Talanta 51 (2000) 395–401

Regenerable immunobiosensor for the chemiluminescentflow injection analysis of the herbicide 2,4-D

Christophe A. Marquette, Loıc J. Blum *Laboratoire de Genie Enzymatique, UPRES-A CNRS 5013-Uni6ersite Claude Bernard Lyon 1, Bat. 308-43,

Bd du 11 no6embre 1918, 69622 Villeurbanne Cedex, France

Received 20 July 1999; received in revised form 30 September 1999; accepted 5 October 1999

Abstract

A semi-automated chemiluminescent competitive immunosensor for the herbicide 2,4-dichlorophenoxyacetic acid(2,4-D) is presented. Anti-2,4-D polyclonal antibodies are directly labelled with horseradish peroxidase allowing ap-iodophenol enhanced chemiluminescent detection. Using antigen immobilised on UltraBind type pre-activatedmembranes, the 2,4-D immunosensor exhibits low non-specific/specific binding ratio (maximum ratio: 5%) of thelabelled antibodies. The quantification of free 2,4-D in water is performed by co-injecting the sample and the labelledantibodies in the flow system, incubating this solution with the antigen immobilised membrane and measuring theamount of specifically bound labelled antibodies. Such an analytical system enables the detection of 4 mg l−1 of freeantigen in 20 min, and the 2,4-D detection is possible in the range 4 mg l−1–160 mg l−1. The immunosensor can beregenerated by simply flowing a chaotropic solution (0.1 M HCl, 0.1 M NaCl, 0.1 M glycine) in the system. Thisregeneration ability enables the achievement of more than 30 measurement cycles of free 2,4-D with the same antigenimmobilised membrane with a good reproducibility (RSD=12.5%). © 2000 Elsevier Science B.V. All rights reserved.

Keywords: 2,4-Dichlorophenoxyacetic acid (2,4-D); Luminol chemiluminescence; Flow injection analysis; Immunosensor

www.elsevier.com/locate/talanta

1. Introduction

Enzyme linked immunosorbent assays (ELISA)on microwell plates are the reference methods inmedical diagnostics for the detection of interestingcompounds. Conversely, the detection of smalltoxic molecules in environmental and farm-pro-duce safety, rarely involves such immunological

measurement but preferred the classical and wellknown chromatographic (HPLC) analysis.

Nevertheless, these two methods are more andmore considered as time-consuming. Conse-quently, an increasing number of immunosensorsare described. Based on the antibody/antigenrecognition, rapid, simple and sensitive methodsare then developed for the measurement of a widetype of target compounds such as bacteria(Yersinia pestis), alphatoxin, ricin, brevetoxin [1]and okadaic acid [2], pesticide such as atrazine [3]and cocaine [4].

* Corresponding author. Tel.: +33-472-431397; fax: +33-472-442834.

E-mail address: loic.blum@univ-lyon1.fr (L.J. Blum)

0039-9140/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.

PII: S0039 -9140 (99 )00298 -2

C.A. Marquette, L.J. Blum / Talanta 51 (2000) 395–401396

In classical ELISA tests, the problem of theduration of the assays is overcome by theconcomitant measurement of multiple assays,and the poor repeatability is compensate for byperforming a calibration within each series ofassays. On the contrary, the aim of immuno-sensor development is the achievement of asystem able to perform a single point determina-tion without calibration between each measure-ment. The challenge is then to obtain animmunosensor which on the one handexhibits good reproducibility and repeatabilityand on the other hand could be regenerated inorder to minimise the consumption of expensivereactants.

Various transduction systems, such as electro-chemistry [1,4–6], potentiometry [7], cap-acitance [8], electrochemiluminescence [9] andchemiluminescence [2,9], have been usedsuccessfully in term of sensitivity and time con-sumed for immunosensor developments.However, the immunobiosensor stability, whichis for a great extent governed by the regen-eration process, is often a critical point thatmay impairs the reproducibility and the re-peatability.

