Structural basis for inhibition of a response regulator of σS stability by a ClpXP antiadaptor.

The stationary phase promoter specificity subunit σS (RpoS) is delivered to the ClpXP machinery for degradation dependent on the adaptor RssB. This adaptor-specific degradation of σS provides a major point for regulation and transcriptional reprogramming during the general stress response. RssB is an atypical response regulator and the only known ClpXP adaptor that is inhibited by multiple but dissimilar antiadaptors (IraD, IraP, and IraM). These are induced by distinct stress signals and bind to RssB in poorly understood manners to achieve stress-specific inhibition of σS turnover. Here we present the first crystal structure of RssB bound to an antiadaptor, the DNA damage-inducible IraD. The structure reveals that RssB adopts a compact closed architecture with extensive interactions between its N-terminal and C-terminal domains. The structural data, together with mechanistic studies, suggest that RssB plasticity, conferred by an interdomain glutamate-rich flexible linker, is critical for regulation of σS degradation. Structural modulation of interdomain linkers may thus constitute a general strategy for tuning response regulators.

In bacteria, the transcriptional reprogramming underlying adaptation to the environment is largely coordinated by σ factors, dissociable promoter specificity subunits that compete with the housekeeping σ70 for limited amounts of core RNA polymerase to target it to specific promoters (Maeda et al. 2000; Gruber and Gross 2003). While some σ factors have specific roles in mediating steps in bacterial development, such as in sporulation in Bacillus subtilis, or in the response to specific stressors, in Escherichia coli and related bacteria, σS (also known as RpoS) is the master regulator for the general stress response (Gruber and Gross 2003). σS mediates wholesale revamping of the transcriptional program and allows cells to adjust to slowed growth (stationary phase), disappearing critical nutrients, or a variety of stressors, including oxidative damage, pH extremes, and changes in osmolarity (Hengge-Aronis 2002; Battesti et al. 2011). The amount of σS is tightly controlled in large part via regulated proteolysis by the ClpXP ATP-dependent protease (Lange and Hengge-Aronis 1991b; Battesti et al. 2011; Mahmoud and Chien 2018). During active growth, σS levels are kept low due to its rapid degradation (half-life of ∼2 min) by ClpXP. Upon encountering stress or in the stationary phase, σS is stabilized to a half-life of >30 min, allowing its levels to rise significantly (Schweder et al. 1996). This regulation is key, as σS controls ∼10%–23% of the genes in E. coli (Lacour and Landini 2004; Patten et al. 2004; Weber et al. 2005; Dong et al. 2008; Dong and Schellhorn 2009; Peano et al. 2015; Wong et al. 2017) and, as a result, is important for many processes, including cross-protection to various environmental stressors, biofilm formation (Adams and McLean 1999) and maturation (Ito et al. 2008), adaptive morphological changes and mutagenesis (Lange and Hengge-Aronis 1991a), the development of antibiotic resistance (Ito et al. 2009), and the persistence and virulence of human enteric and respiratory pathogens as well as animal, insect, and plant pathogens (Dong and Schellhorn 2010).

Efficient σS degradation requires the use of a ClpXP adaptor called RssB (Fig. 1A). RssB is a member of the important family of response regulators found in all bacteria. The N-terminal conserved receiver domain of response regulators is generally subject to phosphorylation by sensor histidine kinases (Gao and Stock 2010); the C-terminal effector domain is frequently a DNA-binding domain whose activity is modulated by phosphorylation of the N-terminal domain (NTD) (Galperin 2006; Gao and Stock 2010). In the case of RssB, the C-terminal effector domain (CTD) was predicted to be a degenerate member of the PP2C phosphatase family (Galperin 2006).

Figure 1.

The σS degradative pathway in E. coli. (A) Under active growth, σS (blue) is loaded by adaptor RssB (pink) onto ClpXP (gray) for rapid degradation. Exposure to stress signals induces the production of unrelated antiadaptor proteins (IraP, IraM, and IraD; green), which inhibit RssB and block σS degradation. RssB phosphorylation (red circle) enhances interactions with σS but is not an absolute requirement for regulated turnover (Peterson et al. 2004; Battesti et al. 2013), consistent with a conformational equilibrium model in which RssB can assume the “on” conformation even in the absence of phosphorylation. RssB oligomerization has not been incorporated into this mechanistic scheme, and the role of desphosphoryation in promoting RssB recycling remains unaddressed. “N” and “C” indicate the NTD (receiver domain) and CTD (effector domain) of RssB, respectively. (B) Crystal structure of the RssB-like protein from Pseudomonas aeruginosa (Protein Data Bank [PDB] ID 3EQ2).

RssB is an unusual response regulator and ClpXP adaptor for several reasons: (1) First, as noted above, most response regulators respond to phosphorylation to regulate DNA binding (Galperin 2006). RssB instead mediates protein–protein interactions as the first response regulator to be identified as a ClpXP adaptor (Bearson et al. 1996; Muffler et al. 1996; Pratt and Silhavy 1996; Zhou et al. 2001). In addition, while phosphorylation of RssB occurs, it is not essential for regulation of RssB activity (Peterson et al. 2004; Bougdour et al. 2006). Within the rssB operon, no cognate histidine kinase has been identified, although some reports have implicated the ArcB kinase in RssB phosphorylation (Mika and Hengge 2005). (2) RssB has exquisite specificity, with σS being the only known substrate that it delivers to ClpXP (Muffler et al. 1996; Pratt and Silhavy 1996; Zhou and Gottesman 1998). (3) Both RssB and σS interact weakly with ClpXP, and the weak interactions are stabilized upon ternary RssB•σS•ClpX complex formation (Zhou et al. 2001). It has also been proposed that RssB may be a substrate-activating adaptor that induces a conformational change in σS to unveil a cryptic ClpXP-binding site in σS required for productive loading (Becker et al. 1999; Studemann et al. 2003). (4) Uniquely, unlike other ClpXP adaptors that are regulated via single cognate inhibitors (also known as antiadaptors), RssB is inhibited by multiple unrelated antiadaptors (IraD, IraP, and IraM) that are induced by specific stress signals (Fig. 1A; Battesti and Gottesman 2013; Joshi et al. 2015). These inhibit σS degradation in response to SOS-independent DNA damage in the case of IraD (Merrikh et al. 2009), Mg2+ starvation in the case of IraM (Bougdour et al. 2008), and phosphate starvation in the case of IraP (Bougdour et al. 2006), allowing intracellular σS to rise. Presumably, the generally low levels of RssB, which are under σS control, and homeostatic feedback coupling (Pruteanu and Hengge-Aronis 2002) are critical for allowing the antiadaptors to efficiently sequester it.

