Evidence against the ‘Y-T coupling’ mechanism of activation in the response regulator NtrC

Y101 rotameric state is kinetically independent from the inactive to active transition

Using ¹⁵N NMR CPMG (Carr-Purcell-Meiboom-Gill) relaxation dispersion experiments originally described by Loria et al.²⁹ and Vallurupalli et al.³⁰, in combination with an independently determined exchange-free transverse relaxation rate³¹, we determined the existence of a second dynamic process in NtrC^R, in addition to the previously described inactive/active transition.² For a resonance undergoing exchange, the transverse relaxation rate (R₂^eff) is equal to the intrinsic rate of relaxation (R₂₀) plus an additional rate of relaxation due to exchange (R_ex). The latter can be suppressed by an external magnetic field, a phenomenon that has extensively been used to quantify conformational exchange processes in the millisecond to high microsecond time regime in CPMG experiments^{29; 32;33}. Typical CPMG relaxation dispersion profiles are shown in Fig. 1 A-C, where increasing CPMG frequencies (ν_CPMG) are applied to suppress the R_ex component of R₂^eff with full suppression at the highest field strength of 1000 Hz that should correspond to the exchange-free transverse relaxation rate.²⁹ These data are routinely fit to an equation describing a two-site exchange process to obtain the corresponding rates. However, we discovered that for a number of residues, the seeming exchange-free relaxation rate determined from the CPMG experiments (R₂β_1000Hz) is actually not the true R₂₀β by measuring the latter via a combination of three TROSY based NMR relaxation experiments as described in detail elsewhere.³¹ These values are significantly lower (shown as values for ν_CPMG→∞) than the one predicted from the CPMG curves, revealing a second dynamic process.

We note that this “hidden” second process could not be unambiguously identified for unphosphorylated wildtype NtrC^R (WT) because its rate of inactive/active interconversion is ~14,000 s⁻¹; too fast to be suppressed by the maximum CPMG field strength of 1000 Hz.² However, for unphosphorylated S85D, Y101F along with WT activated by the phosphate mimic BeF₃⁻, the inactive/active transition is slow enough² to separate the dynamics of the activation step from the faster second process.

By comparing R₂β_1000Hz and R₂₀β we were able to probe on a per-residue basis both dynamic processes. The values for R₂β_1000Hz and R₂₀β calculated for most amides agree well with each other, but a small percentage of residues in S85D, Y101F, and BeF₃⁻ -activated NtrC^R have R₂β_1000Hz values significantly higher than the corresponding R₂₀β (Fig. 1D-F). This provides evidence that in NtrC^R a second faster exchange process is sensed in a subset of residues, which is not fully suppressed in the CPMG experiment, and which does not have any perceptible effect on the CPMG dispersion curves itself. Interestingly, the amino acids with exchange from both processes (orange, Fig. 2G) cluster around Y101 (blue) suggesting that this second faster process is linked to Y101 dynamics. Furthermore, it implies that this movement is unrelated to the inactive/active interconversion, in sharp contrast to the dominant ‘Y-T coupling mechanism’.

Y101L mutant verifies that the faster dynamic process is due to rotamer exchange

In order to test this hypothesis, we mutated Y101 to a leucine. This might be considered a dramatic substitution, as the residue in position 101 is typically conserved as either a tyrosine or a phenylalanine in the NtrC family, with only less than 2% of occurrences as leucines and high conservation across response regulators (Figure S4A). The Y101L NtrC^R mutant could be expressed and purified with no apparent differences in expression levels or stability compared to WT. A backbone amide ¹⁵N-CPMG dispersion experiment was carried out on the Y101L mutant and the residues undergoing exchange were fit globally with a rate of exchange equal to 2,200 ± 300 s⁻¹ (Figure 2A), similar to the rate constants of Y101F.²

Strikingly, the second fast process is absent in Y101L as determined by the fact that the R₂₀β’s and R₂β_1000Hz values from the CPMG experiments are identical (Fig. 2 A, B). Only one residue has a statistically significant difference in value of R₂₀β, residue S13. This serine is located in a loop near the phosphorylation site, next to D10 and D11 that help in the coordination of a Mg²⁺ ion upon phosphate binding. Since S13 is far from position 101, it is reasonable to hypothesize that S13 is not sensing the second fast process in WT, but possibly the binding and release of positively charged sodium ions in the buffer.

