Initially, the primary sequences of human and mouse Mcl1 and PAP were compared. The sequence comparison shows a high sequence identity (88.2%) between mouse and human Mcl1 (see
Figure 1a) and the hydrophobic residues present at the interface region of PAPs (see
Figure 1b) were highly conserved. Therefore, the selected mMcl1 and PAP sequences were used to build ten different mMcl1—PAP model complexes (see
Figure 2 and
Supplementary Figure S2) using the homology-modeling approach that generated 100 models for each complex. Subsequently, the interface regions of the mMcl1—PAP models were closely analyzed. The analysis shows that the highly conserved residues of the PAP models form a hydrophobic face of the amphipathic α-helices (see
Figure 2c). These hydrophobic face residues promote tight binding of PAP to the binding pocket of mMcl1. Therefore, the mMcl1—PAP complexes were used as starting coordinates for the MD simulations.
Previously, several studies have been carried out on different Bcl-2 family proteins using various MD simulation approaches. These investigations revealed (i) the driving force behind the intra-molecular conformational change [
58], (ii) the helix stability [
59,
60], (iii) crucial residues involved in heterodimerization [
61] (iv) crucial molecular properties responsible for complexity [
62,
63,
64] (v) hot-spot residues [
65] (vi) effects of Bim mutants [
66] and (vii) the inter-helical interactions across families [
67]. Based on this information, we attempted to explore the mMcl1—PAP complexes to extend our understanding of the molecular mechanism of heterodimerization by identifying the key features. To achieve this, an extensive sampling of MD simulations was carried out on ten different mMcl1—PAP model complexes in explicit solvent conditions. Subsequently, the MD simulation results were used to calculate the binding free energy (BFE), estimate the per-residue decomposition (PRD) and perform dynamic network analysis.
3.2. The Polar Contacts Estimation at the mMcl1—PAP Interfaces During MD Simulations
Studies demonstrate that the polar contacts formed between the conserved aspartic acid residue present in the PAPs, and the arginine and asparagine residues lining the CBG of Mcl1 maintains complex stability (
Figure 2d and
Supplementary Figure S4) [
12,
21]. Additional residues were also involved, and this strengthens the complex stability, via the hydrogen bonds formation at the interface region (
Supplementary Table S1). Subsequently, the total number of polar contacts at the interface region was estimated over the period of time from the MD trajectories corresponding to each mMcl1—PAP complexes (
Figure 4 and
Supplementary Figure S5).
From
Figure 4, it is observed that the mMcl1—Bak complex (most favorable experimental binding affinity value; K
D = 1.33 nM) exhibited highest number of polar contacts (~11 to 13) at the interface region during the simulation. For the peptide that exhibited relatively weak experimental values such as Puma and NoxaB (K
D = 2.62 and 14 nM), the total numbers of polar contacts were also found decreased (~8 to 10) in comparison with mBak peptide. On contrary, the mNoxaA, mBax, mHrk, and mBik peptides showed the low in experimental values (K
D = 36.9, 39.5, 44.8 and 658 nM). This effect was clearly observed in our result, i.e., the total numbers of polar contacts are significantly reduced (~lesser than 8). Overall, it is observed that the total numbers of polar contacts are higher for the complexes that have higher experimental values, while the number of polar contacts is gradually decreasing, as the experimental binding affinity values decreases.
3.3. Energy Contributions Responsible for mMcl1—PAPs Heterodimerization
The study conducted by Ku. B et al. demonstrated a range of binding affinity values for various PAPs binding to mMcl1 [
22]. All these binding affinity values are reported in nanomolar range. Among those the mBak and mBik peptides showed tight and weak binding affinity values targeting to mMcl1 protein, respectively, while the binding affinity value for mBad was not determined (
Table 1). These experimentally determined binding affinity values provide the advantage to rank the PAPs binding to mMcl1. Due to the fact that the role of Mcl1 is significant in apoptotic regulation, it is necessary to construct novel peptides or non-peptide organic small molecule inhibitors with the potential to downregulate its activity. An efficient approach to attain this requirement is by employing advanced computational programs that can readily predict the BFE values using three-dimensional coordinates.
