In this paper, we studied neural networks of both excitatory and inhibitory populations with inhibition stabilized network (ISN) models. In ISN models, the recurrent excitatory connections are so strong that the excitatory sub-network is unstable if the inhibitory firing rate is fixed; however, the entire network is stable due to inhibitory connections. In such networks, external input to inhibitory neurons reduced their responses due to the withdrawal of network excitation (Tsodyks et al., 1997). This paradoxical effect of the ISN was observed in recent surround suppression experiments in the primary visual cortex with direct membrane conductance measurements (Ozeki et al., 2009). In our work, we used a linearized rate model of both excitatory and inhibitory populations with weight matrices dependent on the locations of the neurons. We applied this model to study surround suppression effects and searched for networks with appropriated parameters. The same model was also applied in the study of spontaneous activities in awake ferrets. Both studies led to network solutions in the ISN regime, suggesting that ISN mechanisms might play an important role in the neural circuitry in the primary visual cortex.

The prototypical prokaryotic channel, KcsA, is model system for mammalian potassium channels. This thesis describes solid-state NMR studies of the conformational dynamics involved in ion binding and channel inactivation. Our studies are conducted on the full-length wildtype channel incorporated into native-like lipid bilayers. We focus on the selectivity filter, which is a key conserved region across all potassium channels and is responsible for the selective passage of potassium ions across cellular membranes. KcsA is known to alter the structure of its pore as a function of the ambient potassium level; at high potassium the selectivity filter adopts a conductive state with high ion occupancy and at low potassium it collapses into a non conductive state with reduced ion occupancy. We report solid-state NMR resonance assignments for ~ 200 15N and 13C atoms in KcsA using two and three dimensional correlation spectroscopy. Using the assignments we characterize the conductive and collapsed states of the selectivity filter and show that the transition between the states occurs at a potassium concentration of 1-15 micromolar. We interpret the shape of the binding curves in terms of the complex equilibria involved in the structural collapse of the filter. Our results describe the detailed structural and kinetic landscape of the selectivity filter as it collapses at low potassium. To gain more mechanistic insight into the physiological role of the collapsed state, we conduct studies of an inactivation resistant mutant of KcsA, E71A. C-type inactivation is the potassium and voltage dependent closure of the outer mouth of the channel and is a clinically important process in mammalian potassium channels. We show that the collapsed state is similar to the inactivated state in terms of structure and chemical shifts, which establishes that inactivation processes can be studied by lowering the ambient potassium level. We also use site-specific chemical shift tensor measurements, to show that glutamic acid 71 is protonated at neutral pH and therefore must have an abnormally high pKa (<7.5). Our studies clarify a previous inconsistency in the literature regarding the pKa of E71 and propose a new model for the role of E71 in channel inactivation. The results have led to insights into the molecular factors that stabilize the collapsed or inactivated state of the channel, and firmly establish KcsA as a model membrane protein system for future studies of structure and dynamics by solid state NMR.

Both protein structure and dynamics are essential to understanding biological function. NMR is a powerful technique for the observation of protein dynamics in that dynamics can be observed site-specifically over a wide range of timescales from picoseconds to seconds. Spin relaxation measurements, including relaxation in the rotating frame (R1ρ), can be very sensitive to exchange processes in proteins, particularly on the millisecond-to-microsecond timescale. Using solid state NMR, few techniques exist that can quantify dynamics on this timescale. Previous R1ρ relaxation measurements in the solid state have utilized reorientation of a dipole tensor to observe dynamics. This application is limited to systems where the nucleus of interest has an attached proton. Relaxation studies using the reorientation of a chemical shift tensor are applicable to a broader range of systems. Furthermore, solid state experiments do not require a change in the isotropic chemical shift as is necessary in solution NMR. We combined R1ρ measurements of the model compound dimethyl sulfone (DMS) with data-fitting routines in Spinevolution to show that R1ρ relaxation due to reorientation of a chemical shift tensor is a large effect in the solid state and these measurements can be used to quantify chemical exchange processes. The temperature dependence of the exchange rates determined with R1ρ measurements is in agreement with other measurements of the dynamics of DMS with various solid state NMR techniques. Deuteration and sparse isotropic labeling were necessary to obtain quantitative results. To distinguish the exchange contribution to relaxation from other effects (R2 relaxation), low temperatures and high spin-lock field strengths were utilized. R1ρ experiments and magic angle spinning (MAS) one-dimensional spectra were used to characterize phosphate ligand binding in the glycolytic protein triosephosphate isomerase. 1D spectra indicated the presence of both isotropic and anisotropic phosphate populations. These states included an unbound state with an isotropic chemical shift tensor, and a protein-bound state in which the anisotropic features are reintroduced through chelation with protein backbone amides. The chemical shift anisotropy tensor of the bound phosphate ligand was fit using spinning sideband analysis of slow MAS spectra and suggest the ligand is in a dianionic state. The temperature dependence of R1ρ measurements indicated a fast dynamic process above the microsecond timescale at physiological temperatures.

