Long-term exposure to different oxygen environments creates a memory that configures C. elegans’ aversion to high CO₂ (Fenk and de Bono, 2017). We therefore hypothesized that C. elegans adapt their O₂ preference depending on previous oxygen experience. To test this, we cultured animals from birth in an atmosphere of either their preferred O₂ concentration (7%) or at 21% O₂, and assayed the speed of locomotion of young (1-day-old) adults when they were exposed to a series of stepwise decreasing concentrations from 21% to 7% O₂ and a final step from 7% to 21% O₂ (Figure 1—figure supplement 1A,B). The reference laboratory strain N2 has acquired a hyperactive version of the neuropeptide receptor NPR-1, which attenuates avoidance of 21% O₂ in the presence of bacterial food (McGrath et al., 2009; Persson et al., 2009). We therefore tested animals bearing the ad609 loss-of-function mutation in npr-1, thus restoring their robust avoidance of high O₂ (Busch et al., 2012). Strikingly, the concentration of O₂ at which the worms were initially cultured determined their responses to the stepwise changes in O₂ concentration: animals initially cultured in 21% O₂ gradually decreased their speed of locomotion as the O₂ concentration decreased from 21% to 7%, as observed in previous studies (Busch et al., 2012); those initially cultured in 7% O_2, however, were highly motile at all concentrations from 21% to 11% and only exhibited dwelling behavior when they reached their initial culture condition of 7% O₂ (Figure 1—figure supplement 1B).

We hypothesized that animals can ‘update’ their O₂ preference if they have recently been switched to a different oxygen environment, and tested this by transferring them from 21% to 7% O₂ or vice versa 12 hr before assaying their locomotory speed in response to stepwise changes in O₂ concentration (Figure 1A). Indeed, when 1-day-old adults cultured at 21% were shifted to 7% O₂ overnight, their locomotory response to intermediate O₂ strongly increased compared to animals cultured at high oxygen throughout, and resembled that of animals that had been cultured at 7% O₂ for their whole life (Figure 1B,C). Conversely, animals shifted from 7% O₂ to a 21% oxygen environment decreased their locomotory speed at intermediate O₂ relative to animals cultured at 7% oxygen throughout, and their response resembled that of animals cultured at 21% O₂ for their whole life (Figure 1B,C). This shows that the tuning of oxygen responses depends on the O₂ conditions animals were cultured in and is plastic according to the most recent O₂ concentration change.

To investigate how age might affect this oxygen-evoked behavior, we cultured animals at either 21% or 7% O₂ up to the adult age of 4, 7, or 10 days and tested them in the same assay as above. In both culture conditions, locomotory speed at 21% O₂ gradually declined with age, from 163 µm/s for 1-day-old to 97 µm/s for 10-day-old animals cultured in 21% O₂ and from 188 µm/s for 1-day-old to 114 µm/s for 10-day-old animals cultured in 7% O₂; these speeds still enabled animals to robustly avoid high oxygen environments, however (Figure 1B,C,E,F,H,I and Figure 1—figure supplement 1C,E). This suggests that the sensory ability itself remains largely intact until day 10 of adulthood but is accompanied by a progressive decline in locomotory ability, consistent with previous studies on the decline of C. elegans nervous system function with age, where locomotory speed typically reaches half-maximal capacity in 7–8 day-old adults (Glenn et al., 2004; Hahm et al., 2015; Hsu et al., 2009; Podshivalova et al., 2017). Response tuning by O₂ culture condition, that is, the higher locomotory speed at intermediate O₂ concentrations of worms cultured at 7% O₂ and lower speed of worms cultured at 21% O₂, is preserved in ageing animals (Figure 1B,C and Figure 1—figure supplement 1C,E). We then tested if experience-dependent plasticity of O₂ responses declines with age, by switching 3-, 6-, and 9-day-old animals to the opposite O₂ environment for 12 hr and recording O₂-evoked locomotion the next day (Figure 1A). Animals that were cultured at high ambient O₂ and switched to 7% started to lose the ability to update their O₂-evoked responses already at day 4 and by day 7 showed no difference to those of animals kept continuously at 21% (Figure 1B,D, Figure 1—figure supplement 1D). By contrast, animals cultured at 7% O₂ largely retained O₂ experience-dependent plasticity even up to day 10 of adulthood and changed their locomotory response to stepwise changes in O₂ concentration after switching to 21% O₂ for 12 hr (Figure 1C,D, Figure 1—figure supplement 1F). We calculated a plasticity index based on the sum of the speed differential across measurement points, normalized to the plasticity in young adults (see Materials and methods for details). The index documents the divergence in plasticity for worms exposed to a high or low oxygen environment (Figure 1D). The decline of cognitive plasticity with age is thus conditional on environmental state.

To test whether the plasticity loss in animals shifted from 21% to 7% O₂ and the plasticity retention in animals shifted from 7% to 21% O₂ may be a consequence of a bias in the assay paradigm, we compared the plasticity of 21% and 7% O₂ cultured animals after animals were shifted from 21% to 7% O₂, or from 7% to 21% O₂ (Figure 1—figure supplement 1G,H). Animals first cultured at 21% O₂ and then cultured for 1 day at 7% O₂ showed significantly less plasticity after the 7%→21% O₂ shift than animals continuously kept at 7% O₂ (Figure 1—figure supplement 1G,I), while animals first cultured at 7% O₂ and then cultured for 1 day at 21% O₂ show significantly higher plasticity after the 21%→7% O₂ shift than animals continuously kept at 21% O₂ (Figure 1—figure supplement 1H,I). This suggests that the decline rate of plasticity is dictated by the accumulated long-term environmental exposure to high or low O₂, rather than the direction of the O₂ shift. In addition, we observed that the plasticity of 7-day-old 21% O₂-raised animals cultured for 1 day at 7% O₂ before the plasticity experiment is significantly higher than that of animals kept at 21% O₂ throughout (Figure 1—figure supplement 1G–I). This suggests that the loss of plasticity in ageing 21% O₂-exposed animals is not irreversible but can be restored by temporary cultivation at low oxygen.

Previous experience also alters spatial oxygen preferences, and, like speed, this spatial preference can be updated by recent O₂ shifts. In a 7–21% O₂ gradient, animals previously cultured at 7% O₂ accumulated in a narrow range of O₂ near 7%, whereas those acclimated to 21% O₂ distributed over a broader range of concentrations between 7% and near 21% O₂, showing attenuated avoidance to high and intermediate O₂ concentrations (Figure 1E,H). These different preferences are maintained in 7-day-old adults (Figure 1F,I). Consistent with the differential decline of O₂-evoked locomotory responses, day 1 adults cultured at either 21% or 7% O₂ both show plasticity in their oxygen preference after switching culture conditions (Figure 1E,H,G,J), but only 7% O₂ cultured worms retained plasticity at day 7 of adulthood (Figure 1F,I,G,J).

Taken together, these results indicate that previous O₂ experience generates a memory that sculpts O₂-evoked behavioral responses. The ability to update this memory declines with age conditional on the previous O₂ environment: it is lost in ageing animals cultured at 21% O₂ but maintained in animals kept at 7% O₂.

