Structural basis for sigma factor mimicry in the general stress response of Alphaproteobacteria

Abstract

Reprogramming gene expression is an essential component of adaptation to changing environmental conditions. In bacteria, a widespread mechanism involves alternative sigma factors that redirect transcription toward specific regulons. The activity of sigma factors is often regulated through sequestration by cognate anti-sigma factors; however, for most systems, it is not known how the activity of the anti-sigma factor is controlled to release the sigma factor. Recently, the general stress response sigma factor in Alphaproteobacteria, σ^EcfG, was identified. σ^EcfG is inactivated by the anti-sigma factor NepR, which is itself regulated by the response regulator PhyR. This key regulator sequesters NepR upon phosphorylation of its PhyR receiver domain via its σ^EcfG sigma factor-like output domain (PhyR_SL). To understand the molecular basis of the PhyR-mediated partner-switching mechanism, we solved the structure of the PhyR_SL–NepR complex using NMR. The complex reveals an unprecedented anti-sigma factor binding mode: upon PhyR_SL binding, NepR forms two helices that extend over the surface of the PhyR_SL subdomains. Homology modeling and comparative analysis of NepR, PhyR_SL, and σ^EcfG mutants indicate that NepR contacts both proteins with the same determinants, showing sigma factor mimicry at the atomic level. A lower density of hydrophobic interactions, together with the absence of specific polar contacts in the σ^EcfG–NepR complex model, is consistent with the higher affinity of NepR for PhyR compared with σ^EcfG. Finally, by reconstituting the partner switch in vitro, we demonstrate that the difference in affinity of NepR for its partners is sufficient for the switch to occur.

Regulation of transcription initiation is the primary strategy used by bacteria to reprogram gene expression to adjust to changing internal and environmental conditions (1). Alternative sigma factors, the RNA polymerase subunits that confer promoter specificity, provide a powerful way to redirect gene expression globally toward specific regulons (2). The largest and most diverse group of alternative sigma factors comprises the group IV (or extracytoplasmic function [ECF]) sigma factors of the σ⁷⁰ superfamily, which are best known for their role in adaptation to environmental changes (3, 4). In contrast to other sigma factors, ECF sigma factors consist only of the two core domains, σ₂ and σ₄, which are required for recognition of the -10 and -35 promoter binding determinants, respectively, in addition to RNA polymerase binding.

ECF sigma factors are often regulated through sequestration by cognate anti-sigma factors, which interfere with their ability to bind the RNA polymerase core enzyme or promoter regions by occluding their respective molecular determinants or by altering their conformations (5–7). The activity of anti-sigma factors must itself be regulated, and this is achieved by a variety of mechanisms, such as direct stimulus sensing followed by conformational changes, regulated intramembrane proteolysis by sensor proteases, or partner switches with anti-sigma factor antagonists (6, 8–11). Although there is considerable understanding of anti-sigma factor regulation for a few systems, this is not the case for the vast majority of systems.

We recently discovered an original anti-sigma factor antagonist, PhyR, involved in a partner-switching mechanism governing the general stress response in Alphaproteobacteria (12–14). PhyR is a member of the response regulator family and is unique in that its output domain shows homology to ECF sigma factors (14, 15). While the PhyR receiver domain (PhyR_REC) has a classic receiver fold with the conserved phosphorylation site and catalytical motifs, the sigma factor-like output domain (PhyR_SL) essentially retains an ECF sigma factor fold but lacks the σ_2.4 region and is degenerate in the σ_4.2 region, which are normally involved in promoter binding in bona fide ECF sigma factors (12, 16). According to the proposed model (12), under unstressed conditions, the sigma factor σ^EcfG is sequestered by its anti-sigma factor, NepR, and the anti-sigma factor antagonist, PhyR, remains in an unphosphorylated inactive state. In response to stress, PhyR becomes phosphorylated and binds to NepR via its sigma factor-like output domain, thereby releasing σ^EcfG from the inhibitory NepR interaction. The σ^EcfG then redirects gene expression toward the general stress response. Because PhyR uses a degenerate sigma factor-like domain to bind to the anti-sigma factor NepR, this partner-switching mechanism was coined “sigma factor mimicry.” Although this term applies at the functional level, no structural data of how PhyR achieves binding to the sigma factor antagonist have been available.

The crystal structure of the unphosphorylated form of PhyR from Caulobacter crescentus shows that the two domains of PhyR, PhyR_REC and PhyR_SL, interact through a polar interface that buries one face of the PhyR_SL domain and the α4-β5-α5 face of the receiver domain (16), a region known to undergo conformational changes on phosphorylation of the receiver domain in other response regulators (17–19). It was proposed that the activation of PhyR releases the PhyR_SL domain to bind the anti-sigma factor NepR (12); however, the mode of binding of NepR to PhyR_SL remains unknown. In the present study, the NMR solution structure of the PhyR_SL–NepR complex of Sphingomonas sp. Fr1 was solved and gives detailed insight into the molecular interactions mediating the complex. The data support the concept of sigma factor mimicry at the atomic level and reveal the basis underlying the partner switch and an unprecedented anti-sigma factor binding mode.

Results

PhyR_SL Domain Contains Determinants of NepR Binding.

In Sphingomonas sp. Fr1, the PhyR-NepR-σ^EcfG signaling cascade controls the general stress response via a partner-switching mechanism that is general to Alphaproteobacteria (12, 13, 20–24). The PhyR_SL domain of Sphingomonas sp. Fr1 was previously shown to be sufficient for binding NepR (20), in agreement with results obtained from Methylobacterium extorquens (12). To assess quantitatively whether the determinants of NepR recognition indeed reside exclusively in PhyR_SL, the affinities of NepR for activated PhyR and for the isolated PhyR_SL domain were compared using isothermal titration calorimetry (ITC). The phosphoryl analog beryllium fluoride (BeF₃) (18, 25, 26) was used to activate PhyR. PhyR-BeF₃ and PhyR_SL were found to bind NepR in a 1:1 stoichiometry with dissociation constants of 31.3 ± 2.7 nM and 5.6 ± 1.0 nM, respectively (Fig. 1 A and B). The higher affinity of NepR for PhyR_SL compared with PhyR-BeF₃ strongly suggests that the determinants of the anti-sigma factor recognition are solely localized in the PhyR_SL domain. Hence, we investigated the structure of the Sphingomonas sp. Fr1 PhyR_SL–NepR complex using liquid state NMR. To distinguish between NepR and PhyR_SL, NepR secondary structures and residues will be indicated with a prime symbol (′) in the following.

