Various Novel Colistin Resistance Mechanisms Interact To Facilitate Adaptation of Aeromonas hydrophila to Complex Colistin Environments

ABSTRACT Aeromonas hydrophila, a heterotrophic and Gram-negative bacterium, has attracted considerable attention owing to the increasing prevalence of reported infections. Colistin is a last-resort antibiotic that can treat life-threatening infections caused by multidrug-resistant Gram-negative bacteria. However, the mechanisms underlying colistin resistance in A. hydrophila remain unclear. The present study reveals four novel colistin resistance mechanisms in A. hydrophila: (i) EnvZ/OmpR upregulates the expression of the arnBCADTEF operon to mediate lipopolysaccharide (LPS) modification by 4-amino-4-deoxy-l-arabinose, (ii) EnvZ/OmpR regulates the expression of the autotransporter gene3832 to decrease outer membrane permeability in response to colistin, (iii) deletion of envZ/ompR activates PhoP/PhoQ, which functions as a substitute two-component system to mediate the addition of phosphoethanolamine to lipid A via pmrC, and (iv) the mlaFD173A mutant confers high-level colistin resistance via upregulation of the Mla pathway. The EnvZ/OmpR two-component system-mediated resistance mechanism is the leading form of colistin resistance in A. hydrophila, which enables it to rapidly generate low- to medium-level colistin resistance. As colistin concentrations in the environment continue to rise, antibiotic resistance mediated by EnvZ/OmpR becomes insufficient to ensure bacterial survival. Consequently, A. hydrophila has developed an mlaF mutation that results in high-level colistin resistance. Our findings indicate that A. hydrophila can thrive in a complex environment through various colistin resistance mechanisms.

Here, we elucidated the colistin resistance mechanisms in A. hydrophila, showing that EnvZ/OmpR is associated with resistance by regulating the arnBCADTEF operon and the autotransporter gene3832, while the mlaF mutant confers a high level of resistance. Bacteria flexibly regulate these mechanisms based on changes in the external environment to enhance survivability. A better understanding of how colistin resistance in A. hydrophila is generated lays a foundation for the development of prevention, control, and treatment strategies.

RESULTS
Serial passage of clinical colistin-susceptible A. hydrophila strain WCX23 on tryptic soy agar with increasing colistin creates highly colistin-resistant strain 23-C-23. The A. hydrophila strain WCX23, which was responsible for a fatal diarrhea outbreak in farm-raised Deinagkistrodon acutus snakes in China, is also pathogenic to mammals (34). In the current study, the colistin-susceptible (Cst s ) strain WCX23 (colistin MIC, 1 mg/liter) was serially passaged on Trypticase soy agar (TSA) with increasing concentrations of colistin sulfate to create a colistin-resistant (Cst r ) strain. MIC assays showed that 23 transfers (23-C-23, MIC = 256 mg/liter) developed high-level colistin resistance (Table 1). Additionally, the colistin MIC of 23-C-23 remained stable after 30 serial passages without colistin (23-C-23-W30) ( Table 1). After 23 passages without colistin, the control WCX23-W23 strain remained susceptible to colistin (Table 1). Resistance testing indicated that WCX23 and 23-C-23 were similarly susceptible to several other antibacterial agents (Table S1). Therefore, the only difference was that 23-C-23 was 256-fold more resistant to colistin than was WCX23.
Mutation in mlaF contributes to high-level colistin resistance by upregulating the Mla pathway. We determined and compared the genome sequences of the A. hydrophila strains WCX23 and 23-C-23 to identify genomic variations caused by colistin. The isogenic pair, WCX23 and 23-C-23, possessed essentially identical genomes, consisting of a chromosome and a plasmid. A single nucleotide polymorphism (SNP) representing one nonsynonymous mutation at 518 bp (c. A518C) was detected within mlaF. This mutation led to a replacement of the negatively charged aspartic acid at codon 173 (p. D173A) in WCX23 by the nonpolar amino acid alanine in 23-C-23. mlaF, which encodes an ABC transporter ATP-binding protein, is involved in lipid transport and is a component of the Mla pathway (phospholipid transport system) (35)(36)(37). We then investigated whether the mlaF mutation contributed to colistin resistance. The wild-type mlaF gene cloned from WCX23 was transformed into colistin-resistant 23-C-23 (23-C-23/mlaF). Colistin resistance was 16-fold lower in 23-C-23/mlaF than in colistin-resistant 23-C-23 cells (Table 1). We also constructed loci harboring nonsynonymous mlaF mutations and introduced them into the colistin-susceptible strain WCX23 (WCX23/mlaF D173A ). This resulted in a 64-fold increase in resistance to colistin on the WCX23 background (Table 1). These data suggest that mlaF D173A is responsible for colistin resistance in A. hydrophila.