To investigate how previous experience and age alter the neural encoding of O₂ levels, we recorded Ca²⁺ responses in URX neurons in vivo by using the Ca²⁺ sensor cameleon YC3.60 when 1- and 7-day-old adults were exposed to stepwise changing O₂ concentrations, with the same O₂ steps as in the assay testing the O₂-evoked locomotory speed. The URX oxygen-sensing neuron pair are tonic receptors that continuously signal ambient O₂ concentration, sufficient to mediate most O₂-evoked behaviors, including the regulation of locomotion, aerotaxis and aggregation (Busch et al., 2012; Macosko et al., 2009). Consistent with previous results (Busch et al., 2012), a 21% O₂ stimulus induced a tonically elevated Ca²⁺ response in URX, whereas at 7% O₂ the intracellular Ca²⁺ concentration dropped to baseline. Responses to O₂ were graded and declined with decreasing ambient O₂ concentration (Figure 2A,B). O₂ concentration in the culture environment profoundly affected the magnitude of O₂-evoked Ca²⁺ responses: in animals grown at 7% O₂, URX displayed significantly higher [Ca²⁺] responses to high and intermediate O₂ concentrations compared to 21% O₂ cultured animals (Figure 2A, red trace and Figure 2B, blue trace, Figure 2—figure supplement 1C).

To test whether the Ca²⁺ responses of URX neurons to O₂ reflect the plasticity seen in O₂-dependent locomotory behavior, we transferred animals from 21% to 7% O₂ or vice versa 12 hr before the Ca²⁺ imaging in one day old adults. Animals shifted from 21% to 7% O₂ showed a significantly increased URX responses to the O₂ stimuli compared to animals kept at 21% O₂ throughout (Figure 2A; Figure 2—figure supplement 1A; Figure 1B). Conversely, 1-day-old animals shifted from 7% to 21% O₂ showed significantly reduced responses to O₂ stimuli compared to animals kept at 7% O₂ (Figure 2B; Figure 2—figure supplement 1A; Figure 1C). There was no significant difference between the URX Ca²⁺ responses in worms shifted from 21% to 7% O₂ (Figure 2A, blue line) and those of animals that have been grown at 7% O₂ only (Figure 2B, blue line; mixed-model ANOVA, p=0.2682). Therefore, Ca²⁺ responses of URX neurons in young adults to O₂ are reprogrammed by switching the animals to a different O₂ environment.

We next tested how age affects URX activity and plasticity. In 7-day-old adults cultured at 7% O₂, Ca²⁺ responses to high O₂ were significantly reduced in magnitude when compared to those of 1-day-old adults, whereas the responses of animals kept at 21% O₂ were maintained at similar levels to those of young adults (Figure 2—figure supplement 1D, Figure 2C, red trace and Figure 2D, blue trace). When the ageing 21% O₂ cultured animals were shifted to 7% for 12 hr, by contrast, the plasticity of the URX response was lost entirely (Figure 2C; Figure 2—figure supplement 1B). In contrast, 7-day adults cultured at 7% O₂ retain the URX plasticity observed in young adults where Ca²⁺ responses to the series of oxygen stimuli are reduced after being shifted to 21% O₂ for 12 hr (Figure 2D; Figure 2—figure supplement 1B).

Taken together, these results show that the URX Ca²⁺ responses are programmed by previous O₂ experience in the culture environment, and display strong experience-dependent plasticity in response to recently changed ambient O₂ concentration. Moreover, previous experience governs the age-dependent decline of both neuronal and behavioral plasticity: exposure to high ambient oxygen causes their decline and loss within 1 week, whereas at lower oxygen levels plasticity is well preserved in 7-day-old animals.

The loss of experience-dependent plasticity in ageing animals cultured in a high O₂ environment may result from either tonically high activity and elevated [Ca²⁺] in the O₂-sensing neurons (Busch et al., 2012) or increased oxidative stress in the organism (Jagannathan et al., 2016).

To test if chronically elevated neural activity is responsible for the accelerated decline of plasticity with age, we inhibited the activity of the oxygen-sensing neurons URX, AQR and PQR long-term while exposing them to the high oxygen condition. To do so, we expressed a chemogenetic tool, the Drosophila histamine-gated chloride channel HisCl1, in the O₂-sensing neurons. Addition of exogenous histamine to C. elegans culture plates enables the rapid, reversible and prolonged inhibition of neurons expressing this channel (Pokala et al., 2014). We cultured pgcy-32::HisCl1 animals and control animals lacking the transgene in 21% oxygen from birth until 6 days old in the presence of 20 mM histamine, and pgcy-32::HisCl1 animals grown in 21% O₂ in the absence of histamine. They were then either switched to 7% O₂ for 18 hr or retained at 21% O₂ (Figure 3A). Animals with inhibited O₂-sensing neurons showed strong plasticity after switching them to 7% O₂, whereas control animals grown without histamine or not expressing HisCl1 showed little plasticity (Figure 3B). Plasticity of the 7-day-old animals with inhibited neurons was nearly as great as that of 1-day-old animals and significantly greater than in either of the control conditions, whereas histamine treatment or HisCl1 expression itself did not prevent the decline of plasticity with age (Figure 3C and Figure 3—figure supplement 2D). In contrast, experience-dependent plasticity of 7-day-old adults cultured at 7% O₂, where the O₂-sensing neurons are not chronically active, was not altered by HisCl1-mediated silencing of the O₂-sensing neurons (Figure 3—figure supplement 1A–C,E).

The chemogenetic silencing of the O₂-sensing neurons shows that high neural activity is necessary to accelerate the decline of plasticity. To investigate if high neural activity is sufficient to accelerate cognitive decline, we performed chronic optogenetic stimulation of the URX O₂-sensing neurons in animals cultured at 7% O₂, using the blue light-gated cation channel Channelrhodopsin (ChR) (Figure 3D–F). We tested 4-day-old adults. At this age, experience-dependent plasticity is sustained in animals grown at 7% O₂, but declines in animals cultured at 21% O₂ (Figure 1D). Animals cultured at 7% O₂ with chronic URX excitation showed significantly reduced behavioral plasticity at day 4 of adulthood, compared to controls (ChR-expressing animals exposed to blue light without the administration of the rhodopsin cofactor all-trans retinal) (Figure 3D,F). In addition, chronic blue light illumination in the presence of all-trans retinal (ATR) did not significantly affect plasticity in control animals without ChR expression (Figure 3E,F). The chronic stimulation only alters the behavior of animals that are switched to 21% O₂ but not that of animals kept at 7% O₂ throughout (Figure 3D). This suggests that chronic optogenetic activation of URX does not alter O₂ responses per se, but specifically changes how the plasticity of the O₂ response declines with age.

To investigate the potential role of oxidative stress, we also performed the behavioral plasticity assay in ageing C. elegans cultured in the presence of the antioxidant N-acetyl cysteine (NAC), which increases resistance to oxidative stress (Desjardins et al., 2017). NAC treatment did not restore behavioral plasticity to 7-day-old animals cultured at high oxygen levels, nor did it have an effect on 1- and 7-day-old adults cultured in 7% O₂ that do show oxygen-evoked behavioral plasticity (Figure 3—figure supplement 1B–F). NAC significantly reduced expression of the oxidative stress reporter gst-4p::GFP, confirming that supplementation with this compound reduces oxidative stress in the assayed strain and culture conditions (Figure 3—figure supplement 1G).

