Development of activity-based anorexia requires PKC-δ neurons in two central extended amygdala nuclei

SUMMARY

Anorexia nervosa (AN) is a serious psychiatric disease, but the neural mechanisms underlying its development are unclear. A subpopulation of amygdala neurons, marked by expression of protein kinase C-delta (PKC-δ), has previously been shown to regulate diverse anorexigenic signals. Here, we demonstrate that these neurons regulate development of activity-based anorexia (ABA), a common animal model for AN. PKC-δ neurons are located in two nuclei of the central extended amygdala (EAc): the central nucleus (CeA) and oval region of the bed nucleus of the stria terminalis (ovBNST). Simultaneous ablation of CeA^PKC-δ and ovBNST^PKC-δ neurons prevents ABA, but ablating PKC-δ neurons in the CeA or ovBNST alone is not sufficient. Correspondingly, PKC-δ neurons in both nuclei show increased activity with ABA development. Our study shows how neurons in the amygdala regulate ABA by impacting both feeding and wheel activity behaviors and support a complex heterogeneous etiology of AN.

Graphical Abstract

In brief

Schnapp et al. identify neurons in the amygdala as essential for the development of activity-based anorexia (ABA), a common animal model for anorexia nervosa (AN). The activity of these neurons increases with ABA development, further indicating their importance in regulating disrupted behaviors associated with AN.

INTRODUCTION

Anorexia nervosa (AN) is a prevalent eating disorder that has the highest mortality rate of any psychiatric disorder.^1,2 It is characterized by abnormal eating behaviors, self-starvation, intense fear of gaining weight, and excessive exercise.³ Surprisingly, while neurons in many brain regions have been identified to regulate eating behaviors,^4–6 those tested so far either have no effect or only a partial effect in preventing AN in animal models.^7–9 Thus, the neural mechanism underlying AN development remains to be determined.

Interestingly, AN is often co-diagnosed with other psychiatric and emotional disorders, such as depression, anxiety, and obsessive-compulsive disorder.^2,3,10–14 These characteristics suggest that extensive interaction between the neural circuits regulating eating behavior and the neural circuits regulating emotion might exist to control AN development. Consistent with this, AN has been associated with elevated neural activity in the amygdala,^15–17 the most well-established brain region for emotional control.¹⁸ However, how neurons in the amygdala regulate the development of AN is still unclear.

Our previous work revealed that neurons expressing protein kinase C-delta (PKC-δ) in two distinct nuclei of the central extended amygdala (EAc), the central nucleus of the amygdala (CeA) and the oval region of the bed nucleus of the stria terminalis (ovBNST), are required in diverse anorexigenic signals and suppress food intake when acutely activated.^19,20 Here, we investigate how this particular subpopulation of neurons in the CeA and ovBNST may regulate the development of AN using a common animal model of AN, activity-based anorexia (ABA), in which rodents develop life-threatening self-starvation and hyperactivity ten-dencies when exposed to a restricted feeding schedule in combination with ad libitum access to a running wheel.^21–23

RESULTS

CeA^PKC-δ and ovBNST^PKC-δ neurons regulate ABA development

To investigate how CeA^PKC-δ or ovBNST^PKC-δ neurons might play a role in the development of AN, we assessed how mice respond to the ABA paradigm when this subpopulation of neurons was ablated. We first focused only on female mice because human AN is observed and diagnosed disproportionally more in females²⁴ and because female mice more consistently developed ABA in our mouse line and paradigm (Figure 5).

Figure 5. Dual ablation of EAc^PKC-δ neurons prevents ABA in both male and female mice (A and B) Survival analysis of WT-FRW mice.

(A) and of Cre-FRW mice (B) (WT: p = 0.0045, male n = 10, female n = 10; Cre: p = 0.93, male n = 7 and female n = 10; log rank test).

(C and D) Left: population mean line plots of BW loss across days of food restriction. Right: weight loss measurement on the day of removal from the experiment, either when the ABA criteria are reached or after day 10. A gray dashed line at 20% weight loss indicates the point when mice have developed ABA and need to be removed from the experiment to prevent death. A sudden change in the line plots for female WT mice (i.e., days 3–4) is due to the removal of half of the initial number of mice from the experiment in 1 day, indicated by a dashed section in the line. Boxplot averages are based on days when more than one WT sample remained in the experiment. See Figure S10 for individual-sample data plots. Unpaired t tests, *p < 0.05, **p < 0.01, ***p < 0.001. Line plot error bars represent mean ± SEM. Boxplot error bars show the data range ± 1.5 times beyond the IQR.

(E and F) Left: total food intake during each day’s feeding period. Right: average food intake across days 2–6 of the experiment.

(G and H) Top: total wheel revolutions per day during baseline before food restriction (food ad libitum, days −5 to −1) and during food restriction (days 1–10).

Bottom: average total wheel activity before and during initial food restriction (days 1–5). An arrow indicates the day when food restriction (FR) was enforced. Two-way ANOVA with Tukey HSD post hoc analysis was used for boxplots comparing male and female mice before and during food restriction.

(I and J) Total wheel revolutions during the light period (ZT0–ZT12). Left: population mean line plots across days of food restriction. Right: average total revolutions during initial food restriction (days 1–5). See Figure S10 for individual-sample data plots and Figure S11 for additional wheel activity parameters.

To ablate the PKC-δ neurons, we stereotaxically injected PKC-δ-Cre mice with a Cre-dependent adeno-associated virus (AAV) expressing caspase (AAV2-FLEX-taCasp3)²⁵ bilaterally into either the CeA or ovBNST or into both simultaneously (CeA + ovBNST) (Figures 1A and 1B). We injected wild-type (WT) mice with the same virus for use as controls. Following behavior experiments, we evaluated and confirmed ablation success for all samples with immunohistochemistry and cell quantification of the labeled neurons (Figures 1A–1D and S1B). Comparison of body weight pre and 3 weeks post virus injection showed that ablation itself did not significantly impact body weight (Figure S1A). Interestingly, even though acute silencing of CeA^PKC-δ and ovBNST^PKC-δ neurons increased food intake in previous studies,^19,20 chronic ablation in either or both nuclei did not increase total daily food intake (Figures S1C and S1F). Additionally, there was no significant difference in total daily wheel activity or body mass in mice with PKC-δ neuron ablation compared with WT mice before any food restriction (Figures S1D, S1E, S1G, and S1H). These baseline data indicate that EAc^PKC-δ neuron ablation does not induce significant changes in body weight, food intake, or wheel activity when the mice are in matching conditions with ad libitum access to a wheel and food.

