3.1. The RssB/σs Complex Docks to the ZBD of ClpX
Initially, to validate the idea that RssB and SspB use a common docking platform on ClpX for delivery of its substrates to ClpXP, we performed a series of adaptor competition experiments in which we monitored the RssB-mediated turnover of σ
s, in the absence or presence of SspB (
Figure 1). Consistent with our initial hypothesis, the RssB-mediated turnover of σ
s was inhibited by the addition of SspB (
Figure 1a, filled symbols). Indeed, in the presence of 15 µM SspB (the highest concentration that we tested,
Figure 1a, filled circles) the half-life (t½) of σ
s was increased by ~9-fold (i.e., from ~2.5 min in the absence of SspB, to ~23 min). To ensure that the SspB-mediated inhibition of σ
s turnover was not simply due to steric hinderance of the pore of ClpX (by SspB) we also monitored the RssB-mediated turnover of σ
s in the presence of the XBR peptide from SspB (SspB
XBR) (
Figure 1b). Consistent with docking of RssB/σ
s to the XBR docking site on the ZBD of ClpX, increasing concentrations of the SspB
XBR inhibited the turnover of σ
s (
Figure 1b). Collectively these data suggest that interaction with the adaptor docking platform on ClpX (i.e., the ZBD) is essential for the RssB-mediated delivery of σ
s. However, during the early stages of this study the structure of RssB from
Pseudomonas aeruginosa was reported (PDB: 3EQ2) and it became apparent that the XBR motif of
PaRssB was integral to the protein fold and hence unlikely to be available for interaction with ClpX. Therefore, we modified our hypothesis regarding the mechanism of σ
s delivery to ClpX, although based on the above results, we maintained a focus on the role the ZBD in this process. Initially, we monitored the ClpP-dependent turnover of σ
s in the presence of either wild type ClpX (
Figure 1c, top panel) or a ClpX mutant (
∆ZBDClpX) which lacked the ZBD (
Figure 1c, lower panel). As expected, σ
s was rapidly degraded (t½ ~2.5 min) by wild type ClpXP in the presence of RssB (
Figure 1c, top panel), while deletion of the ZBD of ClpX (
∆ZBDClpX) completely inhibited the RssB-mediated turnover of σ
s (
Figure 1c, lower panel). Importantly, deletion of this domain does not affect the turnover of adaptor-independent substrates such as SsrA-tagged proteins [
12,
13] demonstrating that
∆ZBDClpX retains full unfoldase and ClpP-docking activities. Collectively, these data demonstrate that the ZBD of ClpX is essential for the turnover of σ
s, likely as a docking platform for RssB. Of note, following consumption of σ
s, full-length ClpX was cleaved into a smaller fragment (
Figure 1c, lanes 5 and 6, upper panel). Interestingly, and consistent with the findings of Houry and colleagues [
45], this “clipping” appears to involve the N-terminal region of ClpX as no such “clipping” was observed for
∆ZBDClpX (
Figure 1c, lanes 5 and 6, lower panel). Next, we asked the question, can the ZBD alone inhibit the RssB-mediated degradation of σ
s, and if so, which component does it interact with? To address these questions, we performed a series of competition assays, in which the RssB-mediated turnover of σ
s by wild type ClpXP was monitored in the presence of increasing concentrations of the ZBD of ClpX (
Figure 1d). Consistent with
Figure 1a, and docking of RssB to the ZBD of ClpX, the turnover of σ
s was inhibited upon addition of increasing concentrations of ZBD (
Figure 1d, filled symbols). These data clearly demonstrate that the ZBD of ClpX is sufficient for interaction with RssB (and/or σ
s), however, it still remained unclear which component(s) was directly involved in the interaction, or indeed how these protein(s) were recognized by the ZBD.
