Multiple conformations of E. coli Hsp90 in solution: insights into the conformational dynamics of Hsp90

Abstract

Hsp90, an essential eukaryotic chaperone, depends upon its intrinsic ATPase activity for function. Crystal structures of the bacterial Hsp90 homolog, HtpG, and the yeast Hsp90 reveal large domain rearrangements between the nucleotide-free and the nucleotide-bound forms. We used small-angle X-ray scattering and recently developed molecular modeling methods to characterize the solution structure of HtpG and demonstrate how it differs from known Hsp90 conformations. In addition to this HtpG conformation, we demonstrate that under physiologically relevant conditions, multiple conformations coexist in equilibrium. In solution, nucleotide-free HtpG adopts a more extended conformation than observed in the crystal, and upon the addition of AMPPNP, HtpG is in equilibrium between this open state and a closed state that is in good agreement with the yeast AMPPNP crystal structure. These studies provide a unique view of Hsp90 conformational dynamics and provide a model for the role of nucleotide in effecting conformational change.

Figure 1. Experimental SAXS data for apo and AMPPNP-HtpG

A) Averaged and scaled solvent-subtracted scattering curves (I(Q)) for full-length HtpG without nucleotide (black) and in the presence of saturating concentrations (10mM AMPPNP, purple). B) Interatomic distance distribution functions (P(r)) calculated from the scattering data shown in panel A (black and purple for apo- and AMPPNP-HtpG, respectively). The P(r) curves are normalized to have equivalent areas under the curve. In solution, apo-HtpG is significantly expanded compared to the AMPPNP state.

Figure 2. Comparing apo HtpG scattering data to the apo HtpG crystal structure

A) Crystal structure for apo HtpG (Shiau, et al., 2006) and the homology model for AMPPNP-HtpG based upon the yeast AMPPNP bound crystal structure (Ali, et al., 2006). The homology model was created by aligning the domains of HtpG to those of the yeast crystal structure. The NTD is shown in blue, the MD in green and the CTD in brown. The second monomer is shown in grey. B) Normalized P(r) curves calculated from the apo crystal structure (red) and the AMPPNP-HtpG homology model (blue) compared to the normalized experimental P(r) for apo HtpG (black). Neither the apo or AMPPNP crystal structures match the experimental SAXS data.

Figure 3. Modeling the apo scattering data by altering the opening angle of the apo crystal structure

A) Schematic showing the setup for the rigid body refinement protocol. The apo crystal structure is treated as three rigid bodies with the CTD dimerization domain’s orientation fixed. The orientations of the NM domains were then allowed to vary while maintaining the two-fold symmetry of the dimer. B) Normalized P(r) showing the fit of the best model (red) to the apo experimental data (black). C) The best fitting model (R = 3.03%) resulting from the rigid body refinement described in panel A. The opening angle between the monomers is ~75° in the apo crystal structure and expands to ~120° in the model, providing a greatly improved estimate of the structure of HtpG in solution.

Figure 4. Comparison of crystallographically-observed NM domain structures to that observed in solution by SAXS

Volume representations of each of the four known NM domain configurations are shown with silhouettes representing the full-length structure that complements each configuration. The normalized P(r) curve for each structure (apo R=11.9%, AMPPNP R=4.06%, ADP extended R=7.64%, ADP compact R=48.8%) is shown in the corresponding color, and the experimental P(r) for NMHtpG is shown in black. While none of the calculated curves completely match the experimental data, the extended AMPPNP structure is best.

Figure 5. Modeling the SAXS data for the NMHtpG

A) Clustering of the solutions from one modeling run of the NM domain data yields three well-populated groups. B) P(r) curves for the top model from each cluster as compared to the experimental data shown in black. All the top models fit the experimental data. C) Structure of the top models whose P(r) functions are shown in the corresponding color in B. D) Alignment of the volumes of the models shown in panel C reveals they all have the same molecular volume. The differences result from rotations about the nearly cylindrical MD.

Figure 6. Modeling the full-length apo SAXS data using different NM domain configurations

A) P(r) curves were calculated for each of the full-length apo models using PRFIT. The middle domain of each model was kept aligned according to the secondary structure. The P(r) curves for the full-length apo models were compared to the experimental full-length apo SAXS data (black ◆). Each color represents a different NM domain configuration calculated from the NM domain experimental SAXS data, as shown in Figure 5. Blue (◆) represents the full-length model using the AMPPNP-bound NM domain. B) The apo models in panel A were further refined using the rigid body modeling procedure implemented in IMP. Each model is shown in the same color as in panel A. C) Residuals of the four different models as compared to the experimental full-length apo SAXS data reveals that one model (green curve) fits significantly better than the others (R=2.6%). D) Structure of the best fitting model. The marked rotations show the transformations required to go from the apo crystal structure to the SAXS model.

Figure 7. Modeling the HtpG:AMPPNP scattering data

A) Normalized P(r) for the AMPPNP-HtpG homology model shown in Figure 2A (red) compared to the experimental P(r) for AMPPNP-HtpG (black). B) Modeling the HtpG:AMPPNP experimental data (black) using a linear combination (red) of 44.5% of the AMPPNP-bound model in Figure 2A and 55.5% of the apo solution structure shown in Figure 6. Even with saturating nucleotide, there is an equilibrium between open and closed states.

Figure 8. Implications for the discussed results on the function of Hsp90

A, B) Potential substrate binding modes of Hsp90. Panel A shows the extended solution structure that could bind large substrates on the order of 130Å. Panel B shows the Progesterone receptor ligand-binding domain (1SQN; (Madauss, et al., 2004) docked to the crystal structure of Hsp90, demonstrating how Hsp90 may bind smaller client proteins. As proposed by (Fang, et al., 2006), the CTD amphipathic helix could readily displace the ligand-gated helix12 of the receptor. C) Model for the effects of ATP on Hsp90 conformation: In the unliganded state HtpG is in an open state that is relatively flat and has a large cleft that is accessible for client protein binding. Once ATP is bound, an equilibrium is established between the open state and a compact closed state.