Purpose:

Fibroblasts expressing the orphan chemokine CXCL14 have been previously shown to associate with poor breast cancer prognosis and promote cancer growth. This study explores the mechanism underlying the poor survival associations of stromal CXCL14.

Experimental Design:

Tumor cell epithelial-to-mesenchymal transition (EMT), invasion, and metastasis were studied in in vitro and in vivo models together with fibroblasts overexpressing CXCL14. An approach for CXCL14 receptor identification included loss-of-function studies followed by molecular and functional endpoints. The clinical relevance was further explored in publicly available gene expression datasets.

Results:

CXCL14 fibroblasts stimulated breast cancer EMT, migration, and invasion in breast cancer cells and in a xenograft model. Furthermore, tumor cells primed by CXCL14 fibroblasts displayed enhanced lung colonization after tail-vein injection. By loss-of function experiments, the atypical G-protein–coupled receptor ACKR2 was identified to mediate CXCL14-stimulated responses. Downregulation of ACKR2, or CXCL14-induced NOS1, attenuated the pro-EMT and migratory capacity. CXCL14/ACKR2 expression correlated with EMT and survival in gene expression datasets.

Conclusions:

Collectively, the findings imply an autocrine fibroblast CXCL14/ACKR2 pathway as a clinically relevant stimulator of EMT, tumor cell invasion, and metastasis. The study also identifies ACKR2 as a novel mediator for CXCL14 function and thereby defines a pathway with drug target potential.

See related commentary by Zhang et al., p. 3476

Translational Relevance

Autocrine fibroblast CXCL14/ACKR2 signaling is shown to induce EMT, migration, and metastasis and to correlate with worse survival in patients with breast cancer. The identification of ACKR2 as a novel component in the signaling of the orphan chemokine CXCL14 is relevant for further biomarker studies and suggests novel targeting opportunities.

Death of patients with breast cancer is almost exclusively due to metastatic disease. Metastasis develops through a multistep process, involving tissue invasion, intravasation, survival in the bloodstream and lymph system, extravasation, and tissue colonization. This study develops recent correlative studies that have indicated clinical relevance, in breast cancer, of stroma-derived expression of the chemokine CXCL14 by demonstrating significant and independent associations between high stromal CXCL14 expression and shorter survival in a population-based breast cancer cohort. Notably, epithelial expression of CXCL14 did not have an impact on breast cancer outcome (1).

During the early steps of metastasis development, tumor cells lose cell-to-cell contacts and epithelial characteristics and instead gain mesenchymal traits that allow them to invade the surrounding tissue and metastasize; a process termed epithelial-to-mesenchymal transition (EMT; ref. 2).

EMT is controlled by distinct transcriptional programs activated by specific transcription factors, including Snail, Slug, Twist, Zeb, and Gsc. Activation of these factors ultimately leads to the loss of epithelial markers including E-cadherin and cytokeratins, and the upregulation of mesenchymal markers, such as vimentin, alpha-smooth muscle actin (αSMA), and matrix-degrading enzymes (3). Although the classical paradigm attributing EMT a crucial role in the process of metastasis has been recently challenged by studies in genetic mouse models, other recent studies including in vivo imaging approaches demonstrated that cancer cells displaying an EMT phenotype give rise to metastases (4–6).

Induction of EMT can occur in a paracrine manner by secreted factors from cells of the tumor stroma, as for example, the cancer-associated fibroblasts (CAF; refs. 7, 8). CAFs constitute a heterogeneous cell population that contributes to cancer growth and spread by secretion of a variety of protumorigenic factors, including soluble factors. Among these CAF-secreted factors implicated in EMT and metastasis are chemokines, proteins of a size between 8 and 14 kDa that stimulate directed cell migration by creating a gradient along which cell types expressing the corresponding receptor travel (7, 9). Chemokines bind to the pertussis-sensitive Gαi-subfamily of G-protein–coupled receptors (GPCR) that engage different signaling pathways including ERK1/2, PI3K/AKT, and calcium signaling (10). Besides the classical chemokine receptors, there is a subfamily of atypical chemokine receptors (ACKR) that are predominantly involved in sequestration of chemokines (11).

In cancer, chemokines are involved in the recruitment of various cell types into tumors and thereby affecting inflammation, angiogenesis, tumor growth, invasion, and metastasis (12). A paracrine chemokine cross-talk between stromal cells and tumor cells, involving effects on EMT, has been demonstrated to enhance formation of metastases (13–16). Expression of certain chemokines in distant tissues has also been reported to determine metastasis formation in specific organs, a process termed metastatic tropism (17).

The orphan chemokine CXCL14, earlier designated BRAK, MIP-2γ, BMAC, or KS1 stimulates migration of various immune cells, including B cells, NK cells, and monocytes, but not T cells (18–21). CXCL14 expression has been shown to be upregulated in CAFs, as compared with normal fibroblasts, in human breast and prostate cancer (22, 23). Experimental studies exploring the function of CXCL14 during tumor progression and metastasis formation have demonstrated context-dependent pro- and antitumoral effects. The reasons for these effects remain largely unknown and could possibly dependent on the cell type that express the chemokine and on the profile of chemokine receptors and ACKRs expressed. Some studies with CXCL14 overexpression in tumor cells have demonstrated antitumoral effects of this chemokine (24). In contrast, tissue culture and mouse cancer model studies of breast and prostate cancer have demonstrated protumoral effects of CXCL14 expressed by stromal fibroblasts, associated with CXCL14-induced changes in fibroblast secretomes (22, 23, 25).

The tumor-promoting roles of CAF-derived CXCL14 have been shown to depend on nitric oxide synthase 1 (NOS1) and involve stimulation of angiogenesis and recruitment of macrophages (25).

Continued exploration of the roles of CXCL14 in tumor biology and possible exploitation of this chemokine as a therapeutic target depend on the identification of critical mediators of CXCL14 signaling, including receptors. This study therefore aimed at providing a better understanding of the molecular mechanism underlying the documented poor survival association of stroma-derived CXCL14.

Cell lines and chemicals

The mouse fibroblast cell line NIH3T3 (and derivatives), the breast cancer cell lines MCF7, SKBR3 MDA-MB-231, 4T1, and Hs578t were cultured in DMEM (Hyclone), supplemented with 10% FBS (Hyclone), 1% glutamine (Hyclone) and 1% penicillin/streptomycin (Hyclone). DMEM-F12 (Hyclone), supplemented with horse serum (Biochrom), 1% glutamine, and 1% penicillin/streptomycin was used for culturing the MCF10-DCIS cell line. Starvation was performed in medium without serum. All cells were maintained at 37°C in humidified air with 5% CO2. NIH-ctr and NIH-CXCL14 fibroblasts have been characterized earlier, and fibroblasts secrete physiologic levels of CXCL14 (22). Cell lines were purchased from ATCC or received from collaboration partners and continually tested for Mycoplasma infection during the study. The identity of the cell lines used was confirmed by short tandem repeat (STR) profiling at Uppsala Genome Center. All experiments were performed with cells of passage 3–20.

Western blot analyses used antibodies against p-ERK (#9101), ERK (#9102), E-cadherin (#3195), Snail (#3879), NOS1 (#4234) (Cell Signaling Technology), β-actin (A1978), and α-tubulin (T5201; Sigma-Aldrich). Primary antibodies detecting E-cadherin (#3195), Cytokeratin 8/18 (#4546), and PDGFβR (#3169; Cell Signaling Technology) together by fluorescent-linked secondary antibodies, were used for immunofluorescence staining of xenograft tumors.

Pertussis toxin was purchased from Sigma-Aldrich and recombinant CXCL14 and CCL5 from R&D Systems and PeproTech.

Alexa Fluor 647–coupled chemokines were purchased from Almac.

RNA isolation, cDNA synthesis, and qRT-PCR analysis

RNA was isolated from xenograft tumors or overnight starved cells using GeneElute Mammalian Total RNA Miniprep Kit (Sigma-Aldrich). cDNA synthesis was performed with the SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen), using PolydT primers, in accordance with the manufacturer's instructions. The qRT-PCR reaction using SYBR Green Universal PCR Master Mix (Applied Biosystems) was performed with the 7500 Real-Time PCR System (Applied Biosystems). The concentration of primers was 200 nmol/L, and expression levels were normalized to the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The primer sequences for primers obtained from Sigma-Aldrich are shown in Supplementary Table S1. Other primers were QuantiTect primers obtained from Qiagen.

Immunoblotting, immunofluorescence, and secretome analyses

Immunoblotting and immunofluorescence analyses were performed as described previously (22). In short, for analysis of CXCL14-induced p-ERK signaling by immunoblotting, overnight serum-starved cells were stimulated with recombinant CXCL14 (R&D Systems or PeproTech) for 7 minutes. For experiments with pertussis toxin, cells were treated with the toxin for 1 hour at 37°C and 5% CO2 prior to CXCL14 stimulation. Cell lysates were prepared and SDS/PAGE was performed followed by transfer to polyvinylidene difluoride membranes (Millipore). Immunoblotting with p-ERK and ERK antibodies (Cell Signaling Technology), diluted 1:1,000, were performed and signals were detected with ImageQuant LAS4000 (GE Healthcare) and quantified using ImageJ (http://imagej.nih.gov.proxy.kib.ki.se/ij).

Analysis of EMT markers was performed for 48–72 hours after stimulation with fibroblast-conditioned medium or coculture of breast cancer cells and fibroblasts at a 1:1 ratio. The conditioned medium was generated by seeding the same number of NIH-ctr and NIH-CXCL14 cells. The next day, medium was changed to DMEM containing low serum (1% FBS) and the conditioned medium was collected for 48 hours, sterile filtered with a 0.2-μm filter. Antibodies for immunoblotting, including E-cadherin and Snail, were diluted 1:1,000 and NOS1 1:500. For immunofluorescence analysis, E-cadherin, Cytokeratin 8/18, and PDGFβR antibodies were diluted 1:100.

