Delineating breast cancer cell interactions with engineered bone microenvironments


  • Anna V Taubenberger,

    1. Institute of Health and Biomedical Innovation, Queensland University of Technology, Kelvin Grove, Australia
    2. Australian Prostate Cancer Research Centre Queensland, Brisbane, Australia
    Search for more papers by this author
    • The first two authors contributed equally to this work.
  • Verena M Quent,

    1. Institute of Health and Biomedical Innovation, Queensland University of Technology, Kelvin Grove, Australia
    2. Department of Obstetrics and Gynecology, University Hospital Erlangen, Erlangen, Germany
    Search for more papers by this author
    • The first two authors contributed equally to this work.
  • Laure Thibaudeau,

    1. Institute of Health and Biomedical Innovation, Queensland University of Technology, Kelvin Grove, Australia
    2. Australian Prostate Cancer Research Centre Queensland, Brisbane, Australia
    Search for more papers by this author
  • Judith A Clements,

    1. Institute of Health and Biomedical Innovation, Queensland University of Technology, Kelvin Grove, Australia
    2. Australian Prostate Cancer Research Centre Queensland, Brisbane, Australia
    Search for more papers by this author
  • Dietmar W Hutmacher PhD

    Corresponding author
    1. Australian Prostate Cancer Research Centre Queensland, Brisbane, Australia
    2. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA, USA
    • Institute of Health and Biomedical Innovation, Queensland University of Technology, Kelvin Grove, Australia
    Search for more papers by this author

Address correspondence to: Dietmar W Hutmacher, Institute of Health and Biomedical Innovation, Queensland University of Technology, 60 Musk Avenue, 4059 Kelvin Grove, Queensland, Australia. E-mail:


The mechanisms leading to colonization of metastatic breast cancer cells (BCa) in the skeleton are still not fully understood. Here, we demonstrate that mineralized extracellular matrices secreted by primary human osteoblasts (hOBM) modulate cellular processes associated with BCa colonization of bone. A panel of four BCa cell lines of different bone-metastatic potential (T47D, SUM1315, MDA-MB-231, and the bone-seeking subline MDA-MB-231BO) was cultured on hOBM. After 3 days, the metastatic BCa cells had undergone morphological changes on hOBM and were aligned along the hOBM's collagen type I fibrils that were decorated with bone-specific proteins. In contrast, nonmetastatic BCa cells showed a random orientation on hOBM. Atomic force microscopy-based single-cell force spectroscopy revealed that the metastatic cell lines adhered more strongly to hOBM compared with nonmetastatic cells. Function-blocking experiments indicated that β1-integrins mediated cell adhesion to hOBM. In addition, metastatic BCa cells migrated directionally and invaded hOBM, which was accompanied by enhanced MMP-2 and -9 secretion. Furthermore, we observed gene expression changes associated with osteomimickry in BCa cultured on hOBM. As such, osteopontin mRNA levels were significantly increased in SUM1315 and MDA-MB-231BO cells in a β1-integrin–dependent manner after growing for 3 days on hOBM compared with tissue culture plastic. In conclusion, our results show that extracellular matrices derived from human osteoblasts represent a powerful experimental platform to dissect mechanisms underlying critical steps in the development of bone metastases.


About 70% of women with advanced breast cancer (BCa) develop bone metastases.[1] The consequences of bone metastases are devastating, causing bone pain, pathological fractures, hypercalcaemia, and nerve compression. These skeletal-related events are the result of pathological bone remodeling induced by metastatic BCa cells resident in bone.[2-4] Herein, BCa cells typically secrete factors, such as parathyroid hormone–related protein (PTHrP) and interleukins (e.g., IL-6, -8, and -11), that directly or indirectly induce maturation of osteoclasts, hence increase bone resorption.[3, 5-7] In return, breakdown of bone matrix releases therein-bound mitogenic factors that promote metastatic cell growth. This fuels a vicious cycle, exacerbating the destructive effects on the skeleton.[5, 8, 9] Currently used treatment modalities in bone metastatic disease include agents that counteract bone degradation, such as bisphosphonates and receptor activator of NF-κB ligand (RANKL) antibodies.[10, 11] However, these therapeutic approaches are at best palliative, and there is no cure for bone metastatic disease at present. To develop more effective therapies, a deeper understanding of factors that facilitate preferential colonization of BCa cells in the skeleton must be gained.

Interactions between metastatic BCa cells and the bone microenvironment, i.e., bone cells, bone matrix, and bone marrow cells, are known to play a critical role in homing, growth, and survival of metastatic BCa cells in the skeleton.[8, 12, 13] The bone matrix consists of an organic matrix that is strengthened by crystalline deposits, primarily calcium and phosphate in the form of hydroxyapatite. The organic bone matrix is composed of different types of extracellular matrix (ECM) proteins, predominantly collagen type I (COL I), and further proteins including osteopontin (OPN), osteonectin, osteocalcin, fibronectin (FN), and bone sialoprotein. Moreover, it is replete with growth factors, such as insulin-like growth factor 2, transforming growth factor β, hepatocyte growth factor, and fibroblast growth factor.[14, 15] Metastatic cancer cells can interact with bone ECM proteins via various cell adhesion molecules, primarily integrins, but also other matrix adhesion receptors such as CD44.[14, 16, 17] In addition to providing a mechanical link to ECM, integrins are important signal transducers.[18, 19] It is well established that integrins synergize with growth factors to regulate intracellular signaling pathways that control cell differentiation, proliferation, survival, and migration,[20, 21] all of which are involved in the development and progression of cancers.[22-24] It has been reported, for instance, that overexpression of αvβ3-integrin increases the potential of the BCa cell line MDA-MB-231 to establish bone metastases in mice.[25] Moreover, transfection with α4β1-integrins could promote the formation of bone metastases in chinese hamster ovarian cells in mice.[26] In addition, several in vitro experiments have demonstrated that α2β1-integrin mediates adhesive interactions between prostate cancer cell and bone matrix.[27, 28] This finding may also be relevant for BCa bone metastasis regarding the high predilection to bone shared by prostate and BCa cells.

