pqiABC and yebST, Putative mce Operons of Escherichia coli, Encode Transport Pathways and Contribute to Membrane Integrity

ABSTRACT

The membranes of single-cell organisms are crucial as the first line of defense. The outer membrane of Gram-negative bacteria is an asymmetric bilayer in which lipopolysaccharides (LPSs) and phospholipids are localized in the outer and inner leaflet, respectively. This asymmetry is important for membrane integrity. In Escherichia coli, the Mla transport pathway maintains this asymmetry by removing phospholipids from the outer leaflet. The MlaD component of this system is a mammalian cell entry (MCE) domain protein, and E. coli has two other MCE domain proteins of unknown function (PqiB and YebT). Here, we show that these two proteins are components of novel transport pathways that contribute to membrane integrity. The pqiAB operon is regulated by SoxS and RpoS. The yebST operon contains pqiAB homologues. Here, we found a third member of the pqi operon, ymbA (pqiC). A PqiB-PqiC complex bridges the inner and the outer membrane, and in other bacteria, pqiBC genes are located in operons together with transporter proteins. We show here that simultaneous deletion of pqiABC and yebST operons in an Δmla background rendered cells more sensitive to SDS-EDTA, and the SDS-EDTA sensitivity of mla mutants was rescued by additional copies of pqiABC. We also found that the yebST operon was induced by a defect in LPS molecules. In conclusion, PqiABC and YebST are novel transport pathways related to the Mla transport pathway and important for membrane integrity.

IMPORTANCE Membranes of bacteria are crucial for stress resistance. The composition of the E. coli outer membrane is asymmetric, with asymmetry maintained by the Mla ABC transport pathway. We propose that the stress-inducible pqiABC operon and homologous yebST operon, both of previously unknown function, encode transport pathway proteins related to the Mla transport pathway. Deletion of these operons rendered cells more sensitive to membrane stress, and additional copies of pqiABC suppressed the SDS-EDTA sensitivity of mla mutant strains. We found that yebS′-′lacZ fusion was activated in mutant strains with defective LPS molecules.

INTRODUCTION

Escherichia coli, a Gram-negative bacterium, has an inner membrane (IM) and an outer membrane (OM). Since these membranes separate essential cellular components from the environment, they are important for stress resistance (1, 2). The OM of E. coli is an asymmetric bilayer in which lipopolysaccharides (LPSs) and phospholipids (PLs) are localized in the outer and the inner leaflet, respectively (3). Lipophilic molecules easily penetrate a bilayer of PLs. Because of the low fluidity of lipid A (the lipid portion of LPSs), the asymmetric OM performs a barrier function against lipophilic molecules (3). Mutant strains with defective LPSs are also sensitive to hydrophilic molecules, presumably due to transient cracks in the OM (3). Therefore, disruption of the asymmetric composition reduces the effectiveness of the barrier and renders cells sensitive to various stressful conditions (4). In unstressed cells, the Mla ABC transport pathway maintains the OM asymmetry by removing PLs from the outer leaflet. Under stressed conditions, two other factors, PldA phospholipase and PagP palmitoyl transferase, also remove PLs from the OM (4).

MlaD, a periplasmic component of the Mla pathway (4), has a mammalian cell entry (MCE) domain. mce genes were originally found in Mycobacterium tuberculosis, and one of them (mce1) enables a nonpathogenic E. coli strain to invade macrophages and HeLa cells (5). Other proteins with MCE domains are found in several transport pathways (MCE proteins are listed at http://www.ebi.ac.uk/interpro/entry/IPR003399). The actual function(s) of these domains is not known, but it was shown that InvX (a 72-amino-acid fragment containing the MCE domain of Mce1) is sufficient for extensive association with HeLa cells (6). M. tuberculosis has four mce operons, and mutations in these operons attenuate the virulence of M. tuberculosis (7).

According to the InterPro Database, E. coli has two other MCE proteins (PqiB and YebT). The pqiB gene is a part of the pqiAB (paraquat-inducible) operon, which is induced by paraquat in a SoxS-dependent manner (8, 9). Paraquat (methyl viologen [MV]) is an agricultural chemical that generates superoxide radical (10). Starvation also induces the pqiAB operon via RpoS (8). The yebT gene is a homologue of pqiB, but they are quite different in length (2,634 bp and 1,641 bp, respectively). yebT forms an operon with the yebS gene, which is highly homologous to pqiA. No functions of these genes or phenotypes of mutant strains have been reported yet. Because the pqiAB operon is conserved in many pathogenic bacteria (including Vibrio cholera, Yersinia enterocolitica, and Pseudomonas aeruginosa) (11) and mce operons encode virulence factors in M. tuberculosis, it is important to identify their functions.

In this study, we obtained several pieces of evidence showing that the pqiABC and yebST operons encode transport pathway proteins that are functionally related to those in the Mla pathway. The localization and structure of these components are suitable for functioning in transport pathways, and the genomic organization of their homologous genes also supports our hypothesis that that these two operons encode transport pathway proteins functionally related to those in the Mla pathway. Deletion of these operons reduced membrane integrity, and additional copies of pqi genes rescued mla mutant strains.

