Modelling of HLA-DQ2 and its interaction with gluten peptides to explain molecular recognition in celiac disease

https://doi.org/10.1016/j.jmgm.2004.12.002Get rights and content

Abstract

Celiac disease (CD) is sustained by abnormal intestinal mucosal T-cell response to gluten and it is strongly associated with HLA class II molecules encoded by DQA1 *0501/DQB1 *02 (DQ2) or DQA1 *03/DQB1 *0302 (DQ8). The in vitro stimulatory activity of gliadin increases after treatment with tissue transglutaminase (tTG) which catalyses the deamidation of specific residues of glutamine to glutamate that can serve as anchors for binding to DQ2 as well as to DQ8 molecules. We modelled the three-dimensional structure of the DQ2 dimer protein, the most frequent in celiac patients, by using a homology modelling strategy, and deposited the model in the Protein Data Bank (PDB). Then, we simulated the interactions of DQ2 with different gluten peptides and the deamidation of specific peptide glutamines in the known p4, p6, p7 and p9 anchor positions, as well as in p1 and p5 positions, and other substitutions for which experimental effects on binding are available by previous experimental studies. By evaluating the energy of interaction and the H-bond interactions, we were able to distinguish what substitutions improve the interaction peptide–DQ2, in agreement with previously published experimental data. By analysing the peptide–DQ2 complex at the atom level, we observed that these glutamate side chains can interact with specific positively charged amino acids of DQ2, absent in other HLA alleles not related to celiac disease. The simulation was also extended to other peptides, related to celiac disease but for which no experimental data exists about the effects of glutamine deamidation. Our results give an interpretation at the molecular level of previously reported binding experimental data and open a new window to gain further insights about peptide recognition in celiac disease.

Introduction

Celiac disease (CD), or celiac sprue, or gluten-sensitive enteropathy, is a multifactorial disorder influenced by both environmental and genetic factors. It is characterized by malabsorption resulting from inflammatory injury to the mucosa of the small intestine after the ingestion of wheat gluten or related rye and barley proteins. It often starts shortly after the first introduction of wheat into the diet, and symptoms include diarrhoea, malabsorption, and failure to thrive, which is due to an inefficient uptake of nutrients by a flattening intestinal epithelium. At present the only effective treatment for the disease is the removal of gluten from the diet, but the reintroduction of gluten in the patient's diet invariably leads to the reappearance of the symptoms [1], [2], [3]. Although the molecular basis of CD is still unclear, the molecular mechanism is considered to involve the binding of gluten peptides to HLA molecules and then the specific recognition by T-cells [2]. Different gluten-derived peptides have been identified as able to be recognized by T-cell clones isolated from biopsies of CD patients [4], [5], [6]. HLA-DQ2 (DQA1 *0501/B1 *0201) is found in the great majority of CD patients, while DQ8 (DQA1 *0301/B1 *0302) is found in most of the remaining patients [1], [2]. The binding of gluten peptides to DQ2 and DQ8 molecules has been experimentally observed and it may be improved by the presence of negatively charged amino acids at known positions in the peptide [1]. Gluten proteins are very rich of glutamine and proline and contain very few negatively charged amino acids, but glutamines may be deamidated to the negatively charged glutamate by the tissue transglutaminase (tTG) enzyme, with the consequence of improving both the binding to DQ2 and the response of T-cell clones [7], [8], [9].

Different studies have been carried out in order to simulate the interaction of gluten peptides with DQ2. Some simulations were started from the knowledge of the experimental model of the DR1 dimer and based on models of the DQ2 molecule obtained by simple modelling procedures as the amino acids substitutions [10], [11]. In two other works, a probably more complete but not described homology modelling procedure was applied starting from the DR1 [12] or DQ8 dimer [13]. The assessment of more efficient modelling procedures as well as the availability of the experimental structure of the DQ8 dimer, which is more similar to DQ2 and represents a better template model than DR1, makes it possible to improve the methods previously applied and to perform more accurate simulations of the interaction of gluten peptides with DQ2, as well as an accurate analysis of the effect of deamidation and other modifications of the peptides.

In our study, we used bioinformatics and biocomputing approaches and tools for predicting the 3D structure of DQ2 molecule, validate it and deposit the model in the PDB [14], in order to make the model available to the scientific community. Then, we simulated the complex between DQ2 and different gluten-derived peptides and their modified forms, to investigate the molecular details of this interaction and the effects of modification of specific peptide amino acids, in order to propose an interpretation at the molecular level of published experimental results concerning the binding of gluten peptides to DQ2 and the effects of deamidation. We also applied the same strategy to other gluten peptides for which a relation to celiac disease has been established, but no experimental evidence exists about deamidation of glutamines. Finally, to find further information about the specificity of DQ2 recognition, we analysed the presence of functional groups in the peptide-binding site, and evidenced the relevance of DQ2 structural peculiarities in comparison to other DQ alleles not related to celiac disease. We also discuss the results of our computational work in comparison to the very recently published article describing the experimental structure of the complex between DQ2 and an immunogenic epitope from gluten [15].

