A series of 3, 5-disubstituted benzimidamides were synthesized and biologically evaluated as potential BACE1 inhibitors. Both the targeted compounds (benzimidamides) and the synthetic intermediates (benzonitriles) were tested for their BACE1 inhibitory activities in a cell-free FRET assay. All the synthesized benzimidamides were active as BACE1 inhibitors and compound 6d showed the lowest IC50 value of 3.35 μm. Molecular docking study proposed a binding mode, which would help to the further optimization on 6d to achieve more potent, BBB penetrant BACE1 inhibitors.
Alzheimer’s disease (AD) is a progressive neurological disease of the brain that leads to the irreversible loss of neurons and dementia. More than 30 million people worldwide currently suffer from AD and by the year of 2050, this number is expected to grow up to more than 100 million (1). Unfortunately, there is no effective therapy currently available, although some acetylcholinesterase (AChE) inhibitors and N-methyl-d-aspartate (NMDA) receptor antagonists are in clinical use to temporarily relieve the symptoms. The major pathological hallmark of AD is the deposition and aggregation of β-amyloid (Aβ) in the brain tissues and hence leads to intracellular formation of neurofibrillary tangles (2). Aβ is produced by the sequential catalytic cleavage of amyloid precursor protein (APP) by two proteases, the β-secretase (β-site APP cleaving enzyme, BACE) and γ-secretases (3). β-Secretase cuts APP first to generate the amino terminus of Aβ. Inhibition of β-secretase decreases the production of all forms of Aβ, including the most pathogenic species, Aβ42 (4). For more than a decade, β-secretase has been regarded as a promising therapeutic target for the development of disease-modifying therapies for AD (5,6). It is a transmembrane aspartyl protease that exists in two isoforms (7). BACE1 located in the central nervous system (CNS) and BACE2 expressed mainly in the periphery (8). In the past decade, considerable efforts have been made to develop BACE1 inhibitors and lots of them were designed as transition state isosteres such as hydroxyethylamine (9,10), hydroxyethylene (11) and statine-based (12,13) peptidomimetic inhibitors. Although some of them are very potent in vitro with IC50S at nM level, the relatively large molecular weight (MW>500) and poor pharmacokinetic properties have prevented them from further development to oral bioavailable CNS drugs.
In the recent years, there has been a boom in the development of non-peptidic β-secretase inhibitors with better pharmacokinetic properties as drug leads (14,15). They were discovered by means of different approaches, such as high-throughput screening (HTS), fragment-based and structure-based strategies (16–19).
In an effort to discover novel small molecular inhibitors for BACE1, we carried out a fragment-based in silico screening in our previous work, which identified several fragments, such as guanidine, quinazolin-2-amine and 1H-indol-7-amine, with the predicted potential to bind at the catalytic aspartic acid dyad in the BACE1 active site (20). Among these fragments, benzimidamide was surmised to bind with Asp32 and Asp228 with a relatively high binding score, and indeed exhibited measurable inhibition against BACE1 in the following fluorescence resonance energy transfer (FRET) assay. Therefore, benzimidamide was selected as the starting point to the novel inhibitor design. We envisioned that a structural extension strategy around the benzimidamide scaffold would allow for direct projection towards the unoccupied subpockets around Asp32 and Asp228. As the first step of optimization, a series of 3-phenylethylbenzimidamides (3-PEBIs) were designed, synthesized and tested for BACE1 inhibitory activities in FRET assay. Results indicated that although very weak, these 3-PEBIs demonstrated detectable inhibitory potencies against BACE1. The most potent compound, 3-(3-propoxyphenethyl)-benzimidamide (11, Figure 1A) generated 10.7% and 16.0% inhibition to the positive control at both the tested concentrations of 1 and 10 μm, respectively (20). Docking studies suggested that the benzimidamide scaffold occupied S1 pocket while the amidine moiety forming multiple hydrogen bonds with the catalytic residues of Asp32 and Asp228 (Figure 1B), whereas the 3-phenylethyl substitution extended into the S3 pocket forming hydrophobic interactions with the residues around, leaving the adjacent subpockets unoccupied (Figure 1B). Herein, we report a further optimization of the benzimidamides using a structure-based approach, magnifying the benzimidamide scaffold to more subpockets than S1 and S3. We were aiming at achieving improved potencies and novel lead structures and validating our ‘structural extension strategy’ in lead compound optimization. In this study, several 3, 5-disubstituted benzimidamides were designed, synthesized and tested for their BACE1 inhibitory potencies. Those discovered inhibitors were also predicted for the binding modes in docking study.
