Host–microbe interactions rewire metabolism in a C. elegans model of leucine breakdown deficiency

Abstract

In humans, defects in leucine catabolism cause a variety of inborn errors in metabolism. Here, we use Caenorhabditis elegans to investigate the impact of mutations in mccc-1, an enzyme that functions in leucine breakdown. Through untargeted metabolomic and transcriptomic analyses we find extensive metabolic rewiring that helps to detoxify leucine breakdown intermediates via conversion into previously undescribed metabolites and to synthesize mevalonate, an essential metabolite. We also find that the leucine breakdown product 3,3-hydroxymethylbutyrate (HMB), commonly used as a human muscle-building supplement, is toxic to C. elegans and that bacteria modulate this toxicity. Unbiased genetic screens revealed interactions between the host and microbe, where components of bacterial pyrimidine biosynthesis mitigate HMB toxicity. Finally, upregulated ketone body metabolism genes in mccc-1 mutants provide an alternative route for biosynthesis of the mevalonate precursor 3-hydroxy-3-methylglutaryl-CoA. Our work demonstrates that a complex host–bacteria interplay rewires metabolism to allow host survival when leucine catabolism is perturbed.

In animals and humans, roughly 80% of dietary leucine is used for protein synthesis and the remainder is broken down by a dedicated metabolic pathway¹ (Fig. 1a). A key enzyme complex in the leucine breakdown pathway, 3-methylcrotonyl-CoA carboxylase (3-MCC; encoded by MCCC1 and MCCC2), catalyses the biotin-dependent carboxylation of 3-methylcrotonyl-CoA (3MC-CoA) to form 3-methylglutaconyl-CoA, ultimately leading to the production of acetyl-CoA, an essential metabolite for lipogenesis and energy production and mevalonate, a key intermediate for the production of isoprenoids (Fig. 1a)^2–4. Leucine breakdown and MCCC1 have been shown to play a role in early adipocyte differentiation, and genetic variants of both MCCC1 and MCCC2 are associated with 3-MCC deficiency, characterized by elevated leucine catabolites but showing highly variable clinical presentation, suggesting that additional factors influence the aetiology of 3-MCC deficiency^5–7.

Fig. 1 |. C. elegans mccc-1 is an orthologue of human methylcrotonyl-CoA carboxylase MCCC1.

a, Leucine degradation pathway. The location of mccc-1, the major enzyme studied here, is shown in bold. C. elegans genes are indicated in colour (blue, leucine breakdown; yellow, branched-chain fatty acid synthesis; pink, mevalonate synthesis; human genes are indicated in black). b,c, Quantification of HMB (b) and HMB-carnitine (c) from exo- and endo-metabolome extracts of control and mccc-1(ww4) mutant animals. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± s.d. Statistical significance was determined using a two-sided unpaired Student’s t-test.

In addition to its cellular metabolic function, leucine and its breakdown product, 3,3-hydroxymethylbutyrate (HMB), are frequently used as dietary supplements to increase muscle building and exercise endurance and performance, as well as in elderly patients with sarcopenia or type 2 diabetes^8–10. In humans, under normal physiological conditions, approximately 5% of leucine is metabolized into HMB¹¹. While both metabolites are generally well tolerated, their effects on physiology have not been comprehensively studied¹².

The disruption of metabolic enzyme function often leads to a hard-to-predict rewiring of metabolism, which can be influenced by environmental factors such as diet¹³. Further, the bacteria that inhabit the gut, known as the intestinal microbiota, greatly influence animal physiology, in part via production of specific metabolites that enter host metabolism^14–16. However, the influence of gut microbiota on central metabolism of the host and on the effects of dietary supplements remains poorly understood. The complex interplay between diet, microbiota, metabolic insults and resulting rewiring is difficult to study in humans.

We, and others, have shown that the nematode Caenorhabditis elegans provides a genetically tractable model system with which to study the consequences of metabolic perturbations^17–20. C. elegans metabolism is highly conserved with humans and, because it is a bacterivore, its diet can be modulated to study bacterial influence on cellular processes^21–25. Therefore, C. elegans and its bacterial diet can be used as a model to study metabolic perturbations and how these are modulated by bacterial influences as models for dietary or microbiota effects.

Here, we used C. elegans to study the physiological and molecular effects of altered leucine catabolism. First, we demonstrate that C. elegans mccc-1 is a functional orthologue of human MCCC1, and that mccc-1 mutant animals exhibit extensive metabolic rewiring, as shown via comparative metabolomics and transcriptomics. Like human patients with 3-MCC deficiency, C. elegans mccc-1 mutant animals accumulate the leucine breakdown products HMB and HMB-carnitine. In addition, we discovered multiple shunt metabolites derived from leucine breakdown, including N-acyl amino acid (AA) conjugates and modular glucosides (MOGLs), whose production may prevent the buildup of toxic leucine catabolites. Notably, while HMB and other leucine catabolites are tolerated by the animal when fed a Comamonas aquatica DA1877 (hereafter referred to as Comamonas) diet, these compounds are toxic in animals fed Escherichia coli OP50. We performed an unbiased bacterial genetic screen that revealed that components of Comamonas pyrimidine biosynthesis are required for the protective effect. We further found that mccc-1 mutant animals upregulate the expression of ketone body metabolism genes as an alternate way to provide precursors for mevalonate synthesis. Altogether, our findings yield insights into the multifaceted metabolic rewiring in animals with mccc-1 deficiency.

Results

C. elegans mccc-1 is an orthologue of human MCCC1

Previously, we isolated four alleles of the mccc-1 gene in a forward genetic screen for mutations that prevent the transcriptional repression of the acyl-CoA dehydrogenase acdh-1 when animals are fed Comamonas²⁶ (Extended Data Fig. 1). C. elegans MCCC-1 is 53% identical to human MCCC1 and the two proteins give reciprocal best BlastP values, suggesting that they are orthologues (Extended Data Fig. 1). We focused on the mccc-1(ww4) mutant strain that has a G150R missense mutation within the biotin carboxylase domain. We compared the metabolome of mccc-1(ww4) mutants to control animals using high-performance liquid chromatography (HPLC) coupled to high resolution mass spectrometry (HRMS). We separately analysed metabolites retained in the animal (endo-metabolome) and metabolites excreted into the medium (exo-metabolome) and found that two metabolites that accumulate in the urine and serum of humans with 3-MCC deficiency, HMB and HMB-carnitine, accumulate in both the endo- and exo-metabolome of mccc-1(ww4) mutant animals²⁷ (Fig. 1b,c and Supplementary Table 1). HMB levels were higher in the exo-metabolome relative to the endo-metabolome, indicating that this metabolite is predominantly excreted, whereas HMB-carnitine was more abundant in the endo-metabolome of mccc-1(ww4) mutant animals. Together with the high similarity in protein sequence, these results indicate that mccc-1 is a functional C. elegans orthologue of human MCCC1.

Conjugation reactions redirect leucine catabolic flux

Next, we investigated the biochemical fate of leucine in mccc-1(ww4) mutant animals by stable isotope labelling with ¹³C₆-leucine (Extended Data Fig. 2a). We focused on metabolites that were isotopically labelled and at least fourfold enriched in mccc-1(ww4) mutants relative to control animals (summarized in Supplementary Table 1). These metabolites were further distinguished as ¹³C₆-metabolites that carry the complete carbon skeleton of leucine or as ¹³C₅-metabolites that are generated after decarboxylation by the BCKDH complex in the second step of leucine catabolism (Fig. 2a).

Fig. 2 |. Novel metabolites accumulate in mccc-1(ww4) mutant animals.

a, Schematic of ¹³C₆-leucine tracing. Each orange dot represents a stable ¹³C isotope. b, Quantification of 3-methylcrotonyl (3MC)-carnitine from exo- and endo-metabolome extracts of control and mccc-1(ww4) mutant animals. c, Structure of MOGLs with positions of the conjugated R or X groups indicated. d, Quantification of mecglu#1. e, Extracted ion chromatograms (EICs) for m/z 285.0944, corresponding to mecglu#1 in extracts of control and mcccc-1 animals, or synthetic mecglu#1 co-injected with extract, as indicated. f, Quantification of 2-ketoisocaproate (2KIC) and 2-hydroxyisocaproate (2HIC). g, Quantification of N-acyl 2HIC-AA conjugates. h, EICs for m/z 244.1554, corresponding to 2HIC-Ile and -Leu in control, mccc-1(ww4) mutant animals and synthetic samples, as indicated. i, C. elegans leucine degradation pathway, with shunt metabolites highlighted in red and ¹³C-labelled moieties in MOGL marked in blue. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± s.d. Statistical significance was determined using a two-sided unpaired Student’s t-test.

¹³C₅-metabolites that were enriched in mccc-1(ww4) animals include HMB and HMB-carnitine, as well as 3MC-carnitine, which was not detected in control animals (Figs. 1b,c and 2b). In contrast, tiglyl-carnitine, the isoleucine-derived isomer of leucine-derived 3MC-carnitine, was not enriched and was instead decreased in mccc-1(ww4) mutant animals, demonstrating the specific accumulation of leucine-derived metabolites in the mutant (Extended Data Fig. 2b). Additionally, ¹³C₅-medium-chain acylcarnitine esters, such as 5-methylhexanoyl-l-carnitine (C7ISO) and 7-methyloctanoyl-l-carnitine (C9ISO) were significantly enriched in the exo-metabolome of mccc-1(ww4) mutant animals (Extended Data Fig. 2c). Increased levels of branched-chain fatty acids are likely the result of shunt metabolites derived from isovaleryl-CoA, which can be diverted during fatty acid elongation by conjugation to carnitine and excretion from the cell (Extended Data Fig. 2c and Supplementary Table 1)²⁸. Other ¹³C₅-labelled compounds include a putative HMB-choline conjugate and S-acyl pantetheine conjugates, the latter likely represent decomposed coenzyme A during sample preparation (Extended Data Fig. 2c and Supplementary Table 1)²⁹.

