Structural Basis for Design of New Purine-Based Inhibitors Targeting the Hydrophobic Binding Pocket of Hsp90

Inhibition of the molecular chaperone heat shock protein 90 (Hsp90) represents a promising approach for cancer treatment. BIIB021 is a highly potent Hsp90 inhibitor with remarkable anticancer activity; however, its clinical application is limited by lack of potency and response. In this study, we aimed to investigate the impact of replacing the hydrophobic moiety of BIIB021, 4-methoxy-3,5-dimethylpyridine, with various five-membered ring structures on the binding to Hsp90. A focused array of N7/N9-substituted purines, featuring aromatic and non-aromatic rings, was designed, considering the size of hydrophobic pocket B in Hsp90 to obtain insights into their binding modes within the ATP binding site of Hsp90 in terms of π–π stacking interactions in pocket B as well as outer α-helix 4 configurations. The target molecules were synthesized and evaluated for their Hsp90α inhibitory activity in cell-free assays. Among the tested compounds, the isoxazole derivatives 6b and 6c, and the sole six-membered derivative 14 showed favorable Hsp90α inhibitory activity, with IC50 values of 1.76 µM, 0.203 µM, and 1.00 µM, respectively. Furthermore, compound 14 elicited promising anticancer activity against MCF-7, SK-BR-3, and HCT116 cell lines. The X-ray structures of compounds 4b, 6b, 6c, 8, and 14 bound to the N-terminal domain of Hsp90 were determined in order to understand the obtained results and to acquire additional structural insights, which might enable further optimization of BIIB021.


Introduction
Heat shock protein 90 (Hsp90) is an adenosine triphosphate (ATP)-reliant molecular chaperone that plays a fundamental role in preserving protein homeostasis by regulating the correct folding,

