Pharmacophore-based design and discovery of (−)-meptazinol carbamates as dual modulators of cholinesterase and amyloidogenesis

Abstract Multifunctional carbamate-type acetylcholinesterase (AChE) inhibitors with anti-amyloidogenic properties like phenserine are potential therapeutic agents for Alzheimer’s disease (AD). We reported here the design of new carbamates using pharmacophore model strategy to modulate both cholinesterase and amyloidogenesis. A five-feature pharmacophore model was generated based on 25 carbamate-type training set compounds. (−)-Meptazinol carbamates that superimposed well upon the model were designed and synthesized, which exhibited nanomolar AChE inhibitory potency and good anti-amyloidogenic properties in in vitro test. The phenylcarbamate 43 was highly potent (IC50 31.6 nM) and slightly selective for AChE, and showed low acute toxicity. In enzyme kinetics assay, 43 exhibited uncompetitive inhibition and reacted by pseudo-irreversible mechanism. 43 also showed amyloid-β (Aβ) lowering effects (51.9% decrease of Aβ42) superior to phenserine (31% decrease of total Aβ) in SH-SY5Y-APP695 cells at 50 µM. The dual actions of 43 on cholinergic and amyloidogenic pathways indicated potential uses as symptomatic and disease-modifying agents.


Introduction
Alzheimer's disease (AD) is an age-related neurodegenerative disorder that causes the majority of dementia in the elderly. Pathologically, AD is characterized by the progressive loss of basal forebrain cholinergic neurons 1 , and neuropathological changes of abnormally accumulated extracellular amyloid-b peptide (Ab) 2 and intracellular tau protein 3 . However, the underlying mechanisms of AD are still poorly understood, which may be attributed to the complex multipathogenic features 4,5 , including amyloidogenic processing of amyloid precursor protein (APP), Ab aggregation, tau hyperphosphorylation, calcium dyshomeostasis, oxidative stress, mitochondrial dysfunction, deterioration of synaptic neurotransmission, and neuronal apoptosis.
Current approved anti-AD drugs are all palliative treatments targeting cholinergic or glutamatergic neurotransmission thereby symptomatically improving memory and cognition in patients. Acetylcholinesterase (AChE) inhibitors ( Figure 1) such as tacrine, donepezil, rivastigmine, and galanthamine, are major palliative treatments available now. Carbamates are classical pseudo-irreversible AChE inhibitors, which bind to AChE catalytic site covalently via carbamylating conserved serine residue, and therefore delay the reactivation of an unbound enzyme. Physostigmine (1) is the first AChE inhibitor separated from natural products, but unacceptable toxicity limits its clinical use. Rivastigmine (14) is the only carbamate AChE inhibitor approved as anti-AD drug on the market.
Over the last decade, much effort has been devoted to amyloidogenesis (APP generation/metabolism) and Ab-induced neurotoxicity. Unfortunately, to date, Ab-directed therapies, such as c-secretase inhibitors and immunotherapies 6 , were too toxic to succeed in clinical trials 7 . It seems that the one-molecule-one-target paradigm is inadequate to address the unmet disease-modifying goal of anti-AD drugs. In view of the multifactorial nature of AD pathogenesis, a "multi-target-directed ligands" (MTDLs) 8 strategy was applied in recent development of modifying treatments for AD [8][9][10] . Single molecule directing toward different biological targets involved in AD etiology showed promising multipotent profiles. Multifunctional AChE inhibitors with anti-amyloidogenic properties have been most widely studied 11 because of their symptom-alleviating ability and disease-modifying potential.
Phenserine (5) 12 , a phenylcarbamate of (À)-eseroline (24), developed by Greig et al. 13 , is a unique multipotent AChE inhibitor. It reduces the levels of APP and Ab via a non-cholinergic mechanism by down-regulating the translation of APP mRNA 14 . The phenylcarbamoyl moiety of phenserine seems to be crucial for its anti-amyloidogenic effect. Although phase III clinical trials of phenserine failed due to lack of efficacy 15 , a redevelopment after correcting some methodological deficiencies 16 might bring to new conclusions.
In our earlier research, bis-(À)-nor-meptazinols [17][18][19] and their derivatives 20 were characterized as dual binding site AChE inhibitors with anti-Ab-aggregation and/or metal-complexing properties. The crystal structure of bis-(À)-nor-meptazinol/AChE complex 18 was resolved, which revealed the binding pose of (À)-meptazinol moiety in the AChE catalytic site. To further explore (À)-meptazinol monomer derivatives, we reported here the design of new carbamates using pharmacophore model strategy with the aim to modulate both AChE activity and amyloidogenesis.
Inhibitory potencies of 42 and 43 for AChE and butyrylcholinesterase (BChE) were tested in vitro. Enzyme kinetic parameters, Michaelis constant (K m ) and maximum velocity of reaction (V max ), were measured on recombinant human acetylcholinesterase (rHuAChE). The association and dissociation rate constants, namely inhibit constant (k i ), dissociation constant (k 3 ), and affinity constant (K D ), were determined using AChE immobilized disk. Anti-amyloidogenic experiments were conducted employing high content screening (HCS) in SH-SY5Y-APP 695 cells, and enzyme-linked immunosorbent assay (ELISA) in the cell culture medium. Mechanisms for the actions of 42 and 43 on reducing APP and Ab 42 levels were further discussed.

