Human antigen R-regulated CCL20 contributes to osteolytic breast cancer bone metastasis

Breast cancer mainly spreads to bone, causing decreased survival of patient. Human antigen R (HuR) and chemokines are important molecules associated with mRNA stability and cell-cell interaction in cancer biology. Here, HuR knockdown inhibited bone metastasis and osteolysis of metastatic breast cancer cells in mice and HuR expression promoted the metastatic ability of cancer cells via CCL20 and GM-CSF. In contrast with the findings for GM-CSF, ELAVL1 and CCL20 expressions were markedly increased in breast tumor tissues and ELAVL1 expression showed a strong positive correlation with CCL20 expression in breast cancer subtypes, particularly the basal-like subtype. Metastasis-free survival and overall survival were decreased in the breast cancer patients with high CCL20 expression. We further confirmed the role of CCL20 in breast cancer bone metastasis. Intraperitoneal administration of anti-CCL20 antibodies inhibited osteolytic breast cancer bone metastasis in mice. Treatment with CCL20 noticeably promoted cell invasion and the secretion of MMP-2/9 in the basal-like triple-negative breast cancer cell lines, not the luminal. Moreover, CCL20 elevated the receptor activator of nuclear factors kappa-B ligand/osteoprotegerin ratio in breast cancer and osteoblastic cells and mediated the crosstalk between these cells. Collectively, HuR-regulated CCL20 may be an attractive therapeutic target for breast cancer bone metastasis.


HuR knockdown inhibits bone metastasis of breast cancer cells.
To determine whether the expression of HuR, an RNA-binding post-transcriptional regulator, is essential for breast cancer bone metastasis, we used an MDA-MB-231 basal-like/triple-negative human breast cancer cell line that has exhibited well-characterized bone tropism in animal models 35,36 and established luciferase-transfected HuR-knockdown (shHuR) cells and control (shNC) cells with corresponding non-specific scrambled shRNA ( Supplementary Fig. 1a). The shNC and shHuR cells were inoculated into the left cardiac ventricles of nude mice, and 6 weeks later, breast cancer metastasis was detected by bioluminescence imaging. The metastasis was substantially inhibited in the shHuR cell-inoculated mice compared with the shNC cell-inoculated mice, as indicated by the bioluminescence images (Fig. 1a). X-ray and three-dimensional (3D) images derived from the micro-computed tomography (μCT) data revealed that osteolytic lesions were notably decreased in the mandibles (Fig. 1b), distal femora, and proximal tibiae (Fig. 1c) of the shHuR cell-injected mice compared with those of the shNC-inoculated mice. Goldner's trichrome staining also showed that HuR knockdown significantly inhibited tumor growth in bone marrow (Fig. 1c,d). Bone resorption in the femora was supported by the reduced number of tartrate-resistant acid phosphatase (TRAP)-positive osteoclasts on the bone surfaces near the tumors in the shHuR cell-injected mice compared with the shNC cell-injected mice (Fig. 1c,e). In the distal femora, the bone volume over total volume (BV/ TV), trabecular thickness (Tb.Th), and trabecular number (Tb.N) were reduced in the shNC cell-inoculated mice compared with the control mice, whereas trabecular separation (Tb.Sp) and the structure model index (SMI) were increased. These alterations were significantly inhibited by HuR knockdown (Fig. 1f). The serum levels of the bone resorption markers TRAP 5b and C-terminal cross-linking telopeptide of type I collagen (CTX) were increased in the shNC-cell inoculated mice, and these increases were also markedly inhibited by HuR knockdown (Fig. 1g). We confirmed that HuR knockdown inhibited tumor growth and osteolysis in a murine intratibial model of breast cancer bone metastasis (Supplementary Fig. 1b-f).
To further verify the in vivo inhibitory effect of HuR knockdown on breast cancer metastasis, we assessed the viability, migration, and invasion of shNC and shHuR MDA-MB-231 cells using MTT, wound-healing, and transwell invasion assays, respectively. Cell viability was decreased by 32.1%, 37.1%, and 49.5% at 24, 48, and 72 h, respectively, in the shHuR cells compared with the shNC cells (Fig. 1h). Scratch wound closure was inhibited by 50.4% (Fig. 1i) and the number of invaded cells was reduced by 66.0% following HuR knockdown (Fig. 1j). These results indicate that HuR expression in breast cancer cells is closely associated with their metastatic potential and production of osteolytic lesions.
