Elevated activity of the mesolimbic dopamine system promotes feeding during pregnancy in mice

Abstract

The pregnancy period is accompanied by increased feeding behavior to accommodate the elevated energy demands associated with fetal growth and development. However, the underlying neural circuitry and molecular mechanisms mediating increased feeding during pregnancy are largely unknown. Here, we utilized a combination of fiber photometry, chemogenetics, and mouse behavioral assays to characterize altered feeding behavior during pregnancy in mice. We uncover that pregnancy increases the average activity of the mesolimbic dopamine system during feeding behavior in mice. VTA dopamine neurons promote increased high fat diet feeding during pregnancy as inhibition of these cells selectively reduces acute high fat diet intake in pregnant mice. Further, pregnant mice exhibit increased sensitivity to food deprivation, an effect which requires activity of the mesolimbic dopamine system. Together, these findings provide a circuit basis mediating altered palatable food intake and sensitivity to negative energy balance during pregnancy in mice.

Highlights

• •Average VTA dopamine response to palatable food is increased during pregnancy.
• •Average VTA dopamine neuron activity is increased in response to negative energy balance during pregnancy.
• •Nucleus accumbens dopamine levels are increased during feeding in pregnant mice.
• •VTA dopamine neuron activity regulates palatable food intake and fast-induced refeeding in pregnant mice.

1. Introduction

Mammals increase food intake during pregnancy to accommodate the elevated energy demands associated with fetal development [1,2]. Ultimately, increased feeding during pregnancy leads to a state of positive energy balance, resulting in elevated fat storage, which provides readily available energy stores for the metabolically demanding process of nursing [1,2]. This increase in feeding is an adaptive response to pregnancy, promoting healthy fetal development and growth and a healthy metabolic response to pregnancy in the mother. Conversely, impairments in energy homeostasis (i.e. excessive caloric intake) during pregnancy are associated with an increased risk of both the mother and her children developing obesity and type 2 diabetes later in life [[1], [2], [3], [4], [5]]. However, the core cellular and molecular mechanisms leading to increased feeding during pregnancy are incompletely understood.

Feeding behavior is initiated by multiplexed neural circuits located in the hypothalamus, midbrain (i.e. ventral tegmental area), and hindbrain [6,7]. Neurons in these regions integrate hormonal and neuroendocrine cues which signal changes in long-term energy stores (i.e. via leptin from fat and ghrelin from the stomach) to initiate food intake during conditions of energy deprivation [6,7]. Following food consumption, meal derived satiety signals and GI distention regulate hypothalamic, midbrain, and hindbrain circuits to terminate ongoing feeding behavior [8]. Together, these processes control energy homeostasis by matching energy intake to changes in energy expenditure (i.e. homeostatic feeding behavior). Animals will also voluntarily consume palatable foods (i.e. foods high in fat and/or sugar) in the absence of energy deprivation, an effect that has previously been described as hedonic feeding behavior [9,10]. While the underlying neural circuitry controlling feeding behavior is extensively studied, the effect of pregnancy on these neural pathways is not well understood.

Among the neuronal pathways controlling feeding, dopaminergic pathways are essential for promoting food seeking and consumption [9,11]. Dopamine deficient mice are aphagic and die of starvation unless provided supplemental nutrition via a feeding tube, suggesting that dopamine promotes food seeking and/or consumption [11]. Consistent with this notion, dopamine neurons in the ventral tegmental area (VTA) are robustly activated during food seeking and consumption of palatable food and exert an important role in signaling the post-ingestive reward associated with the consumption of calories [9,[12], [13], [14], [15]]. VTA dopamine neurons project throughout the brain, including to the nucleus accumbens, amygdala, and prefrontal cortex [9]. In particular, VTA dopamine projections to the nucleus accumbens (NAc; mesolimbic dopamine pathway) represent a critical neural circuit controlling reward seeking behavior, including the seeking and consumption of palatable food [9,10]. Although most data suggest that dopamine promotes food seeking and palatable food intake, pharmacological studies provide contrasting results. For example, many dopamine mimetics are potently anorexic (i.e. cocaine and amphetamines), suggesting an appetite suppressive effect of dopamine transmission [9,16,17]. Thus, the specific role of dopamine in feeding is debated, and the contribution of dopamine to feeding behavior during pregnancy is largely unknown. Here, we utilized a combination of mouse feeding assays, in vivo fiber photometry of dopaminergic transmission, and chemogenetics to characterize the role of mesolimbic dopamine signaling in feeding behavior during pregnancy.

2. Results

2.1. Pregnant mice exhibit hyperphagia for both regular chow and high fat diet

Although pregnancy is known to increase food intake in rodents [1,2], the specific effect of pregnancy on feeding behavior is incompletely understood. To test if pregnant mice exhibit elevated palatable food consumption in the minutes following access to palatable foods, we provided palatable peanut butter (PB) chips to non-pregnant and pregnant ad libitum fed mice. Mice were provided with PB chips daily for 10 min for three days prior to testing to habituate the mice to PB chips and prevent food neophobia. Pregnant mice consumed more PB chip in the minutes following PB chip presentation compared to non-pregnant animals, indicating that pregnant mice exhibit enhanced palatable food intake during acute presentation of palatable foods (Fig. 1A). Next, we quantified daily food intake in non-pregnant and pregnant mice during the consumption of regular chow diet or a 60% high fat diet. Consistent with prior reports [1,18], pregnant mice consume approximately 25% more calories than age-matched non-pregnant mice when provided regular chow diet (Fig. 1B). To characterize palatable food intake in pregnant mice during longer periods of access to palatable high fat foods, we provided ad libitum access of a palatable high fat diet (60% fat) to non-pregnant and pregnant mice. As expected, both pregnant and non-pregnant mice significantly increased their caloric intake following 24-hour access to high fat diet (compared to regular chow intake; Figure 1B,C). However, pregnant mice continued to consume significantly more calories than non-pregnant mice for three consecutive days following HFD access. When calorie intake was adjusted to account for the percent increase in calorie intake on HFD (Fig. 1C, vs baseline regular chow intake), pregnant and non-pregnant mice exhibited a similar level of hyperphagia when provided with high fat diet. Further, in both non-pregnant and pregnant mice, calorie intake is similarly reduced on the second and third day of HFD access, compared to the initial day of HFD access, indicating that pregnant mice appropriately adapt their voluntary food intake in response to long-term access to hypercaloric diets (Fig. 1C).

