Mesenchymal stem cell induced DDR2 mediates stromal-breast cancer interactions and metastasis growth

Summary

Increased collagen deposition by breast cancer (BC)-associated mesenchymal stem/multipotent stromal cells (MSC) promotes metastasis, but the mechanisms are unknown. Here, we report that the collagen receptor Discoidin Domain Receptor 2 (DDR2) is essential for stromal-BC communication. In human BC metastasis DDR2 is concordantly upregulated in metastatic cancer and multipotent mesenchymal stromal cells. In MSCs isolated from human BC metastasis DDR2 maintains a fibroblastic phenotype with collagen deposition and induces pathological activation of DDR2 signaling in BC cells. Loss of DDR2 in MSCs impairs their ability to promote DDR2 phosphorylation in BC cells, BC cell alignment, migration, and metastasis. Female ddr2-deficient mice homozygous for the slie mutation show inefficient spontaneous BC metastasis. These results document a role for mesenchymal stem cell DDR2 in metastasis, and suggest a therapeutic approach for metastatic BC.

Graphical Abstract

INTRODUCTION

Distant metastases occur in ∼20% of patients with breast cancer, originate in all breast cancer subtypes, and are the main cause of death in patients. Inhibiting metastatic spread or halting the growth and invasiveness of established metastasis will improve survival. Although the primary tumor microenvironment was shown to promote breast tumorigenesis, the role of the metastasis-associated microenvironment in supporting metastasis growth and invasiveness has remained elusive.

The stroma associated with breast cancer metastasis consists of extracellular matrix (ECM) proteins including fibrillar collagens, and a cellular component comprised of mesenchymal stem/multipotent stromal cells (MSCs), vessels, and cells of the immune system (Pein and Oskarsson, 2015). MSCs are stromal progenitor cells that participate in tissue maintenance under normal conditions and mediate pathological stromal responses in injury repair and tumorigenesis (Karnoub et al., 2007). MSCs are recruited to the primary tumor site from the bone marrow in response to tumor derived soluble factors to promote cancer progression through mechanisms that involve cytokines and collagen deposition (Berger et al., 2015; El-Haibi et al., 2012; Kalluri and Zeisberg, 2006; Karnoub et al., 2007; Liu et al., 2011; Madar et al., 2013). Likewise, adipose-derived MSCs have been shown to enhance tumor volume of MDA-MB-231 xenografts and to promote cell proliferation and invasion of ovarian cancer (Chu et al., 2015; Millet et al., 2016). However, the underlying mechanisms of MSC function need further investigation.

Fibrillar collagen is an important component of the primary tumor and the metastasis-associated stroma. In human breast cancer and transgenic mouse models, increased collagen in the tumor stroma correlates with development of metastases (Kauppila et al., 1998; Provenzano et al., 2008). In primary tumors, increased collagen enhances stromal stiffness to promote breast cancer cell invasion and provides a path for cell migration, especially at the tumor edges (Conklin et al., 2011; Provenzano et al., 2008). In addition to this mechanical effect, increased collagen initiates intracellular signaling to promote tumorigenesis through specific receptors including integrins and the discoidin domain receptors (DDRs). At present, the role of collagen-initiated signaling at the metastatic site is unknown.

DDR1 and DDR2 are collagen receptor tyrosine kinases; DDR1 is activated by fibrillar (mainly type I and V) and non-fibrillar (type IV) collagen and DDR2 is exclusively activated by fibrillar collagen (Alves et al., 1995; Olaso et al., 2002; Valiathan et al., 2012). In normal tissues, DDR1 is expressed in epithelial cells and DDR2 only in mesenchymal stromal cells, where it regulates ECM synthesis and participates in wound healing (Alves et al., 1995; Olaso et al., 2002; Valiathan et al., 2012). Pathologic DDR2 upregulation has been reported in several human malignancies including breast cancer where it is significantly associated with worse survival (Badiola et al., 2011; Barcellos-Hoff et al., 2013; Kim et al., 2015; Toy et al., 2015; Valiathan et al., 2012; Zhang et al., 2013). In breast cancer cells, DDR2 was shown to maintain Snail1 activity and an epithelial to mesenchymal transition (EMT) phenotype (Zhang et al., 2013). DDR2 expression in cancer cells and in cancer associated fibroblasts is important for metastasis in the mouse mammary tumor virus-polyoma middle T antigen (MMTV-PyMT) model (Corsa et al., 2016). However, the underlying mechanisms leading to DDR2 upregulation in cancer, and the role of MSC-derived DDR2 in breast cancer progression are unknown.

Here, we show that DDR2 protein is concordantly upregulated in MSCs and cancer cells in clinical samples of human breast cancer metastasis to several organs. DDR2 expression in MSCs derived from breast cancer metastasis (Met-MSCs) regulates their stromal phenotype and function, and mediates their effect on breast cancer cell neoplastic functions. Homozygous slie mutated mice with absent ddr2 developed significantly fewer and smaller syngeneic breast cancer metastases compared to heterozygous slie and wild-type mice. Our data reveal that MSC-derived DDR2 initiates a stroma-cancer signaling axis leading to DDR2 upregulation in breast cancer and enhancing growth of metastasis. We provide the foundation to block stromal DDR2 as a potential therapeutic strategy for metastatic breast cancer.

