All presentations are limited to 15 minutes plus 5 minutes for questions (see the program at this link).
Title: Feedback Loops Between Oncogenes & Tumor Suppressor Genes and their Control
Speaker: Baltazar D. Aguda, PhD (DiseasePathways, LLC; Bethesda, MD, USA)
Abstract: Interactions between oncogenes (such as Myc and Ras) and tumor suppressor genes (such as p53 and Ink4a) normally form negative feedback loops. Cancer stem cells may arise when certain steps in these interactions are perturbed. I will first give an overview of the Myc-p53 interactions in glioblastoma and the Ras-Ink4a interactions in pancreatic cancer. A qualitative network model of the Myc-p53 system is then discussed, including an illustration of how the system becomes unstable and how it can be controlled.
Qualitative network modeling of the Myc-p53 control system of cell proliferation and differentiation. Aguda BD, Kim Y, Kim HS, Friedman A, Fine HA. Biophys J. 2011 Nov 2;101(9):2082-91.
Aguda BD, “The Significance of the Feedback Loops between KRas and Ink4a in Pancreatic Cancer,” in Molecular Diagnostics and Therapy of Pancreatic Cancer (Ed: Asfar Azmi). Elsevier Academic Press (2014). http://store.elsevier.com/Molecular-Diagnostics-and-Treatment-of-Pancreatic-Cancer/Asfar-Azmi/isbn-9780124081031/
Title: Full deconvolution of clonal populations in recurrent hematological cancer using Gaussian Mixture Model
Authors: Davide Cittaro, Dejan Lazarevic, Cristina Toffalori, Elia Stupka, Luca Vago
Abstract: We performed Whole Exome Sequencing on five samples collected in eight years during the disease history of in a single patient with hematological cancer (Acute Lymphoblastic Leukemia, Remission and therapy-related Myelodysplastic Syndrome at three different time-points). We identified a joint set of 1201 variants that were present at least in one stage of the disease, for which we calculated allele frequencies corrected by the estimate of tumor purity. These were used to train a set of Gaussian Mixture Models (GMM) allowing for increasing number of classes. We selected the best model using Akaike Information Criterion (AIC) and assigned each mutation to a specific GMM clone. We calculated mutational signatures of each GMM clone according to the procedure proposed by Alexandrov  and estimated distance between them. We identified a single clone common to all relapses but not LLA, we built parent/child relationships with other clones using distance among signatures. Our analysis reveals a branching evolution of clones that putatively diverge in response to the clinical treatment. We show GMM is an effective and unbiased technique that can be applied to deconvolve clonal populations in cancer data only using allele frequencies. Clones can be then characterized by their mutational landscape, this information is sufficient to build clonal relationships when studying the evolution of the disease.
1. Alexandrov, L. B. et al. Signatures of mutational processes in human cancer. Nature 500, 415–421 (2013).
Title: Definition of a computational pipeline for multi level metabolic analysis.
Abstract: Design principles of metabolism have been investigated exploiting several computational frameworks. Among these, constraint-based methods, and in particular Flux Balance Analysis (FBA), have proven to be useful and accurate to calculate the flux of metabolites through reactions of a metabolic network. Despite of this, constraint-based methods alone have not been able to explain mechanics of events and their temporal evolution, suggesting that other in silico methods should be applied. Due to the complex nature of biologic processes, in silico methods should consider multiple approaches to investigate systems. Multi level analysis is to- day a hot research topic in different areas, such as the theoretical formalization of the method and the development of computational tools for the integration of different modeling perspectives. For this reason we are developing a compu- tational pipeline able to perform analyses exploiting, one after the other, three main modeling frameworks for biological systems: constraint-based analysis, network analysis and mechanism-based analysis. Results emerging from the performed analyses will give a synoptic vision of the different properties of the system. In order to validate the developed method and the computational pipeline proposed, we defined a core model of the cellular metabolism in yeast . The first section of the pipeline is a constraint-based analysis performed via FBA techniques maximizing or minimizing a certain physiological aspect; crucial for this task is the optimization of the objective function achieved exploiting ensemble approaches and genetic algorithms. The second part combines results from FBA with network analysis in order to highlight emergent and general properties of the system. In this context we developed a hierarchical clustering analysis exploiting a dendrogram to illustrate how solutions obtained with the genetic algorithm cluster together. Moreover we integrated in this step a visualization of the fluxes on the network and its topological analysis exploiting Cytoscape and the CyFluxViz plug in. The last part, currently under implementation, is devoted to the retrieval of kinetic constants from fluxes and to the mechanistic simulation of the system in order to allow further investigations using methods such as parameter sweep and sensitivity analysis. On the whole, the main goal of this multiple analyses is gaining, from each ex- ploited model formalization (constraint-based, network-based and mechanism- based), a different kind of information in order to widen the knowledge on the system under evaluation.
 Damiani C., et al. ”An ensemble evolutionary constraint-based approach to under- stand the emergence of metabolic phenotypes.” Natural Computing 13.3 (2014): 321- 331.
