Development and application of Hsp90 inhibitors

Heat shock protein 90 has emerged as an important target in several diseases. The present review will discuss our understanding of the role played by Hsp90 in regulating and maintaining the transformed phenotype in cancers and neurodegenerative diseases, as well as recent findings on its roles in fungal and viral infections. It will also update the reader on the preclinical development and clinical translation of Hsp90 inhibitors.

Introduction

As we have currently come to understand, heat shock protein 90 (Hsp90) is a key component of a multichaperone complex with important roles in the development and progression of pathogenic cellular transformation 1, 2, 3, 4. Its wide-ranging functions result from the ability of Hsp90 to chaperone several ‘client proteins’ that play a central pathogenic role in human diseases including cancer, neurodegenerative diseases, and viral infections. At the time of its discovery, however, it was difficult to imagine Hsp90 as a potential therapeutic target. The chaperone is highly abundant in most tissues, comprising over 1% of the total cellular content even in the absence of stress. Further, genetic knockout of Hsp90 is lethal in eukaryotes and over 100 kinases and transcription factors, many of which are crucial to normal cellular growth and survival, have been shown to interact functionally with Hsp90. The initial interpretation of these findings was that a therapeutic window would be difficult, if not impossible, to achieve with Hsp90 inhibitors. This unfavorable view regarding the potential of Hsp90 as a therapeutic target has changed only after the discovery of geldanamycin (GM) (Figure 1). GM was found as part of a phenotypic screen focused on the identification of agents capable of inducing the morphologic reversion of $v$-src-transformed 3T3 cells [5]. It was initially hypothesized that GM was a direct inhibitor of $v$-SRC activity. Whitesell and colleagues later clarified that GM's biologic activity was actually attributable to its ability to bind to and inhibit Hsp90 [6]. Co-crystallization studies demonstrated that GM binds to an ATP/ADP pocket in the N-terminal domain of Hsp90 and that it inhibits Hsp90 function by preventing full chaperone cycling. In cells exposed to GM, unprocessed chaperone-client protein complexes accumulate within the cell leading to recruitment of E3 ubiquitin ligases that target Hsp90 clients for degradation in the proteasome 7, 8.

Much of our understanding of the role of Hsp90 in promoting malignant transformation has been derived from studies using GM and other structurally diverse Hsp90 inhibitors as biologic probes [9]. A surprising observation derived with these agents was that cancer cells were significantly more sensitive to Hsp90 inhibition than non-transformed cells 1, 2, 3, 4, 5, 6, 9. This finding raised the possibility that Hsp90 inhibitors may possess an exploitable therapeutic index and prompted the testing of this strategy using xenograft and transgenic cancer models. These studies, which have used primarily 17-allylamino-17-demethoxy-geldanamycin (17AAG) (Figure 1), a GM derivative with a more favorable profile, but also synthetic small molecule Hsp90 inhibitors, clearly demonstrate that Hsp90 inhibitors possess potent anti-cancer activity at non-toxic doses 10, 11, 12.

Several models have been proposed to explain the selective sensitivity of cancer cells to Hsp90 inhibition 13, 14, 15, 16. One model, akin to the ‘oncogene addiction’ model proposed by Weinstein [13], focuses on the exclusive dependence of some transformed cells on a sensitive Hsp90 client protein. In this model, degradation of a specific Hsp90 client in the appropriate genetic context, for example, BRAF in a melanoma cell with V600E mutant BRAF or Bcr-Abl in chronic myeloid leukemia (CML) 1, 2, 3, 4, results in apoptosis and/or differentiation, whereas its degradation in normal cells leads to little to no effect. As Hsp90 interacts with a large number of oncogenic kinases and transcription factors, this model has been used to justify the clinical development of Hsp90 inhibitors in a broad range of tumor types (Table 1). The importance of this model to the clinical development of Hsp90 inhibitors is that it highlights the relevance of patient selection in the design of Hsp90 inhibitor trials. Unfortunately, only limited animal studies have been performed with these inhibitors to date and, therefore, the ability of 17AAG or other Hsp90 inhibitors to degrade most putative Hsp90 clients in vivo has yet to be explored in relevant animal model systems.

A complementary ‘Hsp90 addiction model’ is based upon the hypothesis that Hsp90 is limiting in many tumor cells because of their increased load of mutated and unfolded proteins 14, 15. An increased requirement for Hsp90 chaperone function in tumors cells may be due to the overexpression of mutated Hsp90 clients or amplification of clients such as HER2. The hypoxic, low pH, and low nutrient conditions found in many tumors may further increase the number of denatured proteins found in tumors, and thus the need for Hsp90 chaperone function. In support of this hypothesis, Kamal et al. have shown that Hsp90 in tumor cells is found entirely in an active complex with co-chaperones, whereas most Hsp90 in normal tissues resides in a free, uncomplexed, or latent state [16]. These data suggest that the sensitivity of some cancer cells to Hsp90 inhibition may not be attributable to their reliance on any individual Hsp90 client or clients, but rather to their increased dependence upon Hsp90 chaperone function. This may explain why therapies that further stress the chaperone system such as anti-angiogenic agents, proteasome inhibitors, cytotoxic chemotherapies, and radiation may broadly synergize with Hsp90 inhibitors even in tumors in which Hsp90 inhibitors alone have little or no activity 1, 2, 3.

Finally, the selective sensitivity of transformed cells for Hsp90 inhibitors may be partly due to the selective accumulation of these drugs in cancer cells [15]. The selective accumulation of Hsp90 inhibitors in tumors has been reported with compounds that display little structural homology and diverse solubility profiles. One possible explanation for this result is that the binding affinity of these compounds may be higher for tumor-derived Hsp90 than recombinant Hsp90 or Hsp90 derived from non-transformed cells because of the higher abundance of activated, co-chaperone bound Hsp90 found in tumors 15, 16.