Anti-Angiogenesis, Gene Therapy, and Immunotherapy in Malignant Gliomas

Malignant gliomas remain associated with poor prognosis and the cause of significant morbidity. In 2010, there was excitement that recent spectacular advancements in our basic understanding of their molecular pathogenesis, angiogenesis and new gene transfer technol‐ ogies will turn the tide in our favor. The negative results of several costly phase III clinical trials are sobering; unfortunately, they take us back to the drawing board in terms of how we can improve our methods and why brain cancer has this incredible ability to resist therapy. This chapter is organized as follows. We start by an overview of the classification and signif‐ icance of malignant gliomas. We proceed to reviewing the molecular pathogenesis of angio‐ genesis and the development of new treatment modalities against anti-angiogenesis targets, some of which were tested in Phase III clinical trials. Before considering immunotherapy strategies and targets for malignant gliomas, we review basic concepts in immunology and discuss the unique immunological features of the central nervous system (CNS). Finally, we discuss gene therapy vectors, strategies, and clinical trials in malignant gliomas. We conclude by an analysis of our current limitations, possible tumor mechanisms for resisting treatments, and what we can do to improve the outcome.


Introduction
Malignant gliomas remain associated with poor prognosis and the cause of significant morbidity. In 2010, there was excitement that recent spectacular advancements in our basic understanding of their molecular pathogenesis, angiogenesis and new gene transfer technologies will turn the tide in our favor. The negative results of several costly phase III clinical trials are sobering; unfortunately, they take us back to the drawing board in terms of how we can improve our methods and why brain cancer has this incredible ability to resist therapy. This chapter is organized as follows. We start by an overview of the classification and significance of malignant gliomas. We proceed to reviewing the molecular pathogenesis of angiogenesis and the development of new treatment modalities against anti-angiogenesis targets, some of which were tested in Phase III clinical trials. Before considering immunotherapy strategies and targets for malignant gliomas, we review basic concepts in immunology and discuss the unique immunological features of the central nervous system (CNS). Finally, we discuss gene therapy vectors, strategies, and clinical trials in malignant gliomas. We conclude by an analysis of our current limitations, possible tumor mechanisms for resisting treatments, and what we can do to improve the outcome.

Summary of angiogenesis
Angiogenesis is the process by which the vascular system is formed through growth of new capillaries from pre-existing vessels (Plate, Scholz et al. 2012). Angiogenesis plays a critical role in key physiologic and formative processes such as embryogenesis, regeneration, and wound healing. Angiogenesis is also involved in various pathologic processes including age-related macular degeneration, rheumatoid arthritis, and tumor growth and development (Wang, Fei et al. 2004).
The process of angiogenesis can be briefly summarized as follows. First, there is vasodilation, in response to nitric oxide, and increased permeability of the existing vessels. This is followed by degradation of the existing vessel's basement membrane. Next, endothelial precursor cells migrate to the area and begin to proliferate and mature into capillaries via a balance of both growth and inhibition. The final steps involve recruitment of vascular smooth muscle cells and pericytes that form a new network of mature vessels (Shinkaruk, Bayle et al. 2003).

Molecular signals of angiogenesis
Although there are numerous factors and signals that contribute to angiogenesis, the chemical signal that seems to play the most critical role in the process is Vascular Endothelial Growth Factor (VEGF). VEGF is a pro-angiogenic growth factor, which is secreted by many cells, including mesenchymal, stromal, and especially tumor cells. VEGF induces the migration of the endothelial precursor cells to sites of angiogenesis and is responsible for their proliferation and differentiation. The VEGF gene is located on chromosome 6p12 and the gene family is composed of five members, namely VEGF-A, VEGF-B, VEGF-C, VEGF-D, and placentalderived growth factor (PlGF). Of these, VEGF-A, B, and PIGF are involved in the development of the vascular system and VEGF-C and D are involved in the development of the lymphatic system (Ahluwalia and Gladson 2010). VEGF primarily signals through its receptor VEGFR2 which is a tyrosine kinase receptor that is expressed by many cells, including endothelial cells, endothelial cell precursors, and tumor cells (Jain, di Tomaso et al. 2007). Other chemical signals that play an important role in angiogenesis are fibroblast growth factor, hepatocyte growth factor (HGF), tumor necrosis factor-alpha (TNF-α), transforming growth factor-beta (TGF-β), angiopoietins, and platelet derived growth factor (PDGF). Their various roles include involvement in extracellular matrix degradation, endothelial proliferation and migration, and neovessel stabilization and maturation (

Angiogenesis in tumors
Seven different cellular mechanisms appear to contribute to tumor angiogenesis: (1) classical sprouting angiogenesis, (2) vascular co-option, (3) myeloid cell-driven angiogenesis, (4) vessel intussusception, (5) vasculogenic mimicry, (6) bone marrow derived vasculogenesis, and (7) cancer stem-like cell derived vasculogenesis ( The role of the remaining four mechanisms in glioma angiogenesis is not yet fully understood. Briefly, vessel intussusception is the process by which a new vessel is formed by internal division of the pre-existing capillary plexus without sprouting through a series of steps that include vascular invagination, intra-luminar pillar formation and remodeling, and splitting (Djonov . It is not the goal of this chapter to study these mechanisms in detail, but instead to provide an overview of angiogenesis in glioma and discuss key molecules involved and possible therapeutic options that target them.
bind to its target receptors on endothelial cells. This neutralization has been shown to have efficacy not only in in vitro studies, but also in in vivo ones.
Recently, two phase III clinical trials investigating Bevacizumab as a first-line treatment for newly diagnosed GBM tumors were completed. Unfortunately, both trials were consistent in showing no statistically-significant prolongation of overall survival time (OS) but there was a slight improvement in progression-free survival time (PFS). The two trials had a similar design, namely double-blinded prospective trials where newly diagnosed GBM patients were randomized to either standard of care with Bevacizumab or with placebo; the standard of care consisted of radiation therapy with adjuvant and concomitant Temozolomide. A total of 637 and 921 adult participants were randomized in the Radiation Therapy Oncology Group (RTOG) and AVAglio trials, respectively. The median OS was 16.1 vs. 15.7 months, in the RTOG trial (p = 0.11). The median PFS was longer in patients who received Bevacizumab, 7.3 vs. 10.7 months (p = 0.004) and 6.2 vs. 10.6 months (p<0.0001) in the RTOG and Avaglio trials, respectively. In addition, the results also showed a higher incidence of adverse reactions in the Bevacizumab arm, including neutropenia, hypertension, and deep vein thromboembolism and pulmonary emboli (Gilbert, Dignam et al. 2013). The AVAglio trial noted delayed time to definitive deterioration in terms of health-related quality of life (p<0.0001) and Karnofsky Performance Scale, and increased time to corticosteroid initiation (HR 0.71, 95% CI 0.57-0.88; median 12.3 vs. 3.7 months) (Henriksson, Bottomley et al. 2013). These results are discouraging and do not justify the use of Bevacizumab in a GBM patient who has had a reasonable surgical resection.
The data support the idea that Bevacizumab may be reserved until the time of recurrence as several prospective phase II clinical trials have shown prolongation of the 6-month PFS rates ranging from 25 to 42.6 percent and median OS times from 6.5 to 9.2 months. However, a significant limitation of these trials is that the comparison was made to historical controls (Friedman, Prados et al. 2009;Kreisl, Kim et al. 2009;Raizer, Grimm et al. 2010).

