Recent findings implicate macrophages and some of their secreted products, especially tumor necrosis factor (TNF), as tumor promoters. Inhibitors of these inflammatory components are currently regarded as potential therapeutic tools to block tumor progression. Here, we show that infiltrating macrophages represented a significant population of nonneoplastic cells within malignant gliomas, in which they were the exclusive producers of TNF. Contrary to the reported pro-oncogenic effects of TNF in other types of solid tumors, glioma-bearing mice deficient in TNF developed larger tumors and had reduced survival compared with their wild-type controls. Histologic examinations revealed that glioma volume was negatively correlated with the number of macrophages and small cavities called microcysts. Overall, our results support the concept that macrophages alter brain tumor development through a TNF-dependent process that culminates in the formation of microcysts. This raises the question of whether anti-inflammatory drugs, such as those commonly administrated to patients with brain cancer, could interfere with antitumor mechanisms.
Macrophages have long been recognized as a critical component of innate immunity against tumors. When appropriately stimulated, they can attack neoplastic cells through contact-dependent interactions as well as the secretion of cytotoxic and cytostatic humoral factors (1). Nonetheless, malignant cells can escape immune surveillance through mechanisms that are not fully understood, leading to tumor growth and metastasis. The fact that tumorigenesis occurs despite the presence of a large number of tumor-associated macrophages has led to the speculation that macrophages contribute to tumor development, presumably by secreting growth signals, angiogenic factors, matrix-degrading proteinases, and immunosuppressors (2–6). This possibility has gained substantial support from the observation that colony-stimulating factor 1, a key determinant of the monocytic lineage, promotes tumor progression by controlling the recruitment and function of tumor-associated macrophages (7–10). Yet, the precise mechanisms by which macrophages can facilitate tumor progression remain to be clearly shown.
The proinflammatory cytokine tumor necrosis factor (TNF) may be one of the ways through which macrophages exert dual functions in tumor biology. This glycoprotein was originally identified as a serum factor mediating the ability of endotoxin to induce hemorrhagic tumor necrosis in mice infected with Bacillus Calmette-Guérin (11). Although TNF can manifest powerful anticancer actions when administered locally at supraphysiologic doses (12), the reality is that it does not kill most types of cancer cells in which apoptosis is turned off by the transcription factor nuclear factor-κB (13). Conversely, studies revealed that TNF promotes tumor invasion and growth in many tissues. For example, it was shown that the number of pulmonary metastases in an experimental fibrosarcoma model increases after pretreatment with TNF (14). It was also shown that TNF-deficient mice develop 10 times fewer skin tumors than wild-type mice after administration of carcinogens (15, 16), and that liver tumorigenesis is markedly reduced in mice that lack TNF receptor-1, but not TNF receptor-2 (17). On the basis of these findings, drugs designed to neutralize TNF and clinically approved for the treatment of autoimmune disorders are currently regarded as potential tools to fight cancer (18, 19).
Thus far, the evidence suggests that anti-TNF therapy might be useful for reducing the growth of glioblastomas, the most frequent and malignant tumors affecting the central nervous system. Derived from glial cells, these tumors are characterized by their rapid growth and ability to diffusely infiltrate the parenchyma. A similarity between glioblastomas and other types of solid tumors can be found in the abundance of tumor-associated macrophages, which represent up to one third of cells found in glioma biopsies (20, 21). TNF is produced in gliomas by different cell types, including macrophages (22). Functional studies showed that TNF stimulates the expression of proangiogenic factors by glioma cells (23, 24), down-regulates the tumor suppressor gene PTEN (frequently altered in human glioblastomas; refs. 25, 26), and induces the production of the matrix metalloproteinase-9 (27), which seems essential for the invasiveness of glioma cells (28). However, all of these observations are derived from ex vivo studies and need to be confirmed in vivo.
