INFECTION AND IMMUNITY, June 1990, p. 1774-17810019-9567/90/061774-08$02.00/0Copyright C) 1990, American Society for Microbiology

Demonstration of Cross-Reactivity between Bacterial Antigens andClass I Human Leukocyte Antigens by Using Monoclonal

Antibodies to Shigella flexneriKRISTINA M. WILLIAMS* AND RICHARD B. RAYBOURNE

Division of Microbiology, Food and Drug Administration, Washington, D.C. 20204

Received 8 January 1990/Accepted 12 March 1990

Bacterial envelope proteins which share immunodeterminants with the human leukocyte antigen (HLA) classI histocompatibility antigen HLA-B27 may invoke spondyloarthritic disease through the process of molecularmimicry in patients expressing this phenotype. Monoclonal antibodies generated by the immunization ofBALB/c mice with envelope proteins of Shigella flexneri type 2a were tested for reactivity against culturedlymphoblastoid cell lines of defined HLA phenotype. As measured by flow microfluorometry, four immuno-globulin M monoclonal antibodies reacted preferentially with HLA-B27-positive lymphocytes (HOM-2, MM) ascompared with a B27-loss mutant line (1065) or cells lacking major histocompatibility complex class I antigen(Daudi, K562). Monoclonal antibodies also reacted with mouse EL-4 cells transfected with and expressing theHLA-B7 gene. Western immunoblot analysis of isolated enterobacterial envelopes demonstrated that thereactive epitope was present on bacterial proteins with an apparent relative molecular mass of 36 and 19kilodaltons. The structural basis for the cross-reactivity of bacterial antigen and HLA-B27 appeared to residein the portion of the HLA molecule that is responsible for allotypic specificity (amino acids 63 through 83), sincemonoclonal antibodies were positive by enzyme-linked immunosorbent assay with synthetic polypeptidescorresponding to this segment.

Most patients who experience arthritic sequelae afterenterobacterial infection express the human leukocyte anti-gen (HLA) class I histocompatibility antigen HLA-B27 (7).Reactive arthritis and Reiter's syndrome are often precededby gastroenteritis involving specific gram-negative patho-gens, most notably Shigella, Salmonella, Yersinia, andCampylobacter species (8). Although an antecedent episodeof infection is not characteristic of ankylosing spondylitis,chronic habitation of the bowel, particularly by Klebsiellapneumoniae, is suspected (21). Of the theories proposed toexplain the expression of a particular MHC haplotype andsusceptibility to postinfectious spondyloarthritic lesions,molecular mimicry emerges as a potential pathogenic mech-anism. Autoreactive humoral and cellular immune responsesmay be induced by persistent bacterial antigens sharingcross-reactive epitopes with the mimicked host protein, inthis case the HLA-B27 molecule.

Serologic cross-reactivity between class I antigens andarthritogenic bacteria has been demonstrated with polyva-lent rabbit and human antisera and monoclonal anti-HLA-B27 antibodies. How'ever, a review of previous studiesreveals descrepant results between laboratories, even whensimilar reagents and techniques are used. For example, usingsera from rabbits immunized with lymphocytes from HLA-B27-positive subjects, Welsh et al. (42) demonstrated immu-nologic cross-reactivity against several gram-negative en-teric pathogens, including K. pneumoniae and Yersiniaenterocolitica; however, Archer (3), using the same antisera,failed to confirm these results. To overcome problems asso-ciated with high levels of nonspecific binding inherent to theuse of whole antisera, several groups, including our own,have used two well-characterized monoclonal anti-HLA-B27antibodies (B27.M1 and B27.M2) to identify cross-reactivebacterial epitopes. Van Bohemen et al. (38) identified a

* Corresponding author.

16-kilodalton (kDa) envelope component of K. pneumoniaeand Y. enterocolitica which reacted with B27.M1 and aB27.M2-reactive 20-kDa antigen of Shigella flexneri 2a.Ogasawara et al. (25) also used B27.M2 to demonstratecross-reactivity with 80- and 60-kDa antigens of K. pneumo-niae K43. No cross-reactivity was seen with seven otherenterobacterial isolates, including S. flexneri, Y. enteroco-litica, or the K77 strain of K. pneumoniae. In prior studies(30), we found B27.M2- and B27.M1-cross-reactive antigensof 36 and 23 kDa that were common to many gram-negativebacteria, not all of which have been etiologically associatedwith reactive arthritis.The present study was undertaken because of the potential

significance of molecular mimicry in the initiation of charac-teristic spondyloarthritic lesions and because of the previousproblems associated with identification of the relevant bac-terial antigen. Rather than using HLA-B27 monoclonal orpolyclonal antisera to identify these antigens, we producedmonoclonal antibodies against an isolate of S. flexneri 2a thatwas cultured from a patient with Reiter's syndrome. Theseantibodies were then used to demonstrate the existence ofcross-reactive epitopes on the native HLA molecule and onsynthetic polypeptides corresponding to the HLA-B variableregion.

