Mitochondrial energy state controls AMPK-mediated foraging behavior in C. elegans

Abstract

Organisms surveil and respond to their environment using behaviors entrained by metabolic cues that reflect food availability. Mitochondria act as metabolic hubs and at the center of mitochondrial energy production is the protonmotive force (PMF), an electrochemical gradient generated by metabolite consumption. The PMF serves as a central integrator of mitochondrial status, but its role in governing metabolic signaling is poorly understood. We used optogenetics to dissipate the PMF in Caenorhabditis elegans tissues to test its role in food-related behaviors. Our data demonstrate that PMF reduction in the intestine is sufficient to initiate locomotor responses to acute food deprivation. This behavioral adaptation requires the cellular energy regulator AMP-activated protein kinase (AMPK) in neurons, not in the intestine, and relies on mitochondrial dynamics and axonal trafficking. Our results highlight a role for intestinal PMF as an internal metabolic cue, and we identify a bottom-up signaling axis through which changes in the PMF trigger AMPK activity in neurons to promote foraging behavior.

Optogenetic dissipation of the PMF unveils a transcellular signaling axis integrating metabolic demand with foraging behavior.

INTRODUCTION

The ability to survey the environment by orchestrating internal metabolic signals with external cues that report the presence of food is an important aspect of animal behavior essential for survival. In Caenorhabditis elegans, environmental stimuli, such as the presence of food, are sensed by the olfactory and gustatory systems, which serve to direct the animals to food sources (1). When C. elegans contact food, they reduce their speed to extend their time at the food source (2). Once the food source is depleted, C. elegans increase their speed and initiate a local search behavior, performing omega-shaped turns and backward movements termed reversals to seek a new food source in close proximity. Over time, worms transition to global search behavior where omega turns and reversals are limited if a new food source is not found, allowing foraging to occur over a larger area using body bends for linear forward movement (3–5). These coordinated behaviors are time-dependent and metabolic status dependent and serve to direct C. elegans toward energy homeostasis.

Food deprivation results in internal energy deficiency, and this metabolic context affects the decision-making processes of many animals (6, 7). Internally, energy homeostasis is maintained in part by AMP-activated protein kinase (AMPK), in concert with a variety of other metabolic signaling pathways (8, 9). When C. elegans are removed from food, they increase speed in search of food via an AMPK-dependent mechanism. Loss of AMPK abolishes this accelerating response, and animals are unable to respond to the lack of food (10–12). Upstream signaling pathways that activate these AMPK-mediated behavioral responses remain unclear.

Food is metabolized through a network of enzymatic processes and ultimately generates adenosine 5′-triphosphate (ATP) via mitochondrial oxidative phosphorylation (OxPhos). During OxPhos, complexes I, III, and IV of the electron transport chain (ETC) pump protons across the inner mitochondrial membrane into the intermembrane space, generating a protonmotive force (PMF). Subsequently, the PMF is used by ATP synthase to produce ATP. Beyond energy supply, the PMF can control cellular function via calcium signaling (13), protein import (14), redox signaling (15), and stress response (16). The PMF also can modulate mitochondrial morphology through fusion and fission events. AMPK initiates mitochondrial fission via phosphorylation of mitochondrial fission factor and facilitates recruitment of pro-fission dynamin-related protein–1 (DRP-1) (17, 18).

Although the PMF is central to cellular function, how the PMF is sensed and communicated throughout an organism is poorly understood. The PMF has a dynamic character that varies in normal physiology and disease as mitochondria respond to cellular cues. Modulation of the PMF is a hallmark of mitochondrial dysfunction associated with pathology, including ischemic conditions and neurodegenerative diseases (19, 20). Loss of PMF decreases the metabolic energy state by reducing ATP production. In C. elegans, intestinal food intake defines the internal metabolic state, and perception of this state involves neuronal plasticity and signaling (4, 6, 21). The energy demands of the animal shape the food-related behavioral response. Available pharmacologic and genetic approaches to manipulate the PMF lack sufficient precision to fully explore cause-and-effect relationships. For example, protonophores, such as carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP), that collapse the PMF, lack tissue selectivity and once applied cannot be reversed. Hence, using these compounds limits dissecting how tissues dynamically communicate food status. Likewise, genetic approaches such as overexpression of uncoupling proteins can cause large-scale metabolic remodeling that may impart the limited ability to interpret experimental outcomes for acute signaling pathways. To address these limitations, we used an optogenetic approach to spatiotemporally control the PMF and assess the causal role of the PMF in regulating intertissue signaling.

In this study, we investigate how food-related behaviors are affected by a tissue-specific optogenetic reduction of the PMF. We demonstrate that foraging behavior is driven by a reduction in the PMF in the intestine but not in neurons. Furthermore, we show that neuronal, not intestinal, AMPK mediates the behavioral response to PMF dissipation. This resulting neuronal PMF loss is communicated primarily by AMPK in serotonergic or dopaminergic neurons and reflected as global search behavior in the presence of food. Last, we characterize whether mitochondrial subcellular localization or fusion-fission dynamics drive these global search behaviors. Together, we show a transtissue signaling axis that coordinates metabolic cues using mitochondria as a central hub to integrate energy perception.

