Development of Highly Selective 1,2,3-Triazole-containing Peptidic Polo-like Kinase 1 Polo-box Domain-binding Inhibitors

Members of the polo-like kinase (Plk) family of serine/threonine protein kinases play crucial roles in cell cycle regulation and proliferation. Of the five Plks (Plk1–5), Plk1 is recognized as an anticancer drug target. Plk1 contains multiple structural components that are important for its proper biological function. These include an N-terminal catalytic domain and a C-terminal non-catalytic polo-box domain (PBD). The PBD binds to phosphothreonine (pT) and phosphoserine-containing sequences. Blocking PBD-dependent interactions offers a potential means of down-regulating Plk1 function that is distinct from targeting its ATP-binding site. Previously, we demonstrated by tethering alkylphenyl chains from the N(π)-position of the His residue in the 5-mer PLHSpT, that we were able to access a hydrophobic “cryptic” binding pocket on the surface of the PBD, and in so doing enhance binding affinities by approximately 1000-fold. More recently, we optimized these PBD-ligand interactions using an oxime ligation-based strategy. Herein, using azide-alkyne cycloaddition reactions, we explore new triazole-containing PBD-binding antagonists. Some of these ligands retain the high PBD-binding affinity of the parent peptide, while showing desirable enhanced selectivity for the PBD of Plk1 relative to the PBDs of Plk2 and Plk3.


Introduction
Members of the polo-like kinase (Plk) family play crucial roles in mammalian cell cycle regulation and proliferation [1]. Proper function of Plks 1-4 requires the coordinated phosphorylation of serine and threonine residues by N-terminal kinase domains (KDs) as well as engagement of protein-protein interactions (PPIs) with phosphoserine (pS)/phosphothreonine (pT)-containing sequences by means of their C-terminal polo-box domains (PBDs) [2]. While Plks 1-3 share significant homology, Plk4 is more distantly related [3,4]. The association of Plk1 over-expression with neoplastic transformation and tumor aggressiveness has defined it as a potentially promising anticancer molecular target [5][6][7][8].
To date, issues of collateral cytotoxicity have arisen for Plk1 kinase inhibitors. This is in part due to a lack of selectivity arising from the general homology among kinase catalytic domains. Given the uniqueness of PBDs to the Plk family, targeting PBD-mediated PPIs may allow down-regulation of Plk1 function with greater kinome selectivity than with inhibitors directed at the KD. However, Plk2 and Plk3 have roles in checkpoint-mediated cell-cycle arrest and maintenance of genetic stability, and they may serve as potential tumor suppressors [3,4,9]. Therefore, in developing PBD-binding inhibitors, it is desirable that they are selective for Plk1 versus Plk2 and Plk3. Because of the high homology among In designing Plk1 PBD-binding inhibitors, we have previously started with the polo-box interacting protein 1 (PBIP1)-derived 5-mer PLHSpT (1) (Figure 1) [7,14]. We found that up to 1000fold enhancement of Plk1 PBD-binding affinity can be achieved by appending alkylphenyl groups from the His N(π)-position (as exemplified by 2a) [15]. A crystal structure of PBD-bound 2a revealed that the alkylphenyl group is situated within a hydrophobic aromatic box defined by residues Y417, Y421, Y481, L478, F482 and Y485. This may be considered as being "cryptic" in nature, since it is revealed by rotation of the Y481 side chain [15]. We have reached the pocket from parent 1 by a variety of approaches, including tethering alkylphenyl groups from the Pro residue [16], an aminoterminal N-alkyl Gly residue [17] and from macrocyclic variants [18,19]. The pocket can also be accessed from more extended peptides, such as the amino-terminal Phe residue of the PBIP1-derived peptide FDPPLHSpTA [20][21][22]. The ability to engage the cryptic pocket has been a critical element of the highest affinity PBDbinding ligands reported to date. Within this context, the pT-2 position arguably represents the most efficient position from which to achieve this access, since it is the most proximal residue to the critical "SpT" recognition motif [23]. By examining a variety of non-proteinogenic amino acid residues at the pT-2 position, we found that the highest affinities were shown by those peptides having alkylation at the His-N(π) position, which provided approximately 50-fold higher affinity than alkylation at the isomeric His-N(τ)position (peptides 2a and 4a, respectively, Figure 1) [24]. Yet, optimizing these interactions has been made difficult due to the tediousness of preparing individual His-N(π)-alkyl analogs. In response to these challenges, we employed an oxime-based post-solid phase peptide diversification strategy that allowed us to screen more than 80 analogs. Ligands such as 2b resulted, which show enhanced Plk1 PBD affinity or selectivity relative to parent 2a [25,26].
