N^α-Terminal Acetylation for T Cell Recognition: Molecular Basis of MHC Class I–Restricted N^α-Acetylpeptide Presentation

Produced by Ag processing and proteasomal degradation of intracellular proteins, polypeptides serve as CTL epitopes presented by MHC class I molecules, which play a critical role in cellular immunity (1). Eukaryotic proteins bearing various posttranslational modifications (PTMs) can generate a group of modified Ags, which contribute to a special repertoire of MHC-associated peptides presented at the cell surface as potential targets for TCR-mediated recognition. A modified peptide may become a new Ag because of the distinguished antigenicity compared with its unmodified homolog. A variety of natural peptide Ags containing modification have been observed that can be immunologically discriminated by T cells from their unmodified homologs as “altered self” (2). Thus, the significance of PTMs on epitopes and the application of modified peptides in vaccine development for immunotherapy against cancer and autoimmune diseases have been increasingly appreciated (3, 4).

The molecular bases of the presentation of peptides with several PTMs by MHC class I molecules have been successfully explicated. For instance, the formyl group on an Nt-formylated peptide binds to the bottom of the peptide-binding groove of H2-M3 (5); both the glycan and the phosphate moieties of the central region of the glycopeptides (6, 7) and the phosphopeptides (8, 9), respectively, are exposed to enable TCR binding, and the deimination (citrullination) of arginine on a peptide presented by two HLA-B27 subtypes induces distinct peptide conformations (10).

N^α-terminal acetylation (Nt-acetylation) is one of the most common PTMs, occurring on the vast majority of eukaryotic proteins. In humans, >80% of the different varieties of intracellular proteins are irreversibly Nt-acetylated by N^α-acetyltransferases, often after the removal of the initiator methionine. Only a subset of the penultimate residues (Ala, Ser, Thr, Cys, and Val) or the retained initiator methionine can be acetylated at the α-amino (NH₂) groups (11). A recent study found that acetylated N-terminal residues of eukaryotic proteins act as specific degradation signals (Ac-N-degrons) that are recognized by specific ubiquitin ligases (12). A subsequent systematic analysis demonstrated that Nt-acetylation can also represent an early determining factor in the cellular sorting for prevention of protein targeting to the secretory pathway (13). These findings suggested that Nt-acetylation–mediated inhibition of secretion could contribute to the retention of proteins in the cytosol where they may subsequently be ubiquitinylated through the specific recognition of their Ac-N-degrons and thereby generating Nt-acetylated proteasomal digestion products (14). Hence, these Nt-acetylated polypeptides in the form of MHC-associated neoantigens stand a good chance to be recognized by T cells. This has indeed been illuminated in an Nt-acetylated MHC class II–restricted peptide derived from myelin basic protein, which stimulates murine T cells to elicit experimental autoimmune encephalomyelitis, whereas the nonacetylated form does not (15). A structural study subsequently suggested that the Nt-acetylation of this peptide is essential for MHC class II binding (16).

For MHC class I, the first Nt-acetylated natural ligand was identified more than a decade ago (17). However, the mode of interaction of this acetylated peptide with class I molecules remained largely enigmatic. To understand this, we determined the crystal structures of a naturally occurring Nt-acetylated self-peptide (NAc-SL9) and two nonmodified variants (SL9 and HL8), respectively, in complex with HLA-B*3901. Taken together with the thermostability analyses of N^α-acetylpeptides complexed with a series of class I molecules of human and murine origin, we elucidated that Nt-acetylation exerts a destabilizing effect on peptide–MHC (pMHC) complex, thereby influencing TCR recognition.

All peptides used in this study were synthesized and purified to >90% by ChinaPeptides (Shanghai, China). Each peptide was stored in lyophilized aliquots at −80°C and dissolved in DMSO to a final concentration of 10–50 mg/ml before use. A summary of the peptide characteristics is presented in Table I.

