14
Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection Julia Yakovleva a , Richard Davidsson b , Martin Bengtsson c , Thomas Laurell c , Jenny Emne ´us b, * a Department of Chemical Enzymology, Faculty of Chemistry, M.V. Lomonosov Moscow State University, Moscow 119899, Russia b Department of Analytical Chemistry, Lund University, P.O. Box 124, SE-221 00 Lund, Sweden c Department of Electrical Measurements, Lund Institute of Technology, Lund University, P.O. Box 118, SE-221 00 Lund, Sweden Received 22 August 2002; received in revised form 25 January 2003; accepted 9 March 2003 Abstract Affinity proteins were covalently immobilised on silicon microchips with overall dimensions of 13.1 /3.2 mm, comprising 42 porous flow channels of 235 mm depth and 25 mm width, and used to develop microfluidic immunosensors based on horseradish peroxidase (HRP), catalysing the chemiluminescent oxidation of luminol/p -iodophenol (PIP). Different hydrophilic polymers with long flexible chains (polyethylenimine (PEI), dextran (DEX), polyvinyl alcohol, aminodextran) and 3-aminopropyltriethoxysilane (APTS) were employed for modification of the silica surfaces followed by attachment of protein A or G. The resulting immunosensors were compared in an affinity capture assay format, where the competition between the labelled antigen and the analyte for antibody-binding sites took place in the bulk of the solution. The formed immunocomplexes were then trapped by the microchip affinity capture support and the amount of bound tracer was monitored by injection of luminol, PIP and H 2 O 2 . All immunosensors were capable of detecting atrazine at the sub-mgl 1 level. The most sensitive assays were obtained with PEI and DEX polymer modified supports and immobilised protein G, with limits of detection of 0.006 and 0.010 mgl 1 , and IC 50 values of 0.096 and 0.130 mgl 1 , respectively. The protein G based immunosensors were regenerated with 0.4 M glycine /HCl buffer pH 2.2, with no loss of activity observed for a storage and operating period of over 8 months. To estimate the applicability of the immunosensors to the analysis of real samples, PEI and DEX based protein G microchips were used to detect atrazine in surface water and fruit juice, spiked with known amounts of the atrazine, giving recovery values of 87 /102 and 88 /124% at atrazine fortification levels of 0.5 /3 and 80 /240 mgl 1 , respectively. # 2003 Elsevier Science B.V. All rights reserved. Keywords: Microfluidics; Immunosensor; Protein G and A; Hydrophilic polymers; Immobilisation; Chemiluminescence 1. Introduction Downsizing of analytical equipment is nowadays a rapidly developing area and has resulted in the concept known as the micro total analysis system (mTAS) (Reyes et al., 2002). Integrated miniaturised assay systems have been applied to the analysis in many different biological areas, including genomics (Sanders and Manz, 2000), proteomics (Wilson and Nock, 2002), multianalyte immunoassays (Silzel et al., 1998), cancer diagnostics (Askari et al., 2001) and others. In respect to the immunoassay field, various micro- fluidic devices have been used to adopt the conventional immunoassays (ELISA) at microscale, including glass chips with embedded microbeads (Sato et al., 2001) or polydimethylsiloxane chips with microfluidic channels (Eteshola and Leckband, 2001). Miniaturised immu- noassays make extensive use of electrokinetic phenom- ena, e.g. electroosmotic flow (McCreedy, 2000; Kricka, 2001) for fluid movement through the microchip chan- nels, and capillary electrophoresis (CE) as a separation technique (Dolnik et al., 2000; Hashimoto et al., 2000). A number of CE-based enzyme immunoassays has been reported (Koutny et al., 1996; Chiem and Harrison, * Corresponding author. Tel.: /46-46-222-4820; fax: /46-46-222- 4544. E-mail address: [email protected] (J. Emne ´us). Biosensors and Bioelectronics 19 (2003) 21 /34 www.elsevier.com/locate/bios 0956-5663/03/$ - see front matter # 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0956-5663(03)00126-X

Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

Embed Size (px)

Citation preview

Page 1: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

Microfluidic enzyme immunosensors with immobilised protein A andG using chemiluminescence detection

Julia Yakovleva a, Richard Davidsson b, Martin Bengtsson c, Thomas Laurell c,Jenny Emneus b,*

a Department of Chemical Enzymology, Faculty of Chemistry, M.V. Lomonosov Moscow State University, Moscow 119899, Russiab Department of Analytical Chemistry, Lund University, P.O. Box 124, SE-221 00 Lund, Sweden

c Department of Electrical Measurements, Lund Institute of Technology, Lund University, P.O. Box 118, SE-221 00 Lund, Sweden

Received 22 August 2002; received in revised form 25 January 2003; accepted 9 March 2003

Abstract

Affinity proteins were covalently immobilised on silicon microchips with overall dimensions of 13.1�/3.2 mm, comprising 42

porous flow channels of 235 mm depth and 25 mm width, and used to develop microfluidic immunosensors based on horseradish

peroxidase (HRP), catalysing the chemiluminescent oxidation of luminol/p -iodophenol (PIP). Different hydrophilic polymers with

long flexible chains (polyethylenimine (PEI), dextran (DEX), polyvinyl alcohol, aminodextran) and 3-aminopropyltriethoxysilane

(APTS) were employed for modification of the silica surfaces followed by attachment of protein A or G. The resulting

immunosensors were compared in an affinity capture assay format, where the competition between the labelled antigen and the

analyte for antibody-binding sites took place in the bulk of the solution. The formed immunocomplexes were then trapped by the

microchip affinity capture support and the amount of bound tracer was monitored by injection of luminol, PIP and H2O2. All

immunosensors were capable of detecting atrazine at the sub-mg l�1 level. The most sensitive assays were obtained with PEI and

DEX polymer modified supports and immobilised protein G, with limits of detection of 0.006 and 0.010 mg l�1, and IC50 values of

0.096 and 0.130 mg l�1, respectively. The protein G based immunosensors were regenerated with 0.4 M glycine�/HCl buffer pH 2.2,

with no loss of activity observed for a storage and operating period of over 8 months. To estimate the applicability of the

immunosensors to the analysis of real samples, PEI and DEX based protein G microchips were used to detect atrazine in surface

water and fruit juice, spiked with known amounts of the atrazine, giving recovery values of 87�/102 and 88�/124% at atrazine

fortification levels of 0.5�/3 and 80�/240 mg l�1, respectively.

# 2003 Elsevier Science B.V. All rights reserved.

Keywords: Microfluidics; Immunosensor; Protein G and A; Hydrophilic polymers; Immobilisation; Chemiluminescence

1. Introduction

Downsizing of analytical equipment is nowadays a

rapidly developing area and has resulted in the concept

known as the micro total analysis system (mTAS) (Reyes

et al., 2002). Integrated miniaturised assay systems have

been applied to the analysis in many different biological

areas, including genomics (Sanders and Manz, 2000),

proteomics (Wilson and Nock, 2002), multianalyte

immunoassays (Silzel et al., 1998), cancer diagnostics

(Askari et al., 2001) and others.

In respect to the immunoassay field, various micro-

fluidic devices have been used to adopt the conventional

immunoassays (ELISA) at microscale, including glass

chips with embedded microbeads (Sato et al., 2001) or

polydimethylsiloxane chips with microfluidic channels

(Eteshola and Leckband, 2001). Miniaturised immu-

noassays make extensive use of electrokinetic phenom-

ena, e.g. electroosmotic flow (McCreedy, 2000; Kricka,

2001) for fluid movement through the microchip chan-

nels, and capillary electrophoresis (CE) as a separation

technique (Dolnik et al., 2000; Hashimoto et al., 2000).

A number of CE-based enzyme immunoassays has been

reported (Koutny et al., 1996; Chiem and Harrison,

* Corresponding author. Tel.: �/46-46-222-4820; fax: �/46-46-222-

4544.

E-mail address: [email protected] (J. Emneus).

Biosensors and Bioelectronics 19 (2003) 21�/34

www.elsevier.com/locate/bios

0956-5663/03/$ - see front matter # 2003 Elsevier Science B.V. All rights reserved.

doi:10.1016/S0956-5663(03)00126-X

Page 2: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

1998; Wang et al., 2001), where glass or silica micro-

fluidic devices were used to conduct the assay procedure.

Another attractive biosensing material is porous silicon

with high surface area, which has been applied for thedevelopment of enzyme microreactors (Laurell et al.,

1996; Thust et al., 1999) and microfluidic immunosen-

sors with immobilised antibodies (Yakovleva et al.,

2002).

Immunosensors explore the principles of heteroge-

neous immunoassay and thus require one of the specific

immunoreaction components to be immobilised on the

solid phase. Different assay formats can be established,including the immobilised antibodies (Yakovleva et al.,

2002) or antigen (Brecht et al., 1998) format. An

alternative immunoassay format to overcome the regen-

eration problems typically encountered in the immobi-

lised antibody format is based on affinity proteins such

as protein G and A. The feature of protein A from

Staphylococcus aureus and protein G from group G

Streptococci to selectively bind the Fc region of a widerange of immunoglobulins (IgG) with no interference to

the antigen binding sites, make them very attractive as

affinity-capture supports for immunobiosensing appli-

cations (Valat et al., 2000).

