Sensory integration of food and population density during the diapause exit decision involves insulin-like signaling in Caenorhabditis elegans

Significance

Animals must respond appropriately to multiple sensory stimuli to make informed decisions. It remains unclear how the nervous system is able to integrate different sensory cues and propagate that information toward making decisions over longer timescales. We use the nematode Caenorhabditis elegans to investigate how sensory integration occurs during the decision to exit diapause, a stress-resistant developmentally arrested state triggered by multiple sensory inputs including food availability and population density. We show how expression of an insulin-like peptide critical to dauer exit reflects the sensory integration process and decision commitment, and we dissect the regulation of this insulin-like peptide’s expression. Our study analyzes the relationship between neuronal activity and neuropeptide expression during a complex decision with diverse sensory inputs.

Abstract

Decisions made over long time scales, such as life cycle decisions, require coordinated interplay between sensory perception and sustained gene expression. The Caenorhabditis elegans dauer (or diapause) exit developmental decision requires sensory integration of population density and food availability to induce an all-or-nothing organismal-wide response, but the mechanism by which this occurs remains unknown. Here, we demonstrate how the Amphid Single Cilium J (ASJ) chemosensory neurons, known to be critical for dauer exit, perform sensory integration at both the levels of gene expression and calcium activity. In response to favorable conditions, dauers rapidly produce and secrete the dauer exit-promoting insulin-like peptide INS-6. Expression of ins-6 in the ASJ neurons integrates population density and food level and can reflect decision commitment since dauers committed to exiting have higher ins-6 expression levels than those of noncommitted dauers. Calcium imaging in dauers reveals that the ASJ neurons are activated by food, and this activity is suppressed by pheromone, indicating that sensory integration also occurs at the level of calcium transients. We find that ins-6 expression in the ASJ neurons depends on neuronal activity in the ASJs, cGMP signaling, and the pheromone components ascr#8 and ascr#2. We propose a model in which decision commitment to exit the dauer state involves an autoregulatory feedback loop in the ASJ neurons that promotes high INS-6 production and secretion. These results collectively demonstrate how insulin-like peptide signaling helps animals compute long-term decisions by bridging sensory perception to decision execution.

A fundamental goal of neuroscience is to understand how nervous systems sense, integrate, and interpret diverse stimuli to deliver the appropriate output. The duration over which these processes occur spans from shorter timescales, such as during reflexive responses to harsh stimuli or chemotaxis toward attractive stimuli (1–5), to longer timescales, such as migration or hibernation decisions in response to changing weather patterns, temperature, and food availability (6–9). We lack a clear understanding of how neuronal activity, which occurs over short timescales of milliseconds to seconds, informs downstream processes that occur over timescales of hours or longer. One such decision that takes place over a longer timescale is the developmental decision to enter (or exit) diapause, a temporarily suspended developmental state that protects against environmental stress and promotes dispersal. Diapause is evolutionarily conserved across metazoans and requires integration of environmental cues to inform a coordinated, organismal-wide decision (10–12)

To better understand how a compact nervous system can interpret sensory signals to coordinate an organismal-wide decision that takes place over multiple hours, we studied the dauer exit decision. During early larval growth, C. elegans choose between two developmental fates depending on environmental conditions (Fig. 1A). Under favorable conditions, larvae undergo reproductive growth, whereas under unfavorable conditions, larvae enter the stress-resistant, long-lived developmentally arrested diapause state known as dauer (13–15). While in the dauer state, C. elegans continually assess their surroundings to detect environmental improvement; when conditions sufficiently improve, animals exit the dauer state and return to the reproductive cycle as late-stage larvae. Of the various environmental inputs to the dauer entry and exit decisions, the strongest input is a ratio of food to pheromone (14). Here, “pheromone” refers to a mixture of secreted dauer-regulating signaling molecules, termed ascarosides, that collectively convey population density (16, 17).

Fig. 1.

Neuropeptides, especially the insulin-like peptide ins-6, are critical for dauer exit. (A) During development, C. elegans makes multiple developmental decisions including whether to enter and when to exit the developmentally arrested state called dauer. This decision depends mainly on food availability, crowding, and temperature. (B) Overview of the dauer exit assay. Animals were induced to become dauer via growth on conditions of high pheromone concentration and high temperature. Nondauers were removed using SDS and surviving dauers were transferred to conditions of intermediate pheromone concentrations and lower temperature to stimulate between 40 and 80% of dauers to exit after 24 h. (C) Dauer exit rates of mutants defective for neuropeptide processing (sbt-1, egl-3, egl-21) or signaling (unc-31, ric-7) compared to that of wild type (WT). Means are written and shown by the orange line, and total number of animals scored is indicated in parentheses. (D) Dauer exit rates of mutants defective for single neuropeptide genes. Data compiled from multiple independent experiments and statistical analyses were only performed between mutant and wild-type samples measured in the same experiment. Bars indicate means. See SI Appendix, Fig. S1 for analysis of individual mutants. For (C) and (D), each dot is the dauer exit % from an assay plate containing 50 to 100 animals each. ****P < 0.0001, ***P < 0.001, **P < 0.01, compared to “WT” by Welch ANOVA with Dunnett’s T3 multiple comparison correction. Comparisons to “WT” were not statistically significant unless indicated otherwise.

The genetic pathways and neuroendocrine signaling mechanisms that govern the dauer entry and exit decisions (18–20) including a cGMP pathway, a TGF-β-like pathway, an insulin/insulin-like growth factor (IGF)-1 signaling (IIS) pathway, and a steroid hormone pathway. While considerably more attention has been paid toward the dauer entry decision, previous studies identified two insulin-like peptides, INS-6 and DAF-28, as regulators of dauer exit that work within the IIS pathway (21, 22).

Despite the wealth of knowledge collected on the C. elegans dauer entry and exit decisions, many fundamental questions remain, including: 1) How does sensory integration of food, pheromone, and temperature (among other possible cues) occur? and 2) How does short-term perception of environmental cues translate into the dauer decisions which take place over the course of hours? To address these questions, we studied the dauer exit decision using genetic and molecular neurobiological approaches. Using an ethologically relevant, pheromone-based assay, we demonstrate an essential role for neuropeptide signaling as a whole in dauer exit and validate INS-6 as important for dauer exit.

