Introduction

After invasion of bacteria, viruses, or toxins, the human immune system uses a special way to detect it. Small parts (peptides) of invading agents are presented by major histocompatibility complex (MHC-I or MHC-II) to the T-cell receptor (TCR). The molecular recognition and complex formation between the peptide (epitope) and MHC (pMHC) is crucial for this detection. These complexes are not only relevant for pathogen detection but also for the development of vaccines that have considered as the most effective method of preventing infectious diseases [1]. All vaccinations work by presenting a foreign antigen to the immune system in order to evoke an immune response. The active agent of a vaccine may be intact but inactivated (“attenuated”) forms of the causative pathogens (bacteria or viruses), or purified components of the pathogen that have been found to be highly immunogenic. Increased understanding of antigen recognition at molecular level has resulted in the development of rationally designed peptide vaccines.

Research on vaccine development tries to find epitopes comprising continuous amino acid sequences that do not cause cross-reactivity with the parent protein or whole microorganism [2] but still bind to MHC and stimulate immune reaction (immunization). In order to achieve this aim, molecular mimicry to the cognate antigen is required for inducing an immune response [2].

The accelerating growth of bioinformatics techniques together with the considerable amount of experimental data has given rise to a new field called immunoinformatics [3] that deals with in silico analysis and modelling of immunological data. Despite that computational identification of potential epitopes that possess molecular mimicry and may activate the immune system is a challenge in immunological research today, this research has been facilitated by the employment of algorithms. In the quest of finding candidates to design epitope vaccines, several approaches have been applied. Some groups have achieved molecular mimicry by using different strategies such as by a trial-and-error approach [4–9], by combinatorial chemistry approaches [10, 11], by structure-based design techniques [12–14], or by coupling the selection of epitopes through immunoinformatics procedures and then validating its binding properties through docking and qualitative MD simulations techniques [15].

The outbreak of a new swine-origin strain of influenza A H1N1 emerged in 2009 caused widespread human infection [16]. Among one of the most important surface proteins, hemagglutinin (HA) has been a promising drug target, which is central to the virus to bind to cell membrane and infect the cells. Unfortunately, because of the high rate of mutation experienced by the virus that confers the ability to escape from either T-cell or antibody recognition, current flu vaccines were unable to defeat the virus. Thus, there is still a need to identify some epitopes amenable to inhibit the function of HA.

Insights into the molecular recognition of epitopes that exhibit affinity to MHC-II (pMHC-IIs) and therefore possess the ability to be presented to TCR could be exploited to facilitate the identification of candidates to design a new peptide-based vaccine. Structural data have provided some initial insights into the substrate preference of MHC-II; generally, peptides are recognized by MHC-I and MHC-II through a group of pockets (Pks), in which the epitopes are accommodated through its side chains of peptide residues. These Pks are classified into two classes: the class comprising Pk1, Pk4, Pk6, and Pk9, which have been identified as major anchors and are localized in solvent-inaccessible regions, and the class comprising Pk2, Pk3, Pk7, and Pk10, which are smaller pockets that function as auxiliary anchors [17]. Molecular dynamics simulations of pMHC-II complexes in soluble conditions have shown that misleading structural and energetic results are present when the entire system (light and heavy α and β) is not taken into consideration [18]. MD simulations of pMHC-II complexes in soluble or membrane environment also showed structural and energetic differences, being the membrane-bound complexes more energetically favored [19]; therefore, this latter approach was chosen in this study.

The aim of the study is to provide a better understanding of the molecular nature of the interaction between MHC-II and peptides on the ns timescale, taking into account the effects of lipid environment (Fig. 1), and based on this analysis discriminate among a group of peptides to the development of a peptide vaccine. In this contribution, the equilibrated interaction energy, the interaction type, and the key residues were investigated. We have performed molecular dynamics (MD) simulations in combination with molecular mechanics generalized Born surface area (MMGBSA) free energy calculations of a homology model of MHC-II that includes the transmembrane regions [19] in complex with four epitopes that have the highest molecular mimicry to the cognate protein (HA) and will thus most likely be good vaccine candidates (Table 1). To this aim, these four epitopes were coupled to the MHC-II through docking procedures to assess their binding conformation with minimum binding energies. Then, to determine their energetic and structural contributions, they were submitted to MD simulation combined with MMGBSA method to determine their affinity. After that, to validate the competence of P1–P4 in vivo, these peptides were synthesized and tested experimentally, revealing that P4 elicited the highest strong humoral immune response to HA of the H1N1 in comparison with P1–P3, suggesting that it could be a good candidate for the development of a peptide-based vaccine and also demonstrating that the theoretical methods used here are able to discriminate between peptide models.