To further confirm the SNP at the mlaF locus and the passage of the colistin-induced strain in which the mlaF mutation was generated, we amplified mlaF from all passaged strains using PCR, then sequenced the amplicons using the Sanger method. We found that the mlaF D173A mutation occurred during passage 13 (23-C-13; colistin MIC, 64 mg/liter) (Fig.  1A), which had high-level colistin resistance ( Table 1). The mlaF D173A mutation was maintained in 23-C-23 after 30 serial passages without colistin (23-C-23-W30; Fig. 1A), indicating its stability in a strain with high colistin resistance even in the absence of colistin.
We constructed structural models for wild-type mlaF and mlaF D173A using the i-Tasser server. The models showed that His-200 serves as the exclusive ATP-binding site residue in WCX23, whereas Arg-18 and Gln-89 were the exclusive ATP-binding site residues in 23-C-23 (Fig. 1B). Moreover, aspartic acid and alanine acid were located at position 173 of MlaF in WCX23 and 23-C-23, respectively, and the number of polar contacts with other atoms in these two amino acids was altered from four to two (Fig. 1C), suggesting that the nonsynonymous mutation had altered MlaF functionality. Moreover, reverse transcription-quantitative PCR (qRT-PCR) confirmed an increase in the expression of MlaABCDEF in the colistin-resistant mutant 23-C-23 ( Fig. 1D), implying that mlaF D173A activated the Mla pathway, which maintains outer membrane lipid asymmetry via phospholipid transport (36,38), ultimately strengthening the barrier function of the outer membrane ( Fig. 2A).
EnvZ/OmpR TCS upregulates expression of the arnBCADTEF operon to mediate lipid A modification by L-Ara4N. TCS activation is a colistin resistance mechanism that is common among bacteria (39)(40)(41)(42). Transcriptome analysis showed that EnvZ/ OmpR was the only TCS in 23-C-23 that was significantly upregulated relative to that in Cst s WCX23, and quantitative RT-PCR further confirmed that the expression of envZ/ ompR mRNA is increased during the development of colistin resistance ( Fig. 3A and B). These data imply that EnvZ/OmpR is involved in colistin resistance in A. hydrophila. We then knocked out envZ/ompR on a 23-C-23 background. The colistin MIC for 23-C-23:D envZ/ompR was 32-fold lower than that for 23-C-23 (Table 1). Thus, the deletion of envZ/ompR in 23-C-23 significantly increased susceptibility, whereas complementation of 23-C-23:DenvZ/ompR with envZ/ompR restored resistance to colistin (Table 1). Moreover, knockout and complementation with envZ/ompR did not significantly alter the susceptibility to other antibiotics (Table S1). This finding underscores the importance of the EnvZ/OmpR TCS to colistin resistance in A. hydrophila.
The arnBCADTEF operon, which is involved in lipid A modification by L-Ara4N, was expressed at higher levels in 23-C-23 than in WCX23 ( Fig. 3C to I), whereas that of pmrC, encoding lipid A PEtN transferase, did not significantly differ (Fig. 3J). These findings suggest that the arnBCADTEF operon-mediated addition of L-Ara4N to lipid A was the more critical LPS modification in A. hydrophila. To further identify whether the EnvZ/OmpR TCS contributes to colistin resistance in A. hydrophila by regulating arnBCADTEF, we compared transcriptomic differences between 23-C-23 and 23-C-23:DenvZ/ompR. Expression of the arnBCADTEF operon was obviously decreased by 5.4-to 28.8-fold in 23-C-23:DenvZ/ompR compared with that in the parent 23-C-23 strain, which was confirmed by qRT-PCR ( Fig. 3C to I).
We characterized the lipid A modifications in these strains using matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry. The WCX23 showed major peaks at mass/charge ratios (m/z) of 1,554 and 1,570, representing a basic lipid A structure and lipid A with an additional hydroxylation, respectively (Fig. S1). Lipid A from 23-C-23 contained two additional peaks at m/z 1,685 and 1,701, which corresponded to an L-Ara4N residue (Dm/z = 1131) added to the lipid A structures at m/z 1,554 and 1,570 (Fig. S1). Deletion of envZ/ompR resulted in the loss of L-Ara4N (131 mass units; Fig. S1), while complementation of envZ/ompR resulted in the addition of L-Ara4N to lipid A (Fig. S1). These results further confirmed that envZ/ompR was responsible for colistin resistance via the arnBCADTEF operon, which mediates the addition of L-Ara4N to lipid A (Fig. 2B).