Figure 1. Simultaneous elimination of PKC-δ neurons in the CeA and ovBNST prevents development of ABA.

(A and B) Left: diagram illustrating bilateral injections of a Cre-dependent virus expressing caspase into the CeA (A) and ovBNST (B) regions of mouse brains. Right: representative histology of WT and PKC-δ-Cre. BLA, basolateral amygdala; AC, anterior commissure.

(C and D) Quantification: average density of CeA^PKC-δ (C) and ovBNST^PKC-δ (D) neurons (number of neurons per micron × 10⁴). Unpaired t tests, ***p < 0.001. CeA-WT, n = 8; CeA-Cre, n = 20; ovBNST-WT, n = 12; ovBNST-Cre, n = 19 mice. Error bars show the data range beyond the interquartile range (IQR), reaching 1.5 times the IQR from the first and third quartiles or to the minimum and maximum values if within this range. Any points beyond this range are outliers.

(E) Timeline of the ABA protocol (created with BioRender).

(F) Survival analysis of all experimental mice. Log rank test with respect to FR (n = 18): FRW WT/no ablation (n = 10, p < 0.001), FRW CeA^PKC-δ ablation (n = 5, p = 0.0084), FRW ovBNST^PKC-δ ablation (n = 5, p = 0.050), and FRW ovBNST^PKC-δ + CeA^PKC-δ ablation (n = 10, p = 0.50). FR, food restriction (includes mice from each ablation group); FRW, food restriction with wheel. All samples are females.

Three weeks post surgery, we applied the ABA assay,²⁶ in which food was given for a limited time each day for 10 consecutive days (Figure 1E). “Food restriction” (FR) included all mice that did not have a wheel; these mice are a standard control group for the ABA assay because they have been shown to drop body weight but survive the limited feeding schedule.²² In contrast, but as expected, all WT mice in the “food restriction with wheel” (FRW) group developed ABA within 10 days (Figure 1F). The criterion for ABA was the point when mice weighed 20% less than their baseline body weight (as determined before food restriction) for 2 consecutive days. While ablation of PKC-δ neurons in either the CeA or ovBNST offered some level of protection from ABA for FRW mice, simultaneous ablation in both regions (CeA + ovBNST) was required for consistent prevention of ABA (Figure 1F). FRW mice with this dual ablation showed a significant difference in survival compared with WT-FRW mice and no significant difference compared with FR controls (Figure 1F). We did not find a correlation between CeA^PKC-δ and ovBNST^PKC-δ neuron expression levels after ablation and day of ABA development (Figure S2). These data suggest that bilateral dual ablation of PKC-δ neurons in two nuclei of the EAc—the CeA and the ovBNST—has a compounding effect to regulate development of ABA.

Dual ablation of CeA^PKC-δ and ovBNST^PKC-δ neurons prevents the reduced body weight and food intake in ABA

To determine how the mice with dual ablation of CeA^PKC-δ and ovBNST^PKC-δ neurons survive the ABA conditions, we first assessed body weight change across days of the experiment as well as corresponding food intake. Consistent with a previous study,²⁷ our baseline data showed that the presence of a running wheel with ad libitum food did not decrease weight loss across days (Figure S1E). FR mice, however, did decrease their body weight across days but not typically to the life-threatening point requiring removal from the experiment (Figures 1F, 2A, and 2B). Therefore, the FR group functioned as a control in our study. As expected, there were significant differences in body weight loss between WT-FRW mice (no ablation) and their respective FR controls (Figure 2A). However, in contrast, Cre-FRW and -FR mice (dual ablation) followed almost identical trends across all 10 days (Figure 2B). Correspondingly, food intake was disrupted and irregular in WT-FRW mice, decreasing across days as the mice lost weight (Figure 2C), while Cre-FRW food intake very closely matched that of the respective Cre-FR control group (Figure 2D). Individual-sample data plots further demonstrate that mice with dual ablation were less susceptible to developing ABA (Figures S3A and S3B). These data show how FRW mice with CeA^PKC-δ + ovBNST^PKC-δ neuron ablation behaved more similarly to their respective FR controls, with adaptive increases in food intake and eventual weight stabilization, thus demonstrating increased “resilience”^28,29 to ABA.

Figure 2. Characteristics of ABA development are mitigated with dual ablation of EAc^PKC-δ neurons.

(A and B) Left: population mean line plots of body weight loss across days of food restriction. Right: weight loss measurement on day of removal from the experiment, either when the ABA criteria are reached or after day 10. A gray dashed line at 20% weight loss indicates the point when mice have developed ABA and need to be removed from the experiment to prevent death. A sudden change in the line plots for WT mice (i.e., days 3–4) is due to the removal of half the initial number of mice from the experiment in 1 day, indicated by a dashed section in the line. WT-FR, n = 7; WT-FRW, n = 10 (A); Cre-FR, n = 7; WT-FRW, n = 10. All samples are females. Boxplot averages are based on days when more than one WT sample remained in the experiment. See Figure S3 for individual sample data plots. Unpaired t tests, *p < 0.05, **p < 0.01, ***p < 0.001. Line plot error bars represent mean ± SEM. Boxplot error bars show the data range ± 1.5 times beyond the IQR.

(C and D) Left: total food intake during each day’s feeding period. Right: average food intake across days 2–6 of the experiment (restriction began after the day 1 feeding period).

(E–G) Total wheel revolutions for WT-FRW mice and Cre-FRW mice per day (E), throughout the light period (ZT0–ZT12) (F) and dark period (ZT12–ZT0) (G). Left: population mean line plots during baseline (food ad libitum, days −5 to −1) and food restriction (days 1–10). Right: average total revolutions before (days −5 to −1) and during initial food restriction (days 1–5 for the total daily and light period, days 1–6 for the dark period). Two-way ANOVA with Tukey HSD post hoc analysis was used for boxplots comparing WT and Cre before and during food restriction. An arrow indicates the day when food restriction (FR) was enforced. See Figures S3 and S5 for individual-sample data plots.

(H and I) Total wheel revolutions during food anticipatory activity (FAA; 4 h preceding presentation of food) (H) and feeding periods (I). Left: population mean line plots for each day of food restriction. Right: average total revolutions during initial food restriction (days 1–5 or 6, respectively).