The utility of 1,2,3-triazoles for introducing conformational constraint in peptidomimetic chemistry has been reported [27][28][29][30][31][32][33][34][35]. A triazole replacement of the imidazole ring in His has been used to prepare constrained His mimics [36]. More recently, triazole-based His mimetics bearing long-chain alkylphenyl groups have been examined within the context of non-peptidic Plk1 PBD inhibitors [37]. However, the best of these constructs showed Plk1 PBD-binding affinities that were 5-to 18-fold less potent than 1 (which itself exhibits 3-orders of magnitude less affinity than 2a) [37]. Herein, we report the use of on-resin azide-alkyne cycloaddition reactions to introduce 1,2,3-triazole functionality into potent lead Plk1 PBD inhibitors based on 2a [15] and 2b [25,26]. The triazole rings were intended either to induce conformational constraint (3a-3d) or to serve as a His mimetic (4b). This work has allowed us to prepare in facile fashion, new ligands that retain the high Plk1 PBDbinding affinity of the parent peptide, while enhancing selectivity for the PBD of Plk1 relative to the PBDs of Plk2 and Plk3. The ability to engage the cryptic pocket has been a critical element of the highest affinity PBD-binding ligands reported to date. Within this context, the pT-2 position arguably represents the most efficient position from which to achieve this access, since it is the most proximal residue to the critical "SpT" recognition motif [23]. By examining a variety of non-proteinogenic amino acid residues at the pT-2 position, we found that the highest affinities were shown by those peptides having alkylation at the His-N(π) position, which provided approximately 50-fold higher affinity than alkylation at the isomeric His-N(τ)position (peptides 2a and 4a, respectively, Figure 1) [24]. Yet, optimizing these interactions has been made difficult due to the tediousness of preparing individual His-N(π)-alkyl analogs. In response to these challenges, we employed an oxime-based post-solid phase peptide diversification strategy that allowed us to screen more than 80 analogs. Ligands such as 2b resulted, which show enhanced Plk1 PBD affinity or selectivity relative to parent 2a [25,26].
The utility of 1,2,3-triazoles for introducing conformational constraint in peptidomimetic chemistry has been reported [27][28][29][30][31][32][33][34][35]. A triazole replacement of the imidazole ring in His has been used to prepare constrained His mimics [36]. More recently, triazole-based His mimetics bearing long-chain alkylphenyl groups have been examined within the context of non-peptidic Plk1 PBD inhibitors [37]. However, the best of these constructs showed Plk1 PBD-binding affinities that were 5-to 18-fold less potent than 1 (which itself exhibits 3-orders of magnitude less affinity than 2a) [37]. Herein, we report the use of on-resin azide-alkyne cycloaddition reactions to introduce 1,2,3-triazole functionality into potent lead Plk1 PBD inhibitors based on 2a [15] and 2b [25,26]. The triazole rings were intended either to induce conformational constraint (3a-3d) or to serve as a His mimetic (4b). This work has allowed us to prepare in facile fashion, new ligands that retain the high Plk1 PBD-binding affinity of the parent peptide, while enhancing selectivity for the PBD of Plk1 relative to the PBDs of Plk2 and Plk3.

