Most malignant tumors evidence infiltration of many macrophages. In this study, we investigated an anti-inflammatory macrophage phenotype (M2) in clear cell renal cell carcinoma (RCC) using CD163 and CD204 as markers. Immunostaining showed a correlation between the number of CD163+ cells and age, sex, nuclear grade, and TNM classification. High infiltration of CD163+ cells was significantly associated with poor clinical prognosis in univariate analysis but not in multivariate analysis. We also carried out in vitro studies to examine cell–cell interactions between macrophages and cancer cells. Culture supernatants from RCC cell lines induced polarization of macrophages toward the M2 phenotype. Coculture of macrophages with cancer cells significantly induced activation of signal transducers and activators of transcription-3 (Stat3) in the cancer cells. Direct coculture of RCC cells with macrophages led to stronger activation of Stat3 in the cancer cells than did indirect coculture using Transwell chamber dishes. Because RCC cells expressed membrane-type macrophage colony-stimulating factor (mM-CSF) on the cell surface, we suggested that this mM-CSF plays an important role in direct cell–cell interactions. Stat3 activation in cancer cells that was induced by coculture with macrophages was suppressed by downregulation of the M-CSF receptor (M-CSFR) in macrophages and by an inhibitor of M-CSFR. In conclusion, investigation of CD163+ tumor-associated macrophages would be useful for assessment of the clinical prognosis of patients with ccRCC. Cell–cell interactions mediated by mM-CSF and M-CSFR binding could contribute to cancer cell activation. (Cancer Sci 2011; 102: 1424–1431)
Macrophages infiltrating tumor tissues, called tumor-associated macrophages (TAMs), have been shown by many studies to contribute to tumor development, neovascularization, and poor clinical prognosis.(1–3) The heterogeneity of macrophage phenotypes has been a focus of recent interest. γ-Interferon induces the classical activation pathway of macrophages, whereas anti-inflammatory mediators, such as interleukin (IL)-10, macrophage colony-stimulating factor (M-CSF), IL-4, and IL-13 evoke alternative activation pathways of macrophages.(4–7) These two types of activated macrophages are referred to as M1 and M2, respectively.(4–7) M2 macrophages produce considerable amounts of angiogenic and immunosuppressive molecules, so these macrophages are a target of anticancer therapy.(8) We previously showed that CD163 is specifically expressed in human macrophages and is a useful marker of M2 macrophages in human pathological specimens.(9) We found that the infiltration of M2 macrophages was closely related to poor clinical prognosis in patients with glioma, lymphoma, and intrahepatic cholangiocarcinoma.(10–12) Other investigators reported similar results for melanoma, leiomyosarcoma, and follicular lymphoma,(13–15) although dense infiltration of M2 macrophages was associated with better clinical prognosis in patients with colorectal cancer.(16)
Clear cell renal cell carcinoma (ccRCC) is the most common cancer of the kidney. Some studies have suggested the significance of TAMs in ccRCC. CD68+ TAMs contribute to angiogenesis in ccRCC by secreting vascular endothelial growth factor, and a high infiltration of TAMs into tumor tissues was correlated with recurrence in patients with RCC.(17,18) Expression of plasminogen activator inhibitor (PAI)-1 in RCC cells was associated with poor clinical prognosis, which suggested a relationship between TAMs and PAI-1 expression.(19) However, no reports have described macrophage heterogeneity in ccRCC. The aim of this study was therefore to evaluate the significance of M2-polarized TAMs in ccRCC by immunostaining tissue microarrays. In addition, we carried out in vitro studies to investigate the detailed mechanisms of cell–cell interactions between cancer cells and cultured macrophages. Many studies have focused on the significance of cell–cell interactions between cancer cells and macrophages for cancer development, but details of the mechanisms of direct interactions between these cells have not been clarified.(20–22) Because membrane-type M-CSF (mM-CSF) on cells is well known to strongly stimulate monocyte/macrophages and osteoclasts by direct cell–cell contact,(23–25) we hypothesized that mM-CSF on the surface of cancer cells might also have an important function in cell–cell interactions and cancer progression, and we therefore carried out in vitro coculture experiments to investigate this possibility.