Dual ablation of CeA^PKC-δ and ovBNST^PKC-δ neurons prevents the light-period hyperactivity in ABA

Another crucial element for the development of ABA is increased wheel activity (i.e., hyperactivity).^30,31 Daily running wheel activity in WT (no ablation) mice was not significantly different compared with Cre mice (dual ablation) before food restriction began (Figure 2E). However, once food was restricted, WT-FRW mice significantly increased their daily wheel activity, while Cre-FRW mice did not have significant alterations in daily wheel activity (Figure 2E and S3C–S3E). WT-FRW mice demonstrated hyperactivity across days of food restriction until terminal removal, whereas Cre-FRW mice tended to show relatively steady wheel activity, suggesting adaptation to food restriction parameters (Figure S3C). Single ablation of either CeA^PKC-δ or ovBNST^PKC-δ neurons did not consistently prevent any of the key ABA phenotypes, suggesting that one of the two nuclei alone does not function with respect to one behavior more than the other (Figures S4A–S4C).

To investigate details of the wheel activity, we examined different time frames for each day: light period, dark period, food-anticipatory activity period (FAA; defined as 4 h preceding food presentation^32,33), and the limited feeding period. WT-FRW mice demonstrated extreme wheel hyperactivity during the light period (“rest/sleep” time), including FAA, but this hyperactivity was absent in Cre-FRW mice (Figures 2F and 2H; see Figures S5A and S5B for individual-sample plots). Additionally, Cre-FRW mice demonstrated a trending decrease of wheel activity during the dark period (“awake” time) across days of food restriction (day 2 vs. day 10, p = 0.00019, unpaired t test), suggesting a level of arousal adaptation that corresponds with the new feeding schedule (Figures 2G and S3C). Consistent with previous results,³¹ time-series data show that WT-FRW mice had strong wheel activity disruptions and abnormalities—with hyperactivity even during the light period (“rest/sleep” time)—1 or 2 days before terminal removal (Figure S5E). In contrast, Cre-FRW mice showed consistent patterns of day/night wheel activity, with moderate activity during the dark period and minimal activity during the light period (Figure S5F). We did not see a significant difference in the wheel activity during feeding time (Figures 2I and S5D).

Due to the strong prevention of the ABA wheel hyperactivity phenotype that occurred with EAc^PKC-δ neuron ablation, we assessed how activating or silencing EAc^PKC-δ neurons influences wheel activity itself. Using chemogenetic methods, we discovered that activating EAc^PKC-δ neurons in fed mice significantly increased the number of revolutions and time active on the wheel compared with control mice during the light-cycle period (Figures S6A–S6D). Correspondingly, silencing EAc^PKC-δ neurons caused a trend in decreased wheel activity (Figures S6E–S6G). These data align with the wheel hyperactivity phenotype that develops with ABA when the neurons are present and, conversely, is eliminated with ablation.

Together, these data demonstrate that Cre-FRW mice with dual ablation of CeA and ovBNST PKC-δ neurons survived ABA conditions primarily by preventing a wheel activity increase upon the introduction of the new feeding schedule as well as by gradually increasing food intake across days.

ABA causes more EAc^PKC-δ neurons to be activated in response to food

Activation of CeA^PKC-δ or ovBNST^PKC-δ neurons suppresses food intake;^19,20 thus, we hypothesized that FRW mice have increased activity of CeA^PKC-δ and ovBNST^PKC-δ neurons after ABA development. To investigate the involvement of the PKC-δ neuron activity in the four different brain regions (bilateral CeA and ovBNST) in mice developing ABA, we monitored c-Fos expression in WT-FRW mice and their respective controls, WT-FR mice, in response to food intake during the ABA paradigm. Interestingly, there were a few FRW mice in this cohort that did not develop ABA within the 10-day experiment, but they were still collected for analysis (FRW non-ABA). On the day when FRW mice reached more than 20% body weight loss from baseline (FRW-ABA), they were perfused for 90 min after presentation of food. FR (non-ABA) mice were similarly collected on days comparable with the FRW mice. The FRW (non-ABA) mice were collected on day 10, the last day of the experiment. Double immunostaining for c-Fos⁺ and PKC-δ⁺ neurons revealed that FRW (ABA) mice had significant increases in both CeA^PKC-δ and ovBNST^PKC-δ neurons expressing c-Fos compared with FR or FRW (non-ABA) mice (Figures 3A, 3B, 3E, and 3F). This increase was observed in both the right and left hemispheres of the CeA and ovBNST, suggesting bilateral importance (Figures S7A and S7B). The total number of c-Fos-expressing neurons in these nuclei, however, was not significantly different (Figures 3C and 3G). Additionally, linear regression analysis indicates a significant negative correlation between food intake and c-Fos⁺ PKC-δ⁺ neurons across groups (Figures 3D and 3H), which aligns with the significantly reduced food intake in FRW (ABA) mice compared with FR (non-ABA) and FRW (non-ABA) mice (Figure 3I) on the day of terminal removal. ABA resistance and susceptibility are clear when comparing the body weight loss and wheel revolutions during days of food restriction for mice in the FRW condition (Figures 3J–3M). The results here are consistent with the previously discovered function of EAc^PKC-δ neurons in suppressing food intake when activated^19,20 and further support the involvement of these neurons in regulating the development of ABA.

Figure 3. Activity of PKC-δ neurons in both the CeA and ovBNST is implicated in ABA feeding behaviour.

(A–C and E–G) Representative histology and quantification of Fos-like immunoreactivity in the CeA (A–C) and ovBNST (E–G) of WT mice in the “food restriction” condition (FR non-ABA; n = 7), ABA-resistant mice with food restriction and wheel (FRW non-ABA; n = 3), and ABA-susceptible mice with food restriction and wheel (FRW ABA; n = 6). One-way ANOVA with Tukey HSD post hoc test. Arrows point to examples of c-Fos and PKC-δ co-expression. All samples are females. Unpaired t tests, *p < 0.05, **p < 0.01, ***p < 0.001. Boxplot error bars show the data range ± 1.5 times beyond the IQR. See Figure S7 for right and left hemispheres analyzed separately.

(D and H) Relationship between c-Fos expression level in CeA^PKC-δ neurons (D) and ovBNST^PKC-δ neurons (H) and food intake on the day of removal from the experiment across groups (BW, body weight); simple linear regression analysis.

(I) Comparison of food intake between groups on the day of removal from the experiment.

(J and K) BW loss on the day of removal (J) and across days of food restriction (K) for ABA-susceptible and ABA-resistant mice.

