A 1 Adenosine Receptors in Microglia Control Glioblastoma-Host Interaction

We report that experimental glioblastoma grow more vigorously in A1 adenosine receptor (A1AR)–deficient mice associated with a strong accumulation of microglial cells at and around the tumors. A1ARs were prominently expressed in microglia associated with tumor cells as revealed with immunocytochemistry but low in microglia in the unaffected brain tissue. The A1AR could also be detected on microglia from human glioblastoma resections. To study functional interactions between tumor and host cells, we studied glioblastoma growth in organotypical brain slice cultures. A1AR agonists suppressed tumor growth. When, however, microglial cells were depleted from the slices, the agonists even stimulated tumor growth. Thus, adenosine attenuates glioblastoma growth acting via A1AR in microglia. (Cancer Res 2006; 66(17): 8550-7)


Introduction
Adenosine is a regulatory nucleoside that is generated in response to cellular stress and damage and is therefore increased during episodes of tissue hypoxia and inflammation.Adenosine is a ubiquitous biological mediator with the capacity to produce both proinflammatory and anti-inflammatory effects in tissues and acts through four distinct cell membrane receptors (A 1 AR, A 2A AR, A 2B AR, and A 3 AR) each with varying ligand affinities, tissue distributions, and signal transduction mechanisms.The A 1 AR has a high affinity for adenosine and has been implicated in both proinflammatory and anti-inflammatory aspects of disease processes.On one side, A 1 AR signaling can promote neutrophil (1) and monocyte (2,3) activation, whereas, on the other side, A 1 AR signaling is involved in anti-inflammatory and protective pathways in neuroinflammation and injury (4) and in cardiac (5) and renal (6,7) injury.Adenosine-mediated anti-inflammatory effects have been studied extensively in macrophages and macrophage cell lines.Adenosine inhibits the production of several proinflammatory cytokines [transforming growth factor-a, interleukin (IL)-6, and IL-8] by lipopolysaccharide-stimulated macrophages and enhances the release of the anti-inflammatory cytokine .Recent studies suggest an anti-inflammatory role for chronic A 1 AR activation by high levels of adenosine in the lung, a surprising and important finding in light of the fact that A 1 AR antagonists are being investigated as a potential treatment for asthma (11).
In the central nervous system (CNS), the A 1 AR is highly expressed on microglia/macrophages and neurons (12).In the latter, A 1 AR is coupled to activation of K + channels (13) and inhibition of Ca 2+ channels (14), both mechanisms that attenuate neuronal excitability.By reducing excitotoxicity, adenosine can act as a neuroprotective factor.Because A 1 adenosine receptors (A 1 AR) are expressed throughout the brain (15), adenosine has the potential to be involved in different brain pathologies.Although A 1 ARs modulate physiologic functions, A 1 AR-deficient mice show no obvious abnormal behavior (16,17).However, exposed to pathophysiologic conditions, such as hypoxia, A 1 AR-deficient mice show more neuronal damage and have a lower survival rate.It was therefore concluded that A 1 ARs are primarily important in mediating effects of adenosine during pathophysiologic conditions (16,17).In the present study, we have addressed the question whether A 1 ARs are involved in another brain disease (i.e., the development of primary brain tumors).We have, in particular, focused on the interaction of the tumor cells with the intrinsic immune cells of the brain, the microglial cells.
Microglial cell cultures.Microglial cultures were prepared from cerebral cortex of newborn C57BL/6 mice as described previously (23).
In brief, the forebrain was carefully freed of blood vessels and meninges.Cortical tissue was trypsinized for 2 minutes, dissociated with a firepolished pipette, and washed twice.Mixed glial cells were cultured for 9 to 12 days in DMEM supplemented with 10% FCS, 2 mmol/L L-glutamine, 100 units/mL penicillin, and 100 Ag/mL streptomycin with medium changes every third day.Microglial cells were then separated from the underlying astrocytic layer by gentle shaking of the flask for 1 hour at 37jC in a shakerincubator (100 rpm).The cells were seeded in six-well plates at a density 10 6 /well.Cultures usually contained >95% microglial cells, which could be stained with Griffonia simplicifolia isolectin B4 (Sigma-Aldrich, Deisenhofen, Germany), a marker for microglia.Cultures were used for experiments 1 to 5 days after plating.Cell medium and supplements were purchased from Seromed/Biochrom (Berlin, Germany).
Stable transfection.Gl261 and F98 glioblastoma cells were either transfected with the pEGFP-N1 vector or with the pdsred2-N1 vector both purchased from Clontech (Heidelberg, Germany) using LipofectAMINE 2000 transfection according to the manufacturer's protocols (Invitrogen).
