Potential of the Synthetic Lethality Principle

See allHide authors and affiliations

Science  15 Nov 2013:
Vol. 342, Issue 6160, pp. 809-811
DOI: 10.1126/science.1244669

Most cancer mutations, including those causing a loss of function, are not directly “druggable” with conventional small-molecule drugs or biologicals, such as antibodies. Thus, despite our growing knowledge of mutations that drive cancer progression, there remains a frustrating gap in translating this information into the development of targeted treatments that kill only cancer cells. An approach that exploits a concept from genetics called “synthetic lethality” could provide a solution. But it has been over 15 years since that framework was proposed (1). Does the synthetic lethality principle still have the potential for treating cancer?

Synthetic lethality, first observed in the fruit fly Drosophila melanogaster almost a century ago, describes a phenomenon where only the simultaneous perturbation of two genes results in a deadly combination. Thus, cancer aberrations that are not readily targetable [e.g., tumor suppressor proteins such as retinblastoma protein 1 (RB1) and p53 (TP53); oncogenes such as RAS and c-MYC] could be indirectly exploited by inhibiting the product of another gene (24). The broader definition of synthetic lethality has also been referred to as “nononcogene addiction” or “induced essentiality” to distinguish it from its classical meaning in genetics. In the budding yeast Saccharomyces cerevisiae, most genes display numerous synthetic lethal interactions (5, 6), which may also apply to many human cancer genes. Furthermore, “passenger” mutations, which do not directly contribute to tumorigenesis, and even rewiring of cellular networks that give rise to a cancerous state, may also be exploited with the synthetic lethal principle. However, to date, only a single synthetic lethal interaction has shown therapeutic promise. Why have synthetic lethal therapies largely failed to deliver?

The proof of principle that the synthetic lethality concept is clinically translatable is the efficacy of drugs that target the single-strand DNA repair enzyme poly(ADP-ribose) polymerase (PARP) in tumors with mutations in the BRCA1 and BRCA2 genes (7). These genes encode tumor suppressor proteins that help repair damaged DNA. The remarkable ability of tumors to acquire resistance to PARP inhibitors by regaining BRCA function shows that PARP-targeting drugs act through a synthetic lethal mechanism (8). This finding triggered an intensive search for synthetic lethal drug targets akin to PARP. In particular, large-scale RNA interference screens (in which RNA molecules block the expression of specific genes) have led to a growing list of potential synthetic lethal gene targets. It is too soon to know if any of these new drug targets can be translated to the clinic. A major obstacle is that genetic and pharmacological perturbations do not always have the same functional outcome. This means that a genetic synthetic lethal relationship may never be realized pharmacologically. Despite this understanding, there is a growing concern that many synthetic lethal interactions are not easily transferred beyond the models used for screening, challenging synthetic lethality as a broadly applicable therapeutic concept. Indeed, only a few synthetic lethal drug targets are in clinical development, and none are as compelling as the BRCA-PARP paradigm.

Context and topology.

Identifying synthetic lethal interactions (when simultaneous mutations in two genes are lethal to the cell) must consider the context of their interactions, such as genetic and epigenetic variability, the microenvironment, and local systemic signals. Topological characteristics (strength, connectivity, degree, and redundancy) of human synthetic lethal interaction networks are also an important consideration, but are largely unknown.


A major obstacle to achieving synthetic lethal therapies is a lack of insight into the first principles that govern the phenomenon in cancer cells. For example, there is little understanding of how variability in genetics, epigenetics, systemic signals, and the microenvironment influences synthetic lethal interactions (see the figure). Too often, a line of investigation focuses on “what works” without considering the context or analyzing unexpected or even contradictory results. Phenotypes are rarely conserved across a panel of biologically diverse replicates (humans), indicating that context matters. This is especially important in cancer, where molecular heterogeneity (contextual variability) is greater than for any other disease. Indeed, delving into biological context can be highly informative and provide therapeutic leads. For instance, colon cancer cells with activating and oncogenic mutations in the gene BRAF fail to respond to BRAF inhibitors. However, these cancer cells display a feedback mechanism that induces signaling by the epidermal growth factor receptor, negating the effects of BRAF inhibition. This “bigger picture” view yielded a rationale to combine BRAF inhibitors with epidermal growth factor inhibitors, a solution that blocked growth of seemingly “drug-resistant” cancer cells (9, 10).