(L and M) Total wheel activity on the 2 full days preceding removal (L) and daily wheel activity across days of food restriction (M) for ABA-susceptible and -resistant mice.

To further investigate the in vivo dynamics of EAc^PKC-δ neurons as mice develop ABA, we injected PKC-δ-Cre mice with a Cre-dependent AAV expressing a genetically encoded calcium indicator (GCaMP6s).³⁴ We targeted both CeA^PKC-δ and ovBNST^PKC-δ neurons (unilateral and opposing sides) and simultaneously implanted fiberoptic cannulas (Figure S8). We collected fiber photometry calcium fluorescence data as mice went through the ABA paradigm (Figure 4A) and compared data from three conditions: “wheel with food ad libitum” (Wheel; before food restriction), “food restriction with wheel” (FRW; after 1-day food restriction), and day when the FRW mouse reached ABA criteria (FRW-ABA). To determine how the EAc^PKC-δ neuron activity changed with feeding, we recorded before and after Zeit-geber Time (ZT) 12, the start of the dark (active) cycle and the time when which mice begin their feeding. The average Z-scored change in fluorescence 15 min before ZT12 and 30 min thereafter revealed elevated calcium fluorescence levels in the FRW-ABA condition relative to both the FRW and Wheel groups during the dark (feeding) period specifically (Figures 4B–4E). The fluorescence change was observed in both CeA and ovBNST nuclei, indicating increased activity of EAc^PKC-δ neurons. Therefore, the in vivo calcium imaging data align with the c-Fos data, further supporting the hypothesis that EAc^PKC-δ neurons play a role in development of ABA.

Figure 4. Activity of PKC-δ neurons in both the CeA and ovBNST is elevated during the feeding period as ABA develops.

(A) Schematic of the fiber photometry imaging and ABA protocol. Bold text signifies the day of calcium imaging: wheel with food ad libitum, after 19 h of food restriction (FRW), and upon reaching ABA criteria (FRW-ABA). The day when the mice developed ABA differed but was no less than after 3 days of FR and no more than after 10 days of FR. All mice were female (n = 4–5).

(B–E) Activity of CeA^PKC-δ (B and D) and ovBNST ^PKC-δ (C and E) neurons before and during the feeding period. Food was presented to FRW and FRW-ABA mice at 0 min when the dark period begins, as indicated by shade along the x axis. Shown is average Z-scored calcium fluorescence before and after ZT12 (B and C) and mean Z score for the 30 min following ZT12 (D and E). Repeated measures one-way ANOVA with Tukey’s multiple-comparisons test, *p < 0.05. Error bars show the data range beyond the IQR, reaching 1.5 times the IQR from the first and third quartiles or to the minimum and maximum values if within this range.

CeA^PKC-δ and ovBNST^PKC-δ neurons function in combination

Our previous results showed that optogenetic activation of the CeA^PKC-δ neurons or ovBNST^PKC-δ neurons suppresses food intake, while chemogenetic silencing of these neurons can block the anorexia induced by the corresponding anorexigenic signals that these neurons mediate.^19,20 Thus, we tested whether bilateral silencing of the PKC-δ neurons in one of the two nuclei would prevent the feeding suppression caused by activation of the other nuclei. We stereotaxically injected AAV-DIO-ChR2-EYFP in one of the nuclei as well as AAV-DIO-hM4Di-mCherry in the other nucleus, both bilaterally (Figure S9). Consistent with our previous studies,^19,20 chemogenetic inhibition of ovBNST^PKC-δ neurons increased food intake compared with saline (Figure S9B), while inhibiting CeA^PKC-δ neurons had a non-significant trend of increased food intake (Figure S9D). When PKC-δ neurons in one of the nuclei were chemogenetically silenced, the simultaneous optogenetic activation of PKC-δ neurons in the other nucleus caused a trend of increase in food intake (right side of graphs, saline vs. clozapine-N-oxide [CNO]), but this was still significantly lower than the no activation control (CNO + activation vs. CNO + non-activation). These results suggest that activation of either CeA^PKC-δ neurons or ovBNST^PKC-δ neurons alone is sufficient to suppress food intake, even in the absence of activity in the other. This observation is consistent with our results showing that single ablation only had a mild attenuation on ABA, while dual ablation prevented ABA.

To further examine how CeA^PKC-δ and ovBNST^PKC-δ neurons might exert their function at the circuit level, we examined downstream projections from the specific PKC-δ neurons in each of these discrete regions simultaneously using Cre-dependent viral tracing techniques (Figure S10A). We alternated between EYFP and mCherry to account for innate differences in fluorescence and found fluorescence largely at the same locations and with similar intensity for both the CeA and ovBNST (Figure S10B). CeA^PKC-δ neurons displayed their strongest projections, in order, at the medial part of the central amygdala (CeM), extended BNST region, and ventrolateral BNST (vlBNST) (Figure S10B). ovBNST PKC-δ neuron projections were similar but with vlBNST the strongest, followed by extended BNST and CeM. Both CeA^PKC-δ and ovBNST ^PKC-δ neurons showed minor terminal fluorescence at the parasubthalamic nucleus (PSTh), ventrolateral medial reticular formation (vlmRt), and lateral parabrachial nucleus (LPB) (Figure S10B). Overall, these data demonstrate how GABAergic CeA^PKC-δ and ovBNST^PKC-δ neurons make similar contributions to intra-circuit inhibitory connections within the EAc (Figure S10C) as well as long-range projections to multiple brain regions, suggesting that removal of one would have some disinhibitory effect, but not as great as if both were removed. We further assessed and confirmed functional connectivity points within the nuclei using patch-clamp electrophysiology with optogenetics (Figure S11).

Dual ablation of CeA^PKC-δ and ovBNST^PKC-δ neurons prevents ABA in both male and female mice

Previous literature has shown that behavior and survival in the ABA paradigm can vary depending on sex.^35,36 Our WT-FRW mice did indeed demonstrate sexual divergence: the females consistently developed ABA quicker and lost weight more drastically than males (Figures 5A, 5C, and S12A). However, in contrast to these WT-FRW mice without ablation, Cre-FRW mice with dual ablation did not demonstrate sexual divergence. Survival analysis indicates no significant difference in probability of developing ABA, while the trend of decrease in body weight was similar between the sexes (Figures 5B, 5D, and S12B).