To analyze the secretome from CXCL14 fibroblasts, a protein array was performed. A total of 7.0 × 105 NIH-CXCL14 and NIH-ctr cells were seeded in 10-cm dishes. The next day, medium was changed from medium with 10% FBS to low serum (1% FBS). Conditioned medium was collected after 48-hour incubation at 37°C, sterile filtered, and stored at −20°C. Aliquots of conditioned medium (1 mL) from NIH-CXCL14 and NIH-ctr fibroblasts were subjected to the proteome profiler (mouse angiogenesis array kit ary015, R&D Systems), performed according to the manufacturer's protocol. Array images obtained were analyzed using the ImageJ software. For all spots, the average background signal from negative control spots was subtracted. The average signal from positive control spots of each membrane was used to normalize the two different types of fibroblast conditions. The relative differences in protein expression between NIH-CXCL14 and NIH-ctr cells were expressed as ratio (fold of NIH-ctr).

In vitro growth, migration, and invasion assays

To study the effect of siRNA and short hairpin RNA (shRNA)-mediated knockdown of ACKR2 on growth of NIH-ctr and NIH-CXCL14 fibroblasts, 2 × 104 cells were seeded per well of a 24-well plate (Sarstedt), in quadruplicates in serum-reduced media. After 3 days of culture, AlamarBlue (Bio-Rad) was used to determine the cell number. A total of 350 μL of a 1:10 dilution of AlamarBlue dye in DMEM was added to each well and cells were incubated at 37°C with 5% CO2 for 2.5 hours. To measure the conversion of the dye, 100 μL was transferred into each well of a white 96-well plate (Costar) and absorbance was measured at a wavelength of 570 nm.

To study cell migration of breast cancer cells, a transwell migration assay was used. A total of 2.5 × 104 breast cancer cells were seeded in transwell inserts (Corning) with an 8.0-μm pore-sized membrane and placed in a 24-well plate (Corning), in duplicates. For analysis of CXCL14 fibroblast–induced migration, 2.5 × 104 CXCL14- or control fibroblasts were seeded in the bottom chamber.

For migration experiments, lasting for 16–24 hours, the inside of the insert was wiped with cotton swabs, washed with PBS, and the cells were fixed in ice-cold methanol. The membrane was cut out and placed on Superfrost Plus slides (Menzel-Gläser) and stained with Vectashield Mounting Medium with DAPI (Vector Laboratories). Quantification of cell migration was performed by counting cell nuclei of the migrated cells. The same principles were used for the invasion assays, with inserts containing a thin Matrigel layer (Corning). Invasion was allowed to occur for 72 hours.

Chemokine binding assay

HEK-293 cells or HEK-293 cells stably expressing ACKR2 (under 200 μg/mL hygromycin selection) were distributed into 96-well plates (2 × 105 cells per well) and incubated with increasing concentrations ranging from 10 pmol/L to 100 nmol/L of Alexa Fluor 647–labeled CCL5 or CXCL14 for 90 minutes on ice. After a washing step, Zombie Green Fixable Viability Kit (BioLegend) was used to gate on living cells only. The experiments were performed in PBS containing 1% BSA and 0.1% NaN3 (FACS buffer). Chemokine binding was quantified by mean fluorescence intensity (MFI) of 10,000 gated cells on a BD FACS Fortessa Cytometer (BD Biosciences).

β-arrestin recruitment assay

Chemokine-induced β-arrestin-1 recruitment to ACKR2 was monitored by Nanoluciferase complementation assay (NanoBiT, Promega) as described previously (26, 27). A total of 5 × 106 HEK cells were plated in 10-cm culture dishes and 24 hours later transfected with plasmids containing human β-arrestin-1 N-terminally fused to LgBiT and ACKR2 C-terminally fused to SmBiT. Twenty-four hours posttransfection, cells were harvested, incubated 30 minutes at 37°C with 200-fold diluted Nano-Glo Live Cell substrate, and distributed into a white 96-well plate (1 × 105 cells per well). β-arrestin-1 recruitment was measured over 25 minutes with a Mithras LB940 luminometer (Berthold Technologies). For each experiment, signal measured with a saturating concentration (300 nmol/L) of the full agonist (i.e., CCL5) was set as 100%.

Bioinformatic analyses

Sequence analyses for novel chemokine receptors started from Pfam family PF00001 (7tm_1), using the 1679 human domains from 1,664 human sequences in the full alignment. A neighbor-joining tree of these was built using scoredist distances with Belvu (28). A subtree of 114 cytokines was cut out. After removing fragment sequences and >99% identical sequences, 32 sequences were left. Following exclusion of DUFFY, not being part of the 23,760 homologs in the Pfam family, a candidate list of 31 candidates was established (Supplementary Fig. S1).

Transfection of siRNA and generation of stable cell lines

For transfection of siRNA, 1.0–2.0 × 105 cells were seeded in 6-well plates (Sarstedt) and transfected for 48 hours with 100 nmol/L siRNA (ACKR2 target sequence; 5′-CTCAATTAGCGTTATTGGCAA-3′; Qiagen) using HiPerFect transfection reagent (Qiagen). To determine the efficiency of siRNA-mediated knockdown of genes of interest, RNA extraction, cDNA synthesis, and qRT-PCR analysis was performed.

Fibroblast derivatives with stable knockdown of ACKR2 were established using shRNA procedures, as described previously (25). In brief, phoenix cells were transfected with 2 unique 29mer shRNA constructs against ACKR2 (gene ID = 59289; shACKR2:A and shACKR2:B) or nontargeting control shRNA (shCtr) in retroviral vectors (Origene). After 48 hours, the supernatant was collected, filtered, and added to NIH-ctr and NIH-CXCL14 cells for 5 hours. Cells were subsequently selected in 30 μg/mL blasticidin for 2 weeks. The knockdown of ACKR2 was confirmed by qRT-PCR. Generation of NIH-ctr and NIH-CXCL14 fibroblasts with a stable knockdown of NOS1 have been described previously (25).

Animal experiments

The animal experiments were conducted in accordance with national guidelines and approved by the Stockholm North Ethical Committee on Animal Experiments. The generation of xenograft tumors was performed as described previously (25).

The analyses of lung metastasis formation were performed after tail-vein injection of 2 × 105 breast cancer cells in three groups of 8-week-old female SCID mice, without any further randomization. The sample size of 10 mice in each group was determined by the 3R criteria together with previous experience. After 4 weeks, mice were sacrificed and lungs were collected, washed in PBS (Hyclone), and snap frozen (for qRT-PCR analysis) or embedded in optimal cutting temperature medium (OCT) and snap frozen (for histologic analysis). No animals were excluded from the study.

For RNA extraction, 1 mL of TRIzol (Life Technologies) was added to the lung tissue and homogenized with a polytron 3 times, each for 10 seconds. A volume of 0.2-mL chloroform was added and the samples were shaken for 15 seconds and incubated for 2 minutes at room temperature. The samples were centrifuged at 12,000 × g for 15 minutes at 4°C and the aqueous phase was placed in a new tube. RNA was precipitated by addition of 0.5 mL 100% isopropanol (Merck) to the samples and incubation for 10 minutes at room temperature. Following centrifugation at 12,000 × g for 10 minutes at 4°C, the RNA pellet was washed with 75% ethanol and air dried before resuspension in nuclease-free water (Ambion). The RNA concentration of each sample was determined using the Nanodrop ND-1000 spectrophotometer (NanoDrop Technologies). cDNA were synthetized and subsequent qRT-PCR analysis was performed as described previously, with human- and mouse-specific primers. Percentages of human cells in mice lungs were quantified as described in Malek and colleagues (29).

Ten frozen sections (10 μm) were made from the OCT-embedded lungs. Five sections were thrown in between each saved frozen section. Stainings were performed with the human-specific antibody Stem121 (Takara Bio) and the positive cells were counted in each section and results are displayed as average number per lung. The tail-vein injections were performed blinded and the analyses of lung metastasis were performed unblinded.

Clinical cohorts

The relation between CXCL14 transcript abundance and EMT as assessed by an EMT gene expression signature was investigated in clinical breast cancer cohorts with publicly available transcriptome data: the Uppsala (30), Stockholm (31), Rotterdam (32), and METABRIC (33) cohorts, as well as The Cancer Genome Atlas (TCGA; ref. 34). Each study site in METABRIC is treated as a separate cohort. TCGA gene expression datasets for breast cancer, ovarian cancer, and prostate cancer was used to investigate the levels of EMT markers and survival associations in CXCL14/ACKR2 subgroups was additionally performed in gene expression datasets for bladder cancer, clear cell renal cell carcinoma (ccRCC), colorectal cancer, esophageal cancer, glioblastoma multiforme (GBM), low-grade glioma, head and neck cancer (HNCC), lung adenocarcinoma, pancreas cancer, and stomach cancer (34).

Gene expression data analysis

An EMT gene expression signature score was derived for each tumor in the panel of clinical cohorts as described previously (35). In brief, the signature was identified from the changes in gene expression shared by upregulation of Gsc, Snail, Twist, and TGFβ1 and by downregulation of E-cadherin. PAM50 intrinsic subtype classification was performed as described previously (36). Individual patient data meta-analysis (IPDMA) of all cohorts with a linear mixed-effects model was performed using the R package nlme [R syntax lme(groupedData(ScaledEMTscore ∼ scaledCXCL14 | cohort, data))]. All gene expression data analysis was performed in R/Bioconductor and SPSS 21.0. The EMT score and CXCL14 data are first centered and scaled to unit SD within cohorts to facilitate comparison between cohorts. Hence, the slope of a linear regression in each cohort in Supplementary Fig. S2A is mathematically equivalent to the Pearson correlation coefficient. This equivalence does not hold in subgroups as in Supplementary Fig. S2B.