The metabolomic analysis of the exo- and endo-metabolomes of mccc-1(ww4) mutant animals additionally revealed accumulation of a series of previously undescribed metabolites whose MS/MS spectra suggested that they represent MOGLs that incorporate 3MC and/or HMB (Fig. 2c). MOGLs comprise a recently discovered family of metabolites built around a core glucose that can be acylated with diverse moieties representing AA, nucleic acid and neurotransmitter metabolism (Fig. 2c)^30,31. More than 20 different MOGLs were ¹³C₅-labelled following ¹³C₆-Leu supplementation, some of which may incorporate multiple 3MC and/or HMB moieties, based on analysis of MS/MS spectra (Extended Data Fig. 2d and Supplementary Table 1). Given that the putative 3MC- and HMB-derived MOGLs were strongly enriched in mccc-1(ww4) mutant animals (Fig. 2d and Supplementary Table 1), we synthesized an authentic standard of one of the proposed MOGLs, with 3MC attached to the anomeric position of glucose (which we named mecglu#1). The synthetic compound exhibited identical chromatographic retention time and MS/MS fragmentation to the natural compound, confirming our assignment (Fig. 2e and Extended Data Fig. 2e). In contrast to most previously described MOGLs^30,31, mecglu#1 and many of the other 3MC- and HMB-derived MOGLs were more abundant in the exo- than in the endo-metabolome of mccc-1(ww4) mutant animals, suggesting that production of 3MC- and HMB-containing MOGLs may represent a mechanism to reduce accumulation of these leucine catabolites in the animal (Supplementary Table 1).

¹³C₆-labelled metabolites enriched in mccc-1(ww4) mutant animals include the known leucine breakdown products 2-ketoisocaproate (2KIC) and 2-hydroxyisocaproate (2HIC) (Fig. 2f)³². Additionally, we detected a series of previously undescribed ¹³C₆-labelled metabolites highly enriched in the endo-metabolome of mccc-1(ww4) mutant animals, whose MS/MS spectra suggested that they represent AA conjugates of 2HIC (Fig. 2g). We first synthesized standards by conjugating either l-Leu or l-Ile with racemic (R/S)-2HIC to demonstrate that stereoisomers of these compounds can be chromatographically separated (Fig. 2h). Subsequent synthetic conjugation of l-Leu to (S)-2HIC yielded a single stereoisomer with identical chromatographic retention time and MS/MS fragmentation as the naturally occurring metabolite (Extended Data Fig. 2f), indicating that this series of metabolites likely represents AAs conjugated to (S)-2HIC. Additionally, these data indicate that stereospecific reduction of 2KIC to (S)-2HIC in C. elegans proceeds with the same stereospecificity as is observed in mammals³³ (Fig. 2h). Notably, 2HIC was conjugated to Leu, Val, Ile, Phe, Met and Arg, an overlapping subset of hydrophobic AAs previously found to be conjugated to 3-hydroxypropionate (3HP) and lactate^34–36 (Fig. 2g and Supplementary Table 1). Whereas 3HP-AAs were more abundant in the exo-metabolome³⁵, the 2HIC-AAs instead primarily accumulated in the endo-metabolome (Fig. 2g). We did not detect AA conjugates of 2KIC, nor did we detect AA conjugates of 3MC or HMB, which were preferentially incorporated into MOGLs and excreted. Notably, the abundance of several other families of Leu-derived metabolites did not differ between control and mccc-1(ww4) mutant animals. For example, N-acetyl- and N-propionyl-Leu conjugates were not significantly enriched in mccc-1(ww4) mutant animals and abundances of branched-chain fatty ethanolamine derivatives were similar between the two strains as well (Extended Data Fig. 2g and Supplementary Table 1).

Collectively, our metabolomic analyses and characterization of several series of shunt metabolites reveals a complex biochemical network of discrete and specific conjugation reactions that redirect leucine catabolic flux in mccc-1(ww4) mutant animals (Fig. 2i).

Genotype and diet reduce leucine breakdown product toxicity

The detection of multiple leucine-derived shunt metabolites in mccc-1(ww4) mutant animals suggests that the accumulation of leucine breakdown products is toxic. To test this hypothesis, we directly supplemented the leucine breakdown products 2KIC, 2HIC, isovalerate, 3MC or HMB, and examined the effect on the development of control and mccc-1(ww4) mutant animals. We found that HMB, 3MC and isovalerate elicit toxicity in control animals fed Comamonas and that isovalerate and 3MC are much more toxic than HMB (Fig. 3a and Extended Data Fig. 3a). Further, these leucine breakdown products elicited stronger toxic effects in the mutants (Fig. 3b). Finally, we found that HMB was also toxic to the three other mccc-1 mutant strains found in our initial screen but did not observe toxicity in either of two mccc-2 mutant strains (Extended Data Fig. 3b).

Fig. 3 |. Leucine breakdown products are toxic and toxicity is modulated by C. elegans genotype and bacterial diet.

a,b, Toxicity of leucine breakdown products as measured by the proportion of animals reaching the L4 stage in control animals (a) and mccc-1(ww4) mutant animals (b) fed Comamonas. c,d, Toxicity of leucine breakdown products using the same criteria in control animals (c) and mccc-1(ww4) mutant animals (d) fed E. coli OP50. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± s.d. Statistical significance was determined using a one-sided unpaired Student’s t-test.

We previously found that the toxic effects of metabolites or therapeutic drugs in C. elegans can be modulated by bacterial diet^23,37,38. We initially performed the experiments described above in animals fed the Comamonas diet, because that diet was used to identify the mccc-1 mutants²⁶. Remarkably, we found that control animals fed E. coli are much more sensitive to 2KIC, 2HIC and HMB, relative to animals fed Comamonas, whereas 3MC and isovalerate were comparatively well tolerated (Fig. 3c). However, mccc-1(ww4) mutant animals were more sensitive to all five compounds when fed E. coli OP50 (Fig. 3d). Finally, while 2KIC and 2HIC are more toxic to animals fed E. coli relative to those fed Comamonas, their toxicity was not altered in mccc-1(ww4) mutants (Fig. 3c,d).

HMB has been deemed non-toxic in mammals^39–41. However, we observed different levels of HMB-elicited toxicity in C. elegans depending on genotype and bacterial diet. Therefore, we determined the metabolic fate of HMB by supplementing control and mccc-1(ww4) mutant animals fed Comamonas with a non-lethal dose (10 mM) of ¹³C₃-labelled HMB and analysing labelled HMB-derived metabolites. In both control and mccc-1(ww4) mutant animals, we detected a putative MOGL containing ¹³C₃-HMB, as well as mono-methyl-branched-chain fatty acids, derived from isovaleryl-CoA, which were further metabolized to sphinganine and lysophosphatidylcholine (LPC) (Fig. 4a–e and Supplementary Table 1). We did not, however, detect labelled 2KIC or 2HIC. These results imply that supplemented HMB is converted into isovaleryl-CoA via the intermediate 3MC-CoA, that is, that the reactions that generate HMB are reversible (Fig. 4f). As isovalerate and 3MC, which can be generated from these metabolites when they lose the CoA, are more toxic than HMB (Extended Data Fig. 3a), and because HMB cannot be degraded via the 3-MCC-dependent route in mccc-1(ww4) mutants, these results may help to explain HMB toxicity. These results also support our finding that 2KIC and 2HIC are equally toxic to control and mccc-1(ww4) animals (fed either bacterial diet) because these two metabolites are synthesized from a branch of the leucine catabolic pathway that does not directly involve mccc-1. In fact, these metabolites are generated upstream of isovalerate and are separated from the reverse reactions by the irreversible metabolic reaction catalysed by the branched-chain ketoacid dehydrogenase (BCKDH) complex (Fig. 2a).

Fig. 4 |. Metabolic fate of HMB.

a, Schematic of ¹³C₃-HMB tracing. Orange dot indicates labelled ¹³C. b, HPLC–HRMS analysis of the endo-metabolomes of control and mccc-1(ww4) mutant animals supplemented with 10 mM ¹³C₃-HMB showing uptake of HMB by C. elegans. Extracted ion chromatograms EICs for m/z 117.0557 and 120.0655 (negative ion mode), corresponding to HMB and ¹³C₃-HMB. The abundance of naturally occurring HMB in control animals (light blue trace) was low compared with mccc-1(ww4) mutants (yellow trace), whereas both control and mccc-1(ww4) mutant animals exhibited similar levels of ¹³C₃-HMB (dark blue and orange, respectively). c, EICs for m/z 303.1050 and 306.1151, corresponding to sodium adducts of the MOGL HMB-glucoside and ¹³C₃-HMB-glucoside (positive ion mode). Only ¹³C₃-HMB-glucoside was observed in control animals (dark blue trace), whereas naturally occurring HMB-glucoside was also observed and more abundant in mccc-1(ww4) mutant animals (yellow trace). d,e, HPLC–HRMS analysis of control animals supplemented with ¹³C₃-HMB or HMB, as indicated. Animals fed ¹³C₃-HMB exhibited ¹³C₃ isotopic enrichment in iso-branched lipids, such as sphinganine (d) and lysophosphatidylcholine (LPC) bearing a C17 acyl group (e), indicating that exogenous HMB can be reduced to isovaleryl-CoA that can feed back into branched-chain fatty acid biosynthesis. ¹³C₁ and ¹³C₂ isotopes represent natural abundance of ¹³C. Similar levels of ¹³C₃-enrichment in iso-branched lipid derivatives were observed in mccc-1(ww4) mutants supplemented with ¹³C₃-HMB. f, Metabolism of supplemented ¹³C₃-HMB in leucine degradation pathway in mccc-1(ww4) mutant animals is indicated in green.

As MCCC-1 is a mitochondrial enzyme, we wanted to determine whether HMB buildup leads to a disruption in mitochondrial membrane potential. We used tetramethylrhodamine ethyl ester (TMRE) staining to examine the mitochondrial membrane potential of control and mccc-1(ww4) mutant animals. We observed a reduction in mitochondrial membrane potential following exposure to 60 mM HMB in both control and mccc-1(ww4) mutant animals fed Comamonas, as indicated by TMRE staining (Extended Data Fig. 4). This suggests a negative impact of HMB on mitochondrial function⁴².

Taken together, the two leucine breakdown products directly upstream of the reaction catalysed by 3-MCC are specifically toxic in mccc-1(ww4) mutant animals fed either bacterial diet and toxicity of all five metabolites is enhanced on an E. coli OP50 diet. These results suggest that there are different mechanisms of detoxification associated with these metabolites, which correlates with the observation of the different novel shunt metabolites described above.

