ABSTRACT
Burkholderia comprises species that are significant biothreat agents and common contaminants of pharmaceutical production facilities. Their extreme antibiotic resistance affects all classes of antibiotics, including polycationic polymyxins and aminoglycosides. The major underlying mechanism is the presence of two permeability barriers, the outer membrane with modified lipid A moieties and active drug efflux pumps. The two barriers are thought to be mechanistically independent and act synergistically to reduce the intracellular concentrations of antibiotics. In this study, we analyzed the interplay between active efflux pumps and the permeability barrier of the outer membrane in Burkholderia thailandensis. We found that three efflux pumps, AmrAB-OprA, BpeEF-OprC, and BpeAB-OprB, of B. thailandensis are expressed under standard laboratory conditions and provide protection against multiple antibiotics, including polycationic polymyxins. Our results further suggest that the inactivation of AmrAB-OprA or BpeAB-OprB potentiates the antibacterial activities of antibiotics not only by reducing their efflux, but also by increasing their uptake into cells. Mass spectrometry analyses showed that in efflux-deficient B. thailandensis cells, lipid A species modified with 4-amino-4-deoxy-l-aminoarabinose are significantly less abundant than in the parent strain. Taken together, our results suggest that changes in the outer membrane permeability due to alterations in lipid A structure could be contributing factors in antibiotic hypersusceptibilities of B. thailandensis cells lacking AmrAB-OprA and BpeAB-OprB efflux pumps.
INTRODUCTION
Burkholderia spp. are extraordinarily resistant to various antibiotics, including polycationic peptides and aminoglycosides (1). As a result of this resistance, infections in cystic fibrosis patients caused by bacteria belonging to Burkholderia cepacia complex (Bcc) and melioidosis caused by Burkholderia pseudomallei complex (Bpc) species present major challenges in clinics (2, 3). Burkholderia thailandensis is a close less-virulent relative of B. pseudomallei, yet this species also can cause infections in certain patients (4).
The high levels of antibiotic resistance in Burkholderia spp. are enabled by a synergistic interplay between active efflux pumps and the low permeability barrier of the outer membrane (OM) with unusual composition (5). The lipid A moiety, the membrane anchor of lipopolysaccharides (LPS) present in the OM of Burkholderia spp., is tetra- and penta-acylated with long C14 and C16 fatty acid chains, some of which are hydroxylated at the C-2 position (6, 7). In addition, unlike in other Gram-negative bacteria, the phosphoryl residues in the headgroup of lipid A and the inner core of LPS of Burkholderia spp. are constitutively augmented by positively charged 4-amino-4-deoxy-l-aminoarabinose (Ara4N) residues (6, 7). Similar Ara4N modifications of LPS are found in other Gram-negative bacteria as well, but these are triggered by the conditions of Mg2+ deficiency and acidic pH to which bacteria are exposed during host invasions and infections (8–10). Interestingly, the Ara4N modification of Burkholderia spp. is essential for bacterial growth, as demonstrated by genetic and bioinformatics studies (11, 12).
In all Gram-negative bacteria, the low permeability barrier of the OM is reinforced by multidrug efflux pumps (5). The genomes of Burkholderia spp. encode multiple efflux pumps (13–16). The resistance-nodulation-division (RND) efflux pumps are the major contributors to intrinsic antibiotic resistance in these and other Gram-negative bacteria. These pumps are composed of three components that span the two membranes and the periplasm. The major RND pump of B. thailandensis and B. pseudomallei is AmrAB-OprA, which is a homolog of the MexXY-OprM from Pseudomonas aeruginosa (17). The mutational inactivation of this pump results in a significant decrease in MICs for a broad range of antibiotics, including aminoglycosides. Similarly, the inactivation of BpeAB-OprB, a P. aeruginosa MexAB-OprM homolog, increases susceptibility to multiple antibiotics, though the spectrum is notably narrower than for AmrAB-OprA (18). In the absence of regulatory mutations, BpeEF-OprC appears to have narrow substrate specificity and confers widespread trimethoprim resistance in clinical and environmental isolates of B. pseudomallei (16, 19). In mutants overexpressing this pump, its substrate spectrum is similar to its MexEF-OprN homolog in P. aeruginosa and includes chloramphenicol, fluoroquinolones, tetracyclines, sulfamethoxazole, and trimethoprim (16, 19–21). Like MexEF-OprN, BpeEF-OprC is positively regulated by at least two transcriptional regulators (16, 19, 20).