Immunofluorescence.Tissue from six animals was used per immunohistochemical experiment (i.e., for each immunohistochemical marker).Three of these animals belonged to the control group (control treatment or wild-type genetic background) and three mice were either treated with pharmacologic compounds or had a deletion in the gene coding for the A 1 AR.All staining was done on 40-Am free-floating sections and immunofluorescent triple labeling was done as described previously (19) with the following primary antibodies: polyclonal mouse, rat, and human anti-A 1 AR (Chemicon, Temecula, CA); guinea pig anti-GFAP (Advanced ImmunoChemical, Long Beach, CA); rabbit anti-GFP (Abcam, Cambridgeshire, United Kingdom); goat anti-GFP (DPC, Bad Nauheim, Germany); mouse anti-dsred (Abcam); rabbit anti-S100h (Swant, Bellinzona, Switzerland); mouse anti-rat CD11b (Serotec, Oxford, United Kingdom); monoclonal mouse anti-O4 (Chemicon); and polyclonal mouse anti-Iba-1 (Chemicon).FITC-, rhodamine X-, or Cy5-conjugated were used as secondary antibodies (all from Jackson ImmunoResearch Laboratories, West Grove, PA).Fluorescent sections were mounted in polyvinyl alcohol with diazabicyclo-octane as an antifading agent.
Confocal microscopy.All confocal microscopy was done using a spectral confocal microscope (Leica, Nussloch, Germany).Appropriate gain and black level settings were determined on control tissues stained with secondary antibodies alone.Overview images were processed with Photoshop version CS (Adobe, San Jose, CA), and colocalization images were processed with Volocity version 2.6.1 (Volocity, Lexington, MA).
Quantification of Iba-1 immunoreactivity.In a blinded study, we did cell counts of Iba-1-immunopositive cells.Three randomly chosen fields in the tumor border of the respective staining were analyzed.Quantitative analysis of Iba-1-positive cells per square millimeter was done using Image-Pro version 5 (Media Cybernetics, Silver Spring, MD).
Organotypical brain slice model.Organotypical brain slice culture preparation was done as described previously by us (24).Brain tissue was derived from 16-day-old male C57BL/6 mice (Animal Breeding Facility, Scho ¨nwalde, Germany).For organotypical brain slice preparations, mice were decapitated and the brain was removed within 2 to 3 minutes and placed in ice-cold (4jC) PBS under sterile conditions.The forebrain was dissected from the brainstem and was glued (cyanoacrylate glue) onto a glass block and cut in the coronal plane into 250-Am sections with a vibratome (VT1000S, Leica Co., Heidelberg, Germany).The brain slices were transferred onto the 0.4-Am polycarbonate membrane in the upper chamber of a Transwell tissue insert (Falcon model 3090, Becton Dickinson, Lincoln Park, NJ), which was inserted into a six-well plate (Falcon model 3502, Becton Dickinson).Thereafter, the brain slices were incubated in 1 mL culture medium per well containing DMEM (Life Technologies, Gaithersburg, MD) supplemented with 10% heat-inactivated fetal bovine serum (Atlanta Biological, Atlanta, GA), 0.2 mmol/L L-glutamine, 100 units/mL penicillin, and 100 Ag/mL streptomycin (medium 1).After overnight equilibration of the brain slices in medium 1, this was exchanged for cultivation medium (medium 2).Medium 2 (100 mL) contained 25 mL heat-inactivated horse serum, 25% of 580 AL bicarbonate (7.5%), 2 mL of L-glutamine solution, 2% of 25 mL HBSS, 2.46 mg/mL of 100 AL insulin (all from Life Technologies), 1.2% glucose (20%, Braun Melsungen AG, Germany), 1 mg/mL of 80 Ag vitamin C (Sigma-Aldrich), 0.8 AL/mL of 1 mL penicillin/streptomycin (Sigma-Aldrich), and 500 AL of 1 mol/L Tris in DMEM.Adenosine and N 6 -cyclopentyladenosine (CPA) were added daily.
Inoculation of Gl261 and F98 cells into rats and mice.Anesthetized CD344 Fisher rats, nestin-GFP, GFAP-GFP, CX3CR1-GFP, A 1 AR wild-type, A 1 AR knockout, and wild-type C57BL/6 mice were immobilized and mounted into a stereotactic head holder (David Kopf Instruments, Tujunga, CA) in the flat-skull position.Approximately 1.5 mm anterior and 1.5 mm lateral to bregma a 1 AL of 30-gauge gas-tight syringe (Hamilton, Reno, NV) was then inserted to a depth of 4 mm and retracted to a depth of 3 mm from the dural surface, and 1 AL GFP-transfected F98, GFP-transfected Gl261, or dsred-transfected Gl261 cell suspension (10 4 cells in 1 AL serumfree DMEM) were injected.Animals were examined daily for alertness, focal motor deficit, gait disturbance, and responses to contact.Intraoperative or postoperative complications were not observed.
