Journal of Immunological Methods 319 (2007) 133–143www.elsevier.com/locate/jim

Research paper

Species and epitope specificity of two 65 kDa glutamatedecarboxylase time-resolved fluorometric immunoassays

Mao Rui a, Christiane S. Hampe b, Chen Wang a, Zhidong Ling a, Frans K. Gorus a,Åke Lernmark b, Daniel G. Pipeleers a, Pieter E.M. De Pauw a,⁎

a Diabetes Research Center, Brussels Free University, Brussels, Belgiumb Department of Medicine, University of Washington, Seattle, Washington USA

Received 3 May 2006; received in revised form 14 November 2006; accepted 21 November 2006Available online 14 December 2006

Abstract

The 65 kDa isoform of human glutamate decarboxylase (GAD65) is a major autoantigen in type 1 diabetes (T1D). In thepresent study, we have developed a sensitive sandwich time-resolved fluorescence immunoassay (TRFIA) for the quantification ofGAD65 in cell extracts, cell media and serum. The monoclonal antibody GAD-6 is used to selectively capture GAD65 but not theslightly larger isoform GAD67, and the utilization of different detecting antibodies with distinct GAD65 epitope specificity allowsmodulating the specificity of the assay. To this effect we have biotinylated a recombinant antigen-binding fragment (rFab) withepitope specificity for the N-terminal region of rat and human GAD65 (rFab N-GAD65) and another rFab that selectively binds tothe middle part of human GAD65 (rFab b96.11). In the assay the biotinylated rFabs are recognized by Europium labeledstreptavidin. The obtained time-resolved fluorescence (TRF) is directly proportional to the concentration of GAD65 over a largemeasuring range (0.1 to N100 ng/mL). Based on total error estimation including both bias and imprecision, the lower limit ofquantitation (LLOQ) of GAD65 in cell extracts is 0.33 ng/mL with the N-GAD65 TRFIA, and 0.10 ng/mL with the b96.11 TRFIA,but the latter is suitable for human GAD65 only, whereas the N-GAD65 TRFIA has equal sensitivity with rat and human GAD65.Specificity was further checked with GAD65/67 fusion proteins, confirming that the presence of intact capture as well as detectionepitope on the analyte is a prerequisite for recognition in both assays. We show that the beta cell-specific marker GAD65 can bequantified in pancreatic cell extracts and in serum, allowing studies on discharge during cell death in vitro as well as in vivo.© 2006 Elsevier B.V. All rights reserved.

Keywords: GAD65; Immunoassay; Type 1 diabetes; Pancreatic beta cells

Abbreviations: GAD65, 65 kDa isoform of GAD; T1D, type 1 diabetes; TRFIA, time-resolved fluorescence immunoassay; GAD67, 67 kDaisoform of GAD; rFab, recombinant antigen-binding fragment; LLOQ, lower limit of quantitation; GADA, glutamate decarboxylase auto antibodies;RIA, radio immunoassay; ELISA, enzyme-linked immunosorbent assay; mAb monoclonal antibody; Eu, Europium; DMSO dimethylsulfoxide;BSA, bovine serum albumin; EDTA, ethylenediaminetetraacetic acid; B-rFab, biotinylated rFab; ULOQ, upper limit of quantitation; FACS,Fluorescence-Activated Cell Sorting; SFSTP, Société; Française des Sciences et Techniques Pharmaceutiques; ANOVA, analysis of variance;NCCLS, National Committee for Clinical Laboratory Standards.⁎ Corresponding author. Diabetes Research Center, Vrije Universiteit Brussel, Medical Campus, Building E 013, Laarbeeklaan 103, B-1090

Brussels, Belgium. Tel.: +32 2 477 50 30; fax: +32 2 477 50 47.E-mail address: pdepauw@vub.ac.be (P.E.M. De Pauw).

0022-1759/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.jim.2006.11.007

134 M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

1. Introduction

The 65 kDa isoform of human glutamate decarbox-ylase (GAD, EC 4.1.1.15) is a key autoantigen in type 1diabetes (T1D) (Baekkeskov et al., 1990). GAD65autoantibodies (GADA) precede or accompany clinicalonset, and their presence in circulation is consideredpredictive of overt disease or later insulin needs (Leslieet al., 1999; Gorus and Pipeleers, 2001; Pihoker et al.,2005). GADA are considered to represent a humoralimmune reactivity against the GAD65 protein ofpancreatic beta cells, which carry specificity in termsof its localization outside the nervous system.

In normal beta cells, GAD65 is reversibly anchored tothe membrane of small synaptic-like microvesicleswithin the pancreatic beta cells by means of eitherpalmitoylation in the NH2-terminal domain (Christgauet al., 1992), specific N-terminal sequences (Solimenaet al., 1994) or both (Shi et al., 1994). When isolated betacells are destroyed by the toxin streptozotocin, the GADenzymatic and immunochemical activity was found to bereleased in the medium (Smismans et al., 1996; Haoet al., 1999). The mechanism and kinetics of thisdischarge has not yet been clarified. It is not knownwhether it occurs in other conditions of beta cell deathand whether it is also observed in vivo. In order toinvestigate these questions, a quantitative assay isneeded with adequate sensitivity and specificity. Amethod to measure the enzyme activity has first beendescribed (Okada et al., 1976), followed by directquantification of the GAD65 protein with a competitiveRIA (Hao et al., 1999) or ELISA (Schlosser et al., 1997).The radioimmunoassay is hampered by its sub-optimalsensitivity (e.g. 1–3 ng/mL detection limit in serum), andthe use of a polyclonal antiserum. The ELISA uses twomonoclonal antibodies (mAb) that simultaneously bindto GAD65 with minimal mutual competition, reaching alower detection limit of 0.03 ng/mL in rat brain extract.This ELISA is however potentially limited by the epitopespecificity of the used mAbs, and its narrow dynamicrange. The aim of the present work was to develop asensitive immunoassay for GAD65 using a set ofrecombinant antigen-binding fractions (rFabs) withknown and distinct – including human – epitopespecificities (Madec et al., 1996; Schwartz et al., 1999;Padoa et al., 2003). The assay uses a sandwich immu-noassay format and time-resolved fluorescence detection(TRFIA), a technique securing a broad dynamic range(Hemmila et al., 1984; Lövgren et al., 1985). It offers theadvantage of varying the detecting Fabs in order toanalyze epitopes that become exposed following shed-ding of GAD65 from damaged or dying cells.

