Phosphorylation decelerates conformational dynamics in bacterial translation elongation factors

Abstract

Bacterial protein synthesis is intricately connected to metabolic rate. One of the ways in which bacteria respond to environmental stress is through posttranslational modifications of translation factors. Translation elongation factor Tu (EF-Tu) is methylated and phosphorylated in response to nutrient starvation upon entering stationary phase, and its phosphorylation is a crucial step in the pathway toward sporulation. We analyze how phosphorylation leads to inactivation of Escherichia coli EF-Tu. We provide structural and biophysical evidence that phosphorylation of EF-Tu at T382 acts as an efficient switch that turns off protein synthesis by decoupling nucleotide binding from the EF-Tu conformational cycle. Direct modifications of the EF-Tu switch I region or modifications in other regions stabilizing the β-hairpin state of switch I result in an effective allosteric trap that restricts the normal dynamics of EF-Tu and enables the evasion of the control exerted by nucleotides on G proteins. These results highlight stabilization of a phosphorylation-induced conformational trap as an essential mechanism for phosphoregulation of bacterial translation and metabolism. We propose that this mechanism may lead to the multisite phosphorylation state observed during dormancy and stationary phase.

Fig. 1. Thermodynamics of the interaction between EF-Tu and pEF-Tu_T382 with nucleotides, EF-Ts and Glu-tRNA^Glu.

(A) Thermodynamic fingerprint of the interaction between EF-Tu and pEF-Tu_T382 with nucleotides. The themodynamic parameters (ΔG, ΔH, −TΔS, and Δc_P) that describe the interaction of EF-Tu with GDP and GTPγS are compared to those for the interaction of pEF-Tu_T382 with GDP and GTPγS in the bar plot. ITC titration of EF-Ts into EF-Tu (B) and pEF-Tu_T382 (C) at 25°C. ITC titration of pEF-Tu_T382 into Glu-tRNA^Glu (D) and EF-Tu into Glu-tRNA^Glu (E) at 25°C.

Fig. 2. Effects of phosphorylation on the conformational state of pEF-Tu_T382.

(A) X-ray structure of the pEF-Tu_T382–GDP complex. All three domains constituting EF-Tu (G domain and β-barrel I and II domains) are highlighted. The switch I region in the β-hairpin conformation is colored in cyan. The 2mF_o-DF_c simulated-annealing omit map (σ = 1) corresponding to pT382 and GDP is shown as a blue mesh. The maps were calculated after removing pT382 and GDP from the model. (B) Close-up view of the electron density map presented in (A). (C) Details of the interactions of pT382 with R59 from the switch I β hairpin and R377 from the β-barrel II domain. (D) Cα representation of the phosphorylated (in cyan) and nonphosphorylated (in ocher) forms of EF-Tu bound to GDP. The two structures have been superimposed by aligning the G domain. (E) X-ray structure of the pEF-Tu_T382–GTP complex. All three domains constituting EF-Tu (G domain and β-barrel I and II domains) are shown as in (A). The disordered switch I region is highlighted. The 2mF_o-DF_c simulated-annealing omit map (σ = 1) corresponding to pT382 and GTP is shown as a blue mesh. The maps were calculated after removing pT382 and GTP from the model. (F) Close-up view of the electron density map presented in (A) for GTP and pT382 (G). The lack of contacts between pT382 and the switch I region due to local disordered is also highlighted. (H) Cα representation of the phosphorylated (in cyan) and nonphosphorylated (in ocher) forms of EF-Tu bound to GTP. The two structures have been superimposed aligning the G domain.

Fig. 3. SAXS-based structural models of EF-Tu and pEF-Tu_T382 in complex with GDP and GTPγS.

Solution structures of pEF-Tu_T382–GDP (A), pEF-Tu_T382–GTPγS (B), EF-Tu–GDP (C), and EF-Tu–GTPγS (D). The model of E. coli EF-Tu–GTPγS in the closed state was reconstructed on the basis of the coordinate 1EXM (66) and is in very good agreement with the experimental SAXS data. The atomic model of the different complexes is superimposed on the ab initio calculated SAXS envelope shown as a blue meshed surface. In each case, the particle dimensions and experimental (in black) and model-derived (in red) SAXS curves are shown below each model. The calculated ab initio envelopes and experimental SAXS data of the complexes of pEF-Tu_T382 with GDP, GTPγS, and GDPNP (table S4) are in good agreement with the crystal structures and strongly indicate that, upon phosphorylation, pEF-Tu_T382 is in an open conformation independent of the bound nucleotide. a.u., arbitrary units.

Fig. 4. EF-Tu conformational dynamics in the presence of nucleotides assessed by spFRET.

