp27 at median values.
Clinical Cancer Research 1203
patients are shown in Fig. 3.
0.4
0.3
0.2
0.1
0.0
0 24 48 72 96 120
Months
p=O.44
Table 4 Univaniate survival analysis in ER+/untreated patients considering hsp27 either as an optimally dichotomized or as a
continuous variable
DFS” OS”
Western
blot study IHC study
Western
blot study IHC study
HR” P HR’ P HR” P HR” p
hsp27’
hsp27” 1.37
0.04
0.04 1.32
0.1 1
0.03 1.17
0.52
0.30 1.17
0.15
0.22
“ Log-rank results.
I’ HR. hazard ratios.‘� PS for hsp27 are from cutpoinl analyses and are corrected for multiple comparisons. Hazard ratios could not be determined.‘I p� for hsp27 are from Cox regression where log(hsp27 + 1 ) was the only explanatory variable. Hazard ratios represent the effect of a 2.7-fold
increase in hsp27 score.
(13�tdjusted = 0.04) or as a continuous variable (using log-transform
P = 0.04) was modestly associated with DFS but was not associ-
ated with OS. Analogous results were obtained with the IHC
dataset; hsp27 expression considered as a continuous variable (us-
ing a log-transform) was associated with DFS (P = 0.03) but not
with OS (P = 0.22). The curves for DFS and OS of ER+/untreated
DISCUSSION
The heat shock or stress-response proteins are a family of
intracellular proteins with many functions, one of which is to
protect cells from a variety of external toxic stimuli (reviewed in
Ref. 1). We first identified hsp27 in breast cancer cells as an
estrogen-regulated protein that is overexpressed in some human
breast cancer specimens (8, 9). We hypothesized that hsp27,
Research. on October 29, 2020. © 1996 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
p=0.03
I� hsp27
High hsp27 (n=105)
p-0.03
Low hsp27 (n=106)
High hsp27 (n=105)
A i.o
0.9
0.8
0.7
0.6
� 0.5
0.4
0.3
0.2
0.1
0.0
B
0.7
0.6
� 0.5
0.4
0.3
0.2
0.1
0.0
0 24 48 72 96 120
Months
C1.o0.9
0.8
0.7
0.6U)� 0.5
0.4
0.3
0.2
0.1
0.0
D i.o
0.9
0.8
0.7
0.6
C,,0 0.5
0.4
0.3
02
0.1
0.0
0 24 48 72
Low hsp27 (n=269)
High hsp27 (n=109)
Low hsp27 (n=269)
High hsp27 (n=209)
96 120
Months
p=O.46p��O.04
0 24 48 72 96 120
Months
Fig. 3 Survival curves (OS and DFS) of ER+/untreated patients as a function of hsp27 expression measured by Western blot (A and B) or IHC (Cand D). The graphs are based on dichotomization of hsp27 at median values.
0 24 48 72
Months
96 120
1204 hsp27 Is Not an Independent Marker in Node-negative Breast Cancer
when present in abundance, would protect tumor cells from a
toxic environment, thereby providing a survival advantage that
would confer a more aggressive clinical course and worse
prognosis for the patient. In a small pilot study using a Western
blot assay, hsp27 overexpression did correlate with a worse DFS
in a group of patients with negative axillary lymph nodes,
despite its association with positive ER, a favorable prognostic
marker (15).
In the present study using a much larger independent
patient sample and two different assay methods, we failed to
confirm our original pilot study. There was agreement between
the Western blot and IHC assays, and with both assays, hsp27
expression correlated, as expected, with ER and PgR positivity.
Neither assay, however, correlated with clinical outcome. Based
on our power analysis, we are confident that we would have
found any clinically significant overall correlation between
hsp27 and outcome in this patient population. This result is in
agreement with several other published studies, although those
typically included patients with positive axillary lymph nodes
who received a variety of adjuvant treatments (16-2 1).
We also divided the cases into four subgroups based on ER
status and adjuvant treatment decision. Analyzing the Western
blot dataset (n = 21 1), hsp27 was correlated with shorter DFS
in ER+/untreated patients. Results from the IHC dataset (n
478) were similar, but subgroup analyses can be problematic,
and the results should be received with caution. However, it is
encouraging that results from two assay methods in only par-
lially overlapping groups of cases were concordant. It has been
known for a long time that hsp27 is estrogen inducible and,
therefore, may be a downstream target in the ER-mediated
growth pathway. Further data also suggest that hsp27 is in-
volved in proliferation processes in breast cancer cells (3, 37).