Herein, we address the question of whether macrophage-derived TNF contributes to brain tumorigenesis by comparing the growth of GL261 glioma cells after orthotopic implantation into TNF-deficient and wild-type mice. Contrary to the evidence presented above, our results indicate that TNF reduces glioma growth and prolongs survival, at least in part, by enhancing the recruitment of phagocytes and the formation of microcysts, raising the possibility that anti-inflammatory drugs, such as those commonly used to manage glioma-associated edema, antagonize antitumor mechanisms.
Materials and Methods
Cell culture. GL261 glioma cells, originally induced by implantation of methylcholanthrene pellets in the brain of a B6 mouse (29), were kindly provided by Dr. Protul Shrikant (Roswell Park Cancer Institute, Buffalo, NY). Cells were cultured in DMEM (Wisent, Saint-Bruno, Quebec, Canada) supplemented with 10% heat-inactivated fetal bovine serum, 2 mmol/L l-glutamine, 110 mg/L sodium pyruvate, 100 U/mL penicillin, and 100 μg/mL streptomycin. Before implantation, cells were harvested by brief incubation at room temperature in 0.25% trypsin (Invitrogen, Carlsbad, CA), then washed and resuspended in Dulbecco's PBS (DPBS, Invitrogen) containing 1 g/L d-glucose.
Ex vivo proliferation and survival assays. GL261 cells were seeded at either 50,000 cells/well in six-well plates or at 500,000 cells/25 cm2 flask and incubated for 24 hours under the conditions described above. The medium was then replaced with fresh medium supplemented with different concentrations of recombinant mouse TNF (contained <1.0 endotoxin unit per microgram of cytokine; R&D Systems, Minneapolis, MN) and goat anti-TNF antibody (R&D Systems). Cells were collected by trypsinization 24 and 48 hours later. Immediately before collection, some cultures were pulse-labeled for 30 minutes with 10 μmol/L bromodeoxyuridine (BrdUrd; BD PharMingen, San Diego, CA). The total number of viable cells was evaluated using an Epic XL flow cytometer (Coulter, Miami, FL) by excluding dead cells by propidium iodide staining. The percentages of apoptotic cells (annexin V+, propidium iodide−) and necrotic cells (propidium iodide+) were estimated by flow cytometry using the Vybrant Apoptosis Assay kit 2 (Molecular Probes, Eugene, OR) according to the manufacturer's protocol. The proportions of cells in each phase of the cell cycle were determined by two methods: (a) nuclei suspensions were prepared by resuspending cells in 0.1% sodium citrate, 0.1% Triton X-100, and 50 μg/mL propidium iodide and analyzed by flow cytometry with Multicycle software (Phoenix Flow Systems, San Diego, CA); (b) after fixation, acid denaturation, permeabilization, and neutralization, cells were stained with a FITC-conjugated anti-BrdUrd antibody and counterstained with 7-amino actinomycin D using a BrdUrd detection kit (BD PharMingen) according to the supplied protocol, then assayed by flow cytometry. The data were analyzed by one-way ANOVA, followed by Tukey-Kramer honestly significant difference (HSD) multiple comparison test using JMP software (SAS Institute, Cary, NC).
Multiplex reverse transcription-PCR. Total RNA was extracted from GL261, BV2 (microglial), and N2A (neuronal) cells with TRIzol reagent (Invitrogen) and its integrity was controlled by agarose gel electrophoresis. The RNA samples were analyzed using a one-step reverse transcription-PCR (RT-PCR) kit (BD PharMingen) with the following primers: TNF, 5′-CCAGAACTCCAGGCGGTGCCTATGT-3′ and 5′-TACAACCCATCGGCTGGCACCACTA-3′; β-actin, 5′-GTGGGCCGCTCTAGGCACCAA-3′ and 5′-CTCTTTGATGTCACGCACGATTTC-3′. These primers were derived from different exons to distinguish between cDNA and genomic DNA amplification. The PCR products (TNF, 330 bp; β-actin, 540 bp) were separated by agarose gel electrophoresis, visualized by ethidium bromide staining, and sequenced to confirm identity.
