Induction of the hypoxic response by targeted genetic manipulations can increase lifespan in C. elegans. These manipulations include decreasing activity of the VHL-1 E3 ubiquitin ligase that targets the transcription factor HIF-1 for degradation under hypoxia³⁰ or by a mutation that stabilizes HIF-1 (HIF-1^P621A)^6,31 (Fig. 1A). We previously found that stabilizing HIF-1 in serotonergic neurons is sufficient to extend lifespan³¹. To determine which serotonergic neurons initiate HIF-1-mediated longevity, we generated strains with a nondegradable HIF-1 variant (HIF-1^P621A)^31,40 expressed under promoters specific to each of C. elegans’ three primary serotonergic neuron types—the ADF, NSM, and HSN neurons⁴¹. These transgenic worms were then crossed into the hif-1 null background to ensure that HIF stabilization in one neuron type could not feedback to modify HIF-1 activity in other cells. When we measured the lifespan of each strain relative to WT and hif-1 knockout controls, we observed that stabilizing HIF-1 in the ADF or NSM neurons significantly extended lifespan by 26% and 23%, respectively (Fig. 1B-C). HIF-1 stabilization in the HSN neurons had a smaller effect but significantly extended lifespan by 9% (Fig. 1D). This result indicates that modifying signaling in any serotonergic neuron is sufficient to induce some level of hypoxic response-mediated longevity. These data suggest that serotonergic neurons can partially substitute for each other to promote longevity in this pathway. In summary, the ADF and/or NSM serotonergic neurons play an essential role in hypoxic response-mediated longevity.

To further test the necessity of the most impactful (for lifespan) serotonergic neurons in hypoxia-mediated longevity, we used transgenic animals in which tph-1, the rate-limiting enzyme in serotonin synthesis, was knocked out exclusively in the ADF or the NSM serotonergic neurons^12,42. We then measured the lifespans of these strains on empty vector (EV) and vhl-1 RNAi to determine the necessity of ADF and NSM serotonergic signaling in vhl-1-mediated longevity. Successful RNAi knockdown was confirmed with qPCR validation (Fig. S1A). We found that serotonin synthesis in ADF but not NSM neurons was required for vhl-1 knockdown to extend lifespan (Fig. 1E-F). This indicates that while HIF stabilization in multiple serotonergic neurons can extend lifespan, ADF serotonin signaling is required for lifespan extension in response to vhl-1 knockdown.

To test if serotonin production in a single neuron type is sufficient for the hypoxic response to extend lifespan, we rescued tph-1 expression in a tph-1 null background using promoters specific to the ADF or the NSM neurons¹². The lifespans of these strains on EV or vhl-1 RNAi were then measured. As expected and consistent with our previous publication³¹, the lifespan of the tph-1 knockout strain was not extended by vhl-1 knockdown (Fig. S1B). Consistent with the ADF neuron’s requirement for hypoxic response-mediated longevity (Fig. 1E-F), serotonin synthesis in the ADF neurons was sufficient for vhl-1 knockdown to extend lifespan while NSM serotonin synthesis was not (Fig. 1G-H). The enzyme fmo-2 is induced in the intestine under hypoxia and is required for lifespan extension by environmental hypoxia, HIF-1 stabilization, or vhl-1 depletion³¹. To test whether ADF-specific HIF-1 stabilization also induces fmo-2, we measured fmo-2 transcription in this strain using qPCR. We found that ADF HIF-1 stabilization significantly induced fmo-2 expression relative to WT controls (Fig. 1I), suggesting a model in which ADF signaling extends lifespan through the same mechanism as vhl-1 knockdown and environmental hypoxia. Together, these data indicate that while HIF-1 stabilization in any serotonergic neuron is sufficient to extend lifespan, only ADF serotonin production is necessary and sufficient for vhl-1-mediated longevity. This suggests that under normal physiological conditions the serotonergic ADF neurons propagate a signal in response to whole body stabilized HIF-1.

