quantitative structure–activity relationships (qsar) of cinnamic acid bird repellents

21
Journal of Chemical Ecology, Vol. 25, No. 12, 1999 QUANTITATIVE STRUCTURE-ACTIVITY RELATIONSHIPS (QSAR) OF CINNAMIC ACID BIRD REPELLENTS R. W. WATKINS,* ,1 J. A. LUMLEY, 2 E. L. GILL, 1 J. D. BISHOP, 1 S. D. LANGTON, 1 A. D. MAcNICOLL, 1 N. R. PRICE, 1 and M. G. B. DREW 2 1 Central Science Laboratory Sand Hutton, York, YO41 1LZ, UK 2 Chemistry Department University of Reading Whiteknights, Reading, RG6 6AD, UK (Received August 17, 1998; accepted August 15, 1999) Abstract—Plants have evolved an array of defense chemicals that inhibit the feeding of vertebrate herbivores and therefore have potential for agricultural and environmental applications to reduce feeding damage. We investigated the relationship between structure and repellency for 14 derivatives of the plant secondary compound, cinnamic acid, using the feral pigeon (Columba livia) as the test species. The mechanism behind the repellent activity of these derivatives is explained by a combination of four descriptors: heat of formation (DH f ), polarizability (XY and YY) and superdelocalizability (Sr). All these parameters are electronic, indicating that changes in electronic distribution within cinnamic acid structures are crucial for activity. This is the first published quantitative structure-activity model for avian repellents, and as a result we can now begin to predict which cinnamic acid derivatives should make effective repellents. The full power of this model to aid the selection and screening of new repellents awaits further experimentation on both related compounds and other avian species. However, this modeling approach promises to provide a more efficient and economic method for prospecting chemical databases for new effective bird repellents. Key Words—Cinnamic acids, bird repellents, quantitative structure-activity relationship, dose-response, pigeon, Columba livia, crop protection, feeding deterrent. *To whom correspondence should be addressed. 2825 0098-0331/99/1200-2825/$16.00/0 © 1999 Plenum Publishing Corporation

Upload: r-w-watkins

Post on 02-Aug-2016

216 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

Journal of Chemical Ecology, Vol. 25, No. 12, 1999

QUANTITATIVE STRUCTURE-ACTIVITY RELATIONSHIPS(QSAR) OF CINNAMIC ACID BIRD REPELLENTS

R. W. WATKINS,*,1 J. A. LUMLEY,2 E. L. GILL,1 J. D. BISHOP,1

S. D. LANGTON,1 A. D. MAcNICOLL,1 N. R. PRICE,1

and M. G. B. DREW2

1 Central Science LaboratorySand Hutton, York, YO41 1LZ, UK

2 Chemistry DepartmentUniversity of Reading

Whiteknights, Reading, RG6 6AD, UK

(Received August 17, 1998; accepted August 15, 1999)

Abstract—Plants have evolved an array of defense chemicals that inhibit thefeeding of vertebrate herbivores and therefore have potential for agriculturaland environmental applications to reduce feeding damage. We investigatedthe relationship between structure and repellency for 14 derivatives of theplant secondary compound, cinnamic acid, using the feral pigeon (Columbalivia) as the test species. The mechanism behind the repellent activity ofthese derivatives is explained by a combination of four descriptors: heatof formation (DHf), polarizability (XY and YY) and superdelocalizability(Sr). All these parameters are electronic, indicating that changes in electronicdistribution within cinnamic acid structures are crucial for activity. This isthe first published quantitative structure-activity model for avian repellents,and as a result we can now begin to predict which cinnamic acid derivativesshould make effective repellents. The full power of this model to aid theselection and screening of new repellents awaits further experimentation onboth related compounds and other avian species. However, this modelingapproach promises to provide a more efficient and economic method forprospecting chemical databases for new effective bird repellents.

Key Words—Cinnamic acids, bird repellents, quantitative structure-activityrelationship, dose-response, pigeon, Columba livia, crop protection, feedingdeterrent.

*To whom correspondence should be addressed.

2825

0098-0331/99/1200-2825/$16.00/0 © 1999 Plenum Publishing Corporation

Page 2: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

INTRODUCTION

The use of chemical repellents has considerable potential as a relatively benignmethod of reducing the damage caused by pest species. However, few effec-tive repellents are currently available for vertebrate pests. This is perhaps dueto the largely empirical strategies used for the selection of chemicals as repel-lents. Screening programs were often targeted at compounds already registeredfor other agricultural or pharmaceutical applications (Woronecki et al., 1981;Brooks and Rowe, 1987; Avery and Decker, 1991). In practice, the effect ofmany such chemicals was found to be short-lived, or the compound was toxicand therefore not widely used.

