Mitogenic signaling and the relationship to cell cycle regulation in astrocytomas

Transcription

1 Journal of Neuro-Oncology 51: , Kluwer Academic Publishers. Printed in the Netherlands. Mitogenic signaling and the relationship to cell cycle regulation in astrocytomas Arnaud Besson and V. Wee Yong Departments of Oncology and Clinical Neurosciences, University of Calgary, Alberta, Canada Key words: PDGFR, EGFR, PKC, cell cycle regulation, Ras, integrin, PTEN/MMAC1/TEP1 Summary The activity and regulation of a number of mitogenic signaling pathways is aberrant in astrocytomas, and this is thought to play a crucial role in the development of these tumors. The cascade of events leading to the formation and the progression from low-grade to high-grade astrocytomas is well characterized. These events include activating mutations, amplification, and overexpression of various growth factor receptors (e.g. epidermal growth factor receptor (EGFR), platelet derived growth factor receptor (PDGFR), c-met), signaling intermediates (e.g. Ras and Protein kinase C (PKC)), and cell cycle regulatory molecules (e.g. mouse double minute-2 (Mdm2), cyclin-dependent kinase-4 (CDK4), and CDK6), that positively regulate proliferation and cell cycle progression. Inactivating mutations and deletions of signaling and cell cycle regulatory molecules that negatively regulate proliferation and cell cycle progression (e.g. p53, p16/ink4a, p14/arf, p15/ink4b, retinoblastoma protein (Rb), and Phosphatase and tensin homologue deleted from chromosome 10 (PTEN)) also participate actively in the development of the transformed phenotype. Several mitogenic pathways are also stimulated via an autocrine loop, with astrocytoma cells expressing both the receptors and the respective cognate ligand. Due to the multitude of factors involved in astrocytoma pathogenesis, attempts to target a single pathway have not given satisfactory results. The simultaneous targeting of several pathways or the targeting of signaling intermediates, such as Ras or PKC, situated downstream of many growth factor receptor signaling pathways may show more efficacy in astrocytoma therapy. We will give an overview of how the combination of these aberrations drive astrocytoma cells into a relentless proliferation and how these signaling molecules may constitute relevant therapeutic targets. Introduction Astrocytomas are the most common primary brain tumor. The World Health Organization (WHO) has defined four categories of astrocytomas: pilocytic astrocytoma (WHO grade I), astrocytoma (WHO grade II), anaplastic astrocytoma (WHO grade III) and glioblastoma multiforme (GBM) (WHO grade IV) [1]. GBMs are the most common and the most malignant of all astrocytomas and are characterized by a very poor prognosis. Based on genetic analyses, two pathways of GBM formation, with distinct pathogenesis, have emerged [2,3]. The first, de novo pathway, is characterized by the absence of p53 mutations, EGFR amplification/ rearrangement, p16/ink4a and p19/arf deletion (on chromosome 9p) [4], and loss of heterozygosity on chromosome 10 (where the PTEN tumor suppressor is located). The second, progressive pathway leads to the development of GBMs characterized by early inactivation of p53, followed by the mutation of Rb, and Mdm2 and CDK4 amplification as late events. The first pathway occurs primarily in older patients and is associated with more aggressive tumors and poor prognosis, while the second pathway occurs preferentially in younger patients and is associated with better prognosis [2,3]. Mitogenic signaling pathways and cell cycle regulation are tightly linked, as growth factor receptor stimulation will initiate signaling cascades that will lead to the activation of the cell cycle machinery, resulting in cell proliferation. The transition from one phase of the cell cycle to another is regulated by the activity of cyclin-cdk complexes (for more details, see ref. [5], and the review by P. Hamel, this issue of

2 246 J. Neuro-Oncol.). Cyclin-CDKs are themselves regulated by two families of cyclin-dependent kinase inhibitors (CKI), the INK4 family (p16, p15, p18, and p19) and the Cip/Kip family (p21, p27, and p57). Growth factor receptor signaling induces the expression of the G1 cyclins (cyclin D). The G1/S transition is regulated by cyclin D-CDK4/6 complexes that phosphorylate and inactivate the Rb protein, releasing the E2F transcription factors, which in turn allow the transcription of genes required for the progression through S phase (such as cyclin E) (Figure 1). Two pathways regulating cell cycle progression and playing a critical role in preventing tumor formation have been identified (reviewed in [6,7]), namely, the Rb and p53 pathways. The Rb pathway is composed of p16/ink4a, CDK4/6, and Rb, and disruption of the normal function of a single component of the pathway (i.e. loss of p16/ink4a or Rb, or amplification of CDK4/6) is sufficient to lead to dysregulated growth control (Figure 1). The p53 pathway is composed of p14/arf (the human homologue of the mouse p19/arf), the product of an alternative reading frame of the INK4a gene also encoding p16/ink4a [6], Mdm2, and p53. p14/arf can activate p53 and promote the degradation of Mdm2, the latter sequesters and promotes p53 degradation, while p53 promotes Mdm2 expression and inhibits p14/arf expression. These complex interactions aim at tightly regulating p53 level and activity (Figure 1). p53 can block cell cycle progression through the induction of p21/waf1/cip1 expression, and plays a critical role in the activation of the DNA repair machinery and induction of apoptosis. Either loss of p14/arf, p53 mutation, or Mdm2 amplification can disrupt the p53 pathway. It is striking that during the pathogenesis of GBM, there is a nearly invariable inactivation of both the Rb and p53 pathways (through the inactivation or amplification of one of its component), leading to loosened cell cycle control [8]. Other genetic alterations target mostly growth factor receptors and their downstream signaling effectors, giving astrocytoma cells the impetus to proliferate. Genetic studies have shown the importance of cell cycle regulatory molecules in the development of astrocytomas. Mutations in the p53 gene have been reported in 30 50% of astrocytomas, in both low and high-grade