Genotoxic therapy and resistance mechanism in gliomas

Abstract

Glioma is one of the most common and lethal brain tumors. Surgical resection followed by radiotherapy plus chemotherapy is the current standard of care for patients with glioma. The existence of resistance to genotoxic therapy, as well as the nature of tumor heterogeneity greatly limits the efficacy of glioma therapy. DNA damage repair pathways play essential roles in many aspects of glioma biology such as cancer progression, therapy resistance, and tumor relapse. O6-methylguanine-DNA methyltransferase (MGMT) repairs the cytotoxic DNA lesion generated by temozolomide (TMZ), considered as the main mechanism of drug resistance. In addition, mismatch repair, base excision repair, and homologous recombination DNA repair also play pivotal roles in treatment resistance as well. Furthermore, cellular mechanisms, such as cancer stem cells, evasion from apoptosis, and metabolic reprogramming, also contribute to TMZ resistance in gliomas. Investigations over the past two decades have revealed comprehensive mechanisms of glioma therapy resistance, which has led to the development of novel therapeutic strategies and targeting molecules.

Introduction

Gliomas account for over 70% of primary brain tumors. Traditionally, gliomas are histologically classified into four grades. The World Health Organization (WHO) grades I and II are characterized as low-grade gliomas, while grades III and IV are high-grade gliomas. Several pioneering studies have shown that mutations in isocitrate dehydrogenase (IDH1/2) are highly frequent genetic abnormalities in glioma, identified in approximately 80% of all WHO grade II/III gliomas and secondary glioblastomas (Cohen et al., 2013). Glioblastoma (GBM, WHO grade IV) is the most common primary brain tumor, accounting for 54% of all gliomas and 14.5% of all brain tumors diagnosed in the United States (Ostrom et al., 2020). GBM is a highly aggressive disease with a poor median survival of approximately 12–15 months, and only 4.7% of patients survive longer than 5 years (Oike et al., 2013; Stupp et al., 2009). The 2016 revised WHO classification of tumors in the CNS combines biology-driven molecular marker diagnostics with classical histological cancer diagnosis. The new classification includes major restructuring of the diffuse gliomas, medulloblastomas, and other embryonal tumors and incorporates new entities that are defined by both histology and molecular features such as glioblastoma (IDH-wild-type and IDH-mutant types), and diffuse midline glioma (Louis et al., 2016).

The chemotherapy for gliomas includes TMZ and other alkylating agents such as carmustine (BCNU), nimustine (ACNU), and lomustine (CCNU) (Fukushima et al., 2009). The clinical benefit of TMZ in patients with gliomas remains limited, primarily due to intrinsic or acquired chemoresistance mechanisms (Sarkaria et al., 2008a). Genomic heterogeneity, a highly infiltrative nature, and distinct mechanisms also enable cancer cells to escape chemotherapy and radiotherapy (Annovazzi et al., 2017). As the standards of care for glioma are mostly based on introducing DNA damage to the cancer genome, the intrinsic DNA repair pathways play essential roles in protecting tumor cells and determining disease outcomes. The activation of DNA repair machinery has been extensively investigated. In the present review, we have summarized the current understanding of DNA repair pathways in glioma therapy resistance. We focused on DNA repair enzymes while including other resistance mechanisms such as programmed cell death, cancer metabolism, and glioma stem cells. We also provide an overview of novel molecular targeting strategies that assist genotoxic therapy in glioma.

DNA damage and repair

The integrity of genomic DNA is maintained by a balance between DNA damage and repair. Cells are constantly exposed to genotoxic factors such as oxidative damage, DNA strand breaks, and replication errors. In eukaryotic cells, many DNA repair mechanisms are established through evolution to limit the impact of DNA damage. Radiotherapy and chemotherapy are commonly used glioma therapies. They trigger the formation of both single- and double-strand breaks. The intrinsic DNA repair pathways are frequently exploited by glioma to handle the impact of DNA damage. Owing to the comprehensive anatomic structure of the blood-brain barrier (BBB), many anti-cancer therapeutics can hardly access intracranial malignancies; hence, genotoxic therapies remain the major treatment options. In the following section, we will discuss several DNA repair machineries that are related to glioma therapy resistance.

1. MGMT in glioma resistance to TMZ therapies

In 2005, Stupp et al. (Stupp et al., 2005) demonstrated that the addition of TMZ to radiotherapy resulted in clinical benefit for patients with newly diagnosed GBM with prolonged overall survival, which led to the approval of TMZ as part of the standard of care for glioma by the United States Food and Drug Administration (FDA). TMZ is almost completely absorbed through the gut, penetrates the BBB, and affects glioma cells. It achieves sufficient concentrations against tumors within the brain, with a concentration between 14.95 and 34.54 μM in brain tumors, about 20% of its plasma concentration (Kaina, 2019). TMZ does not require hepatic metabolism for activation, but is spontaneously hydrolyzed to 3-methyl-(triazen-1-yl) imidazole-4-carboxamide (MTIC), and then splits into monomethylhydrazine and 5-aminoimidazole-4-carboxamide (AIC). These drug metabolites lead to rapid methyl adducts on nucleosides. TMZ results in DNA methylation specifically at the N7 positions of guanine in guanine-rich regions, the N3 positions in adenine or guanine, and the O6 position in guanine. The ratio of these lesions are N7-methylguanine (N7-meG) (80%–85%), N3-methyladenine (N3-meA) or N3-methylguanine (N3-meG) (8%–20%), and O6-methylguanine (O6-meG) (8%). The formation of O6-meG is the major molecular mechanism for TMZ-associated toxicity, as O6-meG-associated nucleotide mispairs result in inhibition of DNA synthesis, and DSB formation (Zhang et al., 2012). Severe DNA damage subsequently activates the ATM/Chk2 signaling cascade and induces p53-associated G2/M cycle arrest (Caporali et al., 2004). Owing to DNA damage, tumor cells eventually undergo apoptosis, autophagy, or senescence (Aasland et al., 2019; Wurstle et al., 2017).

