|Home | About | Journals | Submit | Contact Us | Français|
Differential diagnosis of brain tumor types is mainly based on cell morphology and could benefit from additional markers. The cAMP second-messenger system is involved in regulating cell proliferation and differentiation and is conceivably modulated during cancer transformation. The cAMP second-messenger system mainly activates protein kinases, which are in part docked to cytoskeleton, membranes, or organelles by anchoring proteins, forming protein aggregates that are detergent insoluble and not freely diffusible and that are characteristic for each cell type. The intracellular distribution of the detergent-insoluble regulatory subunits (R) of the cAMP-dependent protein kinase has been examined in mouse and rat glioma cells both in vitro and in vivo by immunohistochemistry. In normal rodent brains, the RIIα regulatory subunit is detergent insoluble only in ependymal cells, while in the rest of the brain it is present in soluble form. Immunohistochemistry shows that in both mouse and rat glioma cell lines, RIIα is mainly detergent insoluble. RIIα is localized close to the nucleus, associated with smooth vesicles in the trans-Golgi network area. Both paclitaxel and vinblastine cause a redistribution of RIIα within the cell. Under conditions that increased intracellular cAMP, apoptosis of glioma cells was observed, and it was accompanied by RIIα redistribution. Also in vivo, detergent-insoluble RIIα can be observed in mouse and rat gliomas, where it delineates the border between normal brain tissue and glioma. Therefore, intracellular distribution of detergent-insoluble RIIα can assist in detecting tumor cells within the brain, thus making the histologic diagnosis of brain tumors more accurate, and may represent an additional target for therapy.
Differential diagnosis of brain tumors requires careful histological examination. Up until now, immunohistochemical markers and molecular biology techniques have been used, but the variable and often elusive expression of different markers by the most malignant tumors has hampered the use of a standardized diagnostic method.
Prognosis depends on malignancy grade and is very poor for higher grade gliomas;1 therefore, additional markers that improve the detection of malignant cells could lead to more accurate identification of these cells and easier tumor classification.
The major focus of research in oncogenesis is on alterations of oncogenes or oncosuppressors. The subsequent dysregulation of different signal transduction pathways may be involved in the pathogenesis of many tumors or in regulating tumor growth.
The cAMP second-messenger pathway is present in all eukaryotic cells, where it controls a large variety of vital cellular functions, ranging from ion channel permeability to gene expression. Observations suggest that the cAMP system is involved in regulating differentiation and proliferation of many normal and transformed cell types.2,3 cAMP analogues are known to inhibit proliferation and to induce differentiation of neuroblastoma and glioma cells in culture.4–6 It is noteworthy that in some cases cAMP has opposite effects on different cell types; for example, in PC12 cells and Sertoli cells, it stimulates cell growth,7,8 but in NIH3T3 cells and adipocytes, it inhibits growth.9,10
cAMP acts mainly through cAMP-dependent protein kinases (PKA), consisting of two regulatory subunits which, upon binding two cAMP molecules, release two catalytic subunits. The catalytic subunits phosphorylate many target proteins, thus modifying their biological activity. Four different isoforms of regulatory subunits have been characterized (RIα, RIβ, RIIα, and RIIβ), each one forming homodimers that bind the catalytic subunits.
Apparently, the four regulatory isoforms have similar biochemical characteristics, and, in the absence of cAMP, their only function seems to be the inhibition of catalytic subunits. In contrast, it is known that PKA regulatory isoforms are differentially expressed in many tissues, with RIβ and RIIβ being restricted mainly to the nervous system and some endocrine glands.11 Moreover, regulatory isoforms may be either soluble or selectively bound to subcellular structures (e.g., cytoskeleton, Golgi apparatus) through a large family of isoform-specific anchoring proteins, A-kinase anchor proteins (AKAPs).12 In this way, a different distribution of PKA in subcellular microdomains can be obtained, depending on the AKAPs and regulatory subunits expressed in a cell.
In cancer cells, the expression of cAMP-dependent protein kinases has been investigated, with different and apparently partially conflicting results.
