Keywords

1 Introduction

Heat shock proteins (HSP) are molecular chaperones that are curial for the maturation, stability and function of other proteins named “client proteins”. Hsp include large Hsp, HSP90, HSP70, HSP60, HSP40 and small HSP. Among which, HSP90 is the most well studied one in the last decades (Garcia-Carbonero et al. 2013). As the most abundant intracellular protein in mammalian cells, HSP90 comprises approximately 1–2% of total cell protein under normal conditions. While under stressed conditions (heat, hypoxia and nutrient deprivation), HSP90 is upregulated to maintain the cellular homeostasis (Barrott and Haystead 2013). There are four HSP90 isoforms in higher eukaryotes, cytosolic HSP90α and HSP90β, endoplasmic reticulum (ER) paralog glucose-regulated protein 94 (GRP94) and mitochondria paralog tumor necrosis factor receptor-associated protein-1 (TRAP1). Despite the high homologies, the four isoforms possess different cellular functions and have different client profiles (Gewirth 2016). In mammalian cells, HSP90 exists as a homodimer, and each monomer is consisted of three domains: N-terminal domain (NTD), middle-domain (MD), and C-terminal domain (CTD). The NTD incorporates an ATP binding site, which hydrolyzes ATP to provide energy for the chaperone function. The MD provides binding sites for clients and cochaperones. The CTD is important for the dimerization. Recent studies suggest that another ATP binding pocket is located at CTD (Miyata et al. 2013). The chaperone function of HSP90 relies on an ATP triggered chaperone cycle. ATP binding leads to transient dimerization of the NTD, then several cochaperones (CDC37, HSP70, P23, HOP, etc.) form a chaperone complex with HSP90 to regulate the folding and release of the client proteins (Calderwood 2013). HSP90 plays important roles in the development of cancer, and HSP90 inhibition is proved to be a promising anti-cancer strategy. To data, HSP90 inhibitors with different mechanisms have been developed, including HSP90 N-terminal inhibitors, C-terminal inhibitors and inhibitors blocking the PPI between HSP90 and cochaperones. These inhibitors block multiple cancer dependent signal pathways and exhibit potent pre-clinical anti-cancer effects both in vitro and in vivo. Moreover, several HSP90 N-terminal inhibitors have entered clinical trails. In this chapter, we will review the development progress of HSP90 inhibitors as anti-cancer agents.

2 HSP90 and Cancer

HSP90 is reported to be over-expressed in tumors two to threefold higher than corresponding non-tumorigenic tissue (Barrott and Haystead 2013). Cancer cells are addictive to the HSP90 chaperone function to prevent the mutated and overexpressed oncoproteins from degradation (Trepel et al. 2010). Among the over 500 client proteins of HSP90 (https://www.picard.ch/downloads/Hsp90interactors.pdf.), many are notorious oncogenic proteins, including tyrosine-kinase receptors (EGFR, HER2, SRC, MEK, etc.), signal-transduction proteins (BCR-ABL, ALK, BRAF, AKT, etc.), transcription factors (AR, OR, HIF1α, P53, etc.), cyclin-dependent proteins (CDK4, CDK6, Cyclin D, etc.), antiapoptotic proteins (BCL2, Survivin), matrix metalloproteinases (MMP-2, MMP-9) and telomerase (hTERT) (Garcia-Carbonero et al. 2013; Verma et al. 2016). These oncogenic client proteins are involved in multiple cancer dependent signaling pathways.

GRP94 is the ER paralog of HSP90. Clinically, high expression of GRP94 is associated with the advanced stage and poor prognosis of some types of carcinomas, including hepatocellular carcinoma, multiple myeloma, gallbladder cancer, et al. (Chen et al. 2015; Chhabra et al. 2015; Rachidi et al. 2015). GRP94 has a limited client protein profile, and some client proteins correlate with the development and metastasis of cancer. Integrins are a major class of GRP94 client proteins. They mediate cell–cell, cell–extracellular matrix communication and are important for cell adhesion and metastasis (Desgrosellier and Cheresh 2010). GRP94 based α-helix peptide blocking the interaction between GRP94 and integrins inhibited cell invasion (Hong et al. 2013). In addition, GRP94 knockdown led to reduce migration and metastasis potential of MDA-MB-231 cells and ROS resistant MCF-7 cells (Dejeans et al. 2012). The maturation and secretion of Insulin-like growth factors (IGFs) are also dependent on GRP94 (Barton et al. 2012; Marzec et al. 2016; Ostrovsky et al. 2009; Wanderling et al. 2007). Both IGF1 and IGF2 play important roles in cancer, and IGF pathway inhibition has been considered as a novel anti-cancer strategy (Brahmkhatri et al. 2015; King and Wong 2016). Bei Liu and coworkers discovered that GRP94 bound and promoted the exportation of Wnt coreceptor LRP6 from ER to cell surface (Liu et al. 2013). Their further study demonstrated that GRP94 deficiency blocked Wnt-LRP-Survivin pathway in human multiple myeloma cells, and induced cell apoptosis and tumor growth inhibition in xenograft models (Hua et al. 2013). Recent studies showed that surface GRP94 interacted with HER2 and facilitated HER2 dimerization. Purine-scaffold GRP94 specific inhibitor PU-WS13, GRP94 based α-helix peptide or anti-GRP94 monoclonal antibody treatment led to decreased HER2 level in SKBr3 cells and exhibited selective therapeutic effects in HER2 positive breast cancers (Li et al. 2015a, b; Patel et al. 2013, 2015).

TRAP1, the mitochondrial paralogue of HSP90, is also reported to be elevated in various human carcinomas, including colorectal carcinoma, prostate cancer, ovarian carcinoma, breast cancer, non-small cell lung cancer (NSCLC) and esophageal squamous cell carcinoma (ESCC). The high TRAP1 expression is associated with the progress and prognostic of these carcinomas (Agorreta et al. 2014; Costantino et al. 2009; Landriscina et al. 2010; Leav et al. 2010; Maddalena et al. 2013; Tian et al. 2014). TRAP1 is involved in the proliferation and apoptosis of cancer cells. In ESCC cells, knockdown of TRAP1 inhibited the proliferation, induced cell cycle arrest and apoptosis. While re-expression of TRAP1 promoted cell proliferation and reduced cell apoptosis (Tian et al. 2014). TRAP1 also promoted neoplastic growth by binding and inhibiting succinate dehydrogenase (SDH). TRAP1 dependent SDH inhibition led to respiratory down-regulation and succinate accumulation. Succinate then stabilized HIF-1α and prompted neoplastic growth (Sciacovelli et al. 2013). Studies in Matteo Landriscina group suggested that TRAP1 was involved in drug resistance of cancer cells. TRAP1 up-regulation was observed in HT-29 colorectal carcinoma cells resistant to 5-fluorouracil (FU), oxaliplatin (l-OHP) and irinotecan (IRI), and was responsible for the multi-drug resistance (Costantino et al. 2009). As the mitochondrial paralogue of HSP90, TRAP1 was also demonstrated locating in ER and involving in ER stress protection and protein quality control of tumor cells. The ER-associated TRAP1 protected MCF7 and MBA-MB-231 cells from paclitaxel-induced apoptosis and was relevant in paclitaxel resistance of these breast carcinoma cells (Maddalena et al. 2013). ER-associated TRAP1 was also responsible for the anthracyclins resistance of breast carcinoma cells (Sisinni et al. 2014). In general, HSP90 isoforms are involved in multiple cancer dependent signaling pathways. HSP90 inhibition is a promising anti-cancer strategy.

3 HSP90 N-Terminal Inhibitors

Inhibitors competitively occupy the NTD ATP binding pocket can block the HSP90 chaperone function. These inhibitors are named HSP90 N-terminal inhibitors.

