Intracranial Microenvironment Reveals Independent Opposing Functions of Host αVβ3 Expression on Glioma Growth and Angiogenesis*

αVβ3 integrins are overexpressed in the host-derived vasculature of glioblastoma multiform (GBM) and are believed to contribute to angiogenesis and tumor growth. To directly address the role of host αVβ3 expression in GBM growth and behavior, we intracranially implanted integrin β3-expressing GBM cells into β3 wild type (WT) or β3 knock out (KO) mice and monitored angiogenesis and growth. GBM in β3 WT animals had a vessel density greater than that in β3 KO animals, consistent with a pro-angiogenic, pro-tumorigenic view of host integrin function. GBM in β3 WT animals, however, were no larger than those in β3 KO animals, because GBM in β3WT animals were infiltrated with a higher number of tumor necrosis factor α-secreting, apoptosis-inducing macrophages than the tumors in the corresponding β3 KO animals. The tumor-suppressive effects of host β3 expression could be reversed by macrophage depletion or by transplantation of bone marrow from β3 KO animals into β3 WT animals, both of which significantly increased tumor growth independently of tumor vessel density. Taken together, these results show that host αVβ3 integrin expression has opposing actions in the intracranial setting, enhancing tumor vascularization and growth while independently enhancing macrophage-mediated tumor elimination. Appropriate management of these functions could lead to enhanced efficacy of anti-integrin based therapies for glioma.

␣V␤3 integrins are overexpressed in the host-derived vasculature of glioblastoma multiform (GBM) and are believed to contribute to angiogenesis and tumor growth. To directly address the role of host ␣V␤3 expression in GBM growth and behavior, we intracranially implanted integrin ␤3-expressing GBM cells into ␤3 wild type (WT) or ␤3 knock out (KO) mice and monitored angiogenesis and growth. GBM in ␤3 WT animals had a vessel density greater than that in ␤3 KO animals, consistent with a pro-angiogenic, pro-tumorigenic view of host integrin function. GBM in ␤3 WT animals, however, were no larger than those in ␤3 KO animals, because GBM in ␤3 WT animals were infiltrated with a higher number of tumor necrosis factor ␣-secreting, apoptosis-inducing macrophages than the tumors in the corresponding ␤3 KO animals. The tumor-suppressive effects of host ␤3 expression could be reversed by macrophage depletion or by transplantation of bone marrow from ␤3 KO animals into ␤3 WT animals, both of which significantly increased tumor growth independently of tumor vessel density. Taken together, these results show that host ␣V␤3 integrin expression has opposing actions in the intracranial setting, enhancing tumor vascularization and growth while independently enhancing macrophage-mediated tumor elimination. Appropriate management of these functions could lead to enhanced efficacy of anti-integrin based therapies for glioma.
Brain tumors affect ϳ100,000 Americans per year and remain among the most difficult of human tumors to successfully treat. Individuals with glioblastoma multiform (GBM), 3 a highly aggressive and angiogenic form of glioma, have a particularly poor prognosis, with an average survival of Ͻ2 years (1)(2)(3). Although surgical resection followed by radiation and chemotherapy have been shown to slightly increase the lifespan of individuals with GBM, new approaches that can selectively target the tumor are needed.
The search for agents that can selectively target glioma cells has led to an interest in the role of integrins in glioma biology. Integrins are transmembrane glycoprotein complexes of noncovalently linked ␣ and ␤ subunits. There are 8 known ␤ subunits that combine with 18 ␣ subunits in a defined manner to create more than 24 unique ␣␤ heterodimers. Although some ␤ subunits combine with a large number of ␣ subunits, others exhibit greater selectivity and combine with single ␣ subunits (4). As an example, the ␤3 subunit combines exclusively with the ␣V subunit in most tissues except in platelets, where it also exhibits a pro-clotting interaction with the ␣IIb integrin subunit (5). Integrin heterodimers each recognize a specific range of ligands in the extracellular matrix or on neighboring cells, and integrin binding plays a key role in adhesion of integrinexpressing cells to the extracellular matrix. Integrin-mediated cellular anchorage, however, is also accompanied by linkage of activated integrins to the actin cytoskeleton and the triggering of a large number of signal transduction pathways that can influence nearly every aspect of cellular behavior (6 -8).