3.2. The C-Terminal Domain of RssB (RssBC) Docks Directly to the ZBD
Given the interaction between ClpX and RssB (or σ
s) is either transient and/or low affinity [
19], we decided to examine a possible interaction between the ZBD and RssB using chemical cross-linking. Specifically, we incubated the ZBD with or without RssB, in the presence of GA. Initially as a control, we examined the crosslinking of the His
6-tagged ZBD alone (
Figure 2a, lanes 1–6). As expected [
15], rapid dimerization of the ZBD was observed in the presence of GA (
Figure 2a, lanes 1–6). These data validate the experimental set-up of our system. Next, we monitored the formation of a complex between His
6-tagged ZBD (using anti-His antisera) and RssB (
Figure 2a, lanes 7–12). Consistent with a specific interaction between ClpX and RssB a crosslinked product was observed when the ZBD was incubated together with RssB (
Figure 2a, lanes 7–12). Significantly, this crosslink was absent from control experiments, i.e., ZBD alone (
Figure 2a, lanes 1–6) and more importantly the MW of the crosslinked band (~45–50 kDa) was equivalent to complex containing a single copy of RssB (~38 kDa) together with a dimer of the ZBD (~12 kDa). To verify the presence of RssB within this crosslink, the same reaction was probed with anti-RssB antisera (
Figure 2b, lanes 7–12). Taken together, these data suggest that (in the absence of σ
s) RssB can still interact with the ZBD of ClpX.
Next, in order to further dissect the interaction between the ZBD and RssB, we again performed chemical crosslinking of the ZBD, however in this case we monitored crosslinking in the presence of either RssB
C or RssB
N. Consistent with
Figure 2a, rapid dimerization of the ZBD was observed in the absence of RssB (
Figure 2d). Interestingly, dimerization of the ZBD appears to be enhanced in the presence of RssB (either full length or C-domain) (
Figure 2a,d, respectively, lanes 10–12). More importantly, addition of RssB
C also resulted in the appearance of two additional crosslinked bands (
Figure 2c, ~30 and 40 kDa), which appear to contain both RssB
C (
Figure 2c) and the ZBD (
Figure 2d). Based on the apparent MW of the crosslinked complexes, we speculated that the smaller complex (~ 30 kDa) represents a heterodimeric complex of RssB
C crosslinked to a single ZBD (within the RssB
C- ZBD
2), while the higher MW complex (~40 kDa) represents RssB
C crosslinked to the ZBD dimer (RssB
C-ZBD
2). Significantly, no cross-linking was observed for ZBD, in the presence of RssB
N (
Figure 2c, lanes 3–8). Finally, in order to identify a potential ZBD docking site on RssB we examined the binding of the ZBD to 13-mers from RssB using cellulose bound peptide library. Analysis of these binding data identified a total of 8 “strong” binding peptides, found in three regions of RssB (
Figure 2e), one was located within the N-terminal domain (peptides 20–23) and two within the C-terminal domain (peptides 81–82 and 90–91), which may represent docking sites for ZBD. Intriguingly, the putative RssB
XBR motif (peptide 109) was not recognized by the ZBD. Consistent with this finding, the RssB
XBR peptide (in contrast to the SspB
XBR peptide, see
Figure 1b) was unable to inhibit the ClpXP-mediated turnover of σ
s (data not shown). Notably, alignment of the “strong” binding peptides revealed a putative consensus binding motif for the ZBD (LKxh, where h = small hydrophobic and x = any amino acid, see
Figure 2f), which is highly similar to the XBR motif of SspB [
12,
46]. Therefore, based on these data, we propose that RssB docks to the ZBD of ClpX (in a competitive manner to SspB), not through the putative C-terminal XBR motif of RssB, but rather via an alternative XBR-like site, possibly defined by peptides 81–82 or 90–91 (discussed later).