Animals. TNF-deficient mice on a B6 × 129S background, their wild-type controls (B6 × 129S), and hemizygous transgenic mice (B6) expressing the green fluorescent protein (GFP) under the regulation of the chicken β-actin promoter were generated from breeders originally obtained from The Jackson Laboratory (Bar Harbor, ME). The genotypes of the TNF-deficient and wild-type mice were confirmed by PCR using DNA extracted from tail biopsies following the protocol provided by the supplier. C57BL/6 mice were purchased from Charles River (Montreal, Quebec, Canada) and adapted to standard laboratory conditions for 1 week before any manipulation. All experiments were done on males aged 2 to 3 months according to procedures approved by our institution's Animal Welfare Committee.
Generation of chimeric mice. C57BL/6 mice were exposed to 10 Gy ionizing radiation using a 137Cs source (Gammacell 40 Exactor; MDS Nordion, Kanata, Ontario, Canada). They were housed in autoclaved cages and treated with antibiotics (Sulfatrim; 1 mg/mL sulfamethoxazole and 0.2 mg/mL trimethoprine in drinking water) for 3 days before irradiation and 3 weeks after. Immediately after irradiation, the mice were injected via a tail vein with 200 μL of a suspension containing 5 × 106 bone marrow cells freshly collected from isogenic GFP mice. The cells were aseptically harvested by flushing femurs with DPBS containing 2% fetal bovine serum using a syringe with a 25-gauge needle. The samples were combined, filtered through 40 μm nylon mesh (Becton Dickinson, Cockeysville, MD), centrifuged, and resuspended in DPBS at a concentration of 2.5 × 107 viable nucleated cells per milliliter.
Orthotopic implantation of glioma cells. Mice were anesthetized and immobilized in a stereotaxic frame. A midline incision was made on the scalp, followed by a circular craniotomy over the right hemisphere, 1.5 mm lateral and 1 mm rostral from bregma. After removal of the dura mater, a 5 μL Hamilton syringe fitted with a 27-gauge beveled needle was advanced into the caudoputamen at a depth of 3.5 mm from the skull surface, and 2 μL of a suspension containing 5 × 104 viable GL261 cells were injected over 2 minutes using a UMPII micropump (World Precision Instruments, Saratoga, FL). After the injection, the syringe was left in place for 2 minutes before being withdrawn very slowly. The wound was sutured and coated with iodine.
Survival analysis. After tumor implantation, mice were monitored daily and killed when any of the following criteria was observed: >20% weight loss, paralysis, or lethargy. The survival time was measured from the day of tumor cell implantation to the day of euthanasia or death. The probability of survival was calculated using the Kaplan-Meier method and the data were compared using the log-rank and Wilcoxon statistical tests.
Tissue preparation. For all protocols except in situ hybridization, mice were anesthetized and transcardially perfused with 10 mL saline, followed by ice-cold 4% paraformaldehyde in phosphate buffer (pH 7.4) over 10 minutes. The brains were removed, postfixed for 4 hours at 4°C, and then cryoprotected overnight in 50 mmol/L potassium PBS supplemented with 20% sucrose. Series of sections through the tumors were cut at 40 μm using a freezing microtome, collected in cryoprotectant [30% ethylene glycol, 20% glycerol, 50 mmol/L sodium phosphate buffer (pH 7.4)], and stored at −20°C until histologic analysis. For in situ hybridization, the following modifications were applied: (a) the fixative was dissolved in borate buffer (pH 9.5) instead of phosphate buffer; (b) brains were postfixed for 48 hours and then cryoprotected overnight in the same fixative supplemented with 20% sucrose; and (c) tumors were cut at 30 μm.
Immunostaining. Immunohistochemistry and immunofluorescence staining were done as previously described (30) using the following primary antibodies: rabbit anti-Iba1 (1:1,000; Wako Chemicals, Richmond, VA), rat anti-CD31 (1:500; BD PharMingen), rat anti-GFP (1:1,000; Molecular Probes), rat anti-CD11b (1:500; BD PharMingen), hamster anti-CD11c (1:500; BD PharMingen), and rat anti-galectin-3 (1:500; American Type Culture Collection, Manassas, VA).