Because integrins play a key role in processes requiring cell-cell interaction such as cell migration and angiogenesis, they also appear to play a key role in tumorigenesis. ␣V␤3 integrin complexes in particular are overexpressed in the proliferating vascular endothelial cells surrounding the tumor (9 -11). This is especially true in GBM, in which ␣V␤3 is overexpressed at the invasive, highly angiogenic edges of the tumor (12). These observations have led to the suggestion that ␣V␤3 integrins may play a role in the angiogenesis and/or growth and invasion of GBM. This concept has been further supported by the demonstration that monoclonal antibodies and small cyclic arginine-glycine-aspartate peptide ligands of ␣V␤3, both of which function presumably as ␣V␤3 antagonists, led to tumor regression in animal models of GBM (13). These studies as a whole have defined ␣V␤3 as a promising target for tumor-selective therapy of GBM.
More recent detailed studies, however, have questioned the established role of tumor and host ␣V␤3 expression in glioma growth. We previously showed that overexpression of ␣V␤3 in intracranially implanted GBM cells did not increase the blood vessel density or size of the subsequent tumors but, rather, suppressed both tumor oxygenation and growth (14). Because the growth suppressive effects of ␤3 expression in the tumor could be overcome by genetic alterations (Akt activation, VEGF overexpression) common to GBM, however, these studies suggested that ␣V␤3 expression in the host, rather than in the tumor, may play a more important role in modulating glioma growth. The effects of host ␣V␤3 expression on tumor growth have been examined, and in these studies host ␣V␤3 expression surprisingly suppressed rather than enhanced the growth of subcutaneously implanted human tumor cells in association with suppression of angiogenesis (15,16). These studies, however, were done in the relatively avascular subcutaneous setting and, therefore, might not adequately address the role of host ␣V␤3 integrin expression in the highly vascular intracranial setting that is home to GBM. Because angiogenesis and tumor growth are highly tissue-specific processes, both of which are greatly influenced by the tumor microenvironment (17,18), we considered the possibility that effects of host ␣V␤3 expression might be tissue-dependent and far different in the intracranial versus the subcutaneous setting. We here report that tumorrelated effects of host ␣V␤3 expression are indeed highly dependent on tumor microenvironment and that in the intracranial setting host ␣V␤3 expression increases glioma vessel density while independently suppressing glioma growth by enhancing TNF␣-secreting macrophage infiltration into the tumor. Manipulation of these independent opposing functions of ␣V␤3 expression functions could lead to enhanced efficacy of anti-integrin-based therapies for glioma.

EXPERIMENTAL PROCEDURES
Cell Lines-U251, an established human GBM cell line, was obtained from the University of California, San Francisco (UCSF) Brain Tumor Research Center Tissue Bank. GL261, a mouse GBM cell line created by intracranial implantation of a methylcholanthrene pellet in a C57BL/6 animal (19), was provided by Dr. Y. Gillespie, University of Alabama at Birmingham. Cells were maintained as monolayers in a complete medium consisting of Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. An immortalized human dermal microvascular endothelial cell line (HDMEC) was provided by Dr. Gabriele Bergers (UCSF Dept. of Neurological Surgery). HDMEC cells were cultured in MCDB131 media (Invitrogen) with 10% fetal bovine serum and 1:100 L-glutamine. All cells were cultured at 37°C in a humidified atmosphere consisting of 5% CO 2 . All reagents for cell culture were obtained from UCSF Cell Culture Facility.
Mice and Tumorigenesis Assay-␤3 integrin knock out (KO) (15) mice were intercrossed with immunodeficient mice lacking the Rag1 gene to generate immunodeficient Rag1 KO mice with or without ␤3 integrin expression (C57BL6/129Sv). The genotypes were confirmed by the polymerase chain reaction as described previously (20). Intracranial injections of U251 and GL261 cells into mice were performed as described previously (14,21). Briefly, 3 ϫ 10 6 cells were stereotactically injected into striatum of age-and sex-matched anesthetized mice (14,22). Mice were sacrificed 35 days after injection. Tumor-bearing brains were sectioned coronally at the point of cellular implantation. After hematoxylin and eosin staining, sections were photographed, and the length (a) and width (b) of the largest tumor cross-sectional areas were determined. To obtain another parameter of widths (c), samples were cut into 10-m serial coronal sections, and every 10th section was stained for hematoxylin and eosin. Tumor volume was calculated using the standard formula (V ϭ length (a) ϫ width (b) ϫ width (c) ϫ 0.52). For subcutaneous studies, 5 ϫ 10 6 cells were injected into the flank, and 25 days after injection the volume was calculated using the standard formula (width ϫ widths ϫ length ϫ 0.52) (23)(24)(25)(26)(27).
Mouse Bone Marrow Transplants (BMT)-Six-week-old recipient female immunodeficient Rag1 KO mice with or without ␤3 integrin expression (C57BL6/129Sv) were lethally irradiated with 8Gy (4Gy ϫ 2 at a 4-h interval) and reconstituted by tail vein injection of 5 ϫ 10 5 BM cells isolated from male mice generated from the same breeding pair of animals and with the same genetic background. These mice were treated with antibiotic water for 14 days. Four weeks after BMT the mice were used for further analyses.
