Expression of Estrogen Receptor (ER) βcx Protein in ERα-positive Breast Cancer: Specific Correlation with Progesterone Receptor¹ Free

Endocrine therapy, such as administration of tamoxifen, contributes significantly to prolonging the disease-free period after breast cancer surgery. Because only about 50% of patients benefit from this therapy, many investigators have tried to define the factors that could predict responsiveness to endocrine therapy (1). ERα³ and PR are still the most reliable markers for choosing treatment with antiestrogens, but about 30% of ERα-positive metastasizing cancers do not respond to tamoxifen (2, 3, 4). The ability to identify this group of nonresponders would be of benefit to both the clinician and the patient.

The second ER, ERβ, was reported in 1996 (5), so its presence was never considered when patient response to tamoxifen was evaluated. ERβ is similar to ERα with approximately 96% and 60% homology in the DNA-binding domains and ligand-binding domains, respectively. The tissue distribution and physiological functions of ERβ and ERα are different (5, 6), and it is thought that ERβ may also have distinct functions in the biology of breast cancer. Several variant forms of ERβ have been reported to date. Among them is ERβcx, a splice variant that utilizes an alternative exon 8. This change in the COOH terminus results in very poor binding to E₂, but ERβcx is capable of heterodimerization with ERα and has a dominant negative effect on ERα function (7, 8). Because the PR gene is one of the representative downstream targets of ligand-activated ERα (6), it seems likely that coexpression of ERβcx and ERα could affect the expression of PR in breast cancer tissues, and this could be one of the conditions that would lead to ERα-positive and PR-negative tumors.

With a specific antibody directed against exon 8 of ERβcx, we performed a detailed analysis of the expression of ERα, PR, and ERβcx in breast cancer.

ERβ expression vector, pRc/CMV-ERβ, was described previously (9). ERβcx expression vector, pRc/CMV-ERβcx, was constructed by recombination of the exon 8 sequence of pRc/CMV-ERβ with the cx sequence amplified by RT-PCR.

MCF-7 cells were transfected with ERβcx expression vector with TransIT LT-1 reagent (Takara, Otsu, Japan). After 1 day in culture, the cells were grown in fresh RPMI 1640 supplemented with 10% FCS containing 1 mg/ml G418 for 10 days. Isolated colonies were trypsinized in metal ring cups, and the cells were further cultured in the presence of 200 μg/ml G418.

Western blot analysis was performed as described previously (9). Briefly, aliquots of 100 μg of cell extract, which was taken from cells stimulated with E₂, were subjected to SDS-PAGE in 10% acrylamide gels, and proteins were transferred to a polyvinylidene difluoride membrane (Millipore, Bedford, MA). Blots were probed with primary anti-ERα mouse polyclonal antibody (C-311; Santa Cruz Biotechnology, Santa Cruz, CA), anti-ERβ rabbit polyclonal antibody (Upstate Biotechnology, Lake Placid, NY), anti-PR mouse monoclonal antibody (633; Dako, Kyoto, Japan), and anti-ERβ rabbit polyclonal antibody (H-300; Santa Cruz Biotechnology). The secondary antimouse or antirabbit antibodies (Bio-Rad Laboratories, Hercules, CA) were conjugated with alkaline phosphatase. Detection was performed using Immun-Star Substrate (Bio-Rad Laboratories) and a Fuji Luminoimage Analyzer LAS-1000 system (Fuji Film, Tokyo, Japan).

Human breast cancer samples were obtained from patients undergoing partial or total mastectomy in Tokyo Metropolitan Komagome Hospital with the informed consent about the usage of resected tumors for research purposes. The ERα content of samples resected during 1999–2000 was routinely evaluated by immunohistochemistry according to the following protocol.

Samples with neoadjuvant tamoxifen treatment were obtained as follows. Before the start of drug administration, core needle biopsies were taken from patients who entered the clinical trial for preoperative tamoxifen treatment. All patients were treated with 20 mg of tamoxifen daily for 3 months and subsequently had surgery. Histological factors were evaluated by pathologists at the National Cancer Center Hospital (Tokyo, Japan). The response of the primary tumors to tamoxifen was evaluated according to criteria established by the Japanese Breast Cancer Society, which are essentially the same as those established by the WHO. Complete response is defined as the disappearance of tumor, partial response refers to a decrease in tumor size of ≥50%, NC indicates a decrease in tumor size of <50% or an increase of tumor size by <25%; PD indicates an increase in tumor size of ≥25%.

