PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Cell Metab. Author manuscript; available in PMC 2010 April 1.
Published in final edited form as:
PMCID: PMC2790375
NIHMSID: NIHMS109125

Critical role for hypothalamic mTOR activity in energy balance

Summary

The mammalian target of Rapamycin (mTOR) promotes anabolic cellular processes in response to growth factors and metabolic cues. The TSC1 and TSC2 tumor suppressors are major upstream inhibitory regulators of mTOR signaling. Mice with Rip2/Cre-mediated deletion of Tsc1 (Rip-Tsc1cKO mice) developed hyperphagia and obesity, suggesting that hypothalamic disruption (for which Rip2/Cre is well known) of Tsc1 may dysregulate feeding circuits via mTOR activation. Indeed, Rip-Tsc1cKO mice displayed increased mTOR signaling and enlarged neuron cell size in a number of hypothalamic populations, including Pomc neurons. Furthermore, Tsc1 deletion with Pomc/Cre (Pomc-Tsc1cKO mice) resulted in dysregulation of Pomc neurons and hyperphagic obesity. Treatment with the mTOR inhibitor, rapamycin, ameliorated the hyperphagia, obesity, and the altered Pomc neuronal morphology in developing or adult Pomc-Tsc1cKO mice, and cessation of treatment reinstated these phenotypes. Thus, ongoing mTOR activation in Pomc neurons blocks the catabolic function of these neurons to promote nutrient intake and increased adiposity.

INTRODUCTION

The mammalian target of rapamycin (mTOR), an evolutionarily conserved serine-threonine kinase, promotes anabolic cellular processes such as protein synthesis in response to growth factors, nutrients (amino acids and glucose), and stress (Biondi et al., 2004; Wullschleger et al., 2006). mTOR exists in two distinct complexes, mTORC1 and mTORC2. TSC1 and TSC2 are tumor suppressors and their gene products form a stable complex. The major function of TSC1/TSC2 tumor suppressors is to inhibit mTORC1 (Inoki et al., 2005), and mutation in either TSC1 or TSC2 leads to constitutive activation of mTORC1; in patients, this results in multi-organ hamartomas. The TSC1/TSC2 dimer has GTPase activating protein (GAP) activity specifically towards the Rheb GTPase, which directly binds and activates TORC1 but not TORC2 (Long et al., 2005; Sancak et al., 2007; Yang et al., 2006). Furthermore, amino acids and cellular energy levels stimulate TORC1, but not TORC2. For the purposes of this report, subsequent statements about mTOR refer to TORC1.

Within the TSC1/2 complex, the TSC1 subunit stabilizes and regulates the GAP activity of TSC2, which inactivates Rheb. During insulin signaling, the PI3K-Akt pathway activates mTOR by phosphorylating and thereby inhibiting TSC2 (Inoki et al., 2002; Manning et al., 2002; Nguyen et al., 2006); the Akt-dependent phosphorylation of PRAS40 may also contribute to mTOR activation (Oshiro et al., 2007; Sancak et al., 2007; Vander Haar et al., 2007; Wang et al., 2007). Energy limitation, such as glucose deprivation, strongly inhibits mTOR, and the glucose deprivation-induced mTOR inhibition is compromised in TSC mutant cells. AMPK, a cellular energy sensor, directly phosphorylates TSC2 to inhibit mTOR. AMPK can also inhibit mTORC1 by phosphorylating Raptor, an essential mTOR1 subunit. Thus, the TSC1/2 complex integrates multiple upstream signals to modulate mTOR activity (Gao et al., 2002; Inoki et al., 2003; Inoki et al., 2002; Manning et al., 2002; Wullschleger et al., 2006).

Although the PI3K-Akt cascade regulates many cellular processes by phosphorylating numerous substrates, including FOXO and AS160, and plays an important role in the hypothalamic control of feeding (Belgardt et al., 2008; Brunet et al., 1999; Kitamura et al., 2006; Kops et al., 1999; Plum et al., 2006; Sano et al., 2003), the function of mTOR activation for appetite regulation by the CNS has not been thoroughly investigated. Recent findings have suggested a potential role for mTOR as a cellular fuel sensor in hypothalamic circuits that regulate energy balance, especially in orexigenic AgRP-expressing neurons, where feeding status prominently regulates mTOR activity (Cota et al., 2006). Increased hypothalamic availability of leucine stimulates mTOR activity and reduces food intake in an mTOR-dependent manner. Hypothalamic mTOR activity is also required for the suppression of feeding by leptin. Thus, mTOR contributes to acute sensing of the nutritional milieu in the regulation of feeding by hypothalamic neural circuits. In contrast, the hyperactivation of mTOR in peripheral metabolic tissues such as muscle and liver during nutritional excess in high-calorie feeding and obesity also contributes to insulin resistance and metabolic dysfunction (Korsheninnikova et al., 2006; Miller et al., 2008). Such chronic activation of mTOR could similarly dysregulate the energy-sensing hypothalamic neurons that control feeding, and perturb whole body energy balance.

We generated mice conditionally lacking Tsc1 in both pancreatic β cells and hypothalamic neurons (Rip-Tsc1cKO mice) using Rip2/Cre, and also lacking Tsc1 specifically in Pomc neurons (Pomc-Tsc1cKO) using Pomc/Cre. Our studies demonstrate hyperphagic obesity in both the Rip-Tsc1cKO and Pomc-Tsc1cKO mice, due to uncontrolled chronic activation of mTOR in both Rip neurons and Pomc neurons. These results suggest that the TSC-mTOR pathway plays crucial roles in the function of appetite-suppressing neural circuits; dysregulated activation of the mTOR pathway in the CNS might be a potential mechanism leading to the development of obesity.

