Blunted Refeeding Response and Increased Locomotor Activity in Mice Lacking FoxO1 in Synapsin-Cre–Expressing Neurons Open Access

The Columbia University Animal Care and Utilization Committee approved all procedures. Normal chow diet included 62.1% of calories from carbohydrates, 24.6% of calories from protein, and 13.2% of calories from fat (PicoLab rodent diet 20, 5053; Purina Mills); high-fat diet (HFD) included 20% of calories from carbohydrates, 20% of calories from protein, and 60% of calories from fat (D12492; Research Diets). We measured weight and length to calculate BMI, and we estimated body composition by nuclear magnetic resonance (Bruker Optics). We generated Syn-specific FoxO single or triple knockouts by mating Syn-Cre transgenic mice with Foxo1^flox/flox mice or Foxo1^flox/flox3a ^flox/flox4^flox/flox mice (19) and genotyped them as previously described (18). We excluded from analyses knockout mice that showed widespread recombination because of stochastic embryonic expression of Syn-Cre. We used cohorts of adult male mice, with 6–8 mice per genotype in most experiments (unless otherwise noted). We used the Rosa-Gfp reporter mice (B6;129-Gt(ROSA)26Sor^tm2sho/J; JAX Stock Number 004077) for lineage-tracing experiments.

We measured food intake with feeding racks (Firma Wenzel). For refeeding experiments, we habituated mice to feeding racks for 3 days, fasted them for 18 h, placed feeding racks 2 h after the start of the light phase, and measured food intake thereafter. We used a TSE Labmaster Platform (TSE Systems) for indirect calorimetry and activity measurements (17). For mice receiving adenovirus injection, we measured the wheel-running activity after the established protocol (20). Briefly, wheel-running activity was monitored remotely by Vitalview (Minimitter, Bend, OR), with counts collected in 10-min bins, and visualized using double-plotted actograms. We measured blood glucose by the One-Touch Ultra meter (LifeScan, Milpitas, CA), insulin and leptin were measured by ELISA, glucagon was measured by radioimmunoassay (Linco Research, St. Charles, MO), plasma free fatty acids and cholesterol were measured by nonesterified fatty acid hazard ratio and cholesterol-E test reagents, respectively (Wako Chemicals, Richmond, VA), and triglycerides were measured by serum triglyceride determination kit (Sigma-Aldrich, St. Louis, MO). Body composition was determined using Bruker Minispec nuclear magnetic resonance (Bruker Optics, Billerica, MA). We performed euglycemic-hyperinsulinemic clamps as previously described (21).

We perfused mice with saline and then with 4% paraformaldehyde. We froze brains in Tissue-Tek O.C.T. Compound (Sakura) and cut 30-μm–thick coronal sections for green fluorescent protein (GFP)-specific immunohistochemistry (Molecular Probes/Invitrogen). We acquired images with a Nikon eclipse microscope.

We extracted RNA with Trizol (Invitrogen) and performed quantitative PCR using SYBR Green I (Roche). Primer sequences are available on request. We used equal amounts of total RNA for reverse transcription and measured threshold cycle for each gene. β-Actin was used as an internal control. Data were quantified by standard delta ratio threshold cycle method.

We extracted MBH in 0.1 N HCl. We measured β-endorphin and α-melanocyte–stimulating hormone (MSH) by radioimmunoassay (17).

We analyzed data with Student t test, one-way ANOVA, or two-way ANOVA. P < 0.05 was considered statistically significant (*P < 0.05; **P < 0.01; ***P < 0.001).

To generate neuron-specific FoxO1 knockout mice, we crossed Foxo1^lox/lox and Syn-Cre mice. Cre-mediated deletion of the loxP-flanked Foxo1 exon 2 resulted in null Foxo1 alleles in Synapsin-expressing neurons (hereafter called Syn-Foxo1). To assess the distribution of Foxo1 ablation in the CNS, we introduced a reporter allele that encodes GFP as a marker of Cre-mediated recombination. GFP immunohistochemistry revealed Cre-mediated recombination across the brain, including cortex, hippocampus, and multiple hypothalamic nuclei. The frequency of Syn-Cre–mediated recombination varied in different brain regions. Hippocampus and dorsal medial nucleus showed the highest fractions of recombined cells, whereas cortex, ventral medial nucleus, and paraventricular hypothalamic nucleus had fewer GFP⁺ cells. The arcuate nucleus showed limited recombination, and the suprachiasmatic nucleus (SCN) showed little if any recombination (Fig. 1A). To provide a quantitative assessment of Foxo1 deletion, we assayed recombination using genomic DNA extracted from various brain regions. These data are consistent with the immunohistochemistry and also show extensive recombination in the brain stem, but not in peripheral tissues (Fig. 1B).

