Short-term Prophylactic Tamoxifen Reduces the Incidence of Antiestrogen-Resistant/Estrogen Receptor–Positive/Progesterone Receptor–Negative Mammary Tumors Free

The treatment and prevention of breast cancer has improved significantly due to the successful use of antiestrogens, more specifically known as selective estrogen receptor modulators (SERM) or down-regulators (SERD), and aromatase inhibitors. However, these drugs have proven to be effective only in clinical cases in which primary tumors express receptors for estrogen (ER+). Because ∼70% of all clinical breast cancer cases are ER+, the use of these antiestrogens has improved overall morbidity and mortality for a substantial number of patients.

Unfortunately, not all breast cancer patients with ER+ disease benefit from such estrogen signaling manipulations. Approximately 30% of metastatic breast cancers are de novo resistant (1), and a majority that are initially sensitive develop resistance with time. This resistance is more frequent in ER+ tumors containing very low (or undetectable) levels of progesterone receptor (ER+/PR−) than in tumors containing higher progesterone receptor levels (ER+/PR+; ref. 2). In addition, in some circumstances in which growth factor and/or growth factor receptor levels are pronounced, treatment with growth factor inhibitors can improve the sensitivity of breast cancers to antiestrogen therapies and therefore promote the efficacy of such treatment (3), pointing to the importance of understanding the mechanisms underlying antiestrogen resistance. One possibility is that resistance to antiestrogen therapy evolves relatively late in tumorigenesis, making early prophylactic administration of chemopreventive SERMs in women at high-risk a means to reduce the subsequent incidence or delay the development of SERM/SERD/aromatase inhibitor–resistant breast cancer.

Although prophylactic treatment with SERMs reduces the development of ER+ breast tumors by ∼50% in women with a high risk of breast cancer development, it is not known if their growth factor or progesterone receptor status influences subsequent breast cancer development. One major hindrance in addressing hypotheses of the underlying mechanisms of antiestrogen resistance has been the lack of appropriate rodent models. Although many mouse models are ideal for studying carcinogenesis due to their relatively short life cycle and the availability of numerous genetically engineered mice that develop mammary cancer, there are few such models that develop ER+ tumors. Furthermore, it has been difficult to study estrogen modulation of mammary tumors in many of these transgenic mouse models because traditional mammary-specific promoters are themselves modulated by estrogen.

The recent development of the NRL-TGFα transgenic mouse lineage, which expresses the ErbB1 HER1 ligand, transforming growth factor-α (TGFα) under the regulation of the NRL promoter, circumvents these limitations. NRL-TGFα mice reliably develop mammary hyperplasias and tumors,the later between 6 and 24 months depending on environmental factors. In this model, the mammary hyperplasias and tumors are ER+ and NRL-directed transgene expression is not modulated by exogenous estrogens or by oophorectomy (4). The tumors develop with progressive loss of progesterone receptor mRNA and protein in the presence of consistently elevated levels of growth factor expression. Thus, the NRL-TGFα mouse model closely simulates the ER+/PR−/growth factor+ phenotype that represents a subset of SERM/SERD/aromatase inhibitor–resistant clinical breast cancers.

Using this mouse model in the present study, we showed that de novo resistance to antiestrogen treatment was present in late stage tumors. Furthermore, we showed that when tamoxifen was given to mice prior to expected lesion development for a 60-day period, there was a significant beneficial effect on tumor incidence and latency as well as hyperplasia characteristics several months later. Our results document an ∼50% reduction in tumor incidence in this model, similar to the reduction rates of ER+ cancer obtained in tamoxifen prevention clinical trials involving women at high risk for developing breast cancer. This study validates the NRL-TGFα mouse model for investigating ER+ mammary tumor behavior and suggests that prophylactic tamoxifen treatment might be helpful in reducing ER+/PR− clinical breast cancers. Our findings open up important opportunities for identifying cellular and molecular predictors of efficacious tamoxifen chemoprevention and/or antiestrogen resistance in this model that may prove applicable to human disease.

