Extended Data Fig. 2: Single-molecule imaging of LIV-BP^WT and LIV-BP^SS reveals kinetic determinants of analogue and digital responses.

a, b, Representative traces of LIV-BP^WT (a) and LIV-BP^SS (b) imaged at 100-ms time resolution in the absence of substrate. c, d, LIV-BP^WT (c) and LIV-BP^SS (d) variants imaged in the presence of leucine at the K_d of each variant (40 nM and 5.6 μM, respectively) and at the indicated time resolution in milliseconds. FRET values from all selected traces (count at top right of each panel) summed into time-dependent population FRET histograms, represented as contour plots. Scale bar is shown at the right. Two distinct FRET peaks are apparent at all time resolutions for LIV-BP^WT; LIV-BP^SS displayed only a single peak at low time resolution (≥100 ms) that resolved into two district populations in the millisecond regime. e–h, LIV-BP^WT and LIV-BP^SS imaged at time resolutions that most-completely sampled the FRET transitions (25 ms and 0.25 ms, respectively) and idealized using the segmental k-means algorithm (Supplementary Methods). Dwell-time distributions of LIV-BP^WT in the low- (e) and high-FRET (f) states, and LIV-BP^SS in the low- (g) and high-FRET (h) states, in the presence of the indicated concentrations of leucine. All experiments were performed at least three times with similar results. i, j, Rate constants derived from maximum likelihood analysis of dwell times in the low- (black squares) and high-FRET (red circles) states, fit to lines to determine ligand association (k_on) (black lines) and ligand dissociation (k_off) (red lines) rate constants for LIV-BP^WT (i) and LIV-BP^SS (j). As expected for a bimolecular interaction, the ligand-binding rate increased linearly with ligand concentrations, whereas the ligand-dissociation rate remained constant. k_on values were similar in LIV-BP^WT and LIV-BP^SS at 77 and 30 μM⁻¹ s⁻¹, respectively, whereas k_off values differed by nearly two orders of magnitude at 4.0 and 212 s⁻¹, respectively. Together, these results are consistent with the observed 100-fold difference in binding affinity in the two sensor variants (Fig. 1c). All experiments were performed at least three times with similar results.