Immunotherapy for cancer in the central nervous system: Current and future directions

ABSTRACT Glioblastoma multiforme (GBM) is the most common primary brain tumor in adults and still remains incurable. Although immunotherapeutic vaccination against GBM has demonstrated immune-stimulating activity with some promising survival benefits, tumor relapse is common, highlighting the need for additional and/or combinatorial approaches. Recently, antibodies targeting immune checkpoints were demonstrated to generate impressive clinical responses against advanced melanoma and other malignancies, in addition to showing potential for enhancing vaccination and radiotherapy (RT). Here, we summarize the current knowledge of central nervous system (CNS) immunosuppression, evaluate past and current immunotherapeutic trials and discuss promising future immunotherapeutic directions to treat CNS-localized malignancies.


Introduction
Glioma is the most common primary malignant brain tumor, accounting for nearly 80% of cases in adults. Glial-derived tumors are classified based on histologic subtype, which include glial fibrillary acidic protein positive (GFAPC) astrocytic tumors, oligodendrogliomas, ependymomas and a mixture of the subtypes. 1 Of these, astrocytic glioma grade IV, otherwise referred to as GBM, is the most common and deadly subtype with a median survival of 14.6 mo post-diagnosis and an average 5-year survival rate of less than 5%. 2,3 Current treatments that combine resection, RT and chemotherapy are unable to prevent tumor recurrence based on residual disease originating from the invading margins/ inoperable surgical bed. Despite previous translational efforts that include new approaches for gene therapy, targeted chemotherapeutics and/or radiotherapeutic modalities, the standard of care for newly diagnosed GBM has remained unchanged for the past 10 y, highlighting the need for better treatment options. Also, there is no standard of care treatment for patients with recurrent GBM. The prevalence of metastatic tumors in the CNS greatly exceeds the number of GBM cases, yet, overall survival (OS) is similarly dismal. In this review, we discuss historical efforts, as well as new and/or expanded approaches that include vaccination, immune checkpoint blockade, adoptive T cell transfer, as well as combinatorial immunotherapy for the rationale design to durably control aggressive tumors in the CNS.

CNS tumors and Immunosuppression
The CNS was originally considered to be an immune-privileged site, in part, based on the superior growth of rat osteosarcoma cells that were intracranially injected into the brain compared to growth subcutaneously or intramuscularly. 4 More recent observations indicate that the CNS is immunospecialized, based on the considerable interaction with the peripheral nervous system and the non-parenchymal ventricles, meninges and subarachnoid space. 5 Inflammatory stimuli, including those induced by brain tumors, increase CNS immunogenicity due to microglial activation and blood-brain barrier (BBB) disruption. 6 The latter occurs secondary to glioma cell invasion of the basement membrane. 7,8 BBB disruption facilitates the drainage and presentation of CNS antigens to the cervical lymph nodes, which primes T cells for homing and infiltration to the tumor parenchyma. Interestingly, the pattern of leukocyte infiltration into GBM is not identical among tumors, with specific genetic subtypes including the mesenchymal profile, possessing higher levels of T cell infiltration. 9 Coincidently, the mesenchymal subtype is almost universally observed in recurrent GBM after standard of care therapy. 10 Commensurate to the inflammatory signals (i.e. cytokines, chemokines, growth factors) that brain tumors induce, are potently immunosuppressive mechanisms that include the tryptophan catabolic enzyme, indoleamine 2,3 dioxygensase 1 (IDO1). This rate-limiting enzyme is expressed in 96% of resected glioblastoma, with the upregulation correlating with a worse patient prognosis. 11,12 IDO1 converts tryptophan into kynurenines, with the latter catabolite mediating inhibition/induction of apoptosis in effector T cells and/or amplification of immunosuppression by CD4 C CD25 C FoxP3 C regulatory T cells (Treg) (Fig. 1). 13 Preclinically, tumor-derived IDO1 is essential for Treg accumulation and immunosuppression, since malignant brain tumors deficient for the enzyme result in spontaneous rejection mediated by a T-cell-dependent mechanism. 