Analytica Chimica Acta 478 (2003) 271–280

Determination of the insecticide fenoxycarb in apple leaf samplesby an enzyme-linked immunosorbent assay

Cristina Giovannolia,∗, Gianfranco Giraudia, Claudio Baggiania, Cinzia Tozzia,Laura Anfossia, Marcello Dolcib

a Dipartimento di Chimica Analitica, Università di Torino, Via P. Giuria 5, 10125 Torino, Italyb Dipartimento di Valorizzazione e Protezione delle Risorse Agroforestali, Università di Torino, Torino, Italy

Received 6 September 2002; accepted 21 November 2002

Abstract

An enzyme-linked immunosorbent assay (ELISA) was used for the determination of fenoxycarb in apple leaf samples.Single step extraction procedures with phosphate–citrate buffered solution containing different amounts of methanol weretested showing that a solvent percentage of 20% (v/v) was the best condition, with recoveries between 85 and 100% inthe working range of 25–500�g kg−1 and a negligible matrix effect. The low detection limit reached, 1�g kg−1 against50�g kg−1 for the recommended liquid chromatographic method, makes the ELISA more suitable for determinations of thefenoxycarb residues in apple leaf samples. The reliability of the ELISA was evaluated by assaying the insecticide in spiked andcontaminated samples by three different approaches: direct determination, standard addition method with a calibration graph,and the dilution test. The corresponding coefficients of variation were, respectively, 11, 22 and 27%. The direct determinationon the (1+ 1) diluted apple leaf extract was used to measure the insecticide residues in samples collected in the north-easternItalian regions of Veneto and Trentino-Alto Adige.© 2002 Elsevier Science B.V. All rights reserved.

Keywords: Ethyl 2-(4-phenoxyphenoxy)ethylcarbamate; Fenoxycarb; Insecticide; ELISA; Apple leaf; Immunoassay; Insegar

1. Introduction

Ethyl 2-(4-phenoxyphenoxy)ethylcarbamate, com-monly known as fenoxycarb, is a recent insect growthregulator insecticide commercially available in differ-ent formulations, the widest used of which is calledInsegar. As widely documented[1–4], fenoxycarbexhibits strong juvenile hormone-mimic activity in in-sects by inhibiting metamorphosis to the adult stage,interfering with the moulting of early instar larvae and

∗ Corresponding author. Tel.:+39-011-6707846;fax: +39-011-6707615.E-mail address: cristina.giovannoli@unito.it (C. Giovannoli).

involving ovicide and delayed larvicide–adulticideeffects in various insect species. Its specific modeof action causes serious damage to the developmentand behaviour of many insects but low toxicity tomammals and some friendly adult insects[5,6]. Forthese reasons, fenoxycarb has recently been mainlyused for insect control in agriculture and productstorage[5,7–9]. In spite of these excellent features,there has recently been some evidence of its harm tosome aquatic species and non-target insects[10–14].Among these non-target insects the silkwormBombyxmori can be identified, which has shown itself to bevery sensitive to fenoxycarb, suspected of causingthe observed inability of theB. mori larvae to spin

0003-2670/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved.doi:10.1016/S0003-2670(02)01510-6

272 C. Giovannoli et al. / Analytica Chimica Acta 478 (2003) 271–280

their cocoon and pupate[2,15–18]. This syndrome,which is causing considerable economic damage tosilkworm breeding, could be connected to the useof the insecticide in fruit plant protection during thegrowth period of the larvae. There is the suspicion,in fact, that the seasonal treatments contaminate notonly the directly treated fruit leaves but also leavesof other types of plants growing in neighbouring ar-eas (e.g. mulberry leaves)[19]. In Italy, damage tosilkworm breeding has led the people involved torequest the complete banning of fenoxycarb in fruitgrowing. In this context, our laboratory has recentlybeen involved in a monitoring program of fruit treeleaf contamination in order to relate the presence offenoxycarb to the silkworm syndrome.

