Browsing by Author "Wood, Kris"

The main contribution of this thesis is to examine the utility of kernel regression ap- proaches and variance component models for solving complex problems in statistical genetics and molecular biology. Many of these types of statistical methods have been developed specifically to be applied to solve similar biological problems. For example, kernel regression models have a long history in statistics, applied mathematics, and machine learning. More recently, variance component models have been extensively utilized as tools to broaden understanding of the genetic basis of phenotypic varia- tion. However, because of large combinatorial search spaces and other confounding factors, many of these current methods face enormous computational challenges and often suffer from low statistical power --- particularly when phenotypic variation is driven by complicated underlying genetic architectures (e.g. the presence of epistatic effects involving higher order genetic interactions). This thesis highlights two novel methods which provide innovative solutions to better address the important statis- tical and computational hurdles faced within complex biological data sets. The first is a Bayesian non-parametric statistical framework that allows for efficient variable selection in nonlinear regression which we refer to as "Bayesian approximate kernel regression", or BAKR. The second is a novel algorithm for identifying genetic vari- ants that are involved in epistasis without the need to identify the exact partners with which the variants interact. We refer to this method as the "MArginal ePIstasis Test", or MAPIT. Here, we develop the theory of these two approaches, and demonstrate their power, interpretability, and computational efficiency for analyz- ing complex phenotypes. We also illustrate their ability to facilitate novel biological discoveries in several real data sets, each of them representing a particular class of analyses: genome-wide association studies (GWASs), molecular trait quantitative trait loci (QTL) mapping studies, and cancer biology association studies. Lastly, we will also explore the potential of these approaches in radiogenomics, a brand new subfield of genetics and genomics that focuses on the study of correlations between imaging or network features and genetic variation.

The dramatic expansion of genomic sequencing methodologies, applications and efforts has empowered our abilities to deepen the conceptual understanding of complex biological processes, including diseases like cancer. Through our accumulated understanding of cancer genomics, targeted therapies, which inhibit the specific driver oncogenes and pathophysiological processes that underlie cancer progression, have been developed. However, in modern precision oncology and therapeutics, cancer drug resistance, both primary and secondary, has greatly limited the potential of targeted therapies to improve patients’ lives. Here, we systematically define combination treatment strategies by using unbiased pharmacological and functional genetic screening approaches to overcome the persistent problem of primary drug resistance in two cancer contexts: (1) epidermal growth factor receptor (EGFR)-driven triple-negative breast cancer (TNBC) and (2) PIK3CA mutant gastric cancer. Particularly, in the first context, using a candidate drug screen, we discovered that inhibition of cyclin-dependent kinase (CDK) 12 dramatically sensitizes diverse models of TNBC to EGFR blockade. Instead of functioning through CDK12’s well-established transcriptional roles, this combination therapy drives cell death through the 4E-BP1-dependent suppression of the translation and consequent stability of driver oncoproteins, including MYC. Further, with mechanistic intent, using a genome-wide CRISPR/Cas9 screen, we identified the CCR4-NOT complex as a major determinant of sensitivity to the combination therapy whose loss renders 4E-BP1 unresponsive to drug-induced dephosphorylation, rescuing MYC translational suppression and stability. Thus, by revealing a long debated EGFR dependence in TNBC, we have identified a therapeutic approach that functions through the cooperative regulation of translation-coupled oncoprotein stability and holds promising translational potential for the treatment of this difficult-to-treat disease subtype. In the second context, despite extensive molecular characterization of gastric cancer, personalized treatment approaches to improve patient survival outcomes are still lacking. Motivated by this unmet need, we performed drug sensitizer screens with a PI3K-alpha isoform-specific inhibitor, BYL719, in multiple PIK3CA wild-type (WT) and mutant cell lines, including those derived from gastric cancers, head and neck squamous cell carcinomas (HNSCCs), and colorectal cancers using a miniaturized CRISPR/Cas9 library targeting key druggable nodes of cellular survival pathways. This work led to the promising findings that intrinsic resistance to PI3K-alpha inhibition specifically in gastric cancer may be mediated by BCL-xL and NEDD9. Sensitization to PI3K-alpha inhibition by BCL-xL specific inhibitor revealed a novel targeted approach for the treatment of EBV+ PIK3CA mutant gastric cancers, thereby overcoming a perplexing obstacle to the effective targeting of PI3K oncogenic dependency in this cancer subtype. Collectively, our work demonstrated the ability and applicability of screening approaches to define the determinants of primary drug resistance in precision cancer therapies across diverse cancer contexts.