RESULTS

ymbA (pqiC), a DUF330 family lipoprotein gene, is a member of the pqiAB operon.

The pqiAB operon consists of two membrane protein coding genes, which are induced by SoxS under oxidative conditions (8). Examination of the DNA sequence showed that the stop codon of pqiB overlaps with the start codon of the downstream gene (ymbA) (Fig. 1A). Therefore, it was predicted that pqiB and ymbA are translationally coupled and form an operon (12). To test this possibility, we constructed reporter plasmids and examined their MV responsiveness. YmbA has a putative lipobox, an export signal peptide for OM lipoprotein (13). We fused the ymbA gene with phoA, an E. coli alkaline phosphatase (AP) gene. PhoA becomes enzymatically active only when exported to the periplasm where the required intrachain disulfide bond can be formed (14). The resultant pPAF-pqiAB-ymbA plasmid contains the sequence from 420 bp upstream of the pqiA gene to the last 5 bp of the ymbA gene (Fig. 1A). AP assays were carried out as described in reference 14. High AP activities were detected from cells expressing this fusion protein, indicating that the putative lipobox acts as an export signal. Furthermore, the expression of the YmbA-PhoA fusion protein was induced by MV, and this response was eliminated by soxS gene disruption (Fig. 1B). In addition, the response to MV was also eliminated by deletion of the SoxS-binding site in the pqiA promoter region (−420 to −305, upstream of the BglII site) (Fig. 1) (9). These results indicate that ymbA is a member of the pqiAB operon. In this study, we hereafter call this gene pqiC.

FIG 1.

Analysis of ymbA (pqiC)′-′phoA reporter fusions. (A) A schematic diagram of the reporter construct. The ymbA gene was fused in frame with the phoA gene without its export signal. The upstream 420 bp of the pqiA gene was also included. In the region upstream of the BglII site, there is a SoxS-binding site. The N-terminal amino acid sequence of PqiC is shown; the putative lipobox is represented in bold. (B) Assays for the translational ymbA′-′phoA fusions. Exponentially growing E. coli cells harboring the ymbA′-′phoA fusion were treated with 50 μM MV for 1 h. The cells were then washed once, and whole-cell extracts were prepared and used for AP activity measurement. Full and soxS::Km^r, BW25113 and JW4023 (soxS::Km^r) harboring the reporter described in panel A; BglII, BW25113 harboring a reporter lacking the SoxS-binding site in the pqiA promoter.

pqiB and pqiC homologues comprise operons with transporter protein genes.

By analyzing homologous genes in other bacteria, we found that in Legionella pneumophila, a human-pathogenic bacterium, homologues of the ABC transporter genes mlaFE constitute an operon with the homologues of pqiB and pqiC genes (Fig. 2) (these data are available in the GenBank database [https://www.ncbi.nlm.nih.gov/gene/19833613]). This fact suggests that PqiB and PqiC may deliver a substrate(s) to a transporter protein. Indeed, like MlaD protein, a substrate-binding protein of the Mla pathway, PqiB protein also has MCE domains. MCE domains were found in substrate-binding proteins of lipid transporter pathways (http://www.ebi.ac.uk/interpro/entry/IPR003399). However, the amino acid sequence of PqiA has no similarity with that of known transporter proteins.

FIG 2.

Schematic diagrams of pqiABC and related loci. (A) Genomic context. Homologous genes are shown using the same graphic pattern. In E. coli mlaFEDCB, pqiABC, and yebST operons, there is one MCE protein gene in each operon (arrow with vertical lines). MlaF and MlaE are components of an ABC transporter, and MlaD and MlaC deliver transported substrates. In L. pneumophila, pqiBC homologues are located downstream of mlaFE-like genes. Unlike in E. coli, a permease component (mlaE) is located in front of an ABC-binding component (mlaF). (B) Localization and functions of product proteins. Mislocalized PLs are removed by the OmpC-MlaA complex and delivered by other components as represented by arrows. Because OmpC and MlaA are expressed from other loci, they are not included in panel A. Predicted localizations of PqiABC and YebST are also shown in this figure. Details are discussed in later sections.

PqiA is an inner membrane protein with six transmembrane segments.

As mentioned above, we did not detect sequence homology between PqiA and known transporter proteins. Therefore, we investigated whether PqiA has any structural similarity with transporter proteins. In the BioCyc Database (http://biocyc.org/), PqiA was described as a protein with eight transmembrane (TM) segments, each with a length of about 20 amino acids. We also detected corresponding hydrophobic segments using several hydropathy analyses (Fig. 3A). However, there is no hydrophilic region between TM segments 5 and 6. Therefore, we considered it likely that these segments together comprise a single long TM segment.

FIG 3.