After identifying ADF neurons as the most central serotonergic neurons in hypoxic response-mediated longevity, we next sought to determine the downstream receptor responding to serotonin signaling from ADF neurons. C. elegans have six known serotonin receptors—SER-1, SER-4, SER-5, SER-7, MOD-1, and LGC-50. Of these six receptors, our previous findings showed that ser-7 expression is required for hypoxia to induce fmo-2 and extend lifespan³¹. SER-7 is a G protein-coupled receptor (GPCR) with high sequence identity to the 5-HT₇ receptor in mammals^43,44. SER-7 is thought to be expressed in 27 of C. elegans’ 302 neurons⁴⁵ and contributes to many aspects of physiology such as reproduction and pharyngeal pumping⁴³. Consequently, broad manipulation of SER-7 signaling is not an ideal method for exclusively modifying aging. To identify more specific targets for lifespan extension within this pathway, we asked whether ser-7 expression in any subset of neurons is sufficient to rescue the hypoxic response in a ser-7 null background.

We created transgenic strains with ser-7 expression under the control of 11 different cell-specific promoters in a ser-7 null background. These promoters were selected to rescue ser-7 expression in the following neuronal populations: whole-body rescue (ser-7p::ser-7), interneuron rescue (glr-1p::ser-7), bioaminergic neuron rescue (cat-1p::ser-7), glutamatergic neuron rescue (eat-4p::ser-7), GABAergic neuron rescue (unc-47p::ser-7), cholinergic neuron rescue (unc-17p::ser-7), sensory neuron rescue (osm-6p::ser-7), GABAergic motor neuron rescue (unc-25p::ser-7), cholinergic motor neuron rescue (acr-2p::ser-7), M3 and M4 neuron rescue (ceh-28p::ser-7), and intestinal rescue (vha-6p::ser-7) in a ser-7 null background. To measure whether ser-7 expression restores a WT-like hypoxic response, we injected each construct into ser-7 null worms crossed with a single-copy transcriptional reporter for fmo-2 (fmo-2p::mCherry). Because fmo-2 is induced by and required for hypoxia-mediated longevity³¹, we utilized its induction as an efficient screening tool to identify components of this longevity pathway.

We first confirmed that ser-7 knockout attenuates vhl-1-mediated fmo-2 induction (Fig. 2A). The whole-body rescue of SER-7 expression under the endogenous ser-7 promoter restored fmo-2 induction to 91% of WT controls (Fig. 2A). Of the 10 cell-specific ser-7 rescues, expression in interneurons, bioaminergic neurons, glutamatergic neurons, or GABAergic neurons fully rescued vhl-1-mediated fmo-2 induction to ≥ 100% of the WT control response (Fig. 2A, pink bars). ser-7 rescue in GABAergic motor neurons or in sensory neurons showed a partial rescue (defined as a 50-99% increase from fmo-2 induction in the ser-7 knockout) (Fig. 2A, orange bars). In contrast, rescuing ser-7 in cholinergic motor neurons, cholinergic neurons, M3 and M4 neurons, or in the intestine did not rescue (Fig. 2A, gray bars). Taken together, 6 of the 10 cell-specific ser-7 rescues at least partially restored vhl-1-mediated fmo-2 induction, including some that have no known overlap in ser-7 expression patterns. Collectively, these data suggest that serotonin receptor ser-7 expression in multiple neurons may be sufficient to convey the hypoxic signal to the intestine and induce fmo-2.

We compiled a table containing the top candidates for the ser-7 neuron based on which rescue constructs restored ser-7 expression (simplified table in Fig. 2B, full table in Fig. S2A). We hypothesized that neurons present in many of the successful rescue constructs would be most important in the hypoxia longevity circuit. Conversely, we hypothesized that candidate neurons within rescues that did not restore a WT-like hypoxia response are less likely to be important (Fig. 2B; Fig. S2A). We acknowledge that this analysis strategy is biased toward neurons rescued in a greater proportion of the transgenic strains, which could lead to false negatives. From this analysis, we identified 10 neurons that were rescued in 3-4 of the 6 successful rescue constructs (Fig. 2B). Five of these neurons (DVB, FLP, M1, RME_LR, and RME_DV) were also present in unsuccessful rescue constructs, so they were not considered for follow-up experiments. Of the remaining five neurons (RIS, AUA, AVC, LUA, and VD_DD), the top candidate was the RIS neuron, because RIS was present in the highest number (4 as shown in Fig. 2B) of successful rescue constructs.

