Introduction

Feeding behavior is essential to animals for their survival and reproduction and is known to be modulated by various internal and external factors: nutritional status, sexual maturity, temperature, seasonality, etc. This behavior is known to be closely correlated with reproduction (Kauffman & Rissman, 2004), which is an essential biological activity important for the animal life. Previous studies reported that nutritional-state modulates reproductive behaviors and functions (Chen et al., 2006; Amirjani et al., 2019; Volk et al., 2017; Lynn et al., 2010). For example, musk shrews show defective sexual behavior under fasted conditions (Temple & Rissman, 2000). In addition, not a few studies demonstrated that fasting-induced low energy condition suppresses reproduction (Evans & Anderson, 2012; Kalra & Kalra, 1996; Kirkwood et al., 1987; Merry & Holehan, 1979; Hasebe et al., 2016). Thus, it has been well investigated how nutritional status resulting from feeding modulates reproduction. On the other hand, it has been reported that some animals change their feeding behavior during the breeding season. For instance, the black seabream migrates to the shallow water during the breeding season (Tsuyuki, 2018; Kawai et al., 2020) when they can get more food, and the white-tailed deer spend more time for feeding under reproductive status (Stone et al., 2017). Such close relationship between reproduction and feeding is thought to be important for biological fitness. However, the regulatory mechanisms for breeding season-dependent feeding behavior are still unknown. One possible reason is that most of the model animals appear to have lost the well-defined breeding season. Although the mammalian models, mice and rats, and teleost model zebrafish, have reproductive cycles of about 4-5-days (Nilsson et al., 2015; Peute et al., 1978), they do not have definite breeding seasonality. Thus, the mechanisms for long-term changes in feeding behavior according to the breeding season have not yet been studied in detail.

Here, as a model animal for the seasonal breeder, we used a teleost fish, medaka (Oryzias latipes). Medaka is a useful model animal, whose reproductive status can be experimentally controlled by day length (Robinson & Rugh, 1943; Egami, 1954) under the fixed appropriate ambient temperature. In the long-day (LD) condition, female medaka becomes reproductive and regularly spawns every day, while it becomes non-reproductive in the short-day (SD) condition. In other words, LD or SD condition can induce breeding or non-breeding season of female medaka, respectively. Thus, medaka enables us to analyze the mechanisms of breeding season-dependent feeding behavior without considerations for possible changes in metabolism and gene expressions due to the changes in ambient temperature.

Regulatory mechanisms of feeding behavior have mainly been analyzed in mammals. These studies reported that some neuropeptides, such as agouti-related peptide (AgRP) and neuropeptide Y (NPY), are involved in the control of feeding and called “feeding-related peptides” as key molecules for the regulation of feeding behavior (Hahn et al., 1998; Aponte et al., 2011; Krashes et al., 2011; Andermann & Lowell, 2017). Teleosts have also been thought to possess regulatory mechanism for feeding similar to mammals. In fact, expression of homologous genes coding for feeding-related peptides have been reported (Rønnestad et al., 2017; Conde-Sieira & Soengas, 2016). On the other hand, although administration of some of them have been suggested to induce feeding behavior in teleosts as well (Rønnestad et al., 2017), their functions in feeding behaviors still remain unclear.

In the present study, to understand mechanisms of breeding season-dependent feeding behavior, we focused on female medaka, which clearly show seasonal changes in breeding conditions by day length (Mitani et al., 2010; Kanda et al., 2008) under the fixed appropriate temperature. We first quantified changes in feeding behavior according to the breeding states and found that female medaka under the condition of breeding season (LD) eat more than those under the condition of non-breeding (SD). Therefore, we searched for genes that show breeding state-dependent changes in expression and found some candidates for feeding-related peptides in medaka. We then analyzed expressions of the candidate genes by using RNA-seq, in situ hybridization, and RT-qPCR, and analyzed phenotypes of gene knockout medaka. These results led us to conclude that AgRP1 plays a key role in the breeding season-dependent changes in feeding behavior via ovarian estrogenic signals.

Results

Feeding behavior of female medaka is upregulated in breeding season

To analyze food intake of male and female medaka in breeding/non-breeding seasons, we first established a method for measurement of food intake in medaka. In brief, we placed medaka in a white cup, fed brine shrimp to medaka in all-you-can-eat style for 10 min, and counted the leftover brine shrimp in the cup with a “shrimp-counter” system (called Japanese “Wanko-soba” like method, Figure 1-Source Code 1 and Figure 1-Supplementary Figure 1). We used this system to analyze food intake of male and female medaka under the breeding condition equivalent to that in breeding season (kept under long day (LD) condition) or under the non-breeding condition equivalent to that in non-breeding season (short day (SD) condition) (Figure 1A; Kanda et al., 2008). We found that female medaka under the breeding (LD) condition eats more than those under the non-breeding (SD) condition (Figure 1B). In contrast to female, in males there was no significant difference in food intake between the breeding and non-breeding condition (Figure 1C). Since these results demonstrated that females, not males, show breeding season-dependent feeding behavior, we focused only on female medaka in the following analyses on neuronal mechanism for breeding season-dependent feeding behavior. Next, to examine which gene products modulate feeding behavior of female medaka in breeding season, we performed mRNA-sequencing (RNA-seq) using the whole brain of female medaka in breeding condition (LD) and non-breeding condition (SD) (Figure 1-Supplementary Figure 2A). Overall, 1025 genes showed significantly different expression between LD and SD female medaka. Figure 1-Supplementary Figure 2B shows a heat map of representative genes mainly related to neuroendocrine system, which were differently expressed between LD and SD female. Among the conventional candidate feeding-related neuropeptides, we identified two kinds of neuropeptides, agrp1 and npyb, both of which showed higher expression in LD than in SD (Figure 1-Supplementary Figure 2C and 2D). Both AgRP and NPY are known to have orexigenic effects mainly in mice (Schwartz et al., 2000; Andermann & Lowell, 2017). Therefore, in the subsequent analyses, we focused on agrp1 and npyb as candidate genes that modulate breeding season-dependent feeding behavior in female medaka.

