Cell lines used were as follows: human triple-negative MDA-MB-231 breast cancer cells (European Collection of Authenticated Cell Cultures, Wiltshire, UK) transfected with green fluorescent protein (GFP) and Luc2 (21) and triple-negative mouse mammary 4T1 and E0771 cells (kindly donated by Dr. Amy Kwan, University of Sheffield, Sheffield, UK, and Dr. Sandra McAllister, Harvard Medical School, USA, respectively) transfected with Luc2 (22, 23). MDA-MB-231 and 4T1 cell lines were maintained in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, UK) supplemented with 10% fetal bovine serum (FBS; Gibco, UK), whereas E0771 cells were maintained in RPMI (Gibco, UK) supplemented with 10% FBS. MDA-MB-231-GFP-Luc2 was used within 30 passages, and 4T1-Luc2 and E0771-Luc2 were used within 20 passages after receipt.

Here, 5 × 10⁴ 4T1-Luc2 cells were grown in DMEM + 10% FBS, washed, and prepared for intracardiac (i.c.) injection into 12-week-old female BALB/c mice (n = 5) (Charles River, UK). Tumour growth was monitored by an IVIS Lumina II system (Calliper Life Sciences, UK). Animals were sacrificed 2 weeks after injection of tumour cells. All tumour-bearing bones were isolated in a sterile environment and dissociated aseptically to collect 4T1-luc cells using Miltenyi tumour cell dissociation kit (Miltenyi Biotech, Germany). Selection of 4T1-luc cells was carried out using 60 μmol/L 6-thioguanine. Cultured colonies were grown and used for further in vivo experiments.

Experiments used 12-week-old female BALB/c wild-type, BALB/c ^fox/- nude, and C57BL/6 mice (Charles River, Kent, UK) kept on a 12-h/12-h light/dark cycle with free access to food and water. All studies were carried out in accordance with local guidelines and with Home Office approval under project licences 70/8964 and P99922A2E, University of Sheffield, UK.

Modelling of human pre-, peri-, and post-menopausal concentrations of oestradiol only was established by ovariectomising (OVX) mice to negate effects of other ovarian hormones, i.e., inhibin [as previously described (24, 25)] prior to supplementation with 17β-oestradiol (Sigma Aldrich, Poole, UK) at 12.5, 1.38, or 0 mg/L via their drinking water. To prevent predicted release of oestradiol from other tissues including adipose tissue as well as increases in FSH, 40 µg/kg/day goserelin (as goserelin acetate, Sigma Aldrich, UK) was administered via subcutaneous injection. Success of OVX ± oestradiol supplementation was assessed by imaging and weighing of mouse uterus at the end of each procedure ( Supplementary Figure S1 ) and checking serum concentrations of oestradiol and FSH by ELISA. For experiments designed to investigate the effects of Zol on the bone microenvironment in the absence of tumours, 100 µg/kg Zol (supplied as disodium salt by Novartis) or phosphate buffered saline (PBS; control) was administered once per week starting 7 days after OVX (n = 5/group). Mice were sacrificed 14 days after OVX, and serum was collected and stored at -80°C for downstream analysis by ELISA, tibiae and femora were fixed in 4% paraformaldehyde (PFA; Fisher Scientific, UK) for μCT analysis before decalcification in 0.5 M EDTA and processing for histology ( Supplementary Figure S2A ).

To investigate the effects of Zol on tumour distribution under established pre-, post-, and peri-menopausal concentrations of oestradiol, 1 × 10⁵ MDA-MB-231-GFP-Luc2 cells were injected into the left cardiac ventricle (i.c.) of BALB/c^fox/- nude mice 3 days after OVX ± oestradiol + goserelin; or 5 × 10⁴ 4T1-Luc2 cells were injected i.c. into BALB/c mice 6 days following OVX ± oestradiol + goserelin; or C57BL/6 mice were injected with 1 × 10⁵ E0771-Luc2 i.c. 7 days following OVX ± oestradiol + goserelin. At 3 days after injection of tumour cells, 100 µg/kg/week Zol or PBS (control) was injected subcutaneously (s.c.) to the corresponding groups (n = 10/group) until the end of the experiment. BALB/c ^fox/- nude, BALB/c, and C57BL/6 mice were sacrificed 28, 15, and 19 days after OVX, respectively. Serum was collected and stored at -80°C for downstream analysis by ELISA, tibiae and femora were fixed in 4% PFA for μCT analysis before decalcification in 0.5 M EDTA and processing for histology ( Supplementary Figures S2B, C ).

