|Home | About | Journals | Submit | Contact Us | Français|
The opportunistic pathogen Pseudomonas aeruginosa causes serious human infections, but effective treatments and the mechanisms mediating pathogenesis remain elusive. Caenorhabditis elegans shares innate immune pathways with humans, making it invaluable to investigate infection. To determine how P. aeruginosa disrupts host biology, we studied how P. aeruginosa kills C. elegans in a liquid-based pathogenesis model. We found that P. aeruginosa-mediated killing does not require quorum-sensing pathways or host colonization. A chemical genetic screen revealed that iron chelators alleviate P. aeruginosa-mediated killing. Consistent with a role for iron in P. aeruginosa pathogenesis, the bacterial siderophore pyoverdin was required for virulence and was sufficient to induce a hypoxic response and death in the absence of bacteria. Loss of the C. elegans hypoxia-inducing factor HIF-1, which regulates iron homeostasis, exacerbated P. aeruginosa pathogenesis, further linking hypoxia and killing. As pyoverdin is indispensable for virulence in mice, pyoverdin-mediated hypoxia is likely relevant in human pathogenesis.
Pseudomonas aeruginosa is a ubiquitous Gram-negative bacterium capable of causing disease in myriad hosts (Chand et al., 2011; Rahme et al., 1995; Tan et al., 1999a). Its ability to infect diverse taxa is attributed to multiple virulence factors: a variety of secreted toxins, siderophores, a quorum-sensing system, and biofilm-formation (Costerton et al., 1995; Smith and Iglewski, 2003; Tang et al., 1996). Human P. aeruginosa infections result in serious complications in burns and eye lesions, and infections can become systemic in immunodeficient patients. Additionally, P. aeruginosa establishes life-long infections in the lungs of patients with chronic obstructive pulmonary disease, diffused panbronchitis, or cystic fibrosis (Govan and Deretic, 1996; Hoiby, 1994; Lieberman, 2003; Lyczak et al., 2000). Finally, this organism remains a stubborn etiological agent responsible for many nosocomial infections (Rosenthal et al., 2010). P. aeruginosa shows high levels of innate antibiotic resistance (De Kievit et al., 2001; Fisher et al., 2005) and outbreaks caused by multidrug resistant strains are on the rise (Obritsch et al., 2005). Exacerbating this, only a few antipseudomonal compounds are currently in development (Bumann, 2008). These factors illustrate the importance of determining the mechanisms of P. aeruginosa virulence and of identifying treatments that may help prevent disease.
Despite ongoing research efforts, the virulence mechanisms underlying many Pseudomonas infection models remain elusive. Unfortunately, no single model, including those in mammals, has succeeded in recapitulating all of the features of P. aeruginosa virulence relevant to human disease, whether chronic or acute. We have utilized Caenorhabditis elegans as a host to develop infection assays for diverse bacterial species, including P. aeruginosa (Powell and Ausubel, 2008). Several features of C. elegans make it desirable for studying host-pathogen interactions, including the ability to easily carry out forward, reverse, and chemical genetic screens, its small size and rapid generation time, and susceptibility to human pathogens. C. elegans-P. aeruginosa infection models are particularly useful, as many P. aeruginosa virulence-related factors are conserved across widely divergent taxa from nematodes to plants to mammals (Kim and Ausubel, 2005; Rahme et al., 1995; Rahme et al., 1997; Tan and Ausubel, 2000). In addition, the human innate immune system shares many characteristics with that of C. elegans, despite the relatively simple immune response pathways of the latter.
Here we report that P. aeruginosa-mediated killing of C. elegans in a liquid pathogenesis format requires the siderophore pyoverdin and the phosphatase activity of the bacterial dual-function, two-component sensor KinB. Unlike other C. elegans pathogenesis assays, known quorum-sensing pathways, intestinal colonization, and phenazines are dispensable for killing in the liquid assay. A library of known bioactive chemicals was used to identify virulence inhibitors. One hit, the iron-chelating compound ciclopirox olamine, implicated iron and the siderophore pyoverdin in virulence, demonstrating the value of querying host-pathogen interactions in the context of a high-throughput, whole-organism approach. Importantly, we show that PA14 triggered a lethal hypoxic crisis in C. elegans that requires the hypoxia-inducing factor HIF-1 for host defense. This hypoxic response is at least partially dependent upon pyoverdin. Combined, these data demonstrate a previously unknown role for pyoverdin in P. aeruginosa virulence that is likely reflected in mammalian infection.
