Kevin Janes, Ph.D. - US grants

DESCRIPTION (Provided by the applicant) Abstract: Cancer is a stochastic disease whose biology has been studied almost exclusively with deterministic approaches. Why? In this application, I propose to exploit the apparent randomness of cellular transformation to uncover new mechanisms involved in tumorigenesis. My focus is the ligandless receptor tyrosine kinase, ErbB2, which is overexpressed in 20-30% of breast cancers and is the target of anticancer drugs such as Herceptin(r) and Tykerb(r). In a 3D in vitro culture model of mammary-acinar morphogenesis, inducible activation of ErbB2 causes hyperproliferative multiacinar structures that in many ways are reminiscent of early-stage breast tumors. Importantly, the penetrance of the phenotype is incomplete-only a random fraction of the cultured acini exhibit the morphogenetic defect when ErbB2 is activated. How this fraction is specified and the mechanism by which a multiacinus initiates are unknown. My hypothesis is that acute differences (dichotomies) in gene expression develop among acini and give rise to the distinct 3D phenotypes induced by ErbB2. The transcriptional dichotomies that exist before the appearance of the multiacinar phenotype will be the ones most likely to control it. However, without seeing the phenotype, it is impossible to know which ErbB2 structures will go on to develop abnormally. To overcome this challenge, we will use a new technique, called "stochastic profiling", that I developed for discovering transcriptional dichotomies in a seemingly uniform cell population. We will apply stochastic profiling to a series of conditional ErbB2 homo- and heterodimer pairs that have different penetrances for the multiacinar phenotype. By mapping the transcriptional dichotomies to the differences in penetrance among dimer pairs, we will link upstream acinus-specific expression programs to downstream morphogenetic heterogeneities. The results from this project could explain mechanistically why only a fraction of ErbB2- overexpressing breast cancers respond positively to ErbB2-targeted therapeutics. Public Health Relevance: Several modern cancer drugs target the ErbB2 protein, but these drugs are effective in only a fraction of cancers that express ErbB2. The research in this proposal combines novel experimental and statistical approaches in an attempt to identify new cancer genes that may be turned on only occasionally by ErbB2. Such genes could link ErbB2 to drug sensitivity, providing new avenues for more-effective anti-cancer therapeutics.

DESCRIPTION (provided by applicant): Coxsackievirus B3 (CVB3) is a leading cause of chronic heart inflammation and heart failure in children and young adults. Persistent CVB3 infection is not associated with the mature virus but instead with latent expression of its RNA genome in cardiomyocytes. Our preliminary observations suggest that the presence of CVB3 RNA disrupts how cardiomyocytes normally respond to inflammatory stimuli, which may be important for establishing a chronically inflamed state. The objective of this proposal is to defin the molecular rewiring that takes place in cardiomyocytes upon chronic infection with CVB3. Our hypothesis is that persistent expression of CVB3 RNA disrupts the posttranscriptional signaling response of cardiomyocytes to proinflammatory stimuli. During the R21 phase, we will develop high-throughput biochemical methods to measure signaling and transcript stability at the systems level. In the R33 phase, these assays will be combined to examine posttranscriptional signal processing in cultured cardiomyocytes expressing CVB3 RNA and stimulated with proinflammatory cytokines. The data will ultimately serve as the basis for a computational-systems model that connects signaling events to gene expression and transcript stability. Aim #1 of the R21 phase is to develop quantitative, systems-level bioassays to measure kinase-phosphatase activation in cardiomyocytes chronically infected with CVB3. Aim #2 of the R21 phase is to develop a real-time, multiplex RNA stability assay for profiling half-lif regulation of NF-?B transcripts in cardiomyocytes. Aim #3 of the R33 phase is to interrogate intracellular dynamics triggered by proinflammatory cytokines and build a predictive, data-driven systems model that captures CVB3-induced cardiomyocyte rewiring. The long-term goal is to use the model to propose novel interventions that can correct posttranscriptional signal processing in patients with persistent CVB3 infection.

