Protective stress response triggered by infection-driven reactive oxygen in C. elegans through p38 MAPK signaling

Previous work in our laboratory demonstrated that C. elegans releases significant amounts of ROS during infection and that expression of oxidative stress response genes such as sod-3 is induced [6]. Additionally, we have shown that cells at the site of the infection, the intestine, display a loss of protein homeostasis indicative of cellular stress [7]. SKN-1 is a transcription factor that responds to cellular stressors, including ROS [8], and we predicted that its activity is likely to be induced as a result of pathogen exposure. We tested genes previously characterized as being regulated by SKN-1, including gst-4, gst-5, gst-7, gst-10 and gcs-1. The gst genes encode for glutathione-S-transferases, which are important for detoxifying reactive compounds by conjugating reduced glutathione to electrophilic centers, while gcs-1 encodes for gamma-glutamine cysteine synthetase heavy chain, a protein involved in glutathione biosynthesis [10], [11], [20]. Expression of these genes was examined by qRT-PCR following a 24-hour exposure of L4 worms to both P. aeruginosa and E. faecalis (Figure 1A). All of the reporter genes were induced between two and five-fold more when the animals were feeding on the pathogens, as compared to those feeding on their normal laboratory food source (E. coli OP50). As shown in Figure 1B, in which exposure of the animals to skn-1 RNAi preceded exposure to E. faecalis, the induction was prevented, indicating that expression of these genes retains dependence on SKN-1 under conditions of pathogen exposure.

The Blackwell and Johnson laboratories have generated several C. elegans strains containing GFP reporter fusions to genes that are markers for SKN-1 activity including gst-4[21], gcs-1[8], [22] and gst-7[20]. We obtained these strains to further examine pathogen-induced SKN-1 activity. L4 animals were exposed to P. aeruginosa, E. faecalis or non-pathogenic E. coli for 18 hours. Fluorescent micrographs were taken, and the animals were scored for low, medium or high expression of the reporters as described in Methods. For each condition, several worms in a representative micrograph are shown (Figures 1C-E, 1G-I), and additionally the quantification of the scoring with statistical analysis is included (Figures 1F and 1J). Figures 1C-F show that significantly more animals had high levels of expression of gst-4::gfp when infected with E. faecalis (Figure 1C) and P. aeruginosa (Figure 1D) than when allowed to feed on E. coli (Figure 1E). Significantly higher expression of gcs-1::gfp was also observed on the pathogenic strains compared to the controls as shown in Figures 1G-J. We also examined the levels of GST-7::GFP using a transgenic strain containing a translational fusion of gfp to the gst-7 gene, and again we observed higher levels when the animals were exposed to the pathogens (Figure S1 in Text S1). To ensure that the expression was SKN-1 dependent, the levels of fluorescence were observed in strains exposed to skn-1 RNAi prior to pathogen exposure; the pathogen-induced increase in fluorescence was abolished, supporting the use of these reporters as read-outs for SKN-1 expression levels (see below, S2E in Text S1 and data not shown). In independent experiments looking at number of transcripts by RNAseq experiments that examine transcript numbers in worms exposed to another human pathogen, Staphylococcus aureus, both gcs-1 and gst-4 were expressed at higher levels compared to animals on E. coli (Javier Irazoqui, personal communication, data not shown). In conclusion, several different experimental methodologies indicate that SKN-1 regulated genes have increased levels of expression in C. elegans fed on pathogenic bacteria.

To investigate whether or not the strength of the SKN-1 response is affected by any known bacterial virulence factors, we tested some well-studied mutants. GacA is a response regulator in P. aeruginosa that strongly affects the virulence of this pathogen and a gacA mutant is greatly attenuated in a variety of hosts including C. elegans[23]. As shown in Figure 2B, a gacA deletion mutant significantly reduced the expression level of the gcs-1::gfp reporter as compared to the isogenic parental strain of P. aeruginosa shown in Figure 2A. One of the GacA-dependent virulence factors is pyocyanin [24], a secreted compound that is redox active and has been shown to oxidatively stress host cells [25] –[27]. We examined a pyocyanin-deficient mutant, which is deleted in phzM, a gene that encodes an enzyme critical for pyocyanin biosynthesis [28], [29]. As seen in Figure 2C, this mutant also results in less gcs-1::gfp reporter expression as compared to wild-type. The quantification of the effects of these P. aeruginosa mutants on gcs-1::gfp expression is shown in Figure 2D. In E. faecalis, FsrB is a component of a quorum sensing system that has been found to affect virulence in every animal model studied, including C. elegans[30], [31]. An fsrB deletion mutant also resulted in less SKN-1 activity, as measured by gst-4::gfp expression compared to the isogenic parental strain (Figure 2E-G). These results show that the level of SKN-1 activity is sensitive to the presence or absence of major virulence factors and that SKN-1 activity may be a way to discern the overall virulence of the infecting organism.