Tumor cell injection into organotypical brain slices.About 10 4 GFPtransfected Gl261 tumor cells within a defined injection volume (0.1 AL) were inoculated into the slices using a syringe mounted to a micromanipulator.An injection canal was formed, reaching 150 Am deep into the 250-Am-thick slice.Then, the needle was retracted by 50 Am, leaving an injection cavity of f50 Am.The cell suspension was slowly injected for 3 minutes; subsequently, the syringe was slowly pulled out in 10-Am incremental steps.To ensure identical experimental conditions, the glioblastomas were always inoculated into the same area.Directly after glioblastoma injection, the tumor cells remained at the inoculation site, which could therefore mark the point of origin for all further movements of these cells.Careful control of the injection procedure ensured that no cells were spilled onto the surface of the slices to avoid migration of cells on the surface rather than through the tissue.
Preparation of clodronate-filled liposomes.Clodronate-loaded liposomes were obtained from GOT Therapeutics (Berlin, Germany) and the Department of Molecular Cell Biology, Free University of Amsterdam and were prepared as described previously by us (24).All liposomes were passed through a 12-Am filter immediately before use to eliminate large lipid aggregates (25).
Intraoperative specimens.Five specimens of neuropathologically confirmed glioblastomas were obtained during planned neurosurgical excisions.The specimens were frozen in liquid nitrogen immediately after excision and stored at À80jC.Sections were immunostained using the above-mentioned primary and secondary antibodies.Fluorescent sections were mounted in polyvinyl alcohol with diazabicyclo-octane as antifading agent.
Quantification of tumor volume.An examiner who was unaware of the slide identity did quantification of tumor volume according to the Cavalieri principle by determining the tumor area in every sixth 40-Am brain slice and then multiplying by 6 Â 40 Am using Image-Pro version 5.
Gelatin zymography.Activity of gelatinases [matrix metalloproteinase (MMP)-2] was analyzed with the gelatin zymography procedure adapted from that of Heussen and Dowdle (26).Gl261 cells were grown as described above and conditioned medium was taken after 24 hours of culturing.Cultivated microglia (culture method described above) were exposed to unconditioned medium (controls) or stimulated with the glioma conditioned medium containing either 100 Amol/L adenosine or conditioned medium alone for 6 hours.Subsequently, after a brief wash of the microglia, cultures were maintained for a further 24 hours in serum-free culture medium.These microglia conditioned media were mixed with sample buffer [0.5 mol/L Tris (pH 6.8), 20% glycerol, 4% SDS, 0.1% Triton X-100] and the samples were loaded into 7.5% SDS gel containing 1% gelatin.After electrophoresis (4jC, 90 V), gel was washed 2 Â 15 minutes in 2.5% Triton X-100 washing solution and incubated overnight in developing buffer [50 mmol/L Tris (pH 7.6), 10 mmol/L CaCl 2 , 50 mmol/L NaCl, 0.05% Brij35].Next, the gel was stained for 30 minutes in 0.5% Coomassie blue solution and later destained with 40% methanol and 10% acetic acid.Gelatinase activity resulted in clear bands on dark blue background and this activity could be quantified.Gelatinase types were identified by molecular weight markers.
Statistical analysis.Data are expressed as mean F SD. Results were analyzed by Student's t test for statistical significance after testing for normal distribution done with SPSS version 11.0 (SPSS, Inc., Chicago, IL).Differences with P < 0.05 and P < 0.01 were considered significant.

Results
A 1 AR deficiency in the host brain promotes the growth of inoculated glioblastoma cells.To study the importance of adenosine receptors for tumor growth in the central nervous system, we inoculated tumor cells into in A 1 AR À/À mice and A 1 AR +/+ littermate controls.With this approach, we deleted the A 1 AR in the host cells but not in the inoculated Gl261 glioblastoma cells.Animals were sacrificed 14 days after Gl261 inoculation and the tumor volume was determined double-blinded in axial section according to the Cavalieri principle.The tumor volume in A 1 AR À/À mice was significantly larger compared with A 1 AR +/+ mice [mean F SD, 2.96 F 0.48 mm 3 for control (n = 14) and 5.21 F 0.53 mm 3 for A 1 AR À/À mice (n = 12); P < 0.01; Fig. 1B].There were no differences in neurologic symptoms (levels of alertness, behavior, and appearance of focal neurologic deficits; e.g., epileptic fits or pareses) between the groups within the 14 days after Gl261 inoculation.
