A

BmsaaRFarsMEa4Cfcw©

K

1

dddm

0d

Biosensors and Bioelectronics 22 (2006) 268–274

Optimizing recombinant antibody function in SPR immunosensingThe influence of antibody structural format and chip surface

chemistry on assay sensitivity

S. Townsend a, W.J.J. Finlay a,b, S. Hearty a,c, R. O’Kennedy a,b,∗a School of Biotechnology, Dublin City University, Glasnevin, Dublin, Ireland

b Biomedical Diagnostics Institute, Dublin City University, Glasnevin, Dublin, Irelandc National Centre for Sensor Research, Dublin City University, Glasnevin, Dublin, Ireland

Received 9 November 2005; received in revised form 5 January 2006; accepted 9 January 2006Available online 17 February 2006

bstract

ackground: Recombinant antibody fragments are valuable tools for SPR-based detection of small molecules such as illicit drugs. However, theultiple structural formats of recombinant antibody fragments are largely uncharacterised with respect to their respective performance in SPR

ensing. We have expressed a model anti-M3G antibody in both scFv and chimeric Fab formats to examine its sensitivity and binding profiles inmicroplate immunoassay format and BiacoreTM. We have further examined the influence of scFv multimerisation, Fab constant region stability

nd SPR chip surface coating chemistry, on anti-hapten SPR assay development.esults: Under optimised competition ELISA conditions, the anti-M3G scFv was found to have an IC50 value of 30 ng/ml, while the most stableab construct exhibited an IC50 value of 2.4 ng/ml. In SPR competition assay on an M3G-OVA-coated SPR chip surface, the two constructsgain differed in sensitivity, with IC50 values of 117 and 19 ng/ml for the scFv and Fab, respectively (the scFv also exhibiting poor linearity ofesponse). However, when the SPR chip surface was directly coated with M3G, both antibody constructs exhibited good linearity of response,imilar high sensitivity IC50 values (scFv 30 ng/ml, Fab 14 ng/ml) and high reproducibility (50 effective regenerations for M3G-OVA, 200 for3G direct). During SPR assay development it was noticed that scFv and Fab constructs gave differing off-rate profiles. Subsequent HPLC,

LISA and electrophoretic analyses then confirmed that a portion of the scFv population multimerises. Bivalent scFv was found to profoundlyffect the dissociation curve for scFv in stringent SPR kinetic analyses, leading to a 40-fold difference in calculated off-rate values (Fab off rate.7 × 10−3 S−1, scFv off rate 1.03 × 10−2 S−1).

onclusion: The structural format of recombinant antibody fragments and chip functionalisation methodology can both profoundly affect the

unction of anti-M3G SPR assay, with direct coating and Fab format proving to be optimal. The confirmation of scFv multimerisation and resultinghanges in SPR kinetics profile, in comparison with a Fab, further suggest that caution must be taken in the interpretation of SPR sensorgrams,hich are commonly used in the ‘affinity ranking’ of scFv panels in which the extent of dimerisation in each sample is unknown.2006 Elsevier B.V. All rights reserved.

pdpa

eywords: Fab; scFv; SPR; Surface chemistry; Kinetics

. Introduction

With the increased importance of regular testing for illicitrug content in clinical or forensic samples, a clear goal is the

evelopment of simple, rapid and highly specific multi-analyteetection methods. Traditional methods for detecting small drugolecules, such as thin layer chromatography (TLC) and high-

∗ Corresponding author. Tel.: +353 1700 5319; fax: +353 1700 5412.E-mail address: Richard.Okennedy@dcu.ie (R. O’Kennedy).

Tma(fes

956-5663/$ – see front matter © 2006 Elsevier B.V. All rights reserved.oi:10.1016/j.bios.2006.01.010

erformance liquid chromatography (HPLC) have become out-ated due to their relative expense, complexity and low through-ut (Hall et al., 1993). Automated ‘real-time’ immunoassays aren emerging candidate to take their place (Yau et al., 2003).o achieve real-time analysis of small molecules, surface plas-on resonance (SPR) biosensors, such as BiacoreTM, are being

pplied to develop highly sensitive competition immunoassays

Rich and Myszka, 2005). The capacity for SPR to monitor label-ree binding events in real time has made it a popular method toxamine molecular activity. However, the efficacy of the biosen-or is heavily dependant on ligand immobilisation methodology

nd Bi

(uott

doa(RuwwswidstatoPtt2

sFfViCmtiscmFitvw1caaercf2atb

f1

faMic2fbraFcTtd

2

2

cueCtWm2

tCtCgeCrPTe(

2

EB2B

S. Townsend et al. / Biosensors a

Hock et al., 2002) and also on the kinetics of the binding reactionsed to perform the assay (Brennan et al., 2003). Heterogeneityf the binding ligand can directly impact on the performance ofhe biosensor and, therefore, homogeneity must be assured tohe greatest extent possible (O’Shannessy and Winzor, 1996).

