Matrix-Targeting Immunotherapy Controls Tumor Growth and Spread by Switching Macrophage Phenotype

The immune system can detect tumors and mount an efficient defense against them, but tumors can evolve to develop effective strategies to evade immune elimination. The importance of the interplay between tumor cells and immune cells in determining whether cancers will progress provided a landmark shift in disease philosophy, opening therapeutic avenues beyond targeting only tumor cells. The first wave of immuno-oncology therapies focuses on tumor-infiltrating lymphocytes (TIL). TILs reliably predict good prognosis and sensitivity to neoadjuvant chemotherapy treatment in many solid cancers (1) and correlate with high mutational burden (2). Therapies that reactivate TILs, via engaging their immune checkpoints, expanding mutation-specific T cells, or injecting antigen-specific T cells, have revolutionized the treatment of cancer, offering an alternative tumor killing mechanism, as well as the possibility of generating long-lasting immunity against the cancer. However, this approach does not work for all patients, nor all types of tumors, and can be associated with severe autoimmune side effects (reviewed in ref. 3).

In contrast to TILs, tumor-associated macrophages (TAM) are frequently associated with poor outcomes in patients, prompting the development of therapies directed against these cells (reviewed in refs. 4–6). Approaches include inhibition of macrophage survival, proliferation, and differentiation by targeting colony-stimulating factor receptor-1 (CSFR-1), blockade of their recruitment by targeting CCR2 (chemokine C-C motif receptor 2), modulating their polarization by targeting macrophage receptor with collagenous structure (MARCO), or their activation by engaging the costimulatory signal CD40. Macrophage-targeting therapies have been hailed as a useful complement for immune-checkpoint therapies such as anti–CTLA-4 and anti–PD-1/PD-L1 treatment, providing a multihit approach to enable parallel lymphocyte reactivation and TAM blockade, as well as synergistically widening T-cell repertoire and activity. However, the clinical efficacy of macrophage-targeting drugs has been mixed (4, 5). These results may, in part, be due to the complexity of the macrophage compartment. Although many cancers are infiltrated by immunosuppressive TAM populations, a variety of other TAM subsets, including tumoricidal macrophages that exert killing or phagocytic activity, metastasis-associated macrophages, and proangiogenic macrophages, also exist (7). It is not clear what dictates TAM heterogeneity but understanding more about the factors in the tumor microenvironment (TME) that influence macrophage phenotype may enable the development of therapies that specifically target pathogenic TAM subsets.

The tissue microenvironment is a key determinant of macrophage phenotype. Tissues comprise a selection of secreted proteins, encompassing extracellular matrix molecules, matrix-associated proteins, growth factors, chemokines and cytokines, and enzymes including proteases (8). These complex 3D networks provide essential context for cell behavior, driving site-specific gene-expression programs to enable geographically adapted cell behavior. Emerging evidence is starting to reveal factors underlying macrophage specialization in the liver, gut, and brain (9) and how macrophage transplantation from the peritoneum to the lungs reshapes tissue-dependent functions and characteristics (10). Macrophage responses to infection and how soluble factors, such as GM-CSF and IL4, can influence macrophage behavior are also known (11). Most solid cancers are made up of a tumor-specific extracellular matrix that has a different composition and organization to that of healthy tissue (12). However, how this altered microenvironment affects macrophage behavior is not fully understood.

Here, we examined tenascin-C, a matrix molecule absent from most healthy tissues, but whose expression in solid tumors is frequently associated with poor prognosis (13). This large multimodular molecule has been shown to modulate angiogenesis (14, 15), stem cell fitness (13), and tumor stiffness (16) during tumor growth and metastasis. Tenascin-C can also shape innate and adaptive immune responses (17), but its role in tumor immunity remains unclear. In this study, we explored the impact of tenascin-C on the immune axis in breast cancer and examined the therapeutic potential of specifically blocking its immunomodulatory action.

Wild-type FVB mice were purchased from Charles River Laboratories, and tenascin-C knockout mice on an FVB background were generated as described in ref. 18. Two- to 3-month-old females were used for the experiments. MMTV-NeuNT female mice (FVB/NCrl background) with a mutated constitutively active form of rat ErbB2 (NeuNT) expressed under the control of the mouse mammary tumor virus (MMTV) regulatory region were purchased from The Jackson Laboratory. All animal procedures were carried out in accordance with the UK Animals (Scientific Procedures) Act 1986 and with the University of Oxford (Clinical Medicine) Ethical Committee approval.

The mouse mammary cancer cell line NT193 was derived from an MMTV-NeuNT tumor (19) and has previously been engineered to express high or low/no tenascin-C by transduction with a lentivirus expressing shRNA against tenascin-C (TNC^–) or control shRNA (TNC⁺). Lentiviral particles with shRNA vectors (Sigma-Aldrich) specific for Tnc were used (sh2, TRCN0000312138, 5′-CCGGGCATCAACACAACCAGTCTAACTCGAGTTAGACTGGTTGTGTTGATGCTTTTTG-3′). Lentiviral particles encoding a nontargeting shRNA vector were used as a control (SHC202V, Sigma-Aldrich; ref. 18). NT193 cells were cultured in DMEM supplemented with 10% inactivated fetal bovine serum, penicillin (10,000 U/mL), streptomycin (10 mg/mL), and puromycin (10 μg/mL; all from Sigma-Aldrich). Cells were maintained at 37°C in a humidified atmosphere of 5% CO₂ and were cultivated for a minimum of 1 week and a maximum of 1 month (roughly 4 passages) before use. Cells were periodically checked for tenascin-C and vimentin expression by Western blotting and tested to be negative for Mycoplasma (Thermo Fisher).

