Evidence of persistent hunger following dietary weight loss in mice

Summary

Over 40% of adults in the United States struggle with obesity, and nearly half have attempted weight loss, yet most efforts result in weight regain. Rodent models have been used to study the physiological drivers of this phenomenon, but it remains unclear whether they exhibit weight regain after sustained weight loss. To address this, we developed a mouse model of reversed dietary obesity (ReDO) that demonstrates susceptibility to weight regain following sustained weight loss. Male diet-induced obese (DIO) mice were calorically restricted (CR) until weight-matched to controls, then pair-fed for 0, 8, or 28 days, or chronically. Upon return to ad libitum feeding, all ReDO groups displayed persistent hyperphagia relative to controls and chronically pair-fed mice. Variability in body weight regain among pair-fed ReDO mice correlated with initial 4-week weight gain on a high-fat diet. This study introduces a mouse model for mechanistic studies of weight loss-associated hyperphagia and provides a potential predictor of weight regain.

Graphical abstract

Highlights

• •Mice exhibited hyperphagia and weight regain for up to a month following weight loss
• •The degree of initial weight gain during high-fat diet exposure predicted weight regain

Nutrition; Physiology

Introduction

Obesity exerts a tremendous public health and economic burden within the US; effective weight loss interventions are in high demand. Bariatric surgery has been an effective means of weight loss for many, yet almost 50%-80% of patients experience some degree of weight regain, with the percentage of regained weight ranging from 16 to 37%.^1,2,3 Additionally, blockbuster drugs, such as the Glp1r agonist Semaglutide, have offered an unprecedented therapeutic benefit, with many losing as much as ∼24% of body weight.⁴ Yet, there is evidence that 67% of the weight loss owed to semaglutide returns within a year of halting the treatment, despite adherence to regimen.⁵ Better understanding the physiological drivers of weight regain may give rise to therapeutic interventions that promote sustained weight loss.

Previous studies involving mouse models of diet-induced obesity (DIO) subjected to caloric restriction (CR), or a return to a chow diet, have observed that mice exhibit a long-lived obesogenic memory in the form of chronic inflammation in the periphery along with hyperphagia-driven weight regain.^6,7,8,9,10 Yet, to date, the documented assessments of weight regain-associated hyperphagia have been initiated immediately following CR, making it unclear whether the heightened hunger exhibited by ReDO mice represents either an expected response to prolonged exposure to caloric restriction, or a drive to re-establish a long-lived upwardly shifted (relative to control mice) homeostatic body weight set point. Thus, we set out to determine the longevity of weight-loss-associated hyperphagia in mice.

Results

ReDO mice exhibit increased hunger immediately following caloric restriction

We first sought to establish a model of DIO followed by ReDO in mice. To this end, we employed both male control mice and those subjected to DIO fed a high-fat diet (60% fat), with both groups maintained on their respective diets for 16 weeks. DIO mice weighed substantially more than control mice and exhibited hyperglycemia (Figures 1A and 1B). DIO mice were then fed either a 40% caloric restriction diet (reversed dietary obesity, ReDO) until their aggregate body weight was matched to that of control mice, which occurred after 19 days of CR (Figures 1A and 1C). Mice were then allowed to consume a chow diet ad libitum to stabilize their weights. Interestingly, ReDO mice exhibited considerable hyperphagia during a 24-h period (Figure 1D). Owing to the slight increase in body weight that occurred during that 24-h period, ReDO mice were calorically restricted for an additional 7 days to match their weight to that of control mice before being pair-fed for the duration of the experiment (Figure 1A). While percent body fat, as measured by dual-energy X-ray absorptiometry (DEXA), was comparable between control mice and ReDO mice (Figure 1E), and ReDO adipocyte diameter appeared similar as well, ReDO mice exhibited a pronounced degree of adipocyte inflammation as indicated by the existence of crown-like structures around their epididymal fat adipocytes (Figure 1F).

Figure 1.

ReDO mice exhibit increased hunger immediately following caloric restriction

(A) Weekly body weight measurements 16 weeks followed by daily body weight measurements during the CR phase and beyond. Arrow indicates body weight following 24-h ad libitum chow exposure.

(B) Fasted glucose. Unpaired t-test (two-tailed); ∗∗∗∗p < 0.0001.