MGMT (O-6-Methylguanine-DNA Methyltransferase) is an enzyme that repairs naturally occurring O6-meG lesions, preventing DNA mismatch and replication errors. In glioma, MGMT plays a major role in the resistance to TMZ and other alkylating agents through the removal of the O6-meG lesion. MGMT repairs the O6-meG lesion in a “suicide” way by transferring the methyl group from the adduct to its cysteine residue, resulting in the degradation of the enzyme (Fig. 1). MGMT knockout mice exhibited higher sensitivity to the toxic effects of TMZ than MGMT wild-type counterparts (Glassner et al., 1999). Inhibition of MGMT with O6-benzylguanine promotes the anti-tumoral activity of TMZ both in vitro (Bobola et al., 1996) and in vivo (Friedman et al., 1995).

Figure 1. TMZ induced DNA damage repair.

TMZ is hydrolyzed to its active metabolite, 3-methyl-(triazen-1-yl) imidazole-4-carboxamide (MTIC), and then splits into monomethyl diazonium ions and 5-aminoimidazole-4-carboxamide (AIC). TMZ causes methyl adducts specifically at the N1-methyladenine (N1meA) and N3-methylcytosine (N3meC) (2%); N7-methylguanine (N7-meG) (80%–85%); N3-methyladenine (N3-meA) or N3-methylguanine (N3-meG) (8%), and O6-methylguanine (O6-meG) (8%). ALKBH2 and ALKBH3 remove N1meA and N3meC, accompanying with oxidative decarboxylation of 2-oxoglutarate (α-ketoglutarate, αKG) to succinate (Suc). MGMT is responsible for repairing O6-meG lesions. MGMT transfers the methyl group from O6-meG to its cysteine residue, resulting in the methylation and degradation of itself. If O6-meG adducts fail to be removed by MGMT, it will mispair with thymine. O6-meG-T mismatches are recognized by MMR components MSH2 and MSH6. MMR components MLH1 and PMS2 further removes the thymine residue and leave the O6-meG adduct unrepaired. The futile repair cycles lead to the accumulation of DNA strand breaks and eventually apoptosis. N7-meG and N3-meA adducts account for about 80% of TMZ-induced DNA adducts. BER system efficiently fixes these lesions. MPG is responsible for the detection and removal of aberrantly N7-meG and N3-meA. APE recognizes the abasic site, cleaves the 5’ end of the DNA, and triggers PARP1 to undergo patch repair. Patch repair can be completed by DNA polymerase, DNA ligase III, and XRCC1 (short patch) or FEN1 and DNA ligase I (long patch).

Owing to the essential role of MGMT in TMZ resistance, the expression level of MGMT is closely related to the efficacy of cancer therapies. Hypermethylation of the promoter region of the MGMT gene epigenetically silences the protein expression. Low expression of MGMT correlates with a better clinical response and prolonged survival (Hegi et al., 2005). The promoter methylation status determines the expression level of MGMT, which could be considered as a biomarker for TMZ responses in patients with glioma (Hermisson et al., 2006; Stupp et al., 2007). Indeed, a clinical data analysis showed that low MGMT methylation could confer some sensitivity to TMZ treatment, providing cut-offs that allow treatment decisions for personalized therapy (Villano et al., 2009).

Besides primary glioma, MGMT also plays critical roles during the disease recurrence. Higher expression of MGMT was found in clinical samples from recurrent GBM compared to matched primary lesions (Storey et al., 2019). The altered expression levels of MGMT in recurrent GBM could be resulted from changes in its promoter methylation status. Lower MGMT promoter methylation correlates with recurrent stage compared to primary tumors. Additionally, MGMT promoter is prone to establish unmethylated status when the cancers proceed to recurrence (Choi et al., 2021). On the other hand, the aberrantly high expression MGMT could also be induced by other factors and further contributes to GBM recurrence. A recent study reveals that MGMT genomic rearrangements lead to MGMT overexpression in recurrent gliomas while MGMT promoter methylation is unaltered (Oldrini et al., 2020).

Since the discovery of MGMT in cancer therapy resistance, molecular targeting strategies have been pursued to overcome MGMT-associated DNA repair. For example, lomeguatrib (a potent MGMT inhibitor) has shown promising efficacy and safety profile in sensitizing solid tumors for TMZ therapy in Phase I and II clinical trials (Middleton et al., 2002). Decitabine (DAC), an inhibitor of DNA methyltransferase that depletes MGMT, resulted in a 12.4-month median overall survival when combined with TMZ therapy in patients with metastatic melanoma, suggesting possible superiority over the historical 1-year overall survival (Tawbi et al., 2013). Moreover, O6-benzylguanine (O6 BG), a nucleotide analog of O6-meG that can be recognized by MGMT, inactivates MGMT enzyme by transferring a benzyl group to the active site of MGMT. O6 BG showed exciting anti-tumoral effects in preclinical studies, but the combination of O6 BG and TMZ in pediatric patients did not improve treatment response in clinical studies (Warren et al., 2005). A recent study by Rahman et al. (Rahman et al., 2019) showed that bortezomib sensitizes glioblastoma to TMZ treatment through the depletion of MGMT mRNA and protein with prolonged animal survival.

2. DNA damage responses in in glioma resistance to TMZ therapies

2.1. Deficient mismatch repair tolerates glioma to TMZ treatment

Mismatch repair (MMR), base excision repair (BER), and nucleotide excision repair (NER) are SSB repair mechanisms. The SSB repair mechanism recognizes and fixes the damaged DNA nucleotide by using the other DNA strand as a template (Jackson and Bartek, 2009). MMR and BER mechanisms have been extensively studied in glioma resistance.

