Antigenic Peptide Loading into Major Histocompatibility Complex Class I Is Driven by the Substrate N-Terminus

Abstract

Open AccessCCS ChemistryRESEARCH ARTICLE1 Mar 2022Antigenic Peptide Loading into Major Histocompatibility Complex Class I Is Driven by the Substrate N-Terminus Mengna Lin, Honglin Xu, Yi Shi and Lin-Tai Da Mengna Lin Key Laboratory of System Biomedicine (Ministry of Education), Shanghai Center for Systems Biomedicine, Shanghai Jiao Tong University, Shanghai 200240 , Honglin Xu Key Laboratory of System Biomedicine (Ministry of Education), Shanghai Center for Systems Biomedicine, Shanghai Jiao Tong University, Shanghai 200240 , Yi Shi *Corresponding authors: E-mail Address: [email protected] E-mail Address: [email protected] Bio-X Institutes, Key Laboratory for the Genetics of Developmental and Neuropsychiatric Disorders, Shanghai Jiao Tong University, Shanghai 200030 and Lin-Tai Da *Corresponding authors: E-mail Address: [email protected] E-mail Address: [email protected] Key Laboratory of System Biomedicine (Ministry of Education), Shanghai Center for Systems Biomedicine, Shanghai Jiao Tong University, Shanghai 200240 https://doi.org/10.31635/ccschem.021.202000657 SectionsSupplemental MaterialAboutAbstractPDF ToolsAdd to favoritesDownload CitationsTrack Citations ShareFacebookTwitterLinked InEmail Major histocompatibility complex class I (MHC-I), a key element of the acquired immune system, plays essential roles in activating CD8+ T cells by recognizing intracellular antigens derived from pathogens and cancer. Assembly of MHC-I and antigen peptides is critical for the antigen presentation on the cell surface. However, the structural dynamics of antigenic peptide loading into MHC-I, at atomistic resolution, is still elusive. Here, by constructing a Markov state model (MSM) based on large scale all-atom molecular dynamics (MDs) simulations with an aggregated simulation time ∼24 μs, we reveal the detailed molecular mechanism underlying the peptide-loading dynamics into MHC-I and identify the key intermediates with associated thermodynamic/kinetic properties. Furthermore, we examine how the chaperone tapasin-binding protein related (TAPBPR) participates in promoting the peptide loading, and the results show that TAPBPR, by binding to the F pocket, allosterically modulates the structures of the distant pocket B, resulting in formation of a peptide-receptive conformation ideal for accommodating the incoming peptide N-terminus. This study provides fundamental structural insights for the peptide loading into MHC-I in both chaperone uncatalyzed and catalyzed contexts. Download figure Download PowerPoint Introduction The presentation of antigenic peptides by major histocompatibility complex (MHC) proteins on the surfaces of antigen-presenting cells is a central process in the mammalian acquired immune system.1–4 MHC, also called human leukocyte antigen (HLA) in humans, consists of two main types of MHC proteins, namely MHC-I and MHC-II. MHC-I captures a broad set of endogenous antigenic peptides with a length of 8–11 amino acids (aa) from the lumen of the endoplasmic reticulum, forming peptide-MHC-I (pMHC-I).5,6 pMHC-I is then transported to the Golgi and cell surface wherein, by interacting with T cell receptor (TCR) from CD8+ cytotoxic T lymphocytes, it induces the TCR signaling cascade.3,4 MHC-I is a heterodimer that is assembled from a polymorphic heavy chain and a non-covalently linked light chain called β2-microglobulin (β2m).7,8 The MHC-I heavy chain is encoded by three genes (HLA-A, HLA-B, and HLA-C in human), and all three are polymorphic, which results in various characteristics of peptide-binding grooves responsible for recognizing specific peptides.9,10 In particular, HLA-A*02:01 (termed as HLA-A2) is the major MHC-I allomorph found in over 50% of the global population.11 To date, extensive X-ray crystallographic structures of the pMHC-I have been obtained.12–17 These structures reveal that the substrate peptide lies in an extended conformation and is embedded in the peptide-binding groove that consists of three domains: helix α1, helix α2, and the β-sheet floor below helices α1 and α2 (Figure 1a).18,19 The helix α2 can be further divided into a shorter helix α2-1 and a longer α2-2 via a turn structure. Notably, two