Macrophages are one of the most abundant leukocyte populations infiltrating tumor tissues and can exhibit both tumoricidal and tumor-promoting activities. In 1989, we reported the purification of monocyte chemoattractant protein-1 (MCP-1) from culture supernatants of mitogen-activated peripheral blood mononuclear cells and tumor cells. MCP-1 is a potent monocyte-attracting chemokine, identical to the previously described lymphocyte-derived chemotactic factor or tumor-derived chemotactic factor, and greatly contributes to the recruitment of blood monocytes into sites of inflammatory responses and tumors. Because in vitro-cultured tumor cells often produce significant amounts of MCP-1, tumor cells are considered to be the main source of MCP-1. However, various non-tumor cells in the tumor stroma also produce MCP-1 in response to stimuli. Studies performed in vitro and in vivo have provided evidence that MCP-1 production in tumors is a consequence of complex interactions between tumor cells and non-tumor cells and that both tumor cells and non-tumor cells contribute to the production of MCP-1. Although MCP-1 production was once considered to be a part of host defense against tumors, it is now believed to regulate the vicious cycle between tumor cells and macrophages that promotes the progression of tumors.
Macrophages play important roles in host defense by presenting antigens to lymphocytes or by participating in immune responses as effector cells. Macrophages that infiltrate sites of inflammation are derived from blood monocytes, which are attracted by locally produced chemotactic factors. A previous focus on the cellular motility and accumulation of nonsensitized inflammatory cells in delayed-type hypersensitivity (DTH) has led to the first description of a lymphokine-migration inhibitory factor.1,2 This was followed by a series of studies in the early 1970s reporting that chemotactic activity for monocytes (lymphocyte-derived chemotactic factor; LDCF) was produced and released by antigen-stimulated sensitized lymphocytes or mitogen-stimulated non-sensitized lymphocytes.3 However, the molecules responsible for these activities remained unidentified.
Infiltration of leukocytes into cancer tissues could also be a result of a host immune reaction against a tumor-specific antigen. Many laboratories explored this possibility both in vivo and in vitro. In a guinea pig transplantable tumor model, antigenically unrelated tumor cells were rejected at skin sites of DTH reactions induced by another tumor cell line.4 This suggested that the immune response to antigenically unrelated tumor cells was nonspecific and mediated by activated macrophages at the skin site. It was also shown that macrophages had the capacity to kill tumor cells in vitro if they were properly activated.5,6 Activated macrophages were therefore assigned a critical role in the destruction of tumors by the host. On the other hand, there was evidence suggesting that tumor-associated macrophages (TAMs) might stimulate tumor growth or connective tissue development.7,8,9 Again, neither the role of TAMs nor molecules attracting TAMs into tumors had been clarified.10,11
In 1989, we and others reported the purification of monocyte chemoattractant protein-1 (MCP-1; also termed by others as monocyte chemotactic and activating factor; MCAF) from culture supernatants of tumor cell lines and activated peripheral blood mononuclear cells (PBMCs).12,13,14 Isolation of MCP-1 greatly contributed not only to inflammation but also to cancer research that continues today, and MCP-1 has become a molecular target for treating many diseases. In this review, the biological activity of MCP-1 is first described, and then, the complex interactions between tumor cells and host cells that result in MCP-1 production in tumor microenvironments are discussed.
MCP-1 is identical to ldcf, tumor-derived chemotactic factor and the product of the gene JE
As a possible mechanism of monocyte/macrophage infiltration into tumors, Meltzer et al. reported the production of MCF by five murine sarcoma cell lines. The molecular weight of tumor cell-derived monocyte/macrophage chemotactic factor (MCF) was in the range of 15 000 on a Sephadex G-100 column, whereas murine LDCF was eluted in the molecular weight region of 40 000, suggesting that tumor cell-derived MCF was different from LDCF.15 Bottazzi et al. followed this report by demonstrating the production of monocyte/macrophage chemotactic activity (MCA) by human and murine tumor cell lines (termed tumor-derived chemotactic factor). Peak MCA was eluted from a Sephadex G-75 column in the molecular weight range of 12 000. Importantly, there was a significant correlation between the amount of MCA and macrophage content in tumors. Similar to tumor cells, human and murine embryo fibroblasts also released MCA into the culture supernatants,9,16 suggesting that tumor cells and non-tumor cells produce the same MCF.
