The Normal Breast Epithelium of Women with Breast Cancer Displays an Aberrant Response to Estradiol

Investigation of the hormonal etiology of breast cancer has traditionally focused on the observation of lifetime events that determine estrogen exposure or on the measurement of serum hormone levels. These studies have clearly shown that cumulative lifetime exposure to endogenous (and perhaps exogenous) estrogen is one of the determinants of breast cancer risk (1); the cellular mechanism for this effect is unclear, but it probably involves the induction of proliferation in breast epithelium (2, 3). Higher serum estradiol levels have also been observed in breast cancer cases than in controls (see Ref. 4 for a review). We have instead chosen a strategy of examining the characteristics of the at-risk target tissue (i.e., normal breast epithelium) and have compared the characteristics of breast epithelium from breast cancer cases, who are at high risk of future second breast primary tumors, to epithelium from benign disease controls, in a population where data on established breast cancer risk factors was prospectively collected.

Given the obligate role of ER² in estrogen response, and the fact that steroid receptor content appears to limit cellular response to steroids (5, 6), we chose to first investigate the ER-α and PgR content of the normal epithelium from women with and without breast cancer. We observed that ER positivity is significantly more frequent in breast cancer cases than in women who do not have breast cancer, but PgR positivity was uniformly high in both groups; these findings have been reported previously on a total population of 376 women (7).

The most important known regulator of the ER is its ligand; estradiol has been shown to down-regulate its receptor in breast cancer cell lines and in the murine mammary gland (8, 9, 10). Studies of normal human breast epithelium from menstruating women and those on oral contraceptives have shown that ER levels vary inversely with estrogen levels, i.e., they are high in the follicular phase, and low in the luteal phase (11, 12, 13). After our initial analysis of breast epithelial ER expression in breast cancer cases and controls (14), we began collecting venous blood samples at the time of surgery to measure simultaneous serum estradiol and progesterone levels and relate these levels to breast epithelial receptor and proliferation levels. We now report data on 121 women, relating serum estradiol and progesterone levels to breast epithelial ER and PgR expression and cell proliferation. These patients comprise a subset of the population of 376 women in whom we have previously analyzed the occurrence of breast cancer in relation to ER expression. The study subjects included in the present report are those on whom matched serum hormone and breast epithelial receptor data were available. Our hypothesis was that high-risk women (i.e., cases, in our population) might demonstrate alterations in breast epithelial response to estrogen. Such alterations in estrogen response may augment the effects of greater estrogen exposure and better explain the associated risk for the development of breast cancer. From the perspective of breast cancer prevention, altered target organ responsiveness may mean that a reduction in estrogen exposure by diet, exercise, or other means needs to be accompanied by strategies to blunt breast epithelial response to estrogen.

Our study subjects were recruited from the women seen at the Breast Care Center at University Hospital (Syracuse, NY). Written informed consent was obtained from women who required breast surgery, according to a protocol approved by the institutional review board. Participating patients agreed to donate a sample of grossly normal breast tissue from the surgical specimen and a venous blood sample drawn at the time of operation. Epithelial samples included in the study showed either morphologically normal epithelium or minimal nonproliferative benign change. Details of recruitment parameters and methods of pathological assessment have been described previously (7).

ER and PgR immunohistochemistry was performed on cryostat sections using the ER-ICA and PgR-ICA kits according to the manufacturer’s instructions. Briefly, freshly obtained breast tissue was embedded in OCT (Miles, Elkhart, IN), snap-frozen, and stored at −120°C. Cryostat sections (5 μm thick) were stained with H&E to screen for the presence of an adequate sample of normal and nonproliferative epithelium (more than 10 acinar or ductal structures). Sections were fixed, and primary antibodies were applied (ER antibody H222 and PgR antibody KD68) for 30 min as per the manufacturer’s instructions, followed by bridging antibody (goat antirat), peroxidase/antiperoxidase complex, a PBS wash, and chromagen (diaminobenzidine). A hematoxylin counterstain was used.

Ki-67 staining was accomplished as follows: paraffin sections were cut at 5 μm and dried overnight at 37°C, deparaffinized, and hydrated, followed by a methanol/peroxide block. The slides were microwaved in a citrate buffer antigen retrieval solution, cooled, blocked with normal goat serum and incubated with primary antibody (Mib-1), followed by biotinylated goat antimouse IgG and peroxidase-conjugated streptavidin label. Diaminobenzidine was used for color development, followed by a hematoxylin counterstain. LIs for ER, PgR, and Ki-67 were calculated by counting an average of 15–20 epithelium-containing fields (a total of 2000 cells), as described previously (7).

