The first OV therapy approved by the FDA in the United States was talimogene laherparepvec (tvec), an oncolytic herpesvirus, for advanced melanoma in 2015 (Conry et al., 2018), paving the way for the use of a host of mutated herpesviruses to treat GBM. Many of these mutated constructs are designed to allow preferential infection and lysis of tumor cells, allowing for viral propagation and stimulation of an immune response via release of tumor antigens (Yin et al., 2017). Two main HSV mutants tested in clinical trials for GBM are G207, capable of replicating in only dividing cells, and HSV1716, which can replicate in both dividing and nondividing cells (Immidisetti et al., 2021). There have been three phase I trials that demonstrated the safety and tolerability of G207 in rGBM and ndGBM patients (Markert et al., 2000; Markert et al., 2009; Markert et al., 2014), and similarly, three phase I trials have shown HSV1716 to be tolerable in ndGBM and rGBM patients (Rampling et al., 2000; Papanastassiou et al., 2002; Harrow et al., 2004). Additionally, there was a phase I trial testing intratumoral and peritumoral injection of HSV1716 after resection in pediatric patients with recurrent gliomas that was terminated due to low recruitment (Pediatric Brain Tumor Consortium, 2016). HSV1716 is currently undergoing testing in a phase II trial for children with recurrent high grade glioma (MD GKF, 2021). An alternate version of G207 named G47Δ is currently undergoing phase I and II studies in rGBM patients in Japan (JPRN-UMIN000002661). There is also a phase I trial currently underway testing M032, a genetically engineered HSV expressing IL12 transgene, in rGBM patients (MD JM, 2021).

Genetically manipulated adenovirus has also become a popular form of viral therapy in treating GBM. ONYX-015 is an adenovirus construct attenuated via deletion at the E1b locus that was one of the first such constructs to be tested in humans (Kirn, 2001). It was investigated in a 2004 phase I study of 24 rGBM patients who received intratumoral injections with a median survival of 6.2 months (Chiocca et al., 2004). However, attention ultimately shifted to other forms of adenovirus constructs for multiple reasons, including a lack of response to ONYX-015 as a single agent for solid tumors in multiple trials (Kirn, 2001).

There is now much work surrounding DNX-2401, a replication-competent adenovirus that is unable to bind healthy cells with intact retinoblastoma pathways and thus selectively binds tumor cells (Philbrick and Adamson, 2019) (p2401). A phase I trial for DNX-2401 tested dose escalation protocols via either single intratumoral injection or permanently implanted catheter followed by tumor resection for both groups and resulted in 5 (20%) patients with over 3 years of survival in the first group and 2 (17%) patients with over 2 years of survival in the second group (Lang et al., 2018). DNX-2401 has been demonstrated to promote a shift towards the M1 macrophage phenotype in the CSF of treated GBM patients compared with controls, suggesting a treatment-mediated protumoral shift in the immune landscape (Van den Bossche et al., 2018). A phase Ib trial has been completed testing the use of DNX-2401 alone versus in combination with interferon gamma (IFN-γ); while IFN-γ did not provide additional survival benefit, DNX-2401 alone appeared to provide an OS-12 of 33% across 27 patients (Lang et al., 2017). The recently completed phase II CAPTIVE/KEYNOTE 192 trial testing DNX-2401 in combination with pembrolizumab in rGBM patients has reported positive results with a median OS of 12.5 months (DNAtrix, 2021).

PVS-RIPO is a replication-competent polio-rhinovirus chimera that selectively infects cells expressing CD155 and demonstrates attenuated neurovirulence via substitution of the native internal ribosome entry site (IRES) for that of rhinovirus (Gromeier and Nair, 2018). Preclinical studies have demonstrated that PVS-RIPO infection may induce dendritic cell and neutrophilic recruitment to the tumor site in vivo (Gromeier and Nair, 2018; Mosaheb et al., 2020). A phase I trial testing intratumoral injection of PVS-RIPO in 61 rGBM patients demonstrated a higher OS rate of 21% at 36 months compared to historical controls, with 2 patients surviving for more than 70 months (Desjardins et al., 2018). The favorable results from phase I have led to two ongoing phase II trials testing PVS-RIPO alone in 122 rGBM patients (UHSCC, 2021) and testing PVS-RIPO in combination with pembrolizumab in 30 rGBM patients in the LUMINOS-101 trial. (Istari Oncology, 2021).

