ORIGINAL PAPER

Development of a fluorescence polarization immunoassayfor the detection of melamine in milk and milk powder

Qiang Wang & Simon A. Haughey & Yuan-Ming Sun &

Sergei A. Eremin & Zhen-Feng Li & Hui Liu &

Zhen-Lin Xu & Yu-Dong Shen & Hong-Tao Lei

Received: 17 October 2010 /Revised: 6 December 2010 /Accepted: 12 December 2010# Springer-Verlag 2011

Abstract A fluorescence polarization immunoassay (FPIA)based on a polyclonal antibody was developed for thedetermination of melamine in milk. To obtain an antibodywith improved sensitivity and specificity, 6-hydrazinyl-1,3,5-triazine-2,4-diamine was coupled to bovine serumalbumin and used as the immunogen for the rabbitimmunization. Three fluorescein-labeled melamine tracerswith different structures and spacer bridges were synthe-sized. The structural effect of the tracers on the assaycharacteristics was investigated. 6-(4,6-Diamino-1,3,5-triazin-2-ylamino)-N-(2-(3-(3′,6′-dihydroxy-3-oxo-2,3-dihydrospiro[indene-1,9′-xanthene]-5-yl)thioureido)ethyl)hexanamide demonstrated better sensitivity than 5-(2-(4,6-diamino-1,3,5-triazin-2-yl)hydrazinecarbothioamido)-2-(6-hydroxy-3-oxo-3H-xanthen-9-yl)benzoic acid and 3-(4,6-diamino-1,3,5-triazin-2-ylthio)-N-(2-(3-(3′,6′-dihy-droxy-3-oxo-3H-spiro[isobenzofuran-1,9′-xanthene]-5-yl)thioureido)ethyl)propanamide. The limit of detection(10% inhibition) of the FPIA was 9.3 ng mL-1 and theIC50 (50% inhibition) value was 164.7 ng mL-1. Theantibody in the FPIA showed 21.2% cross-reactivity to the

fly-killing insecticide cyromazine, but had no cross-reactivity to other natural structurally related compounds.Recoveries, measured in spiked milk and milk powdersamples, ranged from 79.4 to 119.0%. Milk samplesfortified with melamine were analyzed by this method andconfirmed by high-performance liquid chromatography–mass spectrometry. Excellent recoveries and correlationwith spiked levels were observed, suggesting that thisimmunoassay could be applied to the screening ofmelamine residues in milk and milk powder after asimple dilution procedure.

Keywords Melamine . Fluorescence polarizationimmunoassay . Polyclonal antibody .Milk

AbbreviationsBSA Bovine serum albuminCAAT 2-chloro-4,6-diamino-1,3,5-triazineCAC Codex Alimentarius CommissionCR Cross-reactivityDCC N,N′-dicyclohexylcarbodiimideDMF N,N-dimethylformamideEDA EthylenediamineEDC 1-ethyl-3-(3-dimethyl-aminopropyl)carbodiimide

hydrochlorideEDF 1-(2-aminoethyl)-3-(3′,6′-dihydroxy-3-oxo-3H-

spiro[isobenzofuran-1,9′-xanthene]-5-yl)thioureaELISA Enzyme-linked immunosorbent assayFITC Fluorescein isothiocyanate isomer IFPIA Fluorescence polarization immunoassayHPLC High-performance liquid chromatographyLOD Limit of detectionMS Mass spectrometryNHS N-hydroxysuccinimideTLC Thin-layer chromatography

Q. Wang :Y.-M. Sun : Z.-F. Li :H. Liu : Z.-L. Xu :Y.-D. Shen :H.-T. Lei (*)The Institute of Food Quality and Safety, South ChinaAgricultural University,Guangzhou 510642, Chinae-mail: hongtao@scau.edu.cn

S. A. HaugheyInstitute of Agri-Food and Land Use, Queen’s University Belfast,Belfast BT9 5AG, UK

S. A. EreminDepartment of Chemical Enzymology, Faculty of Chemistry, M.V.Lomonosov Moscow State University,Moscow 119998, Russia

Anal Bioanal ChemDOI 10.1007/s00216-010-4599-2

Introduction

Melamine, with an IUPAC name of 1,3,5-triazine-2,4,6-triamine, is an organic base. In its chainlike “polymerized”form, melamine is an industrial chemical and has been usedfor decades in the manufacturing of dishes, plastic resins,flame-retardant fibers, components of paper and paper-board, and industrial coatings [1, 2]. Melamine is notapproved for addition to foods or feeds, nor is it permittedfor use as a fertilizer anywhere in the world. Ingestion ofmelamine may lead to reproductive damage and bladder orkidney stones, which in turn may lead to bladdercarcinogenesis [3]. The maximum amount of melamineallowed is 1 mg kg-1 in powdered infant formula and 2.5mg kg-1 in other foods and animal feed, according to therulings of the Codex Alimentarius Commission (CAC) andthe European Commission [4, 5].