major binding pockets in MHC-I, called pockets B and F, are critical for the peptide recognition and correspond to the binding region of the N- and C-terminus of the peptide, respectively.20–22 These two pockets are responsible for accommodating certain anchor residues from the incoming peptide and are therefore vital for peptide specificities. Importantly, selection of high-affinity antigenic peptides from a pool of peptides is critical for the intracellular assembly of pMHC-I, a process called "peptide exchange" or "peptide editing."23–25 The chaperone tapasin-binding protein related (TAPBPR) is responsible for facilitating the loading of high-affinity peptides into MHC-I.26 Recently, two X-ray crystal structures of TAPBPR revealed the detailed interaction interface between TAPBPR and mouse MHC-I [Protein Data Bank (PDB) id: 5opi and 5wer], wherein one scoop loop of TAPBPR is bound in pocket F, thereby occupying the C-terminal binding site of the incoming peptide.27,28 Likewise, one MHC-I structure in complex with dipeptide Glycine–Leucine (GL) derived from the scoop loop in tapasin (the homologue protein of TAPBPR) reveals that the GL motif indeed binds to the F pocket.29 Notably, the GL binding, together with the C-ter-truncated peptide, can result in concomitant bindings of the GL motif in pocket F and the truncated peptide in pocket B with the same bound conformation as the full-length peptide.29 It thus can be expected that the peptide-loading process into MHC-I likely takes place by initially docking the N-terminus into pocket A or B, followed by the C-terminus binding to pocket F, for example, by replacing the scoop loop of TAPBPR/tapasin. Figure 1 | Energy-minimized pMHC-I structure and methodology diagram used in this work. (a) Two different structural views of pMHC-I after energy minimization. The light chain β2m is in orange; the heavy chain is in grey, with the helices α1 and α2 flanking the peptide-binding groove colored in green and cyan, respectively; the 10-mer substrate peptide, ELAGIGILTV, is shown in magenta. (b) Zoomed-in view of the peptide-binding site, with the sidechains of the peptide highlighted in sticks, as well as two critical peptide binding pockets B and F. (c and d) Key hydrogen-bonding networks for the N-terminal (c) and C-terminal (d) peptide residues, with the HBs represented by black dashed lines. (e) The methodology flowchart of this work. Download figure Download PowerPoint In addition to the static structures of pMHC-I provided by the crystallographic studies, fluorescence,30,31 nuclear magnetic resonance (NMR),32 and biochemical methods11,33 have also been employed to investigate the molecular mechanism underlying the peptide loading into MHC-I. For example, by utilizing the fluorescence anisotropy method, Springer et al.30 observed that the presence of dipeptide GL not only stabilizes the MHC-I structure, but also significantly promotes the high-affinity peptide binding by occupying pocket F. Their further studies indicate that the dipeptide Glycine–Methionine (GM), behaving similarly to GL, can also catalyze the rapid peptide exchange in MHC-I.31 In addition, a recent site-directed mutagenesis study revealed that TAPBPR is responsible for stabilizing MHC-I, and partial or full deletion on the TAPBPR scoop loop can profoundly impair the MHC-I folding.11 Importantly, TAPBPR is proposed to act as a peptide competitor that impedes the binding of low- or moderate-affinity peptides but accepts the high-affinity ones.11,33 Therefore, considering that the peptide N-terminus likely initiates the substrate loading process, it can be hypothesized that occupying the pocket F by TAPBPR/tapasin/dipeptide might allosterically reshape the distant pockets A and B to a conformation that is suitable for peptide N-terminal binding. This conclusion is supported by recent NMR studies that also highlight the significant role of the peptide N-terminus in TAPBPR-mediated peptide loading.32 Despite the above efforts, the atomic-level understanding of the peptide-loading dynamics into MHC-I and how TAPBPR promotes the peptide exchange are still largely unknown due to the low spatiotemporal resolution of the current experimental techniques. Computational simulations can provide atomic insights for the structural