Importance of Core Genome Functions for an Extreme Antibiotic Resistance Trait

Tobramycin-sensitive mutants of A. baumannii AB5075.We initially asked if there were functions other than the known modifying and efflux functions that contribute to the high-level tobramycin resistance of AB5075. To identify such functions at a comprehensive scale, we used transposon sequencing (Tn-seq) to screen for mutations that reduce tobramycin resistance. We grew a saturation level transposon mutant pool (~450,000 mutants) at several subinhibitory tobramycin levels and identified mutants that were lost (see Materials and Methods). The loss of insertions in several resistance genes after growth in the presence of tobramycin is illustrated in Fig. 1A. A plot comparing recovery with and without tobramycin for all of the genes represented in the mutant pool shows that insertions in numerous genes reduce resistance (Fig. 2A; and see Fig. S1 in the supplemental material). On the basis of a mutant fitness decrease based on statistical criteria described in Materials and Methods, the screening identified 105 top candidate resistance genes (Data Set S1). Although the group included the expected accessory tobramycin resistance genes, most (80%) of the genes were novel and belong to the core genome.

• Open in new tab
• Download powerpoint

FIG 1 

Mutations reducing tobramycin (Tob) resistance. Tn-seq read profiles for three regions of the genome with resistance genes are shown for the wild type (AB5075) (A) and the resistance island deletion strain (AB5075ΔRI) (B). Transposon insertion positions and orientations are represented by vertical bars with lengths corresponding to log₂-transformed read counts. Growth in normally subinhibitory tobramycin (~10 to 40% of the MIC) leads to loss of transposon sequence reads for tobramycin resistance genes gigA, gigB, ABUW_0471, and fadB (in AB5075). Note the absence of insertions in pssA under all conditions because it is essential. The vertical scale corresponds to a maximum log₂ (normalized number of reads per site) value of approximately 11. Pre, original mutant pool.

• Open in new tab
• Download powerpoint

FIG 2 

Genome scale identification of tobramycin resistance genes. Tn-seq read counts per gene after growth of mutant pools with and without tobramycin are shown for AB5075 (A) and AB5075ΔRI (B). Values represent the averages of all tobramycin levels analyzed for each strain. Genes whose mutants were significantly selected against during growth in the presence of tobramycin are represented by blue circles (see Materials and Methods). Individual plots of each tobramycin level examined are presented in Fig. S1. The exceptional blue circle near the x-y diagonal in panel B represents a gene whose mutants were significantly selected against at two of the three tobramycin levels analyzed.

Since Tn-seq assays monitor mutants grown in competition, modest reductions in the growth rate are amplified and it can be difficult to distinguish strong from weak phenotypes. Thus, it is critical to validate Tn-seq phenotypes with quantitative assays of individual mutants (15). Accordingly, we evaluated the tobramycin MICs for 437 individual mutants representing the genes of the top group and others with weaker Tn-seq phenotypes (195 genes total). We obtained mutants primarily from the arrayed AB5075 transposon mutant library (14), although we also created gene deletion mutants for five genes for which the phenotypes of different transposon mutants disagreed (see below). Altogether, we found that mutations in 32 genes strongly reduced resistance (≥4-fold MIC reductions) and mutations in another 49 genes led to weaker phenotypes (2-fold MIC decreases) (Table 1; Data Set S1).

Other candidate resistance genes.In addition to genes identified by Tn-seq, we also evaluated the mutant phenotypes of 33 genes potentially contributing to tobramycin resistance on the basis of annotated function. Most of the additional genes were predicted to encode efflux pumps (21 genes) or potential aminoglycoside-modifying functions (5 genes) or were orthologous to tobramycin resistance genes of P. aeruginosa (3 genes) (16–18) (Data Set S1). We assayed multiple insertion mutants for most of these genes, and for seven of them, we also created and assayed deletion mutants. Mutations inactivating 5 of the 33 genes reduced the tobramycin MIC ≥4-fold (Table 1; Data Set S1). The result shows that even extensive Tn-seq screening may fail to identify some mutants with significant phenotypes.

