Ernest Fraenkel, Ph.D. - US grants

Through a grant from the National Science Foundation to the Massachusetts Institute of Technology, six faculty members will collaborate in the purchase and use of high-performance computing equipment for research and education in computational and systems biology. The advent of high-throughput technologies in the life sciences has provided many genome sequences, protein structures and biological interaction networks. The amount of such data will continue to grow, compelling the development of rigorous and quantitative approaches to decipher and understand it. Simultaneously, advances in computing technology are enabling new ways of attacking complex biological problems using modeling and simulation. The projects to be supported cover a wide range of exciting areas, including the study of gene and organism evolution, transcriptional and post-transcriptional gene regulation, molecular signaling, protein conformational modeling, protein design, and the analysis of complex networks. The work will lead to advances in computational methods and provide basic biological insights.

This award will support MIT?s active role in developing computational and systems biology in the United States. The university is establishing novel programs and curricula to train students at the interface of the life sciences, engineering and the physical sciences. The investigators on this award are deeply involved in these activities. Shared computing resources will help attract talented students and provide them with modern, cross-disciplinary training. These students, who come from diverse backgrounds, will assume leadership positions in American universities and companies.

DESCRIPTION (provided by applicant): We propose a Systems Biology approach to map transcriptional regulatory networks and related signaling pathways that are altered in Huntington's disease (HD), a fatal autosomal dominant neurodegenerative disorder. HD is caused by a CAG expansion leading to a polyglutamine extension in the huntingtin protein and is characterized by problems with movement, cognition and behavioral function. Although the genetic basis for the disease is clear, the mechanism by which huntingtin causes the observed symptoms remains enigmatic. Our approach is based on the hypothesis that many of the genes previously linked to HD through proteomic and genetic screens are connected through signaling pathways to many of the transcriptional changes that have been reported in HD studies. Identifying these pathways would provide critical new insights into the molecular changes that underlie the disease, and could lead to novel therapeutic strategies. We have recently developed a technique for identifying such signaling pathways through a combination of computational and experimental methods. In Specific Aim 1 we will map out changes in recruitment of transcriptional regulatory proteins because these proteins lie at the interface between the signaling and expression changes. In Specific Aim 2 we will computationally identify signaling changes "upstream" of these regulators that link the transcriptional changes to the genetic and proteomic data. If successful, this approach will advance knowledge of the etiology of HD and provide a powerful new method for studying many human diseases. PUBLIC HEALTH RELEVANCE: We propose a systems biology approach to understanding the molecular changes that occur in Huntington's disease. This project uses high-throughput experiments and computational modeling to reveal pathways that are altered in the disease and that may lead to new therapeutic approaches.

Administration Component The Administrative Core is responsible for setting the overall direction of the NeuroLINCS center and for ensuring that the resources and components of the Center are optimally utilized. The successful development and evolution of the NeuroLINCS center requires strong interactions between the leaders and co-leaders of each Component and of the center as a whole. Hence, the NeuroLINCS Administrative Component plays a vital role in facilitating these interactions. Moreover, the Administrative Component and its personnel provide the necessary administrative and fiscal oversight to ensure that the NeuroLINCS center is run efficiently. The NeuroLINCS center involves 5 principal sites with defined responsibilities of growing, differentiating and generating new induced pluripotent stem cell lines (iPSCs), performing data generation assays on human brain cells made from iPSCs in response to perturbagens, performing basic analyses and developing cell signatures through integrated data analysis .methods, and establish community interactions. An integrated and highly collaborative group of investigators with expertise in stem cell biology, IPS cells, quantitative molecular phenotyping (omics and single cell imaging) and bioinformatics will work closely together to generate significant and highly predictive cell signatures. The PIs of the NeuroLINCS center are Steven Finkbeiner (Gladstone), Ernest Frankel (MIT), Jeffrey Rothstein (JHU), Clive Svendsen (Cedars) and Leslie Thompson (UCI), who will serve as leaders and co-leaders of components. Each Component has identified co-investigators/collaborators/consultants appropriate for the planned scientific investigations. Component leaders and co-leaders will also be active participants in NeuroLINCS consortium working groups as they are developed to address specific issues. Results of the genetic, proteomic and other characterization conducted by consortium labs will provide important feedback for further enhancement of induction and differentiation protocols and related methodologies and it is anticipated that this collaborative and iterative approach will lead to the broadest success for the study. An Evaluation Program within the NeuroLINCS is in place to determine if the programs supported are meeting the needs of the research community, are efficiently managed, and demonstrably effective and annual objectives and milestones.