Immunohistochemistry-5-m paraformaldehyde-fixed, paraffin-embedded sections or 6-m frozen sections were used for immunohistochemistry. Primary antibodies used for immunohistochemistry were rat anti-F4/80 (Serotec, Oxford, UK), rabbit anti-TNF␣ (R&D Systems, Minneapolis, MN), rabbit antivon Willebrand factor (vWF) (Abcam, Cambridge, MA) and rat anti-endoglin (R&D). All were used at 1:200 dilution except anti-vWF antibody, which was used at 1:800 dilution. For immunofluorescence, FITC-or rhodamine-conjugated secondary antibodies were used at the dilution of 1:200. These slides were counterstained by DAPI and mounted with antifade solution (Prolong, Molecular Probes, Eugene, OR). Vessel density was assessed by counting the number of vWF-or endoglinpositive vessels in 10 independent tumor fields at ϫ400 with the aid of an ocular grid (14). Vessel density was reported as the mean Ϯ S.D.
In Situ Hybridization Combined with Immunohistochemistry-In situ hybridization was performed as described elsewhere (28). Briefly, digoxigenin (DIG)-labeled cDNA probe was prepared by using a PCR DIG Probe Synthesis Kit (Roche Diagnostics). Paraffin-embedded tumor sections were prepared by deparaffinization, rehydration, and digestion with proteinase K for 15 min at 37°C. Hybridization was performed at 42°C overnight in a hybridization mixture consisting of 50 l of deionized formamide, 10 l of salmon sperm DNA (Invitrogen), 10 l of dextran sulfate, 10 l of 50ϫ Denhardt's solution, 10 l of 20ϫ standard saline citrate, and 500 ng of denatured DIG-labeled probe. The DIG-labeled Y chromosome was visualized using anti-DIG fluorescein-conjugated antibody (Roche Diagnostics). To detect the Y chromosome-positive macrophages, slides were initially immunostained for F4/80 expression, after which in situ hybridization was performed.
Flow Cytometry-FITC-conjugated rat monoclonal antibodies (rat anti-mouse anti-F4/80, Serotec; anti-NK-cell, BD Biosciences) were used to stain monocytes/macrophages or NK cells present in the blood of mice, respectively. Blood samples were incubated with a 1:100 dilution of antibody for 1 h at room temperature, treated with red cell lysis buffer (BD Pharmingen) (29), washed twice with phosphate-buffered saline (PBS), resuspended in PBS, and analyzed in a FACScan flow cytometer with CELLQUEST software (BD Biosciences). Aliquots of macro-phages from ␤3 WT and KO mice were sorted and plated onto culture dishes for further analysis.
Macrophage Conditional Supernatant/Western Blot Analysis-Macrophages isolated from ␤3 WT or KO mice were activated by incubation with 100 ng/ml lipopolysaccharide (Chemicon, Temecula, CA). The culture was maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum at 37°C for 24 h. At the end of the culture, supernatants were collected, separated into aliquots, and stored at Ϫ80°C for further analysis. The supernatant from the macrophage culture without lipopolysaccharide stimulation was used as a control. Activated macrophages were harvested and lysed as described previously (14). Whole cell lysate (10 g) was subjected to gel electrophoresis and electroblotted onto an Immobilon-P membrane (Millipore, Bedford, MA). The membrane was blocked in 5% skim milk and incubated with antibodies against TNF␣ (1:1000, R&D Systems, Minneapolis, MN) or ␣ -tubulin (1:2000, Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C. Bound antibody was detected with horseradish peroxidase-conjugated secondary antibody using ECL Western blotting detection reagents (Pierce).
Analysis of Cell Cycle Distribution/Annexin V-FITC Assay-Cells were harvested and washed in PBS and fixed in 70% ethanol at Ϫ20°C. The cells were washed once with PBS followed by incubation in PBS containing 40 g/ml propidium iodine (Sigma) and 200 g/ml RNase A (Sigma) for 1 h at room temperature in the dark. Stained nuclei were then analyzed on FACScan machine (BD Biosciences) with 10,000 events/determination. ModFit LT software (Verity Software House, Inc., Topsham, ME) was used to assess cell cycle distribution. The annexin V-FITC binding assay was done using an ApoAlert Annexin V kit (Clontech, Mountain View, CA) according to the manufacturer's instructions. Briefly, after incubation with macrophage supernatant, U251 cells were washed with PBS and trypsinized. Cells were rinsed and then suspended in binding buffer. After incubation with annexin V-FITC, propidium iodine was added to the cell suspension. Annexin V-FITC positive and propidium iodine-negative cells were counted on FACScan machine (BD Biosciences) with 20,000 events/ determination.
