Fatty acid regulation of feeding in Caenorhabditis elegans reveals the potential ancestral origin of a GLP-1-like multiagonist signaling system

Significance

This study presents a paradigm for understanding appetite regulation in response to metabolic cues using Caenorhabditis elegans. In this organism, the ratio of dietary monounsaturated versus saturated fatty acids (MUFAs/SFAs) is sensed at the level of the endoplasmic reticulum through effects on the membrane stress sensor IRE-1. As a result, an excess of dietary SFAs suppresses PDF-1/PDFR-1 signaling in the brain, which increases dwelling on food and eating. The converse is true for MUFAs. The PDF-1/PDFR-1 signaling system has homology to the whole family of GLP-1/GIP related peptides and their receptors, thus providing insights into links between brain metabolism and development of GLP-1/GIP-related polyagonists therapies for treatment of obesity and associated metabolic disorders.

Abstract

Regulation of food intake in mammals is complex and controlled by an interplay between hedonic and homeostatic signals, including hormones like leptin, which senses fat storage and suppresses food intake. Caenorhabditis elegans lack leptin and leptin receptors but still exhibit controlled eating. Here, we show that in C. elegans eating can be regulated by a balance between saturated and monounsaturated fatty acids interacting with transcriptional pathways regulating lipid synthesis, c-AMP response element binding protein and AMP kinase. This effect is mediated at the endoplasmic reticulum through formation of phospholipids and activation of the IRE-1 sensor in the nervous system, which controls behavior through neuronal serotonin and the G-protein-coupled ligand/receptor pair PDF-1/PDFR-1. We show that this peptide/receptor pair may be an ancestral precursor of the whole family of GLP-1/GIP-related peptides and their receptors. Indeed, administration of a 37 amino acid peptide derived from PDF-1 resulted in a reduction in body weight and improved insulin sensitivity in mice. In worms, signaling through this pathway induced food-leaving behavior on concentrated food and roaming behavior on dispersed food, a state we have termed “food-apathy,” paralleling pharmacologic effects of GLP-1/GIP-related peptides in humans. These findings highlight the potential evolutionary origin of this family of hormones and their receptors, and its link to metabolic and neuronal responses in control of feeding behavior.

Most complex organisms eat intermittently, but require energy continuously for cellular metabolism, especially for the brain, nervous, and cardiovascular systems. To achieve this balance, higher organisms have developed systems for storage of energy in the form of fat and carbohydrate, and regulation of food intake to match energy expenditure. However, in humans and other animals, this process may go awry, leading to obesity or anorexia. In mammals, there are multiple mechanisms regulating appetite and food intake, including effects of gastric fullness, the presence of nutrients in food, hedonic responses to the taste and smell of food, and especially hormonal and neuronal circuitry that responds to states of feeding and fasting (1–4). The peptide hormone leptin, which is produced by fat cells more or less in proportion to adipose mass, is a key component of the feedback regulation between energy reserves and control of food intake (5). In normal individuals, when fat stores are adequate, adipocytes secrete leptin into the circulation which can bind to receptors on AgRP/NPY and POMC neurons in the hypothalamus reducing production of hunger-stimulating hormones, like neuropeptide Y (NPY), and increasing the production of appetite-suppressing hormones like alpha-melanocyte-stimulating hormone (α-MSH), leading to a decrease in appetite and food intake (2). In the hypothalamus, leptin also suppresses the activity of AMP-activated kinase (AMPK), an enzyme which normally senses low ATP levels, a cellular signal of energy insufficiency that acts to increase feeding (6–8).

How metabolic parameters influence food intake or create signals that might stimulate animals to maintain appropriate levels of nutrients or metabolites that mediate essential cellular functions is less clear, but many nutrients play roles beyond simply serving an energy source. For example, incorporation of saturated fatty acids (SFAs) into triacylglycerides (TAGs) depends on coincorporation of the monounsaturated fatty acid (MUFA) oleic acid (OA) (9, 10). MUFAs also compete for incorporation of SFAs into membrane phospholipids, thereby altering membrane properties, including endoplasmic reticulum (ER) membrane function. The importance of such metabolites raises the question of whether regulation of feeding can respond to a need for, or a surplus of, specific nutrients. While there is some evidence that animals have evolved mechanisms to seek out foods containing essential nutrients, these mechanisms are not well understood.

The gastrointestinal tract secretes multiple hormones in response to food intake, including insulin, which promotes energy storage; hormones that potentiate insulin secretion, i.e., incretins, including glucagon-like peptide 1 (GLP-1) and gastric inhibitory polypeptide (GIP); cholecystokinin (CCK), which regulates biliary secretion; peptide YY (PYY), which signals to the brain to reduce food intake by increasing feelings of fullness (satiety), as well as promoting gastric motility and pancreatic exocrine secretion; and amylin, a hormone cosecreted with insulin, which affects gastric motility (11–13). With the exception of insulin, which acts on a receptor tyrosine kinase, these hormones act on a family of closely related G-protein-coupled receptors (GPCRs) to exert their physiological effects (11, 13). Recently, it has become clear that when given at pharmacological doses, GLP-1R agonists (GLP-1Ra) alone or in combination with agonistic (or antagonistic) mimics of GIP and these other related hormones are remarkably effective at reducing food craving and food consumption, thus inducing weight loss (12–16). Increasing evidence indicates that GLP-1R agonist/polyagonists peptides influence appetite by acting in the brain (15), but how these signals are integrated with the leptin-AGRP-NPY system or nutritional factors remains unknown.

Caenorhabditis elegans provide a unique model to search for the metabolic mechanisms that regulate feeding behavior. C. elegans lacks leptin and leptin receptors, but are an animal model that is uniquely amenable to high-throughput genetic analysis for regulation of food consumption (17–19). Starting with an unbiased screen of transcription factors (TFs) involved in metabolic regulation in C. elegans, we identified a group of TFs regulating lipid metabolism that can modulate food intake by affecting food leaving vs. food dwelling behaviors. Disrupting these genes also shifts the relative levels of mono-unsaturated versus saturated fatty acids (MUFAs and SFAs), and supplementation of the C. elegans media with the MUFA oleic acid can suppress the food-dwelling behavior induced by sbp-1 or crh-1 knockdown, acting through a pathway sensed in the endoplasmic reticulum. This in turn acts through downstream pathways involving AMPK, the neurotransmitter 5-HT, and ultimately the worm peptide hormone PDF-1 and its receptor PDFR-1. Interestingly, PDF-1 and PDFR-1 exhibit sequence homology and structural similarity with the family of human incretins, such as GLP-1 and GIP, and their receptors. Consistent with this, administration of a PDF-1 related peptide to diet-induced obese mice resulted in a decrease in body weight and improved insulin sensitivity. Together, these data not only provide insights into mechanisms of metabolic control of feeding, they also demonstrate the potential ancestral origins of GLP-1/GIP polyagonists signaling system that underlies current pharmacological treatment of obesity and demonstrate how nutritional factors may interact with this system.

