Reversal of the ΔdegP Phenotypes by a Novel rpoE Allele of Escherichia coli

RseA sequesters RpoE (σE) to the inner membrane of Escherichia coli when envelope stress is low. Elevated envelope stress triggers RseA cleavage by the sequential action of two membrane proteases, DegS and RseP, releasing σE to activate an envelope stress reducing pathway. Revertants of a ΔdegP ΔbamB strain, which fails to grow at 37°C due to high envelope stress, harbored mutations in the rseA and rpoE genes. Null and missense rseA mutations constitutively hyper-activated the σE regulon and significantly reduced the major outer membrane protein (OMP) levels. In contrast, a novel rpoE allele, rpoE3, resulting from the partial duplication of the rpoE gene, increased σE levels greater than that seen in the rseA mutant background but did not reduce OMP levels. A σE-dependent RybB::LacZ construct showed only a weak activation of the σE pathway by rpoE3. Despite this, rpoE3 fully reversed the growth and envelope vesiculation phenotypes of ΔdegP. Interestingly, rpoE3 also brought down the modestly activated Cpx envelope stress pathway in the ΔdegP strain to the wild type level, showing the complementary nature of the σE and Cpx pathways. Through employing a labile mutant periplasmic protein, AcrAL222Q, it was determined that the rpoE3 mutation overcomes the ΔdegP phenotypes, in part, by activating a σE-dependent proteolytic pathway. Our data suggest that a reduction in the OMP levels is not intrinsic to the σE-mediated mechanism of lowering envelope stress. They also suggest that under extreme envelope stress, a tight homeostasis loop between RseA and σE may partly be responsible for cell death, and this loop can be broken by mutations that either lower RseA activity or increase σE levels.


Introduction
DegP is the major periplasmic protease in Escherichia coli [1][2]. Its proteolytic activity towards outer membrane proteins (OMPs) was initially observed against a mutant LamB protein with a temperature sensitive folding defect [3]. Subsequent studies showed an absolute or conditional requirement for DegP in cells either expressing folding-defective OMPs [4][5] or lacking a factor involved in OMP assembly [6][7][8][9].
Spiess et al. [10] first demonstrated that DegP possesses both protease and chaperone activities. A series of structural and biochemical studies on DegP provided important mechanistic clues [11][12][13][14]. These studies showed that the binding of unfolded substrate proteins to DegP triggers its reversible oligomerization into a cage-like structure and membrane association greatly influences DegP's assembly and activity. A full rescue of the temperature sensitive growth phenotype of DdegP requires DegP's protease activity [10]. The expression of degP is under the control of the s E and Cpx regulons [15][16][17][18], which together constitute the two major envelope stress response pathways [19][20][21]. Both pathways are required to elevate degP expression to overcome the potentially lethal envelope stress caused by aberrant OMP assembly [9].
Compensatory mutations have been isolated that overcome the temperature sensitive growth phenotype of cells lacking DegP. The first of this was isolated by Baird and Georgopoulos [22] by selecting for revertants that grew at 42uC. One of the revertants displayed a cold sensitive growth phenotype and was found to affect a gene named sohA (suppressor of htrA) [22]. The mechanism by which sohA reversed the growth phenotype of DdegP could not be determined. Prior to this study, Kiino and Silhavy [23] identified suppressor mutations of a LamB::LacZ hybrid protein in a locus they termed prlF and hypothesized that it was involved in some aspect of protein localization. The prlF gene, which turned out to be the site of the sohA mutation, together with yhaV, was recently determined to constitute a toxin-antitoxin module [24]. However, despite the new finding, the mechanism by which the sohA/prlF alleles overcome the growth defects of DdegP remains unknown.