Thus, long-term chemogenetic inactivation of O₂-sensing neurons prevents the loss of experience-dependent plasticity in ageing animals cultured in high oxygen concentrations, whereas chronic optogenetic activation of these neurons induces a loss of plasticity in ageing animals cultured at low oxygen. Treatment with the antioxidant NAC to reduce oxidative stress does not ameliorate the loss of experience-dependent plasticity in worms cultured at high oxygen. Together, these data indicate that the amount of neuronal activity governs the decline of cognitive plasticity with age: chronically high neuronal activity causes the decline and eventual loss of the ability of neurons to reconfigure their responses following sensory input changes, compromising their ability to generate behavioral plasticity and adaptation in ageing individuals, whereas neurons that are chronically under-active in the long term are better at retaining neuronal plasticity.

Sustained neuronal excitation and the associated elevated intracellular [Ca²⁺] profoundly affect gene transcription, reconfiguring functional properties of neurons and underpinning experience-dependent plasticity (Flavell and Greenberg, 2008; Tyssowski et al., 2018). To elucidate the molecular and cellular processes responsible for the rapid loss of plasticity with age induced by high neural activity, we investigated whether and how neuronal activity regulates gene transcription in the O₂-sensing neurons during ageing. O₂-sensing neurons are tonically activated by 21% oxygen, and persistently inactive at 7% O₂ (Busch et al., 2012). We therefore studied the transcriptomes of neurons from animals cultured in high (21%) or low (7%) O₂ concentrations as a proxy for neurons at high and low activity state, respectively. We isolated O₂-sensing neurons from 1- and 7-day-old age-synchronized C. elegans populations expressing pgy-37::GFP in the URX, AQR, and PQR oxygen-sensing neurons by enzymic dissociation of worm bodies and fluorescence-activated cell sorting, and determined their transcriptional profiles by RNA sequencing (Figure 4A and Materials and methods; Supplementary file 1). For comparison we also obtained transcriptional profiles of whole C. elegans bodies at each corresponding condition (Supplementary file 1). An interactive presentation of the dataset using the web tool DrEdGE (Tintori et al., 2020) is available (Supplementary file 2; dredge.bio.unc.edu/BuschLab_eLife). For 1-day-old animals cultured in 21% O₂, 7881 genes were differentially expressed in the O₂-sensing neuron samples when compared to whole-worm transcriptomes, of which 4058 genes were enriched in the O₂-sensing neurons (at a false discovery rate (FDR) of <0.1 and a minimum fold change of 1.5). Genes known to be expressed in URX, AQR, and PQR neurons were among the most highly enriched transcripts in neuronal samples from both 1- and 7-day-old animals. These genes included, for example, the atypical guanylate cyclases gcy-32, gcy-33, gcy-34, gcy-35, gcy-36 and gcy-37, which are causally linked to oxygen sensing (Cheung et al., 2004; Gray et al., 2004; Zimmer et al., 2009), and the FMRFamide neuropeptide gene flp-8, which is only expressed in the URX and AUA neurons (Kim and Li, 2004). Tissue enrichment analysis (Angeles-Albores et al., 2016) of all the neuronal samples also indicated that they were enriched for genes known to be expressed in URX, AQR, and PQR neurons (Supplementary file 3). Moreover, a previously published list of genes enriched in the O₂-sensing neurons (Cao et al., 2017; Supplementary file 4) is also significantly enriched in our O₂-sensing neuron samples (normalized enrichment score = 1.79, q = 0). These data confirmed that neuronal RNA isolated from both the 1- and 7-day-old animals in the experiment was highly enriched for transcripts expressed in the O₂-sensing neurons URX, AQR, and PQR. We then did a thorough overlap analysis for differentially expressed genes (DEGs), comparing the O₂-sensing neurons to whole-worm-derived transcripts of 1- and 7-day-old animals, and found an extensive overlap of genes enriched or depleted in the O₂-sensing neurons across age and O₂ culture conditions, suggesting that the expression of genes for neuronal identity remains stable at different culture conditions and life stages (Figure 4B).

When comparing high (21%) and low (7%) O₂ conditions, 301 genes were found to be differentially expressed among the O₂-sensing neuron samples and 248 genes among the whole worm samples in 1-day-old adults, whereas in 7-day-old adults, 264 neuronal and 50 whole-worm genes were differentially expressed. Very few of these DEGs were common to the neuronal and whole-worm gene sets, or to the young and old samples (Figure 4—figure supplement 2A,B). To determine which gene categories were enriched in the differentially expressed gene sets, we performed gene ontology (GO) analysis. As expected, several GO terms were co-regulated in the same direction in the neuronal and whole-worm transcriptomes, which may reflect general cellular responses to ambient O₂ concentration. Specifically, genes involved in neuropeptide signaling were upregulated in both neurons and whole worm transcriptomes of young adults cultured at the high O₂ concentration, whereas several GO terms related to translation were downregulated in both neurons and whole worms at this O₂ concentration (Figure 4C). In 7-day-old worms, only one gene class overlapped between neurons and whole-worm samples (Supplementary file 5). Although the genes expressed in the O₂-sensing neurons that are upregulated at high O₂ overlap by only nine genes between young and old animals (Figure 4—figure supplement 2B), they were enriched in many of the same categories of C. elegans phenotypes as shown by phenotype enrichment analysis (Angeles-Albores et al., 2016; Supplementary file 3). Genes whose expression was significantly increased at 21% O₂ were enriched in phenotypes primarily linked to high locomotory activity, such as ‘forward locomotion increased’ and ‘pausing decreased’ (Supplementary file 3). This is notable as C. elegans display persistently high locomotory activity at 21% O₂ (Busch et al., 2012).

The GO terms that were enriched in genes differentially expressed at high or low O₂ concentration specifically in O₂-sensing neurons were very different in young and old adult worms: no categories were co-regulated in the same way in both age groups. In neurons from young animals, only lipid transport was upregulated by high O₂, whereas most categories that were downregulated were linked to cellular organization and signaling processes (Figure 4D). This suggests that the O₂-sensing neurons largely do not rely on inducing transcription to maintain high neural activity in young adults. By contrast, in neurons from old animals all enriched gene groups were upregulated at high O₂, which signifies chronically high neuronal activity (Figure 4D, Supplementary file 5). These gene groups were related to Ca²⁺-dependent processes and signaling pathways such as cGMP signaling, which is essential for O₂ sensory function (Couto et al., 2013; Figure 4D). We also found that a manually curated set of ‘genes linked to Ca²⁺ signaling’ (Supplementary file 6), which primarily represents genes that modulate Ca²⁺ homeostasis, was upregulated by high O₂ in aged neurons but not in young neurons, and was downregulated in whole worms cultured in high O₂ (Figure 4C). This suggests that the accumulated effect of chronically elevated [Ca²⁺] in tonically active O₂-sensing neurons increases the expression of Ca²⁺ homeostasis-modulating genes.