The TCGA gene expression data (Supplementary Figs. S3 and S4) are displayed as Z-scores obtained from cBioportal. For the breast cancer gene expression dataset, CXCL14low and CXCL14high groups were divided according to the 50 percentile, whereas ACKR2low and ACKR2high expression was determined by fitting a mixture of two normal distributions using the R package mixtools, resulting in a dichotomization below and above the 90.7 percentile. For the TCGA ovarian and prostate cancer gene expression datasets, the ACKR2low and ACKR2high subgroups were divided by the 4th quartile.

Statistical analysis

Statistical calculations were performed using Excel 2011 for Mac (Microsoft Office), R/Bioconductor, or the statistical package SPSS 21.0 (SPSS Inc.). All data are expressed as mean or median values, and error bars represent the SD or SEM. Data that are being statistically compared, relevant for the conclusions, exhibit similar variation. Statistical differences between groups were determined using two-sided, unpaired Student t test or Mann–Whitney U test. Pearson correlation was used to analyze correlations between different parameters. The Kaplan–Meier method and log-rank test method was performed to estimate overall survival. Cox proportional hazards model was used to compare HRs in both uni- and multivariate analyses. The multivariate analysis included known clinical relevant parameters, including T-stage, N-stage, M-stage, and molecular subtypes of breast cancer. Results are presented in the multivariate analysis as HRs including 95% confidence intervals (95% CI). For all analyses, P values below 0.05 were considered significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001). All relevant data are available from the authors.

CXCL14 fibroblasts induce loss of epithelial markers and enhance expression of mesenchymal markers and EMT transcription factors in breast tumor xenografts

The findings of stromal CXCL14 as a poor prognostic marker in breast cancer (1), prompted analyses of the potential effects of CAF-derived CXCL14 on tumor cell invasion and metastasis.

We first studied the expression levels of EMT markers as an indicator of a proinvasive phenotype in xenograft tumors formed following coinjection of the epithelial breast cancer cell line MCF7 and either control fibroblasts or CXCL14 fibroblasts (25).

Immunofluorescence staining of xenograft tumor sections demonstrated a significant loss of tumor cell E-cadherin and Cytokeratin 8/18 in CXCL14 breast tumors, as compared with control tumors (Fig. 1A; Supplementary Fig. S5A; Supplementary Table S2). NOS1, an oxidative stress–induced enzyme, was previously shown to functionally contribute to the protumorigenic actions of CXCL14-expressing fibroblasts (25). EMT markers were therefore also analyzed in MCF7/NIH-CXCL14 tumors with a stable NOS1 downregulation in the fibroblasts. Impaired expression of NOS1 was sufficient to reverse the decrease in epithelial markers induced by CXCL14 fibroblasts in vivo (Fig. 1A; Supplementary Fig. S5A; Supplementary Table S2). Furthermore, CXCL14 fibroblasts also reduced cancer cell Cytokeratin 8/18 levels, in a NOS1-dependent manner, in a xenograft coinjection model of prostate cancer (Supplementary Fig. S5B; Supplementary Table S2).

Figure 1.

CXCL14 affects regulators of EMT and invasion in a xenograft tumor model of breast cancer. A, Xenograft tumors of MCF7 cells coinjected with NIH-ctr or NIH-CXCL14 fibroblasts were stained for E-cadherin. White arrowheads indicate epithelial cells with weak E-cadherin expression. Scale bar, 50 μm. B, qRT-PCR analysis of transcript levels of genes encoding EMT-regulated markers in xenograft tumors. The analysis comprises epithelial marker (E-cadherin, Cytokeratin 18, and Cytokeratin 8), mesenchymal marker [Vimentin, α-SMA (encoded by ACTA2), and MMP2], and the EMT transcription factors Slug and Twist (n = 5). C, Staining of xenograft tumors formed following coinjection of MCF7 cells and NIH-ctr or NIH-CXCL14 fibroblasts with the human-specific antibody Stem121. Budding cells (cluster of up to three cells) in the border of the tumor were counted in 10 vision fields, and results are shown as mean number (nr) of cells/vision field (n = 5). Scale bar, 100 μm. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Error bars, SEM.

Figure 1.

CXCL14 affects regulators of EMT and invasion in a xenograft tumor model of breast cancer. A, Xenograft tumors of MCF7 cells coinjected with NIH-ctr or NIH-CXCL14 fibroblasts were stained for E-cadherin. White arrowheads indicate epithelial cells with weak E-cadherin expression. Scale bar, 50 μm. B, qRT-PCR analysis of transcript levels of genes encoding EMT-regulated markers in xenograft tumors. The analysis comprises epithelial marker (E-cadherin, Cytokeratin 18, and Cytokeratin 8), mesenchymal marker [Vimentin, α-SMA (encoded by ACTA2), and MMP2], and the EMT transcription factors Slug and Twist (n = 5). C, Staining of xenograft tumors formed following coinjection of MCF7 cells and NIH-ctr or NIH-CXCL14 fibroblasts with the human-specific antibody Stem121. Budding cells (cluster of up to three cells) in the border of the tumor were counted in 10 vision fields, and results are shown as mean number (nr) of cells/vision field (n = 5). Scale bar, 100 μm. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Error bars, SEM.

Close modal

Analyses of mRNA levels of EMT markers substantiated these findings and uncovered a reduction of epithelial markers including E-cadherin (CDH1), Cytokeratin 18 (KRT18), and Cytokeratin 8 (KRT8), and increase in mesenchymal markers including Vimentin (VIM), α-SMA (ACTA2), and MMP2 (MMP2), and an increase in EMT transcription factors including Slug (SNAI2), and Twist (TWIST1) in CXCL14-breast tumors (Fig. 1B). Furthermore, analyses of MCF7/NIH-CXCL14 tumors with a stable NOS1 knockdown in fibroblasts demonstrated NOS1 dependency of these gene expression changes (Fig. 1B). However, mesenchymal markers, including Fibronectin (FN1), FAP (FAP), and MMP9 (MMP9) were upregulated in CXCL14 tumors independently of NOS1 expression (Supplementary Fig. S5C).

Additional analyses were performed to investigate whether the CXCL14-dependent EMT phenotype also was associated with a more invasive growth pattern. As shown in Fig. 1C, MCF7/NIH-CXCL14 tumors displayed a more invasive growth pattern with a significantly higher number of budding cells in the tumor periphery, as compared with MCF7/NIH-ctr tumors.

These data collectively demonstrate that cancer cells coinjected with CXCL14 fibroblasts exhibit enhanced EMT and invasion in vivo.

CXCL14 fibroblasts stimulate EMT in vitro and induce a mesenchymal morphology of breast cancer cells

Next, the direct impact of CXCL14-expressing fibroblasts on the modulation of EMT markers in MCF7 breast cancer cells and MCF10-DCIS (ductal carcinoma in situ, DCIS) cells was investigated under in vitro coculture conditions. Western blot analysis demonstrated a reduction of E-cadherin in both MCF7- and DCIS cells after direct coculture with CXCL14 fibroblasts that was not seen with control fibroblasts (Fig. 2A). To confirm that the downregulation occurred in the breast cancer cells and did not reflect changes in fibroblast properties or abundance, immunofluorescence costaining for E-cadherin or Cytokeratin 8/18 together with the fibroblast marker PDGFRβ was performed on MCF7 fibroblast cocultures. As shown in Fig. 2B, there was a specific loss of E-cadherin (left) and Cytokeratin 8/18 (right) in the breast cancer cells when MCF7 cells were cocultured with CXCL14 fibroblasts, but not in the presence of control fibroblasts. Similar findings of altered EMT markers in MCF7- and DCIS cells were detected after treatment with conditioned media (CM) from CXCL14 fibroblasts (Supplementary Fig. S6A and S6B). Another breast cancer cell line, SKBR3, which has lost E-Cadherin, showed increased levels of the EMT transcription factor Snail when treated with CM from NIH-CXCL14 compared with CM from NIH-ctr (Supplementary Fig. S6B).

Figure 2.

CXCL14 fibroblasts induce loss of epithelial marker in breast cancer cells. A, Protein levels of E-cadherin (E-cad) in MCF7 and DCIS cells upon coculture with NIH-ctr or NIH-CXCL14 fibroblasts were detected by Western blot analysis. Representative blots are shown in the top, and quantifications of three independent experiments are shown in the bottom. B, Immunofluorescence of MCF7 cells (green) cocultured with NIH-ctr or NIH-CXCL14 fibroblasts (red) for the markers depicted in the figure. Arrowheads mark sites of loss of E-cadherin (E-cad) or Cytokeratin 8/18. Scale bar, 50 μm. C, Light microscopy pictures of MCF7 cells exposed for 48 hours to conditioned medium (CM) collected from NIH-ctr or NIH-CXCL14 cells (10× magnification). The number of cells with protrusions was counted in five vision fields in three independent experiments. Results are shown as fold of untreated MCF7 cells. unstim, unstimulated. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Error bars, SD.

Figure 2.

