Stanford CCSB Seminar Series

Dynamics of Cancer Cell Response to Molecularly Targeted Therapeutics

Friday December 9, 2011 at 11 am (Li Ka Shing Center Room LK130)
Vito Quaranta, Professor, Cancer Biology, Vanderbilt University

Most studies on molecularly Targeted Therapeutics (TT) are genetic, molecular or
clinical. In contrast, cellular responses are less well characterized. For instance, the
popular GI50 assay limits measurements of TT-response to a fixed time point (~72 hrs.),
with little information on cell behavior before and after, and to cell population averages,
with little information on variable contributions from single cells. To fill this knowledge
gap, we constructed a high-throughput pipeline for tracking single-cell responses to
drug perturbations over several days at 6-min intervals (Extended Time Resolved
Automated Microscopy), and coupled it to image processing and mathematical models
for extracting information from these rich datasets.

Our initial studies focused on three oncogene-addicted cell lines, lung adenoCA PC9,
melanoma A375 and breast CA HCC1954 treated, respectively, with the TT erlotinib
(mutant EGFR), PLX-4720 (V600E mutant B-Raf), and lapatinib (overexpressed Her2).
In each case, the dynamics of TT response revealed a nonlinear process produced by
multiple cell behaviors (cell division, quiescence, and death) that occur simultaneously
but at different rates. Overall, inhibition of cell population growth was predominantly due
to a majority of cells entering quiescence, with only a modest increase in death rate.
Strikingly, a few cells continued to divide even after prolonged exposure to TT drugs.
Tracking the fate of single cells within progeny trees indicated that death, continued
division, or quiescence in response to TT may or may not be maintained within siblings,
suggesting a stochastic component to cellular outcome variability. The molecular basis
for this complex behavior is being characterized by single-cell and population analyses
(partially in collaboration with the Stanford and MIT CCSB, respectively). As this
information accrues, we hope to develop predictive tools to improve TT clinical
performance and to circumvent TT resistance.

Human Acute Myeloid Leukemia: Identification, characterization and targeting of tumor-propagating populations

Friday November 18, 2011 at 11 am (Li Ka Shing Center Room LK101)
Paresh Vyas, Professor, Hematology, University of Oxford, UK

Human Acute Myeloid Leukemia (AML) is the most common aggressive leukemia with a poor outcome. It provides one of the best, and most intensively, studied models of cancer. Identification of recurrent karyotypic abnormalities initially, and then driver mutations and epigenetic changes from focused studies and then using genome wide approaches is providing a near complete genetic description of the bulk tumor. These genetic and epigenetic changes have highlighted the role of nuclear regulators and signal transduction abnormalities in oncogenesis. From a clinical standpoint, genetic markers now define prognosis and response to therapy and have provided evidence of intertumoral molecular heterogeneity.

Despite these scientific advances most AML patients still die of disease often after an initial response. Therapy fails to eradicate all tumor-propagating cells and indeed in many cases selects for chemotherapy resistant tumor propagating subclones. Given the above, a major focus of our work is the identification and characterization of leukemia propagating cells (LPC) in childhood and adult AML. Studies to be presented: Characterization of LPC's in Adult AML, Tracking AML LPC's From Diagnosis Through Therapy, Stepwise transformation of normal blood stem/progenitor cells to preleukemic and leukemic stem cells based on studies in myeloid leukemia of Down Syndrome.

Whole Genome Sequencing and Analysis of Cancer Genomes

Friday October 14, 2011 at 11 am (Li Ka Shing Center Room LK101)
Elaine Mardis, Associate Professor, The Genome Institute, Washington University

The advent of next-generation sequencing technology has enabled an explosion in cancer genomics--from multi-gene assays for mutation discovery to whole genome sequencing and classification of all types of sequence alterations. My talk will focus on the methods, analyses, challenges and discoveries at our Institute, obtained using whole genome sequencing and analysis of tumor/normal paired genomes.

Stem cell self-renewal and cancer cell proliferation

Wednesday September 14th at 9 am (Clark Center S360)
Sean Morrison, Professor, Center for Stem Cell Biology, University of Michigan

Tumor-initiating cells have been suggested to be rare in many solid cancers but common in melanoma, raising the question of whether melanoma is unique among solid cancers and what parameters influence tumorigenic cell frequency. We tested this in mouse malignant peripheral nerve sheath tumors (MPNSTs). Mouse MPNSTs contained a similarly high frequency of tumorigenic cells irrespective of whether they were transplanted into immuno-competent or immunocompromised mice, demonstrating that as long as transplants were performed among histocompatible mice, immune status was not a major determinant of tumorigenicity. In MPNSTs from Nf1+/-;Ink4a/Arf-/- mice, 18% of primary cells and 49% of serially transplanted cells formed tumors after transplantation. However, in MPNSTs from Nf1/p53+/- mice only 1.8% to 2.6% of primary or serially transplanted cells formed tumors, demonstrating that the tumor genotype influenced the frequency of tumorigenic cells that could be detected in some assays. Culture on laminin increased the frequency of tumorigenic MPNST cells from Nf1/p53+/- mice, but not from Nf1+/-;Ink4a/Arf-/- mice, to 19% by binding ß1-integrin-containing receptors. Cells with tumor-forming potential are therefore common in mouse MPNSTs, but tumors of different genotypes, even with the same malignant phenotype or histopathologic classification, require different assay conditions to detect the full range of cells capable of contributing to cancer progression. In MPNSTs, exposure to laminin is a major determinant of whether cells are able to form tumors irrespective of whether the laminin is intrinsically or extrinsically produced.