NT193 cells were lysed using RIPA buffer with protease inhibitor cocktail I (12801640; Fisher Scientific) at 1:1,000, and the protein concentration was determined using a BCA assay (Thermo Fisher). Five percent polyacrylamide gels were prepared, and protein (20 μg/well) was loaded in Laemmli buffer diluted 1/6 (stock solution: 2.5 mL of 0.5 M Tris-HCl pH 6.8; 1.2 g SDS; 2.5 mL β-mercaptoethanol; 5 mL glycerol; 5 mg bromophenol blue). Separated proteins were transferred to a nitrocellulose membrane using a TransBlot Turbo transfer machine (Bio-Rad). Membranes were blocked for 1 hour in blocking buffer (PBS, 0.05% Tween20, 5% BSA) and incubated overnight with primary antibodies recognizing tenascin-C (MTn-12; Sigma-Aldrich) at 1/4,000, vimentin (V5255, Sigma-Aldrich) at 1/2,000 or beta actin (937215; Enzo Life Science) at 1/1,000, and then for 1 hour with secondary antibody (anti-rat IgG A5795; Sigma-Aldrich, anti-mouse IgG P0260; DAKO) at 1/20,000 in blocking buffer. Proteins were detected with ECL Amersham reagent using Amersham Hyperfilm ECL (VWR International) and a Curix 60 film processor (AGFA).

For the orthotopic grafting model, 10 × 10⁶ NT193 TNC⁺ or TNC^– cells were injected into the left fourth mammary gland of 2- to 3-month-old female FVB mice in 50 μL sterile PBS at day 0. Tumor size was measured every 3 days using calipers, and the volume of the tumor was calculated using the formula V = (L × H × W)/2. For the spontaneous model, the apparition of mammary tumors in MMTV-NeuNT mice was monitored from 4 months of age at least once a week, and the experiment started once a tumor became detectable by palpation (day 0). Clodronate liposomes and control liposomes (Liposoma) were prepared according to the manufacturer's recommendations and injected intravenously at 10 μL/g of mouse or intratumorally at 1 μL/g of mouse on days 1 and 7 after engraftment of NT193 TNC⁺ or TNC^– tumor cells. The efficacy of myeloid cell depletion was tested by flow cytometry staining of blood sampled from the tail vein, 7 days after the last injection, by detection of CD45⁺CD11b⁺ cells (using anti-CD45–PE-Cy7 (30F11) and anti-CD11b–PerCP-Cy5.5 (M1/70) from BioLegend). Antibodies recognizing the fibrinogen-like globe (FBG) domain of tenascin-C (anti-FBG; Clone C3, Nascient Ltd.; ref. 20) was given as 6 intraperitoneal (i.p.) injections on days 1, 4, 7, 11, 15, and 18 at 5, 10, or 20 mg/kg in 200 μL sterile PBS. Anti–PD-L1 antibody (10F.9G2; Bio X Cell) was given as 2 i.p. injections of 150 μg in 200 μL sterile PBS on days 1 and 4 (21). Mice were sacrificed when the tumor reached 1.2 cm, and tumors and lungs were collected.

Tissues from paraffin-embedded lung lobes were cut from 46 MMTV-NeuNT mice at 6 different depths separated by at least 500 μm from each other throughout the tissue (sagittal sections; 7 μm) and stained with hematoxylin and eosin to assess lung pathology. The number of metastasis per image were counted, from a minimum of 10 images per section, on an AxioScope.A1 (Zeiss) microscope, at 20× magnification, and their size was measured using ImageJ software.

Tumor tissue was cut into small pieces and digested in RPMI medium supplemented with 5% of inactivated fetal bovine serum, penicillin (10,000 U/mL), streptomycin (10 mg/mL), Liberase TM (500 μg/mL; Roche), and DNAse (100 μg/mL; Roche) for 30 minutes at 37°C under agitation. Cells were plated at 1 × 10⁶ cells/mL in complete DMEM supplemented with 10% of inactivated fetal bovine serum, penicillin (10,000 U/mL), and streptomycin (10 mg/mL) for 5 days. Puromycin (10 μg/mL) was added to select for tumor cells. Adherent cells were collected, and cell lysates prepared in RIPA buffer supplemented with protease inhibitor cocktail I (12801640; Fisher Scientific) at 1/1,000.

Harvested tumors were frozen in liquid nitrogen and mechanically disrupted before being lysed in RIPA buffer supplemented with protease inhibitors, as indicated above. Tenascin-C concentration in the lysates was assessed by ELISA (Tenascin-C Large FnIII-B kit, Demeditec). Two hundred nanograms of protein/well was loaded, and the optic density was measured using a FLUOstar Omega plate reader (BMG Labtech). Tenascin-C concentrations were calculated from a standard curve prepared with recombinant tenascin-C following the manufacturer's recommendations.

Tumor tissue was cut into small pieces and digested in RPMI medium supplemented with 5% of inactivated fetal bovine serum, penicillin (10,000 U/mL), streptomycin (10 mg/mL), Liberase TM (500 μg/mL; Roche), and DNAse (100 μg/mL; Roche) for 30 minutes at 37°C under agitation. Surface and intracellular flow cytometry staining of isolated cells was performed according to standard protocols and analyzed on an LSRFortessa (BD Biosciences). Briefly, 1.10⁶ cells were saturated for 15 minutes in the presence of Fc Block (TruStain FcX, 101319, BioLegend) and incubated for 25 minutes with the surface antibodies mix diluted at 1/200 in PBS 2% FCS. For intracellular staining, cells were then permeabilized and fixed for 30 minutes in the presence of a Fix/perm solution (554722, BD Biosciences) and incubated for 30 minutes with the intracellular antibodies diluted at 1/100 in the perm/wash buffer (554723, BD Biosciences). Surface antibodies: anti–Ly-6C–Brilliant Violet 785 (HK1.4), anti–Ly-6G–Brilliant Violet 650 (1A8), anti-CD206–Alexa Fluor 700 (C068C2), anti-CD11c– Alexa Fluor 594 (N418), anti-CD115–APC/Cy7 (AFS98), anti-CD45–PE-Cy7 (30F11), anti-CD11b–PerCP-Cy5.5 (M1/70), anti-F4/80–Pacific Blue (BM8), anti-CD8α–Pacific Blue (53-6.7), anti-CD3–Brilliant Violet 785 (17A2), anti-CD86–Brilliant Violet 650 (GL-1), anti-TLR4–APC (SA15-21) from BioLegend; anti-IA/IE–FITC (2G9) from BD Biosciences; and anti-CD4–PE-Texas Red (GK1.5) from Abcam. Intracellular antibodies: anti-IL17A–Alexa Fluor 700 (TC11-18H10.1) from BioLegend; anti-IRF5–PE (903430) from R&D Systems; and anti-IFNγ–Alexa Fluor 594 (XMG1.2) from BD Biosciences. Tenascin-C was detected using anti-FBG at 10 μg/mL followed by anti-human IgG-AF488 (BioLegend). Dead cells were stained using a Live/Dead yellow kit (Thermo Fisher). FlowJo software version 10 was used for data analysis. Cell populations were defined as follows: M1-like macrophages: CD45⁺CD11b⁺CD11c⁺F4/80⁺IRF5⁺; M2-like macrophages: CD45⁺CD11b⁺CD11c⁺F4/80⁺CD206⁺; monocytes: CD45⁺CD11b⁺ CD115⁺Ly6C⁺; neutrophils: CD45⁺CD11b⁺CD115^–Ly6C⁺Ly6G^hi; Th1: CD45⁺CD3⁺CD4⁺IFNγ⁺; Th17: CD45⁺CD3⁺CD4⁺IL17A⁺.

