Introduction

Animals are constantly sensing environmental stimuli and changing their behavior or physiology based on their internal state ^1–7. Hungry animals are strongly attracted to food. Immediately after seeing, smelling, or chewing food, even without absorbing nutrients, a burst of physiological changes is suddenly initiated in the body. These responses are known in mammals as the cephalic phase response (CPR) ⁸. For example, a flood of saliva and gastrointestinal secretions prepares hungry animals to digest food ^9–13. Starvation results in lower body temperatures, and chewing food triggers a rapid increase in heat production, demonstrating CPR in thermogenesis ^14–16. However, the underlying mechanisms of how the sensation of food without nutrients triggers the behavioral response remain unclear.

To address this question, we used a relatively simple and versatile model organism, Drosophila melanogaster. Flies exhibit robust temperature preference behavior ^17,18. Due to the low mass of small ectotherms, the source of temperature comes from the environment. Therefore, their body temperatures are close to the ambient temperature ^19,20. For temperature regulation, animals are not simply passive receivers of ambient temperature. Instead, they actively choose an ambient temperature based on their internal state. For example, we have shown that preferred temperature increases during the daytime and decreases during the night time, exhibiting circadian rhythms of temperature preference (temperature preference rhythms: TPR) ²¹. Because their surrounding temperature is very close to their body temperature, TPR leads to body temperature rhythms (BTR) that is very similar to mammalian BTR ^22,23. Another example is the starvation. Hunger stress forces flies to change their behavior and physiological response ^24,25. We previously showed that the hungry flies prefer a lower temperature ²⁶. The flies in a lower environmental temperature has been shown to have a lower metabolic rate, and the flies in a higher environmental temperature has been shown to have a higher metabolic rate ^26–28. Therefore, hungry flies choose a lower temperature and therefore, their metabolic rate is lower. Similarly, in mammals, starvation causes a lower body temperature, hypothermia ⁶. In mammals, body temperature is controlled by the balance between heat loss and heat production. The starved mammals have been shown a lower heat production ^5–7. Therefore, both flies and mammals, the starvation causes a low body temperature.

The flies exhibit robust feeding behaviors ^29,30 and molecular and neural mechanisms of taste are well documented ^2,31–36. Therefore, we focused on taste and temperature regulation and asked how the taste cue triggers a robust behavioral recovery of temperature preference in starving flies. We show in hungry flies that taste without nutrients induces a switch from a low to a high temperature preference. While taste leads to a warmer temperature preference, nutrient intake causes the flies to prefer an even warmer temperature. This nutrient-induced warm preference results in a complete recovery from starvation. Thus, taste-evoked warm preference is different from nutrient-induced warm preference and potentially similar physiology as CPR. Therefore, when animals emerge from starvation, they use a two-step approach to recovery: taste-evoked and nutrient-induced warm preference. While a rapid component is elicited by food taste alone, a slower component requires nutrient intake.

Animals are motivated to eat based on their internal state, such as time of day or degree of hunger. The circadian clock drives daily feeding rhythms ^37–39 and anticipates meal timing. Daily feeding timing influences energy homeostasis and metabolism ^40,41. Circadian clocks control feeding behavior in part via orexigenic peptidergic/hormonal regulation such as neuropeptide Y (NPY) and agouti-related peptide (AgRP) neurons, which are critical for regulating feeding and metabolism ^39,42. Feeding-fasting cycles modulate peripheral organs in liver, gut, pancreas, and so on ⁴³, suggesting that circadian clocks, peptidergic/hormonal signals, and peripheral organs are organically coordinated to enable animals maintain their internal states constantly. We found that clock genes and hunger signals are strongly required for taste-evoked warm preference. The data suggest that taste-evoked response is an indispensable physiological response that represents internal state. Taken together, our data shed new light on the role of the taste-evoked response and highlight a crucial aspect of our understanding of feeding state and energy homeostasis.

Results

Food detection triggers a warm preference

To investigate how food detection influences temperature preference behavior in Drosophila ^17,18, the white¹¹¹⁸ (w¹¹¹⁸) control flies were fed fly food containing carbohydrate, protein, and fat sources (Fig. 1A and S1) and tested in temperature preference behavioral assays (Fig. S1) ^26,44. The flies are released into a chamber set to a temperature range of 16-34°C and subsequently accumulate at their preferred temperature (Tp) at 25.2±0.2 °C within 30 minutes (min) (Fig. 1B: fed, white bar). On the other hand, when w¹¹¹⁸ flies were starved overnight with water only, they preferred 21.7±0.3 °C (Fig. 1A and 1B: overnight starvation (STV), gray bar). Thus, starvation leads to a lower Tp. As we have previously reported, starvation strongly influences temperature preference ²⁶.

Figure 1.

To examine how starved flies recover from lower Tp, they were offered fly food for 5 min, 10 min, 30 min, and 1 hr (Fig. 1A). Immediately after the flies were refed, the temperature preference behavior assay was examined. The assay takes 30 min from the time the flies are placed in the apparatus until the final choice is made. For example, in the case of 5 min refed flies, it took a total of 35 min from the start of refeeding to the end of the assay. After 10 min, 30 min, or 1 hr of fly food refeeding, starved flies preferred a temperature similar to that of fed flies (Fig. 1B, 1F, and S2: statistics shown as red stars, Table S1), suggesting a full recovery from starvation. On the other hand, refeeding after 5 min resulted in a warmer temperature than the starved flies. Nevertheless, Tp did not reach that of the fed flies (Fig. 1B, 1F, and S2). Therefore, refeeding the flies for 5 min resulted in a partial recovery from the starved state (Fig. 1B, 1F, and S2, statistics shown as green stars, Table S1). Thus, our data suggest that food intake triggers a warm preference in starved flies.