Briefly, truncated HLA-B*3901 H chain (residues 1–274) and human β₂-microglobulin (β₂m) were expressed respectively as inclusion bodies using BL21(DE3) strain of Escherichia coli and subsequently refolded and assembled in the presence of peptides as described previously (26, 27). After incubation at 277 K for 24–48 h, the soluble portion of the complexes were concentrated and purified by size exclusion and subsequently by anion-exchange chromatography. The three pHLA-B*3901 complexes were finally concentrated to 4–8 mg/ml (20 mM Tris [pH 8] and 50 mM NaCl) and subsequently crystallized after being mixed in a 1: 1 ratio with crystallization reservoir solution (NAc-SL9: 0.1 M ammonium acetate, 0.1 M Bis-Tris [pH 6.5], and 17% [w/v] polyethylene glycol [PEG] 10,000; SL9: 0.1 M ammonium acetate, 0.1 M Bis-Tris [pH 6.5], and 20% [w/v] PEG 10,000; HL8: 0.2 M sodium formate, and 20% [w/v] PEG 3,350) by the hanging-drop vapor diffusion method at 277 K. For cryoprotection, each crystal was first soaked in reservoir solution supplemented with 17% (v/v) glycerol, and then flash-cooled and maintained at 100 K in a cryostream (28). X-ray diffraction data were collected at beamline BL17U on Shanghai Synchrotron Radiation Facility. Raw data were indexed, integrated, corrected for absorption, scaled, and merged using HKL2000 program (29).

All crystals of pHLA-B*3901 belong to the P2₁2₁2₁ space group. The structures were solved by molecular replacement method using MOLREP program with HLA-B*2705 (Protein Data Bank 2BST) as a search model (30), which were manually rebuilt with COOT program under the guidance of Fo-Fc and 2Fo-Fc electron density maps (31). Restrained refinements were performed with REFMAC5 program (30) and followed by isotropic atomic displacement parameter refinement and bulk solvent modeling in phenix.refine (32). During the process of model building and refinement, the peptides and water molecules were built into unambiguous electron density. The stereochemical quality of the structure was verified with PROCHECK program (33). Different from the NAc-SL9 complex, two clear solutions in both the rotation and translation functions corresponded to the two molecules in one asymmetric unit of SL9 or HL8 complex. A summary of X-ray data collection, refinement, and validation statistics are presented in Table II. Structural figures were produced with PyMOL program (DeLano Scientific) for displaying molecular three-dimensional model and electron density. Unless otherwise specified, figure generation and calculational analysis were performed based on one uniform molecule in the asymmetric units.

The atomic coordinates of the crystal structures of HLA-B*3901^NAc-SL9, HLA-B*3901^SL9, and HLA-B*3901^HL8 have been deposited in the Protein Data Bank (http://www.pdb.org/pdb/home/home.do) with accession number 4O2C (http://pdb.org/pdb/search/structidSearch.do?structureId=4O2C), 4O2E (http://pdb.org/pdb/search/structidSearch.do?structureId=4O2E), and 4O2F (http://pdb.org/pdb/search/structidSearch.do?structureId=4O2F), respectively.

The thermostability of MHC class I complexes formed with N^α-acetylpeptides and nonacetylpeptides were tested by circular dichroism (CD) spectroscopy. All complexes were refolded and purified beforehand as described above and measured at 150 μg/ml in a solution of 20 mM Tris (pH 8) and 50 mM NaCl. CD spectra at 218 nm were measured on a Chirascan spectrometer (Applied Photophysics) using a thermostatically controlled cuvette at temperature intervals of 0.1°C at a rate of 1°C/min between 25 and 90°C. The denaturation curves were generated by nonlinear fitting with OriginPro 8.0 program (OriginLab).

Eukaryotic proteins having an N-terminal serine are most frequently acetylated (11). The acetylserine-containing peptide NAc-SL9 (NAc-SHVAVENAL) in our study was chosen based on its natural presentation by at least three allotypes of HLA-B39 (B*3901, B*3905, and B*3909), and its direct generation by the 20S proteasome as a major breakdown product of the corresponding parental protein (Table I) (17, 18). The structure of HLA-B*3901 bound with NAc-SL9 was solved in this study to a resolution of 1.8 Å (Table II). The overall structure of HLA-B*3901 resembles that of other published class I allotypes, with a H chain containing three extracellular domains (α1, α2, and α3) for binding peptide ligands (Fig. 1A, 1B).