tumors [2]. Mdm2 amplification and overexpression was found in about 10% of high-grade astrocytomas (grades III and IV) [2,8]. Recently, mutation/ deletion of p14/arf were found in 40% of GBMs [8]. p14/arf, p16/ink4a, and p15/ink4b are frequently homozygously co-deleted [8] since INK4a, encoding both p16/ink4a and p14/arf, and the INK4b genes are very close on chromosome 9p21. Mutations in p14/arf, p53, and Mdm2 occur in a mutually exclusive manner [8 10]. CDK4 amplification and overexpression has been detected in approximately 15% of GBMs [4,9], while p16/ink4a is homozygously deleted in 30 60% of tumors [4,8,11]; loss or mutation of Rb was observed in 12 36% of GBMs [4,12 14]. These mutations also occur in a mutually exclusive manner [4,9]. Amplification and overexpression of CDK6, and cyclin D1 and D3 were observed in a small fraction of astrocytomas. In total, inactivation of the Rb pathway occurs in 67 85% of GBM [8,9]. Interestingly, 96% of GBMs with altered Rb pathway had also deregulated p53 pathway [8], indicating that the inactivation of both the Rb and p53 pathways are a crucial event in astrocytic tumor formation. In astrocytomas, the conventional checkpoints of the cell cycle are abrogated due to the inactivation of key players of cell cycle control, and mitogenic signals initiated by mutated, overactive growth factor receptors drive cells into an endless proliferative cycle. The various signaling pathways involved in the pathogenesis of astrocytomas will be described in further details below. The purpose of this review is to present the reader with the current evidence that abnormalities in proliferative signaling pathways, in conjunction with the inactivation of cell cycle control pathways initiate and maintain astrocytoma cells into a permanent proliferative state, a crucial step in tumor formation. We will also discuss how these signaling pathways could be potentially targeted for therapeutic purposes. Growth factor receptors All the growth factor receptors that have been implicated in astrocytoma pathogenesis are receptor tyrosine kinases. Upon ligation of the RTK by its ligand, there is dimerization of the receptor, followed by trans-phosphorylation of the cytoplasmic tail of the receptor on specific tyrosine residues. Phosphorylated tyrosines provide docking sites for Src-Homology-2 domain (SH-2) and phosphotyrosine-binding domain (PTB)-containing proteins, such as Shc, Grb2, and phospholipase Cγ (PLCγ), which in turn recruit and activate additional signaling molecules, thus generating a signaling cascade that leads to activation of transcription, translation and increased growth. Signaling pathways activated by RTK stimulation (Figure 2)

3 247 Growth Factor Receptor EGFR a, P DGFR a, c-met a, c-kit a G1 Proliferation/ Mitogenic signals Cyclins (D1,2,3, E) CDK4 a CDK6 a INK4 family p16/ink4a d p15/ink4b d p18/ink4c p19/ink4d p19/arf d Mdm2 a CDK2 Cip/Kip family p21/waf1/cip1 p27/kip1 p57 p53 d DNA repair machinery Skp2 P P P Rb d Rb d phosphorylation Degradation E2 Frelease S phase genes Cyclin E CDK2 Figure 1. Regulation of cell cycle entry by the Rb and p53 pathways. a = gene amplified, mutated (gain of function), or protein overexpressed in astrocytomas. d = gene deleted, mutated (loss of function), or expression repressed in astrocytomas. Schematic showing the genes that are mutated in astrocytomas involved in growth control, either by controlling mitogenic signaling cascades, or cell cycle progression. Mitogenic signals induce G1 cyclins expression, G1 cyclin-cdk complexes phosphorylate the Rb protein which in turn releases E2F transcription factors, turning on transcription of genes required for entry into S-phase (e.g. cyclin-e). Members of the INK4 family can inhibit CDK4/6, thus preventing hyperphosphorylation of Rb and progression into S-phase. Members of the Cip/Kip family can act both as assembly factors for cyclin-d/cdk4-6 complexes and as cyclin-cdk inhibitors, depending on their stoichiometry. p27/kip1 is phosphorylated by cyclin-e/cdk2, targeting it for ubiquitination and degradation. p21/waf1/cip1 expression can be induced by p53, resulting in cell cycle arrest in G1 or G2. p53 is inhibited/destabilized by Mdm-2, and p53 induces Mdm-2 expression. p19/arf (p14/arf in human) can activate p53 functions and induce Mdm-2 degradation, resulting in p53 stabilization; the expression of p19/arf itself is downregulated by p53.

4 248 Figure 2. Schematic describing some of the mitogenic signaling pathways activated downstream of receptor tyrosine kinases (RTKs). RTK ligation by its cognate ligand induces trans-phosphorylation of the cytoplasmic tail of the RTK, providing docking sites for SH2 and PTB-containing proteins, including PI 3-Kinase, Gab1, Src, PLCγ, Shc, Grb2, and Ras-GAP. PI 3-Kinase activation leads to increased levels of PIP3, which participate in the activation of Gab1, ILK, PDK1, PKB/Akt, and PLCγ. Gab1 participates in the activation of the MAPK and JNK pathways. Phosphorylation of Shc by RTKs induces association with Grb2 and Sos, which in turn activates Ras, leading to activation of the MAPK and JNK pathways. Activated JNK phosphorylates c-jun, increasing AP-1-mediated transcription. MAPK phosphorylates and activates several targets involved in growth control: carbamoyl phosphate synthetase (CPS-II), an enzyme involved in pyrimidine-nucleotide synthesis; the MAPK-interacting kinase Mnk1, which phosphorylates and activates the translation initiation factor eif-4e; transcription factors, such as Elk1, involved in transcription of rapid-response genes, required for growth; p90-rsk, which participates in activation of the Serum Response Factor (SRF), phosphorylates histone H3, thus increasing accessibility of transcription factor to DNA, and inactivates Myt1, a kinase that phosphorylates and inactivates CDK1/Cdc2 kinase, required for G2/M transition; and induces cyclin-d1 expression. Activation of PKB/Akt by PIP3, PDK1 and ILK leads to increased transcription, through inhibition of glycogen synthase kinase 3-β(GSK-3β), which phosphorylates and targets β-catenin for degradation. PKB/Akt also phosphorylates and activates the transcription factor E2F and regulates protein synthesis through the activation of the mammalian target of rapamycin (mtor), which itself activates p70-s6 Kinase.

