|Home | About | Journals | Submit | Contact Us | Français|
Rett Syndrome (RTT) is an autism spectrum disorder caused by mutations in the X-linked gene encoding methyl-CpG binding protein 2 (MeCP2). In order to map the neuroanatomic origins of the complex neuropsychiatric behaviors observed in patients with RTT, and to uncover novel endogenous functions of MeCP2, we removed Mecp2 from Sim1-expressing neurons in the hypothalamus using Cre-loxP technology. Loss of MeCP2 in Sim1-expressing neurons resulted in mice that recapitulated the abnormal physiological stress response that is seen upon MeCP2 dysfunction in the entire brain. Surprisingly, we also uncovered a novel role for MeCP2 in the regulation of social and feeding behaviors since the Mecp2 conditional knockout (CKO) mice were aggressive, hyperphagic and obese. This study demonstrates that deleting Mecp2 in a defined brain region is an excellent approach to map the neuronal origins of complex behaviors and provides new insight about the function of MeCP2 in specific neurons.
RTT is a devastating X-linked neurodevelopmental disorder that affects approximately 1 in 10,000 females (Moretti and Zoghbi, 2006). Patients with classic RTT suffer from a broad array of phenotypes that affect almost every part of the central and autonomic nervous systems (Glaze, 2005; Weaving et al., 2005). These phenotypes include impaired social behavior and communication skills, motor abnormalities, and the development of stereotyped movements (Hagberg, 2002).
The majority (>95%) of RTT cases are caused by mutations in the gene encoding methyl-CpG-binding protein 2 (MeCP2), a protein widely expressed throughout the brain (Williamson and Christodoulou, 2006; Zoghbi, 2005). In addition to classic RTT, patients with MECP2 mutations manifest a variety of neuropsychiatric conditions including autism, juvenile onset schizophrenia, and bipolar disease with mental retardation depending on the type of mutation or the pattern of X-chromosome inactivation (XCI) (Cohen et al., 2002; Weaving et al., 2003). Girls typically present with classic RTT when they have random XCI, whereby half of their cells express a mutant MECP2 allele (Shahbazian et al., 2002b; Takagi, 2001). In contrast, patients with skewed patterns of XCI favoring the wild-type (WT) allele might display only a few features such as autism, tremor, or other neurobehavioral deficits. Because the majority of cells in girls with favorable XCI patterns express the WT allele, the isolated phenotypes might result from loss of MeCP2 function in particular brain regions.
The creation of several mouse models carrying different Mecp2 mutations enabled the recapitulation of many of the phenotypes seen in patients, and has enhanced our appreciation of the breadth of clinical phenotypes that are associated with RTT. Mecp2 null mice display tremor, breathing dysfunction, hind-limb clasping, they are overweight (on a mixed 129, C57BL/6, BALB/c genetic background), and they die by 10 weeks (Chen et al., 2001; Guy et al., 2001). Mice bearing a truncated Mecp2 allele (Mecp2308/Y) are overweight (on a 129 Sv/Ev background, unpublished observation), display motor dysfunction, seizures, stereotypies, altered social behavior, abnormal stress responses, anxiety-like behavior, and die by 15 months of age (McGill et al., 2006; Moretti et al., 2005; Shahbazian et al., 2002a). Conditional deletion of Mecp2 from neurons that express CaMKII-Cre93 throughout the forebrain and midbrain reproduces a large subset of RTT features including abnormal motor coordination, anxiety-like behavior, impaired learning and memory, and weight gain (Gemelli et al., 2006). The overweight phenotype seen in each of the mouse models was initially perplexing given that girls with classic RTT are typically thin, and often suffer from growth failure predominantly because of feeding difficulties due to oropharyngeal dysmotility (difficulty chewing and swallowing) (Motil et al., 1999). However, upon reflection on the clinical course of many patients with MECP2 mutations who do not display classic RTT, we noticed that males with hypomorphic MECP2 mutations and females with milder variants of RTT often become obese (Couvert et al., 2001; Kleefstra et al., 2002; Zappella et al., 2001).