RESULTS

DDR2 is elevated in mesenchymal stem/multipotent stromal cells (MSC) and in cancer cells at the metastatic site

In human samples of breast cancer metastasis to distant sites, we found that DDR2 protein was expressed in tumor cells and in adjacent mesenchymal stromal cells expressing the stem cell marker ALDH-1 (Kusuma et al., 2016), in 17 of 21 (81%) cases (r=0.97, p=0.0001) (Figures 1A–B, S1A, and Table S1).

Figure 1. MSCs are present in clinical samples of distant breast cancer metastasis.

A. DDR2 protein is concordantly expressed in cancer cells and in adjacent mesenchymal stromal cells in human breast cancer metastasis. Immunohistochemical analysis for DDR2 expression in human tissue samples of 21 distant breast cancer metastasis. Shown are representative pictures of metastasis to the bone, brain, small intestine, and liver. Insets show expression of DDR2 in mesenchymal stromal cells (scale bars, 10 µm). Results are tabulated in B.

C. Photograph of Met-MSCs derived from a lymph node metastasis (LN) and liver metastasis (Lv) under phase contrast microscopy (original magnification 100x), and in differential culture conditions. Bone marrow- and adipose-derived MSCs were controls. Met-MSCs demonstrate multipotent differentiation capacity towards bone, adipose, and cartilage. Specific cell stains used were Alizarin Red-S for bone, alcian blue for cartilage, and Oil Red O for adipose. Original magnification 100x (scale bars, 50 µm).

D. Met-MSCs exhibit a normal karyotype. Shown is a karyotype at passage 3.

E. Table shows the results of FACS analysis of Met-MSCs, BM-MSCs, and AD-MSCs demonstrated positivity for CD44, CD90, CD73, and CD105 characteristic of MSCs.

To investigate the relevance of DDR2 expression in MSCs from the metastatic microenvironment in a physiologically relevant model, we isolated MSCs from fresh human breast cancer metastasis to a supraclavicular lymph node and to the liver (LN- and Lv-MSC, respectively). These metastasis associated-MSCs (Met-MSC) had the morphology of CFU-fibroblastic, similar to bone marrow derived MSCs (BM-MSC) and adipose-derived MSC (AD-MSC) (Figure 1C). A hallmark of MSC is their ability to differentiate into multiple mesenchymal cell types. Cell differentiation assays using specific culture conditions, demonstrated that Met-MSCs are capable of multi-lineage differentiation into bone, cartilage, and adipose similar to the control AD-MSCs and BM-MSCs (Figure 1C). Met-MSCs had a normal karyotype at different cell passages (3 to 12 passages) (Figure 1D) and were unable to form tumors when injected into the mammary fat pads of immunocompromised mice (not shown). Flow cytometry analyses revealed surface marker expression characteristic of MSCs as they lacked expression of leukocyte, hematopoietic stem cell, and endothelial progenitor cell markers (CD14, CD45, CD34, and CD133), and expressed CD105 (SH2), CD73 (SH3), CD90 (Thy-1), and CD44 (Figure 1E).

Contact with Met-MSC increases breast cancer cell growth and leads to DDR2 upregulation in breast cancer cells

Using a panel of MSCs and breast cancer cells (BCCs), we found that Met-, AD-, and BM-MSCs exhibit uniformly high levels of DDR2 protein, and low DDR1 and E-cadherin expression, which contrasted with the undetectable DDR2 and high DDR1 expression in BCCs (Figure 2A).

Figure 2. Direct contact with MSCs is sufficient to increase growth and to induce DDR2 expression in BCC.

A. Western blot with the indicated antibodies performed on a panel of MSCs (AD, BM, LN, and DsRED-LN) and on a panel of BCCs (MDA-MB-231, MCF10DCIS.COM, MCF10CA1a, MCF10CA1h, and MCF7).

B–C. A panel of GFP-BCCs as indicated was cultured with DsRED-LN-MSC (LN) in 1:1 ratio, and subjected to tumorsphere formation assays. Shown are representative images of GPF-BCC tumorspheres at 7 days. Size of spheres >50µm was quantified measuring GFP fluorescent pixels using Image J, in triplicate.

D. Tumorspheres of GFP-231 and GFP-MCF10CA1a alone and in cocultures with Ds-RED-LN and AD-MSCs were embedded in Histogel and processed for histology. Arrowheads show central core of collagen, MSC, and elongated breast cancer cells (scale bars, 100 µm).

E–F. Schematic illustrates that GFP-BCC + DsRED-LN-MSC cocultures were sorted to isolate GFP-BCCs for quantitative real-time PCR for DDR2 mRNA. The bar graph shows the ratio of DDR2 mRNA expression between co-cultures and single cultures of BCCs.

To investigate the role of Met-MSCs on the tumorigenic functions of BCCs we established co-cultures of DsRED-labeled Met-MSCs (Lv and LN) with GFP-labeled BCCs of various biological properties (BCC:MSC ratio of 1:1). BM- and AD-MSCs served as controls. While MSC conditioned media had no effect on BCC growth (data not shown), combined cultures of Met-MSCs or AD-MSCs with a panel of GFP-BCCs significantly increased growth and invasion of BCCs (Figure 2B–C, and S1B–C). The morphological details of the combined cultures were evidenced by Histogel preparations. In co-cultures, BCCs aggregated around a central core of MSCs, collagen type I, and elongated BCCs (Figure 2D, S2A–C).

