Introduction

The triple-negative sub-type of breast cancer is a highly aggressive cancer for which treatment options are limited [1]. Expression of the RUNX transcription factors in patients with the triple-negative sub-type of breast cancer correlates with a poor prognosis [2, 3]. Emerging evidence suggests that the role of RUNX proteins in breast cancer is dependent on the specific RUNX factor involved and the sub-type of breast cancer cell [4]. In order to consider the RUNX factors as viable targets in breast cancer therapies it is therefore critically important to determine the role of the different factors in different sub-types of breast cancer. Since CBFβ facilitates the function of all three RUNX transcription factors, establishing its role in determining the phenotype of breast cancer cells is essential [5, 6].

Mutations in CBFβ are amongst the most frequently reported for breast cancer tumours, suggesting a tumour suppressor role for CBFβ in ER+ breast cancer [7, 8]. In contrast, we and others have previously shown that expression of RUNX2 and CBFβ contribute to the metastatic phenotype of triple-negative breast cancer cells [9,10,11,12]. In this context it is therefore the maintained expression of RUNX factor activity that promotes their metastatic phenotype.

Epithelial to mesenchymal transition (EMT) contributes to the progression of metastatic cancer as it enables cancer cells to become migratory and invasive [5d). We found that 7 out of 11 animals injected with the control cells developed tumours in the hind limbs, with numerous cancer-induced bone lesions detected, compared with three out of seven animals injected with CBFβ-CRISPR (Fig. 5d). In addition, the animals receiving the control cells had higher numbers of skeletal lesions (average = 8.1 lesions/mouse) compared with those receiving CBFβ-CRISPR cells (average = 1.7 lesions/mouse), (Fig. 5d). These data demonstrate that loss of CBFβ reduces the ability of MDA-MB-231 cells to metastasise to bone.

Fig. 5: CBFβ contributes to the development of bone metastasis.
figure 5

a MDA-MB-231 cells were transplanted into the inguinal mammary fat pad of CD1-Nude females. Data shown at 4 weeks post-transplantation. Data is presented as mean ± SDM (shNS; n = 5; shCBFβ; n = 7). The difference between the two groups is significant (p < 0.05) as determined using an unpaired Student’s t-test. b Growth curves showing knockdown of CBFβ reduces growth rate in 3D culture. MDA-MB-231, MDA-shNS, or MDA-shCBFβ cells were grown on 3D culture and cells were counted every 2 days. c Knockdown of CBFβ inhibits invasion in 3D co-cultures with osteoblasts. 3D cultures of MC-3T3 osteoblasts were grown for 2 months prior to addition of MDA-MB-231 or MDA-shCBFβ cells. Confocal images were taken after 8 days of co-culture. Cells were stained with phalloidin. MDA cells were identified by GFP fluorescence. d CBFβ silencing reduces tumour growth in bone in vivo. MDA-MB-231 control or CBFβ−/− cells were injected i.c. into 6-week old BALB/c nude mice and tumour growth in the hind limbs analysed 26 days later. 3D reconstruction of tumour-bearing tibia showing the presence of osteolytic lesions. The histogram shows the average number of bone tumours per mouse for each cell line as indicated.

Full size image

Discussion

In this study, we have shown that the RUNX co-regulator CBFβ is essential to drive MDA-MB-231 breast cancer cells through EMT. Maintenance of the mesenchymal phenotype is, at least in part, achieved by regulating the expression of the EMT transcription factor Slug. We also demonstrated that the MET induced by loss of CBFβ is completely reversible by re-expression of CBFβ. These findings are important since they demonstrate that, in principle, regulation of RUNX/CBFβ activity can determine the extent to which triple-negative breast cancer cells differentiate along the epithelial-mesenchymal continuum. In the context of metastasis in vivo this raises the possibility that interactions between cancer cells and the microenvironment influence the activity of RUNX/CBFβ, thereby shifting their phenotype toward the epithelial state and enabling the cells to colonise the new niche. Indeed, the loss of CBFβ significantly reduced the capacity of metastatic cancer cells to invade osteoblast cultures in vitro and to form osteolytic lesions in vivo.