Biological Evaluation
We employed fluorescence polarization (FP) assays to evaluate binding affinities against the isolated PBDs of Plk1, Plk2 and Plk3 (Table 1, Figure S1 in Supplementary Material). Compared with the parent peptide 1 (IC50 = 650 nM) replacement of the His imidazole ring with an alkyne group resulted in an approximate 2-fold loss of Plk1 PBD-binding affinity (17, IC50 = 1000 nM, Table 1). Interestingly, peptide 13 (IC50 = 1100 nM) showed equivalent Plk1 PBD-binding affinity, in spite of the fact that it included a hex-5-yn-1-yl moiety at the His N(π)-position. This group would be expected to partially engage the hydrophobic channel leading to the cryptic pocket. We have previously shown that peptides 2a and 2b exhibit binding affinities (IC50 values of approximately 15 It was our original intent to prepare both 1,4-substituted and 1,5-substituted triazoles as mimetics of the isomeric N(τ)and N(π)-alkylated His analogs, respectively. As stated above, the CuAAC reaction provides a reliable means for selectively assembling 1,4-disubstituted 1,2,3-triazoles [40][41][42]. Accordingly, when we subjected resin-bound alkyne-containing peptide 15 to CuAAC-catalyzed cycloaddition with phenyloctylazide 16, we obtained a peptide following resin cleavage and HPLC purification, whose structure we assigned as 4b (Scheme 3). Alternatively, the ruthenium-catalyzed cycloaddition of azides with alkynes (RuAAC) has been reported to regioselectively yield 1,5-disubstituted 1,2,3-triazoles [43,44]. Based on this, we used the on-resin RuAAC-catalyzed [3 + 2] cycloaddition reaction of the alkyne group of resin 15 and azide 16 with the expectation of obtaining the isomeric 1,5-substituted triazole 18 (Scheme 3). However, the resulting peptide was identical in all respects with peptide 4b ( 1 H-, 13 C-and 31 P-NMR). At this point, a search of the literature revealed that a simple method has been reported, which permits reliable establishment of triazole regio-substitution based on chemical shifts in one-dimensional 13 C-NMR spectra [45]. The C5 signal of 1,4-disubstituted-1H-1,2,3-triazoles characteristically appears at approximately δ = 120 ppm, while the C4 signal of 1,5-disubstituted-1H-1,2,3-triazoles is usually found at δ = 133 ppm. In our case, the products obtained from both CuAAC and RuAAC chemistries provided a diagnostic signal of δ = 123.04 ppm, indicating that the 1,4-substituted triazole (4b) was obtained in both cases.
Peptides 3a-3d represent a series of analogs having 1,4-substituted triazoles tethered from the His N(π)-position by -(CH 2 ) 4 -chains. Similar to 2a and 2b, this results in a total chain extension of 8-units between the His N(π)-nitrogen and the terminal aryl group. We had previously shown that this is an optimal length by examining a series of sequentially lengthened tethers [15]. Introducing a 4-fluoro substituent (3b, IC 50 = 100 nM, Table 1) or 3-chloro-4-fluoro substituents (3c, IC 50 = 130 nM, Table 1) were intended to potentially enhance interactions with the hydrophobic cryptic pocket. However, these did not significantly alter affinity relative to 3a. The reasons for this are not clear. Although peptide 3a was designed to mimic peptide 2a, it shows an approximate 8-fold relative loss of Plk1 PBD-binding affinity (3a, IC 50 = 110 nM, Table 1). In contrast to the marked loss of affinity incurred by introducing the triazole ring to 2a, the triazole-containing mimetic of 2b showed good retention of Plk1 PBD-binding (3d, IC 50 = 25 nM). While peptides 3a-3c contain a single tethered phenyl ring, peptide 3d has a bis-aryl system. The greater extension afforded by this latter arrangement may permit better retention of binding interactions with the cryptic pocket than is afforded by peptides have a single phenyl ring. Importantly, 3d showed extremely high selectivity for Plk1 relative to Plk2 (IC 50 = 5900 nM) and Plk3 (IC 50 = 9900 nM) ( Table 1).