Fig. 1
figure 1

MD simulation box of the pMHC-II complex embedded in a POPC membrane (equilibrated p1MHC-II complex was chosen to represent)

Full size image
Table 1 Binding free energy components of protein–ligand complexes (in units of kcal mol−1)
Full size table

Results and discussion

In order to investigate the more stable conformation between MHC-II and the four epitopes, MD simulations of MHC-II anchored in a POPC membrane (Fig. 1) bound to four immunogenic peptides were carried out in combination with free energy calculations. The investigated peptides were P1 (SSWSYIVETPSSDNGTCYPG), P2 (KTSSWPNHDSNKGVTAACPHAGAKSFYKNL), P3 (KKFKPEIAIRPKVRDQEGRM), and P4 (SLPFQNIHPITIGKCPKYVKSTK).

Previous studies have stated that the epitope recognition process takes place through the well-known MHC-II Pockets (Pk1–Pk11) localized along the recognition binding groove [17, 19], with the Pk1 being particularly important [20, 21]. In order to further investigate the role of the Pk1–Pk11 residues, the binding model between each peptide and MHC-II was obtained through protein–protein docking calculations, and the model with the lowest binding free energy was chosen and then submitted to MD simulation in a lipid environment.

Convergence and equilibrium examination

To evaluate whether the systems reached equilibrium, some structural and energetic properties during the MD simulations (Fig. 2) were monitored, such as area per lipid, the root-mean-square deviation (RMSD) of backbone atoms with respect to the initial structure, the atomic fluctuation (RMSF) of the α-carbon atoms, and the sum gas-phase and solvation free energies. Figure 2a shows that the time evolution of the area per lipid for the four pMHC-II complexes exhibits very high area per lipid values at the beginning of the simulation but decreasing toward a converged area per lipid on the order of 50–60 ns, with area per lipid values in good agreement with those found for other protein–POPC–membrane systems [19, 22]. Figure 2b shows that despite some complexes (P1MHC-II and P4MHC-II) reach equilibrium at the first 40 ns of simulation, the four complexes reach convergence over the last 30-ns MD simulations with RMSD values of 3.54 ± 0.18, 4.53 ± 0.33, 5.1 ± 0.34, and 3.7 ± 0.30 for P1MHC-II, P2MHC-II, P3MHC-II, and P4MHC-II, respectively. RMSD values increased for the four systems when the cytoplasmatic domain (α-domain, residues 215–229 and β-domain, residues 222–237) is included in the analysis, with RMSD values of 5.2 ± 0.30, 5.6 ± 0.34, 5.8 ± 0.30, and 4.85 ± 0.23 for P1MHC-II, P2MHC-II, P3MHC-II, and P4MHC-II, respectively (Figure 1S), pointing out a high mobility for this protein domain. RMSF analysis over the equilibrated simulations (the last 40 ns of MD production runs) time corroborates the high mobility observed for the cytoplasmic domain and also shows that the four complexes exhibit similar conformational mobility but with lower mean RMSF values for p2MHC-II and p4MHC-II complexes, suggesting a higher structural stiffening upon complex formation (Fig. 2c).

Fig. 2
figure 2

Equilibrium and energetic stability observation for MD simulations of pMHC-II complexes. a Area per lipid values of the pMHC-II complexes during the 100 ns of MD simulation. b RMSD values of pMHC-II complexes are shown with respect to the starting equilibrated structure. c Atomic fluctuations (RMSF) of α-carbon atoms of the pMHC-II complexes calculated over the last 40 ns of the 100-ns-long MD simulations. d Sum of gas-phase and solvation free energies for pMHC-II complexes calculated for 100 snapshots extracted at 1-ns intervals from the 100-ns-long MD simulations

Full size image

Free energy calculations

Although RMSD behavior provides information about the time when the system reached equilibrium, fluctuations of effective binding free energies must be converged to obtain reliable effective binding free energies for the pMHC-II complexes. Figure 2d illustrates the time when the four complexes showed constant effective binding free energies. This demonstrates that P1MHC-II and P4MHC-II complexes reach stable effective energies after 40 ns, whereas P2MHC-II and P3MHC-II effective energies become stable after 60 ns, result that is consistent with the RMSD analysis (Fig. 2b). Thus, the following analysis of non-covalent interactions, effective binding energies, and decomposition binding free energies is based on the snapshots obtained after 60 ns of the MD simulations time.