Figure 1.

Pregnant mice exhibit hyperphagia for both regular chow and high fat diet. (A) Peanut butter chip intake in non-pregnant and pregnant mice during an acute 10-minute feeding assay. (B) Daily food intake (in kcal) of non-pregnant and pregnant mice (third trimester) fed a regular chow diet (RC) or a high fat diet (HFD). (C) Change in food intake upon access to high fat diet compared to regular chow food intake in non-pregnant and pregnant mice. (D–F) Pellets consumed (D), meal size (E), and number of meals (F) during fixed-ratio 3 (FR3) food seeking tasks in non-pregnant and pregnant mice. (G) Relationship between the percentage of meals consumed and the size of meals in non-pregnant and pregnant mice during FR3 feeding assays. (H–J) Pellets consumed (H), meal size (I), and number of meals (J) during fixed-ratio 5 (FR5) food seeking tasks in non-pregnant and pregnant mice. (K) Relationship between the percentage of meals consumed and the size of melas in non-pregnant and pregnant mice during FR5 food seeking tasks in non-pregnant and pregnant mice. Data points represent individual mice. Panels A, D, E, F, H, I, and J analyzed with unpaired Students t-test. Panels B and C, analyzed by 2-way ANOVA. Panels in G and K analyzed using multiple t-tests followed by Holm Sidak multiple comparisons' correction. ns (not significant), ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.005.

2.2. Pregnant mice consume larger meals in operant food seeking assays

Having characterized the feeding response to ad libitum access of regular chow or high fat diets, we next utilized home-cage operant feeding devices (feeding experimental device 3; FED3) to describe operant food seeking behavior in non-pregnant and pregnant mice [19]. Non-pregnant and pregnant mice were trained to nose poke on the left nose poke port to receive a single 20 mg food pellet. Following successful training (greater than 70% of nose pokes occurring on the correct nose poke port), we quantified the meal size, meal frequency, and total pellets consumed over 24 h in non-pregnant and pregnant mice during a fixed ratio 3 schedule of reinforcement (3 correct nose pokes leads to 1 20 mg food pellet). In contrast to free access feeding conditions (Fig. 1B), pregnant mice did not consume more pellets than non-pregnant mice during FR3 feeding assays (Fig. 1D). There was, however, a significant shift in meal size and frequency in pregnant animals with pregnant mice consuming larger meals (Fig. 1E, Extended Data Fig. 1) but eating less frequently than non-pregnant mice (Fig. 1F). Pregnant mice exhibited a dramatic shift in the proportion of meals larger than 5 pellets, indicating that pregnant mice shift their meal structure towards larger, and more infrequent meals compared to non-pregnant animals (Fig. 1G). Similar findings were also observed during a fixed ratio 5 schedule of reinforcement (Fig. 1H–K). Therefore, pregnant mice exhibit altered feeding structure during operant food seeking assays, favoring larger and more infrequent meals compared to non-pregnant animals.

2.3. Population activity of VTA dopamine neurons is increased during feeding in pregnant mice

Although palatable food intake and meal structure is altered in pregnant mice (Fig. 1), the underlying neural circuitry mediating these changes is unknown. Given that ventral tegmental area (VTA) dopamine neurons modulate both palatable food intake and meal size in mice [9,10,12], we next utilized in vivo fiber photometry of the average calcium levels in VTA dopamine neurons to characterize changes in VTA dopamine activity in pregnant mice. DAT-cre transgenic mice were injected with a cre-dependent version of the genetically encoded calcium indicator GCAMP6s into the VTA and a fiber optic cannula was positioned into the VTA (Fig. 2A) to record changes in calcium activity in VTA dopamine neurons as a proxy of neuronal activity. Following recovery from surgeries, mice were randomly separated into a pregnant group, which was mated with a male mouse, and a non-pregnant, age-matched control group (Fig. 2B). Given the critical role of VTA DA neurons in palatable food intake we first measured the average calcium activity of VTA DA neurons in response to the presentation and consumption of high fat diet (60% fat) in non-pregnant and pregnant ad libitum fed mice. No difference in the VTA DA calcium response to the presentation of HFD (after presentation and prior to consumption) was observed between non-pregnant and pregnant mice (Extended Fig. 2A–B). As expected, the population activity of VTA DA neurons increased in both non-pregnant and pregnant mice following consumption of HFD (Fig. 2C; Extended Data Fig. 2C–E). The average increase in VTA DA activity was significantly larger in pregnant mice than in age matched non-pregnant animals during HFD consumption (Fig. 2D; Extended Data Fig. 2C–E). However, pregnant and non-pregnant mice exhibited a similar maximum response in VTA DA activity following HFD presentation (Fig. 2A). The increased calcium signal in pregnant mice during HFD consumption was not due to increased HFD intake in pregnant mice, as pregnant and non-pregnant mice consumed a similar amount of HFD during the brief 10-min photometry session (Extended Fig. 2F).

Figure 2.

Population level activity of VTA dopamine neurons exhibits increased responsivity to feeding during pregnancy. (A) Representative image showing expression of FLEX-GCaMP6s virus in VTA with fiber location. (B) Schematic showing experimental setup. (C) Average trace of the calcium signal in VTA dopamine neurons following the consumption of high fat diet. (D and E) Mean change in calcium signal (D) and maximum change (E) in VTA dopamine neurons following HFD consumption in non-pregnant and pregnant mice. (F) Average trace of the calcium signal in VTA dopamine neurons following the consumption of food after a 10 h fast in non-pregnant and pregnant mice. (G and H) Average change in calcium signal (G) and maximum change in calcium signal (H) following consumption of standard chow in non-pregnant and pregnant mice. Data in C and F represents average signal with standard error of the mean from all mice. Data points in D, E, G, and H represent individual mice. Data analyzed with unpaired Student's t-test. ns (not significant), ∗p < 0.05. Scale bar in A (500 μm), Scale bar in C (20 s), Scale bar in F (20 s).

