Research Article

Dual Molecular Signals Mediate the Bacterial Response to Outer-Membrane Stress

See allHide authors and affiliations

Science  17 May 2013:
Vol. 340, Issue 6134, pp. 837-841
DOI: 10.1126/science.1235358

Stress Inside Out

In Gram-negative bacteria, the integrity of the outer membrane is crucial for survival and is an important aspect of resistance to antibiotics. The biogenesis of the major components lipopolysaccharide (LPS) and outer-membrane protein (OMP) of the outer membrane begins in the cytoplasmic compartment, involves export across the inner membrane and transport through the periplasm, and finally requires active insertion into the outer membrane by specialized assembly machines. Lima et al. (p. 837) supply results for a model in which serious defects in LPS biogenesis also create problems for OMP biogenesis, thereby producing the two signals needed to activate the σE stress response pathway.


In Gram-negative bacteria, outer-membrane integrity is essential for survival and is monitored by the σE stress-response system, which initiates damage-repair pathways. One activating signal is unassembled outer-membrane proteins. Using biochemical and genetic experiments in Escherichia coli, we found that off-pathway intermediates in lipopolysaccharide transport and assembly provided an additional required signal. These distinct signals, arising from disruptions in the transport and assembly of the major outer-membrane components, jointly determined the rate of proteolytic destruction of a negative regulator of the σE transcription factor, thereby modulating the expression of stress-response genes. This dual-signal system permits a rapid response to dysfunction in outer-membrane biogenesis, while buffering responses to transient fluctuations in individual components, and may represent a broad strategy for bacteria to monitor their interface with the environment.

The outer membrane (OM) is essential for the survival of Gram-negative bacteria. In Escherichia coli, the σE stress-response system recognizes signals indicative of OM dysfunction and triggers an adaptive response by activating the expression of gene products involved in the biogenesis, transport, and/or assembly of the lipopolysaccharides (LPSs), phospholipids, and outer-membrane proteins (OMPs) that make up the OM, and the proteases and chaperones that maintain or repair OM integrity (1, 2). In this system, the RseA and RseB regulatory proteins and the DegS and RseP inner-membrane (IM) proteases transmit the signal that activates the σE transcription factor (fig. S1). RseA, a single-pass IM protein, has a cytoplasmic domain that binds and inhibits σE and a periplasmic domain (RseAP) that binds RseB (35). After stress, OMPs accumulate in the periplasm, and their C-terminal residues bind to DegS and activate cleavage of RseAP (6), triggering a proteolytic cascade that frees σE to activate gene expression (1). However, a signal that inhibits RseB is also required, because RseB binding to RseAP prevents activated DegS from cleaving RseA (4, 5, 7). The RseB-inhibition signal has not been characterized but may be related to LPS, because alterations in LPS structure activate the σE response (8, 9). We tested and provide evidence here for a model in which intermediates in LPS transport and assembly are the second signal that activates σE.

LPS Antagonizes RseA-RseB Binding in Vitro

We developed an assay for 35S-RseB dissociation from RseAP-agarose and established that 35S-RseB was efficiently eluted by both unlabeled RseB and RseAP but not by unrelated proteins or an abundant bacterial phospholipid (Fig. 1A). LPS purified from an E. coli K-12 strain also had RseB-elution activity (Fig. 1B), as did LPS variants purified from various mutant strains (figs. S2 and S3). LptA, a protein that is part of the periplasmic bridge that shuttles LPS from the IM to the OM (10), inhibited LPS elution of RseB (Fig. 1C), indicating that RseB and LptA compete for LPS. RseB competed more efficiently at 45°C, a temperature that activates the cellular σE response, than at lower temperatures (Fig. 1C). Thus, altered competitive binding could contribute to σE activation in vivo.

Fig. 1 LPS displaces RseA from RseB.

(A) RseB (35 μM) and RseAP (100 μM) eluted 35S-RseB (~1 μM) from RseAP-agarose, whereas buffer controls, bovine serum albumin, (BSA,125 μM), lysozyme (125 μM), or phosphatidylglycerol (PG, 13 mM) did not. (B) LPS from E. coli K12 eluted 35S-RseB from RseAP-agarose (25°C). Data (mean ± SD, n = 3 independent replicates) were fit to a hyperbolic function (Kapp = 270 μM). (C) LptA inhibition of RseB elution by LPS (300 μM) was more efficient at lower temperatures. (D) Structure of E. coli K12 LPS (Kdo, keto-deoxyoctulosonate; Hep, heptose; Gal, galactose; Glc, glucose). The dark gray elements mediate RseB binding.