(C) Body weight measured at the conclusion of CR. One-way ANOVA; ∗∗∗∗p < 0.0001; adjusted p-value for multiple comparisons using Holm-Šídák test.

(D) 24-h food intake compares control and ReDO mice. Unpaired t test (two-tailed); ∗∗∗∗p < 0.0001.

(E) % body fat. One-way ANOVA; ∗∗∗∗p < 0.0001; adjusted p-value for multiple comparisons using Holm-Šídák test.

(F) Representative histology images across control, DIO, and ReDO groups. Scale bars represent 400 μm for all three images. Data are represented as mean ± SEM.

ReDO mice exhibit a persistent hyperphagia and weight regain phenotype

Intrigued by the marked degree of hyperphagia displayed by ReDO mice, we sought to determine if the drive to consume food was owed to the mice having immediately switched from a hunger-inducing state to a post-CR ab libitum state, as opposed to a response that was driven by some enduring orexigenic drive despite being temporarily removed from their exposure to CR. To test this, control or DIO mice were maintained on their respective diets for 20 weeks, at which time mice were subjected to 40% CR for 3 weeks until they were weight-matched to their lean control counterparts (Figures 2A and 2B). Next, ReDO mice were divided into 4 distinct groups, ReDO_0d, ReDO_8d, ReDO_28d, or ReDO_pf, which were pair-fed a chow diet relative to control mice, for 0 days, 8 days, 28 days, or perpetually, respectively (Figure 2A). All ReDO groups, with the exception of ReDO_pf, were allowed to consume a chow diet ad libitum at the end of their paired-feeding regimen.

Figure 2.

ReDO mice exhibit a persistent hyperphagia and weight regain phenotype

(A) Schematic of the experimental design. Male control mice were fed a standard chow diet, while DIO mice were maintained on a high-fat diet (HFD, 60% kcal from lard) for 20 weeks prior to 40% caloric restriction until reversed dietary obesity (ReDO) mouse groups were weight-matched to lean control mice. ReDO mice were pair-fed using a standard chow diet for either 0 days, 8 days, or 28 days, prior to being maintained on an ad libitum chow diet, while ReDO_pf mice were pair-fed chronically.

(B) Body weight across the experiment depicting the control, DIO, ReDO_0d, ReDO_8d, ReDO_28d, and ReDO_pf groups.

(C) Body weight at week 0.

(D) body weight at week 20.

(E) Body weight at week 23.

(F) Body weight at week 31. One-way ANOVA; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001; adjusted p-value for multiple comparisons using Holm-Šídák test.

(G) First 8-day ad libitum food intake for ReDO_0d mice. Unpaired t-test (one-tailed); ∗∗∗∗p < 0.0001.

(H) First 7-day ad libitum food intake for ReDO_8d mice. Unpaired t-test (one-tailed); ∗p = 0.0310.

(I) Fasted serum insulin comparing a subset of control and ReDO_8d mice. One-way ANOVA; ∗∗∗p < 0.001; adjusted p-value for multiple comparisons using Holm-Šídák test.

(J) Pomc gene expression comparing a subset of control and ReDO_8d mice. Data are represented as mean ± SEM.

Mice were pre-assigned to their respective groups at the onset of the experiment, ensuring that all groups started with average weights that were not significantly different (Figure 2C). After having been maintained on their diet for 20 weeks, all DIO groups, including future ReDO groups, weighed significantly more than control mice while not differing among themselves (Figures 2B and 2D). 7 mice appeared to be diet-induced obesity resistant (DIO-R), according to the criteria used in Enriori et al. (2007), as they possessed a body weight that was beyond ±3 standard deviations of the average body weight of control mice, and were therefore removed from the experiment.¹¹

Despite being weight-matched to control mice after CR, all ReDO groups, regardless of the duration of paired feeding, weighed significantly more than control mice 4 weeks after re-exposure to a standard chow ad libitum diet (Figures 2B and 2F). Importantly, ReDO_pf mice did not exhibit an appreciable degree of weight regain, supporting the conclusion that the weight regain in our ReDO model was principally driven by persistent hyperphagia. Indeed, both ReDO_0d and ReDO_8d mice displayed elevated cumulative food intake, 8 days and 7 days, respectively, after the resumption of ad libitum chow-diet feeding (Figures 2G and 2H). Food intake data for the ReDO28d 1-week ad libitum feeding reintroduction are not reported due to a scale malfunction, which precluded accurate measurement.