Pharmacophore modeling
Pharmacophore model generation and validation were performed using 3D QSAR Pharmacophore Generation module and Ligand Pharmacophore Mapping module, respectively, in Discovery Studio v2.5 (DS, Accelrys, San Diego, CA) software package. Carbamatetype AChE inhibitors with comparable IC 50 s tested by Ellman's method 36 and using physostigmine or rivastigmine as positive control were collected from the literature [21][22][23][24][25][26][27][28][29] to generate quantitative pharmacophore hypotheses. The IC 50 values covered a range of three to four orders of magnitude and the activity uncertainty was set 3 as default. All two-dimensional (2D) structures of the compounds were built using ISIS Draw v2.2 (MDL Information Systems, Inc.) and exported into DS to be converted into 3D format. A maximum of 255 conformers were generated for each compound over a 20 kcal/mol range using the BEST conformational analysis method. As an exception, (À)-meptazinol and its carbamate derivatives were calculated by both BEST and CAESAR conformation algorithms. Chemical features including hydrogen bond acceptor (HBA), ring aromatic (RA), positive ionizable (PI) and hydrophobic (HYD) features were selected and each feature was given parameters from a minimum of 1 to a maximum of 5. The minimum interfeature distance was set as a value of 2.50 and the maximum excluded volumes was set to 5. Different weights were assigned to the features and weight variation was the default value of 0.302. Otherwise default parameters were used.

