Download Measurement by Color Video Image Analysis

ICANCERRESEARCH
54,4096-4102.August1, 19941
Detection of Discrete Androgen Receptor Epitopes in Prostate Cancer by
Immunostaining: Measurement by Color Video Image Analysis'
Wayne D. TilIey,2 Sylvia S Lim-Tio,3 David J Horsfall, James 0 Aspinall, Villis R. Marshall, and John M. Skinner
Cancer Cell Biology Uni4 Department of Surgery [W D. T., S. S. L-T., D. J. H., J. 0. A., V. R. M.J and Department of Pathology If. M. S.], School of Medicine, National
Cooperative Research Centre in Sensor Signal and Information Processing, Flinders University ofSouth Australia, P. 0. Box 2100, Adelaide, Set 5001, Australia
ABSTRACT
To determine whether multiple features ofimmunohistochemical stain
ing of the androgen receptor (AR) in prostate cancer could reliably
predict androgen dependence, tumor biopsy specimens from 30 patients
(stages A-D@)were stained using anti.peptide antibodies to the amino- and
carboxyl-termini of the AR. Measurements were made of the mean area
and total amount (i.e., integrated optical density) ofAR staining in at least
20 fields per section using a color video image analysis system, and the
mean
intensity
tumor
of AR staining
cells were
quantitative
derived.
differences
per cell and the percentage
Video
image
in AR staining
analysis
of AR positive
measurement
identified
between the two antibodies,
sug
gesting that this approach may provide a means of identifying receptor
variants
in prostate
tumors.
The AR staining
measurements
were
ann
lyzed by discriminant function analysis to assign individual cases to good
and poor clinical outcome groups. AR staining features measured with a
single antibody (e.g., amino-terminal) were sufficient to predict outcome
following hormonal therapy in stage D2 patients (predictive value, 1.0),
whereas all features of AR staining measured with both antibodies were
required for the entire patient group (predictive value, 0.97). The princi
pal discriminant in both patient groups contributing to the correct assign
ment
of outcome
was the mean
intensity
of AR staining
per celL These
findings suggest that AR staining features measured by video image
analysis have the potential to predict outcome in prostate cancer.
INTRODUCTION
Prostate cancer is now the most frequently diagnosed invasive
cancer and the second most common cause of cancer death in men in
Western societies (1, 2). At present, the majority of prostate cancer
patients are diagnosed with either locally invasive or, more frequently,
disseminated disease (3). Current forms of treatment for patients with
disseminated disease may prolong survival but are essentially only
palliative (4, 5). While an initial response is observed in about 70% of
contributes to the progression of human prostate cancers to an andro
gen-independent state.
It would be a considerable advantage to be able to offer patients
with advanced prostate cancer an alternative form of management
without waiting for evidence of relapse to hormonal therapy. In
contrast to breast cancer where biochemical (i.e., radioligand-binding)
determinations of estrogen and progesterone receptor levels have been
useful in predicting the response of individual tumors to hormonal
therapy (8), the predictive value of AR measurements in prostate
cancer is questionable (9). Radioligand binding to prostate tumor
homogenates does not discriminate between the AR content of ma
lignant versus stromal or benign epithelial cells (10, 11). More re
cently, with the availability of antibodies to the AR (12—15),it has
been possible to use immunohistochemical approaches to circumvent
this problem of AR expression in nonmalignant cells present in the
tumor biopsy tissue while, at the same time, provide some information
regarding the micro-heterogeneity of AR expression within the pros
tate tumor cell population (14—23).
In the present study, we have taken advantage of improvements in
computer-aided color VIA techniques (24, 25) to measure AR staining
features of tumor cells in a test cohort of 30 human prostate cancers.
The specific aims of this study were: (a) to determine whether it is
possible to distinguish between functional (ze., wild-type) and poten
tially nonfunctional (e.g., truncated) forms of the AR in prostate
tumors using antibodies directed against the NH2-terminus and the
COOH-terminus of the receptor (15); and (b) to determine whether
VIA measurements of AR staining using these antibodies are predic
tive for tumor progression (all patients) or outcome following hormo
nal therapy (stage D2 cases only).
MATERIALS AND METhODS
Patient Cohort. All patients accrued into this prospective study were
patients with advanced disease following primary endocrine ablation,
referred
most patients relapse within 3 years and only about 20% survive for
(Bedford Park, SA, Australia) and the Repatriation General Hospital (Daw
Park, SA, Australia) between March 1989 and September 1991. Patients
5 years
(4). The rapid
progression
of prostate
cancer
following
failure
presented
of primary hormone therapy is attributed to androgen-independent
tumor growth. In the androgen-resistant sublines of the R-3327 Dun
ning rat prostate adenocarcinoma and the androgen-insensitive human
to the
combined
with symptoms
Urology
Services
of acute urinary
of
obstruction
carcinoma of the prostate following pathological
resection
prostate cancer cell lines, PC-3 and DU145, androgen independence is
tissue.
The only exclusion
tissue with pathological
Flinders
criteria
were:
evidence of carcinoma;
Medical
and were diagnosed
Centre
with
assessment of transurethral
(a) lack of frozen
biopsy
and (b) loss of patients to
associated with a loss or a decrease in AR@mRNA and protein levels
(6, 7). It is not known whether a change in the expression of the AR
follow-up by the combined Urology Services. Patient details, including disease
stage according to the Modified Whitmore Jewett system (26), are shown in
Table 1.
Received 2114/94;accepted 5131194.
The costs of publication of this article were defrayed in part by the payment of page
based on clinical and biochemical features (i.e., serum PSA levels, bone and
computerized tomography scans, and survival). Serum PSA levels were mess
Patient
charges. This article must therefore be hereby marked advertisement in accordance with
18 U.S.C. Section 1734 solely to indicate this fact.
I This
work
was
supported
by grants
from
the National
Health
and
Medical
Research
Council of Australia, The Anti-Cancer Foundation of the Universities of South Australia,
and the (live and Vera Ramaciotti Foundations.