MATERIALS AND METHODS

Bacterial strains. Bacterial isolates were generously pro-vided as follows: S. flexneri, Shigella sonnei, and Salmo-nella sp. from J. G. Wells, Centers for Disease Control,Atlanta, Ga.; Escherichia coli, K. pneumoniae, and Yersiniapseudotuberculosis from D. T. Y. Yu, University of Califor-nia, Los Angeles, School of Medicine; and Y. enterocoliticafrom Peter Sheldon, Public Health Laboratory ServiceBoard, London, England.

Bacterial envelope preparations. Whole bacterial enve-lopes were prepared by ultrasonication and differential cen-

1774

Vol. 58, No. 6

CROSS-REACTIVITY OF BACTERIAL ANTIGEN AND HLA

TABLE 1. HLA phenotype of cell lines used in testingmonoclonal antibodies

MHC class I surface antigenaCell line

HLA-A HLA-B HLA-C

HOM-2 3 27 1DaudiK562MM 2,24 27,44 2,51065 24 2EL-4 (mouse)EB-lb 7EA-6b 2

a MHC, Major histocompatibility complex.b Transfected mouse cell line (14).

trifugation as described previously (29). Total protein con-

tent was determined by the method of Lowry et al., modifiedfor membrane-associated proteins (24).

Production of monoclonal antibodies. BALB/c female micewere inoculated in multiple subcutaneous sites with 100 ,ugof total protein suspended in 0.1 ml of normal saline. Micewere rested for 1 month and then given a final intraperitonealinoculum of 100 ,ug of total protein. The immunogen for allfusions was whole bacterial envelopes prepared from S.flexneri 2a isolated from a patient with Reiter's syndrome(SSU 6335). Splenocytes were fused with syngeneic plasma-cytoma cells P3x63.Ag8.653) as described by Kohler andMilstein (19) and modified by Kennett (17) and Oi andHerzenberg (26), with polyethylene glycol 1500 (Boehringer-Mannheim Biochemicals, Indianapolis, Ind.) as the fusogen.Positive hybridomas were cloned by using the Epics CS flowcytometer equipped with an autoclone accessory. Antibod-ies were isotyped by radial immunodiffusion by using rabbitanti-mouse subclass-specific antisera (Binding Site, Inc.,San Diego, Calif.). Ascites fluids were produced as previ-ously described (15). Immunoglobulin M (IgM) was purifiedfrom ascites fluids by Sephadex G200 (Pharmacia FineChemicals, Piscataway, N.J.) chromatography followed byhypoosmotic precipitation by dialysis of IgM-containingfractions against 1 mM Tris hydrochloride (pH 7.5).

Synthetic polypeptides. Peptides were synthesized in an

automated peptide synthesizer (model 430A; Applied Bio-systems, Inc., Foster City, Calif.). Peptide from the solid-phase support was deprotected and released by treating theprotected resin-attached peptide with anhydrous hydroflu-oric acid containing 10% anisole for 2 h at 0°C. Peptide puritywas monitored by reversed phase high-pressure liquid chro-matography. Peptides were determined to be >90% pure.ELISA. Hybridomas were screened for antibody produc-

tion by an enzyme-linked immunosorbent assay (EUSA)with 0.05 p.g total bacterial envelope protein applied to eachwell of 96-well flat-bottomed polystrene microdilution plates(Immulon I; Dynatech Laboratories, Inc., Chantilly, Va.).Hybridoma supernatants (100 1d) were pipetted into coatedwells and incubated for 2 h at room temperature. Boundantibody was detected with a 1:1,000 dilution of alkalinephosphatase-conjugated goat anti-mouse IgG (heavy andlight) (Jackson Immunoresearch Laboratories, Inc., WestGrove, Pa.) with p-nitrophenyl phosphate (Sigma ChemicalCo., St. Louis, Mo.) as the chromogen. ELISAs withsynthetic polypeptide were optimized to 10 ,ug/well of pep-tide bound to Dynatech Immulon II microdilution plates.The ELISAs were carried out as described above by using100 1.l of a 2-,ug/ml solution of purified IgM suspended in 1%normal goat serum.

TABLE 2. Specificity of monoclonal antibodies forcultured cell lines

Specificity for the indicated celi lineCell line

A7E2 C7E1O E2B6 FlOE10 D1D8a

HOM-2 + + + + -DaudiK562 - - - - -

MM + + + + -1065 + + + + -EL-4 - _ _ - _EB-1 + + + + _EA-6 - - - - -

a Isotype control.