The low permeability of the OM and active drug efflux are synergistic in their contributions to antibiotic activities. Two kinetic constants define the accumulation and hence the activities of antibiotics in Gram-negative bacteria. The barrier constant is the ratio of the rate of passive influx of antibiotics into the cells to the rate of active antibiotic efflux, whereas the efflux constant is a thermodynamic measure of active efflux efficiency (22). In addition, depending on their biochemical properties, the coexpression of different efflux pumps could lead to either additive or multiplicative effects on antibiotic accumulation and, therefore, antibiotic activities (23, 24). The contributions of pumps that move substrates across the same membrane, either the inner or the outer, will be additive to each other, whereas the effect of efflux across two different membranes is expected to be multiplicative and therefore synergistic. Therefore, analyses of antibiotic activities in mutants lacking single or multiple expressed pumps could highlight the differences and similarities between the pumps. In addition, the separation of contributions of the OM barrier and active efflux could be achieved by controlled hyperporination of the OM, in which the siderophore uptake transporter OrbA is modified by removal of the N-terminal plug and the external loop to create a large OM pore (5, 25). Using this approach, we previously found that B. cepacia ATCC 25416 (BcWT) and B. thailandensis E264 (BtWT) differ dramatically from each other and other species in their permeability barriers and efflux capacities. Hyperporination of the OM of BcWT alone increased the influx of antibiotics sufficiently to overwhelm the efflux pumps. As a result, these cells became hypersusceptible to a broad range of antibiotics despite unchanged active efflux. In contrast, influx remained very slow in the hyperporinated BtWT cells, and the cells remained protected against antibiotics. Only the efflux-deficient cells lacking the two efflux pumps AmrAB-OprA and BpeAB-OprB were hypersusceptible to antibiotics. Furthermore, hyperporination of the OM and inactivation of efflux in B. thailandensis resulted in unexpected species-specific changes in susceptibilities to antibiotics, including aminoglycosides and polymyxin B. These findings led us to conclude that in B. thailandensis the efflux pumps are needed to maintain the proper structure of the OM.
In this study, we analyzed the contributions of the OM barrier and three major RND pumps of B. thailandensis in antibiotic activities and the cell envelope permeability. We report that inactivation of either AmrAB-OprA or BpeAB-OprB increases susceptibility to antibiotics and affects modifications of lipid A in the OM. We conclude that the functions of AmrAB-OprA and BpeAB-OprB are integrated with the structure of the OM of B. thailandensis.
RESULTS
Antibiotic susceptibility of efflux mutants and their hyperporinated derivatives cannot be explained by independent contributions of active efflux and the outer membrane barrier.Previous studies identified three RND-type efflux pumps of B. pseudomallei, AmrAB-OprA, BpeAB-OprB, and BpeEF-OprC, that contribute to clinical antibiotic resistance (16–18). In agreement with this, we found that the inactivation of either one of these pumps in B. thailandensis BtWT changed the susceptibility profile of the cells in a pump-specific manner (Tables 1 and S1). The activities of the majority of tested antibiotics remained unchanged in the Bt34 [Δ(bpeAB-oprB)] and Bt40 [Δ(bpeEF-OprC)] strains. In contrast, inactivation of AmrAB-OprA in Bt36 affected a large number of antibiotics. We also constructed double-knockout mutants lacking AmrAB-OprA either in combination with BpeAB-OprB (Bt38) or with BpeEF-OprC (Bt42) and hyperporinated the OMs of all efflux-deficient cells. The gene encoding Pore was either inserted into a single glmS-linked attTn7 site (low expression [PoreL]) or into both glmS1- and glmS2-linked sites (higher expression [PoreH]) of B. thailandensis (26). As a result, the constructed strains differed in their susceptibilities to antibiotics (Table 1), with the higher potentiation of antibiotic activities in the PoreH than in PoreL strains (Table S2). Hence, the outer membrane barriers in these strains have different permeability properties.
MICs of antibiotics in B. thailandensis strains with various efflux capacities and permeabilities of the outer membrane
We next measured the MICs of various antibiotics and applied clustering analysis to ratios of MICs in efflux-deficient strains with and without hyperporination to MICs in the respective BtWT strains (efflux ratios) and MICs in strains with and without hyperporination of the OM (OM ratios). In addition, the MIC ratios of BtWT to hyperporinated efflux-deficient strains reflect the combined contributions of the entire barrier in activities of antibiotics (barrier ratios).
