Conserved Noncoding Cis-Elements Associated with Hibernation Modulate Metabolic and Behavioral Adaptations in Mice

Abstract

Cis-regulatory elements (CREs) drive phenotypic diversity, yet how CREs are causally linked to function remains largely unclear. Our study elucidates functions for conserved cis-elements associated with the evolution of mammalian hibernation and metabolic flexibility. Genomic analyses revealed topologically associated domains (TADs) enriched for convergent changes in hibernators, including the Fat Mass & Obesity (Fto) locus. In this TAD, we uncovered genetic circuits for metabolic responses and hibernation-linked cis-elements forming regulatory contacts with neighboring genes. Deletions of individual cis-elements in mice differentially altered Fto, Irx3, and Irx5 expression, reshaping downstream gene expression programs and affecting metabolism, torpor, obesogenesis, and foraging in distinct ways. Our findings show how convergent evolution in hibernators pinpoints functional genetic mechanisms of metabolic control, with multiple effects encoded in single CREs.

Understanding the genetic basis of metabolic regulation is crucial for insights into disease, aging, and species-specific metabolic traits (1-4). Extreme metabolic phenotypes, such as obligate hibernation, could be leveraged to help uncover the genetics of mammalian metabolic control. Hibernators gain up to 50% body weight before entering torpor, suppressing metabolism and body temperature to survive months of food scarcity (5-9). In contrast, non-hibernating homeotherms maintain stable metabolism year-round, relying on diverse foraging strategies (10). The repeated evolution of hibernation in mammals suggests shared coding genes with non-hibernators, and that regulatory changes drive key phenotypic differences.

Cis-regulatory elements (CREs) are major drivers of phenotypic diversity (11-15). Genome-wide association studies have linked CREs to human phenotypic differences (16), whereas reporter assays (17) and screens (18, 19) help identify influences on gene expression. However, defining causal CREs shaping complex metabolic phenotypes remains challenging. Topologically associated domains (TADs) are 3D genomic structures that restrict CRE interactions to nearby genes, helping define cis-regulatory programs (20-24). The TAD harboring the Fto locus—the strongest genetic risk factor for human obesity—plays a key role in metabolism (25-28). Our companion study identified conserved hypothalamic CREs that regulate fed, fasted, and refed (FFR) responses in mice and exhibit convergent genomic changes with loss-of-function in hibernators (29). Here, we defined a subset of these "hibernation-linked CREs" within the Fto TAD and demonstrate their function using knockout models. These CREs regulate gene expression and influence hibernation-relevant metabolic traits and foraging.

The Fto-Irx TAD is a Primary Region of Evolutionary Divergence between Hibernating and Homeothermic Mammals

To start defining causal CREs that govern metabolic phenotypes, we first aimed to identify top hibernation-linked cis-elements and cis-regulatory programs for targeted functional studies. Using the parallel hibernator accelerated and deleted regions (pHibARs and pHibDELs) uncovered in our companion study (29), we tested for enrichments in TADs throughout the mouse genome compared to background conserved regions. For comparison, we performed the same analysis for control parallel homeotherm accelerated regions (pHomeoARs) and deleted regions (pHomeoDELs), as previously described (29). From 2153 TADs tested, we found 25 TADs significantly enriched for pHibARs and 580 TADs enriched for pHibDELs (Fig. 1A, blue). In contrast, pHomeoARs and pHomeoDELs are enriched in different TADs than those in hibernators (Fig. 1A, red). These results identify TADs that disproportionately accumulated convergent genomic changes in hibernators at conserved cis-elements (table S1).

Fig. 1. Convergent genomic changes in hibernators identify Fto-Irx TAD cis-elements for functional analysis in knockout mice to affect metabolic gene regulatory networks.

