Dopaminergic neurons in the paraventricular hypothalamus extend the food consumption phase

Significance

Feeding behaviors of wild animals are influenced by the repetitive cycle of feeding phases: Food procurement, consumption, and termination. However, neural circuits controlling the food consumption phase remain unclear. Here, we found that the dopaminergic neurons in the paraventricular nucleus of the hypothalamus (PVH) increased the duration of the food consumption phase. We found that these neurons also increased meal intake by activating the D2 receptor in the lateral habenula (LHb). Furthermore, levels of the rate-limiting enzyme for dopamine synthesis in these neurons were upregulated during obesity and contributed to obesity development driven by epigenetic mechanisms. These findings suggest that the PVH dopaminergic neurons play key roles in enhancing food consumption and obesity induction.

Abstract

Feeding behavior is controlled by various neural networks in the brain that are involved in different feeding phases: Food procurement, consumption, and termination. However, the specific neural circuits controlling the food consumption phase remain poorly understood. Here, we investigated the roles of dopaminergic neurons in the paraventricular nucleus of the hypothalamus (PVH) in the feeding behavior in mice. Our results indicated that the PVH dopaminergic neurons were critical for extending the food consumption phase and involved in the development of obesity through epigenetic mechanisms. These neurons synchronized with proopiomelanocortin neurons during consumption, were stimulated by proopiomelanocortin activation, and projected to the lateral habenula (LHb), where dopamine receptor D2 was involved in the increase in food consumption. In addition, upregulated tyrosine hydroxylase (TH) expression in PVH was associated with obesity and indispensable for obesity induction in mice lacking Dnmt3a. Taken together, our results highlight the roles of PVH dopaminergic neurons in promoting food consumption and obesity induction.

Obesity prevalence is continuously increasing worldwide (1). Excessive energy intake relative to energy expenditure is a direct cause of obesity (2). Therefore, elucidation of the neural mechanisms underlying feeding control and development of hyperphagia is important to control the obesity epidemic. Feeding behavior is controlled by neural networks involving several brain areas, among which the hypothalamic feeding center plays a central role. First-order feeding neurons in the arcuate nucleus (ARC), neuropeptide Y (NPY)/agouti-related peptide (AgRP), and proopiomelanocortin (POMC) neurons send abundant projections to the second-order feeding center neurons in the paraventricular nucleus of the hypothalamus (PVH) (3–5), which comprises heterogeneous groups of neurons, including anorexigenic and orexigenic neurons (5). Most PVH neurons express Sim1 (5), and Sim1-expressing PVH neurons play key roles in feeding regulation and energy balance independently or as part of the melanocortin pathway (4, 6). Feeding behavior is impacted by at least three phases, food procurement (approaching food), food consumption (consuming food), and meal termination (leaving food) (7, 8), and each phase is controlled by different neuronal networks (7). Food procurement phase is mostly controlled by neural circuits starting from the NPY/AgRP neurons (7, 9). Food consumption phase is related to the γ-aminobutyric acid (GABA) neurons in the lateral hypothalamus (LH), which also enhance nonfood consumption (7). The parabrachial nucleus plays a key role in the meal termination phase (7). However, the specific neural circuits involved in the food consumption phase remain unclear.

Recent increase in obesity rate is thought to be influenced by environmental factors, such as food availability, nutrient balance, and physical activity; epigenetic modifications are part of the response mechanisms of the body to environmental changes. Recent studies have shown that epigenetic mechanisms, including DNA methylation in the feeding center, contribute to energy homeostasis (10–13). Gene expression analysis using Sim1-Cre-specific DNA methyltransferase 3 alpha (Dnmt3a) deletion mice has suggested that tyrosine hydroxylase (TH), a rate-limiting enzyme for catecholamine synthesis, is a key downstream gene highly sensitive to aberrant DNA methylation in Sim1-expressing PVH neurons (10). Therefore, we hypothesized that Th-expressing PVH neurons are responsible for the epigenetic induction of obesity. However, the roles of Sim1-Cre-expressing PVH TH neurons remain poorly understood. In this study, we showed that the dopaminergic neurons in PVH play key roles in promoting the feeding behavior during the food consumption phase as well as the epigenetic induction of obesity.

Results

PVH TH Neurons are Unique Dopaminergic Neurons.

First, we histologically characterized the TH neurons overlapping the Sim1-Cre neurons. Throughout the brain, Sim1-Cre-expressing neurons overlapped with TH neurons only in central PVH (Fig. 1A). The number of TH-immunoreactive neurons in central PVH was significantly reduced in the Sim1-Cre-specific Th deletion mice compared to that in the Sim1-Cre/tdTomato mice (18.75 ± 1.73 vs. 3.1 ± 0.73 neurons per section; 83% reduction; Fig. 1B). In other brain areas, such as posterior PVH and zona incerta of the subthalamus (A13 dopaminergic neurons; Fig. 1C), periventricular hypothalamus (A14 dopaminergic neurons), ARC, ventral tegmental area (A10), and substantia nigra (A8; SI Appendix, Fig. S1), no colocalization of Sim1-Cre and TH was observed. Then, we analyzed the TH neurons in central PVH. Almost all [95.96 ± 3.09 (mean ± SEM)%; n = 4 mice] Th-Cre neurons in PVH labeled with mCherry possessed the dopamine marker, dopamine transporter (Fig. 1D), confirming that most Th-Cre-expressing neurons in PVH are dopaminergic neurons. Double immunohistochemistry for TH and glutamic acid decarboxylase 67 revealed that a small portion of TH neurons in PVH (24.92 ± 1.11%, n = 3) were GABAergic neurons (Fig. 1E). The major feeding neurons in PVH, corticotropin-releasing hormone (CRH), thyrotropin-releasing hormone (TRH), oxytocin, nucleobindin 2 (NUCB2), and prodynorphin neurons, were rarely colocalized with TH neurons in PVH (CRH: 8.65 ± 1.84%, n = 3; TRH: 11.86 ± 4.78%, n = 3; oxytocin: 2.44 ± 1.23%, n = 3; NUCB2: 4.11 ± 0.95%, n = 3; prodynorphin: 1.42 ± 0.73%, n = 3; Fig. 1 F–J). Melanocortin 4 receptor was expressed on most TH neurons in PVH (93.99 ± 3.25%, n = 3; Fig. 1K). Therefore, Sim1-Cre-expressing TH neurons are dopaminergic neurons distinct from most other feeding-related neurotransmitter-expressing neurons in PVH and possibly a part of the melanocortin pathway. Dopaminergic neurons in central PVH have not been clearly defined as a specific dopaminergic neuron group in most studies but rather included as a part of the vaguely defined A14 neurons in addition to PeV dopaminergic neurons (14, 15). Moreover, dopaminergic neurons in central PVH have not been well studied, and their roles in feeding and metabolism remain unknown.