A persisting question in biology concerns the exceptional diversity of metabolic enzymes and how they respond to their ligands and dynamic environments with remarkable precision. In humans, the family of biotin-dependent carboxylases holds important roles in intermediary metabolism. Recent years have witnessed significant progress toward understanding these enzymes' roles in homeostatic regulation. However, due to a lack of structural information, their catalytic mechanisms, as well as the macromolecular consequences of their genetic mutations, are still not well understood. This dissertation describes the characterization of two biotin-dependent carboxylases that catalyze essential metabolic transformations in humans and bacteria, using X-ray crystallography to elucidate their structures and biochemical assays to verify their activities. We engineer a novel chimeric variant of propionyl-CoA carboxylase (PCC) and produce the first crystal structure of its 750-kDa α6β6 holoenzyme. This structure reveals the architecture of PCC's twelve catalytic domains and allows the mapping of its disease-associated gene mutations to predict their effects on enzyme stability and catalysis. We also identify and describe a new domain that is integral to maintaining inter-subunit contacts within PCC. Following this, we extend our studies to methylcrotonyl-CoA carboxylase (MCC), another 750-kDa α6β6 holoenzyme that differs from PCC primarily in its substrate preference. The crystal structure of MCC assumes a markedly different configuration from PCC despite the high sequence identity between the two. Theorizing that these enzymes may represent unique lineages in the evolution of the biotin-dependent carboxylases, we apply similar approaches to the study of a third biotin-dependent carboxylase. Our efforts have produced the first two holoenzyme structures of CoA-recognizing biotin-dependent carboxylases, and provide valuable insight for understanding the functions of these vital enzymes.

Neighboring neurons in cat primary visual cortex (V1) have similar preferred orientation, direction, and spatial frequency. How diverse is their degree of tuning for these properties? Are they also clustered in their tuning for the spatial phase of a flashed grating ("absolute spatial phase") or the temporal phase of a drifting grating ("temporal response phase")? To address these questions, we used tetrode recordings to simultaneously isolate multiple cells at single recording sites and record their responses to flashed and drifting gratings of multiple orientations, spatial frequencies, and spatial/temporal phases. We recorded the responses of 761 cells presented with drifting gratings and 409 cells presented with flashed gratings. We found that orientation tuning width, spatial frequency tuning width and direction selectivity index all showed significant clustering. Absolute spatial phase and temporal response phase, however, showed no clustering. We also present an algorithm that improves the performance of spike-sorting algorithms, for use in analyzing cells recorded using tetrodes. A cluster of spikes corresponding to a putative cell obtained through automatic or manual spike sorting algorithms may contain spikes from other cells with similarly-shaped waveforms. Our algorithm preferentially removes contaminating spikes from other cells, thereby decreasing the level of contamination of each unit. We call this procedure "pruning", as it entails removing portions of the cluster that are determined to be more likely to contain contaminating spikes than the cluster as a whole. Testing of the algorithm on data in which "ground truth" is known shows excellent performance, for example on average giving a percentage reduction in false positive spikes 8.2 times the percentage reduction in true positive spikes, and reducing the degree of contamination by an average of about 13%.

This dissertation discusses two main aspects of protein-ligand interaction for G-Protein coupled receptors: structure predictions of the flexible loop domains and docking into these receptors. The prediction of loop structure has been long worked on in the context of native, globular proteins. In this work it is extended to transmembrane proteins, which requires an explicit integration of the lipid bilayer into the loop prediction calculation. In the initial work, this new approach to loop prediction yields highly accurate 3-dimensional structures of the intra and intercellular loops of four G-protein coupled receptors--the A2A adenosine, bovine rhodopsin, β1 and β2 adronergic receptors. For these cases, the loops were predicted in the context of a completely native crystal structure. In subsequent work the approach was extended to work on perturbed cases, where all loops and tails were removed, and side chains near the loop being predicted were in nonnative conformations. Lastly, a full homology model of the β2 adronergic receptor was successfully built from the β1 adronegric receptor as its template. Work on docking into these receptors focuses on the kappa opioid receptor. Known antagonist binders are discriminated from a set of decoy nonbinders via docking calculations. Two new terms were added to the scoring function, WScore to achieve this, based on a detailed molecular understanding of how the receptor works.

The formation of chromatin defines when and where genes will be expressed, from the histone proteins forming the nucleosome and their many post-translational modifications, to the immense number of proteins that bind these modifications. This thesis is comprised of two projects: the first is an analysis of the essential plant homeodomain-containing protein called zinc finger protein 1 (ZFP-1); the second is a study into how RNA interference (RNAi) factors are involved in histone production in C. elegans. I investigate the physical and biological properties of the PHD fingers of ZFP-1 and find that 1) they are essential for viability and 2) they specifically bind to methylated lysine 4 on histone H3. This study has expanded our understanding of the molecular nature of ZFP-1, the C. elegans ortholog of AF10, which has a role in chromosomal translocations promoting leukemia. I also determine that the RNAi factors CSR-1, EGO-1 and DRH-3 are required for proper histone production in C. elegans. Severe histone depletion results from the knockdown of these RNAi pathway components, which explains both the phenotypes of sterility and chromosome segregation defects in early embryos that are associated with mutants of these factors. This discovery explains the well-known, but poorly understood, phenotypes of the RNAi mutants and provides the first evidence for RNAi positively affecting gene expression.