There are several other methods that have been developed to calculate BFEs such as Free-Energy Perturbation (FEP) [
68], Replica-Exchange Free-Energy Perturbation (RE-FEP) [
69], thermodynamic integration (TI) [
70], and umbrella sampling (US) [
71]. These methods are computationally very expensive, they poorly converge and require extensive sampling to explore all intermediate states, while the MM(GB/PB)SA approach can estimate the BFE values relatively fast compared to the above mentioned methods, without sampling the intermediate states. Moreover, our method uses the implicit water model for calculating the free energies, which helps to avoid large fluctuation in the model system and makes it computationally efficient. In addition, our method can rigorously decompose the total BFEs into individual energy components.
The MMPBSA.py [
23] program available in Amber has the advantage to predict the BFE values for the protein and its binding partners using multiple snapshots obtained from MD simulations. Here, the relative BFE values were estimated for ten different modeled mMcl1—PAP complexes that share a common CBG. The BFE values were estimated using 500 snapshots obtained from the last 10 ns of the MD simulations with an even interval of 20 ps. In order to extensively sample the conformational space for the simulated complexes, three different GB models developed by Tsui and Case (igb = 1) [
37] and Onufriev et al. (igb = 2 and igb = 5) [
32,
38,
39] were employed here. It was observed that the predicted BFE values thus achieved were negative for all studied complexes (
Table 1).
Subsequently, these predicted values obtained by the MMGBSA approach were then compared with experimental binding affinity values (
Figure 5).
For igb = 1, the correlation graph (
Figure 5) displays the magnitude of the predicted values in sequential increase with respect to the experimental binding affinity values. Therefore, the
R2 value obtained for the igb = 1 model is as high as 0.92. In contrast, the correlation graph corresponding to igb = 2 and 5 models failed to display the sequential increase for the predicted BFE values with respect to the experimental values. Therefore, the
R2 values were significantly reduced to 0.89 & 0.78 for the igb = 2 and 5 models, respectively. The reason behind this significant decrease in
R2 values is because the igb = 2 and 5 models did not predict the appropriate BFE values (i.e., the energies fluctuated) for NoxaA and NoxaB—mMcl1 complexes (
Table 1) with respect to the experimental values. Moreover, the igb = 5 model did not predict the appropriate BFE value for the mMcl1—Bim complex as well. Instead, all the predicted BFE values obtained using the igb = 2 and 5 models exist within the range. Overall, (i) the predicted BFE values obtained using the igb = 1 model are high, in comparison with the igb = 2 and 5 models, and (ii) the snapshots used to estimate the BFE values exhibit a plausible assumption of the binding conformation for all the mMcl1—PAPs complexes.
Despite the fact that the MMGBSA approach accurately predicts the relative BFE values, it is also reported that the energy values obtained may not converge consistently. Therefore, it is suggested that multiple simulations are required to obtain a reliable BFE value [
43]. Here, to attain a good estimate of the BFE values for the mMcl1—PAP complexes, MD simulations were carried out with ten repeats, individually.
Collectively, the predicted BFE values obtained for the mMcl1—PAP complexes were in good agreement with the experimental values. Here, the correlation coefficient (
R2) method is used to (i) quantify the consistency of the predicted BFE values, (ii) rank the PAPs with respect to the experimental binding energy values, (iii) highlight the inappropriate energies that affected the
R2 values, (iv) highlight the GB model that showed higher
R2 value, and (v) select the GB model that can be considered for further analysis. Note that, the eventual roles of conformational changes are not considered in our BFE calculations. In order to obtain such conformational change, the long range enhanced simulation techniques, e.g., replica exchange molecular dynamics or metadynamics may be required. These simulation techniques are computationally very expensive; therefore, we mainly focused on relative BFE estimation using the MMGBSA method. It has been noticed earlier that the length of the MD simulation is not crucial in the MMGBSA analysis, and even much shorter simulation time than we have used can provide meaningful data [
72,
73].
3.4. Energy Contributions from the Individual Components Responsible for mMcl1—PAPs Hetero-Dimerization
In the current study the igb = 1 model exhibited higher correlation value (
R2 = 0.92). Therefore, the BFE values obtained using the igb = 1 model is further used to acquire the details on the individual energy components responsible for mMcl1—PAP complex formation, while the rest of the models were ignored. These details on energy contributions may be beneficial for novel peptide or non-peptide small molecule synthesis that contain the potential to specifically target Mcl1 to downregulate its activity. Subsequently, to understand the binding process of mMcl1—PAP complexes in detail, the total BFEs were fragmented into the individual energy components (
Table 2).
