Bone is the most common metastatic site for breast cancer. Although the estrogen-related receptor alpha (ERRα) has been implicated in breast cancer cell dissemination to the bone from the primary tumor, its role after tumor cell anchorage in the bone microenvironment remains elusive. Here, we reveal that ERRα inhibits the progression of bone metastases of breast cancer cells by increasing the immune activity of the bone microenvironment. Overexpression of ERRα in breast cancer bone metastases induced expression of chemokines CCL17 and CCL20 and repressed production of TGFβ3. Subsequently, CD8+ T lymphocytes recruited to bone metastases escaped TGFβ signaling control and were endowed with exacerbated cytotoxic features, resulting in significant reduction in metastases. The clinical relevance of our findings in mice was confirmed in over 240 patients with breast cancer. Thus, this study reveals that ERRα regulates immune properties in the bone microenvironment that contributes to decreasing metastatic growth.
This study places ERRα at the interplay between the immune response and bone metastases of breast cancer, highlighting a potential target for intervention in advanced disease.
Bone metastases are a frequent complication of cancer, occurring in up to 70% of patients with advanced breast cancer, and are associated with both high morbidity and elevated mortality (1–3). The progression of bone metastases relies on the ability of the malignant cell colonizing the bone and to modify bone microenvironment allowing the release of bone-stored factors including TGFβ, bone morphogenetic protein, or insulin growth factor, which in turn stimulate bone metastasis progression (2, 4). However, treatments that mainly involved antiresorptive agents of the bone failed to improve the overall survival of patients with cancer even though it inhibited osteoclasts' resorptive activity (3), implying that other mechanisms than the activation of osteoclasts by tumor cells are involved in modulating bone metastasis growth. The immune cells present in the bone and particularly activated CD8+ T lymphocytes can repress the progression of breast cancer osteolytic bone metastasis (5–8). However, whether bone metastasis can influence the activation of immune cells present in the bone and by which mechanisms is totally unknown.
The estrogen-related receptor alpha (ERRα, or NR3B1 according to the Nuclear Receptors Nomenclature Committee, 1999) is overexpressed in 55% of breast tumors (9, 10). Though ERRα shares structural similarities with the estrogen receptors α/β, it does not bind estrogens and no natural ligand has yet been found (11), though several molecules can either increase or decrease ERRα activity, such as the inverse agonists XCT790 or C29 (12, 13). ERRα is mainly involved in the adaptive bioenergetics response (11). In cancer, beside angiogenesis, ERRα is strongly linked to tumor cell invasion (14, 15). Notably, ERRα-positive tumors are associated with more invasive breast cancer and a higher risk of recurrence (9, 14). The overexpression of ERRα in breast cancer promotes tumor growth in the mammary gland and breast cancer metastatic dissemination to the bone (16). However, the role of ERRα in bone metastasis outcome once they have anchorage in the bone microenvironment remains elusive.
In this study, using loss and gain of expression of ERRα, as well as chemical inhibitors, we demonstrated that ERRα enhanced the ability of breast cancer cell established in the bone to recruit activated CD8+ T cells to the bone. In addition, ERRα expression on breast cancer cells repressed their ability to produce TGFβ, a potent immune-suppressive cytokine. Subsequently, TGFβ signaling was impaired in T cells infiltrating the bone, and CD8+ T-cell cytotoxic function exacerbated, leading to metastatic progression. Altogether, our work assigns a totally unexpected role to ERRα, revealing that the expression of this orphan receptor on the breast cancer metastases promotes an efficient antitumor immune response once the tumor cells are settled in bone.
Materials and Methods
The mouse triple-negative breast cancer (TNBC) cell line 4T1 (year 2012; ATCC lot: 58603185-CRL-2539) and human luminal MCF7 (year 2012; ATCC-HTB-22 Lot: 86012803) were obtained from the ATCC. MDA-MB-231/BO2-FRT (BO2) breast cancer cells, a subpopulation of the human MDA-MB-231 breast cancer line (TNBC), were selected for their high efficiency to metastasize to bone (17). These cell lines were tested for authentication by DNA fingerprinting using the short tandem repeat method in 2014. TNBC cell lines and MCF7 were cultured in DMEM or RPMI-1640 (Life Technologies) medium, respectively, supplemented with 10% FBS (Perbio) and 1% penicillin/streptomycin (Invitrogen) at 37°C in a 5% CO2 incubator. Cells lines were tested for Mycoplasma regularly. Mouse and human ESRRA (ERRα) cDNA and the dominant-negative coactivator domain AF2 (AF2) mutants were described previously (16, 18). Briefly, pSRα-ERRαWT and pEcmv-ERRαAF2 or respective empty vectors (CT) constructs were transfected into parental 4T1 cells and cultured for 4 weeks in puromycin (2 μg/mL; Life Technologies). Three independent clones were obtained from pSRα-ERRαWT transfection (4T1-ERRα) and from pEcmv-ERRαAF2 transfection (4T1-ERRαAF2). Two independent clones were obtained from empty vector transfection respectively, pSRα-4T1-CT (4T1-CT) and pEcmv-4T1-CT (4T1-CTaf2). For MCF7 clones, a mix containing 1.5 μg Retroviral pLPCX-Human-ERRαWT, pLPCX-HumanERRαAF2, or empty vector and 0.5 μg pCMV-VSV-G envelope vector (Cell Biolabs) was used previously (16). 4T1-ERRα and 4T1-ERRαAF2 (pool of 3 clones each) cells were treated for 24 hours with the ERRα inverse-agonists XCT790 (Sigma) or C29 (AGV Discovery) at 1 and 5 μmol/L, respectively, as described (12, 13, 16). DMSO was used as a vehicle (Veh).