Two GAD65 assays are now reported with differentspecies and epitope specificities, each providing a largerange of detection and quantification in human as wellas rat beta cell extracts, culture media and serum.

2. Materials and methods

2.1. Antibodies

The capture mAb GAD-6 was a generous gift fromDr.D. Gottlieb, St Louis, MO, USA. It exclusively binds tothe 65 kDa isoform of GAD (Gottlieb et al., 1986) and inparticular to the C-terminal region of the protein (Ujiharaet al., 1994), where rat and human GAD65 have 100%similarity (Butler et al., 1993). Two recombinant antigen-binding fractions were chosen as detecting antibodies.The rFab b96.11 with middle part epitope specificity forhuman GAD65 was cloned (Padoa et al., 2003) from thehuman monoclonal b96.11 antibody kindly donated byDr. Banga, London, UK (Tremble et al., 1997; Schwartzet al., 1999) and the rFab N-GAD65 (rFab144) wasderived from monoclonal antibody N-GAD65 withepitope specificity for the 4–22 amino acid N-terminalregion of human and rat GAD65 (Hampe et al., 2001).K14D10 is a monoclonal antibody against another betacell-specific antigen (Ziegler et al., 1992) from which therFab was recently cloned (Hampe et al., 2005).

2.2. Glutamate decarboxylase reference materials

The recombinant human (rh) GAD65 expressed inSpodoptera Frugipera (sf9) cells was from DiamydDiagnostics AB (Stockholm, Sweden) and served ascalibration standard in both versions of GAD65 TRFIAand for spiking in the recovery experiments. The 35S-rhGAD65, 35S-rhGAD67 and fusion proteins (Hampeet al., 2000) used in the specificity tests were in-vitrosynthesized by transcription/translation of their copyDNA using the L4600 TnT-SP6 reticulocyte system(Promega Benelux B.V., Leiden, The Netherlands) asearlier described (Karlsen et al., 1991; Grubin et al.,1994). Related protein sequences are the N-terminal(N), middle (M) and C-terminal (C) regions of GAD65,and are coded with capitals for GAD65 (NMC), andwith small fonts for GAD67 (nmc) (Hampe et al., 2000).

2.3. Biotinylation

The rFabs were biotinylated in order to be detectablewith the time-resolved fluorescence of Europium (Eu)labeled streptavidin (Suonpaa et al., 1992.) The biotinyla-tion method starting from N-hydroxy-succinimido-biotin

135M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

(NHS-Biotin from Pierce, Rockford, USA) (Bayer andWilchek, 1990) was adapted and optimized for rFabs. Inshort, the rFabwas first conditioned in a Centricon 10 tube(Amicon, Beverly, MA, USA) to a minimal concentrationof 2 mg/mL in 100 mM NaHCO3, pH 8.8. The NHS-Biotin was then added in a proportion of 200 μg/mg rFabstarting from a 10 mg/ml Biotin-NHS solution in DMSO,and the reactionmixture was slightly stirred for 2 h at 4 °Cbefore washing three times with storage buffer (10 mMTris, 150 mMNaCl, pH 8.2). Molecular sieving was thenperformed on a PD-10 column (Amersham BiosciencesAB,Uppsala, Sweden), and the protein peak at 280 nmwascollected and filtered through a sterile 0.2μmMinisart filter(from Sartorius AG, Goettingen, Germany). After deter-mination of the concentration (OD280=1.4 mL cm−1

mg−1), the biotinylated rFabs (B-rFabs)were stored in 50%glycerol at −20 °C. Biotinylation efficiency was checkedby incubating 50 nM of the obtained B-rFab during 1 h in astreptavidin-coated microtiterplate (Perkin Elmer N.V.,Zaventem, Belgium), followed by washing and incubationwith 1 μM of Eu-streptavidin. After a second wash and5 min incubation with enhancement solution, the dissoci-ation enhanced fluorescence counts were compared withthe counts obtained from 50 nM free Eu-streptavidin in thesame conditions, and the ratio between the two measure-ments was considered as equal to n −1 with n=averagenumber of biotins per B-rFab molecule. Binding affinityand specificity of the B-rFabs towards 35S-rhGAD65 waschecked with a radiobinding experiment as describedbefore (Hampe et al., 2001.) Unbiotinylated and biotiny-lated rFab-35S-rhGAD65 complexes were separated byprecipitation of the latter with streptavidin-coated beads(Pierce Biotechnology Inc., Rockford, IL).

2.4. Preparation, culture and treatments of pancreaticcells

Rat and human islet beta cells were prepared andpurified as previously described (Pipeleers et al., 1985).Samples with fixed cell numbers were re-suspended(maximum 250 K cells/mL) in TRFIA assay buffer (seeSection 2.5), and sonicated during 1 min on ice. Theextracts were analyzed for their GAD65 content using thetwo variants of GAD65 TRFIA, and for their insulincontent using a rat insulin RIA fromLinco (Nuclilab, Ede,The Netherlands).

