Journal of Autoimmunity (1997) 10, 77–85

A Detailed Lectin Analysis of IgG Glycosylation,Demonstrating Disease Specific Changes in TerminalGalactose and N-acetylglucosamine

Angela Bond 1, Azita Alavi 1, John S. Axford 1, Brian E. Bourke 2, Felix E. Bruckner 2,Michael A. Kerr 3, J. Douglas Maxwell 4, Karen J. Tweed 1, Michael J. Weldon 4, PierreYouinou 5 and Frank C. Hay 1

1Division of Immunology and AcademicRheumatology Unit, St George’s HospitalMedical School, Cranmer Terrace,London, UK2Department of Rheumatology, StGeorge’s Hospital, London, UK3Department of Pathology, University ofDundee, Ninewells Hospital and MedicalSchool, Dundee, UK4Department of Biochemical Medicine, StGeorge’s Hospital Medical School,Cranmer Terrace, London, UK5Laboratoire D’Immunologie, CentreHospitalier Regional et Universitaire deBrest, Brest, France

Received 19 July 1996Accepted 16 September 1996

Serum IgG from rheumatoid arthritis patients contains a decreased number ofoligosaccharide structures ending in galactose and thus there is an increase inN-acetylglucosamine as the terminal sugar, compared with healthy individ-uals. The relationship between these two sugars varies depending on thedisease examined: IgG from patients with rheumatoid arthritis, juvenile onsetchronic arthritis and Crohn’s disease are at one extreme, and exhibit areciprocal galactose:N-acetylglucosamine relationship, while Sjögren’s syn-drome and osteoarthritis IgG are at the other extreme, exhibiting a parallelincrease in the expression of both galactose and N-acetylglucosamine. Theseresults may occur as a consequence of more than one glycosylation site whichis differentially glycosylated, but more likely by changes in the level ofbisecting N-acetylglucosamine. © 1997 Academic Press Limited

Key words: rheumatoidarthritis, juvenile onset chronicarthritis, Crohn’s disease,infectious endocarditis,Sjögren’s syndrome, SLE,ankylosing spondylitis, ulcerativecolitis, osteoarthritis, IgGglycosylation, galactose,N-acetylglucosamine

Introduction

It is now well established that there are significantglycoform differences in serum IgG from rheumatoidarthritis (RA) patients 1, 2]. Parekh and colleaguesstudied serum IgG from RA using a complex chemicalmethod including controlled hydrazinolysis and gel-permeation chromatography, and showed that thedegree of oligosaccharide galactosylation was lowerthan that of normal individuals [1, 2]. Similar IgGsugar changes are also believed to occur in a restricted

Correspondence to: Angela Bond, Division of Immunology, StGeorge’s Hospital Medical School, Cranmer Terrace, LondonSW17 0RE. E-mail: a.bond@sghms.ac.uk.

770896-8411/97/010077+09 $25.00/0/au960104

group of other diseases where principally, the mainN-glycosylation difference is a lack of outer-arm galac-tosylation expressed as %GO. Crohn’s disease (CD),juvenile onset chronic arthritis (JCA), osteoarthritis(OA) and Mycobacterium tuberculosis (MTB), as wellas systemic lupus erythematosus (SLE) complicatedby Sjögren’s syndrome (SS) may have abnormallygalactosylated IgG [3].Our own method of IgG glycosylation analysis,

which correlates strongly with the biochemicallydetermined %GO, uses the ratio of Bandeiraea sim-plicifolia II (BSII, used to detect terminal N-acetylglucosamine (GlcNAc)) to Ricinus communisagglutinin, RCA I (used to detect galactose). Theseare the two consecutive sugars normally found atthe terminus of complex bi-antennary N-linked

© 1997 Academic Press Limited

78 A. Bond et al.

oligosaccharides associated with the Fc region of IgG.It has previously been thought that the detectableamounts of galactose and GlcNAc should varyinversely with each other, as the absence of terminalgalactose would expose GlcNAc, the next sugar alongthe oligosaccharide chain. In fact a variety of studiesexamining glycosylation changes have based theiranalysis on this pretext, often assuming that themeasurement of one sugar can be used to provideinformation about another sugar [4]. Using our lectinanalysis we have demonstrated that in RA, the detec-tion of these two sugars is consistently found to be ofa reciprocal nature. However, when studying theMRL-lpr/lpr mouse model of arthritis we have evi-dence to indicate that other sugar changes may beoccurring. In this model, as the galactose expressionincreases, so the expression of GlcNAc increases inparallel, and not in a reciprocal way as in RA [5].Interestingly, our earlier studies on IgG from patientswith Sjögren’s syndrome (SS) indicate a similar pat-tern of sugar expression to that found in MRL-lpr/lpr[6]. The culmination of the data so far suggests thatIgG glycosylation changes may be much more com-plex than first believed and so it is important toexamine the expression of the sugars individually, inany given disease. In order to further examine thedifferences in the interrelationship of galactose andGlcNAc, we have set out to measure the expression ofthese sugars independently, in other diseases whichhave previously only been compared in terms of %GO[2]. We have included, to allow for statistical evalu-ation between all the disease groups, our data fromRA, JCA and SS patients [7].We have analysed IgG glycosylation from healthy

individuals and patients from nine different diseasegroups using our glycosylation characterizationmethod in order to reveal specific disease associatedpatterns of sugar changes, as indicated by the inter-relationship of galactose and GlcNAc, and thus clas-sify them into those which are RA-like, those whichare SS-like and those which may have a differentsugar pattern.

