Cancer cell CCR2 orchestrates suppression of the adaptive immune response

Tumors escape immune control via multiple mechanisms (Dunn et al., 2002, 2004). These mechanisms include cancer cell–intrinsic changes that alter how the cancer cell is recognized by the immune system and extrinsic changes that suppress immune cell activities. For example, cancer cells can intrinsically decrease the surface expression of MHC class I, making them effectively invisible to cytotoxic T lymphocytes (CTLs; Garrido et al., 2016). Moreover, cancer cells can up-regulate programmed cell death ligand 1 (PD-L1/B7-H1), which binds the PD-1 receptor on activated T cells, resulting in T cell anergy, ultimately protecting cancer cells against T cell–mediated killing (Chen et al., 2016a). Extrinsic mechanisms include the down-regulation of costimulatory molecules (e.g., CD86) on antigen-presenting cells; the secretion of cytokines that directly inhibit CTLs; and the recruitment of regulatory T cells and myeloid-derived suppressor cells (MDSCs; Igney and Krammer, 2002). In contrast, infiltration of CD103⁺ dendritic cells (DCs) in mice, phenotypically similar to CD141⁺ DCs in humans (Hildner et al., 2008), has emerged as a mechanism by which tumors may be kept under immune control. CD103⁺ DCs are highly efficient at acquiring and processing exogenous antigens, which they directly present on MHC class I molecules to CD8⁺ CTLs. Even a modest accumulation of CD103⁺ DCs in tumors has been associated with improved immune-mediated tumor control (Broz et al., 2014; Roberts et al., 2016).

Chemokine receptors mediate the recruitment of immune cells to sites of inflammation and to tumors. C-C chemokine receptor type 2 (CCR2) is expressed by several bone marrow–derived cell types (including inflammatory monocytes, myeloid precursor cells, and immature DCs), as well as B and T lymphocytes (Lim et al., 2016). CCR2-expressing cells are recruited to sites of inflammation primarily by C-C chemokine ligand 2 (CCL2; Deshmane et al., 2009), although other CCL family members also can activate CCR2. Chemotherapy treatment promotes CCR2-dependent infiltration of tumor-promoting myeloid cells to murine mammary tumors (Nakasone et al., 2012), and CCL2/CCR2–mediated recruitment of CCR2⁺ inflammatory monocytes to the lung has been shown to promote breast cancer extravasation and metastasis in mouse models (Qian et al., 2011). Furthermore, elevated levels of CCL2 in tumors and in serum are associated with advanced disease and poor prognosis in breast carcinoma patients (Lebrecht et al., 2004, 2001; Soria and Ben-Baruch, 2008). These findings have sparked interest in targeting the CCR2 pathway to modulate the innate immune response for therapeutic benefit in cancer. However, CCR2 is also expressed by breast cancer cells, and activation of CCR2 by CCL2 can induce cancer cell migration and survival through Smad3-, p42/44MAPK–, and Rho guanosine triphosphatase–mediated signaling (Fang et al., 2012). In vivo, the potential roles of CCR2 signaling in cancer cells have not been well investigated, largely because they were thought to be minor compared with the roles of CCR2 in myeloid cells.

In this study, we report that CCR2 signaling in cancer cells plays a surprisingly major role in regulating the immune response to murine breast tumors. We show that CCR2 in cancer cells supports immune escape by inhibiting CD103⁺ DC infiltration and maturation and by suppressing CTL activity. CCR2 expression on cancer cells represents a previously uncharacterized mechanism for immune suppression. Thus, our data support the notion that the CCL2/CCR2 axis is an important immune modulatory pathway in cancer, used by both immune cells and cancer cells to orchestrate immune suppression.

To investigate the potential effects of CCR2 on breast tumor growth and metastasis, we crossed Ccr2^−/− mice (Boring et al., 1997) with mouse mammary tumor virus–polyoma middle T (MMTV-PyMT) mice, a model of luminal B breast cancer (Guy et al., 1992; Lin et al., 2003). The Ccr2 genotype did not influence normal mammary gland development (Fig. S1 A) or tumor onset (Fig. 1 A); however, loss of even one allele of Ccr2 significantly reduced tumor growth rates (Fig. 1 B). Consistently, survival time was significantly longer for the MMTV-PyMT;Ccr2^−/− and MMTV-PyMT;Ccr2^+/− mice compared with that of the MMTV-PyMT;Ccr2^+/+ mice (Fig. 1 C). In addition, we noted that the tumors of MMTV-PyMT;Ccr2^−/− and MMTV-PyMT;Ccr2^+/− mice were more cystic, with reduced solid areas, than those of MMTV-PyMT;Ccr2^+/+ mice (Fig. S1, B and C).

The importance of CCR2 signaling for monocyte recruitment is well understood, but CCR2 is also expressed by human breast cancer cells (Fang et al., 2012). In accordance with this report, we detected Ccr2 mRNA and CCR2 protein in the cancer cells of the tumors developing in MMTV-PyMT;Ccr2^+/+ mice (Fig. 1, D and E), and using RNA fluorescence in situ hybridization (FISH), we demonstrated that Ccr2 mRNA is expressed by Krt18-positive breast cancer cells (Fig. S1 D). To determine the relative contributions of cancer cell versus host CCR2 to tumor growth, we isolated primary cancer cells (purity >90%; Fig. S1 E) from MMTV-PyMT;Ccr2^+/+ and MMTV-PyMT;Ccr2^−/− mice and transplanted them orthotopically to mammary glands of either Ccr2^+/+ or Ccr2^−/− syngeneic host mice (Fig. 1 F). Loss of Ccr2 in the host did not alter tumor growth, consistent with our previous report (Nakasone et al., 2012). However, loss of Ccr2 in cancer cells significantly reduced tumor growth rates, leading to approximately twofold longer survival (Fig. 1, G and H). Ccr2 mRNA levels of Ccr2^+/− cancer cells were lower than those of Ccr2^+/+ cancer cells (Fig. S1 F), and tumors derived from Ccr2^+/− cancer cells exhibited the same slow growth as tumors derived from Ccr2^−/− cancer cells (Fig. S1 G). This result suggests that a threshold level of CCR2 expression in MMTV-PyMT cancer cells regulates tumor growth.

CCR2⁺ inflammatory monocytes can promote breast cancer metastasis (Müller et al., 2001; Qian et al., 2011). The overall metastatic burden in the lung—the primary site of metastasis in the MMTV-PyMT model—was, however, very variable in MMTV-PyMT;Ccr2^+/+ mice and was not significantly different from that of MMTV-PyMT;Ccr2^−/− mice (Fig. S1 H). There were also no significant differences in the number of metastatic foci (representing seeding density) related to the Ccr2 genotype (Fig. S1 I). However, we found that the metastatic foci were larger in the lungs of MMTV-PyMT;Ccr2^+/+ mice than in the lungs of MMTV-PyMT;Ccr2^−/− mice (Fig. S1 J), suggesting that CCR2 promotes the growth of secondary lesions.

The tumors from transplanted Ccr2^+/+ or Ccr2^−/− cancer cells were markedly different at the histological level (Fig. 2 A), similar to the autochthonous tumors in the MMTV-PyMT;Ccr2^−/− animals (Fig. S1, B and C). The tumors originating from Ccr2^+/+ cancer cells consisted entirely of microlobules composed of sheets of large, neoplastic epithelial cells. The microlobules were surrounded by small amounts of fibrovascular stroma containing dense collagen, which is characteristic of poorly differentiated MMTV-PyMT tumors. The neoplastic cells had round-to-oval nuclei, prominent nucleoli, scant cytoplasm, and indistinct cell borders, typical of undifferentiated cells. In contrast, the tumors originating from Ccr2^−/− cancer cells were more differentiated. In many areas, the cancer cells contained lipid vacuoles of various sizes. These tumors also had large cystic areas lined by a single or double layer of small, polarized epithelial cells, and the lumina were filled with proteinaceous secretions (Fig. 2 A). Conversely, the tumors originating from Ccr2^−/− cancer cells had a reduced percentage of solid areas compared with those originating from Ccr2^+/+ cancer cells (Fig. 2 B).