General
All reagents except phenyl isocyanate were of commercial quality. Phenyl isocyanate was prepared from aniline and bis(trichloromethyl) carbonate. Rivastigmine hydrochloride standard was available from Sunve (Shanghai) Pharmaceutical Co., Ltd. Melting points were measured in open glass capillary tubes with Thiele-Dennis tube, and were uncorrected. Specific rotation ([a] D ) was determined on a JASCOP-1020 rotatory apparatus. IR data were recorded on an AVATAR 360 FT-IR spectrometer (KBr). NMR data were recorded on a Mercury Plus 400 instrument. Chemical shifts (d) are expressed in parts per million (ppm), relative to tetramethylsilane (TMS) as an internal standard. Signals of active hydrogen disappeared after D 2 O exchange. Mass spectrum was measured on an Agilent 1100 Series LC/MSD 1946D spectrometer. Elemental analysis was tested on vario EL III element analyzer. Purity of the target compound was verified via HPLC. The elution with acetonitrile-0.01 mol/L KH 2 PO 4 (pH ¼4.0) (33:67) was running through a Diamonsil V R C18(2) (200 Â 4.6 mm, 5 mm) column at a flow rate of 1.0 mL/min and at the temperature of 30 C using UV detection at 233 nm.
Synthesis of (S)-3-(3-ethyl-1-methylazepan-3-yl)phenyl dimethylcarbamate (42) To a cooled and stirred mixture of 80% sodium hydride (0.15 g, 5.00 mmol) in 10 mL dry tetrahydrofuran, a solution of (À)-meptazinol (0.40 g, 1.71 mmol) in 10 mL dry tetrahydrofuran was added dropwise. The mixture was stirred in ice-water bath for 30 min, then N,N-dimethylcarbamoyl chloride (195 mL, 2.06 mmol) was added. After stirring at room temperature for 2 h, solvents were removed under reduced pressure. Then, water (20 mL) was added, and the mixture was extracted with EtOAc (15 mL Â 2). Combined organic layer was washed by brine, dried over anhydrous Na 2 SO 4 , and concentrated in vacuo to give a whitish oil 42 (0.52 g, 100% yield). Adding dry HCl-ether to a solution of 42 in dry ether (pH adjusted to 4) afforded the hydrochloride salt 42ÁHCl (0.56 g, 97% yield). Crystallization of 42ÁHCl (0.50 g) from acetone gave needlelike crystals (0. 17   Synthesis of (S)-3-(3-ethyl-1-methylazepan-3-yl)phenyl phenylcarbamate (43) (À)-Meptazinol (0.40 g, 1.71 mmol) was dissolved in anhydrous ether (15 mL), and a piece of Na metal (approximately 5 mg) was added. The mixture was stirred under nitrogen at room temperature for 10 min, then phenyl isocyanate (233 mL, 2.13 mmol) was added. The reaction mixture was stirred at room temperature for 3 h till the starting material had disappeared. 5 mL of H 2 O were added to destroy any trace of remaining unreacted phenylisocyanate and pH was adjusted to 3 by adding 1N HCl. The mixture was washed with ether (10 mL Â 3), basified with saturated Na 2 CO 3 aqueous solution (adjusting pH to 9), and then extracted with ether (10 mL Â 3). The latter ether layer was washed with brine, dried over anhydrous Na 2 SO 4 , and filtered to obtain a clear ether solution of the product. Evaporation of the solvent gave 43

In vitro AChE/BChE inhibition assays
Inhibitory activities of the compounds toward AChE and BChE were evaluated by Ellman's method 36 , employing mice brain homogenate as source of AChE and mice serum as source of BChE. Briefly, 270 mL of a solution of AChE (1:9 w/v homogenate in 0.1 M phosphate buffer (PB), pH 7.4) and 30 mL of a solution of the tested compound (42, 43, or rivastigmine, six to seven concentrations) were mixed adequately. After incubation for 20 min at 37 C, Ellman's reagent (300 mL, 5,5 0 -dithiobis(2-nitrobenzoic acid) (DTNB), 0.5 mM in 0.1 M PB, pH 7.4) and acetylthiocholine iodide (ATCh) (300 mL of 0.5 mM water solution) were added successively, and percent inhibition was determined by absorbance changes at 412 nm detected by UV spectrophotometry compared with control. BChE inhibition assay was similarly carried out using butyrylthiocholine iodide (BTCh) (0.5 mM) as the substrate and BChE (1:19 v/v serum in 0.1 M PB, pH 7.4) as the enzyme source. The concentration of a compound that produced 50% inhibition of the enzyme activity, namely IC 50 value, was calculated by nonlinear least squares regression of the response-concentration (log) curve. Results are reported as the mean ± SEM (standard error of the mean) of IC 50 obtained from at least three independent measures 38 .
Determination of the enzyme kinetic parameters K m and V max 10 mL inhibitors of different concentrations and 50 mL rHuAChE enzyme solution of 0.5 U/mL were mixed and incubated for 20 min, then 75 mL of DTNB solution and 100 mL of ATCh solution (concentration ranging from 0.057 mM to 0.2 mM) were added to the mixture. Enzyme activity was determined right after ATCh was added by modified Ellman's spectrophotometrical method 37 . The K m and V max values for AChE inhibition were calculated by regression analysis of Lineweaver-Burk plots (1/velocity versus 1/[substrate]).

