Dynamic tuning of FRET in a green fluorescent protein biosensor

Abstract

Förster resonance energy transfer (FRET) between mutants of green fluorescent protein is widely used to monitor protein-protein interactions and as a readout mode in fluorescent biosensors. Despite the fundamental importance of distance and molecular angles of fluorophores to each other, structural details on fluorescent protein FRET have been missing. Here, we report the high-resolution x-ray structure of the fluorescent proteins mCerulean3 and cpVenus within the biosensor Twitch-2B, as they undergo FRET and characterize the dynamics of this biosensor with $B_0^2$ -dependent paramagnetic nuclear magnetic resonance at 900 MHz and 1.1 GHz. These structural data provide the unprecedented opportunity to calculate FRET from the x-ray structure and to compare it to experimental data in solution. We find that interdomain dynamics limits the FRET effect and show that a rigidification of the sensor further enhances FRET.

Fig. 1. Cartoon representation of the x-ray structure of Twitch-2B.

(A) The fluorophores of mCerulean3 and cpVenus^cd are depicted as stick structures in red. The calcium ions of the minimal TnC domain are shown as magenta spheres. The disordered linker connecting the original N and C termini of Venus is shown as a dashed line. (B) The distance between the centers of mass of the two fluorophores of Twitch-2B is depicted as a dashed line. (C) The transition dipole moments of the two fluorophores of Twitch-2B as obtained from (15) are depicted as arrows.

Fig. 2. Structural details of Twitch-2B.

(A) Polar interactions between residues of the minimal TnC domain, mCerulean3, and cpVenus^cd are depicted as dashed lines. The residues are shown as sticks. (B) Polar interactions mediated by residues (shown as sticks) from the linkers between mCerulean3 and the calcium-binding domain (in salmon), as well as interactions between cpVenus^cd and the calcium-binding domain (in magenta). (C) Hydrophobic interactions between residues (shown as sticks) from the linkers between mCerulean3 and the calcium-binding domain (in salmon), as well as the linker between cpVenus^cd and the calcium-binding domain (in magenta) with residues (in gray) from the core of the minimal TnC domain. (D and E) Close-up views of the region around N532 of Twitch-2B and of the N532F mutant of Twitch-2B (Twitch-6).

Fig. 3. Histograms of the paramagnetic RDC data.

Predictions of RDC of methyl groups in the two fluorescent protein domains and TnC using the x-ray structure (in purple). The alignment tensor induced by the two dysprosium ions bound to TnC results from the translation of the tensor derived from calmodulin (see Materials and Methods). Experimental RDCs from paramagnetic NMR of Twitch-2B (in green) and Twitch-6 (in magenta). The range is reduced by a factor of 10 and 5 for Twitch-2B and Twitch-6, respectively.

Fig. 4. Ensembles of Twitch proteins consistent with the measured RDCs and FRET efficiencies.

Ensembles for Twitch-2B (left) and Twitch-6 (right) contain six structures each and the largest deviating structures are shown in green and red. States that are not these extreme conformations are transparent.