5 249 include the Ras/Raf/MAPK pathway, the PLCγ /PKC pathway, PI 3-Kinase/Akt, and Src family tyrosine kinases. Ligand binding and receptor activation triggers the endocytosis of the receptor and subsequent degradation in the lysosome. The signaling pathways activated by RTKs will be detailed below. Epidermal growth factor receptor Epidermal growth factor receptor (EGFR) is a 170 kda transmembrane glycoprotein containing two cysteinerich extracellular domains, involved in ligand binding, and an intracellular kinase domain. EGFR belongs to the Erb-B subfamily of RTK formed of 4 members: EGFR/ErbB-1, ErbB-2/HER/Neu, ErbB-3, and ErbB-4. These receptors can homo- or heterodimerize, thereby changing their ligand-binding affinity and specificity; moreover distinct signaling cascades will be triggered by different ligands [15]. Physiological ligands of the EGFR include EGF, Amphiregulin, heparin binding EGF-like growth factor (HB-EGF), and transforming growth factor α (TGFα). The EGFR gene is amplified and the protein overexpressed in 40 50% of glioblastomas [2,9,16 19]. In approximately 50% of these cases, an additional rearrangement occurs, resulting in the overexpression of transcripts lacking exons 2 7 and encoding a truncated receptor (p140/egfr, EGFRvIII, EGFR de2 7, or EGFR) lacking part of the extracellular domain (residues 6 273) [9,18,20]. However, expression of EGFR has been reported in the absence of EGFR amplification, and overall, EGFR expression could be found in 49 62% of GBMs [18,21]. Amplification of the EGFR is associated with poor prognosis and shorter survival. This mutant form of EGFR is constitutively activated, exhibits a permanent low level of tyrosine phosphorylation [22] and is unable to bind its physiological ligands [23,24], possibly because the deleted part of the receptor encompasses the ligand binding domain. EGFR has an extended half-life and seems to escape the normal downregulation mechanisms (by the endosome-lysosome pathway) [22,24] possibly due to its aberrant localization in the endoplasmic reticulum [24]; this is however a controversial issue since EGFR expression was reported elsewhere [22,25] on the plasma membrane. A number of studies have investigated the effect of EGFR overexpression in U87MG glioma cells. Cells overexpressing EGFR do not exhibit a growth advantage in vitro, except under serum starving conditions [26], however, when implanted in the flank or brain of mice, their tumorigenic potential is greatly enhanced compared to the wild type cells [25 28], this growth advantage was associated with increased proliferation rate and decreased apoptosis [26 28]. It may be hypothesized that EGFR expression provides a growth advantage only under adverse growth conditions, such as serum starvation or the conditions encountered in vivo. The reduction in apoptosis observed in vivo and in vitro under serum starving conditions was associated with an increased expression of the anti-apoptotic protein BCL-X L [26]. Moreover, EGFR overexpression conferred a substantial resistance to chemotherapeutic drugs (such as cisplatin) by preventing induction of apoptosis, through the upregulation of BCL-X L which inhibit caspase activation [28]. The proliferative effect of EGFR overexpression has been attributed to an increased activation of the Ras signaling pathway [29]. Ras-GTP was more abundant than in wild-type cells, and Shc and Grb2, two adaptor proteins involved in Ras activation (Figure 2), were constitutively associated with phosphorylated tyrosines on EGFR [29]; moreover, microinjection of inhibitory anti-ras antibody completely inhibited DNA synthesis in both EGFR expressing and wild-type cells. The mitogen-activated protein kinase kinase (MEK), and to a lower extent the p42 and p44 MAP kinases (ERK2 and ERK1), were constitutively activated in NIH 3T3 cells overexpressing EGFR [30]. Interestingly, Chinese hamster Ovary (CHO) cells overexpressing EGFR in vitro have a markedly increased proliferation rate and their distribution between the different phases of the cell cycle resembles that of cells treated with EGF or TGFα [23]. The different phenotypes observed between CHO and U87MG cells overexpressing EGFR may reflect genetic differences between the two lines; mitogenic signaling pathways in U87MG are probably already activated under optimal growth conditions and cannot be further enhanced by EGFR expression. Recently, Holland et al. [31,32] developed a mouse model in which the animals spontaneously develop astrocytomalike lesions. In this model, overexpression of a constitutively activated EGFR ( EGFR with an additional deletion in the C-terminal kinase regulatory domain) alone in cells of the astrocytic lineage was not sufficient to induce tumor formation. However, when overexpressed in mice of INK4A/ARF / background (lacking p16/ink4a and p19/arf), rapid tumor formation was observed; similar results were obtained in mice of p53+/ background when activated EGFR was overexpressed along with CDK4. These results

6 250 clearly demonstrate that in order to be fully oncogenic, EGFR first requires the inactivation of the p53 and Rb pathways that control the cell cycle. Another important aspect of EGFR signaling in astrocytoma pathogenesis is the existence of autocrine or paracrine loops, in which either the receptor expressing cell or the neighboring cells, respectively, express the cognate ligand, thus permanently stimulating the mitogenic pathway (reviewed in [33]). The HB-EGF was abundantly expressed in GBMs and glioma cell lines, and was co-expressed with EGFR in 52% of GBMs [34]. Anti-HB-EGF blocking antibodies reduced glioma cell growth by 30 40%, confirming the existence of an autocrine loop [34], and HB-EGF stimulation efficiently induced glioma cell proliferation in vitro. Also HB-EGF, bfgf, or TGFα stimulation rapidly induced HB-EGF expression [34]. TGFα expression was found in 88% of GBMs [35] and most glioma cell lines [36], and in another study, expression of either or both EGF and TGFα was found in all 62 gliomas analyzed [19]. Transfection of antisense TGFα in U251N glioma cells resulted in decreased growth in serum starving condition, in colony formation in soft agar, and decreased tumorigenicity in nude mice [37]. Together these results indicate the presence of a selfsustained autocrine loop in astrocytomas. Targeting the EGFR in astrocytomas is an appealing therapeutic approach and is being actively investigated. The generation of EGFR-specific antibodies