Hypoxia-inducible factor cell non-autonomously regulates C. elegans stress responses and behavior via a nuclear receptor

The HIF (hypoxia-inducible factor) transcription factor is the master regulator of the metazoan response to chronic hypoxia. In addition to promoting adaptations to low oxygen, HIF drives cytoprotective mechanisms in response to stresses and modulates neural circuit function. How most HIF targets act in the control of the diverse aspects of HIF-regulated biology remains unknown. We discovered that a HIF target, the C. elegans gene cyp-36A1, is required for numerous HIF-dependent processes, including modulation of gene expression, stress resistance, and behavior. cyp-36A1 encodes a cytochrome P450 enzyme that we show controls expression of more than a third of HIF-induced genes. CYP-36A1 acts cell non-autonomously by regulating the activity of the nuclear hormone receptor NHR-46, suggesting that CYP-36A1 functions as a biosynthetic enzyme for a hormone ligand of this receptor. We propose that regulation of HIF effectors through activation of cytochrome P450 enzyme/nuclear receptor signaling pathways could similarly occur in humans.


Introduction
The capacity to sense and respond to oxygen deprivation, or hypoxia, is crucial to normal physiological function and survival of aerobic organisms, which require oxygen to perform respiration and generate energy in the form of ATP. The fundamental importance of a mechanism to detect and react to low oxygen is reflected in the presence of a conserved hypoxia-response pathway in most animal cells. This pathway consists of the transcription factor HIF, or hypoxia-inducible factor, and its negative regulator, the prolyl hydroxylase EGLN, which together mediate a diversity of metabolic and physiological adaptations to hypoxia. The three human EGLNs, which were identified as homologs of the C. elegans protein EGL-9, function as oxygen sensors. In the presence of oxygen, EGLN hydroxylates the HIF a-subunit (HIFa), allowing the von Hippel-Lindau (VHL) E3 ubiquitin ligase to promote HIFa degradation (Maxwell et al., 1999;Jaakkola et al., 2001;Epstein et al., 2001;Ivan et al., 2001). In conditions of low oxygen, HIFa is stabilized and acts with its partner HIFb to drive adaptations to hypoxia through activation of its transcriptional targets (Kaelin and Ratcliffe, 2008;Semenza, 2011;Wang et al., 1995).
The canonical function of the EGLN/HIF pathway is to regulate genes that either increase oxygen availability, for example by promoting erythropoiesis and angiogenesis, or reduce the cellular requirement for oxygen, for example by driving a shift from oxidative phosphorylation to glycolytic metabolism. However, a growing body of work has found roles for the EGLN/HIF pathway in controlling other aspects of animal physiology and behavior. HIF promotes the response to numerous stressors, including infection, proteotoxicity, and oxidative stress (Palazon et al., 2014;Schito and Rey, 2018;Nakazawa et al., 2016;Powell-Coffman, 2010). HIF activation is associated with increased lifespan in C. elegans (Mehta et al., 2009;Zhang et al., 2009;Chen et al., 2009;Lee et al., 2010); this longevity phenotype likely stems from the improved stress resistance associated with HIF activity, as is often the case for pathways regulating longevity (Leiser et al., 2015;Shore and Ruvkun, 2013). The EGLN/HIF pathway also modulates several behaviors of C. elegans following prolonged hypoxia exposure, suggesting a role for this pathway in tuning neural circuit function (Chang and Bargmann, 2008;Pocock and Hobert, 2010;Ma et al., 2012). The mechanisms by which HIF mediates these non-canonical physiological and behavioral changes remain poorly defined.
Here we report the discovery of an endocrine signaling pathway that regulates multiple aspects of physiology and behavior downstream of HIF in C. elegans. From a genetic screen for suppressors of an egl-9(lf) mutant behavioral defect, we identified a cytochrome P450 gene, cyp-36A1, that is required for modulation of egg-laying behavior by the egl-9/hif-1 pathway. cyp-36A1 is transcriptionally upregulated in hypoxia or egl-9(lf) mutants, in which HIF-1 is constitutively active, and appears to be a direct HIF-1 target. cyp-36A1 controls expression of more than a third of HIF-1-upregulated genes, demonstrating that cyp-36A1 acts broadly downstream of hif-1. Regulation of gene expression and behavior by cyp-36A1 occurs cell non-autonomously, and the downstream effector of cyp-36A1 is the nuclear hormone receptor nhr-46, indicating that the likely function of CYP-36A1 is to generate a diffusible signal that controls NHR-46 activity. In addition to modulating behavior and gene expression, cyp-36A1 and nhr-46 mediate multiple forms of stress resistance associated with HIF activation. We conclude that CYP-36A1 and NHR-46 are important downstream effectors of the EGL-9/HIF pathway and function together to regulate a wide range of HIF-mediated physiology.