Fig. 1.

Biophysical characterization of the interaction between NepR and PhyR_SL. K_D determination of the interactions between NepR and PhyR-BeF₃ (A) or PhyR_SL (B). (Upper) Heat changes observed upon injection of 7 μL of a 100-μM NepR solution into a 10-μM solution of PhyR-BeF₃ or PhyR_SL. (Lower) Integrated heat changes following each injection plotted against the molar ratio of NepR to PhyR or PhyR_SL. (C) 2D ¹⁵N-¹H HSQC spectrum of the complex formed by ¹⁵N-labeled PhyR_SL and unlabeled NepR. 2D ¹⁵N-¹H HSQC spectrum of ¹⁵N-labeled NepR alone (D) and in complex with unlabeled PhyR_SL (E). (F) {¹H}-¹⁵N heteronuclear NOE values of ¹⁵N-labeled NepR in complex with unlabeled PhyR, I_rel, vs. the residue position. The location of NepR α-helices is indicated in the Upper part of the figure.

Structure Determination of the PhyR_SL–NepR Complex.

PhyR_SL alone had a tendency to aggregate over time at millimolar concentrations, whereas it showed increased solubility and well-dispersed NMR spectra in complex with NepR (Fig. 1C). NepR alone was disordered in solution, as indicated by poor chemical shift dispersion in 2D ¹⁵N-¹H heteronuclear single quantum coherence (HSQC) (Fig. 1D) and negative {¹H}-¹⁵N heteronuclear NOE values; however, its C-terminal part (residues 28′–56′) undergoes a disorder-order transition upon complex formation with PhyR_SL, as indicated by a well-dispersed 2D ¹⁵N-¹H HSQC spectrum and positive {¹H}-¹⁵N heteronuclear NOE values (Fig. 1 E and F and Fig. S1). The structure of the PhyR_SL–NepR complex was calculated using 2,547 NOE-derived distance constraints, including 205 intermolecular NOEs (Table 1). The ensemble of 15 conformers of the complex was superimposed with a backbone rmsd of 1.10 ± 0.18 Å for the structured regions (PhyR_SL 1–139 and NepR 28′–56′; Fig. 2).

Table 1.

NMR experimental restraints and structural statistics

	PhyRSL	NepR
NMR distance and dihedral constraints
Distance constraints
Total NOE	1,909	433
Intraresidue	552	146
Interresidue	1,357	287
Sequential (|i − j| = 1)	556	167
Nonsequential (|i − j| > 1)	801	120
Intermolecular	205
Hydrogen bonds	68	13
Total dihedral angle restraints	193	48
ϕ	98	20
ψ	95	18
Structure statistics
Violations (mean ± SD)
Distance constraints, Å	1.06 ± 0.77
Dihedral angle constraints, °	1.13 ± 1.08
Maximum dihedral angle violation, °	6.32 ± 2.64
Maximum distance constraint violation, Å	0.33 ± 0.03
Deviations from idealized geometry
Bond lengths, Å	0.004
Bond angles, °	1.56
Average pairwise rmsd*, Å
PhyRSL
Heavy	1.73 ± 0.22
Backbone	1.12 ± 0.21
NepR
Heavy	1.57 ± 0.25
Backbone	0.93 ± 0.20
Complex
Heavy	1.69 ± 0.19
Backbone	1.10 ± 0.18
Ramachandran analysis
Most favored region	67.8%
Allowed region	32.0%
Disallowed region	0.2%

*Pairwise rmsd was calculated among 15 refined structures using residues 3–139 for PhyR_SL and residues 28′–56′ for NepR.

Fig. 2.

NMR solution structure of the PhyR_SL–NepR complex. (A) Amino acid sequences and secondary structure predictions based on the backbone chemical shifts of Sphingomonas sp. Fr1 PhyR_SL and NepR. The α-helices are indicated by cylinders. For PhyR_SL, the σ₂ subdomain is yellow, the linker σ₂–σ₄ is green, and the σ₄ subdomain is gray. NepR is shown in blue. Functional subregions defined by comparison with ECF sigma factors are indicated by solid lines, except for subregion σ_2.4, which is indicated with a dashed line because it is incomplete in PhyR. (B) Ribbon representation of the 15 NMR structures of the PhyR_SL–NepR complex (PDB ID code 2LFW). Flexible regions of NepR (residues 1′–27′ and 57′–62′´) and of PhyR_SL (residues 140–157) are not illustrated. (C) Cartoon representation of the lowest energy model of the ensemble. The coloring and labeling scheme are the same as in A.

Structure of the PhyR_SL–NepR Complex.

PhyR_SL forms an all-helical structure consisting of two subdomains, termed σ₂ (residues 1–70) and σ₄ (residues 97–139) following the nomenclature of the σ⁷⁰ superfamily sigma factors (27). Both subdomains are composed of three helices, α1–α3 for σ₂ and α5–α7 for σ₄, and are separated by a linker that includes an additional helix (α4; Fig. 2). Superimposition of the individual subdomains σ₂ and σ₄ of PhyR_SL with the corresponding regions of a prototypical ECF sigma factor, σ^E [Protein Data Bank (PDB) ID code 1OR7] (28) showed that the PhyR_SL subdomains essentially retain the typical fold of ECF sigma factors, with rmsds of 3.12 Å and 0.80 Å, respectively (Fig. S2), except for helix α4 of bona fide ECF sigma factors, which is absent in PhyR orthologs (16) (Fig. S3).