Our analysis shows that the calculated effective binding free energies (ΔG mmgbsa) for the four pMHC-II complexes without the contribution of the entropy was found to be energetically favorable for all the complexes, indicating that favorable protein–peptide complexes are formed. However, these calculated values overestimate the binding free energy, which is reasonably due to two missing contributions: the lack of entropic contributions, which can be expected to be unfavorable in the case of the flexible peptides, bringing the calculated binding free energy close to the experimental values; the lack of energetic contributions due to conformational changes experienced upon complex formation, which are not taken into consideration when the single trajectory approach is used. Furthermore, although it has been stated that MMGBSA is not able to reproduce the absolute experimental binding free energies accurately, it can give good ranks for the experimental values [23, 24]. Therefore, according to the predicted binding free energies in this study, the P4MHC-II shows the best binding free energy, indicating that this peptide might exhibit a better immunological response than the other three peptides.

According to Table 1, we found that the differences in ΔG mmgbsa between MHC-II and the four peptides are the result of a counterbalance between the polar (ΔE polar) and nonpolar (ΔE nonpolar) contributions, where the major term that favors the binding is the ΔE nonpolar, being the van der Waals interactions (ΔE vdw) the main contributors, whereas that the nonpolar solvation terms (ΔG npol-sol), corresponding to the burial of solvent accessible surface area (SASA) upon complex formation, contribute slightly to binding for the four complexes. The polar interactions (ΔE polar) exhibit a significant unfavorable behavior that opposed the binding, suggesting that this molecular recognition process is hydrophobic in nature.

Structural analysis of the peptide-bound complex

Different structural and MD simulation studies have confirmed that several Pks (Pk1–Pk11) are involved in the epitope MHC-II recognition [15, 17, 19], being the Pk1 pocket, which is formed by bulky hydrophobic residues such as Trp, Tyr, Phe, Leu, and Ile [25], particularly important [20, 26]. A previous report of our research group pointed out that independently of the peptide size, they were stabilized by similar PKs [19]. Outcome is in agreement with the present results where it is observed that, despite the differences in peptide size, the four peptides reached a tight and long-lived conformation in the peptide-binding groove with residues in Pk1, Pk4 and Pk6-7.

To seek initial insights into the binding determinants of the pMHC-II complexes, the most populated conformation for the four complexes was calculated through a cluster analysis for the last 40 ns of MD simulations (see methods). These complex conformations show that despite both the differences in the peptide sequences and peptide sizes, the four peptides were stabilized through hydrophobic interactions by about 26–30 residues distributed along Pk1, Pk4, and Pk6–7 (Figure 2S). Hydrogen bonds (HB) and salt bridges (SB), although in a lesser degree, also contribute in maintaining these complexes. Figure 3 shows in at least three of the four peptide complexes very strong canonical HBs or SBs with an occupancy of at least 40 %, indicated in red and blue. For the P1MHC-II complex, five HBs are formed between the side chain group of P1 and the carbonyl group of E55 (α), between the side chain groups of P1 and E55 (α), between the side chain group of P1 and the carbonyl group of I63 (α), between the side chain groups of P1 and D66 (α), and between the side chain group of K39 (α) and the carbonyl group of P1. For the P2MHC-II complex, two HBs and two SBs are observed, the two former are observed between the side chain group of P2 and the carbonyl group of D66 (α), and between the carbonyl group of P2 and the side chain of Q64 (β), whereas the two latter are formed between the side chain groups of P2 and E11 (α), and between the side chains groups of P2 and D66 (α). For the P3MHC-II complex, one HB and four SBs are formed, where the HB is formed between the backbone atoms of P3 and the side chain group of K71 (β), and the SBs between the side chain groups of P3 and E55 (α), between the side chain groups of P3 and D66 (α), between the side chain groups of P3 and H81 (β), and between the side chain groups of P3 and D28 (β).

