Allosteric communication between ligand binding domains modulates substrate inhibition in adenylate kinase

Abstract

Enzymes play a vital role in life processes; they control chemical reactions and allow functional cycles to be synchronized. Many enzymes harness large-scale motions of their domains to achieve tremendous catalytic prowess and high selectivity for specific substrates. One outstanding example is provided by the three-domain enzyme adenylate kinase (AK), which catalyzes phosphotransfer between ATP to AMP. Here we study the phenomenon of substrate inhibition by AMP and its correlation with domain motions. Using single-molecule FRET spectroscopy, we show that AMP does not block access to the ATP binding site, neither by competitive binding to the ATP cognate site nor by directly closing the LID domain. Instead, inhibitory concentrations of AMP lead to a faster and more cooperative domain closure by ATP, leading in turn to an increased population of the closed state. The effect of AMP binding can be modulated through mutations throughout the structure of the enzyme, as shown by the screening of an extensive AK mutant library. The mutation of multiple conserved residues reduces substrate inhibition, suggesting that substrate inhibition is an evolutionary well conserved feature in AK. Combining these insights, we developed a model that explains the complex activity of AK, particularly substrate inhibition, based on the experimentally observed opening and closing rates. Notably, the model indicates that the catalytic power is affected by the microsecond balance between the open and closed states of the enzyme. Our findings highlight the crucial role of protein motions in enzymatic activity.

Keywords: enzymatic activity; protein dynamics; single-molecule FRET.

Fig. 1.

Structural and kinetic attributes of AK variants. (A) Structure of AK with its three domains. The CORE domain (gray) connects the LID domain (pink) and the NMP domain (blue). The yellow spheres indicate the positions for the attachment of fluorescent dyes. The protein can undergo a conformational change from the open state (left, PDB 4AKE) towards the closed conformation (right, PDB 1AKE). (B) Distribution of in-lysate activity of 1,248 colonies in terms of the screening parameter $v_0^{100}$ / $v_0^{500}$ . WT corresponds to a $v_0^{100}$ / $v_0^{500}$ = 1.38. (C) Distribution of K_I values of purified proteins from library variants with high (top 64), mid (64 around median) and low (Bottom 64) values of $v_0^{100}$ / $v_0^{500}$ measured in-lysate. (D) Conservation score mapped onto the open state with high to low conservation represented as blue to red. The positions with mutants having a K_I ≤ 500 μM for AMP are shown as yellow spheres, while those with K_I ≥ 1,500 μM are shown in green. The four magenta spheres are positions where both inhibited and uninhibited mutations are observed (see SI Appendix, Table S1). The conservation scores were obtained from the ConSurf web server (24, 25). (E) The distributions of the conservation scores in the strong and weak inhibition bins were found to be statistically different (P < 0.01, Student’s t test). The mutations that result in strong inhibition (K_I ≤ 500) usually occur at less conserved positions (2.9 ± 0.6), while inhibition-relieving mutations (K_I ≥ 1,500) occur mostly at highly conserved positions (6.2 ± 0.9). (F) Activity curves of WT AK and selected mutants show increased SI (L107I, L82V) or loss of inhibition (F86W).

Fig. 2.