Apoptosis Analysis-For cell death analysis, DNA fragmentation (terminal deoxynucleotidyltransferase-mediated nick end labeling (TUNEL) staining) was performed by using the FD Apop kit (FD Neurotechnologies, Ellicott City, MD) according to the manufacturer's instructions. Nuclei were counterstained with methyl green.
In Vivo Macrophage Depletion-Macrophages were depleted by intraperitoneal injection of an anti-macrophage antibody (Accurate Chemical and Scientific, Westbury, NY) according to the manufacturer's protocol. Antibody injection was begun 2 days before tumor implantation and was repeated at days 0 (same day as implantation), 2, 5, and 7 and twice a week thereafter until animal sacrifice. Preliminary tests using these antibodies confirmed that this protocol significantly reduced the number of circulating macrophages while not effecting levels of marrow-derived NK cells. As a control, rabbit normal IgG was used in the same strategy.
Transendothelial Migration Assay-To assess in vitro macrophage transmigration ability, we performed transendothelial migration assays (30). Briefly, transwell culture inserts were coated with 50 mg/ml laminin for 30 min. Excess laminin was removed from the inserts and immortalized HDMEC at 10 6 /ml were seeded on the inserts in 100 l of medium. Cells were allowed to grow to confluence on filters for 48 h. The macrophages isolated from ␤3 WT or ␤3 KO mice were adjusted to 10 5 cells/ml, and 100 l of cell suspension were added per insert. Medium (300 l) containing 125 ng/ml monocyte chemoattractant protein-1 was placed into the lower chambers of the transwells, and the inserts were carefully placed into the lower chambers to avoid air bubbles forming at the interface between the underside of the insert and the medium. Migration was then allowed to proceed for 5 h at 37°C, after which the number of cells that had migrated into the lower chamber was determined. For blocking studies, anti-integrin ␤3 antibody (Chemicon) was added to a final concentration of 50 mg/ml. The tubes were then incubated on a shaker for 1 h at 4°C. Before the assay, cells were spun down, washed once in medium, and then resuspended in 300 l of fresh medium. This was then added to three wells per condition. For each experiment, the number of cells that had transmigrated was expressed as the mean value of cells counted in three wells.
Statistical Analysis-All statistical analyses were performed using the Student's t test, with significance defined as p Ͻ 0.05.

Host Integrin ␤3 Expression Has Different Effects on Subcutaneous Versus Intracranial Tumors-Previous studies have
shown that host ␣V␤3 expression suppressed rather than enhanced the growth of human tumor cells in association with suppression of angiogenesis (15,16), although all these studies were done in the relatively avascular subcutaneous setting. To determine how host ␤3 expression might contribute to tumor formation in a vessel-rich brain tumor microenvironment, we implanted human U251 GBM cells subcutaneously and intracranially into immunodeficient (Rag1 KO) ␤3 WT or ␤3 KO mice and monitored the tumor vessel density and size of the resultant tumors. Host ␤3 expression suppressed the subcutaneous growth of U251 cells (Fig. 1A) and also decreased the density of blood vessels in the tumor as detected by immunohistochemistry using antibodies recognizing the endothelial cell markers vWF (Fig, 1B) or endoglin (not shown). These results are consistent with previous reports showing host ␤3-mediated suppression of the growth and angiogenesis of other subcutaneously implanted human tumor cell lines (15,16). In the intracranial setting, however, effects of host ␤3 expression on tumor growth and vascularity were quite different. U251 GBM cells intracranially implanted into ␤3 WT animals formed tumors that were no different in size from those that formed in ␤3 KO animals (Fig. 1C). Furthermore, an examination of vessel density showed that unlike what was seen in the subcutaneous setting, host ␤3 expression enhanced rather than decreased tumor vessel density (Fig. 1D), with the difference particularly apparent in the center (bottom panels, Fig. 1G) versus the invading edge of the tumors (top panels, Fig. 1G). These findings were neither unique to the U251 GBM cell line nor to the implantation of human cells into mice, because when mouse GBM cells (GL261) were intracranially implanted into the same animals used for the U251 cell studies, the mouse GL261 cells formed tumors of equal size in ␤3 WT and ␤3 KO animals (Fig. 1E) just as was noted in studies using human U251 GBM cells. Similarly, the GL261 tumors that formed in the ␤3 WT animals had a greater vessel density than those in the ␤3 KO animals (Fig. 1F). These results suggest that host ␤3 expression influences tumor behavior very differently depending on the site of tumor implantation.