More recently research has focused on identifying compounds that are bio-logically significant to the particular pest we wish to repel (Watkins et al.,1996a). Secondary plant substances (e.g., phenylpropanoids, phenols, and ter-penoids) have a profound effect on food-plant selection by many herbivores.The presence of such chemicals can serve as a warning that the plant is unpalat-able and encourages herbivores to feed on other plant species. The response tothese compounds is mediated by the chemical senses (i.e., taste, olfaction, andtrigeminal chemoreception) and/or postingestional effects (i.e., gastrointestinalmalaise). The aversive response to some chemical repellents is innate, a propertythat is the result of past evolutionary pressures to develop sensitivity to partic-ular odors or tastes. Foods that are toxic often taste bitter or cause irritation tothe buccal cavity, although this relationship is by no means perfect. For exam-ple, mammals show aversive orofacial responses to quinine and capsaicin (thepungent principle of chili peppers) despite having no prior experiences of thesecompounds (Chambers and Bernstein, 1995). Experience can also play a criti-cal role in the response to a repellent. An initial preference for treated food isreversed when the post-ingestional consequences of eating the food are negative.The compound causes some form of transient upper gastrointestinal discomfortor illness such as nausea or vomiting, which the individual then associates withthe taste of the compound or, if the compound has no taste, another salient cuewithin the food (Provenza, 1995). The animal then becomes conditioned to avoidthat cue in future encounters. In agriculture, this latter type of repellency hasbeen successfully used to train livestock to avoid certain plant species (Burritand Provenza, 1990).

These compounds, because of their "evolved" defensive function, are wellsuited for agricultural and environmental applications to reduce feeding dam-age (Mason and Clark, 1992; Jakubas et al., 1992). A small number are alreadyundergoing commercial evaluation in applications to prevent bird damage tocrops (Cummings et al., 1995; Gill et al., 1994, 1998), inhibit nontarget wildlifefrom consuming potentially toxic granular pesticides and chemically treated

2826 WATKINS ET AL.

Page 3: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

seeds (Mason et al., 1993; Watkins et al., 1996b), and prevent gnawing damageto electrical cables by rodents (Kurata et al., 1994).

To date the discovery of chemical repellents has been more by chance thanby design. Such chance discovery has limited the number of economically viableand environmentally safe repellents. Each step in the development of a usefulrepellent is a sifting process with technical, safety, and economic factors all play-ing a role in determining the fate of any candidate compound (Mason and Clark,1992). Thus, if the initial number of candidate compounds is few, the likelihoodof any compound passing through each of the above sieves is also likely to besmall. In order to boost the number of candidate compounds quickly and effi-ciently, we and other research groups have been seeking to generate rule-basedsystems by which we can select or reject a compound as a repellent based simplyon an examination of its chemical structure.

Qualitative models have recently been developed that can predict the repel-lent activity of aromatic, plant-derived compounds for both birds and mammals(Clark and Shah, 1994; Nolte et al., 1993). For avian repellents, these models(Shah et al., 1991; Jakubas et al., 1992; Clark and Shah, 1994) predict that themost critical features, at least for certain classes of compound (e.g., anthrani-lates, acetophenones, benzoates, vanilates, and cinnamic acid derivatives) are thepresence of an electron-rich phenyl ring with electron withdrawing and donat-ing groups that are held within a resonant substitution pattern (Clark and Shah,1994).

The current study set out to build on this previous work and generate aquantitative model with which to describe the relationship between the struc-ture and the activity of bird repellents. It focused on a series of cinnamic acidderivatives for which we had evidence of repellency against birds, mammals, andinvertebrates (Crocker and Reid, 1993; Gurney et al., 1996; Mosson et al., 1996).Metabolites of cinnamic acid, isolated from the flower buds of a bird-resistantvariety of pear tree, have been found to deter birds from feeding (Crocker andPerry, 1990). When these compounds and their derivatives were presented indi-vidually to captive feral pigeons (Columba livia Gmelin) in laboratory no-choicefeeding trials, several proved to be effective feeding deterrents (Crocker et al.,1993; Watkins et al., 1995) that could offer a large margin of safety in termsof their low toxicity (Hoskins, 1984). These studies also showed that despitethe superficial similarity of their structures, the cinnamic acid derivatives dif-fered greatly in their repellency. For example, birds avoided food treated with3,5-dimethoxycinnamic acid and yet were indifferent to 3,4-dimethoxycinnamicacid (Crocker and Perry, 1990). A simple explanation of these differences, how-ever, was not apparent. This paper reports on a subsequent study whose aimwas to establish the dose-response relationships of these compounds and, usingcomputer-aided software, to explore the relationship between their efficacy andkey topological and electronic parameters.

CINNAMIC ACID BIRD REPELLENTS 2827

Page 4: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

METHODS AND MATERIALS

Chemicals. The trans isomers of 2-methoxycinnamic acid (C.A.S. 6099-03-2), 3-methoxycinnamic acid (C.A.S. 6099-04-3), 4-methoxycinnamic acid(C.A.S. 830-09-1), 3,5-dimethoxycinnamic acid (C.A.S. 1132-21-4), 3,4-hydroxycinnamic acid (C.A.S. 331-39-5), 3-methoxy-4-hydroxycinnamic acid(C.A.S. 537-73-5), 3,5-dimethoxy-4-hydroxycinnamic acid (C.A.S. 530-59-6),3-chlorocinnamic acid (C.A.S. 1866-38-2), 3-nitrocinnamic acid (C.A.S. 555-68-0), cinnamic acid (C.A.S. 140-10-3), cinnamamide (C.A.S. 621-78-1), cinnamylalcohol (104-54-1), cinnamaldehyde (C.A.S. 14371-10-9), and 3,5-dimethoxy-benzoic acid (C.A.S. 1132-21-4) were supplied by Aldrich Chemical Company(Gillingham, Dorset, UK). Solvents (HPLC grade) were supplied by Rathburn(Walkerburn, Peebleshire, UK) and potassium dihydrogen orthophosphate byMerck Chemical Company (Lutterworth, Leicester, UK).