Two additional observations from the NMR spectra of the Y101L mutant are worth mentioning. First, the equilibrium of the two conformational substates is shifted more toward the inactive state (Fig. 2C), as revealed by the peak position relative to WT and an activating mutant D86N/A89T.¹ Second, residues with dispersion were not only found on the ‘3445 face’, as in previously studied WT, mutants, and activated-NtrC^R, but were also observed along α-helix 5. By removing the aromatic sidechain from the residue at position 101 in NtrC^R, α-helix 5 now appears to be able to sense the dynamics of the activation transition. We note that chemical shift differences between states do not mean that these residues move, but rather that this nucleus has different electronic environments in the exchanging states. A change in the allosteric network upon mutation of the conserved aromatic group has been observed previously in the receiver domain of RegA, with new dynamics measured in α-helix 1 with a constitutively active tryptophan mutation.³⁴

In summary, NMR experiments on Y101L and BeF₃⁻-activated WT NtrC^R provide confirmation that the interconversion between rotameric states of Y101 is the cause of the discrepancy in R₂₀β and R₂β_1000Hz values in S85D, Y101F, and BeF₃⁻ -activated NtrC^R. The rate of the tyrosine flip is faster than the measured rate of the inactive/active conformational change, even in the activated protein, signifying that ‘Y-T coupling’ is not rate-limiting in the activation mechanism of NtrC^R.

MD simulations identify rotameric exchange as independent from activation dynamics

To gain more insight at an atomistic level on the nature of the fast motion involving Y101, its correlation with T82, and the connection of this motion with the global active to inactive state interconversion, we performed a long unbiased MD simulation of the apo active conformation of NtrC^R using the supercomputer Anton. During the 71 μs long simulation we observe that the side chain of tyrosine 101 is not locked in a single conformation, but alternatively interconverts among the three accessible rotameric states (Fig. 3A). The three rotameric states have relative populations of approximately 0.68, 0.28, and 0.04, for the t, g+ and g− states, respectively, and the typical residence time in the most populated one is on the order of 1 μs (see Fig. S1A). The state with the largest population, state t, not surprisingly corresponds to the χ₁ angle that is most often associated with the active state X-ray structures of response regulators (Fig. 3D, green).

MD simulations can directly address the question of correlated motions that cannot be directly extracted from ensemble-averaged experiments. The simulation shows no strong correlation between the motions of Y101 and T82. The side-chain of T82 remains in the same rotameric state for almost the entire length of the simulation, independent of the position of Y101 (Fig 3B). Interestingly, it is only when Y101 is completely moved away from the active site, as in the minor state g−, that T82 has the freedom to explore all three side-chain rotamers with sizable probability (Fig. S1B). The key result from the simulations is that despite the different conformations attained by Y101 and T82, the rotation of α–helix 4 around its axis, which reports on the crucial change in the solvent exposure of L87, A90, and V91 and is important in downstream signaling, remains in the active conformation throughout the simulation (Fig 3C). This long unbiased MD simulation delivered crucial information about the atomistic details of the Y101 dynamics as well as information about the question of correlated motions, neither of which could be addressed directly by experiments.