To validate whether the RIS neuron is a key signaling cell in hypoxia-mediated longevity, we first genetically ablated the RIS neuron to test its necessity. Our results showed that vhl-1 RNAi does not extend lifespan of RIS ablated worms, indicating that the RIS neuron is completely required for vhl-1 knockdown to extend lifespan (Fig. 2C). To test the sufficiency of RIS signaling, we next rescued ser-7 expression in the RIS neuron in the ser-7 null background under the flp-11 promoter. flp-11 is expressed in the RIS neuron⁴⁶ and uv1 neuroendocrine cells⁴⁷ that do not endogenously express ser-7⁴⁵. We first confirmed that ser-7 knockout worms are partially required for vhl-1 RNAi to extend lifespan (Fig. S2B, Cox Regression, p = 0.003, **). Significantly, ser-7 rescue in the RIS neuron (ser-7; flp-11p::ser-7) restores the lifespan extension by vhl-1 RNAi to the same degree as WT worms (Fig. 2D). These data suggest that ser-7 expression in the RIS neuron only is sufficient for vhl-1-mediated longevity. However, one limitation of this approach is that non-physiological expression of ser-7 in the uv1 cells could play a role in restoration of a WT-like hypoxic response. Together, these data are consistent with a model where hypoxic conditions stabilize HIF-1 in the ADF neurons, which modify serotonin signaling to the SER-7 receptor on the RIS interneuron.

After identifying neuron subtypes involved in serotonin signaling downstream of the hypoxic response, we next set out to identify whether other neurotransmitters act in the pathway. To answer this question, we obtained mutants deficient in the production of each type of neurotransmitter in C. elegans (Fig. 3A) and measured their lifespan on control and vhl-1 RNAi. Out of the six neurotransmitters we tested, blocking the production of GABA (unc-25, Fig. 3B) or tyramine + octopamine (tdc-1, Fig. 3C) attenuated longevity on vhl-1 RNAi. The requirement of GABA is consistent with our finding that ser-7 expression on the GABAergic RIS neuron is also necessary for hypoxic response-mediated longevity (Fig. 2), connecting the serotonergic and GABAergic components of this circuit.

Tyramine (analogous to mammalian adrenaline⁴⁸) is synthesized from tyrosine by tyrosine decarboxylase (tdc-1) and can be further converted to octopamine (analogous to mammalian noradrenaline) by tyramine β-hydroxylase (tbh-1) in C. elegans⁴⁹. Therefore, the tdc-1 knockout is deficient in both tyramine and octopamine synthesis (Fig. 3C) and the tbh-1 knockout (Fig. 3D) is only deficient in octopamine synthesis. Since worms lacking only octopamine (tbh-1, Fig. 3D) were long-lived on vhl-1 RNAi but worms lacking both tyramine and octopamine (tdc-1, Fig. 3C) were not, we can conclude that tyramine is required for the hypoxic response to extend lifespan. To further test whether tyramine but not octopamine is involved in this pathway, we generated a strain in which the primary octopamine-producing neuron, RIC⁴⁹, is genetically ablated. We found that the RIC neuron is not required for vhl-1-mediated longevity (Fig. 3E), further supporting that tyramine but not octopamine is a key neurotransmitter in the hypoxic response longevity circuit.

In addition to octopamine (tbh-1, Fig. S3C), we also observed that acetylcholine (unc-17, Fig. S3A) and glutamate (eat-4, Fig. S3B) were not required for lifespan extension by vhl-1 RNAi. Our lifespans of the dopamine synthesis mutant cat-2 indicated that dopamine is inconsistently required for the hypoxic response to extend lifespan. In three biological replicates, we observed a full requirement, partial requirement, and no requirement for dopamine synthesis in vhl-1 RNAi-mediated longevity. When analyzed together, the net result suggests that dopamine is partially required for vhl-1-mediated longevity (Fig. S3C, Cox Regression for interaction between genotype and vhl-1 RNAi, p < 0.0001, ****). We did not follow up on dopamine due to the inconsistency. Together, the results from this neurotransmitter screen support the conclusion that in addition to serotonin, both GABA and tyramine play a role in hypoxic response-mediated longevity.

