Heat-Shock Protein 90 Inhibitors: Will They Ever Succeed as Chemotherapeutics?

Keywords::

• C-terminal
• heat-shock protein
• heat-shock protein 90 chaperones
• heat-shock response
• Hsp90 inhibitors
• Hsp90 modulators
• stress

Normal cells produce relatively high levels of heat-shock proteins (HSPs), where HSPs account for 1–2% of all proteins in the cell. These HSPs are critical for facilitating protein folding, prevent protein aggregation and assist in cell signaling pathways [1]. HSPs, specifically Hsp90, Hsp70, Hsp40 and Hsp27 participate in a protein-folding cycle and protein-degradation pathways that enable the cell to function under high stress conditions. Rapid reproduction and metastasis require that tumor cells have high levels of these HSPs, where the percentage of HSPs among total protein levels reach up to 5%, depending on the tumor type [2].

Over the past 10 years, 15 different Hsp90 inhibitors have been tested in clinical trials [3–5]. However, only three remain active as mono-chemotherapy regiments [6]. All clinical Hsp90 inhibitors (termed ‘classic inhibitors’) target the Hsp90's N-terminus, and block the binding event between ATP and Hsp90. Inhibiting this ATP–Hsp90 interaction produces a heat-shock response (HSR) [7]. The HSR has been essentially defined as the high-level production of Hsp70, and this production has been used to indicate whether Hsp90 is functionally inhibited. Such Hsp70 induction is thought to augment cellular protein-folding events, and to play an antiapoptotic role, thereby protecting cancer cells from Hsp90 inhibition-induced cell death. The high levels of induced Hsp70 along with other HSR indicators seen when using these classic inhibitors may be responsible for the failure of these clinical trials. Identification of new Hsp90 inhibitors that control it's C-terminus do not produce increased levels of Hsp70 nor do they show other HSR indicators [8–11], therefore instigating a discussion on the role and mechanism by which Hsp70 is upregulated by classical Hsp90 inhibitors.

In this editorial, we propose the following: inhibition of Hsp90 does not automatically trigger an HSR; induction of Hsp70 is not a suitable biomarker for evaluating the effectiveness of Hsp90 inhibition; classic Hsp90 inhibitors maybe nonselective and target other proteins; and molecules modulating the C-terminus of Hsp90 may provide optimal Hsp90 inhibition. These proposals are based on the following published facts: first, classical Hsp90 inhibitors kill cancer cells with GI₅₀ of 50 nM, but they bind to or impact Hsp90 function with IC₅₀ = ~1–2 µM, depending on the assay performed (isothermal calorimetry, surface plasmon resonance, protein binding, protein refolding) [12]. Second, molecules that modulate the C-terminus of Hsp90 typically kill cancer cells at 500 nM to 5 µM, and bind to Hsp90 at similar concentrations as their GI₅₀ values [12–14]. That is, C-terminal modulators have cytotoxicity values that correlate to their biochemical binding affinity. Third, two classic inhibitors, 17-AAG and AUY922, have been compared using cancer tissues, and they have produced clinically distinct phenotypes, suggesting that even classical inhibitors do not function via the same mechanism [15].

Why have so many classical Hsp90 inhibitors failed in the clinic?

Initial results discussing early classical inhibitors, such as Geldanamycin and 17-AAG, demonstrated that they were highly effective at killing cancer cells [16]. Furthermore, the more stable analog, 17-AAG, initially performed well in clinical trials [17]. However, it was discontinued because of its hepatotoxicity [18]. Several reasons appear to be responsible for 17-AAG's failure at clinical trials. First, 17-AAG is also known to target other ATP-binding sites including DNA polymerases, tyrosine kinases and multiple isoforms of mammalian Hsp90 protein (references within [19]). Second, the quinone has toxic properties. Third, perhaps the induced HSR is problematic when aiming to kill tumor cells, as it is a cytoprotective response that triggers many antiapoptotic pathways [8].

Alternatives to 17-AAG have been developed [19], and several promising molecules, including AUY922 [5] and Ganetspib (STA-9090) [20], are in current clinical trials as monotherapies. However, there are still concerns about the HSR triggered by blocking the ATP–Hsp90 interaction, as it provides tumor cells an opportunity to survive. Furthermore, the significant difference (~100-fold) between the efficiency of these molecules in killing cancer cells and their ability to impact Hsp90's function is highly problematic. One argument states that classical Hsp90 inhibitors are more selective for Hsp90 derived from tumor cells [21], than from biochemical Hsp90 used in binding assays. That is, the conformation of Hsp90 protein used when measuring the drug–Hsp90 binding affinity (regardless of the assay type) is different from the Hsp90 protein in tumor cells. However, data supporting the initial study have not been published. It is possible that Hsp90 is in a highly activated state in cancer cells and is not in normal cells, however the concentrations of Hsp90 are higher in cancer cells than normal cells [2], and Hsp90 is more critical for the function of cancer cells than normal cells. How do the facts connect to the results from this one published report?

