Fig. 4. Kinetic analysis of binding of NT to enNTS₁ shows an induced-fit mechanism.

a Single-exponential (red) and double-exponential (blue) fits of an exemplary stopped-flow time trace at 1 µM NT. Fit residuals and goodness of fits assessed with the Akaike Information Criterion (AIC) indicate double-exponential relaxation with two characteristic rates k₁ and k₂. AIC values −2280.7 and −2454.5 of single- and double-exponential fits, respectively, indicate a significantly better fit for double exponentials. b Error-weighted fits of the conformational-selection (CS) (orange lines) and induced-fit (IF) models (blue lines) to the rates k₁ and k₂ obtained from stopped-flow experiments at different total NT concentrations [L]₀ (data points with errors). Data points represent mean values of rates from fits of 12 stopped-flow time traces at 0.5 µM NT; 6 traces at 0.75, 2 and 3.75 µM; 5 traces at 1 and 2.5 µM; 4 traces at 1.5, 10 and 15 µM; 3 traces at 5 µM. Error bars correspond to 95% confidence intervals estimated as 1.96 times standard errors (‘Methods’). For each model, data points are jointly fitted to the model equations for the rates k₁ and k₂ (‘Methods’) with the function Nonlinear Model Fit of Mathematica 13⁷¹ using the rate constants k_e, k_r, and k_- of the CS and IF reaction schemes as fit parameters. The rate constant k₊ is replaced by the experimentally determined K_D value of 6 nM (Supplementary Fig. 11). For the CS model, an unconstrained fit (orange dashed lines) with AIC value 15.0 leads to the fit values k_e = 1.2 ± 0.5 s⁻¹, k_r = 0.6 ± 0.4 s⁻¹, and k_- = 0.005 ± 0.002 s⁻¹ and, thus, to an unrealistically large probability of the active state (R₂) prior to ligand binding. For more realistic unbound-state probabilities of R₂ smaller than 10% imposed by the constraint k_r > 9 k_e, the fit of the CS model (full lines) is poor with an AIC value of 33.8 that is significantly larger than the AIC value of 15.1 for the fit of the IF model. Source data for this figure are provided as a Source data file.