Identification of highly selective type II kinase inhibitors with chiral peptidomimetic tails

Abstract Identification of highly selective type II kinase inhibitors is described. Two different chiral peptidomimetic scaffolds were introduced on the tail region of non-selective type II kinase inhibitor GNF-7 to enhance the selectivity. Kinome-wide selectivity profiling analysis showed that type II kinase inhibitor 7a potently inhibited Lck kinase with great selectivity (IC50 of 23.0 nM). It was found that 7a and its derivatives possessed high selectivity for Lck over even structurally conserved all Src family kinases. We also observed that 7a inhibited Lck activation in Jurkat T cells. Moreover, 7a was found to alleviate clinical symptoms in DSS-induced colitis mice. This study provides a novel insight into the design of selective type II kinase inhibitors by adopting chiral peptidomimetic moieties on the tail region.


Introduction
Most signal transduction processes are mediated through a phosphotransfer reactions catalysed by kinases. However, overexpression or mutation of kinases causes tumour cell proliferation and survival. Therefore, kinases are pursued as invaluable targets and a tremendous amount of effort has been devoted towards the discovery of small molecular kinase inhibitors for the treatment of cancer for decades 1 . Small molecule kinase inhibitors have been classified by binding modes with protein kinases. Type I inhibitors are the most commonly encountered and occupy the ATP-binding site of the active conformation of kinases (i.e. DFG-in conformation). In contrast to type I inhibitors, type II kinase inhibitors recognise the ATP-binding pocket of inactive DFG-out conformation of kinase proteins. Besides ATP competitive inhibitors (e.g. type I, type II), allosteric and covalent inhibitors have also been investigated 1-4 . Although a significant number of small molecular kinase inhibitors have been developed, discovery of selective kinase inhibitors remains challenging. The rationale behind this is that the structure of ATP-binding site in all of the kinase proteins is highly conserved 1 . In addition, designing an allosteric kinase inhibitor, which is the most selective inhibitor, is difficult since it highly relies on an empirical exercise 2a . Although selective kinases have been uncovered using subtle 3-dimensional structural differences among the kinases 5 , discovery of highly selective kinase inhibitors still remains a largely unmet challenge. Type II kinase inhibitors were anticipated to be more selective compared to type I kinase inhibitors in the early phase since the hydrophobic pocket generated by the DFG-out conformation is not quite conserved in contrast to ATP binding pocket and DFG-out conformations are more dynamic. However, type II kinase inhibitors, developed to date have been proven to be largely less selective than type I kinase inhibitors. 2a We noticed that a number of the less selective type II kinase inhibitors possess achiral and limited chemotype tails, which interact with allosteric site of the inactive conformation of kinases. We envisioned that interactions of flat and achiral tail fragments with 3-dimensional structural kinases would be highly limited and resulted in less selective profile 6 . Thus, we were curious whether type II kinase inhibitors containing chiral tails would be more selective among all of the kinases by affording various binding modes 7,8 . We were particularly interested in peptidomimetic structure as tail scaffolds. To our best knowledge, development of selective type II kinase inhibitors that consists of peptidomimetic tails is unprecedented. To investigate how peptidomimetic tail structure affects kinase selectivities, we chose the core structure of GNF-7 (Figure 1) 9 . GNF-7 has been discovered as a type II T315I Bcr-Abl kinase inhibitor and possesses remarkable potencies against many kinases with highly low kinase selectivity. Among the peptidomimetic structure, we were attracted to synthetically easily accessible solution phase turn mimetic libraries 10 .

