Metastatic breast cancer cells overexpress and secrete miR-218 to regulate type I collagen deposition by osteoblasts

Human serum specimens were obtained from voluntarily consenting breast cancer patients between February 2006 and December 2011 at the City of Hope National Medical Center (Duarte, CA, USA) under institutional review board-approved protocols. All 47 patients involved in this study had stage IV disease with or without bone metastases at the time metastatic disease was diagnosed. Among them, 33 patients had bone metastases, in most cases with concurrent metastases to other organs, whereas the other 14 patients had distant metastases to other organs without the involvement of bone. The two groups exhibited balanced age, tumor subtype, and sample collection time. Serum specimens examined in this study were collected at the time metastasis was initially diagnosed or the earliest draw available. Clinical characteristics are summarized in Additional file 1: Table S1. Trizol LS reagent (Thermo Fisher Scientific; Waltham, MA, USA) was used to extract total RNA from ~ 0.5 ml of serum; RNA pellet was dissolved in 10 μl of RNase-free water and subjected to Solexa sequencing and RT-qPCR as previously described [26].

Human breast cancer cell line MDA-MB-231, MCF-7, human non-cancerous mammary epithelial cell line MCF10A, and mouse preosteoblast cell line MC3T3-E1 were obtained from American Type Culture Collection (Manassas, VA, USA) and cultured in the recommended media. The bone-tropic subline of MDA-MB-231 was a kind gift from Dr. T. Yoneda. All cells used herein were tested to be free of mycoplasma contamination and authenticated by using the short tandem repeat profiling method. MDA-MB-231 cells were stably transduced with control and miR-218 lentiviral constructs purchased from GeneCopoeia (Rockville, MD, USA). miRIDIAN miR-218 mimic and the corresponding negative control were purchased from GE Dharmacon (Lafayette, CO, USA). LNA oligonucleotides against miR-218 and control were purchased from Exiqon (Woburn, MA, USA). Cell transfection, reporter assays, production of viruses, infection and selection of transduced cells, as well as flow cytometry for cell characterization, were carried out as previously described [7‐9, 27]. Osteoblast differentiation was induced by 50 μg/ml L-ascorbic acid, 10 mM β-glycerolphosphate, and 0.1 μM dexamethasone for 16–21 days with medium change every 3–4 days. Conditioned medium was collected by passing through 0.45 μm filters. Conditioned medium concentration was performed using Vivaspin 20 columns from Sartorius (Bohemia, NY, USA).

One million bone marrow cells isolated from C57BL/6 mice were bulk cultured in a six-well plate and 40 ng/ml M-CSF was added to induce adherence for 3 days. Osteoclast differentiation was induced by adding 40 ng/ml M-CSF and 100 ng/ml RANKL (R&D Systems; Minneapolis, MN, USA) and culturing the cells for 3–7 days. Tartrate-resistant acid phosphatase (TRAP) staining was performed using an Acid Phosphatase, Leukocyte (TRAP) Kit (Sigma-Aldrich; St. Louis, MO, USA).

PCR primers 5′- GATCAACTCGAGGTACACGGTGGGCTGAGTA and 5′- GATCAAGCGGCCGCCCGTGGCACTCAATCTTTTA were used to clone the wild-type 3’ untranslated region (UTR) of human INHBB. The PCR-amplified fragments were digested with XhoI and NotI and then inserted into the same sites of psiCHECK-2 reporter vector (Promega; Madison, WI, USA) downstream of the Renilla luciferase gene. PCR primers 5′- AATTGCGCCTTCCGAGCACACATAACTCAGATAAGACAGAGACGCAGAGA and 5′- CTCTCTCTCTCTGCGTCTCTGTCTTATCTGAGTTATGTGTGCTCGGAAGG (mutated nucleotides underlined) were used to clone the miR-218-site-mutant of INHBB 3’UTR. Similarly, human Yin and Yang 1 (YY1) 3’UTR was cloned into psiCheck-2 vector using primers 5’-GATCAACTCGAGTTCTCGACCACGGGAAGCA, 5’-GATCAAGCGGCCGCTGAAATTAAGCTACTGGCACTCAA, and mutagenesis primers 5’-AAGAATATGGCAGAACAAGATCTGTCTCAGATGTCTTATTTTCTTTTGTT and 5’-TCTGGACAACAAAAGAAAATAAGACATCTGAGACAGATCTTGTTCTGCCA. The YY1 overexpression plasmid was constructed by cloning the full-length YY1 cDNA amplified by primers 5’-GATCAGAATTCATGGCCTCGGGCGACA and 5′- AATAGGATCCTCACTGGTTGTTTTTGGCCT from MDA-231 cells, and inserting the cDNA into the EcoRI/BamHI sites of pSG5 vector. All constructs were verified by sequencing.