Fig. 3
figure 3

Scheme of hydrogen bonds (HBs) or salt bridges (SBs) formed between MHC-II and the peptide substrates [p1 (a), p2 (b), p3 (c), and p4 (d)] along the last 40 ns of MD trajectories. The peptide orientation is shown from N- to C-terminus (left to right). Canonical HBs or SBs are classified into three groups according to their occupancy: low (occupancy of 10–40 %), moderate (occupancy of 40–80 %), and strong (occupancy of 80–100 %) and are represented as green, red, and blue arrows, respectively (Color figure online)

Full size image

For the P4MHC-II complex, two HBs and five SBs are established. The two HBs are formed between the backbone atoms of P4 and S53 (α), and between the backbone atoms of P4 and the side chain of K71 (β). The five SBs are formed between the side chain groups of P4 and E55 (α) (two salt bridges for this interaction), between the side chain groups of P4 and D66 (α), between the side chain groups of P4 and D66 (β), between the side chain groups of P4 and E11 (α). Overall, these MHC-II residues are lining pockets Pk1, Pk4, and Pk6–7.

Based on this analysis, it is clear that D66 (α), E55 (α), K71 (β), and E11 (α) play an important role in stabilizing most of the complexes, since independently of the residues forming the peptide sequence, the binding is preferentially mediated through electrostatic interactions between these MHC-II residues. According to this analysis, the most stable complex is P4MHC-II complex, given the fact that this is the complex with the highest number of HBs and SBs (Fig. 4). Furthermore, all the residues observed establishing both hydrophobic and electrostatic interactions are in good agreement with those observed in other reports [19].

Fig. 4
figure 4

Summary plot of HBs and SBs formed between MHC-II and the peptide substrates along the last 40 ns of MD trajectories. Only the HBs and SBs with moderate (occupancy of 40–80 %) and strong (occupancy of 80–100 %) are considered

Full size image

Identification of key residues participating in the binding by MMGBSA free energy decomposition

To search for the dominant factors that guide the peptide-binding specificity, the binding free energy was decomposed into the contributions of each residue based on MMGBSA approach and the single trajectory method (see methods). As shown in Table 1, the gas-phase van der Waals and electrostatic interaction energies between MHC-II and the four peptides, especially the latter, determine the difference in the binding affinities of the four complexes. Figure 5 shows that although the four complexes shared similar residue interactions as identified through the hydrophobic and electrostatic interaction analysis (Figure 2S and 3), their contribution to the total effective binding free energy (ΔG mmgbsa) was different, being P3MHC-II and P4MHC-II those that exhibit the largest contributions to the ΔG mmgbsa as demonstrated through the MMGBSA free energy calculations for the four pMHC-II complexes.

Fig. 5
figure 5

Per-residue contribution to the binding effective energy of pMHC-II complexes is depicted as bar plots. Per-residue contributions were calculated by the MMGBSA decomposition method. Residues whose contributions to the effective energy ΔG Eff ≥ −0.1 kcal mol−1 are showed with their respective error

Full size image

In the case of P1MHC-II is observed that the largest contributions to the ΔE total came from the Q18(α), M36(α), F51(α), S53(α), E55(α), N62(α), I63(α), K67(α), and A68(α). Among these residues highlight Q18(α), E55(α), I63(α), and N62(α) that are observed establishing HBs with P1. However, D66(α) and K39(α), for which a considerable energetic contribution could be expected given the finding of strong hydrogen bonds formed by the side chains of P1, do not show any energetic contribution, which is due to the fact that despite these residues they exhibit a favorable electrostatic interaction as calculated by the molecular mechanics force field (ΔE ele). This is cancelled by the unfavorable electrostatic contribution due to desolvation (ΔG ele,sol), whereas the van der Waals contributions (ΔE vdw) and the nonpolar part of the solvation free energy (ΔG npol,sol) scarcely contribute to the binding (Figure 3S, panel A). For P2MHC-II, it can be observed that A52(α), S53(α), F54(α), Y60(β), W61(β), Q64(β), L67(β), and N82(β) are the main contributors to the ΔG mmgbsa, from which only Q64(β) is observed forming a strong HB with P2. Another residues also forming HBs but that contribute scarcely to ΔG total are E11(α), N62(α), D66(α), Y78(β), and N82(β). Energetic contribution is understandable for the case of N62(α), Y78(β), and N82(β), since these three residues form weak hydrogen bonds with P2 (Fig. 3); however, for the case of E11(α) and D66(α), some repulsive effects may contribute to reducing the favorable energetic contribution expected by these residues given their strong electrostatic interactions (Fig. 3), whereas that for Q57(α) and Q9(α), there are no energetic contribution due to the cancellation of the favorable electrostatic interactions as calculated by the molecular mechanics force field (ΔE ele) by the unfavorable electrostatic contribution due to desolvation (ΔG ele,sol) (Figure 3S, panel B).

