Long range dynamic effects of point-mutations trap a response regulator in an active conformation

Abstract

When a point-mutation in a protein elicits a functional change, it is most common to assign this change to local structural perturbations. Here we show that point-mutations, distant from an essential highly dynamic kinase recognition loop in the response regulator Spo0F, lock this loop in an active conformation. This ‘conformational trapping’ results in functionally hyperactive Spo0F. Consequently, point-mutations are seen to affect functionally critical motions both close to and far from the mutational site.

1. Introduction

The two-component signal transduction system is a ubiquitous prokaryotic signaling module [1]. Two conserved components, a histidine kinase (HK) and a response regulator (RR), function as a biological switch, sensing/responding to environmental changes. In response to a signal, the HK transfers a phosphoryl group to an aspartyl pocket in the conserved‘receive’ domain of the RR eliciting an activating conformational change [2].

We have studied Spo0F as the prototypical RR receiver domain and have elucidated both specific and general mechanistic characteristics [3–12]. In its unphosphorylated-inactive form, Spo0F occupies two conformational families in the important β4-α4 loop. This loop is responsible for HK recognition and displays functional dynamics on the millisecond timescale [8]. One of the conformational families overlays with the phosphorylated-active conformation of the protein. Thus Spo0F, in its inactive form, samples its active conformation. We, along with others, proposed that such dynamic sampling of the active conformation was a common feature for RRs [8,10,13,14].

Recently we have been investigating several Spo0F mutants (L66A, I90A and H101A) that are constitutively active [3,4]. These mutants are more readily phosphorylated resulting in a hyper-sporulation phenotype [4,15]. We hypothesized that since these mutants emulate the function of activated Spo0F, they may possess structural characteristics of the activated protein, particularly in the β4-α4 loop. In this work, we solved the NMR structures of the I90A, H101A and L66A hyper-sporulating Spo0F mutants. I90 is in the β4-α4 loop, H101 in β5-strand and L66 in the middle of α3-helix. As discussed below, for each mutant, we found that the β4-α4 loop adopts a conformation more closely approximating the activated form of Spo0F.

While it is tempting to suggest that these structural shifts are the sole reason why the mutants become active, other possible contributions must be assessed. We, and others, have previously highlighted the existence of intra-protein communication networks in RRs [10,16]. These networks use a combination of hydrophobic/salt bridge contacts and concerted millisecond motions to relay information concerning the phosphorylation state of the protein to distant, functionally important regions. In Spo0F L66 (~20 Å from the β4-α4 loop), I90 and H101 are all part of such a network [10]. With these intra-protein communication networks in mind, we hypothesized that the active Spo0F mutants likely demonstrate altered dynamics that significantly contribute to modified function. To investigate this possibility we performed: (i) ‘conformational transitions’ simulations [17,18] on the wild-type protein and each mutant, concentrating on the β4-α4 loop region and (ii) principal components analyses (PCA) on the wild-type protein and each mutant.

The conformational transitions simulations were performed to report on changes in the conformational space being sampled in the β4-α4 loop between wild-type Spo0F and the mutants, and whether any such changes confirm observed phenotypes of the mutant proteins. Complementary PCA studies were performed to account for correlated and/or anti-correlated motions of structural elements within each protein and whether, in the case of the mutants, there is a shift in these motions away from the wild-type condition. If so, this suggests that there are changes in the intra-protein communication networks that affect function. PCA, a standard tool in the field of multivariate analysis, can extract from a set of interrelated variables a much smaller set that retains most of the variation contained in the full set. By setting up a correlation matrix whose elements are the ensemble average of the pairwise products of displacements from their average position of landmarks (such as the Cα positions), PCA helps in identifying (from a structural ensemble) correlations in conformational rearrangements within a protein [19,20].

2. Material and methods

2.1. Preparation of protein samples

Spo0F mutant proteins were expressed and purified as described previously [21].