Optimal cutting temperature–embedded tumor sections (7 μm-thick) were incubated with blocking solution [PBS, 5% goat serum (G9023, Sigma-Aldrich), 5% rat serum (R9759, Sigma-Aldrich), 1% FCS] for 1 hour at room temperature before incubation with primary antibody overnight at 4°C [anti–tenascin-C from Sigma-Aldrich (MTn-12)) at 10 μg/mL; anti–F4/80-PE (BM8) and anti–CD206-APC (C068C2) from BioLegend] both at 1/100. The slides were then incubated with secondary antibody for 1 hour at room temperature (anti-IgG1-AF488, BioLegend) at 1/200, counterstained with DAPI (Thermo Fisher), and embedded with Fluoromount-G (Invitrogen). The fluorescent signal was analyzed with a Zeiss Axio Imager microscope. Five random 20× fields per section were analyzed by ImageJ (NIH) to assess the infiltration of macrophages, as well as their spatial distribution between the tumor nest and the stroma (1 section per tumor, 5 tumors per group). Macrophage subsets and tenascin-C localization in the tumor were also analyzed and displayed as a line scan using FiJi software.

Total RNA was prepared from harvested tumors using TriReagent (Life Technologies) with a tissue homogenizer (Omni International) according to the manufacturer's instructions. RNA integrity was assessed with an Agilent total RNA Pico Kit on a 2100 Bioanalyzer instrument (Agilent Technologies; RINs ranged from 7.6 to 8.2), and ribosomal RNA was depleted with the Low Input RiboMinus Eukaryote System v2 kit (Thermo Fisher). To prepare the sequencing library, the Ion Total RNA-sep kit v2 (Thermo Fisher) was used following the manufacturer's instructions. The libraries were loaded at a concentration of 20 pmol/L on an Ion PI Chip (Thermo Fisher) using the Ion Chef Instrument (Thermo Fisher). Sequencing was performed on an Ion Proton sequencer (Thermo Fisher) with the Ion PI Hi-Q Sequencing 200 Kit (Thermo Fisher). Transcriptomic data were processed using the RNASeqAnalysis plugin from the Torrent Suite Software 5.06, and the reads were mapped using STAR and Bowtie2 (22, 23). The total read maps are available in binary alignment map (BAM) format for raw read count extraction. Read counts were found with the htseq-count tool of the Python package HTSeq (24). Differential analyses were performed using the DESEQ2 package from the Bioconductor framework (25). Upregulated and downregulated genes were selected based on the adjusted P-value cutoff 10%. Data can be accessed via SRA accession number PRJNA587450.

Tumor-infiltrating macrophages were isolated by mechanical dissociation of the tumor tissue and enzymatic digestion with DMEM supplemented with 5 % of inactivated fetal bovine serum, penicillin (10,000 U/mL), streptomycin (10 mg/mL), Liberase TM (500 μg/mL; Roche), and DNAse (100 μg/mL; Roche) for 30 minutes at 37°C under agitation and then purified by sorting CD11b⁺F4/80⁺ cells using a BD FACSAria III. RNA isolation was performed using the RNeasy Microkit (Qiagen). The concentration and purity of the isolated RNA was determined using a NanoDrop One (Thermo Fisher). One hundred nanograms of RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription kit (Thermo Fisher). The real-time qPCR reaction was performed in 384-well plates on a Viia7 Real-Time PCR system (Thermo Fisher) using the TaqMan Universal PCR Master Mix (Thermo Fisher) and the following TaqMan primers (all Thermo Fisher) in triplicate: ARG-1: Mm00475988_ m1, CD86: Mm00444540_m1, MRC1: Mm01329362_m1, TNF: Mm99999068_m1, HPRT: Mm00446968_m1. Data were analyzed using the ΔCt method and expressed as a percentage of HPRT.

T cells were isolated from spleens of naïve female wild-type FVB mice (10–12 weeks of age) by mechanical dissociation of the tissue, purified by negative magnetic sorting (Pan T-cell isolation kit II, 130-095-130 and LS columns, 130-042-401, Miltenyi Biotec), and loaded with 10 μmol/L of CFSE (BD Biosciences) on the first day of coculture. The purity of sorted cells was systematically assessed by FACS to be >95%. Coculturing conditions are described below. Proliferation was calculated using flow cytometry (BD LSRII) and analyzed with FlowJo software (Tree Star) to assess the decreasing mean fluorescence intensity (MFI) of CFSE in T cells on the last day of coculture, as CFSE intracellular content is reduced upon each cell division.

