Structural basis of a protein partner switch that regulates the general stress response of α-proteobacteria

Abstract

α-Proteobacteria uniquely integrate features of two-component signal transduction (TCS) and alternative sigma factor (σ) regulation to control transcription in response to general stress. The core of this regulatory system is the PhyR protein, which contains a σ-like (SL) domain and a TCS receiver domain. Aspartyl phosphorylation of the PhyR receiver in response to stress signals promotes binding of the anti-σ factor, NepR, to PhyR-SL. This mechanism, whereby NepR switches binding between its cognate σ factor and phospho-PhyR (PhyR∼P), controls transcription of the general stress regulon. We have defined the structural basis of the PhyR∼P/NepR interaction in Caulobacter crescentus and characterized the effect of aspartyl phosphorylation on PhyR structure by molecular dynamics simulations. Our data support a model in which phosphorylation of the PhyR receiver domain promotes its dissociation from the PhyR-SL domain, which exposes the NepR binding site. A highly dynamic loop–helix region (α3-α4) of the PhyR-SL domain plays an important role in PhyR∼P binding to NepR in vitro, and in stress-dependent activation of transcription in vivo. This study provides a foundation for understanding the protein-protein interactions and protein structural dynamics that underpin general stress adaptation in a large and metabolically diverse clade of the bacterial kingdom.

Bacteria use a diverse set of regulatory proteins to control gene expression in response to a changing environment. Among these, two-component signaling (TCS) systems and alternative sigma factors (σ) constitute two major classes of transcriptional regulators (Fig. 1 A and B). The recent discovery of PhyR in the α-proteobacteria (1, 2) provides an example of the confluence of TCS and σ-dependent transcriptional regulation in a single polypeptide. PhyR contains an N-terminal σ-like (SL) domain and a C-terminal TCS receiver domain. Stress-dependent phosphorylation of PhyR indirectly activates transcription of stress-response genes (3) through a unique protein partner switching mechanism detailed below (Fig. 1C).

Fig. 1.

PhyR regulatory system integrates features of TCS and alternative σ regulation. (A) Cartoon depicts transcriptional regulation by an archetypal TCS system. (B) Differential regulation of alternative σ factors under specific environmental conditions is a mechanism of transcriptional control. (C) Hybrid PhyR protein contains an N-terminal SL domain (green) and a C-terminal receiver (Rec) domain (red). σ^T (orange) is bound and inhibited by the anti-σ factor NepR (blue) under normal growth conditions. Phosphorylation of PhyR increases its affinity for NepR, which frees σ^T to bind RNAP and regulate transcription during stress.

PhyR-SL has sequence similarity to the EcfG-family of alternative σ factors (4), which are known to function as general stress regulators in the α-proteobacteria (3, 5–7). However, regions σ₂ and σ₄ of PhyR-SL are missing key residues required for interaction with DNA and RNA polymerase (RNAP) (4). This is consistent with the discovery that PhyR does not function as a true σ factor but, rather, indirectly controls gene expression through its interaction with NepR (3), an anti-σ^EcfG protein (3, 4). Specifically, stress-dependent phosphorylation of the PhyR receiver domain is proposed to disrupt its interaction with the SL domain, thereby enabling PhyR-SL to bind NepR (3). Phospho-PhyR (PhyR∼P) thus functions as an anti–anti-σ factor (i.e., a NepR binding factor) that releases σ^EcfG to directly regulate transcription during stress (Fig. 1C). This regulatory model is conceptually similar to the partner switching systems controlling σ^B (8) and σ^F (9) of Bacillus subtilis and other species (10), though the underlying mechanisms differ.

The biochemical, biophysical, and structural underpinnings of PhyR-regulated transcription remain largely uncharacterized. However, a recent high-resolution crystal structure of Caulobacter crescentus PhyR (11) in its unphosphorylated state informs several testable hypotheses centering on the molecular mechanism of PhyR function. As described in other species (3, 6, 7, 12), phosphorylation of C. crescentus PhyR increases its apparent affinity for the anti-σ^EcfG factor, NepR (13). Genetic and biochemical data on C. crescentus PhyR (11, 13) support the aforementioned partner switching model (3), in which PhyR∼P/NepR complex formation during stress activates transcription of the general stress regulon through σ^EcfG (more commonly known as σ^T in C. crescentus).

In this study, we characterize the molecular and structural basis of PhyR function as an anti–anti-σ factor. A crystal structure of the SL domain of C. crescentus PhyR in complex with the anti-σ factor NepR is combined with molecular dynamics (MD) simulations, in vitro binding studies, and in vivo transcription assays to define the functional PhyR-NepR binding interface. NepR binds regions σ₂ and σ₄ of PhyR-SL across the same molecular surface occluded by the receiver domain in the unphosphorylated structure of PhyR. Thus, the receiver domain must undock from the SL domain before NepR binding. We further demonstrate that the dynamic α3-α4 loop–helix region of the PhyR-SL domain is required for high-affinity binding of PhyR∼P to NepR in vitro, and for activation of σ^T-dependent transcription in vivo. This study defines key molecular determinants of the PhyR/NepR/σ^EcfG protein partner switch that regulates the general stress response in the α-proteobacteria.

Results

Structure of the PhyR-SL/NepR Complex.

The PhyR-SL domain is known to bind NepR constitutively when the C-terminal receiver domain is deleted (3). To characterize PhyR binding to NepR structurally, we first coexpressed and purified the isolated PhyR-SL domain (PhyR_ΔRec) bound to NepR. The crystal structure of C. crescentus PhyR_ΔRec in complex with NepR carries a deletion of 13 amino acids (ΔG68-H80) in the α3-α4 loop (LΔ13), which increased protein stability and facilitated crystallization. Crystallographic data are summarized in Table 1. A simulated annealing composite omit map of a region of the PhyR_ΔRec(LΔ13)/NepR complex structure is presented in Fig. S1.

Table 1.

Crystallographic data and refinement statistics

Data collection statistics (SeMet)
Energy, keV	12.66
Resolution range, Å	30-2.1 along b* and c*; 30-2.7 along a*
Unique reflections	17,325
Rmerge†	0.07
〈I〉/〈σI〉	28.2
Redundancy	7.9
Completeness	95.0%†
Phasing statistics‡ (Dmin), Å
	6.0	3.8	3.0	2.6	2.3	2.1	Overall
Figure of merit	0.45	0.46	0.32	0.22	0.17	0.13	0.27
Refinement statistics
Space group	C2221
a, b, c; Å	75.3, 105.3, 97.9
Rcryst§	20.0
Rfree¶	25.1
〈B〉, Å2	36.7
rmsd of bond lengths, Å	0.008
rmsd of bond angles, °	1.16
Ramachandran analysis
Preferred, %	97
Disallowed, %	<1

D_min, the resolution limit of diffraction by the crystal.

