Nutrient Limitation Sensitizes Pseudomonas aeruginosa to Vancomycin

Abstract

Traditional antibacterial screens rely on growing bacteria in nutrient-replete conditions which are not representative of the natural environment or sites of infection. Instead, screening in more physiologically relevant conditions may reveal novel activity for existing antibiotics. Here, we screened a panel of antibiotics reported to lack activity against the opportunistic Gram-negative bacterium, Pseudomonas aeruginosa, under low-nutrient and low-iron conditions, and discovered that the glycopeptide vancomycin inhibited the growth of P. aeruginosa at low micromolar concentrations through its canonical mechanism of action, disruption of peptidoglycan crosslinking. Spontaneous vancomycin-resistant mutants underwent activating mutations in the sensor kinase of the two-component CpxSR system, which induced cross-resistance to almost all classes of β-lactams, including the siderophore antibiotic cefiderocol. Other mutations that conferred vancomycin resistance mapped to WapR, an α-1,3-rhamnosyltransferase involved in lipopolysaccharide core biosynthesis. A WapR P164T mutant had a modified LPS profile compared to wild type that was accompanied by increased susceptibility to select bacteriophages. We conclude that screening in nutrient-limited conditions can reveal novel activity for existing antibiotics and lead to discovery of new and impactful resistance mechanisms.

The outer membrane (OM) of Gram-negative bacteria provides intrinsic resistance to many antibiotics by reducing uptake. With some exceptions, the OM excludes large and charged molecules with a size-exclusion limit of ∼600 Da.¹ The OM of the Gram-negative opportunistic pathogen P. aeruginosa is considered particularly impenetrable, up to 100-fold less permeable compared to that of Escherichia coli.² This characteristic, coupled with multiple efflux systems which extrude molecules that enter the cell, make P. aeruginosa infections difficult to treat. A better understanding of how antibiotics cross the OM may be informative for the development of new strategies to increase drug uptake.

A major contributor to OM impermeability is lipopolysaccharide (LPS), found in the outer leaflet. LPS is composed of a lipid A anchor, an inner and outer core oligosaccharide, and O-antigen of varying length and chemical composition. LPS contributes to membrane stability,³ motility,^4,5 biofilm formation,⁶⁻⁸ and antibiotic resistance.⁹ LPS is highly negatively charged due to the presence of phosphate groups on lipid A and the inner core region.¹⁰ Divalent cations such as Mg²⁺ and Ca²⁺ are essential to neutralize the negative charge and maintain barrier integrity.¹¹ Some antibiotics, such as cationic peptides colistin and polymyxin B, can compete with divalent cations for LPS binding, leading to OM permeabilization.^11,12 The macrolide, azithromycin, acts through a similar mechanism against P. aeruginosa.¹³ However, bacteria can modify their LPS composition to gain resistance. In P. aeruginosa, mutations in the sensor kinase of the two-component PmrAB system lead to the activation of multiple genes, including the arn operon, which catalyzes the covalent addition of 4-amino-4-deoxy-l-arabinose to lipid A.¹⁰ This addition of sugars reduces the binding of cationic antimicrobial peptides. In E. coli, mutations in the O-antigen ligase waaL lead to the incorporation of peptidoglycan precursors that bind vancomycin to prevent entry of the antibiotic into the cell.¹⁴ LPS also serves as a primary receptor for many bacteriophages. In response, bacteria can modify their LPS composition to prevent phage attachment.^15,16 Phages can counter resistance by encoding enzymes that modify host LPS, to facilitate their own uptake or that of related phages. Overall, LPS composition plays an important role in antibiotic and phage susceptibility.

Once an antibiotic gets inside the cell, bacteria respond to the resulting stress in ways that are not yet fully understood. For example, multiple two-component regulatory systems can be activated in response to specific stimuli.¹⁷⁻²⁰ Typical two-component systems are composed of a sensor kinase and a response regulator. The sensor can have both kinase (activating) and phosphatase (deactivating) activities that control the phosphorylation state of the response regulator. The response regulator binds upstream of various genes to modulate their expression. Among the best-characterized two-component systems is CpxA-CpxR in E. coli, which is activated in response to misfolded proteins in the periplasm,²¹ upon overexpression of NlpE,²² or loss of the l,d-transpeptidase LdtD.²³P. aeruginosa has an orthologous system—CpxSR;²⁴ however, the exact repertoire of genes regulated by CpxR and the effects downstream of its activation are not yet fully understood. Activation of the Cpx system is involved in antibiotic resistance through the regulation of porin²⁵ and efflux pump expression,²⁶ plus other mechanisms that remain to be discovered.

We previously discovered that the thiopeptide antibiotics, thiostrepton and thiocillin, cross the OM of P. aeruginosa using siderophore transporters that are upregulated in low-iron media.²⁷⁻³⁰ The thiopeptides synergized with the FDA-approved iron chelator deferasirox (DSX). Prior to our work, the thiopeptides were considered to lack antipseudomonal activity^31,32 because susceptibility assays are typically conducted in nutrient-replete media. These findings suggested that there may be other large natural product antibiotics that could cross the OM in nutrient-depleted conditions. In this work, we screened existing antibiotics reported to have poor or no activity against P. aeruginosa in low-nutrient conditions. We found that the glycopeptide vancomycin had low micromolar activity against P. aeruginosa and that its activity was iron- and copper-dependent. Analysis of vancomycin-resistant mutants revealed mutations that activated CpxSR, concomitantly inducing resistance to β-lactam antibiotics. Further, we identified a mutation in an LPS glycosyltransferase that decreased susceptibility to vancomycin and azithromycin, another large natural product antibiotic, but unexpectedly increased susceptibility to select bacteriophages. These data provide insight into the mechanism of vancomycin uptake in P. aeruginosa and describe resistance mechanisms that can promote changes in resistance to other drug classes or therapeutic alternatives, with implications for the uptake of antibiotics across the OM barrier and effective treatment of P. aeruginosa infections.

Results

Vancomycin Inhibits P. aeruginosa Growth in an Iron- and Copper-Dependent Manner

To identify natural products besides the thiopeptides with possible antipseudomonal activity, we screened 21 commercially available natural product antibiotics at 10 μg/mL for activity against P. aeruginosa PA14 in nutrient-limited 10:90 (10% lysogeny broth: 90% phosphate-buffered saline) medium in the presence of the iron chelator, deferasirox (64 μg/mL) (Figure 1A, Supplementary Table S1). The 21 antibiotics were selected because they were (1) reported to lack antimicrobial activity against P. aeruginosa and (2) had a molecular weight >600 kDa, precluding their passive diffusion through porins.^33,34 Six of the 21 compounds inhibited the growth of P. aeruginosa in 10:90 (Supplementary Table S1). We focused here on vancomycin because it is used clinically to treat Staphylococcus aureus infections that are often associated with P. aeruginosa.³⁵⁻³⁸ The mechanism of action of vancomycin has also been extensively studied using Gram-positive species.