Determination of carbamoylation and decarbamoylation rate constants
The EDA CIM disk was first connected to a syringe pump and equilibrated with 10 column volumes (CV) of triple-distilled water and 10 CVs of 0.5 M Na-PB (pH 8.0). The flow rate was set at 0.5 mL/min. Then, the disk was activated by a 1% glutaraldehyde solution in PB (0.1 M, pH 7.4) (in the dark, 12 h, 4 C). The activated disk was washed with 10 CVs of PB. 2 mL of rHuAChE in PB (10 U/ mL) was then added into the column and left to react overnight (4 C). The Schiff bases were reduced by 5 mL of 0.1 M sodium triacetoxyborohydride (STAB) in PB (2 h at room temperature) and 5 mL of 0.2 M monoethanolamine in PB (3 h at room temperature) 39 . The column was connected to the liquid chromatography system and equilibrated with a mobile phase consisting of 0.5 M NaCl and 10 À4 M DTNB in PB (buffer A) for 10 min. Reference enzyme activity was assessed by injecting 20 mL of saturating substrate (ATCh, 50 mM). The peak area was calculated and noted as A 0 . Then, the mobile phase was switched to the one containing selected concentrations of inhibitor (carbamoylation phase). Aliquots of 20 mL of saturating substrate were injected every 10-20 min and the time-dependent decreasing of peak area (A i ) was monitored down to a plateau. Once a constant plateau was reached, mobile phase was switched to buffer A again (decarbamoylation phase). Aliquots of substrate were kept injecting into the system to monitor the recovery of AChE activity over time. Percent inhibition of enzyme activity [(A 0 ÀA i )/A 0 Â100%] was plotted versus time. The experiment was carried out for two different concentrations of 42 and 43 (50 and 100 nM). The process of AChE carbamoylation and decarbamoylation is generally described as Scheme 1 [40][41][42] .
Calculation of k i and k 3 was performed by applying Perola's mathematical equation 43 . The affinity of the inhibitors toward the enzyme (K D ) was calculated following Equation (1).

Acute toxicity test
Acute toxicity was evaluated in mice of both sexes (20-25 g Kunming mice from experimental animal center of Shanghai Jiao Tong University School of Medicine). All animals were housed in plastic cages with food and water ad libitum and maintained on a 12/12 h light/dark cycle at 22 ± 1 C. They were randomly assigned to one of the five concentrations between 0% and 100% lethal rate according to our preliminary studies (data not shown). Compounds 42 and 43 were dissolved in a 1:10 propylene glycolnormal saline mixture at 0.1 mol/L and then diluted to the final stepwise concentrations with normal saline. Each concentration of the compound was orally or intraperitoneally administered to a group of 10 animals. After two weeks of observation, the lethal rate for each group was measured. A 95% confidence interval for 42 or 43 after oral or intraperitoneal administration was calculated by Bliss method.