2.5. TRFIA

Nunc Maxisorp plates (VWR International BVBA,Leuven, Belgium) were coated with 200 μL per well of a1-μg/mL mAb GAD-6 solution in coating buffer

(100 mM NaHCO3, pH 9.8), covered with Parafilm(Pechiney Plastic packaging, Chicago, IL) and stored at4 °C during at least three days and maximum threeweeks before use. The incubated microtiterplates couldbe stored “dry” for a longer period (3 months) whentreated by washing (two washing cycles on a DELFIAplatewasher (prod. number 1296–026, Perkin Elmer),with a 25 times dilution of “wash concentrate” (prod.number 1244–114, Perkin Elmer), followed by over-night incubation with 6% sorbitol, 0.1% BSA in 50 mMphosphate buffer, pH 7.0. After suction of the sorbitolsolution, the plates were dried at 37 °C for 1 h and thenkept in a dessicator. The day before the assay, a platewas washed twice with “wash” solution, and overnightincubated at 4 °C with 2 g/dL BSA in coating buffer.Just before sample loading, four other washing cycleswere performed. The rhGAD65 (Diamyd DiagnosticsAB, Stockholm, Sweden) was used to prepare cali-brators diluted in TRFIA assay buffer (Tris 50 mM,0.9 g/dL NaCl, 0.05 g/dL bovine γ-globulin, 0.5 g/dLBSA, 20 μM EDTA, 0.05%(v/v) Tween 20, 0.05 g/dLNaN3, pH 7.75) or in GAD65 negative human serum (aserum pool from GADA negative, nondiabetic and ap-parently healthy adult subjects). A 100-μL volume ofassay buffer with 0.5% (v/v) Triton was pipetted in allmicrotiterplate wells, followed by 25 μL in duplicate ofsamples or rhGAD65 calibrator solutions, and themicrotiterplate was incubated for 4 h at roomtemperature on a plate shaker (Perkin Elmer). After a4-cycles repeated washing, 125 μL of 0.50 μg/mL B-rFab (Biotinylated rFab N-GAD65 or rFab b96.11) inassay buffer was incubated overnight at 4 °C in eachwell. After another 4-cycle washing, 125 μL of Eu-labeled streptavidin solution (0.12 μg/mL in assaybuffer) was added to each well and the plate wasagitated for 1 h on the plate shaker at room temperature.After another four times washing, 200 μL of DELFIAenhancement solution (Perkin Elmer) was added, theplate was incubated during 5 min on a plate shaker atroom temperature, and the time resolved fluorescencewas measured on a VICTOR² reader (Perkin Elmer).Curve fitting and calculation of unknowns was per-formed with the “spline smoothed” algorithm of theMulticalc software (Wallac Oy, Turku, Finland).

2.6. Validation procedure

Validation was performed following the SFSTP pro-posals (Hubert et al., 2004). Linearity and parallelismwere tested by diluting extracts of FACS-sorted rat andhuman beta cell preparations in assay buffer. For linearity,the relative error % RE=[(observed / nominal) −1]× 100

Fig. 1. Binding affinity for 35S-rhGAD65 of unlabeled mAbs (opensymbols) and corresponding B-rFabs (solid symbols) directed againstthe N-terminal part (panel A, diamonds) or against the middle part(panel B, squares) of GAD65. Physical separation between bound andfree 35S-GAD65 was performed by precipitation of the formedcomplexes with protein G Sepharose for the mAbs and withstreptavidin-coated beads for the B-rFabs. The obtained dissociationconstants were all between Kd=0.4×10−12 M and Kd =0.8×10− 12 M.Data represent means (range) of two measurements.

Fig. 2. Competition assay between recombinant antigen-bindingfractions (Fabs) and their biotinylated counterparts (B-Fabs) forbinding to 35S-rhGAD65. The rFab directed to the N-terminal part ofhuman GAD65 (rFab N-GAD65) displaced its biotinylated form (B-rFab N-GAD65) (diamonds), and so did the middle epitope specificrFab b96.11 with its biotinylated form (B-rFab b96.11) (squares),whereas no cross-displacement between the two rFabs was observed.No competition occurred with rFab K14D10, an antigen-bindingfragment against a beta-cell-specific antigen that is not GAD65 (blackcircles). Streptavidin coated beads were used for the specificprecipitation of B-rFabs. Data represent means (range) of twomeasurements.

136 M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

was determined at each dilution level. The GAD65concentration found in undiluted extract is the nominalvalue. Parallelism expresses the agreement among theresults obtained at the different dilution levels of a sampleandwas checked by calculating the%CV=(SD /mean)×100. The applied acceptance criteria for immunoassaysare % REb20, and % RSDb20 (Findlay et al., 2000).

The detection limit in a given matrix is estimatedfrom the mean TRF signal plus three standard deviationsof 20 independent measurements, and is based on astandard curve with rhGAD65 calibrator diluted in the

same matrix. Tested matrices are extraction buffer and ahuman serum from a healthy control subject that isGADA negative based on a radio binding assay (Grubinet al., 1994), containing no GAD65 detectable with RIA,and giving an average TRF signal that is less than a 1.2magnitude of the extraction buffer signal (GADnegative serum). The functional sensitivity, defined asthe lowest GAD65 concentration that can be measuredin a given matrix with 95% confidence of a %CV ≤20,was estimated for both assay variants, in assay bufferand GAD negative serum. For this purpose, repeated(n=6 independent assays) measurements were per-formed at different concentrations, resulting in a meanand a standard deviation (SD) for each concentration,with % CV=100×SD / mean. The concentration wherethe %CV equals 20% is graphically determined, and toobtain a 95% confidence, 2SD, i.e. 40% of the foundconcentration is added to obtain the functional sensitiv-ity (Fig. 5). Determination of the Lower and UpperLimits of Quantitation (LLOQ and ULOQ respectively)was based on the estimation of the total error includingboth bias and accuracy (Findlay et al., 2000). The totalerror was estimated at eleven concentration pointsbetween 0.1 and 100 μg/L of spiked rhGAD65. Thetest samples were prepared all at the same time, and keptin aliquots at −80 °C. Each point was analyzed four

Fig. 3. TRFIA calibration curves of rhGAD65 performed with twodifferent detecting antibodies, B-rFab N-GAD65 (♦) and B-rFabb96.11 (■). Error bars are standard deviations (±SD) for triplicateswithin the same assay.

Fig. 4. Dilution linearity and parallelism test on extracts of FACS-sorted β cell preparations in assay buffer; (A) with N-GAD65 TRFIAon 3000 K human beta cells/mL (●) and 100 K rat beta cells/mL (▲),(B) with b96.11 TRFIA on 3000 K/mL human cells (■). For linearity,the relative error (% RE=[(observed / nominal) −1]×100) wascalculated and amounted b10% at each dilution level. For parallelism,the relative standard deviation among dilution levels (%CV=[SD /mean]×100) was calculated and amounted 14.5, 8.2 and 9.3 for therespective cell extracts. The applied acceptance criteria are % REb20,and %CVb20 (Findlay et al., 2000).

