`
`Changes in Estrogen Receptor, Progesterone Receptor, and pS2 Expression in
`
`Tamoxifen-resistant Human Breast Cancer‘
`
`S. R. D. Johnston,’ G. Saccani-Jotti, I. E. Smith, J. Salter, J. Newby, M. Coppen, S. R. Ebbs, and M. Dowsett
`Academic Department of Biochemistry [S. R. D. J., J. S., J. N., M. D.I and the Breast Unit [S. R. D. 1.. I. E. S., J. N.], The Royal Marsden Hospital, Fulham Road, London.
`SW3 6].], England; Istituto di Anatomia ed Istalogia Patalogica, University of Parma. Panna, Italy (G. S-.I.]; and Departments of Pathology [M. C.] and Surgery [S. R. E.].
`Mayday University Hospital, Croydon, Surrey, CR 7 7YE, England
`
`ABSTRACT
`
`Changes in estrogen receptor (ER) expression and function may explain
`the development of tamoxifen resistance in breast cancer. ER expression
`was measured by an immunohistochemical assay, validated for use in
`tamoxifen-treated tumors against a biochemical enzyme immunoassay,
`in 72 paired biopsies taken before treatment and at progression or relapse
`on tamoxifen. Progesterone receptor (PgR) and pS2 gene expression were
`also measured immunohistochemieally as an indicator of ER function.
`Overall the frequency of ER expression was reduced from 37 of 72
`(51%) pretamoxifen to 21 of 72 (29%) at progression or relapse, with a
`significant reduction in the quantitative level of ER (P < 0.0001; Wilcoxon
`signed rank sum test). Tumors treated with primary tamoxifen that
`responded but then developed acquired resistance frequently remained
`ER positive (ER+) at relapse: 16 of 18 (89%) were ER+ pretamoxifen
`(75% of these expressed either PgR or pS2) and 11 of 18 (61%) were ER+
`at relapse (82% continued to express PgR or pS2). In contrast, only 3 of
`20 (15%) tumors that progressed on primary tamoxifen with de nova
`resistance were ER+ pretamoxifen, and all tumors were ER- at progres-
`sion. At progression, 6 of 20 (30%) of these tumors expressed high levels
`of PgR (mean H-score, 98) andlor pS2 (mean, 50% cells positive), despite
`being ER-. In tumors that recurred during adiuvant tamoxifen therapy,
`including locoregional and metastatic lesions, ER expression was signifi-
`cantly reduced from 18 of 34 (53%) in the original primary tumor to 10
`of 34 (29%) at relapse (P = 0.002). PgR expression was likewise signifi-
`cantly reduced in this group (P = 0.001).
`This study confirms that expression of a functional ER in breast cancer
`is a strong predictor for primary response to tamoxifen. Although ER was
`reduced in tamoxifen-resistant tumors overall, the development of ac-
`quired resistance was associated with maintained ER expression and
`function in many tumors, whereas dc nova resistance remained related to
`lack of ER expression. Recurrence during adjuvant tamoxifen was asso-
`ciated with development of an ER/PgR-negative phenotype in some
`tumors. These data imply that separate mechanisms of resistance may
`occur in these different clinical subgroups.
`
`INTRODUCTION
`
`The response to the antiestrogen tamoxifen in human breast cancer
`occurs more frequently in tumors that contain significant quantities of
`ER3 (1). Many tumors that do not respond and thereby demonstrate
`primary de novo resistance to tamoxifen lack detectable ER protein,
`and this generally renders them resistant to other endocrine therapies
`including progestins and aromatase inhibitors. Most of the tumors that
`do respond initially to tamoxifen eventually progress with acquired
`resistance, although clinical evidence suggests that many remain
`sensitive to further endocrine therapies. These observations suggest
`
`Received 2/9/95; accepted 5/23/95.
`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.
`‘ This study was supported by the Cancer Research Campaign. S. R. D. J. is a CRC
`Clinical Research Training Fellow. G. S-J. was supported in part by the following grants:
`CNR ACRO Project, CNR Bilateral 94.02482.CF04, and MURST 40% (Italy).
`2 To whom requests for reprints should be addressed.
`3 The abbreviations used are: ER, estrogen receptor; ER+, ER positive; ER-, ER
`negative; PgR, progesterone receptor; EIA, enzyme immunoassay; IHA, immunohisto-
`chemical assay.
`
`that a basic biological difference exists between tumors with acquired
`tamoxifen resistance and those with intrinsic resistance to the drug.
