|Home | About | Journals | Submit | Contact Us | Français|
Unique features of tumours that can be exploited by targeted therapies are a key focus of current cancer research. One such approach is known as synthetic lethality screening, which involves searching for genetic interactions of two mutations whereby the presence of either mutation alone has no effect on cell viability but the combination of the two mutations results in cell death. The presence of one of these mutations in cancer cells but not in normal cells can therefore create opportunities to selectively kill cancer cells by mimicking the effect of the second genetic mutation with targeted therapy. Here, we summarize strategies that can be used to identify synthetic lethal interactions for anticancer drug discovery, describe examples of such interactions that are currently being investigated in preclinical and clinical studies of targeted anticancer therapies, and discuss the challenges of realizing the full potential of such therapies.
Standard chemotherapies for cancer were initially discovered on the basis of their ability to kill rapidly dividing cells, and thus some of their common side effects — such as hair loss, nausea and immunosuppression — are due to the toxicity to rapidly dividing normal tissues1,2. With the aim of identifying therapies that have greater effectiveness and fewer side effects, cancer research in the past two decades has largely focused on discovering tumour-specific traits that might be exploited for selective targeting.
Many of the resultant targeted anticancer agents that have been discovered and investigated in this time affect cell signalling molecules — such as receptor tyrosine kinases — that have a key role in tumour growth and survival3,4. The development of such therapies is one of the most active areas of drug development, but so far only a few have demonstrated clinical efficacy and received regulatory approval. Nevertheless, these targeted therapies have increased the survival of patients with previously intractable cancers — including chronic myelogenous leukaemia5–9, non-small cell lung cancer10–13, pancreatic cancer 14,15, renal cell carcinomas (RCCs)16–25 and liver cancers26,27 — either as first-line therapies or in patients who have relapsed after standard chemotherapy. However, limitations of the first generation of targeted therapies — including the development of resistance, and on- and off-target toxicities — have become apparent28,29.
One anticancer drug discovery strategy that shows great promise in specifically targeting cancer cells that possess genetic mutations that are not present in normal cells is the exploitation of synthetic lethality30–33. This form of cell killing (also known as conditional genetics) takes its name from classical genetic studies in model organisms such as yeast34, and is based on the interaction of two genes that both contribute, often nonlinearly, to an essential process or processes35,36. When either gene is mutated alone, the cell is viable (FIG. 1a); however, the combination of mutations in both of these genes results in lethality (FIG. 1b). This process is referred to as synthetic lethality because cells with both gene mutations are not viable, and so it is not possible to directly isolate such cells. Nevertheless, various approaches can be used to evaluate and target potential synthetic lethal interactions, as discussed in the next section.
The interactions revealed by synthetic lethality studies can indicate a range of both predicted and unexpected connections. In the most conceptually straightforward scenario, two parallel pathways both contribute to an essential process. Consequently, disruption of a gene in one pathway is non-lethal, as the alternative pathway can sufficiently maintain the essential process, whereas disruption of both pathways is lethal to the cell. In cases in which substantial knowledge of a particular process exists, as in DNA damage repair, some of these synthetic lethal interactions can be predicted without the need for extensive screening37,38. Synthetic lethality, however, need not result from obvious parallel pathways, with some interactions arising from gene products within the same pathway or within the same protein complex. Other associations may include two divergent pathways that are both needed for a response to a cellular insult, or a pathway that is only connected to another pathway as a result of a gain-of-function oncogenic mutation. High-throughput screening may prove to be particularly useful in identifying these more complex, unpredictable interactions.
Synthetic lethal targeting of cancer cells could be therapeutically advantageous to targeting of cancer cells with standard agents, in that only the cancer cells with a specific genetic mutation are killed; that is, it uses a genotype-selective toxin rather than a nonspecific cytotoxin. Cells without the cancer-inducing genotype are unaffected by such targeting, as inhibition of the targeted gene product does not affect cell viability and only the combination of an endogenous gene mutation in the cancer cell and targeted gene inhibition with a small interfering RNA (siRNA) or a small molecule leads to cell death. Exploiting synthetic lethality therefore increases selectivity towards killing tumour cells and, consequently, it enhances the therapeutic index between the tumour and normal tissue.
