Hypothalamic AgRP neurons regulate the hyperphagia of lactation

Abstract

Objective

The lactational period is associated with profound hyperphagia to accommodate the energy demands of nursing. These changes are important for the long-term metabolic health of the mother and children as altered feeding during lactation increases the risk of mothers and offspring developing metabolic disorders later in life. However, the specific behavioral mechanisms and neural circuitry mediating the hyperphagia of lactation are incompletely understood.

Methods

Here, we utilized home cage feeding devices to characterize the dynamics of feeding behavior in lactating mice. A combination of pharmacological and behavioral assays were utilized to determine how lactation alters meal structure, circadian aspects of feeding, hedonic feeding, and sensitivity to hunger and satiety signals in lactating mice. Finally, we utilized chemogenetic, immunohistochemical, and in vivo imaging approaches to characterize the role of hypothalamic agouti-related peptide (AgRP) neurons in lactational-hyperphagia.

Results

The lactational period is associated with increased meal size, altered circadian patterns of feeding, reduced sensitivity to gut-brain satiety signals, and enhanced sensitivity to negative energy balance. Hypothalamic AgRP neurons display increased sensitivity to negative energy balance and altered in vivo activity during the lactational state. Further, using in vivo imaging approaches we demonstrate that AgRP neurons are directly activated by lactation. Chemogenetic inhibition of AgRP neurons acutely reduces feeding in lactating mice, demonstrating an important role for these neurons in lactational-hyperphagia.

Conclusions

Together, these results show that lactation collectively alters multiple components of feeding behavior and position AgRP neurons as an important cellular substrate mediating the hyperphagia of lactation.

Highlights

• •Lactation increases meal size and alters circadian feeding patterns in mice.
• •Lactating mice exhibit increased sensitivity to hunger signals.
• •Lactating mice have an attenuated response to satiety signals.
• •AgRP neurons are activated by nursing in lactating mice.
• •AgRP neurons are required for lactational hyperphagia.

1. Introduction

The increased energy demands associated with milk production and caring for young present an enormous metabolic challenge to mammals [1]. In rodents, this challenge is met by drastically increasing food intake, as both mice and rats increase their food intake by 200–400 percent during lactation [1]. Although lactational-hyperphagia is particularly drastic in rodents, increased feeding is also observed in lactating humans [1]. However, the specific cell types and neural circuitry mediating lactational-hyperphagia are largely unknown.

Importantly, an appropriate feeding response to lactation is critical for long-term metabolic health, since overfeeding during lactation increases the risk of mothers developing metabolic disorders later in life in both rodents and humans [[2], [3], [4], [5], [6]]. Further, the long-term health of the developing pups is dependent on appropriate maternal feeding behavior during pregnancy and lactation [2,3,6] as neural circuitry regulating feeding, metabolism, motivation, and reward continue to develop throughout the lactation period in neonatal mice, a neurodevelopmental period analogous to the third trimester of pregnancy in humans. Consequently, both overnutrition and undernutrition during lactation in rodents or the equivalent developmental period in humans (i.e. third trimester of pregnancy) increases the subsequent risk of developing metabolic and neuropsychiatric disorders later in life [2,3,6]. Thus, understanding the physiological mechanism(s) for increased feeding during lactation is important to the long-term health of both the mother and her children.

Feeding behavior can be dissected into multiple distinct behavioral components, including food seeking, food consumption, and meal termination [7]. Further, feeding is controlled by both homeostatic need (i.e. hunger) and non-homeostatic mechanisms including eating for reward (hedonic feeding) [8], and circadian rhythms which couple feeding behavior to daily energy demands [9]. Although the behavioral mechanisms and neural circuitry controlling discrete components of feeding are well understood, it is unclear which of these components promote the hyperphagia of lactation.

Among the cell types controlling feeding behavior, the hypothalamic agouti-related peptide (AgRP) neurons are well characterized as a critical cell type mediating behavioral, autonomic, and neuroendocrine responses to energy deprivation [[10], [11], [12], [13], [14], [15], [16]]. AgRP neurons directly sense stored energy in the form of leptin from fat [[17], [18], [19]], gut-derived satiety hormones [20,21], and glucose levels in the blood [16,22], and are activated in response to energy deprivation [17,23,24]. Importantly, in addition to sensing the current energy state, AgRP neurons also predict future energy needs to adaptively control feeding behavior according to both current and future energy needs. For example, AgRP neurons are rapidly inhibited in hungry mice by the sight or smell of food [12,21,25,26], and cues predicting impending food availability [25,26]. Conversely, in time-restricted feeding assays the in vivo dynamics of AgRP neurons reflect the periods of expected food delivery, with AgRP neuron activity peaking immediately prior to the time of expected food availability [11]. Ultimately, activation of AgRP neurons leads to the release of the melanocortin receptor antagonist AgRP, GABA, and neuropeptide Y (NPY), which act at downstream brain regions to stimulate feeding behavior [16,27,28]. Consistently, chronic activation of AgRP neurons results in obesity in mice [29,30], while overexpression of agouti-signaling protein (which acts similarly to AgRP) also results in robust hyperphagia and obesity in humans [31].

Although AgRP neurons are well-established cellular regulators of energy homeostasis, the specific role of these cells in lactational hyperphagia is largely unknown. Expression of AgRP mRNA is elevated in multiple conditions associated with increased feeding such as cold exposure [32,33], and food deprivation [16,34], and lactation in rats [35,36]. Although AgRP neurons are activated by low levels of leptin (as observed in lactating rodents) [17], increasing leptin levels to those observed in non-lactating animals only partially reduces the hyperphagia of lactation [[37], [38], [39], [40], [41]], suggesting that altered AgRP activity may not be the sole driver of lactational hyperphagia. In contrast, diphtheria toxin-mediated ablation of AgRP neurons in adult mice results in starvation in both non-lactating and lactating animals [42], suggesting that the hyperphagia of lactation may require functional AgRP circuitry. However, according to recent reports, the starvation phenotype associated with AgRP neuron ablation in adult mice may result from non-specific effects associated with diphtheria toxin administration [43], as other forms of chronic inhibition or ablation of AgRP neurons have no effect on feeding or body weight in ad libitum-fed mice [30,43]. Therefore, it remains unclear if AgRP neurons are required for the hyperphagia of lactation, and it is unknown how the lactational state alters the activity of these cells.

In this study, we first utilized home cage feeding devices [44] to characterize the feeding microstructure and circadian nature of feeding behavior during mid-late lactation in mice (lactation days 7–17). Using a combination of pharmacological and behavioral assays we demonstrate that lactational-hyperphagia is associated with increased meal size, enhanced sensitivity to negative energy balance, reduced sensitivity to anorexic stimuli, and increased hedonic feeding. Finally, using immunostaining, in vivo imaging, and chemogenetics we characterize the role of hypothalamic AgRP neurons in lactational hyperphagia in mice.

2. Results

2.1. Lactating mice consume larger meals and exhibit an altered circadian feeding structure

Lactating rodents drastically increase their food intake to accommodate the metabolically demanding process of nursing [1]. Therefore, we first sought to fully describe the feeding microstructure in lactating and non-lactating mice by utilizing home cage operant feeding devices (feeding experimental device 3, FED3) [44]. These devices attach directly to the mouse home cage and dispense 20 mg food pellets after each successful pellet retrieval, which is recorded by the FED3 system (Figure 1A). Mice must obtain all their food from FED3 devices, allowing for the precise quantification of meal size, frequency, and circadian aspects of feeding behavior in a home cage setting. As previously described, lactating mice consume significantly more food than non-lactating animals (Figure 1B). The increased feeding observed in lactating mice is driven by increased meal size, which is more than double in lactating mice relative to non-lactating control animals (Figure 1C). Surprisingly, lactating mice consume fewer meals than non-lactating animals, an effect that may result from the increased time allocated to nursing and maternal behaviors (Figure 1D). The increase in meal size in lactating mice is accompanied by a drastic shift in the distribution of meal sizes between lactating and non-lactating mice. While non-lactating mice exhibit more frequent “grazing” than lactating mice (i.e. consumption of only a few pellets/meal), most meals in lactating mice are greater than 5 pellets (Figure 1E), indicating a clear shift in feeding structure in lactating mice.

Figure 1.

Lactating mice consume larger meals and have an altered circadian feeding structure. A: Image of the experimental setup for measuring home-cage feeding structure in lactating mice. B–D: Food intake (B), meal size (C), and total number of meals (D) in non-lactating and lactating mice in 24 h. E: Frequency distribution of the different meal sizes of lactating and non-lactating mice. F: Food intake during the light period and dark period. G: Hourly food intake over a 24 h period in non-lactating and lactating mice. Data are shown as mean ± s.e.m. Data in E, F, and G was analyzed with 2-way ANOVA with Sidak's multiple comparisons test. Data in B, C, and D analyzed with Students unpaired t-test. n = 10 lactating mice and 14 non-lactating mice for all panels. ∗∗p < 0.01, ∗∗∗∗p < 0.001.

Next, we characterized the circadian nature of feeding in non-lactating and lactating mice (Figure 1F). Both non-lactating and lactating mice consume most of their food during the dark cycle, and the increased feeding observed in lactating mice is most associated with the dark period (Figure 1F). Interestingly, lactating mice exhibit a drastic increase in feeding specifically at the start and the end of the dark cycle, indicating that circadian cues likely contribute the hyperphagia of lactation (Figure 1G).

2.2. Lactational hyperphagia persists acutely in the absence of the suckling stimulus

Much of the hyperphagia associated with lactation is thought to be mediated by the enhanced energetic demand associated with nursing and milk production [1], the suckling stimulus [45,46], and the metabolic requirements associated with caring for young (i.e. nest building, heat production). To dissect the effects of these factors on feeding during lactation we measured food intake in lactating mice in the presence and absence of pups in a new cohort of mice. In this experiment, food intake was measured in lactating and non-lactating mice during the same time of day (4 pm–10 pm; ZT10-ZT16) in the presence of pups on day one, and in the absence of pups on day 2 (Figure 2A). We reasoned that if lactational-hyperphagia was primarily the result of the acute energetic demand associated with nursing (i.e. milk production and the suckling stimulus), then this hyperphagia should be attenuated by the absence of pups. In contrast, following pup removal, food intake was similar between lactating mice in the presence of pups and lactating mice without pups present (Figure 2B). However, in the absence of pups, lactating mice consume the same number of meals as non-lactating animals, and significantly more meals than lactating mice with pups present in the cage (Figure 2C). Meal size remained similar between actively nursing mice and lactating mice in the absence of pups, with both groups of lactating mice consuming larger meals than non-lactating animals (Figure 2D). Thus, the reduced number of meals observed in nursing mice (Figure 1) is likely mediated by the enhanced time spent nursing/caring for young, since no difference in meal number is observed between lactating and non-lactating mice in the absence of pups (Figure 2C). To confirm that dams were still capable of lactating following acute pup removal, we re-introduced the pups to the dam and measured the body weight of the dam 6 h following pup reintroduction (ZT4-ZT10). We observed decreased body weight of the dam following 6-hours of reintroduction to the pups (Extended Data Fig. 1), suggesting that dams are appropriately feeding the pups after acute separation from the pups.

Figure 2.