To assess the influence of Nt-aceylation in the interaction of the acetylpeptide with MHC class I, we also solved the structures of HLA-B*3901 bound with the nonacetylated peptide SL9 (SHVAVENAL). Both the unmodified and Nt-acetylated peptide bind to HLA-B*3901 in a featured mode with a central bulge revealed by the unambiguous electron density (Fig. 2A, 2B). In both peptides, P2-His and Pc-Leu are completely buried as the primary anchor residues, whereas P3-Val also points downward toward pocket D of the Ag-binding groove and subdominantly contributes to peptide binding (Fig. 1C). The surface-exposed residues in the central region of the peptide interact with the H chain via water-mediated hydrogen (H)-bonds (Table III). Moreover, the side chains of P6-Glu and P7-Asn point upward and both comprise potential TCR contact sites. The third peptide of which we solved the structure in complex with HLA-B*3901 is HL8 (HVAVENAL), a natural truncated analog of NAc-SL9 isolated from the same HLA-B39–bound peptide pool (18). The side chain orientations of HL8 residues are essentially identical to those of SL9 residues except for the absence of P1-Ser (Fig. 2C). Summarizing, there is no change in extended direction of the residue side chains between the NAc-SL9 and SL9/HL8 peptides except the P1 residue (Fig. 1C).

Most peptides presented on MHC class I have their terminal NH₂ groups placed in pocket A (34). The free peptide NH₂ group binds the class I molecules through two conserved H-bonds to Tyr⁷ and Tyr¹⁷¹, respectively, at the bottom of pocket A, essentially in a peptide sequence-independent mode (35). These interactions ensure the lowest Gibb’s free-energy for the peptide binding state, and were observed in the HLA-B*3901^SL9 structure (Fig. 2E). However, in the HLA-B*3901^NAc-SL9 structure, the acetyl group added to the N terminus of P1-Ser is energetically unfavorable for burial within the pocket A, resulting in a unique rotation of the P1 dihedral angle ψ by ∼150°, rotating the acetyl group into the position normally occupied by the P1 side chain. A superposition of the Ag-binding grooves reveals a considerable uplift of the P1 residue, which raises the NAc-Ser Cα-atom ∼1.4 Å compared with its Ser counterpart in SL9. The methyl C-atom of the acetyl in NAc-SL9 is also elevated to 3.0 Å higher relative to the P1-Ser Oγ-atom in SL9 (Fig. 3).

Although NAc-SL9 undergoes conformational changes around the acetylation site, similar interactions to those seen in HLA-B*3901^SL9 are maintained at the bottom of pocket A. The rotation of the P1 residue eliminates the contacts of Tyr⁷ and Tyr¹⁷¹ with the NH₂ group but the contacts are substituted by the hydroxyl group of NAc-Ser (Fig. 2D, 2E). Specifically, the acetyl interacts with a completely conserved Trp¹⁶⁷ residue on the α2-helix of the H chain through an H-bond and indirectly through a water-mediated H-bond (Fig. 2D), both of which seem significant in stabilizing the acetyl binding. The rearrangement of the H-bonding network as well as the rotation of the P1 residue occurring in pocket A are unique features that have never been described in previous reports of MHC class I structures.

In addition to the large rotation of the P1 residue, another substantial elevation (1.2 Å) was observed in the central region of the peptide main chain of NAc-SL9 (Figs. 3, 4A), which shows the highest degree of solvent exposure for direct contact with the TCR. The O-atom of the acetyl interacts with the carbonyl O-atom of P2-His via another water-mediated H-bond (Figs. 2F, 4B). Thus, the acetyl group and the P2 main chain are constrained to each other, which stressfully results in deformation of the peptide main chain and an indirect lift of the P6-P7 residues.