Cancer Genomics and the TCGA project

Friday April 22, 2011 at 11 am (Li Ka Shing Center Room LK120)
David Haussler, Professor, Biomolecular Engineering, UCSC

Large-scale cancer genomics projects such as The Cancer Genome Atlas (TCGA) and the International Cancer Genomics Consortium (ICGC) will sequence tens of thousands of tumors in the next few years, along with matched normal blood samples and other normal tissues. TCGA alone plans to analyze 500 clinically characterized samples from each of 20 different cancer types, detecting frequently mutated genes, common copy number variants, altered gene expression and methylation changes. These data will provide an exceptional resource for identifying new diagnostic targets and predictors of response.

Building on technology developed for the UCSC genome browser, as a data analysis center for TCGA we have developed a cancer genome analysis pipeline and a cancer genomics browser (genome-cancer.soe.ucsc.edu) to interpret cancer genomics data that will aid in the identification of new targets. We reconstruct changes in tumor genomes from tumor sequencing data and use a new approach called PARADIGM, based on factor graphs, to map multiple data types into a single coherent pathway model including thousands of genes and interactions for higher-level interpretation. By transforming raw genomic data to pathway activity levels, PARADIGM provides a comprehensible window into the data that can be coupled to predictors of response to improve accuracy. We discuss the early results of applying this approach to the TCGA data.

Integrative Genomics and Public Data Framework for Basic Research and Clinical Investigations

Friday March 25, 2011 at 11 am (Li Ka Shing Center Room LK130)
Ilya Kupershmidt, Cofounder, VP Products, NextBio

Public data repositories have established themselves as the biggest source of large-scale genomic data compared to any single commercial or academic entity. With the public data paradigm firmly in place the next major challenge facing the research and medical community is how to harness the collective power of this data to accelerate study of disease and design of novel treatments. To address this problem we developed an integrative genomic framework that enables researchers to investigate genes, sequence regions, pathways, as well as your own experimental data within the context of global public datasets. Our framework enables assembly and mining of microarray and nextgen sequencing data from gene expression, mutation, GWAS, epigenetic and DNA copy-number applications. We will discuss how NextBio integrative genomics technology is applied to basic and clinical research initiatives.

Data-driven Personalized and Systems Medicine

Wednesday February 23, 2011 at 11 am (Li Ka Shing Center Room LK130)
Atul Butte, MD, PhD, Chief, Division of Systems Medicine, Department of Pediatrics

Dr. Butte's lab builds and applies tools that convert the billions of points of molecular, clinical, and epidemiological data measured by researchers and clinicians over the past decade into diagnostics, therapeutics, and new insights into disease. Dr. Butte will highlight how using publicly-available molecular data enables the discovery of new gene variants and biomarkers for diseases like diabetes, suggests novel roles for drugs in the treatment of disease, and for the first time allows us to probe the inner commonality across disease. Dr. Butte will also discuss his recent papers on the clinical evaluation of a personal genome and the environment-wide association study (EWAS).

Predicting Cancer Outcomes with Integrative Pathway Modeling

Friday January 21, 2011 at 11 am (Li Ka Shing Center Room LK101)
Josh Stuart, Associate Professor of Biomedical Engineering, UC Santa Cruz

New technologies allow for the simultaneous genome-wide assay of the state of genome copy number variation, gene expression, DNA methylation, and epigenetics of tumorsamples and cancer cell lines. Studies such as The Cancer Genome Atlas (TCGA), Stand Up To Cancer (SU2C), and many more are planned in the near future for a wide variety of tumors. While different genes may be altered in one patient compared to another, integration at the pathway level can provide a coherent view of cancer signatures. Probabilistic graphical models are well-suited for combining multiple sources of data to infer the tumor-specific activity levels of individual genes and signaling outputs in the context of their interactions. Information about a gene is modeled as a set of interconnected variables. The model can then predict the degree to which a pathway.s activities (e.g. internal gene states, interactions, or high-level .outputs.) are altered in the patient using probabilistic inference. The inferred activities can then be used in downstream machine-learning algorithms to predict clinical outcomes such as disease-free survival or the sensitivity (or resistance) to various drugs. If the models faithfully represent disease mechanism it should be possible to use them to identify additional therapeutic points of attack.