Results
A screen for suppressors of the egl-9(lf) egg-laying defect identifies the cytochrome P450 gene cyp-36A1 To identify novel, functionally important HIF effectors, we analyzed the modulation of C. elegans egg laying, the behavior that led our laboratory to discover the first EGLN gene, egl-9, and the first known functional role for any member of the EGLN/HIF pathway (Trent et al., 1983). egl-9(lf) mutants, in which HIF-1 is constitutively active, are defective in egg laying and become bloated with eggs as adults. Although the egg-laying defect of egl-9(lf) mutants is well-established, the downstream effectors of EGL-9 and HIF-1 in regulating egg-laying behavior remain unknown.
We performed a mutagenesis screen to identify genes that act in response to egl-9 to control egg laying. Specifically, we screened for second-site mutations that suppressed the egg-laying defect of egl-9(lf) animals ( Figure 1A). Such suppressors could define genes that function downstream of egl-9. Two isolates from this screen were allelic to hif-1 (Figure 1-figure supplement 1A and B), consistent with a previous observation that hif-1(lf) suppresses the egl-9(lf) egg-laying defect (Bishop et al., 2004) and validating the screen as a means of identifying components of the HIF-1 pathway. A third isolate, n5666, was not allelic to hif-1 and had a G106R missense mutation in the gene cyp-36A1, which encodes a cytochrome P450 enzyme (Figure 1-figure supplement 1A and C). A transgene carrying a wild-type copy of cyp-36A1 fully rescued the suppression by n5666 of the egl-9(lf) egg-laying defect, demonstrating that the mutation in cyp-36A1 is the causative mutation and suggesting that the suppression phenotype is caused by reduction of cyp-36A1 function ( Figure 1B-1E). Confirming these conclusions, a nonsense allele of cyp-36A1 also suppressed the egl-9(lf) egg-laying defect (Figure 3-figure supplement 1B and C). cyp-36A1(lf) single mutants did not exhibit hyperactive egg-laying behavior ( Figure 1F and Figure 3-figure supplement 1A and D), indicating that suppression of the egl-9(lf) egg-laying defect by cyp-36A1(lf) is not a consequence of a nonspecific increase in egg-laying rate. We further showed that cyp-36A1(lf) suppressed the previously reported egg-laying defect of hypoxia-exposed worms (Miller and Roth, 2009), demonstrating a role for CYP-36A1 under physiological conditions of HIF-1 activation (Figure 1-figure supplement 2). We then analyzed the role of cyp-36A1 in regulating other behaviors. We observed that egl-9(lf) mutants have reduced locomotion and defecation rates, both of which were suppressed  Figure 1. The cytochrome P450 gene cyp-36A1 is an effector of the hypoxia-response pathway that regulates behavior. (A) Schematic of the screen design. Stages of embryonic development adapted from Ringstad and Horvitz (2008) and Paquin et al. (2016). (B-F) Distribution of stages of eggs newly laid by adult hermaphrodites, used as a proxy for egg retention time in utero, of animals of the indicated genotypes. (B) Stages of eggs laid by wild-type animals. (C) egl-9 loss-of-function (lf) mutants laid later stage eggs than the wild type (p<0.001, Chi-square test with Holm-Bonferroni correction). (D) cyp-36A1(n5666) suppressed the egg-laying defect of egl-9(lf) mutants (p<0.001). (E) The cyp-36A1(+) transgene, which contains wildtype cyp-36A1, rescued the suppression of the egg-laying defect observed in cyp-36A1(n5666); egl-9(lf) mutants (p<0.001). (F) cyp-36A1(n5666) mutants displayed wild-type egg laying (p>0.05). (G) egl-9(lf) mutants were defective in locomotion rate, and this defect was suppressed by hif-1(lf) and cyp-36A1(lf) mutations. Mean ± SD of n = 3 biological replicates, *p<0.05, **p<0.01 considered significant (Student's t-test with Holm-Bonferroni correction).
Next we observed that cyp-36A1 expression is increased in egl-9(lf) mutants in a hif-1-dependent manner ( Figure 1I), consistent with results from an earlier genome-wide microarray study that identified cyp-36A1 as one of 63 genes regulated by hif-1 in hypoxia-exposed worms (Shen et al., 2005). ChIP-seq of HIF-1 by the modERN project showed HIF-1 binding at two sites near the cyp-36A1 coding region, one 5' to the start of the gene and one in the first intron (Kudron et al., 2018); both of these sites contain the HIF binding motif 5'RCGTG (Kaelin and Ratcliffe, 2008). We conclude that cyp-36A1 is a downstream effector of the hypoxia-response pathway that regulates multiple behaviors and that cyp-36A1 likely is a direct transcriptional target of HIF-1.