Host ␤3 Expression Suppresses Intracranial Growth in a Bone Marrow-dependent Manner-One possible explanation for why host ␤3 expression enhanced intracranial tumor vascularity without increasing tumor size is that host ␤3 expression may have growth-suppressive actions as well as pro-growth, proangiogenic actions. A first clue as to what these growth suppressive actions might be was the observation that the intracranial GBM in ␤3 WT animals, in addition to being highly vascular, also had significantly more infiltration of F4/80-positive macrophages than was noted in the similarly sized, less vascular tumors in ␤3 KO animals (Fig. 2). Although macrophages are known to contribute to tumor elimination (31), our immunohistochemical analysis could not determine the relevance of the infiltrated F4/80-positive cells to tumor growth, the influence of ␤3 expression on macrophage function, or the origin of these cells.
To begin to answer these questions we performed studies in which immunodeficient female Rag1 KO/␤3 WT (or ␤3 KO) animals were lethally irradiated and transplanted with the bone marrow from genetically matched ␤3 WT or ␤3 KO male animals. Four weeks later U251 GBM cells were intracranially implanted into animals with reconstituted bone marrow, after which effects of ␤3 expression in the transferred bone marrow on tumor growth and vascularity were monitored. As shown in Fig. 3A, the size of intracranial tumors in ␤3 WT animals that did not undergo irradiation/BMT (bar 1) was no different from that in ␤3 WT animals that had their ␤3WT bone marrow replaced with that from a matched ␤3 WT animal (bar 3). Consistent with data in Fig. 1C, these tumors were also of a similar size to those that grew in ␤3 KO animals, which received either no irradiation/BMT (bar 2) or were transplanted with ␤3 KO marrow (bar 6). The transplantation of ␤3 WT marrow into ␤3-deficient mice, however, greatly suppressed the growth of intracranially implanted cells (bar 5), whereas the replacement of ␤3 WT marrow with ␤3-deficient marrow greatly enhanced tumor growth (bar 4). These results suggest that in addition to exerting a pro-angiogenic, pro-growth effect, host ␤3 expression also exerts a previously unidentified anti-tumor effect mediated by factors transferable in the bone marrow.
Bone Marrow-derived Macrophages Suppress Intracranial Tumor Growth in a ␤3-dependent Manner-The ␤3 integrin status of the host bone marrow could potentially influence intracranial tumor growth in several ways. The function of marrow-derived B and T lymphocytes is known to be dependent on ␣V␤3 expression (32), and loss of ␤3 expression could block B and T lymphocytes function, thereby leading to enhanced glioma growth. All experiments done in this study, however, were done in Rag1 KO animals that lack B and T lymphocytes, eliminating a possible role for these cells.
The bone marrow has also been suggested to contribute cells to the growing vasculature of tumors (33,34), and it is possible that host ␤3 expression could suppress tumor growth by alter-FIGURE 1. Host ␤3 expression suppresses GBM growth and vessel density in the subcutaneous setting but not the intracranial setting. Human U251 GBM (panels A-D) or mouse GL261 GBM (panels E and F) were subcutaneously (panels A and B) or intracranially (panels C-F) implanted into Rag1 KO mice and allowed to grow for 25 days (subcutaneous studies) or 35 days (intracranial studies). After animal sacrifice, tumors were sectioned coronally, measured for volume (panels A, C, and E) and stained with anti vWF antibody. vWFpositive vessels detected by a rhodamine-conjugated secondary antibody were counted in 10 independent fields (HPF) at ϫ400 with the aid of an ocular grid (panels B, D, F, and G). Insets in panel G, DAPI-stained sections indicating the tumor/brain interface. Results are expressed as the means Ϯ S.D. *, p Ͻ 05.  DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 ing neovascularization. The alterations in tumor growth mediated by BMT, however, were unrelated to vessel density, and as shown in Fig. 3B, tumor vessel density was related to ␤3 status of the host but was unrelated to ␤3 status of the host bone marrow. Consistent with this observation, no significant amounts of cells from the bone marrow of male mouse donors (as assessed by Y chromosome in situ hybridization) were found associated with the tumor vasculature (not shown), suggesting that the bone marrow did not contribute in a significant manner to the generation of tumor vasculature or to host ␤3 effects on blood vessel density.

