Bias in murine IgG isotype immobilisation Implications for IgG glycoform analysis ELISA procedures

JOURNAL OF IMMUNOLOGICAL METHODS EISEVIER

Journal of Immunological

Methods

I97 ( 1996) IO9- I20

Bias in murine IgG isotype immobilisation Implications for IgG glycoform analysis ELISA procedures Richard H.V. Jones, Thomas W. Rademacher, Molecular Mrdicirw Unit. Department

Phillip J. Williams

*

of Molecular Pathology. Unicersi@ College London Medical School, Windeyer Building. 46 Clerelund Street, London WIP 6DB. UK

Received 29 February

1996; revised I6 May 1996; accepted 4 June 1996

Abstract This study investigates immobilisation of murine IgG in various ELBA procedures. Monoclonal murine IgG isotypes and polyclonal IgG from sera were studied. Similar binding curves to plastic were found for all four individual murine IgG isotypes. Single isotypes displayed different affinities for both protein A and protein G, in particular IgGl was poorly and IgG3 strongly bound to both of these proteins. When mixtures of the isotypes were bound to either plastic. protein A or protein G, competition was observed in which IgG3 was dominant. Paradoxically, studies on the binding rates of single isotypes direct to plastic revealed that IgG3 had the slowest binding rate. Heating of bound IgGs resulted in significant but isotypically non-selective losses from the plates. The data demonstrate that despite obtaining equivalent individual IgG isotype binding curves. mixtures of IgG isotypes behave very differently, with competition for binding occurring even on plastic. The IgG isotype levels of murine sera were measured for individual mice, and the capture efficiency of each IgG isotype by protein A determined at different serum dilutions. Comparisons were made between the observed capture levels of IgG isotypes and their known serum levels. At all dilutions tested, greater than expected binding of IgG3, IgG2b and IgG2a was observed. At a serum dilution of l/100 the binding of these three isotypes was increased 16-, 2.Y- and O.Cfold, respectively. These increases were balanced by a decrease in IgGl binding which was the most prevalent serum IgG isotype. The results described above suggest that capture techniques are biased and unlikely to provide a coating of IgG isotypes that accurately reflects that of the serum. This bias is derived from the specificity of the individual isotypes for either protein A or protein G. and the errors further compounded by direct competition between isotypes whatever the capture surface. Induced coalescence of IgG3 may explain the latter observations. Keywords:

Protein A; Protein G: Isotype; IgG: ELISA; Glycosylation

1. Introduction

IgG capture assays have made a significant contribution to the study of disease-associated changes in

* Corresponding author. Tel.: +44-I7 I-580-5 171: Fax: I7 I-380-9497; e-mail: [email protected]. 0022-1759/96/$15.00 Copyright PII SOO22-1759(96)00122-6

i-44-

the galactosylation of IgG (Rook et al., 1988; Filley et al., 1989; Van Zeben et al.. 1994). These assays are performed using protein A or G coated plates to capture IgG from sera (Rook et al., 1988). The IgG can then be fixed to the surface with glutaraldehyde, heat denatured to expose the carbohydrate residues in the Cy2 domain and probed with monoclonal antibodies with specificity for N-acetylglucosamine.

0 1996 Elsevier Science B.V. All rights reserved

110

R.H. V. Jones et al. / Jounml o~lrrlrnlrrlnlngical Methods 197 (19961 10% I20

Variations of the assay include omission of the glutaraldehyde step or the heating step (Tsuchiya et al., 1993) and the use of lectins as carbohydrate probes (&mar et al., 1990; Bodman et al., 1994). Standardisation can be achieved using serum samples in which the carbohydrate content of the IgG has been independently analysed, alternatively simple reactivities can be compared in longitudinal studies. The normal reference parameter (%GO) is obtained by chemical sequencing which determines the percentage of IgG carbohydrate chains in serum IgG samples which terminate in N-acetylglucosamine. The reactivity of either lectins or anti-carbohydrate monoclonal antibodies with human IgG captured from serum samples using protein A coated plates, correlates with %GO determined by chemical

1

sequencing of human IgG purified from the same sample. This correlation is independent of the method of isolation of the IgG (Rook et al., 1991) and the %GO changes in rheumatoid arthritis are present on all human IgG isotypes (Rademacher et al., 1995). In contrast to human IgG we have been unable to detect any correlation between anti-GlcNAc reactivity of protein A captured murine IgG from sera, and the chemically sequenced values of protein A purified murine IgG (Rademacher et al.. 1996; Williams and Rademacher, 1996). As a first step in understanding this lack of correlation, we have examined possible anomalies in the binding of murine IgG to protein A, protein G or plastic. We utilised in-house biotinylated IgG monoclonal antibodies to investigate the differential bind-

2

3

4

Protocol

5

1

Protocol

6

used

5

6

7.4 5.5 5.5 5.5 7.4 7.4

7.4 5.5 5.5 5.5 5.5 7.4

1,,4 Coating protein G to plate Blocking of plate Incubation with sera Heating at 85°C ID12-7/6 incubation overnight Streptavidin/HRP incubation