Flow cytometry. Reactivity of monoclonal antibodiestoward native HLA molecules was assessed by indirectimmunofluorescence with transformed human lymphoblas-toid cell lines or transfected murine cells that express HLAsurface antigen. Daudi and K562 cell lines were obtainedfrom American Type Culture Collection, Rockville, Md.HOM-2 and MM cells (HLA-B27 positive) and an MM-derived HLA-B27 negative mutant (1065) were provided byD. T. Y. Yu. The murine T-cell iymphoma line (EL-4) andEL-4-derived transfectants (EA-6, EB-1) were obtainedfrom Victor H. Engelhard, University of Virginia School ofMedicine, Charlottesville, Va. (14). All cells were main-tained in RPMI medium supplemented with 10% fetal calfserum, 2 mM L-glutamine, and nonessential amino acids.Transfected cells were maintained in the same medium thatalso contained 250 ,ug of G418 (GIBCO Laboratories, GrandIsland, N.Y.) per ml. Methods for fluorescent staining of livecells were performed as previously described (30) with a 1:5to 1:10 dilution of hybridoma supernatant and a 1:10 dilutionof fluorescein isothiocyanate-conjugated goat anti-mouseIgM F(ab')2 fragments (Jackson).Sodium dodecyl sulfate-polyacrylamide gel electrophoresis

and Western immunoblot analysis. Cultured lymphocytes(108) were washed three times by centrifugation for 5 min at500 x g in 0.4 M NaCl-200 mM Tris hydrochloride (pH 7.4).Membrane proteins were extracted by resuspending the cellpellet in 1 ml of the same buffer containing 0.5% NonidetP-40 (Pierce Chemical Co., Rockford, Ill.) and 1 mM phen-ylmethylsulfonyl fluoride (Sigma). After incubation for 30min on ice, extracted cells were pelleted for 1 h at 40,000 xg. Supernatant samples containing 21 ,ul of membrane ex-tract (40 to 50 pg of total protein) were examined byelectrophoresis under reducing conditions by the method ofLaemmli (22) with a 12.5% polyacrylamide resolving gel.Alternatively, pelleted bacterial envelopes were suspendedin sample buffer, and 25-,ul samples (50 ,ug of total protein)were electrophoretically separated under the same condi-tions. Methods for transfer of proteins to nitrocellulosemembranes were as described earlier (34). After blocking for1 h with 1% bovine serum albumin (radioimmunoassaygrade; Sigma) in 0.04 M Tris-0.5 M NaCl (pH 7.5), reactivebands were observed by using a 1:5 to 1:10 dilution ofhybridoma supernatant and a 1:1,500 dilution of peroxidase-conjugated goat anti-mouse IgM. Blots were developed witha solution of 0.05% 4-chlor-1-naphthol ard 0.006% H202.

RESULTS

Antibody-secreting hybridomas were identified by ELISAwith envelope proteins of S. flexneri 2a as the solid-phase

VOL. 58, 1990 1775

1776 WILLIAMS AND RAYBOURNE

U)

LUI

LI.

0

LU

z

Lu

1-I

LU

LOGIo FLUORESCENCE INTENSITYFIG. 1. Immunofluorescent analysis of cultured lymphocytes and transfected murine cells with monoclonal antibodies. Cells were stained

by an indirect method using hybridoma supernatants followed by fluorescein isothiocyanate-conjugated goat anti-mouse IgM F(ab')2fragments as described in Materials and Methods. HLA-cross-reactive antibodies (A7E2, FlOE10) are compared with the non-cross-reactiveisotype control (D1D8).

reactant. Culture fluids with an A405 of >1.0 were thenevaluated for specificity of binding to the cultured lympho-cytes listed in Table 1. Of more than 600 antibacterialantibodies tested by flow microfluorometry, 4 were selectedfor further characterization, based on intensity of surfacefluorescence observed when live cells were stained by an

indirect method. These four antibodies (designated A7E2,C7E10, E2B6, and FlOE10) were all of the IgM isotype. Anadditional antibacterial IgM monoclonal antibody, desig-nated D1D8, was randomly chosen for use as an isotypecontrol.

Results of flow cytometric assays are summarized in Table

iq ;j,i~I.#

if. D1D8

INFECT. IMMUN.

CROSS-REACTIVITY OF BACTERIAL ANTIGEN AND HLA

TABLE 3. Sequences of MHC class I-specific synthetic polypeptides (variable region)

Amino acid sequencePeptide

65 70 75 80

HLA-B*2705a E T Q I C K A K A Q T D R E D L R T L L RHLA-B7 N T Q I Y K A Q A Q T D R E S L R N L R GHLA-B40 E T Q I S K A N T Q T Y R E S L R N L R G

a Formerly designated HLA-B27.1. Contains sequence required for B27.M2 binding (5).

2. HOM-2 cells, which are homozygous at the HLA-B locus,were strongly positive when stained with a saturating con-centration of each of the four test monoclonal antibodies. Bycontrast, Daudi and K562 cells, which lack cell surfaceexpression of class I molecules, were uniformly negative(Fig. 1). MM cells displayed a level of surface fluorescencecomparable to that of HOM-2 cells. The mutant cell line1065, which was derived from the HLA-B27-positive cellline MM but does not express for HLA-B locus antigens,was also tested. Although the antibodies stain MM cellsmore brightly, the 1065 cells demonstrated observable levelsof fluorescence staining (Fig. 1). The transfected cell lineEB-1, which expresses surface HLA-B7, was also positivewith all four test monoclonal antibodies; however, cellstransfected with and expressing HLA-A2 (EA-6) and theparental cell line (EL-4) were negative (Fig. 1).To characterize the molecular basis for antibody activity,

synthetic polypeptides were produced with an amino acidsequence identical to the variable region-spanning residues63 through 83 of the HLA-B*2705, HLA-B7, and HLA-B40heavy chains (Table 3). Monoclonal antibodies were testedfor antipeptide activity by ELISA. All antibodies werepositive with each of the synthetic polypeptides, althoughpreferential binding was observed with the HLA-B*2705peptide and the homologous antigen (S. flexneri 6335 enve-

