Comprehensive mutational scanning of EGFR reveals TKI sensitivities of extracellular domain mutants

The epidermal growth factor receptor, EGFR, is frequently activated in lung cancer and glioblastoma by genomic alterations including missense mutations. The different mutation spectra in these diseases are reflected in divergent responses to EGFR inhibition: significant patient benefit in lung cancer, but limited in glioblastoma. Here, we report a comprehensive mutational analysis of EGFR function. We perform saturation mutagenesis of EGFR and assess function of ~22,500 variants in a human EGFR-dependent lung cancer cell line. This approach reveals enrichment of erlotinib-insensitive variants of known and unknown significance in the dimerization, transmembrane, and kinase domains. Multiple EGFR extracellular domain variants, not associated with approved targeted therapies, are sensitive to afatinib and dacomitinib in vitro. Two glioblastoma patients with somatic EGFR G598V dimerization domain mutations show responses to dacomitinib treatment followed by within-pathway resistance mutation in one case. In summary, this comprehensive screen expands the landscape of functional EGFR variants and suggests broader clinical investigation of EGFR inhibition for cancers harboring extracellular domain mutations.

Abstract 5324: Drug repurposing and genetic screening strategies for effective treatment discovery in soft-tissue sarcomas

Treatment advances for soft-tissue sarcomas have been slow. This is, in part, due to their rarity (accounting for 0.7% of all cancers) and heterogeneity (over 50 different diseases fall under this category). Moreover, preclinical models are scarce, often exhibiting slow growth kinetics, which limits their study by large genetic and pharmacological libraries. Here, we present an update on our efforts to harness the power of patient-partnered research to create a platform for rare cancer drug target discovery as a broadly available community resource. We developed a patient-partnered tissue donation pipeline to enable patients anywhere in the United States to participate and piloted our approach for CTNNB1-driven desmoid tumors. To overcome challenges in tissue heterogeneity during ex vivo culture, we optimized a multiplexed sequencing protocol to quantitatively track changes in tumor cell fraction across hundreds of media formulations. Following this strategy, we were able to verify and expand three cell lines that preserve the CTNNB1 mutations at high purity. To identify potential therapeutics, we completed a 6,750-drug repurposing screen, at 2.5uM in duplicate, in two verified cell line models. After extensive quality control assessments and data integration steps to leverage the power of other large scale drug screens, we selected 263 compounds for follow-up based on potency, selectivity, and association with molecular features associated with desmoid tumors. Approximately 70% of selected compounds were validated by an 8-point, 2-fold dilution, dose-response format with a top concentration of 10uM. Of the confirmed active compounds, 80 showed a strong pattern of selectivity, 20 are FDA approved drugs and 13 investigational compounds show a statistical association with CTNNB1 hotspot mutation status or transcriptomic features associated with desmoid tumors. To prioritize potential therapeutic targets, we tested an efficient CRISPR/Cas9 all-in-one library design. The reduction of the CRISPR/Cas9 library size was achieved via multiple gene- and guide-level strategies, which enables statistically powered gene essentiality interrogation in slow-growing patient-derived models. We tested several plating and infection parameters and developed an optimized pipeline for the rapid introduction of this library into early patient-derived samples. Established cell lines of mesenchymal and non-mesenchymal origin, which have previously been tested by genome-wide libraries, were used to control for library and lineage effects. We are developing a biologist-friendly web portal that will enable the research community to easily interact with models and data produced by this effort. Our study provides evidence that a systematic patient-powered approach can facilitate discovery of therapeutic hypotheses for these understudied diseases.

Citation Format: Mushriq Al-Jazrawe, Kathryn Cebula, Elisabeth A. Abeyta, Haley S. Curtis, Jane K. McIninch, Jaime H. Cheah, James Berstler, Lisa Miller, James Neiswender, Lisa Brenan, Mike Burger, Francisca Vazquez, Jesse S. Boehm. Drug repurposing and genetic screening strategies for effective treatment discovery in soft-tissue sarcomas. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5324.