Results and discussion
3D pharmacophore generation To find out common structural elements necessary for AChE inhibition, quantitative 3D pharmacophore modeling was performed in silico using Discovery Studio v2.5 (DS, Accelrys, San Diego, CA) on 25 carbamate-type AChE inhibitors with diverse scaffolds.
Experimental and estimated activities, fit values, and corresponding error values of the training set molecules are listed in Table 1. Error is the ratio between estimated activity and experimental activity. Estimated activity is predicted based on fit value. Fit value indicates how exactly structural components in a molecule are localized in the center of pharmacophoric feature spheres, and thus represents how well a molecular conformation matches the pharmacophore model. All the training set compounds were predicted in their same order of magnitude except that compound 21 was underestimated with an error of þ10. The most active compound 15 (IC 50  The best pharmacophore model consists of five chemical features (HBA, RA, PI, and two HYD features) and two excluded volumes (Figure 2(a,b)). A pair of green spheres indicates one HBA feature with an arrow showing the direction of hydrogen bond. A pair of orange spheres indicates one RA feature with an arrow showing a normal to the aromatic plane. Red sphere stands for PI feature, and cyan ones are indicative of two HYD regions. Gray spheres represent two excluded volumes that unfavorable steric effects may occur. Spatial disposition of the model features was described in Figure 2(a) and distances between feature centers were labeled in Figure 2(b). Figure 2(c-f) illustrated the mapping of representative training set compounds to the pharmacophore model. Scheme 1. The process of AChE carbamoylation and decarbamoylation. a a EH is the free enzyme, AB is the inhibitor, EH Á Á Á AB is the non-covalent complex and EA is the carbamylated enzyme. k i is inhibit constant and k 3 is dissociation constant.
Features in the pharmacophore model were assigned different weights, which indicated varied importance. HBA (weight: 3.18) is inevitably the most important feature since carbamoyl group is the basis of covalent carbamylation. Thus phenol compound 25 (Figure 2(d)), which failed to fit the HBA feature, showed very low activity; whereas corresponding methylcarbamate 15 (Figure 2(c)) was the most active compound. PI and RA (weight: 1.99) were mapped to amino and phenyl groups, respectively. Protonated amine interacts with aromatic residues in the catalytic site via cation-p interactions, which stabilizes the transition state of inhibitorenzyme conjugate in the process of carbamylation. The distance between HBA and PI, namely O-N distance between the oxygen of carbonyl and the nitrogen of amine, plays an important role in the inhibition of AChE for carbamate-type compounds. As Figure 2(b) illustrated, the optimal distance between HBA and PI was 8.281 Å, indicating a range from 7.281 to 9.281 Å.
Unexpectedly, a small but important region of HYD (weight: 2.58) was identified in the model very near to the PI and RA features, which explained the activity difference between enantiomers. Figure 2(e) showed the alignment of physostigmine (1) to the pharmacophore model. Carbamoyl group and phenol ring overlapped to HBA and RA features, respectively. Nitrogen atom in 1-position instead of 8-position was aligned to the PI feature. (3aS)-methyl group in (À)-physostigmine (1) perfectly matched the HYD region (Figure 2(e)) and thus it shows high potency (IC 50 28 nM 22 ). In contrast, the R-enantiomer (þ)-physostigmine (2) is much less potent (IC 50 9900 nM 22 ). It demonstrated that configurational inversion of the two asymmetric centers (3a and 8a) caused crucial conformational changes which led to pharmacophore model mismatch. When (þ)-physostigmine (2) yielded to map into the HYD region as shown in Figure 2 Table 2. All the test set compounds were predicted in their same order of magnitude. Consequent correlation coefficient of 0.91 for the test set indicated good predictive power of the pharmacophore model.

Mapping of the pharmacophore model into AChE active site
To extend our knowledge of the pharmacophore model from ligand basis to its interactive target, all model features were fitted into AChE active site. A recent study on the X-ray crystal structure of a complex of (À)-bisnorcymserine (Figure 3(a)) and AChE (PDB code 3ZV7) 45 had revealed that a leaving group, (À)-bisnoreseroline (Figure 3(a)), was trapped in the catalytic site. The binding pose of (À)-bisnorphysostigmine (3) was supposed similar to that of (À)-bisnorcymserine since they had a common leaving group. Atom coordinates of the crystal AChE structure were fitted to those of the pharmacophore model through heavy atom superimposition of crystal (À)-bisnoreseroline structure upon pharmacophoric (À)-bisnorphysostigmine (3) conformation (Figure 3(b)).    excluded volumes in the model were aligned with two catalytic triad residues, Ser200 and His440, highlighted as yellow sticks in Figure 3(b,c). HBA was located nearby the hydroxyl of Ser200, wherein covalent bonds were to be formed via carbamoylation. RA feature formed hydrophobic interactions with the backbone of Gly118 and the side chain of His440. PI feature was surrounded by Phe330 and Trp84, also known as the anionic site. The HYD feature near PI showed week interactions with Gly117 and Tyr130. These results verified the reliability of the pharmacophore model in depth, and strengthened its predictive power for carbamatetype pseudo-irreversible AChE inhibitors.

