To gain a comprehensive overview of the role of glo-1 in C. elegans metabolism, we employed a fully untargeted comparison of the metabolomes of a glo-1 null mutant and wild-type worms (Figure 1d). HPLC–high-resolution mass spectrometry (HPLC–HRMS) data for the exo-metabolomes (excreted compounds) and endo-metabolomes (compounds extractable from the worm bodies) of the two strains were analyzed using the Metaboseek comparative metabolomics platform, which integrates the xcms package (Tautenhahn et al., 2008). These comparative analyses revealed that the glo-1 mutation has a dramatic impact on C. elegans metabolism. For example, in negative ionization mode, we detected >1000 molecular features that were at least 10-fold less abundant in the glo-1 exo- and endo-metabolomes, as well as >3000 molecular features that are 10-fold upregulated in glo-1 mutants. For further characterization of differential features, we employed tandem mass spectrometry (MS²) based molecular networking, a method which groups metabolites based on shared fragmentation patterns (Figure 1d, Figure 2—figure supplements 1–4; Wang et al., 2016). The resulting four MS² networks – for data obtained in positive and negative ionization mode for the exo- and endo-metabolomes – revealed several large clusters of features whose abundances were largely abolished or greatly increased in glo-1 worms. Notably, although some differential MS² clusters represented known compounds, for example ascarosides, the majority of clusters were found to represent previously undescribed metabolite families.

In agreement with previous studies (Panda et al., 2017), biosynthesis of most modular ascarosides was abolished or substantially reduced in glo-1 mutants, including all 4′-modified ascarosides, e.g. icas#3 (1) (Figure 1b, Figure 2—figure supplement 5a). Similarly, production of ascarosides modified at the carboxy terminus, e.g. uglas#11 (3) derived from ester formation between ascr#1 (5) and uric acid glucoside (Curtis et al., 2020) (6), and ascr#8 (2), derived from formation of an amide bond between ascr#7 (7) and of p-amino benzoic acid (8), was largely abolished in glo-1 mutants (Figure 1a–b, Figure 2—figure supplement 5a). Metabolites plausibly representing building blocks of these modular ascarosides were not strongly perturbed in glo-1 mutants (Figure 1—figure supplement 2). For example, abundances of unmodified ascarosides, for example ascr#3 (9) and ascr#10 (10), or metabolites representing 4′-modifications, for example indole-3-carboxylic acid (11) and octopamine succinate (12), were not significantly perturbed in the mutant (Figure 1a, Figure 2—figure supplement 5a, Figure 1—figure supplement 2). In contrast, a subset of modular ascaroside glucose esters (e.g. iglas#1 (13) and glas#10 (14), Figure 1e), was strongly increased in glo-1 mutants (Figure 2—figure supplement 5b). These results suggest that glo-1-dependent intestinal organelles function as a central hub for the biosynthesis of most modular ascarosides, with the exception of a subset of ascarosylated glucosides, whose increased production in glo-1 mutants may be indicative of a shunt pathway for ascarosyl-CoA derivatives (Zhang et al., 2015; Zhang et al., 2016; Zhang et al., 2018), which represent plausible precursors for modular ascarosides modified at the carboxy terminus.

Next, we analyzed the most prominent MS² clusters representing previously uncharacterized metabolites whose production is abolished or strongly reduced in glo-1 mutants (Figure 2). Detailed analysis of their MS² spectra indicated that they may represent a large family of modular hexose derivatives incorporating moieties from diverse primary metabolic pathways. For example, MS² spectra from clusters I, II, and III of the positive-ionization network suggested phosphorylated hexose glycosides of indole, anthranilic acid, tyramine, or octopamine, which are further decorated with a wide variety of fatty acyl moieties derived from fatty acid or amino acid metabolism, for example nicotinic acid, pyrrolic acid, or tiglic acid (Figure 2, Appendix 1—table 1; Coburn and Gems, 2013; Stupp et al., 2013). Given the previous identification of the glucosides iglu#1/2 (15/16, Figure 2e) and angl#1/2 (17/18), we hypothesized that clusters I, II, and III represent a modular library of glucosides, in which N-glucosylated indole, anthranilic acid, tyramine, or octopamine (O'Donnell et al., 2020) serve as scaffolds for attachment of diverse building blocks. To further support these structural assignments, a series of modular metabolites based on N-glucosylated indole (‘iglu’) were selected for total synthesis. Synthetic standards for the non-phosphorylated parent compounds of iglu#4 (19), iglu#6 (20), iglu#8 (21), and iglu#10 (22) matched HPLC retention times and MS² spectra of the corresponding natural compounds (Figure 2—figure supplement 6), confirming their structures and enabling tentative structural assignments for a large number of additional modular glucosides, including their phosphorylated derivatives, e.g. iglu#12 (23), iglu#41 (24), angl#4 (cluster II, 25), and tyglu#4 (cluster III, 26) (Figure 2). The proposed structures include several glucosides of the neurotransmitters tyramine and octopamine, whose incorporation could be verified by comparison with data from a recently described feeding experiment with stable isotope-labeled tyrosine (O'Donnell et al., 2020). Similar to ascaroside biosynthesis, the production of modular glucosides is life stage dependent; for example, production of specific tyramine glucosides peaks at the L3 larval stage, whereas production of angl#4 increases until the adult stage (Figure 2—figure supplement 8). Notably, modular glucosides were detected primarily as their phosphorylated derivatives, as respective non-phosphorylated species were generally less abundant. In contrast to most ascarosides, the phosphorylated glucosides are more abundant in the endo-metabolome than the exo-metabolome, suggesting that phosphorylated glucosides may be specifically retained in the body (Figure 2—figure supplement 7).

