Melanocortins integrate a key system in the regulation of food intake in vertebrates. These peptides, exhibiting melanotropic (melanin stimulation) and corticotropic (corticosteroid stimulation) activity encoded in the common precursor POMC. Two discrete groups of neurons in ARC and the caudal region of the nucleus of the tractus solitarius of the medulla produce POMC, mainly processed into α-melanocyte-stimulating hormone (α-MSH) and β-endorphin. Melanocortin signaling is transduced by five different G-coupled receptors (MC1R-MC5R) but only two are conspicuously expressed in the mammalian brain, i.e., MC3R and MC4R. Central activation of these receptors appears to mediate melanocortin effects on energy balance since both MC3R knockout rat and MC4R knockout mice display severe alterations in energy homeostasis (obesity, increased food intake and linear growth in MC4R; hypophagia, increased fat mass and food efficiency but reduced lean mass in MC3R). Accordingly, central administration of MC3/4Rs agonists results in a dose-dependent reduction in food intake in mice but MC4R-deficient mice do not respond to anorectic effects of agonists, suggesting that α-MSH inhibits feeding primarily by activating MC4R (reviewed by Anderson et al., 2016).

Several studies demonstrated that melanocortin system is a key point in the regulation of energy balance in fish (Cerdá-Reverter et al., 2011). Thus, in situ hybridization studies demonstrated the presence of POMC neurons in hypothalamus (Cerdá-Reverter et al., 2003a) but not in the vagal lobe (homolog of the nucleus of the tractus solitarius in mammalian brain). However, more sensitive techniques (quantitative PCR) also reported POMC-A1 expression in the hindbrain (Conde-Sieira et al., 2010b). Immunohistochemical techniques further demonstrated the presence of both α-MSH and β-endorphin in the hypothalamus of several fish species, including zebrafish and tilapia (Forlano and Cone, 2007; Chabbi and Ganesh, 2016) thus suggesting that POMC is mainly processed to these peptides. However, ACTH is also present in the preoptic area of carp (Metz et al., 1994) further suggesting an alternative POMC processing in the fish brain.

Intracerebroventricular (ICV) administration of α-MSH or α-MSH agonists inhibit food intake in a dose-dependent manner in goldfish (Cerdá-Reverter et al., 2003b) and rainbow trout (Schjolden et al., 2009), whereas administration of MC4R antagonists stimulates food intake in satiated animals (Cerdá-Reverter et al., 2003b). POMC-A1 and POMC-C levels increased post-feeding in medaka (Chisada et al., 2014), rainbow trout (Gong and Björnsson, 2014), and Atlantic halibut (Gomes et al., 2015), respectively. In contrast to that expected for a negative regulator of energy balance, POMC hypothalamic levels remained unchanged after 7 progressive fasting days in goldfish (Cerdá-Reverter et al., 2003a), but POMC-A1 mRNA decreased 50% after 28 fasting days in rainbow trout (Leder and Silverstein, 2006). Surprisingly, another study demonstrated that hypothalamic POMC A1, A2 and B levels increased after 118 fasting days in the latter species (Jørgensen et al., 2016). The lack of a consistent POMC regulation by energy balance also occurred in mammalian species. Thus, sheep with about 40% of total body-weight loss did not exhibit changes in POMC mRNA levels (Henry et al., 2001). A similar situation occurred in ewes that lost about 30% body fat (Henry et al., 2000). The constrasting effects of POMC-encoded peptides could explain the paradoxical absence of POMC regulation by energy balance. Thus, central injections of β-endorphin, the C-terminal peptide of POMC, stimulate appetite in the goldfish (De Pedro et al., 1995). It is then plausible that the effects of nutritional status might regulate post-transcriptional POMC-derived peptide levels by selective regulation of prohormone convertases. Under negative energy balance situations, when animals display an enhanced feeding drive, POMC might be processed into β-endorphin. In contrast, a positive energy balance may preferentially drive the processing of POMC into production of the food intake inhibitor α-MSH.

The melanocortin activity may be also modulated by receptor signaling. Accordingly, energy balance, in the absence of an agonist regulation, could up/down regulate the neuronal receptor density in feeding-related areas of the CNS. Studies in the sea bass suggest that hypothalamic MC4R expression remains unchanged after long-term fasting precluding this regulatory pathway (Sánchez et al., 2009). Alternatively, the melanocortin system also exhibits endogenous antagonists competing by binding and activation of the melanocortin receptors. AgRP1 is consistently upregulated by fasting in the hypothalamus of all tested vertebrates including zebrafish (Song et al., 2003), goldfish (Cerdá-Reverter and Peter, 2003), Atlantic salmon (Valen et al., 2011), and sea bass (Agulleiro et al., 2014). It is therefore conceivable that AgRP1 binding to central MCRs might regulate central melanocortinergic activity whereas POMC expression remains constant as a constitutive inhibitor.