RESULTS

Hyperphagic obesity in Rip-Tsc1cKO mice

To study the in vivo function of mTOR activation in metabolism, we set out to analyze mice lacking Tsc1, a major upstream negative regulator of mTOR. Since conventional knockout of Tsc1 leads to embryonic lethality (Kobayashi et al., 2001; Kwiatkowski et al., 2002; Murakami et al., 2004), we generated a loxP-flanked floxed Tsc1 allele (Tsc1lox) to permit tissue specific Tsc1 knockout. Tsc1lox/lox homozygotes (Meikle et al., 2005; Uhlmann et al., 2002) were mated with the Cre transgenic line driven by the rat insulin promoter 2 (Rip) (Choudhury et al., 2005; Kubota et al., 2004; Lin et al., 2004; Nguyen et al., 2006; Stiles et al., 2006). All subsequent experiments were performed using Rip-cre/Tsc1lox/lox (Rip-Tsc1cKO) mice and littermate Tsc1lox/lox (a.k.a. control/wild-type) mice. Western blotting confirmed markedly reduced Tsc1 expression in isolated islets from Rip-Tsc1cKO mice (data not shown). As expected, Tsc1 deletion mediated by Rip/Cre results in increased mTOR activity in β cells, along with increased β cell size and islet mass, and higher insulin content and secretion (Mori et. al unpublished observations).

Interestingly, in addition to their β cell phenotype, Rip-Tsc1cKO male mice displayed increased body weight compared with controls by 6 weeks of age (Figure 1A). By 12 weeks of age, Rip-Tsc1cKO mice were 36% heavier than control animals. The body weight of the Rip-Tsc1cKO female mice was also significantly higher than that of the controls (data not shown). Furthermore, 24 week-old Rip-Tsc1cKO mice displayed increased whole body fat content and increased adipocyte cell size compared with that of wild-type mice (Figure 1B, 1C). No difference in length was observed between the two genotypes (Supplementary Figure 1A). Consistent with the increased adiposity of Rip-Tsc1cKO mice, serum leptin levels were significantly elevated relative to controls in both fasted and fed states in Rip-Tsc1cKO mice at 4 weeks of age, at which the Rip-Tsc1cKO mice have normal body weight (Figure 1D). To examine leptin action in Rip-Tsc1cKO mice, we assessed the effect of administering leptin peripherally for 2 days. This treatment markedly reduced food intake and body weight in control animals (Figure 1E, 1F). In contrast, leptin had little effect on food intake in the Rip-Tsc1cKO mice, suggesting impairment of the leptin-regulated neural satiety circuits in these Rip-Tsc1cKO mice.

Figure 1
Hyperphagic obesity in Rip-Tsc1cKO mice

In order to understand the basis for this increased adiposity, we examined food intake from 4–6 weeks of age, when the difference in body weight between the two genotypes began to appear. We found that food consumption by Rip-Tsc1cKO mice was greater than that of the control animals (Figure 1G), suggesting that the obese phenotype in Rip-Tsc1cKO mice results from increased food intake. Consistently, pair-fed Rip-Tsc1cKO mice displayed similar body weights compared to wild-type mice (Figure 1H). Thus, Rip-Tsc1cKO mice develop hyperphagic obesity, suggesting an important role for mTOR in the regulation of food intake and body weight.

Hypothalamic Tsc1 deletion in Rip-Tsc1cKO mice

Rip-Cre mice express Cre in the hypothalamus during development, resulting in the excision of floxed genes in a diverse set of hypothalamic neurons (Choudhury et al., 2005; Kubota et al., 2004; Lin et al., 2004; Nguyen et al., 2006; Stiles et al., 2006). Although immunoblotting revealed no detectable alteration in Tsc1 protein levels in the whole brain of Rip-Tsc1cKO mice (Supplementary Figure 1B), analysis of the dissected hypothalamus demonstrated reduced Tsc1 protein levels, supporting Cre-mediated deletion of Tsc1 in the hypothalamus of these animals (Figure 2A). Tsc2 expression was also decreased when Tsc1 was deleted (Figure 2A), consistent with the proposed role for Tsc1 in stabilizing Tsc2 (Chong-Kopera et al., 2006; Meikle et al., 2007). We determined the activation status of mTOR by immunoblotting for phosphorylated S6 (pS6). The mTOR-dependent activation of S6K results in S6 phosphorylation, which thus serves as a convenient readout for mTOR activity (Flotow and Thomas, 1992; Pende et al., 2004). Immunoblotting revealed markedly increased pS6 levels in the hypothalamus of Rip-Tsc1cKO mice (Figure 2A). Immunofluorescence analysis confirmed elevated pS6-immunoreactivity (pS6-IR) in many neurons in a number of hypothalamic nuclei important for energy homeostasis (including the arcuate nucleus (ARC), ventromedial hypothalamic nucleus (VMH) and dorsomedial hypothalamic nucleus (DMH)), indicating that mTOR was activated in the hypothalamus of these animals (Figure 2B and Supplementary Figure 2). To test the mTOR-dependence of the increased pS6 in Rip-Tsc1cKO hypothalami, we treated the mice with rapamycin, a specific mTOR inhibitor. While not restored by the peripheral administration of 1mg/kg rapamycin (which ameliorated the effects of Tsc1 deletion in the β cell (Mori et. al unpublished data), higher (6 mg/kg) doses of rapamycin normalized pS6-IR in Rip-Tsc1cKO hypothalami, revealing the mTOR-dependency of the hypothalamic pS6 (Figure 2B).

Figure 2
Hypothalamic Tsc1 deletion and dysregulation of the melanocortin system in Rip-Tsc1cKO mice

Dysregulation of the melanocortin system in Rip-Tsc1cKO mice

The Rip-Tsc1cKO mice display high leptin levels and hyperphagia (Figure 1D, 1G). The inability of elevated leptin levels to mitigate hyperphagia in Rip-Tsc1cKO mice suggests the potential dysfunction of the neural systems that regulate body energy homeostasis in response to leptin in these animals. Quantitative real-time PCR (qPCR) analysis demonstrated that leptin receptor (Lepr) mRNA levels were increased in Rip-Tsc1cKO hypothalami (supplementary Figure 3), however, suggesting that the obesity of the Rip-Tsc1cKO animals was not due to attenuated Lepr expression. In order to identify potential contributing mechanisms underlying the altered feeding and energy balance of the Rip-Tsc1cKO mice, we examined mRNA expression of hypothalamic appetite-controlling neuropeptides that are known to be regulated by leptin (Coll et al., 2007; Flier, 2004). ARC neurons release orexigenic peptides, such as neuropeptide Y (NPY) and agouti-related peptide (AgRP), and anorexigenic peptides such as those derived from pro-opiomelanocortin (POMC). Pomc and Npy mRNA levels were dysregulated in ad libitum-fed Rip-Tsc1cKO mice, and fasting failed to decrease Npy mRNA levels in these mice; Agrp mRNA levels were normal and remained responsive to fasting in these animals (Figure 2C). Both increased Npy and decreased Pomc expression would predict increased feeding in the Rip-Tsc1cKO mice. The appropriate control of Agrp expression in the same neurons that display dysregulated Npy expression is consistent with previous findings demonstrating that the expression of Npy and Agrp are controlled by distinct upstream signals (Bates et al., 2003; Kitamura et al., 2006), and suggests an important role for hypothalamic mTOR in the regulation of Npy, but not Agrp. The dysregulation of neuropeptide expression in multiple populations of ARC neurons is consistent with Rip2/Cre-mediated deletion in multiple populations of hypothalamic neurons, including both anorectic and orexigenic cell types. Other neuronal populations are likely affected, as well, and the anorectic action of the majority of the neurons in this region of the hypothalamus presumably results in a net orexigenic response to the dysregulation of neurons throughout the ARC. These results suggest that uncontrolled mTOR activation blunts the physiological regulation of Pomc and Npy and that dysregulation of anorexigenic neurons, such as the Pomc cells, may contribute to the hyperphagic obesity of the Rip-Tsc1cKO mice.