AMP does not close the LID domain. (A and B) FRET efficiency histograms of the WT protein. (A) The apoprotein (blue trace) mainly adopts an open conformation, occasionally exploring the closed state. The binding of ATP increases the population of the closed state, as demonstrated by histograms at 100 μM (green) and 1 mM (purple) ATP. When 1 mM AMP is added, a significant population of the closed state is attained even at a low ATP concentration of 50 μM (orange). In this experiment, ADP was also added at concentrations that maintained the equilibrium of the enzymatic reaction (Methods). The dashed lines indicate the most likely positions of the open and closed state (0.37/0.72 for the WT, respectively). (B) AMP alone, even at 30 mM (green), leads to very minimal changes in FRET efficiency compared to the apoprotein (blue). ATP-γ-S, a slowly hydrolyzable analog of ATP, binds to the ATP binding site and triggers partial closure at 50 μM (orange). In the presence of both ligands (purple), ATP-γ-S is still bound to the ATP-binding site and is not outcompeted by AMP, which would trigger LID opening. Instead, the binding of AMP to its own site even promotes closure to a minor extent. (C) For the apoprotein, the FRET efficiency histogram of the WT (blue) and all mutants are similar. Shown are F86W (green), L107 (red) and L82V (black). (D) FRET efficiency histogram for all mutants in the presence of 5 mM AMP. Only for L82V a significant shift to higher FRET efficiency is detected. (E) Values for the occupancy of the closed state as a function of AMP concentration were obtained by H²MM analysis. The experiment confirms that the occupancy of the closed state for the WT (blue), L107I (red) and F86W (green) is not significantly altered by AMP alone. In contrast, a partial closure is observed for L82V (black). Asterisks indicate the significance of the deviation of parameters from the apoprotein (***: P < 0.01, **: P < 0.05, *: P < 0.1, no index: P > 0.1, Student’s t test). Error bars are given as the SEM based on three independently prepared samples.

Fig. 3.

ATP-dependent closure of the LID domain. The occupancy of the closed state was determined by H²MM in the presence of either solely ATP (circles) or also 1 mM AMP and appropriate concentrations of ADP to maintain equilibrium, as described in Methods (squares). In all variants, the LID domain closes at high ATP levels. For (A) WT, (B) L107I and (C) L82V, the presence of 1 mM AMP promotes closure at much lower ATP concentrations. In the noninhibited mutant (D) F86W, AMP has no significant effect. Dashed (ATP only) and straight (all substrates) lines indicate fits to a binding model as described in SI Appendix, Supplementary Note 1. Error bars are given as the SEM based on two to three independently prepared samples. As in Fig. 2A, ADP was added to maintain equilibrium; the substrate concentration on the abscissa is the total concentration of ligands for the ATP binding site (i.e., ATP + ADP, SI Appendix, Table S5).

Fig. 4.

AMP accelerates LID domain closing. (A–C) Opening (circles, solid lines) and closing (squares, dashed lines) rates for (A) the WT (blue), (B) L107I (red) and (C) L82V (gray) in the presence of increasing concentrations of AMP and a fixed concentration of ATP (1 mM). Opening rates are not affected by the presence of AMP. The closing rates were fitted to a model described in SI Appendix, Supplementary Note 1. (D) Occupancy of the closed state based on the rates shown in (A–C). Error bars are given as the SEM based on two to three independently prepared samples.

Fig. 5.

A model to explain substrate inhibition. (A) Scheme of the model, which is based on two competing pathways: When ATP binds first (green pathway) to the apoenzyme (E), this results in the ATP-bound state ET. Subsequent binding of AMP leads to a productive closed state (ETM) that undergoes the chemical step of the enzyme in the closed conformation (ETM^C). However, if AMP binds first (red pathway, via EM), subsequent binding of ATP causes the formation of ${\mathrm{E}\mathrm{T}\mathrm{M}}_{\mathrm{i}\mathrm{n}\mathrm{h}}$ , which deviates from the productive state ETM in the orientation of the substrates. It is assumed that the closed state ${\text{ETM}}_{\text{inh}}^{\mathrm{C}}$ does not enable phosphotransfer due to an inappropriate substrate orientation. The corresponding open conformation ${\mathrm{E}\mathrm{T}\mathrm{M}}_{\mathrm{i}\mathrm{n}\mathrm{h}}^{\mathrm{O}}$ must convert to ETM^O to enable catalysis. Binding or reorientation of ATP, which lead to a productive catalytic conformation, cannot occur in the closed conformation (23). (B and C) Simulated activity curves of WT and mutants as a function of (B) AMP and (C) ATP concentrations based on the model (A) and experimentally observed parameters of opening and closing rates. The rate of conversion from ETM^O to ${\mathrm{E}\mathrm{T}\mathrm{M}}_{\mathrm{i}\mathrm{n}\mathrm{h}}^{\mathrm{O}}$ was fixed at 250 s⁻¹. The model shows AMP-mediated SI for WT, L107I and L82V, but not for F86W. No SI was observed as a function of ATP concentration, in accordance with experimental data.