VEGF-trap
VEGF-Trap (drug name Aflibercept) is a recombinant fusion protein that acts as a decoy receptor for VEGF, thereby blocking its interaction with its normal receptors and interrupting the VEGF signaling pathway (Holash, Davis et al. 2002). VEGF-Trap was developed by incorporating domains of both VEGF receptor 1 and VEGF receptor 2 fused to the constant region of human immunoglobulin G1. VEGF Trap has a high affinity for all isoforms of VEGF-A, as well as for PlGF, another pro-angiogenic agent that primarily acts on VEGF receptor 1 (Holash,

Sunitinib
Sunitinib is a small-molecule inhibitor of VEGF receptors 1 and 2, PDGFR alpha and beta, stemcell factor receptor (SCFR), fms-like tyrosine kinase 3 (FLT-3), colony-stimulating factor-1 receptor (CSF-1R), and the RET oncogene tyrosine kinase (RET) (Chow and Eckhardt 2007;Kreisl, Smith et al. 2013). It has FDA approval for use in metastatic renal-cell carcinoma, gastorintestical stromal tumors refractory to imatinib mesylate, and advanced pancreatic neuroendocrine neoplasms (Kreisl, Smith et al. 2013). Recently, a phase II clinical trial was completed investigating the role of continuous daily Sunitinib in recurrent GBM in both Bevacizumab exposed and Bevacizumab naïve patients (Kreisl, Smith et al. 2013). Unfortunately, the results did not demonstrate an improvement in PFS in either population. Recent evidence by Costa et al suggests that silencing of micro-RNA 21 (miR-21), a small, non-coding RNA that regulates gene expression, may enhance the anti-tumoral effect of Sunitinib (Costa, Cardoso et al. 2013).

Nintedanib
Nintedanib (BIBF 1120) is a small, orally available triple angio-kinase inhibitor that targets VEGF receptors 1-3, FGFRs 1-3, and PDGFR alpha and beta. It is still in phase III development, but preclinical models demonstrated effective growth inhibition of both endothelial and perivascular cells when the above listed pathways were simultaneously interrupted (

Vandetanib
Vandetanib is a multi-targeted tyrosine kinase inhibitor of VEGF receptor 2, epidermal growth factor receptor (EGFR) 2, and the rearranged-during-transfection oncogene that results in the simultaneous blocking of several pathways, including angiogenesis (Kreisl, McNeill et al. 2012

Integrins
Integrins are cell surface receptors that play key roles in mediating the migration of endothelial cells. They are receptors for many different extracellular matrix (ECM) ligands and they play an important role in angiogenesis via the processes of integrin-mediated adhesion, migration, proliferation, survival, and differentiation of cells that form the vasculature (Hynes, Bader et al. 1999;Tchaicha, Mobley et al. 2010). The αv integrin subfamily has five members-αvβ1, αvβ3, αvβ5, αvβ6, and αvβ8-and the αvβ8 member, in particular, has been shown in mouse models to be a central regulator of angiogenesis in the developing brain (

Ligands
Notch signaling in host endothelial cells is important for angiogenesis. Recent evidence has shown that delta-like ligand 4 (DLL4), a member of the Notch ligand family, is expressed in tumor cells and can activate Notch signaling in host endothelial cells and can therefore affect the vascular function of tumors. In fact, DLL4 expression appears to be regulated by VEGF and the tumor's hypoxic microenvironment (Patel, Li et al. 2005;Li, Gong et al. 2012). Increased levels of VEGF lead to an up-regulation of DLL4 expression which results in endothelial cells expressing Notch receptors to down-regulate VEGF-induced vessel sprouting and branching and ultimately resulting in productive and efficient angiogenesis (Li, Gong et al. 2012). Furthermore, it has also been demonstrated that blockade of DLL4 can result in non-productive angiogenesis by causing tumor growth inhibition and a decrease in tissue perfusion (Scehnet, Jiang et al. 2007;Li, Gong et al. 2012). Li et al. recently conducted a study to investigate the role of DLL4 in malignant gliomas, specifically in terms of vascular quantity and quality and showed that DLL4 expression was significantly up-regulated in malignant human gliomas as compared to normal brain tissue. Additionally, they also demonstrated that DLL4-positive malignant glioma tissues have increased proliferation of vascular endothelial cells and pericyte recruitment, as compared to DLL4-negative malignant glioma tissue, and that DLL4-positive tissues had a higher vessel maturation index (VMI). These results provide evidence that DLL4 inhibition may alter glioma vessel maturity and, in turn, may improve the effects of antiangiogenic agents (Li, Gong et al. 2012).

Gamma secretase
Gamma secretase is a pre-senilin dependent protease that acts as a regulator of angiogenesis through a series of complex steps that are beyond the scope of this chapter. However, part of its role in angiogenesis is related to Notch signaling (Jain, di Tomaso et al. 2007; Boulton, Cai et al. 2008). RO4929097 is a potent and selective gamma secretase inhibitor of Notch signaling that is being investigated as an anti-tumor agent. Phase I studies have demonstrated safety and phase II studies are underway to assess its role in recurrent GBM when given alone and in combination with Bevacizumab (Tolcher, Messersmith et al. 2012).