Lactational hyperphagia acutely persists in the absence of pups. A: Schematic of experimental setup for experiments shown in B-D. B–D: 6-hour food intake (B), number of meals (C), and meal size (D) in the three groups shown in a (n = 10 non lactating mice, n = 7 lactating mice with pups, and n = 8 lactating mice without pups present). Acute removal of the pups increased the number of meals (C) in lactating mice to the levels observed in non-lactating animals. E: Schematic of experimental approach for data in F, G, and H. F: 24-hour food intake in the three groups shown in e prior to pup removal. G: 24-hour food intake in lactating mice (with pups present), lactating mice one day after pup removal, and non-lactating control animals. 24 h post pup removal significantly reduced feeding in lactating animals, although these mice still exhibited an approximately 50% increase in feeding relative to non-lactating mice. H: 24-hour food intake in the period between 24 and 48 h following acute pup removal from lactating mice (n = 6 non-lactating mice, n = 7 lactating mice with pups present, and n = 7 lactating mice with the pups removed for panels F–H). Food intake in lactating mice returned to the levels of non-lactating mice following 48 h of pup removal. Data points represents individual mice and is shown and mean +/− s.e.m. Data analyzed with one-way ANOVA followed by Sidak's multiple comparison's test. Data in panels A–D and E-H performed in 2 different cohorts of mice. ns (not significant), ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.005, ∗∗∗∗p < 0.001.

Since we observed sustained hyperphagia following pup removal in lactating mice (Figure 2A–D), we next sought to determine the duration of the hyperphagia following removal of the pups. In a new cohort of mice, we weaned the pups from the dam and mapped the feeding response in the days following pup removal. As controls, we also mapped the food intake of lactating mice (with pups present) and non-lactating mice on the equivalent days (Figure 2E,F). On the first day following the removal of the pups, food intake of the dams decreased by approximately 50 percent, but remained about 2× higher than observed in non-lactating mice (Figure 2G). However, between 24 and 48 h after the removal of the pup's, daily food intake returned to the levels observed in non-lactating mice (Figure 2H). Together, these findings (Figure 2E–H) indicate that the hyperphagia of lactation acutely persists in the absence of the suckling stimulus and nursing but returns to the levels of non-lactating mice 48 h following the removal of the pups from the dam.

Figure 5.

Lactation alters the in vivo activity of AgRP neurons. A: Schematic outlining fiber photometry recordings of hypothalamic AgRP neurons in awake behaving mice. B: Representative image of GCAMP6s expression in AgRP neurons (green), with the fiber optic trace shown in white. C: Experimental approach for the fast-refeeding (top panels) and palatable food intake (bottom panels) experiments. D and E: Traces of fluorescent signal in the 60 s prior and following the presentation of food before lactation (D) and during lactation in the same mice (E) mice (n = 8 mice for both D and E). F: Change in fluorescence in the 60 s following food introduction before lactation and during lactation. Lactating mice had a significantly greater decreases in calcium activity following food presentation than prior to lactation (n = 8 mice for pre-lactation and lactation groups). G: Change in fluorescence in the 60 s following eating pre-lactation and during lactation in the same mice. Lactating mice had a significantly greater decrease in calcium activity following food consumption than observed prior to lactation (n = 8 mice). H: Average trace of the calcium signal in non-lactating (blue trace) and lactating mice (red trace) aligned to the start of food consumption (time 0 on x-axis). I: Quantification of the change in calcium signal in the 60 s following food consumption in lactating and non-lactating mice. AgRP neurons in lactating mice exhibit a sustained reduction in fluorescence that is not observed in non-lactating animals (n = 8 mice for lactating group and n = 7 mice for non-lactating group). Data represents mean +/− s.e.m. Data points represent individual mice. Data in F, G, and I analyzed with Mann Whitney test. ∗p < 0.05.

2.3. Lactating mice are hypersensitive to negative energy balance and exhibit enhanced hedonic feeding

Since lactating mice consume more food than non-lactating animals (Figure 1, Figure 2), we next sought to determine the behavioral mechanism(s) mediating increased feeding in lactating mice. We first hypothesized that lactating mice may be more sensitive to signals of negative energy balance, facilitating increased feeding behavior. To test this hypothesis, we performed a series of acute food restriction and re-feeding experiments in a new cohort of lactating and non-lactating mice (Figure 3A,B). Following 90 min of food restriction in the light period (ZT4.5-ZT6), non-lactating mice did not significantly increase their food intake relative to ad libitum fed levels (Figure 3A). In contrast, lactating mice increased their food intake by approximately 2.5-fold relative to ad libitum fed levels, indicating an enhanced orexigenic response to acute food restriction (Figure 3A). To test if this effect is graded by the magnitude of food deprivation, and to establish the kinetics of re-feeding following an acute fast, we fasted non-lactating and lactating mice for 10 h during the light cycle (ZT1-ZT11) and measured acute feeding behavior in the 30 min following re-feeding (Figure 3B). Following re-feeding, non-lactating mice consumed approximately 10 pellets (200 mg), and continued food consumption plateaus after approximately 10 min of re-feeding (Figure 3B). However, lactating mice continued to eat for up to 20 min following re-feeding, with pellet consumption plateauing at approximately 40 pellets (800 mg, Figure 3B). To confirm that lactating mice are hypersensitive to energy deprivation, in a new cohort of mice we administered the gut-derived hunger signal ghrelin (1 mg/kg, i.p., 200ul) or saline (i.p, 200ul) to non-lactating and lactating mice in the ad libitum fed state (ZT6) and measured food intake in the 30 min following saline or ghrelin administration (Figure 3C). Consistent with prior studies, ghrelin transiently increased feeding in non-lactating mice (Figure 3C). However, this effect was enhanced in lactating mice, supporting the hypothesis that lactating mice are hypersensitive to signals of negative energy balance (Figure 3C). Together, these results indicate that the increased feeding observed in lactating mice is accompanied by enhanced sensitivity to signals of energy deprivation and an attenuated/delayed satiety response following re-feeding.

Figure 3.

Lactation alters multiple components of feeding behavior. A: 2-hour food consumption in non-lactating and lactating mice in the ad libitum fed state and following 90 min of food deprivation. Lactating mice increase their food intake following 90 min of food restriction, an effect that is absent in non-lactating animals (n = 5 mice per group). B: Food intake in non-lactating and lactating mice following 10 h of food deprivation. Lactating mice consume significantly more pellets in the 25 min period immediately following re-feeding than non-lactating mice (n = 6 lactating mice and n = 7 non-lactating mice). C: Food intake (30 min food intake) following injections of saline or ghrelin to non-lactating and lactating mice (n = 10 non-lactating mice and n = 9 lactating mice). D and E: High fat diet (D) and peanut butter chip (E) intake in non-lactating and lactating ad libitum fed mice (15 min food intake). Lactating mice consume more HFD and more PB chip than non-lactating mice in the 15 min following access to HFD or PB chips (n = 8 non-lactating mice and n = 5 lactating mice for panels D and E). F: Food intake following injections of saline or liraglutide to non-lactating and lactating mice. Liraglutide significantly reduces feeding in non-lactating mice but not lactating animals (4 h food intake, n = 15 non-lactating mice and n = 9 lactating mice). G: Food intake following injections of saline or PYY to non-lactating and lactating mice. PYY reduced feeding in non-lactating mice but did not alter food intake in lactating animals (2 h food intake, n = 10 non-lactating mice and n = 10 lactating mice). All panels represent different cohorts of mice except for panels d and e, which represent the same cohort of mice. Data represents mean +/− s.e.m. Data in A, B, C, F, and G analyzed with 2-way ANOVA and Sidaks post-hoc test. Data in d and e analyzed with Student's unpaired t-tests. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.005, ∗∗∗∗p < 0.001. ns (not significant).

In addition to homeostatic state (i.e. hunger), feeding can also be driven by the hedonic desire to consume palatable food [8]. Therefore, we next tested if lactating mice have enhanced hedonic feeding by providing acute access to palatable food to ad libitum fed non-lactating and lactating mice (ZT6). In a new cohort of non-lactating and lactating mice we provided acute access to either a high-fat diet (60% kcal from fat) or peanut butter chips (Reese's peanut butter chips) to ad libitum-fed lactating and non-lactating mice (Figure 3D,E). Although non-lactating mice readily consumed high-fat diet during testing, this effect was significantly enhanced in lactating mice (Figure 3D). Similar results were also obtained following presentation of peanut butter chips (ZT6, Figure 3E). Thus, lactation is associated with increased sensitivity to negative energy balance and an enhanced propensity to consume palatable food.

2.4. Lactation is associated with an attenuated response to gut-brain satiety signals in mice

Our prior results indicate that the hyperphagia of lactation is primarily driven by increased meal size (Figure 1). Meal size is regulated by meal-derived satiety signals (i.e. CCK, PYY, and GLP1), which act together with GI distention to terminate ongoing meals [20,21,47,47,48]. Therefore, in a new cohort of lactating and non-lactating mice, we next tested if lactating mice have a normal anorexic response to the gut-derived satiety signals peptide YY (PYY) and glucagon-like peptide 1 (experiments performed from ZT12-ZT14; GLP1, GLP1R agonist liraglutide). As expected, liraglutide administration potently suppressed feeding in non-lactating mice (Figure 3F). In contrast, liraglutide did not significantly affect food intake in lactating dams (Figure 3F), indicating that lactating mice have an impaired anorexic response to GLP1R stimulation. In addition to GLP1, intestinal L cells release peptide YY (PYY) in response to food detection to suppress ongoing meal consumption [49]. Consistent with GLP1 administration, PYY injections reduced food intake in non-lactating mice and this effect was absent in lactating dams (ZT12-ZT14; Figure 3G).

We next sought to determine if the attenuated response to anorexic stimuli in lactating dams is specific to gut-derived satiety peptides, or if lactating mice also exhibit an impaired response to other forms of anorexic stimuli. To address this question, we administrated the pancreas-derived satiety peptide amylin [50] or the anorexic cytokine growth differentiation factor 15 (GDF15) [[51], [52], [53]] to non-lactating and lactating mice (ZT12-ZT14). In contrast to PYY and GLP1, both lactating and non-lactating mice exhibited a similar anorexic response to amylin and GDF15 (ZT12-ZT14, Extended Data Fig. 2). Therefore, lactation is associated with an attenuated response to PYY and GLP1R stimulation, but an intact response to other forms of anorexic stimuli.

2.5. Lactating mice exhibit enhanced AgRP neuronal activity

We next sought to identify a cellular mechanism mediating the hyperphagia of lactation. Among the cell types controlling feeding behavior, hypothalamic neurons in the arcuate nucleus expressing the gene agouti-related protein (AgRP) have been extensively studied in the context of homeostatic feeding behavior [[11], [12], [13], [14], [15],20,27,28,43]. These neurons are activated in response to energy deprivation to increase feeding and reduce energy expenditure [17,23,24]. Further, prior work indicates that mRNA expression of AgRP is increased in lactating rats [35,36]. Therefore, we first utilized immunohistochemical detection of the immediate early gene cfos in transgenic mice with tdtomato expression localized to AgRP neurons (AgRP-Cre x tdtomato reporter mice) to quantify the activity of AgRP neurons in lactating and non-lactating mice (Figure 4). In ad libitum fed mice no difference was detected in cfos positive AgRP neurons between lactating and non-lactating mice (ZT6, Figure 4A). Further, total cfos positive cells in the arcuate nucleus did not differ between non-lactating and lactating mice (ZT6, Extended Data Fig. 3).

