Targeting the molecular chaperone heat shock protein 90 (HSP90): Lessons learned and future directions

Abstract

Due to the critical role of heat shock protein 90 (HSP90) in regulating the stability, activity and intracel- lular sorting of its client proteins involved in multiple oncogenic processes, HSP90 inhibitors are prom- ising therapeutic agents for cancer treatment. In cancer cells, HSP90 client proteins play a major role in oncogenic signal transduction (i.e., mutant epidermal growth factor receptor), angiogenesis (i.e., vascular endothelial growth factor), anti-apoptosis (i.e., AKT), and metastasis (i.e., matrix metalloproteinase 2 and CD91), processes central to maintaining the cancer phenotype. Thus, HSP90 has emerged as a viable tar- get for antitumor drug development, and several HSP90 inhibitors have transitioned to clinical trials. HSP90 inhibitors include geldanamycin and its derivatives (i.e., tanespimycin, alvespimycin, IPI-504), synthetic and small molecule inhibitors (i.e., AUY922, AT13387, STA9090, MPC3100), other inhibitors of HSP90 and its isoforms (i.e., shepherdin and 50 -N-ethylcarboxamideadenosine). With more than 200 ‘‘client’’ proteins, many of them meta-stable and oncogenic, HSP90 inhibition can affect an array of tumors. Here we review the molecular structure of HSP90, structural features of HSP90 inhibition, phar- macodynamic effects and tumor responses in clinical trials of HSP90 inhibitors. We also discuss lessons learned from completed clinical trials of HSP90 inhibitors, and future directions for these promising ther- apeutic agents.

Introduction

Heat shock protein of 90 kDa (HSP90) belongs to the heat shock protein family, a functional class of chaperone molecules that are transcriptionally upregulated by heat and other stressors, and thereby, help protect cells against the damaging effects of cellular stress.1 HSP90 has been highly conserved throughout evolution, is expressed in all eukaryotic cells, and accounts for 1–2% of the total cellular protein load, increasing upon induction from baseline lev- els to 4–6%.1 HSP90 facilitates the maturation, stability, activity and intracellular sorting of more than 200 proteins, called ‘‘clients’’ or ‘‘client proteins’’1,2 (a detailed list of HSP90 client proteins is available at http://www.picard.ch/downloads/Hsp90interactors. pdf). HSP90 client proteins may be defined as proteins that bind HSP90 and whose steady-state levels decrease upon exposure to an HSP90 inhibitor.3 Client proteins of HSP90 impact an array of cellular functions that affect health and disease, including natural and acquired immunity, signal transduction, and intracellular movement of proteins.1 As a molecular chaperone, HSP90 helps nascent proteins adopt their biologically active conformations, cor- rect the conformation of misfolded proteins, and helps incorrigibly misfolded proteins to be removed and degraded by the ubiquitin– proteosome system.1

The HSP90 molecular structure has three major regions: an amino (N)-terminal domain with an adenosine triphosphate (ATP)-binding and hydrolyzing pocket (with ATPase activity) that regulates client protein folding; a middle domain involved in client protein recognition/binding; and a carboxy (C)-terminal domain which directs HSP90 dimerization.1,4 ATP is required for HSP90’s activity. The binding of ATP to HSP90 allows HSP90 to adopt its ‘‘closed’’ conformation, and enables client protein binding/loading. HSP90-bound ATP is then hydrolyzed, and the energy released by ATP hydrolysis enables client protein folding.1 ATP hydrolysis re- sults in the HSP90 dimer transitioning into its ‘‘open’’ conforma- tion and releasing the client protein. The mechanistic operation of several HSP90 inhibitors involves displacement of ATP, and thus, blockade of HSP90’s activity.1

Over 20 co-chaperones regulate HSP90 activity. Some of these inhibit HSP90 ATPase activity [such as HSP70/HSP90 organizing protein (HOP), cell division cycle protein 37 (CDC37) and p23] and others enhance it [such as activator of HSP90 ATPase 1 (AHA1) and CPR6]. In general, co-chaperones that inhibit HSP90’s ATPase activity are more likely to be involved in client loading or the formation of mature HSP90 complexes, whereas those that en- hance the activity are more likely to be activators of the HSP90 conformational cycle.1

Multiple isoforms of HSP90 exist and these include HSP90a and HSP90b in the cytoplasm and nucleus, GRP94 in the endoplasmic reticulum, and TRAP1 in the mitochondria. HSP90a is inducible and its functions include stress-induced cytoprotection and cell- cycle regulation, whereas HSP90b is constitutively expressed and is involved in early embryonic development, signal transduction, and long-term cell adaptation.5 Due to its generally higher levels than HSP90a, HSP90b is the major form of HSP90 involved in nor- mal cellular functions.