To analyze the cell populations from the host in the vicinity of the tumor cells, we studied the distribution of microglial cells and astrocytes in A 1 AR À/À and A 1 AR +/+ mice.Immunoreactivity for the macrophage/microglia marker Iba-1 revealed an accumulation of Iba-1-positive cells at the tumor border (Fig. 1C and D).In A 1 AR À/À mice, the density and number of Iba-1-positive cells was significantly higher compared with wild-type littermates [mean F SD, 291 F 59 cells/mm 2 for A 1 AR +/+ (n = 15) and 547 F 123 cells/mm 2 for A 1 AR À/À mice (n = 15); P < 0.01; Fig. 1C].No differences in the GFAP-positive cell population was observed comparing A 1 AR À/À and A 1 AR +/+ (data not shown).These results imply that A 1 AR modulate tumor growth and that microglial cells are the cellular candidates to mediate this effect.
Figure 1.Gl261 glioblastoma growth in A 1 AR +/+ and A 1 AR À/À animals.A, phase-contrast images corresponding to confocal images in (B-D ) displaying the brain morphology 14 days after tumor inoculation into a A 1 AR +/+ and A 1 AR À/À mouse.White dashed lines, outlines of the tumor; arrows, images (B-D ).B, quantification of Gl261 glioblastoma volumes in A 1 AR +/+ and A 1 AR À/À animals revealed significant larger tumor volumes in knockout animals compared with controls 14 days postoperative (P < 0.01).C, quantification of microglial/ macrophage immunoreactivity showed a higher mean number of immunopositive cells in Gl261-bearing A 1 AR À/À animals compared with Gl261-bearing A 1 AR +/+ animals 14 days postoperative (P < 0.01).D, Iba-1 immunoreactivity was obviously greater in glioblastoma-bearing A 1 AR À/À animals compared with glioblastoma-bearing A 1 AR +/+ animals (control) as illustrated by confocal images.Bar, 100 Am.
A 1 AR are expressed by glioblastoma cells and up-regulated in tumor-associated microglial cells.To study the cell type-specific expression of A 1 AR within glial tumors, we used an specific antibody against A 1 AR in combination with cell type-specific markers.Specificity of the antibody for A 1 AR was investigated by immunocytochemistry on brain tissue from wild-type and A 1 AR À/À mice (see Supplementary Fig. S1).The inoculated tumor cells, either F98 or Gl261 cells, were labeled by stable transfection with enhanced GFP (EGFP) or dsred, brain macrophages/microglial cells with antibody against CD11b, and astrocytes by using a transgenic animal with GFAP promoter-driven EGFP expression.About half of the F98 glioma cells inoculated into Fisher CD344 rats were immunopositive for A 1 AR, indicating that tumor cells proper express A 1 AR (Fig. 2B  and C).Similarly, 60% of the Gl261 cells inoculated into C57BL/6 mice were immunopositive for A 1 AR.Close to and within the tumor, almost all (90%) of the CD11b + cells were immunolabeled for A 1 AR (Fig. 2B and C).In contrast, immunoreactivity of A 1 AR was low in the contralateral hemisphere (data not shown), indicating that microglia associated with the glioblastoma cells increase A 1 AR expression.As a second approach to study A 1 AR expression in microglia, we injected dsred-labeled Gl261 glioblastoma cells into transgenic C57BL/6 mice expressing GFP under the control of the CX3CR1 promoter, thus specifically labeling microglia cells (Fig. 3A).Similar as observed in the rat model, microglia cells in the vicinity of the tumor were immunolabeled for A 1 AR expression (Fig. 3B and C).In contrast, immunoreactivity of A 1 AR was low on the contralateral site (data not shown).
Astrocytes were not immunolabeled for A 1 AR expression close to the tumor zone (Fig. 4A).Because nestin-positive progenitor cells have recently been shown to be attracted by glioblastomas (18), we studied tumors in transgenic animals expressing GFP under the control of the nestin promoter.These nestin-positive progenitor cells were not immunolabeled for A 1 AR (Fig. 4B).In conclusion, A 1 AR are expressed by microglial cells associated with glioblastoma cells and by glioblastoma cells proper.
Human glioblastoma samples exhibit expression of A 1 AR in microglia and tumor cells.In tumor samples from glioblastoma patients (n = 5), A 1 AR-positive cells were abundantly detected (Fig. 4C).Isolectin B4 as a marker for microglial cells was colabeled with A 1 AR in 90% of the cells from the human glioblastoma probes (Fig. 4C).Only in a few A 1 AR-positive cells (<5%) from human glioblastoma probes, we detected coexpression of markers for the oligodendrocyte lineage, such as the glycopeptide O4 (data not shown), for the neuronal lineage marker Neu-N (data not shown), and for the endothelial marker von Willebrand factor (data not shown).GFAP is a classic marker for astrocytes and putative resident stem cells but also labels glioblastoma cells.In the present study, GFAP was found in 15% of A 1 AR-positive cells (data not shown).S100h as a marker for mature astrocytes was not coexpressed with A 1 AR.