In the case of P3MHC-II, it is observed that six residues contribute to the ΔG mmgbsa: F51(α), F54(α), E55(α), V65(α), D28(β), Y30(β), L67(β), Q70(β), Y78(β), and H81(β) from which E55(α), D28(β), Y30(β), Q70(β), Y78(β), and H81(β) form HBs or SBs (Fig. 3), whereas that K71(β) and Q64(β) despite forming strong HBs, at least in the case of K71(β), for both residues their favorable energetic contribution is cancelled by the unfavorable polar desolvation (Figure 3S, panel C).

For P4MHC-II complex, it is observed that E11(α), F51(α), A52(α), S53(α), E55(α), V65(α), H13(β), L67(β), K71(β), Y78(β), and V85(β) are the main contributors to the ΔG mmgbsa. Among these residues, E11(α), S53(α), E55(α), and K71(β) are observed to form strong HBs or SBs (Fig. 3). Although D66(α) and D66(β) exhibit strong electrostatic interactions, these may be decreased by some repulsive effects, reducing their energetic contributions, whereas for N62(β), for which a weak HB is observed, its favorable electrostatic interactions are cancelled not only by the unfavorable electrostatic contribution, but also by a unfavorable contribution to the ΔG mmgbsa (Figure 3S, panel D).

Taking into account only those residues whose energetic contribution to the total effective energy are <−2 kcal mol−1: Q18(α), M36(α), F51(α), and S53(α) for P1MHC-II, A52(α), S53(α), F54(α), Y60(β), W61(β), L67(β), and N82(β) for P2MHC-II, F51(α), F54(α), E55(α), V65(α), D28(β), L67(β), Q70(β), and Y78(β) for P3MHC-II, F51(α), A52(α), S53(α), E55(α), H13(β), L67(β), K71(β), Y78(β), and V85(β) for P4MHC-II. It becomes apparent that the distribution of key residues varies depending on the investigated peptide sequence. Therefore, the peptide sequences of P3 and P4 could be used as template to improve their binding affinity toward MHC-II.

Immunoassay of peptides

The immunogenicity and antigenicity of the four peptides were analyzed by the recognition of the hemagglutinin (A/California/04/2009 (H1N1) 11055-V08h) using the IgG antibodies from the serum of the immunized rabbits by ELISA (Fig. 6). Our results show that the titers of IgG antibodies to the hemagglutinin (A/California/04/2009 (H1N1) 11055-V08h) were significantly higher in the serum of the rabbit immunized with the peptide P4 (SLPFQNIHPITIGKCPKYVKSTK) residues 295–317) (dilution 1:400–1:9600) than those achieved by the other three sera.

Fig. 6
figure 6

Detection of rabbit IgG Ab to the hemagglutinin A/California/04/2009 H1N1 by ELISA. Serial dilutions of sera from rabbits inoculated with peptides conjugated to KLH were added to microplates previously coated with hemagglutinin A/California/04/2009 H1N1 [Cal09 (H1)HA]. The IgG Ab to the hemagglutinin peptides were detected with a secondary Ab specific for rabbit IgG. The serum from rabbit immunized with peptide 4 showed the highest titer (>1:12,800), whereas the serum anti-peptide 2 had a lower titer (1:3200), and the serum anti-peptides 3 and 1 had the lowest titer (1:1600)

Full size image

Based on this result, it is clear that immunological analysis confirmed the theoretical studies, which predict that most favorable interactions are formed between P4 and MHC-II. Because a better interaction probably allows an enhanced peptide presentation by MHC-II molecules and a strong immune response, the pMHC-II complexes with high-affinity peptides might be efficiently presented to T helper cells and therefore induce stronger or more efficient humoral and cell-mediated immune response [27–29]. Thus, the peptide 4 can be an immuno-dominant epitope from the hemagglutinin and an excellent candidate for a peptide-based vaccine.