Hox proteins are a group of homeodomain-containing transcription factors that define the body plan in both vertebrates and invertebrates. Mutations in Hox proteins lead to limb malformations or cancer in humans. Despite having homeodomains with similar sequences and structures, the eight Hox proteins in Drosophila exhibit a variety of DNA-binding specificities when they are in complex with their cofactor Extradenticle (Exd), raising the question of how such diverse specificity is generated. We have identified DNA minor groove shape as a structural determinant for Hox specificity. Using Monte Carlo simulations, we predicted the minor groove widths for Hox-binding sites obtained from a high-throughput experiment - Systematic Evolution of Ligands by Exponential Enrichment with massive parallel sequencing (SELEX-seq). We found that DNA sites selected by anterior Hox proteins have two narrow regions in the minor groove where Hox-Exd binds. In contrast, DNA sites favored by posterior Hox proteins have only one narrow region. Moreover, clustering of Hox proteins based on their preference of DNA minor groove shape reproduced the ordering of Hox genes along the chromosome, suggesting a striking relationship between body axis morphogenesis and nuances in DNA shape. Intrigued by the question of how DNA shape is recognized, we studied the interactions between an anterior Hox protein, Sex combs reduced (Scr), and its preferential DNA sites identified from SELEX-seq. Through structure-based homology modeling, we found that two Arg residues on the N-terminal arm of Scr specifically recognize the two narrow regions in the minor groove of Scr-favored sites, regardless of their nucleotide identities. Our work leads to a new understanding of the structural basis of specific DNA-binding for Drosophila Hox proteins, linking preference of DNA-binding sites to DNA minor groove shape. Our studies on Hox-cofactor-DNA structures revealed highly conserved features of protein-DNA recognition, e.g. Hox's Asn51 forms hydrogen bonds to an adenine, which are essential for Hox-DNA binding. In order to automatically identify this type of important interactions, we developed a computational module based on the functional annotation server MarkUs. This module displays a variety of protein-DNA interactions inside query structure and illustrates their degrees of conservation by comparing query structure with its structural homologs. This functional annotation module provides an effective way to analyze protein-DNA recognition and to identify essential interactions. In this dissertation, Chapter 1 introduces the field of protein-DNA specific recognition from the perspectives of three-dimensional structures, high-throughput experiments, and computational modeling approaches. Chapter 2 introduces the biological background of Hox proteins, focusing on their biological functions, three-dimensional structures, and previous studies on their DNA-binding specificity. Chapter 3 presents the investigation of DNA-binding specificity for Hox-Exd complexes. The role of DNA minor groove width as a structural determinant is demonstrated through Monte Carlo simulations. Chapter 4 describes the homology modeling method for studying DNA minor groove recognition for Scr. The recognition mode of Scr-favored SELEX-seq sequences is inferred through protein-DNA docking and interface optimization. Chapter 5 elucidates the functional annotation module for protein-DNA structures. The functions and features of this module are demonstrated through a case study on a Scr-Exd-DNA structure. Chapter 6 summarizes my research projects described in this dissertation and proposes future directions for studying specific protein-DNA recognition.

Neural circuit assembly requires that the axons and dendrites of the same neuron do not overlap each other while interacting freely with those from different neurons. This requires that each neuron have a unique cell surface identity to that of its neighbors and that neural self-recognition leads to repulsion, a process known as self-avoidance. Self-avoidance is perhaps best understood in Drosophilia, where homophilic recognition between individual Dscam1 isoforms on the cell surface of neurons leads to repulsion between sister dendrites and axons. However, in contrast to Drosophila, where alternative splicing of the Dscam1 gene can generate thousands of isoforms, vertebrate Dscam genes do not generate significant diversity. The most promising candidate to fill this role in vertebrates is the clustered protocadherins (Pcdhs). Despite this hypothesis, little is known about clustered Pcdh proteins and how they interact. The clustered Pcdh genes are encoded in three contiguous gene loci, Pcdha, Pcdhb, and Pcdhg, which encode three related families of proteins, Pcdhα, -β, and -

Characterizing the binding preferences of transcription factors is a major objective in molecular biology. Important processes such as development and responses to environmental stresses are regulated by the interactions between transcription factors and DNA. In this thesis, we address three key issues in the analysis of protein-DNA interactions. First, we demonstrate how transcription factor binding motifs can be inferred from ChIP-seq data by integrating a peak-calling algorithm and a biophysical model of transcription factor specificity. Next, we show that high-resolution DNase I cleavage profiles can provide detailed information about the role that DNA shape plays in protein- DNA recognition. Our analysis reveals the interplay between DNA sequence, methylation status, DNA geometry, and DNase I cleavage. Finally, we construct a model of transcription factor-DNA interaction that allows multiple transcription factors to bind co- operatively and competitively. In addition, the model can also infer transcription factor concentration. As the binding preferences of transcription factors continue to be characterized with a high degree of precision, we anticipate that use of these more realistic models will become more prevalent.