Pharmacophore-based design and synthesis
As we early reported, (À)-meptazinol is a moderate AChE inhibitor (IC 50 41 mM 17 , Table 4) and it binds the enzyme by reversible mechanism. Guided by the pharmacophore model generated above, we selected (À)-meptazinol as the scaffold to build carbamoyl groups on. (À)-Meptazinol dimethylcarbamate (42) and phenylcarbamate (43) (Figure 1) were designed by carbamoylating the phenolic hydroxyl of (À)-meptazinol. They were supposed to be good AChE inhibitors for their perfect superimposition upon the model. Figure 4 showed the geometrical structural similarity between phenserine (Figure 4(a)) and (À)-meptazinol phenylcarbamate (43) (Figure 4(b)), and the spatial overlapping of phenserine ( Figure  4(c)) and 43 (Figure 4(d)) with the pharmacophore model. Although built on azapane scaffold, 43 matched four of the five features in the model just as phenserine did, especially at the very important PI (red) center and HYD (cyan) region. The only nitrogen in azapane ring of 43 resembled the N 1 -nitrogen of phenserine.
AChE inhibitory activities of 42 and 43 were predicted according to ligand pharmacophore mapping, giving an estimated IC 50 of 100 nM for 42 and 560 nM for 43 by BEST algorithm (Table 3). To our knowledge, conformational analysis of (À)-meptazinol derivatives is complicated due to the seven membered ring system. Our early NMR studies 46 on (À)-meptazinol hydrochloride had revealed that two stable conformers were detected in solution, including a lowest energy conformer with phenol group in equatorial orientation and a less favorable energy conformer with phenol group in axial orientation. A conformer in axial orientation was also found in the X-ray crystal structure of bis(9)-(À)-nor-meptazinol/AChE complex (PDB code 2W6C) 18 . It was therefore believed that pharmacophoric conformer of (À)-meptazinol might be the conformer in axial orientation with less favorable energy. To avoid energy minimization in the conformation generation step and to explore more ring conformations, CAESAR algorithm was performed to 42, 43 and (À)-meptazinol. As a result,  (À)-meptazinol was predicted an IC 50 of 18 mM, similar to the experimental activity. By CAESAR algorithm, the estimated IC 50 s of 42 and 43 were 75 nM and 370 nM, respectively (Table 3), and the conformation of 43 fitted to the pharmacophore model was in axial orientation (Figure 4(d)).
Methodology employed for the synthesis of 42 and 43 was illustrated in Scheme 2. 42 was synthesized in quantitative yield by treating (À)-meptazinol with N,N-dimethylcarbamoyl chloride in the presence of sodium hydride at room temperature. The coupling of (À)-meptazinol with phenyl isocyanate at room temperature in the presence of sodium gave 43 in 72% yield. 42 and 43 were prepared as hydrochloride salts for the following in vitro and in vivo assays. Structures of the hydrochloride salts were characterized by [a] D , IR, 1 H NMR, 13 C NMR, MS, and elemental analysis.

Cholinesterase inhibitory potency and selectivity
The synthesized carbamate compounds 42 and 43 were tested in vitro for AChE/BChE inhibitory potency and selectivity (Table  4). Mice brain homogenate and mice serum were used as sources of AChE and BChE, respectively. The dimethylcarbamate 42 inhibited AChE with IC 50 value of 6.93 nM, about 790 times lower than that of rivastigmine (IC 50 5460 nM). The phenylcarbamate 43, showing an IC 50 of 31.6 nM, was about 170 times more potent than rivastigmine and almost five times less potent than 42. Compared with the parent compound (À)-meptazinol, 42 and 43 showed a 5900-fold and 1300-fold increase, respectively, in the inhibition of mice brain AChE. With regard to activities reported by Yu et al. 22 , 42 was four times more potent than physostigmine (IC 50 27.9 nM), while 43 was 1.3 times less potent than phenserine (IC 50 24.0 nM).
As for selectivity, 42 was slightly more selective (twofold) to BChE similar to rivastigmine and physostigmine, while 43 was an AChE-selective inhibitor, showing a twofold selectivity for AChE versus BChE (IC 50 67.1 nM). Similar to phenserine, 43 would have less peripheral side effects and lower acute toxicity than those BChE-selective carbamates, such as physostigmine and 42.