1.4 5.5 5.5 5.5 5.5 5.5

1.4 1.4 7.4 1.4 7.4 7.4

7.4 5.5 7.4 1.4 7.4 1.4

7.4 5.5 5.5 7.4 7.4 7.4

Fig. 1. Effect of pH variation on a protein G based IgG capture ELISA. Protein G was coated onto an ELISA plate in PBS and then blocked with PBT. Following incubation of the plate with sera diluted I/ 100 and heating at 85°C for 15 min, IpG was probed with IDl2-7/6. an anti-GlcNAc monoclonal, then visualised using strepavidin/HRP and ABTS as described in Section 2. Each step was performed at either pH 7.4 (PBS) or pH 5.5 (100 mM sodium acetate) as described on the figure. Each data point represents the mean of three determinations * standard error of the mean.

R.H.V. Jones et al./Journol

of

ImmunologicalMethods 197 (19961 109-120

ing of each of the four murine IgG isotypes either Three different individually or in mixtures. capture/immobilisation methods were used including protein A, protein G and direct binding to plastic. We also investigated whether denaturation of the bound IgG isotypes by heating resulted in selective isotype loss from the plate. This study was then extended to include the analysis of polyclonal IgG isotypes captured on protein A at various serum dilutions.

2.3. Conventional

111

ELISA methodology

Maxi-sorp flat-bottomed 96-well microtitre plates were coated with 50 ~1 of 100 mM sodium carbonate buffer pH 9.2 (coating buffer) containing 125 ng of protein A or protein G at 4°C overnight. Plates were washed with PBS/O.OS% Tween 20 (PT) and blocked with 100 p,l of 1% BSA in PT (PBT) for 1 h at 37°C. Protein A plates were washed with PT and 1.6-400 ng of biotinylated IgG in 50 p.1 of 100 mM

2. Materials and methods

2.5,

(a)

2. I. Materials

2

Nunc Maxisorb 96-well flat-bottomed ELISA plates obtained from Gibco BRL Life Technologies (Paisley, Scotland, UK) were used throughout this study. Protein A. protein G, streptavidin/HRP and biotin ester were obtained from Sigma (Poole, Dorset, UK). Murine IgG isotype specific myeloma proteins were also obtained from Sigma and were as follows; IgGl (MOPC 21~), lgG2a (UPC 10~). lgG2b (MOPC 14 1K) and lgG3 (FLOPC 2 1K), none of these antibodies showed any reactivity to murine IgA, IgM, IgG or any cryoglobulin activity. HRP labelled goat anti-murine IgG isotype antibodies were obtained from Sera-Lab (Crawley Down. Sussex, UK). lD12-7/6, an IgM monoclonal antibody with N-acetylglucosamine specificity, was a kind gift from Professor G.A.W. Rook (University College London Medical School). 2.2. Biotinylation

I

Plastic

+

z

s

+

IgG2a

+

IgGZb

4

lgG3

1

d

8 I;5 8

IgGl

2.5@I 2Plastic 8YC 1.5-

of IgG isotypes

Biotinylation of IgG was carried out at low density by standard techniques (Harlow and Lane, 1988a). Briefly, 200 kg of each IgG isotype were dialysed into 100 mM sodium borate buffer pH 8.5, then diluted to give a final concentration of 1 mg/ml. 10 ~1 of DMSO containing 10 Kg of N-hydroxysuccinimide biotin ester were added and incubation was then carried out for 4 h at room temperature. 10 ~1 of 1 M NH&Cl was then added and after incubation for 10 min samples were dialysed against PBS and stored frozen.

10

100

IgG added per well (ng) Fig. 2. Immobilisation of individual biotinylated murine IgG isotypes directly to ELBA plates. Different amounts of biotinylated IgG monoclonal antibodies of the isotype shown were immobilked directly on to ELISA plates and then visualised (a) without the heating step, or (6) including heating at 85°C for 15 min. Each data point represents the mean of four determinations + standard error of the mean.

glycine/lSO mM NaCl pH 8.2 were added per well, the plates were then incubated for 2 h at 37°C. Protein G plates were similarly handled except that the biotinylated IgG was added in 50 ~1 of 100 mM sodium acetate pH 5.5 as murine IgG shows its greatest affinity for protein G at this pH (Akerstrom and Ejiirck, 1986). In another protein G based capture ELISA measuring anti-GlcNAc reactivity, we found that all steps subsequent to the incubation of sera could be performed in PBS pH 7.4 with no significant effect on the final results (Fig. Il. IgG was coated to some plates directly, 1.6-400 ng of biotinylated IgG were added in 50 pl of PBS for 2 h at 37°C followed by blocking with PBT. All plates were washed with PT then 50 ng of streptavidin/HRP in 50 p1 of PBT added per well and the plates incubated for 1 h at 37°C. Plates were again thoroughly washed and bound IgG visualised by the addition of 100 pg of 2,2-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) (ABTS) in 50 pl of 100 mM citrate phosphate buffer pH 4.1 containing