RESULTS

PMF dissipation induces a transition from the local to global search behavior

C. elegans foraging behavior switches from local to global search mode after a period of food deprivation; however, the mechanism required for the behavioral change is unclear (4). Global search is characterized by increased linear forward movement and decreased omega turns and reversals (Fig. 1A). Because food deprivation may limit metabolites for energy production, we first tested if internal metabolic cues resulting from food deficiency reduce the mitochondrial PMF in vivo (fig. S1). We measured the PMF using TMRM and found decreased TMRM fluorescence as a result of food deprivation. As a more direct approach to decrease the PMF and evaluate its role in food-related behavior, we exposed C. elegans to FCCP, a pharmacological protonophore that dissipates the PMF and prevents ATP production by OxPhos. C. elegans were treated for 24 hours to ensure that the drug had been distributed throughout tissues. C. elegans treated with FCCP were removed from food, and their foraging behavior was assessed. We found that FCCP treatment reduced the frequency of omega turns and reversals, recapitulating the time-dependent change from local search to global search (Fig. 1B). To rule out the possibility that FCCP affects the bacteria digested by worms, we used heat-killed bacteria and obtained the same foraging shift to global search upon FCCP treatment (fig. S2A). Together, these data suggest that the PMF acts as a cue to inform the behavioral switch between local and global search.

Fig. 1. PMF dissipation shifts the foraging behavior from local to global search (G-S) in C. elegans.

(A) Schematic of food-related behavior. (B) Wild-type worms were maintained on nematode growth medium (NGM) plates seeded with bacteria in the presence or absence of 10 μM FCCP. After 24 hours, day 1 adult animals were transferred from seeded plates to unseeded plates, and foraging behavior was assessed. Two-tailed unpaired Mann-Whitney test was applied. Data are means ± SD, n = 51 to 56 animals collected in three independent days. (C) Frequencies of omega turns and reversals in worms ubiquitously expressing mtOFF (eft-3 promoter) recorded off-food. Animals were transferred from seeded plates containing bacterial lawn as food to unseeded plates, pre-illuminated for 5 min, and then the foraging behavior parameters were assessed. ATR and light treatments were applied where indicated. Kruskal-Wallis test with Dunn’s test for multiple comparisons was performed. Data are means ± SD, n = 35 to 40 animals in each condition from four independent days.

Behavioral responses are dynamic, and this pharmacologic approach lacked the temporal resolution to dissect signaling pathways and assess cell specificity. To address this, we used the mitochondria-targeted optogenetic tool mitochondria-OFF (mtOFF) (11), which uses light to depolarize the PMF. mtOFF is a genetically encoded light-activated proton pump that unidirectionally pumps protons (fig. S3A). Similar to FCCP treatment, mtOFF activation dissipates the PMF, leading to a decrease in ATP levels and a compensatory increase in ETC activity (11). However, mtOFF can be targeted to specific tissues and PMF dissipation can be acutely activated or deactivated using light, allowing greater spatiotemporal precision. mtOFF activity requires all-trans-retinal (ATR) as a cofactor, and because C. elegans does not synthesize ATR endogenously, the active optogenetic condition requires both ATR supplementation and light (fig. S3B) (11). Similar to prolonged food deprivation and FCCP, mtOFF activation resulted in a reduction in omega turns and reversals (Fig. 1C). Overall, using both a pharmacologic and optogenetic approach, we demonstrate that PMF dissipation is sufficient to modulate foraging behavior.

Intestinal metabolic state controls foraging behavior

In C. elegans, the intestine absorbs and stores nutrients, while the neurons sense the presence or absence of food. Accordingly, we developed strains expressing mtOFF either in the intestine or in neurons for targeted control of PMF dissipation. First, we used animals with either intestinal (vha-6 promoter) or pan-neuronal (rab-3 promoter) mtOFF expression (11) to test if tissue-specific optogenetic dissipation of the PMF affects cellular energy levels. Previously, we reported that ubiquitous optogenetic PMF dissipation increases compensatory oxygen consumption rates (OCRs) and diminishes whole-body ATP levels (11). We first examined the OCR in whole animals expressing mtOFF in either neurons or the intestine. We found that OCR increased upon intestinal mtOFF activation (Fig. 2A) but not upon neuronal mtOFF activation (Fig. 2B). To test the effect of cell-type–specific PMF dissipation on ATP production, we used a bioluminescence assay to determine the whole-animal ATP levels after tissue-specific PMF dissipation. ATP levels upon intestinal PMF dissipation were reduced (Fig. 2C), while neuronal mtOFF activation did not alter whole-animal ATP levels (Fig. 2D). The intestinal decrease in ATP levels and increase in OCR is consistent with PMF dissipation.

Fig. 2. Tissue-specific PMF dissipation decreases the ATP levels in vivo.

(A and B) Relative oxygen consumption rates of whole worms expressing mtOFF either in the intestine (vha-6 promoter) (A) or neurons (rab-3 promoter, pan-neuronal) (B) normalized to the dark condition. Two-tailed unpaired t test was applied. Data are means ± SD, n = 6. (C and D) Relative ATP levels of whole worms expressing mtOFF either in the intestine (C) or neurons (D) measured by bioluminescence assay normalized to dark conditions. Two-tailed unpaired t test was applied. Data are means ± SD, n = 5 to 6. (E) Representative images of neuronal ATeam displayed using Smart LUT. Anesthetized larvae were imaged, illuminated with mtOFF-activating light, and visualized again. ATR supplementation and light activate mtOFF and reduce the yellow fluorescent protein/cyan fluorescent protein (YFP/CFP) ratio showing reduced ATP levels. Scale bars, 10 μm. (F and G) Relative YFP/CFP ratio in worms expressing mtOFF either in the intestine (F) or neurons (G) normalized to dark condition. The YFP/CFP ratio corresponds with ATP levels. Two-tailed unpaired t test was applied. Data are means ± SD, n = 16 to 20. (H) Frequencies of omega turns and reversals in worms expressing mtOFF in the intestine recorded off-food. ATR and light treatments were present where indicated. Kruskal-Wallis test with Dunn’s test for multiple comparisons was performed. Data are means ± SD, n = 42 to 45 animals in each condition from four independent days. (I) Frequencies of omega turns and reversals in worms expressing mtOFF in neurons recorded off-food. ATR and light treatments were applied where indicated Kruskal-Wallis test with Dunn’s test for multiple comparisons was performed. Data are means ± SD, n = 30 to 38 animals in each condition from four independent days. ns, not significant.