? DESCRIPTION (provided by applicant): Solid tumors are a heterogeneous collection of cells with vastly different capacities to proliferate, metastasize, and resist therapy. Although functionl diversity among individual tumor cells is widely recognized, we do not know how many functional states there truly are, nor is it clear how best to catalog those states in the first plce. The global profile of mRNAs expressed by a cell can suggest its state, provided that the measurements are reliable and can be obtained with minimal disruption of the cell in its native context. To date, neither of these criteria has been achieved for solid tumors. We circumvented the problem by developing a new method, called stochastic profiling, which measures small 10-cell pools of cells microdissected in situ to glean single- cell information through statistical analysis. The 10-cell pools increase the starting material and allow reliable expression profiles to be achieved with samples microdissected in situ. Previously, we have used stochastic profiling to uncover a wealth of single-cell functional states in 3D organotypic cultures of breast epithelial cells. In our answer to PQB4, we seek to address whether stochastic profiling can be directly applied to human or murine solid tumors and yield meaningful information about cancer progression. The hypothesis is that progression is linked to a common subset of regulatory states that change in frequency or identity as solid tumors become more advanced. The aims of this proposal are: 1) To evaluate ex vivo regulatory heterogeneities within genetically engineered small-cell lung cancers at various stages of progression. We will use stochastic profiling with premalignant cells and small-cell lung tumorspheres from mice with Trp53 and Rb conditionally deleted in neuroendocrine cells. 2) To evaluate in vivo regulatory heterogeneities within genetically engineered gliomas at various stages of progression. We will use fluorescence-guided stochastic profiling to evaluate gliomagenesis in mice with Trp53 and Nf1 conditionally deleted in oligodendrocyte precursor cells of the olfactory bulb. 3) To test whether regulatory heterogeneities in human tumors are quantitatively predictive of pathologic stage and grade. We will combine stochastic profiling of breast tumors with partial least squares regression to link single-cell regulatory states to clinical parameters. If successful, this application would set the stage for a long-term goal of identifying all major categories of regulatory heterogeneity in solid tumors. To characterize the functional state of individual tumor cells in context, the answer may be to avoid measuring single cells entirely.

PROJECT SUMMARY/ABSTRACT Perturbation of cancer cells often leads to heterogeneous outcomes, in that most cells exhibit a dominant phenotype, but the rest appear resistant or hypersensitive to the perturbation. If the penetrance of such a phenotype is heritably incomplete, then it becomes extremely difficult to decipher the upstream molecular events that heterogenize the population and cause response variability. By combining quantitative measurements with dynamical models, systems approaches should be useful if provided with a core network of important biomolecules. The daunting hurdle lies in identifying phenotype-relevant regulatory heterogeneities that define the network for penetrance at the single-cell level. Our proposal seeks to exploit a new approach, called stochastic frequency matching (SFM), for elaborating the molecular networks upstream of incompletely penetrant phenotypes. SFM identifies and parameterizes single-cell heterogeneities?which emerge after a uniform perturbation but before the appearance of a variable phenotype?to hone in on regulatory states corresponding to future penetrance. For an onco-phenotype incompletely triggered by ErbB receptor tyrosine kinase signaling in 3D cultured breast epithelia, we implemented SFM using microarrays to uncover a network of critical nucleocytoplasmic regulators. The goals of this proposal are to apply systems approaches to the ErbB nucleocytoplasmic network and adapt SFM more broadly to RNA sequencing of breast cancer patients with ErbB amplification. Based on our provisional SFM results, we hypothesize that ErbB signaling heterogeneously reconfigures the nucleocytoplasmic shuttling state of cells to determine incomplete penetrance of the onco-phenotype. The aims are to: 1) Identify network-level mechanisms for the incompletely penetrant ErbB1:ErbB2 phenotype. 2) Determine whether drivers of incomplete penetrance in 3D define shuttling states in human cancers and promote ErbB2-driven mammary tumors in mice. 3) Sequence and parameterize regulatory-state heterogeneity in HER2+ breast cancers to assemble patient-specific network models of shuttling variability and sensitivity. Drivers of incomplete penetrance are important for understanding transitions during tumor initiation-progression and for developing therapeutic interventions with more reliable patient outcomes. SFM gives the Cancer Systems Biology Consortium a means to identify driver networks in a comprehensive and hypothesis-driven way.