Most studies exploring interactions between BCa cells and bone are based on in vivo experiments; few studies have applied in vitro models. Although numerous studies analyze gene expression and cell migration on two-dimensional tissue culture plastic surfaces, attempts have also been undertaken to mimic more closely the bone microenvironment. Although in vitro models cannot recapitulate the entire metastatic cascade, they are valuable tools to study specific aspects of bone metastasis systematically and under well-defined conditions. For instance, bone colonization by BCa cells has been studied by culturing MDA-MB-231 cells in mouse calvarian cell-derived matrices[29] or human osteosarcoma cell-derived ECM.[30] Several studies have specifically explored adhesive interactions between BCa cells and single proteins of the bone ECM. For instance, BCa cell attachment was qualitatively studied on COL I, bone sialoprotein,[31] vitronectin,[32] OPN, FN, or thrombospondin[33] coated tissue culture surfaces. Furthermore, BCa cell attachment to cryosections of murine vertebrae[34, 35] or bovine cortical bone discs[36] was qualitatively assessed. Taken together, most studies on BCa interactions with bone matrix have investigated human BCa cell interactions with simplified protein coatings, ECM derived from murine or human osteogenic cell lines, or polished bone sections from murine or bovine origin. These approaches, however, may introduce species- or cell line-related biases that possibly impede the interpretation of these findings, e.g., resulting from differences between matrix components laid down by murine and human osteoblasts.[37]

To overcome these limitations, we set out to investigate interactions between human BCa cells and a human primary osteoblast-derived matrix (hOBM) that has been developed based on bone tissue engineering principles and used in our previous work for studying prostate cancer interactions with bone ECM.[38] To our knowledge, this is the first time that ECM derived from human primary osteoblasts is used to study BCa cell interactions. First, we have characterized and validated hOBM generated from different female patients, thereby confirming their bonelike properties. Thereafter, we seeded BCa cells of different metastatic potential onto hOBM and assessed BCa cell morphology. We quantified BCa cell adhesion to hOBM, and investigated the effect of hOBM on the activation of adhesion-related signaling pathways in BCa cells, BCa cell migration and invasion, and BCa cell gene expression.

Materials and Methods

Cell lines

T47D and MDA-MB-231 cells were from the American Type Culture Collections (Manassas, VA, USA). GFP-expressing SUM1315 cells were kindly provided by David Kaplan, and MDA-MB-231BO cells by the University of Texas Health Science Center at San Antonio. T47D, MDA-MB-231, and -BO cells were maintained in DMEM with 10% fetal bovine serum (FBS) and 100 U/mL penicillin/100 µg/mL streptomycin (Invitrogen, Carlsbad, CA, USA). SUM1315 cells were cultured in F-12 medium supplemented with 5 µg/mL insulin, 10 ng/mL epidermal growth factor (Sigma-Aldrich, St. Louis, MO, USA), 10% FBS, and penicillin/streptomycin. SUM1315, MDA-MB-231, and MDA-MB-231BO cells stably expressing short hairpin RNA (shRNA) targeting integrin β1 RNA or control cells expressing shRNA specific for firefly luciferase-GL2 RNA were generated by retroviral transduction. Virus particles were kindly provided by T Kwok.[39] Knockdown of integrin β1 was verified using standard immunoblotting and flow cytometry.

Preparation of hOBM

Human primary osteoblasts (hOB) were obtained from bone specimens of female patients undergoing knee replacement surgery as approved by the ethics committee of Queensland University of Technology and the Prince Charles Hospital (approval number 0600000232). Nonsclerotic, trabecular bone from the tibial plateau was cut into small bone fragments, which were washed with phosphate-buffered saline (PBS) and transferred into 175 cm2 tissue culture flasks (Nunc, Thermo Fisher Scientific, Australia) containing expansion medium (α-minimal essential medium [α-MEM, Invitrogen], 10% FBS, penicillin 100 IU/mL [Invitrogen], streptomycin 0.1 mg/mL [Invitrogen]). hOBs could be collected within 2 to 3 weeks from this explant culture. To prepare hOBM, 6000 hOB of passage 2 to 3 were seeded onto thermanox coverslips (13 mm diameter, Nunc, Thermo Fisher Scientific). Upon reaching confluency, cells were cultured under osteogenic conditions, i.e., in cell culture medium supplemented with 50 µg/mL ascorbate-2-phosphate, 10 mM β-glycerophosphate, and 0.1 µM dexamethasone (all from Sigma). After a 5- to 6-week culture period, highly confluent mineralized cell sheets were formed. These were decellularized by washing them first with sterile double-distilled water, followed by a 10- to 12-minute incubation step in 20 mM ammoniumhydroxide.[40] Decellularized matrices were rinsed carefully with sterile water followed by several washes in PBS and stored at 4°C until further use.

