Repurposing bioactive compounds for treating multidrug-resistant pathogens

Abstract

Introduction

Antimicrobial development is being outpaced by the rising rate of antimicrobial resistance in the developing and industrialized world. Drug repurposing, where novel antibacterial functions can be found for known molecular entities, reduces drug development costs, reduces regulatory hurdles, and increases rate of success.

Aim

We sought to characterize the antimicrobial properties of five known bioactives (DMAQ-B1, carboplatin, oxaliplatin, CD437 and PSB-069) that were discovered in a high-throughput phenotypic screen for hits that extend Caenorhabditis elegans survival during exposure to Pseudomonas aeruginosa PA14.

Methodology

c.f.u. assays, biofilm staining and fluorescence microscopy were used to assay the compounds' effect on various virulence determinants. Checkerboard assays were used to assess synergy between compounds and conventional antimicrobials. C. elegans-based assays were used to test whether the compounds could also rescue against Enterococcus faecalis and Staphyloccus aureus. Finally, toxicity was assessed in C. elegans and mammalian cells.

Results

Four of the compounds rescued C. elegans from a second bacterial pathogen and two of them (DMAQ-B1, a naturally occurring insulin mimetic, and CD437, an agonist of the retinoic acid receptor) rescued against all three. The platinum complexes displayed increased antimicrobial activity against P. aeruginosa . Of the molecules tested, only CD437 showed slight synergy with ampicillin. The two most effective compounds, DMAQ-B1 and CD437, showed toxicity to mammalian cells.

Conclusion

Although these compounds' potential for repurposing is limited by their toxicity, our results contribute to this growing field and provide a simple road map for using C. elegans for preliminary testing of known bioactive compounds with predicted antimicrobial activity.

Introduction

Infections with multidrug-resistant (MDR) pathogens present a serious health concern for immuno-compromised individuals, cancer patients and patients in intensive care units. An estimated 700 000 deaths annually can be attributed to these infections, and this number is predicted to surpass 10 million deaths per year by 2050 [1].

The speed at which pathogens evolve and spread the determinants for antimicrobial resistance has historically outstripped the rate at which new treatments could be devised, often by orders of magnitude. A variety of factors have driven antimicrobial discovery efforts to their lowest levels in the past 100 years [2]. One key reason is that new antimicrobials have an abnormally low expected return-on-investment. Drug discovery, from target choice to FDA approval, can cost upward of $2.5 billion dollars for a single molecular entity [3]. Generally, drug companies rely on offsetting those costs through sales. Several forces conspire to make this more difficult with antibiotics than with medicines targeting other diseases (e.g. autoimmune disorders, metabolic disorders or chronic mental disorders). One is that treatment duration with antibiotics is typically short and rarely repeated. Profitability is also limited by antimicrobial stewardship programs, which seek to prolong the usefulness of a new drug by limiting the use of the new drug as much as possible. In addition, despite the growth of antimicrobial resistance, there are still frequently several treatments available, even if they are less effective. Treatment payees (whether insurance companies or patients paying out-of-pocket) are reluctant to pay an order of magnitude (or more) higher prices for a therapy that is only marginally better. Finally, at least in the USA, there is substantial social pressure for life-saving medicines to be priced affordably, regardless of long-term consequences.

An alternative approach drawing increasing awareness and effort is drug repurposing or repositioning. This technique leverages the information learned during preliminary drug discovery efforts (or even phase I/II/III testing) to ‘jump start’ the process to use the same compound for alternative diseases. Notable examples of the success of drug repurposing include the use of thalidomide as a treatment for multiple myeloma [4], chloroquine for metastatic breast cancer [5], and the antihypertensive monoxodil as a treatment for alopecia [6]. By repurposing molecules that have already been proven safe in humans, many regulatory hurdles can be bypassed [7]. This means less time and money invested, and a greater probability of successfully identifying a safe and effective chemical entity with market potential [8]. Antimicrobials have recently shown several interesting examples of this approach. For example, ibuprofen was shown to have antimicrobial properties and to elicit a therapeutic effect in cystic fibrosis patients [9, 10]. Additionally, entecapone, a medication for Parkinson’s disease, inhibited mycobacterial growth [11]. Identification and characterization of such alternative activities can be extremely useful, especially for molecules with favourable safety profiles and well-studied pharmacokinetics and pharmacodynamics.

