Article pubs.acs.org/jmc
Discovery of Selective and Orally Bioavailable Protein Kinase Cθ (PKCθ) Inhibitors from a Fragment Hit Dawn M. George,*,† Eric C. Breinlinger,† Michael Friedman,† Yang Zhang,∥ Jianfei Wang,∥ Maria Argiriadi,† Pratima Bansal-Pakala,† Martine Barth,⊥ David B. Duignan,† Prisca Honore,‡ QingYu Lang,§ Scott Mittelstadt,‡ Dominique Potin,⊥ Lian Rundell,† and Jeremy J. Edmunds† †
AbbVie Bioresearch Center, 381 Plantation Street, Worcester, Massachusetts 01605, United States AbbVie Inc., 1 North Waukegan Road, North Chicago, Illinois 60064, United States § AbbVie China R&D Center, 5F, North Jin Chuang Building No. 1, 4560 Jinke Road, Pudong New District, Shanghai 201201, P. R. China ∥ WuXi AppTec (Shanghai) Co., Ltd., 288 Fute Zhong Road, Waigaoqiao Free Trade Zone, Shanghai 200131, P. R. China ⊥ Inventiva, 50 Rue de Dijon, 21121 Daix, France ‡
S Supporting Information *
ABSTRACT: Protein kinase Cθ (PKCθ) regulates a key step in the activation of T cells. On the basis of its mechanism of action, inhibition of this kinase is hypothesized to serve as an effective therapy for autoimmune diseases such as rheumatoid arthritis (RA), inflammatory bowel disease (IBD), and psoriasis. Herein, the discovery of a small molecule PKCθ inhibitor is described, starting from a fragment hit 1 and advancing to compound 41 through the use of structure-based drug design. Compound 41 demonstrates excellent in vitro activity, good oral pharmacokinetics, and efficacy in both an acute in vivo mechanistic model and a chronic in vivo disease model but suffers from tolerability issues upon chronic dosing.
■
INTRODUCTION Multiple PKC isoforms exist, each with different functions including roles in cellular signaling, proliferation, differentiation, migration, survival, and death.1 Of the 12 known isoforms, PKCθ and PKCα play fundamental roles in T cell biology. While PKCα is required for T cell proliferation2 and its inhibition could contribute to efficacy in inflammatory disease models, inhibition of this isoform is also associated with undesirable increases in cardiac contractility.3 PKCθ is highly expressed in T cells and plays an essential role in interleukin-2 (IL-2) production, cellular expansion, and activation. PKCθ mediates activation of T cells induced by the T cell receptor and the CD28 co-stimulation pathways. On the basis of this mechanism of action, inhibition of PKCθ is expected to block T cell activation and lead to ablation of T cell responses. PKCθdeficient mice display defects in responses to antigen resulting from defective activation of NF-κB, NFAT, and AP-1 following T cell receptor engagement.4,5 PKCθ-deficient mice are resistant to collagen-induced arthritis (CIA),6 experimental autoimmune encephalomyelitis (EAE),7 and models of asthma8,9 suggesting that PKCθ inhibition alone may be sufficient to drive efficacy in chronic autoimmune and inflammatory diseases. Inhibition of PKCθ was recently demonstrated to enhance suppressive function in regulatory T cells (Tregs),10 suggesting that PKCθ may mediate a negative feedback signal on © XXXX American Chemical Society
regulatory T cell function. Furthermore, blockade of PKCθ by a small molecule inhibitor rescued the function of Tregs isolated from RA patients as well as protected these cells from the inhibitory effects of TNFα, highlighting a potential role for PKCθ inhibition in restoring immune regulation in chronic autoimmune diseases such as RA. These complementary effects (inhibition of T cell activation and enhancement of Treg function) upon PKCθ inhibition are hypothesized to confer superior efficacy in T cell-mediated diseases than either mechanism alone. Given the pivotal and nonredundant role that PKCθ plays in mediating T cell activation and the finding that PKCθ inhibition also enhances Treg function, we specifically targeted this isoform to develop an effective oral therapy for multiple autoimmune and inflammatory disease indications such as RA, IBD, and psoriasis. Although PKCθ has been the focus of intense research over the past 2 decades,11−22 no PKC inhibitor has successfully navigated through clinical trials to market. Novartis’ 2 (sotrastaurin, AEB071, Figure 1)11 inhibits multiple PKC isoforms (IC50: PKCθ 1.0 nM, PKCα 2.1 nM, PKCβ1 2.0 nM, PKCδ 1.3 nM, PKCε 6.2 nM, PKCη 6.1 nM). Most recently 2 Special Issue: New Frontiers in Kinases Received: April 29, 2014
A
dx.doi.org/10.1021/jm500669m | J. Med. Chem. XXXX, XXX, XXX−XXX
Journal of Medicinal Chemistry
Article
has been shown to increase contractility; 27 thus, we hypothesized that improved selectivity over this isoform would be an effective strategy to minimize potential cardiovascular safety concerns. Starting from 19 μM (TRFRET data) fragment hit 1, structure-based drug design investigating substitution at the 4-, 6-, and 7-positions led to the identification of potent compounds that demonstrated excellent PKCθ/α selectivity, good PK/ADME properties, and efficacy in acute and chronic in vivo models.
