HIR-1 Mediates Response to Hypoxia-Induced Extracellular Matrix Remodeling

Inadequate tissue oxygen, or hypoxia, is a central concept in pathophysiology of ischemic disorders and cancer. Hypoxia promotes extracellular matrix (ECM) remodeling, cellular metabolic adaptation and metastasis. To determine how cells respond to hypoxia-induced ECM remodeling, we performed a large-scale forward genetic screen in C. elegans. We identified a previously uncharacterized receptor tyrosine kinase (RTK) named HIR-1 as a key mediator in a pathway that orchestrates transcriptional responses to hypoxia-induced ECM remodeling. Impaired ECM integrity caused by hypoxia or deficiency of the oxygen-dependent procollagen hydroxylases, heme peroxidases or cuticular collagens activates gene expression through inhibition of HIR-1. Genetic suppressor screens identified NHR-49 and MDT-15 as transcriptional regulators downstream of HIR-1. Cellular responses through HIR-1 maintain ECM homeostasis and promote animal adaptation to severe hypoxia. We propose that C. elegans HIR-1 defines an unprecedented type of RTK that mediates responses to hypoxia-induced ECM remodeling by mechanisms that are likely conserved in other organisms. ONE-SENTENCE SUMMARY A regulatory pathway for ECM homeostasis underlies adaptation to hypoxia and re-oxygenation


INTRODUCTION
Oxygen is essential for aerobic metabolism of life. Varying oxygen levels occur in natural environments and in tissues of living organisms, eliciting highly orchestrated organismic and cellular responses to maintain proper metabolic and physiological homeostasis. For example, mammals adjust pulmonary ventilation and blood circulation to improve oxygen delivery to target tissues under hypoxic conditions, and the roundworm C. elegans can navigate for preferred oxygen levels across a gradient of ambient hypoxia and trigger rapid locomotor response upon severe hypoxia and restoration of oxygen (1-3). Hypoxia is also a common pathophysiological condition in human disorders characterized by a low supply of oxygen, including myocardial ischemia, stroke and tumorigenesis (4). Pathological hypoxic conditions can lead to tissue necrosis and fibrosis, degeneration, inflammation and tumor metastasis, driving disease progression and leading to organismic mortality (5).
Animals respond to chronic hypoxia through evolutionarily conserved molecular pathways and cellular mechanisms that regulate gene expression and reprogram metabolism (4,6,7). The hypoxia inducible factor (HIF) is a master transcriptional regulator of hypoxic responses. Genetic studies of C. elegans led to the discovery of the evolutionarily conserved family of HIF hydroxylases (EGL-9 in C. elegans and EGLN2 in humans) that link oxygensensing to HIF-1 activation and transcriptional responses to hypoxia (8,9). Hydroxylated HIF under normoxia is recognized by the Von Hippel-Lindau tumor suppressor and targeted for proteasomal degradation whereas hypoxia causes impaired HIF hydroxylation, leading to transcriptional activation of HIF target genes (10). In C. elegans, the HIF-1 pathway mediates various physiological and behavioral responses to hypoxia (2,6,11,12). The transcriptional targets of mammalian HIF include LOX, LOXL2 and LOXL4 encoding copper-dependent lysyl oxidases that promote cross-linking of ECM components for enhanced tissue stiffness, a proposed trigger of tumor metastasis (5). Hypoxia-induced ECM remodeling, in turn, activates intracellular signaling cascades to regulate cell fate, metastasis and adaptation to hypoxia (13,14). How cells sense and respond to hypoxia-induced ECM remodeling remains undefined.
Collagens are major integral components of ECM that are extensively modified by oxygen-dependent enzymes including lysyl oxidases, lysyl hydroxylases and prolyl hydroxylases. Beyond HIF dependent mechanisms, it is unknown whether hypoxia can alter ECM integrity directly by impairing oxygen-dependent collagen modification and how ECM remodeling might elicit subsequent HIF-independent gene regulation. In mammalian cells, VEGF (vascular endothelial growth factor) can be induced by hypoxia through both HIFdependent and HIF-independent mechanisms (15). Various protein kinases and transcription factors other than HIF respond to hypoxia in a HIF-independent manner to modulate transcriptional responses (16,17). Mitogen activated protein kinases or integrin-linked protein kinases can transduce remodeled ECM signals to influence cell fate and resistance to hypoxia through transcriptional regulation (13,14). C. elegans has also been extensively used to investigate HIF-independent responses that mediate physiological and behavioral adaptation to severe hypoxia (18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28). Nonetheless, the precise roles and cellular mechanisms of HIFindependent transcriptional response to hypoxia-induced ECM remodeling in physiology and diseases remain poorly understood. We sought to identify new genes and pathways that transduce hypoxic signals to gene regulation independently of HIF-1. We generated C. elegans transgenic animals carrying GFP reporters that were robustly activated by hypoxia even in HIF-1-deficient animals. We performed large-scale forward genetic screens to isolate mutants with constitutively activated reporters under normoxia and performed suppressor screens to identify hypoxia-regulated transcription factors. We found that exposure to hypoxia resulted in the remodeling of ECM, which in turn triggers intracellular transcriptional responses by inhibiting a cell transmembrane RTK that we named HIR-1 (Hypoxia Inhibited Receptor tyrosine kinase).