The MIC ratios form four clusters (Fig. 1). The first cluster comprises OM and efflux ratios with the smallest (1- to 4-fold) values. The low effect of hyperporination or efflux inactivation on MICs could be expected when changes in either one of the two barriers do not affect the antibiotic permeation significantly. The OM ratios of hyperporinated BtWT and Bt40 [Δ(bpeEF-OprC)] cells, the efflux ratios of Bt34 [Δ(bpeAB-oprB)] and Bt40 strains, and the OM ratios of all other strains producing PoreL belong to this group. Apparently, these changes in permeability barriers were insufficient to notably reduce the permeation barrier for analyzed antibiotics.
Clustering analysis of MIC ratios. Hierarchical clustering was done using Euclidean distances between logarithms of measured MIC ratios. Ratios of MICs are color coded and visually separated into four clusters numbered at the top. Antibiotics form three groups indicated on the left. BAC, bacitracin; ZEO, zeocin; RIF, rifampin; VAN, vancomycin; NAL, nalidixic acid; LVX, levofloxacin; CIP, ciprofloxacin; AZM, azithromycin; ERY, erythromycin; TET, tetracycline; DOX, doxycycline; AMP, ampicillin; CAR, carbenicillin; CLO, cloxacillin; NOV, novobiocin; CHL, chloramphenicol; TRI, triclosan; HT, Hoechst 33342; MEM, meropenem; TMP, trimethoprim; GEN, gentamicin; AMK, amikacin; KAN, kanamycin; TOB, tobramycin.
The second cluster comprises the efflux ratios of single- and double-knockout strains deficient in AmrAB-OprA. The antibiotics that are most strongly affected by inactivation of AmrAB-OprA include macrolides, zeocin, and aminoglycosides. This result strongly suggest that AmrAB-OprA stands apart from other efflux pumps and its inactivation alone significantly reduces the barrier and efflux constants for several classes of antibiotics.
The OM ratios of all efflux-deficient strains producing high levels of PoreH clustered together and separately (cluster 3). The highest ratios in this cluster are characteristic of antibiotics that are strongly affected by the low permeability of the OM, such as large antibiotics with a mass exceeding 800 Da, e.g., vancomycin or rifampin, and beta-lactams that have targets in the periplasm.
Finally, the efflux ratios of WT-PoreH and PoreH derivatives of single- and double-knockout strains deficient in AmrAB-OprA cluster separately from all other strains (Fig. 1). In this cluster, the ratios of MICs for all tested antibiotics are at least 4-fold. The antibiotics for which the activities are the most affected by these interactions include macrolides, tetracyclines, chloramphenicol, trimethoprim, and zeocin.
Thus, the clustering analysis described above suggested that the effect of hyperporination on activities of antibiotics depends on specific pumps present in B. thailandensis cells. It is particularly profound in the double-knockout Bt38 and Bt42 strains (Fig. 1 and Table S2), pointing to functional interactions between specific efflux pumps and the OM barrier in B. thailandensis cells.
Inactivation of B. thailandensis efflux pumps not only diminishes active efflux of antibiotics but also increases their uptake across the outer membrane.We next analyzed how the interactions between efflux pumps and the OM barrier affect antibiotics with different permeation and efflux avoidance properties.
(i) Antibiotics with activities limited by the OM barrier.In most Gram-negative bacteria, the activities of large antibiotics, such as vancomycin, rifampin, or bacitracin, and those of hydrophilic beta-lactams, such as ampicillin and carbenicillin, are mainly defined by their permeation across the OM. As reported before (5) and seen in Table 1, hyperporination of BtWT-PoreH has no effect on activities of these antibiotics. However, hyperporination of Bt34 [Δ(bpeAB-oprB)] and Bt36 [Δ(amrAB-oprA)] cells decreases the MICs of these antibiotics by at least 128- to 256-fold (Table 1). At the same time, neither AmrAB-OprA nor BpeAB-OprB can extrude these antibiotics, since their MICs in a Bt38 [Δ(bpeAB-oprB) Δ(amrAB-oprA)] mutant are the same as the MICs in Bt34 and Bt36, and deleting both pumps without hyperporination does not decrease the MICs. This result suggests that rather than decreasing their efflux, the simultaneous inactivation of AmrAB-OprA and BpeAB-OprB potentiates the activities of large antibiotics and hydrophilic beta-lactams by increasing their uptake into cells.