Our main interest was to identify LDCF, which is considered to be the mediator responsible for the infiltration of monocytes into sites of DTH reactions. After purification of the first chemokine, interleukin-8 (IL-8)/CXCL8, from the culture supernatants of lipopolysaccharide (LPS)-activated human PBMCs,17 we immediately began to search for MCA in the same culture supernatants and successfully purified a second chemokine, MCP-1, from the culture supernatants of not only mitogen-activated human PBMC but also malignant human glioma cells.12,13 MCF produced by both PBMC and malignant glioma cells had apparent molecular masses of 15 and 13 kDa, respectively, according to SDS-polyacrylamide gel electrophoresis; this difference is likely due to different degrees of N-glycosylation.18 Thus, we demonstrated, for the first time, that non-tumor host cells and tumor cells produce an identical molecule with MCA.
Matsushima et al. 14 simultaneously reported the purification of MCP-1/MCAF from the culture supernatants of activated THP-1 human monocytic leukemia cells. Van Damme et al. 19 also purified MCP-1 from the culture supernatants of double-stranded RNA-activated MG-63 osteosarcoma cells and LPS-activated human monocytes. When the amino-acid sequence of human MCP-1 was compared with those in a database, we noted that the sequence of MCP-1 had significant similarity to the protein encoded by the mouse JE gene that was induced in fibroblasts by platelet-derived growth factor.20 Subsequently, MCP-1 was found to be identical to the product of the human JE gene.18 Finally, Bottazzi et al. 21 confirmed that the MCA they had previously detected in the culture supernatants of tumor cell lines was indeed due to the production of MCP-1 by the cells. Identification of MCP-1, along with IL-8, was the beginning of the ‘chemokine era’.
Biological role of MCP-1
The main activity of MCP-1 is to attract blood monocytes. In an in vitro chemotaxis assay, MCP-1 attracts human blood monocytes at its optimal concentration of 10−9 M, and ~30–40% of input cells respond and migrate.12,13 The magnitude of the response of human monocytes to MCP-1 is similar to that of the bacterial peptide N-formyl-methionyl-leucyl-phenylalanine (FMLF). There is no significant chemotactic response by human neutrophils, although they responded well to FMLF in the same assay.
The presence of the MCP-1 receptor on human monocytes was explored by a ligand-binding assay using radioiodinated MCP-1. Freshly isolated human monocytes possess 1700±600 binding sites per cell with a Kd of 1.9±0.2 nM.22 Similarly, undifferentiated THP-1 cells possess 1175±387 binding sites per cell with a Kd of 1.53±0.35 nM.23 When monocytes were induced to differentiate into macrophages, the number of MCP-1-binding sites was significantly reduced and the cells poorly responded to MCP-1.22 The loss of MCP-1 binding was also detected after the differentiation of THP-1 cells induced by phorbol 12-myristate 13-acetate (PMA).23 The chemokine receptor CCR2 was later cloned as the receptor for MCP-1,24,25 and downregulation of CCR2 expression during the differentiation of monocytes into macrophages was verified by reverse transcription-PCR and flow cytometry using monocytes and THP-1 cells.26 Consistent with the results obtained using in vitro-cultured cells, peritoneal exudate macrophages obtained from animals responded poorly to MCP-1 in a chemotaxis assay.27 These findings support the notion that MCP-1 is a potent chemoattractant for circulating blood monocytes via binding to its receptor CCR2, but is not a potent chemoattractant for differentiated monocyte-derived macrophages.