^†R_merge = Σ_hklΣ_i |I_i − 〈I〉|/Σ_hklΣ_iI_i, for all data greater than −3; completeness to 2.7 Å 〈I〉/〈σ_I〉 greater than 2.

^‡Experimental phases were determined by the Autosol SAD routine in PHENIX using the anomalous signal from selenium. Total figure of merit values are based on experimental phase information (prior) for all reflections.

^§R_cryst = Σ_hkl‖F_obs| − |F_calc‖/Σ_hkl|F_obs| using ellipsoidally truncated and anisotropically scaled reflections as described in Materials and Methods.

^¶R_free uses 1,729 total reflections for cross-validation.

PhyR_ΔRec(LΔ13) (i.e., the PhyR-SL domain) and NepR form a heteromeric protein complex in the crystal (Fig. 2). In this structure, PhyR-SL is a homodimer. Regions σ₂ and σ₄ of each PhyR-SL domain are separated about a flexible loop region (α3-α4) and swapped between two monomers; region σ₂ of molecule A (σ₂^A) dimerizes with region σ₄ of molecule B (σ₄^B), and vice versa. Consequently, each complex contains two copies of PhyR-SL that are organized as follows: σ₂^A σ₄^B and σ₂^B σ₄^A (Fig. 2). Gel filtration on purified PhyR-SL (either WT or the LΔ13 variant) at micromolar concentrations indicates that the SL domain alone likely exists as an open monomer in solution, with the elution volume suggesting an extended conformation. Addition of NepR to these proteins results in formation of the dimeric PhyR-SL/NepR complex (2:2 ratio) observed in the crystal (Fig. S2 B and C).

Fig. 2.

Surface representation of the PhyR_ΔRec(LΔ13)/NepR dimer. Open PhyR_ΔRec(LΔ13) molecule A (light green) dimerizes with open PhyR_ΔRec(LΔ13) molecule B (dark green). Bound NepR is colored blue (PDB ID code 3T0Y).

If we consider each swapped σ₂-σ₄ PhyR-SL dimer in the crystal as a single SL domain, each adopts a structure that is nearly identical (rmsd = 0.75 Å) to the closed SL domain reported in full-length PhyR [Protein Data Bank (PDB) ID code 3N0R] (11). We observe a bundle of seven α-helices with α1, α2, and α3 corresponding to region σ₂ and α5, α6, and α7 corresponding to region σ₄ (Fig. 3B). The major difference in the structure of the SL domain presented here compared with the full-length PhyR protein is the shifted position of helix α4. Specifically, α4 in the PhyR_ΔRec(LΔ13)/NepR complex occupies a position that more closely matches what has been reported in other region σ₄ structures, including Thermatoga maritima σ^A (14), Mycobacterium tuberculosis σ^C (15), Escherichia coli σ^E (16), and M. tuberculosis σ^L (17) (Fig. 5 A and C).

Fig. 3.

Structure of the PhyR_ΔRec(LΔ13)/NepR complex (PDB ID code 3T0Y) (A) Amino acid sequence alignment of C. crescentus (Cc) NepR with orthologous sequences from other α-proteobacteria (Me, M. extorquens; Bj, Bradyrhizobium japonicum USDA110B; Rp, Rhodopseudomonas palustris CGA009; ML, Mesorhizobium loti MAFF303099; SM, Sinorhizobium meliloti 1021; and Ba, Brucella abortus 2308). Residues are highlighted according to degree of conservation (upper right key). The regions of NepR sequence not visible in the electron density maps are marked (dashed line) below the alignment. (B) Ribbon structure of the PhyR_ΔRec(LΔ13)/NepR complex. Residues of NepR are colored by sequence conservation (upper right key). PhyR-SL regions σ₂ and σ₄ are outlined in green, and NepR residues for which there was visible electron density (R30 to E62) are marked. (C) Structural alignment between the full-length PhyR (PDB ID code 3N0R) and the PhyR_ΔRec(LΔ13)/NepR structures reveals overlap between the PhyR receiver domain (red) and the NepR (blue) interaction surfaces. (D) Residue interaction map (<3.5 Å) between NepR (blue) and PhyR_ΔRec(LΔ13) (green). Hydrogen bonds (blue lines) and salt bridges (brown lines) are shown. Dashed lines correspond to interactions that are present in only one of the PhyR_ΔRec(LΔ13)/NepR complexes in the asymmetric unit. Residues are numbered, and their locations in the different helices are annotated.

Fig. 5.

Molecular dynamics simulation of PhyR∼P and structural analysis of helix α4 position. (A) Surface model of the PhyR receiver domain (white) interacting with the SL domain (green cylinders). The position of helix α4 in the unphosphorylated full-length PhyR structure (PDB ID code 3N0R) is shown in red; the position(s) of helix α4 in the PhyR_ΔRec(LΔ13)/NepR complex are shown in orange and yellow (PDB ID code 3T0Y). (B) Conformational change of the α11-β5 receiver loop and the α3-α4 region of PhyR between 0 ns (blue) and 200 ns (red). The receiver (Rec) domain (gray) and SL domain (green) are shown (corresponding heat maps are provided in Fig. S4). (C) Superposition of region σ₄ of the PhyR-SL domain (green) with the structures of region σ₄ of σ^A (T. maritime, PDB ID code 1TTY), σ^C (M. tuberculosis, PDB ID code 208X), σ^E (E. coli, PDB ID code 1OR7), and σ^L (M. tuberculosis, PDB ID code 3HUG) (all in gray coils; the first and last residues of each structure are labeled). The position of helix α4 in the full-length PhyR structure (PDB ID code 3N0R) is rendered as a red cylinder. Two positions of helix α4 in the two PhyRΔRec(LΔ13)/NepR complexes in the asymmetric unit are shown as yellow and orange cylinders (PDB ID code 3T0Y).

