Targeting Tumor Suppressor genes for Cancer Therapy

Abstract

Cancer drugs are broadly classified into two categories: cytotoxic chemotherapies and targeted therapies that specifically modulate the activity of one or more proteins involved in cancer. Major advances have been achieved in targeted cancer therapies in the past few decades, which is ascribed to the increasing understanding of molecular mechanisms for cancer initiation and progression. Consequently, monoclonal antibodies and small molecules have been developed to interfere with a specific molecular oncogenic target. Targeting gain-of-function mutations, in general, has been productive. However, it has been a major challenge to use standard pharmacologic approaches to target loss-of-function mutations of tumor suppressor genes. Novel approaches, including synthetic lethality and collateral vulnerability screens, are now being developed to target gene defects in p53, PTEN and BRCA1/2. Here, we review and summarize the recent findings in cancer genomics, drug development, and molecular cancer biology, which show promise in targeting tumor suppressors in cancer therapeutics.

Introduction

Cancers are genetic diseases. They arise through a multi-stage process, in which germ-line and somatic gene mutations and clonal selection of precancerous cells with the most robust and aggressive growth properties are the primary driving force [1]. Whereas a small portion of hereditary cancer syndromes are caused by germ-line gene mutations, most human cancers are driven by genetic and epigenetic changes in somatic cells. Together, these genetic changes contribute to the hallmarks of malignant cancers, including uncontrolled cell proliferation, resistance to cell death, immortalization, activated invasion, distant metastasis, immune evasion and angiogenesis. Inheritable genetic variations, such as single nucleotide polymorphism, noncoding and coding gene copy number variation, also contribute to cancer development by influencing the susceptibility to malignancy transformation [2,3]. Two classes of genes—oncogenes and tumor suppressor genes—are often targets for mutations and variations during molecular evolution of human cancers. The ultimate goal of cancer therapeutics development is to selectively target cancer mutations that sustain tumor growth and progression. Identification of driver oncogenes in human cancer has led to the development of lead compounds and synthetic drugs that specifically target those driver oncogene-encoding proteins. However, many more driver mutations that cause loss-of-function in tumor suppressor genes had been thought of as undruggable because of technical difficulties or mechanistic complexity. Recent discoveries in human genomics have revealed new therapeutic opportunities for targeting tumor suppressor mutations.

How are driver genes and driver mutations identified in human cancer?

The most exciting development in cancer studies is the emergence of next-generation sequencing technologies that provide exquisite sensitivity and resolution with greatly reduced costs [4,5]. Cancer genomics has been transformed by the application of new technology, greatly expanding our understanding of cancer biology. Remarkable advances in our identification of cancer gene expression profiles and disease-associated alterations of cancer genomes have been reported in the past decade [6,7]. Based on parallel sequencing platforms for target genes, whole cancer genome, whole transcriptome or proteome, bioinformatic and biostatistical approaches are being applied to combine genetic and clinical data to analyze cancer prevention and treatment [8–10]. With the availability of multi-level cancer databases on the horizon for clinical care, a key question now is how to identify mutations that play causal roles in cancer. In other words, what are driver genomic events?

A primary feature of cancers is genetic instability arising from defects in DNA damage signaling and repair pathways [11]. Therefore, a tumor often bears many genetic alterations and intra-tumor heterogeneity [12–16]. Moreover, metastatic tumors often adopt new mutations that do not exist in their primary tumors [17–19]. Distinguishing driver mutations from subsequent passenger mutations relies on their capability to induce cell transformation. However, due to the large number of mutations in cancer cells, this assay is often not feasible. Recent progress in computational biology enables us to analyze the vast amount of data generated by current cancer genomics projects and predict mutations, genes and pathways driving the tumorigenesis [19–22].

