Completed Collaboration & Service Projects



1) Relationship between structural Dynamics and Sequence Evolution; Application to molecular machined such as bacterial chaperonons and HSP70 chaperones

PI: I. Bahar, University of Pittsburgh  Collaborators: Amnon Horovitz, Weizmann Institute; R. Altman, Stanford University; L. Gierasch, University of Massachusetts.

aanmWe have expanded the scope of this project to include Dr. Amnon Horovitz from Weizmann Institute, Israel, an expert in the allostery of chaperonins as well as sequence co-evolution. We are exploring the sequence->structure->dynamics->function mapping of allosteric proteins. Our overarching goal is to establish the computational methodology for simulating the machinery of biomolecular systems on the order of Megadaltons. The molecular machines that we selected for developing and implementing the methodology are bacterial chaperonins (GroEL), molecular chaperones (E. coli Hsp70 DnaK) and their co-chaperones. We study the ATP-regulated mechanisms of intersubunit (GroEL) or interdomain (DnaK) communication, as well as the role of co-chaperonins or co-chaperones in modulating the conformational changes occurring along the allosteric cycle of these machines.

In Year 3 we made progress in quantitative analysis of the evolution of proteins, and in particular in assessing the relationship, if any, between their structural dynamics and sequence evolution. In a recent study performed toward this goal (Mao et al., 2015), we carried out a systematic comparative analysis of current methods for analyzing sequence co-evolution and we proposed a hybrid model that yields high performance. The two metrics for performance were the elimination of false positives between non-interacting proteins, and the verification of tertiary contacts between co-evolving pairs. The methods used in this newly published (that acknowledged MMBioS support) are currently being implemented into our ProDy API.

Publications Resulting from This Work

    • General IJ, Liu Y, Blackburn M, Mao W, Gierasch LM, Bahar I (2014) ATPase subdomain IA is a mediator of interdomain allostery in Hsp70 molecular chaperones PLoS Comp Bio 10: e1003624 PMID: 24831085, PMC4022485
    • Mao W, Kaya C, Dutta A, Horovitz A, Bahar I (2015) Comparative Study of the Effectiveness and Limitations of Current Methods for Detecting Sequence Coevolution Bioinformatics 31: 1929-37 PMID: 25697822; PMC4481699



2) Membrane Proteins Structure and Dynamics Consortium (MPSDC) Computational Core, and AMPA receptors structural dynamics

PI: I. Bahar, University of Pittsburgh   Collaborators: B. Roux and E. Perozo, University of Chicago; K. Schulten and E. Tajkhorshid, University of Illinois at Urbana-Champaign;  I. Greger, University of Cambridge, UK; H. Weinstein, Cornell University

To assess whether ANM-predicted AMPAR NTD pivoting motions can occur in cellulo, the Greger lab mutated K262 to cysteine (K262C), hypothesizing that this mutation would trap the NTD tetramer in a closed conformer via formation of a disulfide bridge. GluA2 wild type (WT) and K262C mutant were expressed in HEK293 cells and protein extracts were analyzed by Western blot. The results showed that our hypothesis was valid, thus lending support to computational predictions. We further examined the dynamics of AMPAR and NMDAR to observe that the differ along a few slow modes only (Figure 1). Overall, our data provided a first glimpse into the dynamic spectrum of AMPAR and NMDARs and delineated conserved mechanisms underlying allosteric communication in iGluRs.

csp2Figure 1. AMPAR to NMDAR conformations differ by rearrangements along a few soft modes. The dark blue bars show the overlap (correlation cosine) between ANM modes of AMPAR and the structural difference vector between AMPAR and NMDAR; the red curve displays the cumulative overlap; the dotted green curve shows the expected cumulative overlap if the modes were equally contributing each. The inset shows the initial structures of AMPAR (left) and NMDAR (right).