ATRX is one of the most frequently altered genes in soft tissue sarcoma, with alterations occurring in 29% of these tumors. However, the role of ATRX in the development and response to cancer therapies in soft tissue sarcoma remains poorly understood. Here, we developed a primary mouse model of soft tissue sarcoma and studied the effect of Atrx deletion on tumor development and therapeutic response. Our findings demonstrate that Atrx deletion regulates tumor development and increases sarcoma sensitivity to radiation therapy. In the absence of Atrx, irradiated sarcomas have increased persistent DNA damage, telomere dysfunction, and mitotic catastrophe. We find that Atrx deleted tumors have impaired cGAS-STING signaling, with accompanying sensitivity to the novel clinical therapy oncolytic herpesvirus. Translation of these results to patients with ATRX mutant cancers could enable genomically-guided cancer therapeutic approaches that improve patient outcomes.

Primary glioblastoma (GBM) is the most common and lethal primary malignant brain tumor, with a median patient survival of only 15 months from the time of diagnosis. GBM is particularly challenging to treat due to its aggressive and invasive nature, and has proven resistant to therapeutic advances, with no significant improvement in outcomes over the past several decades. Understanding of the molecular characteristics of GBM, however, has improved dramatically, with genetic, epigenetic, and transcriptomic classifications now able to divide GBM into subtypes that provide prognostic information and guide the organization of clinical trials. One of the most frequent genetic alterations that has been identified in GBM is homozygous deletion of the methylthioadenosine phosphorylase (MTAP) gene, which occurs in 50% of all GBM cases. Despite its common occurrence, it is unclear what contribution MTAP loss makes in the pathogenesis of GBM or whether this genetic alteration can be used as a therapeutic target.

MTAP is a metabolic enzyme in the salvage pathway of adenine and methionine and its absence results in the accumulation of its metabolic substrate, methylthioadenosine (MTA), within and around tumor cells. MTA is known to inhibit activity of methyltransferases, raising the possibility that MTA accumulation is interfering with regulatory processes within the cell.

We utilized patient-derived GBM cell lines in vitro and GBM xenografts in vivo, to characterize consequence of MTAP deletion in GBM through analysis of DNA methylation, gene expression, and response to therapeutic agents. We show that MTAP loss promotes the formation of glioma stem-like cells through epigenomic dysregulation. We show these epigenetic changes influence gene expression patterns and alter the sensitivity to epigenome-modifying drugs. We also demonstrate that MTAP-null GBM cells are more tumorigenic in experimental models and that patients with MTAP deletion have poor disease outcomes. Finally, we show that targeting metabolic liabilities of MTAP-null cells through inhibition of de novo purine synthesis specifically depletes the therapy-resistant, stem-like cell subpopulation of GBM.

As the final component of this work, we explore the impact of MTA accumulation in the tumor microenvironment. We found that MTA alters the function of immune cells through adenosine receptor signaling, suggesting that modulation of adenosine receptor signaling in GBM may improve the native immune response and the efficacy of immunotherapeutics in the treatment of this disease.

This work thus establishes MTAP deletion as a pathogenic genetic alteration in the process of gliomagenesis by illustrating it’s contribution to the formation of the cancer cell epigenomic landscape, stemness characteristics, growth, and response to therapeutic agents.

Targeted therapeutics are among the most promising approaches for treating diverse forms of malignancies. Indeed, as sequencing prices and technology continue to improve it will be possible to achieve a precise map of each individual cancer’s genomic lesions, providing insights into the best strategies for treatment. However, these approaches will be undermined by cancer’s ability to resist upfront target inhibition (intrinsic resistance) as well as, in cases where tumors are initially sensitive, develop resistance over the course of drug treatment (acquired resistance). The literature to date reveals a problem as complex as the tumor itself; the heterogeneity of cancer as a disease is matched by the myriad ways in which it evades treatment.