In addition to promoting the consumption of palatable foods, VTA dopamine activity is also involved in promoting food seeking behaviors and reinforcement associated with food consumption in energy deprived mice [9,14,15,20]. Therefore, we next tested if the average VTA DA response to standard food consumption differed between non-pregnant and pregnant mice following a brief period of energy deprivation (10 h fast). Both non-pregnant and pregnant mice were fasted for 10 h and presented with familiar standard chow in their home cage while measuring the activity of VTA DA neurons with fiber photometry. No difference in the average VTA DA response was detected between pregnant and non-pregnant mice following the presentation of a food pellet (prior to eating; Extended Fig. 2G–H). In contrast, the average signal intensity in VTA dopamine neurons in pregnant mice following food consumption was significantly higher than in non-pregnant mice (Fig. 2F,G; Extended Data Fig. 2I–K). Further, the maximum increase in VTA DA activity was significantly higher in pregnant mice than non-pregnant animals following consumption of standard chow (Fig. 2H). These differences were not secondary to increased food consumption in pregnant mice as both non-pregnant and pregnant mice consume similar amounts of food during the acute testing session (Extended Data Fig. 2L). Thus, the population activity of VTA dopamine neurons is enhanced in pregnant mice during feeding (Fig. 2).

2.4. Nucleus accumbens dopamine response is increased in pregnant mice during feeding

Although VTA dopamine neurons project to multiple downstream brain regions, the neuronal projection to the nucleus accumbens is particularly important for promoting palatable food intake and food seeking behaviors [9,10,21]. Therefore, we next utilized genetically encoded dopamine sensors [22] and fiber photometry to quantify the levels of dopamine transmission in the nucleus accumbens during feeding behaviors in non-pregnant and pregnant mice (Fig. 3). Genetically encoded dopamine sensors (GRAB-DA) were targeted to the nucleus accumbens (NAc), and a fiber optic cannula was positioned directly above the NAc to quantify changes in dopamine levels with fiber photometry (Fig. 3A,B; Extended Data Fig. 3A). First, we measured changes in NAc dopamine levels prior to pregnancy and during the pregnancy period (or the equivalent time-period in non-pregnant control mice) after providing high fat diet to mice. No difference in the NAc dopamine response was detected between non-pregnant and pregnant mice following the presentation of a high fat pellet (i.e. prior to consumption; Extended Fig 4A–B). The average increase in NAc dopamine levels following HFD consumption was significantly increased in pregnant mice, compared to baseline measurements prior to pregnancy (Fig. 3C–D; Extended Fig. 4C–E). In contrast, the maximum increase in NAc dopamine levels was similar pre-pregnancy and during pregnancy in the minute following HFD consumption (Fig. 3E). Importantly, no difference in the NAc dopamine response to HFD consumption was observed in both time periods in non-pregnant control mice, indicating that these differences likely don't result from order effects associated with repeated experiments (Fig. 3I). Further, a similar trend towards increased NAc dopamine levels was also observed in pregnant mice compared to non-pregnant animals tested on the same day (Extended Data Fig. 4F).

Figure 3.

Dopamine levels in the nucleus accumbens exhibit increased responsivity to feeding during pregnancy. (A) Representative image showing expression of GRAB dopamine sensor in nucleus accumbens with fiber location in nucleus accumbens. (B) Timeline for experiments shown in this figure. (C) Average trace of the nucleus accumbens dopamine signal following high fat diet consumption prior to pregnancy and during the pregnancy period. (D and E) Quantification of the data shown in C, comparing the mean Z-score following HFD consumption (D) and the maximum signal (E) following HFD consumption in non-pregnant and pregnant mice. (F) Average trace of the nucleus accumbens dopamine signal following food consumption after a 10 h fast before pregnancy and during the pregnancy period. (G and H) Quantification of the data shown in F, comparing the mean Z-score following food consumption (G) and the maximum Z-score following food consumption (H). (I and J) Average change in dopamine signal in non-pregnant mice at the two experimental timepoints shown in B following HFD consumption (I) or food intake following a 10 h fast (J). (K and L) Mean change in dopamine signal (K) and maximum change in dopamine signal (L) following the presentation of food to non-pregnant and pregnant mice. Mice were not food deprived prior to food presentation in panels K and L. Data points represent individual mice. All panels analyzed by paired Student's t-test. ns (not significant), ∗p < 0.05. Scale bar in A (300 μm).

Since we observed an increased average VTA calcium response to regular chow consumption in energy deprived mice during pregnancy (Fig. 2), we next measured changes in NAc dopamine levels pre-pregnancy and during the pregnancy period following food consumption after a 10 h fast (Fig. 3F–H). The average increase in nucleus accumbens dopamine levels was significantly greater in the 60 s following food consumption in pregnant mice compared to the pre-pregnancy period (Fig. 3G; Extended Fig. 4G–I). However, mice exhibited a similar maximum change in NAc dopamine levels in the 60 s following food consumption during the pre-pregnancy period and the pregnancy period (Fig. 3H). No change in the average dopamine response was observed in the 60 s following food consumption in non-pregnant mice during the two recording sessions, indicating that these changes do not result from order effects associated with repeated photometry sessions (Fig. 3J). Furthermore, nucleus accumbens dopamine levels were also elevated in pregnant mice compared to non-pregnant control animals on the same testing day (Extended Data Fig. 4J), further validating increased NAc dopamine levels during food consumption in energy deprived pregnant mice.

