The NlpE N-terminal domain is sufficient for resistance to lipoprotein trafficking stress.

E. coli NlpE is produced with two globular domains (43). The N-terminal domain, NlpE_1–101, is homologous to bacterial lipocalin Blc (Fig. 2A). The C-terminal domain, NlpE_121–216, contains an oligonucleotide/oligosaccharide-binding (OB) fold (Fig. 2A). The two domains are joined by a linker region (Fig. 2A). Prior work had solved the crystal structure of a domain-swapped E. coli NlpE dimer (43). This structure led to the prevailing model of NlpE signaling to Cpx: in its inactive (nonsignaling) form, NlpE is proposed to adopt a compact conformation mediated by interactions between N- and C-terminal domains (43). Adhesion cues were suggested to trigger an extended NlpE conformation that enabled it to span the periplasm, allowing the C-terminal domain to activate CpxA in the IM (43). This model implied that the NlpE C-terminal domain is the source of signaling.

Open in a separate window

FIG 2

The NlpE N-terminal domain is sufficient for tolerance of LolB depletion and activation of Cpx. (A) Schematic of NlpE structure in its extended conformation. The N-terminal domain (orange) is joined to the C-terminal domain (blue) via a linker region (black). Sites of truncations are marked with spheres; green spheres indicate truncations that are able to activate Cpx, red spheres indicate truncations that fail to activate Cpx, and gray spheres show the Cys residues in a putatively redox-sensitive CXXC motif. (B) nlpE mutants were tested for their ability to tolerate LolB depletion (− arabinose) in an lpp(ΔK58) ΔosmB background. (C) Relative LacZ levels in ΔnlpE cells harboring a PcpxP-lacZ transcriptional reporter and overproducing plasmid-borne NlpE variants targeted to the OM. (D) Relative LacZ levels in ΔnlpE cells encoding a PcpxP-lacZ transcriptional reporter and overproducing plasmid-borne NlpE(DD) variants targeted to the IM. Data are means ± standard deviations.

Many Gram-negative species produce NlpE homologs that lack the C-terminal domain (43). The extended NlpE conformation model for signaling is inadequate to explain how these NlpE proteins function, since the N-terminal domain is small (approximately 50Å) and unlikely to span the periplasm (>200Å). This phylogenomic comparison suggested that the N-terminal domain may have a discrete signaling function. To test this hypothesis, we constructed strains that produce truncated NlpE proteins, each lacking the C-terminal domain (summarized in Table 1). We used CRISPR-Cas9 gene editing of the native nlpE gene to delete sequences encoding the C-terminal region. The resulting constructs were additionally tagged with FLAG epitopes at their C termini. The full-length nlpE⁺-flag construct exhibited wild-type activity and resulted in a toleration of LolB depletion (Fig. 2B). We found NlpE truncations that either removed the C-terminal domain (NlpE_1–121) or removed both the C-terminal domain and the linker region (NlpE_1–101) resulted in a toleration of LolB depletion (Fig. 2B). Despite the loss of the C-terminal domain, these truncated NlpE proteins were functionally equivalent to the full-length wild-type NlpE. Two larger truncations that included portions of the N-terminal domain (NlpE_1–94 and NlpE_1–81) were nonfunctional and resulted in cells that were as sensitive to LolB depletion as ΔnlpE cells (Fig. 2B). Therefore, our data demonstrate that the NlpE N-terminal domain is sufficient to overcome stress caused by LolB depletion.

The NlpE N-terminal domain is sufficient to activate Cpx signaling.

Prior work established that NlpE overproduction activates the Cpx response (44,–46). We assessed whether NlpE truncations that enable tolerance of LolB depletion could signal to activate Cpx. Each of the truncated nlpE alleles was cloned into pND18, a plasmid that allows l-arabinose-inducible expression of nlpE alleles and that has been used previously to demonstrate Cpx signaling (45). Plasmids were transformed into strain MG3593 that harbors a cpxP-lacZ⁺ transcriptional reporter whose expression is CpxR dependent and lacks the native nlpE gene. We supplemented the medium with l-arabinose to induce overproduction of plasmid-encoded NlpE constructs. LacZ levels in response to induction were measured. As expected, overproduction of full-length NlpE strongly activated the Cpx response (Fig. 2C). We measured equally strong Cpx activation when overproducing truncated NlpE proteins that either lacked the C-terminal domain (NlpE_1–121) or lacked both the C-terminal domain and the linker region (NlpE_1–101) (Fig. 2C). Truncations that included portions of the N-terminal domain were unable to activate Cpx (NlpE_1–94 and NlpE_1–81) (Fig. 2C). Importantly, Cpx activation by full-length NlpE or NlpE lacking the C-terminal domain was entirely dependent on the IM CpxA protein (see Fig. S2).

