Forty female Sprague-Dawley rats, 73 days of age (Harlan-Teklad, Indianapolis, IN, USA) were randomised by weight into two groups of 20. Rats in the tamoxifen group were injected subcutaneously daily with 1 mg/kg tamoxifen (Sigma, St. Louis, MO, USA) dissolved in ethanol and suspended in sesame oil. The vehicle control group was injected daily with an equal volume of ethanol/sesame oil solvent. After 30 days of treatment, lymph-node-free right inguinal mammary glands were harvested, snap frozen and stored at -80°C for subsequent ECM isolation and biochemical analyses. To limit variations in ECM composition that may accompany oestrous cycling in the control group, phase of oestrous cycle was determined by daily cervical lavages and all control rats were sacrificed in the dioestrus 1 phase of the cycle, as previously described 29. Cervical tissue was also collected for histological confirmation of oestrous cycle. The animal experiments were performed in duplicate. All animal procedures were performed in compliance with the AMC Cancer Research Institute and University of Colorado Denver Animal Care and Use Committees and National Institutes of Health (NIH) Policy on Humane Care and Use of Laboratory Animals.

Left inguinal mammary glands (lymph nodes removed) were processed for fibroblast isolation immediately after sacrifice using a protocol provided by Kornelia Polyak which has been previously reported 30. Briefly, the tissue was minced and incubated in collagenase. After incubation, the tissue was processed through a series of centrifugations to separate fibroblasts. Isolated cells were confirmed to be fibroblast-like when they tested negative for pan-cadherin and pan-cytokeratin and positive for vimentin by Western blot analyses (data not shown), and are referred herein as fibroblasts; however, endothelial cell specific markers were not evaluated. Fibroblasts, used at passage three, were maintained in Dulbecco's modified eagle medium (DMEM)/F12 media (Hyclone, Logan, UT, USA) supplemented with 25 mM Hepes (Sigma, St. Louis, MO, USA), 2 mM L-glutamine (Sigma, St. Louis, MO, USA), 50 μg/mL gentamicin (Sigma, St. Louis, MO, USA) and 20% FCS (Sigma, St. Louis, MO, USA).

Transwell 8 μm pore filters in a 24-well plate format (Becton Dickinson, Franklin Lakes, NJ, USA) were overlaid with 50 μL bovine gelatin, type B (Sigma, St Louis, MO, USA) at a concentration of 10 μg/mL and dried overnight. Five × 10⁴ log-phase primary fibroblasts at passage three were suspended in incomplete media and plated into the upper chamber. The lower chamber contained 1% fetal bovine serum as chemoattractant. The number of motile cells, evaluated four hours after plating, was quantified as previously described 15. The assay was performed in quadruplicate and data are expressed as mean ± standard error of the mean (SEM).

Primary mammary fibroblasts were pooled according to group, and 2 × 10⁵ cells were plated into six-well plates pre-coated with 50 μl of 2 μg/ml rat tail collagen (BD Biosciences, Bedford, MA, USA). Fibroblasts were cultured for 10 days in complete media, which was changed every two days. To isolate underlying ECM, fibroblasts were lysed in 20 mM ammonium hydroxide and 0.5% Triton X-100 followed by several washes in PBS and water, according to the published protocol by Cukierman 31. The cell-free underlying ECM was scraped from plates and stored at -80°C for proteomic analysis.

Whole mount mammary glands were fixed in methacarn, processed in a series of ethanol and xylene (Fisher, Fair Lawn, New Jersey, USA), stained with alum carmine (Fisher, Fair Lawn, New Jersey, USA), and bagged in methyl salicylate (Fisher, Fair Lawn, New Jersey, USA), as previously described 32. Cervical tissue was also harvested, fixed in 10% neutral buffered formalin (NBF) for 18 hours, paraffin embedded, cut into 5 μm sections and stained with H&E. To control for morphological differences due to natural variation along the proximal/distal mammary gland axis, mammary tissue associated with the dissected lymph regions of gland number 4 were fixed in 10% NBF for 18 hours and processed for H&E staining as described above. The changes in mammary ductal and alveolar morphology in response to tamoxifen treatment were evaluated in these 5 μm H&E-stained sections as previously described 32.

