Phospho‐dependent signaling during the general stress response by the atypical response regulator and ClpXP adaptor RssB

Abstract

In the model organism Escherichia coli and related species, the general stress response relies on tight regulation of the intracellular levels of the promoter specificity subunit RpoS. RpoS turnover is exclusively dependent on RssB, a two‐domain response regulator that functions as an adaptor that delivers RpoS to ClpXP for proteolysis. Here, we report crystal structures of the receiver domain of RssB both in its unphosphorylated form and bound to the phosphomimic BeF₃ ⁻. Surprisingly, we find only modest differences between these two structures, suggesting that truncating RssB may partially activate the receiver domain to a “meta‐active” state. Our structural and sequence analysis points to RssB proteins not conforming to either the Y–T coupling scheme for signaling seen in prototypical response regulators, such as CheY, or to the signaling model of the less understood FATGUY proteins.

Short abstract

PDB Code(s): 7L9C and 7LCM;

1. INTRODUCTION

Two‐component systems mediate signal transduction in bacteria and consist of an environmental cue‐sensing histidine kinase that phosphorylates a response regulator composed of either a self‐standing “receiver” (REC) domain or a REC domain fused to an effector (EFF) domain. ¹ The REC tunes the activity of the EFF domain in response to phosphorylation of a conserved aspartate within the REC. While early studies conceptualized the REC module in terms of a binary logic element (existing in the phosphorylated ON and a nonphosphorylated OFF form), numerous subsequent studies have documented that REC domains exist in a dynamic equilibrium of multiple conformations, with activation sometimes preceding phosphotransfer. ² , ³ What has become clear is that phosphorylation‐dependent activation often involves small structural changes within self‐standing RECs, or oligomerization and closed‐to‐open transitions within multidomain REC proteins (reviewed by Gao et al. ¹ ). These changes are often unique to each REC subfamily. Thus, despite a common architectural scaffold, a general mechanistic scheme for REC activation does not hold.

Here, we focus on RssB, a two‐domain response regulator (Figure 1a,b) and the poorly understood role of its phosphorylation. RssB, a key element of the general stress response in both commensal and pathogenic γ‐proteobacteria, is best understood in Escherichia coli, where it serves as the only ClpXP adaptor for RpoS, the master regulator of the general stress response. ⁴ , ⁵ , ⁶ , ⁷ RssB delivers RpoS to the ClpXP proteolytic machine for degradation. ⁵ , ⁸ Thus, RpoS levels remain low due to rapid turnover when cells are actively dividing. Upon encountering stress, they increase in large part due to inhibition of RssB by so‐called anti‐adaptors (IraM, IraP, IraD, and IraL), ⁹ , ¹⁰ favoring competition of RpoS with other promoter specificity subunits for core RNA polymerase. ¹¹ This results in large‐scale transcriptional reprogramming. ¹² , ¹³

FIGURE 1.

Overall architecture of IraD‐bound RssB^D58P and RssB^REC in the absence/presence of phosphoryl analog, BeF₃ ⁻. (a) Primary and secondary structure of Escherichia coli RssB^REC. Asp⁵⁸ is indicated with a red circle. Green circles indicate locations at which substitutions result in defects in RpoS degradation. ¹⁶ Purple rectangles indicate residues, which together with Asp⁵⁸ constitute the signaling quintet. (b) Structure of IraD‐bound RssB^D58P variant with REC in cyan and the EFF domain in pink. IraD (gray) docks primarily on the 4‐5‐5 face of RssB^REC. Residue 58 is shown as red sticks. (c) Superposition of apo RssB^REC (pale cyan) and RssB^REC bound to beryllofluoride (dark teal, r.m.s.d. of 0.41 Å). Apo RssB^REC also superimposes well (r.m.s.d. of 0.28 Å) with a previously reported structure of the same domain in the nonphosphorylated state (PDB ID 6Z4C). ¹⁵ (d,e) The phosphorylation site in RssB^REC•BeF₃ ⁻•Mg2⁺ (d) and RssB^REC (e). Mg²⁺ is green, while beryllium is magenta and fluoride yellow. A Polder omit map contoured at 3σ (slate mesh) shows clear electron density for the phosphoryl analog, Mg²⁺ and two water molecules (red spheres). In apo RssB^REC, Asp⁵⁸ hydrogen bonds to Lys¹⁰⁸, Ser⁸⁶, and a water molecule (red sphere)