■
CHEMISTRY Compound 9 was synthesized as described in Scheme 1. Cyclization of aminobromophenol 4 with 2-chloroacetyl chloride provided benzo[1,4]oxazinone 5. Alkylation with ethyl bromoacetate yielded intermediate 6. Treatment with Lawesson reagent to 7 with subsequent cyclization provided bromotriazinone 8. Further transformation through Suzuki chemistry provided 7-substituted triazinone 9. Similar chemistry was applied to effect the syntheses of 4-, 6-, and 7substituted compounds 15a−w as outlined Scheme 2. Benzo[1,4]oxazinones 10 appropriately substituted on the 6- and 7positions were prepared from commercially available starting materials (see Supporting Information). As described previously, alkylation yielded compound 11 and treatment with Lawesson reagent afforded intermediate 12. From this intermediate, two divergent routes were developed. In the first route, reduction of the nitro group generated aniline 13. Reductive amination produced substituted amine 14, and final cyclization with hydrazine provided target triazinones 15. In the alternative approach from 12, the synthetic sequence was reversed to accommodate larger substituents R2 at the C4 position. Initial cyclization of 12 with hydrazine to give triazinone 16 followed by nitro reduction generated aniline 17. Functionalization via reductive amination or acylation delivered compounds 15. The synthesis of 7-iPr substituted analogue (21) was accomplished by later-stage incorporation of the C7 substituent as shown in Scheme 3. Thus, intermediate 15x was brominated to provide 7-bromo analogue 18. Suzuki coupling yielded isopropenyl intermediate 19 followed by hydrogenation to give alkane 20. Subsequent Boc deprotection furnished the desired C7-iPr compound 21. The C7-trifluoromethylated analogues were prepared as described in Scheme 4. Starting
Figure 1. Structures of PKC inhibitors from Novartis and Vertex.
completed phase II clinical trials for psoriasis and transplantation (kidney and liver); however, the compound development was ultimately halted in 2012, as it “did not show sufficient therapeutic benefit over standard of care”.23 A transient increase in heart rate was observed in clinical trials with 2, leading to speculation about the feasibility of developing a safer approach through selective inhibition of one PKC isoform instead of multiple PKC isoforms.24,25 In early 2013, Vertex reported a PKCθ selective pyrazolopyridine 3 (“compound 27”, Figure 1)13 that demonstrated good selectivity against other PKC family members (Ki: PKCθ 0.08 nM, PKCα 356 nM, PKCβI 15 nM, PKCβII 393 nM, PKCδ 16 nM, PKCε 1 nM, PKCγ 725 nM, PKCη 3.5 nM, PKCμ 2420 nM, PKCζ >5000 nM) and activity in an acute in vivo model of staphylococcal enterotoxin B-stimulated IL-2 production. The 200× selectivity observed for PKCθ versus PKCδ was hypothesized to potentially improve the overall safety profile by mitigating the risk of B cell autoimmunity associated with inhibition of the PKCδ isoform. However, no preclinical cardiovascular or toxicology safety data were disclosed and no advancement into the clinic has been reported. Rigel has also announced the nomination of a dual PKCα/PKCθ clinical candidate, R524,26 presenting data arguing that dual α/θ inhibition is required for full efficacy in a model of transplantation. No clinical trials with this compound have been reported, and the structure has not been disclosed. Our focus was to identify highly potent compounds with selectivity for PKCθ over the PKCα isoform. PKCα is a negative regulator of heart contractility, and inhibition of PKCα Scheme 1a
(i) 2-Chloroacetyl chloride, NaHCO3, DME, water, 80 °C; (ii) ethyl 2-bromoacetate, K2CO3, acetone, 56 °C; (iii) Lawesson reagent, toluene, 120 °C; (iv) NH2NH2−H2O, EtOH, rt; (v) potassium dimethylaminomethyltrifluoroborate, Pd(OAc)2, 2-(dicyclohexylphosphino)-2′,4′,6′triisopropylbiphenyl, Cs2CO3, dioxane, water, 100 °C.
