|Home | About | Journals | Submit | Contact Us | Français|
Optimization of a series of highly potent and kinome selective carbon-linked carboxamide spleen tyrosine kinase (Syk) inhibitors with favorable drug-like properties is described. A pervasive Ames liability in an analogous nitrogen-linked carboxamide series was obviated by replacement with a carbon-linked moiety. Initial efforts lacked on-target potency, likely due to strain induced between the hinge binding amide and solvent front heterocycle. Consideration of ground state and bound state energetics allowed rapid realization of improved solvent front substituents affording subnanomolar Syk potency and high kinome selectivity. These molecules were also devoid of mutagenicity risk as assessed via the Ames test using the TA97a Salmonella strain.
The advancement of kinase inhibitors as oncology therapeutics has resulted in approval of a number of effective drugs over the past two decades. Recently, efforts have been directed toward the treatment of autoimmune disorders, particularly rheumatoid arthritis (RA), via kinase inhibition. RA is a chronic inflammatory disease that presents with pathophysiology including synovial hyperplasia, inflammation, and joint destruction. RA is characterized by up-regulation of cytokines and chemokines ultimately leading to inflammation. Treatment paradigms have consisted of palliative therapies to mitigate symptoms including pain and disease-modifying antirheumatic drugs (DMARDs) or biologics with the goal to slow or stop disease progression. Existing therapies are not universally effective for all RA patients, underscoring the need for additional treatment options.
Spleen tyrosine kinase (Syk) is a well-studied enzyme that plays a critical role in immune signaling via Fc and B cell receptors (BCR).1 Preclinical studies demonstrated that pharmacological modulation of Syk (R788 (1), PRT-062607 (2), RO9021 (3), and GS-9973 (4) is efficacious in preclinical RA models (Figure Figure11).2−5 Further, poorly selective Syk inhibitor 1 (52% of 100 kinases tested >100× Syk IC50) has demonstrated efficacy in human clinical trials; however, dosing was limited by observations of hypertension and diarrhea.6 We aimed to develop a selective Syk inhibitor that would obviate the clinical side effects and could be useful for autoimmune and oncological disorders.7−11 As previously described,12 a picolinamide series of Syk inhibitors we developed was halted by mutagenicity risk as indicated by positive results in the Ames assay. Described herein are the design considerations employed to overcome this liability via a scaffold modification affording novel Syk inhibitors consistent with our objective.
Picolinamide 5 was designed via analysis of X-ray crystallographic data with a focus on ligand binding efficiency and ensuring favorable physicochemical properties (Figure Figure22). Carboxamide 5 possessed desirable intrinsic (Syk IC50 = 60 pM) and human whole blood potency (hWB CysLT IC50 = 58 nM). Additionally, kinome selectivity was high (99% of 265 kinases tested >100× Syk IC50),13 and the oral pharmacokinetic profile (rat plasma Clp (Clu) = 33 (710) mL/min/kg, rat T1/2 = 2.7 h, F = 19%) enabled testing in the rat collagen-induced arthritis model, where the compound demonstrated therapeutic efficacy at doses as low as 3 mg/kg QD (Cmax = 0.2 μM, Cmin = 0.007 μM). However, this molecule, and this series of diaminopicolinamides, was plagued with positive results in the Ames test using TA97a Salmonella strain. Since mutation in this strain often reflects intercalation of the test molecule into DNA, we developed a docking model leveraging DNA cocrystals with acridine. Based upon this model, an approach we pursued to ameliorate this activity involved changing the two-dimensional shape of how the molecules are presented, aimed at disrupting the favorable interactions with DNA.
Our initial foray into changing two-dimensional shape focused on altering the vector at which the “solvent front” group was presented; this strategy involved carbon-linking to replace the previous nitrogen-linked moiety. To this end, picolinamide 6 was quickly prepared via Suzuki coupling14 of available intermediates to replace the aminoheterocycle with a C-linked 3-methylpyrazole analogue (Figure Figure33). However, picolinamide 6 was inactive in our enzymatic assay (Syk IC50 > 10 μM, 9% inhibition at 10 μM). We hypothesized that the 3-methylpyrazole may introduce torsional strain with the carbonyl oxygen of the carboxamide, disfavoring the planar geometry desired for efficient interaction with the hinge domain of Syk, and kinases in general. To probe this theory, we pursued computational quantum mechanical approaches to estimate the lowest energy dihedral angles about these two substituents and the energetic penalty required to achieve a fully flat conformation (amide dihedral = 0° and aryl solvent front dihedral = 0°) for binding.15 For picolinamide 6, calculations suggest that the preferred ground state conformation positions the two critical substituents ~30° and −30° out of plane, resulting in an estimated penalty of 2–4 kcal/mol to adopt the planar bound-state conformation, which is reflected in the lack of potency for this analogue.
Considering rational structural modifications that could be implemented to further favor the flat ground state conformation, minimizing the motion and associated energetic penalty required to achieve the flat bioactive state, calculations were executed to examine the roles of different heteroatom substituted cores and solvent fronts. Among those surveyed, the pyrazinecarboxamide core was predicted to favor a smaller dihedral angle, and corresponding lower energetic penalty (0–2 kcal/mol) than the picolinamide. In the picolinamide 9, there is likely increased torsional strain resulting from the two CHs in close proximity compared to the pyrazine 10, where this strain is removed and the conformation may be reinforced by an additional intramolecular hydrogen bond (Figure Figure44). Pyrazinecarboxamide 7 was prepared to test this hypothesis and showed improved potency (Syk IC50 = 1.6 μM). It is worth noting that the cyclohexyldiamine generally introduces a ~10× potency enhancement versus the propanyldiamine, though this matched pair was never made with the methypyrazole solvent front.