HuR influences breast cancer cells via CCL20 and granulocyte-macrophage colony-stimulating factor. Breast cancer-mediated bone loss is triggered by factors secreted from bone metastatic cells, and these factors accelerate osteolysis by promoting crosstalk among cancer cells, osteoblasts, and osteoclasts, creating a vicious cycle 2 . Chemokines greatly contribute to these cell-cell interactions 27 . We attempted to identify chemokines that are regulated by HuR and participate in this cycle using a Human Cytokine Array C6, which is composed of a nitrocellulose membrane spotted with antibodies specific to 60 different cytokines. The levels of CCL20 and granulocyte-macrophage colony-stimulating factor (GM-CSF) were decreased by 57.2% and 65.1%, respectively, in conditioned medium from shHuR MDA-MB-231 cells compared with that from shNC cells, whereas the levels of other cytokines did not indicate substantial inhibition ( Fig. 2a and Supplementary Fig. 2a). In addition, the production of CCL20 and GM-CSF in various breast cancer cell lines, including the luminal B and basal-like/ triple-negative breast cancer subtypes, was also regulated by HuR. Relatively high CCL20 levels were detected in conditioned media from ZR-75-1, MDA-MB-231, and BT549 cells, and HuR knockdown inhibited these levels by 38.2%, 40.1%, and 38.5%, respectively. In addition, high GM-CSF levels were detected that were inhibited by 20.3% and 24.3% by HuR knockdown in MDA-MB-231 and BT549 cells, respectively (Fig. 2b). Furthermore, reduced expression of CCL20 and GM-CSF was observed in tumor tissues from mice intracardially injected with shHuR cells based on immunohistochemical examination (Fig. 2c). However, CCR6 expression did not change by HuR knockdown (Supplementary Fig. 2b).
To clarify whether CCL20 and GM-CSF are critical factors that mediate the function of HuR in breast cancer bone metastasis, we added antibodies specific to these chemokines to culture media of MDA-MB-231 cells. Treatment with up to 500 ng/ml of anti-CCL20 and anti-GM-CSF antibody for 24 h slightly reduced cell viability by 19.3% and 23.4%, respectively (Fig. 2d). Additionally, treatment with these antibodies inhibited cell migration by 14.2% and 45.1% at 5 ng/ml and by 64.8% and 54.7% at 50 ng/ml (Fig. 2e). Cell invasion was also inhibited by 60.1% and 14.3% in response to neutralization with the anti-CCL20 and anti-GM-CSF antibodies at 5 ng/ ml, respectively, and by close to 100% in response to antibody neutralization at 50 ng/ml (Fig. 2f). Furthermore, treatment with the CCL20 and GM-CSF chemokines rescued the metastatic ability of HuR-knockdown cells. Although the treatments with these chemokines did not markedly affect cell viability ( Supplementary Fig. 3a), they did distinctly induce the migration ( Supplementary Fig. 3b) and invasion ( Supplementary Fig. 3c) of shHuR MDA-MB-231 cells in a dose-dependent manner.
HuR has been reported to directly bind to the AREs in the 3′-UTR of GM-CSF mRNA, particularly to the class II AREs, which are characterized by multiple overlapping AUUUA motifs 18,37 . However, it is unknown whether HuR has a role in stabilizing CCL20 mRNA. We found that the human CCL20 mRNA contains three scattered AUUUA motifs, similar to the class I AREs present in early-response gene-encoding mRNAs ( Supplementary  Fig. 4a). RNA-based electrophoretic mobility shift assay (REMSA) was performed using biotin-labeled RNA probes containing a HuR-binding motif in the 3′-UTR of CCL20 mRNA. The results showed that the labeled RNA probes formed a complex with HuR in lysates from the shNC MDA-MB-231 cells and that the formation of (e,f) Migration (e) and invasion (f) capabilities of cells treated with anti-CCL20 or anti-GM-CSF antibody. Cell migration and invasion were measured by wound-healing and transwell invasion assays, respectively. The data are expressed as the mean ± s.e.m. *P < 0.05, **P < 0.01 versus cells without anti-CCL20 or anti-GM-CSF.
this complex was significantly decreased in lysates from the shHuR cells ( Supplementary Fig. 4b). The binding of HuR to RNA probes was blocked in the presence of an excess amount of unlabeled oligonucleotides. These data demonstrate that HuR regulates CCL20 production by directly binding to the 3′-UTR of CCL20 mRNA. Taken together, our results suggest that HuR in breast cancer cells regulates the production of CCL20 and GM-CSF and that the secretion of these chemokines promotes cancer cell migration and invasion.