SKN-1 activity is regulated by localization to the nucleus [8], [32]. If pathogen exposure increases SKN-1 activity, one would expect to see localization of this transcription factor to the nucleus. We obtained a SKN-1B/C::GFP transgenic line used in previous studies to examine SKN-1 activity and localization [8]. As shown in Figure 3, exposure to P. aeruginosa (Figure 3A) or E. faecalis (Figure 3B) causes significant nuclear localization of SKN-1B/C::GFP, similar to what is observed by exposure to paraquat (Figure 3C), a chemical that causes oxidative stress and is documented to promote SKN-1 nuclear localization [8]. In contrast, when feeding on their normal laboratory food source, the animals do not display significant SKN-1B/C::GFP nuclear localization (Figure 3D).

Utilizing chemicals known to generate oxidative stress (paraquat, sodium arsenite and t-butyl peroxide), previous work demonstrated that activation of SKN-1 is dependent on p38 MAPK signaling components [12], [16], [17]. We asked whether or not pathogen-induced activation of SKN-1 is also dependent on p38 MAPK signaling. To investigate this question, the levels of GFP were scored in the gst-4::gfp transgenic worms after RNAi knock-down of the genes-of-interest followed by 18 hours of exposure to E. faecalis (Figure 4). Additionally, we assayed gcs-1::gfp transgenics exposed to P. aeruginosa following RNAi (Figure S2 in Text S1). As a control, the effects of RNAi knock-down on animals exposed to the non-pathogenic control, E. coli, were also assayed and found to be minimal (Table S1 in Text S1).

The p38 MAPK, PMK-1, and the upstream MAPKK, SEK-1, are absolutely necessary to activate SKN-1 as a result of oxidative stress [12]. As shown in Figure 4, reduction in the expression of sek-1 (Figure 4C) or pmk-1 (Figure 4D) by RNAi prior to exposure to E. faecalis resulted in significantly less fluorescence of the SKN-1 dependent reporter, gst-4::gfp, in the worm intestine, with similar levels to the skn-1 RNAi control (Figure 4E). These data suggest that PMK-1 and SEK-1 are also crucial for SKN-1 activation as a result of pathogen exposure. RNAi of these genes had similar effects on gcs-1::gfp transgenics exposed to P. aeruginosa (Figure S2C and S2D in Text S1).

In addition to its role in responding to oxidative stress, it was previously demonstrated that PMK-1 is activated by pathogen exposure and plays a very important role in host defense. The infection protective activity of PMK-1 is dependent not only on SEK-1, but also on the upstream MAPKKK NSY-1 and the Toll/IL-1 receptor (TIR) domain protein, TIR-1 [13], [14]. The role of NSY-1 and TIR-1 in PMK-1 activation under conditions of oxidative stress is less clear and may be dependent on the stressor utilized [12], [16], [17]. To investigate the possible roles of NSY-1 and TIR-1 on SKN-1 activation during pathogen exposure, we exposed the animals to nsy-1 or tir-1 RNAi prior to placing them on pathogen. Knock-down of nsy-1 caused loss of SKN-1 activity on both pathogens as assayed by the gst-4::gfp and gcs-1::gfp reporters, (Figure 4B, Figure S2B in Text S1). In contrast, knock-down of tir-1, resulted in only a non-statistically significant trend towards less expression, suggesting no, to minimal involvement, at best (Figure S3B in Text S1 and data not shown).

To confirm the RNAi results using another methodology, we also examined expression in genetic mutants using qRT-PCR to measure the expression of gst-4 and gcs-1. In null mutants of nsy-1, sek-1 and pmk-1, expression of these two genes was down relative to wild-type worms on both E. faecalis and P. aeruginosa (Figure 4G). In contrast, expression levels were not significantly different in a tir-1 mutant.