MMR is a molecular mechanism that recognizes erroneous insertion, deletion, and mis-incorporations of nucleobases through DNA replication, recombination, and damage. MMR is a highly conserved mechanism that is governed by mismatch repair proteins MutL homolog 1 (MLH1), MutS homolog 2 (MSH2), MSH3, MSH6, and mismatch repair endonuclease PMS2 (PMS2). Alkylating agents such as TMZ result in methylated nucleosides, and mispairs between O6-meG with thymine or cytosine, causing O6-meG:T or O6-meG:C mismatch in the progeny DNA (Sarkaria et al., 2008b) (Fig. 1). Although MMR recognizes these mispairs and removes the thymine residue, the enzymes are not able to remove the O6-meG. Thus, the O6-meG persists in the genomic DNA, mispairs with thymine repeatedly, leading to repetitive rounds of MMR. It has been proposed that the repeating ‘futile repair’ cycle eventually leads to replication forks collapse, DNA double-strand breakages and cell death via apoptosis or autophagy (Villano et al., 2009). A recent study using CRISPR/Cas9 screening confirmed the role of the MMR system in TMZ therapeutic sensitivity (MacLeod et al., 2019). The results highlighted that three weeks after high doses of TMZ treatment in patient-derived glioblastoma stem cells, there is an enrichment for gRNAs targeting four genes involved in the MMR (MLH1, MSH2, MSH6, and PMS2) in resistant cells (MacLeod et al., 2019). It was also reported that a significant decrease in the MSH2, MSH6, and PMS2 protein expression levels was observed in recurrent GBM (Felsberg et al., 2011).

Tumors deficient in the MMR pathway are relatively resistant to genotoxic therapies. Karran et al. (Karran, 2001) proposed that MMR-deficient cells exhibited a 50-fold lower sensitivity towards methylating agents compared to their MMR-proficient counterparts. Cahill et al. (Cahill et al., 2007) reported that loss of the mismatch repair protein MSH6 in human glioblastomas is associated with tumor progression during TMZ treatment. In addition, McFaline-Figueroa et al. (McFaline-Figueroa et al., 2015) also showed that loss of MMR genes MSH2 and MSH6 provided substantial survival benefits to glioma cells. Moreover, mutations in the MMR pathway are specifically enriched in hypermutant glioma recurrence (Wang et al., 2016a). Mutations in MMR proteins may contribute to disease relapse and the hypermutated genotype through error-prone DNA replication, the disabled futile repair cycle, and tolerance of apoptotic changes.

2.2. BER promotes glioma resistance to genotoxic therapies

BER is responsible for correcting small base lesions such as nucleotides with methylation, oxidation, or deamination. At the beginning of BER, a specific DNA glycosylase recognizes the lesion and removes the affected base (Fig. 1). There are at least 11 distinct DNA glycosylases divided into four structurally distinct super families: uracil DNA glycosylases (UDGs), helix-hairpin-helix (HhH) glycosylases, 3-methylpurine glycosylase (MPG), and endonuclease VIII-like (NEIL) glycosylases (Jacobs and Schar, 2012). Apurinic/apyrimidinic endonuclease 1 (APE1) recognizes the abasic site and cleaves the 5’ end of the DNA. There are two repair pathways following cleavage. Short patch (one nucleotide) repair is conducted by DNA polymerase, DNA ligase III, and X-ray repair cross complementing 1 (XRCC1). Long-patch (2–10 nucleotide) repair requires flap structure-specific endonuclease 1 (FEN1) and DNA ligase I (Krokan and Bjoras, 2013).

Although N7-meG and N3-meA adducts account for about 80% of TMZ-induced DNA adducts (Wirtz et al., 2010), they rarely cause cytotoxic effects on cells, as the BER system efficiently fixes these lesions. Several lines of evidence have suggested that the inhibition of BER is synergistic with TMZ treatment. For example, MPG is responsible for the removal of aberrantly N7-meG and N3-meA in mammalian cells (Montaldo et al., 2019). Agnihotri et al. (Agnihotri et al., 2012) found a significant correlation between the expression levels of the MPG and the survival of adult GBM cells under TMZ treatment, indicating a supportive role of MPG in TMZ resistance. Moreover, methoxyamine, an inhibitor of MPG, sensitizes tumor cells to alkylating agents in vitro and in vivo (Agnihotri et al., 2014). It was also shown that the down-regulation of APE1 expression by small interfering RNA or nitroxoline affects the resistance of GBM to TMZ (Cho et al., 2019; Montaldi et al., 2015). Inhibition of DNA polymerase beta (Goellner et al., 2012; Trivedi et al., 2008) or endonuclease non-catalytic subunit 1 (ERCC1) also enhances TMZ efficacy in glioma treatment (Boccard et al., 2015). Goellner et al. (Goellner et al., 2011) found that poly(ADP-ribose) polymerase 1 (PARP1), which functions as a participator and sensor of incomplete BER, is hyperactivated when BER is inhibited. Aberrantly activated PARP1 continuously consumes NAD⁺ to synthesize poly (ADP-ribose) (PAR), leading to ATP depletion and energy-dependent cell apoptosis (Goellner et al., 2011).

PARP1 is thought to play a vital role in the BER pathway by interacting with BER proteins such as DNA ligase III, DNA polymerase beta, and XRCC1 (Dantzer et al., 2000; Heale et al., 2006). These proteins can be modified by PARylation through the action of PARP1. The PAR chain facilitates the recruitment of DNA repair complexes by providing a docking platform (Wei and Yu, 2016). PARP1 is highly expressed in GBM but not in normal brain tissues (Bao et al., 2006; Galia et al., 2012; Rasmussen et al., 2016), indicating that PARP1 is relevant to treatment resistance in glioma (Bao et al., 2006; Galia et al., 2012; Rasmussen et al., 2016). PARP inhibitors have been extensively studied in the past three decades and have shown promising results with increased TMZ sensitivity in GBM and ideal BBB penetrations. There are several PARP inhibitors in Phase I/II clinical trials on GBM patients, including niraparib (NCT04221503), veliparib (NCT03581292), and BGB-290 (NCT03749187).