Effects of Host ␣V␤3 Integrin Expression on Gliomas
A third possibility is that bone marrow-derived cells generated even in Rag1 KO mice (macrophages, NK cells, neutrophils) could directly or indirectly suppress intracranial tumor growth. Because more macrophages were found associated with intracranial tumors in ␤3 WT animals than in ␤3 KO animals, we first considered the possibility that ␤3 WT animals produced more tumor-suppressive macrophages than ␤3 KO animals. Examination of levels of circulating macrophages in ␤3 WT and ␤3 KO animals by flow cytometry, however, showed that levels of F4/80-positive cells were not significantly different in ␤3 WT and ␤3 KO animals (not shown). We, therefore, examined the extent of infiltration of F4/80-positive cells into the tumor-bearing hemisphere and the non-tumor-bearing hemisphere of ␤3 WT or ␤3 KO animals with or without BMT. In all animals examined, no F4/80-positive or bone marrowderived Y chromosome-positive cells were found in the nontumor-bearing hemisphere (not shown). As shown in Fig. 4, A  and B, however, significant amounts of F4/80-positive cells were found in the tumor-bearing hemisphere of ␤3 WT mice that received ␤3 WT bone marrow, and Ͼ80% of these DAPI- (top left, Fig. 4C) and F4/80-positive (top right, Fig. 4C) cells were also Y chromosome-positive (bottom left, Fig. 4C) and, therefore, bone marrow-derived (merge photo, bottom right, Fig. 4C). The number of F4/80-positive, Y chromosomepositive cells in the tumor-bearing hemisphere of ␤3 WT mice with ␤3 WT marrow was no different from that in ␤3 KO mice that received ␤3 WT marrow but was significantly greater than that in ␤3 WT or ␤3 KO animals with ␤3 KO marrow (Fig. 4, A and B). The apparent inability of ␤3-deficient macrophages to reach the tumor was mirrored by alterations in the in vitro migration ability of these cells. In these studies HDMEC were grown to confluence on laminin-precoated polycarbonate filters, after which macrophages isolated from ␤3 WT or ␤3 KO mice were seeded on top of the monolayer and assayed for their ability to transmigrate. As shown in Fig. 4D, ␤3 KO macrophages had less transmigration ability than ␤3 WT macrophages, and only the migratory ability of the ␤3 WT macrophages was inhibited by preincubation with anti-␤3 integrin antibodies (Fig. 4D).
To further verify that ␤3 WT, bone marrow-derived macrophages were responsible for the glioma growth suppression noted in ␤3 WT animals, we used a macrophage-specific antibody to selectively deplete animals of macrophages, after which the effects on glioma growth were monitored. Treatment of ␤3 WT or ␤3 KO animals with a macrophage-selective antibody resulted in a Ͼ90% depletion of circulating macrophages in both groups (relative to circulating macrophage levels in animals treated with an IgG control) and nearly eliminated macrophage infiltration into tumors in both groups (bottom panel, Fig. 5) while not altering the level of circulating NK cells (not shown). Although this macrophage depletion did not significantly effect the growth of U251 cells intracranially implanted into ␤3 KO animals (Fig. 5, far right bars), macrophage depletion significantly increased the growth of U251 cells intracranially implanted into ␤3 WT animals (Fig. 5, first and second  bars). As a whole, these results suggest that macrophages derived from ␤3 KO bone marrow can circulate in the bloodstream but appear to be defective in their ability to cross out of the vasculature and into the tumor and to suppress the growth of intracranially implanted GBM cells.
␤3 WT Macrophages Secrete TNF␣ and Are Associated with Tumor Apoptosis-Although ␤3 WT macrophages migrated to intracranial tumors more effectively than ␤3 KO macrophages, the means by which they suppressed tumor growth once at the site remained unclear. Macrophage infiltration did not appear to slow the growth of tumor cells in a general fashion because the relatively smaller tumors that grew in presence of infiltrated ␤3 WT macrophages had a tumor growth fraction (Ki67 label- ing index) no different from that of relatively larger tumors that grew in the absence of ␤3 WT macrophages (not shown). We, therefore, examined tumors from ␤3 WT or ␤3 KO animals bearing ␤3 WT or ␤3 KO bone marrow for extent of apoptosis. As shown in Fig. 6A, the relatively larger tumors that formed in animals with ␤3 KO bone marrow (␤3 WT or ␤3KO hosts transplanted with ␤3 KO bone marrow) were devoid of TUNEL-positive apoptotic cells. In contrast, a significant number of TUNEL-positive apoptotic cells was present within the relatively smaller tumors that formed in animals with ␤3 WT bone marrow (␤3 WT or ␤3KO hosts transplanted with ␤3 WT bone marrow). Apoptotic cells were particularly apparent at the invading edges of the tumor where the largest amount of bone marrow-derived F4/80-postive macrophages were also seen. Because macrophages can induce apoptosis by secretion of TNF␣ (31), we also immunohistochemically examined tumors from ␤3 WT and ␤3 KO animals bearing ␤3 WT or ␤3 KO bone marrow for the presence of TNF␣-positive, F4/80-positive macrophages. As shown in Fig. 6B, although scattered F4/80-positive macrophages were found in the relatively larger tumors that formed in animals with ␤3 KO bone marrow, no double-positive TNF␣-secreting macrophages were noted. Significant numbers of TNF␣-secreting macrophages, however, were found in the invading, apoptotic regions of the relatively smaller tumors that formed in animals with ␤3 WT bone marrow. The apparent inability of macrophages from ␤3 KO animals to induce apoptosis was also mirrored by alterations in their in vitro TNF␣ secretion. In these studies macrophages isolated from ␤3 WT or ␤3 KO mice were cultured and allowed to secrete