12 Paradoxically, Treg incidence in newly diagnosed patient GBM is a neutral prognostic factor. 14 Importantly, it has yet to be determined whether this finding holds true in recurrent GBM and this may be an important clinical consideration since our laboratory has experimental evidence from a . IDO1 expression is regulated by the Jak/STAT and NF-kB pathways, which is induced by IFNg-and TGF-b-receptor activation, respectively. IDO1 is a cytoplasmic enzyme that metabolizes tryptophan (Trp) to kynurenine (Kyn). Within the GBM cell, Kyn complexes with the aryl hydrocarbon receptor (Ahr), cytoplasmically, facilitating the nuclear translocation and further docking with aryl hydrocarbon receptor nuclear translocator (ARNT) to transcriptionally regulate IL-6, acting as an autocrine loop that amplifies and sustains IDO1 expression. Simultaneously, extracellular Kyn suppresses T effector responses while activating regulatory T cell (Treg; CD4 C CD25 C FoxP3 C ) function through a presumably overlapping mechanism. IDO1 directly activates NF-kB signaling which maintains and/or upregulates TGF-b expression. Increased TGF-b levels upregulate CTLA-4 and GITR expression by Treg. CTLA-4 interacts with B7.1 (CD80) and B7.2 (CD86) on DC, resulting in the induction of IDO1 (in DC) and commensurate downregulation of antigen presentation to T cells. Both GBM and MDSC express TGF-b, which synergizes with PD-L1 to suppress the T cell effector response via interaction with PD-1. Moreover, interleukin-10 (IL-10)-and prostaglandin E2 (PGE2)-expressing MDSC act on their cognate receptors expressed by GBM to ramify Jak/STAT and NK-kB-mediated signaling. DNA released by dead/dying GBM cells is phagocytized by resident DC to activate the STING pathway leading to Type 1 interferon (a/b) expression, supporting increased effectiveness of anti-GBM immunity. PD-1 is highly expressed by tumor-infiltrating cytotoxic T cells and PD-L1 is upregulated on cancer/stromal cells in response to T-cell-secreted IFNg. Blocking the interaction of PD-1-expressing T cells with PD-L1 leads to increased effector function and enhanced GBM immunity. Targets for immunomodulation are shown in red. Note: Although IDO1 expression and signaling are shown in GBM cells, shared signaling patterns are presumed to be present in DC and MDSC as well. T CON : conventional CD4 C FoxP3 ¡ T cell; T REG : regulatory CD4 C FoxP3 C T cell; T C : cytotoxic CD8 C T cell; INCBO24360/NLG919: inhibitors of IDO1; PS1145: inhibitor of the NF-kB pathway; TRX518: humanized monoclonal agonistic antibody for GITR; Ipilimumab: humanized monoclonal antibody for CTLA-4; LY2109761: TGF-b receptor kinase inhibitor; MK-3475/MDX-1106: humanized monoclonal antibodies to PD-1; MEDI4736/MPDL3280A: humanized monoclonal antibodies to PD-L1; Anti-Gr1: mSC-depleting antibody; Daclizumab: humanized anti-CD25 (IL-2Ra); STING: stimulator of interferon genes; TBK1: TANK-binding kinase 1; IRF3/7: interferon regulatory factor 3/7; STAT3: signal transducer and activator of transcription 3; A. 18,72,[84][85][86][87][88][89][90][91][92][93] model of spontaneously forming glioma suggesting that IDO1 functions differently in brain tumors depending on the newly diagnosed vs. recurrent context (unpublished observation). An alternative immunosuppressive pathway that contributes to T cell dysfunction is mediated by interactions between PD-1 and PD-L1, resulting in the loss of T cell effector function. Notably, both human GBM 15 and tumor-infiltrating macrophages 16 express high levels of PD-L1, suggesting the need for multi-cellular targeting for optimal immunotherapeutic benefit. Similar to other malignancies, cytotoxic T cells infiltrating GBM express high levels of PD-1. 17 A third dominant and critical immunosuppressive pathway relevant to brain tumors is mediated by CTLA-4, which simultaneously inhibits effector T cell activation/proliferation and Treg activation/function in GBM. 18 Interestingly, the interaction of CTLA-4 with dendritic cell (DC)-expressed B7, induces IDO1 expression. 19 Thus, it will be interesting to determine whether co-inhibiting CTLA-4 and IDO1 lacks an additive/synergistic impact against brain tumors or if other undiscovered immunosuppressive mechanisms remain independent of the interaction.