Fig. 1.

Paraventricular nucleus of the hypothalamus (PVH) tyrosine hydroxylase (TH) neurons are unique dopaminergic neurons. (A) Sim1-Cre-expressing TH neurons analyzed via TH-immunohistochemistry. TH-immunoreactive neurons (green) colocalized with tdTomato-expressing Sim1-Cre neurons (Left) in the central PVH of Sim1-Cre/tdTomato mice. This TH-immunoreactivity was abolished in the tdTomato-expressing neurons (red) of Th^lox/lox/Sim1-Cre/tdTomato mice (Right). (B) Numbers of TH-immunoreactive neurons per 25-µm thick slice on one side of central PVH in control Sim1-Cre/tdTomato (control; n = 4) and Th^lox/lox/Sim1-Cre (n = 3) mice. Error bar, SEM. ****P < 0.0001. (C) No Sim1-Cre-induced tdTomato expression was observed in the TH neurons (A13 dopamine group) in the posterior part of PVH (PaPo) and zona incerta (ZI) (C) in Sim1-Cre/tdTomato mice. (D) Th-Cre-expressing neurons visualized via injection of AAV-hSyn-DIO-mCherry into PVH (red) and labeling for the dopamine transporter (DAT; green). Double immunohistochemistry for TH with γ-aminobutyric acid (GABA) marker glutamic acid decarboxylase 67 (GAD67) (E), corticotropin-releasing hormone (CRH) (F), thyrotropin-releasing hormone (TRH) (G), oxytocin (H), nucleobindin 2 (NUCB2) (I), prodynorphin (J), and MC4R (K). (Scale bar, 30 μm.) Arrowheads indicate the cells exhibiting both fluorescence signals.

PVH TH Neurons Play an Indispensable Role in the Epigenetic Induction of Obesity.

We further analyzed the relationship between PVH dopaminergic neurons and obesity. As previously reported (10), lack of Dnmt3a in Sim1-Cre neurons induced obesity (Fig. 2A). However, double Dnmt3a and Th deletion in Sim1-Cre neurons abolished the increase in body weight by normalizing the food intake and energy expenditure (Fig. 2A and SI Appendix, Fig. S2), suggesting that PVH dopaminergic neurons are indispensable for the epigenetic induction of obesity. Next, we examined whether the absence of Th in PVH affects the body weight. Under a normal chow diet, Sim1-Cre-specific Th deletion did not affect the body weight, daily food intake, and energy expenditure (Fig. 2B and SI Appendix, Fig. S3), indicating that PVH dopaminergic neurons are dispensable for energy homeostasis under normal dietary conditions. In contrast, when fed a high-fat diet (HFD), Sim1-Cre-specific Th deletion mice exhibited lower body weights compared to the control mice (Fig. 2B), suggesting that Th in PVH contributes to the development of obesity. Because Th expression is highly upregulated in the PVH of Sim1-Cre-specific Dnmt3a deletion mice (10), we hypothesized that this upregulation contributes to obesity development. Consistently, Th expression levels were increased in PVH (Fig. 2 C and D), but not in ARC, zona incerta, and ventral tegmental area (SI Appendix, Fig. S4 A–E), in HFD-induced obese mice. Similarly, monogenetic db/db obese mice exhibited higher Th expression levels, specifically in PVH, compared to the control mice (Fig. 2 E and F and SI Appendix, Fig. S4 F–J). These results suggest that dopamine synthesis in PVH is upregulated during obesity. Interestingly, short-term exposure to HFD (one week; SI Appendix, Fig. S4 K and L), which did not cause significant changes in body weight, did not alter the Th expression levels in PVH. This suggests that Th upregulation is associated with the progression of obesity rather than its onset. Collectively, these data indicate that the PVH TH neurons contribute to the development of obesity by increasing Th expression, possibly mediated by epigenetic mechanisms.

Fig. 2.