In total, we identified 37 genes that strongly contribute to tobramycin resistance in AB5075 (Table 1). All but three of these genes belong to the core genome, the exceptions being the two known tobramycin-modifying functions (aadB and aacA4) and a plasmid gene of unknown function (ABUW_5005). Mutations in six core genes (serA, serB, astE, ABUW_0471, yajC, and ABUW_1451) led to particularly strong phenotypes, individually increasing sensitivity as much as or more than the elimination of both tobramycin-modifying functions (see below). Overall, the results show the profound contribution of core genome functions to high-level aminoglycoside resistance, even when potent antibiotic-modifying enzymes are expressed.

The core genes we identified were also important for resistance to aminoglycosides other than tobramycin. We tested mutants corresponding to 20 of the genes for sensitivity to gentamicin and amikacin and found that nearly all exhibited reduced resistance to the two drugs, with MIC decreases ranging from 1.3- to 21-fold relative to a control insertion strain (Data Set S1).

Allelic variability.In the course of the single mutant studies, we identified several genes for which insertions at some positions strongly increased tobramycin sensitivity, while those at others did not (Data Set S1). Four such genes encode secreted or integral membrane proteins: three RND efflux pump membrane fusion proteins (ABUW_0034/arpA, ABUW_3153, and ABUW_3560) and a MATE family efflux pump (ABUW_3486/abeM). Since aberrant proteins targeted to the membrane can increase aminoglycoside sensitivity (18, 19), we suspected that gain-of-function effects could explain the allelic variability. Indeed, gene deletion mutants of each of the four genes displayed wild-type tobramycin sensitivity, suggesting that the insertion alleles that increased sensitivity did so by gain-of-function effects. However, we have not ruled out the alternative possibility that unlinked second-site mutations are responsible in some cases (Fig. S2).

The plasmid-borne aphA6 gene (ABUW_4087) also showed allelic variability (Fig. S2). This gene is predicted to encode a cytoplasmic aminoglycoside phosphotransferase that does not act on tobramycin (20). We found that a deletion eliminating aphA6 and adjacent IS elements (21) did not change tobramycin sensitivity, again indicating that the insertions increasing sensitivity are gain-of-function alleles (Fig. S2).

Tobramycin-sensitive mutants of a resistance island deletion strain.We next evaluated whether there are intrinsic resistance determinants that function in the absence of tobramycin-modifying functions but not in their presence. To do this, we first created an AB5075 derivative lacking the strain’s two major accessory genome resistance islands (14), including genes for the two tobramycin-modifying enzymes (aacA4 and aadB) (Fig. 3). As expected, the strain (AB5075ΔRI) was much more tobramycin sensitive than its parent, with a 16-fold reduction in the MIC (Table 2).

• Open in new tab
• Download powerpoint

FIG 3 

Deletion of resistance island 2. Resistance island 2 (14) resides on plasmid p1 and contains two genes for modification enzymes predicted to inactivate tobramycin, aadB and aacA4 (red), as well as genes for three other aminoglycoside modification functions (pink). The precise deletion indicated restores the ancestral gene of unknown function (ABUW_4065/ABUW_4043, blue) interrupted by the island (Text S1). Genetic elements: MITE, miniature inverted-repeat transposable element; intI, integrase; aadB, aminoglycoside 2′′-nucleotidyltransferase; cmlA, chloramphenicol resistance protein; aadA, aminoglycoside 3′′-nucleotidyltransferase; strA and strB, streptomycin phosphotransferases; bla_GES-14, class A β-lactamase; aacA4, aminoglycoside 6′-acetyltransferase; dfrA7, dihydrofolate reductase; qacEΔ1, multidrug resistance efflux pump; sul1, sulfonamide resistance protein; orf5, unknown function; tniBΔ1, transposition protein.

We next employed Tn-seq analysis to identify tobramycin resistance determinants in the deletion mutant background by using an approach analogous to that used with the wild type (Fig. 1B and 2B). The screening identified 108 genes whose mutants were strongly depleted during growth with tobramycin (see Materials and Methods and Data Set S1). There was a striking overlap of these resistance genes with those identified in the parent strain (Fig. 4; and see Fig. S3). The findings imply that nearly identical intrinsic resistance functions are important whether or not the tobramycin-modifying functions are present.

