Expression pattern of PRDX-2::GFP and its variation upon H₂O₂ treatment

In budding yeast, the peroxiredoxin Tsa1 is massively induced upon H₂O₂ treatment [25, 26]. To gain insight into the tissue-specific expression of PRDX-2 upon oxidative stress, we first used the PRDX-2 reporter line generated by [29]. However, the level of expression of this strain varies a lot, and transgenics show many fluorescent aggregates ([29], S1 Fig), likely associated with transgene overexpression. Consistently, the strain displays a much stronger resistance to oxidative stress than wild-type animals (S1 Fig), suggesting that overexpressed PRDX-2 construct induces a higher H₂O₂ scavenging capacity in transgenics. Furthermore, it was impossible to identify PRDX-2-expressing cells in this strain, preventing the use of this strain in our study. For these reasons, we created a GFP knock-in line of PRDX-2, using the CRISPR-Cas9 technique [37]. A C-terminus GFP fusion targeting all PRDX-2 isoforms, comprising a linker, was engineered and inserted at the prdx-2 locus (Fig 1A). Three independent knock-in lines were obtained, sharing an identical expression pattern (Figs 1B and S1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRDX-2::GFP knock-in line expression pattern and its variation upon H₂O₂ treatment.

(A) Sketch depicting the PRDX-2::GFP knock-in strategy using CRISPR Cas9-mediated genome editing. The sgRNA target sequence (shown in red) was chosen upstream PRDX-2 STOP codon, to tag all prdx-2 isoforms. Black boxes indicate exons, white boxes untranslated regions. Bar, 100 bases pairs. The resulting last exon of the PRDX-2::GFP fusion protein is shown (right); note the insertion of a flexible linker between PRDX-2 and the GFP, to promote correct folding. After injection, three independent knock-in lines were recovered, sharing the same expression pattern. (B) Spinning-disc confocal projections of a representative PRDX-2::GFP knock-in animal, in 4 body regions (boxed in the worm sketch on top). PRDX-2::GFP expression is observed in the tip of the nose (e1, e3 epithelial cells), in pharyngeal muscle cells (pm), in body wall muscles (m), in the excretory pore cell (EPC), in proximal and distal gut cells (int1 and int9), and in several neuron pairs (I2, CAN, PHA/PHB). Note the different level of PRDX-2::GFP in I2 left and right neurons in panel B1. (C-G) An acute oxidative stress triggers an upregulation of PRDX-2 in the anterior gut, but not in neurons. (C-D) Spinning-disc confocal projections of control or H₂O₂-treated animals in the anterior gut (C) or in the head (D). (E-G) Quantification of fluorescence intensity in controls and in H₂O₂-treated animals in the int1 cell (E), in the EPC (F), and I2 neurons (G). Bars indicate mean and SD. ns, not significant, p>0.05; *p<0.05; ***p<0.001 (t test and Mann Whitney test). Scale bar, 20μm.

https://doi.org/10.1371/journal.pone.0274226.g001

In the PRDX-2::GFP knock-in line, we detected a broad expression of PRDX-2 in various cell types, including the proximal and distal gut, muscles (body wall, vulval and pharyngeal), epithelial cells (vulva, nose tip, hypodermis) and I2 neurons (Fig 1B), as previously reported [8, 27, 28]. In addition, we observed for the first time PRDX-2 expression in the excretory pore cell (EPC), and in two other pairs of neurons; the tail phasmids (PHA/PHB, whose identity was ascertained by DiI staing, S1 Fig) and the excretory canal-associated neurons (CANs), located close to the vulva (Fig 1B). Consistent with this, a high number of prdx-2 transcripts was detected in CANs and in the EPC [38]. Therefore, we hypothesized that the knock-in line faithfully reflects the endogenous PRDX-2 expression, and characterized it further.

PRDX-2::GFP is induced in the anterior gut upon H₂O₂ treatment, but not in I2 neurons

We noticed that many PRDX-2-expressing cells are directly in contact with the environment, such as the EPC, the tip of the nose, the vulva, and neurons, which all possess terminations exposed outside. This pattern is strikingly reminiscent of that detected in animals carrying the HyPer H₂O₂ biosensor after H₂O₂ exposure [39]. Thus, PRDX-2 is expressed in cells in which environmental H₂O₂ penetrates more easily, suggesting a protective role of PRDX-2 in these cells as a peroxidase. Therefore, in the following, we wondered whether all PRDX-2-expressing cells respond similarly to an H₂O₂-induced oxidative stress.

To determine whether PRDX-2 expression changes upon oxidative stress, we exposed animals to different doses of H₂O_2, and PRDX-2::GFP expression was quantified in different cells after spinning-disc confocal acquisitions. We selected H₂O₂ concentrations inducing different physiological responses [8]. A higher PRDX-2 expression was detected in the anterior gut two hours after a 10mM H₂O₂ treatment, but not after a 1mM H₂O₂ treatment (Fig 1C and 1E). The fluorescence was specific to PRDX-2::GFP as H₂O₂-treated controls did not show a higher gut autofluorescence (S2 Fig). Similarly, PRDX-2 expression was only induced in the EPC at 10mM H₂O₂ (Figs 1C and 1F and S2). Thus, our data indicate a dose-dependent PRDX-2 induction—as it occurs at 10mM but not at 1mM. The origin of this difference could arise from the animal behavioral response: at 1mM H₂O₂ the pharyngeal pumping is strongly inhibited to prevent ingestion [8], thereby exposure of gut to H₂O₂. This could explain both the absence of PRDX-2 induction in the gut we report, and the unchanged level of prdx-2 mRNA reported [28], after a 1mM H₂O₂ treatment. In contrast, at 10mM H₂O₂, the nematode should retract back (avoidance response, [8]); but here, as trapped animals cannot escape, they likely swallow a certain amount of H₂O₂, exposing the foregut to a severe oxidative stress which could trigger PRDX-2 induction. A 10mM H₂O₂ treatment of 30min has been shown to induce hyperoxidation of over 50% total PRDX-2 in wild-type lysed worms [40] and this form persists after a 4h recovery period [27]. Although the induction we observed is only based on expression level and not on protein activity, we suggest that PRDX-2 could still scavenge H₂O₂ in the EPC and in the foregut, consistent with the reported protective role of intestinal PRDX-2 [27].

