current medicinal chemistry molecular modelling and qsar ... · qsar studies in the design of...

Current Medicinal Chemistry, 2000, 7, 273-294 273

Molecular Modelling and QSAR of Reversible Acetylcholines-terase Inhibitors

J. Kaur* and M.-Q. Zhang

Department of Medicinal Chemistry, Organon Laboratories Ltd, Newhouse,Lanarkshire ML1 5SH, Scotland, United Kingdom

Abstract: Acetylcholinesterase (AChE) inhibitors are an important class ofmedicinal agents useful for the treatment of Alzheimer’s disease, glaucoma,myasthenia gravis and for the recovery of neuromuscular block in surgery. Torationalize the structural requirements of AChE inhibitors we attempt toderive a coherent AChE-inhibitor recognition pattern based on literature dataof molecular modelling and quantitative structure-activity relationship (QSAR)analyses. These data are summarised from nearly all therapeutically important chemical classesof reversible AChE inhibitors, e.g., derivatives of physostigmine, tacrine, donepezil andhuperzine A. Interactions observed from X-ray crystallography between these inhibitors andAChE have also been incorporated and compared with modelling and QSAR results. It isconcluded that hydrophobicity and the presence of an ionizable nitrogen are the pre-requisitesfor the inhibitors to interact with AChE. However the mode of interaction i.e., the 3-dimensional(3D) positioning of the inhibitor in the active site of the enzyme varies among different chemicalclasses. It is also recognised that water molecules play crucial roles in defining these different 3Dpositioning. The information on AChE-inhibitor interactions provided should be useful for futurediscovery of new chemical classes of AChE inhibitors, especially from De Novo design andhybrid construction.

Introduction The medicinal use of AChE inhibitors has beenknown even longer than the enzyme itself. Already inthe mid-nineteenth century, physostigmine wasrecognised to cause constriction of the pupil andcontraction of the ciliary muscle, thus improving thedrainage of aqueous humour [4]. Today, the mainclinical uses of AChE inhibitors are in anaesthesia [5], inthe treatment of myasthenia gravis [6], glaucoma [7]and Alzheimer’s disease [8]. Table 1 lists a few morerecent examples of AChE inhibitors and their clinicalindications. As more physiological functions of ACh arebeing elucidated, there should arise more therapeuticopportunities for AChE inhibitors [9].

Acetylcholinesterase (EC 3.1.1.7, AChE) isprobably one of the earliest enzymes to have beenstudied for its pharmacological functions. As early as1914, Dale had already suggested that physostigmine(1 ), a naturally occurring alkaloid from the Calabar bean,inhibited an enzyme that catalysed the breakdown ofcholine esters [1]. Indeed, AChE’s vital function is thehydrolytic destruction of the neurotransmitteracetylcholine (ACh), which terminates the impulsetransmission at cholinergic synapses. Throughactivation of either the ionotropic nicotinic receptors orthe metabotropic muscarinic receptors, ACh exertsmany physiological functions both in the periphery andthe central nervous system (CNS) [2, 3], e.g., smoothmuscle contraction, modulation of cardiac rate andforce, motor control, temperature regulation, memoryand pain modulation, etc. Inhibition of AChE results inthe cumulation of ACh and enhanced cholinergictransmission, and has long been an attractive target fordrug development.

The purpose of this paper is to review literature dataon molecular interactions between AChE and itsinhibitors, derived from molecular modelling andquantitative structure-activity relationship (QSAR)analyses. Since irreversible AChE inhibitors aretherapeutically less useful, reports on these inhibitorsare not included in this review. Interested readers arereferred to other sources [10-12]. Attempts will bemade to compare and rationalize modelling and QSARresults with those observed from X-ray crystallographyso to derive a coherent AChE-inhibitor recognitionpattern. This hopefully would be useful for the futuredesign of new AChE inhibitors.

*Address correspondence to this author at the Department of MedicinalChemistry, Organon Laboratories Ltd, Newhouse, Lanarkshire ML1 5SH,Scotland, United Kingdom

0929-8673/00 $19.00+.00 © 2000 Bentham Science Publishers B.V.

274 Current Medicinal Chemistry, 2000, Vol. 7, No. 3 Kaur and Zhang

Table 1. Clinical Uses of AChE Inhibitors

drug structure clinical indication

Physostigmine (1)

H3C

HN

O

O

N

NCH3

H3C

CH3

glaucoma

neostigmine (2) H3CN

O

O

CH3

N+

CH3

CH3H3C

reversal of neuromuscular block; myasthenia gravis

tacrine (3)

N

NH2

Alzheimer’s disease

donepezil (4)

NOMeO

MeO

Alzheimer’s disease

huperzine A (5)

NH

H3C

CH3

OH2N

H myasthenia gravis; Alzheimer’s disease

Structure and Catalytic Mechanism ofAChE

three tetramers of up to twelve catalytic subunits linkedto a collagen tail through disulphide bridges. A secondclass, commonly called the globular forms (Gn), exists asmonomeric (G1), dimeric (G2) and tetrameric (G4)assemblies of catalytic subunits. Different forms ofAChE exhibit tissue-specific distribution, e.g.,asymmetric forms are concentrated in neuromuscularjunctions of mammalian skeletal muscle and in theelectric organs of rays and eels whereas the majorAChE forms in the CNS are the amphiphilic globulartetramer (G4) [16].

AChE is bound to the basement membrane in thesynaptic cleft at cholinergic synapses. The soluble formof AChE is also present in cholinergic nerve terminals,where it seems to have a role in regulating the free AChconcentration, and from which it may be secreted, thefunction of which is so far unclear [13]. It is important tonote that there is a distinct type of solublecholinesterase, butyrylcholinesterase (EC 3.1.1.8,BChE), which has a closely related molecular structure(~53% sequence homology) to AChE but differentdistribution, substrate specificity and functions. BChEalso hydrolyses ACh but unlike AChE it has a muchbroader substrate specificity and broader distribution.The precise physiological function of BChE is still asubject of debate [14].

It is theoretically possible to selectively inhibit oneparticular form of AChE, as so claimed for the action ofSDZ-ENA 713 (6 ) which inhibited preferentially the G4

NO

O

N

6

Two classes of different quaternary structures ofAChE exist in synaptic junctional and extrajunctionalareas, which are distinguishable by their solubilitycharacteristics and hydrodynamic properties [15]. Oneclass exists as asymmetric forms (An) constituted by

Molecular Modelling and QSAR Current Medicinal Chemistry, 2000, Vol. 7, No. 3 275

Table 2. AChE Inhibition ( IC50, µM) by of N-[ω-N’-(Adamant-1’-yl)aminoalkyl]-2 (4’-Dimethylaminophenyl)acetamides (7)

O

HN

NHN7a: n = 77b: n = 97c: n = 11

( )n

AChE source Molecular form 7a 7b 7c Physosti-gmine (1) Donepezil (4)

Torpedo californica Dimer, H2O soluble 10 5.3 3.3 0.046 0.012

Electric eel Tetramer, H2O soluble 28 14 8.9 0.046 0.017

Bovine erythrocytes Oligomers (2-8), membrane bound >100 95 19 0.123 0.008

Bovine brain Tetramer, membrane bound >100 77 20 0.146 0.006

Human erythrocytes Tetramer, membrane bound >100 >100 >100 0.047 0.008

form of AChE [17]. In the brain of Alzheimer patientsthere is a strong reduction of G4-AChE activity but littledecrease of G1–AChE activity as compared to thehealthy brain [18]. By selectively inhibiting G4-AChE,SDZ-ENA 713 was expected to have less potential tocause side effects. Other evidence for possibleselective inhibition of AChE isoforms comes fromobservation that a series of N-[ω -N’-(adamant-1’-yl)aminoalkyl]-2-(4’-dimethylaminophenyl)acetamides(7 ) exhibited significant selectivity towards differentforms of AChE [19] (Table 2).

The role of aromatic residues in the gorge may be tofacilitate diffusion of the substrate to the active centre(aromatic guidance) [20]. The aromaticity may alsopreclude the necessity of displacement of slow-exchanging water molecules at the base of the cleftupon ligand binding and hence it could simply play apassive role. What is definite is that these aromaticresidues play important roles in binding of inhibitors tothe active site, as evidenced by the influence ofphysico-chemical properties of the inhibitors, e.g.,hydrophobicity, electronic effects, etc., to theirinhibitory potency on the enzyme.

Although AChE is polymorphic, activities of catalyticsubunits in the asymmetric and globular forms of theenzyme are similar. In 1991, Sussman et al crystallisedthe glycolipid-anchored homodimer of AChE fromTorpedo californica electric organ and reported itsthree-dimensional structure (at 2.8 Å resolution) [20].The crystal structure established that the subunitscontain 12-stranded mixed β sheet surrounded by 14α helices. They are ellipsoid in shape (45x60x65 Å)and associate as dimers in a four-helix bundle. Atetramer of Electrophorus electricus AChE has alsobeen crystallised [21]. A low resolution structurerevealed a subunit arrangement of a dimer of dimers.

Quantitative and Qualitative SAR ofAChE Inhibitors

In the absence of structural details ofacetylcholinesterase, traditional quantitative structure-activity relationships (QSAR) have been derived to tryto explain variations in biological activities of a givenseries of acetylcholinesterase inhibitors in terms ofphysicochemical, structural, and conformationalproperties of the molecules. Physicochemicalproperties such as the steric, hydrophobic andelectronic features of molecule which have beenshown to be important for enzyme-ligand interactionshave been extensively employed in QSAR analysis. Avariety of descriptors are available to describe thesephysicochemical properties of molecules. An additionalkey interaction for AChE inhibition identified fromQSAR studies in the design of pesticides andinsecticides is a charge transfer interaction betweenprotonated nitrogen and the AChE. QSAR approachescan be employed to guide the design of further potentand biologically attractive AChE inhibitors.

The active site of AChE is a narrow gorge 20 Å indepth and lined up with 14 aromatic residues [20]. Atthe base of the gorge lies the catalytic triad of Glu327,His440 and Ser200 involving the proton shift fromserine OH to the carboxyl of glutamate via the imidazoleof histidine. Formation of the acyl enzyme proceedsthrough formation of a tetrahedral intermediate whichrelaxes back to the trigonal acyl enzyme. Deacylationalso proceeds through a tetrahedral intermediate byattack of the acetyl-AChE bond from an internal waterwhich may be rendered more nucleophilic by aneighbouring carboxyl or imidazole residue.