RpoS proteolysis is also modulated by phosphorylation of Asp⁵⁸ in RssB, but the few existing reports do not agree on its role. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ This prompted us to solve the crystal structure of the RssB REC domain (RssB^REC) in its nonphosphorylated form and bound to the phosphomimetic BeF₃ ⁻. We find only subtle conformational differences between RssB^REC in the presence/absence of Mg²⁺•BeF₃ ⁻, suggesting that deletion of the C‐terminal domain may partially activate RssB^REC even in the absence of a phosphomimic. Such conformational states in which only a subset of residues switch to the conformers seen in phosphorylated RECs or RECs bound to phosphomimetics have been observed in the context of other response regulators, and have been dubbed “meta‐active” states. ¹⁹ , ²⁰ , ²¹ Our analysis also indicates that signaling in RssB is atypical. RssB lacks canonical Tyr/Thr switch residues, and likely does not conform to the Y–T coupling model seen in classic receiver domains such as CheY. ²² Alternative signaling mechanisms, such as seen in the lesser understood FATGUY subfamily ²³ are also unlikely to apply, making RssB a response regulator with an unusual mechanism of action.

2. RESULTS

2.1. The RssB phosphorylation site

We first attempted to determine the structure of full‐length RssB, but our crystallization efforts were hampered by the limited solubility of the full‐length protein construct. Instead, we determined the crystal structure of RssB^REC (residues 1–129) with an intact phosphoacceptor site (Asp⁵⁸) in its nonphosphorylated state and bound to the BeF₃ ⁻ phosphomimetic (Table 1, Figure 1). This has been found to bind and activate numerous response regulators ²⁴ and, due to the great lability in solution of the natural phosphate‐acid linkage, has been widely used in crystallization. RssB^REC has the canonical (βα)₅ REC fold (Figure 1c) and possesses a typical signaling quintet. This includes residues Asp¹⁵, Glu¹⁶, involved in Mg²⁺ binding, but also Lys¹⁰⁸, which reaches toward Asp⁵⁸, and Ser⁸⁶ that stabilizes the Lys¹⁰⁸ amide both in the absence and presence of beryllofluoride (Figure 1d,e). Substitution of Lys¹⁰⁸ results in negative effects on RssB function in vivo ¹⁶ and in vitro, ¹⁵ , ¹⁶ consistent with our structures. Similar defects were observed for a Pro¹⁰⁹ variant (Figure 1a). ¹⁶ Invariant Pro¹⁰⁹ mediates a cis‐peptide bond in the β5‐α5 loop, a conserved feature of many response regulators, ²⁴ , ²⁵ which imposes more stringent constraints on stereochemistry and may explain why the rssB ^P109S allele is defective and temperature‐sensitive. ¹⁶

TABLE 1.

Crystallographic data and refinement statistics

	RssBREC	RssBREC•BeF3 −•Mg2+
Data collection
X‐ray source	Advanced Photon Source 24ID‐E	Advanced Photon Source 24ID‐C
Space group	P3121	P3121
Cell dimensions
a,b,c (Å) a	46.46, 46.46, 95.85	46.33, 46.33, 95.89
α,β,γ (°)	90.0, 90.0, 90.0	90.0, 90.0, 120.0
Mosaicity (°)	0.17	0.14
Wavelength (Å)	0.97911	0.97918
Resolution (Å)	40.0–1.85 (1.92–1.85) a	40.12‐1.91 (1.95‐1.91) a
No. reflections/unique reflections	112,144/10,656	90,612/9,772
Completeness (%)	98.9 (92.52)	99.4 (94.6)
Redundancy	10.5 (6.4)	9.3 (6.7)
<I/σ(I)>	24.77 (2.2)	21.0 (1.7)
R merge	0.049 (0.745)	0.087 (0.992)
R meas	0.051 (0.806)	0.088 (0.994)
R pim	0.0155 (0.300)	0.028 (0.366)
CC1/2	0.999 (0.943)	0.999 (0.690)
Refinement
Resolution (Å)	40.0–1.85	40.12–1.91
No. reflections, working set	10,632 (977)	9,693 (898)
No. reflections, test set	1,052 (96)	962 (86)
R work/R free (%)c	20.04/22.62	20.2/24.5
No. nonhydrogen atoms
Protein	983	987
Water	36	44
B‐factors (Å2)
Protein	45	42
Water	45	44
Ion/ligand	‐	36.5
R.m.s.d.
Bond lengths (Å)	0.008	0.008
Bond angles (°)	1.11	1.14
Coordinate error (Å) b	0.17	0.17
Ramachandran plot analysis
Preferred (%)	99.2	98.4
Allowed (%)	0.8	1.59
Disallowed (%)	0	0
Rotamer outliers (%)	0.91	0
PDB ID	7L9C	7LCM