Sucralose refeeding promotes a warm preference

While only 5 min of refeeding fly food caused hungry flies to prefer a slightly warmer temperature, 10 min of refeeding caused hungry flies prefer a similarly warmer temperature as the fed flies (Fig. 1B, 1F, and S2). Therefore, we hypothesized that food-sensing cues might be important for the warm preference. Sucralose is an artificial sweetener that activates sweet taste neurons in Drosophila ⁴⁵ and modulates taste behaviors such as the proboscis extension reflex ^46–49. Importantly, sucralose is a non-metabolizable sugar and has no calories. Therefore, to investigate how food-sensing cues are involved in warm preference, we examined how sucralose refeeding changes the temperature preference of starved flies. After starved flies were refed sucralose for 10 min or 1 hr (Fig. 1A), they preferred a warmer temperature; however, Tp was halfway between Tps of fed and starved flies (Fig. 1C, F and S2: blue, Table S1). Thus, sucralose ingestion induces a warm preference but shows a partial recovery of Tp from the starved state.

While refeeding the flies with food resulted in a full recovery of Tp from the starved state, refeeding them with sucralose resulted in only a partial recovery of Tp. Therefore, starved flies may use both taste cues and nutrients to fully recover Tp from the starved state. To evaluate this possibility, we used glucose, which contains both sweetness (gustatory cues) and nutrients (i.e., metabolizable sugars) (Fig. 1A), and tested glucose refeeding for 5 min, 10 min, and 1 hr. We found that 10 min or 1 hr glucose refeeding resulted in full recovery of Tp from the starved state (Fig. 1D, 1F and S2: statistics shown in red NS, Table S1) and was significantly different from starved flies (Fig. 1D, 1F and S2: statistics shown in green stars, Table S1). Thus, our data showed that sucralose refeeding induced partial recovery and glucose refeeding induced full recovery from the starved state.

It is still possible that the starved flies consumed glucose faster than sucralose during the first 10 min, which could result in a different warming preference. To rule out this possibility, we examined how often starved flies touched glucose, sucralose, or water during the 30 min using the Fly Liquid-food Interaction Counter (FLIC) system ⁵⁰. The FLIC system assays allow us to monitor how much interaction between the fly and the liquid food reflects feeding episodes. We found that starved flies touched glucose and sucralose food at similar frequencies and more frequently than water during the 30 min test period (Fig. 1G, Table S1). The data suggest that flies are likely to feed on glucose and sucralose at similar rates. Therefore, we concluded that the differential effect of sucralose and glucose refeeding on temperature preference was not due to differences in feeding rate.

Furthermore, to confirm that sweet taste is more important than sugar structure, we instead used fructose, another simple sugar that contains sweetness and nutrients. Glucose and fructose are monosaccharides and a member of hexose and pentose, respectively. We tested fructose refeeding for 10 min and 1 hr. We found that 10 min of fructose refeeding resulted in full recovery of Tp from the starved state (Fig. 1E, 1F and S2: statistics shown in red NS, Table S1), and 10 min and 1 hr fructose refeeding were significantly different from starved flies (Fig. 1E, 1F and S2: statistics shown in green stars, Table S1). The data suggest that sweet taste is more important than the structures of the sugar compounds.

Activation of sweet taste neurons leads to warm preference

To determine how taste elicits a warm preference, we focused on the sweet gustatory receptors (Grs), which detect sweet taste. We used sweet Gr mutants and asked whether sweet Grs are involved in taste-evoked warm preference. Two different sweet Gr mutants, Gr5a^−/−; Gr64a^−/−and Gr5a^−/−;;Gr61a^−/−, Gr64a-f^−/−, are known to reduce sugar sensitivity compared to the control ^47,48,51,52. We found that sweet Gr mutant flies exhibited a normal starvation response in which the Tp of starved flies was lower than that of fed flies (Fig. 2A and 2B, white and gray bars, statistics shown as green stars, Table S1). However, starved sweet Gr mutant flies did not increase Tp after 10 min sucralose refeeding (Fig. 2A and 2B, blue bars, statistics shown as green and red stars, Table S1). These data suggest that sweet Grs are involved in taste-evoked warm preference.

Figure 2.

Sweet Grs are expressed in the sweet Gr-expressing neurons (GRNs) located in the proboscis and forelegs ^52,53. To determine whether sweet GRNs are involved in taste-evoked warm preference, we silenced all sweet GRNs. We expressed the inwardly rectifying K⁺ channel Kir2.1 (uas-Kir) ⁵⁴ using Gr64f-Gal4, which is expressed in all sweet GRNs in the proboscis and forelegs ^47,52,53. Inactivation of all sweet GRNs showed a normal starvation response (Fig. 2C, white and gray bars, statistics shown as green stars). However, flies silencing all sweet GRNs failed to show a warm preference after 10 min of sucralose refeeding (Fig. 2C, blue bar, statistics shown as green and red stars, Table S1). This phenotype was similar to the data obtained with the sweet Gr mutant strains (Fig. 2A and 2B). On the other hand, control flies (Gr64f-Gal4/+ and uas-Kir/+) showed a normal starvation response and a taste-evoked warm preference (Fig. 2D and 2E, gray and blue bars, respectively, statistics shown as green and red stars, Table S1). Thus, our data indicate that sweet GRNs are required for taste-evoked warm preference.

To further investigate whether activation of sweet GRNs induces a warm preference, we used the optogenetic approach, a red light sensitive channelrhodopsin, CsChrimson ^55,56. Starved flies were given water containing 0.8 mM all-trans-retinal (ATR), the chromophore required for CsChrimson activation. These flies were not fed sucralose; instead, gustatory neurons in starved flies were excited by red light pulses (flashing on and off at 10 Hz) for 10 min (Fig. 2F). In this case, although the flies were not refed, the gustatory neurons were artificially excited by CsChrimson activation so that we could evaluate the effect of excitation of sweet GRNs on taste-evoked warm preference.