Studying patients with RTT as well as mouse models of RTT is a daunting task given the incredible range of phenotypes that stem from MeCP2 dysfunction in all neurons. Moreover, it is very difficult to dissociate phenotypes attributable to dysfunction of MeCP2 from secondary compensatory changes. Given the clinical observation that different patterns of XCI likely result in the manifestation of distinct subsets of RTT phenotypes, we proposed that particular RTT phenotypes result from loss of function of MeCP2 in specific neurons and that deleting Mecp2 from small groups of neurons in the mouse will enable us to map specific neurobehavioral phenotypes to discrete brain regions or cell types. Furthermore, we reasoned that this strategy should allow us to uncover novel endogenous functions of MeCP2 that might otherwise be masked in the constitutive Mecp2 null mice that display many phenotypes.
In this work we focus on the hypothalamus because patients with MECP2 mutations display many phenotypes suggestive of hypothalamic dysfunction such as sleep abnormalities, episodes of heightened anxiety, an abnormal physiological response to stress as measured by increased levels of cortisol in the urine, gastrointestinal dysfunction, as well as cardiac and breathing abnormalities (Axelrod et al., 2006; Motil et al., 2006; Motil et al., 1999; Mount et al., 2002; Young et al., 2007). In this study we utilized mice that express Cre recombinase under the control of Sim1 regulatory elements to delete Mecp2 in several regions of the hypothalamus using Cre-loxP technology. We performed a comprehensive battery of behavioral and physiological tests on Mecp2 conditional knockout (CKO) mice and all littermate controls to screen for the presence of both RTT-like and potentially novel phenotypes and we discovered that MeCP2 plays a critical role in pathways important for mediating proper physiological responses to stress as well as both social and feeding behaviors.
We removed Mecp2 from Sim1-expressing neurons by breeding mice carrying a Mecp2 allele flanked by loxP sites (Mecp2flox/+) (Guy et al., 2001), to Sim1-Cre BAC transgenic mice (Balthasar et al., 2005). Sim1 regulatory elements drive Cre expression in transgenic mice during embryonic development and after birth in the paraventricular (PVN), supraoptic (SON) and posterior (PH) hypothalamic nuclei, as well as in the nucleus of the lateral olfactory tract (NLOT) of the amygdala (Balthasar et al., 2005). A small amount of scattered Cre expression also occurs in a few other areas surrounding the hypothalamus and amygdala (Balthasar et al., 2005), however, this is very minimal compared to the robust Cre expression found in the nuclei mentioned above. Fluorescent immunohistochemistry demonstrated a clear reduction of MeCP2 in all areas where Cre is expressed including the PVN (Fig 1. A–B) and SON (Fig 1. C–D) of Mecp2 conditional knock out (Mecp2 CKO) mice. MeCP2 levels were not altered in the suprachiasmatic nucleus (SCN), an area of the hypothalamus where the Cre is not expressed (data not shown).
To screen for the presence of RTT-like phenotypes, we performed a comprehensive battery of behavioral tests on Mecp2 CKO mice and all littermate controls (Table 1). Although Mecp2 CKO mice behaved similarly to their littermates in a variety of paradigms (Fig. S1), they also reproduced a subset of phenotypes typically detected with MeCP2 dysfunction throughout the brain (Table 1). In light of the increased urinary cortisol excretion in patients with RTT (Motil et al., 2006), and the higher serum corticosterone in Mecp2308/Y mice after stress (McGill et al., 2006), we examined endocrine responses to stress in Mecp2 CKO mice. We measured serum levels of corticosterone, in Mecp2 CKO and control mice under basal conditions and after 30 minutes of restraint stress. We found a significant effect of genotype under basal conditions, F(3,28)=5.28, p=0.005 and after stress, F(3,27)=8.70, p<0.001. Tukey’s post hoc comparisons revealed that under basal conditions, Mecp2flox and Mecp2 CKO mice had similar corticosterone levels compared to WT mice, although their levels were higher than mice that carry only the Sim1-cre transgene (p<0.05 for both, Fig. 2A). Strikingly, after 30 minutes of restraint, Mecp2 CKO mice had significantly higher corticosterone levels than all littermate controls and they exhibited a 134% increase in serum corticosterone over WT littermates (p<0.01, Fig. 2A). These results indicate that similar to patients with RTT and Mecp2308/Y mice, Mecp2 CKO mice have an enhanced physiologic response to stress.
