Investigation of Novel pmrB and eptA Mutations in Isogenic Acinetobacter baumannii Isolates Associated with Colistin Resistance and Increased Virulence In Vivo

Description of isolates.

To investigate the mechanisms of colistin resistance, we selected four pairs of A. baumannii isolates obtained during the patients’ hospital course (Table 1). The isolates were collected from patients hospitalized in Germany, Greece, and Spain. With the exception of isolate pair 2, all isolates were recovered while the patients were receiving colistin therapy. The isogenic background of all isolates representing an isolate pair was confirmed by WGS (allele differences, <10). The isolates exhibited an increase in colistin MIC by at least four log₂ dilutions as determined by broth microdilution. The isolates from Spain and Greece all belonged to the international clonal lineage IC2, while isolate pair 4 belonged to IC4. Isolates with low colistin MICs preceded the isolates with high colistin MICs within an isolate pair (Table 1). Isolate pair 1 and 2 were recovered from patients hospitalized in Seville, Spain, and displayed an increase in colistin MICs from 2 to 256 mg/liter. In isolate pair 1, isolates were obtained from one patient, and in isolate pair 2, both isolates were recovered from different patients at day 3 of therapy. Both isolates from isolate pair 3, recovered from different patients hospitalized in Athens, Greece, were considered colistin resistant. Nevertheless, the colistin MICs differed by at least four log₂ dilutions (32 versus ≥256 mg/liter). Isolate pair 4 was recovered from a patient hospitalized in Frankfurt, Germany, and it exhibited a colistin MIC increase from 1 to 256 mg/liter. The patient was infected with both isolates at the same time when admitted to the hospital. All MIC differences were seen consistently in three independent experiments.

Resistome analysis.

In addition, the MICs of 12 different antimicrobials were determined, and the resistome of each isolate was analyzed by ResFinder (Table S1, Fig. S1). In general, all isolates have an MDR phenotype, as they were resistant to at least nine different antimicrobials from six different antimicrobial classes (Table S1). Moreover, the MICs were very similar, and phenotype shifts from susceptible to resistant were only observed in isolate pair 1 for tigecycline. The acquired resistomes were identical in isolate pair 1 and 3, while isolate pair 2 differed in aminoglycoside and sulfonamide resistance determinants and isolate pair 4 differed in aminoglycoside and tetracycline resistance determinants (Fig. S1).

Analysis of pmrCAB, eptA, and other allele differences.

Comparison of the draft genomes revealed differences in the pmrCAB operon between isolates representing an isolate pair. In the four isolates with high colistin MICs, the amino acid substitutions S17R, A28V, and I232T, as well as the 4-amino acid deletion ΔL9-G12, were detected in PmrB (Table 1). Both PmrA and PmrC were identical in isolates of an isolate pair. Moreover, a pmrC homologue (termed eptA) was present in isolate pairs 1, 2, and 3. However, eptA was not detected in isolate pair 4, A. baumannii ATCC 19606, or the pmrAB knockout, as determined by WGS or by PCR. The amino acid substitution R127L was detected in EptA in the colistin-resistant isolate MB_177 of isolate pair 2. Moreover, ISAba1 was found in reverse orientation upstream of eptA in both isolates MB_23C and MB_177. While the insertion site was identical, a single point mutation (A→T) was observed in position 1091, close to the putative promoter region in ISAba1 in isolate MB_177.

Besides mutations in the pmrCAB operon, disruption of H-NS or mutations in/disruptions of genes involved in LPS/lipid A biosynthesis have been associated with colistin resistance. Since no differences were detected, these gene products were considered not to be associated with colistin resistance in our isolates. Furthermore, the plasmid-mediated resistance determinant mcr-1 or its homologues were not present. Other allele differences were detected in all isolate pairs except for isolate pair 4 (Table S2). Isolate pair 1 differed in three alleles, isolate pair 2 in eight alleles, and isolate pair 3 in four alleles. Most of these alleles included hypothetical and uncharacterized proteins, or components of bacterial metabolism, but also two putative transcription regulators differed in isolate pair 2.