Materials and Methods
Specimens. Tissue microarray sections with clinical information for approximately 42 ccRCC cases and 24 additional ccRCC cases were purchased from SuperBioChips Laboratories (Seoul, Korea) and Biomax Informatics (Planegg, Germany), respectively. Nuclear grade and T classification were according to the World Health Organization classification.(26)
Antibodies and immunohistochemistry. Monoclonal antibodies against human CD68, CD163, CD204, and M-CSF were purchased from Dako (Glostrup, Denmark), Novocastra Laboratories (Newcastle upon Tyne, UK), TransGenic (Kumamoto, Japan), and Novus Biologicals (Littleton, CO, USA), respectively. Immunostaining of tissue microarray sections was carried out as previously described.(12) Briefly, after samples were reacted with primary antibodies, they were incubated with HRP-labeled goat anti-mouse or anti-rabbit secondary antibodies (Nichirei, Tokyo, Japan). Reactions were visualized using the diaminobenzidine substrate system (Nichirei). Two pathologists (YK and KO, or YK and HH), who were blinded to information about the samples, evaluated the immunostaining findings, and resulting cell numbers from these evaluations were averaged. Double-immunostaining was carried out as described previously.(12)
Macrophage culture. Peripheral blood mononuclear cells were obtained from five healthy volunteer donors, from whom written informed consent was obtained. The cells were plated in 6-well plates (0.5 × 106 cells/well) for 1 h, and non-adherent cells were removed by gentle washing with PBS. The remaining monocytes were cultured with granulocyte macrophage-colony stimulating factor (GM-CSF) (5 ng/mL; Wako, Tokyo, Japan) for 5 days to induce macrophage differentiation.(9) Adherent cells were all positive for CD68 (data not shown). In this study, GM-CSF differentiated macrophages were used as immature macrophages because M1-type molecules such as CD80, CD86, and IL-12 were not detected in these cells without γ-interferon or lipopolysaccharide (LPS) stimulation. Tumor culture supernatants (TCSs) were added for 2 days to induce M2 polarization, as previously described.(12)
Cell culture. Three human RCC cell lines (Caki-1, ACHN, and 786-0) were purchased from ATCC (Manassas, VA, USA). MAMIYA cells were kindly provided by Prof. Kyogo Itoh of Kurume University (Kurume, Japan). Cells were cultured in DMEM supplemented with 10% FBS and penicillin/streptomycin (Wako).
Cell ELISA and cytokine ELISA. Monocytes were cultured in a 96-well plate with 5 ng/mL GM-CSF for 5 days to induce macrophage differentiation, as described previously.(9) Primary monocyte-derived macrophages were stimulated with 50% TCS or medium alone (control). To study the effects of cell–cell interaction with RCC cells, 1 × 104 MAMIYA cells were added and cultured for 6 days. Expression of CD163 by human macrophages was investigated with the aid of a cell ELISA, as described previously.(9) Interleukin-10 in supernatant was analyzed using an ELISA kit (eBioscience, San Diego, CA, USA).
