Monocyte-derived macrophages promote breast cancer bone metastasis outgrowth

Macrophages potently promote breast cancer progression and act at all steps of the metastasis cascade (Kitamura et al., 2015b; Qian and Pollard, 2010). However, these cells are heterogeneous, and their phenotype and function are tightly regulated by the local tissue environment (Gordon and Plüddemann, 2019). It has become evident that most tissue-resident macrophages originate from embryonic precursor cells that can persist through life. In some cases, they are replaced or complemented by bone marrow–derived precursor cells, monocytes. In contrast, adult macrophages in pathophysiological situations are largely derived from monocytes (Wynn et al., 2013). Several studies, including our own, indicate that monocyte-derived macrophages are critical for tumor growth and metastasis (Afik et al., 2016; Franklin et al., 2014; Lin et al., 2001; Qian et al., 2011). However, in some cases, tissue-resident macrophages can also have an important function in tumor promotion (Zhu et al., 2017). Therefore, a better understanding of the origin and function of macrophage subsets in different cancer types and secondary organs is critical for developing precise treatment to maximize clinical benefit.

Bone metastasis is the most prevalent form of metastasis in breast cancer and accounts for >70% of cases (Coleman, 2001). It causes bone pain, bone loss, fracture, and other skeletal related events that dramatically impair the patient’s quality of life, and it is lethal (Coleman, 2006; Weilbaecher et al., 2011). Although current therapeutic modalities such as radiotherapy, chemotherapy, and antiosteolytic agents can reduce morbidity associated with bone metastasis, these treatments often provide only a minimum survival benefit to patients (Weilbaecher et al., 2011). This indicates that a better understanding of the mechanisms that lead to bone metastasis outgrowth will be critical for the development of effective treatment. Recent studies strongly argue that intricate interaction between tumor cells and the supportive stroma play important roles in the development of this deadly disease (Liu et al., 2019; Price et al., 2016; Sethi et al., 2011; Wang et al., 2015; Wu et al., 2017). For example, the osteogenic niche composed of bone-building osteoblasts has recently been shown to be critical for the survival and colonization of disseminated tumor cells in the bone (Shiozawa et al., 2011; Wang et al., 2015). However, the specific role of macrophage subsets in breast cancer bone metastasis is largely unknown.

In the current study, our data indicate that macrophages are critical for bone metastasis outgrowth in vivo. Combining genetic ablation models of CC chemokine receptor 2 (Ccr2^−/−), IL-4 receptor (Il4r^−/−), and a monocyte add-back approach, we have identified that a population of monocyte-derived macrophages, rather than bone-resident macrophages, promote breast cancer bone metastasis outgrowth in an IL4R-dependent manner.

To determine the infiltration of macrophages in patient bone metastasis samples, we stained a small cohort of patient samples (for patient information, see Table S1) with CD68, a commonly used macrophage marker. All patient samples showed significant amounts of macrophage infiltration (472.8 ± 189.7 macrophages/mm² tumor area; Fig. 1 A). Similar levels of macrophage infiltration detected by Iba1 staining are observed in a widely used in vivo model of bone metastasis using 1833 cells (a subline of MDA-MB-231 human triple-negative breast cancer cells; Kang et al., 2003) in nude mice. This infiltration is also found in bone metastasis derived from MetBo2 cells, a newly developed bone metastasis cell line derived from polyoma middle T oncogene–induced mouse mammary tumor on a syngeneic Friend Virus B NIH Jackson (FVB) background (Fig. 1 B). These results confirm that these two models have a similar level of macrophage infiltration as seen in patient bone metastasis.