To understand how INS-6 production relates to sensory perception, we analyzed the spatiotemporal dynamics of ins-6 expression in response to different environmental cues and found that ins-6 expression in a pair of chemosensory neurons, the Amphid Single Cilium J (ASJ) neurons, reflects both food and pheromone levels during dauer exit. We found that high ins-6 expression in the ASJ neurons reflects commitment to exit the dauer state. Through calcium imaging in dauers, we show that sensory integration of food and pheromone can also be seen at the level of sensory neuron activity: ASJ calcium levels increase in response to food, but this response can be suppressed by adding pheromone. We find that ins-6 upregulation during dauer exit depends on ASJ neuronal activity, cGMP signaling, a CaM-kinase pathway, and is inhibited most potently by the pheromone component ascr#8. Altogether, our data show how the ASJ neurons integrate food and pheromone levels both in the short-term through calcium transients and in the longer term through transcription of ins-6, thereby highlighting how neuropeptides can provide the bridge from ephemeral sensory information to sustained physiological changes.

Results

A Screen of ASJ-Enriched Neuropeptide Genes Validates ins-6 as Critical for Dauer Exit.

Neuronal ablation methods using a laser microbeam or transgenic caspases have shown that the chemosensory ASJ neurons are important for dauer exit (21, 23, 24), but the precise molecular mechanism by which the ASJ neurons promote dauer exit remains unclear. Having previously shown that neuropeptides are collectively required for dauer entry (25), we reasoned that neuropeptides could also be involved in dauer exit. Using a pheromone-based dauer exit assay (Fig. 1B), we tested five mutants defective in neuropeptide synthesis or secretion and found that four out of the five loss-of-function mutants (sbt-1/7BT, egl-3/PC2, egl-21/CPE, and unc-31/CAPS) had significantly lower dauer exit rates than that of wild-type (Fig. 1C), strongly indicating that neuropeptide signaling is required for exit. By contrast, ric-7 loss-of-function mutants, which are impaired for dense core vesicle secretion (26), exited dauer at rates higher than that of wild-type, suggesting that RIC-7 may impair the secretion of different neuropeptides than those affected by the other four mutants.

We hypothesized that the ASJ neurons may utilize neuropeptides to promote dauer exit and tested this by screening for defects in dauer exit rates in mutants lacking ASJ-enriched neuropeptides. We analyzed loss-of-function mutants for eleven insulin-like peptide genes, two neuropeptide-like peptide genes, and two FMRF-amide-like peptide genes that were predicted to be enriched in ASJ based on single-cell RNA-sequencing data (27) (Fig. 1D and SI Appendix, Fig. S1A). Of the fifteen genes tested, ins-6 mutants showed the most severe dauer exit defect as they exited dauer at the lowest rates compared to wild type. ins-32 mutants showed a modest dauer exit defect, while mutants defective for ins-26, nlp-80, flp-15, or flp-34 showed higher than wild-type dauer exit rates, suggesting that the neuropeptides those genes encode may inhibit dauer exit. Future cell-specific knockdown experiments in ASJ will help to ascertain whether ins-6 acts primarily from ASJ vs. other neurons.

ins-6 Transcription and INS-6 Secretion Increase Quickly during Dauer Exit.

We focused on characterizing the expression pattern of ins-6 because loss of ins-6 resulted in the most severe impairment in dauer exit of the neuropeptide genes we tested. To characterize how ins-6 expression responds to environmental improvement during dauer exit, we analyzed ins-6 transcription and INS-6 secretion using genomically integrated fluorescence reporters that were simultaneously integrated into the same strain (Fig. 2A and SI Appendix, Table S1). To measure ins-6 transcription, we constructed ins-6p::destabilized-YFP (dYFP), a transcriptional reporter based off a previous design (21) that fuses both upstream and downstream regulatory regions of ins-6 to dYFP, a modified YFP variant with a shorter half-life that provides higher temporal resolution (28). To measure INS-6 secretion, we constructed ins-6::mCherry, a translational reporter that uses the ins-6 promoter to drive expression of a fusion between the INS-6 propeptide and mCherry, a protein whose secretion can be tracked using a coelomocyte uptake assay (29) in which fluorescence intensity within the coelomocytes reflects the amount of neuropeptide secreted into the body cavity. We built a strain containing simultaneously integrated copies of both these reporters (SI Appendix, Table S1) and concurrently measured YFP signal in head neurons and mCherry signal in coelomocytes (Fig. 2B).

Fig. 2.

ins-6 transcription and INS-6 secretion increase quickly during dauer exit. (A) Design of ins-6 transcriptional and translational reporters. (B) Sample images of merged dYFP and mCherry channels. Left, dauer prior to transfer. Right, dauer 6 h after transfer. Anterior is Left, ventral is facing the viewer. (C–E) Quantification of ins-6p::dYFP transcriptional reporter activity measured in arbitrary units (a.u.) in ASJ (C) or ASI (D) and mCherry signal in the coelomocytes (E) in dauers and following transfer of dauers to favorable conditions (indicated by curved arrow). Plots are from the same set of animals. ns, not significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by Welch ANOVA with Dunnett’s T3 multiple comparison correction when compared to “Dauer.” (F) Representative images of mRNA FISH analysis for ins-6 mRNA in a dauer and an exiting dauer 3 h after transfer (h.a.t.) to favorable conditions. (G) mRNA FISH signal quantification in the ASJ neurons measured in arbitrary units (a.u.). *P < 0.05, **P < 0.01 by Welch ANOVA with Dunnett’s T3 multiple comparison correction when compared to “Dauer.” (H–J) Additional ins-6 reporter activity experiments performed similarly to (C–E), except with additional time points, and LED power and exposure time were lowered relative to the experiments from (C–E) to prevent pixel saturation. “Dauer (early)” and “Dauer” refer to dauer animals obtained 48- and 65-h after incubation, respectively. h.a.t., hours after transfer to favorable conditions. In all plots, individual dots represent one animal. Medians are depicted by the orange bar. Medians and sample sizes are written.

ins-6 transcriptional reporter activity was nearly undetectable in dauers (Fig. 2 C and D), although we did observe a slight increase in reporter activity in the ASJ neurons of animals that recently entered dauer (Fig. 2H), in accordance with previous observations (21) (Discussion). This signal then returned to baseline after animals remained in the dauer state for ~12 h. When these dauers were transferred to favorable conditions to induce dauer exit, ins-6 transcriptional reporter activity in the ASJ neurons increased after just 1 h (Fig. 2C) and continued to increase for multiple hours thereafter (Fig. 2 C and F). ins-6 transcriptional reporter activity also increased in the ASI chemosensory neurons albeit at a slower rate and to a lower maximum relative to the ASJ neurons (Fig. 2 D and G). To validate these results, we performed fluorescence in situ hybridization (FISH) for ins-6 mRNA (Fig. 2 F and G). Consistent with our fluorescence reporter results, our FISH data indicated that ins-6 transcripts increased just 1 h after transfer and continued to increase afterward. Concordant with results using our ins-6 transcriptional reporter, we observed that ins-6::mCherry translational reporter activity in the coelomocytes increased within the first hour following transfer to favorable conditions and continued to increase for multiple hours thereafter (Fig. 2 E and H).