Figure 1.

Vancomycin inhibits P. aeruginosa under nutrient-limited conditions. (A) Schematic for screening compounds against P. aeruginosa. Checkerboard assays with vancomycin + DIBI against (B) PA14 and (C) tonB1::Mar2xT7. The darker the shade of blue, the more growth observed. White squares indicate a lack of growth observed. Checkerboard assays with vancomycin and (D) FeCl₃ and (E) CuCl₂. All checkerboards were averaged from three independent biological replicates.

We first confirmed that vancomycin had activity in combination with other iron chelators using the synthetic 3-hydroxypyridin-4-one polymeric iron chelator DIBI (Denying Iron to Bacterial Infections) that is unable to enter cells due to its high molecular weight (∼9000 kDa).³⁹ Similar to the results with DSX, vancomycin synergized with DIBI, further supporting the connection between iron limitation and antibiotic activity (Figure 1B). Vancomycin alone had a minimal inhibitory concentration (MIC) of 16 μg/mL (11 μM) in 10:90, but the addition of 8.3 μg/mL of DIBI reduced the MIC of vancomycin to 4 μg/mL. DIBI alone has a MIC of 63 μg/mL (7 μM) in 10:90. Since the thiopeptides thiostrepton and thiocillin can cross the OM using siderophore transporters, we tested synergy between vancomycin and DIBI against a tonB1 transposon mutant. The TonB1-ExbBD complex energizes uptake of ligands through OM siderophore transporters and tonB1 mutants are thiopeptide-resistant.²⁹ Susceptibility of the tonB1 mutant to the combination of vancomycin + DIBI was similar to that of the wild type (WT), suggesting that vancomycin uptake did not rely on TonB1-dependent transporters, or that loss of tonB1 was insufficient to reduce activity (Figure 1C). Since vancomycin susceptibility was iron-dependent, we tested the effects of adding back various metal salts, including FeCl₃, CuCl₂, MgCl₂, and CaCl₂ into 10:90. We predicted that FeCl₃ would antagonize vancomycin susceptibility because of the observed sensitization by iron chelators. Further, we predicted that MgCl₂ and CaCl₂ may have a similar effect because the divalent cations stabilize the highly negative charge of the LPS and thus OM integrity. As expected, iron supplementation antagonized vancomycin susceptibility (Figure 1D). Interestingly, copper supplementation also resulted in antagonism (Figure 1E), but neither MgCl₂ nor CaCl₂ antagonized vancomycin susceptibility to the same extent (Supplementary Figure S1). These results suggest that vancomycin susceptibility is mainly Fe³⁺- and Cu²⁺-dependent and that the antibiotic was not simply entering the cell due to compromised OM integrity in 10:90.

Activity Is Vancomycin-Specific and Not due to Decreased Outer Membrane Integrity

We verified that the susceptibility to vancomycin was not associated with membrane permeabilization (Figure 2A). 1-N-Phenylnaphthylamide (NPN) is a weakly fluorescent dye that poorly permeates the OM; however, upon permeabilization of the OM, the dye binds to phospholipids and strongly fluoresces. Polymyxin B, which disrupts the OM, increased fluorescence, whereas vancomycin or piperacillin had no effect on fluorescence compared to the vehicle control.

Figure 2.

Glycopeptide activity is vancomycin-specific and vancomycin acts through its canonical mechanism of action. (A) NPN assay with PA14 in 10:90 with polymyxin B (1X MIC = 1 μg/mL), vancomycin (1X MIC = 16 μg/mL), and piperacillin (1X MIC = 8 μg/mL). NPN fluorescence was measured with excitation at 350 nm and emission at 420 nm. Values for individual biological replicates are shown. Statistical analysis was calculated using a one-way ANOVA followed by Dunnett’s multiple comparison test. ****, p < 0.0001. (B) Structures of vancomycin, oritavancin, teicoplanin, and dalbavancin. Structural compared to vancomycin are highlighted in red. (C) MIC assay with vancomycin and the three glycopeptides against PA14 in 10:90. (D) MIC assay with FITC-vancomycin against PA14 in 10:90. (E) Representative fluorescent microscopy images of cells treated with FITC-vancomycin. Scale bar = 5 μm. (F) Schematic for vancomycin (green circle) binding to the d-Ala-d-Ala moiety of the pentapeptide of lipid II. DacB and DacC remove the terminal d-Ala from the pentapeptide stem to form the tetrapeptide. The cell has both penta- and tetrapeptides and vancomycin can bind to the pentapeptides through H-bonding with the d-Ala-d-Ala residues. In the absence of DacB and DacC, the pool of pentapeptides increases, thus more vancomycin is required to inhibit transpeptidation. MIC assays of PA14, ΔdacB, ΔdacC, and ΔdacBΔdacC treated with (G) vancomycin and (H) piperacillin. All MIC assays were conducted in 10:90 and averaged from three independent biological replicates.

We next tested if other glycopeptides could inhibit P. aeruginosa growth in 10:90 (Figure 2B). Teicoplanin, oritavancin, and dalbavancin all lacked activity up to 64 μg/mL (Figure 2C). We also tested FITC-vancomycin, a fluorescent derivative, and saw no inhibition (Figure 2D). Using microscopy we saw no fluorescence in cells, indicating that the fluorescent analog could likely not cross the OM (Figure 2E). As controls, we included methicillin-resistant S. aureus MRSA USA 300 which was labeled as expected with FITC-vancomycin (Supplementary Figure S2). These results show that the activity is vancomycin-specific and sensitive to structural modifications.

Vancomycin is primarily bacteriostatic against P. aeruginosa but bactericidal during early exponential phase (Supplementary Figure S3A). At 1X MIC (16 μg/mL), where no growth was observed, cells grew similarly to the WT until early exponential phase, when there was a decrease in OD₆₀₀, indicative of bactericidal activity. Time-kill curves showed that at 1X MIC, there was a 1 log increase in CFU/mL compared to the starting inoculum after 24 h (Supplementary Figure 3B). At 4X MIC, vancomycin was bactericidal up to 8 h; however, regrowth occurred by 24 h. These results suggest that while vancomycin has activity under low-nutrient conditions, P. aeruginosa can develop resistance. Therefore, we investigated whether vancomycin was acting through its canonical mechanism of action, how it was taken up, and how P. aeruginosa could become resistant.