In situ hybridization. TNF mRNA was detected by radioisotopic in situ hybridization according to a previously described protocol (31). Briefly, sections were mounted onto gelatin and poly-l-lysine–coated slides, postfixed with 4% paraformaldehyde in borate buffer (pH 9.5) for 20 minutes, digested with 10 μg/mL proteinase K at 37°C for 25 minutes, acetylated with 0.25% acetic anhydride for 10 minutes, and dehydrated. Antisense and sense cRNA probes were transcribed from a 1.3 kb mouse TNF cDNA (Dr. Serge Rivest, Laval University, Québec, Quebec, Canada) in the presence of [35S]UTP and [35S]CTP (Perkin-Elmer, Foster City, CA), purified by phenol-chloroform extraction and ammonium acetate–ethanol precipitation, then applied to the slides at a concentration of 2 × 107 cpm/mL in hybridization buffer containing 50% formamide, 0.3 mol/L NaCl, 10 mmol/L Tris (pH 8.0), 1 mmol/L EDTA, 1× Denhardt's solution, 10% dextran sulfate, 2 μg/mL tRNA, and 10 mmol/L DTT. After incubation at 60°C for 14 hours, the slides were treated with 20 μg/mL RNase A for 30 minutes at 37°C, washed in a solution containing 15 mmol/L NaCl and 1.5 mmol/L sodium citrate for 30 minutes at 65°C, dehydrated, and exposed to a Biomax film (Kodak, Rochester, NY) for 22 hours. After film autoradiography, slides were defatted in xylene, dipped into Kodak NTB2 emulsion, exposed at 4°C for 17 days, developed in Kodak D19 developer for 3.5 minutes at 14°C to 16°C, fixed in Kodak rapid fixer for 5 minutes, counterstained with 0.25% thionin, dehydrated, and coverslipped with a mixture of distyrene, tricresyl phosphate, and xylene (DPX; Electron Microscopy Sciences, Fort Washington, PA).
Combined immunohistochemistry and in situ hybridization. Immunostaining for Iba1 and in situ hybridization for TNF mRNA were done sequentially on the same sections according to the methods described above with the following modifications: (a) tissue pretreatment with hydrogen peroxide was omitted; (b) serum and Triton X-100 were eliminated; (c) antibody solutions were supplemented with 2% heparin sulfate; (d) nickel-3,3′-diaminobenzidine (DAB) solution was replaced with DAB solution composed of 20.5 mg/mL diaminobenzidine, 2 mg/mL β-d(+)-glucose, and 1 μL/mL glucose oxidase in potassium PBS; and (e) thionin counterstaining was omitted.
Volumetric analysis. All quantitative histologic analyses were done by an observer who was blind to the treatment status of the material. Systematically sampled sections (every 10th section through the tumor) were mounted onto gelatin-coated slides and stained with 0.25% thionin. Tumor volume was estimated by the Cavalieri method using Stereo Investigator software (Microbrightfield, Colchester, VT) driving a motorized stage (Ludl, Hawthorne, NY) on a Nikon E800 microscope with a 2× Plan Apochromat objective (numerical aperture 0.1). A 200 μm2 point grid was overlaid on each section and the points that fell within the tumor were counted. Point counts were converted to volume estimates taking into account sampling frequency, magnification, grid size, and section thickness. The data were averaged and analyzed by Student's unpaired t test.
Cell counting. Systematically sampled sections (every 10th section through the tumor) were immunostained for Iba1. The density of labeled cells was estimated by the optical fractionator method using Stereo Investigator software. Tumor tissue, excluding cavities, was traced using a 2 × Plan Apochromat objective (numerical aperture 0.1) and sampled using a 100× Plan Apochromat oil objective (numerical aperture 1.4). The counting parameters were as follows: distance between counting frames, 200 × 200 μm; counting frame size, 50 × 50 μm; dissector height, 10 μm; guard zone thickness, ≥2 μm. Cells were counted only if their nuclei laid within the dissector area, did not intersect forbidden lines, and came into the focus as the optical plane moved through the height of the disector. The data were averaged and analyzed by Student's unpaired t test.