Transforming growth factor beta (TGF-β)
TGF-β is a multifunctional protein that is involved in the regulation of proliferation, differentiation, and survival of many cells, including glioma cells and endothelial cells (Bertolino, Deckers et al. 2005). TGF-β1 and TGF-β2, members of the TGF-β family, stimulate expression of VEGF, the plasminogen activator inhibitor, and some metalloproteinases that are involved in vascular remodeling, angiogenesis, and degradation of the extracellular matrix. Animal models demonstrate that inhibitors of TGF-β signaling reduce viability and invasion of gliomas (Kaminska, Kocyk et al. 2013). Fresolimumab, a human monoclonal antibody that inactivates all forms of TGF-β, is being investigated as a potential therapeutic for glioma (Trachtman, Fervenza et al. 2011).

Topoisomerase I inhibitors
Topoisomerase I is critical for efficient DNA replication and cell division. Topoisomerase I activity is increased in malignant gliomas and inhibitors of topoisomerase I activity, such as Camptothecin, Irinotecan, and the indolocarbazoles, have been tested as potential glioma therapies (Pommier 2006

Oncoproteins
B-cell specific Moloney murine leukemia virus integration site 1 (Bmi-1) is an oncoprotein that plays a role in the development and progression of cancers including breast, lung, prostate, and interestingly, brain (Jagani, Wiederschain et al. 2010). Bmi-1 is a member of the Polycomb gene family of proteins that function as epigenetic silencers of genes that control self-renewal, differentiation, and proliferation; dysregulation of Bmi-1 has been associated with cancer cell proliferation, invasion, and repression of apoptosis or senescence (

Carboxyamidotriazole orotate (CTO)
CTO is a triazole orotate formulation of carboxyamidotriazole (CAI), which is an inhibitor of receptor-operated calcium channel-mediated calcium influx. CTO has anti-proliferative, anti-invasive, as well as anti-angiogenic properties in several human cancer cell lines including glioblastoma (

TRC105
TRC105 is a novel, first-in-class antibody against endoglin (CD 105), an endothelial cell receptor that is essential to angiogenesis and acts primarily through its effects on TGF-β and BMP-9 signaling. A phase I trial conducted by Rosen et al. demonstrated that this drug is welltolerated at clinically relevant doses and multiple phase II trials are ongoing to evaluate its potential role in other malignancies, including malignant glioma (Rosen, Hurwitz et al. 2012).

Thalidomide and lenalidomide
Thalidomide and its analogue Lenalidomide have both been shown to have anti-angiogenic and anti-tumor effects in preclinical models (D' Amato, Loughnan et al. 1994;Short, Traish et al. 2001). The anti-angiogenic effects are thought to be related to a hepatic metabolite that inhibits endothelial growth, although the exact mechanism is unclear (Short, Traish et al. 2001). Additionally, an early clinical trial conducted by Baumann et al. showed that the combination of Thalidomide with Temozolomide appeared to be more effective than Thalidomide alone in the treatment of GBM (Baumann, Bjeljac et al. 2004). Additional studies are underway to examine the role of Thalidomide in combination with other anti-glioma agents.

Tandutinib
Tandutinib (MLN0518) is an active inhibitor of type III receptor kinases with activity against PDGF receptors alpha and beta, FLT3, and c-KIT. Its anti-angiogenic effects appear to be mediated by interruption of PDGF/PDGFR. It is currently being investigated in combination therapy with other agents against malignant glioma (Boult, Terkelsen et al. 2013).

Immunotherapy for malignant gliomas
Our immune system can be viewed as an intricate balance of opposing functions that lead to either immunity or tolerance. Perturbations that disrupt this stable equilibrium could lead to autoimmune disease or tolerance to malignant cells. In general, the immune system has the ability to recognize and to react to foreign antigens, which leads to their removal as well as to the destruction of cells that express them. Before attempting immunotherapy for cancer, one needs to understand the crucial balancing acts of the immune response that eventually lead to a desired outcome; in addition, the central nervous system has unique features that require special considerations.
In this section, our goals are to introduce the readers to the basics of peripheral immunology focusing on how foreign antigens activate the immune response leading to immunity vs. tolerance. Nevertheless, a detailed discussion of immunity is not within our scope; in some disciplines, we will just be scratching the surface. We will detail antigen processing and presentation, T cell priming, with attention to the synapse between T cells and the antigenpresenting cell (APC) and clonal expansion. Because of our interest in brain tumors, we will compare the systemic immune response to that of the CNS, discussing the historical thoughts of the immune privileges of the CNS and more recent evidence of processing of CNS antigens via the glymphatic pathway. The stage will be set for a discussion of immunotherapy for brain tumors, including priming in the periphery, priming in the CNS, and passive transfer of immunity. The last section lists the clinical trials that employ immunotherapy for brain tumors and their proposed modes of action.

Peripheral immunology
As part of the adaptive immune system, antigens enter the body through epithelium and are immediately met in the infected tissue by APCs, most commonly dendritic cells (DCs), which then process the antigens into protein fragments (

T cell priming and activation
T cell priming by DCs induces activation, cytokine secretion, and clonal proliferation (Mempel, Henrickson et al. 2004). For T-cell activation, both the MHC-bound antigen and the MHC itself must be recognized by the T-cell receptor (TCR) and the co-receptor, respectively (Abbas, Lichtman et al. 2014). This process of priming naïve T cells into effector and helper cells occurs in lymphoid organs (Joffre, Segura et al. 2012). DCs prime the T cells during three stages: 1. Contact for exchange of information between the T cell and the dendrite in the lymphocyte pool, 2. The formation of a stable bond followed by secretion of interleukin-2 and interferonγ, and 3. Rapid movement and clonal expansion (Mempel, Henrickson et al. 2004).
The synapse between the T cell and the APC requires the interaction of not only the TCR and MHC, but also adhesion molecules and co-receptors to receive signals from the APC (see section 5.1.3 below). Early on during this process various cytokines are released. Certain cytokines lead to clonal expansion of antigen-specific lymphocytes, some of which become differentiated into effector T cells that can remove infected cells. Others differentiate into memory T cells that serve to remain inactive until re-exposed to the same antigen (Abbas, Lichtman et al. 2014). During future encounters, DC-bearing antigens will migrate to the paracortical region in the lymph node to search for a T cell that recognizes its antigen, ultimately activating clonal expansion (Bousso, 2003).

