Ser253Leu substitution in PmrB contributes to colistin resistance in clinical Acinetobacter nosocomialis

ABSTRACT Infections caused by extensively drug-resistant (XDR) Acinetobacter nosocomialis have become a challenging problem. The frequent use of colistin as the last resort drug for XDR bacteria has led to the emergence of colistin-resistant A. nosocomialis (ColRAN) in hospitals. The mechanism of colistin resistance in A. nosocomialis remains unclear. This study aimed to investigate the mechanisms underlying colistin resistance in clinical ColRAN isolates. We collected 36 A. nosocomialis isolates from clinical blood cultures, including 24 ColRAN and 12 colistin-susceptible A. nosocomialis (ColSAN). The 24 ColRAN isolates clustered with ST1272 (13), ST433 (eight), ST1275 (two), and ST410 (one) by multilocus sequence typing. There was a positive relationship between pmrCAB operon expression and colistin resistance. Further analysis showed that colistin resistance was related to an amino acid substitution, Ser253Leu in PmrB. By introducing a series of recombinant PmrB constructs into a PmrB knockout strain and protein structural model analyses, we demonstrated that the association between Ser253Leu and Leu244 in PmrB was coupled with colistin resistance in ColRAN. To the best of our knowledge, this is the first study demonstrating that the key amino acid Ser253Leu in PmrB is associated with overexpression of the pmrCAB operon and hence colistin resistance. This study provides insight into the mechanism of colistin resistance in A. nosocomialis.


Introduction
A unneglectable number of extensively drug-resistant (XDR) Acinetobacter spp. that have acquired resistance to multiple classes of antibiotics has emerged [1][2][3]. XDR Acinetobacter spp. are resistant to nearly all antimicrobial agents, including carbapenems, cephalosporins, penicillins, aminoglycosides, and fluoroquinolones [4]. In the absence of antibiotic selection, colistin has become a very limited option [5]. In current medical practice, colistin is becoming more frequently prescribed to treat XDR Acinetobacter spp. However, the increased use of colistin has led to the development of colistin resistance in such pathogens [6,7]. Acinetobacter baumannii and A. nosocomialis are the two most frequently observed Acinetobacter clinical isolates [3,8,9]. A. nosocomialis has a higher rate of colistin resistance than A. baumannii [10,11] and the difference is even greater when both species are carbapenem-resistant (45.3% vs. 1.4%) [11][12][13].
The two main colistin-resistance mechanisms have previously been described in A. baumannii. The first mechanism is the modification of the lipid A component in LPS mediated by the pmrCAB operon. The expression of the pmrCAB operon is tightly regulated by the PmrA/PmrB two-component system, which contains a sensor kinase (PmrB) and a response regulator (PmrA) encoded by the same pmrCAB operon. Once the PmrA/PmrB two-component system is activated, phosphorylated PmrA binds to the promoter region of the pmrCAB operon and stimulates the expression of the pmrCAB operon to generate lipid A phosphoethanolamine transferase, which leads to the modification of lipid A [14,15]. This modification reduces the net negative charge of the outer membrane and results in colistin resistance [16]. The second mechanism is the complete loss of LPS expression to impaired lipid A synthesis via mutations in the lpxA, lpxC, and lpxD genes [17][18][19]. Recently, plasmid-mediated colistin resistance (mcr) genes have been found to be associated with colistin resistance [9].
In the case of other Acinetobacter spp., studies regarding the mechanisms of colistin resistance remain limited. In the current study, we focused on the mechanism of resistance in clinical colistin-resistant A. nosocomialis (ColRAN) blood isolates.

Bacterial strains and antimicrobial susceptibility testing
Isolates of A. nosocomialis were collected from clinical blood samples at four medical centres of AntimiCrobial studies in Taiwan Operating Network (ACTION) from 2009 to 2015. The medical centres included Changhua Christian Hospital (CCH), Mackay Memorial Hospital (MKH), Taipei Veterans General Hospital (TVGH), and Tri-Service General Hospital (TSGH) of the National Defense Medical Center (NDMC). This study was approved by the Ethics Committee of the aforementioned centres. The minimum inhibitory concentration (MIC) of colistin was evaluated using the broth microdilution method [20]. Colistin resistance was defined as the presence of an MIC >2 mg/L.

Molecular typing
The clonal relationships of the isolates were determined using multilocus sequence typing (MLST) and pulsed-field gel electrophoresis (PFGE) [11]. All isolates subjected to PFGE were analysed using Bio-Numerics software (Applied Maths) to determine the relatedness of the isolates. Isolates were considered to be of different types if they had more than three DNA fragment differences and a similarity of <80% in dendrogram analysis. MLST was performed using Pasteur schemes, as suggested in a previous study [21]. Allelic profiles (in the order: cpn60-fusA-gltA-pyrG-recA-rplB-rpoB) were used to assign sequence types (STs) to all isolates. eBURST analysis (http:// eburst.mlst.net) was performed to assign each ST to a respective clonal complex (CC) [22].

