Download - Determination of polychlorinated biphenyl compounds in electrical insulating oils by enzyme immunoassay

Analytica Chimica Acta 422 (2000) 167–177

Determination of polychlorinated biphenyl compounds inelectrical insulating oils by enzyme immunoassay

In Soo Kim, Steven J. Setford, Selwayan Saini∗Cranfield Centre for Analytical Science, IBST, Cranfield University, Cranfield, Bedford, MK43 0AL, UK

Received 7 April 2000; received in revised form 21 June 2000; accepted 7 July 2000

Abstract

The development and performance of a competitive indirect immunoassay for the quantitation of polychlorinated biphenylcompounds (PCBs) in insulating oils is described. Reagent preparation and the assay characterisation, optimisation and val-idation steps are described. The dynamic range of the assay for Aroclors 1254 and 1260 in methanol was 30–1000 ng ml−1

with 50% signal inhibition values of 217 and 212 ng ml−1, respectively. Impending legislation in the UK is likely to de-cree that oils containing >50mg ml−1 PCBs be considered contaminated. Assay sensitivity increased with the degree ofPCB chlorination. The assay of structurally related compounds of environmental concern yielded cross-reactivity values of<0.6%. The immunoassay proved reliable for the analysis of diluted transformer oils containing >35mg ml−1 PCB (neat)but over-estimated PCB levels in diluted oils containing<10mg ml−1 of neat analyte. The oils required pre-treatment usingeither solid-phase extraction or washing with KOH–ethanol/sulphuric acid to remove matrix interferents. The analytical per-formance of the assay was compared against a commercially available semi-quantitative immunoassay kit for PCBs in soiland water. © 2000 Elsevier Science B.V. All rights reserved.

Keywords:Polychlorinated biphenyl compounds; Insulating oils; Enzyme immunoassay; Legislation; Organic phase; Extraction

1. Introduction

There is increasing global concern regarding theutilisation and discharge of polychlorinated biphenyl(PCB) compounds into the environment. PCBs areroutinely used as additives in various oil-based prepa-rations and are important in the manufacture ofproducts as diverse as plastics and pesticides. Theubiquitous nature of these materials coupled to theirtoxicity and recalcitrant nature has led to signifi-cant interest in the development of simple, rapid and

∗ Corresponding author. Tel.:+44-1234-752401;fax: +44-1234-752401.E-mail address:[email protected] (S. Saini).

low-cost methods for their analysis in a wide rangeof environmental matrices.

Structurally, a PCB is a chlorinated biphenyl com-pound with the general formula C12H(10−n)Cln. PCBsgenerally occur as mixtures, wheren can vary from1 to 10. The 10 sites available for possible chlorinesubstitution result in 209 possible PCB compounds orcongeners [1]. There is now considerable concern re-garding the presence of PCB congeners in insulatingoils used within large-scale electrical supply systems.The chemical inertness, heat resistance, non-flam-mability, low vapour pressure and dielectric propertiesof PCBs led to their widespread usage as insulator oiladditives from the 1930s until the 1970s when con-cerns regarding their carcinogenicity led to this prac-tice being banned in the US and elsewhere [2]. Despite

0003-2670/00/$ – see front matter © 2000 Elsevier Science B.V. All rights reserved.PII: S0003-2670(00)01068-0

168 I.S. Kim et al. / Analytica Chimica Acta 422 (2000) 167–177

this fact, PCB-contaminated oils are still commonlyencountered partly because some electrical units havenever been refilled with PCB-free insulating oils,whilst those that have may still be contaminated dueto inadequate plant and loading-line cleaning proce-dures. Some PCB contamination is believed to occurdue to the re-use of incompletely reconditioned oil.Since attempts to identify a particular plant as con-taminated have been unsuccessful, the only recourseof action has been to chemically analyse oils.

The current UK action plan [3] dictates that organi-sations with electrical equipment containing more than5 l of oil, contaminated with more than 50mg ml−1

PCB will need to reduce these levels to comply withimminent new regulations. Indeed, the most recentEC Directive on the disposal of PCBs (96/59/EC) de-fines a range of chlorinated diphenyl and terphenylcompounds as PCBs. Consequently, this legal defini-tion considers any preparation containing more than50mg ml−1 of PCB or PCB equivalent to be treatedas a pure PCB preparation.

Current PCB measurement methods are either non-specific or utilise complex laboratory-based instru-mental techniques [4]. The former methods measuregeneral properties of the PCB analyte such as totalchlorine content whilst the latter methods are gener-ally more time consuming and expensive, typicallyrequiring sample preparation, chromatographic sepa-ration and detection. Whilst the specific approachesare reproducible and of high sensitivity, the increasingincidence of PCB compounds in the environment hascreated the need for more rapid, simpler and low-costanalytical procedures. Ideally, these techniques shouldhave the durability and flexibility to be automated andapplied either in the laboratory for routine analysis orin the field where assay speed and simplicity are desir-able for the immediate implementation of appropriateremediation procedures.