The chromatographic techniques for the residueanalysis of fenoxycarb on vegetal samples are gener-ally based on complex and time-consuming extractionand clean-up procedures followed by detection ona reverse phase chromatographic column[20,21].Moreover, the detection limit of the liquid chro-matography (LC) methods is generally not excellent.For example, the recommended method[20], startingfrom an extraction of 50 g of raw material, shows adetection limit of about 50�g kg−1. This means thatvery small concentrations (i.e. at the�g kg−1 level) offenoxycarb in real samples require greater quantitiesof treated raw material in order to be detected. Alsoassuming that this does not modify the typical recov-eries of the procedure and that the determinabilityis not compromised, the practicability of the methodfails when many samples have to be considered.Recently, procedures based on solid-phase microex-traction coupled with LC mass spectrometry (MS) orgas chromatography (GC)-MS have been proposed[22,23] with a gain in terms of sample work-upand sensitivity. However, due to easy or no samplepre-treatment and cheap instrumentation immunoas-say provides a good alternative for trace analysis. Asensitive and selective enzyme-linked immunosorbentassay (ELISA) in the immobilised antigen formathas been recently developed as a result of studieson the structural features that the parent moleculeshould have to raise a specific antiserum against theanalyte[24]. In a previous paper[25], we referred topreliminary work regarding the preparation and char-acterisation of an anti-fenoxycarb antiserum to setup an ELISA for the determination of fenoxycarb on

vegetal samples. The present paper deals with the ap-plication of ELISA to the determination of fenoxycarbin apple leaf samples collected in north-eastern Italywhere economic losses in silkworm breeding are morepronounced.

2. Experimental

2.1. Chemicals and instruments

Fenoxycarb with a purity grade of 99.9% wasobtained from Dr. Ehrenstorfen GmbH (Augsburg,Germany). Stock solutions of the insecticide wereprepared in methanol at 1 mg ml−1 and were storedat −20◦C. Bovine serum albumin (BSA) and tyre-oglobulin (bTG) were purchased from Sigma (St.Louis, MO). All other chemicals were obtainedfrom Merck (Darmstadt, Germany). The rabbit poly-clonal antiserum to fenoxycarb and the conjugatesbetween 2-(4-phenoxyphenoxy)ethanol hemisucci-nate and protein (BSA for the immunogen and bTGfor the solid phase antigen) were prepared accord-ing to the procedure previously described[25]. Thegoat–anti rabbit IgG–horseradish peroxidase conju-gate was obtained from Bio-Rad (Hercules, CA). Thechromogen-substrate reagent was prepared by mixing,immediately before use, equal volumes of a 0.02%(m/v) solution of tetramethylbenzidine dihydrochlo-ride with a 0.005% (v/v) solution of hydrogen perox-ide, both prepared in a 0.075 M citrate buffer, pH 5.Solvents used for extractions (analytical grade) andchromatographic analysis (HPLC quality) were sup-plied by Merck. The Extract-clean Florisil cartridgeswere supplied by Alltech Italia (Milano, Italy). TheC18 monolithic LC column was a Chromolith fromMerck. The LC apparatus (pump L-6200, UV-Visdetector L-4200 and integrator D-2500) came fromHitachi-Merck (Darmstadt).

The microplates were from Nunc (Roskilde, Den-mark), the microplate washer (Novapath Washer), themicroplate incubator and the microplate reader (Mi-croplate Reader 3550) were supplied by Bio-Rad. Allthe apple leaf samples considered were kindly sup-plied by the Italian Silk Growers National Associationand were taken from the north-eastern Italian regionsof Veneto and Trentino-Alto Adige where silk wormgrowth problems were detected. The apple leaf blank

C. Giovannoli et al. / Analytica Chimica Acta 478 (2003) 271–280 273

sample was kindly supplied by the same associationand collected in an area without contamination.

2.2. Sample preparation

Apple leaves, kept in a refrigerator at−20◦C toavoid any degradation, were defrosted, cut into smallpieces and the stalks and ribs removed. This treatmentwas carried out on the whole sample available for anal-ysis. Fifty grams of each sample was homogenisedfor 10 min at high speed in a homogenizer togetherwith 250 ml of 0.1 M phosphate–citrate buffer, pH 7.4,0.05 M sodium chloride, 0.001 M EDTA, 0.1% (m/v)sodium azide, containing different percentages (v/v) ofmethanol (0, 10, 20, 30 and 50%). The homogenisedmixture was first filtered through a wide-mesh steelsieve, then centrifuged in glass tubes at about 500×g

for 15 min. The solution was filtered through a Milli-pore apparatus with a 0.22�m filter cut-off. The finalsolution was kept in glass tubes at 4◦C and analysedwithin a week. The leaf samples prepared as abovementioned were assayed without any other prelimi-nary treatment.