Gene expression of CCL20 rather than that of GM-CSF is highly correlated with gene expression of HuR, distant metastasis-free survival, and overall survival in breast cancer. To evaluate the clinical relevance of our results, we analyzed breast cancer cases from the Cancer Genome Atlas (TCGA) database. The ELAVL1 (the gene that encodes HuR) and CCL20 expression levels were both markedly increased in breast tumor tissues compared with normal breast tissues (Fig. 3a). In contrast, neither CSF2 (the gene that expression (e,f). P values were determined using the log-rank test. Hazard ratios (HRs) with 95% confidence intervals are shown. encodes GM-CSF) nor CCR6 (the gene that encodes the only known receptor for CCL20) expression was upregulated ( Supplementary Fig. 5a). In addition, ELAVL1 expression showed a strong positive correlation with CCL20 expression (Fig. 3b) but not with CSF2 ( Supplementary Fig. 5b) or CCR6 expression ( Supplementary Fig. 5c) in breast cancer subtypes, particularly the basal-like subtype. Kaplan-Meier survival analysis of GSE3494, GSE7390 and GSE26971 showed that metastasis-free survival (Fig. 3c) and overall survival (Fig. 3d) were decreased in the breast cancer patients with high CCL20 expression. In contrast, no correlation was detected between high ELAVL1, CSF2, or CCR6 expression and metastasis-free survival (Fig. 3e, Supplementary Fig. 5d,f) or overall survival (Fig. 3f, Supplementary Fig. 5e,g). These results suggest that breast cancer patients with upregulated CCL20 expression are at a high risk of metastasis.
Anti-CCL20 antibody treatment prevents bone metastasis of breast cancer cells in mice. Although NF-kB-regulated GM-CSF has been implicated in osteolytic bone metastasis in mice injected with MDA-MB-231 breast cancer cells 38 , the role of CCL20 has not yet been determined. To assess whether CCL20 could serve as a novel therapeutic target for bone metastasis of breast cancer cells, we inoculated luciferase-transfected MDA-MB-231 cells into the left cardiac ventricles of nude mice. Then, we intraperitoneally administered either anti-human CCL20 antibody to target inoculated human cancer cell-derived CCL20 or anti-mouse CCL20 antibody to target surrounding non-cancer cell-derived CCL20 to the mice for 6 weeks. MDA-MB-231 cell metastasis was significantly inhibited by treatment with each antibody (Fig. 4a). In particular, treatment with each antibody at 100 μg/kg notably decreased the production of osteolytic lesions in the mandibles (Fig. 4b), distal femora, and proximal tibiae (Fig. 4c) of the MDA-MB-231 cell-injected mice, as shown by X-ray and 3D images derived from μCT scans. In addition, these treatments resulted in considerable decreases in tumor volume and bone destruction, as demonstrated by Goldner's trichrome staining (Fig. 4c,d), and the reduced production of TRAP-positive activated osteoclasts on bone surfaces close to cancer cells (Fig. 4c,e). Intraperitoneal administration of anti-CCL20 antibody at 100 μg/kg inhibited the decreases in the BV/TV, Tb.Th, and Tb.N and the increases in Tb.Sp and SMI compared to those of the vehicle treated-mice (Fig. 4f). The antibody treatments also suppressed the elevated serum TRAP 5b and CTX levels caused by MDA-MB-231 cell inoculation (Fig. 4g). These results suggest that CCL20 is a pivotal factor for osteolytic breast cancer bone metastasis.

CCL20 promotes the metastatic ability of breast cancer cells.