In contrast to peptides 3a-3d, where triazole rings were inserted into His-N(π)-tethered chains to potentially induce conformational constraint proximal to the cryptic binding pocket, peptide 4b represents a triazole mimetic of the His imidazole ring. In our previous efforts to access the cryptic pocket using a variety of amino acid derivatives at the pT-2 position, the highest affinities were obtained using His residues, with alkylation at the His-N(π) position (2a, Figure 1) being significantly preferred to alkylation at the isomeric His-N(τ) position (4a, Figure 1) [24]. In spite of the fact that the 1,4-triazole substitution pattern of 4b does not appear to optimally replicate the geometry of the His-N(π)-1,2-imidazole pattern shown by 2a, its Plk1 binding affinity (IC 50 = 17 nM) equals that of 2a (Table 1). The Plk1 PBD selectivity of 4b is slightly better than 2a against the Plk2 and Plk3 PBDs (690 nM and 3400 nM, respectively).
It is known that auto-inhibitory interdomain interactions between the KD and PBD can result in decreased potencies in assays that employ full-length Plk1 relative to assays that use isolated PBD preparations, which lack a KD component [23]. The selectivity data shown in Table 1 were obtained using fluorescence polarization assays with isolated PBDs. In contrast, Table 2 shows binding data from an ELISA assay employing full-length Plk1 ( Figure S2 in Supplementary Material). Peptides possessing His-N(π)-tethered chains showed an approximate order-of-magnitude potency reduction in the full-length assay relative to the isolated PBD assay (27-fold and 19-fold reductions for 2a and 2b, respectively). However, triazole-containing peptides experienced significantly greater losses of inhibitory potency (160-fold for 4b and 480-fold for 3d). The larger differences may indicate a reduced ability of these peptides to effectively relieve auto-inhibition or to engage the PBD cryptic pocket in the full-length construct.
Although there are no crystals structure of full-length Plk1, which might clarify the mechanisms of autoinhibition, a co-crystal structure of Map205-stabilized isolated Plk1 KD and PBD has been solved (PDB accession code: 4J7B) [46]. In this structure the KD is situated on the face of the PBD opposite the phosphopeptide-binding site. In such an orientation, the KD displaces downward an extended loop of the PBD (residues 490-510) from where it is typically observed in isolated PBD crystal structures with bound phosphopeptides. This conformational change prevents the loop from participating in an extensive network of water-mediated hydrogen bonds with the peptide phosphate group. This may be related to the ability of the KD to inhibit ligand binding to the PBD in full-length Plk1. It is unclear from this how access to the cryptic pocket would be adversely impacted in full-length Plk1 or why the triazole-containing peptides would be more sensitive to these effects. However, it is intriguing that this loop originates from the αB helix (residues 470-489), which forms an important component of the cryptic binding pocket. i See reference [15]; ii See references [25,26]; iii Fold-change relative to isolated PBD value.

General Procedures
As previously reported [26], proton ( 1 H) and carbon ( 13 C) NMR spectra were recorded on a Varian 400 MHz spectrometer or a Varian 500 MHz spectrometer (Varian, Palo Alto, CA, USA) and are reported in ppm relative to tetramethylsilane (TMS) and referenced to the solvent in which the spectra were collected. Solvent was removed by rotary evaporation under reduced pressure and anhydrous solvents were obtained commercially and used without further drying. Purification by silica gel chromatography was performed using Combiflash instruments (Telenyde ISCO, Lincoln, NE, USA) with EtOAc-hexanes or CH 2 Cl 2 -MeOH solvent systems. Preparative high pressure liquid chromatography (HPLC) was conducted using a Waters Prep LC4000 system (Waters, Milford, MA, USA) having photodiode array detection and C18 columns (catalogue No. 00G4436-P0-AX, 250 mm × 21.2 mm 10 µm particle size, 110 Å pore, Phenomenex, Torrance, CA, USA) at a flow rate of 10 mL/min. Binary solvent systems consisting of A = 0.1% aqueous TFA and B = 0.1% TFA in acetonitrile were employed with gradients as indicated. Products were obtained as amorphous solids following lyophilization. Electrospray ionization-mass spectra (ESI-MS) were acquired with an Agilent LC/MSD system (Agilent, Santa Clara, CA, USA) equipped with a multimode ion source. High resolution mass spectrometric (HRMS, ThermoFisher Scientific, Grand Island, NY, USA) were acquired by LC/MS-ESI with a LTQ-Orbitrap-XL at 30 K resolution.