Previous work has shown that about two thirds of the RUNX1 transcriptome is shared with the RUNX2 transcriptome in MCF7 cells [19]. Our finding that CBFβ, RUNX1 and RUNX2 are all necessary for SNAI2 expression suggests that all three factors combine to ensure a sufficient level of Slug is available to maintain the mesenchymal phenotype. This suggests that none of these factors are redundant in this context. This may reflect the need for a threshold level of RUNX factors to be expressed but does not discount the possibility that RUNX1 and RUNX2 also regulate specific subsets of genes. SNAI2 is a well-established EMT transcription factor that appears to be a key target for RUNX transcription factor complexes and is perhaps one of several genes that contribute to the metastatic nature of breast cancer cells [16, 18].

In contrast to these findings, we did not observe significant changes in classical EMT markers when either CBFβ or RUNX1 was depleted in MDA-MB-468 cells. The emerging picture is that RUNX complexes have pleiotropic effects dependent on the cellular context and it is of note that whilst MDA-MB-231 and MDA-MB-468 cells are both “triple-negative” they exhibit different characteristics and gene expression profiles, MDA-MB-231 being of the claudin low sub-type and MDA-MB-468 are basal [32]. Moreover, mutations in RUNX1 and CBFβ are associated with the ER+ sub-type of breast cancer and is therefore predicted to have a tumour suppressor role in this context [7, 8, 33]. Indeed, previous studies have shown that RUNX1 suppresses development of ER+ luminal breast cancer but it is not known how CBFβ contributes in this context [34]. Our finding that depletion of RUNX1 in MDA-MB-231 cells induces MET is in agreement with a previous study in which the expression of miR-378 in MDA-MB-231 cells, which inhibits RUNX1 expression, also resulted in suppression of migration and invasion [35]. However, this is in contrast to the role of RUNX1 in MCF10A and MCF7 cells where RUNX1 expression is required to maintain an epithelial-like phenotype [20, 36]. Taken together these findings suggest that RUNX transcription factors contribute to the epithelial-mesenchymal continuum is a cell-context dependent manner.

Finally, small molecule inhibitors that inhibit the interaction between CBFβ and RUNX have been shown to inhibit colony formation in a basal-like breast cancer cell line [37]. Our findings that CBFβ is essential to maintain the mesenchymal phenotype, and that it contributes to the formation of bone metastases, suggests that in principle inhibiting this complex might maintain metastatic colonies in a less aggressive epithelial state by driving MET. Thus, targeting the RUNX/CBFβ complex in this way might be a viable option to treat a sub-group of triple-negative breast cancer patients.

Methods

Cell lines

Parental MDA-MB-231 expressing GFP were a kind gift from D. Welch, University of Alabama. MDA-MB-231-shCBFβ/RUNX1/RUNX2 were produced using Sure Silencing shRNA plasmids (SABiosciences) as previously described [16]. Lines were authenticated by multiplex-PCR assay using the AmpF/STR system (Applied Biosystems) and confirmed as mycoplasma free. Monolayers were grown in complete medium (DMEM/10% FCS/2 mmol/L L-glutamine/PenStrep 0.4 μg/mL puromycin, 50 μg/mL geneticin, 500 μg/mL hygromycin as required) and maintained in a humidified incubator at 37 °C at an atmospheric pressure of 5% (v/v) CO2/air.

Western blotting

Protein was separated on an SDS–PAGE and transferred to Hybond-C Extra nitrocellulose membrane. Primary antibodies included: β-Tubulin (Abcam, ab6046), Lamin-B1 (Abcam, ab16048), CBFβ (Abcam, ab33516), RUNX1 (Abcam, ab23980), RUNX2 (MBL, D130-3), Snai2 (Cell Signalling, C19G7), FLAG (Sigma, F1804).

For all experiments three biological replicates were performed and densitometry was conducted to calculate average changes using ImageJ software, which is freely available at http://rsb.info.nih.gov/ij/.

Cell scratch assay

Confluent monolayers were scratched on day 0 and medium was changed to serum free. Cells were grown in an AS MDW live cell imaging microscope system at 37 °C 5% CO2 for 48 h. Images were taken every 20 min and 40 views were taken in each well. For all experiments three technical and three biological replicates were performed. Image data analysis was performed using Cell Profile software.

Overlay three-dimensional culture of breast cells

Matrigel (Corning, 354230) was thawed on ice overnight at 4 °C and then spread evenly onto dishes (MatTek P35G-1.0-14-C) or into 24-well plates (Greiner, Bio-one 662892). Cells were resuspended in 3D assay medium (2% Matrigel, 95% DMED, 2%FBS, 1% Pen/Strep, 1% nonessential amino acid, 1% L-glutamine) and plated on to solidified Matrigel. Cells were grown in 5% CO2 humidified incubator at 37 °C. Assay medium was changed every 3 or 4 days. Cells were fixed at day 14. For all experiments three technical and three biological replicates were performed.

