Current progress and future perspectives in the development of anti-polo-like kinase 1 therapeutic agents

Improving specificity through structure-assisted optimization

Further development of Plk1 ATP-competitive inhibitors can be achieved through structure-guided medicinal chemistry targeting unique residues lining the ATP-binding pocket. Comparative analyses of the X-ray co-crystal structures of the catalytic domains of Plk1–3 revealed that the overall shapes of their ATP-binding pockets are similar, and the residues forming the pockets are mostly analogous to one another. Nevertheless, both F58 and R134 residues are unique to Plk1, whereas R57, L132, and R136 residues found in Plk1 and one of Plk2 and Plk3 ATP-binding pockets are semi-specific to Plk1²⁴ (Figure 3). These observations suggest that there may still be explorable chemical space, which can be exploited to further increase Plk1-binding specificity. Notably, volasertib and its parental dihydropteridinone derivative, BI 2536⁴², do not interact with the Plk1-specific F58 and R134 residues but are in contact with the semi-specific R57, L132, and R136 residues and other neighboring residues (C67, L132, and F183) that are somewhat selective against non-Plks (Figure 3). This finding explains in part why they exhibit a high level (~100–1,000-fold) of selectivity against other kinases but show less discriminatory activity against Plk1–3 (IC₅₀ values of 0.83, 5, and 56 nM for Plk1, Plk2, and Plk3, respectively)^25,43.

Figure 3. The binding modes of ADP (left panel, PDB: 3D5W) and volasertib (right panel, PDB: 3FC2) to the catalytic domain of polo-like kinase 1 (Plk1).

Ligand-binding regions that are important for achieving kinase selectivity are shown as colored surfaces. Region I is located near the conserved Asp-Phe-Gly (DFG) motif, which is important in the design of certain allosteric inhibitors. Region II is near the hinge region, and region III is a relatively hydrophobic pocket below the adenine-binding site. Compared to Plk2 and Plk3, residues that are specific (green) or semi-specific (red) to Plk1 are labeled. Notably, the aryl-methoxy group in region II (near L132 of the hinge region) is used to produce moderate selectivity for Plk1 against other Plks⁴⁴.

As noted above, one of the few differences between the ATP-binding pockets of Plk1 and Plk2 occurs near the hinge region (region II in Figure 3), where L132 of Plk1 corresponds to Y161 in Plk2. This difference has been exploited to generate inhibitors that demonstrate ~1,000-fold selectivity for Plk2 versus Plk1⁴⁴. Such work highlights what can be achieved using structure-based drug design to take advantage of small differences in protein-binding sites. An additional effort to target the Plk1-specific F58 and R134 residues that lie at the mouth of the core ATP-binding pocket (Figure 3) could lead to inhibitors with improved selectivity against other Plks. The generation of isoform-selective inhibitors is critically important to elucidate the biological roles of each individual Plk isoform, along with determining the importance of isoform selectivity in preventing the currently observed dose-limiting toxicities. Insights derived from such studies could further reveal the potential therapeutic value of targeting the catalytic domain of Plk1.

Developing allosteric inhibitors

To ensure high selectivity, ATP-competitive small molecules must be designed to maximally engage in receptor-ligand interactions or sterically fill unoccupied space within defined binding pockets. In addition, because allosteric sites are generally less conserved than orthosteric sites, developing strategies to target allosteric sites, if available, can be potentially very useful in deriving specific inhibitors against a target. Close examination of the co-crystal structures of the Plk1 catalytic domain in complex with either ADP or volasertib reveals three ligand-binding regions (regions I, II, and III) that are important for achieving kinase selectivity (Figure 3). Volasertib appears to utilize each of these orthosteric binding regions to achieve selectivity against most clinically important kinases. It is noteworthy that the region I located near the conserved Asp-Phe-Gly (DFG) motif contains a shallow space that could be potentially utilized to further improve the drug–target interaction. Interestingly, a dimethoxybenzeneethanamine derivative, SBE13, is suggested to bind to Plk1 in its “DFG-out” conformation and inhibit the enzyme’s catalytic activity with outstanding specificity^45,46. An allosteric hydrophobic pocket generated in the catalytically inactive, DFG-out conformation of many Tyr kinases and, less frequently, Ser/Thr kinases has been successfully targeted to improve the selectivity of an inhibitor^47–49. However, no structural information is yet available to verify whether Plk1 can indeed embrace a DFG-out conformation and generate an allosteric hydrophobic pocket that potentially allows an ATP-competitive inhibitor to trap the enzyme in a catalytically inactive state.

Another potentially appealing strategy toward achieving target-specific allosteric inhibition of Plk1 could be to block its essential substrate-binding interface. In principle, this may be approached by understanding the nature of the Plk1-substrate binding interactions that are critical for proper mitotic progression. ON01910, a substituted benzyl styryl sulfone, was initially reported as a non-ATP-competitive inhibitor of Plk1 that was thought to interfere with the substrate-binding site⁵⁰. However, subsequent studies showed that it possesses rather poor anti-Plk1 activity and specificity and inhibits multiple kinases^43,51–53. In a broad sense, the anti-PBD agents described below can be classified as target-specific allosteric inhibitors, which inhibit Plk1 function by binding to a specific site distant from the ATP-binding pocket.