To study the factors driving P. aeruginosa strain PA14 virulence, and to facilitate high-throughput screening (N. Kirienko and F. Ausubel, unpublished data), we established a liquid-based killing assay (LK assay) using C. elegans as a host (described in detail in Materials and Methods and summarized in Figure S1). Virtually no host death was observed in the LK assay, as shown by Sytox Orange staining (Moy et al., 2009), when worms were exposed to the normal laboratory food E. coli OP50 (Figure 1A, D), consistent with previous observations that C. elegans can be maintained and grown in liquid medium (Stiernagle, 2006). PA14, on the other hand, killed most worms within 48 hours post inoculation (hpi) (Figure 1B, D). Host death was contingent upon the presence of live PA14, as most worms survived the incubation period when gentamicin was added (Figures 1C). Almost no killing was observed within the first 24 hpi regardless of the initial bacterial inoculum, but a relatively low starting titer of PA14 was sufficient to cause host death by 48 or 72 hpi (Figure 1E), suggesting that time is necessary for the lethal interaction between the host and the pathogen to be established. In contrast to PA14, very high initial concentrations of OP50 were necessary to kill (Figure 1E). Combined, these data show that the LK assay is robust and that killing requires exposure to live P. aeruginosa.
To determine whether LK shares P. aeruginosa-encoded virulence determinants with previously described C. elegans killing assays, we first compared it to agar-based intoxication, referred to as “fast-killing”, which takes place on rich media (Mahajan-Miklos et al., 1999). Fast-killing occurs within 24 hours, and the key determinant is the production of phenazines, a family of toxic small molecules (Cezairliyan et al., 2013; Mahajan-Miklos et al., 1999). Therefore, we tested PA14Δphz (Dietrich et al., 2006), a strain devoid of phenazine biosynthetic enzymes that is dramatically attenuated in fast-killing (Cezairliyan et al., 2013). This strain showed no attenuation in liquid (Figure 2A), indicating that liquid- and fast-killing utilize different mechanisms.
Cyanide-based lethal paralysis is another well-studied virulence mechanism observed during infection of C. elegans with P. aeruginosa PAO1 (Darby et al., 1999). We ruled out the involvement of cyanide in LK by determining that cyanide production under LK assay conditions was negligible for PAO1, MPAO1, and PA14 strains (Figure S2A). Like PA14, both PAO1 and MPAO1 killed C. elegans in the LK assay (Figure S2B). In addition, PA14, PAO1, and MPAO1 strains with lesions in cyanide production genes were indistinguishable from their wild-type parents (Figure S2B).
Agar-based “slow-killing” (SK) takes place on modified C. elegans NGM media (Tan et al., 1999a) and is the most commonly used C. elegans-P. aeruginosa assay. Although the mechanisms underlying pathogenesis in this assay remain unclear, multiple virulence factors, including quorum sensing, are required. An extended panel of PA14 mutants attenuated in SK, both canonical (e.g. gacA, lasR, pqsE, mvfR) and recently described (e.g. kinB, clpA, etc.) was tested in our liquid assay (Table S1, Figure 2A) (Feinbaum et al., 2012; Rahme et al., 1995; Tan et al., 1999b). Of these mutants, only kinB exhibited reduced virulence in LK. This distinguishes our liquid-killing assay from a previously published liquid-based C. elegans-P. aeruginosa killing assay, in which gacA mutants showed strong attenuation (Garvis et al., 2009).
