Integrins and p53 pathways in glioblastoma resistance to temozolomide

Glioblastoma is the most common malignant primary brain tumor. Surgical resection, postoperative radiotherapy plus concomitant and adjuvant chemotherapy with temozolomide (TMZ) is the standard of care for newly diagnosed glioblastoma. In the past decade, efforts have been made to decipher genomic and core pathway alterations to identify clinically relevant glioblastoma subtypes. Based on these studies and more academic explorations, new potential therapeutic targets were found and several targeting agents were developed. Such molecules should hopefully overcome the resistance of glioblastoma to the current therapy. One of the hallmarks of glioblastoma subtypes was the enrichment of extracellular matrix/invasion-related genes. Integrins, which are cell adhesion molecules important in glioma cell migration/invasion and angiogenesis were one of those genes. Integrins seem to be pertinent therapeutic targets and antagonists recently reached the clinic. Although the p53 pathway appears often altered in glioblastoma, conflicting results can be found in the literature about the clinically relevant impact of the p53 status in the resistance to TMZ. Here, we will summarize the current knowledge on (1) integrin expression, (2) p53 status, and (3) relationship between integrins and p53 to discuss their potential impact on the resistance of glioblastoma to temozolomide.


INTRODUCTION
Glioblastoma is characterized by rapidly dividing cells, high degree of vascularity, invasion into the normal brain tissue, and an intense resistance to death-inducing stimuli. Significant advances have been made in understanding the molecular genetics underlying the heterogeneity of glioblastoma and their resistance to therapies. However, standard therapy including surgical resection and radiotherapy with concomitant and adjuvant chemotherapy using temozolomide (TMZ) remains poorly efficient (Stupp et al., 2005). Prognostic and predictive markers are continuously proposed based on large scale genomic data. In the recent years, emphasis has been given to the predictive impact of O 6 -methylguanine-DNA methyltransferase (MGMT) expression/activity, a DNA repair enzyme that protects cells against alkylating drugs such as TMZ (Stupp et al., 2009). The promoter of the MGMT gene is methylated in 40-45% of glioblastoma and the enzyme is not expressed in the majority of these cases (Hegi et al., 2005;Silber et al., 2012). While the contribution of MGMT to TMZ resistance is highly documented, tumors in which MGMT is not the primary determinant of treatment outcome also exist (Carlson et al., 2009;Combs et al., 2011). Integrins have been proposed to play a role in the aggressiveness of gliomas and have been implicated in radio/chemoresistance in different types of tumors (Aoudjit and Vuori, 2012). The p53 protein has been largely studied in gliomas but its prognostic value has not been consistently established. In line with our recent data proposing an α 5 β 1 integrin-p53 axis with potential implication in TMZ resistance (Janouskova et al., 2012), we will summarize here the current knowledge on integrins and p53 status in glioblastoma.

INTEGRINS IN GLIOMA BIOLOGY OF INTEGRINS
Integrins are heterodimeric cell surface receptors that mediate cell adhesion to the extracellular matrix (ECM) and support cell-cell interactions in a multitude of physiological and pathological situations. They are at least 24 known αβ heterodimers formed by a combination of 18 α and 8 β subunits bound non-covalently. Natural ligands of integrins are component of the ECM such as vitronectin, collagen, or fibronectin. Each αβ integrin pair has a defined set of ECM protein (Hynes, 2002). The repertoire of integrins present at the membrane dictates therefore the extent to which a cell will behave on a specific matrix and respond to its environment. Once engaged with the ECM, integrins cluster and recruit various signaling and adaptor proteins to form focal adhesion complexes (Geiger et al., 2001). These complexes activate intracellular downstream signaling pathways including NF-κB, PI3K, Src, or Ras-MAP kinases (Hynes, 2002;Legate et al., 2009). Such pathways regulate functions involved in motility, cytoskeleton organization, adhesion, proliferation, survival, and gene transcription. Integrins link ECM to the actin cytoskeleton through FAK/ILK/SFK/Rho proteins pathway providing the traction necessary for cell motility (Geiger et al., 2001). Integrins regulate the localization and the activity of urokinase-type plasminogen activator (uPA)/uPA receptor (Ghosh et al., 2000;Wei et al., 2007;Bass and Ellis, 2009) and matrix metalloproteinases (MMPs; Lamar et al., 2008;Morozevich et al., 2009) therefore controlling ECM remodeling and the invasive process.
Beside their mechanical functions and despite the lack of intrinsic kinase activity, integrins are true signaling molecules.
Integrins regulate proliferation by controlling the expression of cyclin D1 which permits cells to enter the S-phase of the cell cycle (Fournier et al., 2008). Integrins relay survival or apoptotic signals depending on the surrounding environment. Integrin ligation promote survival through various mechanisms including increased anti-apoptotic proteins (bcl-2, FLIP; Aoudjit and Vuori, 2001b;Matter and Ruoslahti, 2001;Uhm et al., 1999), activation of PI3K-Akt (Aoudjit and Vuori, 2001a) or NF-κB pathway (Scatena and Giachelli, 2002;Courter et al., 2005). Unligated integrins were reported to promote apoptosis through the so-called integrin-mediated death (IMD) a mechanism dependent, or not, on caspases activation in anchorage-dependent cells (Stupack et al., 2001;Jan et al., 2004). However, tumor cells are often IMD-resistant and unligated integrins rather promote anchorage-independent growth, survival, and metastasis than apoptosis (Desgrosellier et al., 2009). Additionally, crosstalks occur between integrins, cytokines, and growth factor receptors. Optimal growth factor stimulation relies on integrinmediated adhesion to an appropriate ECM protein. α v β 3/5 and α 5 β 1 interact with growth factor receptors (VEGFR2, c-Met, FGFR1, PDGFR, EGFR, TIE-2, and IGF-1R) to promote full activation of each receptor and maximal signal transduction (increased MAPK and Akt activity) resulting in enhanced cell migration, proliferation, survival, and angiogenesis (Friedlander et al., 1995;Eliceiri, 2001;Alam et al., 2007;Soung et al., 2010). Integrins were also reported to bind directly growth factor (such as angiopoietins or VEGF) allowing the transduction of information in the absence of the receptor (Carlson et al., 2001;Hutchings et al., 2003). In short, integrins sense, interpret, and distribute information so that cancer cells adjust and respond to their microenvironment.

