Susceptibility to Primary Biliary Cirrhosis Is Associated with the HLA-DR8-DQB1*0402 Haplotype JAMES UNDERHILL, PETERDONALDSON, GARYBRAY,DEREKDOHERTY, BERNARD PORTMA" AND ROGERWILLIAMS Institute of Liver Studies, King's College School of Medicine and Dentistry and King's College Hospital, London SE5 9RS, United Kingdom

although no systematic study of human leukocyte antigen (HLA) haplotypes in affected families has been carried out. Similarities with graft-vs.-host disease and increased expression of HLA class I1 on the bile duct epithelium (€49)suggest the involvement of HLA in the progression of the disease, although this may be an effect rather than the cause of tissue damage. In most autoimmune diseases overrepresentation of one or more HLAs is present-especially of the HLA haplotype Al-B8-DR3 -but until recently no clear association between the disease and HLA markers had been found. All studies to date, including our own unpublished data, have failed to show any significant association with HLA class I (HLA-A and B) (10-16). This contrasts with findings in CAH and primary sclerosing cholangitis, both of which show a significant association with the antigens A1 and B8 (17-20). Several studies have detailed associations between PBC and the immunoregulatory alleles C4B2 (211, tumor necrosis factor (TNF-alNcoI, 5.5 kb) (22) and B-associated transcript (BATIRsaI, 2.7 kb) (231, but to date the greatest attention has focused on HLA class 11. Early studies in Japan (13) and Spain (12) suggested associations with DR2 and DR3, respectively, although three larger studies failed to confirm these observations (14, 15,231. More recently, Gores et al. (24) described associations with DR8 and DR5, and their findings were confirmed by a smaller study from Germany (25). The most likely Patients with PBC exhibit immune abnormalities explanation for these variations between studies is the including increased serum IgM levels ( 11,autoantibodies inherent problems of HLA serotyping. Some antigens to nuclear and mitochondrial antigens (2, 3) (the most particularly DR5, DR6 and DR8-are difficult to define common of these was recently shown to be expressed on because of cross-reactivity and limitations in the availthe membranes of cultured biliary cells of PBC patients ability of monospecific antisera. Recent reports suggest [4]) and abnormal T-cell and complement function that up to 25% of HLA-DR serotyping may be incorrect suggesting an autoimmune pathogenesis (5, 61, albeit compared with more accurate restriction-fragmentwith unknown etiology. PBC carries a marked female length polymorphism (RFLP)HLA genotyping (26, 27). preponderance and isolated reports of familial incidence Taql RFLP analysis with a DRB gene specific comple(7), suggesting a genetic component to the disease, mentary DNA probe can unequivocally define DR8 by the presence of a single, unique 8.7-kb band (Fig. 1A) (28). In addition, the DNA techniques used for HLA Received May 4, 1992; accepted August 3, 1992. typing allow the detection of alleles in other loci that Address reprint requests to: Roger Williams, M.D. FRCP FRCS FRCPE cannot be detected by standard serological methods. FRACP FACP(Hon1, Institute of Liver Studies, King's College Hospital, HLAs are cell surface glycoproteins that function as Denmark Hill, London SE5 9RS, UK. 31/1/41729 part of the system of immune regulation. HLA class I1

In studies to date seeking associations between human leukocyte antigens (HLA)and primary biliary cirrhosis, no class I association but several different class I1 associations have been described. The aims of this study were to reassess the DR associations in primary biliary cirrhosis and to examine for the first time the role of DQB. DRB genotypes were determined on standard Taql restriction-fragment-length polymorphism analysis in 159 white northern European patients with the disease and 162 racially matched local controls. Polymerase chain reaction gene amplification and sequence-specificoligonucleotide analysis were used to determine DQB genotypes in 89 patients and 181 controls. An increased frequency of human leukocyte antigen DR8 was observed in the patient group (11% vs 4%; relative risk = 3.3; p < 0.01). Although we saw an increased frequency of the DQB1*0402 allele (11% vs. 3%; relative risk = 3.5; p < 0.025), this was not significant after correction for multiple testing. The strongest association was with the two-locushaplotype DR8-DQB1*0402(11%vs. 2.2%; relative risk 5.5; p < 0.001). The DRB data reported here confirm the findings of previous studies, although the described association with DR8 is considerably weaker. The weak genetic contribution of human leukocyte antigen in the susceptibility to primary biliary cirrhosis is in contrast to its role in other autoimmune liver diseases. (HEPATOLOGY 1992;16:14041408.)