2.3. Preparation of immunosensitive microplates

The polystyrene microplate wells were coated with0.3 ml of the conjugate 2-(4-phenoxyphenoxy)ethanolhemisuccinate-bTG (reaction mole ratio 20:1) at a con-centration of 2�g ml−1 in a 0.05 M sodium carbonatebuffer, pH 9.6. After an overnight reaction, wells werewashed three times with a 0.05% (v/v) aqueous solu-tion of Tween 20 (washing solution) and then blockedwith 0.3 ml of blocking buffer (0.02 M sodium phos-phate buffer, pH 7.4, 0.15 M sodium chloride, 0.001 MEDTA, 0.1% (m/v) gelatine, 5% (m/v) of sucrose, 4%(m/v) poly(vinylpyrrolidone). After a 1 h reaction atroom temperature, the wells were washed three timeswith the washing solution. The coated microplateswere stored at 4◦C and sealed in a plastic envelopeuntil use.

2.4. Competitive immunoassay procedure

A 0.1 ml of fenoxycarb standards at of 0, 0.1, 0.2,0.4, 2, 4, 20, 40, 200, 400 and 800�g l−1 in a dilu-tion buffer (0.1 M phosphate–citrate buffer, pH 7.4,0.2 M sodium chloride, 0.001 M EDTA, 0.1% (m/v)

gelatine, 20% (v/v) methanol, 0.05% (v/v) Tween20) and 0.1 ml of anti-fenoxycarb antiserum, diluted1:100,000 in the same buffer, were dispensed in du-plicate into each well, coated as previously described.

For the analysis of real samples, 0.1 ml of leaf ex-tract, undiluted and diluted 1+1, 1+3 and 1+7 withthe dilution buffer, was dispensed in duplicate intoeach well (coated as described above) rather than thestandard solutions of fenoxycarb. Since the optimalconditions for the ELISA assay require the use of 20%(v/v) methanol in the dilution buffer, leaf extracts withdifferent amounts of methanol have to be corrected to20% (v/v) by dilution with buffer or methanol beforeanalysis.

The non-specific binding (NSB) was measuredby replacing the diluted antiserum with the diluentbuffer. This substitution allows the measurement ofnon-specific interactions between the labelled an-tiserum and the immunosensitive solid phase andfurthermore, in the case of the sample analysis, thenon-specific interactions of the leaf extract compo-nents with the solid phase.

Wells were incubated overnight at room temper-ature, then washed three times with the washingsolution. For leaf sample analysis, the presence of amatrix effect obliged us to consider different washingsolution compositions. The antibodies to the fenoxy-carb analogue on the solid phase were detected byadding 0.2 ml of goat–anti rabbit IgG–horseradishperoxidase conjugate (diluted 1:4000 with 0.02 Mphosphate buffer, pH 7.4, 0.12 M sodium chloride,0.001 M EDTA, 0.1% (m/v) gelatine, 0.05% (v/v)Tween 20), incubated for 1 h at 37◦C and the wellswashed three times with the washing solution. Fi-nally, 0.2 ml of (1+ 1) chromogen-substrate mixture,prepared as described above, was dispensed into eachwell and incubated in the dark for 30 min at 37◦C. Thecolour development was then stopped by the additionof 0.1 ml of 1 M sulphuric acid and the absorbancewas read at 450 nm and, when out of scale, at 415 nm.

2.5. Washing solution composition

The washing solution used after each reaction stepwas always a 0.05% (v/v) aqueous solution of Tween20 except for the first washing step of the methanolicleaf extract analysis, where different ionic strengths,obtained by the addition of different concentrations of

274 C. Giovannoli et al. / Analytica Chimica Acta 478 (2003) 271–280

sodium chloride to the 0.05% (v/v) aqueous solutionof Tween 20, different surfactant percentages and dif-ferent numbers of washing cycles were considered inorder to define the optimal washing conditions.

2.6. Data analysis

The calibration graph (absorbance versus logfenoxycarb concentration) was obtained throughnon-linear curve fitting by means of a four-parameterlogistic equation[26]. The limit of detection—thelowest concentration of analyte that can be determinedto yield a response statistically different from that ofthe zero dose—was calculated as the concentration offenoxycarb giving an absorbance equal to the meanvalue of 12 repeated experimental measures at thefenoxycarb zero dose plus or minus three standarddeviations[26].

The concentrations of fenoxycarb in leaf extractswere determined by directly reading the concentra-tions correspondent to the sample absorbances onthe calibration graph, after the subtraction of thenon-specific signals of the samples themselves. Theseamounts were corrected by taking into account boththe eventual previous dilution and the effective sam-ple dilution in the well. Each amount determined wasgiven as the mean value of two independent measure-ments. Moreover, in order to express the fenoxycarbamounts as�g kg−1 of treated leaves, the concentra-tion values obtained were multiplied by 5.

