Conformational states and possible ligand binding mechanisms of typical SBPs.

(A) Structural comparison of SBD2 from Lactococcus Lactis (PDB file:4KR510; cyan) and glutamine-binding protein GlnBP from E. coli (pink). SBD2 and GlnBP share 34% sequence identity with a TM-score of 0.90, indicating high structural similarity. (B) Crystal structures of the ligand-free (PDB file:1GGG11; grey) and ligand-bound (PDB file:1WDN12; green) state of GlnBP from E. coli. (C) Sketch of ligand binding via induced-fit (IF) and conformational selection (CS).

Biochemical characterization, fluorescence labeling and thermodynamic characterization of ligand binding of GlnBP.

(A) SDS-PAGE analysis of GlnBP purity with coomassie-staining. Lane 1, molecular mass ladder with sizes of proteins indicated in kDa; lane 2, purified GlnBP WT; lane 3, purified double-cysteine variant GlnBP(111C-192C); lane 4, purified double-cysteine variant GlnBP(59C-130C). (B) SEC was used to further purify the fluorescently-labeled proteins. The protein absorption was monitored at 280 nm (black curve), the donor dye absorption (AF555) at 555 nm, and the acceptor dye absorption (AF647) at 647 nm. The labeling efficiency of AF555 and AF647 were estimated to be about 71% and 59%, respectively. For the solution-based smFRET measurements, the used protein fractions are indicated in grey. (C) Ligand-binding affinities of refolded, unlabeled GlnBP(111C-192C) was determined by ITC with a Kd = 35 ± 5 nM for L-glutamine (mean value from N = 3 with standard deviation), which is in agreement with previous reports57. The free energy of binding was ΔG = - 42.6 kcal/mol with the enthalpy ΔH = -62.3 kcal/mol and entropy contributions - T*ΔS = 19.9 kcal/mol.

smFRET analysis of GlnBP using diffusion-based μsALEX.

(A) Graphical depiction of an E*-S* histogram obtained by μsALEX; panel adapted from ref. 110. Using μsALEX, the stoichiometry S* can be used to separate donor-only (S > 0.8, Donly), acceptor-only (S < 0.3, Aonly), and the FRET molecular species with both donor and acceptor fluorescently-active fluorophore (S* between 0.3-0.8, DA). Bridge artifacts or smearing caused by donor or acceptor photophysics (photoblinking and/or photobleaching) can cause artificial broadening of the FRET population or a shift of the extracted mean apparent FRET efficiency. (B) μsALEX-based E*-S* histograms of the refolded GlnBP(111C-192C) double-cysteine variant labeled with AF555 and AF647. (C, D) Diffusion-based single-molecule analysis and ligand-binding affinity measurements with μsALEX of doubly-labeled GlnBP(111C-192C) (C) and GlnBP(59C-130C) variants (D) at different ligand concentrations. Values provided are mean +/- SD (N = 3). For plotting purposes the concentration of glutamine in the apo state was set artificially to a value of 0.01 nM in the right parts of panels C/D. Data fitting of the fraction of the high-FRET subpopulation as a function of ligand concentration was performed with the Hill equation, which is a valid approximation for describing the bound fraction of GlnBP as a function of glutamine in the case where [GlnBP] << Kd.

Screening GlnBP for rapid within-burst FRET dynamics.

(A) Burst Variance Analysis (BVA) showing a weak signature of within-burst FRET dynamics in the low E* regime. (B) Two-dimensional E* versus S* scatter plots of dwells in mpH2MM-detected states within bursts detected by the Viterbi algorithm. Arrows and adjacent numbers indicate transition rates in s-1. Transitions with rates <100 s-1 are omitted since such long dwells in a state before transitions are improbable to occur within single-molecule bursts with durations <10 ms and are most probably a mathematical outcome of the mpH2MM optimization framework. The dispersion of the E* and S* values of dwells in mpH2MM-detected states are due to the short dwell times in these states, where the shorter the dwell time in a state is, the lower the number of photons it will include, and hence the larger the uncertainty will be in the calculation of E* and S* values of dwells. E* and S* are E* and S* values uncorrected for background, since in mpH2MM all burst photons are considered, including ones that might be due to background. Full analysis shown in Figure S9.

Screening GlnBP for rapid dynamics within single molecule bursts using E-τ and burst-wise FCS analyses.