^aValues in parenthesis are for highest resolution shell.

^bMaximum likelihood estimate.

The octahedral coordination of Mg²⁺ involves Asp⁵⁸, Asp¹⁵, the BeF₃ ⁻ moiety, two water molecules and the D + 2 residue, Ala⁶⁰ (Figure 1d,e). Located two residues downstream of Asp⁵⁸, this augments the active site and likely plays key roles in determining the kinetics of REC autodephosphorylation. These can differ by orders of magnitude between different REC domains and depend heavily on the identity of the D + 2 residue. ²⁶ An Ala at position D + 2 is atypical for a REC. ²⁶ The K + 1 residue, Pro¹⁰⁹, is invariant, while K + 2 (residue 110) is conserved as a branched amino acid (Figure S1). These two residues also affect the kinetics of REC phosphorylation/dephosphorylation but, unlike residue D + 2, they likely do so via an indirect mechanism that affects REC conformational equilibria. ²⁷ Consistent with this, Val¹¹⁰ is located about 12 Å away from the phosphorylation site and does not engage in any hydrogen bonds or van der Waals contacts that could link it directly to the active site.

2.2. Signaling by the RssB^REC domain does not involve Y–T coupling

A highly conserved RssB residue is Asp¹⁰⁴, located in β5, on the 4‐5‐5 face (Figures 2 and S1). In many response regulators, including CheY, OmpR, NtrC, LytR, NarL, this residue is typically replaced by a tyrosine (e.g., Tyr¹⁰⁶ in CheY). ¹ This tyrosine, together with a threonine (Thr⁸⁷ in CheY), is essential for signaling via Y–T coupling. Upon phosphorylation, both Tyr¹⁰⁶ and Thr⁸⁷ of CheY undergo rotameric changes that propagate from the phosphoacceptor site to the 4‐5‐5 face ²² , ²⁸ , ²⁹ (Figure 2c,d), where many regulators bind, and which, in some systems, supports activation by oligomerization. ³⁰ Thus, the tyrosine sidechain points away from the phosphoryl group in the apo state and, upon phosphorylation, it swings inward to bury itself in a pocket vacated by the conserved threonine (or serine) in β4. In some cases, it has been found that the threonine equivalent to CheY Thr⁸⁷ gates and engages the entire β4‐α4 loop and that the phosphorylation‐dependent active conformation of this loop favors burial of Tyr¹⁰⁶, an extended mechanistic model that has been dubbed T‐loop‐Y coupling. ² The presence of an Asp in RssB instead of an aromatic is inconsistent with Y–T coupling. Upon superposition, we also observe no major changes in RssB^REC upon BeF₃ ⁻•Mg²⁺ binding (r.m.s.d. of 0.41 Å). The 4‐5‐5 face, the locus of the largest differences between ON and OFF conformations in REC domains, is remarkably similar, perhaps due to the partial activation of RssB^REC in this truncated form. However, significant conformational differences are observed between the 4‐5‐5 face of RssB^REC and of full‐length RssB^D58P due to tilting of α5 as well as motion of the loop preceding it (Figure 2a). This points to this face of RssB^REC being plastic and subject to regulation.