Materials and Methods
Docking of the virtual molecules was performed using Gold 3.0 (21), a docking program based on genetic algorithm. The 3D structure of BACE1 employed was 2OHN (22), because of its relatively high resolution of 2.15 Å and the structure similarity between its co-crystallized ligand, 4-(4-fluorobenzyl) piperidine and our designed benzimidamides. The protein was prepared in Discovery Studio 2.5 (a), with the amino acids ionized at pH 7.4, hydrogens added, charges calculated and the waters removed. A radius of 12 Å around the co-crystallized ligand was used as the active site. The GoldScore fitness function was applied, and the starting conformation for compound docking was a low-energy conformer generated using modified CHARMm force field-based 3D structure minimization implemented in Discovery Studio’s ‘prepare ligands’ protocol. Re-docking of the native ligand in 2OHN to the binding site was performed to test the docking parameters and check for the correct preparation of structures for use. The RMSD value for the docked and X-ray orientation of the native ligand was 1.064, suggesting that the docking protocol applied was acceptable.
Unless otherwise noted, all reagents were purchased from Acros (Fair Lawn, NJ, USA), TCI (Shanghai, China), Sigma-Aldrich (St. Louis, MO, USA) and used without further purification. Dry solvents were prepared according to standard procedures. Melting points were taken with an X4 apparatus and were uncorrected. 1H NMR spectra were recorded on a Bruker (Bruker BioSpin AG, Fällanden, Switzerland) Avance III 400 MHz system or an AL 300 MHz system. 1H NMR spectra were reported in the following manner: chemical shifts calculated with reference to solvent standards based on tetramethylsilane (TMS), multiplicity (s, singlet; d, doublet; t, triplet; q, quartet; sept, septuplet; m, multiplet), coupling constant (J) in Hz, number of protons and the corresponding attributions. MS data were obtained with MDS SCIEX QSTAR systems. Thin layer chromatography (TLC) analysis was performed on silica gel GF254 purchased from Qingdao Haiyang Chemical Co. (Qingdao, Shandong Province, China) or Merck (Darmstadt, Germany).
The synthesis of the final 3′-alkoxy-5-phenoxybiphenyl-3-carboximidamides is depicted in Scheme 1. The 3-bromo-5-phenoxybenzonitriles (3a–b) were prepared from the 3,5-dibromobenzonitrile (1) and the corresponding phenols (2a–b) in yields of approximately 50% with Ullmann reaction. Suzuki reaction was then employed to form the carbon–carbon bond between two phenyl rings and give the 5-phenoxybiphenyl-3-carbonitriles (5a–d) from the 3-substituted phenylboronic acids (4a–b). Ammonolysis of the nitriles (5a–d) with lithium hexamethyldisilazane (LHMDS) was conducted at 0 °C under N2 atmosphere for 24 h, and the benzimidamides (6a–d) were afforded as free base in yield of 54–88%. The overall yields were 12–27% (Scheme 1).
BACE1 inhibitory evaluation was carried out using a Fluorescence Resonance Energy Transfer (FRET) assay kit supplied by PanVera (kit P2985, Madison, WI, USA).
The principle of the BACE1 FRET assay is as follows: the peptide substrate is synthesized two fluorophores, a fluorescent donor and a proprietary quenching acceptor. The distance between these two groups has been assigned so that upon light excitation, the donor fluorescence energy is significantly quenched by the acceptor through a quantum mechanical phenomenon known as resonance energy transfer. Upon cleavage by the protease, the fluorophore is separated from the quenching group, restoring the full fluorescence yield of the donor. Thus, a weakly fluorescent peptide substrate becomes highly fluorescent upon enzymatic cleavage; the increase in fluorescence is linearly related to the rate of proteolysis.
