Macrophages Kill T9 Glioma Tumor Cells Bearing the Membrane Isoform of Macrophage Colony Stimulating Factor Through a Phagocytosis-Dependent Pathway¹

Macrophages play a complex role in tumor biology; their presence within a tumor can correlate with either tumor destruction or tumor growth (1, 2, 3, 4). Macrophages become cytotoxic for tumor cells in a two-step process (5, 6). Cytokines such as IFN-γ (7), granulocyte-macrophage colony-stimulating factor (GM-CSF) (8), IL-3 (8), TNF (9), and macrophage colony-stimulating factor (M-CSF,³ also known as colony stimulating factor-1) (10, 11, 12) initially prime the macrophages. A secondary triggering signal is supplied either by an Ab, LPS , or taxol (13) allowing the macrophages to kill tumor cells via soluble cytotoxins. TNF, oncostatin-M, hydrogen peroxide, reactive oxygen intermediates (ROI), and reactive nitrogen intermediates (RNI) have all been reported to be possible mediators (14, 15, 16, 17, 18).

Alternative splicing of the M-CSF gene results in different forms of M-CSF being produced (19, 20). The 1.6-kb mRNA translates into a form of M-CSF that stays anchored on the cell membrane (mM-CSF). This isoform is functional since paraformaldehyde-fixed cells stimulate macrophage colony formation when coincubated with bone marrow stem cells (21), but its true physiologic significance remains unknown. When the larger 4-kb or 2.3-kb transcripts are translated, these proteins are cleaved within the secretory vesicle and are released from the cell when the vesicle fuses with the outer membrane. This secreted form of M-CSF (sM-CSF) stimulates the growth and differentiation of macrophages. Many different cell types including tumors are known to produce M-CSF (20); this cytokine may be responsible for the presence of macrophages within breast, ovarian, and brain tumors (22, 23, 24, 25, 26).

Recently, we reported that macrophages directly killed the mM-CSF-bearing tumor cells in an M-CSF dose-dependent manner without the need of secondary triggering signals such as LPS (27). Previously, we were unable to find any evidence of a soluble mediator, indicating that direct cell to cell contact is required for this killing process. Macrophage-mediated killing of the mM-CSF transfectants was completely blocked by 100-fold excess of recombinant M-CSF (27) showing that killing was dependent upon a mM-CSF/M-CSF receptor pathway. In this report, we provide evidence that the putative in vitro killing mechanism of T9 glioma tumor cells expressing the unique membrane isoform of M-CSF includes direct phagocytosis. The significance of this mechanism of tumor cell killing and the potential strategy for a tumor vaccine are also discussed.

Sprague Dawley rats were obtained from either Dr. A. Tarnawski or Dr. S. Szabo (Veterans Affairs Medical Center, Long Beach, CA), who purchased these animals from Harlan Sprague Dawley (San Diego, CA).

Mycoplasma-free cells as determined using the Gen-Probe assay (Fisher Scientific, Pittsburgh, PA) were grown in RPMI 1640 media supplemented with 5% FBS (Gemini, Calabasas,CA) for 2 to 4 days as a monolayer until confluent, when they were passaged 1:6. The conditioned media was filter sterilized through 0.22-μm filters. Transfection and the cloning of the rat T9 glioma cells producing either sM-CSF or mM-CSF was described in Jadus et al. (27).

All chemicals used for this study were purchased from Sigma Chemical Corp. (St. Louis, MO).

Bone marrow cells were cultured in macrophage serum-free media (Life Technologies, Grand Island, NY) using 33% M-CSF transfectant supernatant (28). After one week of culture at 37°C in a humidified 5% CO₂ atmosphere, the media were replaced with fresh 33% conditioned media. All culture materials were disposable plastics and free of endotoxin. Macrophages were removed by washing off the tissue culture media and then incubating the cells in clinical grade irrigation saline (Kendal McGaw Inc, Irvine, CA) for 30 min to 1 h at 4°C. The cells were dislodged using a cell scraper. This procedure results in >95% viability of the macrophages. The identity of the mouse macrophages was >90% positive for Mac-1, Mac-2, Mac-3, and F4/80, as reported previously (28). The identity of the rat macrophages was >90% positive for ED1 and Ox-41, but 25 to 30% positive for ED2, using the Abs obtained from BioProducts for Science (Indianapolis, IN) and >90% positive for rat CD11a, CD11b, CD18, and ICAM using the mAbs from CalTag Inc. (South San Francisco, CA).