INTEGRIN EXPRESSION AND FUNCTION IN GLIOMA
Clustering of transcriptomic data from high grade glioma predicted poor survival in subclasses of tumors overexpressing ECM components such as fibronectin which is the preferred ligand of α 5 β 1 and α v β 3 integrins (Geiger et al., 2001;Freije et al., 2004;Bredel et al., 2005;Colin et al., 2006;Tso et al., 2006). Functional analysis revealed gliomagenesis and glioblastoma networks composed of genes that play a role in integrin signaling including fibronectin, α 3 and α 5 integrins (Bredel et al., 2005). Gingras et al. (1995) investigated glioblastoma for the expression of cell adhesion molecules including integrins that might distinguish tumor from normal adjacent brain tissue. Results showed that glioblastoma expressed α 2 , α 3 , α 5 , α 6 β 1 , and α v β 3 integrins at significantly higher level than normal brain tissue suggesting that these integrins might play a role in the development or the progression of glioma (Gingras et al., 1995). β 8 and α 5 β 1 integrins were commonly expressed in a perinecrotic or perivascular pattern in glioblastoma (Riemenschneider et al., 2005). Higher levels of α 5 and β 3 integrin mRNA were measured in glioblastoma as compared to normal brain or low grade astrocytoma (Kita et al., 2001). Average α v β 3 integrin expression in glioblastoma seemed to exceed those in low grade glioma at the protein level although mRNA levels of both subunits were not discriminative between glioblastoma and low grade glioma (Schnell et al., 2008). In another study, α v β 5 and α 5 β 1 integrins were shown to be expressed at consistently higher levels than α v β 3 integrins in human glioma cell explants (Mattern et al., 2005). We and others showed recently that α 5 β 1 integrin expression in biopsies from patient with glioma correlated with poor prognosis and tumor aggressiveness (Cosset et al., 2012;Holmes et al., 2012;Janouskova et al., 2012).

INTEGRINS IN GLIOMA STEM CELLS
Brain tumors also contain highly tumorigenic and therapeutically resistant pluripotent stem cells referred as glioma stem or initiating cells. The glioma stem cell hypothesis incorporates a model in which only a small subset of cells, the glioma stem cells, can initiate tumor. This hypothesis was confirmed very recently in vivo (Lathia et al., 2011). Elevated levels of α 6 β 1 integrins were found in glioma stem cells and seem to be a reliable new marker to enrich for glioma stem cells (Lathia et al., 2010).
Integrins are implicated at various levels of glioma development and progression. Blocking their functions may affect both tumoral cells and endothelial cells and these characteristics made them attractive therapeutic targets for glioblastoma Goodman and Picard, 2012). Emphasis on α v β 3 integrins has been given recently as cilengitide, their prototypical small peptide antagonist, is currently evaluated in phase III clinical trials in glioblastoma (Tabatabai et al., 2010). Interestingly, the outcome in a phase II trial was particularly good in patients with a methylated MGMT gene promoter (Stupp et al., 2010). Emerging data showing the role of α 5 β 1 integrin in glioblastoma give some hope for new therapeutic propositions in the near future.

