Main

Previously we reported that MMTV-RANK female transgenic mice show sustained proliferation and impaired alveolar secretory differentiation of the mammary epithelium upon pregnancy; however, spontaneous mammary tumours were not observed in aged virgin mice6. In contrast, mammary glands from multiparous MMTV-RANK mice showed spontaneous tumour development (median onset 26.5 months) and exhibited a higher incidence of pre-neoplasias compared with wild-type mice; adenocarcinomas were only observed in MMTV-RANK mice (Supplementary Table 1).

To characterize more fully the role of the RANKL/RANK pathway in mammary tumorigenesis, we induced mammary carcinogenesis using combined treatment with the hormone medroxyprogesterone acetate (MPA) and a carcinogen (7,12-dimethylbenzanthracene (DMBA)) (Supplementary Fig. 1)7. After MPA/DMBA treatment, MMTV-RANK transgenic mice showed a markedly enhanced susceptibility to mammary tumours compared with wild-type mice (Fig. 1a). Adenocarcinoma, adenosquamous carcinoma and adenomyoepithelioma carcinoma histotypes were observed in both strains (Supplementary Fig. 2). Pre-neoplastic mammary lesions were clearly more widely distributed in mammary tissues in MMTV-RANK mice than in wild-type mice (Fig. 1b). Multifocal ductal hyperplasias, multifocal and focally extensive mammary intraepithelial neoplasias (MIN), and multiple carcinomas were frequently present in a single involved MMTV-RANK mammary gland in contrast with focal lesions in the wild-type glands. Whole-mount analysis of mammary glands at a stage before detection of palpable tumours revealed a 100% incidence and high multiplicity of dense pre-neoplastic epithelial foci, consisting of a mix of hyperplasias and MIN, in MMTV-RANK mice (Fig. 1c). In the induced model, increased mammary proliferation was also evident in the MMTV-RANK versus wild-type glands as early as 2 days after the first DMBA treatment, and a significant increase in cyclin D1 immunohistochemistry signal was evident at 4 and 7 weeks after the last DMBA treatment (Fig. 1d).

Figure 1: Decreased latency of mammary tumour formation and accelerated pre-neoplastic changes in mammary glands of MMTV-RANK mice treated with hormone and carcinogen.
figure 1

a, Kinetics of palpable mammary tumour onset in wild-type and MMTV-RANK lines 1 and 2 mice treated with MPA plus DMBA; P-values based on log-rank test. Latency was expressed as weeks after last DMBA treatment. b, Mammary glands from tumour-bearing wild-type and MMTV-RANK line 1 mice after MPA plus DMBA treatment (haematoxylin and eosin staining; original magnification, ×4). Scale bar, 500 μm. c, d, Analysis of mammary glands before tumour formation. c, Whole mounts of inguinal mammary glands from wild-type and MMTV-RANK line 1 mice prepared 7 weeks after last DMBA treatment; arrows indicate dense epithelial foci. Scale bar, 500 μm. d, Proliferation in mammary epithelial cells (BrdU immunohistochemistry, mean ± s.e.m.) and quantification of high cyclin D1 expressing cells (2+ score by immunohistochemistry) in glands from MPA/DMBA-treated wild-type and MMTV-RANK line 1 mice 48 h after the first DMBA treatment and 4 and 7 weeks after the last DMBA treatment. n = 5 mice per group. Independent t-tests were used for statistical comparisons.

PowerPoint slide

Full size image

RANK protein was clearly evident in luminal and abluminal cells of normal mammary epithelium and in the epithelial component of pre-neoplasias and carcinomas in both MPA/DMBA-treated wild-type and MMTV-RANK mice, and as expected, RANK was observed at higher levels in mammary tissue of MMTV-RANK mice (Supplementary Figs 3 and 4). Compared with ductal hyperplasias and adenocarcinomas, RANK expression was relatively lower in MIN lesions from both wild-type and MMTV-RANK mice (Supplementary Figs 3 and 4). In mammary glands from tumour-bearing wild-type and MMTV-RANK mice, RANKL protein was detected within the luminal cells of histopathologically normal mammary epithelium and in the epithelial component of pre-malignant lesions, with consistently higher levels detected in MIN lesions as compared to other pre-malignant tissues (Supplementary Figs 3 and 4). RANKL was also detected within the carcinoma element in the majority of each histotype in wild-type and MMTV-RANK tumours. The lower level of RANK immunoreactivity within MIN may be explained by a ligand-dependent reduction of RANK protein or mRNA, as demonstrated in mammary tissues of MMTV-RANKL mice5 or after treatment of mammary epithelial cells (MECs) with RANKL in vitro6, respectively.