Introduction
Heat shock protein 90 (Hsp90) is an adenosine triphosphate (ATP)-reliant molecular chaperone that plays a fundamental role in preserving protein homeostasis by regulating the correct folding, intracellular disposition, and activity of numerous client proteins [1][2][3]. A significant number of Hsp90 client proteins are oncoproteins, such as receptor tyrosine kinases; human epidermal growth factor receptor 2 (Her2), epidermal growth factor receptor (EGFR), mesenchymal epithelial transition factor receptor (Met), rapidly accelerated fibrosarcoma-C (c-Raf), cell cycle regulators cyclin dependent kinase 4 (Cdk4), and hypoxia-inducible factor 1α (HIF-1α), which are necessary for the malignant transformation and tumor development [4,5]. Under the inimical environment in cancers such as reduced oxygenation and nutrition deprivation, Hsp90 is generally overexpressed 2-10-fold higher than that in normal cells, and forms activated complexes with its clients in cancer cells [6,7]. Therefore, cancer cells are highly dependent on Hsp90 chaperoning function for their proliferation and survival. Moreover, inhibition of Hsp90 chaperone machinery results in proteasomal degradation of various client oncoproteins that are implicated in multiple signaling pathways in cancers, offering merit of circumventing the common drug resistance of classical chemotherapies [8]. In this regard, Hsp90 inhibition represents a legitimate therapeutic approach for the treatment of cancers [3].
Over the past few years, a plethora of Hsp90 inhibitors endowed with anticancer activity has been developed [9][10][11][12][13][14]. Despite the remarkable clinical efficacies of some Hsp90 inhibitors, they were associated with potential off-target and/or HSP90-relevant toxicities, which hampered their further therapeutic application [15][16][17]. For example, tanespimycin (17-AAG) [18] and alvespimycin   [19] (Figure 1), the first-generation geldanamycin analogs, have displayed significant hepatotoxicity, which might stem from their quinone structure [20]. Moreover, visual disorders such as night blindness and blurred vision were reported for some chemically diverse second-generation HSP90 inhibitors such as ganetespib (STA-9090) [21,22]. Moreover, the highly potent antitumor purine derivative BIIB021 [23] (Figure 1) showed a high maximum tolerated dose (MTD), but lacked potency and prolonged response, requiring higher doses to achieve biological effects in clinical trials [24]. These clinically reported drawbacks of Hsp90 inhibitors attracted the curiosity of medicinal chemists, leading them to develop new chemical entities in an attempt to achieve comparable potency, improved target selectivity, and better safety profile. The majority of investigated HSP90 inhibitors are ATP-competitive, occupying the ATP binding pocket at the N-terminal of HSP90 [17]. Guided by the computational analysis of the co-crystal structure/docking investigations of Hsp90 inhibitors, Taldone et al. categorized them on the basis of their interactions with three essential N-terminal binding regions: the ATP-binding pocket A, hydrophobic pocket B, and a solvent-exposed exit pocket C [25]. Besides these three major binding pockets, it was reported that the α-helix 4 region (residues 99-125) in the N-terminal domain of Hsp90 could show ligand-induced conformational changes that contribute to the Hsp90 selective inhibition (Figure 2A,B). The structurally diverse scaffolds of Hsp90 inhibitors showed distinct ligand-induced conformational changes. While most Hsp90 inhibitors such as 17-DMAG and radicicol adopt the open-loop conformation, both purines BIIB021 and PU-H71 adopted the extended α-helix 4 The majority of investigated HSP90 inhibitors are ATP-competitive, occupying the ATP binding pocket at the N-terminal of HSP90 [17]. Guided by the computational analysis of the co-crystal structure/docking investigations of Hsp90 inhibitors, Taldone et al. categorized them on the basis of their interactions with three essential N-terminal binding regions: the ATP-binding pocket A, hydrophobic pocket B, and a solvent-exposed exit pocket C [25]. Besides these three major binding pockets, it was reported that the α-helix 4 region (residues 99-125) in the N-terminal domain of Hsp90 could show ligand-induced conformational changes that contribute to the Hsp90 selective inhibition (Figure 2A,B). The structurally diverse scaffolds of Hsp90 inhibitors showed distinct ligand-induced conformational changes. While most Hsp90 inhibitors such as 17-DMAG and radicicol adopt the open-loop conformation, both purines BIIB021 and PU-H71 adopted the extended α-helix 4 conformation outside of pocket B [26,27]. These insights might aid in optimizing scaffolds to improve target selectivity, reducing the time and cost required to develop novel inhibitors. conformation outside of pocket B [26,27]. These insights might aid in optimizing scaffolds to improve target selectivity, reducing the time and cost required to develop novel inhibitors. Among the arsenal of Hsp90 inhibitors worthy for further structural exploration, BIIB021, the first orally bioavailable inhibitor, is of special interest. BIIB021 exerted potent anticancer efficacy in the N87 gastric tumor model along with its feasible straightforward synthesis [23]. Furthermore, BIIB021 exhibited a strong antitumor effect against various head and neck squamous cell carcinoma (HNSCC) models, with a remarkable ability to sensitize those HNSCC xenografts to radiotherapy [28]. Moreover, BIIB021 has elicited broad application against cancers with acquired multidrug resistance [29]. The X-ray crystal structure of the Hsp90-BIIB021 complex (PDB code: 3QDD) revealed that the 2-amino and 9-pyridyl moieties of purine BIIB021 bind to adenine binding pocket A and the hydrophobic pocket B, respectively ( Figure 2C) [23]. The pyridine ring in BIIB021 was involved in multiple hydrophobic interactions including π-π stacking with the phenyl ring of Phe138, π-σ stacking with isopropyl of Leu107, as well as π-edge contacts with the ring structures of Tyr139 and Trp162. Guided by the structural data, the 7-deazapurine derivative EC144 was developed, where the N7 of BIIB021 was replaced with a carbon atom substituted with 2-methylpent-4-yn-2-ol [26], while the hydrophobic binding moiety, 4-methoxy-3,5-dimethylpyridine, was conserved. Such a study provided EC144 with approximately 20-fold higher efficacy than BIIB021 in mice [26]. However, the impact of structural variation of the 9-pyridine moiety with simple five-membered heterocyclic structures has not been investigated.
To fulfill this structure-activity relationship (SAR) gap, and as a continuation of our efforts to identify potent anticancer chemotherapeutics [11,[30][31][32][33][34][35], we were motivated to synthesize a focused library of N 7 /N 9 -substituted purines. In this study, we focused on the hydrophobic binding pocket B, a newly formed binding pocket of Hsp90ND created by the conformational change. In this line, we conducted a structural exploration study to highlight the tolerability of changing the 4-methoxy-3,5dimethylpyridine of BIIB021 with simpler aromatic and non-aromatic five-ring structures, such as isoxazoline, isoxazole, oxazole, and triazole featuring bulky substituents (ethyl carboxylate, ter-butyl, isopropyl, phenyl) ( Figure 3). We hypothesized that such designed molecules possessing nitrogen and/or oxygen atoms might offer binding mode(s) that are different from that of BIIB021, a variable degree of π-π stacking interactions with pocket B, and their terminal bulky moieties might result in a different configuration of the outer α-helix 4. The target molecules were synthesized and evaluated Among the arsenal of Hsp90 inhibitors worthy for further structural exploration, BIIB021, the first orally bioavailable inhibitor, is of special interest. BIIB021 exerted potent anticancer efficacy in the N87 gastric tumor model along with its feasible straightforward synthesis [23]. Furthermore, BIIB021 exhibited a strong antitumor effect against various head and neck squamous cell carcinoma (HNSCC) models, with a remarkable ability to sensitize those HNSCC xenografts to radiotherapy [28]. Moreover, BIIB021 has elicited broad application against cancers with acquired multidrug resistance [29]. The X-ray crystal structure of the Hsp90-BIIB021 complex (PDB code: 3QDD) revealed that the 2-amino and 9-pyridyl moieties of purine BIIB021 bind to adenine binding pocket A and the hydrophobic pocket B, respectively ( Figure 2C) [23]. The pyridine ring in BIIB021 was involved in multiple hydrophobic interactions including π-π stacking with the phenyl ring of Phe138, π-σ stacking with isopropyl of Leu107, as well as π-edge contacts with the ring structures of Tyr139 and Trp162. Guided by the structural data, the 7-deazapurine derivative EC144 was developed, where the N7 of BIIB021 was replaced with a carbon atom substituted with 2-methylpent-4-yn-2-ol [26], while the hydrophobic binding moiety, 4-methoxy-3,5-dimethylpyridine, was conserved. Such a study provided EC144 with approximately 20-fold higher efficacy than BIIB021 in mice [26]. However, the impact of structural variation of the 9-pyridine moiety with simple five-membered heterocyclic structures has not been investigated.
To fulfill this structure-activity relationship (SAR) gap, and as a continuation of our efforts to identify potent anticancer chemotherapeutics [11,[30][31][32][33][34][35], we were motivated to synthesize a focused library of N 7 /N 9 -substituted purines. In this study, we focused on the hydrophobic binding pocket B, a newly formed binding pocket of Hsp90ND created by the conformational change. In this line, we conducted a structural exploration study to highlight the tolerability of changing the 4-methoxy-3,5-dimethylpyridine of BIIB021 with simpler aromatic and non-aromatic five-ring structures, such as isoxazoline, isoxazole, oxazole, and triazole featuring bulky substituents (ethyl carboxylate, ter-butyl, isopropyl, phenyl) ( Figure 3). We hypothesized that such designed molecules possessing nitrogen and/or oxygen atoms might offer binding mode(s) that are different from that of BIIB021, a variable degree of π-π stacking interactions with pocket B, and their terminal bulky moieties might result in a different configuration of the outer α-helix 4. The target molecules were synthesized and evaluated for their Hsp90α inhibitory activity in cell-free assays. Moreover, the X-ray crystal structures of several members of our target compounds with Hsp90 were determined to obtain reliable structural insights that would lead to a better understanding of the compound's binding mode, therefore paving the way for proper further optimization. for their Hsp90α inhibitory activity in cell-free assays. Moreover, the X-ray crystal structures of several members of our target compounds with Hsp90 were determined to obtain reliable structural insights that would lead to a better understanding of the compound's binding mode, therefore paving the way for proper further optimization.