EnvZ/OmpR regulates expression of autotransporter gene3832 to decrease outer membrane permeability in response to colistin. Genome annotation predicted that gene3832 was an autotransporter. Structural models of gene3832 constructed using the i-Tasser server showed that gene3832 possessed a porin-like structure and comprised b-strands arranged in b-barrels embedded in the outer membrane ( Fig. 4A and B). The increase in gene3832 expression was the third highest in the Cst r strain, 23-C-23, compared with that in the Cst s strain, WCX23 (Fig. S2), suggesting that gene3832 is associated with colistin resistance in A. hydrophila. To test this hypothesis, the gene3832 locus was deleted from the 23-C-23 background. The colistin MIC for 23-C-23:Dgene3832 was decreased 16-fold compared with that for the parent 23-C-23 strain ( Table 1). Complementation of gene3832 in 23-C-23:Dgene3832 restored high colistin resistance (Table 1). These findings confirmed that gene3832 is involved in the colistin resistance of A. hydrophila.   The outer membrane is the first line of bacterial defense against antibiotics. Outer membrane permeability was reduced in 23-C-23 compared with that in WCX23 (Fig. 4C), implying that decreased molecular exchange in the outer membrane is an additional strategy for A. hydrophila to develop colistin resistance. Deleting gene3832 caused an increase in the outer membrane permeability, whereas complementation with gene3832 reduced the outer membrane permeability of 23-C-23 (Fig. 4C), suggesting that gene3832 participates in the downregulation of outer membrane permeability.
Interestingly, gene3832 was the most downregulated gene (236-fold decreased expression) in 23-C-23:DenvZ/ompR compared with that in the parent 23-C-23 strain (Fig. S2). Quantitative RT-PCR further confirmed that the deletion of envZ/ompR decreased the expression of gene3832 mRNA (Fig. 4D). These findings suggest that EnvZ/OmpR is a regulator of gene3832 transcription. Overall, EnvZ/OmpR upregulated the expression of porin gene3832 to decrease outer membrane permeability in response to external colistin, thus reducing the colistin entry into cells, representing another colistin-resistant mechanism mediated by EnvZ/OmpR (Fig. 2C).
Deletion of EnvZ/OmpR activates PhoP/PhoQ as a substitute TCS that mediates PEtN addition to lipid A via pmrC. The expression of phoP/phoQ in A. hydrophila during the development of colistin resistance did not significantly change (Fig. 3K to L), indicating that PhoP/PhoQ is not involved in colistin resistance in A. hydrophila. Moreover, deletion of phoP/phoQ in 23-C-23 showed that susceptibility of the 23-C-23:DphoP/phoQ strain to colistin remained unchanged compared with that of 23-C-23  (Table 1). Therefore, the EnvZ/OmpR TCS functions in colistin resistance, whereas the PhoP/PhoQ TCS does not.

DISCUSSION
The A. hydrophila strains in our studies were resistant to certain antibiotics, including amoxicillin, lincomycin, streptomycin, kanamycin, sulfamethoxazole, and cefradine (Table S1). We identified putative antibiotic resistance genes carried by strains via a BLAST search against the CARD (Comprehensive Antibiotic Resistance Database, http:// arpcard.Mcmaster.ca). The strains were found to harbor multiple resistance genes, including beta-lactamase resistance genes (e.g., cphA7, aqu1, and oca12 in the chromosome and oxa17 and cmy2 in the plasmid), aminoglycoside resistance genes (e.g., aph (39)-Ia, aph(6)-Id, and aph(39')-Ib in plasmid), a sulfonamide resistance gene (e.g., sul2 in the plasmid), and macrolide resistance genes (e.g., mrx and mphA in the plasmid; Table  S2). We speculated that the presence of these resistance genes provided WCX23 and 23-C-23 with resistance to specific antibiotics.