Topological analysis of PqiA protein. (A) Kyte-Doolittle hydropathy plot. The window number is 19. Predicted TM segments are indicated by arrows. Because there is no hydrophilic region between segments 5 and 6, this region is represented as a single TM segment. (B) Topological model predicted from the positive inside rule and the hydropathy plot. There is clustering of basic residues in hydrophilic regions 1, 3, 4, 6, and 8, which suggests cytoplasmic localization of these segments. The fusion points and hydrophobic segments are indicated by gray ovals and cylinders, respectively. (C) Schematic diagram of the PhoA/LacZ fusion system. Both AP and LacZ fusions were constructed for each fusion point. AP is active when it is localized in the periplasm, and BG is active when it is localized in the cytoplasm. (D) Relative activities of pqiA′-′lacZ and pqiA′-′phoA fusions. These fusion genes were placed under the control of an MV-inducible pqiA promoter. Exponentially growing E. coli cells harboring pqiA′-′lacZ and pqiA′-′phoA fusion-expressing vectors were treated with 50 μM MV for 1 h. Relative activities were calculated as described in Materials and Methods. Positive and negative values indicate periplasmic and cytoplasmic localization, respectively. Experiments were conducted at least three times, and representative data are shown.

It is known that in integral membrane proteins, positively charged amino acids (Lys and Arg) are abundant in the cytoplasmic portion, compared with the periplasmic portion (15). Based on this “positive inside rule,” the clustering of basic residues in hydrophilic regions 1, 3, 4, 6, and 8 suggests that these segments are localized in the cytoplasm (Fig. 3B). Most ABC transporters are known to comprise six TM segments (16).

To test our hypothesis that hydrophilic regions 1, 3, 4, 6, and 8 are localized in the cytoplasm, we constructed fusion protein-expressing vectors in which the end of the C terminus of each hydrophilic region was fused to β-galactosidase (BG [LacZ]) or alkaline phosphatase (AP [PhoA]) without its signal sequence. In contrast to AP, BG is enzymatically active only in the cytoplasm. Therefore, by comparing the activities of the two proteins produced by fusion with that of the test peptide at the same point, we can determine the tested region's localization (Fig. 3C). The results revealed higher AP activities at three points (101, 284, 371) and higher BG activities at the other five points. This is consistent with the six-transmembrane model. As predicted, hydrophobic segment 3 does not cross the IM, and the hydrophobic segment 5/6 is a single long TM segment (Fig. 3B and D).

Structural similarity between PqiA and YebS.

Based on the positive inside rule, the membrane topology of YebS protein was predicted to be similar to that of PqiA protein (Fig. 4A). To test our hypothesis that YebS and PqiA are both six-transmembrane proteins with similar topology, we constructed YebS-PhoA-LacZα dual reporters (17). In this construct, the N-terminal segment of LacZ (LacZα fragment) was fused to the C-terminal segment of PhoA. Only when the fusion segment is cytoplasmic is it accessible to the C-terminal segment of LacZ (LacZω fragment) expressed from the isopropyl-β-D-thiogalactopyranoside (IPTG)-inducible ϕ80′lacZΔM15 allele of the host strain. The junction points were chosen based on the alignment of the PqiA and YebS amino acid sequences. Because there were no known conditions for inducing the yebS promoter at the time this experiment was performed, these fusions were placed under the control of the lac promoter.

FIG 4.

Topological analysis of YebS protein. (A) Topological model predicted from the positive inside rule. The fusion points and hydrophobic segments are indicated by gray ovals and cylinders, respectively. (B) Relative activities of yebS′-′phoA′-lacZα fusions. These fusion genes and the ϕ80′lacZΔM15 allele (LacZω fragment) were both IPTG inducible. Exponentially growing E. coli cells harboring yebS′-′phoA′-lacZα fusion-expressing vectors were treated with 0.25 mM IPTG for 1 h. Relative activities were calculated as described in Materials and Methods. Positive and negative values indicate periplasmic and cytoplasmic localization, respectively. 382SF, PhoA-LacZα sandwich fusion at 382Q. Experiments were conducted at least twice, and representative data are shown.

Other than the fusion at 382Q (corresponding to 371M of PqiA), most fusions exhibited activities similar to those of PqiA fusions (Fig. 4B). However, the fusion at 382Q showed high-level BG and medium-level AP activities (Fig. 4B). This result is not consistent with the predicted model or the periplasmic localization of PqiA371M. There are some examples in which the C-terminal segment is known to be necessary for proper localization of the N-terminal segment (18). Therefore, instead of the C-terminally truncated fragment, we inserted the PhoA-LacZα fragment between 382Q and 383I of YebS (producing a sandwich fusion). This fusion protein exhibited much lower BG activity than that of the C-terminally truncated fusion (Fig. 4B). This result suggests that the periplasmic localization of 382Q was unstable without the C-terminal segment. Collectively, these results indicate that YebS and PqiA are both six-transmembrane proteins with similar topology.

Large portion of PqiB was localized in the periplasm.