CloneSifter: enrichment of rare clones from heterogeneous cell populations

BACKGROUND

Many biological processes, such as cancer metastasis, organismal development, and acquisition of resistance to cytotoxic therapy, rely on the emergence of rare sub-clones from a larger population. Understanding how the genetic and epigenetic features of diverse clones affect clonal fitness provides insight into molecular mechanisms underlying selective processes. While large-scale barcoding with NGS readout has facilitated cellular fitness assessment at the population level, this approach does not support characterization of clones prior to selection. Single-cell genomics methods provide high biological resolution, but are challenging to scale across large populations to probe rare clones and are destructive, limiting further functional analysis of important clones.

RESULTS

Here, we develop CloneSifter, a methodology for tracking and enriching rare clones throughout their response to selection. CloneSifter utilizes a CRISPR sgRNA-barcode library that facilitates the isolation of viable cells from specific clones within the barcoded population using a sequence-specific retrieval reporter. We demonstrate that CloneSifter can measure clonal fitness of cancer cell models in vitro and retrieve targeted clones at abundance as low as 1 in 1883 in a heterogeneous cell population.

CONCLUSIONS

CloneSifter provides a means to track and access specific and rare clones of interest across dynamic changes in population structure to comprehensively explore the basis of these changes.

Rhabdoid Tumors Are Sensitive to the Protein-Translation Inhibitor Homoharringtonine

PURPOSE

Rhabdoid tumors are devastating pediatric cancers in need of improved therapies. We sought to identify small molecules that exhibit in vitro and in vivo efficacy against preclinical models of rhabdoid tumor.

EXPERIMENTAL DESIGN

We screened eight rhabdoid tumor cell lines with 481 small molecules and compared their sensitivity with that of 879 other cancer cell lines. Genome-scale CRISPR-Cas9 inactivation screens in rhabdoid tumors were analyzed to confirm target vulnerabilities. Gene expression and CRISPR-Cas9 data were queried across cell lines and primary rhabdoid tumors to discover biomarkers of small-molecule sensitivity. Molecular correlates were validated by manipulating gene expression. Subcutaneous rhabdoid tumor xenografts were treated with the most effective drug to confirm in vitro results.

RESULTS

Small-molecule screening identified the protein-translation inhibitor homoharringtonine (HHT), an FDA-approved treatment for chronic myelogenous leukemia (CML), as the sole drug to which all rhabdoid tumor cell lines were selectively sensitive. Validation studies confirmed the sensitivity of rhabdoid tumor to HHT was comparable with that of CML cell lines. Low expression of the antiapoptotic gene BCL2L1, which encodes Bcl-XL, was the strongest predictor of HHT sensitivity, and HHT treatment consistently depleted Mcl-1, the synthetic-lethal antiapoptotic partner of Bcl-XL. Rhabdoid tumor cell lines and primary-tumor samples expressed low BCL2L1, and overexpression of BCL2L1 induced resistance to HHT in rhabdoid tumor cells. Furthermore, HHT treatment inhibited rhabdoid tumor cell line and patient-derived xenograft growth in vivo.

CONCLUSIONS

Rhabdoid tumor cell lines and xenografts are highly sensitive to HHT, at least partially due to their low expression of BCL2L1. HHT may have therapeutic potential against rhabdoid tumors.

Genotype-Fitness Maps of EGFR-Mutant Lung Adenocarcinoma Chart the Evolutionary Landscape of Resistance for Combination Therapy Optimization

Cancer evolution poses a central obstacle to cure, as resistant clones expand under therapeutic selection pressures. Genome sequencing of relapsed disease can nominate genomic alterations conferring resistance but sample collection lags behind, limiting therapeutic innovation. Genome-wide screens offer a complementary approach to chart the compendium of escape genotypes, anticipating clinical resistance. We report genome-wide open reading frame (ORF) resistance screens for first- and third-generation epidermal growth factor receptor (EGFR) inhibitors and a MEK inhibitor. Using serial sampling, dose gradients, and mathematical modeling, we generate genotype-fitness maps across therapeutic contexts and identify alterations that escape therapy. Our data expose varying dose-fitness relationship across genotypes, ranging from complete dose invariance to paradoxical dose dependency where fitness increases in higher doses. We predict fitness with combination therapy and compare these estimates to genome-wide fitness maps of drug combinations, identifying genotypes where combination therapy results in unexpected inferior effectiveness. These data are applied to nominate combination optimization strategies to forestall resistant disease.