The Mla pathway has been implicated in the maintenance of outer membrane integrity (43,44), which is a critical barrier that restricts the traffic of antibiotics into cells. Moreover, the export of de novo synthesized phospholipids to the outer membrane during cell growth might be driven by Mla (35). The Mla system consists of an MlaFEDB complex in the inner membrane, MlaA-porin in the outer membrane, and the shuttle protein MlaC in the periplasmic region (35)(36)(37). MlaF interacts with MlaB to comprise the cytoplasmic portion of the MlaFEDB complex (44), which drives phospholipid trafficking across the bacterial envelope to maintain outer membrane integrity. Meanwhile, MlaC ferries lipids between the outer and inner membrane complexes, and the MlaA-porin complex extracts lipids for import from the outer membrane. Herein, we confirmed that the mlaF D173A mutation was responsible for colistin resistance. We speculated that the mlaF D173A mutant might lead to enhanced ATP binding, that is, increased mlaF activity would serve to activate the Mla pathway, which would subsequently decrease colistin import.
We further hypothesized that various mutations might randomly occur in functional genes during the development of colistin resistance and that these mutations may either be silent or be beneficial facilitating the survival of bacteria under antibiotic pressure. Ultimately, the advantageous mutants become selected and stably inherited with increasing high-level antibiotics. We observed that only mlaF mutation occurred in the high-level colistin-resistant strain, suggesting that mlaF mutation was the consequence of evolution in response to high-level colistin.
TCSs are key players in bacterial adaptation to environmental changes. Expression of the outer membrane porins, OmpF and OmpC, is regulated by EnvZ/OmpR in some bacteria during the response to adverse environmental conditions, including changes in osmolarity and acidic pH, as well as starvation and virulence (45)(46)(47)(48)(49). Here, we confirmed that EnvZ/ OmpR is also associated with colistin resistance in A. hydrophila. Moreover, in view of the absence of ompF or ompC in A. hydrophila, we confirmed that EnvZ/OmpR regulates the expression of autotransporter gene3832 in response to colistin.
Antibiotics in clinical practice are commonly selected to treat bacterial infections according to agar dilution test results. Using this approach, A. hydrophila is classified as colistin susceptive. However, if treated with a low dose, A. hydrophila might develop relative colistin resistance via the EnvZ/OmpR TCS, eventually leading to therapeutic failure and the generation of a colistin-resistant strain. It is, therefore, not recommended that Aeromonas-associated infections be treated with colistin, which agrees with the results of previous studies (50)(51)(52).
Previous research reported that the PhoP/PhoQ TCS is associated with colistin resistance in E. coli, P. aeruginosa, Salmonella enterica serovar Typhimurium, and K. pneumoniae (14,22,23,32,39,40). Although PhoP/PhoQ is not generally involved in colistin resistance in A. hydrophila, following deletion of envZ/ompR in A. hydrophila, PhoP/PhoQ assumed the task of regulating lipid A modification, thereby leading to colistin resistance. Thus, PhoP/PhoQ may function as a supplementary mechanism of colistin resistance. Therefore, bacteria flexibly adjusted various resistance mechanisms in response to a changing environment.
In terms of lipid A modification, EnvZ/OmpR and PhoP/PhoQ appear to mediate entirely different pathways. We showed that a higher level of colistin resistance was conferred on A. hydrophila by changes in L-Ara4N than by changes in PEtN, according to the colistin MIC of 23-C-23 and 23-C-23:DenvZ/ompR (with L-Ara4N-and PEtN-modified lipid A, respectively). Therefore, colistin resistance is more effective when regulated by EnvZ/OmpR than when regulated by PhoP/PhoQ in A. hydrophila.

MATERIALS AND METHODS
Strains, plasmids, and growth conditions. The A. hydrophila strain WCX23 was isolated from a snake with fatal diarrhea in the city of Wangcheng, Hunan province in 2017 and was identified as ST516 via multilocus sequence typing (MLST) (34). All A. hydrophila isolates were maintained on TSA (Oxoid, Basingstoke, UK) or cultured in tryptone soy broth (TSB; Oxoid), supplemented with antimicrobial agents as appropriate, at 28°C. Escherichia coli strains were routinely grown in lysogeny broth (LB) or on LB agar (Oxoid), supplemented with the appropriate antibiotics as required, at 37°C. Plasmids pKD3 and pCVD442 were obtained from BioVector NTCC Inc. (Beijing, China), and pUC57-Apr, pDS132, and pUC57-sacB-Apr were donated by Y. Zhang (YiHong Biotech Ltd., Shanghai, China).