For detection of fibrillar collagen, 5 μm sections of mammary tissue were stained with sirius red F3B according to published methods 33 and counterstained with Weigert's iron haematoxylin. Mammary gland collagen staining was classified into three grades. Grade 1 glands had fat pads composed primarily (more than 50%) of adipocytes and were free of dense interlobular collagen. Grade 2 glands had mixed fat pad morphology, with both adipocyte and fibrillar collagen-rich regions. Grade 3 glands had stroma dominated by fibrillar collagen. Collagen scoring was performed on coded slides by two independent investigators.

Samples of 4 μm tissue were pretreated in 10 mM sodium citrate at 90°C for 20 minutes. Bromodeoxyuridine (BrdU; BD Immunocytometry Systems, San Jose, CA, USA, for rat tissue; Dako, Santa Barbara, CA, USA, for mouse tissue) and the macrophage marker CD68 (Serotec, Oxford, UK) were detected with a standard avidin biotin complex-peroxidase method with 3,3'-diaminobenzidine as the chromagen.

Mammary ECM isolation was performed based on a previously described protocol 5, using pooled inguinal mammary gland tissue from at least six rats per group. Briefly, frozen inguinal mammary glands, with lymph nodes removed, were pulverised, homogenised and extracted in a high salt/N-ethylmaleimide solution (3.4 M sodium chloride, 50 mM Tris-hydrochloric acid pH 7.4, 4 mM ethylenediaminetetraacetic acid (EDTA), 2 mM N-ethylmaleimide) containing protease inhibitor cocktail (100 μg/ml phenylmethylsulphonyl fluoride, 50 μg/ml each of aprotinin, leupeptin and pepstatin), at 4°C (chemicals purchased from Sigma, St Louis, MO, USA). Homogenates were enriched for ECM by two cycles of centrifugation (relative centrifugal force (RCF)_max 110,000 × g, 30 minutes, 4°C). The ECM-enriched pellets were resuspended in mid-salt/urea solution with proteinase inhibitor cocktail and extracted overnight at 4°C. Samples were pelleted at RCF_max 110,000 × g, and the ECM-enriched supernatants extensively dialysed (MWCO 12–14,000 kDa, Spectrum) against low salt buffer followed by dialysis against sera-free media (DMEM/F12 media (Sigma, St Louis, MO, USA) supplemented with 1 μg/ml gentamicin) at 4°C. ECM were stored on ice at 4°C and used within two weeks of isolation. As reported previously, ECM protein integrity is stable under these storage conditions 5. For the in vitro studies, all experiments were performed with two distinct batches of ECM in at least duplicate.

Aliquots of the mammary ECM isolates were subjected to in-solution tryptic digestion for label-free mass spectrometry-based analysis. Each sample was reduced and alkylated with 5 mM dithiothreitol and 20 mM iodoacetamide after the addition of Rapigest (Waters Technologies, Millford, MA, USA) to 0.1%. Trypsinisation was carried out overnight at 37°C and the reaction stopped by adding formic acid to a final concentration of 1%. Centrifugation was used to remove the hydrolytic detergent and sample zip tipped (Millipore, Temecula, CA, USA) to remove salts.