We next attempted to determine whether Mecp2 CKO mice also exhibit increased anxiety-like behavior. In the open field, there was a significant effect of genotype for the center to total distance ratio, F(3,60)=7.58, p<0.001 (Fig. 2B). Tukey’s post hoc comparison revealed that Mecp2 CKO mice explored the center of an open field significantly less than all controls, a finding consistent with enhanced anxiety-like behavior. The total distance traveled in the open field was similar among Mecp2 CKO and WT mice, indicating that the decreased exploration of the center of the arena by Mecp2 CKO mice was not likely the result of motor impairment (Fig. 2C). It should be noted, however, that we did observe a small but significant decrease in the total distance that the Mecp2 CKO mice traveled throughout the test when compared to mice that carry only the Sim1-cre allele (p<0.01). To confirm that the Mecp2 CKO mice display increased anxiety-like behavior we performed a light-dark box assay, one of the most widely used and well-documented tests for studying anxiety-like behavior (Bourin and Hascoet, 2003). Analysis of the data by two-way analysis of variance (ANOVA) demonstrated a significant effect of genotype (Mecp2flox × Sim1-cre interaction) on both the percentage of time spent in the lit side, F(1,60)=4.08; p=0.048 and the number of transitions between compartments, F(1,60)=5.85; p=0.019. Interestingly, however, one-way ANOVA with Tukey’s posthoc analysis revealed that Mecp2 CKO mice spent a similar amount of time in the lit side of the light dark box compared to all littermate controls (Fig. 2D). Furthermore, the number of transitions between compartments and the latency to enter the dark were not different between Mecp2 CKO mice and WT controls (data not shown). Thus although Mecp2 CKO mice clearly display an increased physiological response to stress and a decreased center to total distance ratio in the open field assay, the light-dark data do not reveal increased anxiety-like behavior in the Mecp2 CKO mice.
Mecp2 CKO mice also behaved abnormally in the partition test, which is designed to evaluate social behavior. In this test, two mice are placed in a cage divided by a clear and perforated plexiglass partition and the amount of time that the test mouse spends near the partition investigating either a familiar or an unfamiliar partner is recorded. Repeated measures ANOVA (genotype × time) revealed a significant effect, F(3,59)=10.65, p<0.001 and Tukey’s post hoc analysis demonstrated that Mecp2 CKO mice (p<0.001) and mice that carry only the Mecp2flox allele (p<0.05) spent more time at the partition compared to control littermates suggesting increased social interaction (Fig. 2E). This Mecp2flox effect is consistent with data reported in a recent study (Samaco et al., 2008). This study demonstrated that mice carrying the Mecp2flox allele express 50% less MeCP2 compared to WT littermates highlighting the importance of including Mecp2flox mice as controls when interpreting data from Mecp2 CKO experiments. Interestingly, we noticed that Mecp2 CKO and Mecp2flox mice engaged in very different behaviors at the partition. Mice that carry only the Mecp2flox allele demonstrated typical social interactions such as sniffing and exploration. In contrast, Mecp2 CKO mice were preoccupied with aggressive behaviors such as tail rattling directed toward the unfamiliar partner.