Contact with MSCs induced DDR2 protein upregulation and over 5-fold mRNA overexpression in a panel of GFP-BCCs compared to GFP-BCC single cultures (Figure 2D–F and S2D). The effect of MSCs on DDR2 overexpression was restricted to cancer cells, as contact with MSCs did not affect DDR2 mRNA in non-tumorigenic MCF10A cells or primary breast cells derived from patients (Figure 2F and S2E). Together, these data demonstrate a direct interaction between MSCs and BCCs, and show that BCCs organize with MSCs and collagen type I. Our data also show that contact with MSCs induces robust DDR2 mRNA and protein upregulation in BCCs, and promotes cancer cell growth and invasion. As we validated our results using multiple MSCs, BCCs, and non-tumorigenic breast cells, our data highlight that DDR2 activation in cancer cells is a common response to contact with MSCs in the tumor microenvironment, which may have therapeutic implications.

DDR2 is essential for the MSC phenotype

Using our panel of MSCs, we found that DDR2 is phosphorylated at Tyr-740, which has been reported essential for protein activation upon collagen binding (Yang et al., 2005). Of note, AD-, BM-, and Met-MSCs displayed upregulation of phosphorylated DDR2(Tyr-740) in nuclear enriched fractions, suggesting that activation induces DDR2 localization in the nucleus (Figure 3A). To investigate the function of DDR2 in MSCs, we generated DsRED-labeled Met- and AD-MSCs with stable shRNA DDR2 knockdown and scrambled controls (Figure 3B–C, and S3A–B). Met- and AD-MSCs shDDR2 displayed downregulation of collagen type I, p-DDR2(Tyr-740), and DDR2 downstream signaling proteins p-ERK-1/2 and p-SRC compared to controls (Figure 3C and S3A–B).

Figure 3. DDR2 is essential for the mesenchymal phenotype and migration of MSCs.

A. Indicated MSCs were lysed and cytosolic (cyto) and nuclear (NE) fractions were isolated for Western blot analyses. The relative distribution of DDR2 and phosphorylated DDR2 (Tyr740) between fractions was determined using densitometry.

B. Human DsRED-LN-MSC (LN) were transduced with lentiviruses expressing shRNAs targeting DDR2 (shDDR2) or control scrambled shRNA (shSCR), and subjected to immunofluorescence to detect DDR2 and p-DDR2(Tyr740). The expression of nuclear p-DDR2(Tyr740) was quantified as white pixels in the nucleus of 50 DS-RED cells per 4 fields/slide, for each condition, using Image J (scale bars, 25 µm).

C–D. Western blot of cells in (B) was performed with the indicated antibodies.

E. Cell migration assays were performed in cells described in (B) using a high throughput microfluidic migration platform to measure the migration distance after 24 h of incubation without media replenishment. For all conditions, 4 replicates (total of 1800 channels) were performed. The Box graphs were plotted using Origin 9.0. The bottom and top of the box are the first and third quartiles, and the band inside the box is the second quartile. The ends of the whiskers represent the 5th percentile and the 95th percentile. The square inside the box indicates the mean, and the x outside the box indicates the minimum and maximum of all of the data.

F–G. Transcriptional profiling analysis of LN-MSC shDDR2 and LN shSCR demonstrates 2259 (greater than 2-fold change) significantly deregulated genes, of which 1200 were downregulated, and 1059 were upregulated. Gene ontology (GO) analysis LN-MSC shDDR2 vs LN-MSC shSCR (F). Heat map representation of microarray data highlights the expression levels of key extracellular matrix regulators, transcription factors, as well as cell adhesion genes (G).

Compared to the fibroblastic morphology of AD- and Met-MSC shSCR cells, AD- and Met-MSC shDDR2 cells were flat and epithelial-like with abundant polygonal cytoplasm, suggestive of a mesenchymal to epithelial transition (MET) (Figure 3B and S3C). These morphological changes were accompanied by downregulation of vimentin, EMT-transcription factors Slug and Zeb1, and EZH2, a regulator of differentiation, and upregulation of cytokeratin-18 and E-cadherin proteins (Figure 3D and S3B). Functionally, DDR2 knockdown significantly reduced the migratory and proliferative abilities of LN-MSC cells compared to controls (Figure 3E and S3D).

We assessed the global impact of DDR2 knockdown on MSCs by transcriptional profiling. In GO analyses, the most significantly represented biological processes between LN-MSC shDDR2 and LN-MSC shSCR were extracellular matrix organization, tissue remodeling, cell migration and adhesion, and cell proliferation (Figure 3F). Among the significantly upregulated genes in LN-MSC shDDR2 compared to LN-MSC shSCR are CDH1, CD36, and DSP which play roles in cell adhesion and stromal-epithelial crosstalk. Among the significantly downregulated genes are the EMT-transcription factors TWIST1 and TWIST2, several collagen and laminin encoding genes including COL6A1, COL6A2, COL6A3, COL4A6, LAMA4, and matrix remodeling genes LOXL1 and PLAU, the adhesion molecule VCAM1, and the epigenetic transcriptional regulator EZH2 and its target genes in mesenchymal cells TIMP3 and GDF6 (Wienken et al., 2016) (Figure 3G). Consonant with our functional observations, LN-MSC shDDR2 cells exhibited downregulation of cell proliferation genes (CDK1, CDK2, MYC, and CCNB1). Collectively, these data document a role for DDR2 in maintaining the mesenchymal phenotype and functional features of Met-MSCs. Our data also demonstrate that DDR2 knockdown induces changes in Met-MSC gene expression profiles leading to reduced collagen deposition and cell migration.