The second DdegP suppressor mutation was obtained somewhat serendipitously and mapped to the cyaA gene [25]. It was concluded that the lower cAMP levels in the mutant cyaA DdegP background overcame the temperature sensitive growth phenotype by inducing the s E and Cpx stress regulons. While it remains to be determined exactly how reduced cAMP levels leads to up-regulation of the envelope stress response systems, another alarmone, ppGpp, was shown to up-regulate s E activity in a RseAindependent manner [26]. In a follow-up study, the authors concluded that ppGpp affects s E activity both directly, by influencing s E -dependent transcription, and indirectly, via altering competition for the core polymerase in favor of s E [27]. From these studies, it is apparent that the metabolic state of the bacterial cytoplasm also influences the envelope stress response systems.
In this study, we sought revertants of a DdegP strain also lacking bamB, a nonessential component of the BAM complex [28][29]. The simultaneous absence of degP and bamB confers a severe growth phenotype even at 30uC [8]. We reasoned that the absence of DegP in conjunction with defective BAM machinery would yield potentially new kinds of suppressors that could give a novel insight into the envelope stress response. Indeed, envZ was identified as one of the genes in which suppressor alterations improved the growth phenotype of the DdegP DbamB strain [30]. EnvZ and OmpR constitute a two-component signal transduction system [31], whose activity is modulated by a newly identified inner membrane protein, MzrA [30,32]. MzrA directly interacts with EnvZ and changes its enzymatic activities so as to elevate the steady-state levels of OmpR,P. Consequently, overexpression of MzrA or certain alterations in EnvZ constitutively activate the EnvZ/OmpR regulon. The activated EnvZ/OmpR regulon directly or indirectly (via inducing the expression of small regulatory RNAs) inhibits the synthesis of several OMPs and reduces envelope stress [30,33].
Here we report two new sites of suppressor mutations that overcome the growth defects of a DdegP DbamB strain. Mutations in rseA constitutively activate the s E regulon and reduce envelope stress in part by lowering the major OMP levels. In contrast, an rpoE mutation, which causes a novel duplication/truncation of the rpoE gene, does not reverse the growth defect by lowering the major OMP levels; instead, it appears to activate a proteolytic pathway that partly compensates for the loss of DegP.

Results
Mutations that overcome the temperature sensitivity phenotype of DdegP One of the ways to understand the consequences of a conditional lethal mutation is by isolating and characterizing revertants that have acquired so-called suppressor mutations and overcome the harmful effects of the original mutation. We have previously shown that a mutant simultaneously lacking degP and bamB displays a synthetic and conditional lethal phenotype at or above 37uC and grows poorly even at 30uC [8]. Because of this phenotype, the double mutant provides an ideal background in which to isolate suppressor mutations that reverse the growth defect, presumably by improving the OMP assembly environment and/or reducing envelope stress.
Liquid cultures of the DdegP DbamB double mutant were grown in LB at 30uC, diluted, spread onto LB agar plates, and incubated at 37uC until colonies appeared (between 24 and 36 hours). Temperature resistant colonies, which arose at a frequency of around 10 27 , were then purified at 30uC and suppressor mutants categorized based on growth robustness and OMP levels. Using these criteria, the suppressors were classified into three groups (Fig. 1). One group included suppressor mutations affecting envZ of the EnvZ/OmpR two-component regulon (Fig. 1, lane 4; [30]).
The second group of revertants were distinct from those carrying missense mutations in envZ in that the levels of all the major OMPs (Fig. 1, lane 2), not just that of OmpF as seen in the envZ mutants, were significantly reduced. Since this phenotype is reminiscent of an rseA null mutation, we determined the nucleotide sequence of the rseAB-rpoE region. Three different types of mutations were found in the rseA gene, resulting in (a) a W33R substitution, which was previously shown to directly affect the RseA-s E binding pocket [34], (b) a frame-shift mutation after the sixteenth codon of rseA, and (c) an IS1 insertion element at nucleotide 454 of rseA. In each case, the mutations presumably abrogated RseA's ability to sequester s E at the inner membrane, thereby elevating the s E -mediated envelope stress response. Thus, rseA mutations suppressed the DdegP DbamB-mediated conditional lethal phenotype, in part, by lowering the load of OMPs in the envelope, and additionally by increasing the synthesis of periplasmic chaperones.
