Scaffold hopping from (5-hydroxymethyl) isophthalates to multisubstituted pyrimidines diminishes binding affinity to the C1 domain of protein kinase C

Protein kinase C (PKC) isoforms play a pivotal role in the regulation of numerous cellular functions, making them extensively studied and highly attractive drug targets. Utilizing the crystal structure of the PKCδ C1B domain, we have developed hydrophobic isophthalic acid derivatives that modify PKC functions by binding to the C1 domain of the enzyme. In the present study, we aimed to improve the drug-like properties of the isophthalic acid derivatives by increasing their solubility and enhancing the binding affinity. Here we describe the design and synthesis of a series of multisubstituted pyrimidines as analogs of C1 domain–targeted isophthalates and characterize their binding affinities to the PKCα isoform. In contrast to our computational predictions, the scaffold hopping from phenyl to pyrimidine core diminished the binding affinity. Although the novel pyrimidines did not establish improved binding affinity for PKCα compared to our previous isophthalic acid derivatives, the present results provide useful structure-activity relationship data for further development of ligands targeted to the C1 domain of PKC.


Introduction
Protein kinase C (PKC) comprises a family of ten phospholipid-dependent serine/threonine kinases [1,2], which regulate several cellular processes including proliferation, migration, cell survival and apoptosis [3][4][5]. Due to its central position in intracellular signaling, PKC is also involved in the pathogenesis of various diseases, including diabetes, cancer, ischemic heart disease and heart failure, some autoimmune diseases, Parkinson's disease and in Alzheimer's disease [2]. The fact that PKC is linked with so many diseases makes it a very attractive subject of research and a potential target for therapeutic discoveries. PLOS  versions of HMI-1a3 and -1b11. We also kept short ethyl derivatives (1d and 1h) to investigate eventual alkyl chain length-dependent loss of activity. The design of the unsymmetrical 2,4,5,6-tetrasubstituted pyrimidines instead focused a deeper investigation on the symmetryrelated activity with compounds featuring the same substituents (2a-f) or different combinations (2g-l) switching them between the ether and ester moieties in positions C4 and C6, respectively.

Modeling
To design a set of pyrimidines we referred to the crystal structure of the phorbol 13-acetate bound PKCδC1B (Protein Data Bank code: 1PTR) [18] and to the knowledge of the key functional groups of the HMIs gained from our previous study [12]. The co-crystallized phorbol acetate forms hydrogen bonds with the amino acids Thr242, Leu251 and Gly253 in the hydrophilic pocket of the C1 domain while it completes the hydrophobic surface of the protein through hydrophobic interactions with Leu251, Leu254 and Met239 (Fig 2A). According to our previous docking study, the HMIs are able to bind to the active site in similar manner showing also a possible additional attractive interaction between Gln257 and the π-electrons of the aromatic core ( Fig 2B). When comparing the previous docking poses with those of the new pyrimidines, the interaction pattern between the ligands and the backbone amino acids of the polar pocket (i.e. Thr242, Leu251 and Gly253) (Fig 2) remain alike. The hydrophobic interactions, instead, show a bit more variation, as the pyrimidines may interact also with for instance Pro241 and Leu250 in addition to Leu251, Leu254 and Met239. (Fig 2C and 2D).

Synthesis
We prepared the symmetrical 2,4,6-trisubstituted pyrimidines in a two to three-step synthesis (Fig 3). We started with an inverse electron demand Diels-Alder reaction reported on related compounds by Duerfeldt, Anderson and coworkers [19,20]. A commercially available diethyl 1,2,3-triazine-4,6-dicarboxylate (3) was reacted with 2-(4-methoxyphenoxy)acetamidine (4) to obtain the 2,4,6-trisubstituted pyrimidine 5 containing a p-methoxyphenyl (PMP)-protected hydroxymethyl moiety at the C2-position. The PMP protection allows the treatment of 5 with different alcohols in the presence of a catalytic amount of sulfuric acid and transesterification of the esters in positions C4 and C6 to give the intermediates 6a-g. Finally, the PMP was easily removed by an oxidative cleavage reaction applying conditions reported by Lee [21] with minor modifications. We treated the intermediates 6a-g with ceric(IV) ammonium nitrate (CAN) to give the desired products 1a-g while the same conditions applied directly to the intermediate 5 gave the final product 1h.
We performed a four to five-step synthesis to obtain the unsymmetrical 2,4,5,6-tetrasubstituted pyrimidines (Fig 4). In the first step, reported on related compounds by Otsuka and coworkers [22], we reacted the commercially available diethyl oxalpropionate (7) and 2-(4-methoxyphenoxy)acetamidine hydrochloride (8) in the presence of triethylamine (TEA) in ethanol to obtain pyrimidine 9 containing a PMP-protected hydroxymethyl moiety in C2-position. The substituted pyrimidine 9 was treated with phosphoryl bromide in N,Ndimethylformamide (DMF) to give the aryl bromide 10 with the C4-position activated for the subsequent nucleophilic substitution. Different alcohols were treated with NaH to generate the respective alkoxides which reacted with intermediate 10 on both positions C4 and the carbonyl moiety to give pyrimidines 11-13 in low yields. Instead, the carboxylic acids 14-18 were formed during the reaction and were isolated for an esterification reaction to give compounds 19-27. The carboxyl groups of 14-18 were esterified with different methods including: 1) treatment with SOCl 2 in an alcohol as a solvent, 2) activation with 1,1'-carbonyldiimidazole (CDI) and treatment with different alcohols in the presence of 1,8-diazabicyclo [5.4.0]undec-7-ene (DBU) and 4-(dimethylamino)pyridine (DMAP) and 3) treatment with trimethylsilyldiazomethane. For the PMP-deprotection step, intermediates 11-13 and 19-27 were treated with CAN to give the final compounds 2a-l.