In this section, we describe the QSAR models andqualitative SAR data for different chemical series of


AChE inhibitors. It has been shown that AChE fromdifferent species may bear differences andconsequently, this may affect the structure-activityrelationships derived. On this basis, we have restrictedour review to QSAR models generated from speciesknown to exhibit high homology for the active siteresidues of AChE. For complete details of the range ofsubstituents/analogues and their activities employed inthe QSAR models presented, the reader is referred tothe original publications.

this QSAR model is that the different R groups forcethe molecule to adopt different modes of interactionwith AChE. The QSAR model does not provide anyinformation about the nature of the interaction betweenthis series of tacrine analogues and AChE. The originalauthors [22] who synthesized this set of 11 N-monosubstituted 1,2,3,4-tetrahydro-9-aminoacridines(Dataset 1) also reached similar qualitative relationshipbetween the length of the R group and AChE activity.However, they concluded that given the comparativelysmall variations (6-fold) in enzyme inhibition for all thevarious alkyl or arylalkyl substituents, that it may bepossible these substituents project away from theprotein having little or no effect in binding.

Tacrine

Tacrine (3 ) (Cognex®) was approved as a drug forthe treatment of Alzheimer’s disease by the FDA in1993. Despite its wide usage, Tacrine (THA) suffersfrom dose limiting hepatotoxic complications, slowpharmacokinetics, and a high incidence of side effects.Although large numbers of analogues have beensynthesized around tacrine (Fig. (1 ), R=H, Ki = 0.05µM,electric eel AChE [22]) in an attempt to identify ananalogue with a more favourable in vivo profile, onlylimited SAR within this series has been reported [22].

Further analogues of tacrine were synthesized [22].Analogues in which both amino hydrogens werereplaced resulted in a less favourable AChE activity ascompared to the monosubstituted analogues. This wasattributed to disubstituted analogues preventing thefavourable planar-like conformation of the ring systemof tacrine being adopted for interaction with theenzyme. Complete aromatization to 9-aminoacridinecreates a fully planar system that is found not to affectbinding. Similarly, condensation or enlargement of theunsaturated ring or addition of pendant groups ontothe aromatic ring are found not to markedly reduce orimprove AChE activity. These results suggest a stronghydrophobic interaction between tacrine analoguesand AChE . In addition, a basic nitrogen is alsosuggested to be important for AChE affinity, althoughthis is not reflected in the QSAR model.

N

NHR1

2

3

45

6

7

8 9

R = H, Me, (CH2)nCH3 n=1-5, (CH2)nR’ [n=1-2 R’=Ph,NH2,CH=CH2]

A series of analogues of 1-hydroxy-tacrineanalogues (Fig. (2 )) were synthesized (Dataset 2) [24].Velnacrine (R=R1=H, IC50= 4µM, Rat Striatum AChE[24]) 1-hydroxy-tacrine, was identified throughmodification of tacrine and is currently in Phase IIIclinical trials [25].

Fig. (1) . N-Monosubstituted 1,2,3,4-Tetrahydro-9-aminoacridines (Dataset 1).

A QSAR model for substituted amino groups on the9 position of tacrine (Dataset 1) was generated in whicha bilinear relationship between the Verloop lengthparameter L with biological activity is observed [23].

N

NHR OH

R1

Log 1/Ki = -0.296(0.136)L +0.569(0.220)log(β x 10 L + 1) + 7.921(0.539) (1 )

n =11 r2 = 0.844 s = 0.142 F3,7 = 12.59 Lopt = 4.903 Ålog β = -4.867

R = H, CH3, C3H7, CH2C6H4R’ [R’ =Me, F,CF3 etc]

Ki range: 0.038µM – 0.22µM (Electric eel AChE) R1 = H, Cl, F, OCH3, CF3

L = Verloop length parameter for R substituent Fig. (2). 9-Amino-1,2,3,4-tetrahydroacridin-1-ol Analogues(Dataset 2).

This unusual bilinear equation (Equation (1)) leadsto Lopt which corresponds to a minimum of activity,rather than maximum in activity. The derived QSARmodel represents an inverse bilinear model whosesignificance and interpretation could not be readilyexplained. However, it was concluded that the QSARmodel suggests that compounds with extreme L valuesare important for activity. One possible interpretation of

The QSAR model generated for the 1-hydroxy-tacrine analogues, in which substituents have beenintroduced into 6 position of the tetrahydroacridine, aswell as at position 9, is given by Equation (2).

Log 1/IC50 = -0.560(1.32)L +0.877(0.263)log(β x 10 L + 1) +1.509(0.386)I-Cl + 7.411(0.528) (2 )


n =30 r2 = 0.861 s = 0.305 F4,25 = 38.83 Lopt = 5.116 Ålog β = -4.869

analogues of physostigmine have been synthesizedand analysized for SAR.

IC50 range: 0.0012µM – 69 µM (Rat Striatum AChE)The first series of analogues of physostigmine (Fig.

(3 )) were synthesized in order to probe the relationshipbetween the N1 substituent and AChE activity [26].

L = Verloop length parameter for R; I-Cl = 1 (R1 = Cl) or 0

(R1 ≠ Cl)

As with the previous QSAR model for tacrineanalogues (Dataset 1), the correlation equationcontains an inverse bilinear term in L. A minimum AChEactivity is observed for substituents R with a length L of5.116 Å. As seen in Equation (2), the indicator variableI-Cl takes a value of 1 when a 6-Cl substituent is presentand 0 when 6-Cl is absent. The presence of the 6-Clsubstituent at R1 is shown to be favourable for highAChE activity. Since a limited number of R1

substituents were considered in the generation of theQSAR model, it is not known whether the 6-Clsubstituent forms a favourable hydrophobic orelectronic interaction with AChE. The authors whosynthesized these 1-hydroxy-tacrine analogues havedrawn similar conclusions about the nature of theinteraction of these analogues and AChE enzyme [24].

A QSAR analysis was performed for this series andEquation (3) was derived [23].

Log 1/IC50 = -1.146 (0.445)log P’ + 7.721 (0.568) (3 )

n = 6 r2 = 0.927 s = 0.275 F1,4 = 51.07

IC50 range: 0.056µM–9.3µM (human erythrocyteAChE)

Log P’ = calculated log P for the ionized compounds.

In this case, the log P’ was found to show an inverselinear correlation with AChE activity, therebysuggesting that less hydrophobic compounds arefavourable for AChE activity. The original authors [27]who synthesized this series of analogues havesuggested that a bulky R1 group reduces activity byinterfering with the binding with the anionic site. Itshould be noted that for this small dataset, onlyhydrophobic substituents have been examined. As willbe observed with other QSARs for AChE inhibitors, logP’ and the shape of R1 may be highly correlated. Inorder to determine whether bulkiness and/orhydrophobicity is important, large polar and largehydrophobic substituents would have to besynthesized and tested. For the only polar substituentR1 = CONHCH3, inactivity (IC50=3.9µM) may result fromthe lack of an ionizable nitrogen within this compound.

For both Datasets 1 and 2, QSAR models aregenerated for which an inverse bilinear relationshipbetween the Verloop length parameter L of R withAChE activity is observed. This suggests thatanalogues with short R substituents bind favourable toAChE, but as the length of the substituent increases to≈ 5 Å the interaction becomes unfavourable and thenfor longer substituents, the interaction with AChEbecomes favourable. Secondly, it was found that log Pand L are highly correlated. This complicates theinterpretation of the QSAR models for tacrineanalogues since it could be that the hydrophobicnature of R is important for activity. A QSAR model has been determined for analogues

of physostigmine in which carbamoyl substituents havebeen modified (Fig. (4 )) [27].

PhysostigmineHN

O

O

N

NR1

H3C

CH3

X

6

5

4

3

2

Physostigmine (1 ) (Fig. (3 ), R1 = CH3) is a naturalproduct which acts as a potent reversible AChEinhibitor. It has been shown to suffer from a short halflife, variable bioavailability, and a narrow therapeuticindex which accounts for its inconsistent clinicalefficacy. In order to improve the in vivo profile ofphysostigmine whilst retaining its in vitro potency (IC50=0.058µM, human erthrocyte AChE [26]), a number of

X = H, CH3, Et, i-C3H7 with various combinations of mono and

disubstitution patterns at the ortho and para positions

R1 = H or CH3

H3C

HN

O

O

N

NR1

H3C

CH3

R1 = H, CH3, CH2R’ [R’=CH=CH2,Ph, CH2Ph], CONHCH3

Fig. (3). N1-substituted Analogues of Physostigmine(Dataset 3).

Fig. (4) . Dataset 4 – Phenserine Analogues.

The R1 substituent is either CH3 or H and thesehave been shown in the previous series of N1-substituted analogues of physostigmine (Dataset 3) tobe favourable for AChE activity. Therefore onlyparameters associated with X have been identified asinfluencing activity (Equation (4)).


Log 1/IC50 =1.251 (0.427)Es6 - 0.984(0.392)π4 + Log 1/IC50 = 2.367(0.346)L - 0.066(0.011)L2 -7.905(0.285) (4 ) 13.924(2.638) (5 )

n = 14 r2 = 0.848 s = 0.384 F2,11 = 30.67 n = 9 r2 = 0.995 s = 0.086 F2,6 = 567.82 Lopt = 18.02 Å

IC50 range: 0.009µM – 1.48µM (human erythrocyteAChE)

IC50 range: 0.03µM – 20 µM (human erythrocyteAChE)

Es6 - Tafts steric parameter for X-substituents in 6position; π4 - hydrophobic parameter for X-substituentsin 4 position.

L = Verloop length parameter for morpholinoalkyl

The Verloop L parameter for the morpholinoalkylsubstituent has been shown to be highly correlatedwith log P thus complicating the interpretation of theQSAR. The QSAR model as represented by Equation(5) accounts for only analogues where n ≥ 4 and thoseanalogues with two ionizable centres. A second QSARmodel was generated in which the calculated logP’shows a bilinear relationship with activity for allcompounds (n=13) of the series (Fig.s (5a) and (5b )).

This QSAR model suggests that orthodisubstitution in this series is unfavourable for activity.The second term in Equation (4) suggests that polar-like substituents at the para position are favoured.Once again the authors [23] highlight the highcorrelation between the descriptors for hydrophobicityand shape for both the 4 and 6 positions.

Physostigmine was then modified [28] toincorporate a second ionizable N via a morpholinemoiety (Fig. (5 )). The authors’ rationale for synthesis ofsuch analogues was that at physiological pH, themorpholino nitrogen will be protonated and can interactwith putative second anionic site on AChE. It hasalready been demonstrated that some of the mosteffective inhibitors are the bis-quaternary compounds[29] such as decathonium.

Log 1/IC50 =0.717 (0.287)log P’ -1.158(0.689)log (βP’+1) + 7.181(0.468) (6 )

n = 13 r2 = 0.772 s = 0.497 F3,11 = 11.44 log P’opt =1.003 log β = -0.792

IC50 range: 0.03µM – 7.9 µM (human erythrocyteAChE)

logP’ = calculated log P for the di-ionized compound

HN

O

O

N

NCH3

H3C

CH3

( )nN

O This result suggests that hydrophobicity of theprotonated species may play an important role in theinteraction with AChE. Given the bilinear nature of theQSAR model, it would appear that there may be twopossible different modes of interaction with theenzyme. At short lengths of n with the exception ofn=2, the presence of a charged species isunfavourable for AChE activity whereas at large valuesof n (n >7), a protonated morpholino moiety isfavourably positioned for interaction with the proposedsecond anionic site on enzyme. The formerobservation is further confirmed by the high AChEaffinity of both physostigmine (Fig. (5b ), R=R1=CH3)and heptylphysostigmine (Fig. (5b ), R=C7H15 R1=CH3),which lack a second protonated nitrogen and arecomparable in length to analogues in Dataset 5 wheren< 7.

n = 2 to 12

Fig. (5a) . MorpholinoalkylcarbamoyloxyeserolineAnalogues.