Beyond the potential for an enhanced therapeutic index, synthetic lethality offers the opportunity to exploit targets that have proved to be challenging to therapeutically modulate by other strategies. Although it is common to attempt to disrupt gain-of-function mutations in oncogenes39–41, tumour growth in many cases is also driven by loss-of-function mutations in tumour suppressor genes. However, restoration of the function of tumour suppressor genes is not simple, as exemplified by the many attempts to restore their function by gene therapy42–44. The identification of synthetic lethal drugs and pathways could enable the exploitation of these gene mutations for selective targeting. In addition, not every oncogene proves to be directly tractable for pharmacological intervention. Here again, synthetic lethality may offer solutions by providing additional targets that take advantage of these driving mutations to provide a clinical benefit.
The utility of a synthetic lethality approach is further enhanced for several reasons. Specifically targeted agents based on synthetic lethality can be used as a form of monotherapy45, in combination with cytotoxic chemotherapy and/or radiotherapy, or in patients with relapsed disease46,47. Furthermore, in some cases, the effect of a synthetic lethal interaction will be enhanced when combined with a genotoxic stress. Agents that induce synthetic lethality could theoretically accentuate the efficacy of cytotoxic therapies at lower doses, thereby decreasing off-target side effects and, consequently, increasing the therapeutic indices of either chemotherapies or radiation treatment. Moreover, as tumour progression is a multi-step process and the driving mutations change in the different stages of tumour growth, synthetic lethality could target a range of temporal mutations that occur from tumorigenesis to metastatic dissemination. Perhaps the greatest benefit of synthetic lethality is its potential to treat metastatic diseases, for which there are currently few effective and selective options.
In this Review, we first consider the advantages and disadvantages of different types of synthetic lethality screens for the discovery of potential anticancer drugs. We then discuss several case studies that exemplify progress in the discovery and development of such agents, including poly(ADP-ribose) polymerase (PARP) inhibitors in patients with BRCA1 and/or BRCA2 mutations, and compounds that could be effective in patients with mutations in VHL, PTEN and RAS.
Hartwell, Friend and colleagues30 were the first to propose the concept of using synthetic lethality screening to identify new anticancer drugs. Recognizing that one of the limiting steps impeding drug discovery was the identification of tumour-selective characteristics, they suggested that loss-of-function mutations — such as those found in DNA repair genes or tumour suppressor genes — could be exploited. In the ideal setting, the first mutation would be a cancer-driving defect and highly conserved evolutionarily from model organisms to humans. Thus, by screening in yeast, synthetic lethal interactions could be identified, either by candidate selection or by genome-wide screening. Choosing putative targets requires both prior knowledge of involved pathways and the ability to generate specific mutations to investigate the effects of various combinations of mutations on these pathways. By contrast, genome-wide screening is a blind, unbiased search that requires large-scale genetic screening technology.
As proof-of-concept, Hartwell et al. 30 performed a small-scale screen of a panel of 70 different isogenic strains from budding yeast with deletions in DNA damage response genes against US Food and Drug Administration (FDA)-approved chemotherapies. For validation of the feasibility of such a screen, the focus was on genetic instability as a basis for drug discovery. The rationale was that genetic instability is a common feature of many tumours, and that the genetic changes that underlie this genetic instability of tumour cells — in particular, defects in DNA damage response and repair pathways — could make tumour cells more sensitive to the effects of some drugs than normal cells. Hartwell, Friend and colleagues30 were able to determine the drug sensitivities of two anticancer agents: cisplatin and mitoxantrone. Cisplatin demonstrated increased specificity for yeast strains that were defective in post-replication repair, whereas mitoxantrone — which functions as a topoisomerase II poison — resulted in increased sensitivity of yeast strains that were defective in double-stranded DNA break repair. This work demonstrated the feasibility of using large genetic screens to identify synthetic lethal interactions.
The value and applicability of synthetic lethality in the context of mammalian cells, especially in the cancer setting, is now more fully recognized. Efforts are no longer limited to model organisms nor are they limited to known essential cancer-driving pathways, such as those involved in genomic instability or DNA damage repair. Considerable technological advances have made it possible to screen for genes involved in synthetic lethal interactions in a mammalian setting (FIG. 2). Notably, the advent of libraries of either siRNAs or short hairpin RNAs (shRNAs), as well as combinatorial and diversity-oriented libraries of small molecules, enables genome-wide investigation of specific mutations in a rapid manner in mammalian cells.
