Sensory behaviors are often flexible, allowing animals to generate context-appropriate responses to changing environmental conditions. To investigate the neural basis of behavioral flexibility, we examined the regulation of carbon dioxide (CO2) response in the nematode Caenorhabditis elegans. CO2 is a critical sensory cue for many animals, mediating responses to food, conspecifics, predators, and hosts (Scott, 2011; Buehlmann et al., 2012; Chaisson and Hallem, 2012). In C. elegans, CO2 response is regulated by the polymorphic neuropeptide receptor NPR-1: animals with the N2 allele of npr-1 avoid CO2, whereas animals with the Hawaiian (HW) allele or an npr-1 loss-of-function (lf) mutation appear virtually insensitive to CO2 (Hallem and Sternberg, 2008; McGrath et al., 2009). Here we show that ablating the oxygen (O2)-sensing URX neurons in npr-1(lf) mutants restores CO2 avoidance, suggesting that NPR-1 enables CO2 avoidance by inhibiting URX neurons. URX was previously shown to be activated by increases in ambient O2 (Persson et al., 2009; Zimmer et al., 2009; Busch et al., 2012). We find that, in npr-1(lf) mutants, O2-induced activation of URX inhibits CO2 avoidance. Moreover, both HW and npr-1(lf) animals avoid CO2 under low O2 conditions, when URX is inactive. Our results demonstrate that CO2 response is determined by the activity of O2-sensing neurons and suggest that O2-dependent regulation of CO2 avoidance is likely to be an ecologically relevant mechanism by which nematodes navigate gas gradients.
Animals from nematodes to humans respond to environmental gases, such as CO2 and O2. CO2 is aversive for many free-living animals but attractive for many parasitic animals, which rely on CO2 for host location (Luo et al., 2009; Chaisson and Hallem, 2012). O2 increases or decreases can evoke avoidance responses in flies and nematodes (Chang et al., 2006; Morton, 2011) and alter foraging and feeding behaviors (Wingrove and O'Farrell, 1999; Cheung et al., 2005; Rogers et al., 2006; Vigne and Frelin, 2010). These responses are critical for survival: exposure to hypercapnia, hyperoxia, or hypoxia can result in reduced neural activity, cell cycle arrest, tumor formation, or death (Wingrove and O'Farrell, 1999; Harris, 2002; West, 2004; Langford, 2005).
The nematode Caenorhabditis elegans detects and responds to changes in environmental CO2 and O2 (Scott, 2011). C. elegans adults migrate away from a CO2 source and toward ∼10% O2 (Gray et al., 2004; Bretscher et al., 2008; Hallem and Sternberg, 2008). However, CO2 response can vary with developmental stage and environmental context. For example, CO2 is repulsive for adults but attractive for dauer larvae (Hallem et al., 2011a), and the behavioral response to simultaneous changes in CO2 and O2 levels is indicative of an interaction between the responses to the two gases (Bretscher et al., 2008; McGrath et al., 2009).
The response of C. elegans to CO2 and many other stimuli is regulated by NPR-1, a polymorphic neuropeptide receptor homologous to mammalian neuropeptide Y receptors (de Bono and Bargmann, 1998; Gray et al., 2004; Rogers et al., 2006; Bretscher et al., 2008; Hallem and Sternberg, 2008; Macosko et al., 2009; McGrath et al., 2009; Reddy et al., 2009). The N2 strain of C. elegans contains an npr-1 allele that confers solitary feeding behavior, whereas the CB4856 Hawaiian (HW) strain contains an npr-1 allele that confers social feeding behavior (de Bono and Bargmann, 1998). N2 animals respond strongly to CO2 but weakly to O2 on food, whereas HW animals appear relatively indifferent to CO2 but respond strongly to O2 on food (Gray et al., 2004; Bretscher et al., 2008; Hallem and Sternberg, 2008). NPR-1 is thought to act by repressing neural activity (Chang et al., 2006; Macosko et al., 2009).