If PqiB and PqiC proteins deliver substrates to a transporter, they should be localized in the periplasmic space. PqiB was predicted to be inner membrane localization, with a single TM segment (Fig. 5A). To determine the localization of PqiB, a C-terminal PhoA-LacZα dual fusion was constructed (17). As shown in Fig. 5B, the PqiB-PhoA-LacZα fusion exhibited high AP and low BG activities, indicating that a large portion of PqiB, including the MCE domains, is exported to the periplasm (Fig. 5E).

FIG 5.

Analysis of localization and interactions of PqiB, YebT, and PqiC proteins. (A) Domain composition of PqiB and YebT proteins. TM, transmembrane domains; MCE, MCE domains; aa, amino acids. (B) Absolute activities of pqiB′-′phoA′-lacZα and yebT′-′phoA′-lacZα fusions. These fusion genes and the ϕ80′lacZΔM15 allele (LacZω fragment) were all IPTG inducible. Exponentially growing E. coli cells harboring fusions were treated with 0.25 mM IPTG for 1 h. (C) Protease accessibility assay of PqiC protein. C-terminally HA-tagged PqiC protein (PqiC-HA) was expressed from the pTrc99A vector. Exponentially growing E. coli cells were treated with 1 mM IPTG for 1 h and then washed and resuspended with Tris-sucrose buffer (pH 8.0). Protease treatments were conducted at 25°C for 30 min. Lanes: 1, 0 min; 2, distilled water; 3, 50 μg/ml proteinase K; 4, 5 mM EDTA; 5, 5 mM EDTA plus 50 μg/ml proteinase K. (D) Yeast two-hybrid assay. The host yeast strain HF7c is a histidine auxotroph, and interaction between AD and BD fusions leads to expression of the histidine synthesis gene. Transformants were spotted on media with or without histidine (+his and −his, respectively). (E) Predicted localization of PqiABC proteins. Experiments were conducted at least twice, and representative data are shown.

yebT is a homologue of pqiB, and its product also has a single TM segment and multiple MCE domains (Fig. 5A). We constructed a YebT-PhoA-LacZα fusion and confirmed its periplasmic localization (Fig. 5B).

PqiC is a periplasmic protein anchored with the OM.

PqiC protein has a signal peptide known as a lipobox (Fig. 1A). At first, lipoprotein precursors are transported to the periplasmic side of the IM. If the second residue is Asp, the lipoprotein remains in the IM. Otherwise, the lipoprotein is transported to the OM (13). Since the second residue is not Asp (Fig. 1A), PqiC is predicted to be localized in the outer membrane. Although all known lipoproteins of E. coli K-12 face the periplasm, surface-exposed lipoproteins have been discovered in some other Gram-negative bacteria (19). Therefore, we conducted a protease accessibility assay to test its periplasmic localization. E. coli cells expressing PqiC protein with a C-terminal hemagglutinin (HA) tag were treated with proteinase K in Tris-sucrose buffer. Whole-cell extracts were analyzed by Western blotting. When cells were intact, PqiC-HA protein was not digested by proteinase K (Fig. 5C, lanes 2 and 3). It is known that EDTA causes permeabilization of the OM, thus allowing access of proteinases to the periplasm (20). In the presence of 5 mM EDTA, PqiC-HA protein was digested by proteinase K (Fig. 5C, lanes 4 and 5). Thus, PqiC-HA protein was not proteinase K resistant or aggregated in the cytoplasm but rather was localized in the periplasm and inaccessible to proteinase K in intact cells (Fig. 5E).

PqiB and PqiC bridges the inner and outer membranes.

Considering their localizations, PqiB and PqiC might interact to bridge the IM and the OM (Fig. 5E). To test this hypothesis, we conducted a yeast two-hybrid assay to investigate whether PqiB and PqiC interact. The host yeast strain HF7c is a histidine auxotroph, and interaction between AD and BD fusions leads to expression of a histidine synthesis gene (HIS3). When histidine was present in the medium, due to the resultant toxicity, GAL4AD-PqiB- or GAL4AD-PqiC-expressing cells grew poorly compared with GAL4AD-expressing cells (Fig. 5D). However, the growth of fusion protein-expressing cells was not affected by withdrawal of histidine from the medium. These results indicate that PqiB and PqiC proteins form not only a hetero-oligomer(s) but also homo-oligomers. This bridge formation is consistent with the substrate delivery hypothesis.

PqiABC and YebST are important for membrane integrity.

Although the pqiABC operon is induced by oxidative stress and under starvation conditions, mutant strains are not sensitive to these stresses (8). E. coli has a homologous yebST operon; however, even simultaneous deletion of the pqiABC and yebST operons did not affect its sensitivity to these stresses (data not shown).

As discussed above, in Legionella pneumophila, mlaFE genes constitute an operon together with pqiBC genes. Therefore, we predicted that pqiABC is functionally related to the Mla transport pathway. In the presence of EDTA, the OM is destabilized due to repulsion of negatively charged LPS molecules. In such a condition, misplaced PLs are deleterious, and mutant strains of mla genes exhibit severe SDS-EDTA sensitivity (4).