Chemical Synthesis
The novel target compounds were synthesized from 6-chloro-purin-2-amine 2, which was generated via chlorination of the imine-protected nucleobase guanine 1 using dimethylformamide (DMF) [36] (Scheme 1). N-allylation of 2 using allyl bromide provided the N 9 -allyl regioisomers 3 in 50% yield. 1,3-Dipolar cycloaddition [37] of the allyl derivative 3 with the appropriate oxime generated the corresponding isoxazolines 4a and 4b. for their Hsp90α inhibitory activity in cell-free assays. Moreover, the X-ray crystal structures of several members of our target compounds with Hsp90 were determined to obtain reliable structural insights that would lead to a better understanding of the compound's binding mode, therefore paving the way for proper further optimization.

Assessment of Hsp90α Binding Affinity Using FP Assay
First, the fluorescence polarization (FP) assay, a well-validated method for high-throughput screening [39], was adopted for determining the binding affinities. The FP assay is based on competition for the ATP-binding pocket of Hsp90α by a test compound and a fluorescently labeled geldanamycin (GM) probe such as the synthesized green or red boron-dipyrromethene (BODIPY)-GM [40] and fluorescein isothiocyanate labeled geldanamycin (FITC-GM) [41,42]. All compounds were tested in a duplicate mode against N-terminal human Hsp90α with BODIPY-GM to calculate their inhibitory constant Ki after 5 h treatment, which was found to be >1.9 µM. Next, we selected an array of compounds for measuring their binding affinity to Hsp90α using the FITC-GM probe at Reaction Biology Corporation (RBC, Malvern, PA, USA) [43]. BIIB021 and 17-AAG were used as references, and the results are summarized in Table 1. The obtained data revealed that the fivemembered ring derivatives exhibited rather weak binding affinities, presumably caused by the lower π-stacking interaction force of the five-membered ring compared to the six-membered ring. The 3tert-butyl isoxazole 6b showed the best binding affinity (95.8% inhibition) among the five-membered ring members. Moreover, 6b elicited a 2.7-fold superior potency to the non-aromatic isoxazoline Synthetic pathway for compounds 10-14.

