ANGPTL2 increases bone metastasis of breast cancer cells through enhancing CXCR4 signaling

Bone metastasis of breast cancer cells is a major concern, as it causes increased morbidity and mortality in patients. Bone tissue-derived CXCL12 preferentially recruits breast cancer cells expressing CXCR4 to bone metastatic sites. Thus, understanding how CXCR4 expression is regulated in breast cancer cells could suggest approaches to decrease bone metastasis of breast tumor cells. Here, we show that tumor cell-derived angiopoietin-like protein 2 (ANGPTL2) increases responsiveness of breast cancer cells to CXCL12 by promoting up-regulation of CXCR4 in those cells. In addition, we used a xenograft mouse model established by intracardiac injection of tumor cells to show that ANGPTL2 knockdown in breast cancer cells attenuates tumor cell responsiveness to CXCL12 by decreasing CXCR4 expression in those cells, thereby decreasing bone metastasis. Finally, we found that ANGPTL2 and CXCR4 expression levels within primary tumor tissues from breast cancer patients are positively correlated. We conclude that tumor cell-derived ANGPTL2 may increase bone metastasis by enhancing breast tumor cell responsiveness to CXCL12 signaling through up-regulation of tumor cell CXCR4 expression. These findings may suggest novel therapeutic approaches to treat metastatic breast cancer.

B reast cancer is the most common cancer type in women, and bone is the most common first site of metastasis in that cancer [1][2][3] . About 83% of patients with advanced breast cancer will develop bone metastases during the course of their disease 4 . The skeletal consequences of metastasis include pain, pathologic fractures, spinal cord and other nerve-compression syndromes, and life-threatening hypercalcemia, all of which cause increased morbidity and mortality 5 . Therefore, it is important to define mechanisms underlying bone metastasis of breast cancer cells.
The ligand of the CXCR4 chemokine receptor is the CXC chemokine stromal cell-derived factor 1 (SDF-1), also known as CXCL12 6 . Binding of CXCL12 to CXCR4 activates intracellular signaling associated with chemotaxis and cell survival 7 and also functions in tumorigenesis and progression of various cancer subtypes 8,9 . CXCL12activated CXCR4 signaling reportedly activates several signaling pathways, such as phosphatidylinositol 3-kinase (PI3K)/Akt and mitogen-activated protein kinase (MAPK), in various cell lines 7 and regulates expression of matrix metalloproteinases (MMPs), which promote destruction of the extracellular matrix and are critical for metastasis 10,11 . ERK signaling induces MMP-13 12,13 , which cleaves collagen type I, which constitutes approximately 95% of bone collagen 14 . We previously reported that ANGPTL2 increases MMP expression and activity in osteosarcoma cells 15 . In breast cancer pathology, CXCL12 derived from various tissues, including bone tissue, preferentially recruits cancer cells expressing CXCR4 and promotes their metastasis to those tissues 16,17 , suggest-ing that CXCR4 suppression in breast cancer cells might be a strategy to decrease bone metastasis. However, molecular mechanisms underlying CXCR4 expression in tumor cells have not been fully clarified.
Angiopoietin-like proteins (ANGPTLs), which possess an N-terminal coiled-coil domain used for oligomerization and a C-terminal fibrinogen-like domain, are structurally similar to Tie-2 receptor ligands known as angiopoietins 18 . However, ANGPTLs do not bind to Tie2 or to its homologue Tie1 and thus function differently from angiopoietins 18 . ANGPTL2 is secreted primarily by adipose tissue in normal conditions 19 . We recently identified ANGPTL2 as a key mediator of chronic inflammation and associated diseases, such as obesity-related metabolic syndrome 19 , cardiovascular disease 20,21 , some autoimmune diseases 22,23 , carcinogenesis 24,25 and tumor metastasis 15,26 . We also demonstrated that suppression of breast cancer cell-derived ANGPTL2 attenuated breast cancer metastasis to lung tissue in vivo using xenograft models created by implanting MDA-MB231 breast cancer cells into the mouse mammary fat pad 26 . We also found that serum ANGPTL2 levels in patients with metastatic breast cancer were significantly higher than those in patients with non-metastatic invasive ductal carcinoma 27 , suggesting that ANGPTL2 promotes breast cancer cell metastasis.
In the present study, we performed RNA sequence analysis of MDA-MB231 cells harboring ANGPTL2 knockdown (MB231/ miANGPTL2) and found that, relative to control (MB231/miLacZ) cells, CXCR4 expression significantly decreased, suggesting that ANGPTL2 contributes to CXCR4 expression in breast tumor cells.