To test if the increased NAc dopamine response during regular chow food consumption in pregnancy is specific to food consumption we also quantified the NAc dopamine levels in non-pregnant and pregnant mice following food presentation after a 10 h fast (prior to the consumption of food). No difference in the NAc dopamine response was detected between non-pregnant and pregnant mice after food presentation in the ad libitum fed state (Fig. 3K–L), or in animals that were fasted for 10 h (Extended Data Fig 4K–L). Thus, pregnant animals do not exhibit a broad increase in NAc dopamine response to food-related stimuli (in the absence of training) and only exhibit an increased NAc dopamine response during the consumption of food.

2.5. VTA dopamine neurons regulate meal size and number in both pregnant and non-pregnant mice

Our prior fiber photometry data (Figure 2, Figure 3) indicate that pregnancy alters the population activity of VTA DA neurons and dopaminergic transmission in the nucleus accumbens during feeding. We thus hypothesized that VTA dopamine neuron activity regulates feeding behavior during pregnancy in mice. To test this hypothesis, we targeted the chemogenetic inhibitor hM4Di or control virus expressing a fluorescent protein to VTA dopamine neurons in DAT-cre mice, and inhibited VTA dopamine neurons during feeding tasks in non-pregnant and pregnant mice (Fig. 4A; Extended Data Fig. 3B,C). Feeding assays were performed using feeding experimental devices (FED3) to quantify changes in food intake, meal size, and meal number following inhibition of VTA dopamine neurons. CNO-mediated inhibition of VTA dopamine neurons did not alter the number of pellets consumed in non-pregnant or pregnant mice which were provided ad libitum access to food (Fig. 4B). As previously described (Fig. 1), pregnant mice consumed larger meals than non-pregnant mice (Fig. 4C). A similar increase in meal size was observed in pregnant mice compared to non-pregnant mice following either saline or CNO injections, indicating that VTA dopamine activity is not required for promoting increased meal size in pregnant mice (Fig. 4C). Inhibition of VTA dopamine neurons led to a similar increase in meal size in both non-pregnant and pregnant mice, and an equivalent decrease in meal number in both non-pregnant and pregnant mice (Fig. 4C,D). No differences between saline or CNO injections were observed for pellets consumed, meal size, or meal number in both non-pregnant and pregnant mice expressing control mCherry virus in VTA dopamine neurons, indicating no off-target effects of CNO or viral expression (Extended Data Fig. 5). Therefore, inhibition of VTA dopamine neurons increases meal size and reduces meal number to a similar extent in both non-pregnant and pregnant mice, and the VTA dopamine neurons are not required for the increased meal size observed in pregnant mice.

Figure 4.

VTA dopamine neurons contribute to increased high fat diet intake and increased sensitivity to negative energy balance in pregnant mice. (A) Representative image of hM4Di-mCherry expression in the VTA. (B) Pellet consumption in non-pregnant and pregnant mice expressing hM4Di-mCherry in VTA dopamine neurons following i.p. injections of saline or CNO. (C and D) Meal size (C) and number of meals (D) following saline or CNO injections in non-pregnant and pregnant mice expressing hM4Di-mCherry in VTA dopamine neurons. (E) Pellets consumed in non-pregnant and pregnant mice expressing either hM4Di-mCherry or control mCherry virus in VTA dopamine neurons following a 10 h fast. All mice were administered CNO 10 min prior to testing for the experiments shown in E. (F) Schmatic showing experimental design for experiments shown in panels G–J. (G) Regular chow food intake in 1 h in non-pregnant and pregnant mice. Although pregnant mice consume more daily calories than non-pregnant mice, no significant difference in 1-hour standard chow food intake is detectable between non-pregnant and pregnant mice during the light period (i.e. during the timepoints in which acute HFD intake was measured in H-J). (H) 1 h intake of high fat diet in non-pregnant and pregnant mice. (I) High fat diet intake in non-pregnant mice expressing either control mCherry virus or hM4Di-mCherry in VTA dopamine neurons following injections of CNO. (J) High fat diet intake in pregnant mice expressing either control mCherry virus or hM4Di-mCherry in VTA dopamine neurons following injections of CNO. Data points represent individual mice. Scale bar in A (300 μm). Data in B, C, D, E, I, and J analyzed by 2-way ANOVA with Sidak's multiple comparisons' test. Data in panels G and H analyzed with unpaired Student's t-test. ns (not significant), ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.001.

2.6. VTA dopamine neurons regulate food seeking in pregnant and non-pregnant energy deprived mice

The average activity of VTA dopamine neurons is increased following food consumption in food deprived pregnant mice (Fig. 2). To test if VTA dopamine activity contributes to increased food seeking in fasted mice during pregnancy, we measured food intake in non-pregnant and pregnant mice following a 10 h fast. Mice expressing either control mCherry virus or hM4Di in VTA dopamine neurons were administered CNO (1 mg/kg) immediately prior to food presentation after a 10 h fast. After an acute fast, pregnant mice consumed more pellets than non-pregnant mice following CNO administration in animals expressing control mCherry virus in VTA dopamine neurons (Fig. 4E). However, this effect was attenuated in mice expressing hM4Di in VTA dopamine neurons, such that food intake levels were not significantly different between pregnant and non-pregnant mice during inhibition of VTA dopamine neurons (Fig. 4E). In contrast to ad libitum fed conditions (Fig. 4B), inhibition of VTA dopamine neurons significantly reduced food intake in both non-pregnant and pregnant energy deprived mice, suggesting that VTA dopamine activity is specifically involved in promoting food seeking for regular chow during conditions of negative energy balance.