We examined the somatic size and projections of the anorexigenic Pomc neurons by immunohistochemical analysis of ACTH, a neuropeptide derived from POMC. Visual inspection suggested the marked hypertrophy of ACTH-IR neuronal soma in the arcuate nucleus of the hypothalamus (ARC) (Figure 3A) along with the diminution of ACTH-IR neuronal projections to the PVN in Rip-Tsc1cKO mice compared to controls (Figure 3B). Thus ablation of Tsc1 by Rip2/Cre likely impairs signaling by Pomc neurons. In order to examine the mTOR-dependence of these effects, we treated animals with rapamycin for 2 weeks (from 2–4 weeks of age), using the dose (6 mg/kg, i.p.) that we had previously found effective in attenuating the increased hypothalamic mTOR activity in Rip-Tsc1cKO mice (Figure 2B). As shown in Figure 3A, this treatment largely corrected the obvious morphological alterations of ACTH-IR cells in the Rip-Tsc1cKO mice, as well as increasing the density of ACTH-IR projections to the PVN in Rip-Tsc1cKO mice (Figure 3B). Consistent with these histologic improvements, the body weight phenotype of the Rapamycin-treated Rip-Tsc1cKO mice improved (Figure 3C).

Figure 3
Rapamycin treatment normalized the morphological alterations of Pomc neurons in Rip-Tsc1cKO mice

Taken together, these data suggest a crucial role for Tsc1 in inhibiting mTOR activity and thereby preserving the normal regulation and function of hypothalamic neurons that modulate energy balance, and further suggest a potential contribution of pathologic changes promoted by mTOR activation in the Pomc neurons of the Rip-Tsc1cKO mice to the hyperphagic obesity of these animals.

Altered Pomc neurons in Pomc-Tsc1cKO mice

While the excision of conditional alleles in the hypothalamus by the Rip2/Cre transgene is well established, the neurons affected by this excision are diverse and poorly characterized, rendering it difficult to pinpoint a single neuronal mechanism underlying the obesity of Rip-Tsc1cKO mice. The alterations observed within the Pomc neurons of the Rip-Tsc1cKO mice suggest that these neurons represent one affected population and that mTOR activation in the Pomc neurons might contribute to the hyperphagic obesity of the Rip-Tsc1cKO mice, however. We thus generated mice harboring a conditional disruption of Tsc1 specifically in Pomc neurons (Pomc-Tsc1cKO mice). The hypothalami of these Pomc-Tsc1cKO animals displayed elevated pS6-IR independent of nutritional status in a discrete set of enlarged ARC neurons (Figure 4A). While the rapid export of neuropeptides from the neural soma in mice prevents the detection of all Pomc neurons by ACTH-IR, the number of ACTH-IR neurons did not differ between Pomc-Tsc1cKO and control animals (Figure 4B, 4D left graph). The ACTH-IR Pomc soma colocalized with the pS6-IR neurons in the Pomc-Tsc1cKO animals, and the size of the ACTH-IR neurons was significantly increased in the Pomc-Tsc1cKO animals, consistent with the known specificity of Pomc/Cre for these neurons as well as the increased mTOR activity that accompanies the deletion of Tsc1 (Figure 4B, 4D). The Pomc-Tsc1cKO mice also displayed decreased ACTH-IR axonal projections to the PVN (Figure 4C, 4D right graph) compared to control mice. Both the colocalization of strong pS6-IR in ACTH-IR neurons and the shift in the size of all ACTH-IR soma in the Pomc-Tsc1cKO mice (supplemental Figure 4A) suggests an efficient excision of Tsc1 in these animals. Thus, mTOR activation cell-autonomously promotes hypertrophy in Pomc soma and abnormal ATCH-IR neuronal projections in Pomc-Tsc1cKO mice.

Figure 4
Dysregulation of hypothalamic Pomc neurons in Pomc-Tsc1cKO mice

Hyperphagic obesity in Pomc-Tsc1cKO mice

We found that Pomc-Tsc1cKO mice weighed significantly more than control mice by 6 weeks of age (Figure 5A), and this increase in body weight reflected increased adiposity, as demonstrated by their increased fat pad mass and circulating leptin levels (Figure 5B, 5C, supplementary Figure 4C). To directly examine leptin action in Pomc-Tsc1cKO mice, we assessed the effect of administering leptin peripherally for 2 days. Though leptin treatment markedly reduced food intake and body weight in control animals, this effect was significantly attenuated in the Pomc-Tsc1cKO mice (Figure 5D, 5E). Furthermore, Pomc-Tsc1cKO mice consumed more food than controls during the period of highly divergent weight gain from 6–8 weeks of age (Figure 5F). These changes were not attributable to alterations in the HPA (hypothalamic-pituitary-adrenal) axis, as circulating corticosterone levels were similar in control and Pomc-Tsc1cKO animals at 4 and 24 weeks of age (supplementary Figure 4B). These data indicate that Tsc1 ablation in Pomc neurons alone induced hyperphagic obesity. Taken together, our results suggest that TSC-mTOR signals in Pomc neurons play a crucial role in the physiology of these neurons and in the regulation of energy homeostasis, and that anabolic mTOR signaling in the Pomc neurons attenuates the catabolic function of these neurons to promote energy uptake and storage.

