A ligand-centric approach to identify potential drugs for repurposing: case study with aurora kinase inhibitors

Introduction

Despite remarkable advancements in understanding the molecular basis of cancer, it remains one of the major causes of death worldwide. However, improved understanding over the decades has led to a paradigm shift in cancer therapy from treatment with cytotoxic chemotherapeutics to more selective targeted therapy focusing on the key molecular players in the cancer cell signaling pathways [ ]. Protein kinases, which are the second most targeted group of drug targets after G protein–coupled receptors, play important role in cancer signaling pathways. More than 30 kinase inhibitors are currently available in the market as approved drugs for treatment of various types of cancer [ ]. Eukaryotic genomes encode protein kinases, such as the Aurora kinases, which promote cell proliferation, survival, and migration. The hallmark of any form of cancer is rapid uncontrolled cell division and loss of apoptosis, which might originate in one part of the body and latter invade to adjoining parts to spread to other organs (metastasis). Therefore, targeting Aurora kinases would be an effective way toward management of many types of cancer as have been seen for the traditional antimitotic agents like taxol and its derivatives, which disrupt the mitotic spindle and thus arrest the cell division process [ ].

The Aurora kinases (A, B, and C) are highly related serine/threonine protein kinases, which transfer the ɣ-phosphate group from ATP to serine/threonine residue of a substrate protein and play important role in mitosis. The ATP binding site is highly conserved within the Aurora kinase subfamily and across all Ser/Thr/Tyr kinases. Thus, targeting this site for protein inhibition is challenging as it leads to undesired off-target effects. However, reports are available where single amino acid difference in the ATP-binding pocket sequence has been exploited to achieve Aurora-Aselective inhibitor over its isoforms [ , ]. Aurora-A localizes to the centrosome from centrosome duplication through mitotic exit and largely plays a role in regulating the centrosome and forming mitotic spindle [ ]. Dysregulation of Aurora-A has been linked to tumorigenesis, and studies have established it as a bona fide oncogene [ ]. The active conformation of Aurora-A is induced by autophosphorylation, which is mediated by several cofactors [ , ]. One of the important autophosphorylation-mediating factors of Aurora-A is TPX2 (targeting protein for xenopus kinesin-like protein 2). Interaction between Aurora-A and TPX2 has been shown to involve three druggable hotspots. These three spots have been explored for allosteric inhibition of Aurora-A, which could potentially be helpful in achieving kinase selectivity ( Fig. 2.1 ) [ ]. Recent studies have shown that TPX2/Aurora-A signaling is a potential therapeutic target in genomically unstable cancer cells [ ]. Aurora-B is a subunit of chromosomal passenger protein and controls accurate chromosome segregation and cytokinesis. The inner centromeric protein and survivin are the major substrates of activated Aurora-B. Experimental reports suggesting the significance of Aurora-B in tumorigenesis are available [ ]. Aurora-C plays role in spermatogenesis and is specifically expressed in testis. The role of Aurora-C in oncogenesis was not very clear [ ]. However, reports in the current decade hint the implication of aberrant expression of Aurora-C kinase in various forms of cancer like breast cancer, prostate cancer, etc. [ ]. Aurora-A/B have been extensively exploited as attractive anticancer drug target since their discovery. A handful of inhibitors for these kinases have been reported in the past few years where some of those either specifically target Aurora-A kinase (MLN8054, MLN8237, etc.) [ ] or Aurora-B kinase (Hesparadin, SU6668, etc.) [ ], while many others have been found to target both Aurora-A and Aurora-B kinases (such as, AZD1 152) [ ]. Reports of pan Aurora kinase inhibitors, i.e., Aurora-A, Aurora-B, and Aurora-C inhibitors (danusertib, PF-03824735, etc.) are also available [ ]. These inhibitors and many other promising Aurora kinase inhibitors have been considered for clinical trial against various forms of cancer that have been elaborately reviewed in a number of articles [ , , , ]. However, currently, no approved drugs targeting Aurora kinases are available. This situation can be largely attributed to the most common challenge associated with any kinase drug discovery programs, i.e., developing selective kinase inhibitors to minimize off-target mediated toxicity [ , , , ]. Efforts to identify more new allosteric inhibitors could be helpful in overcoming the problems of toxicity and achieving selectivity as the allosteric sites are evolutionary less conserved than the orthosteric ATP binding site [ ]. Unfortunately, in general, anticancer drugs (irrespective of targets) have a very high attrition rate (∼95%) [ ] and the lengthy timescale involved in traditional drug discovery process further complicates the situation. Therefore, to meet the unmet medical need, a faster drug discovery process is desirable so that failures are expected to be detected quickly. In this regard, drug repurposing approaches to identify approved molecules that could potentially inhibit Aurora kinases would be helpful. Owing to the established pharmacokinetic and pharmacodynamic profiles of the molecules identified through drug repurposing approaches, the overall complexity in finding a right binding partner for the target of interest is generally lesser as compared to traditional drug discovery process, especially when available state-of-the art in silico techniques are rationally integrated with experimental findings [ , ]. Further, the established clinical safety profile of approved drugs reduces the risk of toxicity-related failures.

