A. Trends in Drug Targets

Many drugs have a defined set of drugged proteins that are important during a broad overview of the drug discovery process. During our analysis, we identified 581 unique drug targets, including protein isoforms. The term target cannot be applied to all entities in our database. For example, the term “target” seems to be inappropriate for oncolytic viruses or hematopoietic stem cells. Thus, the number of agents used to gather targets was smaller than those of the entire dataset and equals 518 drug entities. In our analysis, we did not include drug targets whose pharmacological action designation was “NO” according to the drug bank since there is a high confidence that such a target is not a basis for pharmacological action. However, it must be mentioned that anticancer agents may exhibit their activity not through an initially proposed target (Lin et al., 2019). Thus, we decided not to remove other targets.

The locations of most targets (40.4%) correspond to cellular membrane, which is followed by cytoplasm (26.9%) and nucleus/nuclear envelope (15.7%) (Fig. 5A). The percentage of membrane-localized targets aligns well to the previous analysis undertaken by Yildirim et al. (2007). Other locations are less frequent and account for 17% of the drug-target spectrum combined. Such results are not surprising, considering that many targeted therapies for cancer act on tyrosine kinases that either are receptors and are located on the cellular membrane or represent a cytosolic protein that could be found both in the cytoplasm or nucleus, depending on a particular enzyme. Many other proteins, including enzymes, such as histone deacetylases (HDACs) or poly (ADP-ribose) polymerase (PARP), also demonstrate similar localization and could be found either in the cytoplasm or nucleus.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5

Targets and their distribution. (A) The distribution of target locations. Each target location was manually obtained from the UniProt database and PubMed search. If a target had multiple cellular localizations, we included the most relevant for its functional activity. If the exact function is unknown or hard to determine, we used the location with the largest number of references. Color corresponds to a target location. (B) The general functional activities of the drug targets. Each drug target was manually classified using the UniProt database and PubMed searches. Targets are color-coded corresponding to seven major functional classes. (C) Specific descriptions of each class of drug target presented in part B. Transporters were classified using the Transporter Classification Database. The term “kinase” refers to all kinases, excluding RTK and nontyrosine receptor kinases. Struct., structural; Synth., synthesis.

Using structural and functional properties to classify the targets shows that enzymes are the most abundant target class with almost 41.8% of all targets, and receptors are the second most common class and account for ∼31.5% (Fig. 5B). More specific characterization of the drug targets shows that kinases are the most targeted direction among enzymes as well as all targets (Fig. 5C). The second most commonly targeted enzyme class is represented by metabolic enzymes. Receptors are the third most investigated targets that include receptor tyrosine kinases (RTKs), GPCRs, and immune-related receptors. Interestingly, in comparison with enzymes and receptors, the pools of ligands, transporters, and transcription factors are relatively small.

It should be specified that the abovementioned figures represent a qualitative perspective of the target spectrum in our analysis. We visualized trends for each target class (see additional supplementary materials). Targets demonstrated trends similar to their corresponding drug classes. Interestingly, isoforms of HDACs and vascular endothelial growth factor A (VEGFA) demonstrated steady downward trends, whereas programmed cell death 1 (PD1 or PDCD1) and cytotoxic T-lymphocyte–associated protein 4 (CTLA-4) that are the targets for checkpoint inhibitors showed very noticeable up-trending behavior. Several kinases, including deoxycytidine kinase, BRAF, and Lim kinase also showed increasing presence during the years. Many targets showed no trends and appeared in a very limited number of trials.

B. The Basics, Chemotherapy

Chemotherapy is a staple of cancer treatment, and its primary action is based on cytotoxic effects resulting in cell death. Antimetabolites are the chemotherapeutic agents wherein the primary action hinges on substituting the cellular substrates with the drug, thus resulting in the inhibition of the cellular growth and cell cycle arrest. To date, no antimetabolites have been approved for brain cancer by the FDA. Since 2010, the presence of antimetabolites applied alone or in combination has increased in clinical trials, peaking in 2016 and 2018 with eight trials in each year, respectively. The most common drugs include azacitidine, fludarabine, gemcitabine, and methotrexate. Interestingly, azacitidine has been discovered to possess hypomethylating properties and has been investigated in brain tumors, including isocitrate dehydrogenase (IDH) 1/2 mutated specimens (Federici et al., 2020). The important feature of IDH1 is that it is a prognostic factor of patient survival in gliomas (Sanson et al., 2009), and two of seven clinical trials with azacitidine have the IDH status specified in the inclusion criteria (NCT03684811, NCT03666559). Decitabine is another drug that demonstrated a hypomethylating activity similar to azacitidine in hematologic tumors (Derissen et al., 2013). In brain cancer, this agent has been investigated twice—either in combination with dendritic cell (DC) vaccine (NCT02332889) or with cedazuridine (NCT03922555). The IDH status was noted in the inclusion criteria of the latter trial. The rationale of using decitabine with the DC vaccine is the property of this drug to enhance the expression of NY-ESO-1 antigen in gliomas (Konkankit et al., 2011). This trial, however, was discontinued due to enrolling only one participant (NCT02332889).

Another antimetabolite, fludarabine, was examined 15 times in 13 clinical trials on brain tumors since 2010 and always in combination therapies. It was used in combination with cellular therapies like CAR T cells and T-cell receptor (TCR)-modified T cells; it also appeared frequently with cyclophosphamide, an alkylating agent. A combination of fludarabine with cyclophosphamide represents a lymphodepletion regimen, and it is believed to favor the adoptive T-cell therapy (Salem and Cole, 2010). For instance, the addition of fludarabine to the lymphodepletion chemotherapy improved anti–CD-19-CAR T-cell persistence and survival of patients with blood cancer. Particularly, 16 of 17 patients who received cyclophosphamide with fludarabine and anti–CD-19-CAR T cells demonstrated a complete response (Turtle et al., 2016). Apparently, the high presence of fludarabine is explained by a growing interest in cellular therapies, in particular, CAR T cells and TCR T cells. Another agent, gemcitabine, has been analyzed in six different clinical trials as a part of combination therapies with kinase inhibitors, filgrastim, PARP inhibitors, or chemotherapy. Methotrexate is an antifolate drug that appeared in eight different clinical trials on brain tumors. Mostly, it has been combined with other chemotherapeutic medications.