could lead to the development of therapies similar to those used against HER-2/Neu (humanized anti-her-2 antibody, Herceptin) in breast tumors [38]. However, no therapeutic benefit was observed in a clinical trial using mouse anti-egfr (EMD55900 or MAb 425) administered intravenously, although the antibody was well tolerated [39]. Another trial, using the same antibody administered intratumorally, was interrupted due to an intense inflammatory reaction [40]. Overexpression of antisense RNA to deplete EGFR in U87MG cells led to a 60 80% growth inhibition, a marked decrease in colony formation in soft agar, and accumulation of cells in the G1 phase of the cell cycle and decreased number of cells in S phase [41]. Cells also exhibited a more differentiated morphology, and had increased GFAP expression [41]. Specific targeting of EGFR mrna by a hammerhead ribozyme in NIH3T3 cells overexpressing EGFR led to a marked inhibition of tumor growth in mice, while the growth of ribozyme transfected cells was not altered in vitro [42]. Similarly, inhibition of EGFR expression by retrovirus-mediated transfer of a hairpin ribozyme in U87MG led to a 70% growth inhibition, and 95% decrease in colony formation in soft agar, under serum starving condition [43], consistent with EGFR providing a growth advantage only under adverse conditions. Another interesting strategy developed to inhibit EGFR signaling in glioma cells consists in overexpressing a mutant, truncated form of the Erb-B2/HER/Neu receptor, which lacks the intracellular kinase domain (T691stop) [44]. This kinase-deficient Neu acts as a dominant negative by forming heterodimers with endogenous EGFR and preventing its activation. U87/T691 cells had a decreased proliferative response to EGF stimulation and ability to proliferate under low serum conditions; their ability to form colonies in soft agar was decreased by over 80%, and cells virtually failed to form tumors in vivo [44]. Moreover, in addition to the wild-type (wt) EGFR, this Neu-T691 mutant could also dimerize and inhibit EGFR activity [44]. The development of specific pharmacological inhibitors may also offer therapeutic opportunities. The tyrphostin AG1478 was shown to preferentially inhibit EGFR rather than the wild type receptor [45] and to revert and even potentiate cisplatin-induced apoptosis in U87MG overexpressing EGFR [28]. Thus these inhibitors could be used in conjunction with more conventional chemotherapeutic agents to improve treatment efficiency. Platelet derived growth factor receptor Two types of Platelet derived growth factor receptor (PDGFR) receptors, α and β, have been characterized and can homo- or heterodimerize. They bind ligand dimers (AA, BB, AB) with different specificity: PDGFRα binds both A and B chains of PDGF, while PDGFRβ binds only the B chain (reviewed in [46]). PDGF is secreted as homo- or heterodimer of A and B chains linked by two disulfide bonds. Glioma cell lines commonly express abundant PDGF-A, and both receptor chains [36,47]. Expression of the PDGF-B-chain and the α-pdgf-receptor was found in all astrocytomas, increasing with grade, while expression of the A-chain was found only in GBMs [48]. PDGFRα gene amplification was found in a small subset of GBMs, and interestingly, PDGFRα and EGFR gene amplifications are mutually exclusive [49]. Several studies have demonstrated the existence of an autocrine loop involving PDGF and its receptors in glioma. Disruption of the autocrine loop by overexpression of a dominant-negative form of the PDGF-A ligand reduced U343MG cell growth and decreased

7 251 tumorigenicity when the cells were implanted in mice [50]. Similar results were reported with rat C6 glioma cells overexpressing a dominant-negative, truncated, PDGFRβ [51]. The autocrine loop was successfully disrupted using monoclonal neutralizing antibodies against PDGF-BB (MAb 6D11), inhibiting growth and migration of A172 cells in vitro [52]. More recent evidence for the importance of the PDGF autocrine loop was shown by injecting retroviruses encoding the PDGF-B chain in mice; 40% of these mice spontaneously developed glioblastoma-like lesions weeks after injection [53]. This technique could provide a useful model for the study of brain tumors. Until recently, there was a lack of pharmacological inhibitors specific for PDGFR; however, such inhibitors have been recently characterized and demonstrated some encouraging results towards their therapeutic use. The compound SU101 (leflunomide, SUGEN) could block the entry of PDGF-stimulated cells into the S phase of the cell cycle, without affecting cell cycle progression of EGF-stimulated cells [54]. SU101 selectively inhibited the growth of PDGFRβ-expressing cell lines, and the growth of intracerebrally implanted glioma cells, as well as other tumor types [54]. Another compound, the polyphenol epigallocathechin-3 gallate (EGCG), a major constituents of green tea, specifically inhibited PDGF-BB-induced tyrosine phosphorylation and activation of p42 and p44 MAPK, phosphatidylinositol-3 kinase (PI 3-Kinase), and PLCγ ; and the subsequent induction of the transcription factors c-fos and egr-1 expression [55]. Moreover, EGCG inhibited spheroid formation of A172 cells [55]. Fibroblast growth factor receptor The fibroblast growth factor (FGF) family encompasses about 20 factors, which are 30 70% identical, including the prototypical acidic FGF (afgf/fgf1) and basic FGF (bfgf/fgf2), most FGFs can bind to all four related RTKs of the Fibroblast Growth Factor Receptor (FGFR) subfamily (FGFR1-4) (reviewed in [56]). FGFRs play an important role during development, and are also involved in proliferation and angiogenesis. Virtually all GBMs express high levels of FGFR1 [57 59], and over 90% of astrocytomas express bfgf [58]. It was noted that FGFR1 expression was increasing with tumor grade and that, on the other hand, FGFR2 expression, which was abundant in normal brain and low-grade tumors, was decreasing with the progression to high-grade tumor [59,60]. High expression of bfgf was associated with a marked decrease in survival for children with pediatric highgrade astrocytoma [61]. The abundant expression of both receptor and ligand suggested the existence of a FGF-FGFR autocrine loop in astrocytomas. Inhibition of bfgf expression using bfgf-specific antisense oligonucleotides decreased human glioblastoma cell growth in vitro [60], demonstrating the existence of such a loop. Similarly, antisense oligonucleotides for FGFR1 inhibited glioma