Fine-Tuning of the Cpx Envelope Stress Response Is Required for Cell Wall Homeostasis in Escherichia coli

ABSTRACT The envelope of Gram-negative bacteria is an essential compartment that constitutes a protective and permeability barrier between the cell and its environment. The envelope also hosts the cell wall, a mesh-like structure made of peptidoglycan (PG) that determines cell shape and provides osmotic protection. Since the PG must grow and divide in a cell-cycle-synchronized manner, its synthesis and remodeling are tightly regulated. Here, we discovered that PG homeostasis is intimately linked to the levels of activation of the Cpx system, an envelope stress response system traditionally viewed as being involved in protein quality control in the envelope. We first show that Cpx is activated when PG integrity is challenged and that this activation provides protection to cells exposed to antibiotics inhibiting PG synthesis. By rerouting the outer membrane lipoprotein NlpE, a known Cpx activator, to a different envelope subcompartment, we managed to manipulate Cpx activation levels. We found that Cpx overactivation leads to aberrant cellular morphologies, to an increased sensitivity to β-lactams, and to dramatic division and growth defects, consistent with a loss of PG homeostasis. Remarkably, these phenotypes were largely abrogated by the deletion of ldtD, a Cpx-induced gene involved in noncanonical PG cross-linkage, suggesting that this transpeptidase is an important link between PG homeostasis and the Cpx system. Altogether our data show that fine-tuning of an envelope quality control system constitutes an important layer of regulation of the highly organized cell wall structure.

tions (12). Interestingly, data from a couple of recent studies suggest that the Cpx system may also detect perturbations to the cell wall in E. coli. First, the Cpx response is activated by the simultaneous absence of four nonessential penicillin-binding proteins (PBPs), which are PG-modifying enzymes (13). Second, several genes that belong to the Cpx regulon are upregulated in the presence of antibiotics that block steps of PG synthesis, although in this case, the direct role of CpxR remains to be determined (14). Moreover, Cpx was recently shown to control the transcription of genes implicated in PG regulation (15). The assembly of PG is a critical aspect of bacterial life, since it underlies the fundamental processes of cell elongation (for lateral growth) and division (to build the new poles of the progeny). Therefore, PG synthesis and remodeling are tightly regulated in space and time, in part via the assembly of complex multiprotein machineries known as the elongasome and the divisome (1).
Here we set out to investigate and dissect the potential and intriguing connection between the Cpx system and the PG. We show that Cpx, a system traditionally viewed as monitoring protein homeostasis in the cell envelope, is also intimately connected to PG assembly. Importantly, we demonstrate that precise control of the degree of Cpx activation is critical to maintain cell wall integrity and we identify LdtD, a transpeptidase catalyzing noncanonical PG cross-links, as a major connector between PG homeostasis and Cpx. Finally, we show that the Cpx system responds massively to the mislocalization of the lipoprotein NlpE, a property that can be used to tune the levels of Cpx activation.

Antibiotics specifically inhibiting elongation or division or inactivating MreB induce the Cpx envelope stress response.