RIα loss of heterozygosity is associated with endocrine tumors in the Carney complex: the decrease in RIα makes more PKA catalytic subunits available, enhancing PKA phosphorylating activity, ultimately leading to tumorigenesis. This finding suggests RIα as a candidate tumor-suppressor gene.13,14
In contrast, in other human cancers, an increase in RIα correlates with more aggressive behavior and worse prognosis, whereas decreased expression of RIα correlates with decreased tumor cell growth.15–17 In prostate carcinoma cells, increase in RIIβ expression leads to inhibition of tumor growth, whereas increase in RIα has opposite effects,15 suggesting different involvement of the PKA regulatory subunits in different cancers. In addition, the development of Ewing sarcoma requires the involvement of cAMP response element binding (CREB) proteins, pointing to a cAMP-mediated pathway.18
In the most aggressive glioma, glioblastoma multiforme, protein kinase C is involved in tumor growth.19 Activation of the cAMP pathway, especially through activation of the RII regulatory subunits, causes differentiation and apoptosis of glioma cells.20 AKAP1 was recently found to be upregulated in human glioblastoma specimens, as was phosphodiesterase 1A, a cAMP-degradating enzyme,21 and extracellular PKA catalytic subunit has been used as a cancer marker.22
Our laboratory described a differential distribution in rodent brain of the four PKA regulatory subunits bound to cytoskeleton and membranes and organelles. Docking at particular cell sites, this detergent-insoluble PKA fraction provides a large amount of kinases readily available in the proximity of target proteins, thus allowing more efficient tuning of intracellular signaling. Different brain-cell types can be characterized by PKA intracellular distribution. Detergent-insoluble RIIβ is present in most neural and glial cells,23 whereas insoluble RIα is neuronal and restricted to some brain nuclei.24 Insoluble RIβ is present only in some neuron types, mainly in olfactory-bulb mitral cells and cerebellar Purkinje cells,25 and insoluble RIIα is localized only on the ependymal cells lining the ventricles.23 Each subunit appears at specific times during development,23 suggesting that a different balance between regulatory isoform expression and intracellular localization can be responsible for different properties of the same cell during the normal developmental process. Therefore, it is possible that a different picture emerges when cells have been transformed, acquiring a cancer cell phenotype.
In this article, the distribution of the detergent-insoluble PKA regulatory isoforms in glioma cells is reported, along with the modulation of this distribution by different agents. Two models were investigated: GL261 mouse glioblastoma cells and rat F98 glioma cells.26–28 The cells were analyzed in vitro and in vivo after inoculation in the brain of syngenic hosts.
Cell culture reagents were from Gibco (Milan, Italy). All other chemicals were from Sigma (Milan, Italy), unless otherwise stated.
Mouse GL261 glioblastoma cells, rat F98 glioma, and human Gli36 glioblastoma cells were grown and maintained in monolayer in Dulbecco modified Eagle’s medium (DMEM)/Ham’s F12 (for GL261 cells) or DMEM (for F98) with streptomycin/penicillin (100 U/ml), 10% fetal bovine serum, at 37°C, 5% CO2. Primary glial cell cultures were obtained from telencephalon of 3-day-old mice and maintained in DMEM supplemented with antibiotics and 20% fetal bovine serum.
For pharmacological treatments, GL261 cells were plated on six-well plates. After 48 h, one of the following substances was added to the culture, maintained in DMEM/F12: 500 μM glutamate (as an apoptosis-inducing drug); two clinically used antiblastic agents, 100 nM paclitaxel, 200 nM vinblastine; a series of chemicals affecting the intracellular cAMP pathway, 10 μM forskolin, 500 μM isobutylmethylxantine (IBMX), 100 μM teophyllin, 2 mM caffeine, 500 μM 8-bromo-cAMP, 500 μM 6-dibutyryl-cAMP, and 12 μM H89.
Cells were incubated for 24 h (72 h for paclitaxel treatment), and they were then scraped and cytocentrifuged 10 min at 6,000 rpm on microscope slides, fixed with methanol, and stained with Weigert’s hematoxylin. A total of 600 cells was counted from each slide, and the percentage of apoptotic cells was calculated. Apoptosis was determined by morphological criteria as a reduction in cell volume and condensation of chromatin or cell-membrane blebbing. At least six control and six treated slides were counted for each treatment. Monovariate analysis of variance was used to test the difference in the percentages of apoptotic cells between control and treated cells.