Using a combination of computation and experimental data we have shown that the fast timescale motion of Y101 is a rotamer interconversion that is uncorrelated with the global conformational transition that the ‘3445’ signaling face experiences during the activation/inactivation process in NtrC^R, and not correlated to the T82 dynamics, ruling out a Y-T activation mechanism for NtrC as had been proposed in earlier studies.¹⁹ In this context it is important to point out that the structural description of active and inactive conformers in NtrC^R is quite different from the one typically used when discussing receiver domains of other response regulators. In CheY the structural variation between the active and inactive conformers is more localized than in NtrC and can be described by changes in the positions of a handful of residues, including T82, Y101 and the loop between β-strand 4 and α-helix 4, whereas the conformation attained by α–helix 4 is nearly identical in the two conformers.³⁵ In NtrC^R, a more global conformational change occurs and this global change is essential for signal propagation. For this reason, the rotameric state of Y101 may be less important in the activation mechanism and downstream signaling of NtrC^R as compared to other response regulators, such as CheY.

Comparison between NMR experiments and MD simulations

The MD simulations on activated apo NtrC^R suggest that the second faster process detected by the NMR relaxation experiments may be a rotamer exchange of Y101, in agreement with the NMR data on Y101L. Three bond J-coupling values deliver valuable information about the χ1 rotamer angles of aromatic residues by NMR.³⁶ To test our hypothesis, we determined the ³J_NCγ value for Y101 in BeF₃⁻-activated WT NtrC^R to be 1.6 +/ 0.2 Hz, where gauche χ1 angles cluster around 0.5 Hz and trans angles cluster around 2.4 Hz. Aromatic residues that sample more than one rotamer in the fast NMR time regime will have an intermediate value. The ³J_NCγ for Y101 in the activated state is consistent with the tyrosine sampling multiple rotamers with a higher population of the trans state as seen in the MD simulations (Fig. 3A).

From the NMR relaxation data a rough estimate of the rate constant of about 30,000 ± 12,000 process was determined for BeF₃⁻-activated NtrC^R (see supplemental material). We want to note that it is impossible to separate populations from delta omegas in the fast exchange limit. In principle, one could try to estimate the chemical shift difference from the MD simulations and use this information as input for fitting the NMR data. While qualitatively chemical shift perturbations estimated from the MD simulations using Sparta+³⁷ are in the same regions as detected by NMR (Fig. S2), this approach does not yield reliable quantitative results (Fig. S3), reaching the existing current limits in our methods. In the MD simulations three different rotameric states are being sampled with different lifetimes. First, even with a 71 μs simulation, insufficient sampling hampers accurate estimations of populations and rate constants. Second, a multi-state conformational exchange on this timescale cannot be quantitatively dissected from the NMR experiments. Third, the slower and the faster processes measured by NMR superimpose for many residues, and both processes are not necessarily uncorrelated. For these reasons we prefer not to over interpret our data and only calculated a rough estimate of the second process.

Aromatic rotamer dynamics in proteins has been extensively studied and continues to be a topic of interest. MD and fluorescence studies on rotamer exchange on small peptides have found rates of interconversion for aromatic residues on the nanosecond timescale³⁸, but the variability in packing of aromatic residues in proteins can lead to a large range of ring flipping frequencies, spanning from the minutes to the subnanosecond timescale, even within the same protein.^{39; 40; 41;42} In our MD simulation, F99 does not rotate, whereas Y122 rotates in sub-microseconds within NtrC^R.

Mutating the conserved aromatic does not affect activation and NtrC’s downstream ATPase activity

As previously mentioned, the amino acid corresponding to position 101 in NtrC^R is predominantly an aromatic amino acid in all receiver domains throughout the whole two-component system superfamily. Specifically, for NtrC, that amino acid is a tyrosine or phenylalanine in approximately 90% of the sequences classified as NtrC in the prokaryotic two-component systems database,⁴³ with only few occurrence of other amino acids (Fig S4A). Other subfamilies, such as CheY, CheV and CheB, show a slightly higher variation (Fig S4B-D), but in all analyzed response regulators families, the residue has an aromatic side chain in about 85% of the sequences (Fig S4F). The high level of conservation, together with the available structural information of active and inactive conformers, led researchers to identify the tyrosine as crucial trigger in the activation mechanism of the receiver domain of response regulators.⁸ However we have shown here that the conformational changes that Y101 undergoes are not related to the global conformational rearrangement hallmark of the activation of NtrC.

