Analysis of the Extracellular Proteome of Colistin-Resistant Korean Acinetobacter baumannii Strains.

We analyzed the extracellular proteome of colistin-resistant Korean Acinetobacter baumannii (KAB) strains to identify proteome profiles that can be used to characterize extensively drug-resistant KAB strains. Four colistin-resistant KAB strains with colistin resistance associated with point mutations in pmrB and pmrC genes were analyzed. Analysis of the extracellular proteome of these strains revealed the presence of 506 induced common proteins, which were hence considered as the core extracellular proteome. Class C ADC-30 and class D OXA-23 β-lactamases were abundantly induced in these strains. Porins (CarO and CarO-like porin), outer membrane proteins (OmpH and BamABDE), transport protein (AdeK), receptor (TonB), and several proteins of unknown function were among the specifically induced proteins. Based on the sequence homology analysis of proteins from the core proteome and those of other A. baumannii strains and pathogenic bacterial species as well as further in silico screening, we propose that CarO-like porin is an A. baumannii-specific protein and that two tryptic peptides that originate from CarO-like porin detected by tandem mass spectrometry are peptide makers of this protein.

Introduction

Acinetobacter baumannii is one of the most important opportunistic pathogens among ESKAPE (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, A. baumannii, Pseudomonas aeruginosa, and Enterobacter species) pathogens.[1] Currently, the majority of A. baumannii strains that cause nosocomial infections are multidrug resistant (MDR). Because of their resistance, the bacteria easily colonize the respiratory tract, skin, and wounds and cause pneumonia or sepsis.[2] Therefore, infection with MDR A. baumannii threatens the survival of an immunocompromised patient in the intensive care unit.[3] Global isolation of colistin-resistant A. baumannii strains was recently reported. Frequent outbreaks of these strains are a serious threat because colistin (polymyxin E) is considered as a drug of last resort. Furthermore, most colistin-resistant A. baumannii strains have already acquired other antibiotic resistance genes against useful antibiotics (β-lactams, aminoglycosides, fluoroquinolones, etc.) and have become extensively drug resistant (XDR).[4,5]

Consequently, -omics analyses (genomics, proteomics, and transcriptomic analyses) of MDR or XDR A. baumannii strains have been extensively performed to elucidate the resistance mechanisms of these pathogens.[6−9] Proteomic analysis of the membrane and cell wall fraction of MDR A. baumannii DU202 revealed that stress associated with high concentrations of antibiotics induces specific changes in the proteome. Specifically, imipenem exposure notably increases the levels of various resistance-nodulation-division transporters and penicillin-binding proteins (PonA) and decreases the levels of outer membrane proteins (OmpW, OmpA, and CarO) to prevent the import of antibiotics.[6] Multi-omics analysis of A. baumannii strain with induced colistin resistance (ZJ06-200P5-1) revealed that loss of lipopolysaccharide after mutation of the lpxC gene is the major cause of the colistin resistance of this strain.[7] Lipopolysaccharide loss also results in upregulation of resistance-related protein sets.[10] Genomic and proteomic analyses of colistin-resistant XDR A. baumannii KAB03 revealed that β-lactamase class D blaoxa-23, β-lactamase class C ampC-20, AdeIJK, and PmrAB are the major proteins of antibiotic resistance. Further, targeted mutagenesis demonstrated that A138 and R263 of PmrB and R109 of PmrC are important residues for colistin resistance.[4] The ultimate purpose of such studies is to characterize the MDR or XDR pathogenic bacteria on the genomic or proteomic level and to identify potential targets that might be useful for developing new antibiotic drugs or diagnostic makers.[11−13] In this context, mining of specific protein(s) or proteome sets from clinically useful XDR A. baumannii strains is important.