Though a large number of assays employing protein

A or G have been reported within recent years, most of

them in fact use the affinity proteins as a specific

support for oriented antibody immobilisation (Ander-son et al., 1997; Quinn et al., 1999; Garcinuno et al.,

2000; Pulido-Tofino et al., 2000; Sheikh and Mulchan-

dani, 2001). The affinity capture format, however,

implies that competition between the tracer and analyte

for a corresponding antibody takes place in the bulk of a

solution, and the resulting immunocomplexes are cap-

tured by the affinity support downstream (Gonzalez-

Martinez et al., 1998; Penalva et al., 1999; Bjarnasson etal., 2000; Burestedt et al., 2000). Detection of bound

label in the flow systems reported in literature was

performed fluorometrically (Penalva et al., 1999) or by

biosensor (Nistor et al., 2002). Alternatively, the systems

based on monitoring of unbound label used fluoro-

metric (Gonzalez-Martinez et al., 1998; Bjarnasson et

al., 2000), biosensor (Burestedt et al., 2000) or direct

electrochemical (Kronkvist et al., 1997; Wang et al.,2001) detection. In most cases the affinity proteins were

immobilised on the beads and packed into the column.

In this work, we report the development of micro-

fluidic immunosensors based on affinity protein A and

G immobilised on silicon microchip surfaces by different

chemistries, and chemiluminescence (CL) detection.

Among various techniques applied for label detection

in microfluidic immunoassays, including fluorescence(Chiem and Harrison, 1998; Bernard et al., 2001; Hayes

et al., 2001), electrochemical (Lacher et al., 2001; Wang

et al., 2001) and thermal lens microscope (Sato et al.,

2000), one of the most promising techniques is CL.

Application of CL detection technique offers high

sensitivity and is especially suitable at microscale since

no external light source is needed (Hashimoto et al.,

2000). A similar system was recently published based onimmobilised antibodies (Yakovleva et al., 2002). The

aim of the present study was to optimise the affinity

protein immobilisation on the silicon microchip sur-

faces, to evaluate the assay performance in terms of

stability and sensitivity, and to investigate the imple-

mentation of the immunosensors to the analysis of real

samples.

2. Materials and methods

2.1. Materials

Silicon microchips were fabricated by anisotropic wet

etching technique, as described previously (Laurell et al.,

1996; Drott et al., 1997). The microchips had an overall

dimension of 13.1�/3.2 mm, comprising 42 porous flowchannels, each of those were 235 mm deep and 25 mm

wide. The porous surface structure of the channels was

achieved by anodising in a 1:1 mixture of 40% hydro-

fluoric acid and 96% ethanol with a current density of 50

mA cm�2 for 5 min.

Affinity purified polyclonal anti-atrazine IgG fraction

from sheep serum was generously provided by Dr

Ramadan Abuknesha (King’s College University ofLondon, UK). Protein G (recombinant, E. coli ) was

purchased from CalBiochem (La Jolla, CA, USA).

Protein A, polyethylenimine (PEI, MW 750 000, 50%

w/v aqueous) dextran (DEX, MW 64 000�/76 000, clin-

ical grade), horseradish peroxidase (HRP) type VI-A,

glutaraldehyde (GA) 25% v/v aqueous solution grade I,

N ,N ?-dicyclohexylcarbodiimide (DCC), 1-(3-dimethyla-

minopropyl)-3-ethylcarbodiimide (EDC), N -hydroxy-succinimide (NHS), dimethylformamide (DMF),

succinic acid anhydride (SAA), and 3-aminopropyl-

triethoxysilane (APTS) were purchased from Sigma

(St. Louis, MO, USA). Polyvinyl alcohol (PVA, MW

89 000�/98 000, 99% hydrolysed), ammonia, hydrochlo-

ric acid, hydrogen peroxide and toluene were from

Merck (Darmstadt, Germany). Luminol 97% and p-

iodophenol (PIP) 99% were obtained from Aldrich(Milwaukee, MI, USA). Lysine-monohydrate was sup-

plied from Carl Roth Co. (Karlsruhe, Germany), and

dimethylsulfoxide (DMSO) was from Fisher Co. (New

Jersey, USA). The atrazine hapten derivative, used for

tracer synthesis, 6-{{4-chloro-6-[(1-methylethyl) amino]-

1,3,5-triazin-2-yl}amino}hexanoic acid (iPr/Cl/

(CH2)5COOH) was synthesised according to Goodrow

et al. (1990). Atrazine was obtained from Riedel-deHaen (Seelze, Germany). All other chemicals were of

analytical grade. Microcon YM-10 micro-concentrators

were obtained from Amicon, Inc. (Beverley, MA, USA).

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/3422

Page 3: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

Milli-Q water was obtained from a Milli-Q purification

water system (Millipore, Bedford, USA).

2.2. Buffers and solutions

The carrier buffer was 0.05 M Tris�/HCl, pH 7.4 and

the substrate buffer was 0.05 M Tris�/HCl, pH 9.0.

When not in use, the microchips were kept in 0.1 M

Tris�/HCl buffer, pH 7.0. Borate buffer (BB) was 0.01 M

sodium tetraborate, pH adjusted to 7.0 with 0.5 M boricacid. The regeneration solution was 0.4 M glycine�/HCl,

pH 2.2. The substrate solution was a mixture of 50 ml of

luminol (50 mM in DMSO), 340 ml of PIP (150 mM in

DMSO), 8.5 ml of 30% w/w H2O2 and 50 ml of substrate

buffer. Working solution of the atrazine-HRP enzyme

tracer was prepared from a 1 mg l�1 stock solution by

diluting 10 000 times with the carrier buffer. Calibration

solutions of atrazine of 0.001, 0.01, 0.1, 1, and 10 mg l�1

were prepared from a 5 mg l�1 water stock solution by

diluting with carrier buffer.

2.3. Immobilisation of protein G and protein A on silicon

microchips

Protein G was immobilised on silicon microchip

surfaces modified with four different hydrophilic poly-

mers: (1) PEI; (2) DEX; (3) PVA, and (4) aminodextran

(AMD), while protein A was immobilised only on

surfaces modified with PEI and DEX and APTS, see

Fig. 1. Silanisation with APTS followed by GA activa-

tion (Weetall, 1976) was used as a reference technique.

Prior to any immobilisation, the silicon microchipswere cleaned in a mixture of 25% NH3, 30% H2O2 and

H2O (1:1:5, by volume) at 100 8C for 5 min, followed by

another treatment in a mixture of 37% HCl, 30% H2O2

and H2O (1:1:5, by volume) at 100 8C for 5 min. The

microchips were then carefully rinsed with water,

ethanol and acetone, and dried under a stream of air.

2.3.1. Immobilisation via APTS

The cleaned microchips were silanised in 10% APTS

in sodium-dried toluene (Weetall, 1976). To remove air

bubbles from the microchip pores, vacuum was appliedfor 1�/3 min. The reaction mixture was refluxed for 1 h

at room temperature (RT) in a sealed vessel, protected

from moisture with a drying tube filled with silica gel.

After removal of the silanisation solution, the micro-

chips were rinsed several times with toluene, acetone and

finally rinsed with BB.

The amino-groups on the APTS silanised surface was

activated in 2.5% v/v GA in BB for 60 min at RT, andfinally rinsed with Milli-Q water in order to remove

traces of GA to avoid cross-linking after addition of the

protein solution. The proteins were attached to the

surface by immersing the microchip in 0.5 mg ml�1

protein A solution in BB and reacted overnight at �/

4 8C under stirring. After 12 h the remaining aldehyde

groups were blocked in 10 mg ml�1L-lysine in BB. The

microchips were then carefully rinsed and stored in 0.1M Tris�/HCl buffer, pH 7.0 at �/4 8C until use.

2.3.2. Immobilisation via PEI

The cleaned microchips were immersed in 0.5% v/v

solution of PEI in BB under stirring at RT overnight,

and then rinsed with BB. The amino-groups of the

adsorbed PEI were activated in 2.5% v/v GA in BB for 2

h at RT under stirring. After careful washing with Milli-

Q water and BB, the microchips were placed in 0.5 mg

ml�1 protein G or protein A in BB, and the reaction

was allowed to proceed overnight at �/4 8C. Thefollowing blocking reactions were performed as de-

scribed in Section 2.3.1.

2.3.3. Immobilisation via DEX

Polyaldehyde dextran was obtained by partial oxida-

tion of DEX with sodium periodate according to a

previously described protocol (Penzol et al., 1998).