The process of identifying functional and driver variants can often be divided into three independent, but related, approaches [23]. The first approach is to map gene mutations to annotated functional genomic features, identify their biological or molecular consequences, and then determine whether these mutations have been previously reported in the literature [24,25]. The second approach applies computational methods to predict the nature and magnitude of a mutation’s impact on gene or biological functions. Some bioinformatics tools have been specifically developed to rank somatic tumor mutations, including SIFT [26], Polyphen-2 [27], FATHMM [28], CHASM [29,30], and transFIC [31]. Finally, the third approach depends on biostatistical analyses to uncover positively selected candidates across different cohorts [32]. It is worth noting the difference between driver genes and driver mutations. While driver genes are supposedly enriched to result in driver mutations, the approaches taken to detect these two are quite different. The search for driver mutations usually involves the detection of key amino acid residues for the function of proteins that are known or likely to be associated with tumorigenesis. However, the identification of driver genes relies on picking up the traces or signals left by positive selection across cohorts of tumor samples [32]. It provides an assessment of genes based on the collective analysis of all the somatic mutations across the tumor cohort.

In addition to these bioinformatics tools, predictions can also be made by other means. Gene mutations that have been previously identified as cancer-related, or involved in oncogenic signaling pathways, are more likely to be driver mutations [33]. Since passenger mutations are assumed to occur randomly throughout the genome, whereas driver mutations occur frequently in cancer-related genes, a gene that exhibits higher prevalence of somatic mutations than expected by chance is considered more likely to be a driver mutation, even though it has been recognized that passenger mutations may occur at specific loci in clusters, and driver mutations can exist at low frequency in cancer cells. During molecular evolution of tumorigenesis, a tumor cell often acquires new genetic alterations, in which additional driver mutations promote clonal expansion, through positive selection from enhanced growth or treatment resistance. Under some circumstances, a preexisting passenger mutation may become a driver mutation by contributing to treatment resistance and clonal expansion [34].

Druggability of cancer-associated genes is assessed by bioinformatic analysis of multiple databases

Once the driver mutations are identified in cancers, the primary challenge then becomes drug discovery for clinical intervention, which directs clinical studies to match patients with targeted agents. In particular, resistance to both cytotoxic and targeted drugs continues to emerge, and the genetic complexity and heterogeneity in many cancers is recognized [35]. Therapeutic intervention of driver oncogenic alterations and translational medicine has become a focus of cancer research. In addition to systematic approaches, such as immunotherapy, hormonal therapy, and chemotherapy, the development of new therapies is often based on inhibition of disease-associated molecules or signaling pathways. For example, small chemical inhibitors and monoclonal antibodies has been a great success in the treatment of HER2-positive breast and ovarian cancer. Two inhibitors of the BRAF protein have been approved by the FDA for the treatment of melanoma. Modern experimental research and computational biology have facilitated the search for new druggable targets that potently and selectively modulate the functions of cancer cells.

Given that a large number of driver mutations and genes have been identified from cancer genomics studies, it is extremely important to prioritize candidates from those identified by large-scale genomics initiatives. Systematic, objective and multidisciplinary computational studies allow for effective, unbiased assessments of biologically compelling gene lists for drug discovery [23]. Briefly, the list of candidate genes is first annotated using homology with the targets of approved pharmaceuticals; the molecular and cellular properties of any previously reported small molecules; three-dimensional structure and druggability; and functional class and subcellular localization. Potential drug target screens then need to integrate clinical information from databases, such as The Cancer Gene Atlas (TCGA) or the Cancer Genome Project (CGP) [36,37]. The combined data analyses generate a ranking list of potential targets by incorporation of available supporting evidence for their chemical tractability (Figure 1). The rank and importance of different features will meet eventual requirements.

Figure 1. Overview of the systematic approach to identify and validate new cancer drug targets.

First, new generation sequencing of tumor tissues or cells generate the mutational profiles and other disease-specific alterations of cancer genomes. Second, potential cancer drug target genes are identified by integration and comprehensive analysis of multiple databases. A series computational and experimental approaches will then be applied to validate the potential targets for drug development and test.

