Cutting Edge: HLA-B27 Can Form a Novel β₂-Microglobulin-Free Heavy Chain Homodimer Structure¹

There is a striking association between HLA-B27 and a group of closely related arthritic diseases including ankylosing spondylitis (1). Despite extensive studies, its role in arthritis remains obscure. HLA-B27 is normally found in noncovalent association with β₂-microglobulin (β₂m),⁴-presenting peptides for recognition by CTL. Peptide binds within a groove formed by the α1 and α2 domains of the protein, supported by the α3 domain and β₂m (2).

A direct role for HLA-B27 in arthritis could involve peptide presentation or some unique structural feature (3). The discovery that HLA-B27 transgenic mice develop arthritis in the absence of β₂m and MHC class II implied that “free” HLA-B27 heavy chains might present peptide (4, 5). However, although certain mouse alleles possess this ability (6), no human class I protein has been shown to present Ag in the absence of β₂m.

One unusual feature of HLA-B27 is an unpaired cysteine residue (Cys⁶⁷) in the extracellular α1 domain. Cys⁶⁷ is not essential for maintaining the structure of the molecule (7), although its location above the B peptide anchor pocket (8) would suggest that it influences peptide binding. It has previously been suggested that in vivo modification of Cys⁶⁷ may lead to an autoimmune response (9).

We show that HLA-B27 can form a Cys⁶⁷-dependent heavy chain homodimer (HC-B27) in the absence of β₂m. HC-B27 complexes are recognized by the conformation-specific Ab W6/32. Disulfide-bonded W6/32-reactive-free heavy chains are also present on HLA-B27-transfected T2 cells.

Plasmids pLM1-HLA-B27 and pHN1-β₂m encoding the extracellular domain of HLA-B27 and β₂m were a gift from D. Wiley and D. Garboczi (Harvard University, Cambridge, MA). Plasmid B27B was generated from pLM1-HLA-B27 using primers 5′-TTTGTTGAATTCAGGAGGAAT-3′ and 5′-TTGTGATAAGCTTAACGATGATTCCACACCATTTTCTGTGCATCCAGAATATGATGCAGGGATCCCTCCCAT-3′. Plasmids RAB27S and RAB27SB were generated from pLM1-HLA-B27 and RAB27B using primers 5′-GAGACACAGATCAGCAAGGCCAAGGC-3′ and 5′-GCCTTGGCCTTGCTGATCTGTGTCTC-3′ with the Quikchange system (Stratagene, La Jolla, CA). Proteins were expressed and folded as described (10). Peptide was the HIV-1 gag epitope KWRIIMGLNK. Complexes were purified on a fast protein liquid chromatography system from Pharmacia (Piscataway, NJ); m.w. standards were from Pharmacia.

Gels were blotted onto Hybond-C membrane (Amersham, Little Chalfont, U.K.), stained with HC10 Ab (gift from H. Ploegh, Harvard University), and detected using HRP-conjugated rabbit anti-mouse IgG (Sigma) with ECL reagent (Amersham).

ELISA was developed using HRP-conjugated rabbit anti-mouse IgG (Sigma) with 3,3′,5,5′-tetramethyl benzidine substrate (Sigma, St. Louis, MO).

Cells were labeled with sulfo-NHS-biotin (Pierce, Rockford, IL). Cells were lysed in the presence of 5 mM iodoacetamide. Immunoprecipitates resolved by SDS-PAGE were blotted onto Hybond C and detected as above or using ExtrAvidin-HRP (Sigma) with ECL reagent.

The extracellular domain of HLA-B27 was used in refolding studies. Unlike other alleles (11), HLA-B27 did not fold with β₂m and known epitopes. Reasoning that this might be due to the unpaired cysteine, we mutated Cys⁶⁷ to serine (HLA-B27^Ser67). HLA-B27^Ser67 refolded around β₂m and peptide, allowing us to generate tetramer reagents for T cell studies (12). Further analysis of profiles for refolds of HLA-B27^Cys67 showed an unexpected shoulder of 60–70 kDa. HLA-B27 refolded in the absence of β₂m generated a single 66-kDa peak which resolved as a 33-kDa band by reducing SDS-PAGE (Fig. 1,A). This peak was absent from refolds using mutated HLA-B27^Ser67 (Fig. 1,B). Western blot confirmed that the 66-kDa peak was composed of class I heavy chain (Fig. 2,A), and its altered mobility under nonreducing conditions indicated disulfide bonding. HC-B27 formed in the absence of added peptide but were unstable (Fig. 2 B). Addition of the HIV gag epitope KRWIIMGLNK but not the HLA-B27-restricted influenza NP epitope SRYWAIRTR or the HLA-A2-restricted CMV peptide NLVPMVATV to refolds stabilized HC-B27. Acid elution analysis confirmed that the HIV gag peptide was bound by HC-B27 (data not shown).

