Tubulin is a universally conserved molecule, found in all three domains of life. Tubulin filaments use energy derived from GTP binding and hydrolysis to organize cytoplasm. Although tubulin homologues share the same fold to their monomers and similar protofilament architecture, they assemble filaments with unique geometric designs. Moreover, despite sharing the mechanism of self-assembly, wherein tubulin subunits require GTP binding to polymerize, and their polymerization stimulates GTP hydrolysis, tubulins exhibit distinctive dynamic properties. Together, unique filament architectures and dynamic properties determine a specific set of biological functions each tubulin family performs. The origins of tubulin filament architectural diversity and distinctive dynamic behavior are not well understood.

We have been studying PhuZ tubulins, which are encoded by a few very large Pseudomonas ϕKZ-like bacteriophages. Our studies have shown that PhuZ assembles dynamically unstable spindle-like arrays that organize bacteriophage DNA at the cell midpoint, which somehow facilitates phage infectivity. Moreover, PhuZ monomer structure has the canonical tubulin fold, with a unique, highly conserved and extended C-terminus. The C-terminus mediates protofilament contacts and is critical for polymerization both in vitro and in vivo.

The main focus of this manuscript is the structural investigation of the molecular mechanisms underlying PhuZ dynamic behavior. In an attempt to explain how GTP binding and hydrolysis drives PhuZ filament turnover, we have conducted a number of electron cryomicroscopy studies (cryo-EM) on PhuZ filaments in both pre- and post- hydrolysis states. PhuZ forms a polar three-stranded polymer with an unusual subunit orientation. Its C-terminus guides a cooperative assembly of the three-stranded filament by mediating both longitudinal and lateral interactions. Upon assembly, the longitudinal interface, C-terminus and tubulin fold undergo polymerization-competent rearrangements. We propose that the energy of GTP binding is stored in the displacement of these structural elements. Also, our structural studies have revealed how the energy of GTP hydrolysis is stored in the PhuZ filament lattice. In particular, GTP hydrolysis, initially sensed by the T3 loop and C-terminus, is accompanied by unwinding and supercoiling of the filament lattice. Based on these observations, we propose an assembly/disassembly pathway of PhuZ filament.

Detailed knowledge of folding intermediate and transition state (TS) structures is critical for understanding protein folding mechanisms. For kinetically-stable proteins such as α-lytic protease (αLP) and its family members, their large free energy barrier to unfolding is central to their biological function. Thus their TS structure plays a crucial role in protein function. However, structural information regarding this important state has been completely lacking, mainly because standard techniques to probe TS structure are not realistically applicable for αLP. Therefore, I used the information embedded in the sequence of homologous proteases to discern the physical mechanisms by which kinetic stability can be modulated. This required experimental validation using various biophysical and biochemical techniques, such as mutagenesis, x-ray crystallography, and detailed kinetic analyses.

From these studies, I have shown that the conserved distortion of a sidechain significantly contributes to the destabilization of the TS for a large sub-class of αLP homologs. The strain from this deformation actually provides a biological advantage in that lifetime is greatly extended. This study was the first that shows that sidechain distortion has been shown to be used for a functional purpose and uncovers an unanticipated challenge for structural biology to identify potentially relevant distortions from high resolution structural studies.

My structural and kinetic analysis of a acid resistant αLP homolog, Nocardiopsis alba Protease (NAPase), identified the physical basis for this proteins acid stability, thus providing crucial structural information about unfolding mechanisms and leading to a model for the TS structure for these proteases. This study provided insight into the evolutionary benefits of kinetic stability as a paradigm for generation of extremophilic behavior.

From a similar study of a thermophilic αLP homolog, Thermobifida fusca Protease A (TFPA), I identified a substructure of these proteases, termed the domain bridge, which is used to modulate the degree of kinetic stability. This study refined our model for the unfolding TS, in which the domain bridge undocks and unfolds allowing the two domains of the protease to separate, with the newly formed crevice filling with solvent. These studies represent the first physical understanding of the structural basis for kinetic stability.

α-lytic protease (αLP), a bacterial serine protease of the chymotrypsin family, has evolved both kinetic stability and an extreme unfolding cooperativity in order to limit proteolysis. Trypsin, a well-studied metazoan homolog, is degraded at rates up to 100x faster than αLP even though it has approximately the same global unfolding rate. Previous experimental studies have implicated the interface between αLP's two domains as critical to the unfolding pathway and its cooperativity. To investigate this, I performed multiple high temperature molecular dynamics unfolding simulations on both αLP and trypsin. The simulations revealed a robust unfolding pathway that featured preferential disruption of the domain interface, primarily at three regions: the Domain Bridge, the C-terminal domain β-hairpin and cis-proline turn, and the N-terminus. I developed a metric for measuring global unfolding cooperativity, and it showed correctly that αLP unfolded cooperatively, while trypsin did not. I then applied an information-theory-derived measure of cooperativity developed by Voelz, based on pairs of contacts in the two proteins, to the simulations, allowing me to look at cooperativity at the residue level. By graphing the contact cooperativity as a network, I showed that the αLP network is significantly larger and more connected than that of trypsin, again showing a much higher global unfolding cooperativity. Using only the early parts of the simulations, the cooperativity network highlights contacts broken cooperatively around the transition state ensemble. These network graphs also identify residues that are key centers of cooperativity and if mutated, may disrupt αLP's unfolding cooperativity. Experimental studies are currently underway in the lab to test the hypotheses created from these simulations.