The C-terminal half of NepR forms two helices, α1′ (residues 28′–40′) and α2′ (residues 46′–55′), which are linked by an ordered α1′-α2′ loop, whereas the N-terminal half (residues 1′–27′) is disordered. The helix–loop–helix structure of NepR forms a wide L-shaped clamp that extends along one side of PhyR_SL and contacts both the σ₂ and σ₄ subdomains as well as the linker helix α4 (Fig. 2). The N-terminal half of NepR does not seem to participate in PhyR_SL binding, consistent with its high variability both in length and sequence among NepR orthologs (Fig. S4).

Structural Comparison Between the PhyR_SL–NepR Complex and Unphosphorylated PhyR.

To assess the structural rearrangements in PhyR required for NepR binding, the PhyR_SL–NepR complex and unphosphorylated PhyR from C. crescentus (12) were compared. The conformation of the individual subdomains σ₂ and σ₄ are similar in the two structures: Cα coordinates of subdomains σ₂ and σ₄ in the PhyR_SL–NepR complex can be individually superimposed on PhyR σ₂ and σ₄ Cα coordinates with rmsds of 1.15 Å and 0.98 Å, respectively (Fig. S2). However, in the PhyR_SL–NepR complex, both subdomains are reoriented relative to each other by a rigid body rotation that allows PhyR_SL to generate a continuous hydrophobic interface along σ₂ and σ₄ to which NepR helix α1′ binds (Fig. 3A). Whereas the σ₂–σ₄ linker was described as a dynamic hinge in unphosphorylated PhyR (16), it undergoes an outward movement and positions helix α4 to allow its participation in NepR helix α1′ recognition in the complex. In the PhyR_SL–NepR complex, the position of helix α4 is not compatible with the position of the receiver domain in unphosphorylated PhyR and generates a steric clash with α12 (Fig. 3C). In addition, the surface of PhyR that binds NepR helix α1′ in the PhyR_SL–NepR complex interacts in unphosphorylated PhyR with the α11-β5-α12 surface of the receiver domain (Fig. 3B), the region known to undergo an allosteric transition on receiver domain phosphorylation (17–19). To test whether contacts in the α11-β5-α12 region of PhyR_REC are major determinants of the inhibitory interactions in PhyR, residue E235 in the α11-β5 loop of PhyR_REC, which makes polar contacts with the side chains of residues R19 and Q25 in the PhyR_SL domain, was substituted with alanine in a nonphosphorylatable PhyR allele, PhyR_D194A (20). The mutant protein PhyR_D194A,E235A formed a complex with NepR in the absence of BeF₃ as observed by analytical size exclusion chromatography (Fig. S5). Hence, a conformational change of the α11-β5-α12 surface of the receiver domain on phosphorylation is likely to promote the destabilization of the interaction between PhyR_SL and PhyR_REC, thus making the PhyR_SL interface accessible for binding to NepR helix α1′.

Fig. 3.

Structural comparison between PhyR_SL–NepR complex and unphosphorylated PhyR. (A) Stereoview of the superimposition of the structures of unphosphorylated PhyR (PDB ID code 3N0R) and of the NepR-bound conformation of PhyR_SL (PDB ID code 2LFW). The Cα coordinates for σ₂ were aligned for best fit to reveal differences in the relative orientation of σ₂ and σ₄ in the two structures. The α-helices are represented by cylinders. Structural alignment of the Cα coordinates of σ₂ shows that the σ₄ subdomain in the complex is rotated relative to its orientation in unphosphorylated PhyR by 13° counterclockwise in the plane x,y and by 7° clockwise in the plane x,z with respect to σ_2. The red arrow illustrates the movement of helix α4 on NepR binding. (B) Stereoview of the superposition of the structures of unphosphorylated PhyR and of the PhyR_SL–NepR complex. Only the PhyR_REC domain of PhyR and NepR helix α1′ are displayed. NepR helix α1′ overlaps with the region α11-β5-α12 of PhyR_REC in unphosphorylated PhyR, which is known to undergo an allosteric transition upon receiver domain phosphorylation. Residues important for the allosteric transition are shown as a stick model and are red (T220 and F236). (C) Stereoview of the superimposition of the structures of unphosphorylated PhyR and of the NepR-bound conformation of PhyR_SL. Helix α4 in the PhyR_SL–NepR complex occupies the space of helix α12 from PhyR_REC in unphosphorylated PhyR. In all the panels, the unphosphorylated PhyR crystal structure is cyan, and for PhyR_SL in the complex with NepR, σ₂ is yellow, σ₄ is gray, the σ₂-σ₄ linker is green, and NepR is blue.

Molecular Contacts at the PhyR_SL–NepR Interface.

Two major interfaces are observed in the NepR–PhyR_SL complex, which are largely defined by helices α1′ and α2′ of NepR. The vast majority of contacts are hydrophobic interactions.

At interface 1, NepR helix α1′ folds into an amphipathic cylinder whose hydrophobic face contacts a hydrophobic pocket formed by PhyR_SL helix α5 (corresponding to region σ_4.1 by analogy to bona fide ECF sigma factor), the α5-α6 loop and helix α7 of subdomain σ₄, helix α1 of subdomain σ₂, and helix α4 of the σ₂–σ₄ linker (Fig. 4A). NepR residue L32′ makes the largest number of contacts to PhyR_SL. Y36′ is the only residue that contacts both σ₂ and σ₄, and an aromatic residue is strictly conserved at this position in NepR orthologs (Fig. S4). At interface 1, a single polar contact is observed between R33′ and E108 (Fig. 4A).

Fig. 4.