CXCL14 fibroblasts induce loss of epithelial marker in breast cancer cells. A, Protein levels of E-cadherin (E-cad) in MCF7 and DCIS cells upon coculture with NIH-ctr or NIH-CXCL14 fibroblasts were detected by Western blot analysis. Representative blots are shown in the top, and quantifications of three independent experiments are shown in the bottom. B, Immunofluorescence of MCF7 cells (green) cocultured with NIH-ctr or NIH-CXCL14 fibroblasts (red) for the markers depicted in the figure. Arrowheads mark sites of loss of E-cadherin (E-cad) or Cytokeratin 8/18. Scale bar, 50 μm. C, Light microscopy pictures of MCF7 cells exposed for 48 hours to conditioned medium (CM) collected from NIH-ctr or NIH-CXCL14 cells (10× magnification). The number of cells with protrusions was counted in five vision fields in three independent experiments. Results are shown as fold of untreated MCF7 cells. unstim, unstimulated. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Error bars, SD.

Close modal

Furthermore, treatment of MCF7, DCIS, and SKBR3 breast cancer cells with CM from CXCL14 fibroblasts induced changes in cell morphology. The tumor cells formed filopodium-like protrusions and obtained a mesenchymal-like morphology when cultured in CM from CXCL14 fibroblasts, but not in CM from control fibroblasts or in standard DMEM (Fig. 2C; Supplementary Fig. S6C). These phenotypes induced by CXCL14 fibroblasts were not seen in the mesenchymal metastatic breast cancer cell line MDA-MB-231, a cell line that already has undergone EMT (Supplementary Fig. S6C).

These results demonstrate the ability of CXCL14 fibroblasts to induce an EMT phenotype in certain breast cancer cells, in a paracrine manner independent of cell-to-cell contact.

CXCL14 fibroblasts enhance migration and invasion of breast cancer cells

The induction of EMT suggested functional effects of CXCL14 fibroblasts on breast cancer cells. Thus, we compared the ability of NIH-ctr and NIH-CXCL14 fibroblasts to stimulate the migration and invasion of MCF7, DCIS, and SKBR3 cells. Transwell migration assays were performed to analyze whether the changes in EMT markers and in morphology were accompanied by an increase in cell motility. CXCL14 fibroblasts displayed a stronger ability to stimulate the migration of MCF7, DCIS, and SKBR3 cells, as compared with control fibroblasts (Fig. 3A). We also investigated whether CXCL14 fibroblast–induced EMT and migration could be observed in breast cancer cell lines representing the basal (triple negative) molecular subgroup of breast cancer, including 4T1 cells and Hs578t cells (Supplementary Fig. S7A and S7B). NIH-CXCL14 cells significantly enhanced the migration (Supplementary Fig. S7A) and stimulated EMT (Supplementary Fig. S7B) of 4T1 cells, as compared with NIH-ctr cells. There was a trend toward enhanced migration of Hs578t cells, although not significant, possibly explained by the fact that these cells already have undergone EMT (Supplementary Fig. S7A).

Figure 3.

CXCL14-expressing fibroblasts enhance migration stimulate lung colonization of breast cancer cells. A, MCF7, DCIS, and SKBR3 cells were allowed to migrate toward NIH-ctr or NIH-CXCL14 fibroblasts in a transwell migration assay for 24 hours. Migration was determined by counting DAPI-stained cells that had moved through an 8-μm pore size membrane of the transwell (see details in Materials and Methods). Results are derived from three independent experiments and are presented as fold of MCF7, DCIS, or SKBR3 cells alone. B, MCF7 cells primed for 72 hours in a transwell coculture assay with NIH-ctr or NIH-CXCL14 fibroblasts were injected into the tail-vein of 8-week-old SCID mice (n = 10). C, Lungs were harvested four weeks after injection of the cancer cells. The number of MCF7 cells (human origin) in mouse lungs was semiquantitatively assessed by qRT-PCR using human- and mouse-specific primers (see details in Materials and Methods). D, Lung sections from the tail-vein experiment were stained for the human specific marker Stem121. The number of MCF7 cells in the lung was counted in 10 sections/lung and are depicted as average (n = 10). Arrowheads indicate tumor cells. Scale bar, 100 μm. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Error bars, SEM.

Figure 3.

CXCL14-expressing fibroblasts enhance migration stimulate lung colonization of breast cancer cells. A, MCF7, DCIS, and SKBR3 cells were allowed to migrate toward NIH-ctr or NIH-CXCL14 fibroblasts in a transwell migration assay for 24 hours. Migration was determined by counting DAPI-stained cells that had moved through an 8-μm pore size membrane of the transwell (see details in Materials and Methods). Results are derived from three independent experiments and are presented as fold of MCF7, DCIS, or SKBR3 cells alone. B, MCF7 cells primed for 72 hours in a transwell coculture assay with NIH-ctr or NIH-CXCL14 fibroblasts were injected into the tail-vein of 8-week-old SCID mice (n = 10). C, Lungs were harvested four weeks after injection of the cancer cells. The number of MCF7 cells (human origin) in mouse lungs was semiquantitatively assessed by qRT-PCR using human- and mouse-specific primers (see details in Materials and Methods). D, Lung sections from the tail-vein experiment were stained for the human specific marker Stem121. The number of MCF7 cells in the lung was counted in 10 sections/lung and are depicted as average (n = 10). Arrowheads indicate tumor cells. Scale bar, 100 μm. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Error bars, SEM.

Close modal

Furthermore, invasion of MCF7 and SKBR3 cells through a layer of Matrigel was enhanced by NIH-CXCL14 fibroblasts, as compared with control fibroblasts (Supplementary Fig. S7C).

Together, these results demonstrate that the CXCL14 fibroblast–induced changes in EMT markers are accompanied by an enhanced capacity of breast cancer cells to migrate and invade in vitro.

CXCL14 fibroblasts enhance lung colonization of MCF7 cells following tail-vein injection

The findings of CXCL14-induced effects on migration, invasion, and EMT, together with previous findings revealing a protumorigenic role of CXCL14, prompted in vivo studies to explore the effects of CXCL14 fibroblasts.

Tail-vein experiments monitor the ability of cancer cells to survive in the circulation, extravasate, and colonize metastatic sites. These abilities have previously been linked to EMT (37–39). Therefore, lung colonization of tail-vein–injected MCF7 cells, “primed” in vitro in a coculture format together with CXCL14 fibroblasts or control fibroblasts prior to injection, was studied (Fig. 3B).

Abundance of breast cancer cells in the lungs, 4 weeks after injection, was determined by qRT-PCR analyses with human-specific primers as described previously (29). As shown in Fig. 3C, a significantly higher number of MCF7 cells were detected in the lungs of mice that had been injected with cancer cells “primed” with CXCL14 fibroblasts, as compared with mice injected with control fibroblast-primed cancer cells.

These findings were independently validated by IHC analyses of tissue sections from lungs of mice subjected to tail-vein injection of coculture primed cancer cells. As shown in Fig. 3D, these analyses demonstrated a significantly higher number of breast cancer cells in the lungs of mice that had been injected with NIH-CXCL14–primed breast cancer cells.

These experiments thus demonstrate that CXCL14 fibroblasts, as compared with control fibroblasts, more potently stimulate lung colonization of blood-circulating breast cancer cells.

CXCL14-induced molecular signaling and cellular responses are mediated by the atypical G-protein–coupled receptor ACKR2

Next, we aimed at identifying key signaling components mediating the cellular and protumorigenic effects of the orphan chemokine CXCL14.

We have previously found that CXCL14 enhances MAPK signaling in certain cancer cells and defined them as CXCL14-responsive cell lines (22). Initial experiments, analyzing ERK phosphorylation subsequent to stimulation with recombinant CXCL14, identified a panel of CXCL14-responsive cell lines (MCF7, DCIS, SKBR3, NIH-3T3) and nonresponsive cell lines (MDA-MB-231 and LNCaP) (Supplementary Fig. S8; ref. 22). Treatment of the CXCL14-responsive cell lines MCF7 and NIH-3T3 with pertussis toxin, which specifically inhibits the Gαi subfamily of GPCRs, blocked CXCL14-induced ERK signaling in both cell types (Supplementary Fig. S9). This finding suggests that CXCL14, as other chemokines, signals through the Gαi subfamily of GPCRs.

Next, a sequence alignment approach was initiated to identify GPCRs that exhibit similarities with known chemokine receptors (see “Materials and Methods” for details; Supplementary Fig. S1). CXCL14 is a highly evolutionary conserved chemokine and the absence of chemokine receptor orthologs in species expressing CXCL14 limited the number of potential candidates (40). Furthermore, CXCL14 is a highly selective chemokine for trophoblasts of the placenta, immature dendritic cells, B cells, and NK cells, and does not induce chemotaxis of T lymphocytes. Therefore, mediators of CXCL14 signaling are likely present on trophoblasts, and on certain immune cells, but not expressed on T cells (18–21, 41). These considerations led to the selection for continued studies of 11 candidate proteins from the original list. Expression levels of these receptors were tested in CXCL14-responsive and nonresponsive cell lines (Supplementary Table S3). These results reduced the candidate set to six candidates, which were analyzed in preliminary siRNA experiments that led to continued studies on ACKR2, CXCR4, GPR25, and GPR182.

Initial experiments, using CXCL14-induced ERK phosphorylation as an endpoint, were performed in the CXCL14-responsive MCF7 cells. As shown in Supplementary Fig. S10A and S10B, downregulation of ACKR2 significantly reduced CXCL14-induced ERK phosphorylation. In contrast, downregulation of CXCR4, GPR25, and GPR182 did not affect CXCL14-induced ERK phosphorylation (Supplementary Fig. S10A and S10C). ACKR2-downregulated cells maintained ERK responses after stimulation with CXCL12, indicating specificity of the effects of ACKR2 downregulation on CXCL14 signaling (Supplementary Fig. S10D).