2.3. Homologous repair in gliomas resistance to genotoxic therapy

Several lines of evidence suggest that TMZ resistance in recurrent glioma can be acquired without the up-regulation of MGMT or deficient MMR (Gil Del Alcazar et al., 2016; Happold et al., 2012). This indicates that there are alternative pathways for recurrent glioma tumors to handle genotoxic agents. Increased double-strand breaks (DSB) repair is one of the newly identified mechanisms contributing to acquire TMZ resistance in GBM.

TMZ-induced DNA lesions lead to the stall of replication forks and eventually MMR-dependent DSB (Fan et al., 2013). DSB is among the most severe and lethal types of DNA damage. DSB requires complicated repair mechanisms because of extensive damage and lack of a template to generate the new strand (Cannan and Pederson, 2016). There are two major repair mechanisms involved in DSB repair: homologous recombination (HR) and non-homologous end-joining (NHEJ). HR occurs when DNA strands are damaged during the S/G2 phase of the cell cycle. NHEJ exerts its role when cells lack a second copy of the DNA, and throughout most of the cell cycle, it predominantly acts in the G1 phase. ATM, ATR, DNA-PK, and KU70/80 kinases are the first proteins recruited by the DSB and play important roles in sensing the damage and initiating a downstream signaling cascade (Blackford and Jackson, 2017). The DNA damage response activates subsequent cell cycle checkpoints, which either repair damaged DSB or trigger cell death pathways.

HR DNA repair plays a vital role in glioma resistance genotoxic therapy (Fig. 2). An investigation of the Cancer Genome Atlas (TCGA) revealed that HR deficiency strongly correlated with progression-free interval and better clinical outcomes, indicating an adverse effect of HR DNA repair on GBM treatment (Knijnenburg et al., 2018). Studies have shown more efficient HR DNA repair in glioma-initiating cells than in non-tumor forming neural progenitor cells (Lim et al., 2012). Higher HR DNA repair efficiency was observed in glioma cells compared to neural progenitor cells. Carlos et al. (Buis et al., 2012; Esashi et al., 2005; Tomimatsu et al., 2014) reported that augmented HR DNA repair confers TMZ resistance in recurrent tumors. CDK1/2 facilitates the HR process by phosphorylating BRCA2 and exonuclease 1 (EXO1), or interacting with Double-strand break repair protein MRE11 and nibrin (NBS1). CDK1 and CDK2 are highly expressed in GBM and they induce radioresistance in GBM (Wang et al., 2016b; Zhang et al., 2018). Furthermore, CDK1/2 inhibitors suppress HR efficiency in recurrent glioma tumors and sensitize tumor cells to TMZ (Gil Del Alcazar et al., 2016).

Figure 2. TMZ induced double strands breaks addressed by Homologous Repair.

Unrepaired O6-meG adducts will lead to DSB, which is the most lethal DNA damage. DSB can be sensed by the MRE11-RAD50-NBS1 (MRN complexes), which set in motion a numbers of processes collectively called the DNA damage response to coordinate DNA repair, These sensor complexes recruit and activate kinases including ATM, ATR, DNA-PK and subsequently other modifying enzymes which, through cascades of phosphorylation and ubiquitination events, activate and mobilize a large number of proteins, such as p53, BRCA1, 53BP1 and H2AX. Homologous Repair requires resection by BRCA1, CtIP and BARD1 to generate single-stranded DNA (ssDNA), which immediately becomes coated by RPA and subsequently replaced by Rad51 preparing for strand invasion. Failure of Homologous Repair will result in cell death.

3. Aberrant cell cycle regulation in gliomas resistance to genotoxic therapy

High RNA expression levels of the DNA repair/cell-cycle genes including AURKA, CDC25C, CDC6, CDK1, CENPA, RAD51, RAD54, were revealed in relapsed glioma after radiotherapy alone or radiotherapy plus TMZ (Gobin et al., 2019). Dysregulated activity of cell cycle regulation proteins has been observed in human malignancies, and their aberrant activation is tightly associated with resistance to genotoxic therapy in cancer cells (Ferri et al., 2020; Huang and Zhou, 2020).

3.1. ATM/CHK2 pathway

The G1/S and intra-S checkpoints are controlled by the ATM/p53/CHK2 signaling cascade. ATM/p53/CHK2 induced cell cycle arrest and suppressed tumor formation in a glioma mouse model (Squatrito et al., 2010). Chk2^−/− tumor-bearing mice showed increased GBM resistance to genotoxic therapy with a lower apoptotic index and higher proliferation rate. Consistently, inhibition of checkpoint kinases or ATM sensitized glioma stem cells (GSCs) to ionizing radiation (IR) in vitro (Bao et al., 2006) (Carruthers et al., 2015). The ATM inhibitors AZ32 and AZD1390 result in promising results in glioma xenograft models (Durant et al., 2018; Karlin et al., 2018). They abolish ATM-dependent DNA damage response (DDR) pathway activity, thereby resensitizing radiotherapy-resistant glioma cells to G2/M checkpoint arrest and apoptosis. Both AZ32 and AZD1390 have been specifically designed for effective BBB penetration, granting their future clinical application. A Phase I study is currently conducted to assess the safety and tolerability of AZD1390 administered with radiotherapy in patients with brain cancer (NCT03423628).