To understand why tumors from Ccr2^−/− cancer cells grew more slowly than tumors from Ccr2^+/+ cancer cells, we next examined the rates of proliferation and apoptosis in the tumors collected during the phase when Ccr2^−/− tumors were growth restricted, 5–6 wk after transplantation. There was no difference in proliferation as determined by nuclear Ki67 staining (Fig. 2 C), and primary Ccr2^+/+ and Ccr2^−/− cancer cells also grew similarly in vitro (Fig. 2 D). However, we found a higher percentage of cancer cells undergoing apoptosis in tumors from Ccr2^−/− cancer cells than from Ccr2^+/+ cancer cells (Fig. 2 E). Furthermore, Ccr2^−/− cancer cells were more sensitive to serum-free conditions (Fig. 2 F), consistent with a previous study showing that CCR2 signaling can protect cancer cells from apoptosis induced by serum starvation (Fang et al., 2012).

To determine which pathways in the Ccr2^−/− cancer cells led to reduced tumor growth and increased cell death, we isolated cancer cells from Ccr2^+/+ and Ccr2^−/− transplanted tumors formed in wild-type hosts for equal amounts of time and performed transcriptome profiling by RNA sequencing (RNA-seq). Overall, 520 genes were differentially expressed (adjusted P value <0.05), ∼40% of which were up-regulated in Ccr2^−/− cancer cells compared with Ccr2^+/+ cancer cells. Among the differentially expressed genes, we detected up-regulation of MHC class I (H2-K1) and Tap1, two genes required for antigen presentation, and IFI27, an IFN response gene, in the Ccr2^−/− cancer cells (Fig. 3 A). Vimentin, which is found in poorly differentiated epithelial cells and in mesenchymal cells, was down-regulated in Ccr2^−/− cancer cells. This is consistent with a previous study describing a mesenchymal and invasive phenotype in CCR2-expressing cancer cells (Hu et al., 2019). Additionally, gene set enrichment analysis showed robust expression enrichment of the IFN-γ response hallmark gene set, which included genes involved in antigen processing and presentation, in Ccr2^−/− cancer cells compared with Ccr2^+/+ cancer cells (Fig. 3 B). We also observed a pronounced luminal gene expression signature in the Ccr2^−/− cancer cells (Fig. 3 B), in agreement with the noted histological differences (Fig. 2, A and B). Consistently, gene ontology (GO) term analysis revealed altered expression of pathways involved in (1) regulation of the adaptive immune system and in (2) response to immune-mediated cell killing (Fig. 3 C).

The increased expression of genes associated with IFN response and antigen presentation, together with the prolonged growth delay and increased apoptosis in tumors from Ccr2^−/− cancer cells, is consistent with an ongoing adaptive immune response in the tumors. We therefore determined whether there were any differences in the infiltrating immune cell populations by flow cytometry (for gating strategy, see Fig. S2). There were significantly more CD8⁺ CTLs and fewer CD4⁺ helper T cells in tumors from Ccr2^−/− cancer cells than in tumors from Ccr2^+/+ cancer cells (Fig. 4 A), with no overall change in CD3⁺ T cell infiltration (Fig. S3 A). Increased infiltration of CD8⁺ cells into tumors derived from Ccr2^−/− cancer cells was also evident by immunofluorescence (Fig. 4, B and C). PD-1 is an inhibitory checkpoint surface molecule expressed by activated CD8⁺ T cells (Chen et al., 2016a), and its expression was increased on CD8⁺ T cells from tumors from Ccr2^−/− cancer cells (Fig. 4 D). CTLs kill, in part, through the release of granzyme B, a process that requires lysosomal-associated membrane protein 1 (LAMP-1; also known as Cluster of Differentiation 107a [CD107a]). In tumors from Ccr2^−/− cancer cells, LAMP-1 was increased on the cell surface of CD8⁺ T cells (Fig. 4 E), while granzyme B levels were reduced in CD8⁺ T cells (Fig. 4 F), indicating increased levels of prior CD8⁺ degranulation and therefore cytotoxic activity. Regulatory T lymphocytes (CD4⁺FoxP3⁺ regulatory T cells), which are typically immunosuppressive, were more abundant in tumors from Ccr2^−/− cancer cells (Fig. S3 B).

Tumors from Ccr2^−/− cancer cells had multiple changes in the adaptive immune cell infiltrate, all consistent with an active immune response. Therefore, we tested whether an adaptive immune response contributed to the differences in growth rates between the Ccr2^+/+ and Ccr2^−/− tumors by transplanting cancer cells in parallel into T cell–deficient athymic (nude) or fully immunocompetent mice (Fig. 5 A). In the immunocompetent hosts, tumors from Ccr2^−/− cancer cells grew significantly more slowly than tumors from Ccr2^+/+ cancer cells (Fig. 5 B), as previously observed (Fig. 1 G). In contrast, in the T cell–deficient hosts, the Ccr2^−/− tumors grew at a similar rate as the Ccr2^+/+ tumors (Fig. 5 C). Consistently, tumors derived from Ccr2^−/− cancer cells grew as fast as tumors derived from Ccr2^+/+ cancer cells when CD8⁺ T cells were depleted from wild-type mice (Fig. S3 C). Furthermore, in the T cell–deficient, athymic hosts, tumors from Ccr2^−/− cancer cells no longer had more apoptotic cells than tumors from Ccr2^+/+ cancer cells (Fig. 5 D). Interestingly, the differentiated histological phenotype of the tumors from the Ccr2^−/− cancer cells was maintained in the athymic mice (Fig. 5, E and F). Thus, the histological differences were not the result of an altered adaptive immune response. Together, these data demonstrate that CCR2 expressed by cancer cells enables the tumors to escape the adaptive immune response.

We next speculated that the reduced growth of tumors from Ccr2^−/− cancer cells could be a result of increased sensitivity to CTL-induced cell death, given that the Ccr2^−/− cancer cells showed increased sensitivity to serum-free conditions and that the tumors grew normally when CD8⁺ T cells were depleted. To test this possibility, we used the MMTV-PyMT–mCherry-OVA (chOVA) model (Engelhardt et al., 2012), which is driven by the PyMT oncogene and coexpresses OVA as a model tumor antigen and mCherry for tracking. We generated MMTV-PyMT-chOVA;Ccr2^+/+ and MMTV-PyMT-chOVA;Ccr2^−/− mice and found that similar to the MMTV-PyMT;Ccr2^−/− tumors, the MMTV-PyMT-chOVA;Ccr2^−/− tumors grew more slowly than MMTV-PyMT-chOVA;Ccr2^+/+ tumors (Fig. 6 A). When primary cancer cells were isolated and challenged with activated CTLs isolated from OT-1 mice (where T cells are engineered to recognize an OVA peptide presented by MHC class I), we found that MMTV-PyMT-chOVA;Ccr2^−/− cancer cells were approximately twice as sensitive to CTLs as MMTV-PyMT-chOVA;Ccr2^+/+ cancer cells (Fig. 6 B). Upon treatment with a pharmacological inhibitor against CCR2 (RS 102895 hydrochloride; Mitchell et al., 2013), the MMTV-PyMT-chOVA;Ccr2^+/+ cancer cells became as sensitive to the OVA-specific CTLs as the MMTV-PyMT-chOVA;Ccr2^−/− cancer cells (Fig. 6 C). The CCR2 inhibitor also increased specific cytolysis of the OVA-expressing E0771 murine breast cancer cell line to the OVA-specific CTLs (Fig. S3 D). In contrast, the CCR2 inhibitor had no effect on the specific cytotoxicity of the CTLs to the MMTV-PyMT-chOVA;Ccr2^−/− cancer cells (Fig. 6 C).

