Synthetic Lethal Networks for Precision Oncology: Promises and Pitfalls

Clinically oriented prediction of patient response to targeted and immunotherapies from the tumor transcriptome

BACKGROUND

Precision oncology is gradually advancing into mainstream clinical practice, demonstrating significant survival benefits. However, eligibility and response rates remain limited in many cases, calling for better predictive biomarkers.

METHODS

We present ENLIGHT, a transcriptomics-based computational approach that identifies clinically relevant genetic interactions and uses them to predict a patient's response to a variety of therapies in multiple cancer types without training on previous treatment response data. We study ENLIGHT in two translationally oriented scenarios: personalized oncology (PO), aimed at prioritizing treatments for a single patient, and clinical trial design (CTD), selecting the most likely responders in a patient cohort.

FINDINGS

Evaluating ENLIGHT's performance on 21 blinded clinical trial datasets in the PO setting, we show that it can effectively predict a patient's treatment response across multiple therapies and cancer types. Its prediction accuracy is better than previously published transcriptomics-based signatures and is comparable with that of supervised predictors developed for specific indications and drugs. In combination with the interferon-γ signature, ENLIGHT achieves an odds ratio larger than 4 in predicting response to immune checkpoint therapy. In the CTD scenario, ENLIGHT can potentially enhance clinical trial success for immunotherapies and other monoclonal antibodies by excluding non-responders while overall achieving more than 90% of the response rate attainable under an optimal exclusion strategy.

CONCLUSIONS

ENLIGHT demonstrably enhances the ability to predict therapeutic response across multiple cancer types from the bulk tumor transcriptome.

FUNDING

This research was supported in part by the Intramural Research Program, NIH and by the Israeli Innovation Authority.

Artificial intelligence, molecular subtyping, biomarkers, and precision oncology

A targeted cancer therapy is only useful if there is a way to accurately identify the tumors that are susceptible to that therapy. Thus rapid expansion in the number of available targeted cancer treatments has been accompanied by a robust effort to subdivide the traditional histological and anatomical tumor classifications into molecularly defined subtypes. This review highlights the history of the paired evolution of targeted therapies and biomarkers, reviews currently used methods for subtype identification, and discusses challenges to the implementation of precision oncology as well as possible solutions.

Conditional transcriptional relationships may serve as cancer prognostic markers

BACKGROUND

While most differential coexpression (DC) methods are bound to quantify a single correlation value for a gene pair across multiple samples, a newly devised approach under the name Correlation by Individual Level Product (CILP) revolutionarily projects the summary correlation value to individual product correlation values for separate samples. CILP greatly widened DC analysis opportunities by allowing integration of non-compromised statistical methods.

METHODS

Here, we performed a study to verify our hypothesis that conditional relationships, i.e., gene pairs of remarkable differential coexpression, may be sought as quantitative prognostic markers for human cancers. Alongside the seeking of prognostic gene links in a pan-cancer setting, we also examined whether a trend of global expression correlation loss appeared in a wide panel of cancer types and revisited the controversial subject of mutual relationship between the DE approach and the DC approach.

RESULTS

By integrating CILP with classical univariate survival analysis, we identified up to 244 conditional gene links as potential prognostic markers in five cancer types. In particular, five prognostic gene links for kidney renal papillary cell carcinoma tended to condense around cancer gene ESPL1, and the transcriptional synchrony between ESPL1 and PTTG1 tended to be elevated in patients of adverse prognosis. In addition, we extended the observation of global trend of correlation loss in more than ten cancer types and empirically proved DC analysis results were independent of gene differential expression in five cancer types.

CONCLUSIONS

Combining the power of CILP and the classical survival analysis, we successfully fetched conditional transcriptional relationships that conferred prognosis power for five cancer types. Despite a general trend of global correlation loss in tumor transcriptomes, most of these prognosis conditional links demonstrated stronger expression correlation in tumors, and their stronger coexpression was associated with poor survival.

Paralogous synthetic lethality underlies genetic dependencies of the cancer-mutated gene STAG2

STAG2, a component of the mitotically essential cohesin complex, is highly mutated in several different tumour types, including glioblastoma and bladder cancer. Whereas cohesin has roles in many cancer-related pathways, such as chromosome instability, DNA repair and gene expression, the complex nature of cohesin function has made it difficult to determine how STAG2 loss might either promote tumorigenesis or be leveraged therapeutically across divergent cancer types. Here, we have performed whole-genome CRISPR-Cas9 screens for STAG2-dependent genetic interactions in three distinct cellular backgrounds. Surprisingly, STAG1, the paralog of STAG2, was the only negative genetic interaction that was shared across all three backgrounds. We also uncovered a paralogous synthetic lethal mechanism behind a genetic interaction between STAG2 and the iron regulatory gene IREB2 Finally, investigation of an unusually strong context-dependent genetic interaction in HAP1 cells revealed factors that could be important for alleviating cohesin loading stress. Together, our results reveal new facets of STAG2 and cohesin function across a variety of genetic contexts.