3.2. ATR/CHK1 pathway

ATR acts primarily through CHK1 and controls the intra-S and G2/M checkpoints. ATR-CHK1 checkpoints were found to be upregulated in glioblastoma stem-like cells, suggesting a more effective DNA repair reaction. Eich et al. (Eich et al., 2013) have shown that ATR contributes to the resistance of GBM cells to TMZ in vitro. Knockdown ATR sensitizes GBM cells to TMZ, while ATR depletion has a significantly stronger effect. Knockdown of ATR, but not ATM, abolished phosphorylation of H2A.X, the DSB repair marker, and CHK1 and CHK2.

Ahmed et al. (Ahmed et al., 2015) showed that the ATR-CHK1 pathway is involved in the radioresistance of GSC by enhancing G2 checkpoint arrest. A combination of ATR-CHK1 inhibition and IR decreased the colony-forming ability of GSCs through catastrophic chromosome segregation during mitosis. Gil et al. (Gil del Alcazar et al., 2014) reported that the ATR inhibitor, NVP-BEZ235, potently inhibited the repair of IR-induced DNA damage in the glioma cell line U87vIII derived tumors. The attenuated DDR sensitizes glioma tumors and extends the survival of brain tumor-bearing mice (Gil del Alcazar et al., 2014). It has been reported that another BBB-penetrating ATR inhibitor AZD6738 enhances the therapeutic efficacy of cisplatin in non-small cell lung cancer (NSCLC) xenograft models, and AZD6738 can effectively penetrate the brain, indicating its potential use for treating brain tumors (Frosina et al., 2018; Vendetti et al., 2015). Currently, a Phase II clinical trial aimed at examining the efficacy of NVP-BEZ235 in glioblastoma treatment (NCT02430363). There are multiple clinical studies on AZD6738 in relapsed/refractory cancers. For example, AZD6738 in IDH1 and IDH2 mutated, refractory cholangiocarcinoma, or other solid malignant tumors are under investigation in an ongoing Phase II clinical trial (NCT03878095). Another Phase II trial of AZD6738 alone and in combination with olaparib aims to treat participants with renal cell carcinoma, urothelial carcinoma, all pancreatic cancers, or other solid tumors (NCT03682289). However, primary central nervous system tumors are rarely included in such studies.

4. Resistance of Glioma Stem Cells to genotoxic therapy

Cancer stem cells (CSCs) are a minority population in tumor cells characterized by self-renewal, proliferation, and differentiation (Clarke et al., 2006). These properties confer CSCs the ability to be relatively resistant to conventional chemotherapeutic agents. Similarly, glioma stem cells (GSCs) share many common characteristics with normal neural stem cells (NSCs) such as self-renewal, expression of molecular markers CD133, CD15, L1CAM, and SOX2, and the ability to differentiate into neurons, astrocytes, and oligodendrocytes (Vescovi et al., 2006). GSCs promote tumor heterogeneity and therapy resistance to genotoxic stresses and are thus a key therapeutic target (Prager et al., 2020).

The major players of DDR proteins (ATM, ATR, Chk1, Chk2, PARP1, and RAD51) are upregulated in GSCs (King et al., 2017; Ning et al., 2019). Aberrant G2-M cell-cycle checkpoint activation and increased DNA repair efficiency were found in GSCs. Inhibition of ATM, CHK1, ATR, or PARP abrogated G2-M checkpoint function and resulted in a profound radiosensitization of GSCs (Ahmed et al., 2015). Carruthers et al. (Carruthers et al., 2018) recently revealed high levels of DNA replication stress as a novel mechanism responsible for up-regulating DDR response and radiotherapy resistance in GSCs. Elevated levels of DNA replication stress evoked constitutive activation of the DNA damage response, characterized by increased phosphorylation of ATR, CHK1, and RPA32 in GSC-enriched cultures. Inhibition of the replication stress response with ATR and PARP inhibitors in GSCs reduces neurosphere formation and abrogates radiation resistance (Carruthers et al., 2018).

In addition, high metabolic demands are required to support maintenance of stemness, self-renewal, and proliferation in GSCs. GSCs also exhibit reprogrammed metabolism pathways such as glycolysis, oxidative phosphorylation, and glutaminolysis to support their cell activities more than their differentiated counterparts. For example, GSCs increase glucose uptake by co-opting the neuronal glucose transporter, Glut3 (Cosset et al., 2017; Flavahan et al., 2013). Differentiated tumor cells mainly rely on glycolysis, while GSCs consume less glucose and heavily depend on oxidative phosphorylation through IGF2BP2 regulation and FGFR3-TACC3 gene fusion mechanisms (Frattini et al., 2018; Janiszewska et al., 2012; Vlashi et al., 2011). GSCs can also acquire nutrients from glutamine and acetate as their bioenergetic and growth resources (Tardito et al., 2015). These additional metabolic pathways confer GSCs with an increased radiation resistance when challenged in hostile conditions.

5. Cell death regulation facilitates TMZ resistance in gliomas

Genotoxic agents efficiently clear tumor cells because the treatment-induced DNA damage is commonly translated into programmed cell death, such as apoptosis, senescence, autophagy and ferroptosis. Here we have briefly introduced several major regulated cell death pathways related to chemotherapy resistance. More details could be reached in the review articles which are focused on cell death and chemotherapies (Mohammad et al., 2015; Ricci and Zong, 2006; Sui et al., 2013).