(A) Volcano plots show the odds ratio and p-values for TADs throughout the mouse genome to be enriched for hibernator (blue) and homeotherm (red) convergent genomic changes relative to background conserved regions (CRs). The results in the left plot for pHibARs (blue) and pHomeoARs (red) are based on a chi-square test of a contingency table of AR and CR counts (rows) by hibernator and homeotherm data (columns). The same analysis for pHibDELs versus pHomeoDELs is shown in the right plot. Significantly enriched TADs for hibernator (blue circles) or homeotherm (red circles) convergent genomic changes are shown (q-value<0.05). Other TADs are indicated by grey circles. The Fto-Irx TAD is among the top enriched TADs for both pHibARs and pHibDELs. (B) Barplots show the odds ratio for Irx3 and Irx5 binding site motifs to be located in adult female mouse hypothalamus ATAC-Seq peak sites of open chromatin (FDR 5%) that are specific to fed, 72-hr fasted, versus 12-hr refed conditions compared to peaks that are shared across all three conditions. The results show that Irx3 motifs are enriched in Fed, 72-hr fasted, and 12-hr refed specific peak sites, whereas Irx5 motifs are enriched in 72-hr fasted, and 12-hr refed specific sites. (C) Unsupervised hierarchical clustering (Poisson Dissimilarity Matrix, Ward’s method) on Irx3 Cut&Run binding site profiles in the mouse hypothalamus shows that replicates cluster according to fed, 72-hr fasted, and 12-hr refed condition, indicating metabolic state specific binding profiles (n=4 mice). Horizontal axis shows linkage Euclidean distance. (D) An example of a significant Irx3 binding site peak identified from Cut&Run across all replicates and conditions (q<0.0001, Genrich peak calling). (E) Number of significant differential Irx3 binding sites identified from fed vs. fasted, fasted vs refed and fed vs refed contrasts (q<0.1; EdgeR, n=4 mice). (F) A venn diagram of the numbers of genes contacted by Irx3 binding sites in fed, fasted, and refed conditions as revealed by Cut&Run and H3K27ac+ PLAC-Seq in the mouse hypothalamus. (G) Circos plot displaying genome-wide tracks of Cut&Run significant Irx3 binding sites in the hypothalamus for fed (yellow), fasted (blue) and refed (green) conditions. Arrows in the center connect the Irx3 locus to fed specific, fasting specific, refed specific or shared binding sites. See legend. (H) The plot displays multi-omics data for the Fto-Irx TAD focused on 5 cis-elements showing convergent genomics changes in hibernators (Fto-Irx::hibE1-5 cis-elements) that were selected for functional studies in knockout mice and the genomic size and coordinates for each cis-element. PLAC-Seq for H3K27ac shows the regulatory contacts made by each knocked out cis-element (E1, dark blue; E2, light blue; E3, red; E4, green; E5, purple) and reveals that each cis-element makes multiple long-distance contacts that affect Fto, Irx3, and Irx5, but not Rpgrip1l. Tracks (brown) are shown for the pHibARs and pHibDELs identified in the TAD and show the knockout targeted regions overlap with pHibARs (E1, E3, E4, E5) or pHibARs and pHibDELs (E2). The snATAC-Seq tracks show significant ATAC-Seq peaks in each targeted cis-element identified from single cell multi-omics profiling of fed, 72-hr fasted, and 12-hr refed adult mouse hypothalamus (FDR < 5%). The open chromatin peaks for each targeted cis-element in oligodendrocyte precursor cells (OPCs), astrocytes (Astro), endothelial cells (endo), GABAergic neurons (gaba), Glutamatergic neurons (Glut), Microglia (micro), Other neurons (neuron), and Oligodendrocytes (Oligo).

Among the top pHibAR- and pHibDEL-enriched TADs, we identified the Fto-Irx TAD, which includes Rpgrip1l (RPGR-Interacting Protein 1-Like Protein), Fto (Fat Mass & Obesity alpha-ketoglutarate dependent dioxygenase), Irx3 (Iroquois 3 homeobox transcription factor), and Irx5, building on previous work (31) (Fig. 1A, table S1). Variants in a CRE within FTO’s first intron regulate IRX3 and IRX5 through long-range interactions (26, 30), influencing obesity (30), age-related brain atrophy (31, 32), feeding behaviors (33-36), and brain injury risk (37). We identified 80 pHibARs and 37 pHibDELs within this TAD, highlighting discrete Fto-Irx cis-elements as strong candidates for functional studies of hibernation-related phenotypes.