(A) Two-dimensional histogram of FRET efficiency (E) versus donor lifetime in the presence of acceptor (τD(A)) for apo (left) and holo (right) GlnBP. The FRET populations coincide well with the theoretical static FRET line (black) indicating the absence of conformational dynamics taking place at timescales faster than ms. (B) Analysis of FRET conformational dynamics using burst-wise FCS for apo and holo states on bursts exhibiting photoactive donor and acceptor fluorophores. The fluorescence autocorrelation functions of the detected donor (DDxDD) and acceptor signal (AAxAA) are displayed in green and red, respectively. The fluorescence cross-correlation function between donor and acceptor signals (DDxDA) is shown in black.

Kinetic analysis of ligand binding and dissociation in GlnBP using SPR.

(A) Fitting of maximal responses in sensorgrams from a measurement set with [Gln] concentrations from 7.8 to 1000 nM (data points) with f = c⁄(1 + Kd⁄[Gln]) leads to Kd = 10±1 nM. (B, C) SPR sensorgrams with an association phase of 50 s at the indicated glutamine concentrations [Gln] followed by a dissociation phase of 50 s with [Gln] = 0 in the bulk flow (data points), and fits of the sensorgrams with the reaction scheme (1) for different values of the effective on-rate constant kon (see Methods for details). (D) Rescaled sum of squared residuals versus kon for fits of sensorgrams with different values of [Gln] in the association phase. Note that multiple repeats for the ligand concentrations [Gln] = 15.6 nM, 62.5 nM, and 125 nM are plotted. The two curves with full lines correspond to fits in panels B and C. The 11 curves with dashed lines correspond to the fits in Figure S17.

Dominant relaxation rate kobs of binding pathways.

(A, B) Induced-fit and conformational-selection pathways with conformational excitation and relaxation rates, ke and kr, and with association and dissociation rate constants, k+ and k-, for the binding-competent conformation of the pathway. (C) Dominant relaxation rate, , of the conformational-selection pathways versus ligand concentration [L]. Blue lines represent the exact pseudo-first-order result with S=ke + kr + k+[L] + k− and Kd=k−(ke + kr)⁄k+ke for k−=10 keandkr=9 ke (upper curve) and k−=0.1 keandkr=9 ke (lower curve). The dashed yellow lines represent the approximate result from equation 2. For the induced-fit pathway, the dominant relaxation rate with S as above is monotonously increasing (similar to for ke > k−) and has the limiting value ke + kr at large ligand concentration37, 47.

Crystal structure and dye accessible volume calculations of GlnBP cysteine variants.

(A, C) Crystal structure of the ligand-free (grey structure) and ligand-bound (green structure) GlnBP with the two labeling positions of the respective variants indicated in blue. (B, D) Simulation of accessible volumes for AF555 and AF647 with values of interprobe distinces based on structural predictions (Cß-Cß distances and fluorophore accessible volumes) and experimental values <R>.

Size Exclusion Chromatography (SEC) of refolded GlnBP WT and GlnBP variants.

GlnBP WT and GlnBP double-cysteine variants were unfolded with 6 M Guanidine Hydrochloride and then refolded via dialysis over two days in PBS buffer (pH 7.4, 1 mM DTT). The selected fractions (grey-shaded area) were collected and used for ITC experiments. For the solution-based smFRET measurements, the selected fractions (grey-shaded area) having the best overlap of protein, donor, and acceptor absorption were used. The protein absorption was measured at 280 nm (black curves) and the donor dye (AF555) absorption at 555 nm or donor dye (ATTO 532) absorption at 532 nm. The acceptor dye absorption (red lines) was measured at 647 nm for AF647 and 643 nm for ATTO 643.

Investigating binding affinity of refolded GlnBP WT and refolded GlnBP(59C-130C) using Isothermal Titration Calorimetry (ITC) measurements.

The graphs depict the changes in heat (DP, top) and enthalpy (ΔH, bottom), due to each injection of L-glutamine into the sample cell, as function of time (top x-axis of each graph) and molar ratio of refolded protein and ligand (bottom x-axis), separately. All ITC experiments were repeated three times and performed without fluorophore labeling.

(A) The mean binding affinity of the refolded GlnBP WT is 22 ± 7 nM and the binding stoichiometry is close to 1. (B) The mean binding affinity of the refolded GlnBP(59C-130C) is 31 ± 3 nM and the binding stoichiometry is close to 1.

L-glutamine-induced conformational changes in refolded GlnBP(111C-192C) visualized by μsALEX measurements.