The Cancer Gene Census is a manually curated set of genes that have potentially causative mutations or other genomic abnormalities in human cancers [38]. This data set extracts a large gene list from initial experimental investigation, which is composed of 571 genes or loci in total. Each Census protein was classified into a single class based on its biochemical and cellular features. Halling-Brown et al. developed an integrated bioinformatic tool, canSAR, for drug discovery using biological annotation, chemical screening, RNAi screening, gene expression, copy number variation, and protein structural data [39]. This platform allows for easy access to multidisciplinary data for drug discovery and design. Through analysis of canSAR database, Patel et al. identified a total of 132 potential drug targets that have at least one piece of evidence for chemical tractability, of which 46 have few or no published compounds for targeting [23]. The number of potential druggable proteins increases to 90 by mapping the annotation from homologous targets. While these potential drug targets merit further investigation in drug screen and design, a significant amount of driver genomic alterations in human cancers are associated with non-druggable genes. In particular, given the mutations or deletions in a number of tumor suppressors, it has been a major challenge to restore the activity of tumor suppressors or the associated signaling pathways.

Viral and non-viral vectors restore the expression of tumor suppressor genes

A cancer therapy is supposed to selectively target the genetic lesions that initiate and maintain cancer proliferation and survival, while keep normal cells intact. Although most cancers harbor multiple oncogenic mutations, accumulating preclinical and clinical data now support that many cancers are sensitive to inhibition of single oncogenes, a concept named oncogene addiction [40]. However, a weakness of oncogene-targeting approaches is that they do not address the problem of cancer progression as selected by the recessive phenotypes of genetic instability and apoptotic resistance that arise from loss-of-function defects of tumor suppressors, such as TP53, PTEN and BRCA1 [41]. The most direct way to restore the functions of tumor suppressors is to reintroduce wild type tumor suppressor gene into target cancer cells for expression. Gene therapy for human cancer treatment includes transferring genetic material into a host cell through viral and non-viral vectors, and modulation of tumor cells or tumor microenvironment to reduce tumor angiogenesis, or to increase tumor immune antigenicity [42]. So far, modest success has been achieved with relatively minimal side effects [43,44]. The new generation of viral and non-viral vectors has significantly increased the efficacy of treatment and biosafety [45,46]. Several methods have been developed to facilitate the entry of transgene into target cancer cells. Viral vectors are able to infect cancer cells and transduce the target gene into the host cell as part of their replication process. Non-viral vectors and approaches use physical, chemical, as well as other modes of genetic transfer. These approaches have the advantage of safety and easy modifiability, but sometimes have a lower transfection efficiency compared to viral vectors [46].

The first FDA-approved gene therapy experiment in the United States was conducted in 1990 for a patient with a genetic defect that left her with ADA-SCID, a severe immune system deficiency [47]. Since then, many clinical trials using gene therapy have been conducted for patients with cancer, among which favorable results were reported in patients with chronic lymphocytic leukemia [48,49], acute lymphocytic leukemia [50], brain tumors [51–53], as well as others [54–56]. Several commercially approved medications for gene therapy were released, including ONYX-15 (Onyx Pharmaceuticals) for refractory head and neck cancer [57–59]. Onyx-015 is an adenovirus that was developed with the function of the E1B gene knocked out, so that cells infected with Onyx-015 are incapable of blocking p53 [60–63]. If Onyx-015 infects a normal cell, with a functioning p53 gene, it will be prevented from replication by p53. However, if Onyx-015 infects a p53-deficient cell it should be able to survive and replicate, resulting in selective destruction of these cancer cells. Onyx-015 has been extensively tested in clinical trials [64]. In spite of its safety, limited therapeutic effect has been demonstrated.

Given the high mutation frequency of the tumor suppressor p53 in human cancers, restoration of wild-type p53 function is an intriguing strategy in cancer therapy. This can be achieved by introduction of an intact p53 gene using a viral vector, in most cases an adenoviral vector (Adp53). This gene replacement therapy has been demonstrated to suppress tumor growth because expression of exogenous p53 gene induces cell death and growth arrest in a variety of p53-inactivated tumor cells [65]. An advantage of the adenovirus delivery system is that the vector DNA cannot be integrated into the host cells. Preclinical studies have shown that Adp53 induces tumor regression in various cancers, including head and neck cancer [66], colorectal cancer [67], lung cancer [68], ovarian cancer [69], bladder cancer [70], and prostate cancer [71]. Adp53-based gene therapies were evaluated in many clinical trials. Gendicine is the world’s first Adp53-based gene therapy product approved by the State Food and Drug Administration of China for the treatment of head and neck squamous cell carcinoma [72]. Other Adp53 vectors such as SCH-58500 and Advexin have been developed and used in various clinical trials [73–78]. While Adp53 gene therapy is well tolerated, feasible and exerts promising antitumor effects in some cases, its overall clinical efficacy is not conclusive. No Adp53 therapies have been approved in the USA.