HN

O

O

N

NCH3

H3C

CH3

R

HN

O

O

NR1

R2

R

R = CH3 Physostigmine; IC50=0.031µM ; R = C7H15Heptylphysostigmine; IC50=0.32µM [28]

Fig. (5b) . Heptylphysostigmine and Physostigmine.

A QSAR model for these analogues (Dataset 5)shows a correlation between the Verloop parameter Lfor the morpholinoalkyl substituent and biologicalactivity [23].

R =CH3, (CH2)nCH3 [n=2,3,5,6], Ph, CH2Ph

R1 = CH3, Et, nPr, CH2Ph

R2

= H or CH3

Fig. (6) . Dataset 6 - 8-carba-analogues of physostigmine.


A series of 8-carba-analogues [30] of physostigminecarrying different substituents on both N atoms andvarying the R2 group on the eseroline ring to be eitherH or CH3 were considered for QSAR analysis (Dataset6). Such analogues were synthesized (Fig. (6 )) on thebasis that they should exhibit good metabolic stabilityas compared to physostigmine.

synthesized to explore the SAR of the X substituent(Fig. (7 ))

N

O

OX

( )n N

( )m

Y

Log 1/IC50 = -1.988 (0.469)B1(R) + 0.805(0.30)I-Me +

X = H, NO2, CH3, Cl, OR’ [R’=H,Me,SO2Me], NHCOR’’,COR’’’9.292(0.889) (7 )

Y=H, n=5 and m=1n=9 r2=0.792 s = 0.308 F2,18 = 34.37

Fig. (7). The 2-ω[N-ethyl-N(ω-phenylalkyl)amino]-1H-isoindole-1,3(2H)-diones derivatives (Dataset 7).

IC50 range: 0.045µM – 1.35µM (human erythrocyteAChE)

A QSAR model was developed for this series of 15analogues [23].

B1(R) = Verloop’s bulkiness parameter for Rsubstituent; I-Me =1 (R2=Me) and 0 (R2=H).

Log 1/IC50 = -0.243 (0.131)logP + 1.594(0.57)F +The B1 parameter for R suggests that linear chainsubstituents are favoured and that branchedsubstituents decrease activity. The indicator variable I-Me at R2 suggests that the methyl group is favourablefor activity. Since only R2=CH3 or H were present withinthis series, it would appear that the favourable activity ofthe methyl group may result from a hydrophobicinteraction with the enzyme. Unlike the QSAR modelderived for Dataset 3 of physostigmine analogues(Equation (3)), the R1 substituent within this series isnot important for activity. Closer examination of the R1

substituents (R1 =Me, Et, nPr) present within this seriesreveals that they are all favourable for activity with theexception of the only bulky R1 substituent (R1 =CH2Ph)for which the observed AChE activity is unfavourable.

6.782(0.642) (8 )

n = 15 r2 = 0.821 s = 0.161 F2,12 = 24.49

IC50 range: 0.15µM – 3.4µM (Rat brain AChE)

Log P – calculated hydrophobicity of the molecule; F -electronic parameter for substituent X.

The QSAR suggests that hydrophilic and electronwithdrawing substituents on the phthalimide moiety arefavoured for enzyme activity in agreement with theQSAR model derived by Ishihara [31]. The most activecompound being X=NO2 with an IC50=0.15µM. Therequirement of electron withdrawing substituents wasnot understood but assumed to be required for anelectrostatic interaction with the enzyme. Severalparameters such as the size of X substituent do notappear to be important for activity i.e., bulky groupssuch as X=NHCOCH3 can be accommodated on thearomatic phthalimide ring without loss of activity. In thisseries, only substituents at the 5 position have beensynthesized and tested for AChE inhibition andtherefore the effect of substituents at other positionson the phthalimide moiety remain unknown.Replacement of the benzene group of the phthalimidewith cyclohexane leads to a 7-fold drop in activity whichcannot be accounted for using Equation (8). It hasbeen postulated that the cyclohexane may be stericallyunfavourable for interaction with AChE.

The QSAR models and qualitative SAR datasuggests that physostigmine and carba-8-analogues ofphysostigmine adopt a similar interaction with theenzyme. Within Dataset 6, it is found that n-Pr as a N1substituent is tolerated, extrapolation to Dataset 3would lead to the conclusion that it is the bulkiness ofR1 which is important for AChE activity and not the log Pof the compounds. In conclusion, it would appear thatsteric and/or hydrophobicity of physostigmineanalogues are important for AChE activity. It should benoted that a basic nitrogen within these analogues isalso important for enzyme inhibition although this is notreflected in the QSAR models (Equations (3)-(7)).

Benzylamines Further analogues of 2-ω [N-ethyl-N(ω -phenylalkyl)amino]-1H-isoindole-1,3(2H)-diones havebeen synthesized and tested for AChE inhibition (Fig.(8 )) [31]. The effects of substitution on the phenyl ringof the benzylamino moiety were investigated and areshown in Fig. (8 (I)). The most active Y substituent wasfound to be the 2-OMe (IC50=0.024µM). No solidconclusions could be drawn about the relationship

Ishihara et al. [31] identified a class of benzylamines,2-ω [N-ethyl-N(ω -phenylalkyl)amino]-1H-isoindole-1,3(2H)-diones as potent acetylcholinesteraseinhibitors. Within this class of AChE inhibitors, anumber of analogues (Dataset 7) have been


N

O

OO2N

( )5 NY N

O

OO2N

( )5 N

N

O

OO2N

( )5 N

( )mN

O

O

( )5 N

H3C O

O2N

I ll

lll lV

Fig. (8) . Analogues synthesized to investigate the SAR of the 2-ω[N-ethyl-N(ω-phenylalkyl)amino]-1H-isoindole-1,3(2H)-dioneclass of AChE inhibitors.

between Y and AChE inhibition given the small numberof analogues synthesized within this series (n=8).Replacement of the phenyl ring of the benzylaminomoiety with a cyclohexane ring (Fig. (8 (II))) lead to a 5-fold drop in the AChE inhibitory potency. This resultsuggests that a hydrophobic group at this position inthe molecule is favourable. The lower affinity of thecyclohexyl derivative may result from the lack of π-πinteraction with the enzyme.

On the basis of the hypothesized nature of theinteraction of the above series with AChE, Ishihara et al.[32] proceeded to synthesize a new series in which thephthalimide moiety was replaced with 3-arylpropenamide (Fig. (9 )).

A number of analogues (Dataset 8) in which the Xand R substituents are varied have been analysedquantitatively using the Hansch-Leo approach(Equation 9).

The effect of the chain length of the N-(ω -phenylalkyl)amino moiety as shown in Fig. (8 (III)) hasalso been investigated. The optimal spacer length forAChE activity when m=1 (IC50=0.15µM). Increasing thechain length m to 2 or 4 leads to a drop in AChE affinity(IC50=0.77µM and 1.75µM, respectively). An unusualobservation is that in the analogue for which m=3, anaffinity (IC50=0.16µM) is observed that is comparable tothe analogue in which m=1 (IC50=0.15µM). Oneexplanation for this observation is that the propyl chainmay adopt a conformation that allows good interactionwith the enzyme.

Log 1/IC50 = -0.174 (0.149)π + 0.611(0.249)σ +0.734(0.236)I + 4.991(0.137) (9 )

n = 25 r = 0.919 s = 0.245 F3,21 = 37.93

IC50 range: 0.16µ – 21µM (Rat brain AChE)

π - hydrophobic parameter for X, σ - electronicparameter for X, I = 1 (R= CH3CO) and 0 (R=H)

Equation (9) suggests hydrophilic and electron-withdrawing groups on the aryl moiety and that thepresence of the acetyl group for R is favoured foractivity. The same conclusions could be drawn from theQSAR model generated by Recanatini et al. [23] for thisdataset , the only difference being that logP for thewhole molecule was used as a descriptor forhydrophobicity. Equation (9) is similar to Equation (8)thereby suggesting that the X substituents in thesetwo series of benzylamines interact in a similar mannerwith AChE. The favoured acetyl group for R within thisseries was attributed to its ability to hydrogen bond toan active site residue of the enzyme. Other analogueswere also synthesized within this series to explore itsSAR. It was found that these analogues exhibit similarSAR trends to those observed with analogues of thepthalimide-benzylamines (Fig. (7 )) e.g the ortho OMesubstituent on the phenyl ring of the benzylaminomoiety is favoured for AChE inhibition in both series.

Finally the replacement of the N-ethyl group with N-acetyl (Fig. (8 (IV)) leads to a significantly loss in AChEactivity (IC50>100µM), thereby highlighting theimportance of the basic nitrogen for interaction withAChE enzyme.

N

O

( )5N

RX

X = NO2, NH2, OH, Cl , OMe, SnMe [n=0-2]; meta and/or para

substituted

R = CH3CO or H

Fig. (9) . The [ω[N-ethyl-N-(phenylmethyl)amino]alkyl]-3-Arylpropenamides Series of AChE inhibitors (Dataset 8).


Y

X

O

( )5 NX Y

O

X

NY

l

ll

lll

O

CH2

( )n

N

Fig. (10) . Other members of the benzylamine class synthesized as potential AChE inhibitors.

Further series of compounds were synthesized inwhich the phthalimide moiety in Fig. (7 ) was replacedby benzoyl (I), indanone (II) and cyclohexylphenylketone (III) moieties [33] as shown in Fig. (1 0 ). Thecyclohexylphenyl ketone analogues were found to beinactive and this was attributed to their inability to adoptthe bioactive conformation. Both the indanone andbenzoyl analogues were found to exhibit activity forAChE. From limited analogues for both these classesof compounds, it was found that they show similar SARtrends to phthalimide series. Furthermore, it was shownthat the carbonyl group of the indanone/benzoyl orphthalimide moiety was separately approximately by a 7carbon chain length from the basic nitrogen. It waspostulated that the benzylamine moiety whether linkedto a phthalimide, indanone or benzoyl moiety all adoptthe same mode of binding with the AChE. Noconformational studies have been reported to confirmthat all these classes of benzylamines can adopt similarlow energy conformations.

• The distance between the carbonyl and basicnitrogen corresponds to approximately 7 carbonchain length.