To ascertain the physiological basis for the ReDO mouse weight regain, we sought to determine if these mice, despite being weight-matched to controls, were metabolically dysfunctional. So as not to confound our experiment, we measured fasted serum insulin levels in a subset of ReDO_8d mice immediately prior to ad libitum feeding, which required their euthanasia and removal from the remainder of the experiment. ReDO serum insulin levels were indistinguishable from those of controls (Figure 2I). In these same mice, we measured Pomc mRNA expression in the whole hypothalamus, which trended down in comparison to that of control mice but was non-significant (Figure 2J).

Initial HFD-driven body weight change is correlated with the degree of weight regain

Given the considerable degree of regained body weight stratification within each of our ReDO groups, we sought to identify parameters that could be used to predict the degree of weight regain during future ReDO experiments. Correlational matrices were generated for ReDO_0d, ReDO_8d, and ReDO_28d mice, comparing up to 12 variables (ReDO_28 mice are missing two food-intake-related variables) (Figures 3A–3D and 3G). Given the Gaussian distribution of our data, we generated Pearson’s correlation coefficients and p-values to assess the strength of the linear relationship between two variables. Interestingly, for both pair-fed groups ReDO_8d and ReDO_28d, the initial 4-week body weight (BW) change upon initial exposure to a high-fat diet was strongly correlated with the ultimate degree of body weight regain (Figures 3D, 3E, 3F, 3G and 3H).

Figure 3.

Initial HFD-driven body weight change is correlated with the degree of weight regain

(A) Heatmaps of Pearson’s correlation r values (left) and p values (right) between various parameters for ReDO_0d mice.

(B) Linear regression compares ReDO_0d Initial HFD week 4 BW change versus 1 week regain food intake.

(C) Linear regression compares ReDO_0d Initial HFD week 4 BW change versus 4-week BW regain.

(D) Heatmaps of Pearson’s correlation r values (left) and p values (right) between various parameters for ReDO_8d mice.

(E) Linear regression comparing ReDO_8d Initial HFD week 4 BW change versus 1 week regain food intake (g).

(F) Linear regression compares ReDO_8d Initial HFD week 4 BW change versus 4-week BW regain.

(G) Heatmaps of Pearson’s correlation r values (left) and p values (right) between various parameters for ReDO_28d mice.

(H) Linear regression compares ReDO_28d Initial HFD week 4 BW change versus 4-week BW regain. Food intake columns are missing for ReDO_28d mouse comparisons.

Discussion

Our findings recapitulate and extend those of other groups, revealing that mice with reversed dietary obesity are not only hyperphagic immediately following a CR regimen,⁶ but their hyperphagia lasts for at least a month, despite being diet-matched to control mice during this time. Whereas previous studies have detected various forms of obesogenic memory in the periphery (e.g., persistent adipose tissue inflammation),⁷ it is unclear whether these contribute to weight regain. Attention should be directed toward identifying the key drivers of weight-loss-associated hyperphagia. Various neuronal populations within the hypothalamus, hindbrain, and other brain regions have been established as key regulators of hunger.¹² In particular, leptin receptor-expressing neuronal-types within the hypothalamus, including AgRP and Pomc neurons, as well as Glp1r-expressing cell types within the hindbrain and hypothalamus, have been strongly implicated in the control of hunger and body weight.^{13,14,15,16,17} Of note, in response to DIO, AgRP neurons exhibit an attenuated response to gastrointestinal hormones or nutrients that persists despite weight loss resulting from diet reversal.¹⁸ Studies examining the firing properties of AgRP and other nutrient-sensing neurons in ReDO mice would likely be informative. Moreover, our study, similar to that of earlier work,⁶ failed to detect significant hypothalamic gene-expression profiles indicative of impairments in the function of these neuronal systems, likely owing to the technical difficulties of measuring cell type-intrinsic gene expression profiles in “noisy” whole-tissue preps with an exceptionally high degree of cellular heterogeneity. Future studies should infer the state of hunger-linked neurons in ReDO mice using sensitive cell-type-specific transcriptomic approaches optimized for rare hypothalamic neuronal types.¹⁹

