Abstract
Identification of proteins that markedly vary during early steps of mammary carcinogenesis may help to understand its pathophysiology and to develop a prevention strategy. The expression of total estrogen receptor β (ERβ) protein and of its COOH-terminally spliced variant ERβcx (or ERβ2) was compared in 43 invasive breast cancers and in 39 adjacent normal mammary glands and 26 ductal carcinoma in situ (DCIS). Thirty-six breast cancers were ER positive by radioligand binding assay. The analysis was done by immunohistochemistry on adjacent sections of formalin-fixed, paraffin-embedded tumors using polyclonal anti-ERβ 503 IgY and sheep polyclonal ERβcx antibodies that were previously validated. Nuclear staining was quantified using a computerized image analyzer in selected areas of normal and cancer epithelial cells. Total ERβ expression was high in normal glands, decreased in DCIS (P = 0.0004), and increased from DCIS to invasive tumors (P = 0.029). In contrast, the ERβcx expression was low in normal glands, increased significantly in DCIS (P = 0.0014), and continued to increase in invasive carcinomas (P = 0.0027) in both ERα-positive and ERα-negative tumors. This is the first study showing a significant increase of the ERβcx variant protein in DCIS and invasive breast cancer compared with adjacent normal glands. This contrasts with the decrease of the total ERβ level in the same patients and indicates different mechanisms to explain these variations during mammary carcinogenesis. It also suggests a role of the ERβcx variant in carcinogenesis opposite to the protective effect of the wild-type ERβ1.
- Estrogen receptor β
- splicing
- normal glands
- breast cancer
- ductal carcinoma in situ
- immunohistochemistry
- carcinogenesis
The pathophysiology of sporadic human breast cancer remains poorly understood, although the facilitating role of ovarian hormones seems to be critical (1). The multistep model of carcinogenesis established for colorectal cancer (2) might also be valid in understanding mammary carcinogenesis based on the increased risk of proliferative lesions and of in situ carcinoma as defined by Dupont and Page (3). One approach to understand the first steps of human mammary carcinogenesis is to compare the expression of putative oncogenes and tumor suppressor genes in normal glands, high-risk benign breast disease, and in in situ carcinoma. In addition to early genetic alterations (4), the analysis of protein expression by immunohistochemistry could be decisive. Among the proteins involved in sporadic human breast cancers, those related to estrogen action are good candidates; estrogens are known to be tumor promoters via their mitogenic activity (1). The balance between the levels of the two estrogen receptors, ERα and ERβ (5), their spliced variants (6), and of nuclear receptor coregulators may be important in modulating the mammary tumor promoter effect of estrogens. ERβ is not required for ductal development in mammary gland (7) but it inhibits the activity of ERα after heterodimerization (8, 9). Whereas the level of ERα begins to increase in proliferative mammary gland with atypia (10, 11), the expression of total ERβ begins to decrease in proliferative benign breast disease without atypia to become the lowest in ductal carcinoma in situ (DCIS; ref. 12). The two receptors are differently spliced leading to several variants. The significance of ERα variants in carcinogenesis is not clear but few studies were made at the protein level (reviewed in ref. 13). Two spliced variants of ERβ1 (ERβ2 or ERβcx) and ERβ5 mRNA are expressed in breast cancer (14, 15). ERβcx also expressed as a protein (15, 16) does not bind estrogens but inhibits both ERα and ERβ transcriptional activity by heterodimerization (17).
Using polyclonal antibodies specific to the human ERβcx and to the total amount of ERβ (18), which have been validated previously (12, 18, 19), we have compared by computer-aided immunohistochemistry the expression levels of these proteins in epithelial cells of invasive breast carcinoma, DCIS, and normal mammary glands located in the same section.
Materials and Methods
Patient population. Forty-three primary breast carcinomas, removed by surgery in 1992 at the Montpellier Cancer Center, were selected from the archival files because they contained normal ducts and/or lobules and/or DCIS at the periphery of the tumor. Thirty-six invasive breast cancers were ER-positive by radioligand binding assay and were previously studied (19). Seven additional tumors were chosen as ER negative by immunohistochemistry (ERα antibody, clone 6F11, Novocastra, Newcastle, United Kingdom) and radioligand assay.
