An automated data processing and analysis pipeline for transmembrane proteins in detergent solutions

D. Molodenskiy; H. Mertens; D. Svergun

The automatically generated models using the three modeling paths of the pipeline provide good fits to the experimental data at low angles, thus the overall particle shapes coincide and are well described by the data. The deviations observed at higher resolution ( Figure 7) reflect contributions from the detergents including the complex contrast situation, which is not accurately captured by the simplified modeling procedures used. This intentional simplification was justified for the sake of speed, such that preliminary models are generated that describe the gross structure of PDCs and do not attempt to describe internal inhomogeneties. One of the simplifications we implemented for the MEMPROT modeling was to assume a horizontal detergent torus with a circular geometry, and a fixed ellipticity e = 1. It should be stressed, that further model refinement is recommended in order to obtain more accurate and detailed models, using initial parameters suggested by the pipeline, and as necessary additional penalties as required (such as increased penalty for looseness and/or discontinuity, a higher number of experimental points and dummy beads in the MONSA grid, free ellipticity parameter for MEMPROT, etc). Similar results were obtained from the mechanosensitive channel T2 experimental data obtained at the P12 beamline (Figures 7b-8 ). There is a high degree of similarity between the three types of models generated, and interestingly, MONSA does reconstruct models with a void in the protein phase (green beads in Figure 8 ). This void is maintained also when an increased penalty for the bead discontinuity during minimization is introduced. Upon inspection, the high resolution model of a homologous structure, the MscS ion channel of T. tengcongensis (PDB: 3T9N) also displays a void in the extracellular domain of the structure. Thus the multiphase ab initio reconstruction reliably generates features consistent with those expected for the T2 ion channel. Figure. 8. Automatically reconstructed T2 models by the AMPP pipeline utilizing MEMPROT (hybrid modeling: 3T9N pdb model with found shape of detergent corona) (A), MONSA (multi-phase ab initio with P7 symmetry) (B) and DAMMIF (single phase ab initio, P7 symmetry) (C). The first row shows the side view cross-section of the PDC components that is represented with a section inside the detergent corona; protein beads are green for ab initio models and colored according to atom type for the hybrid model, detergent tails and heads are blue and red, respectively. The second and third rows represent the top and the side view of the models. The reconstructed MONSA model of the T2 channel protein fit the data reasonably well up to s = 0.25 Ã… -1 . The on-the-fly generated MEMPROT model fits only up to s = 0.09 Ã… -1 , which corresponds to the real-space resolution D = 2Ï€/s ~ 60 Ã….