Under reproductive growth conditions, which result in nondauer adult development, our ins-6 transcriptional reporter showed no signal in ASJ throughout development (SI Appendix, Fig. S2A). ins-6 transcriptional reporter activity in ASI (SI Appendix, Fig. S2B) and translational reporter activity (SI Appendix, Fig. S2C) in the coelomocytes peaked in L1 larvae and then decreased throughout development. These results suggest that favorable conditions cause increased ins-6 reporter activity in ASJ specifically in the context of dauer animals.

The observed increase in INS-6 secretion could result from an increase in ins-6 transcription, INS-6 processing, INS-6 packaging, and/or dense core vesicle release. To parse these different factors, we built a translational reporter driven by a constitutively active ASJ-specific promoter, trx-1p::ins-6::mCherry, and again measured coelomocyte uptake of INS-6::mCherry (SI Appendix, Fig. S2D). The mCherry signal in coelomocytes increased when dauers were transferred to favorable conditions in a manner similar to animals bearing the ins-6p::ins-6::mCherry reporter transgene, suggesting that secretion of INS-6 is also controlled at the level of dense core vesicle release.

daf-28 Expression Patterns Suggest a Weaker Role in Dauer Exit versus ins-6.

DAF-28, another insulin-like peptide that promotes nondauer development during the dauer entry decision (30), has previously been shown to be a weaker regulator of dauer exit relative to INS-6 (21, 22). Consistent with those observations, loss of daf-28 gene function did not result in a defect in dauer exit rates in an otherwise wild-type background but did enhance the dauer exit rate defect of ins-6 loss-of-function mutants (Fig. 1D and SI Appendix, Fig. S1B). To understand how the spatiotemporal regulation of daf-28 compares with that of ins-6, we performed similar transgenic reporter experiments to study daf-28 (SI Appendix, Fig. S2 E–H).

A daf-28p::dYFP transcriptional reporter showed virtually no activity in the ASJ neurons (SI Appendix, Fig. S2F) of dauers after being transferred to favorable conditions but did show slight activity increase in the ASI neurons (SI Appendix, Fig. S2G). daf-28::mCherry translational reporter activity measured in the coelomocytes of dauers did not increase even 6 h after transfer to favorable conditions (SI Appendix, Fig. S2H). In contrast, growth under nondauer-inducing conditions resulted in strong daf-28p::dYFP transcriptional reporter activity in the ASI neurons (SI Appendix, Fig. S2G), weaker activity in the ASJ neurons (SI Appendix, Fig. S2F), and strong activity in the coelomocytes (SI Appendix, Fig. S2H), with activity for both reporters peaking after animals reached adulthood. Collectively, our transgenic reporter lines for ins-6 and daf-28 show inverse expression patterns: ins-6 reporter activity increases during dauer exit but less so during reproductive growth, and vice versa for the daf-28 reporters. These findings align with both ours’ and others’ dauer exit assay results in which INS-6 plays a stronger role in regulating dauer exit than does DAF-28 (21, 22).

ins-6 Expression Reflects Commitment in the Dauer Exit Decision.

ins-6 expression increases quickly in all dauers when they are transferred to strongly favorable conditions that stimulate 100% of dauers to exit, but what happens when conditions are more ambiguous such that only a fraction of dauers exit? We transferred dauers to a lower pheromone concentration that induces approximately half of dauers to exit (hereby referred to as “intermediate-pheromone conditions”) and measured ins-6p::dYFP transcriptional reporter activity (Fig. 3A). At 3 h after transfer, all worms showed a slight increase in ins-6p::dYFP signal, while at 6 h after transfer and beyond, a clear bimodality of the population emerges: The worms that would eventually go on to exit dauer [as evidenced by a wider pharynx (Fig. 3B), which is a key morphological characteristic of dauer exit (31)], showed high ins-6 transcriptional dYFP reporter activity, whereas the worms that should remain as dauers showed a return to baseline levels of dYFP signal, resembling the level of signal seen in dauers.

Fig. 3.

ins-6 expression correlates with commitment to exit dauer. (A) ins-6p::dYFP transcriptional reporter activity in dauers before and after transfer (indicated by curved arrow) onto intermediate-pheromone conditions that stimulate half of dauers to exit. (B) Animals from (A) simultaneously had their pharynx widths measured. Shown are correlation plots between ins-6p::dYFP signal and pharynx width. Dotted lines are drawn at the median for each measurement. (C) Dauers were induced under high-pheromone growth conditions and transferred to intermediate-pheromone conditions for the indicated duration before being transferred back onto high-pheromone conditions to prevent dauer exit in noncommitted animals. Animals were scored for dauer exit after 24 h. Included are the following controls: a no-transfer control (NT, light orange) in which dauers were not transferred but instead remained on intermediate-pheromone conditions and a control in which dauers remained entirely on high-pheromone conditions (Ctrl, gray). Bars and numbers indicate dauer exit rates for >100 animals per sample.

To determine whether ins-6 expression could be used to distinguish dauers that are committed versus noncommitted, we compared the time course of ins-6 expression with that of dauer exit commitment (Fig. 3C). To examine dauer commitment, we transferred dauers from high-pheromone, dauer-maintaining conditions to intermediate-pheromone conditions that permitted approximately half the animals to exit dauer. Then, after different time intervals, we transferred the animals back onto high pheromone conditions so that animals that had not committed to exiting dauer would be induced to remain as dauers (Fig. 3 C, Top); this protocol was analogous to previous work that established commitment to exiting dauer (13). Within 3 h of exposure to intermediate-pheromone conditions, approximately half the animals that might exit dauer under such conditions committed to exiting dauer (18% committed, compared to a baseline of 42%), while roughly the total pool of dauers that would eventually exit under intermediate-pheromone conditions committed within 6 h (43%) (Fig. 3C). These observations indicate that while some dauers irreversibly commit in 1 h following transfer to intermediate-pheromone conditions, other dauers can take between 3 and 6 h to commit to the exit decision. This 6-h mark matches the time point in our imaging experiments in which the clearest bimodality of both ins-6p::dYFP signal as well as pharynx width emerges (Fig. 3B), suggesting that such bimodality can reflect decision commitment.

ins-6 Expression Reflects a Food:Pheromone Ratio.