In mammals, ARC POMC neurons also produce CART but hypothalamus is not the only brain area producing this neuropeptide (Elias et al., 1998). CART was isolated form ovine hypothalamus (Spiess et al., 1981) and its expression increased in the rat striatum after the administration of drugs such as cocaine and amphetamine (Douglass et al., 1995). Subsequent experiments revealed that CART neurons are anatomical targets for systemic leptin to induce anorexia (Kristensen et al., 1998). In fish, CART mRNA was characterized in several species (Subhedar et al., 2014) but immunohistochemical localization of the CART peptide in brain was only studied in catfish by using antibodies against rat CART (Singru et al., 2007) or by in situ hybridization in zebrafish (Nishio et al., 2012; Akash et al., 2014). CART mRNA abundance decreased with food deprivation in cod (Kehoe and Volkoff, 2007), goldfish (Volkoff and Peter, 2001a), and Atlantic salmon (Murashita et al., 2009), and increased with re-feeding in channel catfish (Kobayashi et al., 2008) whereas post-prandial changes occurred in channel catfish (Peterson et al., 2012), goldfish (Volkoff and Peter, 2001a), and dourado (Volkoff et al., 2016). As many other genes, CART is duplicated in the teleost genome and four different genes (CART1-4 or CART1, 2a, 2b, 4) encoding CART peptides are reported in zebrafish (Nishio et al., 2012) or up to seven genes in Senegalese sole (Bonacic et al., 2015). All of them are expressed in zebrafish CNS but only CART2 and 4 are produced in the hypothalamus (Akash et al., 2014). Unfortunately, CART/POMC colocalization in the neurons of the tuberal hypothalamus has not been assessed in fish yet. Both CART2 and CART4 are downregulated by fasting in ventrocaudal telencephalon (CART2) and ventral hypothalamus (CART2 and CART4) (Akash et al., 2014), as expected for a negative regulator of the energy balance and thus further suggesting their involvement in the regulation of energy balance in fish. Accordingly, ICV administration of CART peptides inhibit food intake in goldfish (Volkoff and Peter, 2000).

Atypically, melanocortin signaling is not exclusively regulated by the binding of endogenous agonists (see above), since naturally occurring antagonists, agouti-signaling protein (ASIP) and AgRP, compete with melanocortin peptides by binding to MCRs. ASIP and AgRP molecules exhibit a cysteine-rich C-terminal domain essential for the structural and biological properties of the peptide. A basic domain and a proline-rich area precede the cysteine knot in ASIP; AgRP sequences lacks both regions but, in contrast, exhibits a processing site prior to the cysteine domain where the propeptide is cleaved (reviewed by Cerdá-Reverter et al., 2011). In mammalian species, ASIP produced in the ventral skin regulates pigment pattern and binds MC1R and MC4R with similar affinity (Cerdá-Reverter et al., 2005). In contrast, AgRP is mainly expressed in the same NPY-producing neurons in the ARC. In fact, 95% of NPY neurons co-express AgRP (Hahn et al., 1998). AgRP/NPY neurons also produce GABA (Horvath et al., 1997) and regulate the activity of hypothalamic POMC/CART neurons. In addition, the GABAergic innervation from AgRP/NPY neurons to the parabrachial nucleus in the brainstem is able to regulate feeding behavior (Wu et al., 2009). Hypothalamic AgRP sharply increases with fasting, and overexpression in transgenic mice induces overfeeding and obesity but also enhances linear growth (Ollmann et al., 1997). The selective ablation of NPY/AgRP neurons induces a rapid decrease in food intake leading to starvation (Luquet et al., 2005).

Fish exhibit two copies of AgRP named AgRP1 and AgRP2. Few studies to date have precisely localized AgRP1 mRNA in brain regions of teleost fish (Cerdá-Reverter and Peter, 2003; Song and Cone, 2007) whereas others reported AgRP-ir in fish brain (Forlano and Cone, 2007; Agulleiro et al., 2014). All studies identified AgRP1 production within the posterior region of the ventral hypothalamus of the goldfish, zebrafish and sea bass. In the zebrafish, AgRP1 and α-MSH projections strikingly match the nuclei that express MC4R (Cerdá-Reverter et al., 2003b) and MC5R mRNA (Cerdá-Reverter et al., 2003c) in the goldfish. In fact, binding studies with zebrafish receptors demonstrated that AgRP1 acts as a competitive antagonist at MC3R, MC4R, and MC5R, all of them expressed in the brain (Song and Cone, 2007). Pharmacological studies in sea bass have shown that AgRP1 works as an inverse agonist at constitutively activated MC4R (Sánchez et al., 2009). In addition, ASIP1 overexpression working at central MC4R in transgenic zebrafish model results in enhanced feeding, feed efficiency, weight and linear growth but not in obesity (Guillot et al., 2016), in a similar way to the AgRP1 transgenic zebrafish (Song and Cone, 2007). Studies in zebrafish using first developmental stages demonstrated that AgRP1 knockdown results in reduced growth suggesting that AgRP1 suppression of MC4R activity is essential for larval growth (Zhang et al., 2012). Therefore, melanocortin system seems to impose a constitutive break to feeding and growth in fish probably through constitutive activity of MC4R. AgRP1 overexpression is the soundest response to fasting reported in fish, and its hypothalamic expression is essential to counteract MC4R negative effects on energy balance by inverse agonism on the receptor thus driving fish to feeding and concomitantly enhancing fish growth both dependently and independently on feeding levels (Guillot et al., 2016). In fact, AgRP1 mRNA abundance in hypothalamus increased in food deprived zebrafish (Song et al., 2003), goldfish (Cerdá-Reverter and Peter, 2003), sea bass (Agulleiro et al., 2013), and carp (Zhong et al., 2013) but not in Atlantic salmon (Murashita et al., 2009). Post-feeding did not induce changes in AgRP1 hypothalamic mRNA abundance in medaka (Chisada et al., 2014), but the increase in food intake in the GH-transgenic carp is associated with increased values of AgRP1 mRNA (Zhong et al., 2013).