comt-5p::GFP is a robust live reporter induced by hypoxia independently of HIF-1
To identify genes and pathways that mediate HIF-1-independent transcriptional response by hypoxia, we used RNA-seq to compare transcriptomes of wild-type (WT) C. elegans under normoxia (21% oxygen) or severe hypoxia (nearly 0% oxygen) for 2 hours, and egl-9 null mutants (in which HIF-1 is constitutively activated) under normoxia. We identified 72 genes that showed increased expression after hypoxia in WT animals but not in egl-9 null mutants ( Supplementary Fig. S1A). We generated transgenic animals with green fluorescent protein (GFP) driven by the promoters of these genes and focused on one of the reporters constructed, dmaIs1, that showed robust induction of GFP in the hypodermal and intestinal cells within 24 hours in 0.5% O2 ( Fig. 1A and 1B). dmaIs1 carries GFP driven by the promoter of comt-5, which encodes a predicted catechol-O-methyltransferase (Fig. 1C). We crossed animals carrying the dmaIs1 transgene with egl-9(sa307) or hif-1(ia04) loss-of-function (LOF) deletion mutants and, as expected, we did not observe activation of comt-5p::GFP in egl-9 mutants, whereas both WT and hif-1 mutants exhibited increased expression of comt-5p::GFP by hypoxia ( Fig. 1A and 1D).
To verify that endogenous comt-5 was induced by hypoxia independently of hif-1, we quantified its expression by quantitative PCR (qRT-PCR) and found that comt-5 expression was increased by hypoxia in both WT and hif-1(ia04) mutants (Fig. 1F). Other stress conditions that mimic hypoxia in activating HIF-1, including exposure of animals to CoCl2, H2O2, CNor H2S, did not induce the comt-5p::GFP reporter or increase endogenous comt-5 expression (Fig. 1E).
These results identify comt-5 as a specific and robust reporter gene that is activated by hypoxia independently of EGL-9 and HIF-1.