(A and B) Structural models of EF-Tu in the (A) open and (B) closed conformation. Dye attachment sites are Cys⁸² and Cys²²². (C and D) Two-dimensional (2D) histograms of the FRET efficiency E versus the donor fluorescence lifetime τ_D(A) with 1D projections obtained after spFRET analysis of (C) EF-Tu–GDP and (D) EF-Tu–GDPNP. A 15% offset in z was used to reduce noise in the 2D histograms. Overlayed is (solid line) the theoretical relation between E and τ_D(A) in the absence of 1- to 10-ms FRET dynamics calculated using a τ_D estimated from the D-only population, an R₀ = 53 Å, and Gaussian distributed fast dye fluctuations over 6 Å. (E and F) Static versus dynamic PDA analysis of EF-Tu in the presence of (E) GDP and (F) GDPNP. The insets illustrate the substrates used for dynamic PDA analysis [that is, the relative abundance of molecules that coincidently were only in the open (red) or closed (blue) state, and molecules that interconverted from one state to the other during diffusion through the probe volume (green)]. The (relative) area under the green substrate is directly proportional to the chance of observing a molecule that interconverted while diffusing through the probe volume; the larger this area, the more robustly can kinetic rate constants be derived from the data by dynamic PDA. (G) Bar chart of the k_opening and k_closing rate constants. The postfitting 95% relative confidence intervals were very low (<5%) for all parameters, except the k values (smaller than 20% for k > 0.25 s⁻¹). The resulting rate constants are presented as the average ± SD of at least three independent measurements originating from at least two protein purification batches. The significances also apply to the data in (H). (H) Bar chart of the corresponding closed and open state dwell times. Dwell times > 5 ms were not quantified because of the high uncertainty of the corresponding interconversion rate constants (suggesting much slower conformational dynamics). For wild-type EF-Tu, the dwell time in the closed state was only slightly lower (P < 0.05) than that in the open state. GDPNP thus accelerates the rate of opening/closing with respect to GDP, and this frequent sampling of the closed state likely allows the protein to crystallize in this conformation. (I) Equilibrium distance distribution for different variants of EF-Tu, calculated from the k_opening and k_closing rate constants. (J and K) 2D histograms of the FRET efficiency E versus the donor fluorescence lifetime τ_D(A) with 1D projections obtained after spFRET analysis of (J) pEF-Tu_T382–GDP and (K) EF-Tu–GDPNP. (L and M) Static versus dynamic PDA analysis [as in (E)] of (L) pEF-Tu_T382 with GDP or (M) GDPNP. (N to Q) Equilibrium distance distribution for (N) pEF-Tu_T382 and the different EF-Tu mutants (O) EF-Tu_T382E, (P) EF-Tu_T61E, and EF-Tu_R59D (Q).

Fig. 5. EF-Tu phosphorylation is intricately related with the protein conformational cycle.

(A) In the closed state (EF-Tu–GTP complex), all the phosphorylation sites experimentally validated for EF-Tu are either buried or poorly accessible to the solvent. (B) In the closed conformation when bound to aa-tRNA (which is likely the more abundant state of the EF-Tu–GTP complex given the high affinity for aa-tRNA), only two phosphorylation sites are not involved in interactions with the aa-tRNA, and they have very low solvent accessibility. (C) In the open state, four sites become accessible by >25% and three others are partially accessible. This suggests that a modification trapping EF-Tu in an open state would increase the likelihood of additional modifications.

Fig. 6. Regulation of the EF-Tu function in translation by phosphorylation.

During translation, EF-Tu is observed in two predominant conformational states coupled to the nucleotide bound at a given moment. When bound to GTP, EF-Tu is in a closed state compatible with aa-tRNA binding (all three domains, shown in different shades of blue, are tightly packed). When bound to GDP, EF-Tu is in an open state with significantly weaker affinity for aa-tRNA binding (the three domains are more loosely arranged). This open form cannot usher aa-tRNAs to the ribosome for the elongation of protein synthesis. Phosphorylation (represented by a red dot) of EF-Tu traps the protein in its inactive open conformation independent of the nucleotide bound to the G domain. As shown in Fig. 1 (D and E), this decoupling of the conformational cycle from nucleotide binding precludes aa-tRNA binding and strongly inhibits translation. As in the case of the open GDP-bound form, phosphorylated EF-Tu is incompatible with aa-tRNA binding and cannot form a ternary complex to deliver aa-tRNAs to the ribosome (21). In this stable open state, subsequent phosphorylation might be facilitated, establishing a functional barrier preventing recoil from dormancy and other differentiation states.