Briefly, 60 mg DEX was dissolved in 3 ml of Milli-Q

water and then 87 mg of solid sodium periodate was

added during stirring. The reaction mixture was kept atRT for 2 h and protected from light. Finally the oxidised

DEX was dialysed extensively against 3 l of Milli-Q

water at �/4 8C.

The cleaned microchips were silanised, as described in

Section 2.3.1. The dried microchips were immersed in a

fresh solution of polyaldehyde DEX and then allowed to

react overnight at �/4 8C. After attachment of DEX the

microchips were carefully rinsed with Milli-Q water andput in a 0.5 mg l�1 solution of protein G or protein A

dissolved in BB. The protein-coupling step via aldehyde

groups on DEX was then performed as described in

Section 2.3.1.Fig. 1. Structures of the hydrophilic polymers used for silicon

microchips surface modification.

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/34 23

Page 4: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

2.3.4. Immobilisation via PVA

The cleaned microchips were silanised, as described in

Section 2.3.1, and the amino groups of APTS-activated

microchips were then reacted in 2.5% v/v GA in BB for 1h at RT to introduce aldehyde groups on the surface,

followed by rinsing with BB.

A 10 mg ml�1 aqueous solution of PVA was prepared

according to the following protocol; 0.5 g PVA was

suspended in 50 ml of Milli-Q water and allowed to wet

for 5 h. The suspension was heated in a water bath for

approximately 10 min until the solution became trans-

parent and then the pH was adjusted to 1.0 withconcentrated HCl. The polymer was immobilised by

immersing the microchips in the PVA solution, while

stirring for 30 min at RT. After the removal of the

immobilisation solution, the microchips were treated

with 5% GA (pH adjusted to 1.0 with concentrated HCl)

for another 30 min to cross-link the polymer matrix

layer and introduce aldehyde-groups on the surface. The

PVA�/GA-activated microchips were then thoroughlyrinsed with Milli-Q water and put in a 0.5 mg ml�1

protein G solution in BB. The protein attachment was

then carried out as described in Section 2.3.1.

2.3.5. Immobilisation via AMD

AMD was prepared by oxidation of DEX followed by

reductive amination, using a modified protocol reported

earlier (Piehler et al., 1996). The oxidation was carried

out using 20 ml of 0.36 mM DEX aqueous solution and0.15 M sodium periodate. The reaction mixture was

stirred for 2 h at RT, after that the oxidised DEX was

purified by dialysis against 3 l of Milli-Q water for 2

days at �/4 8C.

A 100-fold molar excess (0.7 mmol) of ammonium

chloride was added to the dialysed DEX solution, and

pH was adjusted to 6, followed by heating to 100 8C. A

50-fold molar excess (0.35 mol) of sodium cyanobor-ohydride was added to the reaction mixture in several

portions at 100 8C during 3 days. The AMD product

was purified by dialysis against Milli-Q water for 2 days

at �/4 8C, and then filtered through a 0.45 mm nitro-

cellulose filter. A 10-fold excess of methanol (by volume)

was added to the filtrate (20 ml) and the pH was

adjusted to 4.0 to precipitate AMD. Finally the product

was filtered on a 0.45 mm nitrocellulose filter and dried.To immobilise AMD, the pre-cleaned microchips

were first silanised with APTS, according the protocol

described in Section 2.3.1. The introduced amine groups

were then derivatised to carboxylic groups by succinila-

tion according to the following protocol (Piehler et al.,

1996). The APTS-functionalised microchips were im-

mersed in 10 ml of BB, and 1 g of solid SAA was added.

The pH of the solution was maintained at 6.0 bycontinuous addition of 1 M NaOH, while the reaction

mixture was stirred and SAA dissolved, i.e. after 1 h.

The reaction was allowed to proceed for another hour

and after which the microchips were carefully rinsed

with BB and dried.

Finally, the AMD polymer was immobilised by

immersing the SAA-treated microchips in 15 ml 10 mgml�1 AMD aqueous solution, with pH adjusted to 3

with 2 M HCl, followed by addition of 50 mg of EDC.

The reaction mixture was stirred overnight at RT. The

AMD coated microchips were then carefully rinsed with

Milli-Q water and activated with 5% v/v GA in BB for 2

h at RT under stirring (the GA concentration was higher

than that applied for the PEI activation in order to

cross-link the AMD layer and provide free aldehydegroups capable of conjugating to the proteins). After

washing with BB to remove the excess of GA, the

microchips were immersed in 2 ml of 0.7 mg ml�1 of

protein G solution. The protein immobilisation was

performed as described in Section 2.3.1.

2.4. Synthesis of enzyme tracer

The enzyme tracer was synthesised by attachment ofthe iPr/Cl/(CH2)5COOH atrazine derivative to HRP by

the NHS ester method (Giersch, 1993). The hapten (1

mg), NHS (1.7 mg) and DCC (6.2 mg) were mixed with

130 ml of DMF. The mixture was agitated overnight at

RT. Then the activated ester solution was added drop-

wise to a solution containing 1 mg HPR in 0.5 ml of 0.13

M NaHCO3. The coupling reaction was performed for 3

h, after which the tracer was purified by dialysis against3 l of PBS (five changes of buffer) using a Slide-A-Lyzer

cassette (Pierce, Rockford, IL, USA) with a molecular

cut-off of 10 kDa.

2.5. System set-up and assay procedures

The scheme of the microfluidic system is shown in

Fig. 2a. A CMA Microdialysis 100 syringe pump(CMA/Microdialysis, Solna, Sweden) was used to de-

liver the 0.05 M Tris�/HCl pH 7.4 carrier buffer, at

either 40 or 50 ml min�1. For regeneration of the

immunosensor surface the flow was switched to 0.4 M

glycine�/HCl buffer via a Gilson Minipuls 2 peristaltic

pump (Villiers-le-Bell, France). A Rheodyne six-port

valve (Berkeley, CA, USA) equipped with a 4 ml loop

was used for injecting the samples. The microchipimmunosensor was incorporated in the flow system via

a specially designed flow cell unit (Fig. 2b) made of plexi

glass with the inlet and outlet tubing glued into holes

drilled in the top cover. A thin transparent silicon

rubber membrane was placed between the microchip

and the top cover to prevent leakage. The CL signal was

monitored via a photomultiplier tube (PMT, model no.

HC135-01 UV to visible, Hamamatsu Photonics K. K.,Japan), aligned right above the microchip flow cell with

the surface of immobilised affinity proteins exposed.

The microchip flow cell and the PMT was placed in a

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/3424

Page 5: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

holder to assure correct positioning of the two units and

the whole detection was performed in a ‘black box’ to

shut-out light from the surroundings, as seen in Fig. 2a.

The total volume of the flow system was 20 ml with

connecting PEEK tubing (i.d. 0.25 mm, Alltech, Deer-

field, IL, USA). Data acquisition was performed with a

homemade software.

The assay procedure is shown in Fig. 3 and is based

on the principle of affinity capture competitive immu-

noassay, according to the following steps: enzyme

tracer, atrazine standard or a sample, and anti-atrazine

antibody were mixed off-line and injected directly, or

after off-line pre-incubation for 20�/30 min, into the

system, at 40 ml min�1. The microchip was rinsed with

the carrier buffer for 2 min at 40 ml min�1 to remove

unbound immuno-complexes. Next, the substrate mix-

ture (luminol/PIP/H2O2) was injected at 50 ml min�1

and the enzyme-catalysed chemiluminescent reaction

taking place on the microchip surface was monitored

via the PMT. To complete the assay cycle, removing the

immunocomplex from the affinity proteins, the carrier

buffer was switched to 0.4 M glycine�/HCl buffer pH 2.2

via the Gilson Minipuls 2 peristaltic pump at 50 ml

min�1 for 2�/3 min. The total operation time for oneassay cycle, including rinsing, three repetitive injections

of substrate and regeneration step, was approximately

10 min.

2.6. Analysis of fortified surface water and orange juice

samples

Spiked surface water samples, from Hoje a, a river

that passes through a very active agricultural region in

Lund, Sweden, were applied to the DEX modified

Fig. 2. (a) Scheme of the microfluidic immunosensor manifold: a syringe pump and a peristaltic pump were used for carrier buffer and regeneration

solution at flow rates of 40 and 50 ml min�1, respectively. The sample, containing the enzyme tracer, atrazine standard and antibody, pre-mixed off-

line, was injected through a six-port injection valve, and the chemiluminescent signal was detected by a PMT placed above the flow cell, containing

the microfluidic immunosensor. (b) Detailed view of the plexi glass microchip flow cell and a magnified image of the microchip channel network.

Fig. 3. Scheme of the affinity capture competitive immunosensor format performed in the system shown in Fig. 2.

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/34 25

Page 6: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

protein G immunosensor. The surface water samples

were filtered through a 0.45 mm nitrocellulose filter

(Millipore) and analysed without any pH adjustments.