Synthetic lethal interactions may be no less context dependent than other cellular phenotypes. The opposite actually may be true. In yeast, single gene perturbations that lead to fitness defects (mutations in essential genes) are evolutionarily more conserved than defects induced by two mutations (11). As clonal evolution during tumorigenesis results in a divergence of cell states (through the progressive accumulation of genetic and epigenetic abnormalities in multiple and different genes), context dependencies may be even more pronounced for drugs that exploit synthetic lethality than for those that target “oncogene addiction,” in which only one gene that is required for maintaining malignancy is targeted.

The importance of context is also revealed in the comparison of genes that are differentially essential in two yeast strains (12). It was predicted that some of these conditional essential genes could be explained by mutations in synthetic lethal interactors in the affected strain. However, crossing the two strains and analyzing the frequency of viable spores showed that most of the conditional essential interactions are due to multiple genes acting in concert. This implies that synthetic interactions may also be susceptible to genetic modifiers and that quantitative differences in gene activity (between cell types, tumors, and individuals) can affect synthetic lethality. Moreover, stochastic variation in gene activity in the worm Caenorhabditis elegans can affect the penetrance of synthetic lethal interactions (13). This suggests that context will impact synthetic lethality in human cancer cells and argues for a systematic study of this phenomenon to guide more successful clinical translation. Such studies should focus on identifying co-occurring mutations in tumors that can modify the penetrance of synthetic lethal interactions. This can be achieved, for instance, by testing the synthetic lethal interaction in a large panel of well-characterized cancer cell lines or by using engineered isogenic cell lines that model the variability observed in patients in a defined genetic context. Although these cell culture studies would not directly address the role of the tumor microenviroment, any identified genetic modifiers could serve as biomarkers to predict which tumors are most likely to display the synthetic lethal interaction. Ultimately, performing a genome-wide screen for genetic modifiers that enhance or suppress a specific synthetic lethal interaction could reveal those that are least sensitive to contextual variability, and thus would be more likely to be of general use in the clinic. Results of such suppressor-enhancer screens from a variety of candidate synthetic lethal interactions may reveal more fundamental rules that govern context dependency.

There are several additional fundamental aspects of synthetic lethal interactions in human cells that have yet to be systematically explored. Some of these relate to network topology—the parameters that describe the connectivity structure of synthetic lethal interaction. For example, quantitative insights into the strength of synthetic lethality in human cells are lacking; yet this may determine the eventual therapeutic index—the concentration of a drug required for toxic effects divided by the concentration required for therapeutic effects—that can be achieved in patients. Such knowledge will be instructive for selecting the most promising synthetic lethal interactions for drug discovery and development (4). Indeed, compared to normal cells, cells lacking BRCA1/2 are almost three orders of magnitude more sensitive to PARP inhibitors. By contrast, most other described synthetic lethal interactions in human cells are much less striking, and some may not even conform to the strictest definition requiring that each single gene perturbation alone has no effect on cell viability.

Topological information may point to the best cellular processes for identifying synthetic lethal interactions. Which types of genes or cellular processes tend to display the most interactions (interconnectivity) may instruct functional genomics efforts. Screens in model organisms such as yeast and worm indicate that proteins involved in chromatin regulation such as histone deacetylases display the most frequent genetic interactions (5, 14). One explanation is that transcriptional regulation is well positioned for buffering perturbations by tuning the expression of multiple genes simultaneously. However, there are no experiments to corroborate this hypothesis in human cancer cells. Even for DNA synthesis and repair pathways that represent highly conserved functional modules (an obvious place to look for synthetic lethality), an inventory of synthetic lethal interactions in human cells has still not materialized.

Other largely uncharted topological areas concern the number of synthetic lethal connections between genes and their distribution and redundancy. The complexity of a human cell compared to a yeast cell may suggest that human cells display more redundancy, making them more resilient to perturbations and implying that synthetic lethal interactions would be less frequent. Current and next-generation genomics tools will help to answer these questions.

In the near term, the ability to perform personalized screens for synthetic lethal interactions on ex vivo tissue samples may provide clinically useful knowledge until the long-term goal of better understanding the biological rules can be achieved. Until a thorough understanding of synthetic interactions and the ability to assess their promise is in hand, their validation and translation will remain hit-and-miss. Recognizing the challenges facing gene therapy and immune therapy paved the way for moving from concept to clinical reality, and there is hope that learning the principles that govern synthetic lethal interactions in cancer will do the same.


View Abstract

Navigate This Article