Reproductive female medaka show larger amount of food intake.

(A) Light conditions for breeding and non-breeding status. (B) Food intake (10 mins) of female medaka in LD (orange; n = 7) and SD (blue; n = 7) conditions, normalized to the amount of artemia eaten by medaka in LD (breeding) condition (p = 0.02519). (C) Food intake (10 mins) of male medaka in LD (orange; n = 7) and SD (blue; n = 7) conditions, normalized to the amount of artemia eaten by medaka in LD (breeding) condition (p = 0.6540). Mann-Whitney U test, * p < 0.05. n.s., not significant.

AgRP1, NPYa and NPYb, but not AgRP2, may be the “feeding-related peptides” in female medaka

Medaka has two agrp paralogues, agrp1 and agrp2 and two npy paralogues, npya and npyb, which arose from third round whole genome duplication early in the teleost lineage (Liu et al., 2019; Sundstrom et al., 2008). Therefore, we next examined the anatomical distribution of neurons expressing agrp1, npyb, and their paralogs in the female brain by in situ hybridization (ISH). We found that agrp1-, agrp2-, and npyb-expressing neurons are distributed in local brain regions (Figures 2A-2C and 2E), while npya-expressing neurons are more widely distributed (Figures 2A and 2D) in the brain. The agrp1 neurons were distributed in the nucleus ventralis tuberis (NVT) of the hypothalamus (Figure 2B), while agrp2 neurons were expressed in the telencephalon but not in the hypothalamus (Figure 2C). On the other hand, npya neurons were distributed more widely from telencephalon to hypothalamus (Figure 2D). npyb neurons were distributed locally in the nucleus ventralis telencephali pars dorsalis (Vd) of the telencephalon (Figure 2E).

agrp1-, agrp2-, and npyb-expressing neurons are distributed locally, while npya-expressing neurons are distributed more widely in the brain.

(A) Illustration of the lateral view of medaka brain and distributions of cell bodies expressing each agrp or npy gene. The oblique lines labeled with a-e indicate the level of the frontal sections in a-e. (a-e) Illustrations of frontal sections showing the distribution of neurons expressing each gene. The localization of neurons is indicated in the right half of the illustrations. dDm dorsal region of area dorsalis telencephali pars medialis, Dl area dorsalis telencephali pars lateralis, BO bulbus olfactorius, Vd area ventralis telencephali pars dorsalis, Vv area ventralis telencephali pars ventralis, Vs area ventralis telencephali pars supracommissuralis, lfb lateral forebrain bundle, Vl area ventralis telencephali pars lateralis, TO tectum opticum, NVT nucleus ventralis tuberis, PIT pituitary, NAT nucleus anterior tuberis, NRL nucleus recessus lateralis. (B) agrp1-expressing neurons are localized in NVT. Scale bar: 100 μm. (C) agrp2-expressing neurons are localized in Vv. Scale bar: 100 μm. (D) npya-expressing neurons are distributed in dDm, Dl, BO, Vs, Vl, lfb, NRL, NAT, NVT, and TO. Scale bar: 100 μm. (E) npyb-expressing neurons are localized in Vd. Scale bar: 50 μm. (F) agrp1 (magenta) and npya (green) are distributed in NVT, but the two genes are not co-expressed. (G-J) agrp and npy expressions in the whole brain of female medaka under normally fed condition (Fed; white; n = 7) or 2-week food restricted (FR; gray; n = 8). (G) agrp1 (p = 0.001243) (H) agrp2 (p = 0.9551) (I) npya (p = 0.2319) (J) npyb (p = 0.0003108) Mann-Whitney U test, ** p < 0.01, *** p < 0.001. n.s., not significant.

In mice, agrp is known to be only expressed in the hypothalamus and mostly co-expressed with npy (Hahn et al., 1998), and these AgRP/NPY neurons are known to regulate mammal feeding behavior (Shutter et al., 1997; Broberger et al., 1998; Ollmann et al., 1997; Takahashi & Cone, 2005). In medaka, on the other hand, agrp1 signals were not observed in npya neurons (Figure 2F), although the both genes were expressed in the hypothalamus. These results suggest that AgRP and NPY are not co-expressed in medaka. Since AgRP and NPY of medaka showed different expressing patterns compared with other animals such as mice, we examined whether they act as modulators of feeding. We divided female medaka in LD condition into two groups; one group was kept under normally fed condition (Fed), and the other was kept under 2-week food restricted condition (FR). We then analyzed whole-brain expressions of these four genes. RT-qPCR analysis demonstrated that 2-week FR increased the expression of agrp1 (Figure 2G) but decreased that of npyb (Figure 2J), suggesting that the two peptides are involved in feeding in an opposite manner. On the other hand, agrp2 did not significantly change their expressions between Fed and FR conditions (Figure 2H). Although npya did not significantly change their expressions between Fed and FR conditions (Figure 2I), it may be possible that npya expression changed in a specific brain region, since npya neurons are widely distributed in various brain regions as described above (Figure 2D). Since hypothalamic npy control feeding in mice (Kohno & Yada, 2012), we also examined the npya-expression in medaka hypothalamus by ISH. We counted npya-expressing neurons in each hypothalamic region and compared them between Fed and FR female medaka (Figure 2-Supplementary Figure 1). We found that the total number of npya-expressing neurons in the hypothalamus was significantly larger in Fed compared with FR (Figure 2-Supplementary Figure 1A). Here, significant increase in cell number was observed in NRL and NAT, but not in NVT. Thus, the results suggest that AgRP1, NPYa and NPYb, but not AgRP2, may be the “feeding-related peptides” in female medaka.