Ultimately, antimicrobial discovery needs to keep pace with the development and propagation of new resistance mechanisms. Problematically, pathogens are frustratingly effective at devising ways to avoid antimicrobials, including reduction of drug accumulation (decreased cell permeability, increased efflux, etc.), drug inactivation or modification (phosphorylation, cleavage, etc.), mutation of the site that the antibacterial binds to (mutation, expression of protection proteins, etc.), and even wholesale modification of metabolic pathways [12, 13]. Lateral gene transfer then serves to spread resistance within bacteria of the same environment, including in clinics, in the soil, in water, or even within another organism [14]. Indeed, the environment of the bacteria and their interaction with a host play a large part in acquisition of resistance and persistence of infection. For instance, individuals with cystic fibrosis possess especially viscous mucous within their lungs, facilitating colonization by bacteria [15]. This commonly results in the formation of antibiotic-tolerant biofilms, which create a microenvironment for pathogens that limits access of antibiotics to the bacteria, allowing time for drug resistance to build up over time [15, 16].

New methods and techniques that model the interactions between the host and the pathogen present an opportunity for the identification and validation of new therapeutically useful molecules. Phenotypic screens utilizing infection of whole organisms is one such approach. Unlike biochemical or solely bacteria-based screens, phenotypic screens can tie the outcome of the infection directly to the output of the assay. One such well-characterized model for phenotypic screens is Caenorhabditis elegans. C. elegans has been used extensively as a model for infections, and more recently as a tool for high-throughput chemical screening [17–19]. Death or survival of the worm is often the output, which has a number of distinct advantages. For example, hits are enriched for molecules that effectively interact with either the pathogen through inhibition of growth or virulence, or the worm itself through immune response modulation [17, 20, 21]. In addition, toxic or poorly bioavailable molecules are efficiently eliminated from the hit pool.

Previously in our lab, a high-throughput chemical screen was performed to identify small molecules that attenuated killing of C. elegans by Pseudomonas aeruginosa in a liquid-based pathogenesis assay called Liquid Killing [22]. Worms were sorted into wells with small molecules and P. aeruginosa and incubated for ~2 days. After this time, approximately 60 % of untreated worms would be dead. Small molecules that significantly attenuated worm death were scored as hits. Several classes of hits were identified, including compounds that inhibit the function of the siderophore pyoverdine [23], novel compounds with unknown functions, and known bioactives not generally recognized as antimicrobials. The lattermost category included DMAQ-B1, an insulin mimetic purified from Pseudomassaria sp. [24]; CD437, an activator of retinoic acid receptor-β and -γ [25]; oxaliplatin and carboplatin, FDA approved platinum complexes for the treatment of cancer [26, 27]; and PSB-069, a nucleoside triphosphate diphosphohydrolase inhibitor [28] (Fig. 1a). In this study we examined the ability of these compounds to mitigate the pathogenic activity of three nosocomial pathogens, and assayed the toxicity, the potential synergy, and the specific effects of these molecules to determine their potential for future repurposing efforts.

Fig. 1.

Bioactives rescue C. elegans from P. aeruginosa. (a) Structures for five known bioactive molecules studied in this report. (b) EC and MIC values for bioactives against P. aeruginosa . EC and MIC values were averages calculated from two biological replicates using Student's t-test. For determination of EC values, four wells were used for each compound with 20 worms per well. MIC values were determined using four wells in a 384-well plate. Concentrations above 64 µM were not tested. Also shown is the ratio of MIC to EC. (c) Representative replicate of liquid killing for bioactives. Error bars represent sem. *: P<0.05, **: P<0.01, #: P<0.001. P-values were calculated using Student’s t-test. P-values were also calculated using an ANOVA with a post-hoc Tukey HSD test, which showed siginficance for the following results: DMAQ-B1 at all concentrations, carboplatin at 32 and 16 µM, oxaliplatin at 64, 32, 16, and 8 µM, and PSB-069 at 64, 32, 16 and 8 µM.

