A Method for generating marker-less gene deletions in multidrug-resistant Acinetobacter baumannii

Background Acinetobacter baumannii is an important nosocomial pathogen that has become increasingly resistant to multiple antibiotics. Genetic manipulation of MDR A. baumannii is useful especially for defining the contribution of each active efflux mechanism in multidrug resistance. Existing methods rely on the use of an antibiotic selection marker and are not suited for multiple gene deletions. Results A tellurite-resistant (sacB+, xylE+) suicide vector, pMo130-TelR, was created for deleting the adeFGH and adeIJK operons in two clinical MDR A. baumannii, DB and R2 from Singapore. Using a two-step selection, plasmid insertion recombinants (first-crossover) were selected for tellurite resistance and the deletion mutants (second-crossover) were then selected for loss of sacB. The DNA deletions were verified by PCR while loss of gene expression in the ΔadeFGH, ΔadeIJK and ΔadeFGHΔadeIJK deletion mutants was confirmed using qRT-PCR. The contribution of AdeFGH and AdeIJK pumps to MDR was defined by comparing antimicrobial susceptibilities of the isogenic mutants and the parental strains. The deletion of adeIJK produced no more than eight-fold increase in susceptibility to nalidixic acid, tetracycline, minocycline, tigecycline, clindamycin, trimethoprim and chloramphenicol, while the deletion of adeL-adeFGH operon alone had no impact on antimicrobial susceptibility. Dye accumulation assays using H33342 revealed increased dye retention in all deletion mutants, except for the R2ΔadeFGH mutant, where a decrease was observed. Increased accumulation of ethidium bromide was observed in the parental strains and all pump deletion mutants in the presence of efflux inhibitors. The efflux pump deletion mutants in this study revealed that only the AdeIJK, but not the AdeFGH RND pump, contributes to antimicrobial resistance and dye accumulation in MDR A. baumannii DB and R2. Conclusions The marker-less gene deletion method using pMo130-TelR is applicable for creating single and multiple gene deletions in MDR A. baumannii. The adeFGH and adeIJK operons were successfully deleted separately and together using this method and the impact of each efflux pump on antimicrobial resistance could be defined clearly.


Background
Acinetobacter baumannii, a non-fementing Gramnegative cocco-bacillus, is a frequent cause of nosocomial bloodstream infections and is associated with considerable morbidity and mortality, especially among patients in intensive care or with burns [1]. A. baumannii has become increasingly resistant to multiple antibiotics, including imipenem and meropenem, the carbapenems of choice for treating multidrug resistant (MDR) A. baumannii infections. The incidence of carbapenem-resistant A. baumannii in the United States and Europe is around 54% and 16%, respectively, while the incidence in the Asia/Pacific rim is about 80% [2]. A. baumannii possesses a variety of intrinsic and acquired resistance determinants, including β-lactamases, class D oxacillinases, aminoglycoside-modifying enzymes, outer membrane proteins and active efflux systems [3]. Among its intrinsic resistance determinants, overexpression of the chromosomally encoded active efflux systems of the resistance-nodulation and division (RND) family, such as AdeABC, AdeFGH and AdeIJK pumps, are a mechanism of resistance to a number of antibiotics [4].
The impact of RND pumps to antibiotic resistance in A. baumannii has been demonstrated by inactivating the genes that encode the efflux pumps and the method for gene inactivation involves insertion of an antibiotic resistance gene to select mutants [5][6][7]. Studies using mutants in which RND efflux pump genes have been inactivated have suggested significant overlap in antibiotics that are substrates of the A. baumannii pumps. For instance, derivatives of the MDR clinical isolate BM4454 in which adeABC was inactivated had increased susceptibility to the same antibiotics (fluoroquinolones, chloramphenicol, tetracycline, tigecycline and erythromycin) as inactivation of adeIJK in the same isolate [6]. When both adeABC and adeIJK were inactivated in BM4454, increased susceptibility to ticarcillin, previously not observed in the ΔadeABC mutant or the ΔadeIJK mutant, was seen [6]. Furthermore, overexpression of a pump gene did not always result in an increase in the MIC of the same antibiotics that had increased activity in the pump inactivated mutants. For example, inactivation of adeABC in the MDR clinical isolate BM4454 did not affect its susceptibility to imipenem, amikacin and cotrimoxazole, but overexpressing adeABC in a non-MDR clinical isolate BM4587 increased the MIC of these antibiotics [4]. Therefore, it is possible that inactivation of a gene by inserting an antibiotic-resistance gene may affect the antimicrobial susceptibility of the pump geneinactivated mutants, thus complicating the interpretation of the results.
To address this possibility and to define clearly the impact of each efflux pump on antibiotic resistance, we propose that genes encoding efflux pumps be deleted using a marker-less strategy first described by Hamad et al (2009) for Burkholderia spp. [8]. The suicide vector, pMo130 was modified to carry a tellurite resistance cassette, a non-antibiotic selection marker [9]. The A. baumannii isolates we have tested, including MDR isolates, were sensitive to tellurite and can be counterselected in LB medium containing 30-60 mg/L tellurite. Gene deletion by allelic replacement was selected using a modification of the two-step process described by Hamad et al (2009) [8]. In this study, the adeFGH and adeIJK operons were deleted separately and together in two MDR A. baumannii strains, DB and R2. The adeIJK deletion mutant showed increased susceptibility to nalidixic acid, chloramphenicol, trimethoprim, tetracycline, tigecycline, minocycline and clindamycin, but the deletion of adeL-adeFGH operon had no impact on antimicrobial susceptibility in the two MDR isolates. Genetic and gene expression analyses revealed that the allelic replacement in both MDR strains had occurred. The marker-less gene deletion method we describe is robust and, unlike the creation of mutants by inserting an antibiotic resistance gene, is suitable for deleting multiple genes in MDR A. baumannii.