There are fundamental similarities between screening either siRNA/shRNA libraries or small-molecule compounds to identify synthetic lethal interactions — both approaches frequently use matched, isogenic lines (in which an essential cancer gene has been identified) and a functional readout to assess whether an agent is cytotoxic, cytostatic or has no effect (FIG. 2). Both types of unbiased screens can also reveal unexpected connections that can directly and indirectly advance drug development efforts as well as basic research into our understanding of cancer biology. However, these two types of screens also have distinct although not necessarily mutually exclusive goals. The screening of RNA interference (RNAi)-based libraries can identify genes that are important in a pathway context and thus provide a better understanding of the fundamental biology behind interactions. By contrast, the goal of screening a small-molecule library is typically to obtain candidate compounds for the treatment of a given cancer genotype. Along with differing aims, there are advantages and disadvantages of high-throughput screening of either RNAi-based libraries or small-molecule compound libraries to identify synthetic lethal interactions (BOX 1).
There are certain considerations when screening for synthetic lethal interactions using RNA interference (RNAi) libraries or small-molecule compound libraries. Below, we list some advantages and disadvantages of these two approaches for the identification of synthetic lethal interactions.
RNAi-based screens allow for the direct discovery of unknown gene–gene interactions and pathways. However, the identification of such interactions and pathways does not necessarily lead readily to potential therapeutic candidates.
One approach for RNAi-based genome-wide screening requires a reverse transfection step, in which cells are plated onto pre-seeded plates containing the RNAi library along with a transfection reagent. Following incubation to allow for expression, the cells are then assayed for changes in viability. Off-target toxicity that is inherent to RNAi-based screening may be a source of false readouts. Alternative approaches for screening RNAi-based libraries include using plasmid vectors, in which cells are transfected with DNA encoding shRNA48. Viral methods, such as infection with retroviruses containing targeting sequences, can also be used for conditional inactivation of genes49. These plasmid-based approaches offer another advantage — the inclusion of a DNA ‘bar code’, which is a unique sequence for each shRNA-encoding plasmid that can be amplified from a mixed population. The abundance of each bar code within a pool under different conditions reveals the effect of an individual shRNA on survival and growth without the need to assay each plasmid individually.
Another important consideration is that the primary hits obtained from an RNAi library screen require a second screen to identify the agents that target the identified genes or their products50. If the synthetic lethal interactor is a known gene, then compounds to inhibit its activity may already exist51. Although knowing what the genetic target is for a synthetic lethal interaction could indicate the mechanism of action and aid in the development of small-molecule therapeutics, focusing on the known function of the target gene could also be misleading, as other mechanisms that are independent of the known function of the gene could be responsible for the synthetic lethality. Identifying a pathway involved in synthetic lethal interactions also provides additional targets for small-molecule inhibitors.
In contrast to RNAi libraries, identification of small molecules that demonstrate synthetic lethality in a screen directly provides potential candidates for optimization into lead compounds (for example, by improving potency or pharmacokinetic properties) without additional steps. In addition, the identification of the target of the small molecule can be valuable in structural modelling or for investigating structure–activity relationships to generate additional analogues with improved characteristics. However, this can require substantial additional effort compared to the RNAi-based screens for which the target is already known. A second screen using RNAi may be necessary to reveal the target (or targets).
It should be noted that using isogenic cell lines is not without challenges and is not the only approach that can be taken for synthetic lethality screens. Although this approach is technically and conceptually attractive for identifying synthetic lethal interactions with common, known cancer mutations, the single genetic variation being studied may not actually be the only difference in ‘isogenic cells’, which is a confounding factor in these types of studies. Rather, genetic drift between pairs of isogenic cell lines may result in multiple differences that can alter responses to RNAi or drug treatment. The problem of genetic drift may be especially acute when the mutation of interest results in a defect in DNA repair or genomic instability.
An alternative to isogenic lines has been established by Canaani and colleagues52–54; in this strategy, a single human cancer cell line that is deficient in a gene of interest is used. Complementation of this gene is provided by a low-copy unstable episome expressing this gene. In the context of a drug or RNAi screen, retention of this episome is selected under synthetic lethal conditions, thus revealing novel interactions. Although isogenic lines often provide a valuable avenue for synthetic lethality screens, this work demonstrates that other approaches also have distinct advantages.
To date, synthetic lethality screens have focused on specific, fixed genetic mutations and ignored transient yet unique features that can also be exploited. Conditional synthetic lethality screening demonstrates great potential by using interactions based on temporary situations to further increase the therapeutic index and the selectivity for cancer cells. Conditional synthetic lethality can develop in several different contexts — for example, in response to ionizing radiation, cytotoxic chemotherapeutic agents, or changes in the cellular microenvironment. Recent studies by Bindra et al. 55 and Chan et al. 56 have demonstrated that tumour hypoxia decreases the expression of homologous recombination proteins such as the DNA repair protein RAD51. Therefore, by suppressing the expression of DNA repair proteins, hypoxia conditionally transforms cells into a recombinational-deficient state and consequently they are sensitive to PARP inhibitors57.