Bone metastasis outgrowth leads to macrometastatic lesions that significantly impair patient survival and quality of life. To determine the role of macrophages in bone metastasis outgrowth, we adopted a widely used in vivo macrophage-ablation method of treating mice with liposome-encapsulated clodronate (L-Clod; Van Rooijen and Sanders, 1994; Qian et al., 2009) after tumor detection in the MetBo2 and 1833 models (Fig. 1, C and D). As expected, L-Clod treatment significantly depleted macrophages in bone metastasis without affecting other immune cell populations compared with the treatment with control liposome-PBS (L-PBS; Fig. S1 A). Such macrophage ablation with L-Clod that was performed after detection of established MetBo2 cell bone lesions significantly inhibited metastasis outgrowth in syngeneic FVB mice (Fig. 1 D). Colony-stimulating factor 1 receptor (CSF1R) is a critical lineage receptor for macrophage differentiation and function in vivo (Stanley et al., 1997). Using a specific small-molecule CSF1R inhibitor, BLZ945, we observed a significant inhibition of bone metastasis growth of MetBo2 cells in syngeneic FVB mice compared with vehicle treatment (Fig. 1 E). Similarly, BLZ945 significantly inhibited bone metastasis growth of 1833 cells in nude mice (Fig. 1 F). Consistent with their mode of action, BLZ945 and L-Clod significantly reduced macrophage abundance in bone lesions without affecting other major immune cell populations (Fig. S1 B).

Osteoclasts, specialized cells of the macrophage lineage, have been shown to engage in paracrine signaling with cancer cells that promote breast cancer bone disease (Maurizi and Rucci, 2018). These cells were also depleted by L-Clod treatment (Fig. S2 A). These data revealed that although L-Clod was originally designed to target phagocytic macrophages that actively uptake liposomes (Van Rooijen and Sanders, 1994), it also affects osteoclasts. Therefore, it is possible that part of the metastasis-inhibiting effect observed in our experiments is through inhibition of osteoclasts. Soluble/free clodronate (not encapsulated in liposome) is known to target osteoclasts because of its high affinity to bone. This treatment was effective (Fig. S2 B), as expected, and did not affect infiltration of bone metastasis–associated macrophages (BoMAMs) in our models (Fig. S1 C). Although osteoclast inhibition with free clodronate alone can inhibit bone metastatic growth of MetBo2 cells, the inhibition was 2.4-fold more effective with L-Clod, which depleted both osteoclasts and BoMAMs (Fig. S2 D). Together, these data support the conclusion that BoMAMs can directly promote metastatic outgrowth independently of osteoclasts and that macrophage ablation can effectively inhibit bone metastasis outgrowth in vivo.

Macrophages in most tissues can be derived from both bone marrow monocyte precursor cells and tissue-resident macrophages originated from yolk sac precursor cells (Cassetta and Pollard, 2018; Schulz et al., 2012). Tissue-resident macrophages with a profibrotic phenotype have been shown to promote tumor progression in pancreatic cancer models (Zhu et al., 2017). Bone marrow–resident macrophages express CD169 as a cell surface marker (Batoon et al., 2019; Hashimoto et al., 2013). In our model, CD169⁺ macrophages were mainly associated with tumor-adjacent stroma (Fig. S3 A). These CD169⁺ macrophages could be efficiently depleted in transgenic mice expressing diphtheria toxin (DT) receptor under the control of CD169 promoter (CD169-DTR) upon DT treatment (Fig. S2, B and C), which led to a significant reduction of total macrophage numbers (Fig. S3 B) compared with control treatment with a mutant toxin (Glu52-mutated DT [Glu-DT]) which, because of a point mutation, does not bind to the DT receptor. This ablation did not affect bone metastasis growth (Fig. S3 C). Immunofluorescent staining showed a significant ablation of CD169⁺ macrophages without affecting the majority of tumor-infiltrating macrophages labeled by the lineage marker, Iba1 (Fig. S3 D). This indicates that bone-resident macrophages do not play a major role in our model and strongly suggests that tumor-promoting BoMAMs were derived mainly from recruited monocytes.

Monocytes are heterogeneous, and the two most well-characterized populations in mice are Ly6C-expressing IMs (classical monocytes) and patrolling monocytes (PMs; nonclassic or resident monocytes) lacking Ly6C (Auffray et al., 2009; Geissmann et al., 2010). To understand the origin of BoMAMs, we measured monocyte recruitment after adoptive transfer of each subset. Mouse monocytes were identified by their expression of CD11b and CD115 and were FACS sorted into IMs and PMs based on Ly6C expression. As previously reported, both populations have similar levels of GFP expression in Csf1r–GFP transgenic mice (Qian et al., 2011). We adoptively transferred 10⁵ cells of each population into syngeneic FVB mice bearing bone metastasis of MetBo2 cells or healthy mice as controls (Fig. 2 A). 18 h after adoptive transfer, we determined the number of GFP-positive monocytes in the bone to measure their relative recruitment. Our previous studies have shown that at this time point, similar numbers of donor cells are present in the blood, showing equivalent availability (Qian et al., 2011). In healthy bone morrow, these two populations were recruited to a comparable level. However, in bone metastases, IMs were preferentially recruited, with approximately twofold enrichment (Fig. 2 B).