The kit protocol was followed with minor modification that the total assay volume was reduced to 15 or 20 μL instead of 30 or 40 μL as mentioned in the kit. Assays were carried out in Matrical #MP101-1-PP black V-shape 384-well plate. The reaction rate was monitored at room temperature on a PHERAStar plate reader with excitation and emission wavelengths of 540 ± 20 and 590 ± 20 nm, respectively. A statin peptide derivative (H-Lys-Thr-Glu-Glu-Ile-Ser-Glu-Val-Asn-Sta-Val-Ala-Glu -Phe-OH; Anaspec 23959, Fremont, CA, USA) was incorporated as a positive control, and DMSO was used as negative control. Initial reaction rates were measured, and IC50 values were calculated using a one-site competition model within graphpad prism (GraphPad Software, Inc., San Diego, CA, USA) 5.04 software.
BBB penetration prediction
The capabilities of the acquired compounds to penetrate BBB were predicted using the ADME/T module within Discovery Studio 2.5. The logBB (the logarithm value of brain to plasma concentration ratio) values were calculated based on a quantitative linear regression model derived from over 800 compounds that are known to enter the CNS after oral administration. According to the logBB values, the compounds can be classified into four levels, which are very high penetrants (logBB ≥ 0.7), high penetrants (0 ≤ logBB < 0.7), medium penetrants (−0.52 < logBB < 0) and low penetrants (logBB ≤ −0.52; 23).
Results and Discussion
Docking of the virtual molecules
Results of the docking studies suggested that compared with our previous reported 3-PEBIs (20), the 3, 5-disubstituted benzimidamides were expected to have improved potencies owing to the additional 5-substitutions, which were predicted to extend to the hydrophobic S2′ pocket and stabilize the hydrogen binding between the amidine moiety and the catalytic Asp dyad. Docking fitness scores (Table 1) of the 3-phenyl-5-phenoxylbenzimidamide, namely 5-phenoxybiphenyl-3-carboximidamide analogues (6a–d), were ranging from 60.48 to 61.67, suggesting they could bind to BACE1 with an improved affinity compared with that of compound 11 (Fitness = 54.56).
Table 1. Predicted binding fitness scores and the score composition of the designed 3-phenyl-5-phenoxylbenzimidamide (6a–d).
Compound 6d was among the best solutions from docking (Figure 2). The biphenyl-3-carboximidamide scaffold of 6d was positioned in the S1 pocket, with the amidine moiety forming multiple H-bonds with the Asp dyad. The 5-phenoxy substitution was observed extending to the S2’ pocket, and at the gate of this pocket, two H-bonds were surmised between the oxygen atom of the 5-substitution and Ser25, Tyr71, respectively. A perpendicular conformation was adopted by the biphenyl rings of 6d, leaving one phenyl ring approximately parallel to that of the Tyr 71, so that a π–π stacking interaction could be formed. The 3′-alkyoxy substitution was observed to locate quite close to Thr 72 in the flap region, instead of extending into the S3 subpocket as predicted in the 3-phenylethylbenzimidamide series.
Synthesis of target compounds
The general procedures for synthesis and the 1H NMR data were provided in the Supporting Information (Appendix S1) of this article.
Additionally, we have also tried to reverse the synthetic sequence to introduce the two substitutions from the same starting material of 3,5-dibromobenzonitrile (1) as depicted in Scheme 2. For the Suzuki cross-coupling reaction, the convention of 1 was convenient, but the yield of 5-bromo-3′-hydroxybiphenyl-3-carbonitrile (8a) was only 46%, which was attributed to the generation of a side product (8b) in nearly the same yield. And the yields of Ullmann reactions were almost the same as those in Scheme 1.