Broadly, computer-aided drug discovery/repurposing programs that use structural information can be either target-centric or ligand-centric or a combination of both [ ]. Target-centric structure-based approaches commonly known as structure-based drug design (SBDD) use the information on three-dimensional (3D) structures of the host molecules, which are most often protein targets. Ligand-centric or ligand-based drug design (LBDD) approaches use the information on 2D/3D structures of the guest molecules, which are known to bind to the target of interest and are often small organic compounds [ ]. The most popular SBDD approach involves molecular docking simulations, which focus on sampling the correct conformation of a given ligand in the protein-binding pocket. The predicted poses of each compound in the screening library are then ranked based on a scoring function that aims to calculate the energy of interaction between the binding partners. These scores are one of the determining factors to distinguish between probable weak and strong binders. However, scoring functions are formulated based on many approximations and the conformation sampling can also be inadequate. Hence, the results need to be interpreted cautiously [ , , ]. In the absence of a reliable 3D structure of the host molecules, SBDD approaches cannot be employed. In such situations, LBDD may be the approach of choice. LBDD involves strategies that rely on using molecular fingerprints such as shape, functional groups, and/or other descriptors of known ligand/s as filters to screen databases of chemical compounds for searching new molecules that possess similar fingerprints. However, lack of consideration of target's structural information does not give insights on probable interactions between the binding partners [ , ]. Therefore, integration of LBDD and SBDD approaches are gaining popularity and have been found to perform better in certain instances [ , ]. Besides structure-centric approaches, there are many other computational drug repurposing approaches that are routinely used, such as network-based approaches, pathway mapping, and text mining as reviewed in some of the recent articles [ , ].

In this chapter, we have primarily demonstrated the use of a ligand-centric drug discovery/repurposing approach that exploits the information on the shape and stereo-electronic features of the known binders of Aurora kinases to identify new molecules of similar shapes and chemical features. The underlying idea for such an approach lies in the fact that a ligand which is known to bind to a target of interest is complementary to the shape and electrostatics features of the target binding site. Therefore, if another molecule with similar shape and stereo-electronic features as that of the known binder is identified, it is very likely to fit in the binding site of interest in the target protein and establish similar interactions with the protein as that of the known binder [ ]. This approach is relatively faster than purely target-based drug discovery approaches, and its success over protein-centric approaches taken by docking has been reported in several instances [ , ]. The main advantage of shape-based ligand-centric approaches is that these methods do not require the information on the coordinates of the target and are comparatively faster. Availability of chemical structure of just one active molecule known to bind to the target of interest is enough to employ this technique. Nonetheless, in a study on virtual screening and lead hopping at Wyeth Research [ ], it has been demonstrated that integration of target's information with shape-based approaches improve results. The two-layered approach employed by researchers at Wyeth involved elimination of candidates that not only gave good shape- and feature-based overlay but also clashed with the protein-binding pocket. In the current study, although we have primarily applied ligand-centric approach, at a later stage we have used the target's information by integrating molecular docking simulation to our pipeline in order to predict the binding affinity and binding mode of the selected sub-set of drugs against Aurora-A kinase. Our analysis aided in identifying number of potential Aurora kinase inhibitors from the repertoire of approved drugs, which could be considered for further investigation to examine their anticancer properties. Few selected cases have been presented and discussed here since elaborate discussion of all the findings is beyond the scope of this concise chapter.

Methodology

The overall workflow of the protocol that we have adopted in this study to identify potential Food and Drug Administration (FDA)–approved drugs which could be repurposed for treatment of cancer by targeting against Aurora-A and B kinases is summarized in Fig. 2.2 .