A similar elevation of the peptide main chain also was observed when the two molecules in the same asymmetric unit of the HLA-B*3901^HL8 structure were compared (Fig. 4C). In both molecules, P1-His of HL8 occupies pocket B to serve as the N-terminal anchor residue without the canonical P1 interaction; the NH₂ group of P1-His provides an H-bond with Tyr⁷ but not with Tyr¹⁷¹, and a water molecule bridges the NH₂ group of HL8 and a Tyr¹⁵⁹ on the α2-helix as a substitute for the carbonyl of the absent P1 residue (Fig. 2F). However, two distinct conformations of Arg⁶² were observed in these two molecules (Fig. 4C) (discussed below). One of them has a conformation as the Arg⁶² in HLA-B*3901^SL9, and the other one switches its side chain to stretch over pocket A, and directly contacts the P1-His main chain by H-bonding interactions (Fig. 4D). The interactions appear to have the effect of “pulling” the peptide up out of the binding site creating the same pattern as the water bridge in HLA-B*3901^NAc-SL9 (Fig. 4B).

Comparison of the two HLA-B*3901 molecules in complex with SL9 and NAc-SL9 reveals that dramatic conformational differences in the H chain mainly focus on three residues in the α1-helix [i.e., Glu⁵⁸, Gln⁶⁵, and especially Arg⁶² (Fig. 5A, Supplemental Fig. 1)]. For HLA-B*3901^SL9, the Arg⁶² side chain is captured and tethered down toward the peptide N terminus by an H-bond from the P1-Ser Oγ-atom. For NAc-SL9, in addition to the steric hindrance, the hydrophilic Arg⁶² guanidine group is not compatible with the hydrophobic environment created by the methyl moiety of the acetyl. Therefore, this Arg⁶² has to undergo a local reorientation so that its Cγ-atom faces toward the methyl, thereby forming a hydrophobic interaction. The charged Arg62 guanidine group switches away and pulls the Gln⁶⁵ over via an H-bond. Meanwhile, the flexible Glu⁵⁸ side chain rotates to interact with its own carbonyl O-atom (Supplemental Fig. 1).

For ease of analysis, we have divided the conformational modes of Arg⁶² into different intramolecular locations. The position of the Arg⁶² in the HLA-B*3901^SL9 is defined as position 1, where it is often found binding with TCRs (36), while that in the HLA-B*3901^NAc-SL9 is defined as position 2 (Fig. 5B). This conformational alteration of Arg⁶² also has been observed by others in two HLA-B8 structures (Fig. 5B) (37), one of which bound to a peptide with P1-Gly and its Arg⁶² is at position 1, whereas in the other, the P1-Phe sterically forces the Arg⁶² away to position 2 where it can engage and trigger various TCRs (38). Taking these previous findings into consideration, we hypothesized that in our structure, the acetyl moiety would provide a similar stereochemical effect as the phenyl group of Phe and that the Arg⁶² at position 2 might contact with an incoming TCR. Specially in one of the two molecules in the asymmetric unit of the HLA-B*3901^HL8, Arg⁶² is at position 1, whereas in the other, Arg⁶² extends its side chain to position 3, toward the α2-helix covering the pocket A (Fig. 5B). These alternative conformations reveal the flexibility of the Arg⁶² in the absence of P1 residue. Thus, interestingly, the Arg⁶² with its mobile conformational modes in HLA-B*3901 structures can be seen as a structural knife-switch (Fig. 5C). The aliphatic moiety of Arg⁶² is switched “on” to position 1 for energetic contribution to peptide binding or “off” to position 2 for TCR engagement, respectively. Also, position 3 is in an intermediate state. It is tempting to speculate that these diverse modes might elicit qualitatively different TCR repertoires.

To determine how Nt-acetylation affects the peptide-binding to MHC class I molecules, the thermostability of the three solved pHLA-B*3901 complexes was evaluated. In accordance with previous studies (17), the acetyl in NAc-SL9 exerted a negative influence on Ag-binding and thermodynamic stability of the pHLA-B39 complex relative to SL9, ΔT_m = −9.1°C (Fig. 6A, Table I). Moreover, an excellent correlation was observed between the thermal denaturation curves and the volumes of peptide-binding grooves for all three complexes (Fig. 6A, Table IV). The Nt-acetylation of NAc-SL9 observably enlarges the groove in both volume and surface area. The volume of pocket A in the HLA-B*3901^NAc-SL9 is >50% larger than that in the HLA-B*3901^SL9 because of the side-chain reorientation of the Arg⁶² (Fig. 2G, 2H, Table IV). Also, the elimination of the interactions from Arg⁶² might explain the lowered peptide binding stability. For HLA-B*3901^HL8, one of the alternative conformations of Arg62 in contact with the peptide main chain seems to compensate the absence of the canonical P1 residue for suboptimal peptidic N-terminal interactions as described above (Fig. 4D).