In addition to its monocyte chemotactic activity, MCP-1/MCAF was shown to activate monocytes for tumor cell killing over a concentration range that is optimal for chemotaxis in vitro.14 Indeed, transplantation of Chinese hamster ovary (CHO) cells engineered to express MCP-1 into nude mice led to marked monocyte infiltration, and the transplanted cells failed to form a tumor.28 Furthermore, co-transplantation of MCP-1-producing CHO cells with MCP-1-non-producing CHO cells or HeLa cells also suppressed tumor formation by these cells. When rat tumor cells that were transfected to produce MCP-1 were transplanted into syngeneic rats, the number of macrophages expressing macrophage activation markers, such as Ia antigens, was increased and the rate of tumor growth was significantly reduced.29
There are two types of macrophages based on their polarization states—macrophages activated by LPS±interferon-γ (M1) or Th2-derived cytokines (M2), which are considered to be proinflammatory or anti-inflammatory, respectively.30,31 Similar types of macrophages can be generated in vitro by incubating monocytes in the presence of granulocyte-macrophage colony-stimulating factor (GM-CSF) or macrophage-colony-stimulating factor (M-CSF), respectively.32 As described above, monocytes that infiltrated tumors after injection of MCP-1-transfected tumor cells showed antitumor activity; therefore, they appeared to be in the M1 polarization state. On the other hand, TAMs that accumulated in response to MCP-1 produced in tumor tissues are generally in the M2 polarization state.33 MCP-1 was shown to promote M2 polarization of macrophages in vitro,31 whereas M2 polarization of human monocytes induced by the culture supernatants of human glioma cells was independent of MCP-1.34 Thus, it is debatable whether MCP-1 directly affects macrophage polarization.
To determine the activity of MCP-1 in vivo, we previously injected recombinant rat MCP-1 into rat skin.35 Intradermal injection of 1 μg of MCP-1 induced both intra- and extra-vascular accumulation of mononuclear cells, characterized by the expression of the blood monocyte marker ED1, 3 h after MCP-1 injection. Accumulation of mononuclear cells, most of which were TRPM-3+ and ED3+, characteristic of exudate macrophages, peaked at 12–24 h. These results indicated that MCP-1 also attracts blood monocytes in vivo. However, this MCP-1-induced accumulation of mononuclear cells was transient and did not result in tissue damage. There was also a small increase in the number of OX8+ (CD8+) T cells.
The in vivo function of MCP-1 was further examined in five MCP-1 transgenic mouse models (Table 1).36,37,38,39,40 In a model in which MCP-1 was expressed in the epidermis (model 1) or several organs at high levels (model 2), no monocyte infiltrates were observed in MCP-1-expressing organs. Instead, the numbers of dendritic cells and Langerhans cells were increased in model 1. By contrast, in models 3, 4 and 5, in which MCP-1 expression was targeted to a specific organ, such as the thymus or central nervous system (model 3), type II alveolar epithelial cells in the lung (model 4) or pancreatic islets (model 5), respectively, monocyte infiltration was detected at MCP-1-producing sites; however, there was no sign of monocyte activation or tissue damage associated with the activation of monocytes. Interestingly, additional phenotypes were observed when mice were given exogenous stimuli. Contact hypersensitivity reactions to 2,4-dinitro-1-fluorobenzene were enhanced in the skin (model 1), and inflammatory responses in the lung upon intravenous (i.v.) injection of LPS or yeast wall glucan were also enhanced (model 4). When model 5 mice were crossed with model 2 mice, the resulting mice showed no monocyte infiltration, indicating that high systemic MCP-1 levels could prevent blood monocytes from responding to locally produced MCP-1. From these observations, it can be concluded that blood monocytes infiltrate tissues in response to local production, perhaps a low level of MCP-1, and then differentiate into macrophages in loco. However, their fate, whether they become immune-activating (M1) or suppressing (M2) macrophages, is likely determined by other mediators present at the sites.
In addition to MCP-1, other chemokines also induce monocyte chemotaxis in vitro, including MCP-3/CCL7 MCP-4/CCL13 (MCP-5/CCL12 in mice), and macrophage inflammatory peptide (MIP)-1α/CCL3,41 suggesting that functionally redundant chemokines are present and the loss of MCP-1 might be compensated by other chemokines. However, the infiltration of monocytes in MCP-1-deficient mice in response to a peritoneal injection of thioglycollate, LPS or zymosan was significantly reduced, indicating a non-redundant role of MCP-1 in the infiltration of monocytes during inflammatory responses in vivo.42,43
MCP-1 is also implicated in the recruitment of other cell types. Expression of the MCP-1 receptor CCR2 was detected in various types of cells, such as T lymphocytes, B lymphocytes, natural killer cells, neutrophils,41 hematopoietic stem cells44 and fetal microchimeric cells.45 In unchallenged MCP-1 transgenic mice, minor infiltration of lymphocytes, including CD4+ T cells and CD8+ T cells (model 4 and 5), and B220+ B cells (model 5), was also noted; however, the role of MCP-1 in the recruitment of these cell types in diseases remains unclear.
