Thymoquinone Inhibits Bone Metastasis of Breast Cancer Cells Through Abrogation of the CXCR4 Signaling Axis

Overexpression of chemokine receptor type 4 (CXCR4) has been found to be associated with increased cell proliferation, metastasis and also act as an indicator of poor prognosis in patients with breast cancer. Therefore, new agents that can abrogate CXCR4 expression have potential against breast cancer metastasis. In this study, we examined the potential effect of thymoquinone (TQ), derived from the seeds of Nigella sativa, on the expression and regulation of CXCR4 in breast cancer cells. TQ was found to inhibit the expression of CXCR4 in MDA-MB-231 triple negative breast cancer (TNBC) cells in a dose- and time-dependent manner. It was noted that suppression of CXCR4 by TQ was possibly transcriptionally regulated, as treatment with this drug caused down-regulation of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) activation and suppression of NF-κB binding to the CXCR4 promoter. Pretreatment with a proteasome inhibitor and/or lysosomal stabilization did not affect TQ induced suppression of CXCR4. Down-regulation of CXCR4 was further correlated with the inhibition of CXCL12-mediated migration and invasion of MDA-MB-231 cells. Interestingly, it was observed that the deletion of p65 could reverse the observed anti-invasive/anti-migratory effects of TQ in breast cancer cells. TQ also dose-dependently inhibited MDA-MB-231 tumor growth and tumor vascularity in a chick chorioallantoic membrane assay model. We also observed TQ (2 and 4 mg/kg) treatment significantly suppressed multiple lung, brain, and bone metastases in a dose-dependent manner in a metastasis breast cancer mouse model. Interestingly, H&E and immunohistochemical analysis of bone isolated from TQ treated mice indicated a reduction in number of osteolytic lesions and the expression of metastatic biomarkers. In conclusion, the results indicate that TQ primarily exerts its anti-metastatic effects by down-regulation of NF-κB regulated CXCR4 expression and thus has potential for the treatment of breast cancer.

Overexpression of chemokine receptor type 4 (CXCR4) has been found to be associated with increased cell proliferation, metastasis and also act as an indicator of poor prognosis in patients with breast cancer. Therefore, new agents that can abrogate CXCR4 expression have potential against breast cancer metastasis. In this study, we examined the potential effect of thymoquinone (TQ), derived from the seeds of Nigella sativa, on the expression and regulation of CXCR4 in breast cancer cells. TQ was found to inhibit the expression of CXCR4 in MDA-MB-231 triple negative breast cancer (TNBC) cells in a dose-and time-dependent manner. It was noted that suppression of CXCR4 by TQ was possibly transcriptionally regulated, as treatment with this drug caused down-regulation of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) activation and suppression of NF-κB binding to the CXCR4 promoter. Pretreatment with a proteasome inhibitor and/or lysosomal stabilization did not affect TQ induced suppression of CXCR4. Down-regulation of CXCR4 was further correlated with the inhibition of CXCL12-mediated migration and invasion of MDA-MB-231 cells. Interestingly, it was observed that the deletion of p65 could reverse the observed antiinvasive/anti-migratory effects of TQ in breast cancer cells. TQ also dose-dependently inhibited MDA-MB-231 tumor growth and tumor vascularity in a chick chorioallantoic membrane assay model. We also observed TQ (2 and 4 mg/kg) treatment significantly

