Molecular docking serves as an important tool in modeling protein-ligand interactions. Most of the docking approaches treat the protein receptor as rigid and move the ligand in the binding pocket through an energy minimization, which is an incorrect approach as proteins are flexible and undergo conformational changes upon ligand binding. However, modeling receptor backbone flexibility in docking is challenging and computationally expensive due to the large conformational space that needs to be sampled.

A novel flexible docking approach called BP-Dock (Backbone Perturbation docking) was developed to overcome this challenge. BP-Dock integrates both backbone and side chain conformational changes of a protein through a multi-scale approach. In BP-Dock, the residues along a protein chain are perturbed mimicking the binding induced event, with a small Brownian kick, one at a time. The fluctuation response profile of the chain upon these perturbations is computed by Perturbation Response Scanning (PRS) to generate multiple receptor conformations for ensemble docking. To evaluate the performance of BP-Dock, this approach was applied to a large and diverse dataset of unbound structures as receptors. Furthermore, the protein-peptide docking of PICK1-PDZ proteins was investigated. This study elucidates the determinants of PICK1-PDZ binding that plays crucial roles in numerous neurodegenerative disorders. BP-Dock approach was also extended to the challenging problem of protein-glycan docking and applied to analyze the energetics of glycan recognition in Cyanovirin-N (CVN), a cyanobacterial lectin that inhibits HIV by binding to its highly glycosylated envelope protein gp120. This study provide the energetic contribution of the individual residues lining the binding pocket of CVN and explore the effect of structural flexibility in the hinge region of CVN on glycan binding, which are also verified experimentally. Overall, these successful applications of BP-Dock highlight the importance of modeling backbone flexibility in docking that can have important implications in defining the binding properties of protein-ligand interactions.

Finally, an induced fit docking approach called Adaptive BP-Dock is presented that allows both protein and ligand conformational sampling during the docking. Adaptive BP-Dock can provide a faster and efficient docking approach for the virtual screening of novel targets for rational drug design and aid our understanding of protein-ligand interactions.

Biophysical techniques have been increasingly applied toward answering biological questions with more precision. Here, three different biological systems were studied with the goal of understanding their dynamic differences, either conformational dynamics within the system or oligomerization dynamics between monomers. With Cy3 on the 5' end of DNA, the effects of changing the terminal base pair were explored using temperature-dependent quantum yields. It was discovered, in combination with simulations, that a terminal thymine base has the weakest stacking interactions with the Cy3 dye compared to the other three bases. With ME1 heterodimers, the goal was to see if engineering a salt bridge at the dimerization interface could allow for control over dimerization in a pH-dependent manner. This was performed experimentally by measuring FRET between monomers containing either a Dap or an Asp mutation and comparing FRET efficiency at different pHs. It was demonstrated that the heterodimeric salt bridge would only form in a pH range near neutrality. Finally, with DNA processivity clamps, one aim was to compare the equilibrium dissociation constants, kinetic rate constants, and lifetimes of the closed rings for beta clamp and PCNA. This was done using a variety of biophysical techniques but with three as the main focus: fluorescence correlation spectroscopy, single-molecule experiments, and time-correlated single photon counting measurements. The stability of beta clamp was found to be three orders of magnitude higher when measuring solution stability but only one order of magnitude higher when measuring intrinsic stability, which is a result of salt bridge interactions in the interface of beta clamp. Ongoing work built upon the findings from this project by attempting to disrupt interface stability of different beta clamp mutants by adding salt or changing the pH of the solution. Lingering questions about the dynamics of different areas of the clamps has led to another project for which we have developed a control to demystify some unexpected similarities between beta clamp mutants. With that project, we show that single-labeled and double-labeled samples have similar autocorrelation decays in florescence correlation spectroscopy, allowing us to rule out the dyes themselves as causing fluctuations in the 10-100 microsecond timescale.