Materials and methods

Patients

A total of 264 serum samples from nine patient groupsand a healthy control population were obtained andstored at −20°C. The samples from 100 healthy indi-viduals were obtained from the South Thames BloodTransfusion Centre, London SW17 0QS, and the meanage of the group was 36 years (range 19–64 years).Data were unavailable for two individuals. Twelvesamples from patients with infective endocarditis (IE)were obtained from Ninewells Hospital, Dundee. Themean age of this group was 71 years (range 54–84years).Serum samples were also obtained from patients

attending the Rheumatology clinics at St George’s

Hospital. These were as follows: 29 samples frompatients diagnosed as having RA according to therevised criteria of the American College of Rheuma-tology [8] where the mean age was 54 years (range 36to 76 years); 12 samples from patients diagnosed withsystemic lupus erythematosus (SLE) by the criteria ofthe American Rheumatism Association [9] where themean age was 44 years (range 27 to 68 years); 13 frompatients with osteoarthritis (OA), where the mean ageof this group was 67 years (range 52 to 91 years),although data were unavailable for 1 individual and17 from patients with ankylosing spondylitis (AS), allmeeting the New York criteria [10], where the meanage of this group was 45 years (range 23 to 65 years).A group (n=11) comprising individuals with JCA(n=9) and with adult onset Stills’ (n=2) was alsostudied. The mean age of this group was 20 years(range 2 to 50 years).Seventeen samples from patients with Crohn’s

disease (CD) and 16 with ulcerative colitis (UC)attending the Gastroenterology clinic at St George’sHospital were obtained. The diagnosis of patients wasestablished by standard radiological, histological andendoscopic criteria [11]. The mean ages and rangeswere 41 years (range 30 to 85 years) and 49 years(range 25 to 83 years) respectively.Thirty-seven SS serum samples were obtained from

patients attending the clinic of Internal Medicine,Brest University Hospital Medical School, and allsatisfied the recognised criteria [12] where the meanage of this group was 60 years (range between 31 and84 years).RA, JCA, OA, AS, SLE patients were classified into

two groups, ‘active’ and ‘inactive’ on the basis ofstandard clinical and laboratory parameters i.e. rheu-matoid factor (RF, measured using the Roche COBASMIRA latex immunoturbidimetric method), C-reactiveprotein (CRP), erythrocyte sedimentation rate (ESR)and clinical examination of actively inflamed jointse.g. presence or absence of synovitis.Twelve patients with subacute bacterial endocar-

ditis (IE) had diagnoses confirmed by blood cultureand/or detection of vegetations by echocardiography.UC disease activity was measured using the Trueloveand Witts index [13], and the Crohn’s disease activitymeasured by the Crohn’s disease activity index(CDAI) [14] or by the Harvey Bradshaw index [15].SS disease activity measurement included anti-SSAand anti-SSB antibody titres as well as RF andextra-glandular manifestations.

IgG purification

Ion exchange chromatography was used to purify IgGfrom serum as previously described [16]. Fractionswere collected and the first peak containing IgG wasretained for protein estimation. IgG recovery rangedbetween 48 and 76%; this was determined by singleradial immunodiffusion using anti-human IgG(Amersham International plc, UK) [17]. The ionexchange procedure was specially developed for our

Galactose: N-acetylglucosamine relationship on IgG in disease 79

studies on IgG glycoforms as it is known that otherproteins can contaminate the preparations [18]. Greatcare was taken to check that the preparations werefree of detectable impurities. Sodium dodecyl sul-phate polyacrylamide gel electrophoresis (SDS-PAGE)was used to examine impurities, as previouslydescribed [16]. After transfer onto nitro-cellulose theIg class was confirmed by using specific antisera(Amersham International) and no IgA or IgM wasdetected in the preparation. Total protein was stainedfor using Protogold (Biocell Research Laboratories,Cardiff, UK) and revealed low levels of low molecularweight contaminants which were found not to beglycosylated.

Protein determination

The protein determinations of the IgG preparationswere achieved using the Pierce BCA protein assay(Pierce Europe, The Netherlands).