DDR2-deficient MSCs fail to deposit collagen type I and are unable to promote BCC alignment, migration, invasion, and growth

We hypothesized that MSC-derived DDR2 regulates the migratory and invasive abilities of breast cancer cells. DDR2 knockdown on Met- or AD-MSC blocked their ability to induce DDR2 expression and activation in BCCs compared to controls (Figure 4A and S4A–B). When combined with BCCs, MSC shDDR2 deposited significantly less collagen type I than MSC shSCR (Figure 4B). DDR2 knockdown in Met- and AD-MSCs blocked their migration and invasion promoting effects compared to controls (Figure 4C and S4C). We observed a similar effect upon COL1A1 knockdown in LN-MSCs (Figure 4C). Complementing these studies, pretreatment of Met- or AD-MSCs with active recombinant human DDR2 protein consisting of only the collagen binding extracellular domain was sufficient to inhibit BCC invasion (Figure S4D).

Figure 4. MSC-derived DDR2 regulates collagen type I deposition and is required for BCC alignment, migration, and invasion.

A. Immunofluorescence of GFP-231 single cultures, and cocultures of GFP-231+LN-MSC shDDR and GFP-231+LN-MSC shSCR show that DDR2 and nuclear p-DDR2(Tyr740) are upregulated in 231 cells cultured with LN shSCR, and that culture with LN-MSC shDDR2 reverses this effect. The expression of nuclear p-DDR2(Tyr740) was quantified as white nuclear pixels in 50 GFP-231 cells, 4 fields/slide, for each condition, using Image J (scale bars, 25 µm)

B. Histogel preparations of GFP-231+LN-MSC shSCR and GFP-231+LN-MSC shDDR2 stained for collagen type I and pricosirius, and quantified using Image J (scale bars, 50 µm).

C. High throughput cell migration microfluidic assay and matrigel invasion assay of GFP-231 cells in single culture, and GFP-231 cells in coculture with LN-MSC shSCR, LN-MSC shDDR2, and LN-MSC shCOL1A1. Cocultures were added to the platform to measure the migration distance of GFP-231. For the invasion assay, GFP fluorescence was quantified in 4 fields/well per condition in triplicate, using Image J and shown as mean pixels ± SEM.

D. Immunoflurescence images of 231-GFP cocultured with DsRED labeled LN-MSC shSCR, LN-MSC shDDR2, and LN-MSC shCOL1A1 (upper panel). Twenty-four hours after seeding, the indicated cocultures were subjected to live cell imaging studies. Pictures of GFP-231 cells were taken at 2h of live imaging (lower panel) (scale bars, 50 µm).

E. Bar graph shows the absolute deviation between the angles of alignment formed by at least 50 GFP-231 cells from 5 representative imaged points/slide for each condition in (D).

Live cell imaging studies and co-culture experiments of Met- or AD-MSC with GFP-231 and GFP-436 cells showed that knockdown of DDR2 or COL1A1 in MSCs completely disrupted MSC-BCC cell alignment compared to controls (Figure 4D–E, and S5A–B).

It has been reported that DDR2 upregulation in BCC sustains an EMT phenotype by stabilizing Snail1, which may have implications for metastatic dissemination (Zhang et al., 2013). However, the mechanism leading to DDR2 upregulation in breast cancer is unknown. To explore the role of stromal DDR2 in this process, GFP-231 cells primed with LN-MSC shSCR (231^{LN shSCR}), or with LN-MSC shDDR2 (231^{LN shDDR2}) for 72 hours were isolated by flow cytometry and cultured for several passages (Figure S6A). Priming GFP-231 cells with LN-MSC shDDR2 rescued the spindle morphology and EMT protein profile of GFP-231 cells primed with LN-MSC shSCR (Figure S6B–D). Demonstrating the functional relevance of these findings, 231^{LN shDDR2} displayed reduced migration compared to 231^{LN shSCR} (Figure S6E). We also observed sustained changes in the gene expression profiles of 231^{LN shDDR2} and 231^{LN shSCR} cells leading to greater than 2-fold upregulation of 178 genes and downregulation of 112 genes, with overrepresentation of genes implicated in cell communication, cell proliferation, cytokine production, collagen metabolic process (Figure S6F). Together, these data demonstrate a role for mesenchymal stromal DDR2 expression as a regulator of collagen type I deposition and DDR2 activation in breast cancer cells. Our data also show that DDR2 expression in MSCs regulates breast cancer cell migration and EMT programs, and suggest that contact with DDR2-expressing MSCs induces sustained changes in the gene expression profiles of breast cancer cells.

Using a combination of complementary and independent assays, we found that DDR2 knockdown in Met- and AD-MSCs blocked their growth and proliferation effect on GFP-BCCs compared to controls (Figure S7A–B). A similar effect was observed upon COL1A1 knockdown (Figure S7A–B), and collagen I rescued the reduced growth of GFP-231 cells cultured with LN-MSC shDDR2 or LN-MSC shCOL1A1 compared to controls (Figure S7C).

MSC-derived DDR2 regulates breast cancer growth and metastasis

Combined cultures of GFP-231+LN-MSC formed significantly larger orthotopic primary tumors with increased metastasis compared to GFP-231 alone, and DDR2 knockdown in MSCs rescued the growth and metastatic advantage induced by MSCs on BCCs (Figure 5A–E). DDR2 knockdown in LN-MSC reduced the size of GFP-231 lung metastatic foci compared to controls, and the number of circulating GFP-231 in peripheral blood (Figure 5B, E, G).