2 To whom
requests
for reprints
should
be addressed,
at Department
of Surgery,
address:
Prince
Henry's
Institute
of
Medical
Research,
Monash
Medical
Centre, Clayton, Vic, 3168, Australia.
4 The abbreviations used are: AR, androgen receptor; VIA, video image analysis; PSA,
prostate-specific
antigen; Cl', computerized
tomography;
PBS, phosphate-buffered
saline;
DAB, diaminobenzidene
tetrahydrochioride;
IOD, integrated optical density; DFA, dis
Groups.
Patients were allocated
to two outcome
groups
ured routinely by the Department of Clinical Biochemistry at FInders Medical
Centre using a solid phase, two-site immunoenzymatic assay (Tandem-E PSA;
Hybritech, Inc., San Diego, CA). For this assay, the normal range of serum
PSA levels in healthy men is 0 to 4 ng/ml. In this study, serum PSAlevels were
categorized into three groups: category 1, levels less than 4 ng/ml; category 2,
Flinders Medical Centre, Bedford Park, SA, 5042, Australia.
3 Present
Outcome
levels between 4 and 10 ng/ml; and category 3, levels greater than 10 ng/ml.
The actual outcome groups were defmed as follows. Group A, the good
outcome group, consisted of patients with stable or improving disease. Stable
cases showed no clinical evidence of progression, and the PSA levels remained
criminant function analysis; MA, mean area; MOD, mean optical density; R489, anti
in the original categories. Improving disease cases showed clinical evidence of
COOH terminal AR antibody; U402, anti-NH2 terminal AR antibody.
decreasing
tumor mass and a fall in PSA category.
Group B, the poor outcome
4096
Downloaded from cancerres.aacrjournals.org on June 18, 2017. © 1994 American Association for Cancer Research.
ANDROGEN RECEPTOR EXPRESSION LN PROSTATE CANCER
Table
1 Clinicalstage and treatment details for the 30 prosta
patientsTreatmentRadicalStageaprostatectomy
included in this studyte
VIA measurements were confined to the nuclei within sections of tumor
tissue. Discrimination between malignant and benign tissue areas was based
upon standard histological and cytological features (27). VIA measurements
cancer
were made of the total area stained brown with DAB for AR and the total
RadiotherapyOrchiectomyA@/B1―4B@
amount of AR staining (i.e., the lOD), in each field, for at least 20 fields per
section. The exact number of fields measured was determined by the cumu
lative quotient principle (28) to achieve variance below 5% for the mean area
Cd55D2e16
(equivalent
tonumber)
of tumornucleianalyzed.
Similarmeasurements
were
made for the AR negative, hematoxylin (i.e., blue)-stained
a Patients
@
were
b
staged
disease;
according
serum
to the Modified
PSA
level
within
Whitmore
Jewett
the normal
range
System
(i.e.,
<4 nglml);
tumor-negative regional lymph nodes.
C Clinically
localized
to
the
prostate;
d
spread;
e Metastatic
elevated
disease;
serum
elevated
PSA
variation of these parameters were derived for each tissue section: (a) the mean
area, MA of AR staining per field, which is an estimate of the number of nuclei
elevated
serum
PSA
level
(>4
ng/mI);
negative
bone scan.
@
nuclei. From these
measurements, the following parameters of AR staining and the coefficients of
(26).
level
serum
(>10
PSA
ng/ml);
level
negative
(>10
bone
ng/ml);
positive
positively stained; (b) the lOD per field, which is proportional
to the total
amount of AR staining; (c) the MOD (equals IOD/area of DAB staining),
scan.
bone
scan ±positive Cr scan for lymph nodes.
which is a measure of the average concentration of AR in the positively stained
nuclei; and (d) the percentage of AR-positive nuclei (Le., area DAB/area of
DAB + area of hematoxylin staining). Repeat immunostaining of tumor
sections was reproducible, as assessed by VIA, with coefficients of variation of
group, consisted of progressive disease cases or those who died from prostatic
carcinoma. Progressive disease cases showed an increase in tumor mass or the
development of metastases with or without a change in serum PSA levels. A
less than 5% for the MA and IOD measurements with either antibody (i.e.,
rise in PSA category (i.e., either from category 1 to 2 or 3, or from category
2 to 3) over a 6-month period was also used as an indicator of disease
assign cases to the predesignated (i.e., clinical) outcome groups (see above) on
progression.
The median (range) values for patient age and duration of follow-up for the
U402 and R489).
Statistical Analysis. The statistical technique of DFA (29) was used to
the basis of all staining features measured using VIA. Correlation analysis was
used to compare the results derived for each AR staining feature obtained for
the two antibodies. The statistical significance level (a) was set at a critical
test cohort were: Group A, 77 (66—85) years old, 37 (28—56)months follow
value
up; and Group B, 70 (65—83)years old, 29 (16—53)months follow-up. The
comparable values for the duration of follow-up for the subgroup of 16 stage
D2 patientswere: GroupA, 35 (28—46)months;and GroupB, 29 (16—39)
months.
Tissues and Tissue Processing. Immediately after resection, randomly
probability for the correlation coefficients was significant (i.e., P < 0.05). DFA
selected fragments
of prostatic tissue were embedded
in Tissue-Tek
(P)
of less
than
0.05.
Except
in one
instance,
as indicated
in the text,
the
was performed using the Minitab statistical package (Windows based version;
Minitab, Inc., State College, PA).
