INFECTION AND IMMUNITY, Feb. 1994, p. 484-491 0019-9567/94/$04.00+0 Copyright ©) 1994, American Society for Microbiology

Vol. 62, No. 2

Serum Antibodies from Patients with Ankylosing Spondylitis and Reiter's Syndrome Are Reactive with HLA-B27 Cells Transfected with the Mycobacterium tuberculosis hsp6O Gene HERBERT KELLNER,' JUAN WEN,' JUN WANG,1 RICHARD B. RAYBOURNE,2 KRISTINA M. WILLIAMS,2 AND DAVID TAK YAN YU'* Rheumatology Division, Department of Medicine, University of California at Los Angeles, Los Angeles, California 90024,1 and Immunobiology Branch, Food and Drug Administration, Laurel, Maryland 207082 Received 23 July 1993/Returned for modification 24 September 1993/Accepted 12 November 1993

HLA-B27-related arthritis is probably mediated by an immune response against HLA-B27 complexed with peptides derived from proteins of arthritis-causing bacteria. Immunogenic proteins with a high degree of homology among bacteria, such as in the hsp6O family, are likely candidates. To create such complexes experimentally, we transfected an HLA-B27 cell line with the Mycobacterium tuberculosis hsp6O gene. Because of previous observations that HLA-B27-peptide complexes can be distinguished by antibodies, we tested the transfected cell line with a panel of sera from 24 HLA-B27+ arthritis patients. Significant antibodies were detected in at least eight of the sera. Several cell lines and peptides were used as negative controls to ensure that the antibody reactivity was specific to HLA-B27-peptide complexes. A panel of nine peptides derived from the sequence of the Mycobacterium hsp6O were synthesized and tested. At least three were identified as being responsible for the serological activities. Two major clues concerning the pathogenesis of the HLAB27-related arthritis ankylosing spondylitis (AS) and Reiter's syndrome (RS) are available. The first is that most of the patients are HLA-B27 positive (HLA-B27+) (16). Because a parallel arthritis develops in rats made transgenic with the HLA-B27 gene (7), it is commonly accepted that HLA-B27 itself plays a direct role. The second clue is that, in some patients, the arthritis is induced by infection with certain facultative intracellular bacteria, including species of Salmonella, Shigella, Yersinia, and Chlamydia. The arthritis frequently persists for several years, reflected by a prolonged serological response to the bacteria (6). In some patients, live bacteria or their fragments have been detected in the synovial compartment, perhaps accounting for the selective enrichment of bacterium-responsive T cells in the joints (9). As an MHC class I antigen, HLA-B27 functions to carry small peptides with a well-defined sequence motif. These HLA-B27-peptide complexes are in turn recognized by T-cell receptors (TCRs) to activate CD8+ cytotoxic T lymphocytes (CTL). Such peptides are commonly accepted to be derived from the cytosolic compartments, classically from proteins endogenous to the cells or from viruses (5). Recently, peptides have been reported to be derived from the bacteria Escherichia coli, Listeria monocytogenes, and Salmonella typhimurium as well (14). It is very likely that many intracellular bacteria have this property. Hence, it is important to identify bacterial peptides which activate the CD8+ T cells of HLA-B27+ patients. There are two technical barriers. First, it is difficult to generate panels of CD8+ T cells from a large number of patients for simultaneous comparison. Second, even if panels of CD8+ T cells are available, it is hard to select the particular peptides of promise for extensive testing.