Although sole disruption of the Mla pathway renders cells hypersensitive to SDS-EDTA, we found that only simultaneous deletion of both the pqiABC and yebST operons in the Δmla background affected SDS-EDTA sensitivity (Fig. 6A). This hypersensitivity was complemented by the pqiBC-expressing plasmid (Fig. 6A). However, due to its toxicity, the yebT-expressing plasmid did not complement the hypersensitivity (data not shown).

FIG 6.

Assays for SDS-EDTA sensitivity. Unless otherwise stated, E. coli cultures in stationary phase were spotted onto LB plates containing 0.1% SDS and EDTA. The plates were then incubated at 37°C for 16 h. Expression from plasmid vectors was not induced by additional agents. (A) ΔmlaD and ΔmlaD pqiB yebT strains. These cells harbored pTrc99A-derived vectors. Undiluted cultures were spotted. (B) ΔmlaD ompC (ΔDC) and ΔmlaD pqiB yebT ompC (ΔDBTC) strains. Undiluted cultures were spotted. (C) PqiA- or PqiABC-expressing cells. ΔmlaD cells harbored pBR322-derived vectors. pqiA or pqiABC genes were under the control of the native pqiA promoter. Tenfold-diluted cultures were spotted. (D) PqiBC-, PqiABC- or YebST-expressing cells. ΔmlaD cells harbored pTrc99A-derived vectors. (E) ΔwaaG cells harbored pTrc99A-derived vectors. Undiluted cultures were spotted onto LB plates containing 0.25% SDS. (F) ΔompC cells harbored pTrc99A-derived vectors. Experiments were conducted at least twice, and representative data are shown.

It was reported that the Mla pathway depends on OmpC porin for its function (21). However, even in the ΔmlaD ompC background, deletion of both the pqiABC and yebST operons rendered cells sensitive to SDS-EDTA (Fig. 6B).

Next, we performed several complementation experiments. In these experiments, expression from plasmid vectors was not induced by additional agents, but genomic copies of these genes were intact. Additional copies of pqiABC but not of yebST partially rescued the ΔmlaD mutant strain (Fig. 6C and D). However, the pqiABC-expressing vector did not rescue the SDS sensitivity of the ΔwaaG strain (Fig. 6E). Moreover, the pqiABC-expressing vector rendered the ΔompC strain more sensitive to SDS-EDTA (Fig. 6F). Additional copies of PqiA alone or PqiBC did not rescue the ΔmlaD mutant strain (Fig. 6C and D).

yebST operon was induced by a defect in lipopolysaccharide molecules.

Consistent with the toxicity of YebT, expression of yebS′-′lacZ fusion was very weak under the nonstressed condition (Fig. 7C). However, the fact that deletion of the yebT gene affected SDS-EDTA sensitivity indicated that the yebST genes are not pseudogenes. To search for a factor(s) which controls the expression of the yebST operon, we constructed a mini-Tn10::Tet^r insertion library. BW25113 cells harboring pMC1403-yebS were infected with λ1098 (22) and incubated on Luria-Bertani (LB)–ampicillin (Amp)–tetracycline (Tet)–5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) plates. We thereby obtained several blue colonies, and among them we found three independent insertions in the waa operon (Fig. 7A). It has been reported that waaG and waaP mutants have truncated LPS molecules and are highly sensitive to SDS (23). We measured BG activity of a waaG deletion mutant and confirmed that expression of yebS′-′lacZ fusion was higher than that in the wild-type strain (Fig. 7C). We also found a C-terminal deletion of the yejM gene, which causes a defect in lipid A (a lipid component of LPSs) synthesis (Fig. 7B) (24). LPS molecules are stabilized by bivalent cations, and we found that yebS′-′lacZ fusion was also induced by EDTA (Fig. 7D and E). Most of the other blue colonies were sensitive to SDS, indicating that these clones also had defects in their membranes. pqiA′-′lacZ fusion was not induced by membrane stress conditions (data not shown).

FIG 7.

Screening for yebST-inducing conditions. E. coli cells harboring pMC1403-yebS were infected with a mini-Tn10-containing λ phage (λ1098). Blue colonies were isolated, and λ1098 insertion sites were determined. (A) Approximate positions of mini-Tn10-inserted sites in the waa operon (counted from the start codon of waaQ). Because of the polar effect, the expression of downstream genes was also affected. The downstream part of this operon has a SoxS/MarA-inducible promoter. (B) Approximate positions of mini-Tn10-inserted sites in the yejM gene (counted from the start codon of the yejM gene). (C) Stationary-phase cultures of E. coli cells harboring pMC1403-yebS were diluted 100-fold with flesh medium and incubated at 37°C for 7 h. Experiments were conducted twice. (D) BW25113 cells harboring pMC1403-yebS were grown to stationary phase. Then 4 μl of this culture was spotted onto LB–Amp–X-Gal plates with or without 1 mM EDTA. The plates were then incubated at 37°C for 16 h. (E) BW25113 cells harboring pMC1403-yebS were treated with 1 mM EDTA for 75 min. Experiments were conducted twice.