Convergent genomic changes in hibernators impact hypothalamic gene expression programs for FFR responses (29). The regulatory functions of Fto (an m6A demethylase) and Irx3/Irx5 (homeobox transcription factors) suggest potential for regulating FFR responses. To test this, we analyzed hypothalamic ATAC-Seq profiles from fed, 72-hr fasted, and 12-hr refed mice, and assessed whether condition-specific open chromatin sites were enriched for Irx3 and/or Irx5 motifs relative to shared peaks. Irx3 motifs were enriched across all three conditions, whereas Irx5 motif enrichments were specific to fasted and refed states (Fig. 1B), suggesting roles for Irx transcription factors in FFR gene regulation within the Fto-Irx TAD.

To assess Irx3’s role in FFR regulation, we performed Cut&Run in fed, 72-hr fasted, and 12-hr refed mouse hypothalami to map genome-wide binding sites. Control analyses confirmed specificity, showing enrichment for Irx3 motifs and a cutoff for high-confidence peaks (Fig. S1A-C). Irx3 binding varied according to FFR conditions, revealing adaptive regulatory programs (Fig. 1C-E, S1D). Distal binding sites were linked to genes using H3K27ac+ PLAC-Seq enhancer-promoter contacts (29), demonstrating that FFR-driven changes in Irx3 binding altered downstream gene regulation (Fig. 1F, S2A, Table S2). Thus, Irx3 orchestrates hypothalamic gene programs based on FFR state, positioning the Fto-Irx TAD as a key upstream regulator (Fig. 1G).

We assessed Fto, Irx3, and Irx5 expression across FFR conditions. Single-nuclei (sn) RNA-Seq showed no significant changes between fed and fasted states, but some cell type-specific differences upon refeeding (Fig. S2B). Bulk RNA-Seq revealed increased Irx3 expression during fasting (Fig. S2C) and increased Irx5 with refeeding (Fig. S2D). Overall, Fto, Irx3, and Irx5 expression remains largely stable, whereas Irx3 dynamically regulates numerous targets by altering binding locations in response to metabolic changes.

We sought to identify hibernation-linked cis-elements in the Fto-Irx TAD for targeted functional studies. Using H3K27ac+ PLAC-Seq, snMulti-omics, and pHibAR/pHibDEL datasets (29), we examined the cis- regulatory architecture of this TAD in the mouse hypothalamus. Fto, Irx3, and Irx5 were expressed across major hypothalamic cell types (Fig. S2B). snATAC-Seq revealed pHibARs and pHibDELs in accessible noncoding regions, whereas PLAC-Seq identified genes with regulatory contacts from each cis-element (Fig. 1H).

We identified five hibernation-linked cis-elements with open chromatin in hypothalamic cells and multiple regulatory contacts within the Fto-Irx TAD, suggesting roles in coordinating Fto, Irx3, and Irx5 expression. These were named Fto-Irx::hibE1 through hibE5 (Fig. 1H). Each of them overlapped one or more pHibARs (Table S3) and showed chromatin accessibility in specific hypothalamic cell types. All five formed distinct regulatory contacts across the Fto-Irx TAD, with all interacting with Irx3 and Irx5 promoters, whereas all except Fto-Irx::hibE4 also contacted Fto (Fig. 1H). None of them interacted with Rpgrip1l. To assess their functional roles in gene regulation, metabolism, and behavior, we generated knockout mice for each cis-element (Tables S3, S4).

Deletions of Hibernator-Linked Cis-Elements Alters Hypothalamic Gene Expression and Fasting Responses

To test whether Fto-Irx::hibE elements regulate gene expression, we performed bulk RNA-Seq on hypothalamus from ^−/− and ^+/+ adult female mice across five knockout lines under fed and 72-hr fasting conditions. Examining Fto, Irx3, and Irx5 expression, we found changes in ^−/− mice, except for Fto-Irx::hibE1^−/−, which showed no effect (Fig. 2A-E). Fed Fto-Irx::hibE2^−/− mice had reduced Irx3 and Irx5 expression (Fig. 2B). Fed Fto-Irx::hibE3^−/− mice showed increased Fto and decreased Irx5 expression, whereas fasting reduced Irx3 and Irx5 without affecting Fto (Fig. 2C). In Fto-Irx::hibE4^−/− mice, Irx3 and Irx5 expression decreased after fasting (Fig. 2D), whereas fasted Fto-Irx::hibE5^−/− mice exhibited increased Irx5 expression only (Fig. 2E). No changes were observed in neighboring Rpgrip1l or Crnde. Thus, four out of five Fto-Irx::hibE cis-elements regulate Fto, Irx3, and/or Irx5 expression in a metabolic state-dependent manner. These genes show minimal expression changes in wild-type mice under fasting (Fig. S2B,C), but deletion of these hibernation-linked cis-elements alters their transcriptional response to fed vs. fasted states (Fig. 2A-E), suggesting that loss-of-function mutations in these elements confer new metabolic responses.