These initial findings were extended in two other CXCL14-responsive cell lines with the same endpoint. As shown in Supplementary Fig. S11A–S11E, CXCL14-induced ERK phosphorylation in NIH-3T3 and SKBR3 cells was also attenuated after siRNA-mediated downregulation of ACKR2. Moreover, the enhanced growth of CXCL14 fibroblasts was significantly reduced after downregulation of ACKR2 (Supplementary Fig. S11F). In contrast, no effect on cell growth was observed after downregulation of CXCR4, GPR182, and GPR25 (Supplementary Fig. S11F).

These findings prompted generation of derivatives of NIH-ctr and NIH-CXCL14 cells with stable ACKR2 downregulation with two different ACKR2 shRNAs (Fig. 4A). In agreement with findings above, CXCL14-induced ERK phosphorylation was significantly attenuated in NIH-3T3 cells with stable downregulation of ACKR2 (Fig. 4B and C). As an additional endpoint, CXCL14-induced upregulation of NOS1 was studied. As shown in Fig. 4D and E, stable ACKR2 downregulation reduced NOS1 protein and mRNA levels in NIH-3T3 cells, but not in control cells. Finally, CXCL14-induced proliferation of NIH-3T3 cells was analyzed with regard to ACKR2 dependency. Notably, downregulation of ACKR2 reduced the growth capacity of CXCL14 fibroblasts, whereas no effect of ACKR2 downregulation was detected in NIH-ctr cells (Fig. 4F).

Figure 4.

ACKR2 mediates CXCL14-stimulated signaling in fibroblasts. A, The suppression of ACKR2 expression following introduction of two different ACKR2-targeting shRNA (A and B) in NIH-ctr and NIH-CXCL14 fibroblasts was analyzed by qRT-PCR. Results are obtained from three independent experiments. B, The activation of ERK1/2 signaling following CXCL14 stimulation was monitored by Western blot analysis in NIH-3T3 fibroblasts with and without stable downregulation of ACKR2. C, Quantifications of three independent experiments from B. Analysis of Nos1 transcript (qRT-PCR; D) and Nos1 protein (Western blot) levels (E) in NIH-ctr and NIH-CXCL14 derivatives with or without stable downregulation of ACKR2. E, One representative blot together with the quantification of three independent experiments. F, The growth of NIH-ctr and NIH-CXCL14 fibroblasts with or without stable downregulation of ACKR2 was evaluated by the AlamarBlue assay (see details in Materials and Methods) after culture for three days in serum-reduced medium. The results of three independent experiments are summarized in the figure. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Error bars, SD or SEM.

Figure 4.

ACKR2 mediates CXCL14-stimulated signaling in fibroblasts. A, The suppression of ACKR2 expression following introduction of two different ACKR2-targeting shRNA (A and B) in NIH-ctr and NIH-CXCL14 fibroblasts was analyzed by qRT-PCR. Results are obtained from three independent experiments. B, The activation of ERK1/2 signaling following CXCL14 stimulation was monitored by Western blot analysis in NIH-3T3 fibroblasts with and without stable downregulation of ACKR2. C, Quantifications of three independent experiments from B. Analysis of Nos1 transcript (qRT-PCR; D) and Nos1 protein (Western blot) levels (E) in NIH-ctr and NIH-CXCL14 derivatives with or without stable downregulation of ACKR2. E, One representative blot together with the quantification of three independent experiments. F, The growth of NIH-ctr and NIH-CXCL14 fibroblasts with or without stable downregulation of ACKR2 was evaluated by the AlamarBlue assay (see details in Materials and Methods) after culture for three days in serum-reduced medium. The results of three independent experiments are summarized in the figure. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Error bars, SD or SEM.

Close modal

The above data prompted binding studies to analyze whether CXCL14 directly interacts with ACKR2. However, contrarily to CCL5, a high-affinity ligand of ACKR2, binding of CXCL14 was only weakly detectable at high concentrations on cells overexpressing ACKR2 and was not different compared with cells that lack ACKR2 (Supplementary Fig. S12A). Furthermore, in a β-arrestin1 recruitment assay, a dose-dependent recruitment of β-arrestin1 toward ACKR2 could only be detected upon CCL5, but not upon CXCL14 stimulation (Supplementary Fig. S12B).

In summary, these results derived from analyses of different cell types and using multiple endpoints identify ACKR2 as a critical mediator of CXCL14-induced signaling, although no direct interaction between CXCL14 and ACKR2 could be found.

ACKR2 and CXCL14 expression correlates with an EMT gene expression signature and poor prognosis in clinical datasets of breast cancer

To explore the clinical relevance of the experimental findings of pro-EMT effects of CXCL14 fibroblasts, correlative analyses were performed in publicly available gene expression datasets of breast cancer to investigate potential associations between CXCL14/ACKR2 expression and clinical features (35).

The analyses revealed significant positive correlations between expression of CXCL14 and the EMT gene expression signature in a meta-analysis of nine breast cancer cohorts (Supplementary Fig. S2A). Analysis of the CXCL14:EMT correlation in intrinsic subtypes of breast cancer across all cohorts revealed no major difference among the molecular subgroups of breast cancer, but a slightly stronger correlation in the Basal subgroup (Supplementary Fig. S2B). Correlations between CXCL14 and EMT were not affected by the amount of tumor stroma (Supplementary Fig. S13). This indicates that the CXCL14:EMT correlation truly is driven by cancer cell EMT rather than reflecting stroma abundance.

To extend these studies, additional analyses were performed that focused on relationships between CXCL14 expression and individual EMT-related genes. As shown in Supplementary Fig. S3, CXCL14-high breast cancer displayed, in general, an EMT profile characterized by, for example, reduced expression of E-cadherin and increased expression of EMT transcription factors including SNAI2, TWIST1, and ZEB1 and mesenchymal markers including VIM, ACTA2, FN1, and collagens. On the basis of earlier studies implying a tumor-promoting function of CXCL14 in prostate and ovarian cancer (22, 25, 42), analyses were also performed on the ovarian and prostate cancer TCGA datasets. These demonstrated results similar to those seen in the breast cancer analyses (Supplementary Fig. S3).

On the basis of the experimental studies, these associations were also analyzed in subsets defined by their combined CXCL14 and ACKR2 status. In agreement with a functional link between CXCL14 and ACKR2, the association with the EMT profile was most prominent in the CXCL14high/ACKR2high subgroup (Fig. 5A). This pattern was also seen in prostate and ovarian cancer datasets (Supplementary Fig. S4).

Figure 5.

Patients with breast cancer expressing high levels of CXCL14 and ACKR2 show enhanced EMT and adverse overall survival. A, Z-scores of EMT genes in the TCGA breast cancer gene expression dataset, in patients divided in different subgroups with high or low expression levels of CXCL14 and ACKR2. B, Kaplan–Meier analysis of the CXCL14high/ACKR2high subgroup compared with the rest of the population (n = 1,100 patients). P value was derived from log-rank test and HRs including confidence intervals were derived from univariate Cox Regression analyses.

Figure 5.

Patients with breast cancer expressing high levels of CXCL14 and ACKR2 show enhanced EMT and adverse overall survival. A, Z-scores of EMT genes in the TCGA breast cancer gene expression dataset, in patients divided in different subgroups with high or low expression levels of CXCL14 and ACKR2. B, Kaplan–Meier analysis of the CXCL14high/ACKR2high subgroup compared with the rest of the population (n = 1,100 patients). P value was derived from log-rank test and HRs including confidence intervals were derived from univariate Cox Regression analyses.

Close modal

Survival data in the TCGA datasets was also used to explore survival associations of the four CXCL14/ACKR2-defined subgroups. Initial analyses with all four groups in the breast cancer dataset indicated a particularly poor prognosis of the group with high expression of CXCL14 and ACKR2 (Supplementary Fig. S14). Notably, a significant poor survival association was seen for the combined CXCL14high/ACKR2high group, when contrasted with the rest of the TCGA population (P = 0.01; Fig. 5B). A Cox proportional hazard model revealed an increased risk of death for patients in the CXCL14high/ACKR2high subgroup (HR = 2.494; 95% CI = 1.218–5.104). This poor prognosis association of the CXCL14high/ACKR2high subgroup in breast cancer also remained significant in multivariate analyses with clinicopathologic characteristics, including breast cancer molecular subsets (Supplementary Table S4). Survival correlations of the CXCL14high/ACKR2high subgroup were also explored in publicly available datasets, representing 12 other tumor types (Supplementary Table S5). Besides breast cancer, the CXCL14high/ACKR2high subgroup was significantly correlated to worse survival of low-grade glioma, prostate cancer, clear cell renal cancer, and stomach cancer (Supplementary Table S5).

Together, these correlative analyses support the notion that the previously observed poor prognosis association of stroma-derived CXCL14 in breast cancer is related to a molecular pathway that also involves ACKR2.

Paracrine effects of CXCL14 fibroblasts depend on autocrine CXCL14 signaling

Data presented above do not clearly resolve whether the poor prognosis–associated CXCL14/ACKR2 pathway reflects autocrine CXCL14/ACKR2 signaling, supporting EMT and metastasis in a paracrine manner, or rather reflects paracrine actions of CXCL14-activated fibroblasts that involves ACKR2 in breast cancer cells.

As shown in Fig. 6A and B, CXCL14 fibroblast–induced cancer cell migration and E-Cadherin downregulation was significantly inhibited by knockdown of ACKR2 in fibroblasts. Of note, ACKR2 downregulation in control fibroblasts did not affect migration or E-Cadherin levels of cocultured MCF7 cells.

Figure 6.

