Comparative Phosphoproteomics Reveals the Role of AmpC β-lactamase Phosphorylation in the Clinical Imipenem-resistant Strain Acinetobacter baumannii SK17*

Nosocomial infectious outbreaks caused by multidrug-resistant Acinetobacter baumannii have emerged as a serious threat to human health. Phosphoproteomics of pathogenic bacteria has been used to identify the mechanisms of bacterial virulence and antimicrobial resistance. In this study, we used a shotgun strategy combined with high-accuracy mass spectrometry to analyze the phosphoproteomics of the imipenem-susceptible strain SK17-S and -resistant strain SK17-R. We identified 410 phosphosites on 248 unique phosphoproteins in SK17-S and 285 phosphosites on 211 unique phosphoproteins in SK17-R. The distributions of the Ser/Thr/Tyr/Asp/His phosphosites in SK17-S and SK17-R were 47.0%/27.6%/12.4%/8.0%/4.9% versus 41.4%/29.5%/17.5%/6.7%/4.9%, respectively. The Ser-90 phosphosite, located on the catalytic motif S88VS90K of the AmpC β-lactamase, was first identified in SK17-S. Based on site-directed mutagenesis, the nonphosphorylatable mutant S90A was found to be more resistant to imipenem, whereas the phosphorylation-simulated mutant S90D was sensitive to imipenem. Additionally, the S90A mutant protein exhibited higher β-lactamase activity and conferred greater bacterial protection against imipenem in SK17-S compared with the wild-type. In sum, our results revealed that in A. baumannii, Ser-90 phosphorylation of AmpC negatively regulates both β-lactamase activity and the ability to counteract the antibiotic effects of imipenem. These findings highlight the impact of phosphorylation-mediated regulation in antibiotic-resistant bacteria on future drug design and new therapies.

tigecycline, thus reducing the number of effective therapeutic options (8,9). In epidemiological studies, the incidence rate of carbapenem-resistant A. baumannii in countries such as Australia, Brazil, Singapore, Canada, South Korea, Taiwan, and Thailand is in the range of 47-80% (10). A study showed that 11% of nosocomial isolates of A. baumannii were carbapenem-resistant; resulting in a morbidity and mortality rate of 52% as compared with a rate of 19% of patients infected with carbapenem-sensitive isolates (4,(11)(12)(13). Among the many carbapenem derivatives, imipenem initially was highly effective in the treatment of patients with A. baumannii infections; however, imipenem resistance has been confirmed in 53.7% of Acinetobacter nosocomial infections since the early 1990s (4,14,15). The most common pathways leading to carbapenem resistance are associated with the loss of outer membrane porins, overexpression of efflux pumps, and overproduction of Ambler class B metallo-␤-lactamases, class D oxacillinases, and AmpC cephalosporinase (16 -18). In the case of Acinetobacter-derived cephalosporinase (ADC) 1 , the key upstream insertion sequence (IS) element, ISAba1, provides promoter sequences that confer bacterial resistance to broad-spectrum cephalosporins (3,19,20). In a study of Pseudomonas aeruginosa, the overproduction of AmpC ␤-lactamase exhibited weak carbapenem-hydrolyzing activity and thus contributed to carbapenem resistance in porin-deficient isolates (21). Although the study suggested a link between AmpC ␤-lactamase and carbapenem resistance, the regulatory mechanisms remain unclear.
Kinase-induced protein phosphorylation and phosphataseinduced protein dephosphorylation are crucial for signal transduction in both prokaryotic and eukaryotic species (22)(23)(24)(25)(26). Hence, bacterial phosphoproteomic analysis is a promising and accurate tool to study biological networks, including the mechanisms of antibiotic resistance. In a recent comparative phosphoproteomic study of A. baumannii ATCC17978 and the multidrug-resistant clinical isolate A. baumannii Abh12O-A2, the relationship between phosphoproteins and antibiotic resistance remained unclear because of the lack of biological confirmation (27). In this study, we used two clinical isolates of A. baumannii to establish comparative phosphoproteomic maps and to conduct biological validation to explore the mechanisms of imipenem resistance (28). Phosphoproteomic analysis of A. baumannii SK17 clinical strains was carried out using a shotgun strategy combined with phosphopeptides enrichment techniques and high-performance mass spectrometry, and thus the identified phosphosites were verified by site-directed mutagenesis (23, 29 -31). Our findings clearly show that AmpC ␤-lactamase activity is regulated by phosphorylation and is involved in imipenem resistance.

Bacterial Strains Growth Conditions and Protein Extraction-
The two clinical isolates of A. baumannii used in this study, SK17-S and SK17-R, were originally isolated from the same male patient at a hospital in southern Taiwan (28). Strain SK17-S, which was isolated first, was susceptible to imipenem, whereas strain SK17-R was resistant to imipenem (28). Accordingly, SK17-R harbors plasmid-borne ISAba1-bla OXA-82 , a bla OXA-51 -like gene with the upstream insertion sequence ISAba1 (GenBank; GQ352402.1), which is widely found in A. baumannii isolates (28,32).