The laborious techniques used in traditional antibody pro-uction have led several research groups to investigate the usef recombinant antibody technology, to produce scFv and Fabntibody fragments with high-affinities for small moleculesCharleton et al., 2000; Li et al., 2000; Tout et al., 2001;au et al., 2002; Brennan et al., 2003). The most commonlysed recombinant antibody format is the single-chain-Fv (scFv),hich consists of the variable (binding) regions of an antibody,hich are linked by a peptide chain. The scFv is a popular

tructural format as it can be rapidly constructed, is typicallyell expressed in Escherichia coli and can exhibit high affin-

ty and stability (Barbas et al., 2001). Most scFv molecules areesigned with an Fv-region linking peptide based on glycine-erine repeat sequences of 15–20 residues, which typically leado scFv molecules which are designated ‘monomeric’ (Maynardnd Georgiou, 2000), but in the bacterial periplasm a por-ion of the scFv typically dimerizes, resulting in a mixturef monomeric and dimeric fragments (McGregor et al., 1994;luckthun, 1994). Shortened linker scFv’s can also be generated,

o ensure that the majority of the functional scFv is dimeric andherefore has increased avidity (Wu et al., 1996; Finlay et al.,005).

A number of groups have engineered the variable regions ofcFvs for the expression of Fab fragments (Arndt et al., 1998).ab fragments are an alternative to the scFv and consist of theull antibody light chain (a single polypeptide containing the

� plus the CL region), which is joined by disulphide bond-ng to a second antibody domain comprised of the VH and

H1 regions (Barbas et al., 2001). Fab constructs are obligatelyonomeric and have been used specifically to determine the

rue affinities of scFv fragments, which had been modified forncreased affinity (Rau et al., 2002). However, none of thesetudies has specifically addressed the issues of SPR surfacehip chemistry in relation to different antibody structural for-ats and whether it is most appropriate to convert scFvs toab format, to ensure a monovalent binding population for use

n SPR analyses. Monovalent antibodies are frequently usedo measure binding constants (George et al., 1995) as mono-alency prevents the formation of antigen–bridge complexes,hich result in an increase in binding avidity (MacKenzie et al.,996). The monovalency of ligand–binding interaction is espe-ially important with regard to kinetic analysis on BiacoreTM

nd thus, also in cases where new antibody panels are rankedccording to affinity (Li et al., 2000; Lu et al., 2003; Stacyt al., 2003; Fredericks et al., 2004). Kinetic characterisationeveals the rate at which the immune complex both asso-iates and dissociates, thereby providing a stringent measureor tailoring and selecting reagents (Quinn and O’Kennedy,

001). The interaction between a soluble, monovalent analytend an immobilized monovalent ligand can be interpreted inerms of pseudo-first-order kinetics, but heterogenicity of theinding ligand and possible multivalency can cause deviations

TEcP

oelectronics 22 (2006) 268–274 269

rom pseudo-first-order behaviour (O’Shannessy and Winzor,996).

In the study presented herein, the effects of SPR chip sur-ace chemistry and antibody valency on the function of a modelnti-morphine-3-glucoronide (M3G) SPR assay is investigated.3G is an important potential analyte for routine SPR biosens-

ng; as it is a water-soluble major metabolite of morphine, whichan be found in the urine of opioid drug abusers (Brennan et al.,003). M3G (molecular weight 461.46) is also an ideal modelor small molecule immunosensor development, as it is a sta-le, well-defined antigen for antibody recognition. The variableegions of a previously described anti-M3G scFv (Brennan etl., 2003) were re-engineered for the expression of monovalentab fragments. The resulting scFv and Fab fragments were thenompared using ELISA, SDS-PAGE, HPLC and BiacoreTM.he potential benefits of transforming scFv’s into Fab struc-

ures for kinetic analysis and anti-hapten assay development areiscussed in detail.

. Materials and methods

.1. Anti-M3G scFv conversion to chimeric Fab format

Anti-M3G scFv (Brennan et al., 2003) was converted to ahimeric Fab format by the addition of human constant regions,sing recommended PCR assembly methods (Andris-Widhopft al., 2000; Steinberger et al., 2000; Barbas et al., 2001). HumanH1, C� and C� regions were sourced from the pCOMB3X vec-

or series, which contain standard Fab fragment inserts (Andris-idhopf et al., 2000). Fab constructs were synthesized with theouse V� region in association with both C� (Barbas et al.,

001) and C� regions.To facilitate the construction of the C�-associated Fab,

he reverse primer M-H-CL5-B (CGAGGGGGCAGC-TTGGGCTGACCTAGGACAGTCAGTTTGG) was designed

o amplify the mouse V� sequence and to also introduce a human� sequence tail (italicized). The mouse V� PCR product wasel-purified and combined with purified C� product in overlapxtension PCR to create the final � light-chain construct.himeric heavy chain fragments were then combined with the

espective light chains in a final round of overlap extensionCR, before insertion into the phagemid vector pCOMB3X.his vector allows both phage display-based selection andxpression of functional antibody, tagged with haemagluttininHA) and hexahistidine motifs (Andris-Widhopf et al., 2000).

.2. Phage display selection of functional Fab constructs

The ligated Fab-pCOMB3X constructs were introduced into. coli XL-1 Blue (Stratagene) by electroporation (Gene-pulser,io-Rad). Electroporations were performed at 2.5 kV, 25 �F and00 �. Total transformant numbers were estimated on Luriaertani (LB) agar containing 100 �g/ml carbenicillin (Sigma).

he completed Fab-pCOMB3X constructs were propagated in. coli XL1-Blue and expression of phage-Fab was induced byo-infection with M13 helper phage (New England Biolabs).hage propagation and preparation methods were performed as

2 nd Bi

d2

to(Osis(tbtwiXre2

tuiwoasa

2

pEoic2t3rioabds(at

sbgse

2

pwubpram

aEMww3pcPa

2

ace((0o

wfoeaa1twocwtic1

ii

70 S. Townsend et al. / Biosensors a

escribed elsewhere (Andris-Widhopf et al., 2000; Barbas et al.,001).