Molecular contacts at the PhyR_SL–NepR interface. Schematic (Left) and stereoview (Right) of interface 1 (A) and interface 2 (B) are shown. Side chains at the intermolecular interface are shown as a stick model. Carbon atoms are cyan for NepR and black for PhyR_SL. All oxygens are red, all nitrogens are blue, and all sulfurs are orange. Polar interactions are indicated by red dashed lines in the structure and by red solid lines in the schematic representations, respectively. Hydrophobic contacts are marked by black lines. The coloring and labeling scheme of ribbons indicating backbone positions are the same as in Fig. 2.

Similarly to helix α1′, helix α2′ of NepR at interface 2 folds into an amphipathic cylinder that inserts its hydrophobic surface into a pocket created by the hydrophobic surfaces of PhyR_SL helices α1 and α2 of subdomain σ₂ (corresponding to regions σ_2.1 and σ_2.2, respectively, by analogy to bona fide ECF sigma factor; Fig. 4B). Residues M48′ and L52′ at interface 2 contact both helices α1 and α2 of PhyR_SL. The α1′-α2′ linker also forms contacts at interface 2. One of the residues of this linker, P45′, contacts both helices α1 and α2 and is located close to P12 of PhyR_SL helix α1, which creates a kink in helix α1 responsible for the parallel orientation of the N-terminal part of helix α1 and helix α2. The hydrophobic packing of PhyR_SL helices α1 and α2 is maintained by intramolecular hydrophobic contacts of PhyR_SL F46 with L3, L7, and I39, thus creating a hydrophobic core that stabilizes interface 2 (Fig. S6). In addition to hydrophobic interactions anchoring of NepR helix α2′ to PhyR_SL, helix α2 is supported by two polar interactions between E47′ and R33, and between K54′ and E37 (Fig. 4B).

Although not all residues are strictly conserved, sequence alignments of NepR orthologs show that the amphipathic character of NepR helices α1′ and α2′ is maintained among Alphaproteobacteria (Fig. S4). Among the most conserved residues are those that contact either both subdomains σ₂ and σ₄ in interface 1 (Y36′) or both helices α1 and α2 of σ₂ in interface 2 (P45′ and L52′). In addition, the hydrophobic character of most PhyR residues implicated in the NepR–PhyR_SL interface is conserved in PhyR orthologs (Fig. S3). Taken together, the conservation of key residues making contacts in the NepR–PhyR_SL interface highlights the relevance of the presented structure for the entire class of Alphaproteobacteria and supports an evolutionary conserved binding mode of NepR and PhyR.

Functional Importance of the PhyR_SL–NepR Interface in the Context of Full-Length PhyR.

To test the importance of the identified residues at the PhyR_SL–NepR interface, amino acids in both proteins were substituted with alanine and mutant derivatives were analyzed using the full-length PhyR protein. Using ITC, NepR mutants were tested for their ability to bind PhyR-BeF₃ (Table 2 and Fig. S7) and phyR alleles were analyzed for their capacity to complement a phyR null mutant in phenotypic assays (Fig. 5), which requires formation of the PhyR–NepR complex (20). Western blotting and NMR, respectively, were used to confirm the PhyR mutants were expressed at levels similar to the WT protein and were correctly folded (Fig. 5C and Fig. S8).

Table 2.

Binding affinity of NepR mutants for PhyR

NepR variant	KD, nM	Affinity factor*
WT	31 ± 3	1
M28′A	106 ± 8	3.4
L32′A	>50,000	>1,600
R33′A	1,270 ± 126	40.9
Y36′A	>50,000	>1,600
I40′A	120 ± 8	3.8
E47′A	72 ± 4	2.3
M48′A	>50,000	>1,600
L52′A	3,509 ± 137	112
K54′A	210 ± 48	6.7

ITC data are presented in Fig. S7.

*Affinity factor is the ratio between the K_D of the mutant and the K_D of the WT.

Fig. 5.

Phenotypic stress assays of PhyR mutants. Tenfold dilution series of cultures of a Sphingomonas sp. Fr1 phyR null mutant harboring different phyR alleles were spotted on either nutrient broth (NB) pH 6.9 without NaCl as a control (A) or on NB supplemented with 300 mM NaCl (B). Dilutions are indicated on the left, and the amino acid substitution encoded by each allele, along with its location in PhyR_SL, is shown in the Upper part of the figure. L, σ₂–σ₄ linker. WT indicates a phyR mutant harboring the WT phyR allele (positive control); ΔphyR indicates a phyR null mutant harboring the empty plasmid (negative control). Images were recorded 3 d (A) and 5 d (B) after incubation at 28 °C. (C) Expression level of each mutant allele was monitored by Western blotting using primary rabbit α-PhyR and secondary goat α-rabbit alkaline-phosphatase conjugate antibodies.

Regarding hydrophobic contacts at interface 1, substitutions of residues L32′ and Y36′, located at the center of NepR helix α1′, resulted in proteins without detectable binding to PhyR-BeF₃, and thus confirmed the importance of these key residues. In PhyR, alleles with substitutions of residue V86 of helix α4 and residues L102, L103, and M107 of helix α5, all of which make contact with L32′ and/or Y36′, failed to restore the WT phenotype. In good agreement with the structure, substitutions of NepR residues located at the periphery of interface 1, and therefore involving fewer contacts (M28′ and I40′), resulted in modest decreases in affinity for PhyR-BeF₃ (3- to 4-fold) and substitution of the F13 residue of PhyR, which contacts only I40′, showed no detectable defect in phenotypic assays.

At interface 2, the importance of the core hydrophobic interface was confirmed by a severe decrease in PhyR-BeF₃ binding affinity of NepR mutants M48′A and L52′A, and by the failure of PhyR mutants L7A and L36A to restore the WT phenotype (Fig. 5, Table 2, and Fig. S7). Mutations of residues in PhyR (L3A and L11A) located at the periphery of interface 2 had modest or no effects. Finally, the central role of PhyR F46 in the formation of the hydrophobic core at interface 2, by stabilizing the parallel orientation of helix α1 and helix α2, was also confirmed, because the F46 mutant allele failed to complement a phyR mutant.