Among PRDX-2-expressing neurons, the I2 pair shows the highest expression and possesses terminations exposed to the outside [9]. Therefore, we focused on I2 neurons to investigate whether the neuronal level of PRDX-2 is affected upon H₂O₂ treatment. However, at both concentrations tested (1 and 10mM), the level of PRDX-2 in I2 neurons constantly remained unchanged after H₂O₂ treatment (Figs 1D and 1G and S2).

Thus, we observed an upregulation of PRDX-2::GFP in the anterior gut and in the EPC following an acute oxidative stress, but not in neurons. As Prxs belong to a homeostatic system, an upregulation was shown to be associated with a detoxification function in yeast [26]. By analogy, PRDX-2 might fulfill a peroxidase function in the foregut and in the EPC to protect the animal against environmental aggressions. In contrast, in I2 neurons, PRDX-2 would instead act in H₂O₂ sensing and/or signaling, as proposed [8, 28]. We conclude that the responses of PRDX-2 to oxidative stress are cellular context-dependent.

SKN-1 controls expression of PRDX-2 in the gut, but not in neurons

This prompted us to test whether PRDX-2 expression could rely on different subcellular regulations. By analogy with yeast, we wondered whether PRDX-2 would be controlled by the Yap1 nematode orthologue SKN-1 [5] in cells where PRDX-2 is induced upon oxidative stress, and asked whether such regulation occurs in other cells. We sought to test this hypothesis by inactivating the function of SKN-1 in the PRDX-2::GFP knock-in line. We first used the skn-1(zj15) allele, which specifically inactivates gut-specific isoforms, leaving neuronal isoforms unaffected [41]. In this mutant, the basal expression level of PRDX-2 is reduced in the foregut, but an induction persists after a 10mM H₂O₂ treatment (Fig 2C), suggesting that the neuronal isoforms could mediate this effect. Consistent with this, the RNAi-mediated knock-down of all skn-1 isoforms also triggered a reduction of PRDX-2::GFP signal in the anterior gut, and an absence of induction upon H₂O₂ treatment (Fig 2A, 2B and 2D). This result indicates that SKN-1 activity is essential to regulate PRDX-2 expression in the anterior gut, both at the basal level and under oxidative stress. In agreement with this, SKN-1 was found to bind to the prdx-2 promoter in chromatin immunoprecipitation experiments coupled with high-throughput DNA sequencing (ChIP-seq) [42] (S3 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. SKN-1 function is required for PRDX-2 expression in the gut.

(A-B) Spinning-disc confocal projections of head and anterior gut of PRDX-2::GFP knock-in animals, in control RNAi (A) and in skn-1(RNAi) (B) animals. The right panel shows the foregut of a 10mM H₂O₂-treated animal in both genotypes. Note the low level of int1 PRDX-2::GFP fluorescence in skn-1(RNAi) animal. (C-D) Quantification of the int1 cell fluorescence intensity in control and after a 10mM H₂O₂-treatment, in skn-1(zj15) mutants (C) and in skn-1(RNAi) animals (D). Bars indicate mean and SD. ns, not significant, p>0.05; **p<0.01; ***p<0.001; ***p<0.0001 (t test, ANOVA). Scale bar, 20μm.

https://doi.org/10.1371/journal.pone.0274226.g002

In contrast, in I2 neurons, no change in the PRDX-2 level was detected at the basal level or after H₂O₂ treatment compared to controls, neither in skn-1(zj15) mutants nor in skn-1(RNAi) animals (S2 Fig). With the limitation that RNAi efficiency may be lower in neurons [43], we suggest that additional transcription factor(s) might regulate prdx-2 expression in these neurons. Consistently, several transcription factors were reported to bind to the prdx-2 promoter by ChIP-seq [42] (S3 Fig).

We conclude that skn-1 accomplishes a cell autonomous regulation of prdx-2 in the intestine, but that this regulation may not occur in neurons. Taken together, our data suggest that PRDX-2 might exert a different function depending on the cell type: H₂O₂ detoxification in the foregut, triggered by SKN-1, vs. H₂O₂ perception or signaling in I2 neurons.

PHA neurons respond to H₂O₂ in a prdx-2-dependent manner

As PRDX-2 is expressed in other neurons than I2s, we asked whether PRDX-2 endows these neurons with oxidative stress sensing properties. We focused on PHA tail neurons, as they belong to the phasmid sensory sensilla and respond to many noxious stimuli [12]. We monitored PHA neuron activation by imaging calcium using the GCaMP3 fluorescent sensor [44] expressed under the flp-15 promoter, specific to I2 and PHA neurons [45].