3.3. Structure of the C-Terminal Phosphatase Domain of RssB
To gain structural insight into how RssB might interact with its various different co-factors (ClpX, anti-adaptors or substrate) we attempted to solve the structure of RssB (full-length and individual domains) either alone or in complex with several different partner proteins. Although we were unsuccessful in crystallizing full-length RssB in complex with a partner protein, we did solve the structures of both RssB domains. The structure of the RssB
C was solved at 2.1 Å by multiple isomorphous replacement (see
Table S4). The RssB
C domain structure is a mixed α/β-fold with five α-helices and 11 β-strands and can be further divided into a mediator/connector domain (residues 131–163) connecting RssB
N with RssB
C and the actual C-terminal protein phosphatase (PP2C) domain (residues 164–337). The N-terminal part or connector domain of RssB
C comprises three small α-helices (α6–α8) which are arranged with approximate angles of 90 degrees to each other (see
Figure 3a). These helices are connected to the core domain by salt bridges, hydrogen bonds and hydrophobic interactions, but the conformation of the very first α6-helix is likely misaligned due to the absence of the structuring RssB
N domain. The α8-helix connects the mediating α-helical elements with the C-proximal and structurally conserved phosphatase domain, which consists of 11 β-strands and two α-helices. The core structure of this domain is formed by a β–sandwich comprising two closely interacting β-sheets. One-half of this sandwich domain is formed by the anti-parallel β-sheet of five β-strands in the order β8-β9-β10-β11-β14 (
Figure 3a). The two long helices (α9 and α10) connect the β9 and β10 strands and attach to this side of the beta sandwich domain. The second anti-parallel β-sheet of the sandwich domain is formed by six β-strands in the following sequence: β6-β7-β16-β15-β12-β13. The two β-sheets are stabilized through a hydrophobic core structure formed by conserved aromatic and aliphatic residues.
Next we searched for structural homologs of this domain using the DALI server (
http://ekhidna2.biocenter.helsinki.fi/dali/). From this analysis we found the
E. coli RssB-IraD complex to be the most similar structure, with an r.m.s.d. of 2 Å (for 205 aligned residues, PDB-entry: 6OD1). The second closest structure with an r.m.s.d. of 3 Å is the RssB protein from
P. aeruginosa which was crystallized in different crystal forms either as the full-length protein or an individual domain (PDB-entries: 3F79, 3F7A, 3EQ2 and 3ES2, respectively). In superposition, this structure exhibits a surprisingly weak sequence identity of 16% and shows two additional helices connecting the last two β-strands (see
Figure 3c). The conserved residues (between
EcRssB and
PaRssB) are mostly located in the hydrophobic interface but also include two positively charged surface exposed patches, one of which (
Figure 3b, right panel) is centered around the putative phosphatase site that corresponds to the peptides 90 and 91 from the peptide library and hence may play a role in docking to the ZBD of ClpX.
RssB
C also shares structural similarity with RsbX, another protein phosphatase from
Bacillus subtilis. Although
BsRsbX shared an r.m.s.d. of 2.8 Å with
EcRssB
C (for 155 aligned residues; PDB-entry: 3W41), it only shares 9% sequence identity with
EcRssB
C in structural superposition [
47]. Interestingly, the active site of
BsRsbX contains a metal ion which is coordinated by three Asp residues, however only one of these Asp residues (Asp204) is conserved in
EcRssB. The RssB
C also shows structural homology to a variety of other phosphatase domains in the PDB with r.m.s.d. of ~3 Å and low structural sequence identity of ~10%. The superposition of RssB
C onto the serine/threonine phosphatase from
Streptococcus agalactiae as representative of this class of phosphatases is shown in
Figure 3d.
3.4. The N-Terminal Domain of RssB Adopts the Fold of a Two Component System Regulator
The structure of RssB
N was solved to 2.05 Å by molecular replacement using the PhoP response regulator structure (PDB-entry: 2PL1) as a search model. Overall, the structure consists of five repetitive (αβ)-elements which form a parallel β-sheet (β2–β1–β3–β4–β5–) sandwiched by two helices (α1 and α5) on one side and three helices (α2–α4) on the opposite side (
Figure 4a). A structural search of the PDB identified unpublished structures of
E. coli RssB
N (PDB-entry: 3EOD) and
P. aeruginosa RssB (PDB-entry: 3F7A) to be the most similar models with an r.m.s.d. of 0.4 Å and 1.3 Å, respectively. The next most similar target was the
E. coli RssB-IraD complex (PDB-entry: 6OD1). Superposition of RssB
N onto the RssB/IraD structure revealed an r.m.s.d. of 1.4 Å. While the core structure of this domain is well preserved (with the exception of small deviations around Asp58), there were clear changes in the conformation of the α5 helix across the structures. Interestingly, in contrast to most RRs this helix is somewhat unique as it is predicted to be the structural motif that connects the N- and C-terminal domains of RssB via an extended coiled coil structure (PDB-entry: 3F7A). From analysis of the B-factors of RssB
N (which resemble molecular flexibility) it became obvious that the last helix (α5), in particular, shows higher mobility (see
Figure S1). This is supported by the observation of limited hydrophobic contacts between this helix and the core domain structure. The reason for these high B-factors may either be functional or may have occurred due to the crystal packing where the two monomers are packed in a non-biological orientation. Importantly, mutations in this region have a dramatic effect on the regulation of substrate binding (discussed below).