Macrophage-mediated cytotoxicity studies were performed according to our previously used methods (27). Exponentially growing tumor cells in 10 ml of complete media were internally labeled with 8 μCi of [³H]TdR overnight. The tissue culture media were replaced with fresh media and allowed to incubate a further 1 to 3 h to reduce spontaneous release by the tumor cells. Ten thousand target cells were incubated in 200 μl of media with graded doses of macrophages ranging from 20:1 to 5:1 at 37°C in a humidified 5% CO₂ incubator. To exclude any LPS contamination, polymyxin B (20–30 μg/ml) was added to the cultures. For those experiments using chemical inhibitors of phagocytosis, the chemicals were added at the start of the cytotoxicity assay. Immediately before the supernatants were harvested, cultures were viewed under an inverted microscope to confirm whether tumor cells were present or absent under the various experimental conditions. Afterward, 100 μl of supernatant was removed and placed into 2 ml of scintillation fluid. Maximum release is calculated by taking 10⁴ target cells and rapidly freeze thawing them three times in liquid nitrogen. Specific release is calculated using the standard equation for cytotoxicity reactions (27). For those experiments that used chemical inhibitors, the spontaneous release was obtained from those cultures in which the drug was cultured with the target cells without any effector macrophages. Visual observations from each experiment confirmed the cytotoxicity results. Cytotoxicity data from triplicate cultures at each macrophage:tumor cell ratio are presented as the mean ± SD. Cytotoxicity is not considered relevant if values are ≤ 10% specific release. Data from the cytotoxicity assays were analyzed using Student’s t test on the Sigma Plot Version 5.0 (Jandel Scientific, San Rafael, CA) computer program. Values were considered significantly different at the p < 0.05 levels.

To detect the production of nitric oxide, we used the Griess reaction previously described by Jadus et al.(29) in a 96-well microassay system using an ELISA plate reader. For the positive control, we used macrophages stimulated with activated zymosan. The zymosan (Sigma) was activated by taking 50 mg of the dried particles and incubating it with freshly isolated human plasma for 20 min at 37°C. The zymosan was washed once with PBS and resuspended with 1 ml of the buffer.

To detect the production of hydrogen peroxide we used the scopoletin oxidization technique as desribed by Root et al.(30). Cells were incubated in Krebs-Ringer phosphate buffer in the presence of 26 μM scopoletin, 1.0 unit/ml horse radish peroxidase. Macrophages were incubated with the T9-C2 cells at a 10:1 macrophage:tumor ratio for 105 min. For the positive control, human polymorphonuclear leukocytes were stimulated by activated zymosan for 15 min. The cells were incubated in a quartz cuvette and excited at 350 nm and measured at 460 nm using an Aminco-Bowman Spectrophotofluorometer (Silver Spring, MD). Data are recorded as the change of fluorescence at 460 nm over time (min).

Macrophages were incubated at 1 × 10⁶ cells/ml at 1:1 ratio with the T9, T9-H1, and T9-C2 cells. At 30-min intervals 50 μl of the cell suspension was removed and a cytocentrifuge spin was prepared. The cells were air dried and acetone fixed. The slides were incubated with a goat serum for 1 h to block nonspecific binding. A 1:10,000 dilution of the rabbit anti-Hck Ab (Santa Cruz Biotech, Santa Cruz, CA), which stains the myeloid cells including macrophages (31), or a nonspecific control rabbit serum was incubated for 1 h. The slides were washed 3 times and a biotinylated conjugated secondary goat anti-rabbit IgG Ab was applied for 1 h. The slides were washed and incubated with the ABC reagent (Vectastain, Vector Labs, Burlingame, CA) for 30 min, washed twice, and the diaminobenzidine (DAB) peroxidase substrate was applied and allowed to develop to a brown color (Vector Labs). Cells were counter stained blue with hematoxylin to identify the nucleus of the cell as well as the negatively stained glioma cells.