SPECIFICITY OF ANTI-BACTERIAL ANTIBODIES(Antibody Concentration * 2 ug/ml)

O.D. 406 nm

1.5

0.5

A7E2 C7E1O E2B6 FlOE10 D1D8

HLA-B27 peptide HLA-B7 peptideW HLA-B40 peptide S. flexneri 6335

FIG. 2. Reactivity by ELISA of antibacterial monoclonal anti-bodies with HLA-B synthetic polypeptides and S. flexneri envelopeproteins. Peptides were synthesized according to the sequence oftheir respective variable regions (amino acids 63 through 83). Assays(described in Materials and Methods) used monoclonal antibodiespurified from ascites fluids. Cell-reactive monoclonal antibodieswere positive against all three synthetic polypeptides, althoughstronger reactivity was observed with HLA-B27 peptide and bacte-rial envelope proteins. Control antibacterial monoclonal antibodies(D1D8 isotype control) were negative against synthetic peptides.

lope) (Fig. 2). Antibodies were nonreactive with two syn-thetic peptides of unrelated sequence, which were includedas negative controls (data not shown).Membrane proteins of the HLA-B27-positive cell line MM

were extracted with detergent, resolved by sodium dodecylsulfate-polyacrylamide gel electrophoresis, and transferredto nitrocellulose for Western blot analysis with the HLA-reactive monoclonal antibody (Fig. 3). All monoclonal anti-bodies bound to a protein with a molecular mass of approx-imately 44 kDa, which is identical to that of the MHC classI heavy chain. A protein of similar molecular mass wasidentified by the B27-specific IgM monoclonal antibodyB27.M2, which was included as a positive control. Otherisotype controls were nonreactive by Western blotting.When tested by immunoblotting with electrophoretically

resolved S. flexneri 2a membrane proteins, each of the fourHLA-reactive monoclonal antibodies gave an identical pat-tern of reactivity. A major reactive protein was observed at36 kDa, and a less prominent band was seen at 19 kDa.

A B C D E F G

I 1 6

84

58

48.5

36.5

26.6

FIG. 3. Western blot analysis of detergent-extracted MM cells(HLA-B27 positive) with monoclonal antibodies. Cell membraneproteins were extracted with Nonidet P-40 and subjected to sodiumdodecyl sulfate-polyacrylamide gel electrophoresis and immunoblotanalysis as described in Materials and Methods. Antibacterial mono-clonal antibodies (lanes B through E) and the HLA-B27-specificmonoclonal antibodies, B27.M2 (lane F), reacted with a protein witha molecular mass consistent with that of the major histocompatibil-ity complex class I H chain (44 kDa). Lanes: A, Coomassieblue-stained membrane extract; B, A7E2; C, C7E10; D, E2B6; E,FlOE10; F, B27.M2; G, D1D8 (isotype control).

VOL. 58, 1990 1777

1778 WILLIAMS AND RAYBOURNE

A 3 C D E F

84

8.....

58

48.5

*_9MOW-'4f,W AII

26.6

FIG. 4. Western blot analysis of S. flexneri envelope proteinswith monoclonal antibodies. All of the HLA-reactive monoclonalantibodies reacted with proteins of 36 and 19 kDa; the 36-kDaprotein was the most prominent. Control antibacterial monoclonalantibodies gave a different pattern of reactivity. Lanes: A,Coomassie blue-stained envelope proteins; B, A7E2; C, C7E10; D,E2B6; E, FlOE10; F, D1D8 (isotype control).

Neither protein reacted with antibody D1D8, which was

included as an isotype control (Fig. 4). To determinewhether the cross-reactive epitope was unique to this isolateof S. flexneri, envelope proteins were prepared from a

collection of S. flexneri strains representing most serotypesand from a group of related bacteria, some of which are alsoepidemiologically associated with arthritic disease. Thesepreparations were standardized for protein concentrationand tested by ELISA with the monoclonal antibody. Positiveactivity against bacterial strains was observed with theHLA-reactive monoclonal antibodies (Table 4). The isotypecontrol was specific for S. flexneri 2a. Although the reactionof antibody C7E10 was weaker than that of the remainingthree antibodies, no preferential binding was seen with any

strains tested.