In response to elevated neuronal activity, mammalian neurons induce the expression of hundreds of activity-regulated genes (ARGs), which are thought to orchestrate transcription-dependent neuronal plasticity (Tyssowski et al., 2018). By looking for orthologs in C. elegans of genes in mouse cultured cortical neurons and the visual cortex (Tyssowski et al., 2018), we identified putative ARGs in C. elegans O₂-sensing neurons (Supplementary file 6). Among these ARGs, a subset of delayed primary response genes (dPRGs), whose expression is induced in response to sustained neural activity, was significantly enriched in the set of genes upregulated in high-activity oxygen-sensing neurons from 7-day-old C. elegans when compared to low-activity neurons from worms of the same age (Figure 4D). Our data suggest that functional ARGs are conserved between C. elegans and mammals. Taken together, our findings suggest that neuronal activity regulates gene expression in the O₂-sensing neurons during ageing.

We next identified the gene expression changes with age in O₂-sensing neurons, and asked whether and how neuronal activity alters the age-related transcriptional change. In both low and high O₂ culture conditions, signifying low and high neural activity states, respectively, expression of about 5000 neural genes, ~25% of the genome, significantly changes with age, with about half of them up- or downregulated. In the whole-worm samples, ~2000 genes changed with age, with about half of them up- or downregulated (Supplementary file 1). In the O₂-sensing neurons, ~60% of the genes differentially expressed with age were shared between the high and low activity states in 21% and 7% O₂ cultured animals, respectively (Figure 5A,D). There was also a substantial overlap between the DEGs by age in the whole-worm samples in 21% and 7% O₂ cultured animals, and between neuronally and whole-worm differentially regulated gene sets (Figure 5A and Figure 5—figure supplement 1A). This suggests that those overlapping genes constitute a core set of genes changing expression with age. We thus cross-checked the transcriptome changes with age with the ‘MetaWorm’ dataset (Supplementary file 7) of C. elegans ageing signature genes (Tarkhov et al., 2019) by gene set enrichment analysis, and found that the subset of MetaWorm genes upregulated with age shows significant enrichment in age-related upregulated genes in both O₂-sensing neurons and whole-worm samples; conversely, the subset of MetaWorm genes downregulated with age is significantly enriched in age-related downregulated genes in both neuronal and whole-worm samples (Figure 5—figure supplement 1B, supplementary file 7). Our data thus likely contain many shared gene expression changes that represent a universal transcriptomic signature of ageing regardless of tissue type and culture condition.

Next, we focused on gene expression changes with age that were specific to the O₂-sensing neurons. Consistent with previous findings in the ageing brains of Drosophila, killifish, rat, and mouse (Baumgart et al., 2014; Davie et al., 2018; Ori et al., 2015; Ximerakis et al., 2019), our data show that translational and ribosome biogenesis related genes are upregulated in aged neurons (Figure 5C), comprising the great majority of GO terms upregulated with age. In contrast, translation-related genes in the whole worm transcriptome samples showed consistently decreased expression with age (Figure 5C). By pathway analysis (Supplementary file 8), eIF2 and eIF4 signaling, required for translation initiation, is upregulated in aged neurons regardless of neuronal activity (Figure 5E). Conversely, genes required for neural function are widely downregulated in aged O₂-sensing neurons regardless of activity state. GO categories downregulated with age were dominated by neuronal activity-related terms such as adult locomotory behavior, regulation of membrane potential, G-protein-coupled receptor signaling pathway, neuropeptide signaling or chemical synaptic transmission (Figure 5B).

To explore why neural plasticity is selectively lost in the ageing O₂-sensing neurons experiencing chronically high activity, we then specifically contrasted how the neural transcriptome changes differently with age in animals cultured in 21% or 7% O₂ environments, using gene set enrichment and pathway analysis.

We first examined the genes constituting the C. elegans O₂-sensing machinery by manually curating a gene set based on the literature (Supplementary file 6). We found that expression of the O₂-sensing machinery in neurons declines with age in line with neural function-related GO terms in general. However, their expression is upregulated with high neuronal activity in day 7 adults’ O₂-sensing neurons, but not in day 1 adults (Figure 4D, Supplementary file 5). The aryl hydrocarbon receptor signaling pathway, which controls expression of the soluble guanylate cyclase genes essential for oxygen sensing, is increased with age at high activity, but decreased at low activity (Figure 5F). The activity-related genes we identified in C. elegans showed a significant decrease in dPRG transcript abundance in low-activity O₂-sensing neurons with age, whereas dPRG expression levels in the high-activity neurons were sustained with age (Supplementary file 5). The manually curated set ‘genes linked to Ca²⁺ signaling’, which primarily represents Ca²⁺ homeostasis-regulating genes, was also specifically down-regulated in low-activity O₂-sensing neurons and showed no significant change with age in high-activity neurons (Supplementary file 5). Oxidative stress in neurons generally increases with age and causes their greater susceptibility to oxidative insults and disease (Castelli et al., 2019). Gene set enrichment analysis shows that oxidative stress-related genes and related GO terms were specifically upregulated with age in the high-activity O₂-sensing neurons, whereas only two of these GO terms, such as cell redox homeostasis, showed enrichment with age in the low-activity neurons as well (Figure 5C and Supplementary file 5); no such increase in any of these GO terms was seen in the whole worm transcriptomes. These results suggest that high-activity neurons devote resources to sustaining tonic neuronal excitation in response to high sensory stimulation, as well as dealing with the resulted high neuronal oxidative stress during ageing.

Conversely, GO terms related to learning and signaling downstream from Ca²⁺ (Ca²⁺-mediated signaling) are significantly downregulated with age only in the high-activity but not the low-activity O₂-sensing neurons (Figure 5B, Supplementary file 5). Pathway analysis shows a stronger downregulation of overall calcium signaling and synaptogenesis signaling pathway with age in the high activity state (Figure 5E). These findings suggests that low-activity O₂-sensing neurons kept a better ‘plasticity reserve’ than high-activity neurons during ageing.

Taken together, our data indicate that neural activity alters age-dependent gene expression changes in the O₂-sensing neurons. Specifically, chronically high neuronal activity leads to the continued high expression of genes that encode the core O₂-sensing machinery, Ca²⁺ homeostasis regulating genes, oxidative stress-related factors, as well as activity-regulated genes, while the expression of learning-related genes and Ca²⁺-mediated signaling factors declined. This suggests that in high activity neurons, resources are directed toward maintaining a high firing rate, leaving limited resources for neural plasticity during ageing.

To examine whether the activity-dependent regulation of transcription in O₂-sensing neurons modulates O₂-evoked locomotory responses and O₂ experience-dependent plasticity, we performed cell-specific RNAi in the O₂-sensing neurons to knock down the expression of some of these genes using a transgenic strain that expresses the RNA transporter SID-1 only in these cells, which avoids the RNAi knockdown spreading to other cells or tissues. We first selected 11 genes from top-ranked genes by false discovery rate (FDR) for differential expression in the day 1 adult neuronal transcriptomes, of which four were down- and seven upregulated by neural activity state. Knockdown of 6 of the 11 genes affected either O₂-evoked behavior and/or experience-dependent plasticity in 1-day-old adults (Figure 6—figure supplement 1). Specifically, knockdown of cebp-1 (C/EBP transcription factor), ptp-4 (protein tyrosine phosphatase receptor) or unc-86 (POU homeodomain transcription factor) in the O₂-sensing neurons significantly reduced O₂ experience-dependent plasticity (Figure 6—figure supplement 1G,H,I,M,N). nlp-11 neuropeptide, ptp-4 or sol-1 (glutamate receptor subunit) knockdown altered both the speed responses to the stepwise changes in O₂ concentration of animals kept at 21% O₂ and of animals shifted from 21% to 7% O₂ (Figure 6—figure supplement 1A,H,K,N). Knockdown of nhr-14 (nuclear hormone receptor) altered speed responses to O₂ only in animals kept at 21% O₂ throughout (Figure 6—figure supplement 1D,N). Together, these results demonstrate that genes differentially regulated by neural activity in young adults’ O₂-sensing neurons play a role in the behavior and plasticity evoked by oxygen stimuli.