Our previous studies demonstrated that part of the response of the worm to pathogen challenge is the release of ROS into the intestinal lumen by the dual oxidase Ce-Duox1/BLI-3. Rather than a tangential consequence of cell death, this response is purposeful and protective [3], [6]. Since SKN-1 is known to respond to oxidative stress, and the above-described experiments conducted on the p38 MAPK signaling pathway are most consistent with an oxidative stress response, we postulated that the ROS generated by Ce-Duox1/BLI-3 during pathogen exposure may trigger signaling through the p38 MAPK pathway, resulting in SKN-1 activation. To test this, we knocked down the expression of bli-3 by RNAi in the gst-4::gfp background and then exposed the strain to E. faecalis. As shown in Figure 5A-C, SKN-1 activity is reduced to a level comparable to worms exposed to skn-1 RNAi (Figure 4E), suggesting that the presence of Ce-Duox1/BLI-3 is necessary for SKN-1 activity. We propose that the ROS produced by Ce-Duox1/BLI-3 in response to pathogens and the resulting oxidative stress activates SKN-1 activity.

To test if SKN-1 activity could be rescued in the bli-3 knock-down worms by providing ROS from an alternative source, we exposed the animals to paraquat. Paraquat generates ROS by redox cycling in vivo and has been used in previous studies as a trigger for SKN-1 activity in C. elegans[8]. L4 worms were exposed to either 1/30 bli-3 RNAi or vector control RNAi and were placed for 30 minutes in M9 solution with or without 100 mM of paraquat. As shown in Figure 5E, SKN-1 activity, as measured by fluorescence of the gst-4::gfp fusion, was activated in response to paraquat in worms fed vector control RNAi, in agreement with previous work [8]. Knock-down of bli-3 had no effect on activation of SKN-1 by paraquat (Figure 5G). These data indicate that SKN-1 activation induced by a chemical ROS generator does not require Ce-Duox1/BLI-3, unlike activation by pathogen exposure.

We were unable to perform these experiments on P. aeruginosa using the gfp-expressing transgenics because we discovered that the bli-3 knock-down caused a severe susceptibility phenotype on P. aeruginosa, much more severe than on E. faecalis, and more than half the worms were dead by the 24 hour time point (see below). Instead we used qRT-PCR to look at SKN-1 regulated genes on both E. faecalis and P. aeruginosa at an earlier time point (6 hours). By this methodology, we observed that both gst-4 and gcs-1 expression was significantly reduced in the bli-3 knock-downs compared to wild-type on both pathogens (Figure 5I and 5J). In conclusion, our results support a model in which ROS generated by Ce-Duox1/BLI-3 as a result of pathogen exposure trigger SKN-1 activity.

In previous investigations, loss of skn-1 did not influence susceptibility to P. aeruginosa[15], [19]. One study examined loss of skn-1 by RNAi [19]. However, in this work the animals were not exposed to RNAi until the L4 stage, at which point RNAi is likely less effective since the SKN-1 protein is produced in significant quantities during larval development [8]. We reduced skn-1 expression using RNAi, but began the exposure at the L1 stage and continued exposure through the L4 stage. Following this experimental procedure, the animals exhibited a statistically significant and reproducible susceptibility phenotype to both E. faecalis (Figure 6C, Table S2A in Text S1) and P. aeruginosa (Table S2B in Text S1). We additionally tested the possible role in susceptibility of some individual SKN-1 targets such as gst-4, gst-7 and gcs-1 by reducing the expression of these genes by RNAi. A significant difference compared to control RNAi was not observed, suggesting that these are not the critical SKN-1 targets or more than one gene contributes to SKN-1 pathogen resistance (data not shown).

Another study examined loss of skn-1 by mutation using loss of function alleles zu67 and zu135[15]. Using these same strains, we also examined susceptibility to P. aeruginosa and additionally E. faecalis. In a wild-type background, there was no significant difference in susceptibility, as previously reported; in fact zu135 was slightly more resistant (Figure 6A, 6B, Table S2 in Text S1). However, the zu67 and zu135 strains are sterile and do not produce viable embryos. This eliminates a major mechanism of killing, the internal hatching of the embryos during exposure to pathogen, a process called "bagging" [33], [34]. We found that RNAi of skn-1 did not cause a severe sterility phenotype until the second generation, likely due to its maternal effect [35], so the RNAi experiments described above were not affected by this problem. To render all the strains equally sterile so that they would be directly comparable, we exposed them to cdc-25.1 RNAi prior to pathogen exposure [36], [37]. Under these conditions, both skn-1 mutants were significantly more susceptible compared to wild-type (Figure 6A, 6B, Table S2 in Text S1). Note that the cdc-25.1 RNAi targets germline mitosis/meiosis, while a skn-1 mutation affects cell division within the embryo. Preventing development of the germline is known to increase resistance [34] and likely accounts for the increase in susceptibility observed when the already sterile skn-1 mutants are exposed to cdc-25.1 RNAi. Shivers et al. also attempted to control for sterility by adding FUDR to their assay plates, a compound that prevents cell division [15]. We speculate that this procedure may have affected the virulence of the pathogen, as noted in previous reports [36], [38]. Since the activation of skn-1 is very sensitive to the overall virulence of the pathogen (see Figure 2), this may have confounded the results.