Cavity and blood vessel quantification. Systematically sampled sections (every 10th section through the tumor) were either stained with 0.25% thionin or immunolabeled for CD31. The sections were photographed using a 2× Plan Apochromat objective and a Retiga EX monochrome camera (QImaging, Burnaby, British Columbia, Canada). The photographs (8-bit, grayscale) were converted to Bitmap images with Photoshop (Adobe Systems, San Jose, CA) using a threshold level of 160 and the Batch command. The contour of each tumor profile was traced with the freehand selection tool of ImageJ software (NIH, Bethesda, MD) and the mean gray value of that area was recorded (black = 0, white = 255). The data were averaged, converted to percentages, and analyzed by Student's unpaired t test. Although this method does not provide an accurate estimate of the total area covered by cavities and blood vessels, it permits a meaningful intergroup comparison of the degree of cavitation and vascularization.
Confocal microscopy and cell phenotyping. Confocal images were acquired with a Fluoview confocal microscope (BX-61; Olympus Optical, Tokyo, Japan) and a 100× Uplan Apo oil objective (numerical aperture 1.35) by sequential scanning using a two-frame Kalman filter, z-separation of 0.25 μm and electronic zoom of 2. For each fluorochrome, the upper and lower thresholds were set using the range indicator to minimize data loss through saturation. The images were processed with Photoshop to enhance brightness and contrast. The proportions of cells displaying a given phenotype were evaluated by scoring the colocalization of cell markers using 3-μm-thick confocal images (10 per mouse) randomly and blindly sampled through the neoplastic tissue. A minimum of 350 cells was phenotyped per animal. The data were converted to percentages, averaged, and analyzed by Student's unpaired t test.
ELISA for tumor necrosis factor. Gliomas were grossly dissected 3 weeks after implantation and homogenized in ice-cold lysis buffer (10 μL/mg tissue) containing 10 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, and 1× protease inhibitor cocktail (P8340, Sigma). The homogenates were centrifuged at 55,000 rpm for 20 minutes at 4°C and stored in liquid nitrogen until analysis. Furthermore, conditioned culture media from BV2 and GL261 cells grown for 24 hours under the conditions described above was collected, centrifuged at 15,000 rpm for 10 minutes at 4°C, and stored in liquid nitrogen. The presence of TNF protein in all these samples was determined using an ELISA kit according to the manufacturer's instructions (MTA00, R&D Systems). Each assay was run in triplicate with known standards provided with the kit. Nonconditioned media and homogenates prepared from nonneoplastic brain regions served as controls. The minimum detectable dose of TNF was <5.1 pg/mL.
Recruitment of macrophages in GL261 gliomas. To confirm the assumption that GL261 gliomas are infiltrated by numerous macrophages derived from the bloodstream, as are other types of solid tumors (32), we initially implanted GL261 cells into the brains of lethally irradiated mice transplanted with bone marrow cells expressing GFP. Three weeks later, the animals were killed and brain sections were examined for the presence of bone marrow–derived cells by multiple fluorescence staining and confocal microscopy. Infiltrating GFP+ cells were found in abundance within the gliomas (Fig. 1A-C), but were relatively rare in the adjacent nonneoplastic tissue. Quantification revealed that 95.2 ± 5.4% (mean ± SE) of GFP+ cells in the tumors expressed the monocytic marker Iba1, a calcium-binding protein involved in membrane ruffling and phagocytosis (33, 34), and that 98.4 ± 0.8% of Iba1+ cells expressed GFP (Fig. 1D). Qualitatively similar observations were made using another monocytic marker, CD11b.1
Tumor necrosis factor production by glioma-associated macrophages. To determine whether TNF is expressed in GL261 gliomas, brain sections from nonchimeric mice killed 3 weeks after intracerebral implantation of tumor cells were analyzed for the presence of TNF mRNA by radioisotopic in situ hybridization. As exemplified in Fig. 2A, strong hybridization signals were detected throughout the tumors but not in the surrounding nonneoplastic tissue. Adjacent sections hybridized with a sense riboprobe exhibited no positive signals, confirming the specificity of the method. Next, because different types of cells in human brain tumors have been reported to produce TNF (22), we sought to identify the cellular source(s) of this cytokine in our experimental paradigm. To this end, additional sections were doubly labeled for TNF mRNA and Iba1 by combined in situ hybridization and immunohistochemistry. We observed that TNF mRNA was always colocalized with Iba1, but that only a small fraction of Iba1+ cells within the gliomas expressed this transcript (Fig. 2B and C). To further exclude the possibility that GL261 cells produce TNF, we did RT-PCR on RNA extracted from cultured GL261 cells and found no evidence for TNF gene expression (Fig. 2D). Consistently, TNF protein was detected by ELISA in glioma homogenates (9 pg/mL) and conditioned culture media from BV2 microglial cells (706 pg/mL), but not from GL261 cells. Altogether, these results allow us to conclude that TNF is exclusively synthesized by a subset of tumor-associated macrophages in the model used here.