Acinetobacter baumannii SK17-S was maintained on lysogeny broth (LB) agar plates (BioShop, Burlington, Canada), whereas strain SK17-R on LB agar plates containing imipenem (4 g/ml; USP, Rockville, MD). Protein extracts were prepared by inoculating a single colony of A. baumannii SK17-S into 5 ml of LB medium, and strain SK17-R into 5 ml LB medium containing imipenem (4 g/ml). Both cultures were grown at 37°C with vigorous shaking for 24 h (OD 600 nm ϭ 2.0). The cultures were then transferred at a dilution of 1:100 into flasks containing LB medium with or without imipenem (4 g/ml). After ϳ6 h (OD 600 nm ϭ 0.8), the cells were pelleted at 3000 ϫ g for 15 min at 4°C, washed twice with PBS buffer (pH 7.4), resuspended in fresh lysis buffer (25 mM ammonium bicarbonate, PhosS-TOP phosphatase inhibitor mixture tablets (Roche, Basel, Switzerland), 6 M urea, and 2 M thiourea, pH 8.0), and disrupted by sonication on ice. Cellular debris was removed by centrifugation at 12,000 ϫ g for 30 min at 4°C. The supernatant was recovered, and the protein concentrations were determined using the Bradford assay (Bio-Rad, Hercules, CA).
In-solution Protein Digestion and Phosphopeptide Enrichment-Ten milligrams of crude protein was digested using an in-solution method for phosphoproteomic analysis. The protein extracts were reduced using 10 mM dithiothreitol (Sigma, St. Louis, MO) at 37°C for 1 h and then alkylated with 55 mM iodoacetamide (Sigma) at room temperature in the dark for 1 h. The alkylated proteins were diluted by 4-fold with 25 mM ammonium bicarbonate buffer (pH 8.5) and then incubated overnight at 37°C with trypsin (Promega, Mannheim, Germany) at a dilution of 1:50 (w/w). The tryptic peptides were centrifuged for 10 min at 6000 ϫ g. The supernatants were desalted using SDB-XC StageTips with SDB-XCEmpore disc membranes (3 M Bioanalytical Technologies, St. Paul, MN) (33), eluted in buffer containing 0.1% trifluoroacetic acid (TFA)/80% acetonitrile (ACN), and then dried in a SpeedVac concentrator (Thermo Electron, Milford, MA) to remove any partial salts of the ammonium bicarbonate. The digested peptides were stored at Ϫ20°C until phosphopeptide enrichment.
Phosphopeptides were enriched using hydroxy-acid-modified metal-oxide chromatography (HAMMOC) with 0.5 mg TiO 2 beads (GL Sciences, Tokyo, Japan) packed into 10 l of C 8 -StageTips (34). The custom-made HAMMOC tips were washed with solution A (0.1% TFA, 80% ACN), after which solution B (solution A containing lactic acid (300 mg/ml)) was added as a selectivity enhancer to equilibrate the tips. Each tip contained 100 g of dry digested SK17-S or SK17-R peptides that had been redissolved in solution A and diluted with an equal volume of solution B before loading. Solutions A and B were used to wash the tips and to remove the nonspecific binding of peptides. Sequential elution to obtain pure phosphopeptides was carried out using 0.5 and 5% piperidine (WAKO, Osaka, Japan). The eluted phosphopeptides were acidified in 20% phosphoric acid (WAKO) to pH 2.5, desalted, concentrated as described above, and used in nano-scale liquid chromatography with tandem MS (nanoLC-MS/MS) analysis. 1 The abbreviations used are: ADC, Acinetobacter-derived cephalosporinase SK17-S, A. baumannii SK17-S; SK17-R, A. baumannii SK17-R; TiO 2 , titanium dioxide; TFA, trifluoroacetic acid; ACN, acetonitrile; HAMMOC, hydroxy-acid-modified metal-oxide chromatography; FA, formic acid; CD, circular dichroism; MDRAB, multidrug resistant A. baumannii; IR-MDRAB, imipenem-resistant multidrugresistant A. baumannii; ICU, intensive care units.

NanoLC-MS/MS Analysis-An online nanoLC-MS/MS LTQ-Or-
bitrap Velos (Thermo Electron, Bremen, Germany) equipped with a PicoView nanospray interface (New Objective, Woburn, MA) and coupled with a nanoAcquity system (Waters, Milford, MA) were used for sample analysis throughout this study. Enriched phosphopeptides were loaded onto a 75-m ϫ 250-mm nanoACQUITY UPLC BEH130 column packed with C18 resin (Waters). Elution was carried out at a flow rate of 300 nl/min over a linear gradient of 5-30% ACN in 0.1% formic acid (FA), followed by a sharp increase to 90% ACN and holding at 95% ACN. Instrument control was achieved using Tune 2.6.0 and Xcalibur 2.1 and the HPLC column effluent was directly electrosprayed into the mass spectrometer.