Melamine is usually analyzed by physicochemical meth-ods based on gas or liquid chromatography. High-performance liquid chromatography (HPLC) has been pro-posed for quantitative determination of melamine; the limit ofdetection (LOD) was 0.1 mg kg-1 [6]. Ion-pair liquidchromatography coupled with electrospray tandem massspectrometry (MS) was used to determine residues ofmelamine in chard samples using a C18 reverse-phasecolumn; the LOD was 10 μg kg-1 [7]. Gas chromatogra-phy–MS was also used to detect melamine; the LODproposed was 10 μg kg-1 [8]. The instrumental methodsmeet the requirement of the CAC and are accurate andreliable but expensive, laborious, and time-consuming aswell as requiring professional personnel. To date, immuno-assay technology is increasingly used for screening foodcontaminants owing to its sensitivity, selectivity, timeefficiency, and cost-effectiveness. Over the past 20 yearsthe development of immunochemical methods and theirpotential applications, especially the enzyme-linked immu-nosorbent assay (ELISA), have increased significantly. Ingeneral, ELISA has many advantages over the othertechniques and allows direct analysis of a large numberof samples; however, ELISA is multistep procedure andis time-consuming. Simplifying the assay and minimizingthe analysis time are the primary goals in developingscreening methods for high-throughput sample analysis.The shift from the use of heterogeneous methods (withseparation) to homogeneous methods (without separa-tion) for assay simplification for routine applications hasshown great potential [9].

Fluorescence polarization immunoassay (FPIA) is one ofthe most extensively used homogeneous techniques, andmeets the requirements of a simple, reliable, fast, and cost-effective analysis. FPIA is a competitive immunoassaymethod based on the increase in the polarization of thefluorescence of a small fluorescein-labeled hapten (tracer)

when it is bound by a specific antibody. Recently, the use ofFPIA for the determination of pesticides [10, 11], biologicaltoxins [12, 13], mycotoxins [14, 15], and veterinary drugs[16, 17] in agricultural products and environmental sampleshas been reported.

In a previous study, an ELISA method for the detection ofmelamine was developed [18]. Yin et al. [19] synthesized ahapten, 4,6-diamino-1,6-dihydro-1,3,5-triazin-2-ylaminoace-tic acid, and developed an ELISA for melamine. Acompetitive 1,1′-oxalyldiimidazole chemiluminescent en-zyme immunoassay based on commercial immunoreagentsfor the screening of melamine in milk was also developed[20]. Several commercial ELISA kits for melamine havebecome available, such as the Abraxis melamine plate kit(Abraxis, Warminster, PA, USA), the Romer AgraQuantmelamine sensitive test kit (Romer Labs, Union, MO, USA),and the MaxSignal melamine enzymatic assay kit (BiooScientific, Austin, TX, USA). The LODs for melamine inliquid infant formula and wheat products were usually lessthan 1 mg L-1 and less than 2.5 mg kg-1, respectively [2].Immunoassays for the detection of melamine have also beenreported by other researchers [21–23]. However, no FPIAmethod for the detection of melamine has been publishedthus far. In the present study, the first development of anFPIA for the detection of melamine is described. The assayis based on an improved polyclonal antibody produced by anew melamine hapten (6-hydrazinyl-1,3,5-triazine-2,4-di-amine, hapten A) that has not previously been reported inthe literature and its application for detection of melamine inartificially spiked milk and milk powder samples.

Materials and methods

Reagents and instrumentation

General reagents and organic solvents were of analyticalgrade unless otherwise specified. Melamine, bovine serumalbumin (BSA), 1-ethyl-3-(3-dimethyl-aminopropyl)carbo-diimide hydrochloride (EDC), N-hydroxysuccinimide(NHS), N,N′-dicyclohexylcarbodiimide (DCC), N,N-dime-thylformamide (DMF), fluorescein isothiocyanate isomer I(FITC), and complete and incomplete Freund’s adjuvantswere purchased from Sigma (St. Louis, MO, USA). Ethyl-enediamine (EDA) was obtained from Jingke Company(Guangzhou, China). Silica gel G glass sheets (type GF254,layer thickness 0.2 mm) for thin-layer chromatography(TLC) were obtained from Taizhou Shenghua MaterialCorporation (Zhejiang, China). Microplates (96 wells) wereobtained from Jinchanhua Corporation (Shenzhen, China).2-Chloro-4,6-diamino-1,3,5-triazine (CAAT) and 6-hydrazinyl-1,3,5-triazine-2,4-diamine (hapten A) wereobtained from Shanghai Chaoyan Biotechnology Company

Q. Wang et al.

(Shanghai, China). Cyanuric chloride and cyanuric acidwere obtained from Accela Chembio Company (Shanghai,China). Atrazine was a gift from Shandong ZhongkeQiaochang Chemical Company (Shandong, China). Cyroma-zine was bought from Changzhou Zhineng Animal Pharma-ceutical Company (Changzhou, China). 3-(4,6-Diamino-1,6-dihydro-1,3,5-triazin-2-ylthio)propanoic acid (hapten B)and 6-(4,6-diamino-1,6-dihydro-1,3,5-triazin-2-ylamino) hex-anoic acid (hapten C) were prepared previously [18].

Fluorescence polarization values were recorded using aWallac 1420 VICTOR3 multilabel counter (PerkinElmer,USA). Ultraviolet spectra were recorded with a UV-3010spectrophotometer (Hitachi, Japan).