137M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

times in duplicate within a TRFIA, and this wasrepeated in six independent assays on different days.The bias is defined as the difference between the overallmean for each concentration and the value expectedfrom the dilution of the rhGAD65 for that point.Analysis of variance (ANOVA) was performed, and asspecifically suggested for immunoassays, the Sattert-waite's formula to calculate the degrees of freedom forthe intermediate precision estimate was applied (Findlayet al., 2000). The 95% confidence limits of total errorbased on this ANOVA may not exceed 25% deviationfrom the nominal value, and the concentration wherethis happens is defined as the limit of quantitation.

For assay comparison, a suspension of 106 isolatedrat or human islet cells in 200 μL assay buffer wassonicated and centrifuged, and the recovery of GAD65with TRFIA towards RIA (Hao et al., 1999) wasdetermined. TRFIA is performed after 100 times finalsample dilution in assay buffer, RIA on a five timesdilution in RIA assay buffer. A spiking recovery test wasperformed by repeating this procedure with 60 pgrhGAD65 per 1000 cells added to the cell suspensionbefore sonication. Species specificity was determined bycomparing the recoveries of rat and human GAD65 forboth assay variants. The specificity was further checkedon the molecular level by testing recoveries of therecombinant GAD65/GAD67 fusion proteins as de-scribed in Section 2.2. Interference testing for haemo-lysis, icterus, lipemia, rheumatoid factor and GADAwasperformed according to the EP7-P protocol of theNational Committee for Clinical Laboratory Standards(NCCLS, 1986).

2.7. Statistics

For overall comparison between unrelated groups,Kruskall–Wallis test is performed, except for compar-ison between two unrelated groups, where Mann–Whitney U-test is used. A p-valueb0.05 is consideredsignificant for both tests.

Fig. 5. Between-assay imprecision (%CV) over six independent assaysat different concentrations of rhGAD65 spiked in extraction buffer,

138 M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

3. Results

3.1. Properties of the biotinylated rFabs

Based on testing with streptavidin-coated microtiter-plates, the biotinylation degree was estimated to average6.2 and 6.4 molecules of biotin per rFab molecule for B-rFab N-GAD65 and B-rFab b96.11 respectively (notshown). Radiobinding experiments indicated similar35S-rhGAD65 binding affinity of biotinylated antigen-binding fragments B-rFab N-GAD65 and B-rFab b96.11as compared to their non- biotinylated whole mAbs(Fig. 1). The unbiotinylated rFabs could compete withtheir biotinylated analogs for tracer binding (Fig. 2).Only in the case of B-rFab N-GAD65, the curves inFigs. 1A and 2 show a slight plateau at 60–80% bound35S-GAD65, indicating two-point binding kinetics. TherFab K14D10 directed to a non-relevant beta cell-speci-fic antigen failed to compete, even at high concentrations(Fig. 2). Hence, the biotinylation was considered tominimally affect the binding affinity and specificity ofthe rFabs.

3.2. Standard curves and linearity

Typical standard curves are shown in Fig. 3. Theobtained fluorescence was directly proportional to theGAD65 concentration up to N100 ng/mL, and theb96.11 TRFIA had a higher sensitivity than the N-GAD65 TRFIA as five times less GAD65 was requiredto achieve the same fluorescence signal in the b96.11TRFIA as compared to the N-GAD65 TRFIA. Gooddilution linearity and parallelism were obtained in cellextracts of sorted pancreatic cells for both assays,

Table 1Estimation of detection limits and quantification levels in serum a andextraction buffer (in brackets)

GAD65 [ng/mL] N-GAD65 TRFIA b96.11 TRFIA

Detection limit b 0.28 (0.30) 0.06 (0.12)Functional sensitivity c 0.31 (0.40) 0.10 (0.14)LLOQd 0.33 (0.45) 0.10 (0.15)ULOQd N100 (N100) N100 (N100)a Results are expressed in ng/mL apparent GAD65 obtained by

analyzing 25 μL extraction buffer or GADA-negative human serum,contains undetectable GAD65 with RIA, and has a mean TRF signalthat is less than 1.2 times the one of extraction buffer (in brackets). Allresults are based on standard curves with rhGAD65 calibrator dilutedin the corresponding medium.b The detection limit was estimated from the mean TRF signal+3SD

(n=20).c For the graphical estimation of the functional sensitivity, see Fig. 5.d LLOQ=lower limit of quantitation, ULOQ=upper limit of

quantitation, see Fig. 6 for the graphical estimation.

estimated for the N-GAD65 TRFIA (A) and the b96.11 TRFIA (B).Functional sensitivities were determined as the upper 95% confidencelimit (+ 2SD) of the concentration corresponding to a CVof 20%. Theobtained functional sensitivities amounted to 0.31 μg/L and 0.10 μg/Lrespectively.

because the error relative to the nominal value (RE) wasb10% at all dilution points, and the % CV amongdilution levels was b15% in all cases (Fig. 4).

3.3. Analytical sensitivity and imprecision

Table 1 shows the detection limits for the two assaysdetermined by analyzing 25 μL extraction buffer orGAD65 negative human serum as defined in Section2.6. Five fold lower detection limits can be set when theassay is performed in extraction buffer since this allowsthe use of 125 μL samples without a change in

Fig. 6. Determination of the GAD65-TRFIA limits of quantitationbased on total error estimation, considering both accuracy andprecision. The mean bias (%RE, ●) and the upper (■) and lower(▲) 95% confidence limits based on ANOVA are plotted versus thenominal sample concentration. For immunoassays, acceptable totalerror is less than ±25%, and the concentration where this percentage isexceeded is considered as limit of quantitation (Findlay et al., 2000).The Lower limit of quantitation (LLOQ) for rhGAD65 in extractionbuffer is 0.33 ng/mL for the N-GAD65 TRFIA (A), and 0.10 ng/mLfor the b96.11 TRFIA (B). Upper limits are higher than 100 ng/mL.