Given that NlpE is an OM-targeted lipoprotein, we considered that truncations which fail to activate Cpx signaling may be too short to efficiently access CpxA but could remain otherwise functional for signaling. Therefore, we generated IM-localized NlpE constructs that could facilitate interaction with CpxA in the IM. We mutated the +2 and +3 residues in each of our nlpE alleles to Asp, generating the well-established strong IM retention signal that blocks lipoprotein entry into LolCDE (31). Indeed, the resulting IM localization of NlpE(DD) proteins was previously validated by cellular fractionation (42, 44). As in earlier experiments, we overproduced these constructs in MG3593 ΔnlpE and measured resultant LacZ levels. We detected stronger Cpx activation with IM-targeted full-length NlpE, in agreement with prior work (Fig. 2D) (42). The IM-localized NlpE(DD) proteins either lacking the C-terminal domain [NlpE(DD)_1–121] or lacking both the C-terminal domain and the linker region [NlpE(DD)_1–101] activated Cpx to the same extent as the full-length protein [NlpE(DD)] (Fig. 2D). The NlpE(DD) proteins that included truncations within the N-terminal domain did not activate Cpx [NlpE(DD)_1–94 and NlpE(DD)_1–81] (Fig. 2D). Clearly, the NlpE N-terminal domain is sufficient for Cpx activation; moreover, deletions within this domain abolish its ability to activate Cpx.

The Cpx response protects against inhibitors of lipoprotein trafficking.

Since we had observed that NlpE activation of Cpx was important for resisting lipoprotein trafficking stress caused by LolB depletion, we sought to determine whether this effect was LolB specific or whether other lipoprotein trafficking stresses were also alleviated by NlpE and Cpx. To this end, we exploited two chemical inhibitors of lipoprotein trafficking to induce stress: we used compound 2 (Cpd2), a pyrazole compound that directly inhibits trafficking by interfering with LolCDE function (47), and we used globomycin (Glb), an inhibitor of signal peptidase II (Lsp) that indirectly blocks lipoprotein trafficking by preventing lipoprotein maturation (48,–50).

We examined the effect of ΔnlpE and ΔcpxR mutations to E. coli survival when treated with Glb and Cpd2. Using a wild-type background, we observed ΔcpxR mutants were more sensitive to Glb and Cpd2 (Table 2). Inactivating nlpE caused cells to become more sensitive to Cpd2, but it did not appreciably affect sensitivity to Glb (Table 2). Both Glb and Cpd2 are efficiently effluxed from E. coli. CpxR regulates genes encoding efflux pumps that each work via the TolC OM efflux pore. Hence, one possibility was that Cpx protects against Glb and Cpd2 simply by increasing their efflux. We deleted tolC to inactivate all efflux systems and tested Glb and Cpd2 sensitivity in ΔcpxR and ΔnlpE derivatives. As expected, the ΔtolC mutation resulted in a striking increase in sensitivity to both compounds (Table 2). In this ΔtolC background, ΔcpxR mutants were more sensitive to both Glb and Cpd2 than ΔtolC cpxR⁺ cells. Clearly, CpxR is required to protect against the effects of Glb and Cpd2 in a manner that is independent of its regulation of efflux systems (Table 2). The ΔtolC ΔnlpE strain was more sensitive to Cpd2, but Glb sensitivity was unchanged compared to that in ΔtolC nlpE⁺ cells (Table 2), suggesting that while NlpE was important for resistance to Cpd2, it was not required for Glb resistance. Hence, NlpE is the sole activator of Cpx when LolCDE is inhibited by Cpd2, but NlpE is not absolutely required to trigger Cpx activation in response to Glb treatment. Our data show that the Cpx system is required to protect cells against diverse lipoprotein trafficking stresses, not just LolB depletion.

Cpx senses lipoprotein trafficking stress by monitoring nascent NlpE transiting the IM.

As an OM-targeted lipoprotein, NlpE is hypothesized to signal from the OM. Structural data suggest that the folded N-terminal domain (NlpE_1–101) is too small to span the periplasm to activate CpxA. However, our results demonstrate that NlpE_1–101 is sufficient to activate Cpx signaling. We considered an alternate hypothesis for Cpx activation: NlpE signals lipoprotein stress to Cpx from the IM while it is being acylated and/or awaiting trafficking to the OM. Defects in either process would cause nascent NlpE to accumulate in the IM and correspondingly lead to increased Cpx activation.