In the current study, we selected four colistin-resistant XDR Korean A. baumannii (KAB) strains and used proteomics to analyze their extracellular fractions. We have specifically performed proteomic analysis of secreted proteins (the extracellular protein fraction) because this approach has several advantages for target screening. The first advantage is the relatively low complexity of the extracellular fraction. Therefore, the proteome sets can be relatively easily characterized in search of useful targets. As a second advantage, “real” proteomic information is obtained for the actual secreted samples instead of an analysis involving theoretical bioinformatics tools. In fact, multiple extracellular fractions or outer membrane vesicles contain many cytosolic and membrane-related proteins, which are theoretically irrelevant but have been experimentally identified.[14−16] Finally, secretomic information can be used to predict candidate proteins encountered by the host immune system upon bacterial infection. We then used in silico analytical approaches (sequence homology analysis and reverse vaccinology) to further characterize the identified proteome sets. In conclusion, we propose that multi-omics approaches combined with secretome analysis are a good tool for mining candidate protein sets for diagnosing XDR A. baumannii strains.

Results and Discussion

Susceptibility Profiles of Colistin-Resistant KAB Strains

We collected 36 carbapenem-resistant A. baumannii (CRAB) strains isolated from patients from the intensive care units of five hospitals in five cities of Korea in the years 2015–2016. We selected four colistin-resistant A. baumannii strains from these 36 strains and designated them as A. baumannii KAB strains (KAB01, KAB03, KAB04, and KAB08). A minimal inhibitory concentration test indicated that the four strains were also resistant to β-lactams, cephalosporins, carbapenems, and fluoroquinolones and were XDR (Table S1). Therefore, the four strains were colistin-resistant CRAB. We also selected four colistin-sensitive CRAB strains (KAB02, KAB05, KAB06, and KAB07) as reference strains for a comparative analysis. All secreted CRAB strains showed variable resistance to tetracycline and aminoglycosides.

Genomic Characterization of Colistin-Resistant KAB Strains

Genome sequencing revealed that the genome size of the four colistin-resistant KAB strains ranged from 3.96 to 4.00 Mb, encoding 3919–4026 genes. All A. baumannii strains harbored plasmids, with sizes ranging from 70.8 to 121.4 kb. Each plasmid contained approximately 94–152 coding sequences. The data are summarized in Table S2. Sequence types (STs) of the colistin-resistant KAB strains were defined by using the multilocus ST Oxford method. Multilocus ST analysis categorized the four colistin-resistant KAB strains into three groups. KAB01 and KAB03 represented ST 451 (1-3-3-2-2-142-3); KAB04 represented ST 191 (1-3-3-2-2-94-3); and KAB08 represented ST 208 (1-3-3-2-2-97-3). Antibiotic resistance genes encoded by the KAB genomes were identified by using ResFinder 3.0. A detailed list of antibiotic resistance genes and point-mutations in the pmrBAC operon is presented in Table S3. In the current study, we used these colistin-resistant KAB strains as sample strains to characterize the secretome of colistin-resistant CRAB strains and to mine candidate diagnostic protein sets.

Proteomic Analysis of the Extracellular Protein Fraction of Colistin-Resistant KAB Strains

We assumed that the secreted or extracellular proteins are more useful targets for screening for diagnostic marker proteins to identify XDR strains, including colistin-resistant strains, than the soluble protein fraction. Because the colistin-resistant KAB strains are XDR strains with respect to colistin, meropenem, and amikacin, we cultured these strains in the presence of an antibiotic mixture. We analyzed the extracellular fractions of the four colistin-resistant KAB strains by one-dimensional (1D) liquid chromatography–tandem mass spectrometry (LC–MS/MS) and identified 665–768 proteins in each strain (Table S5-1). Although the protein signature depends on the bacterial strain, we identified 506 common proteins that were present in the extracellular protein fraction of the four colistin-resistant KAB strains (Figure A).