Mechanism of enzyme inhibition and kinetic parameters
Understanding of potent AChE inhibitors' mechanism of action and kinetic parameters is key information to establish the structure-activity relationship and design new compounds for the treatment of AD. The characteristics of AChE activity inhibition by 42 and 43 were revealed by enzyme kinetics assays. The plots of residual enzyme activity versus enzyme concentration at different concentrations of 42 and 43 gave a family of straight lines with a y-axis intercept, suggesting that both compounds were reversible AChE inhibitors. Their enzyme inhibitory properties were further modeled using double-reciprocal plots. Variance of the velocity of control group could be explained as the degradation of enzyme. Increasing the concentrations of 42 and 43 led to a decrease in V max and an unvaried K m (x-intercepts) ( Table 5). 42 decreased the V max by 18% and 29% at the concentrations of 100 and 250 nM and 43 decreased the V max by 10% and 25% at the concentrations of 25 and 50 nM, consistent with the typical characteristics of uncompetitive inhibitors. The observed results showed that both inhibitors bound only to enzyme-substrate complex, not the free enzyme.
The inhibition of AChE by carbamates involves carbamoylation of the enzyme and production of a covalent adduct. The carbamoylated enzyme is then hydrolyzed to regenerate the free enzyme. The process is time-dependent, therefore, determination of the kinetic parameters is of utmost importance to assess time of action. Bartolini et al. 47 has reported that AChE immobilized disk, which could maintain enzymatic activity for about 2 months, was a powerful tool to evaluate both the carbamoylation and the decarbamoylation constants in single experiment. In this study, the well-known pseudo-irreversible AChE inhibitor physostigmine (1) was first selected as a reference compound to verify the reliability of our disk. Percent inhibition of enzyme activity [(A 0 ÀA i )/A 0 Â100%] was plotted versus time. The curve was fitted to Perola's mathematical equation. The calculated k i , k 3 , and K D of physostigmine were (4.78 ± 1.13) Â 10 5 M À1 min À1 , (1.94 ± 0.36) Â 10 À2 min À1 and (4.09 ± 0.22) Â 10 À8 M -1 , respectively, consistent  The data of the carbamoylation and decarbamoylation of AChE by 42 and 43 fitted well to Perola's equation 42 . Figure 5 showed that the immobilized AChE in EDA CIM disk was time-dependently inactivated by 42 and 43 at 50 nM. The carbamoylation half-times of 42 and 43 were found to be 23.5 min and 20.3 min, respectively, longer than that of physostigmine (3.9 min) and rivastigmine (11.4 min) reported by Bartolini et al. 39 . About 2 h and 3 h flushing were required to achieve a complete recovery of AChE activity after complete inhibition by 42 and 43, respectively, similar to physostigmine (2 h) but much shorter than rivastigmine (34 h). The k i , k 3 and K D values of 42 and 43 are shown in Table 6. 43 could bind to and dissociated from AChE faster than 42, indicating that the enzyme was more affinitive to 42 than 43. And, 42 inhibited AChE more strongly than 43. There was no obvious difference of the k i and k 3 values between 42 and 43. The enzyme affinity, carbamoylation and decarbamoylation rates and the duration of the inhibition of 42 and 43 were similar to that of physostigmine, suggesting that both the compounds reacted with the enzyme as a pseudo-irreversible inhibitor, in a way typical to carbamates, through quick formation of an addition complex and subsequent slow decarbamoylation.

Anti-amyloidogenic properties in SH-SY5Y-APP 695 cells
Anti-amyloidogenic properties of 42 and 43 were evaluated with HCS in SH-SY5Y-APP 695 cells 43 . After exposure to 50 mM of 42 and 43 for 16 h, intracellular APP levels were markedly reduced by 23.5% and 26.9%, respectively (Figure 6(a)). Phenserine was reported to produce a 40% decrease of APP level at the same concentration 51 . At a lower concentration of 5 mM, 43 exhibited a 20.0% decrease of APP levels (Figure 6(a)).
Although APP over-expression is a risk factor to AD, APP is still necessary to maintain normal physiological function. So, it will be more important to evaluate anti-amyloidogenic properties by determining Ab lowering effects, especially the neurotoxic form Ab 42 . Application of 50 mM of 42 and 43 to SH-SY5Y-APP 695 cells remarkably reduced the level of Ab 42 by 34.0% and 51.9%, respectively ( Figure 6(b)), keeping the level of less toxic Ab 40 unchanged (data not shown). The Ab lowering effect of 43 was better than that of phenserine (a 31% decrease of total Ab 14 ) at the concentration of 50 mM. Even at a lower concentration of 5 mM, 43 still produced a 30.5% decrease of the Ab 42 level (Figure 6(b)).
The actions of 42 and 43 on reducing APP and Ab 42 levels were very promising, but the mechanism was still complicated. Most AChE inhibitors produce Ab lowering effects by altering APP into non-amyloidogenic pathway 52 . This anti-amyloidogenic property results from post-receptor effects, such as Protein Kinase Ca (PKCa) activation, of the increased ACh level secondary to AChE inhibition. Exhibiting poorer AChE inhibition, 43 produced better APP and Ab 42 lowering properties in comparison with 42, which indicated additional non-cholinergic involvement in the anti-amyloidogenic effect of 43. Phenserine was reported to reduce the levels of APP and Ab via a non-cholinergic mechanism by downregulating the translation of APP mRNA14. Although 43 was less potent than phenserine in reducing APP level, its ability to reduce Ab, especially the most neurotoxic Ab 42 , was much higher than phenserine. It was possible yet still a hypothesis that 43 might have a direct action on the amyloidogenic processing pathway. Further experiments were still needed to clarify the mechanism.