Processes dependent on response regulators are often tunable and plastic and allow for rapid and reversible adaptation to changing environmental conditions, reflecting both positive and negative regulation by histidine kinases, phosphotransfer domains, phosphatases, or other system-specific regulators (Mitrophanov and Groisman 2008; Battesti et al. 2013). Despite decades of research, the structural basis for this regulation has remained elusive, as only a few structures of response regulator complexes, bound mainly by histidine kinases, have been reported (Casino et al. 2009; Herrou et al. 2012; Trajtenberg et al. 2016). In the context of the general stress response, the lack of structural information on RssB itself, antiadaptors, and RssB•antiadaptor complexes has hampered a mechanistic understanding of how multiple antiadaptors recognize the same RssB target to alter its structure and/or dynamics and achieve stress-specific inhibition. Some structural insights have been provided by two structures of a Pseudomonas aeruginosa RssB-like response regulator (Protein Data Bank [PDB] IDs 3EQ2 and 3F7A, deposited by I Levchenko, RA Grant, RT Sauer, and TA Baker, unpubl.) sharing 23% identity with the E. coli protein. These structures, derived from two crystal forms (Fig. 1B; Supplemental Fig. S1A,B), show a dimeric and two-domain structure with an extended helical interdomain linker. However, the P. aeruginosa protein is not functionally equivalent to RssB (Battesti et al. 2013) and differs from E. coli RssB in two important ways: (1) Its active site in the PP2C-like domain is intact. (2) Unlike RssB, its linker features a sequence conducive to parallel coiled-coil formation, stabilized by characteristic “knobs into holes” packing (Supplemental Fig. S1C). Together with the absence of antiadaptor genes encoded in the P. aeruginosa genome, this suggests a different mode of regulation of this homolog and possibly a different function, thus limiting mechanistic comparisons between the two genera.

To shed light on RssB-dependent regulation of σS, here we report the crystal structure of the E. coli IraD·RssBD58P complex. This reveals a compact RssB architecture featuring extensive interactions between its NTD and CTD, promoted by a closed form of a plastic interdomain linker. Using interdomain disulfide engineering and functional studies, we demonstrate that (1) in the presence of IraD, RssB adopts predominantly a closed conformation; (2) IraD contacts both domains of RssB; and (3) the interactions observed in our structure are crucial for blocking σS degradation in vivo and in vitro. Last, we demonstrate that fine-tuning the plasticity of the interdomain glutamate-rich linker plays a critical role in σS turnover and, to a lesser extent, its inhibition by antiadaptors. We propose that linker plasticity is conducive to conformational transitions required for substrate handoff and/or recycling and tunes RssB in response to different stress inputs, which, in the case of IraD, involves stabilizing RssB into a closed inactive form. More generally, structural modulation of interdomain linkers may be a widely used strategy for tuning response regulators in bacteria and possibly fungi and plants, which also rely on response regulators for signal transduction.

Results

Structural overview of the IraD·RssBD58P complex

To gain insight into the mechanism of regulation of the RssB adaptor protein by an antiadaptor protein, IraD, we determined the structure of the RssB·IraD complex. We used the nonphosphorylatable RssBD58P variant for cocrystallization with IraD with the expectation of reducing the heterogeneity of the complexes. This expectation was based on previous work that showed that RssBD58P was more susceptible to IraD inhibition than wild type (Battesti et al. 2013) and suggested that IraD may specifically target and stabilize the “off” (nonphosphorylated) conformation of RssB (Fig. 1A). The RssBD58P variant has been shown to be active in vivo and in vitro, although it is somewhat less active than wild type (Battesti et al. 2013). Additionally, we used a truncated proteolytically stable form of IraD: IraDTrunc (residues 17–130). IraDTrunc is competent for binding both RssB and RssBD58P, as determined by pull-down assays and size exclusion chromatography (SEC) (Supplemental Fig. S2A,B). IraDTrunc also inhibits σS degradation in vivo and in vitro in a reconstituted degradation assay using purified proteins (Supplemental Fig. S2C–F). Using cells coexpressing RssBD58P and IraDTrunc, we isolated an in vivo assembled complex of RssB and IraDTrunc. We determined the structure of this complex using single-wavelength anomalous diffraction at the K edge of selenium at a resolution of 2.6 Å followed by refinement against a higher-resolution (2.0 Å) data set (Table 1). The crystal asymmetric unit featured one compact trefoil-shaped heterodimer (Fig. 2B), consistent with our SEC data (Supplemental Fig. S2B) and a previous report that indicated a 1:1 stoichiometry of IraD to RssB (Micevski et al. 2015).

Table 1.

Crystallographic data and refinement statistics

Figure 2.

Overall architecture of the IraDTrunc·RssBD58P complex. (A) Schematics of E. coli RssB and IraD domain architecture. Indicated are the phosphoacceptor residue D58 and mutations in the RssB switch helix (α8) that lead to resistance to IraD in vivo and/or in vitro (Battesti et al. 2013) characterized in this study. Also shown is IraD, including the IraD recognition helix (a3) and residues mutated here. Dotted lines indicate the boundaries of the RssB switch and IraD recognition helices. (B) Front view of E. coli IraD (residues 18–124; green) bound to the RssBD58P variant (residues 3–337; color-coded as in A). Disordered residues that could not be built reliably into electrondensity are indicated by a dashed curve. P58 (D58 in wild-type RssB) is shown in red stick representation. (C,D) Side view of RssB. The IraD chain has been removed to allow a view of the core of RssB, composed of α1, α5, α8, and α9.

Two lobes of the trefoil belong to RssB and are formed by its two domains: the α/β NTD (RssB NTD), which resembles other receiver domains and contains the phosphotransfer site at the end of strand β3, and the CTD (RssB CTD), which is formed by two β sheets of mixed polarity buttressed on one side by α helices α8 and α9 (Fig. 2B–D). The complex is organized around a core of helices (Fig. 2C,D) that pack against each other through dry interfaces. The receiver domain from full-length RssBD58P overlays well (RMSD of 0.6 Å) over the previously determined structure of the isolated wild-type NTD containing the phosphoacceptor aspartate (PDB ID 3EOD, deposited by I Levchenko, RA Grant, RT Sauer, and TA Baker, unpubl.). However, in our structure, the β3−α3 and β5−α5 loops are well ordered, while in the isolated NTD, they remained unmodeled (Supplemental Fig. S3A,B). Subtle differences affecting the phosphorylation site residues exist, but, due to sequence divergence in switch regions between RssB and well-understood response regulators such as CheY, we exercise caution in interpreting them.

Despite low sequence identity (<10%), the RssB CTD shares strong structural similarity to PPM phosphatases, such as bacterial sporulation phosphatase SpoIIE, consistent with previous predictions (Galperin 2006). The RssB CTD similarity to phosphatase domains extends beyond general architecture and has relevance for RssB regulation. Like the RssB CTD, SpoIIE contains two switch helices embedded within the phosphatase fold, which connects to a regulatory domain via a long α helix (Supplemental Fig. S3C). The switch helices undergo rotation during SpoIIE activation to promote dimerization (Supplemental Fig. S3C,D), donate functional groups, and complete the Mn2+ coordination sphere required for catalysis (Bradshaw et al. 2017). Unlike SpoIIE, E. coli RssB is a pseudophosphatase, as it lacks some of the catalytic aspartates critical for Mn2+ binding. However, an intriguing feature of both proteins is the formation of a hydrophobic network of residues in the interdomain linker and switch helices (Supplemental Fig. S3E). In SpoIIE, these form part of the dimerization interface and are key to function (Bradshaw et al. 2017). Equivalent residues (V120, L124, M128, L139, L222, and L226) (Supplemental Fig. S3F) are not stringently conserved between RssB and SpoIIE but largely maintain their hydrophobicity to establish an extensive long-range network of hydrophobic contacts that link switch helix α8 in the RssB CTD to helix α5 in the RssB NTD. It is striking that a cluster of substitutions in this region (L214H, A216T, L218V, A221T, and Q254K) leads to resistance to IraD regulation but not compromised IraD binding in vivo or in vitro (Battesti et al. 2013). They likely cause a shift in the conformational equilibrium of RssB toward the activated “on” form with high affinity for σS, similar to what is normally brought about by phosphorylation (Fig. 1A).