Glycoform detection by immunoblotting

IgG sugar composition was estimated using our pre-viously described lectin binding assay [16], withreference to standards which had been determinedby controlled hydrazinolysis and gel-permeationchromatography [1]. Equivalent amounts of IgG (4 ìg)were dot blotted onto nitrocellulose where terminal

galactose and GlcNAc residues were detected by theuse of biotinylated lectins. The reproducibility andstandardization of this assay has been confirmed incomparative trials [19]. RCA I was used to detectgalactose and BSII (Vector Laboratories, UK) was usedto detect GlcNAc. Streptavidin–horseradish peroxi-dase (Amersham International) was used as the biotindetector. Samples were measured in triplicate andstandardization was carried out for each experimentto allow for day to day variations; therefore, eachexperiment was referred to the standards and com-plete in itself. Optical density (OD) measurementswere made using a Hewlett Packard desk scanner andanalysed using Collage software (Techgen).The %GO was calculated from the ratio of BSII

binding/RCA I binding (i.e. detection of GlcNAc/detection of galactose) with reference to biochemicallydetermined %GO standards. Coefficient of variance ofthis assay was 15.6%.Standards with known galactose or GlcNAc levels

were not available. IgG measurements of these sugarswere therefore expressed in terms of units. This wasachieved by expressing the OD of each sample as afraction of an internal standard IgG, designated 100units. The assay proved robust and reproducible. Thecoefficient of variance of the galactose and GlcNAcassays were 4.8% and 10.0% respectively.For ease of understanding, the results are described

in terms of ‘levels of sugars’, although determinationof sugars was by lectin binding to exposed residues onthe oligosaccharide chains.

Figure 1. Comparison of IgG %GO from patients with various diseases and normal controls. IgG %GO values were calculatedusing the ratio of BSII binding to RCA I binding (GlcNAc:galactose ratio) with reference to a standard curve. Each individualsample was measured in triplicate. The solid horizontal line represents the mean IgG %GO of the normal control population.Dashed horizontal lines represent the 95% confidence interval (CI) of the normal population. Means and 95% confidenceintervals are shown for each patient group.

80 A. Bond et al.

Statistical analysis

Association analysis between %GO, as well as galac-tose and GlcNAc units, and disease activity wasachieved using regression analysis and Fisher’s exacttest.Separate confidence intervals (of the mean) for each

group were calculated and are illustrated in the fig-ures where appropriate. Multiple comparison pro-cedures were performed; a one way analysis ofvariance followed by Gabriel’s test were used tocompare each group mean with each other groupmean [20]. %GO data were log transformed (base ten)because the distribution was positively skewed andthe variability increased with the mean. Paired datawhen measuring galactose and GlcNAc detectionwere compared by simple linear regression, and 95%confidence intervals for the difference between theslopes of regression lines were used [21], where azero difference between slopes near the middle of aconfidence interval shows no evidence that the twopopulation regression lines have different slopes.

Results

Percentage of IgG oligosaccharide chains lackinggalactose (%GO)

The IgG carbohydrate content from patients withdifferent diseases fell into three groups when com-pared to IgG from normal controls (Figure 1).The disease groups which had significantly raised

mean IgG %GO values when compared with that ofnormal controls were IE, RA, CD (all P<0.01) andJCA (P<0.05). Mean IgG %GO values from SLE, UC,AS and primary SS patients were similar to thatof the normal controls. Significantly reduced mean

IgG %GO values were detected in patients withOA when compared with that of normal controls(P<0.01).

IgG galactose content

In all the disease groups IE, RA, CD, SLE, JCA, UC,AS, SS and OA, the mean IgG galactose detected wasmarkedly less than the mean level of the healthygroup. This drop in IgG galactose was highly signifi-cant (P<0.01) for patients with IE, RA, CD, AS, SS andOA. The lowest mean galactose level detected was onthe IgG from patients with IE (Figure 2). Mean IgGgalactose level of the SLE group was also significantlyreduced compared to the mean IgG galactose level ofthe controls (P<0.05), but for patients with JCA andUC this did not reach significance.

IgG GlcNAc content

In the disease groups IE, RA, CD, SLE, JCA, UC, ASand SS, the mean IgG GlcNAc detected was greaterthan that of the normal controls (Figure 3), reachingsignificance for patients with IE, RA, CD, JCA andUC (P<0.01). Although the mean levels of IgGGlcNAc found in samples from SLE, AS and SSpatients were higher than that found in normal IgG,this did not reach statistical significance. Significantlyless mean IgG GlcNAc was detected on samples fromOA patients than that of normal controls (P<0.01).

Relationship between galactose and GlcNAc

The relationship of galactose to GlcNAc varied foreach disease and was analysed as follows:

Figure 2. Galactose units detected on IgG from patients with various diseases and normal controls. Detection of galactose wasachieved using RCA I. Each individual sample was measured in triplicate. The solid horizontal line represents normal meanIgG galactose level. Dashed horizontal lines represent 95% confidence interval of the normal population. Mean values and95% CI are shown for each disease group.