To further test whether Nt-acetylation is universal in decreasing the pMHC stability, we subsequently carried out a series of thermostability measurements with different pMHC sets (Table I). The thermostability of epitopes with Ala, Thr, Cys, Val, and Met (that can be frequently acetylated in vivo) at P1 residues, respectively, bound to HLA-A*1101, -A*3303, -A*2402, -B*3501, and -A*0201 was compared with that of their Nt-acetylated counterparts. Nt-acetylation of all these peptides causes dramatic reductions of the pMHC thermostability by 8.8°C or more in midpoint transition temperature (T_m) (Fig. 6D–I). The similar phenomenon was observed for HLA-A*0201 and H-2K^b when in complex with a peptide containing a P1-Ser (Fig. 6B, 6C). In conclusion, Nt-acetylation decreases the pMHC stability in a class I allele– and epitope-independent manner.

Nt-acetylation is predicted to have a significant influence on T cell recognition in three ways. First, incorporation of the acetyl group blocks the positive peptide charge, thereby altering the distribution of electrostatic potential and it creates a prominent solvent exposure around the peptide N terminus. This may contribute to putative TCR docking sites, based on previously determined TCR–pHLA complex structures (Fig. 7A, 7B) (36). Second, the conformationally altered central region of the N^α-acetylpeptide may have different interactions with the TCR, and the lift of the peptide main chain may influence the TCR binding and correspondingly the TCR repertoires (41). Third, the conformational changes within the MHC H chain, indirectly caused by the peptide Nt-acetylation, position the negatively charged Glu⁵⁸ slightly outward; most notably, the positively charged Arg⁶² is positioned as a predominant surface residue toward the central region of the peptide, thereby shifting the charge footprint of Ag presentation surface even more (Fig. 7C, 7D). HLA consensus Glu⁵⁸ is retained in most TCR engagement (42), Arg⁶² and Gln⁶⁵ invariably interacts with the TCR CDR1α or CDR3α loops whenever present on the MHC (37, 42). These analyses indicated that not only the acetyl moiety but also the altered residues of MHC class I are highly accessible for interaction with CDR loops of an incoming TCR. To further address this possibility, we superposed the previously solved TCRs in complex with other HLA-B molecules (B*3501 and B8) onto the HLA-B*3901^NAc-SL9 structure (Fig. 7E, 7F). The modeled complex shows close proximity of the CDR1α and CDR3α loops to the acetyl moiety and the MHC residues near the acetylation site, from which the Arg⁶² is engulfed by the CDR loops signifying it as a prominent binding target for TCR recognition.

Our results here provide the structural and thermodynamic insights into the presentation of Nt-acetylated peptides by MHC class I molecules. The structure of the N^α-acetylpeptide in complex with HLA-B*3901 outlines a molecular interpretation of the reduced stability of MHC class I–bound Nt-acetylated peptides and also highlights a potential influence of Nt-acetylation on antigenic identity and T cell recognition. In addition, the structure elucidation of HLA-B*3901, the predominant B39 subtype, also is valuable in studying immune diseases associated with this MHC allele.