BACE1 inhibitory activity
All the synthesized benzimidamides (6a–d) and some of the 3,5-disubstituted benzonitriles (5a–d, 8a–b, 10a–b) were assessed for BACE1 inhibitory activity in FRET assay. For primary screening, compounds were tested at 10 and 1 μm. The % inhibition and IC50 values of the tested compounds can be found in Table 2.
Table 2. The predicted logBB and results of the FRET assay for BACE1 inhibitory activities.
10 μm (% Inhibition)
1 μm (% Inhibition)
aND means the IC50 was not detected because of the poor activity observed in the primary screening at 10 and 1 μm.
The results were consistent with what we expected that compound 6d with the highest docking Fitness score (Table 1) demonstrated the highest potency with the IC50 of 3.35 μm. Compared with the 3-monosubstituted benzimidamide 11, all 3,5-disubstituted benzimidamide analogues 6a–d were active as BACE1 inhibitors with IC50 values <10 μm, generating more than 70% of the response achieved by the positive control at 10 μm level. The concentration–response relationships were conspicuous as shown in Figure 3. To our surprise, some of the tested 3,5-disubstituted benzonitriles, which were expected to be unable to form hydrogen bonds with the Asp dyad because of the lack of hydrogen-bond donors, were also found to have weak potencies, among which 5d generated the lowest IC50 of 9.35 μm. Additionally, compound 8b also showed considerable potency, which is very similar to 5d. The docking result (Figure 4) suggested that no hydrogen bond was formed between 5d or 8b and the catalytic Asp dyad. The position of the cyano nitrogen of 5d was 4.434Å away from that of the amidino nitrogen in 6d and as for 8b, this nitrogen was even further away. The whole molecule of 5d was more positioned in the S1′–S2′ subpocket based on the polar interactions between the 5-substituted oxygen and Arg128, in addition to the π-cation interaction between the 5-phenyl ring and the guanidine nitrogen of Arg128. This π-cation interaction was reserved in the binding mode of 8b, and the hydroxyl substituent on both of phenyl rings in 8b was observed to form four additional hydrogen bonds with Thr 231 and Asn37. The superposition of the docking poses of 5d, 8b and 6d indicated that although much favoured, the direct hydrogen-binding interactions forming between the ligand and the Asp dyad might not be necessarily important for BACE1 inhibition.
Some structure–activity relationships for R1 and R2 variation can be observed from the results. The bulky n-butyl substitution at R2 was more favoured than n-propyl group because of the slightly higher potencies observed in R2 n-butyl-substituted benzimidamides (6b and 6d) than the corresponding n-propyl-substituted analogues (6a and 6c). As expected, the R1 fluoro-substituted benzimidamides (6c and 6d) were found to be twofold more potent when compared to the corresponding non-substituted 6a and 6b. This was even more distinct for the 5-phenoxybiphenyl-3-carbonitriles (5a–d), because the potencies of 5a and 5b were so weak that the IC50s were not detected, whereas the corresponding R1 fluoro-substituted 5c and 5d were much more potent with the IC50s of 80.35 and 9.35 μm, respectively.
BBB penetration capabilities
The predicted logBB values for each compound as presented in Table 2 suggested that all synthesized compounds in our work were BBB penetrable. The synthetic intermediates 5a–d are possibly very high penetrants with logBB values larger than 0.8, and all the targeted 3, 5-disubstituted benzimidamides are high penetrants with logBB values between 0.14 and 0.35. There is a significant improvement in BBB penetration of 3, 5-disubstituted benzimidamides compared with the 3-monosubstituted benzimidamide 11, of which the logBB is only −0.060 and defined as medium penetrant.
Conclusion and Future Directions
Overall, our extension strategy around the benzimidamide scaffold has led to the discovery of 3-phenyl-5-phenoxylbenzimidamide derivatives as moderate BACE1 inhibitors, among which 6d was the most potent one deserving further optimization. Future work will focus on reducing the basicity and improving the potency by derivation on amidine and phenyl rings to achieve the BBB penetrable BACE1 inhibitors.