Therapeutic approaches
Vaccination Therapeutic vaccination against cancer induces and/or rescues unproductive immune responses against tumor antigens intrinsically expressed or cross-presented by stromal cells. 20 This immunity can be generated against mutated peptides, 21 or post-translational modifications. 22 To generate/rescue functional antitumor T cell responses, vaccines co-administer tumor peptide(s) and immuno-stimulatory adjuvant(s) to license DC for activating and expanding tumor-reactive T cells. Determining the optimal peptide(s) for targeting is a challenging task since many tumor-associated antigens are identified as "self" by the immune system. 23 Given the shared neuroectodermal lineage of astrocytes and melanocytes, there is relatively significant overlap of shared tumor associated antigens between GBM and melanoma. 24 This complicates targeting GBM with high specificity given the obvious potential for immunization against normal melanocytes. 24 In practice, however, this phenomenon has not been observed in the majority of previously vaccinated GBM patients. 25 Notably, ex vivo loading of a newly diagnosed GBM patient's DC with six GBM tumor-associated peptides can generate vaccine-specific immune responses that are not associated with an OS advantage. 26 By vaccinating GBM patients with DC loaded with glioma-associated peptides combined with adjuvant poly-ICLC, approximately 60% of patients demonstrate glioma-associated immune responses, with <10 % of recurrent glioma patients demonstrating stable tumor regression. 27 Overall, these studies highlight an important concept suggesting that, stimulating an immune response against exclusively tumor-associated peptides is not sufficient for controlling malignant progression in the majority of patients.
Tumor neoantigens are considered to have higher potential for therapeutic vaccination. These neoantigens are generated during tumor evolution, 28 often resulting in unique targets within individual patients. 23,28 Some neoantigens, however, are present in a higher percentage of GBM, providing rational targets for focusing vaccination efforts against. One of the best characterized neoantigens is the epidermal growth factor receptor variant III (EGFRvIII), which is present in »20-30% of newly diagnosed GBM, 29 carrying an independent negative prognosis for patients who survive >1 y after diagnosis. 30 EGFRvIII is the result of an in-frame deletion leading to a new antigenic junction, 31 capable of inducing both cellular and humoral immunity. 32 Rindopepimut, a 13-amino acid EGFR-vIII peptide vaccine conjugated to adjuvant, is currently utilized for targeting this neoantigen. Phase II EGFRvIII peptide vaccines have demonstrated vaccine immunogenicity and increased OS, with median at approximately 24 mo from diagnosis, compared to historical controls (Table 1). [32][33][34] Survival advantage of treated patients correlate to the magnitude of induced tumor immunity, with tumor relapse occurring with loss of EGFRvIII expression based on immunohistochemical detection. [32][33][34] While promising, these data could also indicate that, sensitivity to EGFRvIII detection by IHC is masked by patient-derived EGFRvIII antibodies or post-translational modification(s) as well as the independent loss due to radiation and/or chemotherapy. 35 A two-arm randomized phase III trial (ACT IV) for recently diagnosed GBM is currently underway to better assess the efficacy of this approach (NCT01480479) ( Table 2). With regard to targeting neoantigens in lower-grade glioma, mutant isocitrate dehydrogenase type 1 (IDH1) is carried by more than 70% of diffuse grade II and III gliomas, 36 and targeting IDH1 by peptide vaccination has shown efficacy. 37 To address tumor relapse from generation of antigenic variants in the process of targeting a single peptide, alternative vaccine approaches have been created to target a broad range of antigens, simultaneously. One exciting approach utilizes heat shock protein (HSP) peptide complexes (HSPPC-96) derived from a GBM patient's resected tumor. Intracellular HSP physiologically binds peptides with extracellular HSP capable of mediating the internalization of HSPPC-96 into APCs for efficient MHC-I and MHC-II presentation of tumor peptides. 38,39 Clinically, HSPPC-96 vaccine generates a tumor-reactive T cell response. 39 In a phase II, single arm trial for surgically resectable recurrent GBM, HSPPC-96 vaccine increased the median OS to an impressive 42.6 weeks, which provides a substantial survival benefit when compared to historical controls. 25 Interestingly, a predictor of poor response to vaccination was lymphopenia at the time of vaccination, a side effect likely attributable to previous chemotherapy, radiation and/or decadron administration. 25 An alternative approach to targeting multiple epitopes, simultaneously, is utilizing pulsed autologous DC with tumor lysate. This approach, identified as DCVax®-L, is currently in a Phase III trial for patients with newly-diagnosed GBM (NCT00045968).