The last group was represented by one suppressor, which was noted to have the unusual ability to suppress lethality without significantly altering steady-state OMP levels ( Fig. 1; lane 3). Through P1 transduction-mediated marker replacement, it was determined this third suppressor could fully correct the temperature sensitive growth defect of a DdegP bamB + strain. Because of these two properties, we chose to investigate this suppressor further in the DdegP bamB + strain and not in the original DdegP DbamB strain, which grows very poorly and prone to accumulating suppressors.
To quantify the extent to which the third suppressor mutation reversed the growth defect in DdegP cells, growth was measured by culturing cells in LB for five hours at 39uC, which is the sub-lethal growth temperature for the DdegP mutant, with cell density measured every thirty minutes. In contrast to DdegP-only cells, DdegP cells containing the suppressor mutation grew at a rate indistinguishable from the degP + strain (Fig. 2). degP + cells containing the suppressor mutation grew just like the degP + parental strain (Fig. 2). It is known that DdegP cells vesiculate profusely [35]. We asked whether the suppressor mutation can reverse this phenotype of DdegP. Membranes and vesicles obtained from cultures grown at 39uC were analyzed by SDS-PAGE and proteins were visualized after Coomassie blue staining (Fig. 3). As expected, DdegP cells released a large amount of vesicles containing a variety of proteins, including OmpC and OmpA. However, the presence of the suppressor mutation almost completely reversed this phenotype of DdegP. Hardly any proteins were visible from the vesicle fraction of wild-type and suppressor-containing cells (Fig. 3).

Identification of the suppressor mutation
Before determining the mechanism of DdegP suppression, we set out to identify the third suppressor mutation. For this, we introduced by P1 transduction a random Tn10 (Tet r ) insertion library into a DdegP strain harboring the suppressor mutation and selected for tetracycline resistant (Tet R ) transductants. Over a thousand Tet R transductants were screened for the reversion to temperature sensitivity (i.e., loss of suppressor) by replica plating at 40uC and 30uC. One colony became temperature sensitive, indicating that either the insertion of Tn10 in a gene confers the growth defect or the Tn10 brought in the flanking wild type DNA and replaced the suppressor mutation. To distinguish the two possibilities, we determined the linkage of Tn10 to the temperature sensitivity phenotype by transducing the Tet R marker into the DdegP strain containing the suppressor mutation. Approximately 50% of the time the resulting Tet R transductants became temperature sensitive, indicating that the Tn10 is closely linked to the suppressor mutation and not inserted in a gene that confers a temperature sensitive growth phenotype.
To conclusively determine the chromosomal location of the Tn10, arbitrarily primed polymerase chain reaction (AP-PCR) was utilized. PCR products obtained after the second, high-stringency reaction were analyzed on a 4% w/v agarose gel, excised and sequenced using the appropriate Tn10 specific primers used in the reaction. Using this method, the Tn10 was found to disrupt yfiF located at 58.59 on the chromosome, at nucleotide 883. A known Kan r insertion in glrK (57.99 on the E. coli chromosome) was utilized to determine which side of yfiF::Tn10 the suppressor mutation was located. DglrK::Kan r was introduced into DdegP suppressor-yfiF::Tn10 cells by P1 transduction, and Kan r transductants were screened for Tet s and temperature sensitive phenotypes. Using this method, the gene order was determined to be glrK-suppressor-yfiF.

Identification of the gene affected by the suppressor mutation
Several sets of primers were designed to amplify 1 to 3 kb long DNA from the 57.93 to 58.53 minute region of the chromosome from wild-type and suppressor-containing strains and the products were sequenced. When amplifying the rpoE-nadB region, we noted the presence of a PCR product from the suppressor strain that was approximately 1 kb larger than that amplified from the wild-type strain (data not shown). Sequence analysis indicated the presence of duplicated DNA in the suppressor strain.