Benzylpiperdines

Sugimoto et al. [34] identified a benzylpiperidine, 1-benzyl-4-[2-N-benzoyl (amino)ethyl]piperdine throughrandom screening as a high affinity AChE inhibitor(X=H, IC50=0.56µM, Mouse brain AChE). A number ofanalogues were synthesized (Fig. (1 1 )); of which a setof analogues were synthesized (Dataset 9) to examinethe effect of substitution on the benzamide moiety[23]. In this dataset of analogues, m=1 and Y=H.

HN

O

X

( )2

N( )m Y

For the benzylamine class of AChE inhibitors, amode of interaction with AChE was hypothesized:

X =ortho,meta and/or para substitutedX = Me, NO2, OMe, CHO, Hal, COMe, SO2CH2Ph

m=1, Y=H• The benzylamino moiety binds to anionic site (via

basic nitrogen) and a hydrophobic site (close toanionic and esteratic sites)

Fig. (11) . 1-benzyl-4-[2-N-benzoyl (amino)ethyl]piperdinederivatives (Dataset 9).

The QSAR equation is:• The phthalimide/indanone/benzoyl moiety bindsto a second distal hydrophobic site

Log 1/IC50 = 0.725(0.479)F + 0.464(0.296)I-P -• Electron withdrawing and hydrophilic

substituents on the phthalimide/indanone/benzoyl moiety forms an electrostatic interactionwith AChE.

0.624(0.337)I-O + 6.354(0.291) (1 0 )

n = 16 r2 = 0.845 s = 0.224 F3,12 = 21.82

IC50 range: 0.039µM-1.0µM (Mouse brain AChE)• The carbonyl group(s) of the

phthalimide/indanone/benzoyl groups forms ahydrogen bond in the active site of the enzyme.

F - electronic field parameter for X; I-P = 1 for para Xsubstituents and 0 for X meta/ortho substituents; I-O


=1 for ortho X substituents and 0 for X meta/parasubstituents.

length is n=2 (IC=0.028µM) and that increasing ordecreasing this spacer leads to a reduction in activity(n=1, IC50= 19µM ; n=3, IC50=0.22µM).

It was found that para X substituents are favoured foractivity whereas ortho substituents are detrimental. Inaddition, electron withdrawing groups are favoured.This result is similar to QSAR models generated for Xsubstituents in the benzylamines class (Equations (8)and (9)). It has also been shown that a significant QSARequation containing log P can be derived. However itwas found that log P represents an alternative to theelectronic parameter F. A wider range of electronic andhydrophobic substituents may result in both effectsappearing in a QSAR model. The highest affinityanalogue was found to be the 4-pyridyl substituent(IC50=0.039µM) within this series.

( )n

N

YO

X

Fig. (12). The benzylpiperidine class of AChE inhibitorslinked to a benzoyl moiety.

Subsequently, Sugimoto et al [35] synthesized aseries of benzylpiperidines bearing a phthalimidemoiety linked via an ethyl spacer to position 4 of thepiperidine ring.

Further modifications within this benzylpiperidineseries (Fig. (1 1 )) were performed in which the amidegroup was replaced in the analogue in which m=1 andY=X=H. It was found that amide nitrogen was notimportant for AChE inhibition (COCH2, IC50=0.53µM)and that carbonyl group was crucial for activity (CH2NH2,IC50>4.6µM). Furthermore, it was also found thatsubstitution of the amide hydrogen could lead to animprovement in AChE inhibiton (CONR R=Me, Et, Ph;IC50’s = 0.17µM, 0.13µM and 0.04µM, respectively).The carbonyl group of the amide is required for activity,presumably for H-bonding to the enzyme.

A QSAR was developed for a set of phthalimides(Dataset 10) where n=2, Y=H and the varying groupwas X [23] as shown in Fig. (1 3 ).

( )n

N( )m Y

O

X

n=2, m=1, Y=H, X = 3 or 4 substituted

X = NO2, NH2, OMe, NHCOR’ [R’=Me,CH2Ph], COR’’

[R’’=NHCH2Ph, Ph]A series of 1-substituted-4-[2-(N-benzoyl-N-

methylamino)ethyl]piperdine analogues weresynthesized [34]. Within this series, it was known thatthe 1-benzyl derivative exhibits an IC50 of 0.17µM forAChE. For a series of substituted benzyl analogues, noobvious SAR trends were apparent. Thecyclohexylmethyl group was found to exhibit moderateAChE activity (IC50= 0.41µM). It was also found thatincreasing the chain length between the basic nitrogenof the piperidine ring and benzyl group from m=1 leadsto inactivity. Substituents such as N-benzoyl whichremove the basic nitrogen in the piperidine ring led toinactivity (IC50= 52µM). In all analogues synthesizedwithin this benzylpiperidine series, the benzamidemoiety was always linked to the benzylpiperdine ring viaan ethyl linker.

Fig. (13) .1-Benzyl-4-(2-phthalimidoethyl) piperdine andderivatives (Dataset 10).

Log 1/IC50 = 0.234 (0.138)MR(4) - 0.417(0.202)MR(3)+ 7.978(0.299) (1 1 )

n = 11 r2 = 0.881 s = 0.260 F2,8 = 29.47

IC50 range: 0.002µM – 0.281µM (Mouse brain AChE)

MR(3) and MR(4) is the molar refractivity of substituentsX at the 3 and 4 positions, respectively.

The authors suggest that interpretation of theQSAR suggests substituents at the position 4 wouldfavour polar interactions whilst at position 3 (ortho tothe phthalimide) bulky substituents are unfavourable.Despite the structural similarity of this series withDatasets 7-9, there appears to be little similarity in theparameters associated with X substituent as beingimportant for AChE inhibition. The lack of an electronicor hydrophobic term in Equation (11) may result fromthe limited variation in these properties with this series.

In a separate study, Ishihara et al. [33] examined aseries of benzylpiperidines linked to a benzoyl moiety(Fig. (1 2 )). Only a small number of analogues weresynthesized. It was found that Y=2-OMe wasunfavourable for AChE activity (IC50=1.9µM) in theanalogue in which n=2 and X=4-pyrrolidino (Fig. (1 2 )).However, the analogue in which n=2, Y=H and X=4-pyrrolidino, a high affinity of AChE was observed(IC50=0.028µM). For X= 4-pyrrolidino and Y=H, thechain length between carbonyl and benzylpiperdinewas investigated. It was found that the optimal chain

Within this series, analogues (X=NO2, Y=H, m=1)were synthesized to allow the investigation of varyinglinker (n) between the phthalimide moiety andbenzylpiperdine. It was found that n=2 (Fig. (1 3 )) is


optimal for activity (IC50=0.03µM) whilst decreasing thelinker to n=0 or 1, a reduction in activity was observed(IC50=2.7µM and 3.0µM, respectively). Furtheranalogues to explore the effect of the Y substituent onactivity were synthesized, but no conclusions could bedrawn.

Log 1/IC50 = -0.107S0 + 0.046IX - 0.151HOMO +6.683 (1 3 )

n = 15 r = 0.905 s = 0.46 F = 16.7

Y = various combinations of meta/para and orthosubstitutions of F,OH,Me,NO2, OMe

With introduction of an indanone moiety in place ofthe phthalimide, E2020 (4) (Donzepil/Aricept®) - anAChE inhibitor currently marketed for the treatment ofAlzheimers disease was identified [36-38].Identification of an achiral AChE inhibitor with a profilesimilar to E2020 (IC50=0.006µM, Mouse brain AChE)(Fig. (1 4 )) would be desirable.

IC50 range: 0.001µM – 5.0µM (Mouse brain AChE)

S0 - non-overlap steric volume of the substitutedbenzyl of an analogue as aligned onto theunsubstituted phenyl, IX - largest principal moment ofinertia and HOMO the highest occupied molecularorbital of the substituted benzyl group.

A 3D QSAR studies of E2020 and its analogueswere generated using the molecular decomposition-recomposition [38] approach to allow the conformationof substructural fragments within E2020 to beaccounted for (Fig. (1 4 )). For Feature 1 (indanone ring)and Feature 4 (benzylpiperidine) - a variety ofdescriptors, some of which are dependent onconformation, were examined.

As S0 for Y becomes larger i.e. the substituentsbecome more bulky, then inhibition decreases. The IXand S0 terms reflects that increasing the bulk in theregion of the benzyl ring leads to a decrease in AChEbinding. Furthermore, the HOMO term suggests that alow electronic density in the aromatic ring is important.

In general terms, it is suggested that the aromaticring of the indanone moiety is less sensitive tosubstitution as bulky substituents can be toleratedwhereas the benzyl group of the benzylpiperidinemoiety is sensitive to the size of the para substituent inparticular. These results show some agreement withthe QSAR (Equations (10) and (11)).

N( )m Y

X

O

( )n

Feature 1: Indanone Moiety Feature 4: Benzylpiperidine MoietyQualitative SAR for the E2020 reveals that the basic

nitrogen and the indanone carbonyl are crucial foractivity. Replacement of the indanone carbonyl with ahydroxyl (IC50=0.3µM) leads to a drop in AChEinhibition. Additionally, the spacers n and m are foundto be important for positioning of key groups forinteraction with the enzyme. Increasing the chain m >1in E2020 a reduction in activity is observed, andincreasing n to 2 (IC= 0.15µM) or decreasing n to 0(IC50=3.3µM) in E2020 leads to a reduction in activity.The basic nitrogen is required in E2020 for activity asreplacement of benzylpiperidine with an N-benzoylgroup leads to inactivity. Furthermore, replacement ofthe benzyl group of the benzylpiperidine with acyclohexylmethyl (IC50=0.009µM) leads to retention ofAChE activity.

Fig. (14). E2020 (X=5,6-diMeO, n=1, m=1, Y=H) and itsanalogues.

The QSAR for Feature 1 is given by:

-Log 1/IC50 = 2.37C4 + 2.86UT - 0.14UT2 -

156.7HOMO -8.25HOMO2 - 740.9 (1 2 )

n = 18 r = 0.804 s = 0.46 F = 4.4

X = various combinations of ortho,meta and/or parasubstituents such as OH,OMe,H,F

IC50 range: 0.002µM- 0.38µM (Mouse brain AChE)

C4 - HOMO out of plane π orbital coefficient of ringcarbon four, UT - dipole moment of the indanonemoiety and HOMO - highest molecular orbital energy.

Molecular Shape Analysis (MSA) was employed todetermine the active conformation of E2020 [39].Using a variety of analogues related to E2020 withvarying activity, it was predicted that the bioactiveconformation has the indanone ring perpendicular tothe piperidine ring as found in the X-ray structure ofE2020. However, the MSA approach suggests that thebenzyl group of benzylpiperidine points towards thecarbonyl of the indanone ring rather than being fullyextended as in the X-ray structure.

The QSAR suggests that the π electron density forthe indanone moiety as reflected by the HOMO and C4parameters in Equation (12) and that a dipole momentof magnitude between 2.8 and 6.6 Debye areimportant for activity.