The result shows that the van der Waals (ΔGvdw), electrostatic (ΔGele) and the molecular mechanics (∆Ggas) components contributed higher favorable energies for the complex formation. Additionally, the (∆Gnon-polar) non-polar energy component also contributed favorable energy to the complex formation, but to a lower extent. In contrast, the polar (∆Gpol) and solvation (ΔGsol) energy component contributed unfavorable energies to the complex formation. Closer observation on all the energies shows that the electrostatic and polar energy contribution for mBax and mBmf complexes displayed significantly low energies in comparison with the other complexes. In summary, the major favorable contributions to the total BFE of the mMcl1—PAP complexes result mainly from the van der Waals (ΔGvdW), electrostatic interactions (ΔGele) and the molecular mechanics (∆Ggas) components, respectively.
3.5. Energy Contributions from PAPs Key Residues Responsible for Heterodimerization
In order to design a novel peptide or non-peptide small organic molecule inhibitor with the capacity to specifically target and downregulate the Mcl1 activity, it is necessary to gain knowledge of the energy contributions of each residue, especially the residues present at the interface region of Mcl1 and its binding partners. Consequently, per-reside decomposition (PRD) analysis was performed using the
decomp module available in Amber.
Figure 6 highlights the residue interaction network (RIN) for the mMcl1—PAP complex interface projected in three-dimensional space. This RIN graph clearly shows five conserved hydrophobic pockets (P1 to P5) of PAPs and their surrounding contacts from mMcl1. Thus, the residues present in the RIN graph were used as the representative contacts for the mMcl1—PAP complexes.
Subsequently, the BFE values were obtained for the mMcl1—PAP complex interface residues using 500 snapshots from the last 10 ns of the MD simulations with an even interval of 20 ps. The result obtained from the decomposition analysis might provide valuable insight for better understanding of the molecular basis of hetero-dimerization. Previous studies explained that the conserved residues of the amphipathic α–helical PAPs form a hydrophobic face, which has the capacity to exhibit tight binding with the CBG of Mcl1 [
16]. Accordingly, the BFEs of the conserved hydrophobic residues of PAPs were extracted from the total energies (
Table 3 &
Figure 7).
Among the highly conserved residues, the leucine present at the P2 position of PAPs contributed the highly favorable energies in comparison with other interface residues (
Table 3). The overall energy contributions are more favorable for all PAPs complexed with mMcl1. This highly favorable energy contribution by the leucine present at the P2 position suggests that this residue plays a crucial role for the mMcl1—PAP complex formation. Additionally, the RIN graph clearly demonstrates that the leucine present at the P2 position of PAP is surrounded with a maximum of four residues—F209, V234, T247 and L248—from mMcl1. These surrounding residues play a vital role in the tight binding. Furthermore, the previous studies demonstrated that the mutation induced at this conserved leucine residue disrupts the binding significantly [
7,
74,
75,
76].
Next, the PRD analysis shows that the residues present at the P4 position of PAPs exhibited relatively less favorable impact in comparison with the residue at the P2 position. The sequence comparison for PAPs (
Figure 1b) reveals that the residues present at the P4 position are all hydrophobic, but not conserved. However, the energy contributions from the PAP residues at the P4 position are more favorable when complexed with mMcl1 (
Table 3). It is important to mention that, among the hydrophobic residues present at the P4 position of the PAPs, the phenylalanine and leucine residues contributed a highly favorable energy. This favorable energy contribution helps to predict that the substitution of bulky aromatic residue such as phenylalanine or aliphatic residue such as leucine at the P4 position of PAPs might be an ideal choice for the mutation to improve the tight binding with mMcl1. This prediction needs further experimental validation. Furthermore, the RIN graph displays that the residues at the P4 position established contact with V197 and V201 from mMcl1 to maintain tight binding.
The isoleucine residue present at the P3 position of PAPs displays high conservation, but the BFE values did not exhibit high energies in comparison with P2 and the P4 positions. In contrast, relatively low energies were observed at this position (
Table 3). Due to inconsistency in the energy pattern, it is difficult to make any plausible assumption for the improvement of novel PAPs binding affinity. It is noted that isoleucine is replaced with leucine in P3 position of mHrk peptide that exhibited −4.08 kcal/mol. The RIN graph highlights that the residue at P3 position of PAPs interacts with two aromatic residues —H205 with imidazole side chain, and F209 with bulky indole side chain—of mMcl1.