NDV, like much of the viruses discussed here, is capable of selectively infecting and inducing lysis in tumor cells from a variety of cancers (Matveeva et al., 2015). There have been three strains developed and tested in clinical trials over the years: MTH-68, NDV-HUJ, and Ulster. The first reported use of the MTH-68 strain to treat CNS tumors occurred in 1999 in pediatric patients with rGBM. This was followed up with a 2004 study that tested MTH-68 in four patients with high grade glioma that resulted in survival rates ranging from five to 9 years (Csatary and Bakács, 1999; Csatary et al., 2004). NDV-HUJ was tested in a 2006 phase I/II trial of 14 rGBM patients that had been refractory to treatment, with minimal Grade I and II toxicities only and three long-term survivors who all eventually progressed either clinically or radiologically (Freeman et al., 2006). In 2001, Schneider et al. tested autologous tumor cells infected with Ulster NDV and subsequently irradiated in 11 patients after surgical resection and found comparable survival compared to patients who received chemotherapy instead of virotherapy (Schneider et al., 2001). A similar approach was taken in 23 ndGBM patients who underwent maximal resection and received ATV-NDV, an anti-tumor vaccine infected with the Ulster NDV strain; the results demonstrated significantly increased rates of OS-1 (91 versus 45%) and OS-2 (39 versus 11%) compared to the control group (Steiner et al., 2004).

Reovirus is a double-stranded RNA virus that preferentially infects and lyses tumor cells in part via overactivated Ras signaling pathways that enhance proteolytic viral disassembly in malignant cells (Coffey et al., 1998; Norman et al., 2004; Alain et al., 2007). The application of reovirus for treating tumors was first accomplished via phase I studies on prostate cancer and cutaneous metastases from systemic cancer before eventually being tested on malignant gliomas by Forsyth et al. (2008) Nine of 12 patients were treated for GBM in a dose escalation protocol with no observable adverse events and survival ranging from 6 to 63 weeks (Forsyth et al., 2008). Similar findings were replicated by Kicielinski et al. when applying Reolysin, a wild-type reovirus, to malignant gliomas in a phase I study (Kicielinski et al., 2014). Results were promising, with survival ranging from 14 to 141 weeks, which prompted the designation of orphan drug status to Reolysin for the treatment of malignant glioma by the FDA in 2015 (Jaime-Ramirez et al., 2017). In vitro studies have demonstrated that reovirus administration induces DC maturation, stimulates proinflammatory cytokine production, including IFN-alpha, TNF-alpha, and IL-6, and increases NK cell cytolytic activity on tumor cells (Errington et al., 2008). Human studies have confirmed the capacity for reovirus to generate a pro-inflammatory environment, with intravenous delivery of reovirus to brain tumor patients being associated with increased CD8⁺ T cell tumor infiltration, likely attributed to the observed increase in CCL3, CCL4, and ICAM expression, which mediate migration to sites of inflammation (Samson et al., 2018). However, tumors from reovirus-treated patients were also noted to have greater expression of PD-1 and PD-L1 immune checkpoint proteins, highlighting a potential response mechanism by the tumor to counteract the stimulated immune system. While reovirus harbors significant therapeutic utility for the treatment of GBM, further characterization of the tumor’s response to infection is required, as well as consideration for combinatorial treatment with immune checkpoint blockades, such as anti-PD-1/PD-L1. As such, reovirus is no longer being investigated as a monotherapy, and further investigation with other treatment modalities is currently underway (Müller et al., 2020).