`Tamoxifen may modulate the expression of ER in hormone-depen-
`dent breast cancer. Although the mechanism of action of the drug is
`to compete with estrogen for the binding site of ER and to inhibit
`estrogen-induced growth, there is evidence in vitro that ER expression
`itself may become up-regulated after tamoxifen (2). However, immu-
`nohistochemical studies have demonstrated that ER expression within
`breast cancers is heterogenous, and theoretically through selective
`pressure tamoxifen could permit the survival of clones of ER- cells
`while inhibiting the growth of ER+ clones. Ultimately, this could
`allow the emergence of an ER- hormone-independent tumor that was
`no longer sensitive to tamoxifen.
`There have been previous studies that have examined the expres-
`sion of ER in tumor biopsies from patients during tamoxifen therapy
`(3~5). In general, these studies found tamoxifen-treated tumors to be
`ER-, supporting the hypothesis of clonal selection. However, the
`most commonly used technique to measure ER in these samples was
`the ligand-binding assay, and tamoxifen may have given false nega-
`tive results due to competition with estrogen for the binding site of
`ER. In addition, this assay requires relatively large quantities of fresh
`tissue, which limits any retrospective comparisons of relapsed with
`primary minors. The more recent development of lHAs with the use
`of mAbs has allowed ER to be measured in paraffin-embedded
`material. These assays have the advantages of detecting tamoxifen-
`bound receptor, requiring very small amounts of tissue, and permitting
`study of the heterogeneity of ER expression within tumors. Several
`groups, including our own, have now validated these assays against
`conventional biochemical techniques (6-8).
`The functional activity of ER may be as important as its level of
`expression, particularly in determining whether endocrine therapy is
`of value. The expression of several proteins is known to be estrogen-
`regulated, including the PgR and the product of the pS2 gene (9, 10).
`ER+ tumors that express PgR have been shown to be more likely to
`benefit from endocrine therapy in the adjuvant setting (11). Likewise,
`pS2 expression in ER+ tumors has been found to improve the
`likelihood of response to endocrine therapy in advanced breast cancer
`(12). II-IAs are now available to measure both of these proteins in
`paraffin-embedded tissue. This allows, therefore, a more complete
`characterization of the ER—related phenotype to be made, which may
`give an indication of the function of ER, in addition to its level of
`expression.
`In this study, we have analyzed ER expression and function, as
`measured by PgR and pS2 expression, in 72 patients with documented
`resistance to tamoxifen. In all patients, a biopsy from the primary
`tumor had been taken before tamoxifen was started, allowing direct
`within-patient comparison of the change in expression of ER, PgR,
`and pS2 in relation to the development of tamoxifen resistance in vivo.
`
`MATERIALS AND METHODS
`
`Patients and Tissue Samples. Seventy-two women with breast cancer who
`progressed during tamoxifen therapy (20 mg daily) were studied. All patients
`attended either the Royal Marsden Hospital or the Mayday University
`3331
`
`|nnoPharma Exhibit 10660001
`
`
`
`ER, PgR. AND pS2 IN TAMOXIFEN-RESISTANT BREAST CANCER
`
`Hospital. Thirty-eight patients had been treated with tamoxifen as primary
`medical therapy when they first presented with breast cancer. In general,
`these were postmenopausal women who were treated with tamoxifen rather
`than surgery because of age, tumor size, or advanced local disease. Of these
`patients, 18 responded to tamoxifen but subsequently relapsed with ac-
`quired resistance, and 20 progressed during initial tamoxifen therapy with
`primary de novo resistance. Response was defined clinically according to
`standard UICC criteria (13) in tenns of change in bidimensional tumor
`measurements (caliper), with a partial response representing a greater than
`50% reduction in the product of the two measurements, and a complete
`response when the tumor was no longer palpable. Progression during
`treatment included tumors in which there was a greater than 25% increase
`in size and those where there was no change in tumor measurement (<50%
`reduction or <25% increase in size). Three of the 20 tumors that progressed
`on primary tamoxifen had no change in tumor measurements for more than
`6 months before progression. Some authors consider these to represent
`clinical “responses,” but for the purpose of this study objective response to
`primary tamoxifen only included those with documented partial of com-
`plete response. In all these cases the tumor at relapse or progression was
`compared with a trucut biopsy taken from the same tumor before tamoxifen
`was started.