Galactose: N-acetylglucosamine relationship on IgG in disease 81

The slopes

In OA and SS IgG the two sugars were found to bepositively associated (indicating that when IgG is richin galactose it is also rich in GlcNAc, or conversely,when the IgG is low in galactose, it is also low inGlcNAc). However with r=0.11 and r=0.26 respect-ively they were not found to correlate significantly(Figure 4). It is important to note that the slopeproduced by the results of the SS IgG was the mostpositive slope of all the groups studied.

Anegative slope (indicating that when IgG is rich ingalactose it has low levels of GlcNAc, or conversely,when the IgG is low in galactose, it has high levels ofGlcNAc), was found for the two sugars when analys-ing IgG from patients in all the other disease groups(Figure 4). The galactose detected in IgG from RA andJCA patients gave a significant negative associationwith the GlcNAc detected on the same IgG samples(r=−0.48, P<0.01 and r=−0.63, P<0.05 respectively).There was a tendency for the two sugars to benegatively related in CD IgG (r=−0.45) but this didnot quite reach significance. Inverse associationsbetween the two sugars on IgG from the AS, SLE, UCand IE groups were found but they were not signifi-cant. In the normal individuals, the relationshipbetween the two sugars gave a slight positive slopebut their expression was not interrelated; despitewide-ranging galactose levels, the GlcNAc levelsremain fairly constant.

Difference between slopes

The differences in the relationship of galactose toGlcNAc detected on individual IgG samples (asshown in Figure 4) between the various diseasegroups was analysed by comparing the slopes of theserelationships (Table 1). The RA IgG, JCA IgG and CDIgG results produced slopes which were significantlydifferent from that obtained when analysing thehealthy group IgG. There was also a significant differ-ence between the slopes resulting from the analysis ofthe RA and JCA IgG compared to that of the SS group(data not shown).

Disease activity associations

Associations between the sugar lectin profile anddisease activity were analysed using appropriate

Figure 3. GlcNAc units detected on IgG from patients with various diseases and normal controls. Detection of GlcNAc wasachieved using BSII. Each individual sample was measured in triplicate. The solid horizontal line represents normal meanIgG GlcNAc level. Dashed horizontal lines represent 95% confidence interval of the normal population. Mean values and 95%CI are shown for each disease group.

Figure 4. Analysis of galactose and GlcNAc on IgG. Galac-tose was detected by the use of RCA I and GlcNAc wasdetected using BSII. Negative slopes for the relationshipbetween the detected galactose and GlcNAc levels areshown in grey; positive slopes for the relationship betweenthe detected galactose and GlcNAc levels are shown inblack. The dashed line represents the relationship betweenthe two sugars in the normal population.

82 A. Bond et al.

activity indices on all patients for whom sufficientinformation was available. We found an associationfor the RA group between disease activity and IgG%GO, P<0.05, where a high %GO value was signifi-cantly associated with active disease. Associationswere also found between IgG galactose and GlcNAcunits and disease activity, where low levels of galac-tose P<0.05 and, conversely, high levels of GlcNAcwere significantly associated with active disease,P<0.05. No correlations were found between totalrheumatoid factor levels (IU) and galactose levels,GlcNAc levels or %GO. This is consistent with otherstudies where serum rheumatoid factor levels did notcorrelate with the percentage of oligosaccharidechains lacking galactose [19, 22].There was no association found between IgG %GO,

galactose or GlcNAc units and disease activity in theCD or UC groups, although 3/4 individuals that hadhigh IgG %GO levels had very highly active disease asreflected by their CDAI value.The patients with endocarditis were all hospitalized

at the time of serum sampling and were receivingantibiotic therapy. In each case where organisms werecultured from blood, they were of the Streptococcalspecies. All patients showed raised ESR and CRPlevels, and 11 out of 12 had normal or raised C3 andC4 levels. Eight out of twelve had raised levels ofimmune complexes detected by a PEG assay [23]. Tenout of twelve were positive for rheumatoid factor.There was no clear correlation between any of theseserological criteria and IgG %GO, galactose or Glc-NAc levels.As there was a wide scatter of results for the SS

group IgG galactose and GlcNAc measurements, wewere very interested in possible associations withvarious serological markers. We were, however,unable to identify any association. No %GO or IgGgalactose/GlcNAc associations with disease activityin SLE, JCA, AS and OA could be determined, due toan insufficient number of inactive and very activeindividuals, on retrospective analysis.It is of note that the JCA IgG sample which gave a

%GO value less than the lower 95% confidence limitof the normal group was from a patient who on

examination of the clinical information was found tobe in complete remission of symptoms at the time oftaking the sample. Similarly, the two SLE IgG sampleswhich also had lower %GO values than the lowerconfidence limit of the normal group, were found tobe from moderately active individuals. There were nodistinguishing clinical parameters that differentiatedbetween the IE patient with the highest IgG %GOvalue and the rest of the IE group.Generally, there was no relationship found between

IgG yield and glycosylation pattern for the diseasegroups. However, there was a small positive relation-ship between the IgG yield and IgG GlcNAc level inthe SLE group, i.e. high levels of IgG GlcNAc werefound in the samples which yielded high IgG levels. Inthe AS group, the pattern was completely reversed, i.e.high levels of IgG GlcNAc were found in the sampleswhich yielded low IgG levels. The significance of thesefindings requires further investigation.