HSP90 and cancer

HSP90 has emerged as a viable target for antitumor drug devel- opment, because HSP90 is important to maintain the cancer phe- notype. HSP90 helps cancer cells overcome multiple environmental stresses, including genomic instability/aneuploidy, proteotoxic stress, increased nutrient demands, reduced oxygen levels, and the need to prevent destruction by the immune sys- tem.3 HSP90 is over-expressed in cancer cells and several of its cli- ent proteins are signaling oncoproteins that represent nodal points in multiple oncogenic signaling pathways.3,6 Signaling oncopro- teins that are also HSP90 clients include mutant cKIT, HER2, mu- tant epidermal growth factor receptor (EGFR), BCR-ABL,2 and BRAF (Fig. 1). Cancer cells that depend on these oncoproteins for survival are sensitive to HSP90 inhibition.7 Indeed, mutant or genetically altered proteins are much more sensitive to HSP90 inhibition than their wild-type counterparts.

HSP90 client proteins are also involved in other hallmark processes of cancer, including induction of angiogenesis, resistance to cell death (anti-apoptosis), and promotion of metastasis.8 For example, HSP90 influences angiogenesis by chaperoning hypoxia- inducible factor-1a (HIF-1 alpha) and vascular endothelial growth factor receptor (VEGFR) in addition to governing nitric oxide syn- thase upregulation. HSP90 chaperones client proteins that are anti-apoptotic, including AKT and survivin. Thus, inhibition of HSP90 could lead to increased apoptosis and enhanced anticancer activity.8,9 Also, HSP90 may promote metastasis through matrix metalloproteinase-2 (MMP-2) activation, digesting extracellular matrix proteins.8,9 Other client proteins of HSP90 that play a role in cell signaling processes include IL6R (JAK/STAT pathway), FAK (integrin pathway), CDK 4, 6, 9 (cell cycling), IkB kinases (NFkB pathway) and APAF-1 (apoptosis).10

The role of HSP90 in malignancy is mediated by its ability to control the stabilization of its oncogenic client proteins, as well as regulate their activated states.11 HSP90 inhibition causes deple- tion of multiple oncogenic client proteins, (for example, mutant ki- nases), leading to blockade of many key cancer causing pathways. Increased expression of HSP90 is associated with disease progres- sion in melanoma, and diminished survival in breast, lung and gastrointestinal stromal tumors (GIST).2 Examples of client pro- teins of HSP90 that play an important role in cancer include the hu- man epidermal growth factor receptor 2 (HER2), mutant epidermal growth factor receptor (EGFR), the BCR-ABL fusion protein, BRAF, the serine-threonine protein kinase AKT/PKB, c-MET, as well as the steroid hormone receptors (estrogen receptor and androgen receptor); thus, it has a putative role in numerous cancers.7,9,12,13

In the ensuing sections of this article, we review HSP90 inhibitors that have transitioned to clinical trials and their outcomes. Addi- tional details, on the antitumor activity of selected HSP90 inhibi- tors by tumor type, and on selected trials of HSP90 inhibitors as mono- or combination-therapy, are available in Tables 1 and 2, respectively.

HSP90 inhibitors

Geldanamycin and its derivatives

Geldanamycin has acted as the gateway for HSP90 inhibitor development following discovery of its destabilizing effects on some HSP90 client proteins in preclinical models.11 Although gel- danamycin’s potentially severe hepatotoxicity was not conducive to its use in the clinic, derivatives of geldanamycin became the first HSP90 inhibitors.

