Novel N-(4,5,6,7-tetrahydrobenzisoxazol-4-yl)amides as HSP90 inhibitors: design, synthesis and biological evaluation

Abstract

Novel N-(4,5,6,7-tetrahydrobenzisoxazol-4-yl)amide derivatives were designed and synthesized as potential HSP90 inhibitors. The synthetic pathway commenced with 6,7-dihydrobenzo[d]isoxazol-4(5H)-ones, utilizing the Ritter reaction as a key step. Molecular docking, molecular dynamics simulations, and MM/GBSA analysis guided the selection of compounds for synthesis and provided insights into the interaction mode of the most active compound with HSP90α. The synthesized compounds exhibited significant antiproliferative effects against breast cancer cell lines ERα+ MCF7 and HER2+ HCC1954. Lead compounds with submicromolar IC₅₀ values, initially synthesized as racemates, were subsequently obtained and tested in their enantiopure forms. In HER2+ HCC1954 cancer cells, the molecular pathways regulated by compound (R)-8n were characterized. Treatment with compound (R)-8n resulted in the pronounced suppression of HSP90-related pathways, including key oncoreceptors (HER2, EGFR, c-MET) and mitogenic kinases (AKT, CDK4). Additionally, compound (R)-8n induced apoptosis, as evidenced by the accumulation of cleaved PARP. The inhibitory effect of compound (R)-8n on the HSP90 pathway was corroborated by molecular modeling and further validated through the observed suppression of client proteins, along with an upregulation of HSP70, a well-established marker of HSP90 inhibition. The activity of compound (R)-8n was associated with cell cycle arrest at the G2/M phases, ultimately leading to dose-dependent cell death. Notably, compound (R)-8n demonstrated substantial selectivity toward breast tumor cells. These findings suggest that N-(4,5,6,7-tetrahydrobenzisoxazol-4-yl)amides represent a promising class of HSP90 inhibitors for anticancer therapy.

N-(4,5,6,7-Tetrahydrobenzisoxazol-4-yl)amides are discovered as novel Hsp90 inhibitors, which show high antiproliferative activity and selectivity toward breast tumor cells.

Introduction

Heat shock protein 90 (HSP90) is a molecular chaperone essential for maintaining proteostasis under both physiological and stress conditions. It plays a crucial role in the folding, maturation, degradation, and activation of client proteins.^1,2 HSP90 functions as a homodimer, with each monomer comprising an N-terminal domain (NTD), a charged flexible linker, a middle domain (MD), and a C-terminal domain (CTD). The NTD contains a nucleotide-binding site and a flexible loop segment known as the “lid.” The MD hosts most of the binding sites for client proteins, while the CTD is responsible for dimerization. Co-chaperones interact with all domains of HSP90, modulating its function.

The HSP90 chaperone cycle is generally described as follows: in its apo state, HSP90 adopts a V-shaped open conformation. Upon recruitment of a client protein, either independently or with the assistance of co-chaperones, a complex is formed between HSP90 and the client. ATP binding to the NTD induces translocation of the lid over the nucleotide-binding pocket,³ leading to a conformational shift into a closed, twisted structure. The energy released from ATP hydrolysis to ADP is then utilized for client protein processing. In the final stage, ADP and the matured client protein are released, restoring HSP90 to its V-shaped open conformation.

To this day, several hundred proteins interacting with HSP90 are known (a list can be found at https://www.picard.ch/downloads/Hsp90interactors.pdf). The increase in volume of knowledge about the complexity of these interactions^4,5 has led to the emergence of a point of view that HSP90 should be considered not as a structurally and functionally homogeneous protein, but as a protein that is shaped by its environment.⁶ Moreover, in recent work it has been shown that nucleotide exchange, but not ATP hydrolysis, is necessary for HSP90 functions in vivo.⁷ All this makes the description of the chaperone cycle above not only very simplified but also partially incorrect.

Additionally, the existence of four isoforms of human HSP90,⁸ which are located in different parts of a cell,⁹ and secretion of HSP90 in the extracellular space (eHSP90)^10–12 make the landscape of this research field extremely complex.

Overexpression of HSP90 in multiple tumors and the presence among its clients the proteins that play important roles in surviving and proliferation of cancer cells (protein kinases, transcription factors, nuclear receptors) made HSP90 an attractive therapeutic target.^13–15 In the last two decades, most of the efforts were directed to the development of small molecules that block ATP-binding pocket of NTD.^16,17 Lately, C-terminal inhibitors¹⁸ and protein–protein disruptors^19,20 have been garnering increasing attention.

Typical medicinal chemistry work dedicated to the development of NTD HSP90 inhibitors includes the search for small molecules that have maximum affinity (usually measured using fluorescence polarization assay) to purified recombinant HSP90 (full-length or only NTD), followed by the evaluation of the most potent binders as antiproliferative agents on different cancer cell lines. Additionally, such articles can contain X-ray structures of the crystals of the NTD of HSP90 in a complex with the small molecules. At the same time, not too much attention is paid to the changes that occur in the cancer cells under the action of the inhibitor. In most cases at best, there are western blots that demonstrate changes in the levels of few well-known HSP90 client proteins, which are essential for surviving and proliferation of cancer cells.

Taking into account all the data discussed above – especially the complexity of HSP90 conformational changes occurring in cells – a crucial question arises: Is using purified recombinant HSP90 as a model and searching for its high-affinity binders the right strategy to combat cancer by targeting HSP90 function? Further supporting the validity of this question is the fact that, as early as 2003, it was demonstrated that tumor-derived HSP90 exhibits a 100-fold higher binding affinity for 17-AAG (Fig. 1) compared to HSP90 from normal cells.²¹ A second important question is whether different small molecules targeting the ATP-binding pocket of the NTD can cause distinct disruptions in HSP90 function due to variations in their affinities for different conformations and/or the distinct conformational changes they induce upon binding.

Fig. 1. The conformations of the NTD lid of HSP90 in the complexes with 17-DMAG²² (pdbid 1OSF) (A), luminespib²³ (pdbid 6LTI) (B), pimistepib analog²⁴ (pdbid 5ZR3) (C), KUNA-111²⁵ (pdbid 7UR3) (D) and structures of these and related inhibitors.