CCL2 and CC chemokine receptor 2 (CCR2) signaling is the major chemoattractant that mediates the recruitment of Ly6C⁺ IMs (Getts et al., 2008; Palframan et al., 2001). Consistent with a previous report (Shi et al., 2011), circulating IMs were significantly reduced in Ccr2^−/− mice compared with Ccr2^+/− or WT littermates (Fig. S4 A), but this reduction was not observed in the bone marrow (Fig. S4 B). The bone colonization efficiency of MetBo2 cells was significantly inhibited in Ccr2^−/− mice, leading to reduced metastatic growth and prolonged survival of mice compared with WT controls (Fig. 2, C and D). Consistent with these data, the bone colonization efficiency of 1833 cells was also significantly inhibited in mice with CCR2 knockout compared with WT controls (Fig. 2, E and F). Furthermore, CCL2 blockade using neutralizing antibody 48 h after tumor cell inoculation significantly inhibited 1833 bone metastasis growth (Fig. 2 G). Because our previous studies indicate that tumor cell metastatic extravasation and seeding mostly occur within 36 h after reaching the distal organ, these data strongly suggest that recruitment of IMs is critical for tumor metastatic outgrowth after seeding. To further test this, we established bone metastasis of MetBo2 cells in Ccr2^−/− mice, which have deficiency in circulating monocytes and delayed tumor growth. Upon bone metastasis detection, add-back of WT IMs via adoptive transfer significantly promoted tumor outgrowth compared with Ccr2^−/− monocytes (Fig. 2 H). Together, these data indicate that CCL2/CCR2-dependent recruitment of IMs is critical for breast cancer bone metastasis outgrowth.

To further characterize the BoMAMs, we performed gene expression profiling of macrophages isolated from bone metastasis and identified 409 up-regulated genes and 527 down-regulated genes compared with macrophages from healthy bone, with a threshold of 1.5-fold change and P < 0.05 (Fig. 3 A). For major macrophage polarization marker genes, macrophages associated with bone metastasis express higher levels of some alternative activation markers (e.g., Il4ra and Il10), but also some of the classic activation markers (e.g., Tnf and Il1b; Fig. 3 B) compared with their counterparts from healthy bone. Further analysis indicated up-regulation of genes related to several tumor-promoting functions, such as angiogenesis, tumor growth, cancer stem cells, immune suppression, etc., in BoMAMs compared with normal counterparts (Fig. 3 C). These results are consistent with previous reported tumor promotion functions of macrophages in other contexts (Cassetta et al., 2019; Ojalvo et al., 2009; Yang et al., 2018). Immunophenotyping of macrophages (defined by CD45⁺CD11b⁺F4/80⁺Ly6C⁻Ly6G⁻SSC^lo; Chow et al., 2011) isolated from bone metastasis of MetBo2 cells and adjacent normal tissue and compared with their counterparts from normal bone using a set of macrophage polarization markers, including CD204, CD206, CD86, CD64, MHCII, and IL4R, indicated that all three macrophage populations have heterogeneous expression of these markers (Fig. 3 D). BoMAMs have higher IL4R expression and increased abundance of CD204⁺ cells (Fig. 3 D). Multiparameter FACS analysis indicated that IL4R and CD204 colocalize and label a distinct population of macrophages that are specific for bone metastasis (Fig. 3 E). Immunofluorescent staining further confirmed that CD204 staining specifically labeled macrophages that infiltrate inside the bone metastasis, and that they have relatively low F4/80 expression compared with the macrophages in adjacent normal bone marrow (peritumor; Fig. 3 F). These CD204⁺ macrophages consistently coexpress IL4R and the lineage marker Iba1, as assessed by double immunofluorescent staining (Fig. 3 G). Importantly, in patient bone metastases samples, CD204⁺ macrophages infiltrated into the tumor in all the cases in our cohort, and the majority of these macrophages also expressed IL4R (Fig. 4 A). Furthermore, in two independent breast cancer bone metastasis datasets, CD204 expression was strongly correlated with a macrophage IL4R target gene score (Palazon et al., 2017; Zhang et al., 2009; Fig. 4 B). Together, these data strongly suggest that IL4R signaling in macrophages plays an important role in breast cancer bone metastasis.