To investigate the mechanisms of CO2 response plasticity, we examined the regulation of CO2 response by NPR-1. We show that HW and npr-1(lf) animals do not avoid CO2 despite showing normal CO2-evoked activity in BAG neurons. However, ablation of URX neurons in npr-1(lf) animals restores CO2 avoidance, suggesting that NPR-1 enables CO2 avoidance by decreasing URX activity. URX is activated by increases in ambient O2 (Persson et al., 2009; Zimmer et al., 2009; Busch et al., 2012), and we show that its O2-sensing ability is required to inhibit CO2 avoidance. We also show that HW and npr-1(lf) animals avoid CO2 under low O2 conditions, when URX is inactive. Our results suggest that CO2 avoidance is regulated by ambient O2 via a pair of O2-sensing neurons, allowing flexible responses to fluctuating levels of environmental gases.
Materials and Methods
C. elegans strains are listed in the order in which they appear in the figures. The following strains were used: N2 (Bristol); DA609 npr-1(ad609); CB4856 (Hawaiian); CX11697 kyIs536[flp-17::p17 SL2 GFP, elt-2::mCherry]; kyIs538[glb-5::p12 SL2 GFP, elt-2::mCherry]; EAH2 gcy-9(tm2816); PS6416 pha-1(e2123); syEx1206[gcy-33::G-CaMP3.0, pha-1(+)]; EAH117 npr-1(ad609); syEx1206[gcy-33::G-CaMP3.0, pha-1(+)]; EAH119 bruEx89[gcy-33::G-CaMP-3.0, ets-8::GFP]; MT17148 flp-21(ok889); flp-18(n4766); PR767 ttx-1(p767); GN112 pgIs2[gcy-8::caspase, unc-122::GFP]; PR679 che-1(p679); MT18636 nIs326[gcy-33::YC3.60]; lin-15AB(n765); AX2047 gcy-8::YC3.60, unc-122::dsRed; XL115 flp-6::YC3.60; CX9592 npr-1(ad609); kyEx2016[npr-1::npr-1 SL2 GFP, ofm-1::dsRed]; CX9395 npr-1(ad609); kyEx1965[gcy-32::npr-1 SL2 GFP, ofm-1::dsRed]; CX9633 npr-1(ad609); kyEx2096[flp-8::npr-1 SL2 GFP, ofm-1::dsRed]; CX9396 npr-1(ad609); kyEx1966[flp-21::npr-1 SL2 GFP, ofm-1::dsRed]; CX9644 npr-1(ad609); kyEx2107[ncs-1::npr-1 SL2 GFP, ofm-1::dsRed]; CX7102 lin-15(n765) qaIs2241[gcy-36::egl-1, gcy-35::GFP, lin-15(+)]; CX7158 npr-1(ad609) qaIs2241[gcy-36::egl-1, gcy-35::GFP, lin-15(+)]; ZG629 iaIs22[gcy-36::GFP, unc-119(+)]; EAH80 iaIs22[gcy-36::GFP, unc-119(+)]; npr-1(ad609); EAH106 bruEx86[gcy-36::G-CAMP3.0, coel::RFP]; EAH114 npr-1(ad609); bruEx86[gcy-36::G-CaMP3.0, coel::RFP]; ZG24 ahr-1(ia3); ZG624 ahr-1(ia3); npr-1(ad609); CX6448 gcy-35(ok769); CX7157 gcy-35(ok769); npr-1(ad609); RB1902 flp-19(ok2460); PT501 flp-8(pk360); PT502 flp-10(pk367); EAH123 npr-1(ad609) flp-19(ok2460); EAH141 npr-1(ad609) flp-8(pk360); EAH140 flp-10(pk367); npr-1(ad609); PS5892 gcy-33(ok232); gcy-31(ok296); EAH127 gcy-33(ok232); gcy-31(ok296) lon-2(e678) npr-1(ad609). In addition, CX7376 kyIs511[gcy-36::G-CaMP, coel::GFP] and EAH115 kyIs511[gcy-36::G-CaMP, coel::GFP]; npr-1(ad609) were used to confirm the results shown in Figure 5A with independent transgenes, and RB1903 flp-19(ok2461) and EAH139 npr-1(ad609) flp-19(ok2461) were used to confirm the results shown in Figure 5C with an independent deletion allele of flp-19. All transgenes were injected into N2, except bruEx89, which was injected into CB4856 to generate EAH119. EAH2 was derived from FX2816 by outcrossing to N2 for five generations. Nematodes were cultured on NGM plates containing Escherichia coli OP50 according to standard methods (Brenner, 1974). C. elegans dauer larvae were collected from the lids of plates from which the OP50 food source had been depleted (“starved plates”) and stored in dH2O at 15°C before use. All nematodes tested were hermaphrodites.