Alkylating agents, another branch of chemotherapy, represent one of a few drug classes that embed the approved drug for brain cancer therapy. These drugs provide cellular cytotoxic effect via crosslinking DNA molecules, causing double-strand breaks, hampering uncoiling, and leading to the apoptosis of a cell. Sixteen drugs have entered clinical trials since 2010 as part of the alkylating agents class, and three of these medicines—temozolomide, cyclophosphamide, and lomustine—were the most frequently applied. In the United States, temozolomide was authorized for use in patients with brain cancer for the first time in 1999, and in 2005, a combination of temozolomide with radiation therapy was approved for newly diagnosed glioblastoma multiforme (https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=overview.process&ApplNo=021029). According to our estimations, temozolomide is the most abundant drug in the whole field and has been tested in 275 different clinical trials in the last 10 years. This alkylating agent has been examined in 209 unique combinations even without taking into account various types of radiation therapy. In the last 10 years, the most common combinations of this drug were with radiation therapy and bevacizumab. Cyclophosphamide is the second most investigated alkylating agent that has been examined in 39 different clinical trials in 48 unique combinations. Before 2016, it was used as a part of combinational chemotherapy regimens to treat brain tumors. Then, with the expanding popularity of cellular therapies, it started to appear in combination with CAR T cells and TCR T cells as a primary component of the lymphodepletion regimen with or without fludarabine.

The third most common alkylating agent identified in our database is lomustine (or CCNU). It has been examined in 21 unique combinations in 24 trials since 2010. Compared with cyclophosphamide, the high presence of this drug is not influenced by the growing popularity of cellular products. In fact, we found no trials investigating the concomitant application of any kind of cell product with lomustine in patients with brain cancer. Initially, lomustine was approved by the FDA for brain tumors and Hodgkin disease in 1976 (https://www.accessdata.fda.gov/drugsatfda_docs/label/2016/017588s043lbl.pdf). This agent is used as a part of a prominent combinational chemotherapy regimen, which is called procarbazine, lomustine (CCNU) , and vincristine (PCV), with procarbazine (another alkylating agent) and vincristine. According to several clinical trials, lomustine demonstrates better results when it is a part of the PCV regimen, especially if used in patients with 1p/19q-codeleted oligodendroglioma (Weller and Le Rhun, 2020). In our data, however, we detected the PCV regimen only in three trials. A phase three POLCA trial started in 2015 is currently investigating PCV in anaplastic gliomas with 1p/19q codeletion (NCT02444000). We also would like to note another alkylating agent, carmustine. Injectable carmustine was approved in the United States for brain tumors in 1977 (https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/017422s055lbl.pdf). Interestingly, we did not identify studies with injectable form of the drug during the analyzed time span. The interesting feature of carmustine is that this drug has been examined as a carmustine sustained-release implant wafer (CASANT wafer) and as a GLIADEL wafer in the last 10 years in five different clinical trials. The purpose of these two implants is to provide prolonged release of carmustine directly into a patient’s surgical cavity after tumor resection. FDA authorized GLIADEL for use in gliomas in adjunction to surgery or radiation therapy in 1996 (https://www.accessdata.fda.gov/drugsatfda_docs/label/2018/020637s029lbl.pdf). Interestingly, we did not find studies with carmustine after 2014, and of five studies presented in the dataset, only one has been completed. In this study, the GLIADEL administration after 5-aminolevulinic acid–supported tumor resection resulted in 15 months of median overall survival of patients with glioblastoma, albeit not without relatively frequent (44%) serious adverse effects (NCT01310868).

Another prominent class of chemotherapy agents is topoisomerase inhibitors. Topoisomerases are vital for enzymes in human cells that regulate DNA supercoiling, thus providing cellular homeostasis during transcription and DNA replication. Generally, there are two types of topoisomerases, type one and type two, which provide DNA uncoiling in a slightly different manner. Both of the two topoisomerase classes are targeted by topoisomerase inhibitors (Pommier, 2013). Seven drugs that could be classified as topoisomerase inhibitors were identified in brain cancer clinical trials since 2010. Irinotecan was the most investigated topoisomerase inhibitor in our database. This is a prodrug, the active metabolite of which, SN-38, acts via inhibiting type one topoisomerases. Irinotecan has been analyzed in 18 combinations in 19 different clinical trials, and the most common combinations of the compound were with temozolomide, kinase inhibitors, and bevacizumab. Additionally, we identified five clinical trials that investigated liposomal and nanoparticle formulations of irinotecan (NCT02022644, NCT02433392, NCT02481960, NCT03086616, NCT03119064). Etoposide is the second most investigated topoisomerase inhibitor that appeared in 26 unique combinations in 18 different clinical trials. This drug acts via inhibiting type two topoisomerases. Interestingly, in comparison with irinotecan, etoposide was combined mostly with chemotherapies (carboplatin, cisplatin, cyclophosphamide) and kinase inhibitors but only appeared once with bevacizumab. Topotecan is the last drug that was investigated relatively frequently in nine trials. Just like irinotecan, topotecan metabolizes into its active metabolite SN-38. Topotecan was applied either as monotherapy or in combination with pazopanib (angiogenesis inhibitor), alisertib (kinase inhibitor), and other agents.

C. Gene Therapy Is Found in Six Classes

Gene therapies are frequently investigated in patients with cancer. We identified six classes of gene therapy in brain tumors: modification of immune effector cells, chemosensitization of tumor cells, immune system recruitment to tumor cells, DNA vaccines, chemoprotection of normal cells, normal gene augmentation, and toxic gene expression in tumor cells. Two of these classes, effector cell modification and DNA vaccines, in turn demonstrated a growing interest from investigators in the last 10 years. Effector cell modification is an introduction of an artificial gene or gene editing of a natural gene in immune effector cells to facilitate the immune response against the tumor. This class includes modifications of immune effector cells, such as CAR T cells, TCR T cells, CAR NK cells, and others.