CYP-36A1 regulates gene expression changes and stress resistance downstream of HIF-1
We sought to determine if CYP-36A1 regulates other HIF-1-dependent processes. Based on sequence identity, CYP-36A1 is most closely related to the CYP2 family of cytochrome P450 enzymes, which function in both detoxification of xenobiotics and metabolism of endogenous molecules (Nebert et al., 2013). CYP2 family members and other CYPs can act on endogenous substrates to generate diffusible signaling molecules that regulate gene expression, such as eicosanoids and steroid hormones (Rendic and Guengerich, 2015;Dennis and Norris, 2015;Evans and Mangelsdorf, 2014). We hypothesized that CYP-36A1 might function in a transcriptional cascade to mediate aspects of HIF-1-dependent gene regulation. We performed an RNA-seq experiment comparing the wild type, egl-9(lf) mutants, egl-9(lf) hif-1(lf) double mutants, and cyp-36A1(lf); egl-9(lf) double mutants. We found that hif-1(lf) suppressed egl-9-dependent gene expression for 93% of egl-9(lf)-downregulated genes and 87% of egl-9(lf)-upregulated genes, indicating that most but not all regulation of transcription by egl-9 occurs through hif-1, consistent with previous work (Angeles-Albores et al., 2018). We further found that 36% of HIF-1-upregulated genes (i.e. genes that are upregulated in egl-9(lf) mutants and suppressed by hif-1(lf)) and 10% of HIF-1-downregulated genes were also regulated by cyp-36A1 (Figure 2A and B and Supplementary files 1 and 2). We focused on the HIF-1-upregulated genes, for which CYP-36A1 function appeared to be more broadly required. Gene ontology (GO) enrichment analysis of these HIF-1/CYP-36A1-upregulated genes suggested a role for cyp-36A1 in regulating stress resistance downstream of egl-9 and hif-1 (Figure 2source data 1). The EGL-9/HIF-1 pathway has previously been implicated in responses to numerous stressors in both nematodes and mammals, with crosstalk occurring between HIF and regulators of the immune response, unfolded protein response, and other stress-response pathways (Palazon et al., 2014;Schito and Rey, 2018;Wouters and Koritzinsky, 2008;Nakazawa et al., 2016;Powell-Coffman, 2010). We tested whether CYP-36A1 is involved in the response to three stressors for which HIF-1 is known to mediate resistance in C. elegans: infection by the pathogenic bacteria Pseudomonas aeruginosa (Darby et al., 1999;Bellier et al., 2009;Shao et al., 2010;Budde and Roth, 2011;Kirienko et al., 2013), tunicamycin-induced ER stress (Leiser et al., 2015), and oxidative stress from tert-butyl hydroperoxide (Bellier et al., 2009). Animals in which HIF-1 is constitutively active because of mutation in egl-9 or the C. elegans VHL homolog vhl-1 are resistant to these stressors relative to wild-type animals: such mutants survive longer when grown on Pseudomonas aeruginosa strain PA14 (Bellier et al., 2009), display reduced tunicamycin-induced growth inhibition (Leiser et al., 2015), and survive exposure to tert-butyl hydroperoxide at a higher rate than the wild type (Bellier et al., 2009). We found that cyp-36A1(lf); egl-9(lf) double mutants are more sensitive than egl-9(lf) mutants to all three of these stressors ( Figure