To formally examine if perturbation of PG synthesis induces the Cpx response, we measured the activity of the cpxP promoter (PcpxP), a reliable reporter of the response regulator CpxR (5, 7), following addition of antibiotics targeting PG assembly. We used concentrations of antibiotics that did not arrest cell growth under the tested conditions, to avoid any secondary toxicity effect. By measuring ␤-galactosidase activity from a chromosomal PcpxP-lacZ fusion, we found that a 1-h treatment with amdinocillin (mecillinam), a ␤-lactam antibiotic inhibiting the PG transpeptidase PBP2 (which causes cell rounding), increased Cpx activity by about 2-fold ( Fig. 1A; see Fig. S1A and S1B in the supplemental material). Similar results were obtained with cephalexin, a ␤-lactam inhibiting PBP3 and leading to cell filamentation. Thus, the specific inactivation of essential PG synthesis components of the elongasome (PBP2) or divisome (PBP3) activates the Cpx response. The rod shape of E. coli is maintained by coupling PG synthesis with the actin-like cytoskeleton element MreB. MreB polymerizes in dynamic short filaments distributed perpendicularly to the long cell axis and closely associates with PG synthases, ensuring uniform PG synthesis along the long axis of the cell (16)(17)(18)(19). Interestingly, depolymerization of MreB by addition of subinhibitory concentrations of the drug A22 [S- (3,4dichlorobenzyl)isothiourea], which impacts PG synthesis and makes the cells become round (20)(21)(22), also activated the Cpx system ( Fig. 1A; see Fig. S1A and S1B). In all of the assays described above, no activity was detected in a cpxR deletion strain, showing that induction of the reporter by ␤-lactams or A22 is fully dependent on the Cpx system (Fig. 1A). Of note, this effect was independent on the OM lipoprotein NlpE (see Fig. S1C), as it is for most known Cpxinducing cues (see Discussion). Because measurement of ␤-galactosidase activity only provides an overview of the Cpx response at the population level, it remained possible that ␤-lactams and A22 highly induce this pathway in a minority of cells within that population, while most cells remain unaffected. Thus, in a second assay, we fused the gene encoding a fast-folding variant of the green fluorescent protein (GFPmut2) to PcpxP-used as single-cell readout for Cpx activity-and quantified the fluorescence intensity in cells imaged after treatment or not with amdinocillin or A22 (Fig. 1B). After 1 h of treatment, cells were round as expected when PBP2 or MreB is inhibited (20,21,23). The distribution of the GFPmut2-associated fluorescence signal (normalized by cell area to account for the cell shape variations) clearly shifted toward higher values when cells were grown in the presence of amdinocillin or A22. This increase was also completely abolished in a ⌬cpxR background (Fig. 1B). Hence, our data clearly demonstrate that the Cpx response is induced upon direct inactivation of the division or elongation machinery by antibiotics targeting essential PBPs or perturbing the spatial organization of PG synthesis.
Cpx activation protects cells exposed to ␤-lactams. Our observations, together with previous data showing that the Cpx twocomponent system is induced in a strain lacking several nonessential PBPs (13), suggest that the Cpx response provides a repair or protective mechanism against insults to PG homeostasis. Hence, we hypothesized that the activation of the Cpx system could offer a fitness advantage to E. coli cells exposed to ␤-lactams. Indeed, we found that turning Cpx off by a cpxR deletion increased E. coli sensitivity to amdinocillin and other ␤-lactams ( Fig. 1C; see Fig. S1D in the supplemental mtaterial). This was not due to a compromised permeability function of the OM since ⌬cpxR cells were not more susceptible than wild-type cells to several antibiotics that cannot readily cross the membrane (see Fig. S1E). On the other hand, Cpx induction by NlpE overexpression rendered cells slightly but significantly more resistant ( Fig. 1C; see Fig. S3 in the supplemental material), which was not observed when a Cpxunrelated OM lipoprotein (RcsF) was overexpressed (see Fig. S1F). Altogether, these data indicate that the Cpx system participates in PG homeostasis by sensing cell wall pertubations and mounting a response to minimize the damage.