Given the potential importance of the RANK/RANKL pathway early in tumorigenesis, we confirmed that MPA pre-treatment resulted in an early and marked induction of RANKL (Supplementary Fig. 5). Expression of both RANK and RANKL proteins was also evident at the time of carcinogen administration (Supplementary Fig. 5). Dual immunostaining of samples from MPA/DMBA-treated mice indicated that RANKL protein was predominately localized in progesterone receptor (PR)-expressing cells, not only in normal luminal epithelial cells (as has been described)5 but also in pre-neoplastic and neoplastic epithelium (Fig. 2a–d). Additionally, some RANKL immunoreactivity was also detected in luminal spaces, perhaps reflecting secreted material or induction via other factors such as prolactin or parathyroid hormone-related peptide3. In human invasive breast carcinoma, RANKL was detected heterogeneously in 11% (6 of 57) of tumours (Fig. 2e) whereas RANK was detected in 6% (3 of 50) of tumours (Fig. 2h). RANK and RANKL were not co-expressed within the carcinoma epithelium of any sample. Unlike the mouse model, RANKL was additionally detected in sporadic infiltrating mononuclear cells present in most tumour stroma (67%; 38 of 57) and in fibroblast-like cells in the stroma of rare tumours (5%; 3 of 57) (Fig. 2f and g, respectively).

Figure 2: RANKL is localized in PR + cells at multiple stages during mouse tumorigenesis, and RANKL and RANK are expressed in invasive carcinoma of the human breast.
figure 2

a–d, Representative examples of immunohistochemistry analysis of RANKL and PR in mammary glands from tumour-bearing wild-type mice after MPA plus DMBA treatment showing normal mammary epithelium (a), hyperplasia (b), MIN (c) and adenocarcinoma (d); RANKL is blue and PR is red and indicated by arrows. RANKL protein was rarely observed within mononuclear cells in the stroma but was detected within the epithelial component of normal mammary epithelium, hyperplasia, MIN and adenocarcinoma. RANKL expression is most frequently localized in PR-expressing cells although there is some RANKL immunoreactivity also detected in luminal spaces distinct from PR-expressing cells. Scale bar, 40 μm. e–h, Representative examples of RANKL or RANK immunohistochemistry (brown) of human invasive breast carcinoma (×20 magnification). e, Intense RANKL staining within the tumour epithelium. f, Scattered mononuclear cells with intense RANKL reactivity. g, Intense RANKL reactivity was detected in spindle-shaped fibroblast-like stromal cells of some tumours. h, RANK was detected in epithelial carcinoma cells in some tumours. Scale bar, 50 μm.

PowerPoint slide

Full size image

To elucidate the operative role of RANKL in mammary tumorigenesis, MPA/DMBA-induced MMTV-RANK mice were treated with a selective pharmacological RANKL inhibitor, RANK-Fc, concurrently with the first DMBA treatment, but 3 weeks after MPA treatment. RANK-Fc significantly delayed formation of palpable mammary tumours (Fig. 3a). Whereas adenocarcinomas were the predominant carcinoma subtype observed in MPA/DMBA-induced tumours in MMTV-RANK mice, RANK-Fc significantly suppressed adenocarcinoma formation such that the spectrum of carcinoma histotypes in mice that eventually developed tumours was evenly distributed (Supplementary Fig. 6). Because we had also observed RANKL and RANK expression in malignant and pre-malignant lesions in wild-type mice, we also tested whether RANKL affected tumour formation in MPA/DMBA-treated wild-type mice. RANK-Fc, initiated concurrently with carcinogen administration, almost completely blocked the occurrence of palpable mammary tumours in wild-type mice (Fig. 3b). The reduced incidence and delay in mammary tumour formation in MPA/DMBA-induced wild-type mice was observed selectively with RANK-Fc treatment, but not after treatment with the bisphosphonate zoledronic acid (Supplementary Fig. 7). The number of F4/80+ myeloid cells in the tumour microenvironment was not altered after RANK-Fc treatment (data not shown). Altogether these data indicate that RANKL inhibition blocks mammary tumour formation via a direct effect on mammary epithelia, distinct from its established ability to inhibit osteoclastogenesis.

