Structure ensembles of free pKID and KID
PTMetaD-WTE method was employed to obtain the structure ensembles of free pKID and KID in solution. The quality of the simulated structure ensembles was evaluated by the prediction accuracy of secondary chemical shifts (δcs). As shown in Supplementary Fig. 1, the predicted chemical shifts are agree well with the NMR measurements18. The RMSE of Cα δcs between simulated and experimental results is 0.44 ppm for pKID and 0.47 ppm for KID. The RMSE of Hα chemical shift for pKID and KID are 0.08 ppm and 0.06 ppm, respectively. The RMSE values are close to the system errors of the chemical shift calculating tool28. The results indicate that the structure ensembles obtained in our work provide a reasonable description of the structural properties of the IDPs. Based on the reliable structure ensemble, we found that both free pKID and KID are mainly disordered in solution, however, some transient helical structures were observed on the αA (residues 120–129) and αB (residues 134–144). The helical propensity of αA and αB were given in Fig. 1a. The N-terminal of KID and pKID have higher helical propensity than those on C-terminus. The helicity of αA are about 50% in both pKID and KID. The average residue-helicity on αB region of pKID (18.9%) is slightly higher than that on KID (14.6%), which is consistent with the experimental observations (Supplementary Table 1)18.
Although the secondary structure compositions in pKID and KID are similar to each other, some obvious differences were observed on their tertiary structures. Based on the residue-residue contact maps (Fig. 1b), more interactions between two terminals (marked by black circle in Fig. 1b) were observed in pKID. The larger amount of residue contact probability indicates more compact structures formed after the phosphorylation. The hydrophobic residues (Leu128, Tyr134, Leu137, Leu138) around the pSer133 in pKID are more likely to form hydrophobic interactions than in the KID. The spatial closed hydrophobic residues would form a HRC. The contact number between the side-chain heavy atoms on the residues Leu128, Tyr134, Leu137, and Leu138 were calculated to quantitatively define the formation of HRC. The conformations with contact number larger than 15 are defined to be HRC structures. As can be seen in Fig. 1c, the HRC structures present on more than 40% conformations of free pKID. On the other side, almost no HRC formed on free KID as the probability of conformations with contact number larger than 8 is close to zero.
The binding free energy landscape of pKID-KIX
To characterize the binding and folding mechanism of pKID and KIX, the free energy landscapes (FEL) along the reaction coordinates were constructed. Figure 2 gives the free energy landscape as a function of the center-of-mass (COM) distances and the number of native contacts between pKID and KIX. Sugase et al.27 proposed that the folding and binding process of pKID-KIX can be described by four-site exchange model, i.e., the free state, the encounter complex, the folding intermediates, and the bound state. All the four stages were clearly present on our simulated FEL, i.e., the free state with the large distance and low contacts (the state marked by F), the encounter complex with distance close to 1.7 nm and the fraction of native contact Q close to 0 (marked by E), the intermediates with many native contacts formed (state I1 and I2) and the bound state with most of the native contact formed (state B). The structure properties of the important free energy minima are given in the following section.
The free state
In the free state, the pKID peptides are far away from KIX, the center of mass distances between pKID and KIX are larger than 25 Å. In the free-state conformations, neither native contacts nor non-native contacts are formed (Supplementary Fig. 2). The secondary structure composition of pKID in this state is similar to the apo-pKID, the helical contents on the αA region is 52.0% and on αB region is 10.2%.
The encounter complex state
Based on the 1H-15N HSQC spectrum and 15N R2 dispersion experiments27, Sugase et al. found ensemble of weak complex which fast exchanging with the free pKID, and these states were defined to be encounter complex. Similar to the experimental observation, the interactions between pKID and KIX in the encounter complex state determined by our simulations are dynamic. Multiple hotspot binding sites present on the KIX and pKID (Fig. 3a), including the hydrophobic interactions between residue Leu128 on pKID with Leu664 on KIX, residue Leu141 on pKID with Val635, Met639 and Leu652 on KIX. In addition, the interactions between the charged residues also contribute to the formation of the encounter complex, such as Arg135 on pKID and Glu648 on KIX, Asp140 on pKID, and Lys659 on KIX. It should be noticed that most of the interactions in the encounter complex are non-native contacts, which indicate the non-native interactions are major force to drive the formation of pKID-KIX encounter complex. On the other side, the phosphorylated serine (pSer133) would not direct contribute to the formation of encounter complex since there is no interactions between the pSer133 and the residues on KIX.
Intermediates and hidden intermediate
Two intermediates (I1 and I2) with similar native contact numbers but different COM distances were observed on the free energy landscape. Many residues of αB regions are correctly anchoring with the native-binding sites on the KIX in both I1 and I2 intermediates, for example, the hydrophobic interactions between the residues Tyr134, Ile137, Leu138, Leu141 on pKID and the residues TyrY650, Ala654, Ile657, Tyr658 on KIX. Besides, the salt bridges interactions between residues Asp140 on pKID and Lys606 on KIX may also contribute to the binding. The residues on αA region are more flexible and less contacts formed with KIX than residues in αB region in the intermediates. The contacts between αA region and KIX are different in I1 and I2, where Arg125 on pKID contact with Glu648 and His651 on KIX I1 (Fig. 3b), however, Arg124 mainly contact with Glu655 in intermediate I2 (Fig. 3c). The non-native contacts also contribute to the stabilization of the intermediates. However, more non-native interactions are formed in the intermediate I2, especially the residues on αA region of I2 (Fig. 3c).