Understanding drug resistance as a whole quickly becomes a problem of scale. Not only is there cancer subtype-associated variation to consider, but also intrinsic and acquired resistance profiles can differ based on the type of inhibitor used and at what node the offending pathway is inhibited. Assigning proper treatments to account for these mechanisms adds an additional layer of complexity as the number of FDA-approved and late-stage clinical candidate molecules increases. Here, when applied appropriately, high throughput methods offer the ability to screen thousands of perturbations in parallel, quickly narrowing the search space for a phenotype of interest.

This work applies such methods to the cell-autonomous complexity of drug resistance and seeks to understand (1) mechanisms by which cancer cells evade drug treatment, (2) design concepts for the most effective combinatorial drugging strategies, and (3) how it might be possible to account for resistance-associated heterogeneity by targeting the evolutionary liabilities of resistant cells. Using a combination of open reading frame (ORF), clustered regularly spaced short palindromic repeats (CRISPR), and pharmacologic screening technologies, this work attains the resolution and throughput necessary to address 1-3 above and begins to unravel the complexity of drug resistance.

Cancer is a diverse set of diseases characterized by genetic and epigenetic alterations that permit growth across diverse environmental contexts. The last decade has led to an explosion of sequencing efforts to define the molecular drivers of proliferation across cancers. This effort has led to the development of small-molecule inhibitors that can block oncogenic drivers and the signaling pathways driving growth. These so called “targeted therapies” have led to better progression-free survival in patients. Despite this early success, it has become clear that with few exceptions, all patients treated with targeted therapies will ultimately relapse. Thus, there is an imminent need to define combination strategies that can be employed to suppress intrinsic or acquired resistance in cancer. Here, we combine functional screening approaches, both pharmacological and genetic, to define apoptosis-inducing precision cancer therapies. Specifically, we utilize a pharmacological screening approach to uncover that breast cancers rely on Mcl-1 and Bcl-XL for survival, and that we can leverage mTOR’s translational control over Mcl-1 to induce apoptosis in PIK3CA mutant breast cancers. Additionally, we utilize CRISPR-Cas9 loss-of-function screening to define the landscape of therapeutic cooperativity in KRAS -driven cancers across diverse tissue types. Further, we leverage this landscape to define principles to rationally design combination therapies to suppress resistance. Lastly, in an effort to define targeted therapeutic strategies for cancers that lack traditional oncogenic drivers, we utilized a pharmacological screening approach to define vulnerabilities associated with dysregulated mitochondrial dynamics proteins in cancer. Collectively, our work has demonstrated the power of functional screening approaches to define apoptosis-inducing anti-cancer precision therapies that combat intrinsic and acquired resistance.

The era of widely applicable sequencing and genomic technologies led to the generation of many large-scale datasets exploring genomes, transcriptomes, or epigenomes of tumors. Availability of a wide range of datasets necessitated the development of new computational analysis approaches to generate novel insights from these datasets and improve our understanding of tumor development and progression.

This work focuses on a variety of graph-based approaches to evaluate their use in cancer genomics. We first focused on a graph-based semi-supervised learning approach called label propagation as a method to generate signaling networks from a gene set of interest. A distance metric based on the concept of maximal common subgraph was then established to quantify the degree of similarity observed across different networks. These two approaches were then combined to examine two separate cancer genomics datasets. Our first application focused on genes frequently altered across patients to build signaling networks that represent genes and pathways that are transcriptionally altered as a result of these mutations. These networks revealed the range of molecular events affected by each mutation and conserved changes observed across networks highlighted the critical signaling pathways tumors dysregulate through distinct alterations. The other area of focus for label propagation was the analysis of a set of melanoma samples resistant to BRAF inhibitors. Evaluation of networks of individual resistant samples revealed signaling changes shared across samples that have similar resistance mechanisms or originated from the same patient. Finally, drug response profiles of a large set of drugs were examined across cell lines belonging to eighteen different tumor types, by building bipartite graphs representing sensitivity patterns of drugs. These bipartite graphs were used to generate drug similarity graphs that revealed shared response profiles of drugs targeting distinct processes, which provided opportunities to refine the annotations of drug targets. Degree distributions of bipartite graphs also revealed drugs connected to exceptional responder cell lines, whose unique genomic profiles nominated potential markers of drug response. Collectively, the studies discussed here emphasize a variety of use cases for graph-based approaches in cancer genomics.