We then validated the most significant differently expressed genes by neural activity in day 7 neurons. They are kdin-1, which encodes the orthologue of the scaffold protein ARMS/kinase D interacting substrate 220 (KIDINS220), skn-1, which encodes the C. elegans orthologue of the transcription regulator Nrf2, and cfp-1, encoding the orthologue of the transcriptional activator CXXC-type zinc finger protein 1. ARMS/KIDINS220 has been implicated in regulating neural excitability and Ca²⁺ homeostasis, and SKN-1/Nrf2 are key regulators of oxidative stress responses and metabolism (Blackwell et al., 2015; Jaudon et al., 2020). Neural expression of all three genes significantly declines with age in 7% O₂ culture but not at 21% O₂ (Supplementary file 1). We knocked down kdin-1, skn-1 and cfp-1 by cell-specific RNAi in the O₂-sensing neurons only, and tested O₂ experience-dependent plasticity in day 2 and day 7 adult worms. kdin-1 and skn-1 knockdown did not affect O₂-evoked locomotory responses or plasticity in day 2 adults, whereas cfp-1 RNAi increased O₂ responses in day 2 adult animals cultured at high O₂, but not in those shifted to 7% O₂ for 12 hr, significantly reducing plasticity (Figure 6A,C). RNAi of all three genes altered O₂-evoked responses in 21% O₂-cultured day 7 adults (Figure 6B). In particular, RNAi of kdin-1 restored O₂ experience-dependent plasticity in 7-day-old 21% O₂ cultured animals, which displayed the same levels of plasticity as at day 2, whereas plasticity significantly declined in control animals (Figure 6B,C). RNAi of skn-1 reduced responses to high O₂ in 7-day-old animals kept at 21% O₂ throughout, but did not affect the responses of those shifted to 7% O₂. cfp-1 RNAi moderately increased O₂ responses of 7-day-old animals shifted to 7% but did not affect those of animals kept at 21% throughout. In contrast, plasticity in 7-day-old animals cultured at 7% O₂ was not significantly affected by the knockdown of kdin-1, skn-1 or cfp-1, although O₂ responses of animals kept at 7% throughout were altered (Figure 6—figure supplement 2). These results suggest that genes whose expression is selectively sustained in ageing neurons kept at a high activity state contribute to the loss of O₂ experience-dependent plasticity with age. kdin-1 has a very specific effect on the decline of plasticity without altering O₂ responses, whereas skn-1 and cfp-1 affect plasticity indirectly through altered O₂ responses in non-shifted animals. Thus, our data indicate that activity-induced neuronal transcriptional alteration plays a profound role in experience-dependent plasticity and its decline with age.

Our transcriptomic profiling of the O₂-sensing neurons suggests that neural activity modulates the changing expression of Ca²⁺ signaling factors with age. Altered Ca²⁺ signaling has been implicated in impaired neural plasticity and age-related cognitive decline (Foster, 2007; Palop et al., 2007). We therefore hypothesized that Ca²⁺ signaling may be an important intermediary between activity state and the decline of plasticity in neurons with age. To test this and gain an understanding of the molecular processes involved, we characterized how four key Ca²⁺-linked genes, cmd-1 calmodulin, tax-6 calcineurin, unc-68 ryanodine receptor and the ncx-5 K⁺-dependent Na⁺/Ca²⁺ exchanger are involved in the age-related decline of behavioral plasticity in oxygen responses. All four genes show age-related downregulation in transcription in our data collected from the O₂-sensing neurons (Supplementary file 1).

The calcium-binding protein calmodulin plays a central role in the regulation of activity-dependent plasticity (Ma et al., 2014) and is highly conserved between C. elegans and humans with 98% amino acid sequence identity. Due to its pleiotropic roles in many physiological processes, we used cell-specific RNAi to knock down cmd-1 in URX, AQR and PQR only, and subjected young (day 1) or old (day 7) animals to the environmental shifts detailed in Figure 1 to test for locomotory speed and experience-dependent plasticity in response to the stepwise changes in oxygen concentrations detailed in Figure 1—figure supplement 1B. In 1-day-old adults cultured at 21% O₂, cmd-1 knockdown strongly reduced the locomotory response to 21% O₂ after downsteps of O₂ concentrations from 21% to 7% O₂, but did not diminish the initial locomotory response to high O₂ (Figure 7A). In animals previously cultured at 7% O₂ throughout and also in those shifted from 7% to a 21% O₂ environment overnight, cmd-1 knockdown reduced locomotory responses to high O₂ (Figure 7B). In contrast, 7-day-old cmd-1 RNAi adults responded normally to 21% O₂ after a period of low O₂ exposure (Figure 7E,F). These results suggest that CMD-1 is essential in young neurons but not in old neurons to recover from a low activity state to promote a strong avoidance response to high O₂.

Cell-specific knockdown of cmd-1 in the O₂-sensing neurons did not affect experience-dependent plasticity in young adults, but significantly restored experience-dependent plasticity in day 7 adults cultured at 21% O₂ (Figure 7C,G), where the O₂-sensing neurons had been chronically stimulated. Lowering cmd-1 levels did not further improve the response and plasticity of animals long-term cultured at 7% O₂ (Figure 7F,G). This suggests that chronically high neuronal activation specifically acts through CMD-1 to result in plasticity loss with age, and that this pathway is not activated when neurons are kept at a low activity state. Lowering neuronal CMD-1 levels can counteract the plasticity decline with age.

To further understand how signaling by Ca²⁺ and calmodulin regulates neural plasticity changes with age, we investigated the Ca²⁺/calmodulin–dependent serine/threonine protein phosphatase calcineurin, the only Ca²⁺-activated protein phosphatase in the brain and a major regulator of synaptic transmission, neuronal excitability and Ca²⁺ homeostasis. Calcineurin is implicated in regulating learning and memory in both mammals and C. elegans (Kuhara and Mori, 2006; Mansuy, 2003). O₂-sensing neuron-specific knockdown of tax-6 in 1-day-old adults altered the response pattern to the stepwise changes in O₂ concentration both of animals cultured at 21% O₂ throughout and of those shifted to 21% O₂ 12 hours before the assay, while animals kept at 7% O₂ prior to the experiment behaved normally. Specifically, worms exposed to 21% O₂ were less able to slow down at intermediate O₂ (between 19% and 14% O₂) (Figure 7A,B). As a result, tax-6 RNAi eliminated plasticity in young animals in the 21% to 7% O₂ shift assay and significantly reduced plasticity in the 7% to 21% O₂ shift experiment (Figure 7D).