In vivo studies were approved by the Italian Ministry of Health and conducted according to the European laws on animal experiments and welfare. A total of 8 × 104 GL261 cells was resuspended in 2 μl DMEM/F12 and injected over 5 min in the caudate nucleus of the syngenic C57/Bl6J mice (1 mm anterior and 1.3 mm lateral to the bregma, 2.5 mm below the pial surface), anesthesized with ketamine (75 mg/kg) and xilazine (20 mg/kg). A total of 8 × 106 F98 cells was resuspended in 8 μl DMEM and injected 1 mm anterior and 1.5 mm lateral to the bregma, 3.7 mm below the pial surface, over 15 min in Fischer rats anesthetized as above. Animals were monitored daily for the appearance of neurological signs and for weight loss. At the appearance of neurological signs, the animals were sacrificed with overdose anesthesia (three times the surgical dose). Brains were immediately excised and frozen in liquid nitrogen.
Cells were scraped and resuspended in 40 volumes of phosphate-buffered saline (PBS; 10 mM phosphate, 150 mM NaCl, 1 mM EDTA, Triton X-100 2%, pH 7.6) with protease inhibitor cocktail (Roche, Milan, Italy) and forced 30 times through a 27-gauge needle. They were then centrifuged 15 min at 6,000 rpm. The supernatant was collected and the pellet was washed again with the same volume of PBS and centrifuged as above; the second supernatant was discarded and the pellet resuspended in the same volume of PBS. Pellets were run on a 12% polyacrilamide gel, then blotted on nitrocellulose membrane, blocked 60 min with 2% bovine serum albumin, incubated overnight with primary antibodies, RIIα or RIα (both made in rabbit; Santa Cruz Biotechnology, Santa Cruz, CA, USA; 1:5,000), then incubated for 4 h with horseradish peroxidase-conjugated antirabbit secondary antibody (Sigma; 1:10,000) and revealed through chemiluminescence (Amersham Advanced ECL, Milan, Italy).
Cells were grown on cover slips, washed in PBS, and frozen. Cells were thawed and air dried, incubated for 5 min in 2% Triton X-100 in PBS at 18°C and fixed in 5% formalin supplemented with 1% Triton X-100 for 1 min at 37°C.
Brains were cut on a cryostat. The 24-μm slices were air dried and postfixed using one of the following protocols: (a) 1 h in 5% formalin at 18°C, followed by 30 min in 2% Triton X-100 in PBS or (b) incubation in Triton X-100 2% in PBS at 18°C for 30 min, followed by fixation in 5% formalin and Triton X-100 2% at 37°C for 1 min. Primary antibodies were incubated overnight. The following antibodies were used: RIIα (rabbit, Chemicon, Temecula, CA, USA), 1:80 on brain slices, 1:100 on cells; RIIβ (rabbit; Chemicon), 1:80 on brain slices, 1:100 on cells; RIIα (rabbit; Santa Cruz Biotechnology), 1:200 on brain slices, 1:500 on cells; RIα (rabbit; Santa Cruz Biotechnology), 1:200 on slices, 1:500 on cells; vimentin, 1:80 on slices, 1:250 in cells (mouse; Sigma); glial fibrillary acidic protein (GFAP), 1:200 on slices, 1:500 on cells (mouse; Sigma); neurofilament 200, 1:200 on slices, 1:500 on cells (mouse; Sigma); neurofilament 68, 1:200 on slices, 1:500 on cells (mouse; Sigma); golgin 97 (mouse; Molecular Probes, Eugene, OR, USA), 1:40 on slices, 1:80 on cells. Secondary antibodies were incubated 30 min at 37°C and were used 1:200 on tissue, 1:500 on cells: anti-rabbit IgG Alexafluor 594 (Molecular Probes), antirabbit IgG fluorescein-conjugate (Sigma), antimouse IgG fluorescein-conjugate (Sigma), antimouse IgG Alexafluor 568 (Molecular Probes). Cell nuclei were counterstained with bisbenzimide (Sigma).
Positive and negative controls were always performed in each session and included control slides for positive immunolabeling; that is, mouse brain sections were incubated with each of the above antibodies, for which the pattern of labeling has already been described for the different areas;23–25 as negative controls, the cells and tumors were reacted omitting the primary antibody, and for controlling background staining, the cells and tumors were incubated with normal rabbit serum and normal mouse serum (data not shown).
For colocalization experiments, after immunohistochemistry, cells or tissue were incubated in the presence of 8-thioacetamidofluorescein-cAMP, which allows the visualization of RIα as reported elsewhere.23–25
After immunohistochemistry, sections were counterstained with hematoxylin and eosin to confirm tumor presence, localization, and extension. A Leica epifluorescence microscope (20×, 40×, 100× objectives) was used for immunofluorescence and histology. Images were captured with the resident software at 782 × 582 pixels with a color digital camera, using the same parameters within each experiment.