2.7. VTA dopamine neurons modulate palatable food intake in pregnant mice

Dopamine levels in the nucleus accumbens are elevated during palatable food consumption (Fig. 3D), and pregnant mice exhibit increased population-level VTA dopamine activity during palatable food consumption compared to non-pregnant mice (Fig. 2D). We thus hypothesized that increased VTA dopamine activity may promote excessive palatable food intake in pregnant mice. To measure acute changes in palatable food intake in non-pregnant and pregnant mice we provided non-pregnant and pregnant mice with standard chow or palatable high fat diet and measured food intake 1-hour later (Fig. 4F–H). Although we were unable to detect increased intake of regular chow in pregnant mice during this acute time-period (Fig. 4G), consistent with prior acute PB chip feeding assays (Fig. 1A), pregnant mice consumed significantly more HFD than non-pregnant mice in acute HFD feeding assays (Fig. 4H). To test if VTA dopamine activity is involved in promoting palatable food intake in pregnant mice, we measured acute HFD intake in non-pregnant and pregnant mice expressing either control mCherry virus or hM4Di in VTA dopamine neurons (Fig. 4I–J). Inhibition of VTA dopamine neurons did not significantly alter HFD intake in non-pregnant mice (Fig. 4I). In contrast, chemogenetic inhibition of VTA dopamine neurons slightly reduced high fat diet intake in pregnant mice (Fig. 4J). Thus, VTA dopamine neurons modulate the consumption of palatable high fat diet in pregnant mice.

3. Discussion

The pregnancy period requires significant increases in energy intake to accommodate the metabolic demands associated with fetal growth and development [1,2]. Although pregnancy is known to increase feeding behavior, the specific neural circuitry and molecular mechanisms mediating increased feeding during pregnancy are largely unknown. Further, the behavioral mechanisms leading to increased caloric intake during pregnancy are incompletely understood. Results presented here are consistent with previous findings showing approximately 25% more food intake during the third trimester of pregnancy compared to age-matched non-pregnant mice [1,18,23]. Although pregnant mice overconsume palatable food when provided acute access to high fat diets (Figure 1, Figure 4H), pregnant mice adapt to daily HFD access in a similar manner as non-pregnant animals and continue to consume approximately 25% more calories than non-pregnant mice when provided ad libitum access to regular chow or HFD (Fig. 1B,C).

Despite consuming more food than non-pregnant mice during free access feeding paradigms (Fig. 1B), pregnant mice did not demonstrate increased food consumption during both FR3 and FR5 operant food seeking assays. Therefore, requiring additional work to obtain food reduced caloric intake in pregnant mice to a level which was indistinguishable from non-pregnant animals. These findings are reminiscent of recent work demonstrating a reduced propensity to diet induced obesity in mice when low levels of work are required to obtain palatable high calorie food (i.e. FR1 or FR3 operant food seeking tasks) [24]. Thus, requiring additional work to obtain food reduces hyperphagia in multiple physiological states associated with positive energy balance (i.e. diet induced obesity and pregnancy). Interestingly, pregnant mice consistently consume larger meals, while eating less frequently during FR3 and FR5 operant assays (Fig. 1). This behavioral approach may provide a strategy to maximize energy intake while reducing the amount of time spent foraging for food. Such a behavioral strategy may be particularly useful for pregnant animals, as pregnancy drastically reduces locomotor activity in rodents [23,25], and likely requires adaptations in meal structure to meet the elevated energy demands associated with this period.

Although pregnant mice exhibit increased meal sizes and reduced meal frequency (Fig. 1), these changes are unlikely to be mediated by ventral tegmental area dopamine neurons since pregnant mice continued to consume larger meals in the absence of VTA dopamine activity (Fig. 4C). However, inhibition of VTA dopamine neurons did overall increase meal size, with similar effects observed in both non-pregnant and pregnant mice (Fig. 4C). These findings are consistent with recent work demonstrating a role for VTA dopamine neurons in regulating the size of ongoing meals [26]. Further work is required to determine the neural circuitry and molecular mechanisms mediating altered meal structure in pregnant mice.

Since dopaminergic circuits are important for promoting food seeking and palatable food intake [9], we hypothesized that mesolimbic dopamine signaling may be altered during pregnancy to promote increased feeding behavior. Fiber photometry data presented here is consistent with this hypothesis as the average photometry signal in VTA dopamine neurons is enhanced following the consumption of palatable food in pregnant mice (Fig. 2C,D). Consistently, downstream dopamine levels in the nucleus accumbens are significantly greater in pregnant mice than non-pregnant animals during high fat diet consumption (Fig. 3C,D). Therefore, the overall evoked activity of the mesolimbic dopamine circuitry is increased during palatable food consumption in pregnant mice. Such a response is consistent with increased acute intake of high fat diets in pregnant animals (Figure 1, Figure 4H). Although the data presented here indicate an overall increase in VTA dopamine neuron activity during food consumption in pregnant mice, fiber photometry experiments do not allow for single cell resolution of activity changes in VTA dopamine neurons. Given the heterogenous nature of VTA dopamine neurons [12], further work is required to map the specific VTA dopamine neurons that are most sensitive to pregnancy. Furthermore, as dopamine release exerts differential effects in distinct downstream projection sites [12], further work is also needed to precisely map the dopaminergic terminal regions most altered by pregnancy. However, data presented here demonstrate that elevated dopamine activity during pregnancy likely contributes to increased palatable food intake in pregnant mice since chemogenetic inhibition of VTA dopamine neurons reduces high fat diet intake in pregnant mice, but not in control non-pregnant animals (Fig. 4I,J). It is important to note, however, that the inhibitory effects of VTA dopamine neuron inhibition on HFD intake during pregnancy are subtle, and thus other neural circuits may also regulate palatable food intake during the pregnancy period. The findings presented here are consistent with a recent report demonstrating that palatable food intake cravings during pregnancy are mediated by increased engagement of D2 dopamine receptors in the nucleus accumbens [27]. Further work is ultimately required to determine the neurophysiological mechanisms mediating increased mesolimbic dopamine activity during pregnancy. Since pregnancy robustly increases the levels of key neuroendocrine hormones (i.e. estrogen and progesterone) which regulate dopaminergic activity [2,28], future studies are also warranted to map the effect of pregnancy-related hormones on the activity of the mesolimbic dopamine system.

