Annals of the Rheumatic Diseases 1995; 54: 216-220

CONCISE REPORTS

Shared amino acid sequences between majorhistocompatibility complex class II glycoproteins,type XI collagen and Proteus mirabilis inrheumatoid arthritis

C Wilson, A Ebringer, K Ahmadi, J Wrigglesworth, H Tiwana, M Fielder, A Binder,C Ettelaie, P Cunningham, C Joannou, S Bansal

Division ofBiomolecularSciences,Immunology Section,King's College,London W8 7AH,United KingdomC WilsonA EbringerK AhmadiJ WrigglesworthH TiwanaM Fielder

Department ofComputingP CunninghamDepartment ofPharmacyS BansalDepartment ofRheumatology,Middlesex Hospital,UCL School ofMedicine,London,United KingdomA EbringerDepartment ofRheumatology,Lister Hospital,Stevenage, Herts,United KingdomA Binder

Department ofChemistry andBiochemistry,The Royal FreeHospital,London,United KingdomC EttelaieDepartment ofBiochemistry,UMDS School ofMedicine,London,United KingdomC JoannouCorrespondence to:Dr Alan Ebringer.Accepted for publication17 October 1994

AbstractObjectives-To show molecular similaritybetween two sequences ofProteus mirabilis(haemolysin-ESRRAL; urease-IRRET)with HLA-DR antigens (EQRRAA) whichare associated with rheumatoid arthritis(RA) and type XI collagen (LRREI),respectively; and, in patients with RA, tomeasure levels of antibody against a16-mer synthetic peptide containing theESRRAL sequence, and the haemolysinand urease proteins ofProteus mirabilis.Methods-The homologous sequencesEQRRAA and ESRRAL were modeliedwith Alchemy III, using the crystallinestructure of DRBl*0101 (HLA-DRl).Sera from 40 patients with RA, 30 withankylosing spondylitis (AS), and 30controls were tested against syntheticESRRAL peptide and the haemolysin ofProteus mirabilis by enzyme linkedimmunosorbent assay. Similar tests werealso carried out on sera from 20 patientswith RA, 40 with AS, and 15 controls,against Proteus mirabilis urease.Results-Molecular modelling of thehomologous sequences ESRRAIJEQRRAAand IRRET/LRREI showed stereochemicalsimilarities. Antibodies to the 16-mersynthetic peptide containing the ESRRALsequence, the haemolysin, and ureaseproteins were significantly increased inRA patients compared with AS patients(p < 0.001) and healthy controls (p < 0.001).No such increases were observedwith three control peptides includingthe EDERAA sequence of DRB1*0402(HLA-DR4/Dw1O), the haemolysin proteinsof Streptococcus pyogenes and Vibrioparahaemolyticus, and the urease ofBacillus pasteurii.Conclusion-The additive effect of theimmune responses to the two Proteusmirabilis antigens, haemolysin (ESRRAL)and urease (IRRET), could be relevant inthe aetiopathogenesis ofRA.

(Ann Rheum Dis 1995; 54: 216-220)

The association between rheumatoid arthritis(RA) and some subtypes of HLA-DR4 is well

established:' a particular region of the DR3 1chain, from position 70-74, encoding aminoacids Gln-Arg-Arg-Ala-Ala (QRRAA) andfound in DRBl*0101 (DR1) and DR4 (sub-types DRB1*0404 (Dw14) and DRBl*0405(Dw15)), has been identified as the molecularsequence responsible for susceptibility to RA.2The sequence closely resembles that found inDRB 1*0401 (DR4/Dw4) individuals, therebeing only one conservative substitution atposition 71, from arginine to lysine (QKRAA);glutamic acid (E) occupying position 69 iscommon to all DR, 1 molecules. Furthermore,the EQRRAA sequence is also found inDRB1*1402 (DR6/Dw16) positive YakimaIndians with RA. Our group has reported anamino acid homology between an outermembrane haemolysin protein of Proteusmirabilis and the susceptibility sequence inHLA-DR1 and DR4 subtypes associated withRA.3 Because RA is a systemic disorder andcartilage destruction is a major feature of thedisease, a search of the protein database wasmade for any sequence homology betweenProteus mirabilis and collagen. An amino acidhomology was identified between Proteusmirabilis urease (IRRET), amino acid residues337-3414 and o2(XI) collagen (LRREI)residues 421-425.5 Type XI collagen is acomponent of hyaline cartilage which iscomposed of three different polypeptide sub-units aol(XI), a2(XI), ot3(XI). Early studiesfrom our laboratory showed that anti-HLA-DR4 tissue typing sera bound significantly toProteus mirabilis compared with non-DR4tissue typing sera; no such interaction wasfound with Escherichia coli.6 In addition, wehave reported an increase in Proteus antibodiesin RA patients7 which has been confirmed bytwo independent groups, from Dublin8 andNewcastle.9 The hexamer sequence ESRRALand the pentamer IRRET of Proteus mirabilisare both hydrophilic, which suggests that theymay be immunogenic.The present study was undertaken to