To image calcium fluxes in neurons, L4 animals were trapped in microfluidic chambers and exposed to H₂O₂ (Figs 3A and S4), and their response was followed in 4D spinning-disc ultrafast acquisitions. This experimental setup, combined with semi-automated image analyses to quantify the mean fluorescence in neurons over time (Supplementary Information), confirmed the activation of I2 neurons upon a 1mM H₂O₂ treatment (Fig 3B and 3D and S1 Movie and S5 Fig), similar to that observed upon exposure to H₂O₂ vapor (ie. 8.82M, [8]). As PRDX-2 reproducibly shows an asymmetric expression level in I2L and I2R (Fig 1B and 1D and detailed description in [46]), each neuron was scored individually to take into account a putative left-right effect. However, this did not impact neuron response as no significant difference in the normalized response of left and right neurons was noticed (Fig 3B and 3L). Importantly, we observed that PHA neurons respond to 1mM H₂O₂ comparably to I2 neurons (Fig 3C, 3D and 3H and S2 Movie and S5 Fig). We then investigated whether PRDX-2 function was necessary for PHA response, using a strong prdx-2 loss of function mutant. In agreement with previous work [8], I2 neurons in prdx-2(gk169) mutants failed to respond to 1mM H₂O₂ (Fig 3E, 3F and 3H and S3 Movie and S5 and S6 Figs). Interestingly, we uncovered that PHA neuron response was severely impaired in prdx-2 mutants, although not completely abolished as in I2s (Fig 3E, 3G and 3H and S4 Movie and S5 and S6 Figs). From these observations, we conclude that both I2s and PHAs responses to oxidative stress require the function of the peroxiredoxin PRDX-2, but the residual response of PHA neurons suggests a lesser requirement of the antioxidant in tail neurons.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. I2 and PHA neurons both respond to 1mM H₂O₂ in a prdx-2-dependent manner, but only PHA neurons respond to 10μM H₂O₂.

(A) Low-magnification DIC picture of L4 animals trapped in the microfluidic device used in all neuron recordings experiments. (B,C,F,G,J,K) Calcium response (visualized by GCaMP3) in I2R/I2L (B,F,J) and in PHAL/PHAR neurons (C,G,K) following H₂O₂ treatment (indicated by the red bar). Average curves show normalized neuron responses (time in seconds), in both left and right neurons (top and bottom curves). N, number of movies quantified for each genotype, in wild-type (B,C,J,K) and in prdx-2 mutants (F,G). (D,E,I) Bar graph showing the fraction of animals responding to the H₂O₂ stimulation in all experiments, classified as high (green), moderate (yellow) or absent (red) responses (see Methods). N, number of movies analyzed. (H,L) Quantification of the calcium response to H₂O₂ in I2 and PHA neurons in controls and in prdx-2 mutants at 1mM (H), and in I2L/R and PHAL/R controls, at 1mM and 10μM H₂O₂ (L). Bars represent the mean (N, number of movies) and error bars SD. ns, not significant, p>0.05; *p<0.05; ***p<0.001; ****p<0.0001 (Mann Whitney and Kruskal-Wallis tests). See corresponding S1–S6 Movies and S5 Fig.

https://doi.org/10.1371/journal.pone.0274226.g003

I2s and PHAs show differences in H₂O₂ sensitivity and in receptors involved

Given the putative role of PRDX-2 as an H₂O₂ sensor, the fact that there was a slight difference in PRDX-2 activity requirement in I2s and PHAs prompted us to analyze whether the head and tail neurons share the same sensitivity to H₂O₂. We thus tested whether I2 and PHA neurons exhibit a response to the very mild dose of 10μM H₂O₂, a dose which induces a less penetrant pharyngeal pumping inhibition (≈ 35% of animals) than that observed at 1mM (≈ 90% of animals, [8]). In our experiments, whereas I2 neurons failed to be activated in most animals at 10μM H₂O₂ (25/29, Fig 3I, 3J and 3L and S5 Movie and S5 Fig), PHA neurons responded in the vast majority of animals (25/27, Fig 3I), in a similar manner than at 1mM H₂O₂ (Fig 3K and 3L and S6 Movie and S5 Fig). In addition, PHA neurons response to 10μM H₂O₂ depends on PRDX-2, as prdx-2 mutants PHA neurons all fail to respond to this low dose (13/13, S7 Movie and S5 and S6 Figs). The requirement of PRDX-2 for PHA response to micromolar H₂O₂ suggests that PRDX-2 is unlikely to transmit the H₂O₂ signal under its hyperoxidized form. Taken together, we conclude that PHA neurons are more sensitive to low doses of H₂O₂ than I2 neurons.

The difference in sensitivity between I2 and PHA neurons may come from distinct molecular mechanisms. To explore this possibility, we first tested which receptors are required in I2s and in PHAs for H₂O₂ perception. We focused on photoreceptors as photosensation is likely to involve the generation of ROS [8, 47]. In C. elegans, light sensing relies on unusual gustatory G-protein-coupled receptors (GPCRs) related to vertebrate photoreceptors: the two nematode closest paralogs LITE-1 and GUR-3 mediate photosensation in ASJ and ASH neurons [47–49], and light and H₂O₂ sensing in I2 neurons [8]. We investigated whether these two receptors are differentially localized in I2 and PHA neurons. We generated a knock-in GUR-3::GFP line, which revealed that GUR-3 is solely expressed in I2 and I4 photosensory neurons (Fig 4A), as previously reported using episomal expression [8]. Such a restricted pattern deeply contrasts with the broad expression domain of LITE-1, which includes phasmid neurons (PHA and PHB), but not I2 neurons, as shown using both translational and transcriptional reporters [8].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. H₂O₂ response involves different receptors and transducers in I2 and PHA neurons.

(A) Spinning-disc confocal projections of a representative GUR-3::GFP knock-in animal, at L1 (left) and L4 stages (right). The GUR-3::GFP signal is detected in I2 neurons and in a single I4 neurons (head panel), but not in PHA/PHB neurons (tail panel). Bar, 10μm. (B,D) Bar graph of the fraction of lite-1, gur-3, pmk-1 and sek-1 mutants responding to the H₂O₂ stimulation, classified as high (green), moderate (yellow) or absent (red) responses. N, number of movies analyzed. (C,E,F) Quantification of the calcium response to 1mM H₂O₂ in I2 and PHA neurons in controls and in gur-3 and lite-1 mutants (C), and in pmk-1 and sek-1 mutants at 1mM (E) and at 10μM H₂O₂ (F). Bars represent the mean (number of movies indicated) and error bars SD; ns, not significant, p>0.05; *p<0.05; ****p<0.0001 (ANOVA and Kruskal-Wallis tests). See corresponding S8–S19 Movies and S7–S10 Figs. (G,H) Blue light triggered different calcium fluxes in I2 and PHA neurons in the three regions analyzed; anterior neurite (green), posterior neurite (red) and soma (blue). The curves represent the mean and shaded error bars indicate SD. See S20–S22 Movies.