Tanespimycin

Tanespimycin (17-allylamino-17-demethoxygeldanamycin, 17- AAG), a geldanamycin derivative, was the first HSP90 inhibitor to be evaluated in humans. Several phase I studies explored different formulations and schedules of tanespimycin as monotherapy, with the most promising clinical result being stable disease (Table 1).14–17 Tanespimycin, by itself, has shown little activity in the phase II setting, with studies generally focusing on specific tumor types, such as BCR/ABL-positive leukemia or HER2/NEU+breast cancer.18 The lack of efficacy seen in these trials was attributed to the sub- optimal doses of tanespimycin used to avoid treatment-related toxicities.

In contrast, tanespimycin has shown activity when combined with biologic or cytotoxic agents (Table 1). In a phase II study com- bining tanespimycin and trastuzumab in 31 patients with HER2+ metastatic breast cancer previously progressing on trastuzumab, rates for overall response and clinical benefit (including com- plete/partial responses and stable disease) were 22% and 59%, respectively.6 In addition, reported medians for progression-free survival and overall survival were 6 and 17 months, respectively, and this study was the first to demonstrate Response Evaluation Criteria in Solid Tumors (RECIST) defined efficacy for tanespimycin in solid tumors.6

Tanespimycin downregulates cell surface ErbB2 expression, possibly driving the response in trastuzumab-refractory patients.19 Tanespimycin has also been tested in multiple myeloma in combi- nation with bortezomib with responses of 41%, 20%, 14% in bort- ezomib-naïve, bortezomib-pretreated, and bortezomib-refractory patients, respectively.20 These data prompted a phase III trial of tanespimycin plus bortezomib for this indication (ClinicalTri- als.gov identifier: NCT00514371).

A drawback of tanespimycin is its benzoquinone sub-group that must undergo reductive metabolism and detoxification by NADPH:quinone oxidoreductase (NQO1) (also called DT-diapho- rase) before it can act against HSP90.21 The presence of the benz- aquinone sub-group in the structure of tanespimycin may contribute to the drug’s greater hepatic toxicity, and constitute a mechanism of drug resistance in patients with mutated/altered NQO1.3 Indeed, tanespimycin’s clinical limitations are modest bio- availability, instability and toxicity.11

Alvespimycin

The second HSP90 inhibitor to enter clinical trials was alvespi- mycin (17-dimethylaminoethylamino-17-demethoxygeldanamy- cin, 17-DMAG). Alvespimycin is less sensitive to and less dependent on NQO1 and has better pharmacokinetic properties3 and fewer side-effects.11,14Alvespimycin mirrors tanespimycin’s activity, with a suggestion of greater potency. In a study of alvespi- mycin administered to 25 patients with advanced malignancies, complete response (CR) occurred in one patient with castration resistant prostate cancer (CRPC), partial response (PR) in one pa- tient with metastatic melanoma, and stable disease (SD) >6 months occurred in three patients with chondrosarcoma, CRPC and clear cell renal cancer respectively.22 However, alvespimycin, like tanespimycin, may be most beneficial in combination with cli- ent protein inhibiting drugs, such as trastuzumab. A recent phase I study of the combination showed a confirmed partial response in a HER2+ breast cancer patient and stable disease >6 months in 6/28 patients.

Fig. 1. Selected examples of HSP90 client protein pathways involved in survival and anti-apoptosis of tumor cells. JAK – Janus family of tyrosine kinases; STAT – Signal transducers and activators of transcription; MAPK/ERK – Mitogen activated protein kinase, Extracellular signal-regulated kinase; PDK – Phosphoinositide-dependent kinase.

