Review: The HSP90 molecular chaperone-an enigmatic ATPase

Abstract

The HSP90 molecular chaperone is involved in the activation and cellular stabilization of a range of 'client' proteins, of which oncogenic protein kinases and nuclear steroid hormone receptors are of particular biomedical significance. Work over the last two decades has revealed a conformational cycle critical to the biological function of HSP90, coupled to an inherent ATPase activity that is regulated and manipulated by many of the co-chaperones proteins with which it collaborates. Pharmacological inhibition of HSP90 ATPase activity results in degradation of client proteins in vivo, and is a promising target for development of new cancer therapeutics. Despite this, the actual function that HSP90s conformationally-coupled ATPase activity provides in its biological role as a molecular chaperone remains obscure. © 2016 Wiley Periodicals, Inc. Biopolymers 105: 594-607, 2016.

Keywords: ATP; client protein; cochaperone; conformational change; molecular chaperones.

Figure 1

HSP90 and its clients. HSP90 engages with a range of ‘client’ protein classes (outer circle) via its interaction with various co‐chaperone proteins or complexes (inner circle), that act as adaptors, simultaneously able to interact with the specific client and the central chaperone itself.

Figure 2

ATP‐binding site. A: ATP/ADP binds into a pocket formed in the helical face of the N‐terminal domain. Cartoon shows secondary structural elements rainbow colored according to reative position within the primary structure of the domain – blue: N‐terminus, red: C‐terminus. B: Molecular surface of HSP90 N‐domain colored according to electrostatic potential blue:+ve, red:‐ve. The adenine base is buried in the binding pocket with the phosphates exposed to solvent. C: Close up of the water‐filled nucleotide‐binding pocket. Polar interactions made by the bound ADP are shown as dashed lines. D: The adenine base makes only a single direct hydrogen bond with the protein, via the side chain of Asp 79. E: The natural product geldanamycin binds to the nucleotide‐binding pocket and is a competitive inhibitor of ATP binding by HSP90. F: The natural product radicicol binds to the nucleotide‐binding pocket and is a competitive inhibitor of ATP binding by HSP90.

Figure 3

HSP90 inhibitors. Chemical structures of a number of synthetic or semisynthetic ATP‐competitive HSP90 inhibitors currently in clinical development.

Figure 4

ATPase coupled molecular clamp mechanism. A: Domain architecture of HSP90. B: ATPase activity of HSP90 C‐terminal truncation mutants. There is a dramatic loss of activity following removal of the C‐terminal dimerization domain. C: Full‐length inherently dimeric HSP90 (left) can be chemically cross‐linked (DMS) regardless of whether it has ADP or an ATP analog (AMPPNP) bound. The C‐domain deleted HSP90 is monomeric when apo or bound to ADP, but is able to dimerize in the presence of AMPPNP demonstrating the presence of a second nucleotide‐dependent dimerization site. D: Schematic of the ATPase‐coupled molecular clamp mechanism, in which ATP binding promotes N‐terminal association to form the active ‘tense’ catalytically active state, which then relaxes on hydrolysis of ATP.

Figure 5

The ATP‐bound state. A: Secondary structure cartoon of the HSP90 dimer in the ATP‐bound closed state. The two protomers (rainbow coloured blue:N → red:C) make a constitutive dimer interaction at the C‐terminus and an ATP‐dependent dimer interface at the N‐terminus. The closed state is stabilized by binding of the co‐chaperone protein P23/Sba1 (see below). B: As (A), but rotated 45º around the vertical axis. The bound AMPPNP is shown as a CPK model. C: Space‐filling representation of the closed HSP90 dimer. The two protomers wrap around each other with a left‐handed twist. D: Close up of the dimerized N‐terminus showing the topological swap of the N‐terminal strand from each protomer. The visible parts of the highly flexible ‘charged linker’ that connects the N‐ and M‐domains is indicated. E: Binding of AMPPNP promotes a conformational change in key ‘switch’ regions in the N‐terminal domain.

Figure 6

Formation of the ATPase active site. A: The conformational changes in the N‐terminal ‘switch’ regions that accompany binding of AMPPNP facilitate docking of the M‐domain and interaction of the conserved arginine residue with the γ‐phosphate of the nucleotide. B: Docking of the M‐domain is accompanied by substantial remodeling of the catalytic loop from its structure in the isolated M‐domain.

Figure 7

Conformational flexibility of HSP90. HSP90 family members display considerable conformational flexibility in the relative orientation of the C‐, M‐, and N‐domains within and between the protomers. Only in the case of the closed N‐terminally dimerized conformation seen in the crystal structures of yeast HSP90‐AMPPNP and TRAP1‐ADP/AlF₄, is there any evident interaction that distinguishes ADP from ATP. PDB codes refer to the corresponding entries in the Protein Databank (PDB).

Figure 8

ATPase‐regulatory co‐chaperones. A: The middle and C‐terminal regions of CDC37 bind to the lid segment in the N‐domain of HSP90 and prevent its closure on ATP‐binding. B: Negative stain electron microscope single particle reconstruction of an HSP90‐CDC37‐CDK4 client protein complex. A single molecule of CDC37 and of the kinase bind asymmetrically to an HSP90 dimer. C: P23/Sba1 binds across the interface between the dimerized N‐terminal domains in the ATP‐bound closed state. D: Close up of P23/Sba1 complex with HSP90. The bound co‐chaperone contacts the three main switch segments – the lid (magenta), catalytic loop (green) and N‐terminal helix/strand (blue) but only in their ATP‐bound ‘tense’ conformations. E: The N‐terminal domain of Aha1 binds to the M‐domain of HSP90 (centre), and promotes the remodeling of the catalytic loop (magenta) from its inactive conformation as in the isolated M‐domain (left) toward the extended conformation seen in the ATP‐bound state (right) in which Arg380 contacts the γ‐phosphate of the ATP. The second HSP90 in the ATP‐bound dimer has been omitted for clarity.