To further test the role of SKN-1 on pathogen susceptibility, we used another approach. If loss of SKN-1 causes a susceptibility phenotype, constitutively active SKN-1 is predicted to increase resistance. We increased SKN-1 activity by two known means. First, we reduced expression of gsk-3 by RNAi, which causes constitutive nuclear localization of SKN-1 and therefore, greater transcriptional activity. By phosphorylation, GSK-3 normally inhibits nuclear localization of SKN-1 [39]. Secondly, we reduced expression of wdr-23, which encodes a WD40 repeat protein that targets SKN-1 to an ubiquitin ligase for proteasomal degradation. Loss of WDR-23 results in increased SKN-1 protein levels and greater output of its transcriptional program [40]. Reducing the expression of both of these genes significantly increased resistance to E. faecalis and P. aeruginosa (Figure 6C, Table S2 in Text S1). As a control, the expression of skn-1 was additionally reduced by RNAi, which abrogated the phenotypes, confirming that the increased resistance was dependent on skn-1. In previous work, loss of gsk-3 or wdr-23 both increased survival upon exposure to oxidative stress, however there was little (wdr-23) to no (gsk-3) concomitant increase in longevity when lifespan was assayed on non-pathogenic E. coli[39], [40]. Therefore, the significant increase in pathogen resistance observed in Figure 6A is not just a byproduct of causing a long-lived phenotype in general. To look at the effect of increased SKN-1 activity by another means, we examined two strains in which SKN-1 is over-produced because the strains carry extra copies of SKN-1::GFP. One strain carried wild-type SKN-1, fused to GFP (Figure 6D, Table S2 in Text S1), whereas the second expressed a mutant form of SKN-1, which is constitutively active (Figure 6E, Table S2 in Text S1) [8], [20]. Both strains exhibited increased resistance to both E. faecalis and P. aeruginosa. The effect was stronger utilizing the strain that produces constitutively active SKN-1 (Figure 6E, Table S2 in Text S1). Overall the data in this section demonstrates that that the level of SKN-1 activity significantly influences how long the worm survives on pathogen; less SKN-1 activity reduces resistance whereas more SKN-1 activity increases resistance.

In previous work, loss of bli-3 was shown to increase susceptibility to E. faecalis[3]. To determine if this effect could be completely dependent on skn-1, phenotypic analysis of the loss of both genes on susceptibility was performed. In a skn-1 background, loss of bli-3 by RNAi caused an increase in susceptibility to E. faecalis compared to skn-1 animals not exposed to bli-3 RNAi (P<0.0001). These data suggest that skn-1 is not completely epistatic to bli-3, ie not all of Ce-Duox1/BLI-3's protective effects are mediated through SKN-1 (Figure 7A, Table S2A in Text S1). However, on P. aeruginosa, we observed some differences (Figure 7B, Table S2B in Text S1). First, we discovered that a bli-3 knock-down caused a profound susceptibility phenotype – much more severe than that observed on E. faecalis in this work and in our previous publication [3]. Loss of skn-1 ameliorated this severe phenotype and showed the same level of susceptibility as a skn-1 mutant plus bli-3 RNAi (P = 0.3429). The data is consistent with skn-1 being epistatic to bli-3 on P. aeruginosa, despite the difference in phenotypic consequences compared to E. faecalis. In this case, loss of skn-1 protects against the severe susceptibility phenotype caused by the loss of bli-3 on P. aeruginosa, even though the skn-1 mutant is still more susceptible than the wild-type strain (P<0.0001). Understanding why there is a difference in the bli-3 phenotypes of animals exposed to these two different pathogens will require further investigation. Overall, the results are consistent with the model shown in Figure 8 that postulates that SKN-1 acts downstream of BLI-3. However, the experiments do not rule out the possibility that SKN-1 and BLI-3 have other roles independent of each other.

C. elegans strains were grown and maintained as previously described [51]. The following bacterial strains were used in this study: E. coli OP50 [52], E. faecalis OG1RF [53], P. aeruginosa PA14 [23]. C. elegans strains used in this study are indicated in Table S3 in Text S1.