RESULTS
OCT
Immunohistochemical examination of AR expression in the human
prostate cancer biopsy specimens using either the anti-NH2 or the
anti-COOH terminal antibody to the human AR revealed a hetero
Louis, MO) coated slides and immediately fixed in 4% formaldehyde (Unilab;
geneous pattern of specific nuclear staining in the majority of tumors
Ajax Chemicals, Sydney, Australia)in PBS (pH 7.3) for 15 mm at room examined. Several distinct patterns of AR staining were observed.
temperature
(20°C).
Pattern 1 consisted of intense, uniform staining in all tumor nuclei
Tissue sections were made permeable in methanol containing 0.3% hydro
with both antibodies (2 of 30). In pattern 2, the majority of tumor
gen peroxide (H2O2)for 5 mm followed by acetone for 1 mm, both at —20°C. nuclei were positively stained by both antibodies but with variable
The sections were washed in two changes of PBS and stored at —20°C
in
intensity of staining (17 of 30). Pattern 2 is illustrated in Fig. 1 by a
specimen storage medium (8.6% w/v sucrose, 14% w/v magnesium chloride,
poorly
differentiated prostate carcinoma stained with the anti-NH2
and 50% v/v glycerol in PBS) until stained. The first and final sections of each
terminal
antibody, U402. Pattern 3 was one of discrete foci of tumor
series were stained
with a standard
laboratory
hematoxylin
and eosin stain for
cells with weak or absent AR staining and adjacent tumor cells or
confirmation of the histological diagnosis by a pathologist (J. M. S.).
apparently benign glands which were more intensely stained (5 of 30).
Androgea
Receptor
Immunohistochemistry.
Immunohistochemical
staining for the AR was performed as previously described (15) using two
Pattern 4 was similar to pattern 3, but the regions of negatively stained
affinity-purified rabbit anti-peptide antibodies that recognize epitopes in the tumor cells were more extensive (6 of 30). A region of a tumor section
NH2-terminal 21 amino acids (designated U402) and the COOH-terminal 20 with no AR staining is illustrated in the inset to Fig. 1. In many
amino acids (designated R489) of the human AR. The antiserum was gener
specimens, there was a degree of weak cytoplasmic staining, not
ously providedby Drs. Carol M. Wilson, Michael J. McPhaul,and Jean D. present in all cells, which possibly is the result of receptor diffusion
Wilson, Departmentof InternalMedicine, Universityof Texas Southwestern
into the aqueous media during thawing and processing of the sections
Medical Center, Dallas Texas.
prior to fixation.
ColorVIA. Theimmunostained
sectionswereexaminedusinganOlympus
VIA was used in this study to measure the area and total amount
BH-2 microscope (X200) coupled to a computer-aided color VIA system
(i.e.,
IOD) of AR staining in nuclei of the tumor biopsy specimens
(Video Pro 32; Leading Edge P/L, Adelaide, South Australia and Leica
Australia P/L). Briefly, images ofAR staining were captured using a Panasonic and, from these measurements, the mean receptor level per nucleus
color CCD video camera (model WV-CL700/A) and digitized using a PV 100 (i.e., MOD) and the percentage of AR-positive tumor cells were
multi-media 16-bit color video digitizer card in an Intel 80486 DX processor
derived. Fig. 2 shows the relationship between the percentage of cells
based personal computer. The resultant digitized image is displayed on a
positive for the AR epitope recognized by the anti-NH2 terminal
SVGA monitorin a 640 x 480 pixel variablewindow with 21-bit resolution. antibody and either: (a) the IOD or (b) the MOD of AR staining for
The image windows used by Video Pro 32 system are composed of up to all 30 prostate tumors examined. A poor correlation was obtained in
640 X 442 pixels and are separated into 8-bit brightness and color values for
both cases (IOD@@2 r@ = 0.5; MODU4O2 r@ = 0.26). Similar poor
each pixel. The range of grey levels for each pixel is from 0 to 255. The system
correlations were also obtained using the anti-COOH terminal anti
can define the color to be measured with great precision. The transmitted light
compound (Miles Scientific, Naperville, IL), frozen, and stored at —80°C
until
processed. Serial 5-sm frozen sections were cut using a Leitz cryostat at
—30°C,
thawed, and mounted onto poly-L-lysine (Sigma Chemical Co., St.
@
intensity was standardized by using a fixed rheostat setting at the microscope
body (IODR4S9
= 0.43; M0DR489
‘@
= 0.05, P > 0.05).
For any
light source which was connected to a voltage-stabilized conditioned power given percentage of AR positivity, both the total amount of AR (i.e.,
line. The stability of light output was measured from time to time using a IOD) and the average content per cell (i.e., MOD) varied by as much
photometer (Leica) as part of the routine VIA quality control procedures (25). as 6- to 7-fold between different tumor specimens (Fig. 2).
4097
Downloaded from cancerres.aacrjournals.org on June 18, 2017. © 1994 American Association for Cancer Research.
@
@
@
:@
m.
-@
,
ANDROGEN RECEFFOR EXPRESSION IN PROSTATE CANCER
1s)i
@:..@
a
@
@
ayt,
@,- .@,_
@‘
I •.aq@i1
@.•*
t,
•$
F.
,@
a
_
O.
1@
@
•‘*
IS
‘@1
•@
i,c
Fig. 1. AR staining in a frozen section of a
poorly differentiated prostate carcinoma showing
nuclear staining of variable intensity. The section
@
was stained for AR with U402 and counterstained
with hematoxylin. Low intensity cytoplasmic stain
ing is present in some of the tumor cells. Bar, 50
sun. Inset, region of a poorly differentiated prostate
@
@
carcinoma with no evidence of AR staining.
-@
I
I
F
.
t
c_
L%_
#@
f.@.
a..
t4V(I@@I@ %.
•::.
@
,‘
$dW&Tøji@,@@f4@
@3@/
@
.,.
d@
‘
.@
1@
@
‘
‘
#.
@
@
I
a
.@
a
1A:1:
@@1r@r
.