Interestingly, the MICs of vancomycin and bacitracin in hyperporinated efflux-deficient mutants of B. thailandensis are the same as the MICs in the hyperporinated B. cepacia BcWT-PoreH strain (Table 1). This result suggests that the uptake of these antibiotics is higher across the OM of B. cepacia than of B. thailandensis.
(ii) Polycationic antibiotics with self-promoted uptake across the OM.Unlike other Gram-negative bacteria, B. thailandensis cells are intrinsically resistant to polycationic peptides and aminoglycoside antibiotics primarily due to the modifications of the OM lipid A with Ara4N (11, 27). As seen in Table 1, aminoglycosides are substrates of AmrAB-OprA but not BpeAB-OprB because the MICs of gentamicin (2 to 4 μg/ml) and other aminoglycosides in Bt36 [Δ(amrAB-oprA)] and in double-knockout Bt38 [Δ(bpeAB-oprB) Δ(amrAB-oprA)] cells are the same. At the same time, the MICs of aminoglycosides in BtWT cells do not decrease upon hyperporination, but MICs in Bt36 (should not be involved in aminoglycoside efflux) do decrease upon hyperporination by 16- to 32-fold, suggesting that the deletion of the BpeAB-OprB pump increases the intracellular uptake of aminoglycosides. Again, the MICs of aminoglycosides in the hyperporinated Bt34-PoreH mutant that lacks the nonaminoglycoside pump BpeAB-OprB but still expresses AmrAB-OprA are the same as the MICs in hyperporinated BcWT-Pore mutant of B. cepacia that has its own aminoglycoside pump homologous to AmrAB-OprA (Table 1) (5).
Interestingly, neither inactivation of efflux pumps nor hyperporination alone was sufficient to make cells hypersusceptible to polymyxin B. The susceptibility to polymyxin B increased only in Bt36-PoreH, Bt38-PoreH, and Bt42-PoreH strains, all lacking amrAB-oprA and producing high levels of the Pore (Table 1). This result provides further evidence that inactivation of AmrAB-OprA and BpeAB-OprB impacts the permeation of polycationic antibiotics into the cells differently.
(iii) Antibiotics with activities affected only by one of the efflux pumps.The comparison of MICs in single- and double-efflux-knockout mutants, both nonhyperporinated and hyperporinated, shows that trimethoprim is not a substrate of AmrAB-OprA or BpeAB-OprB and is almost exclusively affected by the presence of BpeEF-OprC (Table 1). However, the MIC of trimethoprim in the hyperporinated BtWT-PoreH strain (8 μg/ml) is much higher than the MIC in hyperporinated single mutants (Table 1), as if it is a substrate of efflux by a still-present pump. An alternative explanation, in line with the results described above for other antibiotics, is that efflux pumps that are present in BtWT-PoreH affect not only efflux but also decrease the uptake of an antibiotic, which is not their substrate.
Similar effects could be seen for MICs of azithromycin and zeocin that are apparently the substrates of AmrAB-OprA and not BpeAB-OprB (Table 1). Hyperporination decreases the MICs of these antibiotics in Bt36-PoreH mutant lacking only AmrAB-OprA but not in the BtWT-PoreH cells producing both pumps. Also, novobiocin is a substrate of BpeAB-OprB and not AmrAB-OprA, yet hyperporination decreases the MIC in the Bt34-PoreH mutant but not in the wild type BtWT-PoreH.
Thus, for all these antibiotics, the changes in MICs support the conclusion that inactivation of efflux pumps affects the antibacterial activities not only by reducing their efflux but also by increasing their uptake.
(iv) Antibiotics with activities affected by multiple efflux pumps.For the majority of antibiotics, the presence of both AmrAB-OprA and BpeAB-OprB in the nonhyperporinated parent strains has additive effect on MICs. This can be seen from the comparison of MIC ratios between single-pump mutants, which shows the contribution of a still-present pump, and the double-pump mutants versus the MIC ratios between the wild type and the double-pump mutants. For example, for levofloxacin, both pumps contribute 16-fold to the susceptibility of B. thailandensis, as seen from the MIC ratios of Bt34 [Δ(bpeAB-oprB)] or Bt36 [Δ(amrAB-oprA)] (MIC, 1.0 μg/ml for both) versus Bt38 [Δ(bpeAB-oprB) Δ(amrAB-oprA)] (MIC, 0.06 μg/ml) (Table 1). When both the AmrAB-OprA and BpeAB-OprB pumps are present, the MIC of BtWT (2.0 μg/ml) increased by 32-fold compared to the MIC of Bt38. However, when the same ratios are calculated for the hyperporinated strains, the interactions are neither additive not multiplicative and are much higher than that. For levofloxacin, in a hyperporinated background, each pump individually contributes a 4-fold increase in the MIC, as seen from the MIC ratios of Bt34 or Bt36 (MIC, 0.06 μg/ml for both) versus Bt38 (MIC, 0.015 μg/ml), but together, they contribute 64-fold (MIC in BtWT, 1.0 μg/ml). This result shows that in the hyperporinated background the impact of efflux pumps, other than on efflux per se, becomes more prominent. For many antibiotics shown in Table 1, these synergistic effects are much stronger.
