ABSTRACT
Pseudomonas aeruginosa is a biofilm-forming opportunistic pathogen and is intrinsically resistant to many antibiotics. In a high-throughput screen for molecules that modulate biofilm formation, we discovered that the thiopeptide antibiotic thiostrepton (TS), which is considered to be inactive against Gram-negative bacteria, stimulated P. aeruginosa biofilm formation in a dose-dependent manner. This phenotype is characteristic of exposure to antimicrobial compounds at subinhibitory concentrations, suggesting that TS was active against P. aeruginosa. Supporting this observation, TS inhibited the growth of a panel of 96 multidrug-resistant (MDR) P. aeruginosa clinical isolates at low-micromolar concentrations. TS also had activity against Acinetobacter baumannii clinical isolates. The expression of Tsr, a 23S rRNA-modifying methyltransferase from TS producer Streptomyces azureus, in trans conferred TS resistance, confirming that the drug acted via its canonical mode of action, inhibition of ribosome function. The deletion of oligopeptide permease systems used by other peptide antibiotics for uptake failed to confer TS resistance. TS susceptibility was inversely proportional to iron availability, suggesting that TS exploits uptake pathways whose expression is increased under iron starvation. Consistent with this finding, TS activity against P. aeruginosa and A. baumannii was potentiated by the FDA-approved iron chelators deferiprone and deferasirox and by heat-inactivated serum. Screening of P. aeruginosa mutants for TS resistance revealed that it exploits pyoverdine receptors FpvA and FpvB to cross the outer membrane. We show that the biofilm stimulation phenotype can reveal cryptic subinhibitory antibiotic activity, and that TS has activity against select multidrug-resistant Gram-negative pathogens under iron-limited growth conditions, similar to those encountered at sites of infection.
INTRODUCTION
Bacterial pathogens are rapidly evolving resistance to available antibiotics, creating an urgent need for new therapies. Gram-negative bacteria are particularly challenging to treat because their outer membrane limits the access of many drugs to intracellular targets (1). Resistance arises when bacteria accumulate target mutations, acquire specific resistance determinants, increase drug efflux, and/or enter antibiotic-tolerant dormant or biofilm modes of growth (2). Biofilms consist of surface-associated bacteria surrounded by self-produced extracellular polymeric substances (EPS). Biofilm architecture allows for the development of phenotypic heterogeneity that leads to variations in susceptibility as well as the formation of drug-tolerant persister cells (3). Approaches with the potential to preserve current antibiotics include combining them with biofilm inhibitors, resistance blockers (e.g., ampicillin with clavulanic acid or piperacillin with tazobactam), efflux inhibitors (e.g., phenylalanine-arginine beta-naphthylamide [PAβN]), or outer membrane permeabilizers, or coupling them to molecules such as siderophores that are actively imported, so-called “Trojan horses” (4).
Among the bacterial pathogens deemed most problematic by the World Health Organization is the Gram-negative opportunist Pseudomonas aeruginosa (5). It infects immunocompromised patients, particularly those with indwelling medical devices, and it is a major problem for people with severe burns or cystic fibrosis (6). It is intrinsically resistant to many antibiotics and forms biofilms, further enhancing its ability to evade therapy (7). The low permeability of its outer membrane and expression of multiple efflux pumps that extrude a wide variety of substrates, coupled with its propensity to form biofilms, limit the repertoire of effective anti-Pseudomonas antibiotics (8–10). Here, with the aim of identifying potential modulators of P. aeruginosa biofilm formation, we screened a collection of bioactive molecules, including previously FDA-approved off-patent drugs, at 10 μM. During this work, we identified 60 growth inhibitors, plus an additional 60 molecules that stimulated biofilm formation beyond our arbitrary cutoff of 200% of the vehicle control, a phenotype associated with exposure to sub-MIC antibiotics (11, 12). The investigation of one such stimulatory compound, thiostrepton (TS), revealed that it had low-micromolar activity against P. aeruginosa in minimal medium. Through a series of investigations, we showed that TS enters the cell by exploiting iron-dependent uptake pathways. These data show that the biofilm stimulation phenotype can reveal cryptic antibiotic activity when concentrations are too low (or growth conditions not conducive) to inhibit growth.