Figure 1

Proteomic analysis of the extracellular proteome of four colistin-resistant A. baumannii strains. (A) Venn diagram of the data for four colistin-resistant KAB strains (KAB01, 03, 04, and 08). (B) COGs in the core proteome (506 proteins). Each protein was classified by the definition of COG functional categories (A–V).[27]

Cluster of orthologous group (COG) analysis of 506 core extracellular proteins revealed that proteins belonging to post-translational modification, protein turnover, and chaperone (O) groups and defense mechanism (V) groups constituted a major portion of the extracellular proteins (Figure B). Representative extracellular proteins are summarized according to the biological function in Table S4. Two β-lactamases amount to nearly 30% of the extracellular proteome, which are also major proteins of defense mechanism (V) in the COG analysis. The secretion of β-lactamase blaADC-30 by KAB strains was notable, and, amount-wise, the protein comprised the highest proportion of the extracellular fraction; blaOXA-23 was another major secreted β-lactamase, but its abundance of protein was much smaller. We also detected various outer membrane proteins (OmpW, OmpA, OmpH, OprD, and a putative porin) in the extracellular fraction. Specifically, outer membrane proteins encoded by the bam operon (BamABDE) were detected in all KAB strains. One Bam protein, BamA, an outer membrane β-barrel assembly protein, is highly immunogenic. It was suggested as a potential vaccine against MDR A. baumannii because, according to a previous study, BamA immunization results in 60% protection from a lethal bacterial dose.[28,29] In addition, among the identified proteins, 95 proteins (approximately 18.7%) were identified as pathogenic proteins (Table S5-2). We also identified two porins (carbapenem susceptibility porin CarO and CarO-like porin) among the secreted proteins. Other secreted proteins of interest were lytic transglycosylases, which catalyze a nonhydrolytic cleavage of peptidoglycan in the bacterial cell wall. These proteins are involved in the synthesis, remodeling, and degradation of the cell wall. Because of their essential role in bacterial survival, transglycosylases are considered to be potential drug targets.[30] To further understand the proteomic differences between strains in terms of their antibiotic resistance, we performed a comparative analysis of colistin-resistant and colistin-sensitive strains. The comparative analysis revealed that the extracellular fractions of the colistin-resistant KAB strains (KAB01, KAB03, KAB04, and KAB08) contained much more cytoplasmic proteins than those of the colistin-sensitive strains (Figure S1). This suggested that various antibiotics (colistin, amikacin, and meropenem) might facilitate the release of A. baumannii cytoplasmic proteins into the extracellular milieu. Heat map analysis provided detailed data on the strain proteomes (Figure S1). We analyzed three nodes of the heat map (designated as C1, C2, and C3) to characterize the signatures of the two KAB strain groups. The major protein group categories in node C1 (proteins upregulated in the colistin-resistant strains) were transport groups and metabolism of amino acid and nucleotide (E and F) groups. By contrast, nodes C2 and C3 (proteins upregulated in the colistin-sensitive strains) were enriched for cell wall/membrane/envelope biogenesis (M) proteins.

Screening of Specific A. baumannii Genes by Comparative Gene Analysis, in Silico Screening, and MS/MS Analysis of Tryptic Peptides

The identified secreted proteins potentially included vaccine or diagnostic markers. However, sequences of many of these proteins were too similar to the same type of proteins from other pathogenic Gram-negative or Gram-positive bacteria, preventing their use as A. baumannii-specific marker proteins. Examples include β-lactamases blaADC-30 and blaOXA-23, which are frequently identified in K. pneumoniae strains and S. aureus, respectively. Therefore, we used comparative gene homology analysis, in silico screening, and MS/MS analysis in an attempt to select proteins specific for A. baumannii from among the core 506 proteins identified by proteomics (Figure ).

Figure 2

Selection procedure for candidate A. baumannii-specific proteins and tryptic peptides.

For the comparative analysis, we collated gene pools for 1048 strains of six pathogenic bacteria [A. baumannii (123 strains), S. aureus (380 strains), K. pneumoniae (319 strains), Neisseria meningitidis (94 strains), Streptococcus pneumoniae (71 strains), and Haemophilus influenzae (61 strains)] by downloading them from the NCBI genome database. We performed the comparative analysis using BLASTP with an e-value below 1e-20 as the cutoff. According to this criterion, we selected 132 proteins that did not show significant sequence similarity with other pathogenic bacteria as A. baumannii-specific (Figure and Table S5-2).