Also, in our ER-positive subgroup. we found a small but sig-
nificant correlation between hsp27 expression and percentage in
S-phase fraction (data not shown). Adjusting for percentage in S
phase in the prognostic analysis reduces but does not eliminate
the effect of hsp27. Therefore, it is tempting to speculate that
estrogen-induced hsp27 expression could conceivably enhance
the rate of tumor growth and enhance the risk of recurrence in
a patient. The correlation of high hsp27 with shorter DFS but not
OS has been reported previously by Thor et a!. (16) and Love
and King ( I 7). Our hypothesis is that after recurrence, hsp27 is
no longer significant for tumor progression, and its correlation
with and effect on proliferation processes is now overridden by
other mechanisms, such as metastasis or other more aggressive
tumor behaviors. At the moment, there are no data suggesting
that hsp27 is associated with increased tumor metastasis or cell
motility. However, hsp27 is an actin-binding protein and is
associated with actin polymerization (5), and adlin polyrneriza-
lion-associated proteins have been shown to be involved in cell
Research. on October 29, 2020. © 1996 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
8. Moretti-Rojas. I.. Fuqua. S. A. W.. Montgomery. R. A. I.. and
McGuire. W. L. A eDNA for the estradiol-regulated 24K protein:
Clinical Cancer Research 1205
motility (38, 39). We are currently analyzing the melastatic
behavior of breast cancer cells transfected with hsp27 eDNA.
Considering the relatively small sample size of treated
patients, it was difficult to speculate on any influence of drug
sensitivity and/or hsp27 induction by chemotherapy on our
finding in untreated versus treated patients. The direct influence
of adjuvant therapy on the outcome of patients overexpressing
hsp27 could not be evaluated in this study. since the number of
treated lymph node-negative patients with long-term follow up
is very low, and furthermore, the treated patients do not repre-
sent a homogeneous group receiving standardized treatments.
Although the apparent effect on DFS is too weak to be of
practical clinical value, our findings may shed light on the
biological role of hsp27 in breast cancer. These findings may
also help to explain the results published from different groups.
For example. our present studies do not confirm our previous
finding that hsp27 has prognostic power as an independent
marker in lymph node-negative breast cancer in general (15).
However, the patient group analyzed previously differs some-
what from the one analyzed here. In particular, the previous
dataset was smaller but had a greater percentage of untreated
patients with larger tumors and higher S-phase values. The
failure to confirm the prognostic significance of hsp27 in the
overall analyses emphasizes the need to perform validation
studies when characterizing new putative biomarkers.
The data presented here does indicate that hsp27 is not an
independent marker for breast cancer prognosis in axillary lymph
node-negative patients and is, therefore, not a useful tumor marker.
However, it has modest prognostic value in the subgroup of ER+/
untreated patients, which raises new questions about the biological
function(s) of hsp27 in breast cancer cells.
ACKNOWLEDGMENTS
We thank Adrian Lee for critically reading the manuscript and for
many helpful discussions.
REFERENCES
1. Ciocca. D. R.. Oesterreich. S.. Charnness G. C.. McGuire. W. L.. and
Fuqua. S. A. W. Biological and clinical implications of heat shock
protein 27.000 (hsp27): a review. J. NatI. Cancer Inst.. 85: 1558-1570.
1993.
2. Huot, J., Roy, G., Larnbert, H., Chretien. P., and Landry, J. Increasedsurvival after treatment with anticancer agents of Chinese hamster cells
expressing the human Mr 27,()00 heat shock protein. Cancer Res.. 5/.
245-252, 1991.
3. Oesterreich, S.. Weng. C-N.. Qiu, M.. Hilsenbeck, S.. Osborne.
C. K.. and Fuqua. S. A. W. The small heat shock protein hsp27 iscorrelated with growth and drug resistance in htirnan breast cancer celllines. Cancer Res.. 53: 4443-4448. 1993.
4. Miron. T.. Wilchek, M.. and Geiger. B. Characterization of an
inhibitor of adtin polymerization in vinculin-nich fraction of turkey
gizzard smooth muscle. Eur. J. Biochern., 178: 543-553. 1988.
5. Miron, T., Vancornpernolle. K., and Vandekerckhove, J. A 25 kDa
inhibitor of adtin polymerization is a low molecular mass heat shock
protein. J. Cell. Biol., /14. 255-261, 1991.
6. Horwitz. J. Alpha-crystallin can function as a molecular chaperone.
Proc. Nat. Acad. Sci. USA, 89: 10449-10453. 1992.
7. Jakob, U., Gaestel. M., Engel, K., and Buchner, J. Small heat shockproteins are molecular chaperones. J. Biol. Chem.. 268: 1517-1520.
