Midgut CCHa1 regulates feeding behaviour in female Drosophila melanogaster
The initial aim of this study was to examine which enteroendocrine hormones influence food consumption in D. melanogaster. For this purpose, we knocked down each of the identified enteroendocrine hormones in adult EECs and quantified food consumption for 24 h using the capillary feeder assay (CAFÉ) assay. For adult EEC-specific knockdown, we conducted transgenic RNAi using a GAL4 driver (prosV1-GAL4) coupled with tub-GAL80ts (hereafter prosts-GAL4) (Fig. 1a). In this study, we primarily focused on adult virgin females.
Drosophila melanogaster EECs produce multiple peptide hormones, such as Allatostatin A (AstA), Allatostatin C (AstC), Bursicon α (Bursα), CCHa1, CCHamide2 (CCHa2), Diuretic hormone 31 (Dh31), Neuropeptide F (NPF), Orcokinin, and Tachykinin (Tk)19. Among the eight peptide hormones examined, knockdown of CCHa1 and AstA resulted in a significant increase in food intake (Fig. 1b). In subsequent analyses, we decided to focus on CCHa1 because of the remarkable phenotype of the RNAi animals. The hyperphagic phenotype was also observed with the other CCHa1 RNAi line (CCHa1RNAiKK) that targeted a different region of CCHa1 mRNA (Fig. 1c). We also confirmed that knockdown of CCHa1 with prosts-GAL4 suppresses CCHa1 mRNA expression in a gut-specific manner (Fig. 1d, Supplementary Fig. 1a,b). These results suggest that the food intake phenotype is due to the loss of CCHa1 function in EECs.
We further ruled out the possibility that CCHa1 in the central nervous system (CNS), which regulates circadian behavioural rhythms20, affects feeding behaviour. Combining a transgene otd-nls::flp that expresses FLP recombinase only in the brain with tub > FRT > GAL80 > FRT, we were able to restrict nSyb-GAL4 activity to the brain (referred to as nSybBrain-GAL4) (Supplementary Fig. 1c). nSybBrain-GAL4 allowed us to suppress most CCHa1 expression in the brain, but not in EECs (Supplementary Fig. 1c). Nevertheless, the brain-specific knockdown of CCHa1 with nSybBrain-GAL4 did not promote feeding behaviour, indicating that CCHa1 in the brain does not regulate feeding (Fig. 1c). Furthermore, we found that knocking down CCHa1 in the posterior region (R4-5) of the EECs with R65D05-GAL4, a GAL4 driven by a fragment sequence upstream of AstA, increased food intake in a manner similar to prosts-GAL4-dependent CCHa1 knockdown (Fig. 1c). Notably, R65D05-GAL4-mediated CCHa1 RNAi did not affect CCHa1 protein levels in the CNS (Extended Data Fig. 1d), supporting our idea that gut-derived CCHa1 regulates feeding behaviour.
Since the food used for the standard CAFÉ assay contained a mixture of sucrose and yeast autolysate, we conducted a two-choice CAFÉ assay utilizing a solution containing either sucrose or yeast autolysate. CCHa1 RNAi animals consumed more yeast autolysate than control animals, while their sucrose consumption remained unaffected (Fig. 1e), suggesting that the loss of gut-derived CCHa1 selectively impacts appetite regulation for yeast. Since yeast is a source of dietary protein for D. melanogaster, we also examined dietary protein consumption using tryptone, a digested casein, to directly evaluate appetite against proteins. CCHa1 RNAi animals consumed more tryptone than control animals, indicating that gut-derived CCHa1 regulates protein satiety response (Fig. 1f). A short-term feeding experiment with FlyPAD7 for 60 min corroborated these results by demonstrating a similar trend in increased protein feeding in CCHa1 knockdown (Fig. 1g). Thus, our findings imply that gut-derived CCHa1 is pivotal in regulating the protein satiety response in D. melanogaster.