Colonization of the host intestinal tract is thought to be a key pathogenic determinant in SK (Tan et al., 1999a). Mutations in C. elegans that increase resistance to PA14 infection also exhibit reduced intestinal colonization (Evans et al., 2008; Garsin et al., 2003). To investigate whether mutants attenuated in SK show decreased intestinal colonization, we used derivatives of OP50, PA14, and PA14ΔlasR engineered to express dsRed. LasR is a transcriptional regulator mediating a key PA14 quorum-sensing system (Williams and Camara, 2009). As expected, OP50-dsRed showed no intestinal accumulation under SK conditions (Figure 2B), whereas wild-type PA14-dsRed displayed strong colonization (Figure 2C). PA14ΔlasR-dsRed, which is significantly impaired in SK (data not shown), displayed a marked reduction in intestinal colonization compared to wild-type PA14 (Figure 2D). We used the same bacterial reporters to query colonization in liquid, but none of the strains displayed significant intestinal colonization (Figure 2E–G). Colonization was quantified for SK and LK assays and confirmed these results (Figure S2C, D). Together, these data suggest that significantly different mechanisms underlie pathogenesis in the SK and LK assays. This may be partially explained by the much higher bacterial density in SK, where worms are exposed to a saturated lawn of bacteria.
KinB, a two-component sensor histidine kinase that regulates motility, virulence factor production, and biofilm formation, is a key determinant for P. aeruginosa infection of zebrafish (Chand et al., 2011) and is important for virulence in a murine acute pneumonia model (Damron et al., 2012). Intriguingly, zebrafish studies suggested that neither the kinase activity of kinB nor its canonical partner, algB, were required for acute virulence (Chand et al., 2011), but that the phosphatase activity of KinB was critical (Chand et al., 2012). To investigate the role of kinB in LK, we first carried out a time course study to confirm that PA14ΔkinB was attenuated in both LK (Figure S2E) and SK (Figure S2F). Next, we demonstrated that neither algB deletion nor abrogation of KinB kinase activity had a significant effect on pathogenicity in LK (Figure S2G). In contrast, the phosphatase activity of KinB was indispensable for wild-type virulence in LK, as loss of critical catalytic residues strongly ameliorated pathogenesis. These data suggest that PA14 utilizes an acute, rather than a chronic, virulence mechanism for LK.
As an additional method to gain insight into the pathogenic mechanisms involved in PA14-mediated LK, we used an unbiased chemical genetics approach. We tested 1600 known bioactive molecules for their ability to attenuate virulence and obtained 18 hits (Figure S3). The hits included antibiotics (11), topical disinfectants (2), organomercuric compounds (2), chemotherapeutics (2), and a chelating agent (1). The presence of antibiotics among the hits was expected, as PA14 is susceptible to the families of antibiotics identified. Similarly, the chemotherapeutics (5-fluorouracil and cisplatin) interfere with DNA replication, and may function as nonspecific antimicrobials. The identification of the ferric iron-chelating agent ciclopirox olamine (CPX) was unexpected, and suggested that iron molecules may serve an important role in the host-pathogen interaction. We tested whether two additional iron-chelating agents, 1,10-phenanthroline (Phe) and 2,2′-bipyridine, could alleviate PA14 pathology in LK. Phe can chelate both ferric and ferrous iron, while bipyridine (which exhibits the weakest rescue) binds mostly ferrous iron. All three chelators exhibited dose-dependent rescue (Figure 3D), confirming that iron plays an important role in the killing process.
Many microbes synthesize siderophores, small polypeptides with high affinity for oxidized iron, to promote acquisition of this critical nutrient (Crosa, 1989; Schalk, 2008). P. aeruginosa is known to produce two major siderophores, pyoverdin and pyochelin. Siderophore biosynthesis and iron acquisition are known to be required for P. aeruginosa virulence in both plant and mammalian infections (Meyer et al., 1996; Nadal Jimenez et al., 2010; Takase et al., 2000). We hypothesized that disrupting siderophore biosynthesis would mimic the effects of adding a strong chelating molecule, such as CPX or Phe, and would block LK by PA14.
We tested a panel of previously generated PA14 transposon-insertion mutants (Liberati et al., 2006) that affected either pyoverdin or pyochelin biosynthesis. Disruption of pyoverdin but not pyochelin biosynthesis resulted in virulence attenuation (Figure 4A). Although pvd mutants were attenuated in LK, prolonged exposure still resulted in substantial C. elegans death (Figure 4C), suggesting multiple independent virulence mechanisms. None of the pvd mutants tested showed attenuation in SK (Figure 4D). These data suggest that pyoverdin plays a crucial role in P. aeruginosa virulence that is specific to LK.