CYP-36A1 functions cell non-autonomously to regulate gene expression
We next sought to identify the site of action of CYP-36A1. We hypothesized that CYP-36A1 might function cell non-autonomously, as is the case for other cytochrome P450 enzymes that generate signaling molecules (Nebert et al., 2013;Evans and Mangelsdorf, 2014;Gerisch and Antebi, 2004).
We observed cyp-36A1 expression in many tissues, including neurons, intestine, hypoderm, and muscle ( Figure 3-figure supplement 2). To test the hypothesis of cell non-autonomous CYP-36A1 function, we focused on a cyp-36A1-mediated abnormality of egl-9(lf) mutants for which the site of dysfunction is well defined. Specifically, we examined expression of a GFP transcriptional reporter for the gene T24B8.5 (Shivers et al., 2009), which is expressed in only the intestine and based on our RNA-seq data is upregulated in egl-9(lf) mutants in a cyp-36A1-dependent manner. Interestingly, T24B8.5 expression is also upregulated in response to infection, ER stress, and oxidative stress (Shivers et al., 2009;Lim et al., 2014;Park et al., 2009). Expression of the reporter recapitulated the T24B8.5 expression changes observed by RNA-seq: increased expression of GFP was observed in egl-9(lf) mutants, which was suppressed by a second mutation in either hif-1 or cyp-36A1 ( Figure 3A and C and A screen for suppressors of cyp-36A1(lf) identifies the nuclear receptor gene nhr-46 We performed a mutagenesis screen to identify CYP-36A1 effectors that regulate egg-laying behavior, stress responses, and gene expression. We screened for mutations that suppressed both the low P T24B8.5 ::gfp expression and normal egg laying of cyp-36A1(lf); egl-9(lf) double mutants, looking for triple mutants that, like egl-9(lf) single mutants, had high GFP expression and were egg-laying defective. By screening for suppressors of the two abnormalities simultaneously, we were able to focus on effectors of CYP-36A1 rather than finding genes that affect only egg laying or only expression of T24B8.5 independently of the EGL-9/HIF-1/CYP-36A1 pathway. From this screen we identified one nonsense and one missense allele of the nuclear receptor gene nhr-46 ( Figure  , indicating that in addition to nhr-46 at least one other pathway is required to transduce egl-9-mediated modulation of egg laying and stress resistance. For example, in egl-9(lf) mutants or hypoxia, HIF-1 might drive expression of two (or more) targets that act together to inhibit egg laying and promote stress resistance. Thus, cyp-36A1 activation and consequent nhr-46 inhibition would promote egg-laying inhibition and stress  nhr-46 functions tissue-specifically to regulate gene expression and behavior nhr-46 is expressed in many tissues, including neurons, hypoderm, muscle, intestine, and the spermatheca (Feng et al., 2012). Tissue-specific expression of nhr-46 in the intestine, but not in neurons or muscle, rescued the high GFP expression caused by nhr-46(lf), indicating that nhr-46 acts cell autonomously in the intestine to control intestinal expression of T24B8.5 ( Figure 5A and B and Figure 5-figure supplement 1). nhr-46 expression in neurons fully rescued the egg-laying defect of cyp-36A1(lf); nhr-46(lf); egl-9(lf) triple mutants ( Figure 5C), demonstrating that nhr-46 function in neurons is sufficient to regulate egg-laying behavior. nhr-46 expression in intestine, but not muscle, also partially rescued the egg-laying defect of cyp-36A1(lf); nhr-46(lf); egl-9(lf) triple mutants ( Figure 5C). In combination with the tissue-specific CYP-36A1 and HIF-1 experiments described above, these results suggest that a CYP-36A1-regulated cell non-autonomous signal from any tissue can act on NHR-46 in the intestine to drive intestinal T24B8.5 expression and in either the nervous system or the intestine to regulate egg-laying behavior. That nhr-46 can act in either of two different tissues to regulate egg laying suggests that a second intercellular signal, for example a peptide or other hormone, might regulate egg laying downstream of NHR-46.