and dynamic properties of biomolecules and are utilized to investigate almost all the critical biomolecular systems, including peptide-loading complex (PLC), pMHC-I, and TCR.18,34,35 For example, by combining steered molecular dynamics (SMD) simulations and a single-molecule biophysical approach, Wu et al.34 investigated how the dynamics of pMHC-I might be affected by the TCR binding. In addition, extensive conventional molecular dynamics (MDs) simulations have been employed to study the structural dynamics of MHC-I in the presence of peptide or TCR.36–40 Despite that, the external forces applied by the SMD simulations can significantly bias the real transition paths. Moreover, formerly conducted unbiased MD simulations are mainly focused on investigating the structural dynamics of MHC-I or antigenic peptide at equilibrium condition (peptide-bound MHC-I complex) instead of the peptide-loading dynamics. Notably, one can now construct a Markov state model (MSM) based on hundreds of short-time MD simulations (i.e., hundreds of nanoseconds) to reveal the long-timescale dynamic properties for certain conformational changes.41–45 More importantly, both thermodynamic and kinetic properties of the key intermediates involved can be readily obtained by MSM construction.46 Here, by building MSM based on extensive all-atom MD simulations with an aggregated simulation time ∼24 μs, we reveal, at atomic resolution, the molecular mechanism of high-affinity peptide loading into MHC-I from the bulk solution. Our work identifies the key metastable states involved in the peptide-loading process, and pinpoints the key structural transition that limits the overall binding rate. Importantly, docking of the peptide N-terminus into the pocket B in MHC-I is determined to be a critical step for guiding the peptide loading, supported by additional mutant MD simulations. Moreover, we also evaluated the functional roles of TAPBPR in facilitating the peptide exchange by constructing TAPBPR-bound MHC-I complex, upon which additional MD simulations were performed. Our studies provide deep structural insights into the complete peptide-loading dynamics into MHC-I in both chaperone uncatalyzed and catalyzed contexts. This work assists the development of cancer neoantigen immunotherapy. Computational Methods Modeling the pMHC-I structure with a 10-aa peptide as the substrate HLA-A2-activated CD8+ T cells from melanoma patients mainly recognize Melan A26–35 (EAAGIGILTV) and Melan A27–35 (AAGIGILTV), but the immune activation of these two peptides is weak due to weak binding affinities of these peptides to HLA-A2. The anchor-residue-modified 10-aa peptide Melan A26–35 (E LAGIGILTV, altered residue is underlined) not only increases the binding affinity with HLA-A2 but also activates a greater expansion of CD8+ T cells, becoming the focus of therapeutic vaccination targeting melanoma.47–49 In our study, we selected a crystal structure (PDB id: 3qdg) of TCR-ELAGIGILTV-HLA-A2 complex. By removing the TCR molecule from 3qdg, we obtained the pMHC-I to elucidate the molecular mechanism of peptide loading into MHC-I. SMD and MD simulations Based on the constructed pMHC-I model, we initially performed 20 ns SMD simulations (27 SMD trajectories) to obtain the initial peptide-loading path; two different peptide-loading pathways were observed ( Supporting Information Figures S1 and S2). For each path, we chose two SMD trajectories to seed the first two rounds of MD simulations. We found that an apparent free-energy gap appears between the peptide-free and -bound states of MHC-I for the C-ter-mediated loading path ( Supporting Information Figure S3). Therefore, we only kept the trajectories associated with the N-ter-mediated loading path for the third round of MD simulations. Finally, 241 100-ns MD trajectories were collected for the MSM construction. See the Supporting Information text for more details about the setups of SMD and MD samplings. MSM construction and validation Based on the above obtained simulation database, we selected 3888 distance pairs ( Supporting Information Figure S11) as the input feature for time-structure independent component analysis (tICA) implemented in the MSMbuilder-3.8.0 package50–52 to decompose the high-dimensional conformations into the top four