One million bone marrow-derived macrophages were plated into Corning T25 cm² flasks overnight along with complete media containing M-CSF. The macrophages were washed twice with PBS to remove any nonadherent macrophages, and fresh media without any M-CSF were added for 1 to 2 h. Exponentially growing T9-C2 cells (>95% viability) were collagenase treated, and 3 million cells were added to the macrophages at time 0. Filming began when the T9-C2 cells were added to the culture. Macrophages were already attached to the bottom of the flask and the target cells were nonadherent, so it was possible to positively identify both the effector and the target cells. Time lapse photography was performed using a Nikon phase-contrast inverted microscope with the 20× lens using a Sony video camera. The video was recorded on a GYYR VCR. Pictures were taken at appropriate time intervals using a Snappy Video grabber (Play Inc., Rancho Cordova, CA), digitized, and then printed on a color printer.

The annexin V staining kit was purchased from Trevigen Inc. (Gaithersburg, MD). The T9-C2 cells (>97% viable) were incubated with the macrophages for 3 h at 37°C. An aliquot of the T9-C2 cells was killed with 10 mM H₂O₂ for 3 h. The cells were washed in ice cold PBS and labeled with the FITC-conjugated annexin V, as described in the manufacturers directions. The cells were washed and analyzed on the Epics Profile flow cytometer. Data from five thousand cells were collected and analyzed on the Multi2D program (Phoenix Flow Systems, San Diego, CA).

Tumor target cells were stained in a PBS solution containing 5 μg/ml solution of acridine orange at room temperature for 5 to 10 min. The cells were washed 3 times with PBS and allowed to incubate with the macrophages for 2 to 3 h. The cells were loaded onto a microscope slide and cover slipped. The cells were viewed under an Olympus fluorescent microscope. Photographs were taken at various exposures. The slide presented in Figure 8 is overexposed to provide sufficient illumination of the internal structure of the phagocytic macrophage. Living cells’ cytoplasm under UV illumination fluoresce bright green, whereas dead cells killed by either rapid freeze thawing or H₂O₂ fluoresce red under UV light (32).

Two million macrophages or two million tumor cells (T9 or T9-C2) cells were washed in the PKH wash buffer and resuspended in 0.5 ml of the PKH buffer. The cells were added to 0.5 ml PKH buffer containing either 10 μl of either the stock PKH-2 green dye (for the tumor cells) or 10 μl of the stock PKH-26 red dye (for the macrophages) and incubated at room temperature for 5 min. Afterward, 1 ml of FBS was added to stop the reaction for 1 min. The cells were washed three times with PBS. The cells were cultured overnight to assure that all excess dye was removed from the cells. On the next day the cells were mixed together at a 10:1 macrophage:tumor ratio. The cells were examined under the Olympus fluorescence microscope equipped with epifluorescence. Three hundred red macrophages were counted to determine whether they contained a phagocytosized green tumor cell.

The derivation of the mM-CSF transfectant clone, T9-C2, and the secreted sM-CSF clone, T9-H1, were described previously (27, 33). The T9-C2 clone, which possessed 1002 pg mM-CSF/10⁴ cells, was used for these studies while cloned T9-H1 cells, which produced >2000 pg/ml of sM-CSF, were used for the secreted clone. Rat bone marrow-derived macrophages (>90% for CD11a, CD11b, CD18, ED1, and Ox-41) consistently killed only the mM-CSF-transfected T9 clone, T9-C2, in a standard macrophage cytotoxicity assay as shown in Figure 1. Cytotoxicity was observed at all macrophage:T9-C2 cell ratios, while minimal cytotoxicity was observed against either the parental T9 or the cloned T9-H1 cells after 1 or 2 days. Better cytotoxicity was seen against the T9-C2 cells after 2 days of coculture. When paraformaldehyde-fixed macrophages were tested against the T9-C2 clone, no tumoricidal activity was observed, indicating that living macrophages are required for this cytotoxicity (Table I).