T lymphocyte -Antigen presenting cell contact
To begin the process of priming T cells, the DC must physically contact the naïve T cell. This process tends to occur in lymphoid tissue, specifically in the draining lymph nodes, spleen, and Peyer's patches, after infiltration of antigen-APC complexes from peripheral tissue through lymph vessels ( In the presence of antigen, the data supports a three-phase model. First, within a few hours of lymph node entry, contact between naïve T cell-DC with peptide increases in duration, now lasting up to five minutes. Within ten hours of antigen entry into the lymph node, mobility slows dramatically as T cells and DCs form more stable bonds, which will persist from two to twenty four hours. This step also promotes the up-regulation of activation markers. After thirty hours, the bonds separate and this is followed by increased mobility of T cells, corresponding with T cell proliferation (

Naïve T lymphocyte -Antigen presenting cell synapse
Naïve T cells are constantly searching for presented antigen on the MHC-antigen complex of mature DCs, from which the T cell and its receptor will require co-stimulation (Mempel, Henrickson et al. 2004). T cell activation relies on the successful synapse of the T cell receptor (TCR) with the peptide-MHC complex on the APC. Additionally, several signaling complexes must connect between the TCR and the adaptor protein linker for activation of T cells and subsequent filamentous actin (F-actin)-dependent TCR cluster formation (Dustin and Depoil 2011). The role of co-stimulatory and co-inhibitory proteins is to modulate the TCR signal to increase or decrease activation of the T cell or to direct the response of that cell down a particular differentiation pathway (Dustin and Depoil 2011).
Antigen recognition and adhesion involves simultaneous recognition of many molecules. In the receptor layer, antigen recognition occurs by the TCR to the peptide with its co-receptor CD4 or CD8 binding MHC II or MHC I, respectively. This is the first step in the signal cascade. For signal transduction and co-stimulation, several transmembrane signaling molecules, including CD3 and ζ chain, form part of the TCR complex and bind the MHC/antigen complex. Additionally, CD28 (and CTLA-4) on the T cell binds B7-1 (CD80)/B7-2 (CD86) on the APC (Dustin and Depoil 2011). The B7 proteins are created by APC in response to an antigen to ensure that T cells are not activated by self-antigens. This key bond is essential for signaling and thus activation of naïve T cells. Concurrently, the CD40 Ligand on the T cell and CD40 on the APC unite and promote increased production of B7 and secretion of cytokines in the APC in order to further encourage T cell activation ( The co-stimulatory signals play a key role in determining immunity or tolerance. Due to the required co-receptors and signal transduction, many mechanisms are set in place to prevent T cells from activating against self-protein. Through early central tolerance mechanisms designed to prevent autoimmune disease, immature T cells that react to self-proteins undergo apoptosis early in development (Luptrawan, Liu et al. 2008). This activation-induced cell death is assisted through the interaction of Fas, which is expressed everywhere and in high concentration in the thymus, with Fas Ligand on T lymphocytes and NK cells (Maher, Toomey et al. 2002); a similar process results in clonally expanded T cells after they are no longer needed. Similarly, if a T cell encounters an antigen on an APC without the appropriate co-stimulation, it is susceptible to developing tolerance to that antigen such that on future encounters it will ignore it, even if given the appropriate co-stimulation (

Clonal expansion
To amplify activation, T cells and APCs secrete various cytokines. Initially, T cells secrete interleukin-2 (IL-2), which facilitates the binding of IL-2 by augmenting the presence of IL-2 receptors. IL-2, by acting on the T cell that secreted it, supports the production of T cells specific to the antigen. IL-2 is also needed to maintain regulatory T cells (Abbas, Lichtman et al. 2014).
Clonal expansion transpires in 1-2 days, leading to the creation of antigen-specific CD4+ and CD8+ cells. CD8+ cells develop into effector cells that ultimately migrate to the site of infection to interact with the specific antigen to which they were primed. CD4+ cells further develop into T helper I (Th1) and T helper 2 (Th2) lymphocytes. Antigen-exposed macrophages and DCs release IL-12 and natural killer cells secrete interferon-γ, thereby promoting the differentiation of Th 1 cells (Abbas, Lichtman et al. 2014). Th1 cells secrete IL-2, interferon-γ, and lymphotoxin-a which leads to type 1 immunity with enhanced macrophage activation and phagocytosis (Spellberg and Edwards 2001). Interferon-γ also increases the expression of MHC I and II molecules to amplify antigen presentation (Spellberg and Edwards 2001). Th2 cells, stimulated by IL-4, also release IL-4, IL-5, IL-9, IL-10, and IL-13 promoting production of antibodies and type 2 immunity, which minimizes phagocytosis and decreases inflammation (Spellberg and Edwards 2001). In times of overwhelming systemic response or immunosuppression, a type 2 response can supersede the appropriate type 1 response (Spellberg and Edwards 2001).
All the aforementioned steps have to be executed flawlessly to achieve immunity against a tumor or a tumor antigen. The body is set up to have low affinity to self-antigens, as would be expressed on tumor cells (Luptrawan, Liu et al. 2008). Along those lines, dendritic cells, which have been discovered in tumors, play a large role in presentation of tumor antigens; however, that does not necessarily predict the nature of the immune response. In particular, any flaw in dendritic cells, from cross presentation to IL-12 production, will lead to tolerance and impaired CD8+ T cell response to tumors (Joffre, Segura et al. 2012).