To address the possibility that ER, PgR, or Ki-67 is differentially expressed in ducts and lobules, we estimated the ductal and lobular component of the epithelial sample on the H&E-stained sections of each sample analyzed in the study. This was accomplished by counting the number of ×10 fields occupied by epithelial tissue on each section and recording whether these were occupied by ductal or lobular structures or a combination of both. Each tissue sample was therefore assigned one score for ductal structures and one score for lobular structures. The ratio of lobular:ductal area was included in the regression analyses to assess the contribution of lobular versus ductal counts to the ER, PgR, and Ki-67 labeling scores.

Blood samples were obtained immediately before surgery or during the procedure, allowed to clot, and spun. Serum was divided into 100 μl aliquots and stored at −80°C. Total estradiol and progesterone were measured by standard RIA (Coat-A-Count; Diagnostic Products Corp., Los Angeles, CA). This is based on antibody-coated tubes supplied by the manufacturer, to which ¹²⁵I-estradiol or progesterone and patient serum are added; the tube is vortexed, incubated for 3 h at room temperature, and decanted. The estradiol bound to the antibody-coated tubes is then measured by scintillation counting of the decanted tubes. Controls included duplicate runs, inclusion of samples from previously run assays, and serum standards provided by the manufacturer. Serum samples were batched chronologically according to dates of acquisition, and both case and control samples were included in each batch.

The distribution of the variables was explored for normalcy and skewness and to assess possible outliers or coding errors. Because the indices for both receptors and Ki-67 labeling were positively skewed, median values were used to summarize the data, and the Kruskal-Wallis test was used to look for differences in the medians. The relationships between hormone levels and LIs for ER and PgR and Ki-67 labeling were examined by univariate and multivariate linear regression, separately for cases and controls. The differences in regression slopes for a given relationship between cases and controls were tested for significance by the inclusion of an interaction term in the model. Adjustment for age and lobular/ductal composition of the tissue was performed by including these parameters in each linear regression model.

The mean age of the 121 women forming the final study population was 45.6 years; of these, 77 were premenopausal (21 cases and 56 controls), and 44 were postmenopausal (28 cases and 16 controls). Table 1 shows median hormone, ER, PgR, and Ki-67 labeling data for these four groups. ER and PgR labeling was confined to epithelial cells, and staining of stromal cells was not observed. This in accordance with other reports on ER-α and PgR immunohistochemistry in breast epithelium (14, 15, 16).

Table 2 shows the median ER, PgR, and Ki-67 labeling indices in the entire case and control groups and is broken down by predominance of lobular or ductal epithelium in cases and controls. The lobular component of breast epithelium involutes with increasing age, and this is reflected in the age data in Table 2, in which the median age of the patients with predominantly ductal samples was greater than that of patients whose epithelium contained a large lobular component. This was true of both cases and controls, although the difference was more marked in cases, who were, in general, older than the controls. In a logistic regression model, the ratio of lobular:ductal area was strongly associated with case status (P = 0.008). When adjusted for age, this was no longer a significant association (P = 0.134), whereas increasing age was a strong determinant of decreasing lobular area (P = 0.002). The negative correlation of lobular:ductal area to age was also highly significant (r = 0.63; P = 0.0003). Multiple linear regression was performed for the serum hormone relationships to PgR and Ki-67 labeling indices, with adjustment for lobular:ductal area ratio alone and with age.

Correlation of the percentage of ER-positive cells (ER LI) with serum estradiol levels in cases and controls shows that there is a significant difference in the regression slope for these two groups (see Fig. 1). In control women, the ER LI is negatively correlated with serum estradiol, and the proportion of ER-positive cells in the breast declines at higher levels of circulating estradiol. In women with breast cancer, the fraction of ER-positive cells in the breast epithelium displays a nonsignificant positive correlation with serum estradiol. The difference in the regression slopes is significant (P < 0.008). The relationship between ER LI and serum progesterone was not significantly different in cases and controls but was weakly negative in controls, in agreement with data on the suppression of ER expression by progesterone.