.
“a,
‘@sv
,
b
The VIA measurements also identified clear differences in AR
staining obtained with the two antibodies. Fig. 3 shows the relation
ships between the anti-NH2 and the anti-COOH terminal antibodies
for the different parameters of AR staining measured by VIA for all
30 tumor biopsies. Poor correlations between the two antibodies for
each parameter were obtained: (a) mean area of AR staining, MA,
r2
0.2; (b) IOD r@
0.14; (c) MOD r@
0
C
0.29; and (d) percentage
of AR positive cells, r@ = 0.30. Only in the case of the MOD
measurements for both AR antibodies do the cases cluster relatively
close to the regression line. The correlations between the primary
measurements of AR staining (i.e., MA and IOD) for the two anti
bodies shown in Fig. 3 identify several possible subpopulations within
the 30 cases examined: (a) comparable staining with both antibodies;
(b) markedly reduced staining with the anti-COOH terminal antibody
relative to the level measured with the anti-NH2 terminal antibody; (c)
markedly reduced staining with the anti-NH2 terminal antibody rela
tive to the level measured with the anti-COOH terminal antibody; and
(d) markedly reduced staining with both antibodies.
@1
a
0
D
‘C
AR@@ IOD (pixel density units)
The statistical technique of DFA was used to assign the 30 test
cases to the predesignated
clinical
outcome
groups
on the basis of the
VIA measurements of AR staining features. As shown in Tables 2 and
3, theDFAassigned
cases
totheoutcome
groups
witha highdegree
of concordance (97%) with the original clinical assessment. Only a
single case in the poor outcome group was differently classified
(Table 2). The principal discriminant features contributing to the
assignment of outcome groups are shown in Table 3. The most
powerful contributor to the DFA was the mean concentration of AR in
the tumor cells determined by the anti-NH2 terminal antibody (i.e.,
AR@@2MOD). The predictive values (i.e., number of cases correctly
assigned to the predesignated outcome groups by DFA/the total num
ber of cases) for the two antibodies, both individually and in combi
nation, are summarized in Table 4. Any single parameter of AR
staining in the DFA, e.g., the percentage of AR positive cells, did not
enable patient outcome to be accurately predicted, even with the use
of two AR antibodies (jredictive value, 0.63). Similarly, tumor stage
and grade did not accurately predict patient outcome, the predictive
value being only 0.63 when both parameters were included together in
the DFA (data not shown). When all measurements of AR staining
were used in the DFA, patient outcome was correctly assigned for
0.4
ARu402MOD(lOD/pixel)
Fig. 2. Relationship between percentage of AR positive nuclei and (A) IOD and (B)
MOD of AR staining for U402 measured by VIA in 30 prostate tumors.
4098
Downloaded from cancerres.aacrjournals.org on June 18, 2017. © 1994 American Association for Cancer Research.
@
S
1@@ @-
ANDROGEN RECEPTOR EXPRESSION IN PROSTATE CANCER
B
A.
S
6000-
10000@
U)
C
.35
S
0
‘C
0.
S
U)
C
.g 4000
Ca
Ca
0
C
‘C
.
@5000•
0.
0
.
@
S
•.
‘C
S
.
2000
S
S
S
@
A
,@
.
S
@.
‘C
S
,
S
S
IS@
A
I
5000
10000
15000
AR@2 meanarea(pixels)
20( 00
0
S
St
S
I
2000
I
I
4000
6000
ARu4o2IOD(pixeldensityunits)
C
0
0
I
@
ARu4o2
AR@402 MOD (IOD/pixel)
positivenuclei
Fig. 3. Relationship between AR staining with U402 and R489 for: (A) MA of AR staining/field; (B) IOD; (C) MOD; and (D) percentage of AR-positive cells measured by VIA
in 30 prostatetumors.
Table2
assignedby
Comparisonofpredesigna:ed(Le.,clinical)outcomeand outcome
DFA of VIA measurements stainingfor
ofailfeatures
ofAR and hematoxylin
all 30 patientsAssignment
outcomeOutcome
DFAGood
group
22Poor
of
ainic@
21
9
@
8
DFA (Table 5), and when analyzed alone, was not an accurate
indicator of response to hormonal therapy (Table 6). The VIA mea
surements of all AR staining features determined using the anti-NH2
terminal antibody were sufficient to independently assign all stage D2
cases to the designated outcome groups (jredictive value, 1.0; Table
6). The predictive power of the DFA classification using the VIA
measurements of all AR staining features determined with the anti
COOH terminal antibody was only marginally weaker (predictive
value, 0.94). For both antibodies, the mean concentration of AR in the
tumor cells (i.e., AR@402MOD and ARR@9 MOD) was the principal
feature of the AR staining contributing to the DFA (Table 5).
three quarters of the cases (Table 4). The predictive value was further
improved by inclusion of the hematoxylin counterstained AR negative
tumor cell features measured by VIA (Table 4). VIA measurements of
patientsPrincipal
Table 3 DFA for all
all staining features obtained with the anti-COOH terminal antibody
features of staining in tumor nuclei contributing to the assignment of patients
and hematoxylin counterstaining correctly assigned the majority of to outcome groups.Measured
patients to the designated outcome groups (predictive value, 0.87). By
(%)ARU4O2 features
Contribution to DFA
comparison, measurements obtained with the anti-NH2 terminal anti
MODb
54.4
body and hematoxylin staining were less useful in assigning patients
@U4O2 % positive
nuclei
14.7
@R489% positive nuclei
13.7
to these outcome groups (predictive value, 0.67). When all staining
Hematoxylin@@
MOD
11.1
features obtained with both antibodies were included in the DFA, the
Hematoxylin@@9MOD
5.3
predictive value was increased to 0.97.