principle vectors, followed by the k-center method to group the projected simulation into 500 microstates. To interpret the peptide-loading mechanism, the 500 microstates were further grouped into four macrostates by the PCCA+ (PCCA = Perron Cluster Cluster Analysis) algorithm.53 Finally, we evaluated the effects of the correlation lag time and the numbers of microstates on the MSM construction ( Supporting Information Figure S12). The Chapman–Kolmogorov test was also used to validate the MSM ( Supporting Information Figure S13). Refer to the Supporting Information text for more details of the MSM construction, validation, and calculations of the thermodynamic/kinetic properties. Other detailed information, including the setup of the pL2A mutant MD simulations, modeling of TAPBPR-MHC-I complex, and the configurational entropy calculation of peptide, can be found in the Supporting Information text. Results and Discussion Results Extensively sampling the peptide conformations along its loading paths into MHC-I To investigate the molecular mechanism of the peptide-loading dynamics in MHC-I, we initially constructed a pMHC-I based on one crystal structure of HLA-A2 in complex with one 10-aa peptide (ELAGIGILTV) and one TCR molecule (PDB id: 3qdg).54 We removed the TCR molecule and only kept the pMHC-I, which was then subjected to energy minimization. The minimized structure shows that two anchor residues pL2 and pV10 are loaded into the pockets B and F, respectively, by mainly forming nonpolar contacts with the MHC-I residues (Figure 1b). In addition, both the N- and C-terminus of the peptide can establish strong hydrogen-bond (HB) networks with MHC-I. Specifically, the C-terminal residues pT9 and pV10 can form several HBs with the sidechains of MHC-I W147, D77, Y84, T143, and K146 via either the backbone amide or the –COO− group (Figure 1c). The N-terminal residue pE1, in contrast, can interact with the sidechains of MHC-I Y7, E63, Y159, and Y171; the amide groups of pL2 and pA3 interact with MHC-I E63 and Y99, respectively, all via HBs (Figure 1d). Moreover, the remaining part of the peptide, the pG4–pL8 segment, bulges from the binding groove and is largely exposed to solvent, thus no HB forms. Next, to derive an initial peptide-loading pathway, we applied SMD simulations starting from the above modeled pMHC-I to obtain the initial peptide release paths, which, presumably, are equivalent to the reverse process of the peptide-loading dynamics. We conducted 20 ns SMD simulations (27 in total) by varying the pulling direction, pulling force constant, and pulling distance (see Supporting Information text for more details of the SMD setup). Intriguingly, we observed two distinct peptide release paths, which differ in the releasing order for the two peptide termini. That is, for one path the C-terminus releases from the F pocket first, followed by the N-terminus, whereas the other path proceeds in the reverse direction ( Supporting Information Figure S2). For each release path, we then chose two representative SMD trajectories with relatively smaller pulling force constants for the subsequent unbiased MD simulations ( Supporting Information Figure S2). To extensively explore the conformational space for the peptide-loading process, we initially performed two rounds of MD simulations with an aggregated simulation time of 12 μs (see Supporting Information text for more details of the geometrical clustering and MD simulations setups). Our results clearly show that the N-ter-mediated peptide-loading path is energetically more favored than the C-ter-mediated one; for the latter, an apparent high-energy barrier is present ( Supporting Information Figure S3). We therefore only kept the MD simulations associated with the N-ter-mediated peptide-loading paths and performed the third round of MD simulations (see Supporting Information text for details). Finally, 241 100-ns MD trajectories with an aggregated simulation of 24.1 μs were collected for the final MSM constructions (Figure 1e). Insertion of the peptide N-terminus into MHC-I pocket B dictates the peptide-loading dynamics Our MSM reveals four critical metastable states during the overall peptide-loading process, namely S1–S4, among which S1 and S4 correspond to the unbound and bound