Results

Lipid Metabolism and Lipid Intake Regulate Feeding Behavior in C. elegans.

To identify metabolic mechanisms that affect food consumption and feeding-associated behaviors in C. elegans, we used RNA interference (RNAi) to screen a library of transcription factors for effects on food intake as assessed by a bacterial clearance assay (SI Appendix, Fig. S1 A and B and Dataset 1, which describes the various assays used to assess food intake and feeding behaviors in C. elegans). This identified 13 transcription factors for which knockdown reduced food consumption and 6 for which knockdown increased food consumption (SI Appendix, Fig. S1C). Among the genes whose knockdown reduced food consumption, six had been previously suggested to alter some feeding behaviors (20). We were particularly interested in sbp-1 (sterol regulatory element-binding protein, SREBP) and crh-1 (c-AMP response element binding protein, CREB) (Fig. 1A), since both regulate genes involved in fatty acid synthesis and fat storage (21, 22). Including the fat-7 and acs-4 (SI Appendix, Fig. S1D). Importantly, neither crh-1 nor sbp-1 knockdown altered pharyngeal pumping frequency (SI Appendix, Fig. S1E). However, when worms treated with RNAi against crh-1 and sbp-1 were placed in a dish with a circumscribed food lawn, there was an increased fraction of worms which left the food to explore regions of the plate away from the food lawn (Fig. 1 B and C), i.e. whole-body knockdown of crh-1 or sbp-1 produced a “food-leaving” behavior, an action associated with decreased food intake (20, 23). While adverse conditions, like exposure to pathogens or toxins, can also induce food-leaving, the effects of sbp-1 and crh-1 RNAi on food-leaving occurred by independent mechanisms, as shown using worm’s mutant in the immune response and detoxification pathways (SI Appendix, Fig. S1 F–J). Using tissue-specific knockdown of crh-1 or sbp-1 revealed that loss of these transcription factors in hypodermis or intestine, but not in neurons or muscle, was sufficient to increase food-leaving behaviors (Fig. 1D and SI Appendix, Fig. S1K). To determine whether these behaviors correlate with actual food consumption, we directly assessed food consumption in a liquid culture bacteria clearance feeding assay (SI Appendix, Methods and Dataset 2). Indeed, knockdown of sbp-1 or crh-1 in the gut and/or hypodermis reduced food intake in liquid culture (Fig. 1E and SI Appendix, Fig. S1L), consistent with their effects to increase food-leaving on plates with concentrated food (Fig. 1D and SI Appendix, Fig. S1K). Conversely, food consumption was increased by neuronal-specific sbp-1 or crh-1 knockdown (Fig. 1E), with no increase in food-leaving behavior (Fig. 1D).

Fig. 1.

Metabolic cues regulate feeding behaviors. (A) Modulation of food consumption by crh-1 or sbp-1 RNAi whole body knockdown over 5 d from L4 stage to day 5 adult in 24-well plates. (B) Schematic of the food-leaving assay. (C) Food-leaving behavior induced by crh-1 or sbp-1 RNAi knockdown in wild type animals. The fraction worms off food was recorded over an 8 h period from the early L4 stage to the end of time point. (D) Modulation of food-leaving rate by crh-1 or sbp-1 RNAi in the intestine or neuron or muscle tissue. The RNAi-defective mutation sid-1(qt9) was rescued by tissue-specific transgene expression. (E) Modulation of food consumption by crh-1 or sbp-1 RNAi in specific tissues over 5 d from L4 stage to day 5 adult. Approximately 200 animals were assayed per condition. (F) Schematic of the assay used to evaluate food roaming of individual animals on fully covered food lawns from the L4 stage for 16 to 18 h. (G) Modulation of the roaming score by crh-1 or sbp-1 RNAi in the intestine or neuronal tissue. Data were pooled from three independent experiments, and bimodal population are visualized with distinct symbols. In box plots (G), the center line shows the median, with whiskers extending to minimum and maximum values. Dots represent individual animal, n = 15 to 20 per condition, with data pooled across three independent experiments. In (A, C, D, and E), dots represent individual experimental replicates. P values were derived from one-way ANOVA in (A and C), two-way ANOVA with Tukey’s multiple comparison (D, E, and G). Data are mean ± SEM. from at least two or three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

To better understand the modulation of crh-1 or sbp-1 knockdown effects on feeding related behaviors in the nervous system, we employed a “dwelling assay”. In this assay single L4 stage animals are transferred to plates covered with a bacterial food lawn (Fig. 1F) where they can alternate between “dwelling,” a behavior in which worms move slowly and remain in localized areas, which is generally associated with increased food consumption, and “roaming,” a state in which worms move quickly around the plate and do not dwell on food. A roaming score is calculated by overlaying the plate with a grid and quantifying the percentage of total squares an animal enters as it moves across the plate, with a lower score indicating more dwelling and a higher score more roaming (Fig. 1F). A high roaming score could represent searching for better food but also can reflect a loss of interest toward food (24). Neuron-restricted knockdown of crh-1 or sbp-1 (25) dramatically increased dwelling behavior (Fig. 1G), whereas intestinal knockdown of crh-1 or sbp-1 had no significant effect. Modulation of food leaving and dwelling behaviors in response to metabolic signals from different tissues is not contradictory, but rather likely reflects different roles of these tissues in homeostatic mechanisms. In both cases, the behavior correlates with long-term food consumption in the liquid, suggesting that affinity for food is a common element underlying these responses. Thus, altering lipid metabolism in neurons, gut, or hypodermis can differentially affect food intake and shift the balance between food-leaving, food-dwelling, and roaming behaviors.

