|Home | About | Journals | Submit | Contact Us | Français|
Gram-negative bacteria react to misfolded proteins in the envelope through a myriad of different stress response pathways. This cohort of pathways allows the bacteria to specifically respond to different types of damage, and many of these have been discovered to have key roles in the virulence of bacterial pathogens. Misfolded outer membrane proteins (OMPs) are typically recognized by the σE pathway, a highly conserved envelope stress response pathway. We examined the features of misfolded OMPs with respect to their ability to generate envelope stress responses. We determined that the secondary structure, particularly the potential to form β strands, is critical to inducing the σE response in an RseB-dependent manner. The sequence of the potential β-strand motif modulates the strength of the σE response generated by the constructs. By understanding the details of how such stress response pathways are activated, we can gain a greater understanding of how bacteria survive in harsh environments.
The envelope of Gram-negative bacteria mediates interactions with the environment and protects the bacteria from hostile surroundings. Because of the essential nature of the envelope, it is not surprising that Escherichia coli uses at least six stress response pathways to monitor and repair its envelope (10, 20, 21, 27). This cohort of pathways allows the bacteria to specifically and appropriately respond to different types of damage, as well as plays a role in the virulence of pathogenic strains of bacteria (6, 24, 30, 32). Specifically, the σE envelope stress response pathway has been demonstrated to be important for the virulence and survival of pathogenic Salmonella spp., Burkholderia cenocepacia, and Vibrio vulnificus and for the formation of biofilms in Pseudomonas aeruginosa (4, 6, 12, 16). Considering the importance of this stress response pathway to bacterial survival and pathogenesis, it is essential to gain a greater understanding of how the pathway is activated. By further understanding the specific activation signals of the stress response pathways, we can gain a greater understanding of the complex processes that bacteria use to survive in harsh environments.
The σE pathway is highly conserved among Gram-negative bacteria and is the only envelope stress response pathway known to be essential in E. coli (15, 27). This pathway responds to misfolded outer membrane proteins (OMPs) by a cascade of events initiated by the recognition of a misfolded OMP and results in the altered expression of over 60 genes, including the well-characterized periplasmic chaperone and protease DegP (1, 26).
Activation of the σE pathway is initiated when the three carboxyl (C)-terminal amino acids of misfolded OMPs directly interact with the periplasmic PDZ domain of the inner-membrane-spanning periplasmic protease DegS (14, 29). Exposure of the C termini of OMPs in the periplasm occurs when the OMPs are misfolded, whereas the C termini of the majority of properly folded OMPs are contained within the β-barrel structure of the proteins. The sequence of the three C-terminal residues determines the strength of binding with DegS and affects the level of σE activity in vitro, although sequences further upstream of the terminal phenylalanine (Phe) also play a role (14, 28). The C-terminal amino acid of the misfolded OMP is the most critical determinant of DegS binding: Phe is the most common residue at this position, and mutation of the C-terminal amino acid to aspartic acid completely prevents binding to DegS and subsequent activation of the pathway (31). OMP binding changes the structure of the DegS active site so that it can cleave the inner membrane protein RseA, an anti-sigma factor that holds the sigma factor RpoE at the inner membrane (28). The degradation of RseA results in the release of RpoE into the cytoplasm, permitting RpoE to alter the transcriptional profile of the cell to repair the outer membrane.
An additional level of regulation further modulates activation of the σE pathway. The periplasmic protein RseB is a dimeric protein that binds to RseA at the DegS cleavage site and inhibits cleavage of RseA by DegS (1, 11, 18, 19). It is not known how RseB is removed from RseA during σE activation, but in studies of OmpC and OmpX, it has recently been shown that RseB release depends on the C-terminal region of those OMPs (8), suggesting that OMPs contain a second σE activation signal. Characterization of this signal is the first step toward discovering the mechanism of the relief of RseB inhibition.
In this study, the C-terminal 50 amino acids of various OMPs were engineered as C-terminal additions to a periplasmic carrier protein in E. coli, a method that has been previously established for studies of envelope stress responses (8, 21, 31). The σE levels were monitored in rseB+ (wild type) and rseB null mutant (ΔrseB) backgrounds to search for RseB-dependent features. By altering the sequence of the OMP termini, we were able to identify the second OMP signal as a structural motif common to OMPs. Our findings uncover substrate characteristics that govern the exquisite specificity of the σE response so that it is activated only by misfolded OMPs.