2.7. Recovery and dilution test

Matrix effects were investigated by adding fenoxy-carb at various concentrations to several methanolicleaf extracts. Aliquots of leaf extract were spiked withequal volumes of standards of fenoxycarb at con-centrations of 2, 4, 20�g l−1 in the dilution buffer.The spiked samples were dispensed and incubatedinto immunosensitive wells with the anti-fenoxycarbantiserum according to the procedure described.Wells were then washed three times with a wash-ing solution containing 0.25 M sodium chloride. Theanti-fenoxycarb antibodies bound to the solid phasewere detected by the described procedure. The ex-perimental data, subtracted from their correspondingNSBs, was processed by plotting the fenoxycarb con-centration measured on the calibration graph versus

the added amount. Thus, the unknown amount offenoxycarb (present as contaminant in the leaf sam-ples) was calculated by the intercept value of thecorresponding regression line.

The fenoxycarb concentrations determined in theleaf extracts at different sample dilutions (1+ 1,1 + 3 and 1+ 7) were processed by means of thedilution test, which consists of plotting the measuredfenoxycarb versus the inverse of the dilution factor.The regression line slope supplies the insecticideconcentration in undiluted samples.

2.8. LC analysis

The LC determinations were performed accordingto the recommended method[20]. Briefly, 50 g of ap-ple leaf sample was extracted with 200 ml of acetoneand homogenised for 5 min in a mixer. After the fil-tration step, the extract was concentrated to 10–20 mlon a rotary vacuum evaporator, diluted with 400 mlof water and twice extracted with 50 ml of hexane.The organic phases were collected, filtered through asodium sulphate layer and evaporated to dryness. Theresidue was dissolved in 2 ml of the hexane–ethyl ac-etate mixture (75+15, v/v) that was used as eluent toperform a clean-up procedure on a Florisil column.The collected eluate (50 ml) was evaporated. The dryresidue was dissolved in 1 ml of acetonitrile–deionizedwater mixture (1+ 1, v/v) and injected into the liq-uid chromatograph. The elution was performed on areverse phase column by using the mixture 50 mMphosphate buffer–acetonitrile (1+ 1, v/v) as mobilephase and monitoring the absorbance at 228 nm. Therecovery tests were accomplished by spiking organicextract aliquots of the uncontaminated apple leaf sam-ple with appropriate amounts of fenoxycarb to givefinal additions of 250 and 500�g kg−1.

3. Results and discussion

The ELISA used for the determination of fenoxy-carb in apple leaf samples has been developed startingfrom the approach for the determination of the in-secticide in aqueous buffered solution with limitedamounts of methanol previously described[25] andextended to quite complex real samples with a de-tection limit of 1 ± 0.015�g kg−1. The extraction

C. Giovannoli et al. / Analytica Chimica Acta 478 (2003) 271–280 275

procedure from apple leaves was quick and simplebut not selective for fenoxycarb. In consequence,different components present in the leaves wereco-extracted and constituted potential interferents inthe analyte determination. Thus, several experimentswere needed in order to evaluate the matrix effect andhow this would potentially alter the reliability of theimmunoassay.

Although the immunochemical approach appears tobe more attractive than chromatographic methods, acritical point has to be raised. It concerns the difficulty,and practically the impossibility, to validate such animmunoassay by comparison with the recommendedmethod[20], at least in the concentration range ofmost sample contaminations. It is due to the differ-ence in the detection limit between the two methodsand in the level of contamination measured, whichis almost always below the detection limit of the LCmethod. Thus, if in theory the validation could be re-alised by using an appropriate amount of raw sampleto stay above the detection limit of the LC method, inpractice this means treating kg of apple leaves. Thisis hardly practicable and poses serious doubts for thefinal results. Nevertheless, a comparison with the rec-ommended LC method was performed by means ofrecoveries of large insecticide concentrations added tothe uncontaminated apple leaf sample also used to de-fine the optimal methanol percentage of the extractionmixture. For all the other samples considered, insteadof validation through a comparison with a LC method,we tested the reliability and the accuracy of the ELISAby assaying fenoxycarb in real samples with differ-ent approaches: the standard addition method withspiked and contaminated samples, the direct determi-nation with a calibration graph and the dilution test onthe same samples. These comparative studies preventthe presence of unknown matrix effects invalidatingthe fenoxycarb quantification. The approaches we fol-lowed led us to estimate the ELISA performances asfollows.

