2.1. Crystallization conditions

49 mg HEWL (3.2 mmol) was cocrystallized using the batch method with both 3 mg cisplatin (10 mmol) and 3.7 mg carboplatin (10 mmol) as well as 0.04 M sodium acetate buffer pH 4.7 (in 462.5 µl) and 10% NaCl precipitant (in 462.5 µl) with 1 mM DMSO (75 µl) and left at 277 K. An overall sixfold molar concentration of the Pt compounds over that of the protein was used.

2.2. X-ray data collection, structure solution and refinement

For the cryo data collection, the Bruker temperature-control device was set to 100 K and a crystal of 0.15 mm in size after 8 d of crystal growth was cryoprotected using silicone oil and centred on an in-house Bruker PLATINUM 135 detector at a distance of 50 mm from the detector. An X-ray exposure time of 10 s per 0.5° crystal rotation range was used with an X-ray wavelength of 1.5418 Å.

For the room-temperature (RT) data collection, the Bruker temperature-control device was set to 300 K and a crystal of 0.15 mm in size after 10 d of crystal growth was mounted in a 1 mm quartz capillary tube and centred on an in-house Bruker APEX II detector at a distance of 40 mm from the detector. Each data set consisted of a full 360° rotation of the crystal comprising 40 s X-ray exposure per 0.5° crystal rotation range for data sets 1 and 2 and 60 s per 0.5° crystal rotation range for data sets 3 and 4 with an X-ray wavelength of 1.5418 Å. scans were used for data collection in all cases (Tanley, Schreurs, Kroon-Batenburg & Helliwell, 2012).

The 100 K data set was processed using the PROTEUM2 internal software package on the Bruker PLATINUM 135 detector. The 300 K data sets were processed using the APEX II internal software package on the Bruker APEX II detector; all four data sets from the 300 K data collection were processed and integrated separately. The unit-cell parameters were checked for X-ray damage effects, which showed possible significant changes for RT data set 4 (see Supplementary Table S1). All crystal structures were solved using molecular replacement with Phaser (McCoy et al., 2007) and restrained refinement with TLS (except for the 100 K data set, which was solved with restrained refinement only and no TLS) in REFMAC5 (Vagin & Teplyakov, 2010) in CCP4 using the reported lysozyme structure 2w1y (Cianci et al., 2008) as a molecular search model. Model building, adjustment and refinement were carried out using the Coot (Emsley & Cowtan, 2004) molecular-graphics program and REFMAC5 (Vagin & Teplyakov, 2010), respectively, in CCP4. Crystallographic and refinement parameters are summarized in Table 1. Ligand-binding occupancies were calculated using SHELXTL (Sheldrick, 2008) and are given in Table 2.

Table 1 Crystallographic and refinement parameters of all data sets collected using the Bruker internal software programs Data-collection temperature (K) 100 300 300 300 300     Data set 1 Data set 2 Data set 3 Data set 4 PDB code 4gcb 4gcc 4gcd 4gce 4gcf Data reduction   Space group P43212 P43212 P43212 P43212 P43212   Unit-cell parameters (Å, °) a = b = 78.78, c = 36.98, = = = 90.0 a = b = 79.86, c = 37.80, = = = 90.0 a = b = 79.86, c = 37.80, = = = 90.0 a = b = 79.86, c = 37.80, = = = 90.0 a = b = 79.87, c = 37.76, = = = 90.0   Molecular mass (Da) 14700 14700 14700 14700 14700   Molecules per asymmetric unit 1 1 1 1 1   Diffractometer Bruker PLATINUM 135 Bruker APEX II Bruker APEX II Bruker APEX II Bruker APEX II   Crystal-to-detector distance (mm) 50.00 40.00 40.00 40.00 40.00   Observed reflections 1285547 130777 82393 73601 35009   No. of unique reflections 11301 8718 3318 2982 1753   Resolution (Å) 55.71-1.80 (1.85-1.80) 56.47-2.00 (2.05-2.00) 56.47-2.80 (2.87-2.80) 56.47-2.90 (2.97-2.90) 35.74-3.50 (3.60-3.50)   Completeness (%) 99.8 (99.0) 100 (100) 99.9 (100) 99.8 (100) 99.7 (99.2)   Rmerge (%) 0.069 (0.305) 0.202 (0.796) 0.195 (0.575) 0.230 (0.627) 0.278 (0.512)   Mean I/(I) 78.8 (14.27) 12.3 (1.42)# 17.58 (4.5) 14.7 (3.5) 12.6 (5.1)   Multiplicity 113.6 (44.2) 15.0 (6.9) 24.8 (24.8) 24.6. (25.1) 19.9 (20.3)   Cruickshank DPI (Å) 0.14 0.14 0.28 0.28 0.50   Average B factor (Å2) 13.2 26.3 34.7 36.9 50.4 Refinement   R factor/Rfree (%) 17.3/20.9 18.4/24.6 17.3/24.7 16.4/23.9 16.9/23.4   R factor, all 17.5 18.7 17.7 16.7 17.2   R.m.s.d. bonds (Å)/angles (°) 0.019/1.995 0.017/1.794 0.011/1.422 0.010/1.461 0.011/1.369 Ramachandran values (%)   Most favoured 96.8 94.5 96.1 92.9 97.6   Additional allowed 3.2 5.6 3.1 6.3 2.4   Disallowed 0 0 0.8+ 0.8+ 0 #I/(I) is 1.81 in the next resolution shell (2.15-2.05 Å) and is 2.45 in the resolution shell 2.24-2.15 Å. +The disallowed residue in the Ramachandran plot is Gly102, which is part of a loop region of the protein.