a
B
dx.doi.org/10.1021/jm500669m | J. Med. Chem. XXXX, XXX, XXX−XXX
Journal of Medicinal Chemistry
Article
Scheme 2a
(i) 2-Bromoalkyl acid ethyl ester derivative, K2CO3, solvent, heat; (ii) Lawesson reagent, toluene, 120 °C; (iii) NH2NH2−H2O, EtOH, 80 °C; (iv) zinc powder, ammonium chloride, MeOH, rt; (v) ketone or aldehyde, AcOH, MeOH, heat, then NaBH3CN, heat; (vi) for examples with BOC group; HCl, EtOAc, rt; (vii) carboxylic acid derivative, iPr2EtN, HATU, DMF, rt. a
Scheme 3a
(i) n-Bu4NBr3, DCM, MeOH, rt; (ii) 4,4,5,5-tetramethyl-2-(prop-1-en-2-yl)-1,3,2-dioxaborolane, K2CO3, Pd(dppf)2Cl2, water, dioxane, 80 °C; (iii) Pd/C, MeOH, H2 (55 psi); (iv) TFA, DCM, rt.
a
provided methylene azetidine 32 which underwent hydrogenation to 33. Deprotection using TBAF followed by TFA afforded triazinone 34. A synthetic approach employing Suzuki chemistry similar to that used to prepare C7-iPr compound 21 but omitting the reduction step was utilized for the synthesis of C7-Ph analogues 37 and 38. The stereoisomers were separated via chiral SFC purification as detailed in the Supporting Information. Similarly, racemic compounds 15v and 28 were separated into the (R) and (S) isomers (35 and 36; 39 and 40, respectively) by chiral SFC (see Supporting Information for details). 3-Methylazetidine analogue 41 was prepared using a route similar to that for
from substituted aminophenol 22 (prepared as described in Materials and Methods), cyclization with 2-chloroacetyl chloride and alkylation to ester 23 followed by cyclization with hydrazine afforded triazinone 24. Introduction of a SEM protecting group (25), Buchwald−Hartwig reaction to generate amine 26, and two-step deprotection gave triazinone 28. Scheme 5 describes the synthetic approaches to access -O- and -CH2- C6-linked analogues. Starting from common intermediate 25, conversion to the boronate ester followed by oxidation with hydrogen peroxide generated phenol 29. Formation of ether 30 and dual SEM and Boc deprotection afforded triazinone 31. Alternatively, Heck reaction of intermediate 25 C
dx.doi.org/10.1021/jm500669m | J. Med. Chem. XXXX, XXX, XXX−XXX
Journal of Medicinal Chemistry
Article
Scheme 4a
(i) 2-Chloroacetyl chloride, K2CO3, DMF, then ethyl 2-bromopropanoate, Cs2CO3, rt; (ii) Lawesson reagent, toluene, reflux then NH2NH2−H2O, 80 °C; (iii) NaH, SEMCl, THF, rt; (iv) tert-butyl 3-aminoazetidine-1-carboxylate, Pd(OAc)2, Cs2CO3, BINAP, toluene, reflux; (v) TBAF, THF, 80 °C; (vi) TFA, DCM, rt. a
Scheme 5a
(i) Bis(pinacolato)diboron, Pd(dppf)Cl2, KOAc, dioxane, 90 °C, then aqueous H2O2, dioxane, AcOH, rt; (ii) tert-butyl 3-((methylsulfonyl)oxy)azetidine-1-carboxylate, Cs2CO3, DMF, 80 °C; (iii) TFA, DCM, rt, then aqueous NH4OH, rt; (iv) 3-methyleneazetidine-1-carboxylic acid tertbutyl ester, Et3N, CH3CN, tri(o-tolyl)phosphine, Pd(OAc)2, 110 °C; (v) Pd(OH)2/C, MeOH, H2 (50 psi); (vi) TBAF, THF, 80 °C, then TFA, DCM, rt. a