Over the past 3 y, technological advances and clinical discoveries have sparked the development of next-generation vaccines. The first observation from both preclinical subcutaneous fibrosarcoma and clinical melanoma studies demonstrated that CD8 C T cells responsible for eradicating tumors must recognize tumor-specific peptides that have high affinity for MHC-I. [40][41][42] In preclinical subcutaneous fibrosarcomas, peptide affinity determines whether a peptide can be cross-presented by tumor-associated macrophages and thereby serve to optimally stimulate T cells to produce high levels of cytokine in the tumor microenvironment. 40 Recent technological advances  now facilitate these "rejection" antigens to be reliably identified using (i) genome-wide exomic sequencing to find mutations and (ii) peptide affinity algorithms to identify peptides with high peptide-MHC affinity. 41,43 This approach has been validated preclinically demonstrating that, vaccinating against a model "rejection peptide" achieves tumor eradication of aggressive melanoma. 21,44 Creating personalized vaccines to target these predicted rejection antigens is now recognized as a promising approach against non-CNS tumors and should be studied with regard to whether similar efficacy is achievable against aggressive tumors in the CNS.

Checkpoint blockade
Over the past 15-20 y, it has become recognized that inhibitory receptors on T cells play an important role in suppressing T-cell-mediated antitumor responses. 45 These inhibitory receptors are referred to as immune checkpoints due to their role in preventing inappropriate/prolonged  67 Sample size/type of glioma indicate GBM unless otherwise noted. Results/outcomes indicate median unless otherwise noted. Trials were identified on pubmed with the search terms: "glioblastoma" AND "patients" AND "trial," between the years, 2010 and 2015.
e1082027-6 D. C. BINDER ET AL. activation. To date, the checkpoints that have been targeted with the most impressive clinical antitumor responses are CTLA-4 and PD-1. During CD8 C T cell activation, CTLA-4 is upregulated and inhibits further T cell activation and proliferation. 46 CTLA-4 is also expressed on CD4 C T cells where it functions to enhance Treg-mediated immunosuppression. 47 Ipilimumab, a humanized CTLA-4 antibody, was the first FDA-approved immune checkpoint inhibitor.
Much clinical experience with ipilimumab has been in treating metastatic melanoma, in which there is an approximately 2% complete response rate that remains durable. 48 Responses have been observed against both non-CNS and CNS-infiltrated melanoma metastases. 49 Preclinically, mice bearing intracranial glioma and treated with CTLA-4 mAb (clone 9H10) develop robust antitumor immunity without affecting Treg function. 18 Clinically, the administration of ipilimumab for GBM has been limited to a small number of GBM patients in the recurrent setting. More recently, efforts aimed at inhibiting the PD-1/PD-L1 pathway have generated significant interest. Tumorinfiltrating lymphocytes express high levels of PD-1 in most cancers, including GBM, 17 as a result of chronic antigen stimulation by the tumor. 50 When PD-1-expressing T cells interact with PD-L1, T cell effector function is inhibited. 50 PD-L1 is upregulated in GBM through the following mechanisms: (i) oncogenic signaling as a result of PTEN loss, 15 (ii) paracrine signaling, 16 and/or (iii) "adaptive immune resistance" whereby T-cell-secreted IFNg induces PD-L1 expression on neighboring cells. 51 While clinical trials studying PD-1 and PD-L1 blockade are currently recruiting patients for GBM (NCT02337491, NCT02336165), the effectiveness of this approach has been characterized in treating refractory melanoma, providing an objective response rate (ORR) of approximately 15-30% as monotherapy with complete responses restricted to <6 % of patients. 52,53 Since PD-1/PD-L1 does not induce T cell infiltration into tumors, but rather rescues/prevents T cell anergy, it is not surprising that patients associated with the best responses possess higher tumor-infiltrating T cell levels prior to treatment that is co-localized with PD-L1 expression. 