NRL-TGFα mice of line 1372-1 TgN(Nrl-tgfa)29EPS in the FVB/N strain were used in this study as previously described (4), and transgenic mice were identified by PCR of tail DNA using primers that identified specific regions of the transgene and span the final intron of the rat genomic TGFα DNA to the 3′ untranslated region of hGH sequences. Use of the following primers produced a 734-bp product: forward primer ATGGCTCAGGACCACACAACTT and reverse primer ACAGTGCCAAGCAAGCAACTCA. Mice were housed in Association for Assessment and Accreditation of Laboratory Animal Care International–accredited facilities with 14 h on, 10 h off light cycle. Food (Purina Lab Diet 5001) and water were provided ad libitum. All procedures were approved by the Institutional Animal Care and Use Committee. Transgenic mice evaluated in this study were heterozygous for the transgene.

To evaluate the effect of tamoxifen treatment on NRL-TGFα expression in the mammary gland, 8-week-old transgenic mice were implanted with a single 5 mg 4-OH tamoxifen or tamoxifen-free carrier pellet (hereafter called placebo pellet; Innovative Research of America). Three weeks later, mammary glands were isolated, flash-frozen in liquid nitrogen, and stored at −70°C. Quantification of TGFα transgene expression and the housekeeping gene PUM1 was carried out as described previously (4). Briefly, total RNA was isolated from frozen mammary glands and 2.5 μg were reverse-transcribed to cDNA in 25 μL using oligo(dT) (StrataScript First-Strand Synthesis System; Stratagene). Control reactions included no reverse transcriptase controls and controls without RNA (no template). PCR was done using 2× Brilliant SYBR Green QPCR Master Mix (Stratagene) and previously published primers (4). Triplicate cDNA samples were run and average crossing threshold (Ct) values calculated. Reaction efficiencies were calculated to be 1 from standard curves obtained from Ct data of serial dilutions of cDNA. We used PUM1 as our single gene normalizer. Relative gene expressions were normalized to housekeeping gene PUM1 and calculated as 2^{Ct(PUM1) − Ct(gene)}. In preliminary experiments, the mRNA levels of PUM1 as well as three other genes, UBC, RPL19, and PMSC4, were found to be constant in 17 RNA samples from NRL-TGFα mammary glands with and without tumors, indicating that these genes could be used as internal standards (data not shown).

Mice with tumors were implanted with a 5 mg 4-OH tamoxifen (tamoxifen), or placebo pellet, or injected s.c. 5 days/wk with 100 μg of ICI in 100 μL of canola oil or canola oil alone. Fold growth was evaluated after 3 weeks. Tumor size was measured in two dimensions using a calipers and volume estimated using r₁² × r₂ × 3.14/6 in which r₁ ≤ r₂. Fold growth was determined by volume at 3 weeks posttreatment/volume on day 1 of treatment. At the end of treatment, tumor-containing and nontumorous glands were formalin-fixed and subjected to histologic and immunohistochemical evaluation as described below.

NRL-TGFα mice received subcutaneous implants releasing 5 mg of 4-OH tamoxifen (hereafter referred to as prophylactic tamoxifen) over 60 days or containing tamoxifen-free carrier (hereafter known as placebo mice). Mice were maintained under identical housing conditions and every other subject was assigned to the tamoxifen or placebo cohort as they became 8 weeks of age over a 1-year assignment period. Mice were held for tumor development for up to 2 years of age.