states, respectively, and S2 and S3 as two key intermediates (Figures 2a and 2d). Notably, the unbound state is characterized as a conformation where neither the N- nor C-terminus of the peptide has bound to the binding groove in MHC-I, that is, S1. Further transition path theory55 analyses identify one dominant peptide-loading path, S1 → S3 → S4, and one off-path, S1 → S3 → S2. Notably, the peptide binding to MHC-I can significantly rigidify the substrate conformation, reflected from the profound decrease of the configurational entropy of the peptide from S1 to S4 (Figure 2b). Specifically, the peptide in S1 is largely exposed in the solvent and exhibits the largest conformational heterogeneity among all states, thus, no significant HBs exist between MHC-I and the peptide (Figure 3b). The S1 → S3 transition can lead to the insertion of the peptide N-terminus (pE1–pA3) into the pocket B in MHC-I, whereas the C-terminal segment (pT9–pV10) is still largely exposed in the solvent and displays diverse conformations (Figures 2c and 2d). Notably, the S1 → S3 transition can establish six stable HBs between MHC-I and the peptide N-terminus, among which two HBs formed via MHC-I Y99 and Y159 (namely H4 and H8 in Figure 3a) demonstrate the highest HB occupancies (68.1% and 61.4%, respectively). Detailed distance analysis indicates a concurrent formation of the above two HBs in S3, implying their collaborative roles in locking the peptide N-terminus in pocket B (Figure 3c). Further statistical analyses of 239 HLA-A2 structures in complex with different peptides varied in sequence and length show that these above two HBs are conserved in 225 out of 239 crystal structures (Figure 3d), stressing again that Y99 and Y159 are critical for recognizing the N-terminus of the peptide. In addition, four additional HBs can be formed via MHC-I E63 and K66, with the HB-occupancies ranging from 30% to 50% (see H3, H5, H6, and H7 in Figures 3a and 3b). Figure 2 | The MSM identifies four macrostates during the peptide-loading process. (a) Free-energy landscape of the pMHC-I conformations projected onto the two slowest time-structure independent components (tICs), first tIC and second tIC, with each assigned macrostate labeled. (b) Configurational entropy of the peptide for each macrostate calculated using the Schlitter's formula.56 The mean value was averaged over all the microstates that belong to the same macrostate, and the corresponding standard error was calculated. (c) RMSD plots of the N-terminus ("ELA"), linker region ("GIGIL"), and C-terminus ("TV") of peptide, respectively. The RMSD values for these three regions were calculated after superposition of each pMHC-I conformation from the MD simulations onto the peptide-binding groove of the minimized pMHC-I structure (shown in Figure 1a) by the heavy atoms. (d) Representative conformations for four macrostates; each conformation was randomly chosen from the most populated microstate for each macrostate. For S1 and S3, two different conformations are provided. The equilibrium population of each state is given below the structures, and the corresponding MFPTs are provided on the arrows. Download figure Download PowerPoint Then, the following transition S3 → S4 drives the docking of the remaining parts of peptide into the binding site, including the C-terminus and linker regions. In addition to the HB interactions formed via the N-terminus, four extra HBs attributed to the C-terminus appear, among which, the HB between pT9 and the W147 sidechain is significant, with a HB occupancy <95% (H10 in Figures 3a and 3b). In addition, the sidechains of D77, T143, and K146 can also form HBs with pV10 (H9, H11, and H12 in Figures 3a and 3b). Notably, the H9, H10, and H11 are found to be highly conserved in 239 deposited PDB structures of pMHC-I (Figure 3e); the H12, however, is less conserved, likely due to the flexible sidechain of K146 (Figure 3f). Alternatively, S3 can also transit to an off-path state S2 where two aforementioned HBs, H4 and H8, are profoundly impaired, which dramatically increased the structural of the N-terminal region with S3 and S4 (Figures 3a and 3c). two HBs formed between and can be observed only in with the HB occupancies of and and in Figures 3a and 3b). the peptide C-terminus is exposed to the