Figure 3: RANK-Fc inhibits MPA/DMBA-induced mammary tumour formation which is preceded by a protective effect on early pre-neoplastic changes in the mammary gland.
figure 3

RANK-Fc treatment, at 10 mg kg−1 subcutaneously three times per week, was initiated as indicated in Supplementary Fig. 1 and continued until mice were killed. a, Kinetics of palpable mammary tumour onset in MPA/DMBA-treated MMTV-RANK line 1 mice after treatment with PBS or RANK-Fc. b, Kinetics of palpable mammary tumour onset in MPA/DMBA-treated wild-type mice after treatment with PBS or RANK-Fc. Data for panels a and b are expressed as percentage of mice free of palpable tumours after the last DMBA treatment (P-values based on log-rank test). c, d, Distinct regions of interest (normal mammary epithelium, hyperplasia, MIN and adenocarcinoma) of mammary glands from tumour-bearing mice were identified histologically in ≥5 individual mice per group, circumscribed, and each distinct region of interest was separately assessed for proliferation (BrdU immunohistochemistry) or apoptosis (cleaved caspase 3 immunohistochemistry). Data for proliferation (c) and apoptosis (d) in mammary tissue from MMTV-RANK tumour-bearing mice with or without RANK-Fc treatment are expressed as mean per cent ± s.e.m.; independent t-tests were used for statistical comparisons.

PowerPoint slide

Full size image

Analysis of mammary glands from tumour-bearing MMTV-RANK and wild-type mice showed that RANKL inhibition led to a selective reduction in the proliferative index of resident normal mammary epithelium and pre-neoplastic hyperplasias, but not in adenocarcinomas (Fig. 3c), and increased apoptosis was observed in MIN (Fig. 3d). These alterations manifested into a marked reduction in overall mammary epithelial density, as evident in either MMTV-RANK or wild-type mice treated with RANK-Fc (Supplementary Fig. 8). Because our data suggested direct effects on mammary epithelium and protection against early events in tumorigenesis, we analysed pre-neoplasias, cyclin D1 levels8,9 and proliferation in mice at earlier stages, before the detection of palpable mammary tumours. Treatment of MPA/DMBA-induced MMTV-RANK mice with RANK-Fc resulted in a clear suppression of pre-neoplastic lesions and a corresponding reduction in mammary epithelial proliferation at early stages (Supplementary Fig. 8 and Supplementary Table 2). Elevated cyclin D1 has been reported during RANKL-dependent hyperplasia5 and RANKL-dependent normal morphological changes during mammary gland development10; we also demonstrated that RANKL treatment causes increased proliferation6 and rapidly increases cyclin D1 mRNA and protein levels in MMTV-RANK MECs in vitro (Supplementary Fig. 9), indicating that proliferation and cyclin D1 were potential pharmacodynamic markers of RANK-Fc activity in this tissue. RANK-Fc treatment of either MMTV-RANK or wild-type mice caused a very rapid reduction (within 48 h of treatment) in both hormone-induced proliferation and mammary epithelial cyclin D1 levels (Supplementary Fig. 8).