To test our hypothesis, we treated wild-type E. coli cells with sub-MIC amounts of Glb and Cpd2 and then examined the early Cpx response. We used reverse transcription-quantitative PCR (qRT-PCR) to measure the levels of cpxP mRNA following treatment (expression of cpxP is CpxR dependent) (51). We used dimethyl sulfoxide (DMSO) as a vehicle control for mock treatment. We detected a strong increase in cpxP levels in response to either Glb or Cpd2, indicating that the Cpx response was activated (Fig. 3). We hypothesized that NlpE is responsible for the Cpx activation in response to inhibitor treatment. Indeed, in a ΔnlpE strain, cpxP mRNA levels were not increased in response to Cpd2 treatment (Fig. 3). Cpx activation in response to Cpd2 is entirely dependent on NlpE signaling. This finding is consistent with the equivalent sensitivity of ΔcpxR and ΔnlpE mutants to Cpd2: the Cpx response is protective and it requires NlpE for activation. The ΔnlpE mutation only partially impaired Cpx activation following Glb treatment; we still detected increased cpxP transcription, albeit to a lesser extent (Fig. 3). These data showed that while NlpE does activate Cpx in response to stress caused by Glb, unknown additional factors also contribute to Cpx activation to Glb. This finding is also consistent with our MIC data: the Cpx response is protective, but ΔnlpE does not sensitize cells to Glb (like ΔcpxR) because Cpx can apparently be activated some other way. We have ruled out any contributions of Blc or YafY in responding to Glb (see Fig. S3).

Open in a separate window

FIG 3

Inhibitors of lipoprotein biogenesis Glb and Cpd2 activate Cpx through NlpE. Cells were treated with either Glb or Cpd2 lipoprotein trafficking inhibitors (or DMSO vehicle control) for 20 min. RNA was then extracted and subjected to qRT-PCR to quantitate levels of cpxP mRNA. Ksg-treated cells (+Ksg) were treated with a sub-MIC of Ksg for 15 min prior to Glb or Cpd2 treatment (see Materials and Methods). Data are means ± standard errors of the means.

In a parallel set of Cpd2 and Glb experiments, we pretreated cells with the protein synthesis inhibitor kasugamycin (Ksg), which blocks the initiation of translation, stopping the production of new NlpE (and all other proteins) (52). During Ksg treatment, already-synthesized NlpE molecules have time to be secreted, acylated, and trafficked to the OM, but there are no new NlpE molecules going through these steps. Therefore, Ksg-treated cells that are subsequently treated with Glb or Cpd2 inform on signaling originating from OM-localized NlpE. This Ksg treatment approach was recently used to uncover OM-specific signaling in the Rcs stress system (53). Because Ksg-untreated cells continue to synthesize, acylate, and traffic new NlpE, while cells are treated with Glb and Cpd2, signaling can originate from OM-localized NlpE as well as from IM-localized NlpE that is en route to the OM. By measuring cpxP mRNA levels, we found that Ksg pretreatment blocked Cpx activation in response to both Cpd2 and Glb. To activate the Cpx response to these inhibitors, the cell must be synthesizing new proteins (Fig. 3). Since Ksg-treated cells do not activate the Cpx response to Cpd2 or Glb, NlpE signaling that activates Cpx in Ksg-untreated cells must originate from newly synthesized NlpE in the IM. It is clear that OM-localized NlpE molecules do not contribute to sensing lipoprotein trafficking stress caused by Cpd2 or Glb treatment.

NlpE activates Cpx to protect against Cu toxicity.

Our findings indicated that NlpE is a sensor of lipoprotein trafficking stress that activates Cpx to protect the cell. Cpx was previously implicated in resistance to Cu toxicity (54, 55). We wanted to examine whether the ability of Cpx to sense stress caused by lipoprotein trafficking defects or Cu toxicity was due to distinct or linked functions. We also sought to determine whether NlpE was important for resisting Cu stress, since earlier studies yielded conflicting conclusions about its involvement. At first, nlpE (then named cutF) was found to be required for Cu resistance, since nlpE mutations sensitized cells to Cu (56). However, a later study demonstrated that while cpxRA mutants were sensitive to Cu, loss of nlpE had no effect on Cu resistance (55). Given these opposing findings, we reexamined Cu sensitivity in isogenic ΔcpxR and ΔnlpE mutants. We found that ΔcpxR caused a severe sensitivity to 4mM Cu (Fig. 4A). The ΔnlpE mutation also sensitized cells to 4mM Cu, though to a significantly lesser extent (Fig. 4A). Hence, both cpxR and nlpE genes are required to protect against Cu toxicity, though the loss of cpxR makes cells significantly more sensitive to Cu than loss of nlpE. The N-terminal domain of NlpE was sufficient to confer Cu resistance (see Fig. S4).