As in the case of modular ascarosides, the abundances of putative building blocks of the newly identified modular glucosides were not strongly perturbed in glo-1 mutants. For example, abundances of anthranilic acid, indole, octopamine, and tyramine were not significantly affected in glo-1 null animals (Figure 2—figure supplement 9). Notably, abundances of the glucosides scaffold, e.g. iglu#1 and angl#1, were also largely unaltered or even slightly increased in glo-1 mutants (Figure 2—figure supplement 9). In addition, production of some of the identified modular glucosides, e.g. iglu#5, is reduced but not fully abolished in glo-1 worms (Figure 2—figure supplement 6).

To confirm our results, we additionally compared the glo-1 metabolome with that of glo-4 mutants. glo-4 encodes a predicted guanyl-nucleotide exchange factor acting upstream of glo-1, and like glo-1 mutants, glo-4 worms do not form LROs (Hermann et al., 2005). We found that the glo-4 metabolome closely resembles that of glo-1 worms, lacking most of the modular ascarosides and ascarosides detected in wildtype worms (Figure 2—figure supplement 5c). Correspondingly, similar sets of compounds are upregulated in glo-1 and glo-4 mutants relative to wild type, including ascarosyl glucosides and ascaroside phosphates. Compounds accumulating in glo-1 and glo-4 mutant worms further include a diverse array of small peptides (primarily three to six amino acids), consistent with the proposed role of LROs in the breakdown of peptides derived from proteolysis (Figure 2—figure supplement 10; Bird et al., 2009). Taken together, our results indicate that, in addition to their roles in the degradation of metabolic waste, the LROs serve as hotspots of biosynthetic activity, where building blocks from diverse metabolic pathways are attached to glucoside and ascaroside scaffolds (Figure 1a).

Comparing the relative abundances of different members of the identified families of modular glucosides and ascarosides, it appears that combinations of different building blocks and scaffolds are highly specific, suggesting the presence of dedicated biosynthetic pathways. For example, uric acid glucoside, gluric#1 (6), is preferentially combined with an ascaroside bearing a seven-carbon side chain (to form uglas#11, 3), whereas ascarosides bearing a nine-carbon side chain are preferentially attached to the anomeric position of free glucose, as in glas#10 (14) (Artyukhin et al., 2018; von Reuss et al., 2012). Similarly, tiglic acid is preferentially attached to indole and tyramine glucosides but not to anthranilic acid glucosides (Appendix 1—table 1). Given that 4′-modification of ascarosides in P. pacificus and C. elegans require cest homologs, we hypothesized that the biosynthesis of other modular ascarosides as well as the newly identified glucosides may be under the control of cest family enzymes (Falcke et al., 2018; Faghih et al., 2020). From a list of 44 uar-1 homologs from BLAST analysis (Appendix 1—table 2), we selected seven for further study (Figure 3a, Appendix 1—table 3). The selected homologs are predicted to have intestinal expression, one primary site of small molecule biosynthesis in C. elegans (Artyukhin et al., 2018), and are closely related to the UAR-1 gene, while representing different sub-branches of the phylogenetic tree. Utilizing a recently optimized CRISPR/Cas9 method, we obtained two null mutant strains for five of the selected genes (Wang et al., 2018). Mutants for the remaining two homologs, ges-1 and cest-6, had been previously obtained (Appendix 1—table 3). We then analyzed the exo- and endo-metabolomes of this set of mutant strains by HPLC-HRMS to identify features that are absent or strongly downregulated in null mutants of a specific candidate gene compared to wildtype worms and all other mutants in this study. We found that two of the seven tested homologs (cest-1.1, cest-2.2) are defective in the production of two different families of modular ascarosides, whereas cest-4 mutants were defective in the biosynthesis of a specific subset of modular indole glucosides (Figure 3). The metabolomes of mutants for the remaining four cest homologs did not exhibit any significant differences compared to wild type under the tested conditions.

Analysis of the metabolomes of the two cest-2.2 null mutants revealed loss of dauer pheromone component and male attractant ascr#8 (2) as well as of the closely related ascr#81 (27) and ascr#82 (28) (Figure 3b, Figure 3—figure supplements 1 and 2b). Biosynthetically, the ascr#8 family of ascarosides are derived from amide formation between ascr#7 (ΔC7, 7) and folate-derived p-aminobenzoic acid (PABA, 8), PABA-glutamate (29), or PABA-diglutamate, respectively. We did not detect any significant reduction in the production of plausible ascr#8 precursors, including PABA and PABA-glutamate, or ascr#7 (Figure 3b, Figure 3—figure supplement 2b). Biosynthesis of ascr#8, ascr#81, and ascr#82 was recovered in cest-2.2 mutant worms in which the cest-2.2 sequence had been restored to wild type using CRISPR/Cas9 (Figure 3c, Figure 3—figure supplement 3b). These results indicate that CEST-2.2 is required specifically for biosynthesis of the amide linkage between the carboxy terminus of ascr#7 and PABA derivatives, in contrast to the implied functions of UAR-1, CEST-8, CEST-3, and CEST-9.2, which are involved in the formation of ester bonds between various head groups and the 4′-hydroxy group of ascarylose (Falcke et al., 2018; Faghih et al., 2020).