Results

Bioactive molecules are protective against multiple pathogens

As a first step, we tested the potency of our molecules for the ability to limit Liquid Killing. Compounds were tested in the Liquid Killing assay using two-fold serial dilution to identify their minimum effective concentration (EC), defined as the minimum concentration required for statistically significant rescue (Fig. 1b–c, Fig. S1). Four of the compounds rescued at low micromolar concentrations, with CD437 displaying a significantly higher EC. Several different mechanisms have been shown to mediate rescue in C. elegans pathogenesis, including stimulating host defence pathways [20], attenuating pathogen virulence [22] and, most commonly by far, antimicrobial activity [29]. Molecules functioning solely as antimicrobials are often easy to recognize, as the ratio of their MIC to EC is generally between 0.5 and 3 [23]. This result is intuitive, as their ability to protect the host is contingent upon killing the pathogen. For the majority of the bioactives tested, the ratio was >4, suggesting that the antimicrobial activity is not the sole relevant mechanism of rescue.

Previous studies suggest that small molecules that can rescue C. elegans from disease in one assay often have some ability to work against additional pathogens [20, 21]. Therefore, we tested the ability of our molecules to rescue C. elegans from the nosocomial Gram-positive bacteria Enterococcus faecalis and Staphylococcus aureus in liquid-based pathogenesis assays [30]. Four out of five molecules (all except carboplatin) reduced E. faecalis pathogenesis while two (DMAQ-B1 and oxaliplatin) limited the ability of S. aureus to kill worms (Fig. 2a–c).

Fig. 2.

Subset of bioactives rescued C.elegans infected with E. faecalis or S. aureus. (a, b) A representative replicate of C. elegans exposed to E. faecalis (a) or S. aureus (b). Error bars represent sem. (c) Compound EC and MIC values for E. faecalis and S. aureus . EC and MIC values were averages calculated from two biological replicates. For determination of EC values, four wells were used for each compound with 20 worms per well, and Student's t-test was used to determine significance (and hence effect). MIC values were determined using four wells in a 384-well plate. Concentrations above 64 µM were not tested. *: P<0.05, **: P<0.01, #: P<0.001. P-values were calculated using Student’s t-test. P-values were also calculated using an ANOVA with a post-hoc Tukey HSD test, which showed significance for the following results for E. faecalis : DMAQ-B1 at all concentrations, oxaliplatin at 64 and 32 µM, CD437 at 64 µM and PSB-069 at 64, 32 and 16 µM. For S. aureus , the following concentrations were significant: DMAQ-B1 at 64, 32 and 16 µM. For oxaliplatin and PSB-069, 64 µM was significant.

Selected molecules inhibit bacterial growth and reduce pyoverdine levels

Although endpoint MIC assays can provide information regarding the ability of the compounds to prevent growth, information about growth rate is lost. To obtain finer details about the impact of treatment on bacterial growth, bacteria were grown overnight and then diluted into 96-well plates containing four-fold serial dilutions of the bioactive compounds. Aliquots were collected at indicated times, serially diluted ten-fold, and plated to determine accurate colony counts. Bactericidal activity was determined based on the NCCLS definition, which is widely accepted in the field [31–35]. Specifically, molecules were considered bactericidal if treatment caused ≥99.9 % decrease in the initial inoculum (i.e. if the drug reduced c.f.u. ≥3 log10 c.f.u. ml⁻¹). As was observed in the MIC assays, DMAQ-B1 effectively killed all three pathogens (Figs 3a, and S1, available in the online version of this article). Growth of P. aeruginosa was also affected by high doses of oxaliplatin and modestly by carboplatin (c.f.u. reduction ≥3 log₁₀ at 12 h, but not at 24 h) (Fig. 3a–c). The bactericidal activities of DMAQ-B1 and oxaliplatin were confirmed by using propidium iodide to stain dead cells in a GFP-expressing strain of P. aeruginosa PA14 (Fig. S2). DMAQ-B1 and CD437 were particularly efficient against S. aureus , with bactericidal effect at concentrations in the low- to single micromolar range (Fig. 3f–g). All three pathogens demonstrated some rebound growth during treatment with DMAQ-B1, where their minimal numbers were reached earlier and by 24 h growth had resumed. This phenomenon is well-known to occur in P. aeruginosa , where efflux pumps are activated to remove the antimicrobial and facilitate growth. It is often observed with aminoglycosides like gentamicin (Fig. S3). Interestingly, PSB-069 exhibited no antimicrobial activity against any of the pathogens, even at the highest concentrations tested. This suggests that some alternative mechanism of action is involved in its ability to rescue C. elegans from the pathogens.