In C. elegans, the intestine occupies a larger fraction of body volume than neurons (22). We hypothesized that neuronal mtOFF activation reduces ATP in neurons, but this decrease is not sufficient to affect ATP levels in whole animals. To address this, we used a fluorescence resonance energy transfer (FRET)–based biosensor ATeam (23), which allowed us to monitor ATP levels in situ (Fig. 2E). We expressed ATeam in neurons or the intestine and demonstrated functionality using sodium azide to inhibit the ETC (fig. S4). The ATeam emission spectrum overlaps with mtOFF activation wavelengths. However, control conditions demonstrate that the ATeam signal did not activate mtOFF (fig. S5), suggesting the ability to independently monitor and regulate energy balance in living animals with cell-type specificity. Furthermore, the expression of ATeam and mtOFF driven by the same promoter (vha-6) had no effect on levels of mtOFF (fig. S6). Using ATeam, we revealed decreased ATP levels in the intestine (Fig. 2F) and in neurons (Fig. 2G) upon mtOFF activation. These data confirmed that targeted optogenetic PMF dissipation decreases tissue-specific ATP levels in vivo.

We further tested how the tissue-specific PMF changes would be sensed and reflected in food-related behavior. To address this, we quantified the foraging behavior of acutely food-deprived animals. We found that restricting PMF dissipation to the intestine via tissue-specific mtOFF expression caused a shift to global search behavior (Fig. 2H). Despite the central role of neurons in behavior, neuronal PMF dissipation was not sufficient to induce the switch to global search behavior (Fig. 2I). Although PMF dissipation in the intestine drives foraging behavior, the behavior itself is a coordinated process that may involve signaling between distinct tissues.

Neuronal AMPK is required for PMF-induced global search behavior

Reducing the intestinal PMF is sufficient to switch food search behavior. We next wanted to identify molecular components that signal PMF dissipation. AMPK is considered the cellular energy sensor that couples bioenergetic and physiologic responses to changes in cellular energy status (10, 11, 24). A variety of cellular inputs including changes in the adenosine 5′-monophosphate/ATP ratio, reactive oxygen species (ROS), and PMF loss have been shown to result in AMPK activation (11, 25). We tested if AMPK is required for PMF-induced changes in behavior. We used C. elegans strains ubiquitously expressing mtOFF with a mutated AMPK α-catalytic subunit aak-2, which is an ortholog of human PRKAA1 and PRKAA2. The loss-of-function mutation of AMPK [aak-2(ok524)] suppressed the transition to global search behavior following whole animal PMF dissipation (Fig. 3A and table S1) (11). Unexpectedly, however, given our finding that intestinal PMF was sufficient to initiate the response (Fig. 2H), intestine-specific AMPK re-expression did not rescue this effect (Fig. 3B). Together, this suggested a transcellular signaling axis that may communicate metabolic status to regulate behavior.

Fig. 3. Neuronal AMPK is necessary for PMF signaling of foraging decisions.

(A) Well-fed animals were transferred to food-free plates, and foraging behavior parameters were qualified. Reversals and omega turn frequency in mutant AMPK worms and mutant AMPK worms ubiquitously expressing mtOFF were recorded off-food. ATR and light treatments were applied where indicated. Kruskal-Wallis test with Dunn’s test for multiple comparisons was performed. Means ± SD, n = 52 to 69 animals in each condition from five independent days. (B) Frequency of omega turns and reversals in mutant AMPK worms ubiquitously expressing mtOFF with rescued AMPK in the intestine (vha-6 promoter) recorded off-food. ATR and light treatments were applied where indicated. Kruskal-Wallis test with Dunn’s test for multiple comparisons was performed. Data are means ± SD, n = 20 to 42 animals in each condition from four independent days. (C) Frequency of reversals and omega turns in mutant AMPK worms ubiquitously expressing mtOFF with unc-119–driven pan-neuronal AMPK rescue recorded off-food. ATR and light treatments were applied where indicated. Kruskal-Wallis test with Dunn’s test for multiple comparisons was performed. Data are means ± SD, n = 28 to 52 animals in each condition from four independent days. (D) Frequency of reversals and omega turns in mutant AMPK worms ubiquitously expressing mtOFF with UPN-driven pan-neuronal AMPK rescue recorded off-food. ATR and light treatments were applied where indicated. Kruskal-Wallis test with Dunn’s test for multiple comparisons was performed. Data are means ± SD, n = 36 to 43 animals in each condition from six independent days.

Given the role of C. elegans neurons in sensing food (26, 27), we used two different broadly expressed neuronal promoters, unc-119 and Ultra Pan-Neuronal (UPN) (28), to rescue the expression of AMPK in the aak-2(ok524) strain. We found that both promoters rescued the PMF-induced reduction in reversal frequency compared to light control but did not affect omega turn frequency (Fig. 3, C and D). This suggests that omega turns are not signifying AMPK-mediated energy sensing. These findings overall indicate that the intestinal PMF signaling cascade requires neuronal AMPK to facilitate foraging behavior in a transtissue manner.