Taken together, the antibiotic susceptibilities of B. thailandensis strains with varied efflux capacities and permeabilities of the outer membrane suggest that the contributions of active efflux and the OM barrier in activities of antibiotics are interdependent. These analyses further suggest that the two barriers functionally interact with each other and that inactivation of efflux pumps not only diminishes the active efflux of antibiotics but also increases their uptake across the OM.
Outer membrane of B. thailandensis is the defining contributor to the kinetics of active efflux.N-Phenyl-naphtylamide (NPN) is a substrate of various efflux pumps, whose fluorescence is quenched in aqueous solutions and is strongly enhanced by binding to lipids. Therefore, we next used NPN to assess the changes in the permeability barriers of various strains of B. thailandensis and the contributions of different efflux pumps and the OM barrier to the intracellular accumulation of this probe. The kinetics of NPN accumulation was analyzed at different concentrations of the probe (Fig. S1 and S2), and the initial rates and steady-state concentrations were extracted by fitting to a two-step permeation model (Fig. 2 and S3).
Steady-state accumulation levels of NPN in B. thailandensis cells with varied efflux and outer membrane barriers. (A) BtWT and its efflux-deficient derivatives. (B) BtWT and its efflux-deficient derivatives producing Pore from the genes inserted only into one chromosome. (C) BtWT and its efflux-deficient derivatives producing Pore from the genes inserted into both chromosomes.
In the absence of hyperporination, the steady-state accumulation levels of NPN were similar in BtWT and all efflux-deficient mutants, suggesting that the OM and not the active efflux is the major permeability barrier for NPN (Fig. 2A). In agreement with this, hyperporination increased the uptake of NPN into all strains, and the intracellular accumulation correlated with the low and high expression levels of the Pore (Fig. 2B and C and S3). The intracellular levels of NPN in WT-PoreL and WT-PoreH cells were similar and ∼40% higher than in the BtWT (Fig. 2B and C). The NPN uptake into Bt40 and its hyperporinated variants was comparable to that of the BtWT, suggesting that inactivation of BpeEF-OprC alone is not sufficient to increase the NPN permeation into B. thailandensis cells. In contrast, hyperporination of strains lacking either BpeAB-OprB, AmrAB-OprA, or both resulted in the large change in the NPN permeation into cells. In particular, the Bt34-PoreH cells accumulated twice as much NPN as the Bt36-PoreH and Bt38-PoreH cells and up to six times as much NPN as BtWT and Bt40 cells with equal hyperporination (Fig. 2C).
Thus, both AmrAB-OprA and BpeAB-OprB efflux pumps reduce the accumulation of NPN in cells, although BpeAB-OprB appears to be more efficient. As in the case with the antibiotic susceptibilities (Table 1), intracellular accumulation of NPN is dramatically enhanced in hyperporinated efflux-deficient cells.