Results
Deletion of the A. baumannii adeFGH and adeIJK operons To ensure reproducibility of the method, gene deletions were created for the adeFGH and adeIJK operons, separately and together, in two clinical MDR A. baumannii isolates, DB and R2. A suicide vector harboring a tellurite-resistance marker was first created by inserting a 3.26 kb XmaI-digested tellurite-resistance cassette from pwFRT-Tel R into the XmaI site of pMo130 to give pMo130-Tel R [8,10]. In addition to the tellurite-resistance marker, pMo130-Tel R also carries a kanamycin-resistance marker, the reporter gene xylE which converts pyrocathechol to a yellow-colored 2-hydroxymuconic semialdehyde, and a modified sacB gene [8]. Next, DNA fragments of approximately 1 kb upstream and 1 kb downstream of the target region to be deleted was ligated with linearized pMo130-TelR give pMo130-TelR-(Up/Down) ( Figure 1A).
To construct the suicide plasmid for deletion of adeFGH, a 1 kb DNA fragment located upstream of adeF was amplified from R2 genomic DNA using the primer pair: AdeGUp(Not1)F and AdeGUp(BamHI)R ( Figure 1B). The amplimer was digested using Not1 and BamHI and inserted into pMo130-Tel R , creating pMo130-Tel R -adeFGH(UP). Next, another 1 kb fragment located downstream of adeG was amplified using the primer pair: AdeGDwn(BamHI)F and AdeGDwn(Sph1)R and cut with BamHI and SphI, and inserted into pMo130-Tel R -adeFGH(Up), thus creating pMo130-Tel R -adeFGH(Up/Down) ( Figure 1B). The plasmid construct was first introduced in E. coli S17-1 and subsequently delivered into A. baumannii R2 and DB by biparental conjugation. A. baumannii transconjugants (first crossovers) were selected on LB agar containing 30 mg/L tellurite and 25 mg/L gentamicin. These telluriteresistant colonies which carry genomic insertion of pMo130-Tel R -adeFGH (Up/Down) produced yellow colonies when sprayed with 0.45 M pyrocathechol and a 2 kb amplimer corresponding to the size of the ligated Up and Down DNA fragments with the primer pair: AdeGUp(Not1)F and AdeGDwn(Sph1)R, but did not produce any amplimer with the outward-facing primer pair: pMo130Tel F and pMo130Tel R ( Figure 1A) (data not shown). A. baumannii R2 and DB harboring the inserted pMo130-Tel R -adeFGH (Up/Down) construct was cultured in LB broth containing 10% sucrose and passaged daily to select for deletion of adeFGH operon and loss of the sacB gene by a second cross-over and allelic  replacement. Such bacteria, which were white when sprayed with 0.45 M pyrocathechol and were susceptible to 30 mg/L tellurite, usually appeared after the second passage. If the desired gene deletion had occurred, PCR of genomic DNA from these bacteria would produce only a 2 kb amplimer with the primer pair AdeGUp(Not1)F and AdeGDwn(Sph1)R. The same genomic DNA would not give any amplimer using the primer pair: AdeG RTF and AdeG RTR which annealed to the DNA that has been deleted ( Figure 1B). The suicide plasmid for deleting the adeIJK operon was constructed as described above but by first ligating the 1 kb UP fragment and a 0.9 kb DOWN fragment flanking the deletion before inserting into the pMo130-TelR vector ( Figure 1C). The UP and DOWN fragments were amplified from R2 genomic DNA using the primer pairs, AdeJ(UP) PstI F and AdeJ(UP)BamHI R, and AdeJ (DWN)BamHI F and AdeJ(DWN)SphI R, respectively ( Figure 1C and Additional file 1: Table S1). The UP and DOWN fragments were digested with BamHI and ligated together in a 1:1 ratio. The ligated product was amplified using AdeJ(UP) PstI and AdeJ(DWN)SphI R to give a 1.9 kb amplimer which was then digested with PstI and SphI and ligated with pMo130-Tel R linearized with PstI and SphI to give pMo130-Tel R -adeJ(Up/Down). The plasmid construct was introduced into E. coli S17-1 and used for the two-step selection for deletion of the adeIJK operon as described above.