DISCUSSION

Despite the epidemiologic evidence associating gram-neg-ative members of the family Enterobacteriaceae with theseronegative spondyloarthropathies, the mechanism bywhich bacterial factors can initiate and perpetuate thesediseases remains hypothetical. Furthermore, in contemplat-ing pathogenic mechanisms, equal consideration must begiven to genetic factors, such as the presence of the class Ihistocompatibility antigen HLA-B27, which undoubtedlypredisposes individuals to postinfectious arthritic sequelae(6). The results presented in this report demonstrate thatmonoclonal antibodies prepared against an arthritogenicorganism also recognize epitopes present on class I histo-compatibility antigen. This supports the hypothesis that thelink between environmental and genetic etiologic factors is acommon immunodeterminant and that this molecular mim-icry could be of pathologic significance in the induction ofchronic arthritic disease.The four cross-reactive monoclonal antibodies were ini-

tially identified by using cultured cell lines in an indirectmicrofluorometric assay. Although previous investigatorshave used a microcytotoxicity assay to determine HLA

specificity of their monoclonal antibodies (20), we tried toavoid any potential problems associated with complement-binding capabilities of the monoclonal antibodies. The fourmonoclonal antibodies had similar patterns of reactivityagainst the cultured cell lines. Antibodies were completelynegative against cells lacking MHC class I antigens (Daudi,K562, EL-4) and against murine lymphoid cells transfectedwith and expressing only HLA-A2 (EA-6). Monoclonalantibodies were not, however, singularly positive with HLA-B27-positive cells; they were equally reactive with trans-fectants expressing only HLA-B7 (EB-1) and nominallyreactive with cells lacking HLA-B antigen but expressingHLA-A24 and HLA-C2 (1065). This low degree of allelicspecificity indicates that the epitope recognized by theseantibodies consists of nonpolymorphic residues shared byseveral HLA class I antigens. We can also conclude that thisepitope differs from that recognized by B27.M2, which isspecific for B27 and Bw47 (5). Although it appears that ourmonoclonal antibodies bind preferentially with B-locus anti-gens, fluorescence intensity between cultured lymphocytesmay not be directly comparable, because the surface densityof HLA antigen may vary from one cell line to another.

In Western blot analyses against detergent-extracted MM(HLA-B27-positive) cells, all four cross-reactive monoclonalantibodies bound to a single band at an approximate molec-ular mass of 44 kDa. This molecular mass corresponds tothat previously reported for the MHC class I heavy chainand is identical to that observed when monoclonal antibodyB27.M2 is subjected to immunoblot analysis. Based onELISA results with synthetic polypeptides corresponding toamino acids 63 through 83 of the class I heavy chain, wedetermined that the epitope recognized by our monoclonalantibodies was present in this segment. Previous investiga-tors have shown that this polymorphic sequence also con-tains the epitope responsible for specific binding of B27.M2(5) and recognition structures for allospecific cytotoxic Tlymphocytes (2, 41). Although this segment would appearcritical in defining allospecificity of the HLA-B27 molecule,significant structural homology exists between the HLA-B27variable region and the corresponding segment of non-HLA-B27 class I molecules (32). Several nonpolymorphicepitopes within the variable region have been identified byusing monoclonal antibodies generated against lymphocyticcells (1, 11, 27, 28, 31). The existence of such epitopes wouldexplain the reactivity of our four monoclonal antibodies withnon-HLA-B27 class I antigen and in particular with theHLA-B7 and HLA-B40 synthetic polypeptides, which werealso derived from their respective variable regions. Thepresence of such epitopes may also explain the high inci-dence of HLA-B7 CREG (cross-reactive group) antigens inHLA-B27-negative spondyloarthritic patients (4, 18).Using S. flexneri envelope proteins, we observed qualita-

tively identical patterns of immunoblot reactivity for the fourmonoclonal antibodies. The reactive epitope was present on36- and 19-kDa antigen of S. flexneri, but was not restrictedto this species of the family Enterobacteriaceae. StrongELISA activity was observed with all strains tested, evenwith species generally considered nonarthritogenic (e.g., E.coli). Despite the presence of such antigens, additionalbacterial virulence mechanisms (such as the ability to invadehost cells and survive intracellularly) are also characteristicof species with arthritogenic potential. Indeed, the ability ofan organism to establish infection is a prerequisite for hostexposure to and immune system recognition of cross-reac-tive antigens. The occurrence of a cross-reactive epitopethat is present on many organisms is consistent with the

INFECT. IMMUN.