Comamonas pyrimidine synthesis suppresses HMB toxicity

Next, we focused on the bacterial mechanisms that modulate HMB toxicity. One noteworthy difference between E. coli and Comamonas is that Comamonas produces vitamin B12 but E. coli does not. As vitamin B12 has substantial effects on C. elegans metabolism^18,25,37,43, we asked whether supplementation of E. coli with vitamin B12 or disrupting Comamonas vitamin B12 production would modulate HMB toxicity. However, neither supplementation with vitamin B12 in animals fed E. coli (Fig. 5a) nor disruption of its biosynthesis in Comamonas affected HMB toxicity (Fig. 5b).

Fig. 5 |. Vitamin B12 does not attenuate HMB toxicity.

a, Dose–response curves of animals reaching the L4 stage and fed Comamonas, E. coli OP50 or E. coli OP50 supplemented with increasing concentrations of HMB and with or without supplementation of 64 nM vitamin B12. Data represent five Comamonas, six E. coli OP50 and nine E. coli OP50 with vitamin B12 supplemented biologically independent experiments and error bars indicate mean ± s.d. b, Toxicity of titrated HMB in animals fed E. coli OP50, wild-type Comamonas, or two Comamonas mutants cbiA⁻ and cbiB⁻, which cannot synthesize vitamin B12. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± s.d.

There are several mechanisms by which bacteria can affect the host response to supplemented compounds such as metabolites and therapeutic drugs, including differential uptake by the bacteria, which may modulate drug delivery to C. elegans, direct modification of the compound or by providing bacterial metabolites to the animal that result in modulation of compound toxicity^44,45. To discriminate between these possibilities, we first asked whether active bacterial metabolism is required to either enhance (in E. coli) or mitigate (in Comamonas) HMB toxicity in C. elegans. We found that HMB was more toxic when animals were fed dead bacteria compared with animals fed live bacteria of either species (Fig. 6a–c). This result indicates that active bacterial metabolism is required to mitigate HMB toxicity and that active Comamonas metabolism is most protective.

Fig. 6 |. Live bacteria are required to mitigate HMB toxicity.

a, HMB toxicity on live versus dead (kanamycin-killed) bacteria as measured by the proportion of animals reaching the L4 stage. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± s.d. b, Images of C. elegans grown with HMB supplementation and fed live or kanamycin-killed bacteria. Figures represent one of three biologically independent experiments. Scale bar, 1 mm. c, Percentage of animals reaching beyond the L1 stage (percentage L1+) under the condition of 60 mM HMB supplementation, fed E. coli OP50, wild-type Comamonas or their powders. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± s.d.

To identify Comamonas genes involved in the modulation of HMB toxicity in C. elegans we screened a collection of ~5,700 Comamonas transposon mutant strains³⁷ and found five mutant strains that enhanced HMB toxicity in control animals (Extended Data Fig. 5a and Extended Data Table 1). Two of these mutant strains harbour transposons in genes in the Comamonas pyrimidine biosynthesis pathway (pyrC and pyrE; Fig. 7a). Both bacterial mutant strains exhibited reduced growth compared with wild-type bacteria and, while pyrC⁻ mutant growth could be rescued by supplementation with either orotate or uracil, pyrE⁻ mutants could only be rescued by uracil, which corroborates the predicted biosynthetic functions of these enzymes (Extended Data Fig. 5b). We further found that supplementation with either orotate or uracil rescued HMB toxicity in C. elegans fed pyrC⁻ bacteria and likewise, supplementation with uracil rescued HMB toxicity in C. elegans fed Comamonas pyrE⁻ mutants (Fig. 7b and Extended Data Fig. 5c).

Fig. 7 |. Toxicity of HMB is mitigated by Comamonas pyrimidine metabolism.

a, Bacterial pyrimidine biosynthesis pathway. The two mutants found in the screen are indicated in blue. b, C. elegans grown with and without 60 mM HMB and fed E. coli OP50, wild-type or either of the two Comamonas mutants found in the screen that were supplemented with uracil or orotate during bacterial culture. Outlined areas emphasize the rescue of the pyrC and pyrE mutant E. coli by uracil and orotate. Figures represent one of three biologically independent experiments. Scale bar, 1 mm. c, HMB toxicity exhibited by control or mccc-1(ww4) C. elegans strains with combinations of UTP glucose-1-phosphate uridylyltransferases mutations or knockdown and fed either wild-type or ΔgalU mutant Comamonas. Each bar represents the mean value of four biologically independent experiments, indicated by a white dot, with error bar showing the mean ± s.d. Statistical significance was determined using a one-sided unpaired Student’s t-test. NS, not significant. d, Model of UDP-glucose mediated detoxification. C. elegans and Comamonas genes are indicated in blue. Metabolic processes occurring separately in C. elegans and the bacterium Comamonas are distinguished by green dashed lines.

Notably, uracil supplementation did not alleviate HMB toxicity in animals fed E. coli OP50, even though this bacterial strain is an uracil auxotroph (Fig. 7b and Extended Data Fig. 5c). We tested two additional E. coli strains, BW25113 and HT115 and found that C. elegans fed these strains were less sensitive to HMB than animals reared on OP50 and that toxicity was also not suppressed by uracil supplementation in these strains. Further, we found that feeding E. coli pyrE⁻ mutant bacteria did not impact HMB toxicity in the host, even upon uracil supplementation (Extended Data Fig. 6). Collectively, these data show that Comamonas pyrimidine synthesis is required for the protective effect of these bacteria to HMB toxicity in C. elegans. Further, these data indicate that the modest protective effect of live E. coli is independent of pyrimidine metabolism. Therefore, bacteria can protect C. elegans from metabolite toxicity by distinct mechanisms.

Transcriptional metabolic rewiring in mccc-1(ww4) animals

So far, our metabolomic data indicate that there is extensive metabolic rewiring in mccc-1(ww4) mutant animals, that this rewiring functions to detoxify leucine breakdown products, and that bacterial metabolism aides in this detoxification. As metabolism can be rewired by changing the expression of metabolic enzymes and transporters and because metabolic genes are extensively transcriptionally regulated in C. elegans^46–48, we compared the transcriptomes of mccc-1(ww4) mutant and control animals by RNA-seq. Using a threshold of 1.5-fold change and P < 0.05, we identified 156 up- and 939 downregulated genes in the mutant (Supplementary Table 2). While only 99 (11%) of the downregulated genes encode annotated metabolic enzymes, 50 (32%) upregulated genes are associated with metabolism, suggesting that the transcriptional activation of metabolic genes contributes to the metabolic rewiring seen in the mutant animals (Extended Data Fig. 7a).

Upregulated genes in mccc-1(ww4) mutant animals include acdh-1, which agrees with the screen in which the mutant allele was originally identified²⁶. Among the upregulated genes in mccc-1(ww4) mutant animals are four genes encoding predicted UDP-glycosyltransferases (UGTs), enzymes that attach sugars to other molecules for excretion and detoxification⁴⁹. The C. elegans genome encodes 67 UGT enzymes, most of which are completely uncharacterized⁵⁰. The two most highly induced ugt genes in mccc-1(ww4) mutant animals, ugt-53 and ugt-32, showed ~11-fold and eightfold greater expression relative to control animals, respectively (Extended Data Fig. 7b). We generated single mutant ugt-53(ww60) and ugt-32(ww58) animals by CRISPR/Cas9 genome editing (Extended Data Fig. 7c), crossed these animals with mccc-1(ww4) mutants and tested all resulting single and double mutants for HMB sensitivity. While we did not observe a change in HMB toxicity in the ugt mutants, we found that ugt-32(ww58); mccc-1(ww4) double mutant animals were more sensitive to HMB than mccc-1(ww4) alone (Extended Data Fig. 7d,e). Remarkably, deletion of ugt-53 may slightly decrease HMB toxicity (Extended Data Fig. 7e). The sensitivity to HMB was further increased by feeding ugt-32(ww58); mccc-1(ww4) double mutant animals the Comamonas pyrE⁻ mutant (Extended Data Fig. 7d). These results suggests that C. elegans ugt-32 may be upregulated in mccc-1(ww4) mutant animals to detoxify leucine breakdown products. As HMB is toxic at high doses, we were not able to assess MOGL levels in the mccc-1(ww4);ugt mutant animals. Therefore, we measured the levels of excreted 3MC glucosides in the mutant combinations without HMB supplementation. We did not detect a change in any of the 3MC glucosides in any of the mutant combinations (Extended Data Fig. 7f and Supplementary Table 3). These results suggest that other genes are involved as well, likely because detoxification is multifaceted involving different metabolic conjugation reactions.

Comamonas pyrimidine biosynthesis genes pyrC and pyrE are required to mitigate HMB toxicity in C. elegans and, in mccc-1(ww4) mutant animals, HMB is conjugated to glucose to generate MOGLs that are excreted. As C. elegans ugt-32 is mildly involved in alleviating HMB toxicity, and because these genes encode enzymes that conjugate sugars to toxic molecules, we hypothesized that Comamonas pyrimidine biosynthesis may be used to generate UDP-glucose that is then used by the animal to produce MOGLs. Therefore, we tested whether the production of UDP-glucose is necessary for the protection exhibited by Comamonas. UDP-glucose is synthesized by the enzyme UTP glucose-1-phosphate uridylyltransferase. The Comamonas genome has one such transferase (galU), whereas the C. elegans genome encodes two biochemically verified uridylyltransferases (rml-1 and D1005.2)^50–52. We aimed to generate a Comamonas ΔgalU mutant and a C. elegans D1005.2 mutant by CRISPR/Cas9 genome editing (Extended Data Fig. 7g,h). However, while we succeeded in generating a Comamonas ΔgalU mutant strain, we were not able to establish a C. elegans D1005.2 mutant line because the deletions we generated were homozygous lethal. Therefore, we used RNAi to knockdown D1005.2 expression in subsequent experiments. In addition, we used a deletion mutant of the essential gene rml-1 that likely is a partial loss-of-function allele. We found that HMB toxicity increased when we fed animals Comamonas ΔgalU mutant bacteria and that this effect was stronger in mccc-1(ww4) mutant animals (Fig. 7c). We did not observe a further significant change in HMB toxicity upon either additional deletion of rml-1(ok233) or knockdown of D1005.2 (Fig. 7c). However, when we depleted D1005.2 in the mccc-1(ww4); rml-1(ok233) mutant background and fed animals Comamonas ΔgalU mutant bacteria, we observed a slight increase in HMB toxicity. This increase is only moderate likely because we used a partial loss-of-function allele of rml-1 and because we only have achieved partial knockdown of the essential gene, D1005.2, by performing RNAi in animals fed E. coli HT115 before transferring them to the Comamonas diets.