AbstractCancer cells are marked by dysregulated metabolism. Recently, it is increasingly appreciated that oxidative phosphorylation (OXPHOS) is upregulated in certain cancers and can even contribute to resistance to cancer therapies, nominating OXPHOS inhibition as a potential cancer therapeutic target. While a novel electron transport chain (ETC) complex I inhibitor yielded promising results in vitro and in vivo in particular genetic and lineage-dependent contexts, the inhibitor ultimately failed in a high profile clinical trial due to dose-limiting toxicities incurred by targeting a metabolic node that ultimately is fundamental in non-cancerous cells. However, complex III inhibitors are proceeding through clinical trials, indicating that inhibition of the ETC at these nodes can be a safe therapeutic option in the right contexts. Here, we sought to determine the landscape of ETC complex dependences across cancers. We observed general dependence on complex I across cell lines, while dependence on complexes II through V widely varied. In particular, hematological cancers exhibited heightened dependence on complex II. In order to determine what drives this dependence, we performed unbiased computational analyses connecting chemical and genetic inhibition of complex II to identify co-essential genes and pathway. These analyses identified de novo purine synthesis as a strong and unique correlate of sensitivity to complex II. Subsequent metabolomics and mechanistic studies found that complex II directly regulates purine levels and purine synthesis to maintain proliferation. We further linked the oxidation of glutamate, which is generated by rate-limiting glutamine-consuming steps in purine synthesis, as a necessary function of complex II. When complex II is inhibited, glutamate cannot be oxidized and causes negative feedback on glutamine to glutamate steps during purine synthesis to suppress the pathway. Together, these studies have identified an unexpected and fundamental role of complex II in regulating purine synthesis.

The human epidermal growth factor receptor 2, or HER2, is overexpressed in 20-30% breast cancer patients and is associated with aggressive disease. Therapies targeting HER2, including monoclonal antibodies (trastuzumab and pertuzumab), a small molecule kinase inhibitor (lapatinib) and an antibody-drug conjugate (trastuzumab emtansine), have significantly prolonged the overall survival of HER2-positive breast cancer patients. However, almost all patients develop resistance either from the beginning of therapy or with prolonged treatment in two years.

Previous studies to unveil the resistance mechanisms were mainly focused on acquired resistance, culturing cells with HER2 inhibitors and making comparisons to their parental cells. In order to study the mechanism mediating intrinsic resistance, we conducted a loss-of-function genetic screen using a HER2-amplified cell line that is intrinsically resistant to HER2 inhibitors with the purpose to identify synthetic lethal targets. TALDO1, a gene encoding a metabolic enzyme in the non-oxidative pentose phosphate pathway was identified from the screen. Metabolic profiling with isotope-labeled glucose was used to understand the mechanism. The profiling results indicated that TALDO1 was necessary for cellular NADPH generation to combat increased cellular ROS and support synthesis of lipids as a result of HER2 inhibition.

Importantly, the higher expression of TALDO1 is associated with poor response to HER2-targeted therapy in a small cohort of HER2-positive breast cancer patients, suggesting it could potentially serve as a biomarker to predict patient response.

Together our study explained a novel mechanism mediating intrinsic resistance to HER2 inhibition with significant clinical value. Combined inhibition of HER2 signaling and the pentose phosphate pathway may result in a better clinical outcome.

Targeted therapies rarely yield complete tumor responses, and the residual cancer cells that survive upfront treatment act as a reservoir from which eventual resistant disease emerges. Here, we explore several clinically relevant models of disease resistance, with special attention placed on KRAS-driven colorectal cancer (CRC) and EGFR-driven non-small-cell lung cancer (NSCLC).

First, we note that KRAS mutations drive resistance to diverse targeted therapies, including EGFR inhibitors in colorectal cancer (CRC). Through genetic screens, we unexpectedly find that mutant HRAS, which is rarely found in CRC, is a stronger driver of resistance than mutant KRAS. This difference is ascribed to common codon bias in HRAS, which leads to much higher protein expression, and implies that the inherent poor expression of KRAS due to rare codons must be surmounted during drug resistance. In agreement, we demonstrate that primary resistance to cetuximab is dependent upon both KRAS mutational status and protein expression level, and acquired resistance is often associated with KRASQ61mutations that function even when protein expression is low. Finally, we show that cancer cells upregulate translation to facilitate KRASG12-driven acquired resistance, resulting in hypersensitivity to translational inhibitors. These findings demonstrate that codon bias plays a critical role in KRAS-driven resistance and provide a rationale for targeting translation to overcome resistance.