Overactivation of the Cpx system by CpxA mutants leads to growth, division, and shape abnormalities. Because all processes involved in PG synthesis need to be strictly regulated, we asked if tight control of the level of Cpx response is also required for PG homeostasis. To address this question, we took advantage of previously identified mutations in cpxA (cpxA*) that constitutively activate the Cpx response (24). CpxA* variants lack the phosphatase activity exhibited by many sensor kinases of two-component systems (25), which favors the accumulation of the phosphorylated, active form of CpxR in the absence of an inducing signal. We first constructed CpxA* strains by exchanging the cpxA open reading frame (ORF) on the chromosome by a cpxA* construct (encoding either CpxA L38F G415C or CpxA ⌬93-124 ). Both strains showed 10-to 20-fold increased CpxR-dependent activity on average compared to the wild-type strain (this was higher than that upon induction by NlpE overexpression [see below]) ( Fig. 2A). Interestingly, we noticed that both CpxA* strains displayed significant growth defects: mass doubling times were at least twice as long as those of wild-type cells or CpxA* strains in which cpxR had been deleted (Fig. 2B). Thus, overactivation of the Cpx response negatively affects cell growth. Of note, we repeatedly observed  greater variability of Cpx activation levels and doubling times in the CpxA ⌬93-124 strain ( Fig. 2A and B), which could indicate the occurrence of suppressor mutations. Strikingly, CpxA* strains also displayed morphological aberrations (Fig. 2C), which, together with the observed growth defect, are consistent with a general loss of PG homeostasis. First, most CpxA L38F G415C cells looked elongated or filamentous and DNA-free minicells were observed ( Fig. 2C to E), suggesting a division defect and reminiscent of previous observations (see Discussion) (26). Furthermore, additional hallmarks of PG deregulation were observed in CpxA L38F G415C cells, including cell widening (Fig. 2F), irregular cell width (Fig. 2G), and occasional lysis (not shown). Interestingly, all morphological defects were alleviated when the Cpx system was turned off by deleting cpxR (Fig. 2C, D and F), strongly suggesting that it is the overactivation of the two-component system that leads to a global loss of PG homeostasis impacting growth, division, and shape. Further supporting this notion, we found that the CpxA ⌬93-124 clones that highly activated the Cpx response were elongated, whereas the clones with a low Cpx activity (possible suppressors) displayed normal morphological aspects (Fig. 2H). Thus, there was a remarkable correlation between the level of Cpx activation and the extent of morphological aberrations.
Rerouting the lipoprotein NlpE to the inner membrane also overactivates the Cpx system. Even though the growth and morphological defects of CpxA* strains were abolished by deletion of cpxR, it remained possible that this PG deregulation phenotype is somehow specific to Cpx induction by the CpxA* variants. To test this, we sought to stimulate a strong Cpx response using an alternative mechanism. Overexpression of the OM lipoprotein NlpE has frequently been used as a tool to activate the Cpx system (9,10,12). While the molecular mechanism behind this observation remains mysterious, an attractive hypothesis is that a fraction of the overproduced NlpE fails to be sorted to the OM and accumulates in the IM, where it could directly or indirectly activate CpxA (as suggested in references 9 and 27). We reasoned that if that was the case, rerouting the whole population of NlpE to the IM should induce the Cpx response more than overexpression of the wildtype protein. We modified selected residues of the lipobox of NlpE according to lipoprotein sorting rules (28) in order to change its final destination and produce an IM-anchored lipoprotein (NlpE IM ), which was verified by membrane fractionation (see Fig. S2A and B in the supplemental material). We first expressed this mislocalized form in an nlpE deletion strain to avoid any contribution of the native, properly targeted protein. As predicted, NlpE IM turned on the Cpx system more strongly than the OM-localized NlpE (Fig. 3A). This is in agreement with previous data suggesting that IM-targeted NlpE induces the Cpx response (27), although in that case, the activity of a different, nonspecific Cpx reporter was monitored (29). The effect of mislocalized NlpE was not restricted to strains lacking the native copy of nlpE, since similar Cpx activity was observed when the mislocalized NlpE form was produced in a strain carrying the native, chromosomal nlpE (data not shown). Furthermore, in these experiments, NlpE IM was produced at comparable levels to the wild-type protein expressed from the same plasmid, excluding the possibility that the strong Cpx response is due to larger amounts of NlpE (see Fig. S2A). Hence, our data show that the Cpx system is especially sensitive to mislocalized NlpE.