For electrospray ionisation mass spectrometry, each of the samples (2 μL) was injected onto a reverse-phase column using a chilled (9°C) autosampler (Eksigent; Dublin, CA, USA) connected to a high-performance liquid chromatography system run at 0.12 μL/minute before a passive split that resulted in about a 400 nL/minute post-split flow (Aligent; Santa Clara, CA, USA). A gradient of 12% to 30% acetonitrile (40 minutes) was applied over a two-hour run for peptide separation. The column effluent was coupled directly to a LTQ XL Linear Ion Trap mass spectrometer (Thermo/Finnigan; San Jose, CA, USA) with an in-house built nanospray ion source. Data acquisition was performed using the Xcalibur (version 2.0.6) software supplied with the instrument. The 60 minute liquid chromatography (LC) runs were monitored by sequentially recording the precursor scan (MS) followed by three collision-induced dissociation acquisitions (MS/MS). Singly charged ions were excluded from collision induced disassociation (CID) selection. Normalised collision energies were employed using helium as the collision gas. An in-house script was used to create de-isotoped centroided peak lists from the raw spectra (mgf format). These peak lists were searched against the SwissProt (V51.6) database using an in-house developmental Protein Prospector LC Batch-Tag Web (Version 4.25.2, UCSF, San Francisco, CA, USA) and an in-house Mascot server (Version 2.2, Matrix Science, Boston, MA, USA). For searches mass tolerances were ± 0.6 Da for MS peaks (+1 ¹³C option), and ± 0.6 Da for MS/MS fragment ions. Trypsin specificity was used allowing for one missed cleavage. The modifications of met oxidation, protein N-terminal acetylation, peptide N-terminal pyroglutamic acid formation were allowed for. Samples were searched against all rat and mouse entries in the protein database.

Tissues from mammary glands were analysed by Western Blot as previously described 5. The following antibodies were used: polyclonal rabbit anti-rat fibronectin (1:250 Life Technologies, Gaithersburg, MD, USA), polyclonal rabbit anti-laminin (1:500 Novus Biologicals, Littleton, CO, USA), mouse anti-laminin 5 (γ2 chain) monoclonal antibody (1:1000, Chemicon International, Temecula, CA, USA), collagen 1 (1:1000, Abcam, Cambridge, MA, USA) and protein A secondary antibody (1:10,000 Amersham, Piscataway, NJ, USA). Monoclonal mouse anti-glyceraldehyde 3-phosphate dehydrogenase (GAPDH; 1:500 Amersham, Piscataway, NJ, USA) followed by an anti-mouse secondary (1:5,000 Santa Cruz Biotechnology, Santa Cruz, CA, USA) was used for protein loading controls. Signal was obtained using ECL western detection kit (Amersham, Piscataway, NJ, USA). For fibronectin and collagen 1 Westerns, 3.3 μg of respective tissue samples were loaded per lane. For laminin (LN) westerns and zymogen assays, 9.5 μg of respective tissue samples were loaded per lane.

Substrate gel analyses were performed as described 15. Briefly, ECM samples loaded by equal protein were electrophoresed under non-reducing conditions on a 7.5% SDS-PAGE containing 3 mg/ml porcine gelatin (Sigma, St. Louis, MO, USA). Gels were incubated at 37°C for 72 hours in substrate buffer. Proteinase activity was visualised by Coomassie Blue 250 staining. Zymogen activity appears as a cleared band in a dark background. Semi-quantitative data was obtained by scanning densitometry of four independent lanes per condition with data analysed using BioRad Quantity One software (Bio-Rad, Hercules, CA, USA).