Discussion
These studies define a novel molecular genetic pathway that mediates cell non-autonomous regulation of gene expression by the HIF-1 transcription factor. Our genetic analysis indicates that hif-1 activates the cytochrome P450 cyp-36A1, which in turn inhibits the nuclear receptor nhr-46 ( Figure 6A). We speculate that the molecular function of CYP-36A1 is to generate an unidentified hormone that binds and regulates NHR-46, similar to other cytochrome P450 enzymes that function upstream of nuclear receptors (Evans and Mangelsdorf, 2014) and consistent with our observed cell non-autonomous function of CYP-36A1. We propose the following model ( Figure 6B): In wildtype animals, EGL-9 inhibits HIF-1 activity, such that the HIF-1 target cyp-36A1 is not transcribed. The unliganded NHR-46 represses expression of genes that promote stress resistance and inhibit egg laying. In egl-9(lf) mutants, as in hypoxia-exposed worms, HIF-1 is stabilized and drives increased cyp-36A1 expression. A CYP-36A1-generated hormone then binds NHR-46 and antagonizes the repressive function of NHR-46, accounting for the observed negative regulatory relationship between cyp-36A1 and nhr-46. Ligand-bound NHR-46 is likely activated to promote the expression of target genes, by analogy to a well-established mechanism of nuclear receptor regulation in which ligand binding mediates a switch from repressive to activating nuclear receptor function (Evans and Mangelsdorf, 2014). Alternatively, CYP-36A1 might degrade a ligand that activates NHR-46.

Cell non-autonomous regulation of stress resistance by HIF involves multiple pathways
Numerous studies have reported that HIF-1 promotes longevity and stress resistance of C. elegans (Darby et al., 1999;Treinin et al., 2003;Bellier et al., 2009;Mehta et al., 2009;Zhang et al., 2009;Chen et al., 2009;Lee et al., 2010;Shao et al., 2010;Budde and Roth, 2011;Kirienko et al., 2013;Fawcett et al., 2015). Nonetheless, despite substantial interest in the role of this pathway in stress biology, few relevant HIF effectors have been identified. Interestingly, a recent study reported that HIF-dependent serotonin signaling from the nervous system cell non-autonomously drives expression of the xenobiotic detoxification enzyme flavin-containing monooxygenase-2 (FMO-2) in the intestine, resulting in increased stress resistance and consequent extension of lifespan (Leiser et al., 2015). Here we report that a different signal, likely a lipophilic hormone, acts cell non-autonomously downstream of HIF to regulate gene expression and stress resistance. Consistent with the findings of Leiser et al. (2015), our RNA-seq data showed that in egl-9(lf) mutants there is a strong induction of fmo-2 expression and that this induction is suppressed by hif-1 mutation (Supplementary file 2), that is HIF-1 upregulates fmo-2 expression. Notably, this HIF-dependent expression of fmo-2 does not require cyp-36A1: the high fmo-2 expression of egl-9(lf) mutants is not suppressed by cyp-36A1(lf). We therefore suggest that serotonin-mediated fmo-2 expression and

Human cytochrome P450 enzymes might act as mediators of HIFdependent gene expression
We speculate that some human CYPs might serve as mediators of HIF-dependent gene expression changes through a mechanism analogous to that we describe above for CYP-36A1. In support of this hypothesis, previous studies have identified several human cytochrome P450 enzymes that are putative direct HIF targets based on whole-genome ChIP-chip and ChIP-seq analysis (Mole et al., 2009;Schö del et al., 2011). We note that the humoral nature of a CYP-generated molecule would make it a candidate mediator of non-autonomous regulation of hypoxia response by the EGLN/HIF pathway, such as is observed in remote ischemic preconditioning (Cai et al., 2013;Olenchock et al., 2016).