Deletion of pvdA, but not pchD, in P. aeruginosa strain PAO1 causes reduced virulence in several murine infection models (Takase et al., 2000). Similarly to the PA14 pvd and pch mutants, both PAO1ΔpvdA and PAO1ΔpvdF were attenuated in LK, whereas PAO1ΔpchBA, which has been shown to produce virtually no pyochelin (Reimmann et al., 1998), displayed wild-type virulence (Figure 4B).
The simplest explanation for the ability of the three iron chelators to attenuate PA14-mediated LK is that they inhibit PA14 growth. Indeed, the minimal inhibitory concentrations (MICs) for CPX, Phe, and bipyridine in LK media were 25, 50, and 400 μM, respectively, which were about the same as their effective doses in the LK assay (Table S2). These data suggested that the reduced virulence of pvd mutants might be a consequence of poor growth of the mutants under LK assay conditions due to iron-deficiency. However the pvd mutants exhibited wild-type growth kinetics and final growth densities in both nutrient-rich and nutrient-poor conditions (data not shown). An alternative explanation for chelator-mediated rescue is that nutrient-poor media like LK nonspecifically sensitize PA14 to adverse environmental conditions. This seems unlikely, however, because PA14 was at least as resistant to antibiotics in LK as in LB.
While we observed that treatment with metal-chelating compounds rescued worms at low concentrations (Figure S4A), we also noticed that significantly increasing the concentrations of these molecules resulted in host death (Figure S3A for Phe and data not shown for CPX and bipyridine). As the lethal Phe dosage was more than 20-fold higher than the MIC for PA14, it seemed probable that worm killing at this high dose was due to Phe toxicity, not bacterial virulence. Indeed, high concentrations of Phe caused worm killing, even in the absence of PA14 (Figure S3B). These data suggest that iron sequestration causes host death by hijacking oxidized iron required for normal biological functions.
A critical iron-dependent process in aerobically respiring organisms is oxidative phosphorylation. Importantly, there is a direct connection between iron homeostasis and the HIF proteins, a family of hypoxia-inducible transcription factors in both C. elegans and mammals (Peyssonnaux et al., 2007; Romney et al., 2011). HIF-1 is the C. elegans ortholog of the mammalian HIF-1α transcription factor and is crucial for worm survival during hypoxia (Jiang et al., 2001).
Reasoning that treatment with a strong iron chelator like Phe could induce a hypoxic response in C. elegans, we used qRT-PCR to measure expression of the ten most upregulated genes that are dependent on HIF-1 and the ten most upregulated HIF-1-independent genes induced during hypoxia (Shen et al., 2005). Treatment of worms with a lethal dose of Phe in the LK assay with OP50 for 16 hours resulted in an average 60.6-fold upregulation of hypoxia-responsive genes compared to an untreated control (Figure 5A). Mock-treatment, or treatment with a low concentration of Phe, did not result in significant upregulation (1.4- and 3.7-fold on average, Figure 5A and data not shown). Combined, these data suggest that treatment with a high concentration of an iron-chelating compound is sufficient to induce a hypoxic response in C. elegans
The shared molecular function of pyoverdin and Phe (i.e., iron-binding) suggested that pyoverdin synthesized by PA14 might also cause a hypoxic response in LK. qRT-PCR analysis of worms 16 hpi with PA14 showed significant (40.2-fold, on average) upregulation of hypoxic response genes (Figure 5A). C. elegans infected on solid media, however, showed little change in these transcripts, demonstrating specificity of this hypoxic response to LK. Furthermore, the hypoxic response was mitigated when a pyoverdin biosynthesis mutant was substituted for PA14 (Figure 5A for PA14pvdF and data not shown for PA14pvdE and PA14pvdP). These data support a model in which pyoverdin production disrupts host iron homeostasis and triggers a hypoxic response.