The AAA+ ATPase Rubisco activase (Rca) regulates the activity of Rubisco, the photosynthetic enzyme responsible for catalyzing biological carbon fixation. However, the detailed mechanism by which Rca self-association controls Rubisco reactivation activity remains poorly understood. In this work, we are using fluorescence correlation spectroscopy (FCS) to better characterize the thermodynamics of the assembly process of cotton Rca. We present FCS data for Rca in the presence of Mg*ATPgS and Mg*ADP and for the D173N Walker B motif mutant in the presence of Mg*ATP. Our data are consistent with promotion and stabilization of hexamers by Mg*ATPgS and Mg*ATP, whereas Mg*ADP facilitates continuous assembly. We find that in the presence of Mg·ADP, Rca self-associates in a step-wise fashion to form oligomeric and higher order forms, with a strong size dependence on subunit concentration. The monomer is the dominant species below 0.5 micromolar, whereas the hexamer appears to be most populated in the 10-30 micromolar range. Large assemblies containing on the order of 24 subunits become dominant above 40 micromolar, with continued assembly at even higher concentrations. Our data are consistent with a highly dynamic exchange of subunits among oligomeric species of diverse sizes. The most likely ADP-mediated assembly mechanism seems to involve the formation of spiral supra-molecular structures that grow along the helical axis by the step-wise addition of dimeric units. To examine the effect of Mg·ATP on oligomerization, we have generated the D173N mutant of Rca, which binds but does not hydrolyze ATP. In range of 8 and 70 micromolar, 60-80% of Rca is predicted to form hexamers in the presence of Mg*ATP compared to just 30-40% with Mg*ADP. We see a clear trend at which hexamerization occurs at high ATP:ADP ratios and in addition, at increasing concentrations of free magnesium ions to 5 milimolar that results in formation of six subunits. We present an assembly model where Mg*ATP promotes and stabilizes hexamerization at low micromolar Rca concentrations relative to Mg*ADP, and suggest that this results from closed ring hexamer formation in Mg*ATP and open hexameric spiral formation in Mg*ADP .

Protein affinity reagents have aptly gained profound importance as capture reagents and

drugs in basic research, biotechnology, diagnostics and therapeutics. However, due to the

cost, labor and time associated with production of antibodies focus has recently changed

towards potential of peptides to act as protein affinity reagents. Affinity peptides are easy

to work with, non-immunogenic, cost effective and amenable to scale up. Even though

researchers have developed several affinity peptides, we are far from compiling library of

peptides that encompasses entire human proteome. My thesis describes high throughput

pipeline that can be used to develop and characterize affinity peptides that bind several

discrete sites on target proteins.

Chapter 2 describes optimization of cell-free protein expression using commercially

available translation systems and well-known leader sequences. Presence of internal

ribosome entry site upstream of coding region allows maximal expression in HeLa cell

lysate whereas translation enhancing elements are best suited for expression in rabbit

reticulocyte lysate and wheat germ extract. Use of optimal vector and cell lysate

combination ensures maximum protein expression of DNA libraries.

Chapter 3 describes mRNA display selection methodology for developing affinity peptides

for target proteins using large diversity DNA libraries. I demonstrate that mild denaturant

is not sufficient to increase selection pressure for up to three rounds of selection and

increasing number of selection rounds increases probability of finding affinity peptide s.

These studies enhance fundamental understanding of mRNA display and pave the way

for future optimizations to accelerate convergence of in vitro selections.

Chapter 4 describes a high throughput double membrane dot blot system to rapidly

screen, identify and characterize affinity peptides obtained from selection output. I used

dot blot to screen potential affinity peptides from large diversity of previously

ii

uncharacterized mRNA display selection output. Further characterization of potential

peptides allowed determination of several high affinity peptides from having Kd range 150-

450 nM. Double membrane dot blot is automation amenable, easy and affordable solution

for analyzing selection output and characterizing peptides without ne ed for much

instrumentation.

Together these projects serve as guideline for evolution of cost effective high throughput

pipeline for identification and characterization of affinity peptides.