3.1 Pan-HSP90 N-Terminal Inhibitors

Up to now, more than 20 distinct N-terminal HSP90 inhibitors have entered clinical trails. These are pan-HSP90 inhibitors that exhibit similar affinities to the four isoforms. As shown in Fig. 20.1, these inhibitors can be divided into three classes. Geldanamycin analogues are the first-generation inhibitors, including 17-AAG, 17-DMAG, IPI-493 and IPI-504. To solve the unfavorable drug-like properties of the geldanamycin analogues, the second-generation synthetic small molecule inhibitors are developed, including purine derivatives, resorcinol derivatives and benzamide derivatives. The pre-clinical properties and clinical performance of these inhibitors have been thoroughly reviewed. Despite the favorable pre-clinical anti-tumor activity, all these inhibitors exhibit poor clinical efficacy in mono-therapy. Some inhibitors only show moderate anti-cancer effects in specific cancer types when combined with other anti-cancer agents. In 2015, the highly anticipated phase III study of STA-9090 combined with docetaxel for the treatment of advanced stage lung adenocarcinoma was failed. The STA-9090 and docetaxel combination regimen did not result in improved efficacy compared with docetaxel mono-therapy. In clinical trials, these inhibitors exhibited many treatment-related adverse events (TEAEs). The most common side effects were fatigue, visual disorders, gastrointestinal disorders (nausea, vomiting and diarrhoea), anaemia and neutropenia (Bhat et al. 2014b; Biamonte et al. 2010; Garcia-Carbonero et al. 2013). Because of the disappointing clinical performance, the clinical studies of most inhibitors are terminated. Now, only the resorcinol derivative AT-13387 and the benzamide derivative XL888 are active in clinical trials (Fig. 20.2).

Fig. 20.1
figure 1

Structures of representative pan-HSP90 N-terminal inhibitors

Full size image
Fig. 20.2
figure 2

Structures of HSP90α/β selective inhibitors

Full size image

The disappointing clinical performance of these pan-HSP90 N-terminal inhibitors can be partly attributed to the following reasons. (1) The clinical dosages of these inhibitors are limited because of the various adverse events, which lead to the drug exposure levels in patients are insufficient to effectively inhibit tumor cell HSP90. HSP90 four isoforms have different functions and chaperone different client proteins. Pan-HSP90 N-terminal inhibitors inhibit the four isoforms simultaneously, which may lead to mechanism related adverse events. Thus, isoform selective HSP90 N-terminal inhibitors may reduce the adverse events and improve the clinical performance. (2) The pan-HSP90 inhibitors trigger significant heat shock response. The increased cytoprotective HSP (HSP70 and HSP27) partly neutralize the anti-cancer effects. Studies have shown that GRP94 selective inhibitors, TRAP1 selective inhibitors and HSP90 C-terminal inhibitors don’t induce heat shock response. These inhibitors may have more promising usage as anti-cancer agents.

3.2 Isoform Selective HSP90 N-Terminal Inhibitors

HSP90 isoform selective inhibitors only influence partial isoform specific client proteins. Theoretically, isoform selective inhibitors will have reduced adverse effects compared with the pan-inhibitors. In addition, GRP94 selective inhibitors and TRAP1 selective inhibitors don’t induce heat shock response. Though isoform selective inhibition is attractive, it used to be considered very difficult. HSP90 four isoforms are highly homologous. HSP90α possess 93%, 59%, and 47% sequence similarity with HSP90β, GRP94, and TRAP1, respectively. Even worse, the residues in ATP-binding pocket are more heavily conserved (Ernst et al. 2014a). However, isoform selective inhibitors have been developed recently. The isoform selective inhibition can be attributed to the conformation flexibility of the HSP90 ATP-binding pocket.

3.2.1 HSP90α/β Selective Inhibitors

Some HSP90 inhibitors exhibit moderate selectivity for the cytosolic HSP90α/β, such as the benzamide-based SNX2112 and purine-based BIIB021. While some other inhibitors exhibit no selectivity, such as the resorcinol-based NVP-AUY922. As the paralogue of SNX2112, SNX0723 exhibits similar HSP90α/β selectivity with SNX2112. It was used as the probe molecule to analyze the structural basis for the HSP90α/β selective inhibition. In SNX0723 bound HSP90α (Fig. 20.3b), the hydrophobic tetrahydroindolone moiety targeted into the big hydrophobic pocket (Site 1) formed by Leu107, Ile110, Ala111, Phe138, Tyr139 and Trp162. To accommodate this binding mode, the flexible loop sequence 104–111 was induced to form an extended α-helix conformation with the H3 and H4 helix. The residues Leu107, Ile110 and Ala111 covered on the tetrahydroindolone moiety and formed hydrophobic interactions with it. In SNX0723 bound GRP94 (Fig. 20.3c), the equivalent sequence of HSP90α 104–111 in GRP94 (160–167) were disordered. The hydrophobic interactions between tetrahydroindolone moiety and the equivalent residues of Leu107, Ile110, Ala111 in GRP94 were missing, which led to the lower affinity to GRP94 compared with HSP90α/β. The different ligand-bound conformations of HSP90α/β and GRP94 can be attributed to the minimal primary sequence difference. Compared with HSP90α/β, five amino acids (QEDGQ) are inserted in GRP94 at position 126 (HSP90α numbering). For TRAP-1, two amino acid (AE) are inserted at position 129. The amino acid insertions are unfavorable for the ligand induced extended α-helix conformation. The pan-HSP90 inhibitors (such as NVP-AUY922, Fig. 20.3a) can’t induce the extended α-helix conformation, so they exhibit no discrimination for the four isoforms (Ernst et al. 2014a; Gewirth 2016). Compound 1 developed in our lab could conduct both flexible loop conformation and extended α-helix conformation to bind with HSP90α, so 1 possessed moderate HSP90α selective inhibition against GRP94 (Jiang et al. 2016b). These analyses provide structural insights for the design of inhibitors with improved HSP90α/β selectivity.

Fig. 20.3
figure 3

Structural insights of the HSP90α/β selective inhibition. (a) NVP-AUY922 bound HSP90α (PDB ID: 2VCI), the flexible loop sequence 104–111 is colored in magenta. (b) SNX0723 bound HSP90α (PDB ID: 4NH8), the sequence 104–111 on the extended α-helix is colored in magenta. (c) SNX0723 bound GRP94 (PDB ID: 4NH9), the inserted five amino acids (QEDGQ) are colored in yellow. Carbon atoms of the compounds are shown in green. The active sites of the proteins are surfaced according to the hydrophobic state

Full size image

Inspired by the above analyses, researchers in Vertex Pharmaceuticals conducted extensive SAR study focused on SNX0723 to discover more potent and selective HSP90α/β inhibitors. Structure optimization led to cyclopentyl analogue 2, which exhibited improved HSP90α/β selective inhibition activity, with Ki value of 5 nM for HSP90α/β and >10 μM for GRP94 and TRAP1. In human HD patient derived fibroblasts, 2 induced significant up-regulation of HSP70 and degradation of the HSP90α/β clients htt, CDK4 and AKT (Ernst et al. 2014b). The goal of this research was to develop HSP90α/β selective inhibitors for the treatment of CNS disorders, so the anti-cancer effects of 2 were not evaluated. Considering the significant oncogenic clients degradation effects and the HSP90α/β addiction of tumor cells, 2 should have potent anti-cancer effects.

TAS-116 is another HSP90α/β selective inhibitor which has already entered clinical trials. This compound was developed by Taiho Pharmaceutical Co. LTD. through fragment based drug design. TAS-116 possessed excellent HSP90α/β selective inhibition activity with Ki values of 34.7 nM, 21.3 nM, >50 μM, and >50 μM for HSP90α, HSP90β, GRP94, and TRAP1, respectively (Ohkubo et al. 2015). The accurate binding mode of TAS-116 was not reported, while the crystal structure of TAS-116 homolog 3 complexed with HSP90α was released. Considering the similar structure of the two compounds, TAS-116 and compound 3 should take same binding modes. As shown in Fig. 20.4, compound 3 also induced an extended α-helix conformation. In addition to the Site 1 pocket, 3 also targeted into a specific hydrophobic pocket (Site 4) which extended toward the opposite side of the ATP pocket. The excellent HSP90α/β selective inhibition could be partially attributed to the occupation of this pocket (Ohkubo et al. 2012).