DISCUSSION

SoxS is a transcriptional regulator of the AraC/XylS family and induces the expression of a large number of regulon genes that play critical roles in protecting cells from oxidative stress (25). SoxS-activated genes include those for a superoxide scavenging enzyme (sodA), a DNA repair enzyme (nfo), a pyruvate:flavodoxin oxidoreductase (ydbK), a transcriptional repressor of iron uptake (fur), oxidative stress-resistant isozymes of the tricarboxylic acid (TCA) cycle (fumC and acnA), and an efflux pump (acrAB) (26–28).

pqiAB were discovered as SoxS-regulated genes about 2 decades ago, but the function(s) of these genes is still unknown (8). In this study, we obtained several pieces of evidence showing that the pqiABC operon and its homologous yebST operon are transporter pathways related to the mla operon. The six-transmembrane topology of PqiA (Fig. 3B) and the bridge formation of the PqiB-PqiC complex (Fig. 2B and 5E) are consistent with this hypothesis.

Additional copies of PqiACB complemented the ΔmlaD strain, but additional copies of PqiA alone or PqiB-PqiC did not (Fig. 6C and D). This result indicates that PqiA and PqiB-PqiC work together. Because PqiB and PqiC constitute an operon together with transporter proteins in L. pneumophila, we consider it likely that PqiA is a permease component. While sole deletion of the Mla pathway renders cells sensitive to SDS-EDTA, only simultaneous deletion of the pqiABC and yebST operons in a Δmla background affected SDS-EDTA sensitivity (Fig. 6A). This result indicates that these two pathways transport PLs ineffectively or transport different substances that have minor effects on membrane integrity.

Unlike overexpression of pqiB, overexpression of yebT was highly toxic, and additional copies of yebST did not suppress the SDS-EDTA sensitivity of the ΔmlaD strain. The most obvious difference is that there is no pqiC homologue (DUF330) in the yebST operon. Instead, the yebT gene is much larger than the pqiB gene (2,634 bp and 1,641 bp, respectively). Interestingly, Bradyrhizobium elkanii USDA 76 possesses a protein with both DUF330 and MCE domains (BRAEL_RS0100830). (These data are available in the GenBank database [https://www.ncbi.nlm.nih.gov/gene/23084765].) YebT may be derived from such a fusion protein and be able to bridge the two membranes by itself. Considering the fact that the Mla pathway delivers PLs and modifies membrane composition, YebT may also modify membrane composition and thus kill cells when overexpressed. Interestingly, Sutterlin et al. reported the existence of a toxic variant of the mlaA gene (mlaA*) (29). Those authors considered that this toxicity was caused by traffic of lipids from the IM to the OM (the reverse orientation compared with wild-type MlaA). Such retrograde traffic deprives the IM of phospholipids. It is possible that the toxicity of overexpressed YebT is also due to such a retrograde traffic. However, conditions that rescued the mlaA* mutant (deletion of the lpp gene and supplementation of Mg²⁺) did not suppress the toxicity of YebT (data not shown). These results indicate that the causes of toxicity are different or overexpression of YebT is too toxic to be rescued by these conditions.

Our findings that both deletion and overexpression of yebT made cells SDS-EDTA sensitive indicated that it is important to produce the proper amount of YebST. Expression of a yebS′-′lacZ fusion was very weak under the nonstressed condition; however, we found that the yebS′-′lacZ fusion was activated by several conditions that affect LPS molecules (Fig. 7). Therefore, the presence of a moderate amount of YebST seems to be beneficial for cells, especially when there are defects in LPS molecules.

Because even simultaneous deletion of these three pathways did not affect the resistance to oxidative stress conditions, the significance of SoxS-dependent expression of the pqiABC operon is not clear. SoxS also regulates LPS modification genes (waaYZ) (30), indicating the importance of the OM for oxidative stress resistance. It is possible that the growth media used in this study do not contain a transported substrate(s). The fact that the mce operons of M. tuberculosis are virulence factors suggests that it is possible that PqiABC and YebST are necessary for transporting metabolites of the host animals (7).

MATERIALS AND METHODS

Bacterial and yeast strains and plasmids.

Unless otherwise stated, E. coli cells were grown in Luria-Bertani (LB) medium at 37°C with vigorous shaking. When necessary, ampicillin (Amp) (100 μg/ml), kanamycin (Km) (30 μg/ml), tetracycline (Tet) (10 μg/ml), and/or 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) (40 μg/ml) was added. Saccharomyces cerevisiae cells were grown in SD medium (1.7 g/liter yeast nitrogen base, 5 g/liter ammonium sulfate, 20 g/liter glucose, 100 mg/liter adenine, 76 mg/liter uracil, 76 mg/liter l-lysine) or YPDA medium (10 g/liter yeast extract, 20 g/liter tryptone, 20 g/liter glucose, 100 mg/liter adenine) at 30°C with vigorous shaking. When necessary, l-histidine (76 μg/ml), l-leucine (400 μg/ml), and/or l-tryptophan (400 μg/ml) was added. Bacterial strain BW25113 and its derivatives were supplied by the National BioResource Project (NBRP) (31). Double and triple mutant strains were constructed by the standard P1 transduction procedure. pCP20, a site-specific recombinase-expressing plasmid, was used to remove the kanamycin resistance gene from recipient strains (32). Other E. coli K-12 strains and plasmids used in this study are listed in Tables 1 and 2.