It is instead possible that, as suggested by other studies,^{25; 26} the aromatic residue in position 101 is needed thermodynamically in the stabilization of the active state once the protein becomes phosphorylated. To test this we activated Y101L NtrC^R with BeF₃⁻, a ground state phosphate mimic that has been shown to activate NtrC^R to the same extent as phosphate.^{19; 44} An [¹H¹⁵N]-HSQC overlay of BeF₃⁻ -Y101L with BeF₃⁻ -WT protein show similar spectra, with the peak positions confirming that the activated Y101L mutant shifts its population toward the active state similarly to WT (Fig. 4B). BeF₃⁻ -activated Y101L NtrC^R also shows an exchange rate comparable to BeF₃⁻-activated NtrC^R (1,500 ± 600 s⁻¹ and 1,950 ± 290 s⁻¹ respectively) (Fig. 4A). These results imply that stabilization of the active state is also not dependent on the conserved tyrosine residue.

Since we have shown that Y101 is not necessary for the kinetics or thermodynamics of the functional conformational change in the receiver domain of NtrC^R, the next logical possibility to investigate is a potential crucial role of this conserved aromatic residue in the propagation of this structural change to the next step in the signaling cascade. Stabilization of the active conformation upon phosphorylation exposes a group of hydrophobic residues (L87, A90, and V91) on the ‘3445 face’ which have been suggested to be the contact sites between the receiver domain and the central domain of NtrC.¹⁹ This central domain is responsible for both the oligomerization and ATPase activity of the protein, leading to the activation of the σ⁵⁴-holo enzyme transcription factor. To assess the role of Y101 in the signal transduction between receiver and central domain, we measured and compared the ATPase activity of BeF₃⁻ -activated full-length WT and Y101L NtrC using a colorimetric assay that measures the amount of inorganic phosphate released due to ATP hydrolysis (Fig. 4C). The results show that BeF₃⁻ -activated Y101L NtrC has only a small decrease in the rate of ATP hydrolysis compared to WT indicating that an aromatic residue at position 101 is not essential for transducing the signal to the downstream central domain of the protein.

Similar mutational studies on the activity of homologous response regulators have been carried out previously on CheY and OmpR. A study by Zhu et al.⁴⁵ on CheY looked at the effect of mutations on the ability of CheY in E. coli to direct flagella rotation by monitoring swimming and tumbling during chemotaxis. They found that replacement of the tyrosine by a tryptophan led to a hyperactive phenotype. The mutation to a phenylalanine, conversely, created a phenotype that had very low activity, while the leucine, valine, isoleucine, and glycine mutations resulted in no measureable activity at all. In an X-ray crystal structure of a constitutively active mutant, D13K/Y106W, bound to a peptide of its downstream interaction partner, FliM, CheY was found to be bound to FliM and in the active conformation in two of the four molecules in the asymmetric unit, while the other two molecules were unbound and in the inactive conformation.⁴⁶ In the Y106W single mutant structure, however, CheY was found to be in the active form in the absence of peptide.¹² Based on these results, the authors suggested that there might be a low energetic barrier between the inactive and active conformations in CheY, which includes the conserved aromatic residue. This barrier may increase once the protein is bound to its downstream target, FliM.³⁸ In the OmpR system, which is involved in osmotic regulation, a mutation of the conserved tyrosine to a cysteine resulted in a constitutively active protein as monitored by gene expression.⁴⁷ This mutation also caused an increase in the ability of OmpR to bind to its cognate DNA in vitro. Interestingly, in both of these cases, the rate of phosphorylation and dephosphorylation of the mutant proteins remained the same as for WT suggesting that the difference in activity may not be caused by the receiver domain activation but by the domain’s ability to transmit the signal downstream.