Central nervous system immunology
The CNS was once thought to be immunologically-privileged because of its unique immunological features, including 1) lack of immunological surveillance due to low expression of MHC molecules, 2) lack of distinct lymphatic drainage, and 3) protection by the blood brain barrier (BBB), which limits the movement of naïve T cells into the CNS (Chavarria and Cardenas 2013). Nonetheless, the CNS has more recently been discovered to have a finely tuned immune surveillance managed by APC, believed to be microglia, DC, perivascular macrophages and meningeal dendritic cells

Centrally-acting peripheral immune cells
In addition to resident microglia, the primary immune cell in the CNS, peripheral immune cells including peripherally activated T lymphocytes, macrophages, and DC circulate in small numbers within the CNS. They are predominantly in specialized CNS compartments located outside the parenchyma with ability to gain access to the parenchyma through various mechanisms that include the choroid plexus, perivascular or Virchow-Robin spaces, meningeal vessel branch points into the subarachnoid space, and through post-capillary venules (  Similar to peripheral tolerance, if there is insufficient co-stimulatory response, the interaction of Fas ligand (FASL) on microglia and Fas receptor on the T cell leads to activation-induced T cell apoptosis. Microglia also express FAS molecules themselves, which induce apoptosis upon binding FASL (Yang, Han et al. 2010). Nitric oxide released by microglia in response to activation can also potentiate effector cell death (Yang, Han et al. 2010). Furthermore, microglia display B7-H1 molecules which also support immunosuppression by stimulating T cell apoptosis (Yang, Han et al. 2010). Additionally, glycoprotein CD200 down regulates activated microglia (via CD200 ligand on microglia) and perivascular macrophages in the CNS, acting as an anti-inflammatory and serving to keep microglia in a quiescent state (Ousman and Kubes 2012; Chavarria and Cardenas 2013).

The glymphatic pathway
The CNS lacks lymphoid tissue. For appropriate immune surveillance, both antigens and APC must be able to travel to lymphoid tissue, ideally via lymphatic channels, for T cell priming (Romo-Gonzalez, Chavarria et al. 2012). For small and hydrophobic molecules as well as transporter substrates, exit through the BBB is easy. Other substances are cleared from CSF through arachnoid granulations or peripheral lymphatics on cranial nerves. Clearance of large particles and matter deep within the parenchyma is more difficult and is ascribed to a high rate of flow of interstitial fluid (Iliff and Nedergaard 2013). This flow of fluid transports antigens from the brain parenchyma for presentation in cervical lymph nodes (Romo-Gonzalez, Chavarria et al. 2012). Through what has been termed the glio-vascular or glymphatic pathway, interstitial solutes are cleared from the brain to the peripheral lymphatic system via perivascular water channels from the para-arterial CSF influx pathway through the interstitium and along the para-venous clearance route (Iliff and Nedergaard 2013). Once filtered from the CNS, antigens are captured by APCs in the cervical lymph nodes and activate lymphocytes that then migrate to the CNS in search for remaining antigens (Romo-Gonzalez, Chavarria et al. 2012). As noted above, several hypotheses for lymph-like drainage of antigens exist including efferent flow via CSF and interstitial fluid past the optic, trigeminal, and acoustic nerves to the cervical lymph nodes, reabsorption through arachnoid villi into the venous sinuses, and through perivascular APC including macrophages and DC (Romo-Gonzalez, Chavarria et al. 2012).
A more recent study by Jarry et al. showed that adult microglia can cross-present antigen to naïve CD8+ T cells for priming if there is appropriate microglial activation (Jarry, Jeannin et al. 2013). Their study involved injecting naïve T cells into the brain, as the natural presence of naïve T cells in the brain is limited, with restriction of entry to activated T cells instead. Their study suggests that if naïve T lymphocytes are given the ability for entry into the brain, typically during stressful inflammatory illnesses, coupled with the appropriate microglial response, cross-priming of naïve T cells is possible (Jarry, Jeannin et al. 2013).

Glioma-associated microglia
Glioma-associated microglia/macrophages cannot mount a successful anti-tumor T cell response (Yang, Han et al. 2010). Microglia, along with some T lymphocytes, infiltrate gliomas in a pattern that was initially thought to be an immune response against tumor cells but has been more recently realized to actually encourage tumor growth by promoting immunosup- In addition to altering the response of microglial cells, gliomas take an active role in downregulating the immune response. Recent data has shown that reduced phagocytic activity by glioma-associated microglia stems from defective antigen presentation for T cell activation due to decreased MHC II expression as well as suppression of pro-inflammatory cytokine (TNFα) release, especially in high-grade gliomas (

Immunotherapy strategies and targets
Immunotherapy for malignant gliomas is based on various strategies aimed at the induction of anti-tumor immunity. Nevertheless, though curing a mouse from a brain tumor using immunotherapy is rather easy, this goal has proven to be more challenging in humans especially when coupled with the globally impaired immune response and increased tumor tolerance in patients with GBM (Luptrawan, Liu et al. 2008). Because of the aforementioned data, the goal of immunotherapy for gliomas should be not only to activate the cytotoxic T cell response, but also to counteract the active immunological depressive effects by the tumor itself. We will not list an exhaustive search of all immunotherapeutic strategies but will instead discuss an outline of the approaches than can be used. A thorough discussion can be found in Okada et al. (Okada, Kohanbash et al. 2009). Here, we will emphasize the different categories and discuss limitations of immunotherapy.

Priming in the periphery
Initiating an immune response against tumors is typically difficult due to poor antigen presentation and the active immunosuppressive effects by tumor cells (Luptrawan, Liu et al. 2008). Peripheral vaccination has been performed using purified antigen and irradiated genetically modified tumor cells. Through vaccination with a tumor antigen, one hopes to induce an immune response peripherally, which translates to CNS immunity as activated T cells cross the BBB. This goal may be achieved by processing the antigen via APCs at the subcutaneous injection site, migration to lymph nodes, and priming naïve T cells. Nevertheless, choosing an appropriate antigen is crucial so as to avoid an autoimmune response causing encephalitis (Okada, Kohanbash et al. 2009).
Peptide-based vaccines (see Table 1) for glioma epitopes are synthetically derived for specific antigens and run less risk of autoimmune encephalitis. This process has the potential to be individually tailored based on assessment of the patient's peripheral blood for positive response to the various antigens (Okada, Kohanbash et al. 2009). Many antigen epitopes exist and will be briefly covered. Il-13Rα2 appears as a membrane protein in more than 80% of gliomas but not in normal brain tissue, making it a target for immunotherapy ( In addition to using purified antigen as above, whole glioma cells may be used for vaccination (See Table 2). In this process, tumor cells, either autologous or allogeneic, are grown in vitro, irradiated, and injected back into the patient (Wikstrand and Bigner 1980;Zhang, Eguchi et al. 2007). The benefit of whole cell vaccinations is the availability of multiple associated antigens and, specifically, the ones expressed by the individual patient's glioma (Okada, Kohanbash et al. 2009).
As a means of bypassing local antigen presentation at the site of the tumor, DC vaccination has also been a source for many clinical trials with various techniques of uniting the DC with the antigen (See Table 3