The formaldehyde-fixed, paraffin-embedded samples were sequentially cut into 4-μm sections for staining. Antigens were retrieved by boiling the sections in 5% urea buffer for 20 min. The primary antibodies used were 1D5 for ERα, PgR636 for PR (Dako), and ERβ 14C8 (GeneTex, San Antonio, TX) for both the wt and cx forms of ERβ. The amino acid sequence of the peptide used to raise the ERβcx-specific antibody was CMKMETLLPEATME. The peptide was coupled to keyhole limpet hemocyanin by the cysteine at the NH₂ terminus to ensure a much better chance of obtaining a strong immune response. Preimmune serum taken from the sheep before immunization was used as a negative control for the antibody. Samples in which ERβcx mRNA was detected by RT-PCR (data not shown; Refs. 10 and 11) were used as positive controls (Fig. 2, A and B). Tumors in which ERβcx mRNA was not detectable by RT-PCR (data not shown) were used as negative controls for the staining (Fig. 2 C).

Evaluation of staining in entire lesions of 115 tissues and in core needle biopsy samples was done according to the Allred score (12). Briefly, a PS was assigned that represents the estimated proportion of positive tumor cells on the entire slide as follows: none = 0; 1 of 100 = 1; 1 of 10 = 2; 1 of 3 = 3; 2 of 3 = 4; and 1 of 1 (i.e., all of the cells are stained) = 5. An IS is assigned that estimates the average staining intensity of positive tumor cells as follows: negative = 0; weak = 1; intermediate = 2; and strong = 3. The PS and IS are added to obtain a total score [range, 0–8 (Fig. 2, L–N, and Fig. 3)].

Our initial hypothesis was that coexpression of ERβcx and ERα would lead to inactivation of ERα, and this would result in lower expression of PR. To determine whether expression of ERβcx affects the expression of PR in ERα-positive human breast cancer cells, we established two clones of MCF-7 cells (MCF-7/βcx-1 and MCF-7/βcx-2) by stable transfection of the ERβcx-expressing vector. Expression of ERβcx in these transfected cells was monitored by Western blotting with an antibody that recognizes the NH₂-terminal portion of ERβ. This is a region present in both wt ERβ and ERβcx. There was increased intensity of the ERβ bands on Western blotting in transfected cells (Fig. 1). Because it is not possible to have wt ERβ synthesized from ERβcx expression vector, the increase in the NH₂-terminal signal indicates an increase in ERβcx.

When MCF-7/βcx-1 and MCF-7/βcx-2 were compared with the parental MCF-7 cells or with cells transfected with the CMV vector only, it was evident that the amount of PR (PR-A + PR-B; Ref. 13) was reduced in the ERβcx-expressing cells (Fig. 1). In detail, PR-A was obviously reduced, but PR-B seemed to be slightly up-regulated.

To determine whether the findings detected in cell clones hold true in clinical breast cancer, we proceeded to analyze the tissues obtained from breast cancer surgery.

First, for analysis of the expression of ERβcx protein on paraffin-embedded specimens, we developed a specific ERβcx antibody that was raised against the unique COOH-terminal sequence of ERβcx as described in “Materials and Methods.” In samples in which ERβcx mRNA was confirmed to be present by RT-PCR (data not shown; Ref. 10), this antibody detected nuclear localized signals with slight cytoplasmic staining (Fig. 2,A). No such signals were seen with preimmune serum (Fig. 2,B). In samples that had no ERβcx mRNA, there was cytoplasmic but not nuclear staining (Fig. 2,C). Commercially available mouse monoclonal ERβ antibody, 14C8, which should bind both wt ERβ and ERβcx, showed intense nuclear signals (Fig. 2,D) on the same samples used in Fig. 2, A and B. ERβcx antibody also detected nuclear signals with slight cytoplasmic staining on the same area of a sequentially cut slide (Fig. 2 E). From these findings, the nuclear staining with the ERβcx antibody was considered positive for ERβcx.

A total of 115 individual human breast cancer samples, consecutively obtained during 1999–2000, were evaluated as ERα positive by routine immunohistochemical analysis with standard criteria (10% of total cells were positive). Of these, 54 specimens were chosen for the initial experiment because these sections each had an ERα-rich focus (a single independent component consisting of about 500-2000 cancer cells in which ERα was expressed in >80% of the cells as shown in Fig. 2, F and I). Staining and evaluation of these selected foci made it possible to see whether the expression of ERβcx correlates with lack of PR in an almost homogenous ERα-positive group of cells.