Next, we demonstrate that targeted therapies induce DNA double strand breaks and consequent, ATM-dependent DNA repair in tumor cells that survive upfront treatment. This DNA damage response, observed in both laboratory models and human patients, is driven by a pathway involving the sub-lethal activation of executioner caspases 3 and 7 and the downstream caspase-activated DNase (CAD). As a consequence, tumor cells that survive upfront treatment harbor a synthetic dependence on ATM, and combined treatment with targeted therapies and a selective ATM kinase inhibitor eradicates these cells, leading to more penetrant and durable responses in in vitro and in vivo models of EGFR-mutant NSCLC. Finally, rare patients with EGFR-mutant NSCLC harboring co-occurring, loss-of-function mutations in ATM show evidence of extended progression-free survival relative to patients lacking deleterious ATM mutations. Together, these findings establish a rationale for the mechanism-based integration of ATM inhibitors alongside existing targeted therapy paradigms.

Combined, these studies provide mechanistic-based rationale for pharmacological targeting of tumor-specific processes that may overcome intrinsic and/or acquired resistance states, serving as potential novel therapeutic options for genetically defined subsets of cancer patients.

Calcium/calmodulin-dependent protein kinase kinase 2 (CaMKK2) is a calcium-activated regulator of energy homeostasis in many cell types, including neurons and hepatocytes. In these cells, CaMKK2 connects calcium signaling to several ubiquitous energy and metabolism pathways such as AMPK, CaMKI, and CaMKIV. Tumor cells often express CaMKK2 ectopically, commandeering this enzyme to promote survival, proliferation, and metastasis across a wide variety of cancers. Recent work has established additional roles for CaMKK2 in tumor immunity. Notably, expression of this enzyme promotes M2 polarization of tumor-associated macrophages and facilitates the development of myeloid-derived suppressor cells, indirectly promoting tumor growth. Thus, there has been considerable interest in developing CaMKK2 inhibitors, including competitive kinase antagonists and ligand-directed degraders (LDDs), as potential cancer therapeutics. Indeed, the classic competitive inhibitor STO609 has been shown to reduce primary breast tumor growth in mice. However, the role of CaMKK2 in most cellular compartments within the tumor immune environment remains unknown, which is a significant impediment to the clinical development of these agents. To develop a complete understanding of the effects of CaMKK2 inhibition on the tumor environment, we set out to systematically delineate which cells express CaMKK2 and for what purpose.

Our most significant finding is that natural killer (NK) cells upregulate CaMKK2 expression in the tumor environment, conferring a fitness advantage that improves anti-tumor immune activity. We report that CaMKK2 is expressed at low levels in NK cells isolated from murine spleens but that its expression is dramatically increased in NK cells isolated from tumors or exposed to tumor conditions. A series of in vitro functional assays showed that CaMKK2 expression suppresses apoptosis and promotes proliferation of NK cells without affecting the cytotoxic efficacy of individual cells. These findings were highly intriguing as NK cells play a critical role in the anti-tumor immune response, especially in controlling metastatic disease. As such, we demonstrated that NK cell-intrinsic deletion of CaMKK2 increases metastatic progression across several murine models, establishing a critical role for this enzyme in NK cell tumor immunity. Interestingly, ablation of the CaMKK2 protein, but not inhibition of its kinase activity, resulted in decreased NK cell survival, a result which reveals a unique scaffold function for this enzyme. Finally, we identified lactic acid as a key driver of CaMKK2 upregulation under tumor conditions, an interesting finding given the immune suppressive nature of lactic acid. Intracellular importation of lactic acid through the MCT channels, and not general acidification, is necessary for CaMKK2 upregulation in NK cells. These results suggest that CaMKK2 upregulation in tumor-infiltrating NK cells is an adaptive mechanism by which these cells mitigate the deleterious effects of a lactate-rich tumor environment.