Specific Fatty Acids Regulate Food Affinity Behaviors.

sbp-1 and crh-1 encode transcription factors that regulate lipid metabolism (21, 22). Whole body knockdown of sbp-1 decreases fat storage in TAGs (21), a phenotype also observed with crh-1 RNAi (SI Appendix, Fig. S2 A and D). Importantly, both were partially reversed by supplementation of the C. elegans media with oleic acid (OA), a fatty acid whose synthesis depends on FA desaturases (FAT-6/7), which are under control of SBP-1 (Fig. 2A and SI Appendix, Fig. S2A). Supplementation of the media with OA also decreased the fraction of worms leaving the food lawn and suppressed the food-dwelling behavior induced by neuron-specific sbp-1 or crh-1 knockdown (Fig. 2 B and C) but did not suppress toxin-induced food avoidance (SI Appendix, Fig. S2B), suggesting that fat storage and these behavior effects might be linked causally. Supporting this concept, knockdown of either crh-1 or sbp-1 did not reduce fat stores when animals were cultured on plates covered with a full lawn of bacteria, where animals are prevented from leaving their food (SI Appendix, Fig. S2 C and D). This suggests that the decreased fat storage caused by crh-1 or sbp-1 knockdown is secondary to increased food-leaving and the associated reduction in food consumption (Fig. 1 C and E). While knockdown of crh-1 does not affect development of C. elegans into gravid hermaphrodites, knockdown of sbp-1 in the wild type slows development of the worms into adults by ~2 h and a more severe reduction in developmental timing was seen in sbp-1 mutants sbp-1(ep79). Both of these phenotypes could be largely rescued by supplementation with OA or growth of the worms on a full bacterial lawn (SI Appendix, Fig. S2E). Taken together, these data suggest that feeding and feeding-related behaviors observed in worms with sbp-1 or crh-1 knockdown or OA supplementation are due to altered lipid metabolism and not a general impairment of health.

Fig. 2.

Specific fatty acids regulate food affinity behaviors. (A) FA biosynthesis pathway in C. elegans, the enzymes involved, and the effects of the FA on food affinity behaviors. Food affinity was stimulated by supplementation with FAs or knockdown of genes shown in Red. FAs that suppressed food affinity behaviors are in green. Beige indicates mutations or RNAi knockdown or FA supplementation that did not affect food affinity behaviors (SI Appendix, Fig. S2 and Dataset 1). Equivalent mammalian enzymes are indicated in black. (B) Modulation of food-leaving rate by crh-1 or sbp-1 RNAi with or without OA supplementation. (C) Modulation of food roaming score by crh-1 or sbp-1 RNAi in neurons with or without OA supplementation. (D) Modulation of food-leaving rate of the fat-6(tm331); fat-7(wa36) double mutant with or without OA supplementation. (E and F) Modulation of food-leaving rate (E) or food roaming score (F) following supplementation with the indicated dietary fatty acids. In (D), worms were placed on assay plates started at L4 stage and food-leaving rate was read after 8 h. These SCD desaturases have redundant effects in OA biosynthesis. Data were analyzed using two-way repeated-measures ANOVA with Tukey’s multiple comparisons test (B and C), one-way ANOVA in panels (E and F). Data are mean ± SEM. from at 2 to 3 independent experiments, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

To further explore the relationship between dietary fatty acids and sbp-1 or crh-1 knockdown on feeding behaviors, we systematically knocked-down the key enzymes of the de novo lipogenesis pathway or supplemented worms with specific nutrients and assessed feeding-associated behaviors. We found that worms with a double mutation of fat-6/7, which cannot synthesize OA (Fig. 2A), exhibited decreased food-leaving behaviors, and this was reversed by dietary OA (Fig. 2D). Food-leaving was also increased when biosynthesis of OA precursors or the MUFA palmitoleic acid (POA) was impaired (SI Appendix, Fig. S2 F–H). Unlike most other animals, C. elegans also encodes FA desaturases and elongases that convert OA to polyunsaturated FAs (PUFAs) (Fig. 2A) (26), however, genetic or RNAi knockdown of these enzymes failed to induce food-leaving (SI Appendix, Fig. S2 I and J) or to prevent OA from correcting food-leaving stimulated by sbp-1 or crh-1 RNAi (SI Appendix, Fig. S2 K and L). Thus, food affinity behaviors were regulated by deficits in the synthesis of MUFAs, but not PUFAs. Interestingly, feeding of either POA or OA corrected food-leaving induced by deficits in other MUFAs (SI Appendix, Fig. S2M). Food-leaving was also suppressed by the MUFA cis-vaccenic acid and the omega-6 PUFA linoleic acid (LA), but not the omega-3 PUFAs α-LA or the omega-6 derivative of LA, dihomo-gamma-linolenic acid (DGLA) (SI Appendix, Fig. S2 N–Q). Thus, MUFAs, like oleic acid, are critical for maintaining a proper balance of food affinity behaviors in C. elegans, whereas supplementation with palmitic acid (PA), the saturated fatty acid precursor of oleic acid, shifted animal behaviors, inducing food-leaving when the food source was concentrated but dwelling with the food source was scattered (Fig. 2 E and F). Importantly, OA cosupplementation with PA, suppressed PA-induced feeding behaviors robustly and in a dose-dependent manner (SI Appendix, Fig. S2S). Supplementation with stearic acid (SA), the saturated fatty acid precursor of OA, on the other hand, induced food-leaving (Fig. 2A and SI Appendix, Fig. S2R). Thus, food-leaving and dwelling can be triggered by deficits in MUFAs in the periphery (gut, hypodermis) or by increasing levels of SFAs relative to MUFAs in the nervous system. Indeed, just 2 h of OA supplementation corrected the food-leaving that was induced by either crh-1 knockdown or PA addition (SI Appendix, Fig. S2 T–X).

Neuronal IRE-1 Acts As Sensor Integrating Metabolic Signals That Regulate Feeding Behaviors.

How do MUFAs act to modify food affinity behaviors? Two recent studies have shown that supplementation with OA or other MUFAs extends C. elegans lifespan, and this effect is mediated by a network of processes involving incorporation of these FAs into membrane phospholipids and triglycerides (TAGs), which are stored in lipid droplets (LDs) (Fig. 3A) (27, 28). These processes are colocalized at the endoplasmic reticulum (ER) membrane and create a TAG/phospholipid homeostasis network through which FAs can be rapidly incorporated into membrane lipids or TAGs (<2 h) (9, 10, 27), consistent with the rapidity of their effects on food-leaving (SI Appendix, Fig. S2 T–X).