Sample Preparation. Chicken layer pellets (Lillico & Son, Reigate, Surrey,UK) were used as the coating substrate. These pellets are readily accepted bycaptive feral pigeons, provide a balanced diet, are reasonably uniform in sizeand composition, and have sufficient surface area so that an adequate coatingof chemical can be deposited on the surface. Batches of sieved pellets (500 g)were mixed using an industrial food mixer and sprayed with the appropriateamount of compound dissolved in methanol + water (4:1 by volume, 175 ml).The concentration ranges to be tested were determined, where possible, fromprevious feeding studies (Crocker et al., 1993). This mixing process producedan evenly coated batch, with little or no damage to the pellets and no excesssolvent. The pellets were air dried on a stainless steel tray and then sieved toremove any debris. Pilot trials had shown that the application of the methanol+ water solvent alone did not affect the palatability of the pellets.

Sample Extraction. Eight samples (0.5 g) from each treatment preparationwere ground to a fine powder and suspended in 20 ml of methanol + watersolvent (4:1 by volume). The analyte was extracted by sonicating the sample for30 min in a sealed centrifuge tube. After extraction, the sample was centrifuged(10,000 g, 20 min), and the supernatant stored at 4°C. This process was repeatedto ensure complete extraction, the supernatants were pooled, and the total volumemade up to 50 ml.

Sample Analysis. A Waters-Millipore HPLC system was used to separate andquantify the analyte. The system including a WISP autosampler and a Lambda-Max 480 UV detector. A prepacked analytical column (250 x 4.6 mm) and guardcolumn (10 x 3.2 mm) of Kromasil RP-18 (5 um) were employed for the sepa-ration. The mobile phase consisted of: (A) potassium dihydrogen orthophosphatebuffer (pH 7.0, 20 mM), and (B) methanol. The elution profile was: 0–2 min 20%B; 2-24 min, 20-90% B; 24–28 min, 90% B; 28-29 min, 90–20% B; 29-32 min,20% B, at a flow rate of 1.0 ml/min, column temperature of 35°C, and a col-

2828 WATKINS ET AL.

Page 5: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

were weighed before and after the pretrial period. Between the pretrial and trialperiods, the birds received chicken layer pellets ad libitum.

Trial (Short-Term No-Choice Test). The trial with treated food was carriedout on three consecutive days in the week following the pretrial period. Fivebirds were randomly assigned to each treatment and received the same treat-ment on each day. The birds received 30 g of treated food at 08:30 hr and con-

CINNAMIC ACID BIRD REPELLENTS 2829

sumption was measured at 1 -hr intervals between 08:30 and 17:00 hr. Thisprocedure allowed patterns of food consumption to be monitored throughout theday and reduced the possibility of learned association between change of thefood bowl or diet regimen. At 17:00 hr the treated food was replaced with 60 gof untreated food, which was removed at 08:30 hr the following morning. The12-hr light–dark regime ensured that the birds had not fed or drunk for 12 hrbefore they received the treated food (feral pigeons will not feed in the dark; A.Dawson, 1993, personal communication). The birds were weighed before andafter the three-day trial.

Throughout the trial and pretrial periods, equivalent measurements weretaken from a bowl of chicken layer pellets (30 or 60 g, where appropriate) keptwithin the same room. This controlled for the effects of temperature and humid-ity on the weight of food.

Data Analyses: Food Consumption. The percentage change in consump-tion between the three-day pretrial and trial periods was calculated for eachindividual bird. A polynomial regression technique was then used to describethe dose-response relationship for each of the compounds under test. The bestfitting model was determined by analyzing all 14 compounds together, fitting

1 -hr intervals between 08:30 and 17:00 hr and after 24 hr (08:30 hr). The birds

umn pressure of 14–20 MPa. Standards were weighed and dissolved in methanol +water (4:1 by volume) to give various concentrations within the range 0.0125–0.1mg/ml. Calibration curves were plotted based on the linear regression analysis ofpeak area. All samples were filtered through a Millipore filter (FH, 0.45 um). Sam-ple filtrate (20 ul) was then injected onto the column and the concentration of thetest compound determined from its calibration curve.

Pigeons. Feral pigeons were individually housed indoors at 16°C on a 12-hr light–dark cycle (08:30-20:30 hr). They were visually but not aurally isolatedand were supplied with water, grit (Health Grit, Liverine, Grimsby, UK) and food(chicken layer pellets) ad libitum. They were weighed on entry to the room andthree times a week for two to three weeks until the beginning of the experimen-tal period. The experiment was not started until the birds maintained a stable bodyweight and any bird that lost more than 15% of its entry weight was excluded fromthe test. Throughout the experimental period the birds received water and grit adlibitum.