Lipid A composition correlates with strain- and species-specific differences in the permeability barriers.The complexity of relationships between efflux pumps and their interactions with the OM barrier in B. thailandensis suggested that inactivation of efflux pumps has pleiotropic effects. Previously, no significant changes in cell morphology or protein compositions of inner and outer membranes were found for Bt38 and Bt38-PoreH strains (5). Since modifications of LPS have the potential to change the permeability of the OM, we next analyzed the composition of lipid A species in different strains. All lipid A species of B. thailandensis strains contained five major ions at m/z 1,670, 1,686, 1,802, 1,824, and 1,955 (Fig. 3). These m/z species are similar to previously described lipid A species of B. pseudomallei, but there are also some differences (6). The ion at m/z 1,670 represents a penta-acylated bisphosphorylated GlcN disaccharide backbone possessing two C14:0(3-OH) residues and two C16:0(3-OH) residues on the diglucosamine backbone, one acyl-oxo-acyl C14:0 residue, and two phosphoryl moieties (Fig. 4). We identified two chemical modifications of this ion. Hydroxylation is evident by the presence of the ion at m/z 1,686, which is a derivative of the ion at m/z 1,670 with the substitution of fatty acid C14:0 for C14:0(2-OH) (Fig. 4). The addition of Ara4N yields the m/z 1,802 species with an Ara4N residue (Δm/z 131) attached to the phosphate group. Finally, the ions at m/z 1,824 and at m/z 1,955 correspond to penta-acylated species carrying one and two Ara4N residues, respectively, in which one hydrogen atom is substituted with a sodium ion (28). The relative intensities of the five main peaks of lipid A varied significantly between the strains, representing the differences that could potentially contribute to the antibiotic hypersusceptibilities of efflux-deficient strains. However, ion intensity is not absolutely quantitative using matrix-assisted laser desorption ionization (MALDI) due to crystallization and ionization issues.
MALDI mass spectrometry (MS) spectra of lipid A from B. thailandensis and B. cepacia. Red indicates Ara4N modification.
Proposed structures of lipid A and its modifications in B. thailandensis.
We previously found that the permeability barrier of BcWT cells is notably weaker than that of BtWT cells. In fact, hyperporination alone is sufficient to dramatically increase susceptibility to antibiotics in BcWT (5) (Table 1). We next compared the LPS composition of BcWT cells to that of B. thailandensis. The LPS of BcWT was enriched with tetra-acylated m/z 1,576 and penta-acylated m/z 1,728 species containing a single Ara4N substitution, as well as the penta-acylated m/z 1,598 species containing a single phosphate moiety and lacking Ara4N modification (Fig. 3). In contrast to previous reports, our analyses show clearly the presence of m/z 1,955 species. Hence, as with other Burkholderia spp., B. cepacia augments both phosphate groups of lipid A with Ara4N moieties.
Taken together, these results suggest that the differences in both the number of acyl chains and the Ara4N moieties present in lipid A, as well as their relative abundances, could contribute to the differences in the permeability barriers of the Burkholderia species.
DISCUSSION
In Gram-negative bacteria, active efflux acts synergistically with the OM barrier to protect cells from antibiotics and toxic compounds. In this study, we found that in B. thailandensis and likely in other Burkholderia spp., the activities of efflux pumps are functionally linked with the permeability properties of the OM. This finding raises several questions about the assembly and functions of both barriers in Burkholderia spp. and other species.
The activities of all tested antibacterial agents (Table 1) are affected by interactions between the two permeability barriers. The impact of these interactions on the susceptibility of B. thailandensis to polycationic antibiotics stands out because of the known effects of the lipid A modifications to the resistance against these drugs. The aminoglycoside pump AmrAB-OprA is highly conserved in both BcWT and BtWT (87% identity between AmrB components) and shares significant homology with MexXY-OprM in P. aeruginosa PAO1 (74% identity between AmrB and MexY). Although these pumps are constitutively expressed in both genera, their contributions to aminoglycoside and other antibiotic resistance are different. Inactivation of MexXY-OprM mildly potentiates the activities of aminoglycosides in the naturally susceptible P. aeruginosa strains, whereas the overexpression of this pump is the major cause of the adaptive aminoglycoside resistance in clinical isolates. In contrast, Bpc and Bcc species are intrinsically resistant to aminoglycosides, whereas inactivation of AmrAB-OprA makes the cells as susceptible to aminoglycosides (Table 1) as the P. aeruginosa PAO1 strain.
The intrinsic resistance to aminoglycosides in Burkholderia spp. is primarily due to the permeability properties of the OM and the Ara4N modifications present in lipid A. Pseudomonas spp. lack or have a very low level of this modification under laboratory conditions but augment their lipid A with Ara4N and other modifications under acidic-pH and low-Mg2+ conditions (29, 30). In agreement with this, hyperporination of the OM in P. aeruginosa PAO1 and other susceptible Gram-negative bacteria does not potentiate the activities of aminoglycosides, demonstrating that the OM of these species does not hinder the permeation of these antibiotics (5). Even the effect of MexXY-OprM overexpression or deletion on aminoglycoside resistance is strongly affected by the presence of Mg2+ ions that are needed to reduce the permeation of aminoglycosides across the OM (31).