Minimal LPS Fragments Bind RseB and Facilitate Cleavage of RseA

We sought to test whether fragments of LPS containing lipid A or di-[keto-deoxyoctulosonate]-lipid A (Kdo2–lipid-A; Fig. 1D) also bound RseB, but both molecules had very low solubilities, precluding the interpretation of biochemical studies. However, NaOH hydrolysis of lipid A or Kdo2–lipid A, a treatment that partially removes acyl chains, improved solubility and allowed the purification of active fragments (called L-IIA and Kdo2–L-IIA) with masses expected for the retention of two N-linked acyl chains but the loss of all four O-linked acyl chains (Fig. 2A and fig. S4). The activities of both fragments were similar (Fig. 2B), indicating that the Kdo sugars in Kdo2–L-IIA are not essential for RseB elution. Derivatives of lipid A devoid of acyl chains had no RseB-elution activity. LptA also inhibited RseB elution by Kdo2–L-IIA, with inhibition being less efficient at higher temperature (Fig. 2C). Thus, any LPS derivative that contains the phosphorylated N-acetylglucosamine (GlcNAC) disaccharide and N-linked acyl chains of the lipid-A moiety (colored dark gray in Fig. 1D) appears to bind RseB and displace RseAP.

Fig. 2 Lipid-A fragments disrupt RseAP-RseB complexes.

(A) Mass spectrometry and structure of the L-IIA fragment. (B) L-IIA and Kdo2–L-IIA eluted 35S-RseB from RseAP-agarose (25°C). Data (mean ± SEM; n = 2) were fit to a Hill equation, giving Kapp values of ~140 μM (L-IIA) and ~100 μM (Kdo2–L-IIA). (C) Temperature dependence of LptA competition for Kdo2–L-IIA (140 μM) elution of 35S-RseB from RseAP-agarose. Data are means ± SEM (n = 2). (D) L-IIA competed for RseB (3.5 μM) binding to fl-RseAP (60 nM). Data (mean ± SD, n = 5) were fit to a competition equation (median effective concentration ~ 65 μM). (Inset) L-IIA did not change fl-RseAP (60 nM) anisotropy. (E) After premixing RseB (5 μM) and fl-RseAP (60 nM), dissociation was initiated by adding an equal volume of L-IIA (5 mM). The line is a single exponential fit (rate constant = 0.16 s−1).

Although metabolically irrelevant, L-IIA and Kdo2–L-IIA were useful LPS surrogates because they did not scatter light, permitting the use of fluorescence anisotropy to monitor RseB binding (4). L-IIA dissociated a complex of fluorescent RseAP (fl-RseAP) and RseB in a concentration-dependent manner but did not alter the anisotropy of fl-RseAP alone (Fig. 2D), confirming competition between L-IIA and fl-RseAP for RseB binding (fig. S5). Dissociation of the RseB–fl-RseAP complex was complete within 30 s of addition of L-IIA (Fig. 2E), a time rapid enough to account for the kinetics of the cellular σE response after a stress treatment (11).

Purified RseB is a mixture of dimers and tetramers, with only the dimer binding RseA (4, 12). When we added L-IIA to freshly purified RseB dimers and re-chromatographed the mixture, most protein coeluted with L-IIA at an RseB-tetramer position (Fig. 3A). L-IIA alone eluted near the column salt volume (fig. S6). Thus, L-IIA binds directly to RseB, L-IIA complexes with RseB largely persist during the ~1 hour required for chromatography, and L-IIA binding stabilizes RseB in a tetrameric state that does not bind RseA.

Fig. 3

Lipid-A fragments bind RseB/MucB and coactivate cleavage of RseAP/MucAP. (A) After adding L-IIA (2 mM) to RseB (180 μM), the protein eluted from a gel filtration column as expected for an RseB tetramer (~143 kD). SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and silver staining of the "tetramer" peak fraction showed coelution of RseB and L-IIA. See fig. S6 for gels across the entire included volume. (B) SDS-PAGE assay of RseAP (40 μM) degradation by DegS (4 μM trimer) in the presence or absence of RseB (40 μM monomer), OMP peptide (200 μM), and L-IIA (2 mM). (C) After adding L-IIA (2 mM), MucB (180 μM) largely eluted as a tetramer during gel filtration, although a dimeric shoulder was still evident. (D) L-IIA and Kdo2–L-IIA eluted 35S-MucB from MucAP-agarose. Data (mean ± SD, n = 3) were fit to a Hill equation, giving Kapp values of ~120 μM (L-IIA) and ~83 μM (Kdo2–L-IIA). (E) SDS-PAGE assay of MucAP (25 μM) degradation by AlgW (2 μM trimer) in the presence or absence of MucB (40 μM monomer), OMP peptide (75 μM), and L-IIA (2 mM).