Fig. 2. Gene expression alterations in the hypothalamus of Fto-Irx::hibE cis-element knockout mice under fed and 72-hr fasted conditions.

(A-E) The barplots show the RNA-Seq measured expression of Fto, Irx3, and Irx5 in the hypothalamus in each of the Fto-Irx::hibE knockout mouse lines (n=4 mice, two-tailed test (Fto) or One-way ANOVA (Irx3 and Irx5) with Tukey post-test). The H3K27ac+ PLAC-Seq detected regulatory contacts for each cis-element are shown above, along with tracks for pHibARs and the location of each knocked out cis-element (E). Data for mice in a fed (grey box) versus 72-hr fasted condition (orange box) are shown. CPM, counts per million. The results show how each element affects Fto, Irx3, and Irx5 expression in two different metabolic conditions. (F) The heatmap shows the expression of RNA-Seq detected differentially expressed genes (DEGs, FDR < 5%) in the hypothalamus for Fto-Irx::hibE3^−/− mice in fed and 72-hr fasted conditions relative to ^+/+ littermates. (G) The barplots show the numbers of significant DEGs (RNA-Seq, FDR<5%) detected in the hypothalamus of ^−/− versus ^+/+ mice for fed and 72-hr fasted conditions for all 5 different Fto-Irx::hibE cis-element knockout mice (E1-5). (H) The plot shows significant KEGG pathway enrichments from a gene set analysis of significant DEGs in Fto-Irx::hibE2^−/− , Fto-Irx::hibE3^−/− , and Fto-Irx::hibE4^−/− mice.

Given their regulatory roles, changes in Fto, Irx3, and Irx5 expression could influence broader gene expression programs. Supporting this hypothesis, we identified hundreds of differentially expressed genes (DEGs, FDR < 5%) in Fto-Irx::hibE2^−/−, Fto-Irx::hibE3^−/−, and Fto-Irx::hibE4^−/− mice compared to their ^+/+ littermates (Fig. 2F,G). The number of DEGs varied by cis-element and metabolic state (Fig. 2G, Table S5). Fto-Irx::hibE1^−/− and Fto-Irx::hibE5^−/− mice showed only modest effects, suggesting limited influence in the adult hypothalamus (Fig. 2G).

Gene set analysis revealed KEGG pathway enrichments that differed by cis-element and metabolic state (Fig. 2H, Table S6), including signaling, metabolic, and hormonal pathways. snRNA-Seq comparisons showed cell type-specific effects, with Fto-Irx::hibE3^−/− DEGs predominantly affecting GABAergic and other neurons, whereas Fto-Irx::hibE4^−/− DEGs were enriched in endothelial cells and neurons (Fig. S3). Since these mice are germline knockouts, gene expression changes may result from direct hypothalamic effects or indirect influences from other tissues or development. Our findings confirm functional CRE activity for conserved Fto-Irx TAD elements impacted by pHibARs and/or pHibDELs, with single CRE deletions altering metabolic responses and gene regulation.

Hibernation-Linked CREs Modulate Specific Aspects of Metabolism in Torpor

Hibernation is an adaptation for surviving periods of food scarcity. Mice do not hibernate but show brief fasting-inducible torpor bouts (38, 39). We devised an assay of hibernation-relevant phenotypes in mice to explore how hibernation-linked CREs modulate aspects of metabolism and torpor (Fig. 3A). Adult female mice were implanted with internal body temperature (Tb_i) monitors and placed in comprehensive lab animal monitoring system (CLAMS) metabolic cages to assess calorimetric, activity, and feeding changes across a 7-day induced-torpor paradigm. Mice were first fed ad libitum for 48 hrs (pre-torpor), then fasted and exposed to 18°C for 48 hrs (torpor induction), followed by 3 days of ad libitum refeeding at 25°C. Tb_i dropped from ~38°C to ~20°C during torpor and returned to baseline within hours of refeeding (Fig. 3A). This FFR paradigm was used to assess metabolic phenotypes in Fto-Irx::hibE knockout mice.