The median age of patients was 62 years; 77% were postmenopausal as determined by clinical and hormonal analysis. Ninety-one percent of invasive breast cancers were ductal and 9% were lobular. According to a modified Scarff-Bloom-Richardson grading system (19), 19% of tumors were of grade 1, 36% grade 2, and 45% grade 3. Tumor size status was pT1 (44%), pT2 (49%), and pT3/pT4 (2%). Histologic node invasion was in 51% of cases. The median ER cytosolic level, by radioligand assay, was 59 fmol/mg protein (range 0-441), with seven cases with ER < 10 fmol/mg protein. The median progesterone receptor cytosolic level was 55 fmol/mg protein (range 0-576). Ten DCIS structures were of low nuclear grade, 12 of intermediate grade, and 4 of high grade.
Immunohistochemistry. Immunohistochemistry was done using chicken polyclonal ERβ 503 IgY antibodies, which recognize total ERβ proteins (both full-length ERβ and its splice variants), and with the sheep polyclonal ERβcx antibody, raised against the 14 amino acids peptide of the COOH-terminal region: MKMETLLPEATMEQ (from the laboratory of J-A. Gustafsson).
Immunohistochemical study was done on alcohol–formalin-fixed paraffin-embedded sections as described (12, 19). Briefly, the avidin-biotin-peroxidase complex method was applied after a heat-induced antigen retrieval by pressure cooking in EDTA buffer (pH 7) for 15 minutes. 3,3′-Diaminobenzidine was used as chromogen. ERβcx specificity of immunostaining was shown previously (19) by preincubating the ERβcx antibody with a 10-fold excess of ERβcx peptide and with preadsorbed ERβcx antiserum. The immunostaining specificity with ERβ 503 IgY antibodies was shown previously by protein extinction (12, 19). In each experiment, a negative control was done with IgY antibodies from nonimmunized serum (Nordic, Tilburg, the Netherlands) and with preimmune sheep serum for ERβcx. Positive external controls were used in sections of OVCAR cells embedded in paraffin pellet for ERβ and in a breast cancer paraffin block expressing ERβcx.
HER-2/neu protein level was estimated using polyclonal A0485 (DAKO) and monoclonal CB11 (Novocastra) antibodies, with the automated DAKO Autostainer according to the manufacturer's instructions and the DAKO Hercept Test scoring.
Quantitative method. Quantification was done using a computerized image analyzer (Samba 2005 TITN Alcatel, Grenoble, France; refs. 12, 19). This is an objective automatic method of quantification, guided by the pathologist who selected fields containing only epithelial cells (normal or pathologic), with a program quantifying only nuclear staining. Staining of stromal and inflammatory cells was not quantified. Serial, stained slides were analyzed for ERβ and ERβcx. Representative fields of invasive carcinoma and adjacent components of normal and DCIS were analyzed. Results were expressed as the percentage of nuclear-stained epithelial cells and as a quantitative immunocytochemical (QIC) score [= (percentage of surface stained in epithelial cells) ×1 (mean staining intensity) × 10]. The percentage of nuclear staining of negative control, which was usually nil or weak, was subtracted.
Statistical analysis. Quantitative ERβ and ERβcx expressions for pooled comparisons of the three structures were done with the Kruskal-Wallis nonparametric test. The Wilcoxon sign rank test was used for comparisons between paired samples, containing the same components at the periphery of tumor. P values <5% were considered statistically significant.
Results
ERβcx staining distribution. ERβcx expression varied between the patients and from normal tissue to breast cancer in the same patients. Intense and diffuse ERβcx nuclear immunostaining (in brown) was seen in invasive carcinomatous cells (Fig. 1). It contrasted with the weak staining of a few luminal and basal epithelial cells of normal mammary ducts (Fig. 1A and B) or lobules (Fig. 1C). The ERβcx expression in in situ carcinomas was variable, but generally intermediate between normal and invasive carcinoma as shown in two examples (Fig. 1D-F). ERβcx immunoreactivity was also seen in some stromal cells, lymphocytes, and macrophages (Fig. 1). However, these cells were not selected and, therefore, not quantified by the image analyzer. A weak or moderate cytoplasmic staining was considered as nonspecific because, contrary to nuclear staining, it was not eliminated by adding the peptide or using preadsorbed serum (12, 19).