Figure 5. DDR2-deficient MSCs are unable to enhance breast cancer growth and metastasis in vivo.

A. Orthotopic tumor formation assay. Representative macroscopic images of mammary tumors formed by GFP-231, GFP-231+DsRED-LN-MSC shSCR, and GFP-231+DsRED-LN-MSC shDDR2 cells injected the mammary fat pads of NOD/SCID mice (n=10 mice per group). Mammary tumors were stained with H&E, pricosirius, and p-DDR2 (scale bars, 50 µm).

B. Representative images of the lungs of mice in (A) showing GFP- positive and DsRED-positive foci (scale bars, 50 µm). Representative pictures of spontaneous distant metastasis stained with pricosirius and the MSC marker TM4SF1 (scale bars, 50 µm).

C. Growth curve of primary tumors as determined by caliper measurements at the indicated timepoints. Mean ±SEM.

D. Bar graph shows the percentage of mice with extrapulmonary distant metastasis.

E. Quantification of the mean size of GFP-positive lung foci ±SEM (n=10 mice per group).

F. Graph shows quantification of pricosirius staining in the metastasis using Image J. Mean ±SEM of 5, 100x fields counted per condition.

G. Analysis of circulating tumor cells. Left, Standard curve: GFP expression from peripheral blood (0.5ml) from tumor free NOD/SCID mice mixed with 0, 100, 500, 1000 and 2000 231-GFP cells. Right, number of tumor cells from peripheral blood (0.5 ml) of tumor bearing mice was determined using the standard curve. Graph shows # of CTCs x 10² ± SEM, n=6–10 mice per group.

Knockdown of DDR2 in MSCs led to reduced collagen I deposition in primary tumors and metastasis, and decreased the expression of p-DDR2(Tyr740) in the tumor cells compared to controls (Figure 5A, B, F, and S8A, B). Second harmonic generation showed that tumors of GFP-231+LN-MSC shSCR mice had long and aligned fibrils compared to the short and dispersed fibrils of GFP-231+MSC-shDDR2 tumors (Figure S8D–E).

Similar to our observations in human metastasis, GFP-231 lung metastasis contained MSCs detected using the tetraspanin family protein TM4SF1 (Bae et al., 2011) (Figure 5B). We noted the presence of DsRED- LN-MSC in the same lung metastatic foci of GFP-231 cells, independent of DDR2 expression, strongly suggesting that MSCs migrate from the primary tumor to the distant metastatic site (Figure 5B and Figure S8C).

To directly test the relevance of DDR2 expression in the microenvironment on spontaneous breast cancer metastasis in an immunocompetent model, we employed Smallie (slie) mice which carry a ∼150 kb spontaneous autosomal recessive mutation that removes ddr2 (Kano et al., 2008). Mice homozygous for the slie mutation are dwarf in contrast to slie heterozygous and wild type mice (Kano et al., 2008; Labrador et al., 2001). Syngeneic GPF-E0177 mouse mammary carcinomas cells were injected in the mammary fat pads of slie/slie, slie/wt, and wt/wt mice. To investigate the presence of lung metastasis independent from primary tumor growth, we euthanized the mice before the development of palpable mammary tumors. slie homozygous mice developed significantly fewer lung metastases compared to slie heterozygous and wild-type mice (Figure 6A–B). Together, these data document an essential paracrine role for metastasis microenvironment-derived DDR2 in the development and growth of breast cancer metastasis. Our working model is shown in Figure 6C.

Figure 6. DDR2 ablation in the microenvironment reduces breast cancer metastasis.

(A) E0177-GFP mouse breast cancer cells (5×10⁵) were orthotopically injected in the mammary fat pads of slie/slie, slie/wt, or wt/wt mice (10 mice per group). Upper panel: GFP-positive lung foci at 21 days after orthotopic injections. Lower panel: lung sections were stained with H&E to verify the presence of metastatic carcinoma, (scale bars, 50 µm).

(B) The number of lung metastasis per mouse was quantified by measuring GFP- positive foci, n=10 mice per group.

(C) Working model of DDR2 function in breast cancer progression.

DISCUSSION

MSCs derived from the bone marrow and adipose tissue have been shown to promote the growth and metastatic ability of breast cancer and other human malignancies, but the mechanisms are incompletely understood (Barcellos-Hoff et al., 2013; Cuiffo et al., 2014; Del Pozo Martin et al., 2015; Li et al., 2012; Liu et al., 2011). Here, we discovered that the collagen receptor DDR2 is upregulated in breast cancer metastasis-associated MSCs and in metastatic breast cancer cells in clinical samples. Our study shows that the collagen receptor DDR2 mediates MSC-cancer cell cross talk to enhance breast cancer proliferation, migration, and metastasis at least in part through induction of collagen type I deposition and activation of DDR2 signaling in breast cancer.

While studies have focused on the role of the primary tumor stroma in cancer progression (Barcellos-Hoff et al., 2013; Cuiffo et al., 2014; Karnoub et al., 2007; Li et al., 2012; Liu et al., 2011), the functions of the metastasis-associated stroma are largely unknown in part due to the lack of physiologically relevant models. MSCs have been shown to migrate from the bone marrow to the primary and metastatic tumor site in mice, but have only been recently found in human gastric cancer metastasis (Liu et al., 2011; Zhou et al., 2016). We have successfully isolated MSCs from breast cancer metastases, providing evidence that MSCs are present in the breast cancer metastatic niche in humans. Our study directly demonstrates that MSCs migrate with metastatic breast cancer cells to establish the distant metastases. These data advance the understanding of metastasis development and may have implications in the design of anti-metastatic strategies.

