Association of HLA-DR1 with the allergic response to the major mugwort pollen allergen: molecular background

Background Mugwort pollen allergens represent the main cause of pollinosis in late summer. The major allergen, Art v 1, contains only one single immunodominant, solely HLA-DR-restricted T cell epitope (Art v 125-36). The frequency of HLA-DRB1*01 is highly increased in mugwort-allergic individuals and HLA-DR1 serves as restriction element for Art v 125-36. However, Art v 125-36 also binds to HLA-DR4 with high affinity and DR1-restricted Art v 125-36 -specific T cell receptors can be activated by HLA-DR4 molecules. To understand the predominance of HLA-DR1 in mugwort allergy in spite of the degeneracy in HLA/peptide-binding and TCR-recognition, we investigated the molecular background of Art v 125-36 /MHC/TCR interactions in the context of HLA-DR1 compared to -DR4. Results The majority of Art v 125-36 -specific T cell lines and clones from HLA-DR1 carrying, mugwort pollen-allergic donors reacted to synthetic and naturally processed Art v 1–peptides when presented by HLA-DR1 or HLA-DR4 expressing antigen presenting cells. However, at limiting peptide concentrations DR1 was more effective in T cell stimulation. In addition, the minimal epitope for 50% of Art v 125-36 -specific T cells was shorter for DR1 than for DR4. In vitro binding assays of Art v 125-36 mutant peptides to isolated DR1- and DR4-molecules indicated similar binding capacities and use of the same register. In silico simulation of Art v 125-36 binding to HLA-DR1 and -DR4 suggested similar binding of the central part of the peptide to either molecule, but a higher flexibility of the N- and C-terminal amino acids and detachment at the C-terminus in HLA-DR1. Conclusions The predominance of HLA-DR1 in the response to Art v 125-36 may be explained by subtle conformation changes of the peptide bound to DR1 compared to DR4. Computer simulation supported our experimental data by demonstrating differences in peptide mobility within the HLA-DR complex that may influence TCR-binding. We suggest that the minor differences observed in vitro may be more relevant in the microenvironment in vivo, so that only presentation by HLA-DR1, but not -DR4 permits successful T cell activation.

Background CD4+ T helper (Th) cells play a major role in the induction and maintenance of IgE-mediated allergy [1], which is characterized by an excessive Th2 response. The Th2cytokines, interleukin-4 and interleukin-13, lead to the production of allergen-specific IgE antibodies which bind to high affinity receptors on effector cells. Upon renewed allergen contact, cross-linking of these IgE antibodies leads to the release of inflammatory mediators, which cause allergic symptoms. Allergen-specific CD4+ Th cells recognize their cognate peptide via their T cell receptor (TCR) in the context of self major histocompatibility complex (MHC) class II molecules on the surface of antigen presenting cells (APC). Binding of the processed peptides to the human leukocyte antigen (HLA) class II molecules HLA-DR, -DP or -DQ occurs in endosomal compartments and is facilitated by accessory molecules. These HLA class II molecules are highly polymorphic within the population and for many immunogenic peptides degenerate binding to different MHC-class II molecules has been described. This feature is of interest for instance in vaccine development as it would allow presentation of such T cell epitopes on APC within a wide population range [2,3]. Besides, also for TCR a plasticity in ligand recognition, i.e. degeneracy in peptide reactivity as well as cross-recognition in the context of distinct MHC-molecules has been found [4,5].
In this study, we investigated the molecular background leading to the preference of HLA-DR1 in the allergic response to the single immunodominant peptide of Art v 1. Potential differences between HLA-DR1 and -DR4 molecules regarding peptide binding and requirements for TCR recognition were addressed. We compared i) presentation of Art v 1 25-36 by APC expressing either HLA-DR1 or -DR4 to allergen-specific T cell lines (TCL) and clones (TCC) and ii) in vitro binding of Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] to HLA-DR 1 and -DR4 molecules using mutated peptides; furthermore, iii) we performed computer simulations of Art v 1 25-36 -binding to both HLA-molecules using Molecular Dynamics (MD) simulations.
We found that the majority of Art v 1-specific T cells recognize synthetic Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] and variants thereof as well as naturally processed peptides presented by DR1 and DR4 in a similar manner. Differences in minimal epitope recognition suggested structural differences for binding to the MHC peptide binding groove. MD simulation supported this conclusion demonstrating that the peptide is much more flexible at both ends and shows a detached structure at the C-terminus, whereas in DR4 the peptide keeps its extended conformation. Together, the strong bias of HLA-DR1 in the response to the immunodominant epitope of Art v 1 apparently is not primarily due to MHC/peptide binding per se, but rather seems to be based on structural differences in the TCRbinding site of the peptide/MHC (pMHC) complex.