Regarding polar contacts, substitution of NepR residue R33′, located at the center of interface 1, resulted in a 40-fold decrease in PhyR-BeF₃ binding affinity. Replacement of residues E47′ or K54′, located at the periphery of interface 2, resulted in a more modest decrease in PhyR-BeF₃ binding affinity of twofold and sevenfold, respectively (Table 2 and Fig. S7).

These functional data indicate a critical role for hydrophobic contacts in the core of both interfaces, as well as the contribution of polar contacts. These results are in good agreement with the solution structure of the PhyR_SL–NepR complex and the evolutionary conservation of these key residues (Figs. S3 and S4).

Homology Modeling of the σ^EcfG–NepR Complex.

The partner-switching model proposes that PhyR uses its sigma factor-like domain to sequester NepR by mimicking σ^EcfG, implying that PhyR binds NepR with higher affinity than σ^EcfG does (12). Consistently, a dissociation constant of 1.25 ± 0.12 μM was determined by ITC for the σ^EcfG–NepR complex interaction (Fig. 6B), a 40-fold lower affinity than the one observed for the interaction between PhyR and NepR (Fig. 1A). To analyze whether this difference in affinity can be explained at the atomic level and to investigate the structural basis for “sigma factor mimicry” further, a homology model of the σ^EcfG–NepR complex was generated (Fig. 6C) based on the structure of the PhyR_SL–NepR complex. Residues 1–29 and 183–218 of σ^EcfG, which have no counterpart in PhyR_SL, are not essential for NepR binding (Fig. S9) and were not included in the model. In the σ^EcfG–NepR complex homology model, NepR makes contacts centered on helix α1′ residues L32′ and Y36′ (interface 1, Fig. S9) and on helix α2′ residues M48′ and L52′ (interface 2, Fig. S9); schematic views of both interfaces are illustrated in Fig. 6 D and E. As already indicated by the sequence alignment of PhyR and σ^EcfG (Fig. 6A, black arrows, and Fig. S3), although the hydrophobic surfaces are well conserved, fewer contacts were found at the σ^EcfG–NepR interface compared with the PhyR_SL–NepR complex. In particular, the central aromatic residue of NepR α1′ Y36′, which contacts both subdomains in the PhyR_SL–NepR complex (Fig. 4A), forms only one hydrophobic contact with subdomain σ₄ in the σ^EcfG–NepR complex model (Fig. 6D). Polar contacts identified in the PhyR_SL–NepR complex (i.e., between R33′ and E108 and between K54′ and E37) do not have counterparts in σ^EcfG, and are therefore not observed in the σ^EcfG–NepR complex homology model (Fig. 6A, magenta arrowheads). Fewer hydrophobic contacts and the absence of polar contacts in the σ^EcfG–NepR complex homology model might explain the lower affinity of NepR for σ^EcfG compared with PhyR.

Fig. 6.

NepR binds to PhyR_SL and σ^EcfG using conserved hydrophobic determinants. (A) Sequence alignment of σ^EcfG and PhyR_SL of Sphingomonas sp. Fr1. Boundaries of regions σ_2.1, σ_2.2, σ_2.3, σ_2.4, σ_4.1, and σ_4.2 and secondary structure predictions of σ^EcfG of Sphingomonas sp. Fr1 based on the sequence are illustrated above the sequence. Below the sequence, positions of secondary structures of PhyR_SL are illustrated. Residues making contacts in the PhyR_SL–NepR complex are shown with upward pointing arrows (black for hydrophobic contacts and magenta for polar contacts). Residues of σ^EcfG making contacts in the homology model of the σ^EcfG–NepR complex are indicated with downward pointing arrows using the same color code. Residues that lead to disruption of the interaction between NepR and PhyR_SL when substituted are shown with asterisks. (B) K_D determination of the interaction between NepR and σ^EcfG. (Upper) Heat changes observed upon injection of 5 μL of a 150-μM NepR solution into an 18-μM solution of σ^EcfG. (Lower) Integrated heat changes following each injection plotted against the molar ratio of NepR to σ^EcfG. (C) Homology model of the σ^EcfG–NepR complex based on the NMR solution structure of the PhyR_SL–NepR complex. Schematic view of interface 1 (D) and interface 2 (E) observed in the homology model of the σ^EcfG–NepR complex. The color scheme for backbone and side chains in B–D is identical to that in Fig. 4. (F) Bacterial two-hybrid analysis of interactions between NepR mutants and PhyR_SL or σ^EcfG; black bars correspond to PhyR_SL–NepR mutant interactions, and gray bars correspond to σ^EcfG–NepR mutant interactions. (G) Bacterial two-hybrid analysis of interactions between NepR and PhyR_SL mutants or σ^EcfG mutants; black bars correspond to NepR–PhyR_SL mutant interactions, and gray bars correspond to NepR–σ^EcfG mutant interactions. In F and G, β-galactosidase activity reflects the strength of the interaction. Values are given as the percentage relative to the WT protein. Each bar represents the mean value from results of at least three independent cultures, and error bars correspond to the SD.

NepR Binds to PhyR_SL and σ^EcfG Using Conserved Hydrophobic Determinants.