There are no studies on NPY/AgRP colocalization in hypothalamic neurons in fish brain. The NPY family of peptides consists of 36-amino-acid peptides exhibiting carboxy terminal (C-terminal) amidation (Cerdá-Reverter and Larhammar, 2000). The family comprises three different peptides, the NPY, tyrosine-tyrosine peptide, (PYY), and the pancreatic polypeptide (PP). Tetrapod species produce all three peptides, whereas non-tetrapod vertebrates have only NPY and PYY. Teleost fish synthesize two different versions of NPY and PYY but not PP (Sundström et al., 2008). Brain distribution of NPY peptides was reported in detail only in the sea bass by in situ hybridization (Cerdá-Reverter et al., 2000a,b), and only NPY1 expressed in the rostral hypothalamus, where the AgRP1 neurons are localized in this species. Concomitant coexpression in the rostral hypothalamus suggests that both NPY1 and AgRP1 colocalize also in the tuberal hypothalamus, ventral telencephalon and preoptic area of fish.

NPY is the most potent orexigenic factor in vertebrates (Stanley and Leibowitz, 1985) and many studies shown this effect after ICV administration in several fish (López-Patiño et al., 1999; Aldegunde and Mancebo, 2006; Kiris et al., 2007; Yokobori et al., 2012). The mRNA abundance of NPY decreased after meal in goldfish (Kehoe and Volkoff, 2007), zebrafish (Tian et al., 2015), and grass carp (Zhou et al., 2013), though responses were contradictory in other species, such as orange-spotted grouper (Tang et al., 2013), rainbow trout (Gong and Björnsson, 2014), and zebrafish (Chen et al., 2016). Food deprivation in rainbow trout also resulted in decreased mRNA abundance of NPY (Gong et al., 2016b) but also after NPY intraperitoneal (IP) injections in flounder (Li et al., 2016). Accordingly, brain expression is upregulated by fasting in several fish species (Matsuda et al., 2012b) but the number of NPY immunoreactive neurons increase in the hypothalamus and posterior tuberculum after seven days fasting in the goldfish but not in the thalamus (Yokobori et al., 2012). Peripheral leptin is also able to regulate hypothalamic NPY expression since IP injection of recombinant leptin inhibits AgRP1/NPY but upregulates POMC/CART expression in goldfish (Yan et al., 2016). In addition, NPY was suggested to link the reproductive system and the central circuitry regulating energy balance in sea bass (Cerdá-Reverter et al., 1999).

The precise mechanisms connecting changes in glucose or fatty acid sensing systems and the mRNA abundance of orexigenic and anorexigenic factors are mostly unknown. Changes in neuropeptide expression have been associated with the modulation of brain homeobox transcription factor (BSX), forkhead box01 (Fox01), and phosphorylated cAMP response-element binding protein (pCREB) (Diéguez et al., 2011). The actions of these factors would result in the enhancement of CART and POMC, and the inhibition of NPY and AgRP, ultimately leading to decreased food intake (López et al., 2007; Diéguez et al., 2011). How these transcription factors relate to the activity of the different nutrient sensing systems? Several possibilities were suggested in mammals including (1) direct action of malonyl CoA or CPT1c, (2) indirect action through inhibition of CPT-1, (3) modulation by protein kinase B (Akt), AMPK, carbohydrate-responsive element-binding protein (ChREBP), or mTOR, or (4) involvement of ceramides (López et al., 2007; Diéguez et al., 2011; Gao et al., 2013).

In fish, a preliminary study in rainbow trout (Librán-Pérez et al., 2015b) showed that the cell signaling pathways that are dependent on Akt, AMPK, and mTOR activated in hypothalamus of fish fed a lipid-enriched diet. In the same species, a recent study (Velasco et al., 2016b) pointed to a possible role of ceramides in connecting hypothalamic fatty acid sensing systems and neuropeptide mRNA abundance with food intake control. Besides these studies, there is no further evidence in fish about the integration of the metabolic information of different nutrient (glucose, fatty acids, amino acids) sensing systems into a unique pathway regulating transcription factors involved in the production of neuropeptides controlling food intake. A summary of hypothetical relationships is shown in Figure 2.

Insulin and insulin receptors are present in fish hypothalamus (Gutiérrez and Plisetskaya, 1994; Leibush et al., 1996; Caruso et al., 2008). The effects of insulin treatment on food intake in fish are however contradictory. Thus, in rainbow trout IP administration of insulin either resulted in inhibition (Librán-Pérez et al., 2015a) or activation (Polakof et al., 2008a; Conde-Sieira et al., 2010b) of food intake, whereas ICV treatment did not affect food intake in catfish (Silverstein and Plisetskaya, 2000) but induced a decrease in rainbow trout (Soengas and Aldegunde, 2004). The anorectic effects in rainbow trout agree with the increased anorexigenic potential (increased CART and decreased NPY mRNA abundance) observed in hypothalamus after insulin treatment (Librán-Pérez et al., 2015a). Moreover, insulin administration in rainbow trout inhibits the glucosensing system dependent on GK (Polakof et al., 2007a, 2008a; Conde-Sieira et al., 2010b) whereas no clear effects were noted on fatty acid sensing systems (Librán-Pérez et al., 2015a).