p53 PROTEIN IN GLIOMA
p53, the "guardian of the genome," is certainly one of the most widely studied protein in human glioma. Activation of the tumor suppressor p53 by stress signals triggers different cellular programs such as cell cycle arrest, apoptosis, differentiation, DNA repair, autophagy, and senescence through complex network and signaling pathways (Levine and Oren, 2009;Vousden and Prives, 2009;Sullivan et al., 2012). Gaining a better understanding of how transcriptional and non-transcriptional functions of p53 integrate will be of great importance for the proposal of new therapeutic options (Dai and Gu, 2010;Speidel, 2010). Somatic p53 missense mutations are found in approximately 50% of all human cancers. Intensive research on p53 status as a classical molecular marker led to controversial results and non-significant clinical impact, particularly in the glioma field.

p53 STATUS IN GLIOMA
As most mutations in p53 gene led to the accumulation of p53 in the nucleus, nuclear overexpression of p53 was usually considered as a marker of mutation. Several studies showed that the expression of p53 is correlated at 90% with its mutation (Figarella-Branger et al., 2011). Detection of p53 mutation by the yeast functional assay that measures quantitatively mutant p53 alleles and qualitatively the loss of p53 competence was also employed and compared to conventional techniques including DNA sequencing (Tada et al., 1997;Fulci et al., 2000). Overall results indicate that p53 mutations often occurred in low grade gliomas (WHO grade II astrocytoma; Bourne and Schiff, 2010) and thus is a frequent event in the pathological progression of secondary glioblastoma (WHO Grade IV; Gladson et al., 2010). Secondary glioblastoma arise from a preexisting grade II or III astrocytoma in contrast with primary glioblastoma that form de novo. Primary glioblastoma represent about 90% of glioblastoma. p53 gene mutations are present in about 30% of primary glioblastoma, and occur more frequently in secondary glioblastoma (65%; Ohgaki et al., 2004;Zheng et al., 2008). A recent integrated genomic analysis identified four relevant subclasses of glioblastoma (proneural, mesenchymal, neural, and classical glioblastoma). p53 mutation was observed in 54, 32, 21, and 0% of tumors from the proneural, mesenchymal, neural, and classical glioblastoma subtype, respectively (Verhaak et al., 2010). Interestingly, classical glioblastoma benefit from more aggressive therapy regimen than the others (Verhaak et al., 2010). In fact, the prognostic value of p53 status may be reconsidered according to these data.

p53 STATUS AND GLIOBLASTOMA PROGNOSIS
No clear consensus has been reached about the prognostic value of p53 status despite numerous studies ( Table 1). A clear picture remains difficult to draw due to the different techniques used to evaluate p53 (including immunostaining on tumor tissues, direct sequencing of p53 gene, and functional assays) and the complexity of patient cohort composition. Data illustrating an association of p53 with survival always point to a longer survival when p53 is mutated (Tada et al., 1998;Schiebe et al., 2000;Birner et al., 2002;Burton et al., 2002). However, the majority of studies do not validate p53 as an independent prognostic marker for glioblastoma (Kraus et al., 2001;Simmons et al., 2001;Shiraishi et al., 2002;Rich et al., 2005;Ruano et al., 2009;Weller et al., 2009;Levidou et al., 2010;Rossi et al., 2011). Overall it means that the prognostic impact of p53 aberrations is only marginal when considered in a global glioblastoma patient population. Reevaluation of this impact in clinically relevant glioblastoma subpopulations (see above) and association with specific molecular signatures will certainly be of interest in the future.

p53 AND GLIOMA-INITIATING STEM CELLS
Recent studies begin to shed light onto the role of p53 in the regulation of neural stem cells (NSCs). NSCs are self-renewing cells in the central nervous system that can generate both neurons and glia. An elegant study showed that dual inactivation of p53 and PTEN in murine NSC promotes an undifferentiated state with high renewal potential and generates tumors with a high grade glioma phenotype (Zheng et al., 2008). Although the role of p53 in brain tumor stem cells has not been well established, data suggest that loss of differentiation and increase in neurosphere renewal may be linked to the disruption of the p53 pathway in glioma (Molchadsky et al., 2010;Mendrysa et al., 2011;Spike and Wahl, 2011). To achieve a permanent eradication of brain tumors, it is noteworthy that glioma-initiating stem cells have to be considered and in this way their p53 status and functions need to be further explored.