Recently, the use of new pre-activatedpolyethersulfone membranes (UltraBind™) forimmunosensor development was described [2].These membranes allowed to minimise the non-specific binding of antibodies on their surface.They were then used for the development of achemiluminescent flow injection competitive im-munosensor for the detection of the planktonictoxin, okadaic acid. The performances of thisimmunosensor was promising and thus, we inves-tigated the possibility to develop with a similarapproach an analytical system for the detection ofanother small target analyte, the herbicide 2,4-dichlorophenoxyacetic (2,4-D) acid. Such a com-pound needs to be detected in drinking water at avery low concentration (0.1 mg l−1 according tothe European Community specification) [10]and, in order to avoid population intoxication,the analytical procedure must be as fast as possi-ble.

Several 2,4-D immunosensors have been de-scribed, involving different immobilisation sup-

ports such as glass capillary [10], nitro-cellulosemembranes [11], screen-printed electrodes [5],photoactivated collagen membranes [7], glassycarbon [9] or graphite electrodes [12] and differenttransduction systems such as flow amperometry,batch amperometry, pH-sensitive field effect tran-sistor and electrochemiluminescence, respectively.In every case, the quantification of the free 2,4-Dwas performed by the determination of the labelenzyme activity. Depending on whether the anti-bodies involved were monoclonal [5,10,11,13] orpolyclonal [7,12], the detection limits obtainedwere below or above the critical value of 0.1 mgl−1. Nevertheless, only two of these works gavethe possibility to regenerate the immobilisedsensing element [9,10,12] and only the glass capil-lary-based sensor [10] was integrated in a flowsystem.

In the work presented here, the flow injectionsystem developed, integrating a chemilumines-cence-based fiberoptic sensor, enables the sensitiveand rapid competitive immunodetection of 2,4-Din semi-automated conditions. Each step of theimmunoassay procedure, including the regenera-tion of the sensing element was realised in theflow system.

2. Experimental

2.1. Reagents

Peroxidase (HRP, grade I, EC 1.11.1.7., fromhorseradish, 250 IU mg−1) was supplied byBoehringer Mannheim. Luminol (3-aminophthal-hydrazide), 2,4-D, bovine serum albumin (BSA,98% IgG free) were purchased from Sigma. Poly-clonal sheep anti-2,4-D antibodies (5 mg ml−1)were obtained from Europa Bioproducts (UK).All other reagents were of analytical-reagentgrade. All buffers and aqueous solutions wereprepared with distilled demineralised water. Forchemiluminescence measurements, the reactionmedium was a 100 mM Tris buffer containing 30mM KCl, 140 mM NaCl and adjusted to pH 8.5with HCl 6 N. A luminol stock solution was madeas a 5.5 mM solution in 10 mM KOH and storedat 4°C.

C.A. Marquette, L.J. Blum / Talanta 51 (2000) 395–401 397

2.2. Instrumentation and sensor assembly

The chemiluminescence measurements wereperformed in a flow system and recorded on agraphic recorder (Servotrace, Sefram). The flowsystem consisted of a one-channel peristalticpump (model P-1, Pharmacia) connected to asix-way connection valve (model 5011, Rheo-dyne), two injection valves (model 5020, Rheo-dyne) on which a 40 ml and a 200 ml sample loopswere fitted, and a specially designed flow cell witha 250 ml inner volume (Fig. 1A) containing astirring bar (length 8 mm). A liquid core singleoptical fibre from LOT Oriel, France (core diame-ter 5 mm, overall diameter 7 mm) was connectedto the photomultiplier tube of a luminometer(Biocounter M 2500, Lumac). The light intensitywas expressed in arbitrary unit (au). In all mea-surements, the immobilised antigen membrane

was placed in close contact with the plexiglasswindow (Fig. 1B).