For the selection of Fab constructs without PCR-based muta-ion, a bio-panning strategy was employed. The first roundf panning was performed using MaxisorpTM immunotubesNunc) coated overnight at 4 ◦C with M3G-ovalbumin (M3G-VA) conjugate (50 �g/ml). After washing three times with

terile filtered phosphate buffered saline, pH 7.4 (PBS), themmunotube was blocked with 3% (w/v) BSA/PBS (bovineerum albumin in PBS) for 1 h at 37 ◦C. Phage preparation250 �l) plus 1 ml of 1% (w/v) BSA/PBS was then added andhe tube incubated for 1 h at room temperature with mixingy rotation. Non-specific phage was removed by washing theube 10 times with PBS/0.1% (v/v) Tween 20 and 10 timesith PBS. Bound phage was eluted with 1 �g/ml M3G-HCl

n PBS for 30 min. This phage was used for reinfection intoL1-blue E. coli cells and propagation (as above). A second

ound of panning was performed in a single microtitre well, withlution using 10 mg/ml trypsin in PBS (Andris-Widhopf et al.,000).

Phage-Fab suspensions from each round of panning were thenested for specific binding to antigen by direct ELISA analysissing the method of Barbas et al. (2001), with the follow-ng modifications: microtitre plate wells (MaxisorpTM, Nunc)ere coated with 100 �l/well M3G-OVA (10 �g/ml in PBS),vernight at 4 ◦C. Phage binding was detected using anti-HAHRP

ntibody (Roche) in conjugation with o-phenylenediamine sub-trate (Sigma-Fast, Sigma). Absorbances were read after 30 mint 450 nm.

.3. Antibody expression conditions

For the soluble expression of Fab fragments in pCOMB3X,lasmid preparations of single Fab clones were introduced into. coli Top10F’ (Invitrogen) by electroporation before platingn LB agar/carbenicillin plates. Single colonies were inoculatednto 10 ml of SB–G–C (Super Broth/2% (w/v) glucose/50 �g/mlarbenicillin), which were incubated overnight at 37 ◦C with20 rpm orbital shaking. Each starter culture was then usedo seed 500 ml amounts of SB–G–C and incubated for 8 h at7 ◦C/220 rpm. Cultures were then centrifuged at 3220 × g ande-suspended in fresh SB–C (no glucose). The cultures werencubated for 1 h at 25 ◦C before induction with 1 mM IPTGvernight at 25 ◦C and 220 rpm. Cultures were then centrifugeds above and cell pellets resuspended in 20 ml of column loadinguffer (50 mM NaH2PO4; pH 8.0, 300 mM NaCl, 5 mM imi-azole). Cell lysate was prepared by sonication and Fab wasemi-purified using immobilized metal affinity chromatographyNi-NTA, Qiagen). Each Fab preparation was then concentratednd buffer-exchanged into PBS by ultrafiltration (Vivaspin, Sar-orious).

Soluble expression of scFv fragments (in pAK400 expres-ion vector) was carried out as above, but with cultures

eing prepared in TB–G–C–S (Terrific broth plus 2% (w/v)lucose, 25 �g/ml chloroamphenicol, 25 �g/ml streptomycinulphate), followed by TB–C–S (no glucose) during proteinxpression.

bcrfi

oelectronics 22 (2006) 268–274

.4. Enzyme-linked immunosorbent assay (ELISA) analyses

Direct ELISA was performed according to standard ELISArotocol for recombinant antibodies (Barbas et al., 2001)ith the following modifications: scFv and Fab were detectedsing anti-HISHRP (Qiagen) and anti-HAHRP secondary anti-odies, respectively. Bound antibody was detected using o-henylenediamine (o-PD) ELISA substrate. Absorbances wereead after 30 min at 450 nm. This method was used to optimizentibody and coating antigen concentrations before the perfor-ance of competition ELISA.Competition ELISAs were performed separately for each

ntibody construct under conditions determined by directLISAs: microtitre plate wells were coated with 100 �l/well3G-OVA (2 �g/ml in PBS), overnight at 4 ◦C. These platesere then washed three times with distilled water and blockedith 300 �l/well PBS/M (PBS/5% (w/v) milk protein) for 1 h at7 ◦C. ScFv or Fab (50 �l) were then added to each well in theresence of serially-diluted (PBS/2% (w/v) milk protein) con-entrations of free analyte (50 �l) and incubated for 1 h at 37 ◦C.lates were then washed 10 times with distilled water and boundntibody detected as described for direct ELISA.

.5. BiacoreTM assay conditions

Analysis was performed using a BiacoreTM 3000 instrumentnd data analysis was performed using BIAevaluation 3.0 (BIA-ore, Uppsala, Sweden). Research grade CM5 sensor chips weremployed in all analyses with Hepes-buffered saline, pH 7.4,HBS) (10 mM Hepes, 150 mM NaCl, 3.4 M EDTA and 0.025%v/v) Tween 20) as running buffer. HBS was filtered using a.2 �m cut-off filter degassed and syringe filtered (0.45 �m cut-ff) immediately before use.