Section snippets

Protein modelling

Protein sequences of DQ2 chains used for protein modelling were selected from SwissProt database (DQA1 *0501 chain: accession number P01909; DQB1 *0201 chain: accession number P01918). The 3D model of DQ2 was created following the homology modelling procedure [16], [17], [18], [19] already used with success by our group [20], [21], [22], [23], [24]. In brief, searches for sequence similarity within databases were performed with the BLAST program [25]. The searches evidenced that a very high

Homology modelling of DQ2

The sequences of both DQ2 chains have been analyzed by computer programs in order to find similar sequences in databases and perform structural predictions to obtain a theoretical 3D model of the protein. The sequences of both DQ8 chains were found to be the most similar to the corresponding DQ2 chains, in terms of sequence similarity, having 91% of sequence identity for both alpha and beta chains (see Fig. 1). In these conditions, a homology modelling strategy can be applied with very good

Comparison to published experimental structure of DQ2

We modelled the DQ2 structure and deposited it in the Protein Data Bank (PDB code: 1NBN) before the publication and availability of an experimental structure of DQ2 [15]. The comparison between the predicted model and the experimental structure confirms that the homology modelling procedure has been correctly applied. In fact, as shown in Fig. 2B and C, the superposition of chain backbones and secondary structures are very good. The few differences in the secondary structure assigned are mainly

Conclusions

For many years, studies at the molecular level of the interaction between DQ2 and gluten peptides were based on molecular simulations and modelling of DQ2. In order to improve the previous molecular modelling results by using the most recent and reliable approach, we created a model of DQ2 based on a better homologous model and by creating a full atom model instead of simple side chain substitution, and deposited the model in PDB to make it available to the scientific community for further

Acknowledgements

Ph.D. fellowship of Dr. Susan Costantini is supported by E.U.

References (39)

  • H. Arentz-Hansen et al.

    Celiac lesion T cells recognize epitopes that cluster in regions of gliadins rich in proline residues

    Gastroenterology

    (2002)
  • W. Vader et al.

    The gluten response in children with celiac disease is directed toward multiple gliadin and glutenin peptides

    Gastroenterology

    (2002)
  • L.M. Sollid

    Molecular basis of celiac disease

    Ann. Rev. Immunol.

    (2000)
  • L.M. Sollid

    Celiac disease: dissecting a complex inflammatory disorder

    Nat. Rev. Immunol.

    (2002)
  • H. Quarsten et al.

    HLA binding and T cell recognition of a tissue transglutaminase-modified gliadin epitope

    Eur. J. Immunol.

    (1999)
  • A.K. Moustakas et al.

    Structure of celiac disease-associated HLA-DQ8 and non-associated HLA-DQ9 alleles in complex with two disease-specific epitopes

    Int. Immunol.

    (2000)
  • H. Arentz-Hansen et al.

    The intestinal T cell response to a-gliadin in adult celiac disease is focused on a single deamidated glutamine targeted by tissue transglutaminase

    J. Exp. Med.

    (2000)
  • O. Molberg et al.

    Tissue transglutaminase selectively modifies gliadin peptides that are recognized by gut-derived T cells in celiac disease

    Nat. Med.

    (1998)
  • Y. van de Wal et al.

    Selective deamidation by tissue transglutaminase strongly enhances gliadin-specific T cell reactivity

    J. Immunol.

    (1998)
  • Cited by (34)

    • Redesigning the reactive site loop of the wheat subtilisin/chymotrypsin inhibitor (WSCI) by site-directed mutagenesis. A protein-protein interaction study by affinity chromatography and molecular modeling

      2009, Biochimie
      Citation Excerpt :

      CVFF force field developed for InsightII was adopted to assign potentials and charges for the protein complexes. The complex models were minimized by using 500 steps of energy minimization under conjugate gradient algorithm in order to optimize side-chain conformations and to avoid sterical clashes, according to the procedure used in previous similar works [22–25]. The presence of putative intra- and inter-chain H-bonds was analyzed with the HBond tool of InsightII in order to evaluate the enzyme–inhibitor interaction in the examined complexes.

    • Molecular modelling of co-receptor CD8αα and its complex with MHC class I and T-cell receptor in sea bream (Sparus aurata)

      2008, Fish and Shellfish Immunology
      Citation Excerpt :

      The structure of sea bream CD8αα dimer was assembled by superimposing the two best modelled chains of sea bream CD8α1 and CD8α2 to the corresponding murine CD8 chains, in order to obtain the same relative orientation of the two subunits. The CVFF force field within the Discover module of Insight II was used to assign potentials and charges, and a mild energy minimization was applied, by performing 500 steps under the conjugate gradient algorithm, in order to optimise the interaction between the two sea bream protein chains and avoid sterical clashes according to the procedure used in our previous similar work [29]. The modelling of sea bream MHC-I N-terminal region (20–200 amino acid region, accession number: DQ211541.1) was performed using as template the human and murine experimental structures of MHC-I complexed with TCR β (PDB code: 2GJ6, chain A, and 1G6R, chain H) [2,3].

    View all citing articles on Scopus
    View full text