Table 1NSB/A0 ratios obtained for a different number of washing cycles and different composition of the washing solutions

Tween 20(0.05%, v/v)

Tween 20(0.2%, v/v)

Tween 20 (0.05%, v/v)NaCl, 0.25 M

Tween 20 (0.05%, v/v)NaCl, 0.50 M

Cycles 3 5 3 5 3 5 3 5NSB/A0 0.17 0.12 0.24 0.10 0.056 0.052 0.048 0.041

3.1. Washing solution

The first matrix effect revealed was the highernon-specific absorbance values in the real samplescompared to those obtained by utilising the bufferedsolution as diluent of the insecticide standards usedfor calibration. Furthermore, we observed that all themethanolic extracts assayed showed a comparableoverestimation of the NSBs, without any relationshipto the concentration of fenoxycarb really present bothby contamination and by spiking. So, this behaviourcan be ascribed to an amplification of the analyticalsignal related to the sample composition but not to thefenoxycarb itself. Since we are interested in measur-ing the insecticide concentration by direct reading ofthe sample absorbance on the calibration graph, it isimportant to make the non-specific signals of samplesand standards as comparable as possible to reduceany unbalanced matrix effect. For this purpose, thecomposition of the washing solution used after thefirst overnight reaction step was changed by consid-ering increasing amounts of surfactant and sodiumchloride. The quantities considered and the resultsobtained, expressed in terms of NSB/A0 (where A0is the absorbance of the zero standard), are shownin Table 1. While the increase of the surfactant per-centage does not seem to have a positive effect uponthe NSB/A0 ratio, which would be expected to be de-creased, the presence of moderate amounts of sodiumchloride causes a sharp decrease. Also the numberof washing cycles, which have a significant influ-ence in the reduction of this ratio, do not lower theNSB/A0 ratio at values comparable with those mea-sured with the sodium chloride addition. On the basisof these results, the sample analysis was executed bywashing wells (after the first overnight reaction step)three times with a 0.05% (v/v) Tween 20 aqueoussolution containing 0.25 M sodium chloride, sincethe use of a doubled sodium chloride concentrationreduces the ratio NSB/A0 very little. This washing

276 C. Giovannoli et al. / Analytica Chimica Acta 478 (2003) 271–280

solution composition allowed the ratio NSB/A0 to becomparable with that obtained for calibration.

3.2. Optimal extracting solvent percentage

The main goal is to realise the best extraction con-ditions, which would mean extracting as great andas reliable an amount of insecticide as possible. So,although the optimised calibration graph was ob-tained by using a buffer with 20% (v/v), methanol,the presence of such a limited concentration of sol-vent could be unsuitable to solubilise and effectivelyextract the insecticide from the matrix and to give areliable ELISA determination. The more appropriateconcentration of solvent to balance any matrix effectand to produce an effective extractive step was inves-tigated through preliminary studies that were carriedout on an apple leaf sample that certainly was nottreated with the insecticide or contaminated by a drifteffect, so that it represented a genuine blank sam-ple. The leaf extracts obtained by extraction with thephosphate–citrate buffer containing different propor-tions of methanol did not show the same appearance.In fact, the extract obtained without methanol re-mained cloudy after filtration, as happened with theextract which contained 50% (v/v) methanol, thoughin this case only slight precipitation was observed. Asthe presence of particulates can seriously compromisethe insecticide determination, we decided to excludethe sample treatment just mentioned. All the otherextract solutions appeared clear after filtration andremained like this during their storage in the refrig-erator. Among the extractions with 10, 20 and 30%(v/v) methanol, which did not show any precipitation,the best experimental condition is that which warrantsthe lowest matrix effect and the most reproducibleinsecticide recoveries. The evaluation of the presenceof a matrix effect was carried out by comparing theanalytical signal of each leaf extract (Asample), diluted1+1, with the zero standard (A0). The ratio defined asAsample/A0 should be near to 1 when the matrix effectis completely balanced, while values above or below 1indicate the presence of a matrix effect. A pronouncedmatrix effect is observed with a small percentage ofsolvent (10%, v/v), which leads to a ratio greaterthan 1.2, whereas a strong decrease is produced byincreasing the amount of methanol in the extractingbuffer. In fact, the ratiosAsample/A0 measured for leaf

extracts containing 20 and 30% (v/v) methanol are,respectively, 1.05 and 0.9. According to these results,the more appropriate percentage of methanol in theextracting buffer appears to be 20% (v/v).