Figure 5
Hyperphagic obesity in Pomc-Tsc1cKO mice

Rapamycin treatment of Pomc-Tsc1cKO mice

In order to examine roles for altered development and ongoing mTOR activity in the phenotypes of the Pomc-Tsc1cKO animals, we chronically treated two groups of animals with centrally effective doses of rapamycin (6 mg/kg, i.p.; see Figure 3B) or vehicle (Figure 6A, supplemental Figure 5A). In the first group (Treatment A), 2-week-old Pomc-Tsc1cKO and wild-type mice were treated with rapamycin for 5 weeks (from 2 to 7 weeks of age), and then switched to vehicle treatment for an additional 5 weeks (from 7 to 12 weeks of age). Conversely, for Treatment B, the animals were treated with vehicle for the initial 5 weeks, and then switched to rapamycin treatment for the second 5-week period. For each treatment group, some animals were sacrificed at 7 or 12 weeks of age for the analysis of fat pads and hypothalamic histology. We examined average food intake from 4–7 weeks of age and again after the switch in treatment from 7–12 weeks of age.

Figure 6
Ongoing rapamycin treatment ameliorates the hyperphagia and obesity of Pomc-Tsc1cKO mice

The body weight of Pomc-Tsc1cKO male mice was similar to controls during the rapamycin treatment period from 2–7 weeks of age in Treatment A, but the Pomc-Tsc1cKO mice gained significantly more weight following the switch to vehicle treatment after 7 weeks of age (Figure 6B upper). The adiposity (measured by fat pad weight), leptin levels, and food intake for the Pomc-Tsc1cKO mice was also similar to controls throughout the initial rapamycin treatment phase of these animals, but increased along with body weight following the switch to vehicle in Treatment A (Figure 6C–E). Thus, while rapamycin treatment from 2–7 weeks of age normalizes metabolic parameters in Pomc-Tsc1cKO mice, these improvements are lost upon discontinuation of rapamycin.

As for untreated animals (Figure 5), the body weight of Pomc-Tsc1cKO mice increased rapidly compared to controls in the first (vehicle treatment) phase of Treatment B animals (Figure 6B lower); these animals also displayed hyperphagia with increased fat pad weight and leptin levels (Figure 6C–E). The divergence in body weight between the Pomc-Tsc1cKO and control animals ceased upon the initiation of rapamycin treatment at 7 weeks of age, however, and food intake and measures of adiposity all decreased toward control levels during the second (rapamycin) phase of Treatment B (Figure 6B–E). Interestingly, hypothalamic Pomc mRNA expression did not vary by genotype, consistent with the mainly non-transcriptional function of mTOR, and suggesting that the altered Pomc mRNA expression noted in Rip-Tsc1cKO mice was likely due to indirect, rather than cell-autonomous effects of mTOR activation (Supplementary Figure 5B). Similarly, blood glucose levels did not vary among groups, nor were there differences in linear growth (as measured by snout to anus length) between genotypes under any treatment conditions, although this measure was slightly decreased by rapamycin treatment in both genotypes (Supplementary Figure 5C, 5D). Since genetic interference with central melanocortin action generally promotes increased linear growth, the normal body length of the Pomc-Tsc1cKO mice suggests that cell-autonomous mTOR activation interferes with other important aspects of Pomc neuron action.

Thus, the physiologic phenotypes of animals subjected to Treatment A and Treatment B animals suggest an important role for ongoing mTOR activation (as opposed to mTOR-mediated alterations in development) in the hyperphagia and obesity of Pomc-Tsc1cKO mice, since mTOR blockade normalizes the metabolism of adult as well as younger animals, and since these changes are reversed when mTOR inhibition ceases.

Rapamycin normalizes neuronal morphology in Pomc-Tsc1cKO mice

In order to examine the potential alteration of Pomc neurons with mTOR activation and rapamycin-mediated inhibition, we examined the size of Pomc neurons and the density of Pomc projections to the PVN by ACTH immunofluorescence in 7 and 12 week-old animals subjected to Treatment A and Treatment B. While rapamycin treatment from 2–7 weeks of Treatment A normalized the size of Pomc soma and the density of PVN projections in the Pomc-Tsc1cKO mice at 7 weeks of age (Figure 7A, 7B), the size of these neurons increased and their projection density decreased over the subsequent 5 weeks of vehicle treatment (Figure 7A, 7B). While the size of Pomc neurons in vehicle-treated Pomc-Tsc1cKO mice was increased and PVN projections were decreased relative to controls at 7 weeks of age (Figure 7A, 7B), rapamycin administration to these animals from 7–12 weeks significantly reduced their Pomc somatic size and partially improved the detection of ACTH-IR PVN projections in Treatment B (Figure 7A, 7B). Thus, the increased size of Pomc neurons and their decreased projection density to the PVN in Pomc-Tsc1cKO mice is preventable by mTOR inhibition in young animals, but reverts upon the cessation of treatment. While the alterations in Pomc somatic size remain largely reversible by mTOR inhibition in older animals, rapamycin administration to these adult animals does not completely restore ACTH-IR PVN projections, suggesting roles for mTOR activity both in the development and in the ongoing regulation of these projections. In contrast, the ongoing levels of mTOR activity, not developmental processes, largely determine the somatic size of Pomc neurons and metabolic parameters in Pomc-Tsc1cKO mice.

Figure 7
Ongoing rapamycin treatment normalized the morphological alterations of Pomc-Tsc1cKO mice

DISCUSSION

The data presented here suggest that the hyperphagic and obese phenotype of Rip-Tsc1cKO mice likely results from the ablation of Tsc1 in the hypothalamus. Indeed, the β-cell specific (without hypothalamic involvement) knockout of Tsc2 caused hyperinsulinemia but not obesity (Shigeyama et al., 2008). Furthermore, treatment of mice with low dose rapamycin (1mg/kg body weight), which blocks β-cell hypertrophy and hyperinsulinemia but does not normalize hypothalamic pS6-IR, does not prevent the hyperphagia and obesity of the Rip-Tsc1cKO mice (Mori et al., unpublished observations).