In our opinion, the answer to this question is yes. Analysis of HSP90 NTD complexes with various known inhibitors (a similar analysis was previously conducted in the context of isoform selectivity²⁶) reveals that different small molecules can induce distinct conformations of the NTD lid (Fig. 1). An especially intriguing case is the selective HSP90α inhibitor KUNA-111, which exhibits a unique binding mode and does not induce a significant antiproliferative effect.

Although, for now, HSP90 inhibitors are not very successful (none of these compounds was approved by FDA as an anticancer drug), the search for new chemical entities that target the NTD of HSP90 and studying the effects that they (as well as already known inhibitors) induce in cells is still an urgent task. This work describes the development of a novel class of HSP90 inhibitors based on the 4,5,6,7-tetrahydrobenzo[d]isoxazole scaffold. Earlier, benzisoxazoles as HSP90-targeting agents were described in the work,²⁷ partially saturated analogs of benzisoxazoles – in the work,²⁸ and in our previous works^29,30 (Chart 1). The aim of this research was the preparation of new chemical species that would have a different mode of binding to HSP90 than existing inhibitors and their evaluation as anticancer agents.

Chart 1. Structures of HSP90-targeting compounds containing benzisoxazole or partially saturated benzisoxazole core.

Results and discussion

Design and virtual screening

Early we have synthesized 6,7-dihydrobenzo[d]isoxazol-4(5H)-ones 1a³⁰ and 6,7-dihydrobenzo[c]isoxazol-4(5H)-ones 1b³¹ (Chart 2) that contains isoxazole ring substituted with methyl-protected resorcinol fragment. The presence of moieties in their structures that have vital importance for the binding of luminespib to HSP90 and the six-membered ring with three positions (4, 5 and 7) that can be used for additional modifications, makes these molecules perspective scaffolds for the development of new HSP90 inhibitors.

Chart 2. Structures of 6,7-dihydrobenz[d]isoxazol-4(5H)-ones 1a and 6,7-dihydrobenz[c]isoxazol-4(5H)-ones 1b – perspective scaffolds for the preparation of new HSP90 inhibitors.

After analyzing the complex of luminespib with HSP90α (pdbid 6LTI), we hypothesized that introduction of amide group at position 4 can be used as a strategy for preparation of new ligands to the NTD of HSP90 with different from luminespib binding mode. For examination of this hypothesis, we carried out docking of two enantiomeric acetamides (R)-athbzi-2 and (S)-athbzi-2 (Fig. 2). As can be seen from Fig. 2C, both enantiomers fit in the ATP pocket with their isoxazole-resorcinol moiety in a very similar way to luminespib, and the amide bond is placed where the aryl fragment of luminespib is. In predicted docked pose, (R)-athbzi-2 gives hydrogen bond between THR-109 of the lid and oxygen atom of the amide group.

Fig. 2. Structures of new potential classes of HSP90 inhibitors (R)-athbzi (A) and (S)-athbzi (B) and superposition (C) of theoretical poses of (R)-athbzi-2 (green) and (S)-athbzi-2 (pink) with experimental pose of luminespib (teal) in ATP-pocket of the NTD HSP90.

Because the results of docking study looked promising, next we carried out virtual screening of compound libraries of (R)-athbzi and (S)-athbzi (both [d]- and [c]-regioisomers) with different R (496 compounds in total). The libraries were enumerated based on the in-house library of carboxylic acids (R-COOH). The protocol of virtual screening included molecular docking of the compounds into ATP-pocket of the NTD HSP90 followed by 1 ns molecular dynamics simulations of obtained complexes and reassessment of docking affinities with MM/GBSA³² method.

The analyses of the obtained results included comparison of the values of MM/GBSA ΔH with the analogic values for luminespib and visual inspections of docked complexes. When choosing compounds for the synthesis, we also considered that their set covered the maximum diversity of structures. As a result, we selected for the synthesis and antiproliferative evaluation a number of [d]-regiosomeric athbzi amides that in general had better MM/GBSA and MM/PBSA ΔH values than analogic [c]-regiosomers. Only a few [c]-analogues of the most active compounds were planned to be synthesized for comparison. Initially, it was decided to prepare all compounds in racemic form.

Synthesis

For the synthesis of target compounds, we used the approach that had been earlier developed for the preparation of 4-amino- and 4-acylamino-tetrahydroindazoles.³³ At the first stage, carbonyl group in 1a and 1b was reduced by sodium tetraborohydride in EtOH as a solvent with good reaction yields (88 and 80%, respectively) (Scheme 1). The obtained alcohols 2a and 2b were introduced into the Ritter reaction with chloroacetonitrile in the presence of sulfuric acid in acetic acid as a solvent. In such conditions, chloroacetamides 3a and 3b were obtained in 80 and 58% yields, respectively. Subsequent removal of chlororoacetyl group under the action of thiourea³⁴ resulted in the desired amines 4a and 4b.

Scheme 1. Synthesis of potential HSP90 inhibitors 7–9.

In the course of acylation of amines 3a and 3b, a number of amides 5 and 6 were obtained in 47–94% yields. Additional amides 5d and 5j were obtained from chloroacetamide 3a using nucleophilic substitution reaction with methanol in the presence of KOH or with morpholine, respectively. Removal of methyl protection groups from hydroxyls of compounds 5 and 6 using BBr₃ (3 equiv. per OMe) in DCM gave final compounds 8 and 9. Their list is given in the Table 1. The same deprotection step for amines 4a and 4b, followed by treatment with hydrogen chloride in methanol, led to the target compounds 7a and 7b in 45% and 54% yields, respectively.

Table 1. Antiproliferative activity of the synthesized compounds and the reference drug against the human breast cancer cells MCF7 and HCC1954.