Synthesis of fluorescein-labeled tracers

Tracer A

Hapten A (5 mg, 45 μmol) was dissolved in 0.5 mL ofabsolute DMF. Triethylamine (50 μL) and FITC (5 mg,12.8 μmol) were added and the solution was stirredovernight at room temperature. A sample of reactionmixture (50 μL) was isolated by TLC using chloroformand methanol (4:0.4, v/v) as the eluent. The main yellowband at Rf=0.5 was scraped and extracted with 1 mL ofmethanol, thus giving 5-(2-(4,6-diamino-1,3,5-triazin-2-yl)-hydrazinecarbothioamido)-2-(6-hydroxy-3-oxo-3H-xanthen-9-yl)benzoic acid (tracer A) (Fig. 1). The tracer concentrationwas estimated spectrophotometrically at 492 nm, assumingthe absorbance in sodium borate buffer (50 mM, pH 8.0) tobe the same as for fluorescein (ε=8.78×104 M-1 cm-1) [11].

Tracers B and C

FITC (11.7 mg, 0.03 mmol) was dissolved in methanol(1 mL), containing triethylamine (10 μL). EDA (20 mg,0.3 mmol) was added and the mixture was stirred for 1 h atroom temperature. After concentration, the reaction mixturewas subjected to silica gel column chromatography usingethyl acetate and methanol (3:1, v/v) as an eluent. Theyellow band of the column was collected until nofluorescein was eluted. Following this, the column wastreated with anhydrous methanol, and the eluent wascollected and concentrated to dryness to obtain the linker1-(2-aminoethyl)-3-(3′,6′-dihydroxy-3-oxo-3H-spiro[iso-benzofuran-1,9′-xanthene]-5-yl)-thiourea (EDF).

Tracers B and C were synthesized using the active estermethod [10, 11]. Briefly, hapten B (8.6 mg, 40 μmol) orhapten C (10 mg, 40 μmol) was mixed and stirred withDCC (16.0 mg, 80 μmol) and NHS (9.7 mg, 80 μmol) inabsolute DMF (0.4 mL) and the reaction was allowed toproceed overnight. The precipitate was removed bycentrifugation, and EDF (4.5 mg, 10 μmol) was added to

the supernatant. After the color of the solution had changedto yellow, a small portion of the reaction mixture(approximately 50 μL) was subjected to TLC usingchloroform and methanol (4:1, v/v) as the eluent for 3-(4,6-diamino-1,3,5-triazin-2-ylthio)-N-(2-(3-(3′,6′-dihy-droxy-3-oxo-3H-spiro[isobenzofuran-1,9′-xanthene]-5-yl)thioureido)ethyl)propanamide (tracer B) and chloroformand methanol (5:1, v/v) as the eluent for 6-(4,6-diamino-1,3,5-triazin-2-ylamino)-N-(2-(3-(3′,6′-dihydroxy-3-oxo-2,3-dihydrospiro[indene-1,9′-xanthene]-5-yl)thioureido)eth-yl)hexanamide (tracer C) (Fig. 1). The main yellow band atRf=0.8 for tracer B and at Rf = 0.9 for tracer C was isolatedand the concentration was estimated spectrophotometricallyat 492 nm as described above.

Antibody production

Hapten A was coupled to BSA to produce an immunogenfor the immunization. Briefly, 3 mg (20 μmol) hapten Awas dissolved in 100 μL of absolute DMF. Then, haptenA solution was added dropwise with stirring to BSA(10 mg) in 1.9 mL of 0.9% (m/v) NaCl, followed by theaddition of 5 mg (25 μmol) EDC, mixed thoroughly andkept at room temperature overnight. The conjugates werepurified by dialysis against phosphate-buffered saline(0.05 M, pH 7.4) for 3 days with three changes per

Fig. 1 Structures of melamine haptens, fluorescein-labeled tracers,and immunogens. BSA bovine serum albumin, FITC fluoresceinisothiocyanate isomer I

Development of a FP immunoassay for melamine

day. The dialyzed product was spun at 5,000 rpm for15 min, and the supernatant was collected as immunogenA and stored at -20 °C until use.

Two New Zealand rabbits weighing 1.5–2.0 kg wereimmunized five times using immunogen A at intervals of14 days by the Guangdong Medical Laboratory AnimalCenter. Bleeds were taken from the rabbits on the eighthday after immunization, starting 40 days after the firstinjection, and tested for the presence of melamine anti-bodies using an indirect ELISA [18]. The serum obtainedwas divided into aliquots (1 mL), labeled as antibody A andstored at –20 °C until use.

Antibody B was obtained from immunogen B, whichwas prepared by coupling hapten C to carrier protein BSAusing the active ester method [18].

Fluorescence polarization immunoassay

Sodium borate buffer (0.05 M, pH 8.0) with 0.01%sodium azide was used for all FPIA experiments.Standard solutions of melamine and cross-reactants wereprepared by dilution of stock solutions of these com-pounds.

Antibody dilution curves were constructed by mixing thefluorescein-labeled tracer (100 μL, 1 nM solution insodium borate buffer) with the serially diluted antibody(100 μL). THe fluorescence polarization signal (mP) wasmeasured using a Wallac Victor3.