Similar to several REC domains, our structure of RssB
N formed a dimer in the asymmetric unit. However, in contrast to many RRs the RssB
N dimer showed an anti-parallel orientation, which is mainly stabilized by β-strand augmentation via the terminal β2-strands. This orientation is different to the biologically active dimer of several RRs including PhoP from
E. coli. It is also distinct from the parallel arrangement observed in the structure of full-length
P. aeruginosa RssB protein, where the opposite face (formed by α4-β5-α5, here termed the 4-5-5 interface) contact one another. Therefore, to examine if the conserved 4-5-5 interface of
E. coli RssB was involved in dimerization, we modelled the dimer of
EcRssB
N using the dimeric, full-length
PaRssB structure (
Figure 4c). Using this template, the dimeric RssB
N model only yielded a small interacting surface area. This small interface may explain why the two domains interact only weakly in solution and may form alternative dimerization sites. Nevertheless, consistent with a role for the 4-5-5 interface in
EcRssB dimerization, mutation of L106 was previously shown to stabilize dimer formation as determined by chemical cross-linking experiments [
24]. To examine the possibility that a region equivalent to the coiled-coil region of
PaRssB also contributes to the weak dimerization of
EcRssB, we compared the primary sequence of both proteins. Consistent with secondary structure predictions and the recent structure of RssB (in complex with IraD [
23]),
EcRssB lacks most of the key coiled-coil residues found in
PaRssB (
Figure S1b). Collectively, these data suggest that weak dimerization of
EcRssB likely occurs through the 4-5-5 interface but does not involve formation of a coiled-coil domain.
Next we examined which residues on the surface of RssB
N were conserved. From this analysis we identified two conserved regions (see
Figure 4b). As expected, the first conserved region was formed by the loops L1, L3, and L5, which are located at the distal end of the molecule. This conserved patch includes Asp58 (the site of phosphorylation) and is surrounded by additional charged residues (see
Figure 4b and c; Glu14, Asp15, and Glu16) which together coordinate a magnesium ion. The second conserved region forms a groove (
Figure 4b), which centers on the C-terminal part of the protein (including the β5 strand). This groove includes Lys108, which forms salt bridges with Asp58 and Glu14 and is involved in transmitting the phosphorylation signal from the REC domain to the C-terminal output domain (see later).
3.5. The C-Terminal Domain of RssB is Required for σs Docking
Next we asked the question, how does RssB interact with σ
s? To address this question, we examined the interaction of σ
s with RssB (full length and individual domains) using a range of in vitro techniques. Initially we monitored the interaction between RssB and σ
s by co-immunoprecipitation (co-IP) using anti-σ
s antisera (
Figure 5a). As expected, immunoprecipitation of wild type RssB was observed in the presence of σ
s (
Figure 5a, lane 2). Importantly, the level of RssB recovered (in the presence of σ
s), was significantly higher than in the absence of σ
s. The small recovery of RssB (in the absence of σ
s) is likely due to a weak non-specific interaction of RssB with the beads to which the antisera was immobilized (
Figure 5a, lane 1). Next we examined the ability of RssB
N (
Figure 5a, lanes 3 and 4) or RssB
C (
Figure 5a, lanes 5 and 6) to interact with σ
s and determined the relative importance of each domain by quantitating four independent co-IP experiments (
Figure 5b). Similar to wild type RssB, and consistent with an important role for the C-terminal domain in σ
s interaction, RssB
C was co-immunoprecipitated in the presence of σ
s. Importantly, the level of RssB
C recovered (in the presence of σ
s) was significantly greater than in the absence of σ
s (
Figure 5b). In contrast to RssB
C, little-to-no RssB
N was recovered by co-IP using anti-σ
s antisera, suggesting that the N-terminal domain plays only a minor direct role in the interaction. Overall, these data clearly show that RssB
C plays a significant direct role in σ
s binding, however the apparent binding affinity of RssB
C (with σ
s) is substantially compromised given much less RssB
C was recovered by co-IP (in comparison to wild type RssB). Hence, other regions (or conformations) of RssB are likely required for direct interaction with σ
s. Consistent with these data, phosphorylation of RssB (on the N-terminal domain) enhances σ
s binding (see
Figure 6).