Preparation of COL I surfaces

A stock solution of rat tail collagen I (1 mg/mL, Sigma) in 0.1 M acetic acid was prepared. Thermanox (Nunc) coverslips (13 mm diameter) were incubated with 5 µg/mL collagen type I in H2O overnight at 4°C. Coated coverslips were rinsed with PBS before blocking with 2% bovine serum albumin (BSA; Sigma) in PBS for 2 hours at room temperature (RT). Wells were washed twice with sterile PBS before use.[38]

Atomic force microscopy (AFM)-based single-cell force spectroscopy (SCFS)

The AFM (Nanowizard II equipped with CellHesion module, JPK Instruments, Berlin, Germany) used was mounted on an Axiovert 200 inverted light microscope (Zeiss, Thornwood, NY, USA). Cantilevers (NPS, k = 0.06 nN/m; NanoWorld, Neuchatel, Switzerland) were functionalized with concanavalin A overnight[41] and calibrated using built-in procedures of the JPK software. Prior SCFS measurements cells were harvested with TripLE. For integrin β1-blocking experiments, cells were pre-incubated with 6 μg/mL P5D2 (DSHB) for 30 minutes. Experiments were conducted at 37°C using a temperature-controlled chamber (Petridishheater, JPK Instruments) in CO2-independent, serum-free medium (Invitrogen). After attaching a single BCa cell to the cantilever,[41] the cell was probed on COL I or hOBM in constant height mode (closed-loop) using a retraction speed of 5 µm/second and a contact force of 750 pN. F-D curves were analyzed using the JPK image processing software.

Cell migration analysis

40,000 BCa cells were seeded onto TCP, COL I, or OBM and imaged with a Leica (Wetzlar, Switzerland) AF6000 LX wide-field microscope at 37°C and 5% CO2. Differential interference contrast images were recorded every 16 minutes for a total period of 48 hours. Image stacks were analyzed in Image J using the “manual tracking tool.” From the X and Y positions, the mean square displacement of each cell was calculated (equation (1)) using a function written in Igor Pro. Mean square displacements were averaged for each substrate, and plotted as a function of time between steps. In Igor Pro, data were fitted to the persistent random walk (PRW) model (equation (2)),[42] and persistence time P and random motility coefficient μ were extracted,

display math(1)
display math(2)

where <x2> equals the mean square displacement (MSD), N is the number of frames, and n is the number of combined time intervals. At least 64 individual cells from a total of eight different movies were analyzed.


Human primary osteoblast-derived matrix (hOBM) closely resembles native bone matrix

To mimic the bone metastatic microenvironment, we generated cell-free 3D matrices secreted by hOBM isolated from human trabecular bone.[38] Scanning electron microscopy images (SEM) revealed that hOBM consisted of a dense fibrillar meshwork (Fig. 1A) and resembled in their macromolecular assembly native trabecular bone (Supplemental Fig. S1A, B). hOBM had an average thickness of about 5 µm, as determined from confocal z-stacks of fluorescently labeled matrices (Supplemental Fig. 1C). Alizarin Red staining confirmed that hOBM were highly mineralized (Fig. 1B). Immunohistochemical analysis further demonstrated that hOBM contained characteristic bone proteins, such as collagen type I, fibronectin, OPN, osteonectin, and osteocalcin (Supplemental Fig. S1D). All experiments presented here were carried out on hOBM deposited by osteoblasts of at least 3 different patients, and no significant differences in matrix morphology, composition, or mineralization were found. This high reproducibility is in line with our previous findings[38] and unpublished tests comprising in total more than 15 patient samples. Thus, we conclude that hOBM represents a well-reproducible 3D cell culture system resembling the human bone ECM.

Figure 1.

Characterization of hOBM and analysis of BCa cell morphology on TCP, COL I, and hOBM. (A) SEM images of hOBM at two different magnifications. The higher-resolution image reveals the fibrillar composition of hOBM (fibril direction indicated by red arrow). (B) Alizarin Red staining of hOBM indicating a high degree of matrix mineralization. Inset shows an Alizarin Red staining of a control matrix grown under nonosteogenic conditions. (C) Representative phase contrast (row 1) and confocal laser scanning microscopy (rows 2 to 4) images of BCa cells grown for 3 days on hOBM, COL I, or TCP. To visualize f-actin and nuclei, cells were stained using rhodamine-phalloidin (red) and DAPI (blue), respectively. Maximal projections of confocal z-stacks are shown. (D) Polar plot showing the orientation of BCa cells calculated in Metamorph from maximal projections such as shown in (C). Gray area indicates a total angular range of 45°, where 0° is the average orientation angle of all cells.