The third lobe of the trefoil is formed by IraD, which contacts both RssB lobes via a well-ordered recognition helix (a3). Additional contacts with the RssB NTD are via IraD a2 (Fig. 2B). Overall, the IraDTrunc·RssBD58P complex buries a total surface area of ∼2100 Å2, consistent with its stability in solution (Janin et al. 1988). The IraD fold (an antiparallel β sheet flanked by three α helices) resembles the T4 phage protein Gp25 homodimer, with which it shares 15% identity. Gp25 homologs are required for sheath polymerization and attachment of the sheath to the phage baseplate (Basler et al. 2012). Gp25 is structurally related to baseplate components of type VI secretion systems, such as P. aeruginosa HsiF, but details of the interactions of Gp25-like proteins with binding partners have not been defined at atomic resolution. Interestingly, both Gp25 and IraD contain a strictly conserved EPRI motif also found in HsiF (Supplemental Fig. S3G). In HsiF, the conserved glutamate of the EPRI motif is important for type VI secretion in vivo (Lossi et al. 2011), but the precise mechanism behind the functional defect of the variant with a glutamate substitution remains unknown. In crystallized Gp25, the EPRI motif together with a1, the a1–a2 loop, and a3 is part of the dimerization interface (Supplemental Fig. S3H). In IraD, the EPRI motif forms a loop connecting helix a3 to strand b1 (Fig. 2B), both of which make contact with RssB. We note that the region of highest divergence between IraD and Gp25 is the N terminus, which features a long tail in IraD susceptible to proteolysis in vivo; this was deleted in our IraDTrunc construct. This region likely samples multiple conformations. Although not directly at the interface with RssB, this tail appears to modulate RssB inhibition, since IraDTrunc is a more potent inhibitor of RssB than IraD in vitro (Supplemental Fig. S2F).

When bound to IraD, RssB assumes a closed conformation

An unanticipated feature of the IraDTrunc·RssBD58P complex is the compact conformation of RssBD58P (Fig. 2B–D). Despite significant efforts, free E. coli RssB has long resisted structural analyses; therefore, we compared our structure with that of the previously crystallized P. aeruginosa homolog. In our structure, the linker folds into two short helices (α6–α7) (Fig. 2B) that, together with α5, juxtapose the RssB NTD to the CTD. Residues 135–137 preceding α6 were disordered and could not be modeled.

To establish whether the closed conformation seen for crystallized RssBD58P is also populated in solution and in the absence of IraD, we probed conformational dynamics in solution using interdomain disulfide cross-linking. We engineered an RssB variant, RssBCC carrying the S29C and E229C substitutions and the native D58 residue rather than the D58P substitution used for crystallization (Fig. 3A). RssBCC is active, as assessed by its ability to participate in σS degradation in vivo, bind IraD in vivo (Supplemental Fig. S4A,B), and act with ClpXP to degrade σS in vitro (Supplemental Fig. S4C; Supplemental Table S1). We had the following expectation: When bound to IraD, RssBCC will assume a closed conformation that will cross-link the RssB NTD and CTD in the presence of atmospheric oxygen and a catalyst, such as copper (1,10)-phenanthroline (CuPhen). The open conformation observed in the P. aeruginosa homolog is incompatible with formation of the engineered intramolecular disulfide, while the intercysteine distance in IraD·RssBCC is predicted, on geometry considerations, to be conducive to disulfide formation (Fig. 3A,B). In the presence of CuPhen and IraD·RssBCC, we observed clear formation of a faster-migrating electrophoretic species (Fig. 3C, lane 7). This was not apparent with either wild-type or single-variant complexes (Fig. 3D–F, lane 7). When we compared cross-linking of IraD-bound and free RssBCC, we noticed less intramolecular cross-linking in free RssBCC (∼55% vs. 18%) (Fig. 3G, lanes 8,9).

Figure 3.

IraD stabilizes a closed and novel RssB conformation. (A) Model of open RssB obtained using Phyre2 (Kelley et al. 2015) based on the P. aeruginosa protein model (PDB ID 3EQ2) showing the location of the two cysteine substitutions S29C and E229C in yellow CPK models, which are predicted to be too far apart to cross-link. Domains are colored as in Figure 2A. (B) View of the closed IraD-bound RssB conformation highlighting a Cα–Cα distance compatible with a disulfide (Cα–Cα distances range from 5.2 to 6.1 Å) (Trivedi et al. 2009). Domains area colored as in Figure 2A, and residues substituted for cysteine are shown as yellow CPK models. (C–F). Nonreducing SDS-PAGE of CuPhen-catalyzed oxidation reactions with the in vivo assembled purified IraDTrunc·RssBCC complex (C) and wild-type (D) or single Cys variant RssB (RssBS29C [E] or RssBE229C [F]) complexes. Purified protein was first reduced with TCEP, then subjected to oxidation at CuPhen concentrations ranging from 0 to 20 µM for 20 min at room temperature, and blocked with iodoacetamide (IA). A doublet is present only in oxidized IraDTrunc·RssBCC. Boxed bands are shown at higher magnification as insets in the top right corners of the gels. (G) Nonreducing SDS-PAGE of oxidation reactions with purified free and IraD-bound wild type and RssBCC and 15 µM CuPhen. Boxed RssB bands are shown at higher magnification in the top right corner.

Together with LC-MS/MS analysis of in-gel trypsin/AspN-treated IraD·RssBCC (Supplemental Fig. S4D), these data indicate that the engineered disulfide is not only intramolecular but also specific. We also observed intermolecular RssB cross-linking with free RssB and RssBCC, resulting in detection of a minor species consistent with the molecular weight of a dimer (Fig. 3G, lanes 9,10). Thus, we conclude that in the presence of IraD, the closed conformation seen in crystallo also occurs in solution and that the D58P mutation is not required for this. These data also indicate that free RssB exists in a conformational equilibrium of open and closed forms that are predominantly monomeric but can also self-associate into scarce transient dimers.

IraD W64 is key for complex formation

We next sought to define IraD residues critical for RssB recognition. IraD straddles both domains of RssB and interacts primarily through a2 and the recognition a3 helix (Figs. 2B, 4A). The recognition helix is wedged in a groove in between the RssB NTD and CTD and makes contact with the base of the linker helix α5 and β5. The binding interface largely overlaps with the α4–β5–α5 face of RssB (referred to here as the 4-5-5 face in agreement with response regulator nomenclature), which in many response regulators supports dimerization and/or association with binding partners (Gao and Stock 2010). The area buried by IraD binding to the RssB NTD is significantly larger than that buried by the RssB CTD (∼1330 Å2 vs. ∼830 Å2), consistent with two-hybrid analyses and pull-down assays that showed that IraD interacts with the isolated NTD but not the isolated CTD (Battesti et al. 2013; Micevski et al. 2015). Interactions between the RssB NTD and IraD are provided by invariant R115 and D104, which are involved in hydrogen bonds and cation–π interactions with IraD F75 and in hydrogen bonds and salt bridges with IraD R67, Q68, and W64. CTD interactions are conferred primarily by Q254 and N256, which bond to IraD R70, S71, and Q84 (Fig. 4A).