Cmpd	R	MCF7 cells		HCC1954 cells
		IC50, μM	Cell viability at 50 μM	IC50, μM	Cell viability at 50 μM
7a	—	34.4	46	4.0	21
7b	—	>50	56	13.6	20
8a		0.56	19	1.3	18
8b		7.3	45	0.95	26
8c		1.5	21	6.1	41
8d		6.1	43	5.9	23
8e		6.7	35	5.5	32
8f		4.6	21	6.1	11
8g		3.6	28	3.7	29
8h		1.1	16	1.1	15
8i		17.4	30	18.4	20
8j		11.7	44	4.2	15
8k		2.3	33	3.1	27
8l		>50	83	44.7	42
8m		11	45	2.7	29
8n		0.59	18	0.33	35
9a		1.5	24	15.8	43
9b		3.9	16	8.1	15
Luminespib		0.014	9	0.2	8

Evaluation of antiproliferative activity

The molecular classification of breast cancer includes five subtypes: luminal A (ERα+, PR±, HER2-), luminal B (ERα+, PR±, HER2+), HER2-enriched (ERα−, PR−, HER2+), basal-like, and normal breast-like. Among basal-like breast cancers, 55–85% of tumors, as well as a small percentage of normal-like tumors, exhibit a triple-negative phenotype (ERα−, PR−, HER2−), placing them in the most adverse prognostic group alongside HER2+ subtypes.³⁵

Approximately 70% of breast cancer cases are hormone-dependent subtypes, while basal-like breast cancer accounts for 8–20% and is more commonly observed in younger patients. HER2-positive breast cancer occurs in 15–20% of cases, is characterized by aggressive disease progression, and requires a multimodal treatment approach targeting various molecular pathways.^36–38

The effects of the synthesized compounds were evaluated on two breast cancer cell lines: HCC1954 and MCF7. The HCC1954 cell line (HER2-positive) is an epithelial cell line derived from a primary stage IIA, grade 3 invasive ductal carcinoma, obtained from a 61 year-old woman without lymph node metastases.³⁹ The MCF7 cell line (luminal A) is an epithelial-like cell line derived from invasive mammary ductal adenocarcinoma, commonly used as a model for breast cancer research. These cells were isolated from the pleural effusion of a 69 year-old woman with metastatic disease.⁴⁰

The obtained compounds had a significant effect on cell growth. The IC₅₀ values are given in Table 1. Only two compounds in series had IC₅₀ values on MCF7 breast cancer cells exceeding 50 μM. Four compounds showed a medium level of inhibitory activity with IC₅₀ ranging from 11 to 35 μM, the remaining compounds significantly inhibited cell growth with IC₅₀ values less than 10 μM. Two compounds (8a and 8n) were active in the submicromolar range of concentrations, suggesting their potential as antiproliferative agents for MCF7 cells. Although HCC1954 cells are more aggressive than MCF7 cells, the synthesized compounds blocked their proliferation. Only for one compound (8m), IC₅₀ value approached 50 μM, while the other observed IC₅₀ values were below 20 μM. Fourteen compounds showed significant antiproliferative effects, their IC₅₀ values for HCC1954 were less than 10 μM. The most active compound (8n) caused 50% inhibition of HCC1954 cell growth at a concentration of 0.33 μM. Thus, series of novel compounds were effective against luminal A (MCF7) and HER2-positive (HCC1954) breast cancer cells.

Interesting, that [d]-regioisomeric compounds 8a and 8n were more active than their [c]-regioisomers 9a and 9b. Overall, based on the obtained data, it can be concluded that the compounds with simple substituents (CH₃ and CF₃), as well as compounds containing the Het(Ar)–CH₂–CH₂– fragment, exhibited the highest activity.

It is worth noting that the convergence of data on antiproliferative activity with the ΔH values obtained during virtual screening was practically absent. In our opinion, one of the possible reasons for this was the use of too short (1 ns) molecular dynamics simulations.

Synthesis of separate enantiomers of the most active compounds and evaluation of their antiproliferative activity

From the 1980s onwards, regulators have shown a preference for bringing single enantiomers to the market. This approach reduces the risk of adverse effects caused by inactive or harmful enantiomers found in racemic mixtures, enhancing drug safety and efficacy.⁴¹ Therefore, studying the activity and effects of separate enantiomers already in early drug discovery is essential. For this reason, we decided to resolve racemic amine 4a and prepare the most active compounds 8a and 8n in enantiopure form.

The chiral resolution of amine (±)-4a was carried out via salts formation with O,O′-dibenzoyl tartaric acids and crystallization of the salts from acetone (Scheme 2). For determination of absolute configuration and enantiomeric excess (ee) we used NMR analysis of Mosher's amides,⁴² which were synthesized from enantioenriched amines (for details see ESI†).

Scheme 2. Resolution of 3-(5-isopropyl-2,4-dimethoxyphenyl)-4,5,6,7-tetrahydrobenzo[d]isoxazol-4-amine 4a.

Acylation of (R)-4a and (S)-4a with acetic anhydride or with 3-(1H-tetrazol-1-yl)propanoic acid in the presence of EDC led to desired products that next were tested for antiproliferative activity (Table 2).

Table 2. Antiproliferative activity of separate enantiomers of most active amides 8a and 8n.

Cmpd	Structure	MCF7 cells		HCC1954 cells
		IC50, μM	Cell viability at 50 μM	IC50, μM	Cell viability at 50 μM
(R)-8a		0.37	17	0.64	16
(S)-8a		1.2	13	1.5	13
(R)-8n		0.35	19	0.55	18
(S)-8n		0.93	19	1.1	12

In result, it was determined that (R)-enantiomers are more active than (S)-isomers. Surprisingly, separate enantiomers of compound 8n had worst IC₅₀ values on HCC1954 than it does in racemic form. The data of antiproliferative tests determined the choosing of compound (R)-8n as a lead, subsequent in vitro experiments were aimed at its in-depth study on HER2-positive tumor cells.

Identification of active signaling pathways in cells treated with compound (R)-8n

As discussed in detail above, heat shock proteins are molecular chaperones that play a key role in regulating protein homeostasis and the stress response. Inhibition of HSP70, HSP90, and HSF1 is considered a promising approach to cancer therapy. Additionally, signaling pathways associated with chaperones are actively being studied. Protein phosphatase 5 (PP5) is a serine/threonine phosphatase involved in the regulation of various signaling pathways, including the control of cell growth, apoptosis, and stress response.^43,44 In recent years, interest in PP5 has increased due to its potential role, and modulators of this phosphatase's activity have attracted significant attention from researchers.^45–47 Further studies will help identify the most effective proliferation inhibitors and cell death activators for each type of tumor cells. Additionally, an important research focus is the search for compounds that reduce the toxicity of HSP inhibitors in normal tissues.