In addition to palatable food intake, mesolimbic dopamine circuitry promotes increased food seeking in energy deprived animals [9,14]. VTA dopamine neurons express receptors for the hunger hormone ghrelin [20,29,30], which directly activates these cells, and the satiety hormone leptin [29,[31], [32], [33]], which inhibits these cells. Further, acute signals of energy sufficiency such as amylin and glucagon-like-peptide 1 regulate the activity of mesolimbic dopamine circuitry [26,34,35]. Thus, VTA dopamine neurons are well suited to directly respond to signals of energy availability, linking energy state with food seeking behavior and palatable food intake [12,29,[36], [37], [38]]. Consistent with this hypothesis, both food deprivation and artificial stimulation of an energy deprived state via activation of hypothalamic agouti-related peptide (AgRP) neurons increase dopamine levels in the nucleus accumbens during feeding [12,39,40]. Data presented here suggest that pregnancy increases the sensitivity of the mesolimbic dopamine system to energy deprivation. For example, the average VTA dopamine neuron activity is increased in pregnant mice compared to non-pregnant animals in response to food consumption following a 10-hour fast (Fig. 2F–H). Similarly, nucleus accumbens dopamine levels are also elevated in pregnant mice (compared to non-pregnant animals) following an acute fast (Fig. 3F,G). Although chemogenetic inhibition of VTA dopamine neurons does not alter regular chow food intake during ad libitum fed conditions (Fig. 4B), inhibition of these neurons is effective at reducing food intake following an acute 10 h fast (Fig. 4E). This effect occurs in both non-pregnant and pregnant mice, but the magnitude of the effect is larger in pregnant animals (Fig. 4E). In conjunction with the photometry data presented here, these findings suggest that pregnancy shifts the sensitivity of VTA dopamine neurons to signals of negative energy balance. Further work is required to determine the neurophysiological mechanisms mediating altered VTA dopamine responsivity to hunger signals in pregnant mice, which may involve changes in pregnancy related hormones and/or sensitivity to neuroendocrine signals of energy availability.

In conclusion, the average in vivo activity of the mesolimbic dopamine system is enhanced in pregnant mice to increase palatable food intake and the sensitivity of mice to negative energy balance. Given that overfeeding during pregnancy increases the risk of both the mother and her children developing metabolic disorders later in life, our findings suggest that therapies targeting the mesolimbic dopamine system may provide novel pathways to prevent overconsumption during pregnancy.

4. Methods

4.1. Animals

All experiments were approved by the University of Illinois Institutional Animal Care and Use Committee (IACUC). Experiments were performed on female mice (8–16 weeks old). Experiments were performed on C57BL6J (Jax#000664) or DAT-Cre mice (Jax# 020080). DAT-Cre were bred in house by breeding Cre heterozygous mice with C57BL6J mice. Litters were genotyped in house with standard PCR primers for the Cre gene to confirm the transgenic allele: Cre common (5′ GCT TCT TCA ATG CCT TTT GC 3′) and Cre mutant (5′ AGG AAC TGC TTC CTT CAC GA 3’). Prior to experiments mice were group housed in 2–5 mice per cage, in a temperature (20C) and humidity-controlled environment, with a 12 h light/dark cycle. Ad libitum access to food and water was always provided unless specifically mentioned in the text (i.e. during fasting experiments). To generate pregnant and control mice for experiments, mice were bred to a reproductively experienced male mouse for 5 days. Mice were checked daily for a vaginal plug indicating successful mating, which was marked as pregnancy day 1. Control virgin mice were instead paired with a female mouse for 5 days. Following mating, all mice were single caged and body weight and food intake was measured to confirm hyperphagia and weight gain in pregnant mice. In animals subjected to surgical procedures, animal breeding occurred between 2 and 4 weeks following viral injections. All experiments were performed during the third trimester (day 14–20) of pregnancy or the equivalent time-period in non-pregnant control mice. Dopamine fiber photometry and feeding behavior assays were performed on C57/BL6J mice purchased from Jackson Labs that were approximately matched for age. For experiments involving transgenic mice (i.e. DAT-Cre mice), littermates were used as control animals.

4.2. Viral vectors

Adeno-associated viral vectors (AAV) that were used in this study included Cre-dependent GCAMP6s (AAV5-Syn-Flex-GCAMP6s-WPRE-SV40; #100845), Cre-dependent hM4Di (AAV5-hsyn-DIO-hM4Di-mCherry; #44362), Cre-dependent mCherry control virus (AAV5-hsyn-DIO-mCherry; #50459) or Cre-dependent GFP control virus (AAV5-hsyn-DIO-EGFP; #50457), and GRAB dopamine sensor virus (AAV9-hsyn-GRAB-DA2m; #140553). All viruses were purchased from addgene and were injected into the brain at stock concentrations (>1 × 10ˆ12 vg/mL).

4.3. Stereotaxic viral injections and fiber placements

Stereotaxic surgeries were performed as described in our prior studies [41]. Mice were anesthetized with isoflurane and placed in a stereotaxic apparatus (Kopf) with a constant flow of oxygen and isoflurane during surgeries. Mice were administered preoperative carprofen (5 mg/kg, s.c.) for pain prior to injections and for two days post surgeries. To inject virus and/or implant fiber optic inplants, a small incision was made on the skull in the area between bregma and lambda. AAV vectors were injected into the nucleus accumbens or ventral tegmental area using a pulled glass micropipette, which was attached to a micromanipulator (Ronal Tool). Viral injection coordinates for targeting the ventral tegmental area with GCAMP6s (300 nl) or hM4Di/mCherry (200 nl) virus were as follows (from bregma): A/P: 2.8 and −3.1 mm, M/L: +/−0.35 mm, D/V: 4.2 mm (from the surface of the brain). GRAB dopamine sensors were injected into the nucleus accumbens at the following coordinates: A/P:1.2 mm, M/L = −1.0 mm: D/V = −3.5 mm, −3.8 mm and −4.1 mm. Injections of hM4Di/mCherry were bilateral while injections of dopamine GRAB sensors or GCAMP6s was unilateral. For each injection, virus was injected over 5–10 min and left for an additional 5 min before removing the needle.

For fiber photometry experiments, during the same surgery, a fiber optic cannula (200 μm, RWD Biosciences) was implanted directly about the ventral tegmental area or nucleus accumbens viral injection sites. After fiber insertion, the fiber was secured to the skull using dental cement (C&B Metabond). Following surgeries, mice were single caged and returned to housing facilities for at least three weeks before starting experiments.