2.3. Immobilisation of 2,4-D on UltraBind™membrane

In order to covalently immobilise the 2,4-D onpre-activated membranes, the antigen was firstconjugated to BSA, after a carbodiimide activa-tion reaction. The 2,4-D was activated via itscarboxylic acid function by a pre-treatment in1,4-dioxane at the concentration of 1.36 mg ml−1

in the presence of 3.9 mg ml−1 N-hydroxysuccin-imide and 14.8 mg ml−1 N,N %-dicyclohexylcar-bodiimide. After an incubation time of 15 min,the dicyclourea precipitate was eliminated by cen-trifugation and 20 ml of the supernatant wasadded to 500 ml of a 10 mg ml−1 BSA solution in0.1 M carbonate buffer, pH 11. The obtainedsolution was then incubated under stirring for 3 hat room temperature for the coupling process tobe completed. The formed BSA-2,4-D conjugatewas then separated from the non-reacted specieson a desalting chromatography column (SephadexG-25 M). The conjugate was stored in PBS buffercontaining 0.1% sodium azide (w/v) at 4°C.

The BSA-2,4-D conjugate was covalently im-mobilised on a 11 mm diameter preactivated Ul-traBind™ disc by dipping the disc for 10 s in a 1.7mg ml−1 conjugate solution in PBS and drying it15 min at room temperature. The discs were thenwashed, first 10 min in PBS then, 20 min in PBScontaining 4% BSA (w/v) (PBSA), and finally 10min in PBS. The immobilised BSA-2,4-D mem-branes were stored dry at 4°C.

2.4. Labelling of the anti-2,4-D polyclonalantibodies with horseradish peroxidase

Horseradish peroxidase undergone a periodateactivation process of its carbohydrate moiety be-fore reacting with the anti-2,4-D polyclonal anti-bodies. One point four milligrams of peroxidasewas dissolved in 350 ml of distilled water andincubated for 20 min with stirring with 70 ml of21.5 mg ml−1 sodium periodate in water.

One hundred and fifty microlitres of the acti-vated peroxidase solution was then added to 200

Fig. 1. Scheme of the immunochemiluminescent set up. A:flow system; CL, chemiluminescent substrate solution; FC,flow cell; FO, optical fibre; M, measurement buffer; R, regen-eration solution; W, waste. B optical fibre/flow cell interface.

C.A. Marquette, L.J. Blum / Talanta 51 (2000) 395–401398

Table 1Sequence for a measurement cycle with the chemiluminescence-based 2,4-D immunosensor

Step Additional duration (min)Reaction conditions

0.1 M glycine, 0.1 M HCl, 0.1 M NaClRegeneration0.28 ml min−1, 500 rpm 8Buffera, 0.28 ml min−1, 500 rpmSample mixing 10Washing

15Stop flow, 500 rpmIncubationBuffera, 2.8 ml min−1, 1000 rpmWashing 20

Light measurement Buffera, 0.11 ml min−1, 1000 rpm

a 100 mM Tris-buffer, pH 8.5 containing 30 mM KCl and 100 mM NaCl.

ml of the polyclonal anti-2,4-D antibodies stocksolution (5 mg ml−1 in PBS). One hundred mi-crolitres of a 0.1 M carbonate buffer pH 11 wereadded and the solution allowed to react 90 minunder stirring at room temperature. After thattime, the labelling process was complete and theantibody/peroxidase solution was desalted bychromatography (Sephadex G-25 M).

The peroxidase-labelled antibodies were thenseparated from the unbound enzyme with afreezyme purification kit (Pierce). The antibody-enzyme bonds were then stabilised by the additionof 0.2 mg ml−1 sodium borohydride and stored inPBS, pH 7.4 at 4°C.

The peroxidase-labelled antibodies were charac-terised by spectrophotometric absorbance at 278nm (total protein: o278

Ab = [1.4 mg ml−1]−1 cm−1,o278

Pod=0.61 [mg ml−1]−1 cm−1) and 402 nm (per-oxidase specific wavelength: o402

Pod=2[mg ml−1]−1

cm−1). A ratio of about 1:1 in mole of peroxidaseper mole of antibody was found.

