Hsp90 is an attractive drug target in cancer due to its wide range of oncogenic client proteins [1]. Hsp90 helps to overcome various stresses that cancer cells experience by regulating various signalling proteins and pathways. There have already been decades of research in developing Hsp90 inhibitors for treatment of cancer and yet almost no drug successfully passed clinical trials. One of the proposed reasons for the encountered issues is that the drugs were pan-inhibitors, i.e., were not isoform selective [2]. As a result, currently a lot of effort is directed towards developing isoform selective Hsp90 inhibitors.

Hsp90 is a molecular chaperone that is responsible for maintaining proteostasis in the cell. It functions by aiding with the folding, degradation, and activation of various clients. Hsp90 is composed of three domains: N-terminal domain containing an ATP binding pocket, middle domain that binds client proteins, and C-terminal domain that is needed for dimerization. In humans, there are four Hsp90 isoforms: two cytosolic isoforms Hsp90α and Hsp90β, mitochondrial TRAP1, and GRP94, found in the endoplasmic reticulum [3]. The N-terminal domains of the two cytosolic isoforms share over 95 % of identity, with the main difference of their ATP pockets being two amino acids: Ser52 and Ile913 in α and their counterparts being Ala and Leu in β [4]. In spite of this, these isoforms have distinct functions and are regulated differently [5]. Hsp90β is constitutively expressed at significantly higher levels than Hsp90α and thus functions as the general housekeeping chaperone. While Hsp90α baseline expression is lower, it can be induced by cellular stress (oxidative stress, heat, cancer) and also by various agents such as Hsp90 inhibitors [2,6]. The organelle-specific Hsp90 isoforms are significantly less abundant [7] and have more diverged sequences.

Most Hsp90 inhibitors target the ATP binding pocket in the N-terminal domain since ATPase activity is mandatory for effective regulation of folding and maturation of the client proteins [8]. First discovered Hsp90 inhibitor was geldanamycin, a natural product of S. hygroscopicus bacteria [9]. The compound exhibits poor solubility and is too hepatotoxic at therapeutic doses to use in a clinical setting. Thus, many geldanamycin derivatives, such as 17-AAG and IPI-504 [10], were created to avoid these issues and have even entered clinical trials. Later new classes of inhibitors emerged, such as ATP mimetics comprising of the purine scaffold [11] or derivatives of another natural Hsp90 inhibitor – radicicol [12,13].

Dozens of Hsp90 inhibitors have entered clinical trials, all targeting the N-terminus of Hsp90 [14]. Clinical trials [15] revealed that some of these compounds caused hepatoxicity, which limited the upper dose limit, while mild ocular toxicity was observed in a few of the other cases [16]. To date, only one inhibitor has successfully passed clinical trials and was approved for use in humans — pimitespib (TAS-116) for treatment of gastrointestinal stromal tumours [[17], [18], [19]].

In general, cancer cells are much more sensitive to depletion of Hsp90 proteins as compared to normal cells [20]. However, different cancers vary significantly in their molecular biology and response to treatment. In the cancers that do not already overexpress Hsp90, Hsp90 inhibition strategy has been further complicated by the fact that the inhibitors would often cause heat shock response (overproduction of Hsp90 induced by depletion of active Hsp90α), which in turn requires to increase dosage to see efficacy [21]. As a result, in such cases Hsp90β selective inhibitors, which do not cause the significant overexpression of Hsp90 and associated effects while still inducing death of cancer cells [22,23], are more suitable. For example, the recently developed KUNB31, a Hsp90β selective inhibitor, was shown not to induce a heat shock response and might be a solution to the detriments caused by pan-inhibition [23,24]. On the other hand, targeting Hsp90α is more effective in case of Hsp90-overexpressing cancers [5,25].

In our previous work we discovered several Hsp90 inhibitors containing the aryl-resorcinyl-thiadiazole scaffold that exhibit subnanomolar dissociation constants towards human cytosolic Hsp90 isoforms and submicromolar EC50 against various human cancer lines [26,27]. Here we investigate the possible benefits of substituting the five-member ring – thiadiazole – with imidazole in terms of ease of synthesis, selectivity for different Hsp90 isoforms, and their anticancer properties in cancer cell cultures. To gain a more in-depth understanding of ligand isoform selectivity, we constructed a Hsp90α ATP-binding site S52A mutant (Hsp90αN (S52A)) to make it more akin to that of Hsp90β. This allowed us to see whether this single residue is enough to flip the selectivity of our tested ligands or do the other residues also play a significant part in the preferential binding. We further investigated the individual interactions in the Hsp90—ligand complexes using a combination of docking and molecular dynamics (MD) techniques. This allowed us to further our understanding of molecular details of the interactions between the Hsp90 ATP-binding site and resorcinol-based ligands for future drug design.