Open in a separate window

FIG 4

Cu impairs lipoprotein biogenesis and activates Cpx through NlpE. (A) Cultures were serially diluted on LB agar and LB agar supplemented with 4mM CuCl₂. (B) Cultures were grown in the presence of 3mM CuSO₄ to mid-log phase, and levels of cpxP mRNA were measured. Samples were prepared and analyzed together with samples presented in Fig. 3, the DMSO control presented is the same here as in Fig. 3 Data are means ± standard errors of the means. (C) Relative LacZ levels in ΔnlpE cells harboring a PcpxP-lacZ transcriptional reporter and overproducing plasmid-borne NlpE variants targeted to the OM. Data are means ± standard deviations. (D) Cultures were grown to mid-log phase in the presence (Cu +) or absence (Cu −) of 3mM CuCl₂. Lgt and Lnt replete (+) or deplete (−) samples were obtained by growing strains PAP9403 and KA472 in the presence or absence of arabinose; LspA activity was inhibited by treating cells with Glb (LspA −) in comparison to mock treatment (LspA +). Protein samples were taken and probed for Lpp by immunoblotting. Diacyl form Lpp is noted as +2. Lpp forms with signal peptides attached are noted as +SP. Peptidoglycan-bound Lpp forms are noted as * and **. See text for details.

Due to the clear involvement of both NlpE and CpxR in resistance to Cu, we would expect NlpE to activate Cpx when cells encounter Cu stress, as was previously reported. We used qRT-PCR to quantify levels of cpxP mRNA in wild-type E. coli and observed modest activation of Cpx within 20 min of Cu treatment (Fig. 4B), supporting earlier reports (55, 76). Importantly, we determined that this activation was entirely dependent on NlpE, since ΔnlpE cells failed to increase transcription of cpxP when treated with Cu (Fig. 4B).

Given that we detected NlpE-dependent activation of Cpx and that both factors are required for resistance to Cu, we sought to understand how NlpE senses Cu. Within the NlpE N-terminal domain there is a conserved 31-CXXC-34 motif that is in an oxidized state in the crystal structure, forming a disulfide bond (43). The chemically active Cys residues were hypothesized to offer a metal binding site in their reduced state (43). In nascent NlpE secreted from the cytosol, the Cys residues would be in a reduced state. The reaction of Cu with the thiol groups of these Cys residues would preclude disulfide bond formation, and this was proposed to change the structural properties of NlpE so that signaling to Cpx would occur (43). We tested this hypothesis directly by making cysteine to serine substitutions in the CXXC motif.

We used site-directed mutagenesis of an NlpE-encoding plasmid (pND18) to substitute each Cys to a Ser and to make a double Cys-to-Ser mutant. We transformed these plasmids into a cpxP-lacZ transcriptional reporter strain lacking the native nlpE gene (MG3593). NlpE mutant variants were each overproduced, and we assessed their ability to activate the Cpx response by measuring the levels of LacZ produced. We observed no effect of mutating either or both Cys residues; each of the mutants signaled as well as the wild-type protein (Fig. 4C). Clearly, the Cys residues in CXXC are not required for NlpE to activate Cpx, and altering the Cys residues is not sufficient to activate Cpx. Therefore, it is clear that the presence or absence of the C31-C34 disulfide bond is neither inhibitory nor stimulatory for NlpE signaling to Cpx.

Construction of truncated NlpE plasmids and chromosomal alleles.

Plasmid pND18 (45) encoding full-length NlpE was mutagenized using Q5 site-directed mutagenesis (NEB) with the oligonucleotides listed in Table S3 to generate plasmid derivatives encoding truncated NlpE. CRISPR-Cas9 editing was used to generate chromosomal nlpE truncations (73). An nlpE guide RNA was constructed in pCRISPR using primers CRISPR_NlpE_F and CRISPR_NlpE_R, generating plasmid pCRISPRnlpE. NlpE alleles were PCR amplified from plasmids using primers NlpE_ampli_F and NlpE_ampli_R, and the products were cotransformed with pCRISPRnlpE into the recombinogenic E. coli strain HME63 which carried yafC::Tn10 markers (MG3670) (74). The resulting truncations of the native nlpE gene were confirmed by sequencing. The generated nlpE alleles were moved using colinkage with yafC::Tn10.