To examine how CCL20 contributes to bone metastasis of breast cancer cells, we evaluated the effects of CCL20 treatment on the proliferation (Fig. 5a,b), migration (Fig. 5c), and invasion (Fig. 5d) of MDA-MB-231 cells. These processes were all dose-dependently increased in response to the treatment. In particular, cell migration and invasion were markedly increased. Degradation of the extracellular matrix by proteinases, such as MMPs and uPA, is pivotal for the invasion and metastasis of cancer cells. We found that MMP-1 and MMP-2/9 activities were markedly enhanced in conditioned medium from CCL20-treated MDA-MB-231 cells, as demonstrated by collagen and gelatin zymography, respectively (Fig. 5e). uPA activity was also increased by 38.3% in MDA-MB-231 cells treated with 200 ng/ml CCL20 (Fig. 5f). In addition, treatment with 200 ng/ml CCL20 significantly enhanced cell invasion (Fig. 5g) and stimulated the secretion of MMP-2 and MMP-9 (Fig. 5h) in BT549 and HCC38 cell lines, which are basal-like/ triple-negative breast cancer cell lines, as is the MDA-MB-231 cell line. In contrast, the invasiveness of CCL-20-treated MCF-7 and ZR-75-1 luminal breast cancer cells was not enhanced, and no MMP activity was detected. These results demonstrate that CCL20 has greater effects on aggressive and highly metastatic breast cancer cell lines than on less metastatic breast cancer cell lines.

CCL20 induces RANKL expression and mediates interactions between breast cancer cells and osteoblastic cells. RANKL regulates the differentiation and activity of osteoclasts through its receptor
RANK and is thus a key factor for bone metastasis 2, 39 . Breast cancer cells can both produce RANKL directly 40 and affect osteoblastic RANKL expression by secreting osteolytic factors 41 . This induction of osteoclastogenesis results in cancer-related bone loss. We found that CCL20 enhanced the RANKL/OPG ratio by stimulating RANKL expression and secretion and reducing OPG expression and secretion in MDA-MB-231 human breast cancer cells in a dose-dependent manner (Fig. 6a,b). CCL20 also increased the RANKL level in the culture medium of hFOB1.19 human osteoblastic cells but did not change the secreted OPG level (Fig. 6c). Further, the osteoblastic cells produced CCL20 (Fig. 6d), and the antibody-mediated neutralization of osteoblastic cell-derived CCL20 significantly inhibited the invasiveness of MDA-MB-231 breast cancer cells (Fig. 6e). These results suggest that CCL20 produced by breast cancer cells elevates the osteoblastic RANKL/OPG ratio and that osteoblastic CCL20 also contributes to the invasiveness of breast cancer cells.

Discussion
Controlling both the metastatic potential of cancer cells and the vicious cycle of bone metastase-induced bone destruction may be the best strategy to improve the survival and quality of life of patients with bone-tropic cancer. HuR plays crucial roles in essential steps in the metastatic process 21 and is closely associated with the poor prognoses of lung, ovarian, colorectal, and breast cancer patients 23,25,42 . The development and progression of cancer-related skeletal diseases depends on the production of soluble factors within the metastatic niche that cooperate to promote the bone-resorbing functions of osteoclasts and to trigger interactions between cancer cells and surrounding cells 43 . Chemokines are very likely to represent these factors.