We found that metastasis-associated MSCs have a normal karyotype, similar surface marker expression than bone marrow- and adipose-derived MSCs, and exhibit DDR2 pathway activation. DDR2 is a tyrosine kinase receptor expressed in mesenchymal cells, uniquely activated by fibrillar collagens which are main components of the ECM (Alves et al., 1995; Fu et al., 2013; Ikeda et al., 2002; Valiathan et al., 2012). DDR2 participates in ECM remodeling during morphogenesis and tissue repair, as well as differentiation and proliferation (Marquez and Olaso, 2014; Olaso et al., 2002). The profound effect of DDR2 in mesenchymal cell and matrix homeostasis is underscored by the finding that mice homozygous for the slie mutation, which removes ddr2, exhibit dwarfism (Kano et al., 2008; Labrador et al., 2001). Recently, DDR2 was identified as one of 14 genes differentially expressed in bone marrow derived MSCs compared to hematopoietic stem cells (Anam and Davis, 2013). However, the function of DDR2 in MSCs has remained unexplored. In this study, we show that DDR2 is a critical regulator of MSC phenotype, collagen I deposition function, and migration capacity.

Several studies have demonstrated that DDR2 overexpression and signaling activation in breast cancer cells promotes tumor progression, sparking interest in DDR2 tyrosine kinase as a target for the treatment of breast and other malignancies (Badiola et al., 2011; Kim et al., 2015; Valiathan et al., 2012; Zhang et al., 2013). While no DDR2 mutations were identified in breast cancer, our lab and other investigators found that DDR2 is overexpressed in over 50% of invasive breast carcinomas compared to none of the normal breast epithelium. In breast cancer samples, DDR2 overexpression in the cancer cells is associated with high collagen in trichrome stains, and worse patient survival (Toy et al., 2015; Zhang et al., 2013). Of note, DDR2 is also expressed in the stromal cells of the breast cancer microenvironment (Toy et al., 2015; Zhang et al., 2013). A recent study demonstrated that in mice, DDR2 expression in cancer cells and in cells of the host tumor microenvironment including cancer associated fibroblasts, is critical for breast cancer metastasis in the MMTV-PyMT model (Corsa et al., 2016). However, the mechanisms of DDR2 function in the tumor microenvironment are unclear, and the relationship between DDR2 in stromal and cancer cells has not been considered.

We show that direct contact with MSCs is necessary and sufficient to induce DDR2 upregulation in breast cancer cells. DDR2 expression in MSC is required for collagen deposition and leads to increased DDR2 expression and activation in breast cancer cells. This paracrine-autocrine MSC-cancer cell axis results in breast cancer alignment with collagen fibers facilitating migration, invasion, and metastasis. Providing strong evidence for a critical function of DDR2 in metastasis, homozygous slie mice form fewer and smaller breast cancer metastasis that heterozygous and wild type mice.

The ability to upregulate DDR2 in response to MSC-initiated signals appears to be a specific property of cancer cells but not of benign breast epithelial cells. LN-, BM-, or AD-MSCs were unable to induce DDR2 expression, proliferation, migration, or invasion (not shown) in nontumorigenic MCF10A cells or patient-derived primary breast epithelial cells, in contrast to the robust DDR2 upregulation induced in all breast cancer cells tested. These data are in line with our previous report that DDR2 is not expressed in normal breast epithelium from patient’s samples and only overexpressed in cancer (Toy et al., 2015). These results are also in agreement with a study showing that DDR2 upregulation in breast cancer cells does not initiate but is induced during EMT and participates in collagen type I-mediated stabilization of Snail1, promoting breast cancer cell invasion and metastasis (Zhang et al., 2013). Collectively, these data led us to postulate that MSC-derived DDR2 in the primary tumor endows cancer cells that already have initiated the metastatic process with growth and migratory advantage through alignment with collagen. Contact with fibrillar collagen induces DDR2 upregulation and signaling activation in breast cancer cells which stabilizes EMT transcription factors and contributes to metastatic growth and dissemination.

There is substantial evidence that breast cancer cells are highly responsive to signals from MSCs through cytokine networks resulting in increased tumor initiating cells (TICs), tumor growth, metastasis, and MSC homing to the primary tumor site (Li et al., 2012; Liu et al., 2011). Li et al (Li et al., 2012) reported that MSCs induce IL-1 dependent tumor initiation but this effect was not observed on MDA-MB-231 and −453 cells, neither of which secretes IL-1 (Li et al., 2012). Our tumor initiation studies (data not shown) support these observations. Here, we demonstrate that MSCs induce breast cancer growth and metastasis that requires direct cell-cell contact and deposition of insoluble collagen I fibrils and is independent of their effect on TICs. These mechanisms may be interdependent, as contact between MSCs and breast cancer cells was shown to induce activation of cytokines conducive to metastatic dissemination (Karnoub et al., 2007). The cross talk between MSC-derived DDR2 and the cytokine networks in the tumor microenvironment warrants further investigation.

Before these observations, the effects of DDR2 in the metastatic niche on breast cancer progression were largely unknown. Our data suggest that DDR2 conveys important autocrine and paracrine signals to breast cancer cells. Based on our study, targeting DDR2-mediated signaling at the metastatic site may offer a therapeutic opportunity to interrupt breast cancer metastasis growth and invasiveness in distant organs.