Increased glioma growth in tumor necrosis factor–deficient mice. To explore the hypothesis that macrophage-derived TNF promotes glioma growth in the same way it promotes the development of other types of solid tumors (14–17), we implanted GL261 cells into the brains of TNF knockout and wild-type mice (nonchimeric) sharing the same genetic background. Unexpectedly, as shown in Fig. 3A, the survival of glioma-bearing mice lacking TNF was reduced on average by 19% (5 days) compared with wild-type subjects (mean survival time = 22.6 ± 0.9 days versus 27.9 ± 1.1 days, respectively; log-rank and Wilcoxon tests, P ≤ 0.004). To examine whether this observation could be attributed to increased tumor growth, we next did unbiased volumetric analyses on brain sections from TNF knockout and wild-type mice killed 21 days after tumor cell inoculation. Consistent with the observed reduction in survival, the mean tumor volume was 1.9 times larger in the knockout compared with the wild type (55.5 ± 4.8 mm3 versus 29.2 ± 5.1 mm3, respectively; Student's t test, P = 0.002; Fig 3B). Together, these results indicate that TNF does not promote glioma growth, but instead somehow interferes with this process (directly or indirectly).
Fewer glioma-associated macrophages in tumor necrosis factor–deficient mice. Among its numerous functions, TNF governs many facets of macrophage development and activity. To investigate the basis for the increased glioma growth in TNF-knockout mice, we thus asked whether there were differences in the phenotype and frequency of glioma-associated macrophages between TNF-deficient and wild-type mice killed 21 days after tumor cell implantation. As determined by immunostaining, no disparity was detected neither in macrophage morphology (Fig. 4A-E) nor in levels of proteins known to be expressed by activated macrophages [e.g., Iba1 (35), CD11c (36), and galectin-3 (37)].2
Reduced glioma cavitation in tumor necrosis factor–deficient mice. Histologically, gliomas are often characterized by the presence of small cavities or microcysts whose nature and origin are uncertain (39, 40). By further examining the histology of 21-day gliomas from TNF-deficient and wild-type mice, we observed that all of these tumors contained numerous microcysts of variable size with an increased density at their centers (Fig. 5A-D). These cavities were not blood vessels because their walls were found negative for the endothelial marker CD31 as assayed by immunohistochemistry (Fig. 5E and F). By subtracting the area covered by microcysts from the total tumor area using bitmap images in which the cavities appeared white and the neoplastic tissue appeared black (Fig. 5G and H), we found a 1.9-fold (48%) reduction in the degree of cavitation in the knockout compared with the wild type (13.9 ± 1.6% versus 26.7 ± 1.7%, respectively; Student's t test, P < 0.0001; Fig. 5I). Because this method might confound some large-caliber vessels with microcysts, we sought to rule out the possibility that the observed difference in cavitation was due to alteration in vascularization. To this end, similar image analyses were done on adjacent sections immunostained for CD31. No intergroup differences were observed in the area covered by blood vessels relative to the total tumor area (knockout = 14.1 ± 0.8%, wild type = 13.3 ±0.8%; Student's t test, P = 0.52), indicating that TNF does not influence neovascularization in the model used here, and thus validating the observed difference in cavitation. Finally, linear regression analyses revealed a strong positive correlation between cavitation and macrophage density as well as a strong negative correlation between cavitation and glioma volume (Pearson correlation, r = 0.79 and −0.86; P = 0.0005 and < 0.0001, respectively; Fig. 5J and K). Overall, these results suggest that infiltrating macrophages reduce glioma growth, at least in part, through a TNF-dependent process that leads to microcystic degeneration.