Parvovirus, specifically H-1 parvovirus (H-1PV), became the focus of many decades of cancer research after discovering that it possesses a natural tropism for human cancer cells in 1961 (Toolan, 1961). The rat is the natural host of H-1PV, and H-1PV has been shown to be nonpathogenic to humans by failing to produce new virus particles and induce cell lysis in normal, non-transformed cells (Angelova et al., 2015). However, it has been shown to infect and cause cell death in a wide range of cancers, including tumors of the bone, brain, breast, colon, lung, pancreas, and skin, as well as hematological disease such as Burkitt lymphoma, cutaneous T-cell lymphoma, and diffuse large B-cell lymphoma (Angelova et al., 2017; Bretscher and Marchini, 2019). The oncotropism of H-1PV involves a myriad of processes, including but not limited to factors essential for viral entry (PKCalpha, CDK1), replication (cyclin A/CDK2, E2F), and maturation (XPO1, PKB, PKC), as well as deficiency of mechanisms necessary to counter viral infection (type I IFN stress response), in tumor cells (Angelova et al., 2015). In vitro studies of H-1PV found selective killing of glioma cells via a cathepsin-mediated mechanism, which translated to prolonged survival in glioma-bearing rats treated with intratumoral, intravenous, and intranasal H-1PV inoculation, again via elevated cathepsin activation and activity (Di Piazza et al., 2007; Geletneky et al., 2010). The first clinical trial for GBM was then initiated by Geletneky et al., who found H-1PV (ParvOryx01) to be an immunogenic stimulus in patients with recurrent GBM (Geletneky et al., 2017). Treated patients were found to have strong leukocytic infiltration, predominantly CD8⁺ T cells and, to a lesser degree, CD4⁺ T cells. Additionally, a promising finding was that the increased CD8⁺ T cell population did not coincide with a responsive increase in tumor-invading Tregs that is typical of GBM. This is in line with in vitro studies that have found H-1PV to capable of suppressing the activity of Tregs (Moralès et al., 2012). Given these findings, in addition to a case series of patients successfully treated with a combination of NDV, parvovirus, and vaccinia virus, further characterization of the anti-tumor and immune sensitizing effects of H-1PV and investigation of its clinical effects in GBM through a randomized controlled trial is required (Gesundheit et al., 2020).

Replicating retroviral vectors have been harnessed for their ability to deliver ‘suicide genes,’ or genes that encode for proteins capable of converting non-toxic into toxic drugs upon delivery of a prodrug (Li et al., 2021). This was the basis of a retroviral vector encoding for HSV thymidine kinase (HSV-tk), which could convert ganciclovir (GCV) into GCV triphosphate, an inhibitor of DNA replication. HSV-tk was tested in 15 GBM patients who received intratumoral injection of HSV-tk followed by intravenous GCV administration, with 1 long-term survivor at 220 weeks (Ram et al., 1997). Twelve rGBM patients exhibited no serious adverse events in a phase I/II study after receiving intratumoral injection of HSV-tk cells intra-operatively, with 25% of patients living longer than 12 months (Klatzmann et al., 1998). A similar study in 13 ndGBM patients demonstrated significantly elevated soluble Fas ligand and IL-12 levels in serum of HSV-tk treated patients versus controls but did not find increased tumor-infiltrating lymphocytes at the resection cavity or activation of T or NK cells (Rainov et al., 2000). Another phase I/II clinical trial in 48 rGBM patients who received intracerebral injection of HSV-tk demonstrated no serious adverse events or evidence of virus in the serum or tissue at time of repeat resection (Shand et al., 1999). However, a phase III trial studying the effects of HSV-tk and GCV treatment in 248 ndGBM patients found no significant difference in PFS and median survival between treated and control patients, which was hypothesized to be due to poor transduction efficiency (Rainov, 2000).

Toca 511 is a replication-competent retroviral vector that delivers cytosine deaminase (CD) to convert prodrug 5-fluorocytosine (5-FC) to antineoplastic agent 5-fluorouracil (5-FU) (Huang et al., 2013). Toca 511 was tested in 45 high grade glioma patients in a phase I trial that demonstrated a significantly longer OS of 13.6 months compared to control (Cloughesy et al., 2016). These favorable results led to a phase II trial testing Toca 511 to SOC in GBM patients and demonstrated no significant difference in median survival between the two groups (Cloughesy et al., 2020).

Despite the successes seen with immunotherapy against GBM in preclinical models, these results have not translated well in the clinical setting, as fewer than 10% of GBM patients have been shown to respond to immunotherapy (McGranahan et al., 2019) (p2401). As such, it is important to understand the intrinsic and adaptive forms of resistance exhibited in GBM (Table 5).

Several factors contribute to immune evasion in GBM. Unlike other cancers, GBM has been known to display extensive intratumoral heterogeneity (Patel et al., 2014), which confounds efforts to identify high-quality clonal neoantigens regardless of the form of immunotherapy trialed. While GBM has traditionally been classified into proneural, classical, mesenchymal, and neural subsets, single cell RNA-sequencing has revealed cross-over among the subtypes within the same tumor and the presence of hybrid cellular states (Patel et al., 2014). Genomic studies of GBM heterogeneity have revealed the presence of a CD133⁺ chemo- and radio-resistant cancer stem cell (CSC) population responsible for tumor initiation and found to have higher expression in recurrent tumors (Bao et al., 2006; Liu et al., 2006; Tamura et al., 2013; Qazi et al., 2017), but even then, CD133⁺ has failed to be a universal marker of CSCs given the discovery of similarly functioning CD133^- cell populations (Chen et al., 2010).