Both agrp1 and npyb show higher expression levels in LD than in SD female medaka

To further examine the result of RNA-sequencing (Figure 1-Supplementary Figure 2), we compared expression of agrp1 and npyb between the female medaka under the breeding condition (LD) and those under the non-breeding condition (SD), using ISH and whole-brain RT-qPCR (Figure 3). First, we performed whole-brain RT-qPCR and found that the expression level of agrp1 was higher in LD than in SD female (Figure 3A). In ISH experiments, we observed larger number of agrp1-expressing neurons in LD than in SD females (Figure 3B and 3C). Moreover, to investigate the expression levels using ISH, we counted the number of neurons showing ISH signals at each 30-min time point from the beginning of the chromogenic reaction (Figure 3D) and analyzed the ratio of cell number showing ISH signals at 90 min and that at 300 min (saturation point of the reaction) (Figure 3E). We found signals of agrp1 90 min after chromogenic precipitation in both LD and SD conditions, and the relative cell number was not significantly different between LD and SD (Figure 3E). Since the expression level of agrp1 was higher in LD than that in SD (Figure 3A), higher expression of agrp1 under the breeding condition may be due to the increase in the number of neurons expressing agrp1. On the other hand, npyb expression in RT-qPCR was significantly higher in LD than that in SD (Figure 3F), although ISH analysis demonstrated that npyb-expressing cell number was not significantly different between LD and SD (Figure 3G). These results suggest that the expression level for each neuron increased in LD compared with SD. Thus, higher expression of npyb under the breeding condition may be due to the increase of expressions in each neuron expressing npyb.

agrp1 and npyb show higher expression levels in LD than in SD female.

(A) agrp1 expression in the whole brain of LD (orange; n = 8) and SD (blue; n = 8) female medaka. (p = 0.001865) (B) ISH of agrp1-expressing neurons in LD and SD female medaka. Scale bar: 100 μm. (C) The number of neurons expressing agrp1 in LD (orange; n = 5) and SD (blue; n = 5). (p = 0.02828) (D) Time course of the number of neurons showing ISH signals for agrp1 at each 30-min time point from the beginning of the chromogenic reaction in LD (orange; n = 5) and SD (blue; n = 5). The positive neurons were counted from 0 min to 300 min after application of NBT/BCIP (see Material and Methods), and the reaction saturated at 300 min onward. (E) Ratio of cell number at 90 min divided by that at 300 min. (p = 0.2948) (F) npyb expression in the whole brain of LD (orange; n = 8) and SD (blue; n = 8) female medaka. (p = 0.0001554) (G) The number of neurons expressing npyb in LD (orange; n = 4) and SD (blue; n = 4). (p = 0.3429) The upper, middle, and lower bars show the third quartile, median, and the first quartile, respectively. Mann-Whitney U test, * p < 0.05, ** p < 0.01, *** p < 0.001. n.s., not significant.

In juvenile female medaka, expression levels of neither agrp1 nor npyb shows significant change according to the day-length

The results thus far indicates that expressions of agrp1 and npyb are upregulated in female medaka under the condition of breeding season. Since the breeding/non-breeding state is experimentally controlled by day-length (LD/SD) in the present study, we examined which of the two factors, day-length itself or LD-induced sexual maturity, regulates the expression of agrp1 and npyb. Here, we used sexually immature juvenile medaka and compared their whole-brain expressions of agrp1 and npyb under LD/SD conditions using RT-qPCR (Figure 4). We found that expression levels of neither agrp1 nor npyb shows significant difference between LD and SD (Figure 4A and 4B), suggesting that neither of them is regulated directly by day length itself. Instead, the gene expression is suggested to be regulated by LD-induced sexual maturity.

In juvenile female medaka, expression levels of neither agrp1 nor npyb shows significant change according to the day-length.

(A) agrp1 expression in the brain of juvenile female medaka. (p = 0.4179) (B) npyb expression in the brain of juvenile female medaka. (p = 0.3357) LD: orange, n = 8; SD: blue, n = 7. The upper, middle, and lower bars show the third quartile, median, and the first quartile, respectively. Mann-Whitney U test. n.s., not significant.

Estrogen, which is released from mature ovary, may affect the expression of agrp1

Among various factors associated with sexual maturity, estrogens are known to be abundantly released from mature ovary and play important roles in reproductive readiness, sexual behavior, and so on (Jennings & de Lecea, 2020; Naftolin et al., 2007; Adachi et al., 2007; Clarkson & Herbison, 2009; Wintermantel et al., 2006; Micevych & Meisel, 2017; Melo & Ramsdell, 2001). Among the ovarian estrogens, 17β-estradiol (E2) is the major factor important for reproduction (Kanda et al., 2011; Kelly & Qiu, 2010), and the blood E2 concentration of LD-conditioned female medaka is higher than those of SD (Ikegami et al., 2022). Thus, we hypothesized that E2 regulates the expression of agrp1 and npyb under the condition of breeding season. We analyzed the expression of agrp1 and npyb in sham-operated (Sham), ovariectomized (OVX, fish with surgical ablation of the ovary), and OVX medaka with E2-administration (OVX+E) (Figure 5A-5B). The OVX medaka were allowed to survive at least for two weeks to clear the endogenous E2 (Kayo et al., 2020), and spawning of all the Sham medaka were confirmed for three consecutive days. By using whole-brain RT-qPCR, we found that OVX induces significantly lowered agrp1 expression than Sham, and OVX+E shows a tendency to recover agrp1 expression compared with OVX (Figure 5A), which suggests that the ovarian E2 regulates agrp1 expression. On the other hand, the expression levels of npyb did not show significant differences among the three groups (Figure 5B). Therefore, we focused more on the estrogenic regulation of agrp1 expression.

Estrogen, which is secreted from mature ovary, may affect the expression of agrp1.

(A) Relative expression of agrp1 in Sham (white; n = 6), OVX (Ovariectomized medaka, gray; n = 6), and OVX + E (OVX medaka kept in the tank containing E2, yellow; n = 7). Sham vs OVX: p = 0.04310, OVX vs OVX+E: p = 0.05790, Sham vs OVX+E: p = 0.2000. (B) Relative expression of npyb in Sham, OVX, and OVX + E. Sham vs OVX: p = 0.1386, OVX vs OVX+E: p = 0.9991, Sham vs OVX+E: p = 0.0812. The upper, middle, and lower bars show the third quartile, median, and the first quartile, respectively. Steel-Dwass test, * p < 0.05. n.s., not significant. (C) Photographs of brain slices after experiments of double in situ hybridization (agrp1 (green) and estrogen receptors (magenta)). (D) Expanded photograph of agrp1 and esr2a co-expressing neurons (white arrowhead). Scale bars: 50 µm.