Fig. 20.4
figure 4

Binding mode analysis of compound 3 complexed with HSP90α (PDB ID: 3WQ9). The sequence 104–111 on the extended α-helix is colored in magenta. Carbon atoms of the compounds are shown in green. The active sites of the proteins are surfaced according to the hydrophobic state

Full size image

The specific HSP90α/β inhibition of TAS-116 was confirmed at cellular level. In HCT116 cells, TAS-116 treatment induced the degradation of DDR1 (client of HSP90α) and the up-regulation of HSP70, while exhibited no influence on the LRP6/GRP94 complex formation and LRP6 expression level (LRP6 is the client of GRP94). In pre-clinical study, oral administration of TAS-116 exhibited potent anti-tumor activity in various xenograft models. In HER2-expressing NCI-N87 human gastric cancer xenografts, TAS-116 inhibited tumor growth significantly in a dose dependent manner. The T/C value was 47%, 21%, and 9% for dose of 3.6 mg/kg, 7.1 mg/kg, and 14.0 mg/kg, respectively. TAS-116 treatment downregulated the HSP90 clients HER2, HER3, and AKT in tumor tissues, indicting the potent anti-tumor activity was HSP90 inhibition dependent. As results, the PI3K/AKT and MAPK/ERK signaling pathways were inhibited. In FLT3-ITD-expressing MV-4-11 human acute myeloid leukemia xenografts, 14.0 mg/kg/day TAS-116 administration led to tumor shrinkage, accompanied with FLT3-ITD degradation in tumors. In an orthotopic lung tumor model expressing EGFR (L858R/T790M), TAS-116 administration also inhibited tumor growth significantly (Ohkubo et al. 2015).

Multiple myeloma (MM) was reported to be sensitive to HSP90 inhibitors in both pre-clinical models and clinical trails (Kaiser et al. 2010; Patterson et al. 2008; Richardson et al. 2010a, b, 2011; Stuhmer et al. 2008). As a potential therapeutic agent, the pre-clinical anti-MM activity of TAS-116 alone or in combination with other agents was well studied. TAS-116 exhibited potent cytotoxicity in a series MM cell lines and patient derived MM cells, including RAS-mutated MM cells (Suzuki et al. 2015a, b). Moreover, TAS-116 could overcome 17-AAG resistance mechanisms and was more effective than 17-AAG in N-RAS mutated MM cell lines, associated with lower HSP27 induction and more significant degradation of the clients p-C-RAF and p-MEK1/2. In addition, TAS-116 augmented the bortezomib (BTZ) cytotoxicity against MM cells. The synergistic activity was due to inhibition of BTZ-triggered canonical NF-κB activation and enhanced unfolded protein response (UPR). In murine xenografts, both TAS-116 alone and combination with BTZ exhibited potent anti-MM activities. The combination treatment inhibited tumor growth more potently than TAS-116 and BTZ monotherapy. The OS of combination treated animals was significantly prolonged. (Suzuki et al. 2015a). Considering that TAS-116 induced significant degradation of key RAS-RAF-MEK-ERK pathway regulators (p-C-RAF, p-MEK1/2, and p-ERK), the synergistic anti-MM activities of TAS-116 in combination with RAS-RAF-MEK-ERK signaling pathway inhibitors were studied. In RAS mutated MM cell lines, TAS-116 significantly enhanced the cytotoxicity and apoptosis induction effects of tipifarnib (farnesyltransferase inhibitor), dabrafenib (RAF inhibitor), and AZD6244 (MEK inhibitor). In B-RAF-mutated U266 MM cell line, synergistic cytotoxicity of TAS-116 combination with dabrafenib was also observed (Suzuki et al. 2015b). All these results indicting TAS-116 is a potent anti-MM agent, either alone used or combined with other agents.

As a favorable property, TAS-116 exhibited no ocular toxicity in pre-clinical study. In human retinal pigment epithelial ARPE-19 cells that were essential for the support of photoreceptors, TAS-116 was less toxic than 17-AAG (Suzuki et al. 2015a). In NCI-H1975 rat xenografts, TAS-116 exhibited low retina exposure and high tumor/retina drug exposure ratio. In addition, administration of TAS-116 in SD rats didn’t induce cell apoptosis and histologic change in the outer nuclear layer of the retina (Ohkubo et al. 2015).

Inspired by the excellent pre-clinical anti-tumor effects and safety profile, TAS-116 entered clinical trials in 2014 for the phase I dose escalation study enrolled patients with advanced solid tumors. Preliminary clinical data reported in 2015 revealed that two patients (GIST and NSCLC) had partial response (PR) and five patients had stable disease (SD) more than 12 weeks in the qd dose regimen (n = 16) (Shimomura et al. 2015). During the whole study, patients (n = 52) received TAS-116 according to three dose regimens, were 4.8–150.5, 107.5–295.0 and 160 mg/m2 day in qd, qod and qd×5, respectively. The maximum tolerance dose (MTD) for qd and qod regimen was 107.5 and 210.7 mg/m2 day, respectively. Anorexia, AST/ALT/gamma-GTP elevation, night blindness and visual disorders, platelet decrease (grade 3), septic shock, respiratory failure, pneumonia (grade 4) and febrile neutropenia (grade 3) were the observed dose-limiting toxicities (DLTs). For efficacy assessment, in qd, qd×5 and qod groups, PR was achieved by 12.5, 0 and 5.6%; SD was achieved by 43.8, 61.5 and 38.9%; disease control rate (>/=12 weeks) was 31.1%, 46.2% and 22.2%, respectively (Yanagitani et al. 2017). This phase I study revealed the preliminary clinical anti-cancer activity of TAS-116. In the further phase II study in patients with GIST (n = 40), TAS-116 was orally administrated on a 5-days-on/2-days-off schedule. TEAEs were observed in all patients, and 52.5% were grade 3 or higher. Gastrointestinal disorders were the most common TEAEs and could be resolved with dose interruption. Other TEAEs including diarrhea, anorexia, nausea, ALT/AST increase, anemia, fatigue, vomiting, rash and bladder infection. Of the 40 patients enrolled, the median progression-free survival was 4.4 months and no complete response (CR) and PR was achieved. Thirty-four (85%) patients had SD lasting >/= 6 weeks (Kurokawa et al. 2017). Another phase I study is ongoing in US and UK to evaluate the MTD, safety, tolerability and efficacy of TAS-116 in patients with advanced solid tumors (Takanami 2005). Based on these reported clinical data, though TAS-116 was well-tolerated, it still manifested many TEAEs. In addition, no significant anti-tumor activity was observed during the treatment. The clinical safety and efficacy of TAS-116 should be further studied.