compare spatial configurations of EQRRAA/ESRRAL, and LRREI/IRRET sequences forsimilarity using molecular modelling, and todetermine if antibodies to ESRRAL and to thehaemolysin and urease proteins were present inRA patients.

216

on June 9, 2021 by guest. Protected by copyright.

http://ard.bmj.com

/A

nn Rheum

Dis: first published as 10.1136/ard.54.3.216 on 1 M

arch 1995. Dow

nloaded from

http://ard.bmj.com/

Shared amino acid sequences

Patdents and methodsPATIENTSSera were collected from active (erythrocytesedimentation rate (ESR) > 15 mm/Ist h)RA patients attending the RheumatologyDepartment at the Lister Hospital, Stevenage,Herts, and active AS patients attending the ASResearch Clinic at the Middlesex Hospital.The diagnosis of RA was according to theAmerican Rheumatism Association criteriaand that ofAS by the New York criteria.

In the first study (antibodies againstESRRAL and the haemolysin protein), serawere from 40 patients with RA (10 male/30 female; mean age 59 (range 37-77) years;mean ESR 46-2 (SE 4 0) mmllst h), 30 ASpatients (23 males/seven females; mean age49 (27-76) years; mean ESR 30-1 (5 2) mm/1st h), and 30 healthy control subjects (15 male/15 female; mean age 23 (21-27) years).In the second study (antibodies against the

urease protein), patients and controls weredifferent from those in the first study. Theycomprised 20 patients with RA (five male/15 female; mean age 56 (20-75) years; meanESR 59-8 (7 7) mmllst h), 40 AS patients(29 males/i1 females; mean age 45 (21-76)years; mean ESR 25-9 (3-6) mm/lst h), and15 healthy control subjects (mean age 29(21-52) years).

MOLECULAR MODELLINGComparison of space filling models of a pre-dicted ESRRAL sequence of Proteus haemolysinand EQRRAA sequence within DRB1*0101(HLA-DR1) from known crystallographicstructure was carried out to study spatialconfiguration. The ESRRAL model wasconstructed as a helix with torsional anglescorresponding to those observed in the knownEQRRAA sequence and the two sequenceswere superimposed (root mean square (RMS) =0 046). The ESRRAL and EQRRAAsequences were also compared with thesequence EDERAA, present in DRB1*0402(HLA-DR4/Dw10), which is not associatedwith RA.Comparison of space filling models was

made of the predicted IRRET of Proteusmirabilis urease and LRREI of ot2(XI) collagen.Both sequences were constructed as helicesand the two sequences were superimposed(RMS = 0-0 11). Structures were modelledusing Alchemy III (Tripos ASSOC Inc, StLouis, USA).

