A scaffold merging approach to Hsp90 C-terminal inhibition: synthesis and evaluation of a chimeric library

Two previously identified Hsp90 C-terminal inhibitors were merged into a single scaffold that manifested improved Hsp90 inhibitory activity.

Abstract

Inhibition of the Hsp90 C-terminus is an attractive therapeutic paradigm for the treatment of cancer, however the developmental space of C-terminal inhibitors is limited. It was hypothesized that the combination of two previously identified scaffolds into a single structure could provide a platform for which to probe the three-dimensional space within the Hsp90 C-terminal binding pocket. The resulting chimeric compounds displayed anti-proliferative activity at low micromolar concentrations and manifested inhibitory activity in an Hsp90-dependent rematuration assay. Initial structure–activity relationships suggest that this new scaffold binds Hsp90 in a conformation different from that of the parent compounds, and consequently, provides a new opportunity to develop more efficacious inhibitors of the Hsp90 C-terminal binding pocket.

The 90 kDa heat shock protein (Hsp90) is a highly conserved and widely expressed molecular chaperone that plays a role in the maturation and stabilization of more than 200 protein substrates (clients).1 Hsp90 is upregulated in cancer to meet the demands of rapidly growing cells that reside in the harsh microenvironment of a tumor.2 Additionally, many of the Hsp90-dependent client proteins (e.g. Akt, Her2, Raf-1, etc.) are associated with cellular growth and signaling pathways that become dysregulated in cancer and represent well-validated therapeutic targets.3,4 Therefore, inhibition of the Hsp90 folding machinery provides a unique opportunity to simultaneously disrupt multiple oncogenic pathways, and consequently, represents an attractive target to combat the many pathologies exhibited by cancer.5

Within the cell, Hsp90 exists as a homodimer with each monomer possessing both an N-terminal ATPase domain, as well as a C-terminal dimerization domain that contains a second nucleotide-binding pocket.6 In the past decade, 17 inhibitors that target the N-terminus of Hsp90 have undergone clinical evaluation.7 However, N-terminal inhibitors induce both the pro-apoptotic degradation of client proteins and the pro-survival heat shock response at similar concentrations.8 The combination of these effects most often results in cytostatic activity and limits the potential utility of N-terminal inhibitors in the clinic.9 In contrast, inhibitors of the Hsp90 C-terminus can segregate these effects: cytotoxic C-terminal inhibitors promote client protein degradation without induction of the heat shock response,10,11 whereas cytoprotective C-terminal inhibitors induce the heat shock response at concentrations well below that which is needed to induce the degradation of client proteins.12,13 Despite the potential advantages offered by Hsp90 C-terminal inhibitors, they have yet to undergo clinical investigation. The development of more efficacious analogs has been hampered by the lack of a co-crystal structure between the C-terminal nucleotide-binding pocket and an inhibitor.

In 2000, Neckers and coworkers discovered the antibiotic novobiocin (Fig. 1) as the first inhibitor of the Hsp90 C-terminal binding region.14,15 Subsequent SAR studies transformed novobiocin from a DNA gyrase inhibitor into a selective inhibitor of Hsp90 (1).16,17 However, the coumarin core of novobiocin is not amenable to divergent modification and as a result, considerable effort has been placed on the identification of new scaffolds that inhibit the Hsp90 C-terminus.18–21 Prior work has led to the discovery of 2,11 which features a biaryl moiety and a flexible linker attached to the amide side chain, and 3,22 which contains a biphenyl core that serves as a coumarin surrogate. In an attempt to improve potency and expand the chemical space associated with Hsp90 C-terminal inhibition, it was envisioned that these two scaffolds could be combined into a single compound, which would exhibit the interactions manifested by both 2 and 3. In this communication, we report the design, synthesis, and evaluation of such chimeric analogs.

Fig. 1. Hsp90 C-terminal inhibitors.

Retrosynthetic analysis of triphenyl scaffold 4 suggested 5 as a common intermediate that was amenable to late-stage diversification (Scheme 1). Mitsunobu etherification of alcohol 6 with phenol 5 would install the piperidine ring and subsequent coupling of the aniline with acid chloride 7 would produce the requisite amide side chain. The triphenyl core of 5 could be rapidly assembled through sequential Suzuki coupling reactions between bromoresorcinol 8 and phenylboronic acids, 9 and 10.

Scheme 1. Retrosynthetic analysis of triphenyl analogs.