3.2.2 GRP94 Selective Inhibitors

In consideration of the correlation between GRP94 and cancer, many GRP94 selective inhibitors have been developed and exhibited promising pre-clinical anti-cancer effects. As shown in Fig. 20.5, many GRP94 selective inhibitors have been developed. Same as the pan-HSP90 N-terminal inhibitors, GRP94 selective inhibitors can also be divided into three types. As a typical adenosine receptors agonist, purine scaffold NECA was the first discovered GRP94 selective inhibitor. It bound to purified GRP94 with a dissociation constant of 200 nM while exhibited no binding to purified HSP90 (Rosser and Nicchitta 2000). Simple structure modification focused on N-ethyl moiety of NECA didn’t lead to derivatives with higher binding affinity (Immormino et al. 2009). Considering the off-target adenosine receptor pathway activation effect of the purine scaffold and the limited structure modification feasibility, NECA is not a proper lead compound for further development. The cellular GRP94 inhibition effect of NECA has not been reported. In 2013, combined compound library screening with structural and computational analyses led to the discovery of another purine scaffold GRP94 selective inhibitor PU-H54. As shown in Fig. 20.6a, crystal complex showed that PU-H54 induced rearrangements of the GRP94 lid region, and occupied the newly exposed hydrophobic Site 2 pocket. Further structure modification led to PU-WS13, which exhibited more potent and selective GRP94 inhibition activity. PU-WS13 inhibited IGF-II secretion and TLR9 trafficking in cells in a dose dependent manner, and exhibited no effects on HSP70 induction and AKT degradation, indicting GRP94 specific inhibition in cells. The discovery of PU-WS13 provided a new tool to investigate the cellular functions of GRP94. PU-WS13 exposure led to membrane but not cytosolic HER2 reduction in SKBr3 cells, while no same effect was observed in HER2 low-expressing MCF-7 cells. The researchers further demonstrated that GRP94 was important for maintaining the architecture of high-density HER2 formations at the plasma membrane in high HER2-expressing cells. PU-WS13 treatment induced sub-G1 cell cycle arrest, apoptosis and impaired cell viability in various HER2-overexpressing breast cancer cell lines (Patel et al. 2013). In addition, tumor specific retention effect was also observed for PU-WS13, same as the pan-HSP90 inhibitors (Patel et al. 2015). All these results indicated that GRP94 was a promising therapeutic target for HER2-overexpressing breast cancer. The therapeutic effects of PU-WS13 in multiple myeloma and hepatocellular carcinoma were also observed. In various multiple myeloma cell lines, PU-WS13 induced both apoptosis and necrosis, and inhibited cell growth. This was believed to be caused by the Wnt-LRP-Survivin pathway blockade effect of PU-WS13 (Hua et al. 2013). In human HepG2 and PLC hepatocellular carcinoma cell lines, PU-WS13 treatment exhibited significant growth inhibition effect. In HepG2 cells, PU-WS13 lowered the levels of total β-catenin and its downstream target cyclin D1, indicating Wnt signaling pathway blockade. PU-WS13 treated cells also expressed lower levels of hepatocyte growth factor (HGF) and its downstream targets, including p-ERK1/2, p-AKT, and p-STAT3. Further gene expression profile analysis revealed that nine genes involved in cell cycle modulation were down-regulated (Rachidi et al. 2015). All these studies provided proof-of-concept for the anti-cancer effects of GRP94 selective inhibitor PU-WS13 at cellular level. Further pharmacodynamic evaluation in animal models should be conducted to study the anti-cancer effects of PU-WS13.

Fig. 20.5
figure 5

Structures of GRP94 selective inhibitors

Full size image
Fig. 20.6
figure 6

Structural insights of the GRP94 selective inhibition. (a) GRP94-PU-H54 complex (PDB ID: 3O2F). (b) GRP94-4 complex (PDB ID: 5IN9). Carbon atoms of the compounds are shown in green. The active sites of the proteins are surfaced according to the hydrophobic state

Full size image

Another major class of GRP94 selective inhibitors containing resorcinol pharmacophore were developed in Blagg group. Radamide (RDA) is a radicicol-geldanamycin chimera that hybrids the resorcinol moiety from radicicol and the quinone moiety form geldanamycin. Further crystal study in Daniel T. Gewirth group showed that RDA adopted different conformations to bind with HSP90α and GRP94. In the RDA-GRP94 complex, the amide linkage exhibited two orientations. In the cis-amide orientation, the quinone moiety inserted into the GRP94 specific Site 3 and formed van der Waals interactions. In the trans-amide orientation, the quinone moiety was targeted into the solvent area. In the RDA-HSP90α complex, the amide linkage only exhibited a trans-amide orientation and the quinone moiety formed tight hydrogen-bonding network with the surrounded residues. These results indicting GRP94 selective inhibitors could be designed based on the cis-amide orientation RDA (Duerfeldt et al. 2012; Immormino et al. 2009). In the following study of Blagg group, imidazole was chosen as the bioisoster of cis-amide. Hydrophobic phenyl ring was introduced to form π-interactions with Phe199, Tyr200, and Trp223 in the GRP94 specific Site 3 pocket. These efforts led to BnIm, which exhibited 12-fold selectivity for GRP94. BnIm treatment inhibited Toll receptor trafficking and IGF-II secretion in cells. BnIm exhibited no cytotoxicity in MCF7, SKBr3, and HEK293 cells even at 100 μM (Duerfeldt et al. 2012). This phenotype was different with the purine-type GRP94 selective inhibitor PU-WS13, which exhibited selective anti-proliferative activity against the HER2-overexpressing breast cancer cells. In Gabriela Chiosis group’s study, BnIm showed little effect on the HER2 expression in SKBr3 cells. The researchers explained that these different phenotypes between BnIm and PU-WS13 might contribute to the different binding modes to GRP94 (Patel et al. 2015). The discovery of BnIm provided further insights for the design of GRP94 selective inhibitor, but the low activity and selectivity limited its usage as a probe to study the GRP94 cellular function. Further modification of BnIm led to the more potent and selective compound 4 and KUNG29. As shown in Fig. 20.6b, crystal structure analysis showed that 4 bound into the Site 3 pocket of GRP94, confirmed the binding mode of this series. KUNG29 exhibited an IC50 value of 0.2 μM for Grp94 and a 41-fold selectivity over HSP90α. In MDA-MB-231 cells, KUNG29 exhibited potent antimigratory activity without cytotoxicity. Further mechanism study showed that KUNG29 induced integrin α2 degradation, indicting the antimigratory activity was GRP94 dependent. KUNG94 was another resorcinol-based GRP94 selective inhibitor. Inspired by the binding mode of PU-WS13, hydrophobic moieties were introduced at the 5-position of resorcinol to occupy the GRP94-specific hydrophobic site 2 pocket. SAR study led to KUNG94 incorporating a 4-fluorobenzyl moiety manifested ∼10-fold selectivity for GRP94. The IC50 value of KUNG94 for GRP94 and HSP90α was 8 nM and 77 nM, respectively. In RPMI8226 multiple myeloma cell line, KUNG94 manifested a GI50 value of 1.4 μM. The level of GRP94 specific client LRP6, a Frizzled coreceptor, was reduced after treatment. In addition, the antiapoptotic protein Survivin was reduced, indicting KUNG94 activated apoptosis through decreasing the Wnt signaling. Considering that KUNG94 possesses a low selectivity profile and still high inhibition activity to HSP90α, it is difficult to distinguish the cellular effects of GRP94 inhibition and HSP90α inhibition. The selectivity profile of KUNG94 should be further improved (Khandelwal et al. 2017).

Recently, Blagg group developed an benzamide-containing GRP94 selective inhibitor called ACO-1 through structure-based design. In this study, aminocyclohexanol-based clinical HSP90 N-terminal inhibitor SNX2112 was chosen as the rational design template. In order to access the GRP94 specific Site 2 pocket and form additional π interactions with Phe-195, tetrahydroindazolone fragment in SNX2112 was replaced with a pyrrole, and further SAR study led to ACO-1. ACO-1 manifested an IC50 value of 0.44 μM for Grp94 while had no inhibition for HSP90α at 100 μM, displaying over 200-fold selectivity. In PC3-MM2 cells, ACO-1 treatment led to reduced levels of GRP94 specific clients, including integrin α2, integrin αL, Syne2, VAMP2, and Rab10. These clients are important for the migration process of cancer cells. As significant characteristics during cell migration process, filamentous actin (f-actin) localization and integrin trafficking were inhibited in ACO-1 treated cells. In further wound healing scratch assay, ACO-1 potently inhibited the migration of the highly metastatic MDA-MB-231 and PC3-MM2 cells. These results confirmed the anti-migratory activity of GRP94 selective inhibitors (Mishra et al. 2017).