RESULTS
Thiostrepton stimulates P. aeruginosa biofilm formation.We used a previously described P. aeruginosa biofilm assay (13) to screen a bespoke collection of 3,921 bioactive molecules that includes ∼1,100 FDA-approved off-patent drugs and antibiotics (14). The molecules were screened in duplicate at 10 μM in a dilute growth medium consisting of 10% lysogeny broth (LB) and 90% phosphate-buffered saline (here, 10:90) to identify molecules capable of modulating biofilm formation. This medium was chosen to minimize the amount of biofilm formed in the presence of the vehicle control, so that molecules stimulating biofilm formation could be more easily identified. Multiple studies showed that subinhibitory concentrations of antibiotics from a variety of chemical classes and with different mechanisms of action (MOA) stimulate P. aeruginosa biofilm formation, although the specific pathways underlying this response remain unclear (11, 12, 15–18). The hits were divided into planktonic growth inhibitors (60 compounds), biofilm inhibitors (defined as those resulting in ≤50% of the vehicle control biofilm [8 compounds]), or biofilm stimulators (those resulting in ≥200% of the vehicle-treated control biofilm [60 compounds]) (Table S1). The hit rate of ∼3% was relatively high for a primary screen, but all the molecules in this curated collection have biological activity. The hits belonged to a variety of chemical classes and included drugs with nominally eukaryotic targets.
Among the molecules in our screen that stimulated biofilm formation was the thiopeptide antibiotic thiostrepton (TS) (Fig. 1A). This response intrigued us because TS was reported to be ineffective against Gram-negative bacteria (19, 20), likely due to the impermeability of the outer membrane (OM) to large hydrophobic compounds. In light of the need for new therapeutics for P. aeruginosa, and the status of TS as a previously approved drug with a potentially shorter development pathway, we further investigated its possible anti-Pseudomonas activity. In dose-response experiments in 10:90 medium, biofilm levels increased, while planktonic cell density decreased with increasing TS concentrations to 10 μM (17 μg/ml), the maximum that could be tested due to its limited solubility (Fig. 1B).
Thiostrepton stimulates P. aeruginosa biofilm formation. (A) Structure of thiostrepton (TS). (B) TS-stimulated biofilm formation (absorbance of eluted crystal violet at 600 nm, plotted as percentage of the DMSO control on a log10 scale) of P. aeruginosa PAO1 and decreased planktonic cell density (optical density at 600 nm, plotted as percentage of the DMSO control on a log10 scale) in 10:90 medium in a dose-dependent manner, up to its maximum soluble concentration of 10 μM (17 μg/ml). No turbidity was visible below 20% of the control. (C) In VBMM, PAO1 biofilm formation was stimulated by TS, while planktonic cell density decreased below the level of visual detection (20% of the control) at concentrations above 0.63 μM. Assays were performed at least 3 times in triplicate. ****, P < 0.0001.
Growth in minimal medium increases the susceptibility of P. aeruginosa to TS.Environmental conditions can modulate the expression or essentiality of antibiotic targets or alter the availability of particular nutrients (21), leading to changes in susceptibility. We hypothesized that the biofilm response of P. aeruginosa to TS may be the result of nutrient deficiency in 10:90 medium, which was more limiting to P. aeruginosa growth than was M9 minimal medium (see Fig. S1 in the supplemental material). Consistent with this idea, the growth of P. aeruginosa in nutrient-rich Mueller-Hinton broth (MHB) reduced susceptibility to TS (Fig. S2). Growth rates in Vogel-Bonner minimal medium (VBMM) in the absence of TS were similar to those in 10:90 medium (Fig. S1), but in the presence of TS, planktonic cell density was below the level of detection (optical density at 600 nm [OD600], <0.07) at concentrations above ∼1.25 μM (Fig. 1C). These data suggested that nutrient limitation enhances the susceptibility of P. aeruginosa to TS.
The ribosomal methyltransferase Tsr protects P. aeruginosa against TS.The established MOA for TS antibacterial activity is the inhibition of protein translation through direct binding to bacterial ribosomes (22). However, because TS also has antiparasitic and antineoplastic activities (23, 24) we considered the possibility that it might inhibit P. aeruginosa growth in a novel way. To validate the MOA, we expressed a resistance gene, tsr, from a plasmid in P. aeruginosa strains PAO1 and PA14. tsr encodes a 23S rRNA methyltransferase used by TS producer Streptomyces azureus to prevent self-intoxication (25). Tsr methylates the conserved A1067 residue of 23S rRNA, impairing the binding of TS to its target (26). The expression of tsr in trans increased TS resistance of both PAO1 and PA14 compared to vector-only controls (Fig. 2). PAO1 was resistant up to the maximum soluble TS concentration of 10 μM, while resistance of PA14 was significantly increased compared to the control, although not to the same extent as PAO1. These results suggest that TS inhibits growth via its canonical MOA of ribosome binding, implying that it crosses the P. aeruginosa OM to access the bacterial cytoplasm.
Expression of Tsr in trans reduces susceptibility of P. aeruginosa to thiostrepton. Expression of the tsr gene from Streptomyces azureus in trans from pUCP20 in two strains of P. aeruginosa reduced susceptibility to TS in VBMM, suggesting that it inhibits growth via its canonical mode of action, disrupting translation. Left, growth of PAO1 (OD600 plotted as percentage of the DMSO control); right, growth of PA14. No turbidity is visible below 20% of the DMSO control. Each assay was performed at least 3 times in triplicate. ****, P < 0.0001.