Loss of DacB and DacC Leads to Vancomycin Resistance

Vancomycin binds the d-Ala-d-Ala moiety of peptidoglycan pentapeptide stems, resulting in decreased crosslinking and impairment of cell wall integrity.⁴⁰ Therefore, we predicted that mutations that increase pentapeptide levels would result in decreased susceptibility to vancomycin. Previous studies reported that loss of the d,d-carboxypeptidases, DacB (penicillin binding protein 5, PBP5) and DacC (PBP4), increased the pool of pentapeptides in P. aeruginosa (Figure 2F).⁴¹ Deletion of dacB also upregulates expression of the chromosomally encoded β-lactamase, AmpC, increasing the MIC for β lactams such as piperacillin, although the degree of AmpC stimulation depends on the strain.^42,43 Consistent with these data, MICs for both vancomycin and piperacillin increased in a dacB dacC double mutant, suggesting that vancomycin was acting through its expected mechanism of action (Figure 2GH).

Spontaneous Vancomycin-Resistant Mutants Harbor Mutations in the CpxSR Two-Component System

To understand potential mechanisms of uptake and resistance to vancomycin in P. aeruginosa, we screened a library of nonredundant P. aeruginosa Mar2xT7 transposon insertion mutants at 64 μg/mL vancomycin (4X MIC) to identify mutants with increased resistance (Figure 3A). Only three mutants were reproducibly resistant: PA14_64520, PA14_12360, and PA14_12370, each with MICs of 64 μg/mL. PA14_64520 encodes a putative bacterioferritin contributing to iron homeostasis. PA14_12360 and PA14_12370 are tandem uncharacterized genes, suggesting a potential interaction between their protein products or possible polar effects on PA14_12370 from the transposon insertion in PA14_12360. The relationship between the products of these genes and their relation to vancomycin uptake or mechanisms of resistance was unclear; therefore, we also raised spontaneously vancomycin-resistant mutants in both liquid (up to 64 μg/mL) and solid medium formats (128–256 μg/mL) (Figure 3B). We sequenced the vancomycin-resistant mutants arising from liquid cultures, which all mapped to the two-component system sensor CpxS, causing a T163P mutation. Interestingly, all mutants raised on solid medium with 256 μg/mL vancomycin also had mutations in cpxS, suggesting that the CpxSR system is required for high-level vancomycin resistance. In E. coli, at least 10 mutations in the CpxS ortholog, CpxA, have been shown to be activating mutations. We aligned the AlphaFold2 structural models of CpxA and CpxS, which showed that CpxS T163P is located at transmembrane helix 2, a position similar to those of 5 CpxA activating mutations (Figure 4A,B). All of the other mutations identified are also located in positions similar to known CpxA activating mutations (Figure 4C).

Figure 3.

Identifying genes important for vancomycin resistance. (A) Replica plot of interquartile mean normalized growth of the PA14 transposon library screened at 64 μg/mL vancomycin in 10:90. Mutants with growth >2.5 standard deviations (SD) above the mean were considered as resistant. Validated resistant mutants are highlighted in red and labeled with their gene name. (B) Schematic for raising spontaneous vancomycin-resistant mutants in liquid and solid media—Vancomycin (VAN).

Figure 4.

Point mutations in CpxS confer resistance to vancomycin. (A) AlphaFold2 predictions of E. coli CpxA (green) and P. aeruginosa CpxS (blue) overlayed. Activating mutations in E. coli CpxA are highlighted in dark blue, whereas mutations in CpxS in spontaneous vancomycin-resistant mutants are indicated in orange. Structural alignment of (B) transmembrane helix 2 and (C) transmembrane helix 1 near the conserved H-box region between CpxA (green) and CpxS (blue). (D) MIC assay of PA14, ΔcpxR, CpxS T163P, and CpxS T163P ΔcpxR treated with vancomycin. (E) GFP-reporter assay of PA14 WT versus CpxS T163P using a promoterless control or P_mexA to drive GFP expression ns: not significant; ****: p < 0.0001. Statistics were calculated by one-way ANOVA followed by Dunnett’s multiple comparison test in Prism. Individual values for each biological replicate are shown. ΔcpxR and CpxS T163P ΔcpxR complemented with empty vector (pBADGr) or a WT copy of cpxR, and treated with vancomycin in 10:90 (F) with no arabinose or (G) 0.05% arabinose. All assays were conducted in 10:90, and results shown are averaged from three independent biological replicates.

We focused on CpxS T163P because it was recovered from both liquid and solid medium cultures. The CpxS T163P mutant was regenerated in a WT background via allelic exchange and we confirmed that it was resistant to vancomycin (Figure 4D) with MIC > 64 μg/mL. We also complemented a cpxS::Mar2xT7 mutant with WT PA14 CpxS or CpxS T163P and tested for vancomycin susceptibility (Supplementary Figure S4). The cpxS mutant had WT susceptibility to vancomycin with a MIC of 16 μg/mL. Expression of CpxS T163P, but not empty vector or WT CpxS, restored resistance (MIC = 64 μg/mL). In E. coli, mutations in the CpxAR system can be activating or repressing; therefore, to learn whether CpxS T163P is an activating mutation, we examined the expression of GFP under the control of the CpxR-responsive mexA promoter (Figure 4E). MexA is the periplasmic component of the MexAB-OprM efflux pump, and a previous study showed that CpxR binds upstream of mexA at a conserved CpxR box binding motif.²⁶ We compared the expression of GFP in WT PA14 and the CpxS T163P mutant using either a promoterless reporter or a reporter under control of P_mexA. GFP expression was significantly higher in CpxS T163P compared to the WT or a promoterless control, suggesting that CpxS T163P is an activating mutation.

To confirm that the effect of the CpxS T163P mutation was mediated via the response regulator CpxR, we deleted cpxR from WT and CpxS T163P and tested susceptibility to vancomycin (Figure 4D). ΔcpxR had WT susceptibility to vancomycin. Further, loss of ΔcpxR in the CpxS T163P mutant increased susceptibility to vancomycin to WT levels, suggesting that CpxS T163P activates CpxR to confer resistance. We could also complement the ΔcpxR mutants with cpxR in trans on an arabinose-inducible plasmid with a gentamicin resistance marker, pBADGr, which restored resistance to vancomycin in the CpxS T163P mutant even without the addition of arabinose, suggesting that low-level expression from the leaky promoter was sufficient (Figure 4F). Expression of CpxR in trans in ΔcpxR increased the MIC slightly in the presence of arabinose (0.05%), suggesting that there is also a dosage-dependent effect even without CpxS activation (Figure 4G), consistent with limited kinase-independent activation of response regulators by small-molecule phosphodonors.^44,45

Next, we tested whether activation of the CpxSR response was responsible for the observed antagonism between vancomycin + copper and vancomycin + iron. We conducted checkerboard assays with CpxS T163P, ΔcpxR, and CpxS T163P ΔcpxR challenged with vancomycin + CuCl₂ and vancomycin + FeCl₃ (Supplementary Figure S5). WT, ΔcpxR, and CpxS T163P ΔcpxR showed similar profiles of antagonism. CpxS T163P was resistant to vancomycin, but had increased sensitivity to copper compared to other strains. These results suggest that copper and iron antagonize vancomycin activity independently of the CpxSR two-component system.