Fig. 3.

FA regulation of feeding behaviors at the ER. (A) Schematic showing components of the TAG/phospholipid homeostasis network, an essential mediator of OA-induced lifespan extension. Genes shown in red have been analyzed in this study for their effects on feeding behaviors. (B and C) Modulation of food-leaving rate (B) and roaming score (C) by crh-1 or lpin-1 RNAi in either whole body or neuron-specified animals with or without OA supplementation. (D–H) Lipidomics analysis of saturated fatty acid C16:1 and the majority of mono-unsaturated fatty acid, C18:1 in different phospholipids, including PE (D), LPC (E), PI (F), PS (G), and PC (H) following sbp-1 RNAi on the full lawn plates with or without OA supplementation. (I and J) Modulation of roaming score (I) and food-leaving rate (J) with exposure to PA with or without neuronal-specific ire-1 knockdown. (K) Effect of CRISPR mutant IRE-1(S679) to block food-leaving induced by crh-1 or sbp-1 RNAi. P values are derived from two-way ANOVA with Tukey’s multiple comparisons test. Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

We hypothesized that if OA and other MUFAs alter food affinity behaviors through their incorporation into complex lipids, such that inhibiting this TAG/phospholipid homeostasis network, might induce these behaviors without allowing suppression by OA. Indeed, knockdown of enzymes involved in triglyceride and phospholipid biosynthesis and turnover enzymes, including lpin-1 (LPIN), atgl-1 (ATGL-1), and dgat-2 (DGAT-2), induced food leaving behavior that was not rescued by OA (Fig. 3 B and C and SI Appendix, Fig. S3 A–C) (27, 28). Furthermore, RNAi knockdown of the phospholipid biosynthesis enzymes PISY-1 and CEPT-1, the lipid droplet associated protein PLIN-1, acyl-CoA synthetase 4 (ACS-4) that catalyzes the conversion of long-chain fatty acids to their active acyl-CoA forms, and fat storage inducing transmembrane protein-2 (FITM-2) (29, 30) all induced food-leaving behavior that was only partially suppressed by OA (SI Appendix, Fig. S3D). By contrast, interference with the ER-associated degradation (ERAD) mechanism by knockdown of the E3 ubiquitin ligase HRD-1 induced food-leaving that was fully suppressed by OA (SI Appendix, Fig. S3E). ERAD is known to be critical for maintaining ER luminal protein homeostasis and is an essential mediator of OA-induced lifespan extension (28). Thus, while food-leaving can be induced by a disruption of ER homeostasis, this does not interfere with oleic acid’s effect on food-leaving. Rather the suppression of food leaving and dwelling behaviors induced by MUFAs appears to depend on the TAG/phospholipid homeostasis network.

To explore if SFAs alter food affinity behaviors through their incorporation into phospholipids, we performed LC–MS lipidomic analysis on animals that were cultured on full-lawn plates, thus preventing food-leaving behavior. We found that while sbp-1 knockdown did not significantly alter total lipid, TAG, or diacylglycerol (DAG) levels (SI Appendix, Fig. S3 F–I), it did increase the ratio of C16:0/C18:1 (PA/OA) incorporation in membrane phospholipids, including phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylserine (PS), phosphatidylinositol (PI), and Lys phosphatidylcholine (LPC), and this was reversed with OA supplementation (Fig. 3 D–H). These findings support our model that feeding behaviors are regulated by the ratio of SFA/MUFA and that FA incorporation into membrane PLs through this TAG/phospholipid homeostasis network transduces the fatty acid signal at the ER. These findings also agree with studies in which membrane lipid saturation was increased by knockdown of the SBP-1 cofactor MDT-15 and decreased by OA levels (31).

Our observation that dwelling was induced when FA metabolism was perturbed in neurons (Fig. 1G) suggests that the FA signaling pathway we uncovered operates primarily within the brain to regulate behavior. Indeed, food dwelling that was induced by neuron-specific sbp-1 or crh-1 knockdown was suppressed by OA (Fig. 2C), whereas dwelling induced by neuronal knockdown of genes involved in TAG/phospholipid homeostasis, such as atgl-1 and dgat-2, was not rescued by OA (SI Appendix, Fig. S3 A and B). On the other hand, knockdown of the ER membrane sensor IRE-1 in the nervous system, but not in the intestine, prevented PA from inducing dwelling and food-leaving (Fig. 3 I-J and SI Appendix, Fig. S3J), indicating that neuronal ER/IRE-1 signaling is required for the feeding/dwelling response. This requirement appears to be mediated by the IRE-1-dependent mRNA decay pathway (RIDD). Thus, a CRISPR generated IRE-1(S679A) mutant which lacks RIDD function markedly blunted the food-leaving and dwelling behaviors, while XBP-1/PERK mutants did not (Fig. 3K and SI Appendix, Fig. S3 K and L). Together, these results indicate that the relative levels of saturated and unsaturated FAs are sensed through their incorporation into membrane lipids, and this information is transduced by ER/IRE-1 within the nervous system to regulate food affinity behaviors.

FA-Mediated Regulation of Food Affinity Behaviors by AMPK.

Our results raised an important question as to whether other regulators of feeding behaviors act through this FA metabolism-based pathways. In mammals, activation of AMPK in the brain is known to increase food consumption (6–8), whereas in peripheral tissues, activation of AMPK increases glucose uptake and fatty acid oxidation in muscle, while inhibiting lipid synthesis and promoting fatty acid oxidation in liver (6–8). In C. elegans, transgenic expression of a constitutively active AMPKα2 catalytic subunit (CA-AAK-2) is known to increase lifespan (22). Interestingly, we find that whole body transgenic expression of CA-AAK-2 increased food-leaving, resulting in reduced food consumption (Fig. 4 A and B). This occurred without affecting pharyngeal pumping (SI Appendix, Fig. S4A), thus mimicking the effects of whole-body sbp-1 or crh-1 knockdown on feeding. Also, like the latter, these effects of overexpression of CA-AAK-2 were reversed by OA supplementation or knockdown of ire-1 (Fig. 4 C and D), indicating dependence on FA and ER signaling. Importantly, while CA-AAK-2 overexpression in the intestine or muscle did not affect feeding behaviors or food consumption (SI Appendix, Fig. S4C), CA-AAK-2 overexpression within the brain (N-CA-AAK-2) in worms markedly increased dwelling and food consumption (Fig. 4 E and F and SI Appendix, Fig. S4C), and this was suppressed by ire-1 knockdown (Fig. 4G). Thus, the action of AMPK in the brain to increase feeding appears to be evolutionarily conserved. In addition, both the role of AMPK and the effects of knockdown of crh-1 or sbp-1 on feeding behaviors and food consumption depends upon their effects in neurons.