Deletion of Tsc1 specifically from Pomc-expressing cells results in hyperphagia and obesity in Pomc-Tsc1cKO mice, and this phenotype cannot be attributed to alterations in the function of pituitary corticotrophs. Thus, these data suggest that constitutive activation of mTOR in anorexigenic neurons of the hypothalamus dysregulates these neurons and results in hyperphagic obesity. These neurons respond to a variety of signals, including leptin and nutrients, to decrease feeding. A variety of data also suggest an important role for the insulin signaling cascade (including PI3-kinase) in the regulation of neuronal activity and metabolism (Belgardt et al., 2008; Choudhury et al., 2005; Kubota et al., 2004; Lin et al., 2004; Plum et al., 2006). While the activation of mTOR in peripheral tissues interferes with insulin/PI3-kinase signaling, we were not able to detect alterations in PI3-kinase signaling in the Pomc-Tsc1cKO mice (data not shown). The anorectic response to central insulin is modest, and most inactivating mutations in insulin signaling molecules (e.g., Insulin Receptor, PI3-kinase, IRS-proteins, etc.) in POMC neurons produce little change in overall energy balance (Hill et al., 2008; Choudhury et al., 2005; Belgardt et al., 2008; Konner et al., 2007). These findings, along with the dramatic obesity of the Pomc-Tsc1cKO animals, suggest that a direct effect of mTOR activation, rather than interference with PI3-kinase, underlies the obesity of Pomc-Tsc1cKO mice.

Increased mTOR activity cell-autonomously increases the somatic size of Pomc neurons, consistent with the hypertrophic effects of mTOR in other tissues (Meikle et al., 2008; Rachdi et al., 2008; Shigeyama et al., 2008). Increased cell-autonomous mTOR activity also impairs the ACTH-IR projections from the ARC Pomc neurons. mTOR activation appears to modulate the detection of ACTH-IR PVN projections by developmental as well as ongoing processes, since the normalization of projections by rapamycin treatment during development is at least partially reversed upon the cessation of treatment, and 5 weeks of treatment in older animals does not completely normalize projections in Pomc-Tsc1cKO mice. The detection of PVN projections by neuropeptide staining necessarily reflects the expression and axonal content of the neuropeptide, and our data thus reflect the combination of ACTH-IR peptide content as well as the density of axonal projections to the PVN.

Increased Pomc somatic size in Pomc-Tsc1cKO mice is not developmentally programmed, but modulated by ongoing mTOR activity, as is the impaired catabolic action of the melanocortin system in these mice. These data reveal an important role for the post-developmental activation of mTOR in Pomc neurons for metabolic regulation. Note, however, that these data cannot rule out the possibility of a more modest effect of mTOR-dependent developmental alterations in the Pomc system on energy homeostasis, but rather reveal the dominant effect of ongoing mTOR activation in this model system. It is possible that the early activation of mTOR in Pomc and other hypothalamic neurons may alter their normal developmental program, as observed in states of maternal obesity and diabetes. Indeed, ablation of Tsc1 in most neurons using Synapsin-Cre mice (Syn-Tsc1cKO) results in a variety of neuropathological abnormalities, including ectopic and hypomyelinated projections that may result from developmental abnormalities (Meikle et al., 2007).

In contrast to our present data demonstrating that long-term activation of the TSC-mTOR pathway impairs the function of anorectic Pomc neurons, Cota et al. showed that the acute pharmacologic inhibition of mTOR throughout the hypothalamus impairs appetite control- presumably at least in part via the regulation of AgRP neurons (Cota et al., 2006). A number of differences between these studies may underlie the distinct results obtained. While the Cota study essentially examined the effect of intracerebroventricular rapamycin in all hypothalamic (and presumably other) neurons near the 3rd cerebral ventricle, our study isolated the effect of mTOR activation specifically in Pomc neurons. Furthermore, as our unpublished data reveal that neuronal activity cell-autonomously regulates mTOR signaling in hypothalamic neurons (Villanueva and Myers in preparation), the function of hypothalamic mTOR in the ultimate regulation of appetite should depend upon the specific regulation and role of mTOR activity in each neural subtype. As importantly, our examination of a system with chronic activation of mTOR signaling contrasts with the acute mTOR inhibition strategy utilized by Cota, et al. Thus, in our genetic model, the chronic activation of mTOR in anorexigenic hypothalamic neurons may mimic the effects observed in the peripheral tissues of patients and animal models with obesity and type 2 diabetes.

Our study indicates that uncontrolled mTOR activity impairs the development of neuronal projections in a cell-autonomous manner. In contrast, the main effects of mTOR activation in Pomc neurons on somatic size and energy balance are reversible, and chronic ongoing activation of mTOR in the hypothalamic Pomc neurons, such as might be caused by overnutrition, promotes the development of hyperphagic obesity. Such dysregulation may promote hypothalamic dysfunction and further overeating in states of overnutrition and obesity.

MATERIALS AND METHODS

Animals and animal care

Tsc1lox/lox mice, with exons 17 and 18 of Tsc1 flanked by loxP sites by homologous recombination have been described (Meikle et al., 2005; Uhlmann et al., 2002). We generated β cell and hypothalamic specific Tsc1 knockout mice (Rip-Tsc1cKO) by breeding Tsc1lox/lox mice with mice that express the Cre recombinase gene under the control of the rat insulin 2 gene promoter (The Jackson Laboratory). Mice were maintained on the mixed genetic background (C57Bl/6 × 129Sv × BALB/c). We performed experiments using Rip-Cre/Tsc1lox/lox (Rip-Tsc1cKO) mice and littermates Tsc1lox/lox as control (wild-type). Pomc-Cre mice (FVB background) from The Jackson Laboratory were mated with Tsc1lox/lox mice to generate Pomc-Tsc1cKO same as Rip-Tsc1cKO, and were maintained on mix genetic background. Mice were housed on a 12-hour light/12-hour dark cycle in the Unit for Laboratory Animal Medicine (ULAM) at the University of Michigan, with free access to water and standard mouse chow. Animal experiments were conducted following protocols approved by the University Committee on the Use and Care of Animals (UCUCA).

In vivo physiological studies

Blood glucose levels were determined using an automated blood glucose reader (Accu-Check, Roche). Serum insulin and leptin levels were measured by ELISA (Crystal Chem, IL). For a leptin administration study, mice were individually caged. Murine leptin (PeproTech) (2.5mg/kg body weight) was injected intraperitoneally twice daily (at 09:00 and 18:00) for 2 days. Body weight and food intake were measured daily before injection, during the treatment period, and 2-day wash-out period following the injection. Serum corticosterone levels were measured by Hormone Assay & Analytical Services Core, MMPC, Vanderbilt University (NIH grant # DK59637).

