Introduction

Bacteria take up many types of molecule using binding protein-dependent permeases, part of a much larger group of transport systems known as the ATP-binding cassette (ABC) transporters, one of the largest and most diverse protein families known (Higgins 1992). Unlike their eukaryotic counterparts, the bacterial ABC import systems also involve a globular protein family which acts as the initial receptor of the ligand to be brought into the cell. In this family, the one of the most characterized proteins is the periplasmic oligopeptide-binding protein (OppA), which is a 58.8 kDa periplasmic peptide receptor found in Gram-negative bacteria (Monnet 2003). Unlike other members of this protein family, OppA has a broad specificity in the peptide ligands that it binds. Early studies found that the OppA can carry peptides of two to five amino acid residues regardless of their size, charge, and polarity, and therefore has thousands of potential ligands including peptide-based antibiotics (Guyer et al. 1986). Later, the structure of OppA bound to diverse oligopeptide ligands solved by X-ray crystallography revealed a surprising complex mode by which the peptide ligands are enveloped between two lobes (I and III) of the protein and buried deeply in the protein interior (Fig. 1; Tame et al. 1994). The enclosure of ligands in this way is a hallmark of the periplasmic substrate-binding proteins and is normally associated with complementary interactions and exquisite specificity (Tame et al. 1996). However, the exception of the OppA, which possesses a broad specificity, implies a larruping interaction mechanism of this protein with its peptide ligands.

Fig. 1
figure 1

Stereoview of OppA protein–peptide KAK complex (PDB: 1jet). It is shown that the tripeptide ligand is enveloped between the lobes I and III of the protein and buried deeply in the protein interior

Full size image

According to the crystal structures of OppA–peptide complexes, the main chain of the bound peptides is in an extended conformation and forms parallel and antiparallel β-sheet interactions with the OppA that satisfy the hydrogen bonding capacity of the peptide backbones; the N-terminus of the peptides donates a salt bridge to the side chain of Asp419; the C-terminus of the tri-, tetra-, and pentapeptides forms buried salt bridges with the residues Arg413, His371, and Lys307, respectively; the side chains of the peptides are accommodated in voluminous hydrated pockets, where few direct contacts are made with the protein (Picon et al. 2000). This structure picture preliminarily unraveled the origin of the nonspecific nature of OppA binding to sequence-independent peptides: (1) the main chains of diverse peptide ligands share a common interacting mode with the protein receptor, and (2) the binding fluctuations over the capricious peptide side chains can be largely erased by the spacious, adaptable hydrated pockets.

Although the crystallographic information has shed light on the structural basis of OppA–peptide recognition, there still exist a number of problems that need to be addressed. In particular, the quantitative energetic knowledge regarding the free energy contributions of peptide chains and residue sites would be fundamentally valuable for understanding the thermodynamic behavior of this nonspecific binding, but it is completely unavailable from the crystal structures. Recently, Li et al. (2009) have proposed a protocol that incorporates empirical solvation model into hybrid quantum mechanical/molecular mechanical (QM/MM) methodology to dissect and analyze the energetic profile associated with the binding of peptide ligand to protein receptor. This method has been successfully applied to investigate the roles of water molecules and anchor residues in HLA-A*0201–epitope recognition. In this study, the more rigorous Poisson–Boltzmann/surface area (PB/SA) strategy is used instead of empirical solvation model to accurately account for the indirect desolvation effect accompanied with the OppA–peptide binding, and the direct nonbonded interactions involved in this binding are described using the QM/MM scheme. In this way, the energy contributions of different structure moieties of the tripeptide ligands contained in 28 structure-known OppA–peptide complexes are decomposed and examined in detail. The decomposed energy terms are further correlated with the experimentally measured affinities using a structure-based quantitative structure–activity relationship (SB-QSAR) approach. On the basis of these calculations, we would give a systematic discussion around the thermodynamic nature of the broad specificity of OppA–peptide binding.