Six-week-old BALB/c female mice were purchased from Janvier and housed in an a specific pathogen-free facility (ALECS platform; Faculté de Médecine Laennec, Lyon, France). Bone metastasis experiments were performed by inoculating intra-arterially either 4T1-CT (pool of 2 clones) in parallel with 4T1-ERRα (pool of 3 clones) or 4T1-CT(af2) (pool of 2 clones) in parallel with 4T1-ERRαAF2 (pool of 3 clones) cell lines (5 × 105 cells in 100 μL of PBS). Radiographs (LifeRay HM Plus, Ferrania) of animals were taken at 15 days after inoculation using X-ray (MX-20; Faxitron X-ray Corporation). The extent of bone destruction for each animal was expressed in mm2. Animals were sacrificed and hind limbs were then collected for histology and histomorphometric analysis. Tibiae were scanned using microcomputed tomography (Skyscan1076, Skyscan) with an 8.8 voxel size and an X-ray tube (50 kV; 80 mA) with 0.5 μm aluminum filter, and three-dimensional reconstructions were performed with a dedicated visualization software (NRecon and CTVox, and Skyscan; ref. 18). Bone Volume/Tissue Volume: (%BV/TV) were carried out with CTAn (version 1.9, Skyscan) and CTVol (version 2.0, Skyscan) software. Dissected bones were then processed for histologic (Goldner's Trichrome solution staining) and histomorphometric analyses [tumor burden-to-soft tissue volume (%TB/STV); ref. 18]. Depletion of CD8+ T cells was performed by i.p. injection of anti-CD8β (BioXCell, clone Lyt3.2; BE0223). Note that 387.5 μg per mouse were injected 4 times every 2 days, from day 10 after metastasis injection, i.e., when osteolytic lesions start to be detectable.
Mice were handled according to the French Ministerial Decree No. 87–848 of October 19, 1987. Experimental protocols were approved by the Institutional Animal Care and Use Committee at the Université-Lyon1 (France; ethic committee CEEA-55 Comité d'Ethique en Expérimentation Animale-DR2014-44-DR2015-28).
Human sample meta-genomic analysis
Correlation analyses were performed using published datasets downloaded from the Gene Expression Omnibus including primary tumor, no metastases, visceral + bone, or only bone metastasis (GSE12276-GSE2034-GSE2603; n = 248; refs. 19–21). Z-scores were calculated on normalized data of each dataset by subtracting the population mean from individual expression values for each gene and then dividing the difference by the population SD.
The expression levels of several chemokines known to influence T-cell chemoattraction were obtained by qPCR Sign Arrays (Cytokines Array and Inflammation Array). Indeed, two qPCR Sign Arrays, Cytokines and Inflammation Arrays (AnyGenes, CT1M1-IFM1, CliniSciences), were used to quantify expression of cytokines, chemokines, and growth factors. Total RNA was extracted from 4T1-CT and 4T1-ERRα cells, and 2 μg were reverse-transcribed as previously described (16). Real-time PCR was performed according to the manufacturer's instructions. Two heat maps were generated using the heatmap.2 function in the gplots library of R (version 3.5.1). Only regulations that were reproducible between the two arrays are presented.