Recently, it became evident that the metabolic and behavioural characteristics of D. melanogaster exhibit sexual dimorphism21–23. We thus investigated whether the loss of CCHa1 in the gut impacts behavioural phenotypes in male flies. Similar to female flies described above, gut specific CCHa1 knockdown males displayed an increase in feeding compared to controls (Extended Data Fig. 1e). However, contrary to the response in females, gut-specific CCHa1 knockdown in males did not result in a preferential shift to high-protein consumption (Extended Data Fig. 1f). Therefore, these results suggest that CCHa1 in the gut regulates feeding behaviour in both male and female flies, but there appears to be sexual dimorphism, at least in preference.
CCHa1-producing EECs respond to high-protein diets
Recent studies on D. melanogaster have reported that specific nutrients have the ability to stimulate EECs, leading to secretion of enteroendocrine hormones to affect downstream target organs11,13. To investigate effects of protein and sugar on activity of CCHa1+ EECs, we utilized the calcium reporter system, CaLexA24, to facilitate cumulative tracing of Ca2+ signalling activity. Compared with the standard diet, the glucose-only diet did not affect activity of EECs in the midgut (Fig. 2a, b). However, a high-protein diet (HPD), such as a yeast-only diet, 4% yeast-supplemented diet, and 10% peptone- or tryptone-supplemented diet, increased activity of CCHa1+ EECs located in the posterior midgut (Fig. 2a, b). Furthermore, CCHa1+ EECs in the posterior midgut were activated by a yeast-rich, low-carbohydrate diet, indicating that CCHa1+ EECs are more sensitive to the higher yeast ratio in the diet (Extended Data Fig. 2a, b). This HPD-dependent activation of CCHa1+ EECs is consistent with a recent study showing that HPD feeding activates CCHa1+ EECs and affects arousability through CCHa1+ EECs in male flies16. Interestingly, we observed a slight negative correlation between the intensity of GFP and anti-CCHa1 antibody signals in the posterior midgut, suggesting that EEC activation triggers CCHa1 secretion from EECs (Extended Data Fig. 2a-g). In addition, the HPD increased CCHa1 mRNA in the gut (Fig. 2c). These results indicate that CCHa1+ EECs in the posterior midgut respond to an HPD.
To investigate whether manipulation of EEC activity leads to similar effects as CCHa1 RNAi in the gut, we inhibited cellular activity of CCHa1+ EECs by overexpressing the temperature-sensitive allele of dynamin, shibirets, which blocks synaptic transmission and vesicle endocytosis, specifically in CCHa1+ EECs. For this purpose, we employed R57C10-GAL80, which inhibits GAL4 activity in neurons. By combining CCHa1-T2A-knock-in-GAL4 with R57C10-GAL80, we could restrict GAL4 activity to just posterior EECs (Extended Data Fig. 3a). Using these genetic tools, we confirmed that inhibition of CCHa1+ EECs by shibirets results in accumulation of CCHa1 protein in EECs (Extended Data Fig. 3b). Additionally, inhibiting CCHa1 secretion increased feeding amount, and promoted protein feeding, as in CCHa1 RNAi (Extended Data Fig. 3c, d). These findings support our idea that CCHa1+ EECs regulate satiety in response to dietary proteins.
Appetite for proteins is regulated by several humoral factors, including Drosophila insulin-like peptides (Dilps) and female-specific independent of transformer (Fit) in adult female D. melanogaster8,25. Gut-specific CCHa1 RNAi animals showed a reduction of dilp2, dilp5, and fit mRNA expression in the head, compared to controls, whereas dilp3 mRNA expression was increased (Extended Data Fig. 3e). These data are consistent with previous results that the reduction of dilp2 or fit mRNA expression evokes a preference for protein-rich food8,25. In addition, expression of dilp5, which mRNA is reduced by starvation26,27, was also reduced in CCHa1 RNAi animals (Extended Data Fig. 3e). Together, these data suggest that CCHa1 in midgut EECs suppresses protein feeding by activating them in response to excessive dietary protein.