FIGURE 2.

RssB employs neither Y–T coupling nor SdrG‐like signaling. (a) Superposition of RssB^REC (pale cyan) and full‐length RssB^D58P mutant (PDB ID 6OD1, gray). ³⁷ The signaling helix, site of class II mutations as defined by Gottesman and coworkers is colored green. Asp¹⁰⁴ is shown as sticks. (b) Sequence conservation around the PvL[VIM]ISAT motif in the RssB family and equivalent region in SdrG. Residues are colored by charge/hydrophobicity. Substitution of boxed SdrG residues leads to loss of function. ²³ (c,d) Structures of Escherichia coli CheY in OFF state (PDB ID 3CHY) ²⁵ and ON state (PDB ID 1FQW). ⁴¹ Phosphoacceptor is shown in red. The ON state structure was solved by NMR and contains a Mn²⁺ rather than an Mg²⁺ ion. Two Tyr¹⁰⁶ rotamers were observed. Overall r.m.s.d. after superposition of the ON/OFF states is 0.54 Å. (e,f) Sphingomonas melonis SdrG in the OFF state (PDB ID 5IEB) ²³ and ON state (PDB ID 5IEJ). ²³ Phosphorylation site is shown in red. Structures were solved using NMR, and the positions of the phosphomimetic and Mg²⁺ were not determined experimentally. For simplicity, they are not modeled. Overall superposition r.m.s.d. is 1.68 Å

2.3. Comparison to FATGUY response regulators

The lack of Y–T coupling is unusual, but not without precedent. In CheY2, the role of Tyr¹⁰⁶ is taken by Phe⁵⁹ located in the β3‐α3 loop, ³¹ while in NtrC, the rotameric state of the tyrosine is uncorrelated with the global conformational transition during activation. ³² In the NarL subfamily, the tyrosine exists but is constitutively pointing inward regardless of phosphorylation state, ³³ , ³⁴ suggesting it may not be involved in allostery. In response regulators of the FATGUY subfamily, such as SdrG, activation occurs instead via stabilization of the phosphoryl‐aspartate via hydrogen bonding with Lys¹⁰² (Lys¹⁰⁸ in RssB) and Thr⁸³ (Ser⁸⁶ in RssB), which occupies the same position within β4 as the conserved CheY threonine involved in Y–T coupling (Figure 2). The rotation of Thr⁸³ ensues in remodeling of several phenylalanines, in and around a highly conserved PFXFAT⁸³G[G/Y] motif making up the β4 strand (Figure 2b,e,f). ²³ Activation also leads to structuring of the α4 helix and changes in the loop C‐terminal to the phosphoacceptor site. In RssB^REC, RssB^REC•BeF₃ ⁻•Mg²⁺ and the recent structure of IraD‐bound RssB^D58P (Figure 2a), we observe that the equivalent Lys¹⁰⁸ points toward the phosphoacceptor location. This is usually associated with an ON state and is contrasted to the OFF conformer, in which Lys¹⁰⁸ points away (Figure 2e,f). However, neither of the remodeled conserved FATGUY residues is present in the consensus RssB sequence. Instead, they are replaced by a highly conserved PvL[VIM]ISAT⁸⁸ motif spanning β4 and the β4‐α4 loop (Figure 2b), with the phenylalanines being replaced by hydrophobes (Val, Ile, Met). Notably, substitution of Phe⁷⁹, Phe⁸¹, and Phe⁹⁴ lead to functional SdrG defects in vitro and in vivo. ²³ While conservation of the residues in β4—part of the hydrophobic core—may be due to scaffolding reasons, Leu⁸³, Val⁸⁴, and Ile⁸⁵ interact with Cys⁵⁷ next to the phosphorylation site, and may relay phosphorylation status to more distant sites such as the β4‐α4 loop. This loop contains conserved residue Thr⁸⁸ (also dubbed as T + 2 according to the nomenclature of Bourret and colleagues), which modulates dephosphorylation rates in family specific ways. ³⁵ In both of our structures, Thr⁸⁸ hydrogen bonds to residues Ser⁸⁶ and Asn⁹⁰. Consequently, there are minimal differences between the β4‐α4 loops in apo REC and the REC~P‐like form. We cannot rule out that this may be due to using truncated protein constructs for crystallization.