Currently, several methods have been developed to analyze the molecular pathways involved in the cellular response to HSP90 inhibition.^48,49 Markers of HSP90 inhibition include its client proteins as well as other chaperones.^49–51 Among chaperones, the most specific marker of HSP90 inhibition is HSP70.⁴⁹ One of the first studies in this field demonstrated that geldanamycin, a well-known HSP90 inhibitor, induces heat shock protein expression through the activation of HSF1 in K562 erythroleukemic cells.⁵² A subsequent study established differences between high hydrostatic pressure and geldanamycin. In HeLa cells, both high hydrostatic pressure and geldanamycin upregulated HSP70 expression through mRNA stabilization.⁵³ However, unlike high hydrostatic pressure, geldanamycin increased HSF1 activation. Thus, chemical inhibition of HSP90 selectively stimulates HSP70 expression via HSF1.

In our study, compound (R)-8n, at submicromolar and micromolar concentrations, induced HSP70 accumulation in HER2+ HCC1954 cancer cells (Fig. 3A). An increase in HSP70 expression was observed in cells treated with compound (R)-8n at a concentration of 0.09 μM, reaching its maximum at 3 μM.

Fig. 3. Signaling pathways in HCC1954 breast cancer cells treated with (R)-8n. GAPDH antibodies were used as a loading control. LUMI – luminespib (NVP-AUY922) (A). HER2 (ERBB2) and EGFR as key HSP90 clients (diagram of HSP90 signaling pathways created using STRING) (B).

To further confirm HSP90 inhibition, we selected a panel of cell lines with different molecular profiles. The first classification of breast neoplasms, developed in the 20th century, categorized tumors into two groups: hormone-dependent breast cancers, which accounted for the majority of cases, and hormone-independent breast cancers, which made up 20–30% of cases.^54,55 Advances in molecular analysis methods led to an expanded classification, where tumors with expression of steroid hormone receptors (ERα, PR) are classified as hormone-dependent, those with HER2 expression as HER2+, and tumors lacking these markers as triple-negative breast cancers (TNBC).⁵⁶

Our study focused on the aggressive HER2+ breast cancer cell line HCC1954.³⁹ Additionally, we included hormone-dependent breast cancer cell lines MCF7 and T47D, as well as triple-negative breast cancer cell lines MDA-MB-231 and HCC1806. These lines exhibited different levels of HSP70 expression (Fig. S4†). To assess inhibition of the HSP90 axis, cancer cells were incubated with compound (R)-8n, and the increase in HSP70 expression was analyzed using immunoblotting. The hormone-dependent MCF7 cells exhibited a high basal level of HSP70 expression, and compound (R)-8n significantly enhanced it. In T47D cells, where basal HSP70 expression was low, incubation with the compound led to an increase in this marker. Similar trends were observed in the two triple-negative breast cancer cell lines, MDA-MB-231 and HCC1806, where HSP70 expression increased following compound (R)-8n treatment. Thus, the significant accumulation of HSP70 in various cell lines after incubation with compound (R)-8n indicates inhibition of the HSP90-related pathway.

Given the role of HSP90 in various signaling pathways, we assessed the effect of compound (R)-8n on the expression of target proteins, including those involved in apoptosis and proliferation. Luminespib, a well-studied HSP90 inhibitor, was used as a reference drug.⁵⁷ The effect of the compounds on signaling pathways in HCC1954 cells was assessed by immunoblotting (Fig. 3A).

HSP90 regulates the folding of multiple receptor tyrosine kinases that drive rapid cancer progression. These kinases include HER2, c-MET, and EGFR² (Fig. 3B). Their high expression in tumor cells is associated with poor prognosis and increased aggressiveness.⁵⁸ Several drugs have been developed to target individual or multiple kinases. In HCC1954 cells, treated with compound (R)-8n, HER2 expression, a key target, was significantly reduced. The selected HSP90 inhibitor effectively decreased both HER2 expression and activity. The expression level of EGFR remained unchanged, whereas p-EGFR expression was significantly downregulated.

The c-MET receptor, a receptor tyrosine kinase (RTK) from the MET (MNNG HOS transforming gene) family, is expressed in various tumor cells.⁵⁹ Its ligand, hepatocyte growth factor (HGF), plays a key role in tumor growth and metastasis.⁶⁰ c-MET is an HSP90-dependent kinase, and HSP90 stabilizes and interacts with phosphorylated c-MET, whether its activation is ligand-dependent or caused by kinase domain mutations.⁶¹ Compound (R)-8n reduced the expression of total c-MET and its phosphorylated forms p-c-MET (Tyr1003) and p-c-MET (Tyr1234/1235). Thus, compound (R)-8n effectively blocked the activity of three oncoreceptors in HCC1954 cells.

AKT kinase is a key HSP90 client, and its interaction with HSP90 is a major consideration in targeted therapy development.^62,63 Inhibition of HSP90 leads to decrease in AKT kinase activity, sensitizing cells to pro-apoptotic agents. In HCC1954 cells, treatment with compound (R)-8n at various doses significantly reduced AKT expression and activity.

In mammalian cells, CDK4 (cyclin-dependent kinase 4) binds to D-type cyclins and mediates entry of cells into the cell cycle in response to a mitogenic signal. Activation of CDK4 complexes with cyclin D promotes phosphorylation of the retinoblastoma protein (pRb) and related proteins p107 and p130.⁶⁴ The hypophosphorylated form of the pRb binds and sequesters multiple signaling proteins, while its hyperphosphorylation releases these molecules. The key binding partner of pRb is the transcription factor E2F1, which activates genes required for G1-phase progression. Targeted inhibition of CDK4 is a promising approach in cancer therapy. HSP90 plays a crucial role in regulating CDK4 activity. The Cryo-EM structure of the HSP90/CDK4/Cdc37 complex, described in 2016,⁶⁵ was a key breakthrough in understanding the HSP90-CDK4 interaction in both normal and malignant cells. In this study, we found that the selected HSP90 inhibitor compound (R)-8n significantly suppressed CDK4 expression.