Coupled PMF-AMPK signaling in serotonergic and dopaminergic neurons initiates behavioral alterations despite food availability

The PMF is an important cue that transitions behavior from local to global search modalities in the absence of food. We next tested if dissipating the PMF could affect behavior even in the presence of food. C. elegans move slower and stay longer on a food source by decreasing their locomotion measured as the number of body bends on food. We exposed C. elegans to FCCP and found that they exhibited increased body bend frequency on food, consistent with the behavioral effects of food deprivation, the phenotype that we refer to as global search (Fig. 4, A and B). To rule out the effect of FCCP on bacterial food metabolism, we repeated the experiment using heat-killed bacteria and observed an increased body bend frequency (fig. S2B). This is consistent with previous work (11) showing that PMF dissipation increased C. elegans speed despite food availability (fig. S7). Moreover, prior work (11) also revealed that loss of AMPK abolished the PMF-induced global search phenotype, which could subsequently be rescued by pan-neuronal AMPK expression (fig. S7). These results are consistent with the PMF-coupled tissue-specific role of AMPK in regulating foraging behavior autonomously despite food availability.

Fig. 4. AMPK in serotonergic and dopaminergic neurons upon PMF dissipation modulates global search.

(A) Schematic of speed increase of the animal on food initiated by PMF dissipation. (B) Frequencies of body bends on food worms treated with FCCP. Two-tailed unpaired t test was performed. Data are means ± SD, n = 68 animals collected in 3 days. (C) Change of body bends of mutant AMPK worms expressing ubiquitous mtOFF with AMPK rescue in the indicated neurons. The graph shows the optogenetic active condition (+ATR +light) normalized to the mean of the dark condition (+ATR, −light) using the formula: [(+A+L)–mean(−A+L)]/mean(−A+L). The raw data are displayed in figs. S7 and S8. Promoters drive expression as follows: UPN, unc-119, panneuronal; ttx-3, AIY interneurons; flp-15, I2 interneurons; npr-9, AIB interneurons; sra-6, ASH neurons; rig-3, AVA interneurons; gcy-28.d, AIA interneurons; dat-1, dopaminergic neurons; and tph-1, serotonergic neurons. One-way analysis of variance (ANOVA) with Tukey’s test was applied. Data are means ± SD, n = 24 animals collected in 3 days. (D and E) Frequency of reversals and omega turns in mutant AMPK worms ubiquitously expressing mtOFF with rescued AMPK in serotonergic (D) or dopaminergic (E) neurons off-food. Kruskal-Wallis test with Dunn’s test was performed. Data are means ± SD, n = 23 to 29 animals collected in 4 days. (F and G) Change in behavior in response to PMF dissipation. All data collected in four experimental days were used, and the animals that did not perform any omega turn/reversal were excluded. The optogenetic active condition (+ATR, +light) of each worm was normalized to the mean of the corresponding control condition (−ATR, +light) using the formula: [(+A+L)–mean(−A+L)]/mean(−A+L). The raw data are provided in Figs. 1C, 3 [(A) to (D)], and [(D) and (E)]. Kruskal-Wallis test with Dunn’s test was applied.

Because neuronal AMPK is required for PMF-mediated behavioral responses, we next tested which type of neurons is responsible for driving this effect. We screened various subclasses of neurons for their role in AMPK-mediated on-food behavior using cell-type–specific promoters to re-express aak-2, the catalytic subunit of AMPK. First, we confirmed increased speed upon pan-neuronal UPN AMPK rescue in the aak-2(ok524) strain (Fig. 4C and fig. S8), as reported previously using the unc-119 promoter (fig. S7) (11). We then tested whether the re-expression of functional AMPK in interneuron types involved in mitochondrial functions such as ROS signaling (ASH and I2) or food-related behavior (serotonergic, dopaminergic, AIY, AIB, AVA, and AIA) could sufficiently drive on-food behavioral phenotypes. AMPK expression in AIY, I2, AIB, ASH, AVA, and AIA neurons was not sufficient to rescue the global search phenotype. However, AMPK expressed in either serotonergic or dopaminergic neurons rescued the speed deficit (Fig. 4C and fig. S8). Our data suggest a connection between AMPK and serotonergic and dopaminergic neuronal circuits, coupling specific neurotransmitter neuron classes and mitochondrial energy perception to control animal behavior.

Neurotransmitters serotonin and dopamine have been previously shown to regulate food-related behaviors in C. elegans (29–31). We observed the global search phenotype response upon AMPK rescue in either serotonergic or dopaminergic neurons characterized as an increase in the body bends frequency in the presence of food (Fig. 4C). We next tested whether PMF dissipation in these cell types alone would be sufficient to induce the off-food foraging phenotype, observed previously using pan-neuronal AMPK rescue (Fig. 3, C and D). We found that AMPK rescue in either serotonergic or dopaminergic neurons did not restore the PMF-induced reduction in reversal frequency (Fig. 4, C and D). To compare the behavioral response to mtOFF activation between genotypic strains, we first normalized the frequency of reversals and omega turns when mtOFF was activated (+ATR +light) to a control condition where mtOFF was not activated (−ATR +light). This metric accounted for any change in baseline behavior due to genotypic variation and allowed for direct comparisons of the effect of mtOFF activation between mutant or rescue genotypic strains relative to mtOFF activation. The mtOFF effect on foraging behavior was similar when AMPK was rescued in either serotonergic or dopaminergic neurons (Fig. 4, F and G). Together, comparing the relative change revealed mtOFF activation and rescuing AMPK function in the intestine largely returned this local search response (Fig. 4, F and G). Our data indicate that PMF-induced body bends and local search behavior parameter (omega turns and reversals) alterations are communicated via different neuronal classes. Together, these data show that activation of neuronal AMPK initiated by PMF loss is sufficient to change behavior in the presence of food, but not in an energy-deprived environment.