To validate these findings, we performed a series of experiments in which the bait was reversed. In this case, we immobilized Trx-H
6-RssB (wild type, RssB
N or RssB
C) to Ni-NTA-beads and incubated each column with untagged σ
s (
Figure 5c). As expected, σ
s was recovered from the pulldown using full length RssB (
Figure 5c, lane 3). Importantly, reversing the bait reduced the level of non-specific binding of RssB. Consistent with the co-IP experiments, a significant amount of σ
s was also recovered from column bearing immobilized RssB
C (
Figure 5c, lane 7), while only a very small amount of σ
s was recovered from the column bearing immobilized RssB
N (
Figure 5c, lane 5). Taken together, these data validate a direct role for the C-terminal effector domain of RssB in σ
s recognition (
Figure 2b). Finally, to confirm the identification of RssB
C as the primary docking site for σ
s, we performed a series of non-specific cross-linking experiments using GA, in which σ
s was incubated in the absence or presence of RssB
C. In order to identify σ
s specific crosslinks, we monitored the crosslinked proteins using anti-σ
s antisera (
Figure 5d). Consistent with the co-IP and affinity pull-down experiments, a specific cross-link product (~50 kDa), accumulated in the presence of σ
s and RssB
C (
Figure 5d, lanes 8–12), which was absent from the σ
s alone experiment (
Figure 5d, lanes 1–6). Importantly, the MW of this crosslinked product (~50 kDa) was equivalent to the theoretical MW of a heterodimeric complex of σ
s and RssB
C.
3.6. Mutations in RssBN Regulate Phosphorylation Dependent σs Binding
Based on the findings above, the effector domain (RssB
C) is primarily responsible for direct interaction with σ
s, however this domain alone, only appears to account for about one third of the total binding activity/affinity of σ
s (in comparison to wild type RssB). Furthermore, the receiver domain alone, also contributes little to the direct binding of σ
s, yet phosphorylation of this domain appears to play a significant role in the interaction. This suggests that substrate binding to RssB
C is either significantly enhanced (or stabilized) by phosphorylation of the N-terminal receiver domain or alternatively that recognition of σ
s extends beyond docking to RssB
C and likely includes a region of RssB, which is only exposed in the full-length protein and can be enhanced by phosphorylation of the N-terminal domain (such as the linker region between the two domains). However, a systematic in vitro analysis of the effect of phosphorylation and events that contribute to phosphorylation dependent conformational changes in RssB have yet to be examined in detail, as such the importance of RssB phosphorylation remains controversial. Therefore, we compared the rate of σ
s turnover in vitro in the absence or presence of AcP (
Figure S2). Consistent with published findings, the addition of AcP enhanced the RssB-mediated turnover of σ
s. The apparent K
m for the turnover of σ
s by unphosphorylated RssB was calculated to be ~150 nM, which decreased ~3-fold (K
m ~58 nM) in the presence of phosphorylated RssB (RssB~P). These data clearly demonstrate, that although phosphorylation of RssB is not essential for the turnover of σ
s it does alter its affinity for the substrate. Next, we generated a series of mutations within RssB
N to dissect the mechanism of activation by phosphorylation. As expected, mutation of Asp58 (the site of phosphorylation) abolished all phosphorylation-dependent activities of RssB. While the phospho-mimic mutant D58E enabled a “strong” interaction with σ
s (~75% of RssB~P) this binding was no longer activated by the addition of AcP (
Figure 6a, red bars). In contrast, replacement of Asp58 with Lys, not only resulted in a significant reduction to σ
s binding but also resulted in a loss of activation by phosphorylation (
Figure 6a, orange bars). Consistent with a loss of binding to σ
s by both mutant proteins, there was a corresponding loss in the RssB-mediated degradation of the substrate by ClpXP (
Figure 6d). Taken together these data suggest that phosphorylation of Asp58 stabilizes a “substrate-binding” conformation of RssB.
