HLH‐30‐dependent rewiring of metabolism during starvation in C. elegans

Abstract One of the most fundamental challenges for all living organisms is to sense and respond to alternating nutritional conditions in order to adapt their metabolism and physiology to promote survival and achieve balanced growth. Here, we applied metabolomics and lipidomics to examine temporal regulation of metabolism during starvation in wild‐type Caenorhabditis elegans and in animals lacking the transcription factor HLH‐30. Our findings show for the first time that starvation alters the abundance of hundreds of metabolites and lipid species in a temporal‐ and HLH‐30‐dependent manner. We demonstrate that premature death of hlh‐30 animals under starvation can be prevented by supplementation of exogenous fatty acids, and that HLH‐30 is required for complete oxidation of long‐chain fatty acids. We further show that RNAi‐mediated knockdown of the gene encoding carnitine palmitoyl transferase I (cpt‐1) only impairs survival of wild‐type animals and not of hlh‐30 animals. Strikingly, we also find that compromised generation of peroxisomes by prx‐5 knockdown renders hlh‐30 animals hypersensitive to starvation, which cannot be rescued by supplementation of exogenous fatty acids. Collectively, our observations show that mitochondrial functions are compromised in hlh‐30 animals and that hlh‐30 animals rewire their metabolism to largely depend on functional peroxisomes during starvation, underlining the importance of metabolic plasticity to maintain survival.


| INTRODUC TI ON
The ability to regulate metabolism in response to changes in nutrient availability is an evolutionarily conserved mechanism ranging from bacteria to humans. Regulating metabolism by coordinating anabolic and catabolic pathways to sustain metabolic homeostasis ensures prolonged survival during periods of nutrient scarcity. However, rewiring of energy metabolism is a complex and dynamic process encompassing many transcriptional and post-transcriptional regulators. One of the fundamental mechanisms promoting survival in response to starvation is the use of energy stores, for example, via breakdown of lipids. Fatty acids are mobilized from lipid droplets in adipocytes for mitochondrial β-oxidation through either lipolysis or lipophagy; pathways that have both been reported to be crucial for surviving starvation (Martin & Parton, 2006). A major regulator of lipid metabolism during starvation is the conserved basic helixloop-helix transcription factor HLH-30 in Caenorhabditis elegans (C. elegans), an ortholog of the mammalian transcription factor EB (TFEB; Lapierre et al., 2013). HLH-30/TFEB regulates the expression of genes belonging to the coordinated lysosomal expression and regulation (CLEAR) network, which are involved in autophagosome formation, lysosomal biogenesis, lipase function, and fatty acid degradation (Martina et al., 2014;Palmieri et al., 2011;Settembre et al., 2013). HLH-30/TFEB is also a transcription factor-mediating resistance to several stressors besides starvation, including oxidative stress, heat stress, and host defense against pathogen infection (Lin et al., 2018;Visvikis et al., 2014). Additionally, removal of HLH-30/ TFEB impairs the longevity of several long-lived C. elegans mutants (Lapierre et al., 2013) and entry into adult reproductive diapause (Gerisch et al., 2020), and the hlh-30 mutant itself dies prematurely during starvation (Harvald et al., 2017;O'Rourke & Ruvkun, 2013).
In the present study, we have successfully applied a combinatorial metabolomics and lipidomics approach to examine temporal regulation of metabolism during starvation and how HLH-30 regulates metabolism during starvation in C. elegans. Specifically, we find that starvation induces significant and specific changes in the metabolome and in the lipidome of C. elegans. In particular, we find that starvation induces both long-chain acyl-carnitine and cardiolipin levels in wildtype animals, in accordance with enhanced mitochondrial metabolism.
Accordingly, we find that starvation induces oxidation of oleic acid.
Markedly, induction of cardiolipin and acyl-carnitine levels and oxidation of oleic acid upon starvation are completely absent in hlh-30 animals, arguing that HLH-30 is required for induction of mitochondrial β-oxidation during starvation. Interestingly, we find that impaired generation of peroxisomes induces premature death of hlh-30 animals upon starvation, which cannot be rescued by supplementation of exogenous fatty acids. Collectively, we show for the first time, that functional loss of HLH-30 renders C. elegans highly dependent on peroxisomal degradation of fatty acids to survive starvation. Our observations substantiate the importance of metabolic plasticity in order to survive periods of nutrient scarcity.