μsALEX-based E*-S* histograms of the refolded GlnBP(111C-192C) double-cysteine mutants labeled with AF555/AF647 fluorophore pair (A) and labeled with ATTO 532/ATTO 643 fluorophore pair (B). First, the histograms of the apo (no L-glutamine) and holo (500 nM L-glutamine) states of the protein were fitted using a 2D gaussian distribution. Subsequently, these two distributions with variable amplitude were used to fit the intermediate ligand concentrations. Refolded GlnBP(111C-192C) labeled with AF555/AF647 shows an open state at E* = 0.507 and a closed high-FRET state at E* = 0.694 in the presence of a saturating concentration of L-glutamine. Refolded GlnBP(111C-192C) labeled with ATTO 532/ATTO 643 shows an open state at E* = 0.346 and a closed high-FRET state at E* = 0.552 in the presence of a saturating concentration of L-glutamine.

L-glutamine-induced conformational changes in refolded GlnBP(59C-130C) visualized by μsALEX measurements.

μsALEX-based E*-S* histograms of the refolded GlnBP(59C-130C) double-cysteine mutants labeled with AF555/AF647 fluorophore pair. First, the histograms of the apo (no L-glutamine) and holo (500 nM L-glutamine) states of the protein were fitted using a 2D gaussian distribution. Subsequently, these two distributions were used to fit the intermediate ligand concentrations. Refolded GlnBP(59C-130C) shows an open state at E* = 0.735 and a closed high-FRET state at E* = 0.891 in the presence of a saturating concentration of L-glutamine.

Investigating biding affinities of fluorescently labeled GlnBP variants using smFRET measurements.

Binding curves of GlnBP in semilogarithmic fashion of [Gln] vs. bound fraction of protein from μsALEX experiments using the AF555/AF647 dye pair (see Figure S3A and Figure S4). The fraction closed, i.e., the fraction of liganded protein, was determined from the ratio of the area of the high-efficiency peak and the total peak area from the projections in the apparent FRET efficiency. The fraction bound as a function of ligand (L-glutamine) concentration was fitted with the Hill equation using Origin 2016 (Origin Lab Corp, Northampton, MA), with the maximum number of binding sites fixed to 1. All the measurements were repeated three times.

Conformational states of refolded GlnBP variants probed by solution-based μsALEX measurements reveal nearly unchanged conformations.

(A) Apparent FRET efficiency histograms of refolded GlnBP(111C-192C) labelled with AF555/647 in the absence (first row) and presence of L-arginine. (B) Apparent FRET efficiency histograms of refolded GlnBP(59C-130C) labelled with AF555/AF647 in the absence (first row) and presence of L-arginine.

Investigating L-Arginine binding affinity of refolded GlnBP(111C-192C) and GlnBP(59C-130C) variants using Isothermal Titration Calorimetry (ITC) measurements.

The graphs depict the changes in heat and enthalpy with the injection of the L-Arginine against the time and molar ratio of refolded protein and ligand, separately. All ITC experiments were repeated three times and performed without fluorophore labeling. (A) The average binding affinity of the refolded GlnBP(111C-192C) is 421 ± 292 μM. (B) The average binding affinity of the refolded GlnBP(59C-130C) is 737 ± 133 μM. The binding ratio (sites) was manually fixed to N = 1.

Screening GlnBP(111C-192C) for rapid within-burst FRET dynamics.

Confocal-based single-molecule FRET results for GlnBP(111C-192C) doubly-labeled with ATTO 532 and ATTO 643, in the apo state (left panels), near the Kd (middle), and in the holo state (right). (A) Burst Variance Analysis (BVA) showing a weak signature of within-burst FRET dynamics in the low E* regime. (B) Histograms of E* values of bursts, (C) E* versus S* 2D histograms of bursts, (D) 2D scatter plots of bursts classified by mpH2MM, with colors corresponding to which state(s) are present within the bursts as determined with the Viterbi algorithm. Locations of states are given by red circles, and black crosses represent the SD of E* and S* values of dwells within each state. (E) E* versus S* 2D scatter plots of dwells in mpH2MM-detected states within bursts detected by the Viterbi algorithm. Red circles and black crosses are same as in (D). Arrows and adjacent numbers indicate transition rates in s-1 units. Transitions with rates less than 100 s-1 are omitted, since such slow transitions are improbable to occur within single-molecule bursts with durations shorter than 10 ms and are most probably a mathematical outcome of the mpH2MM framework. The dispersion of the E* and S* values of dwells in mpH2MM-detected states are due to the short dwell times in these states, where the shorter the dwell time in a state is, the lower the number of photons it will include, and hence the larger the uncertainty will be in the calculation of E* and S* values of dwells. E* and S* are E* and S* values uncorrected for background, since in mpH2MM all photons within bursts are taken into account, including ones that might be due to background.