E. coli strain ADA600, which has been previously described (5), was used as the wild-type strain to determine levels of σE activity. ADA600 is derived from the MC4100 background and contains an rpoHP3::lacZ chromosomal fusion that is specifically activated by the σE stress response pathway. The ΔrseB strain was created by P1 transduction using the rseB-knockout strain from the Keio collection (BW25113 background) as the donor (2, 22). All cultures were grown at 37°C with shaking in Luria-Bertani (LB) medium containing 250 μg/ml ampicillin and/or 50 μg/ml kanamycin as necessary.
Bacterial alkaline phosphatase (AP) and the OMP C-terminal sequences were PCR amplified from the genomic DNA of E. coli strain BW25113 using the appropriate primers, indicated in Table 1. The phoA gene, encoding AP, was cloned into the expression vector pTRC99a at the SacI and XbaI restriction sites, and then the DNA encoding the OMP C termini, including the stop codon, was inserted using the restriction sites XbaI and HindIII, fusing the OMP peptide to the C terminus of AP. Fusions containing nonnative OMP C-terminal sequences were created by incorporating mutations into the primers. The pTRC99a plasmid encoding wild-type AP, without a C-terminal OMP fusion, was used as a control. All of the constructs were verified by sequencing. The constructs were transformed into ADA600, and expression was induced with 25 μM isopropyl-β-d-thiogalactopyranoside (IPTG; Sigma-Aldrich).
Periplasmic fractions were obtained from cells that were grown to early exponential phase in the presence of 25 μM IPTG. The cultures were inoculated at an initial optical density at 600 nm (OD600) of approximately 0.03. When the cultures reached an OD600 of approximately 0.30, 2 ml of culture was collected, centrifuged at 10,000 × g for 2 min to pellet cells, washed twice in phosphate-buffered saline (PBS), and then treated with 200 μl of 10 mg/ml polymyxin B (Sigma-Aldrich) in PBS for 1 h at 37°C while it was shaken. After the treatment, the cells were pelleted by centrifugation at 10,000 × g for 2 min, and the supernatant, containing released periplasm, was collected. Periplasmic fractions and total cultures were analyzed by SDS-PAGE using a 15% acrylamide gel. Protein concentrations were determined using the Bradford assay following the manufacturer's directions (Invitrogen). To assess total protein content, gels were fixed for 1 h in fixing solution (10% methanol and 7% acetic acid), stained with Ruby stain (Invitrogen) overnight in the dark, and washed for 1 h in fixing buffer. Ruby-stained proteins were detected under UV light and quantified by densitometry using the ImageJ program (NCBI). Immunoblots were performed by transferring unstained gels to nitrocellulose membranes, blocking in Tris-buffered saline (TBS; 50 mM Tris-Cl, pH 7.4, 150 mM NaCl) containing 5% skim milk for 1 h at room temperature, incubating with anti-AP monoclonal antibodies (1:10,000 in TBS containing 1% Tween 20 [TBST]; Sigma-Aldrich) overnight at 4°C, washing six times in TBST for 5 min each, incubating with IRdye antimouse secondary antibodies (1:25,000 in TBST; LiCOR) for 1 h at room temperature, washing six times in TBS, and visualizing using an Odyssey thermal imaging system (LiCOR) according to the manufacturer's instructions. AP activity was also assessed for the periplasmic fractions using the Sensolyte pNPP colorimetric alkaline phosphatase activity kit (AnaSpec) according to the manufacturer's instructions.