Increased expression of pmrC and eptA.

Since amino acid deletions or substitutions were detected in PmrB in all colistin-resistant isolates, we investigated their effect on the expression of pmrC by semi-quantitative reverse transcription-PCR (qRT-PCR). The relative expression was compared between isolates of an isolate pair (Fig. 1A). The expression of pmrC was significantly increased in the colistin-resistant isolates, MB_6C and SG3166. Compared to that in the colistin-susceptible isolates, pmrC was expressed approximately 13× more than that in isolate MB_6C, while in isolate SG3166, a 10-fold increase in expression was observed. Although both isolates in isolate pair 3 were considered colistin resistant, the expression of pmrC was approximately 4× higher in isolate MB_119 compared to that in MB_90. Interestingly, the expression of pmrC was significantly decreased in MB_177 (the colistin-resistant isolate of isolate pair 2), so a different resistance mechanism is suggested in this isolate. Furthermore, the analysis of eptA expression by qRT-PCR revealed slightly increased expression in isolate MB_6C and decreased expression in isolate MB_119 compared to that in the isolates with low colistin MICs (Fig. 1B). In contrast, the expression of eptA was significantly higher in isolate MB_177 compared to that in MB_23C, although in both isolates, ISAba1 was found in reverse orientation upstream of the gene. These findings suggest an association of point mutations in ISAba1 or eptA with increased eptA expression and colistin resistance.

Lipid A modifications.

An increased expression of pmrC or eptA is associated with the addition of PEtN to lipid A. To measure potential modifications of lipid A, LPS and lipid A was extracted from both mid-log-phase and overnight cultures of the isolate pairs and the reference strain A. baumannii ATCC 19606. Analysis by matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF) revealed no significant differences between mid-log-phase and overnight cultures, as very similar mass spectra were obtained. The following major lipid A species were detected in the reference strain ATCC 19606 and all isolate pairs (Fig. 2A to E): hepta-acylated lipid A (m/z 1,910), hexa-acylated lipid A (m/z 1,728), penta-acylated lipid A lacking one fatty acid moiety (m/z 1,530), and tetra-acylated lipid A (m/z 1,404). However, peaks at m/z 1,530 and m/z 1,404 were small and unclear for some isolates (data not shown). The addition of PEtN (theoretical m/z 124) to lipid A is shown by the mass at m/z 2,034 (+1 PEtN) and m/z 2,157 (+2 PEtN). A single PEtN addition was found in isolate pair 1, exclusively in the colistin-resistant isolate MB_6C (Fig. 2B). In contrast, the mass at m/z 2,034 was detected in both isolates of isolate pair 2 (Fig. 2C). This was also observed in isolate pair 4 (Fig. 2E). These unexpected findings suggest a different or additional resistance mechanism in isolates MB_177 and SG3166. Furthermore, the relatively high colistin MICs of both isolates of isolate pair 3 were probably not due to PEtN addition, since a single PEtN addition was detected in isolates MB_90 and MB_119 (Fig. 2D). Interestingly, the addition of two PEtN molecules to lipid A (m/z 2,157) was observed only in isolate MB_119. Moreover, the addition of galactosamine to lipid A, which is indicated by a mass at m/z 2,071, was not observed in any isolate. Although PEtN addition was detected in colistin-resistant isolates, it seemed to be a strain-specific mechanism.

Similar growth kinetics for all isolates.

Mutations in pmrB have been reported to carry a fitness cost (32). We therefore sought to investigate if our colistin-resistant isolates exhibited reduced growth and found that growth rates within an isolate pair did not differ (Fig. S2). In addition, growth rates and doubling times were calculated for all isolates, and this also revealed no differences within an isolate pair (data not shown). Therefore, the acquisition of colistin resistance did not appear to have a fitness cost as measured by growth kinetics.