TS susceptibility increases in the presence of iron chelators.To understand the reason for increased TS susceptibility of P. aeruginosa in VBMM compared to that in 10:90 medium, we considered the differences in nutrient availability between the two medium types. The primary carbon source in 10% LB is amino acids (27), while the carbon source in VBMM is citrate (28). Citrate can chelate divalent cations, including calcium and magnesium, which are important for OM integrity. We hypothesized that this chelation effect may increase OM permeability. To stabilize the OM, we repeated the dose-response assay in VBMM supplemented with 100 mM MgCl2 but saw no effect on susceptibility (Fig. S3A). Since TS is a thiopeptide, we next hypothesized that amino acid limitation during growth in VBMM may increase the uptake of TS, leading to growth inhibition. To test this, we supplemented VBMM with 0.1% Casamino Acids but saw no change in TS susceptibility (Fig. S3B). Further, simultaneous deletion of components of the Opp (Npp) peptide transport system, exploited by other peptide antibiotics for entry (29, 30), and a homologous system, Spp, had no effect on TS susceptibility (Fig. S3C).
We next considered that VBMM was more iron limited than was 10:90 medium, which contains trace iron from yeast extract and peptone. Under iron limitation, bacteria secrete siderophores into the extracellular milieu to scavenge the metal. Specialized receptors then transport siderophore-iron complexes back into the cell. Some antimicrobials, including sideromycins, pyocins, and bacteriocins, use siderophore receptors to access intracellular targets (31–34), and we hypothesized that TS may use this strategy. We compared P. aeruginosa PAO1 grown in 10:90 medium with increasing concentrations of TS alone (Fig. 3A) or with 0.1 μM ethylenediamine-N,N′-bis(2-hydroxyphenylacetic acid) (EDDHA), a membrane-impermeable iron chelator (35) (Fig. 3B). The addition of EDDHA shifted biofilm stimulation and growth inhibition to lower concentrations of TS than with 10:90 medium alone, while supplementation of 10:90 medium plus 0.1 μM EDDHA with 100 μM FeCl3 increased planktonic cell density and reduced biofilm stimulation (Fig. 3C). These data suggest that TS susceptibility is inversely proportional to iron availability, and that TS may exploit siderophore receptors to cross the OM of P. aeruginosa. Consistent with the active import of this antibiotic, mutants lacking the major efflux pumps of P. aeruginosa had near-wild-type susceptibility to TS (Fig. S4).
Thiostrepton activity is potentiated by iron chelators and serum. Biofilm stimulation by TS in 10:90 medium increased with addition of 0.1 μM EDDHA, a cell-impermeant iron chelator, while the further addition of 100 μM FeCl3 increased the concentration of TS required for biofilm stimulation and growth inhibition. At 20% or less of the control, no turbidity is visible. (A) PAO1 growth (OD600) and biofilm (absorbance of CV at 600 nm) in 10:90 medium alone plotted as percentage of the DMSO control on a log10 scale. (B) 10:90 medium plus 0.1 μM EDDHA. (C) 10:90 medium plus 0.1 μM EDDHA and 100 μM FeCl3. Average growth and biofilm values for the lowest and highest concentrations of TS tested are shown in green and blue numerals, respectively, to emphasize shifts caused by the manipulation of iron availability. (D to G) TS activity against PAO1 was also potentiated in 10:90 medium by FDA-approved iron chelators, deferiprone (D) and deferasirox (E), or by 10% heat-inactivated mouse (F) or human (G) serum. Checkerboard assays were plotted as the percent growth of the DMSO control (0, 0 μM at lower left). The highest concentrations of DFP (3680 μM) and DSX (1370 μM) are each equal to 512 μg/ml. Each assay was performed at least 3 times. ***, P < 0.001; ****, P < 0.0001.
The poor solubility of TS has hampered its development as a therapeutic, but these data suggest that its effective concentration could be reduced in the presence of iron chelators. We tested the FDA-approved iron chelators deferiprone (DFP) and deferasirox (DSX) for potential synergy with TS. Checkerboard assays revealed that while neither chelator had activity against P. aeruginosa, both potentiated TS activity (Fig. 3D and E) at concentrations well below those used to safely treat patients, up to 28 mg/kg of body weight/day for DSX or 99 mg/kg of body weight/day for DFP (36). These effects are specific for TS, as DSX failed to synergize with other antibiotics that do not depend on iron availability for uptake (Fig. S5).
High-affinity iron chelation by transferrin, hemoglobin, and lactoferrin is a common strategy used by mammals to restrict the growth of microorganisms (37, 38). We investigated whether serum could also potentiate TS activity. Interestingly, the addition of 10% heat-inactivated mouse or human serum to 10:90 medium markedly decreased the concentration of TS required to inhibit growth, regardless of the presence of DSX (Fig. 3F and G), suggesting that the levels of iron were already very low under these conditions.