To achieve the optimal incubation time, a competitivekinetic curve was plotted. Ninety microliters of the tracersolution (0.8 nM) in sodium borate buffer and 20 μL ofstandard solution (0, 50, 100, and 500 ng mL-1,respectively) were mixed, then 90 μL of working antibodywas added, and the mP value was recorded successivelyversus time beginning at 1 min to obtain the kinetic curves

To construct the competitive calibration curves, the mela-mine standard (20 μL) at concentrations of 0.001–2,000 μgmL-1 was mixed with the labeled tracer (90 μL) and dilutedantibody (90 μL). The mP was measured after an appropriateincubation time, based on the kinetics of the curve. The mPvalues were plotted against the analyte concentration, and a

four-parameter equation, y ¼ A�Dð Þ= 1þ ðx=CÞBh i

þ D,

was used to fit the experimental sigmoidal curve in Origin7.5 for Windows. A and D are the maximal and minimal mP,B corresponds to the slope of the sigmoidal curve, and C is themelamine concentration at 50% of tracer binding (IC50). TheLOD was defined as the standard concentration that inhibited10% of tracer binding, and the detectable concentration rangewas defined as the standard concentration that inhibited 20–80% of tracer binding [24]. Cross-reactivity (CR) wascalculated according to the following equation: CR% ¼IC50ðmelamineÞ=IC50ðstructurally related compoundsÞ½ � � 100%.

Analysis of spiked samples

To evaluate the recovery of the FPIA developed, melamine-free milk and milk powder were spiked with melamine andrecoveries were determined by FPIA. For milk samples(1 mL), 0.5, 1, 5, and 50 μg of melamine standard wereadded, respectively. Each milk solution was centrifuged for10 min (12,000 rpm, 4 °C) to remove the fat in milk. Thelower liquid nonfat phase was diluted 1:30 in sodium boratebuffer before being used in FPIA. For milk powder samples(1 g), 1, 2, 10, and 50 μg of melamine standard were added,respectively. Methanol–H2O (1:1, v/v, 5 mL) solution wasadded to each sample. The mixture was vortexed for10 min, and centrifuged for 10 min (10,000 rpm, 4 °C).The middle clear liquid nonfat phase was diluted 1:12 insodium borate buffer before being used in FPIA. Blanksamples were prepared as described above but were notspiked with melamine. The melamine concentration inspiked samples was calculated after fitting of the standardcurve using the four-parameter logistic model.

To validate the FPIA, correlation studies between methodswere performed on 0.5, 1.0, 5.0, 10.0, and 50.0 mg L-1

melamine-spiked milk samples. Each sample was dividedtwo portions; one was analyzed using FPIA, and the otherwas analyzed by HPLC–MS. The HPLC–MS analysis wascarried out by Guangdong Testing Institute for ProductQuality (Shunde, Guangdong Province, China) using theprocedure described by Yin et al. [19].

Results and discussion

Production of immunoreagents

To obtain a specific and sensitive antibody, it was crucial todesign the suitable hapten structure carefully. The followingconsiderations have been shown to be important indetermining the characteristics of the resulting antibodies:(1) the ability of the hapten to mimic the analyte; (2) thefunctional groups selected for the hapten coupling; (3) thelength of the spacer bridge between the hapten and thecarrier protein; (4) the conjugation method [18, 25, 26]. Inprevious work, hapten C (Fig. 1) was conjugated to BSA asan immunogen for antibody (antibody B) preparation [18].However, antibody B strongly recognized cyromazine inELISA, which is not expected in specific analysis ofmelamine. In this work, a new hapten (hapten A, Fig. 1)was conjugated to the carrier protein BSA for use as animmunogen to prepare antibody A. This hapten wascommercially available and therefore complicated organicsynthesis was not needed. The hydrazine group of hapten Ais more reactive than the –NH2 of melamine. For thisreason, a single zero-length coupling reagent, EDC, was

Q. Wang et al.

used to produce a satisfactory conjugate (immunogen A)for the immunization. Also, without any additional cou-pling reagent, hapten A was linked directly to FITC for useas fluorescein-labeled tracer A. Hapten A was linked toFITC directly, but haptens B and C were covalently boundto EDF using the active ester method [10, 11]. Thus, tracersA, B, and C differed in not only the spacer length, but alsoin chemical structure between the hapten and the fluoro-phore. The spacer length between the hapten and FITC intracer A is the shortest among the three tracers. Since FITChas a big volume, this might result in a stronger sterichindrance to the binding between tracer A and eachantibody. However, the spacer length in tracer C is thelongest, which might make the steric effect weaker andfacilitate the binding between tracer C and each antibody aswell. Besides the spacer length, tracer B had a sulfur atom,which has different electronic characteristics from thehydrazine group in tracer A or the amino group in tracerC. The electronic effect was expected to result in differentability of binding to the antibody. Since all three tracerswould be used as the competing molecules in the FPIA,both the length difference and the electronic differenceamong spacers of tracers were proposed to result indifferent binding strength between the tracer and theantibody.

To produce antibodies to melamine, hapten A–BSA(immunogen A) was injected into two rabbits. After fiveimmunizations, the antibody was collected and the titer wastested by FPIA. The antibody from rabbit 1 demonstrated ahigher titer and better sensitivity in FPIA performed withthe three tracers than the antibody from rabbit 2 (data notshown). Therefore, the antibody from rabbit 1, namedantibody A, was selected for the subsequent investigation.

