Dipeptides promote folding and peptide binding of MHC class I molecules

Dipeptides Support the Folding of MHC Class I.

The human MHC class I allotype HLA-A*02:01 (A2) has been reported to require specific full-length peptide for folding in vitro (8). To confirm this, we expressed A2 in Escherichia coli, dissolved the inclusion bodies in urea buffer, and folded the denatured protein with its light chain, human β-2 microglobulin (hβ₂m), and with or without the A2-specific peptide NLVPMVATV (from human cytomegalovirus pp65; single-letter amino acid code). We then determined the stability of the formed class I–peptide complexes with the “thermal denaturation by tryptophan fluorescence” (TDTF) assay. This label-free method determines the midpoint of thermal denaturation (T_m or “melting temperature”) of a protein by measuring the sharp decrease in fluorescence at 340 nm that occurs when the protein unfolds and the tryptophan side chains are exposed to the aqueous environment. For MHC class I, the T_m is directly correlated to the stability of MHC–peptide complex (14). The A2 molecule folded with NLVPMVATV showed a clear transition with a T_m of 56 °C, but the folding reaction conducted without peptide did not yield any folded protein, even when NLVPMVATV was added just before the TDTF measurement to stabilize any empty folded A2 molecules (Fig. S1A). Thus, without peptide, recombinant A2 molecules do not fold to achieve a peptide-receptive conformation.

Shorter peptides that correspond to fragments of high-affinity peptides have previously been crystallized in complex with class I molecules (12, 13). We therefore hypothesized that short peptides may support the folding of class I molecules but may then be easily removed because of their low affinity for class I. To test this hypothesis, we performed a folding reaction with the dipeptide GL, which corresponds to the last 2 aa of many class I-binding peptides, with a long aliphatic side chain as anchor residue for the F pocket of the peptide binding cleft (15, 16). A2 was incubated in folding buffer with GL, then NLVPMVATV was added to stabilize any folded molecules, and samples were subjected to TDTF immediately. Strikingly, a stable A2–NLVPMVATV complex was obtained with the same T_m as that of A2 folded with NLVPMVATV (Fig. 1A). This indicates that GL had helped the A2 molecule to attain a folded conformation and had then exchanged for the added NLVPMVATV (Fig. 1C). Quantification of the amount of folded protein in the reaction (performed as described in Fig. S1B) showed that 10 mM GL was approximately half as efficient in folding A2 as 10 µM NLVPMVATV. With the dipeptide glycyl-glycine (GG), in contrast, no measurable transition and no folded protein were observed, just as when no dipeptide was added (Fig. 1A). These results suggest that GL, a dipeptide with a long hydrophobic side chain in the C-terminal position, can support the folding of A2 into a peptide-receptive conformation.

We next investigated the effects of other dipeptides with other hydrophobic residues at the second position. The dipeptide glycyl-valine (GV) also promoted folding of A2, but had a smaller effect than GL (Fig. S2A). In contrast, glycyl-alanine (GA) and glycyl-phenylalanine (GF) had no significant effect.

Next, we asked whether the folding enhancement effect of dipeptides extended to the murine class I molecule H-2K^b (K^b). In contrast to A2, K^b folds in vitro without specific peptide and can bind high-affinity peptides afterward, albeit with low efficiency (14). Intriguingly, GL had a dramatic effect on K^b, more than doubling its folding efficiency (Fig. 1B). K^b molecules that were folded with GL subsequently formed a stable complex with K^b-specific FAPGNYPAL (Sendai virus nucleoprotein 324–332) peptide, with a T_m of 51 °C, and with the same efficiency as if K^b had been folded with FAPGNYPAL in the first place. With the dipeptide GG, no enhancement of folding was observed, whereas GV and GF increased the folding of K^b (Fig. S2B). This is in contrast to A2, for which only GV was effective (Fig. S2A), which suggests that the dipeptides assist folding in an allotype-specific manner.

Taken together, these data show that dipeptides with hydrophobic C-terminal anchor residues assist the folding of the recombinant class I molecules A2 and K^b into a stable and receptive conformation that efficiently binds full-length peptides. Among the dipeptides tested, GL was most efficient for both allotypes, and it is evident that a hydrophobic side chain is required for the folding enhancement effect we observed.