CROSS-REACTIVITY OF BACTERIAL ANTIGEN AND HLA

TABLE 4. Reactivity of monoclonal antibodies with enterobacterial envelope proteins

A405Bacterial strainA7E2 C7E1O E2B6 FlOE10 D1D8I

Shigella flexnerila 1235-66la 1921-71lb 6115blb 4343-70lb 5612-732a 6335b2a 947b2a 2747-712a 299032b 120222b 16762b 97683 2146-663 2783-714a 6603-634a 1862-714b 880-694b 1242-705 1170-745 5103-826 2924-716 796-83

1.8451.4841.8011.2231.3281.9091.7951.5050.9201.6391.4371.3621.4301.5571.5701.7601.3371.0021.1641.1661.1631.132

Shigella sonnei64257063

1.7541.890

Salmonella heidelburg 7375b

Salmonella typhimurium 24b

Klebsiella pneumoniae K43

Escherichia coli109083374

Yersinia pseudotuberculosis78b

269b

Yersinia enterocoliticaA2bA6b

1.204

1.690

1.289

1.8351.995

1.9931.180

1.2891.567

0.6520.5970.4110.4540.4780.5170.4060.5740.2690.4700.4300.2970.4860.5680.5310.4540.4450.3470.4060.4410.4280.429

0.4490.602

0.434

0.511

0.337

0.6110.655

0.6980.425

0.5770.763

1.8881.6041.1151.1721.3431.5271.6791.5081.0701.3161.4101.0231.5641.5861.4841.5171.4261.0361.0381.1490.9961.030

1.9331.740

1.342

1.872

1.365

1.7451.885

1.9921.174

1.4961.926

1.2851.3841.2651.0121.4391.4201.3971.1670.6531.4961.0241.0111.1151.1631.1931.2961.9071.2761.8041.4961.3321.532

1.2261.341

1.204

1.226

0.760

1.3291.526

1.6860.946

0.8551.948

0.0860.0570.0560.0500.0481.8361.9471.7351.3890.0750.0540.0490.0510.0520.0490.0700.0720.0580.0480.0630.0550.059

0.0710.055

0.085

0.078

0.058

0.0640.054

0.0660.951

0.0540.074

a Isotype control.b Isolated from a patient with Reiter's syndrome or reactive arthritis.

finding that many species of gram-negative bacteria areepidemiologically implicated in these diseases. Furthermore,many of the 36- to 38-kDa major outer membrane proteins ofmembers of the family Enterobacteriaceae are conservedphylogenetically and considered common antigens (13),making them candidate recognition structures for our cross-reactive monoclonal antibodies. A comparison of the rela-tive mobility in polyacrylamide gels of the bacterial antigenrecognized by B27.M2 indicates that our monoclonal anti-bodies have a different antigen specificity. The molecularspecificity of our antibodies is currently under investigation.

Further studies on humoral and cell-mediated responses ofspondyloarthritic patients which parallel the course of theirdisease are required to determine the pathologic significanceof the 36- and 19-kDa antigens. A vigorous antibody re-sponse against the causative organism is present in allpatients after a gram-negative enteritis; however, those who

develop subsequent arthritis show higher and more persist-ent antibody titers in serum (35, 36, 39). Serum antibodies ofthe IgG, IgM, and IgA classes directed against Yersiniaproteins of 35 to 36 kDa have been reported in patients withreactive arthritis after infection with Y. enterocolitica (12,33, 40), but whether this antigen corresponds to the 36-kDaantigen recognized by our cross-reactive monoclonal anti-body is yet to be determined. The availability of monoclonalantibodies will facilitate purification of this antigen for fur-ther studies with patient specimens.Of greater interest is the potential regulatory role of

bacterially induced HLA class I-reactive antibodies on T-cell responses in a susceptible host. Several monoclonalanti-class I HLA antibodies, in the presence of adherentantigen-presenting cells, can inhibit mitogen- and antigen-induced T-cell proliferation (9, 10, 37). An antibody-medi-ated down-regulation of cellular responses to arthritogenic

VOL. 58, 1990 1779

1780 WILLIAMS AND RAYBOURNE

bacteria could facilitate persistence of bacteria in infectedcells. Indeed, defective cellular immune functions have beenreported in arthritic patients. Leino et al. (23) found thatlymphoproliferative response to bacterial antigen was de-pressed in the peripheral blood mononuclear cells of patientswho developed post-Yersinia reactive arthritis as comparedwith those who recovered from yersiniosis uneventfully. Theimpaired response was not organism specific, since signifi-cant inhibition was obtained after stimulation with eitherYersinia or E. coli antigen. Similarly, peripheral bloodmononuclear cells from patients with reactive arthritis afterSalmonella typhimurium infection showed an impaired lym-phoproliferative response to in vitro stimulation with thecausative organism (16). Hypothetically, HLA-B27-reactiveantibodies could be induced by a common bacterial antigenin all infected patients, but their negative regulatory effectswould be expressed only in patients who are HLA-B27positive or whose class I antigens contain the cross-reactiveepitope.

In conclusion, we identified 36- and 19-kDa bacterialenvelope proteins that share cross-reactive epitope(s) withclass I histocompatibility molecules. Although the mecha-nism by which these or other bacterial antigens precipitatejoint disease remains unknown, the existence of sharedimmunodeterminants lends support to the concept of molec-ular mimicry in disease induction.

LITERATURE CITED1. Antonelli, P., B. Nisperos, P. Gladstone, and J. Hansen. 1983.

Recognition of a unique HLA-B27.7, w22, 17, 14, w63, w46common epitope by two independently derived BALB/c mono-clonal antibodies. Hum. Immunol. 8:296.