All other features―
0.8
The subgroup of patients with advanced disease (i.e., stage D@)was
a
anti-NH2
terminal
androgen
receptor
antibody.
b MOD, mean optical density.
analyzed independently using DFA to determine the ability of the AR
C
@R489' anti-COOH
terminal
androgen
receptor
antibody.
staining features to predict outcome to androgen withdrawal therapy.
d Includes MA, IOD, and the coefficients of variation of all measurements for the AR
The percentage of AR positive cells was a weak contributor to the (both antibodies) and hematoxylin-stained tumor nuclei.
4099
Downloaded from cancerres.aacrjournals.org on June 18, 2017. © 1994 American Association for Cancer Research.
ANDROOENRECEPTOREXPRESSIONIN PROSTATECANCER
Table 4 Summary ofDFA for all staining features
Ability to correctly classify outcome for all 30 patients.
values%
yeAntibody
featuresbA'kJ4O2
nuclei
0.87ARu402
@R489
0.97a
+
coefficientof
Includes
@R4S9
MA,
Predictive
AR +
IOD,
AR feature?
0.50
0.63
0.50
0.77
0.63
MOD,
All
0.67
0.77
the percentage
of AR
positive
nuclei,
and the
variation
measurements.b
for these
Includes
all
AR
staining
features
and
similar
measurements
for
the
hematoxylin
stained,
nuclei.C
AR-deficient tumor
ARu4o2,
anti-NH2
d@
terminal
anti-COOH
androgen
terminal
stageD2Principal
receptor
androgen
antibody.
receptor
antibody.
Table5 DFAfor
features of staining in tumor nuclei contributing to the assignment of patients
to outcome groups.Measured
(%)@j(a
@R489
features
Contribution to DFA
MODE'
54.5
MOD
39.5
ARu4o2 % positive nuclei
3.4
‘@R4S9
1.1
% positive
nuclei
1.2a All other fes@.@d
@
anti-NH2 terminal androgen receptor antibody.
mean optical density.
b MOD,
C
@-‘@R489' anti-COOH
d Includes MA,
terminal
IOD,
androgen
receptor
and the coefficients
antibody.
of variation of all measurements of AR
staining using both
antibodies.Table
featuresAbility
6 Summary ofDFA for all AR staining
to correctly classify outcome for all stage
patients.Predictive
1)2
valuesAntibody
featuresaAR@@y@― % AR + ye nuclei
0.94aARR@J@,C
Includes
MA,
AR
0.69
0.57
IOD,
MOD,
the peroentage
1.00
of AR
positive
of variationfor thesemeasurements.
@
b
anti-NH2
C ARR@,
anti-COOH
DISCUSSION
terminal androgen receptor antibody.
terminal
androgen
receptor
antibody.
nuclei,
and the coefficient
assigning cases to the predesignated clinical outcome groups was the
mean amount of AR per tumor nucleus (MOD). In the study of Sadi
and Barrack (23), mean AR staining intensity was not significantly
different among the 17 stage D tumors examined, but the variance of
staining intensity in their study was significantly greater in the poor
responders to hormonal therapy than in the good responders. While a
statistical difference can be detected in our study between outcome
groups using the Mann-Whitney test to analyze individual parameters
of AR staining (data not shown), as in the case of Sadi and Barrack
(23), the large overlap of values limits the clinical usefulness of that
approach. The technique of DFA, as used in this study, incorporates
all measurements of AR staining in the assignment of individuals into
previously defined groups (e.g., good or bad outcome following
hormonal treatment). In a previous study of the prognostic value of
nuclear shape variables in advanced prostate cancer (30), the Mann
Whitney test was able to define differences between outcome groups
(i.e., response to hormonal therapy), but there was considerable over
lap of individual values diminishing the clinical usefulness of this
statistical approach. DFA in fact showed that it was not possible to
assign individual stage D2 patients to outcome groups using nuclear
shape features (predictive value, 0.65) to a clinically useful degree.
In the present study, the AR staining features determined by either
the anti-NH2 or anti-COOH terminal antibody were almost equally
effective in correctly allocating the 16 stage D2 patients to the desig
nated outcome groups. For both antibodies, the mean concentration of
AR in the tumor cells (MOD) was the principal but not exclusive
parameter contributing to the DFA, suggesting that this feature of AR
staining may evolve as a good indicator of response to hormonal
therapy in future studies. Staining parameters derived from the anti
COOH terminal antibody measurements were more valuable in cor
rectly assigning the entire test group, which included all clinical
stages, to the predesignated outcome groups than those obtained with
the anti-NH2 terminal antibody. Assignment was best for the 30 cases
using a combination of all nuclear staining features measured with the
two AR antibodies and the hematoxylin-only stained cells. This ob
servation suggests that the hematoxylin counterstaining may detect a
feature(s) of the AR-deficient tumor cell nuclei, such as aneuploidy or
variation in numbers of nucleolar organizer regions, which may be
related to tumor progression but, when analyzed independently, are
not good predictors of patient outcome (30—32).
The quantitative differences in AR staining measured with the
anti-NH2 and anti-COOH terminal antibodies possibly explain why
the predictive power of the DFA is increased by using a combination
of staining features obtained with both antibodies. The differences in
AR staining with the two antibodies are most apparent when the mean
area or IOD measurements are examined. These parameters reflect the
number of AR-positive tumor cells and the total amount of AR,
respectively, in the tumor biopsy specimen and identify three possible
subpopulations which are not revealed by the derived parameters of
staining. The main category, showing concordance of AR staining
with the anti-NH2 and anti-COOH terminal antibodies, is that cx
pected for staining with two AR antibodies when used under opti
mized conditions for immunohistochemistry (e.g., antibody excess)
and the structure of the receptor in the tumor samples is normal. This
was the sole category identified in similar studies of human benign
prostatic hyperplasia5 and breast cancer6 tissues stained for AR using
the two antibodies. Collectively, these observations suggest that the
two groups of prostate tumors identified as having a disproportionate
amount of total AR staining with the two antibodies may be of
physiological significance.