The impact of this conditional state of recombination deficiency induced by hypoxia should be common to most solid tumours, and could potentially be enhanced by agents that selectively increase tumour hypoxia by altering their metabolism. For example, treatment of transplanted tumours in mice with dichloroacetate (DCA) increases pyruvate consumption in the mitochondria and total oxygen consumption, which increases tumour hypoxia58. As DCA is already in clinical trials, it represents a practical approach for increasing tumour hypoxia, decreasing the expression of recombination proteins and sensitizing tumour cells to agents such as PARP inhibitors that will induce synthetic lethality59. This approach represents one example of how conditional synthetic lethal interactions can potentially be manipulated and exploited for cancer therapy (FIG. 3).
In this section, selected case studies that exemplify the discovery and development of anticancer agents using synthetic lethality approaches are discussed.
Inhibitors of PARP are one of the first classes of small-molecule compounds that have been identified to interact in a synthetic lethal manner with mutations in the genes encoding proteins involved in DNA repair60–64. The tumour suppressor proteins BRCA1 and BRCA2 are important for the repair of double-stranded DNA breaks via homologous recombination, and mutations in their genes are associated with hereditary forms of breast and ovarian cancers65–68. PARP, however, functions as a sensor to recognize and recruit DNA repair proteins to the sites of single-stranded DNA breaks69. Given that unrepaired single-stranded DNA breaks result in stalled replication forks, and that one of the major roles of homologous recombination is to repair such stalled replication forks, tumours with impaired BRCA1 or BRCA2 function and, consequently, compromised homologous recombination function, could be sensitized to inhibitors of enzymes involved in single-stranded DNA break repair, such as PARP inhibitors70.
Indeed, inhibition of PARP with such compounds was discovered to result in synthetic lethal interactions with mutations in BRCA1 and BRCA2 (REFS 37,38). Bryant et al.37 and Farmer et al.38 demonstrated that inhibition of PARP1 in conjunction with either BRCA1 or BRCA2 deficiency resulted in a failure to repair double-stranded DNA breaks and recombination lesions, leading to apoptosis. Cells that were deficient in BRCA1 or BRCA2 or cells in which BRCA1 or BRCA2 were depleted by shRNAs were sensitive to PARP inhibition, both pharmacologically and genetically. Furthermore, BRCA2-deficient xenografts were also sensitive to PARP inhibitors. The conditional genetic interaction between BRCA deficiency and PARP inhibition provides proof-of-principle that synthetic lethal relationships can be identified for targeting homologous recombination and other DNA repair pathways (BOX 2).
Poly(ADP-ribose) polymerase (PARP) inhibitors are the most clinically developed synthetic lethal targeting agents to date for killing cells with BRCA mutations. BRCA is normally involved in homologous recombinational repair of double-stranded DNA breaks, whereas PARP has a role in single-stranded DNA damage repair (base excision repair). Impairment of a combination of single- and double-stranded DNA break repair should result in synthetic lethality.
The combination of PARP inhibition and loss of BRCA1 or BRCA2 function represents an exciting approach to specifically exploit DNA repair deficiencies in tumour cells. Based on these findings, PARP inhibitors are now being evaluated in clinical trials in patients with breast and ovarian cancers that are characterized by BRCA1 or BRCA2 mutations46,71,72. In addition to breast and ovarian cancers, PARP inhibitors are also being developed by a number of pharmaceutical companies and are being tested clinically in a variety of tumour types (TABLES 1,,22).
As well as potentially having potent activity as single agents in patients with defective BRCA1 or BRCA2 function73, sensitizing tumour cells by targeting DNA repair deficiencies with PARP inhibitors could also allow the use of lower doses of existing cytotoxic drugs or ionizing radiation and, consequently, reduced side effects74,75. For example, standard DNA-damaging drugs for which the resultant DNA lesions are repaired by recombination repair, such as mitomycin C or cisplatin, are ideal candidates for combination therapy with PARP inhibitors in BRCA1 or BRCA2 recombination-deficient cells76–81. Several clinical trials are examining the efficacy of PARP inhibitors with various DNA-damaging agents (TABLES 1,,2).2). Interestingly, Fong et al.82 demonstrated that PARP inhibitors were effective in patients with BRCA1- and BRCA2-mutant ovarian cancers. More importantly, response to PARP inhibition was directly correlated to sensitivity to platinum-based therapies82,83. These findings demonstrate that synthetic lethal interactions can be used to predict clinical efficacy and thus determine which patients are more likely to have a durable response to a particular treatment.
