Dynamics of MHC-I Molecules in the Antigen Processing and Presentation Pathway

Introduction

The class I major histocompatibility complex (MHC-I) proteins display antigenic peptides derived from the processing of intracellular proteins, as well as exogenous antigens acquired from the endosomal pathway. Following peptide-loading, the MHC-I/peptide complex is transported to the cell surface for presentation to cytotoxic T-cells and natural killer cells to enable immune surveillance [1]. Human MHC-I proteins, referred as human leukocyte antigens (HLA), are highly polymorphic, with over 10,000 distinct allotypes identified to date [2]. Despite sharing a highly conserved structural fold [3,4], different HLA allotypes display vastly different peptide repertoires, while allelic sequence diversity at specific residues has been shown to affect interactions with molecular chaperones and T cell receptors [5–7]. Mounting biophysical and functional data highlight the role of protein plasticity in fine-tuning MHC-I interactions and peptide selector function [8–11]. This review presents insights from studies of protein dynamics focusing on both peptide-receptive and peptide-loaded MHC-I molecules, and how such motions impact chaperone recognition and peptide repertoire selection. While conformational changes in the MHC-I peptide-binding groove induced by bound peptides can also modulate T cell recognition [3,12,13], these functional aspects of MHC-I dynamics are discussed elsewhere [14].

Protein Dynamics of peptide-loaded MHC-I

MHC-I molecules are heterotrimers consisting of a heavy chain, an invariant light chain (β₂m) and a bound peptide of 8–15 amino acids long. Polymorphic residues are concentrated in the peptide-binding groove, which results in a tremendous diversity of peptides that can be captured and displayed by different HLA allotypes, to ensure species adaptability to emerging pathogens [15,16]. In some cases, a single amino acid difference can alter the properties of a given HLA allotype and contribute to disease susceptibility [17–21]. In one exemplary system, the class-I protein HLA-B*27:05 (D116) is strongly associated with ankylosing spondylitis, while its closely related subtype, HLA-B*27:09 (H116) is not [22]. The crystal structures of HLA-B*27:09 and HLA-B*27:05 bound to the same peptide showed minimal backbone deviations [23,24]. However, isotope-edited infrared (IR) experiments and molecular dynamics (MD) simulations on HLA-B*27 proteins identified subtype-specific conformational differences between the two variants. From these studies, HLA-B*27:05 was shown to exhibit higher conformational flexibility in the F-pocket region, irrespective of the bound peptide [18,19]. Using nuclear magnetic resonance (NMR) experiments, we have characterized the distribution of residues exhibiting dynamic mobility (on the microseconds to milliseconds timescale) for four different murine and human MHC-I molecules, corresponding to the sampling of “excited-state” conformational states [8]. Through this analysis, we confirmed that MHC-I molecules exhibit allelic-specific dynamic profiles, with main differences in flexibility that are concentrated in the α_2–1 helix, β₅-β₈ sheets, and multiple loops on the α₃ domain (Figure 1A) [8]. Thus, the modulation of protein dynamics in different HLA allotypes offers an additional dimension to regulate function through the sampling or minor conformational states which may differ both locally and globally from the ground-energy structure, provided by X-ray crystallography.

Figure 1.

MHC-I exhibits allele-specific conformational dynamics that influence its interaction with its binding partners. A) microseconds-to-milliseconds dynamics profiles of three MHC-I molecules in decreasing order of flexibility from left to right. Conformational flexibility of methyl probes was measured through ¹³C single-quantum Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion NMR experiments by McShan et al. (2019). The spheres represent methyl groups of Ala, Ile, Leu, or Val on the structure of H2-Dd/P18-I10 (PDB: 3ECB), HLA-A*02:01/TAX (PDB: 1DUZ), and HLA-A*01:01/NRAS^Q61K (PDB: 6MPP). Methyl sites with increased dynamics are shown in yellow, and areas not exhibiting dynamics are shown in gray. B) Protein dynamics extend beyond the peptide-binding groove to the distal α₃ and hβ₂m molecules. Similar to figure 1 A, the yellow spheres are methyl groups exhibiting dynamics [9]. Non-methyl dynamics (K58, W60, and S88) on the nanosecond timescale have previously been reported for hβ₂m bound to HLA-B*27:09 [23, 26]. Residues K58 and S88 interact with T cell CD8 co-receptor and NK cell receptor Ly49A [26]. NMR chemical shift perturbations analysis revealed micro polymorphism and peptide-occupancy modulate W60 conformational heterogeneity [24]. The proximity of W60, the peptide-binding groove, and the TAPBPR hairpin loop (residue 210–213) suggest that TAPBPR could potentially sense peptide occupancy through hβ₂m.