a recurring challenge. A high-throughput screen (HTS) of our corporate compound collection provided a dearth of compelling starting points; hence, a complementary approach, via a TR-FRET screen of a ∼250-member library of fragment-sized kinase hinge binders, was conducted. From this effort, tricyclic triazinone hit 1 (PKCθ TR-FRET IC50 = 19 μM, PKCθ enzyme IC50 = 453 μM) was identified. Although the potency was moderate, the favorable binding efficiency (BEI = 25 based on TR-FRET data) and unique nature of the triazinone scaffold obligated additional investigation. The synthesis of compound 1 was previously described. Two analogues were noted to have moderate efficacy in a carrageenan-induced rat paw edema model,28 but no subsequent findings on this chemotype have been reported. As is frequently the case with fragment-based discovery, we relied on crystallographic data for initial compound design. As the result of two potential donor−acceptor pairings from the triazinone ring (CO/NH or NH/N), modeling of compound 1 against PKCθ suggested multiple potential
compound 39. The absolute configuration of 35 was confirmed by small molecule X-ray (vide infra), and its isomer 36 was assigned by comparison. The absolute configurations of 37, 38, 39, 40, and 41 were initially assigned based on their relative activity, with the R isomer assumed as the more active isomer as predicted by a superior fit in the PKCθ active site. Specific rotation confirmed that the more active isomers had a negative rotation. The absolute stereochemistry of 39 and 41 was later confirmed by stereoselective syntheses prepared using the route described in Scheme 4 starting from chiral 23 (synthesis described in the Supporting Information).
■
RESULTS AND DISCUSSION
PKCθ as a target has tantalized medicinal chemists with the potential to block T cell signaling based on the key role of PKCθ in regulating T cell activation. However, identification of chemical matter with desirable activity, selectivity, and physicochemical properties to enable in vivo dosing has been D
dx.doi.org/10.1021/jm500669m | J. Med. Chem. XXXX, XXX, XXX−XXX
Journal of Medicinal Chemistry
Article
binding modes. Only two solved X-ray structures in PKCθ have been reported, one with staurosporine and one with a maleimide compound: 1XJD at 2.00 Å29 and 2JED at 2.32 Å.30 Internal efforts have shown that PKCθ protein expression, purification, and crystallization are less robust compared to other kinase targets. As a result, PKCα has been utilized as a reasonable crystallographic substitute in combination with PKCθ homology models.11 In our experience, generation of PKCθ and PKCα crystal structures requires significant (minimally, IC50 < 1 μM) PKCθ or PKCα activity. Thus, because of the weak binding affinity of 1 for PKCθ (IC50 = 19 μM) and PKCα (IC50 > 50 μM), efforts to generate a surrogate crystal structure were pursued. Because of the ease of crystallization, focal adhesion kinase (FAK) was selected as a robust crystallographic surrogate. In spite of modest potency for FAK (TR-FRET IC50 = 156 μM), we were able to obtain a binding mode of 1 at high resolution in the kinase ATP binding pocket region (Figure 2, 2.06 Å, PDB accession code 4Q9S).
Figure 3. Compound 9 model in PKCα compared to 2. Red surface = acidic regions. Blue surface = basic regions. White surface = neutral regions.
compound (9) did not deliver a desired improvement in potency (PKCθ IC50 = 16 μM). A related approach, targeting the PKCθ DFG Asp522 (Asp481 in PKCα) in a nearby region in the protein by substitution off the C6-position of the scaffold with an amino azetidine moiety (Figure 4, 15a), met with
Figure 2. Compound 1 crystal structure in FAK (PDB code 4Q9S).