Glands were fixed in 10% neutral buffered formalin at room temperature for 18 to 24 h, dehydrated in ethanols, embedded in paraffin, sectioned (4-6 μm), and stained with H&E. Some tissues were evaluated for ER and PR. ER polyclonal antibody (MC-20; Santa Cruz Biotechnology) and PR polyclonal rabbit antibody (DakoCytomation) were used. Deparaffinized slides were exposed to 0.5% H₂O₂ in methanol, subjected to antigen retrieval consisting of 15 min of boiling in 0.1 mol/L Tris (pH 9.0 for ER) and 15 min boiling in 0.1 mol/L citrate buffer (pH 6.0 for PR), blocked with 0.5% milk in PBS and incubated overnight with 1:1,000 ER or 1:750 PR antibodies, rinsed, and incubated with secondary antibody (BioGenex), and finally rinsed and incubated with horseradish peroxidase streptavidin conjugate. Slides were then developed with 3,3′-diaminobenzidine and counterstained with hematoxylin. Alternatively, deparaffinized slides were subjected to peroxidase blocking reagent (DakoCytomation) followed by antigen retrieval using Rodent DECLOAKER (Biocare Medical) in a Pascal pressure cooker at 125°C for 20 min. Once slides reached room temperature, they were rinsed with distilled water, followed by TBS (DakoCytomation), and protein blocked (DakoCytomation). Primary antibodies were added at 1:250 (ER) and 1:100 (PR) for 40 min and then treated with Rodent on Rodent HRP Polymer (Biocare Medical) for 25 min, 3,3′-diaminobenzidine for 5 min, and DAB Enhancer S196 (DakoCytomation), and counterstained with hematoxylin for 5 min prior to rinsing in water and dehydration through alcohols. Prior to mounting coverslips with Permount, slides received a final rinse in Clear-Rite 3. For negative controls, primary antibody was omitted. Negative controls included an inactivated blocking peptide for ER (Santa Cruz Biotechnology) and omitted primary antibody for PR. Uterine tissue was used as positive control tissue (data not shown). Images were visualized using a Nikon Eclipse E600 microscope and acquired using a SPOT CCD digital camera (Diagnostic Instruments, Inc.) and Metamorph Imaging System 3.5 software, and the photocollage (Supplementary Fig. S1) was assembled using Adobe PhotoShop CS, version 8.0.

For transgene expression and tumor growth analyses, two-tailed Student's t tests were done on log-transformed data (Figs. 1 and 2, respectively). Tumor incidence was determined by dividing the number of mice which acquired tumors by the total number of mice in that treatment group and proportions compared using χ² test (Fig. 3A). For survival analysis, Kaplan-Meier plots were generated and differences were determined using log rank tests (Fig. 3B and C). For analysis of differences in degree of hyperplasias in tumor versus tumor-free mammary glands and in glands from mice that had received tamoxifen or placebo preventive, two-way ANOVA was carried out on nontransformed data (Fig. 4). Differences between groups were considered significant at P < 0.05.

We previously showed that transgene expression in NRL-TGFα mice is not modulated by estrogen in vivo or in vitro or by ovariectomy (4). Therefore, we predicted that transgene expression would not be affected by tamoxifen treatment. In this study, we confirmed that NRL transgene expression is not modulated by the SERM, tamoxifen, when delivered continuously over 3 weeks (Fig. 1).

NRL-TGFα mice with tumors were given tamoxifen or ICI versus placebo and tumor size was monitored over 3 weeks. Neither tamoxifen nor ICI altered the average growth of tumors during treatment as compared with controls (P = 0.8182 and P = 0.8918, respectively), although there were two tamoxifen-treated tumors which had negative growth and two which had more positive growth compared with placebo controls at the end of treatment, suggesting possible tamoxifen sensitivity in some cases and de novo resistance by most. ER or PR levels were not different for the two tumors displaying negative growth compared with the remaining tumors within the tamoxifen group or as compared with the placebo group. However, further study is required to determine if subpopulations of tumors exist that respond differently to tamoxifen treatment (Fig. 2). In addition, histologic assessment of the sizes of microscopic tumors, the degree of epithelial layering, mitotic levels, degree of dysplasia, or ER and PR levels in tumors showed no differences between groups (data not shown).

NRL-TGFα mice receiving a 60-day implant of tamoxifen delivered prior to the development of lesions were followed until tumor development or 2 years of age. We found that, similar to clinical responses to tamoxifen seen with ER+ cases, tumor incidence was reduced by ∼50% in mice that received prophylactic tamoxifen compared with placebo controls (Fig. 3A). Clinical trial data do not as yet reveal whether tumor latency is longer in individuals receiving chemopreventive tamoxifen and we also have been unable to find published preclinical data addressing this issue. One possibility is that tamoxifen delays the onset or progression of precancer changes in individuals at high risk of developing breast cancer. However, in NRL-TGFα mice, time of death due to tumor development was not altered with prophylactic tamoxifen. These results suggest that the tamoxifen response in this model was all or nothing. The SD of tumor latency in placebo-treated NRL-TGFα mice was ∼4 months representing ∼25% of the mean time to tumor development (Fig. 2). This relatively large variation may mask subtle effects of tamoxifen on tumor latency. Tumor latency of TGFα mice that received tamoxifen versus placebo chemopreventive is highly similar as evidenced by similar survival curves (Fig. 3B). Importantly, there is a consistent pattern of tumor incidence throughout the 2-year study indicating that there was no drift associated with genetic, epigenetic, and/or environmental conditions.

