Thiol-based angiotensin-converting enzyme 2 inhibitors: P¹ modifications for the exploration of the S¹ subsite

Graphical abstract

Screening of a metalloprotease library led to the identification of a thiol-based dual ACE/NEP inhibitor as a potent ACE2 inhibitor. Modifications of the P¹ benzyl moiety led to improvements in ACE2 potency as well as to increased selectivity versus ACE and NEP.

Abstract

Screening of a metalloprotease library led to the identification of a thiol-based dual ACE/NEP inhibitor as a potent ACE2 inhibitor. Modifications of the P¹ benzyl moiety led to improvements in ACE2 potency as well as to increased selectivity versus ACE and NEP.

Angiotensin-converting enzyme 2 (ACE2) is a recently identified clan MA, family M2 monocarboxypeptidase with highest homology to the dicarboxypeptidase angiotensin-converting enzyme (ACE, EC 3.4.15.1).1, 2 This membrane-associated and secreted metalloprotease is expressed in heart, kidney, testes, intestine, and lung, and has been implicated in cardiovascular disease, kidney disease, obesity, and lung disease.3, 4, 5

ACE2 processes angiotensin I and the AT₁ and AT₂ receptor agonist angiotensin II to produce angiotensin (1–9) and the mas receptor agonist angiotensin (1–7), respectively, but its exact role in the renin-angiotensin system (RAS) needs to be clarified. One group has reported that C57BL/6 ACE2 (−/−) mice exhibit a severe reduction in cardiac contractility, which is rescued in the ACE (−/−)/ACE2 (−/−) double knockout.⁶ In contrast, another group has disclosed that ACE2 (−/−) mice of 129/SvEv, C57BL/6, and mixed background do not exhibit cardiac contractility defects, but have increased susceptibility to angiotensin II-induced hypertension.⁷ While, a third group has revealed that male ACE (−/Y) mice are more susceptible to heart failure and death after transverse aortic constriction than their normal littermates.⁸ Supporting a role for ACE2 in cardiac function, transgenic mice overexpressing ACE2 in the heart have lower mean arterial pressure with rare focal myocyte vacuolization, myofibril splaying, and nuclear enlargement. Many of these mice develop terminal ventricular fibrillation with lethal arrhythmias.⁹

Furthermore, male ACE2 (−/Y) mice, but not female ACE2 (−/−) mice, also accumulate fibrillar collagen in the renal glomerular mesangium, leading to development of glomerulosclerosis of the kidneys.¹⁰ In addition, ACE2 (−/−) mice exhibit lower body weights than wild type mice with reduced fat mass.¹¹ Moreover, ACE2 is also utilized by the severe acute respiratory syndrome (SARS) coronavirus as the receptor for infection.¹² ACE2 (−/−) mice are resistant to SARS corona virus infection.¹³ Finally, ACE2 (−/−) mice have enhanced vascular permeability, increased lung edema, and worsened lung function in several murine acute respiratory distress syndrome (ARDS) models.¹⁴ With the many potential functions of ACE2, small molecule inhibitors of this enzyme could be utilized to help further define the physiological roles of this protease.

ACE2 belongs to the zinc metalloprotease family and it has been reported that classical ACE inhibitors such as captopril and lisinopril do not attenuate ACE2 enzyme activity. As part of a strategy to discover lead molecules for an ACE2 inhibitor program, a directed screen of ACE2 versus a set of metalloprotease inhibitors from the GlaxoSmithKline compound collection was performed. Surprisingly, although confirming that classical ACE inhibitors like captopril were inactive in the screen, the thiol acid 1a was identified as a potent ACE2 inhibitor (K _{i App}  = 86 nM).¹⁵ This biphenyl analog 1a is a known dual ACE and M13 metalloprotease neutral endopeptidase (neprilysin, NEP, EC 3.4.24.11) inhibitor (ACE K _{i App}  = 30 nM, NEP K _{i App}  = 1.1 nM).¹⁶ It also inhibits the M14 metalloprotease carboxypeptidase A1 (CpA, EC 3.4.17.1, K _{i App}  = 1,200 nM). Speculating that the thiol functioned as a binding group for the active site zinc and that the carboxylic acid served as a recognition element for the enzyme’s monocarboxypeptidase activity, the benzyl and methylene p-biphenyl moieties were surmised to be the P¹ and ${\mathrm{P}}^{1^{\prime }}$ substituents. With this premise, the structure activity relationships of the P¹ position of the lead inhibitor were explored with the goal of improving potency and reducing ACE and NEP inhibitory activity.