Fig. 3. Distinct Metabolic and Thermoregulatory Response Profiles in Fto-Irx::hibE cis-Element Knockout Mice Across Fed-Fasted-Refed Stages of Fasting-Induced Torpor.

(A) The plot shows internal body temperature (Tbi) in mice over 7-days while placed in CLAMS cages for a metabolic assay of processes pertinent to hibernation. The assay has a 48-hr pre-torpor phase (red, ad libitum food + 25°C), 48-hr torpor phase (blue, food deprived + 18°C), and a 72-hr post-torpor ad libitum refeeding phase (orange, ad libitum food + 25°C). Mice show circadian changes with elevated Tbi during active periods (dark cycle, grey bars) and depressed Tbi during inactive periods (light cycle, colored bars). During the torpor phase, Tbi drops, which is indicative of torpor, and then rapidly recovers during refeeding. Dark green line shows the mean, whereas the shaded area shows S.E.M. (B-E) The plots show the metabolic rate (O₂ consumption ml/hr) measured from ^−/−(pink) versus ^+/+ (green) adult female mice in CLAMS cages during the metabolic assay. Mean metabolic rate results (solid line) are shown and S.E.M. (shaded area) for Fto-Irx::hibE1^−/− mice (n=11-14 mice) (B), Fto-Irx::hibE2^−/− mice (n=11 mice) (C), Fto-Irx::hibE3^−/− mice (n=11 mice) (D), and Fto-Irx::hibE4^−/− mice (n=11 mice) (E). Statistical significance in the pre-torpor, torpor, and refeeding phases was determined using a generalized linear regression analysis of the mean metabolic rate for each day and light-dark cycle and tested for a main effect of genotype and an interaction effect between genotype and day/light-dark cycle. Variance due to batch effects from independent CLAMS cohorts was accounted for by a “batch” term in the model. P-values in black show the genotype effect (significant p-values in bold). P-values in blue show the genotype x day/light-dark cycle interaction effect (only significant p-values shown; C, torpor phase). *p<0.05, **p<0.01, ***p<0.001

The induced torpor study focused on three hibernation-linked CREs (Fto-Irx::hibE2^−/−, Fto-Irx::hibE3^−/−, and Fto-Irx::hibE4^−/−), which had the highest numbers of differentially expressed genes, suggesting key regulatory roles. Fto-Irx::hibE1^−/− mice, which showed minimal hypothalamic gene expression changes, were also included due to their location in Fto’s first intron near the human obesity-linked region (26, 40). Adult female ^−/− mice were compared to ^+/+ littermate controls, and regression analysis tested for genotype effects and genotype-by-light-dark phase interactions. Each cis-element had distinct effects on metabolic profiles across torpor phases: Fto-Irx::hibE1^−/− mice had reduced metabolic rate (MR) in the refed phase (Fig. 3B); Fto-Irx::hibE2^−/− mice exhibited increased MR at the end of torpor, with a genotype-light-dark phase interaction (Fig. 3C); Fto-Irx::hibE3^−/− mice showed increased MR in the pre-torpor phase (Fig. 3D); and Fto-Irx::hibE4^−/− mice displayed decreased MR in pre-torpor (Fig. 3E).

Comprehensive metabolic and activity profiling revealed distinct roles for each CRE in modulating metabolism and body weight (fig. S4-S8). Fto-Irx::hibE1 primarily affected Tb_i and energy expenditure during refeeding (fig. S4), whereas Fto-Irx::hibE2 influenced metabolism during deep torpor and emergence (fig. S5). Fto-Irx::hibE3^−/− mice exhibited increased metabolic rates and energy expenditure (fig. S6), whereas Fto-Irx::hibE4^−/− mice showed decreased rates, particularly in pre-torpor (fig. S7). These neighboring CREs thus have opposing effects on metabolic regulation. Only Fto-Irx::hibE1^−/− mice displayed altered body weight, with reduced weight and incomplete recovery post-torpor (fig. S8). Overall, each CRE plays a distinct role in regulating metabolism across pre-torpor, torpor, and refeeding phases.