4.4. Feeding behavioral assays and experiments with FED3 devices

All feeding experiments were performed with feeding experimental devices, except for high fat diet measurements (and initial characterization of regular chow and high fat diet intake in non-pregnant and pregnant mice (i.e. Figure 1A,B)), which were manually measured in the mouse home cage. For manual food intake measurements, a pre-measured amount of food was added to the mouses home cage and the change in the weight of the food was measured the following day. Cages were changed daily during testing to prevent spillage.

FED3 feeding assays were performed as described in our prior study [41]. Briefly, FED3 devices were attached to the side of the mouse's cage, and mice obtained all their food from FED3 devices. Fixed Ratio 1 (FR1) schedule of reinforcement, in which one nose poke on the left poke resulted in dispensing of one 20 mg food pellet, were utilized for all feeding assays, except when FR3 and FR5 assays are specially mentioned in the text and figure legends (i.e. Figure 1). We began collecting experimental data after mice had reached at least 70% correct nose pokes (i.e. 70% of nose pokes occur on the correct left poke vs the incorrect right poke), typically 1–2 days following the start of testing. FR1 was chosen over free access feeding as we observed significantly less food hoarding in FR1 schedule vs free feeding mode during prolonged feeding measurements. For all schedules of reinforcement, feeding patterns were characterized by meal size and meal frequency. We defined a meal as the number of pellets taken with an inter-pellet interval between each pellet less than 5 min, while meal frequency was the number of meals in 24 h. Data was collected every day at ZT6.

4.5. Fiber photometry experiments

Fiber photometry equipment and analysis was performed as described in our prior study [41] Mice were connected to a Plexon Multi-Wavelength Fiber Photometry System (Plexon, 8-61-A-07-A) via a fiber optic patch cord (Plexon, 08-60-A-04-C). Patch cords were attached to the fiber optic implant on the mouse's head via mating sleeves (Plexon). Blue (465 nm) and UV (410 nm) light sources were provided by internal LED drivers which are built into the Plexon fiber photometry system. Fluorescent signals are recorded by the Plexon photometry system which cycles on and off at 30hz sampling window between the 410 nm (isosbestic control signal) and 465 nm (GCAMP6 signal) signals. Analysis of fiber photometry analysis was performed using a custom R code, as described in our prior study [41].

After allowing 2–4 weeks for viral expression and recovery from surgery, all mice were first tested for viral expression by recording the responsivity (for both VTA GCAMP and NAc GRAB sensors) of sensors to acute consumption of a high fat diet pellet. Animals only had access to this diet for this initial 10-minute test. Only animals with noticeable increases in signal in response to HFD were included in subsequent experiments. All mice were subsequently tested in two behavioral assays, which were performed both prior to pregnancy (baseline period) and during the third trimester of pregnancy (or the equivalent time-period in non-pregnant control mice). First, we tested the response of nucleus accumbens dopamine to presentation of a standard chow food pellet in the fed state. Following 10 min of baseline recording, a food pellet was presented to all mice for an additional 5 min. Following this experiment, on the same day, mice were disconnected from the fiber setup and returned to their cages for an additional 10 h. Food was removed from all cages during this 10-h period to test the response to chow presentation in energy deprived mice (food removed from 7am to 5pm). The identical experiment was subsequently performed on the same mice following a 10 h fast. 2–4 days later, the same mice were tested for their response to consumption of palatable high fat diet in the fed state. For these experiments, all mice were habituated to the high fat diet by providing access to HFD in the home cage for 10 min for two days prior to the photometry experiment. On the experimental day (experiment performed during the light period: between 10am and 4pm), mice were again attached to the fiber photometry system and baseline signal was recorded for 10 min. Following baseline recordings, mice were provided with high fat diet for an additional 10 min. For both fast-refeeding and HFD photometry assays a plexon event input generator was used to manually score the start of eating, and these events were directly aligned with fiber photometry calcium traces for analysis. Following this initial testing, all mice were approximately divided into “pregnant” and “non-pregnant” groups. The pregnant group was bred to reproductively experienced male mice following baseline experiments, while the non-pregnant group was paired with a female mouse. The same two experiments (10 h fast and high fat diet presentation) were again repeated on the same mice during either the third trimester of pregnancy or the equivalent time-period in the non-pregnant control group. Changes in calcium response to events (high fat diet consumption or regular chow consumption) were calculated in the 60- or 180-seconds following events (i.e. start of food consumption) compared to the 60 s prior to each event (i.e. prior to food consumption). The change in calcium signal was calculated for each mouse during the two time points in to determine the effect of pregnancy on nucleus accumbens dopamine responses (within subjects' comparison). We also directly compared the response between pregnant and non-pregnant mice on the same test day (i.e. between subjects' comparison). Although we observed stable recordings of dopamine sensor recordings at each time-point in the control non-pregnant mice (i.e. comparing time point 1 to time point 2), mice expressing GCAMP6 in VTA dopamine neurons exhibited signal decay between the two time-points. Therefore, for GCAMP6 experiments we only compared the change in calcium signal in non-pregnant vs pregnant mice during the same testing day (i.e. between subjects’ comparisons).

4.6. Chemogenetic experiments

For regular chow chemogenetic feeding experiments with the FED3 devices (Fig. 4), mice were administered saline (i.p., 200 μl) or CNO (1 mg/kg, i.p.) in a randomized fashion during the light period. Changes in pellets consumed, meal size, and meal number were calculated for each mouse following saline or CNO injections and compared for statistical analysis (repeated measures comparison). For HFD chemogenetic assays, all mice received HFD for 10 min in the two days prior to testing to habituate to HFD presentation. On the testing day, all mice (both mice expressing control mCherry or hM4Di in VTA dopamine neurons) were administered CNO (1 mg/kg) 15 min prior to providing high fat diet to the mice. Consumption of high fat diet was measured 30 min and 1 h following i.p. injections and compared between the mCherry and hM4Di expressing mice. For fasting experiments, after mice were trained on FR1 assays with FED3 devices, all mice were fasted for 10 h during the light period (7am-5pm). Following fasting, all mice were administered CNO (1 mg/kg) 10 min prior to providing FED3 devices to the mice.