Nematodes synthesize GABA in 26 of 302 neurons while tyramine is only synthesized in two cells⁵⁰. As a result, identifying highly specific components of tyraminergic signaling is more feasible than mapping GABAergic signaling components. Therefore, we sought to determine the tyramine-producing neuron(s) and tyramine receptor(s) that contribute to hypoxic response-mediated longevity. In C. elegans, the two canonically tyraminergic cell types (the RIM neuron and the uv1 neuroendocrine cells) express tdc-1 but not tbh-1⁵⁰. To determine which tyraminergic cell(s) act in the hypoxic response pathway, we expressed tdc-1 under promoters specific to either RIM or uv1 in a tdc-1 null background¹⁴. We then measured whether the lifespans of these tyramine rescue strains could be extended by vhl-1 RNAi. Our results showed that vhl-1 RNAi extended lifespan in the RIM rescue strain (Fig. 4A), but not in the uv1 rescue (Fig. 4B). This finding demonstrates that tyramine synthesis in the RIM neuron is sufficient for vhl-1 RNAi to extend lifespan. Notably, both tyramine rescue constructs lived longer than WT controls (Fig. 4A-B). This could indicate that altering tyramine signaling modifies lifespan both within and independently of the vhl-1-mediated longevity pathway.

To identify the tyramine receptor(s) necessary in the hypoxic response-mediated longevity pathway, we used RNAi to knockdown each of C. elegans’ four known tyramine receptors^51–54 (Fig. 4C). Since tyramine receptors are mostly expressed in neurons, and RNAi uptake into some neurons is less efficient than other cell types, we used a strain with enhanced neuronal RNAi uptake (TU3311, unc-119p::sid-1) crossed into vhl-1 knockout worms. This neural enhanced RNAi uptake strain was not used previously in this work for vhl-1 RNAi experiments because vhl-1 RNAi knockdown extends lifespan in WT worms³¹.

The generated vhl-1; unc-119p::sid-1 strain was long-lived compared to the control strain unc-119p::sid-1 (Fig. 4C). Successful RNAi knockdown was validated via qPCR (Fig. S4A). Our lifespan results show that the tyramine receptor tyra-3 was fully required for vhl-1- mediated longevity (Fig. 4C), but the tyramine receptors tyra-2 and ser-2 were not (Fig. S4B-C). Another tyramine receptor, Igc-55, showed a partial requirement for vhl-1-mediated longevity (Fig. S4D). Although there was a lifespan extension by vhl-1 on Igc-55 RNAi (Fig. S4D), we did observe a significant interaction between lgc-55 knockdown and genotype on lifespan (Cox Regression, p < 0.0001, ****), indicating a partial requirement. In summary, these data indicate that tyramine signaling from the RIM neuron to the TYRA-3 receptor is required for hypoxic response-mediated longevity, while the LGC-55 receptor may play a partial role. Interestingly, TYRA-3 is highly expressed on the low oxygen/high carbon dioxide-sensing BAG neuron as well as in the intestine^45,55, the site of fmo-2 induction during hypoxia³¹.

Given that tyra-3 is highly expressed in the canonical low-oxygen sensing neuron BAG⁴⁵, we next explored whether neurons responsible for sensing high- and low-oxygen conditions are essential for hypoxic response-mediated longevity. C. elegans have four oxygen sensory neurons: the URX, PQR, and AQR neurons detect high levels of oxygen⁵⁵, the BAG neuron responds to low levels of oxygen⁵⁶ and high levels of carbon dioxide⁵⁷ (Fig. 5A). We hypothesized that these oxygen sensing cells may perceive hypoxic conditions and help initiate the hypoxic response. Alternatively, these oxygen sensing neurons could modify behavior in response to hypoxia but not play a role in longevity. We found that vhl-1 knockdown was unable to extend lifespan when we genetically ablated the three high-oxygen sensing neurons (URX, AQR, PQR, Fig. 5B). Interestingly, the low O₂/high CO₂ responsive BAG neurons were also required for vhl-1-mediated longevity (Fig. 5C). We next crossed the URX/AQR/PQR ablation worms or BAG ablation worms into C. elegans with HIF-1 stabilized in the ADF neurons. Ablation of either the URX/PQR/AQR or the BAG neurons abrogated lifespan extension in the ADF HIF-1 stabilized worms. This suggests that oxygen sensing neurons act downstream of ADF HIF-1 stabilization in this pathway (Fig. 5D-E).