Progesterone triggers mammary proliferation via two distinct mechanisms: an early, RANKL-independent, direct mitogenic effect on PR+ cells followed by a wave of greater proliferation mediated by the paracrine effect of RANKL on PR− cells4. We observed that pre-treatment with MPA was sufficient to increase both RANKL and proliferation at time of subsequent DMBA exposure. These extremely early proliferative changes, due to either MPA pre-treatment alone or as a result of combined MPA and DMBA treatment, were dependent on RANKL as shown by the significant and rapid reductions upon RANK-Fc treatment (Supplementary Fig. 5). To address the relative contribution of RANKL to the amplification of early progesterone-induced proliferation, we compared the effect of a RANK-Fc dose which maximally inhibits proliferation (10 mg kg−1; Supplementary Fig. 10) to the PR antagonist mifepristone. Injection of unsynchronized cycling wild-type mice with mifepristone resulted in a 98% reduction in progesterone-induced proliferation (P < 0.001 versus control) whereas RANK-Fc reduced proliferation by 82% (P < 0.001 versus control), indicating that RANKL was responsible for the majority of proliferation induced by progesterone at 72 h (Supplementary Fig. 10). Notably, we observed that RANK-Fc had no effect on RANKL levels whereas mifepristone completely blocked progesterone-induced RANKL production (Supplementary Fig. 10), indicating that RANKL inhibition was not directly interfering with progesterone–PR interactions. RANKL treatment of MECs from MMTV-RANK mice stimulated growth in semi-solid media, caused an inappropriate filling of the luminal cavity6 and disrupted normal architecture and apicobasal polarization of three-dimensional acinar cultures (Supplementary Fig. 9), indicating that RANKL can also evoke phenotypic changes consistent with transformation.

We next examined the effect of RANKL inhibition in MMTV-neu transgenic mice, a spontaneous mammary tumour model without an exogenous hormone requirement11. Analysis of virgin cycling MMTV-neu mice at a relatively late stage, but still before detection of palpable mammary tumours (4–5-month-old mice), revealed RANK and RANKL expression in normal mammary epithelium (Fig. 4a). RANK-Fc treatment, beginning at 5 months, had no significant impact on median time to spontaneous mammary tumour onset (Supplementary Fig. 11); however, a significant reduction in total number of mammary tumours quantified at necropsy was observed (Fig. 4b). RANKL inhibition again resulted in reduced pre-neoplasias (Fig. 4c), with correspondent reductions in proliferation and cyclin D1 selectively within normal mammary epithelia, but not in adenocarcinomas (Supplementary Fig. 11). Treatment of MMTV-neu mice with RANK-Fc also significantly reduced the incidence (Supplementary Fig. 11) and number of lung metastases per mouse (Fig. 4d). The reduced metastasis may reflect the reduced primary tumour burden after RANK-Fc treatment, but could also be due to inhibition of the metastasis suppressor Maspin by RANKL12. Despite the lack of hormone receptors in MMTV-neu mammary tumours13 (Supplementary Fig. 12), ovarian hormones (including progesterone) can promote tumorigenesis in these mice selectively during a ‘risk window’, before 5 months of age14. Importantly, we observed that RANKL and PR are expressed in normal mammary epithelia but are not detected in any stage of atypia, including MIN and adenocarcinoma (Supplementary Fig. 12). Therefore, the demonstrated abilities of RANK-Fc to reduce mammary epithelial proliferation, protect against pre-neoplasias, and reduce tumour multiplicity in this spontaneous mammary tumour model are similar to the early stage effects observed in the induced MPA/DMBA model and could conceivably result from inhibition of repetitive cycles of RANKL expression in normal PR+ mammary glands of MMTV-neu mice, driven by progesterone surges at dioestrus15. The potential efficacy of RANKL inhibition at later stages or in hormone-receptor-negative tumours is currently under investigation.