In vitro kinase inhibitory activities
Kinase-inhibitory activities of both 7 and 11 against four selected kinases were assessed by in vitro kinase assay (Table 1). To our delight, both 7a and 11b showed high degree of selectivities among selected kinases even between structurally similar Lck and c-Src (over 10-fold selectivities). Replacement of L-Pro with D-Pro led to lower selectivities between Lck and c-Src (7a and 7b). This is not surprising since stereochemical alteration of Pro at i þ 1 might potentially change 3-dimensional conformation of the structure 10b . The substituents at N-1, C-3, and C-5 positions on the 1,4benzodiazepin-2-one ring significantly affected selectivities. In all cases, the activities on Lck kinases were found to be superior to those on other kinases. Gratifyingly, 11b possessed over 10-fold selectivity for Lck over c-Src. The iso-butyl group at C-3 position in the 1,4-benzodiazepin-2-one surpassed the methyl group in respect of the selectivity (11b and 11d). Also, the 2-butyl group at the C-3 position causes almost no selectivity (11e).
With the exciting initial data in hand, selectivities between Lck and c-Src were investigated on derivatives 7 containing the b-turn mimetic scaffolds (Table 2). Kinase-inhibitory activities of derivatives 7 against Lck and c-Src were assessed by in vitro kinase assay. Interestingly, 7c possessing alternative stereochemistry compared to 7a showed significantly (>10-fold) diminished selectivity. Changing i þ 2 functional groups to cyclopropane (7d and 7e), cyclobutane (7f), and even glycine (7g and 7h) exhibited high selectivities. However, we observed diminished selectivity with benzyl substitution at i þ 2 position (7i). High selectivity was kept with piperidine scaffold at i þ 1 (7j). In contrast to the pyrrolidine and the piperidine groups, addition of the azetidine at i þ 1 (7k) resulted in over 10-fold lower selectivity. The selectivities were not decreased by replacement of iso-butyl with valine at i þ 3 (7l, 7m, 7n, and 7o). Substituents at i significantly affected selectivities (7p, 7q, 7r, 7s, and 7t). High selectivities were observed with iso-butyl and cyclohexyl substitution at i (over 10fold selectivities, 7q and 7s). Compared to 7a, The selectivities of 7v and 7w on Lck and c-Src were highly diminished 10f .
Additionally, we investigated the selectivities of derivatives containing benzodiazepine scaffolds on Lck and c-Src kinases (Table 3). Kinase-inhibitory activities of derivatives 11 against Lck and c-Src were assessed by in vitro kinase assay. The selectivities of 11f was similar to those of its enantiomer 11a and p-fluorobenzyl 11g. It is noteworthy that the benzyl group at N-1 position of the 1,4-benzodiazepin-2-one moiety (11b) was superior to the methyl, allyl, iso-butyl, methoxyethyl, and N-benzylacetamide groups at the corresponding position in terms of selectivities (11h, 11i, 11j, 11n, and 11o). The methyl group in 11k and 11l at C-5 position is slightly more favourable than the phenyl group (11a and 11f) at the corresponding position as regards the selectivity. We observed that introduction of c-turn mimicry 10e at the tail position resulted in almost no selectivity for Lck over c-Src (11p and 11q).
We explored in vitro potencies of 7a and 11b at both 14 mM (Km) and 1 mM ATP concentrations to investigate whether these inhibitors are ATP competitive inhibitors (Table 4). We observed that the IC 50 values were increased dramatically at 1 mM ATP concentration compared to those at Km ATP concentrations. Thus, we concluded that both 7a and 11b are ATP competitive kinase inhibitors.
2.3. Kinome-wide selectivities of 7a described on a kinome phylogenetic tree We were pleased to confirm that 7a possesses higher selectivity for Lck over other 373 kinases at 1 mM concentration compared to GNF 7 (Figure 2, Supplementary Table S1-S4). We also obtained IC 50 values of 7a against five kinases (Lck, DDR1, Fgr, Bmx, and Blk), which were inhibited greater than 70% in the kinome-wide profiling analysis (Figure 2c, Supplementary Table S1). As shown in Figure 2(c), 7a has an IC 50 value of 23 nM against Lck and possesses 5 to 13-fold selectivity over these four kinases. Furthermore, 7a showed more than 10-fold selectivities over other structurally similar Src family kinases (Figure 2d).