The thiol analogs 1a–1s were prepared as depicted in Scheme 1 . The acids 3a–3s were activated in situ via the carbodiimide, converted into the activated esters with the aza-hydroxybenzotriazole, and then coupled to the amine hydrochloride 4 to produce the fully protected amides. Subsequent hydrolysis of the methyl ester, as well as the thioacetate, with lithium hydroxide afforded the thiol acids 1a–1s.¹⁷

Scheme 1.

Reagents and conditions: (a) 2c and 2e, X S, AcCl, NEt₃, dioxane, 0 °C to rt, 15–37%; (b) 2k, 2m, and 2n, X O, AcSH, DIAD, PPh₃, THF, 0 °C to rt, 24–67%; (c) 2a–2b, 2d, 2f–2j, 2l, 2o–2s, X NH, HBr, NaNO₂, H₂O, 0 °C, 25–80%; (d) AcS⁻K⁺, DMF, 0 °C to rt, 11–72%; (e) EDC, HOAt, i-Pr₂NEt, CH₂Cl₂, 48–96%; (f) LiOH·H₂O, THF, H₂O, 30–94%.

The thiol acids 3a–3s were produced by three different routes. Acids 3c and 3e were prepared by reaction of the commercially available thiols 2c and 2e (X S) with acetyl chloride. In contrast, the acids 3k, 3m, and 3n were produced from the alcohols 2k, 2m, and 2n (X O) via the Mitsunobu reaction with thioacetic acid.¹⁸ Alternatively, the acids 3a–3b, 3d, 3f–3j, 3l, and 3o–3s were synthesized from the commercially available amino acids 2a–2b, 2d, 2f–2j, 2l, and 2o–2s (X NH), by diazotization of the amines to afford the bromides, then subsequent displacement of the bromides, by potassium thioacetate, with stereochemical inversion. Some stereochemical erosion, presumably resulting from formation of the α-lactone before nucleophilic trapping, resulted during the Mitsunobu and diazotization reaction steps, but after coupling with the amine 4, the minor diastereomer could be removed via chromatography.

The structure/activity relationships of P¹ analogs are depicted in Table 1 . The (R) isomer 1b derived from l-phenylalanine displays the P¹ benzyl substituent in the unnatural configuration. This orientation of the P¹ substituent was detrimental to the ACE2 inhibitory activity of diastereomer 1b (K _{i App}  = 1400 nM) as well as that of ACE (K _{i App}  = 520 nM). This loss in inhibition is not surprising, since nature has evolved proteases to recognize the natural configuration of physiological peptide substrates. In contrast, NEP (K _{i App}  = 1.3 nM) accommodated the alternate configuration with no loss in potency.

Table 1.

Inhibition of human ACE2, ACE, NEP, and CpA

#	R1	R2	ACE2 Ki Appa (nM)	ACE Ki Appb (nM)	NEP Ki Appc (nM)	CpA Ki Appd (nM)
1a	CH2Ph	H	86	30	1.1	1200
1b	H	CH2Ph	1400	520	1.3	310
1c	H	H	320	16	13	1700
1d	Me	H	6.9	21	23	14,000
1e	Me	Me	2300	3400	730	>50,000
1f	Et	H	1.4	8.6	0.80	9,300
1g	n-Bu	H	1.8	9.3	1.2	8900
1h	i-Pr	H	1.5	200	13	17,000
1i	(R)-s-Bu	H	1.5	490	27	11,000
1j	(S)-s-Bu	H	1.6	200	2.4	15,000
1k	Cyb	H	2.4	620	5.4	12,000
1l	Cyp	H	1.8	700	38	12,000
1m	Cyh	H	65	13,000	1300	12,000
1n	Ph	H	84	>10,000	410	>50,000
1o	i-Bu	H	1.4	3.2	0.28	11,000
1p	CH2t-Bu	H	7.1	2,700	2.6	22,000
1q	CH2Cyh	H	420	840	130	17,000
1r	CH2β-Np	H	550	210	140	32,000
1s	(CH2)2Ph	H	860	93	2.6	5,000

^aInhibition of recombinant human ACE2 activity in a fluorescence assay using 0.4 nM ACE2, 30 μM MCA-Tyr-Val-Ala-Asp-Ala-Pro-Lys(DNP)-OH as substrate in 1 μM Zn(OAc)₂, 100 μM TCEP, 50 mM Hepes, 300 μM CHAPS, and 300 mM NaCl at pH = 7.5. The average percent coefficient of variance for the K_{i App} values was 45%.