Authors: Laura Curti, Domenico Albino, Cecilia Dallavalle, Carlo V. Catapano and Giuseppina M. Carbone. Institute of Oncology Research (IOR) and Oncology Institute of Southern Switzerland (IOSI), Bellinzona, 6500 Switzerland.
Abstract: The ETS transcription factor ERG controls multiple epigenetic effectors in prostate tumors Cancer of the prostate is a leading cause of cancer death worldwide. About 30-50 % of prostate cancers harbor ETS gene rearrangements, the most frequent being the TMPRSS2-ERG gene fusion. However, the role of ERG in prostate cancer progression is still debated. Oncogenic activity of translocated ERG may involve broad transcriptional and epigenetic reprogramming in fully transformed cells. We showed previously that ERG induces directly the expression of the histone methyltransferase EZH2, which is highly expressed in advanced and metastatic prostate cancers and promotes epigenetic gene silencing and dedifferentiation. In line with this finding, we reported that another epigenetic effector, UHRF1, was overexpressed in prostate tumors and contributed to epigenetic silencing of tumor suppressor, tissue-specific differentiation and androgen-regulated genes. UHRF1 expression was frequently associated with EZH2 upregulation and negatively correlated with expression of several tumor suppressor genes. We found that overexpression of UHRF1 was associated with ERG expression in prostate tumors. Consistently, expression of ERG in ERG negative LNCaP cells induced UHRF1 mRNA and protein level. We identify an ETS binding site in the UHRF1 promoter and demonstrated selective binding of ERG by chromatin immunoprecipitation. Thus, UHRF1 is an additional relevant target of ERG with a potentially important role in prostate tumorigenesis. This study uncovers novel epigenetic mechanisms by which ERG fusion can lead to prostate cancer progression.
Cecilia Dalla Valle
Title: MicroRNA-424 promotes transformation and stemness and is associated with aggressive prostate tumors.
Authors: Cecilia Dallavalle, Domenico Albino, Gianluca Civenni, Paola Ostano, Laura Curti, Giovanna Chiorino, Carlo V.Catapano, Giuseppina M. R. Carbone. Institute of Oncology Research (IOR), Bellinzona, Switzerland; Laboratory of Cancer Genomics, Fondazione Edo ed Elvo Tempia, Biella, Italy
Abstract: MicroRNAs play important roles in cell proliferation, differentiation and self-renewal regulating gene expression at a post-transcriptional level. To understand mechanisms controlling prostate epithelial cell differentiation and transformation, we profiled microRNA (miRNAs) expression in tissue samples of human primary prostate tumors (n=45) and normal prostate (n=21). miR-424 was significantly over-expressed in tumours compared to normal prostate and more robustly upregulated in a subgroup of tumours characterized by reduced level of ESE3 and aggressive features (ESE3low tumours). Consistently, miR-424 was at the top list of the miRNA overexpressed in prostate epithelial cells with stable ESE3 knockdown (PrEcESE3kd). Notably, prostate tumours with high miR-424 expression were enriched of epithelial-to-mesenchymal transition (EMT) and cancer stem cell (CSC) transcriptional features, reminiscent of the presence of similar features in ESE3low tumours. MiR-424 was upregulated broadly in other epithelial cancers including gastric, lung and breast cancer. Next, we showed by chromatin-immunoprecipitation (ChIP) that ESE3/EHF directly controlled miR-424 by binding to an ETS binding site in the pri-miRNA promoter and repressing its transcription. Consistent with an oncogenic role of miR-424, inhibition of miR-424 with an antimiR reduced anchorage-independent growth and cell migration. In contrast, stable expression of miR-424 increased anchorage-independent growth and cell migration in cells with low endogenous level of miR-424. Modulation of miR-424 expression affected cancer stem cell properties reducing in vitro prostatospheres formation and the fraction of CD44high/CD24low cells. Furthermore, inhibition of miR-424 in metastatic prostate cancer cells reduced tumor initiation and metastatic spread in vivo. Collectively, these results show for the first time the activation and oncogenic properties of miR-424 in prostate tumors. Targeting miR-424 may be a valid approach to revert stemness in prostate tumours.
Title: Causes and Effects in Analyses of Mutation Rates
Abstract: Mutation is one of the fundamental driving forces of cancer. But are mutations random, spontaneous events, or are they facilitated by stresses, such as carcinogens? I will review mathematics and experiments devised by Luria and Delbrück to answer this question. I will describe experiments I have done using E coli as a model system and point out how biological irregularities can be incorporated into the Luria-Delbrück model. In particular, I will focus on the case study of antibiotic resistance in bacteria, which can arise spontaneously but is also “encouraged” by exposure to antibiotics themselves. Finally, I will discuss how high-throughput sequencing might be used to calculate, in an unbiased way, mutation rates for cells grown under different conditions, as well as the implications of such a method for cancer research.