INTRODUCTION
Breast cancer is the second most common cancer that afflicts women, with an estimated 1.67 million women diagnosed with breast cancer in 2012 Torre et al., 2015). Breast cancer ranked fifth in cancer-associated deaths among all cancers globally in 2012 Torre et al., 2015). While incredible development has been made for the treatment of breast cancers, the treatment of triple negative breast cancer (TNBC) is quite a challenge owing to its destructive features and limited treatment options (Jia et al., 2016;Wang et al., 2018). Often, chemotherapy for breast cancer is ridden with side effects and drug resistance, leading to failure in therapy. Compelling evidence suggests that if breast cancer metastasis to distant site organs can be prevented, then it is an indicator for good prognostic outcome (Alvarez et al., 2010;Perez and Spano, 2012). Interactions between chemokines and their cognate receptors play important roles in tumor metastasis. The interaction between chemokine receptor type 4 (CXCR4) and its cognate ligand stromal-derived factor-1 (SDF1 or CXCL12) plays a crucial role in the regulation of migration and metastasis in a variety of solid tumors including breast cancer (Balkwill, 2004a,b). Moreover, SDF1 has been found to be frequently overexpressed in lymph nodes, lung, liver, and bone marrow (Muller et al., 2001;Keeley et al., 2010). Hence, in the present study, we investigated the effect of thymoquinone on the SDF1/CXCR4 signaling axis.
The master transcription factor nuclear factor kappa-lightchain-enhancer of activated B cells (NF-κB) plays a pivotal role in the development and progression of inflammation-driven diseases including cancer (Dey et al., 2008;Sethi et al., 2008bSethi et al., , 2012Sethi and Tergaonkar, 2009;Shanmugam et al., 2013;Li et al., 2015;Liu et al., 2018;Puar et al., 2018). In human chronic myeloid leukemia cells (KBM-5), TQ was reported to abrogate NF-κB activation and augment cellular apoptosis (Sethi et al., 2008a). Several other studies have shown that TQ can also down-regulate protein kinase B and extracellular receptor kinase signaling pathways (Yi et al., 2008). Woo et al., 2011 reported that TQ can exert a strong anti-proliferative effects in TNBC cells by activating peroxisome proliferator-activated receptor gamma (PPARγ) (Woo et al., 2011). TQ administered intraperitoneally, has been found to be well tolerated up to 22.5 mg/kg in male rats and 15 mg/kg in female rats; whereas for TQ administered orally, the dose was as high as 250 mg/kg in both male and female rats (Abukhader, 2012).
Our prior published data has already indicated that TQ can exert anti-cancer effects on MCF7 breast cancer cells through activation of the PPARγ signaling cascade (Woo et al., 2011). In a recent study TQ was shown to suppresses the proliferation, migration, and invasion of metastatic MDA-MB-321 breast cancer cells by inhibiting the p38 mitogen-activated protein kinase pathway in vitro and in vivo (Woo et al., 2013). Therefore, we postulated that TQ may modulate the expression of CXCR4 and inhibit tumor metastasis in vivo. Our results indicate that TQ can down-regulate CXCR4 expression and CXCL12-induced migration and invasion in breast cancer cell lines through inhibition of NF-κB activation. Furthermore, we tried to investigate the effect of TQ on experimental metastasis that involved the injection of breast cancer cells directly into the systemic circulation (Hwang et al., 2015).
Breast cancer has a greater tendency to spread to the bone marrow of the femora, tibiae, and the mandibular bones. Therefore, intracardiac injection method has been employed to develop osteolytic bone metastasis model (van der Horst and van der Pluijm, 2012; Kumar and Manjunatha, 2013;Lee et al., 2014). On the basis of the crucial role of SDF1/CXCR4 in breast cancer metastasis to femora, tibiae, and mandibles, we systemically injected via an intracardiac route MDA-MB-231-luc + cells to test the ability of TQ to modulate CXCR4 expression and thereby inhibit metastasis to the bones and other organs.

Cell Lines
Human breast cancer MCF7, MDA-MB-231, and BT-549 were purchased from ATCC (Manassas, VA, United States). The MDA-MB-231-luciferase expressing cell line was purchased from Cell Biolabs, United States. The cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and antibiotics penicillin/streptomycin.

Western Blotting
Vehicle or TQ-treated breast cancer cell lines and tumor tissues obtained from the mouse study were lysed in RIPA lysis buffer . Nuclear extracts of TQ treated or control cells were prepared using TransAM nuclear extract kit according to the manufacturer's instructions. Whole cell lysates was resolved on a 10% SDS polyacrylamide gel electrophoresis (SDS-PAGE) gel. The proteins were electro-transferred to a nitrocellulose membrane (Bio-Rad), blocked with Blocking One (Nacalai Tesque, Inc.), and probed with antibodies of interest overnight at 4 • C. The blot was washed with tris-buffered saline with 0.1% Tween-20, exposed to HRP-conjugated secondary antibodies for 1 h, and finally examined by chemiluminescence using Western Bright Sirius HRP substrate (Advansta). Images were captured using a ChemiDoc XRS+ imaging system and analyzed using Image Lab TM software (Bio-Rad, Hercules, CA, United States) as described previously (Chua et al., 2010). Densitometric analysis was done using ImageJ software.

NF-κB Reporter Assay
MDA-MB-231 and BT-549 cells were plated in 6-well plates with 1 × 10 4 cells per well in complete medium. The following morning, cells were transfected with an NF-κB reporter plasmid linked to a luciferase gene or with the dominant-negative IκBα (IκBα-DN) plasmid, co-transfected with a β-galactosidase plasmid (Promega, Madison, WI, United States). Transfections were done according to the manufacturer's protocols using Lipofectamine (Invitrogen, United States). At 24 h posttransfection, cells were treated with TQ (50 µM) for indicated time points and then washed and lysed in luciferase lysis buffer (Promega). Luciferase activity was measured using a Tecan plate reader (Durham, NC, United States) by using a luciferase assay kit (Promega) and was normalized to β-galactosidase activity. All luciferase experiments were done in triplicate as previously described (Shanmugam et al., 2011b).