Cytochrome P450 enzymes might be major players in the hypoxiaresponse pathway
We previously identified another cytochrome P450 gene, cyp-13A12, as acting downstream of egl-9 in a locomotory behavior (Ma et al., 2013). In contrast to CYP-36A1, which is upregulated by HIF-1, CYP-13A12 is downregulated by HIF-1 upon hypoxia exposure or in egl-9(lf) mutants. The downstream effectors of CYP-36A1 and CYP-13A12 are also distinct, as we show here that CYP-36A1 regulates a nuclear receptor that controls transcription, whereas CYP-13A12 generates eicosanoids that act on a seconds-to-minutes timescale unlikely to require gene expression changes. Thus, different cytochrome P450 enzymes can act broadly, through multiple mechanisms, downstream of the EGL-9/HIF-1 hypoxia-response pathway in C. elegans. We propose that cytochrome P450 enzymes might similarly be important HIF effectors in mammals. Polymorphisms in numerous human cytochrome P450 genes have been associated with cardiovascular disease (Elbekai and El-Kadi, 2006;Rowland and Mangoni, 2014), for which HIF plays a protective role (Semenza, 2012), and with cancers (Agundez, 2004), for which HIF contributes to pathogenesis (Semenza, 2012). Furthermore, a study of genetic adaptations in humans to the environmentally hypoxic Tibetan plateau identified well-established members of the HIF pathway and, intriguingly, also noted positive selection at two cytochrome P450 loci (Simonson et al., 2010). Together these observations suggest that the cytochrome P450 family of enzymes is important in a wide range of hypoxia-associated contexts in humans. We suggest the presence of a mechanistic link between the canonical HIF pathway and the function of cytochrome P450 enzymes in humans and posit that an understanding of how these highly druggable enzymes (Schuster and Bernhardt, 2007) control processes downstream of HIF might reveal new therapeutic avenues for treating a broad array of disorders. Cytochrome P450 enzymes require oxygen as a substrate and thus might be regulated by oxygen availability. A long-standing issue in the understanding of HIF function is 'range finding,' that is how activity of the EGLN/HIF pathway is modulated such that it responds to different oxygen set points depending on context to drive a diversity of biological outputs (Ratcliffe, 2013). We hypothesize that the oxygen sensitivity of HIF-regulated cytochrome P450 enzymes might enable them to function in such a range finding mechanism to narrow the range of oxygen concentrations at which a subset of HIF-regulated physiological changes are activated.

Mutagenesis screen for suppressors of egl-9
To screen for suppressors of the egl-9 egg-laying defect, we mutagenized egl-9(n586) mutants with ethyl methanesulfonate (EMS) as described previously (Brenner, 1974). The starting strain contained the P cysl-2 ::gfp (nIs470) transgene, which is highly expressed in egl-9(lf) mutants and served as a reporter for HIF-1 activity (Ma et al., 2012). We used a dissecting microscope to screen the F2 progeny for suppression of the egg-laying defect (i.e. the Egl phenotype), picking (1) adults that appeared less Egl than egl-9(n586) mutants, and (2) eggs laid by the F2 animals that were at an earlier developmental stage than those laid by egl-9(n586) mutants. Screen isolates were backcrossed to determine dominant vs. recessive and single-gene inheritance pattern and crossed to him-8 (e1489); egl-9(sa307) hif-1(ia4) to test complementation with hif-1(lf). The screen allele n5666, which conferred a recessive phenotype and was not allelic to hif-1, mapped between SNPs pkP1052 and rs3139013 on LGI with SNP mapping (Davis et al., 2005) using a strain containing egl-9(n586) introgressed into the Hawaiian strain CB4856 (Ma et al., 2013). Whole-genome sequencing identified a mutation in cyp-36A1 in the n5666 interval, and transgenic rescue demonstrated that this cyp-36A1 mutation is the causative mutation, as described in the text.