Allergen and peptides
Natural Art v 1 [6] was kindly provided by Matthias Egger (University of Salzburg, Austria). Synthetic peptides were obtained from Thermo Electronics (Ulm, Germany). Purity was >90% as confirmed by HPLC.

Allergen-specific T cell lines (TCL) and T cell clones (TCC)
Art v 1 25-36 -specific TCL were established from peripheral blood mononuclear cells (PBMC) by allergen-specific stimulation with either mugwort pollen extract or recombinant Art v 1 (Biomay, Vienna, Austria) as previously described [7]. Art v 1 25-36 -specific TCC were established from these TCL by limiting dilution [7]. The two EBV-transformed B cell lines used as APC expressed comparable amounts of HLA-DR molecules ( Figure 1A) as well as HLA-DQ, -DP-and the co-stimulatory molecules CD80, CD86, CD40 (not shown) on their surface. TCL or TCC (5 × 10 4 cells) were stimulated for 3 days with irradiated (60 Gy) APC and peptides at an optimum concentration of 3 μM, if not indicated otherwise. Proliferation was assessed by 3 H-thymidine uptake within the last 16 hours of incubation. Results are shown in delta cpm (dpm = mean cpm of T cell proliferation with peptide -mean of T cell proliferation without peptide).

HLA-typing
Typing of HLA-DRB1-alleles was performed using a commercial SSO typing kit (DynalRELI SSO HLA-DRB Typing kit, Dynal, Bromborrough, UK). Samples with only a single detectable DRB1-allele were additionally typed by SSP (Dynal All set SSP DR low resolution, Dynal). High resolution typing was performed by nucleotide sequencing (BigDye Terminator Cycle Sequencing Kit, ABI, Foster City, CA).
As the x-ray structure of HLA-DRB1*04:01 has not been resolved yet and HLA-DRB1*01:01 and HLA-DRB1*04:01 show a sequence identity of 91.58%, we modelled the structure of DRB1*04:01 directly on the basis of the previously created HLA-DRB1*01:01 structure. The differing residues were mutated using SCWRL yielding 2 (pMHC) complexes with identical backbone structure and a few side chain mutations in the MHC.
To perform MD simulations we used the software package GROMACS 4 [20]. First, we immersed both complexes into distinct explicit artificial water boxes. Then we employed a steepest descent energy minimisation of the complexes to resolve spatial clashes and simulated warming to 310 K using position restraints. To provide insight into the dynamics of the pMHC complex at atomic scale, we subsequently carried out MD simulations [21] to investigate spatial rearrangements of molecular structures in water for a real time of 30 ns. All further parameters of the simulation were set according to Omasits et al. [22].
We evaluated the resulting trajectories using the root mean square fluctuation (RSMF) which is defined as: M is the number of frames, r i t k ð Þ is particle with number i of complex r at time k andr is the reference structure. This method essentially assesses flexibility of molecules and is implemented in GROMACS in the g_rmsf function. Furthermore, we calculated the solvent accessible surface area (SASA) [23] of the side chains of the peptide by rolling a sphere over them [24]. This method is implemented in GROMACS in the g_sas function. Visual pre-screening was performed using Visual Molecular Dynamics software (VMD, NIH, University of Illinois) [25] and the plug-in vmdICE [26].

Statistics
Significance levels were calculated by using Wilcoxon signed rank test (SPSS 10.0, Chicago, IL).
Taken together, for the T cell recognition of Art v 1 25-36 we found indications for a shorter minimal epitope in DR1, but no marked differences in the position and number of residues crucial for interaction with the two HLAmolecules.

Presentation of naturally-processed Art v 1-peptides by HLA-DR1 and -DR4 molecules
The experiments performed so far had been based on synthetic 12-mer peptides which might bind directly to HLA-DR-molecules expressed on the surface of APC. To compare intracellular loading of processed Art v 1peptides onto DR1-and DR4-molecules, HLA-DR1-or    DR4-expressing APC were incubated with titrated amounts of Art v 1 and used to stimulate Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] specific T cell cultures. As observed before with synthetic peptides, the Art v 1 25-36 -specific TCL and 2 TCC proliferated equally well to DR1 or DR4 at optimum concentrations. However, at lower concentrations DR1 tended to be more effective for T cell stimulation (Figure 3).