Title: TO-DAG: a new graph-based timed model for cumulative cancer progression
Authors: Paola Lecca, Nicola Andrea Casiraghi, Francesca Demichelis
Abstract: The order and the timing at which the somatic alterations occur during cancer progression reveal important information on the underlying biological process with implications for diagnosis, prognosis and treatment. As latest high-throughput technologies provide base level resolution data, the cancer research community gains access to unprecedented comprehensive datasets of genomic alterations in human cancers and prompt to the development of computational models able to capture the process of mutation accumulation with as little assumption as possible. We present a novel computational method named “Timed Oncogenetic Directed Acyclic Graph” (TO-DAG) that infers the graph of the causal dependencies and the waiting times among mutational events from cross-sectional data of genomic alterations in independent human tumor samples. TO-DAG can process very large datasets, does not require a priori assumptions about the order and the timing of mutation event, overcomes the limitations of stochastic memoryless process based methods and allows for computation of conditional probability of multiple events (i.e. it is not limited to pairwise dependencies). Namely, TO-DAG computes the probability of occurrence of each alteration in a pathway as the probability that the alteration occurs when all alterations prior to it have occurred therefore inferring pathways of causal dependencies among genetic alterations reflecting more closely the real non-memoryless dynamics of the mutation accumulation during cancer formation. Once the causal structure of the graph is inferred, the waiting times of the mutation events are estimated as stochastic function of their conditional probability. We present the performance on synthetic data and the networks of causal relationships inferred among mutations affecting genes in prostate cancer  and in melanoma  and discuss them in the light of current knowledge in the genomics of those tumors.
 C. E. Barbieri, S. C. Baca, M.S. Lawrence,F. Demichelis, M. Blattner, J. P. Theurillat, et al. (2012). Exome sequencing identifies recurrent SPOP, FOXA1 and MED12 mutations in prostate cancer. Nat Genet, 2012, 44 (6), 685-689.
 M. Berger,E. Hodis,Y. L. Deribe, M. S. Lawrence, A. Protopopov, E. Ivanova, E., et al., Melanoma sequencing reveals frequent PREX2 mutations. Nature 2012 (485).
Title: Non-coding RNA-based regulation of gene expression in normal and cancer cells
Authors: Giuseppina Pisignano, Sara Napoli, Ramon Garcia-Escudero, Giuseppina M. Carbone, Carlo V. Catapano
Abstract: A variety of epigenetic events, such as DNA methylation, histone modifications and chromatin remodelling, take place during initiation and progression of human cancers. Transcriptional silencing of tumor suppressor genes, like the E-cadherin encoding gene (CDH1), is very frequent in human cancers. Loss of E-cadherin expression triggers epithelial-to-mesenchymal transition (EMT) and acquisition of tumor-initiating properties in epithelial cells. Reduced expression of E-cadherin is associated with tumor progression and poor clinical outcome in many epithelial cancers. However, what drives transcriptional silencing of tumor suppressor genes, like CDH1, is still an open question. Emerging evidence suggest that long non-coding RNAs (lncRNAs) are important players in epigenetic mechanisms and may have relevant roles in transcriptional reprogramming and tumorigenesis. In this study, we investigated the role of promoter-associated lncRNAs (paRNAs) in transcriptional silencing of tumor suppressor genes. paRNAs are defined as lncRNAs originating within a few hundred bases of transcription start sites of a protein-coding gene and have been proposed to act as docking elements for recruitment of epigenetic regulators to the neighbouring gene. We found that bidirectional transcription from distinct initiation sites in the CDH1 promoter generated sense (S) and antisense (AS) paRNAs. The level of S and AS paRNAs dictated the chromatin state and transcriptional activity of the gene. Both in human tumors and cell line models the prevalence of S-paRNAs and low AS/S ratio were associated with low CDH1 expression. The S-paRNA coordinated transcriptional gene silencing by recruiting Argonaute1 and the H3K9 histone methyltransferase SUV39H1 to the CDH1 promoter. Consistently, siRNA-mediated depletion of S-paRNA reactivated CDH1 transcription in low CDH1 expressing cancer cells. This resulted in profound inhibition of cell proliferation and clonogenic capacity. This study reveals a complex RNA-based regulatory network that relies on sequence-specific interactions of paRNAs with Argonaute1 and epigenetic effectors to coordinate transcriptional gene silencing. These findings give also a new prospective and insights into the mechanisms of gene regulation identifying paRNAs as relevant elements in these processes and potential targets for gene modulation strategies.
Title: inferring causal models of cancer progression with cross-sectional data
Comprehensive knowledge of cancer progression is of vital importance for diagnostics, prognostics and the development of targeted therapies. Towards this goal, a huge amount of genomic information has been collected in the last couple of decades from tumor samples. From this information, a set of specific driver events (e.g., genetic mutations, CNVs, etc.) has been identified as relevant in each specific progression. However, despite this flood of information, relatively little is known about the dynamics of cancer progression and the order in which these driving events are likely to occur. The main reason for this state of affairs is that information is usually obtained only at one (or a few) points in time (i.e. cross-sectional data), rather than over the course of the disease. It is an important challenge to extract the essential dynamic information from the available static data. In this presentation, I will describe a novel algorithm for inferring cancer progression from cross-sectional data named CAPRI (CAncer PRogression Inference), which combines insights from several fields, including algorithms in machine learning, theory of causality, and cancer biology.