Activation of CpxS Confers Resistance to Multiple Classes of β-Lactams

In E. coli, CpxA has been implicated in resistance to multiple classes of antibiotics;⁴⁶⁻⁴⁸ therefore, we tested whether CpxS T163P conferred resistance to antibiotics besides vancomycin. The penicillins, monobactams, and cephalosporins showed 4-fold or greater increases in MIC for CpxS T163P compared to the WT (Figure 5A). The mutant also had an elevated MIC of 8-fold for the siderophore antibiotic cefiderocol compared to WT. However, there was no change in MIC for the penems (meropenem and imipenem). Similarly, compounds that inhibit early steps of peptidoglycan synthesis (e.g., d-cycloserine and fosfomycin) or classes of antibiotics with non-cell-wall targets showed no change or only a 2-fold change in MIC. These results suggest that CpxS T163P primarily impacts the periplasmic steps of peptidoglycan synthesis, where vancomycin and β-lactams act. In E. coli, CpxA activation induces changes in cell wall crosslinking and expression of l,d-transpeptidases, which may contribute to resistance.²³ Additionally, the mechanism of resistance may be independent of MexAB-OprM, despite its upregulation in the CpxS T163P mutant, because the MICs for other classes of antibiotics that are substrates for efflux pumps were unchanged. CpxS T163P and WT had similar susceptibility to polymyxin B, suggesting that OM integrity was not impacted.

Figure 5.

CpxSR activation confers resistance to β-lactams. (A) Fold change in MIC of PA14 versus CpxS T163P for the panel of antibiotics tested. Fold changes 4-fold or more are considered resistant. (B) Screening of the WCC P. aeruginosa clinical isolate panel for susceptibility to 64 μg/mL vancomycin in 10:90. C0275, a resistant clinical isolate, is highlighted in red. (C) MIC assay of PA14, CpxS T163P, C0275, and C0098 treated with vancomycin in 10:90. Results for all assays are averaged from three independent biological replicates.

Our data suggested that various CpxS point mutations can activate the CpxSR system. This led us to hypothesize that clinical isolates of P. aeruginosa may harbor mutations in CpxS that differ depending on their diversity, prior antibiotic exposure, and the site of infection. We used vancomycin resistance at 64 μg/mL (4x MIC of WT) (Figure 5B) as a filter to uncover potential CpxS-activating mutations in a set of 96 clinical isolates. One isolate C0275 had ∼40% of control growth even at 64 μg/mL vancomycin, whereas most isolates had <20% of control growth, which was not observable by eye. C0275 had a MIC for vancomycin >64 μg/mL, similar to CpxS T163P, whereas WT PA14 had a MIC of 16 μg/mL. C0098 was used as a control and had similar susceptibility to WT PA14 (Figure 5C). All of the clinical isolates had CpxS sequences similar to those of WT PA14 or PAO1, suggesting that there are other resistance mechanisms that could contribute to the elevated vancomycin MIC in C0275.

WapR P164T Vancomycin-Resistant Mutant Has Altered LPS Composition and Increased Phage Sensitivity in Nutrient-Limited Media

We also investigated another spontaneously resistant mutant raised on 10:90 solid medium containing 128 μg/mL vancomycin. This isolate had a mutation in wapR leading to a P164T substitution. WapR is an enzyme involved in LPS core oligosaccharide biosynthesis that adds an l-rhamnose (α-1,3-linked) to the outer core to provide an attachment site for O-antigen polymerization. Mutants lacking WapR synthesize rough LPS without O-antigen. To understand the effects of the Pro to Thr mutation, we searched the AlphaFold2 structure of WapR using the Dali server.^49,50 The best-matched identified was a chrondroitin polymerase from E. coli K4 (KfoC; PDB: 2Z86).⁵¹ KfoC is a bifunctional glycosyltransferase that catalyzes the elongation of the chrondroitin chain involved in the synthesis of the extracellular layer of the bacterial capsule. WapR aligned with the A2 domain of K4CP bound to uridine-diphosphate glucuronic acid (UDP-GluUA) (Figure 6A). WapR P164 is located in the region that aligns with the K4CP binding cavity for UDP-GluUA, suggesting that the mutation in WapR may affect its function as an l-rhamnosyltransferase (Figure 6B).

Figure 6.

WapR P164T is resistant to vancomycin and azithromycin and has altered LPS profiles. (A) Overlay of the A2 domain of E. coli K4 KfoC (purple) (PDB: 2Z86) and an AlphaFold2 model of PA14 WapR (blue). UDP-glucuronic acid, which binds to KfoC, is highlighted in beige with heteroatoms colored differently. The blue sphere is Mn²⁺. P164 is highlighted in blue. (B) Zoomed-in view of the UDP-glucuronic acid binding site of KfoC overlayed with WapR. P164 is highlighted in red. MIC assays of PA14, WapR P164T, and wapR::Mar2xT7 treated with (C) vancomycin, (D) polymyxin B, (E) piperacillin, and (F) azithromycin. Results are averaged from three independent biological replicates. G. Silver stain of proteinase K-treated crude lysates for LPS from PA14, WapR P164T, and wapR::Mar2xT7. The three strains were grown on LB and 10:90 agar plates. H. Western blot of LPS preparations probed with the monoclonal antibody 5c-7-4 (LPS inner core specific).

