We analyze several proteins interacting with alpha-helical peptides, some of them being known to bind also terphenyl and/or its derivatives. To characterize and compare their surface properties we examine the sequences and the three dimensional (3D) structures of the complexes formed by the protein and the bound peptide. The 3D structures are retrieved from the PDB 37 , the entry codes being presented in Table 1. Most of the structures are crystallographic. Two NMR structures are also used: the C-terminal domain of human centrin 2 in complex with the repeat sequence of human Sfi 1 and the human BCL-XL in complex with the BAK peptide.

*known to be disrupted by terphenyl or its derivatives.

Multiple sequences alignment (Figure 1) shows low sequence identity for the most of the analyzed proteins (shown in Table 2) both for the entire sequences and for the binding areas. The binding areas included all residues of the protein interacting with the alpha-helical peptide. Chicken, human, E. coli and rat calmodulin have very similar sequences (rat, chicken and human calmodulin are 100% identical; E coli has 98% identity with the others). For BCL-XL and human ubiquitin carboxyl-terminal hydrolase MDM2 only those fragments of sequences that are present in the 3D structures are considered. There is a high similarity only between the calmodulin, centrin 2 and troponin C sequences.

The binding area was defined here as all residues of the protein interacting with the helical peptide.

Figure 2 illustrates the complexes’ structures of six alpha-helix binding proteins. In all shown complexes, bulky hydrophobic residues of the bound peptide anchor into the protein binding pocket. Following the sequence similarities we superimposed the alpha-helix binding regions structures of calmodulin, human centrin 2, scherffelia dubia centrin and rabbit troponin C (Figure 3a). Strong structural homology for binding regions is seen following the sequence similarity of these proteins. Figure 3b and 3c illustrate the binding pockets of BCL-XL and human E3 ubiquitin-protein ligase MDM2, respectively.

The interacting residues of the proteins and bound peptides, identified with ContPro 29 , are shown in Figures 1 and 4 and Table 1. The results reveal that usually hydrophobic residues such as TRP, LEU, ILE, PHE, VAL, MET are involved in the interactions. The presence of hydrophobic residues suggests a favorable interaction with terphenyl-like molecules anchoring in the hydrophobic cavities. Most of the residues involved in the interactions between the proteins and alpha-helices are hydrophobic for both partners, as also observed in other studies 38 . We notice several key residues involved in the interaction of the same protein with different peptide partners. For example, in the case of calmodulin, PHE92, MET124, PHE141, MET144 and MET145 are involved in most of the peptides’ interactions. These residues can thus be considered as key for the interaction with terphenyl and its derivatives, or other alpha-helix mimetics. We noticed the presence of MET residues in most of the alpha-helix binding pockets analyzed here. In a recent study, MET residues have not been identified to be a part of hot spot amino acids, in particular in alpha-helix mediated protein interfaces 39 . However, our analysis clearly indicates their presence in positions that are key for the interaction with the alpha-helical partner. Furthermore, Ma and Nussinov 40 have also concluded that the amino acids TRP, MET, and PHE are important for protein-protein interactions. They showed that TRP/MET/PHE residues play roles in the dimerization of the transcriptase (p51/p66) and in cell-fusion processes, including the gp120-CD4 interaction and the gp41 six-helix bundle formation. They suggested that polarizability of MET allows it to assume roles of both hydrophobic and hydrophilic residues 40 . Further, its larger flexibility compared to other hydrophobic residues may facilitate the plasticity of hydrophobic binding pockets allowing to accommodate different ligands 27 .

We used Fpocket 33 and CASTp 32 to calculate geometrical and physicochemical characteristics of the binding pockets taking into account the protein residues interacting with the alpha-helical peptides. The overall hydrophobic character of the binding pockets is again clearly identified. Yet, some specificity is also observed, several pockets show high hydrophobicity score but low local hydrophobic density, or vice versa, demonstrating that the hydrophobic patches are not always regularly distributed in the binding pockets. For example, 1YCR and 3KF9 have similar hydrophobicity scores but high and low calculated hydrophobic density, respectively. The differences of the hydrophobicity distribution are illustrated in Figure 5.