^bInhibition of recombinant human ACE activity in a fluorescence assay using 0.5 nM ACE, 10 μM MCA-Ala-Ser-Asp-Lys-Dap(DNP)-OH as substrate in 1 μM Zn(OAc)₂, 100 μM TCEP, 50 mM Hepes, 300 μM CHAPS, and 300 mM NaCl at pH = 7.5. The average percent coefficient of variance for the K_{i App} values was 58%.

^cInhibition of recombinant human NEP activity in a fluorescence assay using 0.15 nM NEP, 2 μM FAM-Gly-Pro-Leu-Gly-Leu-Phe-Ala-Arg-Lys(TAMRA)-NH₂ as substrate in 1 μM Zn(OAc)₂, 100 μM TCEP, 50 mM Hepes, 300 μM CHAPS, and 300 mM NaCl at pH = 7.5. The average percent coefficient of variance for the K_{i App} values was 43%.

^dInhibition of recombinant human CpA activity in a fluorescence assay using 37 nM CpA, 30 μM Abz-Gly-Gly-Nph-OH as substrate in 1 μM Zn(OAc)₂, 100 μM TCEP, 50 mM Hepes, 300 μM CHAPS, and 300 mM NaCl at pH = 7.5. The average percent coefficient of variance for the K_{i App} values was 38%.

Complete removal of the P¹ moiety caused a modest loss in inhibitory potency versus ACE2 (H, K _{i App}  = 320 nM). This decrease in potency could arise from reduced van der Waals interactions due to the lack of a P¹ substituent, the increased entropic cost for the inhibitor to bind to the protease given the inhibitor’s increased rotational freedom, or a combination of these factors. In contrast to the D-isomer 1b, the ACE activity (K _{i App}  = 16 nM) of the glycine-like analog 1c was not affected by elimination of the P¹ element, while NEP potency (K _{i App}  = 13 nM) of 1c decreased by over an order of magnitude.

The alanine-derived analog 1d (Me, K _{i App}  = 6.9 nM) decreases rotational freedom and was an even more potent ACE2 inhibitor than the starting lead 1a, while increasing selectivity versus both ACE (K _{i App}  = 21 nM) and NEP (K _{i App}  = 23 nM). In contrast, the geminal dimethyl analog 1e (ACE2 K _{i App}  = 2300 nM) decreased inhibitory potency versus not only ACE2, but also ACE (K _{i App}  = 3,400 nM) and NEP (K _{i App}  = 730 nM). Similar to the methyl analog 1d, the linear P¹ ethyl and n-butyl analogs 1f (Et, K _{i App}  = 1.4 nM) and 1g (n-Bu, K _{i App}  = 1.8 nM) were also potent ACE2 inhibitors, but these extensions in P¹ also increased activity versus ACE (1f K _{i App}  = 8.6 nM, 1g K _{i App}  = 9.3 nM) and even more so versus NEP (1f K _{i App}  = 0.80 nM, 1g K _{i App}  = 1.2 nM).

With the hope of improving the selectivity of these thiol-based inhibitors versus ACE and NEP, the effect of branching along the P¹ side chain of this inhibitor class was explored. Branching along the P¹ chain was well tolerated in the S¹ subsite of ACE2. The α-branched iso-propyl 1h (i-Pr, K _{i App}  = 1.5 nM), (R)-sec-butyl 1i ((R)-s-Bu, K _{i App}  = 1.5 nM), (S)-sec-butyl 1j ((S)-s-Bu, K _{i App}  = 1.6 nM), cyclobutyl 1k (Cyb, K _{i App}  = 2.4 nM), and cyclopentyl 1l (Cyp, K _{i App}  = 1.8 nM) analogs and the β-branched iso-butyl 1o (i-Bu, K _{i App}  = 1.4 nM) and neo-pentyl 1p (CH₂ t-Bu, K _{i App}  = 7.1 nM) analogs maintained their ACE2 inhibition. Only the bulkier branched analogs, like the α-branched cyclohexyl 1m (Cyh, K _{i App}  = 65 nM) and phenyl 1n (Ph, K _{i App}  = 84 nM), the β-branched methylene cyclohexyl 1q (CH₂Cyh, K _{i App}  = 420 nM) and methylene β-naphthyl 1r (CH₂β-Np, K _{i App}  = 550 nM), and the γ-branched ethylene phenyl 1s ((CH₂)₂Ph, K _{i App}  = 860 nM), were too large to be accommodated by the ACE2 S¹ subsite.