Small chemical compounds restore the activity of tumor suppressor genes

Tumor suppressors are often repressed by down-expression, protein degradation or inactivation in cancers. In other cases, gene mutation either inactivates the tumor suppressors, or changes their normal functions, a phenomenon that is referred to as loss-of-function. Mutation of the tumor suppressor p53 is the master driver force for tumor growth and metastasis [79]. In response to many stress signals, p53 is activated and transcriptionally induces a myriad of target genes, including both protein-encoding and non-coding genes, controlling cell cycle regulation, DNA repair, senescence, apoptosis, autophagy and metabolism of tumor cells [80–82]. Around half of human cancers harbor mutant p53 and, in the majority of the rest of cancers, p53 is inactivated through different mechanisms. Most genotoxic and chemotherapeutic drugs, such as alkylating agents and platinum-based drugs, trigger DNA damage response and thus activate the p53 pathway, but those effects are non-specific and also cause systemic toxicity [83]. As a major p53 inhibitor, Mdm2 is an E3 ubiquitin ligase for p53 degradation and also binds p53 to inhibit its transcriptional activity [84,85]. In the past decades, much effort has been made to develop Mdm2 inhibitors that can specifically prevent p53 from Mdm2 repression. One of the major developments in this field is the finding of Nutlin, a small molecule disrupting Mdm2-p53 interaction [86]. It has been shown that Nutlin resembles an Mdm2-binding peptide that competitively interacts with MDM2, leading to p53 stabilization and activation [87,88]. The Nutlin derivative RG7112 has been in clinical trials for various types of cancer [89]. In addition to Nutlin, other compounds that inhibit Mdm2 have been developed and tested in Phase I trials, including MI-773, MK-8242, and RO5503781 [90,91].

Most cancer-associated mutations in p53 are missense single base-pair substitutions. These missense mutations are clustered within the central and most conserved DNA-binding domain of p53, among which six hot-spot mutations occur with unusually high frequency [83]. This is in striking contrast to many other tumor suppressors, such as RB1, APC and VHL, in which the primary mutations are deletion or nonsense [92]. Therefore, a long sought goal has been to convert existing mutant p53 proteins into somewhat wild type forms that exhibit normal p53 functions, thereby allowing p53 to carry out its tumor suppressive activity in cancer cells. The most advanced compound so far has been PRIMA-1 (also known as APR-246), which has been shown to reactivate missense mutants (R273H and R175H) of p53 to regain at least some of wild type p53 and thus halt tumor growth [93–95]. However, PRIMA-1 has also been shown to induce cell death independently of p53, hence complicating its clinical applications [96].

Inactivation of Von Hippel–Lindau (VHL) disease tumor suppressor gene VHL is commonly seen in clear-cell renal cell carcinoma (ccRCC). The VHL protein plays a key part in cellular oxygen sensing by targeting hypoxia-inducible factors for ubiquitination and proteasomal degradation [97]. A large portion (55%) of somatic VHL mutations identified in sporadic ccRCC are frame-shift or nonsense mutations that result in loss of VHL function. Nascent pVHL is shuttled from the ribosomal machinery with the assistance of the chaperone protein HSP70 [98]. VHL then forms a ternary complex with the transcription elongation factors C and B, termed the VCB complex. Formation of the VCB complex is mediated by the hetero-oligomeric chaperonin TCP1 ring complex (TRiC) [99]. Incorrectly folded or free forms of VHL are degraded through the ubiquitin–proteasome system, which requires another chaperone, HSP90. As distinct chaperones are responsible for the folding and quality control of VHL, strategies for refolding and stabilization of VHL potentially restore their tumor suppressor activity. The wide-spectrum proteasome inhibitor MG132 is capable of increasing VHL expression levels. A cell-based screen of the Prestwick Chemical Library compounds identified several compounds that up-regulate the levels of a mutant VHL protein (VHLW117A) [100]. The protein levels of VHLR167Q (a recurrent mutation in type 2B VHL disease) dictate its ability to down-regulate HIF2α and suppress tumor growth. The proteasome inhibitors bortezomib and carfilzomib stabilize VHLR167Q and increase its ability to suppress HIF2α [101].

