Design and Synthesis of Novel Meta-Linked Phenylglycine Macrocyclic FVIIa Inhibitors

Abstract

Two novel series of meta-linked phenylglycine-based macrocyclic FVIIa inhibitors have been designed to improve the rodent metabolic stability and PK observed with the precursor para-linked phenylglycine macrocycles. Through iterative structure-based design and optimization, the TF/FVIIa K_i was improved to subnanomolar levels with good clotting activity, metabolic stability, and permeability.

Cardiovascular diseases remain the leading cause of death worldwide in both developed and developing nations. Many of these deaths can be attributed to arterial or venous thrombosis (e.g., myocardial infarction, ischemic stroke, pulmonary embolism, and deep vein thrombosis).¹ In light of this high level of medical need, the discovery of novel and selective antithrombotic therapies with high efficacy and low bleeding liability would be beneficial.

Factor VIIa (FVIIa) resides at the beginning of the extrinsic coagulation cascade.² When blood vessel walls are compromised, tissue factor (TF) is exposed to circulating FVIIa, leading to the conversion of FIX to FIXa. Both FIXa and FVIIa (in conjunction with their cofactors FVIIIa and TF, respectively) cause the conversion of FX to FXa, which in turn converts prothrombin to thrombin. Thrombin then promotes clot formation by converting fibrinogen to fibrin and activating platelets, generating an extended fibrin/platelet network.³

Various blood coagulation factors have been the target of pharmaceutical small molecule inhibitor research; including, most notably FXa (e.g., apixaban and rivaroxaban) and thrombin (e.g., dabigatran).⁴ FVIIa inhibitors are appealing as anticoagulant therapies for several reasons. First, the efficacy and safety of FVIIa disruption has been demonstrated clinically with recombinant nematode anticoagulant protein c2.⁵⁻⁷ Second, it has been demonstrated preclinically that FVIIa inhibition has excellent efficacy and limited bleeding liability.⁸⁻¹⁵ This profile has made FVIIa a tantalizing target for anticoagulant therapy for many years¹⁶⁻²⁰ and has provided the impetus for the investigations detailed herein.

Of particular interest while considering the design of novel FVIIa inhibitors are the two polar residues (Asp189 and Ser190) that reside at the bottom of the primary specificity pocket (S1). Historically, benzamidines are typically required to bind to these residues and confer the required level of potency; however, this comes at a cost of poor permeability and oral bioavailability, necessitating a prodrug strategy.²⁰⁻²³ We have previously reported structures such as 1 that contain weakly basic benzamidine replacements that confer improvements in permeability and oral exposure while maintaining good potency (Figure 1).²⁴⁻²⁶ This series of compounds exhibited many favorable properties; however, low rodent oral bioavailability was a key limitation. Biotransformation studies traced this poor exposure to the carbamate upper linker in these macrocycles, which was highly metabolized in rodent liver microsomes. Therefore, in an effort to design inhibitors that retained favorable potency and selectivity and incorporated improved metabolic stability and oral bioavailability, a program was initiated to modify the macrocyclic upper linker.²⁷

Figure 1.

General structures of FVIIa inhibitors.

The previously reported para-linked macrocycles (e.g., 1) make several critical contacts with the FVIIa active site,²⁴⁻²⁶ leading to the exquisite potency observed. When designing novel macrocyclic inhibitors, we were cognizant of the fact that these interactions would need to be maintained or even improved in order to retain the desired potency. One such interaction that would need to be considered was the hydrophobic contact with the S2 binding pocket, which is primarily made by the P2-methyl group labeled as “A” in structure 1. It was reasoned that macrocycles linked at the meta position, instead of the para position, of the P2-aryl ring could maintain this favorable hydrophobic contact while significantly altering the linker constitution and modulating the metabolic profile. This publication details the design and synthesis of structures of type 2, which provided a novel scaffold for further elaboration.