Table 2GAD65 in beta cell extracts measured with b96.11 and N-GAD65TRFIA a

TRFIA Human (40–50%) Rat (50–70%)

b96.11 N-GAD65 b96.11 N-GAD65

Native GAD65Content (pg/103

beta cells)59.9±7.3 48.5±5.8 6.1±0.8 70.2±4.6

Recovery (%) b 110.7±13.5 89.8±10.7 8.2±1.7 94.4±6.2Spiked rhGAD65

Recovery (%) c 106.3±4.1 86.0±4.6 97.5±5.9 88.4±4.0a Results are mean (range) of two independent experiments. Human

cells are cryopreserved and re-cultured for three days.b Percent of amount measured by TRFIA with rhGAD65 as

calibrator, versus that measured by RIA.c Expressed as percentage of the 60 pg rhGAD65 per 103 cells

retrieved when spiked in the cell suspension before sonication. Thelisted concentration is obtained after subtracting the basal value.

Table 3Recoveries with b96.11 TRFIA and N-GAD65 TRFIA of recombinanthuman GAD65/ GAD67 fusion proteins a

b96.11 TRFIA N-GAD65 TRFIA

NMC 100.0±3.7 100.0±2.8nmc b0.1 b0.28nMc - b0.1 b0.28nMC 22.3±1.1 b0.28NMc 1.6±0.2 4.2±0.3

a The nomenclature is based on rhGAD65=“NMC” and rhGAD67=“nmc”, with “N”=N-terminal, “M”=middle part, “C”=C-terminalhuman GAD65 sequence. The recovery is expressed as mean percen-tage of the average signal obtained with NMC±SD in three indepen-dent experiments.

139M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

background. The inter-assay error was determined atdifferent concentrations of rhGAD65 in extractionbuffer, and in the 0.3–100-ng/mL range the CVs were≤15% (Fig. 5). The functional sensitivity was graph-ically evaluated at 0.31 ng/mL for the N-GAD65 TRFIA(Fig. 5A) and at 0.10 ng/mL for the b96.11 TRFIA (Fig.5B). The experiment was repeated with rhGAD65spiked in GAD negative human serum (not graphicallyshown), and the results in extraction buffer and serumare summarized in Table 1.

3.4. Total error estimation

The total error including both the bias towards thenominal concentration of rhGAD65 and the overallimprecision was determined by repeated measurementsat different rhGAD65 concentrations with both assays.The graphical estimation of the lower level ofquantitation (LLOQ) based on these experiments inextraction buffer is shown in Fig. 6. At the expense of asmaller dynamic range, this LLOQ can be improvedwhen a five times larger sample volume is used. Table 1summarizes the quantification limits for the lowersample volumes.

As expected, the functional sensitivity and the LLOQestimations are more stringent than the detection limitestimation. The data in Fig. 6 further show that for bothassays, GAD65 concentrations up to 100 ng/mL can bemeasured with an acceptable total error.

Table 4GAD65 (N-GAD65 TRFIA) and insulin content in isolated islet cellpreparations (mean over n=4,±SE)

% betacells

GAD65ng/106cells

Insulinμg/106cells

GAD/Insulinng/μg a

Rat — freshly isolatedIslet beta cells N90 118±5 47±4 2.6±0.2Islet non-betacells

b10 8±1 b 1.6±0.4 b 4.3±0.2

Human — culturedIslet cells 25–39 8±3 c 3.6±0.7 c 2.1±0.5a Kruskall–Wallis test revealed no significant difference between the

three cell preparations in GAD/insulin ratio ( p=0.076).b pb0.01.c p=b0.05, versus rat islet beta cells with Mann–Whitney U test.

140 M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

3.5. Recoveries and specificities

When N-GAD65 TRFIA was used to determine theGAD65 content of rat and human beta cells, similarvalues were obtained as with RIA, corresponding torespectively 90 and 94% of the RIA data. This was notthe case for the b96.11 TRFIA where similar values forhuman (111% of RIA data) but not for rat beta cells(8.2% of RIA data) (Table 2) were obtained. WhenrhGAD65 was spiked in these extracts, good recoverieswere found for both assays (Table 2).

In order to determine assay epitope specificities,recoveries of in-vitro synthesized GAD65/67 fusionproteins were measured with both assays (Table 3). Thereactivity of the in-vitro synthesized intact rhGAD65,referred to as “NMC”, is considered as 100%. Asexpected, intact rhGAD67 referred to as “nmc” did notresult in any signal, demonstrating the specificity ofboth assays for the 65 kDa isoform of GAD. Fusionprotein nMC was recognized in the b96.11 TRFIA dueto the presence of both the b96.11 and the GAD6epitope, however the relative reduction in recovery(22.3% compared to NMC) was probably due toconformational differences. As expected, the N-GAD65 TRFIA did not react with this fusion protein.The recoveries for NMc were low (b1.6% and b4%respectively), and the nMc fusion protein was notdetectable (Table 3). Absence of one of the epitopesrequired by the mAbs used in the assay thus leads torecoveries of less than 5%, and absence of both specificepitopes on a fusion protein leads to less than 0.1%recovery. The presence of haemolysis (up to 1000 mg/dL), Intralipid™ (up to 1000 mg/dL), bilirubin (up to30 mg/dL) or GADA did not interfere (b10% change)when a human serum spiked with 3 ng/dL rhGAD65was assayed for rhGAD65 content. With 60 mg/dLbilirubin, an overestimation of 17% was obtained. On a

serum with rheumatoid factor RA=763 IU/mL, arecovery of 101.6 was found.

3.6. Application

To test GAD65 as an intracellular marker of betacells, we compared GAD65 and insulin in extracts ofFACS-sorted human and rat pancreatic cell preparations.Cellular content of GAD65 and insulin and thepercentage of insulin positive cells as determined byimmunocytochemistry vary in parallel (Table 4).

4. Discussion

This study reports on the development of a novelsensitive GAD65 TRFIA. It is based on a sandwich-typeformat where the capture antibody is specific for the65 kDa isoform of GAD, and further epitope specificitycan be modulated by selecting the rFab that serves asdetecting antibody. The chosen rFab is biotinylated inorder to be recognized by Europium labeled streptavi-din, of which the time-resolved fluorescence ismeasured.