X-ray diffraction is the technique of choice to determine the three-dimensional structures of proteins. In this study it has been applied to solve the structure of the survival motor neuron (SMN) proteins, the Fenna-Mathews-Olson (FMO) from Pelodictyon phaeum (Pld. phaeum) protein, and the synthetic ATP binding protein DX. Spinal muscular atrophy (SMA) is an autosomal recessive genetic disease resulting in muscle atrophy and paralysis via degeneration of motor neurons in the spinal cord. In this work, we used X-ray diffraction technique to solve the structures of the three variant of the of SMN protein, namely SMN 1-4, SMN-WT, and SMN-Δ7. The SMN 1-4, SMN-WT, and SMN-Δ7 crystals were diffracted to 2.7 Å, 5.5 Å and 3.0 Å, respectively. The three-dimensional structures of the three SMN proteins have been solved. The FMO protein from Pld. phaeum is a water soluble protein that is embedded in the cytoplasmic membrane and serves as an energy transfer funnel between the chlorosome and the reaction center. The FMO crystal diffracted to 1.99Å resolution and the three-dimensional structure has been solved. In previous studies, double mutant, DX, protein was purified and crystallized in the presence of ATP (Simmons et al., 2010; Smith et al. 2007). DX is a synthetic ATP binding protein which resulting from a random selection of DNA library. In this study, DX protein was purified and crystallized without the presence of ATP to investigate the conformational change in DX structure. The crystals of DX were diffracted to 2.5 Å and the three-dimensional structure of DX has been solved.

Cyanovirin-N (CV-N) is a naturally occurring lectin originally isolated from the cyanobacteria Nostoc ellipsosporum. This 11 kDa lectin is 101 amino acids long with two binding sites, one at each end of the protein. CV-N specifically binds to terminal Manα1-2Manα motifs on the branched, high mannose Man9 and Man8 glycosylations found on enveloped viruses including Ebola, Influenza, and HIV. wt-CVN has micromolar binding to soluble Manα1-2Manα and also inhibits HIV entry at low nanomolar concentrations. CV-N's high affinity and specificity for Manα1-2Manα makes it an excellent lectin to study for its glycan-specific properties. The long-term aim of this project is to make a variety of mutant CV-Ns to specifically bind other glycan targets. Such a set of lectins may be used as screening reagents to identify biomarkers and other glycan motifs of interest. As proof of concept, a T7 phage display library was constructed using P51G-m4-CVN genes mutated at positions 41, 44, 52, 53, 56, 74, and 76 in binding Domain B. Five CV-N mutants were selected from the library and expressed in BL21(DE3) E. coli. Two of the mutants, SSDGLQQ-P51Gm4-CVN and AAGRLSK-P51Gm4-CVN, were sufficiently stable for characterization and were examined by CD, Tm, ELISA, and glycan array. Both proteins have CD minima at approximately 213 nm, indicating largely β-sheet structure, and have Tm values greater than 40°C. ELISA against gp120 and RNase B demonstrate both proteins' ability to bind high mannose glycans. To more specifically determine the binding specificity of each protein, AAGRLSK-P51Gm4-CVN, SSDGLQQ-P51Gm4-CVN, wt-CVN, and P51G-m4-CVN were sent to the Consortium for Functional Glycomics (CFG) for glycan array analysis. AAGRLSK-P51Gm4-CVN, wt-CVN, and P51G-m4-CVN, have identical specificities for high mannose glycans containing terminal Manα1-2Manα. SSDGLQQ-P51Gm4-CVN binds to terminal GlcNAcα1-4Gal motifs and a subgroup of high mannose glycans bound by P51G-m4-CVN. SSDGLQQ-wt-CVN was produced to restore anti-HIV activity and has a high nanomolar EC50 value compared to wt-CVN's low nanomolar activity. Overall, these experiments show that CV-N Domain B can be mutated and retain specificity identical to wt-CVN or acquire new glycan specificities. This first generation information can be used to produce glycan-specific lectins for a variety of applications.