Multiple forms of leptin exist in fishes, with the major form, leptin A, being the most likely candidate for food intake regulation and energy signaling (Gorissen and Flik, 2014). The treatment with either fish or recombinant human leptin in fish usually results in an anorectic response (De Pedro et al., 2006; Murashita et al., 2008; Kling et al., 2009; Vivas et al., 2011; Won et al., 2012; Gong et al., 2016a). This anorexigenic action of leptin appears to be mediated by brain leptin receptors, particularly in the hypothalamus (Kurokawa et al., 2008; Tinoco et al., 2012; Angotzi et al., 2016), where leptin exerts its anorectic action through regulation of neuropeptide expression. Accordingly, the leptin receptor-deficient medaka shows increased food intake accompanied by increased NPYa and AgRP mRNA abundance, and decreased POMC1 mRNA abundance (Chisada et al., 2014). Moreover, ICV injection of leptin reduced NPY mRNA levels in hypothalamus and telencephalon of goldfish (Volkoff et al., 2003), and in the whole brain of grass carp (Li et al., 2010). Peripheral treatment with recombinant salmon leptin A1 in rainbow trout induced a hypothalamic transient reduction and elevation of NPY and POMC-A1 mRNA levels, respectively (Murashita et al., 2011). Moreover, central injection of leptin increased CART-I mRNA levels in hypothalamus of goldfish (Volkoff and Peter, 2001a), and POMC-A1, A2 and B, and CART mRNA levels in rainbow trout (Gong et al., 2016a). On the other hand, the leptin anorectic effects also related in rainbow trout to the activation of central glucosensing systems (Aguilar et al., 2010, 2011).

The action of leptin on food intake regulation depends on the time of hormone administration, indicating a circadian dependence of leptin anorexigenic effects (Vivas et al., 2011). In this sense, leptin-aI and leptin-aII mRNAs in hypothalamus of goldfish show 24-h rhythms under light–dark cycle and scheduled feeding conditions (Tinoco et al., 2014b). Post-prandial increases of hypothalamic and liver leptin-aI expression (Huising et al., 2006; Moen and Finn, 2013; Zhang et al., 2013) are in agreement with the anorexigenic role of this hormone in fish. As expected for an anorexigenic hormone, feeding changes and nutritional status modifies the leptin system in fish, but different responses occurred depending on the feeding regime and species. Thus, leptin genes expression or circulating leptin levels rise with food restriction or starvation, when energy stores decline, and decrease during refeeding in some species (Gorissen and Flik, 2014; Johansson and Björnsson, 2015), but not in others (Huising et al., 2006; Tinoco et al., 2012). These, and other recent studies (Chisada et al., 2014; Londraville et al., 2014; Salmerón et al., 2015; Jørgensen et al., 2016) support that leptin in fish should not be considered as a lipostat signal in contrast to mammals. In this way, zebrafish knockout for leptin suggest the involvement of leptin in glucose homeostasis but not in adipostasis (Michel et al., 2016). Nevertheless, the existence of different leptin paralogs in fish differentially involved in the regulation on energy resources in a species/tissue dependent manner might explain the variety of results.

Feeding shows a rhythmic pattern in a widespread of fish (Madrid et al., 2001), occurring periodically through the 24-h light/dark cycle, but tidal, lunar and seasonal feeding rhythms are also been described in some species (López-Olmeda and Sánchez-Vázquez, 2010). These feeding rhythms, as many other aspects of energy metabolism and homeostatic balance, are coordinated by the phylogenetically well-conserved circadian system, which generates an internal timing that allow animals to anticipate cyclic environmental and endogenous variations, and then, prepare the physiology for upcoming challenges. In the functional organization of circadian system, multiple endogenous clocks, synchronized by environmental and endogenous cues, generate an internal rhythmicity close to 24 h (Albrecht, 2012). The hypothalamus plays a key role in the circadian timekeeping system of vertebrates, acting as an integrative neuronal network. In mammals, a bilateral structure in the anterior hypothalamus, the suprachiasmatic nucleus (SCN), is an autonomously rhythmic nucleus that drives overt circadian rhythms in behavior and physiology, including food intake (Partch et al., 2014). The SCN is formed by clusters of interconnected neurons with individual molecular clock mechanisms inside each cell that function as an autonomous oscillator, and the intercelular coupling among these individual cells is critical for the coordination of endogenous rhythmicity and the outputs to downstream tissues (Evans and Gorman, 2016). Additional neuronal clocks are also present in ARC, ventromedial hypothalamus, and the dorsomedial hypothalamus, and in a variety of peripheral tissues, including liver, adipose tissue, adrenal and muscle (Dibner et al., 2010). In mammals, it is widely accepted that the internal coordination of circadian rhythmicity is established by the temporal signals provided by the master SCN to other hypothalamic nuclei and peripheral organs through a wide variety of direct and indirect pathways that form a network of oscillators. Fish hypothalamus also contains circadian oscillators (Velarde et al., 2009; Patiño et al., 2011; Vatine et al., 2011; Idda et al., 2012; Martín-Robles et al., 2012; Nisembaum et al., 2012; Vera et al., 2013), but no master clock has been yet identified. In contrast, the functional relevance of endogenous clocks in a variety of fish tissues supports the existence of a multiple network of endogenous oscillators in the brain (hypothalamus, diencephalon, pineal, retina, optic tectum, pituitary) and peripheral (liver, gut, gonads, head kidney) locations (Isorna et al., 2017). A summary of the main components in the fish circadian organization is shown in Figure 3.