Figure 4.

AgRP neurons exhibit enhanced sensitivity to negative energy balance in lactating mice. A: Representative image showing expression of cfos (blue) and AgRP neurons (AgRP-cre x tdtomato reporter line, red) in non-lactating (left panel) or lactating mice (middle panel) in ad libitum fed mice. No significant difference was observed between the percentage of AgRP neurons expressing cfos in non-lactating and lactating mice in the ad libitum fed state (right panel, n = 4 non-lactating mice and n = 3 lactating mice). B: Representative image showing expression of cfos (blue) and AgRP neurons (AgRP-cre x tdtomato reporter line, red) in non-lactating (left panel) or lactating mice (middle panel) following a 90 min fast. Lactating mice show an increased percentage of AgRP neurons expressing cfos compared to non-lactating animals following a 90 min fast (n = 7 non-lactating mice and n = 6 lactating mice). C: Representative image showing expressing of cfos (blue) and AgRP neurons (AgRP-cre x tdtomato reporter line, red) in non-lactating (left panel) and lactating mice (middle panel). An increased percentage of AgRP neurons express cfos following a 10 h fast in lactating mice relative to non-lactating animals (n = 6 non-lactating mice and n = 4 lactating mice). Data points represent the average percentage of AgRP neurons expressing cfos for individual mice. Scale bars (50 um). Data analyzed with Student's unpaired t-test. ns (not significant), ∗p < 0.05, ∗∗p < 0.01.

Since AgRP neurons are inhibited by food consumption [20,21,54], we reasoned that the increased food intake in lactating mice may inhibit AgRP neurons, obscuring the ability of cfos to detect increased AgRP neuronal dynamics in the lactational state. Further, given that we observed increased sensitivity to negative energy balance in lactating mice (Figure 3A,B), we hypothesized that AgRP neurons may be more sensitive to signals of negative energy balance during lactation. To test this hypothesis, we quantified the percentage of AgRP neurons that express cfos in non-lactating and lactating mice following a 90-minute fast during the light cycle (ZT4.5-ZT6), a period where lactating mice exhibit enhanced feeding that is not observed in non-lactating animals (Figure 3). In contrast to ad libitum-fed conditions, AgRP neurons exhibited increased cfos levels in lactating mice relative to non-lactating animals (Figure 4B). No difference in total cfos-positive cells in the arcuate nucleus were detected between lactating and non-lactating mice following 90 min of food deprivation (Extended Data Fig. 3). A similar increase in cfos activity was also observed in AgRP neurons following a 10 h fast (fasted from ZT1-ZT11, Figure 4C). In this case we also observed an increase in the total number of cfos positive cells in the arcuate nucleus (Extended Data Fig. 3). Together, these results suggest that AgRP neurons exhibit increased sensitivity to negative energy balance during the lactational state.

2.6. Lactation alters the in vivo activity of AgRP neurons

In addition to being activated by negative energy balance (Figure 4), in vivo imaging experiments demonstrate that AgRP neurons are rapidly inhibited by the introduction of food in food deprived mice [25,26]. This inhibitory response is dependent on the subsequent consumption of calories as AgRP neuron activity returns to baseline levels if food is not subsequently consumed [20,21,25,54]. Thus, we next investigated if these in vivo properties of AgRP neurons are altered in the context of lactation. To monitor dynamic changes in the activity of AgRP neurons in awake-behaving mice we utilized in vivo fiber photometry and the genetically encoded calcium indicator GCAMP6s to record changes in calcium in AgRP neurons in non-lactating and lactating mice (Figure 5A,B). The arcuate nucleus of AgRP-Cre mice was injected with AAV viral vectors expressing a Cre-recombinase-dependent version of the genetically encoded calcium indicator GCAMP6s to selectively express GCAMP6s in AgRP neurons (Figure 5A,B). Three weeks following AAV injections, we performed a second stereotaxic surgery to implant a fiber optic cannula (200um, RWD Biosciences) into the arcuate nucleus (Figure 5A,B). One week following fiber optic implantation we performed fiber photometry experiments to confirm successful targeting of GCAMP6s and placement of fiber optic cannulas. Since prior work has established that AgRP neurons are inhibited upon the sight of food following an overnight fast [25,26], we screened for successful targeting by fasting all fiber-implanted mice overnight and recorded the calcium response following food presentation (Extended Data Fig. 4). As expected, an inhibitory response to food presentation was observed in approximately 60% of implanted mice, which were used for subsequent experiments.

To normalize the relative level of GCAMP6s targeting for future comparisons between lactating and non-lactating groups, prior to mating fiber implanted mice to male mice, we separated all mice into two groups (mice to be mated with males to study AgRP neuron activity during lactation or mice paired with female mice to characterize changes in AgRP neuron activity over time, independent of lactation). Mice from the future lactation group (mated with males) and non-lactating group (paired with females) exhibited the same average change in fluorescence activity upon food introduction prior to mating (or pairing with a female mice) (Extended Data Fig. 4). Mice in the lactating group were subsequently paired with male mice for five days to induce pregnancy and lactation, while control mice were paired with a second female mouse for five days. On the day following mating, we performed additional fiber photometry experiments to determine the baseline inhibitory response to the presentation of food and food consumption following an acute 10-h fast during the light period (Figure 5C). For this experiment, we chose 10 h of food deprivation since we previously observed increased hyperphagia in lactating mice (vs non-lactating animals) following a 10-h daytime fast (ZT1-11, Figure 3), and we reasoned that an overnight or 24-hour fast may maximize AgRP neuron activity in both lactating and non-lactating mice, thus obscuring any potential differences in AgRP neuron activity between these groups. This experiment was repeated in lactating mice between lactation day 7–10, or the equivalent time-point in non-lactating mice, to determine if the inhibitory response to the presentation of food and food consumption is altered in lactating mice (Figure 5C). The inhibitory response of AgRP neurons in the 60 s following food presentation was enhanced during lactation compared to the pre-lactation baseline period (Figure 5D-F, Extended Data Fig. 5a), and compared to non-lactating control mice (Extended Data Fig. 5c). Consistent with this data, lactating mice initiated eating significantly faster following food presentation than non-lactating animals (Extended Data Fig. 6). Next, we aligned the AgRP neuron activity to food consumption to determine if the inhibitory response to food consumption differed between non-lactating and lactating mice (Figure 5G, Extended Data Figure 5b, Extended Data Figure 6a and b). AgRP neurons from lactating mice exhibited an enhanced inhibitory response in the 60 s following the initiation of food consumption, compared to baseline pre-lactating levels (Figure 5G, Extended Data Fig. 5b), and compared to control non-lactating mice (Extended Data Fig. 5D). Importantly, both non-lactating and lactating mice eat for a similar amount of time in this period, suggesting that these differences are not solely secondary to increased food consumption in lactating mice (Extended Data Figure 6D). In contrast, consistent with the lack of baseline differences in AgRP neuron activity in the ad libitum fed state (Figure 4), presentation of food to lactating and non-lactating mice in the ad libitum fed state did not alter AgRP neuron activity in either group of mice (Extended data Fig. 7). To determine if the changes in AgRP neuron activity observed during lactation are due to a potential order effect associated with multiple fiber photometry recordings, we performed the identical experiments on the same days in mice that did not become pregnant and did not lactate. In contrast to lactating mice, a similar inhibitory response in AgRP neurons was observed to both the presentation of food and food consumption in “time-matched” non-lactating mice (Extended Data Fig. 8).

We previously observed that lactating mice consume more palatable food than non-lactating mice (Figure 3). Therefore, we next tested if the response of AgRP neurons to palatable food consumption differs in non-lactating and lactating mice (ZT6, Figure 5C). Additionally, this assay allows for the inhibitory response to food presentation and food consumption to be compared between lactating and non-lactating mice in the absence of negative energy balance (i.e. in ad libitum fed conditions). In this experiment we first habituated mice to presentation of peanut butter chips in their home cage for 10 min for two days prior to testing. On the testing day we provided ad libitum-fed non-lactating and lactating mice with acute access to a palatable peanut butter chip (PB chip) and measured the activity of AgRP neurons during the presentation of a PB chip and during PB chip consumption (Figure 5C). Interesting, both non-lactating and lactating mice exhibited an equivalent decrease in AgRP neuron activity in the seconds following introduction of a PB chip (Extended Data Fig. 9). Since food consumption inhibits AgRP neurons in proportion to caloric intake [20,21,54], we next compared the change in AgRP neuron activity following the initiation of PB chip consumption in non-lactating and lactating mice (Figure 5H,I). At the onset of food consumption, AgRP neuron activity transiently decreased in both non-lactating and lactation mice (Figure 5H). This inhibitory response to PB chip consumption was markedly prolonged in lactating mice (Figure 5H), indicating that the increased palatable food intake associated with the lactational state is accompanied by an enhanced inhibition of AgRP neuron activity (Figure 5H,I). To determine if this enhanced inhibitory response to food consumption is secondary to the increased palatable food intake in lactating mice (Figure 3), we further analyzed the change in AgRP neuron activity during the 20 s period following food consumption, a period when non-lactating and lactating mice consume a similar amount of PB chip (Extended Data Fig. 9). In contrast to the 60 s following PB chip consumption, no difference in calcium signal was detected between non-lactating and lactating mice during the initial 20 s of PB chip consumption (Extended Data Fig. 9). Therefore, palatable food consumption results in a more sustained inhibition of AgRP neurons in lactating mice but does not alter the initial inhibitory response to palatable food consumption.

2.7. Re-introduction to pups activates AgRP neurons in lactating mice

AgRP neurons integrate both homeostatic state (i.e. hunger/satiety) and future energy needs to adaptively control feeding behavior. For example, to increase food-seeking during periods of food availability, AgRP neuronal activity spikes immediately prior to the expected time of food delivery in restricted food access experiments [11]. Further, in addition to being inhibited by food consumption, AgRP neurons are also inhibited by cues that predict the future availability of food in food-restricted mice [25,26]. Given these results, we hypothesized that during lactation AgRP neurons may respond to the presence of pups as a cue of future energy demands, or that the energetic demand of nursing may stimulate AgRP neuron activity to promote feeding. To test these hypotheses, we monitored the activity of AgRP neurons after re-introducing pups to lactating and non-lactating mice. Pups were removed from the dam for 6 h (ZT3-ZT9), upon which they were re-introduced to non-lactating and lactating mice while visualizing AgRP neuron activity with fiber photometry (Figure 6). No change in AgRP neuron activity was observed in the 5 min following pup introduction in both lactating and non-lactating mice (Figure 6A). However, in lactating mice AgRP neuron activity gradually increased in the period between 5 min and 25 min following pup introduction (Figure 6A,B). In contrast, AgRP neuron activity remained stable during this period in non-lactating animals (Figure 6A,B). Upon post-hoc analysis of nursing behavior, the period of increased AgRP neuron activity in lactating mice correlated closely with the time that the mice began nursing the pups, suggesting that the suckling stimulus/the energetic demands of nursing increases AgRP neuron activity (Fig. 6A).