In cest-1.1 null mutants (cest-1.1(null)), biosynthesis of the nucleoside-like ascaroside uglas#1 (30) and its phosphorylated derivative uglas#11 (3) was abolished (Figure 3d, Figure 3—figure supplement 2a). uglas#1 and uglas#11 are derived from the attachment of ascr#1, bearing a seven carbon (C7) side chain, to the uric acid gluconucleoside gluric#1 (6). Production of ascr#1 (5) and gluric#1 (6), representing plausible building blocks of uglas#1 (30), was not reduced (Figure 3—figure supplement 2a). Furthermore, production of uglas#14 (31) and uglas#15 (32), isomers of uglas#1 and uglas#11 bearing the ascarosyl moiety at the 6′ position instead of the 2′ position, was not abolished but rather slightly increased in cest-1.1(null) (Figure 3d–e). These results indicate that CEST-1.1 is required for the formation of the ester bond specifically between ascr#1 (5) and the 2′-hydroxyl group in gluric#1. As in the case of cest-2.2, biosynthesis of uglas#1 and uglas#11 was fully recovered in cest-1.1 mutant worms in which the cest-1.1 sequence had been restored to wild type using CRISPR/Cas9 (Figure 3f, Figure 3—figure supplement 3a).

Sequence alignment with human AChE suggested that serine 213 is part of the conserved catalytic serine-histidine-glutamate triad of CEST-1.1 (Figure 4—figure supplement 1). To test whether disruption of the catalytic triad would affect production of cest-1.1-dependent metabolites, we generated a point mutant, cest-1.1(S213A). As in cest-1.1(null), production of uglas#1 (30) and uglas#11 (3) was fully abolished in cest-1.1(S213A), whereas production of gluric#1 was not affected (Figure 3g).

Previous work implicated cest-1.1 with longevity phenotypes associated with argonaute-like gene 2 (alg-2) (Aalto et al., 2018). alg-2 mutant worms are long lived compared to wild type and their long lifespan was further shown to require daf-16, the sole ortholog of the FOXO family of transcription factors in C. elegans, as well as cest-1.1. Moreover, uglas#11 biosynthesis is significantly increased in mutants of the insulin receptor homolog daf-2, a central regulator of lifespan in C. elegans upstream of daf-16 (Curtis et al., 2020). These findings suggest the possibility that the production of uglas ascarosides underlies the cest-1.1-dependent extension of adult lifespan in C. elegans.

In contrast to our results for cest-1.1 and cest-2.2 mutants, comparative metabolomic analysis of the cest-4 mutant strains did not reveal any defects in the biosynthesis of known ascarosides. Instead, we found that the levels of a specific subset of modular anthranilic acid (33) bearing indole glucosides, including iglu#3 (34) and its phosphorylated derivative iglu#4 (35) were abolished in the cest-4 mutant worms (Figure 3h). Abundances of the putative precursor glucosides, iglu#1 (15) and iglu#2 (16), were not significantly changed in cest-4 (Figure 3i, Figure 3—figure supplement 2c). Notably, production of other indole glucosides, e.g. iglu#6 (36) and iglu#8 (37), was not significantly reduced in cest-4 worms (Figure 3j, Figure 3—figure supplement 4). Biosynthesis of iglu#3 and iglu#4 was restored to wild-type levels in genetic revertant strains for cest-4 (Figure 3i, Figure 3—figure supplement 3c). Therefore, it appears that cest-4 is specifically required for attachment of anthranilic acid to the 6′ position of glucosyl indole precursors, whereas attachment of tiglic acid, nicotinic acid, and other moieties is cest-4-independent (Figure 3j, Figure 3—figure supplement 4). The role of cest-4 in the biosynthesis of the iglu family of modular glucosides thus parallels that of cest-1.1 in the biosynthesis of the uglas ascarosides: whereas cest-4 appears to be required for the attachment of anthranilic acid (33) to the 6’ position of a range of indole glucosides, cest-1.1 appears to be required for attaching the ascr#1 side chain to the 2′ position in uric acid glucosides.

