Uri Ben-David, Ph.D. Assistant Professor, Department of Human Molecular Genetics and Biochemistry at the Tel Aviv University speaks about the AARC 2020 – Abstract NG02: Genomic evolution of cancer models: Perils and opportunities.
Instract
Cancer precision medicine is based on the premise that the drug response is correlated with the unique genomic features of a human tumor. Therefore, its vulnerabilities could be predicted by analyzing the genomic features of a tumor, and appropriate treatment could then be tailored. To achieve this goal, a detailed “dependence map” needs to be established that would allow each tumor’s “Achilles heel” to be predicted based on its genomic characteristics. The cancer community relies heavily on cancer models, such as genetically engineered mouse models (GEMMs), patient-derived xenografts (PDXs), and human cell lines, to try and establish such a cancer dependence chart (CLs). Importantly, xenografts and cell lines are not only used for cancer studies but also to predict a particular patient’s reaction to anti-cancer drugs (i.e., as “avatars” for tumors). Much work has recently been put in their genomic and phenotypic profiling, as cancer models are invaluable for cancer research. However, cancer models evolve, like any biological system. Although much attention has been paid to tumor heterogeneity and genomic instability, the heterogeneity and instability of cancer models – and how their use in cancer research and cancer precision medicine is influenced by the genomic evolution of these models – remain under-explorated. I have researched the extent of genomic evolution in various cancer models, its biological origins, and its functional implications in my postdoctoral work. This thesis consisted of three independent experiments in three major cancer model systems: GEMMs, PDXs, and CLs, which dissected different aspects of genomic evolution. I will explain the key findings of my studies in this talk, and highlight important trends arising from their synthesis. We created the first detailed catalogue of copy-number alterations (CNAs) in cancer GEMMs in a first analysis (Ben-David et al. Nature Communications 2016). Mining this resource, we found that late during breast tumorigenesis, chromosomal aberrations accumulated and observed marked differences in CNA prevalence between mouse mammary tumors initiated by separate drivers. Some aberrations were repeated and peculiar to particular GEMMs, indicating different routes to tumorigenesis depending on the carrier. We narrowed down the critical region of interest in one of the most frequent chromosomal changes in breast cancer (loss of chromosome 1p) using syntene-based oncogenomics of mouse and human results and identified a gene (Sfn) that cooperates with Erbb2 during tumorigenesis of breast cancer. We have experimentally validated that Stratifin (SFN) loss promotes tumorigenesis induced by HER2. This research shows how the natural genomic evolution in GEMMs during tumorigenesis can inform the identification of oncogenic networks that are essential for human cancer pathogenesis. In the second research (Ben-David et al. Nature Genetics 2017), during its derivation and in vivo dissemination, we followed the genomic evolution of PDXs. The PDX copy-number ecosystems have evolved extensively, often driven by strong clonal dynamics, leading to the expansion of pre-existing minor subclones. During derivation and early propagation of the models, clonal dynamics were highest and attenuated at later stages. In independent PDXs developed from the same primary tumor, reproducible changes were observed, suggesting the involvement of selection rather than mere genetic drift. Importantly, the genomic evolution rate in PDXs was close to that observed in CLs derived from patients. The evolution of tumors in PDXs and in patients followed different trajectories. Specifically, repeated cases of aneuploidy in primary tumors eventually vanished in PDXs. The genomic stability of PDXs was related to their response to chemotherapy and targeted drugs, indicating that the genomic evolution of PDXs could affect the accuracy of the PDX-based prediction of the drug response of patients. We compared genomic analyses of 106 CLs grown in two laboratories in a third study (Ben-David et al. Nature 2018) and revealed large genetic diversity. A detailed genomic characterization follow-up of 27 strains of the common MCF7 line of breast cancer cells revealed rapid genetic diversification. With multiple strains of 13 additional CLs, identical results were obtained. Differential activation of gene expression programs and pronounced variations in cell morphology and proliferation have been correlated with genetic changes. Experiments with barcoding showed that CL evolution occurred as a result of positive clonal selection that is highly sensitive to conditions of culture. Single cell-derived clone studies showed that ongoing instability rapidly converted into the heterogeneity of CL. Testing of the 27 MCF7 strains against 321 anti-cancer compounds functionally showed a remarkably disparate drug response: in others, over 80% of compounds that strongly inhibited certain CL strains were fully inactive. In basic cancer research and cancer precision medicine, these results have wide implications for cell line use. Importantly, realistic ways to minimize the risks presented by the genomic evolution of CL and to promote maximally reproducible studies centered on the cell line are suggested. In addition, we suggest in future studies how to constructively expand on this phenomenon (e.g. to research genetic variance forms that can not be easily implemented experimentally, such as significant chromosomal changes). In a fourth study (Ben-David et al. Submitted), before and after the implementation of Cas9, we conducted a deep genomic analysis of human cancer cell lines. Upon introduction to Cas9, specifically in TP53-WT cell lines, we find upregulation of the p53 pathway. At mRNA and protein levels, this upregulation was confirmed. In addition, in the Cas9 cell lines, we found elevated levels of DNA repair transcriptional signatures and verified that the introduction of Cas9 induced DNA damage using γH2AX immunofluorescence quantification. Genetic characterization of 42 pairs of cell lines showed that the introduction of Cas9 could lead to the emergence and expansion of mutations that inactivate p53. Competition experiments with isogenic cell lines TP53-WT/TP53-null verified that the introduction of Cas9 accelerated p53-inactivating mutation selection. Finally, we compared Cas9 activity across 719 human cell lines and found that in TP53-WT cell lines, Cas9 was significantly less active, in line with functional p53 being a barrier to Cas9’s successful expression in human cells. These results have wide implications for the proper use in basic research and in clinical applications of CRISPR/Cas9 genome editing. Following a given, common selection pressure, they also display the profound functional consequences of in vitro genomic evolution. In summary, my postdoctoral studies showed that genomic instability in multiple cancer models had previously been under-recognized, and characterized the implications of this instability for the use of such models in cancer research. This body of study, together, showed widespread genetic evolution in cancer models, far from that seen in patients, resulting in changes in gene expression and disparate drug response. Our studies of cancer models’ stability and fidelity strengthen our biological understanding of the models’ benefits and drawbacks and can help direct their proper implementation. Importantly, cancer models’ genomic heterogeneity often offers a rare opportunity to use these models in innovative and inventive ways (Ben-David at al. Nature Reviews Cancer 2019; Ben-David and Amon Nature Reviews Genetics 2019). This thesis thus illustrates both the risks and the benefits of cancer models’ natural evolution.