Next, to examine how the phosphorylation of Asp58 is signaled to the rest of the protein, we mutated Lys108 (which forms a H-bond with Asp58) (
Figure 6c). Consistent with mutation of Asp58, replacement of Lys108 (with either arginine, aspartate or alanine) not only reduced σ
s interaction significantly, but also abolished the phosphorylation dependent activity of RssB (
Figure 6a, purple bars, data not shown). These data suggest that Lys108 is, not only, critical for the formation of the “activated” complex—that recognizes σ
s with high affinity, but it also plays a critical role in sensing Asp58 phosphorylation. Finally, we speculated, that the signaling of Asp58 phosphorylation would be transferred to the C-terminal via an extended helix composed of α5 (from RssB
N) and α6 (from RssB
C).
Serendipitously, in our search for XBR-like motifs in RssB, we had already generated a mutation with the
116LREM
119 motif located in the α5 helix. To examine the role of this region in σ
s turnover, we replaced the charged residues (RE), central to this motif, with alanine (here termed RE/AA). Initially we examined the ability of wild type or mutant RssB, to bind σ
s and deliver it to ClpXP for degradation, in the presence and absence of AcP. Interestingly, although σ
s recognition by this mutant was completely abolished in the absence of AcP, substrate interaction (and degradation) was partially restored in the presence of AcP (
Figure 6b and
Figure S3). These data suggest that rather than playing a direct role in σ
s binding, the RE motif appears to play an important role in stabilizing the “substrate-binding” conformation of RssB. One interpretation of these data is that the RE motif guides RssB towards the substrate “engaged” state, however activation to the “high affinity” conformation still requires phosphorylation. Importantly, activation by phosphorylation can overcome this defect in σ
s binding. Next, to further dissected this motif, we generated a single point mutant in which Arg117 was replaced with Ala (RssB
R117A) here termed R117A. Consistent with the RE/AA mutant, substrate recognition by R117A was again absolutely dependent on phosphorylation (
Figure 6a, blue bars). Taken together these data suggest that Arg117 is a critical “switch” residue required for σ
s recognition, in the absence of phosphorylation. We speculate that this residue forms an important element in the formation of an “engaged” conformation, possibly via a key interdomain interaction. Notably, the R117A mutant retained the ability to “deliver” σ
s to ClpXP, both in the presence and absence of AcP (
Figure 6e, squares) despite no observable interaction with σ
s in the absence of AcP (as determined by pull-down, see
Figure 6a,b). These data suggest that (i) σ
s preparation only requires a transient interaction with RssB and (ii) activation of ClpX (by RssB) is required for σ
s turnover, and (iii) R117A retains the ability to interact with and activate ClpX for σ
s turnover.
Next, in order to better understand the interaction between σ
s and RssB we performed small-angle X-ray scattering (SAXS). Given the apparent conformational flexibility of RssB, we used the R117A mutant (in the presence of AcP) to limit different conformations of RssB. Preliminary SAXS experiments of RssB alone indicated the presence of high MW scattering particles, consistent with aggregated proteins. To remove these contaminating aggregated proteins, which interfered with the SAXS analysis, samples were fractionated by SEC in-line with SAXS data collection. The SEC elution profile (
Figure 7a), of Trx-RssB and σ
s alone indicated that aggregated protein (Ve ~52 mL,
Figure 7a) was well separated from monomeric Trx-RssB and monomeric σ
s, both of which eluted with a volume of ~90 mL (
Figure 7a, red and blue lines respectively). As expected, the Trx-RssB/σ
s complex eluted earlier (Ve ~82 mL) than either protein alone (
Figure 7a, black line) and importantly was clearly separated from the aggregated protein peak. The SAXS data obtained for Trx-RssB, σ
s and the Trx-RssB/σ
s complex (
Figure 7b, red, blue and black lines, respectively) were analyzed to reveal information about MW and shape (
Table S5). Based on an estimate of the volume of the scattering particle, termed Porod analysis, the MW of Trx-RssB
R117A was determined to be 49 kDa. This MW was similar to the theoretical mass of Trx-RssB, (determined from its amino acid sequence; 52 kDa), and hence is consistent with Trx-RssB forming a well-structured protein composed of globular domains. In contrast, the MW for σ
s (as determined by analysis of the SAXS data) was 45 kDa, which is higher than the theoretical MW of σ
s (38 kDa). This could indicate some self-association and/or a higher apparent volume due to significant conformational averaging. Both Trx-RssB and σ
s had similar maximum dimensions (~105 Å) and very similar
Rg values (~31 Å), consistent with both proteins being elongated and composed of multiple domains. The SAXS analysis of the Trx-RssB/σ
s complex indicated a MW of ~91 kDa (in comparison to the theoretical MW of 92 kDa), which is also in good agreement with the sum of the SAXS-derived mass of each component (45 + 49 = 94 kDa). It also suggests that the factors that caused a high apparent mass for σ
s (self-association and/or conformational averaging) may persist in the complex. However, the
Dmax of the Trx-RssB/σ
s complex is only ~25 Å longer than either of the individual components, suggesting these elongate components bind in a somewhat more globular arrangement.