1993.
control of rnRNA levels in MCF-7 cells. Breast Cancer Res. Treat.. 1/:
155-163. 1988.
9. Fuqua, S. A. W.. Blurn-Salingaros. M.. and McGuire. W. L. Induc-
lion of the estrogen regulated “24K” protein by heat shock. Cancer Res..
49: 4126-4129, 1989.
10. Ciocca, D. R.. Adams, D. P.. Edwards. R. J.. Bjercke. R. J.. and
McGuire, W. L. Distribution of an estrogen-regulated protein with a
molecular weight of 24.0(X) in normal and malignant human tissues and
cells. Cancer Res., 43: 1204-1210. 1988.
I I . Adams. D. J.. and McGuire. W. L. Quantitative enzyme-linked
irnniunosorbent assay for the estrogen-regulated M, 24.0(X) protein in
human breast tumors: correlations with estrogen and progesterone re-
ceptors. Cancer Res.. 45: 2445-2449. 1985.
12. Storm. F. K.. Mahvi, D. M., and Gilchrist. K. Hsp-27 has no
diagnostic or prognostic significance in prostate or bladder cancers.
Urology. 42: 379-382. 1993.
I 3. Tetu, B.. Lacasse, B., Bouchard. H-L.. Lagace. R.. Huot. J.. and
Landry. J. Prognostic influence of hsp-27 expression in malignant
fibrous histiocytorna: a clinicopathological and inirnunohistochernical
study. Cancer Res.. 52: 2325-2328. 1992.
14. Ungar. D. R.. Hailat. N.. Strahler. J. R.. Kuick. R. D.. Brodeur.
G. M.. Seeger. R. C.. Reynolds. C. P.. and Hanash. S. M. Hsp27expression in neuroblastorna: correlation with disease stage. J. NatI.
Cancer Inst.. 86: 780-784. 1994.
15. Tandon. A. K.. Clark. G. M.. Charnness, G. C., Fuqua. S. A. W..
Welch. W. J.. Riehl, R. M.. and McGuire. W. L. Heat shock stress
response proteins: biological and clinical significance in breast cancer.
Proc. Am. Soc. Clin. Oncol.. 9: 84. 1990.
16. Thor, A.. Benz, C.. Moore. D. I.. Goldman. E.. Edgerton. S..
Landry, J.. Schwartz. L.. Mayall, B., Hickey. E.. and Weber. L. A. Stress
response protein (srp-27) determination in primary human breast carci-
nomas: clinical. histologic. and prognostic correlations. J. NatI. Cancer
Inst.. 83: 170-178. 1991.
17. Love. S., and King. R. J. B. A 27 kDa heat shock protein that has
anomalous prognostic powers in early and advanced breast cancer. Br. J.
Cancer. 69.� 743-748. 1994.
18. Darmstrup. L.. Andersen. J.. Kufe, D. W.. Hayes. D. F.. and
Skovgaard Poulsen. H. Irnrnunocytochernical determination of the es-
trogen-regulated proteins Mr 24.000. Mr 52.000 and DF3 breast cancer
associated antigen: clinical value in advanced breast cancer and corre-
lation with estrogen receptor. Ann. Oncol.. 3. 71-77. 992.
19. Tetu, B.. Bnisson. J.. Landry. J.. and Huot. J. Prognostic significance
of heat-shock protein-27 in node-positive breast carcinorna: an irnrnun-
histochemical study. Breast Cancer Res. Treat.. 36: 93-97, 1995.
20. Hurlirnann. J.. Gebhard S.. and Gornez. F. Oestrogen receptor.
progesterone receptor. p52. ERDS. H5P27 and cathepsin D in invasive
ductal breast cancer. Histopathology. 23: 239-248. 1993.
2 1 . Seymour. L.. Bezwoda. W. R.. and Meyer. K. Tumor factors pre-
dieting for prognosis in rnetastatic breast cancer: the presence of p24
predicts for response to treatment and duration of survival. Cancer
(Phila.). 66: 2390-2394. 1990.
22. DressIer. I.. (‘i.. Seamer. L. C.. Owens. M. A. Clark. C. M.. and
McGuire. W. L. DNA flow cytornetry and prognostic factors in 1331
frozen breast cancer specimens. Cancer (Phila.). 61: 420-427, 1988.
23. Wenger. C. R.. Beardslee, S., Owens. M. A., Pounds. G.. Oldaker.
T., Vendelay. P.. Pandian. M. R.. Harrington. D.. Clark. G. M.. and
McGuire, W. L. DNA ploidy, S-phase and steroid receptors in more than
127,000 breast cancer patients. Breast Cancer Res. Treat.. 28: 9 -20.