Growth and morphological defects upon Cpx overactivation are not restricted to CpxA* mutants. Next, we asked if Cpx overactivation upon NlpE mislocalization also affected several PGrelated aspects. First, we found that the growth rate of cells expressing the mislocalized variant of NlpE (NlpE IM ) was slowed down (Fig. 3B), similar to our observations with CpxA* strains (Fig. 2B). This shows that the growth defect occurs as a result of a strong Cpx response, independently of mutations in a core member of the two-component system. Consistently, NlpE IM -carrying cells were also characterized by dramatic morphological aberrations, including cell filamentation, minicell formation, and larger cell width, which were completely abolished in the absence of the response regulator CpxR ( Fig. 3C; see Fig. S2C and D in the supplemental material). All phenotypes induced by mislocalized NlpE were independent of the presence of the chromosomal copy of nlpE (data not shown), ruling out possible effects of either the coexistence of two different forms of the protein or the deletion of the native nlpE.
A member of the Cpx regulon implicated in PG cross-linking contributes to the growth and morphological defects in Cpx overactive cells. Taken together, our data show that turning on the Cpx response to high levels (i.e., more than 10-fold compared to basal activity) produces pleiotropic effects on growth, shape and division, likely reflecting misregulation of PG synthesis and/or remodeling. Since CpxR up-regulates at least three genes known to play a role in shaping the PG structure (8,15,30,31), we hypothesized that some of the observed defects could be alleviated by the deletion of one or several of these genes. To test this, we expressed NlpE IM in single-deletion strains lacking either ldtD (also known as ycbB), slt, or ygaU, three of the identified PG modifier genes controlled by Cpx (8,15). Remarkably, the absence of ldtD, but not of slt or ygaU (data not shown), significantly improved cell growth (Fig. 4A), reduced the amount of filaments (Fig. 4B), and restored normal cell width (Fig. 4C), without preventing Cpx overactivation by the IM-targeted NlpE variant (Fig. 4D). Moreover, mislocalization of NlpE (high Cpx response) rendered wild-type cells more sensitive to amdinocillin (Fig. 4E) and other ␤-lactams (see Fig. S3 in the supplemental material), and deletion of ldtD suppressed this sensitivity (Fig. 4E). Thus, cells with high levels of Cpx activation display growth and mor- phological defects and become sensitized to antibiotics targeting PG synthesis in a manner that depends, at least partially, on ldtD.

DISCUSSION
The Cpx system senses PG damage. The simultaneous absence of four nonessential PBPs turns on the Cpx response, supporting an extended role of this system in monitoring perturbations in the cell wall (13). We now provide compelling evidence for this by showing that Cpx senses inhibition of essential components of both the elongasome and the divisome, as well as the loss of spatial coordination of PG synthesis during growth. However, the molecular nature of the signal that stimulates the Cpx twocomponent system when cells are grown in the presence of PGperturbing compounds like ␤-lactams or the MreB inhibitor A22 Fine-Tuning of Cpx Controls Cell Wall Homeostasis remains to be identified. Cpx likely senses PG damage independently of NlpE since the absence of this lipoprotein did not prevent induction of the Cpx response by ␤-lactams or A22 (see Fig. S1C in the supplemental material), consistent with the fact that NlpE is dispensable for Cpx activation by most inducing cues (5). Actually, it is still unresolved how the kinase activity of CpxA is triggered by the accumulation of misfolded proteins in the envelope, which is considered a major Cpx-activating signal. Clearly, further investigation is needed to shed light on how a protein quality control system such as Cpx is activated by cell wall alterations and other envelope stresses. Cpx overactivation leads to division, growth, and shape defects that are largely dependent on ldtD. A previous study had reported division defects in CpxA* strains, but it was unknown if this phenotype was due to a strong induction of the Cpx response or to side effects of the CpxA* mutations (26). Here we demonstrate that a robust Cpx response triggered either by CpxA* or NlpE mislocalization gives rise to several problems pointing to a general imbalance in PG assembly and remodeling. Moreover, these phenotypes were largely suppressed by deletion of the CpxR-induced gene ldtD, identifying LdtD as a major functional link between the observed PG-related defects and Cpx. Deletion of either slt or ygaU, two other Cpx-induced genes with a known PG-modulating function (15), did not alleviate the morphological aberrations, while double ldtD slt or ldtD ygaU deletions did not improve the compensatory effect of ldtD deletion alone (data not shown), consistent with the primary role of ldtD. However, overexpression of ldtD by itself was not sufficient to recapitulate the morphological and growth defects generated by Cpx overactivation (data not shown). Thus, one or more unidentified Cpxdependent factors may be needed to produce these LdtDdependent phenotypes. Interesting to note, exponentially growing cells lacking ldtD did not trigger the Cpx response (Fig. 4D), contrary to results obtained by Bernal-Cabas et al. using the same reporter (15). The fact that these authors used cells growing in the late exponential phase, a condition known to induce the Cpx system (5), could explain this discrepancy.