The key molecules that establish the functional core of self-sustained circadian oscillators clocks are highly conserved from invertebrates to mammals, including fish (Partch et al., 2014), although important differences are found in fish regarding the additional number of copies of the clock genes resulting from the 3R teleost-specific genome duplication. The generation of cellular circadian oscillations is based on autoregulatory 24-h transcriptional/translational feedback loops of a set of clock genes (Ko and Takahashi, 2006). The proteins BMAL1, CLOCK and its analog NPAS2 act as transcriptional activators via E-box sequences driving the expression of clock controlled genes, some of them encoding other core clock protein repressors (PER and CRY). These PER and CRY proteins form complexes that are translocated into the nucleus where they shut down their own expression by removing the CLOCK(NPAS2)/BMAL1 complexes, defining a main negative loop. Some reinforcing loops are formed by the orphan nuclear receptors from the REV-ERBα-β and RORα-β families, to modulate the rhythmic transcription of Bmal1, Npas and Clock (Guillaumond et al., 2005; Schibler et al., 2015). Recent studies demonstrated that such oscillations in clock genes expression paralleled critical events of chromatin remodeling (Coomans et al., 2015; Masri et al., 2015).

The light/dark cycle is the most effective signal that entrains the brain circadian clocks (Schibler et al., 2015). However, feeding related cues, including feeding time, nutritional inputs, feeding/fasting cycle and diet composition are the dominant synchronizers of peripheral clocks. Such so-called food-entrained oscillators provide animals the adaptive advantage of anticipate periodic meals within a circadian period of entrainment (Stephan, 2002; Mendoza, 2006). The food-entrained oscillators constitute the basis of the food anticipatory activity, an increase in locomotor activity occurring just before feeding time in anticipation of a predictable daily meal. This activity was evidenced in fish (López-Olmeda and Sánchez-Vázquez, 2010; Feliciano et al., 2011; Vera et al., 2013) and mammals (Mistlberger, 2009), and appears to be independent on hypothalamic SCN in mammals (Stephan, 2002) and hypothalamic clocks in fish (Velarde et al., 2009). To date, the endogenous control of the food anticipatory activity is largely unknown, but recent studies suggest that some orexigenic hormones, such as ghrelin, may be involved both in fish (Nisembaum et al., 2014b) and in mammals (Patton and Mistlberger, 2013).

The location of food-entrained oscillators is not yet certain. Neuronal feeding-synchronized clocks are suggested to reside in ventromedial and dorsomedial hypothalamus (Acosta-Galvan et al., 2011), or even in extrahypothalamic brain areas including nucleus accumbens and amygdala in mammals. The search and identification of feeding entrained clocks is a topical subject in fish, and both, peripheral (Nisembaum et al., 2012; Sánchez-Bretaño et al., 2015a) and brain (Feliciano et al., 2011; Vera et al., 2013; Sánchez-Bretaño et al., 2015b) are possible targets. This complex variety of locations of oscillators requires a functional organization among them to generate coordinated responses, and recent data suggest that nutritional cues may act as time givers (Dattolo et al., 2016).

Food intake is a potent phase-resetting agent for peripheral and central clocks, by entraining a number of overt rhythms in animals. The molecular mechanism underlying feeding entrainment remains largely unknown. In mammals, feeding modifies the phase of molecular oscillations in ARC and dorsomedial hypothalamic nuclei independently on the hypothalamic SCN (Feillet et al., 2008). In fish, a daily feeding schedule synchronizes rhythmicity of central (hypothalamus and optic tectum) and peripheral (gut and liver) clock genes (Feliciano et al., 2011; Nisembaum et al., 2012; Vera et al., 2013). The nature of the endogenous signals associated with feeding that may be capable of resetting peripheral clocks is not fully characterized yet, but metabolites, nutrient sensors, hormones, and possibly neuronal signals transmitted from nutrient-sensing areas to peripheral organs are good candidates.

The circadian system controls the expression of essential genes in numerous metabolic pathways (Ribas-Latre and Eckel-Mahan, 2016), but, in turns, it can be synchronized by its own effectors, generating a bidirectional relationship among the circadian clocks and metabolic signals (Challet, 2013).

Little is known about how specific diet components can modulate the circadian clocks, although it appears that periodic availability of circulating nutrients induces the feeding dependent resetting of peripheral and central circadian networks (Asher and Sassone-Corsi, 2015). In this context, ARC is an important target for such metabolic feedback (Uchida et al., 2016). Certain populations of neurons in ARC show rhythmicity in mammals (Ellis et al., 2008), with circadian rhythms of PER2 in vitro (Guilding et al., 2009). The circadian modulation of ARC neurons by circulating metabolites support the timing activation of this center for sensory metabolic information (Van den Top and Spanswick, 2006). The fact that ARC is also able to influence the functionality of the SCN (Yi et al., 2006), suggest that this ARC-SCN reciprocal interaction is essential to maintain a well-balanced metabolic circadian pattern.