Unknown from clinical trials using tamoxifen as a preventive is whether tumor or hyperplasia histotype is altered. In our model, we found that tamoxifen did alter mammary gland histology; mammary glands collected from tamoxifen-treated mice showed a reduced number of hyperplasias compared with glands from mice that received placebo implants (Fig. 4; Supplementary Fig. S1). In addition, the sizes of hyperplasias were reduced, as was the level of cellular atypia and extent of secretory vacuoles (Table 1; Supplementary Fig. S1). Furthermore, the extent of hyperplasias was also influenced by the presence of tumors. Mice with tumors had a larger number and larger hyperplasias than tumor-free mice (Fig. 4; Supplementary Fig. S1). Tumors that developed in mice receiving prophylactic tamoxifen had reduced levels of mitotic cells. Mitotic counts ranged from 1 to 9 per 10 high-power fields compared with a range of 0 to 21 mitotic counts per high-power field in mice that had received placebo implants, nearing but not reaching statistical significance (two-tailed Student's t test; P = 0.08). Immunohistochemical analysis did not reveal any detectable differences in the number of cells stained with ER or PR, or in the intensity of staining of cells making up hyperplasias with and without tumors, or from mice that had received tamoxifen versus placebo. Finally, differences in ER and PR staining were not detected in tumors from mice that received tamoxifen or placebo implants (data not shown).

NRL-TGFα mice that formed tumors reached similar ages regardless of treatment. Mice developed tumors between 11 and 24 months of age (Fig. 3B) and were sacrificed due to the size of the primary tumor. However, in NRL-TGFα tumor- free mice, there was a marked difference in life span between tamoxifen-treated and placebo-treated mice in that that mice which received 60 days of prophylactic tamoxifen lived significantly longer (Fig. 3C; P < 0.0001) than mice that received placebo. Most tamoxifen-treated mice lived close to 2 years of age, whereas death in placebo-treated mice occurred between 1 and 2 years, and was spread uniformly, without a single concentrated period of death in the placebo cohort. In addition, the underlying cause of death generally was not obvious. When mice were found dead or were euthanized, necropsy was carried out to determine the cause. We identified the following causes in six mice: three mice had uterine/vaginal tumors, two mice had enlarged urinary bladders, and another had a lung abscess.

The lack of detectable PR in ER+ clinical breast cancer increases the likelihood of tamoxifen resistance. Additionally, this outcome is correlated with an increased presence of growth factors (5). As predicted by clinical data, established ER+/PR− mammary tumors in NRL-TGFα mice did not respond to tamoxifen therapy or to the pure antiestrogen, ICI 182,780 (clinical drug name is fulvestrant). To the best of our knowledge, the effect of tamoxifen chemoprevention on incidence of clinical cases of ER+/PR− breast cancer has not been determined, but our data clearly show that short-term prophylactic tamoxifen reduces ER+/PR− tumor incidence in NRL-TGFα mice. We aimed to mimic clinical trials in that mice received tamoxifen for a limited period, 60 days in this study, representing ∼8% of a 2-year life span. Tumor incidence in these mice was reduced by 50%, similar to the reduction in invasive ER+ breast cancer in the Royal Marsden Trial National, the International Breast Intervention Study I Trial, and in the National Surgeon Adjuvant Breast and Bowel Project P-1 Trial, which used between 5- and 8-year courses of tamoxifen and which have been followed for up to 20 years (6). One earlier preclinical investigation carried out in the C3H mouse strain, which spontaneously develops mammary cancer due to mouse mammary tumor virus, also found that a limited course of prophylactic tamoxifen was just as effective as a longer course (3 versus 12 months) for reducing tumor incidence (7). These collective findings suggest that a high priority should be to understand how limited exposure to tamoxifen results in a long-term breast cancer–preventive effect. This information could open up new approaches to prevention and be realized as significant reductions in breast cancer morbidity and mortality.