Verification of genomic deletions
Genomic deletions of the adeFGH and adeIJK operons in the mutants were verified by comparing the PCR amplimers obtained from the parental isolates and corresponding pump gene deletion mutants. For the pump gene deletions, PCR using primers flanking the deletion produced a 2-kb amplimer corresponding to the UP and DOWN fragments ( Figure 2, lanes 3,7,11,15,17,19,21 and 23) while a larger wild-type amplimer was obtained using genomic DNA from the parental isolates, R2 and DB ( Figure 2, lanes 1, 5, 9 and 13). For the ΔadeFGH constructs, the deletion was also confirmed using PCR primers that annealed to the deleted region in adeG, whereby a 474 bp amplimer was obtained using genomic DNA from parental isolates ( Figure 2, lanes 2 and 6), but no amplimer was obtained using genomic DNA from the ΔadeFGH deletion mutants ( Transcriptional analysis of the ΔadeFGH and ΔadeIJK deletion mutants RNA was extracted from parental strains and pump mutants cultured during mid-logarithmic growth in the absence of antibiotics. Analysis of the transcripts of the three major RND pumps in A. baumannii showed that the expression pattern of adeB, adeG and adeJ genes in both DB and R2 was similar ( Figure 3). In the absence of any antibiotics, adeIJK was the most highly expressed pump while the expression of adeFGH was the lowest. All three pumps were also about 4-fold more highly expressed in DB as compared to R2 ( Figure 3).
To confirm that the gene deletions had abolished the expression of the efflux pumps, the levels of transcripts of each gene in the adeFGH and adeIJK operons were measured in the deletion mutants and compared with the corresponding transcript levels in the parental strains. Both the DBΔadeFGH and R2ΔadeFGH mutants showed significant reduction (to ≤10%) in the transcript levels for adeF, adeG and adeH when compared to the parental strains ( Figure 4A). Although detectable, the level of adeL transcription in these mutants was also significantly reduced when compared to the adeL transcripts in the parental strains. This was because the genomic deletion had included the putative adeL promoter. Inactivation of adeG in both DBΔadeFGH and R2ΔadeFGH mutants was confirmed by the almost undetectable levels of adeG transcripts ( Figure 4A).
Successful inactivation of adeJ was also similarly confirmed by the absence of adeJ transcripts in the DBΔadeIJK and R2ΔadeIJK mutants ( Figure 4B). A small (See figure on previous page.) Figure 1 Strategy for deleting adeL-adeFGH and adeIJK operons in MDR A. baumannii DB and R2. Panel A, The upstream (UP) and downstream (DOWN) regions (approximately 1 kb) flanking the target genes was cloned into the suicide vector, pMo130-Tel R . pMo130-Tel R was constructed by inserting a 3.26 kb XmaI-digested tellurite-resistance cassette from pwFRT-Tel R into the XmaI site of pMo130. Recombinants obtained after first cross-over were selected for inheritance of tellurite-resistance and xylE + (yellow colonies). These recombinants also do not produce any amplimers with the primer pair pMo130Tel F and pMo130Tel R. During the second cross-over, mutants with gene deletion (1) were selected for loss of sacB by passaging the first cross-over recombinants in media containing sucrose. The second cross-over could also yield parental genotype (2). Deletion of the adeFGH operon (Panel B) and the adeIJK operon (Panel C) showing the positions of the respective UP and DOWN fragments flanking each deletion (striped and hatched boxes, respectively). The locations of the PCR primers used for amplifying the UP and DOWN fragments and for qRT-PCR analysis of gene expression are indicated by black arrows while P1, P2 and P3 (grey arrows) are the locations of predicted promoters for adeFGH operon, adeL, and adeIJK operon, respectively. quantity of adeI transcripts was udetectable in DBΔadeIJK and R2ΔadeIJK mutants, albeit at 56% and 31% of wild-type levels, respectively. This was due to the location of the adeI qRT-PCR primers within the UP fragment, i.e. within the 5' undeleted portion of the adeI gene ( Figure 1C).
Next, we tested the feasibility of our marker-less deletion strategy for creating isogenic mutants carrying a combination of pump gene deletions. We applied this strategy to delete adeIJK in the DBΔadeFGH and R2ΔadeFGH mutants to create DBΔadeFGHΔadeIJK and R2ΔadeFGHΔadeIJK mutants, respectively. As expected, the DBΔadeFGHΔ-adeIJK and R2ΔadeFGHΔadeIJK mutants showed significantly reduced expression of adeL, adeF, adeG, adeH, adeJ and adeK ( Figure 4C). Expression of adeI in DBΔadeFGHΔ-adeIJK and R2ΔadeFGHΔadeIJK mutants was reduced to 38% and 58% of DB and R2 levels, respectively.
Although adeL and the adeFGH operon were expressed in DB and R2, albeit at a lower level that adeB and adeJ, inactivation of adeFGH in both DB and R2 had minimal impact on the MDR phenotype of DB and R2 (Table 1). This is shown by the minimal change in antimicrobial susceptibility between the mutants that had only adeFGH inactivated (DBΔadeFGH and R2ΔadeFGH) and both adeFGH and adeIJK operons inactivated (DBΔadeFGH-ΔadeIJK and R2ΔadeFGHΔadeIJK) ( Table 1). The  DBΔadeFGHΔadeIJK and R2ΔadeFGHΔadeIJK mutants had the same antimicrobial susceptibility as DBΔadeIJK and R2ΔadeIJK mutants, respectively ( Table 1).