To further characterize these aggressive behaviors, we used the resident intruder assay. Prior to testing the response of a resident mouse to an intruder, the test mice were singly housed for two weeks, a procedure believed to enhance territoriality. On the test day, we recorded aggressive behaviors after placing a smaller, unfamiliar, group-housed WT mouse into the home cage of the test mouse. Scores used to denote aggressive behaviors included the number of attacks, mounting, biting, and tail rattling events. Mecp2 CKO mice engaged in significantly more tail rattles (p<0.01) and aggressive attacks (p<0.01) than their control littermates (Fig. 2F, supplemental movie S1, S2). In fact, out of the 24 control littermates tested, not a single tail rattle or attack was recorded during the test period. Notably, the aggression observed in Mecp2 CKO mice appears to be limited to situations when they are stressed, either as a result of being singly housed or as a specific reaction to novel intruders of their territory, since they do not attack their familiar cage mates. Furthermore, they are not aggressive toward the examiner when they are handled.
Mecp2 CKO mice fed regular mouse chow ad libitum exhibited significantly increased body weight after 7 weeks of age (repeated measures two-way ANOVA (Mecp2flox × Sim1-cre), lower-bound F(1,60)=19.59, p<0.001, Tukey’s post hoc p<0.01, Fig. 3 A–B). Mecp2 CKO mice weigh 27% more than their control littermates by 18 weeks and 48% more by 42 weeks (Fig. 3B, S2). We examined linear growth and found that Mecp2 CKO mice also exhibit a small (3%) but significant (p<0.05) increase in body length compared to littermate controls at 42 weeks of age (data not shown).
To determine whether the observed weight gain was due to increased adiposity, we dissected and weighed intra-abdominal gonadal fat pads. We found a significant effect of genotype, F(3,28)=8.91, p<0.001, and Tukey’s post hoc comparison revealed that gonadal fat pad masses were greatly increased in the Mecp2 CKO mice compared to all controls (p<0.01, Fig. 3C) suggesting that their increased body weight is primarily due to an increase in body fat deposition. In order to confirm these findings, we performed dual energy X-ray absorptiometry (DEXA) scans on Mecp2 CKO mice to determine body composition. DEXA analysis indicated that Mecp2 CKO mice had increased total body fat percentages (36% +/− 3%) compared to control littermates (20% +/−1%) at 20 weeks of age (p<0.05, Fig. 3D). These data demonstrate that the increased body weight in Mecp2 CKO mice is primarily due to increased body fat.
To probe the cause of obesity in the Mecp2 CKO mice, we used an automated home cage monitoring system to evaluate daily activity and food intake. Average daily total movement values were not significantly different between Mecp2 CKO mice and control littermates although a trend toward hypoactivity (p=0.066) was noted in the Mecp2 CKO mice (Fig. 4A). Interestingly, on experimental day 1, Mecp2 CKO mice completely lacked the novelty-induced hyperactivity response that is normally seen in WT mice (Fig. 4A) likely due to heightened anxiety. We evaluated oxygen consumption using indirect calorimetry and found that Mecp2 CKO mice have normal basal metabolic rates (Fig. 4B). Together these data suggest that the marked obesity observed in the Mecp2 CKO mice is not secondary to a decrease in energy expenditure. We next evaluated energy consumption by monitoring daily food intake. After a 4-day acclimation period, we analyzed food consumption over a subsequent 4-day period. We found a significant effect of genotype, F(3,26)=13.17, p<0.001, and Tukey’s post hoc comparison revealed that Mecp2 CKO mice are hyperphagic, consuming on average 42% more chow than WT controls (p<0.001, Fig. 4C). Interestingly, Mecp2 CKO mice did not consume extra chow compared to WT mice throughout the 24-hour period (experimental day 1) immediately after they were singly housed in novel cages, suggesting that they experienced stress-induced suppression of food intake during that period (Fig. 4C).
