Posttranslational modification and conformational state of heat shock protein 90 differentially affect binding of chemically diverse small molecule inhibitors

Abstract

Heat shock protein 90 (Hsp90) is an essential molecular chaperone in eukaryotes that facilitates the conformational maturation and function of a diverse protein clientele, including aberrant and/or over-expressed proteins that are involved in cancer growth and survival. A role for Hsp90 in supporting the protein homeostasis of cancer cells has buoyed interest in the utility of Hsp90 inhibitors as anti-cancer drugs. Despite the fact that all clinically evaluated Hsp90 inhibitors target an identical nucleotide-binding pocket in the N domain of the chaperone, the precise determinants that affect drug binding in the cellular environment remain unclear, and it is possible that chemically distinct inhibitors may not share similar binding preferences. Here we demonstrate that two chemically unrelated Hsp90 inhibitors, the benzoquinone ansamycin geldanamycin and the purine analog PU-H71, select for overlapping but not identical subpopulations of total cellular Hsp90, even though both inhibitors bind to an amino terminal nucleotide pocket and prevent N domain dimerization. Our data also suggest that PU-H71 is able to access a broader range of N domain undimerized Hsp90 conformations than is geldanamycin and is less affected by Hsp90 phosphorylation, consistent with its broader and more potent anti-tumor activity. A more complete understanding of the impact of the cellular milieu on small molecule inhibitor binding to Hsp90 should facilitate their more effective use in the clinic.

Figure 1. GA and PU recognize overlapping but distinct Hsp90 populations which are not equally sensitive to protease cleavage

(A and B) Serial challenge of SkBr3 cell lysate (A) or recombinant Hsp90 protein (B) with GA and PU beads. (C and D) Serial challenge of SkBr3 cell lysate (C) or recombinant Hsp90 (D) with GA-beads followed by PU-beads (top) and vice versa (bottom). The GA-bindable Hsp90 population is contained within a larger PU-bindable population. (E) Cartoon of major tryptic cleavage sites in Hsp90, including a site 75 kD from the N-terminus (*) and another that is 40 kD from the N-terminus (**). (F, G) Purified recombinant Hsp90 (F) or Hsp90 from SkBr3 cell lysate (G) were captured by GA- or PU-beads. Samples were digested with increasing concentrations of trypsin, separated by SDS-PAGE, and subjected to Western blotting with Hsp90 antibodies whose epitopes are either in the N-terminal (N) or C-terminal (C) domains

Figure 2. N-terminal Hsp90 inhibitors geldanamycin (GA) and PU-H71 (PU) interact with different sub-populations of N domain undimerized Hsp90

(A-C) Hsp90 was isolated from tumor cell lysates using drug-conjugated agarose beads. Hsp90 levels before (input, in) and after (final, f) chemical precipitation were monitored by Western blot. PU, but not GA, preferentially associates with Hsp90 bound to the co-chaperone p50^Cdc37 (A), and the oncoprotein kinase clients ErbB2 (B) and v-Src (C). (D and E) Pre-incubation of intact cells with the chemically distinct N-terminal inhibitor radicicol prior to PU-bead precipitation blocked Hsp90 and oncoprotein binding, confirming specificity of the PU-bead pulldowns. (F-I) N domains of GA- and PU-bound Hsp90 are in close proximity but are not dimerized. (F) N domain dimerization of yeast Hsp90 determined by DMS cross-linking in the presence of AMPPNP, ADP, GA (Geld) and PU (PU-H71). (G) Average densitometry scans of (F) were normalized against total intensity (AMPPNP, blue tracing; ADP, red tracing; GA, green tracing; PU, purple tracing). The peak representing the slowest migrating band (identified by the black arrow) indicates the degree of N-terminally dimerized Hsp90 protein. (H) Neither GA nor PU alone induce significant conformational changes that would correspond to N-terminal dimerization as assayed by FRET. Time-dependent FRET induced by addition of either 2 mM ATPγS (black tracing), 50 μM GA (green tracing), or 50 μM PU (red tracing) to 400 nM Hsp90 is shown. (I) Fluorescence anisotropy detection of p23 binding to yHsp90. Fluorescein-labeled p23 was excited at 490 nM and emission was recorded at 520 nM. p23 binding to Hsp90, which requires N domain dimerization, is only detectable after addition of 2 mM AMP-PNP (black tracing), but not after addition of 10 μM GA (green tracing), 10 μM PU (red tracing), or 100 μM PU (blue tracing).

Figure 3. Both GA and PU can access Hsp90 whose N domains are in close proximity but are not dimerized

(A) Model of the proposed effect on Hsp90 conformation of enforced N domain proximity achieved by adding coiled-coil domains to the Hsp90 N-terminus. (B) Coiled-coil yeast Hsp90 (coilNMC) has a similar affinity for GA as wild-type Hsp90 (NMC), as determined by BODIPY-GA displacement. (C, Top) Coiled-coil yeast Hsp90 (yCoil) robustly binds both GA and PU, and allows for client capture with GA-beads. (Bottom) Coiled-coil domains strengthen yHsp90 client (Slt2/MPK1) binding.

Figure 4. Post-translational modification of Hsp90 governs drug specificity

(A) Endogenous Hsp90 from 293T cells was isolated with GA or PU beads and Western blotted with antibody to pan-phospho-tyrosine. (B) Endogenous Hsp90 was subjected to GA and PU beads as in (A) and probed with an antibody specific to phospho-tyrosine 197 (Hsp90α). (C) In vitro v-Src-dependent tyrosine phosphorylation of Hsp90 bound to either GA or PU beads. (D, E) Phosphorylation can influence both the binding and isoform specificity of Hsp90 inhibitors. Flag-tagged wild-type and phospho-mimetic mutant Hsp90α and Hsp90β in cell lysates were subjected to GA and PU beads. Phosphorylation of Y197 (Hsp90α) or Y192 (Hsp90β) reduced GA binding without affecting PU binding (D). Phosphorylation of Y301 affected GA binding to Hsp90β only; binding of phospho-Y301-Hsp90β to PU, or phospho-Y309-Hsp90α to either GA or PU was unchanged compared to wild-type (E).