Publications Resulting from This Work

    • Krieger J, Bahar I, Greger IH (2015) Structure, Dynamics, and Allosteric Potential of Ionotropic Glutamate Receptor N-Terminal Domains Biophys J109: 1136-48 PMID: 26255587, PMC4576161
    • Dutta A, Krieger J, Lee JY, Garcia-Nafria J, Greger IH, Bahar I (2015) Cooperative Dynamics of Intact AMPA and NMDA Glutamate Receptors: Similarities and Subfamily-Specific Differences Structure23: 1692-1704 PMID: 26256538, PMC4558295
    • Dutta A, Shrivastava IH, Sukumaran M, Greger IH, Bahar I (2012) Comparative dynamics of NMDA- and AMPA-Glutamate receptor N-terminal domains Structure20: 1838-49 PMID: 22959625, PMC3496038




3) Integration, prediction, and generation of mixed mode information using graphical models, with applications to protein-protein interactions

PIs: C. J. Langmead, Carnegie Mellon University Collaborators: C. Bailey-Kellog, Dartmouth University; N. Ramakrishnan, Virginia Tech;  A. Friedman, Purdue University

This C&SP was an NSF-funded project titled: “Integration, prediction, and generation of mixed mode information using graphical models, with applications to protein-protein interactions” (NSF IIS-0905193). It was collaboration between Carnegie Mellon University, Dartmouth College, Virginia Tech, and Purdue. The project concluded in August, 2013. The focus of this C&SP was to develop novel probabilistic graphical models for modeling protein-protein interactions. Three broad classes of techniques were developed to integrate attribute-value and relational information, integrate statistical and physical information, and utilize probabilistic models generatively. The research has resulted in new graphical models, new algorithms, and the application of those algorithms to the modeling of protein structures and dynamics, including protein-protein interactions. Specific accomplishments include: the first optimal algorithm for learning regularized undirected graphical models of protein sequences; the first method for performing binding free energy calculations for protein-protein interactions via graphical models; the first probabilistic graphical models of molecular dynamics; the first algorithm for learning the parameters of molecular mechanics force fields by minimizing functionals over Boltzmann distributions; the first undirected model of distributions on the hypersphere (for modeling distributions over angles); a game-theoretic method for modeling the emergence of drug resistance-causing mutations in proteins; new algorithms for learning semi- and non-parametric distributions; and a new class of regression models for predicting binding free energies. The project resulted in seventeen publications, three Ph.D. dissertations, and one patent. Primary applications of this research include computer-aided drug design and computer-aided protein engineering. Secondary applications of this work include techniques for identifying functionally important residues mediating binding and allostery.

Publications Resulting from This Work

    • Kamisetty H, Ghosh B, Langmead CJ, Bailey-Kellogg C (2014) Learning Sequence Determinants of Protein: protein Interaction Specificity with Sparse Graphical Models Res Comput Mol Biol8394:129-143 PMID: 25414914, PMC4235964.
    • Kamisetty H, Ghosh B, Langmead CJ, Bailey-Kellogg C (2015) Learning sequence determinants of protein: protein interaction specificity with sparse graphical models J Comput Biol22:474- 86 PMID: 25973864, PMC4449715



5) Modeling immunoreceptor signaling, autophagy, and endocytosis, Hardening Software for Rule-Based Modeling

PI:  J. Faeder, University of Pittsburgh  Collaborator: W. S. Hlavacek, Los Alamos National Laboratory

The fundamental events in FcεRI signaling initiation were considered during the pioneering work of Goldstein, Faeder, Hlavacek and colleagues to establish the utility of rule-based methods for mechanistic modeling of the dynamics of molecular interactions in signaling networks. This led to the creation of BioNetGen, already introduced in the BTRC proposal as a computational framework for modeling biochemical networks. Here, we propose to take advantage of successful initiatives at the BTRC for integration of the BioNetGen rule-based modeling approach into the mesh-based architecture that underlies MCell spatial simulations. In particular, we will focus on the proposed new capabilities for adaptive meshing, which can superimpose well-characterized changes in surface geometry over time during simulations. This unique feature will allow us to develop novel computational models to describe simultaneously slow endocytic vesicle dyanamics and fast signaling dynamics at vesicle surfaces.

Aim 1. To utilize the CellBlender interface to create a framework for cell shape changes that occur during mast cell signaling, such as microvillar-to-ruffled transformation, appearance of signaling patches, and changes associated with enhanced endocytosis.

Aim 2. To evaluate the effects of a changing 3D geometry on signaling output, by incorporating existing Rule-Based Models for FcεRI chemical reaction networks into MCell-BioNetGen and refining the model with new parameters for spatial distributions, rate constants and diffusion characteristics collected through funded initiatives.

During the project we developed MCell-R, which now enables network-free simulations to be performed directly in MCell based on a BioNetGen model specification. We have validated that the simulator gives correct results for a range of FceRI signaling models and also for the bivalent-ligand bivalent receptor, a simple example of unrestricted aggregation. In principle the simulator can now be used to simulate spatial signaling dynamics of any model that can be described in BNGL with any geometry that can be specified in MCell, including a geometry that changes in time. We plan to implement several such models in the future.