Figure 6.

Re-introduction of pups increases AgRP neuron activity in lactating mice. A: Average trace of the calcium signal from non-lactating (blue) and lactating (red) mice following the introduction of pups into the cage after 6 h of pup separation between the dam and the pups (time point 0 on x-axis). B: Quantification of the data in a demonstrating increased AgRP neuron activity following the introduction of pups in lactating mice, with no response observed in non-lactating animals. C: Average trace of the calcium signal from non-lactating (blue) and lactating (red) mice following the introduction of pups into the cage after 16 h of separation between the pups and the dam (time point 0 on x-axis). D: Quantification of the data from c demonstrating increased AgRP neuron activity following re-introduction of pups to lactating mice. E: Schematic showing the timeline for pup removal and re-introduction experiments. F: Re-introduction to pups following 16 h of pup separation between the dam and the litter results in a gradual increase in feeding behavior over the 8 h following pup re-introduction. Data represents mean +/− s.e.m. Data in B and D analyzed with Mann Whitney test. ∗p < 0.05, ∗∗p < 0.01.

Next, we tested if the activation of AgRP neurons following pup re-introduction is sensitive to the amount of time that the dam is separated from the pups. To test this, we repeated the previous pup-reintroduction experiments following 16 h of deprivation from the dam (ZT12-ZT4, Fig. 6C). Like 6-hour deprivation experiments, AgRP neuronal activity was not altered in the 2 min following pup introduction, indicating that the sight of the pups likely does not regulate AgRP neuron activity (Fig. 6C). However, in contrast to 6 h of pup deprivation, AgRP neuron activity was significantly increased within 5 min of pup re-introduction (Fig. 6C). This activity remained elevated for approximately 15 min, upon which the AgRP neuron activity returned to non-lactating levels (Figure 6C,D, Extended Data Fig. 10). Importantly, no change in AgRP neuron activity was observed following pup introduction in non-lactating mice (Figure 6C,D, Extended Data Fig. 10). As observed following 6-hour pup deprivation, the period of increased AgRP neuron activity in lactating dams closely matched the time that the dam was in the nest with the pups (Figure 6C).

Given that AgRP neuron activity is increased following pup introduction, we next mapped the feeding response in lactating mice following re-introduction to pups (pups added to cage at ZT4, 16 h of pup deprivation from the nest; Figure 6E). Interestingly, food intake remained stable in lactating mice in the first 2 h following pup re-introduction and did not increase until approximately 3 h after the introduction of pups (Figure 6F). Further, feeding did not return to the levels observed before pup removal until approximately 8 h after re-introduction of the pups (Figure 6F). Together, these results suggest that the increased AgRP neuron activity upon re-introduction to pups is not immediately associated with increased feeding and may instead act as a signal of future energetic needs, reminiscent of recent findings demonstrating increased AgRP neuron activity in periods associated with low feeding (i.e. the light period) [11].

2.8. AgRP neuron activity regulates lactational hyperphagia

Our prior cfos and fiber photometry experiments demonstrate that AgRP neuronal activity is altered during lactation (Figure 4, Figure 5, Figure 6). However, these experiments do not establish if AgRP neurons are required for promoting the hyperphagia of lactation. Therefore, we next utilized chemogenetic inhibition experiments to test if AgRP neuron activity is required for promoting lactational hyperphagia. AgRP-Cre mice were injected into the arcuate nucleus with the chemogenetic DREADD (designer receptor exclusively activated by designer drugs) inhibitor hM4Di or control virus expressing mCherry fluorescent protein (Figure 7A). DREADD targeted mice exhibited dense expression of hM4Di-mCherry in the arcuate nucleus (Figure 7A), enabling selective inhibition of these neurons. Following viral injections, mice were assigned to non-lactating and lactating groups as described in prior fiber photometry experiments. Approximately 5 weeks later, during the mid-stage of lactation or the equivalent time-period in non-lactating mice, we performed functional feeding assays to access the contribution of AgRP neurons to lactational-hyperphagia (Figure 7). In ad libitum fed conditions inhibition of AgRP neurons with CNO did not affect feeding in non-lactating or lactating mice expressing hM4Di (ZT12-ZT14, Extended Data Fig. 11). These results are consistent with our previous findings indicating no difference in cfos expression in AgRP neurons between lactating and non-lactating mice in the ad libitum fed state (Figure 4). Since we observed increased AgRP neuron activity in lactating mice following acute food restriction (Figure 4), we next repeated chemogenetic inhibition experiments following 90 min of acute fasting during the light period (ZT 10.5-ZT12, Figure 7B). In contrast with the ad libitum state, CNO mediated inhibition of AgRP neurons reduced food intake following 90 min of food restriction in lactating mice expressing hM4Di, and this effect was absent in non-lactating control mice (Figure 7B). No difference in feeding was detected following CNO administration in non-lactating or lactating mice expressing mCherry control virus, indicating that these effects were not mediated by off-target effects of CNO (Figure 7C). Finally, since we previously observed altered AgRP neuron dynamics during high fat feeding (Figure 3), we also tested if the role of AgRP neurons in promoting lactational-hyperphagia is regulated by the palatability of the food. Since a high-fat diet can alter the activity of AgRP neurons [55,56], these experiments were performed at the end of our experimental paradigm (ZT12-ZT14). We administered saline or CNO to non-lactating and lactating mice expressing hM4Di in AgRP neurons and measured consumption of high fat pellets (20 mg pellets, 27% fat). Acute chemogenetic inhibition of AgRP neurons inhibited feeding in lactating mice, normalizing the feeding of the lactating animals to the level observed in non-lactating mice (Figure 7D). In contrast, inhibition of AgRP neurons had no effect on HFD intake in non-lactating mice (Figure 7D).

Figure 7.

Inhibition of AgRP neurons reduces feeding in lactating mice. A: Representative image showing viral expression of hM4Di-mCherry in AgRP neurons. B: Food intake (2-hour food intake) following injections of saline or CNO (1 mg/kg) to non-lactating or lactating mice expressing hM4Di-mCherry in AgRP neurons (n = 10 lactating mice and n = 10 non-lactating mice). CNO administration reduced food intake in lactating mice but did not have an effect in non-lactating mice. C: Same experiment shown in b but for mice expressing mCherry control virus in AgRP neurons. Administration of CNO had no effect on feeding in lactating or non-lactating mice expressing mCherry in AgRP neurons (n = 7 non-lactating mice and n = 5 lactating mice). D: High fat diet food intake (2-hour food intake) in non-lactating and lactating mice expressing hM4Di-mCherry in AgRP neurons (n = 10 lactating mice and n = 10 non-lactating mice). Administration of CNO reduced HFD food intake in lactating mice but not non-lactating animals. E: Average meal size of non-lactating and lactating mice expressing hM4Di-mCherry in AgRP neurons following injections of saline or CNO (calculated from first 2 h following drug administration). Lactating mice consume larger meals than non-lactating mice following saline injections, and this effect is attenuated following CNO-mediated inhibition of AgRP neurons (n = 10 lactating mice and n = 10 non-lactating mice). F: Number of AgRP neurons containing cfos following injections of saline or CNO to non-lactating and lactating mice expressing hM4Di in AgRP neurons. CNO administration reduced the number of cfos positive AgRP neurons in lactating mice, an effect that was not observed in non-lactating animals (n = 5 mice per group). Data represents mean +/− s.e.m. Data analyzed by 2-way ANOVA with Sidak's post hoc test. n.s. (not significant), ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.005. Scale bar (300um).

To determine the behavioral mechanism mediating reduced feeding behavior following AgRP neuronal inhibition in lactating mice we quantified meal size and meal number following AgRP neuronal inhibition (Figure 7E). Consistent with prior experiments, lactating mice treated with saline consume significantly larger meals than non-lactating animals (with no difference in meal number) (Figure 7E). In contrast, in the presence of CNO mediated inhibition of AgRP neurons, no difference was observed in the average meal size of non-lactating and lactating mice (Figure 7E). In contrast, inhibition of AgRP neurons did not alter meal number in non-lactating or lactating mice (Extended Data Fig. 12). Thus, inhibition of AgRP neurons reduces lactational hyperphagia primarily by reducing meal size in lactating dams (Figure 7E).

To confirm successful chemogenetic inhibition of AgRP neurons in lactational hyperphagia assays, we performed cfos immunohistochemical analysis of cfos expression in AgRP neurons following administration of saline or CNO to non-lactating and lactating mice expressing hM4Di in AgRP neurons. Immunohistochemical experiments were performed in a similar manner as described in our behavioral assays (i.e. mice were food deprived for 90 min prior to re-feeding from ZT10.5-ZT12), and brains were processed for cfos immunohistochemistry 2 h after administration of saline or CNO. Since feeding inhibits AgRP neuron activity, mice were not re-fed prior to immunohistochemistry analysis. Consistent with our behavioral results, CNO administration did not significantly alter the number of cfos-positive AgRP neurons in non-lactating mice. In contrast, administration of CNO significantly reduced the percentage of AgRP neurons expressing cfos in lactating mice (Figure 7F).

In a new cohort of mice, we performed additional chemogenetic inhibition experiments to further characterize the role of AgRP neurons in lactational hyperphagia, and to determine if the role of AgRP neurons in lactational-hyperphagia is mediated by altered milk production, maternal behavior, or metabolism (Figure 8). First, we tested if the role of AgRP neurons in promoting lactational-hyperphagia requires the physical act of lactation, by acutely inhibiting AgRP neurons in lactating mice without the pups present (Figure 8A). As observed with the pups present, chemogenetic inhibition of AgRP neurons reduced food intake in lactating mice (ZT12-ZT14, Figure 8A), but not in lactating animals expressing control mCherry protein (Figure 8B), indicating that the role of AgRP neurons in lactational hyperphagia does not require the suckling stimulus (Figure 8A,B).

Figure 8.

Inhibition of AgRP neurons specifically alters feeding during lactation. A: Food consumption in lactating mice expressing hM4Di in AgRP neurons in the absence of pups in the cage. CNO administration significantly reduced food intake in lactating mice in the acute absence of pups (n = 8 mice). B: Same experiment shown in a but for mice expressing control mCherry virus in AgRP neurons. CNO administration had no effect on food intake (n = 11 mice). C: Change in the body weight of the pups (2-hrs after pup reintroduction to the dam) following acute pup removal and re-introduction to the dam. Pups were removed for 0 h (nd), 2 h, and 6 h prior to re-introduction to the dam, and CNO was administered to all mice immediately prior to pup re-introduction. Pups from control mCherry and hM4Di expressing dams gained weight at a similar pace (n = 8 hM4Di mice and n = 11 mCherry mice). D: Schematic of experimental approach for experiments in e and f. Food was removed from lactating and non-lactating mice expressing hM4Di in AgRP neurons, and changes in the body weight of the dam and the pups was measured following 3 h of food deprivation. Saline or CNO was administered to all dams (or non-lactating control mice) immediately prior to food removal. E: Change in the body weight of the dams and non-lactating mice 3 h following food removal. Lactating mice exhibited a drastic reduction in body weight compared to non-lactating animals. However, CNO-mediated inhibition of AgRP neurons did not alter the magnitude of this response (n = 8 non-lactating mice and n = 9 lactating mice). F: Change in the body weight of the pups 3 h after saline or CNO injection in the experiment outlined in D. Pups gained similar amounts of body weight following both saline and CNO injections in lactating dams (n = 9 mice). Data represents mean +/− s.e.m. Data in A, B, C, and E analyzed with 2-way ANOVA with Sidak's post hoc test. Data in f analyzed with paired Student's t-test. ns (not significant), ∗p < 0.05, ∗∗∗∗p < 0.001.