Further understanding of regulatory mechanisms on the expression, activation and stabilization of tumor suppressor proteins will provide foundation to up-regulate their levels in cancer cells and enhance their tumor suppressive activity.

What is synthetic lethality?

There has been a huge amount of interest in therapeutic targeting of tumor suppressor genes for several decades. However, results have been mixed and there are no anticancer agents in clinical applications that directly target tumor suppressors. One reason for the absence is that cancer cells sometimes completely lack a particular tumor suppressor gene, and therefore there is nothing to target. In other cases, tumor suppressor genes are mutated, and restoring the function of a protein is inherently difficult. In addition, it is often a challenge to determine the status of tumor suppressor proteins. PTEN is one of the most frequently disrupted tumor suppressors in cancer. However, loss-of-function mutations in PTEN occurs only in a fraction of PTEN-deficient tumors [102]. PTEN expression is lost as a result of epigenetic and transcriptional silencing in many tumors. Therefore, it is often more important to know the biological consequences caused by defective tumor suppressors instead of the mechanisms for their down-regulation.

The bottleneck to the development of safe and effective cancer drugs is not the inability to identify chemicals that will kill cancer cells. As a matter of fact, a large pool of compounds has been ready for clinical use. However, it is much more challenging to identify potential drugs that specifically kill cancer cells at concentrations that do not harm patients. Currently, most chemotherapy drugs have a relatively narrow therapeutic window. When administered to patients, these drugs also damage rapidly dividing normal cells, such as bone-marrow hematopoietic precursors. Some of these drugs are even toxic to normal cells that are not rapidly dividing, such as doxorubicin (toxic to the heart) [103], and bleomycin (toxic to the lung) [104]. The toxicity and narrow therapeutic window limit their clinical applications. Therefore, it is of great importance to develop cancer drugs that have higher therapeutic indices than those classic cytotoxic agents. Genetic alterations in human cancers include gene activation, amplification, or inactivation and allelic haplodeficiency. These genetic features unique to cancer cells can be exploited to develop therapies that are inherently tumor-specific. One of therapeutic approaches is to identify synthetic lethal gene targets. Synthetic lethality arises when a combination of mutations in two or more genes leads to cell death, whereas a mutation in only one of these genes does not [105]. A large number of studies have explored synthetic lethality approaches to target those maladaptive genetic changes in cancer cells.

‘BRCAness’ in cancer cells leads to synergistic lethality with PARP inhibitors

The most studied synthetic lethality approach is the concept of ‘BRCAness’, which has been well translated to the clinic. BRCAness was introduced by Ashworth and colleagues in order to describe the phenotypic traits that some sporadic tumors share with tumors in BRCA1/2 germline mutation carriers and reflects similar causative molecular abnormalities [106]. A major risk factor for breast and ovarian cancer is inheritance of a mutation in one of the breast cancer susceptibility genes, BRCA1 or BRCA2 [107,108]. BRCA1 and BRCA2 are tumor suppressor genes. They function in homologous recombination (HR), an essential step for high-fidelity repair of double-strand DNA breaks (DSBs). Homologous recombination is the major mechanism for protecting genome integrity in proliferating cells, because other DSB repair pathways, including non-homologous end joining (NHEJ), are error-prone and generate chromosome deletions and translocations [11]. BRCA1-deficient cancer cells are defective in HR, cell cycle checkpoints, and transcriptional control. Recent studies have revealed that cells with dysfunctional BRCA1 or BRCA2, in contrast to cells that have normal BRCA function, are much more sensitive to PARP inhibitors because of their defects in homologous recombination.