TS hijacks pyoverdine receptors FpvA and FpvB.To identify the route of iron-limitation-dependent TS entry into P. aeruginosa, we tested the susceptibility of mutants from the ordered PA14 transposon library (39) with insertions in genes encoding known siderophore receptors, as well as mutants with insertions in uncharacterized OM proteins homologous to siderophore receptors. In VBMM, most mutants had TS MICs (defined as the concentration of TS that resulted in an OD600 of <0.07, equivalent to ∼20% of the vehicle control) similar to those of the PA14 parental strain (Table 1). In contrast, an fpvA mutant encoding the type I pyoverdine receptor had an MIC of 7.5 μM. Growth inhibition was still observed at the highest TS concentration, indicating that the fpvA mutant remained partially susceptible. P. aeruginosa encodes two type I pyoverdine receptors, FpvA and FpvB, with ∼39% amino acid identity (71% similarity). The fpvB mutant was also less susceptible to TS than was the parent strain, with an MIC of 1.3 μM. Based on these patterns of susceptibility, we speculated that TS may use both FpvA and FpvB but that FpvA was the preferred receptor. When we deleted fpvA in the fpvB background, the double mutant had similar resistance levels to the fpvA single mutant (Table 1), but complementation of that mutant with fpvB on a low-copy-number plasmid increased TS susceptibility (Table 1). Together, these data suggest that TS exploits both pyoverdine receptors for entry.
Susceptibility of P. aeruginosa PA14 mutants to thiostrepton in VBMMa
TS is active against clinical isolates.To test whether TS could inhibit growth of a broader range of P. aeruginosa strains, particularly those for which there are fewer antibiotic options, we tested 96 recent clinical isolates for susceptibility to TS in 10:90 medium. While approximately 1 in 10 of those strains had an MIC of ≥5 μM TS (Fig. 4A), a combination of 5 μM TS (8.3 μg/ml) plus 86 μM DSX (32 μg/ml) reduced the growth of most isolates to less than 20% of the dimethyl sulfoxide (DMSO) control (Fig. 4A). We next tested the activity of TS against another MDR Gram-negative pathogen that causes severe infections, Acinetobacter baumannii (40). A. baumannii encodes FpvA and FpvB homologs (Fig. S6), suggesting that it may be susceptible to the thiopeptide. The growth of 6/10 A. baumannii strains in 10:90 medium was reduced to ≤50% of the control with 5 μM TS, while the combination of 5 μM TS and 86 μM DSX reduced the growth of 8/10 isolates of A. baumannii below 20% of the control (Fig. 4B). As reported previously (41), the growth of Escherichia coli (which lacks FpvAB homologs) was unaffected in 10:90 medium even at the maximum soluble concentration of 10 μM (17 μg/ml) TS (Fig. S2). The growth of methicillin-resistant Staphylococcus aureus USA300 (42) was inhibited by <40 nM (32 to 64 ng/ml) of TS in both 10:90 medium and MHB, showing that iron limitation has little effect on TS susceptibility in the absence of an outer membrane.
Thiostrepton inhibits growth of clinical isolates. (A and B) The growth of most clinical isolates of P. aeruginosa (A) and Acinetobacter baumannii (B) (see Table S2 for antibiograms) was inhibited by 5 μM (8.3 μg/ml) TS in 10:90 medium (gray bars), especially when it was combined with 86 μM deferasirox (DSX, 32 μg/ml; black bars). Each assay was performed at least 3 times, and the results are plotted as the percentage of the DMSO-only growth control (OD600) on a log10 scale. Below 20% of the control, no turbidity is visible. Error bars equal one standard deviation.
There are multiple examples of molecules that exploit iron uptake pathways to enter bacteria. Class I microcins, narrow-spectrum antibiotics produced by some Gram-negative species, bind to siderophore receptors and share many of TS’s properties. They are ribosomally synthesized, less than 5 kDa in mass, and cyclic (giving them the nickname “lasso peptides”). Notably, binding of iron by microcins is not a prerequisite for uptake, as some interact with siderophore receptors in an iron-free state. For example, MccJ25, produced by E. coli (43), interacts with siderophore receptor FhuA by mimicking the structure of ferrichrome (44). Although TS has multiple hydroxyls positioned in a manner that could potentially coordinate metals (Fig. 1A), it is unlikely to bind iron based on its inability to decolorize chrome azurol S agar, which changes from blue to orange/yellow when iron is bound by a ligand, in comparison to chelators DFP and DSX (Fig. 5). Further, the structure of TS has been solved both by X-ray crystallography and nuclear magnetic resonance (NMR), and no bound metals were reported (45, 46).
