The EnvZ-OmpR acid stress response acidifies the cytoplasm
We previously determined that S. Typhimurium reduced its intracellular pH in response to external acid stress in an OmpR-dependent manner2, in contrast to previous studies in E. coli that reported a rapid recovery within minutes17,18. We set out to reconcile these differences by measuring the cytoplasmic pH and recovery of S. Typhimurium and E. coli immediately following an external acid shift. In S. Typhimurium, the cytoplasmic pH decreased from 6.80 to 6.35 within 5 min at pHe 5.6 and reached a plateau value of 6.15 after 90 min (Fig. 1a, b). In contrast, the cytoplasm remained unchanged (pH 6.80) when incubated at pHe 7.2. No decrease in cytoplasmic pH was observed in an ompR null strain of S. Typhimurium (pH 6.75), confirming the existence of an OmpR-dependent acidification process2. In E. coli, the initial pH of the cytoplasm was slightly higher (pH 7.13) compared to S. Typhimurium (pH 6.80), but a comparable decrease in pHi from 7.13 to 6.55 occurred during 90 min exposure to acid stress (Fig. 1c, d). In an ompR null strain of E. coli, intracellular pH did not decrease in response to extracellular acid stress, establishing OmpR as a regulator of acid stress in both S. Typhimurium and E. coli (Fig. 1 and refs. 2,19).
Acidification via OmpR repression of lysine decarboxylation
In S. Typhimurium, OmpR prevents recovery from acidification by repressing cadC/BA. CadA encodes a lysine decarboxylase, CadB encodes a lysine/cadaverine antiporter, and CadC encodes a transcriptional regulator. We performed quantitative real-time PCR (qRT-PCR) with E. coli and observed an increase in transcript levels of cadB (4.5-fold) and cadA (3-fold) in the ompR null mutant, suggesting that E. coli OmpR also functions as a repressor at this locus (Supplementary Fig. 1a). OmpR repression resulted from direct interaction at the cadC/BA locus, based on electrophoretic mobility shift assays (EMSAs) (Supplementary Fig. 1b). OmpR did not bind at cadA (Supplementary Fig. 1b), suggesting that cadB and cadA are co-transcribed, as we also observed in S. Typhimurium2.
To determine whether cadC/BA repression was sufficient to prevent recovery from acid stress, we measured the response of an ompR/cadBA null strain, a cadBA null strain, and a cadBA over-expressed strain of E. coli exposed to similar acid stress. As with S. Typhimurium (Fig. 1a, b and ref. 2), OmpR was no longer required for cytoplasmic acidification when cadBA was eliminated, indicating that the CAD system was the main pathway for recovery from acidification between pH 6.1 and 6.5 (Fig. 1c, d). Furthermore, when cadBA was over-expressed in E. coli, OmpR was unable to repress cadC/BA, and the cytoplasm was neutralized.
Overexpression of additional decarboxylases, such as glutamate (gadBA), arginine (speA), and S-adenosylmethionine decarboxylase (speD), during acid stress did not neutralize the pHi (Supplementary Fig. 1c, d), suggesting that cadBA was the major system that eliminated protons upon acid stress in E. coli over the approximate pHi range of 6.1–6.5. This was not surprising, since the glutamate and arginine decarboxylation systems exhibit much lower pH optima (pH 4 and 5, respectively)20, compared to lysine (pH 6.1–6.5)21, S-adenosylmethionine decarboxylase (pH 7.4)22 and ornithine decarboxylase systems (pH 7.0)23. Furthermore, the pHi response of the cadBA null strain was similar to the wild type (Fig. 1). Thus, in both S. Typhimurium and E. coli, OmpR directly repressed cadC/BA to enable cytoplasmic acidification under acid stress.
Osmotic stress acidifies S. Typhimurium and E. coli
EnvZ has been described as an osmosensor, as high osmolality activates OmpR to differentially upregulate ompC and repress ompF (reviewed in ref. 8). Since EnvZ senses and responds to a concentrated cytoplasm2,9, we hypothesized that the cytoplasm might acidify during osmotic stress, indicating that EnvZ is actually functioning as a pH sensor. Using sucrose as the osmolyte, addition of 15% sucrose (w/v; 787 mOsm Kg−1) to the medium at pHe 7.2 led to a decrease in the pHi of S. Typhimurium from 6.80 to 6.45 (Fig. 1a, b). In contrast, the ompR null strain remained near-neutral pHi, decreasing by <0.1 pH unit to 6.75 (Fig. 1a, b). A similar decrease was evident in E. coli, where osmotic stress reduced the pHi from 7.13 to 6.75. The pHi of the ompR null E. coli strain did not decrease in response to sucrose-induced osmotic stress and remained at pHi 7.1 (Fig. 1c, d). Thus, in both S. Typhimurium and E. coli, increasing external osmolality stimulated a prolonged, OmpR-dependent intracellular acidification (Fig. 1).