1404

1405

Vol. 16. No. 6 , 1992

kb

13.81 11-48 11.1 1

r

DNA

samples

L1

- DR8

5. 4.12

y

2.69

u

m

0

1-30

m

0

-

31 60

61 - 90

1.72 1.69

1.42

A

1 2 3 4 5 6 7 8 9

6

-

m-

Y

.

91-120

FIG. 1. (A)Autoradlograph of nine Taql -digested D N A samples probed with the pVRTl complementary DNA DRB gene probe. For each lane. the RFLP pattern of bands revealed by the hybridization of the probe corresponds to the DR type of the individual. ( B I Autoradiograph of 120 DQB PCR products probed with an oligronucleotide probe specific for D Q B 1 4 4 0 1 The probe hybridizes only with those samples carrying that allele, appearing as a dark spot on thcTfilm

glycoproteins are composed of cx- and p-subunits?coded for by paired A and B genes on the short arm of' chromosome 6. At the HLA-DQ and HLA-DP loci, both the A and B genes are polymorphic, but at HLA-DR. variation in the DRB genes alone is responsible for the expressed DR tissue type. Linkage on the same chromosome means that alleles of different HLA loci are inherited together as part of a haplotype. HLA-DR and HIA-DQ are close together on the chromosome and show strong linkage, and although HLA-DR has been investigated in PBC, HLA-DQ has not. The aims of this study were to investigate the relationship between HLA-DR and PBC,with more precise methods to define DR8, to extend the observations to DQB and to examine possible correlations with clinical features of the disease. To this end, we have determined the incidence of HLA-DRB type on RFLP genotyping and the incidence of HLA-DQB alleles on sequence-specific oligonucleotide (SSO) polymerase chain reaction (PCR) amplification products (PCR-SSO analysis) in a series of 159 white northern European patients and controls. PATIENTS AND METHODS Patients. The 159 white European PBC patients comprised 147 women (92%)and 12 (8%)men. The control group of 181 white Europeans was drawn from laboratory and clinical staff the male-to-female ratio was approximately 1:I . The patients' age range was 22 to 77 yr for the women and 31 t o 84 yr for the men, with a unimodal age distribution and a median of 50 yr for men and women. Fifty of the patients were tissue-typed as part of liver transplant workup procedures. The remaining 109

patients were attending the liver follow-up clinic. Three of these patients subsequently underwent liver transplantation for end-stage liver disease. PBC was diagnosed on the basis of characteristic clinical and histologcal features and on the basis of' elevated biliary enzyme levels and antimitochondrial antibody titers more than 1:40. although these were not subtvped. Methods. DRB types were determined on Taql RFLP analysis; DQB types were determined with PCR-SSO analysis as previously described (29).DNA was prepared by washing the defrosted blood sample (anticoagulated with 0.05% EDTA) three times in 60 ml of &stilled water. The remaining pellet was digested in 10 ml of nuclear lysis buffer with protease K (Sigma Chemical Co., Poole. Dorset, U K ) , extracted with phenol/chloroform, dialyzed against TE buffer ( 10 mmol/L Tris LpH 7.41 and 1 mmoliL EDTA) and precipitated with ethanol and sodium acetate. Ten micrograms of genornic DNA was digested to completion with Taql restriction endonuclease (Gibco BRL, Paisley, Scotland), subjected to electrophoresis on 0.7% agarose gels and Southern-blotted onto nylon membrane (Hybond N + ; Amersham Ltd., Little Chalfont, Buckinghamshire, UK), which was probed with a radiolabeled cDNA probe specific for HLA DRB* (suppliedby J. Bidwell). RFLP band patterns were visualized with autoradiography on x-rav film (Amersham Ltd.). [Fig. 1A). Three hundred nanograms of genomic DNA was amplified on PCR using primers for the DQB gene. Aliquots of the product were subjected to electrophoresis on duplicate 1.5% agarose gels and Southern-blotted onto nylon membrane. The membranes were sequentially probed with 13 radiolabeled synthetic oligonucleotides specific for at least one DQB allele. Results were visualised by autoradiography on x-ray film (Fig. 1B).