Fig. 4.

FA-mediated regulation of feeding behaviors by AMPK. (A and B) Modulation of food-leaving rate (A) and food consumption (B) by whole-body expression of a constitutively active catalytic subunit of AMPK, CA-AAK-2. (C) Modulation of food-leaving rate by CA-AAK-2 with or without OA supplementation. (D) Modulation of food-leaving rate by CA-AAK-2 with or without ire-1 knockdown. (E) Modulation of food roaming score by neuro-CA-AAK-2. (F) Modulation of food consumption by constitutively active AMPK in the indicated tissues. (G) Modulation of roaming score by neuro-CA-AAK-2 with or without ire-1 knockdown. (H) Modulation of food leaving rate by CA-AAK-2 with knockdown of peroxisomal beta-oxidation genes. P values are derived from two-tailed Mann–Whitney test (A, B, and E), and two-way ANOVA with Tukey’s multiple comparisons test in (C, D, G, and H). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Dots represent individual animals in (E and G), n = 15 to 20 per condition, with data pooled across two to three independent experiments.

The finding that the catalytic subunits of AMPK, aak-1 and aak-2, were not required for food-leaving behavior induced by sbp-1 or crh-1 knockdown (SI Appendix, Fig. S4D) raised the question of how does AMPK interact with these FA-sensing mechanisms? AMPK signaling is known to inhibit lipid synthesis and increase lipid oxidation (32). Thus, in contrast to the effects of sbp-1 or crh-1 RNAi, whole-body knockdown of CA-AAK-2 reduced fat storage independent of its effects on food-leaving, indicating that the reduction in fat was secondary to altered lipid metabolism (SI Appendix, Fig. S4 E–H). Disruption of peroxisomal β-oxidation by treatment of worms with dhs-28 or dhs-22 RNAi (28) partially suppressed CA-AAK-2-induced leaving (Fig. 4H), suggesting that lipid oxidation contributes to AMPK induction of these behaviors. Thus, AMPK in the brain/neurons regulates food affinity behaviors and food consumption differentially from its effects in other tissues, and this occurs, at least in part, by modulating FA and lipid levels.

Serotonin Signaling Mediates Metabolic Regulation of Food Affinity Behaviors.

In both worms and mammals, feeding behaviors are known to be regulated by the neurotransmitter 5-hydroxytryptamine (5-HT, serotonin) (33–36). In C. elegans, serotonin is primarily produced in three neuron pairs, referred to as ADF, NSM, and HSN (37), with most serotonin production being mediated by the enzyme tryptophan hydroxylase (tph-1) (37, 38). It is known that the level of dwelling seen on a continuous food lawn depends, at least in part, upon serotonin synthesis in the NSM neurons (36), which sense food within the pharyngeal lumen (35). Exposure to bacterial pathogens, on the other hand, induces expression of tph-1, as assessed using a tph-1p::GFP reporter, in the ADF neurons and does not induce food dwelling behavior (SI Appendix, Fig. S5 A-B). These observations add additional evidence that metabolically induced food-leaving is distinct from pathogen-mediated feeding responses.

Using a Cre-Lox deletion transgene (36), we found that expression of tph-1 in the NSM, but not ADF, neurons was required for sbp-1 and crh-1 RNAi induction of food-leaving (Fig. 5A and SI Appendix, Fig. S5C), consistent with most serotonin synthesis in the nervous system being tph-1-dependent. Expression of tph-1 was also required for the dwelling response induced by neuronal AMPK activation (Fig. 5B), as predicted by our model that AMPK regulates food affinity behaviors by acting through the same FA-sensing and ER-based pathway as sbp-1 or crh-1 RNAi. Also consistent with this hypothesis, the level of tph-1p::GFP in the NSM neurons was increased by treatment with crh-1, sbp-1, or dgat-2 RNAi, suggesting that metabolic stimuli that stimulate food-leaving are associated with upregulated serotonin biosynthesis (Fig. 5 C and D and SI Appendix, Fig. S5 D–G). Supplementation with OA suppressed the upregulation produced by knockdown of crh-1 or sbp-1, but did not rescue the effect of dgat-2 knockdown (Fig. 5 C and D), consistent with DGAT-2 being critical in the TAG/phospholipid homeostasis network (Fig. 3A). Feeding of PA rapidly induced tph-1p::GFP expression, and this was suppressed by OA supplementation or ire-1 knockdown (Fig. 5 E–H). Finally, neuron-specific expression of CA-AAK-2, which induces dwelling, upregulated tph-1p::GFP expression in the NSM neurons, in agreement with the role of tph-1 in these neurons in the effects of AMPK on feeding (Fig. 5I). Together these data indicate that imbalanced lipid metabolism activates signals in both the periphery and nervous system to regulate food affinity behaviors, and this occurs through shared NSMs neurons and modulation of 5-HT levels.

Fig. 5.

Regulation of food-affinity behaviors through serotonin signaling. (A) Modulation of food-leaving rate by crh-1 RNAi in wild type or neuron-specific depletion of tph-1 mutants. (B) Modulation of roaming score by the neuronal CA-AAK-2 in wild type or tph-1 mutant. (C). Representative images of tph-1p::GFP expression in NSM neurons of day-1 adult hermaphrodites treated with crh-1, sbp-1, or dgat-2 RNAi with or without OA supplementation from the L1 stage. (D) Quantification of tph-1p::GFP levels in the NSM neurons in (C). (E) Representative images of tph-1p::GFP expression over 16 h PA exposure in animals subjected to ire-1 RNAi. (F) Quantification of the tph-1p::GFP levels in (E). (G) Representative images of tph-1::GFP expression in adult hermaphrodites pretreated with or without PA from L1 to L4 stage followed by exposure to PA or OA for 16 h. (H) Quantification of the tph-1p::GFP levels in (G). (I) Modulation of tph-1p::GFP activation in the NSM neuron by the neuronal activation of AMPK. Dots in (B) individual animals, n = 15 to 20 per condition. Dots in (D, F, H, I) are the average signal intensity of each single neuron per animal, Dots in (A) represents individual experiments replicates. P values are derived from two-way ANOVA with Tukey’s multiple comparisons test in (A, D, and F), two-tailed Mann–Whitney test in (B, H, and I). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

GLP-1-Like Signaling in Metabolic Regulation of Food Affinity Behaviors.

