Pestic. Sci. 1998, 52, 104È110

Quantitative Correlation between MolecularSimilarity and Receptor-Binding Activity ofNeonicotinoid InsecticidesAkira Nakayama* & Masayuki Sukekawa

Odawara Research Center, Nippon Soda Co., Ltd, 345 Takada, Odawara 250-02, Japan

(Received 30 March 1997 ; revised version received 28 July 1997 ; accepted 11 September 1997)

Abstract : Quantitative correlation between molecular similarity and receptor-binding activity of neonicotinoid insecticides such as imidacloprid and acetamip-rid was studied by using a method of similarity index and semi-empiricalmolecular orbital calculations. A series of compounds having an aromatic ringand a cyclic or acyclic amine moiety with an electron-withdrawing group weresubjected to the similarity-activity analysis. Energy-minimum structures and elec-trostatic properties of the molecules were obtained by MNDO-PM3. The elec-trostatic similarity of each molecule compared with the most active compoundswas found to correlate signiÐcantly with the binding activity to nicotinic acetyl-choline receptor (nAChR) in honey bee when the two molecules were superim-posed to maximize the molecular shape similarity by simplex procedure. Thisindicates that molecular similarity in terms of electrostatic properties is impor-tant for activity, as well as superimposability in terms of molecular shape. Aschematic model of interaction between neonicotinoids and nAChR is proposedaccording to the results of similarity-activity analyses. 1998 SCI.(

Pestic. Sci., 52, 104È110, 1998

Key words : neonicotinoid ; similarity index ; receptor-binding activity ; nicotinicacetylcholine receptor ; molecular orbital calculation ; MNDO-PM3; molecularrecognition

1 INTRODUCTION

A number of neonicotinoid insecticides have been dis-covered as agonists of the nicotinic acetylcholine recep-tor (nAChR).1 We have recently developedacetamiprid2,3 as one of the neonicotinoids, which hasan N-cyanoacetamidine structure as its characteristicfeature. In our previous study to predict the active con-formation of acetamiprid,4 we investigated the molecu-lar similarity between acetamiprid and anotherneonicotinoid, imidacloprid,5 which has an N-nitroguanidine structure. It was found that the molecu-lar similarity indices in terms of steric and electrostaticproperties were helpful to understand the bioisosterismof these two insecticides having di†erent structural fea-

* To whom correspondence should be addressed.

tures. This result suggested that the method of similarityindex may be useful as a quantitative measure of molec-ular similarity to account for the structureÈactivity pro-Ðles of neonicotinoid insecticides. Based on such abackground, we have studied the correlation betweenmolecular similarity and receptor-binding activity of aseries of neonicotinoid insecticides. This paper describesthe result of the correlation analysis, and a model ofmolecular recognition at the receptor will be presentedbased on the structureÈactivity analysis.

2 MATERIALS AND METHODS

2.1 Compounds and biological activity

Among a number of neonicotinoid insecticides, we haveinvestigated the compounds having a general structure

1041998 SCI. Pestic. Sci. 0031-613X/98/$17.50. Printed in Great Britain(

Molecular similarity and receptor-binding in neonicotinoid insecticides 105

Fig. 1. Structures of neonicotinoid insecticides studied in this paper. Structural variation and the representatives of developedinsecticides.

shown in Fig. 1, in which an amine moiety with anelectron-withdrawing group and an aromatic ringmoiety are jointed by a methylene, since most of thewell-known neonicotinoids such as imidacloprid,5acetamiprid2,3 and nitenpyram6 have this structuralfeature, and many analogous compounds have beenreported. The receptor-binding activities of neoni-(Ki)cotinoids having various structures in both structuralmoieties were reported by Tomizawa and Yamamoto.1We have employed the values against nAChR ofKihoney bee from their paper, and the pKi (\[log Ki ;

is in M) was used as the index of biological activityKifor the receptor binding.

2.2 Molecular orbital calculation

Structures of neonicotinoids and their model moleculeswere optimized by semi-empirical molecular orbital cal-culations using MNDO-PM37,8 with MOPACprogram.9 Initial structures of the molecules were con-structed by considering the crystal structure ofimidacloprid10 and the result of conformationalanalysis of acetamiprid.4 Then the energy-minimumstructure of each molecule was obtained by geometryoptimization. The molecular electrostatic potentials ofthe molecules were computed from the ESP atomiccharges derived from electrostatic potentials11 by usingthe ESP option in MOPAC.