Previous inability to demonstrate a relationship between AR con
tent measured by radioligand binding assays and response to hormo
nal therapy in patients with metastatic prostate cancer has been
attributed to the diversity of cell types in tumor biopsy specimens (10,
11). The cloning of the human AR and the subsequent development of
antibodies have permitted the use of immunohistochemical methods
to evaluate the androgen-responsiveness of human prostate cancers.
With the exception of a recent report by Sadi and Barrack (23) using
a combination of immunohistochemistry and video image analysis to
determine the relative staining intensity of the AR in 17 patients with
stage D prostate cancer, AR staining intensity has not been objectively
measured. Previous studies have been restricted to the measurement of
the percentage of AR positive tumor cells (16, 18, 19) and/or the
subjective (i.e., observer-dependent) assessment of AR staining inten
sity in human prostate cancer tissues (17, 20—22).Neither method of
assessing AR staining, either alone or in combination, demonstrated a
relationship between AR expression in prostate tumor biopsies and
response to endocrine therapy. Using objective VIA measurements,
the present study has confirmed the earlier findings for the percentage
of AR positive cells. Indeed, our study has shown that the percentage
of AR positive tumor cells is not a good indicator of either the total
amount of AR present in a tumor biopsy or the mean amount of AR
per tumor cell, from which we infer that this staining feature may be
an unsuitable index of the androgen responsiveness of a prostate
cancer. We have shown that the principal parameter for the purpose of
4100
5 M.
A.
Grimbaldeston
6 R. E. Hall
and
and
W.
D.
W.
Tilley,
D.
Tilley,
unpublished
unpublished
observations.
observations.
Downloaded from cancerres.aacrjournals.org on June 18, 2017. © 1994 American Association for Cancer Research.
ANDROGENRECEflOR EXPRESSIONIN PROSTATECANCER
While the reason for the existence of the quantitative differences in
AR staining with the two antibodies in the prostate tumors is not clear,
the differences may reflect tumor-associated changes which are im
portent in progression. There is evidence that some prostate cancer
cells which continue to grow after initiation of anti-androgen therapy
retain the expression of the AR. For example, AR expression was
observed in 80% of prostatic tumor cells in 13 of 17 patients at
varying intervals after the commencement of androgen ablation ther
apy (19). The intensity of staining, however, was not measured.
Similarly, AR expression is retained by androgen-independent mouse
AR antibodies may provide new insights into the role of the AR in the
progression of prostate cancer.
mammary tumors (33). These observations suggest that mechanisms
early versus late endocrine therapy. In: 0. L Andriole and W. I. Catalona (eds.), The
other than the loss of AR expression are involved in the progression
Co., 1991.
5. Santen, R. J. Omi4@al review 37: endocrine treatment of prostate cancer. J. Clin.
Endocrinol. Metab., 75: 685—689, 1992.
of prostate cancer to an androgen-independent
state. A possible
cx
planation is the presence of a mutation in the AR gene in a subpopu
lation of tumor cells which results in aberrant regulation of growth by
steroids. In the case of inherited forms of androgen insensitivity,
subtle changes in the structure of the AR gene may have a profound
effect on receptor function and thus result in major phenotypic ab
normalities during male sexual development (34). Similar mutations
in the AR gene may be important in tumor progression and the
development of androgen independence in prostate cancer. For exam
plc, a mutation in exon 8 of the AR gene in the androgen-responsive
human prostate cancer cell line, LNCaP, results in an altered speci
ficity of the AR molecule for steroid binding and allows stimulation
of cell growth by estrogens, progestins, and anti-androgens in addition
to androgens (35). A mutation in the AR gene has been reported in an
early stage (i.e., stage B) primary human prostate cancer (1 of 26
patients, Ref. 36) and in two tumors in independent studies of endo
crine-resistant, metastatic (i.e., stage D@)disease (1 of 7 patients, Ref.
37; 1 of 9 patients, Ref.38). Similarly, estrogen receptor variants have
been described in human breast cancer tissues (8), and more recently,
the development of male breast cancer has been linked to mutations in
the DNA binding domain of the human AR (39).
While the frequency and clinical significance of AR mutations in
relation to tumor progression in prostate cancer remains to be deter
mined, it is possible that mutations in the AR could explain some of
the differences in AR expression detected by VIA with the two
antibodies
in this study.
For example,
the presence
REFERENCES
1. Boring, C. C., Squires, T. S., and Tong, T. Cancer statistics, 1993. CA Cancer J. ain.,
43: 7—26,1993.
2. Coffey, D. S. Prostate cancer. An overview ofan increasing dilemma. Cancer (Phila.),
71 (Suppl. 3): 880—886, 1993.
3. Scardino, P. T., Weaver, R., and Hudson, M. A. Early detection of prostate cancer.
Hum. Pathol.,23: 211—222,
1992.
4. Kozlowski, J. M., Ellis, W. J., and Grayhack, J. T. Advanced prostatic carcinoma:
Urologic Qinics of North America, Vol. 18, pp. 15—24.
Philadelphia: W. B. Saunders
6. Quarmby, V. E., Beckman, W. J., Cooke, D. B., Lubahn, D. B., Joseph, D. R., Wilson,
E. M., and French, F. S. Expression and localization of androgen receptor in the
R-3327 Dunning rat prostatic adenocarcinoma.
Cancer Res., 50: 735—739,1990.
7. Tilley, W. D., Wilson, C. M., Marcelli, M., and McPhaul, M. J. Androgen receptor
gene expression in human prostate carcinoma cell lines. Cancer Rca., 50: 5382—5386,
1990.