Publications Resulting from This Work

    • Chylek LA, Harris LA, Tung CS, Faeder JR, Lopez CF, Hlavacek WS (2013) Rule-based modeling: a computational approach for studying biomolecular site dynamics in cell signaling systems Wiley Interdiscip Rev Syst Biol Med 6: 13-36 PMID: 24123887, PMC3947470
    • Chylek LA, Harris LA, Faeder JR, Hlavacek WS (2015) Modeling for (physical) biologists: an introduction to the rule-based approach Phys Biol 12: 045007 PMID: 26178138, PMC4526164
    • Lin YT, Feng S, Hlavacek WS. (2019) Scaling methods for accelerating kinetic Monte Carlo simulations of chemical reaction networks. J Chem Phys. 2019 Jun 28;150(24):244101. doi: 10.1063/1.5 096774. PMID: 31255063



6) Distance-dependent structure and function of neuronal dendrites

PIs: T. Sejnowski and T. Bartol, The Salk Institute Collaborator: K. Harris, University of Texas at Austin

The purpose of this CS&P is to collaborate with Kristen Harris' lab to create realistic MCell models of dendritic structure and function.

MCell can simulate the diffusion and interaction of molecules involved in biochemical signaling pathways within the 3D subcellular structure of cells. To do so, surface meshes used to represent cell membranes and subcellular structures must meet very strict geometric standards (e.g. water-tight, non-intersecting, manifold). We help create realistic MCell models of CA1 dendritic spines, in particular to learn how more the accurate serial section tomographic reconstructions of core organelles impact simulations of molecular signaling within dendritic spines.

Recently two papers were published relating to this C&SP. The first paper, published in Frontiers in Synaptic Neuroscience, reported on reconstitution of calcium dynamics in dendritic spines. Nine parameters of the model were optimized within realistic experimental limits by a process that compared results of simulations to published data. Simulations in the optimized model reproduce the timing and amplitude of Ca(2+) transients measured experimentally in intact neurons. Thus, the characteristics of individual isolated proteins determined in vitro could accurately reproduce the dynamics of experimentally measured Ca(2+) transients in spines. The second paper, published in eLife, found that dendritic spines that receive input from the same axon are the same size. This finding allowed estimation of the variability of synaptic plasticity and we found that the amount of information stored at synapses is approximately 4.7 bits, which is an order of magnitude larger than previous estimates.

Publications Resulting from This Work

    • Edwards J, Daniel E, Kinney J, Bartol T, Sejnowski T, Johnston D, Harris K, Bajaj C (2014) VolRoverN: enhancing surface and volumetric reconstruction for realistic dynamical simulation of cellular and subcellular functionNeuroinformatics12:277-89 PMID: 24100964, PMC4033674
    • Kinney JP, Spacek J, Bartol TM, Bajaj CL, Harris KM, Sejnowski TJ (2013) Extracellular sheets and tunnels modulate glutamate diffusion in hippocampal neuopil J Comp Neurol521: 448-64 PMID: 22740128, PMC3540825
    • Bartol TM, Keller DX, Kinney JP, Bajaj CL, Harris KM, Sejnowski TJ, Kennedy MB (2015) Computational reconstitution of spine calcium transients from individual proteins Front Synaptic Neurosci7: 17 PMID: 26500546, PMC4595661
    • Bartol TM, Bromer C, Kinney J, Chirillo MA, Bourne JN, Harris KM, Sejnowski TJ (2015) Nanoconnectomic upper bound on the variability of synaptic plasticity Elife4.pii: e10778 PMID: 26618907, PMC4737657



8) Using generative models of cell organization to investigate tumor cell heterogeneity

PIs: R. Murphy and G. Rohde, Carnegie Mellon University Collaborators: S. Altschuler and L. Wu, University of Texas Southwestern Medical Center

Tumor cell populations have long been known to be complex mixtures of multiple phenotypes. Given the precision with which one is able to measure properties of cellular structure and function, it is often the case that the measure of two cells will differ. Paramount to deriving useful information from a potentially large set of measurements is to be able to determine which differences are statistically significant as well as their biological meaning. The established framework for analyzing the heterogeneity of tumor cell populations involves analyzing these measurements in a high dimensional feature space. Features can be simple and easily interpretable (such as cell size), or complex and difficult to interpret (such as texture features). The goal of this subproject is to try to represent cell images using generative models that are more easily interpretable and could potentially offer more insights into the underlying phenomena that affect the distribution of phenotypes in tumor cells.