| Starvation induced metabolic and lipidomic re-arrangement enhances mitochondrial function in C. elegans
To identify the metabolic response to starvation, we analyzed the temporal response to starvation in C. elegans across a 16 h starvation time course at the mid-L4 stage by metabolomics and lipidomics ( Figure 1).
As previously (Harvald et al., 2017), we analyzed multiple time points within the first 6 h to interrogate the early starvation responses with high resolution. We harvested animals in biological triplicate at each of the 7 time points for extraction of metabolites and lipids, respectively, and for subsequent analyses by MS-based metabolomics and lipidomics ( Figure 1). To optimize lipid extraction from C. elegans, we applied different commonly used lipid extraction methods and performed lipid profiling in positive ionization mode. By using the BUME extraction (Lofgren et al., 2012), we detected 4.963 different molecular features in the apolar phase, while we detected 4819, 4249, 5172, and 5082 molecular features when using Bligh and Dyer (Bligh & Dyer, 1959), Folch (Folch et al., 1957), MMC (Pellegrino et al., 2014), and MTBE (Matyash et al., 2008) extraction methods, respectively ( Figure S1). 4143 features were commonly detected in all tested extraction methods ( Figure S1). Although the MTBE extraction yielded the most features, we also found that this method showed the highest extraction variability. In contrast, the Folch extraction showed not only the lowest variability, but also the lowest number of features ( Figure S1). Based on the overall performance, the BUME extraction method not only showed a high number of features with a low variability, but also provided a polar phase for LC-MS metabolomic analyses of polar metabolites. Furthermore, it showed better recovery of polar lipids such as lysolipids compared with the other methods. Applying the BUME extraction and lipid profiling to our samples, 4063 lipid features in positive and 2258 in negative ionization mode remained after normalization and filtering (detected in all QCs and RSD < 30%). Out of these, 2068 were putatively annotated on the MS 1 level and 427 on the MS 2 level in positive ionization mode, 955 and 118 in negative mode, respectively (Table S1).
We generated volcano plots to visualize to which extend starvation rewires the C. elegans metabolome across a 16 h starvation time course (Figures 2a and S2). This showed that short-term starvation (1 and 2 h) only has subtle effects on the metabolome in wild-type animals, while starvation for 3 h profoundly changes the metabolome (61 significantly altered metabolites in total in negative and positive modes; Table S1).
Further, we found that the abundance of 289, 271, and 501 molecular features was significantly altered after 4, 6, and 16 h of starvation, respectively (Figures 2a and S2, and Table S1). During starvation, lipolysis of triacylglycerols plays an important role as an energy source in both mammals and in C. elegans (Buis et al., 2019;Lee et al., 2014;Martin & Parton, 2006;Murphy et al., 2019;Zaarur et al., 2019). Accordingly, the fatty acids released by lipolysis are subsequently activated to CoAesters and transported into the mitochondria by the carnitine-shuttle system for degradation by β-oxidation. Consistently, we observed that the level of long-chain acyl-carnitines massively increased already after one hour and remained elevated after 16 h of starvation in wild-type animals, while levels of short-chain acyl-carnitines largely remained unchanged (Figures 2a and 3a).
By similar means, we visualized how the C. elegans lipidome alters across the starvation time course (Figures 2b and S3). The volcano plots clearly show that starvation also rewires the lipidome already after 1 h, as 79, and 99 lipid species were significantly up-or downregulated in positive and negative mode, respectively (Figures 2b and S3, and   Table S1). Consistent with previous observations (Lucanic et al., 2011), we also found that the abundance of lipids like N-acylethanolamines (NAEs) decreased in wild-type animals across the starvation time course (Figure 2b). Changed lipids were grouped according to their profile across time. Based on this, lipids could be separated into species changing early within the first time points and later responders, for example, triacylglycerols show changes at 16 h of starvation. Moreover, among the upregulated lipid species we identified seven lipid species in wild-type animals that significantly increased across all time points except after 16 h of starvation. Based on their elution profile and MS-fragmentation pattern, we identified these lipid species to be cardiolipins ( Figure S4). Cardiolipins are major phospholipids almost exclusively located in the inner mitochondrial membrane and required for mitochondrial morphology, mitochondrial membrane dynamics, and energy production not only in C. elegans but also in other eukaryotes (Sakamoto et al., 2012;Sustarsic et al., 2018). Collectively, these results show that starvation induces major re-arrangements of both the metabolome and lipidome, and that mitochondrial functions are enhanced by starvation in wild-type C. elegans.