J774 murine macrophage cells, a generous gift from Dr Douglas Graham, were cultured in DMEM/high glucose medium (Hyclone Laboratories, Logan, Utah, USA) supplemented with 10% heat-inactivated FCS as previously described 34. MCF-12A cells are a non-tumourigenic human immortalised mammary epithelial cell line 35. V12 Ras-transformed MCF-12A cells were previously described 36. MCF-12A and MCF-12A-ras cells were grown in complete media consisting of Ham's F12/DMEM (Gibco, Carlsbad, CA, USA) containing 100 ng/ml cholera toxin (Gibco/BRL, Carlsbad, CA, USA), 0.5 μg/ml hydrocortisone (Sigma, St. Louis, MO, USA), 10 μg/ml insulin (Sigma, St. Louis, MO, USA), 20 ng/ml epidermal growth factor (EGF) (Sigma, St. Louis, MO, USA) and 5% horse serum (Gibco/BRL, Carlsbad, CA, USA). MDA-MB-231 cells (ATCC, Manassas, VA, USA), a human breast cancer cell line, were passaged into nude mice (mammary fat pad) and back out to plastic for at least four cycles as previously described 17. The resulting variant cell line, which was enriched in the ability to grow in the fat pad was carried in MEM Alpha Media (Gibco, Carlsbad, CA, USA) completed with 2.2 g/L of sodium bicarbonate, 1% Hepes, 1% L-glutamine, 10% heat inactivated FCS, 1 μg/ml of insulin, 1% sodium pyruvate and 1% non-essential amino acids.

Ras-transformed MCF-12A and MDA-MB-231 cells were cultured in 3D culture as previously described 5. Briefly, log-phase cells were harvested and overlaid onto 2 mm thick matrix pads at cell concentrations of 4.5 × 10⁴ (MCF-12A) or 1.5 × 10⁴ (MDA-MB-231) or pre-coated with respective matrices with and without 20 μg/mL fibronectin (Collaborative Biochemical, Bedford, MA, USA) before overlaying onto the matrix pad, using a 96-well format (Sarstedt, Newton, NC, USA). The matrix substratum was composed of 50 μL of nulliparous or tamoxifen-treated rat mammary gland ECM (normalised for total protein concentration) mixed 1:1 with Matrigel (BD Biosciences, San Jose, CA, USA) to facilitate polymerisation of endogenous mammary ECM. For control conditions, the pad was composed of Matrigel, without endogenous mammary ECM and controlled for total protein concentration. For MCF-12A cells, 5% horse serum was added to the matrix pads. 3D culture assays were performed in triplicate and each experiment performed in duplicate with two different batches of endogenous mammary ECM.

Five micron sections of 3D organoids capturing top, middle and bottom representative areas of the matrix pad were H&E stained and imaged at 20× using an Aperio Scan Scope T3 System, (Vista, CA, USA) at a resolution of 1 pixel/0.5 u and then down-sampled to a resolution of 1 pixel/2.4 u. Primary identification of cell clusters (organoids) was accomplished using NIH ImageJ (version 1.41o with Java 1.6.0_10.) that had been customised for colorimetric, size and circularity threshold gating steps using custom written plug-ins in the Java programming language. These image analysis algorithms were designed to identify and quantify immunohistochemical and morphological characteristics of 3D organoids. Image pre-processing procedures were applied to identify and remove from subsequent analyses images with scanning and staining errors. Images that passed these filters were analysed using a series of algorithms. The primary image algorithms relied on identifying histochemical features through colorimetric thresholds. Secondary pass morphometric quantifications focused on area and circularity measurements for feature identification. This automated algorithmic-targeted image analysis was utilised to eliminate intra-observer and inter-observer variability and results in a repeatable mathematically quantifiable set of data. The number of organoids analysed was 1456 in the Matrigel group, 1516 in the control mammary ECM group and 709 in the tamoxifen mammary ECM group. Mean cell cluster sizes were compared across treatment groups. Given that the data were not normally distributed, significance was evaluated using the Kruskal-Wallis, non-parametric test. Alternatively, 308 tamoxifen organoids and 241 tamoxifen plus fibronectin organoid images were captured under light microscopy and NIH ImageJ (version 1.41o with Java 1.6.0_10.) utilised to determine the average area per organoid, and differences determined using the Mann-Whitney U Test.