General Procedure A for the Synthesis of Azides 9a-9d and 16
To a solution of bromides 8a-8d or commercially available (8-bromooctyl)benzene (7.0 mmol) in acetone (10 mL) and H 2 O (2.0 mL) was added sodium azide (1.8 g, 28 mmol) and the mixture was stirred (55 • C, 15 h). The reaction was quenched by the addition of H 2 O, extracted with Et 2 O and the combined organic phase was washed with brine, dried (Na 2 SO 4 ), concentrated and purified by silica gel chromatography to provide the target azides 9a-9d, and 16.

Determination of Inhibitory Potency in an ELISA Assay Using Full-Length Plk1
As previously reported [25,26,50] a biotinylated phosphopeptide (sequence: Biotin-Ahx-PMQS(pT)PLN-NH 2 ) was diluted to 1 µM (from a 2 mM DMSO stock solution) in PBS (pH 7.4) and loaded onto the wells of a 96-well Neutravidin-coated plate (Pierce Biotechnology, ThermoFisher Scientific, Waltham, MA, USA) at 100 µL per well for 1 h (background control contained no biotinylated peptide). The wells were washed once with 150 µL PBST (PBS, pH 7.4 + 0.05% Tween-20), and then 100 µL of 1% BSA in PBS (pH 7.4) (blocking buffer) was added for 1 h. A cytosolic lysate-containing transiently expressed myc-tagged Plk1 protein was diluted to 300 µg/mL in PBS (pH 7.4) containing protease/phosphatase inhibitors (Pierce Biotechnology), mixed with competitive inhibitor (from a 10× stock in~4% DMSO/PBS), and allowed to pre-incubate for 1 h (100 µL per well in a 96-well plate, 30 µg total protein). The blocked ELISA plate was washed 2× with PBST (PBS, pH 7.4 + 0.05% Tween-20) (150 µL) and the pre-incubated lysates were added to the plate and incubated (1 h). The wells were washed 4× with PBST (150 µL) and then treated with anti-myc primary antibody (1:1500 dilution in PBS, mouse monoclonal, Pierce Biotechnology) for 1 h. The wells were then washed 4× with PBST (150 µL), and incubated with rabbit anti-mouse horseradish peroxidase (HRP) conjugate [1:3000 dilution in 1% (%w/v) BSA in PBS, Pierce Biotechnology] for 1 h. The wells were then washed 5× with PBST (150 µL) and incubated with Turbo TMB (3,3 ,5,5 -tetramethyl benzidine substrate)-ELISA solution (Pierce Biotechnology) until the desired absorbance was reached (5-10 min). The reaction was quenched by the addition of 2 N aqueous H 2 SO 4 and the absorbance was measured at 450 nm using a BioTek Synergy 2 96-well plate reader. Absorbance was plotted versus concentration (logM) and fit to a non-linear regression analysis using GraphPad Prism 8 software [model: log(inhibitor) vs. response-variable slope (four parameters)]. The calculated IC 50 values presented in Table 2 are from multiple independent experiments and were normalized and averaged to provide values ± SEM.

Conclusions
Presented herein are the design of the triazole-containing conformationally constrained peptides 3a-3d and the His mimic-containing peptide 4b as well as their facile preparation using on-resin azide-alkyne cycloaddition reactions. The resulting peptides were evaluated in FP binding assays using isolated PBDs of Plk1, Plk2 and Plk3 and in ELISA assays against full-length Plk1. Certain of these new ligands retain the high Plk1 PBD-binding affinity of the parent peptide 2a, while having enhanced selectivity for the PBD of Plk1 relative to the PBDs of Plk2 and Plk3. It is interesting that peptides 4b and 3d show significantly greater than anticipated reduced affinities in full-length Plk1 ELISA assays relative to values obtained with the isolated PBD (160-fold for 4b and 480-fold for 3d). The larger differences may indicate a reduced ability of these triazole-containing peptides to effectively relieve auto-inhibition arising from interdomain interactions between the KD and PBD or to engage the PBD cryptic pocket in the full-length construct. These observations are noteworthy, in that they potentially indicate structural interactions of the KD and PBD in full-length Plk1 that are not anticipated by the previous co-crystal structure of isolated KD and PBD in the presence of Map205.