We attempted to determine the role of HuR in breast cancer bone metastasis and to identify novel osteolytic factors with high potentials as therapeutic targets from HuR-regulated chemokines. We demonstrated that HuR expression promoted breast cancer bone metastasis by considerably reducing the tumor burden and osteolysis in murine models inoculated with HuR-knockdown breast cancer cells. These findings were supported by the reduced metastatic potential of HuR-knockdown MDA-MB-231 cells. We further identified that the metastatic ability of HuR-expressing breast cancer cells was related to their secretion of CCL20 and GM-CSF. However, in our analysis using public databases, expression of the GM-CSF-encoding gene CSF2 was not strongly correlated with that of the HuR-encoding gene ELAVL1 in the different molecular subtypes of breast cancer. Additionally, neither distant metastasis-free survival nor overall survival was significantly correlated with CSF2 expression in the breast cancer patients. In contrast, CCL20 expression was significantly correlated with the above analyses. Thus, CCL20 may be a better target for breast cancer bone metastasis than GM-CSF. Recombinant G-CSF/GM-CSF, which has hematopoietic activity, is routinely used to reverse leukopenia due to chemotherapy and radiation and as an adjuvant to boost the immunogenicity of tumor cells for diminishing drug resistance in anti-tumor therapy 44 . In contrast, G-CSF/GM-CSF has also been reported to promote cancer cell dissemination and bone metastasis via increased osteoclast formation and the subsequent homing of malignant cells to bone tissues 38,45 . The roles of CCL20 in cancer development, tumor promotion, and metastasis are linked to the attraction of CCR6-expessing regulatory T lymphocytes to tumor sites expressing CCL20 [46][47][48][49] and to the recruitment of CCR6-expressing cancer cells to metastatic sites with abundant CCL20 50 . CCL20 expression has been reported to be upregulated in osteoblasts and osteoclast precursors co-cultured with multiple myeloma cells, and this upregulation contributes to the formation of osteoclasts and osteolytic bone lesions in multiple myeloma patients 51 . Our data verified that suppression of the functions of CCL20 molecules derived from cancer cells and surrounding non-cancer cells inhibited tumor growth in bone marrow, as well as severe osteolysis, in mice inoculated with MDA-MB-231 cells through the left cardiac ventricles. In addition, CCL20 markedly increased cell migration and invasion and the activities of related proteases in basal-like/triple-negative breast cancer cell lines but not in luminal breast cancer cell lines, suggesting that the targeting of CCL20 may be more effective in patients with aggressive breast cancer.
RANKL and OPG, which are both expressed by osteoblasts/stromal cells, are critical regulators of osteoclast differentiation. The binding of RANKL to its receptor, RANK, on osteoclast precursor cells stimulates osteoclast differentiation; however, its decoy receptor, OPG, blocks osteoclast differentiation by binding to RANKL 52,53 . An abnormally increased RANKL/OPG ratio can lead to excessive osteoclastogenesis and subsequent bone resorption. Recent studies have also reported a critical role of RANKL in the development and progression of breast  cancer 40,54 . We found that breast cancer cells and bone cells influence each other in the bone microenvironment by producing CCL20; the RANKL/OPG ratio was increased in both MDA-MB-231 cells and human osteoblast hFOB1.19 cells exposed to CCL20, and osteoblast-derived CCL20 promoted breast cancer cell invasiveness. Thus, CCL20 may be a key factor for the acceleration of breast cancer metastasis and related bone loss.
Collectively, CCL20, which is regulated by HuR and secreted into the bone microenvironment, enhances the metastatic ability and bone metastasis of breast cancer cells by stimulating cancer cells and osteoblastic cells in both autocrine and paracrine manners. Thus, CCL20 may have high potential as a therapeutic target in breast cancer patients with bone metastasis.

Methods
More details of reagents, antibodies, cell lines, and cell culture conditions were described in Supplementary materials and methods.

In vivo models.
After 5 (for the intratibially injected mice) or 6 weeks (for the intracardially injected mice), the mice were anesthetized for bioluminescence imaging 55 and μCT analysis 55,56 . Quantitative bioluminescence data were expressed as photons/sec/cm 2 /sr. Goldner's trichrome and TRAP stainings and immunohistochemical analysis for HuR, CCL20, and GM-CSF were accomplished in the sections derived from mouse hind limbs as describe previously 55,56  Cytokine and chemokine array. shNC and shHuR MDA-MB-231 cells (5 × 10 5 cells/dish) were seeded in 60-mm culture dishes containing 10% FBS-DMEM and incubated overnight. The cells were washed twice with PBS, cultured in serum-free DMEM for 24 h, and centrifuged at 200 × g for 5 min. Multiple cytokines in the collected conditioned media were detected using a RayBio Human Cytokine Ab Array 6 (Ray Biotech, Norcross, GA) according to the manufacturer's instruction. The membranes used in the array were scanned, and the intensity of each cytokine was analyzed using TINA software (Raytest, Straubenhardt, Germany). The relative cytokine intensities were normalized using the positive control spots on the same membrane.