EXPERIMENTAL PROCEDURES

Cell lines and cell culture

The breast cancer cell lines MDA-MB-231, MCF7, and MDA-MB-436 were obtained from the American Type Culture Collection. MCF10AT progression lines were obtained and cultured as previously reported (Kleer, 2004). The E0771 cell line was derived from a spontaneous mammary tumor in a C57BL/6 mouse and was purchased from CH3 BioSystems (catalog #940001). Human adipose and bone marrow cells were purchased from ScienCell Research Laboratories and maintained following the provider’s instructions. Cells were delivered frozen after being isolated from normal human tissue and being cryopreserved at first passage.

To isolate Met-MSCs, 5 fresh tissue samples from breast cancer metastasis to lymph nodes (n=4) and liver (n=1) were obtained through the Tissue Procurement Service at the University of Michigan (IRB#HUM00050330). The samples were immediately processed in the laboratory. A portion was formalin fixed and paraffin embedded for staining with hematoxylin and eosin (H&E). Another portion was processed to a single cell suspension, by manually mincing the tumor followed by dissociation in a collagenase-hyaluronidase solution for 7 hours at 37°C (StemCell Technologies, #07912). Samples were centrifuged at 350 x g for 5 minutes following the instructions from StemCell Technology to obtain the fibroblast enriched population. Red blood cells were lysed with RBC Lysis Solution (Qiagen, #158902). For further purification, the tissue was treated with Trypsin-EDTA (GIBCO, #25200-056), then DNase1 (StemCell Technologies, #07900) and finally filtered through a 40µm cell strainer. The purified adherent cells exhibited a uniform fibroblastic morphology (Figure 1C) and were cultured in Mammary Epithelial Cell Medium (ScienCell, #7611) completed with Mammary Epithelial Cell Growth Supplement and penicillin-streptomycin (ScienCell, #7652 and #0503). For all the experiments MSCs were passaged 20 times or less in Mesenchymal Stem Cell Medium (ScienCell, #7501). Similar protocols were described to isolate MSCs from human primary ovarian tumors (McLean et al., 2011).

Vectors and Viral Infections

DDR2 and COL1A1 knockdown using stable short-hairpin interfering RNAs in lentivirus were completed as previously reported (Gonzalez et al., 2009). For targeting DDR2 (NM_006182 NCBI) and COL1A1 (NM_000088 NCBI), we used the following shRNA oligoes: for DDR2, ID# NM_006182.x-1694s1c1 and NM_006182.x-834s1c1, for COL1A1 ID#NM_000088.2–795s1c1, corresponding to #TRCN0000001419, #TRCN0000001418 and TRCN0000062558 from SIGMA-ALDRICH, respectively. These oligos were cloned into a pLKO.1-Puro vector and packaged into lentiviral particles at the University of Michigan Vector Core. A lentivirus containing a plasmid encoding a scrambled shRNA oligo was used for control. Cells were transduced and selected for antibiotic resistance with puromycin (Sigma-Aldrich, #P9620).

Differentiation assays

Mesenchymal cell differentiation assays were performed following published protocols (McLean et al., 2011). Briefly, for bone differentiation, cells were plated at 5×10⁴ cells/well of a 6-well plate in either StemPro Osteogenesis Differentiation Media (Invitrogen) or control media and allowed to grow for 14 days, with media changed twice per week. Cells were then rinsed with PBS, fixed with 3.7% formaldehyde, rinsed with water, stained with 2% Alizarin Red S (Sigma-Aldrich) solution (pH 4.2) for 2 minutes, and washed with distilled water. For cartilage differentiation, cells were plated in 5 to 10µl micromass droplets onto a dry plate from a solution of 1.6 × 10⁷ cells/ml, allowed to set for 2 hours, and then incubated in either StemPro Chondrogenesis Differentiation Media (Invitrogen) or control media. Cells were allowed to grow for approximately 14 days with media changed twice per week and then rinsed with PBS, fixed with 4% formaldehyde, rinsed with PBS, stained for 30 minutes with 1% Alcian Blue solution (Sigma-Aldrich) prepared in 0.1 N HCl, and washed with 0.1 N HCl. For adipose differentiation, cells were plated at 5 × 10⁴ cells/well of a 6-well plate, in either differentiation media (StemCell Technologies) or control media, and the media was changed weekly. After 14 to 21 days, cells were fixed with 3.7% formaldehyde, stained with 0.3% Oil Red O (Fisher Scientific) for 1 hour, and washed with water.

Immunoblots

Western blot analyses were carried out with 100µg of whole cell extract derived as previously reported (Gonzalez et al., 2011). Membranes were blocked and incubated with primary antibodies in 4% milk (Sigma-Aldrich, #A3059) in TBS-T (Bio-Rad, #161–0372, with 0.05%) Tween 20) at 4°C overnight. Nuclear and cytoplasmic enriched fractions were isolated as reported (Gonzalez et al., 2009). Mouse monoclonal β-Actin-HRP (Santa Cruz, #47778), anti-GAPDH (Abcam, #ab9484), and rabbit polyclonal anti-Histone H3 (Cell Signaling, #9715) were used to confirm equal loading. Primary antibodies from Cell Signaling included anti-EZH2 (#5246), anti-DDR2 (#12133), anti-DDR1 (#5583), anti-phospho-FAK (Y925) (#3284), anti-FAK (#13009), anti-E-cadherin (#3195), anti-phospho-p44/42 MAPK (ERK1/2) (Thr202/Tyr204) (#4370), anti-p44/42 MAPK (ERK1/2) (#4695), anti-phospho-SRC Family (Tyr416) (#21010, anti-Src (#2109), anti-SLUG (#9585) and anti-Vimentin (#5741); from Abcam anti-Collagen I (#AB34710) and anti-CK18 (#AB189444); from R&D Systems phospho-DDR2 (Y740) (#MAB25382); and from Santa Cruz Biotechnology anti-ZEB1 (#SC-81428).