In the case of GAD65, a series of rFabs (Padoa et al.,2003) are available, each with specific affinity for adifferent epitope of the protein (Madec et al., 1996;Schwartz et al., 1999). As a model for future screeningfocusing on a larger number of epitope specificities, twoof these rFabs were tested for use as detecting antibodyin a TRFIA. The rFab N-GAD65 binds specifically tothe N-terminal amino acids 4–22 of human and ratGAD65 (Hampe et al., 2001; Padoa et al., 2003),whereas the rFab b96–11 selectively recognizes aminoacids 308–365 of the middle part of human GAD65(Tremble et al., 1997; Schwartz et al., 1999). Interest-ingly, the latter rFab competes with T1D related GADAsfor binding to GAD65 (Padoa et al., 2003; Schlosseret al., 2005). Both rFabs were biotinylated with aminimal loss of specific affinity. The observed plateausin the case of B-rFab N-GAD65 may be caused by sterichindrance of biotin at the level of the epitope bindingsite, but we did not observe any consequence of thisphenomenon for the developed assay. Indeed, the B-rFabs function well as detecting antibodies in GAD65TRFIA with overnight incubation. The N-GAD65TRFIA is suitable for measuring rat and humanGAD65, whereas the b96.11 TRFIA can only be usedfor human GAD65 and reacts only poorly (7.4% com-pared to human) with rat GAD65. The difference inspecificity between the two assay variants was furtherdemonstrated with their different reactivities towardsGAD65/67 fusion proteins.

141M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

The assayswere fully validated and provide advantagescompared to previously published techniques. First, thelack of an Fc part on the rFabs avoids unwantedinconveniences like steric hindrance or interferences byheterophyle antibodies, complement or rheumatoid factors(Weber et al., 1990; Kricka, 1999). Second, Fabs areproduced by cloning from hybridoma cultures (Winter andMilstein, 1991) or by isolation from large naïve phagedisplay libraries, bypassing immunization of animals(Winter et al., 1994) and can be modified by site directedrandom mutagenesis (Saviranta et al., 1998) or by geneticengineering (Knappik and Pluckthun, 1994; Lindner et al.,1997), leading to more completely controlled essentialcharacteristics, and less expensive and time-consumingproductionwith high volumetric yields (Carter et al., 1992;Horn et al., 1996; Eriksson et al., 2000). Third, by varyingthe nature of the detection rFab, the epitope specificity ofthe assay can be modulated, hence allowing thedevelopment of species-specific assays that may be ofinterest for xenotransplantation experiments, and facilitat-ing in vitro and in vivo studies on the nature of GAD65epitopes that are detectable after release from damaged ordying beta cells. Indeed, GAD65might be presented to theimmune system under different molecular forms (Schlos-ser et al., 2005). Characterization and quantification ofvarious extra cellular or circulating GAD65 speciesincluding various conformations or fragments might thuslead to a better insight in the presentation of GAD65 as anautoantigen. For instance, we have shown that the N-GAD65 TRFIAwill only work when both the N-terminaland the C-terminal ends of GAD65 are intact, and thusexclusively when the whole protein sequence is present,whereas the b96.11 TRFIA can also detect a GAD65fragment that lacks the N-terminal end of the protein.Fourth, the GAD65 TRFIA is more sensitive than RIA(detection limit of 1–3 ng/mL in serum, Hao et al., 1999),which was in turn more sensitive than an enzymatic assay(Okada et al., 1976). A GAD65-specific ELISA methodhas been reported to achieve a detection limit as low as0.03 ng/mL (Schlosser et al., 1997). However, it is difficultto compare the sensitivities of the latter assay with oursbecause the earlier study determined their detection limit inGAD-free brain matrix instead of in pancreatic tissue orcell extracts, and used less stringent criteria, consideringthe detection limit only and not the functional sensitivity orthe LLOQ. The sensitivity of the TRFIA assay allowsstudies on isolated beta cells that are often restrictedbecause of the relatively low cell numbers that can beprepared for this cell type. With the b96.11 TRFIA,samples containing 750 rat or 1000 human beta cells allowduplicate measurement of cellular GAD65 content. This is100 fold less than for RIA assay (Hao et al., 1999).Finally,

thanks to the utilization of time-resolved fluorescence, theTRFIA achieves a much broader measuring range thanobtained with RIA or ELISA, hence avoiding the need tofrequently repeat out of range measurements. In addition,over most of this range a good interassay imprecision wasobserved (b10% CV).

A disadvantage of the TRFIA is the risk for non-specific binding of streptavidin, as observed with ELISAin other organs than pancreas (Schlosser et al., 1997).Direct labeling of the Fab with Europium is possible, butthis would be on the detriment of the sensitivity becauseremoving the biotin/streptavidin intermediate results in alower amplification of the signal.

At the application level, we show that the TRFIA can beused to quantitatively measure GAD65 in extracts ofhuman aswell as rat beta cells, and this offers the possibilityto study discharge during cell death in vitro. Measurementin serum opens possibilities for studies in vivo.

Acknowledgements

We thank Dr. I.Weets of the Belgian Diabetes Registry(BDR) for providing human sera, and P. Goubert for histechnical assistance in producing rhGAD65. The beta cellbank of the JDRF center for beta cell therapy provided thehuman pancreatic tissues and cells, and FACS-sortedhuman and rat cell preparations were kindly provided bythe Functional Cytomics facility of the Diabetes ResearchCenter (DRC).

The Eli Lilly and Company provided rat and humaninsulin standards. This study was supported by grantsfrom the Belgian Science Policy (Interuniversity Attrac-tion Pole P5/17), Juvenile Diabetes Research Fund(Grant no. 4–2001–434), the Belgian Fund for ScientificResearch (FWO, Grants G0375.00 and G.0517.04) andthe National Institutes of Health (grants DK26190 andDK53004). The Flemish Community Commission(VGC) funded the work of Dr. P. De Pauw and theAmerican Diabetes Association supported the work ofDr. C.S. Hampe.