Assessment of Hsp90α Binding Affinity Using FP Assay
First, the fluorescence polarization (FP) assay, a well-validated method for high-throughput screening [39], was adopted for determining the binding affinities. The FP assay is based on competition for the ATP-binding pocket of Hsp90α by a test compound and a fluorescently labeled geldanamycin (GM) probe such as the synthesized green or red boron-dipyrromethene (BODIPY)-GM [40] and fluorescein isothiocyanate labeled geldanamycin (FITC-GM) [41,42]. All compounds were tested in a duplicate mode against N-terminal human Hsp90α with BODIPY-GM to calculate their inhibitory constant Ki after 5 h treatment, which was found to be >1.9 µM. Next, we selected an array of compounds for measuring their binding affinity to Hsp90α using the FITC-GM probe at Reaction Biology Corporation (RBC, Malvern, PA, USA) [43]. BIIB021 and 17-AAG were used as references, and the results are summarized in Table 1. The obtained data revealed that the five-membered ring derivatives exhibited rather weak binding affinities, presumably caused by the lower π-stacking interaction force of the five-membered ring compared to the six-membered ring. The 3-tert-butyl isoxazole 6b showed the best binding affinity (95.8% inhibition) among the five-membered ring members. Moreover, 6b elicited a 2.7-fold superior potency to the non-aromatic isoxazoline congener 4b (35.4% inhibition). Furthermore, compound 6b showed an IC 50 value of 1.76 µM. Replacing the 3-tert-butyl moiety of isoxazole with less bulky isopropyl showed comparable inhibitory affinity (6c, 77.3% inhibition). However, this compound 6c displayed an affinity twofold higher than 4b and exerted a considerable Hsp90α inhibitory effect with an IC 50 value of 0.203 µM. Inspection of the more extended propionate ester derivatives such as isoxazoline 4a and isoxazole 6a disclosed their low potency (less than 15% inhibition). The phenyl oxazole 8, featuring a lower rotational group, displayed an affinity (48.7% inhibition) twofold higher than the ester derivatives of isoxazole. Compound 9, the positional isomer of compound 8, as well as the p-tolyl acetylene 11 and triazole derivatives 12 and 13 displayed a fluorescence background that interfered with the assay. Interestingly, the sole six-membered ring derivative 14, bearing 4-methylbenzyl moiety, displayed a favorable binding affinity with 95.4% inhibition at the tested dose of 10 µM. Further assay of 14 revealed its promising IC 50 value of 1.00 µM. Replacing the 3-tert-butyl moiety of isoxazole with less bulky isopropyl showed comparable inhibitory affinity (6c, 77.3% inhibition). However, this compound 6c displayed an affinity twofold higher than 4b and exerted a considerable Hsp90α inhibitory effect with an IC50 value of 0.203 µM. Inspection of the more extended propionate ester derivatives such as isoxazoline 4a and isoxazole 6a disclosed their low potency (less than 15% inhibition). The phenyl oxazole 8, featuring a lower rotational group, displayed an affinity (48.7% inhibition) twofold higher than the ester derivatives of isoxazole. Compound 9, the positional isomer of compound 8, as well as the p-tolyl acetylene 11 and triazole derivatives 12 and 13 displayed a fluorescence background that interfered with the assay. Interestingly, the sole six-membered ring derivative 14, bearing 4-methylbenzyl moiety, displayed a favorable binding affinity with 95.4% inhibition at the tested dose of 10 µM. Further assay of 14 revealed its promising IC50 value of 1.00 µM.

In Vitro Antiproliferative Activity
The in vitro anticancer activity of compounds 4b, 6b, 6c, 8, and 14 was evaluated against a panel of three human cancer cells expressing high levels of Hsp90-related client proteins: Her-2 dependent SK-BR-3 breast cancer, estrogen receptor positive (ER+) MCF-7 breast cancer, and EGFR-reliant colon (HCT116) cancer cell line, using BIIB021 and 17-AAG as references. The growth inhibitory effects of the test compounds, at 10 µM dose, were determined after 72 h incubation using the 3-(4,5dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assay. As revealed from the data