NF-κB DNA-Binding Activity Assay
To determine NF-κB activation, nuclear extracts were prepared using a nuclear extraction kit (Active Motif, Carlsbad, CA, United States) according to the manufacturer's instructions. NF-κB DNA-binding activity was analyzed using the TransAM NF-κB p65 transcription factor assay kit (Active Motif, Carlsbad, CA, United States), following the manufacturer's instructions. The enzymatic product was measured at 450 nm with a microplate reader (Tecan Systems, San Jose, CA, United States) as previously described (Shanmugam et al., 2011b).

Chromatin Immunoprecipitation and Quantitative RT-PCR
Human MDA-MB 231 and BT-549 TNBC cells were seeded in 15-cm culture dishes and treated with TQ (50 µM) for 0, 2, 4, 6, and 8 h. At the end of the treatment period the cells were fixed with 1% formaldehyde at room temperature for 10 min and neutralized with glycine. The cells were collected, resuspended in CHIP lysis buffer, and sonicated (Vibra-Cell, TM Newtown, CT, United States). Samples were incubated with protein-G beads that had been pre-incubated with 4-10 µg of anti-NF-κB antibody (Cell Signaling Technology, MA, United States) or negative control IgG (Sigma-Aldrich Co., St. Louis, MO, United States). The next day, immunoprecipitants were washed by using washing buffer and reverse cross-linked at 65 • C. The DNA was then purified by using a PCR purification kit purchased from MACHEREY-NAGEL (Duren, Germany). Finally, the purified DNA was subjected to quantitative RT-PCR and data analyzed.

Scratch Wound Healing Assay
MDA-MB-231 and BT-549 cells were seeded in culture plate containing a culture-insert (Ibidi, Martinsried, Germany) until fully confluent. A cell-free gap of 500 µm was created after removing the culture-insert. After incubation with 50 µM TQ for 12 h, the medium was changed to medium with or without CXCL12. After migration for 8 or 24 h, the wound distance was observed and captured using bright field microscopy and the gap distance was measured using Photoshop software (Manu et al., 2011;Shanmugam et al., 2011b,c). MDA-MB-231 cells were transfected with 50 nmol/L of p65 or control siRNA. The cells were then subjected to wound healing assay either in the presence or absence of TQ (50 uM) for 8 h.

Invasion Assay
An in vitro cell invasion assay was performed using a BioCoat Matrigel invasion assay system (BD Biosciences, San Jose, CA, United States), as described previously (Manu et al., 2013;Shanmugam et al., 2011b,c). MDA-MB-231 cells were transfected with 50 nmol/L of p65 or control siRNA. The cells were then subjected to invasion assay either in the presence or absence of TQ (50 uM) for 8 h.

Determination of Tumor Growth Using a Chick Choriallantoic Membrane Assay
The chick chorioallantoic membrane (CAM) assay was modified from Sys et al. (2013). Briefly, fertilized chicken eggs (Bovans Goldline Brown) were purchased from Chew's Agriculture Pte Ltd., Singapore and placed horizontally in a 37.5 • C incubator with 70% humidity on embryonic day (ED)-0. On ED-3, a sharp weighted tool was used to poke a hole at the apex of the eggshell, and 3 mL of albumin was removed using a 5 mL syringe and 18G needle in order to drop the CAM. The sharp weighted tool was then used to poke a hole in the middle of the egg before using curved surgical scissors to cut a 1 cm 2 hole. The eggs were screened and dead embryos were removed. The hole was then sealed with a 1624W Tegaderm semi-permeable membrane and the egg placed back into the incubator. On ED-7, MDA-MB-231 (0.65 × 10 6 ) cells were mixed with matrigel. Fifty micro liter of the matrigel-cell mixture was placed on the CAM/egg. The hole was then re-sealed with the Tegaderm semi-permeable membrane. Twenty micro liter of DMSO or 25, 50, or 100 µM of TQ was added by pipetting onto autoclaved filter paper disks on ED-10 after the initial ultrasound scan. The tumor volume and tumor vascularity was determined at the 72 h time point in the control and TQ treated groups.