Pretrial Period. The birds were provided with chicken layer pellets (60 g)at 08:30 hr on three consecutive days. Their food consumption was measured at

Page 6: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

polynomial terms, and allowing the parameters to vary for each compound. Therelationship was best described by a second-order polynomial (improvement overlinear fit, F = 49.79, df 1, 392, P < 0.001). Polynomial curves were used becausethey had the flexibility to fit all the compounds studied; more complex nonlinearcurves, such as logistic curves, fitted some compounds better but did not givean acceptable fit with others.

From this regression line the R50 value (Starr et al., 1964; Watkins et al.,1995), analogous to the LD50 values used in toxicological studies and definedas the concentration of chemical required to reduce mean food consumption by50%, was calculated. Confidence limits for the R50 values were determined usinga bootstrapping technique (Efron and Tibshirani, 1993). In all cases, Tukey'sHonestly Significant Difference tests were used to make pairwise comparisonsand isolate significant differences (P < 0.05) between R50 values.

Molecular Modeling Procedure. Initial modeling of all structures and con-formational analysis was carried out using CHEMX (Chemical Design Ltd.,Chipping Norton, UK). Final conformational analysis was completed on the 14compounds using Quanta96/CHARm (Molecular Simulations Inc., San Diego,California) to identify the correct lowest energy conformation of each com-pound. The bond angles, lengths, and torsions were also examined in the Cam-bridge Crystallographic Database (Cambridge Crystallographic Data Centre,Cambridge, UK). A grid search conformational analysis was carried out alongwith the four main torsion angles of the propenoic acid side chain as variables.In the minimum energy confrontation, the propenoic side chain was at a tor-sion angle of 19° to the aromatic ring (a compromise between conjugation andsteric effects) and the polar hydrogen was found to extend out from the structure,synperiplanar with the carbonyl oxygen.

The bond angles, lengths, and torsions were then checked against valuesin crystal structures of cinnamic acid derivatives from the Cambridge Crystal-lographic Database. Fifty-five cinnamic acid derivatives with no substituents onthe propenoic side chain were located, but the torsion angle between the aromaticring and the propenoic acid side chain ranged between 1 and 8°. It is clear thatthe CHARm force field in our case has overestimated the steric effects by com-parison with the conjugation in these bonds. For this reason, these torsion angleswere adjusted to the mean experiment value of 3.6°, but all other dimensionswere taken from the conformational analysis.

The molecular descriptors were generated from CHEMX, ClogP, Cerius2QSAR descriptor and the TSAR software package (Oxford Molecular Ltd.,Oxford, UK) with the Vamp QM calculation module (a MOPAC-type programusing AM1 parameterization). The AM1 (Austin model 1) method (Dewar et al.,1985) is based on NDDO and MNDO, but the overall accuracy is considerablyimproved. The later PM3 method (Stewart, 1989) was not used, as it is thoughtthat AM1 is more accurate for the generation of electronic charges (Dewar et al.,

2830 WATKINS ET AL.

Page 7: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

Smith et al., 1992). In total, 110 independent descriptors were generated, about30% of which were semiempirical molecular properties.

As many of the properties calculated had very different values or ranges,standardization was carried out to allow statistically meaningful analysis of thedata, giving comparable weight to all the parameters involved. This is particularlyimportant for partial least squares or principal component analysis. A correlationmatrix with all descriptors was then constructed to identify any cross-correlatedparameters. When two descriptors had a correlation of r2 > 0.70, then the parame-ter that had a lower correlation with the log (R50) values (dependent variable) wasomitted. This process reduced the number of independent descriptors to 41.

Data Analysis: Structure-Activity Relationships. Various statistical analysismethods were employed to find a suitable descriptive equation, including multipleregression, partial least squares, and principal component analysis. A number ofpredictive equations were selected by these techniques, but all failed to give accept-able cross-validation indices [i.e., r(CV)2; > 0.8]. The problem remained that therewere too many possible combinations of independent variables to select manuallyan appropriate set for accurate prediction. Therefore, the genetic algorithm pack-age in Cerius 3.0 was employed (Molecular Simulations Inc.). This method pro-vides a fast and effective technique by which combinations of independent vari-ables can be matched and predictive equations generated. The technique uses agenetic function algorithm (GFA), which provides the user with multiple mod-els, not only linear polynomials but also higher order polynomials, splines, andGaussians. However, as we were able to obtain high-quality equations with lin-ear polynomials, it was not necessary to use the higher order equations. Use oflinear equations also has the advantage that the equations are easier to interpretin a chemical sense. In the genetic algorithm technique an initial population ofequations is generated with random bias. These are assessed on the basis of a fit-ness function, the best equations are selected and then mutated or combined withothers via the genetic crossover procedure, and those with the best fitness func-tion are retained and the process is repeated. Gradually over many mutations andcrossovers, the best 100 equations are selected and reported together with their cal-culated correlation and cross-correlation coefficients. There are many choices ofgenetic algorithm defaults in the software and the most successful approach wasto continually change the number of crossovers, the mutation probabilities, and/orthe number of terms being added or removed from each equation. Generally around5000-10,000 crossovers and a 50% mutation probability was used.