CsChrimson was expressed in sweet GRNs in the proboscis and legs (all sweet GRNs) using Gr64f-Gal4. These flies showed a normal starvation response (Fig. 2G-2J, white and gray bars, statistics shown as green and red stars, Table S1). Excitation of all sweet GRNs by red light pulses elicited a warm preference, and Tp was intermediate between fed and starved flies, suggesting partial recovery (Fig. 2G, yellow bar, statistics shown as green and red stars, Fig. S2B Table S1). However, neither excitation of Gr5a- (Fig. 2H) nor Gr64a-expressing neurons (Fig. 2I) induced a warm preference (yellow bars, Table S1). While Gr64a-Gal4 is expressed only in the legs, Gr5a-Gal4 is expressed in the proboscis and legs, but does not cover all sweet GRNs like Gr64f-Gal4 ^52,53. Notably, control flies (UAS-CsChrimson/+) did not show a warm preference to red light pulses with ATR application (Fig. 2J). Taken together, our data suggest that excitation of all sweet GRNs results in a warm preference.

We next asked whether the sweet Grs contribute to the nutrient-induced warm preference. We found that all these starved flies did not increase Tp after 10 min of glucose intake (Fig. 2A-2C, green bars, statistics shown as green and red stars, Table S1). All control flies showed normal responses to 10 min of glucose refeeding (Fig. 2D and 2E, green bars, Table S1). The data suggest that the sweet Grs which we tested are potentially expressed in tissues/neurons required for internal nutrient sensing. Notably, we found that flies increased Tp after 10 min of refeeding with fly food containing carbohydrate, fat, and protein (Fig. 2A-2C, orange bars, statistics shown as green and red stars, Table S1). All control flies showed normal responses to 10 min of fly food intake (Fig. 2D and 2E, orange bars, Table S1). The data suggest that gustatory neurons are required for warm preference in carbohydrate refeeding, but not for other nutrients such as fat or protein (see Discussion). Because flies have sensory neurons that detect fatty acids ^57–59 or amino acids ^60–63, these neurons may drive the response to fly food intake. This is likely why the sweet-insensitive flies can still recover after eating fly food (Fig 2A-2C, orange bars).

The temperature-sensing neurons are involved in taste-evoked warm preference

The warm-sensing neurons, anterior cells (ACs), and the cold-sensing R11F02-Gal4-expressing neurons control temperature preference behavior ^26,44,64. Small ectotherms such as Drosophila set their Tp to avoid noxious temperatures using temperature information from cold- and warm-sensing neurons ^17,18,44. We have previously shown that starved flies choose a lower Tp, the so-called hunger-driven lower Tp ²⁶. ACs control the hunger-driven lower Tp, but cold-sensing R11F02-Gal4-expressing neurons do not ²⁶. ACs express transient receptor potential A1 (TrpA1), which responds to a warm temperature >25 °C ^44,65. The set point of ACs in fed flies, which is ∼25 °C, is lowered in starved flies. Therefore, the lower set point of ACs corresponds to the lower Tp in starved flies.

First, we asked whether ACs are involved in taste-evoked warm preference. Because the ACs are important for the hunger-driven lower Tp ²⁶, the AC-silenced flies did not show a significant difference in Tp between fed and starved conditions for only one overnight of starvation²⁶. Therefore, we first extended the starvation time to two overnights so that the AC-silenced flies showed a significant difference in Tp between fed and starved conditions (Fig. 3A, white and gray bars, statistics shown as green and red stars, Table S1). Importantly, longer periods of starvation do not affect the ability of w¹¹¹⁸ flies to recover (Fig. S3).

Figure 3.

To examine whether ACs regulate taste-evoked warm preference, we refed sucralose to AC-silenced flies for 10 min. We found that the Tp of the refed flies was still similar to that of the starved flies (Fig. 3A, blue bar, statistics shown as green and red stars, Table S1), indicating that sucralose refeeding could not restore Tp and that ACs are involved in taste-evoked warm preference. Significantly, even when the AC-silenced flies were starved for two overnights, they were able to recover Tp to the normal food for 10 min and to glucose for 1 hr (Fig. 3A, orange and green bars), suggesting that the starved AC-silenced flies were still capable of recovery.

Other temperature-sensing neurons involved in temperature preference behavior are cold-sensing R11F02-Gal4-expressing neurons ^26,64. To determine whether R11F02-Gal4-expressing neurons are involved in taste-evoked warm preference, we silenced R11F02-Gal4-expressing neurons using uas-Kir. We found that the flies showed a significant difference in Tp between fed and starved conditions, but the flies did not show a warm preference upon sucralose refeeding (Fig. 3B, gray and blue bars, statistics shown as green and red stars, Table S1). As controls, TrpA1^SH-Gal4/+, R11F02-Gal4/+ and uas-Kir/+ flies showed normal starvation response and taste-evoked warm preference (Fig. 3C-3E).

To further ensure the results, we used optogenetics to artificially excite warm and cold neurons with TrpA1^SH-Gal4 and R11F02-Gal4, respectively, by red light pulses for 10 min. We compared the Tp of starved flies with and without ATR under red light. Tp of starved flies with ATR was significantly increased (Fig. 3F and 3G, yellow bars, statistics shown as green and red stars, Table S1) compared to those without ATR (Fig. 3F and 3G, gray bars, Table S1). Therefore, these data indicate that ACs and R11F02-Gal4-expressing neurons are required for taste-evoked warm preference.