9) Large-scale electron microscopy of calcium-imaged neuron populations

PIs: A. Wetzel and G. Hood, Pittsburgh Supercomputing Center Collaborator: D. Bock, Janelia Farm Research Campus, Howard Hughes Medical Institute

During the last year of this project, the biological focus of Bock's work has shifted from localized mouse brain circuits to the mapping of complete drosophila brains with an emphasis on visual circuits. This change did not affect the relation of our collaboration which is focused on the technical aspects of maximum scale serial-TEM data assembly.

Bock's group has improved the throughput of their automated 4 camera TEM system. Although the cameras and their raw data rates are unchanged new mechanisms have been tested in the last year for automatic sample cassette loading. This has substantially improved the ratio of imaging to sample exchange time. The long term goal for this new sample loading method is to enable continuous runs up to 2 weeks for datasets of 15,000 sections and >50 TBytes. Individual sections are imaged in arrays of ~20,000 5 MPixel tiles, typically 112 columns by 196 rows, with a unique stepping pattern to accommodate constraints on the spacing of the optical cameras and lenses. One difficult aspect of these data is the use of the small 5 MPixel tiles which, particularly in the case of hierarchical registration methods, become limited by poor performance of large filesystems when processing small files.

Regular production of datasets at this scale would be difficult to register using our previous method of raw data transmission to the PSC followed by the return of aligned results to Janelia Farm. Therefore we are preparing for Bock's team to run our alignment codes on the large 6,144 core Janelia Farm computing cluster. Wetzel and Hood were part of an April 1-11, 2015 Janelia Farm image registration Hackathon organized by Stephan Saafeld's group. During that trip we were able to test both the AlignTK and SWiFT codes on the Janelia systems using data from Bock's work as well as many others from Janelia and elsewhere (i.e. Winfried Denk and our C&SP10 collaborator Jeff Lichtman). We were also able to adapt a key component of the SWiFT method for use within one of the Janelia registration codes and found that it greatly improved the reliability of registration on difficult regions of Bock's data.

Publications Resulting from This Work

    • Bock DD, Lee WC, Kerlin AM, Andermann ML, Hood G, Wetzel AW, Yurgenson S, Soucy ER, Kim HS, Reid RC (2011) Network anatomy in vivo physiology of visual cortical neurons Nature471: 177-82 PMID: 21390124, PMC3095821



10) Advancing high-throughput thin-section scanning EM to study relationships between neuronal circuit structure and function

PIs: A. Wetzel and G. Hood, Pittsburgh Supercomputing Center Collaborator: J. Lichtman, Harvard University

We have continued work on automated alignment of serial SEM image sets produced using tape collecting ultramicrotome sectioning and wafer mounted imaging procedures developed at Harvard. The primary datasets have been the 16,000 section zebrafish stack, which is the focus of a new CS&P project with the Engert lab and the 114 TB lateral geniculate nucleus (LGN) volume that was previously acquired and is the focus of this C&SP.

The different requirements and resolutions of these datasets have highlighted the need for a variety of approaches to handle different types of biological content and different types of image characteristics. The large number of tape cracks in the LGN dataset has been particularly difficult since the cracks from any section to its neighbors interfere with one another during the alignment process. Even though the detailed cracking pattern is different on each section the fact that most of the cracks have similar sizes and orientations means that they tend to pull the alignment such that we have seen errors on the order of 1 per 1000 sections. We are currently working on a strategy to produce a low-resolution map of the cracks, 1/4th scale, that can then be used to constrain the SWiFT alignment process.

We are entering a new and much higher-throughput phase of EM data capture. Previous datasets, including the zebrafish and LGN, were captured over several months using a Zeiss Merlin microscope and pixel rates up to 20MHz. Lichtman's team recently installed the first Zeiss 61-beam 1.2 Gpixel/sec scanning microscope. This microscope is now being tested with large single sections in the 400GB/section range, which are captured in a hexagonal mosaic pattern. These data will require a substantial number of registration code adaptations. We will test speed and accuracy of these modifications once the new equipment is producing serial spans of at least 100 sections, which is where the power of the SWiFT approach is most useful.