To test if Ccr2 expression in cancer cells also regulated sensitivity to CTLs in vivo, we measured the clonal expansion of T cells, an indicator of active proliferation of antigen-specific T cells, by sequencing the complementarity determining region 3 (CDR3) region of the T cell receptor (TCR) β chain. We found that clonality, a value corresponding to the extent of clonal expansion of T cells (Fig. 6 D), was significantly higher in tumors from Ccr2^−/− cancer cells than in tumors from Ccr2^+/+ cancer cells (Fig. 6 E), while there was no difference in clonality in the spleens of these groups of mice (Fig. 6 F). These results suggested that tumor-recognizing CTLs were actively expanding due to antigen recognition in tumors from Ccr2^−/− cancer cells, but that this expansion did not alter the systemic T cell response. Both Ccr2^+/+ and Ccr2^−/− cancer cells express the PyMT tumor antigen, and Ccr2^+/+ cancer cells formed tumors at the same rate regardless of whether Ccr2^+/+ or Ccr2^−/− cancer cells were transplanted to the contralateral mammary gland (Fig. S3 E). Together, these results suggest that the expression of CCR2 by cancer cells enables the tumors to establish a locally immune-suppressive microenvironment.

CCR2 signaling can activate STAT transcription factors in cancer cells (Chen et al., 2015; Izumi et al., 2013), and altered STAT1 or STAT3 signaling in cancer cells can in turn lead to a less immune-suppressive microenvironment (Ahn et al., 2017; Jones et al., 2016). Therefore, we tested whether STAT activation was different between tumors from Ccr2^+/+ and Ccr2^−/− cancer cells. We found no consistent differences between the activation of STAT1, STAT3, or p65/RELA when whole tumor lysate was assayed by Western blot (Fig. 7 A). By immunohistochemistry, we found that the tumors contained patches of cancer cells with high expression of phospho-STAT1, with more of these patches in the tumors from Ccr2^−/− than from Ccr2^+/+ cancer cells (Fig. 7, B and C). No difference was observed for phospho-STAT3. This suggested a localized, enhanced anti-tumor immune response regulated via STAT1 in tumors from Ccr2^−/− cancer cells compared with those from Ccr2^+/+ cells, consistent with other reports (Avalle et al., 2012).

We next sought to understand how CCR2 signaling in cancer cells resulted in more efficient immune suppression. RNA-seq analysis of the cancer cells showed increased expression of genes involved in antigen presentation in the Ccr2^−/− tumors (Fig. 3 A). Consistently, we found high levels of MHC class I expressed by cancer cells in MMTV-PyMT-chOVA;Ccr2^−/− mice compared with the variable, low to intermediate levels expressed by cancer cells in MMTV-PyMT-chOVA;Ccr2^+/+ mice (Fig. 8 A). Similarly, when cancer cells were transplanted to immunocompetent host mice, expression of MHC class I was higher by Ccr2^−/− cancer cells than by Ccr2^+/+ cancer cells, as shown by immunofluorescence (Fig. 8 B). However, in T cell–deficient, athymic host mice, MHC class I levels were similar on Ccr2^+/+ and Ccr2^−/− cancer cells (Fig. 8 C). These data suggest that MHC class I expression is not directly regulated by CCR2 signaling in the cancer cells, but rather is regulated by the different immune responses between these tumors.

PD-L1 is a potent suppressor of T cell activation and proliferation and is expressed on many types of cancer cells (Freeman et al., 2000). Expression of PD-L1 was significantly higher on Ccr2^+/+ cancer cells than on Ccr2^−/− cancer cells (Fig. 8 D). Since both MHC class I and PD-L1 are IFN-γ response genes, we compared Ifn-g mRNA levels during the early tumor phase (3 wk after transplantation) and found significantly higher levels of Ifn-g, as well as Cxcl9, in whole tumor lysate from tumors derived from Ccr2^−/− compared with Ccr2^+/+ cancer cells by real-time quantitative PCR (RT-qPCR; Fig. S4 A). RNA FISH revealed that the majority of the Ifn-g mRNA was expressed by infiltrating CD8⁺ T cells and that Ifn-g levels were more variable during the growth-restricted phase of tumors than during the early tumor establishment phase (Fig. S4, B–D). In vitro, recombinant IFN-γ treatment resulted in similar up-regulation of Pd-l1 mRNA levels in Ccr2^+/+ and Ccr2^−/− cancer cells (Fig. 8 E). Thus, our data suggest that Ccr2^−/− cancer cells would be able to respond to increased IFN-γ levels in the tumor microenvironment by up-regulating PD-L1. Since we see reduced PD-L1 expression on the Ccr2^−/− cancer cells in vivo in the context of elevated Ifn-g, other factors must repress PD-L1 expression in vivo. Lastly, we tested whether PD-L1 blockade would alleviate immune suppression of the MMTV-PyMT tumors in vivo. However, anti–PD-L1 therapy had no significant effect on the growth of either the Ccr2^+/+ tumors or the low PD-L1–expressing Ccr2^−/− tumors (Fig. 8 F). Thus, although reduced PD-L1 likely contributed to an increased sensitivity to CTLs in tumors from Ccr2^−/− cancer cells, targeting PD-L1 expression was insufficient to reduce growth of the Ccr2^+/+ tumors.

We next examined the infiltration levels of immune populations that might be involved in immune suppression in the Ccr2^+/+ tumors. There was no difference in CD11b⁺MHC class II⁺F4/80⁺ macrophages between tumors from Ccr2^+/+ and Ccr2^−/− cancer cells (Fig. S4 E). There were, however, significantly fewer CD11b⁺MHC class II⁻ myeloid cells in tumors from Ccr2^−/− cancer cells (Fig. S4 F). This population includes subsets of MDSCs, immature myeloid cells with immune-suppressive activity. Specifically, we found fewer granulocytes and granulocytic MDSCs (CD11b⁺CD11c⁻Ly6G⁺Ly6C^low cells) and more inflammatory monocytes and myeloid MDSCs (CD11b⁺CD11c⁻Ly6G⁻Ly6C^high cells) in tumors from Ccr2^−/− cancer cells than in those from Ccr2^+/+ cancer cells (Fig. S4, G and H).

Since DCs are critical for T cell activation, we performed immunofluorescence for CD11c⁺, a marker of DCs, and observed more infiltrating CD11c⁺ cells in the tumors derived from Ccr2^−/− cancer cells than in those derived from Ccr2^+/+ cancer cells (Fig. 9 A). Furthermore, we found more than double the level of CD103⁺ DCs—a DC subtype that is very proficient in cross-presenting antigens to CD8⁺ T cells—in tumors from Ccr2^−/− cancer cells than in those from Ccr2^+/+ cancer cells (Fig. 9 B). The number of CD103⁺ DCs was also increased in the draining lymph nodes (dLNs) for these tumors (Fig. 9 C). A concern when using a transplant system to evaluate immune response is the potential for an immune reaction toward antigens on the transplanted cells. However, there were also more CD103⁺ DCs in the spontaneously developing tumors in MMTV-PyMT;Ccr2^−/− mice than in MMTV-PyMT;Ccr2^+/+ mice (Fig. S4 I). Using RT-qPCR from whole tumor lysate and a panel of markers previously shown to be enriched in CD103⁺ DCs (Broz et al., 2014), we detected more mRNA for Cd11c and from activation markers, including Cd80, Cd86, and Il12b (Fig. 9 D). Consistently, DCs in tumors from Ccr2^−/− cancer cells displayed higher levels of CD86, a surface marker up-regulated on activated DCs as determined by flow cytometry (Fig. 9 E). The difference in CD103⁺ DC numbers was already apparent in the dLNs 2 d after injection of irradiated cancer cells into the mammary fat pad (Fig. 9 F), suggesting a direct signal from the cancer cells. However, infiltration of CD103⁺ DCs was not increased in tumors from Ccr2^−/− cancer cells in T cell–deficient, athymic mice (Fig. 9 G). This finding suggests that a feedback loop between T cells and DCs is critical for CD103⁺ DC recruitment to the tumors.