`
`In an additional 34 patients, tamoxifen had been given as adjuvant therapy
`after initial surgical management, and in these cases the tamoxifen-relapsed
`tumor was compared with the original excised primary tumor. In 15 cases the
`tumor which developed during adjuvant therapy represented a local recurrence
`in the breast, whereas in 19 cases the recurrence was at a different site (11 as
`lymph node metastases and 8 as skin nodules). The demographic data for these
`three groups of patients are shown in Table 1.
`At relapse, mastectomy or excision biopsy specimens were delivered fresh
`to histopathology after resection and were processed immediately. A portion of
`tumor, approximately 200 mg in size, was snap frozen in liquid nitrogen
`immediately after dissection from the breast, and stored at -80°C for analysis
`of ER by EIA. The remainder of the tumor was fixed for approximately 24 h
`in 10% buffered formalin. The tissues were embedded in paraffin wax after a
`routine processing procedure that did not exceed 60°C. Sequential adjacent
`sections (3 p.m) were cut from the paraffin-embedded tumors onto slides
`coated with either 3-aminopropyltriethoxysilane (Sigma Chemical Co.) for the
`ER and PgR assays or poly-L-lysine for the pS2 assay. The sections were air
`dried overnight in an oven at 37°C, and one section was stained with hema-
`toxylin and eosin for light microscope assessment.
`ER IHA. We have previously described this IHA, which has been validated
`against the conventional biochemical EIA (8). In brief, sections were predi-
`gested in 10 mM citrate buffer (pH 6.0) by microwaving (750 W, full power)
`for two 5-min intervals. After blocking endogenous peroxidase activity,
`sections were incubated with monoclonal anti-human ER antibody lD5 (Dako)
`for 2 h (1:100 dilution), rinsed in PBS, and incubated in biotinylated rabbit
`anti-mouse immunoglobulin (Dako) for 45 min (1:100 dilution). After incu-
`bation with horseradish peroxidase-conjugated streptavidin complex (dilution
`1:200) for 1 h, a solution of 0.05% 3,3’-diaminobenzidine (Sigma) dissolved
`in dimethyl forrnarnide plus 100 iii of 30 volumes hydrogen peroxide-100 ml
`PBS was used to develop the peroxidase activity. Previously identified strongly
`ER+ tumors were used as positive controls, with negative controls being
`derived by omission of the primary antibody.
`Ten fields (minimum 500 cells) were chosen at random at X400 magnifi-
`cation for scoring of nuclear staining. Staining intensity was assessed as
`negative, weak, intermediate, or strong (index, 0 to 3), and the percentage of
`cells at each intensity estimated to give an overall “H-score,” ranging from
`
`0-300 (14). We have previously validated our scoring system for this assay
`against the biochemical EIA (ER-EIA) (8). Stroma, normal, and benign epi-
`thelial tissue were excluded, and a tumor was designated ER+ if the I-I-score
`was >20.
`
`PgR and pS2 IHAS. For the PgR assay, no predigestion or microwave
`enhancement was required. The methods were similar to those described
`above, although sections were incubated overnight with a 1:2 dilution of
`monoclonal anti-human PgR antibody (0.1 mg/ml) from the Abbott immuno-
`histochemical kit, followed by a biotinylated rabbit anti-rat antibody at a
`dilution of 1:100 for 45 min. The detection method and scoring system were
`similar to those used for ER.
`The pS2 assay used a mouse anti-pS2 mAb BC6 (gift from Professor P.
`Charnbon, Paris, France), which we have described previously and validated
`against an immunoradiometric biochemical assay (15). Scoring was assessed
`by counting the number of malignant cells with cytoplasmic staining for pS2
`and expressing this as a percentage of the total number of malignant cells with
`the use of a positive cutoff of 10%.
`Comparison of ER II-IA with ER EIA in Tamoxifen-treated Tumors.
`We have already demonstrated the close relationship between ER measured by
`the IHA method described above and by EIA (Abbott) in a separate series of
`119 primary breast cancers (8). It has been shown previously that tamoxifen
`does not lead to false negative results in the EIA (16). To ensure that tamoxifen
`did not interfere with the IHA results in the resistant samples, a separate cohort
`of 33 primary breast cancers from postmenopausal women who had been
`treated with tamoxifen for 2-3 weeks before surgery was studied. In these
`tumors, we measured ER by IHA in paraffin-embedded sections and by EIA in
`frozen tumor samples. A similar comparison between ER IHA and ER EIA
`was also made in a total of 98 tamoxifen-resistant tumors. This cohort com-
`
`prised 40 of the 72 resistant tumors described above, where a frozen sample of
`tumor was also available for EIA. In addition, 58 tamoxifen-resistant tumors
`were available from patients for whom no matched pretreatment biopsy existed
`for the currently reported paired immunohistochemical study, but in whom for
`validation purposes at relapse a frozen sample for EIA could be compared with
`a paraffin-embedded sample for IHA.