In order to conclusively determine the molecular nature of the suppressor mutation, a primer set was designed which consisted of one primer landing upstream of rpoE and reading into rpoE (primer 1) and one primer landing upstream of nadB and reading into nadB (primer 2) (Fig. 4A). Because rpoE and nadB are divergently transcribed, the only way to obtain PCR product is if there was truly a duplication of rpoE or nadB, such that the two ''forward'' primers were able to prime the reaction by facing each other in the chromosome. Indeed, when PCR was performed in the wild-type and suppressor strains, only the suppressor strain gave rise to a fragment using the divergent primer set (Fig. 4C). Another set of primers that points towards each other and lands 157 bp upstream of nadB ATG (primer 3) and 124 bp downstream from the nadB ATG (primer 4) produced an expected 291-bp long PCR fragment from wild type strain (Fig. 4C). However, the suppressor containing strain yielded two fragments of 291 bp and about 1.3 kb (Fig. 4C), confirming the presence of duplicated DNA (Fig. 4B).
The purified 651-bp long PCR product was sequenced using the same primers used for amplification, revealing a novel duplicationtruncation of rpoE (Fig. 4B). The mutation consisted of rpoE truncated at nucleotide 396, corresponding to amino acid F122. After F122, the amino acid sequence continued with the nonnative MVWYA sequence before reaching a stop codon (UAG). In addition to coding for five non-native amino acids, the region immediately after F122 was identical to the DNA sequence from 21 bases upstream of the nadB translation start to the translation start site of the native, full-length copy of rpoE was reached (Fig. 4B). Thus, the native copy of rpoE gene, present downstream of the 39-tructaed copy of rpoE, is likely transcribed by the native promoter as well as the truncated rpoE gene promoter. We refer to Wild type and DdegP cells, with or without the suppressor mutation, were grown overnight at 30uC. Next day, overnight cultures were diluted to 1:100 in flasks containing fresh, pre-warmed, Luria broth and growth was resumed at 30uC and 39uC. OD 600 was measured from bacterial samples withdrawn every 30 minutes. Only 39uC growth curves, obtained from two independent experiments, are shown. All strains grew almost identically at 30uC. doi:10.1371/journal.pone.0033979.g002 the suppressor mutation as rpoE3, named after the suppressor isolate numbered 3.

RpoE (s E ) levels are elevated in the rpoE3 background
Because of the partial duplication of rpoE in rpoE3, we asked whether the steady-state s E levels were elevated in cells harboring rpoE3. Wild-type and DdegP cells containing rpoE3 were grown to late exponential phase, and whole-cell extracts were isolated to analyze s E levels by Western blot using s E -specific antibodies (Fig. 5). In strains containing the rpoE3 allele, s E levels were fourfold (in degP + background) to six-fold (in DdegP background) higher than in their rpoE + counterparts (Fig. 5A). This is likely due to elevated rpoE expression resulting from the rpoE3-mediated promoter duplication (in degP + background) and increased envelope stress (in DdegP background). Despite these increases in s E levels, the steady-state levels of OMPs were not reduced in the rpoE3 mutant (Fig. 1). In comparison, the s E level in a rseA null strain went up by 2.5 fold (Fig. 5A), which is lower than that seen in the rpoE3 strain, yet OMP levels were drastically reduced in the rseA mutant background (Fig. 1). It is noteworthy that despite the existence of a truncated rpoE open reading frame in addition to the full length open reading frame, no small molecular weight bands reacted with the s E -specific antibodies. The reason for this could be that either the antibodies used did not recognize the truncated s E protein, the truncated protein was not produced, or it was highly unstable.