In addition to the extensive phenotypic heterogeneity of cells within a tumor, there is a strong immunosuppressive microenviroment within GBM tumors that remains a major barrier to immunotherapy efficacy (Moserle and Casanovas, 2013). At baseline, GBM patients exhibit a lower number of circulating T cells despite being treatment-naïve (Chongsathidkiet et al., 2018). This phenomenon has been attributed to T cell sequestration in the bone marrow in the setting of GBM as well as other intracranial tumors, due to tumor-mediated internalization of G-protein coupled receptor sphingosine-1-phosphate receptor 1 (S1P1) (Chongsathidkiet et al., 2018). The same study demonstrated reversal of T cell sequestration in murine models of GBM upon inhibiting S1P1 internalization (Chongsathidkiet et al., 2018).

The circulating T cells that are available in the setting of GBM often display dysfunctional phenotypes including but not limited to tolerance and exhaustion (Brooks et al., 1977; Woroniecka et al., 2018). The exhausted phenotype is commonly seen in chronic viral infections and various cancers and has recently been suggested to be irreversible in spite of antigen clearance due to a novel concept referred to as ‘epigenetic scarring’ (Abdel-Hakeem et al., 2021). Persistent proliferation in the setting of chronic antigenic exposure can also lead to shortening of telomeres resulting in T cell senescence (Woroniecka et al., 2018). Importantly, GBM utilizes naturally occurring mechanisms of immune tolerance to promote FasL-mediated peripheral deletion of T cells and recruitment of Tregs via expression of IDO1 on dendritic cells and T cell immunoglobulin- and mucin domain-containing molecule 4 (TIM4) on macrophages (Xu et al., 2011; Choi et al., 2012; Woroniecka et al., 2018). Beyond lymphocytes, the presence of myeloid-derived suppressor cells (MDSC) in both peripheral blood and intracranially has been shown to contribute to immune suppression and tumor progression via expression of arginase, inducible nitric oxide synthase, and reactive oxygen or nitrogen species (Bronte and Zanovello, 2005; Gabrilovich and Nagaraj, 2009; Marvel and Gabrilovich, 2015). Tumor-associated macrophages (TAM) have been similarly implicated via secretion of immunosuppressive cytokines IL10 and TGFβ secondary to induction by CSCs. (Wu et al., 2010; Zhou et al., 2016).

While the central nervous system (CNS) is known to be a site of immune privilege secondary to naturally occurring mechanisms of immune homeostasis such as the blood brain barrier (BBB) and resident microglia (Desland and Hormigo, 2020), the immunosuppression intrinsic to GBM and independent of location in the CNS is clear when comparing the microenvironment and immunotherapeutic results of GBM to that of brain metastases (Friebel et al., 2020). Single cell analysis has revealed that the GBM microenvironment has higher expression of tissue-resident microglia while the metastatic tumor environment has higher expression of tissue-invading leukocytes (Friebel et al., 2020). At the same time, 80% of the leukocytes expressed in the GBM microenvironment were found to be classically immunosuppressive TAMs while most leukocytes found in the metastatic tumor environment were T cells (Friebel et al., 2020). These dichotomous immune findings may help explain why ICIs have seen more success in treating patients with brain metastases from melanoma or NSCLC compared to those with GBM (Goldberg et al., 2016; Tawbi et al., 2018; Kluger et al., 2019).

GBM has been shown to acquire forms of secondary resistance in the setting of recurrence and treatment. Patients with rGBM who initially responded to anti-PD1 immunotherapy have been found to have loss of neo-epitopes and delayed upregulation of immunosuppressive genes upon recurrence (Zhao et al., 2019). Despite the expression of EGFRvIII in 19% of ndGBM, a vaccine trial targeting EGFR found that 82% of tumors had lost EGFRvIII expression upon recurrence (Sampson et al., 2010; Brennan et al., 2013).