Immunochemical techniques offer a simple, low-cost means of routinely and specifically measur-ing compounds in decentralised locations and haveconsequently been used for PCB analysis [5–13].These methods have primarily been developed forthe determination of PCBs in predominantly polarmatrices such as soil, water and milk. Indeed, twocommercially available enzyme-linked immunosor-bent assay (ELISA) test kits for PCBs based on acompetitive-format tube type assay are also widely

available (RaPID assay® and Envirogard®, SDI Eu-rope Ltd., Alton, UK). However, both of these sys-tems have been designed to work in essentially polarenvironments and are only semi-quantitative, yieldingonly a concentrationrange as opposed to a discreetnumerical value. None of the above approaches hasconsidered the analysis of oils for PCBs. Conse-quently, this paper describes the development of arapid extraction and ELISA-based method for thequantitation of PCBs in electrical insulating oils. Akey objective of the work has been to develop a simpleprotocol, amenable to both field-based and automatedlaboratory-based high-throughput analysis. Whilst thelower chlorinated Aroclors are usually predominantin transformer oils, the assay was optimised usingAroclors 1254 and 1260. The optimised procedurewas compared against a commercial PCB ELISA kit.

2. Materials and methods

2.1. Reagents

General chemical and biological reagents were ofanalytical grade and were purchased from Sigma-Aldrich or Merck (both Poole, Dorset, UK). Deionised-reverse osmosis water was purified using an Elgas-tat system (Elga, High Wycombe, UK). Phosphatebuffered saline-Tween (PBST) solutions were pre-pared from 10 mM buffer salts, pH 7.4, 0.15 M NaCland 0.05% (v/v) Tween 20. The carbonate buffer usedwas 0.2 M at pH 9.6. Aroclors 1242, 1248, 1254, 1260and 1262 (Ultra Scientific, North Kingstown, RI) weresupplied as 100mg ml−1 solutions in methanol anddiluted in methanol as required. For cross-reactivitystudies, the structurally similar non-PCB compoundspentachlorophenol (PCP), 2,4-dichlorophenoxyaceticacid (2,4-D), 2,4,5-trichlorophenoxyacetic acid (TC-PAA) and 2-(2,4,5-trichlorophenoxy)propionic acid(TCPPA) from Sigma-Aldrich were used. The Na-tional Grid Co. (Leatherhead, UK) donated an unusedPCB-free transformer oil and five used oils containingAroclors. The used oils had the following gas chro-matographically determined PCB concentrations inmg ml−1: A, 4; B, 35; C, 51; D, 10; E, 6. Oil clean-upwas facilitated using C18 solid phase extraction (SPE)columns containing 500 mg of resin in 3 ml capacitycolumns, from IST Ltd. (Hengoed, Clwyd, UK).

I.S. Kim et al. / Analytica Chimica Acta 422 (2000) 167–177 169

Anti-PCB monoclonal antibody (MAb, ResearchDiagnostics Inc., Flanders, NJ) was diluted 10-fold to500mg ml−1 antibody protein, aliquoted and storedat −20◦C with no significant loss in activity over 1year. Goat anti-mouse antibody-horseradish perox-idase (GAM-HRP) conjugate (Sigma-Aldrich) wassimilarly aliquoted and stored at−20◦C and diluted in10% (v/v) foetal bovine serum (FBS, Sigma-Aldrich)with PBST as required and was stable for 1 month at4◦C. Coating antigen was synthesised by conjugat-ing bovine serum albumin (BSA) to TCPPA (Fluka)using 1-ethyl-3,3-dimethylaminopropylcarbodiimidehydrochloride (EDC) and adsorbed onto microtitrewell walls. Residual adsorption sites were blockedusing SuperBlock PBS buffer (Pierce, Chester, UK).TCPPA and BSA solutions were prepared in 0.1 M2-[N-morpholino] ethane sulphonic acid (MES) buffer,pH 4.5, containing 50% (v/v)N,N′-dimethylformamide(DMF). Conjugate protein concentrations were deter-mined using the Coomassie Plus protein assay (Mod-ified Bradford method, Pierce).o-Phenylenediamine(OPD) substrate was purchased as 5 mg tablets fromSigma-Aldrich and dissolved in 0.15 M citrate buffer,pH 5 (citric acid 7.3 g, Na2HPO4·2H2O 11.86 g).1-Step TURBO TMB was purchased from Pierce andused as received. Hydrogen peroxide stock solution(30% w/w) was from Sigma-Aldrich.

2.2. Instrumentation

Two types of microtitre plate were evaluated:Polysorb® 96-well (A/S Nunc, Roskilde, Denmark)and ImmunoWare® 8-well EIA Strip Plates (Pierce).During incubation, the plates were mechanicallyshaken with an iEMS incubator/Shaker HT (Lab-Systems, Finland) before being washed using aneight-channel manual washer (Nunc). The extent ofcolour development in the plates was determined us-ing a Titertek Multiscan MCC Microwell plate reader(LabSystems). Coating antigen conjugate was recov-ered using a Sephadex G-25 size exclusion column(PD-10, Pharmacia, Sweden).

2.3. Preparation and immobilisation of coatingantigen

Coating antigen was prepared by reacting EDCwith the carboxyl group on the PCB analogue TCPPA.

The amine-reactiveo-acylisourea intermediate wasthen reacted with amine groups on BSA to form astable carrier-protein–analogue conjugate. First, 1 mlof 5 mg ml−1 TCPPA and 400ml of 10 mg ml−1 BSAin MES buffer, pH 4.5, were stirred with 200ml offreshly prepared 10 mg ml−1 EDC in water for 2 h at25◦C. The Sephadex G-25 column was pre-washedwith 0.2 M carbonate buffer, pH 9.6, and loaded withthe reaction mixture. Conjugated material was elutedusing 0.2 M carbonate buffer and collected as 0.5 mlfractions. The protein concentration of each fractionwas determined using the Coomassie protein assayaccording to the manufacturer’s instructions usingBSA as the calibration standard. Fractions 4–8 con-tained the greatest amount of protein and were pooled;the combined protein concentration was 1.3 mg ml−1.The conjugate was diluted to 1 mg ml−1 in 0.2 Mcarbonate buffer and stored at 4◦C until required.The TCPPA:BSA conjugation ratio was determinedphotometrically to be approximately 20:1.