Comparison of tracers and antibodies

Antibody B was produced previously, using immunogen B(Fig. 1), which was prepared with hapten B and BSA [18].To obtain better sensitivity and specificity of the immuno-assay, two antibodies (antibody A and antibody B) and thethree tracers were initially evaluated and compared. The

specificity of an immunoassay generally depends on theantibody employed, whereas the assay sensitivity is alsoaffected by the competitive hapten (tracer) [27]. Thus, theantibody and the tracer were two key factors to consider inthe development of an FPIA. The working antibodyconcentration is an important parameter for FPIA develop-ment and the antibody dilution titer to show 70% or 50%binding response was usually appropriate for FPIA devel-opment [24, 28]. In this study, the antibody workingconcentration was assessed by FPIA using serially dilutedantibody with a fixed concentration of the fluorescence-labeled tracer (1 nM). An antibody dilution titer showing70% binding response to the tracer was chosen as theappropriate one to perform the initial evaluation with thethree tracers. With the fixed tracer concentration (1 nM) toobtain 70% antibody binding, the appropriate antibody Adilution titers were 1:160, 1:210 and 1:150 for tracers A, Band C, respectively. However, the optimized antibody Bdilution titers were 1:300, 1:230, and 1:150 for tracers A, B,and C, respectively. To obtain the same binding (70%), thethree tracers needed different antibody dilution titers,suggesting that the three tracers had different relativeaffinities for each antibody. At the same tracer concentra-tion, low antibody dilution means low relative affinitybetween the tracer and the antibody, and low affinity mayresult in the analyte competing for the antibody bindingsites in a competitive immunoassay, leading to bettersensitivity [29]. Tracer C among the three tracers showedthe lowest relative affinity for both antibody A andantibody B in this case; thus, it could be concluded thattracer C would demonstrate most sensitivity in the FPIA.To confirm this, the ability of melamine to compete withthe three tracers for antibody binding was investigated bymeans of calibration curves.

Figure 2 and Table 1 show the characteristics of the twoantibodies and the three tracers. When combining tracer Cwith antibody A or antibody B, it was found that both IC50

values were the lowest. Comparison of the IC50 of 444.9 ngmL-1 for antibody A and 7,934.0 ng mL-1 for antibody Bimplied that the sensitivity (IC50) had improved by about 18times owing to the choice of hapten used in the immuno-

Table 1 Characteristic comparison of the three tracers and two antibodies

Tracer Antibody A Antibody B

Titer IC50 (ng/mL) LOD (ng/mL) Detectableconcentrationrange (ng/mL)

Titer IC50 (ng/mL) LOD (ng/mL) Detectableconcentrationrange (ng/mL)

Tracer A 1:160 174,196.7 29,872.3 55,345.1–426,973.5 1:300 86,0397 2,455.5 10,427.1–425,374.4

Tracer B 1:210 65,909.9 1,615.9 6,063.9–278,987.6 1:230 22,727 1,546.5 4,321.0–74,348.5

Tracer C 1:150 444.9 36.3 89.3–2,776.7 1:150 7,934 267.0 871.0–46,661.1

LOD limit of detection

Development of a FP immunoassay for melamine

gen. Similar improvements were also found with both theLOD (from 267.0 ng mL-1 with antibody B to 36.3 ng mL-1

with antibody A, approximately 8 times improvement), andthe dynamic concentration range (871.0–46,661.1 ng mL-1

with antibody B and 89.3–2,776.7 ng mL-1 with antibody

A) as well. A heterogeneous immunoassay format can oftenresult in antibodies having a relatively higher affinity forthe analyte [30–33]. Tracer C, with a six-carbon spacer armas a heterogeneous competitor, gave better assay sensitivitythan tracers A and B. This indicated that a longer spacermight lead to a more sensitive assay.

It was interesting to note that the combination of coatedhapten B and antibody B showed better sensitivity than thecombination of coated hapten C and antibody B in ourprevious ELISA work [18], which is not in agreement withtheir performance in the present FPIA. This suggests thatsometimes there is a performance difference betweenhomogeneous direct FPIA and the heterogeneous indirectELISA. Presumably, the short spacer arms of tracers A andB resulted in steric hindrance of tracer binding when thelarge fluorescent dye residue is brought close to theantigen-binding site of the antibody. Since the heterologouscombination of antibody A and tracer C displayed the bestassay sensitivities, this combination was used for furtherinvestigation.

Figure 3 shows the competitive kinetics curves. It can beseen that the kinetic reaction was nearly stable at around4 min for 500 ng mL-1 standard, at around 10 min for 100ng mL-1, but at around 12 min for 50 and 0 ng mL-1.Therefore, it is concluded that different concentrations ofcompetitors need different equilibration times, and higherconcentrations would equilibrate faster. However, evenafter 1 min, the signals showed a significant difference fordifferent concentrations, which means that 1 min isrecordable for the homogeneous immunoreactions. Tominimize the interassay coefficient of variation, a longertime was usually better for manual operation because thesignal at the earlier stage was successively changing. Thus,15 min was selected in the present FPIA.