In Vitro Antiproliferative Activity
The in vitro anticancer activity of compounds 4b, 6b, 6c, 8, and 14 was evaluated against a panel of three human cancer cells expressing high levels of Hsp90-related client proteins: Her-2 dependent SK-BR-3 breast cancer, estrogen receptor positive (ER+) MCF-7 breast cancer, and EGFR-reliant colon (HCT116) cancer cell line, using BIIB021 and 17-AAG as references. The growth inhibitory effects of the test compounds, at 10 µM dose, were determined after 72 h incubation using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assay. As revealed from the data listed in Table 2, most of the examined compounds exerted modest to moderate cytostatic activity over the tested cancer cell lines. Compound 4b showed unpretentious antiproliferative activity against all examined cells, with growth inhibition (GI) less than 30%. Such modest cellular outcomes of those purine derivatives may be attributed to their lower lipophilic characters, which in turn restrict their cellular permeability into cancer cells. While comparing the activity of 3-substituted isoxazole 6b and 6c, we observed that the most hydrophobic isoxazole 6b, with tert-butyl moiety, showed reasonable activity against SK-BR-3 breast cancer (GI = 45.76%). The most active member in Hsp90α assay 6c elicited modest anticancer activity (SK-BR-3; GI = 33.17%, MCF-7; GI = 11.92%). The 2-phenyloxazole-substituted purine 8 exerted reasonable antiproliferative activity, with GI values of 33.27% and 37.46% against MCF-7 and SK-BR-3 cells, respectively. Interestingly, compound 14 showed good cellular activity against the tested cell lines, particularly the SK-BR-3 (GI = 60.93%). In view of GI 50 values, the phenyloxazole 8 showed a GI 50 value of 22.54 ± 0.70 µM towards SK-BR-3. The p-tolyl derivatives 14 exhibited equipotent activity against the breast cancer cells MCF-7 and SK-BR-3, with GI 50 values of 16.22 ± 0.71 µM and 15.24 ± 2.08 µM, respectively. However, all tested derivatives were less potent than the reference compounds BIIB021 and 17-AAG. The binding affinity of the compounds to the N-terminal domain of Hsp90α was measured using the isothermal titration calorimetry (ITC). Some of the tested compounds were not as soluble in the buffer solution, but we were able to analyze the binding affinities for compounds 4b, 6b, 6c, and 14. The results are depicted in Figure 4 and Table 3. They all showed a 1:1 molar binding to the target protein, and the K D value for 17-AAG is comparable to the value reported earlier [44]. The five-membered ring isoxazole derivative 6b, possessing a tert-butyl group, displayed a good binding affinity (K D = 0.44 µM) as well as the isoxazole 6c, featuring an isopropyl moiety with a comparable binding affinity (K D = 0.68 µM). In contrast, compound 4b, the isoxazoline analog of 6b, manifested only a modest activity (K D = 15.6 µM). Interestingly, 4-methylphenyl compound 14, the only six-membered ring derivative, showed the best affinity (K D = 0.41 µM), similar to isoxazoles 6b and 6c.

Colorimetric Determination of ATPase Activity of Hsp90
Since the intrinsic ATPase activity of Hsp90 is pivotal for its chaperone function and as a part of our study, we tested the tendency of compound 14, which exhibited favorable binding affinity with Hsp90, to inhibit Hsp90 by measuring the ATPase activity of Hsp90 along with BIIB021 as a known inhibitor.
As revealed from the data (Figure 5), the ATPase activity of Hsp90 was decreased upon incubation with compound 14 in a pattern comparable to that observed with BIIB021. This finding might emphasize that the binding of compound 14 to Hsp90 would result in inhibition of its ATPase activity.

Western Blot Analysis of HSP90 Key Client Protein Expression and Activation
Furthermore, the effect of compound 14 on cellular biomarkers of the HSP90 inhibition was evaluated to obtain insights about the molecular mechanisms triggering its anticancer activity. MCF7 cells were incubated with various doses of compound 14 (0.5, 1, 5, 10, and 20 µM) for 24 h, and the expression levels of HSP90 client proteins (EGFR, Akt, and Cdk4), Hsp90, HSP70, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were determined by Western blot analysis ( Figure 6). Compound 14 displayed the characteristic signature of Hsp90 inhibitors, where it induced proteasomal degradation of HSP90 oncogenic client proteins (EGFR and Akt) and upregulated the expression level of Hsp70 in a dose-dependent manner, while the levels of GAPDH remained unchanged. The expression levels of EGFR and Akt were remarkably decreased upon the treatment of MCF7 cells with 10 µM and 20 µM of compound 14, respectively. On the basis of these outcomes, it could be suggested that compound 14 can disrupt the HSP90 chaperoning machinery in MCF7 cells.

Structural Analysis of the Complexes Using X-ray Crystallography
To understand the binding of these compounds at the atomic level, we determined the crystal structures of compounds 4b, 6b, 6c, 8, and 14 complexed to the N-terminal domain of mouse Hsp90

Western Blot Analysis of HSP90 Key Client Protein Expression and Activation
Furthermore, the effect of compound 14 on cellular biomarkers of the HSP90 inhibition was evaluated to obtain insights about the molecular mechanisms triggering its anticancer activity. MCF7 cells were incubated with various doses of compound 14 (0.5, 1, 5, 10, and 20 µM) for 24 h, and the expression levels of HSP90 client proteins (EGFR, Akt, and Cdk4), Hsp90, HSP70, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were determined by Western blot analysis ( Figure 6).