Figure 3

Comparative genomic analysis of A. baumannii strains and other pathogenic bacteria. Numbers in parenthesis indicate the numbers of the genome database.

We next performed in silico screening by using antigenicity and transmembrane helix indices to identify candidate proteins of high antigenicity. We subsequently selected 73 proteins (Table S5-2). Finally, we applied the criterion of distinct sequence count of over 5 in the MS/MS analysis to select putative target proteins in real samples; this yielded 29 proteins (Tables and S5-3). Because we performed the MS/MS analysis on the tryptic peptide level, in a further analysis, we focused on identifying accurate target peptides. From among tryptic peptides of the selected 29 proteins, 18 were identified with a high frequency of detection (more than 5 times) in all KAB strains. However, 16 peptides were also identified in A. baumannii ATCC 19606 or did not have sufficient ion match values in the MS/MS spectra. Finally, we selected two peptides (LVSSGSAVTTGDQSLEEAVNAEAR and NWGVFGEVGAYY TGNPTVK) of a hypothetical protein (WP_000866524.1; CarO-like porin) as the final putative target peptide (Figure and Table S5-3).

Table 1

Twenty Nine Putative Target Proteins in the Extracellular Proteome for Diagnosing Colistin-Resistant A. baumannii Strains

					in silico analysis
accession no	description	Mw	pI	cellular localization	coverage	Prot_matchesa	Prot_seq_sigb	antigenicity	TMc
WP_001035103.1	hypothetical protein (protein FilF)	70,818	6.27	extracellular	74.0	230.0	23.0	0.64	0
WP_000777882.1	membrane protein	38,427	5.32	outer membrane	80.3	980.3	16.8	0.85	1
WP_000724999.1	hypothetical protein	36,325	9.49	periplasmic	100.0	486.3	16.5	0.50	0
WP_000362917.1	hypothetical protein	47,411	6.38	outer membrane	82.8	330.8	16.3	0.79	0
WP_000277826.1	lytic transglycosylase	11,5238	9.35	outer membrane	88.5	457.3	15.0	0.65	0
WP_001093907.1	ABC transporter substrate-binding protein	24,317	9.8	periplasmic	98.3	564.3	13.8	0.55	0
WP_000768256.1	long-chain fatty acid transporter	53,090	5.15	outer membrane	77.2	215.0	13.5	0.66	0
WP_000731728.1	putative porin	27,633	4.59	outer membrane	89.5	376.5	13.3	0.82	0
WP_000696031.1	META domain-containing protein	41,182	6.75	periplasmic	84.3	148.5	12.5	0.67	0
WP_001178068.1	hypothetical protein (TolA-binding protein)	32,479	6.12	extracellular	90.7	125.0	12.3	0.69	0
WP_001037943.1	outer membrane protein OmpH	18,699	9.52	periplasmic	100.0	172.0	11.8	0.62	0
WP_000733005.1	carbapenem susceptibility porin CarO	32,095	4.77	outer membrane	66.7	427.8	10.8	0.76	0
WP_000866524.1	hypothetical protein (CarO-like porin)	26,489	4.8	outer membrane	95.3	630.8	9.3	0.56	0
WP_001259423.1	DUF541 domain-containing protein	26,156	7.85	extracellular	84.4	138.0	8.8	0.62	0
WP_000698067.1	membrane protein	25,678	4.49	outer membrane	73.4	76.5	8.5	0.76	0
WP_001009742.1	hypothetical protein	35,015	6.78	periplasmic	96.4	170.5	8.5	0.57	0
WP_000201548.1	DUF3108 domain-containing protein	26,122	9.89	extracellular	93.4	129.3	8.3	0.68	0
WP_000749178.1	copper resistance protein NlpE	17,533	4.87	extracellular	87.1	91.0	8.0	0.69	0
WP_000771345.1	hypothetical protein	14,696	9.54	periplasmic	98.9	175.0	7.3	0.62	0
WP_000720060.1	DUF333 domain-containing protein	15,709	8.83	periplasmic	99.0	113.5	7.0	0.64	0
WP_000932097.1	chorismate mutase	21,177	9.01	periplasmic	99.5	165.8	7.0	0.61	0
WP_000886870.1	hypothetical protein	55,252	5.48	outer membrane	70.5	96.8	6.5	0.55	1
WP_001037893.1	hypothetical protein	11,206	4.71	periplasmic	97.4	78.0	6.5	0.88	0
WP_001238663.1	hypothetical protein	19,127	6.58	periplasmic	97.2	151.0	6.5	0.53	0
WP_001043188.1	hypothetical protein	15,444	9.14	outer membrane	94.3	100.3	6.3	0.69	0
WP_000549819.1	hypothetical protein	18,879	9.04	periplasmic	99.3	96.3	6.0	0.68	1
WP_000480885.1	hypothetical protein	15,404	8.44	periplasmic	96.3	67.0	5.8	0.56	1
WP_001080202.1	DUF1311 domain-containing protein	13,158	9.78	periplasmic	99.6	104.5	5.5	0.54	0
WP_000983397.1	hypothetical protein	21,646	6.82	periplasmic	89.2	29.8	5.3	0.62	0