CAR T cells, along with checkpoint inhibitors, are probably the major breakthroughs in current cancer therapies. Currently, CD19 CAR therapy demonstrated outstanding clinical performance in patients with acute lymphoid leukemia (Sadelain et al., 2017). In one study, the complete response rate in 29 participants was very high at 93% (Turtle et al., 2016). Such impressive results led to the initiation of similar clinical trials in other cancers. Only a single phase one trial investigated CAR T cells in patients with brain cancer in 2010. In contrast, seven clinical trials on CAR T cells have been initiated in 2019. The main idea behind CAR T-cell therapy is an introduction of CAR into a patient’s T cells, propagation of cells in vitro, and subsequent infusion of these cells into a patient to eradicate the tumor. CAR structure includes a binding domain [usually single-chain variable fragment (ScFv)] to recognize a tumor antigen, hinge and transmembrane regions, and signaling domains. CAR structure could vary in terms of signaling domains, and hence there are three CAR generations. Basically, CAR T-cell design, manufacturing, and application are complex and involve rigorous considerations of the binding domain, target selection, signaling domain selection, transmembrane region optimization, gene delivery vehicle, population of T cells, cultivation, and safety. Discussing these factors is beyond the scope of this publication; to see more detailed information on CAR T cells see these comprehensive reviews (Jackson et al., 2016; Sadelain et al., 2017; Rafiq et al., 2020). In brain tumors, we identified several targets for CAR T-cell therapies: mutant EGFRvIII, Cluster of Differentiation 276 (B7-H3), prominin-1 (CD133), EPHA2, ganglioside G2 (GD2), human epidermal growth factor receptor 2 (HER2), mucin 1, and interleukin-13 receptor subunit α-2 (IL13RA2). We also identified single trials investigating CAR T cells against basigin (CD147), EGFR (wild-type), EGFR806, and PD1. In two trials, CARs were targeted using either the chlorotoxin peptide (NCT04214392) or a fragment of the NKG2D receptor from NK cells (NCT04270461). An interesting CAR design that entered clinical trials involved using PD1 as a binding domain to switch the immune inhibitory signal from programmed death-ligand 1 (PD-L1) into activating signaling (NCT02937844). All of the aforementioned antigens have demonstrated their role in brain cancer development and progression. For instance, EGFRvIII, a mutated version of the epidermal growth factor receptor, is the most common oncogenic mutation found in glioblastoma multiform (An et al., 2018). The important and not yet resolved issue is whether the targets listed are suitable for CAR T cells in the clinical settings. In general, we identified 34 different clinical trials investigating CAR cells, and all of them were phase one trials. CAR cells were used 12 times as combinations and 28 times as monotherapies. Most of the trials with CARs did not have a lymphodepletion regimen or did not specify it in the trial description. Several CAR cell designs were derived from memory-enriched T-cell populations since this approach is believed to provide better antitumor responses (Sadelain et al., 2017).

TCR T cells represent an alternative to CAR cells. This approach relies on transferring the specific T-cell receptor gene into recipient T cells to provide an immune response against a desirable antigen. This method requires smaller amounts of antigen on tumor cells for T-cell activation; however, antigens for this therapeutic class are MHC-restricted and not always amenable for targeting. But this is contrasted by the fact that TCR engineering allows targeting intracellular proteins, of which fragments can be presented on MHC molecules (Chandran and Klebanoff, 2019). Unfortunately, TCR T cells failed to demonstrate a more favorable safety profile, as one might expect, over CAR T cells in solid tumors, given that severe toxicities were reported from several clinical trials on different tumor types (D'Ippolito et al., 2019). Currently, only six clinical trials are investigating TCR T cells in brain tumors, with three of these being phase two. In contrast to CAR T cells, TCR T cells were found only in combinations. We detected a lymphodepletion chemotherapy in six out of seven unique combinations involving TCR T cells. CAR NK cells are even rarer in clinical settings. In this approach, the CAR gene is introduced in the genome of the patient’s NK cells instead of T cells. We found only two phase one trials analyzing CAR NK cells in patients with brain tumor. The first trial has an unknown status and investigated anti–mucin 1 CAR NK cells in gliomas and other cancers (NCT02839954). Another trial analyzes the intracranial administration of NK-92/5.28.z cells in HER2 positive glioblastoma. These cells represent an allogeneic NK cell population transduced with the anti–HER2-CD28 CAR gene (NCT03383978). In comparison, in other cancers (especially in hematologic malignancies), CAR NK cells seem to be more studied compared with brain tumors (Wang et al., 2020).

To date, CAR T cells and TCR T cells are one of the fastest-growing areas of in-brain cancer field. However, there are still many concerns as to whether these approaches will work out. Several strategies have been proposed to overcome challenges with CAR T cells in solid tumors (Rafiq et al., 2020). However, only a few of them are currently presented in clinical trials with brain tumors. Combination approaches might be helpful, but their diversity is still sparse. There is an exception, however. In our data, we found four clinical trials analyzing a promising combination of modified effector cells with checkpoint inhibitors (NCT03412877, NCT03726515, NCT04003649). We expect that many more additional CAR T-cell and TCR T-cell designs and strategies, combinations included, will enter clinical trials in the current decade.