Modulation of A 1 AR activity influences glioblastoma growth in organotypical brain slice cultures.To test the functional effect of A 1 AR activity on glioblastoma growth, we employed an organotypical slice model where we injected glioma cells (24) Figure 2. Glioblastoma cells and microglia are the main source of A 1 AR expression within glioblastoma host interaction.F98-GFP cells induced glioblastomas (green ) in the caudate putamen of P50 Fisher CD344 rats, which were examined for cell type-specific markers (red and blue ) 14 days postoperatively.Colocalization of these markers is shown in the micrographs (arrows ) and by three-dimensional reconstructions of a single cells.A, light micrograph of coronal sections through rat brains bearing GFP-expressing glioblastoma give representative example of tumor size and location.B, three-dimensional reconstruction of a Z series through the (A ) labeled region (arrow ) along the Z-X axis (right narrow ) and Z-Y axis (top narrow ), which confirm that both F98-GFP (green ) and microglia/macrophages (identified by CD11b labeling; blue ) indeed colocalize with A 1 AR in this cell, which appears yellowish in F98-GFP cells and lilac in CD11b-labeled microglia.Note that almost every CD11b-labeled cells is strong colocalized by A 1 AR and the three-dimensional overview confirms that CD11B/A 1 AR and GFP/A 1 AR are indeed in the same cell.C, three-dimensional reconstruction of 40-Am image stack of F98-GFP cells (green ) and CD11b-labeled microglia/macrophages (blue ), which both coexpress A 1 AR (red).and could stimulate or inhibit adenosine receptors.Brain slices (250 Am thick) were cultured for 4 days and 10 4 GFP-labeled Gl261 tumor cells were injected (suspended in 0.1 AL) into the tissue (Fig. 5A).The tumor size was evaluated by measuring the area occupied by the fluorescently labeled Gl261 cells.Daily addition of 100 Amol/L adenosine to the cultivation medium showed a significant inhibition of tumor size to 44% after 4 days compared with a nontreated control group [mean F SD, 1.06 F 0.16 mm 2 for the control (n = 10) and 0.46 F 0.14 mm 2 in the presence of adenosine (n = 14); Fig. 5B].The specific A 1 AR agonist CPA (100 nmol/L) had a similar effect on tumor growth [mean F SD, 0.72 F 0.09 mm 2 (n = 57) and 1.11 F 0.1 mm 2 for the control (n = 46); P < 0.05; Fig. 5C].In summary, stimulation of adenosine receptors significantly impairs tumor growth.
Inhibition of tumor growth mediated by adenosine receptors depends on the presence of microglia.Endogenous microglia can be selectively depleted from cultured organotypical slices by a 24-hour treatment with clodronate filled liposomes.The other cell types, neurons, oligodendrocytes, and astrocytes are not affected.As reported previously, activated microglia supported glial tumor growth resulting in significant smaller tumors in microgliadepleted slices compared with control slices (24).This served us as an internal control.Tumor cells were injected 3 days after liposome treatment and the size of the tumor bulk was evaluated in both slices with and devoid of microglia.We compared control slices with those receiving a daily treatment with adenosine and the specific A 1 AR agonist CPA.As expected, depletion of endogenous microglial cells in organotypical brain slice cultures leads to a reduction in tumor growth [mean F SD, 1.25 F 0.23 mm 2 (n = 61) for the control with microglia and 0.80 F 0.15 mm 2 for the control devoid of endogenous microglia; P < 0.05; Fig. 5D].Activation of adenosine receptors in brain slice cultures devoid of endogenous microglia resulted in a larger tumor size [mean F SD, 0.8 F 0.15 mm 2 (n = 45) for the microglia-depleted control and 1.95 F 0.15 mm 2 for 100 Amol/L adenosine (n = 37); P < 0.05; Fig. 5D].Similarly, activation of A 1 AR with the specific agonist CPA (100 nmol/L) in microglia-depleted slices also increased tumor size as studied 4 days after inoculation [mean F SD, 0.8 F 0.15 mm 2 (n = 45) for the microglia-depleted control and 1.7 F 0.23 mm 2 (n = 39) for 100 nmol/L CPA; P < 0.05; Fig. 5D].In an additional approach, we evaluate modulation of tumor growth in brain slices of A 1 AR À/À mice comparing with A 1 AR +/+ mice.As expected, tumor growth in organotypical brain slice cultures of A 1 AR À/À mice exceeded significantly tumor growth in A 1 AR +/+ mice [mean F SD, 1.25 F 0.16 mm 2 (n = 8) for the A 1 AR +/+ control and 2.12 F 0.23 mm 2 for the A 1 AR À/À ; P < 0.05; Fig. 6A].Daily addition of either 100 Amol/L adenosine or 100 nmol/L CPA do not influence tumor growth within brain slice cultures of A 1 AR À/À mice.We conclude that that CPA and adenosine specifically act on A 1 AR and that the tumorreducing effect of adenosine requires the presence of microglia.