After identifying neurons, neurotransmitters, and neuroreceptors acting in the hypoxic response longevity circuit, we wondered how this neural circuit ultimately propagates information about hypoxic conditions to the intestine and induces the pro-longevity gene fmo-2. The C. elegans nervous system does not directly innervate the intestine. Instead, neurosignaling molecules bind to receptors on peripheral tissues through packaging in dense-core vesicles and subsequent release into and diffusion through pseudocoelomic fluid^58,59. These dense-core vesicles are packaged with a wide variety of neuropeptide signals, although there is also evidence that bioaminergic neurotransmitters like serotonin, dopamine, and adrenaline/noradrenaline can also signal through this dense-core vesicle mechanism^59–61.

We first tested if inducing the hypoxic response through vhl-1 RNAi extends lifespan in a mutant strain lacking the dense-core vesicle packaging gene unc-31. We observed that unc-31 was required for vhl-1 knockdown to extend lifespan (Fig. 6A). To examine whether neuropeptide signals carried by dense-core vesicles are involved, we knocked down two neuropeptide processing enzymes—egl-3 and egl-21⁶². Successful knockdown was validated with qPCR (Fig. S5A). Interestingly, egl-21 but not egl-3 was required for vhl-1- mediated longevity (Fig. 6B-C). This could be because egl-21 is the only enzyme known to perform the carboxypeptidase E cleavage step of neuropeptide processing⁶³, while there are four known proprotein convertases with similar function to egl-3^62,64.

We next tested a group of 6 genes corresponding to neuropeptide/receptor binding pairs that blunted hypoxic response-mediated fmo-2 induction (Fig. 6D). These genes, along with their corresponding receptors, were targeted for RNAi and tested for lifespan in TU3311 (unc-119p::sid-1) and TU3311 (unc-119p::sid-1); vhl-1 (ok161) strains. We found that the npr-3 receptor and it’s ligands, flp-15 and flp-21 were not required for vhl-1-mediated longevity. (Fig. S5B-D). However, knocking down the neuropeptide nlp-17 partially blocked vhl-1-mediated longevity (Fig. 6E, Cox regression p < 0.05). Consistently, knocking down nlp-17’s receptors also completely (npr-37, Fig. 6F) or partially (npr-43, Fig. 6G, Cox regression p < 0.001) blocked vhl-1-mediated longevity. Successful knockdown of npr-37 and npr-43 was validated with qPCR (Fig. S5A). The nlp-17 ligand result was validated using a genetic knockout (nlp-17 (ok3461)), which completely prevented vhl-1 knockdown from extending lifespan (Fig. S5E). Together, these data suggest a role for NLP-17 signaling in hypoxic response-mediated longevity.

Taken together, these results support a model in which HIF-1 stabilization in the ADF serotonergic neurons leads to modified serotonin signaling from the ADF to the SER-7 expressing and GABA-producing RIS neuron (Fig. 7). Oxygen sensing neurons act downstream of HIF-1 stabilization in the ADF neuron in the hypoxic response longevity circuit. Tyramine produced by the RIM neuron and neuropeptide (NLP-17) signaling are also required for the hypoxic response to extend lifespan, although it remains unclear whether these signals act upstream, downstream, or in parallel to the serotonergic and GABAergic components in this pathway.

C. elegans were cultured according to established protocols³¹. In brief, worms were grown at 20 °C on standard solid nematode growth media (NGM). Throughout their lifespan, the worms were fed E. coli OP50, except during RNA interference (RNAi) experiments, where E. coli HT115 was used to deliver double-stranded RNA. Transfers of worms were carried out using a platinum wire, unless stated otherwise. The RNAi strains employed are listed in Supplementary Table 1, while the strains used in the experiments are provided in Supplementary Table 2. Genotypes were verified through PCR, and RNAi imaging results were confirmed by sequencing and quantitative PCR (qPCR) before proceeding with the experiments.

