Re-refinement of 4g4a: room-temperature X-ray diffraction study of cisplatin and its binding to His15 of HEWL after 14 months chemical exposure in the presence of DMSO

An addendum is published to Tanley et al. [(2012), Acta Cryst. F68, 1300–1306].

Following our response (Tanley et al 2015) to the article of Shabalin et al. (2015), which reported a rerefinement of our structure of hen egg lysozyme in complex with cis-platin (pdb entry 4g4a, Tanley et al 2012a)), we have corrected the solute molecules that were highlighted. We have also reprocessed our raw diffraction data images, which we had made publicly available (Tanley et al 2013), and extended the resolution from 2.4Å to 1.7 Å (Table 1) using EVAL (Schreurs et al 2010). The omit electron density maps confirm the binding of platinum to both the His15 ND and NE sides of the imidazole ring as also modelled by Shabalin et al (2015) at 2.0Å (PDB code 4yen) and our original paper Tanley et al 2012a and its associated PDB file 4g4a.
At this improved diffraction resolution of 1.7Å for the His15 ND side of the imidazole ring the omit electron density map showed three peaks at the expected geometric positions for the platinum ligands forming a square planar coordination arrangement. The omit map peak heights were 12.2σ, 9.8σ and 8.0σ and from which the assignment of chloride and two ammines respectively were made. The His15 NE position cisplatin binding site omit map showed one clear ligand and, taking account of its peak height (adjusted for the Pt NE occupancy) was assigned to a chloride; another of this platinum atom's ligands is the nitrogen of the Arg14 side chain and, in a later step, involving an omit map, we assigned an ammine. PHENIX_REFINE (Afonine et al 2012) was used for model refinement. Table 2 gives the cisplatin ligand details to the His15 ND and NE sides. The final refined model statistics are given in Table 1. We used the paired refinement method of assessment of the diffraction resolution of Diederichs and Karplus (2013) both manually and also as implemented automatically in Joosten et al (2014).
These calculations indicated an improvement of the 1.7Å model compared to the 2.4 Å model.
The 1.7Å model showed a better agreement against the 2.0Å diffraction data with initial and final Rwork and Rfree values respectively 17.5% vs 18.2% and 20.2% vs 20.4%. We also note that by adding data between 2.0 and 1.7 Å resolution, the number of reflections increased by 26150. Of these 7817 had an I/σ > 1, and 2317 had an I/σ > 2. So, significant new information has been included by adding these data beyond 2.0Å.
For this 1.7Å crystal structure we assessed the stability of the B factor value estimates in Table   2 of our final re-refined structure by refining it in Refmac (Murshudov et al 1997) and the results are displayed in Table 3a. The B value estimates are similar in the two refinement programs. Likewise to assess the stability of the 4yen assigned ligands' B factors we refined 4yen against our new 1.7Å dataset. The results are shown in Table 3b and shows that three of the chloride assignments of the cisplatin, two on the ND side and one on the NE side, have large B factors at both resolutions (the original 2.0Å and now also at 1.7Å resolution) indicative of a ligand with a smaller number of electrons being present, and to which we have instead assigned an ammine at each of those three 4yen chloride positions.
In Table 4, for clarity about what has finally changed in the coordination arrangement of the two platinum atoms of this new study versus the two previous studies, we compare the cisplatin ligand assignments in our 1.7Å crystal structure with 4yen (at 2.0Å) and with 4g4a (at 2.4Å) along with their occupancy values and B factor estimates. The cisplatin ligand occupancies used in this 1.7Å crystal structure are the same as in 4yen. The spread of B factor values in this 1.7Å crystal structure for the cisplatin ligands is now quite small, indicating that reasonable ligand assignments have now been made. We would note that the challenge of differentiating between ammine and chloride, respectively ten versus 17 electrons, is easier at the higher resolution.
Overall, this new model for the platinum coordination on both sides of the His15 imidazole ring at 1.7 Å showed two ammine ligands and one chlorine at the ND side and one chlorine, one ammine and the Arg 14 side chain at the NE side. These ligands refined well i.e. with a relatively smaller spread of B factors for each of the ligands when restrained to the same occupancy value as their respective platinum atoms, compared with the three chlorine atom ligand assignments in 4yen.
Our histidine binding studies of these platin compounds under a wide range of chemical conditions, including high salt and salt free, have revealed a tendency of cisplatin to transform and so a partial i.e. split occupancy of chlorine and ammine at each ligand position to the platinum cannot be ruled out at this diffraction resolution of 1.7Å.     Also the ammines were each approximated by a nitrogen atom alone. The 4g4a is included in this table to explicitly show that these two aspects have now been rectified.