Targeting mTOR in Glioblastoma: Rationale and Preclinical/Clinical Evidence

The mechanistic target of rapamycin (mTOR) drives several physiologic and pathologic cellular processes and is frequently deregulated in different types of tumors, including glioblastoma (GBM). Despite recent advancements in understanding the molecular mechanisms involved in GBM biology, the survival rates of this tumor are still disappointing, primarily due to the lack of efficacious treatments. The phosphatase and tensin homolog (PTEN)/phosphatidylinositol-3-kinase (PI3K)/protein kinase B (AKT)/mTOR pathway has emerged as a crucial player in GBM development and progression. However, to date, all the attempts to target this pathway with PI3K, AKT, or mTORC1 inhibitors failed to improve the outcome of patients with GBM. Despite these discouraging results, recent evidence pointed out that the blockade of mTORC2 might provide a useful therapeutic strategy for GBM, with the potential to overcome the limitations that mTORC1 inhibitors have shown so far. In this review, we analyzed the rationale of targeting mTOR in GBM and the available preclinical and clinical evidence supporting the choice of this therapeutic approach, highlighting the different roles of mTORC1 and mTORC2 in GBM biology.


Introduction
In the last decades, we witnessed important advancements in understanding the molecular mechanisms involved in GBM biology; however, GBM remains one of the deadliest types of tumor worldwide [1]. Indeed, the paradigm of medical treatment of chemotherapy plus radiation therapy has reached an efficacy plateau and several drugs designed to target one of the most deregulated pathways in GBM (PTEN/ PI3K/AKT/mTOR) failed to improve the outcome of these patients [2][3][4]. More in detail, the limited blood-brain barrier penetration and the compensatory activation of collateral signaling pathways caused the failure of PI3K inhibitors [1,4]. Additionally, the lack of mTORC2 inhibition that results in AKT activation led to disappointing results of mTORC1 inhibitors in the clinical trials conducted to date [5,6]. On the other hand, emerging preclinical evidence suggests that targeting mTORC2 might provide an efficacious therapeutic strategy for GBM as it can overcome the limitations of mTORC1 inhibitors and pave the way for a personalized targeted treatment. mTORC1 is activated by at least five cues: growth factors, stress, energy status, oxygen, and amino acid concentration. Growth factors, low energy status, low oxygen level, and DNA damage converge on the tuberous sclerosis complex 1/2 (TSC1/2) that acts as a GTPase-activating protein for the GTPase RAS homolog enriched in brain (RHEB), which in turn directly binds mTORC1 resulting into the stimulation of its kinase activity [7] (Figure 1(b)).
However, it has been demonstrated that prolonged treatment with rapamycin might inhibit mTORC2 assembly by sequestering mTOR in some cell cultures [17]. Differently from mTORC1, less is known about mTORC2 upstream activators; it is triggered by growth factors but does not respond to nutrients [7] (Figure 2(b)). Once activated, mTORC2 drives cell proliferation, motility, and survival primarily through the activation of different AGC protein kinases (Figure 2(c)). In fact, mTORC2 phosphorylates protein kinase C (PKC)δ, PKCζ, PCKγ, and PKCε that are involved in cytoskeleton assembly and cell migration, besides AKT on serine 473 [18][19][20][21] (Figure 2(c)).
Intriguingly, it has recently been demonstrated that mTORC2 triggers the activation of the serum and glucocorticoidregulated kinase 1 (SGK1), which is involved in ion transport and cell survival [22] (Figures 2(c) and 2(d)).
The direct involvement of mTORC2 in GBM biology clearly emerged in a Drosophila glioma model obtained by hyperactivating the epidermal growth factor receptor (EGFR), RAS and PI3K. In this model, it has been observed that RICTOR and mSIN1 loss of function prevented tumor formation [23]. Later on, Bashir et al. established that RIC-TOR overexpression alone was sufficient to promote multifocal infiltrating oligodendroglial tumors in the subventricular zone and lateral ventricles of mice that showed an increased mTORC2 activity that sustained cancer stem cell amplification [24]. In addition to the role in tumor induction, mTORC2 activation is also responsible for GBM growth and progression. Indeed, Gulati el al. observed that treatment of GBM cell lines with rapamycin not only resulted in a timedependent decrease of S6K phosphorylation but also caused a paradoxical increase of AKT phosphorylation on serine 473 which is known to be responsible for cell proliferation. The same authors reported that this paradoxical increase of AKT phosphorylation can be reversed by RICTOR but not RAPTOR knockdown by siRNA [25].