Tumour growth was monitored, in live mice, using an IVIS Lumina II system (Caliper Life Sciences, UK). Here, 30 mg/kg of D-Luciferin (Invitrogen, UK) was injected s.c. 4 min before imaging. Mice were imaged by dorsal and ventral exposure. Hind limbs and internal organs were extracted for additional ex vivo imaging after termination of the experiment. All images were analysed by creating a region of interest (ROI) around the tumour(s); luminescence signal was acquired in radiance (photons/second).

Microcomputed tomography (μCT) analysis was carried out on fixed tibiae using a Skyscan 1172 x-ray–computed microtomography scanner (Skyscan, Aartselaar, Belgium) equipped with an x-ray tube (voltage, 49 kV; current, 200 μA) and a 0.5-mm aluminium filter. Pixel size was set to 4.3 μm, and scanning initiated from the top of the proximal tibia as previously descried (26).

Tibiae and femora were fixed in 4% PFA for a minimum of 48 h prior to decalcification with 0.5 M EDTA at pH 7.0 for 3 weeks, changing the solution at weekly intervals. Bones were embedded in paraffin wax, and sections of 3 µm were taken using a microtome. Osteoclast identification was carried out by tartrate-resistant acid phosphatase (TRAcP) for consistency staining. Osteoblasts were identified as mononuclear cuboidal cells residing in chains along the bone surface. The number of osteoclasts and osteoblasts was analysed per millimetre in the cortical–endosteal bone and trabecular bone surfaces. Leica RMRB upright microscope and OsteoMeasure software (Osteometrics Inc.) were used for osteoclast and osteoblast scoring (27).

Serum concentrations of Tartrate-Resistant Acid Phosphotase 5b (TRAcP 5b), procollagen 1 intact N-terminal propeptide (P1NP), oestradiol, and FSH were measured using commercially available ELISA kits: MouseTRAP™ Assay (Immunodiagnostic Systems), Rat/Mouse P1NP competitive immunoassay kit (Immunodiagnostic Systems), Mouse/Rat oestradiol ELISA (CALBIOTECH), and Mouse FSH (Elabsciences) and standard protocols, respectively.

Total RNA was extracted using an RNeasy kit (Qiagen) and manufacturers’ instructions. For Affymetrix array analysis, RNA was amplified using LunaScript RT SuperMix Kit (New England BioLabs, USA) before being reverse transcribed into cDNA using Superscript III (Invitrogen AB). Gene expression was assessed by analysing the number of cycles (CT) taken for expression of glyceraldehyde 3-phosphate dehydrogenase (GAPDH; control) musculoaponeurotic fibrosarcoma (MAF), macrophage capping protein (CAPG), and PDZ domain containing protein GIPC1 (GIPC1) to be detected using the following assays: Hs00894322 for GAPDH, Hs01533575 for MAF, Hs 05867157 for CAPG, and Hs Hs01149347 for GIPC1 in human MDA-MB-231 cells and Mm00186822 for GAPDH, Mm0300214 for MAF, Mm00179028 for CAPG, and Mm00252127 for mouse 4T1 and E0771. All PCR assays were performed using an ABI 7900 PCR System (Perkin Elmer, Foster City, CA) and Taqman universal master mix (all reagents were purchased from Applied Biosystems via Thermofisher, UK).

Statistical analysis was performed using Prism GraphPad (version 8.0). Data are represented as mean ± SEM. Statistical significance was measured using parametric testing, assuming equal variance. Analysis was by one-way analysis of variance (ANOVA) followed by Newman–Keuls multiple comparison test, and standard unpaired t-tests were used to assess the difference between test and control samples. Statistical significance was defined as p ≤ 0.05. All graphs represent mean ± SEM, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Our hypothesis is that oestradiol-induced changes to the bone microenvironment affect metastatic outgrowth in bone and soft tissue. As female mice have low circulating concentrations of oestradiol compared with humans [65.94 ± 25.28 pg/ml in 14-week-old mice ( Supplementary Figure S3 ) compared with 40–400 in premenopausal women] and do not naturally undergo menopause, we used OVX to establish low baseline concentrations of oestradiol and inhibin. This procedure did not reduce oestradiol release from other endocrine organs; therefore, mice also received 40 μg/kg/day goserelin to further reduce circulating concentrations of oestradiol to 1.01 ± 0.27 pg/ml ( Supplementary Figure S3 ) and prevent OVX-induced release of FSH ( Supplementary Figure S4 ). Mice were subsequently supplemented with 17β-oestradiol in their drinking water to establish concentrations of this hormone that are representative of pre-, post-, and peri-menopausal concentrations observed in women. Supplementation with 12.5, 1.38, or 0 mg/L resulted in serum concentrations of 153.16 ± 18.10, 48.64 ± 18.44, and 1.00 ± 0.27 pg/ml, respectively ( Figure 1A ). These concentrations are comparable to the 40–400, 22–120, and >30 pg/ml found in pre-, peri-, and post-menopausal women. Oestradiol concentrations were not significantly altered following administration of Zol ( Figure 1A ), and serum FSH concentrations remained below the level of detection by ELISA in all mice ( Supplementary Figure S4 ). Removing oestrogen by OVX and administration of goserelin caused withering of mouse uterus and a significant reduction in uterus weight (p < 0.001); this was restored following supplementation with premenopausal concentrations of oestradiol but not with perimenopausal concentrations of this hormone. Uterus appearances and weights were not altered following administration of Zol ( Supplementary Figure S1 ).