Protein–protein interaction network reconstruction and analysis
The protein–protein interaction network with BIOGRID (release 3.4.160) from Homo sapiens with PSICQUIC (Proteomics-Standard-Initiative-Common-QUery-InterfaCe) retrieval (10242018) and Cytoscape environment was used (16). A BIOGRID (https://thebiogrid.org/)-based custom approach was used to define a protein interactome of the following proteins: ESRRA-CCL17-CCL20-OPG-NRIP1-SRC1-SRC2-SRC3-PGC1A-PGC1B-CCR4 and CCR6. The resulting interactome encompasses 911 proteins (hereby defined as “Extended Network of ESRRA, CCL17, CCL20”; reachable on Ndex webserver, https://doi.org/10.18119/N9W891; Supplementary Fig. S1). To determine the connectors between CCL17, CCL20, and ESRRA, a custom approach combining shortest path and connectivity degree analysis was applied to determine a “Minimal Network of ESRRA, CCL17, CCL20” (reachable on Ndex webserver, https://doi.org/10.18119/N9RK5C; containing 101 proteins) acknowledging connections that may support ESRRA signaling (Supplementary Fig. S2; ref. 16). We overlaid and extracted information from the Gene Ontology consortium to pinpoint proteins that are already known to be involved in the immune system process, as well as T- and B-cell homeostasis (GO-IDs: 0002376, 0043029, 0001782) to create “Minimal Network specific to immune response to tumor” (containing 52 proteins). To determine the connectors between ESRRA and CCL17, ESRRA and CCL20, a shortest path was applied to the “Minimal Network of ESRRA, CCL17 and CCL20” (16, 22, 23).
Ex vivo cell preparation
For hind limbs, muscles were removed, and bones were sliced and then incubated at 37°C with a 1/10 solution of collagenase hyaluronidase (Stem Cell) for 2 hours. Bones were then mechanically disrupted with a syringe plunge, filtered, and cells were collected. For lung metastases (LM), lungs were crushed with a syringe plunge on a filter (100 μm; BD Bioscience) and cells were collected. Cells released from lungs and bones were incubated at 37°C in the presence of DMEM (Life Technologies) supplemented with 10% (v/v) FBS (Perbio/Thermo Scientific) and 6-thioguanine (Sigma A4882; 10 μg/mL) for 2 weeks. The cells were then counted after being stained using Crystal Violet (RAL diagnostic 317980).
Cells from spleen and lungs, obtained after mechanical disruption, and flushed bone marrow cells were preincubated with anti-CD16/32 (93 clone, Biolegend) and stained for surface marker for 30 minutes at 4°C with the following antibodies: anti-CD45 (30-F11 clone, BD or eBiosciences), anti-CD3e (145-2C11 clone, BD), anti-CD4 (GK1.5 clone, BD), anti-CD8 (53–6.7 clone, BD or eBiosciences), anti-CD19 (1D3 clone, BD), anti-CD11b (M1/70 clone, eBiosciences), anti-CD11c (N418 clone, eBiosciences), anti-CCR4 (2G12 clone Biolegend), anti-CCR6 (29-2217 clone Biolegend), anti-Ly6C (AL21 clone, BD), anti-Ly6G (1A8 clone, BD), anti-F4/80 (BM8 clone, Biolegend), anti-CD107a (LAMP-1; 1D4B clone, BD), and anti-FasL (MFL3 clone, eBiosciences). For cytokine production, cells were first incubated for 4 hours with PMA (P1585-1MG, Sigma), ionomycin (I0634-1MG, Sigma), and Brefeldin A (00-4506-51, Life Technologies). Intracellular staining was performed with the Transcription Factor Staining Buffer Set (00-5523-00, eBiosciences), according to the manufacturer's recommendations. The following antibodies were used: anti-Foxp3 (R16-715 clone, BD), anti-IFNγ (XMG1.2 clone, BD), anti-GzA (GzA-3G8.5 clone, eBiosciences), anti-GzB (GB11 clone, Invitrogen), anti-Ki67 (11F6 clone, Biolegend), and anti-pSMAD2/3 (D27F4 clone, Cell Signaling Technology) coupled with anti-rabbit A488. CD8+ T-cell depletion was checked by flow cytometry on metastatic bone marrow and spleen using anti-CD8α (53–6.7, e-Biosciences). Data were acquired on a LSR-II (BD Biosciences) and analyzed with the FlowJo software version X.
Determination of ERRα-binding sites
ChIP assays were performed as previously described (16) from MDA-MB231-B02-CT and -ERRα cells (18) using either a monoclonal rabbit anti-ERRα (13826) (Cell Signaling Technology) or a control rabbit IgG(2729) antibody (Cell Signaling Technology). The immune-precipitated genomic DNA was purified using NucleoSpin Clean-up columns (Macherey-Nagel) and analyzed by qPCR. Quantification of ChIP enrichment was calculated relative to input values. Distal and proximal elements of ERRα gene were used as negative and positive controls, respectively (25).