cell growth [62]. Two other studies have provided evidence for a role of FGF in astrocytoma genesis. The rate of brain tumor induction by exposure to N-nitrosoethylurea in rats was markedly increased by the implantation of bfgf-saturated gelfoam in the brain of the animals [63]. However, transgenic mice with glial-specific overexpression of bfgf did not develop tumors but had an increased proliferation and migration of glial cells [64]. Together, these results indicate that, similarly to EGFR, the constitutive activation of the FGF/FGFR mitogenic pathway is not sufficient on its own to induce full transformation and tumor development. However when other alterations are already present (such as the ones provoked by N-nitrosoethylurea, [63]), activation of the FGF/FGFR pathway provides a growth advantage and can lead to tumor formation. Recently, a transcriptional repressor of the bfgf gene was isolated and named regulator of FGF2 transcription (RFT). The RFT-A protein represses FGF2 gene expression, while two splice variants having deletions in the DNA binding domain, RFT-A, and RFT-B, fail to bind the FGF2 promoter and to repress FGF2 gene expression [65]. Interestingly, RFT-A expression was reduced in glioma cells, while levels of the splice variants was increased, the reverse was observed in normal cells, indicating that the FGF2 gene was not efficiently repressed in tumor cells. Moreover, overexpression of RFT-A induced apoptosis in glioma cells [65]. Disruption of the FGF/FGFR autocrine loop using anti-bfgf neutralizing antibodies (MAb 3H3) caused apoptosis of U87MG and U251MG glioma cells in vitro and in tumor xenografts in mice [66]. c-met/hepatocyte growth factor receptor c-met is a RTK formed of a 50 kda extracellular α subunit, and a 140 kda (or 170 kda) transmembrane β subunit with an intracellular kinase domain. Hepatocyte

8 252 growth factor/scatter factor (HGF/SF) is the ligand for c-met. HGF/SF plays a role in the regulation of cell growth, cell motility, morphogenesis, and is a potent angiogenic factor, possibly through the regulation of vascular endothelial growth factor (VEGF) expression [67]. Expression of c-met was found in % of GBMs [68 70], while 53 72% expressed HGF/SF [69,70], and co-expression of both c-met and HGF/SF was found in 40 86% of GBMs [69,70], suggesting the presence of an autocrine loop. Amplification of the c-met gene has also been reported in a small fraction of high-grade astrocytomas. HGF/SF stimulation of glioma cell lines induced progression into S-phase and proliferation could be inhibited by the presence of anti-hgf/sf neutralizing antibodies [70], confirming the presence of autocrine c-met/hgf stimulation in glioma cells. Overexpression of HGF/SF in U373 resulted in increased colony formation in soft-agar and increased tumorigenicity in mice [71]. Recently, depletion of either c-met or HGF/SF was accomplished with a U1 small nuclear RNA/ribozyme antisense in U87 cells (expressing both c-met and HGF/SF) [72]. Depletion of HGF/SF or c-met caused a 17-fold and 11-fold decrease in colony formation in softagar, respectively [72]. The same cells had a substantially decreased capacity to form tumors in mice, and the mitotic index in those tumors was approximately 50% of the control, indicating that disruption of the c-met/hgf autocrine loop impairs cell proliferation in vivo [72]. c-kit/stem cell factor receptor The RTK c-kit and its ligand stem cell factor (SCF), also known as steel factor (SLF) are mostly expressed in cells of the hematopoietic, gonadal, and pigment cell lineages, where they play an important role in development, proliferation, and differentiation. Expression of c-kit and SCF/SLF has been found in brain tumors [73,74], and c-kit levels increased with tumor grade [74]. Most astrocytoma cell lines express both c-kit and SCF/SLF [75,76] and the existence of an autocrine [75] or intracellular autocrine (intracrine, when both the receptor and ligand are expressed intracellularly and interact inside the cell) [76] loop has been suggested; however some of the data reported is conflicting. The c-kit/scf system will require further studies in order to determine whether it plays an important role in the pathogenesis of astrocytomas. Integrins Integrins are a family of cell-surface receptors for extracellular-matrix proteins such as laminin, vitronectin, and fibronectin (reviewed in [77,78]). Sixteen α and 8 β subunits are known so far and they can combine to form at least 24 different integrin receptors. Upon ligation, integrins trigger the formation of focal adhesions, which are large complexes of cytoskeleton, scaffolding, docking, and signaling proteins. Among the latter, focal adhesion kinase (FAK), a cytosolic tyrosine kinase, becomes activated and recruit members of the Src-family tyrosine kinases, the Src-FAK complex in turn activates several signaling pathways such as Ras/MAPK, PI 3-Kinase/Akt, and p130-cas (Crkassociated substrate)/crk (Figure 3). Although integrin ligation is usually associated with cell spreading, migration, and survival, there is accumulating evidence to suggest that integrins play an important role in the regulation of cell cycle and proliferation, these recent findings should grant further studies in astrocytomas. The participation of integrins in proliferative signals came from the finding that FAK activation could mediate ERK/MAPK activation and thus synergize with mitogenic signaling pathways to stimulate cell proliferation [79]. FAK-induced ERK/MAPK activation was dependent on the recruitment of Src by FAK and subsequent Ras activation [80]. Inducible FAK expression in human fibroblasts was correlated with accelerated progression through G1 and S phases of the cell cycle [81]. Conversely, overexpression of dominant-negative FAK inhibited cell cycle progression, this was associated with a decrease in MAPK activation and cyclin-d1 expression, and an increase in p21/waf1/cip1 expression [81]. Recently, integrin-mediated activation of the c-jun N-terminal kinase (JNK) pathway through recruitment of p130-cas by FAK/Src and subsequent activation of Crk, was found to be necessary for cell cycle progression [82] (Figure 3). Activated JNK enters the nucleus and phosphorylate c-jun, which then associates with c-fos to form the AP-1 transcription factor, whose activity is required for transcription of genes important for cell cycle progression. As our knowledge of signal transduction increases, it is becoming more evident that integrins and growth factor receptors can activate the same pathways, and that an extensive cross-talk exists between both types of receptors. Integrins play an important role in the malignant phenotype of astrocytomas, they have been involved in mediating migration, invasion, and in