Although BRCA1 and BRCA2 mutations and their impact on homologous recombination have been well documented in past studies, further studies should be carried out to explore the possibility that tumour cells may have other defects in DNA repair pathways, which could be exploited in a similar manner84,85. For example, although DNA repair pathways are impaired in BRCA2-deficient cells, Feng et al.86 determined that RAD52 mediated an alternative homologous recombinational repair pathway. Depletion of RAD52 substantially reduced homologous recombination and increased cell death independently of BRCA2. However, the combination of RAD52 and BRCA2 deficiencies resulted in a synthetically lethal interaction, demonstrating that targeting alternative homologous recombination pathways is an effective strategy for targeting tumour cells.
To screen for patients who have repair defects and would benefit from this form of targeted therapy, an assay could be used to detect decreased DNA repair protein foci formation in biopsy specimens that have been irradiated ex vivo87. For example, immunofluorescence of the DNA repair proteins BRCA1, FANCD2 or RAD51 (REF. 87) would allow for the detection of tumours with decreased expression of DNA repair proteins that may not necessarily be due to mutations in DNA repair genes, which is a rare occurrence in tumours. Defective or decreased DNA recombination activity should be equally sufficient for synthetic lethal targeting as a therapeutic strategy.
Recently, Martin et al.88,89 identified a pair of potential synthetic lethal interactions with DNA mismatch repair proteins. In a cell line and in several patient samples, MSH2-deficiency correlated with an increase in DNA polymerase β (Pol β), whereas deficiency in MLH1 correlated with an increase in Pol γ88. MSH2 and MLH1 are two genes that encode DNA mismatch repair proteins and are associated with hereditary nonpolyposis colorectal carcinomas through their DNA mismatch repair activities90–94. Pol β95–97 and Pol γ98,99 are two DNA polymerases that are involved in base excision repair. Silencing of either Pol β or Pol γ in an MSH2-negative or MLH1-negative cell line increased nuclear or mitochondrial 8-oxoguanine (8-oxoG) levels, respectively, which indicates a failure to repair oxidized nucleotides.
In a separate report, Martin et al.89 also determined that depletion of PTEN-induced putative kinase 1 (PINK1) and other mitochondrial kinases increased the production of reactive oxygen species, as well as formation of 8-oxoG. Accumulation of the oxidized base,8-oxoG, causes DNA base lesions through incorporation of either cytosine or adenine bases, which can cause GC→TA transversions. Base excision repair is one of the primary methods of dealing with 8-oxoG lesions. Thus, simultaneous inhibition of mismatch repair and DNA polymerase proteins leads to an accumulation of DNA damage by interfering with base excision repair, which results in cell death.
As MLH1 and MSH2 are defective in a number of nonhereditary colorectal cancers, Martin and colleagues propose that Pol β inhibitors could be used for a defined period of time rather than for the chronic treatment of colon carcinoma to limit the development of drug resistance. Furthermore, distinct treatments with a given end point would also prevent the accumulation of additional mutations through inhibition of DNA repair pathways. This study89 highlights that although synthetic lethal interactions may be readily identified, target identification plays only a small part in developing targeted therapies.
VHL is a tumour suppressor gene that follows the Knudson two-hit hypothesis and is associated with a tumour-prone syndrome that is characterized by a unique distribution of tumours, including haemangioblastomas of the central nervous system, retinal angiomas, pancreatic islet cell tumours, phaeochromocytomas of the adrenal glands and renal tumours100,101. In addition, it is estimated that VHL mutation or gene silencing via methylation accounts for over 80% of spontaneous and hereditary kidney tumours102–104. Renal tumours are particularly intractable as they are difficult to diagnose because they do not generally present with clinical symptoms until after they have metastasized to distant sites105. RCC has also been reported to be resistant to both radiation therapy and standard chemotherapeutic agents106. Owing to the strong relationship between loss of VHL function and the development of RCC107, and the lack of curative treatments, renal cancer cell lines that contain VHL mutations are attractive tools for identifying agents that function in a synthetically lethal manner.