in S2 and no apparent HB with MHC-I (Figures and 3b). Figure | HB interactions for the peptide N- and C-terminus responsible for MHC-I binding. (a) The HB networks between MHC-I and peptide in S3, and A of 12 HBs are (namely with each HB represented with a black dashed (b) of each HB in (a) for each macrostate. (c) of the MD conformations for S3, and S4 onto two namely the of H4 and H8 in (d) The of H4 and H8 calculated for 239 crystal structures of pMHC-I, the corresponding for S3 are shown in and The of H9, H10, and H11 and population of H12 for 239 crystal structures of The sequence of the binding peptides in 239 crystal structures of The and the from the N- and respectively. The was by Download figure Download PowerPoint In addition to the HB networks formed between MHC-I and the peptide, pL2 and pV10 are found to be responsible for the N- and C-terminal regions in the pockets B and F, respectively, mainly via forming nonpolar Specifically, during the S1 → S3 pL2 can into a by MHC-I Y7, K66, and (Figure The calculated of of the above and pL2 significantly from in S1 to in S3, and the following (Figure the S1 → S3 transition the N-terminal insertion into the B pocket that limits the overall peptide-loading dynamics (see the we thus that the anchor residue pL2 a role in guiding the peptide docking into MHC-I, with the HBs with MHC-I. This conclusion is supported by the statistical analyses of 239 deposited PDB structures appears in the second of peptides with the highest (Figure analyses of HLA-A2 the peptide length varied from to deposited in the also show that the second of the peptide a ( Supporting Information Figure The C-terminal residue similarly to is responsible for loading the peptide C-terminus into the F pocket, also supported the statistical results (Figure and Supporting Information Figure Notably, the second from the and the first from the are relatively conserved among the HLA-A2 other are found to be highly diverse (Figure and Supporting Information Figure which the HBs between MHC-I and peptide are mainly formed via the peptide than the sidechain atoms. Figure | Key structural involved in the S1 → S3 (a) The transition from S1 to S3 to formation of one stable of pL2 and several MHC-I residues that is responsible for the N-terminal binding. Two N-terminal residues, and and the MHC-I residues involved in the interactions are shown in (b) of the MD conformations onto two first tIC and the of the as in (c) The occupancies of the helices α1, and α2-2 in four (d) of helix α2-1 in one S3 conformation, where four critical HBs are highlighted with black dashed The HB occupancies of the above four HBs for each state (e) The numbers three peptide for each namely the N-terminus ("ELA"), linker region ("GIGIL"), and C-terminus The of each above segment was calculated. The mean value was averaged over all the microstates that belong to the same macrostate, and the corresponding standard error was then Download figure Download PowerPoint the S1 → S3 transition is determined to be the step during the peptide-loading process to the mean first time calculations (Figure by the large energy due to the of several structures and the process. Specifically, the helices α1 and α2-2 no significant between S1 and S3, the helix occupancy of the α2-1 motif profoundly from to after the S1 → S3 equivalent to a of turn (Figure It is that the helices α2-1 and α2-2 are linked by a flexible turn (Figure insertion of the peptide N-terminus into the B pocket likely the of helix whereas it allosterically promotes the between α2-1 and resulting in the of Further structural analyses indicate that α2-1 indeed displays conformations in S3 (Figures and The α2-1 is mainly by the HBs between and and and (namely and in Figure which in turn promotes the formation of one HB between and sidechain in Figure In addition, the interaction between and is also in S3 with S1 in Figure which also to a In contrast, the also plays a key role in the S1 → S3 Importantly, are from the regions of the peptide N-terminus after the S1 → S3 which can a significant of energy The following (i.e., S3 → S4 or S3 → however, the dynamics the peptide N-terminus (Figure We further that the in the S1 → S3 transition is mainly via the N-terminal loading than the of the binding groove by