9 Figure 3. Schematic describing some of the mitogenic signaling pathways activated downstream of the integrin receptors. Integrin ligation to extracellular matrix proteins induces FAK auto-phosphorylation, allowing recruitment and activation of Src. Subsequent phosphorylation of FAK by Src family kinases provides docking sites on FAK for additional signaling proteins, such as PI 3-Kinase and Grb2. Activation of PI 3-Kinase leads to activation of ILK, PDK1, and PKB/Akt (see legend Figure 1). Other targets of PI 3-Kinase include Rac, which in turn activates p21-activated kinase (PAK) and the JNK pathway; and PLCγ which activates PKC. PKC can induce c-fos expression, and activate the MAPK pathway by direct phosphorylation of Raf-1. The FAK/Src complex activates p130cas (CRKassociated substrate), which recruits the adaptor protein Crk. Crk leads to JNK pathway activation, and also binds Sos (not shown), thus participating in Ras/MAPK activation. 253

10 254 protecting cells from apoptosis (reviewed in [83]); recent findings implicate integrins in the control of cell cycle progression and proliferation in normal cells, and it may also be the case in astrocytic tumors. Cytoplasmic signaling proteins PTEN/MMAC1/TEP1 PTEN is a unique protein in that it is a dualspecificity protein phosphatase (towards phosphoserine/threonine and phospho-tyrosine residues although only its tyrosine phosphatase activity has been demonstrated in vivo), and a phosphoinositide 3-phosphatase (reviewed in [84]). Therefore, PTEN can dephosphorylate both proteins and the lipid second messenger phosphatidylinositol-3,4,5 trisphosphate (PIP3), which is the product of the reaction catalyzed by PI 3-Kinase. PTEN has been involved in the regulation of cell proliferation, adhesion, migration, and survival; we will focus here on the roles PTEN play in the regulation of cell cycle progression and proliferation by antagonizing mitogenic signals (Figure 4). PTEN was identified as the putative tumor suppressor frequently mutated on chromosome 10; loss of part or the totality of chromosome 10 is the most common genetic event in GBMs. Mutations/deletions of the PTEN gene were reported in 27 44% of GBMs [85 87]. In another study, PTEN mutations were found to be restricted to primary (de novo) GBMs (32%), and were rarely found in secondary (arising from preexisting low grade tumor) GBMs (4%) [88]. PTEN mutations are found only in high-grade astrocytomas [85 88]. The putative role of PTEN as a tumor suppressor was first demonstrated by its capability of inhibiting growth and reversing the malignant phenotype of glioma cells. Adenoviral-mediated transfer of wt-pten in glioma cells resulted in decreased proliferation, loss of anchorage-independent growth, and loss of tumorigenicity in nude mice [89]. The phosphatase activity of PTEN was found to be necessary to mediate its growth suppressive properties [90]; wt-pten decreased U87 and U178 cell growth by 60 70%, while no reduction in proliferation was observed when a phosphatase inactive PTEN was overexpressed [90]. PTEN was shown to directly dephosphorylate FAK on tyrosine residues, preventing p130-cas recruitment and phosphorylation, this was associated with a decrease in adhesion and spreading; this activity was independent of the phosphoinositide phosphatase activity of PTEN [91,92]. One may speculate that if PTEN can antagonize FAK activation, it may then prevent the activation of mitogenic pathways downstream of FAK such as the MAPK and JNK pathways (see above). Inhibition of Shc phosphorylation by PTEN following EGF stimulation was described in U87 cells, thus preventing subsequent association of Shc with Grb2 and Sos and activation of the Ras/Raf/MEK/MAPK pathway [93]. PTEN dephosphorylate Shc directly [94]. Another study reported conflicting results, as EGF stimulation in PTEN overexpressing U251 cells did not prevent MAPK activation [95], nonetheless, inhibition of MAPK activation was seen following PDGF stimulation and fibronectin ligation of the same cells [95]. Recently, PTEN was shown to inhibit EGF signaling by preventing the recruitment of the docking protein Gab1 to the EGF receptor, which is dependent on PIP3 production by PI 3-Kinase, and inhibiting the downstream activation of the MAPK and JNK pathways [96]. The growth suppressive effect of PTEN in glioma cells is associated with accumulation of cells in G1 and decreased progression into S phase [90,97] and correlated with increased expression of the cyclin-cdk inhibitor p27/kip1 and a corresponding decrease in the G1-S cyclin-cdk kinase activities (cyclin-a, -E, and CDK2) [97]. The upregulation of p27/kip1 was due to inhibition of the serine/ threonine kinase PKB/Akt mediated by PTEN [97]. It was subsequently demonstrated that the PTEN mediated G1 cell cycle arrest through the inhibition of the PI 3-Kinase/Akt pathway was dependent on the phosphoinositide phosphatase activity of PTEN [98]. In order to be activated, PKB/Akt requires phosphorylation on two serine residues by enzymes known as phosphatidylinositol-dependent kinase-1 (PDK1) and PDK2. Integrin-linked kinase (ILK), a serine/threonine kinase associated to integrin β chains, has PDK2 activity and can mediate PKB/Akt activation [99]. PDK1 was also found to exhibit PDK2 activity. Activation of all three enzymes requires the binding of PIP3 to their pleckstrin homology (PH) domain, allowing translocation to the membrane (Figure 4). Thus, regulation of PIP3 levels by PTEN is critical for the activation of PKB/Akt. It is of interest that PKB/Akt activity (of the three isoforms, Akt1-3) is elevated in glioblastoma cells [100,101] lacking a functional PTEN. Given the critical roles played by PKB/Akt in regulating cell survival and proliferation, inactivation of PTEN clearly confers astrocytoma cells a growth advantage. Part of the growth suppressive activity of PTEN could