ELISA STUDIES WITH SYNTHETIC PEPTIDESPeptides were prepared by solid phasesynthesis and analysed for purity by highperformance liquid chromatography. Thetest peptide was LGSISESRRALQDSQR(1 6-mer), which represents amino acidresidues 27-42 of Proteus mirabilis haemoly-sin (hpm B);10 three control peptides,SQKDILEDERAAVDTY (16-mer) DRB1*0402(HLA-DR4/Dw1O), YASGASGAS (9-mer),and DAHKSEVAHRFLDLGEENFKALVL(24-mer), were used to exclude non-specific

binding. Sera were tested against the Proteusmirabilis haemolysin peptide and controlpeptides by enzyme linked immunosorbentassay (ELISA). Briefly, polystyrene microtitreplates (Dynatech, Billingshurst, UK) werecoated with the synthetic peptide (5 0 ,ug/well)overnight at 4°C. After absorption and washingwith phosphate buffered saline (PBS)-Tween,the plates were saturated with 1% bovineserum albumin (BSA)-PBS-Tween andincubated for one hour at 37°C, followed byfurther washing with PBS-Tween. Serumsamples (200 RI at 1/50 dilution in PBS-Tween) were added and the plates incubatedfor 90 minutes at 37°C, followed by washingwith PBS-Tween. Peroxidase conjugated rabbitanti-human class specific IgG, IgA or IgM(DAKO Ltd, Bucks, UK) diluted 1/500 inPBS-Tween was added and the platesincubated for 90 minutes at 37°C then washed.After exposure of the enzyme to substrate (2,2'-azinobis (3-ethylbenz-thiazoline-6-sulphonicacid), Sigma Chemical Company Ltd, Dorset,UK) at room temperature for 20 minutes, thereaction was stopped with sodium fluoride andthe optical density (OD) measured at wave-length 630 nm. All assays were carried out induplicate and under code, so that the status ofeach serum sample under investigation was notknown to the tester.

PURIFICATION AND ASSAY OF THE 63 KDAOUTER MEMBRANE HAEMOLYSIN PROTEINPurification of the 63 kDa outer membranehaemolysin protein was carried out asoriginally described for E coli."1 Sera obtainedfrom the same subjects as above, were testedagainst Proteus mirabilis haemolysin (63 kDa)protein, and against two control haemolysinproteins, obtained from Streptococcus pyogenesand Vibrio parahaemolyticus (Sigma ChemicalCompany Ltd) by ELISA. Briefly, polystyrenemicrotitre plates (Dynatech) were coated withthe haemolysin protein (2X0 ,ug/well) overnightat 4°C. Serum samples were diluted 1/200 inPBS-Tween and the plates saturated with 0 1%BSA-BPS-Tween. The remainder of the assayprocedure was as previously described for thesynthetic peptides.

PURIFICATION AND ASSAY OF 280 KDA UREASEPROTEINPurification of the 280 kDa urease proteinwas by the method of Jones and Mobley.'2Sera obtained from different subjects thanin the first study were tested againstProteus mirabilis urease protein and Bacilluspasteurii urease (Sigma Chemical Company)by ELISA. Polystyrene microtitre plates(Dynatech), were coated with the ureaseprotein (2 ptg/well) and assays carried out aspreviously described.

ResultsMolecular modelling of the two structuresESRRAL and EQRRAA showed a commonsurface of homologous residues. The positions

217

on June 9, 2021 by guest. Protected by copyright.

http://ard.bmj.com

/A

nn Rheum

Dis: first published as 10.1136/ard.54.3.216 on 1 M

arch 1995. Dow

nloaded from

http://ard.bmj.com/

Wilson, Ebringer, Ahmadi, et al

32, 34, 35 and 69, 71, 72 form a motifcommon to HLA-DR and Proteus mirabilisoccupying the same stereochemical space. How-ever, the EDERAA sequence of DRB1*0402(HLA-DR4/Dw10) which is not associatedwith RA, differed from both the ESRRAL andEQRRAA motifs (fig 1). Molecular modellingof the two sequences IRRET and LRREI alsoshowed a common surface of homologousresidues. The positions 338, 339, 340 and 421,422, 423, form a motif common to type XIcollagen and Proteus mirabilis urease occupyingthe same stereochemical space (fig 1). Thecommon surface of homologous residuesobserved in the sequences ESRRAIIEQRRAA

and IRRET/LRREI, which may be involved inimmune interactions, consists of two positivelycharged arginines and one negatively chargedglutamic acid.