We next asked whether a factor secreted by Ccr2^−/− cancer cells was responsible for DC maturation toward the CD103⁺ DC subtype. Indeed, when bone marrow–derived cells (BMDCs) were cultured with conditioned medium from Ccr2^−/− or Ccr2^+/+ cancer cells, the total number of DCs was not different, but a significantly higher percentage of CD11c⁺ cells was positive for CD103⁺ using conditioned medium from the Ccr2^−/− cells (Fig. 10 A). Furthermore, the percentage of CD103⁺CD11c⁺ cells also increased when using conditioned medium from Ccr2^+/+ cancer cells that had been treated with the CCR2 inhibitor (Fig. 10 B), whereas adding the CCR2 inhibitor directly to the bone marrow–derived cultures did not by itself increase the percentage of CD103⁺ cells. Both Flt3 ligand and GM-CSF can stimulate DC maturation toward the CD103⁺ subtype (Broz et al., 2014). Consistently, we found more mRNA for both Flt3l and Csf2 (coding for GM-CSF) in the tumors derived from Ccr2^−/− cancer cells than from Ccr2^+/+ cancer cells by RT-qPCR (Fig. 10 C). Protein arrays showed that the levels of two CCR2 ligands, CCL2 and CCL12, and most other cytokines were similar between MMTV-PyMT;Ccr2^+/+ and MMTV-PyMT;Ccr2^−/− tumors (Fig. S5 A), while IL-16 and granulocyte colony-stimulating factor (G-CSF) were significantly lower in the MMTV-PyMT;Ccr2^−/− tumors at the protein level. When secretions from overnight cultures of the cancer cells under serum-free conditions were analyzed, no significant differences were detected (Fig. S5 B). Thus, from these different assays measuring protein and mRNA, Flt3L, G-CSF, and GM-CSF emerged as candidate factors for future functional studies to determine how CD103⁺ DC maturation is altered downstream of CCR2 signaling in cancer cells.

Lastly, to test the importance of the CD103⁺ DCs in regulating an immune-controlling microenvironment, we transplanted Ccr2^−/− and Ccr2^+/+ cancer cells into Batf3^−/− mice, which lack the basic leucine zipper transcription factor ATF-like 3 (BATF3) critical for CD103⁺ DC maturation. In these mice, the growth of Ccr2^−/− cancer cell–derived tumors was restored to that of the Ccr2^+/+ cancer cell–derived tumors (Fig. 10 D), and as expected, the percentage of infiltrating CD103⁺ DCs was significantly reduced (Fig. 10 E). Furthermore, CD8⁺ T cell infiltration, as well as expression of CD107a and PD-1 by the T cells, was reduced in tumors derived from Ccr2^−/− cancer cells transplanted into Batf3^−/− mice (Fig. 10, F–H). These findings support the notion that CD103⁺ DC and CD8⁺ T cells act together to regulate tumor immunity.

The tumor-promoting effects of CCR2 are well documented. However, they have mostly been attributed to CCR2-expressing immune and other host cells (Lim et al., 2016; Nakasone et al., 2012; Qian et al., 2009, 2011; Wolf et al., 2012). Yet, CCR2 is also expressed by breast cancer cells and possibly up-regulated compared with normal breast epithelial tissue in a subset of breast cancers, and such expression has been correlated with shortened survival (Labovsky et al., 2017). The expression of CCR2 by cancer cells has been proposed to promote tumor growth through increased motility and invasion, cell proliferation, and cell survival (Fang et al., 2012; Hu et al., 2019; Lu et al., 2006). Here, we showed that in the MMTV-PyMT mouse model, CCR2 signaling in cancer cells helps breast tumors establish a locally immune-suppressive microenvironment. Compared with tumors from Ccr2^+/+ cancer cells, the tumors from Ccr2^−/− cancer cells have higher infiltration of CD103⁺ DCs and activated CD8⁺ T cells, as well as lower infiltration of granulocytic MDSCs. Furthermore, the Ccr2^−/− cancer cells have higher surface levels of MHC class I and lower levels of PD-L1. It is likely the combination of all of these changes that results in more effective immune surveillance of the Ccr2^−/− cancer cells and reduced growth of tumors from these cells. To our knowledge, the ability of cancer cell CCR2 signaling to orchestrate tumor immune suppression has not previously been recognized.

Among the most notable changes in the tumor immune microenvironment caused by targeting cancer cell Ccr2 were the increased levels of both CD103⁺ DCs and CD8⁺ T cells in tumors, and the infiltration of these two cell populations was apparently connected. In athymic mice lacking mature CD8⁺ cells, the CD103⁺ DC infiltration was no longer elevated in tumors derived from Ccr2^−/− cancer cells compared with those from Ccr2^+/+ cancer cells. Conversely, in Batf3^−/− mice, which lack the transcription factor BATF3 required for CD103⁺ DC development, CD8⁺ T cell infiltration was equally low for tumors from Ccr2^+/+ and Ccr2^−/− cancer cells. These observations are consistent with several other recent reports intimately linking the function of CD103⁺ DCs and CD8⁺ T cells in tumors. First, CD103⁺ DCs have been shown to secrete chemokines mediating T cell infiltration, including CXCL9 and -10 (de Mingo Pulido et al., 2018; Spranger et al., 2017). Second, secretion of IL-12 from CD103⁺ DCs has been shown to stimulate CD8⁺ T cell function (Ruffell et al., 2014). Our data support these previous studies, showing not only a link between CD103⁺ DC and CD8⁺ T cell infiltration but also increased Cxcl9 and Il-12b mRNA in tumors derived from Ccr2^−/− cancer cells. Altogether, these data support previous work showing that the regulation of CD103⁺ DC infiltration is a critical step in tumor immune surveillance (Broz et al., 2014; Meyer et al., 2018; Roberts et al., 2016).

Our data suggest that CCR2 signaling regulates the secretion of a factor from the cancer cells that affects the polarization of DCs to the CD103⁺ subtype. Several factors have been proposed to induce the polarization of CD103⁺ DCs, including Flt3L and GM-CSF (Mayer et al., 2014), and consistently, mRNA levels for both of these factors were higher in tumors from Ccr2^−/− cancer cells than in those from Ccr2^+/+ cancer cells. However, it is unclear if Flt3L and GM-CSF could be solely responsible for the difference in CD103⁺ DC infiltration between tumors from Ccr2^+/+ or from Ccr2^−/− cancer cells. A recent study reported that G-CSF secreted from MMTV-PyMT cancer cells can inhibit the differentiation of CD103⁺ DCs by down-regulating IFN regulatory factor-8 in DC progenitors (Meyer et al., 2018). Interestingly, G-CSF was reduced in MMTV-PyMT;Ccr2^−/− tumors compared with MMTV-PyMT;Ccr2^+/+ tumors, although secretion of G-CSF from isolated cancer cells was not reduced, suggesting that altered secretion is not a direct effect of CCR2 signaling in cancer cells.

Ccr2^−/− cancer cells had distinct expression of several genes, including many that were consistent with an IFN response. Among these, increased surface expression of MHC class I might explain why Ccr2^−/− cancer cells are more sensitive to T cell–mediated killing than Ccr2^+/+ cancer cells. However, since Ccr2^+/+ cancer cells did not down-regulate MHC class I in T cell–deficient, athymic mice, we speculate that MHC class I expression is not directly regulated by CCR2 signaling in the cancer cells. Rather, the altered expression of IFN response genes, including MHC class I, may be an indirect result of an immune response in the tumors. Altogether, we propose that the cancer cells act on CD103⁺ DCs, leading to a feedback loop between CD103⁺ DCs and CD8⁺ T cells—two immune cell types with highly interconnected functions.