As HIF-1 is necessary for a portion of the hypoxic response observed in LK, we tested whether hif-1 mutations affect sensitivity to PA14 in LK and SK assays. hif-1(ia4) mutants were more susceptible than wild-type worms in the liquid assay (p=1.91e-07) (Figure 5C), but were more resistant than wild-type in the SK assay (p=1.61e-06) (Figure 5D). To test whether hif-1 mutants were non-specifically susceptible to liquid conditions, they were incubated with OP50 instead of PA14. After 3 weeks of incubation, less than 10% of the worms were dead, which was comparable with wild-type N2 worms (data not shown). This suggests that the difference in susceptibility is not merely a consequence of hypoxic conditions in the assay. We also saw no difference between hif-1 mutants and wild-type N2 when infected with Enterococcus faecalis in liquid (data not shown) using a previously described C. elegans-E. faecalis liquid infection assay (Moy et al., 2009). These data, along with the qRT-PCR results described above, suggest that the hypoxic response is specific to LK and that the HIF-1-dependent hypoxic response serves a protective role against P. aeruginosa but not E. faecalis.
As pyoverdin is a secreted non-ribosomal polypeptide, we hypothesized that cell-free filtrates from PA14 cultures would also induce a hypoxic response in the LK assay. Indeed, PA14 filtrates from cells grown to saturation in iron-poor M9 medium (which induces pyoverdin biosynthesis), but not from OP50 cultures, were sufficient to trigger a hypoxic response (Figure 5B). The level of pyoverdin measured in LK media after 44 hpi was comparable to its level in 50% PA14 filtrate, which was used for experiments (data not shown). Pyoverdin biosynthesis is abrogated under iron-replete conditions (Cornelis et al., 2009; Ochsner et al., 1995), and any pyoverdin or other siderophore produced under conditions of iron excess would likely bind the iron present in the medium, limiting the pathogenic potential of the siderophore. Consistent with the hypothesis that pyoverdin is the active factor in the filtrates causing worm death, PA14 filtrate produced from iron-supplemented cultures contained no pyoverdin (Figure 6A) and elicited virtually no hypoxic response (Figure 5B).
We also tested whether the pyoverdin level in filtrates was sufficient to mediate lethality in the LK assay. As expected, little or no death was observed when unconditioned M9 or OP50 filtrates were added to the assay (Figure 6C). In contrast, PA14 filtrate showed significant killing. Addition of exogenous iron during PA14 growth significantly attenuated (p=0) killing by this filtrate, demonstrating that the iron-chelating property of pyoverdin was critical for toxicity.
Pyoverdin also promotes the production of several P. aeruginosa virulence factors, including exotoxin ToxA and the protease PrpL (Lamont et al., 2002). It is possible that iron supplementation during overnight growth diminished production of these pathogenic determinants, thereby decreasing killing. To test this, we split a filtrate produced from an overnight PA14 culture and incubated half with 100 μM iron overnight. We reasoned that exogenous iron would bind to the pyoverdin, decreasing its ability to trigger hypoxia without affecting ToxA, PrpL, or other pyoverdin-dependent virulence factors. Iron acquisition by pyoverdin was verified by monitoring a time-dependent decrease in fluorescence (Figure 6B). This filtrate, PA14/Fe*, which was supplemented with iron after filtration, showed reduced killing in the liquid assay (p=9.21e-07) compared to untreated filtrate (Figure 6C). Taken together, our data demonstrate that pyoverdin-mediated sequestration of host iron is sufficient to induce a hypoxic response and lethality.
P. aeruginosa encodes multiple virulence pathways that may act simultaneously in pathogenesis. To examine the importance of iron sequestration in the context of an active interaction between PA14 and C. elegans, we tested whether iron supplementation had an effect. Addition of 100 μM ferric chloride to the LK assay significantly attenuated host killing (p=0.004), confirming the biological significance of pyoverdin toxicity (Figure S4).