Dipeptides Increase the k_on and Improve Equilibrium Binding of High-Affinity Peptides.

We next asked whether the dipeptides influenced the kinetic association rates (k_on values) of high-affinity peptides. After folding A2 and K^b with dipeptides, we monitored the binding of fluorophore-labeled high-affinity peptides in a time-dependent manner, using fluorescence anisotropy, and calculated the k_on values. We compared them with the k_on values of high-affinity peptides to empty molecules made by folding without any peptide.

The A2 molecule folded with GL showed the fastest binding of high-affinity peptide, with a striking sevenfold increase in the k_on vs. empty folded A2 (Fig. 2A). The dipeptide GV showed a moderate effect and lower total binding, whereas the other dipeptides (GG, GA, and GF) did not show any significant effect, as anticipated from the equilibrium binding data (Fig. 1). Likewise, K^b folded with GL showed sixfold faster peptide binding than the empty folded K^b. The dipeptides GF and GV caused similar increases in the k_on, whereas GA and GG were ineffective (Fig. 2B). GL also increased the k_on values of other high-affinity peptides we tested for A2 and K^b (Fig. S3A), and it also functioned with empty class I molecules (Fig. S3 B and C).

To investigate whether the dipeptide-induced conformation influenced the affinity constant of high-affinity peptide to class I, we performed folding reactions with A2 and K^b with and without GL, and then measured the K_d of the class I molecules to their respective high-affinity peptides, distinguishing bound from free peptide by fluorescence anisotropy. For both A2 and K^b, binding was significantly better in the presence of GL (Fig. 3).

Taken together, the k_on and equilibrium data suggest that GL, and other dipeptides to a lesser extent, promote the rapid binding of high-affinity peptides to A2 and K^b. The simplest explanation for this phenomenon is that GL supports an ordered conformation of these class I molecules that is more peptide-receptive than the empty state.

Binding of the Dipeptide GL to Class I.

We next decided to investigate the molecular mechanism of binding of the dipeptide GL to A2 and K^b, and we followed the hypothesis that GL binds to class I like the two C-terminal amino acids of a full-length high-affinity peptide. In this model, the leucine side chain of GL binds into the F pockets of A2 and K^b and contributes to the binding energy of the dipeptide, which would explain why GG shows no enhancement effect. This hypothesis predicts that the dipeptides leucyl-glycine (LG) and glycyl-d-leucine (GdL, which contains the D enantiomer of leucine) should not support class I folding or rapid association of high-affinity peptide, since, for these two dipeptides, simultaneous binding of the leucine side chain into the F pocket and engagement of the C-terminal carboxylate group in the conserved hydrogen bond network is structurally unfavorable. Indeed, neither LG nor GdL showed the k_on enhancement with A2 or with K^b (Fig. 2 C and D). Thus, the simplest explanation of our data is that the C-terminal leucine side chain and the carboxylate group of GL interact with the residues in and around the F pocket of class I like the C-terminal amino acid of a high-affinity peptide, and that this interaction causes A2 and K^b to improve their peptide binding characteristics.

To investigate whether this mode of binding could explain the enhancement effect of GL, we performed molecular dynamics (MD) simulations. K^b was simulated for 20 ns with the high-affinity peptide FAPGNYPAL and with the dipeptides GL and GG. The starting structure of the simulations was the crystal structure with FAPGNYPAL [Protein Data Bank (PDB) ID code 1KPV], and the dipeptides were introduced by removing the first seven residues of the peptide and then modifying AL to GL or GG, respectively. During the simulations, the side chain of GL remained in the F pocket, whereas GG dissociated completely from the binding groove in the first 2 ns (Fig. 4 A and B). In the absence of any high-affinity peptide or dipeptide, the empty F pocket region of K^b assumes a more closed state, as the α₁ and the α₂ helices approach each other (Fig. 4B, light blue structure). The dipeptide GL held the F pocket open, similar to FAPGNYPAL, suggesting a mechanism for increasing the peptide binding affinity of K^b (Fig. 4A, dark blue structure). No such effect was observed on the A/B or the C pocket regions, which are further removed from the F pocket, the hypothesized binding site of the dipeptide (Fig. 4C, Left). We made the same observation in a simulation of A2, in which GL had a stabilizing effect on the F pocket similar to NLVPMVATV (Fig. 4D, Left). In support of this observation, the binding energy estimates of GL to A2 and K^b calculated from the simulations were significantly more favorable than those of GG (Fig. 4 C and D, Right).