Metastatic BCa cells adopt a spindle-like morphology on hOBM and align along its fiber architecture

To investigate interactions between BCa cells and hOBM, we then seeded a panel of four BCa cell lines of different metastatic potential on hOBM. A nonmetastatic BCa cell line of luminar phenotype, T47D, and three metastatic cell lines of basal phenotype, namely SUM1315,[43] MDA-MB-231, and MDA-MB-231BO, were chosen. The latter is a highly bone-metastatic cell line derived from MDA-MB-231 cells.[44] The assignment of the metastatic potential of these cells lines was based on their propensity to metastasize in immune-suppressed mice.[43, 45] First, we examined the spreading morphology of these BCa cell lines (Fig. 1C) after 3 days of culture on hOBM. Tissue culture plastic (TCP) and COL I, a nonfibrillar coating of collagen type I, the most abundant protein in bone, served as reference culture substrates. Whereas a similar spreading morphology of BCa cells was observed when grown on TCP and COL I, the more aggressive SUM1315, MDA-MB-231, and -BO cells underwent drastic morphological changes and aligned parallel to each other when grown on hOBM. To quantitatively describe these morphological changes, we measured the axial ratio (shape factor), spreading area, and orientation of BCa cells. Consistent with the above qualitative observations, BCa cells predominantly displayed the most rounded and most spread morphology on COL I, whereas they rearranged their cytoskeleton toward a more spindle-like morphology on hOBM (Supplemental Fig. S2). The majority of SUM1315 (65%), MDA-MB-231 (55%), and -BO (70%) cells oriented their longer axis within a narrow angular range of 45° (gray area, Fig. 1D), whereas T47D showed a more random orientation (22%), similar to what was seen for all cell lines on TCP and COL I. Further, staining of hOBM with antibodies directed against FN revealed that the metastatic BCa cells aligned along the fibrillar structures of hOBM (Supplemental Fig. S3).

Metastatic BCa cells show enhanced adhesion to hOBM

BCa cell adhesion to bone matrix has been suggested to play an important role in BCa cell colonization of the skeleton.[46] Thus, we set out to quantify BCa cell adhesion to hOBM in comparison to COL I by AFM-based single-cell force spectroscopy (AFM-SCFS). Briefly, a single living BCa cell was attached to an AFM cantilever[47] (Fig. 2A), brought in contact with hOBM or COL I for a defined contact time, and retracted again, while the force acting on the cantilever was recorded over the traveled piezo distance. From each resulting force-distance (F-D) curve, the maximal force required to detach the cell, the so-called detachment force (Fd, Fig. 2B) was measured to describe overall cell adhesion.[48] Contact times of 5 seconds or 120 seconds were chosen; at this time scale, adhesion can be monitored from initial cell attachment to the formation of more complex adhesion structures.[41, 49] Notably, median detachment forces measured after 5 seconds of contact with hOBM augmented significantly with increasing metastatic potential of the BCa cell lines, ranging from 297 pN detected for T47D cells to 798 pN for MDA-MB-231BO cells (Fig. 2C). Strongly elevated detachment forces were measured for the metastatic cell lines after a 120-second contact to hOBM compared with T47D cells, which indicated a considerable reinforcement of adhesion during this contact period (e.g., about 20-fold for SUM1315) (Fig. 2C). A distinct adhesion pattern was recorded when cells were probed on COL I, whereby the highest detachment forces were—similar to hOBM—detected for MDA-MB-231BO cells (Fig. 2D).

Figure 2.

Quantification of adhesion of BCa cells to COL I and hOBM by AFM-SCFS. (A) Light microscopy image showing the AFM cantilever with an attached BCa cell (encircled) above hOBM. (B) Schematics illustrating the cell adhesion measurement. At the beginning of the measurement, the cell, attached to the force probe (AFM cantilever), is far from the adhesive substrate (hOBM or COL I). Then, by the precise movement of piezo elements, the cell is brought in contact with the adhesive substrate (see gray force-distance curve) and allowed to interact with it for a preset contact time. Then, the cantilever is withdrawn, which results in separation of the cell from the substrate (green). The maximal force required to detach the cell (= detachment force Fd) can be measured from the retraction curve (arrow). (C, D) Detachment forces of BCa cells adhering to hOBM or COL I, respectively, for contact times of 5 and 120 seconds. Numbers denote the total of analyzed F-D curves, numbers in parentheses (below) equal probed cells. Data in (C) and (D) are represented as box-whisker plots. Top and bottom of the boxes indicate the 75th and 25th percentiles, and upper and lower whiskers the 90th and 10th percentiles. Solid lines within boxes denote median, and dotted lines denote mean detachment forces. The number of analyzed F-D curves and probed cells (<n>) is given. Statistical differences among cells analyzed in (C) and (D) were tested using nonparametric tests (Kruskal-Wallis). Contrasts were calculated for the individual pairs. Significant differences (p < 0.05) among pairs are indicated as follows: *, any–T47D; #, any–SUM1315; , MDA-MB-231–MDA-MB-231BO.

BCa cell adhesion to hOBM is mediated by β1-integrins

To get insights into the adhesion receptors mediating cell adhesion to hOBM and COL I, we further examined BCa cell surface and gene expression levels of α2-, α5-, β1-, αv-, and β3-integrins by flow cytometry (Fig. 3A, Supplemental Fig. S4B) and qRT-PCR (Supplemental Fig. S4A), respectively. Hereby, different integrin expression patterns were found among the cell lines. The cell surface levels of α2-integrin correlated well to the respective detachment forces measured on COL I, unlike on hOBM (Fig. 2C, D), suggesting that α2-integrin mediated cell adhesion to COL I but had a less important role in the attachment of BCa cells to hOBM. In contrast, a high correlation between β1-integrin cell surface levels and detachment forces on hOBM was observed, for which we decided to investigate the role of β1-integrins in BCa cell adhesion to hOBM using AFM-SCFS. In the presence of β1-integrin function-blocking antibody (P5D2), detachment forces for T47D, MDA-MB-231, and -BO cells were significantly reduced (Fig. 3B). Although P5D2 had only a minor blocking effect on SUM1315 cells at 5 seconds, it did significantly block adhesion at prolonged contact times (not shown). We further observed that at 6 hours after seeding, P5D2 significantly inhibited spreading of all cell lines on hOBM (Fig. 3C, D). Consistently, knockdown of β1-integrins significantly reduced spreading of the metastatic BCa cell lines. These findings indicate a role for β1-integrins in mediating adhesion and spreading of these cells on hOBM. Because SUM1315, in contrast to the other cell lines, expressed considerable levels of αvβ3-integrins (Fig. 3A) and adhered strongly to hOBM (Fig. 2C), we further investigated the potential role of αvβ3-integrins in SUM1315 cell attachment to hOBM. However, no reduction of cell detachment forces was seen when using a respective blocking antibody (data not shown), which indicated that αvβ3-integrins did not mediate cell adhesion to hOBM.