The current standard of care for GBM can also further exacerbate immune evasion. Dexamethasone, which is commonly used to reduce peri-tumoral edema and temporarily improve neurological symptoms, has been shown to upregulate expression of CTLA-4 on T cells and subsequently dampen patient response to checkpoint blockade (Giles et al., 2018). A unique phenotype of altered monocytes (CD14⁺ HLA-DR^-) has also been identified to represent an immunosuppressive population whose levels increase in response to dexamethasone treatment (Gustafson et al., 2010). TMZ treatment may also result in profound lymphopenia, and in GBM patients, TMZ-induced lymphopenia is worsened by both the absence of a compensatory increase in proliferation-inducing cytokines and the failure of adoptive transfer to increase T cell counts (Karachi et al., 2018). Radiotherapy alone promotes secretion of classically immunosuppressive cytokines IL-6, IL-8, IL10 (Tabatabaei et al., 2017; Authier et al., 2015) but when combined with chemotherapy as in the current SOC, the combination has been shown to severely deplete CD4⁺ and CD8⁺ T cells and increase the proportion of Tregs (Fadul et al., 2011). In a study of 96 patients with high grade gliomas, patients with CD4⁺ counts under 200 at 2 months after therapeutic initiation were found to have significantly shorter survival than those with higher counts (13.1 vs 19.7 months), highlighting the importance of understanding the effects of treatment in contributing to drug resistance in GBM (Grossman et al., 2011).

The use of immunotherapeutic agents discussed here may also contribute to secondary resistance. In melanoma patients treated with anti-PD1 blockade who subsequently relapsed, genomic analysis of paired primary and recurrent tumor revealed alterations in β2 microglobulin and JAK1/2 genes that were the main drivers of acquired PD1 resistance (Shin et al., 2017). It has also been shown in lung cancer that downregulation of one checkpoint may lead to subsequent upregulation of other checkpoint molecules, such as the upregulation of TIM3 in the setting of anti-PD1 blockade (Koyama et al., 2016). In glioblastoma, the use of CAR T cells targeting EGFRvIII has been linked to a compensatory influx of immunosuppressive Tregs into the tumor microenvironment as well as loss of EGFRvIII expression in surgically resected tumors post-treatment (O’Rourke et al., 2017). Similarly, the use of CAR T cells targeting IL13Rα2 was shown to lead to improved survival in preclinical murine models, but recurrent gliomas post-treatment experienced downregulation of the target antigen, suggesting that the use of CAR T cells targeting a singular antigen may select for GBM cells that lack expression and subsequently allow for disease progression (Krenciute et al., 2017).

Many efforts have been directed towards targeting the myeloid compartment, specifically M2-type TAMs and MDSCs, given their wide variety of immunosuppressive functions and dense abundance in the tumor mass (Glass and Synowitz, 2014). CD47, a ubiquitously expressed protein on the surface of GBM cells, interacts with signal-regulatory protein alpha (SIRP α ) on the membranes of macrophages to inhibit phagocytosis by promoting M2 polarization through the PI3K/AKT signaling pathway (Lin et al., 2018; Liu et al., 2020). Inhibiting the CD47/SIRPα anti-phagocytic and pro-M2 axis in GBM has shown promising results by shifting macrophages towards the antitumorigenic M1 phenotype, reducing tumor burden by enhancing macrophage-mediated phagocytosis, and improving survival in xenograft mouse models (Zhang et al., 2016; Gholamin et al., 2017). While anti-CD47 monotherapy has since been shown to be inefficient in immune competent hosts, combinatorial treatment of CD47 blockade and TMZ has led to augmented PD-1 blockade responses and improved survival in murine models (von Roemeling et al., 2020). Its potential and preliminary success as a target for combination anti-PD-1 immunotherapy is hypothesized to be due to increased phagocytosis of tumor cells and induction of the endoplasmic stress response that results in more efficient T cell priming (von Roemeling et al., 2020).

CD47 inhibitors have been tested in several phase I/II clinical trials for hematological and advanced solid malignancies (Jalil et al., 2020). While preclinical studies in mice have shown good tolerance with minimal signs of toxicity, clinical trials have encountered issues regarding hematological toxicity including anemia, leukopenia, and thrombocytopenia (Gholamin et al., 2017; Li et al., 2018). CD47 is ubiquitously expressed by non-cancerous cells of the hematopoietic system, which makes them an alternative binding site for systemically delivered anti-CD47 antibodies (Ishikawa-Sekigami et al., 2006; Hu et al., 2020). Consequently, non-tumor binding of antibodies was found to induce unintentional FcR-mediated phagocytosis of red blood cells leading to toxic anemia (Zhang et al., 2020). Studies have experienced success in mitigating hematological toxicity by delivering a low, priming dose of anti-CD47 antibodies to induce a predictable, transient anemia and compensatory reticulocytosis for non-Hodgkin’s lymphoma (Advani et al., 2018). However, future investigations of applying anti-CD47 treatment to GBM must still consider the possibility of hematological toxicity and develop effective strategies to bypass it.