The FAK−1 cocrystal structure clearly identified the N-NH combination on the triazinone ring as the key hinge binding element to FAK backbone residues Glu500 and Cys502 (Glu459 and Leu461 in PKCθ). On the basis of this crystallographic information, we developed a PKCθ triazinone-bound model that maintained this binding mode and predicted a water-mediated H-bond of the CO with back pocket Thr442, a residue unique to the PKC family. This water molecule was observed in internal PKCα structures with other chemotypes (not shown). We hypothesized that, based on the nonplanar conformation of this simple core combined with the predicted interaction with unique, back pocket Thr442, broad kinome selectivity with this chemotype would be achievable. A structure-based design strategy was developed based on the FAK binding mode. Thus, a modeling overlay of the FAK compound 1 structure with a known structure of 2 in PKCα (PDB accession code 3IW4)11 was utilized for design (Figure 3). An initial approach to drive potency was to target an interaction with the acidic region of the protein with a basic amine. Although several acidic residues line this cavity, initial design focused on targeting Asp508 (Asp467 in PKCα), similar to the interaction observed in the 2 crystal structure. Modeling of a -CH2NMe2 moiety off the C7 position appeared to provide the length and trajectory for optimal interaction; however, this
Figure 4. Model of compound 15a in PKCθ. Red surface = acidic regions. White surface = neutral regions.
greater success (PKCθ IC 50 = 1.3 μM). Comparator compounds replacing the azetidine with a cyclobutyl (15b) or oxetanyl (15c) group were significantly less potent (Table 1, PKCθ IC50 = 38 or PKCθ IC50 > 50 μM, respectively), lending credence to the hypothesis of an amine−Asp interaction. Additional exploration of the C6 position with a variety of amines, via amine or amide linkers, provided compounds 15d− q (Table 1) with a range of potencies, none of which were improved over aminoazetidine 15a. We conjectured that even minor alterations in the trajectory of the amine toward the key Asp interaction could result in a loss in potency. To further improve the potency of these analogues, we targeted a small “lipophilic pocket” proximal to residues Ala407 and Val394 above the plane of the core that could be accessed by substitution adjacent to the carbonyl on the triazinone ring. While the aminoazetidine moiety at C6 was held constant, E
dx.doi.org/10.1021/jm500669m | J. Med. Chem. XXXX, XXX, XXX−XXX
Journal of Medicinal Chemistry
Article
Table 1. Structure−Activity Relationship for C6 Variants
a
Mean from n ≥ 2 experiments.
lipophilic substituents was undertaken, as modeling suggested there was a hydrophobic region that could be accessed for additional potency gain. Substituents such as CH3, F, iPr, and CF3 were tolerated (Table 3) and maintained significant selectivity for PKCθ over PKCα ranging from 17- to 69-fold. These compounds were further profiled in a cellular assay measuring the CD3/CD28-stimulated production of IL-2 from human T-blasts (Table 3). In spite of poor permeability (in all cases 90× for compounds 35 and 39. Stereochemistry had no effect on permeability or unbound microsomal clearance. The (R) stereochemistry of the more active isomer at C4 was later confirmed by small molecule Xray crystallography of C7-methyl derivative 35 (see Supporting
substitution at C4 was explored (Table 2), with initial evaluation of compounds racemic at this position. Addition of Table 2. Structure−Activity Relationship for C4 Variants
IC50 a(μM)
a
compd
R1
PKCθ
PKCα
α/θ
15a 15r 15s 15t 15u
H Me Et cyPr Ph
1.3 0.23 1.0 3.5 >50
7.1 9.4 5.4 >50 >50
4 41 5 >14
Mean from n ≥ 2 experiments.
a methyl group at C4 as in 15r provided a 5-fold boost in potency compared to parent 15a and interestingly demonstrated 41-fold selectivity over PKCα. The C4-Me group of triazinone 15r fits optimally in a small lipophilic pocket (Figure 4), and therefore, larger substituents (Et, cyPr, Ph) at C4 resulted in reduced potency. With enhanced substituents at C4 and C6 apparently identified, subsequent exploration of the C7 position with F
dx.doi.org/10.1021/jm500669m | J. Med. Chem. XXXX, XXX, XXX−XXX
Journal of Medicinal Chemistry
Article
Table 3. Structure−Activity Relationship for C7 Variants
IC50 a(μM) compd
R1
PKCθ
PKCα
α/θ
IC50 a(μM), IL-2
permeabilityb (10−6 cm/s)
CLint,u,c hu/rat (L h−1 kg−1)
15r 15v 15w 21 28
H Me F iPr CF3
0.23 0.11 0.10 0.13 0.067
9.4 4.1 6.9 2.2 2.2
41 37 69 17 33
2.7 0.43 0.31 0.34 0.23
0.42 0.31 0.86 0.28 0.85
2.0/2.9