Osmolyte-induced acidification is distinct from acid stress
Overexpression of CAD did not protect S. Typhimurium or E. coli from acidification by osmotic stress (Fig. 1). Furthermore, elimination of cadB/A in an ompR null strain did not restore intracellular acidification at high osmolality in both S. Typhimurium (Fig. 1a, b and Supplementary Fig. 2a) and E. coli (Fig. 1c, d and Supplementary Fig. 2b), suggesting osmotic stress and acid stress pathways were distinct. However, osmolyte-driven intracellular acidification was sufficient to trigger virulence gene expression of Salmonella pathogenicity island 2 (SPI-2), as evidenced by ssrB transcription and effector SseJ secretion (Supplementary Fig. 3a, b). To identify OmpR targets in response to acid stress, we performed a microarray-based expression analysis and compared the transcriptome of ompR null strains with wild-type strains of S. Typhimurium and E. coli during acid or osmotic stress (see data availability for full results through the GEO).
In our S. Typhimurium microarray, the alternate sigma factor rpoS was upregulated 4.4-fold in the absence of ompR. RpoS was implicated in the acid-inducible exponential-phase acid tolerance response (ATR) of S. Typhimurium24,25,26 and was essential for survival under acid stress27,28. In agreement with the upregulation observed by microarray, the rpoS transcript was up 4.8-fold in the ompR mutant (Fig. 2a), suggesting that OmpR functions as a repressor at rpoS. Furthermore, the rpoS over-expressed strain maintained near-neutral pHi (6.8) in response to osmolytes, but was fully capable of cytoplasmic acidification by acid stress (Fig. 2b, c and Supplementary Fig. 4). Thus, the responses to acid stress and osmotic stress were distinct. Similarly, in the absence of rpoS, OmpR was not required for intracellular acidification at high osmolality (Fig. 2b, c and Supplementary Fig. 4). In response to acid stress, OmpR repressed the cadC/BA operon to maintain an acidic cytoplasm, whereas at high osmolality, an rpoS-dependent pathway was involved in S. Typhimurium acidification (Figs. 1 and 2). OmpR acted directly at rpoS, as determined by EMSAs with purified OmpR protein (Fig. 2d), i.e., OmpR directly repressed rpoS transcription at high osmolality.
To identify rpoS targets involved in intracellular acidification, we examined the microarray results for candidate targets that were downregulated in the ompR null strain and were known to be rpoS-dependent. We identified yghA, encoding a putative oxidoreductase, which was down 39-fold in the ompR null strain at high osmolality compared to the wild-type S. Typhimurium. This was confirmed by qRT-PCR, which indicated comparable downregulation of a yghA transcript (32-fold) in the ompR null strain, as well as in the rpoS over-expressed strain (26-fold), implicating RpoS as a repressor of yghA (Fig. 3a). In contrast, in E. coli, YghA was not involved in osmotic stress-driven acidification, as yghA levels of the wild type and an ompR null strain were comparable (Fig. 3b). As an oxidoreductase, YghA is predicted to oxidize NADH to liberate H+, leading to cytoplasmic acidification. In the absence of YghA, the S. Typhimurium cytoplasm would fail to acidify, due to a reduction in acid production. The pHi of a yghA null strain was 6.8 (i.e., not acidified) and thus, OmpR was no longer required to repress rpoS (Fig. 3c, d). This result suggests a mechanism whereby OmpR relieves rpoS-dependent repression of yghA (which would be activating), to produce and maintain an acidified cytoplasm in response to osmotic stress. Furthermore, when yghA was over-expressed, the pH of the cytoplasm decreased to 6.45. In contrast, the ompR/rpoS/yghA null strain maintained near-neutral pHi (Fig. 3c, d). OmpR did not bind directly to the yghA promoter, as evident from the EMSA (Fig. 3e). This result was further supported by AFM imaging, and the relative height distribution histogram (Fig. 3f). Instead, OmpR directly repressed rpoS transcription (Fig. 2), which increased yghA transcription (Fig. 3a, c, d).
