A comparison of protein models obtained by small-angle X-ray scattering and crystallography

Abstract

The technique of small‐angle scattering can be used advantageously for deriving the shape of proteins in solution. Presently, advanced ab initio approaches allow the reliable automated reconstruction of particle shapes. In particular, the procedures based on simulated annealing such as DAMMIN in combination with tight constraints turn out to be highly qualified for establishing realistic models. The appearance of the models, the comparison of experimental and calculated scattering profiles, and the agreement of both structural and hydrodynamic parameters can be used as evaluation criteria. In addition, the comparison of the SAXS‐based low‐resolution models with high‐resolution crystal structures allows the models under analysis to be checked. Superimposing the SAXS models and the crystallographic structures yields evidence for the validity of the SAXS‐derived models and far‐reaching agreement of the protein structure in the crystal and in solution, both in the case of small molecules and giant multisubunit proteins, and for proteins of quite different shape. Modelling of molecules requires application of a variety of sophisticated approaches and consideration of several precautions. These include the use of substitutes for amino‐acid residues missing in the crystallographic databases, application of special aligning, superimposing, filtering and averaging procedures, drastic reduction of the enormous number of beads to be used for modelling large molecules from crystal data, and the application of hydration contributions. Reconstructions of low‐resolution SAXS‐based shapes and comparisons with high‐resolution X‐ray structures were performed with glycogen phosphorylase, citrate synthase and an annelid haemoglobin. In the latter case, both the dodecameric subunit and the giant 3.5 MDa complex were studied, together with reconstructions of hypothetical complexes from building blocks.

Modeling Complex Biological Macromolecules: Reduction of Multibead Models

“…In the past, we have modeled several proteins of different size and symmetry, tentatively applying different approaches for the required reduction steps [9,10]. Among the proteins tested, the application of the RM and/or CG approaches to monomeric proteins (malate synthase, cellobiose dehydrogenase), dimeric proteins (citrate synthase, phosphorylase), tetrameric proteins (glyceraldehydes-3-phosphate dehydrogenase, lactate dehydrogenase, catalase), and the dodecameric subunit of earthworm hemoglobin [9-11, 13, 15, 17, 18] can be mentioned; HG approaches have been applied advantageously in the case of proteins of hexagonal symmetry such as the giant, 180-meric annelid hemoglobins [11][12][13][14]17].…”

“…Hydrodynamic parameters (s, D, [η]) were obtained by modified versions of García de la Torre's program HYDRO [7,8], applying also the most recent version HYDRO++ [24]. Further details concerning the prediction of structural and hydrodynamic parameters can be found in previous papers [6,17].…”

“…While in the past, the comparison of calculated and experimental data of quite different proteins has established the universal validity of the approaches applied (e.g., [6,[9][10][11][12][13][14][15][16][17][18]), a systematic investigation of the reduction approaches including inherent limitations is missing to date. To test the accuracy of the reduction procedures applied, only the structures of models have to be considered to avoid being influenced by a serendipitous coincidence of predicted and observed quantities.…”

Emerging applications of small angle solution scattering in structural biology

“…Due to the inherent ambiguity of SAS‐based shape modeling, validations of the model using additional experimental data that are “not seen by the model” are endorsed . Comparisons of the SAS‐derived shape models with corresponding known structures indicated excellent agreements for a number of cases . Recently, methods were developed to model RNA structures using experimental SAXS data in conjunction with the structure prediction algorithms …”