Publications Resulting from This Work

    • Morgan JL, Berger DR, Wetzel AW, Lichtman JW (2016) The fuzzy logic of network connectivity in mouse visual thalamus Cell165: 192-206 PMID: 27015312, PMC4808248
    • Hildebrand DGC, Torres RM, Choi W, Tran Minh Quan, Arthur Willis Wetzel, George ScottPlummer, Ruben Portugues, Isaac Henry Bianco, Owen Randlett, Stephan Saalfeld, Alex Baden, Kunal Lillaney, Randal Burns,Joshua Tzvi Vogelstein, Won-Ki Jeong, Jeff William Lichtman, Florian Engert (2016) Whole-brain serial-section electron microscopy in larval zebrafish Nature, revised version in preparation.



11) Morphological and regulatory models of neuronal differentiation

PIs: H Busch and M Böerries, University of Freiburg Collaborators: R Murphy and G Rohde, Carnegie Mellon University

The goal of this project is to develop a spatiotemporal generative model of the changes in cell size, shape and subcellular organization that occur during the differentiation of PC12 cells induced by nerve growth factor. This model will be related to a regulatory model constructed from parallel measurements of RNA expression over the time course of differentiation. The results are expected to lead to the identification of gene expression changes that lead to specific morphological changes and to test their importance for the differentiation process.

Publications Resulting from This Work


12) Generative models of plant organelle distribution and differentiation

PI: K Palme, University of Freiburg Collaborator: R Murphy, Carnegie Mellon University

Plants exhibit the ability to undergo dramatic dedifferentiation and redifferentiation such that both protoplasts (cells extracted from mature plants) and microspores (cells that arise during gametogenesis) can give rise to mature plants under the right circumstances. The goal of this project is to identify the specific changes in protein expression and localization that are associated with such processes, in part by building spatiotemporal generative models directly from images and movies of cells obtained by light microscopy.

Publications Resulting from This Work


13) Microtubule pattern analysis and drug sensitivity

PI: P Giannakakou, Cornell University Collaborator: R Murphy, Carnegie Mellon University

The goal of this new project is to determine whether and how the distributions of microtubules differ between tumor cells sensitive and resistant to drugs that act on microtubules. Initial work has focused on training systems to recognize these patterns, with a complicating factor being the large differences in cell size, shape and unperturbed microtubule pattern between cells from different tumors. Preliminary results indicate that three broad but distinct patterns can be distinguished, corresponding to untreated cells, drug-treated resistant cells, and drug-treated sensitive cells. Future work will focus on using generative models to remove variation due to cell size and shape. Ultimately, this would be used to determine whether a particular tumor is likely to be resistant to a particular drug and thereby to choose an appropriate drug for that tumor.



14) Actin filament patterns and cell motility

PI: J Theriot, Stanford University Collaborators: G Rohde and D Slepčev, Carnegie Mellon University

This is a new project, the goal of which is to quantitatively describe the patterns of actin filaments and related proteins as they relate to cell motility in neutrophils. The project aims to use recently developed methods for measuring the similarity between such patterns, as well as transport-based method for finding correlation between intensity patterns, to decode relationships between different proteins as cells move. Results are expected is to help characterize cell motion as a function of the subcellular protein patterns.



16) Latent Factor Models for Identification of Novel Neuroprotectives for Huntington's Disease

PI: Ivet Bahar and Bing Liu, University of Pittsburgh;  Collaborators: Robert Friedlander, Mark Shurdak, Andrew Stern, D. Lansing Taylor, University of Pittsburgh

This C&SP is supported by NIH award 5R01NS077748-04 (PI: Friedlander) entitled “Functional role of microRNAs in Huntington's disease pathogenesis”. Our focus is the identification of novel neuroprotectives using latent factor modeling (LFM). Firstly, we built a set of known neuroprotectives based on the data in the literature by collecting compounds that were identified in a large scale screen for neuroprotectives in Huntington’s disease (HD) as well as compounds that have been tested in clinical trials for their neuroprotective effect. Then, the LFM-based algorithm learns the most successful latent factor model of the drug-target interactions in the STITCH dataset after assessing the performance of multiple LFM learning algorithms. The LFM of STITCH allowed us to quantitatively assess the interaction profiles of the known neuroprotectives, which we then used to find new compounds toward (i) the discovery of new neuroprotectives by target-based diversification of known neuroprotectives, (ii) the discovery of mechanism of action hypotheses by identifying targets of interest. Specifically, for each target of known neuroprotectives, we identified other chemicals known to be interacting with high confidence and sorted them for maximal dissimilarity to the known neuroprotectives interacting with that target. This method identifies compounds that share the target of interest, but are otherwise as diverse in their interaction profile as possible. Compounds selected will be active if the target of interest is important for the HD etiology, therefore they represent new potential treatment candidates. Furthermore if one or more targets yield many neuroprotectives, that target is implicated as being central to the disease pathophysiology; therefore this method allows us to infer about the proteins regulating the disease phenotype by testing compounds alone.