Chemography and ChemGPS-NP
To compare the physicochemical properties of the novel pyrimidines with those of other PKCα ligands we carried out a chemographic mapping including also the HMIs and some of the most potent PKCα binders (for the complete list of the compounds see Materials and methods and S1 File). We used the ChemGPS-NP Web tool [23][24][25], a principal component analysis-based chemical global positioning system, which allows to plot organic compounds in a two/three-dimensional chemical space assigning a position based on their structurederived physicochemical properties. We converted the structures of the compounds into SMILES (simplified molecular-input line-entry system) and uploaded them to the ChemGPS-NP Web server (http://chemgps.bmc.uu.se) which generated for each of them eight principal components (dimensions PC1-8). Each PC describes different physicochemical properties based on 35 descriptors and the four most significant PCs (PC1-4) represent 77% of data variance. PC1 accounts for size, shape and polarizability, PC2 comprises aromaticity and conjugation properties, PC3 includes lipophilicity, polarity, and H-bond capacity while PC4 represents flexibility and rigidity [24]. We plotted the ligands in a three-dimensional space setting PC1, PC2 and PC3 as the x, y and z axes, respectively, with conical arrows indicating the positive sides ( Fig 5). The full list of the compounds, ChemGPS-NP raw data, SMILES and structures are available in S1 File.
The 3D-plot shows clearly how most of the best binders, the pyrimidines and HMIs are separated by PC2 in 4 bands, then distributed along PC1 by their size and along PC3 by their lipophilicity. In this analysis PC2 is the most significant dimension and, as explained previously, it represents aromatic and conjugation properties of the compounds: the more aromatic rings/ conjugated systems feature in the structure of a compound the higher is the PC2-value the compound obtains. The structures of all the potent binders, except mezerein, 9-decyl-benzolactam-V8 and indolactam-V (Fig 5, cyan and magenta spheres respectively), feature only few π-conjugated systems and no aromatic moieties in both core structure and substituents. This explains why they obtained lower PC2-values compared to the other compounds and thus they aligned together on the most negative side of PC2. The aforementioned three potent binders, which instead did not align with the rest of the ligands with high affinity, present a nonaromatic core but some aromatic features in their substituents that explains their higher PC2-values. All the other compounds feature, instead, an aromatic core which increases their PC2-values to form the two central bands of pyrimidines/HMIs bearing aliphatic substituents while those with aromatic substituents clustered at the most positive side of PC2. Then PC2 highlights clearly the lower aromatic contribution of the pyrimidine ring compared to the phenyl ring with all the pyrimidines separated, with lower PC2-values, from their HMI analogs. The alignment of the pyrimidines, slightly closer to the most active compounds compared to the HMIs, suggested that even better activity might occur. Unfortunately, the biological data did not however support this hypothesis.