Compared to lean control mice, mice with reversed dietary obesity appear to be driven to reclaim an upwardly shifted body weight set point. Precisely when, after initial exposure to a high-fat diet, this set point becomes upwardly shifted remains an open question. Recent work involving an intragastric overfeeding mouse model suggests that after 14 days mice rapidly re-establish a weight that is similar to control mice, suggesting set point shifting occurs after more than 2 weeks of sustained positive energy balance.²⁰ Moreover, another study, using DIO mice that were switched back to chow in a staggered fashion, suggests that set point shifting may occur between 8 weeks and 24 weeks of HFD exposure.²¹

We observed a fair degree of population structure within our data, with some mice regaining more weight and consuming more food than others. Given the low degree of body weight variability within each of the ReDO groups prior to HFD feeding, we suspect that some aspect of high-fat diet exposure unveils a hidden susceptibility to diet-induced obesity. Furthermore, initial HFD-induced body weight gain being positively correlated with weight regain in ReDO_8d and ReDO_28d mice suggests that mice may be uniquely programmed for a particular degree of initial weight gain under various orexogenic conditions, in a manner that is distinct from how they are programmed to maintain their steady-state body weight during long-term chow- or HFD-fed exposure. Indeed, initial pre-HFD body weight and total % HFD-induced weight gain prior to CR were not significantly correlated with weight regain for the ReDO_28d group. We suspect ReDO_0d mice do not show this association between initial HFD-weight gain and weight-regain susceptibility because their post-CR orexigenic drive is so uniformly strong that it masks any within-group differences. The initial 4-week body weight change metric can be leveraged to identify weight-regain susceptible versus resistant mice a priori for future comparative studies.

Limitations of the study

A limitation of this study involved our only including male mice, and future experiments should determine whether female ReDO mice exhibit persistent hyperphagia. Our study did not assess whether ReDO mice pair-fed for longer than 1 month exhibit eventual weight regain, and this should be examined in a future experiment. We did not comprehensively assess the metabolic profile of ReDO mice relative to control and DIO mice, and future investigations should seek to determine whether various endocrine factors are persistently altered in paired-fed ReDO mice. Similarly, this study did not engage in deep transcriptional profiling within the hypothalamus of our groups, and future investigations should seek to identify transcriptional programs that are potentially altered in hyperphagic ReDO mice.

Resource availability

Lead contact

Correspondence and requests for materials should be addressed to and will be fulfilled by the lead contact, Frankie D. Heyward (frankie.heyward@utsouthwestern.edu).

Materials availability

This study did not generate unique reagents.

Data and code availability

• •All raw data are provided in the form of supplementary tables.
• •This study did not utilize code.
• •Any additional information required to reanalyze the data reported in this article is available from the lead contact upon request

Acknowledgments

This work was supported by NIH T32 DK007516 and NIH K01 DK131347 to F.D.H., and R01 DK085171 to E.D.R.

Author contributions

F.D.H. conceived the study, conducted experiments, interpreted results, and wrote the manuscript with input from E.D.R., who supervised and funded the study.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Critical commercial assays
Ultra Sensitive Mouse Insulin ELISA Kit	Crystal Chem	Cat#90080
Experimental models: Organisms/strains
Mouse: C57BL/6J	The Jackson Laboratory	Strain #000664; RRID:IMSR_JAX:000664
Oligonucleotides
Pomc (forward)	This paper	CATAGATGTGTGGAGCTGGTG
Pomc (reverse)	This paper	CAGCGAGAGGTCGAGTTTG
Hprt (forward)	This paper	GGAGTCCTGTTGATGTTGCCAGTA
Hprt (reverse)	This paper	GGGACGCAGCAACTGACATTTCTA
Software and algorithms
Prism	Graphpad software	RRID:SCR_002798
Other
60% High-Fat Diet	Research Diets, Inc.	Cat#: D12492
LabDiet 5008 (Formulab Irradiated Rodent Chow)	LabDiet	Cat#: 5008

