Dissecting Colistin Resistance Mechanisms in Extensively Drug-Resistant Acinetobacter baumannii Clinical Isolates

We deleted pmrA from the genome of a panel of 12 colistin-resistant A. baumannii strains to evaluate the transcriptional regulator PmrA as a potential drug target to rejuvenate colistin efficacy in A. baumannii. The strains in the panel consisted of recently isolated colistin-resistant clinical strains collected from diverse geographical origins. They belong to three distinct and highly successful clonal lineages, the international clone 1 (ST1), international clone 2 (ST2), and ST25 clonal lineages (Table 1) (20–22). All strains were classified as extensively drug resistant according to the criteria of Magiorakos et al. (23). These data underscore the clinical relevance and the diversity of the strain panel. The colistin-susceptible A. baumannii ATCC 17978 strain was included as a reference strain. In all strains, pmrA was deleted by applying a previously described method that allows efficient scarless genome engineering even in extensively resistant A. baumannii clinical isolates (24).

The colistin sensitivity of the parental clinical isolates and their corresponding pmrA knockout mutants (ΔpmrA) was determined by broth microdilution method. pmrA deletion reduced MICs 64- to 1,024-fold in 10 out of 12 initially colistin-resistant clinical isolates (83%), thus restoring susceptibility to colistin (MIC, ≤2 μg/ml) (Table 2). To our surprise, however, two strains (BV94 and BV189) retained colistin resistance even in the absence of pmrA.

We investigated the differences between strains that became susceptible after pmrA deletion and those that remained resistant by analyzing the sequence variations of the PmrAB TCS in the strain panel. The PmrA and PmrB sequences of the A. baumannii AYE, ACICU, and NIPH 146 strains were used as references for ST1, ST2, and ST25 clonal lineages, respectively. Nonsynonymous mutations were found only in the PmrB sensor kinase (Table 2). Interestingly, the two strains with an unaltered PmrB sequence were those that remained colistin resistant after pmrA deletion (BV94 and BV189). Our data suggest that colistin resistance in these two strains is not conferred by PmrA-mediated PmrC overexpression. We confirmed this hypothesis by quantifying the expression of pmrC using quantitative reverse transcription-PCR (qRT-PCR) (Fig. 1). The control strain ATCC 17978 and the two refractory strains BV94 and BV189 showed only marginal levels of pmrC expression. In contrast, the 10 other strains showed pmrC overexpression, and this overexpression was abolished in the ΔpmrA mutant.

Taken together, colistin resistance in A. baumannii is predominantly conferred by mutations in the PmrB TCS sensor kinase that lead to overexpression of PmrC, as shown in 10 out of 12 clinical strains. However, some isolates (2 out of 12 strains in our panel) may use an alternative colistin resistance mechanism independent of PmrA-mediated PmrC overexpression to resist the antibacterial activity of colistin.

Lesho et al. described the presence of the alternative PetN transferase EptA in A. baumannii (17). EptA and PmrC are homologous proteins with 93% amino acid identity, suggesting similar enzymatic activities. However, the role of EptA in A. baumannii colistin resistance is still unclear (17). To investigate the prevalence of eptA in A. baumannii, we took advantage of sequence differences between pmrC and eptA at the N- and C-terminal ends of the open reading frames and designed oligonucleotides (oVT152/oVT153) that can discriminate eptA from pmrC (Fig. 2A; see also Table S1 in the supplemental material). Using these eptA-specific primers, we detected eptA in all our international clone 2 strains but not in international clone 1 strains (Table 3 and Fig. S1). This finding was further confirmed by screening 12 additional isolates from the BioVersys strain collection (data not shown).

Two isoforms of eptA, eptA-1 and eptA-2 (GenBank accession numbers KC700024 and KC700023, respectively) have been described at different locations in the genome of various A. baumannii strains (17). Taking advantage of the different flanking regions, we designed primers able to discriminate eptA-1 from eptA-2 (oVT198/oVT199 and oVT201/oVT202, respectively) (Fig. 2A and Table S1). By genotyping the strain panel, we demonstrated that all strains that belong to the international clone 2 contained eptA-1 and four of them contained an additional copy of eptA-2 (Table 3 and Fig. S1). Interestingly, the PCR products obtained for eptA-2 in BV94 and eptA-1 in BV189 were approximately 1 kb larger than the expected fragment size. Sequencing of the PCR products identified the insertion element ISAbaI upstream of eptA-2 and eptA-1 in BV94 and BV189, respectively. The ISAbaI orientation enabled its strong promoter (P_out) to drive eptA overexpression as previously described for other antibiotic resistance determinants (Fig. 3) (25, 26).