Differential expression of the ERβcx protein. Nuclear positive ERβcx staining (brown) was significantly higher in invasive breast cancer (red arrows) than normal duct (*; A and B) or normal lobule (*; C) where nuclei were mostly blue. In DCIS (blue arrows), the nuclear ERβcx expression was either clearly lower (D) than in adjacent invasive breast cancer (D, inset) or only slightly lower (F) than in adjacent invasive breast cancer (E) as shown in two typical examples. DCIS staining was, however, always higher (D and F) than in normal glands (A-C). ERβcx was also expressed in some stromal and inflammatory cells (black arrows). Magnification, ×200 (A, D, E, F); ×400 (B-C). ERβcx staining specificity was shown previously (19).
Comparison of total ERβ and ERβcx expression levels between normal glands, ductal carcinoma in situ, and invasive carcinomas. When comparing pools of normal tissue and lesions (Fig. 2A; Table 1A), the total ERβ level (percentage of stained nuclei) decreased significantly from normal glands to DCIS (P < 0.0001), confirming our data from another series (12). Total ERβ level also decreased, but to a lower extent, from normal glands to invasive breast cancer (P = 0.002) because it increased from DCIS to invasive breast cancer (P = 0.042). The ERβ variant responsible for this increase was not defined. ERβcx level evaluated in adjacent serial sections was found to vary in opposite direction from normal gland to DCIS than total ERβ (Fig. 2B; Table 1A). ERβcx increased gradually from normal glands to DCIS (P < 0.0001) as well as to invasive breast cancer (P < 0.0001). ERβcx levels increased to a lower extent from DCIS to invasive breast cancer.
Comparison of ERβ (A) and ERβcx (B) levels in adjacent sections between three structures: invasive breast cancer, peripheral normal glands, and/or DCIS. Expression levels were estimated by immunohistochemistry and a computer-aided image analyzer. n = number of each pooled structures. The median value of each pooled structure is represented by a bar (). The different structures in the same tumor section are connected for ER-positive tumors: (- - -, ▵) and for ER-negative tumors: (—, •). The HER-2/neu–positive tumors are indicated by arrows. The P values, determined by the Wilcoxon rank test, indicate the statistical difference between paired structures.
Comparison of ERβ and ERβcx levels in adjacent normal glands (N), ductal carcinomas in situ (DCIS) and invasive breast carcinomas (IBC) in pools of each component (A) and in paired cases (B)
When comparing the pairs of two different structures in the same tumor sections, as shown in Fig. 2 and summarized in Table 1B, the dissociated evolution of ERβ and ERβcx levels from normal to in situ carcinoma and invasive breast cancer were also significantly different.
Similar variations were observed by using the QIC score, which include the staining intensity of nuclei with a significant increase of ERβcx expression from normal glands to invasive breast cancer for the whole population, confirming examples of Fig. 1.
Interestingly, the increase of ERβcx expression was also observed in five of seven ERα-negative invasive breast cancer, suggesting that it was independent of estrogens binding to ERα.
Only three cases of invasive breast cancer, which were ER negative by immunohistochemistry and radioligand binding assay (Fig. 2) and their adjacent high nuclear grade DCIS overexpressed HER-2/neu protein. The ERβcx level in invasive breast cancer was independent of HER-2/neu or ERα-positive status but two ERα-negative and one HER-2/neu positive invasive breast cancer strongly expressed ERβcx (Fig. 2B). In a previous study (19), we showed that expression of ERβ and ERβcx in breast cancer was not correlated to any classic parameters; the only correlation found was between ERβ and ERβcx at the stage of invasive breast cancer.
