Supporting information for Winn et al. (2002) Proc. Natl. Acad. Sci. USA 99 (8), 5361–5366. (10.1073/pnas.082522999)
Simulation Protocols
Set-Up of Simulations. The coordinates of the crystal structure (PDB ID code 1OXA) were checked with WHAT_CHECK (1). Polar hydrogen coordinates were generated with WHATIF by optimizing the hydrogen-bond network (2) (selected asparagine, glutamine, and histidine residues were flipped as appropriate). Nonpolar hydrogen coordinates were generated with the "protonate" module of the AMBER software (3). The net charge on P450eryF was -23e and this was neutralized by placing 23 sodium counter ions in energetically favorable positions [as determined by the GRID program (4, 5)]. The positions of symmetry-related water molecules were calculated with WHATIF. A further 244 water molecules were placed at positions with GRID interaction energies more negative than -10 KJ mol-1. Hydrogen atom positions for all these water molecules were generated by WHATIF. Simulations were performed with the SAPHYR solvation model (6) in which the first two shells of water molecules around the protein are explicitly represented and the surrounding bulk solvent is represented implicitly. For the creation of the water shell around the protein, a preequilibrated box of water was superimposed and all water molecules further than 5 Å from the protein surface were removed. Water molecules in the interior of the protein that were not observed in the crystal structure were also removed.
Ferric heme was parametrized as described in ref. 7. The partial charges on DEB were taken from ref. 8 and were simplified to obtain neutral charge groups comprising 3-6 atoms each.
Molecular Dynamics Simulations of P450eryf.
The system was subjected to 100 steps of steepest descent energy minimization and then heated in three steps, each of 5 ps duration, at 100, 200, and 300 K. A 2-ns simulation (MD-WILD) of the wild-type protein at 300 K was performed. The system was stable over the whole 2-ns period, although some regions, especially some loops on the surface of the protein, exhibited high mobility.Two simulations were performed for an R185M mutant. The starting structure for these simulations was a snapshot at 360 ps of the MD-WILD trajectory. R185 was mutated using WHATIF, and then the system was subjected to 100 steps of steepest descent energy minimization and heated in three steps, each of 5 ps duration, at 100, 200, and 300 K with velocity reassignment every 0.2 ps. The second of these simulations (MD-R185M2) was subjected to a further 5 ps of velocity reassignment at 300 K.
To investigate the solvation properties of R185, simulation MD-SOLV was performed with the substrate, DEB, removed and the active site filled with water molecules. The first solvation shell in the active site cavity was placed using GRID, and the rest of the cavity space was filled by water from a preequilibrated box. The influence of a sodium ion near E173 is of potential interest because its position might be occupied by R185. The sodium ion was placed in this position during the set-up procedure, described above, the position being considered highly favorable according to the GRID map. However, it only requires a rotamer change for the guanidinium of R185 to occupy the same position and it is possible this could be a biologically relevant event. Thus, a 920-ps simulation (MD-ROT) was performed with a solvent-filled active site, as described above, after R185 had been manually rotated through the F/G loop to interact with E173. For this simulation the sodium ion was also manually exchanged with a water molecule near D396.
Supplemental Observations and Comments
Other non bacterial P450s showing an R aligned at the position of R185 include the demethylases. The demethylase family members all have an arginine at this position, normally as the last of three consecutive arginines. However, because most demethylases (but not CYP51) are membrane binding, the three consecutive arginines may assist this binding.
In the crystal structure of CYP51 from Mycobacterium tuberculosis (12), there is an arginine (R194) that is located in an equivalent position to R185 in P450eryF. However, because the position of the long B/B¢ loop in P450eryF is occupied by the F/G loop in CYP51, R194 makes hydrogen bonds to the F/G loop and the G-helix, rather than the B/C loop. Its guanidinium moiety is exposed to solvent and, because of the different lengths and locations of the F/G and B/C loops in the two P450s, it is not clear whether it could undergo a conformational change to hydrogen bond to the B/C loop and take on a similar function role to that of R185 in P450eryF. Interestingly, R198, further down the G-helix in CYP51, does make hydrogen bonds to two backbone carbonyl groups on the B/C loop but is clearly solvent exposed and makes a salt link to D244 on the I-helix.
F193 of P450cam aligns with R185 in the sequence alignment in (9). Interestingly, the side chain of F193 often rotates away from the active site during ligand exit in REMD simulations of P450cam and is observed in an outward pointing orientation in some crystal structures with large ligands (13). Mutation of this residue to alanine also affects substrate binding (14). In P450BM-3, the analogous residue is F205, which lines the access channel, but simulations do not indicate important side-chain rotations.
Thus, from this analysis of P450s, there is little evidence that other P450s exploit an arginine-B/B´ loop interaction to control channel opening in the way P450eryF may do. This may reflect adaptation of the channel opening mechanism to substrate type and the fact that the macrocycles are rather unusual substrates for P450s.