Direct immobilisation of M3G on the sensor chip surfaceas performed as follows: the carboxylmethylated dextran sur-

ace of the sensor chip was activated by mixing equal volumesf 0.1 M NHS (N-hydroxy-succinimide) and 0.4 M EDC (1-thyl-3-(3-dimethylaminopropyl)-carbodiimide hydrochloride)nd injecting the mixture over the sensor chip surface for 8 min atflow-rate of 5 �l/min. The chip surface was then ‘capped’ withM ethylene diamine (pH 8.5) for 10 min (flow-rate 5 �l/min),

o cationise the surface. M3G (100 �g/ml) diluted in EDC/NHSas incubated at room temperature for 3 min, before injectionver the chip surface for 20 min at a flow-rate of 10 �l/min. Toomplete the reaction, unreacted sites on the chip were ‘capped’ith 1 M ethanolamine. The surface was then regenerated five

imes with 40 mM NaOH prior to use. M3G-OVA conjugatemmobilisation was carried out according to standard amineoupling chemistry (Johnsson et al., 1991; O’Shannessey et al.,992).

Antibody samples for BiacoreTM sensor analysis were dilutedn HBS buffer (pH 7.4). Concentration standards of M3G (rang-ng from 318 to 1.25 × 106 pg/ml) were then prepared in HBS

uffer, for use with both scFv and Fab. When using the M3G-oated surface, each concentration standard was incubated sepa-ately with an equal volume of a 1/60 dilution of the Fab (1/120nal dilution) or a 1/25 dilution of the scFv (1/50 final dilu-

nd Bi

ttpcodupfawcacwa

2

t3aTmdcaassH

2

iuawRsr2pflrtsirba1uts

3

amiasuoti(citfirsc

pIfiafncodIusaa2.4 ng/ml (Fig. 1).

In a number of previously described chimeric FAb libraryconstruction methodologies, the light chain V regions (both �and �) of mice (Barbas et al., 2001), rabbits (Rader et al., 2000)

Fig. 1. ELISA inhibtion assay performed on M3G-OVA-coated wells comparingFab ( ) and scFv ( ) binding response versus competing free M3G con-centration. Signal developed using OPD chromogenic substrate and monitored

S. Townsend et al. / Biosensors a

ion) for 30 min at room temperature before being passed overhe chip at a flow-rate of 5 �l/min to examine antibody bindingrofiles. In analyses performed on the M3G-OVA immobilizedhip, Fab was diluted to 1/20 and the scFv to1/8, before carryingut assays as above. Final antibody concentrations were pre-etermined to provide a signal of approximately 300 responsenits (RU) on each respective surface, in the absence of com-eting antigen. M3G and M3G-OVA chip surfaces were bothully regenerated using a single 15 s pulse of 40 mM NaOH atflow-rate of 20 �l/min. A control sample containing no M3Gas included in each run to provide maximal signal for R/R0

alculations (i.e. binding at different free analyte concentrationsre expressed as a proportion of binding in the presence of noompeting analyte). Analyses at each free M3G concentrationere carried out in triplicate on three separate occasions for

ssay variability studies.

.6. scFv multimerisation analyses

To examine the level of multimerisation in scFv prepara-ions, size exclusion HPLC (Phenomenex, BioSep-SEC-S-3000,00 mm × 7.80 mm) was used. All assays were performed atflow-rate of 0.5 ml/min using an injection volume of 50 �l.he mobile phase was PBS. Both scFv and Fab antibody frag-ents were examined at neat dilutions in PBS. HPLC stan-

ards (Sigma) included Bovine serum albumin (BSA); 66 kDa,hicken egg albumin; 45 kDa, carbonic anhydrase; 25 kDa,nd beta lactalbumin; 14.5 kDa. Antibody multimerisation waslso examined by native polyacrylamide gel electrophore-is (Laemmli, 1970). The non-denatured molecular weighttandards used for gel electrophoresis were as described forPLC.

.7. BiacoreTM kinetics analyses

BiacoreTM analysis was performed to determine the bind-ng kinetics of both scFv and Fab. Analyses were performedsing M3G directly coupled to the chip surface, which gener-ted a maximum response signal of 100 RU when saturatedith either scFv or Fab antibody fragments. A Rmax of 100U is optimal for kinetic studies. This provided an appropriateurface for comparative studies and was within the parametersecommended by BiacoreTM for kinetic analysis (Canziani et al.,004). Individual samples consisting of 90 �l of either IMAC-urified scFv or Fab were passed over the chip surface at aow-rate of 30 �l/min. Following analysis, bound antibody wasemoved using 40 mM NaOH until resonance signal returnedo baseline value. The sensorgram obtained from injecting non-pecific antibody fragment in PBS plus 25 mM imidazole dilutedn HBS (mimicking a nickel-purified antibody buffer prepa-ation) was used to normalise the sensorgrams obtained fromoth Fab and scFv binding to M3G. The dilutions for bothntibody fragments used were as follows: 1/10, 1/20, 1/40,

/80, 1/160, 1/320 in HBS. Online reference subtraction wassed throughout and then analysed using BIAevaluation. Massransfer limitations were determined to be negligible (data nothown).