Discussion
In this pilot study, we confirm on another series of samples that the expression of total ERβ protein significantly decreases from normal glands to DCIS (12). However, total ERβ slightly increases from DCIS to invasive cancer. This contrasts with another study (20), suggesting that total ERβ expression decreased in invasive carcinomas but not in DCIS. The reason for this difference is unknown, as well as the nature of the variant that is responsible for the increase in our study. The major new information of our study concerns the evolution of the ERβcx variant that is generally low in normal glands but highly variable in cancer cells with a significant and progressive increase in DCIS and in invasive breast cancer. The dissociated evolution of ERβ and its spliced form, ERβ2, indicates different mechanisms dysregulating the expression of these variants in carcinoma cells and suggests different significance of the two ERβ variants in carcinogenesis.
There are few studies of ERβcx in breast cancer and the significance of its expression is not clear because ERβcx has been shown to have both good (21) and bad (15) prognostic significance. Studies comparing ERβcx in normal mammary glands and invasive breast cancer were mostly done at the RNA level. The increased expression of ERβcx in breast cancer was proposed to be due to inflammatory cells infiltrating the tumor (14). A study showed an increase in the expression of ERβcx protein in invasive breast cancer (54%) compared with a pool of normal tissue (only 9% positive; ref. 15). Our study confirms this finding by comparing the two types of tissue in the same tumor sections and by quantifying only epithelial and cancer cell expression, excluding interference from inflammatory cells and fibroblasts. Furthermore, we show that the increased ERβcx protein expression is already observed in DCIS, suggesting that it may be an early and critical event in mammary carcinogenesis. A similar increase of the ERβcx variant was observed in prostate cancer (22). The function of the two forms of ERβ in breast cancer may be different. Following separate transfection of their cDNA into MCF7 cells, the genes regulated by ERβ2 or ERβ1 are different (23). The association of the decreased expression of ERβ with the increased expression of ERβcx and ERα might lead to an increased sensitivity of transformed mammary cells to estrogen, allowing them to be growth stimulated after menopause by lower estrogen concentrations. It is, therefore, tempting to speculate that the increase of ERβcx, associated with a decrease of ERβ, facilitates mammary carcinogenesis. Increased splicing variants have been described in several cancers (24), including mammary tumors (25). It is frequent that alternative splicing inactivates the wild-type tumor suppressor genes in carcinogenesis (24) and ERβ has been proposed to behave as a tumor suppressor by inhibiting the estrogen mitogenic activity (8, 12, 26). Even if the increased ERβcx/ERβ1 ratio has not yet been proven to facilitate carcinogenesis, it might be very useful as a marker to detect high-risk lesions and facilitate early diagnosis of transformation. The mechanism responsible for the modification of ERβ splicing in carcinogenesis is unknown, but might be specific because the Δ5-6–deleted ERβ variant RNA was specifically decreased in breast cancer compared with normal glands (27). Further studies are required to specify the possible involvement of SR proteins (28), splicing site mutations, or nuclear receptor comodulators controlling transcription and splicing (29). Moreover, differential methylation of the ERβ1, but not the ERβcx promoter gene, might explain the selective decreased ERβ1 level during tumorigenesis (20, 30).
To conclude, this pilot retrospective study points to a differential dysregulation of ERβ1 and ERβcx in in situ mammary carcinoma. The intriguing evolution of this ERβ1/ERβ2 ratio during early steps of mammary carcinogenesis will require more extensive and prospective studies to elucidate several questions concerning its mechanism and significance in the pathophysiology, early detection, and monitoring of breast cancer.
Acknowledgments
We thank Drs. Philippe Rouanet, Bernard Saint-Aubert, Jean Grenier, Frederic Bibeau, François Quenet, and G. Romieu (CRLC Val d'Aurelle, Montpellier, France) for supplying clinical data; Jean-Yves Cance for the skillful preparation of the figures; and Roselyne Lavaill for her excellent technical help.
Footnotes
-
Grant support: Institut National de la Sante et de la Recherche Medicale, Ligue Nationale Contre le Cancer, and Académie Nationale de Médecine (M. Esslimani-Sahla); and grants from Swedish Cancer Society and KaroBio AB (J-A. Gustafsson).
-
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.
- Accepted February 4, 2005.
- Received November 10, 2004.
- Revision received February 3, 2005.