Next, we asked whether temperature-sensitive neurons contribute to the nutrient-induced warm preference. We used the warm- or cold neuron silenced flies (TrpA1^SH-Gal4 or R11F02-Gal4>uas-Kir) and found that all these starved flies did not increase Tp after 10 min glucose intake (Fig. 3A and 3B, green bars, Table S1), but increased Tp after 10 min refeeding with fly food containing carbohydrate, fat, and protein (Fig. 3A and 3B, orange bars, Table S1). Notably, AC-silenced flies increased Tp after 1 hr glucose intake (Fig. 3A, green bars, Table S1). All control flies showed normal responses to both 10 min glucose refeeding and fly food intake (Fig. 3C-3E, green bars, Table S1). The data suggest that temperature-sensing neurons are required for warm preference in carbohydrate refeeding, but not in other foods such as fat or protein (see Discussion).

Olfaction is possibly involved in a warm preference for hungry flies

We also investigated the potential effects of olfaction. We used mutants of the odorant receptor co-receptor, Orco (Orco¹), which has an olfactory defect ⁶⁶. We found that the flies showed a significant difference in Tp between fed and starved conditions, but the flies did not show a warm preference upon sucralose refeeding (Fig. S4A and S4C, gray and blue bars, statistics shown as green and red stars, Table S1). We found that all of these starved flies increased Tp after 10 min of glucose or fly food intake and showed a full recovery after 1 hr of glucose intake (Fig. S4B and S4C, orange and green bars, statistics shown as green and red stars, Table S1). The data suggest that olfaction may be involved in the warm preference in sucralose refeeding.

Internal state influences taste-evoked warm preference in hungry flies

Internal state strongly influences motivation to feed. However, how internal state influences starving animals to exhibit a food response remains unclear. Hunger represents the food-deficient state in the body, which induces the release of hunger signals such as neuropeptide Y (NPY). NPY promotes foraging and feeding behavior in mammals and flies ⁶⁷. While intracerebroventricular injection of NPY induces the cephalic phase response (CPR) ⁶⁸, injection of NPY antagonists suppresses CPR in dogs, suggesting that NPY is a regulator of CPR in mammals ⁶⁹. Therefore, we first focused on neuropeptide F (NPF) and small neuropeptide F (sNPF), which are the Drosophila homologue and orthologue of mammalian NPY, respectively ⁶⁷, and asked whether they are involved in taste-evoked warm preference. In NPF mutant (NPF^−/−) or sNPF hypomorph (sNPF hypo) mutant, we found that the Tp of fed and starved flies were significantly different, showing a normal starved response (Fig. 4A and 4B: white and gray bars, statistics shown as green and red stars, Table S1). However, they failed to show a taste-evoked warm preference after 10 min of sucralose refeeding (Fig. 4A and 4B: blue bars, statistics shown as green and red stars, Table S1). Thus, NPF and sNPF are required for taste-evoked warm preference after sucralose refeeding.

Figure 4.

We next asked whether NPF and sNPF are involved in the nutrient-induced warm preference during glucose refeeding. We found that starved NPF^−/− mutants significantly increased Tp after 1 hr of glucose refeeding (Fig. 4A, dark green bar, statistics shown as green and red stars, Table S1). We also found that starved sNPF hypo mutants significantly increased Tp after 10 min and 1 hr of glucose refeeding (Fig. 4B, green and dark green bars, statistics shown as green and red stars, Table S1). In addition, both NPF^−/− and sNPF hypo mutants increased Tp after 10 min of fly food refeeding (Fig. 4A and 4B, orange bars, statistics shown as green and red stars, Table S1). However, when these flies were fed both glucose and fly food, they did not reach the same Tp as the fed flies. These data suggest that NPF and sNPF also play a role in the modulation of nutrient-induced warm preference.

Factors involving hunger regulate taste-evoked warm preference

Based on the above results, we hypothesized that hunger signals might be involved in taste-evoked warm preference. To test this hypothesis, we focused on several factors involved in the hunger state. The major hunger signals ²⁴, diuretic hormone 44 (DH44), and adipokinetic hormone (AKH) are the mammalian corticotropin-releasing hormone homologue ⁷⁰ and the functional glucagon homologue, respectively ^71,72. We found that DH44- or AKH-expressing neuron silenced flies failed to show a taste-evoked warm preference after sucralose refeeding. However, they were able to increase Tp after 10 min glucose or fly food refeeding (Fig. 4D and 4F, green and orange bars, statistics shown as green and red stars, Table S1), suggesting a normal nutrient-induced warm preference. The control flies (Dh44-Gal4/+, Akh-Gal4/+, and UAS-Kir/+) showed a normal warm preference to sucralose (Fig. 4C, 4E and 4J, blue bars, statistics shown as green and red stars, Table S1), glucose (Fig. 4C, 4E and 4J, green bars, statistics shown as green and red stars, Table S1), and normal fly food refeeding (Fig. 4C, 4E and 4J, orange bars, statistics shown as green and red stars, Table S1). The data suggest that DH44 or AKH neurons are required for taste-evoked warm preference, but not for nutrient-induced warm preference.

Insulin-like peptide 6 (Ilp6) is a homologue of mammalian insulin-like growth factor 1 (IGF1), and ilp6 mRNA expression is increased in starved flies ^73–75. Because Ilp6 is important for the hunger-driven lower Tp ²⁶, the ilp6 mutants did not show a significant difference in Tp between fed and starved conditions for only one overnight starvation. Therefore, we first extended the starvation time to three overnights. We found that the ilp6 loss-of-function (ilp6 LOF) mutant failed to show a taste-evoked warm preference after sucralose refeeding (Fig. 4G, blue bars, statistics shown as green and red stars, Table S1), but did show a nutrient-induced warm preference after glucose refeeding (Fig. 4G, green bars, statistics shown as green and red stars, Table S1). These data suggest that Ilp6 is required for taste-evoked warm preference but not for nutrient-induced warm preference after glucose or fly food refeeding.