Forty eight-week old female homozygous Nu/Nu athymic nude mice (NCI, Frederick, MD, USA) were randomised by weight into two groups of 20. The animals were anaesthetised using isoflurane (Minrad, Inc., Bethlehem, PA, USA). Log-phase MDA-MB-231 cells were mixed with 270 μg/mL of control or tamoxifen rat mammary ECM at a concentration of 1.0 × 10⁵ cells/μL matrix. Of the cell/matrix mix (2 × 10⁶ total cells), 20 μL were back-loaded into a 3/10 cc insulin syringe with a 29 gauge 1/2 inch needle (BD Biosciences, San Jose, CA, USA) and injected into the fat pad of mammary gland number 4 37. For fibronectin experiments, MDA-MB-231 cells were mixed with incomplete α-MEM media with or without 20 μg/ml fibronectin (Collaborative Biochemical, Bedford, MA, USA) just before fat pad injection, with 10 mice per group. Tumour growth was measured using calipers twice weekly. Animals were sacrificed at six weeks post tumour cell injection. Tumours were excised, weighed and final tumour volume calculated (4/3πR²h). Statistical analyses were performed using the Kruskal-Wallis test and the Wilcoxon nonparametric analysis.

Tamoxifen is known to be an oestrogen antagonist in the mammary gland and acts to block proliferation of ER-positive epithelial cells. As expected, in comparison to control rats, the mammary epithelium in the tamoxifen-treated rats had a significant reduction in alveolar development (Figure 1a, upper panels). Morphometric analysis demonstrated that only 25% of vehicle-treated rats had mammary glands lacking well-developed alveoli, whereas 75% of mammary glands from tamoxifen-treated rats had mammary glands lacking alveoli (Figure 1a, graph). Consistent with the loss of mammary alveoli in the tamoxifen-treated rats, mammary epithelial cell proliferation, as measured by BrdU incorporation, was significantly reduced (Figure 1b, upper panels; BrdU quantification is shown in Figure 1b, lower panel). In the uterus, tamoxifen is an oestrogen agonist and one of tamoxifen's known side effects in women is an increase in uterine tumours. As previously reported by others, we found that tamoxifen-treated rats did not cycle through oestrous and the uterine and cervical tissues from these rats displayed proliferative phenotypes (Figure 1c), indicating that, as in women, tamoxifen acts as an oestrogen agonist in the rat uterus 38.

To begin to address the question of whether tamoxifen alters mammary stroma, primary fibroblasts were isolated from mammary glands of control and tamoxifen-treated rats. Fibroblast cells were characterised as positive by immunohistochemistry staining for vimentin and smooth muscle actin and by lack of pan-cadherin staining (data not shown). Fibroblast motility was assessed as a functional marker of cell activity. Although inter-animal variation in fibroblast motility was high, overall mammary fibroblasts isolated from the tamoxifen-treated animals displayed a significant three-fold decrease in transwell filter motility compared with control mammary fibroblasts (Figure 2a).

Fibroblasts are the major producers of mammary interlobular and periductal ECM and the motility data suggested that ECM production might be similarly altered by tamoxifen treatment. To address this question, the ECM secreted by primary fibroblasts in culture and incorporated into a 3D insoluble matrix was harvested and run on a 1D-gel. A prominent band of about 250 kD was decreased in the tamoxifen group (Figure 2b, arrow). This band was excised and found to contain fibronectin by mass spectrometry. These data show that in vitro, mammary fibroblasts isolated from tamoxifen-treated rats are less motile and incorporate less fibronectin into the insoluble matrix than fibroblasts isolated from vehicle control rats, data consistent with tamoxifen treatment causing fibroblast quiescence.

Tissue macrophages are a stromal cell type whose presence correlates directly with breast cancer progression in women and in rodent models 39. To determine whether tamoxifen treatment influenced the macrophage content of the mammary gland, immunohistochemistry analysis of the macrophage lysosomal marker CD68 was performed (Figure 2c, upper panels). Quantification of CD68 staining demonstrated that the CD68 signal was significantly decreased in mammary glands of tamoxifen-treated rats (Figure 2c, graph). Given that ECM proteins can function as attractants for macrophages, we next determined whether isolated ECM preparations from control and tamoxifen-treated rats differentially attract macrophages in a transwell filter assay. Using the isolated mammary matrices as chemoattractants, the motility of J774 cells, a mouse macrophage cell line, was suppressed by tamoxifen ECM in comparison with control ECM (Figure 3). These observations suggest that changes in mammary ECM composition with tamoxifen treatment could provide a plausible mechanism by which macrophage number is reduced in the mammary glands of tamoxifen-treated rats.