A systematic analysis of genetic interactions and their underlying biology in childhood cancer

Childhood cancer is a major cause of child death in developed countries. Genetic interactions between mutated genes play an important role in cancer development. They can be detected by searching for pairs of mutated genes that co-occur more (or less) often than expected. Co-occurrence suggests a cooperative role in cancer development, while mutual exclusivity points to synthetic lethality, a phenomenon of interest in cancer treatment research. Little is known about genetic interactions in childhood cancer. We apply a statistical pipeline to detect genetic interactions in a combined dataset comprising over 2,500 tumors from 23 cancer types. The resulting genetic interaction map of childhood cancers comprises 15 co-occurring and 27 mutually exclusive candidates. The biological explanation of most candidates points to either tumor subtype, pathway epistasis or cooperation while synthetic lethality plays a much smaller role. Thus, other explanations beyond synthetic lethality should be considered when interpreting genetic interaction test results.

Implications of Intratumor Heterogeneity on Consensus Molecular Subtype (CMS) in Colorectal Cancer

The implications of intratumor heterogeneity on the four consensus molecular subtypes (CMS) of colorectal cancer (CRC) are not well known. Here, we use single-cell RNA sequencing (scRNASeq) to build an algorithm to assign CMS classification to individual cells, which we use to explore the distributions of CMSs in tumor and non-tumor cells. A dataset of colorectal tumors with bulk RNAseq (n = 3232) was used to identify CMS specific-marker gene sets. These gene sets were then applied to a discovery dataset of scRNASeq profiles (n = 10) to develop an algorithm for single-cell CMS (scCMS) assignment, which recapitulated the intrinsic biology of all four CMSs. The single-cell CMS assignment algorithm was used to explore the scRNASeq profiles of two prospective CRC tumors with mixed CMS via bulk sequencing. We find that every CRC tumor contains individual cells of each scCMS, as well as many individual cells that have enrichment for features of more than one scCMS (called mixed cells). scCMS4 and scCMS1 cells dominate stroma and immune cell clusters, respectively, but account for less than 3% epithelial cells. These data imply that CMS1 and CMS4 are driven by the transcriptomic contribution of immune and stromal cells, respectively, not tumor cells.

Re-defining synthetic lethality by phenotypic profiling for precision oncology

High-throughput functional and genomic screening techniques provide systematic means for phenotypic discovery. Using synthetic lethality (SL) as a paradigm for anticancer drug and target discovery, we describe how these screening technologies may offer new possibilities to identify therapeutically relevant and selective SL interactions by addressing some of the challenges that have made robust discovery of SL candidates difficult. We further introduce an extended concept of SL interaction, in which a simultaneous perturbation of two or more cellular components reduces cell viability more than expected by their individual effects, which we feel is highly befitting for anticancer applications. We also highlight the potential benefits and challenges related to computational quantification of synergistic interactions and cancer selectivity. Finally, we explore how tumoral heterogeneity can be exploited to find phenotype-specific SL interactions for precision oncology using high-throughput functional screening and the exciting opportunities these methods provide for the identification of subclonal SL interactions.

Identification of Synthetic Lethal Interactions Using High-Throughput, Arrayed CRISPR/Cas9-Based Platforms

Over the past two decades, the concept of synthetic lethality (SL) that queries genetic relationships between gene pairs has gradually emerged as one of the best strategies to selectively eliminate cancer cells. Some of the most successful approaches to identify synthetic lethal interactions (SLIs) were largely dependent on pooled screening formats that require heavy validation in order to mitigate false positives. Here, we describe a high-throughput method to identify SLIs using CRISPR-based strategy that covers, high-throughput production of plasmid DNA preparations, lentiviral production, and subsequent cellular transduction using single guide RNAs (sgRNAs). This method could be adopted to query hundreds of SLIs. As an example, we describe the methods associated with building an interaction map for DNA damage and repair (DDR) genes. The use of multiwell plates and image-based quantification allows a comparative measurement of SLIs at a high-resolution on a one-by-one basis. Furthermore, this scalable, arrayed CRISPR screening method can be applied to multiple cancer cell types, and genes of interest, resulting in new functional discoveries that can be exploited therapeutically.