Initial efforts surrounding the new chemotype 2 required identification of the appropriate upper linker for the macrocycle. The P′ N–H makes a key hydrogen bonding contact to a conserved water (Figure 2A); therefore, linkers were designed to maintain this critical interaction. Molecular modeling also indicated that 14- and 15-membered macrocyclic rings were optimal, which focused our efforts on linkers that would maintain this ring size. As such, a series of linkers of different lengths and constitutions were prepared and evaluated in the TF/FVIIa activity assay (Table 1).¹⁵ Linkers of 4-atoms (3, 6, 7) and 5-atoms in length (4, 5, 8) were prepared, as they were expected to be optimal based on molecular modeling overlays with 1. Within the 4-atom linkers, the position of the oxygen was varied (e.g., 3, 6) until it was discovered that removal of the oxygen altogether (amide linker, 7) resulted in a significant improvement in binding affinity. The lower activity of 4 may originate from strain of the bound conformation in which the ester portion of the carbamate is twisted out of plane, as observed in the crystal structure of compound 4 bound to FVIIa (Figure 2).²⁸⁻³² Removal of the oxygen, as in the amide system 7, eliminates this penalty and possibly accounts for the improved potency (see Figure 3 depicting the crystal structure of 7 bound to FVIIa). Furthermore, heteroatom substitution was evaluated in the 5-atom linker series, which revealed that the urea linker (8) was significantly more potent. In contrast to the carbamates, the urea does not pay a similar thermodynamic penalty to adopt this conformation, which can account for the favorable impact on potency. From these results, the urea (8) and amide (7) linkers were selected for further optimization due to their promising potency. We elected to optimize the urea linker first due to its superior potency relative to the amide linker.

Figure 2.

(A) Crystallographic structure of compound 4 bound to FVIIa determined at 1.9 Å resolution. The Fo-Fc omit electron density map is depicted at 3 RMSD.^33,34 (B) Side view of the strained carbamate substructure. Graphics were generated using the program PyMol.³⁵

Table 1. Upper Linker Optimization.

cmpd	X	TF/FVIIa Ki 25 °C (nM)a	TF/FVIIa Ki 37 °C (nM)a,c
3	–CH2O–	260
4	–CH2CH2O–	190
5	–CH2OCH2–b	23
6	–OCH2–	710
7	–CH2CH2–	<5	58
8	–CH2CH2N(CH3)–		6.3

^aTF/FVIIa K_i values were obtained from recombinant human enzyme and were averaged from two experiments.¹⁵

^bContains 4-OMe on the P2 aromatic ring.

^cDuring the course of optimizations, the assay was modified to 37 °C to better approximate physiological conditions.

Figure 3.

Crystallographic structure of compound 7 bound to FVIIa determined at 1.85 Å resolution. The Fo-Fc omit electron density map is depicted at 3 RMSD.^33,34 Graphics were generated using the program PyMol.³⁵

Molecular modeling revealed the presence of a hydrophobic binding pocket near the C-4 position of the P2-phenyl ring, which the unsubstituted analogues were not able to contact (vide supra) as shown in Figure 4. It was hypothesized that substitution on the urea nitrogen could potentially improve the binding affinity by contacting the hydrophobic binding pocket above the P2 aryl ring. Additionally, the 4-fluoroaminoisoquinoline P1 moiety was incorporated due to its favorable impact on permeability.²⁶ Disappointingly, a variety of substitutions at the urea nitrogen actually led to a decrease in potency (10–13) (Table 2). Unable to improve the potency via interactions with the P2 hydrophobic pocket from the urea nitrogen, further inspiration was drawn from the binding model and substitution at the P2 benzylic position. Modeling suggested that para-methoxy and benzylic substitution might better fill the S2 pocket, so the para-methoxy derivatives with both enantiomers of the corresponding benzylic methyl groups were prepared. Gratifyingly, compound 14 exhibited greatly improved potency when compared to the unsubstituted compound. However, the metabolic stability was poor across the series of urea-based macrocycles and rat oral exposure of compound 10 was extremely low (PAMPA pH 7.4 = 220 nm/s, rat po exposure below the limits of quantitation), which led to the decision to discontinue the work on the urea series and turn our attention to the amide upper linker for further optimization.

Figure 4.

Overlay of crystal structure of the des-fluoro analogues of 1(25) and model of 16 bound in FVIIa. Graphics were generated using the program PyMol.

Table 2. Urea Optimization.

cmpd	X	Y	Z	R	TF/FVIIa Ki 25 °C (nM)	TF/FVIIa Ki 37 °C (nM)	H,RLM t1/2 (min)
9	H	H	H	H	17
10	H	H	H	Me	14	200	5,6
11	H	H	H	Bn	80
12a	H	H	H	CH2cPr	59		5,12
13a	H	H	H	CH2cBu	150
14	OMe	Me	H	Me	<5	3.8	4,4
15a,b	OMe	H	Me	Me	47		6,6

^aRacemic at the phenylglycine center.

^bdes-Fluoro analogue of the aminoisoquinoline.