Community Interactions & Outreach Component The NeuroLINCS Community project will plan to provide resources and tools for a broad user base of basic and clinical scientists. It has a structure to facilitate access to the various genetic and proteomic data sets, the signatures created, and the analysis tools. It is designed to be directed to researchers at the bench, clinicians developing biological disease readouts and those in computational roles. It will incorporate an assessment to demonstrate the utility of the generated resources, methodologies, and analytical tools to LINCS and non-LINCS scientific community. Importantly, it will develop and implement a plan to bring in external collaborators who may have data sets that bear on the development of cell signatures. There is an extensive plan to develop workshops, tutorials, and symposia in conjunction with the use of innovative online technologies for disseminating information to target the major LINCS goals. Finally it will develop bidirectional links with the neuroscience clinical and basic community through a series of collaborations with large National clinical data and tissue-based networks.

Data Analysis & Signature Generation Component Our goal is to generate cellular signatures of human neurons in response to perturbagens. Our studies will focus on human neurons, generated from induced pluripotent stem cells (iPSCs) (i-neurons) obtained from both healthy people and patients with neurodegenerative diseases. The cellular signature will be a composite picture of the molecular properties of a neuron that distinguish the state and determine the behavior of the cell. We will generate three classes of cellular signatures. The first will be static signatures based on quantitative molecular phenotyping involving OMIC analysis of the i-neurons. Analysis of the static signatures will highlight critical signaling pathways that distinguish a cellular response to a perturbagen. The second type of signature will be dynamic signatures generated with a novel high throughput, single cell longitudinal analysis system. Robotic Microscopy (RM). RM will be able to pinpoint critical times in the life of i-neurons as their physiology change in response to perturbagens. Analysis of dynamic signatures will guide selection of time points that will be investigated more in depth with methods that generate static signatures. In turn, elements of these static signatures will be perturbed genetically and analyzed by RM to elucidate the epistatic relationship of the components of a signature and to develop explicit multivariate predictive descriptions of cellular responses to perturbations. The third type of signature will emerge from an integration of the individual signatures using clustering methods and machine learning algorithms. The technology to analyze the data of the cellular signatures will be compatible with those produced at other sites in the LINCS network. A major innovation of our program is the implementation of novel data analysis platforms that will produce signatures that will have greater predictive value of a cell's biology than standard technologies. We will integrate Data Analysis and Data Generation, creating feedback loops to allow the cellular signatures that we generate to influence subsequent data generation. In turn, the use of machine learning algorithms in collaboration with Google will allow us to iteratively refine our signatures to make them more predictive in identifying cause and effect relationships from the cellular signatures.

Data Generation Component We propose three broad experimental aims based around the type of assay, perturbagens and technologies being applied that will overlap across the years. The first will use iPSCs from three disease states (non affected, SMA and ALS) in which we have shown specific phenotypes. We will use an iterative approach by first screening for a number of perturbagens of interest to the broad neuroscience community using cost effective assays including simple cell death models and a highly novel imaging analysis system. The second parallel effort will be to use the same iPSC lines but in this case test a set of known cell modifiers (Glutamate, ER stressor and SOD1 ASO) as perturbagens and perform massive parallel quantitative molecular phenotyping (QMP) to generate robust signatures and to define the responses of motor neuron cultures to these perturbagens. We will then perform QMP on neurons, astrocytes and oligodendrocytes from disease and control cells and in response to the same perturbations as above to elucidate signatures across broadly relevant neural cell types. This data will be compared to motor neuron cultures (where expected disease signature will be) with non motor neuron cultures (where no or a more restricted disease signature is expected) to resolve the question of cell type specificity. We will also generate new iPS lines from post mortem human patient tissues to allow clinical pathological signatures to be incorporated into the LINCS data, providing a unique resource to both the SMA and ALS scientific community and to researchers interested in larger questions relating to the CNS. The third is to bring in disease iPS lines from Huntington's and Parkinson's subjects (from the respective NIH consortia and in coordination with various foundations - see letters of support. Overall) and test the specificity of signatures seen in the motor neuron diseases with other neurodegenerative conditions (both disease and response to perturbagens). All of these studies will be done in close association with the data analysis component community section (being responsive to the needs of the community). Given the speed of discovery in iPSC and new molecule generation, we also aim to be flexible in our design to allow incorporation of breakthrough technologies or drugs should they arise.