Serotonin (5-HT) signaling has been reported to regulate roaming/dwelling behaviors in C. elegans antagonistically to a neuropeptide termed PDF-1 (pigment dispersing factor-1) (36). We found that treatment of worms with crh-1 or sbp-1 RNAi resulted in increases in both tph-1p::GFP (Fig. 5 C-D) and PDF-1p::GFP expression (SI Appendix, Fig. S6 A-B). PDF-1 acts through its G-protein receptor (GPCR) PDFR-1 and has been previously shown to have a role in searching behavior in males (39). Although there is no exact PDF-1 orthologue in humans, in a multiple-sequence alignment, we found that different regions of the PDF-1 peptide show homology to human GLP-1, GLP-2, glucagon, GIP, calcitonin, and amylin, as well as exendin-4 (EX-4), a peptide found in the saliva of the Gila monster that shares ~50% identity with GLP-1 and has been widely used as a GLP-1R agonist in humans (12) (Fig. 6A, and SI Appendix, Fig. S6C). Likewise, the PDF-1 receptor (PDFR-1) has been suggested to be related to the vertebrate calcitonin receptor (40), but also has structural homology to receptors for the glucagon, GLP-1, GLP-2, GIP, amylin, and vasoactive intestinal peptide (VIP) (41) (Fig. 6B and SI Appendix, Fig. S6 D-E). Agonists for the GLP-1R, as well as polyagonists that can activate GLP-1R and GIP receptor, are now major drugs for the treatment of obesity and associated metabolic syndrome (14–16), but no members of this peptide and receptor family have been identified in C. elegans. We hypothesized that PDF-1 and PDFR-1 represent the C. elegans functional counterpart to GLP-1/GLP-1R and related family of peptides and their receptors.

Fig. 6.

Regulation of food-affinity behaviors through GLP-1 receptor-like signaling. (A) Sequence alignment of PDF-1a (G8JYC6) peptide with human GLP1 and related peptides involved in regulating food intake, metabolism, and gastrointestinal function performed using the Clustal Omega program. (B) Structural alignment of PDFR-1 by AlphaFold 2 with the active human GLP-1R (PDB:7DUR). The RMSD between the two structures is 1.000 Å angstroms. (C and D) Modulation of food-leaving rate by sbp-1 RNAi-induced (C) or dietary PA supplementation-induced changes in the SFA/MUFA ratio (D) in wild type and pdfr-1 mutant C. elegans. (E) Modulation of roaming score by crh-1 RNAi in wild type worms, pdfr-1 mutant, or tph-1 mutant. (F) Representative images of pdf-1p::GFP expression in brain neurons following sbp-1 RNAi in wild type worms and tph-1 mutant. Arrows indicate neuronal cell bodies. (Scale bar, 50 µm.) (G) Quantitation of pdf-1p::GFP expression in (F). The intensity of signals was normalized using the average intensity of pdf-1p::GFP of empty vector control RNAi treated worms measured in parallel. Dots represents the average GFP level of individual worm, n = 45 to 60 from three independent experiments. (H) Left: Representative images of pdf-1p::GFP expression in wild type, neuron constitutive AMPK (N-CA-AAK-2) or tph-1 mutant worms. Td-tomato indicates the expression of N-CA-AAK-2. Arrows indicate neuronal cell bodies. (Scale bar, 50 µm.) Right: quantitation of pdf-1p::GFP signal from panels on the Left. The intensity of signals is normalized using the average intensity of pdf-1p::GFP in wild type worms cultured on OP50 measured in parallel. (I–K). Effect of PDF-1 overexpression on roaming score (I), food-leaving rate (J), and food consumption (K) compared to wild type. (L and M) Effects of twice daily injections of saline, PDF1-Pep01a, or Tirzepatide] on cumulative daily food intake with lower dose Pep01a (100 nmol) on the Left and higher dose on the Right (200 nmol) and the body weight change (M). (N) Intraperitoneal glucose tolerance tests. (O and P). Fasting plasma insulin levels and HOMA-IR in mice on high dose Pep01a. (Q) A schematic model illustrating the central control of food intake in C. elegans in which PDF-1/PDFR-1 signaling promote food-leaving and roaming in response to peripheral/neuronal sensing of metabolic status. Dots represent individual assays (C, D, J, and K), the average signals from panneurons of individual worm (G and H), or individual animals in panels (L, and N–P). P values are derived from two-way ANOVA with Tukey’s multiple comparisons test in (C, D, E, and G). (L) Statistical analysis was assessed using a two-tailed Student t test. (M and N) were analyzed by repeated measures two-way ANOVA with Bonferroni’s post hoc test for comparison of individual time points. (O and P) were analyzed by one-way ANOVA. Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Indeed, we found that while loss of PDFR-1 function had no effect on food-leaving or dwelling induced by nonspecific toxins or pathogen exposure (SI Appendix, Fig. S6 F–H), pdfr-1 loss-of-function mutants block the food-leaving effects of whole-body knockdown of sbp-1 or PA supplementation (Fig. 6 C and D), and this was consistent with an increase in PDF-1p::GFP expression level in response to crh-1 or sbp-1 knockdown (Fig. 6 F and G). This increase in PDF-1p::GFP expression was lost, however, in the tph-1(lf) mutant (Fig. 6 F and G). However, the pdfr-1 loss-of-function (lf) mutant did not block food dwelling caused by crh-1 knockdown, but this was lost in the tph-1(lf) mutant (Fig. 6E). Thus, PDF-1/PDFR-1 signaling is functionally opposite to 5-HT signaling in mediating C. elegans dwelling behavior, consistent with previous studies (36).