Paracrine effects of fibroblast-derived CXCL14 depend on NOS1 and ACKR2. A, Migration of MCF7 cells for 24 hours in response to NIH-ctr or NIH-CXCL14 derivatives without (shCtr) or with stable knockdown of ACKR2 (shACKR2:A and shACKR2:B). B, Western blot analysis and quantification of E-cadherin levels in MCF7 cells subsequent to coculture with control (shCtr) or ACKR2-targeting (shACKR2:A) NIH-ctr or NIH-CXCL14 fibroblasts for 48 hours. C, MCF7 cells were allowed to migrate for 24 hours toward NIH-ctr or NIH-CXCL14 derivatives without (shCtr) or with stable knockdown of NOS1 (shNOS1:A and shNOS1:B). D, Western blot analysis of E-cadherin and Snail levels in MCF7 cells following coculture with NIH-ctr or NIH-CXCL14 fibroblasts with or without stable suppression of NOS1 expression. E, Quantification of Western blots as shown in B for E-cadherin (E-cad; left) and Snail (right) expression from three independent experiments. Representative blots are shown, and quantifications are based on three independent experiments. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01 and *, P < 0.05. Error bars, SD or SEM.

Figure 6.

Paracrine effects of fibroblast-derived CXCL14 depend on NOS1 and ACKR2. A, Migration of MCF7 cells for 24 hours in response to NIH-ctr or NIH-CXCL14 derivatives without (shCtr) or with stable knockdown of ACKR2 (shACKR2:A and shACKR2:B). B, Western blot analysis and quantification of E-cadherin levels in MCF7 cells subsequent to coculture with control (shCtr) or ACKR2-targeting (shACKR2:A) NIH-ctr or NIH-CXCL14 fibroblasts for 48 hours. C, MCF7 cells were allowed to migrate for 24 hours toward NIH-ctr or NIH-CXCL14 derivatives without (shCtr) or with stable knockdown of NOS1 (shNOS1:A and shNOS1:B). D, Western blot analysis of E-cadherin and Snail levels in MCF7 cells following coculture with NIH-ctr or NIH-CXCL14 fibroblasts with or without stable suppression of NOS1 expression. E, Quantification of Western blots as shown in B for E-cadherin (E-cad; left) and Snail (right) expression from three independent experiments. Representative blots are shown, and quantifications are based on three independent experiments. P values were derived from unpaired two-sided Student t tests. ***, P < 0.001; **, P < 0.01 and *, P < 0.05. Error bars, SD or SEM.

Close modal

Further evidence supporting autocrine CXCL14 signaling as the driver of the prometastatic effects was provided by analyses of the effects of downregulation of NOS1; an earlier identified downstream component of CXCL14 fibroblast signaling (25). As shown in Fig. 6C, downregulation of NOS1 significantly reduced the ability of CXCL14 fibroblasts to stimulate migration of MCF7 cells. Notably, NOS1 downregulation did not affect cancer cell migration induced by control fibroblasts (Fig. 6C). The reduction of NOS1 signaling also attenuated the CXCL14 fibroblast–induced downregulation of E-cadherin and upregulation of Snail in CM-treated MCF7 cells (Fig. 6D and E).

The reduced paracrine effects of CXCL14 fibroblasts following downregulation of ACKR2 or NOS1 suggest that CXCL14 itself is not promoting EMT, but rather stimulates the expression of EMT regulators in fibroblasts in an ACKR2-/NOS1-dependent manner. To identify such putative EMT-regulating soluble factors derived from CXCL14 fibroblasts, we used a protein profiler and compared the secretome of NIH-CXCL14 and NIH-ctr fibroblasts. Among the factors that are more abundantly expressed by CXCL14 fibroblasts (Supplementary Fig. S15) are proangiogenic factors (e.g., FGF-2, Angiogenin, VEGF-A), supporting the previous notion that NIH-CXCL14 cells stimulate angiogenesis (22), molecules involved in matrix remodeling such as Adamts1, MMP8, TIMP-1, as well as inducers and effectors of EMT including CXCL1, CX3CL1, TIMP-1, FGF2, HGF, and tissue factor.

These data, together with the findings of Fig. 1, which show reduced EMT in the tumors formed after coinjection with NOS1-downregulated CXCL14 fibroblasts, indicate that the promigratory and EMT-modulatory effects of CXCL14 fibroblasts depend on autocrine CXCL14/ACKR2/NOS1 signaling.

This study extends earlier findings that have identified stroma-derived CXCL14 as a poor prognosis factor in breast cancer. Mechanistic and correlative studies together suggest a novel prometastatic pathway composed of autocrine CXCL14/ACKR2/NOS1 signaling in fibroblasts that generates a fibroblast phenotype that supports cancer cell migration, invasion, EMT, and metastasis (Supplementary Fig. S16).

The finding of prometastatic effects of CXCL14 adds to earlier literature indicating the involvement of chemokines in metastasis development. CCL5 secreted from bone marrow–derived mesenchymal cells, recruited to the breast tumor stroma, was identified a key player in promoting tumor cell invasion and development of metastasis in SCID mice (13). In another study, paracrine crosstalk between tumor cells, myeloid cells, and endothelial cells, involving CXCL1 and CXCL2 signaling, was shown to drive metastasis and chemoresistance in the MMTV-PyMT mouse model of breast cancer (14). In addition, in experimental breast tumors, CXCL12 secreted from CAFs was shown to select for clones of cancer cells with a high Src activity, and the ability to specifically form metastasis in bone with a CXCL12-rich microenvironment (15).

Earlier studies have also linked chemokines and their receptors specifically to EMT. A constitutively active form of CXCR4, the receptor for CXCL12, has been shown to be involved in modulation of breast tumor cell EMT markers and to enhance formation of lymph node metastasis in mice (43). CCR7, the receptor for CCL21, and CXCR5 and its ligand CXCL13 have been shown to significantly correlate with EMT markers and enhanced lymph node metastasis of human breast tumors (44, 45). Moreover, a GM-CSF-CCL18–positive feedback loop have been implied in EMT and breast tumor metastasis in mice and associated with worse outcome in patients with breast cancer (16).

The comparison of proteins secreted by NIH-CXCL14 and NIH-ctr cells provided new insight in CXCL14 signaling in fibroblasts and revealed a set of candidates that mediate EMT stimulated by CXCL14-expressing fibroblasts either individually and/or in combination. For example, CXCL1, CX3CL1, HGF, and TIMP-1 and tissue factor have previously been demonstrated to affect EMT and metastasis of breast cancer cells (39, 46–48). Recently, Wang and colleagues identified CCL17 derived from CXCL14-activated fibroblasts as another mediator of CXCL14-stimulated breast cancer EMT and metastasis (49). Furthermore, CXCL14 was shown to act as a chemoattractant for M2 macrophages that are known to promote EMT (16, 50, 51). Together, these findings suggest that CXCL14 operates different axis of stromal signaling shifting the phenotype of stromal cells toward tumor progression and metastasis.

A key finding of this study is the demonstration that CXCL14-induced molecular signaling and cellular responses depend on ACKR2, classified as an atypical chemokine receptor. ACKR2-dependent CXCL14 signaling was shown in cell types of different origin (breast cancer cells and fibroblasts) and the downregulation of ACKR2 almost completely abolished CXCL14-induced effects. These data suggest that ACKR2 is a required component of CXCL14 signaling. Findings of ACKR2 expression on some breast cancer cells (Supplementary Table S3) also suggest the possibility that CXCL14, in certain settings, might be prometastatic or EMT-stimulatory through direct effects on malignant cells. This topic should be further explored in future studies. Continued mechanistic analyses are also warranted regarding the roles of CXCL14/ACKR2 on other steps of the metastatic process than those covered by the analyses of the invasive border (Fig. 1) and the “tail-vein” experiment (Fig. 3).

Studies have suggested the chemokine receptors CXCR4 and GPR85 to directly bind CXCL14 and modulating signaling (49, 52). However, Otte and colleagues demonstrated that CXCL14 does not bind and impact on CXCR4 signaling (53). In line with these data, we found that siRNA-mediated downregulation of CXCR4 did not affect CXCL14-induced MAPK signaling in MCF7 breast cancer cells (Supplementary Fig. S10C) suggesting that the functions of CXCL14 might be more complex than for other chemokines and might require formation of chemokine or receptor heterocomplexes, which may explain the difficulty to precisely define its signaling components (54).

These atypical chemokine receptors have earlier been defined as scavenging receptors that bind chemokines with high affinity but are unable to induce signaling and cell migration (11). Absence of the well-conserved DRYLAIVHA motif (DKYLEIVHA in ACKR2) has been assumed to explain the inability of atypical chemokine receptors to induce downstream receptor signaling subsequent to ligand binding.

In this study, we observe an impact of ACKR2 on the molecular signaling, including MAPK activation (Fig. 4A–C; Supplementary Figs. S10B and S11A–S11E) and cellular functions, including enhanced fibroblast proliferation (Fig. 4F; Supplementary Fig. S11F) in response to CXCL14. Independent recent evidence indeed does support signaling functions for some of the atypical chemokine receptors. These signaling properties were proposed to be possibly cell type dependent as in one study, experiments demonstrated coupling of ACKR3 to Gαi proteins and induction of CXCL12-dependent conformational changes, but no activation of calcium signaling (55). Another study confirmed the binding of ACKR3 to PTX-sensitive Gαi proteins and revealed activation of calcium mobilization, ERK signaling, and AKT signaling and enhanced migration and proliferation, subsequent to CXCL12 binding in rodent astrocytes and human glioma cell lines (56). Earlier overexpression studies have also demonstrated ACKR2-induced calcium mobilization by murine ACKR2 (57). A recent study also proposed G-protein–independent, β-arrestin–dependent, activation of the cofilin pathway [Rac1-p21–activated kinase 1 (PAK1)-LIM kinase 1 (LIMK1) cascade] following ACKR2 stimulation suggesting that ACKR2 is not a totally silent receptor (58). Taken together, these findings challenge the definition of ACKRs as exclusive nonsignaling, chemokine scavenger receptors.