NMR sample preparation

Unlabeled and uniformly ¹⁵N-labeled NtrC^R wild-type, mutant, and BeF₃⁻ activated forms were prepared as previously described. ^{1; 19} S85D and Y101F NtrC^R NMR samples were 0.75 mM in 50 mM Na₃PO₄ buffer, pH 6.75 with 10% D₂O. Y101L, BeF₃⁻ Y101L, and BeF₃⁻ WT NtrC^R samples were 0.3 mM to avoid transient aggregation.

NMR experiments

TROSY ¹⁵N relaxation dispersion experiments were acquired on Varian Inova 600 MHz spectrometer at 298K. Constant time T₂ delays between 60 and 70 ms were used in all TROSY CPMG dispersion experiments,^{29; 60; 61; 62} which roughly yielded 55% of residual signal intensity, and CPMG field strengths between 28 and 1000 Hz were collected. Data was processed using NMRPipe⁶³ and the change in R₂^eff with increasing frequency of CPMG refocusing pulses were fit globally for each mutant to the fast form of the Carver-Richards equation (R₂^eff (ν_CPMG) = R₂₀β(ν_CPMG→∞) + (p_Ip_AΔω²_IA/k_ex)[1-(4ν_CPMG/k_ex)tanh(k_ex/4ν_CPMG)]) in Matlab. The variation of R_ex with an increasing applied radio-frequency field (ν_CPMG) permits the CPMG dispersion experiment to quantitatively determine kinetics (k_ex) for a two-state exchange process^{29; 61} even for highly skewed populations.⁶⁰ Noise estimated from flat residues was used as an input for Monte-Carlo simulations in MatLab to determine the error in the fitted parameters.

In a TROSY-based experiment the exchange free component relaxes with the rate R₂₀β where R₂₀β=R₂₀–η_x+R₁^H/2. η_xy is the cross correlation between the ¹⁵N-¹H dipolar interactions and the ¹⁵N chemical shift anisotropy, which can be measured by determining the ratio of peak intensity for the I_β/I_α using a pulse sequence developed by Wang et al.⁶⁴ R₁^H can be estimated through R₁^H~R₁^HzNz-R₁^N using both a TROSY pulse sequence to calculate R₁^HzNz and a standard experiment to measure the longitudinal relaxation on each amide nitrogen (R₁^N).⁶⁵ The combination of these three experiments allows one to calculate the independently determined R₂₀β.³¹

The exchange-free transverse relaxation time of the TROSY component was determined using pulse sequences designed to measure R₂^α and R₂^β with a 2τ delay of 43.2 ms to ensure a ratio I^β/I^α~0.3 for most residues, and R₁^2HzNz with a τ delay of 21.6 ms.^{31;64 15}N R₁ values were measured using standard experiments⁶⁵ with the following T₁ delays: 0.01, 0.1, 0.5, 0.7, 1.1, 1.5 s. R₁^H values were measured using a TROSY-based R₁^HzNz pulse sequence (L.E. Kay, personal communication) employing a reburp pulse to selectively invert the amide region with following delay times: 0.025, 0.04, 0.06, 0.1, 0.15, 0.2, 0.3, 0.35 s. Data was processed using NMRPipe. Intensities were fit to a mono-exponential decay curve, and uncertainties were measured from duplicate points and an estimate of signal-to-noise (2%).

³J_NCγ couplings for aromatic and residues were obtained from quantitative J-correlation spectroscopy³⁶ in duplicate with a dephasing time of 50 ms.