Fig. 3.

Bioactive compounds have antimicrobial effects. (a–g) Colony counts from serial dilution of bacteria grown for the indicated time in media containing the indicated concentration of compound or solvent. Bacteria were plated on non-selective growth media to determine viable bacterial counts. Data represent the average of three biological replicates. Only compound/pathogen combinations that showed antimicrobial activity at 256 µM or less are shown.

Liquid Killing by P. aeruginosa partially relies upon secretion of the siderophore pyoverdine, which causes worm death by scavenging iron from host mitochondria [36–38]. We have previously shown that limiting the production or the activity of pyoverdine extends survival of C. elegans in Liquid Killing, including upon exposure to multidrug-resistant P. aeruginosa isolates [22, 23, 37, 39]. To determine if any of our molecules may function by inhibiting pyoverdine, we cultured P. aeruginosa in the presence of two-fold serial dilutions of our bioactives and measured production of the fluorescent siderophore (Figs 4 and S4). Three of the compounds, DMAQ-B1, oxaliplatin and PSB-069, inhibited pyoverdine production at concentrations lower than those required for strong bactericidal effect (Fig. 4e). Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) of P. aeruginosa treated with DMAQ-B1 or PSB-069 also showed a modest (~two-fold) decrease in the expression of genes whose regulation depends on pyoverdine (Fig. S5). This effect, whether direct or indirect, may explain the variance of the MIC/EC ratios for these compounds away from what would be expected for antimicrobials.

Fig. 4.

Bioactives do not appear to compromise pyoverdine production. (a–e) P. aeruginosa strain PA14 was grown in 96-well plates using modified M9 medium supplemented with the indicated compound at concentrations ranging from 1 to 128 µM, or solvent control. At 2 h intervals for 30 h, pyoverdine production was measured using fluorescence spectroscopy and normalized to OD₆₀₀. Data shown are average values from four wells per replicate and two replicates were performed.

Another possibility was that the molecules stimulated C. elegans defense pathways. This was tested by treating C. elegans with the compounds in the absence of any pathogen and measuring expression of stress-response pathways using fluorescent reporters. Three canonical host defence pathways known to be involved in the response to P. aeruginosa were tested. These were the DAF-16/FOXO (insulin signalling), SKN-1/Nrf2 (detoxification) and p38 MAPK (innate immunity) pathways [40–42]. The expression of the ESRE mitochondrial surveillance pathway, which is activated by pyoverdine-induced mitochondrial damage [38] was also assessed. Although oxaliplatin and CD437 exhibited small, but significant, activation of the ESRE pathway, activity of the compounds as a group appears to be largely independent of host defence. This suggests that that the molecules’ activity is more likely to depend upon changes to the bacteria rather than the host (Fig. S6).

Biofilms are an important virulence determinant for chronic P. aeruginosa infections [43] and can play an important role in physically limiting access of the antimicrobials to bacteria [44, 45]. Some small molecules are capable of preventing the formation of biofilm, thereby increasing the therapeutic window and, by extension, the effectiveness of antibiotics [46]. We tested the impact of the compounds on biofilm that triggered rescue (oxaliplatin and DMAQ-B1 for P. aeruginosa and CD437 and DMAQ-B1 for S. aureus and E. faecalis ) on biofilm (Fig. 5). Although CD437 limited biofilm formation in S. aureus and E. faecalis and DMAQ-B1 prevented it in S. aureus , the concentrations required were higher than the MIC value, suggesting that biofilm inhibition is not the relevant mechanism for limiting pathogenesis in C. elegans assays.