Another increasingly developing gene therapy is the DNA vaccine class. Being both gene therapy and immune therapy, DNA vaccines are designed to provide a specific immune response against a desired antigen. The difference is the antigen used for immunization. The classic antitumor vaccines employ peptides, lysates, or vesicles, whereas DNA vaccines are DNA molecules. DNA vaccines also require a delivery method, which could be a delivery vehicle like a liposome or a technical device—an electroporator or a gene gun. Once delivered to a living cell, the DNA is supposed to be expressed into an antigen and elicit the immune response. We found that all of the trials that could be referred to DNA vaccines were initiated in 2016 and thereafter. There are four DNA vaccine approaches in five phase one brain tumor clinical trials: anti–VEGFR-2 (VMX01) (NCT02718443, NCT03750071), Wilms tumor gene-1 (WT1)/prostate-specific membrane antigen/human telomerase reverse transcriptase (hTERT) (INO-5401) (NCT03491683), and two neoantigen vaccines (NCT03988283, NCT04015700). VMX01 is an antiangiogenic vaccine against vascular endothelial growth factor receptor and is examined either alone or with avelumab (anti-PD-L1 checkpoint inhibitor). The INO-5401 vaccine is composed of three separate DNA plasmids designed against WT1 antigen, prostate-specific membrane antigen, and hTERT genes. This vaccine is a part of a combinational approach implicating interleukin (IL)-12 DNA, cemiplimab (anti-PD1 checkpoint inhibitor), and temozolomide with radiation therapy. DNA plasmids are intradermally delivered via an electroporation device (NCT03491683) (Diehl et al., 2013). The latter two vaccines should target tumor-exclusive peptides, and they also require electroporation for delivery. It might be favorable to use these medications with other immune therapies, particularly checkpoint inhibitors, and maybe these trials will be undertaken in the future.

D. Checkpoint Inhibitors and Immune Modulators

E. Kinase Inhibitors

The role of kinases in oncology is immense. The list of kinases that are associated with human cancers, the mutations of which are called a “driver,” expands every year leading to the development of new screening approaches to study the cancer kinome (Fleuren et al., 2016). In brain tumors, the driver role is attributed to mutations in several kinases, including EGFR, fibroblast growth factor receptor (FGFR), and the phosphoinositide 3-kinase (PI3K)-AKT–mechanistic target of rapamycin (MTOR) pathway (Huse and Holland, 2010). Kinase inhibitors are small molecules that are designed to target a specific kinase or several kinases to stop tumor growth. The presence of this class was consistently high in trials during the last 10 years, yielding more than ten instances of the class in trials per year. Kinase inhibitors are the most versatile class in our database, with 91 agents investigated, and this is not including angiogenesis inhibitors, many of which are kinase inhibitors by action. Kinase inhibitors target 87 unique proteins (kinases and receptor tyrosine kinases) that account for ∼15% of all targets in our data. The primary and secondary targets include isoforms of several kinases: the PI3K-AKT-MTOR pathway, anaplastic lymphoma kinase, EGFR and ERBB2, FGFR, Janus kinase (JAK), KIT, cyclin-dependent kinases (CDKs), tyrosine-protein kinase Lyn, mitogen-activating protein kinase, EPHA2, tropomyosin receptor kinase, WEE1 kinase, and many other targets. The majority of agents aimed at an MTOR (14 agents), EGFR (11 agents), PI3K isoforms (6–11 agents), KIT (7 agents), anaplastic lymphoma kinase (6 agents), and FGFR isoforms (5–6 agents). Almost half of the targets are drugged by a single agent each.

From a drug perspective, the most investigated agent in the class is everolimus, an MTOR inhibitor that has been detected in 15 clinical trials. A high presence of everolimus is not surprising given that the molecule has been approved in astrocytoma and has potential application in other brain tumors. Everolimus was frequently used in combinations, including temozolomide, sorafenib (a kinase inhibitor with strong antiangiogenesis activity), lenvatinib (VEGFR inhibitor), dasatinib (BCR/ABL and SRC inhibitor), and ribociclib (CDK inhibitor). Frequently investigated kinase inhibitors also include abemaciclib, dabrafenib, nab-rapamycin, pablociclib, and many more drugs. One interesting example is nab-rapamycin. This agent is a nanoparticle albumin-bound rapamycin designed to ease the entering of rapamycin into endothelial and tumor cells. This agent has been detected in one phase two clinical trial exploring its use in five different combinations that include radiation therapy, temozolomide, lomustine, bevacizumab, and proteasome inhibitor marizomib to treat newly diagnosed glioblastoma (NCT03463265).

Despite the diversity of kinase inhibitors, to date almost all of them have not passed to the phase three trial stage yet for the treatment of brain cancer. At the moment, selumetinib, a mitogen-activating protein kinase 1/2 inhibitor, was the only exception that passed to phase three trials and currently is under investigation in low-grade glioma and astrocytoma (NCT03871257, NCT04166409). Currently, this agent has been approved by the FDA to treat neurofibromatosis type I, a genetic disease associated with brain tumor development (https://www.fda.gov/news-events/press-announcements/fda-approves-first-therapy-children-debilitating-anddisfiguring-rare-disease).

Kinase inhibitors are the most saturated class in neuro-oncology, given that particular targets like MTOR are drugged by 14 different agents. Despite the diversity of drugs, kinase inhibitors are rarely combined with each other. We found only 17 unique combinations wherein it contained more than one kinase inhibitor, which is 16% of the total number of unique combinations with this class. Kinase inhibitors are not frequently combined with other therapies either. In our data, 49 of 91 agents were never combined at all, and 19 were in combinations only once, thus making a median number of combinations for a kinase inhibitor being equal to zero. And this is despite the prominent phenomenon of drug resistance, which has been known since the adoption of imatinib, the first small-molecule kinase inhibitor to be approved in oncology (Cools et al., 2005; Rask-Andersen et al., 2014). One of the solutions to overcome drug resistance is to use combinational strategies. Thus, we suggest that this direction has three possible scenarios and room for improvement: The first is exploring additional kinase targets via new agents, the second is exploring new combinations of already existing agents to overcome drug resistance mechanisms, and the third is the combination of these two suggestions. It should be noted, however, that concerns, such as targeting and pharmacokinetic issues like penetration of the BBB, must be addressed as well.