Tibia bearing metastases as well as lungs were fixed in 4% paraformaldehyde (Antigenfix Diapath P0014), embedded in paraffin (Histowax Histolab 00403), and then cut (5 μm sections) on a microtome (Microm HM 350S). Immunocytochemical analyses were performed by incubating tissue sections overnight with goat polyclonal antibody ERRα (V-19, Santa Cruz Biotechnology; 1/40), rabbit polyclonal anti-human/mouse CCL17 (PA5-34515, Thermo Fisher; 1/100), rabbit polyclonal anti-mouse CCL20 (ab139585, Abcam; 1/100), and rabbit polyclonal anti-human/mouse-activated TGFβ3 (ab15537, Abcam; 1/100). Sections were then incubated with horseradish peroxidase–conjugated anti-mouse (K4000, Dako) and anti-rabbit (K4002, Dako) according to the manufacturer's recommendations or anti-goat (sc2020, Santa Cruz Biotechnology; 1/300) antibodies for 1 hour and were detected using 3,3′-diaminobenzidine (K3467, Dako) according to the manufacturer's instructions. Counterstaining was performed using Mayer's hematoxylin (Merck) according the supplier's protocol. Lung sections were made at three different depths for each mouse and stained with hematoxylin and eosin (H&E). Metastasis counting was performed in double blind.
Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling assay
Bone sections were deparaffinized and rehydrated followed by permeabilization with 0.2% triton (T9284, Sigma) and digestion with proteinase K (1 μg/mL; K182001; Thermo Fisher). For positive control, sections were incubated with DNAse I at 1 mg/mL (Sigma, 11284932001). Sections were then incubated with biotin-16-dUTP (Sigma, 11093070910) and terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) enzyme (Sigma, 11767305001) in deoxynucleotidyltransferase buffer [Tris-HCl 125 mmol/L (Euromedex, EU0011), sodium cacodylate 200 mmol/L (Sigma, C0250), BSA 6 mmol/L (Sigma, A7906), and CoCl2 1 mmol/L (Sigma, 15862-1ml-F)] at 37°C for 60 minutes in a humid atmosphere. Sections were washed in stop buffer [300 mmol/L NaCl (Sigma, S3014) and 30 mmol/L NaC6H5O7-sodium citrate (Sigma, 71406)] and blocked with 2% BSA (Sigma, A7906). Sections were then labeled with streptavidin–phycoerythrin (eBiosciences, 12-4317-87) and DAPI (Euromedex, 1050-A) and mounted with Fluoromount [Sigma, F4680-25 mL; upright microscope zeiss axioimager (sip 60549)].
Total RNAs from three independent batches of each clone of 4T1, MCF7 and B02 cell (CT, ERRα, and ERRαAF2) were extracted with Trizol reagent (Life Technologies), and 2 μg were reverse-transcribed using qScript cDNA SuperMix (Quanta-Biosciences). Real-time PCR was performed on a Mastercycler-ep-Realplex (Eppendorf) with primers specific to human and mouse genes (Supplementary Table S1) using Quantifast-SYBR-Green (Life Technologies) according to the manufacturer's instructions. The ribosomal protein RPL32 (L32) gene was used as a housekeeping gene for quantification, and relative results expressed as fold differences equal to 2−ΔΔCt.
Data were analyzed statistically using either the nonparametric Mann–Whitney U test or unpaired t test for in vivo studies [n = 10 mice for each group (bioStaTGV), unblinded studies]. In vivo data on bone were confirmed (n = 3) on smaller groups (n = 4). In vitro assays were repeated at least twice and performed on triplicate samples. Data were analyzed using ANOVA and paired Student t test to assess the differences between groups. All data are presented as mean ± SEM with similar variances between groups. Correlation scores for meta-analysis were calculated using the Pearson correlation coefficient. Statistical significance was determined by GraphPad Prism v5.02 using the two-sided Student t test. All statistical analyses were performed using the GraphPad Prism software. P values less than 0.05 were considered statistically significant.
ERRα expression in breast cancer cells inhibits metastases growth in bones
In order to assess the role of ERRα in breast cancer cells after tumor cell anchorage in the bone microenvironment, BALB/c mice were intra-arterially injected with first a pool of three independent 4T1 tumor cell clones overexpressing ERRα (4T1-ERRα) and a pool of two 4T1 tumor cell clones transfected with empty vector controls (4T1-CT; ref. 16). Remarkably, 15 days later, radiographic analysis revealed that animals bearing 4T1-ERRα tumors had osteolytic lesions that were 70% smaller than those of mice bearing 4T1-CT tumors (9.39 ± 2.6 vs. 3.08 ± 1.45 mm2; Fig. 1A and B). The inhibitory effect of ERRα on breast cancer cell growth was associated with mild bone destruction (Fig. 1C–E). Histologic and histomorphometric analyses also demonstrated the limitation of bone metastasis progression when breast cancer cells overexpressed ERRα (Fig. 1F and G). In clear contrast, when clones expressing a dominant-negative form (4T1-ERRαAF2) with their respective controls clones (4T1-CTaf2) were injected, we found a 60% increase in osteolytic lesions in animals bearing ERRαAF2 tumors compared with control (4T1-CTaf2) mice (3.82 ± 1.99 vs. 9.27 ± 2.064 mm2), leading to their earlier sacrifice, prior 4T1-CT(af2) bone metastasis reached the percentage of osteolysis observed in 4T1-CT (Fig. 1H and I). Concomitantly, increased bone destruction and tumor burden were observed (Fig. 1J–N). Given that 4T1 cells also colonize the lung (16), we analyzed the effects of ERRα expression on the development of LM. As opposed to the bone, both numbers of LM and numbers of breast cancer colonies extracted from the lungs were independent of the expression levels of ERRα in 4T1 cells (Fig. 2A–F). Of note, the ERRα overexpression in LM was observed in animals bearing 4T1-ERRα tumors compared with control groups (Fig. 2G). Taken together, this first set of data reveals that the overexpression of ERRα in breast cancer cells prevents their growth in the bone and suggests that ERRα expression in breast cancer may affect the bone microenvironment to prevent bone metastasis progression.