Oestradiol had significant effects on trabecular bone volume: BV/TV% was higher in the trabecular bone in tibiae of mice following exposure to premenopausal concentrations of oestradiol compared with peri- or post-menopausal concentrations (p < 0.001); BV/TV% was also higher in trabecular bone of tibiae exposed to perimenopausal concentrations of oestradiol compared with pre- or post-menopausal concentrations (p < 0.001 and p < 0.01, respectively) ( Figure 1B ). Administration of Zol caused an increase in trabecular BV/TV% under all concentrations of oestradiol; however, under peri- and post-menopausal concentrations (BV/TV = 13.97% ± 1.13% and 11.45% ± 1.99%, respectively), Zol did not increase bone volume to levels observed under premenopausal oestrogen conditions (BV/TV = 18.70% ± 1.48%; Figure 1B ). These data suggest that oestradiol has a more potent bone-building effect than Zol in the trabecular area of bone. Interestingly, differences in trabecular bone volume observed in OVX mice following supplementation with pre-, peri-, and post-menopausal concentrations of oestradiol were not due to changes in trabecular thickness ( Figure 2A ). Instead, oestradiol appeared to directly influence trabecular number; mice supplemented with premenopausal concentrations of oestradiol exhibited significantly increased trabecular number compared with mice supplemented with perimenopausal (p < 0.001) or postmenopausal concentrations of oestradiol (p < 0.0001). In contrast, administration of Zol increased trabecular thickness in mice supplemented with pre- (p < 0.0001) and peri- (p < 0.001) menopausal concentrations of oestradiol ( Figure 2A ) but did not alter trabecular number ( Figure 2B ), suggesting that oestradiol and Zol affect bone formation via different mechanisms. Trabecular separation was not affected by either oestradiol or Zol ( Figure 2C ). Oestradiol significantly impacted the quality of trabecular bone; trabecular pattern factor was significantly lower under premenopausal concentrations of oestradiol compared with peri- (p < 0.0001) or post- (p < 0.0001) menopausal oestradiol. Interestingly, Zol improved the quality of bone under premenopausal concentrations of oestradiol (p < 0.0001) but had no affect on trabecular pattern factor in trabecular bone under peri- or post-menopausal concentrations of oestradiol ( Figure 2D ).

Effects on cortical bone were less pronounced; cortical thickness was slightly, but not significantly, reduced in mice with peri- and post-menopausal concentrations of oestradiol compared with mice with premenopausal concentrations of oestradiol (cortical bone thickness = 0.047 ± 0.01 mm, 0.47 ± 0.02 mm, and 0.53 ± 0.02 mm, respectively). Furthermore, Zol did not increase cortical bone volume when administered under premenopausal oestradiol but increased cortical volume back to control levels when administered under peri- and post-menopausal oestradiol (cortical bone thickness = 0.55 ± 0.02 mm, 0.57 ± 0.01 mm, and 0.55 ± 0.02 mm, respectively) ( Figure 2E ).

Reduced bone volume observed under postmenopausal concentrations of oestradiol appears to be driven by elevated bone turnover with increased osteoclast number and activity ( Figures 1C, D ) and increased activity of osteoblasts ( Figure 1F ). Alternatively, increased bone volume following administration of Zol was primarily driven by reduced osteoclast activity with only small changes in osteoblast numbers being detected but no changes in activity ( Figures 1C – F ). Interestingly, under postmenopausal concentrations of oestradiol, representing conditions in which adjuvant Zol exerts maximum antimetastatic effects in breast cancer patients, administration of Zol reduced the activity of both osteoblasts and osteoclasts ( Figures 1C, D ), implying that Zol may exert increased antiresorptive effects in trabecular bone when oestradiol is low.