2.4. The importance of helix α5

Temperature factor analysis confirms that the β3‐α3 loop C‐terminal to Asp⁵⁸ displays increased dynamics in RssB^REC in the absence of phosphorylation (Figure S2a). Surprisingly, helix α5—part of the central helical bundle to which both the N‐terminal and C‐terminal domains contribute in full‐length RssB (Figure 2a)—undergoes only subtle changes, notably to the rotameric conformations of Lys¹¹¹, Arg¹¹⁷, and Phe¹²¹ in α5 (Figure S2c–f).

The interactions of α5 residue Arg¹¹⁷ attracted our attention for four reasons. First, in RssB^REC•BeF₃ ⁻•Mg²⁺, the amide of Arg¹¹⁷ hydrogen bonds to the carbonyl of Arg¹¹⁵, while in RssB^REC, Arg¹¹⁷ is stabilized by contacts with Trp ²⁶ (Figure S2c,d). Second, the tilting of α5 we observe in the RssB^REC structure relative to IraD‐bound RssB^D58P would cause a clash with α9 in the closed RssB structure (Figure 2a). α9 is, incidentally, the site of mutations (class II mutations as defined by Gottesman and coworkers and highlighted in green Figure S2b) that activate RssB in the absence of phosphorylation, bypassing the need for acetyl phosphate. ¹⁶ Third, in IraD‐bound RssB^D58P, Arg¹¹⁷ interacts with Gln²⁴⁷, and this interaction dominates the REC–EFF interface (Figure S2b). Last, but not least, an alanine substitution of Arg¹¹⁷ results in loss of interaction with RpoS in the absence of acetyl phosphate, although degradation (and, presumably substrate binding) is only partially compromised in vitro in the presence of acetyl phosphate. ¹⁵ This points to Arg¹¹⁷ serving as a switch that modulates substrate binding.

3. DISCUSSION

As of November 2020, PFAM ³⁶ lists over 324,000 REC sequences and the Protein Data Bank contains many structures of REC domains, mostly in the non‐phosphorylated state. With such a bounty of structural information, one would expect a thorough understanding of how signaling takes place in REC proteins. This is not so, since REC domains are often stringed to a variety of domains supporting different REC–EFF contacts, structures of REC~P‐like states are relatively rare, and reaction kinetics are strongly influenced by subfamily specific residues (e.g., D + 2, T + 2, but not only) outside the conserved architecture of the phosphorylation site. ²⁶

RssB is an unusual response regulator, not in the least because its C‐terminal domain effector domain is a PP2C‐like pseudophosphatase domain that has been co‐opted for protein–protein interactions with ClpXP rather than being used for recognizing DNA. ¹⁵ , ³⁷ Most models of RssB mechanism have been based on RssB sequence similarity with a Pseudomonas aeruginosa response regulator of known structure (PDB IDs 3EQ2 and 3F7A, unpublished), but uncharacterized function. This protein assumes an extended dimeric dumbbell conformation with two domains of one monomer linked by an extended helical linker. Subsequent studies of RssB revealed a compact, very different architecture, different sequence characteristics of the RssB linker and an inability of free RssB to oligomerize at physiological concentrations. ³⁷ In light of this research, the structures presented here suggest that, while RssB may not adopt a dumbbell‐like architecture like that of the P. aeruginosa homolog, its phosphorylation may nevertheless reorganize RssB^REC/RssB^EFF contacts.