Selection of colistin resistance. We selected high-level colistin resistance in the susceptible (colistin MIC, 1 mg/liter) A. hydrophila WCX23 strain by serial daily passages on TSA with increasing concentrations of colistin sulfate. Initially, WCX23 was grown at a sub-MIC of colistin (0.5 mg/liter) at 28°C for 24 h; colonies were then transferred and streaked on the same media. Bacterial growth (including the number, morphology, and size of colonies) was observed to judge the adaptation of bacteria under drug pressure. Colistin concentrations were doubled when vigorous growth was observed. High-level resistance (colistin MIC, 256 mg/liter) that developed after 23 passages (23-C-23) was stable after 30 serial passages in TSA without antibiotics. We also passaged control WCX23 23 times without colistin. All antibiotic-induced strains were maintained separately in TSB with 30% glycerol at 280°C.
Antimicrobial susceptibility testing. The MIC of colistin and other antibiotics was determined using broth microdilution at 35 6 1°C and confirmed in three independent experiments. Antimicrobial susceptibility was defined according to CLSI instructions (53) and EUCAST clinical breakpoints (http://eucast.org).
Whole-genome sequencing and comparative genome analysis. Genomic DNA was extracted using a bacterial DNA kit (Omega Bio-Tek Inc., Norcross, GA, USA) as per the manufacturer's instructions. Strains were sequenced and assembled (Shanghai Majorbio Bio-pharm Technology Co., Ltd., Shanghai, China) as follows. Genomic DNA (5mg) was sheared using g-TUBE (Covaris Inc. Woburn, MA, USA), and then a 10-kb sequencing library was constructed using standard PacBio RS sample preparation instructions and sequenced on the Pacific Biosciences RS II platform (Pacific Biosciences Inc., Menlo Park, CA, USA). A 400-bp paired-end library was also prepared from the same genomic DNA according to Illumina TruSeq DNA sample preparation recommendations and sequenced on HiSeq 2500 platforms (Illumina Inc., San Diego, CA, USA) with a read length of 150 bp. PacBio data assembled using the Hierarchical Genome Assembly Process software generated a onecontig genome and one plasmid. The genome and plasmid were further proofread using HiSeq data and Canu and SPAdes software (54). Finally, a whole-genome assembly was obtained without redundancy. Assembled scaffolds were processed using Glimmer 3.02 and GeneMarkS software to predict genes (55). Thereafter, tRNA and rRNA genes were predicted using rRNAscan-SE v2.0 and Barrnap software, respectively. Predicted proteins were compared with a nonredundant GenBank database using BLASTP, then functionally annotated using the NR, Swiss-Prot, Pfam, EggNOG, Gene Ontology, and KEGG databases, as well as manual supplementation. Clean reads of 23-C-23 were aligned with the WCX23 strain genome using Burrows-Wheeler Aligner software (56). Reads near putative insertions or deletions (InDels) were realigned using Genome Analysis Toolkit (GATK) software to eliminate false-positive SNPs. Small InDels and SNPs detected by VarScan were further filtered to generate high-confidence data. Structural variation was detected using BreakDancer software.
Transcriptome analysis. The concentration and purity of total RNA extracted using Qiagen RNeasy Minikits (Qiagen, Hilden, Germany) were determined using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific Inc., Waltham, MA, USA). rRNA was removed, and cDNA libraries were prepared as previously described (57). The cDNA libraries were sequenced and analyzed (Shanghai Majorbio Biopharm Technology Co., Ltd.) and are described in brief below. Strand-specific cDNA libraries were generated according to the Illumina standard protocol for high-throughput sequencing and sequenced using an Illumina HiSeq. Differentially expressed genes were defined as those with a log 2 fold change (logFC) of .1 (increased expression) or a logFC of ,21 (decreased expression), and statistical significance was set at a P value of ,0.05.
Determination of mRNA expression levels using qRT-PCR. Total RNA was extracted from mid-logphase (optical density at 600 nm [OD 600 ] 0.5 6 0.05) bacterial cultures using the Qiagen RNeasy minikit (Qiagen), according to the manufacturer's instructions. Complementary DNA (cDNA) was generated from total RNA using a random primer hexamer. Table S3 shows the sequences of transcript-specific primers used for qRT-PCR. All reactions were performed in triplicate, and experimental data were normalized to 16S rRNA (GenBank accession number: MN901224) levels and analyzed using the 2 2DDCT method.