PVH TH neurons play an indispensable role in the epigenetic induction of obesity. (A) Body weight of Dnmt3a^lox/lox (n = 11 to 15), Dnmt3a^lox/lox/Sim1-Cre (n = 11 to 15), Dnmt3a^lox/lox/Th^lox/lox (n = 11 to 20), and Dnmt3a^lox/lox/Th^lox/lox/Sim1-Cre (n = 11 to 20) mice. Two-way ANOVA; main effect of group; P < 0.0001. Tukey’s multiple-comparison test; *P < 0.01, ^#P < 0.005, and ^##P < 0.001, Dnmt3a^lox/lox/Sim1-Cre vs. Dnmt3a^lox/lox mice. (B) Body weight of Th^lox/lox and Th^lox/lox/Sim1-Cre mice fed a normal chow diet (n = 7 to 12/genotype) or high-fat diet (HFD; n = 6 to 7/genotype). Two-way ANOVA; main effect of group; P < 0.0001. Tukey’s multiple-comparison test; *P < 0.01, HFD-fed Th^lox/lox vs. HFD-fed Th^lox/lox/Sim1-Cre mice. (C) Th in situ hybridization in the PVH of mice fed chow or HFD. (D) Numbers of Th mRNA-expressing neurons in the PVH of mice fed chow (black; n = 20) and HFD (red; n = 20). (E) Th in situ hybridization in the PVH of male db/+or db/db mice at 20 wk of age. (F) Numbers of Th mRNA-expressing neurons in the PVH of db/+ (n = 5; black) and db/db (n = 7; red) mice. (Scale bar, 50 µm.) *P < 0.05 and **P < 0.01.

PVH Dopaminergic Neurons Increase the Food Intake During the Intermediate Eating Stage.

To analyze the roles of PVH dopaminergic neurons in feeding behavior, we evaluated refeeding after 24 h of food deprivation using I) a feeding behavior analysis system, which measures powdered food consumption from a feeder after several days of acclimation (Fig. 3A), and II) manual weighing of food pellets in the feeder above a single-housed mouse cage (Fig. 3B). Upon the measurement of powdered food consumption, a clear two-phase pattern of food intake was observed in the control mice, with peaks at the beginning and intermediate (60 to 90 min) stages of refeeding (Fig. 3A). A similar two-phase food intake pattern, with a second peak occurring at 60 min, was also observed in the control mice consuming the food pellet; however, the second peak was not as pronounced as that observed with powdered food (Fig. 3B). Mice with Sim1-Cre-specific Th deletion ate less food than the controls approximately 1 h after refeeding onset, as observed with both the food intake measurement systems (Fig. 3 A and B), suggesting that PVH dopamine synthesis is necessary to induce the intermediate eating stage. The percentage of PVH TH-immunoreactive neurons expressing c-Fos increased after 1 h of refeeding, confirming that PVH dopaminergic neurons were activated during the intermediate eating stage (Fig. 3 C and D).

Fig. 3.

PVH dopaminergic neurons increase the food intake during the intermediate eating stage. (A) Intake of powdered food after 24 h of food deprivation in Th^lox/lox (n = 6) and Th^lox/lox/Sim1-Cre (n = 6) mice measured using the feeding, drinking, and activity monitoring system. (B) Intake of food pellets by Th^lox/lox (n = 15) and Th^lox/lox/Sim1-Cre (n = 10) mice from the feeder of a single-housed mouse cage. (C) TH and c-Fos double immunohistochemistry in the PVH of ad libitum fed, 25-h fasted, and 1-h refed mice. Cells expressing both TH and c-Fos are indicated by arrowheads. (D) Percentage of c-Fos-immunopositive neurons among the TH neurons in PVH [n = 6/group; one-way ANOVA (P = 0.003), Tukey’s honest significant difference (HSD) test]. *P < 0.05, **P < 0.01, and ***P < 0.005. Error bars: SEM.

PVH Dopaminergic Neurons are Activated During the Food Consumption Phase.

In nature, feeding typically occurs outside nests in an unsafe open environment under threat (16, 17). To investigate the roles of PVH dopaminergic neurons in more complex feeding behaviors, we placed 24-h food-deprived mice in a rectangular cage with food pellet at the center and recorded their feeding behavior for 15 min (Fig. 4A). Over the entire experimental period, food intake decreased in the Sim1-Cre-specific Th deletion mice (Fig. 4B). To analyze the detailed feeding behaviors, consisting of the food procurement, food consumption, and meal termination phases (7, 8), we measured the duration of eating or contacting food and walking with or without exploration. The total durations of food consumption (Fig. 4C) and food contact (Fig. 4D) were significantly decreased in the Sim1-Cre-specific Th deletion mice. The total time devoted to exploring the chamber was similar (Fig. 4E), but the total duration of walking without exploring was longer in the Sim1-Cre-specific Th deletion mice than in the controls (Fig. 4F). Of note, the time spent in the center zone in the open-field test without any food pellet did not change in these mice (SI Appendix, Fig. S5). These data indicate that PVH dopaminergic neurons play key roles in extending the duration of food contact and consumption and reducing the nonexploratory walking distance.

Fig. 4.

PVH dopaminergic neurons are activated during the food consumption phase. (A) 24-h food-deprived mice were placed in a rectangular open-field with a food pellet on a holder at the center and a water bottle in the corner. Mouse behavior was recorded using a video camera for 15 min. (B) Food intake of 24-h food-deprived Th^lox/lox and Th^lox/lox/Sim1-Cre mice was analyzed during the first 15 min. Duration of food intake (C), food contact (D), exploring (E), and walking (F) by Th^lox/lox (n = 8) and Th^lox/lox/Sim1-Cre (n = 10) mice. Error bar: SEM. (G–L) AAV-syn-FLEX-jGCaMP8s was injected into the PVH of Th-Cre mice (G and J) and arcuate nucleus (ARC) of Agrp-Ires-Cre (H and K) and Pomc-Cre (I and L) mice. After 24-h food-deprivation, the mice were placed in an open-field for 15 min, and GCaMP intensities during feeding were determined in PVH and ARC. (G–I) GCaMP intensities throughout the feeding cycle, from the food procurement and consumption phases to the meal termination phase, are shown. Z-scores were calculated for the entire feeding cycle. Trace lines with error bars (Left) indicate the average GCaMP intensity (±SEM). The onset of each phase is marked as 0 s. Mean z-score of the first 5 s of each phase was compared (Right; n = 30 trials with 10 mice; one-way ANOVA [P < 0.001 (G), P < 0.0001 (H), and P < 0.0001 (I)], Tukey’s multiple comparisons test). (J–L) Change in GCaMP intensity before and after the onset (black line) of each phase is shown (n = 30 trials with 6 to 10 mice/trace line). Z-scores were calculated for the period from 5 s before to 10 s after the onset of each phase. Mean z-scores before (5 s) and after (10 s) the onset of each phase were compared (Right; n = 30). *P < 0.05, **P < 0.01, ***P < 0.005, and ****P < 0.001.