High-throughput screening of shRNA libraries and small-molecule compound libraries have both been used to uncover genetic interactions with mutant VHL108,109. Bommi-Reddy et al.108 used 100 shRNA vectors directed against 88 different kinases to identify those that could inhibit proliferation of cells lacking VHL function. This study found three kinases that were capable of preferentially impairing growth of VHL-mutant cells in a dose-dependent manner: cyclin-dependent kinase 6 (CDK6), hepatocyte growth factor receptor (also known as MET) and dual specificity mitogen-activated protein kinase kinase (MAP2K1). Furthermore, pharmacological inhibition of CDK4 and/or CDK6 reduced the viability of VHL-mutant cells, providing proof-of-principle that screening of shRNA panels can detect synthetic lethal interactions in tumour cells.
In a separate study by Turcotte et al.109, in which 64,000 compounds were screened using a fluorescent cell-based growth assay, one compound — STF-62247 — was found to be selectively cytotoxic to cells lacking VHL. This candidate was then confirmed and validated in more stringent assays, such as clonogenic survival curves. Treatment of VHL-mutant cells with STF-62247 induced autophagy109. The yeast genome deletion pool was used to identify the target of STF-62247; this revealed a previously unknown relationship between Golgi trafficking, autophagy and the loss of VHL function. This study emphasizes the advantages and disadvantages of a small-molecule screen110,111.
From a clinical perspective, although there have recently been important advances in the treatment of RCC with drugs that target vascular endothelial growth factor signalling (which can be upregulated by loss of VHL function) and mammalian target of rapamycin (mTOR), this study also highlights a promising opportunity to improve the effectiveness of strategies to treat RCC by more directly exploiting the presence of the underlying VHL mutations. The recent identification of frequent mutations in the SWI/SNF chromatin remodelling complex gene PBRM1 offers another opportunity for targeting RCC112.
In addition to VHL and BRCA, other loss-of-function tumour suppressor mutations have been examined for synthetic lethal interactions. Several studies have demonstrated a synthetic lethal interaction between mTOR inhibition and PTEN loss113–115. PTEN is a tumour suppressor gene that is frequently mutated in both hereditary and spontaneous cancers. Loss of PTEN increases activation of AKT, which in turn increases proliferation and protein synthesis, and decreases apoptosis. Treatment with CCI-779 (also known as temsirolimus (Torisel; Pfizer)), an mTOR inhibitor, was selectively toxic to PTEN-deficient cell lines as well as PTEN-heterozygous and PTEN-homozygous mouse tumours113,114. CCI-779, an analogue of rapamycin, is a particularly attractive therapeutic compound because it already has FDA approval for the treatment of RCC (see above). CCI-779 decreased cell proliferation and cell size, while increasing apoptosis by inhibiting activation of both AKT and ribosomal protein S6 kinase.
Thomas and colleagues115 further demonstrated that growth arrest caused by CCI-779 correlated with a block in translation of hypoxia-inducible factor mRNA in VHL-deficient cells. Their results suggest that synthetic lethal interactions contribute to the clinical efficacy of some current drugs and, importantly, that an understanding of these interactions can guide their application to those patients who have mutations in their tumours that make them more responsive to these drugs than other patients.
These studies have focused on clinically approved compounds. However, large-scale synthetic lethal screening based on loss of PTEN is complicated by the fact that restoration or overexpression of PTEN to generate matched isogenic cell lines results in cell cycle arrest. Using PTEN-null cells in a cell-based assay of forkhead box protein O1A (FOXO1A) nuclear localization, Kau et al.116 were able to circumvent this limitation by examining a downstream consequence of PTEN loss. In cells lacking PTEN, FOXO1A is inactivated owing to its mislocalization to the cytoplasm, but redirecting FOXO1A to the nucleus can reverse the promotion of tumour growth caused by loss of PTEN. Kau et al. therefore screened for inhibitors of FOXO1A nuclear export, and they found several classes of compounds, including general nuclear export inhibitors and antagonists of calmodulin signalling.
Shen et al.117 recognized that one of the functions of PTEN is to maintain genomic stability by repairing double-stranded DNA breaks via homologous recombination. Thus, in addition to breast and ovarian cancers with BRCA mutations, a variety of tumour types, including endometrial cancers and glioblastomas, should also be considered in clinical studies with PARP inhibitors118–120. These studies on the loss of PTEN also illustrate how downstream functions of genes of interest can be used in the development of synthetic lethality strategies.