Total body fat content was determined using a dual energy X-ray absorptiometry (DEXA) scanner (GE Medical Systems Lunar) according to the manufacturer’s instructions.

Rapamycin treatment studies

Rapamycin (LC laboratories) was initially dissolved in 100% ethanol, stored at −20 °C, and further diluted in an aqueous solution of 5.2% Tween 80 and 5.2% PEG 400 (final ethanol concentration, 2%) immediately before use (Wendel et al., 2004). Rip-Tsc1cKO, Pomc-Tsc1cKO and their control animals were injected rapamycin intraperitoneally (6 mg/kg body weight, every other day) for the periods shown in the each experiment. Both pancreas and brain tissues for WB, RT-PCR and IHC were harvested after 12hr of the last treatment of rapamycin.

Immunohistochemistry

The Brains were harvested from both 24-hr-fasted or ad libitum fed mice at the beginning of light cycle. After transcardiac perfusion with 4% paraformaldehyde, brains were postfixed in 4% paraformaldehyde, then transferred to 20% sucrose solution for overnight. Tissues were frozen on dry ice and cut into 25-μm coronal sections on a sliding microtome, collected in 5 series, and stored in antifreeze solution (50% PBS, 15% ethyl glycol, and 35% glycerol) at −20°C until further use. One series of brain sections was then stained with Phospho-S6 (Ser235/236(IHC preferred) or Alexa Fluor 488 Conjugated; Cell Signaling Technology), and ACTH antibody (from Dr. Parwlow, NATIONAL HORMONE and PEPTIDE PROGRAM), using free-floating method.

Quantification of ATCH-positive cell size in the ARC and projections to the PVN

For quantification of cell size in ACTH positive cells in the ARC and projections to the PVN, we used ImageJ software (NIH, Bethesda MD). All high magnification images were converted into binary files using a unique threshold value that separated positively labeled cells from background. In order to exclude false positives, the average size of cells that fell within a strict set of parameters (larger than 0.2 μm2 and between 0.50–1.00 circularity) was quantified. Average cell size of ACTH(+) cells within matched sections of the ARC were compared. For projection density measurements in the PVN, ACTH(+) stained slices were converted into binary files to separate labeled projections from background tissue. All images for intercomparison in this manner were collected using identical microscope and camera settings on tisse samples processed in parallel. Optical density measurements of projections within a 2.0×2.0 sized circular selection area placed over the left side of each PVN were compared.

Western blot analysis

Tissue extracts were immunoblotted with Tsc1, Tsc2, phospho-S6, S6 (Cell Signaling) and α-Tubulin (Sigma Aldrich).

mRNA quantification by RT-PCR analysis

Hypothalami were harvested immediately from both 24-hr-fasted and ad libitum fed mice at the beginning of light cycle, homogenized in Trizol reagent (Invitrogen). Total RNA was isolated and used for reverse transcription by SuperScript First-Strand Synthesis System (Invitrogen). Quantitative PCR was carried out with specific primers. Leptin receptor (Lepr), Npy, Pomc, Agrp, hypoxanthine phosphoribosyltransferase 1 (Hprt) mRNA levels were measured using the ABI Prism 7300 Real-Time PCR System, with SYBR green reagents (Applied Biosystems Inc.). The expression of each mRNA level was normalized to the expression of Hprt. Primers for real-time RT-PCR were: Lepr sense 5′-GGGTTTATCCATAGTCCTGTCA-3′, Lepr antisense 5′-AAATGTCTGGGCCTCTGTCTCC-3′, Hprt sense 5′-TCATTATGCCGAGGATTTGGA-3′, Hprt antisense 5′-GCACACAGAGGGCCACAAT-3′. Primers for Pomc, Npy, Agrp were described previously (Bates et al., 2003).

Statistical analyses

All data are presented as mean ± s.e.m. and were analyzed by Student’s t-test or analyses of variance (ANOVA). The differences were considered to be significant if P < 0.05.