Hyperporination of the OM in BcWT cells, on the other hand, potentiated the activities of aminoglycosides despite the presence of the fully functional AmrAB-OprA (Table 1). Thus, the activities of aminoglycosides in this species are limited by the OM barrier and increase with increasing the number of Pores and, hence, with the increased influx of antibiotics across the OM. Surprisingly, hyperporination alone was not sufficient to break the OM barrier in BtWT cells (Table 1). Only the inactivation of AmrAB-OprA in Bt36 and double-knockout derivatives reduced the MICs of aminoglycosides and to the same levels as in the susceptible Gram-negative bacteria (5). This contradiction in the effects of AmrAB-OprA and the OM barrier on activities of aminoglycosides is resolved by the differences in the composition of LPS species of BcWT and BtWT (Fig. 3). The LPS of BcWT contains a large fraction of tetra-acylated species, with at least half of them carrying one or two Ara4N moieties. In contrast, the majority of LPS in BtWT are penta-acylated species carrying Ara4N moieties (Fig. 3 and 4). This penta-acylated LPS creates a more rigid outer leaflet in the OM and prevents permeation of aminoglycosides into cells. Importantly, either one of the LPS structures is efficient in protection against aminoglycosides and provide similar high MIC values. This result suggests that although hyperporination increases the influx of aminoglycosides, pores are not the only route of antibiotic permeation. Only the combined contributions of both the hyperporination and the more permeable LPS layer enable potentiation of aminoglycosides in Burkholderia spp. The same trends are also seen with polymyxin B and almost all compounds shown in Table 1, suggesting similar barriers for permeation of all these antibacterials.
Interestingly, Ara4N modifications are thought to be essential for Burkholderia sp. growth (11). Our results, however, show that B. thailandensis can accommodate a broad distribution in the amounts of this lipid A species and that a significant fraction of lipid A could lack Ara4N modification without significant growth defects under tested conditions. Furthermore, recent investigations of B. pseudomallei clinical isolates showed different degrees of Ara4N modification in lipid A as well (7). Since Ara4N is essential for lipid A transport, it is likely that the extent of Ara4N modifications on lipid A is controlled at the surface of the OM.
MATERIALS AND METHODS
Strains and growth conditions.The bacterial strains and plasmids used in this study are listed in Table S4. Luria-Bertani (LB) broth (10 g of Bacto tryptone, 5 g of yeast extract, and 5 g of NaCl per liter [pH 7.0]) or LB agar (LB broth with 15 g of agar per liter) was used for bacterial growth. When indicated, cultures were induced with 0.2% l-rhamnose to induce expression of the Pore protein. For selection, kanamycin (100 μg/ml), tetracycline 50 μg/ml (BtWT), 25 μg/ml (BtWT-PoreL), 10 μg/ml (BtWT-PoreH, Bt36, and Bt36-PoreL), 2.5 μg/ml (Bt36-PoreH), trimethoprim (100 μg/ml), ampicillin (100 μg/ml), and polymyxin B (25 μg/ml) were used.
Deletion of efflux pumps and excision of antibiotic cassettes.Efflux pumps were deleted by natural transformation of assembled PCR products, as described previously (32). Excision of tetracycline and trimethoprim resistance cassettes was done using the pFLPe5 plasmid, as described previously (33). To confirm that the observed changes in susceptibilities to antibiotics are specific to the inactivated pumps, Bt34, Bt36, and Bt38 strains were complemented with genes encoding either the BpeAB-OprB or AmrAB-OprA pump. For this purpose, the respective operons were integrated back onto the chromosomes of the efflux-deficient strains. The antibiotic susceptibility profiles of the complemented Bt34::bpeAB-oprB and Bt36::amrAB-oprA strains were indistinguishable from that of the BtWT cells, whereas the Bt38::bpeAB-oprB and Bt38::amrAB-oprA mutants matched the susceptibility profiles of the Bt36 [Δ(amrAB-oprA)] and Bt34 [Δ(bpeAB-oprB)] mutants, respectively (Table S3). Therefore, changes in susceptibilities to antibiotics in constructed B. thailandensis strains are caused by the loss of the activities of the specific efflux pumps.