DESCRIPTION (provided by applicant): We propose an innovative, systems biology approach to uncover new therapeutic strategies for childhood embryonal tumors. Our project is a collaboration between labs in two separate Integrative Cancer Biology Program (ICBP) centers and a leading hospital-based translational research lab that is not within the ICBP network. Embryonal tumors are the most common central nervous system malignancies in childhood, and there is a pressing need for better therapies. Current survival rates range from 30 - 80%, and nearly all survivors have impaired neurological and neurocognitive function. Extensive genomic analysis of medulloblastomas, the most common embryonal tumors, failed to identify driver genes that could explain the origin of most tumors or suggest new strategies. Nevertheless, these tumors can be grouped into a small number of subtypes that share transcriptional patterns and clinical outcomes. We believe that it is time for a fundamentally new approach that seeks oncogenic driver pathways rather than driver genes. As many different genomic changes can all affect the same driver pathway, such pathways cannot be uncovered by looking for recurring genomic changes. Rather, we will use a systems biology approach to identify these oncogenic driver pathways. We will collect comprehensive datasets in human medulloblastoma tumors and cell lines by measuring mutations, copy number variations, mRNA expression, miRNA expression and epigenomic data. We will then construct network models identifying shared pathways altered across many patients within a subtype. Finally, we will functionally test driver pathways nominated from the network modeling. By merging these diverse genomic and transcriptional data collected from tumors of individual patients, we will have an unprecedented ability to uncover the root causes of cancer, providing new therapeutic strategies. The collective expertise of our collaboration provides a unique environment for solving this critical barrier in cancer, by combining strengths in analyzing genomic data, modeling signaling pathways and transcriptional regulatory networks and clinical expertise in embryonal brain tumors. Together, we will generate and merge all types of transcriptional, genomic and epigenomic data, extract biologically-relevant network models and experimentally validate novel drug targets.

DESCRIPTION (provided by applicant): There is a critical need to define the state and predict the behavior of human brain cells in health and disease. The number of different cell types in the CNS remains undefined, and despite a demographically ordained wave of neurodegenerative diseases, not a single disease-modifying therapy exists. Our knowledge of the CNS and the foundation for intervening rationally in disease would be dramatically advanced by generating quantitative molecular phenotypes essentially cell signatures of human neurons, astrocytes and oligodendrocytes from healthy people and from patients with motor neuron disease, Huntington's disease, and Parkinson's disease. The CNS is so unique that studying non-neuronal cells does not provide much assistance. Despite this desperate need, the inaccessibility of human brain cells meant studying them would have been impossible until the recent discovery of cellular reprogramming and induced pluripotent stem cell technology. Here we propose to form the NeuroLINCS consortium to accomplish these goals. We have handpicked the team to bring in critical expertise in iPSC technology, disease modeling, transcriptomics, epigenomics, metabolomics, whole genome sequencing, proteomics, high content, high throughput longitudinal single cell analysis, other cell-based assays, bioinformatics, statistics and computational biology. In addition, we are collaborating with Google to bring in special expertise in machine learning and the integration of signatures across platforms into highly predictive models of responses to perturbagens. Together, we expect to develop cell signatures of an array of human brain cell types under different conditions that should be broadly applicable to the LINCs community. We also anticipate generating innovative software tools and approaches that will make the signature generating process cheaper, faster, and more reliable. Besides the unique combination of expertise represented within NeuroLINCS, another distinguishing feature is the long track record that its members have of collaborating with each other. That collaborative spirit will be expressed in NeuroLINCS through its significant and multifaceted community outreach programs. These will involve specific and detailed plans to make the data and tools that NeuroLINCS generates available to the community, to interact with other LINCS sites, and to prepare for DCIC and the prospect of disseminating knowledge and resources at scale.