Recently, mutations in cpxA were shown to suppress the lethality caused by the loss of the PG amidase AmiB in Pseudomonas aeruginosa, suggesting a role for Cpx in PG-related aspects of cell division in this bacterium (32). Interestingly, these alleles did not confer growth defects, which could reflect species-specific differences in the strengths of the Cpx responses generated by these CpxA variants or in the connection between Cpx activation levels and PG regulation. How does LdtD contribute to PG-linked defects in division, growth, and shape? LdtD is an L,D-transpeptidase that generates normally poorly abundant cross-links between diaminopimelic acid (DAP) residues (31,33). Consistently, a constitutive, moderate Cpx activation (~10-fold higher than the basal level) increases the overall degree of cross-linkage of the sacculus, partly by increasing the relative amount of DAP-DAP bonds (15,24). In this case, the cell wall is further stabilized (33). Our results suggest that in strains with a disproportionate Cpx response, structural modifications of the PG and the abundance of DAP-DAP cross-links in particular cause defects in cell division, growth, and morphology instead of strengthening the sacculus.
Since the sequential process of cell division is intimately coupled to the assembly and remodeling of the PG structure, a severe imbalance in the different types of bonds in the PG mesh by upregulation of LdtD, and possibly by other Cpx-dependent effects, may have direct disruptive consequences on divisome assembly, stability, or constriction. For instance, DAP-DAP-bridged muropeptides are poor substrates for amidases (34), yet the action of amidases at the septum contributes to the formation of denuded glycan chains that help recruiting the late divisome component FtsN (35), a protein that favors stabilization and constriction of the cytokinetic FtsZ ring driving cell septation (36,37). Thus, an attractive hypothesis to link the observed division defects with LdtD overexpression is that large amounts of DAP-DAP bonds weaken the recruitment of FtsN, leading to a less stable and/or efficient cytokinetic ring. This idea is consistent with the filamentation phenotype of Cpx overactive cells and the mitigation of this effect in an ldtD-deficient strain. Moreover, it is supported by our observations that only a few CpxA* cells that initiated constriction displayed a clear band of GFPmut2-tagged FtsZ (see Fig. S4A in the supplemental material), while most filamentous CpxA* cells displayed dynamic FtsZ clusters or bands (see Fig. S4B), indicating an unstable divisome. We verified that this phenotype was not due to lower FtsZ protein levels (see Fig. S4C).
In addition, we propose that the overproduction of LdtD also impairs cell growth and shape: by causing a strong increase in overall cross-linkage and of DAP-DAP bonds in particular, it could hinder both the incorporation of PG precursors into the preexisting mesh and the release of old material. In another scenario, an excess of LdtD could generate an imbalanced enzymatic composition of the large multiprotein complexes involved in PG remodeling, thereby affecting growth and shape. The degree of Cpx response determines sensitivity to a loss of PG integrity. Our data show that sensitivity to ␤-lactam is affected by the extent of Cpx response. In this context, we propose the following model. As Cpx is turned on up to moderate levels (similar to those obtained when overexpressing NlpE), the tolerance of cells to a loss of PG integrity increases because of a gradual strengthening of the cell wall thanks to Cpx-induced cross-linkage and DAP-DAP bonds. Indeed, the synthesis of L,D-peptide bonds like DAP-DAP is insensitive to ␤-lactams and an increased content of this type of cross-links in the PG has been associated with increased resistance to these antibiotics (38). However, when a threshold level of Cpx activation is reached, LdtD (and possibly additional PG-modifying enzymes) is produced in excess. This generates structural modifications to the PG sacculus, including large amounts of DAP-DAP, such that cells are sensitized to additional PG insults.