Although the synthetically lethal compounds that have progressed furthest in the clinic so far target loss-of-function mutations in tumour suppressor genes, gain-of-function mutations in oncogenes such as RAS can also be targeted. It is estimated that a quarter of all cancers have activating RAS mutations, with half of some specific cancers, such as colorectal carcinoma, having gain-of-function RAS mutations121. A number of synthetic lethality screens with activated RAS have been undertaken to circumvent the pharmacological intractability of oncogenic RAS122,123. Torrance and colleagues124 demonstrated a proof-of-principle synthetic lethality screen using isogenically matched cell lines with and without the mutant KRAS oncogene. Colon carcinoma cell lines containing the mutant KRAS were engineered to express yellow fluorescent protein, whereas the matched cell line in which the mutant KRAS was eliminated by homologous recombination was engineered to express a blue fluorescent protein. These two matched cell lines were co-cultured and screened against approximately 30,000 compounds to identify small molecules that were selectively cytotoxic to cells containing mutant KRAS; this resulted in the identification of four selective compounds. Two previously known RAS pathway inhibitors with unknown mechanisms of action — demethoxyviridin (a wortmannin analogue) and mithramycin — were revealed, thus indicating that this screening method was feasible. Two compounds — triphenyltetrazolium (TPT) and a sulphinylcytidine derivative (SCD) — were also identified as novel RAS pathway inhibitors in this screen.
The authors performed a structural homology search of the two libraries they screened and found several small molecules that were structurally similar to TPT but were not identified in the screen. These compounds did not inhibit the growth of cells with mutant KRAS. Moreover, a known p38 mitogen-activated protein kinase (MAPK) inhibitor — SB203580 — which is structurally similar to TPT, had no selectivity for mutant RAS cells and, likewise, TPT had no effect on p38 MAPK. SCD was further analysed, as it inhibited growth in several additional isogenic cell lines containing mutant KRAS. Animals bearing KRASG13D-mutant tumours showed a reduction in tumour growth when they were treated with SCD, thus demonstrating that a cytosine nucleoside analogue could potentially be used to treat tumours with mutant RAS. The identification of these compounds could provide the basis for synthesizing more specific compounds through the optimization and analysis of structure–activity relationships, although this could be hampered by the lack of knowledge on how these compounds might function, which would require further investigation to identify their mechanism of action.
In an independent screen, Sarthy et al.125 used an siRNA library to identify genes with synthetic lethal interactions with KRAS in the same colon carcinoma cell line. This study found that inhibiting expression of baculoviral IAP repeat-containing protein 5 (BIRC5; also known as survivin) in cells with an activated KRAS oncogene resulted in lower survival rates compared with isogenic cells in which KRAS was inhibited by siRNA. However, although this illustrates the value of genetic screening in identifying the importance of a particular genetic interaction, genetic screening does not necessarily readily indicate a therapeutically viable option with which to progress directly, given that antisense oligonucleotides and siRNA-based treatments are currently limited by delivery methods.
Using a genetic mouse tumour model of non-small cell lung carcinoma, Puyol et al.126 identified a synthetic lethal interaction between KRAS and CDK4. Ablation of CDK4 but not CDK2 or CDK6 reduced the number and grade of KRASG12V-induced lung tumours. Genetic elimination or pharmacological inhibition of CDK4 slowed tumour growth by inhibiting proliferation and inducing cell senescence. This study provides genetic evidence to explain why CDK inhibitors alone have limited efficacy127 and suggests that tumours with specific KRAS mutations may benefit from specific CDK inhibition.
Oncogenic KRAS has also been shown to increase activation of the ataxia telangiectasia and RAD3-related protein (ATR)–CHK1 pathway and cause a subsequent increase in genomic instability128. Inhibition of ATR in the context of oncogenic RAS had two distinct outcomes depending on the degree of impairment. Haploinsufficiency of ATR, or a modest decrease, increased tumorigenesis. A further increase in ATR inhibition caused a decrease in cell viability and a shift to synthetic lethality, thus demonstrating the delicate balance that must be struck when targeting essential pathways.
More recent large-scale genomic screening of oncogenic RAS reveals the complexities in undertaking a synthetic lethality screen122,123. Using two different approaches, Luo et al.51 and Scholl et al.50 identified pathways that could be inhibited in KRAS-mutant cells, which resulted in decreased cell viability. Luo et al. used isogenic cells with either mutant or wild-type RAS, using a second pair of isogenic cells for the screening validation. Their initial screen did not identify an obvious ‘hit’; rather, subsequent computational analysis indicated that the mitotic machinery or the proteasome had a synthetic lethal interaction with mutant RAS. The findings of this screen predicted and determined that cells with mutant RAS are sensitive to agents that target mitotic spindle function, such as the taxanes, or agents that inhibit the proteasome, such as bortezomib (Velcade; Millennium Pharmaceuticals/Centocor Ortho Biotech).