EVs secreted by MDA-MB-231 and derived cell lines were prepared as previously reported [7‐9]. Conditioned medium was first collected after incubating cells in growth medium containing 10% EV-depleted FBS (prepared by overnight ultracentrifugation of medium-diluted FBS at 100,000 × g at 4 °C) for 48 h, and pre-cleared by centrifugation at 500 × g for 15 min and then at 10,000 × g for 20 min. EVs were isolated by ultracentrifugation at 110,000 × g for 70 min, and washed in PBS using the same ultracentrifugation conditions. When indicated, DiI (1,1′-Dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate; Sigma-Aldrich) was added into the PBS at 1 μM and incubated for 20 min before the washing spin, followed by an additional wash to remove the excess dye. The pelleted EVs were resuspended in ~ 100 μl of PBS for cell treatment. For cell treatment, 2 μg of EVs (equivalent to those collected from ~ 5 × 10⁶ producer cells) based on protein measurement using Pierce™ BCA protein assay kit (Thermo Fisher Scientific) were added to 2 × 10⁵ recipient cells. For EV characterization, EVs were subjected to nanoparticle tracking analysis using a NanoSight NS300 (Malvern; Westborough, MA,USA), or further fractionated by gradient separation following a modified protocol [28]. EVs isolated by ultracentrifugation were loaded onto a 5-step OptiPrep (Sigma-Aldrich) gradient consisted of 40, 30, 20, 10, and 5% iodixanol in 20 mM Hepes (pH 7.2), 150 mM NaCl, 1 mM Na₃VO₄, and 50 mM NaF. After centrifugation in a SW 40 Ti rotor (Beckman Coulter; Indianapolis, IN, USA) at 110,000 × g at 4 °C for 16 h, 12 1-mL fractions were collected and washed in PBS by another spin at 110,000 × g for 70 min before Western analysis and RNA extraction for RT-qPCR.

These procedures were performed as described previously [7‐9]. Primers used in RT-qPCR are indicated in Additional file 2: Table S2. The miR-218, miR-140-3p (as internal control for miR-218 in EVs and sera), and U6 primers (as internal control for intracellular miR-218) were purchased from Qiagen (Valencia, CA, USA). An annealing temperature of 57.5 °C was used for all primers.

Protein extracts were separated by SDS-PAGE. Protein detection was performed using the following antibodies: Collagen alpha-1(I) chain carboxy-telopeptide (LF68) (Kerafast; Boston, MA, USA, ENH018), Inhibin βA (Novus, NBP1–30928), Inhibin βB (Thermo Fisher Scientific, PA5–28814), Inhibin α (Santa Cruz Biotechnology; Dallas, TX, USA, SC-22048), YY1 (Abcam; Cambridge, MA, USA, ab109228), Phospho-Smad2 (Ser465/467) (Cell Signaling Technology; Danvers, MA, USA, 3108S), SMAD2 (Cell Signaling Technology, 3122S), Phospho-Smad3 (Ser423/425) (Cell Signaling Technology, 9520S), SMAD3 (Cell Signaling Technology, 9523S), tissue inhibitor of metalloproteinases 3 (TIMP3) (Abcam, ab155749), β-actin (Sigma-Aldrich, A1978), as well as anti-rabbit, anti-mouse, and anti-goat HRP-conjugated secondary antibodies (Santa Cruz Biotechnology).

Bone formation marker P1NP was detected by a rat/mouse P1NP enzyme immunoassay kit (Immunodiagnostic Systems; Boldons, UK). Bone resorption marker C-terminal telopeptide (CTX-1) was detected by a RatLaps EIA kit (Immunodiagnostic Systems).

All animal experiments were approved by the institutional animal care and use committee at the Beckman Research Institute of the City of Hope. Female NOD/SCID/IL2Rγ-null (NSG) mice of 6-month-old were used in this study. EVs from ~ 10⁷ cancer cells were used to treat mice through tail vein injection with 27G needles twice a week for 4 weeks. Serum was collected 1 week after the last EV injection via retro-orbital bleeding.

All quantitative data are presented as mean ± standard deviation (s.d.) unless stated otherwise. Two-sample two-tailed Student t tests were used for comparison of means of quantitative data between two groups. For multiple independent groups, one-way ANOVA with post hoc Tukey tests were used. Values of P < 0.05 were considered significant. Sample size was generally chosen based on preliminary data indicating the variance within each group and the differences between groups. All samples/animals that have received the proper procedures with confidence were included for the analyses. For experiments in which no quantification is shown, images representative of at least three independent experiments are shown.