Remarkably, Cpx plays a critical role in the virulence of bacterial pathogens (39). For several species, the degree of Cpx response changes during the process of infection and is critical for the survival of bacteria inside and outside their host (40)(41)(42), further emphasizing the importance to keep this broad-spectrum stress response system in check.
Conclusions. Our major findings are recapitulated in Fig. 5. First, our study challenges the traditional view of the Cpx envelope stress response as a protein quality control system, by providing clear evidence for a connection between Cpx and PG homeostasis. Moreover, we show that fine-tuning the Cpx stress response is critical to maintain cell wall homeostasis, affecting cell growth, morphology, division, and antibiotic resistance. Our study also identifies ldtD, a Cpx-induced gene involved in noncanonical PG cross-bridging, as a functional link between PG remodeling and Cpx. Finally, we show that modifying the localization of the lipoprotein NlpE within the envelope can be used to modify the levels of the Cpx response.

MATERIALS AND METHODS
Bacterial strains, media, and plasmids. All strains and plasmids, including construction methods, can be found in Table S1 in the supplemental material. Primers are listed in Table S2 in the supplemental material. Cells were grown in LB medium at 37°C, except when indicated. For strains carrying cpxA* mutations, starter overnight cultures were always grown at 30°C to minimize the occurrence of suppressors as suggested in reference 26. To avoid any effect of Cpx activation that starts in the late exponential phase (5,43), all experiments were performed from cultures grown until the early or mid-log phase (optical density at 600 nm [OD 600 ] of Յ0.6) after diluting an overnight inoculum usually 1:1,000 (or never less than 1:500) in order to ensure exit from the stationary phase. Antibiotics were used for plasmid maintenance when appropriate at the following concentrations: ampicillin, 200 g/ml; spectinomycin, 50 g/ml; kanamycin, 50 g/ml; and chloramphenicol, 20 g/ml.
Antibiotic sensitivity assay. To measure sensitivity to ␤-lactams, Sensi-Discs were prepared on the day of the assay by adding a 20-l drop containing 10 g amdinocillin, 10 g ampicillin, or 30 g cephalexin dissolved in water on 6-mm paper discs (Becton Dickinson). To measure susceptibility to other antibiotics (see Fig. S1E in the supplemental material), we used 6-mm paper discs preloaded with 30 g vancomycin, 5 g novobiocin, 10 g bacitracin, 15 g erythromycin, or 25 g rifampin (Becton Dickinson). Cells at an OD 600 of 0.5 (100 l or 1 ml) were mixed with LB top agar (3 or 10 ml), containing spectinomycin when required for the maintenance of pAM238 vectors, and poured on top of an LB agar plate. Sensi-Discs were placed on the solidified mixture at a minimum 3-cm distance from each other and from the plate border. The assay was performed in triplicate for each strain. The diameter of growth inhibition around each disc was measured after overnight incubation at 37°C. When indicated, the relative growth inhibition was calculated as follows: each diameter value was normalized by the average diameter obtained from three replicates of the control strain, and the resulting triplicates of normalized values were then averaged for each tested strain. Bar graphs were prepared using Prism 6 (GraphPad Software, Inc.).