The MS scan range was m/z 300 -2000, and the multistage activation (MSA)-MS/MS top20 method was acquired using the Orbitrap analyzer. A survey scan was performed after accumulation to a target value of 5 ϫ 10 6 ions in the linear ion trap with the resolution set to 60,000 at m/z 400. Peptide ions with charge states Ն2 were selectively isolated to a target value of 5,000 and fragmented in the high pressure linear ion trap by MSA with normalized collision energy of 35%. Neutral loss masses were 97.98, 48.99, and 32.66 Da (single-, double-, and triple-charged phosphopeptides), and the ion selection threshold for MS/MS was 500 counts. The maximum allowed ion accumulation times were 500 ms for full scans and 100 ms for MSA-MS/MS measurements in the LTQ. An activation of q ϭ 0.25 and an activation time of 10 ms were used. For all experiments, standard MS conditions were as follows: spray voltage, 2.0 kV; heated capillary temperature, 200°C; no sheath and auxiliary gas flow; predictive automatic gain control enabled, and an S-lens RF level of 69%.
Data Validation and Classification-The combined database used in this study was generated from the DNA sequencing results of strains SK17-S and SK17-R. Detailed information regarding DNA sequencing and database generation are provided in the supporting information. Raw files were analyzed using the default settings of MaxQuant versions 1.4.1.2 and 1.3.0.5 (http://www.maxquant.org/) combined with the standard MaxQuant contaminants database and searched against the A. baumannii SK17 database. The search criteria applied for phosphopeptide and phosphosite identification were as follows: trypsin digestion and two missed cleavages were allowed; carbamido-methylation of cysteine (ϩ57.0214 Da) was the fixed modification; methionine oxidation (ϩ15.9949 Da), Ser/Thr/Tyr/Asp/His phosphorylation, and protein N-terminal acetylation were variable modifications; each peptide had a minimum of seven amino acids; and the mass tolerance window was 10 ppm for the parent ion and 0.6 Da for the fragment ions. The false discovery rate (FDR) of the peptides, protein groups, and modification sites was set to 1% for the MS/MS spectra automatically processed by MaxQuant for statistical validation and quantification. The minimum Maxquant score for phosphorylation sites was 25. The acceptance criteria of the localization probabilities of all Ser/Thr/Tyr/Asp/His phosphosites were applied, including only peptides with a localization probability Ն75% (calculated based on the post-translational modification (PTM) score) (35). The assembled contigs in fasta format were uploaded to the RAST server (http://rast.nmpdr.org/) to identify the protein encoding, rRNA, and tRNA genes to assign gene functions and thereby classify biological functions (36). The Uniprot and Protein Information Resource databases were used in parallel as complementary methods to further classify each identified protein.
Homology Modeling-A homology modeling approach was used to model SK17 AmpC and to investigate the importance of each conserved Ser residue (Ser-81, Ser-88, and Ser-90) identified in this study. The refined 1.2Å resolution crystal structure of ADC-1, the extended-spectrum class C ␤-lactamase Acinetobacter-derived cephalosporinase (Protein Data bank (PDB): 4net), was used to construct the active site of SK17 AmpC, based on the 99% se-quence identity between these two enzymes (37). The bioinformatics tools PDB, and Discovery Studio 4.0 (Accelrys, San Diego, CA) were used to build a homology model using default protocols. The structure file was visualized using the PyMol molecular graphics system (http://www.pymol.org).

Construction of AmpC Expression Plasmids and Site-Directed
Mutagenesis-The ampC of A. baumannii SK17 was optimized for expression in the multiple antimicrobials-susceptible strain A. baumannii 290 (Ab290) isolated from Taipei Veterans General Hospital in Taiwan in 1999 (38 -40). The conserved forward primer sequence of ISAba1, located upstream of ampC to facilitate efficient production of the enzyme, and the conserved reverse primer of ampC were used in this study (32). The PCR product was cloned between the XbaI and XhoI sites of pYMAb2 and transformed into chemically competent Escherichia coli JM109 cells (39). The primers for ISAba1-ampC wildtype (WT) and the site-directed mutants (S81A, S81D, S88A, S88D, S90A, S90D, S88A/S90A, S88A/S90D, S88D/S90A, S88D/S90D, S81A/S90A, S81A/S90D, S81D/S90A, and S81D/S90D) were synthesized based on the criteria of site-directed ligation-independent mutagenesis (SLIM) (supplemental Table S7) (41). The PCR products of randomly selected isolates were verified by sequencing (Genomics, New Taipei City, Taiwan). For the expression and purification of the WT and mutant forms of AmpC, the respective plasmids were first amplified from E. coli JM109 and then electro-transformed into Ab290 cells.