Neuronal differentiation and cell-cycle programs mediate response to BET-bromodomain inhibition in MYC-driven medulloblastoma

BET-bromodomain inhibition (BETi) has shown pre-clinical promise for MYC-amplified medulloblastoma. However, the mechanisms for its action, and ultimately for resistance, have not been fully defined. Here, using a combination of expression profiling, genome-scale CRISPR/Cas9-mediated loss of function and ORF/cDNA driven rescue screens, and cell-based models of spontaneous resistance, we identify bHLH/homeobox transcription factors and cell-cycle regulators as key genes mediating BETi's response and resistance. Cells that acquire drug tolerance exhibit a more neuronally differentiated cell-state and expression of lineage-specific bHLH/homeobox transcription factors. However, they do not terminally differentiate, maintain expression of CCND2, and continue to cycle through S-phase. Moreover, CDK4/CDK6 inhibition delays acquisition of resistance. Therefore, our data provide insights about the mechanisms underlying BETi effects and the appearance of resistance and support the therapeutic use of combined cell-cycle inhibitors with BETi in MYC-amplified medulloblastoma.

Abstract 1815: Massively parallel identification of conserved drug resistant mutations in kinases

Drug resistant mutations that arise in therapeutic targets often limit clinical responses. However, the discovery of such mutations has historically been performed one gene or mutation at a time, often over decades of experimental and clinical testing, limiting our understanding of conserved mechanisms of drug resistance.

We hypothesized that deep mutational scanning of canonical kinases may expedite this process and identify novel conserved elements that cause drug resistance when mutated (similar to the well-studied “Gatekeeper” residue). To test this, we generated cDNA-expression libraries containing all possible amino acid substitutions in CDK6, CDK4, ERK2, and EGFR. We screened each library against clinically utilized, ATP-competitive small molecule inhibitors. We then mapped the phenotypic data for over 40,000 missense mutations onto the aligned crystal structures of each protein and searched for shared structural attributes associated with drug resistance.

This analysis revealed 4 equivalent amino acid sites whose mutation conferred drug resistance to ATP-competitive inhibitors in all of our screens: the Gatekeeper residue, as well as three uncharacterized residues. One of these sites, which we have termed the “Keymaster”, was additionally found to cause resistance in published data sets of sub-saturation BRAF, HER2, BCR-ABL, and MEK1 mutagenesis screens against their respective inhibitors. We confirmed that drug resistant phenotypes are caused by these alterations utilizing growth assays and protein target phosphorylation detection assays. Mechanistically, we show preliminary evidence that Keymaster-mutant proteins are competent for drug binding, but may display elevated basal activity. Consistent with our findings, we additionally identified mutations at Keymaster residues in reported patient tumors in a number of oncogene kinases, suggesting that Keymaster mutations could be drivers of tumorigenesis, as well as drug resistance. These efforts may prove useful for characterizing somatic kinase mutations of unknown function, designing next-generation therapeutics and deepening our understanding of kinase regulation.

Citation Format: Nicole S. Persky, Desiree Hernandez, Jonathon Cordova, Amanda Walker, Lisa Brenan, Federica Piccioni, Sasha Pantel, Yenarae Lee, Amy Goodale, Xiaoping Yang, Yoichiro Mitsuishi, Mariana Do Carmo, Cong Zhu, Aleksandr Andreev, David E. Root, Cory M. Johannessen. Massively parallel identification of conserved drug resistant mutations in kinases [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1815.

Abstract 954: Predicting synergistic drug combinations and resistance mechanisms from genomic features and single-agent response profiles

Drug combinations promise to improve clinical responses and/or forestall drug resistance. To capitalize on this promise, we need to know which drugs to combine, and which patients to give them to based on the genetic or pathological features of their disease. However, progress towards this goal has been hindered by the infeasibility of performing comprehensive drug-combination studies across thousands of cellular contexts.