2.2. NMR Spectroscopy

NMR experiments were performed at 298K on a Varian INOVA 600. 1.0 – 2.0 mM protein samples in the following buffer were used: 90%:10% or 1%:99% H₂O:D₂O, 25 mM Tris (pH6.9), 50 mM KCl, and 0.02% NaN₃. Sequential assignments were made from HNCACB, CBCA(CO)NH, HNCA, HN(CO)CA, HNCO and HN(CA)CO experiments. Side-chains were assigned from H(CCO)NH, (H)C(CO)NH and HCCH-TOCSY experiments [22–26]. For hydrogen bond information, 100 sequential 12 minute ¹H-¹⁵N HSQCs were recorded to determine exchange protected amide protons. Backbone (ψ and ϕ) dihedral angle restraints were obtained from Cα, Cβ, C′, H_N and N chemical shifts using TALOS [27]. NOEs were obtained from 3D-¹⁵N NOESY-HSQC with 120 and 150ms mixing times and ¹³C NOESY-HSQC with 120ms mixing times. The spectra were processed with NMRPIPE and analyzed with NMRVIEW on LINUX workstations. Molecules were visualized and aligned with PyMOL (http://www.pymol.org).

2.3. Structure Calculations

Protein structure calculations were performed with NOEs, hydrogen bond restraints, and TALOS-predicted ψ and ϕ angles. The programs ARIA (version 1.2) and CNS (version 1.2) were used to calculate solution structures [28,29]. The CNS protocols used simulated annealing with a combination of torsion angle dynamics and Cartesian dynamics with default parameters. The lowest 10 energy structures were further refined in an explicit water solvent by ARIA. Manually assigned NOEs were input as assigned and uncalibrated with respect to distance. The total number of ambiguous NOE restraints allowed for each peak was set to 20 and constrained to 1.8–6.0 Å. The dihedral angle restraints were taken to be 2 S.D. values or at least 20 from the average values. Analysis of Ramachandran plots from PROCHECK [30] showed that 98.0, 98.9 and 97.6% of modeled residues (1–132) were in generously allowed regions or better for L66A, I90A and H101A respectively.

2.4. Structure Preparation and Ensemble Analysis

Structures were prepared for conformational transitions analysis as follows: Structures were initially checked by PROCHECK prior to simulations to ensure good starting structures [30]. Also, GROMACS was used to perform energy minimizations prior to simulation to improve simulation results [31]. For the first step, GROMACS (version 4.0.4) was used for an initial energy minimization (1FSP (model 1 and 6), 1PUX (model 1), 2JVK (model 1), 2JVJ (model 1) and 2JVI (model 1)). The energy minimization was performed with a time step of 2fs with 250 steps at 300°C under the OPLSAA (Optimized Potential for Liquid Simulations all atom) force field. Ensemble analysis was performed through PROCHECK and ProFIT (Martin, A.C.R. and Porter, C.T., http://www.bioinf.org.uk/software/profit/).

2.5. Conformational transitions

The free (1FSP), BeF3 bound (1PUX) and mutant (L66A (2JVK), I90A (2JVJ) and H101A (2JVI)) structures of Spo0F were taken from the Protein Data Bank [3,11,13]. Conformational transitions simulations were performed using the program tCONCOORD. tCONCOORD uses the CONCOORD algorithm for structure generation [17,18]. Despite increases in computer power and recent advances in computer algorithms, MD simulations still remain computationally expensive; furthermore, high-energy barriers are often to intense to overcome within reasonable computational time. A recent approach to deal with these issues is the tCONCOORD method. tCONCOORD uses geometrical considerations for the prediction of protein flexibility. A given input structure is analyzed and translated into a geometric description of the protein. Based on this description, the structure is rebuilt several hundred times, generating an ensemble that allows extraction of essential degrees of freedom - often representing biologically relevant motions. tCONCOORD estimates hydrogen-bond stability via a detailed analysis of the environment and incorporates these data into the constraint definition thereby enhancing the prediction quality of conformational transitions. In addition, tCONCOORD uses the MD simulation program GROMACS [31] to obtain solvation parameters and evaluate the surroundings of a particular hydrogen bond during its construction mode for the prediction of protein conformational flexibility. The structure generation is based on the predefined constraints, such as hydrogen bond-angles, hydrophobics, network and long range constraints. The resulting structures (100 structures) are built starting from random coordinates by iteratively correcting the coordinates to fulfill the constraints. Distances, angles, planarities, and chiralities are corrected simultaneously until all constraints are fulfilled. Each run generates a structure that is completely independent from the previous one.