Bone marrow–derived macrophages (BMM) were isolated by flushing the bone marrow out of the femur of naïve wild-type FVB mice. BMM were then differentiated by plating 2 × 10⁶ cells for 7 days in the presence of GM-CSF (50 ng/mL; PeproTech) and adherent/loosely adherent cells were collected using ice-cold PBS. Given that the tumor-associated macrophage population we identified in our model was CD11b⁺CD11c⁺, we chose GM-CSF–differentiated BMM that are CD11b⁺CD11c⁺ rather than M-CSF–differentiated BMM that only express CD11b. We observed comparable results in coculture systems using both cell subsets.

Tumor-infiltrating myeloid cells were isolated by mechanical dissociation of the tumor tissue from mice and enzymatic digestion, as stated above, for 30 minutes at 37°C under agitation and then purified by positive magnetic sorting of CD11b⁺ cells (CD11b microbeads, 130-049-601 and LS columns, 130-042-40, Miltenyi Biotec). Coculturing conditions are described below.

NT193 tumor cells were detached from tissue culture plates and treated with mitomycin (50 μg/mL; Sigma-Aldrich) to block their proliferation. 1 × 10⁶ tumor cells/mL were cultured in RPMI medium supplemented with 5% of inactivated fetal bovine serum, penicillin (10,000 U/mL), streptomycin (10 mg/mL), and 10 μmol/L of β-mercaptoethanol with BMM, CD11b⁺ tumor–infiltrating myeloid cells, or T cells using all cells at a 1:1 ratio. Tumor cells and BMM or CD11b⁺ tumor–infiltrating myeloid cells were cultured for 5 days, and tumor cells with T cells were cultured for 10 days. Murine IL2 (50 ng/mL; PeproTech), IL7 (50 ng/mL; PeproTech), and LEAF-grade anti-CD3 (BioLegend) at 1 μg/mL were added in T-cell cocultures. The following inhibitors were used: TLR4 inhibitor (TAK242, Invivogen) was added at 5 μmol/L, EGFR inhibitor (Tyrphostin, Sigma-Aldrich) was added at 1 μmol/L, and integrin β1 and β3 inhibitor (Echistatin, Sigma-Aldrich) was added at 50 nmol/L to BMM cocultures every 5 days. In some experiments, LPS was added at 1 ng/mL (L2887, Sigma-Aldrich) on the first day of coculture. Anti-FBG antibody was added to all cocultures every 5 days of coculture at a molar ratio corresponding to 1:10, 1:1, or 5:1 to the concentration of tenascin-C produced (2 ng/mL, 260 ng/mL, or 1.3 μg/mL, respectively) by TNC⁺ tumor cells over 5 days of culture.

Culture supernatants were collected at the end of cocultures as detailed in the culture conditions. Murine IL1β, IL4, IL6, IL8, IL10, IL12, TNFα, and TGFβ (DuoSet ELISA kits, R&D Systems) and tenascin-C (Tenascin-C Large FnIII-B kit, Demeditec) were quantified by ELISA. Supernatants were diluted at 1/2 for IL1β, IL4, IL10, IL12, and TGFβ, at 1/5 for IL6, IL8, and TNFα and at 1/20 for tenascin-C. The optic density was measured using a FLUOstar Omega (BMG Labtech) plate reader, and concentrations were calculated from a standard curve using standards provided in the kit following the manufacturer's recommendations. Flow-cytometric analysis was performed on the cells extracted from cocultures or from harvested mouse organs on a BD LSRII or a LSRFortessa using the following antibodies: anti–CD206-Alexa Fluor 700 (C068C2), anti–CD45–PE-Cy7 (30F11), anti–CD11b–PerCP-Cy5.5 (M1/70), anti–F4/80-Pacific Blue (BM8) anti–CD8α- Pacific Blue (53-6.7), anti–CD3-Brilliant Violet 785 (17A2), anti–CD86- Brilliant Violet 650 (GL-1) from BioLegend; anti–iNOS-PE-eFluor 610 (61-5920-80) from Thermo Fisher; anti-IA/IE–FITC (2G9) from BD Biosciences; and anti-CD4- PE-Texas Red (GK1.5) from Abcam. Dead cells were stained using a Live/Dead yellow kit (Thermo Fisher). Flow cytometry analysis was performed using FlowJo software (Tree Star).

Bulk RNA sequencing (RNA-seq) data of 1,045 primary tumors from The Cancer Genome Atlas (TCGA) Breast Invasive Carcinoma cohort (BRCA-US) were retrieved from the Genomic Data Commons via the TCGA biolinks package in R (version 2.9.4; ref. 26). Transcripts were quantified and normalized by the original authors (27) using the RSEM algorithm (28), and these values were multiplied by 10⁶ to provide transcripts per million (TPM). TPM values were log2(x + 1) transformed prior to analysis. Pairwise correlations of TNC versus all other genes were performed using the Spearman rank test, and P values were adjusted with the Benjamini–Hochberg correction (i.e., false discovery rate, FDR). The relative abundance of tumor-infiltrating immune cells was estimated by computational deconvolution of 22 immune cell types using the CIBERSORT method and LM22 gene signature matrix (29). CIBERSORT was run with 100 permutations and with quantile normalization disabled. Unscaled data are shown. The estimated abundance of “M2 Macrophages” from CIBERSORT analysis, alongside TNC expression, was used to stratify tumors, in each case by selection of the upper and lower 20% of cases. Kaplan–Meier survival analysis using the log-rank test was performed on stratified tumor subsets within the TCGA biolinks package (26). The numbers of patients at risk between the time points of 0, 1,000, 2,000, and 3,000 days, respectively, for each subset were TNC-low, M2-low: 37, 14, 2, 1; TNC-high, M2-low: 39, 23, 9, 1; TNC-low, M2-high: 44, 19, 7, 4; TNC-high, M2-high: 35, 17, 7, 3.