Coating antigen stock solution was appropriatelydiluted in carbonate buffer and 150ml volumes dis-pensed into microtitre plate wells. The wells weresealed and incubated at 25◦C for 4 h to allow adsorp-tion of coating antigen onto the well walls, washedthree times with PBST and excess fluid removed byinversion and rapping on an absorbent towel. Wellswere blocked with SuperBlock PBS buffer for 30 minat 25◦C, then washed and dried as before. The plateswere sealed and stored at 4◦C until required.

2.4. ELISA procedure

A 90ml volume of appropriately diluted anti-PCBMAb in PBST and 10ml of sample or standard inmethanol was added to each TCPPA-coated well. Theplates were then incubated at 37◦C (primary incu-bation) for a given time, then washed and rap-driedas before. Next, 100ml volumes of 1:4000 dilutedGAM-HRP conjugate in FBS-PBST were added toeach well before incubation at 37◦C (secondary incu-bation) for a given time, washing and rap-drying.

HRP activity was determined using either 1-StepTURBO TMB or OPD. The former substrate wassupplied ready for use, whilst the latter was preparedimmediately before required by adding 15ml of 30%(v/v) H2O2 to 15 ml of 1 mg ml−1 OPD in 0.15 Mcitrate buffer. In both cases, 100ml of substrate was


dispensed into each well. The plates were incubatedfor 30 min or until the absorbance (optical den-sity, OD) of the standard zero concentration blankreached 1.0 after which the reaction was stopped byadding 100ml of 2 M H2SO4. The plate was read at492/630 nm. All tests were performed a minimum ofthree times.

2.5. Data evaluation

The logit-log model, the most widely used proce-dure for immunoassay data evaluation [14], was usedto analyse the data. The model represents a continu-ous sigmoidal function with a single inflection point,described by the equation

y = a − d

1 + (x/c)b+ d

wherea represents the maximum current at zero ana-lyte (upper asymptote),b the slope of the linear portionof the sigmoidal curve,c the analyte concentration atmid-point (50% signal reduction), andd the residualcurrent at infinite dose (lower asymptote; backgroundcurrent+ non-specific binding). These constant val-ues were calculated using the curve fit (math) func-tion of Sigma plot for Windows (Jandel Scientific).The practical quantitative range of the ELISA wasdefined as the linear portion of the sigmoidal curve,which was also calculated using Sigma plot. The assaylimit of detection (LOD) was defined as the analyteconcentration corresponding to ODzero analyte− (2 ×S.D.zero analyte). Accuracy is a measure of the system-atic error in the assay and was reported as % bias,calculated as 100%× ((measured value−true value)/true value). Cross-reactivity was calculated as 100%×(IC50-value Aroclor 1254 standard/IC50-value cross-reacting sample). Extraction efficiencies were calcu-lated as 100%× (measured value/expected value).

2.6. Transformer oil preparation

2.6.1. Direct dilutionOil samples were diluted 10-fold in 2-propanol and

50-fold in methanol and assayed.

2.6.2. Liquid–liquid extractionOil samples (100ml) were added to 1 ml of extrac-

tion solvent, either methanol, acetonitrile or dimethyl

sulphoxide (DMSO) and vigorously agitated using abench top vortex mixer for 1 min. After phase sepa-ration, 100ml of the upper phase was further diluted50-fold in methanol and immediately assayed.

2.6.3. KOH–ethanol/sulphuric acid extractionOil samples (100ml) were vortexed with 1 ml of 1 M

KOH in ethanol for 5 min. Next, 1 ml ofn-hexane wasadded and the mixture vortexed for a further minute.After phase separation, the upper hexane phase wastransferred to a fresh glass tube, 1 ml of concentratedsulphuric acid added and the mixture vortexed for1 min. After phase separation, the hexane phase wasremoved and evaporated under nitrogen. The residualmaterial was re-dissolved in 1 ml of 2-propanol, thenfurther diluted 50-, 100- and 200-fold in methanolprior to assay.

2.6.4. Solid phase extractionA C18 SPE column was pre-wetted with 2 ml of

2-propanol followed by 1 ml of 10% (v/v) transformeroil in 2-propanol. Retained materials were eluted with10 ml of 2-propanol then 10 ml ofn-hexane. Each frac-tion was collected separately, mixed well and 100mlfractions evaporated under nitrogen, re-dissolved anddiluted 500- or 1000-fold in methanol and assayed.

2.7. Commercial PCB assay kit

The optimised assay was tested in parallel with asemi-quantitative PCB kit, utilising antibody coatedmicrobeads in tube format (RaPID assay, SDI Eu-rope Ltd.). The kit uses an indirect competitiveassay format with anti-PCB antibody covalentlyimmobilised on paramagnetic particles and usedaccording to the manufacture’s instructions. Thefive PCB-containing transformer oil samples and a100mg ml−1 Aroclor 1260 standard were prepared bythe KOH–ethanol/sulphuric acid method and diluted10,000-fold for the oil samples and 40,000-fold for theAroclor 1260 standard in methanol. Extracts (200ml)were incubated with 250ml PCB–enzyme conjugateat room temperature for 15 min prior to bead recov-ery using a magnetic tray (SDI). The tubes werewashed twice with wash buffer, filled with 500ml ofsubstrate/chromogen and incubated for 20 min. Thereaction was halted using 500ml of 2 M sulphuricacid and the absorbance measured at 450 nm.