All cest homologs selected for this study exhibit domain architectures typical of the α/ß-hydrolase superfamily of proteins, including a conserved catalytic triad, and further contain a predicted disulfide bridge, as in mammalian AChE (Soreq and Seidman, 2001; Figure 4—figure supplement 1). The cest genes also share homology with neuroligin, a membrane bound member of the α/ß-hydrolase fold family, that mediates the formation and maintenance of synapses between neurons (Bemben et al., 2015). Sequence analysis suggests that five of the seven CEST homologs studied here are membrane anchored, given the presence of a predicted C-terminal transmembrane domain (Krogh et al., 2001) (consisting of ~20 residues), with the N terminus on the luminal side of a vesicle or organelle (Figure 4—figure supplement 2). Since the production of all so far identified cest-dependent metabolites is abolished in glo-1 mutants, it seemed likely that the CEST proteins localize to intestinal granules. To test this idea, we created a mutant strain that express cest-2.2 C-terminally tagged with mCherry at the native genomic locus to avoid potentially confounding effects of overexpression. The red fluorescent mCherry was chosen because of the strong green autofluorescence of the LROs (Coburn and Gems, 2013). We confirmed that production of all cest-2.2-dependent metabolites, including ascr#8 (2), ascr#81 (27), and ascr#82 (28) was not significantly altered in cest-2.2-mCherry mutants (Figure 4a), indicating that CEST-2.2 remained functional. Imaging of wild-type adult worms revealed strong green and weaker red autofluorescence in circular features in intestinal cells, consistent with LROs. In addition, cest-2.2-mCherry-tagged worms showed red fluorescence in a distinct set of intestinal granules that showed little if any autofluorescence (Figure 4b, Figure 4—figure supplements 3–4). It is unclear whether mCherry also localizes to the strongly autofluorescent granules, as we cannot distinguish the mCherry signal from the red component of the autofluorescence, given relatively low CEST-2.2-mCherry expression in this non-overexpressing strain. Taken together, it appears that CEST-2.2-mCherry localizes to a subset of intestinal organelles that is partly distinct from the autofluorescent LROs. Further studies are required to determine if CEST-2.2-mCherry co-localizes with other intestinal granule markers, specifically GLO-1 and the lysosomal marker LMP-1.

In addition to C. elegans and P. pacificus, modular ascarosides have been reported from several other Caenorhabditis species (Dong et al., 2020; Kanzaki et al., 2018), including C. briggsae (Dong et al., 2016; von Reuss, 2018). To assess whether the role of LROs in the biosynthesis of modular metabolites is conserved across species, we created two Cbr-glo-1 (CBG01912.1) knock-out strains using CRISPR/Cas9. As in C. elegans, Cbr-glo-1 mutant worms lacked autofluorescent LROs, which are prominently visible in wild-type C. briggsae (Figure 5—figure supplement 1). Comparative metabolomic analysis of the endo- and exo-metabolomes of wild-type C. briggsae and the Cbr-glo-1 mutant strains revealed that biosynthesis of all known modular ascarosides is abolished in Cbr-glo-1 worms, including the indole carboxy derivatives icas#2 (35) and icas#6.2 (36), which are highly abundant in wild-type C. briggsae (Figure 5a; Dong et al., 2016). In addition, the C. briggsae MS² networks included several large Cbr-glo-1-dependent clusters representing modular glucosides, including many of the compounds also detected in C. elegans, for example iglu#4 and angl#4. As in C. elegans, production of unmodified glucoside scaffolds, e.g. iglu#1 (15) and angl#1 (17), was not reduced or increased in Cbr-glo-1 mutants, whereas biosynthesis of most modular glucosides derived from attachment of additional moieties to these scaffolds was abolished (Figure 5b). Taken together, these results indicate that the role of LROs as a central hub for the assembly of diverse small molecule architectures, including modular glucosides and ascarosides, may be widely conserved among nematodes (Figure 5c).

Unless noted otherwise, all reagents were purchased from Sigma-Aldrich. All newly identified compounds were assigned four letter 'SMID's (a search-compatible, Small Molecule IDentifier) for example ‘icas#3’ or ‘ascr#10’. For a list of all compounds referred to in the text and figures, see Appendix 1—table 9. The SMID database (www.smid-db.org) is an electronic resource maintained in collaboration with WormBase (www.wormbase.org). A complete list of SMIDs can be found at www.smid-db.org/browse, and example structures for different SMIDs at www.smid-db.org/smidclasses.

Amino acid sequence of Ppa-UAR-1 was used as previously published (Falcke et al., 2018). BLASTp was run from the WormBase engine at (https://wormbase.org/tools/blast_blat). E-value threshold was set to 1E0. Database was set to WS269 and species was set to C. elegans. Results of BLASTp search are listed in Appendix 1—table 2.

hAChE was aligned with Ppa-UAR-1, CEST-1.1, CEST-2.2, and CEST-4 was done using T-Coffee Multiple Sequence alignment (Notredame et al., 2000). Protein sequences for C. elegans CEST proteins are from WormBase. The AChE sequence was obtained from NCBI (accession number P22303). Amino acids were colored based on chemical properties: AVFPMILW = red (small + hydrophobic), DE = blue (acidic), RHK = magenta (basic), STYHCNGQ = green (hydroxyl + sulfhydryl + amine + glycine). See Figure 4—figure supplement 1 for results.