To analyze the changes in neuronal activity during feeding, we simultaneously performed GCaMP fiber photometry and video recording of the feeding behavior. GCaMP intensity of PVH dopaminergic neurons was highest in the food consumption phase (Fig. 4G and Movie S1). The activity increased immediately at the onset of each feeding phase, followed by a gradual decrease, with the average (Fig. 4J and SI Appendix, Fig. S6 and Movie S1). GCaMP fiber photometry of AgRP and POMC neurons in ARC was performed to compare their activities with that of PVH dopaminergic neurons. Although AgRP neuronal activity was inhibited in the presence of food in a previous study (9), it was repeatedly up- and down-regulated in our open-field feeding chamber according to the feeding phase as follows: Immediate inhibition at the onset of food procurement and food consumption and activation at meal termination (Fig. 4 H and K and Movie S2). Activity of POMC neurons increased concurrently with food consumption and decreased immediately after food termination (Fig. 4 I and L and Movie S3). Therefore, during the food consumption phase, activity pattern of PVH dopaminergic neurons was similar to that of POMC neurons but opposite to that of AgRP neurons.

PVH Dopaminergic Neurons Extend the Food Consumption Phase.

To clarify the roles of PVH dopaminergic neurons, we activated and inhibited PVH dopaminergic neurons using DREADDs and analyzed the feeding behavior via video recording and PVH dopaminergic neuronal activity through GCaMP fiber photometry. Activation of PVH dopaminergic neurons extended the food consumption phase (Fig. 5 A–C and SI Appendix, Fig. S7), whereas inhibition of PVH dopaminergic neurons had the opposite effect (Fig. 5 D–F and SI Appendix, Fig. S7). These data indicate that the PVH dopaminergic neurons are involved in extending the duration of food consumption. Food intake increased upon the DREADD-induced activation of PVH dopaminergic neurons and decreased upon the suppression of PVH dopaminergic neurons (Fig. 5 G–I), confirming that PVH dopaminergic neurons are orexigenic neurons that increase food consumption. Of note, robust drinking behavior was not induced by the chemogenic activation of PVH dopaminergic neurons, indicating that the activation of PVH TH neurons does not induce nonspecific consumption behaviors (18).

Fig. 5.

PVH dopaminergic neurons extend the food consumption phase. (A) AAV-syn-FLEX-jGCaMP8s and AAV-hSyn-DIO-hM3D(Gq)-mCherry were injected into the PVH of Th-Cre mice, and feeding behavior after saline or clozapine N-oxide (CNO) injection in 24-h fasting mice was video-recorded, along with GCaMP8s intensity measurement. (B) Representative GCaMP intensity trace lines in two mice injected with CNO. Blue color highlight indicates the duration for which the mice were in the food consumption phase. (C) Duration of the food consumption phase after the injection of saline (n = 20 mice) or CNO (n = 20 mice). (D) AAV-syn-FLEX-jGCaMP8s and AAV-hSyn-DIO-hM4D(Gi)-mCherry were injected into the PVH of Th-Cre mice, and feeding behavior after saline or CNO injection in 24-h fasting mice was video-recorded, along with GCaMP8s intensity measurement. (E) Representative GCaMP intensity trace lines in two mice injected with CNO. Blue color highlight indicates the duration for which the mice were in the food consumption phase. (F) Duration of the food consumption phase after the injection of saline (n = 14 mice) or CNO (n = 14 mice). (G–I) AAV-hSyn-DIO-hM3D(Gq)-mCherry or AAV-hSyn-DIO-hM4D(Gi)-mCherry was injected into the PVH of Th-Cre mice (G), and cumulative food intake was measured in 24-h fasted mice expressing hM3D(Gq) after the injection of saline or CNO (n = 8/group; two-way ANOVA; main effect of group (P < 0.0001), Sidak’s multiple-comparison test) (H) and mice expressing hM4D(Gi) after the injection of saline or CNO (n = 8/group; two-way ANOVA; main effect of group; P < 0.0001; Sidak’s multiple-comparison test) (I). Error bar: SEM. *P < 0.05, **P < 0.01, ***P < 0.005, and ****P < 0.001.

PVH Dopaminergic Neurons Suppress the Food-Seeking Behavior to Promote Eating.

To determine the specific elements of feeding behavior promoted by PVH dopaminergic neurons, we performed operant conditioning, where mice seek food by choosing the correct touch panel on a LED display; if successful, they obtain a small food pellet delivered from a food dispenser placed on the other side of the LED display (Fig. 6A). Video tracking analysis revealed that the control mice preferred to stay in front of the food dispenser, whereas the Sim1-Cre-specific Th deletion mice were more likely to stay in front of the LED display (Fig. 6B). Sim1-Cre-specific Th deletion mice spent more time in the LED zone and less time in the feeding zone compared to the control mice (Fig. 6 C–E), indicating stronger food-seeking behavior in Sim1-Cre-specific Th deletion mice. Consistently, Sim1-Cre-specific Th deletion mice exhibited better learning ability than the control mice, as indicated by the percentage of correct responses (Fig. 6F). The latency to respond to the LED display was comparable between the two groups (Fig. 6G), implying similar cognitive and physical functions. The long latency in obtaining the food pellet (Fig. 6H) could be due to the downregulation in the food consumption behavior and upregulation in the food-seeking behavior. These data suggest that PVH dopaminergic neurons play a role in suppressing the food-seeking behavior to promote food consumption.