HIR-1 is a cell-autonomous regulator of comt-5 that mediates hypoxia signaling
To identify genes and pathways mediating HIF-independent regulation of dmaIs1, we performed large-scale forward genetic screens using ethyl methanesulfonate (EMS)-induced mutagenesis and isolated many mutants with constitutively activated comt-5p::GFP reporters even under normoxia (Table 1). Genetic linkage analysis together with whole-genome sequencing and subsequent RNA interference (RNAi) of candidate genes identified dma51 as an allele of the previously uncharacterized gene C24G6.2 (hir-1), which encodes a predicted transmembrane protein belonging to the RTK (InterPro scan) superfamily (29) (Fig. 2D). dma51 causes a LOF nonsense mutation W551Stop in HIR-1. We used CRISPR to generate a wholegene deletion allele dma101 for hir-1, which also showed constitutive comt-5p::GFP induction. A similar phenotype was also observed in mutants with partial in-frame deletion (tm3911) or an out-of-frame deletion (tm4098) that covers the exons encoding the intracellular domain of HIR-1 ( Fig. 2A and 2B). A transcriptional reporter hir-1p::GFP revealed ubiquitous expression of hir-1 that was especially strong in the pharynx, hypodermal and seam cells in all developmental stages (Fig. 2E). To determine whether hir-1 regulates comt-5p::GFP cell-autonomously in the hypoderm, where comt-5p::GFP is induced by hypoxia, we generated transgenic animals with hypodermal-specific expression of hir-1(+) driven by the dpy-7 promoter. We observed that transgenic dpy-7p::hir-1(+) extrachromosomal arrays rescued the hir-1(LOF)-induced hypodermal activation of comt-5p::GFP (Fig. 2F). To rule out a possibility that hir-1 works cell non-autonomously, we generated transgenic animals with neuronal-specific expression of hir-1(+) driven by the pan-neuronal promoter ric-19p. We did not observe rescue of the hir-1(-)induced activation of comt-5p::GFP (Fig. S2). These results identify hir-1 as a cell-autonomous negative regulator of comt-5p::GFP.
Because hir-1 inactivation phenocopied exposure to hypoxia in comt-5p::GFP activation, we wondered whether hypoxia directly inhibits HIR-1. We generated a translational GFP reporter of HIR-1 (Fig. 2D) and observed decreased levels of full-length SDS-soluble HIR-1::GFP in hypoxia-treated animals (Fig. 2H). In contrast, a range of partial-length HIR-1::GFP species increased in abundance, suggesting that hypoxia leads to HIR-1 proteolytic processing ( Fig. 2H). Moreover, hypoxia increased the abundance of HIR-1::GFP in the protein pool that did not migrate in the SDS-PAGE, suggesting partial insolubility of HIR-1 induced by hypoxia ( Fig. S2). By direct imaging of HIR-1::GFP, we observed that hypoxia induced formation of HIR-1::GFP foci without affecting the transgenic co-injection marker unc-54p::mCherry in the pharyngeal muscles (Fig. 2G). These data indicate that hypoxia regulates HIR-1::GFP proteins and thereby control gene expression of comt-5 downstream of HIR-1.
These results indicate that impaired hypodermal ECM integrity, which results in dumpy or blistering morphological defects, activates comt-5p::GFP.

Hypoxic stress results in the remodeling of ECM and cuticle disintegration
Since impaired hypodermal ECM integrity increased expression of comt-5p::GFP we wondered whether hypoxia, which also leads to activation of comt-5p::GFP, leads to the remodeling of hypodermal ECM. To test whether collagens proximal to the hypodermal plasma membrane are directly affected by hypoxia, we generated transgenic animals with FLAG-tagged collagen DPY-3. By Western blot analysis we found that the molecular weight of DPY-3::FLAG was markedly altered in animals exposed to hypoxia for 24 hours (Fig. 3D), indicative of altered DPY-3::FLAG cross-linking. In contrast, another cuticle collagen COL-19 did not exhibit altered molecular weight after exposure to hypoxia or genetic conditions increasing expression of comt-5p::GFP, including RNAi against perl-1 and dpy-3 (Fig. S3B). Translational reporter COL-19-GFP normally localizes to the circumferential annular rings and longitudinal alae of the adult exoskeleton and genetic conditions activating comt-5p::GFP lead to disorganization of COL-19-GFP (33). We examined COL-19::GFP in WT animals exposed to hypoxia for 24 hours (nearly 0% O2) and observed disorganized COL-19::GFP distribution proximal to the alae (Fig. 3E).
Cuticle biosynthesis begins in the endoplasmic reticulum (ER) lumen when collagens are hydroxylated by prolyl and lysyl hydroxylases allowing its trimerization, glycosylation and secretion into ECM (34,35). Although there are four ER procollagen hydroxylases in C. elegans (dpy-18, phy-2, phy-3 or phy-4), which are partially genetically redundant (36), the hypomorphic allele of dpy-18 caused disorganization of the COL-19::GFP, which was similar to the pattern observed in worms exposed to hypoxia. We also observed cuticle disorganization in animals with inactivated let-268, the sole C. elegans lysyl hydroxylase, and perl-1 whereas inactivated dpy-3 resulted in altered GFP pattern characterized by a complete loss of the cuticular furrows ( Fig. 3E). Interestingly, LET-268 activity appears to be limited to collagen type IV localized in the basement membrane but not cuticular collagens (37,38). We performed RNAi against let-268 and observed larval arrest with comt-5p::GFP activation (Table S1, Fig. 3A). On the other hand, the RNAi against emb-9 encoding collagen type IV did not increase expression of the comt-5p::GFP (Fig. 3A). The inactivation of let-268 leads to accumulation of the collagen type IV in the ER (38) suggesting that inactivation of let-268 activates comt-5p::GFP indirectly likely through affecting maturation of ECM proteins in the ER. These data indicate that insufficient cuticular collagen modification including hydroxylation mimics hypoxia or hir-1-induced comt-5p::GFP activation in C. elegans.
In addition, WT animals exposed to hypoxia for 24 hours (nearly 0% O2), perl-1, dpy-3 or dpy-18 mutants or let-268 knockdowns by RNAi exhibited exacerbated sensitivity to osmotic stress, characteristic of mutants with disrupted cuticle integrity (39), and is likely caused by increased cuticle permeability to water (Fig. S3C). Furthermore, direct permeability assays with the cuticle impermeable dye Hoechst 22358 (40) revealed intercalation of the dye in the nuclei only in perl-1 and dpy-3 mutants, let-268 knockdowns and hypoxia-treated WT animals but not WT animals under normoxia (Fig. S3D). These data provide multiple independent lines of evidence that cuticle in animals exposed to hypoxia exhibits altered integrity characterized by increased permeability.