Similar to CD47, colony stimulating factor-1 receptor (CSF-1R), a member of the receptor protein tyrosine kinase (rPTK) family, has also been implicated in the differentiation of myeloid cells into M2 macrophages (Dai et al., 2002). In GBM xenograft models, CSF-1R inhibition decreases M2 markers on TAMs but does not deplete the cell population due to tumor-secreted factors, notably GM-CSF and IFN- γ (Pyonteck et al., 2013). However, blockade of CSF-1R has also been shown to increase expression of key chemotactic factors that promote the influx of tumor-infiltrating lymphocytes, a finding consistent with other cancer types (Mok et al., 2014; Ries et al., 2014; Antonios et al., 2017). When inhibition was performed in conjunction with anti-PD-1 and DC vaccination, TIL dysfunction was reversed, suggesting that the development of immune resistance to active vaccination in GBM can be abrogated by combinatorial inhibition of CSF-1R and PD-1 (Antonios et al., 2017). Clinical trials with cabiralizumab, an anti-CSF-1R monoclonal antibody, are currently underway for malignant glioma in conjunction with the anti-PD-1 antibody nivolumab (Brahmer et al., 2016). They have also been performed for other cancers, with a recent phase 1 study in anti-PD-L1 resistant patients with melanoma, kidney cancer, or non-small lung cancer exhibiting low clinical response rates but safe upregulation of pro-inflammatory cytokines and chemokines (Weiss et al., 2021). Overall, CSF-1R inhibitors have been generally well-tolerated, increasing its promise for combinatorial immunotherapy strategies (Cannarile et al., 2017). However, short-lived, asymptomatic elevations in liver enzymes, notably alanine transaminase, aspartate aminotransferase, creatine kinase, and lactate dehydrogenase, have been observed across multiple different studies for different cancers, including GBM (Butowski et al., 2014; Papadopoulos et al., 2017; von Tresckow et al., 2015). Hepatotoxicity is believed to be due to partial depletion of CSF-1R + macrophages/Kupffer cells of the liver leading to reduced physiologic clearance (Ries et al., 2014). While no functional or structural liver damage has been associated with anti-CSF-1R treatment, careful monitoring will be required in future studies.

In addition to targeting the polarization of M2 macrophages, potential adjuvant treatments have also aimed to target the function of M2-type macrophages. Specifically, CD73, an ectonucleotidase preferentially expressed on M2 polarized macrophages, interacts with its upstream signaling molecule CD39 to facilitate the production of adenosine from extracellular ATP (Zanin et al., 2012; Antonioli et al., 2013). Adenosine is a known promoter of tumor proliferation and angiogenesis, and in the case of GBM, it has been implicated in the development of TMZ chemoresistance and CD8⁺ T cell dysfunction via the A_2B adenosine receptor (Takenaka et al., 2019; Yan et al., 2019). CD73 expression on myeloid cells has also been correlated with higher co-expression of the immunosuppressive and protumorigenic CCR2, CCR5, ITGAV, and CSF-1R chemokine receptors (Goswami et al., 2020). Clinical trials targeting these chemokine receptors are underway, but CD73 may be a more relevant target given its high co-expression and the limited success observed thus far. A preliminary study that induced prolonged survival in CD73 knockout models of GBM treated with dual blockade of PD-1 and CTLA-4 highlights the potential of CD73 for combination therapy and need for further investigation (Goswami et al., 2020). Given that clinical trials focusing on targeting CD73 in GBM have yet to commence or are ongoing in other malignances, there is limited awareness of its associated adverse events beyond results from animal studies (Jin et al., 2021). While studies in mice have shown that CD73 plays a role in platelet aggregation and protection of the heart, kidney, and lungs from ischemia, animal studies have shown good tolerability to CD73 blockade (Stagg, 2012; Antonioli et al., 2016; Azambuja et al., 2020). However, these results should not be used to extrapolate responses in humans, and a better understanding of the safety of targeting CD73 will be had following the conclusion of ongoing clinical trials.