Coculture experiment (indirect and direct cell–cell interaction). Coculture experiments were used to compare indirect and direct cell–cell interactions between tumor cells and macrophages. After macrophages were washed in PBS, they were co-incubated with RCC cells for 2 days (each at 2 × 105 cells per well) to determine the existence of direct cell–cell contact. Transwell chamber dishes (Nunc, Rochester, NY, USA) were used for analysis of indirect cell–cell interaction. Rabbit anti-M-CSF polyclonal antibody was used at the concentration of 5 μg/mL.(27) The M-CSFR inhibitor GW2580 (Calbiochem, Nottingham, UK) was used at the concentration of 20 nM. For preparation of paraffin-embedded cell block specimens, cultured cells were detached by means of a cell scraper (TPP, Trasadingen, Switzerland), after which they were fixed in 10% neutral buffered formalin. Cells were suspended in 1% sodium arginate and solidified by addition of 1 M calcium chloride. Viscous solutions containing RCC cells and macrophages were embedded in paraffin in a routine fashion. Sections were deparaffinized in xylene and rehydrated in a graded ethanol series. After reactions of anti-CD204 antibody (mouse monoclonal) and anti-pStat3 antibody (rabbit monoclonal; Cell Signaling Technology, Danvers, MA, USA), samples were incubated with HRP-labeled or alkaline phosphatase-labeled secondary antibody (Nichirei). Reactions were visualized using the diaminobenzidine substrate system and fast blue solution, as described previously.(12)
Flow cytometry. Expression of M-CSF on the cell surface membrane of RCC cells was determined by flow cytometry. Cells were detached using enzyme-free Cell Dissociation Buffer (Invitrogen, Carlsbad, CA, USA) and reacted with anti-M-CSF antibody (PeproTech, Rocky Hill, NJ, USA). Fluorescein isothiocyanate-labeled anti-rabbit antibody (Invitrogen) was used as a secondary antibody.
Small interfering RNA and quantitative real-time PCR. Macrophages were transfected with siRNA against human M-CSFR (sc-29220; Santa Cruz Biotechnology, Santa Cruz, CA, USA) using Lipofectamine RNAi MAX (Invitrogen). Control siRNA (sc-44230; Santa Cruz Biotechnology) was used as a negative control. The expression of M-CSFR mRNA was evaluated by quantitative PCR as described previously.(11) Primers were as follows: M-CSFR, 5′-TGGACACCTATGTGGAGATGAG-3′ (forward) and 5′-CTTGGCTGGAGAAGTGAAGC-3′ (reverse); GAPDH, 5′-GCACCGTCAAGGCTGAGAAC-3′ (forward) and 5′-TGGTGAAGACGCCAGTGGA-3′ (reverse).
Statistics. Statistical analysis was carried out with StatMate III (Atoms, Tokyo, Japan). The simultaneous relationship between multiple prognostic factors for survival was assessed using the Cox proportional hazards model with stepwise backwards reduction. All experimental in vitro data represented two or three independent experiments. Data are expressed as the means ± SD. A value of P < 0.05 was considered statistically significant.
CD163 and CD204 specific for TAMs, but CD68 detected in both TAMs and tumor cells. Because CD163 and CD204 are known to be specifically expressed by CD68+ macrophages,(12) we investigated CD68, CD163, and CD204 expression in ccRCC and non-cancer kidney tissues. Tumor-associated macrophages evidenced strong CD68 expression, but cancer cells in 17 of 42 ccRCC cases also expressed CD68 protein (strongly positive, 4 cases; weakly positive, 13 cases; negative, 25 cases). Immunostaining detected no CD163 or CD204 in cancer cells in all cases (Fig. 1A).
Four RCC cell lines were cultured and expression of these antigens was analyzed by means of immunostaining and Western blot analysis. Immunostaining indicated that two RCC cell lines were weakly positive for CD68 (Fig. 1B), and all cell lines were negative for CD163 and CD204 (data not shown). These antigens were not detected by Western blot analysis (Fig. 1C). In 25 cases in which CD68 was not found in cancer cells, the number of CD68+ cells was more similar to the number of CD204+ cells than the number of CD163+ cells (Fig. 1D). Although almost all CD68+ macrophages were positive for CD204, only a part of CD68+ macrophages were positive for CD163 (Fig. 1E). This result indicates that most CD68+ macrophages express CD204, that CD204 is a more useful marker of TAMs than CD68, and that CD163 is expressed by a subpopulation of macrophages.
