Major histocompatibility complex (MHC) class I and class II proteins play a pivotal role in the adaptive branch of the immune system. Both classes of proteins share the task of presenting peptides on the cell surface for recognition by T cells. Immunogenic peptide–MHC class I (pMHCI) complexes are presented on nucleated cells and are recognized by cytotoxic CD8+ T cells. The presentation of pMHCII by antigen-presenting cells [e.g., dendritic cells (DCs), macrophages, or B cells], on the other hand, can activate CD4+ T cells, leading to the coordination and regulation of effector cells. In all cases, it is a clonotypic T cell receptor that interacts with a given pMHC complex, potentially leading to sustained cell:cell contact formation and T cell activation.

Major histocompatibility complex class I and class II share an overall similar fold. The binding platform is composed of two domains, originating from a single heavy α-chain (HC) in the case of MHC class I and from two chains in the case of MHC class II (α-chain and β-chain) (Figure 1A). The two domains evolved to form a slightly curved β-sheet as a base and two α-helices on top, which are far enough apart to accommodate a peptide chain in-between. Two membrane-proximal immunoglobulin (Ig) domains support the peptide-binding unit. One Ig domain is present in each chain of MHC class II, while the second Ig-type domain of MHC class I is provided by non-covalent association of the invariant light chain beta-2 microglobulin (β₂m) with the HC. Transmembrane helices anchor the HC of MHC class I and both chains of MHC class II in the membrane (Figure 1A).

The groove in-between the two helices accommodates peptides based on (i) the formation of a set of conserved hydrogen bonds between the side-chains of the MHC molecule and the backbone of the peptide and (ii) the occupation of defined pockets by peptide side chains (anchor residues P2 or P5/6 and PΩ in MHC class I and P1, P4, P6, and P9 in MHC class II) (1–4). The type of interactions of individual peptide side-chains with the MHC depend on the geometry, charge distribution, and hydrophobicity of the binding groove. Predicting the affinity of these distinct MHC–antigen interactions for individual allotypes has been a long-standing goal in the community. While good progress has been made in developing and optimizing bioinformatic algorithms to estimate peptide binding to MHC proteins, these in silico predictions, however, still yield false positives (5, 6), and often fail in predicting immunodominance. We argue that understanding the relevance of transient or energetically excited protein conformations that are visited during the equilibrium fluctuations of the molecular structure is important for making good predictions.

In MHC class I, the binding groove is closed at both ends by conserved tyrosine residues leading to a size restriction of the bound peptides to usually 8–10 residues with its C-terminal end docking into the F-pocket (7–9). In contrast, MHC class II proteins usually accommodate peptides of 13–25 residues in length in their open binding groove, with the peptide N-terminus usually extruding from the P1 pocket (10). It has been reported that the interactions at the F pocket region in MHC class I and the P1 region (including the P2 site) in MHC class II appear to have a dominant effect on the presentation of stable pMHC complexes and on the immunodominance of certain peptidic epitopes (11–16). Interestingly, these pockets are located at opposite ends of the binding groove of the respective MHC class I and MHC class II structures (Figure 1B).

The most polymorphic human MHC class I and class II proteins (human leukocyte antigens, HLAs) are each expressed from three gene regions (MHC class I: HLA-A, -B, -C; MHC class II: HLA-DR, -DP, -DQ), which are all highly polymorphic. This allelic variation mainly affects the nature and composition of the peptide-binding groove and thus modulates the peptide repertoire that is presented on the surface by MHC class I or MHC class II proteins for CD8+ or CD4+ T cell recognition, respectively. A good match of the peptide and the MHC binding groove is an important, but certainly not the sole determinant of its presentation. In fact, the formation of a pMHC complex depends on its peptide-loading pathway, in which the selection of peptides is influenced by several factors, such as antigen availability, protease activity, or the availability of chaperones. In addition, for each MHC class, a “catalyst” is available to enhance peptide exchange for certain peptides: tapasin for MHC class I and HLA-DM for MHC class II. These molecules edit the presented peptide repertoire and bias the exchange reaction toward the presentation of thermodynamically stable complexes. Tapasin and HLA-DM thus act similar to typical enzymes by reducing the energy barrier for peptide exchange. However, in the case of HLA-DM and tapasin, no covalent bonds are formed or cleaved during the exchange reaction.