Resistance of cultured GL261 cells to physiologic levels of tumor necrosis factor. TNF could restrict glioma growth by a variety of mechanisms, including stimulation of monocyte recruitment, activation of macrophages, and direct neutralization of glioma cells. To examine this latter possibility, we cultured GL261 cells in the presence or absence of TNF for 24 or 48 hours. The total number of viable cells in each culture was then evaluated by flow cytometry by excluding dead cells by propidium iodide staining. As illustrated in Fig. 6A, GL261 cells were sensitive only to high concentrations of TNF. When cultured with 100 ng/mL TNF, the number of viable cells declined by 25% after 24 hours and 51% after 48 hours (Tukey-Kramer HSD test, P ≤ 0.004). At 10 ng/mL, TNF produced a significant decrease (30%) in the number of viable cells only after 48 hours (Tukey-Kramer HSD test, P = 0.001). This effect was not due to any contaminant because the reduction was completely prevented by the addition of anti-TNF antibody to the culture medium (Tukey-Kramer HSD test, P = 0.001; Fig. 6B). We next asked whether the decrease in the number of GL261 cells resulted from a decline in proliferation, survival, or both. To address this question, GL261 cells were grown for 24 hours in the presence of 100 ng/mL TNF or mock solution, after which the cell cycle and cell death were analyzed by BrdUrd and annexin V labeling and flow cytometry. In TNF-treated cultures, we observed a 13% reduction in the percentage of cells in the S phase of the cell cycle (39.9 ± 1.7% versus 45.7 ± 1.7%; Student's t test, P = 0.032; Fig. 6C) and an equivalent increase in the percentage of cells in the G0-G1 phase (36.2 ± 0.6% versus 32.0 ± 0.6%; Student's t test, P < 0.001), indicating that cell proliferation was slightly reduced by TNF. Furthermore, we found a 2-fold increase in the percentage of cells undergoing apoptosis (Student's t test, P = 0.001; Fig. 6C), whereas no change was observed in the number of necrotic cells. When this experiment was repeated with 10 ng/mL TNF, we noted a 6% decrease in the population of cells in the S phase (50.0 ± 0.7% versus 46.1 ± 0.7%; Student's t test, P = 0.018) but no changes in cell death. Together, these results indicate that TNF does exert cytostatic and cytotoxic effects on GL261 cells in culture, but only at supraphysiologic concentrations, questioning the ability of TNF to directly interfere with glioma cell growth in vivo.
An emerging concept in oncology states that inflammation promotes the growth of transformed cells that have escaped immune surveillance by providing them with a variety of factors required for their migration, proliferation, and survival (2–6, 10). This is supported by studies reporting that mice depleted in macrophages or deficient in cytokines involved in macrophage development were less susceptible to tumorigenesis (7–9, 14–17). Although this concept seems to be true for some cancers such as those of the breast and skin, the present study proposes that it cannot be generalized to all kind of tumors, especially gliomas. The latter develop faster when the population of tumor-associated macrophages is reduced and TNF is absent, suggesting that these inflammatory components do not promote tumorigenesis in the brain as they do in peripheral tissues.