TABLE 1.

Bacterial and yeast strains

Strain	Genotype	Reference or source
DH5αT1R	F− ϕ80lacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17(rK− mK+) phoA supE44 thi-1 gyrA96 relA1 tonA	Invitrogen
BW25113	F− Δ(araD-araB)567 lacZ4787(del)::rrnB-3 λ− rph-1 Δ(rhaD-rhaB)568 hsdR5	39
JW4023	BW25113 soxS::Kmr	31
JW0934	BW25113 pqiB::Kmr	31
JW18222	BW25113 yebS::Kmr	31
JW18223	BW25113 yebT::Kmr	31
JW3160	BW25113 mlaD::Kmr	31
JW2203	BW25113 ompC::Kmr	31
BW25113 pqiB yebT	BW25113 pqiB::Kmr yebT::FRT	This study
BW25113 mlaD pqiB	BW25113 mlaD::Kmr pqiB::FRT	This study
BW25113 mlaD yebT	BW25113 mlaD::Kmr yebT::FRT	This study
BW25113 mlaD pqiB yebT	BW25113 mlaD::Kmr pqiB::FRT yebT::FRT	This study
BW25113 mlaD ompC	BW25113 mlaD::Kmr ompC::FRT	This study
BW25113 mlaD pqiB yebT ompC	BW25113 mlaD::FRT pqiB::FRT yebT::FRT ompC::Kmr
HF7c (yeast)	MATa ura3-52 his3-200 lys2-801 ade2-101 trp1-901 leu2-3,112 gal4-542 gal80-538 LYS::GAL-HIS3 URA3::(GAL417mers)3-CYC1-lacZ	BD Clontech

TABLE 2.

Plasmids used in this study^a

Name	Feature(s)b	Reference or source	Parental plasmid	Oligonucleotide no.
pCP20	Ampr Cmr flp	40
pBR322	Ampr Tetr	41
pTrc99A	Ampr	Pharmacia LKB Biotechnology
pGADT7	Ampr LEU2; GAL4AD fusion	BD Clontech
pGBKT7	Kmr TRP2; GAL4BD fusion vector	BD Clontech
pKK223-3	Ampr	Pharmacia LKB Biotechnology
pMC1403	Ampr; translational lacZ fusion	42
pBR322Δtet	pBR322 without shorter EcoRI-EcoRV region		pBR322
pBR322-pqiA	pBR322 with PpqiA-pqiA coding sequence		pBR322	2, 13
pBR322-pqiABC	pBR322 with PpqiA-pqiABC coding sequence		pBR322	2, 36
pTrc99NK	pTrc99A without shorter NcoI-KpnI region		pTrc99A	45
pTrc99NK-pqiABC	pTrc99NK with pqiABC coding sequence		pTrc99NK	3, 36
pTrc99A-pqiBC(HA)	pTrc99A with pqiBC coding sequence		pTrc99A	25, 36
pTrc99A-yebT(HA)	pTrc99A with yebT coding sequence		pTrc99A	31, 32
pTcr99NK-yebST	pTrc99NK with yebST coding sequence		pTrc99NK	16, 32
pGADT7-pqiB	GAL4AD is fused with PqiB without its TM domain		pGADT7	27, 29
pGADT7-pqiC	GAL4AD is fused with PqiC without its lipobox		pGADT7	34, 37
pGBKT7-pqiB	GAL4BD is fused with PqiB without its TM domain		pGBKT7	27, 29
pGBKT7-pqiC	GAL4BD is fused with PqiC without its lipobox		pGBKT7	34, 37
pPAF	Translational PhoA fusion		pKK223-3	38, 39
pPAF-pqiABymbA	YmbA-PhoA fusion under the control of PpqiA		pPAF	1, 35
pPAF-pqiABymbA (BglII)	pPAF-pqiABymbA without the SoxS-binding site in PpqiA		pPAF-pqiABymbA
pPAF-PqiA	PhoA fusions with PqiA and its C-terminally truncated fragments		pPAF	1, 4–12
pMCT	pMC1403 with rrnB T1T2		pMC1403	43, 44
pMCT-PqiA	LacZ fusions with PqiA and its C-terminally truncated fragments		pMCT	1, 4–12
pPLD	Translational PhoA-LacZα fusion		pPAF	38, 40, 41, 42
pPLD-Plac-YebS	PhoA-LacZα fusions with YebS and its C-terminally truncated fragments under the control of Plac		pPLD	46, 47, 48, 17–24
pPLD-PqiB	PqiB-PhoA-LacZα fusion under the control of lacIq-Ptrc system		pPLD	25, 30, 49
pPLD-YebT	YebT-PhoA-LacZα fusion under the control of lacIq-Ptrc system		pPLD	31, 33, 49
pMC1403-yebS	YebS’-‘LacZ fusion under the control of PyebS (−480, +17)		pMC1403	14, 15

^aNumbers of oligonucleotides used for each plasmid correspond with those in Table S1 in the supplemental material.