IPI-504

Less prone to oxidative stress and more water-soluble than tan- espimycin or alvespimycin, IPI-504 (retaspimycin hydrochloride), the reduced quinone form of tanespimycin, has been evaluated in phase I and II clinical trials in chronic myelogenous leukemia (CML), multiple myeloma, refractory non-small cell lung cancer (NSCLC), and metastatic GIST.24,25 In a phase II trial of IPI-504 in refractory NSCLC stratified by ALK rearrangement status, overall response rates were 66.7% in patients with ALK rearrangements, but only 8.3% in patients without ALK rearrangements.26 Normant et al.2 showed that echinoderm microtubule associated protein- like 4 – anaplastic lymphoma kinase (EML4-ALK), an oncogenic fusion protein was highly sensitive to HSP90 inhibition by IPI- 504 in in vitro and in vivo models. IPI-504 lowered intracellular lev- els of the HSP90 client, EML4-ALK fusion protein and inhibited downstream signaling pathways, including ERK and STAT3 in in vitro models.2 In xenograft nude mouse models of EML4-ALK expressing cells and control cells, significant growth inhibition by IPI-504 was seen in EML4-ALK containing tumors, but not in con- trol tumors, suggesting that EML4-ALK is highly sensitive to HSP90 inhibition by IPI-504.2 The high degree of sensitivity of EML4-ALK to HSP90 inhibition by IPI-504 suggests that patients having cancer subtypes with EML4-ALK versus without EML4- ALK as a key oncogenic driver may be more sensitive to HSP90 inhibition.2 Indeed, in a phase 1/2 study of IPI-504 administered to patients with NSCLC, 2 of 3 patients with the EML4-ALK rear- rangement achieved PR and the third patient achieved SD.27 While the patients in this study were presumably crizotinib naive, an important question follows as to whether HSP90 inhibitors would work in crizotinib resistant patients. However, recent preclinical data have shown that EML4-ALK positive NSCLC cell lines that had acquired resistance to crizotinib still remained highly sensitive to HSP90 inhibition.28 A phase III trial of IPI-504 in GIST (NCT00688766) was spurred by early clinical trials showing partial responses in 22% and stable disease in 78% of patients with refrac- tory GIST, as measured by positron emission tomography (PET) but not by RECIST criteria, confirming the observation that inhibiting HSP90 can lead to KIT degradation, which may potentially reverse imatinib resistance in GIST patients.29 The discrepancy in the lack of response by RECIST criteria, which measure antitumor effect by tumor shrinkage and partial responses seen by PET, which measures fluorodeoxyglucose (FDG) uptake by tumor, may be reflective of the cytostatic nature of HSP90 inhibition in some con- texts. Regardless of the antitumor effect seen, due to the high mor- tality rate in the IPI-504 treatment arm, this trial was recently suspended. IPI-493, a similar derivative in oral form, is also being tested in a phase I study (Table 2).

Synthetic and small molecule inhibitors

Synthetic molecules designed to evade the toxic side effects attributed to the benzoquinone group on geldanamycin-based molecules, have been developed and show promising preclinical antitumor activity.30 XL888 is a novel small molecule inhibitor of HSP90 that has been shown to overcome resistance to BRAF inhib- itors (such as vemurafenib and debrafenib) in preclinical models by inhibiting the expression and/or functional activity of HSP90 cli- ent proteins involved in growth and cell-cycle re-entry (mutated NRAS, PDGFRb, IGF1R, cyclin D1, AKT, ARAF, CRAF, the MAP kinase family member COT) and by inducing the pro-apoptotic HSP90 client protein, BIM.31 One class of drugs includes the pyrazole or resorcinol subunit, which binds to the N-terminal ATP pocket of HSP90, blocks HSP90 ATPase activity, and decreases levels of HSP90-chaperoned oncoproteins, such as BRAF and survivin (Fig. 1).
Several other small molecule HSP90 inhibitors have recently been reported in the clinical setting including AUY922, AT13387, STA9090, and MPC3100. In a phase I study involving treatment with AUY922, extended disease stabilization in a subset of ad- vanced solid tumor patients was noted.32 Further phase II studies investigating AUY922 have shown responses as a single agent in cancers driven by mutated or overexpressed proteins such as EGFR33 or HER34 or responses in combination with other targeted agents such as trastuzumab in HER-2 amplified breast cancer.35 AT13387 in vitro has been the longest-acting HSP90 inhibitor to date, suppressing client proteins longer than 7 days. A phase I study showed an interesting toxicity of reversible visual changes of blurred vision, flashes and delayed light–dark accommodation. AT13387 showed dose proportional pharmacodyamic inhibition of cyclin-dependent kinase 4 (CDK4), RAF-1 and phospho-AKT in peripheral blood mononuclear cells (PBMCs), but no clear activity was demonstrated other than stable disease in follicular thyroid and in a uveal melanoma patient.36 A phase I evaluation of STA- 9090 was noted to be tolerable with dose-limiting toxicities of amylase elevation, fatigue and diarrhea. Preliminary signs of activ- ity in lung, renal, melanoma and GIST tumors were noted.37,38 A phase I trial of the oral agent MPC-3100 demonstrated a dose- limiting toxicity of supraventricular tachycardia in a single patient leading to further dose expansion to clarify the safe dose.