ERRα expression in breast cancer cells increases T-cell antitumor response in the bone
Given the importance of the immune response in the control of tumor growth, in particular metastases (5–8), we next analyzed the effects of the expression of ERRα by breast cancer cells on the bone immune system. It is worth noting that no significant effect was observed on innate cells including dendritic cells, macrophages, with the exception of slight decrease in neutrophils (15%) in the bone colonized by 4T1-ERRα compared with 4T1-CT cells (Supplementary Fig. S3A–S3C). However, we found that metastatic legs of animals bearing 4T1-ERRα tumors contained 5 times more T cells than those colonized with 4T1-CT cells (Fig. 3A). Moreover, in line with the larger 4T1-CT bone metastasis observed (Fig. 1), which developed at the expense of the bone marrow that is largely depicted to sustain all stages of B-cell medullary development (7, 8), the CD19+ compartment was 4 to 5 times lower in mice bearing 4T1-CT bone metastasis compared with 4T1-ERRα bone metastasis (Fig. 3A). Of note, T-cell enrichment was neither observed in the lungs where 4T1-ERRα cells were also anchored (Fig. 3B), nor in the none invaded lymphoid organs such as the spleen (Supplementary Fig. S3D and S3E), implying that the expression of ERRα in breast cancer cells affected the T-cell homeostasis after their anchorage in the bone. Interestingly, the increase in T-cell proportion after 4T1-ERRα cell bone settlement was largely restricted to the CD8+T lymphocyte compartment, with a 3.5-fold increase in their number and percentage in bones bearing 4T1-ERRα tumors compared with bone colonized with 4T1-CT cells (Fig. 3A), whereas the proportion of CD4+ T cells in the bone, including that of Foxp3+ regulatory T cells, was unaffected (Supplementary Fig. S3C).
In order to further characterize the CD8+ T cells overrepresented in the bone after 4T1-ERRα cell colonization, we then analyzed their ability to produce cytotoxic molecules and cytokines. Strikingly, in bone colonized with 4T1-ERRα cells, we found that CD8+ T cells expressed higher levels of FasL and granzyme A and B in association with LAMP1 at their cell surface as well as IFNγ, whereas these molecules were barely detectable in CD8+ T cells from bones bearing 4T1-CT cells (Fig. 3C–E). In total agreement with this exacerbated cytotoxic program of CD8+ T in the presence of 4T1-ERRα breast cancer cells in the bone, we found that a large fraction of 4T1-ERRα bone metastasis underwent apoptosis compared with 4T1-CT cells (Fig. 3F and G). No exacerbated sign of cytotoxic activity was observed in the noninvaded spleen or in the metastatic lungs of mice bearing 4T1-ERRα cells (Supplementary Fig. S4A). Importantly, the depletion of CD8+ T cells was sufficient (Fig. 4A) to increase 4T1-ERRα bone metastasis progression (Fig. 4B–E). This set of data suggests that the expression of ERRα in breast cancer cells influences the CD8+ T-cell homeostasis and increases their antitumor cytotoxic program in the bone allowing the control of the tumor progression.
ERRα expression leads to high levels of CCL17 and CCL20 production in breast cancer cells
The aforementioned data strongly suggest that the expression of ERRα by breast cancer cells influences the bone microenvironment to promote an efficient antitumor response. Interestingly, we failed to find any difference in Ki67 staining between CD8+ T cells evolving with either 4T1-CT or 4T1-ERRα bone metastasis (Supplementary Fig. S4B), suggesting that ERRα expression affected T-cell recruitment to the bone rather than their proliferation in the bone. In order to address this hypothesis, we monitored the expression levels of several chemokines known to influence T-cell chemoattraction and found a 2- and 3-fold upregulation of Ccl17 and Ccl20 in 4T1-ERRα cells, respectively (Fig. 5A and B; Supplementary Fig. S5A and Supplementary Table S2). C29 or XCT-790 were sufficient to inhibit the overexpression of both Ccl17 and Ccl20, whereas no effect was observed in 4T1-ERRαAF2 cells ruling out any off-target effects (Fig. 5C and D; Supplementary Fig. S5B and S5C; ref. 26), arguing in favor of a direct role for ERRα in the control of the expression of these two chemokines (13). Of note, the ability of ERRα to upregulate CCL17 and CCL20 was also observed in other breast cancer cells including MCF7 and MDA-MB-231-B02 cells (Supplementary Fig. S5D–S5E; refs. 16, 18). Interestingly, the upregulation of CCL17 and CCL20, at both gene and protein levels due to ERRα overexpression, was sustained in 4T1-ERRα bone metastasis (Fig. 5E and F) but lost in LM (Supplementary Fig. S5F). These data were reinforced following the analysis of ChIP-seq data revealing binding site for ERRα in the promoter of Ccl17 and in Ccl20 (Supplementary Fig. S6A–S6D).