baosd

oelectronics 22 (2006) 268–274 271

. Results and discussion

Electroporation of E. coli XL-1 Blue with the pCOMB3X/nti-M3G C�-Fab construct yielded 3.1 × 107 total transfor-ants, with a vector self-ligation background of <0.1%. Follow-

ng biopanning, very poor signal was observed in phage ELISAgainst M3G-OVA. In subsequent ELISA characterisation ofoluble antibody, the C�-Fab demonstrated high sensitivity whensed at a 1/200 dilution and exhibited a range of detection (IC50f 2.4 ng/ml), which was similar to that of the parent scFv. Whenhis C�-Fab construct was examined using BiacoreTM, no bind-ng was observed to either the M3G surface or to the referenceOVA) surface. The C�-Fab was therefore converted to a C�-Fabonstruct, as we postulated that using a more compatible pair-ng of light chain constant and variable regions might improvehe structural stability, secretion and thus overall activity of thenal Fab product. In the second Fab panning regime, the secondound for the C�-Fab construct demonstrated a strong ELISAignal. No concomitant increase in signal was observed againstontrol antigen (OVA).

When expressed in parallel, both the scFv and the C�-Fabroduced large quantities of high titre, high activity antibody.n direct ELISA, the IMAC-purified and concentrated (to 1 mlnal volume) antibody preparations provided titers of 1/81,000nd 1/20,000 for scFv and Fab, respectively. ELISA conditionsor each antibody were optimized for maximal sensitivity (dataot shown) and for both scFv and Fab, the optimal well coatingoncentration was determined to be 3 �g/ml M3G-OVA. At anptimal dilution of 1/15,000, the scFv was shown to have aetection range for free M3G of 305–312,500 pg/ml, with anC50 value of 30,000 pg/ml (Fig. 1). The anti-M3G C�-Fab wassed at a 1/5000 dilution, therefore, having several times morepecific activity than that observed for the similarly expressednd purified C�-Fab. The C�-Fab was further shown to havedetection range of 152–19,500 pg/ml, with an IC50 value of

y optical density at 450 nm. Data expressed as A/A0 (i.e. signal at different freenalyte concentrations are expressed as a proportion of signal in the presencef no competing analyte). All analyses were performed in triplicate on threeeparate occasions. Standard deviation at all points examined was ≤10% (meaneviation scFv = 4.8%; mean deviation Fab = 3.2%).

2 nd Bioelectronics 22 (2006) 268–274

aaiatrsbltsTtR

BOfowabOfvwOtf4r

cMa(aS

FcMccod

Fig. 3. BiacoreTM inhibition assays performed on an M3G-coated CM5 chipsurface, comparing Fab ( ) and scFv ( ) binding response versus freeM3G concentration. Data expressed as R/R0 (i.e. signal at different free analyteconcentrations are expressed as a proportion of signal in the presence of nocod

rtEutaccrlsfwMhpds

72 S. Townsend et al. / Biosensors a

nd chickens (Andris-Widhopf et al., 2000) were spliced ontosingle human C� region. This is presumably due to the dom-

nance of the V� in the immunoglobulin repertoires of micend rabbits (Barbas et al., 2001; Popkov et al., 2003), meaninghat V� clones are likely to be infrequent in resulting immuneecombinant antibody libraries. However, the observation pre-ented here that Fab light chain function and overall antibodyinding activity were improved by combining the V� region (iso-ated from a murine scFv library) with a C� region, as opposedo C�, implies that the use of a C� scaffold in Fab library con-truction might improve the isolation of lambda-associated Fabs.he observed improved function of the C�-Fab may be due

o increased light chain V-region stability, as demonstrated byothlisberger et al. (2005).

Having established a functional anti-M3G FAb construct,iacoreTM competition assays were developed using both M3G-VA and M3G-coated SPR surfaces, to determine the range of

ree M3G detection of both the Fab and scFv. A similar rangef free M3G was employed as in the ELISA inhibition assay,ith samples being injected over each chip in random order

nd each injection followed by a regeneration step. Both anti-odies displayed negligible non-specific binding to immobilizedVA and the modified CM–dextran surface, negating the need

or pre-incubation steps of antibodies with either OVA or acti-ated CM–dextran. The linear range of detection for the scFvas found to be 39–156 ng/ml (IC50 of 117 ng/ml) on the M3G-VA surface (Fig. 2) and 9.7–78 ng/ml (IC50 of 30 ng/ml) on

he M3G-coated surface (Fig. 3). The linear range of detectionor the Fab was found to be 4.8–39 ng/ml (IC50 19 ng/ml) and.8–78 ng/ml (IC50 14 ng/ml) on M3G-OVA and M3G surfaces,espectively.

The data described above suggest that the Fab fragment isonsistently more sensitive (∼2–6-fold), in measurement of free3G than the scFv fragment in SPR-based assay (Figs. 2 and 3)

nd also in ELISA assay where it is 12.5-fold more sensitive

Fig. 1). The respective sensitivities for Fab and scFv in ELISAnalysis and SPR assay were highly similar, suggesting that anPR assay for M3G can be designed which is as sensitive and

ig. 2. BiacoreTM inhibition assays performed on an M3G-OVA-coated CM5hip surface, comparing Fab ( ) and scFv ( ) binding response versus free3G concentration. Data expressed as R/R0 (i.e. signal at different free analyte

oncentrations are expressed as a proportion of signal in the presence of noompeting analyte). All analyses were performed in triplicate on three separateccasions. Standard deviation at all points examined was <5% of signal (meaneviation scFv = 1.2%; mean deviation Fab = 0.7%).