Electrotransformation of Plasmid-borne ISAba1-AmpC-The recombinant shuttle vector pYMAb2, which contains the kanamycin resistance determinant and the entire ISAba1-ampC sequence (WT or mutants), or the control vector (pYMAb2 only), was extracted from donor E. coli JM109 cells using a Plasmid Mini Kit (Viogene-BioTek, New Taipei City, Taiwan). These recombinant pYMAb2 plasmids were transformed into the kanamycin-susceptible strain Ab290 by electroporation using a gene pulser electroporator (Bio-Rad) and 2-mm electrode gap cuvettes (7,28,39). Briefly, electrocompetent cells were prepared as follows: overnight cultures of strain Ab290 were diluted 1:100 and grown to the exponential phase (OD 600 nm ϭ 0.5-0.7). The cells were collected at 4°C by centrifugation, washed with ice-cold 10% glycerol in double-distilled water (ddH 2 O) three times, and resuspended in the same buffer. Electroporation was conducted at 2.0 kV with the pulse controller at a parallel resistance of 200 ⍀ and a capacity of 25 F (42). The electroporated cell mixture was suspended in 2 ml of LB broth and incubated at 37°C for 1 h with shaking. The cells were then plated on kanamycin agar plates (25 g/ml) and incubated overnight at 37°C, and the kanamycin-resistant transformants were selected. The ampC of the recombinant plasmids was verified by colony PCR (38,43).
Purification of AmpC and Mutant Proteins-Colonies of strain Ab290 containing the distinct plasmids were grown overnight at 37°C in 100 ml of LB medium containing kanamycin (25 g/ml). Overnight cultures of the transformants were diluted 1:100 into 1000 ml of fresh medium and grown for 6 h (OD 600 nm ϭ 0.8 -1.0). The cells were collected at 4°C by centrifugation at 6000 ϫ g for 25 min at 4°C, re-suspended in cold lysis buffer (20 mM Tris-HCl pH 7.5, 500 mM NaCl, 20 mM imidazole), and then disrupted using a French-press. Cell debris was removed by centrifugation at 8000 ϫ g for 25 min at 4°C; the clear supernatant fluid was filtered through 0.45-m pore size filters (Millipore, Bedford, MA). The hexahistidine-tagged proteins from strain Ab290 carrying wild-type AmpC (AmpC-WT) or mutants (S90A and S90D) were purified from a Ni 2ϩ -NTA column (Sepharose 6 Fast Flow resin, GE Healthcare, Piscataway, NJ; Econo-Pac column, Bio-Rad), and eluted in 200 mM imidazole (44). Fractions containing the desired protein were identified by 12.5% SDS-PAGE, based on the presence of a 43 kDa band, and then concentrated using an Amicon Ultra-15 filter (Millipore).

␤-Lactamase
Assay-To correlate the AmpC phosphorylation status with ␤-lactamase activity, the purified hexahistidine-tagged AmpCs (WT, S90A, and S90D) were treated prior to the assays with alkaline phosphatase (Thermo Scientific, Pittsburgh, PA) or with reaction buffer only for 1 h at 37°C (45,46). The ␤-lactamase activities of phosphatase-untreated (control) and phosphatase-treated (dephosphorylated) recombinant AmpCs were then analyzed using nitrocefin (BioVision, Milpitas, CA) as the substrate. AmpC activity of the purified enzymes was determined spectrophotometrically at 500 nm at room temperature by measuring the rate of hydrolysis (milliOD/ min). The first 10 min of the linear enzymatic data were collected, and the hydrolytic activity in each reaction was measured at a final substrate (nitrocefin) concentration of 100 M and a final enzyme concentration of 0.1 nM (47,48). All kinetic assays were carried out in PBS buffer (pH 7.4) in a microplate reader Paradigm (Molecular Devices, Sunnyvale, CA). The kinetic values were calculated and plotted using GrahPad PRISM 5.0 (GraphPad Software, La Jolla, CA).
Here the assay was carried out as follows: overnight cultures of A. baumannii clinical strains and transformants were spread evenly on Mueller-Hinton (M-H) agar plates (BD Biosciences, Sparks, MD). The antibiotic resistance of each strain was determined using discs containing imipenem (5 or 10 g) and ceftazidime (20 or 30 g). After incubating the plates at 37°C for 24 h, the inhibition zone diameters were measured to the nearest millimeter.
Neutralization of Imipenem by Purified AmpCs-Because the strain SK17-S is imipenem-susceptible, it was used as the test strain in disc diffusion assays to determine the different imipenem-susceptibility profiles conferred by AmpC-WT, S90A, and S90D (51,52). A fixed amount (5 g) of imipenem was mixed with various amounts (4 -7.5 g) of recombinant AmpCs in a final volume of 10 l. These mixtures were incubated for 1 h at 37°C and then loaded onto the discs. Blank discs with imipenem (5 g), purified AmpC-WT protein, or mutant proteins (S90A and S90D) (7.5 g) alone served as the controls. Each disc was placed on MH agar plates containing an overnight culture of SK17-S. Following incubation of the plates at 37°C for 24 h, the differential susceptibility profile of SK17-S to imipenem was determined by measuring the inhibition zones as described previously (53).