Three sequential slides were stained with ERα, ERβcx, and PR antibodies. Fig. 2, I–K, shows representative cases of ERβcx-negative tumors. In the areas that were ERα rich (Fig. 2,I) with only cytoplasmic ERβcx staining (Fig. 2,J), almost all cells were PR positive (Fig. 2,K). However, in fields where there is ERα (Fig. 2,F) and nuclear ERβcx staining (Fig. 2,G), there is no evidence of PR (Fig. 2,H). A summary of the 54 fields examined is shown in Table 1. Not all cases were homogeneously stained with ERβcx and PR antibody as seen in the sample illustrated in Fig. 2, F–K. For this reason and to analyze the data statistically, arbitrary cutoff limits were used. For PR, 10% of positivity was used as the cutoff value. With ERβcx, only the cases with >60% positive staining were categorized as “positive” in Table 1. With these criteria, 14 of 54 foci examined were evaluated as positive for ERβcx.

The presence of ERβcx significantly correlated with the absence of PR staining (P < 0.01, Fisher’s exact probability test). However, in the absence of ERβcx, expression of PR varied.

The results seemed to indicate that the presence of ERβcx in a tumor would be synonymous with ERα-positive and PR-negative breast cancer. However, this relationship was detected only in limited parts of breast cancer, where ERα was abundantly expressed. As is widely known, human breast cancers are generally heterogeneous and are composed of many various cells. We therefore examined the expression of ERβcx in entire lesions in breast cancer specimens and analyzed these data with the clinical information of the patients. We used the Allred score, which is widely used for the evaluation of hormone responsiveness of clinical breast cancer (12).

In total, 115 breast cancer samples were stained with ERα, ERβcx, and PR antibodies, and the proportion and intensity of staining on the entire lesion were scored by the Allred method. Sample results of ERβcx staining are shown in Fig. 2, L–N. An evaluation was made as to whether any characteristics of the patient’s disease could be correlated with the presence or absence of ERβcx (Table 2). Venous invasion of cancer cells was significantly correlated with a ERβcx-negative phenotype, but there was no statistically significant correlation between expression of ERβcx and other factors including PR.

To examine the relationship between ERβcx and PR more closely, Allred scores of ERβcx and PR from individual patients are presented as a scatter plot in Fig. 3 A. The dotted line on the graph indicates the cutoff value decided from the mean values of each receptor’s Allred scores. It is clear from the graph that there are two distinct groups of patients. One group expresses ERβcx and has very reduced levels of PR expression, and the other group expresses both ERβcx and high levels of PR.

From our results, the relationship between ERβcx and PR in clinical breast cancer is not as simple as was found in breast cancer cell lines. To understand the clinical meaning of the results depicted in Fig. 3,A, we acquired biopsies from patients who entered a clinical trial in which they were treated with tamoxifen as their primary therapy before surgery. These samples are rare because this type of treatment is not commonly applied. By evaluation of the response of the primary tumor to tamoxifen, these samples make it possible to evaluate whether the presence of ERβcx and PR affects the response to tamoxifen. Eighteen core needle biopsies from individual patients could be evaluated by immunohistochemistry for ERα, ERβcx, and PR. All samples were rich in ERα expression, as evaluated as 5–8 total points in Allred score. Fig. 3,B shows the distribution of ERβcx and PR scores. Open circles indicate NC or PD (i.e., patients whose primary tumor did not respond well to tamoxifen). The important findings from Fig. 3 B are as follows: (a) eight of nine patients (89%) lacking ERβcx responded well to tamoxifen, although in ERβcx-positive cases, only four of nine (44%) patients responded; and (b) none of the four patients who were ERβcx positive but PR poor showed any response to tamoxifen.

The absence of PR in ERα-positive cancers generally implies that ERα is not active in vivo (4). This inactivity could be due to defects in the receptor itself or to the absence of estrogen. Because mutations or defects in ERα are not common in human breast cancer (14), loss of PR is generally thought to be due to the lack of estrogen. Recent use of an aromatase inhibitor in postmenopausal patients showed that PR-positive cancers sometimes become PR negative after inhibition of aromatase (15). This result suggests that lack of ligands is one important source of the ERα-positive/PR-negative phenotype, although this is not true in all cases.