The present article introduces two completely different approaches. In the first, we used serum antibodies instead of CTL. This strategy is based on the report that, in influenza virus infections in H-2k mice, the hosts generate antibodies, in addition to CD8+ CTL, against virus-derived peptides complexed with H-2k. Furthermore, such antibodies are physiologically relevant, since they also modulate CTL activity (11). That such antibodies exist is not surprising. Monoclonal antibodies that have a TCR-like ability to recognize peptides complexed with MHC class I or class II antigens have already been reported (3, 12). It would be useful to know whether similar antibodies are present in patients with HLA-B27-related arthritis. Our second innovation is that instead of using peptides to generate target cells to study HLA-B27-related arthritis, we created target cells by stably transfecting an HLA-B27 cell line with a bacterial gene. The bacterial protein which is expressed is processed in the same way as an endogenous protein to generate peptides to be complexed with the HLA-B27. The bacterial protein that we selected was the archetypal analog of bacterial stress proteins, Mycobacterium tuberculosis hsp60 (also designated as hsp65 in some publications). This choice was based on the observation that this family of stress proteins is immunodominant among bacteria. Using HLA-B27 cells transfected with this bacterial gene, we were able to identify serum samples with positive antibody reactivities. Three peptides derived from this protein were found to be promising. MATERIALS AND METHODS Cell lines. Mouse H-2k fibroblasts (L929) stably transfected with the human ,32-microglobulin (,B2m) gene (cell line hereafter called 32m-L) or with both the HLA-B7 and the human P2m genes (hereafter called B7-12m-L) were kindly provided by J. Taurog, Dallas, Tex. These cells were resistant to hypoxanthine-aminopterin-thymidine medium. They were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum at 37°C in 5% CO2.

* Corresponding author. Mailing address: 35-40 Rehabilitation Center, Rheumatology Division, Department of Medicine, UCLA, 1000 Veteran Ave., Los Angeles, CA 90024-1670. Phone: (310) 825-5047. Fax: (310) 206-8606.

484

VOL. 62, 1994

ANTIBODIES TO HLA-B27 AND MYCOBACTERIUM PEPTIDES

Plasmids. Plasmid pRL13 was kindly provided by T. Shinnick, Hansen Disease Laboratory, Atlanta, Ga. This plasmid contains a 1.7-kb EcoRI-KpnI fragment from the Mycobacterium tuberculosis heat shock protein 60 (hsp60) gene (18). The insert was subcloned into the polylinker site of plasmid pHSS19 (J. A. Nickoloff, Boston, Mass.), excised by the flanking NotI restriction sites, and then inserted into the eukaryotic expression vector pRC/CMV (Invitrogen Corp., San Diego, Calif.). Plasmid pERU-S2 was a kind gift from R. C. Brunham, Winnipeg, Canada. This plasmid contains a 2.3-kb KpnI-SphI fragment of the Chlamydia trachomatis hsp70 gene (4). The gene was cloned into pRC/CMV by using the XbaI restriction sites. The HLA-B*2705 gene was a kind gift from J. Taurog, Dallas, Tex. A chimeric form of the HLAB*2705 gene, DB27, was obtained from G. J. Hammerling, German Cancer Research Center, Heidelberg, Germany. This chimera contained the genomic DNA encoding the otl and the ct2 domains of the HLA-B*2705 protein ligated to that encoding the c3 domain plus the transmembrane and cytoplasmic segments of the murine H2-Kd protein (1). This chimeric gene was designated DB27. Transfection of mouse L cells. Transfection was carried out by the calcium phosphate method (ProFection Mammalian Transfection; Promega, Madison, Wis.). The first series of transfections were designed to generate HLA-B27-positive geneticin-sensitive L cells. P2m-L cells were transfected with either the B27 or DB27 gene. Hygromycin resistance was provided by cotransfection with pCEP4 (Invitrogen Corp.) at a 1:10 ratio. Colonies which grew with 500 pLg of hygromycin per ml were selected for screening. Expression of B27 or DB27 was tested with the anti-human HLA class I monoclonal antibody W6/32 (13). These cell lines are denoted as B27-p2m-L and DB27-,B2m-L cells, respectively. The second series of transfections were designed to convert bacterial hsp-negative cell lines to positive ones. The M. tuberculosis hsp6o or C. trachomatis hsp7o gene was used. Colonies which grew with 400 ,ug of geneticin per ml were selected for screening. Expression of the hsp60 or hsp70 protein was detected by Western blot (immunoblot). Positive cells were designated MThsp6O and ChThsp7O, respectively. Western blot analyses. To analyze cells transfected with bacterial genes, the cells were solubilized with 0.5% Nonidet P-40 at 107 cells per ml. To analyze recombinant M. tuberculosis hsp60 protein expressed in E. coli, the bacteria expressing the proteins were sonicated, and the sediment was removed by centrifugation at 12,000 x g for 30 min. The intracellular proteins of the L cells or the bacteria were fractionated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to nitrocellulose membranes, and tested with ascites fluids containing the anti-hsp60 and anti-hsp70 monoclonal antibodies IIH9 (United Nations Development Program/ World Bank/World Health Organization Special Programme for Research and Training in Tropical Diseases) and SC107 (provided by R. C. Brunham), respectively, at a dilution of 1:100 by standard techniques. Source of sera. Sera from 11 HLA-B27-positive patients with ankylosing spondylitis (AS), 13 HLA-B27-positive patients with Reiter's syndrome (RS), and 6 patients with rheumatoid arthritis were studied. Diagnosis was made by the attending rheumatologists at the UCLA Medical Center, Los Angeles. No attempt was made to identify the possible causative bacteria. Although patients were regarded as having clinically active disease at the time that blood samples were taken, no attempt was made to verify or ascertain disease activity. Control sera were obtained from six healthy HLA-B27-positive individuals and 12 healthy blood donors of undetermined HLA phenotype