`For the EIA method, the frozen tumor sample was pulverized in a micro-
`dismembrator (Braun Medical, Ltd.) for 1 min after cooling in liquid nitrogen.
`The powdered tumor was reconstituted 1:8 (w/v) in iced tris/molybdate buffer
`[5 mM sodium molybdate, 10 mM monothioglycerol, 1 mM dipotassium chlo-
`ride EDTA, 3 mM sodium azide, and 10 mM TRIS (pH 7.4)], and the
`homogenate was centrifuged at 4°C for 20 min at 2000 X g, after which the
`cytosol fraction was removed and diluted 1:5 in tris/molybdate buffer for
`protein assay with the use of the Bio-Rad method with a bovine y globulin
`standard. An aliquot of the cytosol was diluted to give a protein concentration
`of 1-2 mg/ml. The ER levels in the diluted cytosols were determined with the
`use of the ER EIA kit from Abbott Diagnostics, according to the manufactur-
`er’s instructions, and values >10 frnol/rng protein were regarded as positive.
`Statistics. Comparisons between the semiquantitative scores for the IHA
`and EIA assays were made by linear regression analysis. The mean I-I-scores
`GER and PgR) and the mean percentage positive cells (pS2) were calculated for
`all the pretamoxifen and tamoxifen relapse-positive tumors, and within each of
`the three clinical subgroups. In view of the wide range in values and absence
`of a normal distribution of quantitative data for ER, PgR, and pS2 in each
`group, nonparametric paired analysis of the change in absolute value for each
`parameter was performed with the use of the Wilcoxon signed rank sum test.
`
`RESULTS
`
`Table 1 Characteristics ofpatients with breast cancer who developed
`tamoxifen resistance
`No. of
`patients
`
`Median
`age (yr)
`
`Median time to
`relapse (mo.)
`
`Primary tamoxifen
`Responders: acquired resistance
`Nonresponders: de novo resistance
`Adjuvant tamoxifen
`Loooregional recurrence
`Metastatic recurrence
`
`18
`20
`
`26
`8
`
`74
`72
`
`60
`55
`
`24
`4
`
`27
`14
`
`Comparison of ER IHA with ER EIA in Tamoxifen-treated
`Tumors. The immunohistochemical staining obtained with the IDS
`antibody on paraffin-embedded sections produced clear nuclear stain-
`ing in invasive carcinoma cells (Fig. 1). In our previous comparison of
`the H-score system for ER with the biochemical EIA in 119 untreated
`primary breast cancers, we found a concordance rate of 86%, with a
`positive correlation between the scores (r = 0.605). In the 33 tamox-
`ifen-treated primary breast cancers, the concordance rate was 96%;
`only 1 tumor was EIA+ and IHA—, and the IHA score was borderline
`negative (H-score 16). A strong positive correlation (r = 0.934) was
`shown between the IHA H-score for ER and the EIA value in frnol/mg
`3332
`
`|nnoPharma Exhibit 10660002
`
`
`
`ER. Pgll, AND pS2 lN TAMOXIFEN-RESISTANT BREAST CANCER
`
`0'”)
`
`'****i.->af
`
`l’araffin—embedded section of an inva-
`Fig.1.
`sive ductal carcinoma of the breast stained with
`1D5 mAb against ER. Uniform nuclear staining is
`observed in tumor cells. X 10()().
`
`protein (Fig. 2). From these data it does not appear that tamoxifen
`prevents the detection of ER by the IDS antibody in paraffin-embedded
`sections.