Taken together, these results suggest that Comamonas pyrimidine metabolism may be important to mitigate HMB toxicity in C. elegans by providing the animal with UDP-glucose, which provides the glucose moiety to synthesize MOGLs (Fig. 7d). However, as the effect of the Comamonas ΔgalU mutant is weaker than that of mutants in bacterial pyrimidine metabolism, the latter pathway must be involved in HMB detoxification by other mechanisms as well.

Leucine breakdown supports C. elegans mevalonate synthesis

Our RNA-seq data also revealed significantly altered expression of ketone body metabolism genes in mccc-1(ww4) mutant animals (Fig. 8a and Supplementary Table 2). In contrast with valine and isoleucine, which are propiogenic, leucine is a ketogenic AA, and its catabolites are intermediates for other biosynthetic pathways, including 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA), which is a precursor of mevalonate synthesis (Fig. 8b). Mevalonate is the key intermediate for the endogenous production of isoprenoids, for example, the co-factor ubiquinone, the geranyl and farnesyl moieties serving as protein membrane anchors or isopentenyladenosine, an essential component of tRNA^3,4. HMG-CoA can be derived from leucine but can also be synthesized de novo from condensation of acetoacetyl-CoA with acetyl-CoA by HMG-CoA synthase (HMGS-1), followed by reduction to mevalonate by HMG-CoA reductase (HMGR-1) (Fig. 8b). As the production of HMG-CoA from leucine is impaired in mccc-1(ww4) mutants, we hypothesized that upregulation of ketone body metabolism genes may support mevalonate synthesis. Consistent with this idea, we found that both sur-5 and C05C10.3, which are predicted to produce acetoacetyl-CoA, are more highly expressed in mccc-1(ww4) mutant animals (Fig. 8a). Therefore, we reasoned that, if mevalonate synthesis relied on both leucine catabolism, which is dependent on MCCC-1 activity, as well as ketone body metabolism, simultaneous depletion of both pathways should generate a stronger phenotype than depletion of either pathway alone (Fig. 8c). Previous reports have shown that animals carrying deletions in hmgs-1 or hmgr-1 are not viable, while knockdown of hmgs-1 reduces adult viability and depletion of hmgr-1 does not show a detectable phenotype^53–55. Viability of the hmgr-1 deletion mutant animals can be rescued with low concentrations of mevalonate; however, these animals show reduced reproductive potential. When we knocked down hmgs-1 we also observed reduced adult viability, and this was enhanced in mccc-1(ww4) mutant animals (Fig. 8d). Moreover, the reproductive potential of hmgr-1 knockdown animals was reduced in mccc-1(ww4) mutant animals (Fig. 8e). Further, supplementation with mevalonate, but not the ketone body 3-hydroxybutyrate rescued all of the phenotypes (Fig. 8b,d,e). The increased sensitivity to hmgr-1 RNAi was consistently observed across mccc-1 mutant alleles, but not with mccc-2 mutant alleles (Extended Data Fig. 8). Taken together, C. elegans leucine catabolism is important for mevalonate synthesis and its perturbation is compensated by an increase in ketone body metabolism genes.

Fig. 8 |. Mevalonate biosynthesis is rewired via ketone bodies when leucine breakdown is impaired.

a, Fold change in mRNA levels of differentially expressed C. elegans ketone body metabolism genes in mccc-1(ww4) mutant animals measured by RNA-seq. b, Cartoon depicting leucine breakdown dependent and independent HMG-CoA synthesis facilitating the generation of mevalonate. In the absence of leucine breakdown (mccc-1(ww4) animals), HMG-CoA synthesis depends on precursors derived from ketone body production such as acetoacetate. The dashed line shows that acetyl-CoA generated by HMG-CoA breakdown can be recycled back to form HMG-CoA but both acetyl-CoA and acetoacetyl-CoA can additionally be generated by other cellular pathways, such as tyrosine metabolism. Enzymes active in the absence of MCCC-1 activity are shown in red. Enzymes with elevated expression in mccc-1(ww4) animals are bold. c, Cartoon of changes in C. elegans mevalonate levels between control and mccc-1(ww4) mutant animals (grey) in the presence (black) or absence (red) of hmgs-1 or hmgr-1. d, Bar graphs showing viability of P0 animals exposed to mevalonate biosynthesis gene RNAi and supplemented with 50 mM or 20 mM for 3-hydroxybutyrate (3HB) or mevalonate (Mev), respectively. Outlined areas emphasize the reduced reproductive potential of hmgr-1;mccc-1(ww4) mutant animals. NS, not significant. Each bar represents the mean value of five biologically independent experiments, indicated by a white dot, with error bar showing the mean ± s.d. Statistical significance was determined using a one-sided unpaired Student’s t-test. e, Offspring of P0 animals exposed to mevalonate biosynthesis gene RNAi from P0 animals and supplemented with 50 mM or 20 mM for 3HB or Mev, respectively. Figures represent one of five biologically independent experiments. Scale bar, 1 mm. f, Model of rewired leucine degradation pathway in mccc-1(ww4) mutant animals and a proposed metabolic connection to Comamonas pyrimidine metabolism. Orange arrows indicate rewired metabolic processes in mccc-1(ww4) mutant animals. C. elegans genes are shown in blue.

Discussion

Combining metabolomics, transcriptomics and de novo structure elucidation revealed extensive metabolic rewiring in a C. elegans mutant with perturbed leucine catabolism. This rewiring serves two purposes: to detoxify toxic leucine breakdown intermediates and to support mevalonate biosynthesis by an alternate metabolic route (Fig. 8f).

Our data indicate that the detoxification of leucine breakdown products is multifaceted and relies on several distinct biochemical transformations, resulting in the production of diverse, previously undescribed shunt metabolites, including N-acylated AAs, 2-hydroxyisocapoate AA conjugates, methyl-branched acylcarnitines and MOGLs. Previous studies in humans found that AA conjugations to lactate or fatty acids are catalysed by cytosolic non-specific dipeptidase 2 (CNDP2) or peptidase M20 domain containing 1 enzyme (PM20D1) activities^34,36,56. We therefore speculate that 2HIC-AAs production may involve similar peptidases. However, we did not find any change in expression in putative C. elegans peptidases in our mccc-1(ww4) mutant RNA-seq analysis. Of note, 2HIC toxicity is also modulated by bacterial diet with greater toxicity observed when supplemented animals are grown on E. coli OP50 versus Comamonas, similar to HMB, indicating that metabolic factors produced by the bacteria could also aide in mitigating 2HIC toxicity in the animal. We found that HMB can undergo reverse conversion reactions that lead to the production of potentially toxic intermediates such as 3MC. Future studies will be required to identify the enzymes catalysing these reactions.

Notably, we found that many leucine catabolites, including HMB, are toxic to C. elegans and that this toxicity is modulated by bacterial diet. As active bacterial metabolism is required to mitigate HMB toxicity, it is unlikely that the bacteria modulate delivery to the animal by differential HMB uptake. Instead, our findings suggest that the production of bacterial metabolites such as UDP-glucose may help to mitigate HMB toxicity in the animal. However, because mutants in pyrimidine biosynthesis affect HMB toxicity more severely than mutants required for the bacterial synthesis of UDP-glucose, bacterial pyrimidine metabolism must be involved in additional detoxification mechanisms. Our data further suggest that within the animal, C. elegans UGT enzymes function in the detoxification of leucine breakdown products. We speculate that UGT enzymatic activity would add glucose, potentially from the UDP-glucose generated in bacteria, to toxic leucine catabolites such as HMB, thereby generating MOGLs that are excreted from the animal. However, this model is currently supported by weak genetic interactions, potentially due to the large UGT family in C. elegans, several of which are upregulated in mccc-1 mutant animals and may have compensatory function. Therefore, further work is needed to determine whether bacterial pyrimidine biosynthesis or UDP-glucose and C. elegans UGT enzymes directly modify HMB through the proposed mechanism. It is also possible that Comamonas directly metabolizes HMB in supplementation experiments. However, if so, it is not likely to be a contributing factor to the detoxification of HMB produced by C. elegans because it would have to be significantly taken up from the animal by living bacteria in the gut, which is perhaps not likely to occur. Finally, because vitamin B12 does not affect HMB detoxification, it is clear that Comamonas affects C. elegans metabolism and physiology by both vitamin B12-dependent and independent mechanisms, consistent with our previous findings of drug toxicity^23,38.

Another finding is that perturbation of C. elegans leucine catabolism rewires mevalonate biosynthesis. Because mccc-1 is required for HMG-CoA synthesis from leucine, our data suggest that mevalonate biosynthesis in mccc-1(ww4) mutants is bolstered by increased transcription of ketone body genes providing input to HMGS-1, an alternative route to produce HMG-CoA. While it is well known that leucine catabolism produces HMG-CoA, it has also been proposed that HMB can be directly converted to HMG-CoA via an unknown enzyme⁵⁷. If that were true, mccc-1 loss-of-function would not be expected to affect mevalonate metabolism. However, our data indicate that mevalonate biosynthesis in mccc-1 mutant animals is compensated by an alternate route using ketone bodies, even though these animals produce ample HMB.

Taken together, our work shows that defective leucine breakdown results in activation of a previously uncharted biochemical network of conjugation reactions and the transcriptional compensation for biosynthetic processes. Future studies will determine whether these processes are conserved in mammals, including humans or whether diverse organisms have evolved different solutions to handle the toxic effect of metabolic intermediates such as HMB.