MCP-1 was shown to possess other roles during tumor progression. Human endothelial cells from both the umbilical cord vein and dermal microvasculature expressed CCR2 and migrated in response to MCP-1 in vitro. In vivo, MCP-1 induced angiogenesis in the chick chorioallantoic membrane assay and rat aortic ring sprouting assay in a manner that was independent of inflammatory cells.46 Thus, MCP-1 can directly contribute to angiogenesis via CCR2, but neither MCP-1 −/− nor CCR2 −/− mice have an abnormal angiogenic phenotype.47
Certain types of tumor cells, such as prostate cancer 48 and ovarian cancer cells,49 have been shown to express CCR2. In prostate cancer cells, MCP-1 binds to CCR2 and directly stimulates their proliferation and migration, as well as protects them from autophagic death, greatly contributing to prostate tumorigenesis.48
Regulation of MCP-1 production in tumor microenvironments
Regulation of constitutive MCP-1 production by tumor cells
Nuclear factor of κB (NF-κB) is a transcription factor that regulates the transcription of many genes involved in inflammatory and immune responses.50,51 NF-κB is constitutively activated in most cancer cells,52 resulting in the expression of anti-apoptotic genes and prolonged cancer cell survival. It also regulates the genes involved in the proliferation, invasion and metastasis of cancer cells.52,53 Thus, activation of NF-κB is closely associated with all phases of cancer progression.
The promoter region of the human MCP-1 gene contains multiple cis-elements for different transcription factors (Figure 1). Two NF-κB sites located in the distal region are particularly important for induction of this gene in response to proinflammatory responses, such as LPS and tumor necrosis factor (TNF)-α.54,55,56 We evaluated the role of NF-κB in constitutive MCP-1 production by tumor cells using the U-105MG and U-373MG human glioma cell lines.58 As shown in Figure 2a, U-105MG cells produced an ~40-fold higher level of MCP-1 compared with U-373MG cells. An NF-κB inhibitor, caffeic acid phenethyl ester, markedly inhibited MCP-1 production by both cell lines (Figure 2b); therefore, constitutive MCP-1 production by these cells is dependent on constitutively active NF-κB. Because transcription of the IL-8 gene is also dependent on NF-κB,59 U-105MG cells may produce a higher level of IL-8 than U-373MG cells. In contrast to this hypothesis, U-105MG cells produced an ~10-fold lower level of IL-8 than U-373MG cells. It is well established that transcription of genes is regulated by multiple cis-elements and transcription factors; thus, constitutively active NF-κB is only one mechanism whereby U-105MG cells produce MCP-1 at a high level. Activation of the NF-κB pathway in human MDA-MB-231 breast cancer cells also substantially contributed to expression of MCP-1 mRNA via a cancer-associated membrane glycoprotein, dysadherin.60
Specificity protein 1 (Sp1) regulates transcription of several genes involved in inflammation and tumorigenesis, including vascular endothelial growth factor (VEGF), urokinase plasminogen activator (uPA), uPA receptor and epithelial growth factor receptor.61,62 Sp1 also regulates basal transcription of the MCP-1 gene by binding to a GC-box located in the proximal region of the 5′-untranslated region (Figure 1).54 Therefore, we evaluated the role of Sp1 in constitutive, high-level MCP-1 mRNA expression in U-105MG cells. As we speculated, treatment with mithramycin A, an inhibitor of Sp1, markedly reduced the level of MCP-1 mRNA expression in U-105MG cells (Figure 2c), suggesting that Sp1 overexpression may be another mechanism of the constitutive high-level MCP-1 production by U-105MG cells.
Epithelial to mesenchymal transition (EMT) is a biological process that leads to acquisition of invasiveness and subsequent metastasis of cancer cells. Several molecular processes, including activation of transcription factors, expression of specific cell surface molecules, expression and reorganization of cytoskeletal proteins, production of enzymes that degrade ECM proteins and changes in expression of particular microRNAs, are engaged to initiate and complete EMT.63 It was recently demonstrated that transcription factors, such as Twist 1 or Snail, which induce EMT, have the capacity to induce MCP-1 production in epithelial cells,64,65 suggesting that tumor cells that have undergone EMT may acquire the ability to constitutively produce MCP-1. The expression status of Twist 1 or Snail in U-105MG cells is currently unknown.