To test this hypothesis, we tested the susceptibility of WT PA14, WapR P164T, and a wapR transposon mutant (wapR::Mar2xT7) to vancomycin (Figure 6C). As expected, the WapR P164T mutant had a MIC 4-fold greater than that of the WT. Surprisingly, the susceptibility of wapR::Mar2xT7 was similar to WT. This result suggests that the WapR P164T mutant may not be a loss-of-function mutation. As a control, we also tested the susceptibility of the mutants to polymyxin B (to assess OM integrity) and piperacillin (for peptidoglycan-acting antibiotics) and saw no differences (Figure 6DE). LPS structure modulates antibiotic resistance;¹⁰ therefore, we tested other high-molecular-weight antibiotics identified from our initial screen as having activity under nutrient-limited conditions using PA14, WapR P164T, and the wapR mutant (Figure 6F). Interestingly, the WapR P164T mutant was also 4-fold more resistant to azithromycin compared to the WT, similar to vancomycin. These results suggest that this mutation can also confer resistance to other classes of antibiotics in an LPS-dependent manner.

To further confirm that WapR P164T was not an inactivating mutation, we isolated crude LPS preparations for the three strains grown in LB or 10:90, separated the samples by SDS-PAGE, and visualized LPS by silver staining (Figure 6G). The LPS profile of cells grown in LB was similar to in 10:90 although there appeared to be less long-chain O antigen in 10:90. However, in both conditions, WapR P164T makes long-chain length O-antigen that was comparable to the WT while no polymers in that range were present for wapR::Mar2xT7. The wapR::Mar2xT7 mutant also failed to make short-chain-length LPS, consistent with previous reports.⁵² Additionally, there were differences in the migration pattern of the short + core LPS (Figure 6H). The LPS profiles of the WapR P164T and wapR::Mar2xT7 mutants showed that there were bands corresponding to the LPS core region that migrated faster compared to the WT. This observation is consistent with the fact that WapR modifies the LPS core. Therefore, we probed phenol/ethyl ether-extracted LPS preparations with monoclonal antibody 5c-7-4 that recognizes the inner core. For the WT, a single band was present, whereas the wapR transposon mutant had a single band of decreased molecular weight, corresponding to a truncated core, consistent with previous observations.⁵³ However, two bands were seen for the WapR P164T mutant—one higher-molecular-weight band corresponding to the WT core and a lower-molecular-weight band corresponding to the wapR::Mar2xT7 mutant truncated core. No differences were seen between cells grown in LB or 10:90. Together, these results suggest that the WapR P164T mutation leads to a heterogeneous core phenotype, between WT and the wapR::Mar2xT7 mutant.

The O antigen is important for antibiotic resistance and also acts as a receptor for phages to recognize suitable host cells. We tested PA14, WapR P164T, and wapR::Mar2xT7 for their susceptibilities to LPS-specific phages P2B9, D6, and E6 (Figure 7). As a control, we also tested three phages (B6, P2A3, and P2F10) that use type IV pili as a receptor, using a ΔpilA mutant which is unable to make the major pilin subunit as a negative control (Figure 7, Supplementary Figure S6). We spotted increasing dilutions of each phage on bacterial lawns grown on LB agar and looked for differences in susceptibility. All 6 phages plaqued on WT PA14. When the same set of phages was spotted on WapR P164T, we saw a similar susceptibility pattern to the WT; however, the plaques were clearer and larger. wapR::Mar2xT7 was resistant to phages P2B9, D6, and E6, but susceptible to the pilus-specific phages, B6, P2A3, and P2F10, consistent with the requirement for O-antigen binding for infection. The ΔpilA mutant was susceptible to phages P2B9, D6, and E6 but resistant to B6, P2A3, and P2F10. The larger plaque sizes for WapR P164T appear to be an LPS-independent phenomenon because the same phenotype was also observed for pilus-dependent phage B6. We also tested the CpxS T163P mutant for phage susceptibility. The titers were similar but the plaques were more turbid compared to WT, suggesting that CpxS activation may impact the efficiency of phage replication.

Figure 7.

WapR P164T has larger plaques in LB and increased phage susceptibility in 10:90. Phage plaquing assays with PA14, WapR P164T, wapR::Mar2xT7, ΔpilA, and CpxS T163P in LB and 10:90 agar. Each strain was treated with increasing dilutions of the LPS-specific phages (highlighted in red): P2B9, D6, and E6 and the pilus-specific phage (highlighted in blue) B6. Three biological replicates were conducted and representative plaquing assays are shown.

Despite the differences in plaque sizes between WapR P164T and WT, there were no differences in susceptibility. However, these initial phage assays were conducted in nutrient-replete LB, whereas our antibiotic susceptibility assays were done in 10:90. Therefore, we repeated the phage susceptibility assays in 10:90 agar. Interestingly, on that medium PA14 was resistant to LPS-specific phages D6 and E6 and had a 10⁵-fold decrease in susceptibility to P2B9. However, this was not the case for the WapR P164T mutant. We observed similar plaquing and susceptibility patterns of the LPS-specific phages on WapR P164T in LB and 10:90, although it became resistant to pilus-specific phage B6. The wapR::Mar2xT7, ΔpilA, and CpxS T163P mutants were resistant to all phages tested when grown in 10:90. The differences in phage susceptibility were not due to differences in growth as all strains grew similarly in 10:90 (Supplementary Figure S7). Altogether, these results show that the WapR P164T mutant has altered LPS composition and differences in susceptibility to vancomycin, azithromycin, and phages in rich versus nutrient-limited media.

Discussion

We found that P. aeruginosa is sensitized to the large natural product antibiotic vancomycin when grown in nutrient-limited conditions that may be more representative of the host environment.⁵⁴ Vancomycin activity was antagonized by various metals such as iron and copper, but calcium and magnesium—which help to stabilize the negative charge of the OM—were less important for resistance. One potential explanation for the antagonism between vancomycin and copper is that vancomycin can chelate the metal.^55,56 The functional groups that participate in chelation overlap with those that participate in H-bonding with the d-Ala-d-Ala pentapeptide. Thus, complex formation with copper may inhibit uptake and/or interaction of vancomycin with its target. Copper antagonism of vancomycin activity was independent of the CpxSR pathway, consistent with the chelation hypothesis, although further investigation to support this conclusion is necessary (Supplementary Figure S4). The antagonism of vancomycin activity by iron also suggests that there are alternate mechanisms of resistance yet to be identified but confirms that iron limitation potentiates vancomycin activity.

Our screen of the P. aeruginosa PA14 transposon library to identify genes involved in vancomycin resistance yielded only three mutants in poorly characterized genes. This result suggested that inactivating mutations are not the primary mechanism of P. aeruginosa resistance to vancomycin. Instead, isolation of spontaneously resistant mutants led to the identification of multiple point mutations in CpxS, suggesting that the sensor can be activated through specific substitutions such as T163P, similar to CpxA in E. coli.⁵⁷ Vancomycin resistance conferred by CpxS T163P was dependent on CpxR, as cpxR deletion reduced susceptibility to WT levels.