ufws2cb

eseftdcwa2fatr

ompeting analyte). All analyses were performed in triplicate on three separateccasions. Standard deviation at all points examined was <5% of signal (meaneviation scFv = 1.9%; mean deviation Fab = 2.1%).

eproducible as an optimised plate assay. These data also suggesthat the differences observed between Fab and scFv responses inLISA were not mediated by the different secondary antibodiessed to detect binding. The use of direct M3G conjugation onhe SPR surface appears to be more effective and reliable thann M3G-OVA surface, as it promoted greater linearity in scFvompetition assay response (Figs. 2 and 3). The surface directlyoated with M3G also gave the most sensitive and highly cor-elated recognition of free M3G, by both constructs. Indeed,ogit transformation of the data and subsequent two-tailed t-testhowed that on the surface directly coated with M3G, valuesor Fab and scFv were not significantly different (p > 0.05),hereas, the values for both constructs differed greatly on the3G-OVA surface (p < 0.001). This further suggests that direct

apten-labelled SPR sensor chip surfaces may be more appro-riate for the comparative analysis of the assay sensitivities ofifferent antibody constructs. Interestingly, in both the Fab andcFv assays a lower concentration of antibody was required forse on the directly immobilized M3G chip surface. This sur-ace also exhibited much greater stability after regeneration,ith only 50 regenerations possible for the M3G-OVA conjugate

urface, as previously observed (Brennan et al., 2003), but over00 regenerations possible when using the directly immobilizedhip, suggesting that the directly immobilized M3G surface wille more appropriate for higher throughput assay conditions.

Exactly how the two methods of surface functionalisation,xamined in this study, affect the binding profiles and assayensitivities of differing antibody constructs remains to be fullylucidated. However, it is possible that antigen mobility, sur-ace access and presentation mode may all differ between thewo surfaces. For example, it has been postulated that specificynamic events within the dextran hydrogel may contravene itslassification as a classical ‘brush border’, allowing elasticityithin the dextran layer (Piehler et al., 1999) plus translational

nd rotational mobility of the tethered ligand (Jeppesen et al.,001). Despite these possibilities, we postulate that the main dif-

erence between the two surfaces relies mostly on differences inntigen presentation mode. With multiple M3G moieties beingheoretically present on each individual OVA molecule, a >1:1elationship between antibody and antigen is potentiated, lead-

nd Bioelectronics 22 (2006) 268–274 273

isesc

cohcaeswHhsbemtSmdmaetyMtonswfs

tasbtscbstfac

onlih

Fig. 4. (a) Representative sensogram demonstrating the difference between Fab( ) and scFv ( ) binding response under normal assay conditions describedabove (flow rate 5 �l/min), when examined on the M3G-OVA-coated CM5 chipsurface. On-line reference subtraction was performed against a flow cell coatedwith OVA. An antibody sample buffer mimic, containing no anti-M3G, wasused as a secondary negative control. (b) Kinetics sensograms demonstratingthe difference between Fab ( ) and scFv ( ) binding response whenexamined on the M3G-coated surface (flow rate 30 �l/min). On-line referencesaa

reeaop∼s(psfaptioft1

S. Townsend et al. / Biosensors a

ng to high avidity polyvalent surface attachment in the dimericcFv population. This polyvalent interaction may then be less-ned when the surface is labelled directly with a small moleculeuch as M3G, leading to improved function of the scFv in SPRompetition assay.

While carrying out the comparative ELISA and BiacoreTM

ompetition assays described above, we observed that notnly do the anti-M3G Fab and ‘monomeric’ scFv preparationsave considerably different IC50 values, but they also exhibitonsiderably different dose-response profiles in both ELISAnd BiacoreTM (depending on the sensor surface type). Whenxamined on the M3G-OVA surface, the two antibody con-tructs do not exhibit parallelism in their ranges of response,ith only the Fab giving a clearly linear profile (Fig. 2).owever, their parallelism, linearity and sensitivity becomeighly similar when the same assays are performed on the M3Gurface (Fig. 3). Given that scFv molecules have been shown toe prone to the formation of multimeric intermediates (Griffithst al., 1993; Kortt et al., 1997), while Fabs are obligatelyonomeric (Borrebaeck et al., 1992), we therefore examined

he scFv and Fab preparations by HPLC and non-reducingDS-PAGE to investigate the extent of antibody fragmentultimerisation. When the scFv was examined in HPLC, two

istinct major peaks were observed, corresponding to a majoronomeric scFv peak and a minor dimeric peak, which is in

ccordance with previous findings (Griffiths et al., 1993; Korttt al., 1997; Atwell et al., 1999; Volkel et al., 2001). In contrast,he Fab preparation exhibited only one large peak. ELISA anal-sis of peak fractions from HPLC demonstrated recognition of3G-OVA across both scFv peaks (data not shown), suggesting

he presence of functional scFv dimers, while in Fab analysisnly the major peak exhibited binding to M3G-OVA (dataot shown). In Western blot analysis of antibody preparationseparated by native PAGE, several scFv bands were observedhich migrated both below and parallel to Fab fragments (which

ormed a single band of ∼50 kDa), further confirming that thecFv exists in multiple associated forms (data not shown).