Another group of synthetic HSP90 inhibitors is based on a purine structure. Since the structure of ATP includes adenine, a purine base, HSP90 inhibitors based on a purine structure block ATP binding to HSP90, and inhibit HSP90 function. These synthetic HSP90 inhibitors include PU-H71, which demonstrated putative efficacy in triple negative breast cancer models and downregulated components of the RAS/MAPK pathway, AKT, BCL, and transcrip- tion factors, such as NF-jB, in vitro and in vivo.40 However, this molecule has yet to reach the clinic. BIIB021 is another synthetic purine agent, which binds to the ATP-binding pocket of HSP90 and putatively downregulates HER2 in vitro.41 The first completed phase I study of this molecule has shown activity in chronic lym- phocytic leukemia (CLL), possibly due to depletion of the client protein ZAP70.41 A more potent, intravenous version of BIIB028 is currently being tested in our clinic in a phase I study in advanced solid tumors (Table 2).

Although most drugs targeting HSP90 have focused on its N- terminal ATP binding site (N-terminal inhibitors), a new class of compounds that target HSP90’s C-terminal domain have been developed, and investigators have been able to increase the affinity of these molecules for HSP90. A family of coumarin antibiotics, the most notable of them being DB01051 (novobiocin), binds to the C- terminal domain of HSP90 and disrupts the HSP90 chaperone com- plex, resulting in HSP90 client protein degradation similar in action to the compounds that bind to the N-terminus, but with less deg- radative ability.42 Advantages of HSP90 inhibitors that target the C-terminal domain (C-terminal inhibitors) include less robust acti- vation of heat shock factor 1 (HSF1), which induces HSP90 tran- scription. However, concentrations of novobiocin high enough to inhibit HSP90 have not been reached in vivo, thus several other derivatives of novobiocin such as triazole containing analogues are in development.43

Other inhibitors of HSP90 and its isoforms

Several other types of HSP90 inhibitors have been developed, but most, despite early promise, have yet to reach the clinic. These include shepherdin, a novel peptidomimetic that targets HSP90’s ATP pocket44 and 50 -N-ethylcarboxamideadenosine (NECA), an adenosine analogue, which targets GRP94, the endoplasmic reticu- lum chaperone and isoform of HSP9045 and a stress marker. An- other HSP90 inhibitor is radicicol. Radicicol and its analogues are macrocyclic antibiotics derived from the Monosporium bonorden fungus and, similar to geldanamycin, act on HSP90 through N-terminal domain binding and blockade of ATP.46 However, due to poor pharmacological properties, radicicol has not been developed further. Instead, pochonins, which are a simplified pharmacophore of radicicol, have been shown preclinically to be more promising, inducing client protein degradation and tumor regression at low nanomolar concentrations with improved bioavailability.47

Perhaps the most interesting of the alternative inhibitors of HSP90 is a monoclonal antibody directed against fungal HSP90. A randomized study of amphotericin B alone vs. amphotericin B in combination with Mycograb® (NeuTec Pharma), a human recombinant antibody against HSP90, showed increased efficacy (complete overall response by day 10) in the combination arm in patients with candida infections.48

Lessons learned in oncology clinical trials and future directions for oncology drug development of HSP90 inhibitors

Early trials of HSP90 inhibitors revealed gastrointestinal toxici- ties, such as diarrhea, resulting from most agents. Geldanamycin derivatives demonstrated hepatotoxicity as their dose-limiting toxicity in early clinical trials and in most instances, this has been manageable in phase I and II clinical trials. However, later studies with larger subsets of patients have renewed concern about signif- icant liver toxicity. A randomized phase III study comparing IPI- 504 to placebo in GIST patients was terminated early following an increased rate of death in the IPI-504 arm, with 3 of 4 patients’ deaths attributable to liver failure.49