2.3 Molecular similarity indices

As quantitative measures of molecular similarity, amethod of similarity indices proposed by Richards etal.12h14 was applied to study the electrostatic and shapesimilarity of two molecules. The electrostatic-similarityindex is deÐned by eqn (1), where and are(RAB)13 eA eBelectrostatic potentials of molecules A and B, respec-tively, at a point outside the two molecules superim-posed. The value of the index varies in the range of [1

to ]1, with indicating perfect similarity. TheRAB\ 1shape-similarity index is deÐned in same manner(SAB)14by eqn (2), where and are the volumes of the indi-TA TBvidual molecules A and B, respectively, and C is thevolume commonly shared by the two molecules at thesuperimposition.

RAB\P

eA eB dqNAP

eA2 dqP

eB2 dqB1@2

([1 ^ RAB ^ 1) (1)

SAB\ C/(TA TB)1@2 (0 ^ SAB^ 1) (2)

In the calculation of the similarity index, theRAB ,values of and were calculated at each grid point ofeA eB

intervals within the distance of from the0É5 Ó 3É0 Óedge of the van der Waals surfaces of each molecule,and the space within the van der Waals surfaces of thetwo molecules was excluded from the calculation. Inthis study, the optimum superimposition of two mol-ecules to maximize the shape-similarity index was(SAB)obtained by simplex optimization procedure,15 then theelectrostatic similarity was computed at the super-(RAB)imposition. The calculations of similarity indices andsimplex optimization were performed on SGI IRIS-4Dworkstation.

3 RESULTS AND DISCUSSION

3.1 Similarity–activity correlation in the amine moiety

A series of compounds having various structures in theamine moiety (Fig. 2) were studied to examine thecorrelation between molecular similarity and activityfor cyclic and acyclic amine structures in neonicotin-oids. The 6-chloro-3-pyridylmethyl group in these com-pounds was replaced by methyl to simplify the model,and the geometry and ESP atomic charges of the modelmolecules were obtained by MNDO-PM3. Among thisseries of compounds, the most active (1) having a nitro-

106 Akira Nakayama, Masayuki Sukekawa

Fig. 2. Structural variation in the amine moiety and superimposition of model molecules by simplex procedure. The 6-chloro-3-pyridylmethyl group is replaced by methyl in the model molecules used in the calculations.

ethene structure was chosen as the reference. The modelmolecules of other compounds were superimposed ontothe model of 1 by simplex optimization procedure tomaximize the shape-similarity index then the elec-(SAB),trostatic similarity was computed at the optimum(RAB)superimposition (Fig. 2, Table 1).

The receptor-binding activity and the similarity(pKi)index of compounds with cyclic amine structures(RAB)(1È9) were plotted in Fig. 3, showing excellent corre-lation (r \ 0É940). Addition of acetamiprid (10), whichhas no cyclic amine structure, also gave signiÐcantcorrelation (r \ 0É915). However, another acyclic amine

TABLE 1Receptor-Binding Activity Similarity Indices and(pKi), (SAB

and Atomic Charge at Amino Nitrogen of CompoundsRAB),Having Various Structures in the Amine Moiety

Similarity indicesb Atomic chargec

Compound pKia S

ABd R

ABe Mullikenf ESPg

1 7É68 1É000 1É000 0É0038 [0É31032 7É55 0É901 0É976 0É0365 [0É32733 6É94 0É929 0É971 0É0102 [0É07144 6É51 0É941 0É957 [0É0149 [0É30815 6É44 0É962 0É983 0É0325 [0É26266 3É79 0É855 0É848 0É0332 [0É21807 5É81 0É948 0É916 0É0353 [0É26408 5É16 0É913 0É919 0É0163 [0É24749 4É75 0É919 0É917 0É0018 [0É3158

10 5É16 0É818 0É869 [0É0873 [0É277111 4É32 0É794 0É967 [0É0274 [0É2458

values were taken from Ref. 1.a Kib Calculated for the model molecules shown in Fig. 2.c At amino nitrogen calculated by MNDO-PM3.d Optimized by simplex procedure.e Calculated for the optimum superimposition to maximizeSAB .f Mulliken atomic charge.g Electrostatic potential derived charge.

derivative, nitenpyram (11), was out of the correlation.Although the reason for the outlying plot for 11 is notclear, we speculate that the contribution of tautomericstructural change in 11 may be a possible reason.According to our NMR studies of a nitenpyram analog,a tautomeric form having an imine structure wasobserved in aqueous solution especially in acidic condi-tions (Fig. 4) (Kawai, T. et al., unpublished). However,the tautomeric structural change in the reference com-pound 1 and an N-nitroguanidine derivative was not

Fig. 3. Correlation between electrostatic similarity and(RAB)receptor-binding activity for the amine moiety. The(pKi)correlation coefficient (r \ 0É940) and the regression line werederived from the data excluding compounds 10 and 11.