The sequence comparison of PAPs revealed that the P1 position did not exhibit sequence conservation. Nevertheless, the P1 position of PAPs is fairly occupied by the combination of hydrophobic and polar residues. The average BFE values obtained for the hydrophobic and polar residues are −4.94 and −1.59 kcal/mol, respectively. From this it is clearly understood that the presence of a polar residue at the P1 position severely affects the BFE value approximately by half fold, in comparison with the other hydrophobic residues. Specifically, glutamate in mNoxaB is more polar and acidic in nature over threonine in mHrk, which clearly reflected the BFE values that contributed only a smaller energy than the other. In total, the energy values obtained from the PRD analysis at P1 position clearly show that a hydrophobic residue, particularly a residue that contains a bulky aromatic side chain was highly favored over the other residue types. Moreover, the RIN graph displayed that the residues at the P1 position of PAPs mostly established the tight contact only with M212 residue from mMcl1.
The sequence analysis at the P5 position pointed out that the residues present at this location are similar to the P1 position, i.e., the P5 position is also occupied by the combination of hydrophobic and polar residues. In addition, the aromatic residues were also observed at this position (P5) in comparison with residues present at the other positions (P1–P4). The PRD analysis performed on the residues located at the P5 position exhibited energies in huge variations, i.e., the energies range between −1.71 to −5.89 kcal/mol. In contrast, a general view on the spider plot displays the residues present in P5 position exhibited only a relatively lower impact in comparison with the residues involved at the other positions. Overall, it is difficult to make a plausible assumption due to the inconsistency in the energy pattern. Additionally, the RIN graph displayed that the residue at P5 position of PAPs interacts with the residue that contains bulky aromatic (F299 and F300) and guanidinium (R196) groups of mMcl1, respectively.
It is reported that in addition to the conserved hydrophobic residues the GD doublet present in the P3 + 1 position of PAPs also has a significant role in the structural stability and selectivity [
21]. Therefore, exploring the BFE contributions of these residues is also highly necessary. Accordingly, the BFE values of the GD doublet were extracted from the total energies.
Similarly, aspartate present at the P3 + 2 position of PAPs has a very important role. Several studies have demonstrated that the highly conserved aspartate present at the P3 + 2 position of PAPs constantly form hydrogen bond and salt bridge interactions with carboxamide side chain of asparagine and guanidinium side chain of arginine residues [
17,
21]. These residues are located at the edge of the CBG of Mcl1. Accordingly, the average BFE value for the aspartate residue is approximately −2.0 kcal/mol. The aspartate residue exposed to the surface of PAPs constantly exhibited a moderate impact for the complex formation.
Overall, the predicted BFE values obtained for all the hydrophobic residues (P1 to P5) present at the interface region clearly demonstrate that these residues in PAPs acts as the “key initiating factor” for the complex formation. Furthermore, the rest of the residues present in the peptides may provide collective support to the complex stability.
3.6. Energy Contributions from mMcl1 Key Residues Responsible for Heterodimerization
The study conducted by Ku. B et al. [
22] demonstrated that the PAPs establish tight contact at the CBG present at the surface of mMcl1 (
Figure 2c and
Figure 3), which plays the central role in the binding partner selectivity; however, the mechanism of selectivity remained unresolved. Therefore, exploring the BFE contributions for the residues involved in the CBG—V197, V201, H205, F209, M212, V234, N241, G243, R244, T247, L248, F299, and F300—of mMcl1 is highly necessary. Accordingly, only the BFEs for the selected residues were extracted from the total energies (
Table 4) and compared (
Figure 8).
Initially, an average value was calculated using the BFEs of the individual residues present at the CBG of mMcl1. Subsequently, the average values were used to quantify the binding pocket residues based on the binding strength.
Figure 9 highlights the distinctive segments on the residues involved in the CBG of mMcl1 constructed based on the energy contributions. Among the residues involved at the CBG of mMcl1, R244 contributed the maximum energy consistent in all PAPs, i.e., its average energy value is as high as −5 kcal/mol. These higher energy contributions suggest that the residue present at this specific position plays a predominant role for the binding partner selectivity. Similarly, the average BFE values obtained for an ensemble of residues—H205, N241, G243, T247, F299 and F300—displayed moderate energy contributions, i.e., approximately −2 kcal/mol, respectively.
This result shows that these residues play a significant role in complex formation. Furthermore, the average BFE values obtained for the group of residues—R196, V201 and L248—involved at the CBG exhibited modest energy values, i.e., somewhat more than −1 kcal/mol, respectively. This suggests that these residues are involved in additional support to the complex formation. Likewise, the average BFE values obtained for a couple of residues—V197 and F209—present at the CBG contributed only insufficient or weak energies, i.e., lower than –1 kcal/mol. This illustrates that these residues play a less important role in the complex formation.