Micrographs of differential interference contrast or fluorescence images were taken on golgin and RIIα doubly-labeled cells, with an Olympus Camedia 5050 digital camera mounted on an Olympus BX51 microscope, a 60× oil-immersion objective (1.25 numerical aperture). The following filter sets were selected for fluorescence microscopy: 480-nm excitation filter, 500-nm dichroic mirror, and 515-nm bandpass filter for green fluorochromes, or 540-nm excitation filter, 580 nm dichroic mirror, and 620 nm barrier filter for red fluorochromes. Contrast was enhanced by maximum 10% when necessary, with Corel Photo Paint; final figures were prepared and lettered using Corel Draw 12 (Corel Corporation, Ottawa, Canada).
For immunoelectron microscopy, GL261 cells were fixed in 4% paraformaldehyde and embedded in LR White resin (Electron Microscopy Sciences, Hatfield, PA, USA) as previously described.29 Ultrathin sections were processed for immunocytochemistry by using a rabbit anti-RIIα antibody (Santa Cruz Biotechnology; dilution 1:20) revealed by a secondary 12-nm gold-conjugated antibody (Jackson ImmunoResearch, West Grove, PA, USA).29 Double labeling was performed after golgin 97 (1:20) and RIIα (1:20) antibody incubation with 6-nm and 12-nm gold-conjugated antibodies (Jackson ImmunoResearch). As controls, some grids were treated with the incubation mixture without the primary antibody. The specimens were observed in a Philips Morgagni transmission electron microscope equipped with a Megaview III camera for digital image acquisition.
Different antibodies were tested in cells and tissues. The anti-RIIα antibody from Santa Cruz Biotechnology provided very bright histochemical images in glioma cells, which were almost identical to the fainter images provided by anti-RIIα from Chemicon. In Western blots of normal brain extracts, the Santa Cruz anti-RIIα antibody recognized both RIIα and β isoforms. Labeling was not affected by preincubation with 1 μM RIα. In Western blots of cell and normal brain extracts, anti-RIIα and RIIβ antibodies (Chemicon) recognized only one single band of a slightly different molecular weight. Although concurrent incubation with RIα had no effect, excess RIIα abolished RIIα labeling without affecting RIIβ labeling. Similar results were obtained from immunohistochemistry of the brain sections.
Biochemical fractionation of GL261 glioma cells and primary glial culture in supernatant and in a detergent-insoluble pellet showed that both RI and RII regulatory subunits were present in the pellet (Fig. 1A).
GL261 cells were treated with drugs that either increase the concentration of cAMP or inhibit the catalytic activity of the cAMP-dependent kinases to reveal the possible effect of cAMP modification on cell survival. A statistically significant increase in apoptosis was observed, similar to that elicited by two well-known drugs in clinical use, paclitaxel and vinblastine, as already reported for other cancer cell lines (Fig. 1B).
In both GL261 and F98 cells, only RIIα insoluble aggregates were detected by immunocytochemistry under different fixation protocols, using the different anti-RIIα antibodies. Under the same conditions, RIα aggregates were not detected, even though RIα was present in Western blots, presumably because of a more distributed localization that hindered immunohistochemical detection above background.
RIIα immunolabeling (green; Fig. 1C) was present close to the nucleus (blue), in a restricted region on one side, and apparently was organized as a cluster of many small dots. All the cells were labeled, although with different intensities. A small percentage of cells appeared elongated, with large processes, indicating a possible differentiation of the cell: in these cells, RIIα labeling appeared fainter, although structured in the same way inside the cell.
RIIα appeared to be clearly separated from vimentinlabeled filaments (Fig. 1D), suggesting that RIIα is not associated with this intermediate filament protein. When cells were labeled with anti-golgin 97, a trans-Golgi network marker, RIIα was associated in the same area, although it did not exactly overlap (Fig. 2A). This localization was confirmed by immunoelectron microscopy, which showed that RIIα was associated with smooth vesicles (Fig. 2B), which were also labeled with golgin (Fig. 2C).