Clinical trials have demonstrated that adjuvant Zol reduced subsequent development of tumour relapse in bone independently of menopausal status or serum oestradiol concentrations (16, 17, 19). We therefore used a variety of in vivo models of bone metastasis to see if we could mimic this effect in mice ( Figure 3 ). In immunocompromised BALB/c nude mice, oestradiol had no effect on the number of mice developing tumours in bone following i.c. injection of human MDA-MB-231 cells ( Figure 3Bi ) nor did it affect the volume of these tumours ( Figure 3Ci ). Under immunocompromised conditions, there was a trend toward a reduced number of mice developing bone metastasis in Zol-treated mice, and Zol reduced the size of metastatic deposits under peri- and post-menopausal concentrations of oestradiol ( Figures 3Bi, Ci ). Similarly, in immunocompetent BALB/c mice, treatment with Zol resulted in a trend toward a reduced number of bone metastasis following i.c. injection with mouse mammary 4T1 cells; this did not reach significance ( Figure 3Bii ). However, in this model, oestradiol concentration correlated with increased tumour burden, with smaller tumours detected in the bones of mice supplemented with pre- and peri- menopausal concentrations of oestradiol compared with no-oestradiol (p < 0.1). This increase in tumour volume was reduced following administration of Zol (p < 0.1) ( Figure 3Cii ). Interestingly, the opposite effect was observed in immunocompetent C57BL/6 mice injected with mouse mammary E0771 cells. In C57BL/6 mice, circulating concentrations of oestradiol positively correlated with both number and size of bone metastases. Mice supplemented with premenopausal concentrations of oestradiol developed significantly more bone metastases compared with mice not supplanted with oestradiol (p < 0.01; Figure 3Biii ). In this model, Zol did not reduce bone metastases, instead, under postmenopausal concentrations of oestradiol, mice developed significantly more bone metastases in the Zol-treated compared with the non-Zol-treated groups (p < 0.01; Figure 3Biii ). These data suggest that Zol has the ability to reduce the outgrowth of metastatic tumour cells in bone, but this is dependent on additional factors including genetic background of the host/tumour.

It was hypothesised that under premenopausal conditions, high concentrations of oestradiol may cause tumour cells to move out of the bone microenvironment, driving increased tumour recurrence in soft tissue. Our data show a trend toward reduced numbers of tumours in the soft tissues of mice supplemented with peri- or post-menopausal concentrations of oestradiol compared with mice supplemented with premenopausal concentrations of oestradiol especially in immunocompetent BALB/c and C57BL/6 mice ( Figure 4 ). Treatment with Zol had no effect on the number of tumours that developed in mice but did affect tumour volume ( Figure 4B ).

In immunocompromised BALB/c nude mice, oestradiol had no effect on tumour volume, and metastatic tumour growth outside of the skeleton was unaffected by treatment with Zol ( Figure 4Bi ). However, in both immunocompetent BALB/c and C57BL/6 models, oestradiol concentrations inversely correlated with metastatic tumour growth outside of bone. In both of these models, significantly larger tumour volumes were recorded in mice supplemented with postmenopausal concentrations of oestradiol compared with premenopausal concentrations of oestradiol (p < 0.1; Figures 4Bii, iii ). BALB/c mice treatment with Zol had no effect on tumour growth outside of the skeleton under premenopausal concentrations of oestradiol but significantly reduced tumour volume under postmenopausal concentrations of this hormone (p < 0.01; Figure 4Bii ). In C57BL/6 mice, treatment with Zol significantly increased tumour volume outside the skeleton (p < 0.05) in mice supplemented with premenopausal concentrations of oestradiol but reduced tumour volume under postmenopausal concentrations of oestradiol (p < 0.05; Figure 4Biii ). These data suggest that low circulating concentrations of oestradiol stimulate growth of disseminated tumour cells outside of bone and that Zol is only able to exert antitumour effects on extraskeletal metastasis in low oestradiol conditions. Importantly, these effects appear to be dependent on regulation by the adaptive immune system, as neither oestradiol nor Zol affected extraskeletal metastasis in athymic BALB/c nude mice.