In a manner analogous to the urea upper linkers, it was reasoned that substitution of the P2-aryl ring could improve the potency by interactions with the hydrophobic binding pocket. As such, amide-linked macrocycles containing substituted P2-aryl rings were prepared and evaluated, the results of which are delineated in Table 3. It was discovered that substitution at the C-4 position led to approximately a 10-fold improvement in TF/FVIIa potency (e.g., 18, 21, 22, vs 16). Conversely, substitution at the C-5 position had little to no effect on potency (19), presumably because this position is solvent exposed and substituents are not directed into the binding pocket. Monosubstitution at the C-4 position provided relatively flat potency SAR in both the TF/FVIIa and FVIIa deficient prothrombin time assays (anticlotting assay in human plasma, expressed as the concentration required to prolong clotting time by 2-fold), but it modulated rat metabolic stability (as judged by half-life in rat liver microsomes). Providing an appropriate balance between potency and metabolic stability, the difluoroethoxy substitution (21) was selected for further optimization.

Table 3. Amide Optimization: P2 Aryl.

cmpd	X	R	TF/FVIIa Ki 37 °C (nM)	FVII def. PT EC2X (μM)a	H,RLM t1/2 (min)	PAMPA pH 7.4,5.5 (nM/s)
16	H	H	260	10	13,33	120,67
17	H	F	210	26
18	H	OMe	18	11	14,16	540,350
19	OMe	H	290	14	15,46	290,15
20	H	cPr	77	26	26,5
21	H	OCH2CHF2	27	15	32,64	-,88
22	H	OCH2cPr	26	29	23,23
23	H	OH	92	14
24	F	OMe	410	41	14,8

^aSee ref (15) for a detailed description of the FVII-deficient prothrombin time assay.

Additional modeling suggested that the amide upper linker may not optimally fill the S2 binding pocket, and we hypothesized that a methyl substituent at the benzylic position in the proper stereochemical orientation should improve this binding interaction, analogous to the urea series (vide supra). Both enantiomers of the macrocycle containing the benzylic methyl group were prepared (Table 4), and it was discovered that the (R)-enantiomer (26, assigned by X-ray cocrystal structure, Figure 5) was significantly more potent than the (S)-enantiomer (25), resulting in a 100-fold increase in TF/FVIIa potency over the unsubstituted compound (21) (see Figure 5), with concomitant improvement in the FVIIa deficient prothrombin time assay. Selected compounds from this series were studied in rat oral PK once promising potency, permeability, and metabolic stability was achieved. Compounds 16 and 18 demonstrated very low (3.7 F%) or no rat oral exposure, respectively, despite reasonable permeability (PAMPA assay). While excellent potency against FVIIa and improved metabolic stability were achieved with compound 26, off-target liabilities, namely, poor tissue kallikrein selectivity and poor rat PK, within this series limited its ability to progress further.

Table 4. Amide Optimization: Benzylic Substitution.

cmpd	X	Y	TF/FVIIa Ki 37 °C (nM)	TK1 Ki (nM)	aPC Ki (nM)	FVII def. PT EC2X (μM)	H,RLM t1/2 (min)	PAMPA pH 7.4,5.5 (nM/s)
21	H	H	27	140	1200	15	32,64	–,88
25	Me	H	67	23	5700	23	14,74
26	H	Me	0.24	9.4	39	3.5	31,39

Figure 5.

Overlay of crystallographic structures of the des-fluoro analogue of compound 1(25) (green) and 26 (yellow) bound in FVIIa. Graphics were generated using the program PyMol.

Synthesis of the macrocyclic analogues involved two key steps: (1) Petasis reaction of a functionalized P2 boronic acid to form the phenyl glycine core and (2) macrocyclization of either the lower amide or upper linker amide, carbamate, or urea. Synthesis of the amide macrocycles was typically accomplished via late stage lower amide macrocyclization (Scheme 1). As such, the sequence began with preparation of the appropriately substituted P2 acids according to modifications of literature protocols and known chemistry (27). Formation of the upper linker amide bond was achieved via conversion of the acid 27 to the corresponding acid chloride followed by base-mediated coupling with aniline 28 to afford intermediate 29. The requisite boronic acid 30 was prepared using Suzuki–Miyaura borylation conditions, which set the stage for the key Petasis reaction. Condensation of the protected P1 aniline 31 with glyoxylic acid and boronic acid 30 smoothly provided the desired phenyl glycine core (32) in good yield. Deprotection of 32 to give amine 33 provided the necessary intermediate for the final macrocycle formation. Successful macrocyclization required slow inverse addition of 33 to a solution of the coupling reagents to provide good yields of the desired macrocycle 34. Final chiral separation and deprotection provided compound 18 in 14% yield over two steps. The lower yield of this two-step sequence is primarily attributed to the material loss upon chiral separation (50%) and final HPLC purification of the unprotected aminoisoquinoline.