Earlier studies have identified stromal, but not epithelial, CXCL14 as a bad prognosis marker in breast cancer (1). The correlative data of this study support the notion of functional clinical relevant interaction between CXCL14 and ACKR2. As shown in Fig. 5, the association between CXCL14 and an “EMT profile” is enhanced when ACKR2 status is integrated in patient classification (Fig. 5A). Similarly, the survival association of CXCL14 in the breast cancer TCGA gene expression dataset is only detected in analyses that also consider ACKR2 status (Fig. 5B; Supplementary Fig. S14). Importantly, the combined CXCL14/ACKR2 metric is also a significant marker in multivariate analyses including molecular breast cancer subtypes (Supplementary Table S4). It is recognized that the TCGA-based analyses fail to assign the prognostically relevant ACKR2 expression to the stromal or epithelial compartment. However, the mechanistic studies of the current report suggest that autocrine CXCL14/ACKR2 signaling in the stroma contributes to the survival and EMT associations. Compartment-specific analyses of ACKR2 are prompted by the findings of the current study.

In summary, this study thus identifies a novel potentially druggable CXCL14/ACKR2 pathway involved in breast cancer EMT and metastasis. Important tasks for future studies include development of inhibitory agents for initial testing in experimental breast cancer models and further exploration of relevance of this pathway in other tumor types. Furthermore, these results also encourage continued studies exploring biological activity of ACKRs and to decipher the exact interplay between CXCL14 with classical and atypical chemokine receptors.

No potential conflicts of interest were disclosed.

Conception and design: E. Sjöberg, M. Augsten, A. Östman

Development of methodology: E. Sjöberg, L. Milde, D. Hägerstrand, A. Chevigné

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Sjöberg, M. Meyrath, L. Milde, M. Herrera, J. Lövrot, D. Hägerstrand, M. Bartish, C. Rolny, A. Chevigné, M. Augsten

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Sjöberg, M. Meyrath, L. Milde, M. Herrera, J. Lövrot, O. Frings, M. Bartish, E. Sonnhammer, A. Chevigné, M. Augsten, A. Östman

Writing, review, and/or revision of the manuscript: E. Sjöberg, M. Meyrath, J. Lövrot, D. Hägerstrand, C. Rolny, E. Sonnhammer, A. Chevigné, M. Augsten, A. Östman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Östman