Generation of reporter transgenes and transgenic animals.
To generate EAH119, the gcy-33::G-CaMP3.0 construct from PS6416 was injected into CB4856 at 50 ng/μl along with ets-8::GFP at 50 ng/μl as a coinjection marker. To generate EAH106, a gcy-36::G-CaMP3.0 transcriptional fusion construct was generated by amplifying a 1.0 kb region upstream of the start codon of the gcy-36 gene from N2 genomic DNA using primers that included the following sequences: 5′-gatgttggtagatggggtttgga-3′ and 5′-aaattcaaacaagggctacccaaca-3′. The promoter fragment was then cloned into a modified Fire vector containing the G-CaMP3.0 coding region (Tian et al., 2009). The gcy-36::G-CaMP3.0 construct was injected into N2 animals at a concentration of 25 ng/μl along with 50 ng/μl of coel::RFP as a coinjection marker.
Acute CO2 avoidance assays.
Acute CO2 avoidance assays were performed as previously described (Hallem and Sternberg, 2008; Guillermin et al., 2011; Hallem et al., 2011b). Briefly, ∼10–15 young adults were tested on 5 cm assay plates consisting of NGM agar seeded with a thin lawn of E. coli OP50 bacteria. Gas stimuli consisted of certified industrial mixes (Airgas or Air Liquide). CO2 stimuli consisted of 10% CO2, 10% O2 (unless otherwise indicated), and the rest N2. Control stimuli consisted of 10% O2 (unless otherwise indicated) and the rest N2. Two 50 ml gas-tight syringes were filled with gas: one with CO2 and one without CO2. The mouths of the syringes were connected to flexible PVC tubing attached to Pasteur pipettes, and gases were pumped through the Pasteur pipettes using a syringe pump at a rate of 1.5 ml/min. Worms were exposed to gases by placing the tip of the Pasteur pipette near the head of a forward-moving worm, and a response was scored if the worm reversed within 4 s. Gases were delivered blindly, and worms were scored blindly. An avoidance index was calculated by subtracting the fraction of animals that reversed to the air control from the fraction that reversed to the CO2. Single-worm acute CO2 avoidance assays were performed on L4 or young adult laser-ablated animals (see Fig. 3C) as described above, except that each animal was tested 12 times with >2 min between trials. For each animal, an avoidance index was calculated by subtracting the fraction of trials in which it reversed to the air control from the fraction of trials in which it reversed to the CO2 stimulus. The avoidance index for each genotype or treatment was calculated as the mean avoidance index for each animal of the same genotype or treatment.
CO2 chemotaxis assays.
CO2 chemotaxis assays were performed on young adults essentially as previously described (Bretscher et al., 2008). Briefly, animals were washed off plates and into a 65 mm Syracuse watch glass using M9 buffer. Animals were washed 3× with M9 and transferred from the watch glass to a 1 cm × 1 cm square of Whatman paper. Animals were then transferred from the filter paper to the center of a 9 cm NGM or chemotaxis plate (Bargmann et al., 1993). Gas stimuli were delivered to the plate though holes in the plate lids as previously described (Hallem et al., 2011a; Dillman et al., 2012), except at a flow rate of 2 ml/min. Assay plates were placed on a vibration-reducing platform for 20 min. The number of worms in a 2-cm-diameter circle centered under each gas inlet was then counted, except for Figures 1B and 2A, B, where the number of worms in an area comprising ∼3/10 of the plate under each gas inlet was counted. The chemotaxis index was calculated as follows: (no. of worms at CO2 − no. of worms at control)/(no. of worms at CO2 + control). Two identical assays were always performed simultaneously with the CO2 gradient in opposite directions on the two plates to control for directional bias resulting from room vibration; assays were discarded if the difference in the chemotaxis index for the two plates was ≥0.9 or if <7 worms moved into the scoring regions on one or both of the plates.