Mutagenesis screen for suppressors of cyp-36A1
To screen for downstream effectors of cyp-36A1, we mutagenized cyp-36A1(gk824636); egl-9 (sa307) with ethyl methanesulfonate (EMS). The starting strain contained the P T24B8.5 ::gfp (agIs219) transgene, which has low expression in cyp-36A1(gk824636); egl-9(sa307) mutants and served as a reporter for CYP-36A1 activity. We used a dissecting microscope equipped to examine GFP fluorescence to screen for F2 progeny with high GFP fluorescence and an Egl appearance. The only two isolates failed to complement and were found to be alleles of nhr-46 by whole-genome sequencing and transgenic rescue. The mutant phenotypes of cyp-36A1(lf); n6126; egl-9(lf) were rescued by an nhr-46(+) transgene, demonstrating that the mutation in nhr-46 is the causative mutation and suggesting that n6126 is a loss-of-function allele.

Behavioral assays
To quantify egg-laying behavior, we scored the developmental stages of eggs laid by young adult hermaphrodites as described previously (Ringstad and Horvitz, 2008). Egg-laying defective mutants retain eggs longer in the uterus, thus laying them at later developmental stages. To examine egglaying behavior after exposure to hypoxia, young adult animals were placed in a hypoxia chamber (Coy Laboratory) at 1% O 2 balanced by N 2 for 24 hr, after which the egg-laying assay was performed in normoxia. Control animals were exposed to room air (21% O 2 ) for 24 hr; animals were randomly allocated to 1% or 21% O 2 treatment. Locomotion assays were performed on bacterial food and quantified using a custom worm tracker, as described previously (Paquin et al., 2016). Defecation assays were performed as described previously (Thomas, 1990), counting the number of defecation cycles in ten minutes.

Pseudomonas aeruginosa killing assay
Sensitivity to the Pseudomonas aeruginosa strain PA14 was assayed using the big lawn killing assay as described previously (Reddy et al., 2009). The big lawn killing assay was used to remove any influence of avoidance behavior on survival, as wild-type PA14 avoidance is dependent on normal aerotaxis behavior (Reddy et al., 2009), and egl-9(lf) mutants have previously been shown to display abnormal aerotaxis (Chang and Bargmann, 2008).

Tunicamycin survival assay
Sensitivity to tunicamycin was assayed by placing at least 100 starvation-synchronized L1 animals on NGM plates containing 5 mg/ml tunicamycin (Sigma), made using 10 mg/ml tunicamycin stock in DMSO and seeded with E. coli OP50 bacteria. Survival to the L4 larval stage or later was determined after three days.

t-BOOH survival assay
Sensitivity to tert-butyl hydroperoxide (t-BOOH) was assayed by placing~60 young adult worms on NGM plates containing 7.5 mM t-BOOH, made using 70% t-BOOH solution (Sigma) and seeded with E. coli OP50 bacteria. Survival was evaluated after 10 hr.

Microscopy
Epifluorescence images of P T4B8.5 ::gfp expression were obtained using an AxioImager Z2 upright microscope (Zeiss) and ZEN software (Zeiss). Confocal images of P cyp-36A1 ::gfp expression were obtained using an LSM 800 instrument (Zeiss) and ZEN software. Fluorescence intensity for the P T24B8.5 ::gfp reporter was quantified by measuring average intensity in a 300 mm section of the intestine centered on the vulva using FIJI software. P T24B8.5 ::gfp reporter imaging conditions were optimized for observation of fluorescence in the midbody; fluorescence was not saturated in this region in quantified images.

RNA isolation for qRT-PCR and RNA-seq
All strains were maintained at 22.5˚C for at least two generations without starvation prior to experiment.~150 very young adults (0-1 eggs in uterus) were picked into M9 buffer and allowed to settle. M9 was aspirated, and worms were then rinsed twice with M9 and twice with RNase-free water. Excess liquid was aspirated and the pellet was frozen in liquid nitrogen. RLT buffer (QIAGEN) was added to the frozen pellet, and worms were lysed using a BeadBug microtube homogenizer (Sigma) and 0.5 mm zirconium beads (Sigma). RNA was extracted using the RNeasy Mini kit (QIA-GEN) according to the manufacturer's instructions.
cyp-36A1 mRNA expression analysis by qRT-PCR Reverse transcription was performed using SuperScript III (Invitrogen). Quantitative PCR was performed using Applied Biosystems Real-Time PCR Instruments. Expression levels were normalized to the expression of the ribosomal subunit gene rpl-32.