F. Angiogenesis Inhibitors

Angiogenesis inhibitors have been tested in 175 trials, with the majority of trials represented by the monoclonal antibody bevacizumab being investigated 104 times. The therapeutic exploitation of the highly angiogenic nature of gliomas drove the investigation of this pharmacological class at the beginning of the decade. However, the number of phase one trials declined dramatically from 19 in 2010 to 1 by the end of 2017. Such a decline has likely resulted from the unsatisfactory clinical performance of the key agent—bevacizumab. A recent meta-analysis revealed that antiangiogenic therapy did not provide a significant improvement in overall survival in patients with high-grade glioma (Ameratunga et al., 2018). Moreover, the results of two randomized double blind placebo-controlled trials, (Avastin in Glioblastoma) study and (Radiation Therapy Oncology Group) 0825 study, evaluating a combination of bevacizumab with chemoradiotherapy in patients with newly diagnosed glioblastoma, demonstrated no improvement in the overall survival. In our dataset, five of six trials testing the addition of bevacizumab to standard chemoradiotherapy completed phase two but failed to progress to the subsequent phase, with two trials being terminated. Despite its approval for recurrent glioblastoma by the FDA in 2009, bevacizumab showed no substantial benefits neither alone nor in combination and has not been approved in other brain cancers thus far.

Besides the monoclonal antibody targeting VEGF, there are 27 more agents, among them 21 small molecules, 4 whole antibodies, 1 aptamer, 1 peptibody, and 1 recombinant human protein. Among the most frequent were the α_v-integrin inhibitor cilengtide and the multikinase inhibitors sunitinib and sorafenib. Each of these agents was tested at most in six (clingtide, sorafenib) to seven (sunitinib) trials. The other 24 agents appeared in an even smaller number of investigations. Various chemotherapeutics, including alkylating agents, topoisomerase inhibitors, antimetabolites, platinum drugs, and spindle poisons, have been investigated in combined regimens with angiogenesis inhibitors, comprising 49% of all combinations with antiangiogenic agents in our dataset. Angiogenesis inhibitors are considered to alleviate interstitial edema and favor the delivery of chemotherapeutics to the tumor tissue (Gasparini et al., 2005). However, a combination of antiangiogenic and chemotherapeutic drugs did not show sufficient superiority over chemotherapy alone because of a lack of postulated synergy between these drug classes (Ameratunga et al., 2018). The results of several studies evaluating bevacizumab with various chemotherapy agents were discouraging as well (Rinne et al., 2013). Angiogenesis inhibitors were frequently combined with additional therapies, and the median number of unique combinations for an agent is equal to two. However, only 3 out of 72 unique combinations with bevacizumab (138 for the all class) have reached phase three trials.

G. Vaccines

We previously discussed DNA vaccines that are part of gene therapy. In this section, we review other entities of the general vaccine class that belongs to immune therapy and has been one of the crucial directions of experimental brain tumor therapies for years. By its nature, a vaccine is the best therapeutic approach for simultaneously targeting multiple tumor antigens. This feature is lacking in many other popular therapeutic classes like CAR T cells, thus tumors are prone to develop antigen-escape resistance to such therapies (Majzner and Mackall, 2018). Although simultaneous targeting strategies are proposed for CAR T-cell therapy to overcome this phenomenon (Rafiq et al., 2020), vaccines are better suited for this approach. It is much more feasible to add several peptides into a vaccine composition rather than design several CAR structures. If designed properly, this class could also provide a tumor-specific treatment action, thus avoiding undesired toxicities.

The vaccine is a versatile class; in our data, we identified 67 different vaccine formulations that have been analyzed in 94 clinical trials. On average, approximately seven vaccines were in phase one clinical trials with brain tumors each year. Depending on the structure, the vaccines could be subcategorized into several subgroups: dendritic cell vaccine (36 entities), peptide (20 entities), tumor lysate (5 entities), and antibody and other structures (1 entity each). So, despite manufacturing challenges, DC-based vaccines currently dominate the field. The target spectrum of vaccines is not that versatile given that it is quite a challenge to find a suitable target in solid tumors with a reasonable safety profile and the immune response to be induced. Overall, we count 16 defined targets for vaccines in brain tumor clinical trials. The top five targets include cytomegalovirus antigens, WT1, survivin (baculoviral inhibitor of apoptosis repeat–containing 5), IDH1, and NY-ESO-1. These targets are aimed by three to five different vaccine compositions. There are also less developed targets that include EGFRviii, IL13RA2, melanoma-associated antigen 1, absent in melanoma 2, EPHA2, ERBB2, histone H3 (H3F3A), ganglioside NeuGc-GM3, and tyrosinase-related protein-2. Several vaccines target more than one antigen simultaneously. For instance, one interesting vaccine design leveraged simultaneous targeting of EGFRviii and NY-ESO-1 using the attenuated bacterial cells as a delivery vehicle to dendritic cells (NCT01967758). Many of the vaccines are personalized by design, even those that did not specifically target cancer neoantigens like the Neovac vaccine (NCT03422094). Even pure lysate-based formulations or DCs stimulated with tumor lysates could be called personalized vaccines, so the number of these designs with multiple targets is quite high with 42 agents.

In contrast to kinase inhibitors, vaccines relatively frequently appear in combination with other therapies. The most common combination of vaccines is with alkylating agents and radiation therapy. Frequently, vaccines are combined with adjuvants as well as with checkpoint inhibitors and immune modulators to favor the immunogenicity. In some designs, vaccinations were also combined with the allogenic or autologous transfer of T cells to provide a better immune response. On average, each vaccine has one unique combination (median).

Despite many potential benefits, only a few vaccines have reached phase three trials. One combinational strategy analyzing DC vaccination with T-cell transplantation and stem cells is currently in a phase three trial (NCT01759810). Another study design that reached phase three was composed of DCs immunized with three components: glioblastoma stem-cell RNA, survivin, and hTERT with temozolomide therapy (NCT03548571). In another phase three trial, DCs were stimulated with tumor antigens (NCT04277221). Two trials, however, are already completed, and the outcomes are not promising. The phase three trial of the ICT-107 vaccine composed of DCs stimulated with tumor-specific peptides has been suspended because of lack of funding, which might be associated with unsatisfactory clinical performance (NCT02546102). The other trial investigated rindopepimut, a peptide vaccine targeting EGFRviii, in combination with GM-CSF and temozolomide (NCT01480479). Despite demonstrating very promising results in a phase two trial, rindopepimut showed no benefits compared with keyhole limpet hemocyanin (control peptide) in a phase three trial (Weller et al., 2017). Thus, this direction has not resulted in drug approval, yet there are still many vaccines that might be effective and are under evaluation. Given the growing popularity of checkpoint inhibitors, we expect to see additional clinical trial designs employing the novel CB drugs with vaccination.