| Rewiring of lipid metabolism and induction of β-oxidation upon starvation depend on HLH-30 in C. elegans
Since the transcription factor HLH-30 and its mammalian ortholog TFEB previously have been shown to serve crucial functions F I G U R E 1 Experimental workflow for combined metabolomics and lipidomics of the starvation response in C. elegans. Wild-type C. elegans and the hlh-30 mutant were included in the study. Starvation was induced by transferring animals to plates containing no bacteria. Worms were starved for 1, 2, 3, 4, 6, and 16 h of starvation, prior to harvesting and extraction of metabolites and lipids. The upper aqueous phase was analyzed using LC-MS-based metabolomics, and the lower organic phase was analyzed using LC-MS-based lipidomics | 5 of 13 DALL et AL.
during starvation, dietary restriction, and autophagy in both C. elegans and in mammals (Harvald et al 2017;Lapierre et al., 2013;Murphy et al., 2019;O'Rourke & Ruvkun, 2013;Roczniak-Ferguson et al., 2012;Settembre et al., 2013), this prompted us to examine how functional loss of HLH-30 modulates the metabolome and the lipidome in response to starvation in C. elegans. In contrast to wildtype animals, we only found a limited number of metabolites and lipid species that changed significantly in response to starvation in hlh-30 animals ( Figure 2b and Table S1). Markedly, we found that the level of long-chain acyl-carnitines in hlh-30 animals remained largely unchanged in response to starvation compared to wild-type animals ( Figures 2a and 3a), arguing that fatty acid import into mitochondria is compromised. In keeping with this notion, we also found that cardiolipin levels in hlh-30 animals largely remained unchanged in response to starvation (Figures 2b and 3b). We, therefore, speculate that HLH-30/TFEB is required for biogenesis or for the maintenance of functional mitochondria. Thus, to corroborate these observa- This observation substantiates that HLH-30 is required for induction of fatty acid oxidation during starvation and hence for metabolic adaptation during starvation. Interestingly, we did not observe any overt alterations in mitochondria morphology in hlh-30 animals by Mitotracker staining (results not shown), and collectively arguing that HLH-30/TFEB is required for maintenance of functional mitochondria.

| Fatty acid supplementation rescues premature death during starvation in the hlh-30 mutant
TFEB, the mammalian ortholog of HLH-30, has recently been found to be required for mitochondrial biogenesis, morphology, and functions in skeletal muscle in mice (Mansueto et al., 2017). Although loss of HLH-30 functions does not affect mitochondria morphology in C. elegans (Murphy et al., 2019), the present observations show that HLH-30 is required to support fundamental mitochondrial functions in C. elegans during limited nutritional conditions. Compared with wild-type animals, hlh-30 animals die prematurely during starvation (Harvald et al., 2017). Since mobilization of fatty acids from intestinal lipid stores is required for C. elegans to withstand long-term starvation (Buis et al., 2019), we, therefore, hypothesized that exogenous supplementation of medium-chain fatty acids, which cross the mitochondrial membranes independent of the carnitine-shuttle system, would rescue the premature death of hlh-30 animals. We, therefore, examined survival under starvation conditions by transferring animals to empty plates supplemented with either a mediumchain (lauric acid, C 12:0 ) or a long-chain fatty acid (palmitic acid, C 16:0 ). As previously, we found that hlh-30 animals die prematurely during starvation when compared to wild-type animals. However, when supplemented with lauric acid both wild-type and hlh-30 animals survived significantly longer during starvation when compared to un-supplemented animals ( Figure 4 and Table S2). In fact, hlh-30 animals were completely rescued to wild-type levels by lauric acid. Supplementation with palmitic acid also extended the survival of wild-type animals and surprisingly also of hlh-30 animals, however, not to the same extend as lauric acid ( Figure 4 and Table S2).
Notably, we found that uptake of palmitic acid in to both wild-type and hlh-30 animals under starvation conditions is higher compared with uptake of lauric acid (results not shown), suggesting that the ability to rescue survival of hlh-30 animals is not due to increased uptake of lauric acid.