Comparison of A. baumannii SK17-S and SK17-R Phos-
phoproteomes-Based on the whole shotgun genome sequences of SK17-S and SK17-R, the two strains are highly similar except that the strain SK17-R harbors a plasmid containing the ISAba1-bla OXA-82 gene (28,32). The high genetic similarities between SK17-S and SK17-R suggest that they are suitable for investigating imipenem resistance. Phosphoproteomic analyses of SK17 were conducted using the shotgun LC-MS/MS approach to obtain large-scale data sets from biological triplicate experiments for each strain. Using the robust HAMMOC enrichment method and high mass accuracy LTQ-Orbitrap Velos, we identified 248 unique phosphoproteins with 351 phosphopeptides from SK17-S (supplemental Table S1), as well as 211 unique phosphoproteins with 240 phosphopeptides from SK17-R (supplemental Table S2) (34,44,54). Only 70 phosphoproteins overlapped in these two strains; 178 and 141 uniquely identified phosphoproteins were found in SK17-S and SK17-R, respectively (Fig.  1 and supplemental Table S3). That there were so many more uniquely identified phosphoprotiens than overlapping suggested the basis for the differential phosphorylation-dependent regulation of imipenem resistance between SK17-S and SK17-R. Therefore, these phosphoproteins are potential candidates responsible for imipenem resistance for further validation.
According to site-specific phosphoproteomic analysis of pathogenic bacteria, we found that the numbers of phosphoproteins/phosphosites detected in A. baumannii SK17-S and SK17-R were higher than in other pathogenic bacteria (Table  I). Nevertheless, our results were comparable to those reported for Mycobacterium tuberculosis which has a similar genome size (Table I) (27,34,(55)(56)(57)(58)(59)(60)(61)(62). Accordingly, the phosphoproteins identified from SK17-S accounted for 8.1% all open reading frames encoded in the genome of A. baumannii SK17 compared with 7.5% of M. tuberculosis (56,63). In the phosphoproteomic results of four strains of A. baumannii (ATCC17978, Abh12O-A2, SK17-S, and SK17-R), we found eight overlapping phosphopeptides from the 18 corresponding phosphoproteins (supplemental Table S4). Based on our results for the imipenem-resistant strain SK17-R and the previous results for multidrug-resistant strain Abh12O-A2, with both strains being resistant to carbapenem-type antibiotics, we sought to improve the understanding of phosphorylationdependent antibiotic resistance to carbapenems in A. baumannii (27).
Classification of the Identified Phosphoproteins-To gain insight into the biological functions of the phosphoproteins identified in SK17-S and SK17-R, the phosphoproteins were classified into 18 groups based on their Gene Ontology (GO) assignments (Fig. 1B, 1C, supplemental Table S1 and S2). Approximately 80% of the identified phosphoproteins showed clear functional tendencies (Figs. 1B and 1C). Subcellular localization analysis indicated that most of the phosphoproteins could be assigned to specific locations, whereas only 23 and 17% of the phosphoproteins in SK17-S and SK17-R, respectively, were annotated as unknown (supplemental Fig.  S3). Most identified phosphoproteins were assigned to the cytoplasm, and the phosphoproteins in the two strains showed similar distributions for subcellular localization (supplemental Fig. S3).
To determine the correlation between the identified phosphoproteins and antibiotic resistance, we focused on two functional categories: transport and binding proteins (TBP), and virulence, disease and defense (VDD). Among the identified phosphoproteins from SK17-S and SK17-R are 25 (9%) and 27 (11%), respectively, that are related to TBP, including numerous factors associated with antibiotic pump efflux systems in bacterial antibiotic resistance (supplemental Table S5) (64). Similarly, phosphoproteins classified in the VDD category contributed to pathogenesis and antibiotics resistance, although the phosphoproteins were present in the same percentages (7%) in both strains (supplemental Table S6) (18). Interestingly, one of the identified phosphoproteins, a class C ␤-lactamase (470.191.peg.910; AmpC, cephalosporinase), was previously shown to be responsible for the resistance to multiple ␤-lactam antibiotics (65). Eight phosphopeptides of AmpC ␤-lactamase were identified in SK17-S but only two phosphopeptides were identified in SK17-R, indicating that the low level of AmpC phosphorylation contributed to imipenem resistance in SK17-R (supplemental Table S6). The multiple phosphorylated AmpCs in SK17-S contained three phosphosites, Ser-81, Ser-88, and Ser-90, located in the active-site phosphopeptide AVNSS 81 TIFELGS 88 VS 90 K 91 (Fig. 2).
Effects of AmpC Phosphorylation on Antibiotics Susceptibility-Overexpression of ␤-lactamases (AmpC, TEM, VEB, PER, CTX-M, SHV, OXA, IMP, and VIM) is a common mechanism in the resistance to ␤-lactam antibiotics (66,67). To verify the structure-function relationships of the identified phosphosites, a homology model of AmpC based on ADC-1 (with 99% sequence identity) was generated ( Fig. 3A and supplemental Fig. S4) (37). According to the model, Ser-88 and Ser-90 are located in the catalytic motif S 88 VS 90 K 91 of typical ␤-lactamases (domain 2), whereas Ser-81 and Thr-151 are positioned relatively close to the protein surface, based on their locations in domain 1 and the P2-loop domain, respectively (Fig. 3A) (37, 68). The hydrolytic reaction catalyzed by AmpC ␤-lactamase includes two steps, acylation and deacylation (69,70). First, Lys-91, aligned as Lys-67 in E. coli, may act as a general base and contribute to the deprotonation of Ser-88. The deprotonated Ser-88, aligned as Ser-64, attacks the ␤-lactam ring carbon of the substrate and forms a covalent acyl-enzyme complex in the acylation step (69,70). Second, catalytic water reacts with the covalent linkage in the acyl-enzyme complex leading to release of the hydrolyzed product in the deacylation step (69,70). Additionally, although residues Ser-88 and Lys-91 have been thoroughly examined, the function of the highly conserved Ser-90 remains unknown (51,71,72).