FoxO1 ablation in MBH AgRP/NPY and POMC neurons has striking effects on energy homeostasis and peripheral metabolism (17,18). To assess potential contributions of FoxO1 ablation in these neurons to the overall phenotype of Syn-Foxo1 mice, we determined the extent of Syn-Cre–mediated recombination in these two neuronal populations. To this end, we generated double-transgenic mice in which neurons undergoing Syn-Cre–mediated recombination was labeled red by a Rosa-Tomato allele (Syn-Tom) (22), whereas NPY/AgRP or POMC neurons were labeled green by Npy-Gfp or Pomc-Gfp transgenes, i.e., Syn-Tom;Npy-Gfp (Fig. 2A–E) and Syn-Tom;Pomc-Gfp (Fig. 2F–J), respectively. Double fluorescence demonstrated that Npy-labeled or Pomc-Gpf–labeled neurons had minimal overlap with Syn-Tom neurons (Fig. 2A, B, F, and G). We quantified the findings using fluorescence-activated cell sorting in which double-labeled neurons were distinct from single-labeled neurons (Fig. 2K and L, upper and lower right quadrants, respectively). Syn-Tom neurons accounted for ∼20% of NPY neurons, ∼20% of POMC neurons (Fig. 2M), and ∼40% of hypothalamic neurons (Fig. 2N). We estimated that NPY and POMC neurons combined constitute <5% of total sorted MBH neurons. Thus, we conclude that Syn-Cre primarily targets neurons other than NPY and POMC in the hypothalamus.

We evaluated Syn-Foxo1 mice by measuring plasma metabolites and energy expenditure with different diets. In chow-fed animals, we did not detect differences in body weight and composition, or in plasma triglycerides (Table 1). Measurements of respiratory exchanges failed to reveal alterations of Vo₂ and Vco₂, indicating that basal energy homeostasis was unaltered (data not shown).

However, when we measured the response to fasting and refeeding, we found that Syn-Foxo1 mice fed significantly less after an overnight fast (Fig. 3A). Specifically, during the 6-h refeeding, wild-type and knockout mice consumed 0.3831 and 0.2921 kcal/g body weight (calories normalized by body weight), respectively. To explain this observation, we measured hormonal sensitivity. Serum leptin levels were significantly decreased in fasted and ad libitum–fed Syn-Foxo1 mice (Fig. 3B). We examined leptin signaling by immunohistochemistry with antiphospho-Stat3 (pStat3) antibody in refed mice. The number of cells displaying pStat3 immunoreactivity and the intensity of the signal increased nearly two-fold in Syn-Foxo1 mice (Fig. 3C). Next, we examined whether the increase in leptin signaling was cell-autonomous. To this end, we generated Syn-Foxo1 mice bearing a Rosa-Tom allele and measured pStat3 immunoreactivity in response to refeeding after an overnight fast. The expectation of this experiment was that if the decreased rebound food intake was attributable to increased leptin sensitivity in FoxO1-deficient neurons, then knockout mice should show increased pStat3 signal (green fluorescence) in neurons that were marked by FoxO1 ablation (red fluorescence), giving rise to yellow fluorescence. We observed that most of the pStat3 signal colocalized with recombined neurons, indicating that FoxO1 ablation increases leptin-induced pStat3 generation in a cell-autonomous manner (Fig. 3D).