Estrogens act mainly by interacting with estrogen receptors (Chen et al., 2022). Medaka has three kinds of estrogen receptors; esr1, esr2a and esr2b (Tohyama et al., 2016; Kinoshita et al., 2009), and all of them have been reported to be expressed in NVT (Zempo et al., 2013), in which agrp1 was also expressed (Figure 2B). Therefore, we examined co-expression of these esr genes and agrp1. As shown in Figure 5C and 5D, esr2a signal was clearly co-expressed in some agrp1-expressing neurons, which strongly suggests that E2 affects the expression of agrp1 via esr2a in those neurons of NVT.

agrp1−/− female medaka show a decrease in food intake and in the number of fertilized eggs

Our present experimental evidence thus far highlights the importance of agrp1 as the factor modulating the season-dependent feeding behavior in medaka. To analyze the function of AgRP1 in medaka, we generated knockout medaka of agrp1 (agrp1−/−) by using CRISPR/Cas9. The designed CRISPR guide RNA cleaved targeted sites of agrp1 (exon3, Figure 6-Supplementary Figure 1A), and we obtained agrp1−/− medaka, which has lots of amino acid changes in functional site for AgRP1 (Figure 6-Supplementary Figure 1B). First, the agrp1−/− female medaka appeared skinny and the body weight was significantly lower than that of agrp1+/+ (Figure 6A). In addition, abdominal height of agrp1−/− was also smaller than that of agrp1+/+, while the body length was not significantly different (Figure 6A). We next analyzed food intake of agrp1−/− female medaka in LD condition (breeding). As shown in Figure 6B, we found that LD agrp1−/− female medaka eat less than agrp1+/+. Then, we kept agrp1−/− medaka in LD or SD condition and compared their food intake. In contrast to agrp1+/+, agrp1−/− in LD condition did not show a significant increase in food intake compared with SD (Figure 6C). We also examined whether loss of AgRP1 affects reproductive function. Whereas the agrp1−/− females were fertile, the pairs of agrp1−/− female and agrp1+/+ male resulted in fewer spawned eggs than agrp1+/+ pairs (Figure 6D). In addition, the ovarian size of agrp1−/− appeared to be smaller than agrp1+/+ (Figure 6E left). In particular, since relative ovarian weight normalized by body weight (Gonadosomatic index: GSI) of agrp1−/− female tended to be marginally smaller than agrp1+/+ (Figure 6E right), the smaller body size of agrp1−/− (Figure 6A) may drastically affect ovarian morphology. Since the number of spawned eggs was decreased in agrp1−/− female, we analyzed gene expressions of gonadotropins which should affect ovarian maturation. Oocyte maturation and ovulation are known to be regulated by gonadotropins, follicular-stimulating hormone (FSH) and luteinizing hormone (LH). As shown in Figure 6F, agrp1−/− females showed lower levels of expression of gonadotropin genes (fshb and lhb), which suggests that loss of function of agrp1 impaired breeding season-dependent feeding behavior and led to attenuation of reproductive functions, especially the decrease in number of spawned eggs in the breeding season.

agrp1−/− female medaka show decrease in food intake and the number of fertilized eggs.

(A) Lateral views of representative agrp1+/+ and agrp1−/− female medaka (Left), and body weight, body length, and abdominal height (right) of agrp1+/+ (white; n = 6) and agrp1−/− (gray; n = 6). All the fish are 4-month-old adult medaka. Scale bar: 1 cm. body weight: p = 0.04113; body length: p = 0.3939; abdominal length: p = 0.002165. (B) Food intake (10 min) of agrp1+/+ (white; n = 6) and agrp1−/− (gray; n = 6) female medaka. Each amount of food intake is normalized by the average number of that of agrp1+/+ medaka. (p = 0.004329) (C) Food intake (10 min) of LD agrp1−/− (white; n = 9) and SD agrp1−/− (gray; n = 7) female medaka. Each amount of food intake is normalized by the average number of that of LD agrp1−/− medaka. (p = 0.5953) (D) The number of eggs spawned by agrp1+/+ (white; n = 6) and agrp1−/− (gray; n = 5) female medaka in a day. Each female was paired with a wildtype male. (p = 0.008658). (E) Photograph of ovary in representative agrp1+/+ and agrp1−/− female (left) and the gonado-somatic index (GSI, right). n = 6 of each group. (p = 0.06494) Scale bar: 1 mm. (F) Expression of gonadotropin genes (lhb and fshb) in the pituitary of agrp1+/+ (white; n = 6) and agrp1−/− (gray; n = 6) female medaka. lhb: p = 0.008658); fshb (p = 0.02597). The upper, middle, and lower bars show the third quartile, median, and the first quartile, respectively. Mann-Whitney U test, * p < 0.05, ** p < 0.01. n.s., not significant.

Discussion

In the present study, we took advantage of female medaka which clearly shows breeding season-dependent feeding behavior and found that neuropeptides, agrp1 and npyb, show higher expression under the breeding condition than under the non-breeding condition. We also obtained results to suggest that the expression of both agrp1 and npyb changes depending on nutritional status of female medaka. In addition, ovariectomy and E2 administration changed expression of agrp1 but not npyb, suggesting that increased release of ovarian E2 in breeding season upregulates the agrp1 expression, which results in facilitation of female feeding behavior. Finally, loss-of-function mutation of AgRP1 decreased the amount of food intake and the number of spawned eggs. The present results suggest that breeding season-dependent feeding behavior can be modulated by the increased expression of agrp1 upregulated by increased release of ovarian estrogen in the breeding season (Figure 7). To date, not a few previous reports have shown the influence of nutritional status on reproduction (Evans & Anderson, 2012; Kalra & Kalra, 1996; Kirkwood et al., 1987; Merry & Holehan, 1979; Hasebe et al., 2016). On the other hand, although seasonal breeders have been reported to show changes in feeding behavior during the breeding season (Tsuyuki, 2018; Kawai et al., 2020), its neuroendocrine mechanism have largely remained enigmatic. Our present results may provide a neuroendocrinological model for the mechanisms that play a key role in the control of breeding season-dependent feeding behavior in teleosts.