Myeloid-derived suppressor cells (MDSC), a heterogenous population of CD11b⁺ CD33⁺ HLA-DR^- myeloid cells found in the peripheral blood and tumor mass of GBM patients, have also emerged as a potential target for sensitizing the tumor microenvironment to immunotherapy given their myriad immunosuppressive functions associated with poor prognosis (Almand et al., 2001; Alban et al., 2018). MDSCs have been implicated in GBM tumor progression through the inhibition of T cells, NK cells, dendritic cells, and macrophages, the expansion and differentiation of T regulatory cells, and the promotion of immunosuppressive B cells (Raychaudhuri et al., 2011; Mi et al., 2020). To control their tumorigenic activity, approaches have attempted to target the infiltration, expansion, and activation of MDSCs. Prostaglandin E2 (PGE2) and C-C motif chemokine ligand 2 (CCL2) have both been associated with the recruitment of MDSCs to tumor tissue, while macrophage migration inhibitory factor (MIF) signaling through the chemokine ligand 2 (CXCL2) and MIF/C-X-C motif chemokine receptor 2 (CXCR2) axis has been linked to the differentiation of myeloid cells into MDSCs (Simpson et al., 2012; Chang et al., 2016).

In preclinical models, inhibition of PGE2 production through COX2 inhibitors, specifically acetylsalicylic acid and celecoxib, led to both the suppression of gliomagenesis and reduction of MDSCs in the tumor microenvironment through a decrease in CCL2 (Fujita et al., 2011; Shono et al., 2020). Furthermore, COX2 inhibition in GBM has also garnered interest for its radiosensitizing effects in vivo (Ma et al., 2011). While the enthusiasm for COX2 as a therapeutic target has stagnated due to multiple cohort studies and clinical trials that inversely correlated COX2 inhibitors with survival, its efficacy as an immunotherapeutic adjuvant has yet to be thoroughly investigated and warrants further study (Qiu et al., 2017). Alternatively, it may also be appropriate to shift focus towards PGE2 as a therapeutic target rather than COX2 itself, especially given the broad range of well-known adverse effects associated with COX2 inhibition such as hypertension, congestive heart failure exacerbation, renal impairment, and other cardiovascular events (Mukherjee et al., 2001; Wright, 2002). There are currently no clinical trials focused on targeting PGE2. However, an oncolytic vaccina virus was recently designed to inactivate PGE2 and successfully reduce the number of MDSCs and T regulatory cells in a mouse tumor model as well as potentiate the response to anti-PD1 therapy (Hou et al., 2016). Thus, the safety associated with targeting PGE2 is currently unknown beyond those associated with COX2 inhibition, however, downregulating MDSCs will pose the risk of autoimmune side-effects such as autoreactive T-cells, more so than traditional immunotherapies given the wider spectrum of activity in MDSCs than targeted immune checkpoint inhibitors (Yang et al., 2017).

CCL2 inhibition through anti-CCL2 monoclonal antibodies (mAb) has also experienced success in reducing the population of MDSCs and improving survival in GBM xenograft models (Zhu et al., 2011). Additionally, inhibition of the CCL2 receptor (CCR2) in conjunction with PD-1 checkpoint inhibition extended survival in GBM-bearing mice, highlighting its ability to augment immunotherapy (Flores-Toro et al., 2020). Clinical trial using carlumab monotherapy, a human IgG1_k monoclonal antibody against CCL2, had little success in patients with advanced solid tumors and metastatic prostate cancer, however, results from CCL2 blockade with immune checkpoint blockade have yet to be reported (Pienta et al., 2013; Sandhu et al., 2013). While carlumab clinical trials were well-tolerated with mild-to-moderate adverse events, neutropenia occurred commonly in a multicenter phase 1b study, which may be due to the elimination of CCL2’s anti-apoptotic effect in neutrophils and increase the risk of infection in patients (Brana et al., 2015; Gschwandtner et al., 2019).

The MIF signaling axis has also recently garnered interest as a target given the high levels of the MIF non-cognate receptor CXCR2 expressed on MDSCs and the enhanced CD8⁺ T cell activity observed in the tumor microenvironment following treatment with ibudilast, an inhibitor of MIF-CD74 interactions (Alban et al., 2020). Ibudilast is currently being investigated in clinical trials as an adjuvant to TMZ, given its ability to sensitize GBM cells to TMZ, safely penetrate the BBB, and confer minimal adverse events, with phase 1 studies and clinical trials for other neurological diseases such as multiple sclerosis reporting gastrointestinal side effects, headaches, and depression as the most concerning (Rolan et al., 2008; Fox et al., 2018). However, given the preliminary findings of increased lymphocyte activation, additional clinical trials pairing ibudilast with immune checkpoint inhibitors like anti-PD-1 to maximize cytolytic activity against tumor cells should also be considered.