Our study has several translational implications. First, our findings provide a possible explanation as to why patients with high CCR2 expression in breast cancer cells have a shortened survival time compared with patients with low/negative CCR2 expression (Labovsky et al., 2017). Second, results from an open-label nonrandomized phase I trial using a CCR2 inhibitor against locally invasive pancreatic cancer suggest that the inhibitor enhanced chemotherapy response, reduced macrophage infiltration, and induced an anti-tumor adaptive immune response (Nywening et al., 2016). Our data suggest that the potential clinical benefits of CCR2 inhibitors might be caused by the combined inhibition of CCR2 signaling in both stromal and cancer cells. Interestingly, the loss of even one Ccr2 allele was sufficient to reduce tumor growth in our study, suggesting that partial inhibition of CCR2 could result in clinical benefit. It is unclear why Ccr2^+/− cancer cells phenocopy the Ccr2^−/− cancer cells, but this phenomenon could be due to reduced dimerization of CCR2 in Ccr2^+/− cancer cells, consistent with a prior report (Rodríguez-Frade et al., 1999), or because surpassing a threshold level of CCR2 signaling is required to induce immune suppression. Finally, recombinant G-CSF and GM-CSF are routinely given with certain types of chemotherapy to support the bone marrow and reduce the risk of lethal febrile neutropenia (Mehta et al., 2015). Our data support prior studies suggesting that these cytokines play complex roles in regulating the immune response against tumors: they appear to have opposing effects on CD103⁺ DC differentiation, with GM-CSF stimulating CD103⁺ DC differentiation (Greter et al., 2012; Mayer et al., 2014) and G-CSF inhibiting it (Meyer et al., 2018). Although GM-CSF secretion in the tumor microenvironment potentially enhances immune responses against tumors by stimulating CD103⁺ DC differentiation, this cytokine can also mobilize immune-suppressive MDSCs, leading to suppression of an anti-tumor immune response (Bayne et al., 2012; Pylayeva-Gupta et al., 2012). Therefore, exploring other approaches than administration of recombinant GM-CSF and G-CSF to specifically boost CD103⁺ DC maturation could be an important future translational direction, and determining the exact mechanism that causes inhibition of cancer cell CCR2 signaling to promote CD103⁺ maturation may provide insights into how this process occurs.

We found that tumor onset was similar between MMTV-PyMT;Ccr2^+/+ and MMTV-PyMT;Ccr2^−/− mice, while tumor growth was slower in MMTV-PyMT;Ccr2^−/− mice. These data are very similar to a prior report studying tumor growth in MTMV-Neu mice, where mice null for Ccl2, the ligand for CCR2, also had similar tumor onset yet reduced tumor growth rates, as did mice treated with a CCR2 inhibitor (Chen et al., 2016b). However, unexpectedly, MMTV-Neu mice lacking Ccr2 had accelerated, not reduced, tumor growth in this study, and the authors speculated that this could be due to abnormal differentiation of the monocytes in these mice or unintentional effects of the genetic targeting of Ccr2. There are several possible explanations for the difference between our result using MMTV-PyMT;Ccr2^−/− mice and that of the MMTV-Neu;Ccr2^−/− mice in the previous publication. First, tumors of MMTV-PyMT mice have more stromal components than tumors of MMTV-Neu mice. Second, the genetic background of the mice used in our experiments was C57BL/6, while it was BALB/c in the case of the MMTV-Neu model. Finally, our tumor transplantation experiments strongly suggest that the tumor growth phenotype in our model is caused by CCR2 expression by cancer cells, but it is not clear whether cancer cells in the MMTV-Neu model express CCR2 or whether the phenotype is due to effects from stromal cells.

Despite the data suggesting that CCR2 inhibition has beneficial effects in cancer treatment, some studies have raised serious concerns about targeting the CCR2 ligand. Cessation of anti-CCL2 therapy resulted in increased pulmonary metastasis and decreased survival in a murine model of breast cancer due to rapid mobilization of CCR2-expressing monocytes into the circulation immediately upon cessation of therapy (Bonapace et al., 2014). Therefore, CCR2 inhibition likely will have to be used in combination with other therapies to maximize the eradication of existing cancer cells and to avoid complications of discontinuing the inhibition. In summary, the divergent results across mouse models, genetic backgrounds, and variable experimental approaches to targeting CCL2 or CCR2 suggest that the effects of targeting this axis can depend on tumor subtype, genetic background, and cell types.

Our findings support the notion that the CCL2/CCR2 axis is an immune modulatory pathway in cancer, but with the added layer of complexity that cancer cells can hijack this chemokine receptor pathway to orchestrate immune suppression. In this study, we focused on characterizing the cellular mechanisms responsible for the novel link between cancer cell CCR2 signaling and tumor immune suppression. Preclinical experiments suggest that CCR2 targeting strategies can have detrimental effects, including the promotion of metastasis (Bonapace et al., 2014; Li et al., 2013). Although we observed no effect of anti–PD-L1 antibodies in combination with CCR2 deletion in cancer cells, anti–PD-L1 antibodies in combination with CCR2 inhibition (which also targets CCR2 on myeloid cells) reduced MDSC infiltration and increased survival in a mouse model of glioma (Flores-Toro et al., 2020). Therefore, future work that identifies the precise mechanisms by which CCR2 signaling in cancer cells induce immune suppression in breast cancer may set the stage for developing novel immune modulatory therapies. There are ongoing clinical trials using pharmacological inhibitors of CCR2 in cancer patients, and our data suggest that the potential effects of inhibiting CCR2 signaling in cancer cells should be taken into account when interpreting the results from these trials.

Ccr2^−/−, Batf3^−/−, and OT-1 mice (all on a C57BL/6 background) were obtained from Jackson Laboratory, and athymic (Nu/Nu) mice were obtained from Charles River Laboratory. MMTV-PyMT mice (on C57BL/6 [referred to as BL/6] background) were bred at Cold Spring Harbor Laboratory (CSHL), and MMTV-PyMT-chOVA mice (Engelhardt et al., 2012) were a kind gift from Dr. Mathew Krummel (University of California, San Francisco, San Francisco, CA). MMTV-PyMT and PyMT-chOVA mice were intercrossed with Ccr2^−/− mice. MMTV-PyMT;Ccr2^+/+ and MMTV-PyMT;Ccr2^−/− mice, and similarly MMTV-PyMT-chOVA;Ccr2^+/+ and MMTV-PyMT-chOVA;Ccr2^−/− mice, were from the same mouse colony (mice were littermates, or their parents were littermates). All the BL/6 host mice were purchased from Jackson Laboratory. All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at CSHL and were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Tumor onset was determined by weekly palpation. Once a tumor was detected, its size was measured weekly by caliper, and tumor volume was calculated as (length × width²)/2. For primary tumor growth, mice were either sacrificed at endpoint, when tumors reached 20 mm or ulcerated (whichever came first), or at indicated time points (early phase: 3 wk after transplantation; growth-restricted phase: 5–6 wk after transplantation). For the lung metastasis assay, mice were sacrificed at either of two IACUC-approved endpoints (when tumors reached 25 mm or ulcerated). To determine metastatic burden, we adapted a previously published method (Nielsen et al., 2001). Briefly, lungs were placed in 4% paraformaldehyde (PFA) in PBS in a vacuum desiccator for 1 h. The lungs remained in 4% PFA at 4°C for 48 h, followed by incubation in 20% sucrose in PBS at 4°C for 48 h. Lungs were embedded in Tissue-Tek optimal cutting temperature compound embedding solution (Sakura). The blocks were then placed in a custom-made cutting chamber with razor blade inserts every 2 mm, and the block was cut into 2-mm sections. These sectioned lung pieces were then reembedded in fresh optimal cutting temperature compound to allow for cross-sectional cuts from the entire lung tissue. These cross sections of the lungs were stained with H&E and scanned by an Aperio ScanScope CS System (Leica Biosystems). The metastatic burden was calculated as the percentage of metastasis/lung area and the number of foci/lung area, determined using Aperio eSlide Capture Devices software (Leica Biosystems). Histology was evaluated by pathologist J.E. Wilkinson and scored as percentage of solid areas of tumors.

Aged-matched virgin females (C57BL/6, Ccr2^−/−, Batf3^−/−, or Nu/Nu) 6–12 wk of age were used as hosts for transplantation. Cancer cells were isolated from two or three tumors, each one 8–10 mm in diameter, from MMTV-PyMT;Ccr2^+/+ or MMTV-PyMT;Ccr2^−/− mice. Tumors were mechanically dissociated and digested for 1 h with 1× collagenase/hyaluronidase (10X Solution; Stem Cell Technology) containing DNase I (2 U/ml; Roche) in Roswell Park Memorial Institute (RPMI)–1640 medium supplemented with 5% FBS (VWR Life Science Seradigm). Single cells and debris were removed from the resulting carcinoma organoid preparation by pulse centrifugation in HBSS supplemented with 5% FBS. Purified carcinoma organoids were dissociated into single-cell suspension in 0.05% trypsin with 0.1% EDTA supplemented with 2 U/ml of DNase I for 2–3 min. Single tumor cells were passed through a 100-µm cell strainer (BD Biosciences) and either plated in RPMI-1640 supplemented with 10% FBS for in vitro experiments or washed with PBS and immediately injected into the inguinal mammary glands of host mice (2.5 × 10⁵ in 20 µl of 1:1 PBS/Matrigel; Corning).