Because the PhyR-SL domain adopts a dimeric conformation in this structure, it was necessary to test whether full-length PhyR forms dimers in solution on phosphorylation. To do so, we conducted gel filtration and small-angle X-ray scattering (SAXS) experiments on full-length PhyR in the presence and absence of 25 mM acetyl phosphate (AcP), which has been reported to phosphorylate and activate C. crescentus PhyR (13). At a concentration of ∼300 μM, full-length His-PhyR eluted at a volume consistent with a monomer from a Superdex 75 column (GE Healthcare) in the presence and absence of AcP (Fig. S2A). Additionally, the radii of gyration (R_g) determined by SAXS from 8-μM and 20-μM PhyR samples measured in the presence and absence of AcP, respectively, were consistent with the expected monomeric R_g (Fig. S3). Thus, we conclude that full-length C. crescentus PhyR remains monomeric on phosphorylation.

Although the crystal structure presented here clearly shows that the isolated SL domain is capable of opening about the flexible α3-α4 loop region and forming a homodimer, its binding to NepR occurs in a closed conformation (Fig. 3B). Binding at a 1:1 ratio between PhyR and NepR is supported by gel filtration of the full-length C. crescentus His-PhyR∼P/NepR protein complex (∼300 μM), which elutes at a volume consistent with a 1:1 heterodimer (Fig. S2A). These data provide evidence for a full-length PhyR∼P/NepR binding model in which NepR binds monomeric PhyR-SL in a closed conformation.

Defining the Anti-σ/Anti–Anti-σ Interaction.

The anti-σ factor, NepR, contains 68 amino acids; electron density for the first 29 N-terminal residues and the last 6 residues of the C terminus is not visible in our maps. Despite low sequence conservation of these segments (Fig. 3A), we cannot exclude the possibility that the N- and C-terminal regions of NepR are involved in interaction with the receiver domain or with σ^T. The 33 NepR residues visible in the density maps constitute two α-helices connected by a short, four-residue linker (Fig. 3A). NepR wraps around regions σ₂ and σ₄ of the PhyR-SL domain (i.e., the anti–anti-σ domain) (Fig. 3B), occupying the same molecular surface of PhyR-SL in a nearly identical conformation on both copies in the asymmetric unit (rmsd = 0.44 Å). The portion of NepR that binds PhyR-SL constitutes the region of primary sequence that is most conserved among NepR orthologs from other α-proteobacteria (Fig. 3 A and B).

NepR and the receiver domain of PhyR bind the same surface of PhyR-SL (Figs. 3C and 4). Thus, we conclude that the receiver domain must undock from PhyR-SL on phosphorylation to reveal the NepR binding surface. A presentation of surface hydrophobicity, polarity, and conservation is presented in Fig. 4.

Fig. 4.

Electrostatic (A), hydrophobic (B), and surface conservation (C) maps of the different interaction surfaces of the PhyR_ΔRec(LΔ13) domain, the PhyR receiver domain, NepR, and helix α4.

To validate our crystal structure functionally, we next tested whether key PhyR-SL/NepR interactions observed in the crystal are required for binding in solution, and for stress-dependent activation of transcription in C. crescentus cells. In the crystal structure, residues R15 and R16 of the PhyR-SL domain interact extensively with NepR (Fig. 3D). As such, we mutated both R15 and R16 to alanine and conducted a pull-down binding assay between maltose-binding protein (MBP)-NepR and His-PhyR_ΔRec(R15A-R16A); we could detect no interaction between these two proteins. The same experiment performed with MBP-NepR and WT His-PhyR_ΔRec revealed a strong binding interaction (Fig. 6A). Replacement of WT phyR with these mutant phyR alleles on the C. crescentus chromosome showed that both phyR(R15A-R16A) and phyR_ΔRec(R15A-R16A) are stably expressed but are nonfunctional as assayed by transcription from a σ^T-dependent reporter (Fig. 6A). Thus, the interactions observed in the crystal between NepR and PhyR residues R15 and R16 are necessary for PhyR function as an anti–anti-σ factor.

Fig. 6.

(A) Testing the effect of R15A-R16A mutations in PhyR-SL. (Left) Pull-down binding assay between MBP-NepR and His-PhyR_ΔRec(WT) (lane 1) or His-PhyR_ΔRec(R15A-R16A) [lane 2; unbound protein was present in the column flow-through (FT)]. (Right) Transcription (with or without osmotic stress) from a σ^T transcriptional reporter (PsigU-lacZ) in phyR(R15A-R16A) and phyR_ΔRec(R15A-R16A) strains. A Western blot measuring in vivo stability of these mutant proteins is shown below the bar graph; FixJ is the loading control. (B) Testing the functional role of PhyR region α3-α4 in the C. crescentus transcriptional response to osmotic stress. Boundaries of the engineered α3-α4 loop (LΔ) and helix α4 (Δα4) PhyR mutants used for functional and binding studies are shown. (C) Stress-regulated transcription from PsigU-lacZ in WT and the phyR mutant backgrounds depicted in B (LΔ5, LΔ13, LΔ13 + α4, and Δα4). In vivo stability of mutant PhyR proteins as determined by Western blotting is shown below the bar graph; FixJ is the loading control. (D) Pull-down binding assay of His-PhyR(WT) and His-PhyR α3-α4 loop mutants to MBP-NepR in the presence of AcP; quantification of eluted fractions resolved on SDS/PAGE gel is shown below the gel image. A control binding assay conducted in the absence of AcP is shown on the right. (E) Pull-down binding assay of His-PhyR_ΔRec α3-α4 loop mutants to NepR. Quantification of eluted fractions resolved on SDS/PAGE gel is shown below the gel image. All experiments were conducted in triplicate. Mean values ± SEM are shown.

Structural Dynamics of the PhyR Anti–Anti-σ Factor.

Structures presented previously (11) and herein suggest a model in which conformational changes at the PhyR-SL/receiver domain interface and in the α3-α4 region of the PhyR-SL domain have a role in PhyR function as an anti–anti-σ factor. To assess the dynamics of full-length PhyR, we performed an atomistic MD simulation to 320 ns in a fully explicit, solvated system. This simulation used the high-resolution (1.25 Å) crystal structure of full-length PhyR (PDB ID code 3N0R) phosphorylated at residue D192 as a starting model. We observed significant structural change in two key regions: (i) the loop between α-helix 11 and β-strand 5 of the receiver domain and (ii) the α3-α4 loop and helix α4 of the PhyR-SL domain (Fig. 5B, heat maps are presented in Fig. S4).