2.5. Flow injection immunoassay procedure

All free 2,4-D measurements were performedusing the same procedure (Table 1). Peroxidaselabelled anti-2,4-D antibodies were diluted in thesample mixing cell, filled with 250 ml of PBSA and10 ml of a free 2,4-D solution in PBS. This solu-tion was mixed 5 min at room temperature andinjected in the flow cell by the 200 ml injectionloop. The flowing stream, composed of a 100 mMTris buffer at pH 8.5 and containing 30 mM KCland 140 mM NaCl, was stopped 5 min and thestirring was then fixed at 500 rpm. After thisincubation time, the flow was switched on, at a

rate of 2.8 ml min−1 and the stirring fixed at 1000rpm during 5 min. A 0.11 ml min−1 flow rate wasthen applied, the substrate solution (final concen-tration: 0.22 mM luminol, 0.5 mM H2O2, 0.4 mMp-iodophenol) was injected with the 40 ml injec-tion loop and the chemiluminescent signal wasrecorded.

At that time, the regeneration solution (0.1 Mglycine, 0.1 M HCl and 0.1 M NaCl) was circu-lated at a rate of 0.28 ml min−1 during 8 min,with the stirring fixed to 500 rpm. Afterwards, thecell was washed 2 min by switching on the mea-surement buffer flowing stream at 0.28 ml min−1.The immunosensor was then ready for a newmeasurement cycle.

3. Results and discussion

3.1. Specific and non-specific binding of thelabelled antibodies

The non-specific binding of antibodies is themajor source of problems in the immunosensordevelopment. As a matter of fact, an efficientwashing is generally time consuming, increasingthe total duration of the measurement cycle. In arecent work, we showed that non-specific bindingcould be easily overcome, without extensive wash-ing and saturation steps, by injecting the antibod-ies in a complex matrix solution [2]. In the presentstudy, a 4% BSA (w/v) solution in PBS (PBSA4%) is used as a non-specific binding blocker. Fig.2 presents the signal intensities obtained after theinjection in the flow system of labelled antibodysolutions at different concentrations, from 0.2 to

C.A. Marquette, L.J. Blum / Talanta 51 (2000) 395–401 399

Fig. 2. Specific and non-specific binding of the anti-2,4-Dperoxidase labelled antibodies in the flow cell. Specific () andnon-specific (�) signal. Insert: specific and non-specific signalat low antibody concentrations.

concentrations, from 0.2 to 2.4 mg ml−1. This plotenables the selection of an antibody concentrationvalue compatible with immunosensor require-ments, i.e. an antibody concentration as low aspossible giving a measured signal as high as possi-ble. Thus, an anti-2,4-D labelled antibody concen-tration of 1.2 mg ml−1 was chosen because thesignal obtained at this value, equal to 780 au,corresponded to a signal to noise ratio of 280.This allowed to obtain a high sensitivity for themeasurement of the signal variations as requiredfor the competition reaction quantification.

3.2. Competiti6e measurement of 2,4-D

The competitive measurements of free 2,4-D inwater are performed according to the procedurepresented in Table 1. The length and the flow rateof the different steps were determined previously[2]. More particularly, it was shown that theduration of the sample pre-incubation step withlabelled antibodies and that of the incubation ofthis mixed solution with the immobilised antigenmembrane was long enough for acceptable im-munosensor performances to be obtained. Indeed,the longer the pre-incubation time, the lower thedetection limit. This lapse of time was thus animportant parameter to be optimised.

Samples containing free 2,4-D at different con-centrations, from 0.4 mg l−1 to 160 mg l−1, wereinjected in the sample mixing cell, pre-incubatedduring 5 min with labelled antibodies at a concen-tration of 1.2 mg ml−1 and then injected in theflow cell. For each concentration tested, fourreplicates were performed. The calibration curveobtained and its logit linearisation are presentedin Fig. 3. The logit representation is calculated asdescribed by Eq. (1), where B0 is the maximumsignal obtained in the absence of free 2,4-D, andB the signal obtained in the presence of 2,4-D.

logit= ln� B

[B0−B ]�

(1)

The results show that the presence of free 2,4-Din water could be detected at a concentration of 4mg l−1 and that a 0.4 mg l−1 concentration doesnot induced a significant change in the measuredchemiluminescent signal. The detection could be

48 mg ml−1 in PBSA 4%, using the immunosensorassembled either with an immobilised BSA-2,4-Dmembrane or with a membrane bearing onlyBSA. As it can be seen, the differences betweenthe corresponding specific and non-specific signalsare large, even in the presence of high antibodyconcentrations. The higher non-specific/specificsignal ratio obtained was approximately 5%.Moreover, such a low non-specific binding wasobtained without the need of a saturation step ofthe membrane which should increase the durationof the measurement cycle.