Growth of pump deletion mutants
The optical density at 600 nm measurements of liquid cultures of the parental strains and pump deletion mutants revealed no significant difference in growth kinetics (data not shown). Growth kinetics in the presence of sub-MIC concentrations of EIs were also carried out to simulate conditions in the H33342 accumulation assay (see below) and to ensure no inhibition of growth over a two-hour time period during the assay. These experiments showed that 30 mg/L CCCP and 50 mg/L PAβN did not restrict growth of R2 (data not shown). Viability of all strains was unaffected by H33342 concentrations of 2.5 μM, 5 μM and 10 μM (data not shown).

Accumulation of H33342 by efflux pump gene deletion mutants
Compared with the parental isolate, R2, there was a significant 0.8 fold change in the level of H33342 accumulated at steady state in R2ΔadeFGH ( Figure 5A). Compared with the parental isolate, accumulation of H33342 was significantly increased in R2ΔadeIJK and R2ΔadeFGHΔadeIJK, with a fold change of 1.18 and 1.16 respectively. The mutants created in isolate DB showed a different pattern of accumulation ( Figure 5B). The level of H33342 accumulated at steady state was significantly higher in all three mutants, DBΔadeFGH, DBΔadeIJK and DBΔadeFGHΔadeIJK, compared with the parental strain, with fold-changes of 1.13, 1.26 and 1.22, respectively. Addition of CCCP caused a significant increase in the steady state accumulation of H33342 by all strains ( Table 2). In the R2 isolate and mutants, this increase was most pronounced in R2ΔadeFGH, with a fold increase of 1.46 observed ( Table 2). The parental isolate showed a smaller fold increase of 1.31. R2ΔadeIJK and R2ΔadeFGHΔadeIJK showed the smallest fold changes of 1.09 and 1.10, respectively. In the DB parent and mutant strain, the parental strain DB showed the highest fold increase of 1.51 after addition of CCCP, with the increase in DBΔadeFGH slightly less, at 1.27 (Table 2). DBΔadeIJK and DBΔadeFGHΔadeIJK again showed the smallest fold changes of 1.16 and 1.19, respectively. Addition of PAβN also caused a significant increase in accumulation in all strains ( Table 2). This increase was of a similar fold in the parental strains, R2 and DB, and their mutants.

Accumulation of ethidium bromide by efflux pump gene deletion mutants
It has been shown previously that H33342 and ethidium bromide are substrates of efflux pumps [11]. Therefore, accumulation of ethidium bromide was also measured. Compared with the parental isolate, the fold-change in the steady state levels of ethidium bromide accumulated in efflux pump mutants showed the same pattern as that produced with the H33342 accumulation assay, with levels in R2ΔadeFGH significantly lower than in parental isolate R2 ( Figure 6A), and R2ΔadeIJK and R2ΔadeFGHΔadeIJK accumulating significantly higher levels. Efflux pump mutants DBΔadeFGH, DBΔadeIJK and DBΔadeFGHΔadeIJK accumulated higher levels of ethidium bromide than the parental isolate, DB ( Figure 6C). Addition of both CCCP and PAβN produced a significant increase in the level of ethidium bromide accumulated at steady state in both parental isolates and their mutants and the effect was similar to that seen with H33342.