MCF-7/βcx-1 and MCF-7/βcx-2, which are MCF-7 cells stably expressing ERβcx, showed reduced expression of PR (PR-A + PR-B) without any effect on the level of ERα. These results, together with the findings in clinical samples shown in Table 1 and Fig. 1, F–H, indicate that in addition to the availability of ligands, expression of ERβcx may be another source of the ERα-positive/PR-negative phenotype in a cancer cell.

PR-A and PR-B are products of a single gene transcribed from different promoters. PR-B functions as a transcriptional activator on progesterone-responsive promoters, whereas PR-A acts as a repressor of the function of PR-B and ERα (13). The consequences of the expression ratio of these PRs in breast cancer have not been defined in terms of their contribution to the malignant phenotype and to responsiveness to endocrine therapy (16). In ERβcx transformants, PR-A was dominantly inhibited, but PR-B seemed to be slightly up-regulated. The difference in regulation of PR-A and PR-B by introduction of ERβcx should be investigated in more detail in additional experiments, including an assessment of whether this effect is really through the interaction with ERα or through the action of ERβcx on the promoter of the PR gene.

Mote et al. (17) reported that, with formalin-fixed samples, commonly used antibodies frequently failed to detect PR-B in immunohistochemical analysis. This is thought to be due to masking epitopes on PR-B protein. If this is the case, the loss of PR in our immunohistochemical studies must mean mainly loss of PR-A.

Predicting the response to tamoxifen in breast cancer patients is difficult for the clinician. Although this drug has been widely used as an adjuvant after breast cancer surgery, it is generally used along with chemotherapy (1). Patients who are given tamoxifen alone are usually lower-risk patients and rarely relapse. It is therefore unusual to obtain appropriate samples from poor responders. We were fortunate to obtain core needle biopsies from tumors before treatment with preoperative tamoxifen. Clinical evaluation of primary tumor size after 3 months of treatment allowed us to find poor responders. Although the sample size is small, our analysis revealed that tumors that had ERβcx staining, especially with poor expression of PR, did not seem to be good candidates for tamoxifen treatment. As a supporting finding, we have observed that MCF-7/βcx clones lost the ability to grow in response to estrogen, and tamoxifen could no longer inhibit growth (detailed data will appear elsewhere⁴ with molecular analysis).

In contrast to this type of ERβcx-positive tumor, we also found by evaluation with Allred score in 115 clinical samples that there are ERβcx-positive tumors that are also rich in PR. Because this group showed good response to tamoxifen in core needle analysis (four of five tumors responded), it is possible that the level and function of ERα in these tumors are high enough to overcome the presence of ERβcx, although current experiments were not designed to test this. In addition to this, our initial hypothesis was that expression of ERβcx, if expressed at high levels and coexpressed in breast cancer cells with ERα, can reduce PR levels in those cells. Therefore, the presence of ERβcx does not necessarily permit the conclusion that ERα is repressed. If ERα and ERβcx are not coexpressed, PR can still be expressed.

The opposite concern in the ERβcx-positive/PR-negative group is that the amount and distribution of ERβcx may not be sufficient to silence the widely distributed ERα in some cases. Because the presence of ERβcx seemed to alter the ratio of PR-A:PR-B (Fig. 1), ERβcx may influence PR promoter usage. It is therefore valid to begin to think of ERβcx as a transcriptional regulator in its own right, independent of ERα. If this is the case, ERβcx itself may alter the expression of PR, even when it is expressed at low levels. Of course, we cannot exclude the possibility that expression of ERβcx is more widespread than can be detected by present immunohistochemical methods.

As is often experienced, evidence obtained in cell culture studies does not fully reflect what happens in tissues. However, we believe that the results from our cell culture experiments offer an explanation for why patients whose tumors are ERβcx positive and PR negative do not respond to tamoxifen, whereas those with ERβcx-positive and PR-positive tumors do.

In this study we have presented the first evidence for the presence and pattern of expression of ERβcx protein in clinical ERα-positive breast cancers. We have also found that expression of ERβcx influences PR expression. This information may be useful for identifying patients who will respond to tamoxifen treatment. Further analysis in collaboration with large clinical institutes, which can provide samples in the neoadjuvant setting, will be required to define the clinical significance of our findings.

We thank Makiko Hirose for skillful assistance in these experiments and Prof. Shigetoyo Saji for helpful suggestions.