Apoptosis is a central cell death program that can be induced by genotoxic therapies. Unrepaired DNA damage changes mitochondrial membrane permeabilization through activation of p53 and B cell lymphoma 2 (BCL-2) family proteins such as BAD, BAX, and BID (Fang et al., 2016).

p53 is stabilized and phosphorylated by ATM or ATR pathways in response to TMZ induced DNA damage (Caporali et al., 2004). Activated p53 induces expression of its major transcription target p21. p21 leads to G1 cell cycle arrest through its universal cyclin-dependent kinase inhibitor function (including CDK1 and CDK2) (Georgakilas et al., 2017). p16^INK4a, an inhibitor of CDK 4/6, prevents phosphorylation of pRb through inactivation of CDK4/6 complex, and thus stops cell cycle progression. The p53-p21 and p16^INK4a-Rb axis work together to pause G1 to S phase transition, allowing cells to repair DNA breaks or enter apoptosis process (Georgakilas et al., 2017). Up to 50% of GBM possess p53 inactivation mutations (Liu et al., 2020a). The p53 mutation or loss of functional p53 sensitizes glioma cells to TMZ and BCNU due to the unrepaired DNA breaks (Blough et al., 2011; Dinca et al., 2008; Xu et al., 2005). Similarly, about 78% of GBM tumors are deficient in p16^INK4a-Rb pathway (Cen et al., 2012), suggesting that genotoxic therapy could be effective for GBM therapy.

The BCL-2 family, which plays a key role in regulating apoptosis, balances between pro- and anti-apoptotic processes. The BCL-2 family is divided into three groups: 1) pro-apoptotic proteins such as BAX, BAK, and BCL-Xs; 2) anti-apoptotic proteins BCL-2, BCL-XL, and MCL-1; and 3) pro-apoptotic Bcl-2 homology domain 3 (BH3)-only proteins such as BAD, BID, BIK, and BIM. Numerous studies have revealed the roles of dysregulation of apoptosis pathways in chemoresistance in gliomas. BCL-2 and BCL-XL were highly expressed in GBM (Kouri et al., 2012; Nagane et al., 1998). A study has shown that there is a correlation between BH3-only pro-apoptotic protein expression and overall survival in GBM patients, and the BCL-2/BCL-XL pathway confers resistance in GBM cells against TMZ treatment (Cartron et al., 2012). Apoptotic GBM cells also promote therapy resistance by secreting apoptotic extracellular vesicles (apoEVs) enriched with various components of spliceosomes. Cell cycle gene RNA splicing can be altered in recipient cells, thereby promoting their resistance to radiotherapy or TMZ therapy (Pavlyukov et al., 2018).

Autophagy, senescence, and ferroptosis are several recently identified programmed cell death mechanisms shown to be relevant in glioma resistance under genotoxic stress. Cells segregate damaged cytoplasmic constituents into autophagosomes for lysosomal degradation through autophagy, serving as a pro-survival response to TMZ therapy. Chen et al. (Chen et al., 2015b) found that the autophagy-related protein VAMP8 is elevated in glioma tissues and silencing autophagy-related gene 5 (ATG5) or syntaxin 17 (STX17) can reverse TMZ resistance in VAMP8 overexpressed glioma cells. Filippi-Chiela et al. (Filippi-Chiela et al., 2015) reported that TMZ induces a transient induction of autophagy and cell resistance of the therapy through inhibition of the Akt-mTOR pathway. The inhibition of autophagy can largely increase the level of apoptosis following TMZ treatment (Knizhnik et al., 2013). Targeting autophagy with chloroquine, a classic inhibitor of autophagy, has now emerged as a promising adjuvant therapy for patients with GBM (NCT00486603, NCT02378532, NCT01430351).

Senescence is frequently induced in brain tumors when treated with TMZ, and this is attributed to clinically achievable low TMZ concentrations within the brain (14.95–34.54 μM in brain tumors). TMZ-induced senescence is triggered by the unrepaired DNA lesion O6-meG (Aasland et al., 2019). The massive “senescence-like dormant phase”, rather than efficient apoptosis, can partially explain the resistance of glioma cells, since the senescent cells can re-enter the normal cell cycle when genotoxic stress is withdrawn (Beausejour et al., 2003).

Ferroptosis is an iron-dependent cell death process that is relevant to TMZ-induced cytotoxicity (Xie et al., 2016). The accumulation of reactive oxygen species, lipid peroxidation, and suppression of glutathione peroxidase is regarded as a characteristic of ferroptosis. NADPH oxidase (NOX) and p53 act as positive regulators of ferroptosis by increasing ROS levels and inhibiting the expression of xCT (SLC7A11), a plasma membrane transporter that facilitates glutathione biosynthesis. Glutathione peroxidase 4 (GPX4), heat shock protein beta-1 (HSPB1), and nuclear factor erythroid 2-related factor 2 (NRF2) function as negative regulators of ferroptosis by limiting cellular iron uptake and reducing ROS production (Xie et al., 2016). Zhao et al. (Zhao et al., 2017) reported a strong correlation between high expression of GPX4 and poor prognosis of glioma patients. Downregulation of GPX4 by ALZ003, a curcumin analog, enhances sensitivity to chemotherapy through ferroptosis in TMZ-resistant glioblastoma cells (Chen et al., 2020b). SLC7A11 contributes to TMZ resistance in glioma by repressing ferroptosis. Overexpression of SLC7A11 in tumor cells inhibits ROS-induced ferroptosis, which is induced by salicylic acid sulfapyridine (SAS) inhibition of SLC7A11 activity (Dixon et al., 2012; Jiang et al., 2015). A study showed that the ferroptosis-inducing agent erastin can synlethalize xCT-depleted glioma cells, suggesting that SLC7A11 contributes to TMZ glioma resistance to TMZ by suppressing ferroptosis (Sehm et al., 2016).

6. Cancer metabolism in genotoxic resistance

Gliomas are characterized by altered metabolism. Rewriting cellular metabolism facilitates glioma cell survival, growth, and resistance to therapies. Abnormal cancer metabolism includes antioxidant regeneration, macromolecule synthesis, and NAD metabolism.