Peptide sequence match (PSM) count calculated by MASCOT 2.4.

Count of distinct sequences with significant scores, as calculated by MASCOT 2.4.

Count of transmembrane helix.

Figure 4

MS/MS spectra of a tryptic peptide (LVSSGSAVTTGDQSLEEAVNAEAR) of CarO-like porin from colistin-resistant KAB strains.

Specific Induction of CarO-Like Porin in A. baumannii and Coculture Condition with A549 Cells

To confirm CarO-like porin as a candidate diagnostic protein, we performed Western blotting using secreted fractions and cell lysates of KAB strains. As a result, the CarO-like porins were detected in secreted fractions and cell lysates of all KAB strains but not detected in other pathogenic bacteria (Figure A,B), suggesting that CarO-like porin of KAB strains has unique antigenicity. Genome analysis data revealed that A. baumannii ATCC 19606 has a homologous gene of 76% identity with CarO-like porin. This explains the weak signal in A. baumannii ATCC 19606. Western blotting of cell lysate also showed that induction of CarO-like porin in KAB strains was independent of antibiotic treatment (meropenem) (Figure B). This result was consistent with previous proteomic analysis results.[4] Finally, we performed coculture of A. baumannii KAB03 cells with A549 lung cells in order to know whether CarO-like porin was secreted or not in this culture condition.[31,32] A positive signal was confirmed after 4 h culture, and induction was increased according to the culture time (Figure C).

Figure 5

CarO-like porin secretion by A. baumannii KAB strains and other pathogens. Polyclonal anti-CarO-like porin antibody was used as the primary antibody (1:500; final concentration, 0.1 μg/μL) and anti-rabbit IgG as the secondary antibody (cell signaling, 1:1000). (A) CarO-like porins were detected in the secreted fractions of A. baumannii cultured in LB broth but not detected in other pathogenic bacteria. (B) CarO-like porins were detected in the cell lysate of cultures under both conditions of LB and LB with meropenem. (C) Positive signal was confirmed after 4 h of coculture of A. baumannii KAB03 cells with A549 lung cells.

Conclusions

In the current study, we prepared the extracellular proteomes of colistin-resistant KAB strains and identified the secreted proteins, illustrating the types of proteins released by colistin-resistant KAB strains into the extracellular milieu under antibiotic stress. Cytosolic proteins constituted over 56.9% of all proteins of the extracellular proteome. It is reasonable to expect that a cytosolic protein mixture might have contaminated the extracellular proteome because of disruption or aging of A. baumannii cells. Hence, these proteins could not be fully removed from the extracellular proteome. Nonetheless, these proteins are important because they are encountered by the host immune system upon A. baumannii infection. Therefore, we used the secreted protein list as a basic protein database for screening proteins specific to colistin-resistant A. baumannii strains, following the selection process summarized in Figure . This approach yielded 29 candidate proteins, ultimately identifying CarO-like porin as a candidate diagnostic protein marker with two specific diagnostic peptide fragments. Further, we confirmed that the candidate protein CarO-like porin was A. baumannii-specific and secreted into the culture medium when A549 cells were cocultured with A. baumannii KAB03. This suggests that CarO-like porin is a strong candidate diagnostic protein marker. Originally, we used colistin-resistant KAB strains to screen for specific markers of colistin-resistant strains. However, CarO-like porin is included in the core genes (261 genes) of KAB strains, regardless of the colistin resistance status of the strain (Figure S1D). Therefore, CarO-like porin can be considered as one of the common CRAB markers, regardless of the colistin resistance.