Materials and methods

Structures and affinities of OppA–peptide complexes

The crystal structures of OppA protein in complex with 28 tripeptides, which share a consensus motif Lys-Xxx-Lys comprising a variable central residue Xxx flanked by two lysines, were solved at the atomic resolution level by Tame et al. (Tame et al. 1994, 1995, 1996; Sleigh et al. 1997, 1999; Davies et al. 1999) The central residues Xxx of the 28 tripeptides cover the complete spectrum of 20 coded and 8 non-coded amino acids. The thermodynamic data associated with the formation of these complexes were measured at a standard condition (50 mM sodium phosphate buffer at pH 7.0 and 298 K) using the isothermal titration calorimetry (Sleigh et al. 1999; Davies et al. 1999). Here, the tripeptide sequences, PDB entries, binding free energies ∆G expl, and logarithmic dissociation constants lnK D of the 28 OppA–peptide complexes are collected in Table 1, which will be used as a basic data set for subsequent analyses.

Table 1 The relevant information and experimental thermodynamic data for the 28 OppA–tripeptide complexes
Full size table

Pretreatment of OppA–peptide complex structures

The crystal structures of the 28 OppA–tripeptide complexes were retrieved from the protein data bank (PDB) (Berman et al. 2000), from which the water molecules, metal ions, and other cofactors were removed manually and the hydrogen atoms were added using the programs REDUCE (for coded amino acids) (Word et al. 1999a) and I-INTERPRET (for non-coded amino acids) (Zhao et al. 2007). Subsequently, each complex structure was subjected to an energy minimization procedure with the AMBER03 force field (Duan et al. 2003). Briefly, the maximum number of minimization steps was set to 2,000; the first 300 steps were performed with the steepest descent algorithm and the rest of the steps were performed with the conjugate gradient algorithm. The solvent effect was described using the generalized Born model. The convergence criterion for the rms of the Cartesian elements of the energy gradient was set to 0.2 kcal/mol/Å, as recommended by Hou et al. (2008).

QM/MM analysis of OppA–peptide interactions

At present, one of the most widely applied strategies to implement hybrid QM/MM calculations is the ONIOM method (our own N-layered integrated molecular orbital and molecular mechanics) (Svensson et al. 1996), which enables different levels of theory to be applied to different parts of a molecule/system and combined to produce a consistent energy expression. The objective is to perform a high-level calculation on just a small part of the system and include the effects of the remainder at lower levels of theory, with the end result being of similar accuracy to a high-level calculation on the full system. Previously, the ONIOM-based QM/MM scheme has been broadly used to investigate the interaction profile of proteins with small molecules, including binding activity and nonbonding behavior (Alzate-Morales et al. 2009; Lu et al. 2009; Zhou et al. 2010a, b). However, for the larger ligands such as peptides and oligomers, high-level calculation is also difficult even when adopting the non-electron correlation method in conjunction with the smallest basis set. Therefore, Li et al. recently suggested the use of semiempirical AM1 method—which has been demonstrated to be effective and efficient in the analysis of molecular systems involving weak nonbonded interactions (Dannenberg 1997)—to treat the high-level layer which, in this study, contained the tripeptide ligand and several key protein residues that were experimentally characterized to be fundamental in determining the binding of the OppA–peptide complex (Table 2; Fig. 2), while the remainder of this complex system was included in the MM layer and treated with a lower level of AMBER96 force field (Cornell et al. 1995), in which the partial atomic charges and force field parameters for the non-coded amino acid atoms were pre-obtained using restricted electrostatic potential (RESP) fitting (Comell et al. 1993) and generalized amber force field (GAFF) (Wang et al. 2004), respectively. Here, the hybrid ONIOM method was implemented in the Gaussian03 suite of programs (Frisch et al. 2003), and the electrostatic interactions between the two layers were treated in terms of mechanical embedding scheme to save computational cost.