We hypothesized that the basal gene-transcription state of cancer cell lines, in concert with the cell-viability profiles of single-agent small molecules, might be leveraged to nominate specific synergistic drug combinations and identify mechanisms of drug resistance, eliminating the need to test all possible drug/drug combinations across cellular models. Specifically, we predicted that inhibiting the protein product of transcripts associated with drug resistance to a given small molecule might induce drug synergy.

To test this notion, we analyzed nearly 400,000 drug-sensitivity profiles in >800 cancer cell lines to identify candidate compound-gene pairs. We identified over 100 examples where outlier expression of a single transcript was correlated with resistance to a small molecule. Of these gene/drug pairs, 9 genes represented imminently druggable targets, including established clinically-relevant relationships between the alkylating agent temozolomide and MGMT expression, and between a subset of chemotherapeutics including paclitaxel and the efflux pump ABCB1.

Inhibition of candidate “co-targets”, which included 3 previously characterized relationships and 6 novel relationships, resulted in cell-line-specific synergistic cell killing across multiple cell-line models. For validated compound-gene pairs, exogenous expression of the “co-target” was sufficient to confer resistance. For example, we found that high expression of MGLL, encoding monoglyceride lipase, was uniquely associated with lack of response to the histone lysine demethylase inhibitor GSK-J4. Endogenous or exogenous MGLL expression conferred resistance to GSK-J4, while MGLL-proficient cell lines could be sensitized to GSK-J4 up to 50-fold by co-treatment with an irreversible MGLL inhibitor.

These initial studies highlight the potential of integrating basal gene expression features with small-molecule response to nominate rational candidates for drug combinations. As public repositories of single agent response data from diverse cellular contexts continue to expand, so too will our repertoire of therapeutic combinations. Moreover, this approach permits the parallel identification of genomic features that indicate which patient populations are most likely to benefit from such combinations.

Citation Format: Matthew G. Rees, Lisa Brenan, Patrick Duggan, Cory M. Johannessen. Predicting synergistic drug combinations and resistance mechanisms from genomic features and single-agent response profiles [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 954.

Abstract 1178: Genotype-fitness maps guide targeted therapy combination in lung adenocarcinoma

Targeted EGFR inhibition in lung cancer leads to dramatic responses. Nonetheless, disease evolution to resistance is the rule. As this evolutionary process is fueled by intra-tumoral genetic diversity, a comprehensive mapping of clonal fitness is required to inform strategies to overcome resistance.

To model intra-tumoral diversity in vitro, we performed a genome-wide, over-expression genetic perturbation assay in EGFR-driven non-small cell lung cancer (NSCLC) PC9 cells. Our screen covered 17,255 ORFs (open reading frame constructs) representing 12,728 wildtype and mutated genes, and examined the effects of a first-generation EGFR inhibitor (erlotinib), a third-generation EGFR inhibitor (osimertinib) and a MEK inhibitor (binimetinib) on evolutionary selection, alone or in combination. Specifically, to obtain genotype-to-fitness maps for each drug, we measured clonal abundance up to 4 times during the screen and mathematically resolved their growth behavior. Finally, for each drug, we applied multiple doses to obtain dose-fitness relationships, which may impact tumor evolution in patients due to high inter- and intra-tumoral variability in drug delivery. In total, we have performed 78 genome-wide screens to map the evolutionary landscape of PC9 resistance to targeted therapy.

Erlotinib and osimertinib both result in an overwhelming reduction of fitness, as expected from their clinical benefit. Known clinical resistance mechanisms involving ERBB2, PIK3CA, AXL, and BRAF confer a pronounced fitness advantage in the presence of both drugs, while EGFR (T790M) improves fitness only with erlotinib. Novel resistance mechanisms include alternative tyrosine kinases (NTRK, PDGFRB, CSF1R, KIT), KRAS, G-coupled protein receptors, transcription factors (SOX15, FOXA1), and cellular transporters (ABCG2). Notably, we observe significant divergence in dose-fitness relationships between different resistance mechanisms. For example, PIK3CA confers a modest but persistent advantage across the dose range, in contrast to BRAF which results in a fitness advantage only when sufficiently high drug levels are added.