Discussion

Serum IgG from patients with RA, and from variousanimal models of rheumatoid disease, lacks galactoseas the terminal sugar on the branched bi-antennaryoligosaccharides of the Fc region [1]. Previously, it hadbeen thought that the two possible terminal sugarswould vary inversely with each other; however, theresults of our recent study [7] together with thepresent observations indicate that the expression ofthese sugars changes independently in certain inflam-matory diseases.Our results clearly show that while galactose levels

in normal IgG varied considerably, the levels of Glc-NAc did not change in parallel. In order to study theseobservations further we have analysed, in additionto conventional %GO, the relationship between thedetected galactose and GlcNAc in both inflammatorydiseases and controls.In agreement with previous findings, we found the

mean %GO for RA, CD [3] and JCA IgG [2] weresignificantly raised above that of our normal popu-lation. We have now also shown IE IgG to have

Table 1. 95% CI for the difference between the slopes of the regression lines (for the galactose GlcNAcrelationship, Figure 4), of the disease groups analysed compared with that of the normal population.*indicates those CI where the slopes of the disease groups are significantly different from that of the normalpopulation

Diagnosis 95% Confidence interval Degrees of freedom

IE −0.640 units to 0.181 units 110RA −0.682 units to −0.190 units* 127Crohn’s disease −0.831 units to −0.033 units* 115SLE −0.988 units to 0.136 units 110JCA −1.390 units to −0.261 units* 109UC −1.148 units to 0.015 units 114Ankylosing spondylitis −0.771 units to 0.238 units 115Sjögren’s syndrome −0.029 units to 0.350 units 135OA −0.248 units to 0.376 units 111

Galactose: N-acetylglucosamine relationship on IgG in disease 83

increased mean %GO compared with that of thenormal IgG. This was of particular interest since thiscondition is usually associated with either strepto-coccal or staphylococcal infections, and IgG %GOfrom patients with other bacterial infections such asklebsiella and leprosy have also been found to beraised above normal [3]. We found OA IgG to havesignificantly lowered mean %GO levels comparedwith that of normal controls, which appears to differfrom a previous study based on a biochemical method[1]. There was no significant difference between themean SLE, UC, AS or SS IgG %GO when compared tothat of normal IgG, in agreement with a previousstudy [3].An age association has previously been shown with

IgG %GO in normal individuals [24], but we failed toshow a similar age association in our analysis of IgG%GO in both this study (100 normal individuals) andour earlier studies (measuring 32 and 100 normalindividuals) [25, 26]. It is possible that the distributionof %GO is fairly flat in the age range we haveexamined (19–64 years) with increases in %GO at theextremes as considered by Parekh (1–70 years).In our analysis of the individual terminal sugars,

we found decreased mean IgG galactose in all diseasegroups compared with that of normal IgG. Interest-ingly, this decrease in mean IgG galactose level doesnot always parallel an increase in mean GlcNAc levelsin the various disease groups. For example, OA IgG,where there is a decrease in both mean galactose andmean GlcNAc levels and in SS where despite a signifi-cant decrease in mean galactose levels there is nosignificant change in mean GlcNAc levels comparedto that of normal IgG. These findings reinforce theimportance of measuring the levels of both sugars(galactose and GlcNAc) individually as they showthat the level of one sugar cannot always be predictedbased on the presence/absence of the other. Thechanges in IgG sugars are much more complex thanwe first imagined and this had led us to propose thatdifferent glycosylation patterns may be present indifferent diseases. Indeed, the calculation of ‘putative’%GO may not always accurately reflect the actualpercentage of chains lacking galactose. RA IgG may beused as a standard when the unknown IgG samplesare from RA patients in order to calculate ‘putative’%GO, but our results clearly indicate that this IgGwould be unsuitable for the %GO prediction of IgGfrom other disease groups. As a result of these find-ings, we have initiated a collaborative internationaltrial involving different assays and laboratories toattempt to further investigate these complex glyco-sylation changes.To enhance our examination of these independent

changes of galactose in relation to GlcNAc, were-analysed our data by looking at the slope of therelationship between the detected galactose and Glc-NAc in the various disease groups. This form ofanalysis takes into account the population size andvariation in calculating the confidence intervalsbetween the slopes. Statistical analysis showed thatthe slopes from the results of the three patient groups,RA, JCA and CD, were significantly different from the