Methods

C. elegans strains

C. elegans strains were cultured using standard procedures at 20 °C (ref. 58). N2 (Bristol) was used as the wild-type strain. The VL749 wwIs24(Pacdh-1::GFP + unc-119(+)) strain is the background control strains for the mccc-1(ww2), mccc-1(ww4), mccc-1(ww20) and mccc-1(ww38) alleles (strain VL907, VL1080, VL1520 and VL1521, respectively) and other strains generated herein^17,26. The strains FX22463 mccc-2(tm12463) and FX25273 mccc-2(tm15274) were obtained from National BioResource Project, Japan. FX22463 and FX25273 were backcrossed three times with N2 and crossed to VL749 to generate VL1528 mccc-2(tm12463); wwIs24(Pacdh-1::GFP + unc-119(+)) and VL1529 mccc-2(tm15274); wwIs24(Pacdh-1::GFP + unc-119(+)), respectively. The strains VL1369 ugt-32(ww58), VL1384 ugt-53(ww60), VL1524 D1005.2(ww65)/+ heterozygote and VL1525 D1005.2(ww66)/+ heterozygote were generated by CRISPR-Cas9 genome editing and backcrossed three times with N2 (refs. 59,60). The VL1369 and VL1384 strains were crossed to VL749 to generate VL1410 ugt-32(ww58); wwIs24(Pacdh-1::GFP + unc-119(+)) and VL1411 ugt-53(ww60); wwIs24(Pacdh-1::GFP + unc-119(+)), respectively. VL1410 and VL1411 were further crossed to mccc-1(ww4) to create strains VL1412 mccc-1(ww4); ugt-32(ww58); wwIs24(Pacdh-1::GFP + unc-119(+)) and VL1413 mccc-1(ww4); ugt-53(ww60); wwIs24(Pacdh-1::GFP + unc-119(+)), respectively. The strains MG278 rml-1(ok233) was retrieved from the C. elegans Gene Knockout Consortium. MG278 rml-1(ok233) was backcrossed threes time with N2 and crossed to VL749 or VL1080 to generate VL1451 rml-1(ok233); wwIs24(Pacdh-1::GFP + unc-119(+)) or VL1452 rml-1(ok233); mccc-1(ww4); wwIs24(Pacdh-1::GFP + unc-119(+)), respectively.

Bacterial strains

E. coli OP50, E. coli HT115(DE3), E. coli BW25113 and Comamonas aquatica DA1877 (referred to as Comamonas) were obtained from the Caenorhabditis Genetics Center. The E. coli pyrE⁻ mutant was retrieved from the E. coli Keio collection⁶¹. The Comamonas cbiA⁻, cbiB⁻, valS⁻, recB⁻, pyrC⁻ and pyrE⁻ mutants were retrieved from the Comamonas mutant collection³⁷. The Comamonas ΔgalU mutant was generated using a modified CRISPR-Cas9 genome editing system (see below). Bacterial cultures were prepared in lysogeny broth (LB) from a single-colony inoculation and incubated at 37 °C for 24 h with 200 rpm orbital shaking.

Sequence alignment of C. elegans MCCC-1 to human MCCC1

Protein sequences of human MCCC1 (A0A0S2Z693) and C. elegans MCCC-1 (O45430) were obtained from Uniprot (www.uniprot.org). Sequences were aligned using the online tool Benchling (www.benchling.com).

Culturing C. elegans for metabolomics

First larval stage (L1) animals were cultured on nematode growth medium (NGM) agar seeded with Comamonas approximately 70 h at 20 °C until they reached the gravid adult stage. Eggs were obtained by subjecting gravid adults to buffered bleached solution (20% NaOCl, Fisher, SS290–1), 10% 10 N NaOH and 70% H₂O). Eggs were hatched and synchronized in M9 buffer (22 mM KH₂PO₄, 42 mM Na₂HPO₄, 86 mM NaCl, 1 mM MgSO₄ and 1 L H₂O). Approximately 50,000 L1 animals were placed in a 25-ml Erlenmeyer flask filled with 10 ml S medium (1 l S basal (5.85 g NaCl, 1 g K₂ HPO₄, 6 g KH₂PO₄, 1 ml cholesterol (5 mg ml⁻¹ in ethanol) in 1 l H₂O), 10 ml 1 M potassium citrate, 10 ml trace metal solution, 3 ml 1 M CaCl₂ and 3 ml 1 M MgSO₄) with a concentrated Comamonas bacterial pellet from a 300-ml overnight LB culture. Animals were incubated at 20 °C with 180 rpm orbital shaking for 60 h for checking the developmental stage and/or adding 10 mM pH 6.0-adjusted leucine or HMB. Animals were further incubated for 12 h until they reached young adult stage. Animals and conditioned medium were separated by 264g centrifugation for 2 min and extracted separately. For analysis of animal bodies (endo-metabolome), animal pellets were washed three times with M9 buffer followed by one wash with H₂O. Conditioned medium (exo-metabolome) was centrifuged at 10,000g for 10 min to remove residual Comamonas bacteria. Approximately 7 ml of clarified exo-metabolome was transferred to a new conical tube. The endo- and exo-metabolome samples were snap frozen in liquid nitrogen and stored at −80 °C.

Sample preparation for HPLC–HRMS

Animal bodies (endo-metabolome) and conditioned medium (exo-metabolome) were frozen and processed separately, as described above. For preparation of endo-metabolome extracts, samples were lyophilized for 18–24 h using a VirTis BenchTop 4K Freeze Dryer. After the addition of 1 ml methanol directly to the conical tube in which animals were frozen, samples were sonicated for 5 min (2 s on–off pulse cycle at 90 A) using a Qsonica Q700 Ultrasonic Processor with a water bath cup horn adaptor (Qsonica 431C2). Following sonication, an additional 4–9 ml of methanol was added, depending on sample size, and the extract rocked overnight at room temperature (22 °C). The conical tubes were centrifuged (3,000g, 22 °C, 5 min) and the resulting clarified supernatant transferred to a clean 8- or 20-ml glass vial which was concentrated to dryness in an SC250EXP Speedvac Concentrator coupled to an RVT5105 Refrigerated Vapor Trap (Thermo Scientific). The resulting powder was suspended in 100–250 μl of methanol, depending on sample size, followed by vigorous vortex and brief sonication. This solution was transferred to a clean microfuge tube and subjected to centrifugation (20,000g, 22 °C, 5 min) in an Eppendorf 5417 R centrifuge to remove precipitate. The resulting supernatant was transferred to an HPLC vial and analysed by HPLC–HRMS.

For preparation of exo-metabolome extracts, samples were lyophilized for ~48 h using a VirTis BenchTop 4K Freeze Dryer. Dried material was extracted in 5–15 ml methanol, depending on the sample size and rocked overnight at room temperature. The conical tubes were centrifuged (3,000g, 22 °C, 5 min) and the resulting clarified supernatant transferred to clean 8- or 20-ml glass vials which were concentrated in vacuo and suspended in methanol as described for endo-metabolome samples.

HPLC–HRMS analysis

Reversed-phase chromatography was performed using a Vanquish HPLC system controlled by Chromeleon Software (Thermo Fisher Scientific) and coupled to an Orbitrap Q-Exactive HF mass spectrometer controlled by Xcalibur software (Thermo Fisher Scientific) or by a Dionex Ultimate 3000 HPLC system coupled to an Oribtrap Q-Exactive mass spectrometer controlled by the same software. Extracts prepared as described above were separated on a Thermo Scientific Hypersil Gold column (150 × 2.1 mm, particle size 1.9 μm, part no. 25002–152130) maintained at 40 °C with a flow rate of 0.5 ml min⁻¹. Solvent A: 0.1% formic acid (Fisher Chemical Optima LC–MS grade; A11750) in water (Fisher Chemical Optima LC–MS grade; W6–4); solvent B: 0.1% formic acid in acetonitrile (Fisher Chemical Optima LC–MS grade; A955–4). A/B gradient started at 1% B for 3 min after injection and increased linearly to 98% B at 20 min, followed by 5 min at 98% B, then back to 1% B over 0.1 min and finally held at 1% B for an additional 2.9 min.

The mass spectrometer parameters were spray voltage of −3.0 kV/+3.5 kV; capillary temperature of 380 °C; probe heater temperature of 400 °C; sheath, auxiliary and sweep gas at 60, 20 and 2 a.u., respectively; S-Lens RF level of 50; resolution of 60,000 or 120,000 at m/z 200; and AGC target of 3E6. Each sample was analysed in negative (ESI−) and positive (ESI+) electrospray ionization modes with m/z range 117–1,000. Parameters for MS/MS (dd-MS2): MS1 resolution of 60,000; AGC target of 1E6. MS2 resolution, 30,000; and AGC target of 2E5. The maximum injection time was 60 ms; isolation window was 1.0 m/z; stepped normalized collision energy 10, 30; dynamic exclusion was 1.5 s; and the top eight masses were selected for MS/MS per scan.

HPLC–HRMS RAW data were converted into mzXML file format using MSConvert (v.3.0, ProteoWizard) and were analysed using Metaboseek software v.0.9.9 and normalized to the abundance of ascr#3 (an ascaroside featuring a nine-carbon α,β-unsaturated carboxylic acid side chain) in negative ionization mode or normalized to the abundance of ascr#2 (an ascaroside featuring a six-carbon side chain bearing a ketone functionality) in positive ionization mode as an approximate measure of sample size for replicates from the same genotype. To account for variation between genotypes, metabolites were normalized as a ratio to ascr#3 or ascr#2 and the resulting quotient multiplied by the genotype average (performed independently for endo- and exo-metabolome samples), thereby removing the effect of variation between genotypes. Quantification was performed with Metaboseek software v.0.9.9 (Metaboseek.com) or via integration using Xcalibur QualBrowser v.4.1.31.9 (Thermo Fisher Scientific) using a 5-ppm window around the m/z of interest. Statistical analysis for metabolomics was performed with Metaboseek software v.0.9.9 and with GraphPad Prism v.9.5.0.