Undiluted sample and samples diluted 1:2 and 1:4 withthe carrier buffer were used to construct a calibration

curve for atrazine and to evaluate any potential matrix

effects. To estimate the assay recoveries, the surface

water was spiked with known amount of atrazine to give

the analyte concentrations of 0.5, 1 and 3 mg l�1 and

analysed by DEX modified protein G immunobiosen-

sor. Three assays were performed in triplicate and the

respective CL signals were used to estimate atrazineconcentration by interpolation in the calibration curve,

constructed the same day as the sample measurement.

Recovery values and standard deviations (S.D.) were

calculated for each sample and expressed as a percen-

tage ratio.

In a similar way, orange juice samples from commer-

cial sources were tested for matrix effects. The juice

samples were centrifuged at 10 000 rpm for 5 min, thesupernatant was filtered through a glass microfibre filter

GF/A, (Cat No 1820 047, Whatman, Maidstone,

England). The carrier buffer was used to dilute the

samples to give a juice to carrier buffer ratio of 1:50,

1:100, 1:200, 1:400 and 1:500. Calibration curves for

atrazine were obtained using the analyte standards in

carrier buffer and in diluted orange juice. To determine

the recoveries of atrazine, the juice samples were spikedto give concentrations of 80, 160 and 240 mg l�1

atrazine, and assayed by the PEI modified protein G

immunosensor. Three assays were performed in tripli-

cate and used to estimate the atrazine concentration in

the spiked sample and to calculate the recovery values

and S.D. for each sample. The recovery (R ) values were

calculated according to the formula:

R�CIA � CIA;0

Cspiked

where CIA is the analyte concentration in the spiked

sample evaluated by interpolation in the standard curve,

CIA,0 is the concentration of analyte in the sample

without spiking and Cspiked is the calculated concentra-

tion of the analyte spiked into each sample.

3. Results and discussion

The development of highly sensitive, precise, repro-

ducible, rapid, and cost-efficient assay techniques, im-

plying a fast analysis of a large number of samples, is

general goal that would fulfil the growing demands on

environmental monitoring. Miniaturisation of assays isnow an increasingly useful approach, which allows not

only to detect minute amounts of analyte, but also to

build up array systems for multiplex analysis, and, in

addition, offers experimental simplicity of handling

(Auroux et al., 2002).

Recently we reported microfluidic immunosensors

based on the same approach as presented here, butwith immobilised antibodies, displaying sensitivities in

the low ng l�1 range, thus exceeding those of conven-

tional batch and flow immunoassays previously re-

ported for this analyte (Yakovleva et al., 2002). Our

results indicated that the immunosensor performance

was largely a function of the immobilisation chemistry

used on the silica surface. A comparison of the attach-

ment of antibodies to APTS functionalised silica sup-port and to a hydrophilic layer of PEI, revealed an

appreciable improvement of the assay sensitivity using

the latter method. Significant discrepancies were also

found between the stability, sensitivity, and reproduci-

bility of the antibody coatings formed by the use of

different molecular weight of PEI covalently attached or

physically adsorbed on the silica surface, thus implying

the importance of optimising the immobilisation chem-istry of the sensing surface for the development of

sensitive, stable, and reusable immunosensors.

The immobilised antibody format is, however, af-

fected by the reduction of antigen binding capacity and

hence decreased sensitivity with time due to exposure of

the antibody to harsh regeneration conditions as re-

quired for the desorption of the antigen after each assay

cycle. The present work extends the application of thesesilicon microchips to the affinity capture assay format as

shown in Fig. 3, using protein A or protein G

immobilised in hydrophilic matrices of various poly-

mers, and the same anti-atrazine antibodies as in our

previous work (Yakovleva et al., 2002). This format

extends the flexibility of the assay, because a whole

range of antibodies with different specificity can be

bound and dissociated from protein A and G coatedsurfaces. Thus, the employment of the affinity capture

proteins allows the development of generic assays,

potentially applicable to any analyte.

In this work, different immobilisation chemistries for

protein G and A on silicon microchips were investigated

and the resulting immunosensors evaluated in terms of

stability, reusability and sensitivity, and finally applied

for the analysis of surface water and fruit juice samples.

3.1. Hydrophilic matrices for immobilisation of affinity-

capture proteins

Long flexible hydrophilic polymers are known to

provide a beneficial microenvironment for the immobi-

lisation of various biological macromolecules. Hydro-

philic matrices have been put in use to achieve an

improved performance of batch immunoassays likeELISA (Gregorius et al., 1995), SPR sensor chips (Lofas

et al., 1993), and piezoelectric biosensors (Liu et al.,

2001).

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/3426

Page 7: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

Four different hydrophilic polymers were employed

and compared in this work for immobilisation of the

affinity capture proteins: (i) PEI (MW 750 000 Da); (ii)

DEX (MW 64 000�/76 000 Da); (iii) PVA (MW 89 000�/

98 000, (iv) AMD, and finally (v) APTS used as a

reference to estimate the efficiency of the hydrophilic

polymer modification.

Matrices based on DEXs have for many years been

used in size-exclusion chromatography, and recently

DEX coatings have also been recognised as efficient

supports for immobilisation of antibodies and other

proteins. Carboxymethyldextran (Howell et al., 1998)and periodate-oxidised DEX (Penzol et al., 1998) are the

most widely used DEX linkers for surface modification.

Utilisation of DEX as a coupling matrix for protein A

has been shown to improve the reusability under acidic

regeneration conditions and increase the sensitivity,

binding capacity and specificity (Lu et al., 2000). The

topology of DEX coatings on a silica surface is

dependent on the polymer molecular weight, and thedensity of DEX clusters decreases with the increase of

molecular weight (Tasker et al., 1996).

Antibody immobilisation in a hydrophilic PEI layer

has been reported in a number of publications (Babacan

et al., 2000; Yakovleva et al., 2002), where improved

sensor stability and reusability in comparison with

conventional APTS-functionalised supports were

achieved. The ability of PEI to form a thin polymerlayer irreversibly bound to the surface of silicon dioxide

by direct adsorption makes this polymer very attractive

for fast and efficient surface modification procedure for

antibody immobilisation.

PVA matrix is widely applied for the immobilisation

of enzymes (Rossi et al., 1999) and cells (Khoo and

Ting, 2001), however, the use of this polymer as a

functionalised support in immunoanalytical applicationsis limited by a few examples (Barbosa et al., 2000). The

stabilising effect of PVA on the activity of immobilised

antibodies was reported for long-term storage of ad-

sorbed IgG (Dankwart et al., 1998).

AMD is a less common polymeric support for

immobilisation of biospecific ligands, mainly due to

the elaborate protocol of preparation. Nonetheless,

several applications of AMD for optical immunosensorsbased on immobilised antigens were reported in litera-

ture, demonstrating the lowest levels of non-specific

binding and the maximal binding capacity of AMD-

modified immunosurfaces (Piehler et al., 1996; Brecht et

al., 1998).

Considering the particular feasibility of hydrophilic

polymers to improve the performance of immobilised

biomolecules, we applied different coupling matrices totether affinity proteins to the microchip surface, and

performed a comparative study of stability and sensi-

tivity of the resulting microfluidic immunosensors.

Protein A and G in combination with PEI, DEX,

PVA, AMD polymers or conventional APTS support

were used to establish the most efficient immobilisation

procedure.

3.2. Optimisation of assay procedure

A systematic study of assay parameters (immunor-

eagent concentrations, nature of affinity protein, on-

chip incubation and off-line preincubation time of

immunoreagents) that might affect the antibody�/anti-

gen reaction, and also the binding of immunocomplexes

was performed.

The amount of antibodies is an important factorinfluencing the sensitivity of a competitive immunoas-

say, and lower antibody concentrations should be used

to facilitate a true competition and reach the most

sensitive assay (Hock, 1997). Optimal working antibody

concentrations for APTS, PEI and DEX modified

immunosensors with immobilised protein A and G

were determined by recording the antibody dilution

curves, and summarised in Table 1. As seen from thesedata, protein G immunosensors result in lower optimal

working antibody concentrations compared with pro-

tein A, with DEX modified protein G sensor giving the

lowest value. Table 1 also shows the parameters of

competitive affinity capture assays performed with

atrazine as a model analyte. The values of calibration

curve midpoint (IC50) are consistent with the optimal

antibody concentration obtained for immunosensorswith immobilised protein A and G, displaying the

most sensitive atrazine calibration curve with DEX-

protein G as an affinity capture support. The presented

data also indicate the effectiveness of PEI modification

of silica surface for immobilisation of protein A, as

follows from the comparison of IC50 values for APTS-

and PEI-modified sensors with immobilised protein A.

Based on the optimal antibody conditions, theimmunosensors with protein G were chosen for further

experiments due to lower reagents consumption and

superior assay sensitivity. For additional optimisation of

affinity protein immobilisation procedure we also pre-

pared protein G immunosensors modified with PVA

and AMD. Optimal antibody concentrations for PVA

and AMD modified immunosensors were 5 and 25 nM,

respectively, however, AMD sensor required the appli-cation of higher tracer concentration (9 nM).