For CO2 chemotaxis assays under different O2 conditions, assays were performed as described above inside airtight canisters (OGGI; 13.3 cm × 10.1 cm) with four holes drilled into the lids to insert tubing for gas flow. One hole was used to establish the ambient O2 level, two were used to establish the CO2 gradient, and one was used as an exhaust. A gas mixture consisting of either 7% O2 and the balance N2, or 21% O2 and the balance N2, was pumped into the chamber at a rate of 2.5 L/min for 1 min and then 0.5 L/min for the duration of the assay. The CO2 stimulus (10% CO2, either 7% O2 or 21% O2, balance N2) and control stimulus (7% O2 or 21% O2, balance N2) were pumped into the chamber at a rate of 2 ml/min using a syringe pump, as described above. The assay duration was 25 min.
Dauer CO2 chemotaxis assays were performed as previously described (Hallem et al., 2011a; Dillman et al., 2012). Briefly, assays were performed on chemotaxis plates (Bargmann et al., 1993). For each assay, ∼50–150 dauers were placed in the center of the assay plate. Gas stimuli and gas delivery to the assay plate were as described above, and a chemotaxis index was calculated as described above.
Imaging was performed using the genetically encoded calcium indicators G-CaMP (Zimmer et al., 2009), G-CaMP3.0 (Tian et al., 2009), or yellow cameleon YC3.60 (Nagai et al., 2004). Young adult or L4 animals were immobilized onto a cover glass containing a 2% agarose pad made with 10 mm HEPES using Surgi-Lock 2oc instant tissue adhesive (Meridian). A custom-made gas delivery chamber was placed over the cover glass. Gases were delivered at a rate of 0.8–1 L/min. Gas delivery was controlled by a ValveBank4 controller (AutoMate Scientific). Imaging was performed on an AxioObserver A1 inverted microscope (Carl Zeiss) using a 40× EC Plan-NEOFLUAR lens, a Hamamatsu C9100 EM-CCD camera, and AxioVision software (Carl Zeiss). For YC3.60 imaging, the emission image was passed through a DV2 beam splitter (Photometrics) as previously described (Hallem et al., 2011b). Image analysis was performed using AxioVision software (Carl Zeiss) and Microsoft Excel. The mean pixel value of a background region of interest was subtracted from the mean pixel value of a region of interest containing the neuron soma. Fluorescence values were normalized to the average values obtained in the 4 s before CO2 delivery. For YC3.60 imaging, the YFP/CFP ratio was calculated as previously described (Hallem et al., 2011b). Images were baseline corrected using a linear baseline correction. Traces with unstable baselines before the onset of the CO2 stimulus were discarded.
Ablations were performed on L2 and L3 animals as previously described (Hallem and Sternberg, 2008). Briefly, animals were mounted on glass slides for DIC microscopy on a pad consisting of 5% Noble agar in dH2O with 5% sodium azide as anesthetic. Ablations were performed on a Zeiss AxioImager A2 microscope with an attached MicroPoint laser (Carl Zeiss). Neurons were ablated by focusing a laser microbeam on the cell. Mock-ablated animals were mounted similarly but were not subjected to a laser microbeam. Neurons were identified by both cell position and GFP expression. Loss of the ablated cell was confirmed by observing loss of fluorescence in the adult animal.
Nematodes were anesthetized with 3 mm levamisole and mounted on a pad consisting of 5% Noble agar in dH2O. Epifluorescence images were captured using a Zeiss AxioImager A2 microscope with an attached Zeiss AxioCam camera and Zeiss AxioVision software (Carl Zeiss). To quantify epifluorescence in Figure 4D, all images were taken with the same exposure time. Average pixel intensities in the region of interest were quantified using AxioVision software (Carl Zeiss). Relative intensities were normalized by setting the highest mean intensity value to 1.
Statistical analysis was performed using GraphPad Instat and Prism. All significance values reported are relative to the N2 control, unless otherwise indicated.