The MHC class I HC folds and assembles with β₂m in the lumen of the endoplasmic reticulum (ER). The partially folded heterodimer is then incorporated into the peptide-loading complex (PLC) for peptide binding and exchange. In the PLC, tapasin is a protein that catalyzes, together with other chaperones, the loading of high-affinity peptides derived from proteolysis of endogenously expressed proteins (Figure 1C, left panel) (17, 18). In the absence of tapasin, some class I allotypes (such as HLA-B*44:02) are retained in the ER (tapasin-dependent), whereas other class I proteins (tapasin-independent, such as HLA-B*44:05 and HLA-B*27:09) can bind peptides and travel to the cell surface (19–22). There is no crystal structure of the MHC class I/tapasin complex, but several structural models and mutational studies suggested that tapasin binds two regions in the HC of MHC class I, a loop in the α₃ domain (residues 222–229), and a region of the α₂ domain (residues 128–137) adjacent to the F-pocket (Figure 1B) (18, 21, 23–29).

Major histocompatibility complex class II proteins fold in the ER in complex with a protein called invariant chain (Ii) (30) and are then transported to late endosomal compartments (also coined MHC class II compartment, MIIC). There, Ii is cleaved by cathepsin proteases and a short fragment remains bound to the peptide-binding groove of MHC class II proteins, termed class II-associated invariant chain peptide (CLIP). This placeholder peptide is then normally exchanged against higher affinity peptides, which are derived from proteolytically degraded proteins available in endocytic compartments (Figure 1C, right panel). HLA-DM accelerates peptide exchange, with different allelic variants being more or less susceptible to catalysis. HLA-DM has a highly similar structural fold compared to classical MHC class II proteins, but its closed-up binding groove prevents peptide binding. Crystal structures of HLA-DM in complex with the MHC class II protein HLA-DR1 (31) and in complex with the competitive inhibitor HLA-DO (32) revealed that HLA-DM mainly contacts the α₁-domain of MHC class II proteins close to the P1 pocket and additionally the membrane-proximal β2-domain, in line with previous mutational analyses (Figure 1B) (13, 33–35).

Despite the structural differences between tapasin and HLA-DM as well as their presumably opposite sites of interaction with regard to the orientation of the binding groove, a similar mode of action has been suggested, hinting at a possible convergent evolution of the two exchange catalysts (Figure 1B) (36). A common feature seems to be that both catalysts target regions in the vicinity of those pockets in the peptide-binding groove that are of great relevance for the stability of the respective pMHC complex (11, 13, 15, 31, 37). Furthermore, in both cases, the binding of a high-affinity peptide is able to release the interaction with tapasin/DM (13, 26, 38, 39) and allows for transport of stable pMHC complexes to the cell surface.

While the general hallmarks of antigen processing and editing have been established, the discussion is now moving toward the dynamics of the system, both at the cellular and molecular level. The mechanistic questions relate to a description of how exactly peptides are selected for presentation and how tapasin and HLA-DM catalyze this reaction in an allele-specific manner.

Many allelic variants of MHC class I and MHC class II bound to individual peptide antigens display different biochemical features, but surprisingly, their “ground-state,” i.e., thermodynamically most stable conformations reported by the many available pMHC X-ray structures are very similar. In contrast, increasing experimental and computational evidence of wild type (WT) and mutant MHC complexes over the past years incontestably revealed that changes in conformational dynamics in MHC proteins have to accompany peptide loading and exchange (22, 40–46).