TNF␣ into the supernatant, after which the supernatants were added to U251 cells, and the extent of TNF␣ secretion and supernatant-induced U251 cell apoptosis (% of cells with a sub-G 1 DNA content) were determined. As shown in the top panel of Fig. 7, ␤3 KO macrophages expressed less TNF␣ than ␤3 WT macrophages. The supernatant from ␤3 KO macrophages was also significantly less able to induce apoptosis in target U251 cells Results are expressed as the means Ϯ S.D. * and **, p Ͻ 05. Panel D, for in vitro transendothelial migration assays, freshly isolated macrophages from ␤3 WT or ␤3 KO animals (10 4 cells) were preincubated with control or anti-␤3 integrin antibody (Ab) (50 mg/ml), seeded onto transwell plates covered with HDMEC cells, and measured for their ability to cross the endothelial cell layer in a 5-h period. ␤3 KO macrophages had less transmigration ability than ␤3 WT macrophages, and the migratory ability of the ␤3 WT macrophages was inhibited by preincubation with anti-␤3 integrin antibodies. Results are expressed as the means Ϯ S.D. * and **, p Ͻ 05. than similar supernatant from ␤3 WT macrophages (bottom panel, Fig. 7).
These results as a whole suggest that ␤3 WT macrophages have a greater ability to migrate to the tumor, secrete TNF␣, and to participate in the apoptotic cell death in the tumor. These tumor-suppressive effects in turn appear to counter the enhanced vascular density stimulated by ␤3-expressing, nonbone marrow cells and to explain the different effects of host ␤3 expression in the intracranial and subcutaneous settings.

DISCUSSION
Integrins have long been suspected of playing a contributory role to the vascularization and growth of glioma (10,12). The overexpression of ␣V␤3 in glioma cells and the associated vasculature as well the ability of ␣V␤3 antagonists to block glioma growth has further supported the idea that ␣V␤3 integrin complexes are critical therapeutic targets in the treatment of gliomas. The recent demonstration that tumor-specific ␤3 overexpression suppresses rather than enhances the growth of some glioma (14) as well as the demonstration that host ␤3 expression suppresses the growth of subcutaneous tumors (15) have prompted a re-examination of the role host ␣V␤3 integrins may play in glioma. In the present study we have shown that host ␣V␤3 integrin expression does play a role in glioma growth and angiogenesis but that this role is highly dependent on the microenvironment. We also show that host ␤3 expression has a variety of effects on glioma behavior in the intracranial setting, enhancing both tumor vascularity and growth while at the same time enhancing the ability of macrophages to extravasate, secrete TNF␣, and participate in the suppression of glioma growth. These results suggest a more complex view of the function of host ␣V␤3 expression than previously described.
Although host ␤3 expression has been shown to play a critical and negative role in the angiogenesis and growth of a variety of implanted tumor cell types (14,15), all published studies to date have been performed in the subcutaneous setting, leaving FIGURE 5. Macrophage depletion in vivo enhances GBM growth in ␤3 WT mice but not in ␤3 KO mice. Anti-macrophage antibody or control IgG was intraperitoneally injected into ␤3 WT or ␤3 KO female animals, and 2 days later U251 GBM cells were intracranially implanted into the animals. Antibody injection was repeated at days 0 (same day as implantation), 2, 5, and 7 and twice a week thereafter until animal sacrifice at 35 days post-tumor implantation. Animals were then sacrificed, and brains were sectioned and fixed. Sections were then either stained with hematoxylin and eosin or with anti-F4/80 antibody. The ϫ400 magnifications of brown-staining macrophages shown in the bottom panels are representative of three experiments. Tumor volume was also measured and expressed as a percentage of control (19 Ϯ 4 mm 3 , control IgG). Macrophage depletion significantly enhanced the growth of U251 cells intracranially implanted into ␤3 WT animals but not ␤3 KO animals. Results are expressed as the means Ϯ S.D. *, p Ͻ 05. open the question of how microenvironment influences integrin function. The data presented in this study both confirm previous work as well as highlight the effect of microenvironment on integrin ␤3 function. The present study shows that subcutaneously implanted GBM cells, like previously examined subcutaneously implanted melanoma and lung carcinoma cells (15), exhibit less tumor growth and less angiogenesis in ␤3 WT hosts than in ␤3 KO hosts. Surprisingly, however, the previously demonstrated anti-angiogenic effects of host ␤3 expression were reversed in the intracranial setting. The anti-angiogenic effects of host ␤3 expression in the subcutaneous setting were ascribed to up-regulation of VEGF receptor 2 that occurred in a compensatory fashion in ␤3 KO host endothelial cells but not in ␤3 WT endothelial cells (35). In supplemental data, we confirmed that expression of VEGF receptor 2 is significantly higher in several tissues of ␤3 KO animals than in the same tissues from ␤3 WT animals. The brain, however, exhibited very low levels of VEGF receptor 2 expression, and these levels were not influenced by ␤3 expression. It is possible, therefore, that although endothelial cells in most microenvironments respond to the loss of ␤3 expression by up-regulating VEGF receptor 2 and increasing proliferation and blood vessel formation, brain endothelial cells are incapable of doing so and can, therefore, not initiate angiogenesis in response to loss of ␤3 expression. If this is the case, however, additional mechanisms must exist by which ␤3 expression rather than loss of ␤3 also enhances intracranial tumor vessel density.