Measurements of NAD+/NADH were consistent with these results (Supplementary Fig. 5a, b). In the wild-type strain, yghA transcription increased 32-fold compared to an ompR null strain. Intracellular levels of NAD+ in the wild type (52.3 nM) or an ompR/rpoS (51.3 nM) null strain were higher than an ompR null strain (29.6 nM), a yghA null strain (37.3 nM) or an ompR/yghA/rpoS (35.6 nM) null strain grown at high osmolality (Supplementary Fig. 5a, b). Overexpression of yghA in an ompR null background completely restored intracellular NAD+ to wild-type levels (49.6 nM). Increased NAD+ levels were only observed at high osmolality and not during acid stress, further emphasizing the distinct responses to acid and osmotic stress.
OmpR-dependent acidification occurs at high osmolality
Microarray analysis comparing the wild-type and ompR null E. coli strains indicated that decarboxylase genes were differentially expressed under acid and osmotic stress conditions (Fig. 4a and Supplementary Fig. 1d). To identify the precise pathways, we compared mRNA levels of genes encoding glutamate decarboxylase (gadB and gadA), arginine decarboxylase (speA), S-adenosylmethionine decarboxylase (speD), and ornithine decarboxylase (speF) by qRT-PCR in the wild-type and ompR null mutant (Fig. 4a and Supplementary Fig. 1a). The transcript for speF (fivefold) was the most upregulated, compared to speA (1.5-fold) and speD (1.8-fold) in the ompR null strain at high osmolality, suggesting a repressive role of OmpR (Fig. 4a). In S. Typhimurium, speF levels were identical between the wild-type and ompR null strain, indicating the differing response by OmpR in S. Typhimurium (RpoS repression of yghA) and E. coli (speF) in response to osmotic stress (Figs. 3 and 4b).
Following osmotic stress, the cadBA (Fig. 1c, d), gadBA, speA, and speD overexpression strains all acidified to an extent indistinguishable from wild type, whereas the strain over-expressing speF did not acidify (Fig. 4c, d and Supplementary Figs. 1c, d and 6). This result identified speF as the predominant decarboxylase that functioned to eliminate protons during osmotic stress to maintain pH homeostasis in E. coli. Overexpression of speF in response to acid stress did not affect the cellular response to acid stress (Fig. 4c, d and Supplementary Fig. 6), emphasizing again that the acid and osmotic pathways were distinct. As expected, OmpR was not required for acidification in the absence of speF, and the speF null strain remained similarly acidified compared to the wild type. OmpR bound to the speF regulatory region, as evident from an EMSA (Fig. 4e). Thus, in E. coli, OmpR represses the speF system to prevent recovery from acidification during osmotic stress.
Phosphorylation of EnvZ-OmpR is pH sensitive
When the cytoplasm was acidified from either acid or osmotic stress, OmpR was “activated”, leading to repressive effects on transcription at cadC/BA, speF, or rpoS (Figs. 1–4). We suspected that EnvZ-OmpR phosphorylation might be pH sensitive, as the histidine phosphorylation site must be unprotonated to act as a nucleophile toward ATP29. S. Typhimurium or E. coli EnvZ was phosphorylated with [γ-32P]-ATP and the percentage of EnvZ~P was determined by densitometry (Fig. 5a). EnvZ phosphorylation in S. Typhimurium was less robust (70%) than E. coli EnvZ (100%). At pH 5.5, EnvZ phosphorylation was substantially lower when compared to phosphorylation at pH 7.5 for E. coli (28%) and S. Typhimurium (12%), raising the question as to how does activation of OmpR occur during acid and osmotic stress?