The recoveries of known insecticide amounts addedto the apple leaf blank extract with 10, 20 and 30%(v/v) methanol defined the best extraction condition.The experimental results underline once again theinfluence of the methanol: a small solvent percentage(10%, v/v) led to recoveries in the working range ofthe calibration graph that seldom were >10%, whileoverestimation of the added amount (recoveries be-tween 150 and 200%) was observed with a largesolvent percentage (30%, v/v). Also in this case, themost satisfactory experimental condition appears tobe the one which uses 20% (v/v) methanol, whose re-coveries are about 85±5.2% at 25�g kg−1, 90±3.2%at 100�g kg−1 and 98± 2.8% at 500�g kg−1. Theopposite trends of theAsample/A0 ratio and of therecoveries, observed by changing the methanol per-centage in the apple leaf extracts, have to be consid-ered to be unrelated because the two measurementsare connected with different phenomena (respectivelythe non-specific analytical response due to a com-plex matrix at the zero analyte concentration and theimmunospecific response related to a well-definedanalyte concentration). The recovery data mentionedabove were confirmed by the LC analysis, whoserecoveries were about 75± 6% at 200�g kg−1 and87 ± 7% at 500�g kg−1. The accordance betweenthe experimental results of the LC method and ofthe ELISA performed on the 20% (v/v) methanolicleaf extracts shows the feasibility of using the ELISAmethod for the determination of fenoxycarb in appleleaf samples.

Once the proper extraction conditions were defined,we studied what the best procedure was to follow inorder to assay different insecticide concentrations inleaf samples for routine analysis.

3.3. The recovery of fenoxycarb frommethanolic leaf extracts

The accuracy and reliability of the fenoxycarbdetermination were investigated by spiking five ap-ple leaf extracts as described above. The addedfenoxycarb, the fenoxycarb measured on the calibra-tion graph, the intercept values determined and the

C. Giovannoli et al. / Analytica Chimica Acta 478 (2003) 271–280 277

Table 2Fenoxycarb concentrations in methanolic extracts spiked with fixed amounts of insecticide measured from calibration graph

Sample Fenoxycarb added (�g l−1) Fenoxycarb founda (�g l−1) Interceptb (�g l−1) Fenoxycarb unknownc (�g l−1)

1 1 11.5± 1.72 13.0± 1.3 10.2± 1.6 20.4± 3.2

10 23.8± 1.4

2 1 4.78± 0.092 5.68± 0.19 3.82± 0.38 7.61± 0.8

10 13.3± 2.2

3 1 3.40± 0.362 4.31± 0.38 2.74± 1.1 5.43± 2.3

10 10.3± 0.54

4 1 8.65± 0.122 9.74± 0.45 7.52± 0.1 15.0± 0.2

10 18.6± 0.24

5 1 15.9± 0.822 16.7± 0.23 14.0± 6.0 28.0± 12.0

10 29.6± 0.27

a Samples were diluted 1+ 1 standard additions.b Intercept values of the regression lines in the plot of fenoxycarb added vs. fenoxycarb found.c Fenoxycarb concentrations measured by multiplying the intercept values of the linear regression by two (dilution factor into wells).

unknown fenoxycarb amount present in the apple leafsamples are shown inTable 2. As the recoveries ofdifferent amounts of fenoxycarb from the blank appleleaf extract were good, the recovery tests were per-formed on contaminated spiked samples in order toquantify the unknown insecticide concentration. Thecorresponding regression line has a good correlationcoefficient (>0.99), so that the measured concentra-tion values are directly proportional to the fenoxycarbamounts added. This behaviour excludes the pres-ence of a relevant matrix effect, even if a furtherinvestigation must be made to certify this statement.

3.4. Direct determination based on thecalibration graph

The measurement of the fenoxycarb concentrationby direct reading of the sample absorbance on the cal-ibration graph represents the easiest and quickest wayto perform the analysis on a great number of samples.Nevertheless, this approach has to be corroboratedthrough tests which define the experimental condi-tions that give the most reliable and accurate determi-nations. First of all, it is important to estimate whatdilution of the leaf extracts is most suitable for the de-termination. To fulfil this, the five methanolic extracts

considered above were assayed at different dilutions.Table 3 shows the experimental results obtained,which show that the determinations in the undilutedextracts 1, 4 and 5 greatly exceed those of the 1+ 3and 1+ 7 dilutions, whereas a comparable responsebetween undiluted and diluted extracts is observed forsamples 2 and 3. The overestimation of the insecti-cide in the undiluted extracts may be attributed to thepresence of an unbalanced matrix effect that is notconstant but seems to vary significantly from sampleto sample. In fact, it may be connected with the vari-ety of apple-tree, the age of the leaves and unknownvariables which change from sample to sample. As wecannot definitely know the magnitude of the matrixeffect of a given sample the choice of the experimentalconditions to minimise this effect becomes necessary.So, it is more appropriate to refer to the insecticideconcentrations determined on samples diluted 1+1 or1+3 or 1+7, whose experimental values appear com-parable within the experimental error. But the higherdilutions make the determination difficult on samplescontaining low doses of fenoxycarb. This causes usto neglect dilutions 1+ 3 and 1+ 7. Moreover, thefact that the results of the recovery tests show a goodfit with the samples diluted 1+ 1 justifies the choiceperforming the assay on samples diluted 1+ 1.