Statistical analyses were performed with Prism software (GraphPad). The Student unpaired t test was used to determine the significance of differences from Gaussian data sets. For non-Gaussian data sets, the Mann–Whitney test was used to determine the significance of differences between unpaired sets of data and the Wilcoxon test for paired sets of data. P values <0.05 were considered statistically significant. Data are representative of at least two individual experiments, expressed as the mean ± SEM (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

Tenascin-C expression correlates with dismal prognosis in people with breast cancer (13) and is linked to earlier tumor onset and larger lung metastases in immunocompetent mouse models of this disease (13, 18). A number of potential cellular sources of this matrix molecule are found in the TME. It can be derived from stromal cells (such as cancer-associated fibroblasts; ref. 30), immune cells (31), and tumor cells themselves (32). To dissect the contribution of tenascin-C from different sources to the immune axis in breast cancer, we first grafted mammary tumor cells expressing high tenascin-C (TNC⁺) or expressing low tenascin-C (TNC^–) into the mammary gland of wild-type mice. This model enables specific examination of the impact of tenascin-C expressed by the tumor cell in the context of a wild-type host response to tumor engraftment. As described previously (18), grafting TNC⁺ tumor cells resulted in accelerated tumor growth compared with grafting TNC^– tumor cells (Fig. 1A and B; Supplementary Fig. S1A and S1B), despite comparable cell proliferation in vitro (Supplementary Fig. S1C). We further showed that tumor size correlated with the concentration of tumor-derived tenascin-C following grafting of TNC⁺ cells (Fig. 1C and D). Tumor size did not correlate with total levels of tenascin-C in the tumor following grafting of either TNC⁺ or TNC^– cells (Supplementary Fig. S1D–S1G). These data implied a pathogenic role for tumor cell–derived tenascin-C that is not compensated for by expression of host-derived tenascin-C in TNC^– tumors. Flow-cytometric analysis of TNC⁺ and TNC^– tumors 3 weeks after engraftment revealed a high percentage of TAM in the immune infiltrate, with fewer macrophages per live cells observed in TNC^– tumors compared with TNC⁺ tumors (Fig. 1E). Although TNC⁺ and TNC^– tumors contained comparable numbers of macrophages per tumor volume and a similar proportion of macrophages within the CD45⁺ compartment (Fig. 1F and G), the proportion of TAM correlated with tumor growth only in mice engrafted with TNC⁺ tumor cells (Fig. 1H and I). This effect was specific to macrophages. The proportion of neutrophils and monocytes was not affected by tumor cell expression of tenascin-C (Fig. 1J and K) nor was tumor growth linked to the proportion of these myeloid cell subsets [neutrophils, TNC⁺: R² = 0.05773 (ns), TNC^–: R² = 0.07005 (ns); monocytes, TNC⁺: R² = 0.001424 (ns), TNC⁻: R² = 0.02704 (ns)]. Phagocyte depletion by intravenous administration of clodronate liposomes following engraftment of TNC⁺ or TNC^– tumor cells significantly reduced the growth of TNC⁺ tumors but had little impact on TNC^– tumors. Administration of clodronate liposomes intratumorally almost completely prevented the growth of TNC⁺ tumors (Fig. 1L). Together, these data indicate that the levels of tumor cell–derived tenascin-C within the TME contribute to TAM pathogenicity.

To determine whether tumor cell–derived tenascin-C fostered changes in macrophage behavior, we assessed qualitative differences in TAM phenotype following the engraftment of TNC⁺ and TNC^– tumor cells into wild-type mice. Macrophages from TNC⁺ tumors comprised a higher proportion of CD206⁺ macrophages than TNC^– tumors, whereas the proportion of IRF5⁺ macrophages was comparable (Fig. 2A–C), suggesting elevated M2-like, immune-suppressive macrophages in a tenascin-C–rich TME. Expression of M1 macrophage–associated genes, including Tnf and Cd86, was lower in macrophages sorted from TNC⁺ tumors compared with macrophages sorted from TNC^– tumors, whereas expression of M2 macrophage–associated genes, including Mrc1 (CD206) and Arg1, were significantly elevated in macrophages isolated from a tenascin-C–rich TME (Fig. 2D). Macrophages isolated from TNC⁺ tumors also exhibited lower CD86, MHCI, and MHCII surface expression than TNC^– tumors (Fig. 2E–G), suggesting an immature phenotype restricted to CD206⁺ (Mrc1) subset (Supplementary Fig. S2A and S2B). Coculture of BMM isolated from naïve, wild-type mice with TNC⁺ or TNC^– tumor cells recapitulated these data. Surface expression of CD206 was significantly higher on BMM in TNC⁺ tumor cell cultures and MHCI and MHCII were significantly reduced compared with BMM cocultured with TNC^– tumor cells or to BMM activated with lipopolysaccharide (LPS; Supplementary Fig. S2C–S2E). Cocultures containing either TNC⁺ or TNC^– tumor cells secreted little detectable IL1β, IL4, IL10, and IL12 but secreted more TGFβ and similar IL8 compared with LPS-activated cells. Cocultures containing TNC⁺ tumor cells secreted higher IL6 and expressed lower iNOS than TNC^– cells (Fig. 2H). Immunofluorescent staining of tumors revealed that F4/80⁺ macrophages colocalized with tenascin-C in niches or tracks extending throughout the TME within both TNC⁺ and TNC^– tumors, although these tracks appeared thinner in TNC^– tumors. Costaining with CD206 confirmed a higher number of CD206⁺ TAM in TNC⁺ tumors, whereas fewer CD206⁺ TAM were detected in TNC^– tumors (Fig. 2I–K; Supplementary Fig. S2F). Together, these data demonstrated that macrophage positioning in tenascin-C–rich niches in the TME occurred irrespective of tenascin-C expression by the tumor and that high tumor-derived tenascin-C was associated with phenotypic changes in TAM, favoring an immature CD206⁺CD86^loMHCI^loMHCII^lo phenotype.