Ultrasound Imaging
On embryonic day 10, and after 72 h incubation with or without TQ, the Tegaderm membrane was removed and Aquasonic gel was added onto cling wrap that had been carefully placed over the CAM tumors. Using a VisualSonics Vevo 2100 Imaging system, a 550D transducer connected to a 3D acquisition monitor was used to obtain ultrasound images of the tumors formed on the CAM. Parallel 2D sections obtained were further reconstructed to form 3D images of the tumors. Tumor volumes and percentage of vasculature were calculated using the Vevo Lab 1.7.0 program. On ED-13, after ultrasound imaging, the CAM tumors, along with chick liver (to check for metastasis) were carefully excised and washed in PBS, part of it was snap frozen in liquid nitrogen for molecular analysis while the other part was fixed in 10% formalin overnight at 4 • C, before being embedded in paraffin. The paraffin blocks were then taken for future histopathological analysis.

Intracardiac Experimental Metastasis Murine Model
All procedures involving animals were reviewed and approved by the NUS Institutional Animal Care and Use Committee. Seven week old NCr-Foxn1nu, female mice were purchased from InVivos Pte Ltd., Singapore. NCr-Foxn1nu female mice were intracardially implanted with 1 × 10 6 MDA-MB-231-luc + cell line. Female nude mice (age 7 weeks) were anesthetized with 3-5% isoflurane. On day 0, anesthetized animals were injected with 1 × 10 6 MDA-MB-231-luc + cells suspended in 100 µl sterile Dulbecco phosphate buffered saline (DPBS) into the left ventricle of the heart by non-surgical means . D-Luciferin potassium salt (Gold Biotechnology, St. Louis, MO, United States) at room temperature was dissolved in DPBS without calcium or magnesium to a final concentration of 15 mg/ml. The luciferin solution was sterilized using a 0.22 micron filter and then injected intraperitoneally at a concentration of 150 mg/kg body weight (b.w). The mice were placed in an imaging chamber and imaged both on the dorsal and ventral sides whilst under isoflurane anesthesia (3-5%) using an IVIS TM Imaging System (Xenogen Corporation, United States), for 20 s-3 min as described previously (Rehemtulla et al., 2000;Jenkins et al., 2003a,b;Scatena et al., 2004).
A successful intracardiac injection was indicated on day 0 by images showing systemic bioluminescence distributed throughout the animal. Only mice indicating a satisfactory injection were included in the experimental groups. The mice were imaged once every week for tumor cell metastasis for a total of 4 weeks. Generally, two to three mice were imaged at a time. Imaging and quantification of signals were controlled by the acquisition and analysis software Living Image R (Xenogen R Corporation, United States). Tissues of interest were excised, placed into 24-well tissue culture plates with 300 µg/ml D-luciferin in DPBS and imaged for 1-2 min. Tissues were subsequently fixed in 10% formalin (Sigma, St. Louis, MO, United States) and prepared for standard histopathology evaluation.

Hematoxylin and Eosin Staining
At the end of treatment, mice hind limbs and fore limbs were collected and fixed in 10% neutral buffered formalin solution (Sigma-Aldrich, St. Louis, MO, United States). The limbs' muscle tissue was then removed to expose the bone. The bones were then decalcified by soaking in citric acid buffered formic acid (Sigma-Aldrich, St. Louis, MO, United States) for 96 h. The softened and decalcified bone was then processed and embedded into paraffin wax blocks. The bone samples were then sectioned and stained with hematoxylin and eosin solution (Merck, Germany). Images were taken using an Olympus BX51 microscope (magnification, 40×).

Immunohistochemical (IHC) Analysis of Tumor Samples
Breast cancer tumor cells that had metastasized to the bone, lung, and brain from control and drug-treated groups were fixed with 10% phosphate buffered formalin, processed, and embedded in paraffin. The bone and tumor sections, 5 µM in thickness, were cut and deparaffinized in xylene, dehydrated in graded ethanols, and finally hydrated in water. Antigen retrieval was conducted by boiling the slide in 10 mM sodium citrate (pH 6.0) for 30 min. Sections were incubated overnight with anti-CXCR4 primary antibody (1:100 dilutions). IHC was conducted following the manufacturer's instructions (Dako LSAB Kit) as described previously (Dai et al., 2015;Zhang et al., 2016). Images were taken using an Olympus BX51 microscope (magnification, 40×). Quantitative analysis of IHC images was performed by visual scores between the control and treated images. In this expression quantitation technique, each image was divided into four parts and each part was individually quantitated for the biomarker expression. A cell scored as positive refers simply to the presence of brown staining (peroxidase) in any part of the studied tissue. A negative cell scored refers to no staining or weak staining.