2. Aparicio, P., M. A. Vega, and J. A. Lopez de Castro. 1985. Oneallogeneic cytolytic T lymphocyte clone distinguishes threedifferent HLA-B27 subtypes: identification of amino acid resi-dues influencing the specificity and avidity of recognition. J.Immunol. 135:3074-3081.

3. Archer, J. R. 1981. Search for cross-reactivity between HLA-B27 and Klebsiella pneumoniae. Ann. Rheum. Dis. 40:400-403.

4. Arnett, F. C., Jr., M. C. Hochberg, and W. B. Bias. 1977.Cross-reactive HLA antigens in B27 Reiter's syndrome andsacroiliitis. Johns Hopkins Med. J. 141:193-197.

5. Bjorkman, P. J., M. A. Saper, B. Samraoui, W. S. Bennett, J. L.Strominger, and D. C. Wiley. 1987. The foreign antigen bindingsite and T cell recognition regions of class I histocompatibilityantigens. Nature (London) 329:512-518.

6. Brewerton, D. A., M. Caffrey, F. D. Hart, D. C. 0. James, A.Nichols, and R. D. Sturrock. 1973. Ankylosing spondylitis andHLA-27. Lancet i:904-907.

7. Brewerton, D. A., M. Caffrey, A. Nicholls, D. Walters, J. K.Oates, and D. C. James. 1973. Reiter's disease and HLA-27.Lancet i:996-998.

8. Bunning, V. K., R. B. Raybourne, and D. L. Archer. 1988.Foodborne enterobacterial pathogens and rheumatoid disease.J. Appl. Bacteriol. Symp. Suppl. 65:87S-107S.

9. Dasgupta, J. D., K. Cemach, D. P. Dubey, E. J. Yunis, and D. B.Amos. 1987. The role of class I histocompatibility antigens in theregulation of T-cell activation. Proc. Natl. Acad. Sci. USA84:1094-1098.

10. Dasgupta, J. D., and E. J. Yunis. 1987. Receptor-like role ofHLA-class I antigens: regulation of T cell activation. J. Immu-nol. 139:672-677.

11. Ellis, S., C. Taylor, and A. McMichael. 1982. Recognition ofHLA-B27 and related antigens by a monoclonal antibody. Hum.Immunol. 5:49-59.

12. Gronberg, A., A. Fryden, and E. Kihlstrom. 1989. Humoralimmune response to individual Yersinia enterocolitica antigensin patients with and without reactive arthritis. Clin. Exp.Immunol. 76:361-365.

13. Hofstra, H., and J. Dankert. 1980. Major outer membrane

proteins: common antigens in Enterobacteriaceae species. J.Gen. Microbiol. 119:123-131.

14. Holterman, M. J., and V. H. Engelhard. 1986. HLA antigensexpressed on murine cells are preferentially recognized bymurine cytotoxic T cells in the context of the H-2 majorhistocompatibility complex. Proc. Natl. Acad. Sci. USA 83:9699-9703.

15. Hoogenraad, N. J., and C. J. Wraight. 1986. The effect ofpristane on ascites tumor formation and monoclonal antibodyproduction. Methods Enzymol. 121:375-381.

16. Inman, R. D., B. Chiu, M. E. A. Johnston, S. Vas, and J. Falk.1989. HLA class I-related impairment in IL-2 production andlymphocyte response to microbial antigens in reactive arthritis.J. Immunol. 142:4256-4260.

17. Kennett, R. H. 1979. Cell fusion. Methods Enzymol. 58:345-359.

18. Khan, M. A. 1983. B7-CREG and ankylosing spondylitis. Br. J.Rheumatol. 22(Suppl. 2):129-133.

19. Kohler, G., and C. Milstein. 1975. Continuous cultures of fusedcells secreting antibody of predefined specificity. Nature (Lon-don) 256:495-497.

20. Kono, D. H., M. Ogasawara, R. B. Effros, M. S. Park, R. L.Waldord, and D. T. Y. Yu. 1985. Ye-1, a monoclonal antibodythat cross-reacts with HLA-B27 lymphoblastoid cell lines andan arthritis causing bacteria. Clin. Exp. Immunol. 61:503-508.

21. Kuverski, T. T., M. G. Morse, R. G. Rate, and M. D. Bonnell.1983. Increased recovery of Klebsiella from the gastrointestinaltract of Reiter's syndrome and ankylosing spondylitis patients.Br. J. Rheumatol. 22(Suppl. 2):85-90.

22. Laemmli, U. K. 1979. Cleavage of structural proteins during theassembly of the head of bacteriophage T4. Nature (London)227:680-685.

23. Leino, R., R. Vuento, S. Koskimies, M. Viander, and A. Toiva-nen. 1983. Depressed lymphocyte transformation by yersiniaand Escherichia coli in yersinia arthritis. Ann. Rheum. Dis.42:176-181.

24. Markwell, M. A., S. M. Haas, L. L. Bieber, and N. E. Tolbert.1978. A modification of the Lowry procedure to simplify proteindetermination in membrane and lipoprotein samples. Anal.Biochem. 87:206-210.