Like their 1-day-old adult counterparts, 7-day-old adults cultured at 21% O₂ with tax-6 knocked down showed no plasticity, with no significant behavioral difference to controls of the same age (Figure 7E,H). Remarkably, 7% O₂ cultured animals subject to cell-specific tax-6 RNAi show significantly less age-related decline of O₂-evoked speed responses. Specifically, locomotory speed at high O₂ levels increased compared to controls for animals cultured at 7% O₂ throughout and in those shifted to 21% O₂12 hr prior to the assay (Figure 7F). Plasticity was retained with age in the same way as in 7-day-old control animals cultured at 7% O₂ (Figure 7H). The beneficial effect of tax-6 knockdown was only seen in animals with low neuronal activity in O₂-sensing neurons, suggesting that calcineurin signaling contributes to the decline of O₂-evoked behavior with age under physiological conditions of stimulation. Moreover, since knockdown of tax-6 did not restore plasticity in 7-day-old 21% O₂ cultured animals, the CMD-1-induced plasticity loss is likely to be acting through a calcineurin-independent pathway. Overall, tax-6/calcineurin regulates the tuning of locomotory speed responses to stepwise changes in O₂ concentrations and O₂ experience-dependent plasticity in young adults, and promotes the decline of locomotory responses with age under physiological conditions.

The ryanodine receptor UNC-68 is an endoplasmic reticulum Ca²⁺ release channel and involved in sustaining tonic sensory responses in the O₂-sensing neurons (Busch et al., 2012). Cell-specific knockdown of unc-68 had a similar but weaker phenotype as tax-6 RNAi in 7-day-old animals, where it increased locomotory responses to O₂ in 7% but not 21% O₂ cultured animals and likewise did not alter plasticity compared to controls (Figure 7—figure supplement 1).

ncx-5 encodes a member of the K⁺-dependent Na⁺/Ca²⁺ exchanger family which plays important roles in sustaining long-term Ca²⁺ homeostasis in many biological processes (Sharma et al., 2013). It is orthologous to mammalian NCKX2, which is broadly expressed in the brain and required for long-term potentiation, with NCKX2 null mice showing deficits in learning and memory tasks (Li et al., 2006). In line with a previous report (Sharma et al., 2013), our O₂-sensing neuron-specific transcriptomics data show that ncx-5 is highly enriched in O₂-sensing neurons (Supplementary file 1). High-activity neurons have significantly higher ncx-5 expression than low-activity neurons in day 7 adults, suggesting an involvement in plasticity changes with age. We therefore performed cell-specific RNAi to knock down the expression of ncx-5 in the O₂-sensing neurons.

In young adults, knockdown of ncx-5 does not affect the behavior of animals cultured at 21% O₂ (Figure 7A), and has a slight but statistically significant effect on the responses to the stepwise changes in O₂ levels of animals shifted from 7% to 21% O₂ overnight, suggesting that ncx-5 is involved in adaptation to high neuronal activity (Figure 7B). The knockdown of ncx-5 in O₂-sensing neurons did not affect experience-dependent plasticity of young adults, but significantly reduced it in ageing animals (Figure 7D,H). Day seven animals shifted from 21% to 7% O₂ for 12 hr significantly reduced rather than increased locomotory responses to O₂, resulting in a reversed plasticity index (Figure 7E,H). ncx-5 RNAi in 7-day-old animals cultured at 7% O₂ significantly decreased behavioral plasticity by reducing oxygen responses in animals kept at 7% and increasing responses in animals shifted to 21% O₂ for 12 hr (Figure 7F,H). These results suggest that NCX-5 plays an essential and specific role in sustaining experience-dependent plasticity during ageing and in enabling older animals to adapt to new environments.

We hypothesized that the accelerated behavioral plasticity decline in ageing animals with ncx-5 knockdown may be a consequence of altered O₂-evoked Ca²⁺ responses in the O₂-sensing neurons of young adults. To test this, we performed Ca²⁺ imaging of the URX O₂-sensing neurons in both 2-day and 7-day-old adults (Figure 8, Figure 8—figure supplement 1A–D). Cell-specific knockdown of ncx-5 in the O₂-sensing neurons did not generally alter the response magnitude of the neurons to high O₂ stimuli in either young or old adults (Figure 8—figure supplement 1C,D). Intriguingly, although ncx-5 knockdown did not change the behavioral plasticity of 2-day-old adults, it did reduce plasticity of neuronal Ca²⁺ responses: ncx-5 RNAi animals shifted from 21% to 7% O₂ for 12 hr increased their Ca²⁺ responses to O₂ stimuli, compared to those cultured at 21% O₂ throughout, less strongly than control animals did upon the same treatment (Figure 8A,B, Figure 8—figure supplement 1A,B). Also in the converse experiment, control animals shifted from 7% to 21% O₂ for 12 hr very significantly reduced their responses to high and intermediate O₂ stimuli, whereas with ncx-5 knocked down, URX Ca²⁺ responses to the O₂ stimuli were not significantly different between animals shifted from 7% to 21% O₂ and those cultured at 7% O₂ throughout (Figure 8C,D, Figure 8—figure supplement 1A,B). Therefore, the ability of the O₂-sensing neurons to adapt to changing O₂ environments requires Ca²⁺ clearance by NCX-5 already in young adults. The accumulated effect of lack of NCX-5 during ageing may be the cause of accelerated behavioral plasticity decline. Furthermore, we found that knockdown of ncx-5 in URX elevates baseline [Ca²⁺] at 7% O₂ in low O₂ cultured animals, where the neuron is inactive (Figure 8E). This suggests that impaired Ca²⁺ homeostasis at rest may be responsible for accelerating the plasticity decline. To further explore this, we tested if knockdown of cmd-1, which slows plasticity decline, alters Ca²⁺ homeostasis as well, by performing Ca²⁺ imaging in URX. Ca²⁺ levels were indeed significantly reduced when cmd-1 was knocked down in 1-day-old 21% O₂-cultured animals (Figure 8E, Figure 8—figure supplement 1E). Together, these results suggest that the control of Ca²⁺ homeostasis in young adults by NCX-5 in low-activity neurons and CMD-1 in high-activity neurons determines if behavioral plasticity is retained or lost in aged animals.

Previous studies in C. elegans have highlighted the role of longevity regulation pathways, in particular of the insulin/insulin-like growth factor signaling(IIS)/FOXO pathway, in altering cognitive performance with age (Kaletsky et al., 2016; Kauffman et al., 2010; Murakami et al., 2005). To test if the IIS/FOXO signaling pathway regulates plasticity decline in the O₂-sensing neurons, we performed cell-specific RNAi knockdown of daf-16 in O₂-sensing neurons, but did not observe any significant effect (Figure 7—figure supplement 2A,B), suggesting that IIS/FOXO signaling is not involved in the O₂ experience-dependent decline of plasticity with age. The TAX-6 calcineurin phosphatase is known to modulate C. elegans longevity via the CREB-regulated transcriptional coactivator CRTC-1 (Mair et al., 2011). However, knockdown of crtc-1 in the O₂-sensing neurons affects neither O₂-evoked locomotory responses nor experience-dependent plasticity (Figure 7—figure supplement 2C,D). Thus, these data suggest that neuronal activity state affects age-related plasticity decline independent of lifespan-regulating pathways.

Overall, our results indicate that factors linked to neuronal Ca²⁺ signaling govern the decline of the experience-dependent plasticity in O₂ responses with age.