Microglia increase expression of A 1 AR in coculture with glioblastoma cells.A 1 AR expression immunoreactivity was detected in all cultured, purified microglial cells and in cultured Gl261 and F98 glioblastoma cells.After coculturing Gl261 with microglial cells for 4 days, the A 1 AR immunolabeling in microglial cells was stronger compared with control microglial cells, indicating that the presence of tumor cells up-regulates the expression of A 1 AR in microglia (data not shown).No changes in the degree and pattern of A 1 AR expression immunoreactivity were observed after coculturing Gl261 cells with astrocytes and coculturing microglia with astrocytes (data not shown).
Activation of the A 1 AR prominently decreased the gliomastimulated activity of MMP-2 in microglia.In supernatants of microglia cell cultures, being either stimulated or nonstimulated with glioma conditioned medium, we measured the activity of MMP.Gelatin-degrading MMPs are the main mediators of matrixdegrading activity in gliomas and substantially contribute to glioma cell invasion.In gelatin zymographies, we observed that microglia abundantly release active MMP-2 after stimulation with glioma conditioned medium (Fig. 6B).The glioma-stimulated increase in MMP-2 activity was blunted by costimulation with 100 Amol/L adenosine.

Discussion
Adenosine-impaired tumor growth is mediated by microglia.Our present data indicate that the deletion of functional adenosine receptors, specifically A 1 AR, results in an increase in brain tumor A 1 AR immunoactivity does not exhibit a confine colabeling with reactive astrocytes.B, induced nonlabeled Gl261 glioblastomas are densely enwrapped by neural progenitors cells identified by nestin-GFP labeling.A 1 AR immunoactivity does not exhibit a confine colabeling with neural precursor cells.Bar, 25 Am (A and B).C, specimen of a neuropathologically confirmed glioblastomas was obtained during planned neurosurgical excisions.There is an abundant labeling for A 1 AR in the tumor probe.Colabeling with isolectin B4 as a marker for microglia cells revealed a strict colocalization with A 1 AR immunoreactivity.
growth.This implies that adenosine acting via A 1 AR impairs glioma growth.In the context of glioma, A 1 ARs are prominently expressed by the tumor cells and those microglial cells associated with the glioma cells.In our experimental approach using the A 1 AR-deficient mouse as a tumor host, we highlight the importance of the microglial cells for mediating the A 1 AR effect.This does not exclude that adenosine can also affect the glioma cells directly.The microglial cells were accumulating at the tumor site and this accumulation was even more pronounced in the A 1 AR-deficient mice.The importance of microglial A 1 AR is further supported by our brain slice model where tumor impairment was only observed in the presence of microglial cells.We thereby confirmed the observation that the presence of microglial cells per se is tumor promoting (24).There was no significant change in the population of astrocytes or neural progenitor cells.The latter is of particular interest because we recently showed that neural progenitor cells are attracted to tumors or to gliomas and attenuate tumor growth (18).
Microglia and glioma cells express adenosine receptors.The presence of adenosine receptors has been reported previously on astrocytoma cells (21,23) using an A 1 AR-specific ligand.A recent positron emission tomography study indicates that specific binding sites for A 1 AR ligands are associated with experimental glioblastoma (27).The presence of adenosine receptors on microglia is well established and some functional implications of their activation have become apparent.Cultured rat microglial cells express A 2 ARs, because its specific agonist CGS 21680 triggered the expression of K + channels that are linked to microglial activation (28).In contrast, A 2 AR stimulation in rat microglia triggers the expression of nerve growth factor and its release, thereby exerting a neuroprotective effect (29).Moreover, cyclooxygenase-2 expression in rat microglia is induced by A 2 ARs resulting in the release of prostaglandin (30).Hammarberg et al. (31) provided evidence for functional A 3 ARs in mouse microglia cells, whereas A 1 ARs were not detected in this study.Our results based on our immunocytochemical data, however, indicate that microglial cells express A 1 AR.