Statistical Analysis
Student t-test was used to analyze the data. For in vivo studies, unpaired t-test was used for statistical comparisons between groups. p < 0.05 was considered statistically . The cells were seeded in the top side of the chamber and matrigel was coated as a layer on the bottom side of the chamber. The cells were pretreated with or without TQ (50 µM) for 12 h and the cells were then seeded in the top chamber and placed in a 24 well plate containing either the basal medium only or basal medium containing 100 ng/ml CXCL12 for 24 h. The invaded cells were then fixed and stained with 1% crystal violet solution. Columns represent the number of cells invaded. Bars indicate standard error. * Indicates p-value < 0.05. (C,D) Effect of TQ on migration of MDA-MB-231 and BT-549 cells. A scratch wound healing assay was performed to evaluate the inhibitory effect of TQ on breast cancer cell migration. A chamber containing 500 micron size cell free width was measured on day 0 and after incubation with 50 µM TQ for 12 h, the medium was changed to medium with or without CXCL12 (100 ng/ml). After migration for 24 h, the gap distance of the wound was measured at three different sites. Bars indicate standard error. * Indicates p-value < 0.05.

RESULTS
The present study was designed to investigate the effect of TQ (with structure shown in Figure 1A) on both CXCR4 expression and function in BT-549 and MDA-MB-231 TNBC cells and in a metastatic murine model.

TQ Suppresses CXCR4 Protein Expression in Breast Cancer Cells
We first studied the effect of TQ on the expression of CXCR4 in MCF7 and MDA-MB-231 breast cancer cell lines. We observed that TQ could suppress the expression of CXCR4 in a dose-and time-dependent manner in MCF7 (Figures 1B,C, left panel) and MDA-MB-231 (Figures 1B,C, right panel), respectively.

Down-Regulation of CXCR4 Expression by TQ Is Not Mediated Through Its Degradation
Previous studies have shown that CXCR4 undergoes ubiquitination at its lysine residue, followed by degradation (Kukreja et al., 2005;Shanmugam et al., 2011b); hence, we next examined the likelihood that TQ may augment the rate of CXCR4 degradation through the stimulation of proteasomes. MDA-MB-231 and BT-549 cells were pretreated with ALLN, a proteasome inhibitor, for 1 h before treating with TQ. As shown in Figure 1D, ALLN did not protect TQ-induced degradation of CXCR4, thereby indication that this is an improbable basis for the downregulation of CXCR4 expression by TQ. In addition we also studied the capability of chloroquine, a lysosomal inhibitor, to abrogate TQ-induced degradation of CXCR4, as CXCR4 has been shown to undergo ligand-dependent lysosomal degradation (Shanmugam et al., 2011b). Both the cell lines were pretreated with chloroquine for 1 h before being treated with TQ. Our results indicated that chloroquine at 200 µM only marginally prevented the degradation of CXCR4 (Figure 1E), thereby suggesting that this was possibly not the main pathway for suppression of expression of CXCR4.

TQ Modulates Constitutive Activation of NF-κB in TNBC Cells
The NF-κB pro-survival signaling pathway plays a critical role in breast cancer cell progression and metastasis (Helbig et al., 2003). The promoter region of the CXCR4 gene also has several NF-κB binding sites (Shanmugam et al., 2011b). Therefore it is possible that TQ suppresses NF-κB activation and downregulate CXCR4 expression. We found that treatment with TQ suppressed NF-κB activation in a time-dependent manner (Figure 2A). This result suggests that TQ may down-regulate CXCR4 expression through inhibition of NF-κB activation. However, DNA binding alone is not always associated with NF-κB -dependent gene transcription, suggesting that added controlling steps are involved. Subsequent results also indicated that TQ inhibited NF-κB reporter activity in a time-dependent manner in TNBC cells ( Figure 2B). As both phosphorylation and nuclear translocation are critical for the function of transcription factors, we next evaluated the effect of TQ on nuclear translocation of p65 and phospho-p65 expression by Western blot analysis. We found that TQ dose-dependently suppressed nuclear accumulation of p65 and phospho-p65 expression in the nucleus of MDA-MB-231 cells ( Figure 2E).