• Open in new tab
• Download powerpoint

FIG 4 

Shared core resistance genes in the wild-type and resistance island deletion strains. The overlap of the top resistance genes identified by Tn-seq in AB5075 (105 genes) and AB5075ΔRI (108 genes) is shown. The 130 unique genes represented do not include 4 resistance island genes and 1 resistance gene essential in AB5075ΔRI. A gene was included in the overlap region if (i) it was identified independently in both strains or (ii) the ratio of Tn-seq read counts in the presence of tobramycin for the two strains did not differ significantly from that of the preponderance of AB5075 genes (Fig. S3).

Of the genes that appeared to be necessary for resistance in one strain but not the other (on the basis of the Tn-seq results), the greatest difference was seen in fadB (Fig. 1, right panel). However, when we transferred fadB (and fadA) insertion alleles from AB5075 into AB5075ΔRI, we found that the mutations led to the same MIC decrease (4-fold) in both strain backgrounds (Table 1). Although the basis of the difference in the behavior of the mutants in Tn-seq pools and as individual strains is not clear, the result shows that the contributions of fadAB to resistance do not depend on the tobramycin-modifying functions.

While carrying out these screenings, we were surprised to find four genes that were essential for the growth of AB5075ΔRI, but not that of AB5075, on all of the growth media used. These were genes for dihydrofolate reductase (folA), dihydropteroate synthase (folP), thioredoxin (trxB), and glutathione S-transferase (gst/ABUW_2725). The first three of these genes are paralogs of genes in the resistance islands, suggesting that the resistance island genes can function in place of the corresponding core genes for normal growth. The fourth gene (gst) lacks paralogs in the resistance islands, and the basis of its essentiality in AB5075ΔRI is unclear.

In addition, by phenotype microarray analysis (22), we found that AB5075ΔRI had become sensitive to two boron-containing compounds, boric acid and sodium metaborate. A single resistance island 1 (RI1) deletion mutant was also sensitive, and single transposon mutant analysis of seven genes within RI1 identified sup (ABUW_3659) (annotated as encoding a sulfate permease) as the gene responsible (data not shown).

Single mutant phenotypes in AB5075ΔRI.To provide a direct comparison of core sensitizing mutations in the wild-type and resistance island-lacking strains, we transformed insertion alleles from the wild type into AB5075ΔRI or constructed identical deletion mutations in the two strains. Thirty-one such mutations (corresponding to 24 genes) were examined, and all reduced tobramycin sensitivity comparably in the two strain backgrounds (Fig. 5). The results thus imply that individual core function contributions to resistance are quantitatively similar whether or not the tobramycin-modifying enzymes are expressed.

• Open in new tab
• Download powerpoint

FIG 5 

Direct comparison of resistance mutation effects in the presence and absence of tobramycin modification functions. The tobramycin MIC decreases (ΔMICs) caused by identical insertion or deletion mutations affecting 24 resistance genes were compared in the AB5075 and AB5075ΔRI genetic backgrounds (Pearson’s r = 0.91, n = 31, P = 2.5 × 10⁻¹²). A least-squares regression line of the log₂-transformed ΔMICs is fitted (dashed line).

Importance of phospholipid biosynthesis in intrinsic tobramycin resistance.Three genes with strong mutant phenotypes appear to affect phospholipid synthesis. The first of these encodes a phosphatase (PgpA) that catalyzes the final step in phosphatidylglycerol (PG) synthesis. Although some PG synthesis is probably essential (and indeed, an enzyme acting immediately prior to PgpA [PgsA] is essential) (14), we assume that other phosphatases can partially substitute for a missing PgpA (23). However, the pgpA mutation presumably alters membrane phospholipid makeup (or leads to precursor accumulation) sufficiently to increase tobramycin sensitivity.