Correlation of macrophage infiltration with clinicopathological factors. We then analyzed, by means of the Mann–Whitney U-test (two groups) and the Kruskal–Wallis test (three groups), the correlation between clinicopathological factors and the number of TAMs in 66 cases of ccRCC. The number of CD163+ cells per mm2 was associated with age, sex, nuclear grade, and TNM classification, and the number of CD204+ cells per mm2 was correlated with nuclear grade (Table 1).
Table 1. Clinicopathologic factors and number of macrophages in 66 cases of clear cell renal cell carcinoma
Correlation of macrophage infiltration with clinical prognosis. Infiltration of CD163+ macrophages was significantly correlated with poor overall survival in a univariate analysis (log–rank test, P = 0.006), but multivariate analysis indicated no significant correlation (Table 2, Fig. 2A). There was no correlation between CD204+ macrophages and overall survival (Fig. 2B). Older age, higher histological grade, and positive vascular invasion were also significantly associated with poor overall survival in the univariate analysis but not in the multivariate analysis (Table 2). Similar results were also obtained in other conditions in which the threshold was set as 200 (high, >201 cells/mm2; low, ≤200 cells/mm2) or 300 (high, >301 cells/mm2; low, ≤300 cells/mm2).
Table 2. Univariate and multivariate Cox regression analysis of potential prognostic factors for overall survival in patients with clear cell renal cell carcinoma (n = 66)
Cancer-derived factors induced macrophage differentiation into the M2 phenotype. Cancer-derived factors are believed to induce macrophages to polarize toward an M2 phenotype and promote cancer progression. We therefore evaluated the effects of TCSs from human RCC cell lines on macrophage differentiation in vitro. CD163 expression and IL-10 production served as markers of M2 differentiation, as previously described.(11,12) Exposure to TCSs of all RCC cell lines significantly upregulated macrophage CD163 expression and IL-10 production after LPS exposure (Fig. 3A,B). Direct cell–cell contact with MAMIYA cells by mixed coculture induced significantly elevated expression of CD163 in macrophages (Fig. 3C). Interleukin-10 secretion from macrophages were also strongly increased by direct cell–cell contact with MAMIYA cells (Fig. 3D). Because IL-10 concentrations in TCSs from RCC cell lines were 12.8 ± 1.0 pg/mL (MAMIYA), 18.4 ± 2.2 pg/mL (Caki-1), 17.4 ± 4.7 pg/mL (ACHN), and 15.0 ± 2.8 pg/mL (786-O), macrophages were suggested as the main source of IL-10 in the supernatant. These findings indicated that not only cancer-derived soluble factors but also cell–cell contact with cancer cells induced macrophage differentiation into the M2 phenotype.
Direct cell-cell interaction with macrophages induced significantly increased activation of signal transducers and activators of transcription-3 (Stat3) in cancer cells. Cell–cell interactions between macrophages and cancer cells are well known to cause cancer cell activation.(8,22,26) We therefore investigated the effect of interactions between macrophages and RCC cells by means of the in vitro coculture system, which reflects the in vivo direct contact of TAMs and RCC cells. Stat3 activation was evaluated as a marker of cancer cell activation. Coculture with macrophages significantly stimulated Stat3 activation in RCC cell lines (Fig. 4A). Tumor culture supernatant-stimulated M2 macrophages induced Stat3 activation to a significantly greater degree than did immature macrophages (Fig. 4B). These results indicated that cell–cell interaction between macrophages and cancer cells contributes to cancer cell activation.
Direct contact with RCC cells induced strong activation of macrophages. A coculture experiment with macrophages and MAMIYA cells was carried out using Transwell culture dishes to measure indirect cell–cell interaction. Indirect coculture with macrophages induced Stat3 activation in cancer cells (Fig. 4C). Direct cell–cell interaction, however, caused more significant Stat3 activation in cancer cells (Fig. 4C). Interleukin-10 production by macrophages after coculture with MAMIYA cells was compared in experiments using indirect or direct cell–cell interactions. Direct cell–cell interaction with MAMIYA cells induced greater IL-10 production than did indirect cell–cell interaction (Fig. 3D). This result indicates that direct contact with cancer cells induced strong macrophage activation.