In vitro experiments revealed that MB231/miANGPTL2 attenuates breast tumor cell responsiveness to CXCL12 stimulation by decreasing CXCR4 expression in those cells. We also found that ETS1dependent transcription was important for ANGPTL2-induced CXCR4 expression and that ANGPTL2 increased breast tumor cell invasiveness by activating ERK and MMP-13 expression. Using a xenograft mouse model established by intracardiac injection of tumor cells, we found that mice injected with MB231/ miANGPTL2 cells showed significantly decreased bone metastasis and prolonged survival relative to controls. Finally, we observed a positive correlation of ANGPTL2 and CXCR4 expression in primary tumor tissues from breast cancer patients. These findings suggest that tumor cell-derived ANGPTL2 may increase bone metastasis by enhancing breast tumor cell responsiveness to CXCL12 signaling through up-regulation of tumor cell CXCR4 expression.

Results
ANGPTL2 suppression in MDA-MB231 cells attenuates CXCL12activated CXCR4 signaling and expression. Our previous findings in an orthotopic implantation model showed that ANGPTL2 knockdown in breast cancer cells reduces metastasis to distant tissues, such as lung 26 . To determine what factors downstream of ANGPTL2 might promote metastasis, we compared transcripts in ANGPTL2 knockdown human breast tumor MB231 cells to those in control MB231 cells using an RNA sequencing strategy. To do so, we generated both MB231/miANGPTL2 and control LacZ knockdown (MB231/miLacZ) lines using the Invitrogen BLOCK-iT miR RNAi system 15 , as previously reported 26 . We observed twenty transcripts whose expression was markedly altered (10 upregulated and 10 downregulated) in MB231/miAngptl2 compared to control cells (Fig. 1A, Supplementary Fig. S1). Among them, we focused on the chemokine receptor C-X-C chemokine receptor type 4 (CXCR4), whose expression was decreased in the MB231/miANGPTL2 line. CXCR4 was of particular interest because breast cancer cells expressing CXCR4 reportedly move to secondary sites in lung, liver, bone marrow or lymph nodes where the CXCR4 ligand, the chemokine CXCL12, is produced in at high levels 16,28 . In addition, silencing of CXCR4 in breast cancer cells reportedly reduces metastasis in a mouse xenograft model 29 .
Western blot analysis confirmed that CXCR4 protein levels were significantly lower in MB231/miANGPTL2 compared to MB231/ miLacZ cells ( Fig. 1B and 1C, Supplementary Fig. S2A and S2B). When we assessed cell surface CXCR4 expression by FACS analysis, we found that MB231/miANGPTL2 cells expressed low CXCR4 levels compared to control MB231/miLacZ cells (Fig. 1D). By contrast, CXCL12 expression in both lines was comparable ( Supplementary Fig. S2C).
To compare motility of MB231/miANGPTL2 and MB231/ miLacZ cells, we used a trans-well assay in the presence or absence of exogenous CXCL12. In the absence of CXCL12, MB231/ miANGPTL2 cells showed lower motility than did control MB231/ miLacZ cells (Fig. 1E). In addition, MB231/miANGPTL2 cells treated with CXCL12 showed lower motility than did similarlytreated control cells (Fig. 1F, Supplementary Fig. S3), suggesting that ANGPTL2 loss in these cells decreases their responsiveness to CXCL12, most likely due to downregulation of CXCR4.
Next we used quantitative real-time PCR to compare the effect of CXCL12 on MMP-13 expression in MB231/miANGPTL2 and control MB231/miLacZ cells. In the absence of CXCL12 treatment, MMP-13 expression was lower in MB231/miANGPTL2 compared to MB231/miLacZ cells (Fig. 2C). Treatment of cells with CXCL12 for 8 hours increased MMP-13 expression in MB231/miLacZ controls but not in MB231/miANGPTL2 cells (Fig. 2D). Western blot analysis of CXCL12-treated cells revealed transient ERK1/2 activation in MB231/miLacZ cells but not in MB231/miANGPTL2 cells (Fig. 2E, Supplementary Fig. S5 and S6), suggesting that ERK1/2 activation underlies induction of MMP-13 promoted by CXCL12. To further assess this possibility, we undertook invasion assays of control MB231/miLacZ cells treated with various combinations of CXCL12 or the MKK inhibitor U0126. Quantitative analysis of a given invasion area revealed that the area occupied by invasive MB231/miLacZ cells seen following CXCL12 treatment significantly decreased when the assay was conducted in the presence of U0126 ( Fig. 2F and G). These findings suggest that ANGPTL2 knockdown in MDA-MB231 cells antagonizes CXCL12-dependent ERK1/2 activation and reduces MMP-13-induction required for tumor invasive activity.