Balanced mitochondrial dynamics are essential for a food-related behavior

Mitochondria are dynamic and alter morphology in response to changes in metabolites and cellular stress. Moreover, phosphorylation by AMPK triggers the fission machinery (17, 18). Therefore, we tested the role of mitochondrial fission in facilitating food-related behavior. Using a loss-of-function drp-1(tm1108) mutant, we observed that loss of mitochondrial fission prevented PMF-dependent increase in body bends of fed worms (Fig. 5A). Similarly, we found that PMF dissipation did not affect reversal (Fig. 5B, right), or omega turns frequencies (Fig. 5B, left). Our data indicate that fission is important in PMF sensing in food-related behaviors because the food-searching behavior induced by PMF dissipation was blocked by the loss of mitochondrial fission. Next, we examined the relationship between AMPK and mitochondrial fission using the AMPK activator 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR). We found that AICAR treatment in DRP-1 mutant worms did not affect either body bends on food or reversal and omega turn frequency (fig. S9), suggesting that AMPK activation is upstream of fission, which is consistent with the role of AMPK in inducing fission (17, 18).

Fig. 5. Mitochondrial dynamics regulate the foraging behavior of C. elegans.

(A) Frequencies of on-food body bends of fission mutant. One-way ANOVA with Tukey’s test was used. Data are means ± SD, n = 24 animals collected in 3 days. (B) Frequencies of reversals and omega turns in fission mutant recorded off-food. Kruskal-Wallis test with Dunn’s test was performed. Data are means ± SD, n = 29 to 39 animals from 4 days. (C) Frequencies of on food body bends of fusion mutant. One-way ANOVA with Tukey’s test was used. Data are means ± SD, n = 24 animals from 3 days. (D) Frequencies of reversals and omega turns in fusion mutant recorded off-food. Kruskal-Wallis test with Dunn’s test was performed. Data are means ± SD, n = 24 to 40 animals from 4 days. (E) Change in body bends upon PMF dissipation. All data collected in three experimental days were used. The optogenetic active condition of each genotype (+ATR, +light) was normalized to the mean of the corresponding control condition (−ATR, +light). using the formula: [(+A+L)–mean(−A+L)]/mean(−A+L) from the raw data provided in (A) and (C) and fig. S7. One-way ANOVA with Tukey’s test for multiple comparisons was applied to compare the normalized data. (F) Change in foraging reversals/omega turns upon PMF dissipation. All data collected in four experimental days were used, and the animals that did not perform any reversals/omega turns were excluded. The optogenetic active condition of each genotype (+ATR, +light) was normalized to the mean of the corresponding control condition (−ATR, +light) using the formula: [(+A+L)–mean(−A+L)]/mean(−A+L) from the raw data provided in Fig. 1C and (B) and (D). Data from Fig. 1C (ubiquitous mtOFF) are reproduced for comparison purposes. Kruskal-Wallis test with Dunn’s test for multiple comparisons was applied to compare the normalized data.

Fusion is favored in conditions of acute food deprivation and can be associated with increased OxPhos activity (32–34). We hypothesized that fusion may be involved in the cellular sensing of PMF dissipation. To test this, we crossed C. elegans with mutated fzo-1(tm1133), the ortholog of human MFN2 (mitofusin 2), with worms expressing mtOFF ubiquitously. Without functional mitochondrial fusion, mtOFF activation failed to change body bend numbers (Fig. 5C) or omega turn and reversal frequencies (Fig. 5D). Accordingly, the increase in body bends due to PMF dissipation was lost in the absence of mitochondrial dynamics (Fig. 5E). However, foraging behavior induced by PMF dissipation was affected more through loss of mitochondrial fusion than mitochondrial fission (Fig. 5F). Notably, baseline omega turn and reversal frequencies in the fusion mutant were lower (fig. S10), suggesting that loss of mitochondrial fusion abolishes the ability for worms to sense food regardless of PMF dissipation. Collectively, our results show that the imbalance in mitochondrial dynamics prevents the ability of PMF dissipation to signal food availability.

Intact mitochondrial trafficking in neurons is required for sensing the PMF-induced change in global search behavior

Mitochondrial dynamics are important for trafficking mitochondria into distal regions of the neurons (35) and correct mitochondrial localization is critical for neuronal function and signaling (36). Neurons have unique mitochondrial morphologies depending on subcompartment localization, indicating altered function. Therefore, we tested whether neuronal mitochondrial localization was necessary for PMF-induced behavioral alterations. We used the C. elegans ric-7(n2657) mutant, which lacks an adapter protein required for the anterograde transport of neuronal mitochondria from the cell body to axons (Fig. 6A) (37). Without mitochondria in the axon, we observed no change in speed on food in response to PMF dissipation (Fig. 6B). We next expressed the chimeric protein Mitotruck that links the kinesin-1 transport protein with the outer membrane protein Tom7 to restore mitochondrial transport to axons (37). We found that the PMF-induced global search phenotype was rescued by expressing Mitotruck (Fig. 6 C, D), confirming the importance of energy production in the axons for this behavioral response. Worms with rescued mitochondrial trafficking were slower than control animals at baseline (Fig S10). Overall, we found somatic mitochondria are not sufficient to communicate the PMF-induced energy state to the body.

Fig. 6. Intact transport of axonal mitochondria is essential for PMF-induced speed elevation in fed C. elegans.