Standard tCONCOORD parameters were used [17,18]; maximum distance for hydrogen bonds was set to 2.6 Å with a minimum angle for hydrogen bonds of 140°. Hydrophobic backbone-backbone, sidechain-backbone and sidechain-sidechain protection was used throughout the simulation. Bond flexibility was set to 0.04 Å, angle flexibility was set to 0.1 Å and 9.0 deg, planarity tolerance was set to 0.03, flexibility for restricted dihedrals was set to 0.1%, and flexibility for non-restricted dihedrals was set to 0.6%. The same tCONCOORD procedure was used for all PDBs.

2.6. PCA analysis

PCA was performed on the NMR structure ensembles using THESEUS [19,20]. THESEUS will calculate the principal components of the covariance and correlation matrices for analysis of the major modes of correlated conformational differences within a superposition. In this study, the input files into THESEUS for PCA analysis were the structural ensembles generated from the tCONCOORD studies. The first 3 principal components were determined for each ensemble of structures from the conformational transitions simulations. Only the first few principal components are of practical interest accounting for the majority of correlations in the data, i.e. the largest eigenvalues.

3. Results and Discussion

The first phase of this investigation necessitated solving the NMR structures of the L66A, I90A and H101A Spo0F mutants. NMR structural statistics are given in Table 1 and the structures of all three hyper-sporulating mutants are shown in Fig. 1. Each structure shows the expected global fold, with five parallel β-strands surrounded by five α-helices. These structures were compared to the structures of wild-type inactive-Spo0F and wild-type active-Spo0F-BeF₃.

Table 1.

NMR and refinement statistics for protein structures

	L66A	I90A	H101A*
NMR distance and dihedral
Distance constraints
Total NOE	3582	3238	2640
Intra-residue	654	740	665
Inter-residue	2928	2498	1975
Sequential (|i–j| = 1)	532	590	505
Medium-range (|i–j| < 4)	573	686	524
Long-range (|i–j| > 5)	1823	1222	946
Hydrogen bonds	94	105	93
Total dihedral angle restraints
φ	43	63	75
ψ	43	63	75
Structure statistics
Violations (mean and s.d.)
Distance constraints (Å)	0.036±0.002	0.036±0.002	0.046±0.001
Dihedral angle constraints	0.423±0.090	0.939±0.093	0.711±0.099
Max. dihedral angle	0.513	1.032	0.810
Max. dist. constraint	0.038	0.038	0.47
Deviations from idealized
Bond lengths (Å)	0.00522±0.00013	0.00479±0.00017	0.00631±0.00012
Bond angles (°)	0.684±0.010	0.667±0.019	0.760±0.015
Impropers (°)	1.842±0.087	1.915±0.112	1.890±0.115
Average pairwise r.m.s.**
Heavy	0.73±0.09	1.19±0.16	0.78±0.07
Backbone	0.31±0.04	0.69±0.23	0.36±0.08

^*Relative stability and line broadening of H101A was worse compared to L66A and I90A, nevertheless this still represents 21 restraints per residue.

^**Pairwise r.m.s. deviation was calculated among 10 refined structures over residues 5–118.

Fig. 1.

NMR structures of L66A, I90A and H101A. Ensembles of the 10 lowest energy structures of L66A (A), I90A (B), and H101A (C) shown through stereo images rendered through PyMOL (http://www.pymol.org).

As noted, we have previously shown that in its unphosphorylated-inactive state, the β4-α4 loop of Spo0F exists in two conformations and that one of these conformations overlays with the β4-α4 loop found in the phosphorylated-active form of the protein (Fig. 2A and 2B - inactive Spo0F blue and green and active Spo0F pink). In each mutant, the same circumstance is observed - the β4-α4 loop more closely adopts the β4-α4 loop conformation found in the active protein. As a representative example, the overlay of the β4-α4 loop in H101A Spo0F (orange) with the β4-α4 loop in both inactive-Spo0F and active-Spo0F-BeF₃ is shown in Fig. 2C and 2D. The β4-α4 loop in H101A aligns much more closely with the β4-α4 loop from the active form of Spo0F than the β4-α4 loop from the inactive form. Similar results are observed for L66A and I90A mutants (Supplementary Fig. S1). Clearly, the point mutations result in an activating structural change in Spo0F’s β4-α4 kinase recognition loop. Importantly, two of these residues are removed from that loop. Indeed, L66 is ~20 Å distant from the β4-α4 kinase recognition loop and has practically no exposed surface area.

Fig. 2.