Next we examined cell surface CXCR4 expression in MB231/ ANGPTL2 cells after ETS1 knockdown using ETS1 siRNA ( Fig. 3G and 3H, Supplementary Fig. S8). FACS analysis revealed reduced cell surface CXCR4 expression in ETS1 knockdown versus control cells (Fig. 3I), suggesting a link between ANGPTL2-induced CXCR4 expression and ETS1 expression.    To investigate whether ANGPTL2 increases CXCR4 expression in other breast cancer cells, we generated T47-D lines that constitutively express ANGPTL2 (T47-D/ANGPTL2) and control vector-only lines (T47-D/Control) ( Supplementary Fig. S9A). As anticipated, we found that CXCR4 and ETS1 mRNA expression significantly increased in T47-D/ANGPTL2 compared to T47-D/Control cells ( Supplementary Fig. S9B and S9C). FACS analysis indicated that T47-D/ANGPTL2 cells expressed higher levels of cell surface CXCR4 than did T47-D/Control cells (Supplementary Fig. S9D). These results suggest that ANGPTL2 may enhance CXCR4 expression in breast cancer cells by increasing ETS1 expression.
Suppression of tumor cell-derived ANGPTL2 decreases bone metastasis in an intracardiac inoculation model. Since breast cancer cells expressing CXCR4 tend to move toward regions of high CXCL12 expression 16,28 , we asked whether ANGPTL2 knockdown and concomitant decreases in CXCR4 expression would alter the metastatic potential of MDA-MB231 cells. To do so, we established ANGPTL2 knockdown and control cells harboring a luciferase expression vector (MB231/miANGPTL2/luc and MB231/miLacZ/luc, respectively), as we previously reported 26 . Cellular ANGPTL2 protein levels as well as ANGPTL2 levels in the culture medium were significantly decreased in knockdown compared to control cell lines ( Fig. 4A-4C, Supplementary Fig.  S10). We observed no differences in proliferation in these lines under normoxia or hypoxia in vitro ( Supplementary Fig. S11A). Knockdown and control lines also exhibited no difference in primary tumor size in an orthotopic in vivo implantation model ( Supplementary Fig. S11B). To investigate whether tumor cellderived ANGPTL2 enhances the ability of cells to colonize a metastatic site, we injected MB231/miANGPTL2/luc or MB231/ miLacZ/luc cells into the left cardiac ventricle of immunodeficient mice and assessed metastasis. Three weeks after injection of control MB231/miLacZ/luc cells, we detected bone metastasis. However, bone metastasis was detected one week later (at four weeks) in MB231/miANGPTL2/luc-injected mice (Fig. 4D). At 5 weeks after treatment, metastatic activity was greater in MB231/miLacZcompared to MB231/miANGPTL2/luc-injected mice. When we observed metastatic lesions using H&E staining, we observed significantly decreased bone colonization in MB231/miANGPTL2/ luc-compared to MB231/miLacZ/luc-injected mice: the bonemarrow cavity of MB231/miLacZ/luc-injected mice was almost filled with tumor cells by four weeks after injection. (Fig. 4E, Supplementary Fig. S12). Finally, survival of MB231/miANGPTL2injected mice was prolonged relative to MB231/miLacZ-injected mice (Fig. 4F).