The MD simulation enabled us to propose a detailed model structure of the GL dipeptide bound to the F pocket region of K^b (Fig. 4E). In this model, the N terminus of GL forms a novel hydrogen bond with the side chain of aspartate 77 in the α₁ helix, which may contribute to its binding energy; indeed, k_on experiments with the tripeptide glycyl-glycyl-leucine (GGL) and with acetyl-leucine revealed that the free N terminus of GL is required for the maximum effect of GL (Fig. S3D).

In summary, the simplest explanation for our experimental results is that GL binds to K^b and to A2 like the C-terminal 2 aa of a full-length peptide, and that it thus induces a conformation of class I that is more amenable to binding exogenous full-length peptide.

GL Stabilizes Peptide-Receptive Murine and Human Class I Molecules on the Cell Surface.

At the end of their cellular life cycle, after losing their bound peptide, MHC class I molecules are removed from the cell surface by endocytosis and moved to lysosomes, triggered by a quality control mechanism that is yet uncharacterized (17). We hypothesized that the conformational change brought about by the dipeptides might extend the lifetime of class I molecules at the surface. To test this, we decided to load class I molecules with dipeptides at the cell surface and to observe their persistence. We used transporter associated with antigen processing (TAP)-deficient RMA-S cells and incubated them overnight at 25 °C to accumulate peptide-receptive class I molecules at the cell surface (18–20). We then added Brefeldin A (BFA) to block transport of newly synthesized class I molecules to the cell surface, transferred the cells to 37 °C, kept them in the presence or absence of GL, and measured K^b levels at the surface by flow cytometry with the conformation-specific monoclonal antibody Y3 (Fig. 5A). Without any added peptide, only 25% of K^b molecules were still at the surface after 4 h of incubation, but, in the presence of GL, more than 90% remained on the cell surface. The control dipeptide GA had no significant effect, and, as expected, the control high-affinity peptide SIINFEKL (ovalbumin residues 257–264; amino acid sequence in single-letter code) completely rescued K^b from endocytic destruction.

We next tested whether the GL-stabilized K^b molecules at the cell surface were still peptide-receptive, i.e., could still bind high-affinity peptides. After 4 h of incubation with GL at 37 °C, cells were washed to remove GL from the exterior, and SIINFEKL fluorescently labeled with fluorescein isothiocyanate (FITC) at the lysine side chain (SIINFEK_FITCL) was added to the cells. Excess peptide was washed off, and binding of SIINFEK_FITCL was measured by flow cytometry (Fig. 5B). Intriguingly, the GL-stabilized molecules bound SIINFEK_FITCL, confirming that they were peptide-receptive. Cells incubated with GA, or without dipeptide, bound very little SIINFEK_FITCL. Cells from the same experiments were also investigated by immunofluorescence microscopy (Fig. 5C).

Because GL accumulated K^b molecules at the cell surface in their peptide-receptive form, we next asked whether it would do the same for A2. This is important because peptide-receptive forms of most human class I allotypes cannot be accumulated at the cell surface by a incubation at 25 °C, although exceptions have been described (21–23). We exposed human TAP-deficient lymphoblastoid T2 cells to GL overnight and then tested for binding of the A2-specific peptide, NLVPK_FITCVATV (sequence in single-letter amino acid code; labeled with FITC at the lysine side chain). Strikingly, strong accumulation of peptide-receptive A2 was seen after overnight incubation with GL at 25 °C, but not at 37 °C (Fig. 5D).

Taken together, our results with living cells show that binding of the dipeptide GL can prevent the removal of class I molecules from the plasma membrane, and that GL can thus be used to accumulate peptide-receptive class I at the cell surface.

Preparation and Purification of MHC Class I Heavy Chain and hβ₂m.

A2, H-2K^b (K^b), and hβ₂m were expressed in E. coli, and the denatured proteins were purified from inclusion bodies as described previously (14, 28). Purified proteins (typical final concentrations: A2, 9 mg/mL; K^b, 7 mg/mL; hβ₂m, 15 mg/mL) were stored at −80 °C until use.

In Vitro Folding of MHC Class I Heavy Chain, and hβ₂m.