1406

UNDERHILL ET AL.

HEPATOLOGY

TABLE 1. DRB RFLP

DQB allele showed a significant association with PBC when correction for multiple testing was applied. patients (%) Such strong linkage exists between HLA-DRB and Gene n = 169 n = 89 n = 162 n = 110 x’ HLA-DQB that some alleles are invariably coinherited (31). Although family studies are required for the DR1 48 (30) 47 (29) 0.05 rigorous assignment of haplotypes, in this study we have 33 (21) 46 (29) DR15 2.53 assumed the presence of both DR8 and DQB1*0402 in DR16 l(0.6) 3 (2) 0.98 an individual to represent the haplotype. These antigens 59 (37) DR3/6A 52 (32) 0.89 were found together in 9 of 89 patients (compared with 22 (14) DR3/6B 19 (12) 0.32 4 of 181 controls), giving a relative risk for the two 63 (40) DR4 74 (46) 1.20 locus haplotype of 5.4. No homozygotes for DR8 or DRll 19 (12) 2.19 11 (7) DR12 0.00 DQB1*0402 were found in either group. The PAR is a 2 (1) 2 (1) DR7 33 (21) 40 (25) 0.71 way of expressing the relative risk that takes into DR7/9 26 (16) 4.37 14 (9) account the normal frequency of the marker in the DR8 18 (11) 6.73” 6 (4) population (30). For both DR8 and DQB1*0402, the DRlO l(0.6) l(0.6) 0.00 individual PAR is 0.08, whereas for the combined 52 (32) DR52aJc 59 (37) 0.89 haplotype the PAR is 0.09. DR52b 35 (22) 40 (25) 3.32 We found no significant difference in antigen freDR53 122 (77) 128 (79) 0.36 quencies between those patients who underwent trans16 (18) DR3 25 (23) 0.68 plantation and those who did not. Furthermore, no 27 (30) DR6 23 (21) 2.32 association was detected between HLA-DR or HLA-DQ “p < 0.01. type and sex, age at first examination, stage of disease at first examination, or survival after date of first examination. No correlation was found between any disease SSO probe sequences were as described previously associated with PBC (e.g., scleroderma or sicca syn(291, with four additions: (a) 5’-GAGAGGACCCGGGC drome) and DR8, or with any other HLA. No. of

No. of controls (%)

GGA-3‘ (DQB1*0604,0605,0301,0302,0303);(b) 5’-GGGGT GTGACCAGACAC-3’ (DQB1*0501,0502,0503); (c) 5’-ACCG GCTCGTGCGGGG-3’ (DQB1*0401); and (d) 5’-AGTC TTGTAACCAGACAC-3’ (DQB1*0603,0604). Allele frequencies in patients and controls were compared by means of the x2 test. When previously unreported associations with individual alleles were identified, the Bonferroni inequality method was used to correct for multiple testing. Relative risk (RR) was calculated according to the crossed products ratio method and population attributable risk (PAR) according to the method of Begtsson and Thompson (30). On Taql RFLP, DR3 and DR6 are indistinguishable. Two band patterns are obtained for individuals with these types. One is characterized by an 11.1-kb band (DR52dc; labeled “3/6A”); the other by a 13.0-kb band (DR52b; labeled as 3/6B). Similarly, DR9 and some splits of DR7 share a common band pattern (labeled DR7/9). DR3 and DR6 may be differentiated on the basis of their linkage with known DQB alleles. For those samples in which this differentiation was possible, results are listed at the bottom of Table 1.

RESULTS The only DR antigen to show a significant increase was DR8 (Table 11, which was present in 11%of patients and in 4% of controls (RR = 3.3; p < 0.01). The p value was not corrected for multiple testing because the evaluation of the significance of DR8 was one of the main objectives of the study. In addition, we saw a decrease in the frequency of D R l l (7% vs. 12%) and DR15 (21% vs. 29%) and an increase in the frequency of DR7/9 (16% vs. 9%), although none of the latter were significant after correction for multiple testing. Similarly, only one DQB allele showed substantial variation from control frequencies (Table 2): DQB1*0402 was present in 10% of patients but just 3%of controls (RR = 3.51, although no