To search for miRNAs that might be functionally relevant to breast cancer bone metastasis, we obtained sera from 47 stage IV breast cancer patients with (n = 33) or without (n = 14) bone metastases and performed small RNA sequencing. In addition, to identify miRNAs characteristically secreted by bone-metastasizing breast cancer cells, we profiled the miRNAs in the EVs secreted by the metastatic breast cancer cell line MDA-MB-231 (MDA-231) as well as its bone-seeking variant designated as MDA-231-bone [29]. hsa-miR-218-5p (miR-218) was significantly higher in the sera from patients with bone metastases compared to those without (Additional file 3: Table S3), and was > 5-fold higher in the EVs from MDA-231-bone cells compared to those from parental MDA-231 (Additional file 4: Table S4). These results were confirmed by qRT-PCR using a selected internal reference miR-140-3p, which was consistent among all serum samples tested and between EVs from the two cell lines based on the smRNA-seq data (Fig. 1a and b). In contrast, levels of circulating miR-218 were not significantly different when the same cohort of patients was stratified by the presence or absence of brain metastases (Fig. 1a, right panel). In addition, miR-218 was also expressed at a higher intracellular level in the bone-tropic MDA-231 cells compared to parental MDA-231 and the non-cancerous mammary epithelial cells MCF10A (Fig. 1b). Thus, we focused on miR-218 in the subsequent studies for its potential role in breast cancer bone metastasis.

To determine the specific effects of miR-218 we established a stable cell line of MDA-231 that overexpresses miR-218 (MDA-231-miR-218) or control vector (MDA-231-miR-ctrl) (Fig. 2a). Compared to the control cells, the miR-218-overexpressing cells also secreted a higher level of miR-218 into the EVs. Upon density gradient fractionation of the EVs, miR-218 was found to be enriched in the fractions containing exosomes and with a density between ~ 1.10 and ~ 1.145 g/mL (Additional file 5: Figure S1). When fluorescently labeled EVs from MDA-231-miR-ctrl, MDA-231-miR-218, and MDA-231-bone were added to preosteoblast cells MC3T3, a high EV uptake efficiency was observed with all types of EVs (Fig. 2b). We also detected transfer of miR-218 into recipient MC3T3 cells upon EV treatment in the presence of an RNA polymerase II inhibitor DRB to block endogenous miR-218 transcription (Fig. 2c). COL1A1 has been reported as a direct target of miR-218 in gastric cancer cells and human stellate cells [30, 31]. We found that Col1a1, but not collagen type I alpha 2 chain (Col1a2), was significantly downregulated by EVs from the two high-miR-218-secreting cell lines at the mRNA level (Fig. 2d), and that this effect on type I collagen expression was more dramatic at the protein level (Fig. 2e). Consistent with the lower levels of all forms of type I collagen, the bone formation marker P1NP was also decreased in the conditioned medium (CM) from differentiated MC3T3 treated with high-miR-218 EVs (Fig. 2f). To test the effect of secreted miR-218 in vivo, we injected EVs from MDA-231-miR-ctrl or MDA-231-miR-218 into NSG mice through the tail vein twice a week for 4 weeks. The bone formation marker P1NP was significantly decreased in the serum from mice treated with MDA-231-miR-218 EVs, whereas the bone resorption marker CTX-1 was not affected (Fig. 2g and h). Taken together, cancer-secreted miR-218 encapsulated in the EVs downregulated type I collagen expression and deposition by osteoblasts. In comparison, breast cancer-secreted EVs were found incapable of regulating osteoclast differentiation (Additional file 6: Figure S2).

To identify other genes regulated by miR-218, we performed RNA-seq to compare the gene expression profile in MDA-231-miR-ctrl and MDA-231-miR-218 cells. Among the genes showing significantly different expression between the two cell lines, we found that inhibin beta A subunit (INHBA) was upregulated and inhibin beta B subunit (INHBB) downregulated by miR-218 overexpression (Additional file 7: Table S5). This was also observed at both mRNA and protein levels in parental MDA-231 cells upon transient transfection of a miR-218 mimic (Fig. 3a and b). Similar results were observed in another breast cancer cell line MCF-7, where miR-218 also upregulated INHBA but downregulated INHBB expression (Additional file 8: Figure S3a). In contrast, transfection of an antagomiR against miR-218 into MDA-231 bone cells led to a lower mRNA level of INHBA and a higher level of INHBB (Fig. 3c). Since INHBB is predicted to be a direct target of miR-218 by TargetScan (Fig. 3d), we performed dual luciferase reporter assay using psiCheck2 vector to confirm that the 3’UTR of INHBB containing the wild-type, but not mutated, miR-218 binding site responded to miR-218 (Fig. 3e). In search of the mechanism through which INHBA was upregulated by miR-218, we focused on YY1, a transcriptional repressor that is also a putative miR-218 target (Fig. 3d) and has been associated with Inhba expression in ovaries [32]. YY1 mRNA was downregulated by miR-218 in MDA-231 cells (Fig. 3f). Ectopic expression of YY1 abolished the effect of miR-218 on inducing inhibin βA (Fig. 3g). Furthermore, dual luciferase assay revealed that YY1 was directly targeted by miR-218 through the predicted binding site in the 3’UTR (Fig. 3h). Thus, miR-218 regulates the expression of two TGFβ superfamily cytokines in breast cancer cells by directly targeting YY1 and INHBB.