Circular dichroism identified secondary structure within the HC-B27 fraction, confirming that complexes were not randomly aggregated (data not shown). We studied the structure of HC-B27 using the heavy chain-specific Ab HC10 and the conformation-specific Abs W6/32 and ME1. HC-B27 showed strong HC10 reactivity (data not shown). W6/32 recognizes residues in the α2 helix of native class I complexes (13), and ME1 recognizes an epitope in the α1 helix of HLA-B27 (3). W6/32 reacted comparably with both HC-B27 and HLA-B27^Ser67/β₂m complexes (Fig. 3,A), demonstrating that HC-B27 can maintain some conformation of its peptide-binding groove. HC-B27 consistently failed to bind ME1. Despite low levels of binding, Ab titration showed that HLA-B27^Ser67 heterodimers were recognized by ME1 (Fig. 3 B).

The T2 mutant cell line retains most MHC class I allotypes within the endoplasmic reticulum in a W6/32-unreactive form. T2-HLA-B27 transfectants, however, show unusually high W6/32 reactivity (14, 15), despite low ME1 binding (16). After surface labeling, reducing SDS-PAGE identified β₂m in the W6/32 precipitates of control C1R-HLA-B27 cells, but not in those of T2-HLA-B27 transfectants (Fig. 4,A). Similarly, a 45-kDa heavy chain band was seen in T2-HLA-B27 precipitates but was absent from untransfected cells. High m.w. bands seen in T2-HLA-B27 precipitates may represent the HLA-B27 coprecipitates previously observed for TAP-deficient cells (17). Western blot of T2-HLA-B27 lysates showed that W6/32 precipitates ran at positions consistent with heavy chain tetramers under nonreducing conditions (Fig. 4 B). Tetramerization could be explained by disulfide bonding through both Cys⁶⁷ and a second unpaired cysteine in the cytoplasmic tail of the molecule. Class I heavy chains bonded through this residue have been documented previously (18). This could not have taken place in our refolding studies, which used the extracellular domains of HLA-B27.

Our findings may be relevant to studies that show that HLA-B27 transgenic mice lacking β₂m develop an arthritis the incidence of which can be reduced by treatment with HC10 Ab (5). HC-B27 formation is dependent on disulfide bonding through Cys⁶⁷. Such bonding would almost certainly require unwinding of the α1 helix, consistent with loss of the ME1 epitope. However, HC-B27 complexes are recognized by W6/32 and retain an ability to bind peptide, implying that at least part of the groove structure remains intact. Failure of the influenza NP epitope to bind HC-B27 may be explained by alterations to the peptide-binding groove, and it is possible that HC-B27 binds a subset of classically presented epitopes. An interesting possibility is that partial unwinding of the α1 helix could open the groove to accommodate longer peptides. Long peptides were previously shown to bind a subset of HLA-B27 molecules (19). A partially unwound α1 helix might also resemble the α-chain helix of an MHC class II molecule, raising the possibility of CD4 T cell recognition.

Unlike free heavy chains found on the surface of activated T and B cells that lack W6/32 reactivity (20, 21), we have shown that W6/32-reactive heavy chains can be expressed at the surface of T2-HLA-B27. Altered forms of HLA-B27 may be readily detected on this line due to inhibition of normal class I expression. Similar forms may be present on cells such as C1R-HLA-B27 but at too low a level to be easily detected. In that the T2 line is deficient in TAP-transported peptides, it may be that empty HLA-B27 heavy chains can progress to the cell surface. Alternatively, structural alterations could allow peptides from another source to bind. Progression of heavy chains through the endocytic pathway is consistent with their surface expression in β₂m knockout mice (4) and Ag-processing mutants (16).

Aberrant molecules such as HC-B27 could theoretically trigger disease if, e.g., expressed as a neoantigen under certain conditions. The ME1 epitope can be lost during spondyloarthropathy (22, 23) or bacterial infection (24), and it previously has been suggested that Cys⁶⁷ modification could trigger autoimmunity (11, 25). Alternatively, HC-B27 could act through peptide presentation. It will be important to determine whether HC-B27 is recognized by the immune system and, if so, as an Ag-presenting molecule or as a novel autoantigen.

We have shown that HLA-B27 can form a stable homodimer structure (HC-B27). Despite the absence of β₂m, HC-B27 shares some tertiary structure of its peptide-binding groove with conventional complexes and appears to be capable of peptide binding. Similar structures can be expressed on the surface of HLA-B27-transfected cells. HC-B27 is of particular interest due to its potential role in the pathogenesis of HLA-B27-associated disease.

We thank D. Wiley, D. Garboczi, E. Collins, and M. Bouvier for plasmids and advice; M. Gross for spectrophotometry; V. Cerundolo, T. Elliott, Y. Jones, and N. Zaccai for discussions; and J. Edwards, G. Gillespie, J. Mongkolsapaya, G. Ogg, L. Schimanski, L. Tan, B. Willcox, J. Wilson, and J. Wyer for protocols and reagents.