Obesity is a leading preventable cause of illness and premature death in the United States and much of the developed world (Jia and Lubetkin, 2010). The obesity epidemic is driven in large part by overconsumption of highly palatable food beyond metabolic requirement (Hill et al., 2003; Finkelstein et al., 2005; Swinburn et al., 2009; Kenny, 2011). Increasingly, it is clear that glucose metabolism plays a major role in the acquisition of the rewarding properties of food, independent of taste, and caloric content (de Araujo et al., 2008; Sclafani and Ackroff, 2012). Since brain reward systems constitute a key component in regulating feeding behavior (Saper et al., 2002; Lutter and Nestler, 2009), understanding the neurobiological circuitry underlying food-induced reward is of considerable interest.

Serotonin and dopamine signaling regulate a range of food-related behaviors including food intake across species. In humans, drugs that increase central serotonergic signaling are approved for weight loss (Weintraub et al., 1984; Hurren and Berlie, 2011), whereas decreasing central serotonergic signaling promotes metabolic syndrome and accumulation of central adiposity (Halder et al., 2007). Serotonin mediates a range of food-related behaviors in C. elegans, including activating two feeding motions (pharyngeal pumping and isthmus peristalsis) that regulate intake (Song and Avery, 2012), slowing locomotion, increasing egg-laying rate (Horvitz et al., 1982), and regulating body fat (Srinivasan et al., 2008; Noble et al., 2013). It has been suggested that each of these distinct behaviors is likely mediated by serotonergic signaling via independent (yet related) mechanisms or pathways (Noble et al., 2013), with more and more studies supporting this claim. For example, tryptophan hydroxylase (tph-1) functioning in ADF neurons signaling through the serotonin subtype-5 G-protein-coupled receptor (GPCR) regulates pump rate (and thus food intake; Cunningham et al., 2012). Tph-1 signaling through ADF neurons activates the serotonin-gated chloride channel MOD-1 on AIY and AIZ neurons to modulate aversive olfactory learning (Zhang et al., 2005) and on URX body cavity neurons to decrease adiposity (Noble et al., 2013). Although food-induced reward plays a key role in maintaining feeding behavior, the role of serotonin in mediating food-induced reward is not known. The present studies demonstrate a key role for serotonin acting via 5-HT4 receptor in food-induced reward.

The dopaminergic system also regulates feeding behavior and food reward – palatable food increases levels of dopamine in the nucleus accumbens (NAc; Hernandez and Hoebel, 1988a,b), and the D2 receptor agonist bromocriptine, recently approved to treat type 2 diabetes, decreases food intake (Thanos et al., 2011) and causes weight loss by unknown mechanisms (Meier et al., 1992). Furthermore, antipsychotic D2 antagonists constitute a major cause of diabetes and obesity in this patient population (American Diabetes Association et al., 2004). Despite the major roles these neurotransmitters play in regulating intake, the mechanisms mediating these phenomena remain unknown.

With a nervous system composed of 302 neurons, a short generation time, and easily observable feeding-related behaviors, the model organism C. elegans affords an excellent opportunity to study the mechanisms underlying food-induced reward. As in mammalian neuroendocrine regulatory systems, C. elegans feeding behaviors are controlled by serotonin and dopamine signaling (Sze et al., 2000; Hills et al., 2004). Mechanisms regulating energy balance – including those involved in metabolic pathways, adiposity storage and utilization, and insulin signaling – are also conserved (Kimura et al., 1997; Ogg and Ruvkun, 1998; McKay et al., 2003; Morton et al., 2006; Mullaney and Ashrafi, 2009). Therefore the present study used food-induced odor preference (Kauffman et al., 2010) to investigate the role of glucose metabolism, as well as serotonin and dopamine signaling in the development and modulation of food-induced reward.