`
`In the 98 tamoxifen-resistant tumors, both the ER EIA and ER IHA
`scores for ER+ tumors were lower than scores for the primary ER+
`tumors treated with short-terrn tamoxifen (Fig. 2). The concordance
`rate between the two methods was lower at 66%, which appeared to
`be largely explained by 30 tumors that were EIA+ but IHA-
`(Table 2). However, 25 of these 30 tumors had borderline scores on
`either or both assays; 15 tumors were borderline EIA+ (10-35
`fmol/mg protein) and IHA—, 6 tumors were borderline ll-IA-
`(H-score 5-19) and EIA+, and 2 tumors were both borderline EIA+
`and borderline IHA-. In addition 2 EIA+ tumors were IHA- within
`
`the invasive tumor component but contained benign epithelial cells
`that were strongly positive. The remaining 5 tumors, which were
`completely negative by IHA, had EIA scores which ranged between
`46 and 112 fmol/mg protein. There were 3 tumors that were IHA+ but
`EIA—, but in all 3 cases the EIA score was borderline negative (5-9
`frnol/mg protein). Overall,
`there was weaker positive correlation
`between the two methodologies for the tamoxifen-resistant tumors
`(r = 0.561) compared with either the short-term tamoxifen-treated
`cohort or our previous data on primary untreated tumors.
`
`Table 2 Concordance between ER EIA and ER IHA in tamoxifen-Irealed and
`tamoxifen-re.ri.rIanl human breast cancer
`
`n
`Concordance
`EIA+/lHA+
`EIA+/IHA—
`EIA-/ll-IA+
`EIA-/lHA—
`
`Primary tamoxifen
`33
`96%
`22
`1
`0
`10
`
`Tamoxifen resistant
`98
`66%
`26
`30
`3
`39
`
`200
`
`150
`
`100
`
`IHAER(H-Score)
`
`o
`
`0
`
`Tamoxifen Resistant
`Tumors (n=98)
`concordance = 66%
`r = 0.561
`
`Tamoxifen Treated
`Primary Tumors (n=33)
`concordance = 96%
`
`Change in ER IHA in Tamoxifen-resistant Breast Cancer.
`Before tamoxifen, 37 of 72 (51%) tumors were ER+ by the IHA
`assay, with a mean H-score for ER+ tumors of 90 1 7 (SEM). At
`relapse, only 21 of 72 (29%) tumors were ER+ (mean H-score for
`ER+ tumors, 61 i 9). A direct comparison of the changes in H-score
`between the 72 pairs of samples showed that, overall, there was a
`significant
`fall
`in ER expression at
`relapse (Z value,
`-4.52;
`P < 0.0001, Wilcoxon signed rank sum test).
`Analysis of the three clinical subgroups of tamoxifen resistance
`revealed different patterns of ER expression. In tumors treated with
`primary tamoxifen, ER expression within the same tumor was com-
`pared between the pretreatment biopsy and the tumor at relapse or
`progression. Of those that responded to primary tamoxifen, 16 of 18
`(89%) were ER+ at presentation (mean H-score for ER+ tumors, 70).
`At subsequent relapse in these patients analysis of the same tumor
`showed that 11 (61%) were still ER+ (mean H-score of 66), and in 4
`of these cases the ER score had increased. In total, 5 ER+ tumors had
`become ER-, and 2 originally ER- tumors remained ER- at re-
`lapse. Paired comparison between the 18 presentation and relapse
`samples revealed no significant difference in ER score (Fig. 3).
`Of the tumors treated with primary tamoxifen that progressed on
`treatment (de novo resistance), the majority (17 of 20) was ER- at
`presentation. At progression in the repeat biopsy from the sarrte tumor,
`all 20 samples were ER-. On paired analysis of the ER scores, this
`represented a significant reduction (Z value, -2.67; P = 0.008;
`Fig. 4).
`In the adjuvant group, comparison of ER expression was made
`Fig. 2. Scattergram of EIA versus ll-IA for ER in 33 tamoxifen-treated primary tumors
`and 98 tamoxifen-resistant tumors.
`between the original primary tumor and the recurrent tumor at relapse
`3333
`
`r = 0.934
`
`I00
`
`150
`
`200
`
`250
`
`300
`
`EIA ER (fmol/mg protein)
`
`|nnoPharma Exhibit 1066.0003
`
`
`
`ER, PgR, AND pS2 IN TAMOXIFEN-RESISTANT BREAST CANCER
`
`PgR
`
`
`
`Pro-TAM
`
`TAM Relapse
`
`
`
`Fig. 3. Change in ER, PgR, and pS2 status in 18
`paired samples from patients with acquired tamoxifen
`(TAM) resistance. NS, not significant.