Targeting SMYD3 to Sensitize Homologous Recombination-Proficient Tumors to PARP-Mediated Synthetic Lethality

SMYD3 is frequently overexpressed in a wide variety of cancers. Indeed, its inactivation reduces tumor growth in preclinical in vivo animal models. However, extensive characterization in vitro failed to clarify SMYD3 function in cancer cells, although confirming its importance in carcinogenesis. Taking advantage of a SMYD3 mutant variant identified in a high-risk breast cancer family, here we show that SMYD3 phosphorylation by ATM enables the formation of a multiprotein complex including ATM, SMYD3, CHK2, and BRCA2, which is required for the final loading of RAD51 at DNA double-strand break sites and completion of homologous recombination (HR). Remarkably, SMYD3 pharmacological inhibition sensitizes HR-proficient cancer cells to PARP inhibitors, thereby extending the potential of the synthetic lethality approach in human tumors.

•

SMYD3 phosphorylation by ATM favors the formation of HR complexes during DSB response

•

SMYD3 mediates DSB repair by promoting RAD51 recruitment at DNA damage sites

•

SMYD3 inhibition triggers a compensatory PARP-dependent DNA damage response

•

Co-targeting SMYD3/PARP leads to synthetic lethality in HR-proficient cancer cells

FEN1 endonuclease as a therapeutic target for human cancers with defects in homologous recombination

Synthetic lethality strategies for cancer therapy exploit cancer-specific genetic defects to identify targets that are uniquely essential to the survival of tumor cells. Here we show RAD27/FEN1, which encodes flap endonuclease 1 (FEN1), a structure-specific nuclease with roles in DNA replication and repair, and has the greatest number of synthetic lethal interactions with Saccharomyces cerevisiae genome instability genes, is a druggable target for an inhibitor-based approach to kill cancers with defects in homologous recombination (HR). The vulnerability of cancers with HR defects to FEN1 loss was validated by studies showing that small-molecule FEN1 inhibitors and FEN1 small interfering RNAs (siRNAs) selectively killed BRCA1- and BRCA2-defective human cell lines. Furthermore, the differential sensitivity to FEN1 inhibition was recapitulated in mice, where a small-molecule FEN1 inhibitor reduced the growth of tumors established from drug-sensitive but not drug-resistant cancer cell lines. FEN1 inhibition induced a DNA damage response in both sensitive and resistant cell lines; however, sensitive cell lines were unable to recover and replicate DNA even when the inhibitor was removed. Although FEN1 inhibition activated caspase to higher levels in sensitive cells, this apoptotic response occurred in p53-defective cells and cell killing was not blocked by a pan-caspase inhibitor. These results suggest that FEN1 inhibitors have the potential for therapeutically targeting HR-defective cancers such as those resulting from BRCA1 and BRCA2 mutations, and other genetic defects.

A cancer drug atlas enables synergistic targeting of independent drug vulnerabilities

Personalized cancer treatments using combinations of drugs with a synergistic effect is attractive but proves to be highly challenging. Here we present an approach to uncover the efficacy of drug combinations based on the analysis of mono-drug effects. For this we used dose-response data from pharmacogenomic encyclopedias and represent these as a drug atlas. The drug atlas represents the relations between drug effects and allows to identify independent processes for which the tumor might be particularly vulnerable when attacked by two drugs. Our approach enables the prediction of combination-therapy which can be linked to tumor-driving mutations. By using this strategy, we can uncover potential effective drug combinations on a pan-cancer scale. Predicted synergies are provided and have been validated in glioblastoma, breast cancer, melanoma and leukemia mouse-models, resulting in therapeutic synergy in 75% of the tested models. This indicates that we can accurately predict effective drug combinations with translational value.

Highly Combinatorial Genetic Interaction Analysis Reveals a Multi-Drug Transporter Influence Network

Many traits are complex, depending non-additively on variant combinations. Even in model systems, such as the yeast S. cerevisiae, carrying out the high-order variant-combination testing needed to dissect complex traits remains a daunting challenge. Here, we describe “X-gene” genetic analysis (XGA), a strategy for engineering and profiling highly combinatorial gene perturbations. We demonstrate XGA on yeast ABC transporters by engineering 5,353 strains, each deleted for a random subset of 16 transporters, and profiling each strain’s resistance to 16 compounds. XGA yielded 85,648 genotype-to-resistance observations, revealing high-order genetic interactions for 13 of the 16 transporters studied. Neural networks yielded intuitive functional models and guided exploration of fluconazole resistance, which was influenced non-additively by five genes. Together, our results showed that highly combinatorial genetic perturbation can functionally dissect complex traits, supporting pursuit of analogous strategies in human cells and other model systems.