Fig. 5.

Effect of bioactives on biofilms. (a–c) Bacteria were diluted and grown in static 96-well plates. After 24 h, plates were washed and biofilm was stained with crystal violet, which was solubilized and measured via spectroscopy. Only strain/compound combinations that showed strong growth inhibition are shown. (d) Representative images of stained biofilm after wash and stain. Concentrations ranged from 1 to 64 µM with the last well in each column being a DMSO control. For (a–d), each condition was performed in duplicate, and data shown are the average of two biological replicates. *: P<0.05, **: P<0.01, #: P<0.001. P-values were calculated using Student’s t-test. Error bars represent sem.

CD437 synergizes with ampicillin

Although DMAQ-B1, CD437, oxaliplatin and carboplatin showed varying degrees of antimicrobial activity alone, their potential as therapeutics would be much greater if they had synergy with currently used therapies. To test this, compounds with significant antimicrobial activity against one or more of the pathogens were tested in combination with standard-of-care drugs for P. aeruginosa , E. faecalis and S. aureus (ciprofloxacin, ampicillin and vancomycin, respectively) using a checkerboard assay (Fig. 6a–c). Bacteria were cultured with two-fold dilutions of antimicrobials and test molecules in 384-well plates, incubated for 16 h, and measured for growth. The only synergy observed was a weak interaction between ampicillin and CD437, which displayed a four-fold increase in effectiveness. Although the synergistic relationship increased the antimicrobial activity of the two molecules, it was unclear whether this would translate to increased rescue in C. elegans infection assays with the same pathogen. To test this, a whole-animal in vivo-checkerboard assay was performed with CD437 and ampicillin, this time under liquid-based E. faecalis infection conditions (Figs 6d and S7). No synergy was observed for the drug combination in the context of the infection assay. There are at least two potential explanations. First, the relatively low level of synergy may have been lost under these more stringent conditions. Second, the C. elegans assay, which is scored as living or dead worms, may exhibit decreased sensitivity compared to a more fine-grained bacterial growth assay.

Fig. 6.

Bioactive compounds generally do not have synergy with antimicrobials. (a–d) Synergy landscapes of test compounds in combination with frontline antimicrobials for P. aeruginosa (a), E. faecalis (b) or S. aureus (c). Synergy was defined as ƩFIC<0.5 and is shown in the green area. No interaction was 0.5>4, and antagonism >4. At least two biological replicates were performed and a representative replicate is shown.

CD437 and DMAQ-B1 exhibit toxicity in mammalian cells

One other characteristic important for understanding the potential of a compound is its toxicity. To broadly assess the toxicity of the compounds, C. elegans longevity assays were performed. Young adult worms were placed onto NGM agar plates containing each compound at 25 µM and scored for survival. Of the compounds tested, most showed no signs of toxicity. DMAQ-B1 was a notable exception, showing significant toxicity towards worms (Fig. 7a). Acute toxicity assays were also performed using the human bronchial epithelial cell line 16HBE. Cells were exposed to compounds for 24 h in regular media, and MTT assays were used to assess viability. Recapitulating our data from C. elegans, DMAQ-B1 showed considerable toxicity. Unlike in C. elegans, CD437 was also toxic in this assay. Both compounds were toxic even at concentrations lower than the minimal concentrations tested for rescue against pathogen (Fig. 7b). Although this limits their potential for use, there remains the possibility that their toxicity to eukaryotes and their antimicrobial activity may be separable traits. Active structure-activity relationship studies would need to be performed to further evaluate their potential.

Fig. 7.