Discussion
The two-step deletion strategy we have described was used for creating unmarked deletions in the adeFGH and adeIJK efflux pump operons, separately and together, in two clinical MDR A. baumannii isolates. It is an improvement from the simple method for gene replacement in A. baumannii described by Aranda et al (2010) that uses an antibiotic resistance cassette [12]. To adapt the method first described for use in MDR A. baumannii, we introduced a tellurite resistance cassette into the pMo130 suicide vector created by Hamad et al (2009) to facilitate the selection of MDR A. baumannii transconjugants with the suicide plasmid inserted into the genome, i.e. first crossover products [8]. It was helpful to first ascertain the growth inhibitory concentration of tellurite for the parental A. baumannii strain so the number of transconjugants (first crossover) that are false positives can be minimized by using a suitable tellurite concentration. Passaging the first crossover recombinants in media containing sucrose provided the selection pressure for loss of the plasmid by a second crossover, leading to the formation of white colonies when sprayed with pyrocathechol.
The main advantage of this method, which does not use antibiotic selection for the gene deletion mutants, is its application for generating multiple gene deletions in a single strain as we have demonstrated by creating DBΔadeFGHΔadeIJK and R2ΔadeFGHΔadeIJK mutants. This is particularly important because the majority of A. baumannii strains are MDR or extensively drugresistant (XDR). Other than the MDR strains described in this study, we have also tested this method in a  carbapenem-susceptible A. baumannii strain (data not shown). Un-marked deletion mutants are especially useful for ascertaining the contribution of each efflux pump to MDR as the presence of antibiotic resistance cassettes in the mutants may complicate the interpretation of antimicrobial susceptibility. We believe that the markerless method would allow the impact of each efflux system on antimicrobial resistance to be clearly defined. The contribution of efflux pumps to antibiotic resistance in A. baumannii has been demonstrated with mutants created by gene inactivation/deletion or by creating spontaneous efflux pump overexpressing mutants via selection on antibiotic gradients, but with some inconsistencies in antimicrobial susceptibilities depending on how the genes were inactivated [5]. For example, inactivation of adeABC in a clinical MDR isolate by insertion of a ticarcillin-resistance gene conferred increased susceptibility to aminoglycosides, β-lactams, fluoroquinolones, chloramphenicol, tetracycline, macrolides and trimethoprim [7]. However when adeABC was deleted and an apramycin resistance cassette was inserted in the same MDR isolate, the ΔadeABC mutant showed increased susceptibility to fluoroquinolones, chloramphenicol, tetracycline, tigecycline and macrolides but no change in susceptibility to aminoglycosides, trimethoprim and β-lactams [4,6]. We hypothesized that the antibiotic resistance gene used in the creation of pump gene mutants complicated the interpretation of antimicrobial susceptibility data and hence which agents were putative substrates of each A. baumannii efflux pump.
When adeIJK was inactivated using the marker-less method, the MDR isolates became more susceptible to nalidixic acid, chloramphenicol, clindamycin, tetracycline, minocycline, tigecycline and trimethoprim. It is interesting to note that the DBΔadeIJK and R2ΔadeIJK mutants showed increased susceptibility to nalidixic acid without affecting susceptibility to ciprofloxacin, suggesting AdeIJK may be specific for quinolones but not fluoroquinolones. We also noted that, although the AdeIJK pump confers increased resistance to exactly the same antibiotics in both DB and R2, the host genotype had an influence on the magnitude of resistance to each antibiotic. The successful creation of adeFGH and adeIJK gene deletions, separately and together, in two   MDR A. baumannii isolates demonstrates the robustness of the method and its application across different MDR A. baumannii isolates. The antibiotic substrates revealed with our mutants are in general agreement with those described by Damier-Piolle et al (2008) in which adeIJK was inactivated in an MDR isolate by gene deletion together with insertion of a kanamycin-resistance cassette [6]. However, in our study the DBΔadeIJK and R2ΔadeIJK mutants were also more susceptible to trimethoprim, but not to β-lactams. It should be noted that differences between these studies may be due to the presence of different antibiotic resistance genes on the host genome, e.g. R2 had bla OXA-23 like, bla OXA-51 like genes, bla TEM , bla OXA and bla ADC that confer β-lactam resistance. The MICs of antibiotics for double mutants R2ΔadeFGHΔadeIJK and DBΔadeFGHΔadeIJK were the same as for the corresponding single mutants R2ΔadeIJK and DBΔadeIJK. This was expected, as a single deletion of adeFGH had minimal effect on MICs of antibiotics in either strain. Deletion of adeIJK and of adeFGH in combination with adeIJK in both R2 and DB resulted in a significant increase in steady state accumulation levels of both H33342 and ethidium bromide. This infers reduced efflux in these strains, presumably as a consequence of the removal of the efflux pump AdeIJK. Addition of CCCP to ΔadeIJK and ΔadeFGHΔadeIJK mutants of both R2 and DB significantly increased the steady state accumulation of H33342, suggesting that, despite lacking AdeIJK, these mutants still possess proton gradient dependent efflux activity as a result of another pump system. The addition of CCCP and PAβN had the same effect on the accumulation of ethidium bromide. However, the increase in accumulation observed in these mutants was not as high as that seen with the parental isolates and the adeFGH deletion mutants, supporting the previous finding that efflux is reduced in mutants lacking adeIJK.