Polyadenosyldiphosphate ribose polymerase (PARP) plays a major role in nucleotide excision repair (NER) or base excision repair (BER) when homologous recombination is absent or defective. Inhibition of PARP results in accumulation of single-strand DNA breaks (SSB) in cells, which are converted into DSBs at replication forks [105]. BRCA-deficient cancer cells are unable to utilize the HR to precisely repair those DSBs. Instead, DSBs are repaired via NHEJ or the single-strand annealing pathway of HR with large numbers of chromatid aberrations leading to cell lethality. In a number of clinical trials using PARP inhibitors as monotherapy agents or one regimen in combined therapy, PARP inhibitors have shown promise as a powerful therapeutic tool, especially in the treatment of BRCA-associated breast and ovarian cancers but also in tumors where genes associated with the BRCA signaling pathway may be dysfunctional [109]. These genes include ATM, BARD1, BRIP1, CHEK1, CHEK2, FAM175A, MRE11A, NBN, PALB2, RAD51C, and RAD51D.

Early clinical results suggest that PTEN-deficient cancers may be sensitive to PARP inhibitors. A case report showed significant tumor response to the Olaparib, a PARP inhibitor, in a patient with PTEN-null metastatic endometrial cancer with wild-type BRCA1/2 [110]. A majority of patients with advanced prostate cancer in a phase I trial with the PARP inhibitor Niraparib exhibited significant clinical response, suggesting that PTEN-deficient prostate cancer is likely sensitive to PARP inhibition [111]. In addition, a recent study identified nemo-like kinase (NLK), Polo-like kinase 4 (PLK4) and Monopolar spindle 1 (MLK) as synthetic lethal genes in PTEN-deficient cancer cells [112]. It is conceivable that inhibitors against these kinases may be applied to treat patients with PTEN-deficient cancers.

In the context of DNA damage-inducing chemotherapy and radiotherapy, synthetic lethal interactions can also identify key genes whose inhibition selectively sensitizes cancer cells to genotoxic stress. Genes that are synthetically lethal to mutant p53 have been extremely interesting in cancer research. A recent study has shown that MK2, an upstream kinase activating p38, appears to be essential for survival of p53-deficient non–small-cell lung cancer cells in response to cisplatin-based chemotherapy [113]. In addition, inhibition of CHK2 or ATM in p53-deficient tumors can also directly sensitize cancer cells to chemotherapy [114]. Synthetic lethality screens have now developed from simplistic models to genome-wide screens using short hairpin RNA (shRNA), small interfering RNA (siRNA), and small molecule compound libraries screens. As we understand more about the complexity of cancer cell signaling, an increased number of targets will be identified as potential synthetically lethal candidates for drug development.

Loss of tumor suppressors creates collateral vulnerability

Genomic instability is a hallmark of cancer, which is associated with tumor progression and is considered an active process that drives tumor evolution. There are numerous efforts to manipulate DNA damage responses to selectively induce tumor cell death through catastrophic genomic instability and DNA repair defects. Radiotherapy and chemotherapy take advantage of DNA repair deficiency in cancer cells to induce cell death. However, because they kill fast-growing cancer cells, these DNA damaging agents also can harm proliferating healthy cells, leading to side effects such as fatigue or infection. Various efforts have been made to improve the response to DNA damaging agents and to understand the mechanisms of resistance to current cytotoxic therapeutics [11].

Comprehensive analyses of cancer genome and transcriptome have opened previously inaccessible opportunities for the development of personalized medicine and targeted therapy. Novel driver oncogenes aberrantly magnified or activated in cancers have been identified, which provides new and promising drug targets. However, it remains a challenging question regarding exploitation of loss-of-function mutations or deletions of key tumor suppressors. In most cancers, deletion regions containing tumor suppressor genes often include loss of other neighboring genes. Cancer genomes are characterized by copy number variations, including amplification of driver oncogenes and deletion of tumor suppressor genes. Although these driver events promote malignant tumorigenesis, most gene alterations are passenger events -- the consequences of increased genomic instability in cancer -- and not contributors to tumor development. In most cases, genomic alterations on large fragments of chromosomal DNA are not selected against in molecular evolution of cancer cells, and the copy number alterations of neighboring genes do not result in detrimental biological consequences [6,115,116]. The principle of collateral vulnerability provides an effective treatment strategy for cancers containing such genomic events.