Table 2 The key protein residues in direct interaction with the three residue sites of the tripeptide ligand through salt bridges and hydrogen bonds, as characterized by Tame et al. (1994, 1995, 1996)
Full size table
Fig. 2
figure 2

QM/MM partition scheme for the OppA–peptide complex. The tripeptide ligand (shown in stick style) and the key protein residues (shown in stick and ball style) that directly interact with the ligand are in the QM layer, while the remainder (shown in ribbon style) of this system is in the MM layer. The salt bridges and hydrogen bonds are represented by solid and broken lines, respectively

Full size image

The protein–peptide (or peptide segment) interaction energy (∆E int) in a complex system was predicted according to the strategy proposed by Zhou et al. (2009a). This was accomplished by performing single point energy calculation twice, one on the complex system (E 1) and other on the same system but its members (protein and peptide) were separated by several angstroms of distance each other (E 2). In this way, the interaction energy can be derived as ∆E int = E 1 − E 2.

PB/SA analysis of accompanied desolvation effects

Protein–ligand binding is a desolvation process, with which a substantial free energy change could be accompanied. To count the solvent effects associated with this process, the empirical solvent accessible surface area (SASA) model (Eisenberg and McLachlan 1986) and semi-empirical Poisson–Boltzmann/surface area (PB/SA) model (Kollman et al. 2000) are available. Previously, although Li et al. (2009) have successfully incorporated the SASA method into QM/MM to investigate the binding behavior of HLA-A*0201 protein with its peptide ligands and received satisfactory results for qualitative explanation, for the quantitative purpose as this study the relatively strict PB/SA method could be more reasonable choice.

In PB/SA procedure, the total desolvation free energy (∆G dslv) was estimated from the polar (electrostatic) desolvation energy (∆G polar) and the nonpolar desolvation energy (∆G nonpolar). The polar part was calculated by finite difference solutions to the nonlinear PB equation, as implemented in the DelPhi program (Rocchia et al. 2001), with ionic strength 0.15 M, and dielectric constants 4 for solute and 80 for solvent. A grid spacing of 2.0 grid/Å, in which the longest linear dimension of the solute occupied 60% of the lattice, was used to determine the size of the cubic lattice, and the boundary potentials were set to the sum of Debye–Hückel values (Yan et al. 2003). The PARSE parameter set (Sitkoff et al. 1994) was used for protein and peptide atoms, and a probe radius of 1.4 Å to define the dielectric boundary. The nonpolar contribution was determined by summing up the weighted surface area of whole solute molecule, i.e. ∆G nonpolar = γ·∆A, where γ = 0.00542 kcal/mol Å2 (Kollman et al. 2000), and ∆A is the changes in surface area due to the OppA–peptide binding. We computed the solute’s surface area using Sanner’s algorithm implemented in the MSMS program (Sanner et al. 1996), with a solvent probe radius of 1.4 Å and the PARSE atomic radii parameters (Sitkoff et al. 1994).

SB-QSAR correlation of the energy terms with binding affinities

The SB-QSAR methodology (Zhou et al. 2007) was further employed to ascertain the intrinsic relationship between the calculated energy terms and the experimental binding affinities of OppA–peptide complexes. The first 20 samples as listed in Table 1, of which the central residues of peptide ligands are defined by coded amino acids, were adopted as training set for constructing SB-QSAR models, while the remaining 8 samples with non-coded amino acids at the central positions of the peptides were used as test set for validating the constructed models. Partial least squares (PLS) regression (Wold et al. 2001), which has been broadly applied in the QSAR community, was employed to linearly correlate the energy terms with binding affinities, and we implemented this algorithm using the ZP-explore program (Zhou et al. 2009b). For a constructed SB-QSAR model, its performance in the statistical viewpoint, could be measured quantitatively by the coefficients of determination of fitting (r), leave-one-out (LOO) cross-validation (q), and prediction (r pred) on, respectively, training set, training set, and test set:

$$ r = \sqrt {{\frac{{\sum\nolimits_{i = 1}^{n} {\left( {y_{i} - \hat{y}_{i}^{\text{fitting}} } \right)^{2} } }}{{\sum\nolimits_{i = 1}^{n} {\left( {y_{i} - \bar{y}_{\text{tr}} } \right)^{2} } }}}} $$
(1)
$$ q = \sqrt {{\frac{{\sum\nolimits_{i = 1}^{n} {\left( {y_{i} - \hat{y}_{i}^{\text{cv}} } \right)^{2} } }}{{\sum\nolimits_{i = 1}^{n} {\left( {y_{i} - \bar{y}_{\text{tr}} } \right)^{2} } }}}} $$
(2)
$$ r_{\text{pred}} = \sqrt {{\frac{{\sum\nolimits_{i = 1}^{m} {\left( {y_{i} - \hat{y}_{i}^{\text{pred}} } \right)^{2} } }}{{\sum\nolimits_{i = 1}^{m} {\left( {y_{i} - \bar{y}_{\text{te}} } \right)^{2} } }}}} $$
(3)

where n = 20 and m = 8 are the sample numbers of, respectively, training and test sets, y i is the experimentally determined affinity of sample i, \( \bar{y}_{\text{tr}} \) and \( \bar{y}_{\text{te}} \) are the average values of all the y i in, respectively, training and test sets, \( \hat{y}_{i}^{\text{fitting}} \), \( \hat{y}_{i}^{\text{cv}} \), and \( \hat{y}_{i}^{\text{pred}} \) are the estimated affinities for sample i by, respectively, fitting, cross-validation, and prediction.

Results and discussion

Analysis of the different energy terms involved in OppA–peptide binding

The direct interaction energies ∆E int and indirect desolvation effects ∆G dslv for the binding of 28 OppA–peptide complexes were calculated according to the procedures described in “Materials and methods”, and the resulting values are compiled in Table 3. It is evident that there have been intensive nonbonded contacts between the OppA protein and its peptide ligands, given by the interaction energies ∆E int of most samples larger than −100 kcal/mol—which even achieve to the energy level of conventional covalent bonds. The strong interactions involved in these complexes are not unexpected, since OppA protein can form a number of hydrogen bonds and in particular salt bridges with the peptides, and these electrostatics-driven interactions could be quite effective at the low dielectric peptide-binding site deeply buried in the OppA protein interior where the electrostatic screening is considerably weak. Although the direct interaction of OppA with peptides is remarkable, a large part of this interaction energy are used to pay the unfavorable desolvation penalty due to the burial of polar and charged groups in the low dielectric protein interior. As can be seen, the desolvation free energies, ∆G dslv, are also noticeable and most of them have fallen into the range 60–100 kcal/mol, leading to a substantial discount for the favorable contribution of direct interactions to the binding of peptides to OppA.

Table 3 The calculated energy terms associated with the binding of the 28 OppA–tripeptide complexes
Full size table

The linearly fitting results, as shown in Fig. 3a and b, reveal that the correlations both between calculated interaction energy ∆E int and experimental binding affinity lnK D and between calculated desolvation energy ∆G dslv and lnK D are quite modest, with their correlation coefficients r of, respectively, 0.134 and 0.173, indicating that neither nonbonded interaction nor desolvation effect can independently dominate the binding. Nevertheless, the sum of ∆E int and ∆G dslv presents a good correlation with the lnK D, as suggested by a considerably high r value of 0.652 (Fig. 3c). The compatibility of lnK D with ∆E int plus ∆G dslv (∆E int + ∆G dslv), but not the individual ∆E int or ∆G dslv, implies that the force driving the formation of OppA–peptide complexes is a compromise between the direct interaction and indirect solvent effect. This conclusion is consistent with previous crystallographic studies, which figured out that, in addition to intensive salt bridge and hydrogen bond networks, a large number of water molecules were also observed at the OppA–peptide interfaces (Tame et al. 1996; Rostom et al. 2000). Moreover, it is worth noting here that, though the sum of ∆E int and ∆G dslv is correlated well with lnK D, which is deviated significantly from the experimentally determined binding energy ∆G expl. This phenomenon is interesting but not difficult to explain, if considering that some energy terms such as conformational entropy and conformation change effect were not considered in the present calculation procedure. However, ignoring these additional terms would not essentially influence our conclusions, since their contributions could be roughly regarded as a constant.