Western Blot Analysis of HSP90 Key Client Protein Expression and Activation
Furthermore, the effect of compound 14 on cellular biomarkers of the HSP90 inhibition was evaluated to obtain insights about the molecular mechanisms triggering its anticancer activity. MCF7 cells were incubated with various doses of compound 14 (0.5, 1, 5, 10, and 20 µM) for 24 h, and the expression levels of HSP90 client proteins (EGFR, Akt, and Cdk4), Hsp90, HSP70, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were determined by Western blot analysis ( Figure 6). To understand the binding of these compounds at the atomic level, we determined the crystal structures of compounds 4b, 6b, 6c, 8, and 14 complexed to the N-terminal domain of mouse Hsp90 Compound 14 displayed the characteristic signature of Hsp90 inhibitors, where it induced proteasomal degradation of HSP90 oncogenic client proteins (EGFR and Akt) and upregulated the expression level of Hsp70 in a dose-dependent manner, while the levels of GAPDH remained unchanged. The expression levels of EGFR and Akt were remarkably decreased upon the treatment of MCF7 cells with 10 µM and 20 µM of compound 14, respectively. On the basis of these outcomes, it could be suggested that compound 14 can disrupt the HSP90 chaperoning machinery in MCF7 cells.

Structural Analysis of the Complexes Using X-ray Crystallography
To understand the binding of these compounds at the atomic level, we determined the crystal structures of compounds 4b, 6b, 6c, 8, and 14 complexed to the N-terminal domain of mouse Hsp90 at the resolutions ranging from 1.33 to 1.75 Å. The final statistics on the data collection and refinement are summarized in Table 4. Table 4. Statistics on data collection and refinement of crystal structures of mHSP90ND complexed with compounds 4b, 6b, 6c, 8, and 14 The complex structures confirmed that compounds 4b, 6b, 6c, 8, and 14 are tightly bound to the ATP-binding site of Hsp90, with all atoms well defined in electron density maps ( Figure 7A-E). In all cases, the electron density map gave unambiguous positioning of the compounds as well as the protein atoms near the ligands and water molecules . Compounds 4b, 6b, 6c, 8, and 14 all showed an L-shape conformation, occupying the two binding pockets as expected. The purine ring is located above the β-sheet tightly bound in the binding A pocket, making extensive hydrophobic interactions with protein atoms. Moreover, it makes tight hydrogen bonds to Asp93 and two water molecules that are involved in an intricate hydrogen bond network with protein atoms. This interaction was found in all five structures reported here. The isoxazoline moiety of 4b, isoxazoles of 6b and 6c, and phenyloxazole of 8 bound to the hydrophobic pocket B are oriented in a parallel direction to the phenyl ring of Phe138, isopropyl of Leu107, and perpendicularly to the rings of Tyr139 and Trp162, as seen in the figure. Figure 7F illustrates the superposition of the five investigated compounds. Contrastingly to the purine ring, these aromatic rings showed certain variations in their orientation, with 4b and 8 in particular displaying significant differences. The ITC data (Table 3) of 4b suggested a distinct behavior from the rest. Interestingly, the angle between the purine ring and the isoxazoline in the case of 4b was only 99 • , while it ranged from 107 • to 110 • in others (Figure 8). The non-planar isoxazoline ring of 4b with a bulky substituent at position 3 made the angle between methylene substituents constrained in the hydrophobic pocket B by conserving π-π stacking interaction with phenyl ring of Phe138 without breaking outer helix 4. In the case of 8, the methylene atom connecting purine and phenyloxazole was moved outwards into the solvent by 0.5 Å compared to what was seen in others. However, we were not able to obtain the ITC data on 8 due to its low solubility. The phenyl ring in phenyloxazole of 8 is too bulky to slide sufficiently into the inside of the pocket B. In all five structures, the conformation of the protein atoms in the B pocket remained almost the same, suggesting that the B pocket is highly selective in terms of the tight binding and the rigid space.   Overall, these interactions are quite similar to that of BIIB021, as expected, i.e., the purine made almost identical interactions while the pyridine ring was oriented in a parallel direction to the phenyl ring of Phe138 (π-π stacking distance of 3.7 Å) and isopropyl of Leu107 (π-σ stacking distance of 3.9 Å) and perpendicularly to the rings of Tyr139 and Trp162 ( Figure 2C). In contrast, the crystal structures here pointed out that the isoxazoline derivative 4b, isoxazoles 6b and 6c, and phenyloxazole 8 showed less π stacking interactions, in a variable degree, with Phe138 and Leu107, due to either having lower aromaticity than the benzene ring (6b, 6c, and 8) or even lacking aromaticity entirely (4b).
Thus, these five-membered ring-featuring purines had a relatively less stable conformation than BIIB021. Isoxazoline 4b demonstrated a smaller angle (99°) of the methylene bond between the purine ring and the isoxazoline ring. Moreover, our structural data indicated that the various bulky substituents on heterocyclic rings did not alter the conservation of outer α-helix 4. Interestingly, it was found that the sole synthesized six-membered member 14 showed a similar binding pattern to that of BIIB021, where the p-tolyl moiety in 14 was packed against the side chain of Leu107 through π-σ stacking, in addition to the significant face-to-face stacking with Phe138 residue (π-π stacking distance of 3.8 Å) in the hydrophobic B pocket. Moreover, the tolyl structural domain was in a perpendicular direction to the rings of hydrophobic residues Tyr139 and Trp162, which further contributed to the stabilization of 14 within the ATP-binding pocket of mHsp90ND. In light of these results, it could be concluded that the inhibitor's conformation stability inside the active site pocket would influence its binding affinity and consequently the biological activity.