NPR-1 regulates CO2 avoidance behavior
To investigate the role of npr-1 in mediating CO2 response, we examined the CO2-evoked behavior of N2, HW, and npr-1(lf) animals in both a chemotaxis assay and an acute avoidance assay. We found that N2 animals displayed robust CO2 avoidance in both assays, whereas HW and npr-1(lf) animals were essentially unresponsive to CO2 in both assays (Fig. 1A). Thus, the N2 allele of npr-1 is required for the behavioral response to CO2. CO2 avoidance behavior also requires the CO2-detecting BAG neurons and the receptor guanylate cyclase gene gcy-9, which encodes a putative receptor for CO2 or a CO2 metabolite (Fig. 1B) (Hallem and Sternberg, 2008; Hallem et al., 2011b; Brandt et al., 2012). To test whether npr-1 is required for CO2 detection, we imaged from BAG neurons using the genetically encoded calcium indicator G-CaMP3.0 (Tian et al., 2009). We found that the BAG neurons of N2, npr-1(lf), and HW animals all showed CO2-evoked activity (Fig. 1C), suggesting that npr-1 regulates the behavioral response to CO2 downstream of the calcium response of BAG neurons. The flp-21 and flp-18 genes, which encode NPR-1 ligands, are not required for CO2 avoidance, suggesting that other ligands are required for the regulation of CO2 response by npr-1 (Fig. 1D).
In addition to the BAG neurons, the salt-sensing ASE neurons and the temperature-sensing AFD neurons have been implicated in CO2 detection and avoidance (Bretscher et al., 2011). However, we found that che-1 mutant animals, which lack functional ASE neurons (Uchida et al., 2003), displayed normal CO2 avoidance in both a chemotaxis assay and an acute assay (Fig. 2A) (Hallem and Sternberg, 2008). Both AFD-ablated animals and ttx-1 mutant animals, which lack functional AFD neurons (Satterlee et al., 2001), showed defective CO2 avoidance in a chemotaxis assay but not an acute assay (Fig. 2A) (Hallem and Sternberg, 2008). These results suggest that ASE neurons are not required for CO2 avoidance under our assay conditions and that AFD neurons are required for some but not all CO2-evoked behaviors. By contrast, animals lacking BAG neurons showed a complete loss of CO2 response in both assays, regardless of whether CO2 was delivered in combination with 10% O2, which approximates the preferred O2 concentration of C. elegans (Gray et al., 2004), or 21% O2, which approximates atmospheric O2 concentration (Figs. 1B and 2B). We then imaged from BAG, ASE, and AFD neurons using the calcium indicator yellow cameleon YC3.60 (Nagai et al., 2004). We observed CO2-evoked activity in BAG neurons but not AFD and ASE neurons in response to a 20 s pulse of either 5% or 10% CO2 (Fig. 2C,D). Thus, BAG neurons are the primary sensory neurons that contribute to CO2 response under our assay conditions.
NPR-1 regulates URX neuron activity to control CO2 avoidance behavior
NPR-1 is not expressed in BAG neurons but is expressed in a number of other sensory neurons as well as some interneurons (Macosko et al., 2009). To identify the site of action for the regulation of CO2 response by npr-1, we introduced the N2 allele of npr-1 into npr-1(lf) mutants in different subsets of neurons and assayed CO2 response. We found that expressing npr-1 in neuronal subsets that included the O2-sensing URX neurons (Cheung et al., 2004; Gray et al., 2004) restored CO2 response (Fig. 3A). These results suggest that NPR-1 activity in URX neurons is sufficient to enable CO2 avoidance. However, we cannot exclude the possibility that NPR-1 function in other neurons also contributes to CO2 avoidance.
To further investigate the role of the URX neurons in regulating CO2 response, we ablated URX neurons in both the N2 and npr-1(lf) backgrounds and assayed CO2 avoidance behavior. We found that either genetic ablation of a neuronal subset that includes URX or specific laser ablation of URX in the N2 background had no effect on CO2 avoidance (Fig. 3B,C). However, both genetic and laser ablation of URX in npr-1(lf) mutants restored CO2 avoidance (Fig. 3B,C). Moreover, the response of URX-ablated npr-1(lf) animals was not significantly different from the response of URX-ablated N2 animals in our laser ablation experiment (Fig. 3C). Thus, in npr-1(lf) mutants, URX neurons inhibit CO2 avoidance and removal of URX neurons is sufficient to restore CO2 avoidance. Our results suggest a model in which CO2 avoidance behavior is regulated by URX neuron activity. In N2 animals, NPR-1 reduces URX neuron activity, thereby enabling CO2 avoidance. In npr-1(lf) animals, increased activity of URX neurons inhibits the CO2 circuit, resulting in a loss of CO2 avoidance.