Due to the classical conflict between spatial and temporal resolution, microscopy studies of fast events in living samples are often performed in 2D even when 3D imaging would be desirable and could provide new insights to biological function. This dissertation describes an instant 3D imaging system - a multi-focus microscope (MFM) - which provides high- resolution, aberration-corrected, multi-color fluorescence images of multiple focal planes

simultaneously. Forming an instant focal series eliminates the need for multiple camera exposures and mechanical refocusing, allowing 3D imaging limited only by sample signal strength and the camera read-out rate for a single frame. A module containing the MFM optical components can easily be appended to the camera port of a commercial wide-field microscope. The excellent resolution and sensitivity of MFM is demonstrated on two different 3D biological imaging problems; neuronal imaging in the entire C.elegans embryo and mRNA imaging in cultured mammalian cells.

Hsp90 is a highly conserved, ATP-dependent molecular chaperone that is essential for maintaining the functions of its client proteins. It has been estimated that about 10% of the proteome of a eukaryotic cell interacts with Hsp90. A large subset of this portion consists of protein kinases and steroid hormone receptors, putting Hsp90 as the master regulator of many essential cellular functions. The mechanism of how Hsp90 uses ATP hydrolysis to carry out its function remains unclear. Structural studies of Hsp90 revealed that Hsp90 is a V-shaped homodimer with each protomer composed of three well-folded domains: an ATP-binding N-terminal Domain (NTD), a Middle Domain (MD), and a C-terminal dimerization Domain (CTD). Efficient ATP hydrolysis by Hsp90 requires that the dimer forms a closed state where NTDs are dimerized, forming a closed ”clamp” conformation that is stabilized by ATP binding. The kinetics of forming this stably closed state is not driven by ATP binding, as there are other rate-limiting steps that need to occur within the protein to allow the NTDs to be dimerized. So how does Hsp90 actually use the energy from ATP to remodel its client protein? the focus of this thesis examines the other possibility of this happening during ATP hydrolysis. In the first chapter, I followed up an observation made by a previous graduate student Laura Lavery. She observed that the ATP-bound closed state of a mitochondrial Hsp90 (TRAP1) is asymmetric. The asymmetry is most prominent at the juncture between the MD:CTD interface—one protomer is buckled while the other remains straight, resembling the same conformation previously observed in the symmetric closed ”clamp” state. This buckling happens precisely where conserved binding sites have been mapped for client proteins. This suggests that if conformational changes were to occur due to ATP hydrolysis, subsequent rearrangements of the asymmetric MD:CTD interfaces back to the previously observed symmetric closed state

vii

could be used to drive client protein remodeling. Using a combination of biophysical methods (crystallography, Double Electron-Electron Resonance (DEER), and FRET), we observed that TRAP1 hydrolyzes the 2 bound ATPs sequentially. The buckled protomer hydrolyzes the first ATP, which is then followed by a flip in the asymmetry (the buckled conformation becomes straight and vice versa, on each side), which primes the second ATP for hydrolysis by a buckled protomer. In this model, the MD:CTD interface is guaranteed to undergo remodeling with each ATP hydrolysis and would make efficient use of energy from ATP. The implications for this asymmetric ATP hydrolysis mechanism may also be relevant to other Hsp90s. While we have not observed any other asymmetric Hsp90 structures by itself, several functional Hsp90 complexes seen so far seem to have asymmetric composition/arrangements of their components. In the second chapter, we explore how TRAP1 ATPase activity can be modulated by different divalent cations as co-factors. Despite having two ATP binding sites, the ATPase activity of most Hsp90 homologs appears to be non-cooperative (each site behaves independently from one another). However, we saw that ATPase activity of TRAP1 can be cooperative in presence of calcium, and the activity in presence of magnesium appears to be bi-phasic, with higher activity at low ATP concentrations. This unique behavior of TRAP1 may yet be another adaptation of the Hsp90 machine that has evolved within the mitochondrial matrix environment. Using crystallography, we also discovered that calcium binds to the NTD of TRAP1 unlike previously observed chaperone/calcium/ATP complexes. While the exact biological role for this phenomena is not yet clear, these findings provide a clear molecular basis for the regulation of TRAP1 by calcium. Taken together, the work described in this thesis provide insights into the mechanism of ATP hydrolysis by Hsp90, and a potential role that TRAP1 plays in calcium/magnesium-regulated mitochondrial physiology.

Heat Shock Protein 90 (Hsp90) is a highly conserved molecular chaperone necessary for eukaryotic life. This member of the cellular folding machinery has been extensively studied in recent years due to its intimate link to fundamental biological pathways that govern cellular homeostasis and disease.

Hsp90 undergoes large ATP dependant conformational changes that are necessary for client maturation in vivo. Though much progress had been made in building a model for a functional cycle of Hsp90, the fundamental question of how Hsp90 functions to re-model clients remained unclear.

Several forms of Hsp90 can be found throughout evolution with most higher eukaryotes having four homologs: two in the cytosol (Hsp90α, Hsp90β), one in the endoplasmic reticulum (GRP94) and one in the mitochondria (TRAP1).

Though not extensively studied at the time, I set out to investigate the structure and function of TRAP1, originally inspired by the balance between a higher eukaryotic Hsp90 and a simplified Hsp90 being most similar to the bacterial Hsp90 (bHsp90). With the latter comes the advantage of a more simple system where the large number of cochaperones identified for Hsp90α/β are not known to be required for client interactions with TRAP. Further, unique biology and disease links of TRAP1 were emerging with in vivo data showing that TRAP1 had unique ties to cellular homeostasis through interactions with client protein cyclophilin D (CypD), the master regulator of the mitochondrial permeability transition pore (mtPTP) and thus necrotic cell death.

Through this work I developed several biochemical/biophysical assays and strategies for working with TRAP1 in vitro, and through collaboration elucidated both shared and unique aspects of Hsp90 mechanism. Importantly, this work has provided a new hypothesis for Hsp90 mechanism of action with client proteins, which contributes a new model for the productive use of ATP hydrolysis for client re-modeling.