Figure 4.

IraD residues W64, R67, and R70 serve as a recognition triad. (A) IraDTrunc·RssBD58P, colored as in Figure 2B, with a close-up view of the IraDTrunc–RssBD58P interface. Residues targeted for mutagenesis and P58 are shown as sticks. P58 is colored red. (B) Quantification of pull-down assays of wild-type RssB with purified His6IraD mutants. RssB binding was normalized to the amount of wild-type RssB bound to wild-type His6-tagged IraD. A sample SDS-PAGE gel is shown in Supplemental Figure S5A. Statistical significance (P-value < 0.05) was established using a one-way ANOVA test. Shown are means with SD. n = 3. (****) P < 0.0001. (C,D) Quantification of σS turnover in strain AB007 (low rssB) expressing inducible wild-type and mutant IraD alleles and physiological levels of RssB (C) and strain AB011 (rssB) overexpressing RssB (D). IraD was expressed from induced pQE-IraD, and mutant derivatives were expressed from pST5001–pST5006. Cells were treated with tetracycline, and the remaining σS during the chase was determined as described in the Materials and Methods. Shown are means with SD. n = 3. (E) Inhibition of RssB-dependent σs degradation by wild-type or variant IraD in vitro. σS degradation by ClpXP was performed as described in the Materials and Methods in the presence of 0.1 µM ClpX, 0.2 µM ClpP, 2.4 µM Alexa488-σS, 0.1 µM RssB, and IraD wild type or variant, as indicated. Data from three replicates are presented as mean ± SD. Some error bars are obscured by plot symbols.

Previous studies had demonstrated interactions of IraD with the RssB NTD but not with the RssB CTD (Battesti et al. 2013; Micevski et al. 2015). Our structure, however, suggests important contacts of the recognition helix with both the NTD and the CTD. Using pull-down assays carried out with purified RssB and His-tagged IraD variants, we established that IraD W64, R67, and R70 are critical for the interaction and that the W64R substitution variant is the most defective in interaction (Fig. 4B; Supplemental Fig. S5A). The functional impairment observed here is likely not due to improper folding, as demonstrated using circular dichroism (Supplemental Fig. S5B). We then assessed the function of selected IraD variants in vivo and in vitro. The in vivo assay (see the Materials and Methods) relied on quantifying σS degradation over time following inhibition of protein synthesis via Western blot analysis of induced cells transformed with an IPTG-inducible high-copy plasmid encoding for wild-type or variant IraD carrying substitutions in the critical IraD residues. We carried out these studies in two different strains: AB007 (rssB+) expressing native levels of RssB and AB011 (rssBUP) containing constitutively higher levels of RssB (Bougdour et al. 2006). In both strains, the W64R substitution led to a pronounced functional defect, resulting in little σS stabilization (Fig. 4C,D; Supplemental Fig. S5C). We note that these data were obtained under conditions of IraD overproduction, which is likely to suppress partial defects. The other substitutions caused more subtle defects, which were apparent only in the rssBUP strain at higher levels of RssB. These results are consistent with the multiple interactions of W64 with RssB I94, V105, and D104 observed in our X-ray structure as well as our pull-down assay (Fig. 4B; Supplemental Fig. S5A). To probe IraD function in vitro, we monitored IraD's ability to inhibit degradation of σS in the presence of ClpXP, RssB, ATP, and acetyl phosphate (Fig. 4E), which supports RssB phosphorylation (Bouché et al. 1998). At concentrations of IraD that completely inhibit σS degradation, we observed that IraDW64R inhibited degradation ∼20% (Fig. 4E), consistent with the severe binding defect of this variant in vitro (Fig. 4B) and its lack of function in vivo (Fig. 4C,D). There was also partial inhibition by IraDR67A and IraDR70A both in vivo (Fig. 4C,D) and in vitro (Fig. 4E). Thus, we conclude that R67 and R70 and particularly W64 are important for RssB recognition.

RssB-D104 is key for RssB:antiadaptor complex formation

To dissect binding determinants on RssB, we targeted for mutagenesis residues seen at the IraD–RssB interface within both the RssB NTD (D104 and R115) and the RssB CTD (Q254 and N256), as shown in Figure 4A, and carried out tests of these mutant proteins in vitro and in vivo (Fig. 5). From initial pull-down experiments with lysates coexpressing His-tagged IraD and either wild-type or variant RssB, D104 was identified as critical for the interaction with IraD. Slightly decreased interactions were observed with RssBQ254A, RssBN256A, and the RssBQ254A N256A variant, and no significant decrease was seen with RssBR115A (Supplemental Fig. S6A,B). Further tests focused on RssBD104N, RssBR115A, and the double CTD mutant RssBQ254A N256A. These proteins were purified and shown to have CD spectra similar to wild type (Supplemental Fig. S6C). Their interaction with IraD was tested. RssBD104N showed reduced binding to IraD, RssBQ254A N256A had modestly reduced interaction, and RssBR115A had no detectable decrease in interaction (Fig. 5A; Supplemental Fig. S6D).

Figure 5.

RssB-D104 is key for antiadaptor interactions. (A) Quantification of pull-down assays using purified His-tagged IraD and wild-type or variant RssB. n = 3. Sample SDS-PAGE analyses are shown in Supplemental Figure S6D. Statistical significance was calculated using a one-way ANOVA test. (****) P-value <0.0001; (*) P-value <0.05; (n.s.) P-value >0.05. (B–E). Degradation of σS in cells expressing wild-type or variant RssB in a strain expressing only low levels of antiadaptors (B) or in the presence of specific antiadaptors (C–E). rssB alleles were overexpressed from pHDB3 plasmids (pW6, RssB; pST4002, RssBD104A; pST4003, RssBD104N; pST4004, RssBD115A; pST4007, RssBQ254A N256A; and pBA465, RssBL218V) (see Supplemental Table S3) in which RssB is expressed from its native promoter in low antiadaptor strain ST1005 (B), iraDUP strain AT477 (C), iraPUP strain ST1011 (D), and iraMUP strain BA290 (E) (Supplemental Table S4). (F) In vitro degradation of σS by ClpXP and RssB, RssBD104N, RssBQ254A N256A, or RssBR115A. End point assays within the linear phase of degradation were performed as described in the Materials and Methods in the presence of 0.1 µM ClpX, 0.2 µM ClpP, 2.4 µM Alexa488-σS, and wild-type RssB or RssB variant as indicated. Data from three replicates are presented as mean ± SD and were fit to the Michaelis-Menten equation using Prism. Vmax and KM values are shown in Supplemental Table S1. (G) In vitro IraD inhibition of RssB-dependent σS degradation by ClpXP in the presence of RssB, RssBD104N, RssBQ254A N256A, or RssBR115A. Degradation was performed as described for F using 0.1 µM wild-type RssB or RssB mutant and IraD as indicated. For each RssB protein, degradation in the absence of IraD was set to 100%; 36 pmol (±3 pmol) for RssB, 28 pmol (±1 pmol) for RssBD104N, 27 pmol (±4 pmol) for RssBQ254A N256A, and 28 pmol (±3 pmol) for RssBR115A. Data from three or more replicates are presented as mean ± SD. Some error bars are obscured by plot symbols.