Hibernation-Linked CREs Differentially Modulate High-Fat Diet Weight Gain in Males versus Females

Hibernators exhibit extreme metabolic flexibility, gaining fat before prolonged fasting during torpor. Females prioritize fat storage and longer hibernation for reproductive energy needs, whereas males store less fat, hibernate for shorter periods, and maintain torpor flexibility for competitive advantages (7, 41, 42) . To assess roles for hibernation-linked cis-elements in sex-specific weight regulation, we subjected knockout and wild-type mice to a 16-week high-fat diet (HFD). Two cis-elements (Fto-Irx::hibE1 and Fto-Irx::hibE4) affected weight gain in females (Fig. 4B, H), whereas three (Fto-Irx::hibE2, Fto-Irx::hibE3, and Fto-Irx::hibE4) affected males (Fig. 4C, E, G). Only Fto-Irx::hibE4 influenced both sexes. Fto-Irx::hibE1^−/− females gained more weight, whereas Fto-Irx::hibE2^−/− males gained less. Fto-Irx::hibE3^−/− males had increased body weight but no change to HFD-induced gain. Fto-Irx::hibE4^−/− mice of both sexes were lighter, with reduced HFD-induced gains. Body composition and organ weight analysis showed no differences between knockouts and controls (Tables S7, S8). Overall, hibernation-linked Fto-Irx cis-elements distinctly influence body weight and weight gain in a sex-dependent manner.

Fig. 4. Distinct Effects of Fto-Irx::hibE Cis-Elements on Body Weight and High-Fat Diet-Induced Weight Gain in Males and Females.

(A-H) The plots show the increase in body weight for −/− (purple line) versus +/+ (green line) adult male (A, C, E, G) and female (B, D, F, H) mice on a high fat diet over 16 weeks. Data for Fto-Irx::hibE1 (A,B), Fto-Irx::hibE2 (C,D), Fto-Irx::hibE3 (E,F), and Fto-Irx::hibE4 (G,H) −/− versus +/+ control mice are shown (n shown, bottom right). *p<0.05, **p<0.01, ***p<0.001. Likelihood ratio test of regression models. Mean ± SEM.

The Hibernation-Linked Fto-Irx::hibE3 CRE Influences Foraging Patterns

Metabolism and foraging are biologically intertwined, with metabolism driving energy needs and foraging evolving to meet these demands (43-45). Metabolism also influences aging, and genes in the Fto-Irx TAD are associated with feeding behavior and age-related brain atrophy (31-33, 35, 36, 46). Unlike non-hibernators, hibernators seasonally forage to gain weight and enter torpor during food scarcity, exhibiting enhanced longevity. Thus, hibernation-linked CREs in the Fto-Irx TAD may influence foraging and age-related changes. We tested this in Fto-Irx::hibE3^−/− mice, which showed strong gene expression and metabolic phenotypes, using a naturalistic foraging assay and computational approaches (47-49). The foraging assay consists of an arena attached to a home cage (Fig. 5A) with two phases: (1) a 30-minute Exploration phase, where mice explore a novel environment and find a food patch (Pot 2), and (2) a 30-minute Foraging phase, 4 hours later, where food is relocated and buried (Pot 4). The Exploration phase assesses novelty response, whereas the Foraging phase evaluates memory-driven behavior (Movie S1) (49). Mice perform different round-trip foraging excursions from the home (47-49) (Fig. 5B). Using this approach, we analyzed foraging in male and female Fto-Irx::hibE3^−/− and ^+/+ mice across young (2–3 months) and aged (12 months) cohorts.

Fig. 5. Alterations to Naturalistic Foraging in Fto-Irx::hibE3^−/− Mice with Age.