In our study, the deletion of the adeFGH operon also removed the putative adeL promoter, resulting in reduced expression of adeL. However, both the inactivation of the adeFGH operon and reduced expression of adeL had very little impact on antimicrobial susceptibility when compared to the parental isolates which expressed both adeL and adeFGH operon. This was also true when the antimicrobial susceptibilities of DB and R2 mutants that had both the adeIJK and adeFGH operons deleted were compared with the DB and R2 mutants that had only the adeIJK operon inactivated. In all instances, inactivation of adeFGH had minimal impact on antimicrobial susceptibility when compared to isogenic isolates with functional AdeFGH, indicating that expression of adeL and adeFGH operon was not involved in the multidrug resistance of these clinical MDR isolates. These findings are different to those of Coyne et al, who showed that overexpressing adeFGH in an MDR strain lacking AdeABC and AdeIJK increased the MICs of several antibiotics including chloramphenicol, clindamycin, tetracycline, minocycline, tigecycline, norfloxacin, ciprofloxacin and cotrimoxazole [5]. In that study, the adeFGH operon was overexpressed in a spontaneous drug-resistant ΔadeABCΔadeIJK mutant selected on norfloxacin and chloramphenicol gradient plates. The adeFGH operon was then deleted and a streptomycin-spectinomycin resistance cassette was also inserted to select for the deletion mutant. It is plausible that the process of selecting spontaneous drug-resistant mutants on chloramphenicol and norfloxacin gradients may have created gene duplication and amplification or a mutation in another efflux pump regulator was selected, especially since the inhibition of DNA gyrase by fluoroquinolones induces the SOS response [13]. It is also possible that under the experimental conditions whereby the adeFGH operon was induced and significantly overexpressed, an increase in resistance to chloramphenicol, trimethoprim and clindamycin may be observed. However, further studies are needed to determine the significance and relevance of such conditions to the clinical environment [14]. It was proposed that mutation in adeL results in overexpression of adeFGH operon and hence an increase in antibiotic resistance [5]. It is also possible that mutation in adeL, a LysR-type transcriptional regulator, may affect expression of another efflux pump gene/s or antibiotic resistance determinant. However, in the DBΔadeFGH and R2ΔadeFGH mutants created in the present study, adeL expression was impaired yet there was minimal change in the MICs of antibiotics for the mutants when compared with the parental isolates. This ruled out the possibility that the MDR phenotype of DB and R2 might be due to a mutation in adeL which had an effect on the expression of another efflux pump(s) other than the adeFGH operon. Our data suggests that the activities of the AdeL transcriptional regulator and AdeFGH pump do not contribute to multidrug resistance in DB and R2.
Despite the minimal change in MICs of antibiotics compared with the parental isolate, R2ΔadeFGH showed a significant decrease in accumulation of both H33342 and ethidium bromide, inferring increased efflux in this strain. This may be due to increased expression of another efflux system in order to compensate for the loss of AdeFGH. This could also explain the lack of change in MIC seen with deletion of adeFGH. Previous work in Salmonella enterica serovar Typhimurium has shown that deletion of RND efflux pump genes can lead to compensatory altered expression of other efflux pump genes. For example, deletion of acrB in SL1344 resulted in a 7.9 fold increase in the expression of acrF [15]. An increase in accumulation of H33342 and ethidium bromide was seen in DBΔadeFGH, inferring reduced efflux in this strain, however this difference did not translate into a change in MIC. Addition of CCCP and PAβN had a greater effect on accumulation of H33342 and ethidium bromide in this efflux pump mutant than in mutants lacking adeIJK. A greater fold change in accumulation was seen with both R2ΔadeFGH and DBΔadeFGH than other efflux pump mutants, suggesting that efflux activity is higher in these mutants.
Using the marker-less deletion method, we have demonstrated that AdeFGH and AdeIJK are independent efflux pumps with no common antibiotic substrates. While both adeFGH and adeIJK operons are expressed in MDR A. baumannii, only the expression of adeIJK contributed to increased resistance to nalidixic acid, chloramphenicol, clindamycin, tetracycline, minocycline, tigecycline and trimethoprim. Expression of adeFGH was not the cause of resistance in the clinical isolates of MDR A. baumannii, DB and R2.