RNA-seq library preparation
RNA integrity and concentration were checked on a Fragment Analyzer (Advanced Analytical). The mRNA was purified by polyA-tail enrichment, fragmented, and reverse transcribed into cDNA (Illumina TruSeq). cDNA samples were then end-repaired and adaptor-ligated using the SPRI-works Fragment Library System I (Beckman Coulter Genomics) and indexed during amplification. Libraries were quantified using the Fragment Analyzer (Advanced Analytical) and qPCR before being loaded for single-end sequencing using the Illumina HiSeq 2000.

RNA-seq data analysis
Reads were aligned against the C. elegans ce10 genome assembly using bwa 0.7.5a ) and samtools/0.1.19 (1000 Genome Project Data Processing Subgroup et al., 2009) (bwa aln/bwa samse), and mapping rates, fraction of multiply-mapping reads, number of unique 20-mers at the 5' end of the reads, insert size distributions and fraction of ribosomal RNAs were calculated using dedicated perl scripts and bedtools v. 2.17.0 (Quinlan and Hall, 2010). For expression analysis, reads were aligned against the C. elegans ce10 genome/ENSEMBL 65 annotation using RSEM 1.2.15 (Li and Dewey, 2011) and bowtie 1.0.1 (Langmead et al., 2009), with the following parameters: -p 6 -bowtie-chunkmbs 1024 -output-genome-bam. Raw expected read counts were retrieved and used for differential expression analysis with Bioconductor's edgeR package in the R 3.2.3 statistical environment (Robinson et al., 2010). First, common, trended, and gene-specific read dispersion across sequencing libraries and genes was estimated using the estimateDisp function. Given the small number of replicates, a gene-wise negative binomial generalized linear model (GLM) with quasi-likelihood tests (as implemented in the glmQLFit function) was used to test for differential expression between conditions (Lun et al., 2016). Briefly, this statistical framework works by first fitting the observed and expected distributions of read counts for each gene across conditions using a GLM, which is based on the negative binomial distribution and the observed read dispersion. The significance of biases in read counts is then tested using the quasi-likelihood F-test (implemented in glmQLFTest). This test provides more robust and reliable error rate control at low number of replicates, because it reflects the uncertainty in read distribution better than the likelihood ratio test. Models were fitted across all conditions and relevant differential expression testing was performed using glmQLFTest between pairs of conditions of interest. P values were adjusted for multiple comparisons using the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995). Gene ontology enrichment analysis was performed using GOrilla (Eden et al., 2009), examining genes that were significantly downregulated in cyp-36A1(lf); egl-9(lf) double mutants vs. egl-9(lf) single mutants (i.e. orange circle in Figure 2B) as compared to genes that were at least twofold upregulated in egl-9(lf) mutants vs. wild type (adjusted p value<0.05) and significantly downregulated in egl-9(lf) hif-1(lf) vs. egl-9(lf) (adjusted p value<0.05) (i.e. purple circle in Figure 2B).

Statistical analysis
Chi-square tests were used to compare the distribution of stages of eggs laid by wild-type and mutant animals. Unpaired t-tests were used to compare cyp-36A1 mRNA expression between strains, survival on tunicamycin between strains, survival on t-BOOH between strains, and P T24B8.5 :: gfp reporter fluorescence intensity between strains. Log-rank (Mantel-Cox) tests were used to compare survival of different strains on Pseudomonas aeruginosa. In cases of multiple comparisons, a Holm-Bonferroni correction was applied. Statistical tests were performed using GraphPad Prism software (Graphpad Prism, RRID:SCR_002798) version 7.0a. Biological replicates were performed using separate populations of animals.

Accession numbers
The GEO accession number for the RNA-seq dataset in this paper is GSE108283. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.