Spinal muscular atrophy (SMA) is a neurodegenerative disease that results in the loss of lower body muscle function. SMA is the second leading genetic cause of death in infants and arises from the loss of the Survival of Motor Neuron (SMN) protein. SMN is produced by two genes, smn1 and smn2, that are identical with the exception of a C to T conversion in exon 7 of the smn2 gene. SMA patients lacking the smn1 gene, rely on smn2 for production of SMN. Due to an alternative splicing event, smn2 primarily encodes a non-functional SMN lacking exon 7 (SMN D7) as well as a low amount of functional full-length SMN (SMN WT). SMN WT is ubiquitously expressed in all cell types, and it remains unclear how low levels of SMN WT in motor neurons lead to motor neuron degradation and SMA. SMN and its associated proteins, Gemin2-8 and Unrip, make up a large dynamic complex that functions to assemble ribonucleoproteins. The aim of this project was to characterize the interactions of the core SMN-Gemin2 complex, and to identify differences between SMN WT and SMN D7. SMN and Gemin2 proteins were expressed, purified and characterized via size exclusion chromatography. A stable N-terminal deleted Gemin2 protein (N45-G2) was characterized. The SMN WT expression system was optimized resulting in a 10-fold increase of protein expression. Lastly, the oligomeric states of SMN and SMN bound to Gemin2 were determined. SMN WT formed a mixture of oligomeric states, while SMN D7 did not. Both SMN WT and D7 bound to Gemin2 with a one-to-one ratio forming a heterodimer and several higher-order oligomeric states. The SMN WT-Gemin2 complex favored high molecular weight oligomers whereas the SMN D7-Gemin2 complex formed low molecular weight oligomers. These results indicate that the SMA mutant protein, SMN D7, was still able to associate with Gemin2, but was not able to form higher-order oligomeric complexes. The observed multiple oligomerization states of SMN and SMN bound to Gemin2 may play a crucial role in regulating one or several functions of the SMN protein. The inability of SMN D7 to form higher-order oligomers may inhibit or alter those functions leading to the SMA disease phenotype.

Membrane protein structure is continuing to be a topic of interest across the scientific community. However, high resolution structural data of these proteins is difficult to obtain. The amino acid transport protein, Outer Envelope Protein, 16kDa (OEP16) is a transmembrane protein channel that allows the passive diffusion of amino acids across the outer chloroplast membrane, and is used as a model protein in order to establish methods that ultimately reveal structural details about membrane proteins using nuclear magnetic resonance (NMR) spectroscopy. Methods include recombinant expression of isotope enriched inclusion bodies, purification and reconstitution in detergent micelles, and pre-characterization techniques including circular dichroism (CD) spectroscopy, dynamic light scattering (DLS), and high pressure liquid chromatography (HPLC). High resolution NMR spectroscopy was able to assign 99% of the amide backbone and the chemical shifts provided detailed secondary structure of OEP16 on a per residue basis using the software TALOS+. Relaxation studies explored the intramolecular dynamics of OEP16 and results strongly support the resonance assignments. Successful titration studies were able to locate residues important for amino acid binding for import into the chloroplast as well as provide information on how the transmembrane helices of OEP16 are packed together. For the first time there is experimental evidence that can assign the location of secondary structure in OEP16 and creates a foundation for a future three dimensional structure.

Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) is widely accepted as the world's most abundant enzyme and represents the primary entry point for inorganic carbon into the biosphere. Rubisco's slow carboxylation rate of ribulose-1,5-bisphosphate (RuBP) and its susceptibility to inhibition has led some to term it the "bottle neck" of photosynthesis. In order to ensure that Rubisco remains uninhibited, plants require the catalytic chaperone Rubisco activase. Activase is a member of the AAA+ superfamily, ATPases associated with various cellular activities, and uses ATP hydrolysis as the driving force behind a conformational movement that returns activity to inhibited Rubisco active sites. A high resolution activase structure will be an essential tool for examining Rubisco/activase interactions as well as understanding the activase self-association phenomenon. Rubisco activase has long eluded crystallization, likely due to its infamous self-association (polydispersity). Therefore, a limited proteolysis approach was taken to identify soluble activase subdomains as potential crystallization targets. This process involves using proteolytic enzymes to cleave a protein into a few pieces and has previously proven successful in identifying crystallizable protein fragments. Limited proteolysis, utilizing two different proteolytic enzymes (alpha-chymotrypsin and trypsin), identified two tobacco activase products. The fragments that were identified appear to represent most of what is considered to be the AAA+ C-terminal all alpha-domain and some of the AAA+ N-terminal alpha beta alpha-domain. Identified fragments were cloned using the pET151/dTOPO. The project then moved towards cloning and recombinant protein expression in E. coli. NtAbeta(248-383) and NtAbeta(253-354) were successfully cloned, expressed, purified, and characterized through various biophysical techniques. A thermofluor assay of NtAbeta(248-383) revealed a melting temperature of about 30°C, indicating lower thermal stability compared with full-length activase at 43°C. Size exclusion chromatography suggested that NtAbeta(248-383) is monomeric. Circular dichroism was used to identify the secondary structure; a plurality of alpha-helices. NtAbeta(248-383) and NtAbeta(253-354) were subjected to crystallization trials.