To highlight possible dynamic regions within ground-state crystal structures of human MHC class I and class II proteins bound to a peptide, we performed a global B-factor analysis of all available X-ray crystal structures of human MHC complexes in the absence of any other binding partner. In each structure, we normalized the B-factor values of each alpha carbon (CA) atom to the global mean. Then the variance of all the normalized B-factor values for each CA atom, in 297 human pMHC class I and 41 human pMHC class II structures, was calculated and depicted with a blue to red color spectrum, respectively (heat map on structures of HLA-A*0201 and HLA-DR1, Figures 2A,B). Overall in the binding groove of class I (class II), the α-helix in the α2 (β1) domain displays higher B-factor variation values than the α-helix in the α1 (α1) domain and the β-strands from both domains. Among the pMHCI structures, although B-factor variation values in the N-terminus-proximal helical segments indicate the existence of a certain degree of dynamics, the α2-helical region around the peptide C-terminus displays the highest variation in B-factors (Figure 2A). Among the class II structures, high B-factor variation values are found especially in β-strands 2 and 4 of the α-chain, the 3₁₀ helical region and almost the entire β-chain α-helix (Figure 2B). Our analysis is corroborated by a previous comparison of 91 different pMHC class II crystal structures (47). In this analysis, conformational heterogeneities were observed in three regions: the 3₁₀-helical region (α45–54), the kink region in the β1-helix (β62–71), and the β2-domain (β105–112).

It is known that, to some extent, structural variations can be introduced by variable peptide-binding modes. In this context, peptides longer than 8–10 residues have been reported to bind to the MHC class I binding groove (48–52). To accommodate the increase in length, the peptides have to bulge out, leaving the central residues (between p2 and pΩ) exposed to solvent. This is usually achieved by a kink in the backbone in the middle part of the peptide. Recently, two crystal structures of HLA-A2 bound to15-mer peptides have been solved (53). The two peptides follow a binding mode similar to that of the canonical peptides with two anchor residues in the B and F pockets. While one of the peptides showed a mobile central conformation similar to another reported long peptide (48, 54), the other peptide adopted an unusual rigid β-hairpin secondary structure. Furthermore, although the binding of the N- and C-termini at both ends of the binding groove is conserved in almost all the MHC class I complexes, some exceptions have been reported. For example, in the F pocket of HLA-A2, the C terminal residue of the peptide extends by ~1 Å leading to a significant rearrangement of the pocket with only one of the standard hydrogen bonds (at Thr143) preserved (55). Another example is seen in the HLA-B35 protein, the short N-terminus of the 8-mer peptide does not reach the A pocket. Instead, the hydrogen bonds between the amino group of P1 residue and residue 45 of MHC class I are mediated by a water molecule (56).

Since, in the case of MHC class II proteins, the peptide ligand within the binding groove usually adopts a pseudosymmetrical PPII helix-like conformation, bidirectional binding is theoretically possible. An interesting case represents a crystal structure of a DR1/CLIP complex, in which the peptide binds in a very unusual, inverted orientation (C-terminus close to the P1 pocket). The driving force for this peptide inversion is the formation of three additional H-bonds of P1-close residues and the backbone of the peptide’s C-terminus (57, 58). Biological and biochemical evidence for the existence of other pMHC class II isomers have been described in the context of autoimmunity (59). Since MHC class II proteins have open binding grooves, peptides can protrude outwards and even bind in different registers. In this regard, an insulin B chain-derived peptide (InsB_9–23) was suggested to induce type 1 diabetes (T1D) in a thermodynamically less favored, low-affinity binding register (60, 61).

Apart from variable peptide binding, catalyst binding can induce more significant conformational variation, as seen in DM-bound DR. By designing a P1-anchor-free pMHC class II complex, Pos et al. could increase the affinity for the pMHC class II to DM and solve the crystal structure for the covalently tethered DR1/HA/DM complex (31). In this MHC class II/DM complex, the interaction interface is primarily composed by the α-subunits of DM and DR1 (~65% of the entire interaction surface, Figure 1B). DM binding to DR stabilizes a rearranged conformation in the vicinity of the P1 pocket of DR1 resulting from the absence of critical peptide-MHC class II interactions in this region. The extended region in the DR1α-chain (α52–55) and the 3₁₀-helix adopt an α-helical fold (Figure 3A). Compared to other parts of the pMHC class II structure, it was shown that this site indeed represents a conformationally labile region (35, 46). In addition, the C-terminal part of the β1-α-helix (β86–91) becomes slightly less structured (Figure 3B).