Thiostrepton does not bind iron. To determine whether the uptake of TS by pyoverdine receptors depends on formation of a ferric complex, its ability to decolorize chrome azurol S (CAS) agar was tested. A lack of a color change for TS indicates that in contrast to chelators deferiprone and deferasirox, it is unlikely to chelate iron. The glycopeptide antibiotic vancomycin was used as a negative control. Five microliters of each compound at 2 mg/ml was spotted onto CAS agar, and the plate was incubated at room temperature for 1 h.
DISCUSSION
The natural role of antibiotics has been broadly debated (15, 17), prompting questions such as, are they signaling molecules that are toxic at high concentrations, or weapons used by bacteria to gain an advantage over competitors in their environment? The biofilm stimulation response to subinhibitory concentrations of antibiotics is consistent with both views. At concentrations too low to elicit damage, bacteria show little phenotypic response to antibiotic exposure. As concentrations approach the MIC, the bacteria respond in a dose-dependent manner by ramping up the amount of biofilm produced, detecting either the antibiotics or their effects, which may protect a subpopulation of cells. Above the MIC, antibiotics fall into the deadly weapons category. Biofilm stimulation by subinhibitory concentrations of antibiotics is a common phenomenon among multiple Gram-positive and Gram-negative species and is induced by several drug classes, suggesting that it is not linked to a specific MOA (11, 17, 47). As demonstrated here, this phenotype can be used to identify potential antibiotic activity in the absence of overt killing, a useful feature when screening at a single arbitrary concentration that may be below the MIC for a particular drug-organism combination. Interestingly, we and others (48) found that many drugs intended for eukaryotic targets can impact bacterial growth and biofilm formation, implying that they have deleterious effects on prokaryotic physiology. With a new appreciation of the role of the human microbiome in health and disease, these potential effects should be considered during drug development.
TS, a complex cyclic thiopeptide made by Streptomyces azureus, Streptomyces hawaiiensis, and Streptomyces laurentii, is experiencing a resurgence due to its antibacterial, antimalarial, and anticancer activities (23, 24). It is a member of the ribosomally synthesized and posttranslationally modified peptide (RiPP) class of natural products (49) derived from a 42-amino-acid precursor, TsrA (50). Although the mechanism of its antibacterial activity (inhibition of translation by binding to helices H43/H44 of 23S rRNA) and resistance (methylation of 23S rRNA residue A1067) have been deciphered (26, 51), the way in which this ∼1.7-kDa molecule enters target bacteria is unknown. Our data suggest that TS is actively imported into P. aeruginosa under iron-restricted conditions. Its large mass would impede passive diffusion through the outer membrane, and single, double, or triple mutants lacking the outer membrane components of major efflux systems MexAB-OprM, MexCD-OprJ, and MexEF-OpmD have wild-type TS susceptibility.
Our discovery that TS exploits FpvA and FpvB for uptake into the periplasm explains the resistance of Gram-negative species such as E. coli to this antibiotic, as they lack those proteins. FpvAB homologs are expressed by P. aeruginosa and related pathogens, including A. baumannii, suggesting that TS could have utility as a narrow-spectrum agent. FpvA is also exploited by S-pyocins, 40- to 80-kDa peptide antibiotics produced by competing P. aeruginosa strains, showing that it is an important promiscuous access point for diverse molecules in addition to its pyoverdine ligand (33, 34, 52). The use of multiple pyoverdine receptors by TS may reduce the probability of resistance arising through the mutation of a single receptor, although genome analysis of clinical isolate C0379 that was most resistant to the combination of TS and DSX revealed a wild-type copy of fpvA coupled with an ∼800-bp deletion encompassing the 5′ region of fpvB. Multiple single-nucleotide polymorphisms with unknown effects on function were present in both TS-susceptible and -resistant isolates. A more detailed investigation of P. aeruginosa strains resistant to TS will be needed to understand the most likely routes by which it occurs.
Although TS uses siderophore receptors to cross the P. aeruginosa OM, the way in which this compound transits the cytoplasmic membranes of Gram-positive and Gram-negative bacteria to reach its ribosomal targets remains undefined. The expression of tsr in P. aeruginosa conferred resistance, confirming that TS acts at least in part via its canonical bacteriostatic MOA. While PA14 expressing Tsr was significantly more resistant to TS than was the control, it was more sensitive than PAO1. This difference is not due to the nucleotide polymorphism at the Tsr methylation site on the rRNA, as these residues are conserved between PAO1 and PA14. The reasons for strain-specific differences in susceptibility are unclear, but our data confirm that most clinical P. aeruginosa isolates tested are susceptible to TS, especially when combined with DSX.