2.5(a) 21.5-

u

IgGl

---c

IgGla

+

IgGZb

0.005% H,O, and incubated for 15 min in the dark at room temperature. The reaction was stopped by the addition of 50 pl of 50 mM sodium fluoride and absorbance values measured at 655-490 nm. All results are expressed in terms of absorbance. 2.4. Helli denatltrrd

1gG

In addition some plates were subjected to 15 min heating at 85°C after incubation with the biotinylated IgG isotypes, this step was included in most assays to thermally denature the bound IgG so exposing the Fc N-linked oligosaccharides for analysis (Rook et al.. 1988). Briefly, following IgG capture the plates were washed and 100 ~1 of PBS (protein A or directly coated plates) or 100 mM sodium acetate (protein G plate) added per well. The plates were then floated on a water bath at 85°C for 15 min. Following heat denaturation the PBS or sodium acetate was removed and streptavidin/HRP added as above.

I-

IO00

1

10

100

1000

IgG added per well Cng) Fig. 3. Protein A and G capture of individual biotinylated murk IgG isotypes. Different amounts of biotinylated IgG monoclonal antibodiea of the lsotype shown were captured on ELISA plates using protein A, ((1) without the heating step, or (h) including the 85°C heating step. Similarly IgG isotypes were also captured on ELISA plates using protein G. (~1 without the heating step. or (d) including the 85°C heating step. Each data point represents the mean of four determinations f standard el-l-or of the mean.

R.H. V. Jones et a/./Journal

2.5. Isotype competition

of lr~v~~unolo~ical Metllods 197 (1996) 109-120

ELBA’s

These assays were performed essentially as described under the conventional ELISA methodology. with the exception that each well contained all four murine IgG isotypes. Each biotinylated IgG isotype was pre-mixed with equal quantities of the other three non-biotinylated isotypes (i.e. a total of four possible combinations). This experiment was performed at two different levels, such that there was a total of 2 or 0.5 pg IgG per well (i.e. 0.5 or 0.125 kg of each isotype only one of which was biotinylated). These total IgG levels are similar to those achieved in assays where IgG was present in considerable excess. Quadruplicate determinations were performed.

before, and quantitation (mg/ml) was achieved 2.8. Polyclonal

IgG isotype binding to proteirl A

2.5

(4

0

2.0 Fg total

IgG per well

n 0.5 ~g total IgG per well 2

1.5

Plastic 37oc

dependence

Maxi-sorp flat-bottomed 96-well microtitre plates were coated with 100 ng per well of the individual biotinylated IgG isotypes in either 50 ~1 of 100 mM sodium carbonate buffer pH 9.2 or in 50 ~1 of PBS. The plates were then incubated at 37°C for 2, 6, 18 and 54 min before being washed out with PT and blocked with 100 p,l of PBT for 1 h at 37°C. Binding of IgG isotypes was visualised with streptavidin/HRP and ABTS as described above. A competition ELISA was also performed essentially as described above except that to each well 100 ng of the individual biotinylated IgG isotype and 100 ng of the other three non-biotinylated IgG isotypes were added in PBS and the plates incubated at 37°C for various periods of time.

of IgG isotype serum levels using the on plate standards.

Maxi-sorp flat-bottomed 96-well microtitre plates were coated with 50 ~1 of PBS containing 125 ng of

1

2.6. lsotype binding-time

113

2b

I 3

(b)

1.5

I

Plastic 85OC

0.5

2.7. Determination

qf serum IgG isotype levels 0

Maxi-sorp flat-bottomed 96-well microtitre plates were coated with either 50 ng to 780 pg of myeloma IgG isotype standards or sera, serially diluted 1/lOOOO to l/80000 in PBS, at 4°C overnight, four plates were needed one for each IgG isotype. Plates were washed with PT and blocked with 100 ~1 of PBT for 1 h at 37°C. The respective HRP labelled goat anti-murine IgG isotype was then added in PBT at a dilution of l/1000 and the plate incubated for a further 1 h at 37°C. Visualisation was by ABTS as

k

c 2a

Labelled

I 2b

3

IgG isotype

Fig. 4. Immobilisation of equal mixtures of biotinylated murk IgG isotypes directly to ELBA plates. 500 ng or 125 ng of biotinylated IgG monoclonal antibody of the isotype shown (x axis). were pre-mixed with equal quantities of the other three unlabelled IgG isotypes. These were then immobilised on ELBA plates directly and the bound IgG visualised as described in Section 2 (a) without the heating step or (h) including the 85°C heating step. Each data point represents the mean of four determinations + standard error of the mean.