A feature, which is also present in DM-bound DO, is the intermolecular H-bonds between two conserved residues (DRα W43 and DMα N125). The formation of these critical bonds (as well as other interactions) likely depends on the “flipping out” movement of αW43 from the P1 pocket of DR1 toward DMα (Figure 3A). This movement of DR1 αW43 was suggested to be triggered by partial dissociation of the peptide’s N-terminus or by transient destabilization of contacts of the peptide N-terminus when bound to MHC class II (13, 64). As a consequence, the P1 site is stabilized by a repositioning of two phenylalanine sidechains (DRα F51 and DRβ F89), thereby compensating for the loss of peptide anchors and αW43 from this region (Figure 3B). Incoming peptides have to compete with these repositioned Phe residues for the P1/P2 site in order to be selected for display. Interestingly, structural characteristics of T1D-conferring HLA-DQ alleles were indeed discussed to be linked to decreased DM-sensitivity (62). The analysis of several DQ variants indicated structural differences of the T1D-risk variants DQ2 and DQ8 when compared to DQ1, DQ 6, or DR variants. The decrease in DM-susceptibility of these two proteins was explained by a stabilization of the 3₁₀ helical region (63) (Figure 3C). However, the exact relationship between structural variations in the 3₁₀ helix and DM-susceptibility are not clear, as highly DM-susceptible DR complexes can display a different conformational mode, compared to DM-susceptible DQ alleles (Figure 3D).

The observed changes lead to the question as to how much structural plasticity in these regions preexists in the peptide-loaded form and provide a prerequisite for catalyzed as well as for spontaneous peptide exchange. For example, do pMHC complexes sample conformations observed in simulations of empty proteins or in complex with the catalyst? How would allelic variation affect the distribution of MHC proteins within the conformational space and thereby influence the presented peptide repertoire? Could variation in protein plasticity also account for the association of specific MHC alleles with immune diseases? Since the polymorphic peptide-binding groove of MHC proteins defines its affinity for a certain peptide, substitutions of even a single amino acid may lead to significantly different affinities for individual peptides. In general, the critical factors defining whether a peptide is presented or not are determined at the different levels of antigen processing and presentation such as uptake route, amount, and folding state of the antigenic protein, amenability to proteolytic degradation, and catalysis of the complex, etc. However, at the molecular level, it has been shown that certain polymorphisms shape individual pockets in the peptide-binding groove to optimally present an autoimmunogenic self-peptide (65–67). In other cases, the functional impact of disease-associated polymorphisms remained enigmatic and suggests that dynamics might account for the observed differences.

While simulations and experimental studies vary in the features ascribed to peptide-free MHC proteins, they certainly agree in attributing a substantial degree of dynamics to the peptide-binding groove. Thus, binding of peptides to MHC proteins is of utmost importance for the stabilization of the known MHC fold (40, 68). The lack of a crystal or an NMR structure of peptide-free MHC protein hinders an accurate description of the structural changes upon peptide binding and this is probably due to the ensemble character of the peptide-free conformers. This ensemble character, however, has been probed by computational techniques, as discussed in the following paragraphs.

While the study of empty MHC proteins is of theoretical and conceptual interest, nature has engineered the antigen-presenting system in a way that prevents the accumulation of isolated, non-peptide-bound MHC molecules. Endogenous peptides derived from the proteasome in case of class I or from the invariant chain in the case of class II first bind and eventually are replaced by antigenic peptide. This inherently dynamic process is enabled by intrinsic features of the MHC molecules and several studies suggest that pMHC complexes sample different and transient conformations dependent on the bound peptide and the allelic variant under investigation (46, 64, 71, 72, 95, 100, 106, 113, 114).