Antibodies to Proteus mirabilis haemolysinpeptide LGSISESRRALQDSQR of the IgGclass were significantly increased in RA patients(mean 0-263 (SE 0-011) OD units) comparedwith AS patients (0-162 (0-012) OD units)(t= 5-98, p < 0001) and healthy controls(0K158 (0-014) OD units (95% confidencelimits (CL) of mean ±0 026) (t= 5 90,p < 0-001) (fig 2). The difference between thetitres in AS patients and healthy controls testedagainst the haemolysin peptide was not

:rw,iuX.-} _

V..R?E S R-R .r i;; 1i p.

Lets4 21'

X4 I(XI) Coi1hlen"| oetu 1. H

Figure 1 Comparison ofspacefilling models. A: ESRRAL sequence ofProteus mirabilis haemolysin predictedfromknown crystallographic structure; B: EQRRAA sequence within DRB1 *0101 (HLA-DR1), predictedfrom knowncrystallographic structure; C: EDERAA sequence ofDRB1 *0402 (HLA-DR4/Dw10); D: predicted LRREI sequence ofa2(XI) collagen; E: predicted IRRETsequence ofProteus mirabilis urease.

218

:..4.4 -1 .". k. 4 1.)R I. . 4 1:i,;.4 T. -.,zt'N .." I .-, Jh .:,.,

on June 9, 2021 by guest. Protected by copyright.

http://ard.bmj.com

/A

nn Rheum

Dis: first published as 10.1136/ard.54.3.216 on 1 M

arch 1995. Dow

nloaded from

http://ard.bmj.com/

Shared amino acid sequences

0.50 -

0.40-

030-

Ch

00.20-

0.10-

DRB1*0402 Proteus peptide Proteus(EDERAA) (ESRRAL) haemolysin

*

** +

* *

*0*09~~~~4

*

*

**

* 0

* *0 *m * 994

* 9~~~~

C AS RA C AS RANS NS NS p<

0.001

9

0

* 9* 9

9

* 9* *

9*

* *9** 990

1.00

0.90

0.80

U,* 0.70

O 0.600

0.50

0.40

0.30

0.20

0 10

C AS RANS p<

0.001

Figure 2 Antibody titresfor IgG immunoglobulin incontrols (C) and patients with ankylosing spondylitis (AS)or rheumatoid arthritis (RA) when tested by ELISAagainst the synthetic peptides, EDERAA, ESRRAL andthe haemolysin protein ofProteus mirabilis. OD = Opticaldensity. Bars indicate means. NS = Not significantcompared with controls.

significant. There was no increase in IgA andIgM antibodies against the haemolysin peptide,and no significant reactivity by the RA patientsagainst the DRB1*0402 (HLA-DR4/Dwl0)EDERAA peptide (fig 2) or the other twocontrol peptides used in this study (data notshown).

Titres of antibody to Proteus mirabilis(63 kDa) protein of the IgG class weresignificantly increased in RA patients (0 304(00 13) OD units) compared with AS patients(0- 190 (0 075) OD units) (t=7-11,p

Wilson, Ebninger, Ahmadi, et al

sequences examined in the protein database.The high level of Proteus mirabilis antibodieswas not attributable to non-specific effects ofinflammation, because the AS patients also hadincreased ESR values, yet their levels of Proteusmirabilis antibodies were similar to those foundin healthy control subjects.The ESRRAL and IRRET sequences are

both hydrophilic, and in helical configurationswhere they may be involved in immune inter-actions. It has been reported that Proteusmirabilis may be isolated more frequently fromRA patients than from non-RA patients orhealthy controls,14 although some groups couldnot find such increased Proteus mirabilisisolation.15 The results of this study suggestthat RA patients have a significant increase inantibodies to a Proteus membrane haemolysinsequence containing a hexamer which ishomologous with the RA susceptibility sequence,and this could provide an explanation for the'shared epitope' hypothesis. Moreover, our RApatients also had a significant increase in anti-bodies to Proteus mirabilis urease containing apentamer sequence which is homologous witha sequence in type XI collagen. Furtherinvestigations are required to determine if anti-bodies to ESRRAIJIRRET and haemolysin/urease will react with HLA alleles and collagento produce tissue damage, thereby contributingto the pathogenesis of this disease.We thank the Trustees of the Middlesex Hospital for theirsupport. MF Acknowledges the support of the SERC.