Screening GlnBP(111C-192C) for rapid within-burst FRET dynamics.

Confocal-based single-molecule FRET results for GlnBP doubly-labeled at residues 111 and 192 with AF555 and AF647, in the apo state, near the KD, and holo state. (A) Burst variance analysis showing a weak signature of within-burst FRET dynamics. (B) Histograms of E* values of bursts, (C) E* versus S* 2D histograms of bursts, (D) 2D scatter plots of bursts classified by mpH2MM, with colors corresponding to which state(s) are present within the burst as determined with the Viterbi algorithm. Locations of states are given by red circles, and black crosses represent the SD of E* and S* values of dwells within each state. (E) E* versus S* 2D scatter plots of dwells in mpH2MM-detected states within bursts detected by the Viterbi algorithm. Red circles and black crosses are same as in (D). Arrows and adjacent numbers indicate transition rates in s-1 units. Transitions with rates less than 100 s-1 are omitted, since such slow transitions are improbable to occur within single-molecule bursts with durations shorter than 10 ms and are most probably a mathematical outcome of the mpH2MM framework. The dispersion of the E* and S* values of dwells in mpH2MM-detected states are due to the short dwell times in these states, where the shorter the dwell time in a state is, the lower the number of photons it will include, and hence the larger the uncertainty will be in the calculation of E* and S* values of dwells. E* and S* are E* and S* values uncorrected for background, since in mpH2MM all photons within bursts are taken into account, including ones that might be due to background.

Screening GlnBP(59C-130C) for rapid within-burst FRET dynamics.

Confocal-based single-molecule FRET results for GlnBP doubly-labeled at residues 59 and 130 with AF555 and AF647, in the apo state, near the KD, and holo state. (A) Burst variance analysis showing a weak signature of within-burst FRET dynamics. (B) Histograms of E* values of bursts, (C) E* versus S* 2D histograms of bursts, (D) 2D scatter plots of bursts classified by mpH2MM, with colors corresponding to which state(s) are present within the burst as determined with the Viterbi algorithm. Locations of states are given by red circles, and black crosses represent the SD of E* and S* values of dwells within each state. (E) E* versus S* 2D scatter plots of of dwells in mpH2MM-detected states within bursts detected by the Viterbi algorithm. Red circles and black crosses are same as in (D). Arrows and adjacent numbers indicate transition rates in s-1 units. Transitions with rates less than 100 s-1 are omitted, since such slow transitions are improbable to occur within single-molecule bursts with durations shorter than 10 ms and are most probably a mathematical outcome of the mpH2MM framework. The dispersion of the E* and S* values of dwells in mpH2MM-detected states are due to the short dwell times in these states, where the shorter the dwell time in a state is, the lower the number of photons it will include, and hence the larger the uncertainty will be in the calculation of E* and S* values of dwells. E* and S* are E* and S* values uncorrected for background, since in mpH2MM all photons within bursts are taken into account, including ones that might be due to background.

Effects on conformation of GlnBP(111C-192C) under various conditions.

Due to the high binding affinity of GlnBP for L-glutamine, several control experiments under different conditions were performed to exclude artifacts induced by the reagents present in each set of experiments. The μsALEX experiments of the refolded GlnBP(111C-192C) double-cysteine variant labeled with LD555/LD655 fluorophore pairs were measured in PBS buffer (pH 7.4) using conventional microscope glass slides (A) and using TIRF chamber (B). The PBS buffer containing (C) 40 mM glucose, (D) 50 nM Ni2+, (E) pyranose oxidase/catalase (POC) and (F) protocatechuate-dioxygenase (PCD)/3,4-protocatechuicacid (PCA) was used for the ALEX measurements. (G) The conventional glass coverslips used in μsALEX experiments (top figure) and TIRF chambers (sticky-Slide 18 well, Ibidi; non-sealed chambers: middle panel; sealed: bottom panel) glued on top of PEG-/biotin-PEG-silane microscope glass coverslips used in the TIRF experiments.

Comparing smFRET measurements of biotin-modified dsDNA and GlnBP(111C-192C) using diffusion-based μsALEX versus TIRF microscopy.