GFAP (green), a marker for mature astroglia in the normal brain, had a filamentous appearance in glioma cells (Fig. 3A). Usually, cells that were strongly labeled with GFAP, and hence that were more differentiated, showed a fainter RIIα labeling, and vice versa. Under some conditions, changes in the pattern of RIIα labeling were detected. Treatment with forskolin, an activator of adenylate cyclase, results in an increase in cAMP, triggering differentiation,20 so that the cells acquire an elongated appearance, with many well-developed processes. After forskolin treatment, in many cells RIIα aggregates were not exclusively confined close to the nucleus (Fig. 3B: RIIα, red and GFAP, green).
In cells treated with H89, an inhibitor of the PKA catalytic activity, GFAP labeling was present only in a few cells. RIIα labeling was more dispersed than in control cells but could be observed around the nucleus (Fig. 3C). After treatment with paclitaxel, a microtubule stabilizer, a few cells survived. In these cells, RIIα labeling (Fig. 3D, red) was uniformly distributed around the nucleus, whereas vimentin labeling (green) was weaker than in controls, and filaments could no longer be recognized.
Vinblastine, which inhibits functional microtubule assembly, induced a dramatic redistribution of RIIα, which appeared scattered through the cell cytoplasm (Fig. 4A). Double labeling (Fig. 4B) showed that although RIIα redistributed, vimentin did not seem to be affected by vinblastine treatment.
Hematoxylin and eosin staining of a mouse brain after GL261 tumor growth shows the boundary between glioma and brain parenchyma (Fig. 4C). Similarly, RIIα labeling (Fig. 4D, red) also highlighted the border between the brain and the tumor; after fixation at room temperature and subsequent permeabilization, RIIα appeared restricted to the tumor area. Cells were densely packed and small in size; therefore, in contrast to in vitro observations, RIIα labeling appeared more restricted to a small labeled area in a perinuclear position, present only in the cells within the tumor. At higher magnification (Fig. 5A), mouse tumor labeling consisted of densely packed dots, similar to in vitro labeling. Similar results were observed in a rat glioma cell line (F98) and in human glioblastoma cells (Gli36), confirming that RIIα peculiar distribution can be a characteristic of glioma cell types.
In F98 cell culture, RIIα (Fig. 5B, red) appeared as a cluster of dots restricted to one side of the nucleus (blue). RIIα labeling of the glioma in the rat brain resembled that in the mouse, being located on one side of the cell nucleus (Fig. 5C, red). As in the mouse, at lower magnification (Fig. 5D), RIIα distribution appeared restricted to the tumor mass, thus defining the boundary between the glioma and the brain parenchyma. RIIα showed a similar perinuclear distribution in the human glioblastoma cell line Gli36 (Fig. 5E).
Glioma labeling, as well as its differentiation from the normal brain tissue, was dependent on the fixation procedure and on the primary antibody. In sections fixed in formalin at room temperature and subsequently permeabilized with Triton, bright labeling of the glioma could be observed with Santa Cruz Biotechnology antibodies; the present data show that this labeling was related to the RIIα isoform. Under this condition, no labeling could be observed in the adjacent normal brain tissue, as previously described.23
In contrast, when sections were first permeabilized with Triton and subsequently fixed with formalin, the same antibody showed a bright labeling, both in glioma and in normal tissue. At high magnification, the pattern of labeling was quite different in the brain and tumor areas. Only when permeabilization preceded fixation was labeling with anti-RIIα or RIIβ (Chemicon) antibodies possible, and microscopy showed that glioma labeling was related only to RIIα, whereas RIIβ was present in the brain but not in glioma. These data suggest that RIIβ detection in the brain requires unmasking before fixation. Brain labeling was related to RIIβ, with the exception of ependymal cells, which are the only cell type that shows insoluble RIIα in the normal brain, as already reported.23 These data suggest that RIIα and RIIβ have different biochemical properties and can be detected under different conditions, RIIα in the tumor cells being easily accessible to antibodies and revealed under milder conditions.