H. Metabolic Modifiers

Metabolic modifiers approach cancer cell killing in a different manner. Given that tumor cells possess the aberrant expression of multiple proteins because of mutations, signaling pathway dysregulations, and tumor microenvironment, these cells are amenable to targeting via metabolism (Bi et al., 2020). Metabolic modifiers usually provide antitumor activities by targeting specific metabolic enzymes within cancer cells. The target spectrum is diverse and depends on a tumor type and the molecule itself. In our data, we identified 36 agents that could be referred to this drug class. The presence of metabolic modifiers in phase one trials has increased from one instance in 2010 to four instances in 2019. An interesting feature of the metabolic modifiers class is that it contains many repurposed medications. One such example is metformin, which has been detected in 10 unique combinations in 8 clinical trials. Metformin is a prominent glucose-lowering agent that has been used as first-line therapy for type two diabetes for more than 60 years in Europe. In the United States, it was approved for the same condition in 1995. This agent has been shown to inhibit the mitochondrial complex 1 to alter the cell energy status (Fujita and Inagaki, 2017). Since metformin functions in metabolic disruption as well as inhibits kinase signaling and angiogenesis, this agent has an antitumor activity (Mallik and Chowdhury, 2018). In brain tumor trials, metformin frequently appeared in combination with temozolomide and radiation therapy. In one trial, metformin was used with memantine, mefloquine, and temozolomide. This regimen was well tolerated and demonstrated promising results in newly diagnosed glioblastoma in a phase one clinical trial (Maraka et al., 2019). Metformin is also a part of an experimental metabolic regimen consisting of metformin, atorvastatin (an antihypercholesterolemic agent), doxycycline (a tetracycline-class antibiotic), and mebendazole (a spindle poison). Currently, this regimen is investigated in various types of cancer, including glioblastoma, in the phase three Safety, Tolerability, and Efficacy of Metabolic Combination Treatments on Cancer (METRICS) trial (NCT02201381) (Agrawal et al., 2019).

The second relatively large group of metabolic modifying agents acts via targeting the IDHs. IDH enzymes catalyze the NADP+ transformation of α-ketoglutarate into isocitrate acid, which is tightly associated with the Krebs’ cycle. In gliomas, these enzymes support angiogenesis, the formation of the tumor microenvironment, and invasiveness of cancer cells (Huang et al., 2019). There are two versions of IDH enzymes, IDH1 and IDH2, and both of them are targeted by agents. In our data, we found five drugs that inhibit mutated forms of IDH1: ivosidenib, vorasidenib, olutasidenib, IDH305, and DS-1001. Ivosidenib is the most common of these, being presented in four different clinical trials. It is used either as monotherapy or in combination with nivolumab, which is currently in a phase two trial (NCT04056910). Vorasidenib is an inhibitor of both mutant IDH1 and IDH2 and was identified as a monotherapy in three clinical trials. Relatively recently, a phase three trial enrolling 366 participants has been initiated to analyze the effectiveness of vorasidenib in IDH-mutated gliomas (NCT04164901). The other three agents targeting IDH1 were in phase one trials. IDH2 is less targeted than IDH1, and we found only two agents that act on mutated IDH2: vorasidenib and enasidenib. Enasidenib was analyzed to treat advanced solid tumors, including gliomas, in a single phase one/two trial (NCT02273739). The trial was completed in 2016, but no results have been published since.

Another prominent group of metabolic modifiers is the inhibitors of IDO1. IDO1 is a member of the l-tryptophan kynurenine metabolic pathway and catalyzes the rate-limiting step of converting l-tryptophan into N-formylkynurenine. In cancer, this pathway has multiple roles in tumor immune suppression, cancer cell mobility, and metastasis (Platten et al., 2019). IDO1 is targeted by four agents: indoximod, BMS-986205, PF-06840003, and epacadostat. Indoximod is a methylated tryptophan and the most tested of these four agents that appeared in five unique combinations in three clinical trials. The most investigated combination of indoximod is with temozolomide. PF-06840003 was studied in a single trial, but it was terminated because of a sponsor decision (NCT02764151). Epacadostat and BMS-986205 are in phase one trials and combined with nivolumab with or without temozolomide and radiation. One trial with epacadostat has been recently completed (NCT02327078).

Other metabolic modifiers were much less frequent and mostly stay at the phase one in single trials, with a few exemptions. One is the previously mentioned phase three Safety, Tolerability and Efficacy of Metabolic Combination Treatments on Cancer (METRICS) trial, wherein atorvastatin is combined with metformin, doxycycline, and mebendazole (NCT02201381, NCT02796261). Another phase three trial investigates a combination of eflornithine with lomustine. Eflornithine is an irreversible inhibitor of ornithine decarboxylase that is required for polyamine biosynthesis. Polyamines possess diverse functions within tumor cells and are linked with oncogenic signaling pathways, including the phosphatase and tensin homolog-PI3K-mTOR, WNT signaling, and RAS pathways (Casero et al., 2018).