The inset in Fig. 2 shows the variation of thespecific and non-specific signals at low antibody

Fig. 3. Free 2,4-D calibration curve. Insert: logit linearisation.Each point is the result of four replicate assays. Error barsrepresent the standard deviation values.

C.A. Marquette, L.J. Blum / Talanta 51 (2000) 395–401400

Fig. 4. Operational stability of the chemiluminescent 2,4-Dimmunosensor. Maximum signal B0 (), assay signal B (�)([2,4-D]=40 mg l−1). The arrows show the storage over nightat 4°C. The solid line represents the mean value and the dottedlines represent the mean value9SD.

[9] showed that increasing the pre-incubation timefrom 5 to 30 min allowed to lower the detectionlimit from 200 to 0.2 mg l−1.

The pre-incubation time of the free 2,4-D withthe labelled antibodies in the sample mixing cellhave been then increased to 30 min. Unfortu-nately, such a long pre-incubation time at 30°Cinduces a lost of the antibodies ability to bindspecifically the immobilised antigen, and no inter-pretable results could be obtained.

3.3. Regeneration and stability of the immobilisedantigen membranes

The regeneration of the sensing element is acritical point when considering the use of immuno-biosensors, contrary to the classical ELISA onmicrowell plates with which the reactant immo-bilised on the solid phase is used only once. Theachievement of a regenerable antigen support ap-pears then as an absolute requirement for thedevelopment of a reliable immunosensor. It hasbeen shown previously that the anti-2,4-D anti-bodies/immobilised antigen complex dissociatedonly with high difficulty and required then a 7 minsonication treatment in HCl 0.1 M [9]. With thepresent membrane-based system, the regenerationof the immobilised antigen was obtained by simplyflowing through the cell a chaotropic solutioncomposed of 0.1 M HCl, 0.1 M Glycine and 0.1 MNaCl.

The stability of the 2,4-D immunosensor wasstudied by performing more than 35 successivemeasurements and corresponding regenerations,over a three-day period. For that purpose, themaximum specific binding (B0) of the anti-2,4-Dlabelled antibody was measured, i.e. in the absenceof free 2,4-D, and assays of free 2,4-D at theconcentration of 40 mg l−1 (B) was performedapproximately each ten measurement cycles.

As shown in Fig. 4, the 2,4-D immobilisedsupport appeared to be reusable at least 30 times.No lost of specific binding ability was observedbefore the 33rd cycle and a stable assay signal (B)was obtained throughout this period. Moreover,the two nights of dry storage of the immobilisedantigen membrane have had no effect on thesensor stability.

performed over at least four decades of concentra-tion, and was limited at high antigen concentra-tion by the 2,4-D solubility in water. The mean B0

value for four replicates was 871 au with a stan-dard deviation (SD) of 2 au giving a relativestandard deviation (RSD) equal to 0.2%. For a2,4-D concentration of 4 mg l−1, the mean B valueof four replicates was 811 au (SD=26 au, RSD=3.2%).

In classical immunoassays, the limit of detectionis defined as the amount of free antigen generatinga signal variation equal to at least three times thestandard deviation of the B0 value. According tothis definition and taking into account the experi-mental values of B0 and of the related SD, atheoretical detection limit of the order of 0.1 mgl−1 could be calculated. However, as mentionedabove, a 0.4 mg l−1 concentration did not induceda significant change in the measured chemilu-minescent signal and it appeared more reasonableto consider the 4 mg l−1 2,4-D concentration as thetrue detection limit of the present system.