This work will inform strategies to manipulate the CaMKK2 signaling axis as a therapeutic approach in cancer. Because CaMKK2 acts as a scaffold in NK cells, competitive inhibitors and LDDs are likely to have distinct therapeutic utilities. Our work suggests that competitive inhibitors may be broadly preferable in the clinical setting, and we aim to test this hypothesis in mouse tumor models once suitable LDDs become available. We are also working to identify components of the CaMKK2 scaffold complex and elucidate the downstream pathways, which may yield further insights into how this pathway can be manipulated in NK cells. Moreover, because intracellular lactate import is necessary for CaMKK2 upregulation, there may be liabilities associated with the emerging use of MCT1 inhibitors in cancer therapy, which we intend to explore in mouse models. Our findings also raise the possibility that NK cells can be modified to constitutively express CaMKK2 as a proactive shield against suppressive tumor factors, and we are working to replicate our core findings in human NK cells. Overall, our work holds significant implications for the therapeutic manipulation of tumor immunity.

In 2019, the cornerstone of the clinical management of acute myeloid leukemia (AML) is chemotherapy. In fact, for the majority of patients diagnosed with AML, their best chance at achieving complete remission is a regimen of cytotoxic chemotherapies approved more than three decades ago. Most of those patients will eventually succumb to their cancer. This bleak reality belies an abundance of preclinical and clinical drug development in the myeloid leukemia space. The problem that faces many oncologists today is not whether we have potent, selective drugs, but rather, how our armamentarium of effective chemotherapeutics should be best deployed. To adequately address this question, a detailed understanding of drug mechanism and cellular response alike is required. Unfortunately, drugs tend to be more complicated and less specific than physicians are willing to consider; cancer cells tend to be more adaptable and durable than pharmacologists and medicinal chemists think they are. Our inability to effectively bridge this gap has led some to declare that cancer will never be cured and that targeted therapies, the christened ‘silver bullet’ in the fight against cancer, to be another false hope.

This dissertation seeks to reconcile the nuances of drug mechanism with the adaptability of cancer cells using functional, loss and gain-of-function genomic screens to annotate a series of gene-drug interactions which can then be recast to address the question(s) at hand. Because gene-drug interactions are inherently functional, the data produced by such studies can often be rapidly translated to clinically relevant models. Here we provide four examples, presented as discrete case studies, that illustrate the ability of gene-drug interactions to predict trends in tumor evolution, nominate synergistic drug combinations, and approximate disparate fields of biology. Taken together, the studies presented here highlight the potential for using functional genomics to more fully map the effects of drugs on cancer cells.

In 2019, the cornerstone of the clinical management of acute myeloid leukemia (AML) is chemotherapy. In fact, for the majority of patients diagnosed with AML, their best chance at achieving complete remission is a regimen of cytotoxic chemotherapies approved more than three decades ago. Most of those patients will eventually succumb to their cancer. This bleak reality belies an abundance of preclinical and clinical drug development in the myeloid leukemia space. The problem that faces many oncologists today is not whether we have potent, selective drugs, but rather, how our armamentarium of effective chemotherapeutics should be best deployed. To adequately address this question, a detailed understanding of drug mechanism and cellular response alike is required. Unfortunately, drugs tend to be more complicated and less specific than physicians are willing to consider; cancer cells tend to be more adaptable and durable than pharmacologists and medicinal chemists think they are. Our inability to effectively bridge this gap has led some to declare that cancer will never be cured and that targeted therapies, the christened ‘silver bullet’ in the fight against cancer, to be another false hope.

This dissertation seeks to reconcile the nuances of drug mechanism with the adaptability of cancer cells using functional, loss and gain-of-function genomic screens to annotate a series of gene-drug interactions which can then be recast to address the question(s) at hand. Because gene-drug interactions are inherently functional, the data produced by such studies can often be rapidly translated to clinically relevant models. Here we provide four examples, presented as discrete case studies, that illustrate the ability of gene-drug interactions to predict trends in tumor evolution, nominate synergistic drug combinations, and approximate disparate fields of biology. Taken together, the studies presented here highlight the potential for using functional genomics to more fully map the effects of drugs on cancer cells.