I. Transcription Modifiers: Tumor Targeting at a Different Level

Most of the targeted therapies act on receptors or on the pathway members that transduce the signal from a receptor, ultimately leading to the expression of target proteins. Usually, the last chain in the pathway is represented by a transcription factor that upon biochemical modification becomes activated and translocates into the nucleus to initiate the expression of target genes. A portion of targeted therapies acts on transcription factors or associated proteins rather than other pathway members. An interesting trend is an increase in the number of transcription modifiers in clinical trials, which rose from one agent per year in 2010 up to three agents per year in 2019, with 21 agents cumulatively identified to date. Several of these agents could be subclassified into four groups: E3 ubiquitin-protein ligase Mdm (MDM) 2 inhibitors, bromodomain and extraterminal motif (BET) protein family inhibitors, JAK/signal transducer and activator of transcription (STAT) axis inhibitors, and hypoxia-inducible factor 2α (HIF2A) inhibitors. MDM2 is an E3 ubiquitin-ligase whose primary function is to inhibit the activity of the key tumor suppressor gene TP53. MDM2 is deregulated and overexpressed in various tumors, including glioblastoma (Wade et al., 2013). Three agents that target MDM2 were identified: the peptide ALRN-6924 that inhibits both MDM2 and MDM4, the small molecular inhibitor idasanutlin that blocks the interaction between P53 and MDM2, and AMG 232 that demonstrate similar characteristics to idasanutlin. Each of the three drugs is in phase one clinical trials (NCT03654716, NCT03158389, NCT01723020, NCT03107780). Azurin-derived cell-penetrating peptide p28 exhibits a close action to MDM2 inhibitors. However, in contrast to the previous group, p28 binds directly to P53 to protect it from degradation (Yamada et al., 2013). This agent was well tolerated yet failed to demonstrate promising outcomes in pediatric patients with recurrent or progressive central nervous system tumors in a phase one trial (Lulla et al., 2016).

BET proteins are epigenetic regulators that contain tandem bromodomains (BD1 and BD2) as well as extraterminal and C-terminal domains. The primary function of these proteins is the regulation of the expression activity of various transcription factors and oncogenes, including nuclear factor–κB, Myc proto-oncogene family, E2F2, CDK6, and others. Thus, there is a rationale for using BET inhibitors for cancer therapy (Stathis and Bertoni, 2018). Four BET inhibitors were identified in trials for brain cancer: birabresib, INCB057643, BMS-986158, and CC-90010. However, the actual clinical performance of BET inhibitors in neuro-oncology is questionable. A phase two trial analyzing birabresib (MK-8628) in recurrent glioblastoma was terminated because of a lack of activity of the investigative drug (NCT02296476). In contrast, INCB057643 failed a phase one/two clinical trial in advanced solid tumors because of safety reasons rather than activity (NCT02711137). The other two inhibitors BMS-986158 and CC-90010 are still under examination in recently started phase one trials, so there are not any results yet (NCT03936465, NCT04047303).

Inhibitors of JAK/STAT and HIF2A are not frequent in brain tumor clinical trials, with each of these groups containing only two agents. JAK/STAT is a proliferative signaling pathway that consists of JAK, which in turn activates STAT proteins. Upon activation by phosphorylation, STAT proteins dimerize and translocate into the nucleus to activate expression of target genes (O'Shea et al., 2015). WP1066 and napabucasin, JAK/STAT inhibitors, were investigated in high-grade gliomas in phase one clinical trials (NCT01904123, NCT02315534). However, despite years passed, these agents have not reached phase two. HIF2A is also a transcription factor and is induced under hypoxic conditions common for tumors. It has been discovered that this protein is crucial for glioblastoma stem cells (the most aggressive and proliferating cells in a tumor) and predominantly expressed in this population and also correlates with poor patient survival (Li et al., 2009). Two monotherapies, PT2977 and PT2385, that target HIF2A were identified in our dataset. PT2977 is in phase one, whereas PT2385 has completed the phase two trial (NCT02974738, NCT03216499).

Selinexor is an interesting drug in which the transcription modifying action is based on targeting nuclear transport rather than transcription factors. This molecule inhibits exportin, a protein found on the nuclear membrane, which exports various tumor suppressor transcription factors from the nucleus, therefore hampering their antiproliferative activity and facilitating tumor progression. Selinexor in combination with dexamethasone received accelerated approval in the United States for the treatment of myeloma in 2019 (Syed, 2019). For the treatment of brain tumors, this agent has appeared in four clinical trials. Early trials of selinexor analyzed this agent as monotherapy (NCT01607905, NCT01986348, NCT02323880), whereas a recent trial initiated in 2020 is analyzing this drug in combination with temozolomide (NCT04216329). Despite the fact that the first two trials had been completed, we did not find results.

J. Oncolytic Viruses Are Still in Early Development

Oncolytic viruses are a relatively unique direction for cancer therapy. Oncolytic viruses are defined by their ability to selectively replicate in tumor cells without damaging the normal ones. To date, viruses are designed using either genetic engineering employing immune-modulating and safety modifications or using natural tumor-tropic viruses, such as reovirus (Fukuhara et al., 2016). In 2010, there were no trials analyzing oncolytic viruses in brain tumors; in 2017, this number peaked, reaching three trials per year. We identified 10 unique oncolytic viruses in brain tumor clinical trials. These viruses originated from different viral taxons: Four viruses are derived from herpes simplex virus (HSV); two were adenoviruses; and measles virus, parvovirus, polio virus, and reovirus have only one entity of their origin. Most of the viruses were used as monotherapies. Oncolytic adenovirus AD5-DNX-2401 is the most frequent member of its class and appeared in six clinical trials. This agent was also combined with interferon γ-1b, temozolomide, or pembrolizumab. Combinations with interferon or pembrolizumab might be promising therapeutic designs because of the ability of oncolytic viruses to elicit the antitumor immune response (Fukuhara et al., 2016), which could be enhanced by the immune-modulating activity of interferon γ-1b and pembrolizumab. Currently, a combination of AD5-DNX-2401 with pembrolizumab is one of a few combinations that has reached phase two trials on different types of brain tumors (NCT02798406). Another relatively abundant virus is PVS-RIPO, a modified polio virus, which was analyzed in three clinical trials and reached a phase two stage to be investigated in patients with high-grade glioma (NCT02986178). To date, only the two abovementioned viruses have reached phase two clinical trials.