Synthesis of the proposed analogs commenced upon preparation of 1-benzyloxy-4-bromoresorcinol (8, Scheme 2). Formation of the benzyl ether from 4-bromoresorcinol (11) and benzyl bromide gave the undesired regioisomer 12 as the major product in low yield. However, treatment of 11 with p-toluenesulfonyl chloride resulted in chemoselective formation of the para sulfonate ester.23 Subsequent masking of the ortho phenol with chloromethyl methyl ether gave bromoresorcinol 13, which upon base-catalyzed hydrolysis of the sulfonate ester and installation of the benzyl ether gave 14. Ultimately, the bis-protected bromoresorcinol 14 was subjected to acid-catalyzed hydrolysis of the methoxymethyl acetal to furnish 8 in 81% yield over 4 steps. Suzuki coupling of 8 with 3- or 4-(tert-butoxycarbonylamino)phenylboronic acid (9) gave carbamate-protected anilines, 15a–b. Microwave-assisted conversion of the hindered phenol into a trifluoromethanesulfonate ester and subsequent Suzuki coupling24 with meta-substituted phenylboronic acids 10 yielded the triphenyl compounds, 16a–d. Hydrogenolysis of the benzyl ethers followed by Mitsunobu etherification with 1-methyl-4-hydroxypiperidine yielded 17a–d. Cleavage of the carbamate with trifluoroacetic acid yielded common intermediates 18a–d, which were subsequently coupled with acid chlorides (7a–f) to afford the final products, 4a–x.

Scheme 2. Synthesis of triphenyl analogs 4a–x. Reagents and conditions: a) BnBr, NaHCO₃, CH₃CN, 80 °C, 31%; b) TsCl, K₂CO₃, acetone, 65 °C, then MOMCl, 98%; c) 4 M KOH, EtOH, 60 °C, 99%; d) BnBr, K₂CO₃, acetone, 65 °C, 92%; e) HCl (conc.), THF/MeOH, rt, 91%; f) 3- or 4-(tert-butoxycarbonylamino)phenylboronic acid, Pd(dpppf)Cl₂, K₂CO₃, dioxane, 85 °C, 70–90%; g) N-phenyl-bis(trifluoromethanesulfonimide), K₂CO₃, THF, mw 120 °C; h) 3-(trifluoromethyl)phenylboronic acid or 3-fluorophenylboronic acid, SPhos, Pd₂(dba)₃, K₃PO₄, dioxane, 85 °C, 70–95% over two steps; i) H₂, Pd/C, THF, rt; j) 1-methyl-4-hydroxypiperidine, PPh₃, DIAD, THF, rt, 35–65% over two steps; k) TFA, DCM, rt, quant.; l) Et₃N, DCM, rt, 10–90%.

Upon construction of the library, the chimeric compounds were evaluated for their anti-proliferative activity against MCF7 and SKBR3 breast cancer cell lines. As shown in Table 1, the new chimeric scaffold was efficacious against both cell lines, and manifested EC₅₀ values in the low micromolar range. In general, compounds with trifluoromethyl and fluorine substitutions exhibited similar efficacies (e.g.4avs.4b), which correlated with previously reported SAR trends for 2.11 In contrast to SAR generated for the parent scaffolds,11,22,25,26 neither the location of the amide side chain (meta vs. para) nor its substitutions were drivers for potency (e.g.4qvs.4r and 4jvs.4n, respectively), which suggests the new scaffold may not interact with Hsp90 in the same orientation as the lead compounds.

Table 1. Anti-proliferative activities of 4a–x.

Cmpd	NHCOR	R	R1	MCF7 a (EC50, μM)	SKBR3 a (EC50, μM)
2	—	—	—	2.7 ± 0.40	3.5 ± 0.47
3	—	—	—	1.6 ± 0.15	1.5 ± 0.15
4a	meta	A	CF3	2.3 ± 0.75	1.9 ± 0.66
4b	meta	A	F	4.2 ± 0.21	4.1 ± 0.14
4c	para	A	CF3	2.2 ± 0.96	1.8 ± 0.80
4d	para	A	F	2.1 ± 0.35	4.2 ± 0.12
4e	meta	B	CF3	3.0 ± 0.58	2.9 ± 0.80
4f	meta	B	F	4.4 ± 0.10	3.3 ± 0.98
4g	para	B	CF3	2.6 ± 0.96	2.3 ± 0.95
4h	para	B	F	3.3 ± 0.96	3.2 ± 0.99
4i	meta	C1	CF3	3.9 ± 0.69	3.0 ± 0.43
4j	meta	C1	F	5.6 ± 0.48	5.9 ± 0.97
4k	para	C1	CF3	3.4 ± 0.41	3.5 ± 0.59
4l	para	C1	F	5.1 ± 0.13	5.2 ± 0.21
4m	meta	C2	CF3	2.8 ± 0.22	1.8 ± 0.54
4n	meta	C2	F	5.2 ± 0.25	6.2 ± 0.31
4o	para	C2	CF3	3.0 ± 0.10	2.4 ± 0.51
4p	para	C2	F	4.2 ± 0.10	4.2 ± 0.11
4q	meta	D1	CF3	1.7 ± 0.87	2.2 ± 0.70
4r	para	D1	CF3	2.9 ± 0.99	2.2 ± 0.95
4s	meta	D2	CF3	2.0 ± 0.74	1.9 ± 0.72
4t	para	D2	CF3	1.0 ± 0.21	0.8 ± 0.36
4u	meta	D3	CF3	4 ± 1.2	4 ± 1.1
4v	para	D3	CF3	0.9 ± 0.18	0.5 ± 0.14
4w	meta	D4	CF3	3 ± 1.2	3 ± 1.4
4x	para	D4	CF3	5 ± 1.3	4.4 ± 0.87