ANGPTL2 expression in primary tumor tissues correlates with CXCR4 expression, lymph node metastasis and breast cancer progression. Next, we analyzed a potential relationship between ANGPTL2 and CXCR4 expression in primary tumors from 181 breast cancer patients (Fig. 5A). Patients were divided into two groups based on percentage of ANGPTL2-positive tumor cells at the primary tumor site: the positive group was defined as showing $50% ANGPTL2-positive tumor cells, while the negative group showed ,50%. This analysis showed that ANGPTL2 expression was positively correlated with high CXCR4 expression in specimens (Fig. 5B). ANGPTL2 expression at the primary tumor site was also significantly correlated with tumor size (T-factor), positivity for lymph node metastasis, and severity of breast cancer stage (P , 0.001; Supplementary Table 2). CXCR4 expression at the primary tumor site was correlated only with positivity for lymph node metastasis (P , 0.01; Supplementary Table 3). When we asked whether ANGPTL2 or CXCR4 levels correlated with patient outcome, we found that patients in the ANGPTL2-positive group showed a shortened period of distant relapse-free survival (DRFS) after surgery relative to the negative group ( Supplementary Fig.  S17A). Patients in the CXCR4-positive and -negative groups showed comparable DRFS periods ( Supplementary Fig. S17B). However, patients in the group positive for both ANGPTL2 and CXCR4 showed a shortened DRFS period compared to the other patients group (Fig. 5C). Next, we analyzed the relationship among ANGPTL2, CXCR4 and MMP-13 expression in primary tumors. Both ANGPTL2 and CXCR4 expression levels were positively correlated with high MMP-13 expression in breast cancer patient specimens ( Supplementary Fig. S15 and S16). Patients in MMP-13 positive and -negative groups showed comparable DRFS periods ( Supplementary Fig. S17C). Overall, these findings suggest that ANGPTL2 may promote breast cancer progression, possibly by activating CXCR4 and MMP-13.
We next examined how ANGPTL2 might induce CXCR4. We previously reported that ANGPTL2 binds to the integrin a5b1 receptor, and others have reported that ANGPTL2 binds to leukocyte Ig-like receptor B2 (LILRB2) to promote hematopoietic stem cell expansion 37 . Thus, we asked whether MDA-MB231 cells express either candidate receptor. Flow cytometry analysis showed that LILRB2 was not expressed in MDA-MB231 cells ( Supplementary  Fig. S18A), while integrin a5b1 was ( Supplementary Fig. S18B). When we asked whether treatment of MDA-MB231 cells with a neutralizing antibody for integrin a5b1 could block CXCR4 induction, we observed partial but significant suppression of CXCR4 induction by that antibody (Supplementary Fig. S18C).

Discussion
Here we demonstrate that breast cancer cell-derived ANGPTL2 may play an important role in metastasis of breast cancer cells to bone. Several lines of evidence suggest this possibility. First, we found that ANGPTL2 enhances both CXCR4 and ETS1 expression, suggesting that these outcomes are linked. Our findings also suggest that ANGPTL2 increases bone metastasis by up-regulating CXCR4 in primary tumor cells, thereby enhancing their responsiveness to bone tissue-secreted CXCL12 (Fig. 6A). Furthermore, in bone tissue, CXCL12-activated CXCR4 signaling promoted by ANGPTL2 accelerated osteolysis and bone engraftment most likely by increasing MMP-13 activity (Fig. 6B). Finally, ANGPTL2 expression was positively correlated with CXCR4 expression levels in primary tumor tissue from breast cancer patients.
CXCL12 signaling through CXCR4 plays important physiological roles, as in hematopoietic stem cell homing to the bone marrow www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 9170 | DOI: 10.1038/srep09170 niche 38 . CXCL12 signaling through CXCR4 also plays critical roles in cancer metastasis. In breast cancer, CXCL12 derived from various tissues recruits breast tumor cells that express CXCR4 by increasing their chemotaxis to regions of high CXCL12 expression, such as lymph nodes, lung, liver, and bone tissues 16 . Thus CXCR4 downregulation in breast cancer cells could decrease their metastasis. In  this study, we showed that ANGPTL2 significantly up-regulated CXCR4 expression in breast cancer cells, which likely enhances their responsiveness to CXCL12. We previously reported that ANGPTL2expressing breast cancer or osteosarcoma cells metastasized to lung 15,26 , where CXCL12 expression is high. Therefore, ANGPTL2 activity is likely not limited to bone. However, in this study, bone metastasis by ANGPTL2-expressing cells is often observed in the mouse intracardiac inoculation model. As yet uncharacterized mechanisms that we will address in future studies may govern a preference for bone metastasis.
Molecular mechanisms underlying CXCR4 expression have not been fully clarified, although factors such as HIF1a, NF-kB, and TGF-b reportedly regulate CXCR4 expression 39 . In particular, NF-kB induces CXCR4 expression in tumor cells 32 , and we observed that cell surface CXCR4 expression in breast cancer cells is partially reduced by treatment with an NF-kB inhibitor (unpublished results). This finding suggests the possibility that NF-kB is also activated by ANGPTL2, a conclusion consistent with our previous report that ANGPTL2 activates NF-kB-dependent pro-inflammatory signaling 19,23 .