Illustration of mechanisms of breeding season-dependent feeding behavior in medaka suggested by the present study

(A) Long day (LD) condition in breeding season induces ovarian maturation, which facilitates release of estrogen (E2) from the mature ovary. (B) High concentration of serum E2 increases agrp1 expression in the brain via the estrogen receptors, especially, esr2a. (C) Higher expression of AgRP1 is suggested to activate neural circuitry for feeding, which leads to an increase in food intake and egg spawning of female medaka in the breeding season.

Feeding-related peptides AgRP and NPY in medaka

Here, we demonstrated that female medaka eat more under the condition of breeding season (Figure 1B). Various kinds of neuropeptides in the brain have been suggested to control feeding, and these are generally called “feeding-related peptides” (Funahashi et al., 2003). In the present study, we first used a seasonally breeding model teleost medaka and searched for the “feeding-related peptides” involved in seasonal feeding behavior. A whole-brain RNA-seq analysis using female medaka under the breeding condition (LD) and non-breeding condition (SD) revealed two kinds of feeding-related peptides, agrp1 and npyb, which show different expression levels between LD and SD (Figure 1-Supprementary Figure 2). In mammals, AgRP and NPY are known to have orexigenic function and are co-expressed in hypothalamic neurons (Schwartz et al., 2000; Hahn et al., 1998). Previous studies in mammals (Schwartz et al., 2000; Andermann & Lowell, 2017; Muroi & Ishii, 2016) have suggested neural mechanisms of appetite including functions of AgRP and NPY. However, such mechanisms in non-mammalian vertebrates such as teleosts (Rønnestad et al., 2017; Blanco & Soengas, 2021) have not yet been clarified. Our present study using medaka have shown possible functions of AgRP and NPY in teleost feeding behavior, especially in a breeding season-dependent manner.

Our present study using medaka showed that female medaka express agrp1 in hypothalamus, and food restriction increases the agrp1 expression (Figure 2B and 2G). It has been reported that leptin receptor-knockout medaka show higher food intake and higher expressions of agrp1 and npya than wild type, whereas the expression of agrp2 and npyb remained to be analyzed (Chisada et al., 2014). Zebrafish has also been used as a model animal in teleosts. In zebrafish, food restriction increased agrp1 (Song et al., 2003; Opazo et al., 2018) expression, and transgenic overexpression of agrp1 caused gain of body weight (Song & Cone, 2007), as in mammals (Graham et al., 1997; Adam et al., 2002; Hahn et al., 1998; Ilnytska & Argyropoulos, 2008). It has also been reported that agrp1 knockout zebrafish eat less than the wild type (Shainer et al., 2019), although loss of AgRP in mice showed little effect on food intake (Qian et al., 2002). Our present results and these previous studies strongly support that agrp1 regulates feeding and may act as an orexigenic factor in teleosts. On the other hand, agrp2 neurons were distributed in telencephalon (Figure 2C), which is different from results in zebrafish (Shainer et al., 2019), and food restriction did not significantly affect the agrp2 expression in medaka. It is therefore suggested that agrp2 is not involved in feeding in female medaka.

Furthermore, we demonstrated that npya is expressed in multiple brain regions including hypothalamus in medaka (Figure 2D), which is similar to mammals (Gray & Morley, 1986) and zebrafish (Yokobori et al., 2012; Jeong et al., 2018). We also showed that npya is not co-expressed with agrp1 (Figure 2F) as in the zebrafish (Jeong et al., 2018), suggesting that The relationship between NPY and AgRP of teleosts may be different from that of mammals, in which most of the agrp-expressing neurons co-express npy (Hahn et al., 1998). Moreover, our present study also showed that npyb expression is localized in telencephalon (Figure 2E), which is similar to the previous report using tiger puffer (Kamijo et al., 2011). Previous studies of NPY in zebrafish showed that zebrafish has only one type of NPY (NPYa) (Söderberg et al., 2000; Larsson et al., 2009) and has lost NPYb during evolution. Like other mammals (Marks et al., 1992; Clark et al., 1984; Glenn Stanley et al., 1986; Marks & Waite, 1997; Baldock et al., 2009), food restriction in zebrafish increased the npya expression in the hypothalamus (Song et al., 2003; Opazo et al., 2018), and intracerebroventricular administration of NPYa increased food intake (Yokobori et al., 2012). Although these zebrafish studies suggest that NPYa may increase food intake, it is still debatable, since body weight was not significantly different between npya knockout zebrafish and wild type (Shiozaki et al., 2020). Interestingly, by analyzing both npya and npyb expression in medaka of different nutritional conditions, we found that food restriction decrease the npya-expressing cell number in the hypothalamus (Figure 2-Supplementary Figure 1) and npyb expression level (Figure 2J). These changes in npya and npyb expressions are not consistent with previous studies using other conventional model animals described above (Yokobori et al., 2012; Marks et al., 1992; Clark et al., 1984; Glenn Stanley et al., 1986; Marks & Waite, 1997; Baldock et al., 2009). Our present study may suggest that the function of npy may be different among teleosts. In addition, npyb expression was increased under the breeding condition (LD), while LD female showed increase in food intake (Figure 3F). Thus, decreased expression of npyb by food restriction (Figure 2J) may suggest that the change in npyb expression reflects nutritional condition in medaka. Thus, future study of npya and npyb functions in the control of feeding will be necessary.