Given that CCL17 and CCL20 were reported to attract the fraction of activated CD8+ T cells expressing CCR4 and CCR6 (27), we next monitored the expression of both chemokine receptors on CD8+ T cells from the bone of animals bearing bone metastasis. In total agreement with the ability of 4T1-ERRα cells to sustain their production of CCL17 and CCL20 in the bone, and the activated phenotype of CD8+ T cells (Fig. 3), we found that, in bone colonized by 4T1-ERRα, a large fraction of CD8+ T cells expressed either CCR4 or CCR6 or both (Fig. 5G). Thus, breast cancer cells overexpressing ERRα are endowed with a unique ability to produce high amounts of CCL17 and CCL20 and efficiently recruit activated CD8+ T cells to the bone.
ERRα expression in breast cancer cells reduces their TGFβ3 production and decreases TGFβ signaling in bone metastasis
Because the cytotoxic program of CD8+T cells was largely exacerbated in legs bearing 4T1-ERRα cells, we next assessed the mechanisms by which ERRα overexpression in breast cancer cells increased their cytotoxic function in the bone. To this end, we further investigated the connection between CCL17-CCL20 and ERRα (ESRRA) by choosing a global approach combining bioinformatic analyses of protein interaction networks and transcriptional regulator databases (Supplementary Fig. S1, extended network, https://doi.org/10.18119/N9W891; Supplementary Fig. S2, minimal network, https://doi.org/10.18119/N9RK5C; ref. 28). We created the “Minimal Network specific to immune response to tumor” (containing 52 proteins; Fig. 6A), and by shortest path analysis, we then identified two new ESRRA-CCL17– or ESRRA-CCL20–associated regulators: VCAM and TGFβ3 (Fig. 6B and C). We and others reported that TGFβ signaling in T cells inhibits the cytotoxic differentiation program of CD8+ T cells both in humans and mice (29, 30), and we thus focused on this cytokine. The analysis of Tgfβ3 expression revealed a 70% decrease in 4T1-ERRα compared with 4T1-CT cells (Fig. 6D). Ex vivo bone cultures confirmed that ERRα upregulation in breast cancer cells negatively regulates Tgfβ3 expression with a 55% decrease compared with control bone metastasis (Fig. 6E). Strikingly, immunohistological staining revealed that the production of TGFβ3 was largely decreased in breast cancer bone metastasis overexpressing ERRα (Fig. 6F). In agreement with the decrease of Tgfβ3 expression in 4T1-ERRα, we observed an increase in Tgfβ3 levels in 4T1-ERRαAF2 bone metastasis (Fig. 6D) and a 5-fold increase in Tgfβ3 expression after treatment of 4T1-ERRα cells with the inverse agonist XCT-790 (Fig. 6G). Similar results were observed in MCF7 and MDA-MB-231-B02 human cell lines (Fig. 6H), ruling out an effect restricted to mouse 4T1 cells. Altogether, these results reveal that the overexpression of ERRα represses the expression of TGFβ3 in breast cancer and could thus prevent the breast cancer bone metastasis from creating an immunosuppressive microenvironment provided by TGFβ signal activation in immune cells.
In order to unconditionally confirm that CD8+ T cell evolving in 4T1-ERRα colonized bone escape TGFβ signaling control, we next analyzed CD8+ T cells from metastatic legs for the phosphorylation of SMAD2/3 proteins, which translates specifically the TGFβ signaling activation (31). Clearly, SMAD2/3 phosphorylation was 2 to 3 times lower compared with that of CD8+ T cells from 4T1-CT metastatic legs (Fig. 6I). Thus, in addition to increasing CD8+ T-cell recruitment to the bone, ERRα overexpression in breast cancer cells impairs their ability to produce high amounts of TGFβ3, decreasing TGFβ signaling in CD8+ T cells, a key repressor of their cytotoxic activity and capacity to eliminate cancer cells.