CCHa1 receptor in sNPF neurons regulates sugar/protein feeding balance
We next explored the role of the CCHa1 receptor (CCHa1-R) in controlling protein satiety. Consistent with Fly Cell Atlas data showing enrichment of CCHa1-R expression in the nervous system28, we found that CCHa1-R-T2A-knock-in-GAL4 was predominantly expressed in the nervous system (Fig. 3a). Moreover, nervous system-specific CCHa1-R knockdown by pan-neuronal GAL4 (nSyb-GAL4) exhibited a similar increase in feeding amount and feeding preference shift to gut-specific CCHa1 RNAi (Fig. 3b), suggesting that CCHa1 targets the nervous system to control feeding behaviour.
To identify a specific subset of CCHa1-R+ neurons responsible for suppressing protein feeding, we took a two-choice CAFÉ assay screen with several neuronal GAL4 drivers. Among the neuronal GAL4 lines tested, short NPF (sNPF)-GAL4 driver showed a significant preference shift as observed in nSyb > CCHa1-RRNAi (Fig. 3c). Importantly, knockdown of CCHa1-R in cells involved in feeding preference, such as insulin-producing cells (dilp2-GAL4)25 and corpora cardiaca (Akh-GAL4)13, did not result in a feeding preference shift (Fig. 3c). Moreover, CCHa1-R RNAi in a subset of CCHa1-R+ neurons in the mushroom body regulating sleep16 also exhibited no effects on feeding preference (Fig. 3c). We obtained a similar preference shift with another UAS-CCHa1-R RNAi line (UAS-CCHa1-RRNAiVDRC) (Fig. 3d, Extended Data Fig. 4a).
Since the sNPF-GAL4 labelled more than 100 neurons in the nervous system (Fig. 3e), we tried further restricting sNPF neurons. To this end, we conducted a second screen with sNPF-promoter GAL4 drivers. Whereas the feeding preference against sucrose and tryptone differ between the GAL4 lines, only the sNPFTH-GAL4 driver29 induced a shift in feeding preference from sucrose to tryptone by CCHa1-R knockdown using two RNAi lines targeting different mRNA sequence. (Fig. 3f, g). Therefore, we focused on the sNPFTH-GAL4+ neurons as a target of gut-derived CCHa1 to suppress protein feeding.
We next analysed the expression pattern of sNPFTH-GAL4. Intriguingly, the sNPFTH-GAL4 driver did not label any neurons in the brain or abdominal ganglion, while the GAL4 labelled two or three pairs of enteric neurons located in the hypocerebral ganglion (HCG) 30,31 (Fig. 4a, b). sNPFTH-GAL4+ enteric neurons were co-labelled with anti-sNPF/NPF antibodies. Furthermore, their signals disappeared in sNPF knockdown and sNPF null mutants, indicating that they are sNPF-producing neurons (Fig. 4a, b Extended Data Fig. 5a-c). Notably, enteric sNPF neurons were labelled with the CCHa1-R-T2A-knock-in-GAL4 driver, suggesting that CCHa1-R is expressed in enteric neurons (Extended Data Fig. 5d). Knocking down sNPF in enteric neurons or sNPF genetic null mutants resulted in a significant feeding preference shift, similar to loss of CCHa1/CCHa1-R signalling (Fig. 4c, Extended Data Fig. 5e). Moreover, sNPF knockdown increased yeast autolysate consumption in a short-term one-choice FlyPAD experiment (Fig. 4d). These data suggest that sNPF in CCHa1-R+ enteric neurons are crucial in suppressing protein feeding.
Consistent with the fact that HPD activates CCHa1+ EECs, we found that HPD leads to activation of enteric sNPF neurons using sNPFTH-GAL4-driven CaLexA > GFP flies (Fig. 4e, f). Moreover, CCHa1-R knockdown cancelled the increase in GFP signals in the HPD-fed condition, indicating that CCHa1-R signalling is required to activate enteric sNPF neurons (Fig. 4e, f). Moreover, suppression of the activity of enteric sNPF neurons by shibirets also showed a preference shift to protein (Fig. 4g).