Each of these RssB alleles was tested in vivo for activity and sensitivity to each of the antiadaptors. RssB mutants and the wild-type control were expressed on plasmid pHDB3 from the native rssB promoter, and the degradation of σS was determined after treatment of cells with chloramphenicol to block further protein synthesis. In addition to the alleles discussed above, RssBL218V, a previously studied RssB mutant found to be resistant to all three antiadaptors (Battesti et al. 2013), was included as a positive control. In the strain used in Figure 5B, not exposed to any stress, none of the antiadaptors should be significantly induced (Bougdour et al. 2008). All tested alleles except D104A were active for σS degradation in this strain (Fig. 5B; Supplemental Fig. S6E). To test for sensitivity to IraD in vivo, we also monitored activity in strain AT477 (iraDUP), in which iraD was placed under the control of a tunable arabinose-responsive pBAD promoter. Under these conditions, the degradation of σS was inhibited when wild-type RssB was expressed, but RssBL218V was resistant, as observed previously (Fig. 5C; Supplemental Fig. S7A; Battesti et al. 2013). We observed pronounced resistance to IraD for RssBD104N, RssBR115A, and RssBQ254A N256A, fully consistent with predictions from the cocrystal. We extended our in vivo analysis of antiadaptor sensitivity to IraP and IraM and determined that D104 also plays a key role in sensitivity to IraP but not IraM (Fig. 5D,E).

Purified RssBD104N, RssBR115A, and RssBQ254A N256A were also assessed for their function in σS degradation in vitro (Fig. 5F). As was the case in vivo, these mutants all remain competent for degradation of σS by ClpXP, although they showed slightly decreased activity compared with wild type (Fig. 5F; Supplemental Table S1). As expected, RssBD104N was resistant to IraD regulation (Fig. 5G). Consistent with the effect on IraD function, D104 in the RssB NTD makes multiple interactions with IraD, including van der Waals contacts with W64 and hydrogen bonds with Q68 and R67 (Fig. 4A), and is highly compromised in pull-down assays done with purified proteins (18% ± 3% binding relative to wild type) (Fig. 5A). Like wild type, RssBD104N retains the ability to be stimulated by acetyl phosphate in vitro (Supplemental Fig. S8A), suggesting that, at some level, interdomain communication and conformational switching between an “off” and “on” state are still possible in this variant. However, RssBD104N is more resistant to trypsin than wild type (Supplemental Fig. S8B,C). While the disruptive effect of D104 mutants on antiadaptor action may thus be principally explained by a direct disruption of the binding interface in the case of IraD, in the case of IraP, indirect conformational mechanisms may also be at play. Consistent with important roles for this region of RssB, the RssB NTD carrying a substitution of a neighboring residue (L106D) (Fig. 4A) on the 4-5-5 RssB face was shown previously to be deficient for interaction with IraD, while a variant carrying a substitution on the opposite side (L36D) on the 1-2-2 face (Fig. 2B) was still able to interact (Micevski et al. 2015). RssBQ254A N256A was modestly resistant to IraD (Fig. 5G), consistent with resistance in vivo (Fig. 5C) and a decrease in interaction (Fig. 5A); RssBR115A was not detectably resistant in vitro.

Thus, our data altogether indicate that IraD contacts with the RssB NTD revealed by our structure are critical for function and that regulation by IraD is also modulated by interaction with the RssB CTD.

The glutamate-rich interdomain linker is key for activity and modulates RssB inhibition

One interpretation of our results suggests that long-range interdomain communication and large conformational transitions in the RssB linker are important for antiadaptor function. Consistent with this, a W143R mutation at the N-terminal end of α6 in RssB (Figs. 2B, 6A) compromises IraD regulation (Battesti et al. 2013). In addition, RssBP150T carrying a substitution mapping to the loop connecting α6 to α7 (Fig. 6A) is also resistant to IraD regulation (Battesti et al. 2013). Notably, the α5–α6 loop also contains a low-complexity E134–E137 polyglutamate motif (Supplemental Fig. S1C). Polyglutamate stretches are known to serve as plastic linkages with potential for helix-to-coil transitions due to destabilization of helicity through side chain electrostatic repulsion (Daggett et al. 1991; Krejtschi and Hauser 2011). Consistent with this plasticity, the electron density in this region was of poor quality and could not be modeled reliably. We reasoned that by substituting the linker glutamates with residues with good propensity for α-helix formation, we might be able to stabilize helicity and favor a more open and rigid RssB conformation. We chose a triple alanine substitution, due not in small part to the fact that RssB also possess a low-complexity well-ordered AAAA motif (residues A151–A154) that is located at the N terminus of α7 and is entirely α-helical (Fig. 6A). We chose to mutate three of the four glutamates in the E134–E137 stretch, generating an EAAA motif that has been established to fold as a rigid helical linker in multiple contexts (Chen et al. 2013). This construct is referred to as RssBAAA. In parallel, we also created a RssBGSGS variant in which residues E134–E137 were replaced by a flexible GSGS sequence (Chen et al. 2013). While RssBAAA can promote σS degradation by ClpXP in an in vitro degradation assay, it exhibits significantly less activity than wild type both in vivo and in vitro (Fig. 6B,C; Supplemental Table S1). However, it binds σS as well as wild type in a bacterial two-hybrid assay (Supplemental Fig. S9A). This region may thus provide functionally critical sequence-specific interactions and/or confer the necessary plasticity for productive loading and handoff of σS to ClpXP.

Figure 6.

The interdomain linker is key for σS turnover. (A) The IraDTrunc·RssBD58P structure and a close-up view of the interdomain helical linker (magenta), with residues implicated previously in antiadaptor regulation (Battesti et al. 2013) shown as blue CPK models. Disordered residues, including the polyglutamate motif, are shown as a dashed line. IraD and RssB domains are colored as in Figure 2B. (B) In vitro degradation of σS by ClpXP and RssB, RssBAAA, or RssBGSGS. End point assays within the linear phase of degradation were performed as described in the Materials and Methods in the presence of 0.1 µM ClpX, 0.2 µM ClpP, 2.4 µM Alexa488-σS, and wild-type or variant RssB as indicated. Data from three replicates are presented as mean ± SD and were fit to the Michaelis-Menten equation using Prism. Vmax and KM values are shown in Supplemental Table S1. Some error bars are obscured by plot symbols. (C,D) Quantification of σS degradation in vivo in the presence of wild-type or variant RssB. Plasmid-borne rssB alleles were overexpressed using pHDB3 plasmids (pW6, RssB+; pAT329, RssBAAA; and pBA465, RssBL218V) (see Supplemental Table S3) in low antiadaptor strain ST1005 and iraD strain AT477 as described in the Materials and Methods. (E) In vitro IraD inhibition of σS degradation by ClpXP in the presence of wild-type RssB or RssBAAA. Degradation was performed as described for B using 0.1 µM RssB or RssBAAA and IraD as indicated. Degradation in the absence of IraD was set to 100%; 36 pmol (±3 pmol) for RssB and 18 pmol (±1 pmol) for RssBAAA. Data from three or more replicates are presented as mean ± SD. Some error bars are obscured by plot symbols. (F) Quantification of pull-down analyses with purified immobilized His-tagged IraD and purified wild-type RssB and RssBAAA. Shown are means ± SD of three replicates. (****) P-value <0.0001, Student's t-test. Sample SDS-PAGE is shown in Supplemental Figure S9D. (G) Schematic illustrating the role of RssB plasticity in σS turnover. RssB exists in a dynamic equilibrium of multiple conformations, including an “on” state with high affinity for σS and “off” states with low affinity for σS. Generally, “off” states have mutually inhibitory contacts between the receiver and effector domains, while “on” states generally feature fewer interdomain contacts (Gao and Stock 2010; Corrêa and Gardner 2016). IraD-bound RssB features a “closed” conformation with the interacting NTD and CTD. The interdomain linker containing the glutamate cluster (green) likely functions as a flexing joint to accommodate the binding of multiple inhibitors induced by specific stress signals as well as σS and ClpXP and possibly kinases. The use of a highly plastic linker featuring a glutamate cluster is similar to the mechanism for multiple target recognition in the calmodulin system (Supplemental Fig. S12).