p53 AND TMZ
Despite expressing mainly a wild-type p53 and thus being expected to be sensitive to DNA-damaging agents, primary glioblastoma resist standard therapies including chemotherapy with TMZ. This intriguing observation is in debate and the role of p53 status www.frontiersin.org  Rossi et al. (2011) in response to TMZ has been largely addressed in preclinical studies. Conflicting results have been obtained ( Table 2) and show either an improved capability of TMZ to inhibit cell viability when p53wt is functionally inhibited (Hirose et al., 2001;Xu et al., 2001Xu et al., , 2005aDinca et al., 2008;Blough et al., 2011) or a sensitization of cells to drugs when p53wt is functional (Hermisson et al., 2006;Roos et al., 2007). The former studies suggested that glioma cells with an intact p53 gene are selectively impaired in the proapoptotic functions of p53wt while retaining the potential to mediate relevant DNA repair and cell cycle arrest. Treatment with TMZ induced a persistent cell cycle arrest and an increase in p21 (a cell cycle regulator) in functional p53-expressing cells which showed morphological and biochemical features of senescent cells (Hirose et al., 2001;Martinkova et al., 2010). In cells impaired for p53 function or with a mutant p53, TMZ induced a transient cell cycle arrest and cell death via apoptosis or mitotic catastrophe (Hirose et al., 2001;Martinkova et al., 2010) as well as attenuation of DNA repair (Xu et al., 2005b). When TMZ-triggered apoptosis was reported for both p53wt and p53mutant cells, pathways involved differed with activation of the FAS apoptotic pathway or the mitochondrial apoptotic pathway, respectively (Roos et al., 2007). Thus adverse effects of p53wt activities are increasingly recognized and may participate in chemoresistance of diverse cancers including glioma (Kim et al., 2009;Martinez-Rivera and Siddik, 2012). In one recent report, the effect of p53 status on response to TMZ was explored in glioma-initiating stem cells. It was shown that tumor stem cells are resistant to TMZ when p53 is mutated and sensitive to TMZ when intact (Blough et al., 2011). These data add a new level of complexity in the relationship between p53 status and TMZ sensitivity in glioma.

INTEGRINS AND p53
Although p53 itself is functional in a great majority of primary glioblastoma, inactivation of the p53 signaling pathway occurred in the form of ARF deletions, amplifications of mdm2 or mdm4 leading to p53 signaling alterations in 87% of glioblastoma (Network, 2008). Additionally several oncogenes such as the glioma oncoprotein Bcl2L12 (Stegh et al., 2010) were reported to be overexpressed in p53wt tumors and to impair p53 signaling pathway. Few reports investigated the relationship between integrins and p53 signaling pathways. Both integrin-dependent activation (Lewis et al., 2002) and inhibition (Bao and Stromblad, 2004) of p53 signaling have been suggested in different tumoral settings except gliomas. For example, Stromblad and colleagues demonstrated that α v β 3 integrin impacts negatively on p53wt activity in melanoma cells (Bao and Stromblad, 2004;Smith et al., 2012). We proposed recently that α 5 β 1 integrin plays a similar role in high grade glioma (Janouskova et al., 2012). We demonstrated that overexpression of the α 5 integrin subunit in p53wt U87MG cells impaired the activation of p53 and its transcriptional activity in response to TMZ. Under such conditions, cells became resistant to this alkylating agent. No such effects were found in p53 mutant glioma cell lines. Interestingly, higher levels of α 5 integrin were found in p53wt tumor biopsies than in p53 mutant tumor biopsies suggesting a link between this specific integrin and p53 status in vivo. Our in vitro studies also demonstrated that SJ749 and K34c, two specific non-peptidic antagonists of α 5 β 1 integrin, improved the therapeutic action of TMZ in a p53-dependent way (Martinkova et al., 2010). Molecular pathways involved in the integrin-dependent chemoresistance in p53wt tumors are currently unknown and deserve further studies. Proteins implicated in integrin signaling have been shown to shuttle between the membrane and the nucleus providing a potential mechanism for communication between integrins and p53. In particular FAK, the main kinase activated by integrins, is known to interfere with p53 activity in the nucleus (Lim et al., 2008;Golubovskaya and Cance, 2011). It will also be interesting to discriminate the potential effect of α 5 β 1 integrin on transcriptional and non-transcriptional functions of p53. Our data are the first to demonstrate relationships between an integrin and the TMZ-dependent activation of p53 in glioma possibly explaining the resistance to TMZ of a subgroup of patients. It is tempting to suggest that inhibition of α 5 β 1 integrin by specific antagonists might be an adjuvant treatment to standard therapies in patients expressing high level of α 5 β 1 integrin and p53wt.

CONCLUSION
Despite few and sometimes conflicting data available both on integrin expression and p53 status as prognostic and/or predictive markers for high grade glioma, a reevaluation of their roles is warranted. Due to the growing knowledge on glioblastoma heterogeneity and subclassification, it becomes reasonable to address these questions more accurately in well defined subpopulations of patients. Key issues need still to be addressed before proposing α 5 β 1 integrin expression level and p53 status as relevant biomarkers to stratify group of patients which may be more responsive to TMZ. www.frontiersin.org