IL-6 drives myeloid-based immunosuppressive activity through the induction of PD-L1 expression on MDSCs (Lamano et al., 2019). While IL-6 neutralization has been shown to enhance T cell tumor infiltration, it alone does not sensitize the tumor to immune checkpoint blockade via anti-PD-1 or anti-CTLA-4 (Yang et al., 2021). However, CD40 stimulation in conjunction with IL-6 inhibition and PD-1 and CTLA-4 immune checkpoint blockade did produce clinically relevant responses in syngeneic GBM models, notably the reversal of macrophage-mediated immune suppression and extended survival Yang et al., 2021. This finding supporting dual-targeting of IL-6, CD40, and multiple immune checkpoints emphasizes the paradigm shift towards strategically selecting multiple interrelated targets that increase the possibility of a successful response to immunotherapy. An ongoing clinical trial investigating the addition of tocilizumab, a monoclonal antibody against IL-6, alongside atezolizumab and fractionated stereotactic radiation therapy in recurrent GBM will shed more light on the clinical efficacy of concurrent IL-6 and PD-L1 blockade in GBM. Given the use of tocilizumab for other conditions such as rheumatoid arthritis, adverse effects are minimal, acceptable, and promising for utilization in GBM treatment. Complications mainly consist of neutropenia and rare thrombocytopenia given IL-6 receptors on neutrophils that may bind monoclonal antibodies and lead to opsonization and neutrophil phagocytosis (Wright et al., 2014) (p6), (Smolen et al., 2013).

In addition to cells from the myeloid compartments, cells derived from the lymphoid lineage also possess immunosuppressive functions and serve as potential therapeutic targets, most notable of which are Tregs. Tregs are characterized as a CD25⁺ FOXP3⁺ subset of CD4⁺ cells that divert immune responses away from cytotoxic Th1-mediated responses and towards Th2-mediated responses in part through increasing the expression of CTLA-4 and decreasing the secretion of IL-2 and IFN- γ (Jonuleit et al., 2001; Cosmi et al., 2004). To improve the antitumor immune response, Treg depletion therapy has been attempted in GBM models through combinatorial anti-CXCR4 and anti-PD1 immunotherapy, which have improved survival rates via decreased levels of Tregs and MDSCs, improved CD4⁺/CD8⁺ ratios, and increased levels of pro-inflammatory cytokines (Wu et al., 2019). However, systemic Treg depletion through antibody therapies pose the risk of autoimmunity (Sakaguchi et al., 2006). Thus, recent findings targeting GITR, a surface immunomodulatory receptor highly expressed on GBM Tregs but lowly expressed on systemic Tregs, are especially encouraging for Treg depletion therapies. When targeting GITR, murine Tregs were reprogrammed into CD4⁺ Th1 cells capable of producing IFN- γ and engaging in cytotoxic activity against GBM cells (Amoozgar et al., 2021). Furthermore, anti-GITR and anti-PD1 combinatorial treatment was capable of prolonging survival in multiple GBM murine models, with a subset experiencing complete tumor eradication and immune memory following tumor re-challenge (Miska et al., 2016; Amoozgar et al., 2021). Dual radiation and anti-GITR therapy has also been associated with improved survival via increased CD4⁺ effector T cell infiltration and IFN- γ , IL-2, and TNF- α secretion, suggesting that the two modalities synergize and highlighting the immune sensitizing effects of radiotherapy on the tumor microenvironment (Patel et al., 2016). A clinical trial investigating the use of anti-GITR, anti-PD1, and stereotactic radiosurgery in recurrent GBM was recently initiated and will provide valuable insight into how and to what extent the tumor immune microenvironment responds to treatment.

Despite its immunosuppressive mechanisms documented earlier, radiotherapy can also sensitize the immune response via its ability to increase MHC-1 expression on the surface of tumor cells, leading to better antigen presentation and recognition by cytotoxic T cells (Rajani et al., 2019; Sevenich, 2019). Antiangiogenic therapies have also been considered as a treatment option for overcoming resistance to immune checkpoint therapies in GBM. Dual blockade of VEGF and Ang-2 with concurrent anti-PD-1 treatment has been shown to extend survival, increase cytotoxic T lymphocyte infiltration, and decrease MDSC and Treg abundance (Fukumura et al., 2018; Di Tacchio et al., 2019).

Overall, clinical trials investigating multimodal therapies in GBM that integrate conventional treatments, such as radiotherapy and anti-angiogenic therapy, with novel immunotherapies and strategies for bypassing tumor resistance are currently underway. This combinatorial approach to treatment exemplifies the new frontier of GBM immunotherapy that leverages strategic combinations of multiple treatments to reverse immunosuppression within the microenvironment and maximize the potential of immunotherapy.