Transcriptional analyses by quantitative realtime PCR assays
The transcription levels of pmrC, pmrA, lpxA, and lpxC genes were measured by quantitative real-time PCR assays. Real-time PCR was performed using the QuantiNova SYBR Green PCR Kit (Qiagen, Hilden, Germany). mRNA of the rpoB gene was used as a control and ATCC17903 was used as a reference to the standard expression level.

Mutants of pmrb deletion and expression experiments
The bacterial strains and plasmids used in this study are listed in Supplementary  Table  2. ATCC17903ΔpmrB was an unmarked deletion mutant created as suggested in a previous study [24]. pmrB gene fragments were cloned and generated into a series of recombinant pS01_pmrB clones (ATCC17903, Type 1, and Type 2). Specific nucleotide mutations in the pmrB gene were generated using overlap extension PCR. The mutated pmrB genes were cloned to produce pS01_pmrB (Type 2, Ile243-Ser) and pS01_pmrB (Type 2, Leu244Ser).

Structural modelling analysis
The structural model of wild-type A. nosocomialis PmrB (Type 2) was constructed using SWISS-MODEL [25,26], a fully automated protein structure homology-modelling server, using the Protein Data Bank accession code 4BIV (the cytoplasmic region of Escherichia coli sensor histidine kinase CpxA) as the template [27]. The structural model of the A. nosocomialis PmrB Ser253Leu mutant was generated by Coot using the wild-type PmrB modelling structure [28]. Structural figures were produced using PyMOL (DeLano Scientific, http://www.pymol. org).

Data analysis
Statistical analysis was performed using SPSS version 20.0, and statistical differences among various groups were calculated using the Mann-Whitney test. Differences were considered statistically significant at p <0.05.

Transcription level of the pmrCAB and lpxACD operons
Comparing all ColSAN and ColRAN isolates, the RNA expression levels of the pmrC, pmrA, lpxC, and lpxA genes of the ColRAN isolates were 2.6-, 2.1-, 1.3-, and 1.3-fold higher than those of the ColSAN isolates, respectively ( Figure 1). The transcription levels of the pmrC and pmrA genes in ColRAN isolates were significantly higher than those in ColSAN isolates (both p < 0.05). However, neither lpxC nor the lpxA gene was overexpressed.

The association of PmrAB two-component systems and colistin resistance
Amino acid analysis revealed that the isolates possessed three types of PmrA (A, B, and C types), and six types of PmrB (Types 1-6), including one novel variant (Type 3) (Supplementary Table 3). There was a significant difference in the PmrB pattern between the ColRAN and ColSAN isolates (p < 0.05). The most common combination of the PmrA-PmrB pattern was A-2 (19 isolates), with colistin MICs ranging from 4 to 8 mg/L (Table 1). Among the 19 ColRAN isolates with the A-2 pattern, 16 had pmrC overexpression (84.2%). Therefore, we believe that colistin resistance mediated by pmrC overexpression may be associated with amino acid substitutions in the PmrB sensor sequence (Type 2).

Analysis of amino acid substitutions within the PmrB sensor
After comparison with the reference strain ATCC17903, 15 amino acid substitutions were found in the six PmrB sequences (Supplementary Table 4). The major types of PmrB were Type 2, including ST1272 (12 isolates) and ST433 (seven isolates). There was a distinctive amino acid substitution (Ser253Leu) between type 1 and type 2 PmrB. In comparison to Type 1, the colistin MIC values of Type 2 showed an 8-fold increase. Among the clinical isolates, the transcription level of pmrC in isolates carrying Type 2 PmrB was significantly higher than that in isolates carrying Type 1 PmrB (3.63 ± 1.8 vs. 0.87 ± 0.76, p < 0.05). Hence, it was necessary to investigate whether colistin resistance was associated with this amino acid substitution in PmrB.

Complementation of PmrB mutations
The MIC of colistin against ATCC17903ΔpmrB carrying pS01_pmrB (ATCC17903) was higher than that of the strain carrying pS01 (empty vector) and was associated with a higher expression level of the pmrC gene (Table 2). ATCC17903ΔpmrB transformed with pS01_pmrB (Type 2) showed a higher MIC of colistin than pS01_pmrB (Type 1) (16 vs. 2 mg/L). The transcription levels of the pmrC and pmrA genes in ATCC17903ΔpmrB with pS01_pmrB (Type 2) were significantly higher than those in pS01_pmrB (Type 1). Furthermore, the transcription levels of lpxA and lpxC in the two transformants were not significantly different. Hence, the Ser253Leu point substitution in PmrB appeared to be the critical amino acid substitution causing colistin resistance.