We report that P. aeruginosa causes C. elegans killing in a liquid assay. Unlike previously-reported C. elegans-P. aeruginosa pathogenicity assays (Garvis et al., 2009; Mahajan-Miklos et al., 1999; Tan et al., 1999b), LK is independent of phenazine production, known quorum-sensing pathways, and intestinal colonization. Using a high-throughput screen, we identified a role for iron in PA14-mediated LK, which was confirmed by the attenuated virulence of mutants defective in pyoverdin biosynthesis. As host iron is crucial for multiple biological processes, we reasoned that pyoverdin may be directly functioning as a worm-killing toxin, similar to treatment with a lethal concentration of chelator (Figure 7). qRT-PCR analysis showed strong induction of hypoxic response genes in LK with PA14, which was abolished when pyoverdin mutants were used. Consistent with this, hif-1 mutants, which are sensitized to hypoxia, exhibited increased death in LK but not in SK or during infection with E. faecalis. Finally, we used cell-free filtrates to determine that pyoverdin is sufficient to induce a hypoxic response and killing in the absence of PA14. Filtrate produced from an iron-supplemented culture of PA14 (a condition known to limit pyoverdin production) did not kill the host. Incubation with iron after filtration also alleviated LK. Combined, these data demonstrate that PA14 causes pyoverdin-dependent disruption of host iron homeostasis, triggering a hypoxic response and death. Although pyoverdin is sufficient to mediate LK, as discussed below, other factors are also required for wild-type levels of virulence.
Pyoverdin is critical for P. aeruginosa, as iron acquisition is typically required both for normal growth and for pathogenesis (Meyer et al., 1996; Takase et al., 2000). In addition, iron acquisition by pyoverdin or pyochelin was required for the ‘red death’ phenomenon observed in C. elegans infected with P. aeruginosa strain PAO1 (Zaborin et al., 2009). Pyoverdin biosynthesis is negatively regulated by the presence of intracellular iron (Cornelis et al., 2009), and pyoverdin autoregulates its own production in a feed-forward loop until sufficient iron is acquired. We note that the PA14pvdE mutant, which showed the greatest attenuation of the pvd mutants tested, produces a small amount of fluorescence that is characteristic of pyoverdin (Table S3). We speculate that translation of the truncated PA14pvdE transcript may result in a partially functional protein with a dominant-negative phenotype. Unfortunately, little is known about the function of the pvdE gene product, precluding a definitive explanation of the pvdE mutant phenotype. It is also worth noting that it is unlikely that heme metabolism is altered in this assay, as heme-responsive genes (Severance et al., 2010) showed no significant transcriptional changes by qRT-PCR (data not shown).
Consistent with the hypothesis that pyoverdin may hijack host-assimilated iron in mammals, pyoverdin stimulates bacterial growth in media supplemented with human serum or plasma (Cox and Adams, 1985), suggesting that it can use the ferriproteins in these fluids as a source of iron. Pyoverdin has also been shown to directly remove iron from transferrin (Meyer et al., 1996). The removal of iron from normal iron transport proteins can also cause iron toxicity due to the production of hydroxide radicals (Baldwin et al., 1984; Coffman et al., 1990). Whether pyoverdin might directly cause damage to mammalian tissues by iron removal remains an intriguing and important question.
Observations that P. aeruginosa and other pathogens can elicit a HIF-1-associated hypoxic response in mice and humans are also consistent with a potential role for direct pyoverdin toxicity in mammalian infections. For example, human HIF-1 stabilization has been observed after exposure to P. aeruginosa in intestinal epithelial cells and in bronchial epithelial cells derived from a CF patient (Koury et al., 2004; Legendre et al., 2011). Recent work has also suggested that HIF-1 transcription factor activity is important for optimal innate immune function during bacterial infections in mice (Nizet and Johnson, 2009). HIF-1 activity seems to play an important role, as a tug-of-war takes place during the host-pathogen interaction. Under normoxic conditions, HIF-1 is targeted for degradation by EGL-9 family hydroxylases (Epstein et al., 2001). When iron, an essential EGL-9 co-factor, is depleted (i.e., by treatment with a chelating molecule or siderophore), HIF-1 is stabilized. In turn, P. aeruginosa appears to have developed a mechanism to counteract HIF-1 stabilization by targeting HIF-1α for degradation using the 26S proteasome and its own quorum-sensing machinery (Legendre et al., 2012). Either egl-9(sa307) or egl-9(n586ts) were insufficient to increase survival in the presence of PA14 in LK, as was vhl-1(ok161), the E3 ligase targeting HIF-1 for degradation (N. Kirienko, unpublished data). It is currently unclear whether this is due to HIF-1 stabilization alone being insufficient to promote survival under these conditions, observed pleiotropy for egl-9 alleles during infection (Luhachack et al., 2012), or HIF-1 inhibition via a VHL-1-independent mechanism (Shao et al., 2010). Therefore, the role of HIF-1 stabilization in response to P. aeruginosa infection requires further investigation.