Amplification and sequencing of mlaF by PCR. Fragments of the mlaF gene from WCX23, and all passaged strains, were amplified by PCR using the respective 59-AGAGCGGCGCCCTGTTTAC-39 and 59-CCTTGCGGTTGGCGATGAT-39 primers for mlaF-F and mlaF-R. Product sequences were determined via Sanger sequencing (BIOSUNE Biotechnology Co., Ltd., Changsha, China).
Allelic replacement of the mlaF mutant. DNA fragments upstream and downstream of the mutation were amplified from 23-C-23 (WCX23) genomic DNA. sacB-Apr was amplified using pUC57-sacB-Apr as a template. Table S4 lists all primer pairs. The junction of the upstream fragment sacB-Apr and the downstream fragment was amplified via fusion PCR and cloned into pCVD442. The resulting plasmid was transformed into E. coli b2155 by electroporation and introduced into the recipient strain 23-C-23 (or WCX23) by conjugation. Transconjugants were selected on LB agar plates supplemented with apramycin. The recipient strain (23-C-23/mlaF D173A :sacB-Apr or WCX23/mlaF:sacB-Apr) was confirmed using PCR and sequencing.
The upstream fragment-wild-type mlaF-downstream fragment was amplified via PCR using WCX23 genomic DNA as a template, cloned into the suicide plasmid pDS132 harboring the chloramphenicol-resistance gene, and transformed into E. coli b2155 via electroporation to obtain the donor strain b2155/ pDS132-WCX23mlaF. We similarly constructed b2155/pDS132-23-C-23mlaF D173A from 23-C-23.
Genetic manipulations for gene knockout and complementation. Upstream and downstream DNA fragments were amplified via PCR using Aeromonas hydrophila 23-C-23 genomic DNA as a template. The chloramphenicol-resistance gene cmr was amplified by PCR using pKD3 as a template. Table S5 lists all primer pairs used for gene knockout and complementation. The junction of the upstream fragment, cmr, and the downstream fragment was amplified via fusion PCR and cloned into pCVD442, a suicide plasmid containing the ampicillin resistance gene, to obtain recombinant plasmids that were transformed into E. coli b2155 via electroporation to obtain donor plasmids that were introduced into the recipient 23-C-23 strain via conjugation. Transconjugants were selected on LB agar plates supplemented with ampicillin (100 mg/liter) and chloramphenicol (22 mg/liter). Aeromonas hydrophila 23-C-23:DenvZ/ompR, 23-C-23:D gene3832, and 23-C-23:DphoP/phoQ were confirmed by PCR and sequencing.
Upstream and downstream DNA fragments from 23-C-23 were amplified using PCR to construct the complementation strain. The apramycin resistance gene apr from the pUC57-Apr plasmid was amplified via PCR. The resulting products (upstream fragment-complementary gene-apr-downstream fragment) were connected using fusion PCR and cloned into the suicide plasmid, pCVD442. The resulting plasmid was transformed into E. coli b2155 by electroporation to obtain donor plasmids that were introduced into recipient 23-C-23:DenvZ/ompR or 23-C-23:Dgene3832 strains by conjugation. Transconjugants were selected on LB agar plates supplemented with ampicillin (100 mg/liter) and apramycin (33 mg/liter). Positive clones were then selected on LB agar plates with apramycin (33 mg/liter) and chloramphenicol (34 mg/liter), respectively. The complemented strains 23-C-23:CDenvZ/ompR or 23-C-23:CDgene3832 were screened by growth in LB agar plates with apramycin without chloramphenicol and were further confirmed via PCR and sequencing.
Mass spectrometry analysis of lipid A structure. Lipid A was isolated from strains as previously described (61) and was assessed using MALDI-TOF MS (Bruker Daltonics Inc., Billerica, MA, USA) in the negative ion mode. Data were acquired and analyzed using flexControl version 3.4 and flexAnalysis version 3.4, respectively (Bruker Daltonics).
Outer membrane permeabilization assay. Late-log-phase cells (1 ml) were centrifuged at 12,000 Â g and 4°C, and the supernatant was discarded. Sedimented cells were suspended in 5 mmol/liter HEPES (pH 7.2) containing 0.1% SDS and incubated at 28°C for 10 min. Absorbance was measured at 540 nm.
Statistical analyses. Data were statistically analyzed using GraphPad Prism version 7.0 (GraphPad Software Inc., San Diego, CA, USA). The differences were analyzed using two-tailed unpaired Student's t tests. Results are expressed as means 6 standard deviation unless otherwise mentioned. Statistical significance was set at P , 0.05.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.5 MB. SUPPLEMENTAL FILE 2, XLSX file, 0.03 MB.