As described previously (14), heavy chain–hβ₂m–peptide complexes were prepared by diluting the purified denatured proteins (100–200 µg/mL of heavy chain and 50–200 µg/mL hβ₂m) into 2 mL final volume of folding buffer (100 mM Tris·Cl, pH 8, 0.5 M arginine, 2 mM EDTA, 0.5 mM oxidized glutathione, 5 mM reduced glutathione) and incubating at 4 °C (2 d for K^b, 7 d for A2) with 10 nM to 10 µM high-affinity peptide (NLVMPVATV for A2, FAPGNYPAL or SIINFEKL for K^b) or without peptide. To fold in the presence of dipeptides, 1 to 10 mM dipeptide was added to the folding reaction. After folding, samples were ultracentrifuged at 100,000 × g for 20 min, and protein concentration in the supernatant was measured by the Bradford assay.

TDTF Measurement and Step Height Calculation.

TDTF measurements and T_m determination were performed as described previously (14). For the detection of high-affinity peptide binding to the molecules folded empty or with dipeptides, 10 to 100 µM of NLVPMVATV or FAPGNYPAL was added to the heavy chain–hβ₂m complex and incubated (2 h for A2, and 10 min for K^b) before the TDTF measurements. The relative amount of folded protein was calculated from the step height of the TDTF measurement as follows: tryptophan fluorescence denaturation curves were fitted with a two-state single-transition equation (equation MM10 in ref. 14). At the T_m, the difference between the fluorescence of the folded and the unfolded states was calculated as shown in Fig. S1B.

These step heights were normalized to the step height achieved with high-affinity peptide for each individual experiment, and independent experimental repeats were then averaged.

k_on Measurement by Fluorescence Anisotropy.

k_on values were measured with the fluorescently labeled high-affinity peptides NLVPK_FITCVATV and ILKEK_TAMRAVHGV for A2, and SIINFEK_TAMRAL and SIINFEK_FITCL for K^b [sequences of these peptides in the single-letter amino acid code; fluorescent labels are fluorescein isothiocyanate (FITC) or carboxytetramethylrhodamine (TAMRA)]. Labeled peptide (100 nM) was added to the folded heavy chain–β₂m complex (1.25 µM A2, 900 nM K^b), and k_on values were measured by fluorescence anisotropy (TAMRA λ_ex = 536 nm, λ_em = 580 nm; FITC λ_ex = 494 nm, λ_em = 517 nm) with a Cary cuvette fluorimeter with automated polarizers or an Infinite M1000 PRO plate reader (Tecan). Gel filtration or buffer exchange was performed wherever the k_on determination required the removal of dipeptides or excess high-affinity peptides. To test the effect of GL on empty A2 or K^b, GL was incubated for 10 min with heavy chain–β₂m complex, followed by the anisotropy measurement. All the k_on experiments were done at room temperature (22–24 °C) in 20 mM PBS solution, pH 7.4. k_on values were calculated by using GraphPad Prism software.

Steady-State Anisotropy Measurement and K_d Determination.

After folding class I molecules empty or with GL, 100 nM (final concentration) of NLVPK_FITCVATV (A2) or SIINFEK_TAMRAL (K^b) was added to varying concentrations of heavy chain–β₂m complex and incubated for 3 h (for A2) or 10 min (for K^b). The binding was measured by fluorescence anisotropy using an Infinite M1000 reader (Tecan) in a black 96-well plate. Data were fitted, and K_d values were calculated by using GraphPad Prism.

MD Simulations.