The next question to be addressed is whether CCHa1 secreted from the gut acts directly on enteric sNPF neurons. Therefore, we used the knock-in-Tango system, which enables us to investigate receptor signalling activity32 (Fig. 4h). We employed CCHa1-R-knock-in-Tango with gut-specific CCHa1 knockdown and found that loss of gut-derived CCHa1 reduced CCHa1-R signalling activity in enteric sNPF neurons (Fig. 4i, j). To further explore this axis, we examined the effect of CCHa1 overexpression in EECs. EEC-specific CCHa1 overexpression evoked yeast feeding in the control (sNPF heterozygous mutant) genetic background, while the effect was not observed in sNPF null mutants (Fig. 4k, Extended Data Fig. 6a). These data support our idea that gut-derived CCHa1 suppresses protein feeding through enteric sNPF neurons.
Neuronal sNPF/sNPFR signalling is required for control of feeding preference
We next investigated how enteric sNPF neurons suppress protein feeding. Morphologically, enteric sNPF neurons project to the anterior part of the midgut and HCG neurons including sNPF-negative neurons (Fig. 5a, Extended Data Fig. 7a, b). The postsynaptic marker, DenMark33, was localized in neurites in the gut (Fig. 5a). Furthermore, the presynaptic marker, Syt::GFP, was observed in neurites not only on the gut, but also on the HCG, implying that enteric sNPF neurons stimulate both tissues (Fig. 5a). Expression of the sNPF receptor (sNPF-R) was confirmed by sNPF-R-T2A-knock-in-GAL4-driven nuclear GFP signals. The nuclear GFP was observed in both of sNPF-positive and negative HCG neurons, the brain, and the visceral muscle surrounding the anterior gut (Fig. 5b, Extended Data Fig. 7c). Knocking down of sNPF-R in the nervous system resulted in a shift in feeding preference, whereas sNPF-R RNAi in the muscle had no significant effect on the preference shift (Fig. 5c, Extended Data Fig. 7d, e). Moreover, neuronal sNPF-R knockdown promoted yeast consumption, suggesting that sNPF-R in the nervous system is responsible for suppressing protein feeding (Fig. 5d).
To gain insight into neural circuits downstream of enteric sNPF neurons, we employed the Trans-Tango technique. Remarkably, enteric sNPF neuron-driven Trans-Tango labelled neurons in the HCG by fluorescent reporter mtdTomato, suggesting that the neuronal pathway connects enteric sNPF neurons in the HCG with another set of HCG neurons (Fig. 5e). In contrast, while the corpora cardiaca, labelled with AKH, was located in the vicinity of enteric sNPF neurons, no Trans-Tango signals were observed (Fig. 5e). Since sNPF-RT2A−knock−in-GAL4 was active in brain neurons (Extended Data Fig. 7c), we also investigated neurons labelled by Trans-Tango in the brain. We found that neurons in the superior lateral protocerebrum (SLP) and suboesophageal ganglion (SEG) were labelled as the mtdTomato-positive downstream neurons (Extended Data Fig. 7f). However, similar signals were observed in no GAL4 controls (Extended Data Fig. 7g), suggesting that those signals in the SLP and SEG constitute non-specific activation of Trans-Tango by GAL4 driver. Therefore, we focused on HCG neurons as a target of enteric sNPF neurons.
Enteric sNPF neurons send signals to sugar-sensing Gr43a neurons
There are two types of neurons in HCG, piezo neurons and Gr43a neurons34,35. Piezo neurons are activated by the mechanical stimulus of feeding, while Gr43a neurons are activated by sugar feeding34,35. Immunostaining analysis revealed that a subpopulation of piezo neurons were enteric sNPF neurons, whereas Gr43a neurons were distinct from sNPF neurons (Fig. 6a). To ascertain which neuronal population of HCG neurons is stimulated by enteric sNPF neurons, we knocked down sNPF-R in those HCG neurons. Knockdown of sNPF-R in piezo neurons did not increase the protein and yeast intake in the feeding assay. (Fig. 6b, Extended Data Fig. 8a, b). In contrast, knockdown of sNPF-R in Gr43a neurons led to an enhanced protein preference and increased yeast autolysate consumption, similar to the knockdown of sNPF in the enteric neurons (Fig. 6b, c, Extended Data Fig. 8c). Moreover, GFP Reconstitution Across Synaptic Partners (GRASP) analysis revealed that enteric sNPF neurons and HCG Gr43a neurons probably contact each other through synaptic connections (Fig. 6d). Importantly, the reduced sensitivity to sugar due to knockdown of Gr43a in Gr43a neurons reduced yeast intake and increased sucrose intake (Fig. 6f). On the other hand, activation of Gr43a neurons by a temperature-dependent cation channel, TrpA1, increased yeast intake (Fig. 6g). These data suggest that enteric sNPF neurons regulate neuronal activity of sugar-sensing HCG Gr43a neurons to control the balance of sugar and yeast consumption.