The molecular surface of the receiver domain in which we observe the largest conformational change in our simulation, the α11-β5 loop, is highly conserved and directly contacts the PhyR-SL domain (Figs. 3C, 4, and 5 A and B). This is a region of structure that is known to undergo conformational change on phosphorylation in multiple two-component receiver proteins (18–22). We observe that retraction of the α11-β5 loop from PhyR-SL begins at ∼120 ns, which allows solvent access to the PhyR receiver/SL interface (Fig. 5B and Movie S1).

Within the PhyR-SL domain, the regions of the α3-α4 loop and helix α4 are highly dynamic regions of structure (Fig. 5B). Indeed, among all the regions of PhyR, the α3-α4 loop and α4 have the highest crystallographic B-factors (11) and exhibit the largest structural shifts in our MD simulation. The hydrophobic and highly conserved face of this amphipathic helix (Fig. 4 B and C) is loosely docked against helices α1 and α3 of region σ₂ in the full-length PhyR structure. This same conserved helical face is in a different conformation in the PhyR-SL/NepR complex, where it is docked against helices α1 and α5 (Fig. 5A). This position of α4 in the complex structure is equivalent to what has been reported for orthologous α4 amphipathic helices of classic σ factors (Fig. 5C).

Highly Dynamic α3-α4 Region of PhyR Is Required for Stress-Dependent Regulation of Transcription.

Expression of the isolated PhyR-SL domain (PhyR_ΔRec) is known to sequester NepR, and thus constitutively derepress transcription of the general stress regulon in Methylobacterium extorquens (3). We have confirmed this result in C. crescentus; chromosomal replacement of WT phyR with the phyR_ΔRec allele produces stable protein that consititutively up-regulates transcription from a σ^T-dependent reporter (Fig. 6A). Using σ^T-dependent transcription as a proxy for PhyR-SL/NepR interaction in vivo, we attempted to assess the functional role of the dynamic α3-α4 region in NepR interaction. However, phyR_ΔRec alleles in which we truncated the α3-α4 coding sequence failed to produce stable protein in C. crescentus (Fig. S5), precluding functional analysis.

We next tested α3-α4 function in the context of full-length PhyR. We generated a set of C. crescentus phyR allelic replacement strains in which the α3-α4 loop and α4 coding sequence were removed (Fig. 6B). All these PhyR loop mutants produced soluble and stable protein in vivo (Fig. 6C). Deletion of the first five amino acids of the α3-α4 loop (LΔ5 = ΔQ70-G74) modestly reduced (∼20%) stress-dependent transcriptional activation from a σ^T-dependent reporter. A larger deletion of this loop (LΔ13 = ΔG68-H80) strongly attenuated (∼80%) transcription under sucrose stress. Deletion of the entire loop plus α4 (LΔ13 + Δα4 = ΔL71-R91) or deletion of α4 alone (Δα4 = ΔD84-I92) showed equivalent attenuation of stress-regulated transcription (Fig. 6C). From these data, we conclude that the full α3-α4 loop and helix α4 are required for PhyR to function as an anti–anti-σ^T regulator during stress in vivo.

Intact α3-α4 Region Is Required for NepR Binding to Full-Length PhyR∼P but Not to PhyR_ΔRec.

We next sought to test whether the transcriptional deficiencies of α3-α4 PhyR mutants were a result of a defect in PhyR∼P binding to NepR. Congruent with our observations in the transcription assays described above, a short loop deletion (LΔ5) did not significantly perturb binding of His-PhyR∼P to MBP-NepR in a column pull-down assay. However, deletion of the entire α3-α4 loop (LΔ13), the loop plus α4 (LΔ13 + Δα4), or α4 alone (Δα4) decreased NepR binding by 45–65% under the tested protein concentrations and buffer conditions (Fig. 6D and Materials and Methods). Surprisingly, although an intact α3-α4 region was required for full binding of PhyR∼P to MBP-NepR, we observed no differences in binding between MBP-NepR and His-PhyR_ΔRec or any of the α3-α4 mutant variants of His-PhyR_ΔRec (Fig. 6E). This provides evidence that the dynamic α3-α4 region is not required for high-affinity binding of isolated PhyR-SL to NepR. Rather, an intact α3-α4 region is an important determinant of NepR binding to PhyR-SL only when the PhyR receiver domain is present.

Association and Dissociation Kinetics of the PhyR∼P/NepR Binding Interaction.

Using surface plasmon resonance (SPR), we quantified the association (K_a) and dissociation (K_d) rate constants of full-length His-PhyR binding to MBP-NepR in the presence and absence of the phosphoryl donor, AcP. In the absence of AcP, we observed no binding between immobilized WT His-PhyR and MBP-NepR (across an MBP-NepR concentration range of 250 nM to 2 μM) (Fig. S6F). The addition of AcP to the flow buffer resulted in monophasic binding of MBP-NepR to immobilized His-PhyR(WT), with a calculated equilibrium affinity of 641 ± 92 nM (Table 2 and Fig. S6A). A WT PhyR mutant in which the receiver domain was deleted (His-PhyR_ΔRec) bound MBP-NepR with biphasic kinetics. However, the amplitude of the slow phase was minor (∼10–20%). We attribute this slow phase to a minority of PhyR_ΔRec dimers (which contain 2 His tags) interacting in a nonuniform way with the Ni²⁺ surface of the SPR chip. The major (fast) association and dissociation rates are comparable to WT His-PhyR binding to MBP-NepR. Specifically, we calculated that His-PhyR_ΔRec(WT) binds MBP-NepR with ∼5.5-fold higher affinity (117 ± 18 nM) than His-PhyR∼P. This difference in equilibrium affinity is almost entirely determined by a faster association rate constant (Table 2 and Fig. S6C).

Table 2.

K_a and K_d and calculated equilibrium affinities (K_D) of PhyR/NepR binding

MBP-NepR binding to	Ka, (1/M⋅s) × 105	Kd, (1/s) × 10−1	KD, nM
His-PhyR(WT)∼P	1.6 ± 0.1	1.1 ± 0.1	641 ± 92.6
His-PhyR(LΔ13 + Δα4)∼P	ND*	ND	ND
His-PhyRΔRec(WT)	8.0 ± 1.0	0.9 ± 0.1	117 ± 18.0
His-PhyRΔRec(LΔ13 + Δα4)	11.5 ± 1.0	2.2 ± 0.8	193 ± 55.6

*Binding parameters could not be determined (ND) at the assessed concentrations.