Overall, the predicted BFE values exhibited by the interface residues demonstrate that the higher energy contributing residues may together be responsible for engaging a selective PAP to the CBG of mMcl1. Moreover, the weak and insufficiently energy contributing residues also provided additional support for the PAPs to remain anchored in the CBG of mMcl1.
3.7. Crucial Residues Involved in Allosteric Signal Transmission (AST) in mMcl1—PAPs Complexes
Despite, our MD simulations exhibited the favorable energy contributing residues at the CBG of mMcl1—PAP complexes, the mechanism of the collective internal motion of these residues involved AST remain unclear. This investigation might provide the molecular origin of the regulation, network of interaction across dimer interface, and putative allosteric path from one site to distal functional site [
77]. For this, the advanced post-processing method was applied to MD trajectory (
Figure 10). This method works based on (a) cross-correlations between residues and (b) coarse-grained community network analysis applied to all residues in a protein-peptide complex.
Despite Mcl1 plays a vital role in anti-apoptotic activity, not much is known about how the AST from one site (initiation) to the other site (distal) of the complex occurs. Here, the MD simulation corresponding to the mMcl1—PAP complex that exhibits a high binding affinity value (mMcl1—Bak complex) was selected as the representative. Subsequently, to gain more understanding on the potential AST, the dynamic network analysis was performed using the MD simulation.
3.9. Community Network Analysis to Investigate Collective Internal Dynamics of mMcl1—Bak Complex
Although cross-correlation analysis highlights crucial residues that are involved in mMcl1—Bak complex stability, the analysis was extended to elucidate AST pathway. As network analysis can provide the evidence for allosteric communication between protein–peptide complex [
78], the MD simulations of mMcl1—Bak complex was subjected to (i) all-residue dynamic network analysis and (ii) coarse-grained community network analysis (
Figure 10b). Therefore, all residues in mMcl1—Bak complex were considered for network construction. Here, each amino acid residue was assigned as a node and the pair of nodes that are connected between each other within a certain distance through edges. Subsequently, the entire network was partitioned into clusters of communities (
Figure 10c). The list of residues and the total number of the communities were given in the
Table 5. The analysis produced 17 communities based on consensus correlation matrix.
In order to obtain the long-range communication path, optimal and sub-optimal analysis was carried out. For this, the densely populated communities were alone considered (1, 10–17) while sparsely populated communities were ignored (2–9). The analysis produced ten sub-optimal pathways (
Figure 10c,d). Subsequently, the pathway that connects two remote nodes spanning across the binding interface of mMcl1—Bak complex was alone selected. The selected optimal pathway identified the nodes (I78:P3→L75:P2 (Bak)→V234→V232→V278→T276→Y156→D154 (mMcl1)) from three different communities (13, 14 and 16; ref
Table 5) demonstrating the long-range communication. This indicates that these residues are dynamically stable and have an important role in allosteric activity. Among the residues involved in the predicted pathway, L75, I78 from Bak and V234 from mMcl1 were located at the interface region. Moreover, these residues exhibited more favorable energy contributions from our PRD analysis. Interestingly, majority of the residues involved in the selected pathway are highly hydrophobic, except T276, Y156, and D154.
It is important to mention that until now the Bcl2 family protein structure, particularly Mcl1, was investigated in the carboxy terminal region, while the amino terminal region was ignored due to its intrinsically unstructured nature, comprising the PEST (proline, glutamic acid, serine and threonine) sequence that targets proteasomal degradation, sites specific to ubiquitination, caspase cleavage and most importantly multiple phosphorylation [
79]. As it is evident that the CBG of Mcl1 is located in the carboxy terminal region, where all the members of endogenous PAPs [
22], synthetic peptides with BH3 mimetic [
16], and non-peptide small molecules [
80] bind to elicit subsequent cascade reaction, we took advantage to explore its thermodynamic properties. Based on this fact, it is plausible that the predicted energies from our investigation would vary when the full-length mMcl1 would be considered. Therefore, further experimental data would be needed to validate our findings.
Prior to our simulation, it was not clear which (i) thermodynamic components and (ii) high energy contributing residues are responsible for the mMcl1—PAP hetero-dimerization. Here, we have predicted crucial structural characteristics required for novel peptide or non-peptide organic small molecules synthesis. The results from the current investigation may efficiently provide valuable insight for the novel anti-apoptotic inhibitor design, particularly to down-regulate the activity of Mcl1.