URX neurons are not required for CO2 attraction by dauers
In contrast to C. elegans adults and developing larvae, C. elegans dauer larvae are attracted to CO2 (Fig. 4A) (Guillermin et al., 2011; Hallem et al., 2011a). The dauer is a developmentally arrested, alternative third larval stage that is thought to be analogous to the infective juvenile stage of parasitic nematodes (Hotez et al., 1993). The mechanism responsible for the change in CO2 response valence that occurs at the dauer stage is not yet known. BAG neurons and the putative CO2 receptor GCY-9 are required for CO2 attraction by dauers (Fig. 4A) (Hallem et al., 2011a), suggesting that the same mechanism of CO2 detection operates at the dauer and adult stages. However, npr-1(lf) and HW dauers are also attracted to CO2, indicating that npr-1 is not required for CO2 attraction (Fig. 4B). The lack of requirement for npr-1 at the dauer stage is not the result of altered npr-1 expression in URX neurons because npr-1 is expressed at comparable levels in N2 dauers and developing third-stage larvae (L3s) (Fig. 4C,D). To test whether URX neuron activity is required for CO2 attraction by dauers, we tested whether dauers that lack URX neurons are still attracted to CO2. We found that URX-ablated N2 and npr-1(lf) dauers display normal CO2 attraction (Fig. 4E), indicating that URX neurons are not required to promote CO2 attraction by dauers. Thus, URX neurons control whether CO2 is a repulsive or neutral stimulus in adults, but other mechanisms are required to promote CO2 attraction by dauers.
O2 sensing by URX neurons is required for regulation of CO2 avoidance
The URX neurons are O2-sensing neurons that express O2 receptors of the soluble guanylate cyclase (sGC) family (Cheung et al., 2004; Gray et al., 2004). Whether the URX neurons are also activated by CO2 is unclear (Bretscher et al., 2011; Brandt et al., 2012). To test whether URX neurons regulate CO2 response by directly responding to CO2, we imaged from the URX neurons of N2 and npr-1(lf) animals during CO2 exposure using the calcium indicator G-CaMP3.0. We found that URX neurons are not activated by CO2 (Fig. 5A). URX neurons did appear to show a slight decrease in calcium levels in response to CO2, but whether this decrease is biologically relevant is not yet clear. These results indicate that URX neurons do not regulate CO2 response as a result of CO2-induced activation.
To test whether URX neurons instead regulate CO2 response by responding to O2, we examined the CO2-evoked behavior of aryl hydrocarbon receptor-1 (ahr-1) mutants. AHR-1 is a transcription factor that regulates aggregation behavior and that is required for normal expression of sGC O2 receptors in URX neurons (Qin et al., 2006). We found that ahr-1 mutants respond normally to CO2 and that the ahr-1 mutation rescues the CO2 response defect of npr-1(lf) mutants (Fig. 5B). Thus, regulation of CO2 avoidance by URX neurons of npr-1(lf) animals depends on their ability to sense O2. Furthermore, mutation of the sGC gene gcy-35, which encodes an O2 receptor that is expressed in URX and required for its O2 response (Zimmer et al., 2009), also rescues the CO2 response defect of npr-1 mutants (Fig. 5B). Thus, GCY-35-mediated activation of URX neurons by ambient O2 is required for regulation of CO2 avoidance behavior. Together, these results demonstrate that CO2 response is regulated by ambient O2.
To investigate the mechanism by which URX neurons regulate CO2 response in npr-1 mutants, we examined the role of neuropeptide signaling in the regulation of CO2 avoidance behavior. The URX neurons are known to express FMRFamide-related neuropeptide genes, including flp-8, flp-10, and flp-19 (Li and Kim, 2008). To test whether these neuropeptide genes are required for the regulation of CO2 response, we examined the CO2-evoked behavior of neuropeptide mutants in the npr-1(lf) mutant background. We found that mutation of either flp-8 or flp-19, but not flp-10, significantly rescued the CO2 response defect of npr-1 mutants (Fig. 5C). These results are consistent with the hypothesis that URX neurons modulate CO2 response via a neuropeptide signaling pathway involving flp-8 and flp-19. However, we cannot exclude the possibility that release of flp-8 and flp-19 from other neurons also contributes to the O2-dependent regulation of CO2 response.