(A) The schematic shows the foraging assay design, including the home cage attached to the arena by a tunnel. The arena has 4 potential food patches (Pots1-4) and one has food (Pot2, red) in the 30-min naïve Exploration phase (red pot). In the 30-min Foraging phase 4 hours later, the food is moved from Pot2 and buried in Pot4 (green), such that mice express memory response behaviors of the familiar environment and former food location (Pot2). The diagrams show the key locations in the assay, including zones not involving a Pot or the home (Not-At-Pot, NAP locations). (B) Foraging behaviors are tracked and segmented into discrete round-trip excursions from the home and then decomposed by 52 measures of different behavioral components. (C) The plots show Fto-Irx::hibE3^−/− (orange) and ^+/+ control mice (blue) foraging patterns from the cumulative time spent at each location (seconds) in the Exploration (green) and Foraging (purple) phases for male + female mice, including young adult (8–12-week-old) and (12-month-old) aged adult cohorts (see titles). Bonferroni corrected p-values across the 4 conditions shown (p.Bonf). Bars show mean ± SEM. * p.BH<0.05. TTP, total time at Pot (1-4). (D and E) The schematic depicts how different round trip foraging excursions are grouped into stereotypic foraging excursion types, called modules, using an unsupervised machine learning approach (C). The Venn diagram shows the numbers of foraging modules uncovered from the 156 mice and 7815 excursions analyzed in the Exploration and Foraging phases. The PHATE map shows 2-D embeddings of each excursion colored according to foraging module classification (D). The results show that excursions classified to the same module are closely clustered (see 3-D embeddings in Movie S2). (F) Results for likelihood ratio test comparing two Poisson regression models assessing the interaction between genotype, age, sex, and module on foraging excursions performed by the 156 tested male, female, young, aged, −/− and +/+ mice. Model 1 included additive effects of Genotype and Module, controlling for sex and age differences, whereas Model 2 included their interaction (Module × Genotype). Bonferroni correction for multiple testing in the Exploration and Foraging phases. Df, Model 1 vs. 2 changes in degrees of freedom; n, number of excursions. (G) Post-hoc estimated marginal means analysis of the Foraging phase data identified specific modules where genotype had a significant effect with interactions further modulated by age. The plot illustrates the difference in mean expression between homozygous (−/−; HOM) and wild-type (+/+; WT) mice for significantly affected modules during the Foraging phase, stratified by young and aged adults (BH corrected p < 0.05). The HOM–WT difference in estimated marginal mean expression is represented by dots, with error bars denoting ± standard error (SE). Only modules that show significant differential expression between HOM and WT mice are displayed. *p.adjust<0.05; **p.adjust<0.01. (H) The PHATE map shows the 2-D embeddings for the foraging excursions for module 11 (red) and module 46 (green) that show significant differential expression between ^−/− versus ^+/+ mice in aged adults in the Foraging phase (G). The purple traces show foraging excursion X-Y-Z movement sequences representative of each affected module shaded by movement velocity (purple, see legend). The inset barplot shows the time spent at Pot4 for Module 11 versus 46 excursions expressed by aged mice pooled for genotype and sex (t-test, n=16-21). *p<0.05.

To test our primary hypothesis that Fto-Irx::hibE3^−/− mice exhibit changes to foraging behavior, we measured the amount of time spent at key locations in the assay, including the home, different patches (Pot1, Pot2, Pot3, and Pot4), and zones of the arena not associated with a patch. A multiple regression analysis accounting for sex differences revealed a significant genotype-by-location interaction effect (Bonferroni corrected p-value <0.05) in aged mice during the Foraging phase (Fig. 5C). Post-tests revealed that ^−/− mice spent less time at the novel food patch (Pot 4). No effects were observed in the Exploration phase. Thus, aged Fto-Irx::hibE3^−/− mice exhibited altered spatial foraging patterns.

Fto-Irx::hibE3 Affects Discrete Foraging Excursion Types Across the Lifespan

Many hibernating and non-hibernating mammals are central place foragers, making round-trip excursions that include outbound travel, searching, and a goal-driven return (50, 51). Foraging excursions are discrete behavioral sequences (Fig. 5B) linked to different genetic factors (47-49). We expanded the analysis of Fto-Irx::hibE3^−/− mice to test effects on specific excursion types. The 156 mice profiled expressed a total of 7,851 excursions. Each excursion was decomposed into 52 different measures of gait, movement, context, and more (47-49). Unsupervised hierarchical clustering with our published methods (47-49) identified 76 stereotypic, excursion types, which we call foraging “modules” (Fig. 5D and fig. S9A-F). Forty-five foraging modules were shared across Exploration and Foraging phases, with 13 unique to Exploration and 18 unique to Foraging (Fig. 5D). To validate that behaviorally similar excursions clustered into the same module, we visualized all excursions using PHATE (52) (Fig. 5E; 3D in Movie S2). Each dot represents a single excursion embedded by 52 features and colored by module. Tight clustering of same-module excursions confirms that modules capture related behaviors. We next tested for changes to module expression in Fto-Irx::hibE3 ^−/− mice.

