Therapeutic Strategies for Overcoming Immunotherapy Resistance Mediated by Immunosuppressive Factors of the Glioblastoma Microenvironment

Various mechanisms of treatment resistance have been reported for glioblastoma (GBM) and other tumors. Resistance to immunotherapy in GBM patients may be caused by acquisition of immunosuppressive ability by tumor cells and an altered tumor microenvironment. Although novel strategies using an immune-checkpoint inhibitor (ICI), such as anti-programmed cell death-1 antibody, have been clinically proven to be effective in many types of malignant tumors, such strategies may be insufficient to prevent regrowth in recurrent GBM. The main cause of GBM recurrence may be the existence of an immunosuppressive tumor microenvironment involving immunosuppressive cytokines, extracellular vesicles, chemokines produced by glioma and glioma-initiating cells, immunosuppressive cells, etc. Among these, recent research has paid attention to various immunosuppressive cells—including M2-type macrophages and myeloid-derived suppressor cells—that cause immunosuppression in GBM microenvironments. Here, we review the epidemiological features, tumor immune microenvironment, and associations between the expression of immune checkpoint molecules and the prognosis of GBM. We also reviewed various ongoing or future immunotherapies for GBM. Various strategies, such as a combination of ICI therapies, might overcome these immunosuppressive mechanisms in the GBM microenvironment.


Introduction-Glioblastoma and Its Epidemiological Features
Glioblastoma (GBM) is the most common and lethal malignant brain tumor, classified as grade IV by the World Health Organization (WHO), and reportedly diagnosed in 12.0% of all brain tumor patients [1][2][3]. GBM occurs frequently in people aged 50 to 60 years and is more common in men. Even with the current standard adjuvant therapy using temozolomide (TMZ) and radiation therapy (RT), the median relapse-free survival and median overall survival (OS) of GBM patients are 6.9 months and 14.6 months, respectively, and the 5-year survival rate is less than 10% [4]. Furthermore, in case of recurrence after standard therapy, the median survival time after recurrence is only 1 to 10.8 months, even if conditions such as the Karnofski performance status (KPS), tumor site, tumor volume, and expression of tumor markers are taken into account [5][6][7]. New treatment strategies for GBM patients are being successively developed. For instance, clinical trial of new drugs such as carmustine-loaded polymers (Gliadel wafers) without TMZ [8] and an anti-vascular endothelial growth factor (VEGF) antibody (bevacizumab) added to standard therapies [9] did not significantly improve the OS of initially diagnosed GBM patients.
In GBM, IDH1/2 mutations and MGMT methylation are already considered prognostic and therapeutic markers but further therapeutic efficacy indicators are desired [10][11][12]. Anaplastic astrocytomas (WHO grade III) typically have IDH1 mutations and MGMT promotor methylation is an important prognostic marker [13]. Anaplastic oligodendrogliomas (WHO grade III), in contrast, typically have 1p19q co-deletion in addition to IDH1 mutation and have a better prognosis than GBM, even when they pathologically show necrotic features [14]. Regarding the changes in molecular expression before and after recurrence of GBM, the expression of p53 and EGFRvIII are decreased in recurrent tumor cells [15,16].

Glioblastoma Immune Microenvironment
Immune status biomarkers, such as the pro-/anti-inflammatory phenotype, infiltration rate, and activation status of tumor-infiltrated lymphocytes (TILs) in the tumor microenvironment and expression of immunosuppressive factors by tumor cells, are considered to be prognostic for various cancers (e.g., breast, colorectal) [17,18].