3.2.3 TRAP1 Selective Inhibitors

As TRAP1 involved in multiple processes of cancer development, TRAP1 inhibition has been a promising anti-cancer strategy. Considering that TRAP1 mainly located at mitochondria, mitochondria specific delivery of pan-HSP90 inhibitor is a feasible strategy to inhibit TRAP1. In 2009, Byoung Heon Kang and coworkers developed the first mitochondria-targeted HSP90 inhibitor. Combination of the pan-HSP90 inhibitor 17-AAG with mitochondrial targeting moieties led to Gamitrinibs (G-G1−G-G4, G-TPP). Gamitrinibs significantly accumulated in the isolated tumor mitochondria and induced immediate loss of inner mitochondrial membrane potential and release of cytochrome c. While the unconjugated 17-AAG didn’t exhibit these properties. Gamitrinibs exhibited potent anti-cancer effects. In lung adenocarcinoma H460 cells, Gamitrinibs treatment significantly induced cell death, mitochondrial apoptosis and decreased colony formation. Unlike 17-AAG, Gamitrinibs didn’t induce degradation of HSP90α clients AKT and Chk1. The HSP70 expression level was also not affected, indicating TRAP1 selective inhibition didn’t trigger heat shock response. Gamitrinibs exhibited broad spectrum anti-cancer activity against various cancer cell lines, while exhibited low effects in normal cells, indicting they were selective anti-cancer agents. The anti-cancer effects and safety of Gamitrinibs were further demonstrated in vivo. Gamitrinib (G-G4) administration potently inhibited the growth of human leukemia, breast and lung xenograft tumors without significant body weight loss. In the tumors harvested from Gamitrinib-treated animals, extensive apoptosis and release of cytochrome c in the cytosol was observed. In addition, organs of the Gamitrinib-treated animals exhibited similar histological properties compared with the vehicle group, confirmed the safety of Gamitrinib (Kang et al. 2009). Inspired by these preliminary in-vivo anti-cancer effects and safety profile, the researchers then conducted pre-clinical study of gamitrinibs in advanced prostate cancer. In vitro, Gamitrinibs potently inhibited cell viability in prostate cancer cells. In PC3 xenograft tumors, 10 mg/kg i.p. administration of Gamitrinib (G-TPP) completely inhibited the tumor growth, while 10 mg/kg 17-AAG administration exhibited no effect. Meanwhile, no significant weight loss and histological change of organs were observed in the Gamitrinib-treated group. In bone metastatic prostate cancer models, administration of Gamitrinib significantly reduced bone loss compared with the vehicle group (Kang et al. 2010). The anti-cancer effects of Gamitrinib against advanced prostate cancer were further demonstrated in Transgenic Adenocarcinoma of the Mouse Prostate (TRAMP) models. Long-term treatment TRAMP mice with Gamitrinib (G-G4) suppressed the formation of localised prostate tumors of neuroendocrine or adenocarcinoma origin, while had no effect on prostatic intraepithelial neoplasia or prostatic inflammation. In addition, the metastasis of prostate cancer to liver and abdominal lymph nodes was also inhibited. No significant animal weight loss or organ toxicity was observed during treatment (Kang et al. 2011). These results indicated that Gamitrinibs were effective and safe agents for the treatment of advanced prostate cancer. Synergistic anti-cancer activities of Gamitrinib with other agents were observed in various models. In glioblastoma cells, noncytotoxic concentrations of G-TPP sensitized tumor cells to TNF-α–related apoptosis-inducing ligand (TRAIL). In nude mice carrying intracranial U87-Luc glioblastomas, combination of TRAIL and G-TPP exhibited potent in-vivo antiglioma activity. Mechanically, G-TPP treatment triggered UPR effect. The synergy between TRAIL and G-TPP was attributed to the UPR-related suppression of NF-κB–dependent gene expression (Siegelin et al. 2011). G-TPP also exhibited synergistic anti-cancer effects with doxorubicin. In vitro, combination of G-TPP with doxorubicin exhibited enhanced cytotoxicity and apoptotic inducing effect. This synergistic effect was associated with increased expression of pro-apoptotic CHOP and Bim, and the enhanced mitochondrial accumulation of Bim and Bax. In prostate and breast xenograft models, the combination treatment exhibited more potent tumor growth inhibition compared with single treatment with G-TPP or doxorubicin. In addition, no aggravating cardiotoxic side effect of doxorubicin was observed during treatment (Park et al. 2014a). G-TPP could activate the calcium-mediated UPR in the ER lumen and impair the stress adaptation of cancer cells. Combination treatment with G-TPP and ER stressor thapsigargin augmented interorganelle stress signaling, and exhibited synergistic anti-cancer effects. In relapsed prostate cancer xenograft models, G-TPP and thapsigargin combination potently inhibited tumor growth (Park et al. 2014b). In general, Gamitrinibs manifested potent in-vitro and in-vivo anti-cancer effects either single used or combined with other agents. They are potential anti-cancer agents and provide proof-of-concept for mitochondrial HSP90 inhibition as novel anti-cancer strategy (Fig. 20.7).

Fig. 20.7
figure 7

Structures of TRAP1 selective inhibitors

Full size image

In 2015, Byoung Heon Kang group demonstrated that TRAP1 was the major HSP90 paralogue located in cancer cell mitochondria, with an approximately tenfold higher level than HSP90α/β. In addition, TRAP1 possessed an approximately 32-fold higher ATPase activity. Specific inactivation of TRAP1 alone by siRNA was enough to induce ER UPR in cancer cells. These results indicated that TRAP1 was the predominant target of Gamitrinibs and TRAP1 inhibition was responsible for the potent anti-cancer effects of Gamitrinibs. To design novel TRAP1 selective inhibitor, the researchers resolved the crystal structure of TRAP1 complexed with the pan-HSP90 inhibitor PU-H71. Under the guidance of the crystal structure, another mitochondrial TRAP1 inhibitor named SMTIN-P01 was developed. In SMTIN-P01, the solvent-expose isopropyl amine moiety of the PU-H71 was replaced with the mitochondria-targeting moiety TPP. SMTIN-P01 exhibited dramatic mitochondria accumulation effect compared with the parent molecule PU-H71. In Hela cells, SMTIN-P01 treatment induced significant mitochondrial membrane depolarization. In addition, the expression levels of HSP90α clients (AKT and Chk1) and HSP70 were not affected, indicating no inhibition on the cytosolic HSP90. In various cancer cell lines, SMTIN-P01 exhibited more potent cytotoxicity than the parent molecule PU-H71. These results confirmed that TRAP1 selective inhibition is an effective anti-cancer strategy (Lee et al. 2015a).

Gamitrinibs and SMTIN-P01 inhibit TRAP1 via mitochondrial delivery of pan-HSP90 inhibitors, and provide favorable probes to study the function of TRAP1. Alternatively, TRAP1 selective inhibitors with higher affinity to TRAP1 than the cytosolic HSP90α/β may have better applications. Considering the high sequence homologies and the almost superimposable nucleotide-binding sites of the HSP90 four paralogues, development of TRAP1 selective inhibitors is difficult. Recently, the TRAP1 crystal structures complexed with different ligands were resolved, which provided structural guidance for the rational design of TRAP1 selective inhibitors. As shown in Fig. 20.8a, b, some differences were observed in the ligands bound HSP90α and TRAP1. The lid structure (residues 171–201) in TRAP1 was disordered, and residues Asn171 and Gly202 in the two proteins adopted different configurations. These differences indicate compounds that can stabilize the flexible lid structure may manifest TRAP1 selectivity (Lee et al. 2015a; Park et al. 2017). Based on these analyses, Byoung Heon Kang group conducted structure modification focused on BIIB021 to generate TRAP1 selective inhibitors. The purine scaffold which closed to the disordered lid was changed into pyrazolopyrimidine, then bromopiperonyl was introduced on N-1 position. These efforts led to compound 5, which exhibited modest TRAP1 selective inhibition (IC50 = 79 nM) compared with HSP90α (IC50 = 698 nM). Crystal structure analysis confirmed the structural basis for the selectivity. As shown in Figs. 20.8c, compared with the BIIB021 bound TRAP1, the pyrazolopyrimidine core in 5 formed π-π stacking interaction with the residue Phe201, water-mediated hydrogen bonds were formed between N-2 and Asn171 amide. These additional interactions made great contribution to the TRAP1 selectivity of 5. In Hela cells, 5 treatment induced loss of mitochondrial membrane potential, overproduction of mitochondrial ROS, elevation of CHOP expression and discharge of cytochrome c, which were all signatures of TRAP1 inhibition. 5 possessed potent cytotoxicity against various cancer cells lines. In PC3 xenograft models, 30 mg/kg daily treatment inhibited tumor growth significantly without toxicity. Both in cancer cells and tumor tissues, 5 treatment had no influence on the HSP70 expression level, indicating no heat shock response was induced (Park et al. 2017). TRAP1 selective inhibitors may overcome the heat shock response effect of pan-HSP90 inhibitors (Fig. 20.9).