RESULTS

Dose-Response Relationships. The dose-response relationships were deter-mined for 14 cinnamic acid derivatives (Figures 1 and 2). However, only 13 R50

values could be determined since the value for 3-methoxy-4-hydroxycinnamic

CINNAMIC ACID BIRD REPELLENTS 2831

Page 8: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

FIG. 1. Chemical structure of candidate cinnamic acid and derivative avian repellents.

2832 WATKINS ET AL.

Page 9: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

CINNAMIC ACID BIRD REPELLENTS 2833

FIG. 2. Dose-response curves for the 14 cinnamic acid derivatives over the three-day,short-term, no-choice trial. The relationships between dose and response were describedby second order polynomials. Solid dots depict mean consumption over three daysexpressed as a percentage of pretrial (normal) consumption; (a) cinnamic acid, (b) cin-namyl alcohol, (c) cinnamaldehyde, (d) cinnamamide, (e) 2-methoxycinnamic acid, (f) 3-methoxycinnamic acid, (g) 4-methoxycinnamic acid, (h) 3,5-dimethoxycinnamic acid, (i)3-nitrocinnamic acid, (j) 3-chlorocinnamic acid, (k) 3-methoxy-4-hydroxycinnamic acid(ferulic acid), (1) 3,5-dimethoxy-4-hydroxycinnamic acid (sinapic acid), (m) 3,4-dihydro-xycinnamic acid (caffeic acid), and (n) 3,5-dimethoxybenzoic acid.

acid (caffeic acid) could not be interpolated from the fitted curve without extrap-olating beyond the observed doses (Table 1).

Post-hoc comparisons revealed that cinnamaldehyde was significantly moreeffective (P < 0.05), as measured by its R50 value, than all the other cinnamicacid derivatives with the exception of cinnamyl alcohol. Replacement of the car-boxylic group in cinnamic acid with the aldehyde, amide and alcohol moieties ofcinnamaldehyde, cinnamamide, and cinnamyl alcohol respectively, significantlyincreased repellency (P < 0.05; Table 2). The presence of electron-withdraw-

Page 10: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

2834 WATKINS ET AL.

ing and electron-donating groups increases repellency with the meta-substituted(electron withdrawing) 3-methoxy-, 3-nitro-, and 3-chlorocinnamic acids beingsignificantly more effective (P < 0.05) than the para-substituted (electron-donat-ing) 4-methoxy-cinnamic acid.

Structure-Activity Relationships. The genetic function algorithm producedmany linear regression models, but only one gave r2 > 0.90 with relatively inde-pendent parameters, and this is reported below (Figure 3). This equation, whichuses four electronic parameters [polarizability XY and YY, heat of formation(DHf), and superdelocalizability (Sr)] had an r2 of 0.980.

log(R50) = 2.1305 + 0.1200(polarizability XY) - 0.0296(polarizability_YY)

- 0.0057(DHf) - 0.1498(Sr)

The predictive power of this equation was estimated using a cross-validationtechnique, the model being validated with up to three data points having been

FIG. 2. Continued.

Page 11: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

CINNAMIC ACID BIRD REPELLENTS 2835

FIG. 2. Continued.

removed in each run, and the final value of r(CV)2 was 0.938. Further validationof the predictive equation was undertaken by again using the genetic functionalgorithm to generate regression models on the original data, but after random-izing the R50 values relative to the independent descriptors. This was done inorder to assess the tendency for the algorithm to detect spurious relationships inthe data (Manly, 1991). Ten calculation runs were made with this randomizeddata. Poorly validated regression equations were generated; all resulting modelsproduced poor cross-validation indices with r(CV)2 < 0.6, in many cases < 0.2.

DISCUSSION

The results of any repellent study will be dependent, in part, upon the exper-imental paradigm selected. The standard technique used to evaluate repellencyis to quantify food or water consumption in a choice or no-choice test. Thesetwo paradigms are likely to yield slightly different results: in the choice test the

Page 12: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

2836 WATKINS ET AL.

FIG. 2. Continued.

animal is presented with the choice between a palatable and unpalatable food,while in the no-choice test the animal can either consume the unpalatable foodor go hungry. Thus the no-choice test is a more conservative measure of repel-lency and should therefore provide assurance as to its efficacy in field conditionswhere alternative foods are scarce. The current study was based on a short-termno-choice test carried out over three consecutive days. This test was designedto lessen the effects of food deprivation normally associated with the no-choicetest, such as a reduction in food selectivity (Mason et al., 1989), but provide amore rigorous measure of repellency than that of the two-choice test.

The R50 value was interpolated from the dose-response relationships inorder to categorize the potency of the cinnamic acid derivatives. The R50 val-ues, analogous to the LD50 values used in toxicological studies, provide a use-ful and widely recognized single parameter estimate for comparison of repellentpotencies in structure-activity studies (Schafer and Jacobson, 1983; Schafer etal., 1998). However, it should be noted that this value can only give a partialdescription of a compound's activity: an assessment of the slope, or sensitivity to

Page 13: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

CINNAMIC ACID BIRD REPELLENTS 2837

changes in concentration and the concentration at which maximum repellency isobserved, can provide a fuller description of the compound's repellent properties(Clark and Shah, 1994).

