Information Flow at the Systems Level: Organization and Modeling of Experimental Data Across Multiple Scales of Biological Analysis Raimond L. Winslow Center for Cardiovascular Bioinformatics & Modeling Johns Hopkins University Whiting School of Engineering and School of Medicine (www.ccbm.jhu.edu) Outline Objective – develop new methods for risk stratification and treatment of Sudden Cardiac Death (SCD) Data Collection from the Molecular to Organ level Data Organization Integrative Modeling – A tool for understanding the relationships between molecular events (e.g., changes in gene/protein expression, posttranslational modifications of proteins) and function at the cellular and whole-heart levels Heart Failure is the Leading Cause of SCD Heart Failure MR Imaging of Canine Heart Pre- and Post- Failure Mechanical pump failure leading to reduced cardiac output Diverse origins Common end-stage phenotype The primary U.S. hospital discharge diagnosis Incidence ~ 400,000/year, prevalence of ~ 4.5 million 15% mortality at 1 Yr, 80% mortality at 6 Yr leading cause of Sudden Cardiac Death in the US Chamber Dilation Wall Thinning Data Collection Goal: To understand the molecular basis of sudden cardiac death in human heart failure Experiments (Human, Canine, Rabbit) Gene/Protein Expression Microarrays 2D PAGE Mass Spec (MALDI-TOF, TOF-TOF, SELDI) Cell Membrane Cell Transporter ElectroFunction Physiology Heterologous Expression Systems Whole Cell & Patch-Clamp Recording Cardiac Imaging Ventricular Conduction Ca2+, Na+ MR Diffusion Electrode Tensor Arrays &V Imaging NADH, FADH, Spin-Tagging Vmito, Ca2+mito Modeling & Data Analysis Patient Data Data Organization HTML SOAP IBM WebsphereTM (Not Completed) Web Services Integration (IBM MinelinkTM) Data Analysis & Visualization SQL SOAP SOAP Models Database Federation Software (IBM Information Integrator) (HIPPA) MAGE-DB2 Protein-DB2 IMAGING CLINICAL Integrative Modeling: Relating Molecular Mechanisms of Excitation-Contraction Coupling to Cellular and Whole-Heart Function ~ 10 nm From Katz (1992) Physiology of the Heart Bers (2002) Nature 415: 198-205 Ca2+ Release Channels (RyR) { Vm 10 nm Ca2+ L-Type Ca2+ Channel Ca2+ 2+ Trigger Ca Ca2+ Release } The “Calcium Release Unit” (CaRUs) ~ 10 L-Type Channels and 50 RyR ~5,000 such units in the myocyte ~ independendent Ca2+-I >> Voltage-I Common Pool Models of the Myocyte Existing Myocyte Models Iserca2a Existing myocyte models lump all 5,000 CaRUs into single compartment – => “common pool” models Described as systems of ODEs Reconstruct properties of the AP Common Pool Models Reconstruct the AP 40 Experiment 20 Model 40 20 0 0 -20 -20 -40 -40 -60 -60 -80 -80 -100 -100 0 0.1 0.2 0.3 0.4 0.5 00 0.1 100 0.2 200 0.3 300 0.4 400 0.5 500 EC Coupling and Common Pool Models When Ca2+-I >> Voltage-I Experiment Model Membrane Potential Linz & Meyer (1998) J. Physiol. 513(pt 2): 425-442 Mechanism Unstable APs Ca2+ LCCs Model Prediction Unstable APs (Alternans) Ca2+ { Total Ca2+ Release Lack of Graded Release The Common Ca2+ Pool RyR Integrating from Channels to the Cell: The Local-Control Myocyte Model Greenstein, J. L. and Winslow, R. L. (2002) Biophys. J. 83: 2918-2945 Ca2+ Release Unit Ca2+ Flux from NSR (Jtr) Jxfer,i,1 Jxfer,i,2 Jiss,i,1 Ca2+ Flux to Cytosol JSR RyRs (Jrel) (Jxfer) LCC ClCh Jxfer,i,4 Jiss,i,1, J,2iss,i,2, 4 3 Jiss,i,3 Jxfer,i,3 ,4 (ICaL) (Ito2) 1 ICaL : 5 RyR per Functional Unit 4 functional units coupled via Ca2+ diffusion per Calcium Release Unit (CaRU) ~ 12,500 independent CaRU’s per myocyte (=> ~ 50,000 LCCs per cell) Numerically integrate the ODEs defining the myocyte model over steps Dt, while simulating stochastic dynamics of the CaRUs within each Dt RyR Open Fraction Stochastic Simulation Algorithm 12,500 CaRUs Improved pseudo-random number generator (MT19937) with longer period and improved performance Dynamic allocation algorithm for controlling number of CaRUs Parallel implementation, ~ linear scaling ~1 minute per 1 Sec of activity Model can relate channel level events (e.g., phosphorylation) to whole-cell behavior Local Control Myocyte Model Exhibits Graded Release and Stable APs Experiment 4 40 Wier et al (1994) J. Physiol. 474(3): 463-471 Model Action Potentials Altered Expression of EC Coupling Proteins and the Cellular Phenotype of Heart Failure Altered Gene Expression in End-Stage Canine and Human Heart Failure Normal and Failing APs Kaab et al (1996). Circ. Res. 78(2): 262 Yung et al (2003). Genomics. in press online Experiment Genes Encoding K+ Currents Genes Encoding EC Coupling Proteins Failing Normal ~ 66% { KCNJ12 (IK1) ~ 32% Little Effect on AP and Ca2+ Transient ATP2A2 ~ (62%) NCX1 ~ (75%) { KCND3 (Ito1) Major Effect on AP and Ca2+ Transient Greenstein & Winslow (2002). Biophys. J. 83(6): 2918 Model Failing Normal Relating Effects of PKA-Mediated Phosphorylation of ECCoupling Proteins to Cellular Function EC-coupling proteins are believed to be hyper-phosphorylated in the failing heart Targets and actions of PKA-mediated phosphorylation ( 1mM ISO) – L-Type Ca2+ Channels (LCCs) ∙ Increase LCC availability (~ 2 – 2.5x) ∙ Mode-1, 2 re-distribution (~ 15% Mode-2, ~85% Mode-1) ▪ Increased mean channel open time in Mode-2 (~.5 to 5.0 mSec) – Serca2a Pump (ATP2A2) ∙ Serca2a up-regulated by ~ 3x (Simmerman & Jones Physiol. Rev. 78: 921) − IKr ∙ Increased through reduced inactivation (Heath & Terrar J. Physiol. 