Experimental Procedures

Genomic DNA Preparation and Sequence Analysis of Korean CRAB Strains

Based on their multidrug–resistance profiles, eight strains were selected for whole-genome sequencing from 36 A. baumannii strains isolated from pneumonia patients at the intensive care units of five hospitals in five cities in Korea. The isolates were designated KAB01–KAB03 (from Daejeon), KAB04 (from Busan), KAB05 and KAB06 (from Daegu), KAB07 (from Jeonju), and KAB08 (from Gyeonggi). Genomic DNA of the KAB strains was extracted using an MG genomic DNA purification kit (MG MED, Inc., Seoul, Korea). The extracted DNA was used for the construction of SMRTbell libraries. Whole-genome sequencing was performed using the PacBio RS II platform (Pacific Biosciences, Menlo Park, CA). Raw sequence reads were assembled using PacBio SMRT Analysis (v2.3.0) with the default option of the Hierarchical Genome Assembly Process 3.

Gene Annotation and Bioinformatics

Prokka software and RAST server were used for genome annotation of the reconstructed KAB genomes.[17,18] STs of A. baumannii KAB strains were determined by using a previously described method.[4] Antibiotic resistant genes were identified by using ResFinder 3.0.[19] Pathogenic proteins were identified by using PathogenFinder 1.1.[20] In silico screening of putative antigen proteins was based on three predictions: (i) the number of transmembrane helices; (ii) similarity to host proteins (predictions i and ii were made by using Vaxign);[21] and (iii) antigenicity score (calculated by using VaxiJen v2.0).[22] A proteome database was constructed by using CD-HIT (sequence identity cutoff, 0.9) from 38,779 protein sequences of 10 A. baumannii strains, including KAB01–KAB08.[23] The constructed database contains 4988 homologous protein sequences. The protein database was then used for MS/MS proteomic analysis.

Antimicrobial Susceptibility Assay

The antimicrobial susceptibility of A. baumannii isolates was tested using the Vitek 2 system, as previously described.[24] The susceptibility of A. baumannii isolates was determined by the agar dilution method, in accordance with the guidelines of the Clinical Laboratory Standards Institute (CLSI document m100-s25).

Preparation of Extracellular Protein Fraction

A. baumannii KAB strains and A. baumannii ATCC 19606 were cultured under three culture conditions: (i) strain ATCC 19606 was cultured in Luria Bertani (LB) broth; (ii) strains KAB02, KAB05, KAB06, and KAB07 were cultured in LB broth supplemented with meropenem (16 μg/mL); (iii) strains KAB01, KA03, KAB04, and KAB08 were cultured in LB broth supplemented with meropenem (16 μg/mL), amikacin (16 μg/mL), and colistin (2 μg/mL). The cells were harvested at OD600 = 0.7–0.8. The cultured media were then filtered through a 0.45 mm membrane filter and precipitated in 90% ammonium sulfate solution. The supernatant was centrifuged at 10,000g for 30 min, and the pellets were suspended in 20 mM Tris-HCl (pH 8.0). The protein solution was extensively dialyzed against 10 volumes of 20 mM Tris-HCl (pH 8.0) using dialysis tubing with a molecular mass cutoff of 10,000 Da. The protein concentration was calculated by using the bicinchoninic acid method. Proteins were fractionated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) before protein identification.

