Memprot: a program to model the detergent corona around a membrane protein based on SEC-SAXS data

Abstract

The application of small-angle X-ray scattering (SAXS) to structural investigations of transmembrane proteins in detergent solution has been hampered by two main inherent hurdles. On the one hand, the formation of a detergent corona around the hydrophobic region of the protein strongly modifies the scattering curve of the protein. On the other hand, free micelles of detergent without a precisely known concentration coexist with the protein-detergent complex in solution, therefore adding an uncontrolled signal. To gain robust structural information on such systems from SAXS data, in previous work, advantage was taken of the online combination of size-exclusion chromatography (SEC) and SAXS, and the detergent corona around aquaporin-0, a membrane protein of known structure, could be modelled. A precise geometrical model of the corona, shaped as an elliptical torus, was determined. Here, in order to better understand the correlations between the corona model parameters and to discuss the uniqueness of the model, this work was revisited by analyzing systematic SAXS simulations over a wide range of parameters of the torus.

Keywords: Memprot; SEC–SAXS; membrane proteins; small-angle X-ray scattering.

Figure 1

Model of the complex between the full-atom 2b6p structure and its detergent corona optimized from SEC–SAXS experimental data. (a) Section within the transmembrane plane. (b) Section perpendicular to the transmembrane plane. (c) Overall view.

Figure 2

Algorithm of the Memprot program. The program essentially creates PDB files with the models made of the full-atom protein structure and the parameterized coarse-grained detergent corona, and CRYSOL is called to calculate the SAXS curves. An overall sorting on the χ value is performed to keep the best model.

Figure 3

Illustration of the χ-value sensitivity. The SEC–SAXS experimental curve of AQP0 in DDM is superposed with curves calculated using three different detergent torus parameters, all three of which result in low χ values. The regions with the most pronounced discrepancies are highlighted in the two insets. It clearly appears that the slight differences between the χ values correspond to statistically meaningful differences in the curves. The curve calculated from the bare AQP-0 is shown in light grey as additional information.

Figure 4

Contour plots of χ as a function of different torus parameter pairs while the remaining parameters are kept at their optimum value, as determined by Berthaud et al. (2012 ▶) (a = 30 Å, b = 35 Å, t = 5.5 Å, e = 1.12). See Fig. 1 ▶ for the definition of the parameters. No strong correlations between the fitting parameters appear to exist, except between the parameters a and b, which define the diameter and thickness of the corona, respectively. The dotted line in (a) is a guide for the eye showing the main direction of this correlation.

Figure 5

Mean number of detergent molecules as a function of the parameters a and b. The dotted line is a reproduction of the line in Fig. 4 ▶(a).

Figure 6

The results of multiple runs with variable given electron densities for the hydrophilic (ρ_heads) and the hydrophobic (ρ_tails) parts of the corona. The quantities ρ′_heads and ρ′_tails represent the final electron densities after taking into account the fitted value of the electronic contrast bias (α) by CRYSOL. Geometric parameters and electron densities are mean values for all models with χ within +5% difference from the global minimum, which have essentially nearly indistinguishable scattering curves. The standard deviation for each calculated quantity is indicated below the corresponding mean value. The orange shaded case indicates the worst agreement between the model and the experimental data. The grey-shaded data denote the most self-consistent results, in terms of agreement between the number of detergent molecules calculated either from the hydrophobic or the hydrophilic volumes of the model and in terms of the lowest average contrast bias.

Figure 7

Scattering curves corresponding to corona parameters a = 29.6 Å, b = 35.4 Å, t = 5.6 Å, e = 1.12, e′_heads = 0.512 e Å⁻³, e′_tails = 0.270 e Å⁻³) for the full (2b6p) and truncated (2b6o) structures of aquaporin-0. The respective χ values are 1.31 and 3.79. The curve corresponding to an artificial optimized corona using the truncated form of aquaporin-0 is also plotted. The associated χ value is 3.47, which is still much higher than that for the complex based on the actual 2b6p structure.