Ly6C⁺CCR2⁺ monocytes are mostly CD204⁺ in both healthy and bone metastasis–bearing mice (Fig. 5 A), and bone metastasis–associated monocytes have significantly higher IL4R expression compared with normal counterparts in gene expression profile analysis (Fig. 5 B). Because our data indicate that Ly6C⁺ IM recruitment is enhanced in bone metastasis (Fig. 2 B), we hypothesized that CD204⁺IL4R⁺ BoMAMs are mainly derived from IMs. To test this directly, we performed lineage tracing experiments using adoptive transfer of Ly6C⁺ monocytes FACS sorted from ubiquitous GFP-expressing mice (FVB.Cg-Tg(CAG-EGFP)B5Nagy/J) into syngeneic mice bearing established bone metastasis and recovered GFP⁺ cells for immunophenotyping (Fig. 5 C, left). Compared with endogenous macrophages (GFP⁻), monocyte-derived macrophages (GFP⁺) almost exclusively expressed both CD204 and IL4R and represented tumor-infiltrating BoMAMs, whereas the endogenous macrophages (GFP⁻) were largely IL4R⁻ and had dichotomized expression of CD204 that was consistent with data in Fig. 3 A (Fig. 5 C). These data directly indicate that the majority of IL4R⁺ BoMAMs are derived from recruited IMs.

IL4R has been shown to have multiple functions in different cell types including macrophages, T and B lymphocytes, and nonhematopoietic cells (Ho and Miaw, 2016). To test the role of macrophage IL4R in bone metastasis, we crossed Il4ra^−/− mice (Mohrs et al., 1999) with Rag2^−/− mice that lack mature T and B cells. These mice were then used as bone marrow donors to generate mosaic mice in lethally irradiated Foxn1 nude mice. As a result, these mosaic mice lack mature T and B function, and the IL4R deficiency is limited to bone marrow–derived innate immune cells, predominantly macrophages. The bone colonization potential of 1833 cells was significantly inhibited in mosaic mice bearing Il4ra^−/− bone marrow compared with control mosaic mice containing Il4ra^+/− bone marrow (Fig. 5 D). Furthermore, osteoclast abundance in IL4R^−/− bone marrow mosaic mice was comparable with that in control mice with Il4ra^+/− bone marrow (Fig. S5 A). In contrast, abundance of CD204⁺ BoMAMs was significantly reduced (Fig. S5 B). This suggests that IL4R is important for activation of BoMAMs but not osteoclasts.

Additional experiments were performed to determine Il4 mRNA expression using quantitative PCR on various cell types purified from the bone metastasis environment by FACS sorting. Our data indicate that CD4⁺ T cells have the highest Il4 expression; tumor cells and B cells also have some Il4 mRNA expression; and all other cells type tested did not have detectable Il4 expression. (Fig. S5 C). This suggests paracrine signaling between CD4⁺ T cells and BoMAMs through IL-4, which is consistent with previous studies in primary breast cancer models (DeNardo et al., 2009).

Our data indicate that BoMAMs are derived from Ly6C⁺ IMs. The abundance of monocytes and macrophages in Il4ra^−/− bone marrow mosaic mice is similar to Il4ra^+/− controls with comparable bone metastasis burden (Fig. S5 B). This suggests that IL4R mainly affects BoMAM tumor-promoting function rather than recruitment. Indeed, in similar monocyte add-back experiments as described in Fig. 2 H, IL4R^−/− monocytes were less efficient at promoting bone metastasis growth compared with IL4R⁺/⁺ monocytes, despite equivalent recruitment (Fig. 5 E). Together, these data indicated that macrophage IL4R signaling is critical for BoMAM activation and tumor-promoting function.