The major liability of TS is its poor solubility (53). Smaller more-soluble fragments that retain activity against Gram-positive bacteria and have reduced toxicity for eukaryotic cells have been identified (54), but it is not clear if they would be active against P. aeruginosa or A. baumannii if uptake by the FpvAB receptors requires the intact molecule. Another way to circumvent solubility issues is to reduce the concentration required to kill. Our data show that coadministration of TS with FDA-approved iron chelators DFP or DSX markedly reduces its MIC against P. aeruginosa and A. baumannii. The true potential of TS as an anti-infective may be underestimated, as MIC evaluations are typically performed in rich iron-replete media. Many host environments are iron restricted, particularly in the presence of infection and inflammation (55–57). Our data show that TS is active at low-micromolar concentrations against P. aeruginosa in 10% mouse and human sera, even in the absence of added chelator. The combination of TS with DSX may be useful at sites such as chronically infected lungs, where iron is more abundant (58).
In summary, we showed that biofilm stimulation can be used in high-throughput small-molecule screening to report on subinhibitory antibiotic activity that would otherwise be missed using the usual metric of growth inhibition. In a small screen of fewer than 4,000 molecules at a fixed concentration of 10 μM, we doubled the number of potential antimicrobials identified, finding 60 growth inhibitors plus another 60 molecules that stimulated biofilm formation. This phenotype can indicate potential antimicrobial activity at higher concentrations, or under different growth conditions, as demonstrated here for TS. Stimulation of biofilm matrix production by TS in members of the Gram-positive genus Bacillus was reported previously, and that phenotype was leveraged to identify novel thiopeptide producers in cocultures (59). Those studies, and the data presented here, suggest that monitoring biofilm stimulation (or an easily assayed proxy thereof, perhaps increased expression from biofilm matrix promoters) could allow for more sensitive detection of molecules with potential antibacterial activity during screening, making it a useful addition to the antimicrobial discovery toolkit.
MATERIALS AND METHODS
Bacterial strains and culture conditions.Most bacterial strains and plasmids used in this study are listed in Tables 1 and S2. The PA14 transposon mutants listed in Table 1 were extracted from an ordered library (39). In addition to those strains, we used P. aeruginosa PAO1 (60), Staphylococcus aureus USA300 (42), and Escherichia coli BW25113 (61) for susceptibility studies (Fig. S2) and E. coli strains DH5α and SM10 for cloning and generation of knockouts, as described below. Bacterial cultures were grown in lysogeny broth (LB), 10:90 (10% LB and 90% phosphate-buffered saline) medium, M9 medium, Vogel-Bonner minimal medium (VBMM), or cation-adjusted Mueller-Hinton broth (MHB), as indicated in this study. Where solid media were used, plates were solidified with 1.5% agar. DFP (Sigma-Aldrich) and DSX (Cayman Chemicals) were stored at 4°C until use. TS was stored at −20°C. A 60 mg/ml stock solution of DFP was made in 6 M HCl and Milli-Q H2O (DFP solvent) in a ratio of 3:50. A 20 mg/ml stock solution of DSX was made in DMSO. A 20 mM stock solution of TS was made in DMSO.
Growth curves.PAO1 was inoculated from a −80°C stock into 5 ml LB broth and grown with shaking at 200 rpm for 16 h and 37°C. The overnight culture was subcultured at a 1:500 dilution into 5 different media (LB, 10:90 medium, M9, Mueller-Hinton broth [MHB], and VBMM) and incubated at 37°C for 6 h with shaking at 200 rpm. Each subculture was standardized to an OD600 of ∼0.1 (BioMate 3 spectrophotometer), which corresponds to ∼10e7 CFU per ml, and then diluted 1:500 into the same medium. Six replicates of 200 μl of each sample were added to a 96-well plate, which was incubated at 37°C for 24 h with shaking at 200 rpm (Tecan Ultra Evolution plate reader). The OD612 was read every 15 min for 24 h. The data for the six replicates of each sample were averaged, and the experiment was repeated 3 times. The final data with standard deviations were plotted using Prism (GraphPad).
Biofilm modulation assay.Biofilm formation was assayed as described previously (13), with modifications. Briefly, P. aeruginosa was inoculated in 5 ml of LB and grown at 37°C overnight, with shaking at 200 rpm, and subsequently standardized to an OD600 of ∼0.1 in 10:90 medium. For the initial screen, 1 mM compound stocks in DMSO were diluted 1:100 in a standardized cell suspension (1.5 μl of compound stock in 148.5 μl of cell suspension) to a final concentration of 10 μM. The control wells contained 10:90 medium plus 1% DMSO (sterility control) or a standardized cell suspension plus 1% DMSO (growth control). Biofilms were formed on polystyrene peg lids (Nunc). After placement of the peg lid, the plate was sealed with Parafilm to prevent evaporation and incubated for 16 h at 37°C and 200 rpm. Following incubation, the 96-peg lid was removed, and planktonic density in the 96-well plate was measured at the OD600 to assess the effect of test compounds on bacterial growth. The lid was transferred to a new microtiter plate containing 200 μl of 1× phosphate-buffered saline (PBS) per well for 10 min to wash off any loosely adherent bacterial cells and then to a microtiter plate containing 200 μl of 0.1% (wt/vol) CV per well for 15 min. Following staining, the lid was washed with 70 ml of distilled water (dH2O) in a single-well tray for 10 min. This step was repeated four times to ensure complete removal of excess CV. The lid was transferred to a 96-well plate containing 200 μl of 33% (vol/vol) acetic acid per well for 5 min to elute the bound CV. The absorbance of the eluted CV at 600 nm was measured (BioTek ELx800), and the results were plotted as the percentage of the DMSO control using Prism (GraphPad). Screens were performed in duplicate. Compounds that resulted in <50% of the control biofilm were defined as biofilm inhibitors, while compounds that resulted in >200% of the control biofilm were defined as biofilm stimulators. Compounds of interest were further evaluated using the same assay but over a wider range of concentrations (dose-response assay).