Fig. 4. Tautomeric structural change in a nitenpyram analog.

Molecular similarity and receptor-binding in neonicotinoid insecticides 107

Fig. 5. Structural variation in the aromatic ring moiety and superimposition of model molecules by simplex procedure. The modelmolecules, were used in the calculations.Ar-CH3 ,

TABLE 2Receptor-Binding Activity and Similarity Indices(pKi) (SABand of Compounds Having Various Structures in theRAB)

Aromatic Ring Moiety

Similarity indicesb

Compound pKia S

ABc R

ABd

1 7É68 1É000 1É00012 6É33 0É976 0É97613 7É08 0É915 0É98814 4É92 0É886 0É28615 3É87 0É882 0É54516 5É93 0É971 0É65217 5É61 0É896 0É56718 7É09 0É906 0É939

values were taken from Ref. 1.a Kib Calculated for the model molecules shown in Fig. 5.c Optimized by simplex procedure.d Calculated for the optimum superimposition to maximizeSAB .

Fig. 6. Correlation between electrostatic similarity and(RAB)receptor-binding activity for the aromatic ring moiety.(pKi)The correlation coefficient (r \ 0É919) and the regression linewere derived from the data excluding compound 15.

observed by NMR under similar conditions. The imineform of 11 is less similar to compound 1 (RAB\ 0É647)than the nitroethene structure, so that the receptor-binding activity of 11 may be reduced by taking such atautomeric form in part.

3.2 Similarity–activity correlation in the aromatic ringmoiety

A number of compounds having a variety of aromaticring systems (Fig. 5) have been reported as neonicotin-oids, among which the 6-chloro-3-pyridyl derivative (1)shows the most potent activity. SimpliÐed model mol-ecules of the aromatic ring moiety, were sub-Ar-CH3 ,jected to the molecular similarity analysis, and themodel of the most active compound (1) was chosen asthe reference to which other molecules were superim-posed to maximize the shape similarity The values(SAB).of similarity indices and obtained for the(SAB RAB)optimum superimposition are shown in Table 2. A sig-niÐcant correlation between the activity and the(pKi)electrostatic similarity was observed as shown in(RAB)Fig. 6, in which the 4-pyridyl derivative was an unex-pected outlier. Among the pyridyl derivatives, 3-pyridylcompounds (1, 12 and 13) exhibited potent activity,whereas 2- and 4-pyridyl derivatives (14 and 15) weremuch less active. We speculate that the direction of elec-

Fig. 7. Superimposition of the model molecules of 2-chloro-5-thiazolyl (solid lines) and 6-chloro-3-pyridyl (broken lines)

derivatives to maximize the shape similarity index (SAB).

108 Akira Nakayama, Masayuki Sukekawa

Fig. 8. Superimposition of acetamiprid (10) (solid lines) ontothe reference molecule (1) (broken lines) to maximize the shape

similarity index (SAB).

trostatic or hydrogen-bonding interaction at the aro-matic ring moiety may play an important role for themolecular recognition at the receptor. Such position-speciÐc interactions at the pyridyl nitrogen may be pre-dominant for the receptor binding. The similarity index,

which takes account of the distribution of electro-RAB ,static potentials around whole molecules, might notalways be sufficient as a quantitative measure of theposition-speciÐc interaction.

2-Chloro-5-thiazolyl derivatives, such as compound18, have been known as active analogs of the 6-chloro-3-pyridyl derivative exhibiting potent activity. The

Fig. 9. Correlation between electrostatic similarity and(RAB)receptor-binding activity for the whole molecule. The(pKi)correlation coefficient (r \ 0É941) and the regression line werederived from the data excluding compounds 11 and 15.

