Structural Analysis of a Genetically Encoded FRET Biosensor by SAXS and MD Simulations

Abstract

Inspired by the modular architecture of natural signaling proteins, ligand binding proteins are equipped with two fluorescent proteins (FPs) in order to obtain Förster resonance energy transfer (FRET)-based biosensors. Here, we investigated a glucose sensor where the donor and acceptor FPs were attached to a glucose binding protein using a variety of different linker sequences. For three resulting sensor constructs the corresponding glucose induced conformational changes were measured by small angle X-ray scattering (SAXS) and compared to recently published single molecule FRET results (Höfig et al., ACS Sensors, 2018). For one construct which exhibits a high change in energy transfer and a large change of the radius of gyration upon ligand binding, we performed coarse-grained molecular dynamics simulations for the ligand-free and the ligand-bound state. Our analysis indicates that a carefully designed attachment of the donor FP is crucial for the proper transfer of the glucose induced conformational change of the glucose binding protein into a well pronounced FRET signal change as measured in this sensor construct. Since the other FP (acceptor) does not experience such a glucose induced alteration, it becomes apparent that only one of the FPs needs to have a well-adjusted attachment to the glucose binding protein.

Keywords: coarse-grained molecular dynamics (MD); glucose sensor; green fluorescence protein (GFP); single-molecule FRET; small angle X-ray scattering (SAXS).

Figure 1

The three investigated sensor constructs exhibit structural differences with respect to how the FPs are attached to the glucose binding protein MglB: no linker between MglB and both FPs (no. 1, left), a flexible linker between MglB and the donor FP (no. 2, middle) and a flexible linker between MglB and the acceptor FP (no. 4, right). The corresponding ΔR-values were determined from ensemble Förster resonance energy transfer (FRET) measurements (FRET binding isotherms, see [18]). (a) Data from single molecule measurements (smFRET histograms) were taken from an earlier publication, where these constructs were analyzed in detail [18]. Based on the analysis of the measured small angle X-ray scattering (SAXS) data, the respective Kratky plots (b) and distance distribution functions (c) are presented in the two lower lines. For further details, refer to the Supplementary Materials, Figures S3.2 and S3.3, and Table S2.

Figure 2

The correlation between energy transfer change and change in R_G is shown for the three investigated sensor constructs. Although we do not expect a strict linear relationship between <E> and R_G, the solid lines do not represent fits, but should visualize a general trend.

Figure 3

(a) Schematic presentation of the sensor no. 2 sequence. The flexible linker (purple) and restriction sites (grey) are shown with their respective sequences. The donor FP (mTurquoise2, cyan) is inserted into MglB (green), so the first 11 residues of MglB are in the beginning of the sequence and the main part in the middle. The acceptor FP (Venus) is shown in yellow (for details see Supplementary Materials, Section S1). (b) The experimental SAXS data is shown in blue, the fit of the chosen starting structure (blue FPs in Figure 3c) in orange, the latter is only shown for the glucose-saturated structure. Structures from the simulations with respective minimal χ² values are shown in green. (c) Starting structures for the simulation of sensor no. 2 with MglB in the glucose-saturated state. All structures are aligned to MglB (grey). The different pairs of fluorescent proteins are depicted in orange (sensor 2-1, χ² = 3.0), blue (sensor 2-2, χ² = 3.11), green (sensor 2-3, χ² = 3.75) and red (sensor 2-4, χ² = 4.1). For details see Supplementary Materials, Section S2.6.

Figure 4

(a) Fits of the fluorescence anisotropy r(Δt) as a function of time Δt_SBM for the donor chromophore (as part of mTurquoise2) and for the acceptor chromophore (as part of Venus) attached to MglB (here in the glucose-bound state). The time Δt_SBM is given in ns in the SBM simulation, which does not directly correspond to a physical time scale. The model function for the “wobbling-in-a cone” model (solid red) was fitted to r(Δt) values obtained from the simulation of sensor no. 2. The respective rotational correlation times on a physical time scale are in a time regime of 190–290 ns (further details see Supplementary Materials, Section S2). (b) Relative distance change upon glucose binding between the donor FP and the center of the MglB (left) and between the acceptor FP and the MglB (right), respectively. For the distance calculation we used the center of mass of the respective chromophore of the FPs.