Stensgard, Bridget

Abstract Despite studies that show the antitumor activity of Hsp90 inhibitors, such as geldanamycin (GA) and its derivative 17-allylamino-demethoxygeldanamycin (17-AAG), recent reports indicate that these inhibitors lack significant single-agent clinical activity. Resistance to Hsp90 inhibitors has been previously linked to expression of P-glycoprotein (P-gp) and the multidrug resistant (MDR) phenotype. However, the stress response induced by GA treatment can also cause resistance to Hsp90-targeted therapy. Therefore, we chose to further investigate the relative importance of P-gp and the stress response in 17-AAG resistance. Colony-forming assays revealed that high expression of P-gp could increase the 17-AAG IC50 6-fold in cells transfected with P-gp compared with parent cells. A549 cells selected for resistance to GA overexpressed P-gp, but verapamil did not reverse the resistance. These cells also overexpressed Hsp27, and Hsp70 was induced with 17-AAG treatment. When the GA and 17-AAG resistant cells were transfected with Hsp27 and/or Hsp70 small interfering RNA (siRNA), the 17-AAG IC50 decreased 10-fold compared with control transfected cells. Transfection with siRNA directed against Hsp27, Hsp70, or Hsp27 and Hsp70 also increased sensitivity to EC78, a purine scaffold-based Hsp90 inhibitor that is not a P-gp substrate. We conclude that P-gp may contribute, in part, to resistance to 17-AAG, but induction of stress response proteins, such as Hsp27 and Hsp70, by Hsp90-targeted therapy plays a larger role. Taken together, our results indicate that targeting of Hsp27 and Hsp70 should be exploited to increase the clinical efficacy of Hsp90-directed therapy. [Cancer Res 2008;68(18):7419–27]

Purpose We determined the maximum-tolerated dose (MTD) and the dose-limiting toxicities (DLT) of 17-allylamino-17-demethoxygeldanamycin (17-AAG) when infused on days 1, 8, and 15 of a 28-day cycle in advanced solid tumor patients. We also characterized the pharmacokinetics of 17-AAG, its effect on chaperone and client proteins, and whether cytochrome P450 (CYP) 3A5 and NAD(P)H:quinone oxidoreductase 1 (NQO1) polymorphisms affected 17-AAG disposition or toxicity. Patients and Methods An accelerated titration design was used. Biomarkers were measured in peripheral-blood mononuclear cells (PBMCs) at baseline and on days 1 and 15, and pharmacokinetic analysis was performed on day 1 of cycle 1. CYP3A5*3 and NQO1*2 genotypes were determined and correlated with pharmacokinetics and toxicity. Results Twenty-one patients received 52 courses at 11 dose levels. DLTs at 431 mg/m2 were grade 3 bilirubin (n = 1), AST (n = 1), anemia (n = 1), nausea (n = 1), vomiting (n = 1), and myalgias (n = 1). No tumor responses were seen. 17-AAG consistently increased heat shock protein (Hsp) 70 levels in PBMCs. At the MTD, the clearance and half-life (t1/2) of 17-AAG were 11.6 L/h/m2 and 4.15 hours, respectively; whereas the active metabolite 17-aminogeldanamycin had a t1/2 of 7.63 hours. The CYP3A5*3 and NQO1*2 polymorphisms were not associated with 17-AAG toxicity. The CYP3A5*3 polymorphism was associated with higher 17-AAG clearance. Conclusion The MTD of weekly 17-AAG is 308 mg/m2. 17-AAG induced Hsp70 in PBMCs, indicating that Hsp90 has been affected. Further evaluation of 17-AAG is ongoing using a twice-weekly regimen, and this schedule of 17-AAG is being tested in combination with chemotherapy.

Heat shock protein (hsp)90 functions in a complex chaperoning pathway where its activity is modulated by ATP and by interaction with several co-chaperones. One co-chaperone, p23, binds selectively to the ATP-bound state of hsp90. However, the isolated ATP-binding domain of hsp90 does not bind p23. In an effort to identify the p23-binding domain, we have constructed a series of hsp90 deletion mutants fused with glutathione- S -transferase (GST). Full-length GST-hsp90 is able to bind p23, and also, to chaperone assembly of progesterone receptor complexes. Truncations from the C terminus of GST-hsp90 reveal a C-terminal boundary for the p23-binding domain at approximately residue 490. This fragment contains, in order, the ATP-binding domain, a highly charged region, and 203 residues beyond the charged region. p23 binding is unaffected by deletion of the charged region, indicating that two noncontiguous regions of hsp90 are involved in p23 binding. These regions are only effective when hsp90 is in a dimeric state as shown by loss of p23 binding upon removal of GST or as shown by use of FK506-binding protein12-hsp90 constructs that form dimers and bind p23 only in the presence of a bivalent drug. Thus, p23 binding requires an hsp90 dimer with close proximity between N-terminal regions of hsp90 and a conformation specified by ATP.