Although the native contact numbers are similar for intermediates I1 and I2, the secondary structure compositions are different in the two intermediates. In I1, the C-terminal of αA are basically folded to helical structures (the helicity from 124 to 128 is close to 90%); the helicity of αB in I1 is slightly lower than αA (the helicity of these intermediates shown in Supplementary Fig. 3 and representative structures displayed in Fig. 2). This is consistent with the NMR characterization of the intermediate state27 in which residues 124–128 in αA nearly fully folded and the helicity of residues 133–138 and 141 in the αB region is only about 70%. However, the pKID mainly adopt disordered structure in state I2, only several residues in αB region have ~20% probability forming α-helix27. By analyzing the FEL along αA or αB helicity and native contact number (Supplementary Fig. 4), we found that the structures of I2 are located on the off-pathway regions of the FEL.
In addition, a free energy minimum between the encounter complex and the intermediates was characterized. We denote the state as hidden intermediate (state H) since it has never been uncovered before. Compared with the encounter complex, some native contacts initially formed in state H. The contacts formed in this state might be corresponding to the primary driven force for the binding. The contact analysis showed that the intermolecular contacts between the aromatic residue Tyr658 on KIX and the hydrophobic residues Leu128 and Ile137 on pKID formed in state H, the contact probabilities are 80.3% and 62.1%, respectively. These residues are important to the pKID-KIX binding, which was proved in the single-residue mutagenesis experiment, the Y658A mutation would completely abrogate the complex formation16, the L128A and I137A mutation would increase the dissociation constant (Kd) over tenfold and two orders of magnitude, respectively26.
Fully bound state and free energy barrier
There is a high energy barrier (state T) between the intermediate state and the final-binding state. The pKID further folding to helical structures in this state compared with state I1 and I2, especially for the αB region, which is almost fully folded in state T. The native contacts on the αB regions are almost fully formed, however, the contacts in the αA regions are partially formed with relatively low probability (Supplementary Fig. 2). The results mean that αA haven’t bound to the right position (binding sites) of KIX in the state T. It should be noticed that the native contacts between the phosphorylated serine on pKID and the residues Tyr658 and Lys662 on KIX are formed in this state, with the contact probability 66.8% and 64.2%, respectively, which might be the driving force for αA region binding to the right position on KIX and the fully folding of αA.
In the fully bound state, the residues on pKID bound to the native-binding sites of KIX. The αB region and the C-terminal of αA region (residue 128–132) completely folded, the helicity on these regions are close to 100%. The helicity of N-terminal of αA increase to 40%. The incomplete folding of N-terminus of αA was also demonstrated by Dahal et al.21, they found that the final bound complex of pKID-KIX is partially mobile, with αA loosely bound to the KIX based on the kinetic experiments.
The free energy landscape for KID-KIX binding
Unlike the pKID, the unphosphorylated KID is hard to form stable complex with KIX. The binding affinity of unphosphorylated KID to KIX is about 100 times lower than the pKID21. To describe the binding behaviors of KID and KIX, the FEL as a function of KID-KIX COM distances and the fraction of native contacts (Q) was given in Fig. 4. The Q values is corresponding to the native contact atoms in the structure of pKID-KIX complex. Compared with pKID, KID is hard to form stable complex with KIX since the high free energy in the high Q region. Interestingly, KID and KIX also form encounter complex and some intermediate states with very low native contacts. For the structures in the encounter complex state, the binding sites are variable and transient, which is similar to the state E in pKID binding to KIX. The contact map shows the interactions between KID and KIX in the intermediates were mainly happened in αB region (Supplementary Fig. 5), in which the hydrophobic interactions might dominant the interactions. In the binding process of pKID and KIX, two intermediates with native contacts number close to 80 were observed. However, it is hard to form the similar intermediates in KID-KIX system. Besides, KID is basically incompletely folded and unstructured in all the states interact with KIX. Our results indicate that KID is transiently contacted with KIX and difficult to form stable and ordered complex.
The hydrophobic residue cluster formed in the binding process of pKID
The mutagenesis and kinetic experiments demonstrated that the hydrophobic residues Leu128, Tyr134, Ile137, Leu138, and Leu141 are important in the pKID-KIX binding. Based on our simulations, we found the hydrophobic interactions related to these residues formed prior to the formation of pKID-KIX binding intermediates. On the other hand, the interactions between these residues are absent in the unphosphorylated KID and the binding and folding process of KID and KIX would not proceed after the encountering complex. The results demonstrate the interactions between the hydrophobic residues Leu128, Tyr134, Ile137, Leu138, and Leu141 play important roles in stabilizing and guiding pKID-KIX binding.
By analyzing the structure properties of conformations in the binding process, we found the special structure pattern, i.e., the HRC, also appeared in pKID. The HRC (the number of contact heavy atoms in the side-chain of Leu128, Tyr134, Ile137, and Leu138 are larger than 15) formation probability were projected on the pKID-KIX and KID-KIX binding FEL (Fig. 5a and b), respectively. It can be seen that the average probability of HRC in the unphosphorylated KID is lower than 10%, indicate that the HRC structures are basically absent in the KID. On the other side, the conformations of pKID prefer to form the HRC, especially the conformations in the hidden state H. As the state H plays important role in the binding process of pKID and KIX, the formation of HRC and its anchoring to the binding site on KIX might provide the initial force of correct binding. We inferred that the HRC amplifies the hydrophobic interaction ability of pKID and facilitate the pKID to search for favorable binding sites on KIX and finally fold to the bound state.