Detailed examination of the consequences of CpxSR activation will be necessary to understand the transcriptional responses that result in vancomycin resistance. For example, activation of CpxSR may increase the expression of l,d-transpeptidases, implicated in β-lactam resistance because they bypass the effects of d,d-transpeptidase inhibition.⁵⁸ In P. aeruginosa, 3,3-crosslinking of peptidoglycan catalyzed by l,d-transpeptidases increases in low-nutrient conditions⁵⁹ In E. coli, activation of the CpxS homologue CpxA increases expression of l,d-transpeptidase LdtD.²³ The l,d-transpeptidases are not well characterized in P. aeruginosa, although there are some candidates (e.g., PA14_27180 and PA14_54810) predicted to have the YkuD-like fold characteristic of this family.⁶⁰ In E. coli, CpxR also indirectly decreases expression of dacC and genes involved in LPS biosynthesis;⁶¹ but whether this changes pentapeptide levels is unclear. Some CpxA activating mutants have increased abundance, whereas others do not.²³ Overall, changes in peptidoglycan composition and remodeling may be one way to confer resistance to both vancomycin and β-lactams.

Interestingly, activation of CpxSR through the T163P mutation did not increase resistance to other classes of antibiotics. In E. coli CpxAR is involved in fosfomycin resistance, by repressing transporters of the antibiotic.⁴⁶ These results suggest that the regulatory consequences of activation differ between the two species. P. aeruginosa CpxS T163P was identified previously, in a study investigating the efficacy of sequential antibiotic treatment protocols.²⁴ They found that pretreating cells with carbenicillin for 15 min increased susceptibility to gentamicin compared to treatment with gentamicin alone. However, among the mutations that conferred resistance to this combination was CpxS T163P, which based on our work may have resulted in increased carbenicillin resistance.

Our finding that CpxS mutations confer resistance to β-lactams including the recently FDA-approved siderophore-cephalosporin cefiderocol, but not the penems, has significant implications for the treatment of P. aeruginosa infections. The target of β-lactams is the PBPs; however, the penem subclass also inhibits the l,d-transpeptidases, which may explain why the CpxS T163P mutant remains susceptible to meropenem and imipenem.⁵⁸ Further investigation into the mechanism will be informative. These results, when combined with studies to find additional CpxS-activating mutations, could help guide antibiotic selection for the treatment of P. aeruginosa infections. For example, in polymicrobial infections, methicillin-resistant Staphylococcus aureus (MRSA) is often associated with P. aeruginosa in chronic wounds and the lungs of cystic fibrosis patients, and vancomycin is used in the treatment of MRSA.^54,62−64 However, this work shows that vancomycin can also inhibit P. aeruginosa under nutrient-limiting conditions and select for mutations in CpxS that confer multidrug resistance. Further studies examining the effects of vancomycin on co-cultures will be informative to see if this phenomenon could be clinically relevant. We note that the utility of vancomycin in cystic fibrosis lung infections caused by P. aeruginosa may be limited due to increased iron concentrations in that environment.⁶⁵⁻⁶⁹

Previous work on E. coli showed that cold stress, rather than nutrient limitation, sensitized cells to vancomycin and that truncations in the LPS core restored resistance.⁷⁰ They hypothesized that cold stress negatively affected OM integrity, leading to increased susceptibility to vancomycin, while mutants in LPS biosynthesis were better at maintaining the barrier. Resistance was correlated with truncations with the LPS core or production of heterogeneous LPS populations. In contrast, our work showed that wapR::Mar2xT7, which fails to produce short- or long-chain LPS (Figure 6G) and has a truncated core (Figure 6H), had WT levels of vancomycin susceptibility. These results suggest that in P. aeruginosa, core truncations alone were insufficient to confer resistance. However, intermediate LPS profiles, as seen with WapR P164T, correlated with resistance. P. aeruginosa makes two distinct O-antigens, called A-band or common polysaccharide antigen (CPA), and B-band or O-specific antigen (OSA). However, PA14 has a single amino acid substitution G20R in wbpX that abolishes CPA production.⁸ Therefore, only OSA is produced by the WT. Another possibility is that the OSA is differentially attached to the core compared to WT LPS. More detailed studies of LPS composition in these strains will help to clarify the mechanism of resistance. Interestingly, WapR P164T was also more resistant to the macrolide antibiotic, azithromycin. Like vancomycin, the activity of azithromycin on P. aeruginosa is nutrient-dependent and it directly interacts with LPS by displacing divalent cations.^13,71 However, since the WapR P164T mutant and WT were equally susceptible to polymyxin B, resistance is likely not due to altered OM integrity.

Interestingly, one study reported that a single amino acid mutation in the E. coli O-antigen ligase WaaL conferred vancomycin resistance in both WT and mutants with defects in the OM.¹⁴ Resistance was due to modification of LPS with peptidoglycan subunits that bound vancomycin and reduced its uptake.¹⁴ However, this does not appear to be the case with WapR P164T, as we saw no labeling with FITC-vancomycin. WapR P164T was less susceptible to vancomycin and azithromycin, but more susceptible to LPS-specific phages in nutrient-limited 10:90. This combination of phenotypes may be due to altered LPS composition, changes in levels of expression, or exposure of secondary phage receptors. Further investigation into the effects of nutrient limitation on phage replication cycle and burst size will be informative in understanding the mechanism for the observed differences in phage susceptibility. These observations are important considerations for the use of phages as antibiotic alternatives, as many sites of infection can be limiting for key nutrients such as iron.⁷²

Conclusions

In conclusion, we showed that vancomycin—but not other glycopeptides—has antimicrobial activity against P. aeruginosa in low-nutrient conditions. Vancomycin-resistant mutants harbored activating mutations in CpxS that also conferred resistance to β-lactams, including cefiderocol, but not the penems. A mutation in WapR which conferred resistance to vancomycin and azithromycin but increased susceptibility to phages under nutrient-limited conditions was also identified. Our study highlights how screening in low-nutrient conditions can reveal novel activity for existing antibiotics and shows how investigation of resistance mechanisms may help to guide antibiotic therapies for P. aeruginosa.

Methods

Media and Growth Conditions

All bacteria were cultured overnight in lysogeny broth (LB) at 37 °C with 200 rpm shaking. Strains with pBADGr and P-gfp, cultures were supplemented with 15 μg/mL gentamicin. Subcultures (1:500 dilution of overnight cultures) were grown in 10:90 supplemented with or without gentamicin for 3–4 h until at least OD₆₀₀ at 37 °C with 200 rpm shaking. Arabinose was made as a 20% stock solution in 10:90 and filtered through 0.2 μm filters (Fisher Scientific) before diluting into 10:90 for growth assays.