In accordance with the pull-down assays described above, complete deletion of the α3-α4 loop and helix α4 does not significantly affect the association and dissociation rate constants of MBP-NepR binding to isolated PhyR-SL [His-PhyR_ΔRec(LΔ13 + Δα4)] (Table 2 and Fig. S6D). However, the identical α3-α4/helix α4 deletion in full-length PhyR [His-PhyR(LΔ13 + Δα4)] dramatically decreases the affinity of its interaction with MBP-NepR in the presence of AcP. Although we observed very weak binding at the highest MBP-NepR concentration tested (2 μM), we were not able to determine a binding affinity from these data (Table 2 and Fig. S6B), because concentrations of MBP-NepR beyond 2 μM resulted in strong nonspecific binding to the SPR chip. These kinetic binding data provide additional evidence that the α3-α4 region of the PhyR-SL domain is required for efficient NepR binding to the full-length PhyR protein but not to the PhyR-SL domain alone.

Discussion

PhyR Dynamics and NepR Binding.

We have defined the molecular basis of binding between the SL domain of PhyR and the anti-σ factor, NepR. In its unphosphorylated state, the PhyR receiver domain occludes the NepR binding site on PhyR-SL (Fig. 3C). Structural and biochemical data and MD simulations provide support for a model in which phosphorylation of the PhyR receiver domain results in opening of the PhyR structure (i.e., dissociation of the receiver domain from the SL domain), exposing the NepR binding site on PhyR-SL. MD provides evidence that the α11-β5 region of the receiver domain may have a latch-like function, stabilizing the PhyR receiver/SL interaction in the unphosphorylated state. The α11-β5 fully retracts from contact with the PhyR-SL domain by 200 ns in our simulation, consistent with a model in which PhyR∼P is switching from a closed state to an open state. The result, that α11-β5 undergoes the largest conformational change in the receiver domain in our simulation, agrees with our initial hypothesis (11) that this region plays a key structural role in conformational switching. Concomitant with retraction of α11-β5, we observe that several water molecules infiltrate the binding interface between the PhyR receiver domain and PhyR-SL (Movie S1).

An analysis of different σ/anti-σ complex structures in the Protein Data Bank (PDB) reveals that interaction between a σ factor and its cognate anti-σ factor can occur in a number of different ways (23–25). A flexible linker between regions σ₂ and σ₄ has been described as important for anti-σ interaction but also as a general structural feature required for proper association of σ with RNAP and with -10 and -35 sequences in the promoter (16, 26–29). In the case of PhyR, the SL domain is missing critical sequence required for it to function as a bona fide σ factor (4) yet retains the flexible loop (α3-α4) between regions σ₂ and σ₄. Based on the full-length, unphosphorylated structure of PhyR, we initially proposed (11) that PhyR-SL would open about the α3-α4 loop and that NepR would bind between regions σ₂ and σ₄, much like σ^E and RseA (16). The structure of the complex reported here shows that although regions σ₂ and σ₄ of PhyR-SL are capable of separating about the flexible loop, PhyR-SL binds NepR in a closed conformation. However, we cannot exclude the possibility that opening of regions σ₂ and σ₄ is in some way important for the PhyR-SL/NepR binding process.

Binding Competition Between NepR and the PhyR Receiver Domain.

NepR and the PhyR receiver domain compete for the same binding surface on PhyR-SL. Thus, for this stress regulatory system to function properly, NepR must overcome binding competition from a high local concentration of receiver domain. Even when PhyR is phosphorylated, it is reasonable to presume that NepR must contend with some degree of receiver binding competition. We have presented both functional genetic and biochemical data that support a model in which the dynamic α3-α4 region of PhyR-SL is required for stable binding between PhyR∼P and NepR. We propose that the α3-α4 linker loop may function to reduce interaction between PhyR-SL and the phosphorylated PhyR receiver domain when PhyR is in its “active” form (i.e., when the PhyR receiver domain is not bound to PhyR-SL). The length of the α3-α4 loop is clearly important in controlling the affinity of NepR for full-length PhyR∼P and for activation of σ^T-dependent transcription. Although a short (5 residues) truncation of the loop is tolerated, excision of longer pieces of α3-α4 reduces PhyR∼P binding to NepR and attenuates transcriptional activation on stress insult. These results suggest that a conformational rearrangement requiring the long, dynamic α3-α4 loop is important for regulated binding between PhyR∼P and NepR.

α4 is directly tethered to the α3-α4 loop and is the most inherently dynamic region of secondary structure in PhyR, exhibiting substantial conformational change in our simulations at early time points (Movie S1). This provides evidence that α4 is held in a relatively high-energy configuration within the full-length, unphosphorylated PhyR structure. In the context of the PhyR-SL/NepR complex, α4 occupies an entirely different position, where it is packed against the PhyR-SL surface that interacts with the receiver domain. This structural orientation of α4 more closely resembles what has been reported in other σ structures (Fig. 5C). PhyR phosphorylation, and subsequent disruption of the SL/receiver interface, may free α4 to shift to this position. We propose that this configuration of α4 would stabilize the PhyR∼P/NepR complex by obstructing the receiver domain from interacting with PhyR-SL (Fig. 7).

Fig. 7.

Molecular model of regulated PhyR-NepR binding and σ-dependent transcription in C. crescentus. Under stress conditions a phosphoryl group on PhyK histidine kinase is transferred to the receiver domain (red) of PhyR, inducing structural changes that are transduced to a surface interaction loop (α11-β5) between the receiver domain and the SL domain (green) (1 and 2). Destabilization of the PhyR receiver/SL interface reveals the NepR binding surface; the dynamic α3-α4 loop and helix α4 (yellow) undergo a conformational change that helps to stabilize PhyR in the open state (3). In this open conformation, the PhyR-SL domain is able to bind NepR (blue) stably (4). σ^T (peach) is subsequently freed to bind RNAP (purple) and activate transcription of stress response genes (5).

Materials and Methods

Production of Recombinant PhyR Proteins.