Another signaling molecule involved in the upregulation of MCP-1 expression is Kras.66 Kras mutations occur in one-third of human cancers, including malignant glioma.67,68 A recent study demonstrated that MCP-1 production was upregulated in tumor cells harboring activating Kras mutations.66 This may provide, at least partly, a direct mechanism by which TAMs infiltrate tumors with Kras mutations.
Regulation of MCP-1 production by the host–tumor cell interaction
In addition to constitutive MCP-1 production by tumor cells, MCP-1 production can be further upregulated in tumor cells by inflammatory mediators, such as IL-1, IL-6, TNF-α or transforming growth factor-β.69 In human ovarian cancer cells, autocrine production of TNF-α by cancer cells was shown to stimulate a constitutive network of cytokines, angiogenic factors and chemokines, including MCP-1.70 In a murine chronic colitis-associated carcinogenesis model induced by azoxymethane and dextran sulfate sodium, MCP-1 protein was detected by immunohistochemistry in mononuclear cells, particularly macrophages, infiltrating the lamina propria and submucosal regions and also in endothelial cells at the early phase; and in carcinoma cells at a later phase.71 Production of MCP-1 in this model was almost completely inhibited when hematopoietic cells could not produce TNF-α,72 strongly suggesting that tumors cells activated by macrophage-derived TNF-α are the primary source of MCP-1 in this chemically induced colon cancer model. Thus, TNF-α, among the several cytokines noted above, whether it is produced by stromal cells or tumor cells, is a common molecule that upregulates production of MCP-1 by tumor cells in the tumor microenvironment.
The host–tumor cell interaction also leads to MCP-1 production by stromal cells. Osteoclastogenesis and bone resorption are two independent steps that lead to the development of skeletal metastases and are mutually essential for the establishment of prostate cancer in the bone microenvironment.73 Prostate tumor cells release parathyroid hormone-related protein, which stimulates MCP-1 expression by osteoblasts, and this osteoblast-derived MCP-1 causes increased osteoclastic bone resorption and binds to its receptor on prostate tumor cells, stimulating proangiogenic factor VEGF-A release from tumor cells.74 Thus, MCP-1 derived from stromal cells and osteoblasts, in this case, plays a critical role in cancer development.
The potential roles of MCP-1 in the development of breast cancer have been extensively studied, and significant associations between MCP-1 production and the progression of breast cancer have been found.75,76,77 In a MDA-MB-231 human breast cancer transplantation model, MCP-1 was detected in macrophages, fibroblasts and endothelial cells, and MCP-1 staining was most evident in macrophages. In 128 human breast cancer tissues examined, MCP-1 was detected in both tumor cells and stromal cells, mainly CD68-positive monocytic cells (likely macrophages). Furthermore, macrophage infiltration was significantly correlated with MCP-1 staining in stromal cells, but not in tumor cells.78 These results suggest that MCP-1 produced by stromal cells, but not by tumor cells, plays a critical role in the recruitment of macrophages into breast cancer tissues.
Regulation of MCP-1 production in murine transplantable tumor models
Lewis lung carcinoma model
Lewis lung carcinoma (LLC or LLC1) is a non-small-cell lung cancer (NSCLC) cell line adapted to culture from a lung carcinoma that arose in a C57BL mouse.79 LLC cells are known to produce a significant amount of MCP-1 in vitro.80,81 As described above, it was recently demonstrated that LLC cells harbor the activating Kras G12C mutation, which upregulates MCP-1 production, potentially linking Kras mutations and cancer-associated inflammation.66 MCP-1 production by these cells can be further increased in response to activation by the TLR4 ligand LPS or TNF-α.80,82 Intrapleural injection of LLC cells expressing MCP-1-specific short hairpin (sh)RNA resulted in a reduced level of pleural fluid accumulation,82 and neutralization of MCP-1 by an antibody (Ab) reduced the growth of subcutaneously injected LLC cells.83 These in vivo studies demonstrated the critical role of MCP-1 in the progression of LLC and suggests its role in human NSCLC.