Supplementary Material

Acknowledgments

This work was supported by grants from NIH and DOD (K.L.G), NIH DK57768 and DK56731 (MGM), DK51563 and DK62876 (OAM) and an AHA SDG to H. Mü; H. Mo was supported by the Uehara Memorial Foundation and a mentor-based postdoctoral fellowship from ADA. We thank Marta Dzaman, Wanda Snead, Chris Edwards, for technical advice and assistance, members of the Guan, MacDougald and Myers lab for helpful discussions and assistance.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • Bates SH, Stearns WH, Dundon TA, Schubert M, Tso AW, Wang Y, Banks AS, Lavery HJ, Haq AK, Maratos-Flier E, Neel BG, Schwartz MW, Myers MG., Jr STAT3 signalling is required for leptin regulation of energy balance but not reproduction. Nature. 2003;421:856–859. [PubMed]
  • Belgardt BF, Husch A, Rother E, Ernst MB, Wunderlich FT, Hampel B, Klockener T, Alessi D, Kloppenburg P, Bruning JC. PDK1 deficiency in POMC-expressing cells reveals FOXO1-dependent and -independent pathways in control of energy homeostasis and stress response. Cell Metab. 2008;7:291–301. [PubMed]
  • Biondi CA, Gartside MG, Waring P, Loffler KA, Stark MS, Magnuson MA, Kay GF, Hayward NK. Conditional inactivation of the MEN1 gene leads to pancreatic and pituitary tumorigenesis but does not affect normal development of these tissues. Molecular and cellular biology. 2004;24:3125–3131. [PMC free article] [PubMed]
  • Brunet A, Bonni A, Zigmond MJ, Lin MZ, Juo P, Hu LS, Anderson MJ, Arden KC, Blenis J, Greenberg ME. Akt promotes cell survival by phosphorylating and inhibiting a Forkhead transcription factor. Cell. 1999;96:857–868. [PubMed]
  • Chong-Kopera H, Inoki K, Li Y, Zhu T, Garcia-Gonzalo FR, Rosa JL, Guan KL. TSC1 stabilizes TSC2 by inhibiting the interaction between TSC2 and the HERC1 ubiquitin ligase. The Journal of biological chemistry. 2006;281:8313–8316. [PubMed]
  • Choudhury AI, Heffron H, Smith MA, Al-Qassab H, Xu AW, Selman C, Simmgen M, Clements M, Claret M, Maccoll G, Bedford DC, Hisadome K, Diakonov I, Moosajee V, Bell JD, Speakman JR, Batterham RL, Barsh GS, Ashford ML, Withers DJ. The role of insulin receptor substrate 2 in hypothalamic and beta cell function. The Journal of clinical investigation. 2005;115:940–950. [PubMed]
  • Coll AP, Farooqi IS, O’Rahilly S. The hormonal control of food intake. Cell. 2007;129:251–262. [PMC free article] [PubMed]
  • Cota D, Proulx K, Smith KA, Kozma SC, Thomas G, Woods SC, Seeley RJ. Hypothalamic mTOR signaling regulates food intake. Science. 2006;312:927–930. [PubMed]
  • Flier JS. Obesity wars: molecular progress confronts an expanding epidemic. Cell. 2004;116:337–350. [PubMed]
  • Flotow H, Thomas G. Substrate recognition determinants of the mitogen-activated 70K S6 kinase from rat liver. The Journal of biological chemistry. 1992;267:3074–3078. [PubMed]
  • Gao X, Zhang Y, Arrazola P, Hino O, Kobayashi T, Yeung RS, Ru B, Pan D. Tsc tumour suppressor proteins antagonize amino-acid-TOR signalling. Nat Cell Biol. 2002;4:699–704. [PubMed]
  • Hill JW, Williams KW, Ye C, Luo J, Balthasar N, Coppari R, Cowley MA, Cantley LC, Lowell BB, Elmquist JK. Acute effects of leptin require PI3K signaling in hypothalamic proopiomelanocortin neurons in mice. The Journal of clinical investigation. 2008;118:1796–1805. [PubMed]
  • Inoki K, Corradetti MN, Guan KL. Dysregulation of the TSC-mTOR pathway in human disease. Nat Genet. 2005;37:19–24. [PubMed]
  • Inoki K, Li Y, Xu T, Guan KL. Rheb GTPase is a direct target of TSC2 GAP activity and regulates mTOR signaling. Genes Dev. 2003;17:1829–1834. [PubMed]
  • Inoki K, Li Y, Zhu T, Wu J, Guan KL. TSC2 is phosphorylated and inhibited by Akt and suppresses mTOR signalling. Nat Cell Biol. 2002;4:648–657. [PubMed]
  • Kitamura T, Feng Y, Kitamura YI, Chua SC, Jr, Xu AW, Barsh GS, Rossetti L, Accili D. Forkhead protein FoxO1 mediates Agrp-dependent effects of leptin on food intake. Nat Med. 2006;12:534–540. [PubMed]
  • Kobayashi T, Minowa O, Sugitani Y, Takai S, Mitani H, Kobayashi E, Noda T, Hino O. A germ-line Tsc1 mutation causes tumor development and embryonic lethality that are similar, but not identical to, those caused by Tsc2 mutation in mice. Proceedings of the National Academy of Sciences of the United States of America. 2001;98:8762–8767. [PubMed]
  • Konner AC, Janoschek R, Plum L, Jordan SD, Rother E, Ma X, Xu C, Enriori P, Hampel B, Barsh GS, Kahn CR, Cowley MA, Ashcroft FM, Bruning JC. Insulin Action in AgRP-Expressing Neurons Is Required for Suppression of Hepatic Glucose Production. Cell Metab. 2007;5:438–449. [PubMed]
  • Kops GJ, de Ruiter ND, De Vries-Smits AM, Powell DR, Bos JL, Burgering BM. Direct control of the Forkhead transcription factor AFX by protein kinase B. Nature. 1999;398:630–634. [PubMed]
  • Korsheninnikova E, van der Zon GC, Voshol PJ, Janssen GM, Havekes LM, Grefhorst A, Kuipers F, Reijngoud DJ, Romijn JA, Ouwens DM, Maassen JA. Sustained activation of the mammalian target of rapamycin nutrient sensing pathway is associated with hepatic insulin resistance, but not with steatosis, in mice. Diabetologia. 2006;49:3049–3057. [PubMed]
  • Kubota N, Terauchi Y, Tobe K, Yano W, Suzuki R, Ueki K, Takamoto I, Satoh H, Maki T, Kubota T, Moroi M, Okada-Iwabu M, Ezaki O, Nagai R, Ueta Y, Kadowaki T, Noda T. Insulin receptor substrate 2 plays a crucial role in beta cells and the hypothalamus. The Journal of clinical investigation. 2004;114:917–927. [PMC free article] [PubMed]
  • Kwiatkowski DJ, Zhang H, Bandura JL, Heiberger KM, Glogauer M, el-Hashemite N, Onda H. A mouse model of TSC1 reveals sex-dependent lethality from liver hemangiomas, and up-regulation of p70S6 kinase activity in Tsc1 null cells. Hum Mol Genet. 2002;11:525–534. [PubMed]
  • Lin X, Taguchi A, Park S, Kushner JA, Li F, Li Y, White MF. Dysregulation of insulin receptor substrate 2 in beta cells and brain causes obesity and diabetes. The Journal of clinical investigation. 2004;114:908–916. [PMC free article] [PubMed]