Publications Resulting from This Work



18) Effect of RNA-editing of ADAR1 on the activity of ionotropic glutamate receptors

PI: Ivet Bahar, Mary H Cheng, University of Pittsburgh  Collaborators: QingDe Wang, University of Pittsburgh

This is a C&SP, supported by NIH 5R21CA158650-02 (PI: Wang). The goal of the project is to define the function of RNA editing enzyme adenosine deaminase acting on RNA 1 (ADAR1) in hematopoietic and leukemia stem cells. ADAR1 is an essential protein for embryonic and adult hematopoiesis, while the leukemia cells are more susceptible to the gene deletion that codes ADAR1. The impact of ADAR1 on normal and leukemia stem cells (LSC) remain unknown, and there is a need to develop a therapeutic strategy to eliminate LSCs by targeting ADAR1. Although ADAR1 is identified as an RNA editing enzyme and RNA editing has been shown to play critical roles in stem cells and other biological processes, multiple attempts to date have been unable to identify an editing target that accounts for the death of normal hematopoietic and leukemia cells of ADAR1 knockouts. The R21 seeks a definitive answer whether the editing activity of ADAR1 is necessary for the proliferation and differentiation of LSCs.

ADAR has been found to modify the AMPA receptor GluR2 and G-protein coupling receptors, i.e, SR2C and GABA receptor. The knockout animal models (Wang lab) have shown that this molecule is essential for embryonic survival and tissue homeostasis. However the molecular mechanism is largely unknown. The Bahar lab has initiated a new collaboration with the Wang Lab, to help investigate (a) the effects of the mutations induced upon ADAR1 editing on calcium channeling properties of iGluRs, (ii the effect of specific mutations on the functions of serotonin receptors and potassium channels, using molecular modeling, and (c) mathematical modeling and quantitative assessment of the effect of suppressing cytosolic RNA pathway by ADAR1 on relevant cellular pathways networks.



21) Functional significance of the dynamics of AMPAR extracellular region

Collaborating Investigators: Ingo H. Greger, MRC Lab of Molecular Biology, Ivet Bahar, QingDe Wang, University of Pittsburgh, Tom M. Bartol, Terry J. Sejnowski, Salk Institute

Ionotropic glutamate receptor (iGluRs) are ligand-gated ion channels that allow for the flow of cations into the postsynaptic cell in response to glutamate binding, thus regulating neurotransmission upon depolarization of the cell membrane. Among iGluR subfamilies, AMPAR and NMDAR play a key role in learning and memory, and in particular the AMPAR is essential to rapid neurotransmission and synaptic plasticity. The Greger and Bahar labs have been productively collaborating in recent years on AMPAR dynamics, first using the NTD dimer structures and more recently the intact tetrameric structures. These studies demonstrated that the NTD domains exhibit structural flexibilities comparable to those of AMPAR NTDs. Furthermore, the global modes of motions predicted by ANM (or ProDy) revealed the propensity of homotetrameric AMPAR to assume more compact forms similar to NMDARs. The validity of these modes of motions were confirmed by cross-linking experiments between NTD sites predicted by ANM to come into close proximity. In the new term, we will first adopt ANM-based analysis to characterize the mode spectrum of the heterotetrameric AMPAR. ProDy analysis already revealed that the O ↔ N transition is enabled by a global ANM mode. We will characterize thoroughly the whole spectrum of motions and generate the energy landscape of Glu2/3 heterotetramer, using the recently introduced extension of coMD. Then we will focus on the ECR motions that induce a pore opening (or cooperative twisting) at the TMD and analyze the conformational events that enable the allosteric coupling between the ECR and the TMD with the help of accelerated MD simulations. In the next phase, we plan to examine the significance of GluA2/3 ECR flexibility in adapting to its interactions with auxiliary proteins such as cornichon homologs, TARPs or in forming clusters, which will be further tested/validated with structural and single-particle tracking methods in the Greger lab.