Single agent studies of HSP90 inhibitors have yielded similar side effects, such as fatigue, nausea, diarrhea, and myalgias.15,16,29 Some particular side effects of the small molecule HSP90 inhibitors have emerged, including syncope with BIIB021.50 Also, several HSP90 inhibitors, including alvespimycin23 and AUY92232 have been shown to cause retinal dysfunction. While not all HSP90 inhibitors cause this (STA9090 is an example of HSP90 inhibitor that does not cause visual side effects), visual side effects seem prevalent enough to be a potential class side effect of HSP90 inhib- itors. Possible cardiac effects (supraventricular arrhythmias) have also been seen with MPC-3100 and AUY922.32,39 These adverse events were observed in one or two patients on early phase I stud- ies and a clear association with study drugs needs to be confirmed with larger clinical trials.

HSP90 remains an actively pursued molecular target in oncology drug development. Yet, no HSP90 inhibitor has been FDA- approved to date. The ensuing sections outline strategies that may enhance therapeutic benefit and accelerate the drug approval process for safe and efficacious HSP90 inhibitors.
Personalizing treatments to match patients’ genetic profiles and targeting specific tumors/tumor types are increasingly recognized as ways to increase the effectiveness of HSP90 inhibitors. The greatest therapeutic benefits with HSP90 inhibitors could derive from targeting tumors that have as their oncogenic drivers client proteins that are particularly sensitive to HSP90 inhibition, such as HER2, ALK, EGFR or BRAF. HSP90 inhibitors may be particularly useful to treat cancers such as multiple myeloma in which HSP90 is crucial to buffering the high levels of proteotoxic stress that is characteristic of the disease.3 Workman’s group reported in a pilot study that melanoma patients with NRAS mutation-positive tu- mors were most likely to achieve prolonged stable disease with HSP90 inhibition.51 In addition, studies of HSP90 inhibitors in HER-2 positive breast cancer are examples of trials that have strat- ified patient subgroups with the greatest likelihood of response.52 Such stratification and reporting will facilitate the design of more effective phase II trials.

With the shift in cancer drug development towards targeting specific pathways, it is important to confirm that the target is inhibited at the recommended phase II dose. A host of pharmaco- dynamic biomarkers have been tested in preclinical models, such as HSP72, where induction is a marker.53,54 HSP72 is an inducible isoform of HSP70 and levels of HSP72 increase in response to HSP90 inhibition.55 Thus, induction of HSP72 together with client protein depletion is considered a molecular/pharmacodynamic sig- nature of HSP90 inhibition, and has been observed in clinical trials of patients treated with HSP90 inhibitors.22,55 In addition, it is cru- cial to study relevant client protein depletion in tumors.53 Bio- markers predictive of response are of vital importance if HSP90 inhibitors are to succeed.

It is not clear why HSP90 inhibition has shown clear clinical activity in only a handful of studies, primarily HER2+ breast cancer and EML-ALK translocated NSCLC patients. Why, given the depen- dence of client proteins on HSP90, have studies in BRAF melanoma patients, for example, not fully shown benefit as preclinical models would predict?56,57 Are the current negative trials a consequence of the inadequacy of the current HSP90 inhibitors, our lack of understanding of HSP90, or is trial design a factor? Might acquired drug resistance contribute to the duration of effectiveness of HSP90 inhibitors? Could pharmaceutical inhibition of HSP90 cause cells to rely more on oncogenic pathways and processes that are independent of HSP90? Could HSP90 inhibitors themselves be the problem? HSP90 inhibitors displace co-chaperones like p23, specifically those that target the ATP pocket. Freeman et al. has shown that p23 can disrupt transcriptional regulatory com- plexes.58 It is possible that this will lead to changes in cellular physiology independent of HSP90 binding and actually counteract HSP90 inhibition? Finally, could HSP90s role in normal physiol- ogy itself be problematic? Lindquist et al. has shown that HSP90 ‘‘buffers’’ cells from unwanted somatic alterations, and therefore activates ‘‘cryptic’’ alterations in proteins that potentially could promote, rather than block tumorigenesis (this might be especially relevant in late stage cancer cells with rampant genomic instability and a propensity for somatic changes).59 Future clinical and basic research in this area is crucial to realize the full potential of these agents.