For TS dose-response assays, TS stock solutions were diluted in DMSO, and 2 μl of the resulting solutions plus 148 μl of a bacterial suspension standardized to an OD600 of ∼0.1 in 10:90 medium were added to a 96-well plate in triplicate, as described above. Control wells contained 148 μl of 10:90 medium plus 1.3% DMSO (sterility control) or a standardized bacterial suspension plus 1.3% DMSO (growth control). For experiments with EDDHA with or without FeCl3, each was added (2 μl) as aqueous solutions to reach final concentrations of 0.1 μM EDDHA and 100 μM FeCl3, and the amount of bacterial suspension was adjusted to keep the total well volume at 150 μl. The control for EDDHA and FeCl3 was 2 μl of sterile dH2O. Biofilms were grown for 16 h at 37°C and 200 rpm and then stained and quantified as described above. Assays were performed in triplicate, and the results were graphed using Prism (GraphPad) as a percentage of the DMSO control.
Compounds screened.The biofilm modulation assay was used to screen the McMaster Bioactives compound collection. This curated collection includes off-patent FDA-approved drugs from the Prestwick Chemical Library (Prestwick Chemical, Illkirch, France), purified natural products from the Screen-Well Natural Products Library (Enzo Life Sciences, Inc., Farmingdale, NY, USA), drug-like molecules from the LOPAC1280 (international version) collection (Sigma-Aldrich Canada Ltd., Oakville, Ontario, Canada), and the Spectrum Collection (MicroSource Discovery Systems, Inc., Gaylordsville, CT, USA), which includes off-patent drugs, natural products, and other biologically active compounds. In total, the collection contains 3,921 unique compounds.
Construction of a tsr plasmid for expression in P. aeruginosa.The tsr gene from pIJ6902 (62) was PCR amplified using the 5′-GAATCCCGGGCGGTAGGACGACCATGAC-3′ and 5′-CTTCAAGCTTTTATCGGTTGGCCGCGAG-3′ primers. Both the PCR product and pUCP20 vector were digested with SmaI and HindIII, gel purified, and ligated at a 1:3 molar ratio using T4 DNA ligase. The ligated DNA was transformed into E. coli DH5α and transformants selected on LB agar containing 100 μg/ml ampicillin and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside for blue-white selection. Plasmids from white colonies were purified using a GeneJet plasmid miniprep kit (Thermo Scientific), following the manufacturer’s protocols. After verification by restriction digestion and DNA sequencing, pUCP20 and pUCP20-tsr were each introduced into P. aeruginosa PAO1 and PA14 by electroporation. Transformants were selected on LB agar containing 200 μg/ml carbenicillin.
Serum preparation.Human serum (Corning) and mouse serum (Equitech-Bio) were stored at −20°C. Sera were aliquoted into 5-ml culture tubes by thawing once at 37°C for 30 min with occasional gentle mixing. Culture tubes were frozen at −20°C until use. To make 10% serum solutions, serum was thawed for 10 min at 37°C and then heat inactivated at 57°C for 30 min. Two milliliters of heat-inactivated serum was added to 18 ml of 10:90 medium and gently mixed. This 10% serum solution was used for checkerboard assays.
MIC and checkboard assays.MICs were determined with broth microdilution assays in 96-well plates. Vehicle controls consisted of 1:75 dilutions of DMSO in 10:90 medium inoculated with PA14 or its mutants, as described in “Growth curves,” above. Sterile controls consisted of 1:75 dilutions of DMSO in 10:90 medium. Seven serially diluted concentrations of TS, with 17 μg/ml (10 μM) being the highest final concentration, were set up in triplicate. Tests were done with 1:75 dilutions of each TS concentration in 10:90 medium inoculated with PA14 or its mutants, as described in “Growth curves,” above. Plates were sealed to prevent evaporation and incubated with shaking at 200 rpm for 16 h at 37°C. The OD600 of the plates was read (Multiskan Go; Thermo Fisher Scientific), and after the subtraction of sterile medium values as blanks, it was used to determine the MIC. The final volume of each well was 150 μl, and each experiment was repeated at least three times.