Hypothalamic amino acid signaling regulates satiety and feeding behaviors (23). We measured amino acid sensitivity by immunostaining with antibody to the mTOR substrate, phospho-ribosomal protein S6 (pS6). Syn-Foxo1 mice had increased pS6, predominantly in dorsal medial nucleus (Fig. 3E), a site of Syn-Cre–dependent recombination (Fig. 1A). Using Syn-Tom to label recombined neurons, we sought to determine whether the increase of pS6 was caused by FoxO1 deletion in a cell-autonomous or cell-nonautonomous manner. We saw that increased pS6 staining occurred equally in FoxO1-deleted and FoxO1-intact dorsal medial nucleus neurons, indicating that the effect of FoxO1 ablation is at least partly cell-nonautonomous (Fig. 3F).

The CNS further regulates pancreatic hormone secretion, especially through the adrenosympathetic system (24). We therefore measured insulin secretion induced by the β3-adrenergic receptor agonist, CL-316243. Activating β3-adrenergic receptor quickly increases insulin secretion by pancreatic β-cells and free fatty acid release by adipocytes. After CL-316243 injection, Syn-Foxo1 mice showed consistently lower glucose (Fig. 4A), secondary to increased insulin release (Fig. 4B). Serum free fatty acid levels were comparable with those of controls (Fig. 4C), indicating that the effect of the Foxo1 mutation on insulin release is not secondary to increased lipolysis with β3-adrenergic receptor agonist stimulation. Next, we examined insulin secretion in static incubations of purified pancreatic islets and found small but significant increases induced by low (5 mmol/L) and high (25 mmol/L) glucose concentrations (Fig. 4D). Total insulin content and secretion in response to KCl-induced depolarization were comparable (Fig. 4D). Islet size tended to be larger in Syn-Foxo1 animals, but not significantly larger (Fig. 4E). These data indicate that FoxO1 ablation in Syn-Cre neurons increased sympathoadrenal responses. Using the Syn-Tom reporter, we found no evidence of Cre-mediated recombination in the pancreas of Syn-Foxo1 animals (data not shown), effectively ruling out a direct effect of FoxO1 ablation in β-cells (25). The increased insulin secretion from β-cells in response to adrenosympathetic stimuli did not alter fasted and fed glucagon levels (Fig. 4F). However, Syn-Foxo1 mice exhibited significantly lower circulating insulin levels (Table 1), indicating increased insulin sensitivity.

Male Syn-Foxo1 mice had increased locomotor activity during the dark phase of the light cycle (Fig. 5A). Previous studies have shown that FoxO1 ablation in POMC neurons increased locomotor activity because of increased α-MSH processing and melanocortin signaling (17). Given that only a minority of POMC neurons were targeted in Syn-Foxo1 mice (Fig. 2), the cause of increased locomotion in Syn-Foxo1 mice is likely different from that for Pomc-Foxo1 knockout mice. In fact, α-MSH and β-endorphin levels in the MBH were comparable between wild-type and Syn-Foxo1 mice (Fig. 5B and C), indicating that processing of POMC-derived hormones occurs normally in the knockout mice. The α-MSH is rapidly metabolized in vivo by prolyl-carboxypeptidase (Prcp) (26). The ratio of α-MSH to β-endorphins in Syn-Foxo1 mice was slightly higher compared with that of wild-type mice (0.471 ± 0.025 vs. 0.419 ± 0.014, respectively; P = not significant), consistent with the possibility that slower inactivation of α-MSH may contribute to increased melanocortinergic tone. α-MSH signaling exerts negative feedback regulation on its peptide processing, reducing Pcsk1 expression (17). Pcsk1 and Prcp expression were significantly reduced in the hypothalamus of Syn-Foxo1 mice during the phase of increased locomotor activity (Fig. 5E and F). These data suggest that increased melanocortin signaling in Foxo1 knockout mice contributes to the increased locomotor activity.

We tested this hypothesis by a gain-of-function approach. We reasoned that if FoxO1 loss-of-function increased locomotor activity, then a gain-of-function mutant should reverse this phenotype. We injected adenovirus encoding the constitutively active FoxO1-ADA mutant (9) into the MBH. After this manipulation, we observed increased hypothalamic Foxo1 when compared with mice that received control GFP adenovirus (Fig. 5G). After recovery, we individually housed mice in activity hotels provided with a running wheel and monitored wheel rotations to assess locomotor activity during the 24-h light–dark cycle. Mice injected with FoxO1-ADA showed less locomotor activity during the dark phase (Fig. 5H). Of note, as a negative control, mice with FoxO1-ADA injection into the SCN—a site spared by Syn-Cre recombination—demonstrated no change in activity patterns (data not shown). The SCN is the site of the master clock of the brain, and it known for its role in controlling daily rest–activity rhythms. This result indicates that the activity phenotype is not caused by changes in SCN.