While the studies described in the previous section unambiguously demonstrate the dynamic features of pMHC complexes, the question arises naturally in how far these properties translate into peptide exchange. For pMHCII complexes experimental progress has been made in identifying intermediate or transient conformations of pMHCII with regard to catalyzed or intrinsic peptide exchange (13, 31, 32, 35, 46). For pMHCI molecules the atomistic description of structural changes during peptide exchange mostly relies on MD simulation studies, supported by mutational analysis and circumstantial biophysical evidence (14, 22, 37, 73).

Major histocompatibility complex proteins are encoded by oligogenic and highly polymorphic genes and most polymorphisms map to the regions important for peptide binding. The pMHC complexes display various degrees of flexibility along the binding groove, and these dynamic features seem to correlate with the propensity for peptide exchange. Of interest is the fact that tapasin and DM both bind their MHC targets in regions of enhanced dynamics. Short, destabilized helical segments together with their adjacent structural elements seem to represent the requirements for transient binding of the respective catalyst. The degree of these local flexibilities can be correlated to a higher dependence of a particular pMHC complex on tapasin or DM. Polymorphic substitutions might not only change the binding preference for certain ligands but also the overall stability and dynamics of the corresponding allelic variants. In turn, this will affect the conformational ensemble recognized by the peptide editors and in principle should be able to explain why certain alleles seem to possess a generally lower taspasin or DM dependence. In this way, MHC molecules may become a paradigmatic example of how differences in the dynamic landscape of protein complexes translate into distinct functional outcomes of physiological relevance.

How far have we come and what has to be done to achieve this goal? Figures 4 and 5 summarize the findings described in this review and also emphasize the most daunting questions in the field that need to be answered in order to formulate a unifying concept of antigen exchange. What seems to be clear is that both type of MHC molecules can exchange peptide along two distinct pathways, with the ratio of spontaneous versus catalyzed exchange certainly being different for the allelic protein variants and pMHC complex. While dynamics often correlates with thermodynamic stability, it has yet to be seen which type of motions are critical for catalysis and which structural elements are indispensable for these transitions to occur. For MHC class II molecules, the structure of the MHCII–DM complexes provides a cornerstone (31), and the early intermediates (μs-ms timescale) toward the DM-bound form could be defined (35, 46) (Figure 5). However, there are no structural insights about the replacement of DM by incoming peptide, thus requiring experimental and simulation strategies to follow the fate of DM-prebound MHCII molecules. In the case of MHCI, a tapasin-class I complex structure is required in order to provide a reliable framework for further experimental and theoretical studies. Similarly, characterization of empty MHC molecules will certainly aid in defining the dynamic modes that are explored by the peptide-binding domains. Since it has been shown that empty MHC molecules can be rescued by the chaperoning function of the exchange catalysts (134, 135) and thus the dynamics that occur upon peptide exchange are likely to show features of the empty state. It seems, therefore, highly desirable to compare the two systems on time scales down to a few milliseconds.

For both MHC classes, more sophisticated NMR experiments capitalizing on selective amino acid side-chain labeling protocols are probably required and methods relying on CEST or relaxation dispersion should be able to yield more direct information on the anticipated intermediate states (136, 137). So far, in-depth NMR experiments are restricted to certain stable pMHC complexes and the investigation of other alleles have been hampered by the in-availability of other variants such as the disease-relevant DQ alleles. There is a need to expand the experimental basis of dynamically investigated pMHC complexes in order to test the predictions made on the basis of the dynamic features of just a few alleles. Solutions are most likely to come from protein engineering approaches in combination with the use of different expression systems. The increasing importance of MD simulations arises from the fact that micro-to-milli-second simulations in combination with Markov State Modeling will become more of a standard in the field. This is essential, because the critical intermediates of antigen exchange seem to be populated at this time scale. Once we are able to conceptualize conformational peptide exchange, we will be in the position to better predict MHC peptide occupancies in the context of cellular editing mechanisms and we will understand and be able to manipulate the action of small molecules or biological macromolecules that modulate peptide exchange.

MW, EA, and CF designed and wrote the manuscript. JS, MÁ-B, SS, and FN critically read, discussed, and edited the content prior to submission.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.