TQ Abrogates CXCL12-Induced Over-Expression of CXCR4
Chemokine receptor type 4 is one of the most important chemokine receptors in TNBC. High CXCR4 expressing tumor cells have a greater invasive and metastatic potential, and CXCR4 + cells migrate following a concentration gradient of CXCL12. Overexpression of this receptor has been observed in patients who exhibit advanced stages of cancer and metastasis and correlates with worse prognosis and decreased patient survival (Kato et al., 2003). We next investigated if indeed TQ can suppress CXCL12 induced overexpression of CXCR4 in MDA-MB-231 cells. Interestingly, we found that TQ suppressed CXCL12 induced over-expression of CXCR4 in MDA-MB-231 cells thereby indicating that positive feedback loop between CXCR4 and downstream signaling components may be mediated in an autocrine manner ( Figure 2F).

TQ Attenuates NF-κB Transcriptional Activity by Decreasing Its Binding to the CXCR4 Promoter
Whether the down-regulation of CXCR4 by TQ in TNBC cells was due to suppression of NF-κB activation in vivo was examined by a ChIP assay targeting NF-κB binding in the CXCR4 promoter. We found that TQ suppressed NF-κB binding to the CXCR4 promoter (Figures 2C,D) in both the TNBC cell lines, thereby indicating that TQ inhibits CXCR4 expression by suppressing NF-κB binding to the CXCR4 promoter. According to the quantitative RT-PCR result, compared to the untreated samples, TQ significantly decreased the binding of NF-κB to the CXCR4 promoter and this reduction was noted to occur in a timedependent fashion (Figures 2C,D).

TQ Inhibits CXCL12-Induced Cellular Invasion and Migration
We further elucidated the effect on TQ on CXCL12-induced cell invasion. Using an in vitro invasion assay, TQ treatment suppressed CXCL12-induced invasion in both the cell lines (Figures 3A,B). Furthermore, we performed addition experiment by using p65 siRNA to inhibit the NF-κB signaling pathway and to determine the anti-invasive potential of TQ on MDA-MB-231 cells. Interestingly, we found that upon inhibition of NF-κB, TQ was not able to significantly mitigate cellular invasion ( Figure 3E). Whether the down-regulation of CXCR4 by TQ correlates with cell migration was also examined using an in vitro scratch wound healing assay. We found that both MDA-MB-231 and BT-549 cells migrated faster under the influence of CXCL12 and this effect was abolished on treatment with TQ (Figures 3C,D). Interestingly, we found that upon inhibition of NF-κB using p65 siRNA, TQ was not able to significantly attenuate cellular migration ( Figure 3F).

TQ Abolishes Tumor Growth and Vascular Volume in a CAM Assay Model
We evaluated the effect of TQ on the growth of tumors in a CAM assay model. We found that 3D ultrasound scanning of tumors indicated that TQ could inhibit the growth of tumors in a dosedependent manner. TQ at doses of 50 and 100 µM significantly inhibited the growth of tumors compared to the vehicle control group (Figure 4B), and potently reduced the tumor vascular volume at doses of 50 and 100 µM, thereby indicating complete abrogation of tumor vasculature by TQ ( Figure 4C). However, TQ at the lower dose of 25 µM had no effect on tumor growth and vascularity. Photographic images of CAM vehicle treated and TQ treated CAM tumors are shown in Figure 4A.

TQ Inhibits Breast Cancer Bone Metastasis and Thinning of Bones
To investigate whether TQ can suppress bone metastasis, 8 weeks old nude mice were injected intracardiac with MDA-MB-231luc + expressing cells and then treated with TQ 2 mg or 4 mg/kg b.w. intraperitoneally for 4 weeks. The mice were divided into three groups and the schematic of the TQ dosing regimen is shown in Figure 5B. Bioluminescence images showed a decrease in number of colonies metastasized to distant site organs compared to vehicle control animals ( Figure 5A). A significant dose-dependent, anti-metastatic effect was observed in the lung tissue ( Figure 5C) and in the brain tissue ( Figure 5D). High CXCR4 expressing MDA-MB-231 cells always tend to move to a region rich in CXCL12 such as bone marrow (Fernandis et al., 2004;Hung et al., 2014). By the third and fourth week we detected bone metastasis. At 4 weeks, metastatic activity was greater in control mice compared to TQ treated mice ( Figure 5E).
Remarkably, it was found that TQ significantly suppressed the metastasis of breast cancer cells to the bone (Figure 5E), and also there was a decrease in the metastatic sites such as bone marrow of the femora, tibiae, and mandibles. We observed a decrease in metastatic osteolytic lesions using H&E staining, and a significant decrease in the bone colonization of breast cancer cells (Figure 6B). A similar anti-metastatic effect was also observed in the lung ( Figure 5C) and brain tissues ( Figure 5D). In addition, no significant decrease in the body weight was observed both in the vehicle treated control and TQ treatment group (data not shown). H&E examination of the bone sections revealed prevention of thinning of bones in the TQ treated group as compared to control mice ( Figure 6B).