Nutrient sensors evolved as time-giving signals by exerting different actions on the circadian clocks. This is the case of AMPK, regulated by glucose and fatty acids, which directly regulates CRYs proteins (Lamia et al., 2009), CREB, glycogen synthase kinase 3 beta, PPARs, and the redox state (Oosterman et al., 2015; Ribas-Latre and Eckel-Mahan, 2016). One interesting link between sensing of cellular energy status and circadian clocks is the control of clock genes expression and deacetylation of clock proteins by the sirtuins (SIRT), a (NAD⁺)-dependent class III of histone deacetylases that are well characterized by their numerous effects on intracellular metabolism. Particularly, the SIRT1 and SIRT6 establish functional links between cellular metabolism and circadian clocks physiology in mammals (Masri et al., 2014; Orozco-Solis et al., 2015). The activity of SIRT1 fluctuates through the feeding/fasting cycle (Cakir et al., 2009) and is involved in the cyclic control of cofactors and peptides of circadian clocks by deacetylating BMAL1 and PER2 in liver (Nakahata et al., 2008) and activates the hypothalamic SCN pacemaker in mice (Chang and Guarente, 2013). The high expression of this enzyme in the ventromedial hypothalamus support the role of this brain area on monitoring metabolic signals through SIRT1, at least in mice. Interestingly, these ventromedial neurons expressing SIRT1 project POMC neurons in the ARC, and innervate the outer shell of SCN, providing a basis to integrate and communicate metabolic signals in the hypothalamus (Orozco-Solis et al., 2015). SIRT6 appears to be involved in the cyclic synthesis of lipids and carbohydrates (Masri et al., 2014), which may reflect differential responses to distinct nutritional intakes. Whether these molecular sensors modulate brain and peripheral clocks in fish remains to be elucidated. Preliminary studies in rainbow trout assessed changes in mRNA abundance of SIRT-1 in parallel with the functioning of fatty acid sensing systems (Velasco et al., 2016a,b). The clearly adaptive significance of circadian clocks entrainment by nutrient status is supported in fishes by the feeding-related rhythms driven by the circadian system in digestive enzymes (Vera et al., 2007; Montoya et al., 2010; Nisembaum et al., 2014b). Furthermore, changes described in parameters related to glucose, fatty acid, and amino acid metabolism in brain (Polakof et al., 2007c) and peripheral tissues (Polakof et al., 2007d; Hernández-Pérez et al., 2015) also support the involvement of nutrient status.

It is widely accepted that daily endocrine rhythms are outputs of the circadian system, but recent reports suggest that several hormones may act as inputs in the timing signalization of hypothalamic and peripheral clocks (Challet, 2015; Coomans et al., 2015).

Ghrelin may prove to be an endocrine signal that participates in the integration of gastrointestinal signals by the central clocks. The rhythmic expression of ghrelin transcript in hypothalamus, pituitary and gut (Sánchez-Bretaño et al., 2015b), and the periprandial variations in plasma of goldfish (Blanco et al., 2016a), and in hypothalamus of tilapia, zebrafish and goldfish (Peddu et al., 2009; Blanco et al., 2016b) indicate that ghrelin may work as an output of food-entrained oscillators. Ghrelin receptors are widely expressed in fish brain (Chen et al., 2008; Sánchez-Bretaño et al., 2015b), including those hypothalamic areas involved in the integration of food intake control, as well as the preoptic region and the anterior periventricular nucleus (Kaiya et al., 2010; Sánchez-Bretaño et al., 2015b). The expression of clock genes in these hypothalamic areas (Mazurais et al., 2000; Sánchez-Bretaño et al., 2015c) and the induction of per genes by ghrelin (Nisembaum et al., 2014b) support the role of ghrelin as an input of food-entrained clocks. In vivo (Nisembaum et al., 2014b) and in vitro (Sánchez-Bretaño et al., 2016b) studies carried out in goldfish reinforce the role of ghrelin in the feeding and metabolism-related signaling in hypothalamic and liver clocks. Furthermore, some reports indicate that this hormone may also participate in the regulation of food anticipatory activity in goldfish, as it modifies locomotor activity (Matsuda et al., 2006a), and this locomotor activity is blocked by a ghrelin antagonist (Nisembaum et al., 2014b).

The orexin/hypocretin system is involved in the coordination of rhythmic daily functions, such as feeding and energy balance in fish (Matsuda et al., 2012a). Taking in mind the orexigenic role of this regulator and its stimulatory effect on daily activity rhythms in the absence of other inputs (Nisembaum et al., 2014a), the possible role of orexin system in the physiology of certain brain areas involved in food intake and circadian regulation is intriguing. Orexin fibers found in some central oscillators, such as the pineal and the SCN, and other brain regions related to the regulation of activity/rest cycles (Wong et al., 2011; López et al., 2014). As a rule, orexins exert excitatory effects on various brain nuclei with the exception of a direct hyperpolarisation of clock cells in SCN during the night (Belle et al., 2014).

The orexin system is under the control of the molecular clock, as orexin neurons projecting into the pineal show circadian daily rhythmicity driven by the clock (Appelbaum et al., 2010), and daily variations in the hypothalamic expression of orexin in goldfish related to clock genes oscillations (Hoskins and Volkoff, 2012). Nevertheless, recent studies suggest that this peptidergic system also acts as an input to the clockworks. Particularly, this peptide seems to play an important role in food intake anticipation in fish, as its central administration up-regulates per genes in hypothalamus (central) and foregut (peripheral) clocks (Nisembaum et al., 2014a). The induction of per genes by orexin, as produced ghrelin, questions the existence of synergy between both orexigenic regulators in their actions on the food-entrained clocks.