114

R.H.V. Jones et al./Journal

of immunological

protein A at 4°C overnight. In addition some wells were coated directly with the IgG isotype standards as described above to allow quantification. Plates were washed with PT and then blocked with 100 ~1 of PBT for 1 h at 37°C. Serial dilutions (l/251/25000) of sera from four mice were added in 50 p_l of 100 mM glycine/l50 mM NaCl, pH 8.2 and the plate incubated for 2 h at 37°C wells containing the standards were incubated in the presence of PBT. The plates were then washed and respective HRP labelled goat anti-murine IgG isotype were added in PBT at a dilution of l/ 1000 and the plates incubated at 37°C for 1 h. Goat antibodies were chosen as they give very low background levels, being unable to bind protein A. Visualisation was by ABTS, and quantitation of protein A bound IgG isotypes (ng/well) was achieved using the on plate standards.

Methods 197 (19961 109-120

3. Results 3.1. Single isotype binding studies Initially it was important to show that the four IgG isotype preparations were biotinylated to comparable degrees. In Fig. 2a it can be seen that all IgG isotype preparations bound directly to the plates and showed comparable levels of biotinylation, the assay reaching saturation at levels in excess of 100 ng per well. Heat treatment of the plate (Fig. 2b) had a minimal effect, reducing the quantity of IgG bound to the plate but affecting ail isotypes equally. When IgG was captured using protein A a very different set of data was obtained (Fig. 3a). Firstly assay saturation was achieved with only IgG3 and IgG2a and secondly large differences in the amount

2.5(c) 21.5l-

Protein G 37nc

0.50

I

0 2a

1

1

2b

2a

2b

2a

2b

2.5 -

2.5

(d)

(b)

1 3

21.5

1 Protein

1.5-

A

l_

Protein A 8S’C

0.5 0 1

2a

I 1

2b Labelled

c 3

IgG isotype

Fig. 5. Capture of equal mixtures of biotinylated murine IgG isotypes using protein A and G coated ELISA plates. Either 500 ng or 115 ng of biotinylated IgG monoclonal antibody of the isotype shown (_r axis), were premixed with equal quantities of the other three unlabelled IgG isotypes. These were captured on ELBA plates using protein A. then visualised (a) without the heating step or (h) including the 85°C heating step. Similarly IgG isotypes were also captured on ELBA plates with protein G and visualised (c) without the heating step. or (d) including the 15 min at 85°C heating step. Each data point represents the mean of four determinations f standard error of the mean.

R.H. V. Jones et al. /Journal

of Immunoiogical

of labelled isotypes bound to protein A were apparent, the order being IgG3 > IgG2a > IgG2b with virtually no IgGl bound. The heating step had little effect (Fig. 3b), causing only slight loss of bound IgG. All isotypes were affected equally. Using protein G a different set of apparent isotype affinities emerged (Fig. 3~). IgG3 was the only

Methods 197 (19961 109-120

115

isotype that saturated the assay, with overall apparent affinities for protein G being IgG3 > IgG2b > IgG2a > IgGl, the heating step again resulted in some non-selective losses (Fig. 3d). Very little IgGl (usuaily the major murine IgG isotype) was bound to protein A (Fig. 3a), but this situation was slightly improved with protein G (Fig.

L

(a)

pH 1.2

@)

pH 9.2

1.6

”

1.6

”

(cl 1.6-

0.8-

10

Coating time (min) Fig. 6. Effect of time on IgG isotype immobilisation directly to ELISA plates. ELISA plates were incubated with 100 ng of each biotinylated IgG isotype per well in either (a) PBS or (b) pH 9.2 carbonate buffer for either 2, 6, 18, or 54 min prior to blocking, and then visualised as described in Section 2. 100 ng of biotinylated IgG isotype were also premixed with 100 ng of the other three non-biotinylated isotypes. and then incubated on an ELISA plate in PBS for either 2, 6, 18, 54, or 108 min Cc). Bound IgG detected as described in Section 2. Each data point represents the mean of three determinations.

3~). In addition, while both capture proteins showed very specific IgG isotype selectivity, the immobilisation of the IgG directly onto the plate in PBS appeared to be non-selective (Fig. 2a). 3.2. Isotype competitim

ELISA ‘s

To achieve a more realistic serum like representation of IgG isotype binding, assays were performed with defined quantities of all four IgG isotypes. In these studies one biotinylated isotype was pre-mixed with an equal amount of each of the other three non-biotinylated isotypes before addition to the wells. These experiments permitted free competition between isotypes for the various capture methods. Direct coating of IgG to the plates resulted in maximal binding of biotinylated IgG3 (Fig. 4a). The binding of biotinylated IgG2a or IgG2b was diminished in the presence of an equal mixture of non-biotinylated IgGl. IgG2b and IgG3 or IgGl. IgG2a and IgG3 respectively. Very little biotinylated IgGl was immobilised to plastic when equal amounts of non-

lgG1

IgG2a

biotinylated IgG2a. IgG2b and IgG3 were also present. This result was surprising since IgGl alone was found to bind to plastic with the same relative affinity as the other isotypes (Fig. 2a). Small equivalent losses (approximately 20-309~) of immobilised isotypes occurred following the 85°C heating step, confirming no selective IgG isotype loss (Fig. 4bl. Fi g. 5a shows that essentially no biotinylated IgG1 was captured on protein A coated ELISA plates in the presence of equal amounts of non-biotinylated IgG2a, IgG2b and IgG3. In addition little IgG2a or IgG2b was captured in the presence of a mixture of equal amounts of non-biotinylated IgGl, IgG2b and IgG3 or IgGl , IgG2a and IgG3. respectively. As found for the direct coating experiments IgG3 exhibited the greatest affinity for protein A effectively out competing the other isotypes. The order of binding was found to be IgG3 > IgG2a > IgG2b > > IgG 1. Heating of the plate after capture of the IgG had no effect on the overall pattern IFig. 5b), although losses increased to around 50%. A was found for the similar pattern of competition