The neural circuitry controlling feeding behavior overlaps with circuits controlling maternal behavior [57], and activation of AgRP reduces maternal nest-building behaviors [58]. Therefore, we next tested if chemogenetic inhibition of AgRP neurons impairs maternal behavior and milk production as changes in milk production or maternal behavior could result in reduced feeding behavior in lactating mice. To test if inhibition of AgRP neurons impairs milk production and nursing of the pups we measured the body weight of pups following chemogenetic inhibition of AgRP neurons. Experiments were performed following 0 h, 2 h, and 6 h of separation of the dam from the pups to stimulate maternal behavior in the dam (experiments performed between ZT4-ZT10, Figure 8C). Following this separation period, we administered CNO to lactating dams expressing either hM4Di or control mCherry in AgRP neurons and measured the body weight of the pups 2 h later (Figure 8C). As expected, pups gained weight following 2 h of re-exposure to the dam, and the magnitude of weight gain increased as the duration of deprivation from the dam increased (Figure 8C). Importantly, pups from hM4Di and mCherry targeted dams gained a similar amount of weight following re-introduction of the dam, indicating that inhibition of AgRP neurons does not prevent nursing of the pups (Figure 8C).

In addition to regulating feeding behavior, AgRP neurons also control metabolism by modulating the hypothalamic-pituitary-adrenal axis (HPA axis) [10,18] and modulating substrate utilization and energy expenditure [59,60]. To test if AgRP neurons contribute to altered metabolism in lactating mice, independent of feeding, we administered saline or CNO to lactating and non-lactating mice expressing hM4Di or control-mCherry in AgRP neurons and measured changes in the body weight of the dam and pups after 3 h of food deprivation (ZT6-ZT9, Figure 8D). Thus, changes in the body weight of the dam during this period must reflect altered energy expenditure or metabolism (i.e. changes in neuroendocrine or autonomic function). Lactating mice exhibit a remarkable decrease in body weight following acute food deprivation, losing approximately 2.5 g of body weight in 3 h (Figure 8E; compared to 0.4 g of body weight in non-lactating animals). However, inhibition of AgRP neurons did not affect body weight loss during food deprivation as lactating mice lost similar amounts of weight following saline or CNO injections (Figure 8E). Remarkably, despite the dams losing approximately 2.5 g of body weight during acute food deprivation, the body weight of the pups increased during this period, indicating that lactating dams continue to nurse their pups despite considerable weight loss (Figure 8F). Taken together, AgRP neurons regulate the hyperphagia of lactation without altering maternal behavior or the mobilization of energy reserves during an acute fast (Figure 8D–F).

3. Discussion

Lactation presents an enormous metabolic challenge in mammals that is met by drastically increasing food intake [1]. An appropriate feeding response to lactation is ultimately critical to the long-term metabolic health of both the dam and pups since overfeeding during lactation is associated with an increased risk for obesity and diabetes later in life in rodents [[2], [3], [4], [5], [6]]. Similar detrimental effects of overfeeding have also been observed in the third trimester of humans, which represents a developmental time-period analogous to the lactational period in rodent pups [6]. Despite this, the core behavioral mechanisms and neural circuitry required for promoting increased feeding during lactation are not well understood.

Here we utilized home cage feeding devices (FED3) that allow for the precise calculation of feeding structure in undisturbed lactating mice (Figure 1). By utilizing this approach, we identified that the hyperphagia of mid-lactation is primarily driven by increased meal size, with a concomitant decrease in meal frequency (Figure 1). This decrease in meal frequency is likely due to the increased time associated with nursing and caring for young since meal number in lactating mice is equivalent to non-lactating mice following the acute removal of pups, while meal size remains elevated in the absence of pups (Figure 2). The increased meal size reported here is consistent with prior work demonstrating increased meal size in rats throughout lactation [1,46,61]. However, rats also demonstrate increased meal frequency towards mid-late lactation [1]. These species differences may represent inherent differences in lactational-hyperphagia between rats and mice or a difference in the approaches utilized to calculate feeding dynamics. Consistent with increased meal size, we report here that lactating mice have reduced sensitivity to the acute satiety effects of the gut-derived satiation signals PYY and GLP1 (Figure 3). Further, prior work has demonstrated that in rats the early stage of lactation is associated with an impaired satiety response to the fat-derived adipostatic factor leptin [[37], [38], [39],41,45,62]. Together, these results suggest that an impaired response to satiation signals may represent a core feature promoting the hyperphagia of lactation in rodents. Further work is required to determine the mechanism(s) mediating this impaired satiation response and the downstream neural circuitry that is involved.

In addition to impaired satiety signaling, increased feeding can also be mediated by enhanced sensitivity to signals of negative energy balance and increased hedonic or reward-based feeding. Here, we demonstrate that lactating mice also exhibit increased sensitivity to negative energy balance (Figure 3). For example, while 90 min of food deprivation in the light period does not significantly alter feeding in non-lactating mice, this manipulation drastically increases feeding in lactating mice (Figure 3). Further, administration of the gut-derived hunger hormone ghrelin has an enhanced orexigenic effect in lactating mice vs non-lactating animals (Figure 3). These results are particularly striking given that lactating mice already exhibit profound hyperphagia in baseline conditions. Although the mechanism(s) mediating this enhanced response to negative energy balance are unknown, it may involve an increase in ghrelin receptor (GHSR) mRNA expression in the hypothalamus as prior reports indicate that lactation increases GHSR mRNA expression in the hypothalamus [63,64].

Lactating mice also exhibit increased hedonic feeding, consuming more palatable high-fat diet and peanut butter chips than non-lactating animals in the ad libitum fed state (Figure 3). These results are analogous to recent work in mice demonstrating a critical role for hedonic feeding in the context of the hyperphagia of pregnancy and suggest that links between reproductive state and feeding behavior likely involve both homeostatic and hedonic mechanisms [65]. Additional work is required to determine how the hormonal and neuroendocrine changes associated with lactation communicate with hedonic feeding circuits, particularly given the established effect of reproductive hormones on reward circuitry [[66], [67], [68]].

Although the hyperphagia of lactation is well known, the specific cell types and neural circuits mediating this effect are largely uncharacterized. Given that hypothalamic AgRP neurons are activated in response to caloric deprivation and promote feeding during conditions of negative energy balance, we hypothesized that these cells may contribute to the hyperphagia of lactation. Here, we characterized the neuronal activity patterns of AgRP neurons in lactating mice and determined that AgRP neurons exhibit increased activity following 90 min and 10 h of food deprivation in lactating mice, relative to non-lactating mice (Figure 4). Conversely, no difference in cfos expression was detected in AgRP neurons in the ad libitum fed state (Figure 4). Since AgRP neurons are inhibited by food consumption [20,21,25,54], and lactating mice consume 2–4x more food than non-lactating mice, we suspect that the elevated food consumption associated with lactation may tonically suppress AgRP neuron activity and/or alter the in vivo dynamics of these cells in a manner which cannot be detected with cfos immunohistochemistry. Instead, data presented here suggest that AgRP neurons are more sensitive to negative energy balance in lactating mice, and this increased sensitivity may sustain feeding behavior by altering the dynamics of AgRP neurons (Figure 4).

Prior work indicates that AgRP neurons are rapidly inhibited at the sight of food and following food consumption in food deprived mice [20,21,25,54]. While the rapid inhibition at the sight of food is sensory driven [25,26], the inhibitory response to food consumption depends on the detection of calories in the GI tract, and the magnitude of this inhibitory response correlates with the number of calories consumed [20,21,25,54]. In lactating mice, AgRP neurons exhibit an enhanced inhibitory response to both the presentation of food and food consumption following food deprivation (Figure 5). Further, AgRP neurons exhibit a prolonged inhibitory response to palatable food consumption in lactating mice (Figure 5). Given that AgRP neurons exhibit increased cfos activation following a 10-h fast in lactating mice (Figure 4, an identical experimental paradigm as the fiber photometry experiments outlined here), the increased inhibitory response of AgRP neurons in lactating mice may result from elevated baseline activity in food deprived mice. Consistent with this hypothesis we observed an enhanced inhibitory response to food presentation in 10-hour food deprived mice (Figure 5), but not following presentation of regular chow or peanut butter chips to ad libitum fed mice (Figure 5). However, we cannot exclude the possibility that lactation may also enhance the sensitivity of AgRP neurons to the consumption of calories, resulting in enhanced inhibition of AgRP neurons following food consumption in lactating animals. Such a hypothesis is consistent with our findings that the inhibitory response of AgRP neurons to food consumption remains enhanced in lactating mice during periods when both lactating and non-lactating mice consume similar amounts of food (Extended Data Fig. 6). Thus, the enhanced inhibitory response of AgRP neurons to feeding during lactation may result from elevated baseline AgRP neuron activity following acute food deprivation and/or enhanced sensitivity of AgRP neurons to food consumption. Distinguishing between these possibilities will require further technological approaches such as long-term fiber photometry recordings [11] and/or precise intragastric infusion of calories to non-lactating and lactating mice that are matched for baseline AgRP neuron activity [20,21,54].

Since AgRP neurons are directly sensitive to sensory stimuli associated with feeding, such as audio-visual cues predicting food availability, we hypothesized that the presence of pups may provide a signal of increased energy demands and thus regulate the activity of these neurons. However, AgRP neuron activity is not acutely altered in the immediate seconds following pup re-introduction in lactating mice (Figure 6). Instead, AgRP neurons are activated in the 30 min following pup re-introduction, and the kinetics of this activation depends on the duration of separation between the dam and the pups (Figure 6). For example, AgRP neurons are activated faster following 16 h of pup deprivation than 6 h of pup deprivation, although both periods of pup deprivation ultimately result in a similar magnitude of AgRP neuron activation (Figure 6). Although the specific mechanism mediating increased AgRP neuron activity upon pup re-introduction are unknown, this response is unlikely to be primarily mediated by sensory exposure to the pups since increased AgRP neuron activity does not occur in the first minutes following pup introduction (Figure 6). Instead, increased AgRP neuron activity starts approximately 4 min after pup introduction and is sustained for at least 12–15 min. Further, this increase in activity is relatively well time-locked to the initiation and cessation of nursing behavior in mice (Figure 6). Given that AgRP neuron activation occurs within seconds of the suckling stimulus/entry of the dam to the nest (Figure 6), these results are not consistent with an effect mediated primarily by the energetic demands associated with nursing. For example, the increase in AgRP neuron activity upon entry to the nest (on the order of seconds) likely precedes the energetic costs associated with nursing, which would be expected to take multiple minutes to occur following the initiation of nursing. Further, the increase in AgRP neuron activity does not continue to rise throughout the course of nursing, as would be expected for a signal based on the cumulated energy deprivation associated with lactation (Figure 6). Instead, our results suggest that the suckling stimulus and/or the physical presence of the pups activates AgRP neurons as a signal of impending energetic demand. Such a hypothesis is consistent with the kinetics of food intake following re-introduction of the pups, since food intake does not increase for at least 2–3 h following re-introduction of the pups, despite increased AgRP neuron activity in the minutes following pup introduction (Figure 6). Indeed, recent work indicates that the in vivo activity of AgRP neurons may more appropriately reflect future energetic needs as AgRP neuron activity is elevated prior to eating and in conditions associated with low levels of food intake [11]. Future work is required to identify the synaptic and neuroendocrine mechanism(s) mediating activation of AgRP neurons upon re-introduction of pups to lactating dams.