| Disruption of the carnitine-shuttle system impairs starvation survival of wild-type animals
Our findings support the notion that impaired β-oxidation, caused by, for example, diminished mitochondrial import of long-chain fatty acids, may be the underlying reason for the inability of the hlh-30 mutant to survive during starvation. By RNA-sequencing Harvald et al.
recently profiled the genome-wide response to starvation (Harvald et al., 2017), and found that expression of genes encoding lipases needed for conventional lipolysis of triacylglycerols in lipid droplets (atgl-1) or via lysosomal breakdown (lipl-2 to lipl-4) increased in wildtype animals upon starvation but remained constant or decreased in hlh-30 animals during starvation ( Figure S5). Similarly, the expression of genes encoding enzymes required for activation of fatty acids (acs-2) and for active transport of fatty acids into the mitochondria (cpt-1) is diminished in the mutant upon starvation ( Figure S5). We, therefore, speculated that downregulation of cpt-1 expression would impair the ability of wild-type animals to survive under starvation conditions. Intriguingly, we found that RNAi-mediated knockdown of cpt-1 significantly impairs survival of wild-type animals under F I G U R E 2 Metabolic and lipidomic changes induced by temporal starvation. (a) Volcano plot displaying changes in the metabolome in response to starvation for the wild-type and hlh-30 mutant at each time point. Significant up-or downregulated metabolites in wild-type animals are shown in blue, and in light green for the hlh-30 animals. Regulation of long-chain acyl-carnitines are shown in red and short-chain acyl-carnitines are shown in green for both wild-type and the hlh-30 mutant. Only, metabolites detected in the positive mode are shown. (b) Volcano plot displaying changes in the lipidome in response to starvation for the wild-type and hlh-30 animals at each time point. Significant up-or downregulated lipids in the wild type are shown in blue, and in light green for hlh-30 animals. Regulation of cardiolipins are shown in green and N-acylethanolamines are shown in red for both wild-type and hlh-30 animals. Only, lipid species detected in the positive mode are shown. Metabolites and lipids with a p-value <.05 and a fold-change of >2 or <0.5 were considered to be significantly changed starving conditions compared with the control animals, while cpt-1 knockdown had no effect on survival of hlh-30 animals ( Figure 5 and Table S3). Markedly, lauric acid supplementation fully rescued the effects of cpt-1 knockdown in wild-type animals and extended survival of hlh-30 animals to wild-type levels independent of cpt-1 knockdown. Consistent with the notion that mitochondrial import of long-chain fatty acids depends on a functional carnitine-shuttle system, palmitic acid (C 16:0 ) supplementation did not fully rescue survival of wild-type control animals. Interestingly, palmitic acid supplementation rescued survival of hlh-30 animals under starvation conditions independent of cpt-1 knockdown, indicating that fatty acids can support survival during starvation by being channeled to other energy-producing pathways than mitochondrial β-oxidation.

ATGL-1 mediates lipolysis of intestinal lipid stores during fasting
in C. elegans (Lee et al., 2014). We, therefore, assessed whether the mobilization of stored lipids would affect survival during starvation.  Expectedly, knockdown of atgl-1 in wild-type animals significantly shortened survival compared with its control during starvation but had only minor effects on the survival-span of hlh-30 animals ( Figure 5b). Supplementation with lauric and palmitic acid both extended survival-span of wild-type and hlh-30 animals, yet only lauric acid extended survival-span to wild-type levels. All together, we interpret these observations that mobilization and mitochondrial import of fatty acids from lipid stores are crucial for surviving during starvation, however, only the latter is dependent on HLH-30 in C. elegans.

| Survival of the hlh-30 mutant during starvation is dependent on peroxisomal β-oxidation
Since supplementation of palmitic acid also improved, survival of hlh-30 animals during starvation made us speculate whether hlh-30 animals compensate by using alternative metabolic pathways to generate sufficient energy to survive starvation. Besides mitochondria, peroxisomes are also capable of degrading fatty acids.
Like mitochondrial β-oxidation, peroxisomal β-oxidation catalyzes chain shortening of acyl-CoAs by four enzymatic steps yielding acetyl-CoA. Despite that peroxisomal β-oxidation in C. elegans is mostly known for oxidizing very long-chain fatty acids (VLCFAs) and for the synthesis of ascarosides (Artyukhin et al., 2018), longand medium-chain saturated and unsaturated fatty acids can also serve as substrates for peroxisomal β-oxidation (Poirier et al., 2006).
The first step is catalyzed by the enzyme acyl-CoA oxidase (ACOX), considered to be the main regulator of the flux through the pathway. Interestingly, expression of acox genes in hlh-30 animals is upregulated compared with wild-type animals and sustain upregulated through starvation (Harvald et al., 2017; Figure S5). Therefore, since mitochondrial functions are compromised in hlh-30 animals, we speculated that they rewire their metabolism toward peroxisomes in order to survive starvation. Thus, by RNAi we knocked down prx-5 that encodes a peroxisomal assembly factor required for biogenesis of functional peroxisomes (Weir et al., 2017). Intriguingly, upon knockdown of prx-5, survival of hlh-30 animals during starvation was dramatically decreased when compared to its control, while prx-5 knockdown in wild-type animals did not affect survival during starvation ( Figure 6 and Table S3). Moreover, neither lauric acid nor palmitic acid could rescue the effect of prx-5 knockdown, collectively implying that hlh-30 animals switch their metabolic program toward peroxisomes.
In conclusion, by using a systems-wide analyses we have demon-