Unpaired3 (Upd3) is a Drosophila cytokine that is upregulated under nutritional stress ⁷⁶. We found that the upd3 mutants failed to show a taste-evoked warm preference after sucralose refeeding (Fig. 4H, blue bar, statistics shown as green and red stars, Table S1), but showed a nutrient-induced warm preference after glucose or normal food refeeding (Fig. 4H, green and orange bars, statistics shown as green and red stars, Table S1). These data suggest that Upd3 is required for taste-evoked warm preference, but not for nutrient-induced warm preference after glucose or normal food refeeding.

We also examined the role of the satiety factor Unpaired2 (Upd2), a functional leptin homologue in flies ⁷⁷, in taste-evoked warm preference. The upd2 mutants failed to show a warm preference after refeeding of sucralose, glucose or fly food (Fig. 4I, blue, green and orange bars, statistics shown as green and red stars, Table S1). Thus, our data suggest that factors involved in the hunger state are required for taste-evoked warm preference.

Flies show a taste-evoked warm preference at all times of the day

Animals anticipate feeding schedules at a time of day that is tightly controlled by the circadian clock ⁴¹. Flies show a rhythmic feeding pattern: one peak in the morning ⁵⁰ or two peaks in the morning and evening ⁷⁸. Because food cues induce a warm preference, we wondered whether feeding rhythm and taste-evoked warm preference are coordinated. If so, they should show a parallel phenotype.

Since flies exhibit one of the circadian outputs, the temperature preference rhythm (TPR) ²¹, Tp gradually increases during the day and peaks in the evening. First, we tested starvation responses at Zeitgeber time (ZT)1-3, 4-6, 7-9, and 10-12 under light and dark (LD) conditions, with flies being offered only water for 24 h prior to the experiments at each time point. We found that both w¹¹¹⁸ and yellow¹ white¹ (y¹w¹) flies had higher Tp in the fed state and lower Tp in the starved state at all times of the day (Fig. 5A and 5B: black and gray lines; Table S1). Next, we refed sucralose to starved flies at all time points tested and examined taste-evoked warm preference. While starved y¹w¹ flies showed a taste-evoked warm preference at all time points (Fig. 5B1, gray and blue lines, Table S1), starved w¹¹¹⁸ flies showed a significant taste-evoked warm preference at ZT4-6 and 10-12 (Fig. 5A1, gray and blue lines, Table S1).

Figure 5.

Because starved w¹¹¹⁸ flies showed an advanced phase shift of TPR with a peak at ZT7-9 (Fig. 5A, gray), it is likely that the highest Tp simply masks the taste-evoked warm preference at ZT7-9. We also focused on nutrient- (carbohydrate-) induced warm preference. Starved w¹¹¹⁸ and y¹w¹ flies successfully increased Tp after 10 min of glucose refeeding (Fig. 5A2 and B2, green line, Table S1). Glucose refeeding for 1 hr resulted in Tp similar to that of fed w¹¹¹⁸ flies (Fig. 5A3, dark green line). Because the feeding rhythm peaks in the morning or morning/evening ^50,78, our data suggest that the feeding rhythm and taste-evoked warm preference do not occur in parallel.

Circadian clock genes are required for taste-evoked warm preference, but not for nutrient-induced warm preference

We asked whether the circadian clock is involved in taste-evoked warm preference. We used clock gene null mutants, period⁰¹ (per⁰¹) and timeless⁰¹ (tim⁰¹). Although they showed significant starvation responses (Fig. 5C and 5D, black and gray lines, Table S1), neither starved per⁰¹ nor tim⁰¹ mutants could show taste-evoked warm preference upon sucralose refeeding (Fig. 5C1 and 5D1, blue lines, Table S1). Nevertheless, they fully recovered upon glucose refeeding in LD at any time of day (Fig. 5C2 and 5D2, green lines, Table S1). Therefore, our data suggest that clock genes are required for taste-evoked warm preference, but not for nutrient-induced warm preference.

However, starved per⁰¹ and tim⁰¹ mutants may eat sucralose less frequently than glucose, which could result in a failure to show a taste-evoked warm preference. Therefore, we examined how often starved per⁰¹ and tim⁰¹ mutants touched glucose, sucralose, or water during the 30 min using FLIC assays ⁵⁰ (Fig. S5A-S5C). Interestingly, starved per⁰¹ and tim⁰¹ mutants touched water significantly more often than glucose or sucralose (Fig. S5B and S5C, Table S1). Although starved tim⁰¹ flies touched glucose slightly more than sucralose for only 10 min, this phenotype is not consistent with per⁰¹ and w¹¹¹⁸ flies. However, these mutants still showed a similar Tp pattern for sucralose and glucose refeeding (Fig. 5C and 5D). The results suggest that although the tim⁰¹ flies can eat sufficient amount of sucralose over glucose, their food intake does not affect the Tp behavioral phenotype. Thus, we conclude that in per⁰¹ and w¹¹¹⁸ flies, the differential response between taste-evoked and nutrient-induced warm preferences is not due to feeding rate.

Discussion

When animals are hungry, sensory detection of food (sight, smell, or chewing) initiates digestion even before the food enters the stomach. The food-evoked responses are also observed in thermogenesis, heart rate, and respiratory rate in mammals ^14,15,79. These responses are referred to as the cephalic phase response (CPR) and contribute to the physiological regulation of digestion, nutrient homeostasis, and daily energy homeostasis ¹¹. While starved flies show a cold preference, we show here that the food cue, such as the excitation of gustatory neurons, triggers a warm preference, and the nutritional value triggers an even higher warm preference. Thus, when flies exit the starvation state, they use a two-step approach to recovery, taste-evoked and nutrient-induced warm preferences. The taste-evoked warm preference in Drosophila may be a physiological response potentially equivalent to CPR in mammals. Furthermore, we found that internal needs, controlled by hunger signals and circadian clock genes, influence taste-evoked warm preference. Thus, we propose that the taste-evoked response plays an important role in recovery and represents another layer of regulation of energy homeostasis.