Experimental model and study participant details

All animal experiments were performed with approval from the Institutional Animal Care and Use Committees of The Harvard Center for Comparative Medicine and Beth Israel Deaconess Medical Center (IACUC Protocol #071–2014). 4-week-old (experiment 1) or 5-week old (experiment 2) C57BL/6J male mice from The Jackson Laboratory (Strain #000664) were shipped to our animal facility in groups of weanlings. Upon arrival, mice were group-housed (n = 5), with food and water available ad libitum, on a 12:12 h light/dark schedule [lights on at 6 a.m. (experiment 1) or 7 a.m. (experiment 2); zeitgeber time (ZT) 0]. Mice were acclimated to the housing conditions for at least 1 week. At 6 weeks of age, mice were maintained on either a control, standard chow, diet or a DIO (60% fat by calories from lard) diet (Research Diets, #D12492) for 16 weeks (experiment 1) or 20 weeks (experiment 2). Individual mouse body weights were measured weekly prior to the CR phase of the experiment. One control mouse’s 24-h food intake measurement was omitted owed to it being recorded as 153 g, which is physically impossible (experiment 1). Since only male mice were used in this study, the influence of sex on the investigated phenomena could not be determined.

Method details

Measurement of endocrine profiles

Fasting glucose levels were obtained by fasting mice overnight for 12 h, after which tail blood glucose levels were measured using a glucometer. Serum samples for insulin were taken from trunk blood following euthanasia and decapitation. Insulin analysis was performed via ultra-sensitive mouse insulin ELISA kit (Crystal Chem, Downers Grove, IL).

Caloric restriction and paired feeding

Experiment #1: Prior to CR all mice were group-housed. Control mouse cage-wide food intake was determined for 24 h, and group-housed CR mice were given that amount the following day to commence CR. Mice subjected to CR were weighed daily and given food 30 min to 2 h prior to the onset of the dark cycle. After 1 week (week 17) all mice, in all groups, were singly housed and ReDO mouse daily food intake was measured across two days, after which mice were given 60% of this amount (40% CR) for 12 days until the CR phase was complete. Immediately following CR, ad libitum food intake was measured for 24 h for all mice. ReDO mice were subjected to CR for an additional 7 days prior to being pair-fed for the duration of the experiment.

Experiment 2: After week 20, control, DIO, and ReDO mice were singly housed. Control 24-h food intake was averaged over two days, and 60% was calculated to enforce 40% CR (e.g., 3.45g x 0.60 = 2.1g), which was given to CR mice daily.

Hypothalamic tissue collection and real-time PCR analysis

Mice were food-deprived at the onset of the light cycle (6:00 a.m.) and tissue was collected 7-to-11 h later in Trizol (Thermo Fisher), RNA isolated, cDNA generated, and qPCR performed with normalization to Hprt. Analysis of qPCR was conducted via the 2-ΔΔCT method.

Quantification and statistical analysis

Data are presented as means ± SEM. Analyses used GraphPad Prism, with comparisons via Student’s t-test or one-way ANOVA with Holm-Sidak correction, and correlations by Pearson correlation, given normal data distribution. For all analyses, a p-value of less than 0.05 was considered statistically significant with our employing the following key: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Published: March 5, 2026

Supplemental information

Table S1. Raw data for all figures involving statistical analyses are provided in the spreadsheet associated with this manuscript

(A) Tab 1b: Data corresponding to Figure 1B.

(B) Tab 1c: Data corresponding to Figure 1C.

(C) Tab 1d: Data corresponding to Figure 1D.

(D) Tab 1e: Data corresponding to Figure 1E.

(E) Tab 2c: Data corresponding to Figure 2C.

(F) Tab 2days: Data corresponding to Figure 2D.

(G) Tab 2e: Data corresponding to Figure 2E.

(H) Tab 2f: Data corresponding to Figure 2F.

(I) Tab 2g: Data corresponding to Figure 2G.

(J) Tab 2h: Data corresponding to Figure 2H.

(K) Tab 2i: Data corresponding to Figure 2I.

(L) Tab 2j: Data corresponding to Figure 2J.

(M) Tab 3a_b_c: Data corresponding to Figures 3A–3C.

(N) Tab 3days_e_f: Data corresponding to Figures 3D–3F.

(O) Tab 3g_h: Data corresponding to Figures 3G and 3H.