From the inspection of the R50 values for the test compounds, two con-clusions can be inferred. First and most significantly, the replacement of the—COOH group with other functionalities increases repellency: cinnamalde-hyde > cinnamamide > cinnamyl alcohol > cinnamic acid. Second, that repel-lency is increased by the presence of electron-donating and electron-withdrawinggroups, with the meta-substituted electron-withdrawing groups (e.g., 3-methoxy,3-chloro, and the 3-nitroxy acids) playing the more significant role. Any quantita-tive regression model derived should therefore reflect these simple relationships.

Using the genetic algorithm, we have generated a quantitative model thatcan explain the structure-activity relationship for this series of cinnamic acidderivatives. The power of the model has been confirmed using both cross-vali-dation and randomization tests. Further confirmation of the predictive power ofthis model will await analysis against an independent data set. In our main pre-

FIG. 2. Continued.

Page 14: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

2838 WATKINS ET AL.

diction equation the mechanisms behind the repellent activity are explained bya combination of four descriptors: heat of formation (DHf), polarizability (XYand YY) and superdelocalizability (Sr). It is noteworthy that all these parametersare electronic, indicating that changes in electronic distribution within cinnamicacid structures are crucial for activity.

The first parameter, the heat of formation, describes the sum of four energystates: the energy required to ionize the valence electrons of the atoms involved,Ee l(A); the heat of atomization, DH f(A); electronic energy; and nuclear energy.This descriptor reflects the increasing number of electron-withdrawing groupsbeing added to the cinnamic acid parent structure: as more electron withdraw-ing groups are added to the ring, there are more atoms contributing to the energyterms and consequently the heat of formation increases. Although this relation-ship remains valid for our small set of compounds, it cannot necessarily beapplied to more varied molecules.

The second and third parameters, the polarizability XY and YY are compo-nent vectors to the overall polarizability of the molecule. The tensor term XY

FIG. 2. Continued.

Page 15: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

CINNAMIC ACID BIRD REPELLENTS 2839

describes the ability of the molecule to be polarized in the y direction as a chargeis exerted across the z axis (Figure 4). This reflects a change in charge distri-bution across the acid group, best illustrated by the marked increase in activityobserved when the —COOH functional group of cinnamic acid is replaced bythe amide of cinnamamide.

The last descriptor, superdelocalizability was originally derived as an indexof reactivity of aromatic hydrocarbons. The index is based on the idea that earlyinteraction of the molecular orbitals of reactants can be regarded as a mutualperturbation, the relative energies of orbitals changing together and maintain-ing a similar degree of overlap as reactants approach each other. The index iscalculated by the sum:

FIG. 2. Continued.

Page 16: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

TABLE 1. PARAMETER VALUES AND CURVE-FIT DETAILS FOR DOSE-RESPONSERELATIONSHIPS OF CANDIDATE BIRD REPELLENTSa

Compound

Cinnamic acidCinnamyl alcoholCinnamaldehydeCinnamamide2-Methoxycinnamic acid3-Methoxycinnamic acid4-Methoxycinnamic acid3,5-Dimethoxycinnamic acid3-Nitrocinnamic acid3-Chlorocinnamic acid3-Methoxy-4-hydroxycinnamic acid3,5-Dimethoxy-4-hydroxycinnamic acid3,4-Dihydroxycinnamic acid3,5-Dimethoxybenzoic acid

R50

75.7619.355.12

17.1461.5229.8141.6827.2129.6119.8253.2766.95

b

61.94

SE

5.613.38.91.11.45.13.51.64.526.1

22.4

7.8

95% confidence limits

Lower

64.100

15.158.618.135.323.421.515.442.132.2

54.5

Upper

85.133.61419.56438.648.729.838.323.767.1

108.4

89.6

aConfidence limits for the R50 values were determined using a bootstrapping technique.b3-Methoxy-4-hydroxycinnamic acid (caffeic acid) could not be interpolated from the fitted curvewithout extrapolating beyond the observed doses.

Here the term Sr is the superdelocalizability at position r, ej is the bondingenergy coefficient in the jth molecular orbital (eigenvalue), c is the molecu-lar orbital coefficient at position r in the HOMO (highest occupied molecularorbital) and m is the index of the HOMO. This descriptor has been found tocorrelate highly to the activity of different sets of compounds in many studies(Purdy, 1991; Nandihalli et al., 1992; Karabunarlie et al., 1996).

The equation establishes the electronic effects of substitution on the cin-namic acid aromatic ring as important factors for the resulting activity, corrob-orating earlier observations. Polarizability and superdelocalizability descriptorshighlight the shift in charge distribution across the molecule as the substituentsare changed. Due to resonance or inductive effects through the conjugated cin-namic acid (3-phenyl-propenoic acid) skeleton, changes in the position (ortho,meta, or para) or substituent type will have a marked influence on the distribu-tion of electrons throughout this electron-rich system and thus the repellency ofthe compound.

Having developed a quantitative model for this series of molecule, the nextlogical step is to design compounds that combine and exaggerate those charac-teristics that underlie repellency. The marked increase in repellency is observedwhen the acid functional group is replaced by a less polar one or when electron

2840 WATKINS ET AL.

Page 17: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

CINNAMIC ACID BIRD REPELLENTS 2841

Page 18: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

2842 WATKINS ET AL.

FIG. 3. Quantitative structure-activity model for candidate bird repellents. Plot of ob-served versus predicted log R50 values.

withdrawing groups are substituted on the aromatic ring. These predictions wereconfirmed by the model: 3-chlorocinnamaldehyde should be a potentially strongrepellent. The model corroborates this prediction with a log R50 value of 0.59,although final validation will only be obtained when this compound and othersare made and tested successfully.