We examined the mechanisms of MCP-1 production in LLC tumors by injecting LLC cells into the flank of WT or MCP-1 −/− mice.84 The absence of MCP-1 in host stromal cells (in MCP-1 −/− mice) did not affect the level of MCP-1 detected in the tumors or sera of tumor-bearing mice, indicating that tumor cells are the main MCP-1-producing cells. The high-level MCP-1 production detected in LLC tumors was not due to the constitutive MCP-1 mRNA expression in tumor cells but to the activation of tumor cells in the microenvironment because LLC cells isolated from tumors grown in either wild-type (WT) or MCP-1 −/− mice produced a level of MCP-1 in vitro similar to that found in in vitro-cultured cells.
We next examined whether the interaction between LLC cells and host cells, particularly macrophages, upregulates MCP-1 production by LLC cells. Co-culture with WT mouse macrophages markedly increased the level of MCP-1 mRNA expression by LLC cells, whereas co-culture with macrophages from TNF-α −/− mice or MyD88 −/− mice did not. Neutralization of TNF-α in the co-culture almost completely blocked the effect of WT macrophages to elevate MCP-1 production by LLC cells. Consistent with the in vitro results, expression of MCP-1 mRNA in LLC tumors that grew in TNF-α −/− mice was significantly reduced, and the serum MCP-1 concentrations were lower in tumor-bearing TNF-α −/− mice. Furthermore, the size and volume of tumors in TNF-α −/− mice were significantly smaller than those in WT mice. These results indicate that TNF-α is a critical macrophage-produced mediator that elevates MCP-1 production by tumor cells in vivo. Cordero et al. 85 reported that eiger, the sole member of the TNF superfamily in Drosophila expressed by tumor-associated hemocytes (leukocytes in Drosophila), was necessary and sufficient to trigger TNF-α signaling and express dMMP1 in tumor cells. The chemokine/chemokine receptor system does not appear to be present in the fly.86
It was previously demonstrated that the extracellular matrix protein versican released by LLC cells induces TNF-α production in myeloid cells through activation of TLR2.87 MyD88 is a critical signaling molecule located downstream of TLR2, and we found that macrophages from MyD88 −/− mice failed to elevate MCP-1 mRNA expression by LLC cells in co-culture, leading to the hypothesis that macrophage production of TNF-α may be due to activation of TLR2. However, macrophages from TLR2 −/− mice were as efficient as those from WT mice and elevated MCP-1 expression by LLC cells. Macrophages from TLR4 −/− or IL-1R1 −/− mice or TLR9 −/− mice also elevated MCP-1 mRNA expression by LLC cells. Thus, LLC cells appear to elevate MCP-1 production in macrophages by a mechanism independent of TLR2, TLR4, TLR9 or IL-1R1, involving MyD88. Alternatively, LLC cells may activate macrophages via more than one TLR. Taken together, our results indicate that the interaction of LLC cells with macrophages results in the production of TNF-α by macrophages, and this TNF-α subsequently upregulates MCP-1 production by tumor cells in LLC tumors (Figure 3).84 Additional studies are required to determine the exact mechanism of macrophage activation by LLC cells.
4T1 mouse breast cancer model
As noted above, a study of human breast cancer tissues suggested an important role for host cell-derived MCP-1 in the infiltration of macrophages into tumors.78 4T1 breast cancer cells were established from a spontaneous mammary tumor of Balb/cC3H mice. After orthotopical injection into the mammary pads of Balb/c mice, 4T1 cells form tumors at the injected site and spontaneously metastasize to distant tissues, such as the lung, liver and bone, thus providing a clinically relevant model to characterize the mechanisms that are important for tumor growth and metastasis.88 Furthermore, unlike LLC cells, 4T1 cells are polyclonal and contain mixed populations, providing an experimental model of tumor heterogeneity.89