Glucocorticoids display strong daily rhythms in vertebrates including fish. Cortisol daily rhythms are clear outputs of the circadian system, which are synchronized by the feeding-fasting cycle and feeding time in fish (Isorna et al., 2017). However, little is known about the possible role of glucocorticoids as inputs to the circadian system in fish. Recent reports in goldfish show that cortisol induces per1a and per1b expression, and represses the positive elements (bmal1a and clock) in liver clockwork (Sánchez-Bretaño et al., 2016a). Interestingly, such actions of cortisol on the liver, one of the most sensitive peripheral clockwork, link the entrainment of food-entrained clocks by both, metabolic signals and hormones.

The endocrine stress response in fish is mediated by both the hypothalamic-sympathetic chromaffin cell and the HPI axes. The activation of both systems restores homeostasis by mobilizing fuel to make energy available to cope with increased metabolic demand (Wenderlar Bonga, 1997; Mommsen et al., 1999). A disruption of the feeding behavior is a common feature of the behavioral response to stress in fish (Bernier and Peter, 2001; Bernier, 2006). The activation of HPI in response to stress involves the synthesis of corticotrophin relasing factor (CRF) in the neurons of the preotic area of the CNS that, in turn, stimulates release of adrenocorticotropic hormone (ACTH) from the corticotrophic cells in the adenohypophysis. ACTH binds to the MC2R in the surface of the interrenal cells of the head kidney to stimulate glucocorticoid release into the blood (Wenderlar Bonga, 1997; Mommsen et al., 1999). Cortisol, the main glucocorticoid in fish mediates many effects of stressors on metabolic and behavioral processes (Barton, 2002; Bernier, 2006; Aluru and Vijayan, 2009).

In the fish brain, CRF is abundantly expressed within the dorsal telencephalic structures but also in the preoptic area and tuberal hypothalamus of zebrafish (Alderman and Bernier, 2007). ICV administration induces a dose-dependent reduction in food intake levels in goldfish (De Pedro et al., 1993; Bernier, 2006) that is reverted by the injection of receptor antagonist (De Pedro et al., 1997; Bernier and Peter, 2001). Goldfish treated with glucocorticoid antagonists or inhibitors of cortisol synthesis, display increased CRF mRNA brain levels and reduced feeding, which again can be reverted by CRF receptor antagonist (Bernier and Peter, 2001). It is suggested that CRF is a major physiological transductor of stress effects on food intake in fish (Bernier, 2006). In addition, CRF seems also to regulate social effects since its increased expression is responsible of the reduced feeding levels in subordinate fish (Doyon et al., 2003).

Recent studies demonstrated that MC4R is able to bind ACTH in the presence of the melanocortin receptor accessory protein 2 in zebrafish. In addition, ACTH administration inhibits food intake in zebrafish but only in animals carrying a functional copy of MC4R (Agulleiro et al., 2013). Together with the ACTH production in carp brain (Metz et al., 1994) this suggests that ACTH can also transduce stress information into feeding circuitry downstream of the melanocortin system. In addition, MC4R is highly expressed in the magnocellular neurons of the preoptic area where CRF is synthesized thus suggesting that central melanocortins could modulate the activity of the CRF neurons (Cerdá-Reverter et al., 2003a; Sánchez et al., 2009). Stress conditions not only regulate CRF and melanocortinergic pathways in fish as the activity of many other central systems vary with stressful conditions and/or increased glucocorticoid signaling. Changes in the activity of central monoaminergic systems including serotoninergic and dopaminergic systems occurred after stressful conditions. These central monoaminergic systems can play a key role in the integration of information during the exposition to the stressor but also in organizing the coordinated stress response (Lanfumey et al., 2008). Studies in rainbow trout demonstrated that the severity of the stressors were able to produce a rated stress response, but serotonin central levels were not accordingly gradated suggesting that some other systems must be integrating the magnitude of the stress response (Gesto et al., 2015). Chronic treatment with fluoxetine (inhibitor of serotonin recapture) is able to reduce whole body cortisol levels and display anxiolytic effects in zebrafish (Egan et al., 2009) but, on the contrary, acute stress induced a rapid increase of serotoninergic and dopaminergic activity in the forebrain of rainbow trout (Gesto et al., 2013). Accordingly, both dopamine (Leal et al., 2009) and serotonin (Ortega et al., 2013) inhibit food intake in fish supporting both dopaminergic and serotoninergic pathways as neuroanatomical substrates for the integration of stress effects on behavioral pathways regulating food intake.

The reduced food intake observed in response to stress in fish could also relate to changes in the ability of nutrient sensing systems to modulate food intake. In rainbow trout chronic stress induced a readjustment in the activity of hypothalamic glucosensing mechanisms (Conde-Sieira et al., 2010a; Otero-Rodiño et al., 2015). The readjustment resulted in an inability of the fish to compensate with changes in food intake those of circulating glucose levels as observed in non-stressed fish. The response of hypothalamic mRNA abundance of CART, POMC, and NPY to glucose changed under stress conditions (Conde-Sieira et al., 2010a; Otero-Rodiño et al., 2015). CRF might be involved in the mechanism through which stress influences food intake control (Evans et al., 2004; McCrimmon et al., 2006). Accordingly, CRF treatment of rainbow trout hypothalamus in vitro readjusted glucosensing mechanisms (Conde-Sieira et al., 2011) in a way similar to that observed under stress conditions (Conde-Sieira et al., 2010a).