Bacterial Strains, Phage, and Plasmids

Bacterial strains, phage, and plasmids are listed in Supplementary Table S2.

Molecular Biology

See Supplementary Table S3 for all primers used in this study. All procedures were conducted as previously described.^28,30

Compounds

Supplementary Table S4 lists all compounds used in this study. All antibiotic powders were stored at 4 °C. Stock solutions were stored at −20 °C. Compounds were solubilized in dimethyl sulfoxide or DI H₂O for assays.

MIC and Checkerboards Assays

Broth microdilution MIC assays and checkerboards were conducted as previously described.²⁷⁻³⁰

NPN Assay

The NPN assay was conducted as previously described with modifications.⁷⁴ Briefly, overnight cultures of PA14 in LB were subcultured (1:500 dilution) into 50mL of 10:90 and incubated for 3 h at 37 °C with shaking (200 rpm). Cells were harvested by centrifugation at 3,000G for 5 min and washed three times with PBS. Cells were resuspended in 10:90 to a final OD₆₀₀ of 0.1 with 5 μM carbonyl cyanide m-chlorophenyl hydrazone and 15 μM NPN. Cells were aliquoted into 96-well black plates with flat clear bottoms (Corning). Antibiotics or vehicle controls were added at 75x the final concentration and fluorescence was read immediately on a BioTek Neo plate reader at excitation and emission wavelengths of 350 and 420 nm, respectively.

Fluorescence Microscopy

Microscopy was conducted as previously described.^28,30 Briefly, overnight cultures of PA14 were subcultured in fresh 10:90 (1:500 dilution) and grown at 37 °C with shaking (200 rpm) for 3–4 h. Cells were harvested by centrifugation and washed 3× with 1× PBS then incubated with 10 μM FITC-vancomycin for 30 min at 37 °C with shaking (200 rpm). Cells were harvested by centrifugation and washed 3× with 1× PBS before spotting on a 1% agarose pad in 10:90 and imaged with a Nikon A1 confocal microscope through a Plan Apo 60x (NA = 1.40) oil objective and acquired with a Nikon NIS Elements Advanced Research (V. 5.11.01 64-bit). MRSA USA 300 was grown, labeled, and imaged under the same conditions.

Transposon Library Screening and WCC Vancomycin Susceptibility Testing

All mutants from the nonredundant PA14 transposon insertion library⁷⁵ were transferred into 96-well plates containing 150 μL/well liquid LB media and incubated for 16 h at 37 °C with shaking at 200 rpm. Each mutant was then diluted 1:200 into 10:90 deferrated using FEC-1⁷⁶ to keep the iron concentration consistent between batches of media and incubated for a further 4 h. Mutants were then diluted 1:200 into deferrated 10:90 LB containing 64 μg/mL vancomycin (two replicates) and media without vancomycin (one replicate). Plates were incubated for 16 h at 37 °C with shaking at 200 rpm, then OD₆₀₀ for each well was recorded. OD₆₀₀ values were normalized to the interquartile mean of each plate and each well position.⁷⁷ Growth of >2.5 SD above the mean in the presence of vancomycin was considered a hit. The P. aeruginosa WCC was tested in a similar manner except untreated 10:90 was used.

Spontaneous Vancomycin-Resistant P. aeruginosa

Spontaneously resistant mutants were selected in liquid and solid media. For liquid media, a PA14 overnight culture was diluted 1:500 into 5 mL cultures of 10:90 LB media containing 16 μg/mL vancomycin, or equivalent volumes of sterile H₂O. The cultures were incubated at 37 °C with 200 rpm shaking and inspected daily. When turbidity (growth) was observed, the culture was diluted 1:250 into fresh media containing 2× concentration of compounds. The DIBI concentration was held at 16 μg/mL as higher concentrations prevented growth. Cultures were passaged until growth was observed at 64 μg/mL vancomycin then streaked on LB agar. Single colonies were selected and their genomic DNA sequenced.

For solid media, 10⁷ cells of PA14 subcultured in 10:90 were plated onto 10:90 + 1.5% agar containing 128 and 256 μg/mL in triplicates. Plates were incubated at 37 °C for 48h where colonies appeared. Sixteen colonies in total from the 128 and 256 μg/mL plates were patched onto fresh 10:90 plates with and without vancomycin. For example, a colony that grew from a plate containing 128 μg/mL vancomycin would be patched onto a plate without vancomycin and a plate containing 128 μg/mL vancomycin. Mutants that grew on both plates were streaked for single colonies for MIC testing to confirm resistance to vancomycin (≥4-fold MIC). This process was repeated for the ΔcpxR mutant.

Genomic DNA Isolation and Sequencing

Genomic DNA of vancomycin-resistant mutants was isolated using Promega Wizard Genomic Isolation Kit. Samples were sent to SeqCenter (Pittsburgh) for Illumina sequencing. FASTA files were processed using FASTQ Groomer, Trimmomatic, FASTQ-interlacer, and FASTQ de-interlacer. Mutations were identified using breseq by comparing to the reference genome (accession number GCF_000404265.1).⁷⁸

LPS Isolation (Whole Cell and Phenol/Ethyl Ether Extraction)

LPS from PA14, WapR P164T, and wapR::Mar2xT7 were isolated as previously described with some modifications.⁷⁹ Cells from overnight cultures in LB or 10:90 1.5% agar plates were collected and resuspended in PBS. The OD₆₀₀ was standardized to 2.0, and cells were collected again by centrifugation. Cells were resuspended in 150 μL of lysing buffer (2% SDS, 4% 2-mercaptoethanol, 10% glycerol, 0.1 M Tris-HCl, pH 6.8) and boiled for 10 min. After cooling to room temperature, proteinase K (NEB) (10 μL of 20 mg/mL) was added and incubated at 60 °C for 1 h. Proteinase K-treated preparations were used for silver staining.

For LPS core samples detected by Western blot, prewarmed phenol solution (60 °C) containing 90% phenol (Sigma), 0.1% 2-mercaptoethanol, and 0.2% 8-hydroxyquinoline (Sigma) was added to each lysate (1:1 v/v phenol:lysate) and incubated for 15 min at 60 °C. The tubes were then incubated on ice for 10 min before centrifugation (21,000g for 5 min). The top aqueous layer was transferred to a fresh tube and 500 μL of ethyl ether solution containing 20 mM Tris-HCl, 1 mM EDTA, pH 8.0 was added. Tubes were centrifuged for 1 min at 21,000g and the top ethyl ether layer was removed by aspiration. An equal volume of 2X SDS-PAGE loading buffer was added. All LPS samples were stored at −20 °C.