Study supervision: A. Östman

Members of A. Östman's group are acknowledged for support throughout the studies. Studies were supported by grants from the Swedish Cancer Society, BRECT, the Linné STARGET grant from Swedish Research Council and the KI/AZ-collaborative initiative, the Luxembourg Institute of Health (LIH) MESR (grants 20160116 and 20170113), and the Luxembourg National Research Fund PhD fellows (grants AFR-3004509 and INTER/FWO “Nanokine” - grant 15/10358798). Technical support was provided by the histo-pathology unit of Cancer Centrum Karolinska. Animal experiments benefited from the expertise of the MTC animal facility.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Sjoberg
E
,
Augsten
M
,
Bergh
J
,
Jirstrom
K
,
Ostman
A
. 
Expression of the chemokine CXCL14 in the tumour stroma is an independent marker of survival in breast cancer
.
Br J Cancer
2016
;
114
:
1117
24
.
2.
Kalluri
R
,
Weinberg
RA
. 
The basics of epithelial-mesenchymal transition
.
J Clin Invest
2009
;
119
:
1420
8
.
3.
De Craene
B
,
Berx
G
. 
Regulatory networks defining EMT during cancer initiation and progression
.
Nat Rev Cancer
2013
;
13
:
97
110
.
4.
Zheng
X
,
Carstens
JL
,
Kim
J
,
Scheible
M
,
Kaye
J
,
Sugimoto
H
, et al
Epithelial-to-mesenchymal transition is dispensable for metastasis but induces chemoresistance in pancreatic cancer
.
Nature
2015
;
527
:
525
30
.
5.
Del Pozo Martin
Y
,
Park
D
,
Ramachandran
A
,
Ombrato
L
,
Calvo
F
,
Chakravarty
P
, et al
Mesenchymal cancer cell-stroma crosstalk promotes niche activation, epithelial reversion, and metastatic colonization
.
Cell Rep
2015
;
13
:
2456
69
.
6.
Beerling
E
,
Seinstra
D
,
de Wit
E
,
Kester
L
,
van der Velden
D
,
Maynard
C
, et al
Plasticity between epithelial and mesenchymal states unlinks EMT from metastasis-enhancing stem cell capacity
.
Cell Rep
2016
;
14
:
2281
8
.
7.
Ostman
A
,
Augsten
M
. 
Cancer-associated fibroblasts and tumor growth–bystanders turning into key players
.
Curr Opin Genet Dev
2009
;
19
:
67
73
.
8.
Kalluri
R
. 
The biology and function of fibroblasts in cancer
.
Nat Rev Cancer
2016
;
16
:
582
98
.
9.
Mishra
P
,
Banerjee
D
,
Ben-Baruch
A
. 
Chemokines at the crossroads of tumor-fibroblast interactions that promote malignancy
.
J Leukoc Biol
2011
;
89
:
31
9
.
10.
Thelen
M
. 
Dancing to the tune of chemokines
.
Nat Immunol
2001
;
2
:
129
34
.
11.
Bonecchi
R
,
Savino
B
,
Borroni
EM
,
Mantovani
A
,
Locati
M
. 
Chemokine decoy receptors: structure-function and biological properties
.
Curr Top Microbiol Immunol
2010
;
341
:
15
36
.
12.
Balkwill
F
. 
Cancer and the chemokine network
.
Nat Rev Cancer
2004
;
4
:
540
50
.
13.
Karnoub
AE
,
Dash
AB
,
Vo
AP
,
Sullivan
A
,
Brooks
MW
,
Bell
GW
, et al
Mesenchymal stem cells within tumour stroma promote breast cancer metastasis
.
Nature
2007
;
449
:
557
63
.
14.
Acharyya
S
,
Oskarsson
T
,
Vanharanta
S
,
Malladi
S
,
Kim
J
,
Morris
PG
, et al
A CXCL1 paracrine network links cancer chemoresistance and metastasis
.
Cell
2012
;
150
:
165
78
.
15.
Zhang
XH
,
Jin
X
,
Malladi
S
,
Zou
Y
,
Wen
YH
,
Brogi
E
, et al
Selection of bone metastasis seeds by mesenchymal signals in the primary tumor stroma
.
Cell
2013
;
154
:
1060
73
.
16.
Su
S
,
Liu
Q
,
Chen
J
,
Chen
J
,
Chen
F
,
He
C
, et al
A positive feedback loop between mesenchymal-like cancer cells and macrophages is essential to breast cancer metastasis
.
Cancer Cell
2014
;
25
:
605
20
.
17.
Zlotnik
A
,
Burkhardt
AM
,
Homey
B
. 
Homeostatic chemokine receptors and organ-specific metastasis
.
Nat Rev Immunol
2011
;
11
:
597
606
.
18.
Sleeman
MA
,
Fraser
JK
,
Murison
JG
,
Kelly
SL
,
Prestidge
RL
,
Palmer
DJ
, et al
B cell- and monocyte-activating chemokine (BMAC), a novel non-ELR alpha-chemokine
.
Int Immunol
2000
;
12
:
677
89
.
19.
Kurth
I
,
Willimann
K
,
Schaerli
P
,
Hunziker
T
,
Clark-Lewis
I
,
Moser
B
. 
Monocyte selectivity and tissue localization suggests a role for breast and kidney-expressed chemokine (BRAK) in macrophage development
.
J Exp Med
2001
;
194
:
855
61
.
20.
Shellenberger
TD
,
Wang
M
,
Gujrati
M
,
Jayakumar
A
,
Strieter
RM
,
Burdick
MD
, et al
BRAK/CXCL14 is a potent inhibitor of angiogenesis and a chemotactic factor for immature dendritic cells
.
Cancer Res
2004
;
64
:
8262
70
.
21.
Starnes
T
,
Rasila
KK
,
Robertson
MJ
,
Brahmi
Z
,
Dahl
R
,
Christopherson
K
, et al
The chemokine CXCL14 (BRAK) stimulates activated NK cell migration: implications for the downregulation of CXCL14 in malignancy
.
Exp Hematol
2006
;
34
:
1101
5
.
22.
Augsten
M
,
Hagglof
C
,
Olsson
E
,
Stolz
C
,
Tsagozis
P
,
Levchenko
T
, et al
CXCL14 is an autocrine growth factor for fibroblasts and acts as a multi-modal stimulator of prostate tumor growth
.
Proc Natl Acad Sci U S A
2009
;
106
:
3414
9
.
23.
Allinen
M
,
Beroukhim
R
,
Cai
L
,
Brennan
C
,
Lahti-Domenici
J
,
Huang
H
, et al
Molecular characterization of the tumor microenvironment in breast cancer
.
Cancer Cell
2004
;
6
:
17
32
.
24.
Gu
XL
,
Ou
ZL
,
Lin
FJ
,
Yang
XL
,
Luo
JM
,
Shen
ZZ
, et al
Expression of CXCL14 and its anticancer role in breast cancer
.
Breast Cancer Res Treat
2012
;
135
:
725
35
.
25.
Augsten
M
,
Sjoberg
E
,
Frings
O
,
Vorrink
SU
,
Frijhoff
J
,
Olsson
E
, et al
Cancer-associated fibroblasts expressing CXCL14 rely upon NOS1-derived nitric oxide signaling for their tumor-supporting properties
.
Cancer Res
2014
;
74
:
2999
3010
.
26.
Szpakowska
M
,
Nevins
AM
,
Meyrath
M
,
Rhainds
D
,
D'Huys
T
,
Guite-Vinet
F
, et al
Different contributions of chemokine N-terminal features attest to a different ligand binding mode and a bias towards activation of ACKR3/CXCR7 compared with CXCR4 and CXCR3
.
Br J Pharmacol
2018
;
175
:
1419
38
.
27.
Szpakowska
M
,
Meyrath
M
,
Reynders
N
,
Counson
M
,
Hanson
J
,
Steyaert
J
, et al
Mutational analysis of the extracellular disulphide bridges of the atypical chemokine receptor ACKR3/CXCR7 uncovers multiple binding and activation modes for its chemokine and endogenous non-chemokine agonists
.
Biochem Pharmacol
2018
;
153
:
299
309
.
28.
Sonnhammer
EL
,
Hollich
V
. 
Scoredist: a simple and robust protein sequence distance estimator
.
BMC Bioinformatics
2005
;
6
:
108
.
29.
Malek
A
,
Catapano
CV
,
Czubayko
F
,
Aigner
A
. 
A sensitive polymerase chain reaction-based method for detection and quantification of metastasis in human xenograft mouse models
.
Clin Exp Metastasis
2010
;
27
:
261
71
.
30.
Miller
LD
,
Smeds
J
,
George
J
,
Vega
VB
,
Vergara
L
,
Ploner
A
, et al
An expression signature for p53 status in human breast cancer predicts mutation status, transcriptional effects, and patient survival
.
Proc Natl Acad Sci U S A
2005
;
102
:
13550
5
.
31.
Pawitan
Y
,
Bjohle
J
,
Amler
L
,
Borg
AL
,
Egyhazi
S
,
Hall
P
, et al
Gene expression profiling spares early breast cancer patients from adjuvant therapy: derived and validated in two population-based cohorts
.
Breast Cancer Res
2005
;
7
:
R953
64
.
32.
Wang
Y
,
Klijn
JG
,
Zhang
Y
,
Sieuwerts
AM
,
Look
MP
,
Yang
F
, et al
Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer
.
Lancet
2005
;
365
:
671
9
.
33.
Curtis
C
,
Shah
SP
,
Chin
SF
,
Turashvili
G
,
Rueda
OM
,
Dunning
MJ
, et al
The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups
.
Nature
2012
;
486
:
346
52
.
34.
Cancer Genome Atlas Network
. 
Comprehensive molecular portraits of human breast tumours
.
Nature
2012
;
490
:
61
70
.
35.
Taube
JH
,
Herschkowitz
JI
,
Komurov
K
,
Zhou
AY
,
Gupta
S
,
Yang
J
, et al
Core epithelial-to-mesenchymal transition interactome gene-expression signature is associated with claudin-low and metaplastic breast cancer subtypes
.
Proc Natl Acad Sci U S A
2010
;
107
:
15449
54
.
36.
Parker
JS
,
Mullins
M
,
Cheang
MC
,
Leung
S
,
Voduc
D
,
Vickery
T
, et al
Supervised risk predictor of breast cancer based on intrinsic subtypes
.
J Clin Oncol
2009
;
27
:
1160
7
.
37.
Guarino
M
,
Rubino
B
,
Ballabio
G
. 
The role of epithelial-mesenchymal transition in cancer pathology
.
Pathology
2007
;
39
:
305
18
.
38.
Tsai
JH
,
Yang
J
. 
Epithelial-mesenchymal plasticity in carcinoma metastasis
.
Genes Dev
2013
;
27
:
2192
206
.
39.
Bourcy
M
,
Suarez-Carmona
M
,
Lambert
J
,
Francart
ME
,
Schroeder
H
,
Delierneux
C
, et al
Tissue factor induced by epithelial-mesenchymal transition triggers a procoagulant state that drives metastasis of circulating tumor cells
.
Cancer Res
2016
;
76
:
4270
82
.
40.
DeVries
ME
,
Kelvin
AA
,
Xu
L
,
Ran
L
,
Robinson
J
,
Kelvin
DJ
. 
Defining the origins and evolution of the chemokine/chemokine receptor system
.
J Immunol
2006
;
176
:
401
15
.
41.
Kuang
H
,
Chen
Q
,
Fan
X
,
Zhang
Y
,
Zhang
L
,
Peng
H
, et al
CXCL14 inhibits trophoblast outgrowth via a paracrine/autocrine manner during early pregnancy in mice
.
J Cell Physiol
2009
;
221
:
448
57
.
42.
Zhao
L
,
Ji
G
,
Le
X
,
Wang
C
,
Xu
L
,
Feng
M
, et al
Long noncoding RNA LINC00092 acts in cancer-associated fibroblasts to drive glycolysis and progression of ovarian cancer
.
Cancer Res
2017
;
77
:
1369
82
.
43.
Sobolik
T
,
Su
YJ
,
Wells
S
,
Ayers
GD
,
Cook
RS
,
Richmond
A
. 
CXCR4 drives the metastatic phenotype in breast cancer through induction of CXCR2 and activation of MEK and PI3K pathways
.
Mol Biol Cell
2014
;
25
:
566
82
.
44.
Li
F
,
Zou
Z
,
Suo
N
,
Zhang
Z
,
Wan
F
,
Zhong
G
, et al
CCL21/CCR7 axis activating chemotaxis accompanied with epithelial-mesenchymal transition in human breast carcinoma
.
Med Oncol
2014
;
31
:
180
.
45.
Biswas
S
,
Sengupta
S
,
Roy Chowdhury
S
,
Jana
S
,
Mandal
G
,
Mandal
PK
, et al
CXCL13-CXCR5 co-expression regulates epithelial to mesenchymal transition of breast cancer cells during lymph node metastasis
.
Breast Cancer Res Treat
2014
;
143
:
265
76
.
46.
Wang
N
,
Liu
W
,
Zheng
Y
,
Wang
S
,
Yang
B
,
Li
M
, et al
CXCL1 derived from tumor-associated macrophages promotes breast cancer metastasis via activating NF-kappaB/SOX4 signaling
.
Cell Death Dis
2018
;
9
:
880
.
47.
D'Angelo
RC
,
Liu
XW
,
Najy
AJ
,
Jung
YS
,
Won
J
,
Chai
KX
, et al
TIMP-1 via TWIST1 induces EMT phenotypes in human breast epithelial cells
.
Mol Cancer Res
2014
;
12
:
1324
33
.
48.
Tardaguila
M
,
Mira
E
,
Garcia-Cabezas
MA
,
Feijoo
AM
,
Quintela-Fandino
M
,
Azcoitia
I
, et al
CX3CL1 promotes breast cancer via transactivation of the EGF pathway
.
Cancer Res
2013
;
73
:
4461
73
.
49.
Wang
Y
,
Weng
X
,
Wang
L
,
Hao
M
,
Li
Y
,
Hou
L
, et al
HIC1 deletion promotes breast cancer progression by activating tumor cell/fibroblast crosstalk
.
J Clin Invest
2018
;
128
:
5235
50
.
50.
Cereijo
R
,
Gavalda-Navarro
A
,
Cairo
M
,
Quesada-Lopez
T
,
Villarroya
J
,
Moron-Ros
S
, et al
CXCL14, a brown adipokine that mediates brown-fat-to-macrophage communication in thermogenic adaptation
.
Cell Metab
2018
;
28
:
750
63
.
51.
Linde
N
,
Casanova-Acebes
M
,
Sosa
MS
,
Mortha
A
,
Rahman
A
,
Farias
E
, et al
Macrophages orchestrate breast cancer early dissemination and metastasis
.
Nat Commun
2018
;
9
:
21
.
52.
Tanegashima
K
,
Suzuki
K
,
Nakayama
Y
,
Tsuji
K
,
Shigenaga
A
,
Otaka
A
, et al
CXCL14 is a natural inhibitor of the CXCL12-CXCR4 signaling axis
.
FEBS Lett
2013
;
587
:
1731
5
.
53.
Otte
M
,
Kliewer
A
,
Schutz
D
,
Reimann
C
,
Schulz
S
,
Stumm
R
. 
CXCL14 is no direct modulator of CXCR4
.
FEBS Lett
2014
;
588
:
4769
75
.
54.
Kleist
AB
,
Getschman
AE
,
Ziarek
JJ
,
Nevins
AM
,
Gauthier
PA
,
Chevigne
A
, et al
New paradigms in chemokine receptor signal transduction: Moving beyond the two-site model
.
Biochem Pharmacol
2016
;
114
:
53
68
.
55.
Levoye
A
,
Balabanian
K
,
Baleux
F
,
Bachelerie
F
,
Lagane
B
. 
CXCR7 heterodimerizes with CXCR4 and regulates CXCL12-mediated G protein signaling
.
Blood
2009
;
113
:
6085
93
.
56.
Odemis
V
,
Lipfert
J
,
Kraft
R
,
Hajek
P
,
Abraham
G
,
Hattermann
K
, et al
The presumed atypical chemokine receptor CXCR7 signals through G(i/o) proteins in primary rodent astrocytes and human glioma cells
.
Glia
2012
;
60
:
372
81
.
57.
Nibbs
RJ
,
Wylie
SM
,
Pragnell
IB
,
Graham
GJ
. 
Cloning and characterization of a novel murine beta chemokine receptor, D6. Comparison to three other related macrophage inflammatory protein-1alpha receptors, CCR-1, CCR-3, and CCR-5
.
J Biol Chem
1997
;
272
:
12495
504
.
58.
Borroni
EM
,
Cancellieri
C
,
Vacchini
A
,
Benureau
Y
,
Lagane
B
,
Bachelerie
F
, et al
beta-arrestin-dependent activation of the cofilin pathway is required for the scavenging activity of the atypical chemokine receptor D6
.
Sci Signal
2013
;
6
:
ra30
.

Supplementary data