Assignment-Free Determination of Ligand Binding Sites in Proteins by Solid-State NMR

Solid-state NMR studies of ligand binding sites in proteins traditionally require assignment of the observed resonances to the amino acid sequence. This sequential assignment is time-consuming and constitutes a major bottleneck in protein solid-state NMR. To determine ligand binding sites in proteins whose structures are already known, experimentally measured protein-ligand distances can be analyzed much more rapidly if sequential assignment can be bypassed. Here we present an assignment-free NMR approach for determining ligand binding sites in proteins. We measure 2D ¹³C-¹³C resolved ¹³C-¹⁹F rotational-echo double-resonance (REDOR) spectra that probe protein-ligand proximities and assign the peaks in the 2D spectra to residue types based on the well-known characteristic chemical shifts of amino acids. We simulate the measured REDOR dephasing using a second-moment approximation with an empirically calibrated scaling factor that accounts for experimental imperfections. This efficient REDOR simulation is combined with simulated annealing to rapidly search for ligand positions that agree with the type-assigned REDOR dephasing. We demonstrate this approach on the model protein GB1 and show that the position of a single fluorine can be determined accurately. We then apply this technique to the bacterial transporter EmrE and show that the location of a tetra-fluorinated ligand is within the range of positions found by this assignment-free REDOR approach. This technique should accelerate studies of ligand binding sites in membrane proteins, amyloids, and large protein complexes, and should also be applicable to dynamic nuclear polarization experiments at cryogenic temperature where broad lines may prohibit sequential assignment.