https://doi.org/10.1371/journal.pone.0274226.g004

These differential localizations prompted us to inquire whether mutants in these receptors were still able to trigger a response to 1mM H₂O₂ in I2 and PHA neurons. In gur-3(ok2245) mutants, only PHA neurons were able to respond to 1mM H₂O₂ (Fig 4B and 4C and S8 and S9 Movies and S7 and S8 Figs), providing evidence that GUR-3 function is not essential in PHA neurons for H₂O₂ sensing. In contrast, lite-1(ce314) mutants showed a reciprocal response, with only I2 neurons responding to 1mM H₂O₂ (Fig 4B and 4C and S10 and S11 Movies and S7 and S8 Figs). In conclusion, H₂O₂ likely activates I2 neurons via gur-3 and PHA neurons via its paralog lite-1, respectively. Interestingly, these observations might explain the previously reported impaired avoidance response of lite-1 mutants to 1mM H₂O₂ [8], as PHA neurons are involved in escape behavior [13].

PMK-1 and SEK-1 are required for PHA neurons response to micromolar H₂O₂ but are dispensable in I2 neurons

Since H₂O₂ perception in I2 and PHA neurons involves different receptors and requires the function of PRDX-2 in both cases, we wondered what type of signaling occurs downstream PRDX-2 to trigger neuronal activation upon H₂O₂ stimulation.

As mentioned above, studies in various models have reported that peroxiredoxins can modulate the p38/MAPK signaling pathway to influence cellular decisions, notably in drosophila and mammalian cells [34]. In C. elegans, specifically, the activation of this PRDX-2-PMK-1/p38MAPK cascade allows micromolar doses of H₂O₂ to potentiate the ASH neuron sensory behavior to glycerol [3]. Therefore, we investigated whether the p38/MAPK pathway could be involved in H₂O₂ sensing by analyzing the responses of pmk-1 (p38/MAPK) and sek-1 (MAPKK) mutants in I2 and PHA neurons. As the strongest allele pmk-1(ok811) is homozygous lethal, we had to use the hypomorphic pmk-1(km25) viable mutant, which carries a N-terminal deletion [50]. In pmk-1(km25) mutants, while I2 neurons responded normally to 1mM H₂O₂, PHA neurons showed a slightly milder response to 1mM H₂O₂ compared to controls (Fig 4D and 4E and S12 and S13 Movies and S9 Fig). In sek-1(km4) MAPKK mutants, we observed comparable responses of I2 and PHA neurons to 1mM H₂O₂ to that of controls (Fig 4D and 4E and S16 and S17 Movies and S10 Fig). We conclude that for 1mM H₂O₂ sensing, the p38/MAPK pathway is dispensable at least in I2 neurons, but may play some role in PHA neurons.

In ASH neurons, the PRDX-2-mediated activation of the p38/PMK-1 cascade induces a potentiation of their sensory behavior [3]. Therefore, we asked whether the p38/PMK-1 pathway could similarly promote PHA higher sensitivity to H₂O₂. We examined PHA neurons response to 10μM H₂O₂ in pmk-1 (MAPK) and in sek-1 (MAPKK) mutants. Strikingly, both mutants displayed a very similar phenotype, with an abolished response of PHA neurons to micromolar doses of H₂O₂ observed in pmk-1 (Fig 4D and 4F and S15 Movie and S9 Fig) and in sek-1 mutants (Fig 4D and 4F and S19 Movie and S10 Fig). In I2 neurons, as in controls, both pmk-1 and sek-1 failed to respond to 10μM H₂O₂ (Fig 4D and 4F and S14 and S18 Movies and S9 and S10 Figs), as expected. Taken together, this suggests that the p38/MAPK pathway would be specifically required for PHA neurons hypersensitivity to H₂O₂, but dispensable in I2 neurons.

PHA neurons are photosensory neurons like ASH neurons

Light sensing has been reported for ASJ, ASH and I2 neurons and require either the LITE-1 or the GUR-3 receptor, respectively [8, 47–49]. As PHAs neurons require LITE-1 to respond to H₂O₂ (Fig 4B and 4C and S11 Movie and S7 Fig), we asked whether they could respond to light. To test this, we monitored calcium transients in three neuronal compartments using the GCaMP strain upon stimulating neurons with blue light, as previously done for I2 or ASH neurons [8, 47]. Interestingly, we found that all regions of PHA neurons responded to light, displaying a different response profile than those of I2 neurons (S22 Movie): PHA soma showed a stronger and longer response than I2 soma; PHA posterior neurite responded much slower that in I2 where it exhibits the fastest and strongest response peak, and the anterior neurite also had a slower recovery than in I2 (Fig 4G and 4H and S20 and S21 Movies). Overall, while I2 neurons exhibit a fast photoresponse within 10-15s, PHA neurons photoresponse requires twice longer to return to steady state (approx. 30s). Strikingly, we noticed that PHA neurons profile is highly reminiscent of that reported in ASH neurons photoresponse [47].

It has been proposed that light sensing may involve intracellular H₂O₂ release and peroxiredoxin signaling in both nematodes and yeast [7, 8]. To test whether PHA response to light requires the peroxiredoxin PRDX-2, we analyzed light sensing in prdx-2 mutants. As in H₂O₂ sensing (Fig 3E, 3F and 3H and S3 Movie), we observed that I2 response to light strictly depends on prdx-2 (15/15, S23 Movie and S11 Fig). Surprisingly, we found that PHA photoresponse does not require PRDX-2, as prdx-2 mutants were still able to respond to light (17/20, S24 Movie and S11 Fig). We conclude that light sensing does not involve the same mechanisms in I2 and in PHA neurons, downstream the photoreceptors.