Figure 4: RANK and RANKL are expressed in normal mammary epithelium of MMTV- neu mice, and RANK-Fc decreases spontaneous mammary tumorigenesis and lung metastasis in this model.
figure 4

a, RANK and RANKL expression (immunohistochemistry) in normal mammary epithelium is observed at a stage before palpable mammary tumour formation (4–5-month-old MMTV-neu mice). Scale bar, 100 μm. b–d, Effects of vehicle control (muFc) or RANK-Fc administered beginning in 5-month-old MMTV-neu mice (n = 10 per group). b, Mean number (±s.e.m.) of palpable tumours (all glands) plus tumours identified histologically (inguinal glands) in MMTV-neu mice with or without RANK-Fc. c, Mean number (±s.e.m.) of histologically defined pre-neoplasias in inguinal mammary glands of tumour-bearing MMTV-neu mice with or without RANK-Fc. d, Mean number of lung metastases (±s.e.m.) in tumour-bearing MMTV-neu mice with or without RANK-Fc. Entire lungs were step sectioned at 75 μm and individual metastases identified histologically. Independent t-tests were used for all statistical comparisons.

PowerPoint slide

Full size image

Progesterone may accelerate mammary tumorigenesis in this and other rodent mammary tumour models7,16 via a proliferative expansion of the target cell pool available for carcinogenic damage17; however, defined molecular targets of progesterone function in tumorigenesis have not been identified. Recently, two studies15,18 revealed a critical involvement of RANKL in the expansion of mammary stem cells by progesterone, elevated either at pregnancy or cyclically, during the luteal dioestrus phase. It has been hypothesized that increased numbers and/or activity of mammary stem cells may predispose to mammary cancer by providing a greater target pool for transformation15,18. This suggests that RANKL inhibition may also potentially mediate non-proliferative changes (for example, reductions in mammary stem cells) in addition to the reduced proliferation and delayed onset of pre-neoplasias observed in this study, which collectively reduces mammary tumorigenesis. Progesterone has been linked to the aetiology of human breast cancer due to its mitogenic effects19. The mammary epithelial proliferative index is highest during the luteal phase or in women receiving combined oestrogen plus progesterone hormone replacement therapy (HRT), as compared with women treated with oestrogen alone20. Significant increases in breast density21, breast cancer risk22 and recurrence23 have also been observed when progestins were included in HRT. Our data indicate that the pivotal role of progesterone in the promotion and growth of breast tumours may be mediated by RANKL-dependent effects. The observation of RANKL within human breast tumour stroma (that is, fibroblast-like cells and mononuclear cells), in addition to the tumour epithelium (this study and ref. 24), indicates that RANKL protein expression may not be modulated exclusively by hormone, and other mechanisms may exist to dysregulate and activate this pathway in breast cancer.

Methods Summary

Mouse experimental procedure

MMTV-RANK lines 1 and 2 were established (C57BL/6 background) and maintained as described6. Multiparous females were observed for the development of spontaneous tumours up to 29 months of age (>5 pregnancies for each animal). Induction of mammary tumours by MPA and DMBA were in accordance with the published protocol7. MMTV-RANK line 1 and wild-type C57BL/6 mice were treated with 10 mg kg−1 murine RANK-Fc (Amgen Inc.) or PBS subcutaneously three times per week. Zoledronic acid (Zometa) was dosed at 0.5, 0.2, or 0.025 mg kg−1 weekly. Treatment was initiated with the first DMBA treatment at 9 weeks of age and continued until mice were killed. MMTV-neu (N202 Mul; FvB background) mice were obtained from Jackson Laboratories. 10 mg kg−1 RANK-Fc or muFc (IgG1, both Amgen Inc.) were given three times per week beginning at 5 months of age and continued until mice were killed. Tumours were detected by manual palpation weekly.

Whole-mount analysis, histology and immunohistochemistry

For whole-mount analysis, both inguinal mammary glands were removed at the time specified (either 4 or 7 weeks after the last DMBA treatment), fixed and stained with carmine25. Dense epithelial foci of greater than 300 μm were counted. For histological and immunohistochemistry analysis, tissues were fixed in formalin or Zinc-Tris solution and embedded in paraffin. Sections (4 μm) were stained with haematoxylin and eosin, and assessed according to the Annapolis guidelines26. BrdU, cyclin D1, cleaved caspase 3, RANK and RANKL stained sections were visualized with diaminobenzidine and quantified using Aperio Spectrum software (Aperio Technologies). Dual RANKL- and PR-stained sections were visualized with alkaline phosphatase Ferangi blue (Biocare Medical) for RANKL and permanent red (Dako) for PR. Formalin-fixed, paraffin-embedded tumour specimens (57 samples of invasive carcinoma of human breast) were obtained (Asterand) and immunohistochemistry was performed with M366 (anti-huRANKL monoclonal antibody; Amgen Inc.) or N-1H8 and N-2B10 (anti-huRANK monoclonal antibodies; Amgen Inc.).