The effect of prophylactic tamoxifen on tumor latency has not been scrutinized in clinical trials due to the limited time since the studies were commenced. Such studies will be challenging due to the heterogeneity of the disease in the population. In the present study, tumor latency was not different between genetically identical NRL-TGFα mice that received tamoxifen versus placebo implants. This finding leads us to conclude that the response to prophylactic tamoxifen was all or nothing rather than simply postponing tumor development.

The effect of tamoxifen chemoprevention on tumor phenotype in clinical trials is also unknown. In NRL-TGFα mice, tumor phenotype was not altered with treatment, with one exception; a reduction in mitotic cells was seen with prophylactic tamoxifen which approached statistical significance. Regardless of this finding, tumorigenesis seems to be fundamentally similar in treated versus untreated mice as tumors displayed indistinguishable growth rates, histologic architecture, and similar levels of ER and PR. The fact that prophylactic tamoxifen did not alter tumor phenotype may be due to the unvarying cellular responses to elevated TGFα expression or to the time window in which tamoxifen was delivered. These issues could be clarified by comparing tamoxifen versus placebo outcomes in an inducible mouse lineage in which the TGFα transgene can be induced at different times before, during, and after tamoxifen delivery or in the present NRL-TGFα line used in this study but with delivery of prophylactic tamoxifen at later time points. One consistent finding between our study and other investigations using rodent breast cancer models is that the delivery of tamoxifen prior to tumor establishment seems key to effective prevention (7–11).

Because we know that hyperplasias (usual and atypical) precede tumor development in NRL-TGFα mice in a similar fashion as described in WAP-TGFα mice (12), we evaluated the degree of hyperplasias in mice at necropsy. We found that the extent of hyperplastic disease, measured as the degree of lobularization, was greatly reduced in tumor-free mice (Fig. 4; Supplementary Fig. S1). In addition, hyperplasia load was reduced significantly in glands from mice that received tamoxifen chemoprevention and this reduction occurred to a similar extent in both tumor-free or tumor-laden mice. Furthermore, there was a reduced degree of cellular atypia and vacuoles, markers of atypical hyperplasias, in mice that received chemopreventive tamoxifen compared with placebo mice (Table 1; Supplementary Fig. S1). If tumors arise from hyperplasias, a decrease in hyperplasia size and degree of atypia may be key to reducing tumor incidence. Based on our earlier characterization of tumorigenesis in TGFα mice, this seems to be the case, as we find hyperplasias are (a) preceded by a higher degree of epithelial cell turnover than architecturally normal epithelia within the same mouse, (b) present prior to tumor formation, and (c) coincident with tumor presence (4, 13). By some means, tamoxifen chemoprevention seems to inhibit the net cell gain that comes with high proliferative activity that is intrinsic with mammary hyperplasias in this model. This cannot be accomplished strictly by the direct tamoxifen effect of reducing proliferation as tamoxifen is not present during most of the subject's life, in this mouse study, up to 88 weeks after tamoxifen withdrawal, and in clinical trials for up to 20 years (6).

Hyperplasias may represent precursor lesions or indicate general pathologic changes reflecting an increased risk of breast cancer development. Atypical hyperplasias are found more often in mastectomies or biopsies containing malignancies than in mastectomies or biopsies containing benign lesions (14). Furthermore, low-grade hyperplasias have been found more often in the margins of breast biopsies from women that later developed ipsilateral breast cancers than in women that did not (15). The finding that hyperplastic lesions contain genetic changes that are reflected in invasive cancers in the same individual provides further evidence linking hyperplasias to cancer development (16, 17). A few studies using rodent models of mammary cancer have shown preneoplasia reduction including hyperplasias with tamoxifen chemoprevention, even in cases in which developed tumors are hormone-independent (7-11, 18–21). Importantly, in women, hyperplasias typically have elevated ER+ (22) and therefore represent an excellent target for SERM chemoprevention.