We therefore analyzed the biological role of MCP-1 in the 4T1 breast cancer model (Figures 4).80 4T1 cells constitutively produced MCP-1 in vitro, and its production was significantly increased in response to activation with TNF-α or LPS; however, these cells produced a markedly lower level of MCP-1 than LLC cells. After intra-mammary injection of 4T1 cells in WT or MCP-1 −/− mice, the tumor size at the injected sites was increased at a similar rate in both strains, and there was no difference in their size or weight at 4 weeks. However, the number of metastatic tumors in the lungs of MCP-1 −/− mice was significant smaller than that in the lungs of WT mice, and MCP-1 −/− mice survived significantly longer than WT mice. Host cells in the tumor stroma were the primary source of MCP-1 in this model for the following reasons: (1) MCP-1 mRNA was easily detected in tumors of WT mice, but was hardly detectable in tumors of MCP-1 −/− mice; (2) the serum MCP-1 levels were increased in tumor-bearing WT mice, but not in MCP-1 −/− mice; (3) the MCP-1 mRNA level in tumors of MCP-1 −/− mice was comparable to that expressed by 4T1 cells cultured in vitro; and (4) the MCP-1 mRNA level in tumors of WT mice was clearly higher than that expressed by 4T1 cells activated in vitro with LPS or TNF-α. Although MCP-1 from either hematopoietic or non-hematopoietic cells in the tumor stroma promoted lung metastasis of 4T1 cells, MCP-1 from hematopoietic cells, such as macrophages, was sufficient. MCP-1 promotes lung metastasis of breast cancer cells by recruiting pro-angiogenic monocytes.90
We recently found that the cell-free culture supernatants of 4T1 cells (4T1-sup) directly activate mouse inflammatory macrophages for MCP-1 production (Figure 4).91 4T1-sup also elevates the expression of other MCP genes, such as the MCP-3/CCL7 and MCP-5/CCL12 genes, in macrophages. However, unlike LPS, which highly upregulates the expression of neutrophil-attracting chemokine genes, such as the KC/CXCL1 or MIP-2/CXCL2, 4T1-sup only modestly upregulates the expression of these chemokine genes. A 4T1 cell product with a size of ~20–30 kDa was responsible for the upregulation of MCP-1 mRNA expression by macrophages. Although neutralization of TNF-α in the culture did not affect MCP-1 production by macrophages, neutralization of GM-CSF almost completely abolished the MCP-1 mRNA expression and MCP-1 production induced by 4T1-sup. Furthermore, activation with recombinant GM-CSF markedly increased MCP-1 production by macrophages at a concentration as low as 1 ng/ml. By contrast, neutralization of M-CSF had no effect, and recombinant M-CSF exhibited only modest activity to upregulate MCP-1 expression at 100 ng/ml. Expression of GM-CSF mRNA was detected in 4T1 cells in vitro and 4T1 tumors in vivo. These results indicate that 4T1 cells can directly increase MCP-1 production by macrophages by producing and releasing GM-CSF (Figure 2). An important role of GM-CSF in the progression of human breast cancer was also reported by others.92 However, in our study, neutralization of GM-CSF did not reduce the level of MCP-1 production or lung metastasis in tumor-bearing WT mice, indicating that GM-CSF-independent mechanisms are also involved in increased MCP-1 production in the 4T1 tumor microenvironment.91
As described above, the mouse ortholog of the human MCP-1 gene was cloned from platelet-derived growth factor-activated fibroblasts as the gene JE,20 and human fibroblasts produce MCP-1.18,93 Therefore, fibroblasts are another likely source of MCP-1 in the 4T1 tumor microenvironment. It was previously shown that the products of breast cancer cells increased the expression of MCP-1 by human primary cancer-associated fibroblasts (CAF) in a STAT3-dependent manner.94 The nature of the cancer cell product that activated CAF remains unidentified.
In mammary tumors arising in PyMT mice in which the PyMT oncogene is expressed under the control of the MMTV-LTR promoter, primary tumors were heterogeneously stained for MCP-1, whereas metastatic tumors in the lung were homogeneously stained by immunohistochemistry. In a tumor transplantation model in which human MDA-MB-231 breast cancer cells were orthotopically transplanted into SCID mice, total blockade of MCP-1 using anti-human and anti-mouse MCP-1 antibodies inhibited spontaneous lung metastasis. When MDA-MB-231-derived, metastatic 4173 cells were intravenously injected, either an anti-human or anti-mouse MCP-1 Ab inhibited lung metastasis.90 These results suggest that the primary mammary tumors consist of cells with different degrees of MCP-1-producing capacity and that MCP-1-producing tumor cells may preferentially metastasize to the lung and MCP-1 produced by both tumor cells and stromal cells is important to promote lung metastasis. In fact we found that tumor cell lines established from mammary tumors that arose in C3(1)/SV40 T-antigen transgenic mice95 expressed different levels of MCP-1 mRNA (Figure 5a). Similarly, clones isolated from the original 4T1 cells showed heterogeneous MCP-1 mRNA expression (Figure 5b).