Association between Expression of ICMs and Prognosis in GBM
The immune system is a biological defense mechanism that has evolved to prevent/eliminate the invasion of external enemies (such as bacteria and viruses) but it also eliminates tumor cells. The effector function of immune cells is controlled by ICM expression. When CD8 T cells attack target cells, the first signal of the MHC molecule carrying the target antigen followed by co-stimulative ICMs-such as CD28, ICOS, or CD134 (OX40)-is required [54]. The expression of NR4A1, which inhibits T cell effector function, is also an obstacle to immune activation [55] MHC-deficient cells are targeted and attacked by NK cells. In the absence of these co-stimulative ICMs, however, T cells lapse into an anergic, immune-deficient state. Activated CD8 T cells attack target cells one after another while controlling their own activity by the expression of co-inhibitory ICMs such as cytotoxic T-lymphocyte antigen-4 (CTLA-4), PD-1, and T-cell immunoglobulin and mucin domain 3 (TIM-3) to prevent excessive activation and autoimmune effector activity [54]. On the other hand, tumor cells also utilize these inhibitory ICMs for immune escape. Therefore, novel therapies focused on these inhibitory ICMs and T-cell exhaustion have recently been developed [56]. Among them, inhibitory antibodies of the PD-1/PD-L1 pathway have shown therapeutic effects in various types of cancers, especially in high somatic mutation burden tumors such as melanoma and non-small cell lung cancer [57,58]. The relationship between PD-1/PD-L1 expression and prognosis has been reported for several types of cancers [59].
The prognostic effect of PD-1/PD-L1 expression in the tumor microenvironment of GBM has also been analyzed. Using immunohistochemical analysis, Nduom and coauthors reported that PD-L1 expression was a poor prognosis marker in 94 GBM patients [25]. Similarly, in the analysis of 17 GBM patient samples, Liu and coauthors found that over 10 PD-L1-expressing cell phenotypes in tumor tissue were associated with poor prognosis [23]. However, in the analysis of PD-L1 mRNA expression in 135 specimens (117 initial specimens and 18 local recurrence specimens), no such correlations were found [21]. We have also previously reported that PD-1/PD-L1 expression in primary tumors does not correlate with GBM prognosis [24]. As described above, the correlation between PD-1/PD-L1 expression and patient prognosis is still controversial in GBM [21,23,25]. In addition, infiltration of Tregs and IDO1 expression were reported as poor prognostic factors [60,61]. Thus, such molecules in the tumor microenvironment are potential therapeutic targets [62].

GBM Immunotherapy and Microenvironmental Changes after Recurrence
Immunotherapy for malignant tumors originated in the 1890s, when Dr William Coley discovered the relationship between erysipelas infection and tumor disappearance [63]. Cytokine therapy that administers interleukins and interferons, adoptive immunotherapy that extracts and activates lymphocytes and dendritic cells from the blood, and artificially synthesized cancer vaccine therapy that administers cancer antigens or cancer tissue-processed products added with immune adjuvants have been developed [64]. Our group has previously reported a clinical trial using autologous tumor-specific T lymphocytes [65] and autologous natural killer cells [66] for recurrent malignant glioma. Another report on GBM antitumor therapy compared autologous formalin-fixed tumor vaccine (AFTV), manufactured from autologous formalin-fixed tumor tissue [64], a vaccine with RT [67] and TMZ concomitant with RT standard therapy [68]. In the most recent phase IIa clinical trial, the median OS was 22.2 months and the 3-year survival rate was 38% [68,69]. The number of TILs was increased in recurrent tumor tissues after AFTV therapy compared with initial tumor tissue and the number of Ki-67-positive tumor cells tended to be decreased [70]. Furthermore, the number of PD-1-positive inactivated/exhausted lymphocytes was increased in recurrent GBM tissue, especially in patients treated with AFTV before recurrence ( Figure 3) [24].
Clinical trials on the anti-PD-1 antibody ICI nivolumab failed to show an advantage over control bevacizumab adjuvant therapy for recurrent GBM patients, indicating that the therapeutic effects of ICI alone appear to be limited to recurrent GBM [71]. On the other hand, ICIs may be effective if used as neoadjuvant therapy for resectable, recurrent GBM [72]. M2Mϕ infiltration in the tumor microenvironment and the phosphatase and tensin homolog (PTEN) mutation status may affect the therapeutic effect of ICIs [73]. The efficacy of ICI treatment may thus be improved by combination with other treatments or by patient selection according to tumor mutation genotype. Improving the treatment efficacy of immunotherapy and establishing immune memory will lead to increased long-term survival of patients with GBM. Factors considered as combinational therapeutic targets are described below (Figures 3 and 4).
Cancers 2020, 12, x 5 of 22 combination with other treatments or by patient selection according to tumor mutation genotype. Improving the treatment efficacy of immunotherapy and establishing immune memory will lead to increased long-term survival of patients with GBM. Factors considered as combinational therapeutic targets are described below (Figures 3 and 4).   combination with other treatments or by patient selection according to tumor mutation genotype. Improving the treatment efficacy of immunotherapy and establishing immune memory will lead to increased long-term survival of patients with GBM. Factors considered as combinational therapeutic targets are described below (Figures 3 and 4).