The QSAR of Feature 4 is given by:


The inactivity of some analogues could beexplained by their inability to adopt the bioactiveconformation or by their lack of key functional groupsessential for activity. Such an approach could notexplain the activity profile of the analogues in whichthere is no spacer (n=0 or n=2) between the indanonering and the piperidine ring. For n=0, it was assumedthat inactivity of this analogue results from its inability toadopt a conformation in which its carbonyl and basicnitrogen can overlay with E2020 (n=1). Furthermore,the inactivity of trans-decalin analogue of E2020 couldnot be explained by its conformational behaviour andwas assumed to result from bad steric interactions withAChE. In addition, many of the substituents on thebenzyl ring which result in low activity could beexplained by their inability to adopt the bioactiveconformation. However, in cases such as 2-napthylwhich is inactive, it is found that this moiety can adoptthe correct conformation and is assumed to be inactivedue to a steric clash with the enzyme. These resultssuggest that there is little space around the benzylgroup of E2020 when bound to AChE, in agreementwith the QSAR model (Equation (13)).

n = 15 r2 = 0.440 s = 0.410 F1,13 = 10.22

IC50 range: 0.001µM- 0.10µM (Human ErythrocyteAChE)

σ - Hammett electronic parameter for substituent X

Only a very poor QSAR could be derived from whichno real explanation for activity could be obtained.Dataset 11 contains mainly substituents on position 6on the aromatic ring and show a limited activity/propertyrange. Unlike for Datasets 7-10, electron-donatingsubstituents are favoured on the benzisoxazolemoiety. A qualitative analysis of this series [43]suggests that electron donating substituents favourAChE inhibition, there also appears to be a preferencefor hydrophobic substituents and that steric size ofsubstituents appear not to be important for activity.

The QSAR models for benzylamines andbenzylpiperidines classes of AChE inhibitors showsome similarity. Interpretation of these models wouldsuggest these compounds adopt a similar mode ofinteraction with AChE. The most important effectsinclude an electronic effect exerted by substituents onthe aroyl moiety; whereby electron-withdrawing groupsare favoured. In addition, as the hydrophobicity of themolecules decreases the AChE activity increases.These changes in hydrophobicity may be related tochanges in the hydrophobicity of the X substituents.The benzisoxazoles appear from the limited QSAR toshow a difference in the binding of the X substituentsas compared with the benzylpiperidines andbenzylamines classes. An electron-donating Xsubstituent for the benzisoxazoles has beensuggested to be favoured for AChE activity byincreasing the electron density on the oxygen atom,which may form a hydrogen bond with the enzyme.

As with the benzylamine class of inhibitors, a similarmode of interaction for the benzylpiperdine class withAChE has been hypothesized. On the basis of thismodel, rigidification studies of E2020 led to theidentification of a potent AChE inhibitor, TAK-147 [40]which is undergoing clinical trials as a therapeutic agentfor Alzheimers disease.

Benzisoxazoles

Bioisosteric replacement of the indanone ring inE2020 was attempted using a benzisoxazole moiety inthe search of a AChE inhibitor with an improved in vivoprofile. This led to identification of CP118590 which isbeing evaluated in the clinic. A number of analogues(Fig. (1 5 )) were synthesized in this series [41-43].

The qualitative SAR available for these AChEinhibitor classes also suggest that they all sharefunctional groups (aromatic, basic nitrogen andcarbonyl groups) which have been shown to beimportant for activity. However, there is also evidenceindicating that there may be a few localised differencesin their binding profiles. In particular, there appears tobe a difference in the optimal chain length between thebasic nitrogen group and carbonyl group for thebenzoyl linked to a benzylamine moiety (n=5, m=1, 7carbon chain length) and benzoyl linked to abenzylpiperidine moiety (n=2, m=1, 6 carbon chainlength). These observations suggest that the differentenvironments of the basic nitrogen within these seriesleads to a different interaction of the benzylaminomoiety with AChE. This could also account fordifferences in effect of Y substituent on AChEinhibition between these two series. Furthermore,comparison of the phthalimide-benzylpiperidine series

NO

N

Y

X 45

6

7

X = H, Me, OH, OMe , NH2, Br, CN, CONH2, NHCOR’ [R’=Me, Ph etc]

X = 5 and/or 6 or 7 substituted; Y =H

Fig. (15). The benzisoxazole class of AChE inhibitors(Dataset 11).

For Y=H, a QSAR model were derived to explain theeffect of X substituent on AChE activity for Dataset 11is given in Equation (14).

Log 1/IC50 = -1.274 (0.861)σ + 7.987(0.231) (1 4 )


reveals that a longer chain between the carbonyl andbasic nitrogen as compared to the benzoyl andindanone-benzylpiperidine series. This could not beexplained by MSA analysis but suggests a differentmode of binding.

were active (IC50> 100µM, rat brain AChE) and this ledto the hypothesis that either the conformationalconstraints imposed by the fused ring system ofHuperzine A or the steric and/or hydrophobic nature ofboth of the other rings are important for AChEinhibition.

Huperzine A Further analogues were synthesized in order toinvestigate the contributions of the three carbonbridge, endocyclic double bonds and methyl group. Anoverlay of HupA and ACh suggests that the threecarbon bridge in Hup A would not be necessary foractivity. However, it was found that the three carbonbridge essential for AChE activity based on the activityof the Hup A analogue (IC50=>100µM, rat brain AChE)in Fig. (16A (II)). This result suggests that ACh and HupA adopt different modes of binding within the enzyme’sactive site. Two analogues of HupA (Fig. (16A (III))) inwhich the double bond between C7 and C8 is removedand the C7 methyl is equatorial or axial weresynthesized. Both of these analogues were found toexhibit high AChE (equatorial Me (IC50=0.9µM) > axialMe (IC50=1.6µM)) albeit significantly lower than HupA.In order to understand these results, the steric fields ofboth the equatorial and axial methyl substituted Hup Aanalogues were modelled. It was found that equatorialmethyl analogue mimics the shape of Hup A better thanthe axial methyl. The lower affinity of these analoguesas compared HupA could be explained by differencesin electrostatic potential about 3 carbon bridge [44].

Huperzine A (5) (Fig. (1 6 )) represents a new class ofpotent, reversible AChE inhibitor which is undergoingclinical trials for Alzheimers disease. Understanding theSAR of Huperzine A (HupA) is important in order to finda similar more easily synthesizable AChE inhibitor with abetter in vitro (IC50=0.07µM, rat brain AChE (44)) and invivo profile as HupA.

NH

H3C

CH3

OH2N

H

7

610

23

45

11

9

1

8

Fig. (16) . The structure of Huperzine A (HupA).

NH

OHRNNH

H3C

R

OH2N

H

l (a) H (b) Me

R = equatorial or axial CH3lll

NH

CH3

H3C H2NO

ll

Other qualitative SAR data reveals that substitutionof C7 methyl with a phenyl leads to inactivity; this hasbeen interpreted as little room being available forsubstitution on three carbon bridge. Conversion ofpyridone to phenyl leads to inactivity suggesting thatpyridone moiety can form H bonds with AChE.Replacement of C11-ethylidene with both methyl andpropylidene leads to a drop in activity - the former is duelack of van der Waals contacts and latter due to stericsize of the substituent which clash with enzyme.Replacement of NH2 with either NMe2 or CNH2 resultsin inactivity as does conversion of pyridone tomethoxypyridine [45]. All these analogues were testedfor AChE inhibition using the rat brain and all exhibitedactivity IC50>10µM.

Replacement of pyridone ring with pyrimidone ringleads to poor activity (IC50 >10µM) as compared toHupA. For pyrimidones, the additional nitrogen hasbeen suggested to be electrostatically unfavourable foractivity [46].

Fig. (16A) . Huperzine A analogues.

No QSAR studies have been published for HupA,however modelling studies have been performed andqualitative SARs derived. These are presented below.

Based on the SAR of HupA, it would appear thathydrophobic and hydrogen bonding interactions withenzyme are important for its activity. Furthermore, anionizable nitrogen is also required for activity.

To delineate the pharmacophoric elements ofHuperzine A [44], conformational flexibility ofaminomethyl substituted pyridones were investigated(Fig. (16A (I))). Neither of these compounds A or B


QSAR Summary based on the use of 2D descriptors. There is noaccounting of conformational properties of inhibitorsand the fact that inactivity may result from the inability ofa molecule to adopt the bioactive conformation.Furthermore, no rationalization of the binding affinitiesof enantiomers is possible.

Examination of the QSAR models and qualitativeSAR data derived suggests that the different classes ofAChE inhibitors adopt unique binding modes with theacetylcholinesterase enzyme.

However, generally speaking, it is observed that allinhibitors tend to be of hydrophobic in nature assuggested by the presence of log P in the majority ofQSAR models. In addition, it can be seen that allclasses of inhibitors contain an ionizable nitrogenalthough the importance of a basic nitrogen for AChE isnot reflected in the QSAR models generated. The pre-requisites of hydrophobicity and an ionizable nitrogenfor AChE inhibition has since been confirmed by thecrystal structure of AChE-inhibitor complexes [20, 47]and enzyme docking studies. Hydrophobicity is a keyfeature of AChE inhibitors given that the active site ofthe enzyme is hydrophobic, lined with aromaticaminoacids. The ionizable nitrogen appears to beimportant for anchoring the inhibitors within the activesite via cation-π interaction between the ionizednitrogen and aromatic Trp 84 residue in the enzyme’sactive site.

It has been well established that the most importantfactors for a favourable interaction between a drug andits specific biological target are best represented by 3Dproperties - a 3D geometric fit of the ligand to a bindingsite (both ligand and target are in low energyconformations), a complement of electrostatic potentialsurfaces, the formation of charged and/or neutralhydrogen bonds between functional groups andhydrophobic interactions between lipophilic surfaces.Only one example of 3D-QSAR modelling [38, 39] wasemployed prior to publication of the 3D structure of theAChE.

Molecular Modelling of Acetylcholine-sterase Inhibitors

Since the publication of the crystal structures ofboth the native AChE [20] and the ligand-AChEcomplexes [47] structure-based drug designapproaches and 3D QSAR have been employed toboth attempts to design novel AChE inhibitors andrationalize existing SAR for AChE inhibitors. In thisSection, enzyme-ligand models will (1) be utilized toprovide information about the bioactive conformations,(2) to be employed to rationalize the SAR data and (3)to assess QSAR models for various classes of inhibitorsand (4) to be compared to experimental crystalstructures (where available).

The common features described above representvery general requirements for AChE inhibitors. Theinteraction of the different chemical classes are veryspecific to each class as dictated by the functionalityand their 3D positioning within each molecularframework. It is further observed that the molecularframework for each class in its entirety is crucial foractivity; localised changes result in fine-tuning of theglobal inhibitor-enzyme interaction.

Given the structural diversity of AChE inhibitors andthe comparison of the individual QSARs for eachseries, it can be seen that it would be impractical tostructurally align all known inhibitors in any unbiasedway and generate a meaningful QSAR model.