When a co-deleted passenger gene is a member of a redundant multigene family for an essential housekeeping function, Muller et al. suggested that homozygous loss of these passenger genes creates collateral vulnerabilities unique to those cancer cells [117]. It is known that many essential functions in cells are carried out by several homologous genes with overlapping functions [118]. This functional redundancy allows cells to be viable after accidental loss of one homologue. However, complete loss of all homologues will cause cell death [119,120]. Therefore, when one redundant essential gene is deleted, inhibition of the remaining non-deleted homologue(s) results in the complete loss of the particular activity in cancer cells and eventually causes cell death, while normal cells remain intact (Figure 2). In their study, the glycolytic gene enolase 1 (ENO1) was found to be deleted in glioblastoma (GBM). However, cancer cells are viable owing to the expression of a redundant gene, ENO2 [117]. Knockdown of ENO2 selectively inhibits growth, survival and the tumorigenic potential of ENO1-deleted GBM cells. The enolase inhibitor phosphonoacetohydroxamate (PhAH), which targets ENO1 and ENO2, has much higher levels of toxicity on ENO1-deleted GBM cells in comparison with ENO1-intact GBM cells or normal astrocytes. In addition to ENO1, their analysis of TCGA data set of GBM identified eight homozygous deletions targeting genes involved in essential cell activities, including H6PD, KIF1B, NMNAT1, UBE4B, ACO1, KLHL9, PANK1, and KIF20B. These essential genes are co-deleted with tumor suppressor genes, such as PTEN and ARF. Pharmacologic inhibition of their redundant gene products will create therapeutic vulnerability in GBM. It is anticipated that many more passenger genes co-deleted with tumor suppressor genes will be identified as potential drug targets in human cancers. For instance, ARID1A, a key component in the SWI/SNF complex, is frequently mutated across a variety of human cancers. Using a broad screening approach, Helming and colleagues identified ARID1B, an ARID1A homolog that is mutually exclusive with ARID1A in SWI/SNF complexes, as a therapeutic target in treating ARID1A-mutant cancers [121]. Although ARID1A and ARID1B are frequently co-mutated in cancer, ARID1A-deficient cancers always retain at least one functional ARID1B allele, suggesting a unique functional dependence with one another. Loss of ARID1B in ARID1A-deficient cancer cells destabilizes SWI/SNF and impairs cell proliferation. Inhibiting ARID1B is a new approach to treat ARID1A-mutant cancers.

Figure 2. Homozygous co-deletion of passenger genes with a tumor suppressor gene generates therapeutic vulnerabilities in cancer.

Homozygous deletion of a tumor suppressor gene (TSG) often leads to loss of the neighboring essential gene (ESG1). This makes the tumor cells dependent on ESG2 that is functionally redundant with ESG1 in the same complex or pathway. Normal cells will survive treatment with a drug that inhibits ESG2, because of their high expression of ESG1. However, tumor cells only express ESG2, which renders them highly sensitive to the treatment.

Partial loss or inactivation of tumor suppressors is much more frequent in cancers than homozygous loss or complete inactivation of these genes [116]. Similarly, hemizygous loss of tumor suppressor genes often involves multiple neighboring genes that may not contribute to cancer development. The loss of neighboring essential genes renders cancer cells highly vulnerable to further suppression of those genes (Figure 3). To identify cancer specific vulnerabilities from those essential genes in the deletion region, Nijhawan and colleagues performed integrated analyses of genome-wide copy number and RNAi profiles, from which 56 genes were identified as potential drug targets [116]. Knockdown or inactivation of one gene inhibited the proliferation of cells harboring partial copy number loss of that gene. They named these identified genes as CYCLOPS (copy number alterations yielding cancer liabilities owing to partial loss) genes, which are enriched for spliceosome, proteasome, and ribosome components. The highest-ranked CYCLOPS candidate, PSMC2, is a key component of the 19S regulatory complex in the 26S proteasome, which is responsible for degradation of unfolded substrate proteins [122,123]. Normal cells express excess PSMC2, which forms a complex with PSMC1, PSMD2, and PSMD5 and acts as a reservoir protecting cells from PSMC2 suppression [116,124]. Cells harboring partial PSMC2 copy number loss lack this complex and are killed by PSMC2 suppression.