Figure 3.

(A) Cell surface levels of integrin subunits (α2, α5, β1) or integrin heterodimers (αvβ3) measured by flow cytometry. (B) Detachment forces of BCa cells adhering to hOBM (5-second contact) in presence of 6 µg/mL β1-integrin function-blocking antibody (P5D2). Data are represented as box-whisker plots. Top and bottom of the boxes indicate the 75th and 25th percentiles, and upper and lower whiskers the 90th and 10th percentiles. Solid lines within boxes denote median, and dotted lines denote mean detachment forces. Data sets recorded in presence or absence of the blocking antibody were compared using a nonparametric test (Mann-Whitney). Significant differences (p < 0.05) among pairs are indicated by *. As above, numbers of analyzed F-D curves/probed cells are given. (C) Mean spreading area (± standard deviation) of cells treated with P5D2 normalized to control cells. At least 90 cells were analyzed per cell type/condition. (D) Spreading of control and β1-integrin knockdown cells evaluated after 6 hours on hOBM. Given is the percentage of spread cells. β1-integrin knockdown was confirmed by Western blot analyis (left). Data sets of control and β1-integrin knockdown cells were compared using a t test. Significant differences (p < 0.05) among pairs are indicated by *.

BCa cell attachment to hOBM stimulates phosphorylation of FAK and ERK1/2

Integrins synergize with growth factors to regulate intracellular signaling pathways that control cell differentiation, proliferation, survival, and migration.[50] Intracellular signaling molecules downstream of integrins include focal adhesion kinase (FAK) and the nonreceptor tyrosine kinases Src and Fyn, which mediate the ligand-mediated activation of mitogen-activated protein kinase (MAPK) and PI3K signal transduction pathways.[51, 52] Western blot and densitometric analysis showed the highest levels of phosphorylated FAKTyr397 and extracellular signal-regulated kinase (ERK1/2Thr202/Tyr204) on hOBM for all cell lines except MDA-MB-231BO. The latter showed the highest FAKTyr397 phosphorylation levels on COL I, and comparable ERK1/2 phosphorylation levels on COL I and hOBM (Fig. 4A, B). High levels of FAK phosphorylation in MDA-MB-231BO cells grown on COL I were consistent with the strong adhesion by these cells to COL I. Increased FAK expression and activity in BCa specimens are frequently associated with poor prognosis.[53] FAK activity modulates not only tumor cell growth and survival but also directs cell migration, which requires coordinated and dynamic regulation of adhesive contracts and the cytoskeleton networks they connect with.[51] Although hOBM did not stimulate BCa cell proliferation compared with TCP or COL I for any of the analyzed cell types (Supplemental Fig. S5), we next set out to investigate BCa cell migration.

Figure 4.

Analysis of integrin downstream signaling pathways on COL I and hOBM. (A) Representative Western blot of BCa cell lysates after 2-hour attachment to TCP, COL I, or hOBM. Total and phosphorylated (Tyr397) levels of FAK and total and phosphorylated levels of ERK1/2 are shown. α-tubulin served as loading control. (B) Quantification of FAK and ERK1/2 phosphorylation levels. Total and phosphorylated levels were measured by densitometry of Western blots and normalized to loading controls. The ratio of phosphorylated to total levels is shown (n = 3 each). Significant differences (p < 0.05; one-way ANOVA) among pairs are indicated as follows: , TCP–COL I or TCP–hOBM; , COL I–hOBM.

BCa cells show directed cell migration within hOBM

Thus, we conducted time-lapse videomicroscopy on the three substrates over a period of 48 hours (Supplemental Movies S1–4). We observed that the majority of BCa cells (>95%) migrated on hOBM in a mesenchymal migration mode (opposed to amoeboid movement), in accordance with the elongated spreading morphology we had observed before. To quantitatively describe cell migration, individual cells were tracked in Image J (Fig. 5A, C), and their mean square displacements were calculated (Fig. 5B, D). These were fitted to a persistent random walk model, and the migratory coefficient and persistence time were extracted (Table 1).[42] The higher persistence times measured on hOBM versus COL I and TCP indicated that the cells migrated directionally on hOBM but randomly on COL I and TCP. MDA-MB-231 and SUM1315 showed the largest displacement and the most persistent migration on hOBM. MDA-MB-231BO cells were less motile than the parental cell line, whereas T47D cells barely advanced (Fig. 5C, Table 1).

Figure 5.

Analysis of BCa cell migration on COL I and hOBM. (A, C) Representative tracks of individual BCa cells migrating on COL I or hOBM, respectively, recorded in time-lapse movies over 48 hours. (B, D) Mean square displacements (MSD) calculated for BCa cells (n = 64). Dotted lines represent fits by the persistent random walk (PRW) model.