Advances in understanding protein-ligandinteractions have led to the development of newmolecular modelling approaches such as 3D QSAR astypified by COMFA (Comparative Molecular FieldAnalysis) [48]. Three-dimensional quantitativestructure-activity relationships (3D-QSAR) methodsthat computes steric, electronic and occasionallyhydrophobic interactions for a series of ligands on aregular 3D lattice of probe atoms. The quantitativeresults are tabulated and appropriate statisticalmethods yield a QSAR equations highlighting key 3Dfeatures of the ligand responsible for activity. Thisapproach is extremely sensitive to the initial alignmentof the inhibitors under consideration. The alignmenteffect on the results has been well documented [49].

The QSAR approaches can be employed todevelop guidelines for target synthesis, and,retrospectively to explain unusual SAR data. For thepublished AChE inhibitors, little evidence exists for theuse of QSAR as a predictive tool in the design of noveland/or more potent AChE inhibitors. Despite the vastamount of synthetic efforts invested in searching forAChE inhibitors, there appears to be little QSARanalyses reported in the literature; much of the analysishas been of a qualitative nature. However, therationalization of physicochemical properties of AChEinhibitors important for enzyme inhibition represents astarting point upon which new and more potent AChEinhibitors can be designed.

Numerous structure-based drug design algorithmshave been published [50]. It is beyond the scope ofthis review to compare and contrast the differencesbetween algorithms and their potential effects on theresults of protein-ligand docking studies. For details ofOne of the limitations of 2D QSAR analysis is that it

is more suitable for congeneric series of inhibitors and


the docking approaches used for AChE inhibitors, thereader is referred to each paper. In SYSDOC [51] , theneed for an additional term to describe charge-transferinteraction was assessed but it was found that thisnature of interaction can be implicitly described byLennard-Jones and H-bonding terms.

greater (IC50= 0.0004µM, rat brain AChE) was observedas compared with tacrine (IC50= 0.60µM, rat brainAChE). This chain length leads to a distance ofapproximately 16 Å between ring nitrogens which hasbeen suggested as being favourable in bisquaternaryinhibitors for AChE inhibition. Further hybrid analoguesof Tacrine with Huperzine A [54] (Fig. 1 7 (II)) have alsobeen identified as showing improved affinity ascompared with tacrine.Tacrine

The crystal structure of the tacrine-AChE complexwas determined at a resolution at 2.5Å [47]. It wasfound that the aromatic system of THA is stackedagainst Trp84 in a π-π interaction, its ring nitrogen formsa hydrogen bond with His 440 and its amino nitrogen isbound to a water molecule. The implications of theQSAR models (Equations (1) and (2)) have not beenrationalized within this publication.

Benzylamines and BenzylPiperidines

E2020 has been examined in a number of enzyme-ligand docking studies. Using the SYSDOC software,the protonated form of E2020 were docked into theAChE [55]. Three likely binding sites were identified.Two of the binding sites (Fig. (1 8 )) show that benzylgroup of the benzylpiperidine is positioned deep downin the AChE gorge, interacting through π-π stackinginteraction with the aromatic residues Trp 84 and Phe330. The protonated nitrogen forms a cation-πinteraction with Trp 84, the indanone carbonyl forms ahydrogen bond with the enzyme and the phenyl groupof the indanone moiety is shown to interact with Trp279 with the methoxyls exposed to solvent. Thedifference between the binding sites is the differencein the hydrogen bonding pattern. The third binding siteshows E2020 binding in a reverse orientation in whichthe benzylpiperdine is positioned at the top of theAChE gorge as compared to other binding sites. Thetwo binding modes described above in which thebenzylpiperdine is positioned deep in the AChE gorgeshow agreement with the QSAR models derived forE2020 and its analogues.

The SYSDOC software was employed to examinethe binding of Tacrine (THA) to the Torpedo AChE [51]to check whether the software could reproduce thecrystal structure of the THA-AChE complex. A lowenergy complex of tacrine bound to AChE showedgood agreement with crystal structure. A second lowenergy enzyme-docking model was observed for THA;it was found that THA was bound to a secondperipheral site (Trp 279) at the opening to AChE gorge.At present there are no crystal structure for thisinteraction mode of THA with AChE, however this isconsistent with the experimental finding of a peripheralsite for THA [52].

On the basis of the crystal structure and the THA-AChE docking studies, it was proposed that two tacrinemolecules linked by a methylene chain may prove to bemore potent than tacrine by allowing bindingsimultaneously at both the catalytic and peripheral sites[53]. Using molecular modelling docking approaches, itwas found that an approximate chain length of 9methylene units should be favourable for AChE affinity.This analogue (Fig. 1 7 (I)) and a few others weresynthesized and tested for AChE inhibition. It wasfound that for n=7, an enzyme inhibition of 1000-fold

In a second docking study [56], the enzyme-ligandcomplexes of Torpedo Californa AChE and E2020analogues were determined using full forcefieldcalculations allowing complete conformational flexibilityin both the ligand and enzyme.

As with SYSDOC calculations, this energy baseddocking approach also suggests the possibility of

NHN

(CH2)n

HN N N

CH3

H2Nn = 7-10

l - Tacrine-Tacrine Hybrids ll - Tacrine-Hup A Hybrid

Fig. (17) . Tacrine hybrid analogues.


Fig. (18) . SYSDOC Model of E2020 with AChE (red - binding site 1 interactions ; blue - binding site 2 interactions).

multiple binding modes for E2020. A similar orientationwithin the AChE active site gorge is observed as shownin Fig. (1 8 ). However a number of alternate bindingmodes are also observed of which many show anumber of common interactions with the AChE but alsosome different localised interactions with AChE. Anunusual binding mode is observed whereby the E2020is aligned in the reverse orientation within the bindingpocket (indanone moiety at the bottom of the gorge).This orientation has been ruled out given that there islittle space for benzyl substituents unlike for theindanone moiety - based on QSAR models (Equation(13)).

could not be explained by either docking or QSARstudies.

In a third docking study [57] the crystal structure ofthe Torpedo AChE was modified to represent thehuman AChE. The only differences in the active siteresidues is the existence of Tyr at 330 rather than Phein the human AChE. Compounds with either aphthalimide, benzoyl or indanone moiety at one endand benzyl moiety connected to a tertiary ammoniumgroup at the other end were docked into the AChE.

For the phthalimide analogues shown in Fig. (7 ), thebenzyl group is deep in the active site pocket andorthogonal to Trp 84, the carbonyl group of thephthalimide moiety forms a hydrogen bond to Tyr 121of the enzyme and the phthalimide moiety is at the topof the gorge forming a π-π interaction with Trp 279. Theenzyme-ligand models show that electron withdrawingsubstituents are favoured on aromatic phthalimidesince they are exposed to solvent and since theyreduce the electron density of the aromatic moietyallowing a stronger interaction with Trp 279.Replacement of aromatic phthalimide with cyclohexanealters conformation and is too bulky leading to its pooractivity. There is little space around the benzyl groupwhich explains the inactivity of p-OMe. The o- and m-OMe activity results from hydrogen bonding to AChE.When the length m between the ammonium group andbenzene ring (Fig. (8 (III)) is made longer, the activitydecreases for m=2 but not for m=3. For the analoguewhere m=2, bad steric clashes with AChE is observedwhereas for more flexible analogue where m=3, areasonable conformation can be adopted withoutincurring bad contacts with the enzyme. Replacementof phthalimide with benzoyl or indanone moieties leadsto compounds that retain AChE activity. Similar

In this study, the E2020 analogues for which thenumber of methylene units between the indanonemoiety and the benzylpiperidine is 0 or 2 (Fig. (1 4 ))have been docked in AChE in order to rationalize theiractivity. The activity profile of these E2020 analoguescould not be explained by 3D QSAR studies. For n=0,it is found that benzyl group of the benzylpiperidinemoiety is at the bottom of the gorge parallel to Trp 84.The lower activity of this compound as compared toE2020 is attributed to the inability of the indanonemoiety to interact with Trp 279. The analogue of E2020where n = 2, has weaker activity than E2020. It wasfound that this analogue adopts a similar bindingorientation within the active site of AChE as E2020 butcan not fulfil as many strong interactions as observedwith E2020.

The inactivity of the trans-decalin analogue ofE2020 was assumed to be result from steric clasheswith AChE in QSAR studies. Docking models of thisanalogue with AChE shows that this compound can besterically accommodated within the enzyme active site.The inactivity of the trans-decalin analogue of E2020


enzyme-ligand models as observed for phthalimideanalogues could explain these observations.

bond could be formed between benzoyl C=O andAChE Tyr 121 OH in a low energy conformationthereby explaining its low activity. The inactiveanalogue in which n=2, no conformation could beobtained where the piperidine and benzoyl moietiescould be accommodated sterically within the active siteof AChE.

Docking of the 3-arylpropenamide series (Fig. (9 ))reveals that their vinyl moiety and not the benzene ringinteracts with Trp 279. The higher activity of the acetylgroup at R as compared to H is suggested to haveresulted from favourable van der Waals interaction ofacetyl methyl group with AChE. No hydrogen bondbetween the acetyl group and AChE was observed,which was suggested to explain the higher activity byQSAR models (Equation (9)).

For benzylpiperidines with an indanone moiety (Fig.(1 4 )), a similar model of enzyme-ligand interaction aswith benzoyl series was observed. Forbenzylpiperidines with a phthalimide group (Fig. (1 3 )),an extra carbon between phthalimide and piperidine isrequired to allow the molecule to adopt a low energyconformation which can adopt a similar mode ofinteraction to its benzoyl and indanone equivalents.

Representative compounds of the benzylpiperidineseries were also docked into AChE. For the benzoyl-piperidine analogues (Fig. (1 2 )), a model of theirinteractions with AChE shows that the benzoyl groupinteracts with Trp 279, the carbonyl moiety forms ahydrogen bond with Tyr 121 and benzyl group with Trp84. For the latter interaction, a stacking interaction withTrp 84 is observed for benzylpiperidine whereas anorthogonal interaction was observed for the N-alkylclass. This result explains the difference in the SAR forthe aromatic substituents (Y) between these twoseries. Furthermore, the piperidine nitrogen was foundto be shifted 1.5 Å closer to the peripheral site ascompared to the equivalent nitrogen in benzylamineseries. This explains the differences in optimal chainlength for the two series as observed in the qualitativeSAR analysis.

The crystal structure of E2020 bound to AChE wasrecently published [58], after the publication of theenzyme-docking studies described above. It was foundthat E2020 binds along the active site gorge. Principalinteractions with the enzyme occur through the benzyl,piperidine nitrogen and indanone moieties (Fig. (1 9 ))which shows agreement with SAR and enzyme-liganddocking studies. The main interactions arehydrophobic and stacking in nature and are observedwith the conserved aromatic residues which line thegorge. Interestingly, E2020 forms no direct contactwith the protein via hydrogen bonds/salt bridges; allhydrogen bonds are mediated via water molecules.The role of solvent is shown to be important and wasnot accounted for in the simulated ligand-enzymemodels.