3. Results and discussion

3.1. ELISA optimisation

3.1.1. Optimisation of coating antigen and primaryantibody concentration

Optimisation was achieved using the checkerboardtitration method in which serial dilutions of bothcoating antigen and primary antibody were incubatedtogether. Since no sample was present, 100ml ofprimary antibody was used. Primary and secondaryincubation times of 1 h and OPD substrate (incuba-tion temperature 25◦C) were used in all tests. Thehighest OD readings were recorded at a coatingantigen and primary antibody concentration of 10and 1.25mg ml−1, respectively. These concentrationswere correspondingly employed in subsequent assaysunless otherwise stated.

Both the Polysorb and Immunoware plates werecoated with 10mg ml−1 TCPPA-BSA and incubatedwith 5, 2.5, 1.25, and 0.625mg ml−1 primary antibodyaccording to the assay procedure used for antibodyoptimisation. The Immunoware plate was found togive ca. 40% higher response than the Polysorb at thesame primary antibody dilution, indicating greaterprotein binding capacity, and was used in subsequentexperiments.

Fig. 1. Comparison of assay performance using OPD and TMB as enzymatic substrates. Antigen coating, 10mg ml−1; primary antibody,1.25mg ml−1; GAM-HRP, 1:4000; OPD or TurboTMB, 1 mg ml−1. Primary/secondary incubations, 1 h; substrate incubation, 15 min. Errorbars= S.D. (n = 4).

3.1.2. Comparison of OPD and TMB substratesA 90ml volume of 1mg ml−1 MAb in PBST and

10ml of serially diluted Aroclor in methanol (800,600, 400, 200, 100, 50 and 0 ng ml−1) were addedto TCPPA-coated Immunoware plates and the im-munoassay procedure followed. Bound label activitywas determined using either OPD or TMB. OPD ex-hibited ca. three-fold stronger absorbance signal thanTMB (Fig. 1) and hence was used in subsequent ex-periments, although, unlike TMB, it is carcinogenicand less convenient to prepare.

3.1.3. Optimisation of incubation timesPrimary antibody (100ml of 1.25mg ml−1) was in-

cubated in TCPPA-coated plates for 30, 60, 90, 120,150, and 180 min, with subsequent assay steps beingperformed as before (secondary incubation for 1 h).Equilibrium binding was achieved at a primary incu-bation time of 150 min. This relatively long incubationtime highlights the problem of solid-phase (hetero-geneous) immunoassays compared to solution-based(homogeneous) assays, the limiting factor being thetime taken for antibody to diffuse from the bulk so-lution and achieve equilibrium binding with the solidphase support. The commercially available homoge-neous RaPID assay and Envirosys assay (SDI) testkits have incubation times of ca. 30 min. By using


Table 1Logit-log data for the five Aroclor standards assayed using theoptimised ELISA

Aroclor a (OD) b (OD ml ng−1) IC50 (ng ml−1) d (OD)

1242 1.816 0.932 803 0.0251248 1.851 0.938 332 0.0991254 1.933 1.044 217 0.1091260 1.986 1.077 212 0.1491262 1.987 0.983 120 0.125

an elevated incubation temperature of 37◦C and amechanical shaker/incubator, the primary incubationbinding equilibrium was achieved in 30 min. Underthe same conditions, the secondary incubation timeachieved equilibrium in 50 min, although an incuba-tion time of 30 min (75% of equilibrium OD value)was used in subsequent studies for reasons of assayspeed. The heterogeneous PCB assays reported in theliterature have incubation times of hours [15,16].

3.2. ELISA characterisation

3.2.1. Dose response of selected AroclorsThe performance of the fully optimised PCB

ELISA assay was determined for each of the standardAroclors (n = 8). Aroclor concentrations of 50,000,10,000, 5000, 1000, 500, 250, 125, 62.5, 31.25 and15.625 ng ml−1 were tested, with a methanol blank.The results are shown in Table 1. Thec-values(IC50-values) indicate that the assay is most sensitivetowards Aroclor 1262, and showed reduced sensitivitywith decreasing degree of chlorination. The supplierstates that the primary antibody has a specificity of112, 100, 67 and 40% to Aroclor 1260, 1254, 1248and 1242, respectively. Thus, it would appear thatthe antibody reacts more favourably with the morehighly chlorinated (higher numbered) PCBs. However,

Table 2Optimised ELISA detection limits in ng ml−1 for Aroclor 1254 and 1260a

Series 1 2 3 4 5 6 7 8 Mean LOD± S.D.

Aroclor 1254b 19.6 16.5 22.7 16.7 33.2 33.0 26.4 31.4 24.9 ± 7.1Aroclor 1260c 18.4 38.0 5.3 49.0 45.8 35.1 63.0 51.1 38.2 ± 18.7

a Eight separate assays were performed over a 7-month time period. Each calibration graph was constructed from a 11-point plot(n = 4) and the resultant linear equations given asy and r2.

b y = 2.425− 0.793x; r2 = 0.998.c y = 1.956− 0.610x; r2 = 0.984.