IgG2b

IgG3

Murine IgG isotype Fig. 7. IgG isotype levels of four mice sera as determined by ELISA. Mouse 3 WI\ a female DBA-I, the rest were male DBA-I. all hera were obtained from mice aged between 14-16 weeks old. All four mice were arthritic at the time of serum collection (induced with bovine type II collagen and Freunds complete adjuvant). numbers l-3 were moderately severe while number 4 had very severe arthritis. The bound serum IgG isotypes were detected with goat HRP-labelled murine IgG isotype specific polyclonal antisera. The relative anti-GlcNAc reactivities (absorbance values) of the sera l-4 were 0.059, 0.086. 0.181 and 0.195 respectively. Each data point represents the mean ot three determinations.

R.H. V. Jones et al. / Journal of Immunological

117

Methods 197 (19961 109-120

capture of the different isotypes on protein G coated plates, as shown in Fig. 5c and Fig. 5d. The order of binding was IgG3 > > IgG2a = IgG2b > IgGl . 30

3.3. Isotype binding-time

+ IgGl ._....._ +...... IgG?&

dependence

This study was performed in an attempt to understand the data in Fig. 4a, which showed decreased binding of IgGl direct to plastic in the presence of the other isotypes. Contrary to expectations when the individual isotypes were added in PBS binding of IgG2a, IgG2b, and IgGl occurred rapidly and to a similar degree. However, while eventually reaching the same level of saturation IgG3 appeared to bind more slowly than the other isotypes (Fig. 6a). When immobilising the IgG isotypes directly to the plate in pH 9.2 ‘coating’ buffer the anomalous behaviour of lgG3 was even more pronounced (Fig. 6b). In contrast, the time dependant competition ELISA on plastic, showed the following isotype binding affinities IgG3 > IgG2b > IgG2a > IgGl, which were apparent even after two min (Fig. 6~). As expected this data shows great similarity with that illustrated in Fig. 4a. 3.4. Studies with polyclonal

20

.___ 0

IgGj

10

0 Mouse 2 20

GJ E z 3 5 a T3

10

O 4 / k.

Mouse 3

,i/

g $ ,M

20

8’ /

10

\\

,I’

serum IgG

The IgG isotype levels of four DBA/l murine sera are presented in Fig. 7. Mouse number 4 (male with severe type II collagen induced arthritis) had considerably elevated levels of all four IgG isotypes. The other three mice exhibited very similar isotype levels with IgGl > IgG2a > IgC2b > IgG3. Fig. 8 shows the quantities of each IgG isotype in sera from the four individual mice when bound to protein A at various dilutions, with results expressed as ng IgG isotype per well. It is important to realise that this data was obtained using IgG isotype specific detecting antibodies and not biotinylated antibodies. IgGl, the predominant isotype, was very poorly captured at all dilutions tested. At the highest serum dilution IgG2a was the only isotype readily detected. IgG2b and IgG3 exhibited the highest affinity for protein A in this type of assay. as bound levels generally increased steadily as serum dilution decreased. These isotypes were also able to displace the IgG2a which is normally present in serum at a much higher concentration. Mouse serum 3 gave

IgG2b

__..

____b____

,*

-.........._

,..’

.i”

*

(’ ,,.......+_,,..w

,,,__,

:

:

\

: i o”” ..A ._.“..

0 Mouse 4 20 a ,,a’ . . . . . . .

o”’ ..A ‘-“““.....~.._...,..._.._ o...j::..,, ,,(.....

10

o../-”

8

0 10000

1000

100

Serum dilution Fig. 8. Effect of serum dilution on relative IgG isotype capture by protein A based ELISA. This figure shows the amount of each specific IgG isotype bound to an ELBA plate via protein A at various serum dilutions. The same mouse serum samples as in Fig. 7 were used, mouse serum sample numbers are given in the top right corner of each graph. The protein A bound serum IgG isotypes were detected with goat HRP-labelled murine IgG isotype specific polyclonal antisera. Each data point represents the mean of three determinations.

II8

R.H.V.

Jones et al. /Joumal

qj’lmmunological

exceptionally high levels of IgG3 binding, while mouse sera 4 gave exceptionally high levels of IgG2b binding at the important I/ 100 dilution level used in many assays. A comparison of Figs. 7 and 8 clearly shows that at no serum dilution were the expected ratios of IgG isotypes observed.