To identify prominent ECM proteins that compose the rat mammary gland stroma and which might be differentially expressed with tamoxifen treatment, ECM proteins were extracted from mammary glands of control and tamoxifen-treated rats and identified by tandem mass spectrometry. Prominent ECM proteins identified included fibronectin, collagen chains α-1(I), α-2(XIV) and α-2(I), laminin chains α4, β1, β2 and γ1, nidogen 1 and fibrillin 1. In addition, numerous matrix proteoglycans were identified in the mammary ECM, including decorin, perlecan, lumican, biglycan, mimecan and periostin (Table 1).

To determine whether the relative abundance of the identified mammary ECM proteins differed with tamoxifen treatment, several of the ECM proteins identified by proteomics were evaluated by Western blot. Mammary tissue from tamoxifen-treated rats had decreased levels of fibronectin (Figure 4a), data consistent with the in vitro fibroblast data described in Figure 2b. In addition, tamoxifen treatment resulted in a modest increase in the basement membrane protein laminin 1, and a decrease in total laminin 5 levels (Figure 4a). Western blot analysis revealed that tamoxifen treatment increased levels of the intra-lobular and inter-lobular ECM protein collagen 1 (Figure 4b, left panel). Tissue collagen deposition, as measured by Picro-sirius red staining, confirmed elevated levels of fibrillar collagen in the mammary stroma of tamoxifen-treated rats (Figure 4b, right panels and Table 2). Further, Western blot analyses demonstrated a generalised decrease in mammary ECM proteolysis with tamoxifen treatment, as evidenced by an increase in the ratio of high to low molecular weight species of fibronectin, laminin 1, laminin 5 and collagen 1 (Figures 4a, b).

Activity of MMP-2, an MMP found in high levels in virgin rat mammary glands and implicated in mammary ECM proteolysis, was evaluated in mammary tissue of control and tamoxifen-treated rats. Total MMP-2 and active MMP-2 were found to be decreased in mammary tissue of tamoxifen-treated rats (Figure 4c, left panel). Quantitation by densitometry showed about a 45% decrease in total MMP-2 (Figure 4c, upper right panel) and more than a 85% reduction in the cleaved active form of MMP-2 (Figure 4c, lower right panel). These data show that the decrease in ECM proteolysis in mammary glands of tamoxifen-treated rats correlated with concurrent decreases in MMP-2 levels and activity.

As confirmed by Western blot analyses, prominent differences between control and tamoxifen mammary ECM composition were the presence of high levels of intact fibrillar collagen in mammary tissue of tamoxifen-treated rats and the presence of partially proteolysed collagen and elevated levels of fibronectin in control mammary glands. The question of whether these ECM components influence macrophage motility was evaluated next. To reconstitute the high levels of collagen 1 observed in tamoxifen ECM, 10 μg/ml fibrillar collagen 1 was added to control ECM. The addition of fibrillar collagen was found to suppress J774 macrophage cell motility compared with control ECM alone (Figure 5a). Conversely, the addition of 10 μg/ml denatured collagen to tamoxifen mammary ECM was found to stimulate macrophage motility (Figure 5b). Given that fibronectin levels were higher in control ECM than tamoxifen ECM, the effect of 10 μg/ml fibronectin on macrophage motility was also investigated. We found that the addition of fibronectin to tamoxifen ECM had a modest but statistically significant stimulatory influence on J774 cell motility when compared with tamoxifen ECM alone (Figure 5c). Cumulatively, these data suggest that the high level of intact collagen and reduced level of fibronectin in the mammary tissue of tamoxifen-treated rats contributes to the reduced number of macrophages observed in mammary glands of tamoxifen-treated rats.