NHR-49 and MDT-15 mediate comt-5 transcriptional response to hypoxia
To identify specific transcription factors that drive comt-5p::GFP expression in response to hypoxia and ECM remodeling, we sought second-site suppressor mutations of the most penetrant comt-5p::GFP-activating mutation dma11 isolated from EMS screens ( Table 1). The gene defined by dma11 remains as yet unidentified. Nonetheless, we isolated two independent suppressing alleles dma53 and dma54 and used linkage analysis and RNAi phenocopying to identify them as mutations of mdt-15 and nhr-49, respectively. dma53 is a nonsense mutation leading to a premature stop codon in mdt-15, and dma54 is a missense mutation G33R in nhr-49 (Fig. 4B). NHR-49 and MDT-15 are transcriptional regulators that physically interact to regulate lipid homeostasis (41). Protein sequence analysis of orthologous nuclear hormone receptors revealed that Gly 33 is in the conserved DNA binding domain among all examined sequences, including the vertebrate orthologue HNF4 (Fig. 4D). To verify that LOF of nhr-49 also suppressed hir-1-induced comt-5p::GFP activation, we crossed the nhr-49 null allele nr2041 with hir-1 mutants and found that the hir-1; nhr-49 double mutants exhibited suppressed comt-5p::GFP. Moreover, hypoxia did not activate comt-5p::GFP in nhr-49 or mdt-15 null mutants whereas the gain-of-function mutations nhr-49(et7) and mdt-15(et14) showed constitutively activated comt-5p::GFP under normoxia (Fig. 4A). Western blot and qRT-PCR analysis confirmed the requirement for NHR-49 and MDT-15 in the activation of comt-5p::GFP in hir-1 mutants (Fig. 4C). qRT-PCR analysis also confirmed that both nhr-49 and mdt-15 were required for hypoxic induction of comt-5 (Fig. 4E). These data indicate that transcriptional activation of comt-5 by hypoxia or LOF of HIR-1 requires NHR-49 and MDT-15.
NHR-49 and MDT-15 orchestrate lipid homeostasis in C. elegans (42) and mediate response to changes in lipid metabolic cues and temperature (11,42). We generated a transgenic strain expressing nhr-49::Venus under the ubiquitous rpl-28 promoter. We did not observe altered pattern of the subcellular localization of NHR-49::Venus after hypoxia, suggesting that NHR-49 is not regulated at the level of nuclear translocation. In addition, we observed that comt-5p::GFP induction by hir-1 inactivation was limited to adulthood and increased with age, whereas nhr-49 gain-of-function mutation or hypoxia activated comt-5p::GFP in all developmental stages (Fig. S4A). These results support that NHR-49 is essential for HIR-1-dependent transcriptional response but plays broader roles than HIR-1.