Although there have been hints of activity with single agent HSP90 inhibitors, these limited responses should not discourage future development, given their obvious molecular functionality and potential to enhance the activity of other molecules.25 HSP90 inhibitors may be most effective in combinations, as evidenced by the combination of tanespimycin with trastuzumab since HSP90 inhibitors alone may only be cytostatic. Thus, the develop- ment of these HSP90 inhibitors will likely need close development of client protein inhibitors, e.g., RAF inhibitors. In fact, it is increas- ingly believed that combination therapies targeting parallel signal- ing pathways that regulate ‘‘hallmark’’ processes that are absolutely essential for the survival and thriving of cancer cells may provide better therapeutic outcomes in cancer.

Additionally, drug combinations including an HSP90 inhibitor may help overcome resistance to targeted inhibitors. For example, a recent study by Roue et al.60 showed that the HSP90 inhibitor, IPI-504 helps overcome resistance to the proteosome inhibitor, bortezomib in mantle cell lymphoma. Intrinsic and acquired resis- tance to bortezomib are associated with up-regulation of the pro- survival chaperone, BIP/GRP78, which forms a complex with HSP90 for stabilization. Exposure of cells to the combination of IPI504 and bortezomib resulted in dissociation of the BIP/GRP78 complex, BIP/GRP8 depletion, inhibition of the unfolded protein re- sponse and increase in mitochondrial depolarization by NOXA, a pro-apoptotic protein.

An additional strategy for combination studies in the future is to inhibit multiple heat shock proteins. Inhibition of HSP90 activates the heat shock response, which induces production of several heat shock factors, including heat shock factor 1 (HSF1), members of the HSP70 family and HSP27, which have been shown to reduce apop- tosis and sensitivity of HeLa cells to tanespimycin.62 HSP90 inhibi- tion leads, through a negative feedback loop, to activation of the heat shock transcription factor, HSF1, which causes transcriptional induction of HSP70, HSP27, and to a smaller degree, HSP90, all of which protect cancer cells from apoptosis.3 Thus, HSF1 is believed to limit the activity of HSP90 inhibitors. Accordingly, silencing HSF1, HSP70 and HSP27, has been shown to cause a marked in- crease in the sensitivity of cancer cells to HSP90 inhibition, and induction of apoptosis.3 In support of this model for the attenua- tion of cell death following inhibition of HSP90, modulating HSF1 activity has been shown to increase apoptosis in response to tanes- pimycin.63,64 Also, simultaneously reducing the expression of the HSP70 isoforms, heat-shock cognate 70 (HSC70) and HSP72, in- duces proteosome-dependent degradation of HSP90 client pro- teins, G1 cell-cycle arrest, and extensive tumor-specific apoptosis.65 Importantly, simultaneous silencing of HSP70 iso- forms in nontumorigenic cell lines does not result in comparable growth arrest or induction of apoptosis, indicating a potential ther- apeutic window.65 Also, silencing of HSP90 cochaperones, such as p23, AHA1, CDC37 (these HSP90 cochaperones are known to be ex- pressed in cancer cells) increases cancer cell sensitivity to HSP90 inhibitors.3 Thus, combinatorial targeting of HSP70 isoforms/ HSP90 cochaperones and HSP90 is a potentially attractive ap- proach. However, a caveat with combining HSP70 and HSP90 inhibitors is that the toxicity of this combination is unclear, and re- mains to be understood.