DISCUSSION This is the first report of an association between PBC and HLA-DR8 made with HLA genotyping methods. In addition, we report for the first time an association with HLA-DQB1*0402, although none of the 14 alleles we tested for showed a significant variation from control frequencies after correction for multiple testing. The increase in the observed frequency of DQB1*0402 mirrors that of DR8 because 9 of the 10 DR8 patients typed had both antigens as haplotypes (vs. four of six DR8 controls). These data confirm the findings of DR8 associations in previous studies (24, 25). However, the frequency of DR8 in the patient group was lower than that found by these studies: Gores et al. (24) detected a frequency of 30.1% (vs. 4.7%), and Manns et al. (25) stated a frequency of 36%(vs. 3.6%). This difference may be due to the difficulty in assigning DR8 by serology. This is particularly relevant in the study by Gores et al. because DR5 was reduced in their patients and DR5 cross-reacts with DR8. The RFLP typing method used in this study allows unequivocal assignment of these antigens. The lower frequency of DR8 in our patients compared with those in the American or German series may reflect differences in the populations studied, the criteria of diagnosis or, more probably, the accuracy of the techniques used. However, criteria for diagnosis of PBC are well accepted, and although there is some racial variation between British, German and American white populations we would not expect the differences in HLA frequencies to be so pronounced. In this study, we were unable to confirm the previously reported negative associations with DR5 (24) or DR2 (321, though the same antigens were present in

1407

t11,\ ('LASS I I XSSO('lAT1ON Ih P I K

Vol 16. No 6 . 1992

TABLE2. DQB PCR SSO analysis HLA allele

Patients

( ( < )"

Controls

("rP

9

DQB1*0501 DQB1*0502 DQB1*0503 DQB1*0504 DQB1*0601 DQB1*0602 DQB 1*0603 DQB1*0604 DQBl4201 DQBl*O301 DQBlt0302 DQB1=0303 DQBl*0401 DQB1*0402

:10

:+1 i

49

27

1.27 0.12 1.63 0.26 0.32 2.91 1.42 0.00 0.04 0 57 3.35 0. 13

RR for DR8 "n "n

= ~

3.3 and PAR

:

1 > 1

III

3 9 I 1 . 2i I1 12

28 ti ti

12

47

H3

46

_.,r,

"S

,59

I9

21

3:1 32

4

'I

1)

0

58 13 1

11

6

I c,

!I

I

1

1;

19 $1

I-,

9

0 08: l i H for I)HQIA0402

~

;J 5 and PAR

I

8 1

-

j.25

3

0.08; RK for DRR-DQB1*0402

~

5.4 and PAR

~

0.09

83. 161.

reduced frequencies in our patient group. Despite an increased frequency of DR719, we were also unable to confirm the strong association with DR7 found in the study by Lindgren et al. 1321. Unfortunately, Taql RFLP alone cannot differentiate DR9 from DR7. The low frequency of DR8 in the patient group and correspondingly low PAR value indicate that although DR8 is a risk factor in the development of the disease, this may only be true for a small proportion of patients. A possible explanation for the relatively weak association is that DR8 is a marker for a susceptibility gene i n the HLA region. The higher relative risk of the DRS-DQB1*0402 haplotype would support this hypothesis. The lack of association with HLA class I suggests that any locus contributing to disease susceptibility would not lie in this (telomeric) direction on the chromosome relative to HLA-DR. However, t.he HLA class I11 region also lies telomeric to HLA-DR, and several strong asociations with genes in this reeon have been reported, including C4B ( 21J and TNF-cuI 22). This is of particular interest because patients have been shown to have abnormal circulating complement levels. Neither the C4B2 nor TNF-tdNcoI 5.5-kb markers associated with PBC are known to be in linkage disequilibrium with HLA-DR8 or DQB*0402. Curiously, in the Danish study, the TNF variant was associated with the autoimmune haplotype A1 -B8-DR3 (22,. The HLA-DQ genes map immediately centromeric to HLA-DR and are in very close linkage, making it difficult to separate candidate susceptibility genes. as we have seen. Further in the centromeric direction is HLA-DP, separated from HLA-DR and HLA-DQ by several other genes. Recent discoveries show that non-HLA genes in this region have functions related to antigen processing and presentation (33).Even so: recombination between DR/DQ and DP occurs with a frequency of less than 3%. such that the entire range of HLA and related genes is almost always coinherited as a complete haplotype (34). An alternative explanation for the differences between