General synthetic procedures

Unless stated otherwise, all reactions were carried out under argon (Ar) atmosphere in flame-dried glassware. All commercially available reagents were used as purchased unless otherwise stated. All solvents were dried over activated 3 Å molecular sieves for a minimum of 24 h unless used in reactions containing aqueous reagents. Solutions and solvents sensitive to moisture and oxygen were transferred via standard syringe and cannula techniques. Reactions were cooled with iced water or dry ice–acetone baths or heated with mineral oil baths depending on reaction temperature. Unless stated otherwise, all chemicals and reagents were purchased from Sigma-Aldrich. l-isoleucine tert-butyl ester hydrochloride, l-leucine tert-butyl ester hydrochloride and trifluoroacetic acid (TFA) were purchased from TCI. The 4-dimethylaminopyridine (DMAP) was purchased from Fluka. Dichloromethane (DCM), ethyl acetate, hexanes and methanol were purchased from Fisher Scientific. Thin-layer chromatography was performed using J. T. Baker Silica Gel IB2F plates. Flash chromatography was performed using Teledyne Isco CombiFlash systems and Teledyne Isco RediSep Rf silica columns. All deuterated solvents were purchased from Cambridge Isotopes. Nuclear Magnetic Resonance (NMR) spectra were recorded on a Bruker AV 500 (500 MHz) spectrometer at Cornell University’s NMR facility. ¹H NMR chemical shifts are reported in ppm (δ) relative to residual solvent peaks (7.26 ppm for CDCl₃ and 3.31 ppm for CD₃OD). ¹H NMR chemical shifts are reported as follows: chemical shift, multiplicity (s, singlet; d, doublet; t, triplet; m, multiplet), coupling constants (Hz) and integration. ¹³C NMR chemical shifts are reported in ppm (δ) relative to residual solvent peaks (77.16 ppm for CDCl₃ and 49.00 ppm for CD₃OD). All NMR data processing was conducted using Mnova v.14.2.3 (https://mestrelab.com/).

Chemical syntheses

(2S,3R,4S,5S,6R)-3,4,5-Trihydroxy-6-(hydroxymethyl) tetrahydro-2H-pyran-2-yl 3-methylbut-2-enoate (1).

Based on a previously reported procedure⁶², we added N,N-dimethylformamide (5 ml) to a flask of α-d-glucose (551 mg, 3.06 mmol, 1.0 eq.), 3,3-dimethylacrylic acid (398.0 mg, 3.98 mmol, 1.3 eq.) and triphenylphosphine (1.2 g, 4.59 mmol, 1.5 eq.) under argon and cooled to 0 °C, before adding diisopropyl azodicarboxylate (0.9 ml, 4.59 mmol, 1.5 eq.) to the flask. The reaction mixture was stirred at 0 °C for 1 h and warmed up to room temperature. After 17 h, the mixture was concentrated in vacuo. Flash-column chromatography on silica using a gradient of 0–50% methanol in DCM afforded mecglu#1 (1) as a colourless oil (86.0 mg, 11%).

¹H NMR (500 MHz, methanol-d₄).

δ (ppm) 5.73 (m, 1H), 5.45 (d, J = 8.1 Hz, 1H), 3.80 (dd, J = 1.9, 12.0 Hz, 1H), 3.68–3.61 (m, 2H), 3.39 (dt, J = 8.7 Hz, 1H), 3.36–3.32 (m, 2H), 2.15 (d, J = 0.9 Hz, 3H), 1.90 (d, J = 1.0 Hz, 3H).

¹³C NMR (125 MHz, methanol-d₄).

δ (ppm) 166.4, 161.0, 116.2, 95.1, 78.7, 78.0, 73.9, 71.1, 62.3, 27.5, 20.5.

N-(2-Hydroxyisocaproyl)-l-isoleucine tert-butyl ester (2).

We added 4 ml dichloromethane to a vial containing 2-hydroxyisocaproic acid (45 mg, 0.34 mmol, 1.0 equiv.), l-isoleucine tert-butyl ester hydrochloride (98 mg, 0.44 mmol, 1.3 equiv.), 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) hydrochloride (224 mg, 1.17 mmol, 3.4 equiv.) and DMAP (83 mg, 0.68 mmol, 2.0 equiv.). The resulting solution was stirred at room temperature overnight and concentrated in vacuo. Flash-column chromatography on silica using a gradient of 0–100% ethyl acetate in n-hexane afforded 2 (86 mg, 85%).

¹H NMR (500 MHz, chloroform-d).

δ 4.51 (ddd, J = 8.8, 4.6, 1.9 Hz, 1H), 4.19 (ddd, J = 9.7, 3.3, 2.3 Hz, 1H), 1.96–1.84 (m, 2H), 1.68–1.53 (m, 2H), 1.49 (s, 9H), 1.30–1.17 (m, 2H), 0.99–0.93 (m, 12H).

¹³C NMR (126 MHz, chloroform-d).

δ 174.2, 174.0, 171.1, 170.9, 82.2, 82.2, 70.9, 70.6, 56.5, 56.4, 44.0, 43.9, 38.2, 38.2, 28.1, 25.3, 25.2, 24.6, 24.6, 23.5, 23.5, 21.4, 15.4, 15.4, 11.7.

N-(2-Hydroxyisocaproyl)-l-isoleucine (3).

We added trifluoroacetic acid (TFA) (0.52 ml, 6.8 mmol, 100 equiv.) to a solution of 2 (20 mg, 0.068 mmol, 1 equiv.) in 1 ml of DCM. The resulting solution was stirred for 1 h and then concentrated to dryness in vacuo, yielding 2HIC-l-Ile (3, 21 mg, 127%).

¹H NMR (500 MHz, methanol-d₄).

δ 4.43 (t, J = 5.6 Hz, 1H), 4.10 (td, J = 9.8, 3.7 Hz, 1H), 2.00–1.92 (m, 1H), 1.92–1.83 (m, 1H), 1.61–1.47 (m, 3H), 1.31–1.21 (m, 1H), 0.99–0.95 (m, 12H).

¹³C NMR (126 MHz, methanol-d₄).

δ 176.2, 176.1, 173.0, 173.0, 70.0, 69.9, 56.0, 55.9, 43.5, 43.2, 37.4, 37.3, 24.8, 24.7, 24.2, 24.1, 22.5, 22.5, 20.4, 20.3, 14.6, 14.6, 10.5.

N-((S)-2-hydroxyisocaproyl)-l-leucine tert-butyl ester (4).

We added 4 ml dichloromethane to a vial containing (S)-2-hydroxyisocaproic acid (46 mg, 0.35 mmol, 1.0 equiv.), l-leucine tert-butyl ester hydrochloride (106 mg, 0.47 mmol, 1.3 equiv.), EDC hydrochloride (229 mg, 1.20 mmol, 3.4 equiv.) and DMAP (89 mg, 0.73 mmol, 2.0 equiv.). The resulting solution was stirred at room temperature overnight and concentrated in vacuo. Flash-column chromatography on silica using a gradient of 0–100% ethyl acetate in n-hexane afforded 4 (72 mg, 68%).

¹H NMR (500 MHz, methanol-d₄).

δ 4.39 (t, J = 7.3 Hz, 1H), 4.08 (dd, J = 9.6, 3.7 Hz, 1H), 1.92–1.83 (m, 1H), 1.73–1.66 (m, 1H), 1.65–1.62 (m, 2H), 1.59–1.50 (m, 2H), 1.49 (s, 9H), 0.99–0.94 (m, 12H).

¹³C NMR (126 MHz, methanol-d₄).

δ 176.2, 171.9, 81.5, 69.9, 51.0, 43.5, 40.6, 26.8, 24.7, 24.1, 22.5, 21.8, 20.6, 20.4.

N-((S)-2-Hydroxyisocaproyl)-l-leucine (5).

We added TFA (0.53 ml, 6.9 mmol, 100 equiv.) to a solution of 4 (21 mg, 0.069 mmol, 1 equiv.) in 1 ml of dichloromethane. The resulting solution was stirred for 1 h and then concentrated to dryness in vacuo. Flash-column chromatography on silica using a gradient of 0–100% methanol in DCM afforded (S)-2HIC-L-Leu (5, 19 mg, 113%).

¹H NMR (500 MHz, methanol-d₄).

δ 4.39 (t, J = 7.1 Hz, 1H), 3.98 (dd, J = 9.6, 3.7 Hz, 1H), 1.80–1.82 (m, 1H), 1.60–1.56 (m, 3H), 1.48–1.35 (m, 2H), 0.88–0.82 (m, 12H).

¹³C NMR (126 MHz, methanol-d₄).

δ 176.3, 174.4, 70.0, 50.1, 43.4, 40.6, 24.7, 24.1, 22.5, 22.0, 20.5, 20.4.

Developmental rate assay

For the developmental rate assay, overnight bacterial cultures were concentrated by resuspending the bacteria in H₂O after centrifuge at 2,500g for 30 min. E. coli was concentrated fivefold, Comamonas, cbiA⁻, and cbiB⁻ were concentrated tenfold, and Comamonas pyrC⁻ and pyrE⁻ mutants were concentrated 20-fold. Then, 0.2 ml concentrated bacteria was seeded and dried onto 35-mm NGM plates.

Animals were fed a diet of E. coli or Comamonas for three generations before assay. Approximately 100 L1 animals were added to their respective bacteria-seeded 35-mm NGM plates containing various concentrations of pH 6.0-adjusted HMB, 3-methylcrotonate, or isovalerate. L1 animals were incubated for 70 h at 20 °C and post-L4 stage animals were counted. The percentage of post-L4 developed animals was calculated compared with the non-supplement condition of same bacterial diet. For the vitamin B12 supplement condition, we used 64 nM adenosyl cobalamin (Sigma-Aldrich, C0884). We obtained dose–response curves using the following equation that was fit to the biological replicate datasets:

$$y=\text{Bottom}+(\text{Top}-\text{Bottom})/{(1+10}^{\left(\right(\text{logLD}50-X)\times \text{HillSlope}))}$$

For starved and kanamycin treated dead bacterial diet, L1 animals were incubated for 70 h at 20 °C and post-L4 stage animals were counted to obtain the percentage animals that developed beyond this stage.

TMRE staining

L1 animals were cultured on NGM agar seeded with Comamonas approximately 30 h at 20 °C until they reached the third larval (L3) stage. Animals were treated with 10 μM TMRE in S Basal buffer (44 mM KH₂PO₄, 5 mM K₂HPO₄ and 0.1 M NaCl in H₂O) and incubated at room temperature with gentle rocking for 45 min, covered in foil. Animal pellets were washed two times by 161g centrifugation for 1 min in S Basal buffer. After washing, animals were placed in NGM plates without TMRE for 30 min. Animals were mounted on a 2% agar pad to assess mitochondrial morphology and measure TMRE intensity of the intestinal mitochondria.