slope of the normal group. Galactose and GlcNAc inIgG from these disease groups are inversely related, adirect indication that as the level of galactosedecreases, there is, as expected, an increase in the levelof the next sugar along the oligosaccharide chain,GlcNAc. This inverse relationship is also observedwith UC, AS, IE and SLE IgG sugars, but is notsignificant in comparison with that of normal IgG.Interestingly, however, in the SS and OA IgGs, the

two sugars have, in fact, a slight positive associationarising from parallel changes in both sugars. Thispositive association can be explained in one of threeways: increase in the overall glycosylation, differentialglycosylation or variations in the level of bisectingGlcNAc. Increased overall glycosylation could beinfluenced by the presence of other glycosylation sitesin addition to the one in the Fc region, and would bedependent on the presence of the peptide glycosyl-ation motif Asn-X-Ser/Thr in the Fab domain. Thismay not be applicable in this instance, however, as wehave no evidence to indicate an increase in totalglycosylation, as demonstrated by the sum of thegalactose and GlcNAc content in our SS IgG samples.Differential glycosylation has been demonstrated [27]in other glycoproteins and could be due to variationsresulting from differential processing [28] of the Fabsugars. It would be possible, therefore, to have un-altered Fc sugars while Fab sugars vary. Thus Fabsugar variation may be related to particular, disease-associated, antibody specificities. Finally, as bisectingGlcNAc is attached to the C4 position of the â-mannosyl residue between the bi-antennary oligo-saccharide arms, exposed bisecting GlcNAc may bedetected while, at the same time, terminal GlcNAccould be masked by galactose, thus maintainingapparent constancy in GlcNAc expression.It is also feasible that the observed pattern of lectin

binding may be due to targeting of particular oli-gosaccharide structures. For example, we have evi-dence to indicate that BSII binds very effectively tocompletely agalactosylated IgG (using monoclonalIgG, a gift from Professor Roy Jefferies, BirminghamUniversity). Thus, a small proportion of fully agalac-tosylated antibodies within the total serum IgG couldresult in enhanced BSII binding while still maintain-ing normal galactosylation over the majority of themolecules in the serum. In this respect, it is interestingthat isolated, specific antibodies have been found tohave very low galactosylation, in both autoimmunedisease and normal immune responses [29].It is intriguing to note that the same reciprocal

relationship found between galactose and GlcNAc inRA IgG, was also found in IgG from diseases whichare associated with bacterial infections. In RA, Proteushas been implicated [30], CD may be associated withMycobacterium paratuberculosis [31], and there is evi-dence to suggest that AS is linked with klebsiella [32].A reciprocal sugar relationship was also seen in IE,where the infection is most frequently associated witheither streptococcal or staphylococcal species [33]. Onthe other hand, SS IgG glycosylation is very differ-ent from that found in RA, where a parallel relation-ship between galactose and GlcNAc is found. It is

84 A. Bond et al.

interesting that viruses, and not bacteria, are believedto be associated with this disease, namely cytomega-lovirus [34] and Epstein-Barr virus [35]. This raises thepossibility that the RA-like or SS-like carbohydrateprofiles found in our present study may be the conse-quence of a bacterial or viral infection. Evidence insupport of this comes from a previous study showingthat the galactosylation of IgG from patients with viralassociated infections, such as parvovirus, mumps,HIV, rubella and glandular fever, may be similar tothat of normal IgG (although this was in terms of%GO, and the size of the groups was very small) [3].In RA patients, changes in IgG glycosylation are

related to disease activity, and it is interesting that wehave previously shown that IgG immune complexesfrom these patients are especially enriched with aga-lactosyl IgG [36]. These complexes may well containIgG antiglobulins which are self-associating. This maybe a reflection of preferential agalactosylation ofautoantibodies as shown in autoimmune haemolyticanaemia [29]. We are now examining isolated auto-antibodies from SS patients to see if there is pref-erential abnormal glycosylation of autoantibodiescompared with total IgG.In conclusion, our data reveal the importance of

measuring sugar moieties as separate entities becauseeach disease group IgG has a unique carbohydrateprofile and furthermore indicates the presence ofdisease specific glycosylation changes which may beuseful diagnostic or prognostic tools.

AcknowledgementsThis research has been supported by the Arthritis andRheumatism Council. We thank Professor M. Bland,Medical Statistics, St George’s Hospital MedicalSchool, for his statistical advice, Deborah M. Johnson,Department of Rheumatology, St George’s HospitalMedical School, for her help with patient details,Pauline Rudd, Glycobiology Unit, Department of Bio-chemistry, University of Oxford, UK, for her gift ofGO standards, and Dr E. Wawrzynczak of the Instituteof Cancer Research: Royal Cancer Hospital who sup-plied us with the Ricinus communis agglutinin.