Although by using this experimental methodology we did not directly demonstrate the interaction of the four peptides with MHC-II molecule, there is indirect evidence that the peptides bind to MHC-II in vivo. When rabbits were immunized with the peptides, they produced specific antibodies of the IgG isotype. The production of IgG antibodies requires activated T helper cells that recognize the antigen (epitope T) in association with the MHC-II protein on the surface of antigen-presenting cells [30–33].

Conclusions

In order to understand the structural and energetic properties required to obtain favorable pMHC-II complexes, we employed an integrated computational approach by combining protein–protein docking, MD simulations, MMGBSA binding free energy calculations, and MMGBSA per-residue decomposition analysis. Starting from the MHC-II and four peptides previously designed, four potential complexes were predicted by protein–protein docking, in which each peptide was accommodated along the peptide-binding groove and stabilized by the canonical binding pockets. Although it could be expected that these complex conformations would be good enough to discriminate among different peptides, it is known that other more certain methodologies are needed to perform a more reliable peptide selection. Therefore, MD simulations and free energy calculations were performed for the four different complexes. The analysis of the non-covalent interactions along the MD simulations indicated that despite the peptide size and sequence, several canonical HBs and SBs are formed between the peptides and the MHC-II binding site cleft. The free energy decomposition revealed that van der Waals interactions and the nonpolar part of solvation free energy dictate the binding strength of the peptides, whereas the binding specificity is determined by electrostatic interactions and the polar part of solvation free energy. Result is consistent with the outcome of the free energy decomposition for most of the residues, which revealed that, in general, P3 and P4 have the largest contribution to the effective free energy. These two latter peptides revealed strong salt bridges whose favorable electrostatic interactions, as calculated by the molecular mechanics force field (ΔE ele), outweigh the unfavorable electrostatic contribution due to desolvation (ΔG ele,sol).

Calculations of this type are very useful to discriminate among residues stabilizing the complex, and based on this, it is clear that some mutations could be introduced in P3 and P4 in order to increase their affinity toward MHC-II, especially by introducing mutations that create new HBs or SBs. It is encouraging that, although the present study is based on only a complex conformations predicted through theoretical approaches, the theoretical methodology was able to discriminate between the four peptides proposed, those with a weak humoral immune response (P1–P3) and those with a strong humoral immune response to hemagglutinin H1N1 (P4), a result that not only is in line with the experimental data obtained here, but also suggests that P4 could be used for the development of a peptide-based vaccine. Furthermore, because this peptide sequence is located in a conserved region, it might induce antibodies that are cross-reactive to other influenza subtypes.

The simulations of the pMHC-II complexes show that the four peptides are inserted along the peptide-binding groove in a variety of conformations, which is guided by the electrostatic charge distribution between the receptor and the peptides (data not showed). We identified several energetic and topological factors that affect the peptide-binding. The physical insights concerning the conformational propensities of epitopes inserted into MHC-II provided by this work will facilitate the design and selection of peptide vaccine candidates, which may have a higher chance of success than those chosen only using qualitative empirical approaches, and has the advantage that only a limited number of peptides needs to be synthesized and tested in an immunological assay. This work is currently under way in our laboratory

Materials and methods

Construction of the complexes

Four pMHC-II complexes between MHC-II and four epitopes were constructed through docking procedures. The three-dimensional (3D) structure of MHC-II was taken from a previous study, where the transmembrane domain was modeled through homology modeling methods [19]. The sequences of P1 (SSWSYIVETPSSDNGTCYPG, residues 80–99 from PDB ID: 3UYX), P2 (KTSSWPNHDSNKGVTAACPHAGAKSFYKNL, residues 125–154 from PDB ID: 3UYX), P3 (KKFKPEIAIRPKVRDQEGRM, residues 211–231 from PDB ID: 3LZG), and P4 (SLPFQNIHPITIGKCPKYVKSTK, residues 295–317 from PDB ID: 4JTV) were identified by employing a similar combination of methods described elsewhere [15], and their structures were constructed using I-TASSER server [34]. The blind protein–protein docking studies were performed using the Cluspro 2.0 server [35, 36], and the models with the lowest energy were selected.