Regarding the role of tumor cell-derived MCP-1, treatment of Met-1 cells (a cell line generated from a PyMT-induced mammary tumor) with an anti-MCP-1 Ab shortly before i.v. injection reduced the number of lung metastases via inhibition of metastasis-associated macrophage recruitment.90 In our study, 4T1 cells isolated from lung metastases of WT mice expressed various levels of MCP-1, whereas cells isolated from lung metastases of MCP-1 −/− mice tended to express high levels of MCP-1,80 supporting the hypothesis that a population of tumor cells expressing higher levels of MCP-1 has a better chance to metastasize to the lung. It is important to note that lung or bone metastasis of MDA-MB-231 cells was inhibited by the blockade of MCP-1, but liver metastasis of Met-1 cells was not,90 indicating that MCP-1-independent mechanisms are involved in the metastasis of mammary tumor cells to the liver.
Similar to 4T1 tumors, MCP-1 mRNA expression was detectable in mouse B16 melanoma growing in WT mice, but was almost undetectable in tumors in MCP-1 −/− mice, indicating that stromal cells are the major MCP-1-producing cells in B16 tumors.84 A similar observation was made by others.96 However, the host-tumor cell interaction regulating MCP-1 production in B16 melanoma remains uncharacterized.
Tumors are wounds that do not heal;97 thus, there is no secret to the production of the chemokine MCP-1 in the tumor microenvironment. Because MCP-1 production in tumors facilitates the accumulation of immune-suppressive, tumor-promoting TAMs, it has become a molecular target for cancer treatment.98 The effects of the blockade of the MCP-1/CCR2 pathway on the progression of breast cancer or prostate cancer have been tested in clinical trials using anti-MCP-1 antibodies or CCR2 antagonists alone or in combination with other agents, and positive responses have been reported.99,100 On the other hand, it was recently reported that interruption of anti-MCP-1 Ab treatment markedly increased lung metastasis of breast cancer cells and subsequent death in mouse syngeneic cancer models. Further analysis of this model led to the finding that interruption of Ab treatment resulted in monocyte release from the bone marrow, an increase in cancer cell mobilization from the primary tumor, blood vessel formation and proliferation of metastatic cells in the lung in a manner dependent on IL-6 and VEGF. Inhibition of MCP-1 along with IL-6 prevented the promotion of lung metastases after interruption of anti-MCP-1 Ab therapy.101 Thus, blockade of the MCP-1/CCR2 pathway needs to be carried out with caution, and additional information is clearly needed for an effective cancer therapy using MCP-1 or CCR2 antagonists.
Over the past several years, we attempted to determine the mechanisms regulating the production of MCP-1in the tumor microenvironment using transplantable mouse tumor models. Our results, along with those by others, indicate that the production of MCP-1 is a consequence of the interaction between host cells and tumor cells and that both tumor cells and host stromal cells can contribute to the production of this chemokine. TNF-α, a well-known tumor promoter, is a frequently found mediator that upregulates production of MCP-1 by tumor cells, whereas tumor cell-derived mediators, such as GM-CSF, act on host cells, such as macrophages, to induce production of MCP-1 and other cytokines/chemokines. The serum MCP-1 level is increased in some cancer patients and may correlate with the progression of the disease, but its meaning remains unclear. It would be helpful if we could determine from the serum MCP-1 level the nature of the host–tumor cell interaction that occurs inside a tumor. The information that we have obtained to date is still limited, and additional studies are obviously needed. Our goal is to identify the mechanisms of the tumor cell–stromal cell interaction to better understand the tools tumor cells use for their progression and to provide a new means to save the lives of patients with cancer.
I am grateful to Drs Hideo Hayashi, Edward J Leonard, Kouji Matsushima and Joost J Oppenheim for their invaluable input during my studies on inflammation. I am also grateful to Drs Naoya Yuhki, Shuji Tanaka and Ettore Appella, and Ms Elizabeth A Robinson for their critical collaborations for the identification and cloning of MCP-1, and to Dr Ji Ming Wang and the members of the Laboratory of Molecular Immunoregulation, NCI, for their discussion and support.