We tested the effects of RseB, L-IIA, and OMP peptide on DegS cleavage of RseAP in vitro (Fig. 3B). As expected (46), DegS alone did not cleave RseAP, the addition of OMP peptide activated cleavage, and the further addition of RseB inhibited this cleavage. However, the addition of L-IIA to OMP peptide restored robust DegS cleavage of RseAP in the presence of RseB. Other combinations did not restore cleavage. L-IIA alone did not activate DegS, confirming that it acts to inhibit RseB.

LPS Plays an Evolutionarily Conserved Role

Most γ and β proteobacteria have RseA, RseB, and DegS orthologs, with MucA, MucB, and AlgW being their functional equivalents in Pseudomonas aeruginosa, a distant relative of E. coli (12). We tested the generality of our findings by examining the interaction of E. coli LPS fragments with the P. aeruginosa proteins. L-IIA largely converted MucB dimers to tetramers (Fig. 3C) and eluted 35S-MucB from a MucAP-affinity column (Fig. 3D), supporting conservation of an LPS-mediated displacement mechanism. L-IIA also allowed OMP-activated AlgW to cleave MucA in the presence of MucB (Fig. 3E). Thus, in both the E. coli and P. aeruginosa systems, an LPS molecule and OMP peptide mimic the signals that result in RseA/MucA cleavage in vivo by inactivating RseB/MucB and activating DegS/AlgW, respectively.

σE Activation by Wild-Type and Mutant LPS

Our results predict that increased periplasmic LPS should activate the cellular σE response. After synthesis on the IM, LPS is shuttled over an Lpt protein bridge and inserted into the OM by the LptD/E translocon at the distal end of the bridge (1315). The LptD∆330-352 variant has fewer functional bridges because of the loss of key disulfide bonds and should cause LPS accumulation in the periplasm (1517). We found that the lptD∆330-352 allele increased expression of a β-galactosidase reporter under σE transcriptional control in a strain requiring inhibition of RseB but not requiring an OMP signal (degS∆PDZ) and in a strain requiring both inhibition of RseB and an OMP signal (degS+) (Fig. 4A and fig. S7). Although RseA cleavage neither requires nor is activated by OMPs in degS∆PDZ cells, because the autoinhibitory PDZ domain of DegS is missing, RseB normally keeps σE activity low (4, 7, 18). Thus, increased σE activity in the lptD∆330-352 degS∆PDZ strain supports a model in which the accumulation of periplasmic LPS relieves RseB inhibition. The activation of σE in lptD∆330-352 degS+ cells suggests that OMP intermediates also accumulate (6, 7), probably as a consequence of defects in LPS assembly. We also decreased the level of properly disulfide-bonded LptD by deleting the DsbA disulfide oxidoreductase (16, 17). Although the ∆dsbA mutation affects many proteins in addition to LptD, it increased σE activity in degS∆PDZ and degS+ strains (fig. S8).

Fig. 4 Activation of a σE-dependent rpoHp3-LacZ reporter under nonstress conditions by LPS-biogenesis mutations.

(A) σE activity was enhanced by lptD∆330-352 in degS+ and degS∆PDZ backgrounds. β-gal, β-galactosidase. (B) σE activity was enhanced in a degS+ strain by numerous LPS-biosynthesis mutations (see fig. S9 for structures). (C) Same as (B) but degS∆PDZ. DegS∆PDZ has lower RseA-cleavage activity than OMP-activated DegS (18), probably accounting for the lower σE activities in (C) as compared to (B) (also see fig. S10). (D) σE activities were similar in ∆rseB degS∆PDZ strains producing wild-type LPS and some of the strongest LPS-biosynthesis mutations from (C). In all panels, data are means ± 1 SD (n ≥ 4).

Using degS+ and degS∆PDZ strains, we tested σE activity in a panel of 11 LPS-biosynthesis mutants to determine whether they relieved RseB inhibition and/or activated DegS. The mutant LPS variants all contained the minimal RseB-interaction motif (fig. S9). Nine mutants elevated σE activity in degS+ and degS∆PDZ strains (Fig. 4, B and C), supporting a model in which activation is linked to LPS inhibition of RseB and to the generation of an OMP signal that activates DegS. In a ∆rseB degS∆PDZ strain in which an OMP signal is not required and negative regulation by RseB is abrogated, the most strongly inducing LPS biosynthesis mutations did not further activate σE (Fig. 4D).

