(A) Conformation of 4-OHT (magenta) docked onto the crystal structure of the N-terminal ATP-binding domain of human Hsp90α (yellow mesh, pdb code 1uy8). 4-OHT occupies the nucleotide binding pocket as illustrated by the superimposed structure of human Hsp90α N-terminal domain (not shown for clarity–pdb code 1byq, backbone RMSD 1.45 Å) co-crystallized with ADP (white). Unlike ADP, the 4-OHT aliphatic chain extends deep into the pocket, and makes multiple hydrophobic contacts with the lid which acts as a structural switch between the active (green–pdb code 2cg9) and inactive (red–pdb code 1uy8) conformation of the enzyme. (B–D) The effects of 4-OHT and raloxifene on Hsp90 ATPase activity are shown in low salt (40 mM KCl, red) and high salt (400 mM KCl, blue) buffer. All ATPase activities refer to the ATPase inhibited by 20 µM geldanamycin (refer to Materials and Methods) and are relative to that of yeast Hsp82, which has a measured geldanamycin-inhibited ATPase activity of 4.85 (±0.19) pmol Pi min⁻¹ µg⁻¹. Data shown are averages of three to five repeats. Error bars refer to the standard deviation. Stars on selected bars indicate that the average value obtained for that experiment is statistically different from the average obtained in the absence of the drug using the Z-test at the 95% confidence limit.

(A) Raw ATPase data showing the effect of 4-OHT on yeast Hsp82 as monitored by the decrease in NADH absorbance. (B) Effect of geldanamycin on human Hsp90α ATPase activity in the presence of 52 µM 4-OHT in high salt buffer. (C) Effect of 4-OHT on the ATPase activity of E. coli ClpX in low salt buffer. (D) Effect of 4-OHT on the ATPase activity of human Hsp70 in low salt buffer. All ATPase activities given refer to total activity and not to the geldanamycin-inhibited ATPase. They are given relative to the geldanamycin-inhibited ATPase activity of yeast Hsp82 [4.85 (±0.19) pmol Pi min⁻¹ µg⁻¹]. Data shown are averages of three to five repeats. Error bars refer to the standard deviation. Stars on selected bars indicate that the average value obtained for that experiment is statistically different from the average obtained in the absence of the drug using the Z-test at the 95% confidence limit.

Various concentrations of ATP, ADP, raloxifene, and 4-OHT were injected over a surface immobilized with yeast Hsp82. The sensorgrams shown have been double referenced. The steady state responses were fit, where possible, using nonlinear regression to a single class of binding site model to obtain the K_d values indicated. Numbers in parenthesis give the standard errors on the K_d obtained from the fits.

Various concentrations of ATP or ADP were injected over a biosensor chip with immobilized Hsp82N using running buffer containing no drug (top panels), 25 µM raloxifene (middle panels), or 25 µM 4-OHT (bottom panels). The sensorgrams shown have been double referenced. The steady state responses were fit as in Figure 3.