In contrast to the tissue-specific differences in ␤3 effects on angiogenesis, the present studies suggest that ␤3 expression provides a microenvironment-independent, suppressive action on tumor growth. As in the subcutaneous setting, host ␤3 expression suppressed the growth of intracranial tumors, although this effect was masked by the pro-tumorigenic effects of host ␤3 expression on angiogenesis. Studies in the subcutaneous setting suggested that the anti-tumor effects of host ␤3 expression were mediated by bone marrow-derived cells (15), although the identity of these cells was not demonstrated. In the present study we clearly show that the ability of host ␤3 expression to suppress intracranial tumor growth was associated with the bone marrow and specifically with macrophages, which in ␤3 WT animals were able to extravasate and migrate into the tumor, where their presence was linked to TNF␣ secretion and tumor cell apoptosis. The function of ␣V␤3 in macrophages is not well defined, although integrins appear to be involved not only in the cellcell contact required for macrophage extravasation and migration but also for the recognition and killing of tumor cells (30,31). The reduced ability of ␤3 KO macrophages to cross a layer of endothelial cells in vitro as well as their inability to localize to the tumor in vivo suggests that the loss of ␤3 expression interferes with macrophage function at several levels although not with the production of the circulating macrophages themselves. Finally, although macrophage function was linked to the effects of ␤3 on tumor growth and macrophage depletion studies were shown to selectively deplete macrophages and not NK cells, the present studies cannot rule out the possibility that other bone marrow-derived cells such as neutrophils may also contribute to host ␤3 effects of glioma growth and angiogenesis. More detailed studies of the effects of these cells are likely to be of value.
Although the present work helps define the role of host ␤3 expression in the control of tumor growth and angiogenesis, it also has implications for tumor therapy. In the subcutaneous setting, ␤3 expression appears to contribute only to tumor suppression, and loss of ␤3 expression appears to only stimulate tumor growth and angiogenesis. In this type of setting there would appear to be little therapeutic value for a true ␤3 antagonist, although molecules that function as ␤3 agonists, as has been suggested for existing integrin-targeted therapeutics (36), would be predicted to be of value. In the more complex intracranial glioma setting, host ␤3 expression has a more mixed function. In the intracranial setting, ␤3 agonists could stimulate the elimination of the tumor by the innate immune system, although this might come at the cost of increased angiogenesis and growth of the tumor. ␤3 antagonists might FIGURE 7. ␤3 WT macrophages preferentially secrete TNF␣ and induce apoptosis in glioma cells. Macrophages isolated from ␤3 WT or KO mice were activated by 24-h incubation with lipopolysaccharide (LPS, 100 ng/ml), after which the supernatants were collected, analyzed for TNF␣ secretion by Western blot (top panel), or added to U251 cultures. U251 cell apoptosis was then analyzed by fluorescence-activated cell sorter and compared with that noted in control cells incubated with supernatant from unstimulated ␤3 KO macrophages, with cells exhibiting a sub-G 1 DNA content considered to be apoptotic (bottom panels). Results are expressed as the means Ϯ S.D. *, p Ͻ 05. DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 block tumor angiogenesis but would also suppress tumor elimination. It seems likely that appropriate use of well defined integrin targeting agents could minimize the growth-enhancing effects of the compounds while enhancing the growth suppressive effects of integrin expression. The identification of the effect of host ␤3 expression in the intracranial setting is a first step in this direction.