Bacterial killing

E. coli OP50 and Comamonas were cultured in 50 ml LB medium in a 250-ml flask at 37 °C for 24 h with 200 rpm shaking. The bacterial culture was transferred to 50-ml tube. Transferred bacteria was pelleted by 3,000g centrifugation for 30 min. Supernatant was discarded and bacterial pellet was resuspended in 100 ml H₂O containing 250 μg ml⁻¹ kanamycin. Resuspended bacteria were evenly separated in two 250-ml flasks and incubated for 24 h with 200 rpm shaking. For checking bacterial survival, 1 ml of 100 ml culture was aliquoted and washed three times by 1 ml H₂O. Washed bacterial culture was plated on LB agar and incubated at 37 °C for 24 h. When no viable colonies were detected, the killed bacteria were stored at 4 °C. Before seeding killed bacteria on NGM agar, they were washed three times with 100 ml H₂O.

Bacterial powder from E. coli OP50 or Comamonas was produced as described previously³⁸. In brief, a single bacterial colony was cultured overnight and subsequently diluted 1:100 in LB medium without antibiotics to reach the log phase (OD_600nm = 0.8 to 1.0). The bacterial cultures were collected and washed three times with sterile water. Each pellet was weighed and resuspended in sterile water at a concentration of 1 gram of wet weight per 25 ml. The bacteria were then disrupted using a Microfluidics M-110P lab homogenizer set to 22,000 psi for ten cycles. The disrupted cells were flash-frozen and lyophilized using a Labconco FreeZone 2.5-l benchtop freeze dryer/lyophilizer until completely dry. The dried bacterial powder was then dissolved in sterile water at a concentration of 50 mg ml⁻¹. The prepared powder solution was plated on LB agar to ensure no bacterial growth.

Microscopy

For TMRE stained mitochondria, 3–4 animals in the third larval (L3) stage were mounted onto a single slide and imaged together within 5 min using a ZEISS LSM800 confocal microscope at ×63 magnification to minimize time spent for each slide as the prolonged scoping time can lead to mitochondrial fission and loss of the network.

For Nomarksi pictures, synchronized L1 animals were grown in NGM agar plates with each metabolite supplemented for 72 h for measuring the number of post-L4 animals at P0 and 144 h for observing F1 embryonic lethality and larval development at 20 °C. Brightfield images of each plate were taken using a Nikon Eclipse Ci (microscope) and Nikon DS-Ri2 (Camera) at ×20 magnification.

Bacterial mutant screens

Primary screens were performed in 96-well plates containing modified NGM agar (substitute K₂HPO₄ buffer (pH 6.0) to PBS (Gibco, 10010023) containing 60 mM HMB and 10 μg ml⁻¹ gentamycin). Bacterial mutants were cultured from frozen glycerol stocks in 1 ml LB medium with gentamycin and incubated at 37 °C for 24 h with 200 rpm shaking. Then, 50 μl of bacterial mutant overnight culture was seeded in each well of a 96-well plate and dried in an aseptic hood. Approximately 20 L1 animals were seeded in each well and incubated for 70 h at 20 °C. Comamonas mutants that did not support C. elegans development in the presence of HMB were considered hits. The mutant collection was screened twice.

All hits were retested using approximately 50 L1 animals grown on 35 mm NGM agar plates plus the corresponding antibiotics and containing 60 mM HMB. Comamonas hits that retested were sequenced to identify the location of the transposon insertion as described previously³⁷.

For C. elegans developmental rate assays (see above) with the Comamonas mutants, bacteria were cultured without supplement, with 10 mM NaOH (solvent control), 1 mM uracil or 1 mM orotic acid, and incubated at 37 °C for 24 h with 200 rpm shaking. Cultured bacteria were centrifuged for 30 min at 2,054g. Supernatant was decanted and bacterial cells were resuspended in H₂O. Larval development was captured after 20 °C incubation for 70 h on 35-mm NGM agar plates containing 0 or 60 mM HMB by using Nikon DS-Ri camera in Nikon Eclipse Binocular Ergonomic microscope at ×20 magnification.

Bacterial growth rate

Bacterial growth rate was measured in flat-bottom 96-well plates. For different nucleobase supplementation, 5 μl diluted bacterial cultures (OD_600nm = 0.4) were inoculated to 0.2 ml LB medium containing 1 mM of uracil, adenine, thymine, guanine, cytosine or orotate dissolved in 10 mM NaOH. Plates were incubated at 37 °C for 24 h with 200 rpm orbital shaking. Bacterial growth was monitored every 30 min by measuring OD_600nm in an EON microplate spectrophotometer.

RNA extraction, RNA-seq and data analysis

L1 animals were cultured on NGM agar seeded with Comamonas approximately 70 h at 20 °C until they reached the gravid adult stage. Gravid adults were bleached and synchronized eggs were obtained as described above. Approximately 500 L1 animals were incubated at 20 °C until they reached at young adult stage. Then, 250 young adults were picked and transferred to 0.5 ml TRIzol (Thermo Fisher, 15596018). Animals were immediately frozen in liquid nitrogen and stored at −80 °C for RNA extraction. Animals in TRIzol were freeze-cracked in liquid nitrogen and warmed to 37 °C three times. RNA was separated from TRIzol by adding 50 μl of 1-bromo-3-chloropropane (MRC, BP 151) and purified by following manufacturer’s protocol of Directzol RNA miniprep kit (Zymo research, R2050). Extracted RNA was sent to BGI genomics for sequencing, read mapping and analysis. Differentially expressed genes were identified with thresholds of 1.5-fold change with PPEE < 0.05 of statistical significance, where PPEE indicates the post probability of equally expressed⁶³ genes. In addition, metabolic differentially expressed genes were annotated and categorized according to the annotations used in the iCEL1314 genome-scale metabolic network model^21,48,64 and pathway enrichment analysis using WormPaths was used to detect significantly induced or repressed metabolic pathways⁶⁵.

CRISPR/Cas9 genome editing

To delete ugt-32 and ugt-53 in C. elegans, we modified a previously described CRISPR-Cas9 genome editing protocol⁵⁹. In brief, to delete each gene, we injected two sgRNAs targeting both 5′ and 3′ ends of each ugt coding region. dpy-10 sgRNA was co-injected as a co-CRISPR marker with dpy-10(cn64) ssDNA as a repair template⁶⁶. After injecting 100 animals, F2 rollers were screened and genomic DNA PCR was performed to each roller animal to detect knockout alleles of ugt genes. Each ugt deletion mutant was outcrossed three times with N2.

To delete D1005.2 in C. elegans, we designed to induce gene deletion through non-homologous end-joining repair of double strand breakdown that was described in a previously reported CRISPR-Cas9 genome editing protocol⁶⁷.

To delete the galU genomic region in Comamonas, we modified a Pseudomonas CRISPR-Cas9 genome editing method⁶⁸. We found a single UTP glucose-1-phosphate uridylyltransferase (galU) orthologue in the Comamonas genome³⁷. sgRNA was designed to target 5′ end of galU open reading frame and it was incorporated to pACRISPR plasmid by Golden gate assembly⁶⁸. We predicted that the Comamonas genome does not encode the DNA end binding protein, KU or a DNA ligase, and therefore, we predict that it does not use non-homologous end-joining DNA repair machinery^37,69. Therefore, to enable homologous recombination, adjacent DNA sequences of both the 5′ and 3′ end of the galU genomic region were provided as a DNA repair template by Gibson assembly. For Gibson assembly, the vector was linearized by PCR from a circular plasmid followed by a DpnI restriction enzyme digest (NEB, R0176) to remove the methylated circular plasmid DNA template.

We found that Comamonas is resistant to β-lactam antibiotics. Therefore, we replaced the antibiotic selection marker, amp^R in the pACRISPR plasmid with a gentamycin resistant aacC1 marker which originates from a Pseudomonas transposon, Tn1696 by Gibson assembly^70,71. This Amp^R replaced plasmid, pACRISPR-gent was validated by plating E. coli DH5α having pACRISPR-gent plasmid on LB agar plates containing different antibiotics. Cas9 nuclease coding sequence was encoded in pCasPA plasmid with λ-Red recombination system proteins that increased efficiency of homologous recombination⁶⁸.

To prepare Comamonas for electroporation, a single colony was inoculated in 5 ml LB medium and incubated at 37 °C for 24 h with 200 rpm orbital shaking. Overnight culture was diluted 50-fold in fresh 5 ml LB medium and incubated until OD_600nm = 0.6. This diluted secondary culture (1.5 ml) was washed three times with H₂O at room temperature. After a third wash, bacteria were resuspended in 0.1 ml H₂O and mixed with 0.5 μg plasmid DNA. Prepared bacteria and plasmid DNA were electroporated at 15 kV cm⁻¹ in a Gene Pulser Cuvette (Bio-Rad). After electroporation, bacteria were transferred in a 15-ml round-bottom tube and immediately 1 ml prewarmed (37 °C) super optimal broth with catabolite repression (SOC) medium was added. Cells were incubated at 37 °C for 2 h with 200 rpm orbital shaking and seeded on LB agar plates containing antibiotics (10 μg ml⁻¹ tetracycline for pCasPA; 10 μg ml⁻¹ gentamycin for pACRISPR-gent) and incubated at 37 °C for 2 days with parafilm sealing. The presence of plasmid was validated by antibiotic selection, colony PCR and plasmid PCR.