The crystal structure of K^b [PDB ID code 1KPV] in complex with the high-affinity peptide FAPGNYPAL served as the starting structure for the MD simulations. The GG and GL starting complexes were created, deleting appropriate atoms from this structure. The ILKEPVGHV crystal structure [PDB ID code 3GSO (29)] represents the starting structure for the A2 bound to the full length peptide, whereas its modification defines the A2 GG complex. The A2 GL complex was modeled based on the SLFNTIAVL crystal structure [PDB ID code 2C7U (30)]. By using the Amber simulation package (31), each complex was placed in an octahedral TIP3 (32) water box (leaving at least 9 Å between protein and the box boundaries) and neutralized with sodium (14/15/14) and chloride (6/7/6) ions in a parm03 force field (33). Each complex was energy minimized, positionally restrained (25 kcal⋅mol⁻¹⋅Å⁻²), and heated from 100 to 300 K in three 0.1-ns simulation runs. After resolving the restraints in five MD simulation runs, each complex was equilibrated for 1 ns and simulated for 20 ns, using a time increment of 2 fs. The particle mesh Ewald method was used to account for long-range electrostatic interactions (34). Binding energy calculations were performed with the Molecular Mechanics Poisson–Boltzmann Surface Area tool of the Amber simulation package (35). The binding energy of the GL– and FAPGNYPAL–K^b complexes was calculated from 5-ns trajectories. Because, for the calculation of the binding energy, only the bound peptide K^b states are relevant, the free energy was determined evaluating only the very beginning of the GG K^b simulation (0.4 ns) before peptide dissociation. The calculated binding energies can be considered only estimates because conformational entropy changes caused by the restriction of the peptide conformation into the bound form are not included. A snapshot of the equilibrated simulation represents the structure of GL bound to K^b (Fig. 4E).

BFA Decay Assay.

RMA-S cells were cultured at 25 °C overnight in CO₂-independent medium to accumulate peptide-receptive class I molecules at the cell surface. The next day, 5 × 10⁵ cells per well were treated with 20 mM dipeptide (GL and GA), with 20 µM SIINFEKL, or kept without peptide, and transferred to 37 °C in the presence of 10 µg/mL BFA (Invitrogen) for 4 h. Cells were harvested at different time points and washed twice with 0.01% NaN₃, and the presence of cell surface K^b was measured with the MAb Y3 and anti-mouse IgG labeled with Alexa Fluor 488 (Invitrogen/Molecular Probes) secondary antibody. Samples were analyzed by using a CyFlow Space (Partec) flow cytometer.

Flow Cytometry with Labeled Peptide.

RMA-S cells were cultured and treated similarly as in the BFA decay assay. As a positive control, instead of SIINFEKL, 1 µM SIINFEK_FITCL was used. After 4 h of incubation at 37 °C, cells were washed twice with 0.01% NaN₃ in PBS solution to remove the dipeptide or full-length peptide. SIINFEK_FITCL 1 µM was added to detect the peptide-receptive cell surface K^b molecules, and they were incubated for 10 min and washed twice with PBS solution. K^b-bound SIINFEK_FITCL was detected by flow cytometry.

Immunofluorescence Microscopy.

RMA-S cells were treated similarly as in the BFA decay assay. After incubation with secondary antibody, cells were washed twice with PBS solution and transferred to a coverslip coated with poly-l-lysine. Then, they were incubated at 37 °C for 30 min and fixed with paraformaldehyde [2% (wt/vol) in PBS solution, pH 7.4]. The coverslip was mounted onto the microscope slide with Mowiol (Applichem). Cells were observed with a Zeiss LSM 510 confocal microscope.

Accumulation of Cell Surface A2 Molecules.

Human TAP-deficient lymphoblastoid T2 cells were incubated overnight at 25 °C or 37 °C with 20 mM dipeptide (GL or GA), 5 µM NLVPK_FITCVATV, or without peptide. Next day, cells were washed two times with PBS solution, resuspended in medium containing 10 µg/mL BFA, and incubated with 5 µM NLVPK_FITCVATV for 30 min at room temperature. After incubation, cells were washed with PBS solution, and A2-bound NLVPK_FITCVATV was detected by flow cytometry.

Peptides.

Sequences of all peptides are given in the standard single-letter amino acid code. The fluorescent dyes attached to the peptides were fluorescein isothiocyanate (FITC) and carboxytetramethylrhodamine (TAMRA). The high-affinity peptides SIINFEKL, FAPGNYPAL, and NLVPMVATV, and the fluorescently labeled peptides NLVPK_FITCVATV, ILKEK_TAMRAVHGV, SIINFEK_FITCL, and IRAAPPPK_FITCF were from Genecust, purified by HPLC to >90% (wt/wt) purity. The dipeptides GL, GG, GA, GV, GF, LG, and acetylated leucine were procured from Bachem or Sigma-Aldrich. SIINFEK_TAMRAL, FAPK_TAMRANYPAL, GGL, GL-NH₂, and G-dL were synthesized and donated by Hubert Kalbacher (Tübingen University, Tübingen, Germany).