Computational protocol: Crystal structure of O-methyltransferase CalO6 from the calicheamicin biosynthetic pathway: a case of challenging structure determination at low resolution

Similar protocols

Protocol publication

[…] Crystals of CalO6 were grown by vapor diffusion in hanging drops made by mixing 1 μL of protein with 1 μL of crystallization buffer [0.1 M NaCl, 0.1 M Bis-Tris (pH 5.8, adjusted with HCl at room temperature), and 1.10-1.25 M ammonium sulfate], incubated over 1 mL of crystallization buffer. Rod shaped crystals of CalO6 (0.1 mm × 0.1 mm × 0.3 mm in size) grew in 3 days at 21 °C and were transferred to cryoprotectant solution [0.1 M NaCl, 0.1 M Bis-Tris pH 5.8 (adjusted with HCl at room temperature), 1.10-1.25 M ammonium sulfate, 15 % (v/v) glycerol] by increasing glycerol concentration in steps of 3 %, then incubated there for 40 min and quickly immersed into liquid nitrogen. SeMet-substituted CalO6 displayed much lower solubility (precipitated at ~3 mg/mL) than unsubstituted CalO6 due to a large number of Met residues. Crystals of SeMet CalO6 took 2-3 weeks to grow and did not diffract well enough (resolution >4 Å, streaky reflections) to be useful for phasing. Likewise, CalO6 did not form suitable crystals in the presence of 1-2 mM SAM, SAH, or 2-5 mM substrate N-acetylcysteamine orsellinic acid (SNAC-OSA; used with or without SAH), nor did we observe these ligands in the electron density map upon soaking them into crystals of CalO6 grown in their absence. These ligand concentrations were several-fold higher than the previously reported Km values (0.3 mM for SAM and 1.3 mM for SNAC-OSA) [], which ensured that most CalO6 is in a ligand-bound form at the conditions of the reported activity assays. However, we could not exclude a possibility that the ligand binding was disfavored in the crystallization solution. Ethyl mercury phosphate (EMP)-derivative crystals of CalO6 were prepared by soaking crystals of native CalO6 in the cryoprotectant solution containing 2 mM of EMP overnight prior to flash-freezing in liquid nitrogen. A number of other mercury, platinum, tantalum, and other reactive and inert heavy metal salts were tried, but did not yield useful derivatives.X-ray diffraction data were collected at beamline X-12 at the National Synchrotron Light Source at the Brookhaven National Laboratory and processed with HKL2000 []. The diffraction was highly anisotropic, with the useful data extending only to a modest-low resolution (Table ), making structure determination challenging. The anisotropy analysis by the anisotropy server [] indicated that the data were strongly anisotropic (the spread in values of the three principal components of scale factors is 33.62 Å2), with resolution limits of 3.6 Å, 3.6 Å, and 3.1 Å along three principal component axes. The crystals were not merohedrally twinned, as analyzed by using XTRIAGE [] program in PHENIX suite []. In addition, EMP derivative crystals were highly non-isomorphous with native CalO6 crystals. Three-wavelength data set was collected with the EMP derivative, but due to rapid crystal decay in the X-ray beam, only the data set collected at 1.007 Å was used for structure determination. The anomalous signal was measurable to 4.1 Å according to XTRIAGE output; with 25 % of strong (>3σ) intensities displaying strong anomalous signal (the magnitude of the Bijvoet intensity difference over 3σ) in the lowest resolution shell and 5 % of strong intensities with strong anomalous signal at resolution ~4.1 Å. Four mercury sites were found by SOLVE [], but the resulting electron density map quality was insufficient for model building.Molecular replacement (MR) was attempted by using the data collected with the native CalO6 crystal with all available crystal structures of different CalO6 homologues as search models, by using PHASER [] and MOLREP []. The only structure that yielded a molecular replacement (MR) solution was that of aclacinomycin-10-hydroxylase RdmB (PDB ID: 1QZZ) [], the only known non-methyltransferase in this structural family. Specifically, only the C-terminal domain of RdmB as a search model yielded an MR solution; neither searching with the full-length RdmB nor searching with the N-terminal domain of RdmB after placing the C-terminal domain were productive. The resulting electron density map was not of high enough quality for model building. However, the phase provided by the MR solution when used with the anomalous difference signal from the EMP derivative SAD data yielded 2 mercury sites in the anomalous difference Fourier map. With these mercury sites as an input, we used AUTOSOLVE [, ] in PHENIX package [], to combine the MR and the SAD phases to find 5 additional sites and yield an interpretable electron density map with the figure of merit of 0.46 after density modification (Fig. ). The electron density for the missing N-terminal domain was clearly discernible in the difference density map. The structure was then built by ~50 cycles of iterative model building with Coot [] and refinement with REFMAC [] by using the EMP derivative data set. Tight geometric restraints were used in REFMAC to prevent divergence and preserve proper bond geometry, which also kept R and Rfree values similar to each other. The resolution cut-off was chosen as 3.4 Å, as no map or statistic improvement was achieved upon including higher resolution data in the refinement, also consistent with the anisotropy analysis. Using the data extending to 3.1 Å in resolution after the ellipsoidal truncation by the anisotropy server did not lead to improvement either. Potential twinning in R3 space group to mimic apparent R32 was excluded based on the Britton plot analysis by XTRIAGE program []. Refinement of a model that contained two CalO6 monomers per asymmetric unit with the data reduced in R3 did not yield further improvement in map quality or refinement statistics. All mercury sites were located near sulfur atoms of the Cys residues of the refined structure, confirming the proper residue register. The data collection and refinement statistics are given in Table . Due to an apparently complex non-isomorphism and very high anisotropy, the native CalO6 crystal data did not improve the resolution or map quality, even after molecular replacement with individual domains as search models.Fig. 2 […]

Pipeline specifications

Software tools PHENIX, Molrep, Coot
Applications Small-angle scattering, Protein structure analysis
Chemicals Mercury, S-Adenosylmethionine, Hydroxyl Radical