522: 391) – IKs • Increased ~ 2x (Kathofer et al J. Biol. Chem. 275: 26743) Use the local-control model to understand consequences of this hyperphosphorylation at the cellular level Develop Model Using Data on b1-Adrenergic Agonists Effects on APs and Ca2+ Transients Action Potentials Ca2+ Transients Control Iso (1 mM) “Baseline Model” – Serca2a and K+ current changes – Mode-1,2 redistribution – Increased availability Early After-Depolarizations in Response to LCC Phosphorylation EAD Frequency % Mode 2 # EADs # APs 0 0 100 7.5 2 100 15 5 100 Early After-Depolarizations (EADs) are thought to trigger polymorphic ventricular tachycardia Rate of occurrence of EADs is increased in myocytes isolated from failing hearts No EADs in the absence of Mode 2 gating => rate of EAD generation increases with increased Mode-2 gating EAD Generation is Stochastic Identical initial conditions, but different random number seeds produces different realizations of LCC and RyR state transitions => stochastic gating of LCCs triggers EADs Initiation of Stochastic EADs by Increased Mode-2 Gating Mode 2 Current Mode 1 Current Long Mode-2 open time increases likelihood of clustered random Mode-2 LCC openings Spontaneous, near simultaneous openings of a sufficient number of LCCs gating in Mode 2 generates inward current Resulting depolarization re-activates LCCs gating in Mode 1, producing an EAD Novel hypothesis regarding generation of EADs Integrating from Cell to Ventricular Function: DTMR Imaging of Ventricular Anatomic Structure Diffusion Tensor MR Imaging (DTMRI) x DTMRI 3x3 diffusion tensor Mi(x) Hypothesis – The principle eigenvector of Mi(x) is aligned with fiber direction at point x Fox and Hutchins (1972). Johns Hopkins Med. J. 130(5): 289-299 DTMRI vs HISTO Fiber Angles Holmes, A. et al (2000). Magn. Res. Med., 44:157 DTMRI Fiber Angles In Cross Section Scollan et al (2000). Ann. Biomed. Eng., 28(8): 934-944. Imaging Procedure fixed Myocardium 3-D FSE DTMRI 256 x 256 x 100 imaging volume 350 mm in-plane, 800 mm out-of-plane resolution Fiber orientation estimates at ~ 3 * 106 voxels Finite Element Models of Cardiac Ventricular Anatomy User selects number of volume elements/nodes Matlab GUI for visual control of the fitting process All imaging datasets, FE models, and FEM software are available at www.ccmb.jhu.edu Epicardial Fibers – FEM Model Endocardial Fibers – FEM Model Modeling Electrical Conduction in the Cardiac Ventricles EADs Can Trigger Ventricular Arrhythmias Reaction-Diffusion Equation v( x, t ) 1 1 I ion (v( x, t )) I app ( x, t ) M i ( x)v( x, t ) , x H t Cm b 1 From Ionic Models From DTMRI EADs Trigger Reentry and Polymorpic VT Winslow et al (2000). Ann. Rev. Biomed. Eng., 2: 119-155 “Closing the Loop” on Whole-Heart Experiments and Models 256 Epicardial Electrode Array Measure Electrode Positions MR Image and Model Ventricular Anatomy “Closing the Loop” on Whole-Heart Experiments and Models (cont.) Electrically mapped and DTMR imaged 4 normal and 3 failing canine hearts – 256-electrode sock array, ~ 5mm electrode spacing Complete anatomical and electrical reconstruction performed on one normal canine heart Experiment Model Winslow et al. (2002). Novartis Foundation Symposium 247: In Silico Simulation of Biological Processes, pgs. 129-150, John Wiley & Sons, Ltd. 2002. Summary Use of a “hierarchy of models”, each developed to address problems at different levels of biological organization, is important Individual stochastically gating channels Cell models Tissue/whole heart models The detailed spatial arrangement of ion channels in the cardiac myocyte has a profound effect on cell and whole heart function Stochastic effects at low molecule copy number – ~10 – 100 free Ca2+ ions in the diadic space at the peak of the Ca2+ transient – Continuum models may not be valid – Dynamics of Ca2+ ions become important Importance of the interplay between modeling and experiment – Whole heart models have been used exclusively in the “predictive” mode – Methods now exist for coupling whole-heart experiments and models Acknowledgements Supported by the NIH (HL60133, HL70894, HL61711, HL72488, P50 HL52307, NO1-HV-28180, ), the Falk Medical Trust, the Whitaker Foundation, the D. W Reynolds Foundation and IBM Corporation