Fig. 3
figure 3

Plots of experimental binding affinity lnK D against calculated interaction energy ∆E int (a), desolvation energy ∆G dslv (b), and the sum of ∆E int and ∆G dslv (c)

Full size image

Furthermore, we herein give a preliminary discussion on the apparent relationship between the calculated energy terms and the physicochemical properties of amino acids, and the further detailed analysis of the energy contributions from each residue position of tripeptide ligands will be addressed in next section. From Table 3, it can be seen that weaker interactions ∆E int are commonly associated with the bigger nonpolar residues at the central position of the tripeptides. For example, the KLK, K-Hph-K, and K-Nap-K with hydrophobic and/or aromatic central residues possess the lowest ∆E int values as −76.54, −82.24, and −87.63 kcal/mol, respectively. In contrast, the peptides with charged central residues, such as KKK, KRK, and K-Dab-K, exhibit a stronger interaction with the OppA protein. Visual analysis of complex structure models unraveled that both nonpolar and polar central residues, especially those of bulk, are in, more or less, steric contact and/or collision with the neighboring OppA atoms, leading to a slight distortion on their side-chain rotamers. However, this unfavorable effect could be paid by long-range electrostatic attractions of the central residues with protein context, if the central residues are charged, or could be counteracted by favorable desolvation effects of the central residues buried in protein interior, if the central residues are hydrophobic. In addition, although the charged residues could keep effective interaction with OppA, they inevitably incur severe desolvation penalty. This can be rationalized by the relatively large desolvation free energy ∆G dslv generally associated with the peptides possessing polar central residues, albeit the ∆G dslv difference between polar and nonpolar is not significant as compared to ∆E int. Sum up, the more polar the residues are, the stronger/larger the electrostatic interaction/desolvation penalty is, and vice versa. In this way, the total binding energies of different peptides, no matter their central residues are polar or nonpolar, bulky or small, and charged or uncharged, are basically in a similar level, giving rise to the OppA protein capable of binding sequence-independent peptides.

Decomposition of total energy terms into individual residue sites

In order to examine in detail the energy contributions of the three residue sites in the tripeptides, we here broke down the peptide bonds (and capped by hydrogen atoms) of the peptide ligands in complexes and separately computed the interaction energy and desolvation effect of each residue binding to the OppA protein. The protocol used for calculating the energy components for individual residues was the same as that for whole peptides, and the resulting values are tabulated in Table 3. As can be seen, for most samples the sums of individual energy terms are not equal to the total energy values obtained from the whole peptides (ΔE int ≠ ΔE p1int  + ΔE p2int  + ΔE p3int and ΔG dslv ≠ ΔG p1dslv  + ΔG p2dslv  + ΔG p3dslv ). It could be attributed to the fact that the cooperation between the peptide residues was overlooked when the energy contributions were calculated using isolated residues. However, this discrepancy is not significant because the sums of individual terms are, albeit not equal to, roughly consistent with the total values.