Priming in the brain
Fathallah-Shaykh et al. showed that priming in the brain elicits an anti-tumor response leading to destruction of the brain tumor as well as to anti-tumor systemic immunity in animals (Fathallah-Shaykh, Gao et al. 1998). The basic mechanisms for eliciting such an immune response in the CNS are detailed above. One possible method consists of injecting DCs directly into the tumor; the goal is to enhance local antigen processing followed by glymphatic drainage and priming in cervical lymph nodes (Luptrawan, Liu et al. 2008). Early preliminary results in humans are encouraging. In a study of 10 patients with glioma, half received subcutaneous vaccination of pulsed DC with autologous tumor lysate and the other half received both subcutaneous vaccine and intra-tumoral injection of immature autologous DC. On follow-up imaging, the patients who received both therapies showed diminution of contrast-enhancing tumor (Yamanaka, Yajima et al. 2003;Luptrawan, Liu et al. 2008). A phase I/II trial including 24 patients with Grade III or IV glioma at first recurrence evaluated the safety and benefits of DC immunotherapy given either via subcutaneous injection near a cervical lymph node or both subcutaneously and intra-tumorally via an Ommaya reservoir. The study revealed that patients with both intratumoral and intradermal administrations had a longer survival times than patients with intradermal administration only (Yamanaka, Homma et al. 2005; Luptrawan, Liu et al. 2008). Another method, which has shown promise in animal models was used by Choi et al and involves the injection of chimeric antigen receptors-transduced T cells targeting EGFR variant III into mice gliomas. The results show a dose-dependent increase in survival, while at the same time sparing cytotoxicity to normal brain tissue (Choi, Suryadevara et al. 2013).

Passive transfer of immunity
In passive immunotherapy, the patient is given effector cells or molecules. Such therapies include monoclonal antibodies, radio-nucleotides that are conjugated to monoclonal antibodies, coupled toxins, and T cells.

Transfer of monoclonal antibodies
The use of monoclonal antibodies (see Table 4) for CNS targets necessitates overcoming important barriers (Okada, Kohanbash et al. 2009); for instance, the size of monoclonal antibodies, around 150kDa, impairs their diffusion into the CNS. However, evidence suggests that the BBB both in normal patients and those with malignancy tolerates the entry of monoclonal antibodies (Chen and Mitchell 2012). Additionally, antibodies bound to the tumor boundary layer create a concentration gradient that makes it difficult for additional antibodies to permeate against a concentration gradient, essentially not being able to reach the core of the tumor. This option may be more valid for use in conjunction with surgical resection and convection enhanced delivery (CED) where the agent of choice is given at high pressure and in bulk through an intracranial catheter into the brain tumor and parenchyma (Okada, Kohanbash et al. 2009). As opposed to using diffusion, this method uses bulk flow and has been implemented in several clinical trials. While bypassing the BBB and limiting systemic toxicities, a limitation of this method is that it can be slow and thus difficult to deliver high volumes of molecules (Bobo, Laske et al. 1994 Radio-immunotherapy (RIT) via radionucleotides conjugated to monoclonal antibodies is another technique for targeting specific tumor antigens (see Table 5). This technique delivers localized, cytotoxic radiation to tumor cells resulting in cell death. This method is used concurrently with surgical resection into the surgical cavity. Specifically, antitenascin has been most studied for RIT due to high prevalence of the glycoprotein tenascin on the surface of high-grade gliomas, including 90% of GBMs (Zalutsky 2004). Duke University has developed the specific antibody 81C6, which has shown promise when given into the tumor cavity concurrently with resection (Zalutsky, Moseley et al. 1989;Okada, Kohanbash et al. 2009). Several other clinical trials have used similar approaches with RIT and glycoprotein tenascin. Other targets include the DNA/Histone H1 complex, the extra domain B of fibronectin, and the alpha chain of the IL-2 receptor (Okada, Kohanbash et al. 2009).

Transfer of ligands (cytokines)
Via coupled targeted toxins, cytokines fused with toxins can be delivered to tumor cells (see Table 6). Specifically, IL-4R and IL-13Rα2 expression is increased in high-grade gliomas making them ideal targets for chimeric fused proteins. For these chimeras, pseudomonas exotoxin is fused to IL-4 and IL-13, creating IL4-PE and IL13-PE, respectively (Debinski, Obiri et al. 1995;Joshi, Leland et al. 2001). These proteins are then delivered via CED (Okada, Kohanbash et al. 2009). By combining toxins with cytokines, one can target tumor receptors and induce cytotoxicity. Additional chimeras have been made using diphtheria toxin, which bonds to transferring, and TGFα, which binds to Pseudomonas exotoxin.

Transfer of cells
For the adoptive transfer of tumor-reactive autologous cytotoxic T lymphocytes (see Table 7), various techniques are used to create an antigen-specific receptor on a CD8+ T cell that can prompt T cell activation (Okada, Kohanbash et al. 2009). This process has previously been used in conjunction with IL-2 infusion for the treatment of melanoma. Antigen-specific cytotoxic T cells from peripheral blood or from tumor nodules are isolated from the patient. The T cells will then undergo clonal expansion in vitro with specificity for tumor antigen, possibly with the aid of IL-2. These cells are then returned to the patient where they would in theory perform cytotoxic responses upon recognition of tumor-associated antigen in the brain parenchyma (Okada, Kohanbash et al. 2009). In terms of usage in gliomas, the first steps would be creating a library of highly avid cytotoxic T cell clones from which to build highly selective TCR gene pairs to create transgenic cytotoxic T cells. Adoptive therapy is not limited to CD8+ T cells and has been also tried using NK cells and CD4+ T cells, both of which can be similarly removed, expanded, and injected back into the patient. Blancher et al treated 13 GBM by recombinant IL-2, with and without lymphokine activated killer cells, given directly via a catheter into the tumor resection bed. Unfortunately, the treatment had no effects on tumor progression. The adverse reactions included cerebral edema, confusion, and fever (Blancher, Roubinet et al. 1993).
Some obstacles with the adoptive process include creating T cells with TCR of appropriate avidity (Okada, Kohanbash et al. 2009). Further difficulties arise with T cell reproduction; many of these specialized T cell populations are thought to be terminally differentiated and thus unable to propagate long-term existence (Wherry, Teichgraber et al. 2003). Additionally, these transgenic T cells must also overcome the immunosuppressive features of GBM and, in fact, do so better than natural T cells due to the ability to manipulate them and strengthen them with specific chemokines and integrin receptors (Okada, Kohanbash et al. 2009).