While average food intake did not significantly differ between sexes in each group (Figures 5E, 5F, S12C, and S12D), there was variance in the wheel activity (Figures 5G, 5H, S12E, and S12F). Consistent with previous studies,^37,38 there was a slight increase in daily wheel activity in females compared with males for both groups before food restriction started (Figures 5G, 5H, S12E, and S12F). However, while the male and female WT-FRW mice increased their differences in wheel activity upon food restriction, the difference was minimized upon food restriction for Cre-FRW mice. Accordingly, the wheel activity during the light period was significantly different between male and female WT-FRW mice but not with Cre-FRW mice (Figures 5I, 5J, S12G, and S12H). FAA, specifically, was also significantly different between sexes for WT-FRW mice but not for Cre-FRW mice (Figures S13A and S13B). While dark-period wheel activity was increased in females of both experimental groups compared with the respective males for the first 5 days of food restriction, the activity for Cre-FRW females decreased and plateaued to become aligned with males for the latter half of food restriction days (Figures S13C and S13D). Additionally, there was no difference in the wheel activity for the sexes of either group during the feeding period (Figures S13E and S13F). Overall, decreased wheel activity during the light period indicates decreased susceptibility to develop ABA, which is seen in a portion of the male WT-FRW mice as well as in a majority of both male and female Cre-FRW mice with dual ablation. Therefore, these data suggest that ablation of EAc ^PKC−δ neurons offers a level of resilience to ABA development, particularly regarding the wheel hyperactivity phenotype, in both male and female FRW mice that typically occurs only in male WT-FRW mice.

DISCUSSION

Although functional alterations of brain regions associated with AN have been observed in human neuroimaging studies,^15–17 and neurons regulating some of the ABA phenotypes have been identified in animals studies,^{7,9,17,29,33,39–43} details of neural mechanisms that might cause AN are still being uncovered.^22,23,44

It has long been known that AN is comorbid with emotional conditions and, therefore, that development of the disorder may be attributed to the neural circuits that control emotions, especially those in the amygdala regions.^45,46 Consistent with this theory, neuroimaging studies in humans have suggested that the function of the amygdala or amygdala-associated mesolimbic brain regions is altered in those with AN.^15–17 However, whether neurons in the amygdala regulate ABA was not known. In fact, only recently have neurons in the amygdala, especially the CeA and BNST regions, been demonstrated to regulate eating behavior and eating suppression in anorexigenic conditions.^{19,20,47–52} CeA^PKC-δ neurons are preferentially activated by anorexigenic signals such as satiation, visceral malaise and nausea, and bitter taste.¹⁹ On the other hand, the ovBNST^PKC-δ neurons are preferentially activated by anorexia signals related to inflammation or sickness, such as inter-leukin-1b(IL-1b), lipopolysaccharides (LPSs), and tumor necrosis factor alpha (TNF-a).²⁰ Here, we demonstrated that single ablation of CeA^PKC-δ neurons or ovBNST^PKC-δ neurons had only a mild effect in attenuating ABA, while dual ablation of the PKC-δ neurons simultaneously in these two nuclei prevented ABA development. Notably, all key phenotypes of ABA were attenuated to a level indistinguishable from their respective controls: life-threatening body weight loss, insufficient food intake, overall running wheel hyperactivity, and increased FAA. Thus, our results suggests that ABA development involves contributions from a combination of multiple anorexigenic factors rather than a single factor.

Limitations of the study

Increased wheel activity during the light period, including FAA, is prevented with dual ablation of the PKC-δ neurons, suggesting that these neurons might be involved in circadian rhythm^27,32 or food-entrained oscillators.^53,54 This finding aligns with specific expression of the circadian clock protein PER2 in the CeA and ovBNST⁵⁵ and changes in expression that occur with the estrous cycle,⁵⁶ thus potentially accounting for the sex differences we observed. It is still unclear whether the neural activity or some molecular mechanisms of these neurons are critical for ABA development. Future work will involve clarifying how these PKC-δ neurons and PER2 or other clock proteins may interact to regulate the circadian disruption that occurs with the ABA paradigm.

In summary, our study provides evidence that malfunction of neural circuits in the amygdala contributes to ABA development and demonstrates that they may be a more relevant and robust therapeutic target in the treatment of AN. We also propose a multiorigin possibility for ABA development, which suggests that treating AN requires consideration of combining multiple factors or targeting multiple brain regions.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Haijiang Cai (haijiangcai@arizona.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• Data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mice

We crossed PKC-δ-Cre C57BL/6 mice (generated by David Anderson’s lab⁵⁷) with wildtype (WT) C57BL/6 mice from the Charles River Laboratory to get PKC-δ-Cre and WT littermate mice; the same mouse line as was used in previous studies from our lab.^19,20 The genotype of offspring that we generated and used from these mice were determined by PCR of genomic tail DNA. Adult male and female mice were used. The average weight of an adult mouse was 21g or 27g for females and males, respectively. Stereotaxic survival surgery was performed when mice were 2–3 months old. Littermates of the same sex were randomly assigned to experimental groups, and also depending on genotype (WT vs. PKC-δ-Cre). Behavior experiments were done before mice reached 24 weeks of age. Mice were group housed until the activity-based anorexia (ABA) or running wheel experiments began. All mice were housed on a 12-h light (4 a.m., ZT0)/dark (4 p.m., ZT12) cycle, with ad libitum access to water and rodent chow, except for during the ABA experiment food restriction and food intake tests. The ambient temperature and humidity of the colony room were monitored daily and kept at 68–74°F and 30–70%, respectively, according to the Guide for Care and Use of Laboratory Animals. All animal care and experimental procedures were in accordance with ethical regulations, conducted according to the National Institutes of Health guidelines for animal research, and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Arizona.

METHOD DETAILS

Virus and tracer

For Cre-dependent ablation, we used rAAV2-FLEX-taCasp3-Tevp, a virus generated by Dr. Nirao Shah’s lab.²⁵ For Cre-dependent anterograde tracing, we used rAAV2-EF1a-DIO-EYFP and rAAV5-hSyn-DIO-mCherry, generated by Dr. Karl Deisseroth’s lab and by Dr. Bryan Roth’s lab, respectively. For fiber photometry calcium imaging, we used rAAV5-Ef1a-DIO-GCaMP6s from Dr. Karl Deisseroth’s lab. For optogenetic activation, we used rAAV2-EF1a-DIO-ChR2-EYFP generated by Dr. Karl Deisseroth’s lab. For chemogenetic silencing and activation, we used rAAV5-hSyn-DIO-hM4Di-mCherry and rAAV5-hSyn-DIO-hM3Dq-mCherry, respectively; both generated by Dr. Bryan Roth’s lab. These viral constructs were deposited and packaged into viral vectors either at the University of North Carolina (UNC) Viral Vector Core or Addgene at a titer of 4–6×10¹² genome copies per mL. Upon arrival to our lab, the virus stocks were aliquoted and stored at −80°C until use.