Online Methods

Mouse experimental procedure

All animal procedures were approved by and performed under the guidelines of Amgen’s Institutional Animal Care and Use Committee (IACUC). MMTV-RANK lines 1 and 2 (C57BL/6 background) were established and maintained at Amgen under specific pathogen-free conditions as previously described6. Multiparous females were observed for the development of spontaneous tumours up to 29 months of age (>5 pregnancies for each animal). Induction of mammary tumours by MPA and DMBA were in accordance with the published protocol7 in which 5–6-week-old female mice were implanted subcutaneously with a 90-day-release 50 mg MPA pellet (Innovative Research of America) and replaced at expiration. One milligram of DMBA (Sigma-Aldrich) was administered at weeks 9, 10, 12, 13 of age by oral gavage. Tumours were detected by manual palpation. For treatment studies, MMTV-RANK line 1 and C57BL/6 wild-type mice were treated with 10 mg kg−1 muRANK-muFc (Amgen Inc.) or PBS subcutaneously three times per week or zoledronic acid (Zometa) dosed at 0.5, 0.2, or 0.025 mg kg−1 weekly. Tumours were detected by manual palpation weekly and latency was expressed as weeks after last DMBA treatment. Mice were killed when a palpable mass exceeded 100 mm2 or the third MPA pellet expired. For initial tumorigenesis studies, wild-type and MMTV-RANK lines 1 and 2 mice were killed when a palpable mass exceeded >200 mm2 or when mice appeared moribund. MMTV-neu (N202 Mul; FvB background) mice were obtained from Jackson Laboratories. The unactivated neu proto-oncogene, under control of the MMTV promoter, results in mammary tumours of highly uniform morphology and lung metastasis after a long latency11. Ten milligrams per kilogram RANK-Fc or muFc (IgG1, both Amgen Inc.) were given three times weekly beginning at 5 months of age and continued until mice were killed. Tumours were detected by manual palpation weekly. All mice were injected intraperitoneally with 1 mg per 100 μl BrdU (Sigma) 2 h before euthanasia. For progesterone treatment experiments, mice were given subcutaneous progesterone (1 mg) in sesame oil daily. Mifepristone (10 mg kg−1 in sesame oil subcutaneously) was given daily at least 2 h preceding progesterone treatment. RANK-Fc was given at 10 mg kg−1 subcutaneously every 48 h; the first dose of RANK-Fc was administered 24 h before progesterone. Animals were killed and analysed for BrdU and RANKL immunohistochemistry 72 h after the first progesterone treatment. Demonstration that the 10 mg kg−1 dose of RANK-Fc optimally blocked MPA-induced mammary proliferation was confirmed by a dose–response comparison of 1, 3, 10, 30 and 100 mg kg−1 RANK-Fc in vivo (Supplementary Fig. 10). Treatment was initiated with the first DMBA treatment at 9 weeks of age and continued until mice were killed. Previous studies have established the specificity of RANK-Fc for RANKL in vitro, via multiple independent direct ligand-binding assays and in vivo, using either transgenic overexpression or pharmacological treatment with RANK-Fc27,28.