11 255 Figure 4. PTEN in the negative regulation of signal transduction pathways. PTEN can directly dephosphorylate tyrosine residues on FAK and Shc, thus preventing the activation of the pathways activated downstream of these proteins. The phosphoinositide phosphatase activity of PTEN, by dephosphorylating the PI 3-Kinase product PIP3, prevents the activation of proteins such as Gab1, PLCγ, PDK1, PKB/Akt, and ILK; and the activation of the signaling pathways downstream of these proteins. Potentially, PTEN could antagonize the activation of all the proteins dependent on PIP3 for activation, therefore all the PH domain containing proteins could be negatively regulated by PTEN. be mediated through the inhibition of PKB/Akt and preventing it to phosphorylate and activate p70 S6 kinase, a protein that plays a key role in the regulation of protein synthesis. Through the regulation of PIP3 levels, PTEN was shown to suppress the activation of PLCγ and ILK [101], probably by decreasing the availability of PIP3 for binding to their PH domain, thus preventing proper localization of the enzymes at the membrane. Inhibition of PLCγ by PTEN is likely to affect PKC activation (see below). Recently, PTEN-induced growth arrest was shown to rely on the presence of a functional Rb protein [102], as tumor cell lines Saos-2 and C33A, both lacking Rb, are not sensitive to G1 arrest by PTEN, which could be restored by re-introducing Rb in those cells [102]. PTEN inhibited the phosphorylation of Rb, a step required for transition into S phase, and decreased cyclin D1 expression [102]. PTEN-induced cell cycle arrest was relieved by PI 3- Kinase overexpression, and also by the overexpression of the small GTP-binding proteins Rac and Cdc42. Although the mechanism by which Rac and Cdc42 prevent PTEN-induced growth arrest was not investigated

12 256 [102], it is possible that both proteins act through the activation of the JNK pathway. PI 3-Kinase, Rac, and Cdc42 activities have all been demonstrated to induce cyclin-d1 expression [ ], this could represent a mechanism by which PTEN-induced growth arrest could be abrogated. The glycogen synthase kinase-3β (GSK-3β) was shown to mediate the phosphorylation of cyclin-d1 on threonine 286, targeting it for proteolytic degradation [106], however, inactivation of GSK-3β by PKB/Akt could provide a mean of cyclin- D1 stabilization. Also, mammalian target of rapamycin (mtor), a protein activated by PKB/Akt, was shown to increase cyclin-d1 mrna translation [107]. Overall, the tumor suppressor PTEN regulates many cellular functions, including proliferation, cell cycle progression, migration/motility, and apoptosis, by either directly dephosphorylating proteins such as FAK and Shc, or by modulating the levels of the PI 3-Kinase product PIP3, thereby potentially inhibiting all the pathways activated by PIP3 (Figure 4). The critical roles of PTEN in the regulation of multiple phenotypes render its inactivation in tumors, and particularly astrocytomas, important to provide a growth advantage. Ras Ras is a member of the small GTP-binding protein family; there are several Ras proteins in mammals (H-Ras, N-Ras, and K-Ras). In its GDP-bound state, Ras is inactive, while binding of GTP activates Ras and its GTPase activity (Figure 2) (reviewed in [108,109]). Ras is regulated by GDP-exchange factors (GEFs) (such as Sos) which facilitates the transition from GDPbound to GTP-bound, and by GTPase-activating proteins (GAPs) (such as RasGAPs, 4 isoforms known in mammals: p120-gap, GAP1, GAP1-P4BP, and neurofibromin) which increase the GTPase catalytic activity, thus facilitating the return to a GDP-bound, inactive, state. PI 3-Kinase can also mediate Ras activation. Once activated, Ras recruits the serine/threonine kinase Raf to the membrane, where it can become fully activated by a mechanism that remains unclear, and phosphorylate MEK. Activated MEK phosphorylates the MAPK proteins (p44/erk1 and p42/erk2) which in turn phosphorylate a variety of targets, including transcription factors such as Elk-1. Thus, Ras is a key molecule in the activation of the MAPK pathway downstream of RTKs. Other targets activated by Ras include the JNK pathway (Figure 2) [110], Rac and Rho, and PI 3-Kinase. Activating mutations of Ras, resulting in decreased GTPase activity, are found in approximately 30% of all human cancers. However, such mutations were not detected in astrocytomas. Other mutations in the Ras pathway are activating mutations of GEFs or inactivating mutations of GAPs, both resulting in increased Ras activity. Although Ras is not mutated in astrocytomas, the overexpression and/or constitutive activation (through mutations or autocrine loop) of several RTKs is likely to result in increased Ras activation. Thus, being activated downstream of many cellular receptors, Ras constitute a target of choice for therapeutic purposes. Activated Ras (GTP-bound) levels were elevated in GBMs specimens when compared to normal brain tissues [111,112]. Astrocytoma cell lines exhibit an increased amount of active Ras-GTP (20 30%), similar to that observed in v-h-ras transformed fibroblasts, when compared to non-transformed NIH 3T3 cells (5%) [112]. Ras activation correlated with the stimulation of RTKs such as EGFR and PDGFR [112]. U373 overexpressing a dominant-negative Ras (Ras-N17) had a 50% decrease in EGF-induced MAPK activation, decreased proliferation and anchorage-independent growth in soft-agar [112]. The farnesyl transferase inhibitor (FTI) L-739,749, which inhibits the enzyme involved the post-translational modification of Ras (required for proper localization of Ras at the cell membrane), inhibited U87 cells proliferation by approximately 50% [112]. The growth inhibitory effect of the FTI L744,832 was associated with decreased entry into S phase and accumulation in G2/M, and with cell death accompanied with induction of the pro-apoptotic proteins Bax and Bak and of the anti-apoptotic protein Bcl-2 [113]. It will be interesting to see whether FTIs show efficacy on astrocytoma cells in vivo and in clinical trials. In a phase I clinical trial, the FTI R115/777 demonstrated mild toxicity [114] and awaits phase II trial. FTIs have given some very promising results in mice models with distinct types of tumors [115]. However, one may question the specificity of inhibition obtained when using FTIs, as numerous cellular proteins are farnesylated, and the anti-tumorigenic effect of FTIs may not be entirely due to Ras inhibition. For example, the protein Rho-B, which is farnesylated in vivo, was shown to mediate the growth suppression induced by the FTI L739,749 in Rat1 cells transformed with a myristylated Ras (and thus insensitive to FTI); while Ras-transformed cells expressing a myristylated Rho-B were resistant to growth suppression induced by the FTI [116].