H₂O₂ sensing in head and tail neurons relies on different mechanisms

Here, we describe how two pairs of sensory neurons located in the head and the tail of C. elegans, namely I2s and PHAs, contribute to exogenous H₂O₂ and light sensing. Compared to previous reports, our study relies on a PRDX-2::GFP knock-in line more closely reflecting endogenous expression level, in comparison to overexpression often observed with extrachromosomal arrays. While classical methods do not enable precise control of the environment such as application of a stress, we carried out neuron response experiments using the microfluidic technology, allowing live imaging of immobilized animals upon simultaneous exposure to a controlled oxidative stress.

We found that PHA tail neurons can elicit a response to a micromolar range of H₂O₂, whereas I2 head neurons cannot, suggesting that distinct molecular mechanisms may account for this difference. Accordingly, while the peroxiredoxin PRDX-2 is essential for H₂O₂ sensing in both I2 and PHA neurons, a different transmembrane receptor is required to transduce the signal: I2 neurons use the gustatory receptor GUR-3, while PHA neurons would require its paralogue LITE-1. Another difference lies in the p38/MAPK activity requirement: while dispensable in I2 neurons, it would be specifically required in PHA neurons to confer their hypersensitivity to H₂O₂ (lack of response of PHA neurons to 10μM H₂O₂ in sek-1/MAPKK and in pmk-1/MAPK mutants, Fig 4). Finally, although nematodes were known for a long time to avoid light when the light pulse was applied on their tail [48], our work provide the first evidence, to our knowledge, that PHA tail neurons act as photoreceptor cells.

Overall, our data are consistent with previous findings unveiling the existence of two distinct modes of response to oxidative stress in C. elegans: a direct response in peripheral tissues such as the gut, and a neuronally-regulated response relying on synaptic transmission [51]. Specifically, we uncovered that a harsh oxidative stress (10mM H₂O₂) triggers PRDX-2 induction in the anterior gut and in the EPC (Figs 1 and 2), while lower doses of H₂O₂ triggers either I2 and PHA neurons activation (1mM), or only PHA neurons activation (10μM) (Fig 3). Interestingly, both types of response involve the peroxiredoxin PRDX-2, which behaves differently in the two cellular contexts: PRDX-2 would be cell-autonomously induced by SKN-1 in the intestine, but likely not in neurons (Fig 2). Therefore, we propose that PRDX-2 might act as a peroxidase in the gut, as proposed [27], whereas it could function as a H₂O₂-signaling molecule in neurons, as suggested for I2 neurons [8]. Importantly, H₂O₂ response in I2 and PHA neurons requires the joint function of PRDX-2 and a receptor, as each mutant individually cannot respond (Figs 3 and 4). In addition, former rescue experiments indicated that light response requires the activity of PRDX-2 and GUR-3 specifically in I2 neurons [8]. Based on all these data and recent studies shedding light on how H₂O₂ is sensed in plant and animal cells [34, 52], we propose new hypotheses regarding H₂O₂ signaling in neurons which are depicted in Fig 5 and described below, to integrate our findings with recent results from the literature.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Hypothetical model of H₂O₂ sensing and signaling in C. elegans I2 and PHA neurons based on this and on previous work.

(A-B) Hypothetical model of H₂O₂-induced neuronal activation, based on our data and on previous studies. Schematic drawing of a presynaptic button in I2 (A) and PHA (B) neurons, illustrating the presumptive H₂O₂-PRDX-2-mediated neuronal activation in both cases (modelled in S12 Fig). High doses of H₂O₂ (1mM) are sensed by both neurons, but only PHA neurons respond to 10μM H₂O₂, as illustrated in red. In this model, H₂O₂ would freely diffuse through the neuron plasma membrane and oxidize PRDX-2, presumably leading to LITE-1 or GUR-3 activation. Receptor activation is likely relayed by cGMP signaling, resulting in the opening of voltage-gated calcium channels (in grey) and neurotransmitter release (glutamate in both cases), triggering an adapted response. In PHA neurons, the PMK-1/p38MAPK pathway is additionally required to promote neuronal response to micromolar doses of H₂O₂, potentially through OSM-9 phosphorylation, as observed in ASH neurons [3]. See Discussion for further details and bibliographic references.

https://doi.org/10.1371/journal.pone.0274226.g005

A presumptive model of H₂O₂ sensing in C. elegans neurons

In both I2 and PHA neurons, we favor the hypothesis that cytosolic PRDX-2 rather than the transmembrane receptor would be the neuronal H₂O₂ sensor, based on the following observations: i) in many cases, H₂O₂ signaling is mediated by oxidation of cysteines in redox-regulated proteins [53]. Alternatively, redox signaling often relies on peroxiredoxins acting as sensor and transducer of H₂O₂ signal [54], as thiol modifications would be much faster when catalyzed by peroxiredoxins [55–57], due to their abundance and their high reactivity to H₂O₂ [58]. Here, the striking abundance of PRDX-2 in PHA, and especially in I2 neurons reinforces this idea, as well as the defective response of prdx-2 mutants to both doses of H₂O₂ tested (Figs 3 and S6). Consequently, we propose that H₂O₂ signaling in I2 and PHA neurons would involve redox signaling through PRDX-2. ii) Concerning receptor topology, GUR-3 and LITE-1 present a higher conservation in their intracellular domains, with conserved cysteines only found intracellularly and in a transmembrane domain (S12A Fig). This structure is more reminiscent of that of the vertebrate transmembrane protein GDE2, which is activated intracellularly by the Prdx1 peroxiredoxin [59], than of the plant H₂O₂ sensor, the HCPA1 receptor, whose activation involves direct oxidation of extracellular cysteines [52]. Here, GUR-3 and LITE-1 receptor topology does not support the hypothesis of a direct oxidation by H₂O₂ on the extracellular domain, but rather suggests intracellular signaling as in the case of GDE2 (S12B and S12C Fig). We thus propose a scenario in which H₂O₂ would diffuse through the neuron plasma membrane and oxidize PRDX-2. Oxidized PRDX-2 or its disulfide form (PRDX-2^ox/S-S) would in turn react with the cysteines of GUR-3 or LITE-1, triggering receptor activation, and I2 or PHA neuron response (Figs 5 and S12). These hypotheses will require future experimental validation.