We also investigated whether phosphorylation of AmpC cephalosporinase can regulate the ability to hydrolyze ceftazidime (a third-generation cephalosporin), because ceftazidime-resistant A. baumannii has increased enormously in the past decade (50). Interestingly, most of the strains, including SK17-S, SK17-R, S81A, S81D, S90A, and S90D, were resistant to ceftazidime, indicating the clear regulatory role of Ser-90 phosphorylation of AmpC in imipenem-specific resistance (supplemental Table S8). Among the analysis of double mutants containing Ser-81 and Ser-90, only the S81D/S90D strain was sensitive to both imipenem and ceftazidime (supplemental Table S8). Single-or double-mutagenesis screening of the catalytic residue Ser-88 (S88A, S88D, S88AS90A, S88AS90D, S88DS90A, and S88DS90D) yielded both imipenem-and ceftazidime-sensitive phenotypes; thus, Ser-88 phosphorylation is likely not involved in antibiotic resistance (supplemental Table S9). Taken together, our findings indicate that Ser-90, but not Ser-81, plays a regulatory role in imipenem resistance through phosphorylation/de-phosphorylation (Table II and supplemental Table S8). Because of the location of Ser-90 in the catalytic motif S 88 VS 90 K 91 , its phosphorylation may be involved in regulating AmpC ␤-lactamase activity. Thus, we conducted kinetic analysis of ␤-lactamase  Effect of AmpC Ser-90 Phosphorylation on ␤-Lactamase Activity and the Ability to Neutralize Imipenem-To gain insight into the regulation of AmpC phosphorylation and enzyme activity, phosphatase was used to remove covalently bound phosphate groups from recombinant AmpCs (46). Kinetic analysis showed that the S90A mutant protein exhibited greater ␤-lactamase activity with an ϳtwofold higher k cat /K m value compared with AmpC-WT, suggesting that Ser-90-dephosphorylated AmpC increased enzyme activity (Fig. 3B and supplemental Table S10). Interestingly, similarly enhanced activity was observed for the phosphatase-treated AmpC-WT and S90A mutant proteins, indicating that Ser-90 phosphorylation of AmpC negatively regulated ␤-lactamase activity (Fig. 3B). This conclusion is supported by the fact that we were unable to obtain kinetic values for the mimic phosphorylated S90D mutant protein, as its ␤-lactamase activity was dramatically reduced (Fig. 3B and supplemental Table S10). Additionally, the difference in enzyme activities of the phosphatase-treated and phosphatase-untreated S90D proteins was relatively low, suggesting that the Ser-90 phosphosite is crucial for the phosphorylation network of global AmpC (Fig.  3B). Based on the circular dichroism spectra, all AmpC mutants showed similar components in their secondary structure compared with AmpC-WT (supplemental Fig. S5).
Imipenem acts as an inhibitor of AmpC ␤-lactamase, and the covalent bond between the catalytic serine residue and imipenem molecule has been determined by x-ray crystallography (51,71,72). Therefore, formation of an AmpC-imipenem covalent complex inhibits both AmpC enzyme activity and the efficacy of imipenem against bacteria. Given the proximity of Ser-90 to the imipenem-bonding site that includes Ser-88, we proposed that Ser-90 phosphorylation of AmpC may influence covalent bond formation. To better understand the efficacy of imipenem for recombinant AmpCs in vivo, neutralization assays were carried out. The results showed that the imipenem-induced clear zones could be neutralized by 5.5 g and 7.0 g of the S90A and AmpC-WT proteins, respectively ( Fig. 4 and supplemental Fig. S6). In agreement with the results of the kinetic assay, the neutralization of imipenem was achieved with a lower dose of S90A protein than of
AmpC-WT because of the higher enzyme activity of the S90A protein ( Fig. 4 and Supplemental Fig. S6). Therefore, overexpression of the S90A protein exhibited greater protection of strain SK17-S against imipenem compared with AmpC-WT. For the discs containing S90D protein, the sensitivity of strain SK17-S to imipenem was similar to that of the control (imipenem only), demonstrating that Ser-90-phosphorylated AmpC failed to show neutralizing activity against imipenem ( Fig. 4 and supplemental Fig. S6). Additionally, the predicted interactions between the four AmpCs (WT, phosphorylated-Ser-90, S90A, and S90D) and imipenem complexes from the docking experiments supported our finding that the S90D protein may be unable to bond covalently to imipenem (supplemental Figs. S7-S10). In summary, the phosphorylation/ dephosphorylation of Ser-90 of AmpC has large and specific influences on not only imipenem susceptibility in vivo but also ␤-lactamase activity in vitro. Moreover, Ser-90-dephosphorylated AmpC regulated imipenem neutralization, conferring protection to the imipenem-susceptible strain SK17-S. DISCUSSION The virulence of the Acinetobacter species is based on their adherence, colonization, and invasion of human epithelial cells, but the mechanisms of the virulent factors and antibiotic resistance remain unclear (5). Although AmpC ␤-lactamase is known to mediate the resistance to multiple ␤-lactam antibiotics, the mechanisms underlying its regulation are poorly understood (16). In this study, we used a shotgun approach to conduct comparative phosphoproteomic analysis of imipenem-susceptible and imipenem-resistant strains of A. baumannii SK17, as well as site-directed mutagenesis to determine whether AmpC phosphorylation is involved in regulating ␤-lactamase activity and imipenem resistance.