To address whether mammary ECM isolated from tamoxifen-treated rats inhibits metastatic-attributes of epithelial tumour cells in vitro, the effects of control and tamoxifen ECM on mammary tumour cell invasion, motility and haptotaxis were evaluated. In a transwell filter assay, the invasiveness of Ras transformed mammary epithelial MCF-12A cells was reduced two-fold on tamoxifen ECM compared with control ECM (Figure 6a). Differences in motility were not observed (data not shown). These functional analyses were extended to the highly metastatic human breast cancer MDA-MB-231 cells. MDA-MB-231 cells had reduced motility on the tamoxifen ECM (Figure 6b), but no change in invasion (data not shown). The haptotactic properties of the isolated mammary ECM were evaluated using mammary ECM as the chemoattractant. MDA-MB-231 cells displayed a three-fold reduction in haptotactic properties towards tamoxifen ECM (Figure 6c). Cumulatively, these data suggest a partial suppression of the metastatic phenotype by tamoxifen ECM.

To further characterise the effects of these ECM matrices on tumour cell behaviour, 3D cell culture assays were performed. When plated onto control ECM, MCF-12A-Ras cells displayed an aggressive phenotype characterised by disorganised structuring and invasive filapodia (Figure 7a, left panel). In contrast, when plated onto tamoxifen ECM, these same cells formed compacted smooth-surfaced multicellular clusters lacking invasive filapodia (Figure 7a, right panel). H&E stained 4 μm sections of 3D organoids confirmed the observation that tamoxifen ECM induced small spheroid shaped organoids (Figure 7b). Quantitative morphometric analyses of organoid size demonstrated that the average size of organoids plated onto tamoxifen ECM was 2.6-fold smaller than organoids on control ECM (Figure 7c). Even very aggressive MDA-MB-231 cells had a subtly altered 3D phenotype on the tamoxifen ECM. These cells were more clustered and cuboidal in appearance than when plated on the control ECM, indicating induction of an epithelial-like morphology (data not shown). The 3D culture data corroborate the transwell filter assay data and show that the tamoxifen ECM fails to support tumour cell motility and invasion.

To evaluate the contributions of individual ECM proteins to tumour organoid formation, reconstitution experiments were performed with fibronectin and collagen 1. The addition of high molecular weight fibronectin to tamoxifen matrix was found to consistently increase organoid size in 3D culture, but the increase in size was modest (Figure 7d). Further, an increase in local invasion was observed with fibronectin (data not shown). These data suggest that an increase in fibronectin alone does not wholly account for the aggressive phenotype displayed by MCF-12A-Ras cells plated on control ECM. Of interest, the addition of fibrillar collagen to control matrix disrupted 3D organoid formation, thus the contribution of collagen to the suppressive microenvironment could not be evaluated in this model (data not shown).

Based on the demonstration that tamoxifen ECM reduced the aggressive characteristics of two transformed breast cancer cell lines in multiple cell culture models, we investigated whether tamoxifen ECM could decrease tumour promotion in vivo. For this study, MDA-MB-231 cells were mixed with mammary ECM from control and tamoxifen-treated rats and injected into the inguinal fat pads of female athymic mice. By seven weeks after tumour cell injection, tumour incidence was the same between groups; however, mice injected with tumour cells plus tamoxifen ECM had tumours that were three times smaller than those injected with tumour cells mixed with control ECM (Figure 8a). To determine whether the elevated levels of fibronectin in control mammary ECM influence tumour size in this xenograft model, as suggested by the 3D organoid data, MDA-MB-231 cells were treated with 20 μg/ml fibronectin before mammary fat pad injection (Figure 8b). Tumour size was increased by fibronectin across all time points, however, because of a large variation in tumour size within a group, these data did not reach statistical significance.