In contrast to the increased PDF-1p::GFP expression observed in worms with whole body crh-1 or sbp-1 knockdown, neuronal expression of constitutively active AMPK (N-CA-AAK-2) strongly suppressed PDF-1p::GFP expression levels, and this was dependent on TPH-1 (Fig. 6H). These findings suggest that PDF-1 functions as a downstream integrator of peripheral and neuronal metabolic cues to promote food-leaving or roaming behavior, thereby reducing food intake in C. elegans. In support of this model, we found that transgenic overexpression (OE) of PDF-1 in worms resulted in an increase in food-leaving rate similar to that observed in response to whole-body crh-1 or sbp-1 knockdown (Fig. 6J). PDF-1 overexpression also counteracted the effects of increased 5-HT and increased roaming behavior when food was dispersed (Fig. 6I). Thus, worms with PDF-1 overexpression exhibited both enhanced food leaving and increased roaming, ultimately leading to reduced food consumption (Fig. 6K). We have termed the unique combination of food leaving in the assay on a circumscribed food lawn and increased roaming on the plate covered with bacteria as “food apathy,” since the worms neither dwell on food when it is dispersed in the environment and actually actively leave food when it is concentrated in one spot. Consistent with this model, worms with CRIPSR-generated constitutively active PDFR-1 [designated PDFR-1(gf)] phenocopied the effects of PDF-1 overexpression, exhibiting food apathy with both increased food-leaving and increased roaming behaviors (SI Appendix, Fig. S6 I–K). PDFR-1 (gf) worms also consumed less food compared to wild type and pdfr-1(lf) strains and exhibited lower fat-storage (SI Appendix, Fig. S6 L and M). These actions of PDF-1/PDFR-1 signaling mimic the effect of GLP-1R agonists and GLP-1/GIP receptor polyagonists in humans.

To confirm that PDFR-1 is acting as a GLP-1R in worms we tested the effect of 6,7-dichloro-2-methylsulfonyl-3-N-tert- butylaminoquinoxaline (referred to as “Compound 2”), an ago-allosteric modulator of the GLP-1R that can both act as an agonist on its own and also enhance the effects of natural ligand GLP-1 (42) in C. elegans. Indeed, Compound 2 has been shown to covalently bind to C347 in transmembrane domain TM6 of GLP-1R inducing receptor activation (43). Using AlphaFold3 based DiffDock-L binding assay and SwissDock analysis, we found that Compound 2 can fit into a similar pocket in PDFR-1 (SI Appendix, Fig. S6N). More importantly, treatment of C. elegans with Compound 2 triggered a rapid increase in food-apathy behaviors (SI Appendix, Fig. S6 O–Q), mimicking its effect in mammals (44). Importantly, this effect was lost in worms with loss of function mutations in either PDF-1 or PDFR-1. Taken together, these data suggest that in the nervous system, in response to dietary changes, such as MUFA insufficiency, 5-HT inhibits PDF-1, and this results in increased food dwelling and increased food consumption. Conversely, when changes in nutrient intake or lipid metabolism in the peripheral are perceived by the ER/IRE-1 sensor in the nervous system, 5-HT signaling interacts synergistically with PDF-1 to promote food-leaving and reduced food intake (Fig. 6Q). This PDF-1/PDFR-1 mediated signal results in a state of food apathy characterized by both increased roaming and increased food-leaving, leading to a decrease of food intake and decreased stored fat (Fig. 6Q).

As an initial attempt to more directly assess whether the PDF-1/PDFR-1 pathway might mimic human GLP-1/GIP agonists, we synthesized a 37 amino acid peptide corresponding to amino acids 26 to 63 of PDF-1 (SI Appendix, Fig. S7A) and evaluated its effects in vitro on cells expressing human GLP1R or GIPR and in vivo in high fat diet-obese mice (SI Appendix, Fig. S7F). In vitro, although considerably less potent than the mammalian peptides and the synthetic dual GLP1-GIP agonist Tirzepatide, this peptide showed clear stimulation of cAMP accumulation in cells expressing the human GIPR and some stimulation in cells expressing GLP-1R (SI Appendix, Fig. S7 B-E). When injected into high fat diet-induced obese mice in vivo, this peptide at low and high doses also induced a trend to a reduction in food intake (Fig. 6L) and a small, but significant, reduction body weight compared to saline injected mice (Fig. 6M and SI Appendix, Fig. S7G). More impressive, however, it also resulted in a significant improvement in glucose tolerance (Fig. 6N) and a near normalization of hyperinsulinemia and insulin resistance assessed as HOMA-IR (Fig. 6 O and P). Thus, the effects of GLP-1, the GLP-1 receptor, and related members of this ligand–receptor family observed in humans are foreshadowed by PDF-1/PDFR-1 in worms, indicating that food aversion and improved insulin sensitivity by these hormonal signals is an ancient, conserved function.

Discussion

In humans, regulation of food is a complex process involving both hedonic mechanisms, driven by the pleasure associated with eating, and homeostatic signals, primarily mediated by leptin secretion from adipose tissue and its effects at the hypothalamus to inhibit neurons in the arcuate nucleus that express orexigenic neuropeptides like AgRP (agouti-related peptide), which promote hunger, while simultaneously activating anorexigenic neurons expressing POMC (pro-opiomelanocortin) that decrease appetite (2, 45). However, in obesity there is leptin resistance (46), and thus most pharmacological therapy has resorted to use of other agents, most importantly GLP-1 receptor agonists or polyagonists for GLP-1R and several closely related hormones, including GIP, glucagon, and amylin, even though these are not normally major hormones in the control of eating behavior (16, 47). In the present study, starting with unbiased genetic screening in C. elegans for pathways affecting food intake, we identified a pathway in which the relative availability of MUFAs and SFAs, acting through the endoplasmic reticulum stress sensor IRE-1, interfaces with serotonin and AMPK signaling to control the behavior of the animals to dwell on food and eat versus leave food and roam. We have termed the combination of roaming in the presence of a full lawn of food and actively leaving a concentrated source of food “food-apathy”. We show that induction of “food apathy” is ultimately mediated through signals of a G-protein-coupled receptor (PDFR-1) and its cognate ligand (PDF-1). Most importantly, we show that this primitive signaling system is homologous to the GLP-1/GIP family of peptides, including glucagon, calcitonin, and amylin, and their receptors, and that a peptide containing a partial sequence of C. elegans PDF-1 can influence body weight and insulin sensitivity in mice. Thus, these pathways appear to represent the ancestral origin of this food consumption regulating pathway.