In contrast, α-branching of the P¹ moiety was poorly tolerated by the ACE S¹ subsite, resulting in substantial decreases in ACE inhibitory activity for the α-branched iso-propyl 1h (K _{i App}  = 200 nM), (R)-sec-butyl 1i (K _{i App}  = 490 nM), (S)-sec-butyl 1j (K _{i App}  = 200 nM), cyclobutyl 1k (K _{i App}  = 620 nM), cyclopentyl 1l (K _{i App}  = 700 nM), cyclohexyl 1m (K _{i App}  = 13,000 nM), and phenyl 1n (K _{i App}  = >10,000 nM) analogs. The S¹ subsite of NEP is more accepting of α-branching in P¹ substituents than the ACE enzyme, but less liberal than ACE2. The α-branched iso-propyl 1h (K _{i App}  = 13 nM), (R)-sec-butyl 1i (K _{i App}  = 27 nM), (S)-sec-butyl 1j (K _{i App}  = 2.4 nM), cyclobutyl 1k (K _{i App}  = 5.4 nM), cyclopentyl 1l (K _{i App}  = 38 nM), cyclohexyl 1m (K _{i App}  = 1300 nM), and phenyl 1n (K _{i App}  = 410 nM) analogs lost some inhibitory activity versus NEP relative to the ethyl analog 1f (K _{i App}  = 0.80 nM). Thus, α-branching at the P¹ group enhanced selectivity for ACE2 protease inhibition relative to ACE and NEP. The (R)-sec-butyl analog 1i (K _{i App}  = 1.5 nM) has 330-fold selectivity versus ACE and 18-fold selectivity versus NEP, while the cyclopentyl 1l (K _{i App}  = 1.8 nM) has 390-fold selectivity versus ACE and 21-fold selectivity versus NEP.

In comparison to α-branching, both ACE and NEP S¹ subsites were more permissive to β-branching in P¹ groups. Thus, the iso-butyl 1o analog is not only a potent ACE2 inhibitor (K _{i App}  = 1.4 nM), but also a low nanomolar ACE (K _{i App}  = 3.2 nM) and subnanomolar NEP inhibitor (K _{i App}  = 0.28 nM). ACE is less tolerant of bulky β-branching, with the neo-pentyl 1p analog exhibiting potent dual ACE2 (K _{i App}  = 7.1 nM) and NEP (K _{i App}  = 2.6 nM) inhibition with good selectivity over ACE (K _{i App}  = 2700 nM). Larger β-branched substituents like the methylene cyclohexyl 1q (ACE2 K _{i App}  = 420 nM, ACE K _{i App}  = 840 nM, NEP K _{i App}  = 130 nM) and methylene β-naphthyl 1r (ACE2 K _{i App}  = 550 nM, ACE K _{i App}  = 210 nM, NEP K _{i App}  = 140 nM) attenuate inhibitory activity at all three enzymes. Finally, as the ethylene phenyl analog 1   s shows, like ACE2 (K _{i App}  = 860 nM), ACE (K _{i App}  = 93 nM) was less tolerant of γ-branching in P¹ groups, while NEP (K _{i App}  = 2.6 nM) accepted it. None of these thiol-based inhibitors was a potent inhibitor of CpA.

A model of the thiol 1i 19, 20 docked into the active site of ACE2 based on the recent X-ray crystal structure²¹ is shown in Figure 1 . It provides insight into the SAR of the thiol series. In addition to binding to the imidazole nitrogens of ³⁷⁴His and ³⁷⁸His and the carboxylate of ⁴⁰²Glu, the active site zinc coordinates the thiol of the inhibitor. Also, the terminal carboxylate of the inhibitor forms electrostatic interactions with the guanidine of ²⁷³Arg and accepts hydrogen bonds from the imidazoles of ³⁴⁵His and ⁵⁰⁵His. Moreover, the carbonyl of ³⁴⁶Pro accepts a hydrogen bond from the amide nitrogen of the inhibitor, while the amide carbonyl is stabilized by a hydrogen bond to ⁵¹⁵Tyr. Furthermore, the ${\mathrm{P}}^{1^{\prime }}$ methylene biphenyl substituent of 1i forms significant lipophilic interactions with the quite large ${\mathrm{S}}^{1^{\prime }}$ channel composed of the lengthwise canal between the two subdomains, including residues ²⁷⁴Phe, ²⁷⁶Thr, ³⁴⁶Pro, ³⁶⁷Asp, ³⁷⁰Leu, ³⁷¹Thr, and ³⁷⁴His. This biphenyl substituent is hypothesized to occupy a different part of this large pocket than the carboxyl inhibitor co-crystallized in 1R4L. The P¹ (R)-sec-butyl group of the inhibitor forms van der Waals interactions with the S¹ pocket composed of ³⁴⁷Thr, ⁵⁰⁴Phe, ⁵¹⁰Tyr, and ⁵¹⁴Arg. As shown in the close-up of the S¹ subsite depicted in Figure 2 , ⁵¹⁰Tyr forms part of a pocket which the α-branched methyl of the P¹ (R)-sec-butyl group occupies. Larger substituents are disfavored as they would require movement of the phenol side chain of ⁵¹⁰Tyr. Thus, this model agrees with the SAR, since the S¹ subsite can tolerate cyclobutyl and cyclopentyl substituents, but is less accepting of a cyclohexyl group.