Publications Resulting from This Work

    • Krieger J, Bahar IGreger IH (2015) Structure, Dynamics, and Allosteric Potential of Ionotropic Glutamate Receptor N-Terminal Domains Biophys J 109: 1136-48 PMID: 26255587, PMC45761612
    • Dutta A, Krieger J, Lee JY, Garcia-Nafria J, Greger IHBahar I (2015) Cooperative Dynamics of Intact AMPA and NMDA Glutamate Receptors: Similarities and Subfamily-Specific Differences Structure 23: 1692-1704 PMID: 26256538, PMC4558295
    • Dutta A, Shrivastava IH, Sukumaran M, Greger IHBahar I (2012) Comparative dynamics of NMDA- and AMPA-Glutamate receptor N-terminal domains Structure 20: 1838-49 PMID: 22959625, PMC3496038
    • Lee JY, Krieger J, Herguedas B, García-Nafría J, Dutta A, Shaikh SA, Greger IH, Bahar I. (2019) Druggability Simulations and X-ray Crystallography Reveal a Ligand-binding Site in the GluA3 AMPA Receptor N-terminal Domain. Structure 27: 241-252. 
    • Krieger J, Lee JY, Greger IH, Bahar I. (2019) Activation and Desensitization of Ionotropic Glutamate Receptors by Selectively Triggering Pre-existing Motions. Neurosci Lett 700: 22-29 PMID: 29481851 PMCID: 6107436
    • Lee JY, Krieger J, Herguedas B, García-Nafría J, Dutta A, Shaikh SA, Greger IH, Bahar I. (2019) Druggability Simulations and X-Ray Crystallography Reveal a Ligand-Binding Site in the GluA3 AMPA Receptor N-Terminal Domain. Structure. 27(2):241-252 PMID: 30528594



23) CaMKII Structural Dynamics

Collaborating Investigators: John Kuriyan, University of California-Berkeley, Mary Kennedy, Caltech, Ivet Bahar, James R. Faeder, University of Pittsburgh, Tom M. Bartol, Terry J. Sejnowski, Salk Institute

This C&SP will focus on studying mechanisms of function of the complex biomolecular assembly of CaMKII and its dynamics. We recently showed that the human alpha isoform, CaMKIIalpha, literally acts as a device for propagating the activation by calcium signaling. This is achieved through subunit exchanges between activated CaMKII and holoenzyme assemblies that have not been activated. The spread of activation through these exchanges now suggests that activation can be maintained long after the disappearance of the initial signal. This is remarkable because it reveals a possible mechanism by which CaMKII assists in the induction of long-term memory in the brain. It is clear that the phosphorylation of the calmodulin-recognition element triggers the exchange. But in addition to localized interactions, there is something unique about the overall architecture of CaMKII so that it accommodates different oligomerization states. Starting from early crystal structures and EM images, we have seen that the holoenzyme could assume both dodecameric and tetradecameric forms. Our new observations now show that they coexist. An intriguing observation was: why does the vertical dimer dissociate, rather than a single subunit, or a horizontal dimer? ANM results unambiguously show that the dodecamer structure indeed favors the vertical dimers to be released. The type of analysis performed by ProDy can open the way to a better understanding of CaMKII dynamics. Additionally, coevolution analysis may give us additional information on interfacial interactions that modulate subunit exchange. These studies will help further understanding the dynamics and interactions of CaMKII.



25) Efficient parallel sampling at multiple scales using the weighted ensemble strategy

Collaborating Investigators: Lillian T. Chong, University of Pittsburgh, Daniel Zuckerman, Oregon Health and Science University

There is a "silicon ceiling" that ultimately limits many, if not most, types of dynamical biological simulations. That is, even the world's most powerful computers cannot generate sufficiently long simulations, whether for atomistic models of proteins or for realistic models of cell behavior. In many cases, the key events may occur beyond simulation timescales - such as protein folding, conformational transitions of proteins, assembly of protein complexes, or transitions of cell behavior from healthy to pathological states. The work in this C&SP will continue the development of WESTPA, a tool for controlling other software tools: it orchestrates up to thousands of trajectories run natively by other software at any scale (e.g., Gromacs, Amber, BioNetGen, MCell) using a "weighted ensemble" strategy. Not only does WESTPA parallelize the use of dynamics engines - but because of the statistical process by which trajectories are added and removed, WESTPA can obtain estimates of key kinetic as well as equilibrium observables in significantly less computing time than would be required in ordinary parallelization. Our aims are to improve the ease of use and interoperability of WESTPA; to improve its performance and reliability; to demonstrate the effectiveness of WESTPA through a series of "showcase" examples from molecular to cellular scale using a variety of dynamics engines; and to improve instructional materials based on the showcase examples.