Tp is determined by the taste cue

Starved flies increase Tp in response to a nutrient-free taste cue (Fig. 1C and 1F), resulting in a taste-evoked warm preference. We showed that silencing of ACs or cold neurons caused a loss of taste-evoked warm preference (Fig. 3A-3E), and that excitation of ACs or cold neurons induced a taste-evoked warm preference (Fig. 3F and 3G). The data suggest that both warm and cold neurons are important for taste-evoked warm preference: while ACs are required for the hunger-driven lower Tp ²⁶, both ACs and cold neurons are likely to be important for this taste-evoked warm preference.

The hunger-driven lower Tp is a slower response because starvation gradually lowers their Tp ²⁶. In contrast, the taste-evoked warm preference is a rapid response. Once ACs and cold neurons are directly or indirectly activated, starved flies quickly move to a warmer area. This is interesting because even when ACs and cold neurons are activated by warm and cold, respectively, the activation of these neurons causes a warm preference. Given that the sensory detection of food (sight, smell, or chewing food) triggers CPR, the activation of these sensory neurons may induce CPR. Tp in sucralose-refed hungry flies is between that of fed and starved flies (Figs. 2 and 3), making it difficult to detect the smaller temperature differences using the calcium imaging experiments. Therefore, we speculate that both ACs and cold neurons may facilitate rapid recovery from starvation so that flies can quickly return to their preferred temperature - body temperature - to a normal state.

Internal state influences taste-evoked warm preference

We show that mutants of genes involved in hunger and the circadian clock fail to show taste-evoked warm preference, suggesting that hunger and clock genes are important for taste-evoked warm preference. At a certain time of day, animals are hungry for food ^50,78. Thus, the hungry state acts as a gatekeeper, opening the gate of the circuits when hungry flies detect the food information that leads to taste-evoked warm preference (Figs. 4 and S6, blue arrow). While most of the hunger signals we focused on are important for taste-evoked warm preference, some hunger signals are also required for both taste-evoked and nutrient-induced warm preferences (Fig. 4). Notably, sensory signals contribute to both taste-evoked and nutrient-induced warm preferences (Figs. 2, 3 and S6, blue and green arrows). Thus, taste-evoked warm preference and nutrient-induced warm preference differ at the internal state level, but not at the sensory level. This idea is analogous to appetitive memory formation. Sweet taste and nutrients regulate the different layers of the memory formation process. Recent evidence suggests that the rewarding process can be subdivided; the sweet taste is for short-term memory and the nutrient is for long-term memory ^80–82. The data suggest that the taste-evoked response functions differently from the nutrient-induced response. Therefore, taste sensation is not just the precursor to nutrient sensing/absorption, but plays an essential role in the rapid initiation of a taste-evoked behavior that would help the animal survive.

How do hunger signals or clock genes contribute to taste-evoked warm preference?

The hunger signaling hormones/peptides studied in this project are important for taste modulation. For example, mammalian NPY and its Drosophila homolog NPF modulate the output of taste signals ^49,83,84. The AKH receptor is expressed in a subset of gustatory neurons that may modulate taste information for carbohydrate metabolism ⁸⁵. Therefore, the hunger signals are likely to modulate the downstream of the sensory neurons, which may result in a taste-evoked warm preference.

Circadian clock genes control and coordinate the expression of many clock-controlled genes in the body ^86–89. Therefore, we expect that the absence of clock genes will disrupt the molecular and neural networks of homeostasis, including metabolism, that are essential for animal life. For example, taste neurons express clock genes, and impaired clock function in taste neurons disrupts daily rhythms in feeding behavior ⁹⁰. Temperature-sensing neurons transmit hot or cold temperature information to central clock neurons ^91–95. Therefore, the disrupted central clock in clock mutants may respond imprecisely to temperature signals. There are many possible reasons why the lack of clock gene expression in the brain is likely to cause abnormal taste-evoked warm preference.

In addition, hunger signals may contribute to the regulation of circadian output. DH44 is located in the dorsomedial region of the fly brain, the pars intercerebralis (PI), and DH44-expressing neurons play a role in the output pathway of the central clock ^96,97. Insulin-producing cells (IPCs) are also located in addition to DH44-expressing neurons ^98–101. IPCs receive a variety of information, including circadian ^97,102 and metabolic signals ^73–75,77. and then transduce the signals downstream to release Ilps. Both Upd2 and Ilp6, which are expressed in the fat body respond to metabolic states and remotely regulate Ilp expression ^73–75,77. Insect fat body is analogous to the fat tissues and liver in the vertebrates ^103,104. Therefore, each hunger signal may have its specific function for taste-evoked warm preference. Further studies are needed to describe the entire process.

Taste-evoked warm preference may be CPR in flies

The introduction of food into the body disrupts the internal milieu, so CPR is a necessary process that helps animals prepare for digestion. Specifically, in mammals, taste leads to an immediate increase in body temperature and metabolic rate. Starvation results in lower body temperatures, and chewing food, even before it enters the stomach, triggers a rapid increase in heat production, demonstrating CPR in thermogenesis ^14,15.

Starved flies have a lower Tp ²⁶. Because Drosophila is a small ectotherm, the lower Tp indicates a lower body temperature ^19,20. Even when the flies do not receive food, the sweet taste and the excitation of sweet neurons induce starved flies to show a warm preference, which eventually leads to a warmer body temperature. In fact, CPR is known to be influenced by smell as well in mammals ⁹. We have shown in flies that olfactory mutants fail to show a warm preference when refed sucralose (Fig. S4). Starvation leads to lower body temperatures, and food cues, including taste and odor, rapidly induce a rise in body temperature before food enters the body. Thus, the taste-evoked warm preference in Drosophila may be a physiological response equivalent to one of the CPRs observed in mammals.