Regulatory T Cells (Tregs) as a Therapeutic Target of GBM
Tregs that mainly differentiate from the naïve T-cell fraction to suppress excessive activation of T cells are described as CD25 + FOXP3 + CD4 T cells and function to suppress antitumor immunity by effector T cells in the tumor microenvironment [74]. Tregs are classified by molecular marker expression level and by examination of their functions; thus, it has been reported that the fraction of FoxP3 high CD45RA-CD25 high cells has strong immunosuppressive abilities, including expression of co-inhibitory ICMs, consumption of autoexpanding IL-2, production of inhibitory cytokines, consumption of nutrients, and killing activity against effector T cells [75]. For many types of malignant tumors, the proportion of Tregs in the tumor and peripheral blood and the ratio of Tregs to CD8 cells correlate with the prognosis [76][77][78][79][80][81][82][83]. In GBM, there are opposite reports regarding the relationship between Treg proportion and the CD8/Treg ratio with prognosis [84][85][86][87]. On the other hand, Treg accumulation in the tumor margins [88] and inhibitory ability in murine models have been reported [89], indicating that Tregs can also be therapeutic targets in GBM. Combination therapy targeting angiogenesis using VEGF and angiopoietin-2 inhibitors combined with anti-PD-1 antibody inhibited Treg and MDSC infiltration and increased proliferation and anti-tumor activity of glioma-infiltrating CD8 T cells [90]. Anti-CD25 antibody, anti-CCR4 antibody, anti-CXCR4 antibody, anti-CTLA-4 antibody, IL-10, and TGFβ inhibitors have been clinically developed as Treg depletion therapy ( Figure 4) [91][92][93][94]. Combination therapy using Treg depletion with anti-CD25 or anti-CXCR4 antibodies and ICIs was shown to be effective in a murine glioma model [94,95]. Regarding Treg-depleting antibodies, methods of suppressing the onset of autoimmune diseases by adjusting the administration method and period are also under investigation [96,97]. However, these antibody therapies are systemically effective and may cause autoimmune adverse effects [75,96,97]. In addition, there are several types of Tregs: a fraction called FOXP3 low 'fragile or non' Tregs with low immunosuppressive function and a truly functional fraction called FOXP3 high 'effector' Tregs. Development of a therapy targeting these effector Tregs is required [96][97][98].