Aroclors 1254 and 1260 were selected for fur-ther experimentation due to their relative abun-dance in insulating oils. The practical quantitativerange of the ELISA for these two compounds was30–1000 ng ml−1 for both Aroclor types, with theIC50-values positioned centrally on the calibrationgraph. The linear calibration profiles and regressiondata for these Aroclors is shown in Table 2.

The assay LODs for Aroclors 1254 and 1260 weredetermined for eight separate assays (n = 4) per-formed during a 7-month period. Mean LODs of 24.9± 7.1 and 38.2 ± 18.7 ng ml−1 were recorded forAroclors 1254 and 1260, respectively. No significantdecrease in assay sensitivity was observed over the7-month period, indicating a high degree of assayrepeatability.

In order to calculate inter-assay precision, the as-say was performed on three separate days. In eachassay, the samples were analysed in triplicate to al-low calculation of intra-assay precision. The datafor 600 ng ml−1 Aroclor 1254 are reported here.Intra-assay precision varied from 1.4 to 6.3% rela-tive S.D. (R.S.D.), whilst inter-assay precision variedfrom 2.1 to 6.3% R.S.D. with a mean value of 3.8%.The overall average measured concentration was637.9 ng ml−1, yielding a bias of+6.3%. Shah et al.[17] recommends that the total R.S.D. of an assayshould not exceed 15% (20% at the lower limit of theassay range). Using these definitions, the optimisedELISA can therefore be considered both accurate andprecise for the analysis of Aroclor 1254 standard.

3.2.2. Cross-reactivityThe following stock solutions: 2,4-D, 5 mg ml−1;

PCP, 10 mg ml−1; TCPAA, 2.5 mg ml−1; TCPPA,2.5 mg ml−1, were serially diluted in methanol to fi-nal concentrations of 4.9, 78.1, 2.4 and 2.4mg ml−1,respectively, and assayed using the optimised ELISA


procedure (n = 4). The resultant logit-logc-valueswere 572, 1294, 52.9 and 78.4 mg ml−1 for 2,4-D,PCP, TCPAA and TCPPA, respectively. Thus, all ofthese potentially cross-reacting compounds exhibitedless than 0.4% cross-reactivity relative to Aroclor1254. As expected, TCPAA and TCPPA had thehighest cross-reactivities due to their close structuralsimilarity to the TCPPA-BSA coating antigen, al-though the binding interaction was sufficiently lowto permit application of the assay in most situations.These findings suggest there is a significant struc-tural difference between free TCPPA and conjugatedTCPPA and that the PCB–antibody binding reactionis affected by factors other than PCB proximal struc-ture. The very low cross-reactivity exhibited by thehighly chlorinated PCP suggests that PCB–primaryantibody binding is influenced by both the proximaland distal structure of the antigen. 2,4-D, chosen forits structural similarity in the proximal region of thePCBs, had a cross-reactivity of<0.06%.

3.2.3. Organic solvent tolerance of ELISAPCBs, like many environmental contaminants, are

relatively hydrophobic and as such are best extractedfrom the sample matrix using organic solvents. Whilstfor assay purposes it is possible to evaporate the sol-vent and re-constitute in the aqueous phase, it is prefer-able for reasons of assay simplicity to perform theanalysis directly within the organic solvent extract.This factor introduces a paradox since antibodies, bytheir nature, will favour the aqueous phase. However,it is becoming widely appreciated that antibodies, likeenzymes, can retain biological activity in organic sol-vents. Since the purpose of this study was to developa simple immunoassay to measure PCBs in insulatingoils, the performance of the assay in the presence ofa number of organic solvents was assessed. Since theassay requires a wash step prior to the measurementprocess, it is only the primary incubation that is per-formed in the organic phase.

Ten microlitre volumes of acetonitrile, propanol,ethanol, methanol and PBST were pipetted (n =4) into individual antigen-coated microtitre wellscontaining 90ml of primary antibody and the opti-mised assay procedure performed. Taking the OD492response of the PBST control to be 100%, the corre-sponding responses for 10% (v/v) solutions of ace-tonitrile (least polar), propanol, ethanol and methanol

(most polar) were 86, 75, 49 and 75%, respectively.It is apparent that the presence of organic solvent de-creases antibody–PCB binding and that the extent ofantibody binding is not solely dependent upon solventpolarity.

The latter observation does not concur with otherstudies [18,19] that show a correlation between decrea-sing antibody–antigen binding affinity and increasingsolvent hydrophobicity. However, Giraudi and Bag-giani [20], using a testosterone–anti-testosterone testsystem, found that binding affinity was more influ-enced by solvent molecular mass rather than solventpolarity, suggesting that binding inhibition is related tothe ability of the solvent to displace water from aroundthe antigen. It has been demonstrated that some anti-bodies, either free or immobilised, are able to retaina residual binding activity in 90% (v/v) methanol andethanol and in methanolic solutions containing 50%(v/v) acetone, diethyl ether or benzene [21]. Since thebinding of hydrophobic antigens to antibody bindingsites may be via hydrophobic interactions, loweringthe polarity of the surrounding water miscible solventwill result in the reduced effectiveness of these bind-ing mechanisms. Interestingly, despite reductions insignal intensity, assay sensitivity can actually be im-proved in the presence of some polar organic solvents[22].