To evaluate the effect of cancer cell Ccr2 on the anti-tumor immune response locally versus systemically, contralateral transplantation was conducted with cancer cells with different Ccr2 genotypes transplanted into each inguinal mammary gland (2.5 × 10⁵ in 20 µl of 1:1 PBS/Matrigel). Cancer cells were isolated as described above, and both C57BL/6 and Ccr2^−/− mice were used as hosts.

To transplant irradiated cancer cells, freshly isolated primary MMTV-PyMT;Ccr2^+/+ or MMTV-PyMT;Ccr2^−/− cancer cells were plated in a 10-cm Petri dish (2 × 10⁶ per dish) in RPMI with 10% FBS and received 80 Gy of irradiation in the Gammacell 40 Exactor (Best Theratronics). After irradiation, cancer cells were washed with PBS and injected into the inguinal mammary glands of host mice (2.5 × 10⁵ in 20 µl of PBS). To generate cancer cell–conditioned medium, primary cancer cells from either MMTV-PyMT;Ccr2^+/+ or MMTV-PyMT;Ccr2^−/− tumors were plated in a 10-cm dish (2 × 10⁶ per dish) in DMEM/F12 medium plus 10% FBS and 1% penicillin/streptomycin, with or without CCR2 antagonist (RS504393; 10 µM; Sigma-Aldrich) for 2 d. The conditioned medium was collected and centrifuged at 300× g for 10 min, and the supernatant was used in experiments.

Tumor cells from MMTV-PyMT;Ccr2^+/+ or MMTV-PyMT;Ccr2^−/− mice were transplanted as described above. For anti–PD-L1 treatment, on days 15, 18, 21, and 24 after transplantation (after tumors had formed), mice received 200 µg of anti–PD-L1 antibody by intraperitoneal injection (Clone 10F.9G2; Bio X Cell) or control rat IgG2b antibody (Bio X Cell; Winograd et al., 2015). For CD8⁺ T cell depletion, mice were injected intraperitoneally with 200 µg of anti-CD8a antibody (Clone 2.43; BE0061; Bio X Cell) or control rat IgG2b antibody (Clone LTF-2; BE0090; Bio X Cell) every 3 d starting from day 0 and until day 30. Tumor size was measured biweekly, and mice were sacrificed at the IACUC-approved endpoint.

Tumors were harvested and mechanically dissociated for 30 min with collagenase D (2 mg/ml; Sigma-Aldrich) and DNase I (4 U/ml) in RPMI. For flow cytometry of cells from lymph nodes, the lymph nodes were forced through a 100-µm cell strainer, and flow-through cells were collected for experiments. Cells were resuspended in 1 × HBSS supplemented with 0.5% BSA and centrifuged at 300 ×g. The cell suspension was filtered through a 70-µm cell strainer, and red blood cells were lysed using red blood cell lysing buffer (Sigma-Aldrich) for 1 min at room temperature.

For flow cytometry staining, cells (1 × 10⁶) were incubated with mouse Fc Block (clone 2.4G2; BD Biosciences) for 10 min on ice, and then stained with the appropriate antibodies to surface markers at 4°C for 30 min in the dark or permeabilized and stained with intracellular antibodies overnight (see below for antibodies). Cell viability stain Zombie Red (BioLegend) was used to differentiate between dead and live cells. The stained populations were analyzed using an LSR II flow cytometer (BD Biosciences; see Fig. S2 for gating strategies) and FlowJo software (Version 10; BD Biosciences). Antibodies: CD45-eFluor 450 (Clone 30-F11), CD11c-PE-Cy7 (Clone N418), MHC II (I-A/I-E)–APC-eFlour780 (Clone M5/114.15.2), CD103-FITC (Clone 2E7), CD274 (B7-H1, PD-L1)–PE-Cy7 (Clone MIH5), CD3-FITC (Clone 17A2), CD8-eFluor450 (Clone 53–6.7), CD107a (LAMP-1)–PE (Clone 1D4B), IFN-γ–PE (Clone XMG1.2), Granzyme B–FITC (Clone NGZB), γδ TCR-PE-Cy5 (Clone GL3), and FoxP3-PE (Clone NRRF-30) were all from eBioscience; CD4-PerCP-Cy5.5 (Clone BM9), CD69-PE (Clone H1.2F3), F4/80-PerCP-Cy5.5 (Clone BM9), MHC I (H-2)–PE (Clone M1/42), CD326 (EpCAM)-APC (Clone G8.8), and CD86-BV510 (Clone GL-1) were all from BioLegend; CD45-APC (Clone 30-F11), Ly6G-FITC (Clone 1A8), and CD11b-PE (Clone M1/70) were from BD Biosciences; EpCAM-APC (Clone CI:A3-1) was from AbD Serotec; and CCR2-flourescein (Clone 475301) was from R&D Systems. Prior to Granzyme B staining, cells were incubated for 2 h with Brefeldin A (#B6542; Sigma-Aldrich).

Tumors (8–10 mm in diameter) were isolated from MMTV-PyMT;Ccr2^+/+ and MMTV-PyMT;Ccr2^−/− mice and immediately snap-frozen in liquid nitrogen. They were then homogenized in PBS with the addition of protease inhibitors (Promega). Proteome Profiler Mouse Cytokine Array Kit, Panel A (R&D Systems) was used according to the manufacturer’s instructions. Films were scanned, and spots were analyzed using ImageJ software.

Proliferation of cancer cells isolated from MMTV-PyMT;Ccr2^+/+ and MMTV-PyMT;Ccr2^−/− mice was measured using the CellTiter 96 AQueous One Solution Cell Proliferation Assay (MTS; Promega) according to the manufacturer’s instructions. In this assay, the MTS tetrazolium compound (Owen’s reagent) was bioreduced by metabolically active cells into a colored formazan product. The quantity of formazan product was measured in a 96-well plate by absorbance at 490 nm.

Total RNA from cancer cells (cultured 48 h in RPMI supplemented with 10% FBS) or primary tumors was extracted using an RNAeasy kit (Qiagen), following the manufacturer’s instructions. RNA was quantified using a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific). Reverse transcription was performed with an ImProm-II Reverse Transcription System (Promega) for the cancer cells or the RevertAid First Strand cDNA Synthesis system (K1622; Thermo Fisher Scientific) for the primary tumors, with the following cycling conditions: 5 min at 70°C, 10 min at 4°C, 5 min at 25°C, 60 min at 42°C, and 15 s at 70°C using 1 µg total RNA extract for 20 µl of final volume. RT-qPCR analysis was performed using a TaqMan gene expression assay (Thermo Fisher Scientific) with the following specific primers: Ccr2: Mm00438270_m1; Pd-l1: Mm00452054_m1; Cd8a: Mm01182107_g1; Ifn-g: mm01168134_m1; Pd-1: mm01285676_m1; Gzmb: mm00502528_m1; Ctla4: mm00486849_m1; Prf1: mm00812512_m1; Clec9a: mm00554956_m1; Cxcl9: mm00434946_m1; Cxcl10: mm00445235_m1; Cxcr3: mm00438259_m1; Cxcr4: Mm01996749_s1; Cd11c: mm00498698_m1; Cd40: mm00441891_m1; Cd80: mm00711660_m1; Cd86: mm00444543_m1; Ccl5: mm01302427_m1; Il-12b: mm01288989_m1; Tim3: mm00454540_m1; Csf2: mm01290062_m1; Flt3l: mm00442801_m1; Flt3: mm00439016_m1; Ccl2: mm00441242_m1; Csf3: mm00438334_m1; Csf1: mm00432686_m1; and Tbp: mm01277042_m1.