Anchoring the complexes into the membrane

pMHC-II complex orientation with respect to the membrane was performed using OPM (Orientations of Proteins in Membranes) server [37]. A rectangular pre-equilibrated palmitoyl oleoyl phosphatidylcholine (POPC) phospholipids bilayer was generated for each system using the membrane builder tool of CHARMM-GUI [38, 39] 110.56 × 110.5 Å (xy). In order to place the receptor into the bilayer, the replacement method was used [38, 40, 41]. The membrane consisted of about 340 POPC phospholipids with ~170 on the top and bottom, respectively (Fig. 1). The size along the Z-axis (136.592 Å) was determined by specifying the thickness of bulk water from the protein extent along Z (Z center = −20.172 Å). The membrane–receptor complexes thus obtained were solvated and neutralized using the solvation and autoionize modules (Ion placing method) of CHARMM-GUI membrane builder [39]. The ionic strength was kept at 0.15 M by NaCl, and we used TIP3 water model. The all-atom models of each system were generated by using the CHARMM force-field parameters (http://mackerell.umaryland.edu/CHARMM_ff_params.html). Then, these files were converted into files that conform to the Lipid11 naming convention using charmmlipid2amber.x script included in the AMBER 12 package [42]. Lipid11 is the Amber lipid force field that includes parameters from the general amber force field (GAFF) [43, 44]. The topologies for the pMHC-II complexes were constructed by the Leap module using the ff99SB and Lipid11 force field [42, 43]. Then, these structures were minimized and heated with sander by the following three steps. First, the systems were energy-minimized in 10,000 steps under the NVT ensemble, with position restraints (the force constant was set to 500 kcal−1 mol−1 Å−2) on both the protein and lipids to allow for relaxation of the water molecules. After this equilibration step, the system was slowly heated in the NPT ensemble through two sequential runs from 0 to 300 K while keeping the protein atoms and lipids restrained (the force constant was set to 10 kcal−1 mol−1 Å−2). First, the system was heated from 0 to 100 K using the Langevin thermostat, and the second phase of heating consisted of slowly increasing the temperature to the production temperature (300 K). The positions and velocities were read from the previous temperature heating, and an anisotropic Berendsen weak-coupling barostat [45] was also used to equilibrate the pressure, whereas the Langevin thermostat was used to equilibrate the temperature.

The MD simulations were performed with pmemd.cuda AMBER 12 executable [42], which allowed to accelerate explicit solvent particle mesh Ewald (PME) calculations [46] through the use of GPUs [47, 48]. 100-ns-long MD simulations were run with the input files without position restraints under periodic boundary conditions (PBC), and using a NPT ensemble at 300 K, the electrostatic interactions were calculated via the PME method [46], and a 10-Å cutoff was used for the van der Waals interactions. The bonds between the heavy atoms and hydrogens were constrained with the SHAKE algorithm [49]. Temperature was controlled here using Langevin dynamics, while pressure is controlled using a semi-isotropic constant surface tension to maintain a specific area per lipid. The collision frequency was 1.0 ps−1. The pressure was maintained at 1 bar, and the pressure coupling constant was set to 1 ps. The time step of the MD simulations was set to 2.0 fs, and the coordinates were saved for analyses every 1 ps.

Analysis of MD trajectories

The “ptraj” module of Amber 12 was used for analyzing the root-mean square deviation (RMSD) between structure pairs and the root-mean square fluctuations (RMSF) about the mean position of atoms. The area per lipid was calculated using the “cpptraj” module of Amber 12 by dividing the area of our simulation box ((box X length) × (box Y length)/(number of phospholipids per layer). Before RMSD and RMSF calculations, overall translational and rotational motions were removed. RMSD were calculated for all heavy atoms, and RMSF values were calculated only for the α-carbon atoms. Average conformations were obtained through a cluster analysis using the kclust algorithm that belongs to the MMTSB toolset http://mmtsb.scripps.edu/software/mmtsbtoolset.html; then, from the most populated cluster, the conformation with the lowest RMSD to the cluster center (centroid) was selected. Hydrogen bonds were defined by a distance cutoff of 3.0 Å and an angle cutoff of 120°, and only considered if their occupancies attained >10 %. Images were generated using LIGPLOT version 4.5.3 [50], pepDraw http://www.tulane.edu/~biochem/WW/PepDraw/, and structural representations were prepared using PyMOL version 0.99 [51].