Position 1

The N-terminal residues of all the 28 tripeptide ligands in our sample set are the charged lysine (Lys1). Crystal structures revealed that the backbone α-amino and side chain ε-amino of this residue can separately form a well-defined salt bridge with the Asp419 of OppA protein. In addition, a hydrogen bond was observed between the backbone amides of tripeptide Lys1 and OppA Cys417 (Fig. 2; Sleigh et al. 1997). According to the optimized complex models, these three nonbonded interactions of Lys1 with OppA are quite conserved and their configurations were basically unchanged during the MM minimization procedure, indicating that the energy contributions of Lys1 to the peptide binding in different complexes are significant but vary slightly. This could be rationalized by the relatively large average values (−55.62 and 43.56 kcal/mol) and considerably small standard deviations (7.14 and 7.01 kcal/mol) of the ΔE p1int and ΔG p1dslv over all samples. In this respect, we concluded here that, at least for the samples in our data set, the position 1 of tripeptide ligands performs the role as a stabilizer but not a specificator; it confers a large part of stabilization energy to the OppA–peptide binding but does not give rise to much judgment for different peptide ligands.

Position 3

Just like position 1, the position 3 of all sample peptides is occupied by the lysine residue (Lys3). Previous studies figured out that the conformations of Lys3 are shown to be somewhat more variable and disordered than that of Lys1, albeit this conformational fluctuation is insignificant (Tame et al. 1996). According to our calculations, the average values ± standard deviations of ∆E p3int and ∆G p3dslv are, respectively, −38.68 ± 8.77 and 40.60 ± 6.21 kcal/mol—they can be nearly completely counteracted by each other—manifesting that the Lys3 can only throw a slight effect on the binding. The less impact of Lys3 than Lys1 on the complex’s formation could be at least in part attributed to its absence of a salt bridge with the OppA as compared to Lys1 (Fig. 2). However, this difference is partly erased from the total energy values due to the additional desolvation penalty. Sum up, if considering the solvent effect, Lys3 dominates neither stabilization nor specificity for the peptides binding to OppA. Thus, it is more suitable to be viewed as an adaptor or filler in the complexes.

Position 2

The position 2 is completely variable in our sample panel, which presents 20 coded and 8 non-coded amino acids. Therefore, as might be anticipated, both direct interaction energies ∆E p2int and indirect desolvation effects ∆G p2dslv have a larger variance over all the samples than that at positions 1 and 3. From Fig. 4a it can be seen that, generally speaking, the polar and small, nonpolar amino acids exhibit stronger nonbonded interactions with context than those with bulky, nonpolar side chains, such as Ile, Leu, and Trp, which are electrostatically inactive but would introduce significant steric hindrance to the binding. For example, use of small probe technique (Word et al. 1999b) to detect atomic packing behavior in OppA–peptide KWK complex can apperceive intensive van der Waals contacts and collisions at the binding interface of OppA with the side chain of tryptophan residue in KWK (Fig. 5). This atomic overlapping, albeit is not severe, would more or less undermine the compatibility between receptor and ligand and hence results in a low ∆E p2int value—as predicted to be −2.54 kcal/mol. Furthermore, the indirect solvent effect, as shown in Fig. 4b, is readily accordant with the polarity of amino acids; there are significant desolvation penalties (∆G p2dslv  > 0) for polar and charged residues due to wrapping them into the low-dielectric OppA interior during the binding processes, whereas the desolvation effects for most nonpolar amino acids, especially those bulky nonpolar amino acids, are favorable, given by their ∆G p2dslv values < 0. Although ∆E p2int and ∆G p2dslv appear to be much variable, the fluctuations of ∆E p2int and ∆G p2dslv roughly present a complementary profile and, consequently, the sum binding energies at the position 2 (∆E p2int  + ∆G p2dslv ) vary insignificantly through the entire sample set (Fig. 4c). In this way, the apparent contribution of position 2 to the specificity of OppA–peptide binding is reduced considerably.