(A) Schematic showing the ric-7 mutation blocks mitochondrial transport leading to the lack of mitochondria in axons. (B and C) Frequencies of on food body bends of worms with mutated ric-7(n2657) (B) and with rescued mitochondrial transport to axons (C). One-way ANOVA with Tukey’s test was used. Data are means ± SD, n = 24 animals from 3 days. (D) Change in body bends upon PMF dissipation. All data collected in three experimental days were used. The optogenetic active condition of each genotype (+ATR, +light) was normalized to the mean of the corresponding control condition (−ATR, +light) using the formula: [(+A+L)–mean(−A+L)]/mean(−A+L) from the raw data provided in (B) and (C) and fig. S7. Data from fig. S7 (ubiquitous mtOFF) are reproduced for comparison purposes. One-way ANOVA with Tukey’s test was applied. (E) Frequency of reversals and omega turns in ric-7(n2657) mutant recorded off-food. Kruskal-Wallis test with Dunn’s test was performed. Data are means ± SD, n = 23 to 31 animals from four independent days. (F) Frequencies of reversals and omega turns of mutant ric-7(n2657) with rescued mitochondrial transport to axons recorded off-food. Kruskal-Wallis test with Dunn’s test was performed. Data are means ± SD, n = 20 to 32 animals from four independent days. (G and H) Change in omega turns/reversals upon PMF dissipation. All data collected in four experimental days were used, and the animals that did not perform any omega turns/reversals were excluded. The optogenetic active condition of each genotype (+ATR, +light) was normalized to the mean of the corresponding control condition (−ATR, +light) using the formula: [(+A+L)–mean(−A+L)]/mean(−A+L) from the raw data provided in Fig. 1C and (E) and (F). Data from Fig. 1C (ubiquitous mtOFF) are reproduced for comparison purposes. Kruskal-Wallis test with Dunn’s test was applied.

We demonstrated the importance of axonal mitochondria in signaling behavior in the presence of food, and we next asked whether the neuronal localization of mitochondria is required in the absence of food. Consistent with the body bend results, we did not observe an effect of PMF on local search behavior in worms lacking axonal mitochondria (Fig. 6E). However, the lack of axonal mitochondria reduced the foraging behavior at baseline, indicating disrupted food sensing in this background (fig. S10). Furthermore, expressing Mitotruck was not sufficient to restore the PMF-induced reversal and omega turn frequency pattern (Fig. 6F). Accordingly, the PMF-induced change in omega turns (Fig. 6G) was smaller in worms with disrupted and rescued mitochondrial trafficking. Reversal frequency did not change in either strain upon PMF dissipation (Fig. 6H). Overall, our data highlight the mitochondrial PMF as an important mediator in communicating the energy status to affect behavior.

DISCUSSION

C. elegans explore their environment for food, the availability of which informs foraging behaviors that influence the probability of finding and staying on a food source (10). We optogenetically dissipated PMF to identify signaling pathways through which this process is entrained. Using a combination of genetic and pharmacologic techniques, our results demonstrate how alterations in the tissue-specific metabolic state modulate foraging behavior in C. elegans. The transition between local and global search foraging behavior, in which worms increase their search range in the absence of food, was found to be selectively driven by the PMF in the intestine. Moreover, neuronal perception of this internal metabolic state was shown to involve AMPK and axonal trafficking of mitochondria.

Our data provide insight into how tissue-specific PMF dissipation affects the energy state in C. elegans. The PMF has many roles in the cell and affects a wide range of cellular processes. Rapid fluctuations in the PMF have been demonstrated as signals affecting dendritic spine stabilization in primary hippocampal neurons (38). Moreover, using the yeast Saccharomyces cerevisiae, PMF dissipation has been associated with cell cycle regulation, affecting the orchestration of the cell cycle (39). These findings highlight the importance of the PMF in dictating cellular energy status and various downstream outputs. The wide range of PMF-dependent cellular outputs is most likely dependent on the cell type and the intracellular location of mitochondria. Accordingly, a recent study (40) demonstrated that individual C. elegans neurons had distinct energetic needs and capabilities to meet their cellular energy demands. Overall, precise control of the PMF will be important to fully dissect cause and effect relationships in energy sensing pathways.

We showed that acute intestinal PMF dissipation is sufficient to control the foraging behavior of acutely food-deprived animals (Fig. 2H). PMF dissipation in the intestine affects bioenergetic parameters in the whole body (Fig. 2, A and C) because the intestine occupies approximately 35% of the body volume compared to 1% taken up by neurons (22). Behavioral states may be initiated by acute energy deficiency in the intestine sensed by the AMPK pathway in neurons (and possibly other tissues) as AMPK mutation is capable of switching behavior (Fig. 3A) (11). Alternatively, we show a possible cross-talk between the intestinal metabolic state and neuron-integrated behavioral response. Accordingly, it has been reported that the intestinal insulin-like peptide INS-31 may transmit the low-energy state of the intestine to other tissues, including neurons (6). Recent data also showed that food uptake regulates on-food behavior via neuronal (homolog of human cytokine transforming growth factor-β) daf-7 expression in C. elegans (41). Neuronal circuits are important in integrating metabolic states and behavioral changes. However, information on how organisms integrate sensed internal metabolic states and food decisions as well as behavioral changes is still unfolding (42–44). Because acute neuronal PMF dissipation does not significantly affect the behavioral response off food (Fig. 2I), we hypothesize that a prolonged PMF dissipation may be required. The majority of neuronal ATP production is provided by mitochondrial OxPhos, and the loss of mitochondrial function is associated with cell death and neurodegeneration. Direct control of the PMF is sufficient for survival in response to ETC dysfunction (45) and models of neurodegeneration (46).