We next investigated whether the comparatively mild phenotype seen in Syn-Foxo1 mice could be explained by compensatory actions of the two closely related isoforms, FoxO3 and FoxO4 (19). To this end, we generated triple-knockout mice using Syn-Cre (Syn-Foxo1,3,4). mRNA measurements in hypothalamus and hippocampus from wild-type and Syn-Foxo1,3,4 mice revealed significant reductions of Foxo1 and Foxo3, whereas Foxo4 was undetectable (Fig. 6A). Syn-Foxo1,3,4 and wild-type mice showed comparable body weight when fed the normal chow diet (Fig. 6B). When fed HFD after weaning, triple-knockout mice gained less weight compared with wild-type mice (Fig. 6C), likely because of reduced fat accumulation, as indicated by body composition studies (Fig. 6D).

Next, we determined whether the knockout protected from weight gain induced by HFD feeding. We used mice backcrossed onto B6 background to minimize individual variation. Wild-type mice rapidly gained weight when fed HFD, whereas knockout mice consistently maintained a significantly lower body weight (Fig. 6E). We analyzed body composition after 4 months of HFD feeding and found that knockout mice had significantly less fat mass compared with wild-type mice (Fig. 6F). We compared these mice with triple Syn-Foxo1,3,4 mice, and we found that the extent of protection from HFD-induced fat accumulation was the same. The triple-knockout mice fed HFD had lower body weight than wild-type mice (33.8 ± 0.8 g; knockout mice: 31.3 ± 0.7 g; P = 0.035). Then, we analyzed energy balance in this cohort by indirect calorimetry. Interestingly, knockout mice had significantly increased Vo₂ (Fig. 6G), locomotor activity, and energy expenditure during both light and dark phases (Fig. 6H and I). Knockout mice had respiratory quotients comparable with those of wild-type mice (Fig. 6J). In conclusion, FoxO1 ablation in Syn-Cre–expressing neurons increased energy expenditure in HFD-fed animals and partly offset diet-induced weight gain.

POMC and AgRP neurons in the arcuate nucleus are critical CNS components for controlling metabolic homeostasis. We have reported that knocking out FoxO1 in either POMC or AgRP neurons improves energy and glucose homeostasis (17,18). Curiously, we found that only few POMC and AgRP neurons undergo Syn-Cre–mediated recombination (Fig. 2), yet Syn-Foxo1 mice have an improved metabolic profile (Fig. 3). To distinguish the collective contribution of POMC and AgRP neurons from that of other CNS neurons, we characterized the phenotype of mice with combined FoxO1 knockout in both POMC and AgRP neurons. In contrast to Syn-Foxo1 mice, Pomc;Agrp-Foxo1 mice fed normal chow exhibited a lean phenotype, with lower body weight (Fig. 7A), decreased fat mass (Fig. 7B), and increased lean mass (Fig. 7C). Furthermore, during the 24-h light cycle Pomc;Agrp-Foxo1 mice showed significantly reduced Vo₂ (Fig. 7D) and elevated respiratory quotient but unaltered locomotor activity (Fig. 7E) compared with wild-type mice.

Considering this energy partitioning phenotype, we analyzed hepatic glucose production in Pomc;Agrp-Foxo1 mice by euglycemic-hyperinsulinemic clamp studies. Glucose infusion rates trended higher in Pomc;Agrp-Foxo1 mice (Fig. 7G), likely because of increased rates of glucose disposal (Fig. 7H), but neither difference reached statistical significance. Interestingly, hepatic glucose production was unchanged (Fig. 7I), unlike in Agrp-Foxo1 mice (18). Based on these results, we conclude that FoxO1 function in POMC and AgRP has concordant effects to increase energy expenditure but has opposing actions on hepatic glucose production, as intimated by a previous study (21). The metabolic functions of FoxO1 in Syn-Cre–expressing neurons appear to be distinct from those in AgRP and POMC neurons.