PROJECT SUMMARY/ABSTRACT Perturbation of cancer cells often leads to heterogeneous outcomes, in that most cells exhibit a dominant phenotype, but the rest appear resistant or hypersensitive to the perturbation. If the penetrance of such a phenotype is heritably incomplete, then it becomes extremely difficult to decipher the upstream molecular events that heterogenize the population and cause response variability. By combining quantitative measurements with dynamical models, systems approaches should be useful if provided with a core network of important biomolecules. The daunting hurdle lies in identifying phenotype-relevant regulatory heterogeneities that define the network for penetrance at the single-cell level. Our proposal seeks to exploit a new approach, called stochastic frequency matching (SFM), for elaborating the molecular networks upstream of incompletely penetrant phenotypes. SFM identifies and parameterizes single-cell heterogeneities?which emerge after a uniform perturbation but before the appearance of a variable phenotype?to hone in on regulatory states corresponding to future penetrance. For an onco-phenotype incompletely triggered by ErbB receptor tyrosine kinase signaling in 3D cultured breast epithelia, we implemented SFM using microarrays to uncover a network of critical nucleocytoplasmic regulators. The goals of this proposal are to apply systems approaches to the ErbB nucleocytoplasmic network and adapt SFM more broadly to RNA sequencing of breast cancer patients with ErbB amplification. Based on our provisional SFM results, we hypothesize that ErbB signaling heterogeneously reconfigures the nucleocytoplasmic shuttling state of cells to determine incomplete penetrance of the onco-phenotype. The aims are to: 1) Identify network-level mechanisms for the incompletely penetrant ErbB1:ErbB2 phenotype. 2) Determine whether drivers of incomplete penetrance in 3D define shuttling states in human cancers and promote ErbB2-driven mammary tumors in mice. 3) Sequence and parameterize regulatory-state heterogeneity in HER2+ breast cancers to assemble patient-specific network models of shuttling variability and sensitivity. Drivers of incomplete penetrance are important for understanding transitions during tumor initiation-progression and for developing therapeutic interventions with more reliable patient outcomes. SFM gives the Cancer Systems Biology Consortium a means to identify driver networks in a comprehensive and hypothesis-driven way.

PROJECT SUMMARY/ABSTRACT Roughly 85% of triple-negative breast cancers are categorized as basal-like or claudin-low carcinoma, molecular subtypes with especially poor prognosis and limited treatment options. Triple-negative breast cancers frequently harbor mutations in DNA-surveillance pathways; consequently, their overall genomic heterogeneity has been extensively characterized. By comparison, much less work has been done on the cell biology of triple-negative breast cancer. Despite the recognized histological nonuniformity of triple-negative tumors, we have only a rudimentary inventory of the types of signaling and transcriptional regulatory states that single basal-like and claudin-low cells can adopt. The long-term goal of this work is to identify and characterize the major cell-to-cell regulatory heterogeneities in triple-negative breast cancer. The current application focuses on growth-differentiation factor 11 (GDF11), a diffusible factor that is heterogeneously regulated in 3D organotypic cultures of claudin-low breast epithelial cells. Functional GDF11 bioactivity is lost in clinical cases of advanced triple-negative breast cancer, and addition of GDF11 to invasive claudin-low and basal-like cancer lines strongly suppresses invasion into basement membrane ECM. The hypothesis is that GDF11 acts a local breast-epithelial cue for proper lobular architecture, which is suppressed nongenetically during triple-negative breast cancer progression. The aims of this proposal are: 1) To identify the signaling and transcriptional mechanisms that mediate GDF11-induced phenotypes in triple-negative breast cancer. 2) To define the key steps of GDF11 misregulation in triple-negative neoplasms. 3) To determine the impact of GDF11 on progression and metastatic colonization of triple-negative tumors. The diversity of regulatory states enables triple-negative breast cancer cells to switch and adapt rapidly during tumor progression and the evolution of drug resistance. A complete inventory of regulatory states and their transitions could one day be harnessed by novel therapies that reset intratumor regulatory heterogeneity to delay progression or resistance.