54 The most promising outcomes related to immune checkpoint inhibition have been achieved through combinatorial CTLA-4/PD-(L)1 blockade, [55][56][57] which is consistent with these pathways providing non-redundant T cell inhibition. In a recent randomized control trial for untreated advanced melanoma, dual CTLA-4 and PD-1 blockade provided an improved ORR (58%) compared to monotherapy CTLA-4 (19%) and monotherapy PD-1 (44%). 57 Interestingly, dual CTLA-4 and PD-1 blockade was found to be superior compared to PD-1 monotherapy in treating PD-L1-negative tumors, but not PD-L1-positive tumors, suggesting that CTLA-4 blockade induces T cell infiltration into tumors. 57 Consistent with these findings in melanoma, preclinical models of GBM demonstrate high rates of survival when treated with simultaneous PD-L1 and CTLA-4 blockade, as compared to the respective monotherapies. 58 Clinically, trials aimed at GBM patient treatment with ipilimumab and nivolumab (humanized PD-1 mAb) are already underway (NCT02311920, NCT02017717). In addition, several clinical trials enrolling patients with brain metastases are also in progress, including studies using PD-1 mAb alone and CTLA-4 combined with PD-1 mAb (NCT02374242, NCT02320058).
In addition to PD-1 and CTLA-4, therapeutic modulation of other immune inhibitory and stimulatory pathways is currently being evaluated preclinically and in early-phase trials (Table S1). Blocking inhibitory receptors LAG-3 or TIM-3 in combination with PD-1 blockade provides impressive preclinical tumor control in non-CNS tumor models 59,60 . Dual LAG-3 and PD-1 blockade is currently being tested against multiple non-CNS solid tumors in a Phase I trial (NCT01968109). Modulating both inhibitory and stimulatory immune pathways may also be a promising approach as dual CTLA-4 blockade and ICOS stimulation provides improved antitumor control against preclinical murine melanoma and prostate cancer. 61 This strategy may also be effective in GBM, as triple therapy with RT combined with CTLA-4 inhibition and 41BB stimulation provides improved tumor control compared to each dual therapy. 62 Adoptive T cell therapy Previously described therapeutic approaches endeavor to rescue or induce endogenous T cell responses, while adoptive T cell therapy provides an alternative strategy that involves expanding tumor-specific autologous T cells, ex vivo, followed by venous infusion into the same individual. Tumor-reactive T cells are isolated from (i) peripheral blood, (ii) surgically resected tissue or (iii) generated by transduction of the patient's autologous T cells with vectors encoding T cell receptors (TCR) or chimeric antibody receptors (CAR). 63 The capacity of adoptive T cell therapy to eradicate a large established tumor burden has been demonstrated with the re-infusion of tumor-infiltrating lymphocytes specific to melanoma, 64 as well as CAR-based treatment for CD19 C B-cell malignancies. 65 In GBM patients, adoptive T cell therapy has been used to target human cytomegalovirus (CMV) antigens expressed by tumor cells. [66][67][68] A recent study treating 11 recurrent GBM patients with infusions of autologous adoptively transferred CMV-specific T cells led to a median OS of >57 weeks, with four patients remaining progression-free throughout the study period. 67 Longer progression-free survival (PFS) was associated with decreased expression of checkpoint receptors on T cells suggesting that, maintaining effector function of adoptively transferred T cells is required for a durable clinical response. 67 A clinical trial investigating CMV adoptive T cell therapy is ongoing (NCT00693095).
Utilizing CAR T cell adoptive therapy for GBM patients is a logical 'next step' for autologous therapy. CAR consist of an extracellular antibody domain fused to a T cell cytoplasmic signaling domain. Preclinical glioma CAR studies targeting HER2 and the previously described EGFRvIII reported impressive results. 69,70 Clinical trials targeting both antigens are ongoing (NCT02209376, NCT01109095, NCT01454596), as well as a CAR trial targeting IL13Ra2 (NCT02208362). Future studies should focus on identifying additional tumor-specific antigenic targets shared among patients and/or developing an approach to personalize CAR technology to each patient's tumor antigen profile.