Table 2 The Pt-occupancy values (%) calculated from SHELXTL (Sheldrick, 2008) Based on Tanley, Schreurs, Kroon-Batenburg, Meredith et al. (2012), the sigmas on these occupancies are estimated to be 5%.   N binding site N binding site 100 K data collection 53 40 300 K data collection       Data set 1 65 64   Data set 2 64 64   Data set 3 67 56   Data set 4 71 59

3.1. Binding at His15

Cisplatin rather than carboplatin is observed bound to both the N and N atoms of the imidazole ring of His15 in both the 100 and 300 K data sets. Fig. 1 shows the details of this binding of cisplatin at 100 K (Fig. 1a) and for the first data set collected at 300 K (Fig. 1b). In the N binding site, clear anomalous difference electron density is observed for the Cl atoms in the 100 K structure, with each atom being distinct, whereas for the 300 K data set 1 the Cl atoms are less well defined, but some anomalous electron density is also observed for them here. However, the N binding site is more difficult to assign as either cisplatin or carboplatin. In the 100 K data set a distinct Cl atom is observed, but the anomalous difference density for this atom is weak. Data set 1 collected at 300 K shows less detail in the N binding site; hence, only a Pt ion can be assigned into the density. The other three data sets collected at 300 K were collected to lower resolutions (2.8, 2.9 and 3.5 Å); they also all show clear binding to both binding sites (Supplementary Fig. S2) supported by anomalous difference electron density for the Pt ion, but the data are of poorer quality owing to overall radiation damage. Hence, the data are too poor to provide detailed information as to the coordinated atoms beyond the Pt ion. Hence, it is difficult to assign whether cisplatin or carboplatin is present, but owing to the observation of cisplatin binding in the 100 K structure it is assumed that cisplatin rather than carboplatin is also bound in the 300 K structure. A whole cisplatin molecule was initially centred on the Pt position for both the N and N binding sites and its coordinated atoms were removed based on electron density not being observed. The Pt-His N distances were not restrained during refinement. The occupancy values of the Pt ions at each binding site in the 100 K and the four 300 K data sets are given in Table 2.

Figure 1 Cisplatin binding to the N and N atoms of His15. (a) 100 K data collection at 1.8 Å resolution, (b) 300 K data set 1 at 2.0 Å resolution. The 2Fo - Fc map (in purple) is at the 1.5 r.m.s. cutoff level and the anomalous difference density map (in orange) is at the 3 cutoff level. The N and N atoms in the imidazole ring are labelled, together with the Pt and Cl atoms of the bound cisplatin moiety. The Cl occupancies calculated from SHELXTL are 73 and 51% for Cl1 and Cl2 in the 100 K data set and 78 and 68% for Cl1 and Cl2 in the RT data set 1. Cl3 and Cl4 have occupancies of 25 and 36%, respectively, in the 100 K data set.

The data set at 100 K has an extra atom visible near the Arg14 residue (Fig. 1a). This is assumed to be likely to be a Cl atom. HEWL has the six usual chloride-binding sites and in this 100 K data set these have anomalous difference electron-density peak heights of between 5 and 11, with the new site having a peak-height value of 9.6. However, the 2F_o - F_c electron density for this Arg14 residue is weak and thus it is difficult to assign this density to the Arg14 side chain or to a molecule of water. Assuming the presence of an arginine, the B factors are between 24 and 34 Å².