Overexpression of ESRRA (ERRα)in patient tumors is associated with high levels of CCL17, CCL20, and low levels of TGFβ3 expression
Finally, we addressed the relevance of our data obtained in mice to the human pathology. We performed a meta-analysis on 248 patients including luminal and triple-negative breast tumors (TNBC) split into four groups: all tumors (n = 248), patients without metastases (No Mets, n = 121), patients that had visceral and bone metastases (Visceral+Bone metastases, n = 53), and patients that had only bone metastases (Bone Only, n = 74). As in mice, the ESRRA (ERRα) expression was positively correlated with that of CCL17 and CCL20 and inversely proportional to that of TGFβ3 in patients with metastases restricted to the bone (Bone Only) with luminal and TNBC tumors (Table 1A and B). Correlations were also identified in No Mets, All, and Visceral + Bone groups of luminal or TNBC patients (Table 1A and B). In addition, ERRα expression was not associated with LM in two cohorts of patients with breast cancer (16). Thus, this set of data from human sample analyses strongly suggests that, similarly to mice, the overexpression of ERRα in human breast cancer cells allows them to create an immune-efficient environment in the bone by increasing the production of chemokines capable of attracting activated T cells to the bone and decreasing the production of TGFβ essential for repressing the cytotoxic activity of T cells.
|.||All (N = 248) luminal .||No mets (N = 121) .||Visceral + bone (N = 53) .||Bone only (N = 74) .|
|Correlation with ESRRA .||R .||P value .||R .||P value .||R .||P value .||R .||P value .|
|All (N = 110) triple negative||No mets (N = 50)||Visceral + bone (N = 53)||Bone only (N = 16)|
|Correlation with ESRRA||R||P value||R||P value||R||P value||R||P value|
|.||All (N = 248) luminal .||No mets (N = 121) .||Visceral + bone (N = 53) .||Bone only (N = 74) .|
|Correlation with ESRRA .||R .||P value .||R .||P value .||R .||P value .||R .||P value .|
|All (N = 110) triple negative||No mets (N = 50)||Visceral + bone (N = 53)||Bone only (N = 16)|
|Correlation with ESRRA||R||P value||R||P value||R||P value||R||P value|
Note: Meta-analysis of public datasets (GSE12276-GSE2034-GSE2603; n = 248) revealed a positive correlation between the expression of ESRRA and CCL17 and CCL20 and a negative correlation with TGFβ3 expression levels in luminal (A) and triple-negative (B) breast tumors; correlation scores were calculated using the Pearson correlation coefficient. P values less than 0.05 were considered statistically significant and are in bold.
Cancer cells adapt to the microenvironment, shaped by their own doing, which in turn influences their fate. This interplay is particularly important for cells forming metastases, which leave their primary microenvironment to settle in a new, second one. Here, we revealed that the level of ERRα expression on breast cancer metastases promotes their ability to condition an efficient antitumor CD8+ T-cell response selectively in the bone.
CD8+ T cells have been described as critical inhibitors of bone metastases. Indeed, in mice, the alteration of CD8+ T-cell development after metastases implantation in the bone, or the deprivation of CD8+ T cells, was reported to increase tumor growth (5, 6). Osteoclasts have been depicted to secrete chemokines that can attract CD8+ T cells (32). However, the regulation of the bone metastasis burden by CD8+ T cells seems totally independent of the osteoclast activity (6). Our study reveals that the cancer cells per se can influence both the recruitment and the cytotoxic activity of the CD8+ T cells in the bone. Moreover, the ability of the breast cancer metastases to condition the immune response in the bone can be in part orchestrated by the levels of expression ERRα on the breast cancer and potentially to the sensitivity of metastases to the ERRα ligand(s). The selective effects of ERRα expression in breast cancer on the tumor burden of bone metastasis and antitumor response in the bone strongly suggest that unlike the lung, the bone could constitute a microenvironment with high levels of the ERRα ligand(s) that so far remain uncharacterized. Another alternative is that the lung, but not the bone, could be highly enriched in inhibitors of ERRα signaling or negative regulators of ERRα expression that remain to be identified. Thus, the ability of the metastases to induce or not a potent immune response maybe dictated by both the tumor per se and the microenvironment where it is anchored. In the case of breast cancer cells, we propose to place ERRα at the core of this interplay between metastases and their new microenvironment.