DMAQ-B1 and CD437 show signs of toxicity. (a) Wild-type, young adult C. elegans were placed on NGM media supplemented with compounds at 25 µM. Survival was scored daily until no survivors remained. Only DMAQ-B1 was found to statistically significantly differ from DMSO controls (P<0.001). A representative replicate is shown. Each compound was tested using three plates of ~60 worms. Two biological replicates were performed. (b) 16HBE cells were seeded into 96-well plates and allowed to attach before media was replaced with serum-free medium containing compound at the indicated concentration. Data shown represent percent viability (normalized to DMSO controls) and are the average of three biological replicates. Error bars represent sem. *: P<0.05, **: P<0.01, #: P<0.001. Statistical significance was calculated using a log-rank test (a) or t-test.

Discussion

In this study, we assayed the antimicrobial and anti-virulence properties of five known bioactives. These compounds were selected based on their ability to reduce the pathogenesis of P. aeruginosa . Of the molecules characterized in this study, carboplatin, oxaliplatin, DMAQ-B1 and CD437 appeared to act as an antimicrobial for at least one of the three pathogens. Our investigation led to several interesting conclusions. First, despite its identification screening in P. aeruginosa, CD437 was more effective against E. faecalis and S. aureus . Second, DMAQ-B1 and CD437 had the best profile as antibiotics, but also exhibited the most toxicity in our assays. Third, although the MIC/EC ratio of the majority of the compounds was relatively high for antimicrobials, their effect on bacterial growth was significant. This higher ratio could indicate that they are working through multiple mechanisms rather than just a straightforward inhibition of bacterial growth. This type of multifunction drug may be more difficult for bacteria to evolve resistance against than a classical antimicrobial. Finally, although a subset of molecules rescued against multiple pathogens, the mechanism for rescue against different pathogens likely differs. The MIC/EC ratios suggest that the compounds were more likely to function as antimicrobials for E. faecalis and S. aureus . The likeliest explanation for this is that the P. aeruginosa assay is based on intoxication rather than infection, while the assays for E. faecalis and S. aureus require infection. To rescue, compounds need to contravene the relevant disease-causing phenomenon or promote general host health. This reinforces the value of using multiple screening platforms for drug discovery.

Interestingly, CD437 was previously identified by another group looking for anti-infectives using C. elegans and S. aureus [21]. Much like them, we saw a strong antibacterial effect from the CD437. However, CD437 was much less effective for preventing S. aureus pathogenesis in C. elegans in our hands. The likeliest explanation for these discrepancies is the differences between the experiments. For example, the C. elegans experiments differed in host (glp-4 vs glp-4; sek-1) and bacterial strains ( S. aureus strain MRSA131 vs S. aureus strain MW2), timing, and even the host's diets varied (E. coli strain OP50 vs E. coli strain HB101). We have previously shown that switching the diet of C. elegans between the E. coli strains OP50 and HT115 significantly changes its susceptibility to P. aeruginosa PA14 [47], perhaps a similar phenomenon exists for S. aureus . Ultimately, much more work is needed to determine the potential of CD437 to treat S. aureus .

The other high-potential compound was DMAQ-B1. Since this compound includes a quinone group, there is some concern that it will disrupt function of the electron transport chain [48], oxidize heme, lyse erythrocytes [49], and cause the production of reactive oxygen species [50]. The toxicity of this compound is somewhat more controversial, with publications from two labs suggesting that it is non-toxic [51–53] while others have shown considerable toxicity [54].

However, DMAQ-B1 also clearly has antimicrobial activity. As far as we are aware, this activity was not previously reported. The indole and quinone moieties present within the molecule have been shown to possess varying levels of antimicrobial activity in other circumstances [55, 56]. Although, DMAQ-B1 has been used in vivo, the high potential for toxicity makes it an unappealing drug candidate [54]. Studies aimed at reducing the toxicity of DMAQ-B1 have generated safe analogues with toxicity 128-fold less than the original molecule but that retain, or improve, the insulin-stimulating capacity of the original [54]. As an antimicrobial, however, the same molecular function that causes bacterial death may be what causes the toxicity to the host. As a first step to address this question, it may be worth testing an analogue where the quinone moiety has been removed. Unfortunately, we have been unable to identify a commercial supplier for such a compound, but it remains an active area of research in our lab.