C. elegans were grown at 20°C and kept on NGM plates seeded with E. coli OP50 bacteria as food source as previously described (Brenner, 1974). Previously described strains are AX204 (npr-1(ad609) X) and AX1864 (npr-1(ad609) X; dbEx(pgcy-32::YC3.60)) (Busch et al., 2012). KL24 (npr-1(ad609) X; cipEx41(pgcy-32::ChRcodopt-mCitrine)) was generated by ballistic gene transfer of a pExpr plasmid containing a Channelrhodopsin-mCitrine fusion gene, both codon optimized for C. elegans, under the gcy-32 promoter for expression in the O₂-sensing neurons. KL71 (sid-1(pk3321) him5(e1490)V; npr-1(ad609)X) was generated by crossing AX204 with TU3596 (sid-1(pk3321) him-5(e1490) V; lin-15B(n744) X). KL123 (npr-1(ad609) X; iaIs25[Pgcy-37::GFP + unc-119(+)]) was generated by crossing AX204 with ZG610 (iaIs25[Pgcy-37::GFP + unc-119(+)]). KL92 (sid-1(pk3321) V.; npr-1(ad609) X.; cipIs7[pgcy-32::sid-1::SL2-GFP] V) was generated by miniMos injection (Frøkjær-Jensen et al., 2014) as follows: A sid-1 full length genomic fragment was cloned into the miniMos plasmid pCFJ907, under the gcy-32 promoter that drives expression in the O₂-sensing neurons. The pExpr sid-1 was injected at 10 ng/µl to AX204, together with co-injection markers pGH8 (Prab3::mCherry::unc-54 3’UTR) at 10 ng/µl, pCFJ90 (Pmyo-2:mCherry:unc-54 3’UTR) at 2.5 ng/µl, pCFJ104 (Pmyo3::mCherry::unc-54 3’UTR) at 10 ng/µl, pRF4 at 20 ng/µl, and pCFJ601 (Peft-3::Mos1 transposase) at 50 ng/µl. 500 µl 25 mg/ml (diluted in water) G418 antibiotic was directly added to the 6 cm NGM plate the day after injection, and the worms were grown at 25°C. Inserted transgenes were identified by screening for surviving animals that lost the co-injection markers as seen by the absence of fluorescent protein expression in neurons and muscle. For identifying the integration sites in the chromosome, inverse PCR was performed as described (Frøkjær-Jensen et al., 2014). This strain was then crossed with KL71. KL269 (npr-1(ad609)X; cipIs35[pgcy-32::HisCl1 codon optimized::SL2 GFP]) was also generated by miniMos injection, and a codon-optimized single copy of HisCl1 was inserted. KL217 (npr-1(ad609) X; cipIs31[pgcy-32::cmd-1 sense fragment::let-868 3'UTR; pgcy-32::cmd-1 antisense fragment::let-868 3'UTR; ccRFP]) was generated by injecting sense and antisense cmd-1 fragments (40 ng/µl each) cloned from genomic DNA, co-injected with punc-122::RFP at 20 ng/µl followed by UV integration. KL317 (sid-1(pk3321) V; npr-1(ad609) X; cipIs7[pgcy-32::sid-1::SL2-GFP] V; dbEx(pgcy-32::YC3.60 lin-15(+))) was generated by crossing KL92 to AX1864. KL325 (dvIs19[(pAF15) gst-4p::GFP::NLS] III; npr-1(ad609) X) was generated by crossing SPC167 (dvIs19[(pAF15) gst-4p::GFP::NLS] III; skn-1(lax120) IV) to AX204.

Aerotaxis assays were done essentially as described previously (Gray et al., 2004). For each assay, 150–200 L4s were picked the night before the day 1 adults assay, or then transferred to newly seeded plates every 2 days for the 7-day-old adults assay. For the assay, animals were picked to 9 cm NGM plates seeded two days before assay to form a thin rectangular bacterial food lawn, and a 30 × 15×0.2 mm polydimethylsiloxane (PDMS) microfluidic assay chamber placed over the animals, connected to two syringes at each end to pump 7% O₂ or 21% O₂, respectively, into the chamber at a pumping rate of 1.5 ml/min. After 30 min, the location of animals in the chamber was recorded with a Point Grey Grasshopper3 CMOS camera mounted on a Leica stereo microscope for distribution analysis and chemotaxis index quantification. The aerotaxis index is defined as (number of animals at high O₂ area – number of animals at low O₂ area)/total animal number.

Ratiometric Ca²⁺ imaging was performed on a Nikon Eclipse Ti-E Inverted compound microscope with a CoolLED pE4000 fluorescence light source using a 40x Apochromat λS LWD water immersion lens, following an established protocol (Busch et al., 2012).For imaging in day one adults, L4 animals were picked the day before imaging. For day seven adults imaging, L4 animals were picked and transferred to fresh culture plates every 2 days. For imaging, 2% agarose pad (diluted in M9) was used. Individual animals were glued on pad with Dermabond Topical Skin Adhesive, leaving the nose free of glue. A drop of food was put next to the nose, M9 buffer was applied alongside the worm, and glued worms were covered with a polydimethylsiloxane (PDMS) microfluidic chamber. Gases at specific O₂ concentrations were generated by a gas mixer (MCQ Gas Blender 100) and collected into individual sealable syringes. 435 nm fluorescence light intensity was set at 1% and exposure time was set at 300 ms. The filter set used was Chroma 59017 for dual CFP/YFP recording. An emission splitter (Cairn Research) was used to separate the cyan and yellow emission light with an FITC/Alexa Fluor 488/Fluo3/Oregon Green dichroic, ET480/40 m filter for CFP emission and ET535/30 m filter for YFP emission. Images were captured with an ORCA-Flash 4.0 V2+ camera at 2 frames per second. Imaging data were analyzed using Neuron Analyzer, a customwritten Matlab program, as previously described (Laurent et al., 2015).

RNAi by feeding was performed as previously described (Kamath and Ahringer, 2003) with slight modifications. RNAi plates were supplemented with 100 μg/ml carbenicillin and 1 mM IPTG. Plates were poured freshly and stored at 4°C for no more than 1 month. Bacterial strains (RNase-deficient E. coli HT115) expressing dsRNA were obtained from the Ahringer library and confirmed by sequencing. For the unc-68 RNAi clone which is not available from the library, an unc-68 gene fragment cloned from genomic DNA was inserted into L4440 plasmid by double digestion using XhoI and Not I, and this plasmid was then transformed into HT115 bacteria. Primers used for unc-68 fragment PCR cloning from genomic DNA were CGCGCTCTCGAGGCTACATGGTGATTGCCTCA and TAATTTGCGGCCGCTGCTCCCATTT CACATGAGT. Bacteria were grown on LB plates with carbenicillin and tetracycline at 37°C overnight. Bacterial liquid culture with 100 μg/ml carbenicillin was grown at 37°C with shaking at 200 rpm for 6–8 hr. The liquid culture was seeded on the plates, left to dry at room temperature for 2 days. L4 stage animals were picked to the RNAi plate. After 48 hr, adults were transferred to the second RNAi plate seeded with the same bacteria, and their progeny was used for the assays. For RNAi by feeding experiments in young adults we assayed 2-day-old adults instead of day 1 adults because RNAi knockdown is more efficient at this age (data not shown). KL92 was used for behavioral assays and KL317 for Ca²⁺ imaging; in both strains, expression of the RNA transporter sid-1 was limited to the O₂-sensing neurons by transgenic expression from the gcy-32 promoter to achieve cell-specific knockdown (Calixto et al., 2010). All the RNAi feeding animals were cultured at 20°C. The RNAi clones used for RNAi feeding experiments are listed in Supplementary file 9.