Although the activity of AgRP neurons is altered in lactating mice (Figure 4, Figure 5, Figure 6), it is unclear if altered AgRP neuron activity is required for promoting the hyperphagia of lactation. Chemogenetic experiments performed here indicate that AgRP neuron activity is acutely required to promote the hyperphagia of lactation (Figure 7). For example, although inhibition of AgRP neurons did not alter food intake in ad libitum fed mice, AgRP neuron inhibition significantly reduced food intake following 90 min of food deprivation in lactating, but not non-lactating animals (Figure 7). This inhibition of feeding was associated with reduced meal size, indicating that the increased meal size associated with lactation may in part result from elevated AgRP neuron activity (Figure 7). Consistent with this hypothesis, increased AgRP neuron activity delays satiation by inhibiting the activity of parabrachial CGRP neurons controlling meal termination [69]. Thus, the elevated AgRP neuronal dynamics associated with lactation may delay satiation in lactating animals by inhibiting downstream neurons signaling meal termination, although further work is required to confirm this hypothesis. Importantly, the ability of AgRP neuron inhibition to reduce feeding behavior in lactating mice is not a result of secondary changes in milk production, nursing behavior, or altered energy metabolism as inhibition of AgRP neurons does not alter these processes in lactating mice (Figure 8). Together, these results demonstrate that AgRP neurons show enhanced sensitivity to negative energy balance in lactation (Figure 4, Figure 5), and that dynamic AgRP neuron activity is acutely required for promoting the hyperphagia of lactation (Figure 7, Figure 8). Further work is required to determine if the AgRP neurons exhibit a unique role in linking lactation to increased feeding, or if other orexigenic cell populations are also involved. Additionally, in future studies it will be important to establish if AgRP neuron activity is required for long-term metabolic changes associated with lactation and for the proper development of pups. However, here we uncover an important role for AgRP neurons in promoting the hyperphagia of lactation and provide a cellular entry point for characterizing the link between the lactational state and feeding behavior.

4. Methods

4.1. Animals

All experiments were approved by the University of Illinois Institutional Animal Care and Use Committee. Littermate mice of approximately equal age and body weights were used as experimental and control groups in all experiments (age 8–16 weeks). Experiments were performed on C57BL6J mice (Jax#000664) or AgRP-Cre mice (catalog # 012899, The Jackson Laboratory). For experiments with AgRP-Cre mice, mice were bred in house (heterozygous Cre+ x C57BL6J WT mice), and litters were genotyped upon weaning by collecting ear tissue samples, isolating genomic DNA, and running polymerase chain reaction (PCR) with primers for the Cre gene to confirm the presence of Cre allele. Primers that were used for PCR techniques are AgRP-cre Common (100uM to 10uM) (5′ GCT TCT TCA ATG CCT TTT GC 3′), and AgRP-cre Mutant (100uM to 10uM) (5′ AGG AAC TGC TTC CTT CAC GA 3’). Before any experiments, female mice were caged in groups of two to five mice per cage. All the mice were housed in humidity and temperature-controlled (19–21 °C) cages and their room was under 12 h light/dark cycle where the light cycle begins at 6 A.M. (ZT0) and dark cycle begins at 6 P.M. (ZT12). ZT0-ZT12 indicates the light period, while ZT12-ZT24 indicates the dark period. All mice had access to regular chow (ad libitum) and water, unless otherwise noted in the text and figure legends. Experiments were performed during the mid to late stage of lactation (i.e. lactation day 7–19). Experiments were not performed in the first week following birth to allow the dam to recover from the birthing process. For behavioral experiments lactating mice had between 4 and 8 pups per litter. All experiments were performed with different cohorts of animals unless otherwise noted in the text and figure legends.

4.2. Viral vectors

Adeno-associated viral (AAV) vectors that were used included Cre-dependent GCAMP6s (AAV5-Syn-Flex-GCaMP6s-WPRE-SV40), Cre-dependent hM4Di (AAV5-hsyn-DIO-hM4D-mCherry), and Cre-dependent control virus (AAV5-hSyn-DIO-EGFP). All viral vectors were purchased from Addgene.

4.3. Stereotaxic viral injections and fiber optic implants

Before beginning stereotaxic surgery, mice were subjected to anesthesia inside of an isoflurane chamber and then positioned in a stereotaxic frame (Kopf) with a constant flow of isoflurane and oxygen. Following administration of preoperative carprofen (5 mg/kg), the incision area was sterilized by iodine and 70% ethanol and stereotaxic surgeries was performed as previously described [70,71]. AAV vectors were injected into arcuate nucleus of the hypothalamus (ARC) by using a micromanipulator attached to pulled glass pipette (Ronal Tool). Viral injection coordinates for ARC were A/P: −1.30 mm, and −1.80 mm from bregma; M/L: −0.30 mm, and +0.30 mm; D/V: −5.65 mm to −5.80 mm from the surface of brain. In each injection site 250 nl of virus was injected for a total injection volume of 1 ul per mouse. For Gcamp6s AAV injections 300 nl of GCAMP6 virus was injected at each site for a total injection volume of 1.2 ul per mouse. After injections, the incision area was closed using tissue adhesive glue. Following stereotaxic injections, mice were returned to their home cages and monitored for 10 days.

For fiber photometry experiments, mice with GCaMP6 viral vector injected into their ARC were subsequently implanted with a fiber optic cannula 2–3 weeks after AAV injections in a second surgery. 200 um fiber optic cannulas (0.66 NA, Plexon) were implanted into ARC with the following coordinates A/P, −1.60 mm; M/L, 0.0 mm; D/V, −5.90 mm. After successful insertion of fiber optics, fiber optic implants were secured to the skull by using dental cement (C&B Metabond). The incision site was glued by using tissue adhesive and mice were returned to their cages for post-surgical recovery period.

4.4. Food intake studies

All mice were single caged for at least one week prior to performing feeding measurements. Feeding studies were performed using feeding experimental devices, except for data in Figure 2, Figure 3E-H, which were collected manually in the mouse home cage. For manual food intake measurements, a set amount of food was weighed daily, and food intake was measured by weighing the food after different times following i.p. injections. These experiments were performed at the start of the dark cycle (i.e. 6 pm CST), except for ghrelin injection experiments which were performed during the light period (12 pm CST).

4.5. FED3 feeding assays

To measure feeding structure in mice we utilized FED3 feeding devices. For FED3 feeding assays there were no regular chow in mice cages, so all food had to be consumed by using FED3 devices. FED3 devices have two nose poke sites for poking (correct poke; left) or incorrect poke (right poke) and third site (in the middle of correct and incorrect poking sites) where a 20 mg food pellet is dropped upon successful completion of a correct nose poke. To reduce food pellet hoarding and more accurately quantify food intake, all experiments were performed on an FR1 schedule of reinforcement. In this case one correct nose poke (left poke) results in the release of one 20 mg food pellet into the food pellet hole. The removal of the food pellet is directly sensed by the FED3 devices, allowing for precise measurements of food consumption.

To train mice to use FED3 devices, mice were single caged for one week prior to testing and a FED3 device was provided to each mouse by directly attaching the device to the side of a typical mouse cage (see Figure 1). First, mice were acclimated to FED3 devices for one day on free feeding mode. In this mode each pellet retrieval resulted in an additional pellet retrieval (regardless of nose poke behavior). However, in our experimental setting, this mode resulting in hoarding of pellets (i.e. many pellets located in the cage that were not consumed), resulting in inaccurate food intake quantification. Therefore, after 1 day of free feeding training, the feeding mode was changed from free feeding to FR1 mode which required one correct poke (left poke) to deliver food pellet. Experiments were initiated once all mice reached more than 70% correct poking (typically 1–2 days). After every experiment, data was exported from FED3 devices to where the number of consumed pellets, number of meals, and meal sizes were calculated using the fed3viz software. A meal was defined as the number of pellets taken by the mouse where the maximum inter-pellet interval between each pellet was 5 min.

4.6. Pup removal feeding studies

Pup removal experiments were performed in the late period of lactation (lactation day 14–17). For pup removal feeding experiments pups were removed from the cage and placed into new clean cages with a heat lamp. The protocol involved three groups (Figure 3A–D): non-lactating female mice, lactating mice with pups present, and lactating mice from which pups were temporarily removed. For the acute pup removal group, pups were removed from the dam for 6 h during the light period (10 am–4 pm) and returned to the cage with the original dam at 4 pm. Food intake was measured during the same 6 h for all three groups. Total pellets consumed, meal size, and meal number was calculated from 10 am to 4 pm in lactating dams with the pups removed, lactating dams with the pups present, and non-lactating female control mice using FED3 devices.

To map the dynamics of feeding behavior following the removal of the pups (Figure 3E–H) we first measured 24-hour food intake in non-lactating female mice, and two groups of lactating dams with the pups present. On the following day, pups were removed from one of the groups of lactating dams and 24-hour food intake was measured for an additional two days. Food intake was measured manually by weighing a pre-measured amount of food daily. Cages were changed daily to reduce the crumbs associated with eating.

4.7. Negative energy balance feeding studies

Feeding experiments to assess the response of non-lactating and lactating mice to negative energy balance were conducted using automated home cage feeding devices (FEDs). Prior to experiments, mice were acclimated to the FEDs over a 2–3-day period to familiarize them with the operation of the devices (see FED3 methods section). The experiments were performed between lactation days 15–18. For the 90-minute fasting study, the mice were fasted by deactivating the FEDs for 90 min just before the onset of dark cycle. Immediately after the fasting period, the FEDs were reactivated to allow food consumption, facilitating the measurements of post-fast feeding behavior. In contrast, the ad libitum group had continuous access to food as their FED devices remained activated throughout, allowing for continuous recording of feeding data. For the 10-hour fasting protocol, the FEDs were turned off from 7 am to 5 pm. Following this extended fasting period, the devices were reactivated for 30 min to measure immediate post-fasting food intake.

4.8. Ghrelin feeding assays

Ghrelin feeding assays were performed in the mouse home cage by manually measuring the amount of consumed food. Non-lactating and lactating mice received intraperitoneal injections of saline at 12 pm to record baseline acute food intake measurements, which were recorded 30 min following saline injections. Following baseline saline injections mice were randomly assigned to receive either saline (200ul saline, i.p.) or ghrelin injections (1ug/g in 200ul saline, i.p.) at 12 pm. These groups were then flipped on the following day to counterbalance the effects and the food intake on the last two days was calculated for each mouse following saline or ghrelin injections for statistical analysis.