Are other factors at play with classical inhibitors?

The difference between the cell-based GI₅₀ values of those inhibitors versus their IC₅₀ values against Hsp90 binding is a concern for any drug development program, and there is no successful drug precedent where the killing efficacy is 100-fold higher than the biochemical binding affinity. Usually the affinity is higher for the biochemical target as nonspecific binding events do not occur. A common explanation for this discrepancy is that these types of ‘Hsp90 inhibitors’ are hitting multiple targets. Supporting this statement are the high IC₅₀ values reported using numerous types of assays [8,11], which measure the Hsp90 activity in cancer cell lysates. The Hsp90 protein would presumably have the same conformation in cancer cell lysates as in tumor tissues yet their IC₅₀ values remain high. Thus, the argument that the molecules are more effective at targeting Hsp90 in cancer cells appears to be without merit [21].

Supporting the concerns outlined above is the work published by Butler et al. [15], which showed that it was inaccurate to use Hsp70 as a biomarker to demonstrate Hsp90 inhibition. Butler demonstrated that only AUY922 and not 17-AAG showed antiproliferative and proapoptotic activity in human prostate tissue, despite both molecules acting similarly in eternal cancer cell lines. Furthermore, both molecules induced equivalent levels of Hsp70 protein in the human prostate tumor, but only AUY922 inhibited tumor proliferation and induced apoptosis. These data not only indicate that Hsp70 is not an effective biomarker for showing that Hsp90's antiproliferative and proapoptotic effects are suppressed, but also that 17-AAG's ineffectiveness in clinical trials maybe related to its inability to effectively control Hsp90 and its signaling pathways in clinically relevant tissue. However, AUY922 and other recent Hsp90 inhibitors relating to this class of molecules, still exhibit the same problematic properties as 17-AAG. Specifically AUY922 induces cell death at low nanomolar concentrations but binds to Hsp90 and modulates its protein-folding function with micromolar affinity. Thus, N-terminal Hsp90 inhibitors appear to be nonselective, with each inhibitor producing a unique phenotype that involves regulating the HSR pathway and/or other apoptotic inducing pathways through multiple unknown mechanisms.

How do C-terminal modulators compare with classical inhibitors?

Investigation of C-terminal inhibitors has been accomplished primarily from two groups: Blagg [11] and McAlpine [8–10]. Together their work has produced a significant volume of data showing that C-terminal modulators do not trigger a HSR. Furthermore, Blagg's (termed ‘KU inhibitors’) and McAlpine's molecules (termed ‘SM inhibitors’) show similar binding affinity for Hsp90 relative to their ability to kill cancer cells. Recent work defining how the mechanism differs between the classical versus SM inhibitors shows that there are numerous differences between these two classes [8]. Specifically, 17-AAG results in a 70-fold increase in mRNA encoding for inducible Hsp70, whereas SM122 does not. In fact, SM122 significantly decreases the expression of mRNAs that encode for Hsp70 and Hsp27 over a 24-h period, two heat-shock proteins that are critical in the HSR.

Comparison of HSP expression levels in the cancer cells treated with either KU inhibitors, SM inhibitors or 17-AAG have shown that both KU [11] and SM [8–10] decrease Hsp70 and Hsp27 protein expression in cancer cells. However, 17-AAG causes a fivefold increase of both proteins after treatment. Taken together, these data support the following statements: first, the C-terminal Hsp90 inhibitors are likely selective for Hsp90 given that their cell killing effects correlate to their binding affinity for Hsp90. Second, molecules that modulate the C-terminus do not impact mRNA or protein expression levels of Hsp70 and Hsp27, which indicate that they do not induce an HSR. Thus, C-terminal modulators may be the optimal approach for inhibiting Hsp90, as they show selectivity for Hsp90 rather than impacting the entire HSR pathway.

In summary, there is a need to widen the view on how classical Hsp90 inhibitors function. Investigation on whether they target alternative pathways, and specifically how they impact the HSR pathway is critical. Targeting multiple components in the HSR pathway would explain why these classical inhibitors kill cancer cells with such extraordinary potency, induce biomarkers associated with heat-shock, but do not regulate Hsp90 with high affinity. Investigation of alternative mechanistic pathways that are activated by these classical inhibitors would provide critically important answers. Detailed mechanistic analysis focusing on the classical inhibitors selectivity for tumor-associated Hsp90 versus normal tissue is also essential and would provide a route for potentially improving their clinical impact. Finally, development of potent and highly selective C-terminal modulators of Hsp90 may ultimately be the most promising route for new clinical inhibitors.

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending or royalties.

No writing assistance was utilized in the production of this manuscript.

Additional information

Funding

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending or royalties.