mTORC2 is also involved in the induction of the Warburg effect, a metabolic process by which tumor cells metabolize glucose via the aerobic glycolysis also in the presence of sufficient oxygen levels to supply the macromolecular demand of rapidly growing cells (Figure 2(d)). Indeed, by stimulating AKT phosphorylation on serine 473, mTORC2 triggers the expression of the glucose transporter type 4 (GLUT4) and the activation of the glycolytic enzyme hexokinase 2 (HK2) and phosphofruttokinase-1 (PFK-1) [26][27][28] ( Figure 2(c)). Moreover, mTORC2 phosphorylates and inactivates class IIa histone deacetylases (HDACs) that causes Forkhead box O (FOXO)1 and FOXO3 acetylation, resulting in c-MYC release from the suppressive miR-34-c [29] (Figure 2(c)). More recently, it has been shown that the acetyl coenzyme A (acetyl-CoA) derived from glucose and lactate metabolism is used by GBM cells to induce RICTOR acetylation that results in mTORC2 activation; this mechanism creates an autoactivation loop by which mTORC2 triggers cell proliferation and growth, bypassing growth factoractivated upstream signaling and rendering GBM cells resistant to receptor tyrosine kinase inhibitors [30]. Additionally, mTORC2 has recently been reported to confer resistance to the alkylating agent cisplatin via nuclear factor kappa-lightchain-enhancer of activated B cell (NF-κB) activation in an AKT-independent way [31] (Figures 2(c) and 2(d)). In this study, Tanaka et al. demonstrated that NF-κB signaling is upregulated in GBM cell lines and in GBM patients treated with rapamycin and that the inhibition of both mTORC1 and mTORC2 with the mTOR kinase inhibitor PP242 sensitizes EGFRvIII-mutant tumors to cisplatin-induced cell death, confirming the direct involvement of mTORC2 activation in chemotherapy resistance [31]. As previously mentioned with regard to mTORC1, mTORC2 also participates in lipid synthesis (Figure 2(d)). In GBM, the cleavage of SREBP1 that occurs on Golgi membranes and is triggered by mutant EGFR has been demonstrated to be rapamycin insensitive. By contrast, mTORC2 induces SREBP1 cleavage through AKT-dependent and AKT-independent mechanisms, resulting in the expression of acetyl-CoA carboxylase, fatty acid synthase and acyl-CoA synthetase genes, which are all involved in fatty acid and cholesterol synthesis [32,33].

Preclinical Data and Clinical Application of mTOR Kinase Inhibitors
Despite the pivotal role of mTOR in GBM which is now widely recognized, the first generation of mTORC1 inhibitors gave disappointing results in clinical trials. The magnitude of this failure is ascribable to the pharmacodynamic properties of these compounds, as they only target mTORC1, leading to an incomplete inhibition of mTORC1 downstream targets and to the deregulation of a negative feedback following mTORC1 inhibition that results in PI3K reactivation [10]. In addition, the lack of activity against mTORC2 represents another major clinical limitation to the efficacy of rapamycin analogs (rapalogs) [34]. To overcome these limitations, a new generation of ATP-competitive mTOR kinase inhibitors has been developed. These compounds include in vitro and in vivo, respectively [36,37]; moreover, the same authors demonstrated that the sensitivity to CC214 compounds is significantly increased in the presence of EGFRvIII and PTEN loss and that the pharmacologic inhibition of autophagy induced by CC214 sensitizes GBM cells to cell death, preventing a cytostatic effect [37]. Furthermore, Kahn et al. demonstrated that AZD2014 enhances the radiosensitivity of glioblastoma stem cells (GSCs) in vitro and under in vivo orthotopic conditions by inhibiting mTORC1/2 [38]. Additionally, our group has recently demonstrated that the treatment of genetically different GBM cell lines with PP242 but not with rapamycin induces a dramatic and permanent reduction of AKT phosphorylation on serine 473 that not only counteracts tumor growth and invasiveness but also prevents GSC proliferation. Moreover, we also proved that mTORC2 activation is independent from PI3K, as the irreversible inhibition of PI3K with wortmannin is not able to prevent mTORC2 activation, which is evaluable analyzing mTOR phosphorylation on serine 2481 [39].