Fig. 6.

PVH dopaminergic neurons suppress the food-seeking behavior to promote eating. (A) Operant conditioning was performed in a rectangular chamber with a touch panel LED display on one side and food dispenser on the other side, along with video recording. (B) Representative heatmap showing the duration and locations of a single mouse during the 15-min test period. (C) Average time spent in front of the LED display by Th^lox/lox (n = 12) and Th^lox/lox/Sim1-Cre (n = 22) mice during the 15-min test on the first day of the experiment. *P < 0.05 via Mann–Whitney test. (D and E) Time spent in the zone near the food dispenser (two-way ANOVA; main effect of group (P < 0.0001), Sidak’s multiple-comparison test) (D) and LED display (two-way ANOVA; main effect of group (P < 0.0001), Sidak’s multiple-comparison test) (E) was measured using infrared sensors for Th^lox/lox (n = 13) and Th^lox/lox/Sim1-Cre (n = 14) mice throughout the experimental period. (F–H) Percentage of correct responses to get rewards (two-way ANOVA; main effect of group (P < 0.0001), Tukey’s multiple-comparison test) (F), latency for mice to touch the panel [two-way ANOVA; main effect of group (P = 0.7861)] (G), and latency to get reward after touching the correct touch panel [two-way ANOVA; main effect of group (P < 0.0001), Tukey’s multiple-comparison test] (H) for Th^lox/lox (n = 13) and Th^lox/lox/Sim1-Cre (n = 14) mice during the 15-min test. Error bar: SEM. *P < 0.05, **P < 0.01, ***P < 0.005, and ****P < 0.001.

ARC AgRP and POMC Neurons Activate the PVH Dopaminergic Neurons.

As a large proportion of PVH-feeding neurons receive projections from the first-order feeding neurons, such as the NPY/AgRP and POMC neurons in ARC (4, 5), we investigated whether PVH dopaminergic neurons are part of the NPY and melanocortin pathways. Histological analysis revealed that the NPY and POMC fibers surrounded the cell bodies of TH neurons in PVH (Fig. 7 A and D). In addition, NPY and POMC fibers were attached to the cell bodies and fibers of TH neurons in PVH (Fig. 7 B, C, E, and F). To analyze the functional relationship between the NPY/AgRP and POMC neurons in ARC and dopaminergic neurons in PVH, we performed Th-FLPo-specific GCaMP fiber photometry of PVH dopaminergic neurons and examined the response to the optogenetic activation of ARC POMC and AgRP neurons (Fig. 7G and SI Appendix, Figs. S8 and S9). Unexpectedly, optogenetic activation of ARC AgRP and POMC neuronal fibers in PVH increased the activity of PVH dopaminergic neurons (Fig. 7 H–M), suggesting that the activation of ARC AgRP and POMC neurons leads to the activation of PVH dopaminergic neurons. AS the activity pattern of PVH dopaminergic neurons is similar to that of POMC neurons at the onset of the food procurement and consumption phases and AgRP neurons at the onset of the meal termination phase (Fig. 4 G–L), PVH dopaminergic neurons may be activated by POMC and AgRP neurons during these feeding phases.

Fig. 7.

ARC POMC neurons activate the PVH dopaminergic neurons. (A–C) TH immunohistochemistry (red) in the PVH of NPY-hrGFP mice. (D–F) TH and POMC double immunohistochemistry in the PVH of wild-type mice. Fluorescence microscopy images at low magnification (A and D), confocal microscopy images at high magnification (B and E), and Z-stack projections of confocal microscopic images (C and F). Arrowheads indicate the conjunctions of neurons. [Scale bar, 30 µm (A and D), 5 µm (B and E), and 1 µm (C and F).] (G-M) AAV-Syn-FLEX-rc(ChrimsonR-tdTomato) was injected into the ARC and AAV-Ef1a-fDIO-GCaMP6f was injected into the PVH of Th-FLPo/Agrp-Ires-Cre or Th-FLPo/Pomc-Cre mice (G). Representative GCaMP intensity trace lines in the PVH TH neurons of Th-FLPo/Agrp-Ires-Cre (H) or Th-FLPo/Pomc-Cre (K) mice. Gray color highlight indicates 10 s of optogenetic activation of AgRP or POMC fibers in the PVH. Trace lines with error bars indicate the average GCaMP intensities (±SEM) in Th-FLPo/Agrp-Ires-Cre (n = 12 trials with 4 mice) (I) and Th-FLPo/Pomc-Cre (n = 12 trials with 4 mice) (L) mice during optogenetic activation. Heatmap showing the percentages of normalized z-scores of GCaMP6f intensity in the PVH TH neurons during the optogenetic activation of AgRP (J) and POMC fibers (M) in the PVH.

PVH Dopaminergic Neurons Extend the Food Consumption Phase and Project to the Lateral Habenula (LHb).