The fitness landscape for binimetinib resistance appeared to be largely orthogonal to that of the EGFR inhibitors, with decreased fitness noted for many tyrosine kinases as well as KRAS mutations. MEK inhibitor resistance results from RAF family member overexpression (ARAF, BRAF, RAF1) and MAPK activation. The orthogonal resistance landscapes of EGFR inhibition and MEK inhibition translate into highly synergistic effects in drug combination, with effective prediction of combinatorial fitness from the single-agent fitness landscapes using generalized linear models. These results validate this approach as a systematic method to address the combinatorial problem of optimizing drug combinations and doses to directly anticipate and address cancer evolution.

Citation Format: Asaf Zviran, Patrick Bolan, Lisa Brenan, Amy Goodale, Denisse Rotem, Viktor Adalsteinsson, Federica Piccioni, Cory Johannessen, Dan Landau. Genotype-fitness maps guide targeted therapy combination in lung adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1178.

Phenotypic Characterization of a Comprehensive Set of MAPK1/ERK2 Missense Mutants

Tumor-specific genomic information has the potential to guide therapeutic strategies and revolutionize patient treatment. Currently, this approach is limited by an abundance of disease-associated mutants whose biological functions and impacts on therapeutic response are uncharacterized. To begin to address this limitation, we functionally characterized nearly all (99.84%) missense mutants of MAPK1/ERK2, an essential effector of oncogenic RAS and RAF. Using this approach, we discovered rare gain- and loss-of-function ERK2 mutants found in human tumors, revealing that, in the context of this assay, mutational frequency alone cannot identify all functionally impactful mutants. Gain-of-function ERK2 mutants induced variable responses to RAF-, MEK-, and ERK-directed therapies, providing a reference for future treatment decisions. Tumor-associated mutations spatially clustered in two ERK2 effector-recruitment domains yet produced mutants with opposite phenotypes. This approach articulates an allele-characterization framework that can be scaled to meet the goals of genome-guided oncology.

Abstract A18: Predicting synergistic drug combinations from genomic features and single-agent response profiles

Drug combinations promise to improve clinical responses and/or forestall drug resistance. To capitalize on this promise, we need to know which drugs to combine, and whom to give them to based on the genetic or pathological features of their disease. However, accomplishing this goal has been precluded by the infeasibility of performing comprehensive drug-combination studies across thousands of cellular contexts. We hypothesized that the basal gene-transcription state of cancer cell lines, in concert with the response profiles of hundreds of single-agent small molecules, might be leveraged to nominate synergistic drug combinations, eliminating the need to test all possible drug/drug combinations across cellular models. Specifically, we predicted that inhibiting the protein product of transcripts associated with drug resistance to a given small molecule might induce drug synergy. To test this notion, we analyzed public cell-line drug-sensitivity data to identify candidate compound-gene pairs. We identified 7 examples in which outlier expression of a druggable candidate protein was associated with lack of single-agent response. Inhibition of 6/7 candidate co-targets resulted in cell-line-specific synergistic cell killing across multiple cell line models, validating the overall approach. For example, consistent with clinical findings, we found that high expression of the MGMT gene, encoding O-6-methylguanine-DNA methyltransferase, was uniquely associated with response to alkylating agents such as temozolomide, and that combination of the MGMT inhibitor O-6-benzylguanine with temozolomide resulted in synergistic killing. These initial studies highlight the potential of integrating basal gene expression features with small-molecule response to nominate rational candidates for drug combinations. As public repositories of single-agent response data from diverse cellular contexts continue to expand, so too will our repertoire of therapeutic combinations. Moreover, this approach permits the parallel identification of genomic features that indicate which patient populations are most likely to benefit from such combinations.

Citation Format: Matthew G. Rees, Lisa Brenan, Amanda Walker, Cory M. Johannessen. Predicting synergistic drug combinations from genomic features and single-agent response profiles [abstract]. In: Proceedings of the AACR Precision Medicine Series: Opportunities and Challenges of Exploiting Synthetic Lethality in Cancer; Jan 4-7, 2017; San Diego, CA. Philadelphia (PA): AACR; Mol Cancer Ther 2017;16(10 Suppl):Abstract nr A18.