As this was an RNAi library screen, the potential for a cancer therapeutic relied on developed or known compounds. For example, the authors used a previously identified small-molecule inhibitor of polo-like kinase 1 (PLK1)129, a mitotic kinase, to demonstrate that mitotic pertubation resulted in death of cells with mutant RAS. However, Scholl et al.50 identified a single gene, STK33 (REF 130), for which there was no developed inhibitor. Rather, this RNAi screen identified a previously unknown interaction between STK33 and mutated KRAS. Similarly, the genetic interplay between nuclear factor-κB (NF-κB) and oncogenic KRAS could potentially be targeted, as upstream regulators of NF-κB have been shown to be essential for KRAS expression131 and transformation132, whereas NF-κB signalling is essential for lung adenocarcinoma development133.
In another study, Dolma et al. screened two chemical libraries for synthetic lethal interactions with oncogenic HRASV12 in engineered tumour cells derived from primary fibroblasts134. This screen identified a novel compound, erastin, which selectively killed tumorigenic cells expressing HRAS and simian virus 40 (SV40) small T oncoprotein through non-apoptotic cell death. In order to determine the mechanism of genotype-specific cyto-toxicity, two subsequent approaches were taken: a suppressor screen with a battery of bioactive compounds of previously determined function and a functional screen based on affinity purification135. The suppressor screen utilized 2,000 biologically active compounds with known targets and/or functions. Several antioxidants (for example, α-tocopherol, butylated hydroxytoluene and β-carotene) prevented erastin-induced death. Immobilization of an erastin analogue to a solid support followed by mass spectrometry revealed that erastin interacted with specific isoforms of voltage-dependent anion channels (VDACs). These two techniques showed that mitochondrial VDACs are responsible for an oxidative, iron-dependent, non-apoptotic cell death in response to erastin. More importantly, these studies identified a synthetic lethal interaction between VDAC and oncogenic HRASV12, the more common KRAS mutants and a promising lead compound.
Using a similar cell system of immortalized and transformed fibroblasts, Wang and colleagues136 identified agonists of DR5 — a receptor for the tumour necrosis factor-related apoptosis-inducing ligand (TRAIL) — that selectively induced apoptosis in cells overexpressing the MYC oncogene. Non-transformed primary cells were retrovirally infected to overexpress the MYC oncogene in order to understand the conditions in which MYC sensitizes cells to apoptosis. These cells were subjected to diverse stimuli. Many of these stimuli, such as DNA-damaging agents or serum starvation, decreased cell proliferation and cell cycle arrest. By contrast, TRAIL receptor activation decreased cell viability in cells that overexpressed MYC by inducing apoptosis. More specifically, using antagonistic monoclonal antibodies against different TRAIL death receptors, the authors showed that activation of DR5 resulted in MYC-selective apoptotic cell death. Treatment of MYC-overexpressing cells with an Aurora B kinase inhibitor, VX-680, initially kills cells by inducing apoptosis, but prolonged killing can occur via the autophagic cell death pathway137. These findings demonstrate that gain-of-function oncogenic mutations can be indirectly targeted by a synthetic lethality approach (TABLE 3).
The search for targeted therapies is a burgeoning field and synthetic lethal interactions represent a very promising means of selectively killing tumour cells. Furthermore, they could allow the exploitation of differences between tumour cells and normal cells that have previously been considered to be pharmacologically intractable. Future synthetic lethality screens could also be used to investigate mechanisms to exploit epigenetic phenomena, the tumour microenvironment and stromal–tumour interactions. Future conditional synthetic lethality screens should further investigate tumour hypoxia as a common physiological feature of solid tumours that can be targeted therapeutically. Other screens could focus on identifying genetic interactions to enhance radiotherapy or current cytotoxic chemotherapies. Overall, synthetic lethality screening could allow the field of oncology to shift its focus from ‘first, hasten to help’ back to the historical guiding medical principle of ‘first, not to harm’.
This work was supported by the following grants from the US National Cancer Institute: NCI-CA-67166 (A.J.G), NCI-CA-88480 (A.J.G) and NCI-CA-123823 (D.A.C). It was also supported by a grant from Action to Cure Kidney Cancer (A.J.G.). We apologize to colleagues whose work we failed to cite.
Competing interests statement
The authors declare competing financial interests: see Web version for details.
Denise A. Chan’s homepage: http://radonc.ucsf.edu/faculty/radiobiologists/chan_d.html
Amato J. Giaccia’s homepage: http://med.stanford.edu/profiles/Amato_Giaccia
ALL LINKS ARE ACTIVE IN THE ONLINE PDF