`
`ER4-Va:
`Man Score:
`Wilcoxon:
`
`ll (61%)
`16 (89%)
`66
`70
`NS (p=0.12)
`
`
`
`PgR+ve:
`Mean Score:
`Wileoxon:
`
`10 (56%)
`ll (61%)
`5'7
`82
`NS (p=0.ll)
`
`pszwe:
`Mean Score:
`Wileomn:
`
`ll (61%)
`to (56%)
`45%
`53%
`NS (p=0.90)
`
`on tamoxifen. Of the original primary tumors, 18 of 34 (53%) were
`ER+ (mean H-score of ER+ tumors, 103). At progression on tamox-
`ifen only 10 of 34 (29%) were ER+, and in these tumors the mean
`H-score was reduced to 57. Paired comparison showed a highly
`significant reduction in ER expression between the primary and
`relapsed tumor in this group (Z value, -3.77; P = 0.0002). Of the 18
`ER+ tumors, 9 became ER-, and in 6 the ER score was reduced by
`more than 50% (Fig. 5). Of these original 18 ER+ primary tumors, 12
`recurred on tamoxifen with locoregional tumor and 6 with metastatic
`skin nodules. In these patients, a greater number of locoregional
`recurrences than metastatic tumors remained ER+ (8 of 12 versus 2
`of 6; Table 4).
`Change in PgR and pS2 in Tamoxifen-resistant Breast Cancer.
`Overall, there was no significant difference in the frequency of PgR+
`or pS2+ tumors between the pretreatment and tamoxifen-resistant
`groups (Table 3). The mean scores for the PgR+ or pS2+ tumors
`were similar, although paired comparison of the change in score for all
`72 cases showed a significant reduction in PgR (P = 0.01) but not pS2
`expression (P = 0.49).
`Of the 18 primary treated tumors that developed acquired resist-
`ance, 11 (61%) were PgR+ and 10 (56%) were pS2+ at presentation.
`All of these tumors that expressed PgR or pS2 were also ER+. At
`relapse, 10 of 18 (56%) tumors were PgR+. Whereas the mean
`H-score in these tumors was lower at relapse compared with preta-
`moxifen (57 versus 82), paired comparison showed no significant
`difference (Fig. 3). Two tumors that were originally PgR— became
`
`PgR+ at relapse, one of which was strongly PgR+ (H-score = 61)
`despite being completely ER- both at presentation and at relapse. pS2
`expression was likewise unchanged in these 18 tumors, with 11 (61%)
`tumors pS2+ at relapse (Fig. 3).
`Of the 20 tumors that progressed on primary tamoxifen, 3 (15%)
`were PgR+ and 1 (5%) was pS2+ before therapy (all but one of these
`tumors was ER+). Whereas all tumors were ER- at progression, 6
`(30%) were PgR+, of which 3 were also pS2+ (Fig. 4). Four of these
`tumors had zero scores for ER, PgR, and pS2 in the pretamoxifen
`biopsy. The quantitative level of both PgR and pS2 expression in these
`ER- tumors at progression was relatively high (mean PgR H-score,
`91; mean pS2 score, 50% positive cells).
`In sequential adjacent
`sections from one of these tumors (Fig. 6), populations of invasive
`carcinoma cells that were negative for ER stained positive for PgR
`and pS2.
`In the adjuvant group, 13 (38%) of the original primary tumors
`were PgR+, and 12 (35%) tumors pS2+. There was a reduction in the
`expression of PgR in the relapsed tumor (38 to 12%), which was
`highly significant on paired analysis (Z value, -3.29; P = 0.001). In
`contrast, the frequency of pS2 expression did not change (Fig. 5). Ten
`of the 12 pS2+ tumors at relapse were local or nodal recurrence,
`whereas only 2 cases were metastatic tumors (both ER+). In addition,
`4 of the 12 recurrent pS2+ tumors were ER-, including one that also
`expressed PgR. However, all the ER+ recurrences during adjuvant
`tamoxifen (5 local, 3 regional, and 2 metastatic) expressed either PgR
`(n = 3) or pS2 (n = 8; Table 4).
`
`ER
`
`Wilcoxon: Wilcoxon:
`
`eP=0_o0s
`
`NS (p=0.44)
`
`ER+ve:
`Mean Score:
`
`3 (15%)
`99
`
`0 (05)
`0
`
`PgR-we:
`Mean Score:
`
`3 (15%)
`75
`
`6 (20%)
`91
`
`pS2
`
`100
`
`75
`
`Fig. 4. Change in ER, PgR, and pS2 status in 20 paired
`samples from patients with de nova tamoxifen (TAM)
`resistance. NS, not significant.