Combination approaches
Optimal immunotherapy approaches must provide immune activation while, simultaneously, countering inhibitory checkpoint blockade signals. Moreover, it is now recognized that single modality immunotherapy has limitations that can be overcome by multi-targeted strategies. Some of the promising immunotherapeutic combinations will be further discussed.

Radiation, DNA sensors and immune checkpoint blockade
Combining ablative radiation with immune checkpoint blockade is a promising immunotherapeutic combination. While radiation was previously viewed as immunosuppressive, preclinical tumor models have demonstrated that hypofractionated ablative radiation can generate tumor regression that is T cell dependent. 71 The mechanism accounting for this effect likely relies on: (i) radiation-induced tumor inflammation and cell death, (ii) DC that phagocytize "released" cancer cell DNA capable of activating the Stimulator of IFN genes (STING) pathway, (iii) increased type 1 IFN-licensed DC that prime tumor-specific T cells and (iv) reactive T cells that home to and engage the tumor with strong effector function. 72 Type I IFN appears to be essential for antitumor immunity, with intratumoral injection of a STING agonist significantly improving tumor control following radiation in experimental models. 72 While the impact of combined radiation and STING activation has yet to be confirmed in CNS tumor models, it is notable that immune-mediated control of glioma outgrowth is dependent on STING-mediated induction of type 1 IFN. 73,74 Accordingly, glioma patient prognosis is dictated, in part, by type 1 IFN single nucleotide polymorphisms (SNPs). 75 Collectively, these findings suggest that immune-modulating approaches utilizing a combination of RT and STING agonists may be promising to combat tumors in the CNS.
For both CNS-and non-CNS-resident tumors, combined RT and immune checkpoint blockade has demonstrated increased effectiveness when compared to RT alone. In a mouse orthotopic glioma model, combining radiation with anti-PD-1 provides an additive effect that improves OS, when compared to either therapy administered individually. 76 As a mechanism accounting for the enhanced effectiveness of combinatorial treatment, radiation-induced inflammation results in PD-L1 upregulation on cancer cells, macrophages and DC. 77 Similarly, combinatorial anti-CTLA-4 and RT leads to tumor control in a preclinical model of breast cancer. 78 Notably, the latter combination has thus far yielded a less impressive impact on OS when compared to combinatorial RT and PD-(L)1 blockade. 77 More recently, it was reported that control of preclinical melanoma is optimal when simultaneously treating with RT, anti-PD-(L)1 and anti-CTLA-4, when compared to dual therapy. 79 Each modality induced a unique immune activating profile with RT expanding the TCR repertoire, anti-CTLA-4 inhibiting Treg function and increasing the Tc/Treg ratio and anti-PD-(L)1 preventing T cell exhaustion/dysfunction in tumors. 79 Interestingly, RT combined with anti-CTLA-4 and anti-4-1BB induces similar antitumor activity, with the latter agonist causing direct stimulation to cytolytic T cells, resulting in an increased level of survival and T cell infiltration when compared to dual therapy. 62 Clinically, combining RT and checkpoint blockade was recently tested for the first time in a phase I trial. Patients received three doses of hypofractionated radiation to a single metastatic melanoma lesion followed by anti-CTLA-4 treatment. While median OS was <11 mo, local tumor control was achieved in the irradiated lesions for all 12 patients analyzed. 79 Although CNS metastases were not targeted in this trial, local tumor control of melanoma brain metastases has been reported in a case series using both whole-brain RT (30 Gy/10 fractions) and stereotactic RT (20-24 Gy/1 fraction) for patients who received RT following a course of ipilimumab. 80 Based on the strong promise of radiation combined with checkpoint blockade to achieve local tumor control in CNS and non-CNS tumors, future preclinical and clinical GBM studies should investigate how to optimize this approach. For melanoma brain metastases, two phase II trials combining RT approaches with ipilimumab for brain metastases are currently underway (NCT02115139, NCT02097732).