PARP is involved in chromatin remodeling, transcription, replication, recombination, and protein modifications.⁶⁶ PARP cleavage is considered a hallmark of apoptosis, with its accumulation indicating the pro-apoptotic potential of a drug. In this study, we observed cleaved PARP expression, a characteristic marker of apoptosis, in HCC1954 cells treated with compound (R)-8n (Fig. 3A).

The cell cycle of the HCC1954 cell line was analyzed by measuring DNA content using flow cytometry (Fig. 4). HCC1954 cells were incubated with compound (R)-8n (0.125–2 μM) for 48 hours. After treatment with a propidium iodide (PI)-containing buffer, cell cycle distribution was assessed. Cells incubated with DMSO served as the control group.

Fig. 4. Cell cycle distribution of HCC1954 cells treated by compound (R)-8n at concentrations of 0 (control, DMSO), 0.125, 0.5 and 2 μM for 48 h.

As shown in Fig. 4, the fraction of cells in the sub-G1 phase (containing fragmented DNA, indicative of dead cells) increased in a dose-dependent manner, reaching 35% at 2 μM. Simultaneously, the fraction of cells in the G1 phase decreased dramatically from 44% in the control group to 12% in the group treated with 2 μM. Moreover, compound (R)-8n promoted the accumulation of cells in the G2/M phases, suggesting cell cycle arrest at the G2/M boundary.

Thus, the cytotoxic activity of compound (R)-8n is associated with inducing cell cycle delay in the G2/M phases, followed by dose-dependent cell death.

Molecular modeling study

As it was mentioned in the Introduction section, the NTD of HSP90 has different conformations in complexes with different small molecules. Initially at virtual screening stage, we carried out docking of compound (R)-8n into NTD with closed lid conformation (Fig. 1B). Next we additionally carried out docking into NTD with conformations from Fig. 1A, C and D. It was found that compound (R)-8n also can fit with expected pose into NTD with open lid conformation (Fig. 1A). Having two possible complexes of (R)-8n with HSP90α, we conducted 1000 ns dynamics simulations of both and calculated ΔH values using the MM/GBSA method. As a result, we obtained values (Table 3) from which it can be assumed that the formation of a complex with the closed lid conformation is more likely. Analogous calculations were carried out for experimental complex of luminespib with HSP90α (pdbid 6LTI).

Table 3. MM/GBSA ΔH values of binding compound (R)-8n luminespib to NTD of HSP90α.

Compound	ΔH, kcal mol−1 (closed lid conformation complex)	ΔH, kcal mol−1 (open lid conformation complex)
(R)-8n	−44.50	−43.79
Luminespib	−50.26	—

Analysis of the obtained trajectories shows that flexible –(CH₂)₂-tetazolyl fragment changed its position during simulation (as reflected in the RMSD chart (Fig. 5A)). However, the overall complex remained stable, as seen in the MM/GBSA ΔH chart (Fig. 5B). Interaction mode of (R)-8n with HSP90α NTD from the simulation snapshot at t = 185.5 ns is shown in Fig. 5C and D. The arylisoxazole fragment has the same interactions with protein as luminespib does. The carbonyl of the amide group forms a hydrogen bond with ASN51, and tetrazole ring has pi-stacking with ASP54.

Fig. 5. The time dependence of the RMSD of (R)-8n. Non-hydrogen atoms of (R)-8n and the backbone atoms of M^pro were used in the calculation (A); the change of MM/GBSA ΔH value during the MD simulation of the predicted complex of (R)-8n with HSP90α (B); 3D representation of the predicted binding pose of (R)-8n in complex with N-terminal of HSP90α (hydrogen bonds shown as yellow dash lines) (C); 2D representation of intramolecular interactions of (R)-8n with N-terminal of HSP90α (D).

Assessment of selectivity

Testing the toxicity of chemical compounds on fibroblasts is a widely used method for assessing the safety of various substances, including pharmaceuticals, cosmetics, and industrial chemicals.^67–69 Fibroblasts, as connective tissue cells, play a crucial role in maintaining tissue structural integrity and exhibit high sensitivity to toxic effects, making them a suitable model for in vitro experiments. These cells are easily cultured under laboratory conditions, allowing for experiments on a stable cellular model. Fibroblasts also respond to a wide range of toxic agents, including those causing oxidative stress and membrane damage, making them ideal for evaluating the cytotoxicity of various classes of new compounds.

The survival comparison between fibroblasts and breast cancer cells is shown in Fig. 6 compound (R)-8n began to exert antiproliferative effects at a concentration of approximately 0.1 μM. Half-maximal inhibition of cancer cell growth was observed at compound (R)-8n concentrations of 0.35–0.55 μM, while fibroblast viability remained high. As the compound (R)-8n concentration increased above 1 μM, significant cancer cell death was observed, with fibroblasts demonstrating resistance to the effects and showing a lower degree of cell death. Thus, the compound (R)-8n exhibits high selectivity toward cancer cells (Fig. 6).

Fig. 6. Antiproliferative activity and toxicity of compound (R)-8n. The breast cancer cells and fibroblasts were treated with the compound and after 72 hours their survival was evaluated by MTT test.

Conclusions

Our study aimed at discovering a new class of antiproliferative agents. The integrated use of in silico and in vitro approaches allowed us to characterize N-(4,5,6,7-tetrahydrobenzisoxazol-4-yl)amides as potent compounds that inhibit the proliferation of breast cancer cells related to two molecular subtypes. Extended experiments on HER2+ HCC1954 cells revealed the potential of N-(4,5,6,7-tetrahydrobenzisoxazol-4-yl)amides as inhibitors of the HSP90 pathway. The effect of the lead compound, (R)-8n, on key oncoreceptors (HER2, EGFR, c-MET) and mitogenic kinases (AKT, CDK4) was established. Furthermore, accumulation of the apoptosis marker, cleaved PARP, was observed in HCC1954 cells treated with the compound. The compound (R)-8n not only exhibited significant anti-proliferative effects but also demonstrated a relatively high selectivity for tumor cells. Further research is required to analyze the stability, toxicity, and pharmacological characteristics of the new HSP90 inhibitor.