Checkerboard assays were set up using 96-well plates in an 8-well by 8-well format. Two columns were allocated for vehicle controls and two columns for sterility controls. Vehicle controls contained 2 μl DMSO plus 2 μl DFP for checkerboards with TS and DFP or 4 μl DMSO for TS and DSX. One hundred forty-six microliters of 10:90 medium or 10% serum was then inoculated with PA14 or PAO1, as described in “Growth curves,” above, and added to the wells. Sterile controls contained the same components in 10:90 medium or 10% serum without cells. Serial dilutions of TS, with 17 μg/ml being the highest final concentration, were added along the y axis of the checkerboard (increasing concentration from bottom to top), whereas serial dilutions of DFP or DSX, with 512 μg/ml being the highest final concentration, were added along the x axis (increasing concentration from left to right). The final volume of each well was 150 μl, and each checkerboard was repeated at least three times. Plates were incubated and the final OD600 determined as detailed above.
Clinical isolate testing.Clinical isolates of P. aeruginosa and A. baumannii were inoculated from −80°C stocks into 200 μl LB broth and grown with shaking at 200 rpm for 16 h at 37°C in Nunc 96-well plates. The overnight cultures were subcultured (1:25 dilution) into 10:90 medium and grown with shaking at 200 rpm for 2 h at 37°C. Vehicle controls consisted of 4 μl of DMSO, 144 μl of 10:90 medium, and 2 μl of subculture. Sterile controls consisted of 4 μl of DMSO and 146 μl of 10:90 medium. Test samples consisted of 2 μl of TS (final concentration, 5 μM [8.3 μg/ml]), 2 μl of DMSO (or DSX; final concentration, 86 μM [32 μg/ml]), 144 μl of 10:90 medium, and 2 μl of subculture. The final volume of each well was 150 μl, and each checkerboard was repeated at least three times. Plates were incubated with shaking at 200 rpm for 16 h at 37°C, and the OD600 was measured (Multiskan Go; Thermo Fisher Scientific). The results were plotted as the percentage of the control (wells containing only DMSO) using Prism (GraphPad).
Chrome azurol S plate assay.Chrome azurol S (CAS) agar plates were prepared as described by Louden et al. (63). All components were purchased from Sigma except for agar, NaOH, NaCl (BioShop), Casamino Acids (Becton, Dickinson), and glucose (EMD Millipore). Stock solutions of TS, vancomycin (VAN), DSX, and DFP were standardized to 2 mg/ml. Five microliters of 2 mg/ml compound was spotted on a plate and incubated at room temperature for 1 h prior to photographing the plate.
Generation of efflux mutants.Deletion mutants lacking the outer membrane components of the 4 major resistance-nodulation-division (RND) efflux systems of P. aeruginosa (MexAB-OprM, MexXY-OprM, MexCD-OprJ, and MexEF-OpmD) were generated as reported previously (60). Briefly, the pairs of primers listed in Table S3 were used to amplify regions up- and downstream of the gene to be deleted. The PCR products were digested with the restriction enzymes indicated in the primer sequences and the resulting fragments ligated into the suicide vector pEX18Gm. After DNA sequencing validation of the constructs, they were introduced into E. coli SM10 for biparental mating into P. aeruginosa PAO1. Mating mixtures were plated on Pseudomonas isolation agar containing 200 μg/ml gentamicin (Gm) to counterselect the donor. Gm-sensitive double recombinants were selected on LB agar, with no salt and with 5% (wt/vol) sucrose. Gm-sensitive deletion mutants were identified by PCR and validated by DNA sequencing of the deletion junction.
ACKNOWLEDGMENTS
We thank Gerry Wright for access to strains from the Wright Clinical Collection, David Heinrichs for the gift of EDDHA and helpful discussions, and Neha Sharma, Andrew Hogan, Amanda Veri, and Victor Yang for assistance with method development.
This work was supported by Natural Sciences and Engineering Research (NSERC) grant RGPIN-2016-06521 and by Ontario Research Fund grant RE07-048. M.R.M.R. and U.N. held Ontario Graduate Scholarships, M.R. was supported by an NSERC Undergraduate Summer Research Award, S.K.-P. was supported by a Summer Studentship from GlycoNet, and H.A. was supported by a Summer Studentship from Cystic Fibrosis Canada.
FOOTNOTES
- Received 4 March 2019.
- Returned for modification 25 March 2019.
- Accepted 23 June 2019.
- Accepted manuscript posted online 1 July 2019.
Supplemental material for this article may be found at https://doi.org/10.1128/AAC.00472-19.
- Copyright © 2019 American Society for Microbiology.