To analyze the projection sites of PVH dopaminergic neurons, AAV-hSyn-DIO-EGFP, an anterograde AAV tracer, was injected into the PVH of Th-Cre mice. Projection sites existed both inside and outside the hypothalamus (Fig. 8A and SI Appendix, Fig. S10). Next, using the dopamine sensor, dLight1.1, we analyzed dopamine secretion at the projection sites while recording the feeding behavior. During the food consumption phase, dopamine was released in LHb, paraventricular nucleus of the thalamus (PV), nucleus accumbens (Acb), basolateral amygdala (BLA), and periaqueductal gray (PAG), but not in the lateral septum (LSI) (Fig. 8 B–E and SI Appendix, Fig. S11 and Movie S4). To analyze the dopamine released from the PVH dopaminergic neurons, PVH dopaminergic neurons were activated by DREADDs to examine dopamine release at the projection site. Activation of PVH dopaminergic neurons extended the dopamine release duration in LHb (Fig. 8 F–H), whereas inhibition of PVH dopaminergic neurons had the opposite effect (SI Appendix, Fig. S12). However, activation of PVH dopaminergic neurons did not extend the dopamine release duration in PV (Fig. 8 I–K). These data indicate that PVH dopamine is released, at least in LHb, during the food consumption phase. To confirm that the activation of PVH dopaminergic neurons induces the release of dopamine in LHb, dLight1.1 intensity in LHb was analyzed, and the response to optogenetic activation of PVH dopaminergic neurons was examined. Optogenetic activation of PVH dopaminergic neurons increased dopamine release in the LHb of ad libitum-fed mice, whose access to food was temporarily blocked during the experiment (Fig. 8 L–O), suggesting that PVH dopaminergic neurons release dopamine in LHb. We further analyzed the dopamine receptor subtypes in LHb mediating the orexigenic effects of PVH dopaminergic neurons (Fig. 8P). Although the injection of the D1R antagonist, SCH23390, into LHb did not alter the food intake induced by the DREADD activation of PVH dopaminergic neurons (Fig. 8Q), injection of raclopride, a D2R antagonist, into LHb decreased the food intake (Fig. 8R), suggesting that D2R in LHb partly mediates the orexigenic effects of PVH dopaminergic neurons. LHb is known to induce aversive food behavior (19), and projection of the glutamatergic neurons in LH to LHb negatively regulates the consumption of palatable calorically dense liquids (20). D2R-mediated inhibitory signals (21) possibly suppress aversiveness to promote food intake.

Fig. 8.

PVH dopaminergic neurons project to the lateral habenula (LHb) to extend the food consumption phase. (A) Anterograde tracer AAV-hSyn-Flex-Axon-EGFP was injected into the PVH of Th-Cre mice, and EGFP fluorescence was observed in multiple areas of the brain, including LHb, paraventricular nucleus of the thalamus (PV), periaqueductal gray (PAG), nucleus accumbens (Acb), and, basolateral amygdala (BLA), and lateral septum (LSI). Higher magnification images are shown on the Right side of each image. (Scale bar, 100 µm.) Representative trace lines of dLight1.1 intensity in mice injected with AAV-CAG-dLight1.1 into LHb (B) and PV (C) during feeding after 24-h fasting. Colored highlights indicate the duration of the food consumption phase. Heatmaps showing the percentages of the normalized z-scores of dLight1.1 intensity in LHb (D) and PV (E). The onset of food consumption is marked at 0 s (bar). (F–K) AAV-hSyn-DIO-hM3D(Gq)-mCherry was injected into the PVH and AAV-CAG-dLight1.1 was injected into the LHb (F) or PV (I) of Th-Cre mice. dLight1.1 intensity was measured after saline or CNO injection in 24-h fasted mice, along with the video-recording of the feeding behavior. Trace lines with error bars indicate the average dLight1.1 intensities (±SEM) in LHb (saline: n = 30 with 6 mice; CNO: n = 30 with 6 mice) (G) and PV (saline: n = 30 with 6 mice; CNO: n = 30 with 6 mice) (J) before and after the onset (bar) of the food consumption phase. Area under the curve (AUC) of the dLight1.1 intensity trace line in LHb (saline: n = 30 with 6 mice; CNO: n = 30 with 6 mice) (H) and PV (saline: n = 30 with 6 mice; CNO: n = 30 with 6 mice) (K) during the first 5 s after the administration of saline or CNO. (L–O) AAV-Syn-FLEX-rc(ChrimsonR-tdTomato) was injected into the PVH and AAV-CAG-dLight1.1 was injected into the LHb of Th-Cre mice (L). Representative trace line of dLight1.1 intensity in the LHb of Th-Cre mice is shown, with gray highlight indicating the 10-s optogenetic activation of PVH TH neurons (M). Average dLight1.1 intensity in LHb (n = 12 trials with 4 mice) during optogenetic activation (N). Heatmap showing the normalized z-scores of dLight1.1 intensity in LHb neurons during the optogenetic activation of PVH TH neurons (O). (P and R) AAV-hSyn-DIO-hM3D(Gq)-mCherry was injected into the PVH of Th-Cre mice, and a guide cannula was implanted into LHb (P). Saline or dopamine receptor D1 antagonist SCH23390 (0.2 µg; saline: n = 11; SCH23390: n = 5; two-way ANOVA; main effect of group; P = 0.1785) (Q), or dopamine receptor D2 antagonist raclopride (0.3 µg; saline: n = 11; raclopride: n = 5; two-way ANOVA; main effect of group (P < 0.0001), Sidak’s multiple-comparison test) (R) was injected into LHb before the intraperitoneal injection of CNO. Cumulative food intake was subsequently quantified. Error bar: SEM. *P < 0.05, **P < 0.01, and ****P < 0.001.