Since ins-6 transcriptional reporter activity responded differently when dauers were transferred to no-pheromone versus intermediate-pheromone conditions (compare Fig. 2C to Fig. 3A), we asked whether ins-6 expression responds to a food:pheromone ratio—the principal metric by which dauers decide to exit (14). Dauers bearing the ins-6 transcriptional reporter were transferred onto plates containing differing amounts of pheromone and of a food signal (Fig. 4A). We chose crude yeast extract as our food signal because of the following reasons: It promotes dauer exit in our pheromone-based assay (SI Appendix, Fig. S4A), it is more easily incorporated into agar plates, and because yeast extract is the source material from which a purified food signal was derived in the original dauer exit studies demonstrating the importance of a food:pheromone ratio during dauer exit (14).

Fig. 4.

ins-6 expression reflects food and pheromone integration. (A) ins-6p::dYFP transcriptional reporter activity in dauers before transfer (“Dauer”) and 3 h after being transferred to varying concentrations of food signal (yeast extract) and pheromone. Note that the 0.05% w/v pheromone + 0.05% w/v yeast extract sample is shared between the Left and Right halves of the graph. (B and C) ins-6p::dYFP transcriptional reporter activity from ASJ (Left) and ins-6::mCherry translational reporter activity from coelomocytes (Right) were measured in wild-type dauers (B) or daf-7 loss-of-function dauers (C) before (“Dauer”) and after being transferred to plates with different amounts of food and pheromone (“+,” 0.5% yeast extract or 0.5% pheromone; “−,” 0% yeast extract or 0% pheromone). ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by the Kruskal–Wallis test with Dunn’s multiple comparison correction when compared to “Dauer.” For all graphs: Each dot represents one animal. Medians are depicted by the orange bar. Medians and sample sizes are written.

We found that ins-6 transcriptional reporter activity correlated positively with food:pheromone ratios (Fig. 4A), consistent with ins-6 being up-regulated under favorable conditions during dauer exit. Using live OP50 instead of yeast extract was not more effective in increasing transcriptional reporter activity (SI Appendix, Fig. S4B). To assess whether food or pheromone was the stronger input, we transferred dauers to plates containing either no pheromone or high amounts of pheromone and either no food signal or high amounts of food signal and analyzed ins-6 transcription and INS-6 secretion using our transcriptional and translational reporters, respectively (Fig. 4B). At high pheromone levels, both transcriptional and translational reporter activity remained low regardless of whether high yeast extract was present, indicating that pheromone can suppress the ability of food signal to increase INS-6 production and secretion.

Having determined that ins-6 is up-regulated under conditions that promote dauer exit, we asked whether this upregulation happens early during dauer exit (i.e., happens due to sensory perception of environmental improvement), or whether it might be an effect of the dauer exit process. To explore this possibility, we analyzed our ins-6 transcriptional and translational reporters in a daf-7 mutant background. These animals are defective for daf-7, which encodes a TGF-β-like growth factor (20, 32). These mutants constitutively form and remain dauers at 25 °C even under no-pheromone conditions. We repeated the experiment shown in Fig. 4B but kept the transferred worms at 25 °C in order to prevent dauer exit. Despite all animals remaining as dauers, these daf-7 mutants exhibited near-identical ins-6 regulation with regard to both transcription and secretion when compared to that in a wild-type background (Fig. 4C). INS-6 production and secretion is thus up-regulated in response to a high food:pheromone ratio that promotes dauer exit, even in animals incapable of exiting dauer, and such upregulation is not simply a downstream effect of dauer exit.

Calcium Imaging in Dauers Shows Integration of Food and Pheromone in ASJ Neurons.

To evaluate how the ASJ neurons respond to food and pheromone, we monitored calcium dynamics in dauers expressing the genetically encoded calcium indicator GCaMP6s (33) in the ASJs (34). We exposed dauers to a food signal (the supernatant from an overnight culture of OP50 grown in LB), pheromone, or a combination of both (Fig. 5A and SI Appendix, Fig. S5A). We found that the ASJ neurons were activated in response to food signal at both stimulus onset and slightly after stimulus offset. The ASJs did not respond to pheromone, but we observed that when the ASJs were exposed to a mixture of food signal and pheromone, the response looked nearly identical to the response to pheromone alone. This finding suggests that ASJ calcium levels integrate food and pheromone. Although the food signal we used differed between calcium and transcriptional reporter imaging experiments, we found that LB (which contains 0.5% w/v yeast extract) increased calcium levels in the ASJs of L4 larvae (SI Appendix, Fig. S5C), consistent with yeast extract being an activator of ASJ neuronal activity.

Fig. 5.

ASJ neuronal activity shows sensory integration and is necessary for ins-6 upregulation. (A and B) Calcium traces of ASJ in wild-type (A) or unc-31 loss-of-function mutant (B) dauers in response to pheromone, food (bacterial supernatant), or a mixture of both. Shown are the mean ± SEM from 10 to 12 ASJ neurons from six animals. Individual traces can be found in SI Appendix, Fig. S5. (C) ins-6p::dYFP transcriptional reporter activity in the ASJ neurons was measured in wild-type or unc-31 loss-of-function mutant dauers (with or without various rescue constructs that express unc-31 cDNA panneuronally or specifically in ASJ) 3 h after transfer to favorable conditions. ns, not significant, **P < 0.01, ****P < 0.0001 by one-way ANOVA with Dunnett’s multiple comparisons correction compared to the unc-31 mutant strain without rescue. (D) ins-6p::dYFP transcriptional reporter activity in ASJ was measured in wild-type or unc-13 loss-of-function mutant dauers before and 3 h after transfer to favorable conditions. ns, not significant, by the unpaired t test. (E) Dauers expressing the histamine-gated chloride channel in the ASJ neurons (trx-1p::HisCl-T2A-mKate2) were transferred to favorable plates with or without histamine and imaged 3 h later for ins-6p::dYFP transcriptional reporter activity in of the ASJ neurons. ***P < 0.001 by the Mann–Whitney U test. (F) Image of a mosaic animal in which the trx-1p::HisCl-T2A-mKate2 extrachromosomal array only expressed in one ASJ. (G and H) Dauers expressing the red-shifted rhodopsin Chrimson (G) or the capsaicin-activated cation channel TRPVI (H) in ASJ were subjected to the stimulation conditions shown while on high-pheromone conditions and imaged two (F) or three (G) hours later for ins-6p::dYFP transcriptional reporter activity in the ASJ neurons. ns, not significant. *P < 0.05 by the Mann–Whitney U test. In all graphs: Each dot represents one animal. Medians are depicted by the orange bar. Medians and sample sizes are written.