Heterologous expression of WT and mutant variants of C. crescentus PhyR and NepR was carried out in E. coli Rosetta (DE3) pLysS (Novagen). Protein was expressed from genes cloned into pET28c (Novagen), pETDuet-1 (Novagen), or pMALc2g (New England Biolabs) (strain, plasmid, and primer information is provided in Tables S1 and S2). Three classes of expression strains were obtained: pETDuet-1 strains coexpressing His-tagged PhyR or PhyR_ΔRec with the NepR protein (growth in LB + ampicillin, 100 μg/mL), pET28c strains overexpressing His-tagged PhyR or PhyR_ΔRec alone (growth in LB + kanamycin, 50 μg/mL), and a pMALc2g strain expressing an MBP-NepR fusion protein (growth in LB + ampicillin, 100 μg/mL).

Protein Expression and Purification.

Liquid cultures for expression of recombinant WT and mutant PhyR and NepR were induced at an OD₆₆₀ of 0.8 (37 °C, 220 rpm) by adding 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG; Gold Biotechnology). After 3 h, cells were harvested by centrifugation at 8,000 rpm for 20 min at 4 °C. Cell pellets were resuspended in 10 mM Tris⋅HCl (pH 7.4), 150 mM NaCl, and 10 mM imidazole (Fisher Scientific) with 5 μg/mL DNase I and 80 μg/mL PMSF (Sigma–Aldrich), and disrupted by two passages in a French pressure cell; the cell debris was removed by centrifugation at 14,000 rpm for 20 min at 4 °C.

For protein purified by nickel affinity chromatography (GE Healthcare), after loading the clarified lysate on a preequilbrated column, three washing steps were performed using 10 mM, 30 mM, and 75 mM imidazole Tris⋅NaCl buffers followed by elution with 500 mM imidazole Tris⋅NaCl buffer. The protein solution was then dialyzed against 10 mM Tris⋅HCl (pH 7.4) and 150 mM NaCl buffer to remove imidazole.

For purification of MBP-NepR fusion protein, an amylose resin column (New England Biolabs) was first equilibrated with 20 mM Tris (pH 7.4) and 200 mM NaCl. Cell lysate was loaded and washed five times with three column volumes of equilibration buffer. Protein was eluted with equilibration buffer supplemented with 10 mM maltose.

When necessary, purified proteins were concentrated using a centrifugal filter [3 kDa molecular weight cutoff (MWCO); Amicon–Millipore]. All purification steps were carried out at 4 °C. The protein purity was assessed by resolving the different fractions by 14% (wt/vol) SDS/PAGE gels.

Crystallization of the PhyR_ΔRec(LΔ13)/NepR Complex.

Multiple attempts to crystallize NepR complexed with the WT SL domain of PhyR failed. We postulated that the disordered α3-α4 loop in the SL domain might perturb crystal packing. As such, we coexpressed NepR with a mutant of the PhyR-SL domain that is missing residues 68–80 [PhyR_ΔRec(LΔ13)]; these residues have been defined previously as the disordered α3-α4 loop in the full-length PhyR structure (11). Protein was expressed and purified as described above. All crystallization attempts were carried out using the hanging-drop, vapor-diffusion technique. The protein concentration was 30 mg/mL. Initial crystallization screening was carried out in 96-well microplates (Nunc). Trays were set using a Mosquito robot (TTP LabTech) and commercial crystallization kits (Nextal–Qiagen). The drops were set up by mixing equal volumes (0.1 mL) of the protein and the precipitant solutions equilibrated against 75 mL of the precipitant solution. After manual refinement (in 24-well plates; Hampton Research), the best crystals were obtained at 14 °C with the following crystallization solution: 100 mM 2-(N-morpholino)ethanesulfonic acid (MES) (pH 6.0), 20% PEG 2000 MME, and 200 mM NaCl. The drops were set up by mixing 4 μL of protein and 1 μL of the precipitant solutions equilibrated against 500 μL of precipitant. Crystals grew to their final size in 10 to 15 d and were soaked for 1 min in a precipitant solution containing 5 mM β-mercaptoethanol and 20% glycerol (Fisher) as a cryoprotectant before flash-freezing in a cryoloop (Hampton Research) (30). To produce selenomethionine (SeMet; Sigma–Aldrich) protein for experimental phase determination, proteins were expressed in defined medium as previously described (31).

Crystallographic Data Collection and Processing.

Crystal diffraction data were collected at a temperature of 100 K on beamline 21-ID-D (Life Sciences Collaborative Access Team, Advanced Photon Source) using a MAR Mosaic 300 detector and an oscillation range of 1°. Diffraction images were reduced using the HKL 2000 suite (32). Diffraction data revealed that the crystals belonged to the orthorhombic space group C222₁, with cell dimensions a = 75.4 Å, b = 105.3 Å, and c = 97.9 Å. Diffraction from these crystals was moderately anisotropic (D_min = 2.1 Å along b* and c* and 2.7 Å along a*, where Dmin is the diffraction limit of the crystal). The reflection set was ellipsoidally truncated and anisotropically scaled before refinement using the method of Strong and colleagues (33).

Phasing and Refinement.

Diffraction from a single native crystal of the PhyR_ΔRec(LΔ13)/NepR protein complex containing SeMet was measured at an energy of 12.66 keV (0.979 Å) and phased by single-wavelength anomalous dispersion (34). Six selenium sites were located within the asymmetric unit using the Autosol SAD routine in PHENIX (35). Two PhyR_ΔRec(LΔ13)/NepR complexes were present per asymmetric unit. The initial PhyR_ΔRec(LΔ13)/NepR structure was built de novo from these experimental maps using the PHENIX AutoBuild routine. Manual model building, solvent addition, and refinement of this initial structure were conducted iteratively using Coot (36) and phenix.refine (Table 1). The structure was refined to a final R_cryst of 20.0% and R_free of 25.1%; the Rcryst and Rfree residuals are defined in the legend of Table 1. Coordinates of C. crescentus PhyR_ΔRec(LΔ13)/NepR have been deposited in the Protein Data Bank (PDB ID code 3T0Y).

Gel Filtration Chromatography.