In the next step, different protein G immunosensors

were used to investigate the immunoreagent incubation

time. The assays were performed with or without on-

chip incubation of antibody and labelled antigen, thus

influencing the kinetics of immunocomplex affinity

capture by protein G (Fig. 4a and b). The on-chip

incubation of the immunocomplex did not have aconsiderable influence on the binding response obtained

with the PEI and DEX sensors (Fig. 4a), whereas 2�/5

min incubation was required for stable and intense

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/34 27

Page 8: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

response using the PVA and AMD sensors (Fig. 4b).

The importance of on-chip incubation for the PVA and

AMD sensors may indicate the steric hindrance arising

from densely cross-linked polymer layers, which inter-

feres with the binding of the large immunocomplex.The effect of off-line pre-incubation of the antibody,

enzyme tracer and analyte on the assay performance was

investigated using the protein G based-immunosensors

and the optimised on-chip incubation conditions. Atra-

zine calibration curves were run by 20 min pre-incuba-

tion of the components or by immediate injection of the

immunoreagents to the microfluidic system. As seen in

Table 2, the assay parameters varied dramatically for

PEI and DEX based assays, leading to a much better

limit of detection (LOD) and curve midpoint (IC50)

values without any pre-incubation compared with the

PVA based assay. Off-line pre-incubation for 20 min

resulted, however, in a significant drop in sensitivity for

the PEI and DEX based assays. Taking into considera-

tion that antibody affinity constant toward the labelled

antigen is usually higher than that toward the target

analyte (Hatzidakis et al., 2002) one can assume that the

equilibrium for the formation of immunocomplex dur-

ing the off-line pre-incubation will shift toward the

binding of enzyme tracer, thus leading to a decrease in

assay sensitivity. For the PVA microchip, the observed

sensitivity was practically equal for both 20 min pre-

incubation and immediate injection of the components.

This fact may account for the influence of immunocom-

plex on-chip incubation, which is necessary to obtain the

intense and reproducible CL signals. Thus, in the case ofPVA microchip the equilibrium shift toward the forma-

tion of immunocomplex with enzyme tracer may occur

during the on-chip incubation step. As a result, the

observed sensitivity for PVA microchip is lower than

that for PEI and DEX microchips, and not dependent of

off-line pre-incubation of the components.

3.3. Regeneration and short- and long-term stability

Regeneration of immunosurfaces is a key factor inimmunosensor development, which implies, in regard to

affinity capture surfaces, the desorption of bound

immunocomplex without affecting the activity of im-

mobilised proteins or damaging the bonds between the

protein, modifying polymer layer, and silica surface.

In comparison with immobilised antibodies, protein

A and G are known to require more drastic conditions

for desorption of bound material, whereas the stabilityof immobilised affinity proteins is much higher than for

antibodies (Gonzalez-Martinez et al., 1998; Gonzalez-

Martinez et al., 1999; Quinn et al., 1999). Based on our

previous findings (Yakovleva et al., 2002), the best

recovery of immobilised antibody layer was achieved by

Table 1

Optimal antibody concentrations and curve midpoint values (IC50) for different protein A and G immunosensors

Microchip immunosensor APTS-protein A PEI-protein A DEX-protein A PEI-protein G DEX-protein G

[Ab] (nM) 5 5 5 4.2 3.1

IC50 (mg l�1) 0.4719/0.132 0.1349/0.013 �/ 0.0889/0.012 0.0729/0.005

Optimal antibody concentrations were determined from antibody dilution curves as values, corresponding to 70% binding of enzyme tracer. IC50

values were determined from atrazine calibration curves, obtained at optimal antibody concentration. Tracer concentration was 4.5 nM in all

experiments. The conditions were as follows: carrier buffer 50 mM Tris�/HCl (pH 7.4), substrate buffer 50 mM Tris�/HCl (pH 9.0), regeneration

solution 0.4 M glycine�/HCl (pH 2.2), carrier flow rate 40 ml min�1, substrate flow rate 50 ml min�1.

Fig. 4. Effect of on-chip incubation time on the immunocomplex binding response. A mixture of antibody and enzyme tracer taken at optimal

concentrations (Table 1 and Section 3.2) was injected to the microfluidic system (4 ml). The mixture was either allowed to pass through the affinity

capture surface (40 ml min�1) without incubation (peaks 1) or the carrier flow was stopped when the mixture reached the immunosensing surface, and

incubated for 2�/5 min (peaks 2). The flow cell was washed for 2 min, and a CL signal after triplicate substrate injection (50 ml min�1) was registered.

(a) Shows the peak pattern for protein G immunosensors modified with DEX and PEI. (b) Shows the peak pattern for protein G immunosensors

modified with AMD and PVA.

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/3428

Page 9: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

the application of acidic pH, and, especially by using

glycine�/HCl buffer (0.4 M, pH 2.2). The same buffer

was used in this work to remove the bound immuno-

complex from protein A and G sensing surfaces.

For comparison of the reusability and thus short-term

stability of immunosurfaces based on different poly-

mers, the immunocomplex binding response was mon-

itored after several repetitive binding-desorption cycles.

Fig. 5a and b show the regeneration profiles given by

different immunosensors based on protein A and G,

respectively. Each binding cycle was accompanied by

three substrate injections in order to monitor the signal

reproducibility, which can be impaired by dissociation

of antibody�/affinity protein complex. As seen in Fig.

5a, the protein A coatings displayed poor reproduci-

bility due to the elution of bound immunocomplex

within one binding cycle (triplicate substrate injection)

by the carrier flow. The protein G coatings were more

reproducible, with the best short-term stability exhibited

by PEI and DEX modified immunosensors (Fig. 5b).

Such behaviour of protein A and protein G may be

explained by different affinity of these proteins to the

sheep IgG used in this work (Hermanson et al., 1992).

Since protein A coatings did not display sufficient

reproducibility and stability within regeneration experi-

ments, further studies were performed with protein G

immunosensors. Estimation of the immunosensor long-

term stability after long periods of storage and use was

based on the following criteria: (i) the ability to maintain

a stable zero analyte dose signal and (ii) the reproduci-

bility of the whole calibration curve and its parameters.

Fig. 6 shows the immunocomplex binding response

obtained on different days for protein G immunosensors

modified with PEI, DEX, and PVA. As seen from these

data, the immunocomplex binding response observed

for PEI and DEX based immunosensors is consistent

throughout a period of 260 days. No decrease of surface

binding capacity was registered for these immunosen-

sors, and in addition, a minor activation effect was

observed after long period of storage (between day 30

and 180). The PVA immunosensor displayed a gradual

decrease of binding response already within a 2-week

Table 2

Influence of off-line pre-incubation on assay parameters using PEI, DEX and PVA modified protein G immunosensors

Immunosensor No off-line pre-incubation Off-line pre-incubation for 20 min

PEI DEX PVA PEI DEX PVA

LOD (mg l�1) 0.006 0.005 0.065 0.15 0.17 0.17

IC50 (mg l�1) 0.089/0.02 0.089/0.03 0.859/0.31 1.229/0.16 0.969/0.11 0.749/0.08

Slope 0.599/0.08 0.419/0.12 0.629/0.28 0.969/0.11 1.199/0.08 1.299/0.16

Atrazine calibration curves with or without off-line pre-incubation were run on the same day. The curve midpoint (IC50) and a slope were extracted

from a four-parameter logistic equation used to fit the sigmoidal curve. LOD values were defined as the analyte concentration corresponding to 90%

tracer binding. Antibody concentrations for PEI, DEX and PVA modified immunosensors were 4.2, 3.1 and 5 nM, respectively. On-chip incubation

for PVA immunosensor was 2 min. PEI and DEX immunosensors were used without on-chip incubation. Tracer concentration and flow conditions

were the same as in Table 1.

Fig. 5. Short term stability of immunosensors. The intensity of CL signal after three consecutive cycles of binding (injection of antibody and enzyme

mixture; 4 ml; 40 ml min�1) and desorption (a flow of 0.4 M glycine�/HCl, pH 2.2; 3 min; 50 ml min�1) is plotted vs. time. (a) Shows the

immunocomplex binding response for protein A immunosensors. On-chip incubation for 2 min, antibody and tracer conditions were as in Table 1.

(b) Shows the immunocomplex binding response for protein G immunosensors. No on-chip incubation was performed for PEI and DEX modified

sensors, antibody and tracer conditions were as in Table 1. On-chip incubation for 2 min was applied for PVA and AMD modified sensors, antibody

concentrations were 5 and 25 nM, respectively. Tracer concentration was 5 nM.

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/34 29

Page 10: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

period, thus showing much lower stability than PEI and

DEX based immunosensors.

The monitoring of atrazine calibration curve varia-

bility with time for DEX and PEI modified immuno-

sensors containing immobilised protein G is presented in

Fig. 7a and b, respectively. Atrazine calibration curves

coincide very well for the immunosensor modified with

DEX (Fig. 7a), whereas for PEI more dramatic increase

of the IC50 value and thus decrease of sensitivity was

observed after 215 days of storage and use (Fig. 7b).