In Figure 1A and 4G, RNA was extracted from L4 larvae exposed to E. faecalis OGIRF, P. aeruginosa PA14 and E. coli OP50 for 24 hours. In Figure 5I and 5J, RNAi treated eri-1(mg336) worms were exposed to E. faecalis OGIRF, P. aeruginosa PA14 and E. coli OP50 for 6 hours. The RNA was extracted using Trizol (Invitrogen) as indicated by the manufacturer. Samples were treated with DNase I to remove DNA contamination using the Turbo DNA free kit (Applied Biosystems) as described by the manufacturer. qRT-PCR was performed on an ABI 7500 instrument using the Power SYBR Green RNA-to-C_T 1 step kit (Applied Biosystems). Comparative C_T method was used to determine fold changes in gene expression normalized to act-1[20]. Primers used are listed in Table S4 in Text S1.

RNAi was induced by feeding L1 worms through L4 stage with bacteria producing dsRNA to target genes. RNAi expressing clones were obtained from the C. elegans library (Geneservices, UK) [54], [55]. All clones were verified by sequencing. Clones absent in the library were constructed as follows. Briefly, RNA was extracted from C. elegans L4 larvae using Trizol (Invitrogen) according to the manufacturer's protocol. cDNA was synthesized using SuperscriptII reverse transcriptase (Invitrogen) with oligodT and random hexamer primers. Gene specific primers were used to amplify regions of target genes, cloned into the vector pL4440 [56] and transformed into E. coli HT115(DE3). Constructs were verified by sequencing. Sequences of gene specific primers are listed in Table S5 in Text S1. To induce bli-3 knockdown, the bacterial strain expressing bli-3 RNAi was diluted in a 1∶30 ratio using the vector control. Double RNAi knockdowns were obtained by mixing bacteria expressing dsRNA to each target gene in a 1∶1 ratio. To render worms sterile prior to killing assays, larvae were exposed to cdc-25.1 RNAi.

To investigate the expression of gst-4::gfp, gcs-1::gfp and GST-7::GFP, worms were exposed to E. faecalis, P. aeruginosa and E. coli strains for 24 hours at 25°C and paralyzed with 1mM levamisole. Anesthetized worms were mounted on 2% agarose pads and imaged using an Olympus IX81 automated inverted microscope and Slidebook (version 5.0) software. The levels of GFP expression were scored as previously described [8], [12]. Briefly, little or no expression of GFP, expression of GFP in the anterior or posterior of the worm and expression throughout the intestine of the worm are categorically indicated by low, medium and high respectively for gst-4, gcs-1 and gst-7. To determine the effect of hydrogen peroxide and paraquat on gst-4::gfp expression, worms were exposed to 5mM hydrogen peroxide or 100mM paraquat in M9 for 20 and 30 minutes respectively, then transferred to seeded NG plate to recover for 4 hours before imaging. As controls, worms were exposed to the same period of time in M9. Higher background expression was observed in the control worms using the latter procedures.

SKN-1B/C::GFP expression was analyzed by fluorescence microscopy in worms exposed to E. faecalis for 24 hours and P. aeruginosa for 6 hours. Imaging was performed using the FITC, TRITC, DAPI and YFP filter sets to exclude the signal from autofluorescence in the worms. Percentages of worms indicating the degree of nuclear localization in the intestinal cells were scored as previously described [8], [12]. Briefly, no nuclear localization, localization of SKN-1B/C::GFP in the anterior or posterior of the worm and nuclear localization of SKN-1B/C::GFP in all intestinal cells are categorically indicated by low, medium and high, respectively. All fluorescence microscopy experiments shown were independently repeated at least three times.

Killing assays were generally performed as previously described [28], [30], [57]. Briefly, for E. faecalis killing assays, E. faecalis OG1RF grown in Brain Heart Infusion (BHI) medium for 5 hours was seeded on BHI plates and incubated at 37°C for 24 hours. While for P. aeruginosa killing assays, P. aeruginosa PA14 was cultured in Luria broth (LB), seeded on slow-killing plates and incubated first for 24 hours at 37°C and then for 24 hours at 25°C. A total of 90 L4 larvae were transferred to three replica plates. Worms were scored as live and dead at various points along the time course.

After scoring the fluorescent micrographs, statistical differences were determined by Chi square and Fisher's exact tests using GraphPad Prism version 5.0 (GraphPad Software, San Diego, CA). Each experimental condition was compared pairwise to the control condition. P-values of <0.05 were considered to be statistically significant. Statistically significant differences are indicated in the figures with asterisks next to the experimental condition. Kaplan-Meier log rank analysis was used to compare survival curves pairwise and to calculate the median survival. P-values <0.05 were considered to be statistically significant.