References1. Parekh R.B., Dwek R.A., Sutton R.A., Fernandes D.L.,

Leung A., Stanworth D., Rademacher T.W., MizuochiT., Taniguchi T., Matsuta K., Takeuchi F., Nagano Y.,Miyamoto T., Kobata A. 1985. Association ofrheumatoid arthritis and primary osteoarthritis withchanges in the glycosylation pattern of total serumIgG. Nature 316: 452–457

2. Parekh R.B., Ansell B.M., Isenberg D.A., Roitt I.M.,Dwek R.A., Rademacher T.W. 1988. Galactosylation ofIgG-associated oligosaccharides: reduction in patientswith adult and juvenile onset rheumatoid arthritis andrelation to disease activity. Lancet i: 966–969

3. Parekh R., Isenberg D., Rook G., Roitt I., Dwek R.,Rademacher T. 1989. A comparative analysis ofdisease-associated changes in the galactosylation ofserum IgG. J. Autoimmunity 2: 101–114

4. Axford J.S. 1994. Glycosylation and rheumatic disease:More than icing on the cake. J. Rheumatol. 21:1791–1795

5. Bond A., Cooke A., Hay F.C. 1990. Glycosylation ofIgG, immune complexes and IgG subclasses in theMRL-lpr/lpr mouse model of rheumatoid arthritis.Eur. J. Immunol. 20: 2229–2233

6. Bond A., Alavi A., Hay F.C., Axford J.S. 1993.Immunoglobulin G glycosylation changes occur inrheumatoid arthritis and primary Sjögren’s syndromebut due to different mechanisms. Arthritis Rheum. 36:(suppl) S242

7. Bond A., Alavi A., Axford J.S., Youinou P., Hay F.C.1996. The relationship between exposed galactose andN-acetylglucosamine residues on IgG in rheumatoidarthritis, juvenile chronic arthritis and Sjögren’ssyndrome. Clin. Exp. Immunol. 105: 99–103

8. Arnett F.C., Edworthy S.M., Block D.A., McShane D.J.,Fries J.F., Cooper M.S., Healy L.A., Kaplan S.R., LiangM.H., Luthra H.S., Medsger T.A. Jr, Mitchell D.M.,Neustadt D.H., Pinals R.S., Schaller J.G., Sharp J.T.,Wilder R.L., Hunder G.G. 1988. The AmericanRheumatism Association 1987 revised criteria for theclassification of rheumatoid arthritis. Arthritis Rheum.31: 315–324

9. Tan E.M., Cohen A.S., Fries J.F., Masi A.T., McShaneD.J., Rothfield N., Schaller J.G., Talal N., WinchesterR.J. 1982. The 1982 revised criteria for the classificationof systemic lupus erythematosus. Arthritis Rheum. 25:1271–1277

10. Bennett P.H., Burch T.A. 1967. New York symposiumon population studies in the rheumatic diseases: newdiagnostic criteria. Bull. Rheum. Dis. 17: 453–458

11. Lennard-Jones J.E. 1989. Classification of inflammatorybowel disease. Scand. J. Gastroenterol. Suppl. 170: 2–6

12. Vitali C., Bombardieri S., Moutsopoulos H.M.,Balestrieri G., Bencivelli W., Bernstein R.M., BjerrumK.B., Braga S., Coll J., de Vita S., Drosos A.A.,Ehrenfeld M., Hatron P.Y., Hay E.M., Isenberg D.A.,Janin A., Kalden J.R., Kater L., Konttinen Y.T.,Maddison P.J., Maini R.N., Manthorpe R., Meyer O.,Ostuni P., Pennec Y., Prause J.U., Richards A., SauvezieB., Schiödt M., Sciuto M., Scully C., Shoenfeld Y.,Skopouli F.N., Smolen J.S., Snaith M.L., Tishler M.,Todesco S., Valesini G., Venables P.J.W., Wattiaux M.J.,Youinou P. 1993. Preliminary criteria for theclassification of Sjögren’s syndrome. Results of aprospective concerted action supported by theEuropean community. Arthritis Rheum. 36: 340–347

13. Truelove S., Witts L. 1960. Cortisone andcorticotrophin in ulcerative colitis. BMJ 1: 464–467

14. Best W.R., Becktel J.M., Singleton J.W., Kerr F. Jr. 1976.Development of a Crohn’s disease activity index.National Co-operative Crohn’s disease study.Gastroenterology 70: 439–444

15. Harvey R.F., Bradshaw J.M. 1980. A simple index ofCrohn’s disease activity. Lancet i: 514

16. Bond A., Jones M.G., Hay F.C. 1993. Human IgGpreparations isolated by ion-exchange chromatographydiffer in their glycosylation profiles. J. Immunol.Methods 166: 27–33

17. Hudson L., Hay F.C. 1989. Single radial immuno-diffusion (SRID). In Practical Immunology, 3rd edn. L.Hudson, F.C. Hay, eds. Blackwell ScientificPublications, Oxford. pp. 230–233