Biology
We tested the compounds for binding to the C1 domains of PKCα with a 96-well plate filtration assay as described earlier, at a concentration range of 0.2-30 μM [12,26]. To our surprise, none of the new compounds displaced [20-3 H]phorbol-12,13-dibutyrate ([ 3 H]PDBu) as efficiently as HMI-1a3. The comparison of the displacement ability between the compounds 1ac, 1e, 1f and 2a-c aimed to reveal a correlation between the length of the linear side chain and the binding affinity ( Fig 6) (raw data available in S2 File). The differences, however, were very low and no trend can be established. Compounds 1d, 1h, 2g, and 2k demonstrate that the core structure requires longer alkyl side chains on both sides to achieve detectable binding. Surprisingly, the corresponding pyrimidine version of HMI-1a3, 1g, could not displace [ 3 H]PDBu at the concentration range used (Fig 7) (raw data available in S2 File). However, the HMI-1b11 analog 1e was one of the most effective novel compounds to displace [ 3 H]PDBu from PKCα. Its affinity was however considerably lower than that of HMI-1b11 determined in our previous work [12]. Compounds 2a and 2l showed the strongest concentration dependence (Fig 7). In terms of lipophilicity, most of the novel compounds showed a lower clogP value compared to HMI-1a3 (Fig 6).

Discussion and conclusion
The C1 domain of PKC represents a potential target for discovery of therapeutic drugs for diseases with unmet medical needs [6]. Plant and animal derived natural C1 domain ligands, such as phorbol esters and bryostatins, show high affinity and biological activity but they are not optimal drug candidates as their complex chemical structures make their synthesis tedious. In our previous work, we have demonstrated that simple 5-hydroxymethyl isophthalic acid derivatives exhibit promising biological activity [9,10,12,13]. The lipophilicity values for the HMIs (clogP 6-7) are however higher than the Lipinski's drug-like lipophilic property value (logP 5) [27] and therefore, we endeavored to synthesize a new set of compounds with reduced lipophilicity and retained/increased binding affinity.
In the present study, we designed and developed a novel set of PKC C1 domain-targeted pyrimidines. Despite their similarity to the HMIs in terms of structure and predicted binding mode, they were not able to displace [ 3 H]PDBu from the C1 domain of PKCα at the concentration range tested. Surprisingly, not even 1e and 1g showed similar binding to PKCα as the corresponding HMI-1b11 and HMI-1a3, respectively. This overall outcome was not expected based on the docking model of the pyrimidines, which returned docking scores in the same range as for the HMIs and suggested the same binding interactions. In addition, the chemographic data from the ChemGPS-NP 3D-plot displayed the pyrimidines slightly closer to the more potent binding compounds suggesting, at first, a possible activity. In line with the negative feedback from the biological data, a reinterpretation of the chemographic study highlights that (1) the overall lower affinity of the HMIs, compared to more potent ligands, may be due to the aromatic/planar nature of the core structure; (2) the presence of aromatic substituents have no effect or may favor the affinity; and (3) the scaffold hopping towards a heterocycle, pyrimidine in this case, caused the loss of activity.
In addition to PKCα, the HMIs bind to PKCδ and other protein families containing a DAG-responsive C1 domain (e.g. β-chimaerin, protein kinase D1 and myotonic dystrophy kinase-related Cdc42-binding kinase [MRCK]) at comparable affinities [9]. The present work demonstrates the binding affinities of the pyrimidines only for PKCα. As many other C1 domain ligands, these compounds might show substantial differences in binding affinity towards different PKC isoforms or single C1 domains [28]. However, due to the analogy with the HMIs we expect that the almost complete lack of binding of the pyrimidines to PKCα may indicate only weak or no affinity to other C1 domains as well. This is why we did not proceed to characterize the binding of pyrimidines for those.
To improve the affinity and selectivity of C1 domain ligands Ohashi and coworkers recently presented a novel set of dimeric DAG-lactone derivatives [17]. These dimeric lactones showed no enhanced binding affinity to the full-length PKCα or -δ compared to their monomeric constructs, and they indicated higher lipophilicity (clogP values: 10.7-16.7). However, they showed stronger binding to the individual PKCδC1B domain than the monomer. Physiological relevance of this finding is unclear, as affinity for the full-length protein was not increased. Elhalem and coworkers studied the C1 domain selectivity of indololactones, bearing a heterocyclic ring at the sn-1 or sn-2 position, for PKCα, -δ and Ras guanine nucleotide-releasing protein (RasGRP1) [29]. They demonstrated selectivity for RasGRP1 over PKCα when the indole ring is in the sn-2 position of indololactones [30]. Binding affinity for PKCα, -δ and RasGRP1 as well as selectivity for RasGRP1 decreased when substituted at the sn-1 position compared to the sn-2 position. These results encourage a further pharmacophore optimization for the design and synthesis of novel C1 domain targeted ligands to achieve improved binding affinity and selectivity for PKCs and other C1 domain-containing targets.
Taken together, we demonstrated previously that the isophthalate derivatives show affinity for the C1 domains of cPKCs and nPKCs and possess promising biological activities in cell culture models related to cancer and Alzheimer's disease. In an attempt to improve the aqueous solubility of the C1 domain ligands, we prepared a set of 2,4,6-trisubstituted-and 2,4,5,6-tetrasubstituted pyrimidines, bearing similar hydrophobic substituents as the isophthalates, and quantified their binding to PKCα. We can conclude that the novel pyrimidine analogs did not establish improved binding affinity for PKCα compared to the most promising isophthalates and the lower binding affinity of the isophthalates, compared to more potent ligands, may correlate to the aromatic/planar nature of their core structure. Results presented here, however, provide useful SAR data for further development of ligands targeted to the C1 domain of PKC.