Involvement of Other Neurons in Regulating ins-6 Expression within ASJ.

We tested whether regulation of ins-6 in response to food and pheromone signals in the ASJ neurons depends on signals received from other neurons by measuring ins-6 transcriptional reporter activity in unc-31/CAPS or unc-13 loss-of-function mutant backgrounds, which are considered defective for neuropeptide secretion and for neurotransmitter release, respectively (29, 35). While loss of unc-31 abrogated ins-6 upregulation (Fig. 5C), loss of unc-13 did not produce a noticeable effect (Fig. 5D), suggesting that neuropeptide signaling, but not fast-acting synaptic transmission, is essential for ins-6 upregulation. Using an ASJ-specific promoter to express unc-31 cDNA in the unc-31 mutant background partially rescued the low levels of ins-6 transcriptional reporter activity (Fig. 5C), implying positive autoregulation: Each ASJ neuron might secrete a peptidergic signal to be received by itself or its bilateral partner to increase ins-6 expression. The lack of a full rescue in two of three independent lines suggests additional neurons may contribute to ins-6 upregulation by secreting neuropeptides in an unc-31-dependent manner. In support of an autoregulatory model, we performed calcium imaging in the ASJ neurons of unc-31 mutants and found that the responses to food, pheromone, and a mixture of the two remained largely unchanged compared to those of wild type (Fig. 5B and SI Appendix, Fig S5B). Collectively, these results imply that when it comes to calcium response, each ASJ neuron possesses the molecular machinery to detect and integrate food and pheromone signals.

In accordance with ins-6 partaking in a feedback loop among ASJ neurons, we found that in daf-2 loss-of-function mutants, which are defective for the insulin/IGF-1 receptor (IGFR) ortholog DAF-2, ins-6 transcriptional upregulation was impaired (SI Appendix, Fig. S5D). Transfer of daf-2 mutant dauers to no-pheromone conditions resulted in a small increase in ins-6 transcriptional reporter activity compared to that of wild-type dauers (compare SI Appendix, Fig. S5D to Figs. 2C and 4B).

Activity Dependence of ASJ on ins-6 Upregulation.

We evaluated the relationship between neuronal activity and upregulation of ins-6 during dauer exit. To test whether ins-6 upregulation depends on neuronal activity, we chemogenetically silenced the ASJ neurons using cell-specific expression of a histamine-chloride gated channel (36) during dauer exit by transferring dauers to no-pheromone conditions with or without histamine and measuring ins-6 transcriptional reporter activity. We found that inhibiting ASJ activity severely repressed ins-6 upregulation (Fig. 5E), and that this effect occurs cell autonomously since, in a mosaic animal in which only one ASJ bears the HisCl transgene, only the nontransgenic ASJ showed an increased ins-6p::dYFP signal (Fig. 5F)

We also tested whether chemogenetic or optogenetic activation of ASJ could induce ins-6 expression even under high pheromone conditions (Fig. 5 G and H). Optogenetic stimulation of ASJ via the red-shifted rhodopsin Chrimson (37) or capsaicin-induced chemogenetic stimulation of ASJ via the TRPV1 channel (38) both slightly increased ins-6 transcriptional reporter activity, but to a much lesser extent when compared to dauers transferred to favorable conditions (compare Fig. 5 G and H with Figs. 2C and 4B). These loss- and gain-of-function experiments demonstrate that neuronal activity promotes ins-6 expression during dauer exit.

Molecular Regulation of ins-6 in ASJ during Dauer Exit.

We sought to characterize the molecular components responsible for ins-6 upregulation during dauer exit. Since cGMP functions as a messenger in a variety of ASJ-mediated behaviors including dauer development, pathogen avoidance, hydrotaxis, and phototaxis (39–43), we tested mutants defective for genes in the cGMP pathway. Mutations in daf-11 (encoding a transmembrane guanylyl cyclase), tax-2, or tax-4 (which encode subunits of a cGMP-gated channel) (44, 45) all abrogated ins-6 transcriptional reporter activity in the ASJ neurons of dauers transferred to favorable conditions (Fig. 6A). Addition of the nonhydrolyzable cGMP analog, pCPT-cGMP, increased ins-6 transcriptional reporter activity to high levels even in the presence of pheromone (Fig. 6B). Collectively, these results demonstrate that cGMP signaling is both necessary and sufficient for ins-6 upregulation during dauer exit.

Fig. 6.

ins-6 upregulation is regulated by cGMP signaling and ascr#8. (A) daf-11, tax-2, and tax-4 loss-of-function mutant dauers were transferred to favorable conditions and measured for ins-6p::dYFP transcriptional reporter activity in ASJ 3 h later. ****P < 0.0001 by the Mann–Whitney U test for daf-11. ***P < 0.001, ****P < 0.0001 by the Kruskal–Wallis test with Dunn’s multiple comparison correction for tax-2 and tax-4 compared to “WT.” (B) Dauers were transferred to high pheromone plates containing 1 mM pCPT-cGMP or H2O control and measured for ins-6p::dYFP transcriptional reporter activity in ASJ 3 h later. ****P < 0.0001 by the Mann–Whitney U test. (C) Dauers were transferred to plates containing ascarosides at 250 nM or 1 µM and measured for ins-6p::dYFP transcriptional reporter activity in ASJ 3 h later. ns, not significant, **P < 0.01, ****P < 0.0001 by the Kruskal–Wallis test with Dunn’s multiple comparison correction compared to “EtOH Ctrl.” (D) srw-97; dmsr-12 loss-of-function mutants were transferred to ascr#8 (1 µM) or EtOH control plates and measured for ins-6p::dYFP transcriptional reporter activity in ASJ 3 h later. ns, not significant, ***P < 0.001 by the Mann–Whitney U test. (E) Synchronized cultures were grown in the presence of food for 3 d and then additional bacterial food was either given (“Fed”) or withheld (“Starved”). 24 h later, cultures were prepared and subjected to HPLC analysis to measure the amount of ascr#8 present. Each dot is one biological replicate. For (A–D): Each dot represents one animal. Medians are depicted by the orange bar. Where applicable, medians and sample sizes are written.