FIG. 4. The z and y axes of the electronic parameter tensors.

Page 19: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

CONCLUSION

A simple regression equation has been derived and statistically validatedthat confirms activities of our test compounds and can predict the activity of puta-tive cinnamic repellents. This is the first published quantitative structure-activitymodel for avian repellents and as a result we can now begin to predict whichcinnamic acid derivatives (e.g., 3-chlorocinnamaldehyde) should make effectiverepellents. Quantitative structure-activity models have been developed by Clarkand colleagues (unpublished) to explain the repellency of acetophenone andanthranilate derivatives using charge and topological descriptors. Further experi-mentation is required to assess the commonality of these models and to test theirpower in selecting new repellents from a more diverse set of compounds and fora range of avian species. However, this modeling approach promises to providea more efficient and economic method for prospecting chemical databases fornew effective bird repellents.

Acknowledgments—We thank the Pesticide Safety Directorate and the Countryside Manage-ment Division of the British Ministry of Agriculture Fisheries and Food for funding the work andthe EPSRC for the studentship of James Lumely. I would also like to thank Clare Grey and JoanneGurney for their assistance with the experiments and Roger Quy and Dave Cowan for their assistancein the experimental design and comments on the manuscript.

REFERENCES

AVERY, M. L., and DECKER, D. G. 1991. Repellency of fungicidal rice seed treatments to red-wingedblackbirds. J. Wildl. Manage. 55:327-334.

BROOKS, J. E., and ROWE, F. P. 1987. Commensal Rodent Control. World Health Organisation,Geneva.

BURRIT, E. A., and PROVENZA, F. D. 1990. Food aversion learning in sheep: Persistence of condi-tioned taste aversion to palatable shrubs (Cerocarpus montanus and Amelanchier alnifolia). J.Anim. Sci. 68:1003-1007.

CHAMBERS, K. C, and BERNSTEIN, I. L. 1995. Conditioned flavor aversions, pp. 745-773, in R. L.Doty (ed.). Handbook of Olfaction and Gustation. Marcel Dekker, New York.

CLARK, L., and SHAH, P. 1994. Tests and refinement of a general structure-activity model for avianrepellents. J. Chem. Ecol. 20:321-339.

CROCKER, D. R., and PERRY, S. M. 1990. Plant chemistry and bird repellents. Ibis 132:300–308.CROCKER, D. R., and REID, K. 1993. Repellency of cinnamic acid derivatives to rooks and

chaffinches. Wildl. Soc. Bull. 21:456–460.CROCKER, D. R., PERRY, S., WILSON, M., BISHOP, J., and SCANLON, C. 1993. Repellency of cinnamic

acid derivatives in rock doves. J. Wildl. Manage. 57:113-122.CUMMINGS, J. L., AVERY, M. L., POCHOP, P. A., DAVIS, J. E., DECKER, D. G., KRUPA, H. W., and

JOHNSTON, J. W. 1995. Evaluation of a methyl anthranilate formulation for reducing bird damageto blueberries. Crop Prot. 14:257-259.

DEWAR, M. J. S., ZOEBISCH, E. G., HEALY, E. F., and STEWART, J. J. P. 1985. AM1: A new generalpurpose quantum-mechanical model. J. Am. Chem. Soc. 107:3902-3909.

DEWAR, M. J. S., HEALEY, E. F., YUAN, Y. C., and HOLDER, A. J. 1990. Comments on a comparisonof AM1 with the recently published PM3 method. J. Comp. Chem. 11:541-542.

CINNAMIC ACID BIRD REPELLENTS 2843

Page 20: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

EFRON, B., and TIBSHIRANI, R. J. 1993. An Introduction to the Bootstrap. Chapman & Hall, London.GILL, E. L., SERRA, M. B., CANAVELLI, S. B., FEARE, C. J., ZACCAGNINI, M. E., NADIAN, A. K.,

HEFFERNAN, M. L., and WATKINS, R. W. 1994. Cinnamamide prevents captive chestnut-cappedblackbirds (Agelaius ruficapillus) from eating rice. Int. J. Pest Manage. 40:195-198.

GILL, E. L., WATKINS, R. W., COWAN, D. P., BISHOP, J., and GURNEY, J. E. 1998. Cinnamamide anavian repellent reduces wood pigeon damage to oilseed rape. Pestic. Sci. 52:159-164.

GURNEY, J. E., WATKINS, R. W, GILL, E. L., and COWAN, D. P. 1996. Non-lethal mouse repellents:Evaluation of Cinnamamide as a repellent against commensal and field rodents. Appl. Anim.Behav. Sci. 49:353-363.

HOSKINS, J. A. 1984. The occurrence, metabolism and toxicity of cinnamic acid related compounds.J. Appl Toxicol. 4:283–291.