Structural analysis of PmrB with critical ser253leu substitution
To address the structural effect of PmrB with the critical amino acid substitution of Ser253Leu on colistin resistance, a structural model of the Type 1 PmrB homodimer was generated by the homology-modelling server using the CpxA structure as the template (Figure 2a and b). The Ser253Leu substitution of PmrB resulted in two possible orientations of the side chain of the Leu253 residue. In the first orientation, the side chain of Leu253 was too close to the side chain of Leu244 on the opposite subunit of the PmrB homodimer, causing a steric collision (Figure 2c). In the second orientation, the side chain of Leu253 was close to the side chain of Ile243 on the opposite subunit of the PmrB homodimer, increasing hydrophobic interactions between these two residues (Figure 2d).
To identify the critical amino acid interacting with Leu253, two types of mutations (Ile243Ser and Leu244Ser) were constructed in Type 2 PmrB. ATCC17903ΔpmrB transformed with pS01_pmrB (Type 2, Leu244Ser) showed significantly lower MICs of colistin than pS01_pmrB (Type 2) (2 vs. 16 mg/L) ( Table 2). Comparing the two transformants, the transcription levels of the pmrC and pmrA genes were related to the MIC of colistin. Summarizing the results of the structural modelling and MIC assays, Leu253 in PmrB was possibly related to colistin resistance by pushing Leu244 on the opposite subunit of the PmrB homodimer.

Discussion
The emergence of colistin resistance in Acinetobacter spp. is a serious clinical threat. However, an indepth investigation of the mechanism of colistin resistance in ColRAN is still lacking. This is the first detailed description of the mechanism of resistance caused by an amino acid substitution (Ser253Leu) in PmrB. Under normal circumstances, colistin targets the anionic lipid A portion of LPS and binds to phospholipids in bacterial cell membranes, leading to changes in the permeability of the outer cell membrane and leakage of cell contents (Figure 3a). The Ser253Leu substitution in this study was further identified to increase the interaction with Leu244 on the opposite side of PmrB and promote conformational changes in PmrB dimers. This interaction within PmrB drives the expression of the pmrCAB operon via PmrA phosphorylation. The PetN transferase encoded by the pmrC gene increases and adds the positively charged PetN to lipid A. Finally, the positively charged lipid A of LPS prevented the binding of cationic colistin (Figure 3b).  Among the ColRAN isolates found in this study, 12 isolates (54.2%) belonged to ST1272, and eight isolates (33.3%) belonged to ST433. We traced the source of our isolates by searching the MLST database and found that ST1272 is a novel ST identified in this study. ST433 was originally isolated in Taiwan in 2003. In contrast, ST410 was found to be the most dominant type of clinical carbapenem-resistant (CR) A. nosocomialis isolates in Taiwan [29]. Notably, in our unpublished study analysing the correlation between MLST types and carbapenem resistance, we found that the carbapenem resistance rates among different MLST types of A. nosocomialis isolates were 4.8, 100, and 46.7% for ST410, ST433, and ST1272, respectively.
These results suggest that A. nosocomialis, with different MLST types under antibiotic stress, has different evolutionary capabilities. Unknown fitness advantages exist in ST1272 and ST433, and these advantages probably drive them to develop colistin resistance in the hospital setting and under antimicrobial selection pressure [6,7,13]. In this study, six ColRAN isolates (MIC = 4-8 mg/ L) carried Type 2 PmrB (three), Type 5 PmrB (one), and Type 6 PmrB (two), and did not show overexpression of pmrC. Some unidentified factors may have affected the expression of pmrCAB in these clinical isolates. Among the three isolates carrying type 2 PmrB, without overexpressed pmrC, colistin resistance was probably due to mechanisms other than pmrCAB activation. Previous studies have identified that other  genetic determinants conferred colistin resistance in A. baumannii, including chromosome-mediated mechanisms (the eptA gene) and plasmid-mediated mechanisms (mcr genes) [30,31]. To identify the cause of colistin resistance, we also performed whole genome sequencing of these ColRAN isolates and found no relevant mechanism (data not shown). Further in-depth investigations are required to elucidate the mechanism of colistin resistance in these isolates.
In the PmrB Ser253Leu mutant model, Leu244 and Ser253/Leu253 residues were located on the respective alpha-helices of dimerization and the histidine phosphotransfer (DHp) domain. The side chain of the Leu253 residue was too close to the Leu244 residue on the opposite subunit of the PmrB homodimer, which then pushes out the DHp domain and causes helical bending of the histidine kinase DHp domain. In previous reports focused on the A. baumannii two-component system-mediated regulation of the AdeRS two-component system, the helical bending of the histidine kinase DHp domain, caused by a highly conserved proline residue (Pro154 of AdeS) in the proximity of phosphorylated histidine residues (His149 of AdeS), was determined to be essential for histidine kinase activation [27,32,33]. Combining the results of structural modelling and MIC assays, we speculated that Leu253 residues present a van der Waals repulsive force to the Leu244 residue in the PmrB Ser253Leu mutant, resulting in the constitutive helical bending conformation of the PmrB DHp domain, leading to the activation of PmrB and hence colistin resistance.

Conclusion
To the best of our knowledge, this is the first study to identify a new Ser253Leu substitution in PmrB and mediate colistin resistance in clinical A. nosocomialis isolates. The effect of this substitution in the PmrBactivated pmrCAB operon created by the potential association between these two amino acids (Leu253 and Leu244) led to colistin resistance. The substitution position of Ser253Leu is key for a deeper understanding of the mechanism of colistin resistance in ColRAN isolates. The discovery of this result provided an indepth investigation of the mechanism of colistin resistance and shed light on drug development for fighting such pathogens.