Celaj et al. introduce “X-gene” genetic analysis (XGA), a strategy for modeling complex systems by engineering and profiling highly combinatorial genetic perturbations. They apply XGA to 16 yeast ABC transporters, revealing many high-order genetic interactions. Neural network models yielded intuitive functional models and illuminated an ABC transporter influence network, supporting application of XGA to other organisms and processes.

A Humanized Yeast Phenomic Model of Deoxycytidine Kinase to Predict Genetic Buffering of Nucleoside Analog Cytotoxicity

Knowledge about synthetic lethality can be applied to enhance the efficacy of anticancer therapies in individual patients harboring genetic alterations in their cancer that specifically render it vulnerable. We investigated the potential for high-resolution phenomic analysis in yeast to predict such genetic vulnerabilities by systematic, comprehensive, and quantitative assessment of drug–gene interaction for gemcitabine and cytarabine, substrates of deoxycytidine kinase that have similar molecular structures yet distinct antitumor efficacy. Human deoxycytidine kinase (dCK) was conditionally expressed in the Saccharomyces cerevisiae genomic library of knockout and knockdown (YKO/KD) strains, to globally and quantitatively characterize differential drug–gene interaction for gemcitabine and cytarabine. Pathway enrichment analysis revealed that autophagy, histone modification, chromatin remodeling, and apoptosis-related processes influence gemcitabine specifically, while drug–gene interaction specific to cytarabine was less enriched in gene ontology. Processes having influence over both drugs were DNA repair and integrity checkpoints and vesicle transport and fusion. Non-gene ontology (GO)-enriched genes were also informative. Yeast phenomic and cancer cell line pharmacogenomics data were integrated to identify yeast–human homologs with correlated differential gene expression and drug efficacy, thus providing a unique resource to predict whether differential gene expression observed in cancer genetic profiles are causal in tumor-specific responses to cytotoxic agents.

Genetic interaction networks in cancer cells

The genotype-to-phenotype relationship in health and disease is complex and influenced by both an individual's environment and their unique genome. Personal genetic variants can modulate gene function to generate a phenotype either through a single gene effect or through genetic interactions involving two or more genes. The relevance of genetic interactions to disease phenotypes has been particularly clear in cancer research, where an extreme genetic interaction, synthetic lethality, has been exploited as a therapeutic strategy. The obvious benefits of unmasking genetic background-specific vulnerabilities, coupled with the power of systematic genome editing, have fueled efforts to translate genetic interaction mapping from model organisms to human cells. Here, we review recent developments in genetic interaction mapping, with a focus on CRISPR-based genome editing technologies and cancer.

Integrating molecular networks with genetic variant interpretation for precision medicine

More reliable and cheaper sequencing technologies have revealed the vast mutational landscapes characteristic of many phenotypes. The analysis of such genetic variants has led to successful identification of altered proteins underlying many Mendelian disorders. Nevertheless the simple one‐variant one‐phenotype model valid for many monogenic diseases does not capture the complexity of polygenic traits and disorders. Although experimental and computational approaches have improved detection of functionally deleterious variants and important interactions between gene products, the development of comprehensive models relating genotype and phenotypes remains a challenge in the field of genomic medicine. In this context, a new view of the pathologic state as significant perturbation of the network of interactions between biomolecules is crucial for the identification of biochemical pathways associated with complex phenotypes. Seminal studies in systems biology combined the analysis of genetic variation with protein–protein interaction networks to demonstrate that even as biological systems evolve to be robust to genetic variation, their topologies create disease vulnerabilities. More recent analyses model the impact of genetic variants as changes to the “wiring” of the interactome to better capture heterogeneity in genotype–phenotype relationships. These studies lay the foundation for using networks to predict variant effects at scale using machine‐learning or algorithmic approaches. A wealth of databases and resources for the annotation of genotype–phenotype relationships have been developed to support developments in this area. This overview describes how study of the molecular interactome has generated insights linking the organization of biological systems to disease mechanism, and how this information can enable precision medicine.

This article is categorized under:

Translational, Genomic, and Systems Medicine > Translational Medicine

Biological Mechanisms > Cell Signaling

Models of Systems Properties and Processes > Mechanistic Models

Analytical and Computational Methods > Computational Methods