Discovering the Structural Basis for Biological Function at the Molecular and Cellular Levels
Summary: David Agard's research is focused on elucidating the mechanism of Hsp90 chaperone function and its role in human disease, microtubule nucleation, and centrosome structure and on the structure and cell biology of phage-encoded tubulins.
Mechanism of Hsp90 Function
In eukaryotes, the ubiquitous Hsp90 molecular chaperone facilitates the folding and activation of a broad array of proteins important in cell signaling, proliferation, and survival. Unlike other molecular chaperones, Hsp90 preferentially stabilizes near-native-state structures, aiding the dynamic assembly and disassembly of signaling complexes. Hsp90 is thus an important therapeutic target. Our goal is to understand Hsp90 action and the structural basis for its requirement by its substrate client proteins and to use this information in the discovery of novel small-molecule modulators.
We solved the x-ray structures of the Escherichia coli Hsp90 in the apo and ADP states; a structure of the yeast ATP state was determined concurrently in Laurence Pearl's lab (Institute of Cancer Research, London). Through small-angle x-ray scattering and single-particle electron microscopy (EM), we demonstrated that bacterial, yeast, and human Hsp90s have a conserved 3-state ATP conformational cycle and that the open-closed equilibrium is species specific, reflecting optimization for the unique requirements of each species. These results profoundly affected models of the Hsp90 conformational cycle, suggesting where client proteins may bind and how nucleotide binding and hydrolysis propel the chaperone through conformational changes that lead to the release of client proteins.
Our current efforts focus on elucidating the structural basis for Hsp90-client and Hsp90-cochaperone interactions and understanding how Hsp90 remodels its clients. This is challenging as Hsp90 has a strong preference for partially folded states, which are normally only sparingly populated. To remedy this, we developed a model system based on a largely unfolded, but nonaggregating, mutant of staphylococcal nuclease (SN). By 15N-NMR (nuclear magnetic resonance) we defined the binding site for SN at the interface between middle and C-terminal Hsp90 domains, showed that a 25- to 30-residue SN segment, corresponding to the most folded region within SN, is responsible for the interaction and remains structured as an α-helix when bound. Kinetic experiments show an unexpected coupling between conformational transition states in both client and Hsp90. Notably, model and natural clients accelerate the rate-limiting conformation transitions in Hsp90, thereby speeding ATP hydrolysis.
We are simultaneously pursuing several classes of natural clients, including the glucocorticoid receptor (GR), kinases, and E3 ligases and have obtained cryoEM structures and mechanistic information on Hsp90:Cdc37:kinase at ~12-Å resolution and Hsp90:Hop at 16 Å; we are making progress on the Hsp90:Hop:Hsp70:GR complex. The human mitochondrial Hsp90 (TRAP1) is linked to Parkinson's disease and cancer. We have solved its atomic structure, revealing an unexpected asymmetric ATP state that likely has significant functional importance. We are using this system to discover small-molecule allosteric modulators.
Mechanisms of Microtubule Formation and the Role of γ-Tubulin Complexes as Nucleators
The spatial and temporal regulation of tubulin polymerization into microtubules (MTs) is a central question in cell biology. Our goal is to understand, in atomic detail, the molecular mechanisms underlying dynamic MT behavior and MT nucleation.
Of key importance is the structural and functional analysis of γ-tubulin complexes, which act in vivo to nucleate MT growth. Although central to all MT nucleation, the γ-tubulin small complex (γTuSC) is a surprisingly poor MT nucleator. We have determined the structure of the isolated yeast γTuSC by single-particle EM. We can assemble γTuSCs into either ring complexes or filaments, and we have recently determined the cryo-EM structure of the filament at ~8-Å resolution. These structures have 6.5 γTuSCs/turn, resulting in the display of 13 γ-tubulins, explaining how pairs of γ-tubulins present within γTuSCs can nucleate 13-protofilament MTs. Although yeast γTuSC assemblies can form spontaneously, they are only stable at low pH. However, Spc110p, which links γTuSC to the spindle pole body (SPB), stabilizes the assemblies to physiological conditions. Thus, γTuSC assemblies are only formed at the SPB, ensuring high fidelity of MT nucleation. The cryoEM structure also reveals that the γ-tubulins within each γTuSC are too far apart to efficiently nucleate MTs, indicating that allosteric activation is required for potent MT nucleation. By engineering a closed form, the EM resolution can be extended to 6.5 Å, allowing a detailed model to be built from a distantly related human accessory protein. This is being used to help discover the allosteric regulator and to understand the role of post-translational modifications.
As part of our efforts to understand the molecular basis of MT nucleation, we solved the atomic structures of human γ-tubulin complexed with GTP and GDP, providing the first eukaryotic GTP/GDP pair. These structures provided two key insights: γ-tubulin forms MT-like lateral interactions independent of nucleotide, and γ-tubulin remains in a curved conformation independent of nucleotide, contrasting sharply with the prevailing allosteric hypothesis for activation of tubulin assembly by GTP. Solution studies (conformation-specific ligand binding and small-angle x-ray scattering on αβ-tubulin confirm that it too remains in a curved conformation independent of nucleotide. As a major paradigm shift, we proposed that the lattice and not the nucleotide is the allosteric effector. In this lattice model, GTP acts only to tune the longitudinal affinity. Lattice metastability is determined not by GTP hydrolysis but by the mechanical spring constant for straightening. This new view has a dramatic impact on understanding MT formation. To explore this, we have developed new first-principle kinetic models of MT assembly and shown how a spring penalty is required to match observed behavior.
The Structure and Cell Biology of Bacteriophage Tubulins
In collaboration with Joe Pogliano (University of California, San Diego), we have discovered a family of tubulins encoded by very large bacteriophage (~300 Kb) and have determined the crystal structure of the first phage cytoskeletal element—a tubulin family member known as PhuZ. While longitudinal interactions are well conserved in all tubulins, the crystal structure, which mimics longitudinal packing in the filaments, revealed a gap between molecules. Instead, filaments are formed by a novel C-terminal extension that directly links molecules in the filament. In vivo observation of infected cells reveals that PhuZ forms a spindle-like structure that positions phage particles at the cell midline. Altering PhuZ dynamics leads to phage mispositioning and a significantly decreased burst size. In a continued collaboration, we now focus on understanding the cell and structural biology underlying this remarkable observation.
Grants from the National Institutes of Health provided partial support for the microtubule/centrosome efforts. An NIH grant partially supports structural studies on Hsp90. A UCSF SPORE grant partially supports Hsp90 small-moleculr discovery efforts. A grant from the National Science Foundation is funding a collaborative effort (University of California, San Francisco/Lawrence Berkeley National Laboratory/GATAN) to develop a next-generation direct-detect electron camera for high-resolution cryo-EM.
As of November 02, 2012