TQ Reduces the Expression of NF-κB Regulated Gene Products in the Tumor Tissues
NF-κB regulates the expression of several genes that are involved in tumor growth, metastasis, and angiogenesis (Li et al., 2015). A dose-dependent decrease in the expression of NF-κB-regulated proteins (CXCR4 and Ki-67) in the lung and brain tissues was observed by immunoblot analysis (Figure 6A). As shown in Figure 6C, TQ treatment inhibited CXCR4 expression in bone (top panel), lung (middle panel), and brain tissues (bottom panel), that was assessed by IHC analysis. A marked decrease in the expression of CXCR4 was observed in the femur of mice, and a similar effect was observed in the lung and brain tissues treated with TQ (4 mg/kg b.w.).

DISCUSSION
The current study was designed to investigate the anti-metastatic effect of TQ, and its underlying molecular mechanisms of action in TNBC cell lines in vitro and in an intracardiac metastatic murine model. CXCR4 over-expression has been associated with tumor cell proliferation, invasion, and metastasis. TQ was found to inhibit the expression of CXCR4 in MDA-MB-231 and MCF7 breast cancer cell lines, irrespective of the cell type and HER2 status. We further observed that TQ inhibited constitutive NF-κB DNA binding and activation in both the TNBC cell lines and significantly suppressed NF-κB dependent luciferase expression. These results are in line with previous work showing that TQ can abrogate NF-κB activation pathway (Sethi et al., 2008a;Siveen et al., 2014b) in diverse cells. Interestingly, TQ mediated down-regulation of CXCR4 did not occur by proteolytic degradation of the receptor but rather by a decrease in the number of transcripts. In addition, inhibition of the CXCR4 receptor was associated with down-regulation of migration and invasion of TNBC cells upon stimulation with its ligand CXCL12. Numerous reports have previously suggested that CXCR4 is over-expressed in a variety of tumors including gastric, ovarian, pancreatic, melanoma, renal, cervical, colon, and hematological malignancies (Manu et al., 2011(Manu et al., , 2013Shanmugam et al., 2011b). However, there are no substantial studies elaborating upon the detailed mechanism(s) of CXCR4 over-expression in tumor cells (Kulbe et al., 2005).
Genetic mutations in von Hippel Lindau tumor suppressor gene (Staller et al., 2003), PAX3-and PAX7-FKHR gene fusion (Libura et al., 2002), angiogenic growth factors (Bachelder et al., 2002;Schioppa et al., 2003), deregulated NF-κB (Zlotnik, 2006), and pro-inflammatory cytokines (Kulbe et al., 2005) have all been implicated in CXCR4 over-expression. Recent evidence implicates that EGFR, c-erbB2 and its encoding gene HER2/neu, can also regulate CXCR4 expression at the post-transcriptional level in breast cancer cells (Benovic and Marchese, 2004;Arya et al., 2007). Moreover, studies have implicated CXCR4 over-expression to cancer metastasis and are an indicator of poor prognosis and overall patient survival (Xu et al., 2015). A high level of expression of CXCR4 was significantly associated with a greater risk of bone metastasis. Bone metastasis occurs significantly more often in CXCR4 positive patients (13.1%) as compared to CXCR4 negative patients (2.4%; P = 0.008) (Hung et al., 2014). In a recent article, Ashour et al. (2014Ashour et al. ( , 2016 demonstrated that TQ can also inhibit NF-κB phosphorylation and its regulated gene products, such as interleukin-8 and chemokines, in a dose-and time-dependent manner in hepatocellular carcinoma cells. Therefore, CXCR4 appears to be a novel therapeutic target for the treatment of metastatic breast cancer. It has also been suggested that CXCR4 may be degraded by ligand-dependent lysosomal degradation (Fernandis et al., 2004) and by atrophin-interacting protein 4 mediated polyubiquitination and subsequent degradation (Bhandari et al., 2007).
CXCL12 is the only ligand for CXCR4 and acts as autocrine/paracrine growth factor for several cancers.