The present study suggests that ETS1-dependent transcriptional regulation is important for ANGPTL2-induced CXCR4 expression. ETS1 activity is also reportedly important for transformation of endothelial cell phenotypes from quiescent to angiogenic 40 and for regulation of angiogenesis by pro-angiogenic factors, such as VEGF 41 , angiotensin II 42 or TGF-b 43 . Therefore, tumor cell-derived ANGPTL2 may up-regulate ETS1-dependent activation of proangiogenic genes, promoting tumor angiogenesis. ANGPTL2 also exhibits direct pro-angiogenic activity 24 . Taken together with these findings, ANGPTL2 likely plays both direct and indirect roles in tumor angiogenesis in the case of breast cancers, and these effects may underlie enhanced bone metastasis.
We also found that increased MMP-13 expression seen following CXCL12 treatment was significantly attenuated in ANGPTL2 knockdown cells. Many reports indicate that elevated circulating MMP-13 levels are associated with progression and metastasis of various cancers and specifically correlate with decreased overall patient survival and increased lymph node metastasis in breast cancer 44 , increased bone metastasis in renal cell carcinoma 45 , and poor prognosis of nonsmall cell lung and colorectal cancers 46,47 . MMP-13 cleaves type I collagen, which constitutes almost all collagen found in bone. In breast cancer pathology, MMP-13-selective inhibitors show effective therapeutic effects against cancer-induced bone osteolysis 48,49 . Recently, we reported that ANGPTL2 enhances tumor cell invasion by increasing expression of MMP-1, MMP-9, and MMP-13 In osteosarcoma cells 15,21 . Thus, ANGPTL2 secretion from cancer cells likely contributes to MMP activity in those cells.
In summary, here we show that breast cancer cell-derived ANGPTL2 may play important roles in bone metastasis by enhancing responsiveness of breast cancer cells to bone tissue-secreted CXCL12 through upregulated tumor cell CXCR4 expression. Furthermore, our findings suggest that in bone tissue, CXCL12-activated CXCR4 signaling in breast cancer cells induced by ANGPTL2 also accelerates osteolysis and bone engraftment. These studies could form the basis of new therapeutic strategies to antagonize bone metastasis of breast cancer cells.

Methods
Cell culture. Human breast adenocarcinoma cell lines MDA-MB231 and T47-D were purchased from the American Type Culture Collection (ATCC). Cells were confirmed to be identical to cells registered with ATCC by comparison with the database of the JCRB Cell Bank. MDA-MB231 cells were cultured in Leibovitz's L15 (Wako, Osaka, Japan) medium supplemented with 10% FCS under 100% air without CO 2 , and T47-D cells were cultured in RPMI-1640 (Wako) medium supplemented with 10% FCS under 5% CO 2 /95% air.
Proliferation assay. Proliferation assays were performed using the Cell Counting Kit-8 (Dojindo, Kumamoto, Japan) in normoxia (under 100% air without CO 2 ) and hypoxia (1% O 2 without CO 2 ) conditions. 100 ml cell suspensions (5 3 10 4 /ml) with 10% FCS medium was added to a 96-well plate and incubated for 6 hour at 37uC. Then, 10 ml CCK-8 reagent was added to each well, cells were incubated 2 hours at 37uC, and optical density at 450 nm was measured. Similar assays were performed after 24, 48, and 72 hours. The growth rate was expressed as a value relative to hour 0.
MicroRNAi silencing of ANGPTL2 expression. MDA-MB231 cells were transfected with ANGPTL2-specific small interfering RNA vectors and control vectors 26  To evaluate differential expression, expression data was normalized, and gene annotations were added using RegionMiner with Genomatix Genome Analyzer (Genomatix, Munich, Germany) software. The Normalized Expression value (NEvalue) was based on the following formula: NE~10 7 Ã number of reads region =(number of reads mapped Ã length region ): 52 To compile NE-value information for an individual gene, mean NE-values were calculated from all gene transcripts, as were log2-transformed fold-changes between MB231/miANGPTL2 and MB231/miLacZ cells. RNA-seq data were deposited in the DDBJ Sequence Read Archive (DRA). Accession number: DRA002859.
Quantitative real-time PCR. Total RNA was extracted from cells using TRIzol reagent (Invitrogen). Deoxyribonuclease-treated RNA was reverse-transcribed with a PrimerScript RT Reagent Kit (Takara Bio, Otsu, Shiga, Japan). PCR was performed using SYBR Premix Ex Taq II (Takara Bio). Oligonucleotides used for PCR are listed in Supplementary Table 1. PCR products were analyzed with a Thermal Cycler Dice Real Time system (Takara Bio), and relative transcript abundance was normalized to that of RPS18.