The orexins/hypocretins are neuropeptides belonging to the incretin gene family of peptides and are involved in many physiological processes in fish, including feeding (Hoskins and Volkoff, 2012), locomotion/rest cycles (Volkoff, 2012; Nakamachi et al., 2014; Nisembaum et al., 2014a) and reproduction (Wong et al., 2011). The cDNA sequences encoding for preproorexin in fish contains orexin A and orexin B, with a high degree of homology with other orexins in vertebrates (Kaslin et al., 2004). Orexins are present in numerous fish species with a broad distribution. Thus, orexin ir-cell bodies, preproorexin mRNA, and orexin proteins are abundant in hypothalamus and preoptic area of most species so far studied (Volkoff, 2015), areas associated with the control of food intake in fish.

Orexins stimulate food intake in all fishes investigated so far. Earlier studies demonstrated this effect in goldfish (Volkoff et al., 1999; Facciolo et al., 2011; Matsuda et al., 2012a), and zebrafish (Novak et al., 2005; Yokobori et al., 2012). Currently, it is generally accepted that orexins are potent orexigenic regulators in fish (Hoskins and Volkoff, 2012; Matsuda et al., 2012a; Penney and Volkoff, 2014). The stimulation of appetite by orexin injections in many of the species studied is consistent with increases in preproorexin brain mRNA expression levels displayed after different periods of food deprivation (Novak et al., 2005; Abott and Volkoff, 2011; Volkoff, 2014). In goldfish, the presence of orexin-like-ir in the hypothalamic nucleus recessus lateralis is induced by fasting condition (Nakamachi et al., 2006). These stimulations of the hypothalamic orexin system by fasting suggest that this peptide play a role in long-term regulation of feeding in fish. Thus, short-term periprandial changes in the expression of orexin occurred in fishes. Orexin expression in the brain increases 1 h before the scheduled mealtime in Mexican blind cavefish (Wall and Volkoff, 2013); and hypothalamic preproorexin expression peaks around mealtime, and decreases after feeding in Atlantic cod (Xu and Volkoff, 2007) and orange-spotted grouper (Yan et al., 2011). These results suggest that orexin may be a signal preceding food supply under scheduled feeding conditions. Recently, it has been shown that orexin-A enhances locomotor activity in goldfish (Nakamachi et al., 2014), and synchronizes locomotor activity in goldfish maintained under 24 L and fasting conditions (Nisembaum et al., 2014a), which allowed authors to suggest that orexin might mediate the locomotor activity entrainment by food schedule in this teleost.

Orexin system is a good target to investigate possible interactions among food intake regulators. Central administration of orexin increases the expression of hypothalamic NPY in goldfish (Volkoff and Peter, 2001b; Nisembaum et al., 2014a), and in the orange spotted grouper (Yan et al., 2011). Desensitizing the orexin system by treatment with high doses of orexin A results in a decrease in NPY-induced feeding (Volkoff and Peter, 2001b), while the blocking of NPY receptors reduces feeding induced by orexin A (Kojima et al., 2009). In addition, ICV co-administration of orexin-A and NPY results in a synergistic orexigenic effect in goldfish (Volkoff and Peter, 2001b; Volkoff et al., 2003). Thus, a functional interdependence may exist between orexin-A and NPY in the stimulation of food intake. On the other hand, ICV orexin-A treatment stimulates ghrelin mRNA 1 h post-injection in goldfish foregut. This agrees with the increase of diencephalic ghrelin induced by orexin-A in this teleost (Miura et al., 2007), and demonstrates that the interaction between these two peptides takes place not only into the hypothalamus, but also suggests a brain-gut communication (Unniappan et al., 2004; Nakamachi et al., 2006). Neuroanatomical evidences support the interactions above described among orexin-A and other appetite-regulating peptides, as NPY and ghrelin (Volkoff et al., 2003; Miura et al., 2007). Recent studies suggested a role of tyrosine hydroxylase in the regulation of food intake and locomotor activity in vertebrates (Wall and Volkoff, 2013). The increased mRNA expression levels of this enzyme induced by fasting or orexin in cavefish may suggest the involvement of catecholamines in the actions of orexin (Abott and Volkoff, 2011).

Melanin-concentrating hormone (MCH) characterized in chum salmon as a circulating-cyclic heptadecapeptide involved in color change (Kawauchi et al., 1983). MCH is mainly produced within the tuberal hypothalamus, stored in the pituitary neural lobe, and released under adaptation to a light-colored background (Cerdá-Reverter and Canosa, 2009). MCH is a potent orexigenic factor inducing weight gain in mammals (Qu et al., 1996). In fish, however, contradictory results were obtained, such as in goldfish (Matsuda et al., 2006c). Data reported in barfin flounder showed increased hypothalamic MCH expression after food deprivation (Takahasi et al., 2004) and enhanced somatic growth after white background adaptation (Yamanome et al., 2005). Transgenic medaka overexpressing MCH exhibit lightened body color, but development, growth, feeding behavior and reproduction do not remarkably differ from non-transgenic siblings (Kinoshita et al., 2001).

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.