^bAmp^r, ampicillin resistance; Cm^r, chloramphenicol resistance; Tet^r, tetracycline resistance; Km^r, kanamycin resistance.

Construction of reporter vectors.

The DNA fragment coding for alkaline phosphatase (nucleotide 63 to the stop codon of the phoA gene [33]) was amplified from the chromosome of E. coli BW25113 using primers that introduced SalI and HindIII restriction sites at the ends of the product. The forward primer also had recognition sites for XbaI, XhoI, KpnI, PstI, and BamHI. The phoA gene fragment was ligated into pKK223-3 digested with SalI and HindIII. The resulting plasmid was named pPAF (for phoA fusion). The DNA fragment coding for the PhoA-LacZα fusion protein was obtained by overlap extension PCR of the set of fragments (the phoA fragment without the signal sequence and stop codon and the lacZα fragment encoding amino acid residues 4 to 60 amplified from the chromosome of E. coli MG1655) (34). The phoA-lacZα gene fragment was ligated into pPAF in place of the phoA gene fragment. The resulting plasmid was named pPLD (for phoA-lacZ dual reporter). The rrnB T1T2 coding sequence was amplified using pKK223-3 as a template (35). This fragment containing additional XbaI, HindIII, and KpnI sites (included in the forward primer) was inserted into the EcoRI and BamHI sites of pMC1403 to yield pMCT (for pMC1403 with terminators).

Construction of mini-Tn10::Tet^r insertion library.

The stationary-phase cultures of BW25113 harboring pMC1403-yebS were infected with λ1098 (22) and incubated at 40°C on LB plates supplemented with Amp, Tet, and X-Gal for 16 h.

Assays for PhoA and LacZ activities.

β-Galactosidase (BG [LacZ]) activity was assayed as described in reference 36. Alkaline phosphatase (AP [PhoA]) activity was assayed as described in reference 14 with slight modification. Briefly, to prevent the activation of AP localized in the cytoplasm, cultures were transferred to 1.5-ml microcentrifuge tubes containing iodoacetamide to yield a final concentration of 1 mM, and all buffers were supplemented with 1 mM iodoacetamide. Assays were conducted at 28°C. Normalized activities were calculated as the ratio of the absolute activity of a fusion to the strongest absolute activity among the series. Then Ln (normalized AP activity/normalized BG activity) was calculated (17).

Protease accessibility assay.

IPTG (0.1 mM) was added to a growing culture of BW25113 harboring a PqiC-HA-expressing vector, and this culture was incubated for an additional 1 h. The cells were collected by centrifugation, washed, and resuspended in Tris-sucrose buffer (10 mM Tris-HCl [pH 8.0], 250 mM sucrose) at a final concentration giving an optical density at 600 nm (OD₆₀₀) of 1.5. When necessary, 5 mM EDTA and 50 μg/ml proteinase K (Nacalai) were added. After incubation for 30 min at 25°C, 5 mM phenylmethanesulfonyl fluoride (PMSF) and an equal volume of 2× PAGE sample buffer were added, and then the samples were heated at 95°C for 5 min. SDS-PAGE was carried out under reducing conditions using 10% (wt/vol) gels. Proteins were electrophoretically transferred to nitrocellulose membranes. The membranes were blocked with 5% (wt/vol) nonfat milk with phosphate-buffered saline–Tween 20 (PBS-T) (0.05% [vol/vol] Tween 20 in PBS) and then incubated in PBS-T containing antibodies (90 min at room temperature [RT] or overnight at 4°C). Monoclonal anti-HA mouse IgG antibody (Nacalai) and goat anti-mouse IgG antibody (Invitrogen) were diluted in the PBS-T at a 1:5,000 ratio. The membranes were washed with PBS-T using the following conditions: rinse three times, incubate for 15 min, and incubate for 5 min.

Yeast two-hybrid assays.

SD medium has the above-listed ingredients with the appropriate supplement(s): histidine and leucine (no tryptophan), histidine (no tryptophan or leucine), or no additional supplement (no tryptophan, leucine, or histidine). GAL4BD fusion-expressing vectors were transformed into S. cerevisiae strain HF7c using the lithium acetate-polyethylene glycol (PEG) method (37). The transformants were spread on no-tryptophan plates and incubated at 30°C for 2 days. GAL4AD fusion-expressing vectors were transformed into HF7c harboring GAL4BD fusion-expressing vectors by the same procedure. The transformants were spotted on a no-tryptophan/no-leucine plate or a no-tryptophan/no-leucine/no-histidine plate and incubated at 30°C for 3 days (38).

Assay for SDS-EDTA sensitivity.

Stationary-phase cultures of E. coli cells were serially diluted 10-fold in 0.85% NaCl solution, and 4 μl of each culture was spotted on an LB plate containing SDS and EDTA. These plates were then incubated at 37°C for 16 h.