Reference

Baekkeskov, S., Aanstoot, H.J., Christgau, S., Reetz, A., Solimena,M., Cascalho, M., Folli, F., Richter-Olesen, H., De Camilli, P.,Camilli, P.D., 1990. Identification of the 64 K autoantigen ininsulin-dependent diabetes as the GABA-synthesizing enzymeglutamic acid decarboxylase. Nature 347, 151.

Bayer, E.A., Wilchek, M., 1990. Protein biotinylation. MethodsEnzymol. 184, 138.

Butler, M.H., Solimena, M., Dirkx Jr., R., Hayday, A., De Camilli, P.,1993. Identification of a dominant epitope of glutamic aciddecarboxylase (GAD-65) recognized by autoantibodies in stiff-man syndrome. J. Exp. Med. 178, 2097.

142 M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

Carter, P., Kelley, R.F., Rodrigues, M.L., Snedecor, B., Covarrubias,M., Velligan, M.D., Wong, W.L., Rowland, A.M., Kotts, C.E.,Carver, M.E., 1992. High level Escherichia coli expression andproduction of a bivalent humanized antibody fragment. Biotech-nology (N. Y.) 10, 163.

Christgau, S., Aanstoot, H.J., Schierbeck, H., Begley, K., Tullin, S.,Hejnaes, K., Baekkeskov, S., 1992. Membrane anchoring of theautoantigen GAD65 to microvesicles in pancreatic beta-cells bypalmitoylation in the NH2-terminal domain. J. Cell Biol. 118, 309.

Eriksson, S., Vehniainen, M., Jansen, T., Meretoja, V., Saviranta, P.,Pettersson, K., Lovgren, T., 2000. Dual-label time-resolvedimmunofluorometric assay of free and total prostate-specificantigen based on recombinant Fab fragments. Clin. Chem. 46, 658.

Findlay, J.W., Smith, W.C., Lee, J.W., Nordblom, G.D., Das, I.,DeSilva, B.S., Khan, M.N., Bowsher, R.R., 2000. Validation ofimmunoassays for bioanalysis: a pharmaceutical industry perspec-tive. J. Pharm. Biomed. Anal. 21, 1249.

Gorus, F.K., Pipeleers, D.G., 2001. Prospects for predicting andstopping the development of type 1 of diabetes. Best. Pract. Res.Clin. Endocrinol. Metab. 15, 371.

Gottlieb, D.I., Chang, Y.C., Schwob, J.E., 1986. Monoclonal antibodies toglutamic acid decarboxylase. Proc. Natl. Acad. Sci. U. S. A. 83, 8808.

Grubin, C.E., Daniels, T., Toivola, B., Landin-Olsson, M., Hagopian,W.A., Li, L., Karlsen, A.E., Boel, E., Michelsen, B., Lernmark, A.,1994. A novel radioligand binding assay to determine diagnosticaccuracy of isoform-specific glutamic acid decarboxylase anti-bodies in childhood IDDM. Diabetologia 37, 344.

Hampe, C.S., Hammerle, L.P., Bekris, L., Ortqvist, E., Kockum, I.,Rolandsson, O., Landin-Olsson, M., Torn, C., Persson, B.,Lernmark, A., 2000. Recognition of glutamic acid decarboxylase(GAD) by autoantibodies from different GAD antibody-positivephenotypes. J. Clin. Endocrinol. Metab. 85, 4671.

Hampe, C.S., Lundgren, P., Daniels, T.L., Hammerle, L.P., Marcovina,S.M., Lernmark, A., 2001. A novel monoclonal antibody specificfor the N-terminal end of GAD65. J. Neuroimmunol. 113, 63.

Hampe, C.S., Wallen, A.R., Schlosser, M., Ziegler, M., Sweet, I.R.,2005. Quantitative evaluation of a monoclonal antibody and itsfragment as potential markers for pancreatic beta cell mass. Exp.Clin. Endocrinol. Diabetes 113, 381.

Hao, W., Daniels, T., Pipeleers, D.G., Smismans, A., Reijonen, H.,Nepom, G.T., Lernmark, A., 1999. Radioimmunoassay forglutamic acid decarboxylase-65. Diabetes Technol. Ther. 1, 13.

Hemmila, I., Dakubu, S., Mukkala, V.M., Siitari, H., Lovgren, T.,1984. Europium as a label in time-resolved immunofluorometricassays. Anal. Biochem. 137, 335.

Horn, U., Strittmatter, W., Krebber, A., Knupfer, U., Kujau, M.,Wenderoth, R., Muller, K., Matzku, S., Pluckthun, A., Riesenberg,D., 1996. High volumetric yields of functional dimeric minianti-bodies in Escherichia coli, using an optimized expression vectorand high-cell-density fermentation under non-limited growthconditions. Appl. Microbiol. Biotechnol. 46, 524.

Hubert, P., Nguyen-Huu, J.J., Boulanger, B., Chapuzet, E., Chiap, P.,Cohen, N., Compagnon, P.A., Dewe, W., Feinberg, M., Lallier, M.,Laurentie, M., Mercier, N., Muzard, G., Nivet, C., Valat, L., 2004.Harmonization of strategies for the validation of quantitativeanalytical procedures. A SFSTP proposal—Part I. J. Pharm.Biomed. Anal. 36, 579.

Karlsen, A.E., Hagopian, W.A., Grubin, C.E., Dube, S., Disteche,C.M., Adler, D.A., Barmeier, H., Mathewes, S., Grant, F.J.,Foster, D., 1991. Cloning and primary structure of a human isletisoform of glutamic acid decarboxylase from chromosome 10.Proc. Natl. Acad. Sci. U. S. A. 88, 8337.

Knappik, A., Pluckthun, A., 1994. An improved affinity tag based onthe FLAG peptide for the detection and purification of recombinantantibody fragments. Biotechniques 17, 754.

Kricka, L.J., 1999. Human anti-animal antibody interferences inimmunological assays. Clin. Chem. 45, 942.

Leslie, R.D., Atkinson, M.A., Notkins, A.L., 1999. Autoantigens IA-2and GAD in Type I (insulin-dependent) diabetes. Diabetologia 42, 3.