Brain reward systems activated by highly palatable foods are major regulators of feeding behavior (Saper et al., 2002; Lutter and Nestler, 2009; Zheng et al., 2009; Shomaker et al., 2010; Kenny, 2011). Given the growing obesity and type 2 diabetes epidemics, the mechanisms by which reward from highly reinforcing foods – including glucose – increase consumption is of considerable interest. Furthermore, although both serotonin and dopamine play a role in neuroendocrine signaling and regulate feeding behaviors across species, the underlying molecular mechanisms remain unknown. This paper reports several novel observations regarding molecular mechanisms underlying the re-enforcing effects of food in C. elegans. First, we demonstrate for the first time that glucose is rewarding to C. elegans, based on the rapid acquisition of preference of a neutral odor when paired with glucose. Second, we demonstrate that serotonin mimics glucose and food-induced reward, acting through 5-HT4 receptor. We also demonstrate for the first time that dopamine also mimics reward produced by glucose and that the D2 receptor agonist reduces this reward. Of particular interest we report that both food withdrawal and antagonism of glucose metabolism prior to pairing enhance food-induced reward.

Not all food sources signal similarly as nutrient signals. Foods high in sugar are particularly reinforcing (Berthoud et al., 2011), and glucose is particularly rewarding compared to other sugars, such as fructose, as well as other nutrient sources, such as the amino acid serine (Ren et al., 2010; Sclafani and Ackroff, 2012). Studies completed in our lab (data not shown) have demonstrated that similarly to glucose the amino acid leucine increases chemotaxis after training, whereas serine does not, demonstrating that, as in mammals, in C. elegans not all nutrients are equally reinforcing. Here we demonstrated that novel finding that glucose increases odor preference similar to OP50, supporting the hypothesis that the reinforcing effects of glucose are conserved across species.

Hunger increases reinforcing effects of food (i.e., “hunger is the best sauce”). For example, fasting increases ventral striatum activation after consumption of highly palatable food in humans (Goldstone et al., 2009). Similarly, nutritional deprivation enhanced food-induced odor preference in C. elegans (Figure 1D). 2DG induces hunger in humans (Thompson and Campbell, 1977) and other mammals (Bergen et al., 1996), possibly by inhibiting neuroendocrine sensing of glucose. Similarly, in C. elegans 2DG in the presence of OP50 enhanced food-induced odor preference (Figure 1D) similar to nutritional deprivation, indicating that glucose sensing also plays a role in mediating effects of nutritional state on reinforcing properties of food.

When nutritional resources are abundant, insulin-like ligands signaling through DAF-2 leads to DAF-16 phosphorylation and its consequent sequestration in the cytoplasm (Henderson and Johnson, 2001; Lee et al., 2001; Lin et al., 2001). Conversely, when food is scarce, DAF-2 signaling decreases, leading to DAF-16 translocation to the nucleus and subsequent gene expression (Ogg et al., 1997; Honda and Honda, 1999; Henderson and Johnson, 2001; Garsin et al., 2003; Libina et al., 2003). Since genetic inactivation of daf-2 enhanced, whereas inhibition of daf-16 inhibited, food-induced odor preference, we conclude that reduced activity of insulin-like signaling mediates at least some of the effect of nutritional deprivation to increase the reinforcing properties of food (Figures 1E,F). These findings support those previously demonstrated by Kauffman et al. (2010); however in contrast to these authors’ findings that daf-2 and WT animals learn at a similar rate when trained after a fast, here we show that in the absence of fasting prior to training, daf-2 mutants exhibit enhanced odor preference development compared to WT animals. This finding indicates that the daf-2 mutation mimics starvation – this effect of enhanced odor preference development compared to WT animals is perhaps then masked when both groups are fasted prior to conditioning.