Experimental

Chemistry

Full detailed synthetic procedures and characterizations of all new compounds are given in the ESI.† Here, synthetic route details for compound 8n are provided.

General procedure for the synthesis of compounds 2a–b

3-(5-Isopropyl-2,4-dimethoxyphenyl)-6,7-dihydrobenzisoxazol-4(5H)-ones 1a–b (2.26 g, 7.17 mmol, 1 eq.) were dissolved in EtOH (48 mL) and treated with NaBH₄ (543 mg, 14.3 mmol, 2 eq.). The reaction mixture was stirred overnight at room temperature. After the completion of the reaction, EtOH was removed by rotary evaporation. The residue was partitioned between saturated aqueous NH₄Cl (100 mL) and EtOAc (100 mL), the organic layer was separated. The aqueous layer was extracted with EtOAc (3 × 30 mL). The combined organic phases were dried over Na₂SO₄ and concentrated under reduced pressure. The residue was purified by column chromatography on silica gel (30% EtOAc/PE) to give alcohols 2a–b.

3-(5-Isopropyl-2,4-dimethoxyphenyl)-4,5,6,7-tetrahydrobenzo[d]isoxazol-4-ol 2a

Yield 88%. White solid, mp 45–50 °C. ¹H NMR (500 MHz, CDCl₃) δ 7.32 (s, 1H), 6.55 (s, 1H), 4.68 (t, J = 3.4 Hz, 1H), 3.89 (s, 3H), 3.87 (s, 3H), 3.25 (hept, J = 6.9 Hz, 1H), 3.10 (br. s, 1H), 2.87 (ddd, J = 17.1, 5.6, 3.2 Hz, 1H), 2.70–2.56 (m, 1H), 2.22–2.09 (m, 1H), 2.01–1.87 (m, 2H), 1.75–1.65 (m, 1H), 1.20 (d, J = 3.4 Hz, 3H), 1.19 (d, J = 3.4 Hz, 3H). ¹³C NMR (126 MHz, CDCl₃) δ 169.74 (C), 159.38 (C), 159.28 (C), 155.74 (C), 130.60 (C), 129.16 (CH), 116.21 (C), 110.24 (C), 96.22 (CH), 61.45 (CH), 56.72 (CH₃), 55.71 (CH₃), 31.01 (CH₂), 26.52 (CH), 23.04 (CH₂), 22.76 (2CH₃), 17.61 (CH₂). HRMS (ESI⁺): m/z calc'd for C₁₈H₂₃NO₄Na [M + Na]⁺: 340.1525, found 340.1518.

General procedure for the synthesis of compounds 3a–b

Chloroacetonitrile (2.0 mL, 32.0 mmol, 4 eq.) was added to a solution of alcohols 2a–b (2.54 g, 8.0 mmol, 1 eq.) in glacial acetic acid (8 mL), followed by sulfuric acid (4.3 mL, 80.0 mmol, 10 eq.). The resulting reaction mixture was stirred at 60 °C for 25 h. The reaction mixture was cooled to 0 °C, poured into a vigorously stirred 10% aqueous solution of NaOH (100 mL) at 0 °C and extracted with CHCl₃ (3 × 30 mL). The combined organic layers were dried over Na₂SO₄. After removal of the solvent in vacuum, the residue was purified by column chromatography on silica gel (35 → 45% EtOAc/PE) to give products 3a–b.

2-Chloro-N-(3-(5-isopropyl-2,4-dimethoxyphenyl)-4,5,6,7-tetrahydrobenzo[d]isoxazol-4-yl)acetamide 3a

Yield 79%. Light brown solid, mp 133–137 °C. ¹H NMR (500 MHz, CDCl₃) δ 7.21 (s, 1H), 6.43 (s, 1H), 6.27 (br. d, J = 7.9 Hz, 1H), 5.13 (dd, J = 13.3, 7.5 Hz, 1H), 3.87 (s, 3H), 3.80 (s, 3H), 3.78 (d, J = 15.1 Hz, 1H), 3.51 (d, J = 15.1 Hz, 1H), 3.23 (hept, J = 6.9 Hz, 1H), 2.80–2.74 (m, 2H), 2.22–2.12 (m, 1H), 2.05–1.88 (m, 2H), 1.65–1.54 (m, 1H), 1.19 (d, J = 6.9 Hz, 3H), 1.15 (d, J = 6.9 Hz, 3H). ¹³C NMR (126 MHz, CDCl₃) δ 169.60 (C), 165.06 (C), 159.48 (C), 159.26 (C), 156.50 (C), 129.40 (C), 128.12 (CH), 113.34 (C), 109.23 (C), 94.74 (CH), 55.79 (2CH₃), 44.03 (CH), 42.36 (CH₂), 30.43 (CH₂), 26.37 (CH), 22.95 (CH₃), 22.82 (CH₂), 22.75 (CH₃), 20.15 (CH₂). HRMS (ESI⁺): m/z calc'd for C₂₀H₂₅ClN₂O₄Na [M + Na]⁺: 415.1400, found 415.1396.

General procedure for the synthesis of compounds 4a–b

Thiourea (690 mg, 9.08 mmol, 2 eq.) was added to a solution of compounds 3a–b (1.78 g, 4.54 mmol, 1 eq.) in EtOH (35 mL). The resulting reaction mixture was heated under reflux for 24 h and then EtOH was removed by rotary evaporation. The residue was dissolved in CHCl₃ (100 mL) and washed with 10% aqueous solution of NaOH (30 mL). The organic phase was dried over Na₂SO₄, filtered and evaporated under reduced pressure. The residue was purified on silica gel (5 → 10% MeOH/CHCl₃) to provide amines 4a–b.