4. Discussion In the present study we have shown that the four biotinylated murine IgG isotypes bind to plastic ELISA plates with equal ability. However, in the presence of other competing isotypes we have found that the binding of specific isotypes to either plastic, protein A or protein G was dramatically altered. The data demonstrates that despite exhibiting near identical binding curves to plastic, components will not be bound (and detected) with equal efficiency when present in a mixture. While we would have anticipated protein A and G based assays to exhibit some bias on addition of IgG in excess, the observed isotype competition for immobilisation to plastic was unexpected. Surprisingly, the pattern of competition on plastic differed only slightly from the patterns found for protein A or protein G. In all three capture methods, IgGl showed the weakest and IgG3 the strongest apparent binding affinities. It is difficult to compare our results with conclusions drawn from the previous Iiterature on the relative affinities of the murine IgG isotypes with protein A or G (Harlow and Lane. 1988b). For example. the studies of Ey et al. (1978) and Seppal’ri et al. (198 1) report the ability of pH to selectively elute the different murine IgG isotypes from protein A affinity columns. However it is not clear how conclusions on relative binding constants can be drawn from these experiments. Akerstriim et al. (1985) reported the percentage of IgG bound to Sepharosecoupled protein A under conditions where the concentrations of IgG and protein A were near equimolar. No binding constants were calculated and no statistical differences were found between the percent IgG bound for the different murine isotypes although a hierarchy of relative binding was proposed. Eliasson et al. (1988) calculated the amount of mouse monoclonal IgG required to give 50% inhibition of rabbit polyclonal IgG reactivity towards

Methods

197 (19961

109-120

protein A or G. These apparent inhibition constants differed dramatically fro,“’ the suggested hierarchy of affinities proposed by Akerstriim et al. (1985). The only detailed study of association constants that we are aware of is the report by Biedermann et al. (199 1) which found that the association constant of murine IgGl to protein A depended on the support used. However. two distinct association constants were always found. The high affinity site varied from 1.0 X 10’” 1 mol-’ to 1.7 X 10” 1 mall’ and the low affinity site varied from 1.0 X lo6 1 mol-’ to 5.9 X IO3 1 mol-‘. Recovery from the columns also ranged from 57% down to 2%. Interestingly only 0.3-3% of the protein A molecules were able to bind IgG at high affinity. Our assays were performed under conditions of limiting protein A and protein G with IgG in excess. This is the opposite of column affinity purification where immobilised protein A or protein G will be in excess. At present it is not clear why IgG3 was able to compete-out the binding of IgGl, and to a lesser extent IgG2a and IgG2b binding to plastic. In contrast. the data in Fig. 6a suggests that IgG3 on its own binds to plastic more slowly than the other isotypes. One possible explanation for this apparent discrepancy is that non-IgG3 isotypes act as a ‘seed’ for multiple high affinity interactions with IgG3, possibly mediated through the Fc region. It has previously been observed that murine IgG3 is ‘sticky’. and possesses a novel Fe-mediated mechanism that gives rise to increased self-associations (Greenspan and Cooper, 1992). It is interesting that the hierarchy of affinities to protein A in this study is the same as the quoted affinities (IgG3 > IgG2a > IgG2b > > IgGl) for murine IgG isotypes for the human FcyIII low affinity receptor (Van de Winkel and Capel, 1993). The observations reported here question the accepted view that the differential binding of the isotypes to protein A and protein G, and possibly Fey receptors, solely reflect a hierarchy of binding affinities. That is, these assays may be measuring the relative propensity for the different isotypes to form multimers, which give rise to high affinity multivalent binding, with IgG3 the most prone to aggregation and IgGl the least. Biotinylated IgGs enabled us to assess losses, following the 85°C heating step that is commonly used in assays to expose the carbohydrate residues.

R.H.V. Jones et al. / Journal of immunological

The present data suggests that anti-IgG y chain detection showing losses of > 90% were exaggerated due to epitope denaturation (unpublished observations). Heating the plate had no selective effect, causing losses in the region of 30-50% for all captured IgG isotypes. In order to assess the behaviour of polyclonal IgG isotypes in a serum matrix, four mice sera were studied. Although individual sera contained quantitatively different amounts of the four IgG isotypes, we observed that IgG3 and IgG2b generally exhibited increased binding, while the amount of IgG2a bound was near expected values, and IgG1 bound poorly. The experiments also show that our data obtained using monoclonal antibodies is similarly reflected in experiments using polyclonal IgG. Further, there is no evidence of any interference from serum components. Interestingly the total amounts of IgG captured differed (range of 20-40 ng) between the different mouse sera, especially at the I/ 100 dilution at which IgG is present in considerable excess. Mice in which elevated levels of IgG were seen also had elevated anti-GlcNAc reactivity. The increase in total IgG bound to the plate may reflect the presence of IgG species in these arthritic mouse sera which bind to low affinity sites on protein A, stacking of IgG molecules due to complex formation (self-associations) or cross-linking of additional IgG molecules to the monomer surface via IgM rheumatoid factors. With respect to measuring the glycosylation status of murine IgG by protein A or protein G capture assays, some of the observations in this paper raise obvious concerns. This is especially so as differences in murine and human IgG isotype specific glycosylation have been observed (Williams, 1996). Furthermore, errors in the measurement of glycosylation status may be compounded, since it is known that distinct IgG isotypic responses are associated with specific infections and disease models. For example, bacterial and viral infections often promote Thl responses which may lead to elevated IgM and IgG2a production (Coutelier et al., 19881, whereas parasitic infections are characterised by Th2 responses and elevated IgGl and IgE levels (Ferrick et al., 1995). In addition unusually high levels of IgG3 (> 5 mg/ml) have been observed in the murine MRL/lpr lupus model (Takahashi et al., 1991), and have been clearly implicated in certain aspects of the disease