Figure 4 shows a typical optimal calibration curve. Theconcentration of tracer C in the FPIA was 0.8 nM and the

10-1 100 101 102 103 104 105 106 107

10-1 100 101 102 103 104 105 106 107

140

160

180

200

220m

P

Melamine / ng mL-1

Tracer A Tracer B Tracer C

(a)

160

180

200

220

240

260

280

300 Tracer A Tracer B Tracer C

mP

Melamine / ng mL-1

(b)

Fig. 2 Initial fluorescence polarization immunoassay (FPIA) calibra-tion curves for melamine using the three traces in combination withantibody A (a) and antibody B (b). Each point of the curve representsthe mean ± the standard deviation (SD) of three assays. mPfluorescence polarization signal

0 5 10 15

150

160

170

180

190

200

210

220

230

mP

Time / min

0 ng/mL 50 ng/mL 100 ng/mL 500 ng/mL

Fig. 3 The competitive kinetic curves of tracer C and antibody A

100

101

102

103

104

140

160

180

200

220

mP

Melamine / ng mL-1

IC50 =164.7 ng mL-1

Fig. 4 FPIA calibration curve for melamine using antibody A andtracer C. Each point of the curve represents the mean ± SD of threeassays

Q. Wang et al.

Table 2 Cross-reactivity (CR) of antibody to melamine and related compounds determined by fluorescence polarization immunoassay (FPIA)

Compound StructureIC50

(nM)

CR

(%)

Melamine 1.31 100

Cyromazine 6.17 21.2

Hapten C 8.85 14.8

CAAT 20.31 6.4

Hapten A 23.36 5.6

Hapten B 41.50 3.2

Ammeline 81.01 1.6

Ammelide 1584.76 0.1

Cyanuric

acid6569.74 0.02

Cyanuric

chlorideND 0.01

Atrazine ND 0.01

CAAT 2-chloro-4,6-diamino-1,3,5-triazine, ND not determined (the IC50 was too high to be available because the calibration curve could not befitted)

Development of a FP immunoassay for melamine

reaction time was 15 min. The LOD (10% inhibition) formelamine was 9.3 ng mL-1, the IC50 value was 164.7 ngmL-1, and the working range (20–80% inhibition) was24.1–964.4 ng mL-1. Therefore, the performance of thisFPIA could meet the maximum melamine amount allowedin foods and animal feed set by the CAC and the EuropeanCommission [4, 5].

Specificity

The CRs for several compounds structurally related tomelamine were tested and calculated as the ratio of the IC50

value of the melamine standard to the IC50 of thecompounds tested. The results are presented in Table 2.Among all the structurally related compounds tested in thepresent study, the CR of melamine to antibody A washighest in FPIA format assays as expected, followed bycyromazine (21.1%), hapten C (14.8%), CAAT (6.4%),hapten A (5.6%), hapten B (3.2%), ammeline (1.6%),ammelide (0.1%), and cyanuric acid (0.02%). Cyanuricchloride and atrazine showed 0.01% or less CR. Comparingthe CRs of melamine and CAAT and ammeline with theirstructures, even when one –NH2 was changed to –Cl or –

OH, the CR decreased from 100% for melamine to 6.4%for CAAT or 1.6% for ammeline. This suggested that thethree –NH2 groups in the triazine ring are crucial structuralfactors to determine the affinity of the analyte for theantibody. Additionally, the CR of atrazine was very muchlower than that of CAAT. Probably, the steric hindrance,resulting from two alkyl groups in triazine, should beresponsible for the CR change. However, the CR of CAATwas found to be higher than that of ammeline, which couldbe due to the different electronic effects of –Cl and –OH onthe triazine ring. The difference between the CR ofcyromazine and that of hapten C can be attributed to thenegative charge of the acid anion. In one case reported, theinfluence of the negative charge was still noticeable aftereven changing the spacer length from two atoms to sevenatoms [34]. The difference between cyromazine and haptenA may be attributed to the positive charge of the aminocation of hydrazine.

In a previous report [18], with hapten C used as thecoating antigen (hapten C–ovalbumin conjugate), antibodyB had an extremely high CR to hapten C (1,652.9%) andcyromazine (267.6%). On the basis of the analysis of thestructure–activity relationship, it was concluded that theelectronic effect resulted in more influence than the stericeffect of the spacer in melamine ELISA, and that therewould probably be a decrease in the CR if hapten A wereused in an immunogen for antibody production [18]. Toverify this, in the present work antibody from hapten Awasprepared and hapten C was used as a competitor (tracer).The CR of the antibody A to cyromazine and hapten Cdecreased from 267.6% and 1,652.9% in ELISA to 21.2%and 14.8% in this FPIA (Table 2), respectively. This was ingood agreement with our previous conclusion [18].