Inter-assay precision for both PEI and DEX based

immunosensors (R.S.D. within the linear range of the

calibration curve) was below 15% over 8-month period,

thus evidencing that both immunosensors are capable of

long-term use.

An interesting observation is that PEI and DEX

modified protein G immunosensors exhibited similar

and very high long-term stability, although the PEI

modification is performed by physical adsorption of thepolymer on the silica surface, while DEX modification is

based on covalent attachment of the polymer to the

microchip surface. Immobilisation of antibodies on a

PEI layer performed in our previous work (Yakovleva et

al., 2002) resulted in immunosensor lifetime of 1 month,

whereas this work demonstrates that the significant

prolongation of PEI based immunosensor lifetime may

be achieved by making use of the affinity capture assayformat.

3.4. Analytical performance

Investigation of the analytical characteristics of the

atrazine assay involved a comparison between different

hydrophilic polymers used for modification of silicon

microchip surface and immobilisation of protein G. The

parameters of the calibration curves obtained with PEI-,

DEX-, PVA-, and AMD-protein G immunosensors are

given in Table 3. As seen from these data, application ofdifferent matrices for protein G immobilisation signifi-

cantly influenced the assay performance, where the PEI-

and DEX modified immunosensors were found to be the

most sensitive. A correlation was found between the

optimal antibody conditions (Table 1 and Section 3.2)

and the sensitivity of the calibration curve. The PEI and

DEX microchip modification procedures resulted in the

lowest optimal immunoreagent concentrations and mostsensitive atrazine calibration curves, thus being the most

suitable choices for affinity protein immobilisation.

Fig. 6. The long-term stability of PEI, DEX, and PVA based

immunosensors with immobilised protein G. Immunocomplex binding

responses at zero analyte doze were measured on different days and

plotted vs. time over. Average signals from three repetitive binding-

desorption cycles are shown. The conditions were the same as in Fig. 5.

Fig. 7. Variability of atrazine calibration curves obtained with protein G immunosensors modified with PEI and DEX. Calibration curves

constructed on different days were fit to a four parameter logistic equation: y�/(A�/D)/[1�/(x/C)B]�/D, where A and D are the maximal and minimal

signals of the assay, respectively, B corresponds to the slope of the sigmoid and C is the concentration of atrazine resulting in 50% inhibition of

binding of the enzyme-tracer to antibody (IC50). (a) Atrazine calibration curves using the DEX based immunosensor. (b) Atrazine calibration curves

using the PEI based immunosensor. Each curve is an average of three replicates, performed on each day. (�/) Represents the IC50 values (right y-

axis) plotted vs. time (top x-axis), extracted from the corresponding four-parameter curves. Antibody and tracer conditions were as in Table 1. No

on-chip incubation or off-line pre-incubation was performed.

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/3430

Page 11: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

Despite the different analytical characteristics dis-

played by the protein G immunosensors, the LODvalues of all assays with exception for the PVA-based

were below the threshold level for single pesticide

content in drinking water according to official European

Union (EU) requirements (0.1 mg l�1). PEI and DEX

modified protein G immunosensors were chosen for

further analysis of surface water and fruit juice samples,

considering their good stability (Figs. 6 and 7) and that

the obtained IC50 values were around 0.1 mg l�1.

3.5. Analysis of spiked surface water samples

Surface water collected from a small river in an

agricultural region in southern Sweden was used for

this study. In order to investigate the matrix effect,

surface water samples were spiked with atrazine to give

different analyte concentrations between 0 and 10 mg

l�1. Crude surface water and surface water diluted with

50 mM Tris�/HCl buffer at various ratios were used toprepare the spiked samples and construct atrazine

calibration curves, see Fig. 8. No calibration curve

could be obtained with atrazine standards prepared in

crude surface water, the assay resulted in very low light

intensity, indicating strong matrix influence. Water

sample diluted 1:2 in carrier buffer lead to the calibra-

tion curve being shifted to the range of higher analyte

concentrations, with zero analyte dose signals approxi-mately 1.5 times lower than that obtained in the carrier

buffer. Almost identical calibration curves were ob-

served for Tris�/HCl and surface water diluted with

buffer 1:4.

Surface water was fortified with atrazine added at 0.1,

1 and 3 mg l�1 and then diluted 1:4 with carrier buffer,

i.e. after dilution the obtained concentrations fell into

the assay working range. The mean recovery data for thethree spiked samples were 87�/102% as shown in Table

4, thus confirming that 1:4 sample dilution in the carrier

was sufficient to minimise interference for surface water

matrix, however, as seen in Table 4, the precision,

especially for higher concentrations, was rather low.

These finding must be taken into account for assay

optimisation in any future work.

3.6. Analysis of spiked fruit juice samples

Food matrix effects are well-known phenomena inimmunoassays, and hence the influence of orange juice

on the performance of the DEX-immunosensor was

investigated in order to estimate the applicability to

food analysis. The atrazine recovery and matrix effects

were tested in orange juice in a similar way as for surface

water. Orange juice was diluted by various factors (1:10,

1:50, 1:200, 1:400 and 1:500) with carrier buffer. The

influence of juice dilution on the zero analyte dose signalis shown in Fig. 9. As can be seen, at least 1:400 dilution

is required in order to eliminate the matrix effects. The

obtained results indicate that the presented protein G

Table 3

Assay characteristics for protein G immunosensors modified with different hydrophilic polymers

Hydrophilic polymer PEI DEX PVA AMD

LOD (mg l�1) 0.006 0.010 0.202 0.051

IC50 (mg l�1) 0.0969/0.018 0.1309/0.027 0.8569/0.307 1.6049/0.597

Dynamic range (mg l�1) 0.014�/1.120 0.014�/0.822 0.173�/4.580 0.077�/13.58

R.S.D.intra (%) 4.98 5.62 10.84 5.75

R.S.D.inter (%) 15.33 11.45 13.26 23.10

The LOD was calculated as atrazine concentration corresponding to 90% tracer binding. The dynamic range was defined by the analyte

concentration inhibiting the maximum signal by 20�/80%, and the IC50 value was estimated as the analyte concentration that inhibits 50% of tracer

binding. The intra-assay precision (R.S.D.intra) was calculated as the average R.S.D. between three replicates within one selected calibration curve,

using all points of the assay dynamic range. The inter-assay precision (R.S.D.inter) was calculated as the R.S.D. between three different averaged

assays performed on different days, including all points (N�/3) in the dynamic range of the assays. Antibody conditions for PEI and DEX modified

immunosensors were as in Table 1, without on-chip incubation. Antibody concentrations for PVA and AMD modified immunosensors were 5 and 25

nM, respectively, on-chip incubation 2 min. Tracer concentration for PEI, DEX and PVA sensors was 4.5 nM, for AMD sensor 9 nM. No off-line

pre-incubation was performed in all experiments. Flow conditions were as in Table 1.

Fig. 8. The surface water matrix influence on the DEX immunosensor.

Atrazine calibration curves were constructed using surface water at

various dilutions or the carrier buffer (50 mM Tris�/HCl, pH 7.4)

fortified with atrazine at 0.01, 0.1, 1 and 10 mg l�1. The plotted signals

are averages of three measurements. The conditions are as in Fig. 5.

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/34 31

Page 12: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

immunosensors with chemiluminescent detection are

very sensitive to interference from the complex vegetable

matrices.

In order to estimate atrazine recoveries, orange juice

was spiked with atrazine at 80, 160 and 240 mg l�1 to

give analyte concentrations within the assay working

range after a 1:400 dilution in the carrier buffer. The

spiked samples were analysed using the calibration curve

obtained the same day. Table 5 shows the results

obtained for each fortified juice sample determined in

triplicate with recoveries between 88�/124%. In this case,

however, better precision was obtained since R.S.D.

values did not exceed 7%. For environmental samples

the US EPA recommendation is that the recovery

should fall between 70�/120% (Krotzky and Zeeh,

1995) and based on this the sensor is suitable for

atrazine quantification in juice.

4. Conclusions

Microfluidic immunosensors based on affinity pro-

teins immobilised on silicon microchip surfaces and CL

detection were developed. Various hydrophilic polymer

coatings were employed for immobilisation protein A

and G and used for comparative study to establish the

best activity, reusability and sensitivity of the sensor.

Affinity capture assay based on protein A or G

represents a generic tool, applicable for a wide range

of analytes.