Membrane-type M-CSF on cancer cells involved in macrophage activation by direct cell–cell contact. We next focused on the interaction of mM-CSF and M-CSF receptor (M-CSFR, known as c-fms or CD115) in direct cell–cell interaction of macrophages and cancer cells. Surfaces of cancer cells from all 42 ccRCC patients showed M-CSF expression (weak expression in 15 patients and strong expression in 27 patients) (Fig. 5A). The M-CSF expression had no correlation with the number of TAMs or clinical prognosis (data not shown). Flow cytometry analysis detected mM-CSF on cell surfaces in all RCC cell lines; data for MAMIYA cells are shown in Figure 5(B). Studies also showed that siRNA silenced the M-CSFR of macrophages (Fig. 5C). Coculture was then started. Downregulation of M-CSFR in macrophages significantly suppressed IL-10 production by direct cell–cell contact with MAMIYA cells (Fig. 5D). The M-CSFR inhibitor GW2580 (Fig. 5E) and neutralizing antibody against M-CSF (Fig. 5F) also inhibited IL-10 production. Activation of Stat3 in cancer cells was inhibited by silencing of M-CSFR in macrophages (Fig. 5G) and GW2580 (Fig. 5H). These results indicated that mM-CSF on cancer cells plays an important role in cell–cell interaction by inducing strong macrophage activation through binding of mM-CSF with M-CSFR.
In this study, we investigated three macrophage-specific antigens, CD68, CD163, and CD204, in ccRCC tissues. All populations of macrophages express CD68, which many studies use as a pan-macrophage marker.(12) Certain populations of human macrophages express CD163 and CD204, which are considered markers for the M2 phenotype.(4,12) However, some differences do exist between CD163+ and CD204+ cells. Almost all infiltrated macrophages in ccRCC express CD204 as well as CD68, which indicates that CD204 is suitable as a pan-macrophage marker rather than as a marker for the M2 phenotype in ccRCC. Although some studies using CD68 as a marker of TAM have been published,(17–19) our observation that not only macrophages but also cancer cells expressed CD68 indicates that CD68 is not a good marker of macrophages in ccRCC. Although CD68 is known to be expressed in epithelial cells and involved in macropinocytosis,(26,28) the function of CD68 in cancer cells has been unclear. In the present study, the number of CD163+ cells was significantly associated with poor clinical prognosis, whereas the number of CD204+ macrophages was not significantly related to poor clinical prognosis. These findings indicate that CD163 antigen might be a better marker of the M2 anti-inflammatory phenotype in ccRCC tissues.
Cancer-derived factors including IL-6, IL-10, transforming growth factor-β (TGF-β), and M-CSF are believed to induce macrophages to polarize toward an M2 phenotype,(29) and we therefore investigated the effects of TCSs from human RCC cell lines on macrophage differentiation in vitro. Exposure to TCSs significantly upregulated macrophage expression of CD163 and LPS-induced production of IL-10, which are molecules associated with the M2 phenotype. Direct coculture with cancer cells caused strong expression of CD163 and production of IL-10, and this indicates that M2 polarization is significantly induced by direct contact with cancer cells. Although we could not specify which cytokines were involved in M2 polarization in this study, unknown tumor-derived factors are believed to induce M2 macrophages and contribute to cancer-related immunosuppression by promoting the secretion of anti-inflammatory cytokines such as IL-10. Previous studies showed a clear correlation between M2 macrophages and regulatory T cells.(11,30–32) Dense M2 populations in tumors could create favourable immunosuppressive conditions for antitumor immunity.