studies is that those patients with DR8 and DQB1*0402 may represent a discrete clinical subgroup of PBC more common in the United States than in Great Britain. We have been as yet unable to identify any discrete clinical features in patients with or without DR8. HLA may be involved in the pathogenesis of a disease in several ways. The HLA molecule itself, perhaps with a tissue-specific antigen bound, may be the target of immune attack or may be the receptor by which a pathogen gains access to a cell. Current dogma suggests that HLA alleles vary in their ability to present antigen I 35 ). Thus differing alleles may have differing affinities for a range of peptides, including those of putative pathogens. The disease process itself may be determined by the extent of response to a particular peptide. This is likely to involve interaction with one of the mitochondrial antigens - for example, the E2 antigen, which has recently been found to be expressed on cultured biliary epithelial cells from patients with PBC (4). Although these data confirm an immunogenetic predisposition to PBC, the large number of patients without DR8 or DQB1*0402 suggests that other factors are involved in disease pathogenesis. Such factors may be genetic - including both immunological and metabolic genes - or environmental, perhaps causing the expression of E2 on biliary epithelial cells. KEFEKENCES 1 Fiezi T. Immunoglobulins in chronic liver disease. Gut 1968:9:

193-198. 2 Berg PA, Doniach D. Roitt IM. Mitochondrial antibodies in primary biliary cirrhosis: localization of the antigen to mitochondrial membranes. J Exp Med 1967;126:277-299, 3 Mackay IR, Gerschwin ME. Molecular havis of mitochondria1 autoreactivity in primary hiliary cirrhosis. Immunol Today 1989; 10:315-316. -1 Joplin K, Lindsay J G , Johnson GD, Strain A, Neuberger J . Membrane dihydrolipoamide acetyltransferase E2 on human hiliary epithelial cells in primary hiliary cirrhosis. Lancet 1992: 33993-94

1408

UNDERHILL ET AL.

5. Walker JG, Doniach D, Roitt IM. Serological tests in the diagnosis of primary biliary cirrhosis. Lancet 1965;1:827-831. 6. Christensen E, Crowe J , Doniach D, Popper H, Ranek L, Rodes J, Tygstrup N, et al. Clinical pattern and course of disease in primary biliary cirrhosis based on analysis of 236 patients. Gastroenterology 1980;78:236-246. 7. James OFW, Myszor M. Epidemiology and genetics of primary biliary cirrhosis, Prog Liver Dis 1990;9:523-536. 8. Ballardini G, Bianchi FB, Doniach D, Miriakian R, Pisi E, Bottazzo GF. Abberant expression of HLA DR antigens on bile duct epithelium in primary biliary cirrhosis: relevance to pathogenesis. Lancet 1984;2:1009-1013. 9. Nakanuma Y, Naoko K. Expression of HLA DR antigens on interlobular bile ducts in primary biliary cirrhosis. Hum Pathol 1991;22:431-436. 10. Galbraith RM, Eddleston ALWF, Smith MG, Williams R, McSween RNM, Watkinson G, Dick H, et al. Histocompatibility antigens in active chronic hepatitis and primary biliary cirrhosis. BMJ 1974;3:604-605. 11. Ercilla G, Pares A, Arriga F, Brugera M, Castillo R, Rodes J , Vives J. Primary biliary cirrhosis associated with HLA DR3. Tissue Antigens 1979;14:449-452. 12. Hamlyn AN, Adams D, Sherlock S. Primary or secondary sicca complex? Investigation in primary biliary cirrhosis by histocompatibility testing. BMJ 1980;281:425-426. 13. Miyamori H, Kato Y, Koboyashi K, Hattoi N. HLA antigens in Japanese patients with primary biliary cirrhosis and autoimmune chronic active hepatitis. Digestion 1983;26:231-217. 14. Bassendine MF, Dewar PJ, James OFW. HLA DR antigens in primary biliary cirrhosis: lack of association. Gut 1985;26: 625-628. 15. Johnston DE, Kaplan MM, Miller KB, Conners CM, Milford EL. Histocompatibility antigens in primary biliary cirrhosis. Am J Gastroenterol 1987;82:1127-1129. 16. Modena V, Marengo C, Amoroso A, Rosina F, Costatini P, Bellandon P, Coppo R, et al. Primary biliary cirrhosis and rheumatic diseases: a clinical immunological and immunogenetical study. Clin Exp Med 1986;4:129-134. 17. Mackay IR, Morris PJ. Association of autoimmune chronic hepatitis with HL-A1:8. Lancet 1972;2:793-795. 18. Chapman RW, Varghese Z, Gaul R, Pate1 G, Kokinon N, Sherlock S. Association of primary sclerosing cholangitis with HLA B8. Gut 1983;24:38-41. 19. Donaldson PT, Farrant JM, Wilkinson ML, Hayllar KM, Portmann BC, Williams R. Dual association of HLA DR2 and DR3 1991;13: with primary sclerosing cholangitis. HEPATOLOGY 129-133. 20. Donaldson PT, Doherty DG, Hayllar KM, McFarlane IG, Johnson PJ, Williams R. Susceptibility to autoimmune chronic active hepatitis: human leukocyte antigens DR4 and Al-Bg-DR3 are independent risk factors. HEPATOLOGY 1991;13:701-706.