As described above, we found that agrp1, npya, and npyb change expression levels in response to nutritional status. Among these three genes, we suggest that agrp1 most probably affects relatively long-term feeding in breeding season, which agrees well with the recent studies in mice showing the function of AgRP as a long-term orexigenic factor. In mice, it has been reported that intracerebroventricular administration of AgRP increases food intake for one week (Hagan et al., 2000), and stimulation of receptors expressed in AgRP neurons triggers AgRP release, leading to an increase in food intake for three days (Nakajima et al., 2016). Our present study also suggests a long-term (seasonal) orexigenic effect of AgRP in teleosts and may also provide an important insight into the understanding of common regulatory mechanisms of feeding by AgRP among various animal species.

High concentration of E2 in breeding season facilitates agrp1 expression

Our results suggest that agrp1 and npyb show higher expressions under the breeding condition (LD) (Figure 3), but the experiments using juvenile female medaka (Figure 4) showed that expression levels of these two genes do not change according to the day-length itself but to the LD-induced sexual maturity. In addition, the present results indicates that the ovarian estrogen E2 upregulates agrp1 expression mainly via the estrogen receptors esr2a that are co-expressed in some population of agrp1 neurons in the hypothalamic nucleus NVT (Figure 5). Since LD female medaka (breeding) shows high blood concentration of E2 (Ikegami et al., 2022), this pathway may be important for breeding season-dependent feeding behavior. Especially, in teleost, main egg protein for nutrition is vitellogenin, whose expression is also facilitated by E2 (Tohyama et al., 2017). Taken together, it is suggested that E2 may synchronously regulate amount of food intake and female-specific reproductive signals (vitellogenin production and oocyte maturation), which plays a key role in reproductive success in oviparous animals.

In mammals, previous studies have reported on inconsistent effects of ovary and E2 on feeding.

Ablation of ovary caused suppression of food intake in mice (Yu et al., 2020), whereas it caused no change in rats (Roesch, 2006). On the other hand, administration of E2 decreased food intake in both mice (Yu et al., 2020) and rats (Roesch, 2006). It should be noted that these laboratory rodents only exhibit short estrous cyclicity and have lost breeding seasonality, and the blood E2 concentration drastically changes in a few days (Nilsson et al., 2015). Thus, it is possible that the control mechanisms of feeding may be different between animals with short estrous cyclicity and those with breeding seasonality.

Furthermore, our present study suggests different control mechanisms of feeding between the animals with breeding seasonally and those without. Here, we showed that E2 directly modulates agrp1 expression via esr2a receptors co-expressed in the agrp1 neurons (Figure 5C), while in mice, AgRP/NPY neurons are reported to be suppressed by E2 indirectly via esr1 (erα)-expressing Kiss1 neurons located in the hypothalamic arcuate nucleus (Qiu et al., 2018; Dubois et al., 2016; Yang et al., 2017). On the other hand, in medaka, expressions of all kinds of estrogen receptors are reported to be localized in NVT (Zempo et al., 2013), in which agrp1 expression is also localized (Figure 2B). In addition, esr2a has been reported to be involved in the feedback regulation of follicle stimulating hormone in the pituitary and in the development of oviduct, and esr2a knockout females are completely infertile (Kayo et al., 2019). Our hypothesis that estrogen signaling via esr2a affects agrp1 expression may highlight another important function of esr2a for reproduction, while a possibility still remains that esr1- and esr2b-expressing neurons also affect agrp1 expression indirectly.

AgRP1 changes feeding behavior depending on LD-induced sexual maturity, which causes the increase in food intake in breeding season

Since our results thus far indicate the importance of agrp1, which shows upregulated expression directly stimulated by the ovarian E2 in breeding season, we examined phenotypes of agrp1 knockout (agrp1−/−) medaka. We found that agrp1−/− medaka under the condition of breeding season eat less (Figure 6B) and spawn a smaller number of eggs (Figure 6D) than WT. Furthermore, the agrp1−/− females did not show significant difference of food intake in LD and SD (Figure 6C). These results strengthen our hypothesis that agrp1 is involved in the increased food intake in breeding season. Furthermore, agrp1−/− female displayed light body weight (Figure 6A), accompanied by smaller ovary (Figure 6E) and low level of expression of gonadotropins, fshb and lhb (Figure 6F), which are considered to have caused smaller number of spawned eggs. All of these results support our hypothesis that AgRP1 plays an important role in breeding season-dependent feeding behavior, which culminates in normal reproduction.

In summary, by using a seasonal breeder medaka, we found evidence to suggest that long day-length facilitates ovarian maturation and E2 release, which upregulates agrp1 expression of hypothalamic neurons to activate neural circuitry for feeding behavior and boost oocyte maturation. We propose that this kind of positive feedback control may be important for animals that spawn many eggs every day in breeding season (Figure 7); medaka needs plenty of food for production of many eggs. Indeed, previous study showed a need for high food intake for reproduction (Hasebe et al., 2016). It is expected that future studies will elucidate whether or not the present findings in medaka are applicable to other seasonal breeders as well.

Material and Methods

Animals

Female and male wild-type d-rR medaka (Oryzias latipes) and agrp1 knockout (agrp1−/−) medaka were maintained in pairs or shoals at 27℃. Fish were fed three times a day with brine shrimp and flake food (Otohime B-2; San-u Fish Farm, Osaka, Japan). Their reproductive status was controlled by day-length (long day (LD; 14-hour light/10-hour dark): reproductive, short day (SD; 10-hour light/14-hour dark): non-reproductive). The light-on time was 8:00 AM. We used juvenile medaka (∼5 weeks after fertilization) and adult medaka (>3 months after fertilization). Female medaka, which spawned at least three consecutive days, were used as reproductive ones. For the analysis of the effect of food restriction, reproductive and non-reproductive female medaka were fasted for 14 days. Note that all medaka survived after food restriction. For the analyses of food intake, we food-restricted medaka for 6 hours after 10-min feeding in the morning and sampled their whole brains for subsequent experiments. 2-week food-restricted medaka were sampled at the same time as the other fed medaka. All experiments and fish maintenance were conducted in accordance with the Guidelines for Proper Conduct of Animal Experiments (Science Council of Japan) and the protocols approved by the Animal Care and Use Committee of Graduate School of Science, the University of Tokyo (Permission number, 17-1, 20-6) and the Animal Care and Use Committee of Graduate School of Agriculture, Tokyo University of Agriculture and Technology (Permission number, R05-15, R06-27).