General Methods
All solvents and reagents were purchased from commercial suppliers and were used directly without purification. The reactions, applying the standard Schlenk techniques, were monitored using the thin layer chromatography (TLC) plate (Merck, silica gel 60 F254). The final compounds were purified by column chromatography using silica gel (Merck, 230-400 mesh) and the eluent solvents as indicated. The purity of all final compounds was >95%. 1 H and 13 C NMR charts were recorded on a Bruker Avance 400 or 300 MHz spectrometers (Bruker, Rheinstetten, Germany), using the appropriate deuterated NMR solvent. Chemical shifts (δ) are given in parts per million (ppm) upfield from the internal standard tetramethylsilane (TMS). The singlet, doublet, triplet, and multiplet peaks were indicated as s, d, t, and m, respectively, and the coupling constants (J) are reported in hertz (Hz). High-resolution mass spectra (HRMS) were recorded on either JMS 700 (Jeol, Tokyo, Japan) or Waters Acquity UPLC/Synapt G2 QTOF MS mass spectrometer (Waters, MA, USA). Overall, these interactions are quite similar to that of BIIB021, as expected, i.e., the purine made almost identical interactions while the pyridine ring was oriented in a parallel direction to the phenyl ring of Phe138 (π-π stacking distance of 3.7 Å) and isopropyl of Leu107 (π-σ stacking distance of 3.9 Å) and perpendicularly to the rings of Tyr139 and Trp162 ( Figure 2C). In contrast, the crystal structures here pointed out that the isoxazoline derivative 4b, isoxazoles 6b and 6c, and phenyloxazole 8 showed less π stacking interactions, in a variable degree, with Phe138 and Leu107, due to either having lower aromaticity than the benzene ring (6b, 6c, and 8) or even lacking aromaticity entirely (4b).

Synthesis and Structural Characterization
Thus, these five-membered ring-featuring purines had a relatively less stable conformation than BIIB021. Isoxazoline 4b demonstrated a smaller angle (99 • ) of the methylene bond between the purine ring and the isoxazoline ring. Moreover, our structural data indicated that the various bulky substituents on heterocyclic rings did not alter the conservation of outer α-helix 4. Interestingly, it was found that the sole synthesized six-membered member 14 showed a similar binding pattern to that of BIIB021, where the p-tolyl moiety in 14 was packed against the side chain of Leu107 through π-σ stacking, in addition to the significant face-to-face stacking with Phe138 residue (π-π stacking distance of 3.8 Å) in the hydrophobic B pocket. Moreover, the tolyl structural domain was in a perpendicular direction to the rings of hydrophobic residues Tyr139 and Trp162, which further contributed to the stabilization of 14 within the ATP-binding pocket of mHsp90ND. In light of these results, it could be concluded that the inhibitor's conformation stability inside the active site pocket would influence its binding affinity and consequently the biological activity.

General Methods
All solvents and reagents were purchased from commercial suppliers and were used directly without purification. The reactions, applying the standard Schlenk techniques, were monitored using the thin layer chromatography (TLC) plate (Merck, silica gel 60 F254). The final compounds were purified by column chromatography using silica gel (Merck, 230-400 mesh) and the eluent solvents as indicated. The purity of all final compounds was >95%. 1 H and 13 C NMR charts were recorded on a Bruker Avance 400 or 300 MHz spectrometers (Bruker, Rheinstetten, Germany), using the appropriate deuterated NMR solvent. Chemical shifts (δ) are given in parts per million (ppm) upfield from the internal standard tetramethylsilane (TMS). The singlet, doublet, triplet, and multiplet peaks were indicated as s, d, t, and m, respectively, and the coupling constants (J) are reported in hertz (Hz). High-resolution mass spectra (HRMS) were recorded on either JMS 700 (Jeol, Tokyo, Japan) or Waters Acquity UPLC/Synapt G2 QTOF MS mass spectrometer (Waters, MA, USA).