Hoping to parlay the potency gain achieved moving from the picolinamide to the pyrazinecarboxamide with a solvent front group predicted to further favor a planar disposition, indole 8 was synthesized.16 Calculations suggested that the indole substituted pyrazinecarboxamide should strongly favor a flat ground state conformation, resulting in a 0–0.5 kcal/mol energetic penalty to adopt the bound conformation. In addition, the energy minimum at the presumed bound conformation is predicted to become much deeper than alternative states, which likely reduces entropic penalties as well. This is likely influenced by an intramolecular hydrogen bond between the more polarized NH of the indole and the carbonyl of the carboxamide. Gratifyingly, indole 8’s doubly stabilized planarity afforded a significantly improved intrinsic potency (Syk IC50 = 5 nM). Interestingly, the predicted ground-state geometry for 8 suggests a pseudo 7-membered ring where the bond angle increases to accommodate an intramolecular hydrogen bond between the indole NH and the carboxamide carbonyl (Figure S1).15
In order to further understand the binding interactions of these compounds, as well as interrogate the computational predictions, we solved the X-ray crystal structure of compound 8 bound to the kinase domain of human Syk (Figure Figure55). Similar to other carboxamide-based Syk inhibitors,17−208 binds to the hinge region via bidentate hydrogen bonding to the backbone of Glu449 and Ala451. The indole region extends toward solvent and packs against the side chain of Met450 at the top of the hinge, while NH groups of the diamine region form interactions with the catalytic Asp512 and surrounding backbone carbonyls from Arg498 and Asn499. As predicted, the core is planar with a pseudo 7-membered ring geometry stabilized by an intramolecular hydrogen bond between the indole NH and the carboxamide oxygen with a short N–O distance of 2.6 Å (Figure Figure55, magenta dashed line).
Seeking to further optimize the C-linked carboxamides, introduction of the cyclohexyldiamine moiety to afford pyrazinecarboxamide 11 delivered the expected potency enhancement (Syk IC50 = 500 pM) approaching the previously achieved intrinsic activity of the N-linked picolinamide 5. Further, indole 11 featured excellent kinase selectivity (99% of 101 kinases tested >100× Syk IC50, LRRK2 IC50 = 30 nM) and human whole blood potency (hWB CysLT IC50 = 130 nM) with a favorable pharmacokinetic profile in rat Clp (Clu) = 44 (255) mL/min/kg, rat T1/2 = 2.8 h, F = 19%) (Figure Figure66). As previously described, all molecules tested from the C-linked carboxamides were negative in the mutagenicity assay in strain TA97a Salmonella with rat S9 up to testable concentrations.12 Additional data to support the effectiveness of the shape change in ablating interaction with DNA were gathered via consistent results in a UV–vis perturbation assay and a DNA unwinding assay. We subsequently examined a nitrogen transposition of pyrazinecarboxamide 11 to afford benzimidazole 12. Interestingly, picolinamide 12 showed a loss in intrinsic potency (Syk IC50 = 7 nM) but a maintained selectivity (97% of 101 kinases tested >100× Syk IC50, LRRK2 IC50 = 130 nM, CHK2 IC50 = 160 nM, TSSK3 IC50 = 420 nM), human whole blood potency (hWB CysLT IC50 = 93 nM), and rat pharmacokinetic (Clp (Clu) = 134 (277) mL/min/kg, rat T1/2 = 4.8 h) profile.
Encouraged by the successes of exploiting intramolecular hydrogen bonds, we investigated the use of weaker H-bond donors and other configurations predicted to maintain a planar geometry. Benzofuran 13 lost ~100× potency (Syk IC50 = 530 nM) versus its indole analogue 8. Styrene 14 had comparable intrinsic affinity, ~3× loss (Syk IC50 = 24 nM) versus the corresponding benzimidazole 12. Aryl ether 15, interestingly predicted to prefer a planar configuration, affords an equipotent inhibitor (Syk IC50 = 9 nM) to benzimidazole 12. Alkyne 16 was readily prepared but demonstrated a substantial potency loss (Syk IC50 = 990 nM), likely due to negative interaction of the phenyl moiety with the hinge region of Syk (Table 1).
The combined efforts of this scaffold redesign ultimately delivered pyrazinecarboxamide 17,21 which possessed the overall profile we were seeking (Syk IC50 = 700 pM) including high selectivity (99% of kinases >100×, 101 kinases tested, LRRK2 IC50 = 33 nM), potent cell functional activity (hWB CysLT IC50 = 62 nM), and a favorable pharmacokinetic profile in preclinical species (rat plasma Clp (Clu) = 25 (100) mL/min/kg, rat T1/2 = 3.5 h, 29%F). Further, indole 17 posed low risk for ion channel activity and DNA interaction as assessed by preclinical safety assays (hERG IC50 = 22 μM and UV–vis = negative). In summary, careful consideration of designs to bias toward the bioactive conformation delivered alternative chemotypes that maintained favorable drug-like properties while overcoming several safety issues.
We would like to acknowledge Jaren Arbanas and Uwe Mueller for routine in vitro pharmacology support, and Bruce Adams and Bridget Becker for NMR structural support.
The complex between the human Syk kinase domain and compound 8 has been deposited in the Protein Data Bank (PDB) with accession code 5TIU.
All authors have given approval to the final version of the manuscript.
The authors declare the following competing financial interest(s): All authors are employed by Merck & Co., Inc., who supported and funded these research efforts.