The GSGS linker mutant RssBGSGS also affects RssB function (Fig. 6B). RssBGSGS has reduced ability to bind σS in a bacterial two-hybrid assay (Supplemental Fig. S9A) and is incapable of promoting degradation of σS in vivo (data not shown) or in vitro (Fig. 6B), highlighting the importance of the residues in the linker. RssBGSGS is much more susceptible to trypsin proteolysis than wild type (Supplemental Fig. S9B), consistent with decoupling of the RssB NTD from the CTD and a more solvent-accessible flexible linker. We observed no evidence of altered oligomerization relative to wild-type RssB in either of these constructs by gel filtration or chemical cross-linking, and wild-type RssB was determined to be monomeric by SEC-MALS (SEC with multiangle light scattering) (Supplemental Fig. S10). Thus, protection of wild-type RssB does not stem from oligomerization but rather distinct intramolecular rearrangements. In vivo, the reduced activity of RssBAAA makes sensitivity to antiadaptors more difficult to assess, but it appears to be sensitive to regulation by all antiadaptors (Fig. 6C,D; Supplemental Fig. S11). In vitro, RssBAAA is modestly more resistant to IraD inhibition than wild type (Fig. 6E). Finally, while overall expression of RssBAAA is lower than for wild-type RssB, complicating the assessment of IraD binding in lysates (Supplemental Fig. S9C), pull-down assays with purified protein demonstrate a modest defect in binding IraD (Fig. 6F; Supplemental Fig. S9D).

Discussion

General stress responses in all bacteria are pivotal for adaptation to the environment as well as for mediating host–microbe interactions and often connect signaling by phosphorylation to σ-factor regulation. This interplay often relies on modulating the pools of free σ via anti-σ factors, which prevent RNA polymerase holoenyzme formation by sequestration of the σ subunit, and antianti-σ factors, which in turn sequester the anti-σ (Gruber and Gross 2003). In both Gram-positives and Gram-negatives, kinases and phosphatases often regulate σ/anti-σ/antianti-σ networks (Gruber and Gross 2003). However, E. coli and related species use σS degradation instead as a major way of partitioning the transcriptional space when mounting a general stress response. This process is dependent on a response regulator (RssB), which plays a catalytic role; i.e., it is not degraded by ClpXP (Fig. 1A). Nonphosphorylatable substitutions at the phosphoacceptor location of RssB, such as D58P and D58A, support σS degradation in vivo and in vitro, suggesting that phosphorylation is not the major mode of regulation (Peterson et al. 2004; Battesti et al. 2013). Instead, most regulation is provided by antiadaptors IraD, IraP, and IraM, which titrate RssB to prevent σS destruction. These have been proposed to use distinct modes of RssB recognition (Bougdour et al. 2006; Battesti and Gottesman 2013; Battesti et al. 2013; Micevski et al. 2015). Our study elucidates the mechanism of RssB inhibition upon DNA damage by IraD and, importantly, provides a more general mechanistic framework for understanding the functional modulation of response regulators.

First, we demonstrated that IraD binding involves direct interactions with both the RssB NTD and CTD, dependent on a novel compact form of the linker that juxtaposes these domains (Fig. 2). The IraD interaction with both the RssB NTD and CTD was unexpected. Previous work had implicated only the NTD in interactions with IraD (Battesti et al. 2013; Micevski et al. 2015). Indeed, we found that contacts with the RssB NTD are more extensive and critical to function. We identified one residue within the receiver domain, D104, which, when mutated, abolishes IraD binding and significantly reduces function in vitro and in vivo (Fig. 5). On the IraD side, we pinpointed residues R67, R70, and especially W64 as playing key roles in RssB recognition (Fig. 4).

Second, we found that free RssB is a mixture of predominantly monomeric “open” and scarce “closed” forms enabled by the plasticity of its interdomain linker (Fig. 3). We emphasize that the “open” conformation may not be structurally identical to the conformation of the crystallized P. aeruginosa homolog, which may not functionally substitute for E. coli RssB (Battesti et al. 2013). The P. aeruginosa linker sequence is significantly different from that of RssB. First, it does not contain a glutamate cluster or a helix breaker, such as P150 (Supplemental Fig. S1C), arguing against an extended one-helix linker for E. coli RssB. In addition, the dimeric P. aeruginosa structure features a more rigid coiled–coil stabilized linker (Supplemental Fig. S1A). While RssB oligomerization can be detected by cross-linking (Fig. 3G; Supplemental Fig. S10A), it cannot be detected by SEC-MALS even at concentrations above physiological (Supplemental Fig. S10B) and does not appear to be phosphorylation-dependent, leaving the role of RssB oligomerization unclear.

ClpXP-dependent σS degradation consists of several sequential steps: σS binding to RssB, transfer of σS to ClpXP, unfolding of σS by ClpX, translocation of unfolded σS to ClpP for proteolysis, and RssB recycling. Regulation can thus be achieved at any of these steps and could involve more than just binding to RssB to prevent its association with σS. It could include modulating RssB–σS (Battesti et al. 2013) or RssB–σS–ClpXP interactions and/or RssB recycling. This is likely the reason behind the difference in sensitivity of some of our in vivo and in vitro assays, which we interpret as being due to competition with σS (Battesti et al. 2013) and possibly different extents of phosphorylation. While IraD and IraM interactions with RssB are largely phosphorylation-independent, IraP binding is disfavored by nonphophorylatable (D58P) substitutions and is competitive with σS for RssB binding; both σS and IraP may share partially overlapping RssB-binding interfaces, including or close to the 4-5-5 face (Battesti et al. 2013). Importantly, in agreement with previous work (Battesti et al. 2013), our data indicate that requirements for antiadaptor binding are not sufficient for antiadaptor sensitivity (Fig. 6). In the case of IraD, the RssB NTD confers enough affinity for IraD for formation of a complex stable enough to be detected by pull-down assays, yet variants carrying mutations beyond the NTD (RssBR115A and particularly RssBQ254A N256A) bind well but are partially resistant to inhibition in vivo (Fig. 5A,C). R115 is located at the N terminus of α5, which connects directly to the linker helices and likely modulates its conformation and NTD/CTD interactions. Q254 and N256 are located at the C terminus of the switch helix that packs against α5, the site of mutations (class II mutations) (Battesti et al. 2013) that lead to constitutive activation and bypass the need for phosphorylation. This is likely due to disruption of the helical RssB core (Fig. 2C,D), which is conducive to σS binding, which would normally be favored by phosphorylation (Battesti et al. 2013). Overall, these earlier findings and our data point to a two-step IraD-binding mode. First, IraD gets recruited via high-affinity contacts with the RssB NTD, which may then trigger closure of the molecule or capture of transient closed forms, leading to formation of fully inhibitory contacts with the RssB CTD. This closed inactive state does not resemble the P. aeruginosa-like open structure of RssB, as proposed earlier by Dougan and coworkers (Micevski et al. 2015), and is incompatible with the head-to-head dimerization seen in the P. aeruginosa homolog because of IraD docking to the 4-5-5 face.

