Phenotype Transitions in Small Cell Lung Cancer
- Phenotype Transitions in Small Cell Lung Cancer
The goal of our U01 project to develop rational systems approaches to treat cancer. Small-cell lung cancer (SCLC) is arguably the most aggressive form of lung cancer, and while modern treatment strategies are initially highly effective, tumors quickly relapse, resistant to further therapies. We seek to decode the gene-regulatory processes that underlie SCLC’s phenotypic response to treatment, and develop clinically viable strategies to reprogram those circuits, while keeping cells locked into drug-sensitive states. Our major driving questions are: How does SCLC heterogeneity contribute dynamic adaptation to treatment? What regulatory mechanisms govern this response, and how are they coordinated? How can these regulatory mechanisms be manipulated to drive SCLC cells toward therapeutically desirable phenotypes? What molecular species can be targeted therapeutically to affect these mechanisms? To this end we will integrate bioinformatics, mathematical modeling, and experimentation to provide a systems-level understanding of SCLC drug response and optimal treatment strategies.
Vito Quaranta, M.D.
Vito Quaranta, M.D. is the Director of the Quantitative Systems Biology Center, and Professor of Cancer Biology at Vanderbilt University School of Medicine. For over a decade he has been implementing cutting-edge interdisciplinary effort melding mathematics, engineering, computation and biology to solve the problem of cancer invasion and metastasis. The Quaranta laboratory applies both theory and experimentation to frame cancer as a complex adaptive system that responds to perturbations, such as drug treatment, by evolving over multiple biological and temporal scales. The laboratory is comprised of a mix of experimentalists, engineers, statisticians, and mathematicians combining experimental and modeling approaches iteratively towards a systems-level understanding of cancer. A major focus of the laboratory is the development of single-cell methodologies to evaluate the mechanism of action of targeted therapy in cancer, based on the merging of automated time-lapse microscopy with image analysis and computational modeling. Another is the dynamics of transcription factor and signaling networks that define and maintain cell identity, and ultimately contribute to forming the phenotypic landscape of the tumor microenvironment. A recent achievement was the development of methods that measure drug-response dynamics of heterogeneous cancer cell populations as they emerge from single-cell behavior.
Carlos F. Lopez, Ph.D.
Carlos F. Lopez, Ph.D. is an Assistant Professor in the Department of Cancer Biology at Vanderbilt University. He is a computational biologist with a distinguished record in promoting accessibility of mathematical models to biologists and developing new tools and approaches. His group is the main contributor to the PySB modeling framework, a shareable programming language environment that enables encoding biological processes as computer programs for automatic mathematical representation, simulation, and analysis. He is an active member of the Vanderbilt-Ingram Cancer Center, a member of the Faculty Board for computational resources at Vanderbilt (ACCRE), and a member of the Steering Committee for the Initiative to Maximize Student Diversity (IMSD). He is also a member of the Vanderbilt Quantitative Center for Systems Biology, the Center for Structural Biology, and the Institute for Chemical Biology.
Our project is articulated into two Aims:
Identify key transcriptional regulators of phenotypic transitions and drug-sensitivity in neuroendocrine (NE), mesenchymal-like (ML), and hybrid SCLC phenotypes.
The goal of this Aim is to reverse-engineer the regulatory circuits underlying SCLC heterogeneity and drug response, and develop intervention strategies to control the circuit’s dynamics. TFs induce and maintain cell identity and can be used to reprogram cells by activation and/or inhibition of small groups of TFs (such as in induced pluripotent stem cells). In previous work (Udyavar et al, 2016), we showed that canonical neuroendocrine/epithelial (NE) and mesenchymal-like (ML) SCLC phenotypes correspond to attractors (stable states, as in the bottoms of energy wells) of a core TF regulatory network, as well as identifying novel “Hybrid” single-cell phenotypes (Figure 1). Here, we will expand our TF network to describe these hybrid phenotypes, and integrate data from multiple sources including human tumors and tissue microarrays (TMAs). Based on publically available transcriptomic datasets and TMAs, we employ bioinformatic techniques including Consensus Clustering, WGCNA, ARACNe, and Gabi (manuscript in preparation), as well as database mining, to identify molecular processes characterizing heterogeneous SCLC phenotypes, and uncover networks for TFs which regulate expression of those processes. Single-cell experiments are used to validate the phenotypic character of individual SCLC cells.
Boolean simulations of this TF network will reveal underlying regulatory mechanisms which stabilize single-cell heterogeneous SCLC phenotypes, and provide an in silico representaiton which we will explore to find cocktails transcription factors which reprogram cells from untreatable attractors into treatable ones. This is a computationally expensive task which must take into account uncertainty in the network topology, TF regulatory interaction rules, and stochastic effects in order to prioritize the most likely candidate reprogramming targets. We have therefore developed GPU/CPU hybrid algorithms to address this need and run the simulations in high-performance computing environments.
The final outcome of this Aim will be a mechanistic model that describes the regulation of NE, ML, and hybrid phenotypes based on TF gene regulatory dynamics, that will guide further exploration and experimental validation.
Identify signaling perturbations that promote SCLC shifting to phenotypes maximally susceptible to killing under drug treatment.
While Aim 1 seeks to uncover underlying regulatory mechanisms of SCLC phenotypic heterogeneity and drug response, it is well-known that TFs are difficult, and often impossible, to target directly in the clinic. Here we exploit the fact that intracellular signaling pathways, which are commonly targeted in the clinic, can be used to indirectly manipulate the activity of TF networks.
To attain this goal, we will first identify and experimentally validate signaling pathways which are heterogeneously active across distinct SCLC phenotypes. Heterogeneous WGCNA gene co-expression modules are found which are enriched for expression of distinct signaling pathways (Table 1). CyTOF will be used to validate the activity of these at the single-cell level, providing a high resolution map of SCLC signaling pathway activity. Using this map, we will track the single-cell response of SCLC cells to drug treatment. Our group has developed a metric of drug sensitivity based on the net Drug Induced Proliferation rate (DIP rate), which we will use to estimate the effects of selection for distinct phenotypes. Coupled with this, we have implemented a Stochastic Phenotype Transition Proliferation (SPTP) model to quantify relative roles of selection and plasticity, by tracking single-cell drug response population shifts. We will complement this with unsupervised and semi-supervised logistic regression algorithms to link molecular drivers to cellular drug response. This information, will enable us to prioritize optimal combinations of chemotherapy, paired with signaling perturbations which drive cells toward states sensitive to that specific chemo-agent. These combinations will be tested in vitro, and the best candidates will be validated in in vivo mouse models.
With the completion of this Aim, we will have developed of a map of SCLC phenotypic heterogeneity, coupled with signaling-induced transition paths and chemo-agent sensitivity. This map will provide clinically actionable signaling drivers which can push SCLC cells toward states which maximize chemotherapy efficacy.