Discussion

Here, we showed that PVH dopaminergic neurons constitute a previously uncharacterized distinct neuronal population in PVH that is indispensable for the epigenetic induction of obesity and upregulated in obese individuals. Functionally, PVH dopaminergic neurons were involved in extending the intermediate eating stage, specifically the food consumption phase. Qualitatively, PVH dopaminergic neurons suppressed the food-seeking behavior, thus promoting food consumption. PVH dopaminergic neurons seem to be activated by POMC neurons during the food consumption phase, highlighting the short-term role of POMC neurons in the feeding behavior. The projection of PVH dopaminergic neurons to LHb and D2R-mediated responses mediate the feeding-enhancing effect of PVH dopaminergic neurons.

Dopaminergic Neurons in PVH Constitute a Distinct PVH Neuronal Population.

In this study, we focused on PVH Sim1-Cre-expressing dopaminergic neurons. Various neuronal populations in PVH are involved in the control of feeding behavior (5). A previous (22) study also showed that PVH dopaminergic neurons exhibit a limited overlap with other neurons in PVH, such as CRH, TRH, oxytocin, AVP, NUCB2, and prodynorphin neurons. This distinction from other neuronal populations is consistent with the previously reported single-cell RNA-seq data of hypothalamic neurons (23). Dopaminergic neurons often coexpress GABA (23), but only a small proportion of dopaminergic neurons colocalized with GABA in PVH in this study, consistent with a previous report (24). Therefore, PVH dopaminergic neurons differ from most known PVH neurons and form a discrete neuronal population in PVH.

Functions of PVH dopaminergic neurons are also distinct from those of most other known feeding-related neurons in PVH. Unlike the PVH dopaminergic neurons, primary function of Sim1-Cre neurons is to suppress food intake and prevent obesity, as shown in the mouse models with genetically ablated, chemogenetically inhibited, or Mc4r-deficient Sim1-Cre neurons (6, 25, 26). In these models, the phenotype of PVH dopaminergic neurons is possibly masked by the satiety effects of other predominant Sim1-Cre neurons, including those expressing glutamate, prodynorphin, BDNF, and oxytocin (4, 27–29). PVH neurons encompass a transcriptionally diverse population with various calcium response patterns, showing both activation and inhibition during hunger-induced and hedonic eating, as shown via single-cell analyses (5). This suggests that not all PVH neurons function as satiety neurons. Furthermore, home-cage foraging and feeding-related behaviors, such as sniffing and digging, are enhanced by the chemogenetic activation of PVH neurons and suppressed by their chemogenetic inhibition (30), supporting the role of PVH neurons in promoting the feeding behavior.

PVH Dopaminergic Neurons are Indispensable for the Epigenetic Induction of Obesity.

Increase in body weight of Sim1-Cre-specific Dnmt3a deletion mice was abolished by double Dnmt3a and Th deletion in Sim1-Cre neurons. Double-deletion mice exhibited energy intake and expenditure patterns comparable to those of wild-type mice. Therefore, lack of weight gain in these mice was possibly due to normal energy intake and expenditure. Similarly, normal energy intake and expenditure observed in Sim1-Cre-specific Th deletion mice fed a normal chow diet indicate that Th expression in PVH is not necessary for controlling the normal energy balance. Hence, Th expression in PVH and dopamine synthesis are dispensable for normal energy balance but significant contributing factors to obesity development. Consistently, PVH Th expression levels were markedly elevated in both diet-induced and genetically induced obese mouse models, and HFD-induced obesity was partly suppressed by Sim1-Cre-specific deletion of Th.

Th appears to be a key target gene for DNMT3A and DNA methylation, as indicated by its highest upregulation in the genome-wide analysis of the PVH of Sim1-Cre-specific Dnmt3a deletion mice (10). This upregulated Th expression is also observed in the pancreatic β-cells of progenitor β-cell-specific Dnmt3a deletion mice as well as mice fed a chronic HFD (31), underscoring the sensitivity of Th expression to aberrant DNMT3A-induced DNA methylation.

In vitro studies have shown that Th expression is affected by the level of DNA methylation of the Th promoter region (32). The drastic increase in Th expression observed with aberrant DNA methylation can be attributed to genomic imprinting, a phenomenon in which one parental allele is preferentially expressed because of epigenetic modifications. Th is located within a genomic imprinting cluster locus (33) and imprinted in some tissues, including the placenta and specific neurons (34–36). Therefore, Th is an important effector of Dnmt3a in various tissues, including PVH.

PVH Dopaminergic Neurons Extend the Intermediate Eating Stage.

This study highlights the pivotal role of PVH dopaminergic neurons in enhancing food intake during the intermediate eating stage. Two different measurement methods for food intake resulted in different time-dependent food intake amplitudes in control mice, possibly due to differences in food form (solid vs. powder), acclimation period (several weeks vs. several days), and food accessibility (above vs. below the cage). Nevertheless, both methods showed a discernible difference in food intake between control and PVH-specific Th deletion mice approximately 60 min after feeding onset, i.e., the intermediate eating stage. Furthermore, c-Fos expression data support the activation of PVH dopaminergic neurons during the intermediate eating stage. As excessive food consumption is the primary contributor to obesity (2), the observed increase in food intake during the intermediate eating stage possibly underlies hyperphagia in obesity.

Three-Phase Feeding Behavior Analysis.

Using an open-field feeding behavior analysis system, we clearly distinguished among the three feeding phases of mouse behavior by observing the dynamic changes in neural activity in the PVH dopamine, AgRP, and POMC neurons at the onset of each phase. Although dynamic changes in AgRP and POMC neuronal activities have been reported to occur immediately after the sensory detection of food availability in the chamber (9), our analysis in an open-field revealed changes in the activities of AgRP and POMC neurons at the onset of each feeding phase, which persisted even after food availability detection. Because an open-field implies risk-taking behavior in contrast to the home cage environment, modulation of self-preserving neural circuits to hypothalamic feeding circuits (37) possibly underlies the cyclic activation of feeding neurons according to the feeding phase. Consistent with a previous report (38), activity patterns of AgRP and POMC neurons contrasted each other at each phase.