Modeling
We docked our 22 compounds to the crystal structure of the C1B domain of PKCδ (PDB ID: 1PTR) using Glide of Schrödinger Maestro [31] with SP parameters. The targeted binding site was defined by the mass center of the co-crystallized ligand, phorbol 13-acetate, which was also used as a reference compound in docking. Prior to the docking, the target protein was prepared with Maestro's Protein preparation tool, and 3D coordinates of the compounds were calculated by Schrödinger's LigPrep utilizing Epik to generate protonation states. For scoring, we used Glide's "docking score".

Syntheses
All reagents were acquired from Sigma-Aldrich (Schnelldorf, Germany), Fluorochem (Hadfield, United Kingdom) and Fluka (Buchs, Switzerland), and were used without further purification. All reactions in anhydrous conditions were conducted using dry solvents in oven-dried glassware under an inert atmosphere of dry argon. The progress of chemical reactions was monitored by thin-layer chromatography on Silica Gel 60 F254 aluminum sheets acquired from Merck (Darmstadt, Germany), visualized under UV light (254/366 nm) and stained with phosphomolybdic acid (10% w/v in EtOH). Microwave reactions were performed with a Biotage Initiator + SP Wave Microwave Synthesizer (Uppsala, Sweden). Flash SiO 2 column chromatography was performed with an automated high performance flash chromatography Biotage Sp1-system equipped with a 0.1-mm path length flow cell UV-detector/recorder module (fixed wavelength 254 nm) or with a Biotage Isolera™ Spektra Systems with ACI™ and Assist (ISO-1SW Isolera One) equipped with a variable UV-VIS (200-800 nm) photodiode array (Uppsala, Sweden), and the indicated mobile phase gradient. 1 H, 13 C and 19 F NMR spectra (also available in S1 Appendix including 13 C HSQC, 13 C HMBC and 15 N HMBC 2D NMR spectra) were acquired on a Bruker Ascend 400 MHz-Avance III HD NMR spectrometer (Bruker Corporation, Billerica, MA, USA) as solutions in CDCl 3 . Chemical shifts (δ) are reported as parts per million (ppm) relative to the solvent peaks at 7.26 and 77.16 ppm for 1 H and 13 C NMR respectively. Multiplicities of peaks are represented by s (singlet), d (doublet), t (triplet), q (quartet), quint (quintet), and m (multiplet). Visual features of peaks including broad (br) or apparent (app) are also indicated. In 13 C NMR data, peaks referring to two symmetrical carbons (sym, 2C) or two different carbons with overlapping signals (2C) are also indicated. All spectra were processed for recorded FID files with MestReNova 11.0.4 software (Mestrelab Research, Santiago de Compostela, Spain). Low resolution mass (MS-APCI) analyses were performed on a MS Advion expression 1 CMS spectrometer equipped with an APCI ion source and an Atmospheric Solids Analysis Probe (ASAP) and the data was reported for the molecular ions [M+H] + . Exact mass and purity (>95%) of all tested compounds was confirmed by LC-MS analyses with a Waters Acquity 1 UPLC system (Waters, Milford, MA, USA) equipped with an Acquity UPLC 1 BEH C18 column (1.7 μm, 50 × 2.1 mm, Waters, Ireland), an Acquity PDA detector and a Waters Synapt G2 HDMS mass spectrometer (Waters, Milford, MA, USA) via an ESI ion source in positive mode. High resolution mass (HRMS-ESI) data was reported for the molecular ions [M+H]  General procedure I: Acid-catalyzed transesterification. Compound 5 was dissolved in alcohol (13-16 equiv) and heated to 100˚C for 3 h in the presence of a catalytic amount of H 2 SO 4 (0.1 equiv). Complete dissolution occurred while heating. The reaction was quenched by adding a saturated solution of NaHCO 3 in water and the mixture was extracted with EtOAc. The organic layers were combined, and the solvent was evaporated under reduced pressure at 40˚C. The residual alcohol was removed by vacuum distillation. The crude residue was purified by flash column chromatography with appropriate eluents and a gradient.