Microscopy

2D culture

Images were collected on a Zeiss Axioimager.D2 upright microscope using a 10× objective and captured using a Coolsnap HQ2 camera (Photometrics) through Micromanager software v1.4.23. Specific band pass filter sets for DAPI, FITC and Cy5 were used to prevent bleed through from one channel to the next.

3D culture

Images were collected on a Leica TCS SP5 AOBS inverted and upright confocal microscopes. Images were collected using PMT detectors with the following detection mirror settings; [FITC 494–530 nm; Texas red 602-665 nm; Cy5 640-690 nm] using the [488 nm (20%), 594 nm (100%) and 633 nm (100%)] laser lines, respectively. When it was not possible to eliminate cross-talk between channels, the images were collected sequentially. When acquiring 3D optical stacks the confocal software was used to determine the optimal number of Z sections. Only the maximum intensity projections of these 3D stacks are shown in the results. Images were then processed and analysed using Fiji ImageJ (http://imagej.net/Fiji/Downloads) [14], which is freely available online.

For all experiments three technical and three biological replicates were performed.

Immunofluorescence

For cell grown on coverslips, fixing and permeabilisation was performed in 4% paraformaldehyde (Sigma) and 0.1% Triton-100 (Sigma) before blocking in 1% Bovine Serum Albumin. The cells were then incubated with the primary antibodies overnight at 4 °C at a dilution of 10 µg/µl). pERM (Cell Signalling Technology; Antibody #3141), anti-integrin avβ6 (Abcam, ab97588). Alexafluor secondary antibodies (Invitrogen) were used at a 1/200 dilution. The coverslips were mounted on glass slides using mounting medium with DAPI (Invitrogen P36965). For cells grown in Matrigel, following fixing and permeabilisation as detailed above non-specific staining was blocked using cells were blocked using IF Buffer (7.7 mM NaN3, 0.1% bovine serum albumin, 0.2% Triton X-100, 0.05% Tween-20 in PBS) + 10% goat serum. Cells were stained using Phalloidin for F-actin (Sigma, P1951) and mounted using DAPI (Invitrogen).

For all experiments three technical and three biological replicates were performed.

Mammosphere culture

Mammosphere culture was carried out as previously described [38]. Spheres >50 µm were counted on day 5. For all experiments three technical and three biological replicates were performed.

Quantitative reverse transcription PCR

RNA was extracted using the Qiagen RNAeasy kit according to manufacturer’s instructions and quantified on the Nanodrop spectrophotometer (Thermo). Real time one step qRT-PCR was carried out using the QuantiTect SYBR® Green RT-PCR Kit (Qiagen) according to manufacturer’s instructions before analysis on the 7900 PCR machine (Applied Biosystems). A table of the primers used can be found in Supplementary Table 1. For all experiments three biological replicates were performed.

Inducible Cell line production

Mouse CBFβ-FLAG and ER was ligated into pcDNA3.1/Hygro(-) vector producing pcDNA3.1/Hygro(-)-CBFβ-FLAG-ER. Stable lines were made using this vector and cells were transfected using Lipofectamine according to manufacturer’s instructions (Fig. S1).

Nuclear/Cytoplasmic separation

Cells were resuspended in 400 μl of ice cold Buffer A (10 mM HEPES pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM DTT, 0.5 mM PMSF) with the addition of complete mini-EDTA-free protease inhibitor cocktail (Roche) and incubated at 4 °C for 15 min. Cells were lysed by addition of 10% NP-40 (Sigma) before centrifuging at 4 °C and removal of the cytoplasmic extracts in the supernatant. The pellet was then resuspended in ice cold Buffer B (20 mM HEPES pH 7.9, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT and 1 mM PMSF) containing protease inhibitors and vortexed vigorously for 45 min at 4 °C. Nuclear proteins were collected from supernatant following centrifugation at 4 °C. For all experiments three biological replicates were performed.

Invasion assay

Matrigel Matrix (Corning, 354230) was diluted to final concentration of 300 μg/mL in cold coating buffer (0.01 M Tris (pH8.0), 0.7% NaCl) before being added to invasion chambers (Corning Cat, 353097) and left to set overnight at 37 °C. 2 × 104 cells were added to each chamber in serum free medium and 0.75 mL complete medium was added to the wells. Cells were allowed to invade 24 h in cell culture incubator. Invading cells were fixed and permeabilised with 4% PFA (Electron Microscopy Sciences, 15713-S) and 0.1% Triton (Sigma). Non-Invading cells were removed using a cotton swab. Cells were stained with Crystal violet solution. For all experiments three technical and three biological replicates were performed.