Mechanisms of HIR-1 regulation by hypoxia-induced ECM remodeling
The genetic evidence connecting hypoxia, ECM integrity and hir-1 suggests that exposure to hypoxia induces remodeling of the ECM, leading to inhibition of HIR-1 and subsequent activation of comt-5p::GFP. The intracellular domain of HIR-1 is structurally similar to the proto-oncogene receptor RET and fibronectin growth factor (FGF) receptors, with conserved catalytic sites essential for autophosphorylation (Fig. 5A). We wondered whether the HIR-1 kinase activity could be inhibited by depletion of ATP due to hypoxia. We exposed animals to rotenone, an inhibitor of the oxidative phosphorylation, but did not observe increased expression of comt-5p::GFP. The extracellular domain of HIR-1 resembles immunoglobulin-like fold (a.a. 80-446) including fibronectin type III-like fold (a.a. 84-170), domains common in many receptors previously identified as interacting with ECM proteins (43,44). FGF4 is a ligand implicated in mediating ECM sensing to regulate trophoblast stem cell fate (13). Inactivation of C. elegans FGF-encoding gene let-756 by RNAi led to constitutive activation of comt-5p::GFP under normoxia, phenocopying hir-1 LOF mutants (Fig. 5B). We generated an HA-tagged translational reporter of let-756 and observed hypoxia-induced change in LET-756 molecular weights (Fig. 5C). These findings indicate co-regulation of HIR-1 and LET-756, supporting the notion that HIR-1 acts with FGF-like proteins to sense ECM remodeling upon hypoxia, which would attenuate binding of FGF-like ligands and thereby inactivating HIR-1.
Collagens are the main structural components of the ECM and cuticle, and the epicuticle contains lipids that regulate its permeability (46). We found that hir-1 mutants exhibited increased expression of genes involved in lipid metabolism, including acs-2, a target gene of NHR-49 and key regulator of fatty acid homeostasis (42) (Fig. S5B). LOF of hir-1 led to disrupted cuticle integrity, supporting that hir-1 mutants exhibit more permeable cuticles ( Fig.   6B and 6C, and S3F). The cuticle defects of hir-1 mutants was not fully rescued by nhr- 49 LOF, indicating that defects in ECM integrity of hir-1 mutants under normoxia involves abnormal activation of additional unidentified transcription factors (Fig. S4B).
We next tested whether the genetic program mediated by HIR-1 signaling pathway was essential for resistance to severe hypoxic stress. We exposed 1-day old WT and hir-1(dma101) mutants to anoxia for 40 hours and subsequently compared their locomotion behavior after reoxygenation. All tested animals were in suspended animation-like state unable to respond to external stimuli (mechanic touch or UV light) during anoxia and immediately after reoxygenation, which was followed by behavioral recovery in crawling (defined as displacement) and pharyngeal pumping. We found that the displacement of re-oxygenated WT animals gradually recovered, whereas hir-1 mutants had markedly severe locomotion defects that did not improve even after 24 hours of re-oxygenation and culminated in enhanced rates of organismic death (Fig. 6D-F). These results indicate that HIR-1 plays key roles to control a broad genetic program that guards against maladaptive cuticle collagen and ECM homeostasis for animal survival upon exposure to severe hypoxia (Fig. 6G).