Increased virulence in isolates with high colistin MICs.

Colistin resistance is reportedly associated with decreased fitness and virulence (32–33). Therefore, we also investigated the virulence of these isolate pairs in the G. mellonella model of infection. The larvae were infected with the A. baumannii isolates, and survival, as well as melanization, was assessed (Fig. 3 and 4). Interestingly, the colistin-resistant isolates exhibited increased virulence (Fig. 3). In isolate pair 1, the survival rate of the larvae infected with MB_2 was 40% after 48 h, while none survived when infected with MB_6C (P ≤ 0.021). After 96 h, the survival rate of larvae infected with MB_177 was also lower (30%) compared to that of those infected with MB_23C (60%) (P ≥ 0.04). Although both isolates of isolate pair 3 are colistin resistant, a difference in virulence was also observed. While 50% of larvae infected with isolate MB_90 survived, the survival rate of larvae infected with MB_119 was only 20% after 96 h (P ≤ 0.04). No significant differences were detected in comparison of larvae infected with the isolates of isolate pair 4; both isolates resulted in low survival rates of approximately 30% after 96 h, although a higher percentage of larvae were killed in the first 24 h when infected with the colistin-resistant isolate SG3166 (P ≤ 0.17).

Similar results were observed for melanization, which is used as an indicator for morbidity. Although all A. baumannii isolates induced high levels of melanization, higher melanization scores were observed in larvae infected with isolates with higher colistin MICs (Fig. 4). Larvae infected with the colistin-resistant isolate MB_6C displayed increased melanization over the entire time course (24 h to 96 h). Infection with isolates MB_177 and MB_119 led to unchanged or decreased rates of melanization after 24 h, but was elevated after 48 to 96 h. In contrast, larvae infected with isolate pair 4 revealed lower melanization scores for the colistin-susceptible isolate after 24 h and very similar and constant melanization scores between 48 and 96 h for both isolates, indicating no significant difference in morbidity. Both the decreased survival rates and the elevated melanization scores indicate a greater virulence in the colistin-resistant A. baumannii isolates MB_6C, MB_117, and MB_119.

Introduction of pmrAB variants into A. baumannii ΔpmrAB.

To further analyze the association of pmrB mutations with colistin resistance, a pmrAB knockout strain was created (ΔpmrAB). The pmrAB variants from isolates MB_2 and MB_6C, as well as that from ATCC 19606 as a control, were introduced into ΔpmrAB, and colistin MICs were determined (Table S3). Surprisingly, the colistin MICs were 0.5 to 1 mg/liter, regardless of the pmrAB variant introduced into the knockout strain. To confirm the expression of pmrAB on the plasmid and pmrC in ΔpmrAB, gene expression was analyzed by qRT-PCR. Large numbers of pmrA and pmrB transcripts were detected in all cases, indicating expression of both genes on the plasmid (data not shown). Moreover, the expression of pmrC was significantly increased in ΔpmrAB complemented with pmrAB from isolate MB_6C (ΔL9-G12), whereas the expression did not differ in ΔpmrAB complemented with pmrAB from ATCC 19606 and isolate MB_2 (Fig. 5). Slightly increased pmrC expression was detected in the wild-type ATCC 19606 compared to ΔpmrAB, indicating a basal expression of pmrC.

Since pmrC was overexpressed in ΔpmrAB harboring pJN17/04::pmrAB-6, the modification of lipid A was analyzed by mass spectrometry (Fig. 6). The same major lipid A species already described for the isolate pairs (m/z 1,404, 1,530, 1,728, and 1,910) were also detected in the pmrAB knockout strains. The addition of a single PEtN molecule indicated by the mass at m/z 2,034 was observed only in ΔpmrAB complemented with pJN17/04::pmrAB-6C (Fig. 6D), indicating that the pmrB mutation resulted both in increased pmrC expression and lipid A modification. Since the colistin MICs did not change, a different, hitherto unknown factor contributing to resistance or strain dependency are strongly suggested.