25. Ogasawara, M., D. H. Kono, and D. T. Y. Yu. 1986. Mimicry ofhuman histocompatibility HLA-B27 antigens by Klebsiellapneumoniae. Infect. Immun. 51:901-908.

26. Oi, V. T., and L. A. Herzenberg. 1980. Immunoglobulin produc-ing hybrid cell lines, p. 351-372. In B. B. Mishell and S. M.Shiigi (ed.), Selected methods in cellular immunology.

27. Parham, P., P. Antonelli, L. A. Herzenberg, T. J. Kipps, A.Fuller, and F. E. Ward. 1986. Further studies on the epitopes ofHLA-B7 defined by murine monoclonal antibodies. Hum. Im-munol. 15:44-67.

28. Parham, P., C. J. Barnstable, and W. F. Bodmer. 1979. Use ofa monoclonal antibody (W6/32) in structural studies of HLA-A,B, C antigens. J. Immunol. 123:342-349.

29. Portnoy, D. A., H. Wolf-Watz, I. Bolin, A. B. Beeder, and S.Falkow. 1984. Characterization of common virulence plasmidsin Yersinia species and their role in the expression of outermembrane proteins. Infect. Immun. 43:108-114.

30. Raybourne, R. B., V. K. Bunning, and K. M. Williams. 1988.Reaction of anti-HLA-B monoclonal antibodies with envelopeproteins of Shigella species: evidence for molecular mimicry inthe spondyloarthropathies. J. Immunol. 140:3489-3495.

31. Rebai, N., and B. Malissen. 1983. Structural and genetic analy-ses of HLA class I molecules using monoclonal xenoantibodies.Tissue Antigens 22:107-117.

32. Szots, H., G. Reithmuller, E. Weiss, and T. Meo. 1986. Completesequence of HLA-B27 cDNA identified through the character-ization of structural markers unique to the HLA-A, -B, and -Callelic series. Proc. Natl. Acad. Sci. USA 83:1428-1432.

33. Toivanen, R., R. Tertti, R. Lahesmaa-Rantala, T. H. Stahlberg,and K. Granfors. 1988. Factors associated with the developmentof reactive arthritis. Br. J. Rheumatol. 27(Suppl. 2):46-51.

34. Towbin, H., T. Staehelin, and J. Gordon. 1979. Electrophoretictransfer of proteins from polyacrylamide gels to nitrocellulose

INFECT. IMMUN.

CROSS-REACTIVITY OF BACTERIAL ANTIGEN AND HLA

sheets: procedures and some applications. Proc. Natl. Acad.Sci. USA 76:43504354.

35. Truli, A. K., C. G. Eastmond, G. S. Panayi, and T. M. S. Reid.1986. Salmonella reactive arthritis: serum and secretory anti-bodies in eight patients identified after a large outbreak. Br. J.Rheumatol. 25:13-19.

36. Truli, A., A. Ebringer, G. Panayi, R. Ebringer, and D. C. 0.

James. 1984. HLA-B27 and the immune response to enterobac-terial antigens in ankylosing spondylitis. Clin. Exp. Immunol.55:74-80.

37. Turco, M. C., M. De Felice, L. Corbo, G. Morrone, R. Mertels-mann, S. Ferrone, and S. Venuta. 1985. Regulatory role of amonomorphic determinant of HLA class I antigens in T cellproliferation. J. Immunol. 135:2268-2272.

38. Van Bohemen, C. G., F. C. Grument, and H. C. Zanen. 1984.Identification of HLA-B27M1 and -M2 cross-reactive antigensin Klebsiella, Shigella and Yersinia. Immunology 52:607-610.

39. Van Bohemen, C. G., R. J. Lionarons, P. van Bodegom, H. J.

Dinant, J. E. Landheer, A. J. J. M. Nabbe, F. C. Grumet, andH. C. Zanen. 1985. Susceptibility and HLA-B27 in post-dysen-teric arthropathies. Immunology 56:377-379.

40. Van Bohemen, C. G., A. J. W. van Alphen, 0. G. Zanen-Lim,A. J. Dekker-Saeys, and H. C. Zanen. 1983. Antibodies againstcell envelope antigens of Yersinia enterocolitica in reactivearthritis and ankylosing spondylitis. Br. J. Rheumatol.22(Suppl. 2):83-84.

41. Vega, M. A., A. Ezquerra, S. Rojo, P. Aparicio, R. Bragado, andJ. A. Lopez de Castro. 1985. Structural analysis of an HLA-B27functional variant: identification of residues that contribute tothe specificity of recognition by cytolytic T lymphocytes. Proc.Natl. Acad. Sci. USA 82:7394-7398.

42. Welsh, J., H. Avakian, P. Cowling, A. Ebringer, P. Wooley, G.Panayi, and R. Ebringer. 1980. Ankylosing spondylitis, HLA-B27 and Klebsiella. I. Cross-reactivity studies with rabbitantisera. Br. J. Exp. Pathol. 61:85-91.

VOL. 58, 1990 1781