The axis from gut-derived CCHa1 to HCG neurons influences the urea cycle and ammonium metabolism in HPD conditions
Given that loss of CCHa1 or sNPF resulted in dysregulation of protein feeding, we expected that loss-of-function animals would substantially influence HPD-associated metabolism. To explore the role of CCHa1 and sNPF in adapting HPD, we performed metabolomic analysis using Liquid Chromatography-tandem Mass spectrometry (LC-MS/MS) and compared metabolic profiles between SD and HPD conditions, as well as between control and EEC-specific CCHa1 RNAi animals. In both genotypes, HPD led to significant metabolic changes, including increased levels of some amino acids (Fig. 7a, Extended Data Fig. 9a). Strikingly, HPD increased levels of glycine, methionine, and arginine, which are important in removing nitrogen from amino acids via the urea cycle and creatine-creatinine metabolic pathway (Fig. 7b). Pathway analysis also revealed that arginine biosynthesis, pyrimidine metabolism, and aminoacyl-tRNA biosynthesis are significantly affected, suggesting that amino acid metabolism and degradation are influenced by HPD feeding (Extended Data Fig. 9b).
In comparing control and CCHa1 RNAi animals, we found a significant difference in levels of metabolites related to the urea cycle, a part of the arginine biosynthesis pathway, which detoxifies ammonium by incorporating it into ornithine and metabolizing it to citrulline36. Under HPD, CCHa1 RNAi animals showed an increase in citrulline and an increasing trend for argininosuccinate, the metabolite from citrulline (Fig. 7b, Extended Data Fig. 10a-g). Moreover, several metabolites of the creatine metabolism pathway, which functions in conjunction with the urea cycle37, also increased (Extended Data Fig. 10h-n).
Considering the accumulation of these intermediate products, we surmised that the urea cycle in gut-specific CCHa1 knockdown animals would not be able to adequately cope with amino acid metabolism. Therefore, we measured ammonium levels in whole-body samples to confirm whether the CCHa1 RNAi affects ammonium detoxication. As expected, HPD increased ammonium levels in both genotypes, but CCHa1 RNAi animals exhibited a more significant increase in ammonium levels under HPD feeding (Fig. 7c). Moreover, knocking down sNPF or CCHa1-R in enteric sNPF neurons also increased ammonium levels (Fig. 7d, Extended Data Fig. 11a). Taken together, these findings suggest that the gut-neuronal axis by gut-derived CCHa1 and enteric sNPF neurons is essential to balance protein intake and urea cycle capacity.
The axis from gut-derived CCHa1 to HCG neurons influences longevity in HPD conditions
Finally, we examined whether the axis from gut-derived CCHa1 to HCG neurons affects longevity in HPD conditions, as urea cycle disorders, leading to hyperammonemia, have a significant impact on survival in mammals38. Indeed, we found that gut-specific CCHa1 RNAi animals exhibited a slightly reduced lifespan under SD compared to controls (Fig. 7e). In contrast, under HPD conditions, their lifespan was more significantly shortened (Fig. 7e). Moreover, enteric sNPF neuron-specific sNPF knockdown also shortened lifespan on HPD (Fig. 7f). Taking the above results together, we revealed that the relay from gut-derived CCHa1 to HCG Gr43 neurons via enteric sNPF neurons suppresses excessive protein feeding, and as a result, excess ammonia production, which contributes to systemic adaptation to HPD.