Fig. 20.8
figure 8

Structural insights of the TRAP1 selective inhibition. (a) HSP90α-BIIB021 complex (PDB ID: 3QDD). (b) TRAP1-BIIB021 complex (PDB ID: 4Z1G), the NTD is intercepted from the TRAP1NM. (c) TRAP1-5 complex (PDB ID: 5Y3N), the NTD is intercepted from the TRAP1NM. The conserved water molecule is shown as red spheres, hydrogen bonds are indicated by magenta dashed lines. Carbon atoms of the compounds are colored in green. N171 and F201 in (c) are shown in magenta sticks, other residues are shown in yellow sticks. The active sites of the proteins are surfaced according to the hydrophobic state

Full size image
Fig. 20.9
figure 9

Structures of novobiocin and representative derivatives

Full size image

4 HSP90 C-Terminal Inhibitors

Another ATP-binding pocket is located at the HSP90 CTD. Some inhibitors are reported to occupy this C-terminal ATP-binding pocket and exhibit promising anti-cancer effects.

4.1 Novobiocin and Derivatives

As the first discovered HSP90 C-terminal inhibitor, novobiocin provides a proof-of-concept for HSP90 C-terminal inhibition. In 2000, Neckers and coworkers discovered that novobiocin exhibited HSP90 inhibition activity in a different manner compared with the HSP90 N-terminal inhibitors geldanamycin and radicicol (Marcu et al. 2000b). Further studies demonstrated that novobiocin bound to a second ATP-binding pocket which located at the HSP90 CTD (Marcu et al. 2000a; Soti et al. 2002). In SKBr3 cells, novobiocin treatment reduced p185erbB2, mutated P53, and RAF-1 protein levels in a dose-dependent manner (Marcu et al. 2000b). To improve the anti-proliferative effects and simplify the structure of novobiocin, Blagg group conducted extensive SAR study and discovered many derivatives with potent anti-cancer effects. As representative, KU-135 exhibited dramatically improved anti-proliferative activity against Jurkat T Lymphocytes and melanoma cells compared with novobiocin. KU135 treatment induced significant mitochondria-mediated apoptosis and G2/M cell cycle arrest in these cells. Moreover, HSP90 client degradations were observed without obvious HSP70 induction (Samadi et al. 2011; Shelton et al. 2009). KU675 is another novobiocin derivative with potent anti-cancer effects. In both androgen-dependent and -independent prostate cancer cell lines, KU675 exhibited potent anti-proliferative activity and cytotoxicity. Interestingly, KU675 exhibited some HSP90α selectivity compared with HSP90β (191 vs 726 μM, K d in an intrinsic-fluorescence-spectra based binding assay). In prostate cancer cells, KU675 induced significant degradation of HSP90α specific client proteins (Survivin, B-RAF), but had no obvious effect on the HSP90β specific CXCR4, which confirmed the HSP90α selectivity. In addition, KU675 manifested Hsc70 binding affinity with a K d value of 76.3 μM, which associated with the degradation inducing effects of the Hsc70 client proteins Myb and Sp1. KU675 didn’t trigger heat shock response in PC3-MM2 cells, and exhibited a modest down-regulation of HSP (Hsc70 and HSP27) in the LNCap-LN3, LAP C-4, and C4-2 cells (Liu et al. 2015). Compound 6 was a recently developed novobiocin derivative in Blagg group. 6 potently inhibited the proliferation of SKBr3 and MCF-7 cells with IC50 of 0.14 and 0.64 μM, respectively (Zhao et al. 2015). David Montoir and coworkers developed another novobiocin derivative 7, which showed potent anti-proliferative properties in MCF-7 and MDA-MB-468 breast cancer cells (Montoir et al. 2016). Though the above novobiocin derivatives showed potent anti-proliferative activity in cancer cells, the in-vivo anti-tumor efficacy was not evaluated. Recently, our group developed a novobiocin analogue 8 incorporating a 6-acylamino-2-aminoquinoline scaffold. 8 potently inhibited the proliferation of MCF-7 and SKBr3 cells. In MCF-7 cells, 8 inhibited cell migration and induced significant apoptosis. We further evaluated the in-vivo anti-tumor effects of 8 in 4T1 mice breast xenografts. 8 potently inhibited tumor growth and lung metastasis potential without obvious toxicity, indicating 8 was an effective and safe anti-tumor agent (Jiang et al. 2017). All the above novobiocin derivatives inhibited HSP90 without heat shock response.

4.2 Deguelin and Derivatives

Deguelin is a natural rotenoid isolated from several plant species, such as Mundulea sericea. In recent years, the potent cancer chemo-preventive and anti-tumor activities of Deguelin against various cancer types have been demonstrated (Lee et al. 2008; Thamilselvan et al. 2011; Udeani et al. 1997). Though some studies indicate that the potent anti-cancer effects of Deguelin is related to the inhibition of several cell signaling pathways, such as PI3K-AKT and GSK-3β/β-catenin, its accurate binding target is still unclear (Chun et al. 2003; Thamilselvan et al. 2011). In 2007, to explore the anti-cancer mechanisms of Deguelin, Oh and coworkers observed that Deguelin could reduce the expression of HIF-1α in a HSP90 inhibition-mediated mechanism. Additionally, Deguelin could induce the degradation of HSP90 clients (CDK4, AKT, eNOS, MEK1/2, and mutant P53) in a time- and dose-dependent manner. ATP-Sepharose assay suggested that Deguelin could bind to HSP90 competitively with ATP, and no activity of Deguelin and its analogues was observed in HSP90 N-terminal inhibitor based fluorescence polarization assay, indicating Deguelin might bind to the HSP90 C-terminal ATP binding pocket (Chang et al. 2012; Oh et al. 2007). In order to elucidate the essential skeleton for the potent HSP90 inhibition activity and discover novel and safe C-terminal inhibitors, Chang and coworkers conducted extensive SAR studies for Deguelin. These efforts led to the discovery of some simplified derivatives with potent anti-proliferative activity (Chang et al. 2012). In the further pre-clinical evaluation, the C-ring truncated L80 and B, C-ring truncated SH-1242 exhibited potent anti-tumor effects in various cancer cell lines and xenograft tumor models (Lee et al. 2015b, 2016). It is reported that the potential Parkinsonism-like neurotoxicity of Deguelin may limit its clinical application (Caboni et al. 2004). Compared with Deguelin, L80 and SH-1242 displayed significantly reduced cytotoxicity against normal cells. In addition, SH-1242 didn’t produce obvious Parkinsonism-like toxicity in rat brains. These results indicated that L80 and SH-1242 were safe anti-cancer agents. In the mechanism study, results of ATP-Sepharose assay and fluorescence-based equilibrium binding assay indicated that the HSP90 inhibition activity of the derivatives was mainly mediated by the direct binding with HSP90 CTD. Docking modeling suggested that the derivatives could bind into the C-terminal ATP-binding pocket and form key interactions with the residues Ser677 and Lys615. Further modification of L80 and SH-1242 led to the discovery of the third generation Deguelin derivatives (compound 9 and 10) (Kim et al. 2015, 2016). The discovery of these novel Deguelin derivatives enriched the structural diversity of HSP90 C-terminal inhibitors and verified the druggable ability of HSP90 C-terminal domain.