To support the homology model experimentally, we first analyzed the fold of NepR in both complexes by comparing the 2D ¹⁵N-¹H HSQC spectra of NepR in complex with σ^EcfG and PhyR_SL, respectively. Both spectra were closely related (Fig. 7 A and B), suggesting that NepR adopts a similar conformation in both complexes. As observed for the PhyR_SL–NepR complex, the N-terminal part of NepR seems to be unfolded in the σ^EcfG–NepR complex and apparently does not contribute to the interaction, which is in agreement with mutagenesis data (below). Next, alanine point mutations of residues in NepR, PhyR_SL, and σ^EcfG were constructed and interactions were tested in a bacterial two-hybrid assay (29) (Fig. 6 F and G). Substitutions of NepR residues L32′, Y36′, M48′, and L52′, all located in the hydrophobic core of interface 1 or 2, impaired binding to both PhyR_SL and σ^EcfG, whereas substitutions of residues M28′ or I40′, located at the periphery, had no effect. These results indicate that NepR uses the same hydrophobic determinants to bind both proteins. The C-terminal part (28′–56′) of NepR was sufficient for recognition of both PhyR_SL and σ^EcfG, suggesting that the N-terminal part of NepR, which is unfolded in the PhyR_SL–NepR complex, does not provide additional contacts in the σ^EcfG–NepR complex. In a complementary approach, hydrophobic residues conserved in PhyR_SL and σ^EcfG were substituted and the resulting mutants were tested for their interaction with NepR (Fig. 6G). The results obtained for PhyR_SL and σ^EcfG were similar: Substitutions of residues located in the hydrophobic core of both interfaces 1 and 2 affected binding to NepR, whereas mutations of residues at the periphery only slightly destabilized the interaction. However, substituting hydrophobic residues was more detrimental in the case of the σ^EcfG–NepR interaction; this is in line with the lower number of hydrophobic contacts observed in the homology model, and thus their greater relative contribution to protein binding.

Fig. 7.

Observation of the NepR partner switch in vitro. (A) 2D ¹⁵N-¹H HSQC spectrum of the complex formed by ¹⁵N-labeled NepR and unlabeled σ^EcfG. (B) 2D ¹⁵N-¹H HSQC spectrum of the complex formed by ¹⁵N-labeled NepR and unlabeled PhyR_SL. (C) Titration of the complex formed by ¹⁵N-labeled NepR and unlabeled σ^EcfG by addition of unlabeled PhyR_SL. NMR spectra are colored according to the ratio σ^EcfG/PhyR_SL: 1:0 (blue), 1:0.35 (magenta), 1:0.70 (green), and 1:1.05 (red). (D) Titration of the complex formed by ¹⁵N-labeled NepR and unlabeled PhyR_SL by addition of unlabeled σ^EcfG. NMR spectra are colored according to the ratio σ^EcfG/PhyR_SL: 0:1 (red) and 1.5:1 (black). (E) (Left) Expansion of the series of spectra in C showing the amide proton-nitrogen (H-N) signals of E47′ in the σ^EcfG- and PhyR_SL-bound states. (Right) 1D projections along the proton chemical shift axis for each spectrum of the titration series of the expansion shown to the left. The resonances corresponding to the PhyR_SL-bound and σ^EcfG-bound populations are indicated by a black square and an asterisk, respectively. (F) Same representation as that in E for the H-N signals of G29′.

As mentioned above, polar contacts appear unique to the PhyR_SL–NepR interaction. In agreement, substitution of R33′ or K54′ of NepR by alanine resulted in destabilization of the PhyR_SL–NepR interaction, whereas the interaction between σ^EcfG and NepR was not or only slightly affected (Fig. 6F). As expected, reciprocal substitutions in PhyR_SL (E37A, E108A, and [E37A, E108A]) affected the interaction with NepR (Fig. 6G). Interestingly, when acidic amino acids were introduced at the corresponding positions in σ^EcfG, the resulting mutants (L66E, G151E, and [L66E, G151E]) bound NepR more tightly than the WT protein did, suggesting that, indeed, polar contacts contribute to the higher affinity of NepR for PhyR compared with σ^EcfG.

Altogether, these data show that the same NepR residues are taking part in both complexes and suggest that NepR binds PhyR and σ^EcfG in a similar way. A combination of polar interactions and a higher number of hydrophobic contacts in the PhyR–NepR complex may explain the 40-fold difference in affinity between NepR–PhyR and NepR–σ^EcfG.

PhyR_SL Disrupts the σ^EcfG–NepR Complex in Vitro.

The higher affinity of NepR for PhyR compared with σ^EcfG could account for the partner-switching mechanism. However, this notion only holds true if σ^EcfG was not present in substantial excess over PhyR, because greater concentrations of σ^EcfG could compensate for the higher affinity of NepR for PhyR. To test for cellular concentrations of PhyR and σ^EcfG, semiquantitative Western blots were used. The data indicate that PhyR and σ^EcfG are present at roughly similar concentrations (the σ^EcfG/PhyR ratio was estimated to be 1:1.4; Fig. S10). Although we were not able to determine the cellular concentration of NepR, it is intuitive to assume that it must be less than the cumulative concentration of PhyR and σ^EcfG for the switch to occur; accordingly, genetic experiments have shown that overexpression of nepR is sufficient to abolish the responsiveness of the cascade, presumably because both PhyR and σ^EcfG are titrated by NepR in this condition (12, 20). We thus attempted to reconstitute the partner switch in vitro using equimolar concentrations of proteins. NMR titration of the σ^EcfG–NepR complex by PhyR_SL was performed by following the resonances of ¹⁵N-labeled NepR upon addition of unlabeled PhyR_SL (Fig. 7). Here, PhyR_SL served as a mimic for the constitutively active form of PhyR. Monitoring the chemical shift changes of isolated resonances of NepR allowed us to assess complex formation of NepR with PhyR_SL or σ^EcfG differentially (Fig. 7 E and F). Upon addition of unlabeled PhyR_SL and as long as σ^EcfG was present in excess, two sets of NepR resonances were observed, corresponding to the σ^EcfG-bound and PhyR_SL-bound conformations, respectively (Fig. 7 E and F). At equimolar concentrations of σ^EcfG and PhyR_SL, the NepR spectrum matched that of the PhyR_SL–NepR complex, indicating that all NepR molecules were bound to PhyR_SL. In a control experiment, addition of 1.5 equivalents of σ^EcfG to the PhyR_SL–NepR complex induced no changes in NepR resonances (Fig. 7D).

