Theses Doctoral

Identifying and Targeting Master Regulators of Drug Resistance in Lung Adenocarcinoma through Network Analysis of Tumor Transcriptomic Data

Griffin, Aaron Timothy

Lung cancer is the leading cause of cancer-related mortality each year in the United States. The majority of patients diagnosed with lung adenocarcinoma (LUAD), the most common histological subtype of lung cancer, present with locally advanced or widely metastatic disease. The American Society of Clinical Oncology (ASCO) and the National Comprehensive Cancer Network (NCCN) recommend that these patients receive systemic antineoplastic treatment tailored to the molecular features of each patient’s disease. LUAD tumors which contain an activating genomic alteration in the gene which encodes the Epidermal Growth Factor Receptor (EGFR) are more likely to respond to treatment with a tyrosine kinase inhibitor (TKI) directed against the EGFR oncoprotein rather than cytotoxic chemotherapy or immune checkpoint inhibitors. However, up to 25% of patients diagnosed with metastatic EGFR-mutated LUAD tumors will not derive clinical benefit from treatment with the current standard-of-care third-generation EGFR TKI osimertinib.

The biochemical mechanisms underlying this phenotype of primary TKI-resistance are poorly understood; as a result, alternative treatment strategies for these patients are limited and no FDA-approved method exists for predicting TKI-resistance at the time of diagnosis. To address this unmet need in the field of thoracic oncology, we have implemented a novel systems biology paradigm developed in the Califano Laboratory which focuses on identifying and targeting key transcriptional regulatory proteins responsible for the initiation and maintenance of tumor phenotypes which we refer to as Master Regulators (MRs). This novel paradigm for precision oncology relies on network analysis of tumor transcriptomic data which can be accurately measured with RNA sequencing (RNA-Seq) technology. In the same way that the turnover of biochemical substrates to products can be used to measure the activity of a metabolic enzyme in vitro, our systems biology approach integrates the differential expression of a transcriptional regulator’s targets to measure differential protein activity in vivo in a manner akin to a highly multiplexed gene reporter assay.

To facilitate the accurate measurement of transcriptional regulatory protein activity from tumor transcriptomic data, we have developed three novel computational algorithms. The Algorithm for the Reconstruction of Accurate Cellular Networks version 3 (ARACNe3) reverse-engineers context-specific transcriptional regulatory networks from gene expression data using an information theoretic framework. The algorithm for Modulator Inference by Network Dynamics version 3 (MINDy3) reverse-engineers context-specific transcriptional modulatory networks from gene expression data using an information theoretic framework.

Finally, Nonparametric analytical Rank-based Enrichment Analysis (NaRnEA) performs gene set analysis under the framework of null hypothesis significance testing using an optimal null model for gene set enrichment derived in accordance with the information theoretic Principle of Maximum Entropy. In order to identify and target MRs of primary TKI-resistance in EGFR-mutated LUAD, we collaborated with clinical oncologists and pathologists at the Herbert Irving Comprehensive Cancer Center (HICCC) to perform RNA-Seq on pretreatment formalin-fixed paraffin-embedded (FFPE) biopsies from 50 EGFR-mutated LUAD tumors. These tumors were divided into an exploratory cohort (n1 = 31) and a validation cohort (n2 = 19) based on the time of sample acquisition and subsequent gene expression profiling. The EGFR-mutated LUAD tumors were classified as TKI-resistant (r1 = 17, r2 = 7) or TKI-sensitive (s1 = 14, s2 = 12) based on patient progression free survival on treatment with a first- or second-generation EGFR TKI.

Our network analysis of these tumors’ transcriptomic data revealed two subtypes of primary TKI-resistance which were represented in both the exploratory cohort and the validation cohort. These subtypes of TKI-resistant EGFR-mutated LUAD tumors exhibited distinct transcriptional regulatory protein activity signatures with respect to cohort-matched TKI-sensitive EGFR-mutated LUAD tumors. Using one-of-a-kind drug perturbation gene expression profiles for the LUAD cell line NCIH1793 obtained by the High-Throughput Screening center at the Columbia University Irving Medical Center (CUIMC), we implemented an updated version of the OncoTreat algorithm to identify FDA-approved and investigational compounds capable of targeting the MRs of each TKI-resistant EGFR-mutated LUAD tumor. To validate these drug predictions, we obtained the TKI-resistant EGFR-mutated LUAD cell lines NCIH2172 and NCIH1650. A CRISPR/Cas9 knockout screen performed in these cell lines by researchers in the Califano Laboratory at CUIMC revealed that knocking out MRs for the major subtype of TKI-resistance significantly abrogated the growth of these cell lines; additionally, this effect was amplified for NCIH1650 in the setting of combined treatment with osimertinib. Based on the results of this screen, we plan to validate the drugs prioritized by OncoTreat to target the major subtype of primary TKI-resistance in NCIH2172 and NCIH1650 alone and in combination with osimertinib to evaluate for single-agent and dual-agent efficacy in vitro.

To provide additional validation for this analysis, we collaborated with researchers at the Memorial Sloan Kettering Cancer Center (MSKCC) to create patient-derived xenograft (PDX) models of TKI-resistant EGFR-mutated LUAD tumors. A total of six EGFR-mutated LUAD tumors were obtained from the Rudin Laboratory at MSKCC based on clinical history suggestive of TKI-resistance; these EGFR-mutated LUAD tumors were subsequently implanted into NOD scid gamma (NSG) immunocompromised mice in the Kung Laboratory at MSKCC to generate PDX models. Once the EGFR-mutated LUAD tumors had engrafted, researchers in the Califano Laboratory at CUIMC performed RNA-Seq on microdissected tumor from each model.

Using a novel transcriptomic machine learning biomarker trained on our exploratory cohort and tested on our validation cohort to predict primary TKI-resistance from MR protein activity, we predicted that all six EGFR-mutated LUAD PDX models would demonstrate primary TKI-resistance. These predictions were further reinforced by differential protein activity and OncoTreat analysis which revealed that these EGFR-mutated LUAD PDX models strongly recapitulated the MRs of the major subtype of primary TKI-resistance. Based on this analysis and future in vitro validation, we plan to conduct a preclinical trial using these EGFR-mutated LUAD PDX models to evaluate the drugs prioritized by OncoTreat to target the MRs of the major subtype of primary TKI-resistance alone and in combination with osimertinib. If successful, this novel paradigm for precision oncology could motivate the development of a biomarker-guided clinical trial for patients diagnosed with metastatic EGFR-mutated LUAD at high risk for developing primary TKI-resistance.

Files

This item is currently under embargo. It will be available starting 2025-02-08.

More About This Work

Academic Units
Cellular, Molecular and Biomedical Studies
Thesis Advisors
Califano, Andrea
Degree
Ph.D., Columbia University
Published Here
February 22, 2023