Increasing central serotonergic signaling promotes satiety and weight loss in humans (Weintraub et al., 1984; Hurren and Berlie, 2011) and non-human mammals (Edwards and Stevens, 1991; Vickers and Dourish, 2004), whereas impaired serotonin signaling causes hyperphagia and obesity in a 5-HT_2c knockout mouse model (Tecott et al., 1995). In C. elegans, serotonin is known to signal the presence of food – similarly to being exposed to the worm food source OP50 bacteria, exogenous serotonin slows locomotion and increases pump and egg-laying rate (Horvitz et al., 1982; Srinivasan et al., 2008). Interestingly, tph-1 expression only in ADF neurons supports aversive but not attractive learning, while tph-1 expression in both NSM and ADF neurons supports both aversive and attractive learning, suggesting that NSM neurons play a role in “evaluating attractive components of food-related signals” (Zhang et al., 2005).

Our data indicate a role for serotonin released from NSM neurons in food-induced odor preference (Figure 2). Anatomically, NSM neurons are well suited to convey the presence of nutritional resources through neuroendocrine signaling. NSM neuron sensory endings extend into the pharyngeal lumen and NSM axons extend into the pseudocoelom where released molecules can signal in an endocrine fashion (Albertson and Thomson, 1976; White et al., 1986; Altun and Hall, 2009). Furthermore, NSM neurons are secretory and modulate feeding behavior (Avery and Horvitz, 1989). Since genetic inactivation of ocr-2, which blocks serotonin synthesis in ADF neurons, and inactivation of mod-5, which blocks serotonin reuptake in AIM and RIH neurons, fail to block food-induced odor preference, whereas inactivation of unc-86, which eliminates serotonin synthesis in NSM and HSN neurons, blocks food-induced odor preference, these data are consistent with the hypothesis that food-induced odor preference may be mediated by serotonin released from NSM neurons (Figures 2D–F). Furthermore, 5-HT4 receptor appears to be the key receptor mediating these effects of serotonin (Figure 2F). More studies utilizing cell-specific reconstitution are needed to validate these hypotheses.

Dopamine is also a key regulator of feeding behavior (Hernandez and Hoebel, 1988a,b). In mice, ablation of tyrosine hydroxylase and thus dopamine produces deficits in both motor output and motivated behavior, and animals die due to failure to initiate feeding (Zhou and Palmiter, 1995; Szczypka et al., 2000). Delivery of the TH gene primarily to the ventral striatum, which regulates reward, rescues lethality and restores feeding without improving motor deficits (Szczypka et al., 1999). Conversely, the dopamine D2 receptor agonist bromocriptine, approved for use to treat diabetes because of its ability to reduce hyperglycemia in individuals with type 2 diabetes mellitus (Barnett et al., 1980), decreases adiposity in both humans (Meier et al., 1992) and other mammals (Cincotta et al., 1991, 1997). As in mammalian systems, in C. elegans dopamine acts through D1- and D2-like receptors that signal through orthologs of the major mammalian G-proteins (Gaq and Gao, respectively) to direct locomotion (Jansen et al., 1999; Suo et al., 2002; Chase et al., 2004; Allen et al., 2011). Since exogenous dopamine, like OP50, increased odor preference, while exogenous bromocriptine blocked both food- and glucose-induced odor preference (Figures 3A–C), we conclude that D1- and D2-like signaling antagonistically drive food-induced reward, with predominant activation of lower affinity D1-like receptors enhancing reward when dopamine levels are high, and predominant activation of higher affinity D2-like receptors decreasing reward when dopamine levels are low (Avery and Horvitz, 1989; Missale et al., 1998; Marcellino et al., 2012).

Genetic inactivation of cat-1, which decreases synaptic dopamine and serotonin function (Sulston et al., 1975; Duerr et al., 1999), inhibited food-induced odor preference, whereas partial inactivation of cat-2, leading to deficiency in dopamine alone, failed to inhibit food-induced reward statistically, although clearly there was a trend toward reduced reward (Figures 3D,E). These observations support that serotonin and dopamine signaling function in parallel to mediate food-induced odor preference. Furthermore, since nutrition deprivation enhanced odor preference produced by food, dopamine, and serotonin similarly, we hypothesize that nutrition deprivation enhances learned odor preference through a post-synaptic mechanism (Figure 3F). However, this hypothesis is subject to the caveat that the synaptic localizations of these receptors have not been definitively demonstrated in C. elegans.