The σE response pathway activity levels were determined by lacZ reporter assays, as previously described (5, 22, 25). Briefly, cultures were incubated in LB medium at 37°C until the cultures reached an OD600 of 0.03, induced with 25 μM IPTG, and allowed to grow until they reached an OD600 of 0.3. Samples from each culture (400 μl) were diluted to 1 ml total volume using ice-cold Z buffer (60 mM Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, 1 mM MgSO4, and 50 mM β-mercaptoethanol) and lysed by adding 1 drop of 0.1% SDS and 2 drops of chloroform and vortexing for 30 s. The lysed cultures were incubated in a water bath (27°C, 10 min), before adding 0.2 ml of 4 mg/ml ortho-nitrophenyl-β-galactoside in Z buffer. The reaction was stopped by adding 0.5 ml of 1 M Na2CO3, and the absorbances at 420 nm and 550 nm were measured. The Miller units were determined using the following equation: 1,000 × (OD420 − 1.75 × OD550)/(t × v × OD600), where t is the time between adding the substrate and stopping the reaction, v is the volume of culture used in the reaction, and OD600 is the density of the culture at the time that the sample was removed. Each sample was tested in a minimum of three independent experiments, and values were normalized to the value for the empty vector control for the appropriate background.
The Student's t test assuming unequal variance was used to assess statistical significance. A P value of <0.05 was considered statistically significant. Data represent the means ± standard deviations.
Activation of the wild-type σE depends on two parameters: activation of the DegS protease and removal of the RseB inhibitor protein. In strains lacking rseB, DegS activation is the only factor involved in σE activation. Therefore, differences in σE activity between wild-type and ΔrseB strains are due to the effects of RseB. Following this logic, we devised experiments to test for features of OMPs that are required for the relief of RseB inhibition. In order to elucidate the feature(s) of the upstream OMP sequences that regulates the extent of σE activation in a living cell, we used a chimeric protein fusion strategy that has been effective in similar studies (8, 21). We created expression plasmids encoding a periplasmic carrier protein, alkaline phosphatase (AP), fused with the 50 C-terminal amino acids of various OMPs. We chose to fuse peptides to AP instead of cytochrome b562, the previously used carrier protein, because AP is more easily tracked through activity assays and immunoblotting with commercially available antibodies. The synthetic constructs featured OMP sequences with previously described high (PhoE and OmpC), moderate (BtuB), and low (FecA and FepA) affinity for DegS, as well as one OMP that is predicted not to bind to DegS (GspD) on the basis of the lack of a terminal Phe (14).
We assessed σE activity in E. coli expressing the AP-OMP chimeric proteins and encoding a lacZ reporter for σE promoter activity. As expected, significantly increased σE activity was detected in cells expressing constructs terminating in a native Phe (Fig. 1A), which is consistent with previously published reports. We also tested an AP-GspD construct modified to terminate in YYF (GspD-YYF). OMP sequences terminating in YYF are known to bind strongly to and activate DegS (31). We found that GspD-YYF did not activate the σE pathway in the wild-type background (Fig. 1A). We considered whether it was RseB, a periplasmic protein known to bind to RseA and prevent its degradation by DegS, that prevented GspD-YYF from activating the σE pathway. It has been previously reported that RseB recognizes an upstream signal in OMPs and can prevent σE activation even in the presence of a DegS-activating signal (8). Indeed, when we tested the GspD-YYF construct in a ΔrseB background, it was able to activate the σE pathway (Fig. 1B). Together these results indicated that σE activation requires two specific and independent signals: one consisting of the C-terminal Phe and one that is upstream and RseB dependent. The native C terminus of GspD lacks both signals.
To determine if poor expression of GspD-YYF in the wild-type strain was the cause for the lack of σE activation, periplasm was collected from the strains expressing the various constructs. The AP activity in these periplasmic preparations was significantly above the levels detected in the vector control strain (data not shown), indicating that at least the AP portion of the construct was both present and properly folded. By further analyzing the periplasm using SDS-PAGE followed by immunoblotting for AP, we determined that expression of the GspD-YYF constructs produced proteins that migrated slower than the AP construct in both wild-type and ΔrseB strains (Fig. 2A). Therefore, we could conclude that the GspD-YYF construct was expressed in the periplasm of the wild-type strain but did not activate the σE pathway.