P. aeruginosa pyoverdin mutants may also exhibit reduced virulence in mammalian infection models because they are deficient in virulence factor production. Pyoverdin regulates the production of a variety of virulence factors, including ToxA and the protease PrpL in addition to its own biosynthesis (Lamont et al., 2002; Vasil and Ochsner, 1999). Interestingly, PrpL is capable of mediating degradation of several secreted proteins, including lactoferrin and transferrin (Wilderman et al., 2001), which may mediate iron release for siderophore-mediated acquisition. Such degradation has been reported in patients with chronic lung conditions, including CF patients (Britigan et al., 1993). In this way, pyoverdin regulates the production of virulence factors that cause the release of iron from host proteins in addition to taking iron directly from them.
In contrast to pyoverdin, disruption of pyochelin, the other major siderophore of P. aeruginosa, had little effect on LK. This is consistent with observations in mammals, where pyochelin disruption alone does not result in attenuation but does show a synthetic interaction when combined with pyoverdin removal (Takase et al., 2000). Unfortunately, double mutants defective in both pyoverdin and pyochelin biosynthesis exhibited severe growth defects in LK medium (data not shown), limiting the conclusions that could be drawn. One possible explanation that has been put forth for this synthetic interaction is that both siderophores play roles in virulence, but that pyoverdin is more important (Ankenbauer et al., 1985; Takase et al., 2000).
Although we have shown that pyoverdin is an important virulence factor in the C. elegans LK model, PA14 pvd mutants are still able to kill the nematodes, suggesting that other virulence factors remain to be discovered. An important P. aeruginosa virulence factor in mammalian models is alginate, which serves as a prognosticator for outcomes in CF patients (Ramsey and Wozniak, 2005). A key regulator in alginate production is KinB, which has been implicated in C. elegans SK, zebrafish, and mouse infection models (Chand et al., 2011; Damron et al., 2012; Feinbaum et al., 2012) and is also required for LK (this study). KinB’s role in alginate production, however, does not appear to be relevant for LK. Alginate is more commonly involved in chronic infections and is not thought to be necessary for acute virulence (Yorgey et al., 2001), and a recent report suggests that KinB may be key in mediating the transition between acute and chronic P. aeruginosa infections (Chand et al., 2012). Also, it is unlikely that KinB plays a major role in regulating pyoverdin production in LK since kinB mutants synthesize normal levels of pyoverdin (data not shown). Further characterization of KinB in P. aeruginosa virulence will be necessary to identify its role(s) in pathogenesis, both in C. elegans and in vertebrates.
P. aeruginosa represents a serious risk to patients who have CF, immunodeficiencies, or are in hospital settings. The ubiquity of P. aeruginosa ensures that this bacterium will remain an obstacle to human health, and its propensity to develop antimicrobial resistance demonstrates that increased understanding of its virulence is critical and that novel treatments are necessary. The identification of novel drug leads requires the development of assays that are amenable to high-throughput screens, similar to the LK assay presented here. Model systems like this are of limited value, however, unless the mechanisms underlying virulence are understood. Our development of a PA14 liquid-killing platform, and the demonstration that C. elegans killing is directly mediated by pyoverdin toxicity, should prove useful for Pseudomonas researchers and will further our understanding of complex host-pathogen interactions.