Interestingly, although immunogen Awas prepared usinghapten A, it is surprising that hapten A does not show ahigh CR; indeed, the CR was found to be lower than theCR to melamine, cyromazine, hapten C, and CATT in theFPIA format. The steric effect of the hydrazine group inhapten A on the antibody recognition appeared to be much

Table 3 Recovery of melamine from spiked sample determined by FPIA

Sample Spiked level Observed value Recovery (%, n=3) Meanrecovery (%)

CV (%) Mean CV (%)

Milk 0.5 mg L-1

1.0 mg L-1

5.0 mg L-1

50.0 mg L-1

0.40±0.08 mg L-1

0.84±0.14 mg L-1

4.93±0.60 mg L-1

52.80±13.44 mg L-1

79.4±15.783.6±13.698.5±12.0

105.6±26.9

92.9 19.7216.3212.1825.45

18.4

Milk powder 1.0 mg kg-1

2.0 mg kg-1

10.0 mg kg-1

50.0 mg kg-1

1.19±0.27 mg kg-1

1.73±0.35 mg kg-1

10.41±1.55 mg kg-1

55.39±10.80 mg kg-1

119.0±26.886.6±17.7104.1±15.5110.8±21.6

105.4 22.5220.4014.9319.49

18.9

CV coefficient of variation

0 10 20 30 40 50 60

0

10

20

30

40

50

HP

LC /

mg

L-1

FPIA / mg L-1

r2=0.9986

y=0.7638x+0.7078

Fig. 5 Correlation between the result from FPIA and that from high-performance liquid chromatography (HPLC) for milk samples

Q. Wang et al.

smaller than that of the cyclopropyl group in cyromazine.Owing to the hydrazine amino group of hapten A beingconjugated in immunogen A it was therefore converted to aamide group by reaction with the carboxyl groups on BSAunder the condensing action of EDC, leading to differentelectronic states between free and conjugated hapten A.Thus, we presume that the electronic effect of the hydrazineamino group plays a more important role than the sterichindrance in this case, resulting in the weaker recognitionbetween the free hapten A and antibody A.

Determining the contribution that each part of thetriazine molecule makes to antibody binding is importantfor understanding the phenomenon of CR between relatedcompounds. Our CR study using the FPIA shows that thethree –NH2 in the triazine ring must be present for antibodybinding to occur, and the differences in antibody recogni-tion of different triazine analytes could be caused by stericor/and electronic factors. As the ELISA showed, theelectronic effect still appeared to influence more theantibody recognition in this FPIA.

Analysis of spiked samples

Immunoassays are susceptible to interferences with differ-ent components existing in some matrices [33–35]. Matrixeffects were estimated in this research using two types ofmatrix systems, including milk and milk powder purchasedin local markets. As reported previously, the most commonways to reduce such matrix effects are selective extraction(“cleanup”) and dilution to bring the interfering substancesbelow a concentration that would interfere with the assay[35]. Cleanup is relatively time-consuming and expensiveand would counteract the advantage of the FPIA method inbeing a rapid and simple screening application. The dilutionapproach was successfully applied to extract melaminefrom milk and milk powder samples for the ELISA method[19].

In this study, milk was prepared by centrifugation toget the defatted solution before dilution; melamine inmilk powder was extracted by methanol–H2O solution.After extraction, the optimum dilution factor of theextracts of milk and milk powder samples using sodiumborate buffer was determined. It was found that with 30-fold and 12-fold dilutions with sodium borate buffer formilk and milk powder, respectively, the diluted extractsshowed a similar mP to the signal of sodium borate buffer(data not shown). The two matrix samples were spiked withdifferent levels of melamine to determine the recovery andcoefficient of variation. The results are summarized inTable 3. The FPIA gave satisfactory recovery that rangedfrom 79.4 to 119.0%. The means coefficients of variationfor milk and milk powder were 18.4% and 19.3%,respectively. Although the coefficient of variation is a little

high, this immunoassay is a screening method and also hasan undoubted practical advantage over methods requiringin-depth sample purification.

Correlation studies between methods were carried out onfive milk samples each spiked with melamine. All sampleswere analyzed by both FPIA and HPLC-MS. Samples wereidentified by a code number, so the melamine concentrationwas unknown to the analysts. Furthermore, FPIA analysiswas performed in our laboratory according to the proceduredescribed earlier, but HPLC-MS analysis was performed ina certificated laboratory (Guangdong Testing Institute forProduct Quality, Shunde, Guangdong Province, China), andthe results were compared only after the all the analyses hadbeen completed. The analytical results obtained with thesetwo types of samples by the two methods are shown inFig 5. Linear regression analysis yielded an excellentcorrelation between methods (r2=0.99 and slope 0.76)(Fig. 5). This suggested that the FPIA developed could beused for the accurate detection of melamine withoutcomplicated sample cleanup.

Conclusions

In conclusion, to improve the performance of anantibody to melamine, a new hapten strategy was utilizedand the antibodies obtained were found to have high titerand better sensitivity and specificity than our previousantibodies. The FPIA developed is able to detectmelamine in milk at levels of regulatory relevance, withaccuracy and precision comparable to those obtainedwith the reference method. After the sample treatment,the immunoassay takes 15 min to performed. An issue ofmajor importance is the ability of this FPIA to detectmelamine without complicated cleanup. This FPIA maybe easily included as a complementary method inmelamine regulatory programs. FPIA could be consid-ered to be a valuable method for screening purposes.Overall, this work should reasonably contribute toincrease the acceptance of FPIA methods among analyt-ical chemists involved in melamine residue analysis infoods.

Acknowledgements This work was supported by the NationalDepartment Public Benefit Research Foundation (201003008-08)and China Guangdong Provincial Science and Technology Projects(cgzhzd1005, 2009B040500002).