Extending previous studies (Tsalik and Hobert, 2003; Luedtke et al., 2010), these results are consistent with a basic model mediating food reward in C. elegans. In the presence of an odor, AWC neurons are inactive, and signaling by downstream interneurons increases the dwell time near a food source (Chalasani et al., 2007). Also during exposure to food, we hypothesize that food- or glucose- sensing causes NSM neurons to humorally release serotonin, which signals through 5-HT4 receptor expressed on the RIB interneuron (Tsalik et al., 2003), in turn acting on AVA neurons to modulate ventral motor neurons, leading to forward and backward movement keeping the animal close to the odor or food source (future cell-specific studies would be necessary to validate this proposed circuit). The D1-like receptor is also expressed on the RIB interneuron (Sanyal et al., 2004), plausibly enhancing the activity of this neuron by dopamine. When an odorant is removed, AWC neurons become activated leading to downstream ventral motor signaling that induces turning behavior (Chalasani et al., 2007), which is a main feature of local area search behavior. Removal of a food source would likely correspond to odorant removal – during this time we suggest that hunger enhances activity at 5-HT4 receptor and D1-like receptors on the RIB interneuron (further cell-specific studies would be necessary to validate this hypothesis). In mammals, D1- and D2-like receptors are in different populations of neurons in the nucleus accumbens. D2-like receptors are present on the RIC interneuron (Suo et al., 2009). RIC interneurons receive input from CEP mechanosensory neurons concerning the presence of bacteria (White et al., 1986; Sawin et al., 2000) and synapse on to the AVA command interneuron which modulates ventral motor neurons. Alternatively, CEP may signal through the D2-like receptor expressing SIA interneuron (Suo et al., 2009), which lacks synaptic output but does form gap junctions with the RIB interneuron.

Food-induced reward is regulated by a complex interaction of environmental inputs, such as nutrient availability, and internal inputs, such as nutrient selection, current nutritional state, neuroendocrine signaling, and prior experience (Zheng and Berthoud, 2008). Determining how serotonin and dopamine receptor signaling regulate food-induced reinforcement is more tractable in a nervous system containing 302 neurons than in mammalian neuronal systems. Investigating this circuitry in C. elegans affords an opportunity to study a simple form of food reward learning that is easily observable on the level of the whole organism. For example, genes regulating food-induced reward may be identified by high-throughput RNAi screens. Because of the known conservation of the many pathways regulating adiposity and food intake between mammals and C. elegans, studies such as these can lead to better, and more specific, therapeutic interventions that modulate food-induced reward.

While the present studies are informative, a number of caveats are pertinent to the interpretation of these results, and future studies should clarify these caveats. For example, future studies would carry out more extensive dose–response relationships to allow more precise interpretation of the likely targets mediating these effects. Similarly, while the present studies clearly demonstrate a role for dopamine in mediating reinforcing effects of food and glucose in C. elegans, future studies using strains with genetic ablation of specific dopamine receptors (e.g., assessing effects of food and glucose-induced reward in dop-1, dop-2, dop-3, or mod-1 mutant strains, which might reveal a possible involvement of distinct dopamine receptor subtypes). Similarly, the present studies are based on efficacy of manipulations to increase preference for butanone, although butanone is already somewhat reinforcing compared to ethanol. Future studies might examine efficacy of these reinforcing effects on more neutral stimuli which have no reinforcing properties. Other future studies might examine effects of more specific agonists or antagonists of the C. elegans 5-HT4 receptor to corroborate the genetic studies reported here.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ES designed, oversaw and carried out most of the studies, and wrote the manuscript. HD and SL assisted with some studies. JS contributed some C. elegans strains and provided advice. CM designed the studies and wrote the manuscript. All authors contributed to the article and approved the submitted version.

ES was supported by the American Diabetes Association Clinical Scientist Predoctoral Training Award and the Ruth L. Kirschstein National Research Service Award for Individual Predoctoral Fellows. Studies funded by NIDDK.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.