3-(5-Isopropyl-2,4-dimethoxyphenyl)-4,5,6,7-tetrahydrobenzo[d]isoxazol-4-amine 4a

Yield 72%. Light brown oil. ¹H NMR (500 MHz, CDCl₃) δ 7.35 (s, 1H), 6.51 (s, 1H), 4.05–4.01 (m, 1H), 3.89 (s, 3H), 3.86 (s, 3H), 3.24 (hept, J = 6.9 Hz, 1H), 2.75–2.69 (m, 2H), 2.10–2.01 (m, 2H), 1.91–1.81 (m, 1H), 1.50–1.41 (m, 1H), 1.32 (br. s., 2H), 1.20 (d, J = 6.9 Hz, 3H), 1.18 (d, J = 6.9 Hz, 3H). ¹³C NMR (126 MHz, CDCl₃) δ 167.83 (C), 159.31 (C), 159.27 (C), 156.14 (C), 129.73 (C), 128.58 (CH), 118.23 (C), 110.49 (C), 95.27 (CH), 55.95 (CH₃), 55.64 (CH₃), 44.99 (CH), 33.07 (CH₂), 26.46 (CH), 23.00 (CH₂), 22.91 (CH₃), 22.69 (CH₃), 20.12 (CH₂). HRMS (ESI⁺): m/z calc'd for C₁₈H₂₄N₂O₃Na [M + Na]⁺: 339.1685, found 339.1681.

General procedure for the synthesis of compounds 5 and 6. Procedure A

A round-bottomed flask was charged with dry DCM (1 mL), 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (EDC·HCl) (55 mg, 0.29 mmol, 1.3 eq.), and DMAP (38 mg, 0.31 mmol, 1.4 eq.) or hydroxybenzotriazole (HOBt) (3 mg, 0.02 mmol, 0.1 eq.) under argon atmosphere. The reaction flask was cooled to zero degree in an ice bath, and the carboxylic acid (0.24 mmol, 1.1 eq.) was then added. After five minutes of stirring, a solution of the amine (70 mg, 0.22 mmol, 1 eq.) in dry DCM (1 mL) was added into the flask. The ice bath was removed, and the reaction mixture was allowed to stir at room temperature until the starting material was consumed completely as indicated by TLC (1–2 h). The reaction was quenched with HCl (1.0 M, 2 mL), and the organic layer was separated from the aqueous layer. The aqueous layer was then extracted with DCM (2 × 2 mL). The organic layers were combined and dried over Na₂SO₄. After removing the solvent with a rotary evaporator, the crude product was purified by silica gel column chromatography using either PE/EtOAc or CHCl₃/MeOH as eluent.

N-(3-(5-Isopropyl-2,4-dimethoxyphenyl)-4,5,6,7-tetrahydrobenzo[d]isoxazol-4-yl)-3-(1H-tetrazol-1-yl)propanamide 5n

Synthesized according to the general procedure A from amine 4a and 3-(1H-tetrazol-1-yl)propanoic acid using DMAP. The crude product was purified by column chromatography on silica gel (5% MeOH/CHCl₃). Yield 90%. Light yellow solid, mp 70–75 °C. ¹H NMR (500 MHz, CDCl₃) δ 8.67 (s, 1H), 7.16 (s, 1H), 6.34 (s, 1H), 5.63 (br. d, J = 7.4 Hz, 1H), 5.02 (dd, J = 12.6, 6.9 Hz, 1H), 4.59–4.49 (m, 2H), 3.91 (s, 3H), 3.64 (s, 3H), 3.21 (hept, J = 6.9 Hz, 1H), 2.76–2.65 (m, 2H), 2.55–2.45 (m, 1H), 2.29–2.19 (m, 1H), 2.12–2.03 (m, 1H), 1.94–1.82 (m, 2H), 1.55–1.43 (m, 1H), 1.16 (d, J = 6.9 Hz, 3H), 1.14 (d, J = 6.9 Hz, 3H). ¹³C NMR (126 MHz, CDCl₃) δ 169.67 (C), 167.80 (C), 159.52 (C), 158.95 (C), 156.33 (C), 143.80 (CH), 129.46(C), 127.94 (CH), 113.12 (C), 109.12 (C), 95.09 (CH), 56.00 (CH₃), 55.76 (CH₃), 43.94 (CH), 43.70 (CH₂), 34.98 (CH₂), 30.44 (CH₂), 26.37 (CH), 22.81 (CH₃), 22.76 (CH₃), 22.74 (CH₂), 19.88 (CH₂). HRMS (ESI⁺): m/z calc'd for C₂₂H₂₈N₆O₄Na [M + Na]⁺: 463.2070, found 463.2063.

General procedure for the synthesis of compounds 8 and 9

To a dry round-bottomed flask equipped with a stir bar, amide 5 or 6 (1 eq.) was added, followed by the addition of dry DCM (0.1 M). The flask was filled with argon, after which BBr₃ (3 eq. per OMe group) was added slowly at 0 °C. The reaction mixture was stirred at room temperature until the starting material was consumed completely as indicated by TLC (72–96 h). The reaction was quenched with saturated solution of NaHCO₃, and the organic layer was separated from the aqueous layer. The aqueous layer was then extracted with EtOAc. The organic layers were combined and dried over Na₂SO₄. After removing the solvent with a rotary evaporator, the crude product was purified by silica gel column chromatography or preparative TLC using either PE/EtOAc or CHCl₃/MeOH as eluent.