DISCUSSION
How cells in multicellular organisms sense hypoxia-remodeled ECM to promote cuticle homeostasis and facilitate animal adaptation to severe hypoxia is unknown. From genetic screens, we identified HIR-1 as a key mediator of the transcriptional response to hypoxiainduced ECM remodeling (47,48). Altered expression pattern of GFP-tagged HIR-1 in animals exposed to hypoxia suggests that HIR-1 is directly regulated by hypoxia-induced ECM remodeling. The observed HIR-1::GFP foci are reminiscent of aggregation and proteolytic processing, a previously observed effect of hypoxia on other proteins in C. elegans (27,28). We do not exclude a possibility of receptor clustering, internalization or block of the membrane trafficking that would affect activity of HIR-1 or its interacting partners. HIR-1 does not possess the discoidin domain, which is responsible for interaction with collagens to sense ECM remodeling (45). Our genetic evidence suggests that LET-756 homologous to mammalian FGFs is a likely HIR-1 ligand, whose binding to HIR-1 might be modulated by hypoxia-induced ECM remodeling. How precisely HIR-1 is regulated by ECM remodeling in coordination with its ligand and, in turn, transduces intracellular signaling for ECM homeostasis awaits further studies.
The integrity and composition of ECM are sensitive to oxygen availability due to oxygendependent enzymatic activities of prolyl and lysyl hydroxylases in the ER as well as dual oxidases in the extracellular space (31,49). We found that procollagen hydroxylase or dual oxidase inactivation by RNAi phenocopied hypoxia-induced activation of comt-5p::GFP. LOF phenotypes of these oxygen-dependent enzymes suggest that they likely mediate hypoxic response directly, whereas their HIF-induced increase in expression might constitute a homeostatic response to eventually restore their activity in the cell. Depending on the tissuespecific range of oxygen levels to which these enzymes are sensitive, procollagen hydroxylases and extracellular oxidases likely act as cellular oxygen sensors that mediate responses to varying degrees of hypoxia. We propose that such oxygen sensing in the ER or extracellular space acts in parallel to cytosolic EGLN oxygen sensors to mediate HIF-independent transcriptional programs to induce ECM remodeling and inhibit HIR-1 to maintain ECM homeostasis and animal survival under severe hypoxic conditions. Hypoxia-induced ECM remodeling occurs in cultured tumor cells (5) and ECM surrounding bone marrow tumor cells exhibit increased stiffness that triggers metastasis (50).
Increased stiffness of the ECM due to enhanced crosslinking of ECM proteins is mediated by increased expression of lysyl hydroxylase and lysyl oxidase and increased collagen deposition (5,51,52). Similarly, we also found hypoxia increased the expression of genes encoding procollagen lysyl oxidase, ER hydroxylases, and collagens in C. elegans, indicating conserved transcriptional responses that promote homeostatic ECM remodeling, which may have been coopted by tumors to facilitate hypoxic survival and metastasis. ECM remodeling during aging through differential regulation of collagen-encoding genes contributes to extension of C. elegans longevity (53), consistent with the notion that homeostatic ECM remodeling mediated by the HIR-1 pathway promotes cellular and organismic resistance to hypoxic stresses.
We found that proper HIR-1 regulation is important for C. elegans adaptation to hypoxic stress and re-oxygenation while its downstream transcriptional effectors include not only ECM components but also a non-ECM gene comt-5. comt-5 is predicted to encode a catecholaminedegrading enzyme, and its upregulation by hypoxia likely help decrease catecholamine levels to alleviate the toxicity of their oxidized forms under hypoxia, while its constitutive expression in hir-1 mutants under normoxia is likely detrimental. Dopamine targets peripheral tissues to maintain xenobiotic stress resistance in C. elegans (54), however, whether upregulation of comt-5 is required for proper resistance to hypoxic and/or xenobiotic stress awaits further studies. Although we used primarily comt-5::GFP reporters for gene and pathway discoveries, transcriptomic analysis revealed that HIR-1 regulates a broad genetic program responsible for remodeling of ECM, altering cuticle integrity and reprograming lipid metabolism.
Our findings suggest that hir-1-and nhr-49-mediated reprogramming of lipid metabolism likely also contributes to hypoxic survival in addition to ECM homeostasis. Hypoxic regulation of genes involved in cuticular lipid synthesis can promote cuticle permeability and adaptation to hypoxia in Arabidopsis thaliana (55). Lipid metabolic reprogramming also contributes to tumor development and hypoxic resistance in mammals (56). Upstream of NHR-49, the regulatory axis from oxygen, ER hydroxylases, and collagen to RTKs appears to share features common in both C. elegans and humans. We thus propose that the HIR-1 pathway is evolutionarily conserved to mediate cellular response to hypoxia and ECM remodeling in diverse organisms.
Numerous human RTKs, including FGFRs, have been implicated in driving tumor survival and metastasis (57,58), and ECM remodeling is essential for the progression of cancer, tissue fibrosis and many other diseases involving ECM dysregulation (59). Given the central role of HIR-1 in mediating cellular response through ECM remodeling to promote resistance to hypoxia, the human counterpart of HIR-1 may, once verified, be a promising therapeutic target for solid tumors that survive severe hypoxia and metastasize through ECM remodeling.