Food intake assay

Each 6-hour food-restricted medaka was put into a white cup with 100 mL breeding water and was habituated for 5 min. Then, we fed medaka by application of 200 μL aliquots of food water containing brine shrimp in all-you-can-eat style and serve another aliquot once done with it, which is repeated N times (like the Japanese “Wanko soba”; so-called Japanese “Wanko soba” method). 10 min after the start, we stopped feeding medaka, took 10 mL out of the breeding water and transferred it to a conical tube. The conical tube was frozen overnight, and the leftover brine shrimp sunk in the bottom were counted by “shrimp-counter”. We counted the number of brine shrimp in the 200 μL solution three times before and after the experiments, and the average number was used. The food intake was calculated as follows.

(Food Intake)

= (The average number of brine shrimp in the solution) * (number of aliquots, N) - (number of leftover brine shrimp sunk in the bottom) * 10

“Shrimp-counter” system

The number of shrimps in the solution was counted using OpenCV3 library (Intel, Santa Clara, CA) run under a python script (Figure 1-Source code 1). This script was run under Anaconda 4.4.0 for windows running python 3.5.

RNA-sequencing

We collected two whole brains of LD or SD female in one tube (Note that pituitary was confirmed not to be included), and extracted total RNA by using NucleoSpin® RNA Plus kit (MACHEREY-NAGEL, Düren, Germany). cDNA was obtained by KAPA Stranded mRNA-Seq Kit (Nippon Genetics Co., Ltd, Tokyo, Japan) and KAPA Library preparation kit (Nippon Genetics Co., Ltd). Then, it was applied to a next generation sequencer Hiseq 2500 (Illumina, San Diego, CA), following the standard protocol of Illumina system. We selected the candidate genes judging from transcripts per million (TPM) for expression value in the obtained data using CLC workbench. We made volcano plot using R (R core team 2023) and Rstudio (2023) and colored dots, which indicate P-value < 0.05 and |log FC| > 1. In addition, we made a heatmap of genes related to neuroendocrine system using DESeq2 (Love et al., 2014).

Histological analysis of the distribution of agrp- and npy-neurons in the brain

To analyze the distribution of agrp- and npy-expressing neurons, we performed in situ hybridization (ISH) for agrp and npy on frozen sections of reproductive medaka. In brief, medaka was anesthetized (FA100, Bussan Animal Health Co., Ltd., Osaka, Japan), and its brain was picked up and fixed with 4% paraformaldehyde (PFA)/PBS. After incubation with 30% sucrose/PBS, brains were embedded in 5% low melting agar/ 20% sucrose/ PBS and sectioned at a thickness of 25 µm. The sections were hybridized with agrp1 (ENSORLG00000000398, 177 bases), agrp2 (ENSORLG00000029106, 303 bases), npya (ENSORLG00000004649, 288 bases) and npyb (ENSORLG00000007880, 288 bases) specific digoxigenin (DIG)-labeled RNA probes and performed nitro blue tetrazolium (NBT)/ 5-bromo-4-chloro-3-indolyl-phosphate color-reaction (BCIP) after wash and incubation with anti-digoxigenin antibody (catalog. no. 11093274910; Roche; RRID: AB_514497) as previously reported (Zempo et al., 2013). Photographs were taken with a digital camera (DFC310FX; Leica Microsystems, Wetzlar, Germany) attached to an upright microscope (DM5000B; Leica Microsystems).

Histological analysis of agrp1-, npy- and estrogen receptor (esr)-expressing neurons

To examine whether agrp1-expressing neurons co-express npya and esr, we prepared agrp1 fluorescein-labeled RNA probe and carried out double ISH as previously reported (Umatani et al., 2022). esr DIG-labeled probes were kindly given by Dr. Kayo (Kyoto Univ.), and we used npya DIG-labeled probe described in the previous paragraph. In brief, we made brain sections as described above and applied both agrp1 fluorescein-labeled and each DIG-labeled RNA probes. Signals for npya, esr1, esr2a and esr2b were visualized by incubation with anti-digoxigenin antibody (catalog. no. 11207733910; Roche; RRID: AB_514500) and TSA Plus Cy3 System (TSA-Plus Cyanine 3 system, catalog. no. NEL744001KT, Akoya Biosciences, Marlborough, MA). After inactivation of Cy3 system by 3% H2O2, we applied peroxidase-conjugated anti-fluorescein antibody (catalog. no. 11426346910, Roche; RRID: AB_840257) on sections and performed TSA Plus biotin system (catalog. no. NEL749A001KT, Akoya Biosciences). Then, signals for agrp1 were visualized by Alexa 488 conjugated streptavidin (catalog. no. S11223, Invitrogen). For counter-staining of cell nuclei, DAPI in PBS was applied on section. Fluorescent images were acquired with a confocal laser-scanning microscope (AXR, Nikon, Tokyo, Japan) using excitation and emission wavelengths of 405 nm and 429-474 nm for DAPI, 488 nm and 512-526 nm for Alexa 488, and 561 nm and 571-625 nm for Cy3, respectively. These were photographed at Tokyo University of Agriculture and Technology for Smart Core facility Promotion Organization.