In-Gel Digestion and Proteomic Analysis Using LC–MS/MS

Silver-stained gels were destained in a solution of 30 mM potassium ferricyanide and 100 mM sodium thiosulfate. The gels were then rinsed several times with distilled water to stop the destaining reaction. For Coomassie Brilliant Blue staining, the gels were destained in a destaining solution (50% acetonitrile and 10 mM ammonium bicarbonate).[25] The destained gels were rinsed with distilled water, followed by a rinse in 100% acetonitrile. After drying using a speed vacuum concentrator, the gels were incubated in a solution of 10 mM dithiothreitol and 100 mM ammonium bicarbonate at 56 °C, followed by incubation in 100 nM iodoacetamide to reduce and alkylate S–S bridges. The gels were then washed with two or three volumes of distilled water by vortex-mixing and completely dried using a speed vacuum concentrator. Tryptic digestion was performed in 50 mM ammonium bicarbonate at 37 °C for 12–16 h. For optimal digestion, the final trypsin concentration was adjusted to 10 ng/mL. Tryptic peptides were obtained by two extraction steps in 50 mM ammonium bicarbonate and 50% acetonitrile containing 5% trifluoroacetic acid. The resulting peptide extracts were pooled and lyophilized in a vacuum concentrator and stored at 4 °C. Tryptic peptides were suspended in 0.5% trifluoroacetic acid prior to further analysis. A 10 μL tryptic peptide sample was loaded onto an MGU30-C18 trapping column (LC Packings) to enrich peptides and remove chemical contaminants. Concentrated tryptic peptide solutions were eluted from the column and directed onto a 10 cm × 75 μm I.D. C18 reverse-phase column (PROXEON, Odense, Denmark) at a flow rate of 300 nL/min. Peptides were eluted by a gradient of 0–65% acetonitrile for 80 min. All MS and MS/MS spectra were acquired in a data-dependent mode using an LTQ-Velos ESI ion trap mass spectrometer (Thermo Scientific, Waltham, MA). After each full MS scan (m/z range, 400–2000), three MS/MS scans of the most abundant precursor ions in the MS spectra were performed. The MS/MS data were analyzed by using MASCOT 2.4, with a cutoff value corresponding to a false discovery rate of 1%. Protein quantities were calculated by using the exponentially modified protein abundance index and are denoted as mol %. For each sample, three technical replicates were performed, and the proteins identified at least two times in each triplet were used for further analysis. The mass spectrometry data have been deposited to the ProteomeXchange consortium via the PRIDE partner repository with the data set identifier: PXD015959.

Coculture of A549 Cells with A. baumannii KAB03 and Preparation of Secreted Proteins

A549 cells purchased from ATCC were cultured in RPMI-1640 medium (Hyclone) supplemented with 10% fetal bovine serum containing penicillin (100 U/mL) and streptomycin (100 μg/mL) at 37 °C and 5% CO2. A. baumannii KAB03 cells were cultured in LB medium containing 4 μg/mL colistin at 37 °C. A549 cells were seeded into a 100 mm culture dish at a density of 6.3 × 106/mL; approximately 4 × 106/mL of KAB03 was added to the dish. The cells in the culture dishes were incubated for 2, 4, or 6 h in RPMI-1640 serum-free medium at 37 °C and 5% CO2. A549 cells and A. baumannii KAB03 cells were separated from the culture medium by centrifugation at 5000g for 30 min. The supernatant containing the secreted proteins was saturated with 80% ammonium sulfate at 4 °C for 2 h and then precipitated at 150,000g for 1 h. Ammonium sulfate was then removed from the precipitated protein samples by dialysis (vivaspin 5000 MWCO; Sartorius, Göttingen, Germany) by using over 10 volumes of 25 mM Tris-HCl (pH 8.0).

Western Blotting

Western blotting was performed as previously described.[26] Briefly, about 30 μg of secreted proteins was separated by 12% SDS-PAGE. Polyclonal anti-CarO-like porin antibody was used as the primary antibody (1:500; final concentration, 0.1 μg/μL). After incubating with the anti-rabbit IgG secondary antibody (cell signaling, 1:1000), the blots were visualized using the Bio-Rad ChemiDoc system.