Discussion
In spite of the strong association of the allergic immune response to Art v 1 with HLA-DR1, we have previously shown that the in vitro binding of its single major T cell epitope Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] to HLA-class II molecules and its recognition by CD4+ T cells is degenerate [7,8]. In addition to the vast majority of Art v 1 25-36 -specific, DR1-restricted TCC, a few TCC from DR1-negative patients have been isolated that were restricted by DR15, DR16 or DR3. Although DR4 in general has similar peptide binding characteristics [11] as DR1 and binds Art v 1 25-36 which can be cross-recognized by DR1-restricted TCC in vitro [10], no Art v 1 25-36 -specific DR4-restricted TCC have been isolated so far. Therefore, we expected differences in Art v 1 25-36 -peptide binding and T cell activation between HLA-DR1 and -DR4 and compared anchor residues and minimal epitopes of the peptide to elucidate the molecular background for the DR1-biased T cell response to Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] in more detail. To find further support for structural differences between Art v 1 25-36 bound to HLA-DR1 or HLA-DR4 we also applied MD simulations to the respective pMHC complexes. As there was no crystal structure available for HLA-DR4, our computer simulations on peptide/HLA-DR4 interactions were performed on a modeled structure based on the x-ray structure of HLA-DR1. Nevertheless, this model of HLA-DRB1*04:01 can be assumed to be reliable, as the sequence identity between DR1 and DR4 is above 90% and models based on a template with a sequence identity >50% are generally regarded as highly accurate [28]. Furthermore, major changes in the fold of MHC molecules are extremely unlikely since the MHC superfamily shares a common overall shape [29].
First, we confirmed T cell cross-recognition of Art v 1 25-36 presented by APC expressing DR4. All T cell cultures used in this study were derived from DR1-positive donors and had most probably been selected by this MHC class II restriction element. Still, the vast majority of TCC (82%) recognized Art v 1 25-36 presented by DR4 indicating degenerate TCR recognition ( Figure 1B). However, at limiting peptide concentrations DR1expressing APC outperformed DR4-expressing APC in their stimulating capacity ( Figure 1C). Natural processing of antigens generates a pool of peptides with different length (mostly 15-18 aa) [30,31] and their intracellular loading into MHC-class II molecules underlies editing by HLA-DM [32]. Similar to our synthetic, 12-mer Art v 1-peptides, the peptides resulting from natural processing were also loaded efficiently into both DR1-and DR4 molecules as detected by the response of Art v 1 25-36 -specific T cells (Figure 3). Here again, DR1-expressing APC performed better at lower peptide concentrations than DR4-expressing APC.
Next, differences in the binding behaviour of Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] to DR1-and DR4-molecules and T cell stimulatory capacity were addressed. In vitro binding assays with isolated HLA-molecules and single aa mutants of Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] confirmed peptide-binding to DR1-molecules with high affinity and residue I27 as p1-anchor [8]. Moreover, we found a similar binding affinity and again I27 as p1-anchor for binding to DR4-molecules ( Table 1). The latter was consistent with our computer simulation data. Notably, the residues at p4, p6, or p9 (E30, A32, G35) which in general can bind to additional secondary binding pockets in the groove of DR1 and DR4, had no measurable further supporting effects for Art v 1 25-36 -binding to either DR molecule, again emphasizing the anchor role of I27. A and G by nature do not provide suitable side chains for anchoring and also proved not to be relevant for T cell activation ( Figure 2B). On the other hand, E30 (p4) turned out to be important for most T cell responses ( Figure 2). This finding might indicate that, instead of interacting with the HLA-molecule, E30 may interact with the TCR together with its neighbouring aa W29 (p3). However, the calculated surface exposition (SASA) of E30 (p4) obtained by computer modelling was relatively low ( Figure 4F) and rather suggested interaction with the MHC molecule. In this case, the simultaneous presence of the all-dominant I27 in the A30-mutant peptide apparently prevented in vitro detection of secondary MHC-binding residues.
Overall, the position and number of Art v 1 25-36residues crucial for T cell reactivity was similar in response to DR1-or DR4-molecules and focussed on the 5 aa stretch from I27 to K31. The indispensible residues E28 (p2) W29 (p3), K31 (p5) are located in the core of the minimal epitopes found. Less frequently Q33 (p7) and H34 (p8) also formed part of the minimal epitopes ( Figure 2A). Due to their increased flexibility and higher surface exposition of H34 in DR1 ( Figure 4E,F), these aa may contribute to a more intense TCR-interaction. An influence of peptide flanking residues (p-1 and p10) on T cell responses was also observed in a few experiments, but was not restricted to either DR1-or DR4-presentation.