Neuronal response involves the opening of calcium channels, which are different in I2 and PHA neurons: I2 activation depends on UNC-2 and UNC-36 voltage-gated calcium channels [9], while PHA neurons require the cyclic nucleotide-gated channel TAX-4 and the vertebrate TRPV5 nematode equivalent OSM-9 [12]. Of note, the other TRPV5 channel subunit OCR-2 may also act in PHAs, as it is expressed in PHAs (S13 Fig, [60]) and it functions together with OSM-9 [61]. In conclusion, although closely related and both able to sense H₂O₂ and light, GUR-3 and LITE-1 receptor signaling likely involves different downstream transducers. Except for PRDX-2 requirement, I2 and PHA neurons would use distinct molecular pathways to transduce H₂O₂ response.

A striking parallel between PHA and ASH neurons

Our data uncover a high sensitivity of PHA neurons to micromolar doses of H₂O₂ that is not seen in I2 neurons and requires PRDX-2 and p38/MAPK activity. Strikingly, micromolar doses of H₂O₂ were also reported to activate the p38/MAPK pathway via PRDX-2 in ASH neurons, leading to the phosphorylation of the OSM-9 TRPV sensory channel, thereby increasing its sensitivity [3]. Although our study and the latter do not elucidate how PRDX-2 triggers PMK-1 activation, recent evidence sheds light on this process: in both mammalian and drosophila cells, H₂O₂ induces transient disulfide-linked conjugates between the MAP3K and a typical 2-Cys peroxiredoxin [34]. Similarly, in C. elegans, PRDX-2 could activate the MAPKKK NSY-1, as NSY-1 function is required in ASH neuron [3], and may be expressed in PHAs (unidentified cells of the tail, D. Moerman, WormBase). Based on these observations and on the fact that OSM-9 is expressed [62] and likely required in PHA neurons [12], it is possible that PMK-1/p38MAPK activation might increase PHA neurons’ sensitivity to H₂O₂ through the downstream phosphorylation of OSM-9, via AKT-1, as in ASH neurons (Fig 5). Alternatively, by analogy with recent work showing that DLK/p38MAPK signaling controls LITE-1 stability in ASH neurons, PMK-1 could also control LITE-1 turnover by RAB-5-mediated endocytosis [63]. These assumptions, which will necessitate further experimental testing, are supported by the observation that ASHL/R and PHAL/R are both descendants of ABplp or ABprp in the nematode cell lineage, and hence may express similar sets of genes following lineage-specific priming [64].

The fact that PHA, PHB and ASH were found in the same neuron cluster [65] confirms that they share similar molecular signatures. To illustrate the parallel between ASH and PHA polymodal nociceptors, we retrieved the set of genes expressed in these neurons [65], and analysed which ones were specifically enriched in these neurons in comparison to all other neurons (S13 Fig, data in S1–S3 Tables). This analysis revealed that ASH and PHA not only express many genes specific to ciliated neurons, as expected (eg. che-3, nphp-4, che-11, ifta-1, C33A12.4 and R102.2, [66]), but also a number of common receptors (eg. ocr-2, osm-9, ida-1, casy-1, ptp-3 or pdfr-1), although some of them are more enriched in ASH (dop-2, nrx-1, snt-5, sue-1). Intriguingly, their neuropeptide profile appears different for the two neuron pairs (PHA enriched in flp-7, flp-4, flp-16, nlp-1 ins-18, while ASH strongly express flp-13 and npr genes), suggesting that beyond their functional similarity, ASH and PHA may trigger different types of intercellular communication. Finally, this analysis highlights the fact that several genes with unknown function are highly enriched in both ASH and PHA (eg. F27C1.11, W05F2.7, tos-1, cab-1), pointing to their potential role in neuronal function.

Taken together, our analyses highlight the common features shared between PHA and ASH polymodal nociceptive neurons, as formerly noticed [13], since they both: i) display a higher sensitivity dependent on the p38/MAPK pathway (Fig 4D–4F and [3]), ii) exhibit a similar response profile to light (Fig 4H and [47]), iii) require the photoreceptor LITE-1 for lightsensing [47] or H₂O₂ sensing (Fig 4B and 4C), iv) trigger avoidance [13], and v) share a close molecular signature (S13 Fig and [65]). A recent report indicates that ASJ neurons are also required for H₂O₂ avoidance [6], and unravels the behavioral mechanisms allowing C. elegans to find a suitable niche, owing to the interplay between H₂O₂ and bacteria in its environment. Overall, these observations led us to propose that nematodes might integrate the environmental redox signals from at least four different pairs of neurons (I2, ASH, PHA and ASJ) in order to trigger an appropriate dose-dependent physiological response.