Transwell migration assay. Assays were performed using a Costar Transwell Permeable Support system with 8.0 mm pore size. In brief, 5 3 10 4 cells were suspended in 1% FCS media and placed in the upper chamber of a transwell plate without ligand, while the lower chamber contained 1% FCS media plus CXCL12 at a final concentration of 50, 100 or 200 ng/ml (R&D System, Minneapolis, MN, USA). After 18 h incubation at 37uC, cells on the upper surface of the filter were removed by wiping with cotton swab, and cells that had migrated to the lower side were fixed with formalin and stained with Giemsa and haematoxylin. Migration was quantified using 6 standardized microscopy fields per membrane.
Invasion assay. MB231/miLacZ and MB231/miANGPTL2 cells were cultured in 96well plates (ESSEN ImageLock plates) according to the manufacturer's instructions (ESSEN IncuCyte ZOOM) and visualized using a real-time cell imaging system (IncuCyte TM live-cell, ESSEN BioScience, Inc). In brief, plates were coated with a thin layer (50 ml) of collagen-1 (300 mg/ml) (R&D System) and placed in a 37uC incubator with 5% CO 2 overnight. Then, 100 ml of the cell suspension (1 3 10 5 cells/ml) was added to each well, plates were incubated 6 h, and a scratch was made in the plate using a 96-pin WoundMaker TM . 50 ml of a 53 neutralization solution (DMEM with 1.875% bicarbonate buffer and 25% FBS) plus a working stock of 200 ml of 3.75 mg/ ml collagen-I in 20 mM acetic acid were mixed and 100 ml of the mixture was added to each plate for 30 min. Then, 100 ml of culture media was added to each. Images were automatically acquired and registered by the IncuCyte TM software system (ESSEN BioScience). Typical kinetic updates were recorded at 2 h intervals over a 24 h period. CXCL12 (200 ng/ml final concentration, R&D Systems) was added to the neutralization solution with or without 20 mM of the MKK inhibitor U0126 (Merck KGaA, Darmstadt, Germany).
Animal studies. All experiments were performed according to guidelines of Institutional Animal Committee of Kumamoto University. For intracardiac injections, subconfluent tumor cells were harvested, washed in PBS, and resuspended at a concentration of 1 3 10 6 cells/ml. Non obese diabetic/severe combined immunodeficient (NOD/SCID) Janus kinase 3 knockout (NOJ) mice 26 were anesthetized by isoflurane and placed in the supine position. With a 29-gauge needle, 1 3 10 5 cells were injected into the left ventricle of anesthetized mice after visualization of arterial blood flow into the syringe. Metastasis was monitored by bioluminescence imaging for 5 weeks after injection. D-Luciferin (100 mg/g) was injected into back skin of anesthetized mice before imaging, and images were then acquired using a NightOWL II LB 983 System (Berthold Technologies, Bad Wildbad, Germany). Luminescence was calculated using IndiGO2 software. For the orthotopic implantation model, subconfluent tumor cells were harvested, washed in PBS, and resuspended at 1 3 10 7 cells/ml. Female 8-12 week-old NOJ mice were anesthetized by isoflurane, and for the breast cancer model, cells (1 3 10 6 ) in 100 ml PBS were injected into their mammary fat pads.
Human studies. The study group included 181 specimens of female primary invasive duct breast cancer cases from patients diagnosed between January of 1989 and December of 1996 at Kumamoto University. This study was approved by the Ethics Committees of Kumamoto University. Written informed consent was obtained from each subject.
Evaluation of immunostaining. Each patient sample was evaluated by two independent investigators blinded to the clinical pathology parameters. We divided the 181 total samples (including 178 assessed for MMP-13, as 3 were excluded due to sample loss) into two groups, based on the median percentage of ANGPTL2-, CXCR4-or MMP-13-positive invasive cells. ANGPTL2-or MMP-13 positive groups were defined as having $50% positive cells; the CXCR4-positive group was defined as having $25% positive cells.
Treatment of cells with integrin a5b1 antibody. Parental MDA-MB231 cells were maintained in serum-free Leibovitz's L15 for 24 hours. Then cells were incubated with a5b1 antibody (25 mg/ml) for 24 hours. www.nature.com/scientificreports