Cloning and integration of OrbA expression constructs into the attTn7 site.The sequences of the primers used in this study are shown in Table S5. To construct pUC18T-mini-Tn7T-Tp-RHA, the rhaRS-PrhaBAD promoter was amplified with RhaR NsiI REV and P-rhaBAD SpeI/NcoI REV. The PCR product and pUC18T-miniTn7T-Tp were digested with NsiI/SpeI and ligated using T4 ligase, following the manufacturer’s protocols (NEB). Positive clones were selected on LB agar (LBA) plates containing 100 μg/ml trimethoprim and confirmed positive clones by restriction digestion. The B. thailandensis OrbA gene encodes a homologue of the Escherichia coli FhuA siderophore receptor protein. The gene encoding OrbA was modified similarly to the synthetic FhuAΔC,Δ4L gene (35), in which the periplasmic cork domain and the extracellular loops have been removed. This synthetic orbA(OrbAΔcork,Δ4loop,9His) gene was synthesized by the GenScript Corporation and cloned into pUC57 between the SacI and HindIII sites. An NcoI site was engineered to encode the start codon for cloning into pUC18T-mini-Tn7T-Tp-RHA, which encodes optimized Shine-Dalgarno sequences in frame to their NcoI sites.
The orbA gene was integrated onto the chromosome of B. thailandensis via mini-Tn7 cassettes by triparental mating with recipient B. thailandensis, RHO3/pTNS3, and SM10(λpir+) carrying pUC18T-based mini-Tn7 cassettes. Mating was carried out by growing strains on LB agar plates with appropriate selection markers. Overnight cultures were diluted 1:100 for recipient and RHO3/pTNS3 or 1:20 for SM10(λpir+) carrying pUC18T-based mini-Tn7 cassettes into fresh LB containing appropriate selection markers and grown at 37°C to an OD of ∼0.2. Cells were collected by centrifugation at 3,220 × g for 20 min and gently resuspended in LB to concentrate 100×. Ten microliters of each strain was mixed in a sterile centrifuge tube, and 10 μl of triparental mixture was spotted and allowed to adsorb onto a dried LBA plate containing 0.5 mM diaminopimelic acid (DAP). Mating reaction mixtures were incubated at 37°C for 16 h. Spots were scraped off the LBA plates and resuspended in 500 μl of LB. One hundred microliters of resuspended mating reaction mixture was plated onto LBA containing 25 μg/ml polymyxin B and 100 μg/ml trimethoprim. Integrants were screened by stable growth on LBA plates containing 25 μg/ml polymyxin B and 100 μg/ml trimethoprim and confirmed by PCR using either B. thailandensis glmS1-Dn or B. thailandensis glmS2-Dn along with Tn7L primers to confirm integration into either glmS1- or glmS2-linked attTn7 sites, respectively.
Drug susceptibility assay.The susceptibility of Burkholderia spp. to different classes of antibiotics was determined by employing a 2-fold broth dilution method, as described before (25).
N-Phenyl-naphtylamide accumulation assay.An NPN uptake assay was done by monitoring its fluorescence in B. thailandensis cells expressing OrbA ΔC-Δ4L(Pore) induced with 0.2% rhamnose, as described previously (5). The data were fitted to an exponential equation in the form of F = A1 + A2[1 − exp(−kt)], where A1 represents the amplitude of the initial fast uptake of NPN, and A2 and kt are the amplitude and rate, respectively, associated with the slower uptake. All measurements were done in duplicate, and at least two independent experiments were done for all strains.
Extraction of lipid A and structure determination.Burkholderia sp. cells were grown overnight in 20 ml of LB broth at 37°C with aeration, as described above. Cells were collected by centrifugation at 1,600 × g, washed 2 to 3 times with phosphate-buffered saline (PBS) (pH 7.0), and kept frozen at −80°C until analyses. For extraction of lipid A, an ammonium hydroxide-isobutyric acid-based procedure was used (34). Samples were analyzed on a Bruker microflex system in negative-ion mode, as described before (6). Data were processed with the flexAnalysis software.
ACKNOWLEDGMENTS
Studies in H.I.Z.’s laboratory are sponsored by the Department of the Defense, Defense Threat Reduction Agency (grant HDTRA1-14-1-0019), and by the NIH/NIAID grant R01AI132836. H.P.S. was supported by the NIH/NIAID grant U54 AI065357.
FOOTNOTES
- Received 8 May 2019.
- Returned for modification 11 July 2019.
- Accepted 28 July 2019.
- Accepted manuscript posted online 5 August 2019.
Supplemental material for this article may be found at https://doi.org/10.1128/AAC.00956-19.
- Copyright © 2019 American Society for Microbiology.