The present findings also suggest that patients with central nervous system tumors could benefit from interventions designed to enhance macrophage recruitment and activity. The development of such treatments will require a better understanding of the regulatory mechanisms that govern the migration of monocytes into gliomas and their conversion into phagocytes with antitumor properties. Presently, therapy for brain cancer includes the administration of dexamethasone, a glucocorticoid agonist, to reduce edema and the resultant increased intracranial pressure (41). The use of indomethacin, an inhibitor of cyclooxygenases, has recently been proposed as a mean to augment postnatal neurogenesis after cranial irradiation (42). Thus, an important question that warrants further investigation is whether gluocorticoids, nonsteroidal anti-inflammatory drugs, or TNF antagonists (e.g., Infliximab, Etanercept) may accelerate the development of established gliomas, interfere with antitumor immunotherapy, or even place individuals at risk for developing brain tumors. Before this question can be fully answered, it may be suitable to restrict the use of these substances to the minimum required for the management of edema so that any interference with macrophage activity and other antitumor mechanisms is minimized.
Our results contrast strikingly both with clinical studies indicating that glioma-associated macrophages correlate with poor prognosis (38) and with basic studies showing that tumor-associated macrophages contribute to tumor growth (7, 8). On the first hand, considering the highly variable nature and origin of human brain tumors obtained from the clinic, the reported positive correlation between macrophage number and glioma malignancy does not necessarily suggest a role for macrophages in glioma growth, but rather may suggest that aggressive gliomas are somewhat more immunogenic than low-grade gliomas. On the other hand, the pro-oncogenic effects of macrophages observed in peripheral organs were attributed, at least in part, to the ability of these cells to promote tumor vascularization (7). This is likely because macrophages not only produce angiogenic factors by themselves, but also stimulate surrounding cells through TNF and other proinflammatory cytokines to secrete angiogenic molecules (2, 4, 6, 10). In the present study, surprisingly, we report that the absence of TNF and the consequent depletion in tumor-associated macrophages had no impact on the degree of glioma vascularization. Therefore, the discrepancy between our results and previous studies could be explained by the fact that angiogenesis in gliomas is not regulated by macrophages as opposed to angiogenesis in peripheral tumors. Further investigations will be required to determine the basis for this difference, and whether additional factors could be responsible for the dual role of macrophages in tumorigenesis.
To our knowledge, the present work provides the first clue linking microcyst formation with macrophage activity in gliomas. Until now, the mechanism by which microcysts are produced in brain tumors has not been explained. It has been postulated that they may arise from either necrosis or a deficiency in the blood-brain barrier, causing extravasation of plasma. Here, we report that the number of macrophages and that of microcyst positively correlate with each other, but negatively with the volume of gliomas. This is consistent with the general view that low-grade human gliomas vary with respect to cell density and microcystic changes, whereas glioblastomas tend to be more homogeneous and densely packed (39, 40). It could be that macrophages recognize and destroy a fraction of glioma cells under certain circumstances, leaving behind only digested material and cavities. Although this possibility needs to be experimentally confirmed, we propose that the degree of cavitation, which could be estimated by in vivo imaging or histology, may serve as a prognostic indicator over the course of a given treatment.
In conclusion, this study provides evidence in agreement with the classic view that macrophages can attack tumor cells rather than the recent concept that they contribute to tumorigenesis. We have learned that macrophages promote their own recruitment into the brain by producing TNF, an observation that has also been reported by other investigators using a model of multiple sclerosis (43). Thus, we speculate that TNF reduces the growth of brain tumors by enhancing the recruitment (and perhaps the activation) of macrophages, but it remains to be shown whether the relatively weak sensitivity of glioma cells in culture toward TNF is further reduced or, conversely, amplified in a more complex in vivo system due to interactions with other signaling pathways. The greatest challenges are now to confirm the beneficial role of macrophages in spontaneously arising gliomas, to unveil the mechanisms by which these phagocytes recognize and kill glioma cells (e.g., cell-to-cell interactions, presentation of antigens to CTLs, and secretion of cytotoxic and cytostatic molecules), and to explore the possibility of intervening in these processes for therapeutic purposes.
Note: L. Vallières was a scholar of the Fonds de la Recherche en Santé du Québec.
Grant support: Canadian Institutes for Health Research, Canadian Foundation for Innovation, National Alliance for Research on Schizophrenia and Depression, and Cancer Research Society.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Maurice Dufour, René Labrecque, and Nicolas Vallières for technical assistance with flow cytometry, mouse breeding, and genotyping, respectively.