As preclinical studies confirmed the superiority of mTOR kinase inhibitors compared to rapalogs and have revealed the efficacy of mTORC2 inhibition in counteracting GBM growth, invasiveness, and GSC proliferation, the mTOR kinase inhibitors AZD8055 (AstraZeneca) and OSI-027 (OSI Pharmaceuticals) have already entered clinical trials [40] (Figure 3). A phase I study of AZD8055 (AstraZeneca) in advanced solid malignancies (NCT00973076) and in recurrent GBM (NCT01316809) has completed the recruitment, and results are eagerly awaited; a phase I study of OSI-027 (OSI Pharmaceuticals) in advanced solid tumors and lymphoma started in 2008 and is now completed (NCT00698243). As single agent, OSI-27 has shown to be well tolerated and evidence of activity has emerged [41].

Evaluation of mTOR Activation in GBM Patients
Despite PTEN/PI3K/AKT/mTOR pathway is considered a hallmark of GBM and the inhibition of this pathway represents to date an interesting strategy against this lethal tumor [1,42], the direct evaluation of this pathway activation in GBM patients is not routinely performed or standardized and results are still controversial. In this context, the first analysis of PTEN/PI3K/AKT/mTOR pathway activation in patient specimens was carried out on 45 untreated primary GBM; immunohistochemistry analysis revealed that PTEN loss correlated with AKT activation and that in turn AKT phosphorylation significantly correlated with mTOR, FOXO1, FOXO3a, FOXO4, and S6 activation [43]. Moreover, a tight relation between EGFRvIII expression and the activation of PI3K downstream targets has been also observed in this study [43] (Table 1). Later on, Chakravarti et al. analyzed the expression of total PI3K, AKT, and S6K in 92 gliomas with different malignancy degree by western blot and did not find any difference in the total expression of PI3K, AKT, and S6K between GBM and non-GBM tumors. However, despite the total expression of these protein kinases was unchanged between groups, the authors reported that PI3K, AKT, and S6K phosphorylation was significantly higher in GBM versus non-GBM tumors. Moreover, the levels of PI3K, AKT, and S6K phosphorylation were inversely related to the expression of the cleaved caspase 3 and correlated with radiation resistance and an adverse outcome [44] (Table 1). The activation status of AKT and in addition of NF-κB and STAT3 and their correlation with tumor grade has also been analyzed in 259 diffuse gliomas by Wang et al., by microarray and immunohistochemistry. These authors described consistent AKT and NF-κB activation in tumor samples but not in astrocytes or oligodendrocytes of normal brain cortex and cerebellum; moreover, these authors reported a positive correlation between AKT, NF-κB activation, and tumor grade but not between them and STAT3 [45] (Table 1). In another study, Riemenschneider et al. reported the colocalization of TSC2, mTOR, 4EBP1, S6K, S6, and STAT3 phosphorylation with AKT activation, although only TSC2, S6K, and S6 phosphorylation has been found to correlate with AKT activation [46] (Table 1).
Contrariwise, Fiano et al. did not find any correlation between AKT phosphorylation, cyclin D1, p27/Kip1, and PTEN or EGFR mutations in 65 GBM surgical samples [47]. Instead, Hlobilkova et al. found a strong relation between EGFR expression and tumor grade in 89 samples of glioma with different malignancy degree but they observed comparable levels of AKT phosphorylation between lowand high-grade gliomas [48] (Table 1).