Compelling evidence indicates that macrophages play important roles in cancer metastatic progression and that targeting macrophages could have significant therapeutic benefit (Cassetta and Pollard, 2018; Quail and Joyce, 2017; Ruffell and Coussens, 2015). However, metastasis still remains a major cause of death. In particular, in breast and other cancers, outgrowth of bone metastasis represent a major clinical challenge (Mundy, 2002). The current study provides direct evidence that macrophages are critical for breast cancer bone outgrowth. Macrophage ablation or inhibition of CSF1 receptor potently inhibited bone metastasis outgrowth of both mouse and human breast cancer cells in immune-competent and -deficient mice, respectively. These data suggest that macrophage targeting might be an effective therapeutic approach in treating breast cancer bone disease.

Macrophages can be derived from embryonic precursor cells and bone marrow monocytes (Jacome-Galarza et al., 2019). In the current study, our data strongly suggest that monocyte-derived BoMAMs are critical for bone metastasis outgrowth in an IL4R-dependent manner. BoMAMs express low levels of F4/80 in immunofluorescent staining, which is one of the characteristics of monocyte-derived macrophages (Schulz et al., 2012). Consistent with this data, a deficiency in CCR2-dependent monocyte recruitment significantly inhibited bone colonization efficiency, while the add-back of WT monocytes restored bone metastasis outgrowth. Embryonic precursor cells give rise to tissue-resident macrophages in almost all tissues (Perdiguero and Geissmann, 2016). Studies by Frenette’s group and others suggest that CD169⁺ macrophages represent such bone-resident macrophages, and that these control the perivascular hematopoietic stem cell niche (Chow et al., 2013). In the current study, depletion of these macrophages using CD169-DTR mice did not affect metastasis outgrowth. Consistent with this result, we found that the majority of the resident macrophages do not express IL4R, a receptor that is critical for bone metastasis outgrowth. Together with the data from Ccr2^−/− mice and the monocyte add-back rescue experiments, this indicates that, at least in the models tested here, bone metastasis–promoting macrophages are predominantly derived from recruited IMs. Interestingly, the contribution of tissue-resident macrophages can vary among different secondary organs and cancer types. For example, resident macrophages have been shown to play an important role in tumor progression of primary pancreatic adenocarcinomas (Zhu et al., 2017). Therefore, distinct macrophage subsets may play dominant roles in different cancer types and different secondary organs. This suggests the need for precisely targeted therapies to maximize patient benefit.

It is interesting that adoptive transfer of WT IMs, but not Ccr2^−/− counterparts, can significantly promote bone metastasis growth in Ccr2^−/− mice. These data suggest that CCR2-dependent monocyte recruitment is critical for bone metastasis outgrowth. Our previous studies showed a similar mechanism of monocytes in promotion of breast cancer lung metastasis seeding and extravasation (Qian et al., 2011). In both cases, tissue-resident macrophages did not seem to play major tumor supporting roles. Interestingly, Kuffer cell and microglia, tissue-resident macrophages of liver and brain respectively, have been shown to play important roles in the process of breast cancer metastasis to these two target organs (Bowman et al., 2016; Gordon et al., 2017; Hoshino et al., 2015; Zhu et al., 2017). Together, these data underline macrophage heterogeneity as an important component of tissue microenvironment (the “soil”) that determines tissue-specific metastatic mechanisms. Of note, the CCL2/CCR2 axis may also regulate the phenotype (Sierra-Filardi et al., 2014) or function of macrophages, as has been shown in melanoma, lung, and liver cancer models (Kitamura et al., 2015a; Li et al., 2017; Xu et al., 2019). Our data cannot exclude the possibility that a similar mechanism may also exist in breast cancer bone metastasis after the recruitment of IMs.