The linear regression analysis of ER against serum estradiol presented above was then adjusted for the ratio of lobular:ductal area. The significantly inverse relationship between serum estradiol and ER LI in the control population was preserved in this analysis (P = 0.015), as was the significance of the interaction term (P = 0.001). Thus, the difference between cases and controls in the relation of serum estradiol levels to ER expression in normal breast epithelium is not explained by differences in either age or lobular and ductal composition of the tissue.

The breast epithelium of cases displays a statistically significant positive correlation between PgR expression and the levels of both estradiol and progesterone. Controls display a marginal, nonsignificant positive relationship between PgR expression and both serum estradiol and serum progesterone. The regression slopes between cases and controls do not differ significantly (see P for interaction terms in Table 3).

Adjustment for age in these analyses did not cause any major changes in these relationships. When the regression analyses were controlled for age and the ratio of lobular:ductal area, there was still a significant positive relationship between serum hormone levels and PgR expression in cases, but not in controls. The difference between case and control regression slopes remained nonsignificant (see Table 3).

There was no significant relationship between serum estradiol levels and Ki-67 labeling in either cases or controls. However, the relationship between Ki-67 labeling and serum progesterone was significant for both cases (r = 0.297; P < 0.045) and controls (r = 0.295; P < 0.021), with very similar regression slopes (P for the interaction term = 0.984). However, after controlling for estradiol levels in a multiple regression analysis, there was a significant positive relationship between serum progesterone and Ki-67 labeling (coefficient = 0.23; P < 0.007) in the entire study population, but when the analysis was stratified by case-control status, progesterone significantly predicted Ki-67 labeling in controls (P = 0.027), but not in cases (P < 0.118). The relationship between Ki-67 labeling and serum progesterone was no longer significant when adjusted for age and lobular:ductal ratio in case women. In controls, this adjustment caused a slight decrease in the strength of the association, which was still of borderline significance (Table 3).

Because there is evidence pointing to ligand-independent activation of the ER by epidermal growth factor (17), cyclin D1 (18), and mitogen-activated protein kinases (19), we examined the relationship between ER expression and cell proliferation. We found a nonsignificant direct relationship between ER LI and Ki-67 LI in cases, and a significant inverse relationship in controls. The difference between the regression slopes was again significantly different (P = 0.047, see Fig. 2). When adjusted for age and ratio of lobular:ductal area, the difference between case and control regression slopes was still of borderline statistical significance (P = 0.066).

Because estradiol and progesterone have a synergistic effect on some of the tissue parameters we have measured (e.g., both hormones down-regulate ER, and epithelial proliferation is increased in the luteal phase of the cycle), we next examined the relationships of ER, PgR, and Ki-67 expression to the product of the two hormones (E*P). Again, we saw an inverse relationship between ER LI and E*P in controls (r = −0.2), and a direct relationship in cases (r = 0.2). The P for the difference in coefficients was 0.019. There was also a significant difference (P = 0.02) in PgR LI related to E*P in that a strong positive relationship exists in cases (r = 0.5), but not in controls (r = 0.02). With regard to the Ki-67 LI, there was no difference in the relationship between cases (r = 0.3) and controls (r = 0.2).

Previous studies of ER levels in human breast tissue have relied on menstrual cycle dates and hormone levels inferred from the menstrual cycle phase. However, there is considerable interindividual variation in hormone levels for given phases of the menstrual cycle, and the occurrence of anovulatory cycles often goes undetected. By measuring serum estradiol and progesterone levels on the day of acquisition of the breast tissue sample, we have been able to relate the precise hormone level to hormone receptor expression and cell proliferation in the normal breast epithelium of benign disease controls and breast cancer cases. Our present observations on the relationship of serum estradiol levels to breast epithelial ER and measures of estrogen responsiveness such as PgR expression and cell proliferation are the first indications of an enhanced response of morphologically normal epithelium to estrogen stimulation in women with breast cancer.

We find that in normal breasts, rising serum estradiol suppresses ER expression, has no discernible effect on PgR expression, and has no significant relationship with Ki-67 labeling. In women with breast cancer, ER expression does not fall with rising estradiol levels, and estrogen response in the form of PgR expression is enhanced. There is no direct relationship between serum estradiol and Ki-67 labeling in either cases or controls, but ER expression is inversely related to cell proliferation in control women only. These data provide in vivo evidence that the normal down-regulation of ER by estradiol is important in suppressing the proliferative response to estrogen and in maintaining a homeostatic balance. The significant negative relationship between the Ki-67 LI and ER LI in controls persists when serum estradiol is controlled for (P < 0.039).