Efficiency of plating assays.

Efficiency of plating assays was used to determine the relative sensitivities of strains to CuCl₂ or growth in the absence of l-arabinose (in strains with arabinose-dependent lolB expression). Assays were performed by preparing serial dilutions (2-fold or 10-fold) of overnight cultures (standardized by A₆₀₀) in 96-well microtiter plates before replica plating onto LB agar and selective medium and then incubating plates overnight at 37°C.

MIC determination.

Overnight cultures were diluted (10⁵ cells/ml) and transferred to a 96-well plate. Two-fold serial dilutions of either compound 2 or globomycin were prepared in DMSO and added to each of the cell suspensions. Plates were incubated statically, overnight at 37°C, and A₆₀₀ was measured using a BioTek Synergy H1 plate reader. The MIC was determined as the lowest concentration that completely inhibited growth.

β-Galactosidase assays.

LacZ levels were measured as previously described (75). Briefly, overnight cultures of strains carrying NlpE-encoding plasmids were subcultured 1:100 for 2h in medium supplemented with 0.01% arabinose to induce NlpE production. An equivalent number of cells (as determined by A₆₀₀) was collected and resuspended in 100μl before being permeabilized with 30μl 0.1% SDS and 40 μl chloroform. Z-buffer (830μl) was added and ortho-nitrophenyl-β-d-galactopyranoside (ONPG) hydrolysis was measured every 1min over a 15-min kinetic assay, allowing V_max to be calculated.

Depletion of Lgt and Lnt.

To deplete Lgt or Lnt, strains harboring arabinose-dependent expression of Lgt or Lnt (PAP9403 and KA472, respectively) were grown overnight in LB supplemented with 0.2% l-arabinose. Strains were subcultured 1:100 in fresh LB broth without arabinose. Cultures were grown at 37°C for 2h and then back diluted 1:50 in fresh LB broth and grown for 2 to 4h to an A₆₀₀ of ~0.4 to 0.6. Control cultures grown in the presence of l-arabinose were grown for ~2h to an A₆₀₀ of ~0.4 to 0.6.

Treatment with Glb and Cpd2.

Cultures of an arabinose-resistant MC4100 derivative (MG3178) were grown overnight and then subcultured 1:100 in fresh LB. Sublethal amounts of Cpd2 (0.25× MIC, 2.5μg/ml), Glb (0.25× MIC, 2.5μM), or CuCl₂ (3mM) were added. Subcultures were grown at 37°C to an A₆₀₀ of ~0.5 to 0.6 (~2h). Cell samples were taken and analyzed by immunoblotting.

Whole-cell lysate preparation and immunoblotting.

Equivalent cell densities (normalized by A₆₀₀) were pelleted by centrifugation (10,000×g for 5min). For Cu-treated cells, absorbance measurements were made at A₅₀₀ to minimize interference by Cu. Cells were solubilized in Bugbuster (EMD Millipore) with added Benzonase and incubated at room temperature for 15min. Samples were diluted 1:2 with 2× Tricine-SDS sample buffer (Novex) with 4% β-mercaptoethanol (BME). Lysates were incubated at 100°C for 5min and then resolved on 16% Tricine gels (Novex) with Tricine-SDS sample buffer (Novex) at 125V for 2.5h. Resolved proteins were transferred to 0.2-μm nitrocellulose membranes which were then probed with polyclonal rabbit anti-Lpp (from the Silhavy laboratory collection of antiserum raised against denatured proteins, used at 1:800,000).

This work was supported by startup funds from Emory University School of Medicine (to M.G.) and by grant 1R35 GM118024 to Thomas J. Silhavy (Princeton University).

We thank Jim Imlay (University of Illinois) for valuable discussions and Geraldine Laloux and Jean-Francois Collet (Université Catholique de Louvain) for exchanging data. We also thank Sarah McLeod (Entasis Therapeutics) for supplying compound 2, Tim Meredith (Pennsylvania State University) for providing the Lnt depletion strain, Nienke Buddelmeijer (Institut Pasteur) for providing the Lgt depletion strain, and Tom Silhavy (Princeton University) for providing anti-Lpp antiserum.