Our results showed that tumor cell–derived tenascin-C promoted an immune-suppressive, protumoral, M2-like macrophage response. However, both stromal and immune cells within the TME can also synthesize tenascin-C. Studies showing that mice lacking tenascin-C fail to effectively recruit and polarize M1-like macrophages in response to experimental myocardial infarction or aortic constriction (33, 34) prompted us to investigate whether host-derived tenascin-C promoted antitumoral macrophage behavior. To address this, we grafted mammary tumor cells expressing high tenascin-C (TNC⁺) into the mammary gland of wild-type mice or into the mammary gland of tenascin-C–null mice. This model enables examination of the impact of tenascin-C expressed by the host without changing the tumor cell–derived tenascin-C. Flow-cytometric analysis revealed fewer macrophages in tumors grafted into tenascin-C–null mice compared with tumors grafted into wild-type mice (Fig. 3A). Macrophages from tumors grafted into knockout mice comprised a higher proportion of CD206⁺ macrophages than wild-type mice (Fig. 3B) and exhibited lower MHCII surface expression (Fig. 3C). Expression of M1 macrophage–associated genes, including Cd86, Irf5, and Nos2, was lower in tumors isolated from tenascin-C knockout mice compared with wild-type mice, whereas expression of the M2 macrophage–associated gene Arg1 was elevated in tumors from knockout mice (Supplementary Table S1). Immunofluorescence staining revealed that F4/80⁺ macrophages colocalized with tenascin-C tracks in tumors grafted into both wild-type and tenascin-C–null mice, whereas costaining with CD206 confirmed a higher number of CD206⁺ TAM in tumors from tenascin-C knockout mice compared with wild-type mice (Fig. 3D and E). Together, these data indicated that host-derived tenascin-C drove an M1-like TAM response comprising a CD206^loMHCII^hi phenotype.

Our data indicated that TAM function differently within TMEs in which tumor- versus host-derived tenascin-C dominated and suggest that high tenascin-C expression by tumor cells subverted host defense responses by switching the TAM phenotype from an antitumoral phenotype toward a protumoral phenotype. To further examine the consequences of TAM reprogramming by tumor-derived tenascin-C, we assessed whether the altered phenotype and cytokine profile of macrophages following grafting of TNC⁺ and TNC^– tumor cells into wild-type mice affected macrophage effector function. We cocultured purified splenic T cells from naïve wild-type FVB mice with TNC⁺ or TNC^– tumor cells in the presence or absence of naïve wild-type BMM. T cells cultured with tumor cells alone exhibited low proliferation, which was comparable between TNC⁺ and TNC^– tumors. However, T cells cultured with TNC⁺ tumor cells and BMM proliferated significantly less than T cells cultured with TNC^– tumor cells and BMM (Fig. 4A and B). Similarly, CD11b⁺ cells sorted from TNC^– tumors were associated with significantly more T-cell proliferation than CD11b⁺ cells isolated from TNC⁺ tumors (Fig. 4C), suggesting that TNC⁺-associated macrophages were less efficient at driving T-cell proliferation. CD11b⁺ cells sorted from TNC⁺ tumors drove significantly more IL17 synthesis in culture with naïve T cells than CD11b⁺ cells isolated from TNC^– tumors. IFNγ remained unchanged (Fig. 4D and E), consistent with a higher proportion of Th17 cells in TNC⁺ tumors compared with TNC^– tumors (Fig. 4F and G). Together, these results support a dual role for macrophages programmed by tumor cell–derived tenascin-C in the generation of Th17-based inflammation, accompanied by a phenotype that fails to support T-cell proliferation.

We next investigated the mechanism by which tumor-derived tenascin-C affects macrophage behavior. Tenascin-C is a large multimodular molecule comprising a series of different domains that each exert specific effects on cell behavior through distinct receptor binding capabilities, including integrins, epidermal growth factor receptor (EGFR), and TLR4 (17). Surface expression of MHCII (Fig. 5A) and IL6 synthesis (Fig. 5B) by wild-type BMM in coculture with TNC⁺ tumor cells was rescued to levels exhibited by wild-type BMM cocultured with TNC^– tumor cells by adding increasing doses of TAK242, a small-molecule inhibitor of TLR4 signaling. Inhibition of RGD-dependent integrin binding using echistatin or inhibition of EGFR kinase activity using tyrphostin had no effect on the BMM phenotype (Supplementary Fig. S3A–S3D). Addition of TAK242 also rescued the proliferation of T cells in coculture with BMM and TNC⁺ tumor cells (Fig. 5C and D). Antibodies that prevent the FBG domain of tenascin-C from binding to and activating TLR4 (20) recovered expression of MHCII by BMM (Fig. 5E), decreased IL6 production (Fig. 5F), and rescued CD8⁺ T-cell proliferation (Fig. 5G and H) in cocultures with TNC⁺ tumor cells. Together, these data indicated that FBG activation of TLR4 was necessary for tumor-derived tenascin-C–mediated macrophage polarization in vitro.

We next determined whether blocking FBG activation of TLR4 affected macrophage phenotype in vivo, and if this affected tumor growth and spread. We titrated 3 doses of anti-FBG, starting treatment 24 hours after mammary gland engraftment of TNC⁺ tumor cells into wild-type mice, based on dosing from preclinical models of arthritis (20). To benchmark the effects of anti-FBG, we used anti–PD-L1 due to the reported efficacy in patients with breast cancer (35), selecting a dosage known to reduce tumor volume in both spontaneous and grafting murine models of breast cancer (21, 36). Antibody treatment was well tolerated, and mice exhibited no side effects (Supplementary Fig. S4A). Treatment with anti-FBG led to a dose-dependent reduction in the growth of primary tumors compared with isotype control–treated mice (Supplementary Fig. S4B) and had comparable efficacy to anti–PD-L1 treatment (Fig. 6A). Treatment with 10 mg/kg of anti-FBG did not affect the size of metastases observed in the lungs but was associated with reduced incidence of lung metastasis, with fewer metastases observed in animals that received anti-FBG compared with isotype control antibody (P = 0.058; Supplementary Fig. S4C). Anti-FBG treatment also significantly increased expression of TLR4 and MHCII on TAM, decreased Th17 cell abundance, and induced a trend toward a higher proportion of tumor-infiltrating CD8⁺ T cells (Fig. 6B–G). Anti-FBG treatment did not influence the localization of F4/80⁺ cells within tenascin-C–rich tracks in the TME, but these cells were less clustered within the TME. Anti-FBG treatment was also associated with fewer CD206⁺ macrophages that were restricted to the periphery of the tumor compared with isotype control–treated tumors (Fig. 6H).