`
`1 (5%)
`60%
`
`3 (15%)
`50%
`
`Mean Score:
`
`NS (p=0.l4)
`
`3334
`
`|nnoPharma Exhibit 1066.0004
`
`
`
`ER, PgR, AND pS2 IN TAMOXIFEN-RESISTANT BREAST CANCER
`
`ER
`
`
`
`
`
`Fig.5. Change in ER, PgR, and pS2 status in 34
`paired samples from patients with resistance to adju-
`vant tamoxifen (TAM) therapy. NS, not significant.
`
`ER1-ve:
`Mean Score:
`Wlleoxon:
`
`18 (53%)
`103
`
`I0 (29%)
`
`‘ p=0.0002
`
`57
`
`Pglhve:
`Mean Score:
`
`Wicoxon:
`
`I3 (38%)
`53
`
`4 (12%)
`41
`
`- p=0.00l
`
`pS2+ve:
`Mean Score:
`Wileoxon:
`
`12 (35%)
`48%
`
`I1 (35%
`55%
`
`NS (p=0.57)
`
`DISCUSSION
`
`oophorectomy, aromatase inhibitor, or chemotherapy, whereas a sig-
`nificant fall in ER was observed in those treated with tamoxifen.
`
`The expression of ER within the majority of human breast cancers
`is heterogenous (17). Immunohistochemical studies with mAbs to ER
`have identified mixed populations of ER+ and ER- cells in human
`breast carcinomas (18). A potential consequence of prolonged endo-
`crine therapy could be the clonal selection of ER-, presumably
`hormone-insensitive cells from within an originally heterogenous
`ER+ tumor (19). One mechanism for relapse after successful endo-
`crine therapy,
`therefore, might be the emergence under selective
`pressure of tamoxifen of ER- horrnone-resistant tumors.
`Several groups have studied ER content in sequential tumor biop-
`sies after intervening endocrine therapy. Allegra et al. (3) were among
`the first to report that whereas the ER content was similar between
`either multiple metastatic sites or over time without intervening ther-
`apy, a significant fall
`in ER content followed endocrine therapy.
`Taylor et al. (4) showed in 26 patients with advanced breast cancer
`that the ER content of metastatic skin deposits fell in both responding
`and nonresponding patients after 2-3 months of endocrine therapy.
`Hull er al. (5) demonstrated a significant decrease in tumor ER levels
`after tamoxifen, but not in patients in whom the second biopsy was
`taken more than 2 months after discontinuing the drug. However, in
`all of these studies a ligand-binding assay was used to measure ER,
`and it is probable that at the time of the second biopsy, receptor
`occupancy by tamoxifen resulted in a false negative ER assay for
`many tumors.
`Another confounding variable is sequential comparison between
`different metastatic deposits of tumor. In a more recent study, ER was
`measured in the same tumor before and after systemic therapy in 63
`patients with large operable primary breast cancer (20). No significant
`change in ER concentration was seen in those treated with surgical
`
`Table 3 Overall change in frequency of ER, PgR, and p82 expression and mean score
`in 72 paired biopsies taken before and at relapse on tamaxrfen“
`Pretamoxifen
`Tamoxifen relapse
`
`ER+
`No. (%)
`Mean l-I-score:SEM
`
`PgR+
`No. (%)
`Mean H-score .+. SEM
`
`37 (51)
`90:7
`
`27 (38)
`66 2 9
`
`21 (29)
`61:9
`
`20 (28)
`64 1 9
`
`However, again it was concluded that this was due to interference by
`tamoxifen or its metabolites in the ligand-binding assay. The impact
`of such interference was recently demonstrated in a study where ER
`was measured by both ligand-binding assay and IHA in tumors from
`34 patients on tamoxifen (21). ER was detected more frequently by
`immunohistochemical compared with ligand-binding assay, again im-
`plying that receptor occupancy by tamoxifen may interfere with the
`ligand-binding assay.
`From our study it appears that tamoxifen does not interfere with our
`IHA for ER. In the subset of tamoxifen-treated primary tumors,
`comparison with the biochemical EIA (unaffected by ligand interac-
`tion) showed a 96% concordance rate and strong positive correlation
`for the immunohistochemical H-score for ER. These data suggest that
`tamoxifen does not reduce or inhibit ER detection by 1D5 antibody,
`which is targetted against the NH2-terminal end of the receptor, away
`from the ligand-binding region. In the 98 tamoxifen-resistant tumors
`that were studied by both EIA and IHA, lower scores were observed
`by both assays (Fig. 2). The lower concordance rate between the two
`assays may largely be explained by borderline scores on either or both
`assays (Table 2). The technical difference between measuring ER in
`a tumor homogenate and on a histological section means that such
`discrepancies at the detection threshold for each assay are to be
`expected. However, 5 tumors were completely IHA- but clearly
`EIA+. In these tumors, for example, mutations or conformational
`changes within certain ER domains could explain why antibodies
`directed toward the NH,-terminal epitopes (1D5) may not bind,
`whereas antibodies directed towards the COOH-terminal epitopes
`(H222) bind strongly.