Protein oligomeric state was assessed using gel filtration chromatography. Three hundred microliters of purified His-PhyR(WT) (10 mg/mL, ∼300 μM), His-PhyR(WT)/NepR (10 mg/mL, ∼300 μM; purification steps carried out in the presence of 25 mM AcP and 5 mM Mg²⁺), His-PhyR_ΔRec(WT)/NepR (5 mg/mL, ∼100 μM), and His-PhyR_ΔRec(LΔ13)/NepR (5 mg/mL, ∼100 μM) complexes was loaded onto a Superdex 75 HR 10/30 gel filtration column (GE Healthcare). Running buffer was 10 mM Tris (pH 7.4) and 150 mM NaCl. Buffer to measure elution volume of PhyR(WT)∼P protein and PhyR(WT)∼P/NepR complex was supplemented with 25 mM AcP and 5 mM MgCl₂, which has previously been reported to catalyze phosphorylation of C. crescentus PhyR (13). Gel filtration experiments were also carried out on purified His-PhyR_ΔRec(WT) and His-PhyR_ΔRec(LΔ13) proteins in the absence of bound NepR. Because these truncated variants of PhyR were unstable during the concentration steps and gel filtration, 10 mM Tris (pH 7.4) and 150 mM NaCl buffer was supplemented with 200 mM imidazole, which improved solubility. Three hundred microliters of each recombinant protein (10 mg/mL, ∼600 μM) was loaded on the gel filtration column. The protein composition of the column fractions was assessed by 14% SDS/PAGE gel.

SAXS.

SAXS data were collected at Advanced Photon Source beamline 18-ID (Biophysics Collaborative Access Team). Unphosphorylated PhyR in 10 mM Tris (pH 7.5) and 150 mM NaCl was suspended in a 1-mm capillary at a final concentration of ∼20 μM, and SAXS was measured from this sample. Scattering from PhyR under phosphorylating conditions [10 mM Tris (pH 7.5), 150 mM NaCl, 5 mM MgCl₂, 25 mM AcP) was measured at a final concentration of ∼8 μM. Scattering data were recorded on an Aviex CCD detector, and data analysis was conducted using a custom SAXS analysis plug-in implemented in Igor Pro (WaveMetrics).

Engineering Allelic Replacement Strains.

C. crescentus CB15 strains in which the WT phyR allele was replaced with different phyR mutant alleles were built using a double-recombination gene replacement strategy (37). pNPTS138 carries the nptI gene to select for single integrants on kanamycin and the sacB gene for counterselection on sucrose. Transformation of C. crescentus and sucrose counterselection for allelic replacement were carried out as described previously (38). PCR and Sanger sequencing of the PCR product confirmed allele replacement. Primer, plasmid, and strain information is provided in Tables S1 and S2.

Stress Response Transcriptional Assays.

It is known that transcription of sigU is up-regulated by the general stress σ factor σ^T on osmotic or oxidative stress insult (5). The plasmid pRKLac290-PsigU, which contains the sigU promoter transcriptionally fused to lacZ, was conjugated into WT C. crescentus CB15 and strains in which the WT phyR allele was replaced with full-length phyR variants (R15A-R16A, LΔ5, LΔ13, LΔ13 + Δα4, and Δα4) and phyR_ΔRec variants (WT, R15A-R16A, LΔ5, LΔ13, LΔ13 + Δα4, and Δα4). Each mutant was evaluated for its ability to activate transcription from the PsigU-lacZ reporter fusion.

All strains were grown in peptone/yeast extract (PYE) medium (+1 μg/mL tetracycline) at a starting OD₆₆₀ of 0.05 (30 °C, 220 rpm) and stressed by adding 150 mM sucrose to the culture medium (5) at the beginning of the experiment. At OD₆₆₀ ≈ 0.25, β-galactosidase activities were measured in triplicate as previously described (39). Stability of mutant proteins was assessed by Western blot analysis.

Western Blot Analysis of WT and Mutant PhyR Proteins.

The effect of deleting different phyR regions (receiver domain, loop, and α4) on PhyR protein stability was assessed by Western blot analysis. phyR and mutant variants thereof were cloned into pMT585 (40) containing the sequence for a human influenza HA epitope tag at the 5′ end of the multiple cloning site. Inserts were obtained by NdeI/EcoRI digestion of pCR-BLUNT II-TOPO plasmids (Invitrogen) carrying the different versions of phyR and directly ligated in pMT585-HA plasmid. These clones yielded xylose-inducible, N-terminal HA protein fusions (primers, plasmid, and strain information is provided in Tables S1 and S2). The different pMT585-HA-phyR fusion plasmids were conjugated into a C. crescentus CB15 ΔphyR background from an E. coli TOP10 donor strain.

Western blots were performed as follows. After 5 h of induction with 0.15% d-xylose (Fisher) (30 °C, shaken at 220 rpm, initial OD₆₆₀ = 0.05, final OD₆₆₀ = 0.5), 50 mL of PYE (+5 μg/mL kanamycin) culture was pelleted (8,000 rpm for 10 min) and resuspended in 1.5 mL of 10 mM Tris (pH 7.4) and 150 mM NaCl buffer. Each sample was French-pressed and centrifuged at 14,000 rpm for 20 min to remove cell debris. Ten microliters of supernatant was mixed with 3 μL of SDS loading buffer, incubated at 95 °C for 5 min, resolved on a 14% SDS/PAGE gel, and blotted onto a 0.2-μM PVDF membrane (Millipore). Blotting and antibody incubation were performed as previously described (38). Briefly, after transfer and blocking steps, the membrane was incubated with a mix of mouse monoclonal anti-HA (Sigma–Aldrich) and rabbit polyclonal anti-FixJ (loading control) primary antibodies. Anti-mouse and anti-rabbit HRP secondary antibodies (ThermoFisher Scientific) were used for detection. Western blots were developed with SuperSignal West Femto chemiluminescent substrate (ThermoFisher Scientific).

PhyR/NepR Copurification/Pull-Down Assays.

MBP-NepR from the clarified lysate of a 50-mL culture pellet (Protein Expression and Purification) was bound to 600 μL of amylose resin. A 50-mL wash of the beads was performed with 20 mM Tris (pH 7.4) and 200 mM NaCl buffer. One hundred microliters of MBP-NepR beads was mixed with 500 μL of His-PhyR(WT) purified protein (concentration of ∼50 μM). This same protocol was applied to the His-PhyR “loop” mutant proteins [PhyR(LΔ5), PhyR(Δα4), PhyR(LΔ13), and PhyR(LΔ13 + Δα4)]. In each sample, 1 mL of 20 mM Tris (pH 7.4) and 200 mM NaCl buffer supplemented with 5 mM MgCl₂ and 25 mM AcP was added to induce phosphorylation of PhyR. After 1 h of incubation at room temperature, beads from each sample were individually washed with 15 mL of buffer three times. For the elution of the different MBP-NepR/His-PhyR complexes, 100 μL of 20 mM Tris (pH 7.4) and 200 mM NaCl buffer, plus 10 mM maltose, was added to the beads. Fifteen microliters of each sample was loaded on a 14% SDS/PAGE gel, and spots corresponding to MBP-NepR and His-PhyR proteins were analyzed by ImageJ (National Institutes of Health) (41). The quantity of PhyR protein and mutant variants thereof was normalized to the amount of MBP-NepR on the gel. The same protocol was applied to study the binding of MBP-NepR to the His-PhyR_ΔRec(WT) and the His-PhyR_ΔRec(R15A-R16A) proteins.