Strain KL123 was used for cell-specific RNA-sequencing. URX, AQR and PQR were strongly labeled with GFP in this strain in both young and old adults (Figure 4—figure supplement 1A,B). Weaker fluorescent labeling was also occasionally detected in the AVM mechanosensory neuron and the ASI chemosensory neurons (Figure 4—figure supplement 1A,B and data not shown). AVM and ASI did not respond to oxygen stimuli (Figure 4—figure supplement 1C–E), making it unlikely that they would contribute to activity-dependent transcriptional changes under different oxygen conditions. To avoid contamination of the 7-day adult samples with RNA of their offspring, we fed worms with bacteria inducing RNAi against mex-3, which causes embryonic lethality with near 100% penetrance (Kamath et al., 2001). We confirmed that the mex-3 RNAi treatment had no impact on either experience-dependent plasticity or the differential decline of plasticity in 7-day-old animals (Figure 4—figure supplement 1F–H).

Animals were synchronized by bleaching for two rounds. Eggs were plated on OP50 plates and L3 stage animals were transferred to mex-3 RNAi plates to avoid egg hatching. To obtain day 7 adults, adults at day 2 were washed off by M9 and transferred to new mex-3 RNAi plates for further cultivation, and at day 4 were washed off and transferred to new OP50 plates (Figure 4A).

Cell dissociation and FACS sorting procedures were performed as previously described (Kaletsky et al., 2016) with modifications. For each sample, 10,000–15,000 worms were washed off by M9 and transferred to 1.5 ml Eppendorf tubes, followed by washing with M9 buffer five times. 300 µl lysis buffer (200 mM DTT, 0.25% SDS, 20 mM HEPES pH 8.0, 3% sucrose) was added and worms were incubated for 5–6 min until the head of worms became blunt. Worms were then washed rapidly with M9 buffer five times. 200 µl room temperature pronase (15 mg/ml, dissolved in 340mOsm L15) was added and worms were incubated for 8–15 min. During pronase incubation, a pellet pestle motor was used every 2–3 min for 10 s each time to dissociate and homogenize worm cells. After each round of using the pellet pestle motor, dissociation was monitored by checking 2 µl of the pronase suspension under the microscope. When the majority of big worm chunks were no longer visible, 800 µl ice-cold BSA (0.2%)/L-15 (340mOsm) was added to stop the reaction. At this point, 10 µl of the dissociated worm solution of each sample was transferred to 300 µl RNA extraction buffer (PicoPure RNA isolation kit (Arcturus Bioscience)) and kept on ice. This solution was subsequently used to determine the whole-worm transcriptome. Cells were pelleted by centrifugation at 9600 g for 5 min at 4°C and then resuspended using 1 ml ice-cold BSA (0.2%)/L-15 (340mOsm), and immediately passed through a 10 µm filter for sorting.

For FACS sorting, age-matched AX204 worms were used as negative control for setting gates. For KL123 (npr-1(ad609) X; iaIs25[Pgcy-37::GFP + unc-119(+)]), 5’000–20’000 GFP positive cells from each sample were sorted into 300 µl extraction buffer for RNA extraction using the PicoPure RNA isolation kit (Arcturus Bioscience). RNA samples with a RIN score >7.0 were used for cDNA synthesis. cDNA was generated using the Ovation RNA-seq system V2. Libraries were generated using TruSeq DNA Nano gel free library preparation kit. NovaSeq 50PE was used for all samples. Four independent biological replicates per condition were sequenced.

Reads were trimmed using Cutadapt (version cutadapt-1.9.dev2) (Martin, 2011). Reads were trimmed for quality at the 3’ end using a quality threshold of 30 and for adapter sequences of the TruSeq DNA kit (AGATCGGAAGAGC). The reference used for mapping was the Caenorhabditis elegans (build WBcel235) genome from Ensembl. Reads were aligned to the reference genome using STAR (version 2.5.2b) (Dobin et al., 2013) specifying paired-end reads and the option --outSAMtype BAM Unsorted. Reads were assigned to features of type ‘exon’ in the input annotation grouped by gene_id in the reference genome using featureCounts (version 1.5.1) (Liao et al., 2014). Genes with biotype rRNA were removed prior to counting. featureCounts assigns counts on a ‘fragment’ basis as opposed to individual reads such that a fragment is counted where one or both of its reads are aligned and associated with the specified features. Strandedness was set to ‘unstranded’ and a minimum alignment quality of 10 was specified. The raw counts were filtered to remove genes consisting predominantly of near-zero counts, filtering on counts per million (CPM) to avoid artefacts due to library depth. Differential gene expression analysis was performed using DESeq2 (Love et al., 2014) (version 1.18.1 with R version 3.4.2) at a default false discovery rate (FDR) of 0.1 as used in previous studies (Thomson et al., 2017), and additionally with a minimal fold change of 1.5 to exclude smaller changes in transcript abundance mostly seen in the whole-worm samples. Pathway analysis was performed using Ingenuity Pathway Analysis (Qiagen). Gene set enrichment analysis was performed in GSEA (Broad Institute) and WormBase Tissue Enrichment Analysis (Angeles-Albores et al., 2016). GSEA version 2–2.2.1 was used and gene sets containing more than 500 genes or less than five genes were excluded. Only gene sets with p<0.05 and q < 0.25 were considered as significantly enriched, as done in previous studies (Ximerakis et al., 2019). The normalized enrichment score (NES) was defined previously (Subramanian et al., 2005).

KL325 was used for gst-4p::GFP fluorescence intensity quantification. NAC (5 mM) treatment was initiated from hatching. Worms were immobilized by 100 mM NaN₃ on a 2% agarose pad. Images were captured on a Zeiss Axio Imager Z1m with a 10x objective, and images analyzed by ImageJ. Specifically, a threshold was applied to each picture to highlight the fluorescent whole worm and the mean value of this highlighted area was used for statistical analysis.

Unpaired t-test with Holm-Sidak correction for multiple comparisons was performed to examine the plasticity difference between 21% and 7% O₂ animals at day 4, day 7, and day 10 of adulthood. For O₂-evoked speed responses, mixed model ANOVA was performed to examine the effect of condition shift (21% to 7% O₂ or 7% to 21% O₂ the night before assay) and [O₂] steps (21%, 19%, 17%, 14%, 11%, 7%, and 21% O₂) on worms’ speed responses. Mann-Whitney U test was used for aerotaxis index comparison. For O₂-evoked speed responses in RNAi feeding assays, mixed model ANOVA with Holm-Sidak test was conducted to examine the effect of gene knockdown and [O₂] steps on speed responses. O₂-evoked Ca²⁺ responses were analyzed using mixed model ANOVA, and the time period used for analysis is indicated in corresponding figure legends. Plasticity indexes were compared by unpaired t-test. One-way ANOVA with Holm-Sidak test was used for fluorescence intensity comparison. GraphPad Prism was used for statistical analysis.