`Overall, the frequency and quantitative expression of ER appears to
`be reduced in tamoxifen-resistant
`tumors. The paired comparison
`between the primary and resistant tumor suggest that ER expression
`and function may change in association with certain types of tamox-
`ifen resistance. In the patients treated with primary tamoxifen, the
`clinical response was strongly correlated with ER status; 89% of
`responders compared to only 15% of nonresponders were ER+ at
`presentation. In the biopsy from the same tumor taken at relapse or
`progression on tamoxifen, 61% of the responding group who had
`developed acquired resistance remained ER+, whereas all of the
`nonresponders who progressed on treatment were ER-. In those with
`acquired resistance, paired comparison of the quantitative scores
`showed no significant change (Fig. 3). Futhermore 82% of these
`tamoxifen-resistant tumors that were ER+ at relapse still expressed
`3335
`
`pS2+
`26 (36)
`23 (32)
`No. (%)
`51 2 5
`52 3 5
`Mean percentage positive 1* SEM
`“ ER and PgR scores are the mean H-scores for those positive cases; pS2 scores are the
`mean percentage positive cells in positive cases.
`
`|nnoPharma Exhibit 1066.0005
`
`
`
`
`
`ER, PgR, AND pS2 IN TAMOXII-"EH4-RESlS'l‘AN'I' BREAST CANCER
`
`Table 4 ER expression and function during adjuvant tamoxifen therapy
`
`Site of tumor recurrence
`local breast
`Regional nodes
`Metastatic
`TOTAL
`
`No.
`15
`11
`8
`34
`
`ER+ (%)
`8 (53)
`4 (36)
`6 (75)
`18 (53)
`
`Primary tumor
`ER +
`+
`PgR +
`—
`pS2 —
`+
`1
`2
`1
`0
`2
`3
`4
`5
`
`+
`+
`+
`3
`2
`0
`5
`
`1»
`—
`—
`2
`1
`1
`4
`
`ER+ (%)
`5 (33)
`3 (27)
`2 (7.5)
`10 (29)
`
`Recurrent tumor
`ER +
`+
`PgR +
`—
`pS2
`—
`+
`1
`3
`1
`2
`0
`2
`2
`7
`
`+
`+
`+
`1
`0
`0
`1
`
`+
`—
`—
`0
`0
`0
`0
`
`itive antagonism by tamoxifen, thus supporting tumor regrowth. Clin-
`ically, it is well known that previous objective response to tamoxifen
`significantly increases the chance of response to estrogenic depriva-
`tion (25), and in part this may be explained by maintained expression
`of a functional ER pathway.
`Primary tumors that progressed on tamoxifen with de novo resist-
`ance invariably lacked estrogen receptor. However, six of these ER-
`tumors were found to express PgR and/or pS2 at progression. Com-
`parison with the pretamoxifen biopsy showed that four of these
`tumors had been completely negative for for ER, PgR, and pS2
`(Fig. 6). The staining for the three parameters was performed on
`adjacent 3-p.m sections, such that PgR and pS2 expression was
`detected in a cell population that was immunologically ER-. Whereas
`previous studies have attributed this ER—/PgR+ phenotype in tamox-
`ifen-treated tumors to a false negative ER assay (26), a similar
`observation was recently noted by Encamacion et al. (21) with the use
`of an IHA with 3 different mAbs. They noted 6 of 30 patients with
`documented tamoxifen resistance in whom the tumor was clearly
`ER— by both ligand-binding and IHAs but strongly positive for PgR.
`This phenotype has been found to be associated with increased levels
`of a variant form of ER mRNA in which exon 5 of the ER gene, which
`codes for the hormone binding domain, is spliced out during tran-
`scription (27). In vitro data have shown that transfection of this variant
`mRNA, which codes for a constitutively active truncated ER protein,
`results in high levels of PgR expression and hormone-independent
`growth in MCF—7 breast cancer cells (28). Theoretically, such cells
`might