We also tested the cGMP-dependent kinase EGL-4/PKG (46, 47) and found that egl-4(n479) loss-of-function mutants had significantly lower ins-6 transcriptional reporter activity in ASJ, although we could not rescue ins-6 reporter activity in ASJ using cell-specific rescues for ASJ (trx-1p) (48), ASI/AWA (gpa-4p) (49), or AWC (ceh-36Δp) (50) (SI Appendix, Fig. S6A), suggesting that EGL-4 may act in other cells.

To test the role of neuron activity-dependent genes, we measured ins-6 transcriptional reporter activity in loss-of-function mutants for cmk-1, encoding the C. elegans homolog of calcium/calmodulin-dependent kinase CaMKI (51), and crh-1, encoding the C. elegans homolog of the CREB transcription factor (52). Both genes have been previously implicated in regulating expression of neuropeptide-encoding genes and DAF-7, the dauer-inhibiting TGF-β-like growth factor (43, 49, 53, 54). crh-1 mutants showed only a small decrease in reporter activity, while cmk-1 mutants displayed a substantial loss in ins-6 transcriptional reporter activity (SI Appendix, Fig. S6B). Similar to our results with egl-4, neuron-specific expression of cmk-1 in ASJ, ASI/AWA, or AWC failed to rescue ins-6 reporter activity in the ASJ neurons (SI Appendix, Fig. S6C), suggesting that CMK-1 may act in other cells. Moreover, we observed that cmk-1 mutants were slow to form dauers: Even after an additional 1 to 2 d of incubation time, only ~10% of animals formed dauers while the remaining still remained as L2 or L2d larvae, indicating there may be pleiotropic effects caused by loss of cmk-1 that indirectly affects ins 6 expression.

Dauer pheromone comprises multiple ascarosides (nematode-specific glycosides of the dideoxy sugar ascarylose) of which ascr#2, ascr#3, ascr#5, and ascr#8 are considered potent inducers of dauer entry (16, 17). To determine which ascaroside components of the dauer pheromone modulate ins-6 expression during dauer exit, we transferred dauers to plates supplemented with individual ascarosides at concentrations previously reported to induce dauer entry (55–57) (Fig. 6C). We found that ascr#8 most effectively suppressed ins-6p::dYFP expression, particularly at 1 µM, while ascr#2 mildly suppressed ins-6p::dYFP expression. ascr#3 and ascr#5 did not have a statistically significant effect at either 250 nM or 1 µM.

Since ascr#8 most strongly suppressed ins-6 transcriptional reporter activity in dauers, we asked whether ascr#8 conveys population density and/or nutritional status by measuring ascr#8 levels in well-fed versus starved liquid cultures via liquid chromatography–mass spectrometry (LCMS). We found that ascr#8 levels were unquantifiable in well-fed cultures but highly abundant in starved cultures (Fig. 6E), indicating that ascr#8 likely signals population-wide starvation. We also tested the GPCRs DMSR-12 and SRW-97, previously shown in C. elegans males to be involved in reception of ascr#8 in the context of attraction to hermaphrodites (58), for their roles in regulating ins-6 expression (Fig. 6D). A double mutant defective for both dmsr-12 and srw-97 showed a wild-type ins-6 transcriptional reporter activity response to ascr#8, suggesting alternative receptors may be responsible for ascr#8 response in the context of dauer exit.

Discussion

ins-6 Highlighted among a Screen of ASJ-Enriched Neuropeptides.

Having tested fifteen neuropeptide genes enriched in the ASJ chemosensory neurons, we found that ins-6 loss-of-function mutants exited dauer at the lowest rates compared to wild type (Fig. 1D). While ins-6 has previously been implicated in dauer exit (21, 22), this work specifically tests the role of INS-6 function in an otherwise wild-type background for dauer exit. Loss-of-function mutations in ins-32 also resulted in a strong dauer exit defect, indicating that INS-32 may act alongside INS-6 to promote dauer exit. Loss-of-function mutants for neuropeptide genes such as ins-26, nlp-80, flp-15, and flp-34 had higher dauer exit rates than those of wild type, suggesting that those neuropeptide genes may be involved in inhibiting dauer exit. The decision to exit dauer may thus be partially calculated based on a ratio between exit-promoting versus exit-inhibiting neuropeptides and growth factors.

While the C. elegans genome includes over 40 genes encoding insulin-like peptides, ins-6 stands out as being the only gene whose encoded peptide INS-6 has been shown to biochemically bind and activate the human insulin receptor with high affinity despite bearing key structural differences such as the lack of the conserved C peptide compared to human insulin (59, 60). Such biochemical data provide one possible explanation for the potency of INS-6 in regulating dauer exit. Yet, the fact that ins-6 loss-of-function mutants can still exit dauer under low-pheromone conditions indicates that INS-6 likely acts alongside other insulin-like peptides (such as DAF-28 and possibly INS-32) to induce dauer exit.

ins-6 Expression Dynamics: Implications and Discrepancies with Previous Studies.

Using both fluorescence reporter assays (Fig. 2 C–H) and mRNA FISH (SI Appendix, Fig. S2 F and G), we show that ins-6 expression and INS-6 secretion increase in just 1 h after transfer of dauers onto favorable conditions. In that same time frame, we could not observe any behavioral or morphological changes that would distinguish these dauers as exiting, suggesting that ins-6 acts early in the dauer exit process. Our ins-6 transcriptional reporter showed high activity in the ASJ neurons specifically during dauer exit and virtually no activity when animals were raised under reproductive growth conditions leading to nondauer adults (SI Appendix, Fig. S2 A–C). These data contrast with previous single-cell RNA sequencing data and mRNA FISH data showing that ins-6 transcripts are abundant in the ASJ neurons of L4 larvae raised under reproductive growth conditions (27, 61). These discrepancies suggest that our ins-6 reporters contain regulatory regions responsible for dauer-specific transcriptional activity while lacking additional regulatory regions that promote expression in the ASJ neurons during nondauer-inducing growth.