CPR is essential because both starved mammals and starved flies must rapidly regulate their body temperature to survive

As soon as starved flies taste food, the sensory signals trigger CPR. They can move to a warmer place to prepare to raise their body temperature (Fig. S6, blue arrows). CPR may allow flies to choose a more hospitable place to restore their physiological state and allow for a higher metabolism, and eventually move on to the next step, such as foraging and actively seeking a mate before competitors arrive. Thus, CPR may be a strategy for the fly’s survival. Similarly, starvation or malnutrition in mammals leads to lower body temperatures ^6,7,105, and biting food triggers heat production, which is CPR ^14–16. Thus, while starvation in both flies and mammals leads to lower body temperatures, food cues initiate CPR by increasing body temperature and nutrient intake, resulting in full recovery from starvation. Our data suggest that Drosophila CPR may be a physiological response equivalent to CPR observed in other animals. Thus, Drosophila may shed new light on the regulation of CPR and provide a deeper understanding of the relationship between CPR and metabolism.

Acknowledgements

We are grateful to Drs. Anupama Dahanukar, Hubert Amrein, Greg Suh, and Paul A. Garrity and the Bloomington Drosophila Fly Stock Center for the fly lines. We thank members of the Hamada laboratory for critical comments and advice on the manuscript, Dr. Richard A. Lang and members of his laboratory for comments and kindly sharing reagents, Dr. Satoshi Namekawa for kindly sharing reagents, and Matthew Batie and Nathan T. Petts for design and construction of the behavioral assays and red-light illumination apparatus. This research was supported by a RIP funding from Cincinnati Children’s Hospital, JST (Japan Science and Technology)/Precursory Research for Embryonic Science and Technology (PRESTO), the March of Dimes, and NIH R01 grant GM107582, NIH R21 grant NS112890, NIH R35 GM152154 grant, and NIH R34 grant NS132843 to F.N.H. S.H.S. is a Latin American Fellow in the Biomedical Sciences supported by the Pew Charitable Trusts and by NIH K99 grant NS133470. Research in the laboratory of J.C.C. is supported by NIH R01 grant DK124068.

Contributions

F.N.H and Y.U designed the experiments. Y.U, E.N., T.N. and J.S. performed the temperature preference behavioral assays and data analysis. Y.U., S.H.S., S.S.U. and J.C. performed the feeding experiments and the data analysis. F.N.H and Y.U wrote the manuscript.

Declaration of Interest

The authors declare no competing interests


Materials and Methods

All flies were reared under 12 hr-light/12 hr-dark (LD) cycles at 25°C and 60-70% humidity in an incubator (DRoS33SD, Powers Scientific Inc.) with an electric timer (light on: 8am; light off: 8pm). The light intensity was 1000-1400 lux. All flies were reared on custom fly food recipe, with the following composition per 1 L of food: 6.0 g sucrose, 7.3 g agar, 44.6 g cornmeal, 22.3 g yeast and 16.3 mL molasses, as previously described ²⁶. white¹¹¹⁸ (w¹¹¹⁸) and yellow¹ white¹ (y¹w¹) flies were used as control flies. EP5Δ; Gr64a¹ (Gr5a^−/−; Gr64a^−/−) was kindly provided by Dr. Anupama Dahanukar ⁴⁷. R1; Gr5a-LexA; +; ΔGr61a, ΔGr64a-f (Gr5a^−/−; Gr61a^−/−, Gr64a-f^−/−) were kindly provided by Dr. Hubert Amrein ^51,52. Dh44-Gal4 was kindly provided by Dr. Greg Suh ⁴⁸. TrpA1^SH-gal4 was kindly provided by Dr. Paul A. Garrity ⁴⁴. Other fly lines were provided by Bloomington Drosophila stock center and Vienna Drosophila Stock Center.

Temperature preference behavioral assay

The temperature preference behavior assays were examined using a temperatures gradient, set from 16-34°C and were performed for 30 min in an environmental room maintained at 25°C /60%-70% humidity, as previously described ²⁶. Because starved flies showed a lower preferred temperature (Tp), the temperature regulation was lower than usual (Umezaki et al. 2018 Current Biology). We prepared a total of 40-50 flies (male and female flies were mixed) for fed condition experiments and 90-100 flies for overnight(s) starved and refed condition experiments for one trial. Flies were never reused in subsequent trials. In the starved and refed experiments, we prepared twice the number of flies needed for a trial because almost half of them died from starvation stress. Others climbed on the wall and ceiling (starved flies are usually hyperactive ¹⁰⁶, even though we applied slippery coating chemicals (PTFE; Cat# 665800, Sigma or byFormica PTFE Plus, https://byformica.com/products/fluon-plus-ptfe-escape-prevention-coating) to the Plexiglass covers.

Behavioral assays were performed for 30 min at Zeitgeber time (ZT) 4-7 (light on and light off are defined as ZT0 and ZT12, respectively), and starvation was initiated at ZT9-10 (starved for 1, 2 or 3 O/N are 18-21 hr, 42-45 hr or 66-69 hr, respectively). As for STV1.5 for Fig. S3, starvation was initiated at ZT1-2 (26-29 hr). For the starvation assays, the collected flies were maintained on our fly food for at least one day and then transferred to plastic vials containing 3 mL of distilled water, which was absorbed by a Kimwipe paper. For refeeding experiments, starved flies were transferred to plastic vials containing 2 mL of 2.8 mM sugar solution (sucralose water/glucose water). Sugar solution is absorbed by half the size of a Kimwipe. The details of the starvation period are described in the following section (see “Starvation Condition”).