Microscopy image acquisition and analysis. Live cells were spotted on 1% agarose pads (prepared with phosphate-buffered saline [PBS]) between a glass slide and a coverslip. When indicated, cells were stained with 5 g/ml DAPI (4=,6-diamidino-2-phenylindole) (Sigma-Aldrich) and 5 g/ml FM4-64 (Life Technologies) just before imaging. Image acquisition, analysis, and processing were performed as described previously (45) using filter sets 31, 49, and 46 (Carl Zeiss) to image FM4-64, DAPI, and GFPmut2-associated fluorescence, respectively. We used a parameter set modified from algorithm 1 of the MicrobeTracker suite (46) to obtain subpixel cell outlines. Quantitative analysis and plots from the MicrobeTracker data were done on MATLAB (MathWorks, Inc.) using homemade scripts. The values of cell width were obtained by dividing the cell area by the cell length.  ) improves resistance to ␤-lactam antibiotics, while cells unable to activate the system ("OFF" [⌬cpxR]) are more sensitive than wild-type cells with basal Cpx response (WT), defining an "unprotected zone" and a "comfort zone" based on the degree of Cpx activation. However, excessive activation of the Cpx system by mislocalized NlpE (NlpE IM ) or by CpxA* mutants gives rise to a loss of PG homeostasis, manifested by cell division and shape defects, slower growth, and increased sensitivity to ␤-lactams ("danger zone"). Cell morphology and the induction level of the Cpx response relative to the wild type are represented schematically.
Growth analysis. Single colonies were used to inoculate overnight cultures, which were diluted 1:500 in round-bottom 96-well plates in 200 l LB medium with antibiotics when appropriate for plasmid maintenance. Absorbance measurement was performed at 600 nm every 5 min in a Synergy H1 microplate reader (BioTek) with constant orbital shaking at 37°C. Doubling time, t, was calculated as t ϭ log 2 /k, k being the growth rate calculated as the maximum slope from 10 consecutive points on the semilogarithmic curves. Bar graphs were prepared using Prism 6 (Graph-Pad Software, Inc.) Membrane fractionation. Overnight cultures were diluted 1:500 and grown at 37°C. For each strain, 400 ml of cells was harvested when the OD 600 reached 0.6 and cell fractionation was performed using a two-step sucrose gradient as described previously in order to separate the IM from the OM (45). ImageJ (http://imagej.nih.gov/ij) was used to quantify the percentage of NlpE detected in each fraction.
Western blotting. Proteins from exponentially growing cultures were precipitated with trichloroacetic acid as described previously (47), solubilized in 1ϫ nonreducing Laemmli buffer (48) (with the volume for each sample adapted to normalize all samples by OD 600 ), and boiled before being loaded on precast NuPAGE Bis-Tris gels (Life Technologies). Western blotting was performed using standard procedures with the following primary antibodies: anti-FtsZ (rabbit polyclonal serum; AgriSera), anti-PtsI, anti-NlpE, anti-Lpp (rabbit sera; CER Group, Marloie, Belgium), or anti-DsbD␣ (49) followed by a horseradish peroxidase (HRP)-conjugated anti-rabbit antibody (Sigma). Chemiluminescence signal was imaged with a GE ImageQuant LAS4000 camera (GE Healthcare Life Sciences).
Statistical analysis. Statistical tests were performed using Prism 6 (GraphPad Software, Inc.). For Fig. 1A and Fig. S1B in the supplemental material, we used an unpaired one-tailed Student's t test with assumed equal variances (with each condition compared to the untreated control). For Fig. 1C and 4A and Fig. S1C, we used an unpaired two-tailed Student's t test with assumed equal variances for compared groups. For Fig. 3B and 4E and Fig. S1F, we used an unpaired, one-way analysis of variance (ANOVA) with Sidak's multiple comparison test, with a single pooled variance. For Fig. S1A, S1D, and S1E and Fig. S3 in the supplemental material, we used a two-way ANOVA with Sidak's multiple comparison test. The reported P values for all ANOVA tests were adjusted for multiplicity. For all statistical tests, P values of Ͻ0.05 were considered statistically significant.