In conclusion, these data suggest that the major aspect of the partner-switching mechanism is the higher affinity of NepR to PhyR compared with σ^EcfG, and that a posttranslational mechanism is sufficient to explain the partner switch in vivo.

Discussion

Based on genetic and biochemical evidence, the anti-sigma factor antagonist PhyR has been proposed to function in a partner-switching mechanism via sigma factor mimicry by using its N-terminal PhyR_SL to interact with the anti-sigma factor NepR, thus releasing the bona fide sigma factor σ^EcfG (12). In complex with PhyR_SL, NepR forms two helices that contact the solvent-accessible surfaces of PhyR_SL σ₂ and σ₄, as well as the short helix α4 of the σ₂–σ₄ linker. By comparison, in the X-ray crystal structures of both the Escherichia coli σ^E–RseA complex (28) and the Rhodobacter sphaeroides σ^E–ChrR complex (30), the anti-sigma domain (ASD) of RseA and ChrR is composed of four helices, the first three of which are sandwiched between σ₂ and σ₄, occluding region σ_4.1, where the RNA polymerase β flap-tip-helix would bind, whereas helix 4 of the ASD occludes the RNA polymerase β coiled-coil binding surface of σ₂ (7, 31–34) (corresponding to σ_2.1 and σ_2.2; Fig. 6A). Thus, although NepR contacts the same functional regions required for RNA polymerase binding in true sigma factors (σ_2.1, σ_2.2, and σ_4.1), it is not sandwiched between σ₂ and σ₄ but uses a distinct binding mode compared with the anti-sigma factors ChrR and RseA, forming a clamp along the surface of its partner (Fig. 8). Whether this mode of binding at the surface is specific to the PhyR–NepR–σ^EcfG partner switch or represents a more general sigma/anti-sigma factor interaction remains to be determined.

Fig. 8.

Comparison between ECF sigma factor–anti-sigma factor and PhyR_SL–NepR complex structures. (A) X-ray crystal structure of R. sphaeroides σ^E–ChrR complex (PDB ID code 2Z2S), wherein the cupin domain of the anti-sigma factor is not illustrated. (B) X-ray crystal structure of E. coli σ^E–RseA complex (PDB ID code 1OR7). (C) Lowest energy NMR structure of the complex formed by the PhyR_SL domain and the anti-sigma factor NepR (PDB ID code 2LFW). In A–C, helices are represented by cylinders. The subdomains σ₂ and σ₄ are yellow and gray, respectively, and the anti-sigma factor is shown in blue. The positions of specific subdomain regions σ_2.1, σ_2.2, and σ_4.2, which are important for RNA polymerase multi-subunit binding, are indicated.

Homology modeling of the σ^EcfG–NepR complex, together with experimental data, suggest that the binding interfaces between σ^EcfG–NepR and PhyR_SL–NepR are similar: (i) comparative analysis of the binding of NepR mutants to PhyR_SL and σ^EcfG indicates that the same hydrophobic molecular determinants are required for both interactions; (ii) the reciprocal approach using PhyR_SL and σ^EcfG mutants also demonstrates the importance of hydrophobic residues conserved between both proteins; and (iii) the 2D ¹⁵N-¹H HSQC spectrum of NepR in the σ^EcfG–NepR complex is closely related to the spectrum of NepR in the PhyR_SL–NepR complex, suggesting that the C-terminal part of NepR adopts a similar conformation in both complexes, whereas the N-terminal part appears to be disordered. Although the N-terminal part is not required to bind both proteins, it is required for functionality in vivo (22) and might play a role in expression, solubility, or stability, perhaps because of the presence of charged residues, a common feature of NepR orthologs. In conclusion, although elucidation of the structure of the σ^EcfG–NepR complex will be necessary to address this point unambiguously, our data give support for the concept of sigma factor mimicry at the structural level.

The partner switch model implies that NepR must bind to PhyR with higher affinity than to σ^EcfG. The measured affinity of the interactions between NepR and σ^EcfG or NepR and PhyR indeed indicate that NepR preferentially binds phosphorylated PhyR, with a 40-fold higher affinity than σ^EcfG. At the structural level, this may be achieved by the lower density of hydrophobic contacts and the absence of polar interactions in the σ^EcfG–NepR complex compared with the PhyR_SL–NepR complex structure. In fact, we showed that PhyR_SL can disrupt the σ^EcfG–NepR complex in vitro when the three partners are present in equimolar amounts. Although demonstrated with PhyR_SL, which shows higher affinity to NepR than the phosphorylated form, these data suggest that the difference in affinity is sufficient to make the switch feasible.

In conclusion, based on our structural and functional work, we propose that on stress, PhyR phosphorylation triggers binding to NepR by inducing destabilization of the polar interface between PhyR_REC and PhyR_SL, rendering interface 1 available for binding of NepR. The activated form of PhyR is then able to titrate NepR from its inhibitory interaction with σ^EcfG, by providing a hydrophobic interface mimicking the NepR binding interface of σ^EcfG supplemented by additional polar contacts. Released σ^EcfG then redirects transcription toward stress genes (Fig. 9). This model illustrates how allosteric coupling in the PhyR response regulator governs the partner switch.

Fig. 9.

Structural model of the partner-switching mechanism governing the general stress response in Alphaproteobacteria. (Upper Left) Schematic diagram of the unphosphorylated form of PhyR and cartoon representation of the homology model of PhyR from Sphingomonas sp. Fr1. The position of helix α4, the polar interaction between PhyR_REC and PhyR_SL, and the position of the PhyR_REC α11-β5-α12 face are illustrated. A red arrow shows the movement of the region α11-β5-α12 upon phosphorylation. (Upper Right) Schematic diagram of the phosphorylated form of PhyR in complex with NepR, with the black surface illustrating the hydrophobic surfaces exposed for the binding of NepR, and a model of the phosphorylated form of PhyR in complex with NepR. (Lower Left) Schematic diagram and homology model of the sigma factor σ^EcfG in complex with NepR, based on the PhyR_SL–NepR solution structure. (Lower Right) Schematic diagram of the σ^EcfG–RNA polymerase multi-subunit core complex bound to the promoter of a stress gene. Positions of the -35 and -10 promoter elements are indicated. For all panels, PhyR_REC is shown in cyan, the σ₂ subdomain of PhyR_SL or σ^EcfG is shown in yellow, the σ₄ subdomain of PhyR_SL or σ^EcfG is shown in gray, and the σ₂–σ₄ linker encompassing helix α4 is shown in green. NepR is colored blue.

