Computational protocol: Three-dimensional structures of two heavily N-glycosylated Aspergillus sp. family GH3 β-d-glucosidases

Similar protocols

Protocol publication

[…] Both proteins were crystallized using hanging-drop vapour diffusion. AfβG was crystallized at 13 mg ml−1, mixed in a 1:1 volume ratio with well solution consisting of 21% polyethylene glycol (PEG) 1500, 25% ethylene glycol and 0.1 M MIB, a PACT screen buffer (Molecular Dimensions; consisting of sodium malonate, imidazole and boric acid in a 2:3:3 molar ratio) at pH 5.0. A crystal was harvested into liquid nitrogen, without the need for additional cryoprotectant, using a nylon CryoLoop (Hampton Research). Data were collected to 1.95 Å spacing on beamline I24 at Diamond Light Source and were processed using MOSFLM (Leslie & Powell, 2007) and scaled with AIMLESS (Evans & Murshudov, 2013). The space group was P212121. The structure was solved using programs from the CCP4 suite (Winn et al., 2011). Molecular replacement was employed using MOLREP (Vagin & Teplyakov, 2010), with the structure of the Thermotoga neapolitana homologue (PDB entry 2x40; Pozzo et al., 2010) as a search model; structure solution was performed prior to the publication of the structure of the more closely related homologue. The structure was rebuilt using Coot (Emsley et al., 2010) interspersed with maximum-likelihood refinement using REFMAC (Murshudov et al., 2011). The refined model contained a dimer of the protein, which was the most favourable assembly as calculated by PISA (Krissinel & Henrick, 2007), in the asymmetric unit (residues 21–863 of chains A and B), with each subunit having nine glycosylation sites ranging from 1–11 residues in length (45 and 46 sugar monomers were associated with chains A and B, respectively), as well as 39 ethylene glycol molecules, seven imidazole molecules and 1527 water molecules. The final R and R free were 0.15 and 0.17, respectively.Crystals of AoβG were obtained from drops of protein at 20 mg ml−1 mixed in a 1:1 volume ratio with well solution consisting of 30% PEG 400, 0.2 M magnesium chloride, 0.1 M HEPES pH 7.5. A crystal was harvested as above, and data were collected on the ID14.2 beamline at the ESRF to 1.95 Å resolution and were integrated with MOSFLM and scaled with AIMLESS. The space group was either P212121 or P21212 (one of the crystallographic axes lay along the rotation axis and unambiguous space-group assignment was not possible from inspection of systematic absences). The structure was solved employing the refined structure of AfβG as the search model in space group P212121 and was refined following similar methods to those used for AfβG to an R and R free of 0.22 and 0.25, respectively. The final model contains two protein dimers (AB and CD, again as calculated by PISA) in the asymmetric unit (chain A consisting of residues 23–861, B of 23–860, C of 23–861 and D of 23–860), with the following numbers of glycosylation sites: chain A, ten sites (with 40 sugar monomers); chain B, nine sites (42 sugar monomers); chain C, ten sites (41 sugar monomers); chain D, nine sites (34 sugar monomers). There are a PEG molecule and a Cl− ion associated with each chain. There are two Mg2+ ions in chains A and C which are coordinated to a water molecule which forms hydrogen bonds to both Asp722 O and NAG1202 O7; two other Mg2+ ions interact with waters which are hydrogen-bonded to both Asp558 (in chains A and D) and NAG1601 O7 in symmetry-related molecules. In addition, there are a PEG molecule and a phosphate ion associated with chain A, and 2117 water molecules.Coordinates and X-ray data for both structures have been deposited in the PDB. Details of X-ray data-quality and structure-refinement statistics are given in Table 1. [...] One of the features of these fungal enzymes is the presence of extensive and interacting N-glycans (see below), which posed particular challenges for correct refinement. All sugars forming N- and O-glycans are expected to be in the lowest energy 4 C 1 chair conformation, with the exception of a couple of l-pyranoside rings, where the 1 C 4 conformation is often preferred. No sugars of the latter type are present in the β-d-glucosidase structures described here. However, when working with poorer than atomic resolution data, most model-building and refinement software do not use energy-minimization techniques, but rather include a set of geometric restraints that approximate the correct chemistry. These restraints define ideal values and their respective acceptable deviations for bond lengths, angles, planes, chiral volumes and torsions, with the first four being the only ones actively used by default in the existing versions of both REFMAC5 and Coot. While a chair conformation has neither bond length nor angle strain, there are in addition a number of higher energy conformations that also show minimal or no strain. Any refinement process that exclusively minimizes the deviations from ideal bond lengths and angles can lead to sugar models in such higher energy conformations after attempting to fit them to featureless or incomplete electron-density maps, as has recently been demonstrated (Agirre, Davies et al., 2015).In the present study, all NAG (N-acetyl-β-d-glucosamine), BMA (β-d-mannopyranose) and MAN (α-d-mannopyranose) sugar monomers were imported from dictionaries created with ACEDRG into Coot and showed the expected initial 4 C 1 conformation. ACEDRG was used because it has been reported (Paul Emsley, personal communication) to produce geometric targets for bond lengths and angles that approximate well the values expected by MOGUL (Bruno et al., 2004), which approximate real chemistry better than the classic Engh and Huber values (Engh & Huber, 1991) used to build the REFMAC5 monomer library (Vagin et al., 2004). Torsion restraints had to be activated in order to keep a 4 C 1 conformation, but at present ACEDRG produces generic torsion values corresponding to the different combinations of hybridizations along the restrained bond (e.g. 60° for sp 3–sp 3). For this reason, the generic torsion values were replaced by ones measured from the lowest energy conformer, which ACEDRG calculates using RDKIT (http://www.rdkit.org). Torsion restraints were activated in REFMAC5 using a keywords file containing lines beginning with ‘RESTR TORS INCLUDE RESI’ and ending in ‘NAG’, ‘BMA’ and ‘MAN’.Manual rebuilding was performed between refinement cycles using Coot with the same custom library file input as used for REFMAC5. Torsion-angle restraints were enabled in the Refinement and Regularization Parameters window.The stereochemistry and conformation of the sugars were checked between refinement and rebuilding cycles in order to ensure chemical correctness. The software Privateer (Agirre, Iglesias-Fernandez et al., 2015) was used to this effect, but was extended to generate linear glycan descriptions including those presented here (Tables 2 and 3) and to produce script files for generating an interactive list of detected issues that could be used in subsequent sessions of model rebuilding with Coot, loading maps (calculated from 2mF o − DF c and OMIT mF o − DF c coefficients) and activating torsion restraints automatically. Privateer is distributed by CCP4 starting from the v.6.5 release. […]

Pipeline specifications

Software tools iMosflm, CCP4, Molrep, Coot, REFMAC5, RDKit
Applications Small-angle scattering, Protein structure analysis
Organisms Aspergillus fumigatus, Hordeum vulgare
Chemicals Cellobiose, Glucose, Hydrogen, Mannose