Previously published studies have indicated that antimetastatic effects of Zol correlate with that tumour cell expression of the transcription factor MAF (28), whereas other studies have shown that high expression of GIPC1 and CAPG is associated with antitumour activity of Zol in breast cancer patients (24). We therefore investigated the expression profile of these genes in our breast cancer models. Using Real Time QPCR, MAF was not detected within 30 amplification cycles in bone homing, MDA-MB-231, 4T1, or E0771 cells, indicating that this gene is not expressed in these cell lines. In contrast, CAPG and GIPC1 are highly expressed in the bone homing clones used for current experiments, with no significant differences detected ( Figure 5 ). GIPC1 and CAPG were detected at 15.22 ± 3.12 and 21.80 ± 1.15 cycles, respectively, compared with 15.22 ± 3.12 for GAPDH in MDA-MB-231 cells; in 4T1 cells, GAPC1, CAPG, and GAPDH were detected at 17.14 ± 2.11, 20.34 ± 2.55, and 16.18 ± 2.33 cycles, respectively; and in E0771 cells, GAPC1 and CAPG were detected after 19.225 ± 3.112 and 23 ± 3.562 cycles compared with 18.21 cycles for GAPDH.

Evidence from the literature suggests a strong link between bone turnover and metastatic outgrowth in bone (25, 29–31). We therefore investigated whether differential effects of oestradiol and Zol on bone metastasis observed between BALB/c/BALB/c nude and C57BL/6 mice ( Figure 3 ) could be explained by effects of this hormone/drug on bone turnover. As previous experiments showed that Zol does not affect oestradiol-induced changes in cortical bone ( Figure 2E ) but has profound effects on oestradiol-induced changes in trabecular bone ( Figures 1 , 2 ), which is also the area in which tumours form in mouse models of bone metastasis, we focused our analysis on trabecular bone volume. μCT analysis of trabecular bone from the tibiae of mice revealed that administration of premenopausal concentrations of oestradiol had similar effects on all mouse strains, resulting in BV/TV% of 20.36 ± 1.06 in BALB/c nude, 21.82 ± 0.91 in BALB/c, and 20.22 ± 2.0 in C57BL/6 ( Figure 6 ). Reducing oestradiol had significant effects on bone turnover, with the greatest effects observed in BALB/c nude mice in which perimenopausal concentrations of oestradiol reduced trabecular bone volume to 6.38% ± 0.45%, and this remained at a similar volume (8.08% ± 0.70%) under postmenopausal oestradiol. Immunocompetent BALB/c and C57BL/6 exhibited a more gradual response to changes in oestradiol, with significant reductions in bone volume being recorded between mice supplemented with pre- and peri-menopausal concentrations of oestradiol (p < 0.0001 for BALB/c and C57BL/6) and further reductions observed between mice supplemented with peri- and pre-menopausal concentrations of this hormone (p < 0.001 and p < 0.01 for BALB/c and C57BL/6, respectively). Interestingly, in tumour-bearing mice, administration of Zol only increased bone volume in C57BL/6 mice under postmenopausal concentrations of oestradiol ( Figure 6 ). In immunocompromised BALB/c nude and immunocompetent BALB/c mice, reduced bone volume observed in mice with peri- and post-menopausal concentrations of oestradiol compared with premenopausal concentrations of this hormone correlated with increased activity of osteoblasts ( Figures 7Ai, ii ) and osteoclasts ( Figures 7Bi, ii ), indicating increased bone turnover. Oestradiol did not appear to alter activity of either osteoclasts or osteoblasts in C57BL/6 mice on termination of this experiment ( Figures 7Aiii, Biii ). Interestingly, administration of Zol reduced osteoblast activity under pre- and peri-menopausal concentrations of oestradiol in BALB/c nude (p < 0.001 and p < 0.01 for pre- and peri-menopausal, respectively) and BALB/c (p < 0.01 and p < 0.001 for pre- and peri-menopausal, respectively) mice. Conversely, in C57BL/6 mice, Zol reduced osteoblast activity under peri- and post-menopausal concentrations of oestradiol (p < 0.001 and p < 0.001, respectively; Figure 7A ). Zol significantly reduced osteoclast activity in both background strains of immunocompetent and immunocompromised independently of circulating concentrations of oestradiol ( Figure 7B ). These differences observed in the regulation of osteoblast activity may, in part, contribute to the opposing antitumour effects observed in the bones of BALB/c and C57BL/6 mice following administration of Zol ( Figure 3 ).