Acidification does not require OmpR phosphorylation
Canonically, response regulators are phosphorylated by their cognate kinases and phosphorylation drives dimerization. The phosphorylated dimer is an active complex that binds DNA with high affinity and interacts with RNA polymerase to activate transcription. However, some orphan response regulators that lack cognate kinases activate transcription without phosphorylation by mimicking an active interface via dimerization30; (see ref. 11 for a recent review). We reasoned that there might be an acid-dependent conformational change in OmpR that was independent of phosphorylation. We also knew from our previous measurements that EnvZc had higher affinity for OmpR compared to OmpR~P31,32, which would promote an interaction. We measured the cytoplasmic pH of S. Typhimurium in response to acid and osmotic stress in a D55A OmpR mutant that was incapable of phosphorylation33 in the presence of the EnvZ kinase. Cytoplasmic acidification occurred in the D55A mutant background in response to either acid or osmotic stress, indicating that OmpR phosphorylation was not required for acidification (Supplementary Fig. 7a, b). Intriguingly, this phosphorylation-independent activation was completely dependent on the presence of the EnvZ kinase2. We therefore examined whether an EnvZ mutant that was incapable of interacting with OmpR (EnvZc C277Y)10 was still capable of acid-dependent OmpR repression (Fig. 5b, c). In the presence of the EnvZc C277Y mutant, the S. Typhimurium cytoplasm was no longer acidified in response to acid stress (Fig. 5b, c). Thus, OmpR must interact with EnvZc to repress CAD, even though phosphoryl transfer was not involved.
Gel filtration profiles of EnvZc, EnvZc C277Y, OmpR, (OmpR + EnvZc), and (OmpR + EnvZc C277Y) established that in isolation, both wild-type EnvZc and the EnvZc C277Y mutant existed as dimers, whereas OmpR was monomeric (Supplementary Fig. 8a, b). In the presence of wild-type EnvZc, OmpR addition generated a higher molecular weight complex, representing two OmpR molecules bound to an EnvZc dimer (Supplementary Fig. 8a). OmpR dimerization upon interaction with an EnvZc dimer is in agreement with our previous findings using fluorescence cross-correlation spectroscopy10. Addition of OmpR to the EnvZc C277Y mutant generated profiles that were identical to the individual proteins in isolation, i.e., no interacting complex was evident (Supplementary Fig. 8b) and ref. 10. OmpR is unusual among response regulators, because it can bind DNA in the absence of phosphorylation34, presumably because some percentage of the OmpR population exists in an activated, dimeric state. We thus examined whether the oligomeric state of OmpR was affected by acidic conditions (Supplementary Fig. 8c). The percentage of OmpR dimers increased substantially (to 63–73%), compared to neutral pH, where dimers were undetectable (Supplementary Fig. 8c). Thus, acid pH can also drive an OmpR conformational change that results in unphosphorylated OmpR dimers. In vivo, this process is driven by interaction with its kinase EnvZ.
Acid pH stimulates OmpR binding to DNA
How does acidification affect OmpR/DNA interactions? Most DNA-binding assays are sensitive to acidic pH34,35. For this reason, we turned to AFM to visualize OmpR interactions with the ompC promoter of S. Typhimurium (Fig. 5d) and E. coli (Fig. 5e) or the cadBA promoter (Supplementary Fig. 9). OmpR was added to a solution buffered to an identical pHi that we measured during acid stress (Fig. 1), indicated in the panels. Acid pH alone did not stimulate OmpR aggregation, as evident in the panels that contain OmpR protein in the absence of DNA (Supplementary Fig. 9a). Addition of OmpR at pH 6.1 (S. Typhimurium) led to an increase in binding to DNA, compared to addition of OmpR at pH 6.8 (Fig. 5d and Supplementary Fig. 9b). At pH 6.8, a localized binding of OmpR to ompC DNA was visible as specific, discrete foci (see Supplementary Fig. 9d for quantitation of binding and other controls). Addition of OmpR~P prepared from acetyl phosphate phosphorylation was in-between the level of binding observed at pH 6.8 and 6.1 (Fig. 5d). A similar stimulation of OmpR binding to DNA in acidic conditions was also evident in E. coli (Fig. 5e and Supplementary Fig. 9c, f), although the level of acidification was less (6.5 vs. 6.1). In the presence of EnvZ, acidic conditions further increased OmpR binding to the ompC promoter compared to OmpR alone (Fig. 5d, e; right panels, and Supplementary Fig. 9e, g). The simplest interpretation of this result is that in vivo, acid pH promotes an activating conformation of EnvZ9,36 that promotes interaction with OmpR, inducing an OmpR conformational change that stimulates dimer formation and favors OmpR binding to DNA. This process is stimulated by OmpR contact with EnvZ, in an interaction that does not involve phosphorylation. Thus, an EnvZ- and acid-dependent conformational change is sufficient to stimulate OmpR binding to DNA (see “Discussion” section).