The protein sequence of Ppa-UAR1 was submitted to an NCBI BLASTp search (Altschul et al., 2005) (restricted to species C. elegans, conditional compositional BLOSUM62, gap open cost:11, gap extension cost: 1, word size: 6) using Geneious software (Biomatters Inc). The top BLAST hits by E-value up to and including ace-3 were selected, and only the best scoring transcript variant was kept for each protein sequence hit. A total of 28 sequences were then imported into MEGA7 (Kumar et al., 2016) and aligned using MUSCLE (Edgar, 2004) (settings: gap open penalty: −2.9, gap extend 0, hydrophobicity multiplier 1.2, max. iterations 8, clustering method for all iterations: UPGMB, minimal diagonal length: 24). From this alignment, an Maximum Likelihood tree was built based on the JTT matrix-based model (Jones et al., 1992). Initial trees were built by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model assuming uniform substitution rates across positions. Phylogeny confidence was tested using 200 bootstrap replications. The tree with the highest log likelihood (−22299.9282) is shown. At each branch, the percentage of bootstrap replicates containing the same branching event is denoted. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. The evolutionary history was inferred by using the Maximum Likelihood method based on the JTT matrix-based model (Jones et al., 1992). The tree with the highest log likelihood (−22299.9282) is shown. The percentage of trees in which the associated taxa clustered together is shown next to the branches. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, and then selecting the topology with superior log likelihood value. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. The analysis involved 28 amino acid sequences. All positions containing gaps and missing data were eliminated. There were a total of 427 positions in the final dataset. Evolutionary analyses were conducted in MEGA7 (Kumar et al., 2016; Felsenstein, 1985).

Wild-type (N2) and glo-1(zu437) null animals were provided by the Caenorhabditis Genetics Center (CGC), which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440). cest-2.2 mutant strains integrating N-terminal (mCherry-cest-2.2) or C-terminal mCherry (cest-2.2-mCherry) were generated by SunyBiotech. Generation of C. elegans and C. briggsae null mutants and revertants as well as generation of the cest-1.1 point mutant is described below. See Appendix 1—table 3 for a complete list of strains used in this study.

CRISPR/Cas9 mutagenesis was performed as in Wang et al., 2018. Briefly, C. elegans strain N2 was gene-edited by insertion of a 43-base-pair insertion that disrupts translation Appendix 1—table 8. Independent homozygous mutants were picked among the progeny of heterozygous F1 progeny of injected hermaphrodites and given distinct unique allele names. Reversion of mutants was accomplished in the same way.

The C. briggsae glo-1 mutants sy1382 and sy1383 were both created using the briggsae adaptation of the STOP-IN cassette method as described in Cohen and Sternberg, 2019 and Wang et al., 2018. Both strains were made using a successful insertion of the STOP-IN cassette into the middle of the first exon using the guide AACAAATCTCCGGATGATTG. To detect the insertion, we used forward primer GGGTGACCGCCCATTTATTG and reverse primer AAAGGCGCACATCTTGCTTC.

cest-1(dp683) was generated as previously described (Paix et al., 2015). Briefly, daf-2(e1368) mutant animals were injected with in-vitro-assembled Cas9-crRNA-tracrRNA complexes targeting cest-1.1 and the dpy-10 co-CRISPR gene and two 100 bp repair oligonucleotides containing the desired cest-1.1 mutation and the dpy-10(cn64) co-CRISPR mutation (Arribere et al., 2014). Sequences of the cest-1.1 crRNA and repair oligonucleotide are 5’ acctacCGCTACTATCATAC 3’ and 5’ GAAATTGAAAACTTTGGAGGAAATAAAAACAGAATTACATTGGCAGGGCATGCCGCTGGAGCAAGTATGATAGTAGCGgtaggtcacataaatgatacatttttg 3’, respectively. F1 Rol progeny of injected animals were picked and screened for the presence of the cest-1.1(dp683) mutation after egglay. F2 broods of F1 Rol animals that were heterozygous for cest-1.1(dp683) were screened for animals that were homozygous for cest-1.1(dp683) and either wild-type or heterozygous for cn64 at the dpy-10 locus. Subsequent broods were screened for wild-type dpy-10 animals to remove the co-CRISPR mutation.

To image, gravid adult C. elegans were transferred to an agarose pad on a glass slide with 10 µM of levamisole to immobilize the worms. Microscopic analysis was performed using a Leica TCS SP5 Laser Scanning Confocal Microscope. Green autofluorescence was excited at 488 nm and the emission detector was set to 490–540 nm. mCherry was excited with 561 nm and the emission detector was set to 590–650 nm. Worms were imaged using the 100x objective.

0.5 mL of 2 µM Lysotracker Deep Red (Thermo Fisher 1 mM stock in DMSO) was added to a 6 cm NGM plate seeded with 0.1 mL of E. coli OP50 and incubated in the dark for 24 hr at 20°C. L4 larvae of C. briggsae were added to the plate and allowed to grow in the dark for 24 hr at 20°C. To image, C. briggsae were transferred to an agarose pad on a glass slide with 10 µM of levamisole to immobilize the worms. Microscopic analysis was performed using a Zeiss Axio Imager Z2 florescence microscope with Apotome.