To compare the possible effects of multimerisation (andherefore multivalency) on the BiacoreTM profiles of the scFvnd Fab constructs, kinetic analysis was performed for each con-truct. Kinetic analyses were performed on the directly immo-ilised M3G surface rather than the M3G-OVA surface. Underhe optimised competition assay conditions on the M3G-OVAurface, the Fab exhibited faster association and dissociation rateonstants than the scFv (Fig. 4a). In kinetic analyses the contrastetween the two constructs was even more noticable, with thecFv exhibiting a considerably different dissociation profile tohat of the Fab (Fig. 4b). Although association rates are similaror both antibody constructs, the anti-M3G Fab fragment showssignificantly faster dissociation rate of 4.7 × 10−3 S−1, when

ompared to 1.03 × 10−2 S−1 for the anti-M3G scFv.These observations are of particular interest in the context

f anti-hapten recombinant antibody development. Recombi-

ant antibody technology can lead to the rapid development ofarge panels of antigen-specific antibodies of differing sensitiv-ty. To accelerate the screening of these panels, several groupsave employed kinetic screening of initial clones by SPR, to

ostS

ubtraction was performed against a flow cell, which was activated and cappeds for the M3G-coated cell. An antibody sample buffer mimic, containing nonti-M3G, was used as a secondary negative control.

apidly identify the highest affinity binders (Li et al., 2000; Lut al., 2003; Stacy et al., 2003; Fredericks et al., 2004). How-ver, the purpose of this SPR screening is typically to identifyntibodies for subsequent use in competition assays, which relyn antibody recognition of free molecule in solution. It has beenreviously shown that ‘monomeric’ antibody fragments (scFv,90% monomeric, by HPLC) demonstrate a more rapid dis-

ociation profile than ‘dimeric’ (∼90% dimerized) fragmentsGrant et al., 1999). However, our observations suggest that aredominantly monomeric scFv form can still exhibit a con-iderably slower dissociation rate than a truly monomeric Fabragment. Due to the common occurrence of multimerisation asresult of elevation of antibody concentration and/or sample

H (Arndt et al., 1998), any individual scFv sample is likelyo contain an unknown proportion of dimeric antibody, whichs functionally bivalent and can lead to a profound change inff-rate kinetics profiles. Indeed, it has been shown that dif-erent antibodies with linkers of the same length may dimeriseo different extents (Griffiths et al., 1993; Raag and Whitlow,995; Wu et al., 1996). It is therefore arguable, that the ranking

f scFv molecules by off-rate kinetics on hapten-coated SPRurfaces may be misleading and may give an unreliable estima-ion of binding affinity for free hapten, even under the stringentPR conditions designed for kinetics analyses. If kinetic rank-

2 nd Bi

imrtwmralfoitau

A

IuswI

R

A

A

A

B

B

BCC

F

F

G

G

G

H

HJ

JK

LL

L

M

M

MOO

P

P

P

QRR

RRR

S

S

T

V

74 S. Townsend et al. / Biosensors a

ng of recombinant anti-hapten antibody fragments presupposesonomeric binding interaction, and the ranking is performed to

apidly isolate antibodies which are good for competition assay,hen this method will only be truly reliable for Fab constructs,here monomeric binding status is guaranteed. The influence ofultimeric scFv is starkly illustrated in Fig. 4b, where the off-

ate apparent in the kinetic profiles of the Fab and scFv constructsre strikingly different, leading to a 40-fold difference in calcu-ated off-rate values. This suggested higher affinity of the scFvor M3G is not borne out, however, in subsequent assay devel-pment, as the Fab can actually out-perform the scFv moleculen both plate and SPR assay formats (Figs. 1 and 2). Moreover,he scFv and Fab constructs are found to have highly similarffinity for the free M3G molecule (as judged by IC50 values)nder optimised assay conditions (Fig. 3).

cknowledgements

The authors thank Prof. Carlos Barbas at the Scripps Researchnstitute, for providing the pCOMB3X phage-display vectorssed in this study. The authors also thank Dr. Michael Parkin-on for his assistance in data analysis. Funding for this studyas provided by Enterprise Ireland and Science Foundation

reland.

eferences

ndris-Widhopf, J., Rader, C., Steinberger, P., Fuller, R., Barbas, C.F., 2000. J.Immunol. Meth. 28, 159–181.

rndt, K.M., Muller, K.M., Pluckthun, A., 1998. Biochemistry 37 (37),12918–12926.

twell, J.L., Breheney, K.A., Lawrence, L.J., McCoy, A.J., Kortt, A.A., Hudson,P.J., 1999. Protein Eng. 12 (7), 597–604.

arbas 3rd, C.F., Burton, D.R., Scott, J.K., Silverman, G.J., 2001. Phage Displaya Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold SpringHarbor, NY.

orrebaeck, C.A., Malmborg, A.C., Furebring, C., Michaelsson, A., Ward, S.,Danielsson, L., Ohlin, M., 1992. Biotechnology 10 (6), 697–698.

rennan, J., Dillon, P., O’Kennedy, R., 2003. J. Chromatogr. B 786, 327–342.anziani, G., Klakamp, S., Myszka, D.G., 2004. Anal. Biochem. 325, 301–307.harleton, K.A., Moyle, S., Porter, A.J.R., Harris, W.J., 2000. J. Immunol. 164,

6221–6229.inlay, W.J.J., deVore, N.C., Dobrovolskaia, E.N., Gam, A., Goodyear, C.S.,

Slater, J.E., 2005. Clin. Exp. Aller. 35 (8), 1040–1048.redericks, Z.L., Forte, C., Capuano, I.V., Zhou, H., Vanden Bos, T., Carter, P.,

2004. Protein Eng. Design Sel. 17 (1), 95–106.eorge, A.J., French, R.R., Glennie, M.J., 1995. J. Immunol. Meth. 183 (1),

51–63.