13 257 Targeting of the Ras/MAPK pathway with the oncolytic virus reovirus has given some promising results, leading to U87 tumor regression in severe combined immuno-deficiency (SCID) mice [117]. For reovirus to be infectious, the Ras/MAPK pathway must be activated. Indeed, Ras/MAPK activation results in the inhibition of the double-stranded RNA-activated protein kinase (PKR), thereby allowing viral protein synthesis and virus multiplication, making tumor cells with activated Ras pathway very susceptible to viral infection, while normal cells were not infected [117]. Protein kinase C Protein kinase C (PKC) is a family of 12 serine/ threonine kinases that has been divided in three subgroups, based on the requirement of each isoform for activation (reviewed in [118]): Ca 2+, diacylglycerol (DAG), and phosphatidylserine are required for full activation of the conventional PKCs (PKC α, β1, β2, and γ ), while Ca 2+ is not required for activation of the novel PKCs (δ, ε, η, θ, ν, and µ/pkd), and both Ca 2+ and DAG are dispensable for atypical PKC activation (PKC ι/λ and ζ ) [118]. The substrate specificity of PKC is thought to be mediated by a set of anchoring proteins and by a tight regulation of their subcellular localization [118]. PKC is activated downstream of many membrane receptors, including RTKs and integrins and is involved in many cellular processes such as proliferation, cell cycle progression, apoptosis, migration/motility, invasion, and angiogenesis. PLCγ is one of the main enzyme involved in PKC activation, it cleaves phosphatidylinositol 4,5 bisphosphate (PIP2) into DAG and inositol 1,4,5 trisphosphate (IP3) (Figure 2). PLCγ is activated by RTKs (by binding to tyrosine phosphorylated residues on the receptor), PI 3-Kinase, and FAK. PKC may participate in mitogenic signaling by activating the Raf/MEK/MAPK pathway, indeed, PKC can directly phosphorylate and activate Raf-1 [119], although Ras appears to be required to recruit Raf to the membrane [120]. There is substantial evidence to suggest a critical role for PKC in the malignant phenotype of astrocytomas (reviewed in [121,122]), and PKC may constitute a relevant target for astrocytoma therapy since it is at the convergence of many signaling pathways. In astrocytomas, PKC has been involved in the regulation of proliferation, cell cycle progression, invasiveness, apoptosis, and angiogenesis [121,122]. PKC has been involved in the downregulation of EGFR signaling [ ], through phosphorylation of the juxtamembrane domain of the receptor (T654), resulting in decreased affinity of the receptor for its ligand [124] and inhibition of its tyrosine kinase activity [123,125]. PKC-induced down-modulation of EGF signaling was recently shown to be MAPK-dependent [125]. Thus, PKC could have an anti-proliferative role by inhibiting EGF signaling. However, recent evidence suggest that although activation of PKC potently inhibits subsequent ligand-induced EGFR signaling; in cells that are pre-treated with EGF, PKC-mediated phosphorylation of the receptor stabilizes ligand receptor interactions, and results in intensification of EGFR signaling [126]. Therefore, in the astrocytoma context, where an autocrine EGF/EGFR loop exists and the receptor is permanently stimulated, PKC may not exert an inhibitory role on EGF signaling, but may synergize in the mitogenic response to EGF stimulation. Although PKC is not mutated nor amplified, high expression levels of several isoforms have been reported in astrocytomas cell lines, including PKC α, ε, and ζ [127,128]. Analysis of PKC isoform expression in tumor samples gave conflicting results, and should be re-examined in the future. PKC activity was markedly increased in astrocytoma cell lines when compared to non-transformed astrocytes and directly correlated with their growth rate [129]. PKC activity was also elevated in astrocytoma specimen [130]. Depletion of PKC with the phorbol ester phorbol-12-myristate-13- acetate (PMA) (phorbol esters are potent activators of conventional and novel PKC isoforms) resulted in a decreased growth rate [129]. Inhibition of PKC with a relatively non-specific inhibitor (staurosporine) or with its more specific analog CGP resulted in a marked decrease in proliferation associated with decreased entry into S-phase [ ]. Interestingly, the mitogenic effect of serum addition or stimulation with EGF and FGF on A172 or U251 glioma cells was completely abrogated by PKC inhibition [130,131], suggesting that mitogenic signaling in astrocytoma cells is mediated by a PKC-dependent pathway. In another study, CGP inhibited proliferation of 9 glioma lines; this was associated with a decrease in S-phase and accumulation in the G2/M phase, and with induction of apoptosis [132]. CGP also inhibited tumor growth of U87-MG and U373-MG implanted in the flank of nude mice [132]. The growth inhibition and cell cycle arrest caused by CGP and Ro (another staurosporine derivative) was later associated with decrease of CDK2 and CDK1/Cdc2 associated kinase activities and decreased