Others have reported ins-6 transcriptional reporter activity in the ASJ neurons of dauer animals (21), while we saw modest reporter activity in recently formed dauer animals and virtually no ins-6 expression in dauer animals that had already been dauers for ~12 h or longer (Fig. 2 C, G, and H). Our results suggest that there may be a temporary increase in ins-6 expression upon entering dauer that then fades over time as animals remain in the dauer state. Therefore, the levels of ins-6 expression measured in dauers may depend on both the age of the dauer as well as the half-life of the molecular substrate analyzed (e.g., a stable fluorescent protein, a destabilized fluorescent protein, or mRNA).

Model of ins-6 Regulation during Dauer Exit.

Our analysis of ins-6 regulation led us to propose the following model of how ins-6 expression is regulated during dauer exit (Fig. 7). GPCRs for food signals activate DAF-11 which increases cGMP levels and opens cGMP-gated heterodimeric TAX-2/TAX-4 cation channels to allow calcium through. The resulting calcium influx could lead to a combination of calcium-dependent transcription of ins-6 as well as exocytosis of INS-6-containing dense core vesicles, which has canonically been associated with high calcium levels (29, 62–65). Pheromone likely acts upstream of cGMP signaling because addition of pCPT-cGMP increases ins-6 expression even under high-pheromone conditions (Fig. 6B). Therefore, pheromone could prevent calcium influx by blocking food-based activation of DAF-11.

Fig. 7.

Model for ins-6 upregulation during dauer exit. In each ASJ neuron, GPCRs specific to food signals trigger a G-protein signaling pathway that activates the transmembrane guanylyl cyclase DAF-11 to increase cGMP production. GPCRs specific to pheromone (including the ascarosides ascr#8 and ascr#2) inhibit this activation. cGMP increases calcium influx through CNG-gated channels such as TAX-2/TAX-4, and the resulting calcium increase can both activate calcium-dependent transcription and promote secretion of neuropeptides, which may include INS-6, in a process dependent on UNC-31/CAPS. cGMP might also activate cGMP-dependent transcription pathways to directly up-regulate ins-6 expression. The secreted neuropeptides bind to receptors on other neurons (including the ASJ neurons) and, upon reaching a certain threshold, triggers a positive feedback loop that further increases production and secretion of INS-6 from the ASJ neurons. This positive feedback promotes commitment to exit the dauer state by causing an irreversible buildup of INS-6 production and secretion. INS-6 then signals to downstream tissues that execute the relevant dauer exit gene programs. Dashed gray lines indicate processes for which the intermediary components are unknown.

We ruled out that calcium-dependent transcription specifically in the ASJ neurons could be mediated by cell-autonomous activity of the calcium/calmodulin-dependent kinase CaMKI (SI Appendix, Fig. S6C), although CaMKI could be acting in other cells. We speculate that additional calcium-dependent kinases or phosphatases such as UNC-43 and TAX-6•CNB-1, the C. elegans orthologs of CAMKII and Calcineurin, respectively, may be involved as they have previously been shown to regulate the IIS pathway through the DAF-16/FoxO transcription factor (66). It remains to be seen whether DAF-16 regulates ins-6 expression.

cGMP may also increase transcription of ins-6 via a calcium-independent mechanism, as has been shown for other neuropeptide- or growth factor-encoding genes via EGL-4/PKG (43, 54). Our results implicate EGL-4/PKG as possibly important for ins-6 upregulation, but our cell-specific rescue experiments suggest that EGL-4/PKG may not act cell autonomously in the ASJ neurons (SI Appendix, Fig. S6A). A cGMP-dependent, calcium-independent mechanism of ins-6 upregulation would help explain why optogenetic or chemogenetic activation of the ASJ neurons was only sufficient to slightly increase ins-6 expression under high-pheromone conditions (Fig. 5 G and H), since calcium alone would be unable to activate cGMP-dependent pathways. Further molecular studies will be required to fully elucidate the components of the calcium- and cGMP-dependent pathways involved.

Positive Feedback as a Driver of Decision Commitment.

Our observations suggest that ins-6 expression levels can mark decision commitment. When dauers are transferred to intermediate-pheromone conditions, a clear bimodality emerges in ins-6 transcriptional reporter activity around the same time that dauers commit to exiting (Fig. 3). The increase in ins-6 transcriptional reporter activity in exit-committed dauers likely results from a feedback loop that depends on UNC-31 and DAF-2/IGFR (Fig. 5D and SI Appendix, Fig. S5D). We propose that after dauers encounter more favorable conditions, the ASJ neurons produce and secrete small amounts of neuropeptides (which may include INS-6) that bind to receptors on other neurons (including on the ASJ neurons themselves) and trigger a positive feedback loop that promotes further production of INS-6. Our observations that mutating unc-31 nearly abrogated ins-6 expression during dauer exit but spared the ASJ calcium response to food and pheromone (Fig. 5 B and D) suggests that calcium increases are involved upstream of this putative positive feedback loop. Our finding that chemogenetic silencing of the ASJ neurons (which likely inhibits calcium influx) prevents ins-6 upregulation (Fig. 5 E and F) further suggests that neuronal calcium may help trigger this positive feedback loop. Such a rampant, irreversible increase in INS-6 production and secretion likely promotes commitment to exit the dauer state by acting on downstream tissues to induce relevant dauer exit-related genetic programs via insulin-like signaling (60).

The nature of this proposed positive feedback loop would explain two features of the dauer exit decision. First, it would explain the all-or-nothing nature of dauer exit in which dauers committed to exiting must proceed fully with decision execution since rampant positive feedback would hinder suppression of that same signaling pathway. Such positive feedback mechanisms have previously been reported in developmental decisions involving commitment including in yeast, flies, frogs, and worms (67–70). Second, the positive feedback loop would help connect the timescale discrepancy between sensory perception (over seconds) and decision execution (over hours). Dauers must be exposed to favorable food:pheromone ratios for a sufficiently long period of time to trigger the level of positive feedback required for decision commitment.

Ethological Significance of ins-6 Expression Dynamics in ASJ.