Whole-mount analysis, histology, immunohistochemistry and mRNA analysis of mammary tissue

Whole mounts of inguinal mammary glands were used to analyse gross morphological changes at the specified times (4 or 7 weeks after the last DMBA treatment) after fixation in Clarke’s solution and stained with carmine25. Dense epithelial foci of greater than 300 μm were counted. Haematoxylin and eosin stained sections (4 μm) of inguinal mammary glands from MMTV-RANK, wild-type, or MMTV-neu tumour-bearing mice were used to analyse for the presence of pre-neoplastic lesions and overall mammary epithelial density. Tumour histotype was assessed from haematoxylin and eosin stained tissue according to the Annapolis guidelines26. For immunohistochemistry, antigen heat retrieval with DIVA (Biocare Medical) was used on paraffin-embedded sections and sections were incubated with antibodies against muRANKL (AF462; R&D Systems), PR (Dako), CK14 (Covance), CK5 (Abcam 32118), CK18 (Chemicon B3457), cyclin D1 (Lab Vision clone SP4), cleaved caspase 3 (Cell Signaling Technologies catalogue no. 9661) or BrdU (Accurate). For RANK staining, tissues were fixed in Zinc-Tris solution, embedded in paraffin and antigen heat retrieval with DIVA (Biocare Medical) was used. Sections were incubated with muRANK antibody (AF692; R&D Systems). Antigen–antibody complexes were detected with streptavidin horseradish peroxidase (Vector Laboratories) and visualized with diaminobenzidine (Dako). Sections were counterstained with haematoxylin. Dual RANKL and PR stained sections were visualized with alkaline phosphatase Ferangi blue (Biocare Medical) for RANKL and permanent red (DAKO) for PR. The specificity of anti-muRANK and anti-muRANKL antibodies was confirmed by correlation of the immunohistochemistry signal with mRNA expression6 and, additionally for anti-RANKL, by demonstrating the absence of staining in skeletal tissues obtained from RANKL-deficient mice. For quantification of BrdU, active-caspase 3, cyclin D1, RANK, RANKL and PR immunostaining, slides were scanned at ×20 using an Aperio ScanScope (Aperio). Algorithms within Spectrum software (Aperio) were used at the default settings to score staining (minimum 500 cells per analysis). MPA-plus-DMBA-treated mammary tissues stained for cyclin D1 were analysed for the percentage (±s.e.m.) of elevated cyclin-D1-positive cells (2+ score) and quantified using Aperio Spectrum software. Total per cent cyclin-D1-positive cells was quantified in Aperio in MMTV-neu treated mice. Expression of survivin (Birc5), RANK and RANKL mRNA was determined via reverse transcriptase polymerase chain reaction (RT–PCR) and primer probe sets from Applied Biosystems.

Tumour transplant experiments

8-week-old female C57BL/6 mice were implanted with freshly prepared tumour fragments as follows. On experimental day 0, a donor mouse bearing a MMTV-RANK tumour that arose from a tumour fragment implantation was harvested and diced into approximately 2 mm × 2 mm fragments. An incision was made under ketamine/xylazine anaesthesia and a fragment was implanted into a pocket of the mammary fat pad. Mice were also implanted with a 90-day-release MPA pellet (50 mg per pellet, Innovative Research of America). Treatment began on day 0 and consisted of 10 mg kg−1 RANK-Fc or PBS three times per week subcutaneously and continued until termination of the study at day 53. All tumours from this study were harvested at day 53 into neutral buffered formalin for pathology analysis.

Cyclin D1 analysis of primary MECs from mammary glands

Cultured MECs from normal mammary glands were prepared as described6. RNA was harvested from MMTV-RANK MECs grown in three-dimensional cultures in differentiation media with or without RANKL for 24 h. Gene expression analysis of cyclin D1 was performed as described6. Cyclin D1 expression was assessed by immunohistochemistry 72 h after RANKL treatment.