Acute toxicity
The LD 50 values of 42 and 43 were tested in mice after intraperitoneal (i.p.) and oral (p.o.) administration, and corresponding results are reported in Table 4. As the doses of 42 and 43 escalating, peripheral cholinergic side effects such as salivation, twitch, and incontinence were observed. 42 showed high acute toxicity (LD 50 1.4 mg/kg) after i.p. administration. When administered orally, 42 (LD 50 12 mg/kg) was almost three times less toxic than physostigmine (LD 50 4.5 mg/kg) 54 , although 42 showed four times higher potency than physostigmine in in vitro test. The LD 50 of 43 (73 mg/kg, p.o.) was 12-24 times higher than that of rivastigmine (3-6 mg/kg, p.o.) 53 . If administered intraperitoneally, 43 (LD 50 45 mg/kg) was slightly less toxic compared with phenserine (25 mg/kg) 55 . Thus, 43 showed low acute toxicity and deserved further studies in cholinergic impairment animal models to evaluate its in vivo cognitive enhancement function.

Conclusions
In summary, (À)-meptazinol carbamate derivatives were designed based on a 3D pharmacophore model built using 3D QSAR Pharmacophore Generation module in DS from 25 known carbamate-type AChE inhibitors. The best pharmacophore model consists of five chemical features (namely HBA, RA, PI, and two HYDs) and two excluded volumes. The existence of a HYD region near the PI feature have been recognized as essential chemical characteristics in the model to differentiate enantiomers. Merging of carbamoyl groups onto the (À)-meptazinol scaffold generated new bifunctional ligands with dual actions on both cholinesterase and amyloidogenic pathways.
The synthesized compounds 42 and 43 were verified as nanomolar cholinesterase inhibitors in in vitro assay. 42 and 43 showed uncompetitive inhibition and reacted with the enzyme as a pseudo-irreversible inhibitor, such as typical carbamates, through quick carbamoylation and subsequent slow decarbamoylation. 42 (IC 50 6.93 nM) was more potent in inhibiting AChE than 43, and was slightly selective to BChE (two-fold). In acute toxicity test, 42 had lower LD 50 values (12 mg/kg, p.o.) and showed more peripheral cholinergic side effects. However, the phenylcarbamate 43 was more promising and exhibited significant anti-cholinesterase and anti-amyloidogenic properties.
43 (IC 50 31.6 nM) was 170 times more potent than rivastigmine in inhibiting AChE, and was 1.3 times less potent than phenserine. 43 exhibited a twofold selectivity for AChE, therefore milder peripheral side effects and lower acute toxicity were observed for 43 (LD 50 73 mg/kg, p.o.). 43 also showed Ab lowering effects (51.9% decrease of Ab 42 ) superior to phenserine (31% decrease of total Ab) at the concentration of 50 mM. Even at a lower concentration of 5 mM, 43 still reduced APP level by 20.0% and reduced Ab 42 by 30.5%. The dual actions of cholinesterase inhibition and anti-amyloidogenesis indicated a potential use of 43 as symptomatic and disease-modifying agent for the treatment of AD, which deserved further studies in cholinergic impairment animal models.