485

TABLE 1. Synthetic peptides Peptide

Derived from hsp6O A B C D E F G H I Known to bind to HLA-B27 J K L M N

Sequence

ARRGLERGL RRGLERGLN GRNVVLEKK VREGLRNVA LRNVAAGAN KRGIEKAVE RRKAMLQDM HRIEDAVRN ERLAKLAGG

Analog

Sequence"

Al AERGLRNVA Dl VEEGLRNVA Fl

KEGIEKAVE

Ml RRYQKSTEL RRISGVDRY RIQRGSGRAFVTIGR

REYQKSTEL

SRYWAIRTR RRIKEIVKK

RRVKEVVKK

0 " Residues that differ from those of the parent peptide are in boldface type.

from the American Red Cross, Los Angeles, Calif. HLA typing was done by the dye exclusion microcytotoxicity method at the HLA typing laboratory of UCLA. Purification of serum IgG. Serum immunoglobulin G (IgG) was isolated by protein A columns (Acti-Disk separation and purification cartridge; FMC, Pine Brook, N.J.). One milliliter of filtered serum (0.22-,um filter) diluted 1:4 with 0.01 M phosphate buffer, pH 7.2, was loaded onto a column, washed with 0.01 M phosphate buffer, and eluted with 0.1 M glycine HCl, pH 2.8. The recovered fraction was immediately neutralized and concentrated (Centricon microconcentrators; Amicon, Beverly, Mass.) to a final volume of 1.0 ml. The optical density at 280 nm was measured, and the final concentration of eluted IgG was adjusted to the serum level of 1,250

mg/dl. Flow microfluorocytometry analysis. From 3 x 105 to 5 x 105 cells in the log phase of growth were incubated for 60 min with a serum. After being washed twice with Hanks' balanced salt solution supplemented with 1% bovine serum albumin and 0.2% azide, the cells were incubated for another 60 min with 1 [lI of the phycoerythrin-conjugated F(ab')2 fragment of goat anti-human IgG or anti-human IgM (Jackson Immunoresearch Labs, West Grove, Pa.). All incubations were carried out at 4°C with shaking. Finally, the cells were washed twice and fixed with 2% paraformaldehyde. Analysis was done with a Becton

Dickinson flow cytometer and the version B FACScan Research software (San Jose, Calif.). A total of 2,500 cells per sample were analyzed. Live cells were gated by their lightscattering characteristics. Flow cytometry (FACS) values were calculated on a linear scale and expressed as mean fluorescence intensities. Peptide synthesis and loading onto cells. The sequences of the peptides synthesized are shown in Table 1. All peptides were made by F-moc chemical synthesis, carried free amino and free carboxyl termini, and were purified by high-pressure liquid chromatography (17). The compositions were verified by mass spectrometry. Peptides were incubated with cells at 1 mg/ml at 37°C overnight and washed before being assayed with antibodies. Statistics. Multiple values were calculated as mean ± standard error of the mean (SEM). Comparisons were made either by the Mann-Whitney test or by regression analysis.