To exclude reactive nitrogen intermediates (RNI) and reactive oxygen intermediates (ROI) from killing the mM-CSF targets, we used various chemical inhibitors that prevent the actions of these short-lived cytotoxins. l-N^ω-nitroarginine methyl ester (l-NAME), catalase, and superoxide dismutase were added to cultures of macrophages killing the T9-C2 cells (Fig. 2,A). In none of these cultures was there any inhibition of the cytotoxicity displayed against the mM-CSF-transfected cells. There were no signs that nitric oxide was being produced by the macrophages in reacting to the T9-C2 cells as shown by the Griess reaction (Fig. 2,B). Here the macrophages reacting without the tumor cells produced as much NO as did the macrophages responding to the T9-C2 cells. Likewise, there were no signs that hydrogen peroxide was being produced when the macrophages reacted to the T9-C2 cells (Fig. 2 C). This work confirms our previous report that these intermediates do not prevent the macrophages from killing these cells (27).

Preliminary visual studies suggested that macrophages were capable of phagocytosizing tumor cells bearing mM-CSF. To investigate whether phagocytosis was involved with the killing, we used several known chemical inhibitors of phagocytosis to prevent the killing of the T9-C2 cells. In Figure 3, iodoacetate, 2-deoxyglucose, cytochalasin B, and gadolinium chloride significantly (p < 0.05) prevented the macrophages from killing the T9-C2 clone. Visual observation of these cultures revealed that the macrophages were still adherent and viable; this eliminated the trivial possibility that these chemicals had killed the macrophages. Most interesting, the microtubule inhibitor, colchicine, actually enhanced macrophage-mediated killing of the T9-C2 cells.

Phagocytosis can take 1 to 3 h; to demonstrate that macrophage-mediated phagocytosis of T9-C2 cells can occur within this time, we used the anti-Hck Ab in an immunohistochemical staining technique. Figure 4,A shows a typical M-CSF-activated macrophage. All macrophages are positively stained (brown) by the anti-Hck Ab while the T9-C2 tumor cell (Fig. 4,B) stained only with the hematoxylin counter stain. Within 1 h of coculture, macrophages began to phagocytosize mM-CSF tumor cells (Fig. 4, C and D) and by 2 h phagocytosis can be completed (Fig. 4 E). During this time, the T9-C2 cells appear uniform and do not show any signs of membrane disruption. No phagocytosis was observed when either the parental T9 or the sM-CSF-transfected T9-H1 cells were incubated together with the macrophages over the same time.

To confirm that macrophages were phagocytosizing the mM-CSF transfectant cells, photocinematography was done. Figure 5 shows the sequence of events occurring when an adherent macrophage was engulfing the T9-C2 cloned cell. Macrophage:T9-C2 tumor cell contact was detected within 20 min and was strongly contacting the macrophage at 83 min (Fig. 5,A). Phagocytosis of the intact T9-C2 cell was underway by 102 min and showed intramacrophage stress elements immediately adjacent to the glioma cell being internalized (Fig. 5,B). By 108 min (Fig. 5,C) complete phagocytosis occurred. The engulfed T9-C2 cell was still intact and highly refractile. Finally the T9-C2 cell quickly swelled; then the T9-C2 cell lost membrane integrity and ruptured within the macrophage (Fig. 5 D). At this time there were no signs of membrane blebbing characteristic of apoptosis. It is at this time that the tumor cell is killed.

We next determined the number of macrophages that were capable of killing the mM-CSF-transfected tumor cells by phagocytosis. We used the stable membrane dyes, PKH-2 and PKH-26, to label the T9 and T9-C2 cells green and the macrophages red. The cells were mixed at a 10:1 macrophage:tumor ratio for 24 h to assure that macrophage cytotoxicity was occurring with the mM-CSF-transfected cells. By flow cytometry, macrophages had bound to both T9 and T9-C2 cells as evidenced by dual fluorescence staining by 2 h. This finding was not unexpected since this technique has been used for effector-target conjunction by various cells. However, under the fluorescence microscope, there were more macrophages that phagocytosized the T9-C2 cells (13 cells/300 macrophages) than phagocytosized the control T9 tumor cells (2 cells/300 macrophages).