Phosphoproteomic Comparison of A. baumannii Strains-Because of the biased nature of data-dependent acquisition (DDA) used in conventional shotgun proteomic studies; lowabundance peptides are unlikely to be sequenced in a complex mixture of peptides. Therefore, our findings may represent only high-abundance phosphoproteins because of the highly dynamic process of phosphorylation in living cells and the settings used in the LC-MS/MS method in DDA mode. Western blot analysis of lysates from SK17-S and SK17-R revealed high levels of phosphorylation on Ser/Thr/Tyr residues (Supplemental Fig. S11B, S11C, and S11D). Interestingly, Ser and Tyr showed different phosphorylation levels in SK17-S and SK17-R (supplemental Fig. S11A, S11B, and S11D). A total of 178 and 141 phosphoproteins were uniquely identified in SK17-S and SK17-R, respectively, but only 70 phosphoproteins overlapped between the two strains ( Fig. 1A and supplemental Table S3). The observation of a higher number of unique phosphoproteins between SK17-S and SK17-R compared with the number of overlapping phosphoproteins is consistent with the different phosphorylation patterns observed in Western blot analysis. Although our results FIG. 4. Neutralization of imipenem by recombinant AmpCs. The relative susceptibility of A. baumannii SK17-S to imipenem in the presence of recombinant AmpCs (WT, S90A, and S90D) were identified by the imipenem neutralization assay. The discs contained increasing amounts (4 -7.5 g) of recombinant proteins and imipenem (5 g). Imipenem susceptibility was determined by measuring the diameters of the bacterial growth inhibition zones. All values were normalized to the mean value of the imipenem-only disc. Scale bars correspond to 10 mm. identified relatively high abundant phosphoproteins that may be linked to imipenem resistance, further biological validations are required to enhance the significance of our findings. Based on the Western blotting and phosphoproteomic analyses results, our study contributes to understanding of the potential phosphoproteins involved in imipenem resistance mechanisms.
Recently, a study comparing A. baumannii ATCC17978 and Abh12O-A2 pointed out potential virulence-related phosphoproteins in this emerging pathogen (27). Our results revealed many of the same phosphoproteins whose biological functions may be correlated to antibiotic resistance, such as pathogenesis (e.g. protein tyrosine kinase), virulence (e.g. KdpD, KdpE), drug resistance (e.g. ABC-type multidrug transport system), and stress response protein (e.g. superoxide dismutase) (see supplemental Table S11) (13,27). Hence, the overlapping phosphoproteins and phosphopeptides identified from the four A. baumannii strains may represent the most reliable candidates involved in carbapenem resistance (supplemental Table S4).
Phosphorylation of AmpC at Ser-90 Regulates Imipenem Susceptibility and ␤-Lactamase Activity-Previously, it has been reported that phosphorylation-dependent regulation of the ␤-lactamase modulated its enzyme activity and secretion. In Myxococcus Xanthus, the transmembrane protein kinase Pkn2 can phosphorylate threonine residues of class A TEMtype ␤-lactamase, thus controlling its enzyme activity and localization (66,67). The Ser-88 phosphosite of AmpC was identified in a phosphoproteome study of E. coli; however, the regulatory networks of phosphorylation-mediated AmpC in bacterial signal transduction remain unclear (27,54). Previous studies only revealed a connection between overexpression of AmpC and ␤-lactam-resistant pathogens (21,92). Our results may provide insight into the phosphorylation-dependent regulation of AmpC in imipenem-resistant A. baumannii.
The identification of two phosphorylated serine residues (Ser-88 and Ser-90) in the catalytic motif S 88 VS 90 K 91 of AmpC suggested that these residues regulate enzyme activity and substrate recognition (Fig. 3A). Thus, strains carrying a substitution at the catalytic Ser-88 were sensitive to both imipenem and ceftazidime, which agrees with the results of studies showing that the mutation and phosphorylation of Ser-88 significantly impaired AmpC enzyme function (supplemental Table S9) (71,93). Moreover, our antibiotic susceptibility results for mutants S90A and S90D revealed that the regulation of imipenem resistance was mediated by AmpC Ser-90 phosphorylation (Table II and supplemental Table S8). The S90A mutant showed both enhanced ␤-lactamase activity and the ability to neutralize imipenem; in contrast, the S90D mutant caused the imipenem-sensitive phenotype because of the inability to neutralize imipenem (Fig. 3B, 4, supplemental Table S10, and supplemental Fig. S6). The inability of the S90D mutant to neutralize imipenem suggests that Ser-90-phosphorylated AmpC interferes with the AmpC-imipenem interaction (51). Here, we provided the first evidence highlighting the significance of phosphorylation-mediated regulation of AmpC with respect to the imipenem resistance of A. baumannii.