GUR-3 and LITE-1 receptors mediate both light and H₂O₂ sensing

In yeast, light sensing relies on the peroxisomal oxidase Pox1 which triggers light-dependent H₂O₂ formation, the latter being sensed by the Tsa1 peroxiredoxin and transduced to thioredoxin for subsequent signaling [7]. Nematodes, unlike yeast, require a photoreceptor in addition to the antioxidant for lightsensation: GUR-3 in I2 neurons [8], and LITE-1 in ASJ and ASH neurons [47, 49]. Here we showed that both I2 and PHA respond to light, but with a different profile. Despite its unusual membrane topology, LITE-1 has been shown to encode a bona fide photoreceptor, whose photoabsorption depends on its conformation [67]. However, whether light directly activate the neuron photoreceptor or triggers intracellular H₂O₂ release and signaling is still unclear. In agreement with the fact that H₂O₂ inhibits LITE-1 photoabsorption in vitro [67], it has been shown that a H₂O₂ pretreatment reduces LITE-1-mediated photoresponse in ASH neurons [47]. Consistent with these reports, our data illustrate that GUR-3 and LITE-1 have a dual function in both light and H₂O₂ sensing. In I2 neurons, as in yeast, redox signaling could be involved in transducing the light signal, as PRDX-2 is strictly required for light sensing (S11 Fig, [8]). In contrast, PHA neurons can respond to light without PRDX-2, indicating that the LITE-1 photoreceptor tranduces the light signal without a redox relay. This difference could explain our observation that I2 neurons respond faster to light than PHA and ASH neurons (Fig 4G and 4H and [47]). Whether this difference strictly depends on the photoreceptor (LITE-1 in PHA, ASH vs GUR-3 in I2) and/or on its downstream signaling cascade is an open question.

Finally, it is noteworthy that I2 and PHA neurons rely on peroxiredoxin for H₂O₂ and/or light sensing, while ASJ and ASH rely on thioredoxin: TRX-1 is required for LITE-1-dependent photosensation in ASH neurons [47], and is expressed in ASJ photosensory neurons [68], which respond to H₂O₂ [6]. Altogether, this further underlines the importance of a redox signaling relay involving antioxidants of the peroxiredoxic cycle in nematode photosensory neurons.

In conclusion, our work illustrates that nematodes can sense various concentrations of H₂O₂ through sensory neurons located in the head and the tail, using either partially different (I2 and PHA), or similar molecular mechanisms (ASH and PHA). This set of neurons also confer light sensing to the nematode with a distinct speed in the response (fast in I2, slow in PHA and ASH, [47]). While I2 neurons seem more specialized in sensing oxidative stress [8, 10], PHA and ASH can detect many other stimuli [12], but all of them can trigger avoidance. To tackle the complex question of how these neuronal inputs translates into behavior, a recent method was developed [69], allowing the simultaneous recording of behavioral and neural responses of C. elegans to salt concentrations changes. These observations may help uncover whether and how inputs from head and tail oxidative stress sensory neurons integrate to allow nematodes to quickly and appropriately react to a change in the environment.

Generation of plasmids and transgenic strains by CRISPR/Cas9-mediated genome editing

C. elegans strains (listed in Supplementary Information) were maintained as described [70]. PRDX-2::GFP knock-in strain was generated by CRISPR/Cas9-mediated genome editing, using a DNA plasmid-based repair template strategy [37]. For both PRDX-2 and GUR-3 knock-ins, a C-terminal GFP fusion was generated, comprising a flexible linker between the coding region and GFP to allow correct folding of the fusion protein. A combined small guide-RNA/repair template plasmid was built using the SAP Trap strategy [71]. Phusion DNA polymerase was used to amplify by PCR 5’ and 3’ prdx-2 and gur-3 homology arms (HAs) from N2 genomic DNA, using primers containing SapI restriction sites and silent mutations to prevent Cas9 re-cleavage (all primers sequences listed in Supplementary Information). After purification, 5’ and 3’ HAs and sgRNA oligonucleotides were assembled into the destination vector (pMLS256), together with the flexible linker (from pMLS287), GFP and the unc-119 rescuing element (from pMLS252), in a single SapI restriction-ligation reaction, as described [71]. Prior to transformation in DH5α cells, a sabotage restriction was performed with SpeI to digest empty destination vectors but not the desired assembly constructs, which were subsequentially verified by restriction digest analysis and sequencing. All plasmids used for injection were purified using a DNA Miniprep Kit (PureLink, Invitrogen), or a DNA midiprep kit (Macherey Nagel). For PRDX-2::GFP knock-in, a plasmid mix containing combined sgRNA/repair template plasmid (50 ng/μl), Cas9-encoding pSJ858 (25ng/μl) and co-injection markers (pCFJ90 at 2.5ng/μl; pCFJ104 at 5ng/μl, and pGH8 at 5ng/μl) was injected in the germline of unc-119(ed3) animals [37]. For GUR-3::GFP knock-in, an injection mix containing purified Cas9 protein (IDT) associated with tracrRNA and crRNA (guide RNA), GUR-3 repair template, and co-injection markers was injected in unc-119(ed3) animals, according to IDT online protocols for C. elegans. Plates containing 2/3 injected F0 animals were starved, chunked on fresh plates, for candidates screening (attested by the presence of wild-type non fluorescent animals). Knock-in events were validated by PCR on homozygous lysed worms (QuantaBio AccuStart II GelTrack PCR SuperMix), using primers annealing in the inserted sequence and in an adjacent genomic region not included in the repair template. The PRDX-2::GFP strain was outcrossed 5 times to N2 wild-types.

RNA interference

RNAi experiments were performed by feeding using the Ahringer-MRC feeding library [72]. Animals fed with the empty vector L4440 served as a negative control. The efficiency of each RNAi experiment was assessed by adding an internal positive control, zyg-9(RNAi), which induces embryonic lethality.

Spinning-disk confocal microscopy acquisitions and fluorescence intensity measurements

For live imaging, animals were anesthetized in M9 containing 1mM levamisole and mounted between slide and coverslip on 3% agarose pads. Synchronized L4 animals were treated for 30min in a 96-well flat bottom plate, in 50μl of M9 containing 1mM or 10mM H₂O₂. Treated animals were transferred using a siliconized tip on a freshly seeded plate to recover, and imaged 1h30 to 2h later. Spinning-disk confocal imaging was performed on a system composed of an inverted DMI8 Leica microscope, a Yokogawa CSUW1 head, an Orca Flash 4.0 camera (2048*2018 pixels) piloted by the Metamorph software. Objective used were oil-immersion 40X (HC PL APO, NA 1.3) or 63X (HCX PL APO Lambda blue, NA 1.4). The temperature of the microscopy room was maintained at 20˚C for all experiments. Z-stacks of various body regions were acquired with a constant exposure time and a constant laser power in all experiments. Maximum intensity projections were used to generate the images shown. Fluorescence intensity measurements in int1, I2 and EPC cells were performed using the Fiji software, by manually drawing a region of interest (ROI) around the cell (int1, EPC), or applying a threshold (I2 neurons), background was subtracted and average pixel intensity was quantified.