Of the two affinity proteins used in this study, protein

G resulted in better reusability and long-term stability of

the immunosensors compared with protein A. The

matrix of the hydrophilic polymer was an other para-

meter influencing the stability and sensitivity of the

assay, where the PEI and DEX modified microchips

offered the most sensitive detection. The other polymers

used here, PVA and AMD, resulted in worse reusability

and sensitivity of protein G immunosensors. Addition-

ally, the assay characteristics were found to be substan-

tially dependent on the incubation conditions, where

off-line pre-incubation of the tracer, analyte and anti-

body led to an appreciable loss in sensitivity. Based on

the optimised immobilisation chemistry and assay con-

ditions, we developed microfluidic immunosensors for

determination of atrazine, with calibration curve mid-

points around the EU threshold lever for drinking water

(0.1 mg l�1) and long-term stability over a period of 8

months.

Due to the high sensitivity provided by PEI and DEX

coated protein G immunosensors, they were applied for

atrazine determination in surface water and juice

samples. Surface water matrix effects could be mini-

Table 4

Recoveries in surface water fortified with atrazine

Fortification concentration (mg l�1) Found concentration (mg l�1) S.D. Recovery (%9/R.S.D.)

0.5 0.510 0.028 1029/15

1 0.880 0.279 889/32

3 2.623 0.684 879/26

Surface water was first spiked with known amounts of atrazine and then diluted 1:4 with 50 mM Tris�/HCl buffer (pH 7.4) and analysed using the

DEX immunosensor. Averages of three replicates obtained by interpolation into a calibration curve constructed the same day as the measurement

was performed. Antibody, tracer and flow conditions were as in Table 1, no on-chip or off-line incubation was performed.

Fig. 9. The orange juice matrix influence on the PEI immunosensor is

shown. Tracer signals in the presence of zero analyte doze are plotted

at different dilutions of juice sample with 50 mM Tris buffer, pH 7.4.

Averages of three measurements are shown.

Table 5

Recoveries in orange juice fortified with atrazine

Fortification concentration (mg l�1) Found concentration (mg l�1) S.D. Recovery (%9/R.S.D.)

80 99.1 6.9 1249/7

160 163.5 10.8 1029/7

240 210.0 13.0 889/6

Orange juice was first spiked with known amounts of atrazine and then diluted 1:400 with 50 mM Tris�/HCl buffer (pH 7.4) and analysed using the

PEI immunosensor. The calibration curve and sample assay performed in triplicate the same day were used to calculate the average recoveries and

R.S.D. of the samples. Antibody, tracer and flow conditions were as in Table 1, no on-chip or off-line incubation was performed.

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/3432

Page 13: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

mised by a 1:4 dilution of the analysed sample, and the

recovery of atrazine in surface water spiked with

concentrations between 0.5 and 3 mg l�1 was 87�/

102%. The application to the analysis of orange juicesamples required at least 1:400 dilution, with recoveries

between 88 and 124% for juice spiked with concentra-

tions between 80 and 240 mg l�1 atrazine. The total

assay time was 10 min including, sample injection, three

substrate injections and regeneration.

High sensitivity, versatility, suitability for point-of-

care analysis, high sample throughput and possibility to

build up array systems are important advantages ofmicrofluidic immunosensors, which make these kind of

devices very promising for the future application in

many fields, including potential multiple biological

information and environmental multiplex analysis using

channel arrays with different sensing elements immobi-

lised. For the future we anticipate further development

of a multi-assay platform, making use of different

immobilised biorecognition elements, such as receptors,antibodies, enzymes and cells.

Acknowledgements

The authors acknowledge financial support from the

European Commission (INCO Copernicus project ER-

BIC15-CT98-0910), the Swedish Foundation for Strate-

gic Environmental Research (MISTRA), the Swedish

Council for Forestry and Agricultural Research (SJFR),and the Swedish Research Council (Vetenskapsradet).

Authors also wish to thank Dr Ramadan Abuknesha

(King’s College University of London, UK) for provid-

ing us with anti-atrazine antibodies and Dr Oleg

Kolyasnikov (Department of Chemical Enzymology,

Faculty of Chemistry, Moscow State University) for

useful discussions. Sven Hagg (Department of Analy-

tical Chemistry, Lund University) is acknowledged forproviding us with a home made software for data

acquisition.

References

Anderson, G.P., Jacoby, M.A., Ligler, F.S., King, K.D., 1997.

Effectiveness of protein A for antibody immobilization for a fiber

optic biosensor. Biosens. Bioelectron. 12 (4), 329�/336.

Askari, M., Alarie, J.P., Moreno-Bondi, M., Vo-Dinh, T., 2001.

Application of an antibody biochip for p53 detection and cancer

diagnosis. Biotechnol. Prog. 17 (3), 543�/552.

Auroux, P.-A., Iossifidis, D., Reyes, D.R., Manz, A., 2002. Micro total

analysis systems. 2. Analytical standard operations and applica-

tions. Anal. Chem. 74 (12), 2637�/2652.

Babacan, S., Piravnik, P., Letcher, S., Rand, A.G., 2000. Evaluation of

antibody immobilization methods for piezoelectric biosensor

application. Biosens. Bioelectron. 15 (11�/12), 615�/621.

Barbosa, G.H., Santana, E.M., Almeida, A.M., Araujo, A.M.,

Faitibello-Filho, O., Carvallo, L.B., Jr, 2000. The use of filter

paper plasticised with polyvinyl alcohol-glutaraldehyde in ELISA.

Braz. J. Med. Biol. Res. 33 (7), 23�/827.

Bernard, A., Michel, B., Delamarche, E., 2001. Micromosaic immu-

noassays. Anal. Chem. 73 (1), 8�/12.

Bjarnasson, B., Chimuka, L., Onnerfjord, P., Eremin, S., Jonsson, J.-

A., Johansson, G., Emneus, J., 2000. Enzyme flow immunoassay

using a protein G column for the screening of triazine herbicides in

surface and waste water. Anal. Chim. Acta 426 (2), 197�/207.

Brecht, A., Klotz, A., Barzen, C., Gauglitz, G., Harris, R.D., Quigley,

G.R., Wilkinson, J.S., Sztajnbok, P., Abuknesha, R., Gascon, J.,

Oubina, A., Barcelo, D., 1998. Optical immunoprobe development

for multiresidue monitoring in water. Anal. Chim. Acta 362 (1),

69�/79.

Burestedt, E., Nistor, C., Schagerlof, U., Emneus, J., 2000. An enzyme

flow immunoassay that uses beta-galactosidase as the label and a

cellobiose dehydrogenase biosensor as the label detector. Anal.

Chem. 72 (17), 4171�/4177.

Chiem, N.H., Harrison, D.J., 1998. Microchip systems for immunoas-

say: an integrated immunoreactor with electrophoretic separation

for serum theophilline determination. Clin. Chem. 44 (3), 591�/598.

Dankwart, A., Muller, J., Hock, B., 1998. Stabilisation of enzyme

immunoassay for atrazine. Anal. Chim. Acta 362 (1), 35�/45.

Dolnik, V., Liu, S., Jovanovich, S., 2000. Capillary electrophoresis on

microchip. Electrophoresis 21 (1), 41�/54.

Drott, J., Laurell, T., Rosengren, L., Lindstrom, K., 1997. Porous

silicon as the carrier matrix in microstructured enzyme reactors

yielding high enzyme activities. J. Micromach. Microeng. 7 (1), 14�/

23.

Eteshola, E., Leckband, D., 2001. Development and characterisation

of an ELISA assay in PMDS microfluidic channels. Sens. Actuat. B

72 (129), 133.

Garcinuno, R.M., Fernandez, P., Perez-Conde, C., Guttierez, A.M.,

Camara, C., 2000. Development of a fluoroimmunosensor for

theophylline using immobilized antibody. Talanta 52 (5), 825�/832.

Giersch, T., 1993. A new monoclonal-antibody for the sensitive

detection of atrazine with immunoassay in microtiter plate and

dipstick format. J. Agric. Food Chem. 41 (6), 1006�/1011.

Gonzalez-Martinez, M.A., Morais, S., Puchades, R., Maquieira, A.,

Marco, M.-P., Barcelo, D., 1998. Automation of a heterogeneous

enzyme immunoassay for atrazine. Comparison of three immobi-

lization supports. Fresenius J. Anal. Chem. 361 (2), 179�/184.

Gonzalez-Martinez, M.A., Puchades, R., Maquieira, A., 1999. On-line

immunoanalysis for environmental pollutants: from batch assays

to automated sensors. Trends Anal. Chem. 18 (3), 204�/218.

Goodrow, M.H., Harrison, R.O., Hammock, B.D., 1990. Hapten

synthesis, antibody development, and competitive inhibition en-

zyme-immunoassay for s-triazine herbicides. J. Agric. Food Chem.

38 (4), 990�/996.

Gregorius, K., Mouritsen, S., Elsner, H.I., 1995. Hydrocoating: a new

method for coupling biomolecules to solid phases. J. Immunol.

Methods 181 (1), 65�/73.

Hashimoto, M., Tsukagoshi, K., Nakajima, R., Kondo, K., Arai, A.,

2000. Microchip capillary electrophoresis using on-line chemilumi-

nescence detection. J. Chromatogr. A 867, 271�/279.