s E must be released into the cytoplasm in order to become active. One possible explanation for the increase in s E levels without a concomitant reduction in OMPs in the rpoE3 mutant is that, since rseA itself is a member of the s E regulon, the rpoE3 mutation increases the level of membrane-bound s E rather than the soluble s E , thus limiting the s E response. To test this, cells were grown at 37uC to late exponential phase, lysed by French press, and soluble and insoluble fractions were separated by centrifugation at 100 000 g for an hour. s E from the two fractions were analyzed by Western blot using anti-s E antibodies ( Fig. 5B and C). Purity of the two fractions was affirmed by probing for TolC (insoluble) and MalE (soluble) proteins. Surprisingly, s E levels increased in both fractions ( Fig. 5B and C). Quantification revealed that soluble s E levels increased 1.38 fold (Fig. 5B) and insoluble s E levels decreased 0.78 fold (Fig. 5C) in the rpoE3 DdegP strain compared to the rpoE3 strain. Thus, despite the increase of s E in both fractions, there appears to be a small but preferential increase in the soluble active s E form in the rpoE3 DdegP strain. This is expected, since additional envelope stress caused by the loss of DegP would trigger a further release of s E from the envelope to combat stress.
rpo3 triggers a modest s E response Based on the results obtained above, we hypothesized that the suppressor mutation exerted its effect by slightly activating the s E  response without reducing the steady-state OMP levels (Fig. 1). To test this, rpoE3 was introduced into cells containing a RybB::LacZ fusion. RybB, whose expression is controlled by s E , encodes a small regulatory RNA responsible for down-regulating synthesis of several OMPs, including OmpC, OmpF and OmpA [36][37][38].
RybB::LacZ activity was determined from wild type, rpoE3, DdegP and rpoE3 DdegP cells grown at 39uC to mid-to lateexponential phase (Fig. 6A). In cells harboring the rpoE3 allele, RybB::LacZ activity increased by about 10% relative to wild type, indicating only a slight activation of the s E response at this growth temperature. The DdegP cells, which grew slightly slower than the parental strain (Fig. 2), showed a 30% increase in the RybB::LacZ compared to the parental strain. In the rpoE3 DdegP double mutant, which does not show a growth defect, an additive effect was observed: RybB::LacZ activity increased by 50% over wild type. This indicates that the two mutations elevated rpoE expression by independent mechanisms: rpoE3 directly increased s E levels, while the DdegP mutation indirectly activated s E by causing envelope stress. Only a modest increase in RybB::LacZ expression would help explain why there was no significant reduction in steady-state OMP levels in the rpoE3 DdegP background compared to the parental rpoE + degP + strain (Fig. 1). In contrast, the RybB::LacZ activity in an rseA null background went up almost six-fold (Gerken and Misra, unpublished data). This increase correlates well with a significant reduction in the OMP levels ( Fig. 1). We also measured RybB::LacZ activity from cells grown at 37uC since there is a possibility that the relatively small difference observed between wild type and mutant strains grown at 39uC could be due to the basal RybB::LacZ activity in wild type strain being already high at 39uC. The results from 37uC growth experiments were similar to those obtained from cells grown at 39uC, with a relative increase of 20% (rpoE3), 20% (DdegP), and 70% (rpoE3 DdegP) (Fig. S1). The additive effect of rpoE3 and DdegP was once again observed, indicating the two mutations activate RybB::LacZ expression by independent mechanisms. rpo3 restores envelope homeostasis by lowering DdegPmediated activation of the Cpx pathway Both Cpx and s E pathways are important in reducing envelope stress caused by aberrant b-barrel OMP assembly [9]. The Cpxmediated upregulation of degP is critical for this reduction in envelope stress [9]. Consistent with this view, expression of cpxP, a prototypical member of the Cpx regulon, is modestly upregulated in DdegP cells grown at 39uC (Fig. 6B; [9]). We asked whether the elevated s E response by rpoE3 will help reduce the activated Cpx response observed in DdegP cells. As shown in Fig. 6B, the DdegPmediated elevated CpxP::LacZ activity returned to the wild type level in the rpoE3 DdegP double mutant background. These results showed the complementary nature of the two envelope stress response pathways controlled by the s E and Cpx systems.

Potential activation of a DegP-independent proteolysis pathway in the rpoE3 mutant
One of the major functions of the s E response is to promote the proper folding and/or destruction of aberrantly folded OMPs in the envelope. In a DdegP background, rpoE3 obviously cannot simply increase degP transcription to compensate for increased stress and abolish the temperature sensitive growth phenotype. We therefore asked whether rpoE3 caused the activation of another proteolytic pathway, thus allowing for normal growth in a DdegP background.