Colocalization experiments indicated that, under the same conditions, RIα detergent-insoluble aggregates were never detected in glioma cells in culture and in the tumor (data not shown); they were instead observed in neurons of some regions of the normal mouse and brain, as previously described.24
The role of the cAMP pathway in cell differentiation and oncogenesis is still unclear. Strict regulation of the RI/RII ratio seems to be crucial for a normal cell cycle. Some human tumors have a predominance of RI over RII, the extent of RI increase being related to malignancy. Induction of differentiation correlates with restoration of a normal RI/RII ratio.16
Among brain tumors, glioblastoma is the most diffuse and the one with the poorest prognosis.1 Up until now, only a few molecular markers, such as intermediatefilament-associated proteins, laminin-8, nestin, epidermal growth factor receptor, and tenascin,30–34 have been used to differentiate glioma cells from adjacent parenchyma, but none of them is unique to glioma cells, making both therapeutic molecular targeting and immunohistochemical diagnosis problematic. Prompted by these considerations, we screened mouse and rat glioma cell lines for the presence and intracellular localization of the four PKA regulatory subunits, an issue apparently never addressed before. Although earlier reports pointed to the presence of only RII in glioma cells,35,36 the present experiments show detergent-insoluble RI isoforms in Western blot, but only RIIα hot spots can be detected by immunohistochemistry. This result can be explained by more distributed RI localization, which does not reach a local intensity higher than the immunofluorescence background.
The RIIα isoform presents an unusual intracellular organization in glioma cells from mice, rats, and humans: instead of being soluble, it is mainly compartmentalized in detergent-insoluble form, conceivably localized in the trans-Golgi network area, as shown by golgin-97 colocalization.37
The functional meaning of this supramolecular organization is not clear: in principle, aggregates of insoluble regulatory subunits may either concentrate catalytic subunits close to target proteins or act as traps for free catalytic subunits, thus creating phosphorylation-free microdomains, or they may bind free cAMP molecules, thus decreasing their intracellular diffusion. Also unclear is the reason that the insoluble aggregates in glioma cells consist of RIIα, while in the rest of the brain they consist mainly of the RIIβ isoform.
In breast cancer and in prostate carcinoma cell lines, paclitaxel, forskolin, and vincristine activate PKA, which hyperphosphorylates Bcl2, leading to apoptosis.38 A similar mechanism may underlie the cAMP-induced cell death observed in the present experiments in glioma cell lines. In glioma cells, paclitaxel also induces a reorganization of vimentin intermediate filaments, an effect apparent in our paclitaxel-treated cells as well.39
Aggregates of detergent-insoluble RIIα are not specific for glioma cells; immunolabeling of ependymal cells was previously described in the normal brain, although with a different localization, at the base of the cilia, and with morphology suggesting a different structure.23 RIIα insoluble aggregates are not detectable in the rest of the brain. Moreover, the structural basis of RIIα aggregates in gliomas is, presumably, different from those of RIIβ in the brain. It is noteworthy that a different sensitivity to detergent treatment is evident: RIIβ aggregates in the brain can be visualized only if permeabilization is made before fixation, whereas RIIα in tumor cells do not need such treatment. In cell culture, RIIα aggregates seem more typical of undifferentiated glioma cells, as suggested by GFAP and vimentin colocalization.
In contrast to normal brain parenchyma, in which RIα and RIIβ insoluble aggregates are detected, mouse and rat glioma cells present a striking cluster of insoluble RIIα. This is consistent with a report on human neuroblastoma SH-SY5Y cells that showed an increase in RIIα after cAMP activation.40 Moreover, in a different glioma cell line, an increase in cAMP induced differentiation and apoptosis,20 suggesting a mechanism common to different glioma models. In humans, some proteins, including AKAPs and phosphodiesterases, are modulated in glioblastoma specimens,21 suggesting involvement of the cAMP signaling system in glioblastoma.
The presence of RIIα in gliomas is possibly correlated with genetic abnormalities on chromosomes 7 and 17 detected in glioblastoma specimens.41,42 We emphasize the novel result that these abnormalities span the PKA-, RIα-, RIβ-, and RIIβ-coding genes, suggesting possible upregulation of RIIα to compensate for other PKA regulatory isoform deficits.
In conclusion, the present data show that in glioma cells, RIIα is distributed at particular sites, unlike its distribution in normal brain tissue. Moreover, glioma cells are sensitive to cAMP modifications. The presence of a detectable biochemical marker in glioma cells could possibly assist in diagnosis and points to an additional pathway that can be targeted for therapy.
This work was supported by Ministero dell’Istruzione, dell’Università e della Ricerca (MIUR), Progetti di Ricerca di Interesse Nazionale (PRIN) 2005 (A.Car.), Fondo per gli Investimenti della Ricerca di Base (FIRB) 2001 (A.Cav.), and the University of Padova (C.M.-C.).