For each mouse, the number of foraging excursions per module was quantified. Regression analysis controlling for sex and age revealed a significant genotype-by-module interaction during the Foraging Phase (Bonferroni corrected p-value <0.0001), but not during Exploration (Fig. 5F), confirming phase-dependent altered foraging in Fto-Irx::hibE3^−/− mice with an orthogonal method. Estimated marginal means analysis revealed module expression differences between ^−/− and ^+/+ mice, with post hoc tests identifying seven affected modules in young mice and two in aged mice (Fig. 5G).

In aged ^−/− mice, Module 11 expression decreased whereas Module 46 increased, with PHATE clustering and movement traces showing these modules represent distinct behavioral sequences. Module 46 involved less time at Pot 4 than Module 11 (Fig. 5H), contributing to the overall reduction in Pot 4 time observed in aged ^−/−mice (Fig. 5C). Young ^−/− mice exhibited a different subset of affected modules, driving age-related differences (Fig. 5G, S10). Thus, Fto-Irx::hibE3 regulates specific foraging modules across the lifespan, modulating age-dependent food patch interactions. As foraging in hibernators supports torpor preparation and recovery (53), these findings show that a single hibernation-linked CRE can coordinate metabolism, body weight, and foraging behavior.

Discussion

In this study, we used hibernator-homeotherm comparative genomics to define how non-coding CREs regulate metabolism and behavior. The Fto-Irx TAD showed a high accumulation of convergent genomic changes in hibernators, with hibernation-linked cis-elements overlapping accessible chromatin sites in the hypothalamus and forming regulatory contacts with Fto, Irx3, and Irx5. Irx3 emerged as a key transcription factor coordinating gene regulatory programs during different FFR metabolic states. Using five knockout mouse lines, we found that each hibernation-linked cis-element in the Fto-Irx TAD modulates Fto, Irx3, and Irx5 expression differently, with downstream effects on hundreds of genes. Deleting individual cis-elements shifted Fto, Irx3, and Irx5 expression from being largely stable in wild-type mice to exhibiting metabolic flexibility with more changes between fed and fasted states.

These molecular changes translated into distinct metabolic and behavioral phenotypes. Each cis-element uniquely influenced metabolic rate, body temperature, energy expenditure, and locomotion across torpor stages, with some also exhibiting sex-specific effects on high-fat diet-induced weight gain. Fto-Irx::hibE3 modulated naturalistic foraging by reducing novel food patch interactions during memory-influenced foraging in aged mice and altering a subset of foraging excursion modules across the lifespan. Previous studies linked CREs to anatomical evolution (11-13, 15); our findings demonstrate their roles in shaping complex metabolic and behavior phenotypes.

Homeotherms rely on stable metabolism and foraging to survive seasonal scarcity, whereas hibernators adapt through metabolic flexibility and energy conservation (5-9). We identified specific TADs, including the Fto-Irx TAD, as key drivers of metabolic divergence between hibernators and homeotherms due to their enrichment in convergent accelerated evolution sites (pHibARs) and deletions (pHibDELs). These regions serve as prime candidates for further studies on cis-element contributions to metabolic adaptations and their implications for human health. Whereas TADs are largely conserved, their internal loops evolve more rapidly and contribute to species-specific gene expression differences (20-23). Within the Fto-Irx TAD, we identified 80 pHibARs and 37 pHibDELs, marking cis-elements conserved in homeotherms but altered in hibernators. These elements likely function in diverse tissues, cell types, and physiological contexts, influencing Fto, Irx3, and Irx5 gene networks. Our companion study found that most accelerated and deleted regions in hibernators reflect loss-of-function effects on cis-elements conserved across homeotherms (29). Here, knockout mouse data show the effects of eliminating these sites in mice. Our findings support expanded studies of pHibAR- and pHibDEL-impacted cis-elements, individually and in combination, to better understand loss-of-function effects in hibernators and refine genetic models of mammalian metabolic regulation.