Analysis of the periplasmic preparations from all of the strains expressing constructs that activated σE revealed that these constructs migrated through the SDS-polyacrylamide gel faster than expected, similar to the migration of the unmodified AP carrier (Fig. 2B). Since the AP activity observed for all of these constructs suggested that the amino-terminal region was intact, the faster migration of the σE-activating constructs may have been due to cleavage of the C-terminal OMP tail, or it could be explained by a novel conformation of these nonnative proteins. If there was a cleavage event, we reasoned that it must have occurred after entry into the periplasm because σE activation by the C terminus did occur. Nevertheless, the unexpected migration patterns of OmpC, PhoE, BtuB, FecA, and FepA prevented us from determining the exact expression levels of these σE-activating constructs; therefore, we could not conclude whether one of these constructs was intrinsically better than another at activating the σE pathway.
To gain insight into the upstream, RseB-dependent signal, we compared the primary and secondary structures of the C-terminal 50 amino acids of GspD with those of the σE-activating OMPs OmpC, PhoE, BtuB, FecA, and FepA (Fig. 3A). Secondary structures for the σE-activating OMPs were obtained from the crystal structures deposited in the Protein Data Bank (www.pdb.org). Because the structure of the GspD C-terminal domain is not known, the secondary structure was predicted using the PSIPRED program (17). We noted that all of the σE activators contain at least one β-strand motif (BSM) longer than 6 amino acids, while GspD does not. Consequently, we hypothesized that the C-terminal BSM of particular OMPs was the RseB-dependent signal.
Like GspD, the C terminus of OmpA is also predicted to be devoid of BSMs. An OmpA construct was tested for σE activity, and like GspD, it activated the pathway only when both its C terminus was mutated to YYF and this construct was expressed in a ΔrseB background (data not shown). Although this profile agrees with the hypothesis that BSMs are needed for σE activation, the OmpA-YYF construct proved to be unstable in both the wild-type and mutant backgrounds, and we could not rule out the possibility that construct instability prevented the OmpA-YYF construct from activating the σE pathway in the wild-type background.
To directly test our hypothesis that BSMs could be involved in RseB-dependent σE activation, we engineered a series of constructs with alterations in the number or sequence of the BSMs (Fig. 3B). We found that deleting the sole BSM of the FepA construct completely abolished its ability to induce σE activity in the wild-type background but not in the ΔrseB background (Fig. 4A and B, FepA truncate). The OmpC construct contains 3 different BSMs. Deletion of the C-terminal BSM from the OmpC construct reduced, but did not abolish, σE activity in both wild-type and ΔrseB backgrounds (Fig. 4A and B, OmpC truncate), suggesting that one or both of the remaining upstream BSMs was able to relieve RseB inhibition. These data provided the first evidence that the presence of a BSM is critical to RseB-dependent σE signaling. Statistically significant signaling by the truncated OmpC construct further suggested that the exact sequence of the signaling BSM need not be that of the native, C-terminal BSM and that the signaling BSM need not be immediately proximate to the terminal Phe.
Next, we tested whether the C-terminal BSM of OmpC was sufficient for RseB-dependent activation of the σE pathway and whether its ability to form a linear strand was critical to this activity. We created a shortened AP fusion construct, using only the 20 C-terminal amino acids of OmpC, thus including only the C-terminal BSM of OmpC (OmpC20) (Fig. 3B). This chimeric protein was able to elicit a σE response in both wild-type and ΔrseB backgrounds (Fig. 4A and B). The linearity of the BSM was disrupted by a proline mutation at G362 (OmpC20 proline). OmpC20 proline was not able to activate the σE pathway in the wild-type strain but was able to activate it in the ΔrseB background (Fig. 4A and B). These data support the conclusion that a linear, C-terminal BSM is critical to RseB-dependent signaling. It should be noted that the BSMs in these fusions are unlikely to be part of a β sheet or barrel as they are not in the context of a full-length OMP, yet they can still serve as a recognition motif.
To further investigate the sequence specificity of different BSMs on signaling, we reengineered the fusion with GspD so that the C terminus of the GspD construct was replaced with each of the BSMs from the OmpC construct and a C-terminal DegS-activating motif (YQF) (Fig. 3C). The presence of any one of the three OmpC BSM sequences caused significantly higher σE activities than the vector alone (Fig. 5A, BSM1, BSM2, and BSM3). Although the BSMs activated σE to different extents in a wild-type background, the σE activity was similar for all three of the OmpC BSMs in a ΔrseB background (Fig. 5B). Similar σE activity levels could be the result of exceeding the detection limit of the assay. To ensure that we were within dynamic range of the reporter, the experiments were also conducted at lower induction levels (0 and 5 μM IPTG), and we found that activation remained similar between the constructs (data not shown).