Having demonstrated the role of intestinal PMF in AMPK-mediated foraging behavioral shift, we further tested the tissue-specific impact of AMPK. We re-expressed AMPK via an extrachromosomal array, which may not match endogenous AMPK expression levels in each cell type tested. Using this standard approach, we revealed that, neuronal, not intestinal, AMPK rescued the PMF-induced global search phenotype (Fig. 3 C, D). Supporting our findings, neuronal AMPK has been reported to modulate neuronal activity to signal corresponding food-related responses in C. elegans (10, 11) and in Drosophila melanogaster (47).

Here, we show that PMF dissipation enhances speed and initiates global search foraging behavior by altering the internal metabolic state. Intriguingly, we found that AMPK re-expressed in serotonergic or dopaminergic neurons is sufficient to restore the global search behavior. While previous studies have reported the role of serotonergic and dopaminergic neurons in modulating food behavioral responses (43, 44, 48–52), we have shown that AMPK activation in these neurons is also required. Overall, we have shown that PMF dissipation reduces ATP production (Fig. 2, C, F, and G) and is mediated by AMPK signaling (Fig. 3A). However, we are unable to distinguish whether AMPK is activated by PMF dissipation directly or via other signals such as ATP decline or possibly ROS or calcium ions.

Mitochondria respond to changes in metabolic state, and this results in alterations in the mitochondrial dynamics, trafficking, and function, necessary for maintaining energy homeostasis. Mitochondrial shape and function are closely linked with internal metabolic state. The mitochondrial morphology is dynamically regulated by a balance between fission and fusion, which is crucial for energy metabolism (53). Moreover, AMPK recruits DRP-1 by phosphorylating mitochondrial fission factor (17, 18). Thus, our data indicate that DRP-1 is necessary for mitochondrial fission downstream of AMPK activation upon PMF dissipation. Mitochondrial fission is essential for mitochondrial quality control and mitophagy, whereas mitochondrial fusion is required for mitochondrial repair, renewal, and improved mitochondrial network connectivity (18, 54, 55). Our study shows that the dissipation of the PMF activates both mitochondrial fission and fusion machinery downstream of AMPK. Reports show that an overall balance in mitochondrial fission and fusion processes is more important for energy homeostasis than the preponderance of either one (56).

Mitochondrial fission is crucial for the maintenance of axonal mitochondria (57, 58). Moreover, mitochondrial distribution and the presence of mitochondria in axons are required for synaptic plasticity, the release of neurotransmitters, and ROS signaling (37, 42, 59). Consistent with these reports, we show that axonal mitochondria are required for AMPK-mediated perception of internal metabolic state and signaling global search response. Furthermore, AMPK activation is region-specific. Activation in the soma does not alter mitochondrial transport within the spatially isolated distal axon, indicating that axons can sense and modulate response to local energetic perturbations (60). Overall, these findings highlight the theme that AMPK senses the mitochondrial metabolic state and that mitochondrial dynamics and distribution are important in response to the low-energy state.

We focused on behavioral changes induced by PMF dissipation to characterize intertissue communication of the low-energy state. Our findings revealed that the PMF acts as a key element in the regulation of energy-related behaviors. These experiments show that mitochondrial status can influence behavioral decisions. PMF dissipation reduces internal energy availability and shifts the behavior toward starvation phenotypes in the absence and presence of food. Overall, our work links mitochondrial function to feeding habits and solidifies the role of mitochondria as the central signaling hub coordinating metabolic decisions.

MATERIALS AND METHODS

Experimental design

The main objective of the study was to investigate the role of mitochondrial energy deficiency in neurons and the intestine in C. elegans behavior. We used an optogenetic model called mtOFF expressed either in the intestine or in neurons. Upon activation by light and ATR supplementation, mtOFF reduces the metabolic energy state by dissipating the PMF. Using various mutant C. elegans strains, we then tested the association of optogenetic PMF dissipation with other energy-dependent components of metabolism. Using an in vitro bioluminescence assay, an in situ ATeam biosensor probe, and in vivo with a Clark-type electrode, we determined the bioenergetic parameters in worms expressing tissue-specific mtOFF. We then quantified how these bioenergetic changes and mutations manifest in foraging behavior.

C. elegans maintenance and development of transgenic strains

Worms were grown on nematode growth medium (NGM) seeded with OP50 Escherichia coli lawn as a food source at 20°C. Where indicated, we heat-killed the OP50 by boiling for 10 min. After 10 min, the bacteria were allowed to cool before plating. ATR was supplemented, where indicated, through E. coli food at a final concentration of 100 μM as previously published (61). The mtOFF strains (ubiquitous, intestinal, and neuronal expression) were generated using the mmCRISPi method (11, 62). Transgenic C. elegans strains used in this study were generated by microinjection of DNA plasmids, as previously described (63). The complete list of strains is available in table S1.

ATP measurement–bioluminescence assay

Synchronized day 1 adults on seeded NGM plates (+/−ATR) were exposed to light (2.1 mW/cm² 590-nm light-emitting diode (LED) array with STOmk-II stimulator by Amuza) for 4 hours at 20°C. The control +/−ATR groups were kept in the dark at 20°C. The worms were then collected into 1.5-ml tubes using M9 buffer and centrifuged (1 min, 2000g), the supernatant was removed, and the washing was repeated. The samples were immersed three times in liquid nitrogen and then boiled for 15 min and left on ice for 5 min. Last, the samples were centrifuged (14,800g, 10 min), and the supernatants were used to measure ATP and protein levels. Protein concentration in the samples was determined using the Folin-phenol method. ATP bioluminescence kit (Invitrogen Molecular Probes, A22066) was used to measure the ATP levels in the samples according to the manufacturer’s instructions. The relative fold change in ATP levels is shown after normalization to control dark groups.