Interestingly, we found that CXCL12 was able to increase the levels of CXCR4 in TNBC cells. In a recent study, it has been reported that the overexpression of both CXCR4 and CXCL12 mRNA in glioblastoma cancer stem cells (GBM-CSC) can control proliferation, invasion and angiogenesis. CXCL12 secretion from CXCR4 expressing cells supports the occurrence of autocrine/paracrine mechanism in GBM CSCs, in human meningioma and pituitary adenoma cells (Gatti et al., 2013). Furthermore, in another study by Borrello et al. (2005), it was reported that transforming oncogene RET/PTC1 in human papillary thyroid carcinoma can induce the expression of CXCL12 and its receptor CXCR4, thus triggering autocrine proliferation of thyroid carcinoma cells (Borrello et al., 2005). Our results also demonstrate that TQ can substantially suppress CXCR4 expression in TNBC cells and this modulation was observed to occur at the transcriptional level and not mediated by either ligand-dependent lysosomal or proteasomal degradation. Because TQ was observed to modulate inhibit both NF-κB DNA binding and reporter gene expression, we further investigated the effects of TQ on CXCL12-induced invasion and migration in TNBC cell lines. We found that pre-incubation with TQ completely blocked CXCL12-induced migration and invasive potential of TNBC cells. Furthermore, TQ failed to abrogate inhibit invasion or migration of cells in p65 deleted cells.
We also investigated for the first time the potential effect of TQ on breast cancer cell bone metastasis in a breast cancer mouse model. CXCL12 signaling via CXCR4 has an important physiological role in hematopoietic stem cell homing to the bone marrow (Balkwill, 2004b;Xu et al., 2015). This signaling axis also plays a critical part in breast cancer cell metastasis by increasing chemotaxis to regions of high CXCL12 expression such as bone, lung, and brain (Arya et al., 2007). Therefore, down-regulation of CXCR4 in triple negative metastatic breast cancer cells could greatly decrease their metastatic potential. In our study we also determined the anti-cancer effects in a tumor-bearing chick embryo model. The CAM assay has been used extensively as a model for studying angiogenesis due to the accessibility and visibility of the developing blood vessels (Siveen et al., 2014a). However, recent studies have shown that tumor cells can grow readily on the CAM in an immunocompromised environment (Sys et al., 2013).
Tumor bearing chick embryos have been used to study anti-cancer effects of drugs (Abe et al., 2011;Busch et al., 2013). In our study, we implanted breast cancer cells onto the CAM and evaluated the effects of TQ on tumor growth and vascularity. We found that TQ dose-dependently inhibited tumor growth and potently reduced the tumor vascular volume. Furthermore, we found that intraperitoneal administration of TQ at doses of 2 or 4 mg/kg b.w for 4 weeks could inhibit the growth and metastasis of breast cancer tumors. TQ significantly suppressed lung, bone, and brain metastasis, as evidenced by a decrease in bioluminescence signals in these organs. It is of note that patients with terminal breast cancer exhibit severe bone destruction associated with osteolytic lesions (Mundy, 2002). We further evaluated the inhibitory effect of TQ on osteolytic bone metastasis in nude mice inoculated with MDA-MB-231-luc + cells into the left cardiac ventricle. Bone metastasis was found to be significantly suppressed in the TQ treated group, and the number of osteolytic lesions in the mandibles, femora, and tibiae were markedly decreased, as indicated by bioluminescence imaging. Western blot analysis of tumors obtained from the lung and brain also showed a decrease in the expression of CXCR4 and Ki67. In fact, upon TQ treatment, there was a marked reduction in thinning of bones, as evidenced by H&E staining. In addition, IHC analysis of lung, brain, and bone tumors revealed inhibition of CXCR4 expression. Overall, TQ was well tolerated at the doses tested and was effective in suppressing the metastasis of TNBC cells. In conclusion, TQ inhibited NF-κB regulated CXCR4 expression, migration, as well as invasion of the TNBC cells, and markedly suppressed bone metastasis by suppressing NF-κB activation cascade. Further studies are warranted as TQ as its clinically relevant doses are yet to be determined.

AUTHOR CONTRIBUTIONS
GS, APK, KHBT, LL, RH, KSA, AC, TA, MS, and SA conceived the project. GS, APK, KHBT, LL, RH, and KSA designed the experiments. MS, AH, CW, YY, and APK carried out the experiments and analysis the data. MS, GS, AKo, APK, KSA, and FA wrote and corrected the paper.

FUNDING
This work was supported by NUHS basic seed grant to KHBT. APK was supported by grants from National Medical Research Council of Singapore, NCIS Yong Siew Yoon Research Grant through donations from the Yong Loo Lin Trust. APK and RH were also supported by the National Research Foundation Singapore and the Singapore Ministry of Education under its Research Centers of Excellence initiative to Cancer Science Institute of Singapore, National University of Singapore.