In this work, we describe the metabolic and energy balance phenotypes associated with deletion of transcription factor FoxO1, a key metabolic sensor, in Syn-Cre–expressing neurons that are found scattered in the cortex, hippocampus, hypothalamus, and brain stem. Syn-Foxo1 mice display a catabolic energy balance phenotype associated with increased sensitivity to hormone and nutrient signaling within the CNS and increased locomotor activity, possibly attributable to low α-MSH turnover leading to enhanced melanocortin signaling. The heterogeneity of Syn-Cre–mediated recombination, as assessed by lineage tracing, limits our ability to assign the observed phenotype to a specific neuronal subtype. Nonetheless, we used three alternative approaches to begin to map the functional neuroanatomy of FoxO functions in the CNS. First, lineage-tracing approaches show that only ∼20% of NPY/AgRP and POMC neurons display Syn-Cre–dependent recombination, effectively ruling a major contribution of these key neuronal populations to the energy balance and metabolic phenotypes of Syn-Foxo1 mice. We supported this conclusion by analyzing mice with combined ablations of FoxO1 in AgRP and POMC neurons, and by showing that the phenotypes differ. Second, concurrent ablation in Syn-Cre neurons of the three insulin-regulated FoxO isoforms, FoxO1, FoxO3, and FoxO4, also conferred a catabolic energy expenditure phenotype, with improved glucose metabolism and resistance to HFD, demonstrating that FoxO functions likely overlap, similar to previous observations (6,27). Third, we used a gain-of-function approach to map the increased activity phenotype to the MBH. We conclude that in addition to their established role in neuropeptide synthesis and processing, FoxO1, FoxO3, and FoxO4 have anabolic functions in other areas of the CNS.

A new finding of the present work is the increase in insulin secretion observed in response to a β3-AR agonist. This likely reflects an effect of FoxO1 on sympathetic activity, possibly because of loss of FoxO1 function in the brain stem (28,29). Our conclusion is that the integrative neuronal function of FoxO1 extends beyond the arcuate nucleus, suggesting that broad-based inhibition of FoxO1 function in the CNS can promote hormone sensitivity and prevent metabolic disease.

Key components of insulin and leptin signaling pathways have been studied using pan-neuronal ablation approaches (30). Our data emphasize that there is no such thing as a truly pan-neuronal knockout, and that caution should be exerted in comparing knockouts performed with different Cre drivers. Mice with Nestin-Cre–driven insulin receptor knockout (31) develop diet-induced obesity with increases in body fat, mild insulin resistance, increased plasma leptin, and hypertriglyceridemia, demonstrating that neuronal insulin receptor function affects energy balance. With the caveats stated, it is reassuring that the overall direction of the metabolic changes is consistent with the opposing cellular actions of insulin receptor and FoxO (5), especially because the latter acts as a relay node for multiple extracellular signals (32). Similarly, Syn-Cre neuron–specific leptin receptor deletion gave rise to obesity associated with increased plasma glucose, insulin, and leptin in a manner that was directly related to the extent of hypothalamic leptin receptor ablation (33). These data lend support to previous work indicating that FoxOs act to integrate insulin with leptin signaling by virtue of their ability to antagonize STAT3 signaling in response to decreasing levels of phospho-Akt (13). Conversely, neuron-specific knockout or knockdown of protein tyrosine phosphatase-1b, a negative regulator of insulin and leptin signaling, improves insulin and leptin sensitivity and protects from diet-induced obesity (34,35). Nonetheless, there are telling differences between the phenotypes resulting from protein tyrosine phosphatase-1b and FoxO ablation, for example, with regard to peripheral metabolism and locomotor activity. These data indicate that FoxO relays multiple pathways acting not only through Akt but also possibly through other kinases. The nature of these kinases and their target sites on FoxO1, as well as the physiologic consequences of their activation on FoxO1 function, represent important areas for further research. In addition to insulin and leptin, FoxO1 mediates the effects of other key hormones on metabolism, such as ghrelin (36). We found in our studies that the diverse, if internally consistent, effects of FoxO ablation on energy homeostasis impinge equally on insulin and leptin signaling, sensitizing the body to both. More challenging is the question of which cellular mechanisms and neuronal relay systems are used to effect these actions.