HEPATOLOGY

21. Briggs DC, Donaldson PT, Hayes P, Welsh KI, Williams R. A major histocompatibility complex class 111 allotype, C4B2, associated with primary biliary cirrhosis. Tissue Antigens 1987;29:141-145. 22. Fugger L, Morling N, Ryder LP, Platz P, Georgson J , Jackobson BK, Svejgaard A, et al. Nco I RFLP of tumor necrosis factor alpha region in primary biliary cirrhosis in healthy Danes. S a n d J Immunol 1989;30: 185-189. 23. Fugger L, Morling N, Ryder LP, Jakobson BK, Andersen V, Oxholm P, Dalhott K, et al. RFLP of two HLA B associated transcripts genes in five autoimmune diseases. Hum Immunol 1991;30:27-3 1. 24. Gores GJ, Moore SB, Fisher LD, Powell FC, Dickson ER. Primary biliary cirrhosis: association with class I1 major histocompatibility complex antigens. HEPATOLOGY 1987;7:889-892. 25. Manns MP, Bremm A, Schneider PM, Notghi A, Gerken G, Prager-Eerbele M, Stradmann-Bellinghausen B, et al. HLA DRw8 and complement C4 deficiency as risk factors in primary biliary cirrhosis. Gastroenterology 1991;101:1367-1373. 26. Bidwell JL, Bidwell EA, Savage DA, Middleton D, Klouda PT, Bradley BA. A DNA RFLP typing system that positively identifies serologically well defined and ill defined HLA DR and DQ alleles, including DRwlO. Transplantation 1988;45:640-646. 27. Opelz G, Mytilineos J , Scherer S, Dunckley H, Trejaut J , Chapman J, Middleton D, et al. Survival of DNA HLA DR typed and matched cadaver kidney transplants. Lancet 1991;338:461-463. 28. Bidwell J . DNA-RFLP analysis and genotyping of HLA-DR and DQ antigens. Immunol Today 1988;9:18-23. 29. Doherty DG, Donaldson PT. HLA DRB and DQB typing by a combination of serology, RFLP and PCR SSO probing. Eur J Immunogenet 1991;18: 111-124. 30. Bengtsson BO, Thompson G. Measuring the strength of associations between HLA antigens and disease. Tissue Antigens 1981; 18:356-363. 31. Morel C, Zwahlen F, Jeannet M, Mach B, Tiercy J-M. Complete analysis of HLA-DQB1 polymorphism and DR-DQ linkage disequilibrium by oligonucleotide typing. Hum Immunol 1990;29: 64-77. 32. Lindgren S, Kockum I, Lernmark A, Sundkvist G, Wassmuth R, Eriksson. HLA DR gene polymorphisms in primary biliary cirrhosis [Abstract]. J Hepatol 1991;13(suppl 2):165. 33. Trowsdale J , Campbell RD. Complexity in the major histocompatibility complex. Eur J Immunongenet 1992;19:45-55. 34. Ronningen KS, Spurkland A, Markussen G, Iwe T, Vardtal F, Thorsby E. Distribution of HLA class I1 alleles among Norwegian Caucasians. Hum Immunol 1990;29:275-281. 35. Rudensky AY, Preston-Hurlburt P, Hong SC, Barlow A, Janeway CA Jr. Sequence analysis of peptides bound to MHC class I1 molecules. Nature 1991;353:622-627.