IL4R has been shown to induce alternative activation of macrophages (Varin et al., 2010), Ig production in B cells (Yanagihara et al., 1995), activation of type 2 T helper cells, and signaling in nonhematopoietic cells (Li et al., 2009). Previous studies have illustrated the role of IL4/IL4R signaling in tumor invasion and response to treatment in primary breast cancer (Andreu et al., 2010; Venmar et al., 2014) and glioma models (Pyonteck et al., 2013). The current study revealed that IL4R signaling is critical for the polarization and protumor function of BoMAMs. Together, these data suggest that the IL4/IL4R pathway can be a promising target in neutralizing tumor-promoting macrophages.

In summary, we have elucidated a novel mechanism that promotes breast cancer bone metastasis outgrowth through a specific subset of CD204⁺IL4R⁺ macrophages derived from CCR2-recruited IMs. This population of BoMAMs is preserved in patient bone metastasis samples. This mechanism can be potentially targeted to treat breast cancer bone disease.

All animal procedures were performed based on the UK Animals Act 1986 following local rules and approved project license or in accordance with National Institutes of Health regulations concerning the use and care of experimental animals, with approval by the Albert Einstein College of Medicine Animal Use Committee. FVB/N, athymic nude (outbreed, mixed background) and albino-b6 mice were bought from Charles River. B6.129S4-Ccr2tm1Ifc/J (Ccr2^−/−), B6.Il4ratm1Sz/J(Il4ra^−/−), B6(Cg)-Rag2tm1.1Cgn/J (Rag2^−/−), and FVB.Cg-Tg(CAG-EGFP)B5Nagy/J (FVB-eGFP) mice were bought from Jackson Laboratory. CD169-DTR mice were kindly provided by Prof. Paul Frenette (Albert Einstein College of Medicine, New York, NY). Csf1r-EGFP (MacGreen) mice on the C57BL/6 background were obtained from Dr. David Hume (University of Edinburgh, Edinburgh, UK). Ccr2^−/− mice and CD169-DTR mice were crossed with athymic mice. Il4ra^−/− mice were crossed with Rag2^−/− mice. For generation of bone marrow mosaic mice, recipient mice received 8 Gy irradiation at the age of 3 wk and rested for 5 h before an i.v. injection of bone marrow cells from donor mice. The mice were allowed to recover for 3 wk before metastasis assays were performed. All the in vivo experiments were repeated at least twice with more than three mice in each group.

The bone metastases samples were collected from nine patients diagnosed with grade IV breast cancer from 2016 to 2017. This study followed standard guidelines and was approved by the Ethics Committee of the Second Affiliated Hospital of Zhejiang University School of Medicine, Hangzhou, China. Written informed consent was obtained from all participants. Multiplex staining was performed with slides cut from formalin-fixed and paraffin-embedded patient bone metastasis samples. Their demographic information, including sex, sampling site, and molecular phenotype, is shown in Table S1. For PanCK staining, slides were retrieved in 0.125 mM EDTA solution (pH 8.0) at 97°C for 45 min in a pressure cooker. After cooling, samples were incubated in 30% vol/vol hydrogen peroxide for 30 min and blocked. Staining was performed using Alexa Fluor 488 Tyramide SuperBoost Kit (Thermo Fisher Scientific; B40941) via the protocol recommended by the antibody providers. For multiplexing staining, the method described previously was used (Prost et al., 2016) with protocols provided by the manufacturer. Anti-human primary antibodies used were PanCK (AE1/AE3, Thermo Fisher Scientific), CD204 (Abnova; SRA-E5), CD68 (Dako; PG-M1), and IL-4Ra (R&D Systems; MAB230). Other products used were serum-free block (Dako; X0909), Qred (1/750, Thermo Fisher Scientific; Q10363), Avidin/Biotin blocking kit (Vector; SP-2001), and EcoMount (Biocare Medical; EM897L). Samples were scanned and unmixed using Polariz multispectral slide scanner (Vectra) and analyzed using Nuance.