C. elegans strains
Animals were maintained under standard procedure with nematode growth media (NGM) plates unless otherwise stated. Bristol strain N2 was used as wild type and Hawaiian strain CB4856 was used for the linkage analysis of the mutants (60,61). hir-1 null alleles were generated by CRISPR to induce double stranded breaks and subsequent non-homologous end joining caused a deletion of hir-1. Feeding RNAi was performed as previously described (62).

Genetic screens
Stereo-epifluorescence dissecting microscope (Nikon SMZ18) was used to isolate mutants with constitutive expression of comt-5p::GFP reporters after ethyl methanesulfonate (EMS)-induced mutagenesis, as described previously (2,11). Mutants were mapped genetically by single nucleotide polymorphisms-based linkage analysis using the Hawaiian C. elegans strain CB4856 and then were sequenced by whole-genome sequencing to obtain lists of candidate genes.
Genes with putative causal mutations were verified by RNA interference that can phenocopy mutants and the causality of mutation was subsequently confirmed by transformation rescue of mutants with wild-type alleles as transgenes.

Environmental stress assays
Hypoxia chamber with ProOx110 oxygen controller (Biospherix) was used for hypoxia stress with 0.5%-21% oxygen concentrations. Hypoxia incubator chamber (Applied StemCell) with constant nitrogen flow delivering was used for achieving severe hypoxia with nearly 0% oxygen (anoxia stress). In hydrogen peroxide stress assay, animals were placed on the plate with 10 mM peroxide in the NGM and observed in the 1-48-hour interval for GFP activation. For determining effects of HIF-activating compounds on comt-5p::GFP, 5 mM CoCl2, 5 mM KCN containing NGM plates were used. For assaying hydrogen sulfide, 0.1 mg of NaHS powder was placed onto NGM agar plate (10 ml of agar) with sealed lid by parafilm to prevent leaking of the released H2S gas. Animals were subsequently screened for the GFP induction in the 1-48-hour interval, and for the qRT-PCR analysis collected after 2 hours of exposure.

Western Blot analysis
Animals were lysed in the Laemmli sample buffer (BioRad) with reducing agent betamercaptoethanol followed by boiling the samples for 10 minutes. The worm lysates were separated by SDS-PAGE and subsequently detected by GFP Goat polyclonal antibody (Fisher scientific -AF424) with Histone H3 antibody (AbCam -ab1791) as a loading control. For mCherry-tagged EMB-9, we used mCherry antibody 16D7 (Thermo Fisher Scientific M11217).

RNA-seq data analysis
The prinseq-lite software (0.20.4) was used (64) to trim and filter raw reads. Reads longer than 30 bp together with minimum quality score >15 were used for subsequent analyses. The Pairfq script was used for separation of paired and single reads. Clean reads were mapped to the C.

Imaging
Animals were mounted onto 2% agarose pad with 10 mM sodium azide and imaged with EVOS FL auto digital microscope for epifluorescence imaging or the confocal Leice SPE microscope for high-resolution col-19::GFP confocal imaging. At least 3 biological replicates (≥10 animals for each replicate) were used for quantification of the designated phenotype.

Hoechst staining
Animals were placed into liquid drops containing 2 µg/ml of the Hoechst 22358 dye diluted in M9 buffer for 15 minutes. Then the animals were picked into fresh M9 drops and subsequently placed onto 2% agarose pad with 10 mM sodium azide for imaging by the confocal Leica SPE microscope. At least 3 biological replicates (≥10 animals for each group) were used for quantification of stained animals.

Anoxia sensitivity assay
Animals were grown at 25°C for two continuous generations in non-starving non-stressed conditions. 1-day old adult hermaphrodites were placed on the NGM plates into the anoxia chamber for 40 hours. Animals were subsequently screened for their paralysis/movement in the indicated time intervals. WormLab system (MBF Bioscience) was used for quantification of the displacement, moving average speed and tracking based on the mid-point position. At least 3 biological replicates (10 animals per assay) were used for statistical analysis.