The potential source for extracellular adenosine is most likely ATP, which is released activity dependent from presynaptic and postsynaptic terminals of neurons and also from glial cells (32).In the extracellular space, adenosine is generated from ATP after dephosphorylation by specific ectoenzymes (e.g., CD39 and CD73).These ectoenzymes represent a highly organized enzymatic cascade for the regulation of nucleotide-mediated signaling.They control the rate of nucleotide (e.g., ATP) degradation and nucleoside (e. g. adenosine) formation (33).Microglial cells express specific ectonucleotidase isoforms, CD39 and CD73, which are not expressed by any other cell type in the brain.Due to this specific expression, both molecules have served as microglia-specific markers even long before their functional importance has been recognized (34)(35)(36).Figure 6.A, tumor growth in organotypical brain slice cultures of A 1 AR À/À mice exceeded significantly tumor growth in A 1 AR +/+ mice.Daily addition of either 100 Amol/L adenosine or 100 nmol/L CPA does not influence tumor growth within brain slice cultures of A 1 AR À/À mice.B, gelatin zymography for MMP-2 activity was analyzed from supernatant obtained from microglial cultures (left ), from microglial cultures stimulated with Gl261-conditioned medium (middle ), and from microglia cultures mixed with supernatant from Gl261 cultures and stimulated by 100 Amol/L adenosine (right ).Modulation of Gl261-GFP-induced glial tumor growth was evaluated in organotypical brain slice cultures after 4 days.A, fluorescence microscope images illustrate a compact Gl261-GFP tumor in all three planes 1 day after inoculation.Tumor size after 4 days depicts the measurability of tumor bulk size.B, daily treatment of slice cultures with 100 Amol/L adenosine leads to a significant tumor inhibition after 4 days.C, daily treatment of Gl261-GFP-bearing brain slice cultures with specific A 1 AR agonist CPA revealed significant influence on glial tumor growth, indicating a functional role of A 1 AR expression in the process of glioblastoma-host interaction.Bar, 100 Am (A, left) and 250 Am (A, right ).To evaluate the possible role of microglial cells in the A 1 AR-mediated influence on glial tumor growth, we depleted selectively microglia in organotypical brain slice cultures as described.D, depletion of microglia in Gl261-bearing brain slice cultures reversed significantly the observed effect of A 1 AR stimulation in microglia-containing slices (P < 0.05).
A 1 AR deficiency leads to an increased microglial density at gliomas.Our data indicate that loss of A 1 AR leads to an increase of tumor-associated microglia, which can be due to infiltration and/ or proliferation.The role of adenosine for microglial proliferation remains controversial.One study reports that adenosine stimulates the proliferation of microglial cells through a mechanism that involves the simultaneous stimulation of A 1 AR and A 2 AR (37).By contrast, adenosine has been reported to inhibit proliferation of microglial cells: phorbol 12-myristate 13-acetate-stimulated microglial proliferation is reduced following treatment with an A 1 AR agonist (38).Moreover, adenosine receptor stimulation by the A 1 AR can also cause microglial apoptosis (39).
Adenosine levels in the extracellular fluid are lower in human glioma tissue compared with control tissue (i.e., 1.5 and 3 Amol/L, respectively).These values were obtained from human gliomas of high-grade malignancy and measured by brain microdialysis coupled to high-performance liquid chromatography (25).Whether this rather small difference causes the accumulation of microglia close to tumors is speculative.
A 1 AR mediates tissue protection.Recent studies support the idea that adenosine receptors; specifically, the A 1 AR are good targets for drug development in several diseases that affect the CNS (40).A 1 AR deficiency aggravates experimental allergic encephalomyelitis (4) and it has been repeatedly shown that adenosine can protect tissues against the negative consequences of hypoxia or ischemia (41) mainly by acting on A 1 AR.Hence, survival after a hypoxic challenge may be reduced if A 1 ARs are absent or blocked (17).
The tissue protective effect of A 1 AR has been implicated in experimental paradigms using the A 1 AR-deficient mice.In a model of renal ischemia and reperfusion injury, A 1 AR-deficient mice exhibited an increase production of proinflammatory mediators and showed an increased renal injury (6,7).Similarly, in a model of experimental allergic encephalomyelitis, A 1 AR deficiency led to an increased neuroinflammation and demyelination and also augmented axonal injury (4).Both studies concluded that A 1 AR serves anti-inflammatory functions that regulate subsequent tissue damage.Furthermore, MMP-9 and MMP-12 are significantly elevated in A 1 AR-deficient mice (4).Indeed, MMPs play an important role in glioma progression; as we showed recently, expression of MMPs by microglia has an effect on tumor growth (24).