Notably, our data suggest that PVH dopaminergic neurons are activated by both POMC and AgRP neurons. Considering the simultaneous activation of PVH dopaminergic neurons and POMC (food procurement and consumption phases) or AgRP (termination phase) neurons, PVH dopaminergic neurons are possibly activated by both POMC and AgRP neurons. AgRP neurons, but not POMC neurons, are inhibitory in nature (39–41). The mechanism by which the projection of AgRP neurons to PVH dopaminergic neurons activates them requires further investigation. POMC neurons are well-known satiety neurons (42–45); however, the satiety effect of POMC neurons in ARC is chronic, not acute (42, 43). A previous study indicated that POMC neurons promote acute feeding induced by cannabinoids (46). POMC-induced activation of PVH dopaminergic neurons possibly extends the food consumption phase via an acute effect.

PVH Dopaminergic Neurons Extend the Food Consumption Phase.

In this study, we demonstrated that PVH dopaminergic neurons played a critical role in extending the food consumption phase. Although the food procurement phase is mainly controlled by the NPY/AgRP neurons in ARC (7, 9, 47, 48), specific neurons governing the food consumption phase have not yet been revealed. Although LH neurons have been implicated in the food consumption phase because of increased food intake after their electrical stimulation (49), a subset of GABAergic LH neurons has been reported to enhance consumption regardless of the caloric content (50, 51), suggesting their involvement in generalized consummatory or compulsive-like feeding behavior. Through preferential localization during operant conditioning of PVH-specific Th deletion mice and chemogenic activation and inhibition of PVH dopaminergic neurons, PVH dopaminergic neurons critically extend the food consumption phase. In addition, PVH dopaminergic neurons mediate food-specific consummatory feeding because chemogenic activation of PVH dopaminergic neurons does not induce robust water drinking. GCaMP fiber photometry revealed that PVH dopaminergic neurons were also activated at the onset of other feeding phases. However, the specific roles of PVH dopaminergic neurons in these phases need to be investigated in future studies.

Projection of PVH Dopaminergic Neurons to LHb Mediates the Enhanced Feeding Behavior.

Our findings indicate that PVH dopaminergic neurons project to LHb, increase the duration of dopamine release in LHb during the food consumption phase, and increase food intake through D2R in LHb. Supporting our results, the medial subdivision of LHb (LHbM) receives projections from PVH, according to retrograde analysis (52) and expresses TH-immunopositive fibers and D2R mRNA (53). LHb is associated with negative valence (19, 54), and excitatory input into LHb from glutamatergic neurons in LH suppresses the consumption of calorically dense liquids, causing aversiveness (20). This pathway also responds to bitter taste (55), and its inhibition contributes to hunger-induced tolerance to bitter taste downstream of AgRP neurons (56). Inhibition of Npy1r-expressing LHb neurons enhances stress-induced palatable food consumption (57). These data indicate that the suppression of LHb neurons enhances the food intake. As D2R generally induces inhibitory input through Gα_i/o-mediated signaling (21), dopamine release from PVH dopaminergic neurons to D2R in LHb possibly suppresses the negative valence and extends the food consumption phase. The feeding-enhancing effect of D2R in LHb is consistent with previously reported D2R knockout mouse phenotypes, which are hypophagic and exhibit reduced body weight compared to the control mice (58). Additionally, positron emission tomography measurements of human subjects using a radiolabeled D2R antagonist have indicated that dopamine release in the habenular complex occurs at a relatively early phase and is related to a high desire for food (59), further supporting the feeding-enhancing effect of dopamine release in LHb. Overall, these data support the pivotal role of D2R in LHb in enhancing the feeding behavior.

Technical Limitations.

One technical limitation of this study was the use of Sim1-Cre mice for PVH-specific deletion of Th. We confirmed that Sim1-Cre was not expressed in TH neurons in most brain areas outside central PVH. However, Sim1-Cre was expressed in various other regions, including some peripheral tissues. Additionally, as suggested by our findings, Th expression can be influenced by both external and internal conditions. Th expression patterns observed under normal diet conditions may change under other conditions. Therefore, we cannot entirely exclude the possibility that Th expression is altered in regions beyond PVH in Sim1-Cre-specific Th knockout mice, thus impacting their long-term phenotypes, including feeding behavior.

In summary, PVH dopaminergic neurons downstream of the melanocortin pathway extended the food consumption phase through projections to LHb, playing an indispensable role in the epigenetic induction of obesity.

Materials and Methods

Video Recording and Analysis of Feeding Behavior.

After 24 h of food deprivation, a mouse was placed in a 29 cm × 39 cm chamber, with a food pellet placed vertically in a folder at the center of the chamber and a water bottle placed in the corner at ZT12. Video recording was performed from the top using an IP camera, (rlc-811a; Reolink). The duration of each behavior was measured by analyzing the recorded videos. We defined mouse behaviors as follows in Fig. 4. Bout refers to biting food, food contact includes both sniffing and biting food, exploring refers to actively searching for something like food, and walking refers to moving around without searching. The three phases of the feeding behavior were determined as follows: The onset of the food procurement phase was when the mouse entered a 6-cm radius circle with a food pellet placed at the center to approach the pellet, the food consumption phase began when the mouse started sniffing or biting the food, and meal termination was defined as the point at which the mouse moved away from the food pellet. All behavioral experiments and analyses were conducted by individuals blinded to the genotypes and injected AAVs.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.