Mouse bone metastasis samples were fixed by 4% wt/vol paraformaldehyde for 30 min, incubated with 30% wt/vol sucrose overnight, and embedded in optimal cutting temperature compound at −20°C. Then bone was shaved to expose bone marrow, washed in PBS, and blocked with 5% wt/vol BSA and anti-mouse CD16/32 antibody (BD Biosciences; 1:200). Primary and secondary antibodies were stained in the dark. The bone was imaged by spinning-disk confocal microscopy. 3D images were generated via Volocity. FACS antibodies as described below and anti-Iba1 (WDR2342) were used for staining. Tartrate-resistant acid phosphatase (TRAP) staining was performed for osteoclast quantification. An Acid Phosphatase, Leukocyte (TRAP) Kit (Sigma-Aldrich) was used based on manufacturer’s guidelines. Imaging was performed with Axioskop 2 (Zeiss).

MetBo2 is an in house–selected bone homing clone of the mouse breast cancer cell line Met-1, which was derived from polyoma middle T oncogene tumor as previously described (Qian et al., 2009). Human breast cancer cell line 1833 (bone metastatic clone of triple-negative breast cancer cell line MDA-MB-231) was a gift from Prof. Joan Massagué (Memorial Sloan Kettering Cancer Center, New York, NY). Cells were cultured in DMEM (Gibco) supplemented with 10% vol/vol heat-inactivated FBS (Gibco) and 1% vol/vol penicillin-streptomycin (Gibco). Cells were maintained at 37°C with 5% vol/vol CO₂ in an incubator. All cells were negative for mycoplasma (Applied Biological Materials Inc.). Short tandem repeat tests were performed for human cells to make sure of their authenticity.

Bone metastasis was generated through intracardiac injection of 10⁵ tumor cells into 4-wk-old female mice of strains described above. Bioluminescent signal was recorded twice a week using the Photon IMAGER Optima system (Biospace) or IVIS spectrum (PerkinElmer). The region of interest was quantified by Photon IMAGER software or IVIS Living Imaging v4.3.1. Macrophage depletion was performed by i.v. injection of liposome-encapsulated clodronate (1 mg/mouse; Liposoma) twice a week. BLZ945 was given as daily gavage (200 mg/kg body weight; MedChemExpress). CCL2-neutralizing antibody or the control antibody (20 mg/kg body weight; Ortho BiotechOncology) was administered twice a week. Osteoclast depletion was performed by i.p. injections of free clodronate (dichloromethylenediphosphonic acid disodium salt; Sigma-Aldrich; 1 mg/mouse) on day 0, day 1, and then twice a week after tumor was detected.

For adoptive transfer, 1 × 10⁶ monocytes (CD45⁺CD11b⁺CSF1R⁺Ly6C⁺) from the bone of WT, Ccr2^−/−, or Il4ra^−/− were sorted and injected via intracaudal injection to mice with detectable bone metastasis. Mice were imaged right before monocyte transfer and on day 1, 3, 7, and 10 for quantification of tumor growth. For lineage tracking, 1 × 10⁶ GFP⁺ monocytes (from FVB-eGFP or Csf1r-EGFP mice) were injected i.v. into mice with late-stage bone metastasis. Bone and blood were collected 48 h after injection.

For mice with bone metastasis, bone was separated into tumor and peritumor areas based on the bioluminescent signal. Samples were then digested in DMEM supplemented with Librease TL, Librease DL, DNase I, and hyaluronidase (1:100 each) on a shaker at 37°C for 30 min. Blood was drawn by cardiac puncture with heparin (Sigma-Aldrich). Spleen was mashed using the plunger from an insulin syringe into a 40-μM cell strainer (Falcon). Red blood cell lysis buffer (eBioscience) was used to remove red blood cells. Cells were blocked with anti-mouse CD16/32 antibody (BD Biosciences; 1:500) before staining. Surface markers were stained for 30 min on ice and washed with PBS. Cells were sorted on Aria II or Fusion or analyzed on Fortessa (BD Bioscience). Antibodies used were as follows: CD11b (M1/70), Ly6G (1A8), Ly6C (HK1.4), F4/80 (BM8), CCR2 (FAB5538A), CD45 (30-F11), CD64(X54-5/7.1), CD86 (GL-1), CD115 (AFS98), CD169 (SER-4), CD206 (C068C2), MHCII (M5/114.15.2), MSR-1 (FAB1797P), mouse IgG1 (MOPC-21), rat IgG2a (RTK2758), hamster IgG (HTK888), B220 (RA3-6B2), CD3 (17A2), CD4 (RM4-5), CD8a (53–6.7), CD19 (6D5), and NK1.1 (PK136).