Limitations of immunotherapy
The limitations of immunotherapy for malignant gliomas include: 1) physical obstacles of drug administration due to the BBB, 2) direct and indirect down-regulation of the immune response by gliomas, 3) the high mutation rate of the tumor, which will select for tumor cells that do not express the target of the immune response. In fact, cancer genomes are unstable as evidenced by microsatellite instability of the tumor cells, which aides in tumor evolution and progression (van de Kelft and Verlooy 1994; Yip, Miao et al. 2009;Milinkovic, Bankovic et al. 2012). Additional limitations include difficulty in monitoring the tumor response to treatment because inducing an inflammatory response may create MRI changes that mimic tumor growth. Immunotherapy is also complicated by the common use of steroids, which suppress the immune system.

Gene therapy
Gene therapy as it relates to malignant gliomas is based on tumor-specific introduction of genetic material for the purpose of treatment. It involves direct injection of a gene transfer vector or vector producing cells (VPC) into the tumor itself or into the cavity left after resection. Although preclinical studies have been quite promising, unfortunately therapeutic response to gene therapy clinical trials remains low (Tobias, Ahmed et al. 2013). Three classes of genetic therapy treatment have taken center stage over the last several decades: prodrug/suicide genes, oncolytic viruses, and gene immunotherapy. Although each is its own distinct entity, they all facilitate delivery of genetic material through the use of one or more vectors as described below.

Retroviral vectors
Retroviruses and retroviral vector producing cells (RVPCs) may be used to deliver specific genes to glioma cells; they are perhaps the most widely-studied class of vectors for treatment of GBM. This class of virus is advantageous in that its transduction is limited to rapidly dividing cells, meaning that normal brain cells remain unaltered. However, the transduction rate is low secondary to rapid inactivation of free retroviral vectors by complement as well as a lack of movement of virus to sites distant to the injection. It should be noted that transduction of circulating cells by vectors may occur, thus putting the patient at risk of cancer initiation via insertional mutagenesis (Barzon, Zanusso et al. 2006).

Adenoviral vectors
Adenoviruses belong to a family of 90-100 nm non-enveloped viruses made up of a nucleocapsid and double-stranded linear DNA. They account for roughly one tenth of all upper respiratory tract infections in children, infecting the host via introduction of their genome into the nucleus of the host organism's cells where the viral DNA remains free. This is in opposition to the retroviral mechanism involving incorporation of genetic material into the host cell's genomic structure.
Adenovirus enters the host cell by way of 2 distinct sets of interactions. Firstly, the knob domain of the virus's fiber protein binds to the cell receptor (either CD46 or coxsackievirus adenovirus receptor). This is followed by the interaction of a specialized motif in the penton base protein with an integrin molecule, which prompts internalization of the virus via an endosome. Thereafter the capsid components dissociate and the virion is released into the cytoplasm. Viral DNA enters the nucleus via the nuclear pore, later associating with histones. Following nuclear invasion, the viral genome is reproduced along with the host cell's DNA. However, the progeny of the original host cell will not carry the newly-introduced viral DNA. This necessitates numerous rounds of viral introduction in the treatment of cancer (Doloff and Waxman 2013).

Reoviral vectors
The genome of Reoviridae is segmented, double-stranded RNA, and the virus has the ability to make use of a non-functional protein kinase R (PKR) pathway in glioma cells to allow for viral replication. This is advantageous as the virus does not require genetic engineering. Other advantages include small size (70-80nm) and an absence of known consequent encephalitis in humans (Clarke, Debiasi et al. 2005).

Nonviral vectors
There are several nonviral vectors either currently in use or being considered for use in gene therapy such as synthetic vectors, nanoparticles, and stem cells/progenitor cells. From this group, perhaps the most studied is the liposome (included in the category of nanoparticles). Cationic Liposomes are easy to produce, have relatively low immunogenicity and toxicity, and typically exhibit long-term stability (Tobias, Ahmed et al. 2013).

Prodrug activating genes/suicide genes
Prodrug/suicide genes represent an ingenious wing of gene therapy. The basis of this antitumor modality is introduction of genes, either into the host genome or the intranuclear milieu, which imparts susceptibility to a subsequent therapeutic agent. The vectors themselves are genetically modified to produce an enzyme which converts a prodrug, given systemically, into toxic metabolites which act specifically on the malignancy.
Perhaps the earliest/most-studied example of prodrug/suicide gene utility when addressing gliomas is that of Herpes Simplex Type 1 Thymidine Kinase (HSV-tk). After incorporation of this gene into tumor cells (often residual cells status-post resection) and the endothelium of their vasculature, the host is treated with an antiviral such as gancyclovir (GCV). HSV-TK phosphorylates the prodrug of GCV into its active compound, whose mechanism of action involves DNA cross-linking, which leads to cell death. Following treatment with GCV, there may also be an observed "bystander effect" which involves the killing of non-transduced adjacent cells or even distant cells via immune response (T Cells, NK Cells) and toxic metabolites received via gap junctions (Ram, Culver et al. 1997