Previous studies of human mammary gland morphology have pointed to the potential importance of lobular architecture as an indicator of the degree of differentiation of the mammary gland (17). One recent report from the same group of investigators, based on an analysis of 12 breast samples, presented data that suggest that the expression of hormone receptors and the rate of cell proliferation vary according to the lobular architecture (18). These results are very intriguing, but they have not been reproduced by other groups. Analysis by lobule type will be part of our future efforts in this area, but we do not believe that it currently represents the standard analytic approach in this field. Our analysis of ductal and lobular area shows that age is a strong determinant of the specific glandular compartment distribution in the breast, as has been widely observed by breast pathologists. However, not all of the variation in ductal and lobular composition of our epithelial samples could be explained on the basis of age, and therefore we adjusted for both age and the ratio of lobule:duct area. These adjusted analyses show that the finding of an inverse relationship between serum estradiol levels and ER LIs in control women is very robust and is essentially unchanged by the adjustment. The case-control difference in regression slopes for estradiol and ER is also substantially unchanged. The negative relationship between Ki-67 and ER LI in controls (r = 0.27; P = 0.03) was weakened by adjusting for age and lobule:duct ratio, but the difference between cases and controls remained of borderline statistical significance (P = 0.066).

These data seem biologically plausible, because if down-regulation of ER were not to occur, increased availability of estradiol would be accompanied by an increased sensitivity of target cells and a markedly enhanced estrogen effect. It appears that in control women, down-regulation of ER is a key event in restraining estrogen response in the presence of rising estradiol availability. In high-risk women, on the other hand, the failure of ER down-regulation by estradiol is accompanied by a trend toward a positive relationship between ER expression and proliferation. Additionally, we see no correlation between estradiol levels and proliferative response. This raises the possibility that ligand-independent activation of ER by epidermal growth factor or other mitogens that has been observed in breast tumors (19, 20, 21) is already operative in the high-risk epithelium of these women.

PgR levels are directly and significantly related to both estradiol and progesterone levels only in case women, even when adjusted for age and lobular:ductal area. Controls show no significant relationship, again pointing to enhanced hormone responsiveness in the cases.

Although the correlations observed are weak on the whole, even when statistically significant, they are consistent with relationships that would be expected based on laboratory data developed in far more stringent and controlled conditions than is possible in an essentially heterogeneous clinical population. Nevertheless, the differences we have observed in the relationships between these parameters in cases and controls point out that future studies should look for other such differences in responses and relationships, rather than for summary values for the parameters themselves.

The location of the tissue sample in our patients was not constantly related to the tumor location; 21 of the 50 case samples came from women who had breast-sparing treatment of their breast cancer, and the remainder came from mastectomy specimens. Samples for immunochistochemistry were chosen based on the best quality and quantity of normal epithelium rather than the proximity or distance from the tumor. It is possible that distance from the tumor could influence the epithelial response to hormones; our study is too small to analyze these subsets, but we will pursue this possibility in future investigations.

Factors that may contribute to the variability of the data presented here include differences in free versus bound estradiol levels, variations in local breast estradiol levels related to aromatase activity in the breast, and differences in estradiol metabolism in cases and controls. All of these factors have been reported to differ between breast cancer cases and controls (22, 23, 24, 25). However, higher free estradiol, more bioavailable estradiol, or the presence of more potent estradiol metabolites should all lead to lower rather than higher ER expression in high-risk women in the presence of normal estrogen signaling and regulation of ER. Finally, several ER variants have been demonstrated in breast tumors lacking portions of either the hormone-binding domain or the DNA binding domain and are postulated to be responsible for the antiestrogen resistance of some ER-positive tumors (26). Other described ER alterations include superactive forms of the ER, such as those described in duct hyperplasia samples (27). The significance of these findings relative to the evolution of neoplasia is uncertain but raises the question of whether the dysregulation of ER expression that we see in our cases is related to structural alterations in the ER.

In summary, we have found a dissociation of ER response to estradiol and of proliferative response to ER levels in the benign breast epithelium of women with breast cancer, whereas breast epithelium from unaffected women displays significant negative relationships between these parameters. These changes probably reflect some of the earliest events in estrogen-related breast carcinogenesis. Explanation of these changes on the molecular level will lead to new insights into hormonally related breast cancer etiology and will have applications in the arena of breast cancer prevention.