Hatzidakis, G.I., Tsatsakis, A.M., Krambovitis, E.K., Spyros, A.,

Eremin, S.A., 2002. Use of L-lysine fluorescence derivatives as

tracers to enhance the performance of polarization fluoroimmu-

noassays. A study using two herbicides as model antigens. Anal.

Chem. 74 (11), 2513�/2521.

Hayes, M.A., Polson, N.A., Phayre, A.N., Garcia, A.A., 2001. Flow-

based microimmunoassay. Anal. Chem. 73 (24), 5896�/5902.

Hermanson, G.T., Mallia, A.K.P., Smith, P., 1992. Peptide antigens,

antibodies, and immunoglobulin binding proteins. In: Immobilized

Affinity Ligand Techniques. Academic Press, San Diego, pp. 210�/

251.

Hock, B., 1997. Antibodies for immunosensors. Anal. Chim. Acta 347

(1�/2), 177�/186.

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/34 33

Page 14: Microfluidic enzyme immunosensors with immobilised protein A and G using chemiluminescence detection

Howell, S., Kenmore, M., Kirkland, M., Badley, R.A., 1998. High

density immobilization of an antibody fragment to a carboxy-

methylated dextran-linked biosensor surface. J. Mol. Recognit. 11

(1�/6), 200�/203.

Khoo, K.M., Ting, Y.P., 2001. Polyvivyl alcohol as an immobilisation

matrix*/a case of gold biosorption. Water Sci. Technol. 43 (11),

17�/23.

Koutny, L.B., Schmalzing, D., Taylor, T.A., Fuchs, M., 1996.

Microchip electrophoretic immunoassay for serum cortisol. Anal.

Chem. 68 (1), 18�/22.

Kricka, L.J., 2001. Microchips, microarrays, biochips and nanochips:

personal laboratories for the 21st century. Clin. Chim. Acta 307

(1�/2), 219�/223.

Kronkvist, K., Lovgren, U., Svensson, J., Edholm, L.E., 1997.

Competitive flow injection enzyme immunoassay for steroids using

a post-column reaction technique. J. Immunol. Methods 200 (1�/2),

145�/153.

Krotzky, A.J., Zeeh, B., 1995. Immunoassays for residue analysis of

agrochemicals: proposed guidelines for precision, standardization

and quality control. Pure Appl. Chem. 67 (12), 2065�/2088.

Lacher, N.A., Garrison, K.E., Martin, R.S., Lunte, S.M., 2001.

Microchip capillary electrophoresis/electrochemistry. Electrophor-

esis 22 (12), 2526�/2536.

Laurell, T., Drott, J., Rosengren, L., Lindstrom, K., 1996. Enhanced

enzyme activity in silicon integrated enzyme reactors utilizing

porous silicon as the coupling matrix. Sens. Actuat. B 31 (3),

161�/166.

Liu, Y.C., Wang, C.M., Hsiung, K.P., 2001. Comparison of different

protein immobilization methods on quartz crystal microbalance

surface in flow injection immunoassay. Anal. Biochem. 299 (2),

130�/135.

Lofas, S., Johnsson, B., Tegendal, K., Ronnberg, I., 1993. Dextran

modified gold surfaces for surface plasmon resonance sensors:

immunoreactivity of immobilized antibodies and antibody-surface

interaction studies. Coll. Surf. B: Biointerf. 1 (2), 83�/89.

Lu, H.-C., Chen, H.-M., Lin, Y.-S., Lin, J.W., 2000. A reusable and

specific protein A-coated piezoelectric biosensor for flow injection

immunoassay. Biotechnol. Prog. 16 (1), 116�/124.

McCreedy, T., 2000. Fabrication techniques and materials commonly

used for the production of microreactors and micro total analytical

systems. Trends Anal. Chem. 19 (6), 396�/401.

Nistor, C., Rose, A., Farre, M., Stoica, L., Wollenberg, U., Ruzgas,

T., Barcelo, D., Gorton, L., Emneus, J., 2002. In-field monitoring

of cleaning efficiency in waste water treatment plants using two

phenol-sensitive biosensors. Anal. Chim. Acta 456 (1), 3�/17.

Penalva, J., Gonzalez-Martinez, M.A., Puchades, R., Maquieira, A.,

Marco, M.-P., Barcelo, D., 1999. Immunosensor for trace deter-

mination of Irgarol 1051 in sea water using organic media. Anal.

Chim. Acta 387 (3), 227�/233.

Penzol, G., Armisen, P., Fernandez-Lafuente, R., Rodes, L., Guisan,

J.M., 1998. Use of dextrans as long and hydrophilic spacer arms to

improve the performance of immobilized proteins acting on

macromolecules. Biotechnol. Bioeng. 60 (4), 518�/523.

Piehler, J., Brecht, A., Geckeler, K.E., Gauglitz, G., 1996. Surface

modification for direct immunoprobes. Biosens. Bioelectron. 11

(6�/7), 579�/590.

Pulido-Tofino, P., Barrero-Moreno, J.M., Perez-Conde, C., 2000.

Flow-through fluoroimmunosensor for isoproturon determination

in agricultural foodstuff. Evaluation of antibody immobilization

on sold support. Anal. Chim. Acta 417 (1), 85�/94.

Quinn, J., Patel, P., Fotzpatrick, B., Manning, B., Dillon, P., Daly, S.,

O’Kennedy, R., Alcoter, M., Lee, H., Morgan, M., Lang, K., 1999.

The use of regenerable, affinity ligand-based surfaces for immuno-

sensor applications. Biosens. Bioelectron. 14 (6), 587�/595.

Reyes, D.R., Iossifidis, D., Auroux, P.-A., Manz, A., 2002. Micro total

analysis systems. 1. Introduction, theory and technology. Anal.

Chem. 74 (12), 2636�/2636.

Rossi, A., Morana, A., Lernia, I.D., Di Tombrino, A., De Rosa, M.,

1999. Immobilization of enzymes on spongy polyvinyl alcohol

cryogels: the example of beta-galactosidase from Aspergillus

oryzae . Ital. J. Biochem. 48 (2), 91�/97.

Sanders, G., Manz, A., 2000. Chip-based microsystems for genomic

and proteomic analysis. Trends Anal. Chem. 19 (6), 364�/378.

Sato, K., Tokeshi, M., Odake, T., Kimura, H., Ooi, T., Nakao, M.,

Kitamori, T., 2000. Integration of an immunosorbent assay system:

analysis of a secretory human immunoglobulin A on polystyrene

beads in a microchip. Anal. Chem. 72 (6), 1144�/1147.

Sato, K., Tokeshi, M., Kimura, H., Kitamori, T., 2001. Determination

of carcinoembyonic antigen in human sera by integrated bead-bed

immunoassay in a microchip for cancer diagnosis. Anal. Chem. 73

(6), 1213�/1218.

Sheikh, S.H., Mulchandani, A., 2001. Continuos flow fluoro-immu-

nosensor for paclitaxel measurement. Biosens. Bioelectron. 16 (9�/

12), 647�/652.

Silzel, J.W., Cercek, B., Dodson, C., Tsay, T., Obremski, R.J., 1998.

Mass-sensing, multianalyte microarray immunoassay with imaging

detection. Clin. Chem. 44 (9), 2036�/2043.

Tasker, S., Mathijs, G., Davies, M.C., Roberts, C.J., Schacht, E.H.,

Tendler, S.J.B., 1996. Molecular resolution imaging of dextran

monolayers immobilized on silica by atomic force microscopy.

Langmuir 12 (26), 6436�/6442.

Thust, M., Schoning, M.J., Schroth, P., Malkoc, U., Dicker, C.I.,

Steffen, A., Kordos, P., Luth, H., 1999. Enzyme immobilisation on

planar and porous silicon substrates for biosensor applications. J.

Mol. Catal. B 7 (1�/4), 77�/83.

Valat, C., Limoges, B., Huet, D., Romette, J.-L., 2000. A disposable

protein A-based immunosensor for flow-injection assay with

electrochemical detection. Anal. Chim. Acta 404 (2), 187�/194.

Wang, J., Ibanetz, A., Chatrach, M.P., Escarpa, A., 2001. Electro-

chemical enzyme immunoassays on microchip platforms. Anal.

Chem. 73 (21), 5323�/5327.

Weetall, H., 1976. Covalent coupling methods for inorganic support

materials. Methods Enzymol. 44, 134�/138.

Wilson, D.S., Nock, S., 2002. Review: functional protein microarrays.

Curr. Opin. Chem. Biol. 6 (1), 81�/85.

Yakovleva, J., Davidsson, R., Lobanova, A., Bengtsson, M., Eremin,

S., Laurell, T., Emneus, J., 2002. Microfluidic enzyme ummunoas-

say using silicon microchip with immobilized antibodies

and chemiluminescence detection. Anal. Chem. 74 (13), 2994�/

3004.

J. Yakovleva et al. / Biosensors and Bioelectronics 19 (2003) 21�/3434