Table 1. Migratory Parameters Extracted From Persistent Random Walk Fits of Mean Square Displacements of BCa Cells Migrating on COL I and hOBM
Random motility coefficient µ (µm2/min)Persistence time P (min)Random motility coefficient µ (µm2/min)Persistence time P (min)
T47D2.8 ± 0.00 ± 221.2 ± 0.0477 ± 39
SUM13157.4 ± 0.1324 ± 1419.8 ± 0.2920 ± 24
MDA-MB-23110.6 ± 0.00 ± 026.2 ± 0.5934 ± 40
MDA-MB-231BO7.0 ± 0.0142 ± 106.0 ± 0.1419 ± 25

Metastatic BCa cells invade hOBM and upregulate matrix metalloproteinases (MMPs)

We next set out to study cell invasion into the fibrillar network of hOBM. SEM images taken of BCa cells growing on hOBM for 3 days indicated that a major proportion of SUM1315, MDA-MB-231, and -BO cells were entangled in FN-containing hOBM fibers. In contrast, T47D cells were exclusively located on top of hOBM (Fig. 6A, arrows). To quantify the numbers of cells that had invaded hOBM, we recorded z-stacks of phalloidin/DAPI-stained BCa cells on hOBM, where FN was visualized by immunofluorescent labeling. Analysis of respective cross sections confirmed that no T47D cells had invaded hOBM, whereas about 80% of SUM1315, MDA-MB-231, and -BO cells were partly covered by fibronectin fibrils (Fig. 6B).

Figure 6.

Studying cell invasion into hOBM. (A) SEM images of BCa cells growing on hOBM at day 3. Arrows highlight cells that are either located on top of hOBM (T47D) or entangled within the fibrillar meshwork of hOBM. (B) Orthogonal sections of confocal z-stacks recorded of phalloidin-rhodamine and DAPI-stained BCa cells after 3-day culture on hOBM. FN in hOBM is stained by immunofluorescence (green). (C) Relative expression of MMP-2, -9, and -10 (calculated as: 2^ctgene of interest- ctgeometric mean of house-keeping gene) measured by qRT-PCR. Normalized values are presented. (D) Gelatin zymography using supernatants collected after 48-hour BCa cell culture on TCP or hOBM under serum-free conditions. The molecular weights of the molecular markers are shown on the left. One percent serum and MMP-2 and -9 standards were loaded as references. Controls (right) are medium supernatants in absence of BCa cells on TCP or hOBM. (Below) Densitometric quantification of zymography bands corresponding to the active forms of MMP-2/-9 (n = 3). Normalized values are presented. MMP-2/-9 levels of cells grown on TCP and hOBM were compared using a t test (p < 0.05).

Cell migration through mesenchymal compartments is believed to be partly mediated through active proteolysis.[54] qRT-PCR indicated that the nonmetastatic T47D cells had only negligible mRNA levels of metalloproteinases MMP-2, -9, and -10. In contrast, high MMP-2 and -10 levels were detected for SUM1315 (Fig. 6C). In accordance with previous reports, MDA-MB-231 and -BO cells expressed significant amounts of MMP-9 but not MMP-2.[17] Notably, expression of MMP-2 and -10 by SUM1315 cells was significantly increased on hOBM compared with TCP. Conversely, a decrease of MMP-9 mRNA levels was detected on hOBM for MDA-MB-231 cells. To investigate the functional forms of MMP-2 and -9, gelatin zymographies were employed (Fig. 6D). hOBM alone contained significant levels of pro-MMP-2 but no active MMP-2, such as previously described for the matrix of bone marrow–derived fibroblasts.[55] Active MMP-2 was only detected in supernatants of SUM1315 cells cultured on hOBM but not TCP. Similarly, levels of active MMP-9 were significantly higher for MDA-MB-231 and -BO cells cultured on hOBM than on TCP (Fig. 6D). The observed changes in the active levels of MMP-2 and -9 motivated us to further test their role in BCa cell invasion into hOBM using 20 µM of the potent broad-sprectrum MMP inhibitor GM6001, which also targets MMP-2 and -9. After 3 days of culture on hOBM, no differences in the invasion of SUM1315, MDA-MB-231, and -BO cells were seen (Supplemental Fig. S6). This indicates that although the levels of MMP-2 and -9 were modulated by hOBM, their proteolytic actions were not responsible for BCa cell invasion into hOBM.

BCa cell growth on hOBM increases their OPN and PTHrP expression levels

It is well established that bone metastatic BCa cells can produce bone matrix proteins, such as OPN, bone sialoprotein,[56] or osteoprotegerin.[57] To investigate changes in gene expression upon BCa cell growth on hOBM, we performed qRT-PCR for a set of candidate osteogenic genes, such as osteopontin (OPN), osteonectin, osteocalcin, osteoprotegerin, and bone sialoprotein. Osteocalcin and bone sialoprotein were not detectable at the mRNA level for any of the analyzed cell lines (not shown). SUM1315 cells expressed relatively high levels of OPN and osteonectin compared with the other BCa cell lines. MDA-MB-231BO cells had relatively high levels of osteoprotegerin (more than fourfold compared with MDA-MB-231) (Supplemental Fig. S7), which is in line with previous reports showing that osteoprotegerin is an indicator of high bone metastatic potential.[57] Notably, OPN mRNA levels of both SUM1315 and MDA-MB-231BO cells increased significantly when cells were grown on hOBM compared with TCP and COL I (Fig. 7). Interestingly, this increase was β1-integrin-dependent because β1-integrin knockdown cells showed an attenuated response to hOBM (Fig. 7). MDA-MB-231 cells had only negligible OPN mRNA levels (Supplemental Fig. S7). We further analyzed factors implicated in the osteolytic phenotype of BCa bone metastasis, such as IL-6 and -8. However, we did not detect any significant changes in the expression of these genes (not shown).