Vaccination and immune checkpoint blockade
Therapeutic vaccination may fail if the strategy does not optimally expand tumor-reactive T cells and/or vaccine-generated T cells lose effector function in the immunosuppressive tumor microenvironment. 81 PD-1/PD-L1 interactions likely dampen vaccine responses by two mechanisms: (i) in the draining lymph node where vaccine adjuvant-induced inflammation results in PD-L1 expression on antigen-presenting cells that inhibits maximal expansion of vaccine-generated T cells, 82 and (ii) in the tumor itself whereby "adaptive immune resistance". 51 is generated by T cells secreting IFNg that induces PD-L1 upregulation on neighboring cells leading to T cell anergy. Thus, combining vaccination with PD-1/PD-L1 antibody blockade is likely to provide a synergistic effect. In support of this, long-established preclinical melanomas resistant to dual PD-L1 and CTLA-4 blockade are eradicated by vaccination in approximately 33% of mice, but eradicated by vaccination combined with anti-PD-L1 in 80% of mice. 21 In an independent Table 3. High priority questions for increasing immunotherapeutic efficacy against tumors in the CNS.

Preclinical
Do inhibitors that co-target IDO1 and IDO2 provide superior efficacy when compared to monotherapy? Will inhibitors of tryptophan catabolism complement other immunotherapies? Which capacity of IDO1 is more important for immunotherapeutic efficacy: signal transduction modifier vs. tryptophan catabolism? What is the best approach for further identification of ubiquitous GBM-specific neoantigens for translation into vaccine and/or adoptive T cell therapeutic approaches? Is there an optimal vaccination approach to generate functional T cell responses and is this GBM subtype-specific (i.e. responsiveness in classical vs. mesenchymal, newly diagnosed vs. recurrent)? Do different GBM subtypes possess correlative mutational frequencies that associate with responsiveness to immunotherapy? Will survival outcomes be enhanced with combinatorial approaches (vaccine § RT § checkpoint blockade § STING activation)?

Clinical
Will GBM respond to immune checkpoint blockade? What is the best approach for identifying patient cohorts that will benefit from immunotherapy? What is the best approach for monitoring treatment effectiveness in GBM patients to immunotherapy (i.e. peripheral blood markers, tryptophan metabolic profiling or IHC markers in the tumor)? What is the best approach to limit brain swelling following immunotherapy? Is bevacizumab an alternative to decadron that can be easily added without defusing effectiveness? preclinical study of subcutaneous tumors, vaccination combined with PD-(L)1 and CTLA-4 inhibition led to improved tumor rejection and mouse survival, when compared to dualand mono-therapeutic treatment. 83 Clinically, the combination of peptide vaccination and PD-1 blockade is currently being evaluated in patients diagnosed with melanoma (NCT01176474). Since the majority of prior studies have been performed in non-CNS tumor models, future preclinical and clinical studies should evaluate these treatment approaches in patients with GBM.

Conclusions
GBM is a highly immunosuppressive tumor that is refractory to traditional therapies and difficult to treat based on its anatomical location. Metastatic tumors in the brain, with a prevalence of >20 :1 compared to GBM, also present much treatment challenge. Past immunotherapeutic efforts for brain tumors have predominantly focused on therapeutic vaccination that has achieved promising immune activity and clinical responses. However, durable responses remain rare highlighting the need to further test existing promising approaches including gene therapy (supplemental text, Table S2), develop next-generation therapeutics (i.e. IDO inhibitors/STING agonists,) and test novel immunotherapeutic combinations (Table 3). Because antitumor immune responses occur in the context of inflammation, the possibility for tumor-and therapy-induced inflammation to cause additive/synergistic brain swelling and neurologic compromise must be recognized. While Decadron is routinely used to counter brain swelling, its use is restricted to low doses in immunotherapeutic trials as it is also extremely immunosuppressive. Next-generation CNS immunotherapies, if more efficacious, may carry an even higher risk for brain swelling and neurological compromise, thus identifying non-immunosuppressive anti-inflammatory approaches is important. Utilizing bevacizumab, a VEGF neutralizing antibody that secondarily decreases inflammation, is one such approach currently being explored in combination with GBM immunotherapy (NCT02336165, NCT01814813). CNS immunotherapy has a bright future in this current "golden age" of immunotherapy . Future studies should focus on providing patients with this battery of ever-evolving options, while also recognizing that CNS malignancies have unique immunosuppressive phenotypes that need to be specifically targeted.

Disclosure of potential conflicts of interest
No potential conflicts of interest were disclosed.