Microfabrication and microfluidic chip preparation

The microfluidic chip original design was inspired by the Wormspa [73], but pillars distances were adapted to trap L4 animals, and multiple series of traps were included to increase the number of experiments per chip (S4 Fig). A master mold was made by standard soft photolithography processes by spin-coating a 25μm layer of SU-8 2025 (Microchem, USA) photoresist at 2700 rpm for 30sec on a 3" wafer (Neyco, FRANCE). Then, we used a soft bake f 7min at 95°C on hot plates (VWR) followed by a UV 365nm exposure at 160 mJ/cm² with a mask aligner (UV-KUB3 Kloé®, FRANCE). Finally, a post-exposure baking identical to the soft bake was performed before development with SU-8 developer (Microchem, USA). Then, the wafer was baked at 150°C for 15min to anneal potential cracks and strengthen the adhesion of the resist to the wafer. Finally, the master mold was treated with chlorotrimethylsilane to passivate the surface.

Worm microchannels were cast by curing PDMS (Sylgard 184,10:1 mixing ratio), covalently bound to a 24 × 50 mm coverslip after plasma surface activation (Diener, Germany), and incubated 20min at 60°C for optimal adhesion. The chip was perfused with filtrated M9 solution through the medium inlet using a peristaltic pump (Ismatec), until complete removal of air bubbles. Worm loading was performed with the pump set at a low flow rate (<30 μl/min), through a distinct inlet (S4 Fig): 10–15 young L4 animals (synchronized by bleaching 48h prior to each experiment) were picked in a siliconized Eppendorf tube (Sigmacote SL2, Sigma Aldrich) containing M9, and perfused into the traps. The loaded chip was carried to the microscope while being still connected to the pump by a gravity flow (preventing animals to escape) until the microfluidic chip was installed on the microscope stage.

Calcium imaging

I2 and PHA neuronal response was monitored using the calcium sensor GCaMP3 expressed under the flp-15 promoter as in [8]. To image H₂O₂ response, young L4 animals trapped in microfluidic chips were imaged using the confocal spinning disc system described above with the 20X air objective (HC PL APO CS2, NA 0.75). The microfluidic chip allowed the simultaneous recording of up to 3 animals per experiment (S4 Fig). Z-stacks of 10–15 images (10 μm spacing) were acquired every 2s (using the stream Z mode), for 350 time points. Exposure time was 50ms and laser power set on 40%. The device was perfused with M9 medium throughout the experiment using a peristaltic pump set at 80 μl/min, and H₂O₂ was perfused (at 10μM or 1mM in M9) for 100 time points (3min20s) after an initial recording of 35–45 time points. Movies were computationally projected using MetaMorph, and data processing (including movie registration, neuron segmentation and tracking over time) was conducted with a custom-developed Matlab program detailed below and available at https://github.com/gcharvin/viewworm (a user guide is provided in Supplementary Information).

To image light response in I2 and PHA neurons, L4 worms were mounted on a slide covered with 3% agarose pads in M9 supplemented with 1mM levamisole. Video-recordings were performed on the spinning-disc microscope using the 40X oil objective. Animals were exposed to blue light (485nm) while their neuronal response was simultaneously recorded in stream mode (10 frames/sec, single Z, 100ms exposure, laser 100%) for 30sec.

Calcium response analyses

For H₂O₂ response analyses, sequences of images were spatially realigned with respect to the first image of the timeseries in order to limit the apparent motion of the worm in the trap and ease the tracking of neurons of interest. This image registration process was performed using standard 2D image cross-correlation by taking the first image as a reference. Then, we used a machine-learning algorithm (based on a decision tree) to segment pixels in the fluorescent images. For this, we took a series of image transforms (gaussian, median, range filters) as descriptors for the classifier, and we trained the model on typically 10 frames before applying the result to the rest of the time series. This segmentation method appeared to be superior to simple image thresholding, which is inadequate when dealing with fluorescent signals that vary both in time and space (the brightness of two neurons is quite different). Next to the segmentation procedure, we tracked the identified neurons using distance minimization, and we quantified the mean fluorescence signal in each neuron over time. Last, fluorescence data corresponding to individual animals were pooled after synchronization from the time of exposure to H₂O₂ and signal normalization. This image analysis pipeline is available at https://github.com/gcharvin/viewworm and a tutorial for use is included in Supplementary Information. As some movies could not be quantified due to uncontrolled animal movements, a visual classification of neuronal responses (high, moderate, absent) was made by comparison to successfully tracked movies.

For light response analyses in I2 and PHA neurons, the same image processing pipeline as in [8] was used, except that ROI were manually drawn in Fiji.

Statistical analyses

For pairwise comparisons of data sets with a normal distribution (or N>30), p-values were calculated using an unpaired two-tailed Student test, and the Welch correction was applied when samples variance was not homogeneous. When distributions were not normal (Anderson-Darling and Shapiro Wilk tests not satisfied), a Mann Whitney test was used. Multiple comparisons were analyzed using a one-way ANOVA with Bonferroni’s correction (for normal distributions), or a non-parametrical Kruskal-Wallis test followed by a Dunn’s multiple comparisons test. Statistical analyses were conducted with the GraphPad Prism9 software.

Mean are represented and error bars indicate standard deviation (SD) in all figures. The data presented here come from at least three independent experiments. For p values, not significant p>0.05; *p<0.05; **p<0.01; ***p<0.001: ****p<0.0001.