Culturing began by chunking C. elegans or C. briggsae onto 10 cm NGM plates (each seeded with 800 µL of OP50 E. coli grown to stationary phase in Lennox Broth) and incubated at 22°C. Once the food was consumed, the cultures were incubated for an additional 24 hr. Each plate was then washed with 25 mL of S-complete medium into a 125 mL Erlenmeyer flask, and 1 mL of OP50 E. coli was added (E. coli cultures were grown to stationary phase in Terrific Broth, pelleted and resuspended at 1 g wet mass per 1 mL M9 buffer), shaking at 220 RPM and 22°C. After 70 hr, cultures were centrifuged at 5000 G for 1 min. After discarding supernatant, 24 mL H₂O was added, along with 6 mL bleach, 900 µL 10 M NaOH and the mixture was shaken for 3 min to prepare eggs. Eggs were centrifuged at 5000 G, the supernatant was removed, and the egg pellet washed with 35 mL M9 buffer twice and then suspended in a final volume of 5 mL M9 buffer in a 50 mL centrifuge tube. Eggs were counted and placed on a rocker and allowed to hatch as L1 larvae for 24 hr at 22°C. 70,000 L1 larvae were seeded in 25 mL cultures of S-complete with 1 mL of OP50 and incubated at 220 RPM and 22°C in a 125 mL Erlenmeyer flask. After 72 hr, cultures were fed an additional 1 mL of OP50 and incubation continued. After an additional 48 hr, worms were spun at 1000 G 5 min and spent medium was separated from worm body pellet. Separated medium and worm pellet were flash frozen over liquid nitrogen until further processing. At least three biological replicates were grown for all mutant strains. Mutants were grown with parallel wildtype controls, and biological replicates were started on different days.

Lyophilized pellet and media samples were crushed and homogenized by shaking with 2.5 mm steel balls at 1300 rpm for 3 min in 30 s pulses while chilled with liquid nitrogen (SPEX sample prep miniG 1600). Thus powdered media and pellet samples were extracted with 15 mL methanol in 50 mL centrifuge tubes, rocking overnight at 22°C. Extractions were pelleted at 5000 g for 10 min at 4°C, and supernatants were transferred to 20 mL glass scintillation vials. Samples were then dried in a SpeedVac (Thermo Fisher Scientific) vacuum concentrator. Dried materials were resuspended in 1 mL methanol and vortexed for 1 min. Samples were pelleted at 5000 g for 5 min and 22°C, and supernatants were transferred to 2 mL HPLC vials and dried in a SpeedVac vacuum concentrator. Samples were then resuspended in 200 μL of methanol, transferred into 1.7 mL Eppendorf tubes, and centrifuged at 18,000 G for 20 min at 4°C. Clarified extracts were transferred to fresh HPLC vials and stored at −20°C until analysis.

40,000 synchronized L1 larvae were added to 125 mL Erlenmeyer flasks containing 30 mL of S-complete medium. Worms were fed with 4 mL of concentrated OP-50 and incubated at 20°C with shaking at 160 RPM for: 12 hr (L1), 24 hr (L2), 32 hr (L3), 40 hr (L4) and 58 hr (gravid adults). For preparation of starved samples, each of the stages was starved for 24 hr after reaching their desired developmental stage in S-complete without OP-50. After incubation for the desired time, liquid cultures were centrifuged (1000 x g, 22°C, 1 min) and supernatants were collected. Supernatant was separated from intact OP-50 cells by centrifuging (3000 x g, 22°C, 5 min) and the resulting supernatants (exo-metabolome) were lyophilized. Lyophilized samples were homogenized with a dounce homogenizer in 10 mL methanol and extracted on a stirring plate (22°C, 12 hr). The resulting suspension was centrifuged (4000 g, 22°C, 5 min) to remove any precipitate before carefully transferring to an LC-MS sample vial. Three biological replicates were started on different days.

High resolution LC-MS analysis was performed on a Thermo Fisher Scientific Vanquish Horizon UHPLC System coupled with a Thermo Q Exactive HF hybrid quadrupole-orbitrap high-resolution mass spectrometer equipped with a HESI ion source. 1 μL of extract was injected and separated using at water-acetonitrile gradient on a Thermo Scientific Hypersil GOLD C18 column (150 mm x 2.1 mm 1.9 um particle size 175 Å pore size, Thermo Scientific) and maintained at 40°C. Solvents were all purchased from Fisher Scientific as HPLC grade. Solvent A: 0.1% formic acid in water; solvent B: 0.1% formic acid in acetonitrile. A/B gradient started at 1% B for 5 min, then from 1% to 100% B over 20 min, 100% for 5 min, then down to 1% B for 3 min. Mass spectrometer parameters: 3.5 kV spray voltage, 380°C capillary temperature, 300°C probe heater temperature, 60 sheath flow rate, 20 auxiliary flow rate, one spare gas; S-lens RF level 50.0, resolution 240,000, m/z range 100–1200 m/z, AGC target 3e6. Instrument was calibrated with positive and negative ion calibration solutions (Thermo-Fisher) Pierce LTQ Velos ESI pos/neg calibration solutions.