Experiments involved DREADD targeting of VTA dopamine neurons were peformed on two cohorts of mice. High fat diet feeding experiments were performed on both separate cohorts of mice, and combined data from both cohorts are plotted in Figure 4. Fasting experiments were only performed on one cohort of animals. Representative viral locations from all mice for both cohorts are shown in extended Figure 3.

4.7. Post-hoc validation of viral and fiber optic placement

Following the completion of behavioral experiments on virally targeted and fiber implanted mice, all mice underwent trans-cardiac perfused to check for the location of virus and fiber optic placement. Perfusions and post-hoc fixation/cryopreservation were performed as described in our prior study [41]. Brain sections covering the VTA or NAc were obtained in 40 μm sections using a cryostat (Leica CM3500) and mounted onto glass slides to check for viral location and fiber optic placement. Only mice with correct targeting of virus into the VTA/NAc and fiber optic placement in VTA/NAc were included in experiments.

4.8. Statistical analysis

Specific statistical tests are outlined in the figure legends. Data that was normally distributed was analyzed with parametric statistical tests, while data that was not normally distributed was analyzed with non-parametric tests. Data was analyzed using Graphpad Prism.

CRediT authorship contribution statement

Tanya Pattnaik: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis, Conceptualization. Benjamin Wang: Writing – review & editing, Methodology, Investigation, Formal analysis. Patrick Sweeney: Writing – review & editing, Writing – original draft, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization.

Funding sources

This work was funded by the University of Illinois and the National Institute of Health (R01HD113522 to PS).

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Patrick Sweeney reports financial support was provided by University of Illinois Urbana–Champaign. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Extended Data Fig 1.

Pregnant mice consume larger meals (A) Cumulative distribution of meals across different meal sizes in pregnant and non-pregnant mice in FR3. (B) Cumulative distribution of meals across different meal sizes in pregnant and non-pregnant mice in FR5. Area under the curve of the distributions compared using t-test. ∗p < 0.05.

Extended Data Fig. 2.

Pregnant mice have an increased average VTA DA neuron response to food stimuli (A) Average z-score post 5s of HFD drop in non-pregnant and pregnant mice. (B) Average photometry trace showing VTA calcium response to HFD drop in adlib mice. Heat map showing average VTA calcium signal during consumption of HFD in (C) non-pregnant and (D) pregnant mice. (E) Average z-score during 20s of initial HFD consumption in non-pregnant and pregnant mice. (F) Time spent eating the high-fat diet by the non-pregnant and pregnant mice in the 10-minute fiber photometry recording session. (G) Average z-score post 5s of RC drop after 10h of fast in non-pregnant and pregnant mice. (H) Average photometry trace showing VTA calcium response to RC drop in fasted mice. Heat map showing VTA calcium signal during consumption of regular chow (RC) after 10h fast in (I) non-pregnant and (J) pregnant mice. (K) Average z-score during 20s of initial RC consumption in fasted non-pregnant and pregnant mice. (L) Amount of regular chow eaten by non-pregnant and pregnant mice during the first 10 min of refeeding. Data points represent individual mice. Data represented as mean ± SEM. Data analyzed using unpaired student's t test.

Extended Data Fig. 3.

Representative viral locations and fiber optic placements for nucleus accumbens GRAB dopamine photometry experiments. (A) Schematics showing the location of the viral GRAB sensor and fiber location in the pregnant and non-pregnant mice for GRAB fiber photometry experiments. Spread of the virus is shown in green, while location of fiber optic is shown as grey line. (B) Schematics showing the spread of viral expression of the inhibitory DREADD in the pregnant and non-pregnant mice for chemogenetic studies. The spread of the virus in the VTA is shown in red. (C) Microscope image showing DIO-mCherry virus (in red) and TH antibody (green). Colocalization marked with white arrows.

Extended Data Fig. 4.

Nucleus accumbens dopamine signal is enhanced in pregnant mice during feeding. (A) Average z-score 5 s after HFD presentation to adlib fed pregnant and pre-pregnant mice. (B) Average GRAB photometry trace during HFD presentation to adlib pregnant and pre-pregnant mice. Heat map showing dopamine release in response to HFD consumption in adlib mice (C) before they have been mated and (D) during pregnancy. (E) Average z-score in the 20s after initial consumption of HFD by pregnant and pre-pregnant mice. (F) Comparison of average z-score during consumption of HFD in pregnant and non-pregnant mice in the same session. Heat map showing dopamine release in response to regular chow (RC) consumption in fasted mice (G) before they have been mated and (H) during pregnancy. (I) Average z-score in the 20s after initial consumption of RC by fasted pregnant and pre-pregnant mice. (J) Comparison of the average z-score while refeeding in pregnant and non-pregnant mice in the same session. (K) Average z-score 5 s after RC presentation to 10h fasted pregnant and pre-pregnant mice. (L) Average GRAB photometry trace during RC presentation to 10h fasted pregnant and pre-pregnant mice. Data points represented individual mice. Data represented as mean ± SEM. Data in F and J analyzed using unpaired student's t test. Data in A, E, I, K analyzed using paired t-test.

Extended Data Fig. 5.

CNO administration does not alter feeding behavior in control mice expressing eGFP in VTA dopamine neurons. (Left) Number of pellets taken in 4 h following administration of either saline or CNO in control eGFP injected non-pregnant and pregnant mice. (Middle) Average meal size in the 4 h following administration of either saline or CNO in the eGFP injected non-pregnant and pregnant mice. (Right) Number of meals in the 4 h following administration of either saline or CNO in the eGFP injected non-pregnant and pregnant mice. Data represented as mean ± SEM. Data points represent individual mice. Data analyzed by 2-way ANOVA. ∗p < 0.05, ∗∗p < 0.01, ns (not significant).

Data availability

Data will be made available on request.