4.9. Hedonic feeding studies

To measure the acute feeding response to palatable food in lactating and non-lactating mice we provided acute access of non-lactating and lactating mice to a high fat diet (HFD, 60%kcal in fat) or peanut butter chips in their home cage. A premeasured amount of HFD or PB chips were placed on the bottom of the mouse's home-cage in a small petri dish. During testing, mice only had access to these diets for 15 min (i.e. their normal food was removed from the cage). Fifteen minutes after presentation of palatable food the remaining amount of palatable food was measured, and the food consumption was calculated by subtracting this amount from the original amount of food. Following the conclusion of the 15-min palatable food intake assays, mice were returned to their home cage and provided ad libitum access to regular chow. For PB chip assays mice were first habituated to the PB chips in their home cage for 10 min the day before testing, while HFD intake studies were performed without prior habituation to the HFD pellets. HFD and PB chip consumption assays were performed on the same group of mice with one week separating the two assays.

4.10. Satiety peptide feeding experiments

For satiety peptide experiments, peptides were administered in the following dosages and methods: GDF-15 and Liraglutide were both given subcutaneously at doses of 10 ug/kg and 100ug/kg, respectively. Amylin and PYY (Peptide YY) were administered intraperitoneally, each at a dosage of 10 ug/kg. Each peptide was administered at the beginning of the dark cycle, a period when mice are most active and feeding behaviors are naturally heightened. This timing aimed to capture the maximal anorexic effect of each peptide. Following administration, food intake was measured at 2 h post-injection. Saline (200 uL) served as the vehicle control and was used to administer each peptide, ensuring consistent delivery volumes across all treatments. Baseline food intake was first established for each mouse by administering saline and measuring food intake 2 h later. Following baseline saline injections, mice were injected with satiety peptides at the same time period and food intake was recovered 2 h later. To ensure accuracy and consistency during the experimental analysis, food intake was monitored over a 24-hour period and compared against baseline measurements taken during habituation days. Only when food intake and body weight returned to baseline (saline injected levels) was a subsequent peptide injection administered to the mice.

4.11. DREADD feeding studies

DREADD feeding studies were performed on non-lactating and lactating female mice. Experiments were performed using FED3 devices. First, we tested the effect of saline or CNO administration of food intake in ad libitum fed conditions. Following two days of habituation to daily saline injections, mice were administered saline or CNO in a counterbalanced fashion 30 min prior to the start of the dark cycle (i.e. 5:30 pm, lights out at 6 pm). Food intake was subsequently measured 2 h following injections of saline or CNO. On the following day mice were all injected with saline to ensure that any lasting effects of CNO injections had dissipated. On the next day we performed 90-minute food restriction assays on the same group of mice. For 90-minute food restriction experiments, food was removed from 4 to 5:30 pm and saline or CNO was administered in a counterbalanced fashion as previously described. Food intake was recorded for an additional 2 h following saline or CNO injections. Following an additional injection of saline in ad libitum fed conditions, we repeated the 90-minute food restriction experiments but with high fat diet pellets provided in the FED3 devices. Following the completion of this assay cfos immunohistochemistry experiments and confirmation of viral expression was performed.

Feeding experiments in the absence of the suckling stimulus (i.e. Figure 8A,B) were performed on a new cohort of mice. In this experiment, pups were removed from 6 pm to 8 pm (start of the rodent dark cycle) and saline/CNO was administered to all mice at 5:30 pm in a counterbalanced fashion. Subsequent food intake was calculated with FED3 devices during the period of pup deprivation (i.e. 6 pm–8 pm). Following this experiment, after a two-day recovery period, additional pup deprivation experiments and food deprivation experiments were performed on this cohort of mice (see further description below; i.e. DREADD maternal behavior and metabolic studies).

4.12. DREADD maternal behavior and metabolic studies

DREADD maternal behavior and metabolic studies (Figure 8) were performed following acute pup removal feeding experiments. For these experiments, the change in body weight of the pups (mCherry and hM4Di expressing mice) was calculated 2 h following no separation between the dam and the pups, 2 h of separation of the dams and pups, or 6 h of separation of the dams from the pups. Body weight change of the pups was calculated during the same time of day for each of the pup separation periods.

Following acute pup removal experiments, we performed acute food removal experiments by removing food for 3 h from non-lactating and lactating mice expressing hM4Di in AgRP neurons. Mice were administered either saline or CNO on consecutive days in a counterbalanced fashion and the change in the body weight of the dam and pups was calculated following 3 h of food deprivation.

4.13. Perfusion, sectioning, and immunohistochemistry

Following in vivo experiments virally injected mice were perfused with 10% formalin, and their brain was removed via cardiac perfusion. Removed brains were transferred to 10% formalin solution for further fixation for 24 h. The next day, fixed brains were transferred into increasing amount of sucrose solutions which are 10%, 20%, and 30%. Fixed brain tissues were kept inside each increasing sucrose concentration solution approximately 24 h. Brain sections containing the arcuate nucleus (40 um thickness) were obtained by sectioning with a Lecia (CM3050 S) cryostat. Viral expression was confirmed for all mice injected with GCAMP6s or hM4Di via fluorescent microscopy. Only mice with confirmed viral expression were included in subsequent analysis. For immunohistochemistry, collected sections were placed into 24-well plates with 500 ul blocking buffer which contained 2% bovine serum albumin and 0.1% TWEEN 20. After incubation of brain sections in blocking buffer on shaker for 2 h at room temperature, blocking buffer was removed and diluted primary antibody with blocking buffer was added to each well containing brain sections (cfos, 9F6 rabbit mAb, 1:500, Cell Signaling Technology). Then, sections were placed on a shaker (4 °C) and incubated for 24 h. After incubation, primary antibody solutions were removed, and brain sections were washed with 500 ul ultra-pure 1× PBS solution for 10 min on a shaker at room temperature and this step was repeated for two more times. Following washing steps, secondary antibody (goat anti-rabbit IgG (H + L), Alexa Fluor 647) was prepared in a concentration of 1:500. Sections inside of the secondary antibody solution were placed on shaker for an additional 2 h at room temperature. After incubation with secondary antibody, washing step were repeated by using ultra-pure 1× PBS on a shaker at room temperature (3 washes for 10 min each). Finally, brain sections were mounted on superforst glass slides (Thermo Fischer Scientific), and images were captured by confocal microscopy (Zeiss LSM 700).

4.14. c-fos quantification and analysis

Cfos expression in the arcuate nucleus was captured by confocal microscopy and processed through Image J software. Approximately 3–5 images of the arcuate nucleus (across the anterior–posterior axis of the arcuate nucleus) was taken for each mouse. For each of these images the total number of AgRP+ cells, total number of c-fos signals, and total number of AgRP neurons expressing cfos were calculated using the cell counter plugin in Image J. For each mouse, the average number of each of these parameters (i.e. number of cfos cells and number of AgRP neurons expressing cfos) from all the analyzed sections was calculated for each mouse for statistical analysis.

4.15. Fiber photometry experiments

Mice were tethered to a Plexon Multi-Fiber Photometry system (Plexon, 8-61-A-07-A) through the compatible multi-fiber patch cable (Plexon, 08-60-A-04-C). For all experiments, 3–10 min of baseline recordings were performed prior to stimulus presentation (i.e. food drop, introduction of pups, etc). Blue (465 m) and UV (410 nm) light sources were provided by LED drivers from the Plexon Multi-Fiber Photometry system (Plexon, 8-61-A-07-A), which alternately cycle on and off in every 30hz sampling window to allow for orthogonal measurements of active GCaMP6 and isosbestic control signals. These fluorescent signals were collected via the installed photometric fiber and multi-fiber patch cables and detected via the installed fluorescence camera in the Plexon Multi-Fiber Photometry system. The average intensity of the 410 and 465 nm signals were collected in every 30hz sampling window and exported as .csv files for subsequent analysis in custom built R code. Photometry signals were analyzed in the context of behavioral, experimental, or feeding events (e.g. food presentation, or pup introduction). Behavioral and experimental events were recorded manually and aligned to photometry recordings using Plexon Digital Input Generator (Plexon, 07-04-A-03-B/27901) attached to the Plexon Trigger Box (Plexon 07-12-A-00-P3/92304).

4.16. Normalization of signal prior to mating

One week following implantation of the fiber optic cannula's all mice were fasted overnight to screen for successfully viral targeting and fiber optic placement. Following 3 min of baseline recording a food pellet was introduced to all mice. Only mice that exhibited a reduction in calcium signal were included in subsequent studies. To approximately normalize the baseline signal between non-lactating and lactating mice, the change in signal was calculated for each mouse and mice were evenly distributed into lactating and non-lactating groups with the same average fluorescent change to the food pellet observed in each group. Mice in the lactating group were paired with male mice for five days while mice in the non-lactating group were paired with a female mouse for five days.

4.17. Fiber photometry feeding studies in non-lactating and lactating mice

One day following mating, mice were fasted for 10 h (from 7 am to 5 pm) to measure the baseline response to presentation of food and food consumption following a 10 h fast. Following 5 min of baseline recordings a standard chow food pellet was presented to each mouse. Changes in fluorescence were calculated in the 60 s following the presentation of food and the initiation of food consumption (with the 60 s before food presentation or food consumption serving as the baseline signal). Following this assay mice were monitored daily for body weight changes and the identical photometry experiment was performed during the mid-late stage of lactation or the equivalent time-point in the non-lactating time-matched control group. Changes in the calcium response to food presentation and food consumption were compared for each mouse at these two timepoints (within-subjects comparision). Additonally, the change in calcium response was compared between baseline-matched non-lactating mice and lactating mice on the same testing day (between-subjects comparision).

Following one day of rest, PB chip presentation experiments were performed on the same group of non-lactating and lactating mice. For this experiment, mice were provided 5 min of access to the PB chip on the day prior to testing. On the testing day, 5 min of baseline recordings were performed upon which a PB chip was introduced to the testing arena. The latency to consume the PB chip and the time eating the PB chip was calculated using the Plexon digital input generator, which was used to align behavioral events to the calcium signal. Calcium signal was recorded for an additional 10 min upon which the PB chip was removed, and all mice were returned to their home-cage with ad libitum access to regular chow food. Changes in calcium response were compared between lactating and non-lactating mice on the same day of testing (i.e. between subjects comparisions) in response to both the presentation of the PB chip and the initiation of eating.

4.18. Pup reintroduction fiber photometry experiments

Following the completion of peanut butter chip photometry recordings non-lactating and lactating mice were returned to their home cages for an additional three days, upon which pup reintroduction fiber photometry experiments were performed. On the first day of testing, pups were removed from the lactating dam for 6 h (9 am-3 pm). Following pup removal, both non-lactating and lactating mice were connected to the fiber photometry system and baseline recordings were performed for 5 min. At this point, the pups from each dam were introduced into the cage and calcium activity was recorded for an additional 25 min. Pups were also introduced to non-lactating animals. The approximate start and stop time of nursing was recorded by post-hoc examining of video recordings and aligned to the fiber photometry trace. Following experiments, all mice were returned to their home cage and provided ad libitum access to food for three additional days prior to performing 16-hour pup deprivation experiments. For 16-hour pup deprivation experiments, pups were removed from the dam overnight and provided moist chow, water, ensure (liquid food), and a heating lamp. The next day identical fiber photometry experiments were performed as described for 6 h of pup deprivation. During both pup reintroduction photometry recordings, all mice had ad libitum access to food.