Immunofluorescence analysis of primary MECs from mammary glands

Mammary glands were dissected from day 16.5 pregnant mice as described6. Briefly, tissues were minced in DMEM/F12 (Gibco) containing antibiotics, and incubated with 0.3% collagenase (Roche), and 2.5 U ml−1 dispase (Invitrogen) at 37 °C for 90 min with trituration every 30 min. Crude digests were filtered through 500 AM Nitex mesh (Technicon) to remove undigested clumps, and centrifuged at 100g to collect epithelial spheroids. Pellets were washed five times in DMEM/F12 with antibiotics and 5% fetal bovine serum (BCS) and collected each time with a low-speed spin. Spheroids were plated on 10-cm plates in growth medium (DMEM/F12 with antibiotics, 10 µg ml−1 insulin, 10 ng ml−1 EGF, and 5% BCS), for 2 to 5 days before being harvested for immunofluorescent staining (Cellomics) or for three-dimensional culture. The culture medium was changed every other day. For three-dimensional cultures, first passage primary mammary epithelial cells (MECs) were harvested by trypsinization. Cells were counted and replated in MatrigelR (Collaborative Research) coated onto 8-well chamber slides (LabTek) and growth medium was added. After 24 h, the medium was changed to differentiation medium (DMEM/F12 with antibiotics, liquid media supplement ITS (Sigma), 3 μg ml−1 prolactin (Sigma), 1 μg ml−1 hydrocortisone (Sigma) with or without murine RANKL-LZ (200 ng ml−1); medium was changed every day.

Acinar structures were stained as previously described6. MEC organoids were cultured in Matrigel in 8-well chamber slides. At day 8 the medium was removed and acini were fixed in 2% paraformaldehyde and permeabilized using 0.5% Triton X-100 before blocking. Cells were incubated with the primary antibodies rat anti-integrin α6 (1:150; Chemicon MAB1378) and rabbit anti ZO-1 (1:100; Zymed, 61-7300) overnight at room temperature and then with Alexa-conjugated secondary antibodies (Molecular Probes) for 45 min at room temperature and DRAQ5 (Alexis) for nuclear staining. Slides were mounted with Prolong Gold Antifade Reagent (Molecular Probes). Confocal analysis was performed using the Zeiss confocal microscopy system equipped with argon and HeNe lasers. Images were captured using LSMTM version 5 software (Zeiss).

Immunohistochemistry of human breast cancer samples

Anti-human RANK and RANKL immunohistochemistry was performed on sections from formalin-fixed, paraffin-embedded tumour specimens (57 samples of invasive carcinoma of human breast: 39 ductal carcinoma, 6 lobular, 3 mucinous, 9 non-specified) prepared with heat retrieval in Diva buffer or citrate buffer pH 6 for 20 min. Immunohistochemistry was performed with an automated staining method using Lab Vision or Dako Autostainers. The primary antibody concentration was 0.75 μg ml−1 for anti-huRANKL monoclonal antibody (M366; Amgen) and 5 μg ml−1 for anti-human RANK (N-1H8 and N-2B10) with goat anti-mouse secondary antibody. An isotype control mouse IgG1 and appropriate secondary antibody was performed for each specimen and each antibody. Antibody reactivity was visualized by diaminobenzidine enhanced by tyramide amplification. A formalin-fixed giant cell tumour of the bone and transfected cell lines were used as positive controls for the anti-RANKL and anti-RANK antibodies. Specificity of the anti-human RANKL and anti-human RANK monoclonal antibodies were further substantiated by flow cytometry and western blot analysis of positive and negative control cells. The low frequency of human tumours with RANKL-positive tumour epithelia precluded any correlative analysis with PR expression. RANKL-positive fibroblast-like stromal cells were only observed within a dense lymphoid environment. Antibody staining for RANK and RANKL was scored on a scale of 0 to 3 for intensity (0 = no staining, 1 = weak signal, 2 = moderate stain, 3 = intense stain) and for the percentage of positive cells. A complex score was used to analyse the epithelial component of tumours and was equal to the percentage of positive cells times the intensity score. A tumour was considered positive above a threshold of 10. For RANKL only the most intensely staining cells ( = 3) were counted.

Statistical analyses

The log-rank test was used to assess differences in kinetics of tumour formation. Fisher’s exact test was used for categorical analysis of differences in tumour phenotype. One-way ANOVA with Bonferroni’s post-test was used to evaluate differences in immunohistochemistry-quantified sections and to evaluate differences in other multiple group comparisons. For all other numerical comparisons, independent t-tests were used. For all tests, P < 0.05 was considered significant.