(A) Schematic view of dsDNA labeled with Cy3B and ATTO 647N for smFRET characterization on PEGylated coverslips. (B) Typical μsALEX-based E*-S* histograms of the biotin-modified dsDNA labeled with Cy3B and ATTO 647N. (C) Representative fluorescence time trace of respective single emitter of the biotin-modified dsDNA sample under continuous wave excitation of ∼500 μW at 532 nm and the FRET histograms of all analyzed molecules and the FRET histograms of all measured molecules combined. (D) Schematic view of the refolded GlnBP(111C-192C) labeled with ATTO 532 and ATTO 643 for smFRET characterization. (E) Typical μsALEX-based E*-S* histograms of the refolded GlnBP(111C-192C). (F) Representative fluorescence time trace of respective single emitter of the refolded GlnBP(111C-192C) under continuous wave excitation of ∼500 μW at 532 nm and the FRET histograms of all analyzed molecules. Additional data for each condition is shown in Figures S13/S14.

Representative fluorescence time traces of respective single emitter of biotin-functionalized DNA labeled by maleimide-modified derivatives Cy3B and ATTO 647N (13 bp inter-dye distance). All measurements were done in oxygen scavenging buffer (3 U/mL of pyranose oxidase, 90 U/mL of catalase and 40 mM glucose, PBS buffer, pH 7.4). Laser power: 500 μW.

Examples of fluorescence time traces of respective single emitter of refolded GlnBP(111C-192C) labeled by maleimide-modified derivatives ATTO 532 and ATTO 643. All measurements were done in PBS buffer, pH 7.4 and 2 mM Trolox. Laser power with continuous 532 nm excitation: 200 μW.

(A) and (D) Size Exclusion Chromatography (SEC) of SBD(T369C-S451C) and SBD(T159C-G87C). The selected fractions (grey-shaded area) were collected and used for the solution-based smFRET measurements. The selected fractions (grey-shaded area) having the best overlap of protein, donor, and acceptor absorption were used. The protein absorption was measured at 280 nm (black curves) and the donor dye (ATTO 532) absorption at 532 nm. The acceptor dye absorption (red lines) was measured at 643 nm for ATTO 643. (B) and (E) Typical μsALEX-based E*-S* histograms of the SBD(T369C-S451C) and SBD(T159C-G87C). (C) and (F) Representative fluorescence time trace of respective single emitter of the SBD(T369C-S451C) and SBD(T159C-G87C) and the FRET histograms of all measured molecules.

SPR sensorgrams at the indicated glutamine concentrations, and fits of the sensorgrams for different values of the effective on-rate constant, kon, as in Figure 6B/C. The rescaled sum of squared residuals versus kon for these fits is shown by dashed lines in Figure 6D.

Exemplary plots of the dominant relaxation rate versus ligand concentration [L] for rate parameters consistent with Eq. (S4). Full lines represent the exact solution given in the caption of Fig. 7, and dashed lines represent the approximate solution for sufficiently small [L] based on Eq. (S1). In (A), the effective off-rate resulting from the exemplary parameters is koff = 0.5 ke. In (B), the effective off-rate is koff ≈ ke because the unbinding process is dominated by the opening of the closed ligand-bound conformation with rate ke for k− ≫ 20 kr as in this example. The limiting value of at large ligand concentrations is ke + kr.

The structure of holo GlnBP with optimized docking of glutamine.

The figure reports the optimized results of docking glutamine onto the crystal structure of GlnBP in holo form, after the glutamine substrate was removed from the structure, and presented back as a docking ligand using the SwissDock web server. From left to right: (i) the glutamine is docked onto the correct binding pocket within the closed conformation of GlnBP, (ii) amino acid side chains are wrapping the docked glutamine from all directions, (iii) and indeed the protein surface covers the docked glutamine, and (iv) the residues covering the docked glutamine seem to carry a net negative charge.

Optimized docking of glutamine to GlnBP in its open and closed conformations.

Using the SwissDock web server, the molecule glutamine was docked onto the crystal structures of GlnBP in its open (pdb:1GGG) and closed (pdb:1WDN; with the glutamine substrate taken away) conformations, and the optimized docking sites as well as the calculated dissociation constant are shown (dissociation constant is calculated out of the binding energies reported in the docking results). The preferred docking of glutamine is the same site within GlnBP. The difference is that while in the open conformation glutamine binds to one domain with the other as a distant domain, in the closed conformation the other domain closes on top of the docked glutamine. Following the calculated binding energies from the optimized docking results, while the dissociation constant of glutamine to GlnBP is 20 μM in the open conformation, in the closed conformation it is 230 nM.