For this, we employed AcrA L222Q , an unstable variant of AcrA that is rapidly degraded in stationary phase grown cultures at 37uC, but not at 30uC, in a DegP-dependent manner [39]. Because of its instability properties, AcrA L222Q makes an ideal substrate to assay envelope proteolysis activity in vivo. The chromosomal allele of acrA expressing AcrA L222Q was introduced into cells harboring rpoE3 and lacking tolC (DtolC destabilizes the mutant AcrA protein). In order to test for proteolysis in this background, degP was also removed by a degP::Kan r insertion.
Cells were grown overnight in LB broth at 37uC and 30uC, after which samples were equalized according to final cell densities and boiled in SDS sample buffer. Samples were analyzed by Western blot using antibodies recognizing AcrA. At 30uC, the presence of degP::Kan r or rpoE3 did not significantly affect AcrA L222Q levels (Fig. 7, lanes 1-4). However, at 37uC AcrA L222Q levels increased six folds in a degP::Kan r background compared to the degP + strain (Fig. 7, lanes 5 and 7). In an rpoE3 degP + background, the protein was also degraded due to the presence of intact DegP (Fig. 7, lane  6). Interestingly, the presence of rpoE3 in the degP::Kan r background caused a three-fold reduction in the AcrA L222Q levels compared to that present in the degP::Kan r background alone (Fig. 7, lane 8), reflecting decreased stability of the protein, presumably due to an increase in proteolysis in the envelope. Consistent with the involvement of a s E -mediated, DegPindependent proteolytic pathway, AcrA L222Q levels were also reduced in a DrseA DdegP background where s E was fully activated (Fig. S2A). acrA is not known to be a member of the s E regulon. Indeed, we have previously shown that the wild-type AcrA protein level did not change in a bamA mutant background in which the s E response was strongly induced [40]. Nevertheless, to eliminate the possibility that the expression of the acrA gene is somehow strongly affected by the induction of s E in the rpoE3 background, we created an isogenic set of strains to those mentioned above, except that this set contained the wild type acrA gene. Western analysis data showed that at 37uC in the rpoE3 degP::Kan r background, the level of wild type AcrA decreased only about 10% compared to the rpoE + DdegP cells (Fig. S2B). In contrast, AcrA L222Q levels were reduced over 300% in the rpoE3 degP::Kan r background compared to rpoE + degP::Kan r cells (Fig. 7). These data support the notion that a post-synthesis event, most likely an activated proteolytic pathway of the s E regulon, is primarily responsible for the observed destabilization/degradation of AcrA L222Q in the absence of DegP. This pathway, in part, contributes to the mechanism of suppression by rpoE3.

Discussion
One of the hallmarks of the fully activated s E system is the severely reduced major OMP levels resulting from high expression of the s E -controlled small regulatory RNAs, RybB and MicA [36][37][38]. Indeed, rseA mutations that constitutively activate the s E stress response pathway also dramatically lower the major OMP levels (Fig. 1). Even though the rpoE3 mutation significantly elevates s E in the cell-over two and fourfold compared to DrseA and rpoE + strains, respectively-no significant changes in the levels of major OMPs, OmpA and OmpC, were observed, indicating that the s E stress response pathway was not fully activated. This was also reflected by a mere 10% and 20% increase in s E -controlled RybB::LacZ activity at 39uC and 37uC, respectively. We suspect that in the rpoE3 background the continued presence of RseA at presumably slightly higher levels prevents s E from becoming fully active. In the rpoE3 DdegP background, the s E pathway is more active than in the rpoE3 or rpoE + background, based on the elevated RybB::LacZ activity and a complete reversal of the vesiculation and temperature sensitive growth phenotypes of DdegP. Yet, the absence of a significant effect on the major OMP levels suggests the extent of s E activation still remains lower in the rpoE3 background than in a DrseA background.