Elevated ROS levels are one of the mechanisms of TMZ-mediated cytotoxicity. Glutathione (GSH) and GSH-related enzymes are among the most important antioxidant components that protect cells from toxicity due to chemotherapy. Cysteine is the rate-limiting substrate for GSH synthesis, and its uptake is regulated by xCT (encoded by SLC7A11 gene). GSH levels are highly correlated with drug resistance in gliomas. Zhu et al. (Zhu et al., 2018) reported higher levels of GSH and glutathione reductase in TMZ-resistant glioma cells than in sensitive cells. Glutathione depletion by the glutathione inhibitor, buthionine sulfoximine sensitized TMZ-resistant glioma cells to TMZ sensitive, both in vitro and in vivo (Rocha et al., 2014). It was shown that xCT is highly expressed in glioma cells (Polewski et al., 2016). Suppression of xCT increased ROS accumulation and compromised glutathione generation, leading to sensitivity to oxidative and genotoxic TMZ stress (Chen et al., 2015a). NRF2 is a transcriptional factor that regulates antioxidant-related genes and cellular redox (Liu et al., 2020c; Yu et al., 2020). High expression of NRF2 induced by TMZ also confers TMZ resistance to gliomas, and inhibition of NRF2 resensitizes glioma cells to TMZ (Cai et al., 2019; Liu et al., 2020b; Rocha et al., 2016; Zhang and Wang, 2017).

Glutamate can be catalyzed into α-ketoglutarate (αKG) and contributes to redox homeostasis with antioxidant GSH production (Wang et al., 2010). Glutamine is converted to an ammonium ion and glutamate by glutaminase (GLS). GLS has been proposed to mediate resistance to GBM therapy. Elevated GLS and glutamate levels were found to promote resistance of GBM cells under mTOR inhibition (Tanaka et al., 2015). Gu et al. (Gu et al., 2017) found that mTORC2 protects glioma cells from cellular stress by repressing cystine uptake and glutathione synthesis in gliomas via inhibition of xCT. The diversity of metabolic adaptations contributes to therapeutic resistance and promotes survival or proliferation. Indeed, mTORC2 inhibition can effectively repress TMZ-resistant gliomas (Miyata et al., 2017).

NAD is an essential co-factor in DDR through PARP DNA repair. PARP1 recruits downstream DDR proteins by synthesizing poly ADP-ribose (PAR) chains, requiring NAD as a substrate (Zhou and Wahl, 2019). PARP inhibitors compete with NAD for the catalytic pocket of PARPs, increasing the radio-sensitivity of gliomas (Hurtado-Bages et al., 2020). NAMPT, the NAD biosynthetic enzyme, is highly expressed in glioblastoma tumors and is associated with overall poor survival in such patients (Gujar et al., 2016). Treatment with NAMPT inhibitors sensitizes glioblastoma cells to TMZ (Feng et al., 2016; Sampath et al., 2015). Aberrantly activated PARP1 by BER inhibition continuously consume NAD to synthesize PAR chains. Overconsumption of NAD leads to ATP depletion and energy-dependent cell apoptosis (Goellner et al., 2011).

7. IDH mutation associated TMZ sensitivity

Isocitrate dehydrogenase enzymes (IDH) are essential enzymes that participate in several major metabolic processes such as the Krebs cycle, glutamine metabolism, lipogenesis, and redox regulation (Tang et al., 2020). Approximately 80% of WHO grade II/III gliomas are IDH mutations (Han et al., 2020; Liu et al., 2019). IDH mutations cluster in the active sites of the enzymes (R132 for IDH1, R140, or R172 for IDH2) (Cohen et al., 2013). Mutant IDH heterodimers catalyze α-ketoglutarate (α-KG) into D-2-hydroxyglutarate (D-2-HG). D-2-HG disrupts the demethylation process in histones and DNA by inhibiting the activities of Lysine-specific demethylase 4 (KDM4) or Ten-eleven translocation methylcytosine dioxygenase (TET). IDH mutations predict better disease outcomes and greater sensitivity to chemotherapy in low-grade gliomas and secondary glioblastomas (Minniti et al., 2014).

Recent investigations have revealed functional changes in DNA repair pathways in the context of IDH mutants and D-2-HG. For example, D-2-HG leads to a genome-wide hypermethylation phenotype, which may trigger MGMT promoter methylation, thereby sensitizing glioma cells to radiotherapy and chemotherapy (Minniti et al., 2014). In addition, D-2-HG inhibits ALKBH2 and ALKBH3, the enzymes mainly responsible for the removal of N1-meA and N3-meC induced by TMZ. Deletion of mutant IDH allele or overexpression of ALKBH2 or AKLBH3 would rescue D-2-HG-induced DNA damage and sensitization to alkylating agents (Wang et al., 2015). The clinical relevance of these findings is supported by a previous report that ALKBH2 confers resistance to TMZ in GBM cells (Johannessen et al., 2013). Moreover, a recent investigation showed that D-2-HG compromises homologous repair through epigenetic regulation. Inhibition of KDM4A and KDM4B by D-2-HG blocks DNA repair protein recruitment to sites of DNA damage (Sulkowski et al., 2017). Inhibition of KDM4A induces hypermethylated H3K9me3 and the widespread hypermethylation of H3K9 masks the specific local H3K9me3, thus blocking the recruitment of factors required for the homologous repair process (Sulkowski et al., 2020). Consequently, impaired homologous repair increases IDH mutant glioma cell sensitivity to genotoxic therapies. Overall, glioma with the IDH mutant appears to exhibit a distinctive biological pattern. More effort is urgently needed to identify the unique therapy response and resistance mechanisms in this disease cluster.