Conclusions
The marker-less gene deletion method we have described is useful for creating gene deletions in MDR A. baumannii. Deletions of the adeFGH and adeIJK efflux pump operons, separately and together, were created in two clinical MDR A. baumannii isolates to demonstrate the robustness of the method. Even though both adeFGH and adeIJK operons are expressed in MDR A. baumannii, only the expression of adeIJK contributed to increased resistance to nalidixic acid, chloramphenicol, clindamycin, tetracycline, minocycline, tigecycline and trimethoprim. Expression of adeFGH was not the cause of resistance in the clinical isolates of MDR A. baumannii, DB and R2. This method allows the impact of each efflux system on antimicrobial resistance to be clearly defined.

Bacterial strains, plasmids and culture conditions
Bacterial strains and plasmids used in this study are listed in Table 3. Acinetobacter baumannii R2 (TTSH6013 654325/06) and DB (DB15354/07) were clinical isolates from a collection by the Network for Antimicrobial Resistance Surveillance, Singapore. According to the interim standard definitions for acquired resistance, both DB and R2 are classified as MDR as they are non-susceptible to ≥1 agent in ≥3 antimicrobial categories (aminoglycosides, fluoroquinolones, carbapenems, tetracycline, extended spectrum cephalosporins, folate pathway inhibitors) [17]. DB and R2 carry and express bla OXA-23-like and bla OXA-51-like , do not carry bla OXA-24-like and bla OXA-58-like (data not shown). A. baumannii and E. coli were cultured under aerobic conditions at 37°C in Luria-Bertani Miller (LB) agar or LB broth (Becton Dickinson, Cockeysville, U.S.A.). Antibiotics used were at the following concentrations for E. coli: kanamycin, 10 mg/L; tellurite 6 mg/L; and for A. baumannii: tellurite, 30 mg/L.

DNA manipulations
Bacterial genomic DNA was extracted using a rapid procedure described by Pitcher et al [18]. Plasmid DNA was extracted using GeneAid Hi-Speed Plasmid Mini kit (GeneAid, Taiwan). Standard PCR amplifications were performed with Biotools DNA polymerase (Biotools, Spain). All primers used for PCR were synthesized by 1 st Base Singapore and are listed in Additional file 1: Table  S1. Electrocompetent cells were prepared from 6 ml overnight bacterial culture according to the procedure described by Choi et al (2005) [19]. Electroporation was carried out by placing 100 μl electrocompetent cells and 3 μl plasmid DNA in a sterile cuvette (0.1 cm electrode gap, Bio-Rad) and pulsed at 1.8 V using settings for bacteria in a Bio-Rad MicroPulser.
The plasmid, pwFRT-Tel R , was digested with XmaI and the 3.265 kb fragment carrying the tellurite-resistance cassette was isolated and ligated with XmaI-linearized pMo130 to produce the suicide plasmid, pMo130-Tel R . The orientation of the tellurite-resistance cassette insert shown in Figure 1A was ascertained by digesting the plasmid with Xho1 and BamHI which gave a 4.161 kb and a 5.231 kb band. An insertion of the tellurite-resistance cassette into pMo130-Tel R in the opposite orientation would have produced two bands of 1.150 kb and 8.242 kb.

Conjugative transfer
E. coli S17-1 donor strain harboring the respective pMo130-TelR-(Up/Down) constructs and the A. baumannii recipient strains were cultured overnight at 37°C in 2 ml LB (supplemented with kanamycin for the donor E. coli strain). Aliquots of 0.2 ml each of donor and recipient cells were added to a microfuge tube containing 1.2 ml of LB and washed twice with 2 ml LB each time. The cells were then suspended in 30 μl LB medium and added on to a sterile 0.45 μm cellulose nitrate filter paper (Sartorius Stedim, NY, U.S.A.) on LB agar and incubated at 30°C for 16 h. The cells were washed off from the filter by adding 0.4 ml of 0.9% NaCl. Aliquots of 0.1 ml were plated onto LB agar containing tellurite (30 mg/L) and gentamicin (25 mg/L) and incubated at 37°C for at least 16 h. Gentamicin was added for counter-selection against the donor cells.

RNA analysis and quantitative real-time PCR (qRT-PCR)
RNA was extracted from mid-log phase bacteria prepared by inoculating 10 ml Luria-Bertani (LB) broth Miller (1 st BASE Pte Ltd, Singapore) with an overnight culture (1:50) and incubating at 37°C, with shaking at 120 rpm, until OD 600 = 1.0. Triplicates of culture volumes containing two OD 600 units (~2x10 9 cells) were centrifuged at 3,000 g for 10 min to harvest the cells. The cells were lysed by adding 1 mL of TRIzol ® (Invitrogen, Carlsbad, CA) to the cell pellet and RNA was extracted according to the manufacturer's protocol. Contaminating DNA was removed by treating the RNA sample with Ambion ® TURBO™ DNase (Invitrogen) and cDNA was synthesized using random hexamer primers and TaqMan ® Reverse Transcription Reagents (Invitrogen) according to the manufacturer's protocol.
Primers for qRT-PCR of efflux pumps genes were designed using PrimerQuest (Integrated DNA Technologies, Coralville, IA) and are listed in Additional file 1: Table  S2 (Supplementary files). qRT-PCR was performed using KAPA SYBR ® FAST Universal 2X qPCR Master Mix (Kapa Biosystems Inc., Woburn, MA) using 1X ROX (High) reference dye, 500 nm primers and~10 ng cDNA in a total volume of 20 μL and the transcripts were detected using Applied Biosystems 7300 Real-Time PCR system (Applied Biosystems ® , Carlsbad, CA). 16S rRNA was used for normalization of the qRT-PCR gene transcripts. qRT-PCR was performed twice for each of the triplicate RNA extracts. Data from each quantitative run was exported from the 7300 System software and analysed using 2 -ΔΔCt calculations [20].