Bacterial isolates.

In total, 8 A. baumannii isolates were investigated. The isolates demonstrated an increase in colistin MIC by at least four dilution steps. In addition, isolates representing isolate pairs were required to demonstrate fewer than ten allele differences in core genome multilocus sequence typing (cgMLST). The isolates were obtained from four patients hospitalized in Frankfurt (Germany; n = 1), Athens (Greece; n = 1), and Seville (Spain; n = 2). The isolates from Greece and Spain were recovered from respiratory tract or stool samples obtained from patients with ventilator-associated pneumonia (VAP) between 2012 and 2015 as part of the MagicBullet clinical trial (9, 18). The isolates from Germany were recovered from wound and respiratory samples obtained in 2016 from a stroke patient.

Antimicrobial susceptibility testing.

Susceptibility to colistin was determined by broth microdilution using premanufactured freeze-dried microtiter plates (Micronaut S Colistin plates; Merlin Diagnostika, Bornheim, Germany). The MIC was determined in cation-adjusted Mueller-Hinton broth (CAMHB) according to the international standard ISO 20776-1 (19). Colistin resistance was defined as an MIC of >2mg/liter, using the EUCAST resistance breakpoints for A. baumannii. Susceptibility to tetracycline, gentamicin (Sigma–Aldrich, Steinheim, Germany), amikacin, ciprofloxacin, meropenem, minocycline, rifampin, tigecycline (Molekula, Newcastle upon Tyne, UK), levofloxacin (Sanofi Aventis, Frankfurt, Germany), azithromycin (Pfizer Pharma GmbH, Berlin, Germany), and erythromycin (AppliChem, Darmstadt, Germany) was determined by agar dilution following the current CLSI guidelines (20). All susceptibility tests were repeated at least three times independently. E. coli ATCC 25922, P. aeruginosa ATCC 27853 and Staphylococcus aureus ATCC 29213 were used as controls.

WGS and sequence data analysis.

All A. baumannii isolates were subjected to WGS (21). Sequencing libraries were prepared using the Nextera XT library prep kit (Illumina GmbH, Munich, Germany) for a 250-bp paired-end sequencing run on an Illumina MiSeq sequencer. The resulting FASTQ files were quality trimmed and assembled de novo using Velvet with SeqSphere⁺ v.3.0 software (Ridom GmbH, Münster, Germany). The assembled genomes were aligned to the A. baumannii reference strain ACICU to determine differences in specific open reading frames in each draft genome sequence, as well as for cgMLST (21), and were further analyzed using ResFinder for resistome analysis (22). Besides the amino acid sequence of PmrCAB, the sequences of H-NS and of the LPS/lipid A biosynthesis components LpsB, LpxABCD, LpxHKLM, and KdtA/WaaA were compared between the isolates and to those of the reference strain ACICU.

PCR-based methods.

All primers used in this study were designed using Primer3 software (http://bioinfo.ut.ee/primer3/) and are listed in Table 2. Crude cell lysates or genomic DNA were used as the templates for PCR. Differences determined by WGS in pmrCAB and eptA between isolates of an isolate pair were confirmed by Sanger sequencing (LGC Genomics GmbH, Berlin, Germany).

Semiquantitative reverse transcription-PCR.

To measure the expression of pmrCAB and eptA, qRT-PCR was performed as previously described, using rpoB as reference gene (25). The specificity of the pmrC and eptA primers were confirmed by PCR. All qRT-PCRs were done in triplicates using freshly prepared RNA and cDNA and were repeated independently at least three times. The relative expressions of pmrCAB and eptA were compared with either the isolates representing an isolate pair or with the pmrAB knockout strain, and statistical analysis was done by performing an unpaired t test using the recorded absolute values.