3.2. Absorbed X-ray radiation dose for the 300 K crystal study

The X-ray diffraction data collected at 300 K were partitioned into four data sets, as described in §2.2. This data-collection strategy was used to confirm how stable the cisplatin/carboplatin molecule is when bound to a histidine residue in a model protein, as it is known that these drugs are radiation sensitizers and thus are used in conjunction with radiation therapy. This crystal was not `back-washed' before data collection; hence, the solvent channels of the protein crystal contained Pt ions. Taking into account explicitly the solvent content as well as the protein content, and the bound cisplatin and DMSO molecules to the protein, the absorbed X-ray radiation dose is 0.36 MGy for data sets 1 and 2 and 0.53 MGy for data sets 3 and 4. This thus gives an overall absorbed radiation dose of 1.78 MGy for this crystal. The dose calculation and equations used are given in Supplementary Table S3. Supplementary Fig. S3 shows the F_o - F_c electron-density maps after rigid-body refinement for all four data-set runs but with data sets 1, 2 and 3 cut at 3.5 Å resolution (the resolution limit of data set 4) and Supplementary Fig. S4 shows data sets 1 and 2 cut at 2.9 Å resolution (the resolution limit of data set 3). These figures confirmed that the electron-density detail for each data-set experiment is exactly the same, which was confirmed by the Pt-occupancy values (Table 2). The occupancies do differ slightly, but they are not greatly different and are in the range ±5% (Tanley, Schreurs, Kroon-Batenburg, Meredith et al., 2012); this is thus in agreement with the principle that cisplatin is a radiation sensitiser and is stable during repeated X-ray irradiation. Furthermore, we observed that its binding to histidine remained throughout this X-ray dose.

4.1. Cisplatin rather than carboplatin binding is preferred

A threefold molar excess of both cisplatin and carboplatin was used in the cocrystallization with HEWL, giving an overall sixfold molar excess of the Pt compounds over the protein. Both the 300 and 100 K X-ray diffraction data sets showed evidence for binding of cisplatin over carboplatin in the N binding site, as indicated by the presence of anomalous difference electron density for the Cl atoms bound to the Pt ion (Figs. 1a and 1b). Cisplatin also bound to the N binding site in the 100 K structure (Fig. 1a), as again indicated by the presence of anomalous difference electron density for a Cl atom. However, for the 300 K X-ray diffraction data sets it was difficult to assign which Pt compound bound to the N atom owing to the deteriorating quality of the data under repeated X-ray irradiation, meaning that only a Pt ion could be modelled into the density and no specific coordinated species could be assigned. However, owing to the fact that cisplatin rather than carboplatin was observed to be clearly bound in the 100 K structure, it can be deduced that cisplatin does have an overall higher binding affinity for the His side-chain N and N atoms compared with carboplatin in this model study. For the binding of two Pt ions to occur to the imidazole ring of this histidine side chain, both of the N atoms are sp²-hybridized with N lone pairs in the plane of the imidazole ring, providing two N atoms to which a metal centre can bind (Tanley, Schreurs, Kroon-Batenburg, Meredith et al., 2012). The usual N-hydrogen of one of the histidine N atoms has to be removed and this could be facilitated by the crystallization conditions used, which contained both chloride and acetate ions.

4.2. Absorbed X-ray radiation dose of the 300 K crystal

This study also looked at the stability of cisplatin towards continued X-ray irradiation once bound to the histidine side chain. Under continued X-ray irradiation in the 300 K study, the crystal absorbed 1.78 MGy of radiation (Supplementary Table S3) over the four X-ray diffraction data sets collected, and the electron-density maps for the bound cisplatin molecules looked very similar even at the differing resolutions of each data set (Supplementary Figs. S3 and S4). The results confirmed that cisplatin is stable over prolonged exposure to X-ray irradiation as it does not dissociate from the histidine residue, meaning that its relative binding affinity to this model protein is fixed. This result was backed up by the Pt-occupancy values for each 300 K X-ray diffraction data set (Table 2), which do not vary significantly (they are within ±5%) between each run. Hence, the relative binding affinity of cisplatin to HEWL, our model protein, has not been affected by the quantity of X-ray radiation absorbed.