We continued to test how miR-218-mediated differential expression of inhibin β subunits in breast cancer cells potentially influences the surrounding cancer cells and preosteoblasts to respectively establish autocrine and paracrine signaling. We treated serum-starved MDA-231 or MC3T3 with EV-depleted CM collected from miR-218-transfected MDA-231 cells or anti-miR-218-transfected MDA-231-bone cells. To our surprise, CM from miR-218-overexpressing breast cancer cells induced SMAD2/3 phosphorylation in MDA-231 cells (Fig. 4a), whereas the same CM treatment to MC3T3 cells suppressed the level of phospho-SMAD2/3 (Fig. 4b). Similarly, CM from miR-218-overexpressing MCF-7 cells also induced SMAD2 phosphorylation in untransfected MCF-7 cells (Additional file 8: Figure S3b). Conversely, anti-miR inhibition of miR-218 in CM-producing MDA-231-bone cells led to reduction of SMAD2/3 phosphorylation in cancer cells (Fig. 4a) and de-repression of SMAD2/3 signaling in preosteoblasts (Fig. 4b) at the 2-h time point.

To explain the differential response of SMAD signaling in cancer or bone stromal cells to CM from breast cancer cells with varying levels of miR-218, we concentrated the CM from MDA-231 cells and found that the inhibin βA subunit was secreted at a higher level by miR-218-transfected MDA-231 cells compared to control group while inhibin βB was undetectable in the concentrated CM (Fig. 4c). Moreover, we also detected increased secretion of activin A in the CM from MDA-231 cells transfected with miR-218 (Fig. 4d). Interestingly, concentrated CM from MC3T3 preosteoblast cells contained inhibin α subunits (Fig. 4e), which was not expressed by MDA-231 or MCF-7 breast cancer cells (Additional file 7: Table S5, Fig. 4e, Additional file 8: Figure S3c). We thereby hypothesized that the presence of inhibin α in the environment led to a functional switch of cancer-secreted inhibin βA from activin-mediated SMAD2/3 activation to inhibin-mediated suppression. Indeed, addition of recombinant inhibin α into the CM from miR-218-overexpressing cells reversed the activation of SMAD2/3 phosphorylation in recipient cancer cells (Fig. 4f), whereas adding neutralizing antibody against inhibin α into the CM to treat MC3T3 cells led to de-repression of SMAD2/3 signaling (Fig. 4g). Therefore, secreted inhibin βA, which is upregulated by miR-218 in breast cancer cells, autocrinally activates SMAD signaling in MDA-231 cancer cells but paracrinally represses this pathway in bone stromal cells, due to the different expression status of inhibin α in the environment.

We next examined the downstream effect of cancer-secreted inhibin βA on preosteoblasts. Tissue inhibitor of metalloproteinases 3 (TIMP3) has been shown to inhibit the activity of ADAMTS2, which is the processing enzyme of type I procollagen during osteoblast differentiation [33]. We found that Timp3 was upregulated in differentiating MC3T3 cells following the treatment with EV-depleted CM from miR-218-overexpressing MDA-231 cells, and was downregulated when miR-218 was inhibited in CM-producing cells (Fig. 5a and b). As a result, the processing of type I procollagen was suppressed during osteoblast differentiation when miR-218 level was high in breast cancer cells, and vice versa (Fig. 5a). To confirm that this effect was caused by higher level of inhibin βA in the CM from miR-218-overexpressing MDA-231 cells, which contributed to the upregulation of Timp3 expression, we added neutralizing antibody against inhibin βA into the CM and detected restoration of Timp3 level in differentiating MC3T3 (Fig. 5c and d). Taken together, we show that paracrinal inhibin βA secretion by breast cancer cells, which is regulated by miR-218, alters the processing of procollagen during osteoblast differentiation through regulating Timp3 expression.