Galactose: N-acetylglucosamine relationship on IgG in disease 85

18. Hardy R.R. 1986. Purification and characterisation ofmonoclonal antibodies. In Handbook of ExperimentalImmunology. Volume I, Immunochemistry, 4th edn.D.M. Weir, ed. Blackwell Scientific, Oxford

19. Sumar N., Bodman K.B., Rademacher T.W., DwekR.A., Williams P., Parekh R.B., Edge J., Rook G.,Isenberg D.A., Hay F.C., Roitt I.M. 1990. Analysis ofglycosylation changes in IgG using lectins. J. Immunol.Methods 131: 127–136

20. Bland M. 1995. Comparisons of means after analysis ofvariance. In An introduction to medical statistics, 2ndedn. M. Bland, ed. Oxford University Press, New York.pp. 175–176

21. Gardner M.J., Altman D.G. 1989. Calculatingconfidence intervals for regression and correlation. InStatistics with confidence, 1st edn. M.J. Gardner, D.G.Altman, eds. The Universities Press, London. pp 34–49

22. Tomana M., Schrohenloher R.E., Koopman W.J.,Alarcón G.S., Paul W.A. 1988. Abnormal glycosylationof serum IgG from patients with chronic inflammatorydiseases. Arthritis Rheum. 31: 333–338

23. Kerr M.A., Wilton E., Naama J.K., Whaley K. 1986.Circulating immune complexes associated withdecreased complement mediated inhibition of immuneprecipitation in sera from patients with bacterialendocarditis. Clin. Exp. Immunol. 63: 359–366

24. Parekh R., Roitt I., Isenberg D., Dwek R., RademacherT. 1988. Age-related galactosylation of the N-linkedoligosaccharides of human serum IgG. J. Exp. Med.167: 1731–1736

25. Axford J.S., Sumar N., Alavi A., Isenberg D.A., YoungA., Bodman K.B., Roitt I.M. 1992. Changes in normalglycosylation mechanisms in autoimmune rheumaticdisease. J. Clin. Invest. 89: 1021–1031

26. Hunt K., Bond A., Alavi A., Mrazek I., Sharrock C.,Hay F.C., Axford J.S. 1991. In healthy individualsincreased agalactosylated IgG is found in males andthose with HLA DR 2 and 7 gene expression. Br. J.Rheum. 30: 67

27. Zamze S.E., Ashford D.A., Wooten E.N., RademacherT.W., Dwek R.A. 1991. Structural characterization of

the asparagine-linked oligosaccharides fromTrypanosoma brucei type II and type III variant surfaceglycoproteins. J. Biol. Chem. 266: 20244–20261

28. Alavi A., Axford J.S. 1996. The glycosyltransferases. InAbnormalities of IgG glycosylation and immunologicaldisorders, 1st edn. D.A. Isenberg, T.W. Rademacher,eds. John Wiley, Chichester. pp. 149–169

29. Thompson S.J., Leader K.A., Lastra G.C., Elson C.J.1996. The role of agalactosyl IgG in rodent models ofautoimmune disease and in the pathogenesis ofrheumatoid arthritis. In Abnormalities of IgGglycosylation and immunological disorders, 1st edn. D.A.Isenberg, T.W. Rademacher, eds. John Wiley,Chichester. pp. 131–148

30. Wilson C., Ebringer A., Ahmadi K., Wrigglesworth J.,Tiwana H., Fielder M., Binder A., Ettelaie C.,Cunningham P., Joannou C., Bansal S. 1995. Sharedamino acid sequences between majorhistocompatibility complex class II glycoproteins, typeXI collagen and Proteus mirabilis in rheumatoidarthritis. Ann. Rheum. Dis. 54: 216–220

31. Sanderson J.D., Moss M.T., Tizard M.L.,Hermon-Taylor J. 1992. Mycobacterium paratuberculosisDNA in Crohn’s disease tissue. Gut 33: 890–896

32. Ebringer A. 1992. Ankylosing spondylitis is caused byKlebsiella. Spondyloarth. 18: 105–121

33. MacSween R.N.M., Whaley K. 1992. Cardiovascularsystem. In Muir’s textbook of pathology, 13th edn.R.N.M. MacSween, K. Whaley, eds. Edward Arnold,London. pp. 440–523

34. Shillitoe E.J., Daniels T.E., Whitcher J.P., Strand D.V.,Talal N., Greenspan J.S. 1982. Antibody tocytomegalovirus in patients with Sjögren’s syndromeas detected by enzyme linked immunosorbent assay.Arthritis Rheum. 25: 260–265

35. Wolf H., Haus M., Wilmes E. 1984. Persistence ofEpstein-Barr virus in the parotid gland. J. Virol. 51:795–798

36. Bond A., Kerr M.A., Hay F.C. 1995. Distinctoligosaccharide content of rheumatoid arthritis-derivedimmune complexes. Arthritis Rheum. 38: 744–749