Ab initio shape reconstruction of the SAXS data supports this with both isolated components being relatively narrow and elongate (
Figure 7c,d), while the complex is considerably wider at one end (
Figure 7e). An atomic resolution structure is not available for σ
s, however based on homology to other
E. coli sigma factors (free or in complex with RNA polymerase; PDB: 1SIG, 1L9U, 1KU2 and 3IYD), σ
s is expected to be highly flexible and likely to change its structure upon complex formation. Model building using rigid body refinement of linked structural fragments against the SAXS data was attempted, but in the absence of an experimental structure for σ
s and with long inter-domain linkers, no consistent solutions emerged. Thus, the X-ray crystal structures of RssB
N and RssB
C (see
Table S4) together with the structure of Trx, were manually docked into the Trx-RssB envelope, using the relatively small Trx domain as a reference point. Trx was used to orient the organisation of the complex and positioned in the narrow protrusion at one end of the envelope (RssB
N and RssB
C were too large to occupy this site), RssB
N was placed proximal to Trx, and RssB
C was positioned in the remaining distal unoccupied electron density (
Figure 7c). Assuming Trx-RssB maintains a similar domain arrangement in the complex, comparison of the ab initio shape envelope of the Trx-RssB/σ
s complex with free Trx-RssB was performed (
Figure 7e). This suggests that σ
s may wrap around the end of Trx-RssB, distal to Trx, making contact with both the N- and C-terminal domains of RssB. While other orientations may be possible, this arrangement, where σ
s binding seems to be mediated predominately through the C-terminal domain of RssB, is consistent with in vitro pull-down and GA cross-linking experiments.
Finally, we examined the mechanism of delivery of σ
s (by RssB) to ClpX(P). Previously, it was proposed that the N-terminal region of σ
s contains a ClpX-docking motif [
9,
48] and exposure of this motif (by RssB) was required for delivery of σ
s to ClpX. To confirm this possibility, we generated a deletion mutant of σ
s which lacked the first 50 residues of σ
s. Initially, we tested the ability of RssB to deliver wild type or mutant σ
s to ClpX for degradation by ClpP (
Figure 8a). Consistent with the findings of Hengge and colleagues [
48], deletion of the first 50 residues of σ
s abolished the RssB-mediated turnover of σ
s (
Figure 8a, compare lanes 1–5 with lanes 6–10). Next to ensure that removal of 50 residues from the N-terminus of σ
s did not compromise RssB interaction, we examined the ability of ΔNσ
S to compete with the RssB-mediated turnover of wild type σ
S. Importantly, the RssB-mediated turnover of σ
S was inhibited in the presence of equimolar concentrations of ΔNσ
S (
Figure 8a, blue squares). This turnover was further inhibited in the presence of a 2-fold (
Figure 8a, red triangles) and 5-fold (
Figure 8a, green diamonds) excess of ΔNσ
S. Indeed, in the presence of a 5-fold excess of ΔNσ
S the half-life of σ
S increased from ~2.5 min to ~ 20 min. In order to confirm that ΔNσ
S retained the ability to interact with RssB we performed a series of pull-down experiments (
Figure 8b). Consistent with the substrate competition experiments, equal amount of wild type or ΔN σ
S were recovered from immobilized RssB (
Figure 8b, compare lanes 4 and 7, respectively). Collectively, these data suggest that RssB is required for the release of the N-terminal ClpX binding motif, which is recognized by ClpX.