Using probes specific for eptA or pmrC, we quantified their respective expression levels in our strains. eptA was 550- and 25-fold higher expressed in BV94 and BV189, respectively, than the homologous isoform pmrC in the control strain ATCC 17978 (which does not contain eptA) (Fig. 2B). This eptA overexpression was not altered in the ΔpmrA mutant strains, suggesting that eptA expression in both strains is independent of the PmrAB TCS. These data suggest that an ISAbaI-driven eptA overexpression may represent an alternative and PmrAB-independent colistin resistance mechanism in A. baumannii clinical strains.

To validate the hypothesis that ISAbaI-driven eptA overexpression confers colistin resistance in A. baumannii clinical isolates, we deleted eptA-1 in the clinical isolates BV189 and BV94 and determined MIC values. Indeed, BV189 (which carries ISAbaI upstream of eptA-1) became colistin susceptible upon eptA-1 deletion, indicating an essential role of ISAbaI-eptA-1 in conferring colistin resistance in this strain (Table 4). In contrast, BV94, carrying both eptA-1 and eptA-2 isoforms but carrying an ISAbaI insertion only upstream of eptA-2, remained colistin resistant after deletion of eptA-1, suggesting the key role of ISAbaI insertion for colistin resistance in A. baumannii. To further confirm the importance of ISAbaI, we constructed the double mutant BV94_{ΔeptA-1/ΔeptA-2} and evaluated its colistin susceptibility. A 4-fold MIC reduction was observed in the BV94_{ΔeptA-1/ΔeptA-2} mutant compared to BV94 and BV94_ΔeptA-1, indicating that eptA-2 with an upstream ISAbaI is involved in the colistin resistance mechanism of BV94. However, we were surprised to see that BV94_{ΔeptA-1/ΔeptA-2} remained resistant to colistin with a MIC of 16 μg/ml, indicating that there must be yet another colistin resistance mechanism present in this isolate.

We genotyped the BV94_{ΔeptA-1/ΔeptA-2} double mutant and confirmed the successful deletion of eptA-1 and eptA-2. However, we detected the presence of at least one additional eptA copy (eptA-3) in the double mutant (Fig. S2). We performed a fusion primer and nested integrated PCR experiment (FPNI-PCR) to amplify the genomic flanking regions of the additional eptA-3 variant (27). Sequencing revealed the ISAbaI insertion element and the gene ABK1_2603 present upstream and downstream of eptA-3, respectively (Fig. 3). This eptA-3 gene context in BV94 was identical to the eptA-2 gene context present in the A. baumannii strain 1656-2 (GenBank accession number NC_017162). However, further upstream there were marked differences. ISAbaI-eptA-3 in BV94 was adjacent to the gene ABK1_3144, while ISAbaI-eptA-2 in 1656-2 was adjacent to a different gene (Fig. 3). We could not determine the downstream flanking region of ISAbaI-eptA3 ABK1_2603 in multiple attempts. Nevertheless, the 9-bp target site duplications (TSD) created by ISAbaI transposition could not be identified directly outside the ISAbaI upstream eptA-3 (25). In contrast, TSD were present next to the right and left inverted repeats of the ISAbaI upstream eptA-1 and eptA-2, which is consistent with a single transposition event. These observations indicate that ISAbaI did not insert upstream eptA-3 in a single transposition event, and therefore, ISAbaI-eptA-3 in BV94 might be a result of an ISAbaI-eptA-2 cassette gene duplication, implying that the ISAbaI-eptA colistin resistance determinant is contained in a mobile genetic element.