Last, we show that the RssB polyglutamate motif is crucial for σS degradation even in the absence of antiadaptors. While substitution of the polyglutamate motif for a highly flexible GSGS sequence compromises σS degradation (Fig. 6) and σS binding (Supplemental Fig. S9A), introduction of a more rigid helical AAA sequence has a less severe effect, reducing activity in vivo and in vitro but not affecting σS interactions (Fig. 6B–D; Supplemental Fig. S9A). These data argue that the rigidity and reorganization of the linker might be particularly important for fulfilling stringent steric requirements for substrate loading/handoff to ClpXP and RssB recycling. Is the linker then important for regulation by antiadaptors? While RssBAAA appears to be slightly impaired in binding IraD in vitro in the absence of σS (Fig. 6F), it remains sensitive in vivo to all antiadaptors (Supplemental Fig. S11). Further structural studies of RssB, particularly in complex with σS, will be required to fully answer this question, but a simple explanation for this observation is that RssBAAA can still interact well enough with IraD through its 4-5-5 face in its intact receiver domain and that this binding event is sufficient to achieve partial inhibition.

To conclude, our study supports earlier proposals that antiadaptors use distinct mechanisms for inhibiting RssB (Battesti et al. 2013; Micevski et al. 2015) and places decades of genetics and biochemical work within a mechanistic framework. None of the previously described mutations that compromise IraD regulation (Battesti et al. 2013) map to the interface with IraD. Two mutations (W143R and P150T) (Fig. 6A) map to the linker sequence; others, the class II mutations, map to an ancestral switch element (Fig. 2A,B) not only present in Ser/Thr phosphatase folds but universally conserved in proteosomal-like proteases (Bradshaw et al. 2017). For activation, these proteases reposition their switch helices upon docking of a regulatory “cap” (Sousa et al. 2000). In SpoIIE, phosphatase activation involves dimerization partially mediated by these switches. In RssB, mutations in this area likely lead to intramolecular rearrangements of the helical core. We highlight the polyglutamate motif within the RssB linker as an important element for σs turnover (Fig. 6). Its collapse into shorter α helices and >70 Å displacement of the effector domain (Fig. 2B) are unusual among the response regulators that have been crystallized thus far (Gao and Stock 2010) but are reminiscent of the conformational changes in calmodulin. In free form, calmodulin consists of a dumbbell architecture with a central helical linker connecting Ca2+-binding EF hands (Babu et al. 1985), as shown in Supplemental Fig. S12A. Upon binding to multiple physiological targets, calmodulin undergoes compaction (Meador et al. 1993) via local unfolding of the central helix and its splitting into three shorter helices (Supplemental Fig. S12B,C). Strikingly, like RssB, calmodulin also features a highly conserved glutamate cluster within its helical linker, which has been observed either disordered or in a helical conformation (Supplemental Fig. S12B,C). This study and other work have led to the concept of linker plasticity as underlying the ability of calmodulin to assume conformations capable of recognizing dozens of sequence-dissimilar targets. Charge reversal of this glutamate cluster by mutagenesis differentially affected activation of various calmodulin targets (Craig et al. 1987), demonstrating its functional importance. Like calmodulin, RssB interacts with sequence-dissimilar antiadaptors, kinase, σS, and ClpX. Although only three antiadaptors have been identified, the list might be more extensive, reflecting the diversity of environmental stress signals. In fact, σS is stabilized by overexpression of transcriptional regulator AppY in strains deleted for iraD, iraM, and iraP (Bougdour et al. 2008), supporting the existence of additional RssB inhibitors. The RssB linker could then serve as an expansion joint that could be adjusted to achieve multiple conformations, possibly allowing for combinatorial and tunable interactions of RssB with kinase, antiadaptors, ClpX, and σS (Fig. 6G). Thus, altogether, our work not only reveals the mechanistic underpinnings of IraD-mediated RssB inhibition but extends the remarkable plasticity of response regulator architectures, particularly in complex with modulators, and points to RssB proteins as sharing elements of the calmodulin paradigm.

Materials and methods

Protein expression and purification

E. coli K12 iraD and rssB genes were amplified by PCR from plasmids pQE80-IraD and pT7-RssB (Battesti et al. 2013), digested, and ligated into pSKB2 (a generous gift from Dr. S. Burley) and pOKD5 (a generous gift from Dr. H. Wu) to generate plasmids pVF14 (His6IraD), pVF18 (His6IraDTrunc), pVF16 (RssBD58P), and pBD1 (RssB). IraD and RssB mutations were introduced by standard PCR-based site-directed mutagenesis. Plasmids were verified using DNA sequencing. Oligonucleotide sequences used for plasmid construction are listed in Supplemental Table S2. Bacterial plasmids and strains are listed in Supplemental Tables S3 and S4.

For overexpression of IraD•RssB and IraDTrunc•RssBD58P complexes, plasmids were transformed into BL21(DE3)-competent cells. Cells were grown in Lennox broth to an OD600 of 0.6–0.8 at 37°C in the presence of 25 μg/mL kanamycin and 50 μg/mL carbenicillin. This was followed by induction with 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) and further incubation for 4 h at 30°C with shaking. For expression of selenomethionyl-substituted protein, pVF16/pVF18 were cotransformed into Met-auxotroph strain B834(DE3) (EMD Millipore) and grown in SelenoMet medium (Molecular Dimensions).

Hexahistidine-tagged IraD•RssB protein complexes were purified by Ni2+-affinity chromatography followed by cleavage of the affinity tag with PreScission protease and further purification by anion exchange chromatography and SEC on a Superdex 200 Increase 10/300 column (GE Healthcare).

Crystallization, structure determination, and refinement

The IraDTrunc•RssBD58P complex was crystallized at room temperature using sitting drop vapor diffusion from a protein solution mixed at a 1:1 ratio with 0.2 M MgCl2, 7.5%–12.5% polyethylene glycol 8000, and 0.1 M Tris (pH 8.5). Crystals were cryoprotected in well solution supplemented with 30% ethylene glycol and flash-cooled in liquid nitrogen.

Data were collected at the Advanced Light Source beamline 5.02. The SAD data set was collected at the peak of the Se K-absorption edge. Both data sets were indexed, reduced, and scaled using HKL2000 (Otwinowski and Minor 1997). Anomalous scatters were located using HySS, and phasing was done using the Phaser-EP module in Phenix (Adams et al. 2002) followed by density modification with RESOLVE (Adams et al. 2002). This procedure yielded electron density maps in which ∼75% of the protein chains could be built automatically using Autobuild (Adams et al. 2002). The partial model was refined against the intensities of the higher-resolution data set, also with Phenix (Adams et al. 2002), using a riding hydrogen model. Reflection and refinement statistics are shown in Table 1.