Docking into Lck binding site
Docking of 7a and 11b with Lck kinase by long time (3 ls) molecular dynamics (MD) simulation revealed that turn peptidomimetic scaffolds were located at an allosteric binding site. Also, hydrophobic groups of the peptidomimetic scaffolds interacted with the allosteric helix of the binding site (C-helix in Figure 3). The results reveal that the amide moiety of 7a participates in hydrogen-bonding networks with Glu288 in Lck. Additionally, the phenyl group of 11b interacts with Phe354 by p-p stacking interaction. The core scaffolds of 7a and 11b were interacted with kinases hinge region through similar binding poses of the crystal structure of imatinib with kinases (Supplementary Figure S1). The Asp from the "DFG-motif" interacts with the amide functional group of 7a and 11b. Moreover, phenyl moieties on the linker region of 7a and 11b interacts with Lys273 by cation-p interaction.

Watermap application
Next, we further analysed binding sites using WaterMap application 11,12 . With WaterMap results, many hydration sites with unstable energy (0 > kcal/mol) were found in the allosteric binding site (Supplementary Figure S2). Therefore, when 7a was located at the unstable hydration sites, high binding free energy compensation could be obtained. Five hydration sites with a high energy of over 3.00 kcal/mol were found in Lck, and the peptidomimetic scaffold of 7a occupied these sites ( Figure 4). We envisioned that the energetically unstable regions at the allosteric site of Lck could be compensated with interaction of terminal turn peptidomimetic tail of 7a, which would contribute to high selectivity of 7a for Lck kinase.
With the interesting WaterMap calculation result, we attempted to explain the IC 50 difference of 7a and its enantiomer 7c on Lck by the WaterMap application method ( Figure 5). The U-shaped tail of 7a is located in the allosteric binding site on Lck. Also, amino acids at i and i þ 3 of 7a tail interact with allosteric helix. However, turn structure of 7c is not fully located at the allosteric binding site of Lck, which is expected to be highly important for  kinase selectivities. Although the amino acid at i of 7c tail interacts with allosteric helix, the amino acid at i þ 3 is extruded outside and exposed to solvent. Thus, the tail structure of 7c could have high fluctuation, which results in lower selectivities ( Figure 5).

Inhibitory effect of 7a on Lck activation
We investigated whether Lck activation was affected by 7a in Jurkat cell line ( Figure 6). Western blot analysis revealed that 7a markedly reduced the phosphorylated Lck levels at tyrosine 394 residue in a concentration-dependent manner in anti-CD3-treated Jurkat T cells. These results suggested that 7a inhibited the anti-CD3-activated Lck, similar to positive control A770041.

In vivo experiment with dextran sulphate sodium (DSS)induced colitis model
Inflammatory bowel disease (IBD), which is a chronic and immune-mediated disorder of the gastrointestinal tract encompasses Crohn's disease (CD) and ulcerative colitis (UC) 13 . Although the exact cause of IBD is unclear, it is widely accepted that an excessive immune response against normal components of microflora results in IBD. Especially, excessive T cell activation plays a pivotal role in mucosal damage in both CD and UC 14 . Lck plays a crucial role in activation of TCR-linked signal transduction pathways, leading to T cell activation and proliferation 15 . Additionally, it is reported that overexpression of Lck leads to IBD 15 . Hence, we evaluated the potential of our selective Lck kinase inhibitor 7a for IBD treatment with dextran sulphate sodium (DSS)-induced colitis model. DSS administration induces acute colonic damage, and changes in clinical parameters can be monitored 16 . To determine the recovery effect of 7a in DSS-induced colitis, we assessed the clinical symptom including disease activity index (DAI) and colon length. The DAI scores were evaluated by body weight loss, stool consistency, and occult/gross bleeding (Table 5). During the administration of DSS (4%) for 7 days, DAI values were significantly increased ( Figure 7a). We discovered that 7a treatment (5 mg/kg, i.p.) improved the symptom changes at the end of experiments ( Figure 7b). In addition, the colon length of DSS-treated group was significantly shorter than that of the vehicle-administered control group (8.17 ± 0.32 cm vs. 4.25 ± 0.38 cm, p < 0.001), while 7a treatment (5 mg/kg, i.p.) recovered the DSS-induced colon shortening (4.25 ± 0.38 cm vs. 5.47 ± 0.60 cm, p < 0.05, Figure  7(C), D).