LC−MS RAW files from each sample were converted to mzXML (centroid mode) using MSConvert (ProteoWizard), followed by analysis using the XCMS (Smith et al., 2006) analysis feature in METABOseek (metaboseek.com). Peak detection was carried out with the centWave algorithm (Tautenhahn et al., 2008), values set as: 4 ppm, 320 peakwidth, 3 snthresh, 3100 prefilter, FALSE fitgauss, 1 integrate, TRUE firstBaselineCheck, 0 noise, wMean mzCenterFun, −0.005 mzdiff. XCMS feature grouping values were set as: 0.2 minfrac, 2 bw, 0.002 mzwid, 500 max, 1 minsamp, FALSE usegroup. METABOseek peak filling values set as: 5 ppm_m, 5 rtw, TRUE rtrange. Resulting tables were then processed with the METABOseek Data Explorer. Molecular features were filtered for each particular null mutant against all other mutants. Filter values were set as: 10 to max minFoldOverCtrl, 15000 to max meanInt, 120 to 1500 rt, 0.95 to max Peak Quality as calculated by METABOseek. Features were then manually curated by removing isotopic and adducted redundancies. Remaining masses were put on the inclusion list for MS/MS (ddMS2) characterization. Positive and negative mode data were processed separately. In both cases we checked if a feature had a corresponding peak in the opposite ionization mode, since fragmentation spectra in different modes often provide complementary structural information. To acquire MS2 spectra, we ran a top-10 data dependent MS2 method on a Thermo QExactive-HF mass spectrometer with MS1 resolution 60,000, AGC target 1 × 10^6, maximum IT (injection time) 50 ms, MS2 resolution 45,000, AGC target 5 × 10^5, maximum IT 80 ms, isolation window 1.0 m/z, stepped NCE (normalized collision energy) 25, 50, dynamic exclusion 3 s.

Peak integration data from HPLC-MS analysis were log-transformed (Karpievitch et al., 2012) prior to statistical analysis. Significance of differences between average peak areas were then assessed using unpaired t-tests.

For the differential featuresidentified above, MS² data was acquired. To generate the MS² molecular network, Metaboseek version 0.9.6 was used. Using the MS2scans function, differential features were matched with their respective MS² scan, using an m/z window of 5 ppm, and a retention time window of 15 s. To construct the molecular network, tolerance of the fragment peaks was set to m/z of 0.002 or 5 ppm, minimum number of peaks was set to 5, with a 2% noise level. Once the network was constructed, a cosine value of 0.8 was used, and the number of possible connections was constrained to 5.

The serine hydrolase list was reported previously (Chen et al., 2019). From this list, sequences were inputted into Geneious Prime (version 2020.1.2 Biomatters). Sequences were aligned using Clustal Omega, neighbor joining alignment. Dendrogram tree was generated using the Geneious Tree Builder; Genetic distance model Jukes-Cantor, Tree build method UPGMA, no outgroup, Bootstrap resampling, random seed 508,949, 300 interactions, support threshold of 1. CEST enzymes were colored red and PPA-UAR-1 was colored blue (Figure 1—figure supplement 1).

To a solution of anthranilic acid (33, 300 mg, 2.18 mmol) in 4 mL of THF and H₂O (1:1), Boc-anhydride (521 mg, 2.39 mmol) was added, and 2 M NaOH was added to the mixture until pH 10 was reached. The reaction mixture was stirred at room temperature. After 23 hr, the solution was concentrated in vacuo, and 15% citric acid aqueous solution was added until pH 4 was reached. The white precipitate was filtered off and dried under vacuum to provide 2-((tert-butoxycarbonyl)amino)benzoic acid (SI-1, 497 mg, 96%) as a white solid. ¹H NMR, 600 MHz, chloroform-d: δ (ppm) 10.06 (s, 1H), 8.47 (dd, J = 8.7, 0.9 Hz, 1H), 8.08 (dd, J = 7.9, 1.5 Hz, 1H), 7.57 (dt, J = 7.9, 1.5 Hz, 1H), 7.03 (dt, J = 7.2, 1.2 Hz, 1H), 1.55 (s, 9H).

To a stirred solution of N-(tert-butoxycarbonyl)anthranilic acid (Krueger et al., 2008) (SI-1, 10 mg, 0.042 mmol) in dimethylformamide, 1-(3-dimethylaminopropyl)−3-ethylcarbodiimide hydrochloride (EDC·HCl, 20.1 mg, 0.105 mmol) was added. The mixture was stirred at room temperature for 5 min, and 4-dimethylaminopyridine (DMAP, 18.1 mg, 0.105 mmol) and N-β-glucopyranosyl indole (iglu#1, 15, 9.8 mg, 0.0351 mmol) were added. The reaction mixture was stirred at room temperature. After 5 hr, the mixture was concentrated in vacuo to yield a viscous oil, which was dissolved in 1.4 mL of a 5:2 mixture of dichloromethane and methanol. Trifluoroacetic acid (TFA, 0.5 mL) was added slowly and the reaction mixture was stirred at room temperature. After 3 hr, the mixture was concentrated in vacuo. Preparative HPLC provided a pure sample of iglu#3 (34, 0.8 mg, 5.7%). See Appendix 1—table 4 for NMR spectroscopic data of iglu#3.