Flexible residues have also been observed at the interface between the heavy chain and the non-covalently bound β₂m subunit [9]. This observation is consistent with previous NMR studies showing that β₂m residues (K58, W60, and S88) that are localized at the heavy chain interface of HLA-B*27:05 and HLA B*27:09 varied in conformational flexibility on the nanoseconds timescale (Figure 1B) [25,26]. The crystal structures of TAP binding protein-related (TAPBPR)/H2-D^d complex showed that the previously mentioned dynamic sites are located on a β₂m loop (58–60) that is positioned directly beneath the F-pocket near residue 116 of the heavy chain, and is also in the vicinity of the “jack hairpin” of the TAPBPR (residues 210–213) [27,28]. Furthermore, the TAPBPR C-terminal domain forms an 1g domain trimer with β₂m/α₃. Consequently, it is plausible that peptide-editing chaperones can sense peptide occupancy in the MHC-I groove through perturbations of dynamics at the MHC-I floor and the β₂m interface, which form direct interactions with the chaperone. While no large-scale conformational changes are observed between cryogenic crystal structures of MHC-I, such motions can be still sampled in an aqueous environment at physiologically relevant temperatures [29,30]. In addition to TAPBPR the α₃ and β₂m domains form binding sites for other important receptor molecules [31]. Notably, residues K58 and S88 of β₂m have been crystallographically observed to interact with the CD8 co-receptor, the murine Ly49A NK cell receptor — the human LILRs homolog and leukocyte immunoglobulin-like receptors [25,32,33]. Taken together, these results suggest that allosteric communication of the peptide-binding groove with distal sites of the structure, particularly with respect to α₃/β₂m domain orientation, could modulate MHC-I interactions with other co-receptors and molecular chaperones.

These studies paint a picture where peptide-loaded MHC-I molecules vary in mobility range that extends beyond the peptide-binding groove. They suggest that domain reorientation between the peptide-binding groove, the α₃ domain, and β₂m provides a plausible mechanism employed by MHC-I molecules to fine-tune their interactions with chaperones and co-receptors [34]. However, the study of global domain motions in silico is challenging due to the much longer timescales involved (milliseconds to seconds) [35]. Here, obtaining experimental parameters reporting on global domain orientation, like residual dipolar couplings (RDCs) and Paramagnetic Relaxation Enhancements (PREs) by NMR, or distance distributions by double electron-electron resonance (DEER) paramagnetic resonance (EPR) spectroscopies, can shed light on the emerging link between long-range allosteric communication and MHC-I function [36,37].

Dynamics of peptide-deficient MHC

Peptide-deficient MHC-I’s intrinsic instability has prevented detailed structural characterization of the peptide-receptive state [38,39]. An attempt to study empty HLA-C*07:02 by NMR revealed broadened and overlapped peaks in the peptide-binding groove, which are characteristics of conformational exchange between different conformations at the milliseconds timescale [40]. However, empty molecules are not necessarily unfolded. MD simulations combined with tryptophan fluorescence experiments showed that empty molecules have varying degrees of structure in the α₁ and α₂ helix [39,41], Furthermore, the conserved 3₁₀ helix has been shown to adopt "locked" and "unlocked" forms which impact the conformation of the A/B pocket [42], Recently, a series of crystal structures of peptide-deficient HLA-A*02:01 were published [43]. Analysis of these peptide-free structures revealed that synergistic side chain interactions in the A/B and F pocket resulted in the groove adopting two alternate conformations, referred to as "open" and "closed". Molecular dynamics (MD) simulations of these structures further showed that interconversion between the open and closed form in the F pocket affects peptide conformations in the adjacent binding pockets [43]. The peptide-binding groove is highly flexible and coordinated movements of the side chains in the groove orchestrate peptide-MHC interactions. In short, the biophysical characterization of peptide-free MHC-I is consistent with results from atomistic MD simulations.