  • Long X, Lin Y, Ortiz-Vega S, Yonezawa K, Avruch J. Rheb binds and regulates the mTOR kinase. Curr Biol. 2005;15:702–713. [PubMed]
  • Manning BD, Tee AR, Logsdon MN, Blenis J, Cantley LC. Identification of the tuberous sclerosis complex-2 tumor suppressor gene product tuberin as a target of the phosphoinositide 3-kinase/akt pathway. Molecular cell. 2002;10:151–162. [PubMed]
  • Meikle L, McMullen JR, Sherwood MC, Lader AS, Walker V, Chan JA, Kwiatkowski DJ. A mouse model of cardiac rhabdomyoma generated by loss of Tsc1 in ventricular myocytes. Hum Mol Genet. 2005;14:429–435. [PubMed]
  • Meikle L, Pollizzi K, Egnor A, Kramvis I, Lane H, Sahin M, Kwiatkowski DJ. Response of a neuronal model of tuberous sclerosis to mammalian target of rapamycin (mTOR) inhibitors: effects on mTORC1 and Akt signaling lead to improved survival and function. J Neurosci. 2008;28:5422–5432. [PMC free article] [PubMed]
  • Meikle L, Talos DM, Onda H, Pollizzi K, Rotenberg A, Sahin M, Jensen FE, Kwiatkowski DJ. A mouse model of tuberous sclerosis: neuronal loss of Tsc1 causes dysplastic and ectopic neurons, reduced myelination, seizure activity, and limited survival. J Neurosci. 2007;27:5546–5558. [PubMed]
  • Miller AM, Brestoff JR, Phelps CB, Berk EZ, Reynolds THt. Rapamycin does not improve insulin sensitivity despite elevated mammalian target of rapamycin complex 1 activity in muscles of ob/ob mice. American journal of physiology. 2008;295:R1431–1438. [PubMed]
  • Murakami M, Ichisaka T, Maeda M, Oshiro N, Hara K, Edenhofer F, Kiyama H, Yonezawa K, Yamanaka S. mTOR is essential for growth and proliferation in early mouse embryos and embryonic stem cells. Molecular and cellular biology. 2004;24:6710–6718. [PMC free article] [PubMed]
  • Nguyen KT, Tajmir P, Lin CH, Liadis N, Zhu XD, Eweida M, Tolasa-Karaman G, Cai F, Wang R, Kitamura T, Belsham DD, Wheeler MB, Suzuki A, Mak TW, Woo M. Essential role of Pten in body size determination and pancreatic beta-cell homeostasis in vivo. Molecular and cellular biology. 2006;26:4511–4518. [PMC free article] [PubMed]
  • Oshiro N, Takahashi R, Yoshino K, Tanimura K, Nakashima A, Eguchi S, Miyamoto T, Hara K, Takehana K, Avruch J, Kikkawa U, Yonezawa K. The proline-rich Akt substrate of 40 kDa (PRAS40) is a physiological substrate of mammalian target of rapamycin complex 1. The Journal of biological chemistry. 2007;282:20329–20339. [PMC free article] [PubMed]
  • Pende M, Um SH, Mieulet V, Sticker M, Goss VL, Mestan J, Mueller M, Fumagalli S, Kozma SC, Thomas G. S6K1(−/−)/S6K2(−/−) mice exhibit perinatal lethality and rapamycin-sensitive 5′-terminal oligopyrimidine mRNA translation and reveal a mitogen-activated protein kinase-dependent S6 kinase pathway. Molecular and cellular biology. 2004;24:3112–3124. [PMC free article] [PubMed]
  • Plum L, Ma X, Hampel B, Balthasar N, Coppari R, Munzberg H, Shanabrough M, Burdakov D, Rother E, Janoschek R, Alber J, Belgardt BF, Koch L, Seibler J, Schwenk F, Fekete C, Suzuki A, Mak TW, Krone W, Horvath TL, Ashcroft FM, Bruning JC. Enhanced PIP3 signaling in POMC neurons causes KATP channel activation and leads to diet-sensitive obesity. The Journal of clinical investigation. 2006;116:1886–1901. [PubMed]
  • Rachdi L, Balcazar N, Osorio-Duque F, Elghazi L, Weiss A, Gould A, Chang-Chen KJ, Gambello MJ, Bernal-Mizrachi E. Disruption of Tsc2 in pancreatic beta cells induces beta cell mass expansion and improved glucose tolerance in a TORC1-dependent manner. Proceedings of the National Academy of Sciences of the United States of America. 2008;105:9250–9255. [PubMed]
  • Sancak Y, Thoreen CC, Peterson TR, Lindquist RA, Kang SA, Spooner E, Carr SA, Sabatini DM. PRAS40 is an insulin-regulated inhibitor of the mTORC1 protein kinase. Molecular cell. 2007;25:903–915. [PubMed]
  • Sano H, Kane S, Sano E, Miinea CP, Asara JM, Lane WS, Garner CW, Lienhard GE. Insulin-stimulated phosphorylation of a Rab GTPase-activating protein regulates GLUT4 translocation. The Journal of biological chemistry. 2003;278:14599–14602. [PubMed]
  • Shigeyama Y, Kobayashi T, Kido Y, Hashimoto N, Asahara S, Matsuda T, Takeda A, Inoue T, Shibutani Y, Koyanagi M, Uchida T, Inoue M, Hino O, Kasuga M, Noda T. Biphasic response of pancreatic beta-cell mass to ablation of tuberous sclerosis complex 2 in mice. Molecular and cellular biology. 2008;28:2971–2979. [PMC free article] [PubMed]
  • Stiles BL, Kuralwalla-Martinez C, Guo W, Gregorian C, Wang Y, Tian J, Magnuson MA, Wu H. Selective deletion of Pten in pancreatic beta cells leads to increased islet mass and resistance to STZ-induced diabetes. Molecular and cellular biology. 2006;26:2772–2781. [PMC free article] [PubMed]
  • Uhlmann EJ, Wong M, Baldwin RL, Bajenaru ML, Onda H, Kwiatkowski DJ, Yamada K, Gutmann DH. Astrocyte-specific TSC1 conditional knockout mice exhibit abnormal neuronal organization and seizures. Ann Neurol. 2002;52:285–296. [PubMed]
  • Vander Haar E, Lee SI, Bandhakavi S, Griffin TJ, Kim DH. Insulin signalling to mTOR mediated by the Akt/PKB substrate PRAS40. Nat Cell Biol. 2007;9:316–323. [PubMed]
  • Wang L, Harris TE, Roth RA, Lawrence JC., Jr PRAS40 regulates mTORC1 kinase activity by functioning as a direct inhibitor of substrate binding. The Journal of biological chemistry. 2007;282:20036–20044. [PubMed]
  • Wendel HG, De Stanchina E, Fridman JS, Malina A, Ray S, Kogan S, Cordon-Cardo C, Pelletier J, Lowe SW. Survival signalling by Akt and eIF4E in oncogenesis and cancer therapy. Nature. 2004;428:332–337. [PubMed]
  • Wullschleger S, Loewith R, Hall MN. TOR signaling in growth and metabolism. Cell. 2006;124:471–484. [PubMed]
  • Yang Q, Inoki K, Kim E, Guan KL. TSC1/TSC2 and Rheb have different effects on TORC1 and TORC2 activity. Proceedings of the National Academy of Sciences of the United States of America. 2006;103:6811–6816. [PubMed]