Platinum complexes such as oxaliplatin and carboplatin are useful for cancer treatment as they crosslink DNA strands and disrupt replication [57]. While neither compound exhibited substantial toxicity against either C. elegans or mammalian cells, high concentrations were required to elicit their antimicrobial effects. Interestingly, we found that these compounds were particularly effective at reducing the virulence of P. aeruginosa . Previous studies have similarly found a susceptibility of P. aeruginosa to cisplatin, another platinum-complex used to treat cancer [58, 59]. To date, it is unclear how the selectivity towards this Gram-negative pathogen is being generated, since both Gram-negative and Gram-positive bacteria possess robust DNA repair responses like the SOS response [60]. Since platinum complexes show preferences to cross-link certain DNA sequences [61], one possible explanation is that these sites may be more common in P. aeruginosa . A more likely possibility is that the pathogen has differential uptake or removal of platinum complexes, perhaps due to poor recognition by efflux pumps or reduced permeability through Gram-positive cell walls. Alternatively, cisplatin has been shown to disrupt biofilm [59], which may partially underlie its efficacy in mitigating pyoverdine pathogenesis; early steps in biofilm production have been shown to be required for pyoverdine production [62].

Repurposing bioactives to treat bacterial infections shows promise for the rapid identification and development of safe and effective therapies, but it does introduce other challenges. For example, the compounds targeted for repurposing will, by definition, have been developed for other purposes. As such, dosing and off-target effects require careful analysis and attention. If these concerns can be allayed, however, repurposing may be the best way to address the impending crisis of antimicrobial resistance. In addition, our findings add to the conversation about pathogenesis and highlight that, even in simple model organisms, it is a phenomenon that emerges from the concerted activities of the host, the pathogen and their interactions.

Methods

C. elegans and bacterial strains

Worms were maintained on E.coli OP50 at 15 °C [63]. Experiments were performed on young adult worms unless otherwise stated. The strain glp-4(bn2) was used for most of the C. elegans experiments [64]. For the defence pathways' reporter experiments, the following strains were used: AY101 [acIs101, |pDB09.1(pF35E12.5::GFP); pRF4(rol-6(su1006))|] [65], CL2166 [dvIs19, pgst-4::GFP] [66], TJ356 [pdaf-16::DAF-16::GFP] [67], WY703 [fdIs2, |3X-ESRE::GFP; pFF4[(rol-6(su1006)|] [68]. For bacterial and infection assays, P. aeruginosa strain PA14 [69], P. aeruginosa strain PA14-GFP [23], S. aureus strain MRSA131 [70] and E. faecalis strain OG1RF [71] were used. For all assays and overnight cultures, S. aureus was cultured in Tryptic Soy Broth (TSB) media, E. faecalis in Brain Heart Infusion (BHI) media, and P. aeruginosa in Luria-Bertani (LB) media unless otherwise stated.

Liquid-based killing assays

This method has been described previously [22]. In brief, young adult glp-4(bn4) worms were sorted into 384-well plates with 20 worms/well. Overnight cultures for the three pathogenic bacteria were diluted to an OD₆₀₀ of 0.03 ( P. aeruginosa and S. aureus ) or 0.04 ( E. faecalis ) in a solution of S Basal and their respective media. Final media concentrations in the wells were 41 % slow killing media, 10 % BHI and 10 % TSB for P. aeruginosa , E. faecalis and S. aureus killing assays, respectively. For P. aeruginosa liquid killing, MgSO₄ and CaCl₂ (300 µM), and cholesterol (1.6 mg ml⁻¹) were added to the wells. Two-fold serial dilutions of molecules in DMSO and S Basal were added to the wells for a final concentration of 1 % DMSO. Plates were then incubated at 25 °C for 40–48 h ( P. aeruginosa ), 76–90 h ( E. faecalis ) or 85–100 h ( S. aureus ) to allow for killing of the worms. Next, plates were washed and stained for 10 h with Sytox Orange at 1 µM followed by additional washing and imaging.