Methods

Plasmid Construction, Expression, and Purification.

Details on plasmid construction, expression, purification, and sample preparation are provided in SI Methods.

NMR Measurements.

All NMR measurements were performed in NMR buffer [10 mM sodium phosphate (pH 6.8), 50 mM NaCl) with either 90% H₂O/10% D₂O (vol/vol) or 100% D₂O (for 3D ¹³C-edited/¹³C-filtered NOESY experiments) at 303 K using Bruker AVIII-500 MHz, AVIII-600 MHz, AVIII-700 MHz, and Avance-900 MHz spectrometers (all equipped with cryoprobes). The data were processed using Topspin 2.1 (Bruker) and analyzed with CARA software (http://www.cara.nmr.ch/) (35). Sequence-specific backbone and side chain assignments were achieved using 2D ¹H-¹⁵N HSQC, 2D ¹H-¹³C HSQC, 3D HNCA, 3D HNCO, 3D HNCOCA, 3D HNCACO, 3D CBCACONH, 3D HNCACB, 3D HC(C)H TOCSY, 3D H(C)CH total correlation spectroscopy (TOCSY), 3D NOESY ¹H-¹⁵N HSQC, 3D NOESY ¹H-¹³C_aliphatic HSQC, and 3D NOESY ¹H-¹³C_aromatic HSQC experiments for each protein. All NOESY experiments were recorded with a mixing time of 80 ms. Intermolecular NOEs were obtained using 3D ¹³C F1-edited/F3-filtered NOESY-HSQC experiments for two different samples (36). In each sample, either NepR or PhyR_SL was ¹³C,¹⁵N-labeled and the other component was unlabeled. The {¹H}-¹⁵N heteronuclear NOE experiment was recorded on an AVIII 500 MHz spectrometer in an interleaved fashion, alternatively recording one increment for the reference and one increment for the NOE spectrum. A relaxation delay of 2 s and a ¹H presaturation delay of 3 s were used in the NOE experiment, whereas a 5-s relaxation delay was used in the reference experiment (37).

Structure Calculation and Refinement.

The nearly complete resonance assignments of PhyR_SL in the complex formed the basis for peak picking and NOE assignment in the 3D NOESY spectra obtained with labeled PhyR_SL using UNIO-ATNOS/CANDID software (38, 39) in combination with structure calculation using CYANA (40). A preliminary ensemble of structures with the lowest energy target function and a list of automatically assigned NOE-derived distance restraints were obtained for PhyR_SL. A similar approach was followed to obtain initial structures of NepR in the complex. In addition to the NOE-derived distance constraints, dihedral angles constraints were derived for selected residues in regions of secondary structures with TALOS+ using backbone chemical shifts as input (41). Hydrogen bonds were identified by inspection of the structures, and hydrogen bond constraints were added to the constraint list for those amides showing slow hydrogen-deuterium exchange (observed after 2 h of exchange in a ¹⁵N-¹H HSQC experiment). Intermolecular NOEs were manually assigned in the two 3D ¹³C F1-edited/F3-filtered NOESY ¹H-¹³C HSQC spectra, and NOE-derived distance constraints identified in the initial structure calculations of PhyR_SL and NepR with UNIO/ATNOS-CANDID at corresponding positions in the nonfiltered 3D NOESYs were eliminated from the constraint lists for the individual proteins. Structures of the protein–protein complex were calculated using CYANA (40) by adding the manually assigned intermolecular NOEs, dihedral angles, and hydrogen bonds to the edited UNIO/ATNOS-CANDID restraints. For each CYANA run, 200 independent structures were calculated and the 20 best resulting structures were refined with the SANDER module of AMBER 7.0 using a simulated annealing protocol (42). Analysis of refined structures was performed using PROCHECK (43).

Analytical Size Exclusion Chromatography and Western Blotting.

Analytical size exclusion chromatography and Western blotting were performed as previously described (20). Details on semiquantitative Western blot analysis are given in SI Methods.

ITC.

ITC experiments were performed on a VP-ITC instrument (Microcal). The calorimeter was calibrated according to the manufacturer’s instructions. Both proteins were dialyzed against the same buffer: in the NMR buffer for the PhyR_SL–NepR complex titration; in 10 mM sodium-phosphate buffer (pH 6.8), 100 mM NaCl, 5 mM BeCl₂, 100 mM NaF, and 5 mM MgCl₂ for the PhyR–NepR complex titration; and in 50 mM Tris⋅HCl (pH 8), 100 mM NaCl, and 2.6 mM β-mercaptoethanol for the σ^EcfG–NepR complex titration. The protein concentrations were determined using the optical density absorbance at 280 nm, and the NepR protein concentrations were confirmed using the Bradford assay. All the titrations were performed at 25 °C. Raw data were integrated, normalized for the molar concentration, and analyzed using Origin 7.0 (OriginLab) according to a 1:1 binding model.

Phenotypic Stress Assays.

Phenotypic stress assays were performed as previously described (20).

Bacterial Two-Hybrid Complementation Assays.

Bacterial two-hybrid complementation assays were performed as previously described using E. coli strain BTH101 (20).

Sequence Alignment, Homology Modeling, and Structure Visualization.

Details on sequence alignment, homology modeling, and structure visualization are provided in SI Methods.