For cancer cells, RT-qPCR was performed on three independently isolated cancer cell populations and in triplicate for each sample. To determine the effects of IFN-γ on Pd-l1 expression, cancer cells were cultured with the indicated concentration of IFN-γ (#485-MI-100; R&D Systems) for 48 h before isolation of RNA. For tumor samples, RT-qPCR was performed on at least five primary tumors from different mice and in triplicate for each sample. Relative quantitation was performed with the 2^(−ΔΔCT) method using β-Actin or Tbp expression for normalization and MMTV-PyMT;Ccr2^+/+ cancer cells or tumor samples as a reference sample.

RNA ISH was performed on PFA-fixed paraffin-embedded tissue sections using an RNAscope Chromogenic 2.0 Detection Kit (Advanced Cell Diagnostics [ACD], Inc.). A Ccr2 probe was custom designed by ACD, Inc. (#436261-C2) and used according to the manufacturer’s instructions. We confirmed that no signal was found with the probe using Ccr2^−/− tissues. RNA FISH for Ccr2, Krt8, Cd8a, and Ifn-g was performed on fresh frozen tissue sections using an RNAscope Fluorescent Multiplex Assay Kit (ACD, Inc.) with the following probes (by ACD, Inc.): Ccr2: 436261-C2; Krt18: 424531-C1; Cd8a: 401681-C1; and Ifn-g: 311391-C3. They were used according to the manufacturer’s instructions. For Ccr2 RNA FISH, we confirmed that no signal was found with the probe using Ccr2^−/− tissues.

PFA-fixed paraffin-embedded tissue sections were deparaffinized and rehydrated following standard protocols, stained with Gill’s Hematoxylin (REF6765008; Thermo Fisher Scientific) for 3 min, washed in tap water, rinsed with Nu-Clear II (REF6769009; Thermo Fisher Scientific), and then rinsed again first with tap water and then with Bluing Reagent (REF 6769001; Thermo Fisher Scientific). Slides were then stained with eosin (HT110180; Sigma-Aldrich), dehydrated following standard protocols, air-dried, and mounted with Cytoseal 60 (#831016; Thermo Fisher Scientific).

To stain for cytokeratin 5 and cytokeratin 8, paraffin sections were deparaffinized and rehydrated, and antigen retrieval was performed by boiling slides in Tris EDTA buffer (10 mM Tris base and 1 mM EDTA, pH 9.0) for 6 min in a pressure cooker. The slides were blocked with 1× blocking buffer (PBS containing 2.5% BSA and 5% goat serum) and Fc receptor blocker (Innovex Biosciences) before incubating with anti–cytokeratin 5–Alexa Fluor 488 conjugated antibody (ab193894; 1:200 dilution; Abcam) and anti-Cytokeratin 8-Alexa Fluor 405 conjugated antibody (ab210139; 1:200 dilution; Abcam) overnight at 4°C. Sections were counterstained with ToPro-3 (1:1,000 dilution; Life Technologies).

For all other stainings, frozen tissue sections were incubated with 1x blocking buffer (5% goat serum and 2.5% BSA in PBS) and Fc receptor blocker. Sections were incubated with rabbit anti-CD3 polyclonal antibody (1:500 dilution; Abcam), rat anti–MHC class I monoclonal antibody (Clone ER-HR52, 1:100 dilution; Abcam), rat anti-PyMT antibody (Ab15085, 1:200 dilution; Abcam), rabbit anti-cleaved caspase-3 (Asp175; D3E9; #9579; 1:200 dilution; Cell Signaling Technology), anti–CD8a–Alexa Fluor 488 (Clone 5.3–6.7), anti–CD11c–Alexa Fluor 488 (Clone N418), or anti–CD326 (EpCAM)-APC (Clone G8.8) in 0.5× blocking buffer overnight at 4°C. Secondary antibodies (not used for primary antibodies conjugated to Alexa Fluor 488) anti–rabbit-Alexa568 or anti–rat-Alexa568 (1:150 dilution; Life Technologies) were used for detection, and sections were counterstained with DAPI (1:200 dilution; Life Technologies). Images were collected at 40× magnification using a Leica SP8 confocal or an AX10 microscope and an AxioCam HRc camera (Zeiss) and were analyzed using Volocity software (Version 6.3.0; PerkinElmer).

Sections were deparaffinized and rehydrated. For phospho-STAT3 and Ki67 staining, antigen retrieval was performed by boiling slides in 10 mM sodium citrate buffer (pH 6.0) for 6 min in a pressure cooker, while for phospho-STAT1 staining, antigen retrieval was done in Tris EDTA buffer (10 mM Tris base and 1 mM EDTA, pH 9.0). The slides were blocked with 3% hydrogen peroxide, 1× blocking buffer (PBS containing 2.5% BSA and 5% goat serum), Fc receptor blocker, and finally avidin/biotin blocking buffer (SP-2001; Vector Laboratories). Slides were then incubated with primary antibodies overnight at 4°C in 0.5× blocking buffer: rabbit anti–Ki67/MK167 (1:1,000 dilution; NB110-89717; Novus International), rabbit anti–phospho-STAT1 (1:200 dilution; Cell Signaling Technology), or rabbit anti–phospho-STAT3 (1:500 dilution; Cell Signaling Technology). After incubating with secondary biotinylated goat anti-rabbit IgG antibody (1:500 dilution; BA-1000; Vector Laboratories) for 1 h at room temperature, slides were incubated with avidin-conjugated horseradish peroxidase (PK-6100; Vector Laboratories) for 30 min, and the signals were detected by a 3,3′-diaminobenzidine substrate kit (SK-4100; Vector Laboratories). Lastly, sections were counterstained with hematoxylin. TUNEL staining was performed on deparaffinized sections to detect late-stage apoptotic cells using the ApopTag peroxidase in situ apoptosis detection kit (Millipore) according to the manufacturer’s instructions. Images were scanned by an Aperio ScanScope CS System (Leica Biosystems), and positive nuclei were counted using Aperio eSlide Capture Devices software (Leica Biosystems).

Ccr2^+/+ or Ccr2^−/− cancer cells were isolated from tumors and sorted as EpCAM⁺CD45⁻CD31⁻ live cells (using 7-AAD Viability Staining Solution; Thermo Fisher Scientific) 4 wk after orthotopic transplantation into the mammary glands of wild-type C57BL/6 mice. Total cancer cell RNA was extracted using a TRIzol extraction protocol. cDNA libraries were prepared using the Ovation RNA-Seq System V2 (NuGEN Technologies) according to the manufacturer’s protocol and sequenced using a NextSeq 500 instrument (Illumina) to obtain 388 million 76-bp single-end reads. Reads were mapped to the reference mm9 mouse genome using Spliced Transcripts Alignment to a Reference (Dobin et al., 2013) and were determined using HTSeq (Anders et al., 2015). Differential gene expression analysis was performed using DESeq2 (Love et al., 2014) with a false discovery rate cutoff of 5%. Hierarchical clustering was performed by sample and gene using normalized and log₂-transformed gene expression values of differentially expressed genes. All of the analyses described above were performed on the Bioinformatics Shared Resource Galaxy server at CSHL. Gene set enrichment analysis (Subramanian et al., 2005) was performed on a preranked gene list sorted by log₂FoldChange value against v5.1 Hallmark gene sets using default parameters. GO term enrichment analysis was performed using GOrilla PANTHER (Eden et al., 2009). The RNA-seq data have been deposited in the Sequence Read Archive with accession no. PRJNA605445 (https://www.ncbi.nlm.nih.gov/Traces/study/?acc=PRJNA605445).