1 Nepom G T, Byers P, Seyfried C E, et al. HLA genesassociated with rheumatoid arthritis. Identification ofsusceptibility alleles using oligonucleotide probes.Arthritis Rheum 1989; 33: 15-21.

2 Watanabe Y, Tokunaga K, Matsuki K, et al. Putative aminoacid sequence of HLA DRB chain contributing torheumatoid arthritis susceptibility. J Exp Med 1989; 169:2263-8.

3 Ebringer A, Cunningham P, Ahmadi K, Wrigglesworth J,Hosseini R, Wilson C. Sequence similarity between HLA-DR1 and DR4 subtypes associated with rheumatoidarthritis and Proteus/Serratia membrane haemolysins.Ann Rheum Dis 1992; 51: 1245-6.

4 Jones B D, Mobley H L T. Proteus mirabilis urease:Nucleotide sequence determination and comparison withjack bean urease.J Bacteriol 1989; 171: 6414-22.

5 Kimura T, Cheah K S E, Chan S D H, et al. The humancs2 (XI) collagen (COLIIA2) chain molecular cloning ofcDNA and genomic DNA reveals characteristics of afibrillar collagen with differences in genomic organisation.Biol Chem 1989; 264: 13910-6.

6 Ebringer A, Khalafpour S, Wilson C. Rheumatoid arthritisand Proteus: a possible aetiological association. RheumatolInt 1989; 9: 223-8.

7 Ebringer A, Ptaszynska T, Corbett M, et al. Antibodies toProteus in rheumatoid arthritis. Lancet 1985; ii: 305-7.

8 Rogers P, Hassan J, Bresnihan B, Feighery C, Whelan A.Antibodies to Proteus in rheumatoid arthritis. Br 7Rheumatol 1988; 27 (suppl 2): 90-4.

9 Deighton C M, Gray J W, Bint A J, Walker D J. Specificityof the Proteus antibody response in rheumatoid arthritis.Ann Rheum Dis 1992; 51: 1242-4.

10 Uphoff T S, Welch R A. Nucleotide sequencing of theProteus mirabilis calcium independent hemolysin genes(hpm A and hpm B) reveals sequence similarity with theSerratia marcescens hemolysin genes (shl A and shl B).J7Bacteriol 1990; 172: 1206-15.

11 Schnaitman C. Outer membrane proteins of Escherichiacoli OIII. Evidence that the major protein of Escherichiacoli OIII outer membrane consists of four distinct poly-peptide species. J Bacteriol 1974; 118: 442-53.

12 Jones B D, Mobley H L T. Genetic and biochemicaldiversity of ureases of Proteus, Providencia, andMorganella species isolated from urinary tract infection.Infect Immun 1987; 55: 2198-203.

13 Takeuchi F, Kosuge E, Matsuta K, et al. Antibody to aspecific HLA DR 1P sequence in Japanese patients withrheumatoid arthritis. Arthritis Rheum 1990; 33: 1867-8.

14 Ebringer A, Wilson C, Ahmadi K, Corbett M, Rashid T,Shipley M. Rheumatoid arthritis as a reactive arthritis toProteus infection: prospects for therapy. In: Machtey I,ed. Progress in Rheumatology, vol 5. Petach Tiqva, Israel:Rheumatology Service, Hasharon Hospital, GoldaMedical Centre, 1993; 77-83.

15 McDonagh J, Gray J, Sykes H, Walker D J, Bint A S,Deighton C M. Anti-Proteus antibodies and Proteusorganisms in rheumatoid arthritis: A clinical study. Br JRheumatol 1994; 33: 32-5.

220 on June 9, 2021 by guest. P

rotected by copyright.http://ard.bm

j.com/

Ann R

heum D

is: first published as 10.1136/ard.54.3.216 on 1 March 1995. D

ownloaded from

http://ard.bmj.com/