In contrast to DdegP cells grown at 39uC, a greater s E and Cpx activation is observed in cells with impaired OMP assembly pathways stemming from the loss of a major periplasmic chaperone, SurA, or the presence of a defective BAM complex [9,40]. It is conceivable that differences in the biochemical nature of OMP assembly intermediates and their levels account for some of the disparities in envelope stress response in different genetic backgrounds. For example, it is well established that during aberrant OMP assembly, exposure of the C-terminal OMP residues in the periplasm activates the s E pathway by promoting RseA proteolysis [41]. This likely occurs in a background lacking SurA or expressing a defective BAM complex. In contrast, there is no evidence that OMP assembly is significantly defective in DdegP cells, since we do not see any decrease in OmpA and OmpC levels from the envelopes. Nevertheless, it is likely that there is some increase in the nascent, unassembled OMPs that have fallen off the proper assembly pathway and are normally captured and/or degraded by DegP. Also, the levels of other unstable/unfolded envelope proteins, including those that normally reside in the periplasm, must increase in DdegP cells grown at 39uC. Thus, envelope stress is likely to be created in DdegP cells-enough to induce to a robust stress response through activating the s E or Cpx pathways. The predominant response to this stress appears to be the release of outer membrane vesicles, but this is clearly not sufficient when growth temperature reaches 40uC where viability of DdegP cells becomes severely compromised. The fact that rpoE3 completely reverses the phenotypes of DdegP suggests that it must lower the levels of stress-causing envelope proteins. However, unlike the rseA null mutations, which significantly lower OMP levels, the effect of rpoE3 on OMPs alone in not sufficient to account for its protective phenotype.
Since OMP levels are not reduced by rpoE3, alternative stress responses must account for the reversal of the DdegP phenotypes.
Our results indicate that up-regulation of a DegP-independent, s E -dependent proteolytic pathway can partly accounts for this reversal. To test this pathway, we used a mutant AcrA protein, AcrA L222Q , which is rapidly degraded in a DegP-dependent manner [39]. In the absence of DegP, the mutant AcrA protein level rose significantly, but in the presence of rpoE3, AcrA L222Q levels went down again. The level of AcrA L222Q was also reduced in the absence of RseA, indicating the involvement of s E . With no reduction in wild type AcrA levels in the rpoE3 background, it appears that a post-synthesis step, most likely involving increased proteolysis, accounts for the reduced AcrA L222Q level. We do not currently know the identity of the protease(s) responsible for AcrA L222Q degradation in the absence of DegP in the ropE3 and DrseA strains.
Overexpression of SohB and DegQ from plasmids has been reported to rescue the temperature sensitive growth phenotype of DdegP [42][43]. However, neither sohB nor degQ is known to be regulated by s E and there is no experimental data showing SohB is actually a protease. yfgC, a gene regulated by s E , is predicted to encode a periplasmic metalloprotease [44]. However, deletion or overexpression of YfgC did not influence AcrA L222Q stability or the DdegP phenotypes (data not shown). It should be noted that activation of a proteolytic pathway likely represents only a component of the broader defensive response triggered by s E . Therefore, we believe that the collective action of increased proteolysis and other defensive measures most likely account for the rpoE3-mediated reversal of the DdegP phenotypes.
The fact that mutations required to increase s E levels and rescue the DdegP phenotypes are either in rseA, a gene encoding for the negative regulator of s E , or the rpoE gene itself suggests that even under high envelope stress conditions the amount of free s E becomes limiting. If growth at high temperatures without DegP indeed leads to the accumulation of misfolded OMPs, then these accumulated OMPs should trigger the DegS/RseP-mediated proteolytic pathway to degrade RseA and elevate free s E levels [45]. The fact that DdegP cells begin to die at 39uC suggests that the defensive response system may have maxed out, perhaps because newly released and activated s E allows for a greater synthesis of RseA, which in turn re-captures s E . In other words, under extreme envelope stress, a tight homeostasis loop between RseA and s E may partly be responsible for cell death. This loop can be broken by mutations that either lower RseA activity or increase s E levels. An alternative means to overcoming the growth phenotype of DdegP without directly influencing the RseA-s E loop would be through mutations that influence the cAMP-CRP [25] or EnvZ/OmpR [30] pathway. The prlF/sohA mutations likely provide yet another mechanism to overcome the temperature sensitive growth phenotype of DdegP.