278 C. Giovannoli et al. / Analytica Chimica Acta 478 (2003) 271–280

Table 3Fenoxycarb concentrations measured in methanolic leaf extractsat different dilutions

Sample Dilution Fenoxycarb measured (mg l−1)

1 1 + 0 70 ± 81 + 1 15.4± 0.81 + 3 15.1± 4.21 + 7 15.9± 2.9

2 1 + 0 7.92± 0.51 + 1 6.01± 1.21 + 3 4.84± 0.11 + 7 7.52± 4.2

3 1 + 0 6.25± 0.61 + 1 4.84± 0.61 + 3 9.21± 0.041 + 7 10.3± 4.0

4 1 + 0 147± 951 + 1 13.2± 1.31 + 3 15.6± 2.51 + 7 13.8± 5.3

5 1 + 0 235± 1201 + 1 26.2± 1.91 + 3 16.6± 2.11 + 7 18.0± 3.5

The experimental values (mean± S.D.), corrected by the dilutionfactor, supplied the insecticide concentrations expressed as�g l−1.

3.5. The dilution test

The results of the dilution test, shown inTable 4,confirm the good agreement with the determinationexecuted on samples diluted 1+ 1. This agreementcan be observed for all the samples, though sample 3gives results which are more scattered.

Finally, the fenoxycarb determinations carried outby the different approaches mentioned above are com-pared inTable 5. The good agreement (supported byt-test values) of the experimental data suggests that all

Table 5Comparison between fenoxycarb concentrations measured by different approaches

Sample Direct determination (�g kg−1) Standard addition method (�g kg−1) Dilution test (�g kg−1)

1 77 ± 4 102± 16 76± 1.52 30 ± 6 38 ± 4 27 ± 83 24 ± 3 27 ± 11 13± 9.54 66 ± 6.5 75± 1 63 ± 7.55 131± 9.5 140± 60 149± 25

The concentrations are expressed as�g kg−1 leaves. The direct determination was performed on samples diluted 1+ 1.

Table 4Fenoxycarb concentrations measured by means of the dilution test

Sample Fenoxycarb (�g l−1) slope,m ± S.D.

1 15.3± 0.32 5.41± 1.63 2.64± 1.94 12.7± 1.55 29.9± 5.1

The linear regression was performed by using the mean valuesobtained for the differently diluted leaf extracts (excluding theundiluted one) on the calibration graph.

three methods can be used for fenoxycarb determina-tion. Taking into account the percentage coefficientsof variation, the direct determination from the calibra-tion graph (with a sample dilution of 1+ 1) appearsto be more precise (11 against 22% for the standardaddition and 27% for the dilution test), besides beingthe most rapid.

3.6. Application of the ELISA

Once the ELISA was optimised, the goal was toscreen a consistent number of apple leaf samples,whose results are shown inFig. 1. The best period tocollect samples is late spring or early summer, whenall treatments are about to be concluded.

All the apple leaf samples considered were poten-tially contaminated by unknown amounts of insecti-cide, with the exception of samples 1–3 and reported inTable 6. Sample 1 was not directly treated with Insegar(the pesticide formulation commercially available) sothat contamination could only be indirect. Samples 2and 3 were treated with Insegar, respectively, 1 dayand 15 days before collection. Therefore, the concen-tration of insecticide determined on these samples canbe considered as a reference. As a consequence, a level

C. Giovannoli et al. / Analytica Chimica Acta 478 (2003) 271–280 279

Fig. 1. Fenoxycarb determination on apple leaf samples collected in the north-eastern Italian regions of Veneto and Trentino-Alto Adige. Theconcentrations measured were in the ranges: 4.6–627�g kg−1 (1996), 4.7–217�g kg−1 (1997), 1.4–13�g kg−1 (2000) and 9.4–63�g kg−1

(2001).