A second set of mutations appears to affect PssA, the enzyme responsible for the synthesis of phosphatidylserine (PS) from CDP-diacylglycerol and serine. Although mutations in pssA are absent from the Tn-seq mutant pools because the gene is essential, insertions in the gene transcriptionally upstream of it (ABUW_0471) markedly reduced tobramycin resistance (MIC decreases of up to 64-fold). The reduction appears to be due to a polar effect on pssA on the basis of two observations. First the ABUW_0471 insertion alleles were phenotypically variable, with insertions closer to the 5′ end of the gene stronger than those closer to the 3′ end, a hallmark of nonsense-mediated polarity (24). A similar pattern is evident in the Tn-seq data (Fig. 1, middle panel). Second, an in-frame ABUW_0471 deletion mutation designed to be nonpolar did not reduce tobramycin resistance. We thus assume that decreased expression of pssA in the ABUW_0471 insertion mutants limits the production of PS (and its decarboxylation product phosphatidylethanolamine) and that the defect increases tobramycin sensitivity.

Finally, mutations in a serine biosynthesis gene (serB) led to enhanced tobramycin sensitivity, again presumably because of reduced PS synthesis. The mutants grew slowly on unsupplemented nutrient growth medium (LB), suggesting that the level of serine in the medium is not sufficient to fully supplement their auxotrophies. Two other serine pathway genes were either fully (serC) or borderline (serA) essential (14) (Fig. S4), presumably because of reduced phospholipid synthesis. Two very slowly growing serA mutants available in the arrayed AB5075 transposon mutant library displayed 32-fold decreases in the tobramycin MIC relative to that of the parent strain, further documenting the need for serine synthesis for intrinsic resistance.

Other than ABUW_0471 (see above), 24 of the genes in Table 1 reside in potential operons in which polar effects on genes downstream are possible. In nearly all of these cases, downstream genes were not essential and insertion mutants of them did not show altered tobramycin sensitivity on the basis of Tn-seq, suggesting that polar effects were not responsible for the sensitivity phenotypes (not shown). However, in four cases, immediate downstream genes were essential, and we cannot rule out the possibility that polar effects contribute to the observed sensitivity phenotypes. The four genes were pgpA (upstream of glmU and glmS), yajC (upstream of secD and secF), trpA (upstream of accD, folC, and ABUW_0625), and lptE (upstream of holA).

Highly sensitized strains lacking multiple core resistance functions.We sought to determine whether it is possible to create highly tobramycin-sensitive AB5075 strains by combining multiple mutations compromising core resistance. Indeed, several double and triple mutants had as much as 256-fold increased tobramycin sensitivity despite expressing the tobramycin-modifying functions (Table 2).

Mutations increasing tobramycin resistance.Our screenings identified several genes in which insertion mutations increased resistance to tobramycin (Data Set S1). Two of the genes encode a two-component regulator (PhoBR) controlling the bacterial response to inorganic phosphate limitation (25). Assays of six individual transposon mutants from the arrayed mutant library found that insertions in each gene increased the tobramycin MIC 4-fold (Data Set S1). This result implies that PhoBR activity reduces the tobramycin resistance of the wild type. To test explicitly whether PhoBR activation by phosphate limitation reduces resistance, we assayed the tobramycin MICs for strains grown on a defined medium supplemented with different levels of phosphate (Table 3). We found that wild-type strains were indeed sensitized by phosphate limitation, whereas phoR mutants were not. This finding confirms that activation of the PhoBR regulator by phosphate limitation enhances tobramycin sensitivity in AB5075. The finding also suggests that the increased tobramycin sensitivity of Pst phosphate transporter mutants (Table 1) could be explained by their activation of PhoBR (25).

Mutations that impair electron transport can result in enhanced aminoglycoside resistance in other bacteria (26), but we did not identify such mutations in our screenings. However, in AB5075, many of the genes encoding central electron transport functions (including nuoA to nuoN [NADH dehydrogenase], cyoA to cyoE [cytochrome oxidase], and ubiquinone biosynthesis genes) were essential under the growth conditions used in our screenings. Thus, the effects of corresponding mutations on tobramycin resistance could not be evaluated.