To assess the efficacy of anti-FBG treatment in a more heterogeneous model that better recapitulates spontaneous tumor progression, we used MMTV-NeuNT FVB mice, which develop mammary tumors and lung metastases around 6 months of age. Upon tumor detection, mice were treated with anti-FBG alone, anti–PD-L1 alone, the 2 antibodies in combination, or their respective isotype controls. Treatment with anti-FBG alone led to a reduction of tumor growth comparable to treatment with anti–PD-L1 alone (Fig. 6I), whereas combination treatment led to significantly enhanced, and sustained, inhibition of tumor growth (Fig. 6J). Treatment with anti-FBG alone or together with anti–PD-L1 led to a significant reduction in the percentage of TAM and augmentation of the proportion of CD8⁺ T cells compared with the isotype group and to treatment with anti–PD-L1 alone. MHCII was upregulated in the combination group alone (Supplementary Fig. S5). Finally, treatment with anti-FBG alone or together with anti–PD-L1 reduced the size of lung metastases compared with the isotype group and to treatment with anti–PD-L1 alone (Fig. 6K and L). Together, these results demonstrated that anti-FBG treatment could reduce both primary tumor growth and metastasis and that combination treatment with anti–PD-L1 results in significantly improved outcome compared with either antibody alone, with anti-FBG providing an effect on tumor spread that was not obtained by checkpoint inhibition alone.

To investigate whether the tenascin-C–dependent pathogenic TAM phenotypes we observed in the mouse were conserved in human disease, bulk RNA-seq data of 1,045 human breast tumors from TCGA were analyzed. Pairwise correlations of tenascin-C revealed significant positive associations with macrophage-related genes (CD68 and CSF1R) and genes associated with the polarization of macrophages toward a protumoral M2-like phenotype (CD163, CD204, CD206, and TGM2), but not with other genes, including markers of cell proliferation (Ki67) or M1-like phenotype (IRF5; Fig. 7A). To gain insight into the composition of the tumoral immune compartment, CIBERSORT analysis was performed to computationally deconvolve the estimated abundances of 22 different immune cell types. The predicted immune infiltrate was predominantly composed of M2-like macrophages and a reciprocally low infiltration of proinflammatory M1-like macrophages (Fig. 7B), consistent with published literature (7). Stratification of tumors based upon tenascin-C expression and predicted M2-infiltration indicated that patients with combined tenascin-C–low and M2-low status were completely protected up to 10 years after diagnosis, whereas prognosis was poorer for patients whose tumors contained higher tenascin-C and M2 macrophages (Fig. 7C). Together, these data implied an orthologous role for tenascin-C–mediated switching of TAM phenotype that may contribute to the progression of human disease.

High numbers of TAM correlate with poor prognosis in people with breast cancer (37). However, the factors driving TAM pathogenicity in this disease are not well understood. In this study, we showed that host-derived tenascin-C promoted antitumoral immunity in murine models of breast cancer, via recruitment of proinflammatory macrophages to the TME. In contrast, synthesis of this extracellular matrix protein by tumor cells promoted a phenotypic switch in macrophages that drove the prevalence of tumor-supportive TAM in the TME. We identified that engagement of TLR4 by the FBG domain of tumor cell–derived tenascin-C was required for tenascin-C–mediated macrophage polarization toward this phenotype in vitro and showed that targeting this interaction in vivo was sufficient to inhibit tumor growth and metastasis, conferring sustained therapeutic benefit in combination with immune-checkpoint inhibition. Finally, we observed that people with breast tumors with combined low tenascin-C expression and low M2 macrophage infiltrate exhibited improved prognoses, suggesting that tenascin-C–mediated evasion of immune elimination may also be relevant to human cancer.

As in human disease, high numbers of TAM infiltrate the TME in murine models of breast cancer, where they contribute to primary tumor growth and metastasis (7). Here, we showed that TAM pathogenicity was dependent on expression of tumor cell–derived tenascin-C within the TME and that engraftment of mammary tumor cells expressing high tenascin-C into wild-type mice favored an immature, M2-like (CD206^hiMHCII^lo) TAM phenotype compared with engraftment of cells expressing low tenascin-C. Grafting TNC⁺ or TNC^– tumors cells into wild-type mice enabled us to specifically modulate tumor-derived tenascin-C. As a result, total tenascin-C concentrations in the TME were comparable following grafting of either type of tumor cell. However, the ratio of tumor- and host-derived tenascin-C was altered, with a shift toward more tumor cell–derived tenascin-C proving pathogenic. The proportion of infiltrating myeloid cells, and macrophage positioning within the tumor, were comparable following grafting of either TNC⁺ or TNC^– tumor cells, whereas high tumor cell–derived tenascin-C promoted elevated numbers of CD206^hi macrophages. These results contrast data showing that tenascin-C drives an M1-like macrophage phenotype in models of cardiac pathology, with more CD206^hi macrophages observed in the hearts of tenascin-C–null mice, with higher expression of M2- and lower expression of M1-associated genes compared with wild-type mice (33, 34). This highlights a context specificity of tenascin-C modulation of macrophage behavior that is mirrored in vitro. Tenascin-C can induce aspects of both M1- and M2-like cell behavior, as well as fibrotic signaling, proangiogenic properties, or differentiation into foam cells, depending on the experimental setup (17). Macrophage behavior in tissues is defined by integrating multiple signals derived from the microenvironment (9), and our study highlights the importance of the cellular source of these signals.