Chromatin immunoprecipitation (ChIP)

ChIP was performed as previously described [39]. ChIP-PCR was performed using Quantitect SYBR green (Qiagen). The primers used can be found in Supplementary Table 1. For all experiments three biological replicates were performed.

CRIPSR-Cas9 mediated gene deletion

CBFβ and RUNX1 gene knockout was performed using a double nickase CRISPR-Cas9 strategy as described previously [31]. Guide-RNA sequences were designed using E-CRISP to minimise off target effects [24]. Cells were Fluorescence-activated cell sorted (FACS) for GFP-Cas9 expression 48 h after transfection and grown up from single colonies prior to genomic DNA PCR and western blot screening.

Mammary fat pad xenografts

3 × 106 MDA-MB-231 cells were transplanted in matrigel into the inguinal mammary fat pad of 12 week old CD1-Nude females (Charles River, UK). Mice were randomised to receive shNS or shCBFβ-KO cells to give groups of comparable weight/age. The same investigator (SMM) transplanted all cells into the recipients.

Animals were excluded if they failed to grow a tumour to clinical endpoint, and/or exhibited unrelated general ill health within the duration of the experiment.

Caliper measurements were carried out throughout by technical staff blinded to the expected outcome of the experiment to assess tumour volume which was calculated using the formula ½(length × width2).

This experiment was carried out in dedicated animal facilities under project licence 60/4181 with adherence to the Animal (Scientific Procedures) Act, the European Directive 2010 and local ethical approval (University of Glasgow). No randomisation was required.

Bone tumour growth studies

Tumour growth studies used 6–8 week old female BALB/c nude between 13 and 18.4 g (Charles River, Kent, UK). Experiments were carried out in accordance with local guidelines and with Home Office approval under project licence 70/8799, University of Sheffield, UK. 12 mice per group were injected with 1 × 105 MDA-MB-231 control (2014-8-044) or CBFβ-CRISPR knockout cells (2015-6-010 CRISPR) via the left cardiac ventricle to generate tumours in bone [30]. Mice were randomised to receive control or CBFβ-KO cells to give groups of comparable weight/age. Mice were removed early from the study if they showed luciferase signal in the chest only (indicating a missed injection) or if the mice developed hind limb paralysis within the first 48 h. These parameters were pre-defined before the experiment commenced.

Animals were culled 26 days following tumour cell injection and hind limbs collected for analyses of tumour growth and associated bone lesions in tibiae and femurs.

Analysis of bone lesions

Hind limbs were fixed in 4%PFA and scanned by μCT prior to decalcification in 1%PFA/0.5% EDTA and processing for histological sectioning. μCT analysis was carried out using a Skyscan 1272 × -ray-computed μCT scanner (Skyscan, Aartselar, Belgium) equipped with an x-ray tube (voltage, 50 kV; current, 200uA) and a 0.5-mm aluminium filter. Pixel size was set to 5.99 μm and scanning initiated from the top of the proximal tibia or distal femur. Lytic, tumour-induced bone lesions were counted manually for each bone and performed by a technician being unaware of anticipated outcome of the experiment.

Statistical analysis

Data is represented as mean +/− SD, n = 3 unless otherwise stated.

Statistical significance was measured using parametric testing, assuming equal variance, unless otherwise stated, with standard t-tests for two paired samples used to assess difference between test and control samples. An asterisk (*) indicates 0.01 < P < 0.05; ** indicates 0.001 < P < 0.01; *** indicates P < 0.001; **** indicates P < 0.0001; N indicates 0.05 < P when compared to control.

Power calculations were performed for mammary fat pad experiments. Using 80% power and 95% confidence, 25% practical difference and 15% coefficient of variation we anticipated that 8-10 mice was required for each cohort and so n = 10 animals per cohort were transplanted.

Power calculations were also performed for bone tumour growth assays based on the minimum number of animals required to obtain statistically significant data in a factorial ANOVA design were based on our extensive previous studies: Metastasis is known to develop in the hind limbs of 80–90% of mice injected with control MDA-MB-231 cells, for studies predicted to decrease metastasis (or metastatic lesions) by 70%, a minimum of six mice per group is required to obtain 80% power with 10% error.