The effect of solvent on the physiochemical prop-erties of the antigen may also contribute to the overallbinding process. For example, it has been found thatdecreased antibody–antigen binding affinity is appar-ent in solvents that are most suitable for antigen dis-solution. The TCPPA-BSA coating antigen complexis soluble in pure aqueous solution by virtue of thehydrophilic BSA carrier protein. It is probable thatthe TCPPA molecule retains its non-polar nature inthe conjugated form, thus, this non-polarity partly ex-plains the relatively high antibody–PCB binding ob-served in the presence of propanol and acetonitrile.However, it should be noted that there may also be adifference in PCB and non-conjugated TCPPA solu-bility in polar and non-polar environments. Althoughthe most favourable binding characteristics were ap-parent in acetonitrile and methanol, the latter solventwas selected for further study due to its widespreadusage in environmental sample extractions [23]. Thebehaviour of antibodies in organic solvents has beenreviewed by Setford et al. [24,25].


4. Detection of PCBs in transformer oil

4.1. Evaluation of extraction methods

The performance of the assay for quantitatively de-termining PCBs in insulating oils was investigated. Inorder to minimise the matrix effect, a number of alter-native sample preparation strategies were examined.Since a primary aim of the work was to develop anautomatable assay, a simple sample preparation pro-cedure was required. Tests were performed on unusedPCB-free transformer oil either used as supplied orspiked with 100mg ml−1 of Aroclor 1254 or 1260.Calibration graphs were prepared using the same stan-dards in methanol. All tests were repeated four times.

4.1.1. Direct dilutionThis was the simplest and hence the most easily au-

tomatable sample preparation method. The blank andspiked (1254 and 100mg ml−1) transformer oils wereimmiscible in methanol and thus required pre-dilutionin 2-propanol co-solvent before methanol dilution toyield a homogeneous solution. An ELISA PCB con-centration of 10.13±0.16mg ml−1 was determined inthe spiked oil, equating to a recovery of ca. 10% ofthe expected value. No significant difference was ob-served between the spiked and blank oils.

4.1.2. Liquid–liquid extractionExtraction efficiencies of 19.6, 21.7 and 19.8% were

obtained for methanol, acetonitrile and DMSO, re-spectively, for the 500-fold diluted PCB spiked (1254and 100mg ml−1) transformer oil. Assay repeatabilityvaried from 1.4 to 3.0% R.S.D. for the three solventsused. Increased extraction time, as well as solvent typehad no significant effect on PCB recovery.

Table 3ELISA based measurement of unused transformer oil spiked with 100mg ml−1 Aroclor 1260 standard and extracted using a C18 SPEcolumna

Dilution Eluate Measured (ng ml−1) S.D. (ng ml−1) Expected (ng ml−1) Efficiency (%)

1:500 2-Propanol 119 27 200 59.5Hexane 56 5.6 200 28.0

1:1000 2-Propanol 66 6.6 100 66.0Hexane 38 1.9 100 38.0

a Calibration curve-fit correlation (n = 8): r2 = 0.988; y = 1.591− 0.481x.

4.1.3. KOH–ethanol/sulphuric acid extractionThis approach, previously used for animal fats

digestion [26], was examined as a means of reducinginterferent co-extraction by degrading the long chainaliphatic interferents present in the transformer oil ma-trix. The final diluted methanol extracts were assayedby the ELISA method. The spiked transformer oil hada PCB level of 100mg ml−1 Aroclor 1254 equatingto PCB concentrations of 200, 100 and 50 ng ml−1

for the 500-, 1000- and 2000-fold diluted samples,respectively, assuming a 100% recovery of analyte.The actual recorded values were 179±14, 97±8, and37 ± 2 ng ml−1, respectively, equating to extractionefficiencies of 89, 97, and 74%. The 1:2000 dilutedsample, with an expected Aroclor 1254 concentrationof 50 ng ml−1, was at the lowest limit of the assayrange and may account for the decreased extractionefficiency observed. The higher extraction efficienciesrecorded for the 1:500 and 1:1000 diluted samplesindicates that the KOH–ethanol/sulphuric acid extrac-tion method, with appropriate extract dilution, mayrepresent a satisfactory means of minimising matrixeffects.

4.1.4. Solid phase extractionThe ELISA results for unused oil spiked with

100mg ml−1 Aroclor 1260 and extracted using a C18SPE column are shown in Table 3. At 500-fold di-lution, the 2-propanol phase contained 60% of theexpected PCB concentration whilst the subsequentlyeluted hexane phase contained 28% of the expectedPCB concentration giving a combined Aroclor 1260recovery of 87.5%. At the 1000-fold dilution, the2-propanol and hexane phases contained 66 and 38%of the expected PCB concentration, respectively,equating to an overall recovery of 104%. The S.D.


values suggest that there was no significant differencein Aroclor 1260 recovery at each dilution indicatingthat SPE column treatment may be an effective wayof sample pre-treatment.

Since matrix effects are a prime cause of problemsin PCB analysis in complex samples [27], numer-ous sample preparation methods have been reported.Most methods employ a combination of liquid–liquidextraction followed by florisil, silica gel, aluminiumoxide or benzenesulphonic acid column adsorptionchromatography. Shu et al. [28] report a method inwhich over 98% of the components of an oil matrixwere removed by DMSO liquid–liquid extraction,with further removal of 80% of the remaining oilcomponents by liquid chromatography. Although re-coveries in excess of 90% were recorded, the methodrequires a time-consuming and instrumentally com-plex chromatographic step. In this study, similar PCBrecoveries were achieved using rapid, simple au-tomatable methods based on KOH/ethanol treatmentor SPE column methods. The high sample dilutionsused in these latter approaches should act to minimisematrix effects in the subsequent sensitive ELISAstep.