1993.
24. Clark. G. M.. DressIer, L. G.. Owens. M. A.. Pounds. G.. Oldaker.
T., and McGuire. W. L. Prediction of relapse or survival in patients with
node-negative breast cancer by DNA flow cytornetry. N. EngI. J. Med..320: 627-633. 1989.
25. Tandon, A. K.. Clark. G. M.. Chamness. G. C.. Ullrich. A., and
McGuire, W. L. Her-2/neu oncogene protein and prognosis in breast
cancer. J. Clin. Oncol.. 7: 1 120-1 128, 1989.
Research. on October 29, 2020. © 1996 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1206 hsp27 Is Not an Independent Marker in Node-negative Breast Cancer
26. Tandon, A. K., Clark, G. M., Chamness, G. C., and McGuire, W. L.
Cathepsin D and prognosis in breast cancer. N. Engl. J. Med., 322:
297-302, 1990.
27. Adams, D. J., Hajj, H., and Edwards, D. P. Detection of a Mr 24,000
estrogen-regulated protein in human breast cancer by monoclonal anti-
bodies. Cancer Res., 43: 4297-4301, 1983.
28. Allred, D. C., Tandon, A. K., Clark, G. M., Brown, R. W.,
Abbondanzo, S. L., and MeGuire, W. L. Comprehensive evaluation of
prognostic factors by immunocytochemistry on extremely small samples
(50 mg) of “pulverized” breast carcinomas. Breast Cancer Res. Treat.,
16: 182, 1990.
29. Ciocca, D. R., and Bjercke, R. J. Inmiunocytochemical techniques
using monoclonal antibodies. Methods Enzymol., 121: 562-579, 1986.
30. Molina, R., Ciocca, D. R., Tandon, A. K., Allred, D. C., Clark,
G. M., Chamness, G. C., Gullick, W. J., and McGuire, W. L. Expressionof Her-2/neu oncoprotein in human breast cancer: a comparison of
immunohistochemical and Western blot techniques. Anticancer Res.,
12: 1965-1972, 1992.
31. Elledge, R. M., Ciocca, D. R., Langone, G., and McGuire, W. L.
Estrogen receptor, progesterone receptor, and Her-2/neu protein inbreast cancers from pregnant patients. Cancer (Phila.), 71: 2499-2506,
1993.
32. Cox, D. R., and Oakes, D. Analysis of survival data. New York:
Chapman and Hall, 1984.
33. Hilsenbeck, S. G., and Clark, G. M. A practical method of P-value
adjustment. Stat. Med., 15: 102-1 12, 1996.
34. Hilsenbeck, S. G., Clark, G. M., and McGuire, W. L. Why do somany prognostic factors fail to pan out? Breast Cancer Res. Treat., 22:
197-206, 1992.
35. Bielka, H., Hoinkis, G., Oesterreich, S., Stahl, H-J., and
Benndorf, R. Induction of the small stress protein, hsp25, in Ehrlich
ascites carcinoma cells by anticancer drugs. FEBS Lefts., 343: 165-
167, 1994.
36. Dunn, D. K., Whelan, R. D., Hill, B., and King, R. J. Relationship
of hsp27 and oestrogen receptor in hormone sensitive and insensitive
cell lines. J. Steroid Biochern. Mol. Biol., 46: 469-479, 1993.
37. Mairesse, N., Horman. S., Mosselmans, R., and Galand, P. Anti-
sense inhibition of hsp27 production affects growth rate and cytoskel-
eton organization in MCF-7 cells. J. Cell. Biochern., 19B: 100, 1995.
38. Condeelis, J. Life at the leading edge: the formations of cell pro-
trusions. Ann. Rev. Cell Biol., 9: 41 1-444, 1993.
39. Stossel, T. P. On the crawling of animal cells. Science (Washington
DC), 260: 1086-1094, 1993.
Research. on October 29, 2020. © 1996 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1996;2:1199-1206. Clin Cancer Res S Oesterreich, S G Hilsenbeck, D R Ciocca, et al. cancer patients.prognostic marker in axillary lymph node-negative breast The small heat shock protein HSP27 is not an independent
Updated version
http://clincancerres.aacrjournals.org/content/2/7/1199
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/2/7/1199To request permission to re-use all or part of this article, use this link
Research. on October 29, 2020. © 1996 American Association for Cancerclincancerres.aacrjournals.org Downloaded from