Methods 197 (19961 10%120

119

pathogenesis (Takahashi et al., 1991; Fulpius et al., 1993). We have found in the DBA/ 1 mouse, elevated levels of IgG3 (up to 1 mg/ml by RID, data not shown) with IgG2b at lowest abundance. Some groups pre-purify IgG prior to lectin analysis (Tsuchiya et al., 19931, avoiding the need for IgG capture by protein A or protein G. However, the use of protein A or protein G affinity columns to purify IgG can still result in low yields and some isotypic selectivity. While IgGl is often the major murine isotype found in sera, high losses of IgGl may be experienced when using protein A immunoaffinity columns for purification, this is consistent with our observations on IgGl capture by protein A in the present study. While the purification yield can be enhanced by the use of high pH (> 8.5) and high salt (3 M NaCl) loading (Harlow and Lane, 1988c), purification methods such as salt fractionation, or gel filtration and ion exchange chromatography, are more likely to be isotypically non-selective (Bond et al., 1995; Bond et al., 1990). In any event the data presented in this paper, suggests that isotype competition may be a serious problem if these purified preparations are probed following immobilisation to plastic. In conclusion, the data reported here using both polyclonal IgG and isotype specific murine IgG myeloma proteins, suggest that IgG capture or immobilisation techniques used in ELISAs are biased and unlikely to provide a coating of IgG isotypes that accurately reflect that of the starting preparation. This bias derives from the affinity of individual isotypes for either protein A or protein G, and the observation that whatever the capture surface, isotype competition plays a strong part in determining the final proportion of each isotype eventually bound.

Acknowledgements

The Molecular Medicine Unit gratefully acknowledges the financial support of the Arthritis and Rheumatism Council, the Medical Research Council and The Sir Jules Thorn Charitable Trust. P.J.W. was in receipt of an Emily Le Rossignol fellowship. We wish to thank Graham Rook for helpful discussions.

120

R.H.V. Jones et al./

Journal

qf Immunological

References Akerstriim, B. and Bjiirck, L. (19861 A physiochemical study of Protein G, a molecule with unique immunoglobulin G-binding properties. J. Biol. Chem. 261. 10240. Akerstrom. B., Brodin. T.. Reis, K. and Bjdrck, L. (19851 Protein G: A powerful tool for binding and detection of monoclonal and polyclonal antibodies. J. Immunol. 135. 2589. Biedermann, K., Sabater. M.. Sorensen, J., Fiedler, H. and Emborg, C. (19911 Quantitative binding studies of a monoclonal antibody to immobilized protein-A. Bioseparation 2, 309. Bodman, K.B., Hutchings, P.R., Jeddi, P.A., Delves, P.J., Rook, G.A., Sumar, N., Roitt, I.M.. Lydyard, P.M. (1994) IgG glycosylation in autoimmune-prone strains of mice. Clin. Exp. Immunol. 95. 103. Bond, A., Cooke, A. and Hay, F.C. (19901 Glycosylation of IgG immune complexes and IgG subclasses in the MRL-lpr/lpr mouse model of rheumatoid arthritis. Eur. J. Immunol. 20. 2229. Bond. A.. Kerr, M.A. and Hay, F.C. (19951 Distinct oligosaccharide content of rheumatoid arthritis derived immune complexes. Arthritis Rheum. 38, 744. Coutelier, J.P., Van de Logt. J.T.M., Heessen, F.. Vink, A. and vansnick, J. (19881 Virally induced modulation of murine IgG antibody subclasses. I. Exp. Med. 168. 2373. Eliasson, M., Olsson. A., Palmcrantz. E., Wiberg. K.. InganPa. M.. Guss. B., Lindberg, M and Uhlen, M. (19881 Chimeric IgG-binding receptors engineered from staphylococcal protein A and streptococcal protein G. J. Biol. Chem. 263, 4323. Ey, P.L., Prowse, S.J. and Jenkin, C.R. (19781 Isolation of pure IgGl. IgG2a and IgG3 immunoglobulins from mouse serum using protein A-Sepharose. Immunochemsitry 15. 429. Ferrick, D.A., Schrenzel. M.D., Mulvania. T., Hsieh, B.. Ferlin, W.G. and Lepper. H. (1995) Differential production of interferon-y and IL4 in response to T,, and T,? stimulated pathogens by yS T cells in viva. Nature 373, 255. Filley, E., Andreoli, A.. Steele, J.. Waters, M., Wagner. D.. Nelson, D.. Tung, K., Rademacher. T.. Dwek, R. and Rook, G.A. (19891 A transient rise in agalactosyl IgG correlating with free interleukin 2 receptors, during episodes of erythema nodosum leprosum. Clin. Exp. Immunol. 76. 343. Fulpius. T.. Spertini, F., Reininger, L. and Izui, S. (19931 Ig heavy chain constant region determines the pathogenicity and the antigen binding activity of rheumatoid factor. Proc. Natl. Acad. Sci. USA 90, 2345. Greenspan, N.S. and Cooper, J.N. (19921 Intermolecular cooperativity: a clue to why mice have IgG3. Immunol. Today 13, 164. Harlow, E. and Lane. D. (1988al Labeling antibodies. In: E. Harlow and D. Lane tEds.1, Antibodies a Laboratory Manual. Cold Spring Harbor Laboratory. New York, p. 341. Harlow. E. and Lane, D. (1988bl Reagents. In: E. Harlow and D. Lane (Eds.1, Antibodies a Laboratory Manual. Cold Spring Harbor Laboratory, New York. p. 618.