4.1.5. Optimised ELISA versus a commercial assaykit

A comparison of the ELISA results obtained forthe optimised quantitative ELISA system and com-mercial RaPID assay system for the five used trans-former oil samples (A–E) and the standard sample isshown in Table 4. The oil samples were prepared by

Table 4Optimised quantitative ELISA system vs. a commercial PCB test kit for the analysis of five PCB-contaminated transformer oils and a100mg ml−1 Aroclor 1260 standard after extraction using the KOH–ethanol/sulphuric acid method

Sample Measured value (ng ml−1) R.S.D (%) Expected valuea (ng ml−1) Efficiency (%)

ELISAb RaPIDc ELISAb RaPIDc ELISAb RaPIDc ELISAb RaPIDc

Aroclor 1260 169 2.2 2.5 10.7 200 2.5 84.5 88.0A 24 0.45 4.7 5.5 8 0.4 436.3 112.5B 62 0.82 1.6 12.8 70 3.5 88.6 23.4C 89 4.53 8.9 6.0 102 5.1 87.3 88.8D 26 0.50 2.8 11.9 20 1.0 130.0 50.0E 25 0.378 4.7 6.6 12 0.6 208.3 63.0

a Value expected for 100% extraction efficiency. Based on gas chromatography data.b ELISA calibration (n = 4): r2 = 0.988; y = 1.870− 0.593x.c RaPID assay calibration (n = 2): r2 = 0.945; y = 0.121− 0.097x.

KOH–ethanol/sulphuric acid extraction and the PCBlevels determined simultaneously by both assay meth-ods. The calibration equations for the two assay meth-ods are also given in Table 4. A stronger linear cor-relation was observed for the optimised assay (r2 =0.988) compared to the commercial test kit (r2 =0.945).

The standard Aroclor sample, oils C and B withPCB concentrations of 100, 51 and 35mg ml−1, re-spectively, yielded efficiencies of 84.5, 87.3 and88.6%, respectively, using the optimised ELISAmethod and thus were in good agreement with theexpected PCB values (Table 4). However, for oils A,D and E, which had significantly lower PCB values(4–10mg ml−1), the ELISA assay significantly over-estimated the PCB concentration by 130–436%. Itshould be noted that these latter oils had PCB levelsat the lowest limit of the optimised ELISA analyti-cal range. Whilst further refinement is necessary toimprove the assay performance at this lower limit, itshould be noted that impending regulation in the UKwill identify samples with PCB concentrations in ex-cess of 50mg ml−1 to be classified as contaminated.By overestimating the PCB levels in contaminatedsamples will minimise the number of genuinely con-taminated oils being graded as ‘PCB-free’. Ultimately,to eradicate this problem, the assay LOD must be im-proved. This could be achieved by employing moreeffective clean-up procedures to allow the analysisof less dilute oil samples, reducing the quantity ofprimary antibody used in the assay or by selectionof a different higher affinity antibody. Alternatively,


an increase in sample volume may improve assay sen-sitivity.

The RaPID assay data did not display theconcentration-response trends observed with the op-timised ELISA. The Aroclor 1260 standard and oilsA and C yielded efficiency values of 88.0–112.5%,whilst oils B, D and E, with efficiency values of23.4–63.0% under-estimated the PCB levels expectedin the oils. It is interesting to note that the ratioof the efficiency of the optimised ELISA over theRaPID assay varies from 2.60 to 3.87 for all of theoils tested excluding oil C. The cause of this generalover-estimation by the optimised ELISA test (andunderestimation by the commercial kit) has to beelucidated but may, in part, be due to the differentantibody clones used.

The RaPID assay system had a total assay timeof 35 min compared to 75 min for the optimisedELISA. However, the latter system operates within amicrotitre plate format and hence is more amenableto automation than the magnetic bead assay. Whilstthe assay time of the current system is acceptablefor most applications, it could be reduced by directlyconjugating the enzyme label to the primary anti-PCBantibody thereby obviating the secondary incubationstep. A preliminary evaluation indicates that this as-say format has a similar analytical performance to thecurrent ELISA with an assay time of<50 min. Thisassay format would be simpler-to-use and hence moreattractive, but the influence of the solvent and matrixon enzyme label activity requires further attention.

5. Conclusions

The fully optimised assay had a dynamic range of30–1000 ng ml−1 with 50% signal inhibition values of217 and 212 ng ml−1 for the Aroclors 1254 and 1260(abundant in insulating oils) in methanol. The assayproved more sensitive to the more highly chlorinatedPCB standards tested. The acceptable limit of PCBs inelectrical plant insulating oils has yet to be defined inthe UK, but it would appear that impending legislationwill select a value ca. 50mg ml−1, which is close to theLOD of the developed assay. An improvement in assayLOD is therefore desirable. This could be achieved us-ing an anti-PCB antibody of higher PCB-binding affin-ity or by improving the sample preparation method

to remove greater quantities of interferents, allowinglower sample dilutions to be employed.