TABLE 3Receptor-Binding Activity and Similarity Indices(pKi) (SABand of Compounds Having Various Structures both inRAB)

the Amine and the Aromatic Ring Moieties

Similarity indicesb

Compound pKia S

ABc R

ABd

1 7É68 1É000 1É0002 7É55 0É941 0É9873 6É94 0É952 0É9784 6É51 0É977 0É9495 6É44 0É876 0É9476 3É79 0É886 0É8077 5É81 0É928 0É9048 5É16 0É926 0É8969 4É75 0É932 0É891

10 5É16 0É871 0É89811 4É32 0É833 0É92812 6É33 0É983 0É99113 7É08 0É956 0É99614 4É92 0É917 0É83815 3É87 0É932 0É89116 5É93 0É982 0É91717 5É61 0É942 0É90318 7É09 0É938 0É97819e 5É14 0É893 0É905

values were taken from Ref. 1.a Kib Calculated for the whole molecules.c Optimized by simplex procedure.d Calculated for the optimum superimposition to maximizeSAB .

e

optimum superimposition of the model molecules of 2-chloro-5-thiazolyl and 6-chloro-3-pyridyl derivatives isshown in Fig. 7, through which the molecular similarityof these two molecules may(RAB\ 0É939, SAB\ 0É906)be understood.

3.3 Similarity–activity correlation in the whole molecule

The molecular similarity in the whole molecule was alsostudied for compounds with variation both in amineand aromatic ring moieties. Structures of each moleculewere fully optimized by MNDO-PM3, and then super-imposed onto the reference molecule (1) by simplex pro-cedure to maximize the shape similarity The(SAB).optimum superimposition of acetamiprid (10) onto thereference (1) is shown in Fig. 8 as an example. The elec-trostatic similarity obtained for the optimum(RAB)superimposition and the receptor-binding activity (pKi)are listed in Table 3, and plotted in Fig. 9. Equation (3)was obtained as a regression equation for the data in

Molecular similarity and receptor-binding in neonicotinoid insecticides 109

Table 3, indicating the signiÐcant correlation betweenand computed for the whole molecules. In thispKi RAB

equation, n represents the number of compounds, r thecorrelation coefficient, s the standard deviation and Fthe ratio of regression and residual variances. The Ðgurein parentheses is the 95% conÐdence interval.

pKi\ 18É16 (^3É61)RAB [ 10É87

n \ 17, r \ 0É941, s \ 0É382, F1, 15 \ 115É112 (3)

The above equation was obtained for the data setexcluding compounds 11 and 15 which were outlyingfrom each regression line in the analyses for amine andaromatic ring moieties, respectively. The regressioncoefficient (r \ 0É864) for the data including 11 and 15was considerably lower than in eqn (3). The plots forthese two compounds in Fig. 9 deviate from the regres-sion line as well as in Figs 3 and 6, reÑecting speciÐcstructural features, such as tautomeric structural changeand the position of pyridyl nitrogen.

3.4 Model of molecular recognition at nAChR

As a model of receptor-binding of neonicotinoids,Tomizawa and Yamamoto postulated that the aminonitrogen which is positively charged by an electron-withdrawing group plays the role of the cationic nitro-gen in protonated nicotine at nAChR.1 According totheir hypothesis, we have examined the correlation

between the atomic charge of the amino nitrogen andthe receptor-binding activity. Mulliken atomic chargeand ESP charge at the amino nitrogen in the modelmolecules of compounds 1È11 were obtained byMNDO-PM3 (Table 1). There was, however, no signiÐ-cant correlation between the atomic charges and activ-ity. This suggests that the point charge on the nitrogenis not, on its own, sufficient to account for the receptor-binding activity.

On the other hand, the electrostatic similarity of mol-ecules was signiÐcantly correlated with the receptor-binding activity, when each molecule was superimposedonto the most active one so as to maximize the molecu-lar shape similarity. This indicates that the molecularsimilarity in terms of electrostatic properties is impor-tant for the activity as well as the superimposability interms of molecular shape. The molecular electrostaticpotentials of compound 1 are shown in Fig. 10, in whichnegative electrostatic potentials are distributed aroundthe nitro group in the amine moiety and positive poten-tials are on the opposite side. The similarity in such adistribution of electrostatic potentials was quantitat-ively evaluated by the similarity index, and the indexwas found to correlate signiÐcantly with the receptor-binding activity. The negative potentials around thepyridyl nitrogen in the aromatic ring moiety are alsocharacteristic, and the position-speciÐc interactionbetween the pyridyl nitrogen and the receptor seems toplay a signiÐcant role for the binding. We thereforepropose an alternative model where the electrostatic

Fig. 10. Molecular electrostatic potentials of the cyclic amine moiety (right) and the pyridine ring (left) in compound 1. Contourvalues are in unit of 10~3 au.