Lindner, P., Bauer, K., Krebber, A., Nieba, L., Kremmer, E., Krebber, C.,Honegger, A., Klinger, B., Mocikat, R., Pluckthun, A., 1997.Specific detection of His-tagged proteins with recombinant anti-Histag scFv-phosphatase or scFv-phage fusions. Biotechniques 22, 140.

Lövgren, T., Hemmilä, I., Pettersson, K., Halonen, P., 1985. Time-resolvedfluorimetry in immunoassay. Chapter 12. In: Collins, W.P. (Ed.),Alternative immunoassays. JohnWiley&Sons Ltd, NewYork, p. 203.

Madec, A.M., Rousset, F., Ho, S., Robert, F., Thivolet, C., Orgiazzi, J.,Lebecque, S., 1996. Four IgG anti-islet human monoclonalantibodies isolated from a type 1 diabetes patient recognizedistinct epitopes of glutamic acid decarboxylase 65 and aresomatically mutated. J. Immunol. 156, 3541.

NCCLS, 1986. Interference testing in clinical chemistry; proposedguideline. NCCLS EP7-P, Villanova. ISBN: 1-56238-020-6.

Okada, Y., Taniguchi, H., Schimada, C., 1976. High concentration ofGABA and high glutamate decarboxylase activity in rat pancreaticislets and human insulinoma. Science 194, 620.

Padoa, C.J., Banga, J.P., Madec, A.M., Ziegler, M., Schlosser, M.,Ortqvist, E., Kockum, I., Palmer, J., Rolandsson, O., Binder, K.A.,Foote, J., Luo, D., Hampe, C.S., 2003. Recombinant Fabs of humanmonoclonal antibodies specific to the middle epitope of GAD65inhibit type 1 diabetes-specific GAD65Abs. Diabetes 52, 2689.

Pihoker, C., Gilliam, L.K., Hampe, C.S., Lernmark, A., 2005.Autoantibodies in diabetes. Diabetes 54 (Suppl 2), S52.

Pipeleers, D.G., in't Veld, P.A., Van de Winkel, M., Maes, E., Schuit,F.C., Gepts, W., 1985. A new in vitro model for the study ofpancreatic A and B cells. Endocrinology 117, 806.

Saviranta, P., Pajunen, M., Jauria, P., Karp, M., Pettersson, K.,Mantsala, P., Lovgren, T., 1998. Engineering the steroid-specificityof an anti-17beta-estradiol Fab by random mutagenesis andcompetitive phage panning. Protein Eng. 11, 143.

Schlosser, M., Hahmann, J., Ziegler, B., Augstein, P., Ziegler, M.,1997. Sensitive monoclonal antibody-based sandwich ELISA fordetermination of the diabetes-associated autoantigen glutamic aciddecarboxylase GAD65. J. Immunoass. 18, 289.

Schlosser, M., Banga, J.P., Madec, A.M., Binder, K.A., Strebelow, M.,Rjasanowski, I., Wassmuth, R., Gilliam, L.K., Luo, D., Hampe,C.S., 2005. Dynamic changes of GAD65 autoantibody epitopespecificities in individuals at risk of developing type 1 diabetes.Diabetologia 48, 922.

Schwartz, H.L., Chandonia, J.M., Kash, S.F., Kanaani, J., Tunnell, E.,Domingo, A., Cohen, F.E., Banga, J.P., Madec, A.M., Richter, W.,Baekkeskov, S., 1999. High-resolution autoreactive epitopemapping and structural modeling of the 65 kDa form of humanglutamic acid decarboxylase. J. Mol. Biol. 287, 983.

Shi, Y., Veit, B., Baekkeskov, S., 1994. Amino acid residues 24–31 butnot palmitoylation of cysteines 30 and 45 are required for mem-brane anchoring of glutamic acid decarboxylase, GAD65. J. CellBiol. 124, 927.

Smismans, A., Ling, Z., Pipeleers, D., 1996. Damaged rat beta cellsdischarge glutamate decarboxylase in the extracellular medium.Biochem. Biophys. Res. Commun. 228, 293.

Solimena, M., Dirkx Jr., R., Radzynski, M., Mundigl, O., De Camilli, P.,1994.A signal locatedwithin amino acids 1–27 ofGAD65 is requiredfor its targeting to the Golgi complex region. J. Cell Biol. 126, 331.

143M. Rui et al. / Journal of Immunological Methods 319 (2007) 133–143

Suonpaa, M., Markela, E., Stahlberg, T., Hemmila, I., 1992.Europium-labelled streptavidin as a highly sensitive universallabel. Indirect time-resolved immunofluorometry of FSH and TSH.J. Immunol. Methods 149, 247.

Tremble, J., Morgenthaler, N.G., Vlug, A., Powers, A.C., Christie, M.R.,Scherbaum, W.A., Banga, J.P., 1997. Human B cells secretingimmunoglobulin G to glutamic acid decarboxylase-65 from anondiabetic patient with multiple autoantibodies and Graves'disease: a comparison with those present in type 1 diabetes. J. Clin.Endocrinol. Metab. 82, 2664.

Ujihara, N., Daw, K., Gianani, R., Boel, E., Yu, L., Powers, A.C., 1994.Identification of glutamic acid decarboxylase autoantibody hetero-geneity and epitope regions in type I diabetes. Diabetes 43, 968.

Weber, T.H., Kapyaho, K.I., Tanner, P., 1990. Endogenous interferencein immunoassays in clinical chemistry. A review. Scand. J. Clin.Lab Invest., Suppl 201, 77.

Winter, G., Milstein, C., 1991. Man-made antibodies. Nature 349, 293.Winter, G., Griffiths, A.D., Hawkins, R.E., Hoogenboom, H.R., 1994.

Making antibodies by phage display technology. Annu. Rev.Immunol. 12, 433.

Ziegler, B., Lucke, S., Kohler, E., Hehmke, B., Schlosser, M., Witt, S.,Besch, W., Ziegler, M., 1992. Monoclonal antibody-mediatedcytotoxicity against rat beta cells detected in vitro does not causebeta-cell destruction in vivo. Diabetologia 35, 608.