Higher levels of AKT, mTOR, and S6K phosphorylation in high-grade glioma compared with low-grade glioma have been also reported by Li et al., who analyzed 87 tissue samples and found that the percentage of patients with high AKT, mTOR, and S6K phosphorylation, as detected by immunohistochemistry, was greater in grades III and IV than in grades I and II glioma [49] (Table 1). Similarly, Korkolopoulou et al. analyzed the expression of mTOR, S6K, and 4EBP1 phosphorylation in 111 tissue samples (grades II-IV) by immunohistochemistry and validated their analysis in 3 primary GBM cell cultures by western blotting analysis [50] (Table 1). They found that while normal tissues had no positivity for all the proteins considered, none of the tumor samples was negative for mTOR, S6K, or 4EBP1 phosphorylation. With the exception of one sample, tissues that showed mTOR positivity were also positive for S6K staining and all mTOR-positive specimens also showed 4EPB1 coexpression. Moreover, these authors demonstrated that 4EBP1 phosphorylation increased with the histological grade and that mTOR phosphorylation was higher in grade III/IV glioma compared with grade II. Of note, no significant difference in terms of S6K phosphorylation was reported according to tumor grades. They next sought to investigate the correlation between the level of mTOR/S6K/4EBP1 phosphorylation and survival and found that 4EBP1 expression was an independent adverse prognostic index in all the cohort analyzed, while the increased level of mTOR phosphorylation correlated with disease-free survival (DFS) [50] ( Table 1). More recently, Machado et al. demonstrated that mTOR expression was significantly higher in wild-type IDH1 primary GBM, when compared to healthy tissue, and was also higher compared with R132H IDH1-mutant GBM [51]. In addition, these same authors proved that mTOR phosphorylation on serine 2448 and S6 phosphorylation on serine 240 and 244 were increased in wild-type IDH1 GBM compared with R132H IDH1-mutant GBM [51] (Table 1).
Compared with mTORC1 and its upstream and downstream targets, the evaluation of mTORC2 activation in GBM patients is further underestimated. The activation of mTORC2 has been evaluated in 5 GBM cell lines, in 31 tumor samples and 5 normal brain tissues by Masri et al. [52]. These authors observed that the expression of activated mTORC2 was quite undetectable in normal brain tissue while it was high in the tumor cell lines they analyzed. Consistently, they observed increased RICTOR expression and extent of AKT phosphorylation on serine 473 only in tumor cells. Of note, RICTOR overexpression appeared to be independent of PTEN status. In accordance with the in vitro data, these same authors found that the 86% of tumor samples had RICTOR overexpression and 70% of them showed high mTORC2 activity [52] (Table 1). More recently, Alvarenga et al. described RICTOR expression and AKT phosphorylation on serine 473 in 195 patients with brain tumors (38 grade I, 49 grade II, 15 grade III, and 93 grade IV astrocytoma) and correlated AKT activation with overall survival (OS) [53]. They did not find any differences in AKT phosphorylation on serine 473 between low-grade glioma and normal brain tissue but they observed a significant increase in AKT phosphorylation in GBM patients compared with normal brain tissue; moreover, the increased expression of activated AKT correlated with a reduced OS [46]. These same authors analyzed RICTOR expression, and although they did not find an increased expression of this mTORC2 component between normal brain tissue and all grade astrocytoma, they observed nuclear localization of RICTOR in GBM that might suggest a change of its binding partner and a possible implication in tumor progression [53] (Table 1).

Discussion and Future Perspective
As discussed, mTOR pathway is certainly one of the most compelling mechanisms driving GBM biology. However, to date, there are still some cruxes that need to be unraveled to translate the encouraging preclinical results reported in the clinical management of GBM patients.
First of all, the direct evaluation of mTOR pathway activation in GBM patients is not routinely performed and results are affected by discrepancies due to different methodologies of quantification applied by distinct laboratories, which makes it challenging to select those patients that, based on their molecular profile, might mainly benefit of treatment with mTOR kinase inhibitors. As it is now widely accepted that genetic background influences and predicts the outcome of targeted therapy, one of the future challenges for GBM treatment will be to improve the precision and reproducibility of molecular analysis of mTOR pathway activation and standardize this evaluation among laboratories.
Another concern regarding the use of mTOR kinase inhibitors in clinic is their potential immunosuppressive activity, as the impairment of the immune system is widely known to favor tumor growth and progression. However, assays of adaptive immune functions of 1-3 weeks in leukemia revealed that the anticancer properties of PP242 are dominant over its immunosuppressive activity compared with those of rapamycin and another mTOR kinase inhibitor known as PI103 but the outcome of prolonged treatment with this compound and other mTOR kinase inhibitors remains to be investigated [54]. Additionally, as most of the clinical trials evaluating mTOR kinase inhibitors in GBM are ongoing, data regarding the tolerability and the safety profile of these drugs are still not available and the results of these studies will further help to define the role of mTOR as therapeutic target, hopefully providing a new tool against GBM.

Conflicts of Interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.