SDS-PAGE and Western blot

SDS-PAGE and Western blot were conducted as previously described^28,30 with some modifications. For the detection of the inner LPS core, isolated LPS was separated on a 15% SDS-PAGE gel at 90 V for 10 min followed by 200 V for 1 h. After transferring the LPS to a nitrocellulose membrane and blocking for 1 h with 5% skim milk, the blot was incubated overnight with mAb 5c-7-4 (specific for the inner core; 1:100 dilution in PBS).⁸⁰ Isolated LPS was also separated on a 12.5% SDS-PAGE gel at 120 V for 1.5 h with the same transferring and blocking steps. The next morning, the blot was washed 3× with PBS for 5 min/wash and incubated with 1:500 α-mouse-alkaline phosphatase in PBS for 1 h. The blot was washed 3× with PBS for 5 min/wash and rinsed briefly with DI H₂O before detection with BCIP and NBT. Bands appeared after 5–15 min and imaged with an Azure 400 imaging system.

Silver Stain

Silver stain was conducted as previously described with modifications.⁷⁹ 4 μL of each LPS preparation were loaded onto a 12.5% polyacrylamide gel and separated for 1.5 h at 120 V. The gel was then incubated in EtOH:acetic acid (40%:10%) overnight on a shaking platform at room temperature. The next morning cells were treated with periodic acid solution (40% ethanol: 10% acetic acid: 0.7% periodic acid) for 30 min then washed 3× with DI H₂O for 10 min each. Silver stain solution (2 mL NH₄OH, 28 mL 0.1M NaOH, 115 mL DI H₂O) was added, and the gel was incubated for 30 min. The gel was then washed two times (6 min/wash). Overwashing led to destaining of the high-molecular-weight LPS. Gels were developed immediately with developing solution (100 μL 37% formaldehyde, 10 mg of citric acid, 200 mL of DI H₂O) until bands could be visualized. Development was stopped by putting the gel in 10% acetic acid. Gels were imaged using an Azure 400 imaging system.

Phage Isolation, Purification, and Plaquing Assays

Environmental phage lysates isolated and amplified using P. aeruginosa PA14 were serially diluted (10^–1 to 10^–8) in phage buffer (68 mM NaCl, 10 mM Tris-HCl (7.5), 10 mM MgSO₄, 10 mM CaCl₂). Five microliters of each dilution was spotted onto the prepared plates. Phage spots were air dried for 10 min with the lid on. Plates were incubated inverted for 18 h at 37 °C. Phage lysate dilutions that produced visible plaques were used for phage plaque purification. Ten microliters of serially diluted phage lysate and 100 μL of culture were added to 10 ml of top agar (0.6% agar) and poured onto pre-set 1% LB agar. Plates were incubated inverted for 18 h at 37 °C. Plates that had 2–30 plaques were used for plaque purification. A pipette was used to touch the center of a plaque, followed by resuspension in 100 μL of phage buffer. This process was repeated two to three times or until uniform plaque morphology was observed.

Phage plaque assays were conducted as previously described.⁸¹ Briefly, bacteria were grown at 37 °C overnight then subcultured in LB or 10:90 (1:100 dilution) and cultured until the OD₆₀₀ reached 0.3 for LB and 0.2 for 10:90. One hundred μL of the LB subculture or 200 μL of the 10:90 culture was mixed with 10 mL of LB + 0.6% agar or 10:90 + 0.6% agar. Phage stocks were serially diluted 10-fold in phage buffer (68 mM NaCl, 10 mM Tris-HCl (7.5), 10 mM MgSO₄, 10 mM CaCl₂) and 5 μL of each dilution was spotted. Plates were allowed to dry with the lid on for 10 min, inverted, and incubated at 37 °C overnight. The next day plates were imaged. Each experiment was repeated at least three times, representative plates are shown.

Structural Comparisons and Phylogenetic Analyses

High-confidence structural models of CpxA, CpxS, and WapR were generated using AlphaFold2.⁵⁰ The AlphaFold2 model of WapR was used to look for similar proteins using Dali.⁴⁹ The structure of KfoC was retrieved from the Protein Data Bank (PDB: 2Z86).⁵¹ Structural alignments were conducted using ChimeraX.⁸²

Glossary

Abbreviations Used

NPN

1-N-phenylnaphthylamide

CPA

common polysaccharide antigen

DIBI

denying iron to bacterial infections

LPS

lipopolysaccharide

MRSA

methicillin-resistant Staphylococcus aureus

MIC

minimal inhibitory concentration

OSA

O-specific antigen

OM

outer membrane

PBP

penicillin binding protein

UDP-GluUA

uridine-diphosphate glucuronic acid

WT

wild type

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsinfecdis.3c00167.

• Checkerboard assays with PA14 treated with vancomycin and MgCl₂ or CaCl₂ (Figure S1); fluorescence microscopy images of WapR P164T and MRSA USA 300 treated with FITC-vancomycin (Figure S2);growth curve and time-kill assays of PA14 treated with vancomycin in 10:90 (Figure S3); PA14 cpxS::Mar2xT7 expressing CpxS T163P in 10:90 + 0.25% arabinose treated with vancomycin (Figure S4); checkerboard assays with ΔcpxR, CpxS T163P, and CpxS T163P ΔcpxR with vancomycin and CuCl2 or FeCl3 (Figure S5); WapR P164T treated with pilus-specific phages in LB and 10:90 (Figure S6);growth curves of PA14, WapR P164T, wapR::Mar2xT7, ΔpilA, and CpxS T163P in 10:90 (Figure S7); antibiotics screened under low-nutrient conditions against P. aeruginosa (Table S1);strain list (Table S2); primer list (Table S3); and compound source list (Table S4) (PDF)

Author Contributions

D.C.K.C. and L.L.B. designed the experiments and wrote the draft. D.C.K.C. performed the fluorescence microscopy, silver stain, western blot, GFP-promoter assays, phage assay, clinical isolate screening, and analysis experiments. K.D. and D.C.K.C. conducted the transposon mutant screen. D.C.K.C. designed primers, and D.C.K.C. and K.D. made the mutants. D.C.K.C. and K.D. conducted the checkerboards and antibiotic susceptibility assays, and raised spontaneous resistant mutants. M.F. assisted with the primary screen for antibiotics with increased activity under low-nutrient conditions. H.H. made the pilA mutant, isolated phages from environmental samples, and assisted with phage plaquing and silver stain assays. I.Q. purified phage P2B9 for phage studies.

The authors declare no competing financial interest.