xMDFF: molecular dynamics flexible fitting of low-resolution X-ray structures

Abstract

X-ray crystallography remains the most dominant method for solving atomic structures. However, for relatively large systems, the availability of only medium-to-low-resolution diffraction data often limits the determination of all-atom details. A new molecular dynamics flexible fitting (MDFF)-based approach, xMDFF, for determining structures from such low-resolution crystallographic data is reported. xMDFF employs a real-space refinement scheme that flexibly fits atomic models into an iteratively updating electron-density map. It addresses significant large-scale deformations of the initial model to fit the low-resolution density, as tested with synthetic low-resolution maps of D-ribose-binding protein. xMDFF has been successfully applied to re-refine six low-resolution protein structures of varying sizes that had already been submitted to the Protein Data Bank. Finally, via systematic refinement of a series of data from 3.6 to 7 Å resolution, xMDFF refinements together with electrophysiology experiments were used to validate the first all-atom structure of the voltage-sensing protein Ci-VSP.

Keywords: molecular dynamics flexible fitting; xMDFF.

Figure 1

xMDFF re-refinement characteristics for six structures. The structures with diffraction data between 4 and 4.5 Å resolution had been deposited earlier in the Protein Data Bank and were used as initial models for xMDFF refinement. Improvements were evaluated using the R _free values of the xMDFF-refined structures (shown in red) against their respective starting values taken from the deposited structure (blue). In addition, the increase in the percentage of favored Ramachandran angles owing to xMDFF refinement and other structural statistics, as summarized by the overall MolProbity score, were used to measure the improvement in structural geometries. Further details of the refined structures are provided in Table 1 ▶.

Figure 2

Overview of the xMDFF workflow. Amplitudes F _obs from the X-ray diffraction data are combined with computed phases ϕ_calc from a tentative model to produce an electron-density map Φ(r). Biasing forces f ^fit derived from a Φ(r)-dependent potential are used to flexibly fit the phasing model into the map, yielding structures most consistent with the electron density. With the most recently fitted structure as a new phasing model, density maps are synthesized and flexibly fitted until a structure is found with sufficiently low R _free, providing the best possible interpretation of the diffraction data.

Figure 3

Test xMDFF refinement of the ‘closed’ conformation of d-ribose-binding protein using its ‘open’ conformation as the initial phasing model. Starting with an initial model of the open conformation (PDB entry 1urp; cyan) of the d-ribose-binding protein (a) and diffraction data from the target closed structure (PDB entry 2dri; orange) at resolutions ranging from 3.5 to 5 Å, xMDFF refinements were performed. (b) The xMDFF-refined closed structure (red) shows excellent agreement with the target at all resolutions (5 Å resolution shown with a final R _free of 0.34, down from an initial 0.56). Overall improvement was characterized using the R _free and cross-correlation of the final xMDFF structure and the r.m.s.d. of the final xMDFF structure to the target. Structural refinement is suggested by all three measures as well as the all-atom statistics provided in Table 1 ▶.

Figure 4

Refinement of the highly flexible region in the case of PDB entry 1xdv. Substantial density improvements are observed in a flexible region of 1xdv, illustrated by the difference in the density map between the initial (blue) (a) and xMDFF-refined final (red) (b) structures; local cross-correlations increase from 0.47 to 0.63, implying a more unambiguous placement of the atoms.

Figure 5

xMDFF refinement of the voltage-sensing protein Ci-VSP. (a) A MUFOLD-predicted homology model (cyan) was used as an initial phasing model in xMDFF; this model has an r.m.s.d. of 6 Å from an independently refined Ci-VSP structure (orange; Li et al., 2014 ▶). Ci-VSP includes a transmembrane (TM) domain and was crystallized with an antibody (Ab). Inset: the placement of the S4 helix within the TM region which determines the voltage-gating capabilities of the protein is of particular interest. (b) xMDFF refinement with 4 Å resolution diffraction data produced a final structure (red) 2.6 Å away from the independently refined Ci-VSP structure (orange). Inset: the placement of the S4 helix in the Down position and the alignment of the regularly placed voltage-sensing Arg residues is in agreement with the independently refined model (Li et al., 2014 ▶).