The volumes of the detected pockets in the peptide-binding regions computed with CASTp are given in Table 3. The average volume of the sub-cavities present at the PPI interfaces found by Fuller et al 41 was ~60 Å ³ . Sonavane & Chakrabarti 42 found PPI pocket volumes to be up to ~330 Å ³ . We found similar volumes to those reported in Bourgeas et al. 43 . Taking into account the various algorithms and different concepts for binding pocket definition, such differences for the computed volumes can be expected. Several small cavities are present in the binding region (seen in Figure 2 and Figure 5), as it has been previously observed for other targeted PPI interfaces 39 . For the proteins studied here, the presence of several small hydrophobic cavities in the alpha-helix binding region seems to be a typical surface feature guiding the anchoring of hydrophobic residues from the peptide side. Such characteristics can also facilitate targeting PPI mediated by alpha-helices by small molecules containing hydrophobic anchors (as terphenyl or other mimetics).

Further, we decided to explore the roughness of the alpha-helix binding sites. The methodology implemented to calculate the fractal surface dimensions, used for the roughness evaluation, is illustrated in Figure 6 for the global surface roughness of chicken calmodulin. The fractal global surface dimension and the fractal local surface dimension for the binding site of chicken calmodulin are calculated to be D_S=2.238; ± 0.006 and D_L= 2.616 ± 0.072, respectively. The global and local fractal dimensions for the other proteins are given in Table 4. Our results and other previously published data 44 45 46 47 suggest that the global fractal dimension of protein surface is about 2. The local surface fractal dimensions for the binding cavities are computed to be larger than the global surface fractal dimensions for all studied proteins. This reflects the higher roughness of the binding site and its more complex shape and that can be considered as important for ligand binding. The most important differences between D_S and D_L are obtained for human calmodulin (2VAY), centrin (3KF9, 2K2I), BCL-XL (1BXL), MDM2 (1YCR) and troponin C (1A2X). It has been experimentally demonstrated that human calmodulin 18 , BCL-XL 19 20 and MDM2 21 22 interact with terphenyl or its derivatives. Recently, we suggested a possible binding of terphenyl 2, which mimics the relative positions of the side chains of residues TRP848, LEU851, LEU855 of the XPC peptide, into human centrin 2 following our energetic and conformational flexibility analysis performed for the alpha-helical peptide-binding pocket of centrin 2 27 . The D_L value for the peptide-binding site of troponin C shows rougher surface than the entire protein, similarly to the above listed terphenyl-binding proteins.

Taking into consideration the sequence and structural homology of troponin C and calmodulin and other physicochemical similarities of the binding sites as discussed above, we decided to probe putative terphenyl binding into troponin C. We performed docking of terphenyl 2 into the peptide-binding sites of calmodulin and troponin C using AutoDock. The best scored docking poses are shown in Figure 7. The terphenyl orientations in the best scored poses correspond to the position of the bound alpha-helical peptides shown in Figure 2. The predicted interaction energies of -7.98 and -8.18 kcal/mol for terphenyl binding in calmodulin and troponin C, respectively, suggest favorable interactions with the two proteins.

In the light of the results obtained here, it is now interesting to discuss the physicochemical properties of known PPI modulators, such as terphenyl. In a previous work 10 we gathered a set of 66 PPI inhibitors among which some terphenyl derivatives and other inhibitors of alpha-helix mediated PPI were present. In that work we demonstrated the more hydrophobic character of these compounds but also their bigger size. Interestingly, we also showed the importance of a critical number of aromatic bonds and some specific molecular shapes (T-shaped, star-shaped, or L-shaped compounds), among which some correspond to terphenyl derivatives. The present work therefore confirms that such genuine properties on the ligand side seem to be cavity-driven, and that these small molecules must possess certain properties in order to efficiently modulate an alpha-helix mediated PPI and to mimic the native partner and its properties.