While the type of food consumed affects satiety and long-term energy balance in higher organisms, nutritional signals play a more important role in lower organisms like C. elegans. For example, when fed a diet high in monounsaturated fatty acids, worms show normal feeding behaviors, whereas on a diet high in saturated fatty acids they show increased food-leaving, i.e. moving away from a spot of concentrated food. However, with fatty acid supplementation, there is also reduced roaming (i.e., increased dwelling) on full-lawn plates, and this behavior is generally associated with eating. This suggests that in the worm the relative ratio of fatty acid subtypes provides an indicator of food quality, as well as signals to seek better food through food-leaving versus promoting a dwelling behavior to ensure food intake. Using tissue-specific RNAi, we find that this FA sensing pathway operates primarily within the nervous system. Although it is not clear which neurons function as the metabolic sensor, at a mechanistic level these nutrient effects are mediated, at least in part, as responses to membrane stress, with IRE-1 inducing a transcriptional program to promote restoration of ER membrane homeostasis (48). Our results indicate that this molecular sensor has also been co-opted during evolution to transduce information on membrane conditions into behavioral responses.

The FA sensing pathway we uncovered also responds to the low-energy signal molecular AMPK, with neuronal AMPK activation inducing dwelling and food consumption, mirroring its effects in mammals. Indeed, it is known that a hypothalamic AMPK signal stimulates feeding in mammals (6), but whether this is regulated by effects of fatty acids on the endoplasmic reticulum and serotonin is unknown. The fact this response is linked to a GLP-1R-like signaling pathway in worms indicates that this ancestral pathway was designed to provide an immediate indicator of cellular energy availability, while in higher organism the leptin signal from adipose stores provides this information, but over a more chronic timespan. It will be interesting to determine if the fatty acid/ER-based pathway we identified in worms might also be sensitive to the relative abundance of other metabolites or conditions, thus providing a broadly responsive mechanism of metabolic behavioral regulation.

GLP-1 and GIP were initially identified in mammals as incretins, i.e., peptides released from the GI tract in response to glucose ingestion that potentiated glucose-stimulated insulin release from β-cells (49). Therapeutically GLP-1R agonists were developed to mimic this effect and increase insulin release by the pancreas in the treatment of type 2 diabetes. However, it was soon recognized that at therapeutic levels, GLP-1Ra also had effects to reduce food intake, and as a result, they have become major weight loss drugs (15). Poly-agonist peptides that target GLP-1R along with related receptors (15), including the GIP receptor, glucagon receptor, amylin, and even the calcitonin receptor, are in clinical trial or clinical use, and are even more potent in their effects on weight loss. However, how such agents influence food consumption and what relationship this appetite regulation has to normal physiology remains unknown (15, 50). Our findings in C. elegans have revealed a paradigm that may help answer this question, namely that an early physiological function of signaling through this receptor family within the brain is indeed to control of eating behaviors in response to the metabolic sensing pathway. PDFR-1 is a class-B (secretin-family) GPCR and shares structural features with mammalian incretin receptors, including GLP-1R. In C. elegans, PDF-1 expression and signaling are modulated by upstream signals from nutrients, such as fatty acid, as well as serotonin and AMPK.

Previous studies in C. elegans have identified three class-B (secretin-family) GPCRs: PDFR-1 (also called SEB-1), SEB-2, and SEB-3. While few functions has been identified for the latter two, PDF-1 binding to PDFR-1 has been shown to be involved in many functions, including locomotor behavior, reproduction, mating, both arousal and sleep-like behaviors, circadian rhythm, food-related associative learning, food-leaving behavior, and even lifespan (51). Importantly, in terms of metabolism, we find that activation of PDF-1/PDFR-1 system in worms reduces food-dwelling and increases food-leaving, a state we have termed “food apathy”. In the mouse, the two corresponding effects of GLP-1R agonist therapy appear to be mediated through separate areas of the brain and may arise from different inputs (52). This is consistent with our finding that the metabolic signals regulating dwelling and food-leaving in C. elegans arise in the nervous system but also get input from peripheral tissues. Although these behavioral endpoints conceptually parallel aspects of GLP-1R agonist action in mammals, it might not be a functional equivalence across species. In mammals, it is possible that effects of dietary fatty acids, such as MUFAs, might act to synergize with the effects of GLP-1R agonists to maximize satiety. Indeed, diets rich in the MUFA oleic acid have been reported to reduce deposition of abdominal adipose tissue and, in some cases, decrease food intake (53). Thus, more careful studies of this interaction in mammals might open additional therapeutic options to couple specific diets with GLP-1R agonist therapy. In either case, the identification of this PDF/PDFR-1 pathway that may represent the origin of the actions of the family of GLP-1/GLP-1R hormones and help explain the powerful effects of these agents on food intake. Indeed, in vitro, using a 37 amino acid peptide derived from the sequence of PDF1, we find effects to stimulate human GIP and, to a lesser extent, GLP-1 receptors. In addition, in vivo this peptide induced a small, but significant, reduction in body weight, and, more importantly, markedly improved insulin sensitivity in diet-induced obese mice. This is similar to GLP-1 receptor agonists, which have been noted to improve insulin sensitivity, often before significant weight loss occurs (54). It is also worth noting that this peptide was not sequence-optimized or modified in ways used to slow degradation or prolong circulating half-life, such as those used in therapeutic GLP1R/GIP agonists/polyagonists. Also, since 37 amino acids represents only one partial sequence from the 88 amino acid PDF-1 peptide, it is possible that this backbone, with further modifications such as those made in other GLP-1/GIP agonists, may provide a unique scaffold on which to develop new agents for combating obesity, diabetes, and the many other disorders that appear to be modified by GLP-1 agonist therapy, as well as strategies for minimizing their side effects. Future studies in GIPR and GLP-1R knockout mice would also be interesting to better define the mechanism and sites of action of these peptides.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (PDF)

Dataset S02 (PDF)

Dataset S03 (PDF)

Dataset S04 (PDF)

Data, Materials, and Software Availability

OmicsData data have been deposited in UCSD MassIVE (MSV000096286) (55). Study data are included in the article and/or supporting information.