4.19. Fiber photometry analysis

Analysis of fiber photometry data was performed in R unless otherwise stated. Raw fiber photometry signals were normalized to the isosbestic control via the function $\Delta F/F_0=\left(F-F_0\right)/F_0$, where $F$ is the photometry signal from active GCamp6 from stimulation at 465 nm, and $F_0$ the photometry signal obtained via excitation at the GCamp6 isosbestic point, 410 nm. Where $\Delta F/F_0$ was plotted as a percentage, i.e. $\%\Delta F/F_0$, this is simply the original $\Delta F/F_0$ scaled by a factor of 100. Differences in photometry signal bias between 465 nm and 410 nm signals were corrected by changing the offset of the control signal so that the arithmetic mean of $\Delta F/F_0$ prior to experimental perturbation (i.e. food presentation or pup removal) was zero. Z-scores were calculated via the function $Z=\left(\Delta F/F-\mu \right)/\sigma$, where $\mu$, and $\sigma$ are the arithmetic mean and standard deviation of the $\Delta F/F$ trace as standard. Where photometry signals in response to events are combined, the plotted values (Z-score, $\Delta F/F_0$, or $\%\Delta F/F_0$) are the mean of the combined normalized signal, and error bars given by the standard error of the mean (SEM).

The fiber photometry analysis presented here was based on the suggestions outlined in a recent review article highlighting the advantages and disadvantages of fiber photometry analysis approaches [72], and a prior study analyzing the response of AgRP neurons to peanut butter chips (an assay performed in this paper) [73]. Although the exact ways to analyze fiber photometry results are debated [72], we chose to analyze the data by Z-scoring to best account for the mouse-to-mouse variation in signal quality associated with differences in GCAMP expression and/or fiber placement between different animals [see 72 for additional advantages/disadvantages of this analysis strategy]. This analysis approach is well suited for comparing differences in the calcium response across two different groups of animals (i.e. comparing the response to peanut butter chips in lactating vs non-lactating mice, or the response to pup re-introduction in non-lactating vs lactating mice) [73].

4.20. Data analysis and statistics

Experiments were performed during the mid-late lactational period for all experiments (lactation day 7–19). For some feeding assays data was excluded if FED3 devices jammed during testing, which occurred during approximately 5% of total FED3 studies. Normality tests were performed on all data to determine if data was normally distributed. Based on results of normality test if data had normal distribution parametric tests were applied. Non-parametric tests were used if data was not normally distributed. Only mice with confirmed viral expression were included in statistical analysis. All the data were analyzed by using GraphPad (Prism 10) software, and specific statistical tests are outlined in the figure legends.

Funding

This work was supported by National Institute of Health-National Institute of Diabetes and Digestive and Kidney Diseases Grant R00DK127065 (PS), the Foundation for Prader-Willi Research (PS), Brain and Behavior Research Foundation Grant 100000874 (PS), and National Institute of Health- Eunice Kennedy Shriver National Institute of Child Health and Human Development Grant R01HD113522 (PS).

CRediT authorship contribution statement

Kerem Catalbas: Writing – review & editing, Writing – original draft, Methodology, Investigation, Data curation, Conceptualization. Tanya Pattnaik: Writing – review & editing, Methodology, Investigation, Data curation, Conceptualization. Samuel Congdon: Writing – review & editing, Methodology, Investigation. Christina Nelson: Writing – review & editing, Methodology, Investigation. Lara C. Villano: Writing – review & editing, Investigation. Patrick Sweeney: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization.

Declaration of competing interest

P.S. owns stock in Courage Therapeutics. All the other authors declare no competing financial interests.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Extended Data Fig. 1.

6 h of pup deprivation does not impair nursing behavior. Change in the body weight of the dams following 2 h of re-introduction of the dam to the pups. Lactating dams lose a significant amount of weight following re-introduction of the pups, suggesting that they continue to nurse the pups. Data represents individual mice and mean +/− s.e.m. Data analyzed by paired Student's t-test. n = 8 mice. ∗∗∗∗p < 0.001.

Extended Data Fig. 2.

Lactating mice exhibit a normal anorexic response to amylin and GDF-15. Left panel: 2-hr food intake following administration of saline or amylin to non-lactating and lactating mice. Amylin produced a similar anorexic effect in both non-lactating and lactating mice (n = 10 non-lactating mice and n = 10 lactating mice). Right panel: 2-hr food intake in non-lactating and lactating mice following administration of GDF-15 or saline. GDF-15 produced a similar anorexic effect in both non-lactating and lactating mice (n = 10 non-lactating mice and n = 9 lactating mice). Data represents individual mice and mean +/− s.e.m. Data analyzed by 2-way ANOVA with Sidak's post hoc test. ∗∗p < 0.01, ∗∗∗p < 0.005.

Extended Data Fig. 3.

Lactating mice have higher total cfos levels in the arcuate nucleus following a 10-h fast. Left panel: Total number of cfos positive cells in the arcuate nucleus in the ad libitum fed state in non-lactating and lactating mice. No difference was observed in the total number of cfos positive cells between non-lactating and lactating mice (n = 4 non-lactating mice and n = 3 lactating mice). Middle panel: Total number of cfos positive cells in the arcuate nucleus following 90 min of fasting in non-lactating and lactating mice. No difference was observed in the total number of cfos positive cells between non-lactating and lactating mice (n = 4 non-lactating mice and n = 4 lactating mice). Right panel: Total number of cfos positive cells in the arcuate nucleus of non-lactating and lactating mice following a 10-h fast. Increased cfos expression in the arcuate nucleus was observed in lactating mice relative to non-lactating animals (n = 6 non-lactating mice and n = 4 lactating mice). Data represents individual mice and mean +/− s.e.m. Data analyzed by Students unpaired t-test. ns (not significant), ∗p < 0.05.

Extended Data Fig. 4.

Lactating and non-lactating groups exhibit similar baseline levels of calcium signal in fiber photometry experiments. Change in calcium signal in the non-lactating and lactating groups of mice in the 60 s following presentation of a food pellet following an overnight fast. Both groups of mice had a similar inhibitory response to food presentation, indicating similar baseline levels of calcium signals in the two groups (n = 8 mice for both groups). Data represents individual mice and mean +/− s.e.m. Data analyzed by Mann Whitney test. ns (not significant).

Extended Data Fig. 5.

Lactating mice exhibit an increased inhibitory response to chow compared to non-lactating mice. a: Change in calcium response shown in delta F/F0 in the 60 s following the presentation of chow in mice prior to lactation and the same mice during lactation. b: Change in calcium response shown in delta F/F0 in the 60 s following consumption of chow in mice prior to lactation and the same mice during lactation. c and d: Change in calcium response in non-lactating and lactating mice (between subjects comparison) in the 60 s following the presentation of chow (c) or the consumption of chow (d). Data points represent individual mice. Data in a and b analyzed by Wilcoxon test. Data in c and d analyzed by Mann–Whitney test. ∗p < 0.05, ∗∗p < 0.01.

Extended Data Fig. 6.

Fiber photometry signals in non-lactating and lactating mice aligned to the start of eating. a and b: Z-scored calcium signal in the 60 s prior and 60 s following the start of eating before lactation and during lactation in the same mice (n = 8 mice per group). c: Latency to start eating during photometry experiments following a 10-h fast prior to lactation and during lactation in the same mice (n = 8 non lactating mice and n = 8 lactating mice). Lactating mice have a significantly reduced latency to eat during testing relative to prior to lactation. d: Total time spent eating in the same mice before lactation and during lactation in the 60 s following the start of eating. Both non-lactating (prior to lactation) and lactating mice eat for a similar amount of the time in the first 60 s of food consumption. Data points represents individual mice and mean +/− s.e.m. Panels c and d analyzed by Students paired t-test. n.s. (not significant), ∗∗∗p < 0.005.

Extended Data Fig. 7.

Non-lactating and lactating mice have a similar response to the drop of a food pellet in the ad libitum state. Change in calcium signal in AgRP neurons from non-lactating (baseline measurements prior to lactation) and lactating mice that are fed ad libitum (n = 8 mice per group). No difference in the response to the food pellet was detected between the two conditions. Data represents individual mice and mean +/− s.e.m. Data analyzed by Mann Whitney test. n.s. (not significant).

Extended Data Fig. 8.

Non-lactating mice show no difference in calcium response at baseline and in the time-matched “lactation” time-point. Left panel: Change in calcium signal in the 60 s following the introduction of a food pellet (after a 10 h fast) in the same group of mice at baseline and during the lactational period in time-matched mice. No difference was detected between the inhibitory response to food presentation at either time-points. Right panel: Change in calcium signal in the 60 s following the initiation of eating in the same mice at baseline and during the time-matched lactational period (in non-lactating animals). No difference was detected between the inhibitory response to food consumption at either time-points. N = 5 mice per group. Data analyzed with Mann–Whitney test. Data represents individual mice and mean +/− s.e.m. n.s. (not significant).

Extended Data Fig. 9.

Lactating and non-lactating mice exhibit a similar inhibitory response to the first 20 s of eating. a: Graph of the average calcium signal from non-lactating and lactating mice aligned to the presentation of the PB chip (time-point 0). b: change in calcium signal in non-lactating and lactating mice in the 20 s following the presentation of the PB chip. No difference was detected between the non-lactating and lactating mice during this time-period. c: Change in calcium signal in non-lactating and lactating mice in the first 20 s of PB chip consumption (a time-period where lactating and non-lactating mice eat for the same amount of time). No difference was detected between non-lactating and lactating mice in the first 20 s of food consumption. Data points represent the average (+/− s.e.m.) of all mice in panel a and individual mice in panels b and c. Data analyzed by Mann Whitney test. n.s. (not significant). n = 7 non-lactating mice and n = 8 lactating mice.

Extended Data Fig. 10.

Example traces of signal channel and isosbestic control channel during pup reintroduction experiments. Top panel: Example trace of signal channel (465 nm) and isosbestic control channel (410 nm) in a lactating mouse during pup reintroduction following 16 h of deprivation of pups from the dam. Bottom panel: Example trace of signal channel (465 nm) and isosbestic control channel (410 nm) in a non-lactating mouse following introduction of the pups to the mouse.

Extended Data Fig. 11.

Inhibition of AgRP neurons has no effect on feeding in non-lactating and lactating ad libitum fed mice. Food intake (2-hr food intake) in non-lactating and lactating mice following saline or CNO injections. No difference in food intake was detected between saline and CNO injections in both the non-lactating and lactating mice. Data represents mean +/− s.e.m. n.s. (not significant). n = 10 lactating mice and n = 10 non-lactating mice.

Extended Data Fig. 12.

Inhibition of AgRP neurons does not alter meal number in non-lactating or lactating mice. Meal number following injections of saline or CNO to non-lactating and lactating mice (fasted for 90 min prior to testing). CNO did not alter meal number in either group of mice. Data represents individual mice and mean +/− s.e.m. Data analyzed by 2-way ANOVA. n.s. (not significant).

Data availability

Data will be made available on request.