HRMS (ESI) m/z: [M - H]^- calcd for C₂₁H₂₁N₂O₆^- 397.13938; found 397.14017.

To a stirred solution of nicotinic acid (7.3 mg, 0.059 mmol) in a mixture of dimethylformamide and dichloromethane (1:1), EDC·HCl (28.4 mg, 0.148 mmol) was added. The mixture was stirred at room temperature for 30 min, before DMAP (18.1 mg, 0.148 mmol) and N-β-glucopyranosyl indole (iglu#1, 15, 13.8 mg, 0.0494 mmol) were added. The reaction mixture was stirred at room temperature for 20 hr, the mixture was concentrated in vacuo, and flash column chromatography on silica using a gradient of 0–25% methanol in dichloromethane afforded iglu#5 (SI-2, 2.5 mg, 13.9%) as a colorless oil. See Appendix 1—table 5 for NMR spectroscopic data of iglu#5.

HRMS (ESI) m/z: [M + H]⁺ calcd for C₂₀H₂₁N₂O₆⁺ 385.13941; found 385.14038.

To a stirred solution of tiglic acid (5.0 mg, 0.050 mmol) in a 1:1 mixture of dimethylformamide and dichloromethane, EDC·HCl (23.9 mg, 0.125 mmol) was added. The mixture was stirred at room temperature for 30 min, and DMAP (15.2 mg, 0.125 mmol) and N-β-glucopyranosyl indole (iglu#1, 15, 11.6 mg, 0.0416 mmol) were added. The reaction mixture was stirred at room temperature for 22 hr and then concentrated in vacuo. Flash column chromatography on silica using a gradient of 0–30% methanol in dichloromethane afforded iglu#7 (SI-3, 2.5 mg, 11.3%) as a colorless oil. See Appendix 1—table 6 for NMR spectroscopic data of iglu#7.

HRMS (ESI) m/z: [M + H]⁺ calcd for C₁₉H₂₄NO₆⁺ 362.15981; found 362.16025.

To a suspension of pyrrole-2-carboxylic acid (6.0 mg, 0.054 mmol) in dichloromethane, oxalyl chloride (14 µL, 0.163 mmol) was added slowly, followed by dimethylformamide (1 µL, 0.0129 mmol). The mixture was stirred at room temperature for 18 hr and then concentrated to dryness in vacuo. The residue was re-dissolved in dimethylformamide (2 mL) containing N-β-glucopyranosyl indole (iglu#1, 15, 10.8 mg, 0.0387 mmol). Triethylamine (45 µL, 0.324 mmol) was added, and the reaction was stirred at 35°C for 7 days. Subsequently the mixture was concentrated in vacuo, and flash column chromatography on silica using a gradient of 0–30% methanol in dimethylformamide afforded iglu#9 (SI-4, 1.5 mg, 10.4%) as a colorless oil. See Appendix 1—table 7 for NMR spectroscopic data of iglu#9.

HRMS (ESI) m/z: [M + H]⁺ calcd for C₁₉H₂₁N₂O₆⁺ 373.13941; found 373.14026.

To a stirred solution of Boc-AA (2 mg, 0.00 84 mmol) in dimethylformamide, 1-(3-dimethylaminopropyl)−3-ethylcarbodiimide hydrochloride (3.9 mg, 0.0203 mmol) was added. The mixture was stirred at room temperature for 5 min, and 4-dimethylaminopyridine (2.5 mg, 0.0203 mmol) and angl#1 (17, 2 mg, 0.0068 mmol) were added. The reaction mixture was stirred at room temperature. After 5 hr, the mixture was concentrated in vacuo. The crude product was dissolved in 0.55 mL dichloromethane and methanol (10:1), and trifluoroacetic acid (500 µL) was added slowly. The reaction mixture was stirred at room temperature for 3 hr and then was concentrated in vacuo, affording angl#3 (SI-5).

HRMS (ESI) m/z: [M + H]⁺ calcd for C₂₀H₂₃N₂O₇⁺ 403.14998; found 403.15100.

p-Aminobenzoic acid (Chem-Impex) (8) was dissolved in warm dichloromethane (DCM) containing triethylamine (0.1 eq). EDC·HCl (Amresco Biochemicals) (1 eq), and di-tert-butyl glutamate (1 eq) was added to the reaction mixture. N,N-Dimethylaminopyridine (1.1 eq) was then added and the resulting mixture was stirred at room temperature for 24 hr and then extracted with ethyl acetate. The combined extractswere dried with sodium sulfate and evaporated to dryness in vacuo. The crude product was dissolved in DCM, and trifluoroacetic acid was added (100 eq). The reaction was then stirred for 6 hr at room temperature. TFA and DCM were evaporated off to yield crude PABA-glutamate (29). ¹H NMR, 600 MHz, methanol-d₄: δ (ppm) 7.93 (d, J = 8.6 Hz, 2H), 7.37 (d, J = 8.5 Hz, 2H), 4.61 (dd, J = 5.0, 9.3 Hz, 1H), 2.09–2.28 (m, 4H).

NMR spectra appendix. NMR spectra of synthetic intermediates and newly identified metabolites.