Oxygen consumption measurement

Oxygen consumption of synchronized day 1 whole worms was determined using a Clark-type oxygen electrode (S1 electrode disc, DW2/2 electrode chamber, and Oxy-Lab control unit, Hansatech Instruments, Norfolk, UK). Worms maintained either with or without ATR were transferred to 1.5-ml tubes using M9 buffer and collected by centrifugation (2 min, 2000g). Worms were washed once more in M9 buffer and moved to the chamber containing 0.5 ml of continuously stirred M9 buffer. After basal respiration measurements, worms were exposed to light for 2 min (0.39 mW/mm², 540- to 600-nm GYX module, X-Cite LED1 by Excelitas, Waltham MA). FCCP was then added at a final concentration of 160 μM to achieve maximal respiration. Last, sodium azide was added at a final concentration of 40 mM to quantify nonmitochondrial oxygen consumption. After each measurement, the whole sample was collected and the protein concentration was determined using the Folin-phenol method.

Confocal microscopy

Staged L3-L4 expressing C. elegans–optimized ATeam (64) were anesthetized with 10 mM tetramisole in M9 buffer and placed on a 2% agar pad. For the control condition, sodium azide was added to the solution at a final concentration of 10 mM. A Nikon A1R HD microscope with 10× air objective for worms expressing ATeam in the intestine and to compare intestinal mtOFF expression between strains and 20× air objective for worms expressing ATeam in neurons was used to visualize the worms. ATeam was imaged at a single excitation (405 nm) and two emissions (485 and 538 nm). To use ATeam, worms were imaged, exposed to the epi-fluorescent light (74.2 mW/cm²) for 5 min, and imaged again. Images were analyzed using Fiji ImageJ software. The regions of interest were drawn around either the intestine or neuronal ring based on ATeam/mtOFF expression to quantify the intensity of fluorescence. Subsequently, the percentage difference between the fluorescence intensity before (baseline signal) and after illumination was calculated. The background signal was subtracted manually.

Off-food foraging behavior assay

Well-fed synchronized day 1 and day 2 adult worms (+/−ATR) were transferred to unseeded plates. Then, one plate from each group (+/−ATR) was pre-illuminated with default microscopic white light for 5 min, and one plate from each group (+/−ATR) was pre-illuminated with mtOFF activating light (0.14 mW/mm², 540- to 600-nm GYX module, X-Cite LED1 by Excelitas, Waltham MA) as well. Videos were recorded for 6 min at the same illumination settings. To measure locomotion upon drug exposure, L4 worms were transferred 24 hours before the experiment to seeded NGM plates containing 10 μM FCCP or 100 μM mdivi-1. For AICAR exposure, day 1 adults were moved to the seeded NGM plate containing 1 mM AICAR 4 hours before the experiment. Subsequently, the number of omega turns and reversals regardless of their direction and length was manually counted. Worms that were in the scene for less than 1 min or did not move for more than 1 min were excluded from the analysis.

On-food locomotion assay–body bends

Synchronized day 1 and day 2 adult worms (+/−ATR) were transferred to the freshly seeded plate, and the number of body bends per 15 s was counted. One plate from each group (+/− ATR) was exposed to default white microscopic light, and the rest of the worms were additionally treated with mtOFF-inducing light (0.265 mW/mm², 540- to 580-nm excitation filter MVX10 Fluorescence MacroZoom dissecting microscope by Olympus powered by an X-Cite 220 V mercury bulb by Excelitas). To measure locomotion upon drug exposure, L4 worms were transferred 24 hours before the experiment to seeded NGM plates containing 10 μM FCCP or 100 μM mdivi-1. For AICAR exposure, day 1 adults were moved to the seeded NGM plate containing 1 mM AICAR 4 hours before the experiment.

In vivo ΔΨm measurement

Staged L4 were incubated with 100 nM TMRM and 12 μM MitoTracker Green FM on seeded NGM plates for 24 hours. Worms were then washed three times with M9 buffer and centrifuged for 1 min at 1000g, the supernatant was discarded between these wash steps to clear worms of residual dye. The worm pellet was then transferred to seeded and unseeded plates 1 hour before imaging and incubated at 20°C. For imaging, worms were anesthetized in 10 mM tetratmisole and mounted on 2% agarose pads. Texas Red and green fluorescent protein filter sets were used to record images on an epifluorescence microscope (Zeiss microscope equipped with Colibri 7 LED and Axiocam 705 camera). TMRM fluorescence intensity was assessed using a 10× air objective with excitation and emission wavelengths set at 548 and 573 nm, respectively. Data were analyzed by drawing regions of interest around the head of each individual worm using Fiji ImageJ software. Background fluorescence was subtracted.

Statistics

Data were analyzed using GraphPad Prism (9.5.0.). Data are presented as means ± SD. D’Agostino-Pearson test was used to test the normality of the data. Nonparametric tests were used for all frequencies of omega turns and reversals because most of the data did not pass the normality test. To compare behavioral data quantitatively among strains, all nonzero values for omega turns and reversals were considered when calculating relative change to the mtOFF metric. Specific statistical tests are described in the figure legends. Statistical significance is defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Supplementary Materials

This PDF file includes:

Figs. S1 to S10

Table S1