In addition to increasing the recruitment of activated CD8+ T cells to the bone, the overexpression of ERRα on breast cancer cells also decreased their ability to produce high amounts of TGFβ3. Depletion experiments confirmed that CD8+ T cells are key antitumor immune cells whose activation and recruitment are controlled by the levels of ERRα expression on bone metastasis. All forms of TGFβ have been reported as potent immune-regulators and share a common receptor (33). Although TGFβ1 is predominant in the immune system, TGFβ3 is mainly produced by muscles, bones but also by various cancer cells (34). The repression of TGFβ3 production in ERRα breast cancer cells subsequently affects TGFβ signaling in CD8+ T cells present in the bone. TGFβ signaling represses the expression of numerous transcription factors associated with cytotoxicity, as well as T-Bet, a key inducer of IFNγ (35). Therefore, overexpressed ERRα breast cancer cells that settle in the bone are unable to sustain an immunosuppressive microenvironment based on high levels of TGFβ signaling in T cells and repression of cytotoxic program and IFNγ production. Interestingly, IFNγ also contributes to the suppression of bone metastasis. Indeed, IFNγ has been reported to reduce both RANKL expression and osteoclast formation, counterbalancing the aberrant bone resorption, which facilitates tumor growth (36). Concomitantly, inhibition of bone resorption also leads to the decrease in TGFβ release from the bone matrix (4), thus potentially contributing to amplifying the activation of T cells including their production of IFNγ.
It is likely that effector CD8+ T cells that reach the bone metastasis have previously been primed in the draining lymph nodes or by the spleen-presenting antigens from the primary tumor and/or the metastases. As in the primary tumor, the activated CD8+ T-cell population in contact with the bone metastasis is actually heterogeneous and composed of cells recently activated and activated memory cells. Interestingly, in both mice and humans, the fraction of CD8+ T cells that expresses CCR6 and CCR4 has been depicted to rapidly mount an efficient response, corresponding to activated/memory T cells (27). Once implanted in the bone, we found that the breast cancer overexpressing ERRα has unique ability to sustain high expression of CCL17 and CCL20 and low expression of TGFβ3, thus attracting the activated/memory CD8+ T cells whose antitumor cytotoxic function is magnified by the lack of repression by TGFβ signaling.
In conclusion, this study assigns an unsuspected role for ERRα expression in breast cancer on the bone immune system that conditions the bone metastasis growth outcome, providing the mechanistic basis for understanding how ERRα expression in breast cancer can affect the bone microenvironment and reduce bone metastasis growth. ERRα seems to appear at the core of this interplay between breast cancer metastases and their new environment, integrating signals from the microenvironment to develop an efficient antitumor response. Therefore, we propose to consider ERRα expression on breast cancer as a biomarker predictive of bone metastasis response to immunotherapies and/or as a good prognosis marker in bone metastasis progression once established, opening the path toward to the clinical use of ERRα agonist to relieve patients with ERRα-positive bone metastasis after primary tumor resection.
Disclosure of Potential Conflicts of Interest
P. Clézardin reports receiving honoraria from the speakers' bureau of Amgen. No potential conflicts of interest were disclosed by the other authors.
Conception and design: J.C. Marie, E. Bonnelye
Development of methodology: M. Bouchet, C. Boyault, M. Proponnet-Guerault, L. Bouazza, S. Geraci, S. El-Moghrabi, C. Benetollo, J.C. Marie, E. Bonnelye
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Bouchet, A. Lainé, M. Proponnet-Guerault, L. Bouazza, S. Geraci, S. El-Moghrabi, M. Duterque-Coquillaud, J.C. Marie, E. Bonnelye
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Bouchet, A. Lainé, C. Boyault, E. Meugnier, C.W.S. Kan, H. Hernandez-Vargas, M. Duterque-Coquillaud, J.C. Marie, E. Bonnelye
Writing, review, and/or revision of the manuscript: A. Lainé, C. Boyault, H. Hernandez-Vargas, Y. Yoshiko, M. Duterque-Coquillaud, P. Clézardin, J.C. Marie, E. Bonnelye
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Bouchet, M. Proponnet-Guerault, M. Duterque-Coquillaud, J.C. Marie, E. Bonnelye
Study supervision: J.C. Marie, E. Bonnelye
This work was supported by the National Center for Scientific Research (CNRS) to E. Bonnelye, the National Institute of Health and Medical Research (INSERM), the University of Lyon1, La Ligue Nationale (Drôme; E. Bonnelye), Inserm-Transfert (E. Bonnelye), and La Ligue Nationale contre le cancer labellisation EL2016 (J.C. Marie). M. Bouchet was supported by the French National Cancer Institute (INCa); A. Lainé, J.C. Marie, and P. Clézardin by the Labex DEVweCAN (ANR-10-LABX-61); M. Proponnet-Guerault and A. Lainé by La Ligue Nationale contre le cancer; and C.W.S. Kan by the Marie-Curie-Individual-Fellowship (655777-miROMeS). The authors thank J. Ribiero, C. Scote-Blachon, G. Amorim, Geoffrey Vargas, E. Gineyts, A. Flourens, and M. Gervais for technical help and A. Emadali and CeCIL, ALECS, flow cytometry, and histology platforms of CRCL (Faculté de Médecine Laennec, Lyon, France) for their assistance. The editing work and advice of B. Manship are acknowledged.
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.