How to reconcile our results with previous studies
Our results with the I-switch and BCECF-AM in E. coli and S. Typhimurium (this work and ref. 2) differ from previously published studies on the response of E. coli18,37 and S. Typhimurium38,39 to acid stress using pHluorin, a pH-sensitive fluorophore. In addition to using a different fluorophore, some measurements were performed in media with poor buffering capacity using HCl to initiate acid stress. We lowered the pHe of M63 media supplemented with casein hydrolysate from 7.5 to 5.5 by adding 8.5 mM HCl and measured the pHi of E. coli using a similar acid-induction strategy, but measured the response with BCECF-AM and fluorescence microscopy as before. In both S. Typhimurium and E. coli, the cytoplasm was acidified, indicating that the differing results were not a result of different stress conditions (Supplementary Fig. 10a, b).
Plasmid expression of pHluorin is heterogeneous
We next examined whether differences were due to the use of different fluorophores. Most previous studies measured the average fluorescence of cultures in solution, whereas our measurements were performed in single cells. We transformed S. Typhimurium and E. coli with ratiometric pHluorin (pH-sensitive GFP) and examined single cells by confocal microscopy after acid or osmotic shock (Fig. 6a). It was immediately apparent that the cells were extremely heterogeneous with respect to pH values in both S. Typhimurium and E. coli. In fact, in many of the cells, there was no apparent response to acid stress, while in others, there was acidification, but to varying levels. If one determined the average pH of this population in solution, it might very well appear to have “recovered” from acid stress. This level of heterogeneity was not observed in single cells using BCECF-AM (this work) or using the I-switch2. Thus, expressing pHluorin using the arabinose-inducible pBAD promoter is not a good indicator for measuring intracellular pH, because it is not uniformly distributed40. Constitutive expression of pHluorin was homogeneously distributed, but the fluorescence signal was weak, as previously reported41.
MC4100 does not acidify during acid or osmotic stress
Another explanation for the differences might be that different strains were employed. We used the sequenced E. coli strain MG165515, whereas many previous studies used MC410017. We measured the response of MC4100 to acid stress, using our approach (Fig. 6b, c). MC4100 was slightly acidified initially compared to MG1655 (pH 6.80 vs. 7.13, respectively), which was comparable to the pHi of S. Typhimurium (Figs. 1 and 6). In response to pHe 5.6, E. coli MC4100 maintained its cytoplasmic pH (pHi = 6.83) throughout the experiment (Fig. 6b, c). The pHi remained essentially unchanged whether the pHe was 7.2, 5.6, or induced by 15% (w/v) sucrose (Fig. 6b, c and Supplementary Fig. 10a, b). In contrast, addition of osmolytes actually increased the pHi (from 6.8 to 7.15), as reported18. This result was opposite to what we observed with E. coli MG1655. Cytoplasmic acidification in the non-pathogenic E. coli Nissle 1917, a widely used probiotic strain14, was comparable to E. coli MG1655 and S. Typhimurium (Supplementary Fig. 10d).
Sodium benzoate is not a good clamping agent
Most significantly, previous studies generated a standard curve using uncouplers that were reported to collapse ΔpH6,17,18,38,39. Sodium benzoate was commonly employed for this purpose. Unfortunately, sodium benzoate was not effective at setting pHe = pHi (Fig. 6d); the standard curve differed dramatically from curves generated from nigericin clamping. To examine this issue further, we used BCECF-AM to measure the pHi of E. coli MG1655 in response to various external pH conditions in the presence of 30 mM sodium benzoate. It was evident that at pHe = 5.6, the pHi was 6.3; at pHe = 6.0, the pHi was 6.2 and at pHe = 7.2, the pHi was 5.8 (Fig. 6e). Thus, using sodium benzoate at neutral pHe, the intracellular pH would be presumed to have been neutralized, when it was actually quite acidic. These values were not strain dependent, i.e., they were similar for MC4100 and MG1655 (Fig. 6e, f). Thus, it was evident that sodium benzoate was not an effective clamping agent. This finding also explains why previous measurements of S. Typhimurium pH were substantially different from ours38,39 and why E. coli was reported to return to a neutral pHi6,17,18 after an acid stress. More recent measurements in S. Typhimurium using our methods (although not performed in single cells)42 now are in keeping with our previous findings2.