We finally deleted the DNA fragment between ABK1_3144 and ABK1_3143 containing eptA-3 to confirm that ISAbaI-eptA-3 was responsible for the high residual colistin resistance in BV94_{ΔeptA-1/ΔeptA-2}. In addition, we performed PCR-based eptA genotyping on the resulting triple mutant BV94_{ΔeptA-1/ΔeptA-2/ΔeptA-3} to exclude the presence of yet another eptA copy (Fig. S2). The loss of all 3 eptA isoforms in BV94_{ΔeptA-1/ΔeptA-2/ΔeptA-3} rendered this triple mutant susceptible to colistin, indicating that colistin resistance in BV94 was entirely conferred by the overexpression of EptA-isoforms (Table 4).

We have shown that colistin resistance was mediated in 10 out of 12 analyzed clinical strains by PmrA-mediated overexpression of PmrC. In the remaining two strains, ISAbaI-driven EptA expression conferred colistin resistance. Taken together, in all tested clinical isolates colistin resistance was mediated by the overexpression of PetN transferases, suggesting that inhibition of these homologous enzymes with small molecules may have the potential to overcome colistin resistance in A. baumannii. Chin and colleagues recently suggested that the acetyl-galactosamine (GalNAc) deacetylase NaxD plays a role in colistin resistance in A. baumannii (28). In this report, the expression of NaxD, which was regulated by the PmrAB TCS, mediated galactosamine (GalN) addition to lipid A, conferring colistin resistance in A. baumannii. We performed additional experiments to exclude the possibility that the colistin resensitization observed in our clinical isolates after deletion of pmrA was based on a modulation of naxD expression and not pmrC expression. We first confirmed the PmrAB-controlled naxD expression based on qRT-PCR data for BV191 and BV191_ΔpmrA. BV191 has a mutated PmrB that likely triggers PmrA-mediated pmrC overexpression (Table 2). Similarly, naxD expression was 15-fold higher in the colistin-resistant strain BV191 than the susceptible strain ATCC 17978 (Fig. 4). In BV191_ΔpmrA, lacking the response regulator PmrA, pmrC and naxD overexpression was abolished, confirming that both genes were regulated by the PmrAB TCS. Notably, pmrC overexpression was 20-fold higher than naxD overexpression, suggesting a minor contribution of NaxD compared to PmrC in colistin resistance. To confirm the major role of PmrC in PmrA-mediated colistin resistance and to exclude that another PmrA-regulated gene, such as naxD, is involved in colistin resistance, we directly deleted the effector pmrC from the genome of BV191. The loss of PmrC rendered BV191 susceptible to colistin (MIC of 0.5 μg/ml) and resulted in a similar phenotype as in BV191_ΔpmrA (Table 2). In contrast, naxD was still 15-fold overexpressed in BV191_ΔpmrC (Fig. 4). This result suggests that overexpression of naxD is not sufficient to confer colistin resistance in BV191 and indicates that PmrC is the main effector of PmrA-mediated colistin-resistant A. baumannii strains, such as BV191.

The A. baumannii reference strain ATCC 17978 and 12 extensively drug-resistant A. baumannii clinical isolates from the BioVersys proprietary strain collection were used in this study. The microdilution method was used to determine MICs according to the CLSI guidelines (33). Multiple locus sequence type (MLST) was determined according to the Pasteur scheme using specific primers (source: http://pubmlst.org/abaumannii/) (20). Oligonucleotides used in this study are listed in Table S1 in the supplemental material.

Scarless deletions of pmrA, pmrC, and the eptA isoforms were performed using a two-step recombination method previously described (24).

DNA fragments corresponding to 700-bp up- and downstream genomic regions of the genes to be deleted were amplified by PCR and cloned in the multiple cloning site of the knockout platform pVT77. Oligonucleotides oVT49/oVT50 and oVT51/oVT52 were used to amplify the up- and downstream regions, respectively, of pmrA. The resulting DNA fragments were ligated and introduced into pVT77 previously digested by EcoRI and BamHI. Similarly, oligonucleotides oVT235/oVT236 and oVT237/oVT238 were used to amplify the flanking regions of eptA-1, and oligonucleotides oVT305/oVT306 and oVT307/oVT242 were used to amplify the flanking regions of eptA-2. The resulting DNA fragments for eptA-1 and eptA-2 were introduced into pVT77 previously digested by XhoI and XbaI using NEBuilder HiFi DNA assembly (New England Biolabs). For eptA-3 deletion, the genomic regions flanking the duplicated cassette were amplified using oVT390/oVT391 and oVT392/oVT393. The resulting DNA fragments were cloned into pVT77 previously digested with EcoRI and XbaI using NEBuilder HiFi DNA assembly. Last, the flanking regions of pmrC were amplified using oVT324/oVT325 and oVT326/oVT327, and the resulting DNA fragments were cloned into pVT77 previously digested with KpnI and PstI using NEBuilder HiFi DNA assembly.