(A) overlay of inactive wt Spo0F (two structural families - blue and green) with active wt Spo0F-BeF₃ (pink); (B) expansion of the β4-α4 loop (same color scheme); (C) same as (A) but now including H101A Spo0F strcuture (orange) - representative of hypersporulation mutants; (D) expansion of the β4-α4 loop (same color scheme).

The question of whether the dynamic makeup of the protein is also affected by the point-mutations was then addressed. The conformational transitions simulation data for wild-type Spo0F (Fig. 3A) shows that the β4-α4 loop has significant propensity for motion. In fact, this region displays the largest motional proclivity in the protein (neglecting the termini). Conversely, conformational transitions data for each mutant shows that motional propensity and amplitude is significantly reduced in the β4-α4 loop (L66A shown in Fig 3B, I90A and H101A in Supplementary Fig. S2). Coupling these data with our mutant Spo0F structural studies - where the β4-α4 loop more closely approximates active Spo0F - we can say that the mutants are seemingly ‘conformationally trapped’ in a conformation that promotes function. Their ability to sample inactive conformations in this region is significantly reduced.

Fig. 3.

(A) Conformational transitions simulation of inactive wt Spo0F. The two structures with the greatest RMSD over the simulation are shown in green and red. Black arrows highlight the magnitude of this difference in the β4-α4 loop. (B) Conformational transitions simulation of constitutively active L66A Spo0F - same color scheme as (A). (C) Principal components for residues in kinase recognition cluster 6. Symbols are as follows: blue “” Spo0F; red “” Spo0F-BeF₃; orange “” L66A; purple “” I90A; black “*” H101A.

PCA of the conformational transitions data for wild-type and mutant proteins provides insight into how the intra-protein communication networks propagate information. The aim here was to see whether the mutants demonstrated internal motional traits that were shifted from inactive wild-type motions and, if so, whether these shifted traits possessed any characteristics of the active wild-type motions. PCA is an ordination method that allows a cloud of data points to be evaluated by rotating it such that the maximum variability is visible and correlated clusters are evident. The PCA data here are presented in 3D, with each axis corresponding to the first three principal components of motions the proteins are experiencing in the simulation. In this application, this presentation allows ‘motional clustering’ features to be visually detected readily.

Using Fig. 3C as a representative example, we see principal components for residues in the previously reported kinase recognition cluster 6 [3]. Kinase recognition cluster 6 (encompassing the β4-α4 loop) represents a set of residues identified using covariance methods that are present in the RR-HK interface. The wild-type Spo0F peaks tend to cluster in a specific region (bottom/center), illustrating an ‘inactive’ PCA signature. The same is true of the Spo0F-BeF₃ peaks, which cluster in a different region (top/left), illustrating an ‘active’ PCA signature. Cluster 6 residues from the mutants all shift away from the inactive Spo0F cluster, indicating that a change in relative motions within each mutant compared to the wild-type has occurred. The PCA characteristics of the mutant proteins no longer closely resemble the PCA characteristics of the wild-type protein. This implies that the general motional characteristics of the mutants have changed with respect to the wild-type protein. For the most part, the cluster 6 residues migrate towards the active Spo0F-BeF₃ cluster, implying a more ‘active’ condition, though this correlation is not large. The same effect is seen for other clusters of residues in Spo0F (Supplementary Fig. S3). Most importantly, the PCA data show that even single residue mutations notably modify the internal dynamic character of the protein. Consequently the intra-protein communication networks are also altered, seemingly in a fashion that contributes to the ability of the mutant Spo0F proteins to be active.

4. Conclusions

Here, we have determined the NMR structures of the constitutively active, hyper-sporulating L66A, I90A and H101A Spo0F mutants. From a structural standpoint we show that seemingly innocuous single residue point-mutations distal to the functionally important β4-α4 loop result in dramatic structural modifications in that region that affect its biological role in HK recognition. Furthermore, conformational transitions investigations and PCA studies show that such localized point-mutations alter the global dynamic composition of Spo0F such that (a) the β4-α4 recognition loop becomes more trapped in its active conformation and (b) the protein’s intra-protein communication network is globally modified and shifts away from the wild-type form. These effects may be common across the response regulator family. In general, because of established intra-protein communication network involving concerted motions, point-mutations may affect function by perturbing both structure and dynamics, even when those mutations are far removed from interaction sites and/or are buried.