Stereotaxic survival surgery

All mouse surgeries were performed using aseptic techniques with a stereotaxic frame (Model 1900 Stereotaxic Alignment System, Kopf Instruments). Injection coordinates (in mm) relative to midline, bregma, and skull surface at bregma were followed as (x, y, z): ovBNST (±1.13, +0.30, −4.10) and CeA (±2.85, −1.40, −4.70). Viruses were microinfused through a pulled-glass micropipette with 20–50 μm tip outer diameter connected with a Nanoliter Injector (Nanoliter 2010, World Precision Instruments) at a rate of 10–20 units per minute (1 unit = 4 nL). After each injection, the micropipette was left in place for 5 min to allow for diffusion of the liquid, followed by withdrawal of 5 units at the same rate, 0.3 mm in the +z direction to remove any unwanted spread of virus. Injection volume per injection location depended on the virus: taCasp3 was 55 units (220 nL), GCaMP6S was 100 units (400 nL), hM4Di and hM3Dq was 55 units (220 nL) for ovBNST and 100 units (400 nL) for CeA, and anterograde tracer was 40 units (160 nL). The caspase and hM4Di/3Dq virus were injected bilaterally and the virus for tracing experiments and calcium imaging were injected unilaterally. For the optogenetic tests, optical ferrule fibers (200 μm in diameter) were implanted bilaterally ~0.5 mm above the injection coordinates. After ferrule fiber implantation, dental cement (C&B Metabond) was used to secure the fiber to the skull. For postoperative care, mice were given water with 1.2% Septra antibiotics for 5–7 days, as well as injected intraperitoneally with ketoprofen (5 mg/kg) daily for 3 days. Mice were allowed three weeks after virus injection surgery for recovery and viral expression before used for behavioral experiments or euthanized for tracing analysis.

Activity-based anorexia (ABA)

Stereotaxic surgery was performed on mice undergoing the ABA protocol (based on Welch, 2018²⁶) for the ablation experiment. After 2–3 weeks for recovery and gene expression, mice were individually housed, and a wheel was added to their respective home cages. Control mice without a wheel were individually housed at the same time. Food and water were given ad libitum for the next 5–7 days, while establishing a baseline of wheel activity, body weight, and food eaten. Total food and body weight of each experimental mouse was measured daily during the hour before start of light off/active period. Mice were habituated to the experimenter handling during this time, as well. MED Associates Inc low-profile Wireless Running Wheels (ENV-047) with the respective Wheel Manager Software was used for continually recording running data. After establishing the baseline and acclimating the mice to the new environmental condition for 5–7 days, we started food restriction. On the first day, mice were given ad libitum food for 5 h, starting at the beginning of the dark cycle. On the second day, mice were given ad libitum food for 4 h, still starting at the onset of the dark cycle period. From the third day until the end of the study, mice were given ad libitum food for 3 h, again starting at the beginning of the dark cycle.

Each day, mice—and their respective wheels—were transferred from their home cage with bedding to an empty cage, for the sake of measuring the food consumed most accurately. More than enough pre-weighed regular chow (NIH-31, Zeigler Bros, Inc) was provided in the empty cage (~10 g total). Water was given ad libitum, both during the feeding period and during the food restriction period. After the feeding period, mice were weighed and transferred back into their original home cages with the wheel if they had one. Red light was used in the dark room to prevent disruption to the light cycle and circadian rhythm. In another room, total food left was then measured for each sample, and calculated from what was given to determine amount eaten. The empty cages were cleaned and used again for the rest of the experiment. Ambient temperature of the room was unchanged from normal conditions.

For ethical purposes, mice were removed from the experiment when body weight loss exceeded 20% of their respective baseline weight two days in a row, when measured before given food. If the mouse’s body weight remained under 20% after the feeding period on the first day of that measurement, it was also removed. Mice were monitored on the day after reaching 20% and were removed from the experiment if they appeared to be in critical condition or reached 20% loss before group measurement time, in order to prevent unexpected deaths. Therefore, since mice were removed at various times on their second day of being below 20% weight loss, the final day of running wheel activity data was not included in analysis for all mice who developed ABA. Removal from the experiment (i.e., “terminal removal”) meant mice were given ad libitum food immediately, while the wheel was taken out and a cardboard house was placed in the home cage.

For the c-Fos experiment, food restricted mice with a wheel (FRW) were perfused 90 min after presentation of food (ZT 13:30) on the day in which they reached >20% loss from baseline body weight (i.e., ABA criteria). FRW mice that did not reach this body weight loss by end of experiment (FRW Non-ABA) were perfused in the same way, but on day 10. Food restricted mice without a wheel (FR Non-ABA) were perfused on matched days.

In vivo fiber photometry calcium imaging

Stereotaxic surgery was performed on PKC-δ-Cre mice to target both ovBNST and CeA PKC-δ neurons (unilateral on opposing sides). After GCaMP6s virus was injected (see “stereotaxic survival surgery” for virus protocol details), fiber optic cannulas were immediately implanted 0.2 mm above location of virus injection. After at least 3 weeks for recovery and gene expression, mice were habituated to individual housing, running wheel, handling, and fiber photometry set-up before any recording. For each recording session, mice were briefly put under isoflurane anesthesia (less than 2 min) to attach and remove the optic fibers to implanted cannulas. Mice were given at least 45 min to recover before a recording session started. Using a commercial fiber photometry system (R821, RWD Life Science), calcium fluorescence activity data were collected with 470 nm and 410 nm laser beams. The 410 laser was used for motion control and to account for background noise. The fluorescence was collected at 30 frames per second. The in vivo recording data were collected for 90 min while mouse was in an open-top home cage. Baseline (wheel only) data were collected 3–4 times for each mouse before starting food restriction. No significant changes were observed across days for this baseline condition and, thus, the data were combined for each mouse. ABA criteria was when mouse was at critical body weight loss of ≥18% from baseline at ZT10. All samples were female and younger than 6 months. Experimental mice were perfused, followed by brain perfusion, slicing, and imaging to confirm the EAc^PKC–δ neurons were accurately targeted with virus and optic implant.