For analogues where the chain length between thebenzoyl group and piperidine is n=3, no hydrogen

Fig. (19) . The interactions between E2020 and AChE as observed in the crystal structure (E2020 in Dark Green).


One unexpected observation is that the principalbinding site for protonated nitrogen of AChE inhibitorsis usually the indole ring of Trp 84. In E2020, it wouldappear as though Phe 330 may play a key role. Theselectivity profile of E2020 for AChE over BuCHE isattributed to interactions with Trp279 and Phe 330.

its pharmacophoric elements, this led to severalerroneous docking studies [46]. Saxena et al. [59]proposed a carbonyl of HupA points towards oxyanionhole and that the primary group interacts with Glu 199.Ashani et al.[60] suggested that Huperzine NH2 andendo- and exo-cyclic groups interacts with Trp 86 andTyr 337 in human AChE (Trp 84 and Phe 330 inTorpedo Californa AChE), and that the pyridoneheteroatoms interact with aminoacids distal to Tyr 337.

Benzisoxazoles

Molecular dynamics simulations of a benzisoxazoleanalogue (X= 6-NHCOCH3; Fig. (1 5 )) in a solvatedactive site of the crystal structure of Torpedo CalifornaAChE was performed [42]. Key interactions observedincluded a critical interaction between the protonatedpiperidine and negatively charged side chain of Asp 72of AChE, a hydrogen bond between thebenzisoxazole oxygen and NH of Phe 288, the N-benzyl group forms an off-center π stacking with Trp84. Assuming the binding mode in which theprotonated nitrogen of the benzisoxazole interacts withAsp 72 in AChE, the acetyl substituent is primarilyobserved as being exposed to solvent and thebenzisoxazole oxygen forms a hydrogen bonds withPhe 288 mediated via water molecule. This enzyme-inhibitor model explains the requirement of electron-donating substituents (Fig. (1 5 )) on the benzisoxazolering in the QSAR study (Equation (14)). Electron-withdrawing substituents will increase the electrondensity on the oxygen of the benzisoxazole ring andthereby allowing a stronger hydrogen bondinginteraction with Phe 288 via water.

Using the SYSDOC software [51], Huperzine A wasdocked into the crystal structure of the native AChE. Inthis docking study, no a priori assumption was madeabout the binding site of Hup A on AChE. It was foundthat Huperzine A can bind to the bottom of gorge ofAChE close to the catalytic domain as well as close tothe opening of the gorge near the peripheral site. Atthe bottom of the gorge, Huperzine A could adoptthree different binding modes with AChE as shown inFig. (2 0 ).

The possibility of three possible catalytic subsites isnot fully understood but it has been suggested thatlack of solvent or dynamic interactions in such asimulation may be responsible. The low affinityperipheral site has been suggested to be important forthe build up concentration of HA at gorge entrance.The HupA-AChE docking model in which Hup A bindsclose to the catalytic site of AChE has been usedsuccessfully to rationalize the SAR of some Hup Aanalogues.

Huperzine AAfter the publication of the studies described

above, a 2.5 Å resolution crystal structure of HuperzineA complexed to Torpedo Californa AChE was solved[61]. The crystal structure reveals that despite beingsuch a strong AChE inhibitor, Huperzine A makes few

Initial docking studies assumed a plausibleorientation for HupA being parallel to AChE based on

Fig. (20) . SYSDOC Model of Huperzine A with AChE (red - catalytic binding site I,II and III interactions ; blue - peripheral siteinteractions).


Fig. (21) . Crystal Structure of HupA with AChE (Hup A in Dark Green, Catalytic Triad in Pink).

direct contacts with the enzyme (Fig. (2 1 )). With threepotential donor and acceptor sites on the ligand, it isinteresting to observe that only one hydrogen bond isseen between the pyridone oxygen and Tyr 130. Thisexplains the importance of this carbonyl group and whyits substitution to methoxy group yields inactivity. Thering nitrogen and protonated nitrogen are hydrogenbonded via solvent to the protein. The protonatedamino group interacts with Trp 84 and Phe 330 (Tyr inhuman AChE allow hydrogen bonding to occur). Thisobservation explains the species difference activity ofHuperzine A.

introduced at C10 in HupA (IC50=0.002µm, FetalBovine Serum (FBS) AChE); the correspondingequatorial methyl was observed to decrease activity by1.5-fold (IC50=0.03µm, FBS AChE) . Substituentslarger than methyl were found to lead to a drop inactivity - the implication being that there was a stericclash between the inhibitor and AChE. Docking studies[62] reveals that axial methyl at C10 position of HupApoints towards a small hydrophobic pocket whilst theC10 equatorial faces into a hydrophilic pocket. Thisadditional axial methyl group at C10 increases the logPof molecule and results in favourable additionalhydrophobic binding interactions with enzyme ascompared to HupA.In general, hydrophobic interactions with aromatic

residues in active site of AChE are observed. Thereappears to be little room for additional groups onHuperzine A without resulting in steric clashes with theenzyme. The crystal structure of the AChE andHuperzine A has been used to explain the SAR ofHuperzine A and its analogues. Interestingly as withother AChE inhibitors, no binding to the peripheral sitewas observed in the crystal structure.

On basis of crystal structure, the pyridone ring ofHupA was replaced by catechol and phenol moieties[63]. It was found that although these analogues arepredicted on the basis of the crystal structure of HupA-AChE to be active, they were both found to be inactive(Ki > 60µM, FBS AChE). This observation can notreadily be explained by SAR or docking studies. It maybe speculated that the crystal structure represents onestatic viewpoint of the interaction of an inhibitor with theenzyme and that role of solvent within the active site isnot understood.

C10 analogues of HupA were synthesized andtested for their ability to inhibit AChE [62]. A 8-foldactivity increases when an axial methyl group was


Miscellaneous AChE Inhibitors Even with the crystal structure of AChE-inhibitorcomplexes, the problem of predicting AChE - inhibitorinteractions for new chemical series is complicated bythe flexibility of the ligands and enzyme, the role ofsolvent and the size of the AChE active site. Enzyme-docking models have been shown in general to be ableto provide a reasonable prediction of ‘likely’ bindingmode of inhibitors such as E2020 and Huperzine A ascompared to the crystal structure. A priori assumptionsabout the nature of an inhibitor-AChE interactions maylead to erroneous models as observed with the initialdocking studies of Huperzine A. In these models, it wasassumed that Huperzine A adopts a similar bindingmode to ACh and subsequently, the docking studiesshow a very different interaction of HupA with theenzyme than observed in the crystal structure.Interestingly, solvent has been shown to be importantin mediating interactions between inhibitors such asE2020 and Huperzine A and the enzyme in the crystalstructure; and its role is often ignored in modellingstudies.

Finally, a 3D QSAR COMFA study has beenreported for a number of different chemical classes ofAChE inhibitors [64]. The alignment of the threeinhibitors - edrophonium, tacrine and decathonium inthe AChE active site has been employed as a templateupon which other structurally analogous AChEinhibitors have been superimposed.

A total of 60 AChE inhibitors, includingedrophonium, neostigmine and physostigmine, werestructurally aligned on the basis of the knowledge ofseveral crystal inhibitor complexes. These alignedcompounds were then subject to COMFA analysis. Across-validated correlation r2 of 0.93 was obtained. TheCOMFA steric fields were found to explain thedifferences in the AChE activity of the inhibitors. Thesteric contours observed are compatible with the activesite environment - unfavourable steric regions areoccupied by His 440, Ser 226, Glu 199, Ser 200 andPhe 288 whereas favourable steric regions do notcoincide with residues from the enzyme. Theelectrostatic contours were found also to correlate wellwith regions in the active site of AChE.

The enzyme-based models have also been usefulin rationalizing the QSAR models generated in order tounderstand further the AChE-inhibitor interactions. Forexample, the activity profile of E2020 analogues inwhich the length of the chain (n) was varied betweenthe indanone moiety and piperidine ring could not beaccounted for using 2D or 3D QSAR approaches.Whereas protein-docking models offered a reasonableexplanation. Such information available from QSAR andenzyme-docking models has been used successfullyto predict more potent analogues within a particularseries. There are however examples whereby enzyme-docking models have been unsuccessful at explainingthe activity profiles of compounds.

Development of such a COMFA model would beextremely difficult in the absence of crystal structures ofdifferent AChE-inhibitor complexes. Furthermore, thepredictivity of such a model will be confined to chemicalseries employed in the study or limited to new serieswhich can bind within the same region of the bindingsite. The authors suggest that docking algorithms maybe employed in the identification of new chemicalseries of AChE inhibitors and 3D QSAR models can beemployed to prioritize compounds for synthesis andtesting. The rationale for this type of approach is basedon the premise that the scoring functions in dockingalgorithms to prioritize active compounds are poor andthat the 3D QSAR models may be better. However, noevidence is provided in this publication to confirm thisstatement and no new chemical classes are suggestedas potentially potent AChE inhibitors.

The existence of a crystal structure has beenemployed successfully and unsuccessfully to designmore potent analogues of known inhibitors e.g. tacrineand Huperzine A, respectively. It must realised that thecrystal structure of inhibitor-AChE complexes offers astatic representation of enzyme-inhibitor binding.However, it is observed that the structure of AChEdoes not appear to undergo significant conformationalchanges upon binding various sizes of inhibitors.Therefore, the crystal structure represents a goodstarting point for designing new AChE inhibitors.

Summary of Molecular Modelling ofAcetylcholinesterase Inhibitors.

Crystallographic studies have shown that the AChEactive site is a deep narrow gorge lined by 14 aromaticresidues which penetrates into the enzyme wideningits base where the catalytic triad featuring Ser200,His440 and Glu327 are positioned. Thecocrystallisation of the enzyme with various inhibitorshas led to detailed information about the bindingmodes of various classes of inhibitors.

Despite the availability of the crystal structure ofAChE, it is interesting to note that there are nopublications which discuss the identification of a newchemical series of AChE inhibitors. Most of the dockingstudies have focused on AChE inhibitors which wereidentified prior to the crystal structure of AChE beingsolved. Techniques such as De Novo drug design,structure-based design in conjunction with


combinatorial chemistry approaches may be the futureof identifying new AChE inhibitors.

[17] Enz, A.; Amstutz, R.; Roddeke, H.; Gmelin, G.; Malanowski, J. InProgress in Brain Research; Cuello, A.C., Ed., ElsevierPublishers, 1993, Vol. 98, pp431-438.

Conclusions[18] Ogane, N.; Giacobini, E.; Struble, R. Brain Res., 1992, 589, 307.