Hanging drop assay

As previously described (Hsiao et al., 2012), hanging drop plates from 3D Biomatrix (Perfecta3D, HDP1384-8) were coated in 0.1% Pluronic F108 (BASF Co., Ludwigshafen, Germany) solution for 1 h and UV sterilized for 30 min. The single BCCs and co-culture with MSC groups were prepared at 160 cells/µL to obtain a 5000 cell spheroid per hanging drop. All hanging drop plates were sandwiched between a 96-well plate lid and plate filled with distilled water and maintained in a humidified incubator (37°C, 5% CO₂) for 8 days. Spheroids were monitored by phase contrast and fluorescence microscopy (Nikon TE-300), and co-culture media was exchanged every other day using a liquid handling robot, which removes 8 µL and adds back 10 µL to accommodate routine droplet evaporation, for 8 days.

High Throughput Microfluidic Cell Migration

The cell migration assays were performed using a microfluidic migration platform (Burgos-Ojeda et al., 2015; Chen et al., 2015). To achieve higher throughput, the design was modified to have 450 migration channels per device, and the migration channel is designed to be 5 µm in height and 30 µm in width. Before cell loading, PBS (Gibco 10082) was used to prime the device for one hour, and the cell culture media flowed through the channel for one hour for better cell adhesion and viability. Cells were trypsinized, centrifuged, and re-suspended to a concentration of 4×10⁵ cells/ml for loading into the device. For on-chip co-culture of MDA-MB-231 and Met-MSCs, 1:1 mixture was used in cell loading. After cell loading, the cell suspension was replaced by cell culture media. The microfluidic chip was put into an incubator, and migration distance was measured based on the final cell position after 24 h of incubation without media replenishment. The images were analyzed by MATLAB code automatically. Cells were identified based on their fluorescence, and debris was ignored by their small size. For all conditions in this work, 4 replicates (totally 1800 channels) were performed. The Box graphs were plotted using Origin 9.0. The bottom and top of the box are the first and third quartiles, and the band inside the box is always the second quartile (the median). The ends of the whiskers represent the 5th percentile and the 95th percentile. The square inside the box indicates the mean, and the x outside the box indicates the minimum and maximum of all of the data.

Animal Studies

All procedures were conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee at the University of Michigan (UCUCA#PRO 00005009). For tumorigenicity experiments, ten-week-old severe combined immunodeficient mice (Jackson Laboratories) were used. GFP-MDA-MB-231 cells alone or in combination with equal amounts of DsRED-LN-MSC shSRC or DsRED-LN-MSC-shDDR2 groups were orthotopically injected into the mammary fat pads at a concentration of 1×10⁶ cells in 30 mice (n=10 per group). Tumor size was measured twice a week until tumors reached 2 cm³ (tumor volume= (length x width²)/2), at which time mice were sacrificed.

Smallie (slie) mice, which contain a spontaneous deletion in Ddr2, were initially obtained from the Jackson Laboratory on a C57BLS/J (BKS) background (BKS(HRS-Ddr2^slie/JngJ). Mice were bred with C57BL/J6 (B6) mice for at least 10 generations. Unlike mice on the BKS background, which are sterile when homozygous (slie/slie), homozygotes on the B6 background bred normally with normal litter sizes. Female and male mice were genotyped using previously defined PCR primers and maintained on a normal chow diet until sacrifice at 3 or 5 months for analysis of skeletal and marrow fat phenotypes. Slie mice were dwarf. All animal studies were approved by the University of Michigan Committee on the Use and Care of Animals (UCUCA) and conformed to all guidelines and regulations for the protection of animal subjects. Mice were housed in specific pathogen-free AAALAC-certified facilities. After genotyping, 10 week-old slie/slie, slie/wt, and wt/wt were used to study spontaneous metastasis. Syngeneic 5×10⁵ GPF-E0177 cells were orthotopically injected into the mammary fat pads of slie/slie, slie/wt, and wt/wt (n=10 per group). Mice were sacrificed after 21 days of the injections before palpable tumors appeared. Lung metastases were identified by GFP fluorescence microscopy right after collecting the tissues at necropsy. The number and size of the metastasis per mice per group was quantified using Image J. The presence of metastases was confirmed by histology.

Statistical Analysis

Results are reported as mean ± SD or mean ± SEM unless otherwise noted. Comparisons between two groups were performed using an unpaired two-sided Student’s t test (p < 0.05 was considered significant).

In Brief.

By isolating human mesenchymal stem cells from breast cancer metastasis, Gonzalez et al. identify a pathway initiated by stromal DDR2, a unique receptor tyrosine kinase activated by fibrillary collagen, that mediates DDR2 activation in breast cancer cells and induces metastasis.

Highlights.

MSCs are present in human breast cancer metastatic microenvironment

DDR2 regulates MSC phenotype, collagen deposition function, and migration

DDR2 activation in breast cancer depends on stroma-derived DDR2

Inhibition of DDR2 in the tumor microenvironment reduced metastasis