npr-1(lf) and HW animals avoid CO2 under low O2 conditions
The URX neurons are activated when the ambient O2 concentration increases from 10% to 21% (Zimmer et al., 2009; Busch et al., 2012). This response consists of both phasic and tonic components: a large initial increase in calcium transients is followed by a smaller sustained increase that continues until O2 levels return to 10% (Busch et al., 2012). The fact that URX neurons remain active at high O2 levels but are inactive at low O2 levels led us to hypothesize that npr-1(lf) and HW animals might avoid CO2 under low O2 conditions, when URX neurons are inactive. We therefore examined the responses of npr-1(lf) and HW animals to CO2 under low O2 conditions by reducing the ambient O2 concentration to 7% for the duration of the CO2 chemotaxis assay. We found that, at 7% ambient O2, both npr-1(lf) and HW animals displayed CO2 avoidance behavior that was comparable with that of N2 animals (Fig. 6A). Thus, npr-1(lf) and HW animals are indeed capable of responding robustly to CO2. However, CO2 response in these animals is regulated by ambient O2 such that CO2 is repulsive at low O2 concentrations and neutral at high O2 concentrations.
The BAG neurons, which are activated by CO2, are also activated by decreases in ambient O2 from 21% to <10% (Zimmer et al., 2009). This raised the possibility that BAG neurons could cell-autonomously integrate responses to O2 and CO2, thus contributing to the O2-dependent regulation of CO2 response. To test this possibility, we examined the ability of animals that lack the soluble guanylate cyclase genes gcy-31 and gcy-33, which are expressed in BAG neurons and are required for the O2-evoked activity of BAG neurons (Zimmer et al., 2009), to respond to CO2 at low ambient O2. We found that gcy-33; gcy-31 mutants responded normally to CO2 at low ambient O2 in both N2 and npr-1(lf) animals (Fig. 6B), indicating that the O2-sensing ability of BAG is not required for the O2-dependent regulation of CO2 response. Consistent with these results, the BAG neurons were recently shown to play only a minor role in the chronic response to ambient O2 (Busch et al., 2012). Thus, regulation of CO2 response by ambient O2 is not a result of cell-intrinsic signaling within BAG but instead requires a pair of designated O2-sensing neurons.
Our results demonstrate that URX neurons control CO2 response by coordinating the response to CO2 with the response to ambient O2. In npr-1(lf) animals, O2-dependent activation of URX neurons determines CO2 response such that CO2 is repulsive at low ambient O2 but neutral at high ambient O2 (Fig. 6C). Moreover, our results are consistent with the hypothesis that URX neurons regulate the activity of the CO2 circuit via a neuropeptide signaling pathway that involves the FMRFamide-related neuropeptide genes flp-19 and flp-8. By contrast, in N2 animals, the URX neurons do not inhibit CO2 avoidance at high ambient O2 as a result of the presence of NPR-1 (Fig. 6C). NPR-1 does not constitutively silence the URX neurons of N2 animals because the URX neurons of N2 animals are activated by increases in ambient O2 and ablation of URX in N2 animals alters O2 response (Zimmer et al., 2009). However, our results suggest that NPR-1 may reduce URX neuron activity in N2 animals such that URX neurons no longer inhibit the CO2 avoidance circuit. Alternatively, it is possible that NPR-1 activity is dynamically regulated by its neuropeptide ligands such that it is active under some conditions but not others, or that the URX neurons of N2 animals are sufficiently activated but are incapable of regulating CO2 avoidance as a result of differences in neural connectivity or signaling between N2 and npr-1(lf) animals.