Markerless pmrAB knockout and cloning of pmrAB.

A pmrAB knockout was created as previously described by Stahl et al. (26), using A. baumannii ATCC 19606 as the parent strain. The upstream and downstream regions of pmrAB were integrated into plasmid pBIISK_sacBkanR using the In-Fusion HD EcoDry cloning kit (TaKaRa Clontech, Saint-Germain-en-Laye, France) and primers J76 to J79 (Table 2). The successful deletion of pmrAB was confirmed by WGS. The pmrAB variants of A. baumannii ATCC 19606 as control and of isolates MB_2 and MB_6C were amplified using the primer pair K58 and K59 (Table 2), and the shuttle vector pJN17/04 was linearized as previously described (27). pmrAB was integrated into the plasmid using the In-Fusion HD EcoDry cloning kit. The plasmids pJN17/04::pmrAB-19606, pJN17/04::pmrAB-2, and pJN17/04::pmrAB-6C were introduced into A. baumannii ΔpmrAB by electroporation (27). Successful transformants were confirmed by PCR and maintained on Luria-Bertani (LB) agar supplemented with 10 mg/liter kanamycin (Molekula, Newcastle upon Tyne, UK).

Lipid A extraction and mass spectrometry.

For LPS and lipid A extraction, all isolates and the reference strain A. baumannii ATCC 19606 were cultured in 250 ml of LB medium at 37°C and 220 rpm. To determine if the composition of lipid A varied between growth phases, the bacteria were cultured overnight or until the mid-log phase was reached (optical density at 600 nm [OD₆₀₀], 0.7 to 0.9). The cultures were centrifuged at 4,000 × g for 15 min, and the cell pellet was washed twice with phosphate-buffered saline (PBS). Approximately 200 to 400 mg of the cell pellets were subjected to LPS extraction following the Tri Reagent method with three additional chloroform extraction steps (28). Mass spectrometry analysis was done using a MALDI-TOF mass spectrometer (Bruker Biflex III; Bruker Daltonics, Bremen, Germany). Crude lipid A extract was resolved in chloroform:methanol:H₂O (4:4:1 [vol/vol]) and subjected to an ultrasound bath for 20 min, followed by a centrifugation step at 13,000 rpm for 2 min. Aliquots of 1 μl sample were mixed 1:1 with the matrix (5-chloro-2-mercaptobenzothiazole and ammonium-EDTA), spotted directly onto the MALDI target plate, and dried in air. Mass spectra were measured using negative ion mode within a mass range of 500 to 3,000 Da (29).

Growth kinetics.

Growth kinetics were performed in triplicates as previously described (27). Briefly, a McFarland suspension of 0.5 generated from a fresh overnight culture was used to inoculate 10 ml Mueller-Hinton (MH) broth. The cultures were incubated for six hours at 37°C and 200 rpm, and a sample was taken every hour. Samples were serially diluted in cold saline and plated onto MH agar, and CFU per milliliter values were determined after overnight incubation.

In vivo virulence in G. mellonella.

The in vivo virulence of the A. baumannii isolates was investigated using the G. mellonella invertebrate model (TruLarv; Biosystems Technology, Exeter, UK) as previously described (30). For each A. baumannii isolate, 10 larvae were infected by injection with 10⁵ CFU/larva of the respective isolate into the left proleg. PBS was used in the control injections. The larvae were incubated aerobically at 37°C and scored for survival (live/dead) at 0, 24, 48, 72, and 96 h postinfection. Moreover, the melanization scores were recorded as an indicator of morbidity (31). All assays were performed in triplicates and results were pooled for analysis. Statistical analysis was carried out using an unpaired t test at 96 h, as described previously (30).

Accession number(s).

All raw reads generated were submitted to the Sequence Read Archive of the National Center for Biotechnology Information (NCBI) under the accession numbers SRP131448 and SRP155178, as well as under the BioProject accession number PRJNA482774.