We excluded the possibility that the macrophages first killed the T9-C2 cells and then phagocytosized them by two independent methods. First, the annexin V staining technique was used. Macrophages were incubated together in the presence of the T9-C2 cells for 3 h at a 1:1 ratio. The cells were then tested for the presence of extracellular phosphatidylserine. Annexin V binds phosphatidylserine, normally found on the inner membrane of living cells. Dead cells will bind to annexin V because the phosphatidylserine gains access to the exterior membrane. As shown in Figure 6, macrophages that interacted with the T9-C2 cells (middle panel) did not show any difference from that of the T9-C2 cells alone (top panel). The T9-C2 cells killed with H₂O₂ were positively stained for annexin V (bottom panel). We also performed this experiment with a macrophage:tumor ratio as high as 10:1 that was incubated for 24 h to show that there was no difference in the amount of annexin V staining after the macrophages killed the T9-C2 tumor cells. Figure 7 shows that there was no increase in the amount of annexin V fluorescence between the macrophages responding to the T9-C2 cells as compared with the macrophages responding to the T9 parental cells.

Second, we used the acridine orange staining technique. Living T9-C2 cells were first loaded with acridine orange and then cultured with the macrophages for 2 to 3 h. We have reproducibly found macrophages that have engulfed bright green T9-C2 cells as shown in Figure 8. Here a living T9-C2 cell is found within the macrophage while another living T9-C2 cell is found in close proximity to the macrophage. From these two experiments it appears unlikely that the macrophages have killed T9-C2 cells first and then phagocytosized them.

Macrophages can be considered a “double edged sword” in tumor biology; these cells are commonly found within various tumors in response to M-CSF produced by these tumors (22, 23, 24, 25, 26). Macrophages stimulate tumor growth by releasing growth or angiogenic factors or by acting as immunosuppressor cells (34, 35, 36, 37, 38, 39). Other studies concluded that macrophages were beneficial for the host’s survival (40, 41, 42, 43, 44, 45). For macrophages to become tumoricidal in vitro they must be stimulated twice (5, 6). First, cytokines prime the macrophages, while secondary signals allow the macrophage to kill the tumor cell. It could be speculated that this “double edged sword” effect is explainable by the two-signal model. When macrophages receive the priming signal, they promote tumor growth and metastases. Whereas, when both signals are received the macrophages mediate tumor regression. One way to tip the balance toward a tumoricidal response is to devise a molecule that delivers both cytotoxic delivery signals simultaneously to the macrophage. Membrane M-CSF may be a candidate molecule for this dual function. Macrophage cytotoxicity against rat T9 glioma cells occurred only when the tumor cells expressed the membrane isoform of macrophage colony stimulating factor (mM-CSF).

In this report, we have reproduced our previous in vitro findings that macrophages will kill T9 tumor cells bearing mM-CSF but not the parental T9 glioma cells or the sM-CSF-transfected T9 cells in the absence of exogenous LPS (Fig. 2 and Table I). We present evidence that bone marrow-derived macrophages are killing the mM-CSF-transfected target cells through a novel phagocytic-dependent mechanism. Chemical inhibitors of phagocytosis such as cytochalasin B, 2-deoxyglucose, iodoacetate, and gadolinium chloride (Fig. 3) prevented the macrophages from killing our mM-CSF target cells. To eliminate simple cytoskeletal disruption from being the mechanism of tumoricidal macrophages, colchicine was added to other macrophages. Colchicine prevents microtubule polymerization in various cells, just as taxol does. Taxol also supplies the secondary cytotoxic signal to macrophages for tumor cytotoxicity (13), suggesting that augmentation of the mM-CSF-specific killing can be achieved with secondary signals.