Targeting particular HSP90 isoforms and understanding post- translational modifications that influence HSP90’s function can provide additional avenues to enhance therapeutic benefit. For example, Kang et al.66 showed that a fraction of HSP90 localizes to the mitochondria as TNFR-associated protein (TRAP1) in tumor but not normal cells, and that TRAP1 inhibition could lead to selec- tive apoptosis in tumor cells. This finding could provide an avenue for therapeutic intervention in the future. Also, WEE1-induced phosphorylation of HSP90 on its tyrosine 100 residue increases HSP90’s ability to chaperone several oncogenic kinases, including HER2, SRC, CRAF, CDK4, and WEE1 itself, and decreases its ability to bind inhibitors such as geldanamycin and tanespimycin.3,67 A greater understanding of phosphorylation and other posttransla- tional modifications of HSP90 could provide additional strategies to enhance the effectiveness of HSP90 inhibitors and provide great- er understanding of novel resistance mechanisms.3

It is also possible that the schedule of HSP90 inhibitors has not been optimized. Perhaps, more frequent dosing could be explored together with associated pharmacokinetic studies. However, this strategy needs to be weighed in the context of possible benefits and the potential risk of increased toxicities. Future studies would benefit from investigating the effects of prolonged HSP90 inhibi- tion, and, importantly, ways to overcome potential negative conse- quences of HSP90 inhibition. For example, prolonged HSP90 inhibition could cause increased mutation rates in germ cells, where HSP90 is required to suppress transposon activity, through the actions of its client protein, PRMT5 an arginine methyl trans- ferase that suppresses transposon mobility.3,68 Also, HSP90 chaper- ones a few proteins with tumor suppressor activity, such as interferon regulatory factor 1 (IRF1), LATS1 and LATS2 kinases, which activate the Hippo tumor suppressor pathway, and a mutant form of retinoblastoma protein that retains approximately 50% of the tumor suppressor activity of wild-type retinoblastoma pro- tein.3 Thus, prolonged HSP90 inhibition could lead to deregulation of tumor suppressor pathways.
Clinical trial designs may ultimately be critical in determining if any clear clinical benefit is derived from HSP90 inhibitors. Ran- domized discontinuation studies that measure non-progression of disease or enrichment studies of patients based not only on his- tology but also molecular phenotype may yield better clinical results.69

Beyond cancer

The potential use of HSP90 inhibitors transcends cancer, and in- cludes treating resistant fungal infections and neurologic disor- ders. In vitro studies showed that HSP90 contributes to azole and echinocandin resistance, two classes of antifungal drugs, in the fungal pathogens Candida albicans, Aspergillus fumigatus, and terre- us.70,71 The key mediator of HSP90-dependent resistance in these pathogens is the client protein calcineurin, a protein phosphatase that regulates the stress exerted by exposure to azoles and echino- candins.70 A study combining HSP90 inhibitors and azoles rescued an invertebrate host, Galleria mellonella larvae, from lethal C. albi- cans infections and an echinocandin rescued larvae from lethal A. fumigatus infections.71 Furthermore, inhibition of C. albicans HSP90 expression enhanced the therapeutic efficacy of an azole in a mouse model of disseminated candidiasis.71

HSP90 may also contribute to the etiology of some neurodegenerative diseases such as Alzheimer’s, Parkinson’s, and Hunting- ton’s.72 Specifically, in Alzheimer’s disease, transformation is characterized by tau protein abnormalities, leading to an accumu- lation of hyperphosphorylated and aggregated tau.72 Pharmaco- logic inhibition of HSP90 significantly reduced intracellular levels of the disease-associated phosphorylated tau species pS202/T205 and pS396/S40473 and HSP90 inhibition decreased the level of ac- tive tau in transgenic tau mice.74 Similar to malignant transforma- tion, neurons undergoing a degenerative process utilize HSP90 to maintain the functional stability of aberrant proteins, whether mu- tated or overactivated, suggesting HSP90 inhibition as a therapeu- tic approach for increasing the survival of affected neurons.

The recent trial of Mycograb® showing improvement in patients with resistant candida is a first step in confronting resistant fungal infections by HSP90 inhibition.48 Studies in Alzheimer’s disease and other neurologic diseases have yet to begin, but if preclinical data predict future clinical benefit, HSP90 inhibition is a promising strategy.The pace of development of HSP90 inhibitors continues to accelerate at the bench and in the clinic. The lessons learned from the first early trials in cancer have provided insights on toxicity, efficacy and limitations of these new molecules and serve as a stepping stone forward in other disease types.