Immunohistochemistry and histology

Immunostaining and histology analyses were performed in order to check (1) the level of ablation of PKC-δ neurons after the ABA experiments, (2) c-Fos expression in PKC-δ neurons, and (3) virus expression for the tracing experiments. All mice were deeply anesthetized with ketamine/xylazine and perfused with 20 mL PBS followed by 20 mL of 4% paraformaldehyde (PFA) in PBS. Mice in the c-Fos experiment were perfused 90 min after presentation of food. The brains were then extracted, post-fixed in 4% PFA overnight at 4°C, rinsed with PBS, and then sectioned with a vibratome (Leica, VT1000S). Brains from mice that underwent the ABA experiment were sliced at 100 mm thickness, while brains from the viral tracing were sliced at 150 mm thickness. The brain slices with the viral tracing were mounted on glass slides and imaged. The brain slices from ABA mice were stained with antibodies to tag all neurons (NeuN) and neurons expressing PKC-δ. Brain tissue slices were stained with primary antibody at 4°C overnight, in a blocking solution containing 5% donkey serum and 0.5% Triton X-100. After three rounds of 5–10 min washes in PBS with 0.1% Triton X-100 solution, the tissues were incubated in secondary antibodies in the PBS-0.1% Triton X-100 at room temperature for 1–2 h. Tissue slices were then washed for three times for 5–10 min in PBS before being mounted on glass slides. Vectashield mounting medium was added before placing coverslips on top. Imaging was done using a ZEISS AxioZoom v16 Fluorescent Microscope with Apotome 2 Structured Illumination Module for optical sectioning.

Primary antibodies used were rabbit anti-PKC-δ (Abcam, ab182126, 1:1000), guinea pig anti-NeuN (Fisher/Sigma, ABN90MI, 1:1000), and guinea pig c-Fos antibody (SYSY, 266 308, 1:5000). Secondary antibodies used were Alexa Fluor 488 donkey anti-rabbit IgG (Jackson Immuno Research Inc. 711-545-152, 1:500) and Alexa Fluor 594 donkey anti-guinea pig IgG (Jackson Immuno Research Inc. 50-194-3535, 1:500).

Wheel activity with chemogenetic activation and silencing

Stereotaxic surgery was performed on Cre and WT (control) mice to target both ovBNST and CeA PKC-δ neurons (see “stereotaxic survival surgery” for details). After 2–3 weeks for recovery and gene expression, mice were habituated to individual housing for at least a week and running wheel and handling for at least four days before the test. Baseline wheel activity was monitored, during which mice were provided with ad-libitum food and water. For activation, mice were injected with 0.1 mg/kg of Clozapine-N-oxide (CNO; Enzo life science-Biomol, BML-NS105–0005, freshly dissolved in 0.9% NaCl saline to a concentration of 0.1 mg/mL) at ZT6 (light period). For silencing, mice were injected with 5 mg/kg of CNO (dissolved in 0.9% NaCl saline to a concentration of 1 mg/mL) at ZT17 (dark period). All mice were in fed state. Wheel activity was continuously recorded, but data for analysis started 20 min after the final IP injection. All samples were female.

Food intake with chemogenetic silencing and optogenetic activation

After three days of habituation for at least 20 min each day, mice were food-deprived, with water provided ad libitum, one day before test. Mice were briefly anesthetized with isoflurane and coupled with optic fibers and Clozapine-N-oxide (CNO) IP injection (Enzo life science-Biomol, BML-NS105–0005, freshly dissolved in 0.9% NaCl saline to a concentration of 1 mg/mL) at 5 mg/kg. Saline was injected as vehicle control. After at least 25 min of recovery, optogenetic activation was performed and food intake was measured in a 20 min feeding session. The light was delivered by a blue laser (Shanghai DreamLaser: 473 nm, 50 mW. The measured power was 5 mW at the optic fiber tip before coupling with the implanted ferrule fiber. The calculated light power in the stimulated brain region is ~5 mW/mm²) just after the mice were introduced into the testing cage. To be consistent with previous experiments,^19,20 15 Hz, 10 ms light pulses were used to activate ovBNST^PKC–δ neurons, while 5 Hz, 10 ms light pulses were used to activate CeA^PKC–δ neurons. No difference in food intake were observed between male and female mice after these manipulations,^19,20 thus both male and female mice were pulled together in this experiment.

Brain slice electrophysiological recordings

Coronal brain slices were sectioned at 250 μm thickness with a vibratome (Leica, VT1000S), using the artificial cerebrospinal fluid (ACSF) containing 126 mM NaCl, 1.6 mM KCl, 1.2 mM NaH₂PO₄, 1.2 mM MgCl₂, 2.4 mM CaCl₂, 18 mM NaHCO₃, 11 mM glucose (oxygenated with 95% O₂ balanced with CO₂ for at least 15 min before use). The brain slices were then immediately transferred to NMDG-HEPES recovery solution (93 mM NMDG, 2.5 mM KCl, 1.2 mM NaH₂PO₄, 30 mM NaHCO₃, 20 mM HEPES, 25 mM Glucose, 5 mM Sodium Ascorbate, 2 mM Thiourea, 3 mM Sodium Pyruvate, 10 mM MgSO₄, 0.5 mM CaCl₂, 300–310 mOsm, titrated with 10 N HCl to adjust PH to 7.3–7.4, 32°C–34°C) to recover 15 min. Brain slices were then transferred to ACSF (room temperature) and were recorded 1 h later in an electrophysiology rig equipped with a fluorescence microscope (Olympus BX51), MultiClamp 700B and Digidata 1550A1 (Molecular Devices). The patch pipettes with a resistance of 5–8 MUΩ were pulled with P-97 Sutter micropipetter puller and filled with an intracellular solution (135 mM potassium gluconate, 5 mM EGTA, 0.5 mM CaCl₂, 2 mM MgCl₂, 10 mM HEPES, 2 mM MgATP and 0.1 mM GTP, pH 7.3–7.4, 290–300 mOsm). Recordings were sampled at 10 kHz, filtered at 3 kHz, and analyzed with pCLAMP10. For the optogenetic stimulation, a laser (Shanghai DreamLaser, 473 nm, 50 mW) was used to deliver light pulses (1–2 mW/mm² at the tip, 2 ms) to trigger action potentials in neurons expressing ChR2 or to induce IPSC in the postsynaptic neurons.