We also constructed a chimeric protein incorporating 7 amino acids from a BSM of a cytosolic protein, MscS, again terminating with a C-terminal DegS-activating motif (YQF). The BSM from MscS is part of a parallel, cytoplasmic β barrel in the native protein (3). This construct activated σE to a significant but low extent in both backgrounds (Fig. 5A and B, MscS). Together, these results suggest that a variety of BSM sequences can function as the second signal and that they need not be from an antiparallel β barrel, as is found with OMPs. However, the exact sequence of the BSM plays a very important role in the strength of the σE response, either through direct interactions with RseB or because of differences in stability. Because these fusion constructs activated σE to a similar extent in the ΔrseB background, it is unlikely that the sequences affect their interaction with DegS; thus, the differences in activity levels in the wild-type background are most likely due to their interaction with RseB. These data conclusively show that BSMs are necessary for a σE response in an RseB-dependent manner.
On the basis of immunoblots of periplasm preparations from wild-type cells, we noted that the major species of the BSM-containing constructs (OmpC, PhoE, BtuB, FecA, FepA) migrated slightly further than the constructs that did not contain BSMs (GspD and GspD-YYF) (Fig. 2A and B). The different migrations could be due to either an altered conformation of the constructs or degradation. Incubation in 4 M urea did not alter the migration of the BSM-containing constructs (data not shown), suggesting that the altered migration is not due to the conformation of the constructs. The faster migration of BSM-containing constructs matched that of the untagged AP carrier protein, suggesting that all or nearly all of the OMP tag was removed from the carrier protein.
The different migration patterns were observed in both wild-type and ΔrseB backgrounds (Fig. 6A). This result indicated that RseB does not target the constructs for degradation and further showed that the differences in the levels of σE activation by the same construct in wild-type and ΔrseB strains are not due to differences in protein stability or folding. Interestingly, as shown in Fig. 6B, the altered migration of the construct did not correlate with σE activity levels. Specifically, the GspD-YYF construct showed a slower migration profile, even in the ΔrseB background, where it activates the σE pathway. Additionally, the OmpC construct migrated faster, even when the C-terminal Phe was mutated to Asp (OmpC F367D), which has previously been shown to prevent it from activating the σE pathway by disrupting the DegS-activating motif (31). The altered GspD constructs, BSM1 to BSM3 and MscS, also migrated faster than the native GspD construct and the GspD-YYF construct (Fig. 6B and data not shown). Taken together, these data suggest that the exposed, unfolded BSMs target the constructs for degradation. We analyzed the constructs expressed in a strain lacking degP and in other protease-deficient strains but were unable to identify a protease responsible for the altered migration (data not shown).
Our data show that the structural motifs of misfolded OMPs play a role in the σE response. Specifically, (i) a BSM within the misfolded protein is required for σE activation in a wild-type strain, (ii) the BSM does not need to originate from an antiparallel β barrel or from a transmembrane protein, and (iii) modulation of the σE response at least partially depends on the exact sequence of the BSM. We demonstrated that in a ΔrseB strain, a BSM is not required for σE activation but a terminal DegS-binding motif must be present. Unstructured BSMs may serve an additional role in targeting polypeptides for degradation independently of the σE pathway.
We considered how bacteria would benefit from a second σE-activating signal. Most notably, a second signal would increase its specificity. The σE stress response is quite dramatic, activating a regulon of over 60 genes (26); thus, its activation costs the cell a substantial amount of energy and causes a dramatic change in envelope characteristics. Using BSMs as the second signaling requirement would increase the specificity in the response to misfolded OMPs, likely distinguishing the activation of the σE stress response from the signaling of other envelope stress response pathways. OMPs form β barrels, making BSMs a common feature of these proteins. A BSM second signal would reduce misactivation of the σE pathway in response to denatured or proteolyzed proteins that happened to terminate in Phe, the C-terminal DegS-activating motif.