Myeloid-Derived Suppressor Cells (MDSCs), M2 Macrophages (M2Mϕs), and Regulatory B Cells (Bregs) as Therapeutic Targets of GBM
MDSCs are detected as a fraction of CD11b + CD33 + HLA-DR − cells in humans and are classified into M-MDSC CD11b + CD33 + HLA-DR − /CD14 + CD15 − cells and PMN (polymorphonuclear)-MDSC CD11b + CD33 + HLA-DR − /CD14 -CD15 + cells [99]. The tumor infiltration and blood level of MDSCs correlate with the prognosis of many types of cancers [99]. MDSCs accumulate in the tumor microenvironment and suppress anti-tumor immunity by signals-such as cytokines, chemokines, and extracellular vesicles-secreted from tumor cells [100,101]. These bone marrow-derived cells cause immunosuppression of the tumor microenvironment not only by humoral factors such as IL-10 and TGFβ, but also by ICMs, extracellular vesicles, and nutrient consumption [101]. In GBM, several reports revealed increased MDSC levels in the blood or tumor tissue as a poor prognostic factor [84,[102][103][104]. CCL2 produced by murine GBM cells recruits CCR4 + Tregs and CCR2 + Ly-6C + mMDSCs [105] while MIF produced by tumor stem cells activates MDSCs, resulting in an immunosuppressive tumor microenvironment [105]. In addition, CD49d mRNA levels in tumor tissue, suggesting CD49d + MDSC/TAM/Treg levels, strongly correlate with GBM prognosis [103]. On the basis of such reports, combination immunotherapy for MDSC inhibition has been developed. Representative examples are described below. Combination therapy using a CCR2 antagonist (CCX872) and anti-PD-1 antibody to inhibit MDSC recruitment into the tumor microenvironment for KR158 (HGG-like cells) and 005GSC (stem-like cells) murine models increases IFNγ + TILs and prolongs overall survival [106]. Combination therapy using anti-PD-1 antibody and an inhibitor of CXCR4, a receptor for CXCL12/SDF-1 that contributes to the maintenance of tumor stem cells, also increases local infiltration of CD4/8 T cells by suppressing Treg and MDSC tumor invasion and prolongs the survival of tumor-bearing mice [94]. CD200 is required to maintain bone marrow cell homeostasis; however, it may cause exacerbation due to increased MDSC infiltration during tumor growth. Therefore, combination of a CD200 synthetic peptide that leads to production of anti-CD200 antibodies in the organism and inhibits MDSCs with a tumor vaccine has been developed [107]. Since IL-6 produced by glioma cells promotes PD-L1 expression of MDSCs, the immunosuppressive function of MDSC is suppressed by IL-6 KO or an anti-IL-6 antibody and the antitumor effect on glioma cells is enhanced by combination therapy using anti-PD-1 antibodies [108]. αPD-L1-LNP, an anti-PD-L1 antibody-conjugated lipid nanoparticle (LNP) containing a CDK inhibitor (dinaciclib), demonstrated a therapeutic effect in murine glioma models. It eliminated tumor-associated bone marrow cells (TAMCs) localized to the tumor microenvironment after RT by inducing apoptosis [109]. Owing to the strong immunosuppressive potential of MDSCs and their prognostic impact on GBM, the development of a combination therapy to inhibit MDSC infiltration is desirable.
Among immunosuppressive cells, M2Mϕs have a wide variety of functions that promote tumor growth and therefore cause strong immunosuppression when recruited into the tumor microenvironment at the early stage of tumor growth [110]. M2Mϕs enhance PD-L1 and IDO expression by antibody-dependent cellular phagocytosis [111]. It is expected that the effects of ICIs can be enhanced by combination with an M2Mϕ inhibitor. In GBM, B7-H4 expression on tumor cells and tumor-associated macrophages (TAMs) results in the maintenance of glioma progenitor cells and the formation of an immunosuppressive tumor microenvironment [29] while miRNA-21 contained in exosomes produced by TAMs increases the production of PDCD4, SOX2, STAT3, IL-6, and TGF-β1 in GBM cells that engenders TMZ resistance [112]. Conversely, exosomes derived from glioma stem cells (GSCs) promote M2 polarization and PD-L1 expression in Mϕs [113]. Hence, tumor cells and Mϕs mutually regulate the survival environment. Additionally, large amounts of Mϕs infiltrate GBM tissues and are derived from monocytes rather than microglia [114,115]. IDO expressed by TAMs and tumor cells metabolizes tryptophan to kynurenine, and kynurenine inhibits T-cell immunity while stimulating Mϕ aryl hydrocarbon receptor (AHR) expression. This AHR signal increases CCR2 and CD39, an ATP/ADP-degrading ectonucleotidase, expression on TAMs, resulting in increased CCL2-induced TAM recruitment into the tumor microenvironment and subsequent rise in the environmental adenosine concentration by the ectonucleotidase activity of CD39/CD73; all of which contributes to tumor progression by suppressing T-cell immunity [31]. CD73, an AMP-degrading ectonucleotidase that functions in cooperation with CD39, is highly expressed by TAMs infiltrating GBM and is expected to be a target for combination immunotherapy [116]. In addition to treatments aimed at inhibiting the function of Mϕs, treatments aimed at inhibiting Mϕs accumulation or M1 conversion have also been developed. However, a colony-stimulating factor-1 receptor (CSF-1R) inhibitor aimed at obstructing accumulation of Mϕs that did not alter the infiltration number of TAMs suppressed M2 function and improved prognosis in a murine glioma model [117]. In addition, combination therapy using a dendritic cell (DC) vaccine and an anti-PD-1 antibody with the CSF-1R inhibitor prolonged OS in a murine model via promotion of infiltration and activation of TIL [118] IPI-549, an inhibitor of phosphatidylinositol 3-kinase γ (PI3Kγ), an intracellular signal of M2Mϕs, suppresses M2Mϕs by selectively inhibiting PI3Kγ (IC 50 :16 nM) and inducing the PI3Kδ-dominant M1 phenotype [119]. Tumor growth inhibitory effects of IPI-549 have been confirmed in combination with anti-PD-1 antibodies in lung, breast, and head/neck cancer models [120]. We have also demonstrated the antitumor effect of combination therapy using anti-PD-L1 antibody and IPI-549 in a TMZ-resistant glioma-initiating murine model [121]. Furthermore, using human GBM tissue, we revealed an M2Mϕ infiltration increase during recurrence after intervention with immunotherapy as compared with the primary tumor [121]. Triple combination therapy using oncolytic herpes simplex viruses (oHSV, G47D) expressing murine IL-12 (G47D-mIL12) with anti-CTLA-4 and anti-PD-1 antibodies also induces Mϕ infiltration and M1-like phenotype polarization, contributing to the eradication rate in a murine model using GSCs via increasing the CD8 + /CD4 + FOXP3 + ratio [122].