JAKUBAS, W. J., SHAH, P. S., MASON, J. R., and NORMAN, D. M. 1992. Avian repellency of coniferyland cinnamyl derivatives. Ecol. Appl. 2:147-156.

KARABUNARLIE, S., MEKENYAN, O. G., KARCHER, W., RUSSOM, C. L., and BRADBURY, S. P. 1996.Quantum chemical descriptors for estimating the acute toxicity of substituted benzenes to theguppy and fathead minnow. Q. S. A. R. 15:311-320.

KURATA, M., ICHIKAWA, Y, TOYA, M., TAKAHASHI, I., and OKUI, Y. 1994. Resin moulding compo-sition for preventing damage by animals. U.S.A. Patent No. 5,322,862.

MANLY, B. F. J. 1991. Randomization and Monte Carlo Methods in Biology. Chapman & Hall,London.

MASON, J. R., and CLARK, L. 1992. Nonlethal repellents: The development of cost-effective, practicalsolutions to agricultural and industrial problems, pp. 115-129, in J. E. Borrecco and R. E. Marsh(eds.). Proceedings of the Vertebrate Pest Conference. University of California, Davis.

MASON, J. R., AVERY, M. L., and OTIS, D. L. 1989. Standard Protocol for Evaluation of RepellentEffectiveness with Birds. Denver Wildlife Research Center, Lakewood, Colorado.

MASON, J. R., CLARK, L., and MILLER, T. P. 1993. Evaluation of a pelleted bait containing methylanthranilate as a bird repellent. Pestic. Sci. 39:299-304.

MOSSON, H. J., WATKINS, R. W., and EDWARDS, J. P. 1996. The cinnamic acid derivative cinna-mamide as a repellent against the vine weevil Otiorhynchus sulcatus (Coleoptera; Curculion-idae). Mitt. Biol. Bundesanst. H. 316:95–100.

NANDIHALLI, U. B., DUKE, M. V., and DUKE, S. O. 1992. Relationships between molecular-proper-ties and biological activities of o-phenyl pyrrolinocarbamate and piperidinocarbamate. J. Agric.Food Chem. 40:1993-2000.

NOLTE, D. L., MASON, J. R., and CLARK, L. 1993. Nonlethal rodent repellents—differences in chemi-cal structure and efficacy from nonlethal bird repellent. J. Chem. Ecol. 19:2019-2027.

PROVENZA, F. D. 1995. Postingestive feedback as an elementary determinant of food preference andintake in ruminants. J. Range Manage. 48:2-17.

PURDY, R. 1991. The utility of computed superdelocalizability for predicting the LC50 values ofepoxides to guppies. Sci. Total Environ. 109:553-556.

SCHAFER, E. W., and JACOBSON, M. 1983. Repellency and toxicity of 55 insect repellents to red-winged blackbirds (Agelaius phoeniceus). J. Environ. Sci. Health A 18:493-502.

SCHAFER, E. W., BOWLES, W. A., and HURLBUT, J. 1998. The acute oral toxicity, repellency andhazard potential of 998 chemicals to one or more species of wild and domestic birds. Arch.Environ. Contam. Toxicol. 12:355-382.

SHAH, P. S., CLARK, L., and MASON, J. R. 1991. Prediction of avian repellency from chemical struc-ture: The aversiveness of vanillin, vanillyl alcohol and veratryl alcohol. Pestic. Biochem. Phy-siol. 40:169-175.

SMITH, D. A., ULMER, C. W., and GILBERT, M. J. 1992. Structural studies of aromatic amines andthe DNA intercalating compounds meta-amsa and ortho-amsa: Comparison of MNDO, AMIand PM3 to experimental and abinitio results. J. Comp. Chem. 13:640–650.

2844 WATKINS ET AL.

Page 21: Quantitative Structure–Activity Relationships (QSAR) of Cinnamic Acid Bird Repellents

STARR, R. I., BESSER, J. F., and BRUNTON, R. B. 1964. Bird repellency: A laboratory method forevaluating chemicals as bird repellents. J. Agric. Food Chem. 12:342-344.

STEWART, J. J. P. 1989. Optimisation of parameters for semiempirical methods. J. Comp. Chem.10:209-264.

WATKINS, R. W., GILL, E. L., and BISHOP, J. D. 1995. Evaluation of cinnamamide as an avian repel-lent: Determination of a dose-response curve. Pestic. Sci. 44:335-340.

WATKINS, R. W., GILL, E. L., and COWAN, D. P. 1996a. Plant secondary chemicals as non-lethalvertebrate repellents, pp. 186-192, in R. M. Timm and A. C. Crabb (eds). Proceedings of theVertebrate Pest Conferencce. University of California, Davis.

WATKINS, R. W., MOSSON, H. J., GURNEY, J. E., COWAN, D. P., and EDWARDS, J. P. 1996b. Cinnamicacid derivatives: Novel repellent seed dressings for the protection of wheat seed against damageby the field slug, Deroceras reticulatum. Crop Prot. 15:77-84.

WORONECKI, P. P., DOLBEER, R. A., and STEHN, R. A. 1981. Response of blackbirds of mesurol andsevin applications on sweet corn. J. Wildl. Manage. 45:693-701.

CINNAMIC ACID BIRD REPELLENTS 2845