Protein dynamics in the antigen processing pathway

The chaperone-mediated peptide loading process is a quality control checkpoint which ensures that MHC-I molecules with poorly docked peptides are prevented from reaching the cell surface [44–46]. Newly synthesized, nascent MHC-I molecules associate with the peptide-loading complex (PLC) for stabilization and peptide selection [47]. The PLC is composed of multiple subunits, with Tapasin linked to the ERp57 disulfide isomerase providing the main catalytic peptide exchange activity [48,49]. The structure of the Tapasin-ERp57 complex revealed that Tapasin adopts an L-shaped form, and comparisons of the Tapasin domains in the asymmetric unit suggest that the molecule exhibits significant interdomain flexibility [50]. A recent analysis of the intermolecular interactions from MD simulations of cancer variants of ERp57 and Tapasin has shown that missense mutation at the ERp57/Tapasin interface significantly impacts the mobility of the C-terminal domain of Tapasin [11]. Furthermore, MD simulations using an atomic-resolution model built from the PLC cryo-EM density have shown that interactions with calreticulin can constrain the relative orientations of the Tapasin N-and C-terminal domains in the context of the PLC [48,51,52]. The Tapasin C-terminal domain forms direct contact with the class I MHC CD8-loop located on the α₃ domain (residues 223–229), suggesting that mutations at these sites may affect peptide-loading and selection of MHC-I [53].

With respect to recognition by Tapasin, the intrinsic protein mobility of MHC-I is important [21]. HLA-B*44:05 is capable of binding to peptides and exhibits cell surface expression in the absence of Tapasin, while the opposite is true for HLA-B*44:02 [54]. MD simulations of these molecules in a peptide-receptive form revealed that the α-helices flanking the peptide-binding groove of HLA-B*44:05 are conformationally less heterogeneous compared to HLA-B*44:02 [20,41,55]. In conjunction with previous data, this supports that a high degree of Tapasin dependency correlates with increased protein dynamics near the F pocket [56]. The MD simulations suggest that Tapasin-dependent MHC-I molecules, which are less stable, transiently adopt a minor conformation which can be recognized by Tapasin. While the function of Tapasin has been extensively studied in vivo, the lack of high-resolution structural data of the Tapasin/MHC-I complex has made it difficult to determine the precise molecular mechanism for peptide-loading and exchange [34,49,57].

Molecular insights into understanding MHC-I recognition and peptide-selection process have come from studying the Tapasin homolog TAPBPR. The molecular chaperone TAPBPR shares ∼22% amino acid sequence identity with Tapasin, and the two molecules employ a similar binding mode to engage MHC-I [45,58,59]. However, TAPBPR functions independently of the PLC and knockout of TAPBPR generally has minimal impact on MHC-I cell surface presentation on a Tapasin wildtype background [60]. There are two available X-ray structures for TAPBPR in complex with murine MHC-I molecules [27,28]. Comparison between the mouse H2-D^d TAPBPR-bound and unbound states revealed structural remodeling due to widening of the MHC groove by displacement of the α_2–1 helix and the floor of the peptide-binding groove, that is also accompanied by long-range domain remodeling which reveals a cryptic binding site at the α₃/β₂m interface [27]. In addition, residues R66 and Y159 in the peptide-binding groove form polar interactions across the empty A/B pocket where the peptide normally binds [47]. Furthermore, methyl-based NMR and MD simulations revealed that TAPBPR also dampens dynamics, likely, by altering the ensemble of sidechain rotameric states sampled by residues across the entire groove, with a more pronounced effect was observed for the α_2–1 helix [9]. Altogether, TAPBPR employs the intrinsic MHC-I mobility to stabilize a peptide-deficient conformation and induce an open conformation of the α_2–1 helix which enables annealing of high-affinity peptides to the MHC-I groove.