For expression of Cas9 nuclease and λ-Red recombination system proteins, Comamonas was inoculated in 5 ml LB medium containing antibiotics (10 μg ml⁻¹ tetracycline for pCasPA and 10 μg ml⁻¹ gentamycin for pACRISPR-gent) and incubated at 37 °C for 24 h with 200 rpm orbital shaking. Overnight culture was 1:20 diluted in fresh 20 ml LB medium with antibiotics and incubated until OD_600nm = 0.6 in 125 ml flask for aeration. Then, a 20-ml culture was washed three times with H₂O at room temperature. After the third wash, the bacterial pellet was resuspended in 20 ml bacterial M9 plus amino acids, l-arabinose and isopropyl β-d-1-thiogalactopyranoside (IPTG) (1× M9 salt, 0.4% glucose, 1 mM MgSO₄, 1 mM CaCl₂, 0.002% histidine, 0.006% leucine, 0.003% lysine, 0.002% methionine, 0.5% adenine sulfate, 20 mg ml⁻¹ l-arabinose and 2 mM IPTG). Resuspended bacteria were incubated at 37 °C for 48 h with 200 rpm orbital shaking. Then, 50 μl of the bacterial culture was seeded on LB agar plates containing antibiotics and incubated at 37 °C for 2 days with parafilm sealing. Each single colony was validated by colony PCR and sequencing of isolated genomic DNA.

To remove the pCasPA and pACRISPR-gent plasmids from Comamonas ΔgalU mutants, a single colony was inoculated in 5 ml LB medium without any antibiotics and incubated at 37 °C for 24 h with 200 rpm orbital shaking. The overnight culture was diluted to 1:10,000 in LB medium and 100 μl of diluted culture was seeded on LB agar containing no antibiotics or selective antibiotics and incubated overnight at 37 °C. Five colonies from non-antibiotic-containing LB agar were tested to verify the absence of the plasmids by plating each colony on LB agar containing tetracycline or gentamycin for antibiotic selection.

RNA interference

Animals were fed a diet of E. coli HT115 for three generations before assay. RNAi clones were cultured in LB containing 50 μg ml⁻¹ ampicillin at 37 °C for 20 h and seeded on NGM agar with 2 mM IPTG (Fisher Scientific). Plates were dried and incubated at room temperature for 48 h. After 2 days, approximately 100 synchronized L1 animals were plated, followed by incubation at 20 °C for 72 h for counting P0 adult viability and continuously incubated further 72 h to examine F1 embryo viability. For the D1005.2 RNAi experiment, animals were exposed to vector or D1005.2 RNAi for more than two generations before being fed with Comamonas or Comamonas ΔgalU mutant diets.

Statistic and reproducibility

Statistical analysis was performed using a Student’s t-test. The data were analysed using the Data Analysis Toolpak in Microsoft Excel to compare the means and s.d. between groups. Data distribution was assumed to be normal, but this was not formally tested. Individual data points are shown in the figures. No data were excluded from the analyses. No statistical methods were used to pre-determine sample sizes. For HPLC–HRMS we used 50,000 animals, which has been shown to provide enough material to give a detectable reading for the assay. For developmental/lethality we used ~100 worms, which has been shown to provide a sample size which will produce statistically significant results³⁷. For the Comamonas mutant screen we used 10–20 worms per well to prevent the bacteria from being consumed before the end of the experiment. For the RNA-seq we used 250 worms per condition, which provides the minimum amount of total RNA to generate a statistically significant number of reads. Multiple biological replicates for each experiment were conducted as mentioned in the figure legends. For all experiments, batches of C. elegans for the specified genotype under each condition were randomly selected from experimental plates and assayed as appropriate. Data collection and analysis were not performed blind to the conditions of the experiments as blinding is not relevant to our study; knowing the genotype or condition did not bias the study.

Extended Data

Extended Data Fig. 1 |. Alignment of human (H.s.) MCCC1 and C. elegans (C.e.) MCCC-1 protein sequences.

a, Alignment of protein sequences, asterisks indicate reported amino acid alterations in human 3-MCC deficiency patients. Four C. elegans mccc-1 mutant alleles are in bold, and their altered amino acids are marked below C. elegans MCCC-1. Biotin carboxylase and biotinyl-binding domains are marked as black and grey underbars, respectively. b, Cartoon of the C. elegans MCCC-1 protein showing the mutant alleles.

Extended Data Fig. 2 |. Conjugation products of leucine and its catabolites.

a, A schematic of Leu supplements provided to animals during C. elegans culture in liquid medium used to generate exo- and endo-metabolome samples. b, c, Bar graphs showing tiglyl-carnitine (b), isovaleryl-carnitine, 5-methylhexanoyl(C7ISO)-carnitine, 7-methyloctanoyl(C9ISO)-carnitine, HMB-choline, and HMB-pantetheine (c) levels in control and mccc-1(ww4) mutant animals. d, e, f, HPLC–HRMS analysis of mecglu#72: 3-methylcrotonyl, HMB and mecglu#: 3-methylcrotonyl (2x) (d), mecglu#1 (e), 2HIC-Leu (f). g, Bar graphs showing N-acetyl-Leu and N-propionyl-Leu levels in control and mccc-1(ww4) mutant animals. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± standard deviation. Statistical significance was determined using a two-sided unpaired Student’s t-test.

Extended Data Fig. 3 |. HMB, isovalerate, and 3MC toxicity.

a, Dose–response curves of control animals fed Comamonas supplemented with HMB, isovalerate or 3MC. Data represent five HMB, and three each for IV and 3MC supplemented biologically independent experiments and error bars indicate mean ± s.d. b, Toxicity of 60 mM HMB in C. elegans control, mccc-1(ww2), mccc-1(ww4), mccc-1(ww20), mccc-1(ww38), mccc-2(tm12463) or mccc-2(tm15274) mutant animals fed Comamonas. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± standard deviation. Statistical significance was determined using a one-sided unpaired Student’s t-test.

Extended Data Fig. 4 |. Mitochondrial membrane potential decreases in C. elegans fed Comamonas supplemented with HMB.

a, Box plot representing the intensity of TMRE staining in C. elegans control or mccc-1(ww4) mutant mitochondria with or without HMB supplementation and fed a Comamonas diet. The vertical lines (whiskers) extend from the minimum to the maximum scores, excluding outliers, which are represented by circles above the maximum value. The box itself delineates the interquartile range (IQR), capturing the middle 50% of scores, with the bottom and top edges indicating the 25th and 75th percentiles, respectively. A horizontal line within the box marks the median value, providing a clear indication of the distribution’s centre. Outliers, indicating data points significantly above the maximum score, underscore variations beyond the typical range of observations. Data represents control (n = 53), mccc-1(ww4) mutants (n = 50), control (n = 54) and mccc-1(ww4) mutants (n = 54) with 60 mM HMB supplementation from three biologically independent experiments. n.s indicates statistically not significant p-value. Statistical significance was determined using a two-sided unpaired Student’s t-test. b, Representative images of TMRE stained mitochondria. Figures represent one of three biologically independent experiments. Scale bar, 10 μm.

Extended Data Fig. 5 |. Comamonas pyrimidine biosynthesis is required to mitigate HMB toxicity.

a, Schematic of Comamonas transposon mutant screen with 60 mM HMB. b, Bacterial growth of wild-type and mutant strains and supplemented with nucleobases as indicated. c, Bar graphs of C. elegans grown with and without 60 mM HMB and fed wild-type or either of the two Comamonas mutants supplemented with uracil or orotate during bacterial culture. Each bar represents the mean value of three biologically independent experiments,indicated by a white dot, with error bar showing the mean ± standard deviation.

Extended Data Fig. 6 |. Supplementation with uracil does not alleviate HMB toxicity in C. elegans fed an E. coli diet.

a, Toxicity of titrated HMB in animals fed E. coli OP50, HT115, BW25113 or Comamonas supplemented with 1 mM uracil during bacterial culture. b, Toxicity of titrated HMB in animals fed E. coli OP50, or Comamonas pyrE⁻ mutants supplemented with uracil during bacterial culture. c, Toxicity of 40 mM HMB in animals fed E. coli BW25113, E. coli pyrE⁻ mutant, Comamonas, or Comamonas pyrE⁻ mutants supplemented with uracil during bacterial culture. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± standard deviation. Statistical significance was determined using a one-sided unpaired Student’s t-test.

Extended Data Fig. 7 |. UTP glucose-1-phosphate uridylyltransferase gene expression is increased in mccc-1(ww4) mutant animals and partially protects against HMB toxicity.

a, RNA-seq data of differentially expressed metabolic genes in mccc-1(ww4) mutants fed Comamonas. b, Fold change in mRNA levels of upregulated C. elegans ugt genes in mccc-1(ww4) mutants measured by RNA-seq. c, Cartoon of ugt-32 and ugt-53 deletion alleles created by CRISPR/Cas9 genome editing. d, e, Toxicity of 20 mM (d) or 40 mM (e) HMB toxicity in six different C. elegans strains fed wild-type or pyrE mutant Comamonas as measured by the proportion of animals reaching the L4 stage. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± standard deviation. Statistical significance was determined using a one-sided unpaired Student’s t-test. n.s indicates statistically not significant p-value. f, Bar graphs showing exometabolomic mecglu# levels in six different C. elegans strains fed Comamonas. Each bar represents the mean value of three biologically independent experiments, indicated by a white dot, with error bar showing the mean ± standard deviation. Statistical significance was determined using a two-sided unpaired Student’s t-test. g, h, Cartoon of Comamonas galU deletion (g) and C. elegans D1005.2 deletions (h) generated by CRISPR/Cas9 genome editing.

Extended Data Fig. 8 |. mccc-1 mutants, but not mccc-2 mutants are vulnerable to hmgr-1 RNAi knockdown.

Offspring of P0 animals exposed to hmgr-1 RNAi with or without mevalonate supplementation. Figures represent one of three biologically independent experiments. Scale bar, 1 mm.

Extended Data Table 1 |.

Comamonas mutant strains that modify HMB toxicity in C. elegans

Comamonas mutant	Disrupted gene product by transposon insertion	Homologous gene in E. coli K-12MG1655
Plate number/ well location in library
14A7	Valyl tRNA synthase	ValS
39D7	Exodeoxyribonuclease V beta	recB
41H7	Orotate phosphoribosyltransferase	pyrE
42H11	Dihydroorotase	pyrC
51D3	Intergenic region	Not applicable

Data availability

All data are fully available without restriction, RNA-seq data are available with Gene Expression Omnibus accession no. GSE233533. The HPLC–MS/MS data generated during this study have been deposited in the MassIVE database under accession code MSV000093813 (https://doi.org/10.25345/C5F47H501). Source data are provided with this paper.