REMSA. shNC and shHuR MDA-MB-231 cells (1 × 10 6 cells/plate) were seeded into 100-mm culture dishes and cultured in DMEM supplemented with 10% FBS. Cell lysates were prepared using RIPA buffer containing 1 mM phenylmethylsulfonyl fluoride (PMSF, Millipore, Billerica, MA) and protease inhibitor cocktail. The cell lysates were centrifuged at 22,000 × g for 15 min at 4 °C, and the protein concentrations in the supernatants were determined using a BCA protein assay. The sequence of the probe for the HuR-specific binding was 5′-AUUUUAUGUUAUUUAUAGCUGUAGGUUU-3′ (563-590 nt). The probes were synthesized as single strands that were labeled at the 3′-end with biotin. Unlabeled oligonucleotides were used as competitors. REMSA was performed using a LightShift Chemiluminescent EMSA Kit (Pierce) according to the manufacturer's protocol.
Public database analysis. ELAVL1, CCL20, CSF2, and CCR6 mRNA expression levels in normal and breast cancer tissues were retrieved from the TCGA Research Network website. Correlations of ELAVL1 expression with CCL20, CSF2, and CCR6 expression were reanalyzed using Statistical Package for the Social Sciences (SPSS) software (IBM, Endicott, NY). Additional details of the study are available on the TCGA Research Network website. Kaplan-Meier plots for the overall survival and metastasis-free survival of breast cancer patients were generated using the GSE3994, GSE7390, and GSE26971 datasets with Kaplan-Meier Plotter (http://kmplot.com/analysis/) 58 . The patient samples were split into a high expression group and a low expression group for each gene using the following setting of best cutoff. The JetSet best probe set was used to analyze ELAVL1, CCL20, CSF2, and CCR6 expression.
BrdU incorporation. MDA-MB-231 cells (1 × 10 4 cells/well) were seeded into a 96-well plate and cultured in serum-free DMEM supplemented with CCL20 at the indicated concentrations for 24, 48, and 72 h. The cells were then treated with BrdU (10 µM) for 2 h. The amount of BrdU incorporated into the DNA of proliferating cells was quantified using a BrdU labeling and detection kit (Roche Diagnostics) according to the manufacturer's protocol.
Zymography and uPA activity. BT549, HCC38, MCF-7, MDA-MB-231, and ZR-75-1 cells (5 × 10 5 cells/ dish) were each seeded into 60-mm culture dishes and incubated with the indicated concentrations of CCL20 in serum-free medium for 24 h. The culture media were collected by centrifugation at 200 × g for 5 min. MMP activities in the collected media were determined on 8% polyacrylamide gels containing 0.8 mg/ml gelatin (for MMP-2 and MMP-9) or 1 mg/ml collagen (for MMP-1) as described previously 55 . The gels were stained with 0.1% Coomassie Blue R-250, and MMP activities were assessed by analyzing clear bands against a blue background. uPA activity was determined using a uPA activity assay kit (Chemicon, Billerica, MA) according to the manufacturer's instruction.
Western blot analysis. shNC and shHuR MDA-MB-231 cells (5 × 10 5 cells/dish) were each cultured in 10% FBS-DMEM for 24 h. Total cell lysates were prepared using RIPA buffer containing 1 mM PMSF and protease inhibitor cocktail. The lysates were centrifuged and the protein concentrations in the supernatants were measured using a BCA kit. Equal amounts of each lysate (30 μg) were separated on sodium dodecyl sulfate-polyacrylamide gels. Target proteins were detected using a primary antibody (1:1000) against HuR, RANKL, OPG, or β-actin as described previously 55, 56 . Detection of secreted RANKL and OPG. MDA-MB-231 (1 × 10 4 cells/well) and hFOB1.19 cells (1 × 10 5 cells/well) were treated with CCL20 at the indicated concentrations in serum-free media for 24 h, and the media were collected. The RANKL and OPG levels in the culture media were determined using commercially available ELISA kits (EIAab, Guangguguoji, China) according to the manufacturer's instructions.
Statistical analysis. The data are expressed as the mean ± s.e.m. Statistical analyses were performed using one-way ANOVA and Student's t-test to detect significant differences between two groups. A P < 0.05 was considered statistically significant. In public database analysis, the data retrieved from the TCGA website were reanalyzed using SPSS (IBM, Endicott, NY) to determine Pearson's correlation coefficients (r) between ELAVL1 expression and CCL20, CSF2, and CCR6 expression. Hazard ratios (HRs) with 95% confidence intervals and P values were calculated using the log-rank test.
Data availability. All data generated during this study are included in this published article and its