Metho&

197 (1996) 109-120

Harlow. E. and Lane, D. (1988~1 Storing and purifying antibodies. In: E. Harlow and D. Lane (Eds.1, Antibodies a Laboratory Manual. Cold Spring Harbor Laboratory, New York, p. 3 I 1. Rademacher, T.W., Jones, R.H.V. and Williams, P.J. (19951 Significance and molecular basis for IgG glycosylation changes in rheumatoid arthritis. Adv. Exp. Med. Biol. 371. 193. Rademacher. T.W.. Jaques. A. and Williams. P.J. (19961 The defining characteristics of immunoglobulin glycosylation. In: D.A. Isenberg and T.W. Rademacher (Eds.1. Abnormalities of IgG Glycosylation and Immunological Disorders. Wiley. Chichester, England. p. 1. Rook, G.A.W., Steele, J. and Rademacher, T.W. (19881 A monoclonal antibody raised by immunising mice with group A streptococci binds to agalactosyl IgG from rheumatoid arthritis. Ann. Rheum. Dis. 47, 247. Rook, G.A.W.. Steele, J.. Brealey, R., Whyte, A., Isenberg, D., Sumar. N., Nelson, J.L., Bodman, K.B., Young, A.. Roitt, I.M.. Williams, P., Scragg. I.. Edge, C.J.. Arkwright, P.D., Ashford, D.. Wormald. M., Rudd, P., Redman, C.W.G., Dwek, R.A. and Rademacher. T.W. (19911 Changes in IgG glycoform levels are associated with remission of arthritis during pregnancy. J. Autoimmun. 4, 779. Seppalli. I., Sarvas, H.. PCterfy. F. and Miikela, 0. ( 198 Il. The four subclasses of IgG can be isolated from mouse serum by using protein A-Sepharose. Stand. J. Immunol. 14, 335. Sumar. N., Bodman. K.B., Rademacher, T.W., Dwek. R.A., Williams. P.J., Parekh, R.B., Edge, J., Rook. G.A.W.. Isenberg, D.A., Hay, F.C. and Roitt. I.M. (19901 Analysis of glycosylation changes in IgG using lectins. J. Immunol. Methods 131. 127. Takahashi. S., Nose, M., Sasaki, J.. Yamamoto, T. and Kyogoku. M. (1991) IgG3 production in MRL/lpr mice is responsible for development of lupus nephritis. J. Immunol. 147, 5 IS. Tsuchiya, N.. Endo. T., Matsuta, K., Yoshinoya. S.. Takeuchi. F.. Nagano. Y.. Shiota. M., Furukawa, K., Kochibe, N.. Ito. K. and Kobata. A. (19931 Detection of glycosylation abnormality in rheumatoid IgG using N-acetylglucosamine-specific Pscrrhyrellrl r~elutino lectin. J. Immunol. 15 I. I 137. Van de Winkel. JGJ. and Cape], PJA. (19931 Human IgG Fc receptor heterogeneity: molecular aspects and clinical implications. Immunol. Today 14, 215. Van Zeben, D., Rook, G.A.W.. Hazes. J.M.W., Zwinderman, A.H.. Zhang. Y.. Ghelani, S., Rademacher. T.W. and Breedveld. F.C. (19941 Early agalactosylation of IgG is associated with a more progressive disease course in patients with rheumatoid arthritis: results of a follow up study. Br. J. Rheum. 33. 36. Williams. P.J. t 1996) Analytical methodsl: Biochemical. In: D.A. Isenberg and T.W. Rademacher tEds.1, Abnormalities of IgG Glycosylation and Immunological Disorders. Wiley. Chichester, England. p. 45. Williams. P.J. and Rademacher. T.W. (19961 Glycosylation of IgG isotypes in murine collagen induced arthritis. Stand. J. Immunol.. in press.