The current accepted level of free 2,4-D in wateris 0.1 mg l−1 (European Community). Conse-quently, the detection limit obtained with thepresent system appeared insufficient. As men-tioned above, the pre-incubation time is the mainparameter to be optimised to obtain low detectionlimit. Previous works on 2,4-D immunodetection

C.A. Marquette, L.J. Blum / Talanta 51 (2000) 395–401 401

The standard deviation for the first 33 measure-ments of the maximum signal (B0) was equal to12.5%.

4. Conclusion

The present work demonstrates the efficiency ofthe proposed membrane-based system for regener-able immunobiosensor developments. The ap-proach developed in this study enables the easyachievement of a rapid and sensitive immunosen-sor for the detection of the herbicide 2,4-dichlorophenoxyacetic acid. Indeed, the detectionof 2,4-D was obtained within 20 min, which is alow assay duration when considering immuno-chemical assays.

The performances of the present immunosensorwith a detection limit of 4 mg l−1, could beconsidered as insufficient since the level actuallyaccepted by the European Community is 0.1 mgl−1. Nevertheless, keeping in mind that the anti-bodies used are of polyclonal type, an im-munosensor using monoclonal antibodies mightbe able to reach the 0.1 mg l−1 critical detectionlimit. Indeed, 2,4-D immunoprobes using mono-clonal antibodies, usually exhibited lower detec-tion limits (0.1 mg l−1 [5,10,11] and 0.01 mg l−1

[13]), than those involving polyclonal antibodies(1 mg l−1 [7], 40 mg l−1 [12]).

Finally, the operational stability of the antigenimmobilised membranes and the reproducibilityof the method were demonstrated by the ability ofthe sensor to perform, with the same sensinglayer, more than 30 measurement cycles with astandard deviation of 12.5%. Most of the studiesconcerning the imunodetection of 2,4-D did notproposed a regeneration of the sensing layer (im-mobilised antibody or antigen) [5,7,11,13] and

only few realizations provide this possibility[9,12]. The present work which enables the inte-gration of the immunosensor in a flow injectionsystem and the easy regeneration of the immo-bilised compound appears then to be an interest-ing development.

The use of pre-activated polyethersulfone mem-branes for the sensing layer elaboration in associ-ation with a chemiluminescent detection systemappears to be powerful for the development ofregenerable immunobiosensors. The total automa-tion of the system, in order to obtain a betterreproducibility, and the extent to other analytesand other test formats, such as sandwich tests, arenow under investigation.

References

[1] R.M. Carter, G.J. Lubrano, G.G. Guilbault, Life Chem.Rep. 11 (1994) 271.

[2] C.A. Marquette, P.R. Coulet, L.J. Blum, Anal. Chim.Acta 398 (1999) 173.

[3] E.F. Schipper, S. Rauchalles, R.P.H. Kooyman, B. Hock,J. Greve, Anal. Chem. 70 (1998) 1192.

[4] C.G. Bauer, A.V. Eremenko, A. Kuhn, K. Kurzinger, A.Makower, F.W. Scheller, Anal. Chem. 70 (1998) 4624.

[5] S. Kroger, S.J. Setford, A.P.F. Turner, Anal. Chem. 70(1998) 5047.

[6] J. Wang, P.V.A. Pamidi, K.R. Rogers, Anal. Chem. 70(1998) 1171.

[7] S.M. Khomutov, A.V. Zherdev, B.B. Dzantiev, A.N.Reshetilov, Anal. Lett. 27 (1994) 2983.

[8] C. Berggren, G. Johanson, Anal. Chem. 69 (1997) 3651.[9] C.A. Marquette, L.J. Blum, Sens. Actuators B 51 (1998)

100.[10] M. Wilmer, D. Trau, R. Renneberg, F. Spener, Anal.

Lett. 30 (1997) 515.[11] P. Skladal, T. Kalab, Anal. Chim. Acta 316 (1995) 73.[12] B.B. Dzantiev, A.V. Zherdev, Biosens. Bioelectron. 11

(1996) 179.[13] A. Dzgoev, M. Melcklenburg, P. Larsson, B. Danielsson,

Anal. Chem. 68 (1996) 3364.

.