In vivo function of Hsp90 is dependent on ATP binding and ATP hydrolysis

Abstract

Heat shock protein 90 (Hsp90), an abundant molecular chaperone in the eukaryotic cytosol, is involved in the folding of a set of cell regulatory proteins and in the re-folding of stress-denatured polypeptides. The basic mechanism of action of Hsp90 is not yet understood. In particular, it has been debated whether Hsp90 function is ATP dependent. A recent crystal structure of the NH2-terminal domain of yeast Hsp90 established the presence of a conserved nucleotide binding site that is identical with the binding site of geldanamycin, a specific inhibitor of Hsp90. The functional significance of nucleotide binding by Hsp90 has remained unclear. Here we present evidence for a slow but clearly detectable ATPase activity in purified Hsp90. Based on a new crystal structure of the NH2-terminal domain of human Hsp90 with bound ADP-Mg and on the structural homology of this domain with the ATPase domain of Escherichia coli DNA gyrase, the residues of Hsp90 critical in ATP binding (D93) and ATP hydrolysis (E47) were identified. The corresponding mutations were made in the yeast Hsp90 homologue, Hsp82, and tested for their ability to functionally replace wild-type Hsp82. Our results show that both ATP binding and hydrolysis are required for Hsp82 function in vivo. The mutant Hsp90 proteins tested are defective in the binding and ATP hydrolysis-dependent cycling of the co-chaperone p23, which is thought to regulate the binding and release of substrate polypeptide from Hsp90. Remarkably, the complete Hsp90 protein is required for ATPase activity and for the interaction with p23, suggesting an intricate allosteric communication between the domains of the Hsp90 dimer. Our results establish Hsp90 as an ATP-dependent chaperone.

Figures

Figure 1

Structure of the NH2-terminal domain of Hsp90 with bound ADP-Mg. View into the nucleotide-binding pocket. The residues D93 and E47, critical for nucleotide binding and hydrolysis, respectively, are colored in yellow. Image was prepared with the programs MOLSCRIPT (Kraulis, 1991) and RASTER3D (Merrit and Murphy, 1994).

Figure 2

Mutations in the nucleotide-binding site of Hsp90 abolish the function of Hsp90 in vivo. (A) Yeast strain ΔPCLDa/α that expresses wild-type Hsc82 from the plasmid pKAT6 (containing the URA3-marker) was cotransformed with wild-type HSP82 (lane 1), HSP82His6EEF (lane 2), HSP82(E33A)His6EEF (lane 3), HSP82(E33D)His6EEF (lane 4) and HSP82(D79N) His6EEF (lane 5). After growth to mid-log phase in liquid SD/ −Trp/−Ura, cell lysates were prepared, adjusted to equal protein concentrations and His6EEF tagged proteins quantitatively absorbed with Ni-NTA beads followed by SDS-PAGE and Coomassie staining (lanes 1–5). Lanes 6–10 correspond to lanes 1–5 and show an immunoblot with the EEF-specific antibody. Note that only the band at ∼90 kD (and two degradation products at ∼60 and ∼50 kD, generated during isolation) are specifically recognized. (B) The cotransformants described in A were restreaked on 5-FOA plates at 25°C to select for cells that had lost the original wild-type plasmid and were rescued from lethality by a functional hsp82 protein. Only wild-type Hsp82, Hsp82His6EEF, and the E33D mutant were able to support growth.

Figure 3

Mutations in the nucleotide binding site of Hsp90 affect ATP binding and hydrolysis differentially. (A) Binding of ATP to the NH2-domains of wild-type and mutant Hsp82 in the presence and absence of GA (see Materials and Methods). Averages of three independent experiments are shown with error bars. (B) A typical plot of ATP hydrolysis versus time at a concentration of 0.05 mM ATP and 10 μM Hsp82. ATPase activity was measured with [α32P]ATP at 30°C as described in Materials and Methods. The ATPase activity of Hsp82 (closed circles) and the E33D mutant (closed triangles) is specifically inhibited by GA (open circles and open triangles, respectively). In contrast, the E33A (closed squares) and D79N mutants (open squares) are devoid of GA-inhibitable ATPase activity. (C) Double reciprocal plot of the rate of ATP hydrolysis versus ATP concentration. Km values for wild-type Hsp82 (closed circles) and for the E33D mutant (closed triangles) were determined as 172 μM and 520 μM, respectively.

Figure 4

The complete Hsp90 protein is required for ATPase activity. ATP hydrolysis versus time at 0.10 mM ATP and 10 μM Hsp82 proteins at 30°C is shown for full-length Hsp82 (FL82, closed circles), the NH2 domain of Hsp82 (N82, closed triangles) and for ΔC82, lacking the COOH-terminal residues 600–709 (closed squares).

Figure 5

Requirement of full-length Hsp90 with an intact ATP-binding site for the interaction with p23 in vivo. (A) Schematic representation of the domain structure of human Hsp90 established by limited proteolysis (Stebbins et al., 1997) and of the fragments used in the two-hybrid assay. (B) Different fragments of Hsp90 in the pAS2-1 bait vector were cotransformed in S. cerevisiae (strain Y190) with the cDNA encoding human p23 in the pACT2 target vector (see Materials and Methods). To monitor protein–protein interactions, cotransformants were restreaked on SD/−Trp/−Leu/−His plates containing 25 mM 3-AT and incubated for 5 d at 30°C. Cell growth was only observed with full-length Hsp90 protein but not for any of its subfragments or for the vector control without insert. (C) Full-length wild-type Hsp90 and the mutants E47A, E47D, and D93N were analyzed for their ability to interact with p23 by two-hybrid assay as above. Cell growth was observed for wild-type Hsp90, the E47A and E47D mutants, but not for the D93N mutant.

Figure 5

Requirement of full-length Hsp90 with an intact ATP-binding site for the interaction with p23 in vivo. (A) Schematic representation of the domain structure of human Hsp90 established by limited proteolysis (Stebbins et al., 1997) and of the fragments used in the two-hybrid assay. (B) Different fragments of Hsp90 in the pAS2-1 bait vector were cotransformed in S. cerevisiae (strain Y190) with the cDNA encoding human p23 in the pACT2 target vector (see Materials and Methods). To monitor protein–protein interactions, cotransformants were restreaked on SD/−Trp/−Leu/−His plates containing 25 mM 3-AT and incubated for 5 d at 30°C. Cell growth was only observed with full-length Hsp90 protein but not for any of its subfragments or for the vector control without insert. (C) Full-length wild-type Hsp90 and the mutants E47A, E47D, and D93N were analyzed for their ability to interact with p23 by two-hybrid assay as above. Cell growth was observed for wild-type Hsp90, the E47A and E47D mutants, but not for the D93N mutant.

Figure 6

p23 binds the ATP-bound state of purified Hsp90 in vitro, not the ADP-bound and nucleotide-free forms. Wild-type and mutant Hsp82 proteins (200 μg each) were bound to Ni-NTA agarose via their His6 tags and incubated for 45 min at 30°C with yeast p23 (100 μg) and different combinations of nucleotides and GA as indicated (lanes 2–22). In lane 1 (control) Ni-NTA beads without bound Hsp82 were used. Protein was eluted from the beads with imidazole and analyzed by 15% SDS-PAGE and immunoblotting for the T7 immunotag of p23 (see Materials and Methods).