Using immunohistochemical techniques with the anti-Hck Ab, we observed that macrophages can phagocytosize the mM-CSF transfectants within 2 h (Fig. 4). Time-lapse photography verified these observations (Fig. 5). Once the T9-C2 cell was phagocytosized, the tumor cell first swelled and then ruptured suggesting osmotic lysis, as opposed to cell shrinkage seen in apoptosis (46). Previous coculture experiments excluded soluble cytotoxic factors as likely mediators responsible for the killing of the T9-C2 cells. No evidence of prior cell killing was observed when we used an annexin V staining technique (Figs. 6 and 7). We have observed that when the T9-C2 cells are first stained with acridine orange, macrophages have engulfed green living T9-C2 cells (Fig. 8). This suggests that macrophages do not kill the mM-CSF transfected tumor cells before phagocytosis.

Others have shown mixed results with M-CSF transfection in various tumor models. Myeloma cells transfected with sM-CSF grew as an ascites: macrophages were found within the ascites but failed to stop tumor growth (47). This showed that M-CSF acted as a chemoattractant for macrophages and did not elicit any direct tumoricidal activity, perhaps by not allowing the macrophage to physically contact the tumor cell. In three other cases, M-CSF transfection resulted in tumor rejection (48, 49, 50) and immunized the mice against the parental tumor. Those studies did not address the exact mechanism of tumor cell killing; it would be of interest to determine whether those M-CSF transfectants had any membrane-associated M-CSF.

If macrophages are directly killing the mM-CSF-transfected tumors in vivo, then the possibility exists that these macrophages should act as an Ag-presenting cell to T cells. T9-C2 cells, when implanted into rat brains, were rejected and most rats were long-term survivors, whereas all rats implanted with the T9 and T9-H1 cells died by 35 days.⁴ Two days after implantation, a large number of macrophages/microglial cells were found physically contacting the T9-C2 cells; relatively few isolated T9-C2 glioma cells were present when compared with a sM-CSF-transfected clone T9-H1. By Day 6, lymphocytes were found within the implantation site and very few T9-C2 tumor cells were still present. Those rats that survived the challenge of the mM-CSF tumor cells also resisted the challenge of the malignant parental T9 tumor cells but not the challenge of an unrelated rat breast adenocarcinoma cell. Adoptive transfer of CD3⁺ splenocytes from rats that rejected the mM-CSF transfectants protected naive rats from a lethal challenge with the parental T9 cells. We believe that immunization against this tumor cell occurred because the macrophages/microglial cells first killed the mM-CSF-transfected tumor cells. These macrophages then presented the phagocytosized tumor Ags to the T cells, which stimulated the immune response.

In summary, we have found that M-CSF-activated bone marrow-derived macrophages killed T9 cells expressing the membrane isoform of M-CSF. This cytotoxicity was mediated by direct phagocytosis of the mM-CSF tumor cells. Macrophages treated with chemical inhibitors of phagocytosis did not kill the mM-CSF-bearing tumor cells. Immunohistochemistry and time-lapse photography confirmed that the macrophages were phagocytosing the tumor cells within 3 h. Using an acridine orange staining technique we observed that living T9-C2 glioma cells were found within the macrophage, suggesting that the macrophages did not kill the mM-CSF-transfectant tumor cell before phagocytosis.

We thank Marie Flack for doing the immunohistochemistry. We especially thank Dr. J. Wu (Veterans Affairs Medical Center, Long Beach, CA) for purchasing some much needed supplies and plasticware due to shortage of our funding; and Drs. Tarnawski and Szabo for the rats we used in this work. We also acknowledge the help supplied to us by Terry Berger of the Coulter Corporation for setting up our flow cytometer for the PKH-2 and PKH-26 dyes. We are also indebted to Dr. Katherine Muirhead of SciGrow Inc. for many useful suggestions and much advice concerning the use of PKH-2 and PKH-26 dyes. We thank Dr. John Hiserodt (University of California, Irvine) for many insightful discussions. We thank Dr. Dennis Anjo of California State University-Long Beach for allowing us to use the fluorometer to measure hydrogen peroxide.