486

INFEC7. IMMUN.

KELLNER ET AL. B27-2m-L-celis

A

82m-MThsp-L-cels |

1200

1200

1000

1000

800g

800

B00

600

400

400

200

C

200

__m_U-,, 3

1

2

5 4

9

7

13

71

8

6

1O

12

15 14

0

17 18

16

2

T .

.

3

5

7

6

4

8

9

.

10

12

13

14

15

16

17

18

1400z

I1400

1 B27-Mrn-ChThsp-L-clls

B27-B2m-MThsp-L-cells

D

1200

.?

1000

.000

c .)

c

800

0 a

600

0)

sO

0

400

.00

200

40 1

5*W 7 3 13 15. A a~~* a*v

I1

2

4

6

8

10

12

14

17

18

5

3

I

. . 16

2

7

4

9

10

I11

13

12

Is

14

17

16

is

FIG. 1. Reactivity of sera from 12 H[A-B27-undetermined controls (sera 7 to 18) and six HLA-B27-positive controls (sera 1 to 6) with mouse L929 fibroblasts stably transfected with the genes for (A) HLA-B27 and human 12m, (B) HLA-B27, human ,32m, and C. trachomatis hsp70, (C) human ,2m and M. tuberculosis hsp60, or (D) HLA-B27, human 12m, and M. tuberculosis hsp60. Serum reactivity was quantitated by immunofluorescence. Background fluorescence values with the second antibodies alone were less than 100.

RESULTS Testing assay parameters. Because preliminary experiments demonstrated that most human sera contained a significant amount of antibodies against L cell surface antigens not relevant to HLA-B27, all sera were absorbed with B27-fp2m-L cells. Next, a dose-response preliminary experiment with 0.001 to 25 ,ul of serum was carried out, and the volume of serum used per sample was determined to be optimal at 1 RI.l. Lastly, to test sample variation, triplicates of B27-32m-MThsp6O-L cells were assayed with two serum samples. Because the variability was very low, the standard deviations being less than 5% of the mean values, serum samples were assayed singly in all subsequent experiments. Testing control subjects. Sera from 12 healthy HLA-B27undetermined and six HLA-B27-positive controls were tested with B27-,B2m-MThsp6O-L cells (Fig. 1D). The mean fluorescence intensities ± SEM were 47 ± 46 and 3 + 0.4, respectively. These values were not different from the corresponding values obtained when B27-,B2m-L cells (Fig. 1A) and 12m-L cells (results not shown) were tested. In addition, sera from six patients with rheumatoid arthritis

also tested against B27-p2m-L, P2m-L, and B27-pi2mMThsp6O-L cells. The mean fluorescence intensities ± SEM were low for all three cell lines: 18 ± 6, 18 ± 5, and 14 + 9, respectively. All values for the control sera were in the same range when the second antibody was used alone. Reactivity of sera from AS and RS patients. The reactivity of sera from 11 AS patients (sera 3, 4, 5, 8, 12, 13, 17, 18, 20, 21, and 22) and 13 RS patients (sera 1, 2, 6, 7, 9, 10, 11, 14, 15, 16, 19, 23, and 24) to B27-,32m-MThsp60-L cells was tested (Fig. 2D). For the RS patient sera, fluorescence intensities ranged from 9 to 1,370 (mean ± SEM, 296 + 116). They were significantly higher than the corresponding values for the control sera (P < 0.0001). The four highest observed values were 1,370, 1,018, 649, and 222. For the AS patient sera, fluorescence intensities ranged from 23 to 850 (mean ± SEM, 326 ± 98). These values were again higher than the corresponding values for the control sera, with an equally high statistical significance (P 0.0001). The five highest values were 850, 813, 777, 453, and 336 (Fig. 2D). In subsequent sections concerning control cells, data for AS and RS patient sera were analyzed together. were