Adenosine-regulated glioma invasion is due to the activity of extracellular proteases.Matrix degradation by MMPs is an important prerequisite for glioblastoma invasion (42).It was shown previously that A 1 AR activation on microglia/macrophages inhibits the production of not only cytokines, such as interleukin-1h, but also MMPs, such as MMP 12 (4).In our present study, we observed that activation of the A 1 AR prominently decreased the gliomastimulated activity of MMP-2 in microglia.Above, we have described that A 1 AR-deficient microglia exhibits increased chemoattraction toward gliomas and that the enhanced accumulation of microglia is associated with increased tumor size.The zymographies suggest that A 1 AR blockade on microglia may facilitate their attraction toward the tumor because these cells have higher matrix-degrading activity and may therefore faster arrive at the lesion.However, this increased matrix degradation by microglia may also create an environment, in which glioma cells can easier invade into the surrounding parenchyma and thereby facilitates tumor growth.
Results from this study show that the A 1 AR plays an antitumorigenic role mediated by microglia cells in the development of glioblastomas.If we better understand how the pathways of A 1 AR signaling modulate glioblastoma development, it may ultimately lead to concepts to reduce the progression of this disease.

Figure 3 .
Figure 3. Verification of A 1 AR expression on microglia cells in CX3CR1-GFP-labeled microglia knock-in mice.Gl261-dsred cells induced glioblastomas (red ) in the caudate putamen of P25 CX3CR1-GFP mice (microglia-labeled CX3CR1-GFP cells; green ), which express A 1 AR cell type-specific marker (blue ) 14 days postoperatively.Colocalization with CX3CR1-GFP cells (microglia ) is shown by three-dimensional reconstructions of single cells.A, phase-contrast image displaying the brain morphology 14 days after tumor inoculation into CX3CR1-GFP mouse.White dashed lines, outlines of the tumor; arrows, confocal images.Fluorescence microscope images in different magnification illustrate the strong association of CX3CR1-GFP-labeled microglia to the Gl261-dsred-induced glioblastoma especially in the tumor margin.B, labeling A 1 AR expression revealed a clear colocalization of CX3CR1-GFPlabeled microglia (green ) and A 1 AR immunoactivity (blue ).C, three-dimensional reconstruction of a Z series through CX3CR1-GFP-positive cell along the Z-X axis (right narrow) and Z-Y axis (top narrow ), which confirms that A 1 AR (blue ) and GFP (green ) are indeed present in the same cell (light blue ).Bar, 75 Am (A ), 25 Am (B), and 7 Am (C ).

Figure 4 .
Figure 4. Distribution of A 1 AR expression in human glioblastoma probe and glioblastoma-induced knock-in mice labeled for GFAP and nestin.To examine the possible role of reactive astrocytes and neural precursor cells in the process of glioblastoma development, Gl261-dsred cells induced glioblastomas (red) in the caudate putamen of P25 GFAP-GFP mice and nonlabeled Gl261-induced glioblastomas in the caudate putamen of P25 nestin-GFP (green ) were evaluated 14 days postoperatively for A 1 AR immunoactivity.Phase-contrast images corresponding to confocal images displaying the brain morphology 14 days after tumor inoculation.White dashed lines, outlines of the tumor; black dashed boxes, confocal images.A, Gl261-dsred-induced glioblastomas (red ) are surrounded by reactive astrocytes identified by GFAP-GFP labeling (green ).A 1 AR immunoactivity does not exhibit a confine colabeling with reactive astrocytes.B, induced nonlabeled Gl261 glioblastomas are densely enwrapped by neural progenitors cells identified by nestin-GFP labeling.A 1 AR immunoactivity does not exhibit a confine colabeling with neural precursor cells.Bar, 25 Am (A and B).C, specimen of a neuropathologically confirmed glioblastomas was obtained during planned neurosurgical excisions.There is an abundant labeling for A 1 AR in the tumor probe.Colabeling with isolectin B4 as a marker for microglia cells revealed a strict colocalization with A 1 AR immunoreactivity.

Figure 5 .
Figure 5.A 1 AR modulation influences glial tumor growth observed in organotypical brain slice cultures.Modulation of Gl261-GFP-induced glial tumor growth was evaluated in organotypical brain slice cultures after 4 days.A, fluorescence microscope images illustrate a compact Gl261-GFP tumor in all three planes 1 day after inoculation.Tumor size after 4 days depicts the measurability of tumor bulk size.B, daily treatment of slice cultures with 100 Amol/L adenosine leads to a significant tumor inhibition after 4 days.C, daily treatment of Gl261-GFP-bearing brain slice cultures with specific A 1 AR agonist CPA revealed significant influence on glial tumor growth, indicating a functional role of A 1 AR expression in the process of glioblastoma-host interaction.Bar, 100 Am (A, left) and 250 Am (A, right ).To evaluate the possible role of microglial cells in the A 1 AR-mediated influence on glial tumor growth, we depleted selectively microglia in organotypical brain slice cultures as described.D, depletion of microglia in Gl261-bearing brain slice cultures reversed significantly the observed effect of A 1 AR stimulation in microglia-containing slices (P < 0.05).