The reciprocal copurification experiment was conducted by coexpressing the WT and His-PhyR_ΔRec variants with NepR from pETDuet-1. Each complex was purified from a pellet from 50 mL of culture (Protein Expression and Purification). After nickel affinity immobilization with 100 μL of Ni²⁺ agarose resin and a 50-mL wash of each complex with 10 mM Tris (pH 7.4), 150 mM NaCl, and 10 mM imidazole buffer, each sample was eluted with 50 μL of 200 mM imidazole Tris⋅NaCl buffer and resolved by 14% SDS/PAGE gel for analysis (loading volume of 15 μL). Spots corresponding to NepR protein and to the different PhyR_ΔRec proteins were analyzed by ImageJ. The quantity of NepR protein that was bound to affinity-immobilized PhyR-SL proteins was normalized to the amount of His-PhyR_ΔRec on the gel. All copurification/pull-down experiments were conducted in triplicate.

SPR Binding Assays.

The kinetics of WT and mutant PhyR binding to NepR were measured using a Bio-Rad ProteOn XPR six-channel SPR instrument. Preparation of the proteins for SPR assays was performed as follows: 50 mL of culture containing E. coli Rosetta (DE3) pLysS expressing either His-PhyR (WT or LΔ13 + Δα4) or His-PhyR_ΔRec (WT or LΔ13 + Δα4) was grown (37 °C, 220 rpm) to an OD₆₆₀ of 0.5 and induced for 4 h with 1 mM IPTG. Pellets were resuspended in 1.5 mL of 50 mM Tris (pH 7.4), 150 mM NaCl, and 5 mM MgCl₂ buffer and disrupted two times by French press. After 10 min of centrifugation at 14,000 rpm, the supernatant was collected and filtered with a 0.22-μm filter (ThermoFisher Scientific). Each crude extract was diluted at a ratio of 1:4,000 and loaded onto a nitrilotriacetic acid Bio-Rad sensor chip and washed for 5 min with the previous buffer supplemented with 0.05% Tween 20.

MBP-NepR protein was purified by amylose affinity chromatography as described above. Different concentrations of MBP-NepR were tested to identify the ideal response range of the SPR chip, which was loaded with the different His-PhyR variants. Protein interaction between MBP-NepR and His-PhyR (WT or LΔ13 + Δα4) proteins was assayed under both phosphorylating conditions [using 50 mM Tris (pH 7.4), 150 mM NaCl, 5 mM MgCl₂, 5 mM AcP, and 0.05% Tween 20 buffer] and nonphosphorylating conditions [50 mM Tris (pH 7.4), 150 mM NaCl, 5 mM MgCl₂, and 0.05% Tween 20 buffer]. All protein dilutions were carried out in the buffer condition being tested.

As a control, we confirmed that the MBP tag alone is not able to interact with His-PhyR (WT or LΔ13 + Δα4) or His-PhyR_ΔRec (WT or LΔ13 + Δα4) proteins, using the same protocol as above (Fig. S6 E and F).

PhyR/NepR and PhyR_ΔRec/NepR raw binding data were analyzed in the ProteOn software suite using the kinetic-Langmuir and the kinetic-heterogeneous ligand data analysis options, respectively. All SPR assays were conducted in triplicate.

MD Simulations.

The crystal structure of full-length PhyR (PDB ID code 3N0R) was solvated in an aqueous KCl ionic solution to create a neutral system. A total of 49,493 atoms are in the model. After solvation, energy minimization was carried out in VMD (44) to remove any local strain and bad contacts. To prepare the system for MD, a predynamic equilibration of 1,000 steps was implemented to bring the system to a pressure of 1 bar and a temperature of 300 K. A simulation of 10 ns with the unphosphorylated protein was carried out to validate the methodology and ensure the stability of the system. A phosphate group was added to the structure on residue D192, and the phosphorylated protein was simulated for 320 ns. All simulations were carried out using the NAMD 2.7 b2 scalable MD software (42) with the all-atom CHARMM force field PARAM22 + CMAP (43). The temperature in the system was set to 300 K and maintained using Langevin dynamics on all nonhydrogen atoms; the damping coefficient was set to 0.5 ps⁻¹. The simulation was run with periodic boundary conditions. The particle mesh Ewald method was activated to calculate full system electrostatics at a grid point density of 1.5/Å. Pressure in the system was maintained with a Nosé–Hoover piston at every 200-fs window; the decay time scale was set to 100 fs. The trajectory was propagated with a multiple time-step scheme of 1 fs; bonded forces were evaluated at every step, nonbonded forces were evaluated every two steps, and long-range electrostatics were evaluated every four steps. The real-space, short-range electrostatics and van der Waals Lennard–Jones interactions were smoothly switched off in the interval from 10–12 Å. All calculations were preformed on the TeraGrid systems NCSA Abe and TACC Ranger. Images and trajectories were analyzed using VMD (44), PyMol (45), and RasMol (46). Original Tcl and MATLAB scripts were written for atomic distance calculations and contour maps.

Sequence Alignment and Protein Visualization Methods.

Protein sequence alignments were carried out in Clustal W2 (www.ebi.ac.uk/clustalw/). PhyR ribbon structure rendering, electrostatic potential surfaces calculation (using the generate-vacuum electrostatics function), and visualization of the hydrophobic surfaces (according to the Eisenberg hydrophobicity scale (47) were performed with PyMOL. Visualization of the conserved residues in the different structures was performed using the Consurf server (48). PyMol and PDBe protein interfaces, surfaces, and assemblies (PISA) have been used to define the residue interaction map between NepR and PhyR_ΔRec(LΔ13).