4.3 Dihydropyrimidinones

Drug design based on the endogenous ligand is an efficient strategy to discover specific inhibitors of the targets. As a successful case, the purine class HSP90 N-terminal inhibitors were designed based on the endogenous ligand ATP (Biamonte et al. 2010). It is reported that the HSP90 CTD nucleotide binding pocket preferred to interact with GTP and UTP (Soti et al. 2003). Therefore, GTP and UTP can serve as the leads to design novel HSP90 C-terminal inhibitors. Inspired by the structural analogy between UTP and the privileged heterocyclic core 3,4-dihydropyrimidin-2-(1H)-one (DHPM), Maria Strocchia and coworkers synthesized several DHPM-based compounds. In Surface Plasmon Resonance (SPR) based binding analysis, seven compounds possessed high binding affinity with the immobilized HSP90α. Compound 11 exhibited moderate cytotoxic effects, which was in line with the SPR affinity. Western-blot analysis showed that compound 11 could induce the HSP90 clients RAF-1 and p-AKT degradation, while the HSP90 and HSP70 levels were not affected. These results verified that compound 11 was an inhibitor of HSP90. Limited proteolysis and oligomerization assays confirmed compound 11 interacting with the C-terminal domain of HSP90α. Further docking analysis showed that compound 11 bound with the region located at the dimerization site interface. Compound 11 was the first reported non-nature-inspired HSP90 C-terminal inhibitor (Strocchia et al. 2015). In order to explore the SARs systematically and discover more potent HSP90 C-terminal inhibitors, researchers conducted further structural modifications and developed the second (12) and third generation (13) derivatives based on the DHPM core (Terracciano et al. 2016a, b). The two derivatives exhibited improved anti-proliferative activity, especially compound 13. It inhibited the growth of A375 cancer cells potently with an IC50 value of 2.1 ± 0.3 μΜ. These DHPM-based HSP90 C-terminal inhibitors provided new options for the treatment of cancer and the discovery process provided reference for the design of novel HSP90 C-terminal inhibitors.

4.4 EGCG and Derivatives

Epigallocatechin gallate (EGCG) is the major polyphenol component in green tea which exhibits broad anti-cancer and chemo-prevention effects. Previously study showed that EGCG could inhibit proliferation and induce apoptosis via blocking multiple essential survival pathways in human cancer cells, including JAK/STAT, MAPK, PI3K/AKT, Wnt and Notch (Singh et al. 2011). In 2005, Gasiewicz group first reported that EGCG exhibited chemo-preventive effects through directly binding to HSP90 (Palermo et al. 2005). Then they further demonstrated that EGCG could inhibit the activity of HSP90 through proteolytic footprinting and immunoprecipitation. Further ATP-agarose pull-down assay declared EGCG bound to the C-terminal ATP binding pocket and interacted with amino acids (538–728) same as novobiocin (Yin et al. 2009). In pancreatic cancer cells, EGCG could impair the HSP90 super-chaperone complex and downregulate the HSP90 clients (AKT, Cdk4, RAF-1, Her-2, and pErk) (Li et al. 2009). In addition, EGCG inhibited the expression of HSP70 and HSP90 in MCF-7 cells and CT26 xenograft models by inhibiting the expression of the transcription factors HSF1 and HSF2 (Tran et al. 2010). As results, EGCG inhibits HSP90 chaperone function with no heat shock response.

To improve the anti-proliferative activity and drug-like properties, Anuj Khandelwal and coworkers conducted structure modification on EGCG and illustrated the SARs. The systematical investigation of the four rings and the linker between C- and the D-rings led to discovery of some derivatives with enhanced anti-proliferative activity, such as 14 (IC50, 4 μM in MCF-7 cells). 14 exhibited more than 18-fold improved anti-proliferative activity against MCF-7 cells compared with EGCG. It could also induce the HSP90 clients HER2, RAF, and p-AKT degradation in a dose-dependent manner, indicating the potent anti-proliferative activity was HSP90 inhibition dependent (Bhat et al. 2014a). Though the accurate binding modes of EGCG and 14 are unclear, they provide novel scaffolds for HSP90 C-terminal inhibitors.

4.5 Derrubone and Derivatives

Derrubone is a natural isoflavonoid isolated from the Indian tree Derris Robusta (East et al. 1969). It was found to inhibit HSP90 in 2007 by Hadden and coworkers through high-throughput screening based on luciferase refolding inhibition. Derrubone inhibited HSP90-dependent refolding of denatured firefly luciferase with an IC50 value of 0.23 ± 0.04 μΜ. It also exhibited low micro-molar inhibition activity against various cancer cell lines (IC50 in MCF-7 and SkBr3 cancer cell lines are 9 and 12 μM, respectively). Meanwhile, the HSP90 clients HER2, RAF-1, AKT and ERα were dose-dependently degraded. Additionally, Derrubone could stabilize the HSP90/CDC37/kinases complex and then prevent the chaperone cycle from progressing (Hadden et al. 2007). Through docking study, Khalid and coworkers predicted that Derrubone might bind to the HSP90 C-terminal ATP-binding pocket and interact with the residues Leu665, Leu666, and Leu694 (Khalid and Paul 2014).

Hastings and coworkers reported the SAR study of Derrubone as HSP90 inhibitors and got three derivatives (15–17) with improved anti-proliferative activity. Derivative 17 possessed the most potent anti-proliferative activity. Meanwhile, the derivatives could induce the degradation of the clients HER2 and RAF dose-dependently (Hastings et al. 2008). In some extent, further modification can be conducted on Derrubone to improve the HSP90 inhibition and anti-proliferative activity.

4.6 Silybin and Derivatives

Silybin, a flavonoid natural product, is the major active constituent in the silymarin that is extracted from milk thistle seeds. In recent years, the anti-cancer effects and chemo-preventive efficacy of Silybin was demonstrated in various cancer types, such as skin, prostate, breast, lung, etc. (Deep and Agarwal 2007; Ting et al. 2013). Encouraged by the promising effects in pre-clinical models, silybin entered clinical trials for safety and pharmacodynamic study (Flaig et al. 2007, 2010; Hoh et al. 2006). The anti-cancer and chemo-preventive mechanisms of silybin have been well studied, including growth inhibition, angiogenesis inhibition, chemo-sensitization, and metastasis inhibition (Desgrosellier and Cheresh 2010). Considering that silybin could block the cell cycle and decrease the Cyclin D1, CDK4 and CDK6 levels (well known HSP90 clients), Zhao and coworkers hypothesized silybin might be a HSP90 inhibitor (Zhao et al. 2011; Zi and Agarwal 1999). The following study showed that silybin could inhibit HSP90-dependent refolding of denatured firefly luciferase (IC50, 0.23 ± 0.04 μΜ) and induce HSP90 clients (HER2, RAF-1, and AKT) degradation in a dose dependent manner. Riebold and coworkers demonstrated that silybin could bind to HSP90 C-terminal through an ELISA assay. Further NMR spectroscopy analysis suggested that silybin bound to the C-terminal ATP-binding pocket, same as novobiocin (Riebold et al. 2015).

Zhao and coworkers conducted SAR study of silybin. The phenol-removal derivatives 18, 19, and 20 exhibited significant improved cytotoxicity against MCF-7 and SKBr3 cancer cell lines. Western blot analysis verified these derivatives induced HSP90 client proteins degradation without heat shock response (Zhao et al. 2011).

5 HSP90 Inhibitors Blocking HSP90-CDC37 Protein-Protein Interactions (PPI)

Cochaperones are important components of the HSP90 chaperone machine. They form HSP90-cochaperone complex and regulate the HSP90 chaperone function. Inhibition of the HSP90-cochaperone PPI is a promising strategy to block the HSP90 chaperone function. CDC37 is the kinases specific HSP90 cochaperone. Theoretically, inhibiting the HSP90-CDC37 PPI might specifically block the maturation of the kinase clients, while have no influence on other client proteins. Thus, HSP90-CDC37 PPI inhibitors are speculated to reduce the mechanism related side effects of HSP90 N-terminal inhibitors (Trepel et al. 2010).