References

1. Andersen WC, Turnipseed SB, Karbiwnyk CM, Clark SB, MadsonMR, Gieseker CM, Miller RA, Rummel NG, Reimschuessel R(2008) J Agric Food Chem 56:4340–4347

Development of a FP immunoassay for melamine

2. Garber EAE (2008) J Food Prot 71:590–5943. Ogasawara H, Imaida K, Ishiwata H, Toyoda K, Kawanishi T,

Uneyama C, Hayashi S, Takahash M, Hayashi Y (1995)Carcinogenesis 16:2773–2777

4. Word Health Organization (2010) International experts limit mela-mine levels in food. New guidance to help improve food safetyprovided by UN food standards commission. http://www.who.int/mediacentre/news/releases/2010/melamine_food_20100706/en/print.html. Accessed 6 Jul 2010

5. European Commission Decision 2008/921/EC, amending Deci-sion 2008/798/EC

6. Zhang J, Wang Y, He Y, Jiang T, Yang H, Tan X, Kang R, Yuan Y,Shi L (2010) Anal Biochem 397:212–217

7. Sancho JV, Ibanez M, Grimalt S, Pozo OJ, Hernandez F (2005)Anal Chim Acta 530:237–243

8. Xia X, Ding S, Li X, Gong X, Zhang S, Jiang H, Li J, She J(2009) Anal Chim Acta 651:196–200

9. Smith DS, Eremin SA (2008) Anal Bioanal Chem 391:1499–150710. Kolosova AY, Park JH, Eremin SA, Kang SJ, Chung DH (2003) J

Agric Food Chem 51:1107–111411. Shim WB, Yakovleva ME, Kim KY, Nam BR, Vylegzhanina ES,

Komarov AA, Eremin SA, Chung DH (2009) J Agric Food Chem57:791–796

12. Zezza F, Longobardi F, Pascale M, Eremin SA, Visconti A (2009)Anal Bioanal Chem 395:1317–1323

13. Goryacheva IY, Rusanova TY, Burmistrova NA, Saeger S (2009)J Anal Chem 64:768–785

14. Shim WB, Kolosova AY, Kim YJ, Yang ZY, Park SJ, EreminSA, Lee IS, Chung DH (2004) Int J Food Sci Technol 39:829–837

15. Nasir MS, Jolley ME (2003) Comb Chem High ThroughputScreen 6:267–273

16. Gasilova NV, Eremin SA (2010) J Anal Chem 65:255–25917. Zhang SX, Wang ZH, Nesterenko IS, Eremin SA, Shen JZ (2007)

Int J Food Sci Technol 42:36–4418. Lei HT, Shen YD, Song LJ, Yang JY, Chevallier OP, Haughey SA,

Wang H, Sun YM, Elliott CT (2010) Anal Chim Acta 665:84–90

19. Yin WW, Liu JT, Zhang TC, Li WH, Liu W, Meng M, He FY,Wan YP, Feng CW, Wang SL, Lu X, Rimo X (2010) J Agric FoodChem 58:8152–8157

20. Choi JH, Kim YT, Lee JH (2010) Analyst 135:2445–245021. Wang ZY, Ma X, Zhang LY, Yang WJ, Gong LM, He PL, Li Z

(2010) Anal Chim Acta 662:69–7522. Liu JX, Zhong YB, Liu J, Zhang HC, Xi JZ, Wang JP (2010)

Food Control 21:1482–148723. Kim B, Perkins LB, Bushway RJ, Nesbit S, Fan T, Sheridan R,

Greene V (2008) J AOAC Int 91:408–41324. Murtazina NR, Eremin SA, Mozoleva OV, Everest SJ, Brown AJ,

Jackman R (2004) Int J Food Sci Technol 39:879–88925. Vallejo RP, Bogus ER, Mumma RO (1982) J Agric Food Chem

30:572–58026. Goodrow MH, Hammock BD (1998) Anal Chim Acta 376:83–9127. Carlson RE (1995) Immunoanalysis of agrochemicals. In: Nelson

JO, Karu AE, Wong RB (eds) Emerging technologies. ACSsymposium series 586. American Chemical Society, Washington,pp 140–152

28. Nartova YV, Ermolaeva TN, Fleisher MR, Abuknesha R, EreminSA (2008) J Anal Chem 63:499–505

29. Wortberg M, Goodrow MH, Gee SJ, Hammock BD (1996) JAgric Food Chem 44:2210–2219

30. Choi MJ, Lee JR, Eremin SA (2002) Food Agric Immunol14:107–120

31. Roucairol C, Azoulay S, Nevers MC, Créminon C, Grassi J,Burger A, Duval D (2007) Anal Chim Acta 589:142–149

32. Paknejad M, Rasaee MJ, Mohammadnejad J, Pouramir M,Rajabibazl M, Kakhki M (2008) J Toxicol Sci 33:565–573

33. Kolosova AY, Park JH, Eremin SA, Park SJ, Kang SJ, Shim WB,Lee HS, Lee YT, Chung DH (2004) Anal Chim Acta 511:323–331

34. Zeck A, Weller MG, Niessner R (1999) Fresenius J Anal Chem364:113–120

35. Skerritt JH, Amita Rani BE (1996) Immunoassays for residueanalysis. In: Beier RC, Stanker LH (eds) Food safety. ACSsymposium series 621. American Chemical Society, Washington,pp 29–43

Q. Wang et al.