K. PARP Inhibitors: Now beyond Breast Cancer

PARP inhibitors are small molecules that are designed to inhibit the enzyme PARP. PARP is a crucial enzyme participating in the DNA base excision repair, homologous recombination, and nonhomologous end-joining processes that are crucial for cell homeostasis. PARP inhibitors demonstrated impressive clinical performance in BRCA1/2 mutated tumors, and their application seems to expand beyond these cancers (Tangutoori et al., 2015). In our data, we identified seven unique PARP inhibitors that were analyzed in 19 clinical trials. Of these seven agents, olaparib, pamiparib, and veliparib were the most common. Olaparib was found in eight clinical trials. If combined, this drug was used with temozolomide with or without radiation as well as with cediranib or durvalumab. Pamiparib and veliparib were used in three clinical trials each. To date, PARP inhibitors mostly are in the phase two stage. However, veliparib is currently under evaluation with or without temozolomide in glioblastoma/gliosarcoma patients in a phase two/three trial (NCT02152982).

L. Radiopharmaceuticals: A Targeted Radiation Therapy

Another interesting class is represented by radiopharmaceutical agents. This is a type of radiation therapy that is drug-based, compared with standard radiation therapy. Despite being relatively rare, this class has demonstrated increasing interest, especially in the last 3 years. We count eight unique agents identified as radiopharmaceuticals. This class could be subcategorized, depending on the structure: one group is radiolabeled antibodies, and another is radiolabeled small molecules or peptides. In our database, two drugs are small molecules, four drugs are radio-antibody conjugate, two are peptides, and one agent is a macromolecular complex (liposomal rhenium Re186). In the majority of trials, radiopharmaceuticals appeared as monotherapies. Currently, many agents have not progressed beyond the phase one clinical trial stage. An exception is yttrium Y 90-DOTA-tyr3-octreotide, a small radiolabeled peptide that has reached phase two clinical trials and currently is under evaluation in several tumors, including medulloblastoma (NCT02441088, NCT03273712). Also, lutetium Lu 177 dotatate, a radiolabeled peptide, is tested in two phase two trials on inoperable/progressive meningioma (NCT03971461, NCT04082520).

M. Embryonic Pathway Inhibitors

Embryonic pathway inhibitors represent a specific branch of targeted therapies designed to target the oncogenic pathways that are predominantly active during the embryonal development. These pathways include Hedgehog, Notch, and Wnt signaling. These pathways also are commonly attributed to the concept of cancer stem cells and may play a significant role in tumor development (Takebe et al., 2015). We identified seven agents that function as embryonic pathway inhibitors. Depending on the target, these agents could be subdivided into two groups: γ-secretase inhibitors (Notch signaling) (RO4929097, MK-0752) and smoothened (SMO) inhibitors (Hedgehog signaling) (LEQ506, sonidegib, vismodegib, glasdegib, ZSP1602). All of these seven drugs are small molecules. In several trials, these agents were combined with kinase inhibitors, angiogenesis inhibitors, alkylating agents, and radiation. The number of agents has declined sharply from eight molecules in 2010 to one in 2018 in phase one trials.

One might ask the reason for such a decline, and probably this is related to the overall uncertainties and limitations in the cancer stem-cell concept (Pastrana et al., 2011; Rahman et al., 2011; Medema, 2013; Tarasov et al., 2019b). Currently, none of the agents in the class reached phase three trials. Consistently, the percentage of discontinued trials (terminated, suspended, withdrawn) for embryonic pathway inhibitors is one of the largest among all therapeutic classes (Supplemental Fig. 1). However, there are six clinical trials that are still recruiting participants and may successfully demonstrate the rationale in using embryonic pathway inhibitors for brain cancer therapy.

N. Histone Deacetylase Inhibitors: Reaching the Third Stage

HDACs represent a family of regulatory enzymes that control the functioning of various cellular processes through the modification of gene expression. HDACs function via removing the acetyl groups from N-acetyl lysine in histones. Upon deacetylation, histones bind with DNA, thus preventing its transcription. Generally, there are 18 HDACs that demonstrate distinct expression patterns and perform slightly different functions. For further information, read these publications (Medema, 2013; Ho et al., 2020; Verza et al., 2020). Nine HDAC inhibitors were identified in trials for neuro-oncology. The most investigated agent is vorinostat that has been tested in 10 trials, which is followed by valproic acid (six trials) and panobinostat (four trials). HDAC inhibitors were analyzed both alone and in combinational regimens, which included temozolomide, radiation therapy, and, rarely, bevacizumab or another agent. Of all agents, only vorinostat and valproic acid have reached phase two/three and phase three clinical trials, respectively, and are under evaluation in combinations in several types of brain tumors (NCT03243461, NCT01236560, NCT02265770). Vorinostat was analyzed in combination with temozolomide and bevacizumab, whereas valproic acid was combined with either temozolomide or another chemotherapeutic agent. An interesting example of HDAC inhibitors is tinostamustine, a fusion molecule composed of an alkylating agent bendamustine and vorinostat. To date, this agent is under evaluation with radiation therapy in a phase one trial (NCT03452930). MTX110 is a water-soluble panobinostat administered as a nano formulation through a convection-enhanced delivery system. Loading a drug into water-soluble nanoparticles allows to effectively use the convection-enhanced delivery system to bypass the BBB and avoid systemic toxicity in comparison with the poorly soluble oral formulation, which failed to penetrate the BBB (Singleton et al., 2018). In our dataset, the earliest trials of panobinostat analyzed this agent without a delivery system.

Despite two drugs that have demonstrated promising performances, the overall number of agents in phase one trials declined from five in 2010 to one in 2019. This class exhibits the fourth largest percentage of discontinued clinical trials among all therapeutic classes. It should be noted, however, that the terminated trials are relatively old, and only one trial, which analyzed the combination of vorinostat, erlotinib, and temozolomide, has failed because of issues with treatment (toxicity) (NCT01110876). Thus, this downward tendency might be a data artifact, taking into account the small number of all trials in total (27) and a small number of trials per year (5 or fewer).