Table 6Fenoxycarb determined in apple leaf samples not directly treated(1), treated for 1 day (2) and 15 days (3) before collection

Sample Fenoxycarb (�g kg−1)

1 43 ± 52 517± 663 139± 30

of contamination comparable with that of sample 1can be ascribed to indirect contamination due to treat-ment performed in neighbouring areas, while a levelof contamination of hundreds of�g kg−1 can be re-lated to direct treatment of the sample. Moreover, thegreater the concentration of fenoxycarb, the more re-cent the treatment can be considered. A relevant aspectis that a sharp decrease of the fenoxycarb concentra-tion measured in the last 2 years was observed, withsome samples (omitted inFig. 1) showing contamina-tion below the ELISA detection limit.

4. Conclusions

The experimental work showed that the applicationof the ELISA for the determination of fenoxycarb in

apple leaves shows itself to be more attractive thanthe standard method which uses a LC determinationpreceded by quite complex and time-consuming ex-traction and clean-up steps. The main advantages ofthe immunometric approach are higher sensitivityand easier sample pre-treatment, both connected toits specificity. Due to the difficulty of validating theELISA by comparison with the conventional tech-nique, because of the lower sensitivity of the latter, acomparative analysis of the results obtained througha different approach was needed. The level of re-producibility reached and the satisfactory recoveriesmake the ELISA an alternative to pursue at thesecontamination levels.

References

[1] S. Dorn, M.L. Frischknecht, V. Martinez, R. Zurfluh, U.Fischer, Z. Pflanzenkrankh. Pflanzenschutz 88 (1981) 269.

[2] S. Grenier, A.M. Grenier, Ann. Appl. Biol. 122 (1993) 369.[3] M. Schneider, G. Wiesel, A. Dorn, J. Insect Physiol. 41 (1995)

23.[4] I. Ujvary, G. Matolcsy, I. Belai, F. Szurdoki, K. Bauer, L.

Varjas, K.J. Kramer, Arch. Insect Biochem. Physiol. 32 (1996)659.

[5] C.R. Worthing, R.J. Hance, The Pesticide Manual, A WorldCompendium, ninth ed., British Crop Protection Council, UK,1991, p. 375.

280 C. Giovannoli et al. / Analytica Chimica Acta 478 (2003) 271–280

[6] A.C. Barr, B. Abbitt, R.A. Fiske, J.T. Jaques, H.R. Maynard,J.C. Reagor, J. Veg. Diagn. Invest. 9 (1997) 401.

[7] M.G. Solomon, J.D. Fritzgerald, J. Hortic. Sci. 65 (1990) 535.[8] H. Vogt, Biol. Agric. Hortic 15 (1997) 241.[9] J.P. Edwards, J.E. Short, L. Abraham, J. Stored Prod. Res.

27 (1991) 31.[10] P.B. Key, G.I. Scott, J. Environ. Sci. 29 (1994) 873.[11] A.J. Hosmer, L.W. Warren, T.J. Ward, Environ. Toxicol.

Chem. 17 (1998) 1860.[12] T.X. Liu, T.Y. Chen, Fla. Entomol. 84 (2001) 628.[13] W. Wakgari, J. Giliomee, Int. J. Pest Manage. 47 (2001) 179.[14] J.N. Tasei, Apidologie 32 (2001) 527.[15] S.G. Dedos, H. Fugos, Insect Biochem. Mol. Biol. 29 (1999)

723.[16] H. Monconduit, B. Mauchamp, Arch. Insect Biochem.

Physiol. 40 (1999) 141.[17] S.G. Dedos, H. Fugos, Zool. Sci. 18 (2001) 771.

[18] D. Bharathi, Y. Miao, J. Adv. Zool. 22 (2001) 56.[19] A. Arzone, M. Dolci, F. Marletto, C. Minero, Biosci. Biotech.

Biochem. 59 (1995) 1318.[20] R.P. Haenni, P.A. Mueller, Analytical Methods for Pesticide

and Plant Growth Regulators, vol. XVI, Academic Press, NewYork, 1988, p. 21.

[21] M. Michel, A. Krause, B. Buszewski, Pol. J. Environ. Stud.10 (2001) 283.

[22] Z. Wang, B. Hennion, L. Urruty, M. Montury, Food Addit.Contam. 17 (2000) 915.

[23] M.L. Reyzer, J.S. Brodbelt, Anal. Chim. Acta 436 (2001) 11.[24] F. Szurdoki, A. Szekacs, H.M. Le, B.D. Hammock, J. Agric.

Food Chem. 50 (2002) 29.[25] G. Giraudi, C. Giovannoli, C. Baggiani, I. Rosso, P. Coletto,

M. Dolci, G. Grassi, A. Vanni, Anal. Commun. 35 (1998)183.

[26] IUPAC, Pure Appl. Chem. 67 (1995) 2065.