Atagunduz et al (1) suggest that self peptides derived from cartilage proteins can be bound by HLA–B27 molecules, serve as targets for autoimmune cytotoxic T cells, and play a role in the pathogenesis of ankylosing spondylitis (AS). They identified a single peptide (C34, DRASFIKNL) which was stimulatory for HLA–B27–restricted CD8+ T cells from some AS patients. However, as the authors point out, an N-terminal acidic residue has so far not been observed in peptides eluted from HLA–B27 complexes (2), making it crucial to demonstrate the unequivocal existence of the HLA–B27/C34 complex in order to establish C34 as an immunodominant candidate arthritogenic peptide (1).
Atagunduz et al attempt to provide the above-described evidence using a refolding protocol for HLA–B27 complexes (Figure 2 in their report). This protocol differs from the commonly employed procedure (3) in several important aspects, however, most notably the absence of a redox system. With the standard procedure, crystallization of a wide range of HLA/peptide complexes, including various HLA–B27 molecules (ref. 4 and references therein), has been achieved. This is not the case with the system used by Atagunduz et al (1). Chromatographic separation of the protein mixture yielded a minute peak at 13.7 ml, which the authors regarded as the refolded, trimeric HLA–B27/C34 complex. According to the column manufacturer, this position corresponds to a MW of ∼30,000 and may therefore not represent an intact HLA heavy chain/β2-microglobulin/peptide complex (expected MW ∼45,000). Insufficient description of the experimental details also does not allow for an assessment of its composition as judged by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (1). Therefore, the authors' assumption that the product peak represents an HLA–B27/C34 complex remains unsubstantiated (3).
In addition, structural comparisons (Figure 1) indicate why an acidic N-terminal residue has never been found among HLA–B27–bound peptides, while Arg or Gly is preferred (2). Substituting Asp for Arg at peptide position 1 is feasible in principle, as judged from the x-ray crystallographic structure of B*2705/TIS (protein data bank accession code 1w0v) (4). However, inspection of the 2 binding modes (Figure 1) reveals that the accommodation of peptides with an N-terminal Asp will probably not permit a stable interaction of such ligands in vivo, making it unlikely that the C34 peptide can serve as a viable target for cytotoxic T cells. Furthermore, considering the entire C34 sequence, with its lack of an Arg in the middle of the peptide and a C-terminal residue (Leu) that is well tolerated by all HLA–B27 subtypes studied to date, it appears very unlikely that C34 would be differentially bound or displayed by HLA–B27 subtypes. These subtypes differ in their association with AS, as described in the arthritogenic peptide hypothesis (refs. 2 and 4 and further references therein).
In summary, the lack of biochemical evidence for the existence of an HLA–B27/C34 complex in vitro, as well as our structural and modeling analyses, do not support a role for C34 as an immunodominant candidate arthritogenic peptide in the context of HLA–B27–associated AS (1).