TAPBPR functions on specific MHC-I allotypes through the recognition of distinct protein conformational states. NMR analysis revealed a highly conserved binding mode across different MCH-I proteins involving the α₂ helix along with the α₁ helix, β₅-β₈ strands, and α₃ domain [9]. The α₂ helix on the MHC-I groove plays an important role in TAPBPR chaperone activity, and "helix breaking" proline substitutions at positions 112, 156, and 166 decreased TAPBPR recognition in situ [8]. In addition, introducing a disulfide bond at the F pocket (Y84C and A139C), which limits the mobility of the α_2–1 helix, also abrogates TAPBPR interactions (Figure 2) [8]. Conversely, mutations of residues along the α₁ helix, probed using deep saturation mutagenesis follow in parallel by bi-fluorescence complementation and surface expression assays, can be tolerated with respect to TAPBPR binding but not with respect to surface expression [8]. This is consistent with a recent report using local frustration and evolutionary trace analysis to analyze over 8,000 HLA allotype homology models [61]. In this work, highly conserved and minimally frustrated sites are localized in the α₂ helix and F pocket, further highlighting the importance of this region for interactions with chaperones. In summary, MHC-I molecules can sample a broad range of conformations at the milliseconds timescale in an allotype-specific manner. TAPBPR acts by selectively recognizing epitopes present in a subset of these conformations to promote folding and loading of high-affinity peptides.

Figure 2.

A conceptual model for chaperone recognition of the MHC-I conformational landscape. The vertical axis is free energy, and the horizontal axis shows the potential conformations that MHC-I can adopt. Unfolded or misfolded, α₂ helix (denoted in red) has been shown to result in compromised TAPBPR recognition in situ [8]. However, disruption of the α₁ helix does not influence TAPBPR binding [8]. MHC-I loaded with high affinity peptides may also interact with TAPBPR in an allotype-dependent manner, through the sampling of a minor, open conformation of the α_2–1 helix. Introducing a disulfide linkage (Y84C/A139C) that restricts the mobility of the α_2–1 helix abolishes TAPBPR binding in vitro. Figure adapted from McShan et al. (2019).

Taken together, molecular chaperones recognize a minor conformational state which is sampled by some nascent MHC-I allotypes. Binding of TAPBPR to these molecules further induces a widening of the peptide-binding groove, leading to an enhancement in suboptimal peptide dissociation. Subsequent binding of high-affinity peptides induces a closed conformation of the α_2–1 helix which triggers chaperone dissociation from the pMHC complex (Figure 3) [57].

Figure 3.

A conceptual energy landscape depicting the progressive loading of an MHC-I molecule with a high-affinity peptide. MHC-I molecules can bind to low-affinity peptides forming a suboptimal-loaded complex that is unstable and conformationally heterogeneous. Molecular chaperones, like TAPBPR, can catalyze low-affinity peptide dissociation by widening the groove. Ultimately, the binding of a high-affinity peptide quenches MHC-I dynamics, stabilizes the MHC-I molecule and displaces TAPBPR.

Conclusion

The process of peptide-loading and selection on MHC-I molecules is highly complex and dynamic. Biophysical studies have provided insights into the role of protein dynamics in different steps of the antigen processing and presentation pathway (Figure 3). However, there are still areas that warrant further investigations. Analysis has been primarily focused on model systems using HLA-A and HLA-B subtypes, thus does not encompassing the entire allelic human landscape. For example, HLA-C differs from the other two families in terms of surface expression and its ability to interact with NK cells, but has been largely understudied at the structural level [62,63]. Also, most studies exploit recombinant molecules lacking membrane association and glycosylation. Yet, most MHC-I molecules undergo extensive, complex glycosylation at the conserved Asn86 position, which has been suggested to modulate conformational dynamics of proteins in the PLC [51,64]. Insights gained from probing these questions using complementary biophysical, structural and high-throughput functional techniques will ultimately lead to a detailed, mechanism-focused understanding of this essential immune surveillance process.

Highlights.

• MHC-I molecules exhibit allotype-dependent conformational flexibility, which is critical for their function.

• Peptide-deficient MHC-I adopts distinct groove conformations relevant for peptide binding.

• Molecular chaperones locally and allosterically modulate dynamics to edit the peptide repertoire.