F I G U R E 6
The ability of the hlh-30 mutant to survive starvation is dependent on peroxisomal β-oxidation. Survival-span showing the effect of RNAi-mediated knockdown of prx-5 on the survival of wild-type and hlh-30 to mutant during starvation. Starvation was induced at start L4 stage by transferring worms to empty plates. Both strains were supplemented with either a medium-chain fatty acid, lauric acid (C 12:0 ) or a long-chain fatty acid, palmitic acid (C 16:0 ) throughout the experiment. Survival was monitored every day. Survival analysis was carried out using the Kaplan-Meier estimator and p-value was calculated using log-rank test in GraphPad Prism 6  survival of wild-type animals during starvation but not of hlh-30 animals. However, fatty acid supplementation extends lifespan of both wild-type and hlh-30 animals, further supporting that mobilization of fatty acids from lipid stores provides metabolic energy to support organismal lifespan. Consistent with our findings, Macedo et al. recently reported that the abundance of certain cardiolipin species increases in response to dietary restriction (Macedo et al., 2019), which increased further upon loss of LIPL-5.
Collectively, this study provides a comprehensive analysis of temporal starvation responses that combine both metabolomic and lipidomic analyses. Our data highlight the relevance of combining global profiling analyses to further understand how metabolism is regulated and how an organism adapts to nutritional changes. As exemplified by our comparison of the effect of starvation on metabolites and lipids in wild-type and hlh-30 animals, the genetic tractability of C. elegans and combined with RNA interference, shows that the nematode system serves as an excellent framework to delineate conserved mechanisms whereby specific metabolic pathways regulate starvation responses.

| C. elegans strains and maintenance
The wild-type N2 Bristol and the hlh-30 mutant (tm1978, a kind gift from Dr. Marlene Hansen, Sandford-Burnham Medical Research Institute) were used and cultivated under standard conditions and handled as described (Harvald et al., 2017).

| Survival-span assay
Worms were synchronized and grown until L4 stage on NGM seeded plates. For RNAi treatment, worms were transferred to plates containing IPTG seeded with the respective HT115 RNAi bacteria clone.
At L4, worms were transferred to empty plates to induce starvation conditions. For supplementation with fatty acids, worms were transferred to plates containing either 40 µM lauric acid (C 12:0 ) or palmitic acid (C 16:0 ). 12 worms were placed on each plate and scored every day. A worm was scored dead when it was unresponsive to gentle prodding.

| Mitotracker staining and microscopy
Eggs from synchronized adults were grown until L4 stage on NGM seeded plates. Worms were stained as previously described (Ruiz et al. 2019

| Sample harvest for mass spectrometry
Worms were synchronized as described above and grown on NGM

| BUME extraction
Samples were extracted by using the BUME method as described (Lofgren et al., 2012)

| Lipid analysis, data processing, and statistical analysis
Lipids were analyzed as previously described (Witting et al., 2014).
Briefly, lipids were separated on a Waters Acquity UPLC (Waters)

| Metabolomics analysis and data processing of MeOH/ACN/H 2 O extracted metabolites
Metabolites were analyzed according to (Sustarsic et al., 2018) with minor alterations to the gradient. In brief, metabolites were analyzed with LC-MS using reverse phase (RP) separation. 5 μl were injected using an Agilent 1290 Infinity HPLC system (Agilent

| β-oxidation assay
The method for measuring fatty acid oxidation was applied as described previously (Elle et al., 2012).

This work was supported by The Danish Council for Independent
Research, Natural Sciences (6108-00268A). We gratefully acknowledge technical help from Ditte Neess and Alice Dupont Juhl and scientific discussions with Marta Moreno-Torres.

CO N FLI C T O F I NTE R E S T
None declared.