N-(3-(2,4-Dihydroxy-5-isopropylphenyl)-4,5,6,7-tetrahydrobenzo[d]isoxazol-4-yl)-3-(1H-tetrazol-1-yl)propanamide 8n

The crude product was purified by column chromatography on silica gel (5% MeOH/CHCl₃). Yield 62%. White solid, mp 144–147 °C. ¹H NMR (500 MHz, DMSO) δ 9.65 (br. s, 1H), 9.54 (br. s, 1H), 9.22 (s, 1H), 8.27 (br. d, J = 7.6 Hz, 1H), 7.07 (s, 1H), 6.43 (s, 1H), 4.99 (dt, J = 6.8, 3.0 Hz, 1H), 4.63–4.44 (m, 2H), 3.06 (hept, J = 6.9 Hz, 1H), 2.81–2.71 (m, 1H), 2.70–2.59 (m, 2H), 2.49–2.42 (m, 1H), 1.84–1.74 (m, 1H), 1.73–1.58 (m, 3H), 1.06 (d, J = 6.9 Hz, 6H). ¹³C NMR (126 MHz, DMSO) δ 169.45 (C), 167.65 (C), 159.49 (C), 156.81 (C), 154.71 (C), 143.97 (CH), 126.53 (CH), 125.85 (C), 111.90 (C), 105.63 (C), 102.60 (CH), 44.06 (CH₂), 41.22 (CH), 34.86 (CH₂), 29.09 (CH₂), 25.67 (CH), 22.62 (CH₃), 22.51 (CH₃), 21.92 (CH₂), 17.22 (CH₂). HRMS (ESI⁺): m/z calc'd for C₂₀H₂₄N₆O₄Na [M + Na]⁺: 435.1757, found 435.1749.

Cell lines and antiproliferative activity and selectivity evaluation

The MCF7, T47D, MDA-MB-231, HCC1806 and HCC1954 breast cancer cell lines were obtained from the ATCC. Human skin fibroblasts (hFB-hTERT) were generously provided by E. Dashinimaev and V. Tatarskiy.⁷⁰ MCF7 cells were cultured in DMEM medium supplemented with high glucose and 10% fetal bovine serum (FBS; HyClone), while hFB-hTERT, T47D, MDA-MB-231, HCC1806, and HCC1954 cells were maintained in RPMI medium containing 10% FBS and vitamins (PanEco). All cell lines were incubated at 37 °C in a humidified atmosphere with 85–90% humidity. Cell proliferation was assessed using a modified MTT assay (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide; Applichem) as previously described in ref. 71 and 72.

Cells were seeded at densities of 4 × 10⁴ MCF7 cells, 5 × 10⁴ HCC1954 cells, or 6 × 10⁴ hFB-hTERT cells per well in 24-well plates (Corning) in 900 μL of medium. Compounds were initially dissolved in DMSO (Applichem) at a concentration of 5 mM and subsequently diluted to the desired working concentrations in the medium. Compounds were added in 100 μL aliquots to each well 24 hours post-seeding, and cells were incubated for an additional 72 hours. Following incubation with compounds at concentrations ranging from 0.1 to 50 μM, the medium was replaced with MTT reagent dissolved in medium to a final concentration of 0.2 mg mL⁻¹. After a one-hour incubation, supernatants were discarded, and formazan crystals were solubilized in 350 μL of DMSO per well. The plates were gently agitated, and absorbance was measured at 571 nm using a MultiSkan plate reader (ThermoFisher). Cell viability was calculated by subtracting the blank value (absorbance in wells without cells) from each measurement. IC₅₀; values were determined using GraphPad Prism software.

Immunoblotting

HCC1954 cells were plated in 100-mm dishes (Corning), and after 24 hours, compound (R)-8n (0.09–3 μM) was added to fresh medium. Cells were harvested after 24 hours of treatment with compound (R)-8n. Cell extracts were prepared by washing cells twice with phosphate-buffered saline, followed by 10 minute incubation on ice in modified lysis buffer containing 50 mM Tris-HCl (pH 7.5), 1% Igepal CA-630, 150 mM NaCl, 1 mM EDTA, 1 mM DTT, 1 mM sodium orthovanadate, 1 mM NaF, and protease inhibitors (aprotinin, leupeptin, pepstatin at 1 μg mL⁻¹ each), as previously described.⁷³ Protein concentration was determined by the Bradford assay.⁷⁴ Proteins were separated on 10% SDS-PAGE under reducing conditions, transferred to nitrocellulose membranes (Cytiva), and processed according to a standard protocol. To block nonspecific binding, membranes were treated with 5% nonfat milk in TBS buffer (20 mM Tris, 500 mM NaCl, pH 7.5) containing 0.1% Tween-20, and incubated overnight at 4 °C with primary antibodies. Primary antibodies for p-HER2, HER2, p-EGFR, EGFR, p-c-MET (Tyr1003), p-c-MET (Tyr1234/1235), c-MET, p-AKT, AKT, CDK4, HSP90, HSP70, and cleaved PARP were obtained from Cell Signaling Technology, with GAPDH as a loading control. The expression of HSP70 was additionally analyzed in the MCF7, T47D, MDA-MB-231, and HCC1806 cell lines (Fig. S4†). Horseradish peroxidase-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch) was used as the secondary antibody. Signals were detected using ECL reagents following the protocol of Mruk and Cheng⁷⁵ and visualized on an ImageQuant LAS4000 system (GE HealthCare). Fig. 3B was created using the STRING database.⁷⁶

Cell cycle analysis by flow cytometry

Cell cycle distribution was analyzed by measuring DNA content as described in ref. 29. HCC1954 breast cancer cells were seeded in 60-mm dishes and treated with the indicated concentrations of compound (R)-8n for 48 hours. Cell pellets were incubated with a PI buffer containing 50 μg mL⁻¹ propidium iodide (PI), 100 μg mL⁻¹ RNase A (Sigma-Aldrich), 0.1% sodium citrate, and 0.3% NP-40 (VWR Life Science) for 30 minutes in the dark. Cell cycle data were acquired using a CytoFLEX flow cytometer (Beckman Coulter) in the PerCP-A channel. Data analysis was performed using CytExpert (Beckman Coulter) and Microsoft Excel.

Data availability

The authors confirm that the data supporting the findings of this study are available within the manuscript and its ESI.†

Author contributions

Conceptualization and methodology, Yu. P. and A. S.; chemical synthesis, N. V. and D. P; NMR investigation, N. V and D. P; in vitro investigation, A. S., D. I. S., D. V. S., A. K., A. M.; molecular modeling studies Yu. P.; data curation, Yu. P. and A. S.; writing – original draft, review, editing, and visualization, N. V., A. M., Yu. P., D. I. S., and A. S.; resources, Yu. P. and A. S.; supervision, Yu. P., F. L. and A. S.; funding acquisition, Yu. P. and A. S. All authors have read and agreed to the published version of the manuscript.

Conflicts of interest

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.