Calculation of effective binding free energies and per-residue contributions

Binding free energies were calculated using the MMGBSA method [52–54] following the “single trajectory approach.” Under this methodology, the conformations of the binding partners are taken from the pMHC-II complexes obtained along the MD simulation; therefore, the energetic contributions due to conformational changes are neglected; however, this approach leads to a drastic reduction in the statistical uncertainty of the free energy components [52].

Before the calculations, all counterions and water molecules were stripped from the snapshots for each system, 400 snapshots at time intervals of 100 ps from the last 40-ns simulations runs, using a salt concentration of 0.1 M and the Born implicit solvent model of 2 (igb = 2) [55]. The analyses were performed using the MMPBSA.py script [54] provided in the Amber 12 package [42]. The binding free energy (ΔG bind) of each complex can be calculated as:

$$\begin{aligned} \varDelta G_{\text{bind}} & = G^{\text{complex}} - G^{\text{receptor}} - G^{\text{ligand}} \\ & = \varDelta E_{\text{MM}} + \varDelta G_{\text{GB}} + \varDelta G_{\text{SA}} - T\varDelta S \\ \end{aligned}$$

where ΔE MM is the gas-phase interaction energy between the receptor and ligand, which includes the van der Waals (ΔE vdw) and the electrostatic (ΔE ele) interaction energies. ΔG GB and ΔG SA are the electrostatic and nonpolar contributions to desolvation upon ligand binding, respectively, and −TΔS is the entropy contribution resulting from the changes in the degrees of freedom of the solute molecules, which were not considered here; therefore, our values reported for the MMGBSA calculations should be named effective binding free energies ΔG Eff rather than absolute binding free energies.

In order to detect the hot spot residues, the absolute binding free energies were decomposed into the contribution of individual residues using the theory of free energy per-residue decomposition [38].

Rabbit immunization

Four rabbits were immunized with each peptide conjugated to KLH on three different days. On day 1 (first immunization), 1 mg of peptide plus complete Freund’s adjuvant (Sigma Chemical Co.) was administered through the subcutaneous route. On day 8 (second immunization), 1 mg of peptide plus incomplete Freund’s adjuvant was administered through the subcutaneous route. On day 15 (third immunization), 1 mg of peptide in 5 ml of saline solution was administered through the intramuscular route. Seven days after the last immunization, the rabbits were anesthetized with pentobarbital, and serum samples were obtained from blood extracted by cardiac puncture and stored at −70 °C.

Detection of hemagglutinin H1N1 using serum of rabbits immunized with study peptides

Anti-peptide levels were evaluated in serum samples from the immunized rabbits. The antibody levels in the serum samples were determined using an indirect enzyme-linked immunosorbent assay (ELISA). Briefly, 96-well plates were coated with hemagglutinin (A/California/04/2009 (H1N1) 11055-V08h) in carbonate bicarbonate buffer (15 mM Na2CO3, 35 mM NaHCO3, at pH 9.6). The plates were incubated for 2 h at 37 °C and washed three times with 0.05 % Tween-20 in PBS (PBST). Blocking was performed by treatment with PBST plus 6 % fat-free milk and by further washing with PBST. Each sample was tested in duplicate. Serum samples from the four immunized rabbits were placed in serial dilutions, starting from a 1:100 to a dilution 1:16,000. The plates were incubated overnight at 4 °C and washed with PBST; then, 100 µl of goat anti-rabbit IgG (Pierce, Rockford, Ill) was added per well, and the plates were incubated for 2 h at room temperature. The plates were washed with PBST, and the enzymatic reactions were started by adding substrate solution (0.5 mg of o-phenylenediamine per ml, 0.01 % H2O2 in 0.05 M citrate buffer [pH 5.2]). After 15 min, the reactions were stopped with 25 µl of 2.5 M H2SO4 and the absorbance at 492 nm (A492) was measured in a Multiscan Ascent (Thermo Labsystems) microplate reader. The end point titer was defined as the reciprocal of the highest analyzed dilution that gives a absorbance value above 0.5 [56].