Figure 3. Hemizygous co-deletion of essential genes with a tumor suppressor gene generate therapeutic vulnerabilities in cancer.

Hemizygous deletion of a tumor suppressor gene (TSG) leads to partial loss of the neighboring essential gene (ESG). This genomic event results in decreased abundance of the ESG-encoding proteins in the cancer cells. Therefore, cancer cells are killed by much lower concentrations of the ESG-inhibiting drugs than normal cells that express higher levels of ESG from both copies of this gene.

A tremendous effort has been made to restore p53 activity in cancer therapies. However, no effective p53-based therapy has been successfully translated into clinical cancer treatment due to the complexity of p53 signaling. It is well documented that p53 level and activity are primarily controlled by post-translational modifications [80]. Indeed, our analysis of The Cancer Genome Atlas (TCGA) revealed that there is no correlation between p53 gene copy numbers and protein expression levels [125]. Consequently, identification of vulnerabilities conferred by TP53 deletion or mutation is a major challenge to target p53 aberrancy in human cancer. In our recent study, we demonstrated that genomic deletion of p53 frequently encompasses neighboring essential genes, rendering cancer cells with hemizygous p53 deletion vulnerable to further suppression of such genes [125]. POLR2A is identified as such a gene that is virtually co-deleted with p53 in human cancers. Hemizygous loss of p53/POLR2A occurs in 53% of colorectal cancers (CRC), 62% of breast cancers, 75% of ovarian cancers, and 41% of pancreatic cancers. The POLR2A gene encodes the catalytic subunit of RNA polymerase II complex, which is specifically inhibited by α-amanitin, a cyclic 8-aa peptide toxin found in the death cap mushroom (Amanita phalloides) [126]. POLR2A expression levels are tightly correlated with its gene copy numbers in the human cell. Normal cells with two alleles of POLR2A express excess POLR2A, protecting cells from POLR2A suppression, and one allele of POLR2A is sufficient to support cancer cell survival and proliferation. However, further suppression of POLR2A selectively inhibits proliferation, survival and tumorigenic potential of cancer cells with hemizygous TP53/POLR2A loss in a p53-independent manner, which creates a therapeutic window between normal and cancer cells. As a specific inhibitor for POLR2A, previous clinical applications of α-amanitin have been limited because of its liver toxicity [127]. Free α-amanitin causes apoptosis and necrosis of hepatocytes by interacting with the hepatocyte-specific transporting protein OATP1B3. However, α-amanitin is no longer a substrate for OATP1B3 when coupled to antibodies [128,129]. Therefore, α-amanitin-based antibody drug conjugates (ADCs) are highly effective therapeutic agents with significantly reduced toxicity. Our study has shown that low doses of α-amanitin-conjugated anti-EpCAM (Epithelial Cell Adhesion Molecule) antibody lead to complete tumor regression in murine models of human CRC with hemizygous deletion of POLR2A. These preclinical studies provide the foundation for future clinical trials [125].

Conclusions and outlook

Around 80% of detected cancer-related mutations are in tumor suppressor genes. However, targeting tumor suppressors has been long thought to be challenging or even unlikely in some cases. Despite many failed attempts in the past, scientists have been hoping to find novel approaches and techniques to restore the level and activity of tumor suppressors or the signaling pathways they initiate. In particular, advances in cancer genomics have greatly improved the prospects for new drug discovery and re-evaluation of existing drugs and compounds. In a new era of precision medicine, many efforts are anticipated to improve comprehensive assessment of various subsets of heterogeneous cancer by integrating advanced genome sequencing, new bioinformatic and biostatistical tools, and humanized preclinical models. Drug toxicity and resistance are key issues to address in drug identification and development. Moreover, it remains to be determined whether the targeted therapy can be effectively combined with immunotherapy, such as immune checkpoint blockage therapy. To this end, we will need to better understand what contributes to host immune response during tumor destructions. New techniques, including single circulating tumor cell sequencing, exosomal analysis, and advanced tumor imaging, will facilitate personalized medicine. These breakthroughs in each aspect will provide unprecedented opportunity for targeted cancer therapy.