The analysis of diluted insulating oils indicated thatdirect oil analysis was an unreliable method of samplepreparation due to the presence of matrix interfer-ents. Two sample preparation methods were foundto be suitable for interferent removal — solid sor-bents and washing with KOH–ethanol then sulphuricacid. The latter method, combined with the optimisedELISA process, proved reliable for the analysis oftransformer oils containing >35mg ml−1 PCB whenneat, but over-estimated PCB levels in oils containing<10mg ml−1 neat analyte. The assay was comparedto a commercially available semi-quantitative PCBmagnetic bead format ELISA. Sample clean-up wasby the KOH–ethanol/sulphuric acid method. Bothmethods yielded reliable data for the quantification ofan Aroclor 1260 standard in methanol, but for fourof the five oils tested, the ELISA method determinedthe PCB levels to be 2.6–3.9 times higher than thecommercial test kit. The major benefit of the opti-mised ELISA over the commercial kit was simplicityof automation, allowing high sample throughputs.Preliminary data shows that reduced assay times areachievable by conjugating the assay label directly tothe primary antibody.

Acknowledgements

The authors wish to thank the National Grid Com-pany, Leatherhead, UK, for the supply of insulatingoil samples and Dr. J.A. Bolbot for his many helpfuldiscussions.

References

[1] K. Ballschmiter, M.F.Z. Zell, Fresnius J. Anal. Chem. 302(1980) 20.

[2] T.O. Rouse, Mineral insulating oil in transformer oil, IEEEElect. Insulat. Mag. 14 (1998) 6–16.

[3] UK Action Plan for the Phasing Out and Destruction ofPolychlorinated Biphenyls and Dangerous PCB Substitutes,Dept. Environment (PCBs), March 1997.

[4] A.L. Alford-Stevens, Environ. Sci. Technol. 20 (1986) 1194.[5] M.I. Luster, P.W. Albro, G. Clark, K. Chea, S.K. Chaudhary,

L.D. Lawson, J.T. Corbett, J.D. McKinney, Toxicol. Appl.Pharmacol. 50 (1979) 147.

[6] W.H. Newsome, J.B. Shields, Int. J. Environ. Anal. Chem.10 (1981) 295.


[7] M. Sisak, M. Franek, K. Hruska, Anal. Chim. Acta 311(1995) 415.

[8] P.T. Charles, D.W. Conrad, M.S. Jacobs, J.C. Bart, A.W.Kusterbeck, Bioconjugate Chem. 6 (1995) 691.

[9] M.A. Roberts, R.A. Durst, Anal. Chem. 67 (1995) 482.[10] C.Q. Zhao, N.A. Anis, K.R. Rogers, R.H. Kline Jr., J. Wright,

A.T. Eldefrawi, M.E. Eldefrawi, J. Agric. Food Chem. 43(1995) 2308.

[11] J.C. Johnson, J.M. Van Emon, Anal. Chem. 68 (1996) 162.[12] T.S. Lawruk, C.E. Lachman, S.W. Jourdan, J.R. Fleeker, M.C.

Hayes, D.P. Herzog, F.M. Rubio, Environ. Sci. Technol. 30(1996) 695.

[13] S. Bender, O. Sadik, Environ. Sci. Technol. 32 (1998) 788.[14] D. Rodbard, J.A. Cooper, In Vitro Procedures with

Radioisotopes in Medicine, International Atomic EnergyAgency, Vienna, 1970, pp. 659–673.

[15] T. Vo-Dinh, A. Pal, T. Pal, Anal. Chem. 66 (1994) 1264.[16] Y.-W. Chiu, R.E. Carlson, K.L. Marcus, A. Karu, Anal. Chem.

67 (1995) 3829.[17] V.P. Shah, K.K. Midah, S. Dighe, I.J. McGilveray, J.P. Skelly,

A. Yacobi, T. Layloff, C.T. Viswanathan, C.E. Cook, R.D.McDowall, K.A. Pittman, S. Spector, Pharm. Res. 9 (1992)588.

[18] A.J. Russell, L.J. Trudel, P.L. Skipper, J.D. Groopman,S.R. Tannenbaum, A.M. Kilvanov, Biochem. Biophys. Res.Commun. 158 (1989) 80.

[19] W. Stocklein, A. Gebbert, R.D. Schmid, Anal. Lett. 23 (1990)1465.

[20] G. Giraudi, Baggiani, Biochem. Biophys. Acta 1157 (1993)211.

[21] S. Matsuura, Y. Hamano, H. Kita, Takagaki, J. Biochem. 114(1993) 273.

[22] W.F.M. Stocklein, F.W. Scheller, R. Abuknesha, Sens.Actuators B 24 (1995) 80.

[23] T.M. Spittler, Environ. Samp. Hazardous Wastes 267 (1984)37.

[24] S.J. Setford, S. Kroger, A.P.F. Turner, Analysis 27 (1999)600.

[25] S.J. Setford, Trends Anal. Chem., in press.[26] S. Tanabe, N. Kannan, A. Subramanian, S. Watanabe, R.

Tatsukawa, Environ. Pollut. 47 (1987) 147.[27] K. Li, T. Caron, J.R. Jocelyn-Pare, M. Lanoy, C. Chiu,

Proceedings of the Electric Power Research Institute, 1991,pp. 48.1–48.3.

[28] Y.Y. Shu, J.E. Dowdall, C. Chiu, R.C. Lao, Int. J. Environ.Anal. Chem. 60 (1995) 185.

Download - Determination of polychlorinated biphenyl compounds in electrical insulating oils by enzyme immunoassay

Top Related