RseB as a Periplasmic LPS Sensor

Our results support a model in which LPS prevents RseB from blocking RseA cleavage. Intact LPS or LPS fragments containing a portion of the lipid-A scaffold bind RseB and release RseA, allowing its cleavage by DegS. Functionally, detection of part of the lipid-A moiety of LPS ensures that RseB can sense and respond to many mislocalized LPS species, including wild-type LPS, incompletely synthesized LPS molecules that reach the periplasm, and LPS variants lacking O-linked acyl chains or core sugars that support E. coli viability but decrease the ability of the OM to protect against cytotoxic agents.

Mutant LPS molecules with the fewest core sugars and phosphates activated σE to the highest levels in rseB+ cells (Fig. 4, B and C, and fig. S9). By analogy with the activation observed when LPS transport was disrupted by partially disabling LptD, we propose that these mutations disrupt synchrony between one or more discrete steps in LPS biosynthesis or transport or OM insertion in ways that increase periplasmic LPS (Fig. 5), which could transit to RseB either through free diffusion or hand-off. During exponential growth, the flux of LPS through the periplasm is ~70,000 molecules min−1. Changes in the coordination of individual biogenesis and transport steps that diverted ~10% of these molecules would result in a ~40 μM min−1 rise in periplasmic LPS (19). Likewise, environmental stress could affect the integrity or function of the Lpt machinery for LPS transport or OM insertion, causing periplasmic accumulation of LPS. Temperature stress might have additional direct effects. LptA, a key component of the transenvelope bridge (Fig. 5) (10, 15), competed less efficiently with RseB for binding to LPS at 45°C, suggesting that heat shock facilitates the transfer of LPS from LptA to RseB, providing a "feed-forward" mechanism for σE activation.

Fig. 5 Dual-signal model for σE activation.

LPSs and OMPs accumulate in the periplasm when stress interferes with the normal pathways of transport and/or OM insertion. Periplasmic LPS binds RseB, freeing RseA to be cleaved by OMP-activated DegS, which initiates the σE stress response.

Dual-Signal Logic and σE Control

Two signals are required for DegS cleavage of RseA when RseB is present in vitro. An OMP signal activates proteolysis by DegS, whereas an LPS signal prevents RseB from blocking the cleavage of RseA. The importance of both signals is also clear in certain genetic backgrounds in vivo. For example, mutations expected to increase periplasmic LPS activate the σE response in degS∆PDZ cells, which require RseB inhibition but not OMP activation. Similarly, the expression of OMP signals alone is sufficient to strongly induce the σE response in ∆rseB cells (7). This requirement for dual signals is reminiscent of an "AND" operation in symbolic or digital logic.

σE activates the transcription of genes for the chaperones and machines that transport and insert both LPS and OMPs into the OM. Crosstalk between these pathways could add an additional layer of regulation (2, 13, 20, 21). If defects in LPS biogenesis created problems with OMP biogenesis and vice versa, then both signals needed to activate σE would be produced when either pathway is perturbed. Indeed, rfaC, rfaF, and rfaP LPS mutations activate σE in strains requiring only RseB inhibition (degS∆PDZ rseB+, Fig. 4C), only OMP activation (degS+ ∆rseB, fig. S11), or both (degS+ rseB+, Fig. 4B). Supporting this idea, the OMs of rfaC, rfaF, and rfaP strains have significantly reduced levels of OMPs (2224), suggesting problems with OMP biogenesis. Conversely, the LptD component of the LPS translocon is transported and inserted into the OM by the same chaperones and machinery as other OMPs (16, 25), and thus defects in OMP biogenesis would soon lead to problems in LPS transport and assembly. Additionally, the combination of AND logic and cross-signaling permits the σE response to reflect the stress level. Low levels of off-pathway OMPs and LPS may be constantly present in the periplasm, because basal cleavage of RseA by DegS is essential for survival (26). Thus, transient increases in either signal would result in an increased but buffered response, whereas a full and metabolically expensive response would be mounted only when RseB and DegS detect high concentrations of off-pathway LPS and OMPs, a condition indicative of extensive dysfunction in OM biogenesis.

Most unicellular organisms are enclosed by cell walls or envelopes that provide both a barrier to and an interface with the environment. These external structures are built from components that are synthesized in the cytosol but are assembled outside of the cytoplasmic membrane. Monitoring the flux of these molecules in the vicinity of the membrane, coupled with dual-signal logic to buffer spurious responses, may be a common regulatory strategy enabling precise homeostatic control of these critical barrier structures.

Supplementary Materials

Materials and Methods

Figs. S1 to S12

Table S1

References (27–32)

References and Notes

  1. Acknowledgments: We thank B. Cezairliyan, S. Chng, D. Kahne, C. Raetz, and D. Six for discussions and strains. Supported by NIH grants AI-16892 and GM-36278. S.L. was supported by an NIH postdoctoral fellowship (AI-084442). The data reported in this paper are presented in the main text and supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article