Biofilm formation and pyoverdine production

Biofilm formation assay was adapted from a previously published method [72]. Overnight cultures were diluted to a final concentration of 0.02 OD₆₀₀ in their respective media (TSB for S. aureus , M9 for P. aeruginosa and BHI for E. faecalis ). Overall, 100 µl of solution was added to each well in a 96-well plate with two-fold dilutions of small molecules starting at 64 µM. Plates were incubated at 37 °C for 24 h, media was aspirated, and wells were washed 3× with PBS. Crystal violet stain was used to measure biofilm after solubilization in 40 µl of 30 % acetic acid (absorbance at 570 nm).

Pyoverdine production was assessed spectrophotometrically, based on its intrinsic fluorescence (Ex 405 nm; Em 460 nm). A Cytation5 multimode plate reader/imager (BioTek, VT) was used.

Synergy assessment

Synergy was determined by calculating ƩFIC for each molecule combination [73]. MICs were calculated for each molecule alone and in combination unless MIC was not within concentration range tested (>64 µM). We then used the Lowe model of synergy to calculate ƩFIC with the equation ƩFIC=FIC_A+ FI_CB=C_a/MIC_a+C_b/MIC_b where C_a and C_b are the concentration of molecules A and B, respectively, when used in combination and MIC_a and MIC_b are the MIC values for the molecules alone. A ƩFIC score of <0.5 was classified as synergy, 0.5–4 no interaction and >4 antagonistic.

Checkerboard assays

Overnight cultures for each bacterial strain were diluted in their respective media to an OD₆₀₀ of 0.03 final concentration in the wells. Bacterial solution was added to a 384-well plate with four wells/condition and a final volume of 80 µl/well. Two-fold serial dilutions of compounds were made in DMSO with at least three sub-MIC concentrations tested for each molecule. Platinum complex molecules were diluted in water with DMSO added to each well for a final concentration of 1 % DMSO in every well. Plates were incubated at 37 °C for 16 h before determining presence of bacterial growth by measuring optical density.

c.f.u. assays

Overnight cultures of bacterial strain were diluted 1 : 2000 in their respective media and added to 96-well plates. Molecules were diluted four-fold in DMSO with the exception of platinum complexes, which were diluted in H₂O with DMSO added afterwards to ensure 1 % DMSO in all wells. Plates were incubated at 37 °C and c.f.u. ml⁻¹ were measured at 0, 4, 8, 12 and 24 h. c.f.u. were measured by addition of 10 µl of culture into 90 µl of water followed by serial dilutions and plating of 5 µl onto an LB-agar plate. Plates were incubated for 24 h at 37 °C before counting colonies.

Longevity assay

Compound plates were made by adding molecules to molten nematode growth media (NGM) agar. Concentrated E. coli OP50 were dropped onto plates as a food source for the worms. Young adult glp-4(bn4) worms were picked onto plates and scored every other day for survival until no worms remained alive. Worms were scored as dead if they failed to move when prodded with a platinum wire. Worms that left the plate were not scored.

MTT assay

Lung bronchial epithelium 16HBE cells were seeded in a 96-well plate at 10 000 cells/well. Cells were allowed to adhere for 24 h before being treated with serial dilutions of compounds. Cells were treated for 24 h. Then media was aspirated and MTT reagent (8 mg ml⁻¹) added for 3.5 h. Media was then aspirated and 100 µl of DMSO was added before reading absorbance at 570 nm. Cells were grown in RPMI media with 10 % FBS and penicillin-streptomycin.

Statistical analysis

For most analyses, Student's t-test was used, and a P-value of 0.05 was considered the cutoff for statistical significance. For the determination of drug treatment in C. elegans rescue, an ANOVA was used with a post-hoc Tukey test (and a Bonferroni correction) to identify conditions that led to significant differences in rescue. In the latter case, α remained 0.05. Calculations were performed in Microsoft Excel 2007 using the Real Stats implementation, available at www.real-statistics.com.