The cloned knockout plasmids were transformed in E. coli conjugative strain MFDpir to proceed with the construction of markerless deletion in A. baumannii, as previously described (24). Briefly, after conjugation, genomic plasmid integration was selected on LB agar plates containing 100 μg/ml sodium tellurite. Clones were screened for up- or downstream integration by PCR using primer oVT8, which anneals on the plasmid, and oVT91, oVT243, oVT311, or oVT328, which anneals upstream of pmrA, eptA-1, eptA-2, or pmrC, respectively. For eptA-3, clones were screened using primers oVT8/oVT396 and oVT174/oVT397 for up- and downstream integration, respectively. Clones containing up- and downstream plasmid integrations were transferred on LB agar plates containing 1 mM isopropyl-β-d-1-thiogalactopyranoside and 200 μg/ml 3′-azido-3′-deoxythymidine to select for plasmid removal from the genome. Clones were screened for gene deletion and plasmid removal by PCR using primers oVT91/oVT92, oVT243/oVT244, oVT246/oVT311, oVT396/oVT397, and oVT328/oVT14 for pmrA, eptA-1, eptA-2, eptA-3, and pmrC, respectively. The genomic gene deletions were finally confirmed by DNA sequencing (Microsynth AG, Balgach, Switzerland).

A genomic DNA sequence including pmrA and pmrB was PCR amplified from all the strains of the panel using oVT91 and oCK292, and the PCR products were sent for sequencing (Microsynth AG, Balgach, Switzerland). The genotyping of eptA isoforms was performed by PCR using eptA-specific primers oVT152 and oVT153, which anneal on all eptA isoforms but not on pmrC (Fig. 2A). PCR using primers oVT198/oVT199 and oVT201/oVT202, which anneal on the flanking sides of eptA-1 and eptA-2, respectively, were used to discriminate between eptA isoforms (Fig. 2A).

Quantitative reverse transcription-PCR was performed as previously described (24). The specific expression of the PetN transferases encoded by pmrC and eptA was evaluated using oVT162/oVT163 and oVT164/oVT165 primers, respectively (Fig. 2A). The expression of naxD was evaluated using oVT314/oVT315 primers. Expression levels were normalized to that of the housekeeping gene rpoD using the comparative threshold cycle (ΔΔC_T) method. The expression of rpoD was evaluated using rpoD-qRT-F/rpoD-qRT-R primers.

Fusion primer and nested integrated PCR was performed as previously described (27). This method relies on a three-step PCR using arbitrary degenerated oligonucleotides fused to known adaptors and three sequence-specific oligonucleotides, which consist in our case of eptA-specific oligonucleotides. FPNI-PCR experiments were performed on the BV94_{ΔeptA-1/ΔeptA-2} mutant with two sets of three eptA-specific oligonucleotides, oligonucleotides oVT343, oVT344, and oVT345 and oligonucleotides oVT340, oVT341, and oVT342, to identify the sequence up- and downstream of the new eptA copy, respectively. The degenerated primers and the known adaptor primers were directly taken from the previously described method (27). Briefly, the first round of PCRs was performed using the degenerated primers and oVT343 for upstream identification and oVT340 for downstream identification. The second round of PCRs was performed with the first adaptor primer FSP1 and oVT344 for upstream identification and oVT341 for downstream identification. The last round of PCRs was performed with the second adaptor primer FSP2 and oVT345 for upstream identification and oVT342 for downstream identification. The brightest and most distinct PCR products obtained were sent for sequencing (Microsynth AG, Balgach, Switzerland).