Antimicrobial susceptibility testing
Minimum inhibitory concentrations (MICs) of antibiotics were determined using the agar doubling dilution method according to BSAC standard methodology [21]. MICs of imipenem and meropenem were determined by E-test (Biomerieux, Hampshire, UK).

Measurement of growth kinetics
Bacterial strains were grown with aeration in LB broth at 37°C overnight. Bacterial cultures were diluted 1:100 in sterile Luria Bertani (LB) broth and 100 μl of this suspension was added to each well of a clear 96 well microtitre tray. Optical density (OD) at an absorbance of 600 nm was measured over 16 hours in a BMG FLUOstar Optima (BMG, UK) at 37°C. The BMG FLUOstar is sensitive to an OD600 of between 0.0 and 4.0 and reproducibility is ±0.010 for the OD range of 0.0-2.0 (www.bmglabtech. com). Each experiment included three biological replicates and three technical replicates of each bacterial strain. Differences in generation times and final OD at 600 nm were calculated using a Student's t-test. P values ≤0.05 were considered as significant.
For assessment of toxicity of EIs and H33342, bacterial strains were grown with aeration in LB broth at 37°C overnight. A 4% inoculum (120 μl in 3 ml) of bacterial Invitrogen S17-1 Genotype: recA pro hsdR RP4-2-Tc::Mu-Km::Tn7, Gm S [16] Plasmids pMo130 Suicide plasmid, xylE + , sacB + , Km R [8] pwFRT-Tel R Donor of tellurite resistance cassette [10] pMo130-Tel R pMo130 plasmid containing 3.26 kb XmaI-digested telluriteresistance cassette from pwFRT-Tel R This study pMo130-Tel R -P8(UP/DWN) pMo130-Tel R containing a 1 kb UP fragment (promoterless adeL) and 1 kb DOWN fragment (3' partial adeH) This study pMo130-Tel R -adeJ(Up/Down) pMo130-Tel R containing a 1 kb UP fragment (5' partial adeI) and 0.9 kb DOWN fragment (3' partial adeK) This study culture was added to fresh LB broth. This suspension was incubated with aeration at 37°C until the culture reached an OD at 600 nm of 0.6 (= 10 8 cfu/ml). Cells were harvested by centrifugation at 2200 g for 10 min at room temperature and resuspended in 3 ml sterile LB broth at room temperature. The OD at 600 nm of the suspension was measured and adjusted to 0.5 to standardize the number of bacterial cells in each culture and to simulate the conditions used in the H33342 accumulation assay. The bacterial suspension (196 μl) was added to each well of a clear 96 well microtitre tray, along with 4 μl of EI and 20 μl H33342 at the required concentrations (see Results). OD at an absorbance of 600 nm was measured over 16 hours in the BMG FLUOstar OPTIMA (BMG, UK) at 37°C. Each experiment included three biological replicates and three technical replicates of each bacterial strain. Differences in generation times and final OD at 600 nm were calculated using a Student's t-test. P values ≤0.05 were considered as significant.

H33342 bis-benzamide accumulation in efflux pump deletion mutants
Accumulation of H33342 was measured and data analysed as described previously [11]. The level at which maximum fluorescence was reached and remained unchanged within the time period of the assay was taken as the steady state accumulation level. The fold change in fluorescence of mutants compared to the parental clinical isolate in the presence and absence of efflux pump inhibitors (EI) was calculated. Student's t-tests were carried out to compare the accumulation of H33342 by the mutant with the parental strain, R2; P values <0.05 were taken as significant. Each assay was repeated 3 times with 3 biological replicates.

Ethidium bromide accumulation in efflux pump deletion mutants
Ethidium bromide assays were carried out in the same way as the H33342 accumulation assay, except that cultures were resuspended in 1 M sodium phosphate buffer with 5% glucose. A 1 mM ethidium bromide stock solution was prepared and 20 μl was injected to give a final concentration of 0.1 mM in the assay. Fluorescence was measured over 117 minutes at excitation and emission wavelengths of 530 nm and 600 nm, respectively, in a FLUOstar OPTIMA.

Additional file
Additional file 1: Table S1. Description of primers used for PCR and DNA sequencing. Table S2. List of primers used for quantitative real-time PCR.
Authors' contributions IA and PS jointly constructed the adeFGH and adeIJK deletion mutants and carried out the molecular genetic studies. GER performed the dye accumulation antimicrobial susceptibility assays. THK provided the MDR A. baumannii isolates, characterized the bla OXA sequences in DB and R2. KLC conceived the study. LJP and KLC participated in the design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.