Figure 7.

Relative expression levels of osteogenic markers. qRT-PCR was performed in triplicate with cDNA from BCa cells after 3-day culture on TCP, COL I, or hOBM. (A) Relative ITGB1 (β1-integrin gene) expression was compared among control and β1-integrin knockdown cells grown on TCP. (B, C) Relative OPN expression levels for SUM1315 and MDA-MB-231BO cells (controls and β1-integrin knockdown cells) that had been grown for 3 days on TCP, COL I, or hOBM. Data were normalized to control cells. GAPDH and β-actin were used as housekeeping genes. All data are presented as mean ± standard deviation. Relative expression levels of TCP, COL I, and hOBM groups were compared using a one-way ANOVA test. Significant differences are indicated as follows: , TCP–hOBM; #, COL I– hOBM. Significant differences between respective controls and β1-integrin knockdown cells are indicated by *.


In the present work, we set out to unravel BCa cell interactions with an engineered bonelike microenvironment produced by human primary osteoblasts. Using a cell-free bonelike matrix allowed us to dissect the influence of the bone extracellular matrix on BCa cells without the complexity of cell-cell interactions. This is of particular interest because the extracellular matrix components of bone have been proposed to be an important mediator of tumor cell adhesion, migration, and invasion at the metastatic site.[58]

First, we have shown that the metastatic BCa cell lines adopted an elongated spreading morphology on hOBM, aligning along its fibers. This observation is in line with previous reports, where MDA-MB-231 aligned along fibers of tumor-associated fibroblast-derived matrices.[59] Thus, the morphological changes observed may be indicative of a more general cellular response to fibrillar protein assemblies, disregarding the differences in composition of fibroblast- or osteoblast-derived ECMs. In addition, Castello-Cros and colleagues reported that different from MDA-MB-231, MCF-7 cells did not align along their fibroblast matrix.[59] This observation is in line with our observation that T47D cells, having a similar luminar phenotype as MCF-7s, did not elongate along the fibrils of hOBM after 3 days but formed small cell colonies. This ability of the metastatic BCa cells versus nonmetastatic BCa cells to respond to fibrillar cues of the respective matrices may be related to epithelial-mesenchymal transition (EMT)-related changes undergone by SUM1315, MDA-MB-231, and -BO cells, such as evident by loss and gain of E-cadherin and vimentin, respectively (Supplemental Fig. S4D). Interestingly, for T47D cells, a decrease in the expression of the epithelial marker cytokeratin 8 was detected after 3 days of culture on hOBM (not shown). Taken together, these observations emphasize the epithelial-mesenchymal plasticity of these cells in the hOBM microenvironment, which is in agreement with what we have previously observed for prostate cells on hOBM.[38]

Concomitantly with the elongated spreading morphology adopted by BCa cells on hOBM, the metastatic BCa cells showed a more effective directional migration on hOBM compared with T47D cells. Directional migration of fibroblasts along similar fibrillar assemblies of cell-derived matrices has been previously described.[60] The study reports significant differences in cell migration on TCP versus cell-derived matrices, where the latter ones only supported persistent cell migration. Importantly, the directed cell migration observed for our BCa cells on hOBM shared hallmarks of the migration of BCa cells in vivo. Condeelis and colleagues showed by multiphoton microscopy on mice with green fluorescent protein-labeled tumors that metastatic BCa cells are also moving linearly along ECM fibers.[61] In contrast to these early stages of tumor progression, however, little is known about the migration pattern of metastatic BCa cells within the bone microenvironment. We hypothesize that metastatic BCa cells also would show a similar migration pattern when they meet fibrillar ECM arrangements, such as found in the bone matrix in vivo.

Whereas the enhanced directional migration of the metastatic BCa cells fits to the elongated spreading morphology they had adopted on hOBM, differences in migration rates among the different cell lines may relate to the differences in their adhesion to hOBM, such as observed in our AFM-SCFS experiments. Although adhesion is required to form integrin complexes at the leading edge of migrating cells, high adhesion can also limit cell motility.[62] Using AFM-SCFS, we further showed that the metastatic BCa cells adhered more strongly to hOBM in a β1-integrin–dependent manner. These findings are in line with previous studies, in which MDA-MB-231 cell attachment to bovine[36] and murine bone discs[33] was studied. Previously, αvβ3-integrin has been suggested to be a key adhesion receptor involved in BCa metastasis,[22] albeit it may have a more important role during BCa cell homing rather than colonization.[25] Among the BCa cell lines used in our study, only SUM1315 cells showed considerable levels of αvβ3-integrin, and a functional blocking antibody directed against αvβ3-integrin did not reduce cell adhesion forces to hOBM. Hence, we conclude that αvβ3-integrins did not mediate cell adhesion to the hOBM. To our knowledge, we have quantitatively shown for the first time that metastatic BCa cells adhere via β1-integrins to bone ECM. Because enhanced cell adhesion correlated with the bone-metastatic potential of BCa cells, our data suggest that increased cell adhesion of BCa cells to the bone ECM is relevant to the process of bone colonization by BCa cells, and may therefore present a target for the development of therapies interfering with BCa bone colonization.