110 Akira Nakayama, Masayuki Sukekawa

Fig. 11. A hypothetical model of molecular recognition ofneonicotinoid insecticides at nAChR.

interaction is vital for molecular recognition at nAChR,as shown in Fig. 11.

As well as the electrostatic properties, the molecularshape similarity should be important for molecularrecognition, since the optimum superimposition of mol-ecules in terms of molecular shape gave a signiÐcantelectrostatic parameter in the similarity-activityanalysis.

4 CONCLUSION

As described above, the receptor-binding activity ofvarious neonicotinoid insecticides was found to correl-ate quantitatively with the molecular similarity index.This indicates that the method of similarity index ishelpful not only to understand the bioisosterism butalso to provide useful descriptors in structureÈactivityanalyses concerning molecular similarity. A schematicmodel of interaction between neonicotinoids andnAChR was drawn according to the result of similarityÈactivity analyses for a series of compounds shown inFig. 1. Further studies including other nAChR agonistssuch as nicotine and its analogs are of interest, and bio-logical and structural information of nAChR itselfshould also be expected to emerge to help furtherunderstanding of binding mechanism of the insecticides.

ACKNOWLEDGEMENTS

We are grateful to Dr W. G. Richards (OxfordUniversity) for providing us with invaluable informa-tion of molecular similarity analysis. We appreciate the

excellent data on biological activities measured andpublished by Dr M. Tomizawa and Professor I. Yama-moto (Tokyo University of Agriculture). Thanks arealso given to many colleagues of Nippon SodaCompany for their support to this work.

REFERENCES

1. Tomizawa, M. & Yamamoto, I., StructureÈactivityrelationships of nicotinoids and imidacloprid analogs.Nihon Noyaku Gakkaishi (J. Pestic. Sci.), 18 (1993) 91È8.

2. Hopkins, W. L., Ag Chem New Compound Review, Vol. 13.Ag Chem Information Services, Indianapolis, 1995, pp. 59.

3. Nippon Soda Co., Ltd, Amine derivatives. PCT WO 91/04965, 1991.

4. Nakayama, A., Sukekawa, M. & Eguchi, Y., Stereoche-mistry and active conformation of a novel insecticide, ace-tamiprid. Pestic. Sci., in press.

5. Shiokawa, K., Tsuboi, S., Iwaya, K. & Moriya, K., Devel-opment of a chloronicotinyl insecticide, imidacloprid.Nihon Noyaku Gakkaishi (J. Pestic. Sci.), 19 (1994) 329È32.

6. Minamida, I., Iwanaga, K., Tabuchi, T., Aoki, I., Fusaka,T., Ishizuka, H. & Okauchi, T., Synthesis and insecticidalactivity of acyclic nitroethene compounds containing aheteroarylmethylamino group. Nihon Noyaku Gakkaishi(J. Pestic. Sci.), 18 (1993) 41È8.

7. Stewart, J. J. P., Optimization of parameters for semi-empirical methods I. Method. J. Comp. Chem., 10 (1989)209È20.

8. Stewart, J. J. P., Optimization of parameters for semi-empirical methods II. Applications. J. Comp. Chem., 10(1989) 221È64.

9. Stewart, J. J. P., MOPAC Ver.6, Quantum ChemistryProgram Exchange, Program 455, Indiana University,Bloomington, IN.

10. Born, L., The molecular and crystal structure of imid-acloprid (phase 2). PÑanzenschutz-Nachr. Bayer, 44 (1991)137È44.

11. Besler, B. H., Merz, Jr, K. M. & Kollman, P. A., Atomiccharges derived from semi-empirical method. J. Comp.Chem., 11 (1990) 431È9.

12. Richards, W. G. & Hodgkin, E. E., Molecular similarity.Chemistry in Britain 24 (1988) 1141È4.

13. Hodgkin, E. E. & Richards, W. G., Molecular similaritybased on electrostatic potential and electric Ðeld. Int. J.Quantum Chem. Quantum Biol. Symp., 14 (1987) 105È10.

14. Meyer, A. Y. & Richards, W. G., Similarity of molecularshape. J. Comput. Aided Mol. Design, 5 (1991) 427È39.

15. Nelder, J. A. & Mead, R., A simplex method for functionminimization. Comput. J., 7 (1965) 308È13.