8. Epigenetic factors in glioma pathogenesis and treatment

Epigenetic regulations are tightly linked to glioma pathogenesis. The unique epigenetic patterns in glioma, such as DNA/histone methylation and chromatin remodeling, have been studied for novel molecular targeting strategies to reverse glioma-associated epigenetic shifts. The epigenetics in glioma has been covered by several elegant review articles recently (Cheng et al., 2019; Dong and Cui, 2019; Gusyatiner and Hegi, 2018). Here we highlight several known mechanisms that glioma-associated epigenetic shift that affects genotoxic therapy.

Glioma-CpG Island Methylator Phenotype (G-CIMP) is a glioma-specific DNA methylation pattern (Ruiz-Rodado et al., 2020). G-CIMP is featured with genome-wide hypermethylated CpG islands. G-CIMP mostly occurs in IDH-mutated gliomas. IDH mutation is sufficient to establish the unique DNA methylation pattern in gliomas and most IDH-mutant gliomas are characterized with G-CIMP (M and Wojtas, 2019; Turcan et al., 2012). One of the mechanism is IDH mutant gliomas produce oncometablites such as 2-hydroxyglutarate (2-HG) and these oncometablites represses TETs-mediated DNA demethylation process (Xu et al., 2011). Hypermethylated CpG islands play important roles in GBM development and could be a prognostic factor for treatment. Most G-CIMP+ patients exhibit a better prognosis compared to G-CIMP− patients (Malta et al., 2018). As previously described, the hypermethylation status of MGMT promoter has been extensively studied in glioma. MGMT methylation is related to a better prognosis of the tumor. Tumor suppressors, such as p16^INK4a, p14ARF and RUNX3 are also inactivated due to DNA methylation and thus promote GBM tumor development (Kanu et al., 2009).

Since DNA methylation causes tumor suppressor inactivation, the DNA Methyltransferases (DNMTs) are targeted for treatment of gliomas. The DNMT inhibitors 5-Azacitidine (5-Aza-CR) and 5-aza-2′-deoxycytidine (5-Aza-CdR) are two promising drugs targeting DNMTs (Palii et al., 2008). They incorporate into DNA during replication as analogues of cytosine. Their binding with DNMTs will reduce soluble DNMT levels and reverse DNA methylation status. Currently there are several ongoing clinical trials using these two DNMT inhibitors to treat gliomas (NCT03572530, NCT02332889, and NCT02940483).

Histone modifications mediated by histone deacetylases and histone demethylases are dysregulated in Glioma. Several histone deacetylases such as HDAC1, HDAC3, HDAC5 and HDAC9 (Staberg et al., 2017; Yang et al., 2015) have been revealed to be overexpressed in adult GBM. The high expression of HDACs is indicated to be correlated with a poor patient prognosis (Lee et al., 2017; Wang et al., 2017). Inhibition of HDAC has shown effective anti-cancer effects. Several clinical studies are currently ongoing to evaluate the therapeutic value of HDAC inhibitors including Valproic acid (NCT01817751 and NCT03048084), and Entinostat (NCT03048084) (Chen et al., 2020a). Histone demethylases are also shown abnormal expression in gliomas and could contribute to tumor development. The histone demethylase KDM1 (Amente et al., 2015), KDM2B (Staberg et al., 2018), JMJD2A (Li et al., 2020) is overexpressed in glioma samples. The high expression of these Histone demethylases promotes the proliferation of glioma cells and indicating shorter overall survival.

Interestingly, the oncometablite D-2HG produced by IDH mutation have been reported to repress several histone demethylases including KDM4A, KDM7A, and KDM2B (Liu et al., 2020a). KDM4A repression by IDH mutation leads to impaired DNA damage response to genotoxtic agents (Sulkowski et al., 2020). The histone demethylase inhibition effects can partially explain glioma patients with IDH mutation have a better prognosis.

Chromatin remodeling factors, such as lymphoid-specific helicase (LSH) and E2F1 are also found up-regulated in GBM and promote the development of GBM tumors (Xiao et al., 2017). SWI/SNF complex is reported involved in the maintenance of stemness in glioma cells (Hiramatsu et al., 2017).

9. Conclusions

The development of new treatments to improve the outcomes of patients with glioma has been heavily pursued in recent decades. TMZ has been used as the standard chemotherapy for glioma for over a decade. However, its efficacy is limited by a series of resistance mechanisms. Understanding these mechanisms, developing novel modalities of therapy, and exploring new combinations of TMZ to overcome these resistances is an important topic in glioma therapy.

The mechanisms of glioma drug resistance to TMZ are highly comprehensive and require further investigation. The different resistance mechanisms such as DNA damage repair pathways, heterogeneity of cancer cell populations, newly discovered death mode, and reprogrammed cancer metabolism indicate the need for combinatorial therapies targeting multiple signaling pathways. In addition, targeting altered metabolism might be a promising strategy for this challenging disease. Mutated IDH and its metabolic product, D-2-HG, lead to a distinctive pattern of the Krebs cycle, redox homeostasis, energy metabolism, and resistance mechanism, suggesting that selective therapy could be possible for this subtype of glioma.

List of nonstandard abbreviations used in the text

BCL-2

B cell lymphoma 2

BER

base excision repair

D-2-HG

D-2-hydroxyglutarate

DDR

DNA damage response

DSB

double-strand breaks

GBM

Glioblastoma

GSCs

glioma stem cells

GSH

Glutathione

HR

Homologous repair

IDH

Isocitrate dehydrogenase enzymes

MGMT

O6-methylguanine-DNA methyltransferase

MMR

Mismatch repair

N3-meA

N3-methyladenine

N3-meG

N3-methylguanine

N7-meG

N7-methylguanine

NHEJ

non-homologous end-joining

NRF2

nuclear factor erythroid 2-related factor 2

O6-meG

O6-methylguanine

PARP1

poly(ADP-ribose) polymerase 1

SLC7A11

Solute Carrier Family 7 Member 11

SSB

single-strand breaks

TMZ

temozolomide

α-KG

α-ketoglutarate