How does our model relate to what C. elegans may encounter in its natural habitat during dauer exit? In its natural life cycle, C. elegans cycles through periods of favorable conditions and unfavorable conditions (boom and bust), and dauer exit plays a critical role in supporting the transition from bust to boom. Expression levels of ins-6 can convey two important pieces of information: presence of food and absence of pheromone [including ascr#8, which is secreted by starved populations (Fig. 6E)]. We envision that, in the wild, C. elegans larvae enter dauer as they grow up in an environment crowded by pheromone-secreting worms and a depleted food source. To exit dauer, C. elegans would need to travel away from its crowded environment and find a new food source. The presence of food alone is insufficient because high pheromone levels indicate that, even if food were present, it would be depleted too quickly given the abundance of other worms. Our results are consistent with this scenario in that high pheromone levels suppress food response at both the calcium and ins-6 expression levels (Figs. 4B and 5A).

Sensory Integration in a Single C. elegans Sensory Neuron.

Our genetic and calcium imaging experiments indicate that ins-6 upregulation during dauer exit can occur cell autonomously within ASJ. unc-13 loss-of-function mutants defective for synaptic transmission had close to wild-type ins-6 upregulation in ASJ (Fig. 5D), while unc-31 loss-of-function mutants defective for neuropeptide transmission had low ins-6 expression levels that could be partially rescued by ASJ-specific expression of unc-31 cDNA (Fig. 5C) but retained wild-type calcium responses to food and pheromone (Fig. 5B). Our data suggest that each ASJ neuron alone can integrate food and pheromone at both the level of neuronal calcium and gene expression, which highlights a stark difference between how C. elegans integrates sensory inputs compared to insects and vertebrates. The compact nervous system of C. elegans requires that it efficiently use its few neurons to discriminate against the vast array of sensory signals. Accordingly, C. elegans employs multiple polymodal sensory neurons that express multiple GPCRs to detect different stimuli (71), such as ASH which can detect a variety of noxious stimuli, AWA which responds to numerous chemoattractants (72, 73), and AWC which responds to both odors and temperature (74, 75). While vertebrates and insects do utilize polymodal neurons, most notably in nociception (76–78), sensory integration in such organisms is typically performed via a multilayer model in which modality-specific neurons converge onto higher-order brain regions (79, 80), a key example being the “one neuron—one receptor” principle observed in mammals and flies wherein each olfactory neuron class usually only expresses one olfactory receptor, and signals from each become integrated in further regions of the olfactory bulb (81, 82). Our results help us better understand how a single neuron performs complex computations of multiple sensory inputs and underscores how the computational power of nematode nervous systems (all about 300 neurons) is vastly underestimated by cell number alone.

Materials and Methods

Note: For fully detailed materials and methods, please see SI Appendix, Supplemental Materials and Methods.

C. elegans Strains and Maintenance.

C. elegans strains were derived from the wild-type strain N2 (Bristol) and were cultured according to standard laboratory conditions. A list of strains used in this study, including their genotypes and origins, can be found in SI Appendix, Table S1.

Dauer Entry Induction.

Dauer animals were obtained by picking young adult hermaphrodites onto plates supplemented with high concentrations of pheromone and allowing the adults to lay eggs. After ~5 to 9 h, adults were removed, additional food was provided, and plates were incubated at 25.5 to 26.1 °C for 60 to 72 h.

Dauer Exit Assay.

Dauers were selected for via SDS incubation and transferred onto plates containing a low amount of pheromone that induces roughly half of wild-type dauers to exit. After 24 h, dauer exit was scored according to morphological and behavioral criteria.

Transgenic Strain Construction.

Plasmids were microinjected into the gonads of young adult animals according to standard protocols (83, 84). Contents of injection mixtures can be found in SI Appendix, Table S1. Select transgenes were integrated according to standard X-ray irradiation protocols and outcrossed at least three times prior to use (85). Details for the construction of ins-6 reporter strains and loss-of-function mutants via CRISPR can be found in SI Appendix.

Microscopy and Image Analysis.

Animals were immobilized on a 4% ultrapure agarose pad in 5 mM levamisole in H2O. Imaging was performed on a Zeiss AxioImager2 equipped with a Colibri 7 for LED fluorescence illumination and an Axiocam 506 Mono camera (Carl Zeiss Inc.). Fluorescence quantification and pharyngeal width measurements were processed using FIJI. Quantification of secreted INS-6::mCherry or DAF-28::mCherry in coelomocytes was performed as described previously (29).

mRNA FISH.

FISH was performed similarly to a previously used protocol (61). ins-6 mRNA probes (20 total; see SI Appendix, Table S2 for probe sequences) were designed using the Stellaris RNA-FISH Probe Designer 2.0 (LGC Biosearch Technologies). FISH imaging was performed on an upright Zeiss LSM880 microscope (Carl Zeiss Inc.) equipped with a 594 nm laser using a Zeiss 63× oil objective. Image quantification was performed using FIJI.

Calcium Imaging.

To image dauers, we adapted an existing microfluidic chip design (86) by incorporating a thinner opening for the worm to accommodate dauers. Fluorescence was recorded with a spinning disc confocal microscope (Dragonfly 200, Andor) and a sCMOS camera (Photometrics Kinetix) that captured fluorescence from GCaMP6.0s (33) in the ASJ neurons of KP9672 (SI Appendix, Table S1). ΔF/F₀ was calculated for each stimulus–response, where F₀ was the average fluorescence value during the 2.5 s before delivery of the stimulus.

Perturbation Assays.

For optogenetic assays using Chrimson (37, 87), dauers were transferred onto pheromone-containing plates supplemented with 500 µM ATR or EtOH and placed in the Wormlab Tracking system (MBF Biosciences) equipped with a red LED light (stimulus regime: 100 ms on, 1,000 ms off at 100% intensity for 2 h).

Ascaroside Metabolomics.

3-d-old synchronized cultures containing day 1 gravid adults were provided either additional bacterial food or S-complete as a control and allowed to grow for a further 24 h. Cultures were collected, pelleted, frozen, lyophilized, and subjected to HPLC analysis according to the protocol detailed in SI Appendix to measure the amount of ascr#8 present.

Statistical Analysis, Plotting, and Figure Design.

Plots were designed and statistical tests were performed using Prism 10.0 (GraphPad). Illustrated figures were designed and drawn using Affinity Designer (Serif).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.