Phosphorylation of AmpC at Ser-90 may Modulate the Interaction between AmpC and Imipenem from Docking Calculations-Numerous studies of prokaryotic and eukaryotic enzymes have shown that conserved phosphosites stabilize the phosphate group in the catalytic site to modulate conformational changes during substrate binding, thereby achieving phosphorylation-mediated signal transduction (94 -96). As a covalent inhibitor of class C ␤-lactamases, imipenem may initially noncovalently bind to AmpC and then form a nonhydrolyzable covalent intermediate in the acylation step, thus inactivating enzymatic function (51). Based on the substrateassisted catalysis model, imipenem lacking the equivalent nitrogen on its 6(7)␣ substituent is trapped in the acyl complex state without the deacylation step (51,69,70). Ser-90 phosphorylation may perturb the initial noncovalent binding step indirectly when AmpC encounters imipenem by altering the charge state or shape of the binding pocket (51,97).
Our docking calculations of four AmpCs (WT, phosphorylated-Ser-90, S90A, and S90D) in complex with imipenem revealed their pocket shapes and the relative distances between imipenem and the catalytic residues in the motif S 88 VS 90 K 91 (supplemental Fig. S7 and S8). Lys-91, which plays an electrostatic role, may contribute to the deprotonation of Ser-88O␥ to attack the ␤-lactam ring carbon of imipenem during the acylation step (51,69). Next, Lys-91N forms hydrogen bonds with Ser-88O␥ and the carbonyl oxygen O-7 of imipenem (Imi O-7) in the nonhydrolyzable AmpCimipenem complex (51,69). From the minimized pockets of AmpC-WT and the S90A mutant, the pose of imipenem suggests that the orientations are favorable for Lys-91N to form hydrogen bonds with Imi O-7, and Ser-88O␥ to attack the ␤-lactam ring carbon of imipenem (supplemental Fig. S7A, S7C, S8A, S8C, and S9B). Additionally, the Lys-91N in the models of AmpC-WT and the S90A mutant are close to Ser-88O␥, with distances of 3.4Å and 2.7Å, respectively, suggesting that Lys-91N facilitates the deprotonation of Ser-88 (supplemental Fig. S8A and S8C). However, the pose of imipenem in the pockets of the phosphorylated-Ser-90 and S90D mutant may reflect the loss of hydrogen bonding between Lys-91N and Imi O-7, as well as the inability of Ser-88O␥ to attack the ␤-lactam ring carbon of imipenem (supplemental Fig. S8B, S8D; S9A, and S9C). In the phosphorylated-Ser-90 model, rotation of Lys-91N toward the phosphate group of phosphorylated-Ser-90 suggested that Lys-91N loses the ability to act as a general base for Ser-88O␥ as well as its hydrogen bonding interaction with Imi O-7 (supplemental Fig.  S8B and S9A). Moreover, the S90D mutant model showed the predicted structural impediment around the pocket and the loop, between the B8 and B9 strands from domain 1, in the active site (supplemental Fig. S7D, S9C, and S10). Taken together, both the phosphorylated-Ser-90 and S90D mutant models revealed an unfavorable pocket shape or orientation of Lys-91N for acylation, which agrees with the results of high imipenem sensitivity and the inability of neutralizing imipenem from the mutant S90D (Table II and Fig. 4). Our docking results revealed that Ser-90 phosphorylation of AmpC may modulate the pocket shape or the orientation of Lys-91N, which may indirectly interfere with the interaction between AmpC and imipenem.
Protein phosphorylation plays an important role in the regulation of a wide variety of cellular functions in both prokaryotic and eukaryotic cells. Bacterial protein phosphorylation is required for the biosynthesis of capsular polysaccharides, biofilm development, virulence, and antibiotic resistance, particularly through two-component signaling systems (44,55,98,99). Thus, it has been demonstrated that His and Asp phosphorylation regulate the functions of pathogenesis, such as adherence, motility, enhancement of toxicity, quorum sensing, capsule formation, and drug resistance (100). Recently, in cancer drug development studies, it has been suggested that changes in phosphorylation stoichiometry are helpful for characterizing the kinase-dependent pathway in gefitinib-resistant lung cancer cells (101). Our study showed that Ser-90 phosphorylation of AmpC may be part of the kinase/ phosphatase machinery conferring imipenem resistance, ␤-lactamase activity, and the ability to neutralize imipenem in A. baumannii. Thus, upstream cognate kinases/phosphatases involved in the phosphorylation-mediated regulation of AmpC ␤-lactamase activity may serve as targets for the design of effective antibacterial agents (102,103). Therefore, drugs designed to interfere with bacterial protein phosphorylation cascades may provide an alternative way for eliminating pathogenic bacteria in the post antibiotic-resistant era.