Skeleton brief of our procedure for predicting the binding poses and energies
To understanding the functional response of DNA-binding proteins, the procedure for our predicting the binding mode of a p53 protein with an arbitrary DNA sequences and estimating the binding energy in a molecular vibratory environment in vivo is developed. As shown in Fig. 1, our procedure is divided into two stages: (1) binding mode prediction and (2) energy calculation, which are described in detail below.
(1) Prediction the complex of p53 protein and each DNA sequence.
In the case of docking simulations of DNA for p53 protein, the number of conformations that must be performed is too enormous, in addition, it might be extrapolated that the value of docking score as an index could not be different based on the experimental facts reported by Y. Itoh10. Therefore, in order to predict how DNA-binding proteins recognize and bind to their target DNA, we imported the concept of not a docking simulation but a conformational homology (Fig. 1A). That is, using the previously reported crystal structures, we predict binding configurations by considering the positional relationships between the residue loci at the DNA-binding interface of the P53 protein and the binding motifs (5'- C(A/T)(T/A)G − 3'). In concrete terms, at first, from the PDB database, we extracted the structures where binding between the p53 protein and DNA sequence is observed, and superposed the conformations of the amino acids in P53 protein on the binding interface so that the root mean square deviation (RMSD) is smaller. Similarly, the binding motifs on the DNA structure parts were superposed as well. As results, we were able to shed light on the average binding between the p53 protein and the binding motif part. Based on the binding position of recognition motif in above average binding structure, the initial binding conformation is created by replacing it with the DNA of the target sequence. The binding structure predict here does not have to be a precise predicting structure. This is because the energy minimization will be performed with molecular mechanics (MM) for this predicted structure and the after thermodynamical behavior in biological environment will be sampled later with molecular dynamics simulations.
(2) Estimation the DNA binding energy to p53 protein.
As shown in Fig. 1B, in order to sample the molecular behavior of the created p53-DNA complex in biological environment (water phase and 310 K: ~37 °C), water molecules and counter ions (Na+) are placed around the predicted structure to mimic the biological environment. The purpose of placing counter ions is to neutralize the total charge across the MD space to be considered. After then, energy minimization of this complex structure is performed. It makes sampling stably the molecular behavior possible. Using the stabilized structure as the initial structure, molecular dynamics simulations are performed for thermodynamic sampling, which is carried out after an elevated temperature process from 0 K to 310 K and equilibration process for a sufficient amount of time. Using this molecular sampling, we calculate the binding energy, which is estimate from the energy difference between the complex formation state and the dissociation state. This makes it possible to obtain the binding energy as a distribution in this calculation process.
Validation Study using antitumor drug incorporating to DNA.
(1) Antitumor drug and test sequences used for verifications
To validate our process, we selected the curious compound, FTD, which possesses a highly effective antitumor potency11–13. The structural difference between FTD and normal thymidine is only a substituted group, where is methyl group or trifluoro methyl group (Fig. 2A). The combination of FTD and tipiracil hydrochloride has already been approved as a cancer treatment by various institutions including the FDA (US Food and Drug Administration)14. FTD has been reported to induce p53-dependent sustained arrest during G2 phase15. Therefore, there might be changes in the binding affinity of FTD-incorporated DNA to p53 protein. From our previous study, as shown in Fig. 2B, it is presumed that thymidine adjacent to adenine and guanine having a part of electron rich structure is likely to replace FTD in DNA replication16. This is because the fluorine atoms at the trifluoromethyl group cause an attractive interaction with an electron-rich molecular orbital via a halogen bonding.
In fist validation case, normal type and FTD incorporating type of the BAX response elements’ sequences binding to p53 were used (Fig. 2C-D). In these figures about p53-binding sequences, some yellow halftone screening means the area of p53 recognition motif. At this validation case, the thymidine in BAX sequence surrounded by electron rich base was substituted to FTD as shown in Fig. 2D.
In second validation case, we used p53-binding sequences extracted from cancer cell lines that had become resistant to FTDs due to continuously exposed to it over a long period of time. As shown in Fig. 3A, for above sequence extraction, the resistant cells were disrupted and then captured using a consensus binding sequence for the p53 protein. The complementary strand was added based on the captured DNA sequence and is shown in Fig. 3B. Based on our previous theoretical predictions16, we validated the sequences with FTD substitutions at thymidine positions adjacent to electron-rich bases (Fig. 3C).
(2) Thermodynamical stabilization and binding energies
In accordance with the proposed method described above, the binding complex structures of p53 tetramer were predicted to each DNA, which is BAX sequence (Supplementary Conformational data 1), its sequence incorporating FTD (Supplementary Conformational data 2), p53 binding sequence from FTD resistant cell (Supplementary Conformational data 3), and its binding sequence incorporating FTD (Supplementary Conformational data 4). For these complexes, the conformational structures were sampled with molecular dynamics method in biological temperature (~ 37 °C = 310 K) after the equilibration for sufficient time. Figure 4A-B shows the RMSDs of the DNA positions calculated for these sampling structures after superposing the part of p53 protein structures, for BAX sequence and binding sequence from FTD resistant cell, respectively. The variation for a normal type of DNA and a FTD containing DNA are shown as blue line and orange line, respectively. In the BAX sequence, DNA incorporating FTD was found to have lower thermal oscillations than normal type DNA, while normal type DNA from FTD-resistant cells, in contrast, showed increased thermal stability compared to DNA incorporating DNA. Due to these differences in thermal stability, Fig. 5A-B shows the difference in the binding energy distributions fitted to a Gaussian type function, in which the blue line and orange line show the distribution for a normal type of DNA and a DNA incorporating FTD, respectively. The pre-fit distributions and each p-values are shown in Figure S1, S2 as box plots. In this test cases the binding energies were calculated at the MM level with Amber Force Field, which might not be quantitative, but it does qualitatively represent the property. As the DNA recognition probability of p53 has been reported to be very low10, the binding energy difference caused by sequence differences is also expected to be very small. In fact, the increase in binding energy caused by FTD incorporated DNA was also shown to be slight. However, this means that FTD-incorporated DNA binds to DNA-binding proteins more strongly and is easier to recognize than normal DNA. On the other hand, in p53-binding sequences extracted from FTD-resistant cells, we found that the FTD-containing DNA greatly reduced its binding affinity compared to normal DNA. In other words, it could be suggested that the randomly substituted thymidine to FTD, in resistant cells, might have acquired resistance to the FTDs by entering a position that prevents them from binding to DNA-binding proteins. Using this procedure, the binding energy could be calculated for any mutation-containing p53 in the same way. Not only the case of p53, it also allows us to analyze in terms of the structural and energetical views.