Publications Resulting from This Work

    • Zwier MC, Adelman JL, Kaus JW, Pratt AJ, Wong KF, Rego NB, Suárez E, Lettieri S, Wang DW, Grabe M, Zuckerman DM, Chong LT (2015) WESTPA: an interoperable, highly scalable software package for weighted ensemble simulation and analysis J Chem Theory Comput 11: 800-809. PMID: 26392815; PMC4573570
    • Suárez E, Pratt AJ, Chong LT, Zuckerman DM (2016) Estimating First Passage Time Distributions from Weighted Ensemble Simulations and non-Markovian Analyses Protein Science 25: 67-78. PMID: 26131764; PMC4815309
    • Suárez E, Pratt AJ, Chong LT, Zuckerman DM (2016) Estimating First Passage Time Distributions from Weighted Ensemble Simulations and non-Markovian Analyses Protein Science 25: 67-78. PMID: 26131764; PMC4815309



26) GPCR signaling and interactions. Learning from SILAC/proteomics

Collaborating Investigators: Kunhong (Kevin) Xiao, Ivet Bahar, James R. Faeder, University of Pittsburgh

This focus of this C&SP will focus on the development and application of quantitative proteomics and systems biology approaches to study GPCR (G protein couple receptor) structures, their signaling networks, and regulation. The approach will be to use mass spectrometry-based quantitative proteomics, in combination with systems, chemical, and structural biology to study the receptor structure-function relationship, macromolecular interactions, and the expression and posttranslational modifications of signaling molecules downstream of GPCR. The ultimate goal of this research is to provide a better understanding of structure-function relationship of GPCRs and their related proteins and provide tools and resources for structure-based drug design.



28) Dynamic modulation of interferon binding affinity as a mechanism to regulate interferon receptor signaling

Collaborating Investigators: Gideon Schreiber, Weizmann Institute, Ivet Bahar, James R. Faeder, University of Pittsburgh

Type I interferons (IFNs) are multifunctional cytokines that mediate/induce diverse cellular responses, including both innate and adaptive immune responses, stimulation of antiviral responses, and cancer surveillance, upon forming a ternary complex with two surface receptors, IFNAR1 and IFNAR2. The activities of IFN-a subtypes correlate with their affinities to bind to IFNAR1 and IFNAR2. While the Schreiber lab made seminal contributions to understanding the molecular basis of IFNARs, the mechanism of regulation of differential IFN activities through interactions with IFNAR1 and 2, remains unclear. Our integrated computational (TR&D1) and experimental preliminary studies point to the significance of the intrinsic dynamics in modulating binding affinity. We adopted a closely integrated computational/experimental strategy that yielded promising results, which we are currently further pursuing and that illustrate the adaptability of proteins to different bound states or to sequence variations/mutations via their softest modes of motion. Further cross-linking, fluorescence quenching and gene induction experiments will be conducted in the Schreiber lab, in close coordination with TR&D1 computational studies at the Bahar lab.

Publications Resulting from This Work

  • Li H, Sharma N, General IJ, Schreiber G, Bahar I. (2017) Dynamic Modulation of Binding Affinity as a Mechanism for Regulating Interferon Signaling. J Mol Biol 429: 2571-2589 PMID: 28648616 PMCID: 5545807



32) Integrating compartmental rule-based modeling into VCell

Collaborating Investigators: James C. Shcaff, Mikhail L. Blinov, Ion I. Moraru, Lew Loew, University of Connecticut, James R. Faeder, University of Pittsburgh

This C&SP is a collaboration with The National Resource for Cell Analysis and Modeling (NRCAM), which develops new technologies for modeling cell biological processes. The technologies are integrated through Virtual Cell (VCell), a problem solving environment built on a central database and disseminated as a client-server application. It will develop methods to simulate multi-molecular interactions at a sub-cellular scale for problems where molecular shape and cellular geometry influence the system behavior. MMBioS investigators will collaborate to integrate compartmental rule-based modeling into the VCell platform.


Copyright © 2020 National Center for Multiscale Modeling of Biological Systems. All Rights Reserved.