For all RNAi knockdowns, worms were exposed to the RNAi treatment for two generations to achieve optimal knockdown. All RNAi clones were sourced from the Vidal RNAi library. Each RNAi clone was sequence-verified.

For RNAi validation experiments, RNA was isolated from day 1 adult worms that had been grown on RNAi for two generations. For measuring fmo-2 induction in the hif-1(ia4); ADF::HIF-1S strain, worms were grown on OP50, synchronized, and collected at day 1 of adulthood. RNA extraction was performed using the Direct-zol RNA Miniprep Kit (Zymo), and cDNA synthesis was carried out with the Biorad iScript cDNA Synthesis Kit. Gene expression was assessed using the SYBR Green quantitative RT-PCR (qRT-PCR) system (BioRad), with mRNA levels normalized to the housekeeping genes cdc-42 and Y45FD10.4. Primers for RNAi validation were designed to target the 3’ UTR of each gene to avoid amplification of the bacterially produced RNAi. Gene expression was quantified using standard curves.

Lifespan assays were conducted following previously described protocols³¹. Briefly, 10-15 gravid adults were transferred to NGM plates for a three-hour timed egg lay, after which they were removed. Once the progeny reached day 1 of adulthood, 60-80 worms were placed on NGM plates containing 33 µL of 150 mM fluorodeoxyuridine (FUdR) and 100 µL of 50 mg/mL ampicillin per 100 mL of NGM. FUdR inhibits progeny development, while ampicillin prevents bacterial contamination. NGM + FUdR + Ampicillin plates were seeded with concentrated bacteria (5x for OP50-fed lifespan assays). At least two plates per strain and condition were used for each lifespan replicate. Worms were considered dead when they failed to respond to a gentle touch with a platinum wire under a dissection microscope. Lifespan data were recorded at least three times a week until all worms were dead. To prevent escape, a barrier of 75 µL of 100 mM palmitic acid (Sigma-Aldrich) dissolved in 100% ethanol was applied along the edges of each lifespan plate. Data were analyzed using R version 4.3.1 and visualized in Adobe Illustrator 2022.

RNAi lifespan assays were performed similarly to standard lifespan assays, with modifications to the initial timed egg lay (TEL) procedure and food concentrations. To ensure maximal RNAi knockdown, worms were first TEL’ed for 3 hours on RNAi plates. The progeny from this TEL were left on RNAi plates to develop into gravid adults and then used for a second TEL under the same RNAi conditions. Progeny from the second-generation TEL were then used for the lifespan assay. RNAi bacteria (HT115) are ampicillin-resistant and can grow slowly on AMP-containing lifespan plates. To maintain consistent bacterial availability throughout the lifespan assay, RNAi plates were seeded with bacteria at 2x concentration, starting from an optical density (OD₆₀₀) of 3.0.

Fluorescent images for this study were captured using a Leica M165F fluorescent microscope, controlled by Leica Application Suite X (LASX) software. A minimum of 15 worms per condition were imaged at a magnification of at least 70x. For image quantification, individual worms were carefully separated to ensure they did not touch. Custom R code⁷⁶ was then used to create a pixel mask of each worm from the brightfield image and measure the fluorescent intensity of that region in the corresponding fluorescent image. The background fluorescence was subtracted from the mean fluorescence values. Data were analyzed using R version 4.3.1 and visualized in Adobe Illustrator 2022.

The height of all bar plots represents the mean value for each condition, with error bars indicating the standard error of the mean (SEM). For comparisons involving more than two conditions, one-way ANOVA followed by post-hoc Tukey HSD tests (2-tailed, unpaired) were used to assess interactions between variables and determine statistical differences between conditions. Significance levels are indicated as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. For lifespan assays, statistical comparisons of survivorship curves were performed using the survfit survival analysis function in R. The log-rank test was used to compare two survival curves. To assess interactions across multiple experimental variables and for comparisons involving more than two survival curves, a Cox regression analysis was conducted using the survivalMPL package in R. For experiments with multiple biological replicates, p-values were adjusted using the Bonferroni correction. Exact sample sizes (Ns) for each experiment are provided in the source data files, and minimum sample sizes for each plot are specified in the figure legends.