Within the TME, tenascin-C can be produced by stromal and immune cells, and by tumor cells. To determine whether macrophages responded differently to tenascin-C from different sources, we used an alternative experimental setup in which we grafted TNC⁺ tumors into wild-type or tenascin-C knockout mice, where tumor cell–derived tenascin-C expression was kept constant. We observed a higher proportion of tumor-infiltrating macrophages in wild-type mice compared with knockout mice, consistent with the association of tenascin-C expression with macrophage infiltration in a range of other human (38–40) and murine pathology (13, 34, 41, 42), and its ability to promote macrophage migration in vitro (43). Also, in our model, a greater proportion of macrophages exhibited a CD206^hiMHCII^lo phenotype in the absence of host-derived tenascin-C, consistent with data that indicate a role for host-derived tenascin-C in polarizing proinflammatory cell behavior (33, 34). Together, these data indicate that tenascin-C made by the host comprises part of innate defense, facilitating macrophage infiltration and organization, and favoring an aggressive M1-like phenotype, but that deposition of tenascin-C by incipient tumor cells acts locally to switch the phenotype of these macrophages to prevent elimination by the immune system.

To further investigate the mechanism by which tumor-derived tenascin-C switched TAM phenotype, we cocultured TNC⁺ or TNC^– mammary tumor cells together with naïve, wild-type BMM. This model revealed that the BMM phenotype induced by tumor cell–derived tenascin-C could be reversed by inhibiting TLR4 signaling and by blocking the ability of the FBG domain of tenascin-C to bind to and activate TLR4. We also showed that treating mice with grafted TNC⁺ tumors, or mice that develop spontaneous mammary tumors, with anti-FBG could effectively reduce tumor size and spread to the lungs. TAM contribute to primary tumor growth and metastasis by stimulating angiogenesis, inducing tumor cell migration, invasion, and intravasation, promoting epithelial-to-mesenchymal transition, and suppressing antitumor immunity (44–46). Although our coculture assays revealed some insight into the functional capabilities of macrophages programmed by TNC⁺ tumor cells, suggesting an immune-suppressive phenotype that failed to support optimal T-cell proliferation and favored Th17 cell polarization, the precise pathogenic nature of these cells in vivo remains to be elucidated. Reduced Th17 cell abundance in the TME of TNC^– tumors and in anti-FBG–treated grafted tumors is consistent with data from murine models of arthritis (20). This may be of interest for further investigation given that IL17 is associated with poor prognosis in breast cancer patients (47), favors PD-L1 expression and lung metastasis, and reduces antitumor immunity in murine models of breast cancer (48). We also found here that anti-FBG therapy reduced primary tumor burden to a degree comparable to anti–PD-L1 treatment, an effect that was sustained following combination treatment, but only treatment with anti-FBG reduced the size of lung metastases. These findings are consistent with data showing that high expression of tenascin-C in breast cancer is associated with marked acceleration of metastases to the lungs (13). However, whether tenascin-C–programmed TAM are directly, or solely, responsible for enhanced tumor spread within mammary tumors that comprise mixed populations of TLR4-expressing cells is not yet known, and further investigation of the impact of anti-FBG treatment on the tumor as a whole is warranted.

The prognostic value of TAM in breast cancer is not straightforward (37), and data indicate that combining phenotypic markers with TAM location may provide better diagnostic sensitivity (49). Further investigation of how matrix molecules, derived from either host or tumor, dictate macrophage distribution throughout the TME may highlight novel mechanisms underpinning these clinical observations. Our data suggest that tumor- and host-derived tenascin-C contribute differently to TAM behavior, necessitating further detailed molecular analysis of tenascin-C from different cellular sources. In tumor tissue where tenascin-C expression is low or absent, we observed a protective effect, with no deaths by 10 years, in people classified as tenascin-C–low, M2-low. Altogether, our data highlight one way in which tumor cells evade immune elimination: by synthesis of the immunomodulatory matrix molecule tenascin-C, which switches TAM phenotype.

K.S. Midwood is a consultant for, reports receiving a commercial research grant from, and has ownership interest (including patents) in Nascient Ltd. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Deligne, R. Carapito, I. Velazquez-Quesada, Z. Sun, G. Orend, K.S. Midwood

Development of methodology: C. Deligne, D. Murdamoothoo, Z. Sun

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Deligne, D. Murdamoothoo, M. Gschwandtner, W. Erne, T. Loustau, A.M. Marzeda, R. Carapito, I. Mazzier, Z. Sun

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Deligne, D. Murdamoothoo, A.N. Gammage, T. Loustau, N. Paul, K.S. Midwood

Writing, review, and/or revision of the manuscript: C. Deligne, D. Murdamoothoo, A.N. Gammage, M. Gschwandtner, W. Erne, T. Loustau, A.M. Marzeda, I. Velazquez-Quesada, I. Mazzier, G. Orend, K.S. Midwood

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Deligne, R. Carapito

This work was supported by grants from Worldwide Cancer Research (14-1070), the Medical Research Council, Nascient Ltd., the Kennedy Trust for Rheumatology Research, the Austrian Science Fund (FWF; J-4102), the Institut National contre le Cancer (INCa, TENPLAMET), Ligue Regional contre le Cancer Grand Est to G. Orend, fellowships to D. Murdamoothoo (Fondation ARC pour la Recherche sure le Cancer), I. Velazquez-Quesada (Conacyt Mexico), Z. Sun (Chinese Scholarship Council), and W. Erne (French Ministry of Research), a Clarendon Fund Scholarship to A.N. Gammage, and a Versus Arthritis Senior Fellowship to K.S. Midwood (20003). The authors would like to express their gratitude to the staff of the Biological Safety Unit at the Kennedy Institute of Rheumatology, in particular Elizete Araujo, Dan Andrew, and Mino Medghalchi, to Richard Corderoy at the OQF, and to the staff of the GENOMAX sequencing facility of INSERM U1109.

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.