Although we found similar affinities and structures (i.e. the same register) for Art v 1 25-36 -binding to DR1 and DR4 (Table 1), for more than 50% of the T cells the minimal epitope was shorter for the DR1-bound peptide (Figure 2A). In silico, differences in peptide flexibility, indicated by RMSF, became obvious between Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] bound to HLA-DR1 or HLA-DR4 with advanced simulation time. In DR1 the peptide was more flexible at both ends and detached from the MHC-molecule at the Cterminus ( Figure 4A-C). In essence, the parts of the Art v 1-peptide with lower ranges in RMSF, i. e. less flexibility, reflect the common core of the minimal peptide regions as defined by our in vitro experiments for DR1 (IEWEK) or DR4 (CIEWEKAQH). Thus these peptide ranges might define the parts of Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] buried in the MHC-binding groove. Further indications for structural differences between the Art v 1 25-36 /DR1 or -DR4 complex were the overall discrepancies obtained for surface exposition (SASA, Figure 4D). As compared to DR4, the C-terminal aa residues in DR1 were more solvent-exposed, while the N-terminal residues were less exposed. In addition, K31 at position p5, classically the central solvent TCR-exposed residue in MHC class II [33], showed a markedly higher SASA value in complex with DR1 than with DR4. These differences may indicate better access of the TCR to the core and C-terminus of the Art v 1-peptide in DR1.
Our simulations were based on the crystal structure of the 13-mer peptide HIV Gag p24 34 [34][35][36][37][38][39][40][41][42][43][44][45][46][47][48] containing 2 additional C-terminal aa residues (A, T). While the 13 N-terminal residues of the peptide show an essentially identical spatial arrangement as the Art v 1-complex at the initial configuration [16,26], these two residues reach out of the MHC binding groove and build a hairpin turn which has been reported to be essential for T cell recognition of this HIV Gag p24-epitope [16,34]. Intriguingly, this hairpin structure resembles certain substates of the detached C-terminal end of the Art v 1 25-36 / HLA-DRB1*01:01 complex as shown in Figure 5. Within 61 published crystal structures of pMHC class II molecules that contain at least short peptide flanking regions, we found in about 10% of them structures divergent from the canonical stretched conformation such as detachments, loop-like structures or hairpin turns. These structures most likely are important for T cell reactivity. We have previously reported effects of such protruding peptide segments on T cell activation for peptide flanking regions [34] and altered peptide ligands with high sequence similarity [35]. Thus, we suggest that the C-terminal Art v 1 25-36 -residues protruding from the peptide binding groove are relevant for interactions with TCR in vivo.
An alternative interpretation for the lack of DR4responses to Art v 1 in the population may be a potential negative influence of flanking regions in naturally processed peptides selectively on HLA-DR4. However, the fact that DR1-restricted T cells also respond well to naturally processed, DR4-presented Art v 1 (Figure 3) at optimum concentrations indicates that these effects would only be of minor relevance.
In summary, compared to DR4, DR1 allowed T cell activation of shorter minimal epitopes within Art v 1 [25][26][27][28][29][30][31][32][33][34][35][36] and at lower peptide concentrations. However, the differences obtained in our experimental assays may not clearly explain the high association of Art v 1 25-36recognition and HLA-DRB1*01. We used allergenspecific effector T cells with a very high sensitivity for antigen-specific activation in our study. However, the threshold level for activation of naïve T cells is much higher. Therefore, we suggest that at much lower allergen-concentrations, i.e. at conditions present in the human body, the observed differences become much more relevant in vivo, so that DR1-but not DR4expressing APC are capable of activating cognate T cells. This assumption would be backed by the fact that we could not establish DR4-restricted TCC from any of our DR4-expressing mugwort-allergic patients.

Conclusions
We used the T cell response to the immunodominant epitope of Art v 1 as a model to study p/MHC/TCR interactions in order to explain why a peptide with degenerate MHC-class II binding behaviour causes primarily HLA-DR1-restricted immune responses. The combination of experimental and in silico data suggested that TCR recognition of this peptide in the context of DR1 may be promoted by distinctive structural features in the pMHC complex leading e.g. to superior TCR accessibility and preferential priming in vivo.

Competing interests
BB is consultant for Biomay. All other authors declare that they have no competing interests.
Authors' contributions BJS and BK conceived of the study; GF and BB participated in its design and interpretation; BK did the computer simulation studies; BJS and DVH carried out the T cell experiments; IF and GF performed HLA-typing and analysed the data; BM did the HLA-binding assays; WS contributed with materials and supervision; CE provided clinical expertise; BJS and BK prepared the manuscript; BB and GF helped to draft the manuscript. All authors read and approved the final manuscript.