Scheme 1. Representative Synthesis of Compound 18.

(COCl)₂ (1.1 equiv), DMF (0.1 equiv), DCM, 30 min; DMAP (0.1 equiv), TEA (1 equiv), 28 (1.5 equiv), 16 h.

KOAc (2.5 equiv), bis(neopentyl glycolato)diboron (1.4 equiv), PdCl₂(dppf) (0.1 equiv), DMSO, 1,4-dioxane, 90 °C, 3 h.

CHOCO₂H·H₂O (1 equiv), 31 (1 equiv), DMF, MeCN, 60 °C, 16 h.

TBAF (10 equiv), THF, 3 h.

BOP (2 equiv), DMAP (4 equiv), DIPEA (10 equiv), DMF, DCM, 10 h inverse addition.

Chiral HPLC.

TFA.

Efficient access to the urea containing macrocycles was more difficult and required extensive optimization of the synthetic route to access the final compounds in good yield. Scheme 2 delineates this optimized route. Beginning with the amine 35, installation of a Teoc protecting group and alkylation of the resulting amine generated intermediate 38. Boronic acid installation (39), Petasis reaction, and subsequent amine formation proceeded smoothly to generate intermediate 40. Removal of the Teoc protecting group provided 41 and set the stage for the key macrocyclization reaction. Extensive optimization eventually revealed that formation of the activated carbamate under carefully controlled conditions (5 equiv of p-NO₂Ph-OCOCl and 10 equiv of pyridine, 0 to 23 to 50 °C) could effect smooth urea formation in good yield to give 42. Chiral separation and deprotection provided compound 10.

Scheme 2. Representative Synthesis of Compound 10.

p-NO₂-PhOTeoc (1 equiv), DIPEA (1.2 equiv), MeOH, 2 h.

NaH (1.5 equiv), DMF, 15 min; MeI, 16 h.

KOAc (2.5 equiv), bis(neopentyl glycolato)diboron (1.1 equiv), PdCl2(dppf) (0.05 equiv), DMSO, 1,4-dioxane, 90 °C, 4 h.

CHOCO₂H·H₂O (1 equiv), 31 (1 equiv), DMF, MeCN, 60 °C, 16 h; 37 (1.3 equiv), DIPEA (5 equiv), BOP (1 equiv), 23 °C, 3 h.

TBAF (10 equiv), THF, 3 h.

Pyr (10 equiv), p-NO₂PhOCOCl (5 equiv), DCM, 0 to 23 to 50 °C, 16 h.

Chiral HPLC.

TFA.

In conclusion, a series of factor VIIa inhibitors were designed and prepared in order to address the poor rodent metabolic stability and oral exposure of compound 1. The meta-linked macrocyclic series of factor VIIa inhibitors was designed based on careful modeling of the VIIa active site. Using the information garnered through this analysis, the potency of these novel inhibitors was improved from submicromolar to subnanomolar levels with concurrent improvement in rodent metabolic stability; however, the compounds were not able to solve the larger issues of tissue kallikrein selectivity and poor rodent PK, which precluded further advancement of the series.

Glossary

ABBREVIATIONS

FVIIa

factor VIIa

TF

tissue factor

H,RLM

human, rat liver microsome metabolic stability

PAMPA

parallel artificial membrane permeability assay

TK1

tissue kallikrein-1

aPC

activated protein C

PK

pharmacokinetic

FIX

factor IX

FIXa

factor Ixa

FVIIIa

factor VIIIa

FX

factor X

Fxa

factor Xa

DMF

N,N-dimethylformamide

DCM

dichloromethane

DMAP

dimethylamino pyridine

DMSO

dimethyl sulfoxide

TBAF

tetra-N-butylammonium fluoride

THF

tetrahydrofuran

BOP

(benzotriazol-1-yloxy)tris(dimethylamino)phosphonium hexafluorophosphate

DIPEA

diisopropylethylamine

HPLC

high performance liquid chromatography

TFA

trifluoroacetic acid

HATU

1-[bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxidhexafluorophosphate

NMM

N-methylmorpholine

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.6b00375.

• Synthesis and characterization data of key compounds (PDF)

The authors declare no competing financial interest.