Pull-down assays

For pull-down assays with IraD variants, 67 pmol of purified wild-type or variant His6IraD was mixed with 160 pmol of purified wild-type RssB at a concentration of 1.6 µM. Proteins were incubated for 20 min on ice in binding buffer (20 mM Tris at pH 8.0, 40 mM imidazole, 250 mM NaCl, 5% glycerol, 2 mM β-mercaptoethanol). The sample was applied to HisPur Ni-NTA resin (Thermo Scientific), equilibrated in binding buffer, and incubated for 20 min at 4°C with gentle rotation. Beads were washed three times with binding buffer, and then immobilized protein was eluted from the beads with elution buffer (20 mM Tris at pH 8.0, 500 mM imidazole, 250 mM NaCl, 5% glycerol, 2 mM β-mercaptoethanol). Samples were analyzed by SDS-PAGE followed by staining with Coomassie blue. For pull-down assays with RssB variants, His6IraD was coexpressed with wild-type or variant RssB, and pull-downs were performed on clarified lysates as described above.

For functional studies with purified RssB, we overexpressed wild-type and variant RssB constructs and purified them using nickel-affinity followed by tag cleavage with Prescission protease (where applicable) and SEC on a Superdex 200 10/300 column (GE Healthcare). Pull-down analyses were performed as above except that the concentration of RssB variants applied to beads was increased to 6 µM. Purified RssB constructs were verified using MS.

Interdomain disulfide engineering and cross-linking reactions

For engineering disulfide bonds that would trap RssB in the crystallographically observed closed conformation, we used Disulfide by Design (Dombkowski 2003). This analysis suggested S29 and E229 as good candidates for cysteine substitutions, and these were introduced in our wild-type RssB plasmid. Complexes containing RssB cysteine mutants or wild-type RssB were purified like the IraDTrunc•RssBD58P complex used for crystallization.

For cross-linking studies, purified protein was first fully reduced by incubation with 5 mM TCEP for 1 h on ice, and buffer was subsequently exchanged to 0.2 M NaCl and 20 mM Tris (pH 8). Cysteine oxidation by atmospheric oxygen was carried out immediately afterward at a protein concentration of 0.25 mg/mL in the presence of the CuPhen catalyst at concentrations of 0, 5, 10, 15, and 20 µM for 20 min at room temperature; quenched by alkylation with 50 mM iodoacetamide and SDS-loading buffer; and analyzed on 10% SDS-PAGE gels. Cross-linking in Figure 3G was performed with 15 µM CuPhen.

In vitro σS degradation assay

ClpX (Grimaud et al. 1998), ClpP (Maurizi et al. 1994), and σS (Zhou et al. 2001) were purified as described previously. σS was labeled with Alexa fluor 488 carboxylic acid, succinimidyl ester (Thermo Fisher) as recommended by the manufacturer. The degree of labeling was 2 mol/mol. Protein concentrations given are for σS, RssB monomers, ClpX hexamers, and ClpP tetradecamers. Degradation reactions (30 µL) were assembled in buffer B (20 mM Tris at pH 7.5, 100 mM KCl, 0.1 mM EDTA, 5 mM DTT, 5% [v/v] glycerol) containing 10 mM MgCl2, 2 mM ATP, 25 mM acetyl phosphate, 0.1 µM ClpX, 0.2 µM ClpP, 2 µM fluorescent σS, and RssB as indicated. After incubation for 20 min at 23°C, 70 µL of buffer B containing 15 mM EDTA was added to stop the reactions. Degradation products were isolated with Nanosep 10-kDa MWCO ultrafiltration devices (Pall; prewashed with 20 mM Tris at pH 7.5, 100 mM NaCl, 0.01% Triton X-100) by centrifugation at 14,000g for 10 min. Eluent fluorescence was measured using a Tecan M200 Pro plate reader with excitation and emission wavelengths of 490 and 525 nm, respectively. σS degradation by ClpXP in the absence of RssB (1.9 pmol ± 0.2 pmol) was subtracted from the data shown. Degradation reactions containing IraD were performed as described above using 0.1 µM wild-type RssB or RssB mutant and IraD as indicated. Data analysis was performed using Prism (Graphpad Software). The degradation end point of 20 min was within the linear phase of degradation as indicated by time-course experiments using the reaction conditions described above.

Monitoring in vivo σS degradation

To test the IraD wild-type or mutant activity, strain AB011 (expressing high levels of RssB) or AB007 (expressing physiological levels of RssB) was transformed with pQE80L-IraD, mutant derivatives of pQE80L-IraD, or pQE80L (vector) plasmid (Supplemental Table S2), and transformants were selected at 37°C on LB-agar containing 50 μg/mL ampicillin. A single colony was inoculated into LB containing 100 µg/mL ampicillin and grown overnight at 37°C with shaking. Overnight cultures were diluted 1:500 into fresh LB medium containing 100 µg/mL ampicillin. When cells reached an OD600 of 0.3, IraD was induced from the pQE80L IraD plasmid with 100 µM IPTG for 15 min. Global protein synthesis was inhibited by adding 400 µg/mL tetracycline. Where appropriate, 1-mL samples were harvested at various times (0, 10, 20, and 30 min), and TCA precipitation was performed. Samples were resuspended in 75 µL of sample buffer (containing 50 mM Tris-HCl at pH 6.8, 0.2% bromophenol blue, 20% glycerol, 2% SDS) after normalization to cell density.

To test wild-type or variant RssB activity, strains (listed in the figure legends and described in Supplemental Table S4) were transformed with pHDB3 RssB or mutant derivatives containing rssB under its native promoter or with the pHDB3 vector plasmid, and transformants were selected at 37°C on LB-agar containing 50 µg/mL ampicillin. A single colony was inoculated into LB containing 100 µg/mL ampicillin and grown overnight at 37°C with shaking. The overnight culture was diluted 1:500 into fresh LB medium containing 100 µg/mL ampicillin and grown to an OD600 of 0.3. Global protein synthesis was inhibited by adding 200 µg/mL chloramphenicol. One-milliliter samples were harvested at various times (e.g. 0, 5, 10, and 20 min). For measurement of IraD resistance, strain AT477 (ΔrssB iraDup), expressing IraD from a pBAD promoter, was grown to OD 0.2 and induced with 0.2% arabinose for 30 min before adding chloramphenicol. IraP and IraM are expressed from constitutive promoters in strains ST1011 (ΔrssB iraPUP) and BA290 (ΔrssB iraMUP) and thus were not induced. TCA precipitation was performed, and samples were prepared as described above.

Western blot analysis of σS accumulation

Samples were separated by electrophoresis on Nu-PAGE 12% Bis-Tris gel (Invitrogen). Proteins were transferred to nitrocellulose membrane using iBlot Gel transfer block (Life Technologies) and probed with polyclonal rabbit anti-σS (1:5000 dilution; Wickner laboratory stock), polyclonal rabbit anti-IraD (1:5000 dilution; Gottesman laboratory stock), and mouse monoclonal anti-EFTU (1:10,000 dilution; LSBio, LS-C128699). Blots were developed with CDP-Star substrate with Nitro-Block-II enhancer (Thermo Scientific) visualized with ChemiDoc (Bio-Rad). Quantification was performed using ImageJ software (National Institutes of Health).

Coordinates and reflection intensities have been deposited in the Protein Data Bank under PDB ID 6OD1.