Adenovirally-mediated therapy
It should also be noted that unlike retroviral vectors, adenovirus can transduce both dividing and non-dividing cells. The majority of adenoviruses used for this purpose are E1-deleted adenoviral vectors, which may be injected at a higher titer than RVPCs; however high doses may indeed lead to serious side effects, including confusion, seizures, fever, leukocytosis, and hyponatremia that appear to be secondary to immune response to the vector (Trask, Trask et al. 2000). This same immune response lowers the yield of viral delivery but also aids in tumor reduction (Trask, Trask et al. 2000;Lang, Bruner et al. 2003). Notably, the adenoviral vector may be found transiently in blood but has not been found as a replication-competent entity.
Preliminary clinical data suggest that adenoviral mediated gene transfer of suicide genes (AdvHSV-tk) may have clinical utility ( Despite the aforementioned promising results from a small number of patients, the Phase III international open-label, randomized ASPECT clinical trial, which studied the intra-operative administration of adenoviral-HSV-tk followed by GCV (n=124) as compared to resection and standard of care alone (n=126), was not positive. Unfortunately, the data revealed no difference between the groups in terms of OS; furthermore, more patients in the experimental group had one or more treatment-related adverse events than those in the control group ( Pleuripotent neural stem cells procured from the subgranular zone of the hippocampus and the areas surrounding the lateral ventricles have the ability to migrate to areas of parenchymal damage (Luskin 1993). Neural stem cell clones may migrate to areas of tumor infiltration and thus were examined as vehicles for delivery of suicide genes, cytokines, or tumor necrosis factor-related apoptosis-inducing ligand (TRAIL); there is evidence of potential efficacy in animal models but no clinical utility data yet (

Tumor suppressor gene replacement
A well-documented characteristic of GBM is its inherent inactivation of the p53 tumor suppressor gene. Animal trials have shown that re-introduction of the wild-type p53 gene is pro-apoptotic leading to increased sensitivity to current modalities of treatment such as chemo-and radiotherapy. A Phase 1 trial of adenoviral gene transfer of intra-tumoral wildtype p53 in recurrent malignant glioma proved to be safe, but the transfected cells were not found in a radius large enough to be therapeutically effective (Lang, Bruner et al. 2003).

Oncolytic gene therapy
The realm of oncolytic virus therapy involves the use of replication-competent viruses with the ability to selectively replicate and kill cancer cells, with or without gene transfer. This is in opposition to prodrug/suicide gene therapy which makes use of replication-incompetent modalities. In order to combat the inefficiency of suicide gene therapy, oncolytic treatment employs tumor-specific, conditionally replicating viral vectors (Tobias, Ahmed et al. 2013).
The mechanism of action involves viral replication which eventually leads to lysis of the host tumor cell and subsequent release of additional copies of competent virus which may lead to further tumor reduction. This method is tumor-specific as it makes use of either attenuated viruses containing inactivated genes which replicate in tumor cells only, or viruses with replication-essential genes in tumor-specific promoters (Chiocca 2002). This method employs herpes simplex virus (HSV), Adenovirus, Reovirus, Poliovirus, Newcastle Disease Virus (NDV), and Measles virus.

Oncolytic HSV-1
Herpes Simplex is an enveloped, doubled-stranded DNA virus which exhibits inherent action upon the human nervous system; it can replicate in both active and quiescent cells. Consequently, safety was an original concern with this viral vector. Approximately 8 different HSV-1 genes have been altered or deleted to promote tumor specificity and lower collateral CNS damage (Tobias, Ahmed et al. 2013). There are two strains of replication-competent HSV-1 which have been significantly studied: G207 and HSV1716. G207 is the more widely-examined of the two and possesses a mechanism of action involving alteration of the gene which produces ribonucleotide reductase. In a recent phase 1B clinical trial, patients received injections of this virus both before and after tumor resection. Although viral replication was observed, treatment efficacy was sparse (Markert, Liechty et al. 2009

Oncolytic adenoviruses
Adenoviruses carrying mutations in E1A or E1B can also act on GBM via oncolysis. Their mechanism of action involves tumor-specific binding and inactivation of apoptotic proteins like pRB family and p53. Of note, adenovirus is inherently non-neurotropic, which may lend itself to superior safety versus HSV. One adenovirus, ONYX-015, has been found to preferentially replicate in p53 deficient cells secondary to its deletion for p53-inactivating protein E1B-55K. In one clinical trial it was injected into the surgical cavity after resection and found to have no serious adverse effects; however; almost all patients involved in the trial had progression of their GBM (Chiocca, Abbed et al. 2004). It should also be noted that Geoerger et al found human xenografts to be responsive to ONYX-015 without correlation to their p53 status (Geoerger, Grill et al. 2003).

Oncolytic NDV
NDV is an avian paramyxovirus, which does not harm humans except for rare pulmonary infection in poultry farmers; certain strains harm neoplastic cells via a currently unknown mechanism (Reichard, Lorence et al. 1992). Interestingly, NDV also has pleiotropic immunemodulatory properties (Schirrmacher, Haas et al. 1999). It should be noted that treatment with NDV necessitates starting at a low dose as there have been examples of treatment-related death with NDV PV701 and solid cancers (Pecora, Rizvi et al. 2002). MTH-68/H, a live attenuated oncolytic viral strain of NDV, has shown promising results is a small number of GBM patients (Csatary, Gosztonyi et al. 2004).

Oncolytic reoviruses
Reovirus is a double-stranded RNA-containing virus that replicates in GBM cells because of a hyperactive ras signaling; it distinctively does not replicate in normal brain cells. A phase I clinical trial of intrattumoral administration of genetically unmodified virus was well tolerated by patients with recurrent malignant gliomas (Forsyth, Roldan et al. 2008). Further studies involving reovirus are currently underway.

Gene immunotherapy
Treatment of gliomas with immune therapy is based on harnessing of the patient's T-Cell mediated response to tumor cells. Typically, gene-immune therapy falls into the category of priming in the brain by the transfer of cytokine genes, like IL-2, IL-4, IL-12, and interferons gamma and beta (

Conclusion
The aforementioned negative results of several key phase III clinical trials in GBM demonstrate that current proof of efficacy in preclinical models is a necessary but not sufficient condition for clinical utility. The consistency in obtaining negative results in GBM is remarkable. How can we improve and what do we do to turn the tide in our favor?
It is becoming evident that the phenotypes of GBM are not created by few solitary molecules but rather by dynamic networks with positive and negative loops that react and respond to a therapeutic intervention. The good news is that these networks are finite dimensional. Furthermore, because of the instability of cancer genomes, random mutations are introduced in the population of rapidly dividing glioma cells; hence, a particular therapy could merely delay growth by selecting a resistant subpopulation. We suggest that we should elevate the threshold by mandating stringent criteria before proceeding to very costly phase III clinical trials, as follows.

1.
Phase II clinical trials must include a control arm with appropriate stratification instead of historical controls.

2.
Preclinical models must include proof of feasibility in at least 8-10 different cell lines/ animal models.