Signal transducers and activators of transcription-3 is one of the main transcription factors that promote cancer cell survival, angiogenesis, and immunosuppression in the cancer microenvironment.(33) In patients with RCC, activation of Stat3 in cancer cells is correlated with poor clinical prognosis.(34)In vitro studies with RCC cell lines indicated that Stat3 inhibitors induced apoptosis and growth arrest of cancer cells.(35) Inhibition of Stat3 in murine RCC models suppressed cancer growth by inducing cancer cell apoptosis and by reducing the number of immunosuppressive cells such as myeloid-derived suppressor cells and regulatory T cells.(35) Treatment with sunitinib, a tyrosine kinase inhibitor, reversed immune suppression in patients with RCC by reducing the myeloid-derived suppressor cell population.(36) Because sunitinib inhibits Stat3 activation, reversal of immune suppression may be caused by Stat3 inhibition in the cancer microenvironment.(37) Our present study suggested that TAMs, especially M2 polarized TAMs, contributed to Stat3 activation in the cancer microenvironment by secreting several cytokines, which were strongly induced by direct cell–cell contact with cancer cells. Inhibition of cell–cell interaction between cancer cells and TAMs could thus be an effective target for anticancer therapy by causing reversal of immunosuppressive conditions in patients with RCC.
Cell–cell interactions between macrophages and cancer cells are well studied and believed to contribute to cancer development, invasion, and metastasis.(8,30,38) Previously, Ikemoto et al. showed that coculture of RCC cells with macrophages induced macrophage activation which caused cancer cell proliferation, and macrophage-derived cytokines, such as IL-6, tumor necrosis factor-α, and IL-1β, were considered to be involved in cancer cell proliferation.(39,40) However, details of the mechanisms of cell–cell contact have not been clarified. It is well known that mM-CSF induces strong activation of monocyte/macrophages.(23,24) In bone metabolism, for example, osteoclasts express mM-CSF along with other membrane ligands, such as RANK, and mM-CSF induces differentiation of monocytes into osteoclasts by direct binding to M-CSFR.(25) Zheng et al.(41) recently reported that direct cell–cell contact between macrophages and myeloma cells protected myeloma cells from chemotherapy drug-induced apoptosis, whereas cell–cell interaction without direct contact did not protect the myeloma cells. In the present study, we showed that direct cell–cell contact with macrophages strongly induced Stat3 activation in RCC cells and that mM-CSF–M-CSFR binding induced significant macrophage activation. The overexpression of M-CSFR in TAMs has been shown in some types of malignant tumor.(42–44) Patsialou et al.(45) showed that epidermal growth factor and TGF-β secreted by M-CSFR-activated macrophages induced cell–cell interactions between breast cancer cells and macrophages. Although these cytokines were not evaluated in this study, IL-1β, IL-6, IL-10, tumor necrosis factor-α, epidermal growth factor, and TGF-β are thought to be involved in the cell–cell interaction. Taken together, we suggest that direct cell–cell interaction induces macrophage differentiation into M2 phenotype and the secretion of these cytokines, which in turn activates the Stat3 signal in cancer cells and promote the cancer progression.
In conclusion, we found that CD163+ M2 TAMs contributed to poor clinical prognosis in patients with ccRCC and that TAMs induced Stat3 activation in cancer cells through direct cell–cell interactions. The number of CD163+ TAMs, as well as nuclear grade, vascular invasion, and TNM classification, could be useful as pathological prognostic factors predicting poor clinical prognosis, and CD163 might also aid evaluation of the M2 activation status of TAMs or immunosuppressive conditions in the cancer microenvironment. We also showed that the binding of mM-CSF on cancer cells to M-CSFR on TAMs contributes to Stat3 activation in cancer cells by direct cell–cell interaction. Blocking of M-CSFR signaling, and thereby reversing immunosuppressive conditions, may be a promising approach for anticancer therapy in patients with ccRCC.
We thank Ms Yui Hayashida, Ms Emi Kiyota, Mr Osamu Nakamura, and Mr Takenobu Nakagawa for their technical assistance. This study was supported in part by Grants-in-Aid for Scientific Research (B20390113, 21790388) from the Ministry of Education, Culture, Sports, Science, and Technology of Japan and by Okukubo Memorial Find for Medical Research at Kumamoto University.
None of the authors has any conflict of interest to declare.