Conclusion
In conclusion, we discovered highly selective type II kinase inhibitors by introducing chiral turn peptidomimetic moieties on the tail region for the first time. It turned out that 7a, a novel type II kinase inhibitor, is a potent and exceptionally selective Lck inhibitor. Based on kinome-wide selectivity profiling data, it was confirmed that 7a possesses high selectivity. Kinases selectivities were highly affected by subtle changes of the substituents of peptidomimetic scaffolds. To the best of our knowledge, it has never been reported that low selectivity of a type II kinase inhibitor is dramatically enhanced by adopting chiral peptidomimetic tail. The western blot analysis revealed that 7a is capable of inhibiting Lck activation in anti-CD3-treated Jurkat T cells. Finally, we discovered that 7a could alleviate of clinical symptoms in DSS-induced colitis mice. This study may shed a bright light on the design of selective type II kinase inhibitors by adopting chiral peptidomimetic tails.

General procedures
Unless otherwise stated, reactions were performed in flame-dried glassware under a nitrogen atmosphere using dry solvents. Reaction progress was monitored by thin-layer chromatography (TLC). Purified water was obtained using a Barnstead NANOpure Infinity UV/UF system. Brine solutions are saturated aqueous solutions of sodium chloride. Commercially available reagents were purchased from Sigma-Aldrich, Acros Organics, Combi-Blocks, TCI or Alfa Aesar and used as received unless otherwise stated. Reaction temperatures were controlled by an IKAmag temperature modulator unless otherwise indicated. TLC was performed using E. Merck silica gel 60 F254 precoated glass plates (0.25 mm) and visualised by UV fluorescence quenching, KMnO 4 , or Ninhydrin staining. Silicycle SiliaFlash P60 Academic Silica gel (particle size 0.040-0.064 mm) was used for flash column chromatography. 1H NMR spectra were recorded on Bruker 400 MHz and 600 MHz spectrometer and are reported relative to residual CDCl 3 (d 7.26 ppm),  operating on the sodium D-line (589 nm), using a 100 nm pathlength cell and are reported as: [a] D T (concentration in g/100 ml, solvent). High resolution mass spectra (HRMS) were obtained from Waters SYNAPT G2 TOF with a Waters Multimode source in electrospray ionisation (ESIþ), atmospheric pressure chemical ionisation (APCIþ), or mixed ionisation mode (MM: ESI-APCIþ).

General Procedure B for synthesis of amides 11
To a solution of aniline 8 (1.20 equiv) and carboxylic acid 9 or 10 (1.00 equv) in DMF (0.100 M) was added HATU (2.00 equiv) and i-Pr 2 NEt (5.00 equiv) at 23 C. The reaction mixture was stirred for 16 h at 23 C. The reaction was diluted with CH 2 Cl 2 and quenched by addition of sat. NaHCO 3 . The phases were separated and the aqueous phase was extracted with EtOAc. The combined organic layer was washed with brine, dried over MgSO 4 , and concentrated in vacuo. The residue was purified by column chromatography (1:20 MeOH:CH 2 Cl 2 ) on silica gel to afford amide 11 (10-20% yield).

Induction of colitis by dextran sulphate sodium (DSS) and treatment
Colitis in mice was induced by providing water containing 4% (w/ v) DSS for 7 days. Mice were randomly divided into 5 groups (n ¼ 6/group, Figure 8) as follows: control group treated with vehicle; DSS plus vehicle group exposed to 4% DSS and treated with vehicle; the other 3 groups consist of mice receiving 4% DSS was treated with 5-ASA (75 mg/kg/day, p.o.) as a positive control or 7a (1, 5 mg/kg/day, i.p.) daily for 7 days.

Assessment of the disease activity index (DAI)
To calculate the severity of colitis, body weight, stool consistency, and occult/gross bleeding of all mice were assessed. DAI score was measured every day according to the following table (Table  5). The colon length was measured at end of the experiment.

Statistical analysis
Results are expressed as the mean ± SE of triplicate experiments with similar patterns. Statistically significant values were compared using ANOVA and Dunnett's post hoc test, and p values of less than 0.05 were considered statistically significant.