Additional detailed methods are presented in the supplemental Experimental Procedures available online.
glp-4 worms were synchronized by hypochlorite isolation of eggs from gravid adults, followed by hatching of eggs in S-basal. L1 larvae were transferred to NGM plates seeded with OP50 and incubated at 15°C for 16h and then transferred to 25°C for 48h. Worms were rinsed from the plates and washed in S-basal. 350μL of an overnight PA14 LB culture was spread on an SK plate and incubated at 37°C for 24h then at 25°C for 24h. Bacteria were scraped from SK plates and resuspended in S-basal supplemented with 5μg/mL cholesterol. The titer was determined by spectrometry, and adjusted to a final OD600 = 0.03. 17μL of SK media and 33μL of resuspended bacteria were added per well. 18 worms were dispensed into each well of a 384-well plate using a COPAS BioSort robot (Union Biometrica). Plates were sealed with gas-permeable membranes (Diversified Biotech) and incubated at 25°C without agitation at 80–85% humidity for 44–48 hours. Subsequently, worms were washed five times using a BioTek ELx405 microplate washer and Sytox Orange (Invitrogen) was added to a final concentration of 0.7μM. After 18–24h plates were imaged in both brightfield and Cy3 fluorescence channels using an IXMicro automated microscope (Molecular Devices). Worm survival was scored by automated image analysis using CellProfiler (Kamentsky et al., 2011; Moy et al., 2009). The pipeline consisted of illumination correction (to flatten lighting abnormalities), adaptive intensity thresholding (to identify worms in brightfield images), and normalization of fluorescence area to brightfield area. Size- and contrast-based filters were used to exclude debris and other artifacts.
For assays using N2 or hif-1(ia4) mutants, worms were prepared via hypochlorite egg preparation as above, but L1 larvae were transferred to NGM plates seeded with RNAi targeting the cell-cycle gene cdc-25.1 to ensure the sterility of adults used for LK assays.
Screening was performed at the National Screening Laboratory for the Regional Centers of Excellence in Biodefense and Emerging Infectious Diseases (NSRB) at Harvard Medical School. 384-well plates (Corning #3712) were filled with 25μL media (see above) using a WellMate Microplate Dispenser (Thermo Scientific). 0.3μL of compound in DMSO were pin transferred into each well using an Epson Compound Transfer Robot system, and PA14 and nematodes were added. Each plate had DMSO- and gentamicin-treated wells to serve as negative and positive controls, respectively (Figure 3). Compounds were tested at 20μg/mL and chemicals that exhibited rescue in duplicate plates (>3 SD from DMSO control) were considered hits.
For RNA collection, LK assay was performed as described above, except that volumes were adjusted to 150μL and the assay was performed in 96-well plates with 40 worms per well. Worms from two 96-well plates were combined 16 hpi (all conditions except filtrates) or 24 hpi (filtrates) and washed twice with S-basal. RNA purification and qRT-PCR were performed as previously described (McEwan et al., 2012), except that fold-changes were calculated using a ΔΔCt method. Primer sequences are available upon request.
Bacteria were inoculated into M9 medium as described above and grown 18–20 hours at 37°C with agitation. Bacteria were pelleted by centrifugation at 4,000g for 30 minutes. Supernatants were decanted, and sequentially filtered through 0.45, 0.45, and 0.20μm filters. Pyoverdin production was determined spectrophotometrically by measuring emission at 460nm with an excitation of 400nm. Filtrates were stored at 4°C for up to 24h prior to use.
We are grateful to L. Rahme, L. Dietrich, D. Newman, D. Hung, I. Schalk, S. Lory, C. Manoil, and D. Haas for providing bacterial strains and the Caenorhabditis Genetics Center for worm strains. We also wish to express our gratitude to V. Mootha and V. Cracan for helpful discussions. We thank members of the Ausubel and Ruvkun labs for critical reading of the manuscript and experimental advice. Finally, we wish to express our gratitude to S. Chiang and the staff of the NSRB for providing resources and facilities for high-throughput screening. This study was supported by the Massachusetts Biomedical Research Corporation Tosteson Postdoctoral Fellowship Award to NVK and by the following grants from the National Institutes of Health: F32 AI100501 awarded to NVK; R01 AI085581 awarded to FMA, R01 GM095672 awarded to CW, T32 DA013911 awarded to DRK, and U54 AI057159 awarded to NSRB.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.