The exact mechanism by which RseB recognizes BSM signals remains unclear because it does not appear to be a simple binary interaction. RseB binds RseA at the DegS cleavage site, protecting it from proteolytic cleavage, and upon signaling, RseB likely is released from RseA, thereby allowing cleavage of RseA and activation of the σE pathway (19). An N-terminal domain of RseB has a potential protein-binding pocket that shares structural homology to the lipoprotein-binding protein LppX (18, 19). Previous studies have suggested that RseB may bind to misfolded proteins directly (9, 11, 13, 23), and it is tempting to speculate that the BSM directly binds to the LppX-like pocket, possibly by interacting with the saddle-shaped β-sheet structure (18). However, in vitro studies by others and ourselves using purified components have shown that a 10- or 20-amino-acid mimic of the OmpC C terminus was unable to remove RseB that was bound to RseA (7, 13; data not shown). Further, it has been shown that OmpC expression did not remove RseB from RseA in a ΔdegS sup+ strain (13). These results suggest that RseB interacts with the BSM indirectly, requiring other factors, either in the substrate polypeptide or in the envelope, to do so.
Although the BSM and DegS-activating C terminus are critical to σE activation, these are not the only factors that determine the strength of the stress response. Illustrating this, OmpC20 and BSM3 induce different stress response levels, despite having identical BSM and DegS-activating sequences (Fig. 4 and and5).5). The links between the carrier protein and the σE-activating factors of these two constructs differ in both sequence and length, and it is not clear which of these differences is important for σE activation. The amino acid sequence could provide binding sites that modulate σE activation, or a longer linking sequence could prevent steric hindrance by the AP carrier protein. However, it is clear that the activating strengths of the BSMs can be compared only if they are fused to the same carrier protein and separated by a similar distance, preferably by the same linker sequence. When comparing activation by different constructs, it is also important to note that DegS activation is affected by residues upstream of the C-terminal YXF motif (29). Thus, if BSMs are to be compared, they should be sufficiently upstream of the C terminus so that DegS activity is not affected. The wild-type σE activity of the OmpC truncate construct shows that the BSM does not need to be adjacent to the DegS-activating motif, and thus, the BSM can be completely independent of the DegS-activating residues.
BSM-containing constructs appeared to migrate faster than constructs that lacked BSMs in an RseB- and σE-independent manner (Fig. 6). Degradation of the OMP tag may cause these altered migration levels, as suggested by three lines of evidence. First, AP activity and antibody recognition are retained in response to these constructs, indicating that the N terminus remains largely intact. Second, 4 M urea did not affect the migration, suggesting that the proteins are completely denatured (data not shown). Third, using an in vitro assay, we found that the AP-OmpC construct that we purified from the periplasm was unable to activate DegS, indicating that the C terminus was altered (data not shown). We have not been able to identify a protease responsible for the altered migration levels; thus, we cannot conclusively determine that degradation is the cause. However, the BSM-dependent faster migration patterns offer further evidence that BSMs are recognized in vivo.
In sum, this study supports a model whereby the σE stress response pathway is stimulated by a dual interaction of a C-terminal Phe that binds DegS, as reported previously, and a BSM that relieves RseB-mediated inhibition. Recognition of the BSM signal by RseB allows DegS to cleave RseA, whereas without relief of RseB-mediated inhibition, a misfolded polypeptide cannot signal DegS to cleave RseA. This explains why OmpA-YYF and GspD-YYF did not activate σE in the wild-type strain but did activate a response in the ΔrseB mutant strain. It is not yet clear how RseB senses the BSM, but it seems that the sequence of the BSM is important, either through direct interactions or by affecting the stability of the misfolded protein, as exposed C-terminal BSMs also appear to target misfolded proteins to a pathway of partial degradation independently of σE activity. Consequently, this study has uncovered new principles governing the molecular specificity in the activation of bacterial envelope stress pathways.
We thank current and past lab members for helpful discussions.
This work was funded by grants R01AI064464 and R01AI079068 from the National Institutes of Health.
Published ahead of print on 9 September 2011.