Quantitative Real-time Polymerase Chain Reaction (RT-qPCR)

A Whole brain or a pituitary was collected from each medaka and total RNA was extracted by using FastGeneTM RNA basic kit (Nippon Genetics Co., Ltd) according to the manufacturer’s instructions. For the juvenile medaka, we checked their sex as previously reported, and two samples of the same sex were mixed and used as one sample. Total RNA samples were reverse transcribed by FastGeneTM cDNA synthesis 5× ReadyMix OdT according to the manufacturer’s instructions. For the analyses of the brain, 1μL of cDNA diluted with 10-fold MQ was mixed with KAPA SYBR Fast qPCR kit and amplified with Lightcycler96 [Roche; 95℃ 150s (95℃ 10 sec, 60℃ 10 sec, 72℃ 15 sec) × 45 cycles]. For the analysis of the pituitary, 1 μL of cDNA diluted with 5-fold MQ was mixed with KAPA SYBR Fast qPCR kit and amplified with Lightcycler96 [Roche; 95℃ 150s (95℃ 10 sec, 60℃ 10 sec, 72℃ 10 sec) × 45 cycles]. The data was normalized by housekeeping gene, ribosomal protein s13 (rps13). Primer sequences were as follows:

AgRP1 RT-PCR F1 CCAATTTCCAGTCACCGAAG

AgRP1 RT-PCR R1 CTGGGTCCAACACAGAATCA

AgRP2 RT-PCR F1 TTGTTGTGCTTCTTGCTGCT

AgRP2 RT-PCR R1 ACAGAGCTCCAAACGGTGTC

NPYa SE CTCATCACAAGACAGAGGTATGGG

NPYa AS GGGTTGTAACTTGACTGTGGAAGTG

NPYb SE CTGCCTGCTCCTCTGTTTTTTCTC

NPYb AS CACAGTGTCTGGGTTGTCTCTCTTTC

qPCR FSHb Fw new TGGAGATCTACAGGCGTCGGTAC

qPCR FSHb Fw new AGCTCTCCACAGGGATGCTG

qPCR LHb Fw new AGGGTATGTGACTGACGGATCCAC

qPCR LHb Rv new TGCCTTACCAAGGACCCCTTGATG

RPS13 SE GTGTTCCCACTTGGCTCAAGC

RPS13 AS CACCAATTTGAGAGGGAGTGAGAC

Sham operations, ovariectomy, and E2 administrations

Ovariectomy and E2 administration were performed according to a previous study (Kayo et al., 2020). Briefly, reproductive female medaka were anesthetized with 0.02% MS-222 (Sigma-Aldrich, St. Louis, MO) and their ovaries were excised via intraperitoneal operation. After checking that all Sham females spawn, we prepared three tanks; two tanks contained 7-8 OVX medaka, and one tank contained Sham medaka in 2 L breeding water in it. We dissolved β-Estradiol 1.4 mg in 1 mL EtOH (E2 stock) and dispensed 2 μL of E2 stock or the same amount of 100% Ethanol for the control tank. Ethanol or E2 containing water were changed every day. After the steroid treatment for 5 days, the medaka were anesthetized, and their whole brains were collected for RT-qPCR analysis.

Generation of agrp1 KO medaka lines

We generated agrp1 KO medaka lines by using CRISPR/Cas9. Cas9 mRNA and tracer RNA were purchased from Integrated DNA Technologies (IDT, Coralville, IA). The guide RNA sequence for digestion by CRISPR/Cas9 complex was “CCTCACCAGCAGTCCTGCCTGG”.

Mixture of Cas9 protein, tracer RNA, CRISPR RNA, GFP mRNA diluted with PBS and 0.02% phenol red (final concentration: Cas9 protein; 500 ng/μL, tracer RNA; 100 ng/μL, CRISPR RNA; 50 ng/μL, GFP mRNA; 5 ng/μL) was injected into the cytoplasm of one- or two-cell-stage embryos (F0). To obtain homozygous transgenic offspring, the carriers were crossed with each other.

Measurement of body size and ovary size

We took photograph of fish body from the lateral side by using a digital camera MC120HD (Leica) and calculated the abdominal and body length by ImageJ. For the analysis of body length, we measured the length from the mouth to the base of the tail. Gonadosomatic index (GSI) was calculated as ovary weight/ body weight*100.

Statistics

For the statistical analysis, we used Kyplot 5.0 software (Kyence, Tokyo, Japan) or R software (R core team 2023) with R studio (version 2023.06.0+421). For the comparison of the TPM, we used Student’s t-test. For the comparison of agrp1 and npyb expressions in OVX and E2 administrated medaka, Steel-Dwass test was used for multiple comparison of the expression level. In the other experiments, we used Mann-Whitney U test. In all statistical analysis, significance levels were described as follows: *p < 0.05, **p < 0.01, and ***p < 0.001.

Data availability

All data supporting the findings of our present study and its supplementary information are available in the article or from the corresponding author upon reasonable request.

Code availability

The code of “Shrimp-counter” system is available in the present study, Figure 1-Source code 1 (see Methods).

Acknowledgements

We thank Drs. Mikoto Nakajo (Osaka Med and Pharm Univ), Soma Tomihara (Nagahama Inst of Bio-Sci and Tech), and Kana Ikegami (Kitasato Univ) for helpful discussion. We also thank Dr. Daichi Kayo (Kyoto Univ.) for his kind supply of esr DIG-labeled probes. In addition, we deeply appreciate Dr. Hiroyuki Takeda (U Tokyo) for his continued kind support and encouragement during our experiments. We also thank Dr. Yutaka Miura (Tokyo Univ of Agri and Tech) for his kind support of our experiments. As for the help of animal care, we thank Ms. Hisako Kohno, Miho Kyokuwa, Hiroko Tsukamoto, and Risa Nakaba. We thank Ms. Maiko Matsuda (Vesper studio) for her kind gift of medaka illustrations to us. The present study was supported by JSPS KAKENHI (JP21J20864 and JP22KJ0597 to YT, JP26221104 and JP21K06262 to YO, JP20H03071 to CU).

Author contributions

Conceptualization: Y.T. and C.U. ; Investigation: Y.T., S.T., H.W., S.Ki., S.Ka, and C.U. ; Methodology: Y.T., H.W., T.K., S.Ka., and C.U. ; Resources: Y.O., Y.H., and C.U.; Writing—original draft: Y.T. and C.U.; Writing—review and editing: Y.T., S.Ka., Y.O., Y.H., and C.U.; Project administration: C.U.; Supervision: S.Ka., Y.O., Y.H., and C.U.; and Funding acquisition: Y.T., S.Ka., Y.O., Y.H., and C.U.

Competing interests

The authors declare no competing interests.