Moreover, we observed that signaling proteins, such as FAK and ERK1/2, were modulated by hOBM. Remarkably, T47D, SUM1315, and MDA-MB-231 cells significantly increased their levels of FAK and ERK1/2 phosphorylation when seeded onto hOBM compared with TCP and Col I. In contrast, the bone metastatic MDA-MB-231BO showed the least changes on hOBM, which may indicate that they may have undergone changes related to these signaling pathways during the in vivo selection process in the bone microenvironment. We speculate that culturing the less bone metastatic BCa cells for prolonged periods on hOBM may induce changes associated with the bone metastatic phenotype in these cells, a hypothesis that we will address in our future work.

We further showed that the metastatic BCa cell lines expressed MMP-2, -9, and/or -10 in contrast to T47D cells. MMPs have been recently described as master regulators of the vicious cycle of bone metastasis.[63] We demonstrated that active forms of MMP-2 were increased in supernatants of SUM1315, such as MMP-9 protein levels were elevated in supernatants of MDA-MB-231 and -BO cells on hOBM. Although MMP-2 mRNA and protein levels in culture supernatants were directly correlated, discrepancies were observed for MDA-MB-231 and -BO cells. This may be explained by the fact that the amount of secreted proteases can be regulated beyond the transcriptional level by mechanisms controlling their secretion, turnover, and processing by other proteases.[64]

Besides gelatin, MMP-2 and -9 can also degrade fibronectin, laminin, vitronectin, and different types of collagens.[65] We speculate that cellular interactions with components of hOBM are responsible for modulating MMP levels, which is in line with previous reports showing that interactions with FN[66] or TGF-β,[67] both highly expressed in the native bone microenvironment and hOBM,[38] can lead to an increase of active MMP-9. However, invasion of BCa cells into hOBM could not be blocked by the MMP inhibitor GM6001, which also targets MMP-2 and -9. Still, the finding of upregulation of MMP-2 and -9 levels is interesting beyond the aspect of ECM degradation and invasion. Over the past years, increasingly further important functions of MMPs have been acknowledged, such as processing and activation of matrix-bound growth factors, which increase their bioavailability. Highly relevant targets of MMP-2 and -9 with respect to BCa bone colonization include VEGF and TGF-β.[65]

We have further seen significant changes in the expression of OPN upon SUM1315 and MDA-MB-231BO cell growth on hOBM, which were modulated by β1-integrin expression. Interestingly, upregulation of OPN in SUM1315 cells growing on hOBM was accompanied by upregulation of MMP-2 (Fig. 6) and integrin αvβ3 (data not shown), both interaction partners of OPN. Upregulation of OPN within the bone microenvironment may have important implications for the formation of osteolytic lesions in vivo, e.g., by promoting the differentiation and activity of osteoclasts.[68] Indeed, a role of OPN in the formation of osteolytic lesions has been recently demonstrated in vivo, where administration of nanoparticles loaded with OPN-antisense oligonucleotides successfully reduced the incidence and size of osteolytic lesions in nude rats.[69] Using hOBM as an experimental platform, future work could further investigate the mechanisms leading to OPN upregulation, e.g., testing specifically the role of TGF-β and other growth factors present on gene expression of BCa cells cultured within hOBM.

Together, we have detected significant differences in the responses of BCa cells to hOBM and 2D reference substrates TCP and COL I, which were chosen because they are still most commonly used in cancer research. Apart from its chemical cues that affect cell adhesion, the macromolecular structure of hOBM permitted the study of cell migration and invasion. From our results, it can be concluded that hOBM more closely resembles in its chemical and structural composition the physiological bone microenvironment than 2D protein coatings. Using a cell-free matrix further allowed us to specifically study the effects of the bone extracellular matrix on gene expression and protein levels. Thus, we conclude that hOBM represents a well-characterized and validated cell culture platform that allows one to systematically study cellular processes relevant to BCa bone colonization.


All authors state that they have no conflicts of interest.


We thank D Loessner and the cancer model group for discussions, T Friis and K Schrobback for help with qRT-PCR; the Australian National Fabrication Facility, the Analytical Electron Microscopy Facility, and C Theodoropoulos for LSM and SEM assistance; the University of Texas Health Science Center for providing MDA-MB-231BO cells; and D Kaplan for providing SUM1315 cells. The monoclonal antibodies P5D2, MPIIIB10, and HFN 7.1, developed by EA Wayner, M Solursh/A Franzen, and RJ Klebe, respectively, were obtained from DSHB developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biology, Iowa City, IA 52242. Retrovirus particles for the generation of stable β1-integrin knockdown cells were kindly provided by T Kwok. The project was supported by Australian Research Council (ARC), DAAD, and the Deutsche Forschungsgemeinschaft (DFG).

Authors' roles: Study design: DWH, VR, AT, and JAC. Data collection: VR, AT, and LT. Data analysis/interpretation: AT and VR. Drafting manuscript: AT and VR. Revising manuscript content: DH, VR, AT, and JC. Approving final version of manuscript: DWH, AT, and JC. AT and VR take responsibility for the integrity of the data analysis.