8. McOuire,W. L, Chamness,G. C., and Fuqua,S. A. Estrogenreceptorvariantsin
clinical breast cancer. Mol. Endocrinol., 5: 1571—1577,
1991.
9. Barrack, E. R., and TIndaII,D. J. A Criticalevaluation of the use of androgen receptor
assays to predict the androgen responsiveness of prostatic cancer. Ping. Gin. Blot.
Res., 239: 155—187,
1987.
10. Goreic@
L S., Lamm,D. L, Ramzy,I., Radwin,H. M.,andSham,S. A. Androgen
receptors in biopsy specimens of prostate adenocarcinoma. Heterogeneity of distal
bution and relation to prognostic significance of receptor measurements for survival
of advanced cancer patients. Cancer (Phila.), 60: 211—219,1987.
11. Martin, P. M., La Goff, J. M., Brissct, J. M., Ojasoo, T., Husson, J. M., and Raynaud,
J. P. Usesandlimitationsof hormone,receptorandenzymeassaysinprostatecancer.
In: 0. P. Murphy,S. Khoury, R. Kuss, et aL, (eds.), ProstateCancerPartA: Research,
Endocrine Treatment and Histopathology. New York: Alan R. Lisa, 1987.
12. Chang, C. S., Whelan, C. T., Popovich, T. C., Kokontis, I., and Liao, S. T. Fusion
proteins containing androgen receptor sequences and their use in the production
of poly- and monoclonal anti-androgen receptor antibodies. Endocrinology, 125:
1097—1099,1989.
13. van La.ar, J. H., Voohorst-Ogink, M. M., Zegers, N. D., Boersma, W. J. A., Claassen,
E., van der Korput, J. A. 0. M., Ruizeveld de Winter, J. A., van der Kwast, T. H.,
Mulder, E., Trapman, J., and Brinkmann, A. 0. Characterization of polyclonal
antibodies against the N-terminal domain of the androgen receptor. Mol. Cell.
Endocrinol., 67: 29—38,1989.
14. Sar, M., Lubahn, D. B., French, F. S., and Wilson, E. M. lmmunohistochemical
localization of the androgen receptor in rat and human tissues. Endocrinology, 127:
3180—3186,1990.
15. Husmann, D. A., Wilson, C. M., McPhaul, M. J., Tilley, W. D., and Wilson, 3. D.
of AR mutations
similar to those reported in individuals with either the partial or
16.
complete form of the androgen insensitivity syndrome, which can
result in NH2-terminal or COOH-terminal truncations of the receptor
17.
(34, 40), may contribute to the differences in staining intensity de
scribed in this study. The recent report (41) of both wild-type and a
functionally impaired Mr 87,000 form of the AR which is truncated at
18.
the NH2-terminus in normal genital skin fibroblasts raises the possi
biity that multiple forms of the AR might be present in both normal
and malignant androgen target tissues. An altered ratio of the wild
19.
type to an NH2-terminal truncated form of the AR in prostate cancer
is one possible mechanism currently being examined to explain the
20.
increased level of AR staining measured in four of the prostate tumors
with the anti-COOH terminal antibody relative to the level measured
21.
with the anti-NH2 terminal antibody.
In summary, objective assessment of AR staining in prostate cancer
22.
tissues using VIA in combination with DFA is potentially a powerful
method for predicting disease progression and/or response to hormo
23.
nal therapy. The present study has been performed on a relatively
small trial group of prostate cancer patients, and clearly it is necessary
24.
that a larger prospective study is conducted to confirm our findings.
An important aspect of the current study has been the measurement of
25.
AR staining with two antibodies since this has provided a strategic
method to identify prostate tumors containing AR which may be 26.
structurally abnormal. Further investigation of the basis of the quan
27.
titative differences in the levels of AR staining observed with the two
4101
Antipeptide antibodies to two distinct regions of the androgen receptor localize the
receptor protein to the nuclei of target cells In the rat and human prostate. Endocri
nology, 126: 2359—2368,1990.
Masai, M., Sumiya, H., Akimoto, S., Yatani, R., Chang, C. S., Liao, S. S., and
Shimazaki, J. Immunohistochemical study of androgen receptor in benign hyperplas
tic and cancerous human prostates. Prostate, 17: 293—300,1990.
de Winter, J. A., Trapman, J., Brinkmann, A. 0., Boersma, W. J., Mulder, E.,
Schroeder, F. H., Claassen, E., and van der Kwast, T. H. Androgen receptor hetero
geneity in human prostatic carcinomas visualized by immunohistochemistry. J.
Pathol., 160: 329—332,1990.
Sadi, M. V., Walsh, P. C., and Barrack, E. R. Immunohistochemical study of
androgen receptors in metastatic prostate cancer. Comparison of receptor content and
response to hormonal therapy. Cancer (Phila.), 67: 3057-3064,
1991.
van der Kwast, T. H., Schalken, J., Ruizeveld de Winter, J. A., van Vroonhoven,
C. C., Mulder, E., Boersma, W., and Trapman, J. Androgen receptors in endocrine
therapy-resistant human prostate cancer. mt. J. Cancer, 48: 189—193,1991.
Brolin, I., Lowhagen, T., and Skoog, L. Immunocytochemical
detection of the
androgen receptor in fine needle aspirates from benign and malignant human prostate.
Cytopathology, 3: 351—357,1992.
Chodak, 0. W., Kranc, D. M., Puy, L A., Takeda, H., Johnson, K., and Chang, C.
Nuclear localization of androgen receptor in heterogeneous samples of normal,
hyperplastic and neoplastic human prostate. J. Urol., 147: 798—803,1992.
Miyamoto, K. K.. McSherry, S. A., Dent, 0. A., Sar, M., Wilson, E. M., French, F. S.,
Sharief,Y., and Mohler,J. L Immunohistochemistry
of the androgenreceptorin
human benign and malignant prostate tissue. J. Urol., 149: 1015—1019,
1993.