In a previous report, the Nt-formyl group on an Nt-formylated peptide binds to the bottom of the peptide-binding groove of the murine MHC class I H2-M3 playing an anchoring role for MHC class I binding (Supplemental Fig. 2A) (5). In our study, the methyl and carbonyl groups of the acetyl are rotated upwards like two arms that push the peptide-binding groove open (Fig. 2G, Supplemental Fig. 2B), thereby altering its immunogenicity at the expense of the pMHC stability. The thermostability we tested from seven human and one murine complexes indicates a general feature of N^α-acetylpeptide in weakening the binding affinity to MHC class I, which could be revealed by the gel-filtration chromatography of pMHC refolding assays as well (Supplemental Fig. 3). Their instability would partially explain why, as yet, such epitopes are rarely found. Within N-terminal residues of eukaryotic proteins, Ser is the most frequently acetylated in vivo (11). The Ala, Thr, Cys, and Val residues can also be Nt-acetylated and have small side chains like Ser. Thus, the rotation of P1 residues observed in the pHLA-B*3901 complex with an acetylated P1-Ser could very well be a general mode in N^α-acetylpeptide binding. In contrast, the long side chain of Met precludes it from being rotated into pocket A, but a certain reorientation is presumed to take place in the acetylated P1-Met based on the thermal instability (Fig. 6H). Besides the accommodation of the acetyl moiety, Nt-acetylation is presumed to decrease the stability of the pHLA-B*3901 complex as a result of the conformational switch of the Arg⁶². Arg⁶² in the α1-helix is largely conserved in almost all HLA-B and -C allotypes (Table V). For other HLA class I (Table V, Fig. 8), the long charged side chains of the residues in position 62 (Glu⁶² of A24 and Gln⁶² of A11 and so on) also may interact with the acetyl. Hence, the residue in position 62 plays a key role in the interaction between acetyl group and the H chain, which may influence not only the N^α-acetylpeptide binding to HLA molecules but also the TCR docking.

The discoveries that intracellular proteins with Ac-N-degrons are inhibited from being secreted (13) and then are degraded via ubiquitylation (12) raise many questions on the biological significance of acetylation-mediated proteolysis (14). The Nt-acetylated peptides with the size of MHC class I ligands (8–11 aa) as neoepitopes for CD8⁺ T cells, represent one of the possible roles of the Nt-acetylated digestion products. The vast armory of intracellular proteins that are frequently Nt-acetylated can create a large pool of N^α-acetylpeptides for Ag presentation and T cell surveying. The Nt-acetylation potentially impacts the TCR-MHC interaction in three different aspects: 1) the direct interaction of the solvent-exposed acetyl moiety; 2) the altered conformation of the central region of the peptide main chain; and 3) the conformational switches of the MHC residues. The Nt-acetylation creation of a distinctive pMHC landscape and participation in a potential binding element for TCR engagement described in our results highlights needs for further investigation into the N^α-acetylpeptide–specific TCR repertoires.

Although the apparent instability of MHC class I molecules for binding Nt-acetylated self-ligands is likely to be a mechanism of immune tolerance, it cannot yet rule out the possibility that Nt-acetylation of some self peptides are not present during the T cell precursor negative selection in the thymus thus allowing autoreactive T cells to migrate to the periphery. Once in the periphery, T cells encounter the N^α-acetylpeptides, and it is then viewed as “foreign,” which will result in autoimmunity. Conversely, if some Nt-acetylated self-Ags or peptides arise temporally in the course of T cell development, the unmodified form of such peptides under disordered or pathological conditions will become neoantigens to which the immune system has never been exposed or tolerated in the periphery. Certain processes that induce Nt-acetylation for proteolysis, such as apoptosis, do take place in the thymus (3).

Furthermore, there are many examples of viral protein Nt-acetylation that normally uses the Nt-acetylation system of the host eukaryotic cells (11). For instance, the N-terminal peptides detected by Hutchinson et al. (45) from PB1, PA, NP, M1, M2, NS1, and NEP of influenza A virus and from PA, NP, M1, NS1 and NEP of influenza B virus can be acetylated as well as eukaryotic proteins. According to our estimate, some other viral N^α-acetylpeptides also could be important MHC class I–associated epitopes displayed during virus infections (Supplemental Table I). Thus, an acetylated form of the N-terminal epitope corresponding with a certain viral Ag designed into epitope vaccines may gain an antivirus utility. To identify more acetylated epitopes requires further exploration of the immunopeptidome, and current algorithms of MHC-associated epitopes prediction should be improved to take account of Nt-acetylation. N^α-Acetylpeptides represent a library of new potential vaccine candidates that are of considerable interest to be unearthed in the future.

We thank Yue Liu for excellent suggestions regarding the study, Joel Haywood for comments on the manuscript, and the staff at the Shanghai Synchrotron Radiation Facility beamline 17U for assistance.

The authors have no financial conflicts of interest.