A recent survey of wild C. elegans strains revealed that the HW allele of npr-1 is the natural variant, with the N2 allele having arisen during laboratory culturing (McGrath et al., 2009). HW animals were previously thought to be virtually insensitive to CO2 (Hallem and Sternberg, 2008; McGrath et al., 2009), raising the question of whether CO2 avoidance is exclusively a laboratory-derived behavior. Our results demonstrate that HW animals do indeed display robust CO2 avoidance, but this behavior is restricted to low O2 conditions. Wild C. elegans adults have been found in fallen rotting fruit and in the soil under rotting fruit, where O2 levels are lower and CO2 levels are higher than in the atmosphere (Felix and Duveau, 2012). Inside rotting fruit, C. elegans occupies microhabitats replete with bacteria, fungi, worms, insects, and other small invertebrates (Felix and Duveau, 2012). In this context, fluctuating levels of CO2 and O2 likely serve as important indicators of food availability, population density, and predator proximity (Bendesky et al., 2011; Milward et al., 2011; Scott, 2011). Suppression of CO2 avoidance at high ambient O2 may allow worms to migrate toward rotting fruit, which emits CO2. Once inside the low O2 environment of rotting fruit, CO2 avoidance may allow worms to avoid cohabitating predators or overcrowding. Thus, O2-dependent regulation of CO2 avoidance is likely to be an ecologically relevant mechanism by which nematodes navigate gas gradients.
In addition to CO2 response, a number of other chemosensory behaviors in C. elegans are subject to context-dependent changes in sensory valence (Sengupta, 2012). For example, olfactory and gustatory behavior exhibits experience-dependent plasticity, in which chemicals that are attractive to naive animals become neutral or repulsive after prolonged or repeated exposure in the absence of food (Sengupta, 2012). Olfactory plasticity occurs as a result of altered signaling in the AWC olfactory neurons (Tsunozaki et al., 2008), and salt plasticity occurs as a result of altered signaling in the ASE gustatory neurons and the downstream AIA and AIB interneurons (Tomioka et al., 2006; Adachi et al., 2010; Oda et al., 2011). Similarly, O2 preference is modulated by prior O2 exposure and the presence of bacterial food as a result of altered signaling in a distributed network of chemosensory neurons (Cheung et al., 2005; Chang et al., 2006). Our results suggest that CO2 response is modulated by ambient O2 via the activity of a pair of O2-detecting neurons that interact with the CO2 circuit downstream of CO2 detection by BAG neurons (Fig. 6C). The neurons that act downstream of BAG and URX to control CO2 response have not yet been identified. A number of interneurons receive synaptic input from both BAG and URX (White et al., 1986), and it will be interesting to determine whether any of them play a role in CO2 avoidance.
CO2-evoked behaviors in insects are also subject to context-dependent modulation. For example, the fruit fly Drosophila melanogaster is repelled by CO2 when walking (Suh et al., 2004) but attracted to CO2 in flight, a valence change that is modulated by octopamine signaling (Wasserman et al., 2013). In addition, both CO2 repulsion by walking D. melanogaster and CO2 attraction by mosquitoes can be suppressed by food odorants, which directly alter the activity of the CO2 receptor (Turner and Ray, 2009; Turner et al., 2011). Insects as well as many other animals, both free-living and parasitic, occupy microhabitats where environmental levels of O2 and CO2 vary greatly as a function of food or host availability, population density, and microorganism composition. Thus, it will be interesting to determine whether the control of CO2 response by O2-sensing neurons is a conserved feature of gas-sensing circuits.
M.A.C. was supported by a National Science Foundation Graduate Research Fellowship (Grant No. DGE-0707424) and a Eugene V. Cota-Robles Fellowship. S.R. was supported by the National Institutes of Health National Institute of General Medical Sciences training Grant GM08042 and the UCLA-Cal Tech Medical Scientist Training Program. E.A.H. is a MacArthur Fellow, an Alfred P. Sloan Research Fellow, a Rita Allen Foundation Scholar, and a Searle Scholar. This work was supported by a National Institutes of Health R00 Grant to E.A.H (Grant No. R00-AI085107). We thank Cori Bargmann, Alon Zaslaver, Paul Sternberg, Jo Anne Powell-Coffman, Miriam Goodman, Maureen Barr, Ikue Mori, Mario de Bono, Shawn Lockery, Shohei Mitani, and the Caenorhabditis Genetics Center for C. elegans strains; Cori Bargmann, Alon Zaslaver, and Paul Sternberg for plasmids; Lars Dreier, Michelle Castelletto, Alvaro Sagasti, Doug Black, and Keely Chaisson for critical reading of this manuscript; and Joe Vanderwaart for insightful discussion of this manuscript.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Elissa A. Hallem, University of California, Los Angeles, MIMG 237 BSRB, 615 Charles E. Young Drive East, Los Angeles, CA 90095.