Sadi, M. V., and Barrack, E. R. Image analysis of androgen receptor immunostaining
in metastatic prostate cancer. Heterogeneity as a predictor of response to hormonal
therapy. Cancer (Phila.), 71: 2574—2580,1993.
Jarvis, L R. The microcomputer and image analysis in diagnostic pathology. Microsc.
Res.Tech.,21: 292-299,1992.
Skinner, 3. M., Thao, Y., Coventry,
B., and Bradley, J. Video image analysis in
pathology. Din. Markers, 11: 53—70,1993.
Khoury, S. Staging of prostatic cancer. In: S. Khoury, C. Chatelain, 0. Murphy, and
L Denis(eds.),Urology:ProstateCancer,pp. 149—153.
Boissy-le-Cutté,
France:
FIIS, 1990.
Bocking, A., Kiehn, J., and Heinzel-Wach,
M. Combined histological
grading of
Downloaded from cancerres.aacrjournals.org on June 18, 2017. © 1994 American Association for Cancer Research.
ANDROGEN RECEFrOR EXPRESSION IN PROSTATE CANCER
prostate carcinoma. Cancer (Phila.), 50: 288—294,1982.
steroid binding characteristics and response to anti-androgens. Biochem. Biophys.
28. Aherne, W. A., and Dunnill, M. S. Preparation oftissues: sampling. In: Morphometry,
pp. 19—32.London: Edward Arnold, 1982.
Res.Commun.,173:534-540, 1990.
29. Hand, D. 1. Linear discriminant function. In: Discrimination and Classification, pp.
71—95.
Chichester: 1. Wiley and Sons, 1981.
36. Newmark, J. R., Hardy, D. 0., Tonb, D. C., Carter, B. S., Epstein, J. I., lasses, W. B.,
Brown, T. R., and Barrack, E. R. Androgen receptor gene mutations in human
prostate cancer. Proc. Nati. Acad. Sci. USA, 89: 6319—6323,1992.
30. Miller, J., Horsfall, D. J., Skinner, J. M., Rao, D. M., Leong, A. S. Y., and Marshall,
37. Culig, Z., Hobisch, A., Cronauer, A. C. B., Hittmair, A., Radmayr, C., Eberle, J.,
V. R. Nuclear shape and prognosis following orchiectomy in stage D2 prostate
Bartsch, 0., and Klocker, H. Mutant androgen receptor detected in an advanced-stage
prostatic carcinoma is activated by adrenal androgens and progesterone. Mol. Endo
crinol., 7: 1541—1550,
1993.
38. Suzuki, H., Sato, N., Watabe, Y., Masai, M., Seino, S., and Shimazaki, J. Androgen
carcinoma. Prostate, in press, 1994.
31. Hussain, M. H., Powell, I., Zaki, N., Maciorowski, M. S. Z., Sakr, W., KuKuruga, M.,
Visscher,D.,Haas,0. P.,Pontes,E.,andEnsley,J. F. FlowcytometricDNAanalysis
of fresh prostatic resections: correlation with conventional prognostic parameters in
patients with prostate cancer. Cancer (Phila.), 72: 3012—3019,1993.
32. Schned, A. R. Nucleolar organiser regions as discriminators
for the diagnosis of
well-differentiated adenocarcinoma of the prostate. Arch. Pathol. Lab. Med., 117:
1000—1004,1993.
33. Darbre, P. D., and King, R. J. Progression to steroid insensitivity can occur irrespec
tive of the presence of functional steroid receptors. Cell, 51: 521—528,1987.
34. McPhaul, M. 1., Marcelli, M., Tilley, W. D., Griffin, J. E., and Wilson, J. D. Androgen
resistance caused by mutations in the androgen receptor gene. FASEB J., 5: 2910—
2915, 1991.
35. Veldscholte, J., Ris-Stalpers, C., Kuiper, G. 0., Jenster, G., Berrevoets, C., Claassen,
E., van Rooij, H. C., Trapman, J., Brinkmann, A. 0., and Mulder, E. A mutation in
the ligandbindingdomainof the androgenreceptorof humanLNCaPcellsaffects
receptor gene mutations in human prostate cancer. J. Steroid Biochem. Mol. BioL, 46:
759—765,1993.
39. Wooster, R., Mangion, J., Eeles, R., Smith, S., Dowsett, M., Averill, D., Barreu, L P.,
Easton, D. F., Ponder, B. A, and Stratton, M. R. A germline mutation in the androgen
receptorgene in two brotherswith breastcancerand Reifensteinsyndrome.Nat.
Genet., 2: 132—134,1992.
40. Zoppi, S., Wilson, C. M., Harbison, M. D., Griffin, J. E., Wilson, J. D., McPhaul,
M.J.,andMarcelli,M.Complete
testicular
feminization
causedbyanamino-terminal
truncation of the androgen receptor with downstream initiation. J. Clin. Invest., 91:
1105—1112,
1993.
41. Wilson, C. M., and McPhaul, M. J. A and B forms of the androgen receptor
are present in human genital skin fibroblasts. Proc. NatI. Acad. Sd. USA, 91:
1234—1238,
1994.
4102
Downloaded from cancerres.aacrjournals.org on June 18, 2017. © 1994 American Association for Cancer Research.
Detection of Discrete Androgen Receptor Epitopes in Prostate
Cancer by Immunostaining: Measurement by Color Video Image
Analysis
Wayne D. Tilley, Sylvia S. Lim-Tio, David J. Horsfall, et al.
Cancer Res 1994;54:4096-4102.
Updated version
E-mail alerts
Reprints and
Subscriptions
Permissions
Access the most recent version of this article at:
http://cancerres.aacrjournals.org/content/54/15/4096
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 June 18, 2017. © 1994 American Association for Cancer Research.