Our structure of apo RssB^REC reveals a meta‐active conformation that resembles to a great extent RssB^REC•BeF₃ ⁻•Mg²⁺. The structural differences between RssB^REC within IraD‐inhibited RssB^D58P and RssB^REC•BeF₃ ⁻•Mg²⁺ primarily involve helix α5 and the loop preceding it, while differences between apo RssB^REC and RssB^REC•BeF₃ ⁻•Mg²⁺ are maximal in the β3‐α3 loop, C‐terminal to Asp⁵⁸. While a full understanding of RssB mechanism will require the structure determination of full‐length RssB and RssB~P bound to its various partners, our work allows us to reconsider the role of phosphorylation in the RpoS general stress response. Phosphorylation, by either kinases or small phosphodonors, is generally believed to promote RpoS binding and turnover by ClpXP. ¹⁴ , ¹⁶ More recent studies have proposed that, on the contrary, phosphorylation reduces RssB affinity for RpoS, and consequently the rate of RpoS degradation. This model was based on equilibrium dissociation constant determination for core RssB‐RpoS complexes comprised of RssB^REC and region 3 of RpoS (residues 163–220), and a less than twofold decrease in the affinity in the presence of acetyl phosphate. ¹⁷ This model is at odds with multiple observations, including the requirement for acetyl phosphate in promoting RpoS‐RssB ¹⁶ and RpoS‐RssB‐ClpXP complex formation, ⁵ and in stimulating RpoS degradation in an in vitro system. ¹⁶ , ³⁷ It is, however, easily explained by our determination that removing the EFF domain partially activates RssB^REC independent of phosphorylation, obfuscating the effects of phosphoryl transfer on the full‐length protein.

We suggest that RssB conforms to neither the Y‐T‐coupling model for signaling seen in many response regulators, nor to the signaling mechanism used by FATGUY proteins during the alphabacterial stress response. RssB deregulation achieved by a single amino acid substitution of Arg¹¹⁷ in helix α5 severely impairs function. ¹⁵ This, together with the major conformational differences between IraD‐bound RssB^D58P (in an OFF state) and RssB^REC point to helix α5 as key for signal propagation to the EFF domain. Consistent with this, mutations in the interdomain linker C‐terminal to α5 (W143R, P150S, E135A E136A E137A) have also been demonstrated to affect the adaptor function of RssB and its ability to be regulated by anti‐adaptors. ¹⁶ , ³⁷ Without doubt, further understanding of the regulatory RssB system will require a multipronged approach, including determination of structures of full‐length RssB and RssB complexes, which, more than two decades since the discovery of RssB's role as an RpoS regulator, ⁵ , ⁶ , ³⁸ remain eagerly awaited.

4. METHODS

4.1. RssB^REC production and crystallization

RssB^REC was purified as a hexahistidine tag RssB (residues 1–129) fusion by a succession of nickel‐affinity chromatography, tag cleavage and dialysis followed by gel filtration on a Superdex 75 HiLoad column 16/60 (GE Healthcare) preequilibrated with 20 mM HEPES pH 7, 150 mM NaCl, 10 mM MgCl₂, and 2 mM TCEP. The sample was crystallized against 0.2 M lithium sulfate, 22% PEG 3,350, and 0.1 M Tris–HCl pH 8.8 in sitting drop format. Cryoprotection was achieved by supplementation of the well solution with 25% ethylene glycol. RssB^NTD•BeF₃ ^−•Mg²⁺ was crystallized in conditions similar to those identified for RssB^NTD.

4.2. Structure determination and refinement

Diffraction data were processed using XDS ³⁹ and phased using molecular replacement with the homologous domains of the structure with PDB ID 6OD1 in Phaser. ⁴⁰ Iterative refinement and building were performed in PHENIX and Coot. Diffraction images have been deposited in the SBGrid Data Bank as dataset number 815 (doi: 10.15785/SBGRID/815) and 818 (doi: 10.15785/SBGRID/818). PDB IDs are listed in Table 1.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

Jacob Schwartz: Purified protein and obtained crystals, and Jonghyeon Son: Collected diffraction data and refined structures. Alexandra M. Deaconescu: Designed the overall study, collected diffraction data, solved and refined structures, and wrote the manuscript with figure contributions from Christiane Brugger.

Supporting information

Appendix S1 Supporting information

Schwartz J, Son J, Brugger C, Deaconescu AM. Phospho‐dependent signaling during the general stress response by the atypical response regulator and ClpXP adaptor RssB. Protein Science. 2021;30:899–907. 10.1002/pro.4047

DATA AVAILABILITY STATEMENT

Data and reagents are available from the corresponding author upon reasonable request.