Colorimetric Determination of ATPase Activity of Hsp90
In brief, ATPase activity assay was performed by mixing Hsp90α solution (5 µL, 6 µM in assay buffer), 20 µL assay buffer, ATP solutions (10 µL, 4 mM in assay buffer), and 5 µL of tested compounds (14 and BIIB021) to yield final concentrations of 50 µM for 30 min at the room temperature using the QuantiChrom ATPase/GTPase Assay Kit (BioAssay Systems, Hayward, CA, USA). The reaction was terminated by adding 200 µL reagent buffer. Followed by incubation with reagent buffer for 30 min at the room temperature, the absorbance was measured at 620 nm using a Biotek-Synergy Neo-Plate Reader (BioTek Instruments, Winooski, VT, USA).

Western Blot Assay
MCF7 cells were seeded in 60 mm culture dishes, incubated overnight, and treated with the indicated concentrations of compound 14 or geldanamycin for 24 h. Cells were then harvested and lysed on ice by radioimmunoprecipitation assay (RIPA) buffer (23 mM Tris-HCl (pH 7.6), 130 mM NaCl, 1% nonidet P-40 (NP-40), 1% sodium deoxycholate, 0.1% SDS) supplemented with protease and phosphatase inhibitor cocktails, and 30 g of lysate perlane was separated by SDS-PAGE, followed by transferring to a polyvinylidene fluoride (PVDF) membrane (Bio-Rad, Hercules, CA, USA). The membrane was blocked with 5% skim milk in TBS-T, and then incubated with the corresponding primary antibody (EGFR, Akt, Hsp90, Hsp70, or GAPDH). After binding of an appropriate secondary antibody (Santa Cruz, CA, USA) coupled to horseradish peroxidase, proteins were visualized by enhanced chemiluminescence (ECL) according to the instructions of the manufacturer (GE healthcare, Chicago, IL, USA).

Crystallization and Data Collection
The purified protein of mHsp90ND was concentrated to 25 mg/mL in preparation for crystallization. Initial screening for the crystallization was carried out by using 96-well Intelli plates (Hampton Research) and Hydra II Plus One (MATRIX Technology Ltd, Toronto, Canada) robotics system at 295 K and optimized by the hanging-drop vapor diffusion in 24-well plates with a 2:1 mixture of protein solution and reservoir solution. Diffraction quality crystals of mHsp90ND in complex with 4b, 6b, 6c, 8, and 14 were obtained using the hanging-drop vapor diffusion method by mixing an equal volume of protein solution with reservoir solution. For the inhibitor ternary complex, the inhibitor was added at a 1:5 molar ratio and incubated for 1 h at 277 K prior to crystallization. Diffraction quality crystals of the inhibitor complex were obtained by mixing 25 mg/mL protein solution and a reservoir solution with 1.0-2.0 M ammonium sulfate and 0.1 M Tris-HCl (pH 8.5) at 22 • C, with the crystals appearing in 1 day. All diffraction datasets were collected at a temperature of 100 K on the DSC Quantum 315 CCD area

Conclusions
In this investigation, we report the synthesis and biological evaluation of a focused library of N 7 /N 9 -substituted purines as potential Hsp90 inhibitors. Our main objective was to explore the influence of replacing the 4-methoxy-3,5-dimethylpyridine, the hydrophobic binding moiety of BIIB021, with various simplified five-membered N/O-containing rings on π-π stacking interactions within pocket B of Hsp90 and the configuration of outer α-helix 4. Considering both biochemical and cell-based findings, the isoxazole derivatives 6b and 6c and the p-tolyl derivative 14 displayed promising Hsp90α inhibitory activity, with IC 50 values of 0.203-1.76 µM. Moreover, the p-tolyl containing purine 14 elicited favorable anticancer potency over the Hsp90-relevant cancer cells MCF-7, SK-BR-3, and HCT116. The crystal structures and ITC of mHsp90 N-terminal domain complexed with compounds 4b, 6b, 6c, 8, and 14 were determined, and their binding to the N-terminal domain of Hsp90 was pointed out. Furthermore, the structural findings revealed significance of conserving the aromaticity at N 9 -substituent of purine to make a proper angle between the purine ring and the aromatic ring (107 • -110 • ) and achieve good binding affinity and hence considerable Hsp90α inhibitory activity. It could be concluded that the conformations of the protein atoms in the hydrophobic B pocket of Hsp90 remain constant in its ligand complex structure by tight binding and conserving outer helix for their biological activity. Considering these biological and crystallographic findings, we plan to extend our efforts in a further structural optimization study of BIIB021 based on binding mode and conformation of ligands at hydrophobic pocket B.