Figure 1.

A model of the thiol compound 1i bound to the active site of ACE2 based on a X-ray co-crystal structure of a carboxylate inhibitor bound to ACE2 (PDB code 1R4L). The ACE2 carbons are colored cyan with inhibitor 1i carbons colored green. The semi-transparent gray surface represents the molecular surface, while hydrogen bonds are depicted as yellow dashed lines. Several residues were removed for visual clarity. This figure was generated using PYMOL version 1.0 (Delano Scientific, www.pymol.org).

Figure 2.

A close up of the P¹ pocket of the model of the thiol compound 1i bound to the active site of ACE2 based on a X-ray co-crystal structure (PDB code 1R4L). The ACE2 carbons are colored cyan with inhibitor 1i carbons colored green. ⁵¹⁰Tyr is colored in orange to highlight its putative role in the selectivity of 1i. The semi-transparent gray surface represents the molecular surface. This figure was generated using PYMOL version 1.00 (Delano Scientific, www.pymol.org).

A model of the inhibitor 1i 19, 20 docked into the active site of ACE, focusing on the S¹ subsite, based on the X-ray structure of lisinopril with ACE,²² is shown in Figure 3 . From the model, it is apparent that the methyl group of the (R)-sec-butyl P¹ moiety of inhibitor 1i bumps into the enzyme, helping explain the 330-fold selectivity for ACE2 versus ACE. The branching nature of ⁵¹⁸Val (atom CG1) makes this portion of the S¹ pocket smaller than in ACE2. ⁵¹⁸Val is the homologous residue in ACE to ⁵¹⁰Tyr in ACE2. This ⁵¹⁰Tyr to ⁵¹⁸Val substitution from ACE2 to ACE and resulting alteration of the S¹ pocket volume helps explain the reduced ability of ACE to accommodate α-branching in P¹ substituents.

Figure 3.

A model of the thiol compound 1i bound to the active site of ACE using the X-ray co-crystal structure of ACE and lisinopril (PDB code 1O86). The ACE carbons are colored magenta with inhibitor 1i carbons colored green. ⁵¹⁸Val is colored orange to highlight its hypothesized role in the selectivity of 1i. The semi-transparent gray surface represents the molecular surface. This figure was generated using PYMOL version 1.00 (Delano Scientific, www.pymol.org).

Although X-ray structures of NEP co-crystallized with a thiol-based inhibitor (PDB code 1Y8J)²³ and an analog with a ${\mathrm{P}}^{1^{\prime }}$ bi-phenyl moiety (PDB code 1R1H)²⁴ are available, as NEP belongs to a different metalloprotease family (M13 vs M2), no reasonable model of these thiol inhibitors bound to NEP could be determined to help explain the increase in NEP selectivity of α-branched P¹ analogs. Likely protein movement to accommodate potent NEP inhibitors like 1o is required. The movement of multiple amino acid residues in proteins is difficult to predict accurately.

In summary, a series of α-thiol amide-based inhibitors of ACE2 with varied substituents at the P¹ position were synthesized. Inhibitors containing linear alkyl P¹ moieties were some of the more potent analogs in the ACE2 enzymatic assay. The smaller α-branched P¹ substituents, exemplified by inhibitors 1i and 1l, maintained similar potencies, and increased selectivity versus ACE and NEP. Information gained from these studies has proven to be useful in the design of other ACE2 inhibitors. These inhibitors will be reported in due course. Furthermore, a potent pan ACE/ACE2/NEP inhibitor 1o and a potent dual ACE2/NEP inhibitor 1p have been identified. These tools may prove useful in further defining the roles these proteases play in the RAS cascade.