Peptidomimetics

Since the initial discovery of a pentameric p-Thr-containing peptide, PLHSpT, as a specific Plk1 PBD-binding ligand⁵⁵, several high-affinity peptide-derived inhibitors, including 4j (Figure 4), have been generated^56–58. These efforts have revealed that three distinct structural elements are critical for achieving high affinity and specific binding to the Plk1 PBD (Figure 5). These are 1) a phosphoepitope-recognition pocket containing two basic residues (His538 and Lys540), which binds to Ser-Xxx motifs, where Xxx is p-Thr (or several-fold lower-affinity p-Ser) or a suitable anionic mimetic^55,59, 2) an adjoining Pro-binding region that engages portions of the bound ligand N-proximal to the Ser-p-Thr dipeptide motif⁵⁵, and 3) a hydrophobic channel, which is capable of boosting the binding affinity ~500–1,000-fold without compromising Plk1 PBD specificity^56–59. However, despite the fact that these peptide-based inhibitors exhibit extremely high affinity and specificity in vitro, their utility in cellular contexts is greatly decreased by poor membrane permeability and limited bioavailability. Efforts to increase cellular uptake have been directed at the reduction of anionic charge through charge masking and prodrug protection⁶⁰, macrocyclization^61,62, and reducing peptide character⁶³.

Figure 4. The structures of peptidomimetic and small molecule polo-box domain (PBD) inhibitors.

Figure 5. The structural model of polo-box domain (PBD) in complex with 4j.

The X-ray cocrystal structure of the PBD + 4j complex (PDB: 3RQ7) shows a “Y-shaped” binding pocket composed of three discrete but interlinked binding modules—namely, a p-Thr/p-Ser-binding module (violet), a Pro-binding module (yellow), and a hydrophobic channel (green). Residues highlighted in red are specific to Plk1 PBD. Inhibitors designed to bind to more than one of the three binding modules could possess a superior binding specificity because of the specific requirement of the shape and geometrical arrangement of their binding moieties (see text for details).

Small molecule inhibitors

Paralleling the development of peptidomimetics as described above are efforts to generate anti-Plk1 PBD agents that have yielded a wide range of small molecule inhibitors, including thymoquinone (TQ) and poloxin⁶⁴, poloxipan⁶⁵, purpurogallin (PPG)⁶⁶, T521⁶⁷, Cpd161⁶⁸, and (−)-epigallocatechin⁶⁹ (Figure 4). As expected, if they interfered with the function of Plk1 PBD, these inhibitors induce mitotic block and apoptotic cell death, at least at the cultured cell level. While X-ray crystallographic data suggested the binding nature of how TQ and its analogue poloxime bound to the p-Thr/p-Ser-binding module⁷⁰, the assignment of these ligands within the phosphoepitope-recognition module appears to be disputable because of the poor resolution of the reported crystal structures. In addition, whether TQ and poloxin can be used as a template for structure-assisted drug design and hit-to-lead optimization remains in doubt because of their low Plk1 PBD specificity and non-specific alkylating activity with other cellular proteins^68,71,72. Indeed, covalent protein reactivity and other chemotypes, such as polyhydroxylated aromatic compounds (e.g. PPG), are thought to be some of the common artifacts that lead to the isolation of pan assay interference compounds (PAINS)⁷³ (as discussed in depth in a recent review⁷⁴).

Multidirectional approaches to identify a lead with druggable properties

Even though currently available peptide-derived and small molecule inhibitors have provided valuable information in understanding the nature of Plk1 PBD-dependent interactions and their cellular functions, it remains uncertain whether they can be used as leads for developing anti-PBD agents. Therefore, while these inhibitors are being further developed, additional endeavors could be considered in parallel in the hope of generating a new class of inhibitors with more druggable properties. Identifying promising leads possessing outstanding biochemical selectivity (i.e. specificity and affinity) and druggable physicochemical properties (e.g. solubility, bioavailability, chemical functionality, etc.) will likely be a critical step for successful lead optimization and ultimate discovery of anti-Plk1 PBD therapeutic agents.

Exploiting the shape and geometrical arrangement of three adjacent binding modules of the Plk1 PBD-binding pocket to achieve high specificity

Now it is clear that, unlike the nondescript nature of many PPI surfaces, the Plk1 PBD furnishes a well-defined binding interface that may make it more amenable for structure-based drug discovery. Accumulated data show that the affinity and specificity of Plk1 PBD binding are dependent on the ability of a ligand to interact with three adjacently placed but biochemically distinct binding modules^55,56 that collectively form a “Y-shaped” binding pocket (reviewed in Lee et al.²⁴) (Figure 5). Future anti-Plk1 PBD drug discovery could take advantage of the Y-shaped interaction interface, so that ligands are optimized as three discrete but interconnected binding-module platforms.

Recent work has demonstrated that suitably designed p-Thr mimetic derivatives optimized for the phosphoepitope-recognition module can achieve several-fold enhancement in PBD-binding affinity⁷⁵. In addition, structural elaboration of hydrophobic channel-binding functionality and improvement of the interactions with Plk1-specific residues found in the surroundings of the hydrophobic channel have significantly increased overall ligand affinity and selectivity^76,77. These findings suggest that the binding affinity and specificity of each of the three binding modules are determined independently of one another. Since a ligand capable of simultaneously binding to two or more binding modules must additionally conform to the exact geometrical arrangement of these modules, anti-Plk1 PBD agents generated through the Y-shaped, three-binding-module platform (Figure 5) may likely possess uncommon physical shapes and chemical properties that may allow them to reach a high level of specificity. Multiple modular bindings discussed here may also help diminish the probability of developing drug resistance.