^aValues represent the mean ± SD of at least three separate experiments performed in triplicate.

In light of the relatively flat SAR trends observed for this chimeric scaffold in the anti-proliferation assay, the compounds were evaluated in an Hsp90-dependent luciferase rematuration assay. Briefly, PC3-MM2 cells expressing firefly luciferase were heated to denature luciferase before incubation with either DMSO or a proposed Hsp90 inhibitor.27 After incubation, luciferase activity was measured to determine whether the compounds inhibited the refolding of luciferase, a process dependent upon Hsp90.28 As shown in Fig. 2A, analogs containing the p-bromo substitution were the most efficacious of the benzamide-containing compounds, regardless of whether the amide was in the meta or para position (4i–l). In contrast to the benzamide-containing compounds, the triazolamide-containing compounds (4q–x, Fig. 2B) exhibited a preference for meta-substitution. Additionally, with the exception of t-butyl-substituted analogs, 4w–x, the triazolamide-containing compounds were more efficacious than the benzamide-containing compounds. Together, these data suggest the triazole ring may participate in hydrogen bonding interactions that require substitution at the meta position.

Fig. 2. Inhibition of heat-denatured luciferase refolding. Values are RLU relative to DMSO and represent the mean ± SD of at least two separate experiments performed in triplicate. (A) Benzamide-containing analogs 4a–p. (B) Parents 2 and 3 and triazolamide-containing analogs 4q–r.

Interestingly, there were little trends between the anti-proliferative and luciferase rematuration assays. Indeed, despite the entire library manifesting good anti-proliferative activity, only a subset inhibited the Hsp90 protein-folding machinery. Regardless, p-bromo benzamides 4i–l and meta-substituted triazolamides 4q, 4s, and 4u displayed increased inhibition of luciferase rematuration compared to parent compounds 2 and 3, validating the potential use of this scaffold as a starting point for the development of more efficacious inhibitors.

To further characterize the most efficacious compounds, the levels of Hsp90-dependent client proteins from MCF-7 cells treated with compounds 4q, 4i, and 4k were examined by western blot analyses. As shown in Fig. 3, the level of Hsp90-dependent client proteins (HER2, CDK4, ER-α, and cyclin D1) decreased in a concentration-dependent manner, while the levels of actin, which is not dependent upon Hsp90, remained constant. The levels of Hsp90 and Hsp70 also remained constant, indicating that these compounds did not induce the heat shock response, a hallmark manifested by Hsp90 C-terminal inhibitors that exhibit anti-cancer activity. Compounds 4i and 4k appear to be more efficacious than 4q at inducing client protein degradation, despite exhibiting a similar potency in the luciferase rematuration assay.

Fig. 3. Western blot of Hsp90-dependent client proteins from MCF7 cell lysate upon treatment with triphenyl compounds. L represents a concentration equal to 0.5-fold of the anti-proliferative EC₅₀. H represents a concentration equal to 10-fold of the anti-proliferative EC₅₀.

In summary, a library of chimeric small molecules was designed, synthesized, and biologically evaluated in Hsp90-dependent cancer cell lines. The anti-proliferative activities manifested by the triphenyl analogs against both MCF7 and SKBR3 cells were similar to the parent compounds. The ability of select analogs to inhibit the refolding of denatured luciferase in PC3-MM2 cells confirm these molecules inhibit the Hsp90 protein folding machinery. In addition, the degradation of Hsp90-dependent client proteins treated with 4q, 4i, and 4k occurred without induction of the heat shock response, which is consistent with Hsp90 C-terminal inhibition. The ease with which this new triphenyl scaffold can be rapidly constructed and divergently modified provides an excellent opportunity to explore the C-terminal binding pocket and to develop more efficacious inhibitors. Such studies are currently underway and will be reported in due course.