QUANTIFICATION AND STATISTICAL ANALYSIS

All WT and PKC-δ-Cre mice that were injected with the taCasp3 virus (see virus and tracer section for details) were perfused, and brains were extracted for analysis of ovBNST^PKC–δ and CeA^PKC–δ cell population (see immunohistochemistry and histology sections for details). Data from WT mice that underwent food restricted (FR) and food restricted with wheel (FRW) conditions for the c-Fos experiment were collected and analyzed in the same way. For the ovBNST, 3–4 brains sections that included anterior, middle, and posterior ovBNST regions were analyzed and averaged per animal. Similarly, for the CeA, 6–8 brain sections that included anterior, middle, and posterior CeA were analyzed per animal. The number of PKC-δ cells per area of interest (ovBNST and CeA) were quantified using the multi-point tool/cell counter plug-in in FIJI/ImageJ. Samples with partial ablation were included, as this ablation technique is not always absolutely complete, or to the same degree in the exact same parts of the region targeted (ovBNST and/or CeA). To determine and calculate the percentage of PKC-δ cells expressing c-Fos in the ovBNST and CeA, the cells simultaneously expressing PKC-δ and c-Fos were counted with the channels tool, divided by the total PKC-δ cells in the same region, and multiplied by 100.

As much of the brains as possible from the PKC-δ-Cre mice injected with tracing virus were sliced with the vibratome and imaged with the AxioZoom (see previous sections for details). The FIJI/ImageJ software was used to measure the level of fluorescence. For each region, the measurements were averaged across all slices in which they appeared. The Mouse Brain In Stereotaxic Coordinates⁵⁸ was used for reference to identify location of fluorescence. The regions were grouped and designated as follows: bed nucleus of the stria terminalis (BNST, oval and ventral-lateral; plates 29–31), extended BNST (Ext BNST, plates 32–36), extended amygdala and medial/capsular central amygdaloid nucleus (EA-CeC/M; plates 37–40), central amygdaloid nucleus (CeA, CeM, CeL, CeC; plates 41–46), parasubthalamic nucleus (PSTh; plates 46–51), ventrolateral medial reticular formation (vlmRt, plates 55–62), lateral parabrachial nucleus (LPB, plates 76–79). No difference in fluorescence was observed between male and female mice, thus sexes were combined in this analysis.

Fiber photometry calcium fluorescence data were pre-processed and analyzed using the RWD program. The photometry signal was calculated as ΔF/F = (F – F₁)/F₀. F is the target fluorescence (470). F₀ is the median during a baseline time period, which was designated to be 60 s preceding each event (lights off and food presented at ZT12). F₁ is the motion fitted 410 data curve. The Z score was calculated according to a standard calculation: Z score = (x-mean)/std, where x is DF/F at a given time point and mean and std are the mean and standard deviation of the baseline time window, respectively.

All other data were analyzed with RStudio or GraphPad Prism Software. Line plot error bars represent mean ± s.e.m. Boxplots show interquartile range (IQR), ranging from 25^th to 75^th percentile, with the bar in the box representing 50^th percentile (median). Top/bottom whiskers are the largest/smallest value within 1.5 times IQR above/below the 75^th and 25^th percentile, respectively. Outside values beyond either end of the box and error bars are outliers. Unpaired t test was used to compare two groups with one variable. Wilcoxen-Rank Sum test was used if data distribution was not normal, according to the Shapiro-Wilk test. One-way ANOVA was used to compare three or more groups with one variable, and two-way ANOVA for groups with more than one variable: both with Tukey HSD post-hoc analysis. Survival analysis plots display the Kaplan-Meier estimate of time-to-event (i.e., development of ABA) with right censoring method to account for subjects that had not developed ABA by the end time point (day 10). Log rank test was used to determine if there was a statistically significant difference in survival curves between groups. Shading around curves represent the 95% confidence intervals for the point estimates. A p value less than 0.05 was considered significant.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-PKC-δ, monoclonal	Abcam	Cat# AB182126; RRID: AB_2892154
Guinea pig anti-NeuN, polyclonal	Fisher/Sigma/Millipore	Cat# ABN90; RRID: AB_11205592
Guinea pig c-Fos, polyclonal	Synaptic Systems	Cat# 266 308; RRID: AB_2905595
Alexa Fluor 488 donkey anti-rabbit IgG	Jackson Immuno Research Inc.	Cat# 711-545-152; RRID: AB_2313584
Alexa Fluor 594 donkey anti-guinea pig IgG	Jackson Immuno Research Inc.	Cat# 706-586-148; RRID: AB_2340475
Bacterial and virus strains
rAAV2-FLEX-taCasp3-Tevp	UNC (Yang et al.)25	N/A
rAAV2-EF1a-DIO-EYFP	UNC, Deisseroth	N/A
rAAV5-hSyn-DIO-mCherry	Addgene, Roth	ID# 50459
rAAV2-EF1a-DIO-ChR2-EYFP	UNC, Deisseroth	N/A
rAAV5-hSyn-DIO-hM4D(Gi)-mCherry	Addgene, Roth	ID# 44362
rAAV5-hSyn-DIO-hM3D(Gq)-mCherry	Addgene, Roth	ID# 44361
rAAV5-Ef1a-DIO-GCaMP6s	UNC, Deisseroth	N/A
Chemicals, peptides, and recombinant proteins
Clozapine-N-Oxide	Enzo Life Sciences	Product # BML-NS105
Experimental models: Organisms/strains
Mouse: C57BL/6. Strain: PKC-δ-Cre	David Anderson (Haubensak et al.)57	N/A
Mouse: C57BL/6	Charles River Laboratory	N/A
Software and algorithms
RStudio	Posit	https://posit.co/downloads/
GraphPad Prism Software	GraphPad	https://www.graphpad.com/scientific-software/prism/
R821 Tricolor Multichannel Fiber Photometry System	RWD Life Science	https://www.rwdstco.com/product-item/r821-tricolor-multichannel-fiber-photometry-system/
FIJI-ImageJ	ImageJ	https://imagej.net/software/fiji/
Med Associates Wheel Manager	Med Associates	https://med-associates.com/product/wheel-manager-data-acquisition-software/
ZEN microscope software	ZEISS	https://www.zeiss.com/corporate/en/home.html

Highlights.

• Amygdala PKC-δ neurons contribute to both feeding and activity phenotypes of ABA

• Activity of PKC-δ neurons in both amygdala nuclei is increased after ABA development

• Food intake is suppressed when one nucleus is activated while the other is silenced

• Amygdala PKC-δ neurons are involved in the sexual divergence of ABA