Fig. 4
figure 4

Direct interaction energies ∆E p2int (a), indirect desolvation effects ∆G p2dslv (b), and sum binding energies ∆E p2int  + ∆G p2dslv (c) of different amino acid types present at the second position of the tripeptide ligands

Full size image
Fig. 5
figure 5

Intensive steric contacts and collisions between the central residue Trp of tripeptide KWK and its neighboring OppA residues. The steric contacts and collisions were detected using PROBE program (Word et al. 1999b) and are shown in this picture as dots and spikes, respectively

Full size image

SB-QSAR modeling and analyses

SB-QSAR methodology was further employed to statistically model the linear relationship of energy values with affinities for this OppA–peptide system. The independent variable matrix X consists of the six decomposed energy terms, ∆E p1int , ∆G p1dslv , ∆E p2int , ∆G p2dslv , ∆E p3int , and ∆G p3dslv , and the dependent variable vector y is the experimentally determined binding affinities, lnK D. The results stemming from PLS modeling between the X and y were prettily good. As shown in Fig. 6a, the calculated lnK D agree well with corresponding experimental values. On the training set, the PLS model with four principal components (PCs) can explain a large ratio of variance for observed y, with coefficients of determination of fitting (r) and cross-validation (q) of 0.836 and 0.658, respectively; on the test set the performance of this model was also satisfactory with predictive r pred of 0.782. All of these definitely claimed that (1) the calculated energy contributions by the combination of QM/MM and PB/SA methods exhibit a strong linear correlation with the experimental binding affinities, and (2) the PLS model based on the first 20 tripeptides comprised of coded amino acids could be used to reliably predict other samples, no matter they are made up by coded or non-coded amino acids.

Fig. 6
figure 6

a Plot of PLS-calculated against experimental binding affinities, lnKD. b Variable importance in the projection (VIP) of PLS model

Full size image

The largest deviation of calculated from experimental lnK D value occurred on the test sample K-Dap-K, of which the central residue Dap possesses a small, charged side chain ( ). According to model’s prediction, this peptide should present an active binding behavior to OppA because of the low steric hindrance and strong electrostatic attraction. However, its experimental affinity was not as high as the predicted value owing to some unknown factors, leading to the overestimation by PLS model.

Furthermore, variable importance in the projection (VIP) (Wold et al. 2001) gives a straightforward insight into the relative importance of these six energy terms involved in the PLS model (Fig. 6b). As can be seen, indirect desolvation effect seems to be more decisive than direct nonbonded interaction in the OppA–peptide binding, which is compatible with crystallographic information because the strong nonbonded interactions are mainly formed on the invariant peptide’s backbones (Picon et al. 2000) and therefore provide only little judgment between the different peptide ligands. In addition, variation in desolvation effects, although which is relatively significant, is also largely discounted as the peptide’s side chains are accommodated in a spacious, plastic hydrated pocket (Tame et al. 1994, 1995, 1996) that can partially cancel the specific solvent effects arisen from the capricious side chains of central residues.

Conclusions

Exploring the complicated binding behavior of flexible biopolymer ligands to their cognate receptor is definitely a challenge but can provide much valuable information about thermodynamic properties associated with the recognition process. In this study, we attempt to understand the structural basis and energetic landscape of OppA protein interacting with its peptide ligands and to, at least in part, explain why the OppA protein can bind sequence-independent peptides? To achieve this, we employed a combination of ONIOM-based QM/MM computation and PB/SA solvent model to ascertain the binding mechanism of 28 structure/affinity-known OppA–tripeptide complexes. We demonstrate that: (1) in the tripeptides, the N-terminal residue confers significant stability but little specificity for the OppA–peptide binding; the C-terminal residue contributes neither stability nor specificity to the binding; and the variable central residue provides a large ratio of specificity for the binding, but which is considerably discounted if considering solvent effect. (2) Direct interactions are mainly formed on the invariant backbones of peptide ligands and thus contribute very little to the specificity, while the variation in desolvation effects due to the capricious side chains of central residues is erased largely by the voluminous, adaptable hydrated pocket. (3) The bulky central residues would incur intensive steric collisions from neighboring OppA atoms. This unfavorableness could be partly compensated from the favorable desolvation effect when the bulky residue is hydrophobic (nonpolar) or the favorable long-range electrostatic attraction when the bulky residue is charged (polar).