The simple genetic cause of Huntington?s disease contrasts starkly with the vast number of pathways that are affected by the mutation. Some of these pathway-level changes may persist even if the mutated allele of the disease-causing gene (HTT) can be corrected through gene therapy or related methods. During the first granting period, our analysis of HD models identified several potential therapeutic directions, including ones closely tied to epigenetics (the transcriptional regulators NEUROD1, WNTand ELK-1), as well as pathways that interact with epigenomic changes (energy metabolism and lipid biochemistry). Some of these effects were restricted to particular cell types in the brain. We also found evidence that mutant HTT (mHTT) expression causes neurodevelopmental impairments, changing the distribution of cell types in the brain. We and others have also identified a significant number of genetic variants in the human population for which there is significant support for an impact of that variant on HD age of onset (AOO). In the current proposal, we examine the therapeutic potential of interventions based on these findings. We will target these pathways in mice, measuring how interventions alter transcription, the epigenome, signaling and metabolomics. A critical innovation is our use of single-cell and spatially resolved methods to examine how responses to mHTT and therapeutics vary among different types of cells. Equally important, we will differentiate specific cell types from induced-pluripotent stem cells (iPSC) in vitro to examine cell-type specific effects in human cells. Using an approach based in systems biology we will look for common pathways that are affected by the genetic AOO modifiers, the candidates from our prior grant period and leads from the literature. Our approach is highly innovative, as it uses cutting edge experimental methods with single-cell and spatial resolution to reveal aspects of HD that cannot be detected in homogenates. We also computationally integrate multi-omic data (genomics, epigenomics, transcripts, proteins and metabolites) from the individual cells and brain regions to uncover therapeutic pathways. The research is highly significant, as it seeks to guide therapeutic discovery for an invariably fatal neurodegenerative disease. We expect that the impact of our work will extend beyond HD, by providing a model for how to measure and model cell-type specific neurodegeneration to identify therapeutic approaches.

Alzheimer's disease is the most common neurodegenerative disorder and is characterized clinically by cognitive dysfunction and pathologically by the formation of extracellular amyloid plaques and intraneuronal deposition of aggregated tau into neurofibrillary tangles. Although rare forms of the disorder are caused by highly penetrant mutations in autosomal dominant genes, the pathogenesis of more common forms of the disease remains incompletely understood. A thorough understanding of the basic mechanisms driving loss of neuronal integrity during aging would provide a crucial underpinning for efforts focused on identifying the pathways mediating neurodegeneration in Alzheimer's disease and related disorders. Thus, to provide a comprehensive knowledge of mechanisms driving brain degeneration in higher eukaryotes we will take advantage of the power of forward genetics in Drosophila to outline, in an unbiased fashion, mechanisms controlling preservation of neuronal function during aging. Then, to relate our findings to human disease directly we will integrate, using a systems biology approach, networks derived from eQTL analysis and RNA sequencing data from a unique and high-quality resource of laser captured temporal neurons from patients with Alzheimer's disease and carefully age- and sex-matched control patients without neurological disease. We will further discover pathways relevant to disease by performing an integrated metabolomics and phosphoproteomic analysis in Drosophila models relevant to Alzheimer's disease, namely human tau and Aß transgenic animals. The resulting networks, including previously undiscovered, or hidden, nodes will be tested for their causal relationship to neurodegeneration in Drosophila in well-characterized models of tau and Aß neurotoxicity. Our studies will thus discover on a genome scale novel mechanisms driving neurodegeneration and will provide a census of those mechanisms most likely to underlie cell death in Alzheimer's disease and related disorders. The genes and pathways we discover can then be examined in mechanistic detail in the appropriate mammalian models.

SUMMARY We propose developing a systems-biology approach to understand interactions between tumor and immune cells and their clinical implications. Our work will focus on medulloblastoma, a malignant pediatric brain tumor in which our team has extensive expertise. We and others have shown that medulloblastoma tumors are sites of immune activity despite the blood-brain barrier. However, the clinical consequences of these immune cells are unclear and there is little information that might guide development of therapeutics that modulate these immune cells. Our innovative strategy combines single-cell methods, including single-cell proteomics, with a sophisticated computational analysis. In Aim 1, we map the landscape of tumor-immune interactions using sequencing and imaging methods on human samples. Aim 2 builds a causal model of the molecular interactions that govern interactions among cell types in medulloblastoma, determines the clinical correlates of these cells, and identifies potential therapeutic targets. Aim 3 maps the tumor-immune environment in well-validated mouse models of the disease, and builds computational models for mice parallel to those for humans. Hypotheses from Aim 2 that are likely to translate well to the mouse models are then tested for their effects on tumor growth and survival. The mouse results are used to update the computational models and refine the therapeutic strategies. We expect that successful completion of this project will have a substantial impact on medulloblastoma therapeutics. Further, the methods we develop will catalyze research of interactions between immune cells and many other tumor types beyond medulloblastoma.