Bacterial strains, media and reagents
Escherichia coli K-12 bacterial strains used in this study were constructed in an MC4100 background [46] and are shown in the Table S1. Luria-Bertani broth (LB) and Luria-Bertani agar (LBA) were prepared as previously described [47]. When required, media were supplemented with 50 mg ml 21 ampicillin, 25 mg ml 21 kanamycin, 12.5 mg ml 21 chloramphenicol, 10 mg ml 21 tetracycline and 0.2% L-(+) arabinose (Sigma). All other chemicals were of analytical grade. Bacterial growth curves were performed by diluting overnight cultures 1:100 into 25 ml fresh pre-warmed media supplemented with appropriate antibiotics. Cultures were vigorously shaken at 39uC in baffle-bottom flasks in a water incubator with cell density checked every 30 min.

DNA manipulation
Standard bacterial genetic methods were carried out as previously described [47]. Chromosomal deletion of degP, followed by the removal of the antibiotic resistance cassette was carried out using the l-red mediated gene deletion method as previously described [48]. Primer sequences are available upon request.
The unknown Tn10 marker was identified by the arbitraryprimed PCR (AP-PCR) method [49][50]. The first, lowstringency round of PCR was carried out with wild-type cells and cells containing the unknown Tn10 marker (see below). That round of PCR used both ARB1 and ARB6 primers [50], as well as either Tn10 left or Tn10 right primers to amplify fragments on either side of the Tn10. First round products were analyzed on an 0.8% agarose gel, and fragments unique to the Tn10 strain were excised using a QIAquick gel extraction kit (QIAGEN) and subjected to a second, high-stringency round of PCR using the appropriate Tn10 primer and ARB2 [50]. Amplified products were cleaned up using a QIAquick PCR purification kit and sequenced using the appropriate Tn10 primer or ARB2. Sequences were analyzed to obtain non-Tn10 sequence, and unassigned sequences were analyzed using BLASTn to determine the location of the Tn10.

Cell fractionation
Fresh cultures were grown by sub-culturing overnight cultures into fresh media with appropriate antibiotics at a starting OD 600 of 0.025. Cell pellets, collected from freshly grown cells, were resuspended in a lysis buffer containing 20 mM Tris-HCl pH 7.5, 2 mM phenylmethylsulfonylfluoride (PMSF), 10 mM MgCl 2 and 25 mg ml 21 DNase I. Samples were passed twice through a French pressure cell, followed by low-speed centrifugation of lysates to remove unlysed cells. Crude lysates were the centrifuged at 100,000 g for 1 h at 4uC to pellet membranes. Periplasm was extracted using the gentle osmotic shock method [51]. Prior to French Press lysis, cell pellets were resuspended in periplasm extraction buffer (10 mM Tris-HCl pH 7.5, 500 mM sucrose, 10 mM EDTA and 0.2 mg ml 21 lysozyme) and incubated on ice for 30 min. Samples were centrifuged at 100,000 g for 15 min and periplasm drawn off as soluble fraction. Spheroplast pellets were then washed with cold 10 mM Tris-HCl pH 7.5 and subjected to French press lysis as described above. After high speed spin, cytoplasm was harvested from the soluble fraction. Membranes pellets were routinely washed in 10 mM Tris-HCl pH 7.5 and resuspended according to cell density in the same buffer.
Enzymatic assays b-galactosidase activity was determined by an established method [52], modified for use with microtiter plates. Kinetic analysis of b-galactosidase activity was carried out using a Versamax microtiter plate reader (Molecular Dynamics). Activity was calculated as the rate of ONPG (ortho-Nitophenyl-b-Dgalactopyranoside) cleavage (OD 420 ) normalized to cell density (OD 562 ) in each well. All assays were performed in quadruplicate.