[19] Perola, E.; Cellai, L.; Brufani, M. Bioorg. Med. Chem. Lett., 1998,8, 575.There have been significant effort invested towards

the discovery of novel acetylcholinesterase inhibitorsduring the past decades. During this time, severalstructural classes of inhibitors have been developedand a few have undergone human clinical trials andhave been marketed for a variety of diseases. Thepursuit of inhibitors for human acetylcholinesterasecontinues with the aim of identifying drugs with a moreattractive biological and/or chemical profile to currentlymarketed drugs. With detailed knowledge about AChEinhibitor-enzyme binding modes from QSAR models,enzyme-docking models and crystallography, newchemical classes of inhibitors may be designed in thefuture, specially from De Novo design and hybrids.

[20] Sussman, J.L.; Harel, M.; Frolow, F.; Oefner, C.; Goldman, A.;Toker, L.; Silman, I. Science, 1991, 253, 872.

[21] Schrag, J.D.; Schmid, M.F.; Morgan, D.G.; Philips, G.N.; Wah,C.; Tang, L. J. Biol. Chem., 1988, 263, 9795.

[22] Steinberg, G.M.; Mednick, M.L.; Maddox, J.; Rice, R. J. Med.Chem., 1975, 18, 1056.

[23] Recanatini, M.; Cavalli, A.; Hansch, C.; Chemico-BiologicalInteractions., 1997, 105, 199.

[24] Shutske, G.M.; Pierrat, F.A.; Kapples, K.J.; Cornfelt, M.L.;Szewczak, M.R.; Huger, F.P.; Bores, G.M.; Haroutunian, V.;Davis, K.L. J. Med. Chem., 1989, 32, 1805.

[25] Velnacrine Maleate, Mentane®, Drugs Fut., 1993, 19, 709.

Acknowledgements[26] Yu, Q.S.; Atack, J.R.; Rapoport, S.I.; Brossi, A. J. Med.

Chem.,1988, 31, 2297.

[27] Chen, Y.L.; Nielsen, J. Hedberg, K., Dunaiskis, A.; Jones, S.;Russo, L.; Johnson, J.;Ives J.; Liston, D. J. Med. Chem.,1992, 35,1429.

We would like to thank Dr David Rees for his criticalreading of this manuscript.

References[28] Alisi, M.A.; Brufani, M.; Filocomo, L.; Gostoli, G. Bioorg. Med.

Chem. Letters., 1995, 5(18), 2077.

[29] Main, A.R.; Biology of Cholinergic Function, 1975, 269.[1] Dale, H.H. J. Pharmacol. Exp. Ther., 1914, 6, 147.

[30] Chen, Y.L.; Nielsen, J.; Hedberg, K.; Dunaiskis, A.; Jones, S;Russo, L.; Johnson, J.; Ives, J.; Liston, D. J. Med. Chem.,1992,35, 1429.

[2] Caulfield, M.P.; Birdsall, N.J.M. Pharmacol. Rev., 1998, 50, 279.

[3] Colquhoun, L.M.; Patrick, J.W. Adv. Pharmacol. 1997, 39, 191.

[31] Ishihara, Y.; Kato, K.; Goto, G. Chem Pharm. Bull.,1991, 39(2),3225.

[4] Argyll-Robinson, D. Edinburgh Med. J., 1863, 8, 815.

[5] Rees, D.C.; Hill D. Annual Rep. Med Chem., 1996, 31, 41.[32] Ishihara, Y.; Kato, K.; Goto, G. Chem Pharm. Bull.,1991, 39(2),

3236.[6] Wittbrodt, E.T. Arch. Intern. Med., 1997, 24, 157.

[33] Ishihara, Y,; Miyamoto M,; Nakayama T,; Goto G. Chem Pharm.Bull.,1993, 41(3), 529.

[7] Millard, C.B.; Broomfield, C.A. J. Neurochem., 1995, 64,1909.

[8] Taylor, P. Neurology 1998, 51(1 Suppl 1), S30.[34] Sugimoto, H.; Tsuchiya, Y.; Sugumi, H.; Higurashi, K.; Karibe

,N.; Iimura, Y.; Sasaki, A.; Kawakami, Y.; Nakamura, T; Araki,S.; Yamanishi, Y.; Yamatsu, K. J. Chem Med.,1990, 33, 1880.

[9] Hucho, F. J. Neural. Transm. Suppl., 1995; 46, 59

[10] Makhaeva, G.F.; Filonenko, I.V.; Yankovskaya, V.L.; Fomicheva,S.B.; Malygin, V.V. NeuroToxicology, 1998, 19, 623. [35] Sugimoto, H.; Tsuchiya, Y.; Sugumi, H.; Higurashi, K.; Karibe,

N.; Iimura, Y.; Sasaki, A.; Nakamura, T; Araki, S.; Yamanishi, Y.;Yamatsu, K. J. Chem Med.,1992, 35, 4542.[11] Maxwell, D.M.; Brecht, K.M. Chem. Res. Toxicol., 1992, 5, 66.

[12] Hansch, C. In Computer Aided Molecular Design; ReynoldsC.N., Holloway M.K., Cox A.K., Eds.; American ChemicalSociety; Washington, DC, 1995, pp. 281-291.

[36] Sugimoto, H.; Iimura, Y.; Yamanishi, Y.; Yamatsu, K. Biorg. Med.Chem. Letters.,1992, 2(8), 871.

[37] Sugimoto, H.; Iimura, Y.; Yamanishi, Y.; Yamatsu, K. J. Med.Chem.,1995, 38, 4821.[13] Greenfield, S.A. Cell Mol. Neurobiol., 1991, 11, 55

[14] Brown, S.S.; Kalow, W.; Pilz, W.; Whittaker, M.; Woronick, C.L.Adv. Clin. Chem. 1981, 22, 1.

[38] Cardozo, M.G.; Iimura, Y.; Yamanish, Y.; Yamatsu, K.; HopfingerA.J. J. Med. Chem.,1992, 35, 584.

[15] Inestrosa, N.C.; Perelman, A. Trends Pharmacol. Sci. 1989, 10,325.

[39] Cardozo, M.G.; Kawai, T.; Iimura, Y.; Yamanishi, Y.; Yamatsu,K.; Hopfinger, A.J. J. Med. Chem.,1992, 35, 590.

[16] Fernandez, H.L.; Moreno, R.D.; Inestrosa, N.C. J. Neurochem.1996, 66, 1335.


[40] Ishihara, Y.; Hirai, K., Miyamoto, M.; Goto, G. J. Med.Chem.;1994, 37, 2292.

[52] Pomponi, M.; Giardina, B.; Gatta, F.; Marta M. Med. Chem. Res.,1992,2, 306.

[41] Villalobos, A.; Butler, T.W.; Chapin,D.S.; Chen, Y.L.; Ives, J.L.;DeMattos, S.B.; Jones, S.B.; Liston, D.R.; Nagel, A.A.; Nason,D.M.; Nielsen, J.A.; Ramirez, A.D.; Shalaby, I.A.; Frost White,W. J. Med. Chem,;1995, 38, 2802.

[53] Pang, Y-P.; Quiram, P.; Jelacic, T.; Hong, F.; Brimijoin, S. J.Biol. Chem. 1996, 271(39), 23646

[54] Badia, A.; Banos, J.E.; Camps, P.; Contreras, J.; Gorbig, D.M.;Munrz-Torrero, D.; Simon, M.; Vivas, N.M. Bioorganic & Med.Chem. 1998, 6, 427.[42] Villalobos, A.; Blake, J.F.; Biggers, C.K.; Butler, T.W.; Chapin,

D.S.; Chen, Y.L.; Ives, J.L.; Jones, S.B.; Liston, D.R.; Nagel, A.A.;Nason, D.M.; Nielsen, J.A.; Shalaby, I.A.; Frost White, A. J. Med.Chem.,1994, 37, 2721.

[55] Pang, Y-P.; Kozikowski, A.P. J. Comp. Aided Mol. Design.; 1994,8, 683.

[56] Ionue, A.; Kawai, T.; Wakita, M.; Iimura, Y.; Sugimoto, H.;Kawakami, Y.J. Med. Chem., 1996, 39, 4460.

[43] Villalobos, A.; Blake, J.F.; Biggers, C.K.; Butler, T.W.; Chapin,D.S.; Chen, Y.L.; Ives, J.L.; Jones, S.B.; Liston, D.R. J. Med.Chem.,1997, 105, 199.

[57] Ishihara, Y.; Kuntz, I.D. J. Med. Chem., 1997. 37, 3141.

[44] Kahn, M.; Nakanishi, H.; Chrusciel, R.A.; Fitzpatrick, D.;Johnson, M.E. J. Med. Chem., 1991, 34, 3399. [58] Kryge, G.; Silnab, I.; Sussman, J.L. J. Physiology (Paris), 1998,

92, 191.

[45] Kozikowski, A.P.; Xia, Y.; Reddy, J.; Tuckmantel, W.; Hanin, I.;Tang, X.C. J.Org. Chem., 1991, 56(15), 4636. [59] Saxena et al. Prot. Sci., 1994, 3, 1770.

[60] Ashani, Y.; Grunwald, J.; Kronman, C.; Velan, B; Shafferman, A.Mol. Pharmacol., 1994, 45, 555.

[46] Kozikowski, A.P.; Campiani, G.; Nacci, V.; Sega, A.; Saxena, A.;Doctor, B.P.; J. Chem. Soc. Perkin Trans.I , 1996, 1287.

[61] Raves, M.L.; Harel, M.; Pang, Y-P.; Silman, I.; Kozikowski, A.P;Sussman, J.L. Nature Structural Biology., 1997, 4(1), 57.

[47] Harel, M.; Schalk, I.; Ehretsabatier, L.; Bouet, F.; Goeldner, M.;Hirth, C.; Axelsen, P.H.; Silman, I.; Sussman, J.L. Proc. Natl.Acad. Sci. USA, 1993, 90, 9031. [62] Kozikowski, A.P.; Campiana, G.; Sun, L-Q.; Wang, S.; Saxena,

A.; Doctor, B.P., J. Am. Chem.Soc., 1996, 118, 11357.[48] Cramer, R.D. III.; Patterson, D.E.; Bunce, J.D. J. Am. Chem.Soc., 1988, 110, 5959. [63] Campiana, G.; Kozikowski, A.P.; Wang, S.; Ming, L.; Nacci, V.;

Saxena, A.; Doctor, B.P. Bioorg. Med. Chem. Letters., 1998, 8,1413.

[49] Folkers, G; Merz, A.; Rognan, D.; In 3D QSAR in Drug Design;Kubinyi H, Ed.; Escom, Leiden, 1993pp583-618.

[64] Cho, S.J.; Garsia, M.L.S.; Bier, J.; Tropha, A. J. Med. Chem.,1996, 39, 5064.

[50] Reich, S.H.; Weber, S.E.; Perspectives in Drug Discovery andDesign., 1993, 1(2), 371

[51] Pang, Y-P.; Kozikowski, A.P. J. Comp. Aided Mol. Design., 1994,8, 669.

current medicinal chemistry molecular modelling and qsar ... · qsar studies in the design of...

Documents