ATPase activity assays

Full-length dimeric WT and Y101L NtrC with 3 alanine mutations in the c-terminal domain to prevent higher order oligomerization upon activation⁶⁶ were grown as described previously.⁶⁶ Cells were resuspended in buffer (50mM TrisOAc, pH 8.2, 50mM KCl, 10% v/v glycerol, 0.02% NaN₃, and 0.4mM TCEP) and lysed using 1mg/mL lysozyme (Sigma) at 25°C for one hour. Cleared lysates were precipitated by a 0-35% (NH₄)₂SO₄ saturation and were further purified by ion exchange chromatography using a Q Sepharose Fast Flow column (50 mM TrisHCl, pH 7.5, 20 mM NaCl, 0.4 mM TCEP) with a 1M NaCl gradient over 8 column volumes for elution. Rates of ATP hydrolysis were determined using a colorimetric malachite green assay⁶⁷ in order to detect released inorganic phosphate (linear range 0.2-40 μM). Reactions were carried out in triplicate in a 96 well plate using malachite green reagents (Bioassay Systems) and KH₂PO₄ as the standard. Briefly, 1 μM final dimeric NtrC was mixed in with assay buffer (50 mM HEPES·KOH, pH 7.3, 100 mM KOAc, 8 mM Mg(OAc)₂, and 5 mM β-ME) in the presence and absence of 200μM BeCl₂ and 5 mM NaF and incubated at room temperature for 5 minutes. Following the addition of 0.1mM equimolar Mg²⁺ATP, pH 7.0, 80 μL aliquots were taken at 30 second intervals and quenched in 20 μL of the reaction solution (0.337 g Malachite green in 750 mL of HPLC grade H2O combined with 10.5 g ammonium molybdate (Sigma) in 250 mL of 4N HCl). The formation of the reactant malachite green/molybdate/Pi complex was measured by absorbance at 620 nm after incubation at room temperature for 30 minutes to allow for color development.

Molecular simulations

The active structure of the receiver domain of NtrC, refined in a previous study in our lab,⁶⁸ was used as starting structure for the simulation. The system was parameterized with the CHARMM 22-protein all-atom force field with the CMAP backbone energy correction included^{69; 70} using the psfgen plugin in VMD.⁷¹ The system was solvated in a rectangular box with TIP3 water molecules and neutralized with 7 sodium ions. The simulation box therefore contained approximately 24,600 atoms. Periodic boundary conditions were applied to the simulation box. After energy minimization, the simulation box was gradually heated to 310 K with a time step of 1 fs while gradually releasing positional restraints in a MD simulation of 2 ns with the software NAMD 2.7b2.⁷² The system was equilibrated for an additional 70 ns in the NPT ensemble (T = 310 K, P = 1.01325 bar) with the software NAMD, using the Langevin dynamics method⁷³ for temperature control, and the combined Langevin piston Nose-Hoover method^{74; 75} for controlling pressure.

Further testing and the production run were then performed on the supercomputer Anton, designed and produced by DESRES research.⁷⁶ The final configuration of the NAMD trajectory was converted to run on Anton using a set of scripts provided by the staff at PSC and tests were performed to evaluate performance and accuracy and to tune the parameters for the run. The final parameters used for the production runs on Anton were a timestep of 2.5 fs, a 32×32×32 FFT mesh and a 1:1:2 RESPA scheme. With those parameters, the energy drift rate in test NVE simulations was approximately 0.017 Kcal/mol/Dof/us, which for our system (~ 50000 dof) matches DESRES published results for similar systems.⁷⁷ In all subsequent runs the systems were coupled to a Nose-Hoover thermostat.

The performance of the runs was about 12 μs/day. After a first exploratory run at 310K (data not shown) for ~12 μs the system was reintroduced to NAMD, re-equilibrated to 320K, reconverted to Anton and simulated at 320K for approximately 71.5 μs.

The pseudoangle used for describing the rotation of the helix along its axis in Fig 3C is defined in⁶⁸ as the dihedral angle identified by the position of the four center of mass of the following atom groups:

• Group1: residues Val8, Val9, Thr32, Phe33, Leu52, Ser53, Ile80

• Group2: residues Leu87, Asp88, Ala89, and Ala90

• Group3: residues Val91, Ser92, Ala93, and Tyr94

• Group4: residues Asp88, Val91 and Ser92

Δδ¹⁵N were computed for Fig. S2 and Fig. S3 with Sparta+³⁷ for the three possible exchange processes: t ⬄ g−; g+ ⬄ g−; and average t,g+ ⬄ g−. For each of the single states (t, g+ and g−) the chemical shifts have been averaged over 5000 frames extracted from the simulation.