Tissue preparation and antibodies were the same as those described in FACS analysis. Inflammatory monocytes (Ly6C⁺) were sorted from bones bearing metBo2 or healthy controls. Total RNA was extracted from these sorted macrophages (RNeasy Mini kit; Qiagen), and its quality was determined using Pico Chip with a 2100 Bioanalyzer (Agilent Technologies). Samples where then submitted to the Albert Einstein College of Medicine microarray facility for labeling and hybridization on Affymetrix MoGene 2.0 ST chips. Differential gene expression analysis was performed using R package limma (Ritchie et al., 2015). Genes with fold-change more than ±1.5 and P < 0.05 were considered differentially expressed. Original data were deposited in the Gene Expression Omnibus (accession no. GSE152795).

Immune cell populations and tumor cells were sorted from MetBo2 bone metastasis samples. RNA was extracted by RNeasy Plus Micro Kit (Qiagen). Primers were Il4ra forward, 5′-TGT​GCC​AAA​CGT​CCT​CAC​AGC​A-3′, and reverse, 5′-TCC​AGG​CAT​CGA​AAA​GCC​CGA​A-3′.

Bioinformatics analysis was performed with R and Bioconductor. Gene annotation was performed using the Database for Annotation, Visualization, and Integrated Discovery. For correlation analysis, three individual datasets with breast cancer bone metastasis samples (GSE14020 and GSE54323) were used. IL4 target score was calculated using differentially regulated genes in IL4-treated macrophages reported by Gupta et al. (2018) and single-sample gene set enrichment analysis (Barbie et al., 2009) and the gene set variation analysis package (GSVA; Hänzelmann et al., 2013). Pearson correlation analysis was used to analyze the correlation between IL4 target scores and IL4R and CD204 expression.

Statistical analysis was performed with GraphPad Prism (v7). Two-tailed Student’s t test and ANOVA followed by multiple comparison was used to calculate statistical significance. The difference was considered significant with a P value <0.05.

Fig. S1 indicates that macrophage depletion assays specifically ablate macrophages but do not affect other bone marrow immune cells, and neither assay affects the growth of abdominal metastasis. Fig. S2 demonstrates that L-Clod depletes both bone marrow macrophages and osteoclasts, showing a better inhibition of bone metastasis growth than osteoclast-specific depletion. Fig. S3 shows that CD169⁺ bone-resident macrophages do not play a predominant role in breast cancer bone metastasis outgrowth. Fig. S4 demonstrates that the deficiency of Ccr2 leads to a significant reduction of IMs in blood but does not affect monocyte populations in bone marrow. Fig. S5 indicates that Il4ra deficiency specifically affects CD204⁺ BoMAMs in the 1833 bone metastasis model. Also, IL-4 is produced by T cells, B cells, and tumor cells in bone metastasis. Table S1 shows clinical information for breast cancer bone patients.

This work was supported by Cancer Research UK Career Development Fellowship C49791/A17367, European Research Council starting grant 716379to B.-Z. Qian; National Natural Science Foundation of China grant 81872150; National Cancer Institute cancer grants P30CA013330 and RO1CA131270 and Wellcome Trust grant 101067/Z/13/Z to J.W. Pollard; and an Edinburgh Global Research scholarship to R.-Y. Ma. This work was undertaken in the Medical Research Council Centre for Reproductive Health, which is funded by Medical Research Council Centre grant MR/N022556/1.

Author contributions: B.-Z. Qian and R.-Y. Ma designed the experiments and wrote the manuscript. R.-Y. Ma, H. Zhang, X.-F. Li, and A.D. Lam performed animal experiments. C.-B. Zhang, G. Tagliavini, S. Prost, and Z. Wang helped with tissue staining. C. Selli and A.H. Sims contributed to the bioinformatics analysis. Z. Ye, H.-Y. Hu, and T. Ying provided human samples or materials used in this study. J.W. Pollard and B.-Z. Qian supervised this project. All authors reviewed and approved the manuscript.