BMDCs were isolated and cultured as previously described (Inaba et al., 1992). Briefly, bone marrow was obtained from female C57BL/6 mice by flushing the femur and tibia with 2 ml of HBSS using a 1-ml insulin syringe with a 29G × 1/2 needle. The cells were washed with HBSS twice by centrifugation at 250 ×g for 8 min and then suspended in BMDC culture medium (RPMI-1640 medium [Thermo Fisher Scientific] containing 10% FBS, 20 ng/ml recombinant mouse GM-CSF [BioLegend], 50 U/ml penicillin, and 50 µg/ml streptomycin). The cells were then plated in 10-cm culture plates (2 × 10⁶ cells per Petri dish) and kept at 37°C in a 5% CO₂ incubator for 3 d, and then split at day 4 into two plates. For activation of BMDCs with cancer cell–conditioned medium, the nonattached BMDCs were harvested at day 6 by gently pipetting the cultures with medium. This BMDC population was then rinsed three times in HBSS and suspended in (1) culture medium, (2) conditioned medium from MMTV-PyMT;Ccr2^−/− or MMTV-PyMT;Ccr2^+/+ cancer cells, or (3) conditioned medium from MMTV-PyMT;Ccr2^+/+ cancer cells cultured either with or without CCR2 antagonist (10 µM; RS504393; Sigma-Aldrich). After incubation with conditioned medium for 24 h, nonadherent BMDCs were harvested for flow cytometry.

Cytotoxic CD8⁺ T cells were isolated from spleens of OT-I transgenic mice by magnetic labeling and separation (Miltenyi Biotec). Briefly, the spleen was forced through a 100-µm cell strainer, and flow-through cells were collected. After red blood cells were lysed, the splenocytes were washed and incubated with CD8a⁺ T Cell Biotin-Antibody Cocktail, mouse (Miltenyi Biotec) at 4°C for 5 min, followed by incubating with anti-biotin microbeads (Miltenyi Biotec) at 4°C for 10 min. The cell suspension was applied to an LS column (Miltenyi Biotec) in a magnetic field, and the flow-through (negatively selected) CD8⁺ T cells were collected. The isolated CD8⁺ T cells were cultured on plates coated with anti–mouse-CD3 (BioLegend) in RPMI-1640 containing 10% FBS, 50 µM of β-mercaptoethanol, 20 ng/ml mouse IL-2, 2 ng/ml mouse IL-7, and 1% penicillin/streptomycin for 6–9 d. Primary cancer cells were isolated as described above from MMTV-PyMT-chOVA;Ccr2^+/+ or MMTV-PyMT-chOVA;Ccr2^−/− mice and cultured overnight with or without CCR2 antagonist (10 µM; RS504393) and OVA peptides (1 µg/ml) in DMEM/F12 medium with 10% FBS. The E0771 cancer cell line was infected with lentivirus expressing luciferase and OVA SIINFEKL peptide fused to the C-terminus of enhanced GFP. Transduced cells were sorted by flow cytometry for enhanced GFP–positive cells. For chromium (⁵¹Cr) release assay, cancer cells (1,000 per well) were incubated with 50 µl of ⁵¹Cr solution (activity 1 mCi/ml; PerkinElmer) plus 50 µl of RPMI medium with 10% FBS for 2 h at 37°C. After the cells were rinsed three times with RPMI plus 10% FBS, 50 µl of cancer cells (1,000) were plated into each well of a 96-well V-bottom plate, followed by the addition of 50 µl of CD8⁺ T cells at a ratio of 10:1. After 4 h of incubation at 37°C, 50 µl of supernatant from each well was transferred to miniature vials, and radioactivity was determined by a Beckman scintillation counter (PerkinElmer). The percentage of specific lysis was calculated using the standard formula [(experimental − spontaneous release)/(total − spontaneous release) × 100] and expressed as the mean of triplicate samples.

Snap-frozen tumor samples were homogenized in radio immunoprecipitation assay buffer with cOmplete Protease Inhibitor Cocktail Tablets (Roche) and phosphatase inhibitor cocktail (Thermo Fisher Scientific), and lysis was performed on ice for 30 min. Cell lysates were centrifuged for 15 min at 12,000 ×g at 4°C. After centrifugation, the supernatant was mixed with 4× loading buffer plus 2.5% of β-mercaptoethanol and loaded on 8% SDS/PAGE gels. Proteins with similar molecular weight were detected using blots from separate gels loaded with an equal amount of samples. Protein bands were transferred to polyvinylidene difluoride membranes (Bio-Rad) at 90 V for 2 h in a Bio-Rad Mini Trans-Blot system. The membrane was incubated sequentially with different primary antibodies: anti-p65, anti–p-p65, anti–p-STAT1, anti–p-STAT3, and anti-STAT3 (all from Cell Signaling Technology) and anti-STAT1 and anti–β-Actin (Santa Cruz Biotechnology). Secondary antibodies were from LI-COR. Protein detection was performed using the Odyssey imaging system (LI-COR).

Mammary glands were isolated and spread out on glass slides. Then, the mammary glands were fixed in Carnoy’s fixative (60% ethanol, 30% chloroform, and 10% glacial acetic acid) for 2–4 h at room temperature, followed by 15-min washes first in 70% and then in 50% ethanol, with a final rinse in double-distilled water. The glands were then stained with carmine red stain (2.5 g of aluminum potassium sulfate, 1 g of carmine, and dH₂O to a final volume of 500 ml, and boiled for 25 min) for 16 h or until the fat pad was a uniform pink color. After staining, the mammary glands were de-stained in 1% solution of 1N HCl in 70% ethanol. De-stained glands were rinsed twice for 15 min in 70% ethanol followed by 15-min washes first in 90% and then in 100% ethanol. The mammary glands were then placed in xylene for 2 h for clearing and were mounted.

All statistical analyses were performed using GraphPad Prism Version 6 or newer software. Data were analyzed using Kaplan-Meier survival analysis, Student’s t tests, ANOVA, χ² test, and log-rank (Mantel-Cox) test, as indicated in the figure legends, with an α of 0.05. For tumor volume analysis, one-way ANOVA followed by Sidak’s multiple comparisons test; Welch’s ANOVA followed by Dunnett’s T3 multiple comparisons test; or Student’s t tests were performed as indicated in the legends. The number of sampled units, n, is indicated in the figure legends.

Fig. S1 (related to Fig. 1) shows CCR2 expression in spontaneous MMTV-PyMT tumors and lungs. Fig. S2 shows the flow cytometry gating strategy for immune cells. Fig. S3 (related to Figs. 4 and 5) shows T cell infiltration and depletion in tumors derived from Ccr2^+/+ or Ccr2^−/− cancer cells. Fig. S4 (related to Figs. 8 and 9) shows the activation of T cells and the infiltration of myeloid cell in tumors derived from Ccr2^+/+ or Ccr2^−/− cancer cells. Fig. S5 (related to Fig. 10) shows cytokine analysis in tumor lysate and conditioned medium.

We thank Laura Maiorino for drawing the included illustrations.

This work was supported by funds from the CSHL Cancer Center (P30-CA045508) and the Department of Defense (W81XWH-14-1-0078) to M. Egeblad; from the Charles and Marie Robertson Foundation to X.-Y. He; from the Lustgarten Foundation, the National Cancer Institute, and the Cedar Hill Foundation to D.T. Fearon; from the Rita Allen Foundation and the V Foundation for Cancer Research to C.O. dos Santos; and a Starr Centennial Scholarship from the Watson School of Biological Sciences to E. Bružas. R. Yan is a George A. and Marjorie H. Anderson Fellow. M. Egeblad, C.O. dos Santos, and L. Van Aelst were supported by a grant from the Simons Foundation to CSHL. Support was also provided by the State of New York (contract no. C150158), and the opinions, results, findings and/or interpretations of data contained therein do not necessarily represent the opinions, interpretations, or policy of the State.

Author contributions: M.R. Fein, X.-Y. He, and M. Egeblad designed the experiments. M.R. Fein and X.-Y. He performed the in vitro experiments; M.R. Fein, C.O. dos Santos, and E. Bružas isolated RNA from cancer cells and performed the RNA-seq analysis. M.R. Fein and X.-Y. He performed the animal experiments; A.S. Almeida, M.R. Fein, A. Pommier, and X.-Y. He performed flow cytometry; J.E. Wilkinson analyzed the histological sections; and M.R. Fein, X.-Y. He, and A. Eberhardt performed immunohistochemistry and ISHs. X.-Y. He, A. Pommier, R. Yan, and M.R. Fein designed and performed the chromium release experiments; and A. Eberhardt, M.R. Fein, and X.-Y. He performed qPCR. L. Van Aelst and D.T. Fearon provided experimental advice. M.R. Fein, X.-Y. He, and M. Egeblad wrote the manuscript.