W

Y

oelectronics 22 (2006) 268–274

rant, S.D., Porter, A.J., Harris, W.J., 1999. J. Agric. Food Chem. 47 (1),340–345.

riffiths, A.D., Malmqvist, M., Marks, J.D., Bye, J.M., Embleton, M.J., McCaf-ferty, J., Baier, M., Holliger, K.P., Gorick, B.D., Hughes-Jones, N.C.,Hoogenboom, H.R., Winter, G., 1993. EMBO J. 12 (2), 725–734.

all, J.C., Van Deynze, T.D., Struger, J., Chan, C.H., 1993. J. Environ. Sci.Health 28 (5), 577–598.

ock, B., Eifert, M.S., Kramer, K., 2002. Biosen. Bioelectron. 17, 239–249.eppesen, C., Wong, J.K., Kuhl, T.L., Israelachivilli, J.N., Mullah, N., Zalipsky,

S., Marques, C.M., 2001. Science 293, 465–468.ohnsson, B., Lofas, S., Lindquist, G., 1991. Analyt. Biochem. 198, 268–277.ortt, A.A., Lah, M., Oddie, G.W., Gruen, C.L., Burns, J.E., Pearce, L.A.,

Atwell, J.L., McCoy, A.J., Howlett, G.J., Metzger, D.W., Webster, R.G.,Hudson, P.J., 1997. Protein Eng. 10 (4), 423–433.

aemmli, U.K., 1970. Nature (Lond.) 227, 680–685.i., Y., Cockburn, W., Kilpatrick, J.B., Whitelam, G.C., 2000. Biochem. Bio-

phys. Res. Commun. 268, 398–404.u, D., Shen, J., Vil, M.D., Zhang, H., Jimenez, X., Bohlen, P., Witte, L., Zhu,

Z., 2003. J. Biol. Chem. 278 (44), 43496–43507.cGregor, D.P., Molloy, P.E., Cunningham, C.C., Harris, W.J., 1994. Mol.

Immunol. 312, 219–226.acKenzie, C.R., Hirama, T., Deng, S.J., Bundle, D.R., Narang, S.A., Young,

N.M., 1996. J. Biol. Chem. 271 (3), 1527–1533.aynard, J., Georgiou, G., 2000. Ann. Rev. Biomed. Eng. 2, 339–376.’Shannessy, D.J., Winzor, D.J., 1996. Analyt. Biochem. 236, 275–283.’Shannessey, D.J., Bringham-Burke, M., Peck, K., 1992. Analyt. Biochem.

205, 132–136.iehler, J., Brecht, A., Hehl, K., gauglitz, G., 1999. Colloids Surf. B: Biointerf.

13, 325–336.luckthun, A., 1994. In: Rosenberg, M., Moore, G.P. (Eds.), Handbook of Phar-

macology, 13. Springer–Verlag, Berlin, pp. 269–315.opkov, M., Mage, R.G., Alexander, C.B., Thundivalappil, S., Barbas III, C.F.,

Rader, C., 2003. J. Mol. Biol. 325 (2), 325–335.uinn, J.G., O’Kennedy, R., 2001. Analyt. Biochem. 290, 36–46.aag, R., Whitlow, M., 1995. FASEB J. 9 (1), 73–80.ader, C., Ritter, G., Nathan, S., Elia, M., Gout, I., Jungbluth, A.A., Cohen, L.S.,

Welt, S., Old, L.J., Barbas, C.F., 2000. J. Biol. Chem. 275 (18), 13668–13676.au, D., Kramer, K., Hock, B., 2002. Analyt. Bioanalyt. Chem. 372, 261–267.ich, R.L., Myszka, D.G., 2005. J. Mol. Recogn. 18 (1), 1–39.othlisberger, D., Honegger, A., Pluckthun, A., 2005. J. Mol. Biol. 347,

773–789.tacy, J.E., Kausmally, L., Simonsen, B., Nordgard, S.H., Alsoe, L., Michaelsen,

T.E., Brekke, O.H., 2003. J. Immunol. Meth. 283, 247–259.teinberger, P., Sutton, J.K., Rader, C., Elia, M., Barbas 3rd, C.F., 2000. J. Biol.

Chem. 275 (46), 36073–36078.out, N.L., Yau, K.Y.F., Trevors, J.T., Lee, H., Hall, J.C., 2001. J. Agric. Food

Chem. 49 (8), 3628–3637.olkel, T., Korn, T., Bach, M., Muller, R., Kontermann, R.E., 2001. Protein Eng.

14 (10), 815–823.u, A.M., Chen, W., Raubitschek, A., Williams, L.E., Neumaier, M., Fischer, R.,

Hu, S.Z., Odom-Maryon, T., Wong, J.Y., Shively, J.E., 1996. Immunotech-nology 2 (1), 21–36.

au, K.Y.F., Lee, H., Hall, J.C., 2003. Biotechnol. Adv. 21, 599–637.