In this study, we found that mice with FoxO1 knockout in Syn-Cre–expressing neurons have increased melanocortin signaling tone, manifested as increased locomotor activity. The finding that constitutively active FoxO1 expression in MBH, but not in SCN, reduces locomotor activity highlights the physiological importance of FoxO1 function in MBH. MBH includes several neuronal populations that are critical for metabolic regulation. For example, FoxO1 activation in POMC and AgRP neurons results in anabolic function, but the underlying effectors and molecular mechanisms are starkly different (12–15,17,18). In POMC neurons, FoxO1 knockout promotes leanness, decreases food intake, increases leptin sensitivity, and increases melanocortin signaling. We have proposed that the key effector of these actions is carboxypeptidase, the protease responsible for cleaving POMC to α-MSH and β-endorphin, which are normally suppressed by FoxO1 (17). In AgRP neurons, FoxO1 regulates expression of AgRP itself (13) and also of the orphan G-protein-coupled receptor Gpr17, whose pharmacological modulation affects food intake and, hence, shows promise as a target of antiobesity drugs (18). Unlike ablation in POMC neurons, ablation in AgRP neurons also had a substantial effect on hepatic glucose production, adding one important potential therapeutic benefit to CNS-targeted FoxO inhibition (18). The ventral medial nucleus is also a site of FoxO1 action (16). Mice lacking FoxO1 in ventral medial nucleus steroidogenic factor-1–expressing neurons are lean because of increased energy expenditure and have increased insulin sensitivity, suggesting that the overarching role of FoxO as a mediator of energy balance extends to multiple CNS sites.

The adult mammalian brain also contains neural stem cells that can give rise to neurons, astrocytes, and oligodendrocytes (37). FoxO1, FoxO3, and FoxO4 cooperatively regulate neural stem cell homeostasis and play important roles in neural stem cell proliferation and renewal, possibly through their regulation of cellular metabolism (27).

FoxO isoforms include FoxO1, FoxO3, and FoxO4. Based on estimates from quantitative PCR analysis, FoxO1 is the major isoform expressed in the brain. FoxO3 is expressed at lower levels, whereas FoxO4 expression is virtually undetectable. These different FoxO isoforms could have functional redundancy, and other isoforms may compensate for the loss of one isoform. For example, this is the case in the liver, where triple ablation of Foxo genes causes more pronounced effects than the single FoxO1 knockout with fasting hypoglycemia, increased glucose tolerance and insulin sensitivity, and decreased plasma insulin levels (6). This paradigm appears to hold true in the brain, because our data show that triple FoxO knockouts in the brain are better protected from weight gain induced by HFD, indicating potentially redundant contributions of other FoxO genes to CNS regulation of energy balance.

Understanding the integrative function of FoxO1 in the CNS is crucial for harnessing its therapeutic potential for metabolic diseases. The data presented in this study are reassuring in this regard because we observed no untoward effect of single or triple FoxO ablation. Further, we saw that the effects linked to Syn-Cre ablation are qualitatively similar to those observed in more restricted knockouts (16–18), allaying fears that broad-based FoxO inhibition might adversely affect cellular survival, behavior, and overall energy homeostasis.

This study was supported by National Institutes of Health grants DK58282 and DK63608 (Columbia Diabetes Research Center), American Diabetes Association mentor-based fellowship, and Berrie Fellowship Award (H.R.).

No potential conflicts of interest relevant to this article were reported.

H.R. designed and conducted experiments, analyzed data, and wrote the manuscript. L.P.-M. designed and conducted experiments, analyzed data, and wrote the manuscript. R.G.-J. conducted glucose clamp experiments. T.Y.L. and J.Y.K.-M. conducted experiments. G.H. analyzed data. S.L.W. and R.S. designed experiments and analyzed data. D.A. designed experiments, analyzed data, and wrote the manuscript. D.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank members of the Accili laboratory for insightful data sessions and for critical reading of the manuscript. The authors thank Thomas K. Kolar and Ana M. Flete Castro (Department of Medicine, Columbia University) for excellent technical support. The authors thank Dr. Matthew Butler (Department of Psychology, Columbia University) for helping with wheel-running experiments.