Celastrol is the first discovered HSP90-CDC37 PPI inhibitor. In 2008, Duxin Sun lab demonstrated that celastrol could disrupt the HSP90-CDC37 PPI via immunoprecipitation assay. In Panc-1 cells, celastrol treatment induced HSP90 client proteins (CDK4 and AKT) degradation significantly. As results, celastrol inhibited Panc-1 cell growth and induced apoptosis. In Panc-1 xenograft mice, celastrol treatment inhibited tumor growth potently. These studies provided proof-of-concept for HSP90-CDC37 inhibitors as anti-tumor agents (Kannaiyan et al. 2011). Further study revealed that celastrol covalently bound to CDC37 via a conjugate addition between the Michael reaction acceptor substructure and cysteine residues on CDC37, and then blocked the HSP90-CDC37 interaction (Sreeramulu et al. 2009a). Recently, our lab studied the SARs of celastrol derivatives as HSP90-CDC37 PPI inhibitors and discovered CEL20, which exhibited improved HSP90-CDC37 disruption activity and anti-proliferative effects (Jiang et al. 2016a). Some other agents are reported to be HSP90-CDC37 blockers, including Withaferin A, FW-04-806, Apigenin, Sulforaphane and Kongensin A (Wang et al. 2017a). All these agents are natural products incorporating covalent-binding moieties and the accurate HSP90-CDC37 blocking mechanisms are unclear and their further applications are limited. Thus, rational designed HSP90-CDC37 PPI inhibitors with high affinity and clear mechanisms are urgently needed (Figs. 20.10, 20.11, 20.12, 20.13 and 20.14).

Fig. 20.10
figure 10

Structures of Deguelin and representative derivatives

Full size image
Fig. 20.11
figure 11

Structures of dihydropyrimidinones as HSP90 C-terminal inhibitors

Full size image
Fig. 20.12
figure 12

Structures of EGCG and representative derivative

Full size image
Fig. 20.13
figure 13

Structures of Derrubone and representative derivatives

Full size image
Fig. 20.14
figure 14

Structures of Silybin and representative derivatives

Full size image

The crystal structures of HSP90N-CDC37M/MC complex have been resolved and provides structural insights for the rational design of HSP90-CDC37 PPI inhibitors. As shown in Figs. 20.15, two interaction regions are observed on the HSP90N-CDC37M interface. The residues on CDC37M α4 and α5 helix form hydrophobic interactions with the HSP90N H4 helix. Mutation results suggest that Leu205 is critical for the HSP90-CDC37 PPI. Additionally, the CDC37M α1 and α2 helix and the connecting loop interact with the H5 helix of HSP90N (Roe et al. 2004; Sreeramulu et al. 2009b). Recent calculation results in our lab suggest that residues Lys160, His161, Met164, Leu165, Arg166 and Arg167 contributed the most to the binding free energy of the complex, especially the Arg167 (Song et al. 2015).

Fig. 20.15
figure 15

HSP90-CDC37 PPI interface (PDB ID: 2K5B). HSP90N and CDC37M are colored in blue and magenta, respectively. Sequence 160–170 on CDC37M is shown in red, α-helix numbers are indicated

Full size image

Peptides derived from the PPI interface are favorable leads for the development of PPI inhibitors. Based on the crystal structures and calculation analysis, our group developed several peptides derived from CDC37 that could block the HSP90-CDC37 PPI. Pep1 was an 11-peptide directly extracted from CDC37 (sequence 160–170). Pep1 bound to HSP90N with K D value of 6.90 μM and inhibited the HSP90 ATPase with IC50 of 3.0 μM. In addition, Pep1 interfered the HSP90-CDC37 PPI in a biolayer interferometry (BLI) based competitive assay (Wang et al. 2015). Further optimization led to the pentapeptide Pep5, which exhibited slightly improved HSP90N binding affinity with K D value of 5.99 μM. GST-HSP90N pull down assay further demonstrated that Pep5 could block HSP90-CDC37 PPI (Wang et al. 2017b). The discovery of Pep5 further confirmed that the residues His161, Met164, Leu165, Arg166 and Arg167 on CDC37 were critical for the HSP90-CDC37 PPI. Thus, small molecules disrupting the interactions between these residues and HSP90 can block the HSP90-CDC37 PPI. Based on these results, a pharmacophore model was generated and used for virtual screening to discover HSP90-CDC37 PPI inhibitors. Fortunately, we discovered compound VS-8 exhibited moderate HSP90-CDC37 PPI inhibition activity with IC50 value of 76.85 μM in Homogeneous Time-Resolved Fluorescence (HTRF) assay. Further optimization of VS-8 led to compound 21, which exhibited threefold improved activity with IC50 value of 27.40 μM. Compound 21 bound to HSP90 with K D value of 40.4 μM. In GST-pull down assay, 21 blocked HSP90-CDC37 interaction in a dose dependently manner. Moreover, 21 possessed moderated anti-proliferative activity in cancer cells and induced HSP90 kinase clients AKT and CDK4 degradation (Wang et al. 2017b). As the first reported non-natural small molecule HSP90-CDC37 PPI inhibitor, 21 can serve as a probe to study the HSP90-CDC37 PPI (Fig. 20.16).

Fig. 20.16
figure 16

Structures of HSP90-CDC37 PPI inhibitors

Full size image

6 Conclusions

HSP90 isoforms are involved in multiple processes of cancer development, and HSP90 inhibition has been a promising anti-cancer strategy. Till now, HSP90 inhibitors with different mechanisms have been developed. Many pan-HSP90 N-terminal inhibitors have entered clinical trials. Despite the potent pre-clinical anti-cancer effects, these inhibitors exhibit disappointing clinical efficacy. In addition, many serious adverse effects are occurred during the treatment, such as the ocular toxicity. The adverse effects may limit the clinical dosage, which is a possible reason for the poor clinical efficacy. Heat shock response triggered by the pan-HSP90 inhibitors may be another reason. Thus, inhibitors with reduced adverse effects and no heat shock response may bring breakthrough for the HSP90 inhibitor development. HSP90 four isoforms have different functions and client protein profiles, so isoform selective inhibitors may have reduced adverse effects. TAS-116 is a HSP90α/β selective inhibitor which has entered clinical trials, and moderate clinical anti-cancer effects are observed in patients with advanced solid tumors. For further clinical studies, TAS-116 should be evaluated in more cancer types to find the most sensitive patient populations. Moreover, the heat shock response effect should be a concern of TAS-116, which may neutralize the anti-cancer effects. GRP94 selective inhibition and TRAP1 selective inhibition don’t trigger heat shock response, may be more promising. GRP94 inhibitor PU-WS13 exhibited favorable anti-proliferative activity in HER2-overexpressing breast cancer cells and multiple myeloma cells. Additionally, some GRP94 inhibitors possess anti-migratory potential. The anti-cancer effects of GRP94 selective inhibitors should be further evaluated in pre-clinical study. Studies of Byoung Heon Kang and coworkers demonstrate TRAP-1 selective inhibition is a promising anti-cancer strategy. Though Gamitrinibs and SMTIN-P01 possess potent anti-cancer effects, their clinical application may be limited. Compound 5 exhibits moderated TRAP1 selectivity compared with cytosolic HSP90, the selectivity profile should be improved in further modification. HSP90 C-terminal inhibition doesn’t induce heat shock response, is another promising HSP90 inhibition strategy. Most reported HSP90 C-terminal inhibitors are natural product analogues, and their accurate binding modes are still unclear. In future study, more efforts should be done to reveal the binding modes of the C-terminal inhibitors. This will promote the development of HSP90 C-terminal inhibitors. HSP90-cochaperone PPI inhibition is an attractive strategy to inhibit the HSP90 chaperone function. CDC37 is a cochaperone which specifically regulate the maturation of kinase client proteins. Theoretically, HSP90-cochaperone PPI inhibition will only block the maturation of kinase clients and have no influence on other client proteins. Thus, HSP90-CDC37 PPI inhibitors can reduce the mechanism related adverse effects of the pan-HSP90 inhibitors. All previously reported HSP90-CDC37 PPI inhibitors incorporate a covalent-binding moiety and the HSP90-CDC37 PPI inhibition activity is covalent-binding based. Thus, rational designed HSP90-CDC37 PPI inhibitors are urgently needed. Our group have explored the hot-spots of the HSP90-CDC37 interface and discovered compound 21 as the first small-molecule HSP90-CDC37 inhibitor. 21 manifests moderate HSP90-CDC37 PPI blocking activity, and the activity should be further improved. Moreover, the anti-cancer effects of HSP90-CDC37 PPI inhibitors should be further studied. In general, HSP90 chaperone function inhibition is a promising anti-cancer strategy, but there is still a long way to go.