After the 30-min behavioral assay, the number of flies whose bodies were completely located on the aluminum plate was counted. Flies whose bodies were partially or completely located on the walls of the Plexiglass cover were not included in the data analysis. The percentage of flies within each one-degree temperature interval on the apparatus was calculated by dividing the number of flies within each one-degree interval by the total number of flies on the apparatus. The location of each one-degree interval was determined by measuring the temperature at 6 different points on the bottom of the apparatus. Data points were plotted as the percentage of flies within a one-degree temperature interval. The weighted mean of Tp was calculated by summing the product of the percentage of flies within a one-degree temperature interval and the corresponding temperature (e.g., fractional number of flies x 17.5°C + fractional number of flies x 18.5°C +……… fractional number of flies x 32.5°C). If the SEM of the averaged Tp was not < 0.3 after the five trials, additional trials were performed approximately 10 times until the SEM was < 0.3.

Microsoft Excel (Home tab>Conditional formatting tool>3-color scales and data bars) was used to create heat maps to show the distribution of flies in each experimental condition. The averaged percentages of flies that settled on the apparatus within each one-degree temperature interval were used to create the heat maps. Each scale value is as follows; minimum value: 0, midpoint value: 15%, and maximum value: 60% for w¹¹¹⁸. Minimum value: 0, middle value: 10%, and maximum value: 45% for Gr64fGal4>CsChrimson.

Starvation conditions and recovery

Most of the flies were starved for 2 overnights (O/N). Because some flies (e.g. ilp6 mutants) show starvation resistance and seem to be still healthy even after 2 O/N of starvation. We had to starve them for 3 O/N to show a significant difference in Tp between fed and starved flies. On the other hand, some flies (e.g., w¹¹¹⁸ flies) are very sick after 3 O/N of starvation, in which case we only had to starve them for 1 day. Therefore, the starvation conditions we used for this manuscript are from 1 to 3 O/N.

First, we confirmed the starvation period by focusing on Tp which resulted in a statistically significant Tp difference between fed and starved flies; as mentioned above, flies prefer lower temperatures when starvation is prolonged ²⁶. Therefore, when Tp was not statistically different between fed and starved flies, we extended the starvation period from 1 to 3 O/N. Importantly, we show in Fig. S3 that the duration of starvation does not affect the recovery effect. Furthermore, w¹¹¹⁸ flies cannot survive 42-49 or 66-69 hours of starvation.

Temperature preference rhythm (TPR) assay

For the TPR assays, we performed temperature preference behavior assays in different time windows during the daytime (zeitgeber time (ZT) or circadian time (CT) 1-3, 4-6, 7-9, and 10-12) as described previously ^21,107. Because starvation duration directly affects flies’ Tp ²⁶ starvation was initiated at each time window to adjust the starvation duration at each time point, which means flies were starved for 24 hrs or 48 hrs but not 1 or 2 O/N. Each behavioral assay was not examined during these time periods (ZT or CT0-1 and 11.5-12) because of large phenotype variation around light on and light off.

Furthermore, insulin levels were shown to peak at 10 min and gradually decline ¹⁰⁸. Also, how quickly the flies can consume food is unclear. These factors may influence temperature preference behavior. Therefore, to minimize these effects, we decided to test the temperature preference behavioral assay immediately after the flies had eaten the food.

Optogenetic activation

For the optogenetic activation of the target neurons for behavioral assays, the red-light-sensitive channelrodopsin, UAS-CsChrimson, was crossed with each Gal4 driver. Flies were reared on fly food at 25°C and 60-70% humidity under LD cycles in an incubator (DRoS33SD, Powers Scientific Inc.) with an electric timer. After the flies emerged, adult flies were collected and maintained on fly food for 1-2 days. The next day, flies were transferred to water with or without 0.8 mM all-trans retinal (ATR; #R2500, Sigma) diluted in dimethyl sulfoxide (DMSO; #472301, Sigma) for 2 O/N. To activate flies expressing UAS-CsChrimson crossed with Gal4 drivers, we used a 627-nm red light-emitting diode (LED) equipped with a pulsed photoillumination system (10 Hz, 0.08 mWmm-2). Flies were exposed to pulsed red light for 10 min, which corresponds to the refeeding period. This photoillumination system was used in an incubator (Sanyo Scientific, MIR-154) and followed by temperature preference behavioral assays.

The R11F02-Gal4>uas-CsChrimson flies do not develop into adults and die in the pupal stage. Therefore, the Gal4/Gal80ts system was used to restrict uas-CsChrimson expression. The Gal80^ts is a temperature sensitive allele of Gal80 that causes Gal4 inhibition at 18°C and activation at 29°C ¹⁰⁹.The tubGal80^ts; R11F02-Gal4>uas-CsChrimson flies were reared on fly food at 18°C. Emerging adult flies were collected and kept on fly food at 29°C. The next day, flies were transferred to water (starved condition) with or without 0.8 mM ATR for 2 O/N. Starved flies with or without ATR application were exposed to pulsed red light for 10 min (equivalent to the refeeding period) and then immediately loaded into the behavioral apparatus for behavioral assays to measure their Tp. All flies were treated with ATR after they had fully developed into the adults. This means that Gal4-expressing cells were activated by red light via CsChrimson only at adult stages.

Feeding assay

To measure individual fly feeding, we used the Fly Liquid Interaction Counter (FLIC) system ⁵⁰. Groups of 1-2 day old male and female flies were starved for 24 hours starting between ZT4-5 (12-1pm). Individual flies were then loaded into FLIC monitors. Flies were acclimated to the monitors for 30 min with access to water in the feeding wells. At the start of the feeding study, water was replaced with 2.8 mM sucralose or 2.8 mM glucose solution (equivalent to 5% glucose concentration) and the number of licks (touches) was recorded for 30 min. Water, sucralose, or glucose water was administered individually in separate experiments. Assays were performed on ZT4-7. To account for the potential confounding effect of startle response when food is changed, wells where water was replaced with new water were used as a control. FLIC raw data were analyzed using the FLIC R code master (Pletcher Lab, available at https://github.com/PletcherLab/FLIC_R_Code). Lick counts were obtained and summed in 5-min window bins, while cumulative licks were obtained by successively summing the licks in these bins.