Cells that form preneoplasias seem to be an important target of SERM chemoprevention, not only in preclinical models but in one critical clinical trial: women in the tamoxifen arm of the National Surgical Adjuvant Breast and Bowel Project Breast Cancer Prevention Trial (P-1) realized a 50% reduction in ductal carcinoma in situ compared with women receiving placebo (23). In spite of these findings, the influence of SERMs on the incidence of usual or atypical hyperplasias has had very little scrutiny in clinical studies, with one exception; there were no detectable differences in the presence of usual and/or atypical ductal hyperplasias in one small study involving ∼50 women who received a 1-year course of either tamoxifen or placebo (24). The influence of estrogen signaling in hyperplasia development is intuitive based on what is known about the mitogenic properties of estrogens in breast as well as the following clinical observations: (a) a recent evaluation of the Women's Health Initiative randomized trial of estrogen and progestin reported an increased risk of benign breast disease in women that took these hormones (25) compared with women receiving placebo (26). The concern is that women with benign breast disease have an increased risk of breast cancer with hormone replacement (27). (b) Tamoxifen has been shown to reduce breast density in women, and the major determinants of breast density at baseline were menopausal state, body mass index, and previous history of atypical hyperplasias (28). (c) Women at high risk of breast cancer development due to heredity had high incidences of hyperplasias found on mastectomy, however, incidence was lower in a subgroup that earlier had undergone oophorectomy (29). Taken together, these data predict that hyperplasia development is influenced by estrogens and can be prevented or abrogated with antiestrogens. These findings in women and in the aforementioned studies involving rodents stress that evaluations of hyperplasia incidence and degree should be evaluated in SERM breast cancer prevention trials. Additionally, it would be important to correlate reductions in hyperplasias with ER and tumor status.

An unexpected result of this study was that mice receiving prophylactic tamoxifen remained tumor-free and lived longer than placebo-receiving tumor-free mice. Tumor-free mice died from causes other than mammary tumor development and displayed few clues as to cause, although the underlying process(es) seem to be modulated by estrogen. Whether this is a phenomenon specific to this model, the FVB/N inbred mouse strain, or can be more generally applied is unknown. The effect of tamoxifen chemoprevention on life span in women is currently unclear and will likely remain as such for many years. However, statistical modeling of tamoxifen's effects on life span in women in prevention trials has been carried out (30). These authors present a theoretical model which predicts that younger women (ages 30-49) receiving tamoxifen may live longer than placebo arm age-matched women, assuming that the negative side affects of tamoxifen, pulmonary embolism, and endometrial cancer are removed upon treatment cessation. This benefit also applied to women >50 years of age, but only when it was assumed that cancer incidence did not change with time since receiving tamoxifen. The appropriateness of these statistical models to women in SERM prevention trials will be revealed in the future and our studies support these future evaluations.

In conclusion, identifying individuals who will respond to SERM chemoprevention is a high priority as is discerning the molecular or cellular targets of SERM and general markers of tumor outcome. In addition, understanding the basis of why some breast cancer cases respond to SERM chemoprevention whereas others do not would greatly enhance the field of breast cancer prevention. NRL-TGFα mice develop ER+/PR− tumors that are resistant to tamoxifen and ICI, but tumor development can be prevented by a short-term prophylactic course of tamoxifen. The cellular and molecular targets of tamoxifen, as well as the biomarkers for tumor outcome, can be studied in this mouse model because only half of the mice respond to tamoxifen chemoprevention. Mammary epithelial cell subtypes that evolve into hyperplasias likely represent the targets of tamoxifen action in this mouse model. Investigations which reveal these targets in this and other preclinical models will shed light on the mechanism by which short-term tamoxifen chemoprevention leads to long-term tumor prevention. In addition, our data suggest that evaluation of outcomes in women enrolled in SERM prevention trials should include specific tracking of the effects of tamoxifen chemoprevention on the incidence of ER+/PR− tumors and the presence of hyperplasias and precancers as well as its effects on life span.

No potential conflicts of interest were disclosed.

We are grateful to Dr. Lois J. Heller for critical review of this manuscript and acknowledge St. Luke's Hospital, Duluth, MN for their generous research support. We also acknowledge Rebekah Bolstad for managing tissue collections and Shannon Ament-Yellowbird for evaluating mitotic and apoptotic counts in tumors as well as Barbara Elmquist and Emily Heid for excellent histological and immunohistochemical preparations.