By receiving extracellular vesicles produced by MDSC and M2Mϕs, B cells express inhibitory
ICMs such as PD-L1 or CD155, thereby forming an immunosuppressive tumor microenvironment as Bregs [53]. Tumor growth was significantly suppressed by using CD20 antibody or B-cell knockout in a glioma murine model in vivo [53]. Since extracellular vesicles derived from MDSC or M2Mϕs control the immunosuppressive ability of Bregs, a therapeutic strategy for targeting MDSC, M2Mϕs, or extracellular vesicles may be effective.

Ongoing Clinical Trials
A search of the NIH website (https://clinicaltrials.gov/) conducted on January 27, 2020 for ongoing immunotherapy clinical trials for GBM revealed 61 clinical trials. Excluding observational and non-interventional studies, 58 were intervention trials related to immunotherapy for GBM (Table 1). Registration status is 14 for "Active, not recruiting" and 2 for "Active, not recruiting / Has results", 2 for "Enrolling by invitation" and 39 for "Recruiting". Twenty-two of these trials were ICI-related and 11 were in combination with other immunotherapies such as adoptive immunotherapy including dendritic cell therapy, oncolytic virus therapy, IDO inhibitors, and vaccine therapy. Although there is some overlap with the abovementioned, there were 21 adoptive immunotherapies including dendritic cell therapy, TIL therapy, and CAR-T therapy; 11 oncolytic virus therapies; 4 IDO inhibitors; and 5 vaccine therapies.
Numerous phase I-II clinical trials of state-of-the-art treatments, such as a DC vaccine, ICI combination, chimeric antigen receptor (CAR)-T cell therapy, a cytomegalovirus (CMV) pp65 vaccine, and recombinant human IL-7-hybrid Fc NT-I7 (IL-7-hyFc, GX-I7), are ongoing. Phase III clinical trials included nivolumab and phase II/III clinical trials included DC therapy. Among these, immunosuppressive cells, including Tregs, MDSC, and M2Mϕs in the tumor microenvironment, are often the standard therapeutic target; however, a search of the UMIN website for clinical trials (https://upload.umin.ac.jp/) revealed that, in Japan, no active clinical trials related to immunotherapy for GBM are currently being conducted.

Future Perspectives and Conclusions
The most crucial point in the treatment of GBM is recurrence prevention after surgery and the development of immunotherapies to expand treatment options is welcomed. However, single immunotherapy is often insufficient in GBM and leads to formation of a 'cold tumor', since the immune microenvironment in GBM varies so greatly compared to other types of cancers. Analysis of the GBM microenvironment shows that patients with poor prognosis often have infiltration of immunosuppressive cells such as Tregs, MDSC, and M2Mϕs (Figure 3). A neoantigen vaccine using comprehensive gene analysis [123], CAR-T/NK cell therapy [124][125][126], and oncolytic virus therapy [122,127] are also considered promising. In addition to combinations of chemotherapies, radiotherapies, and ICIs already available for other malignant tumors [128], these ongoing or future therapies targeting the above-mentioned cells or immunosuppressive function will be forthcoming. It is also important to establish minimally invasive measurement methods to assay changes in immune status, such as liquid biopsy using the blood or cerebrospinal fluid of GBM patients [129].
In conclusion, we have here reviewed epidemiological features, tumor immune microenvironment, and associations between ICM expression and GBM prognosis. We have also reviewed the various ongoing and future immunotherapies for GBM. Various strategies, such as combinations of ICI therapies, will overcome these immunosuppressive mechanisms in the immune microenvironment of GBM.

Conflicts of Interest:
The authors have no conflicts of interest directly relevant to the content of this article. As indirect relevance, some of the materials for the AFTV described in this review article was provided by Cell-Medicine, Inc. (CMI), which is a venture company for research and development of immunotherapy born from RIKEN (The Institute of Physical and Chemical Research) and University of Tsukuba in Japan. T.M. is a member of CMI.