The success of machine learning (ML) in many applications where large-scale data is available has led to a growing anticipation of similar accomplishments in scientific disciplines. The use of data science is particularly promising in scientific problems involving processes that are not completely understood. However, a purely data-driven approach to modeling a physical process can be problematic. For example, it can create a complex model that is neither generalizable beyond the data on which it was trained nor physically interpretable. This problem becomes worse when there is not enough training data, which is quite common in science and engineering domains. A machine learning model that is grounded by explainable theories stands a better chance at safeguarding against learning spurious patterns from the data that lead to non-generalizable performance. This is especially important when dealing with problems that are critical and associated with high risks (e.g., extreme weather or collapse of an ecosystem). Hence, neither an ML-only nor a scientific knowledge-only approach can be considered sufficient for knowledge discovery in complex scientific and engineering applications. This project is developing novel techniques to explore the continuum between knowledge-based and ML models, where both scientific knowledge and data are integrated synergistically. Such integrated methods have the potential for accelerating discovery in a range of scientific and engineering disciplines. This project will train interdisciplinary scientists who are well versed in such methods and will disseminate results of the project via peer-reviewed publications, open-source software, and a series of workshops to engage the broader scientific community.

This project aims to develop a framework that uses the unique capability of data science models to automatically learn patterns and models from data, without ignoring the treasure of accumulated scientific knowledge. Specifically, the project builds the foundations of knowledge-guided machine learning (KGML) by exploring several ways of bringing scientific knowledge and machine learning models together using pilot applications from four domains: aquatic ecodynamics, climate and weather, hydrology, and translational biology. These pilot applications were selected because they are at tipping points where knowledge-guided machine learning can have a transformative effect. KGML has the potential for providing scientists and engineers with new insights into their domains of interest and will require the development of innovative new machine learning approaches and architectures that can incorporate scientific principles. Scientific knowledge, KGML methods, and software developed in this project could potentially be extended to a wide range of scientific applications where mechanistic (also known as process-based) models are used.

This project is part of the National Science Foundation's Harnessing the Data Revolution (HDR) Big Idea activity.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

PROJECT SUMMARY The regulation and mis-regulation of sphingolipid metabolism in acute myeloid leukemia (AML) occurs at multiple bio-molecular levels?gene mutation, enzyme abundances and activities, as well as sphingolipid precursors and products. The innovative aspect of the Program Project has systematically collected deep biomolecular profiles of accrued patient samples and relevant AML cell lines. The new Systems Metabolomics Core (Core C) will help the larger P01 project team to generate testable hypotheses by integrating the systems-level datasets from individual research projects and the other research cores. The goal is to inform critical Project-level decisions involving sphingolipid-targeted therapies. Core C will pursue the following Specific Aims: 1) evaluate linear and nonlinear predictive models of outcome and response based on patient- specific biomolecules and (cyto)genetics; 2) test data-driven predictive models of outcomes and response based on patient-specific, gene?transcript?lipid?activity signatures; and 3) integrate gene, transcript, and metabolite profiles of AML patients with a reconstructed sphingolipid module of the human metabolic network. Core C leadership is comprised of experts in data-driven cancer modeling, metabolic network reconstruction, and systems biology. Therefore, the Core is poised to adapt its approaches to the changing needs of the individual research projects over the second term of the P01.