Feeding behavior is regulated by several signaling molecules and hormones (for a recent review see (Oswal and Yeo, 2007)). One major signal the body uses to sense energy balance is leptin; a soluble hormone secreted by adipocytes to communicate body fat deposition information to the brain. To investigate possible mechanisms responsible for mediating the hyperphagia in Mecp2 CKO mice, we measured serum leptin levels. We did not detect any significant difference in serum leptin at 6 weeks of age, but by 42 weeks, we found a significant effect of genotype F(2,17)=31.48, p<0.001, and Tukey’s post hoc comparison revealed that Mecp2 CKO mice had a 4-fold increase (p<0.001) in leptin compared to WT controls (Fig. 4D). The observation that Mecp2 CKO mice are obese in the context of elevated leptin suggests that they are able to produce and secrete leptin but are insensitive to its effects. Furthermore, because leptin levels are normal at 6 weeks, prior to the onset of obesity, we conclude that the increased leptin levels likely result from increased fat deposition in the older mice and are not due to direct leptin regulation by MeCP2.
Since the melanocortin pathway is an important mediator of leptin signaling in the hypothalamus (Oswal and Yeo, 2007), we next evaluated the expression of several members of this pathway including the melanocortin 4 receptor (Mc4r), pro-opiomelanocortin (Pomc1), agouti-related protein (Agrp), and neuropeptide Y (Npy) by in situ hybridization in Mecp2 CKO mice and control littermates. POMC and AgRP/NPY neurons in the arcuate nucleus (ARC) are activated and repressed by leptin, respectively, and they project to the PVN where POMC neurons stimulate, and AgRP/NPY neurons block the activity of MC4Rs (Haskell-Luevano and Monck, 2001; Schwartz et al., 1997). Proper MC4R signaling in the PVN is critical for the regulation of food intake (Balthasar et al., 2005). RNA in situ hybridization revealed that expression of Mc4r in the PVN (Fig. 5 A–B), and Pomc1 (Fig. 5 C–D) and Agrp in the ARC (Fig. 5 E–F), was not altered in Mecp2 CKO mice. Quantitative reverse transcriptase PCR (QRT-PCR) using cDNA from hypothalamus confirmed that Mc4r, Pomc1 and Agrp mRNA levels were not significantly different between Mecp2 CKO mice and control littermates (data not shown). In contrast, Npy expression was reduced in the ARC but greatly induced in the DMH of Mecp2 CKO mice (Fig. 5 G–H). Increased expression of Npy in the DMH occurs in several hyperphagic models of reduced MC4R signaling including the agouti yellow mutant (Ay), Mc4r −/− mice, and WT rats during lactation, suggesting that Npy in the DMH may be involved in feeding regulation mediated by MC4R signaling (Chen et al., 2004; Kesterson et al., 1997). The robust expression of Npy in the DMH of Mecp2 CKO mice suggests that although Mc4r, Pomc1 and Agrp RNA levels are all normal, MC4R signaling may be disrupted in Mecp2 CKO mice.
We next sought to determine whether any downstream components of the MC4R signaling pathway were altered in Mecp2 CKO mice. RNA in situ hybridization revealed that Bdnf, a neurotrophic factor important for weight regulation (Rios et al., 2001) and proposed to function as a downstream mediator of MC4R signaling (Tsao et al., 2007; Xu et al., 2003), was significantly decreased in the PVN of Mecp2 CKO mice compared to WT littermates (Fig. 6 C–D). In contrast, Bdnf expression in Mecp2 CKO mice was comparable to littermate controls in several nuclei of the amygdala (Fig. 6 A–B) and in the ventromedial nucleus of the hypothalamus (VMH) (Fig. 6 E–F). To quantify Bdnf expression, the in situ images from 4–7 mice per genotype were pseudo colored using the Celldetekt protocol to determine the relative expression level on a per cell basis (Carson et al., 2005), and subsequently the total number of Bdnf-expressing cells was counted throughout the PVN (Fig. 6G), and the VMH (Fig. 6H). We found a 50% decrease (p<0.05) in the number of cells that express Bdnf in the PVN of Mecp2 CKO mice (Fig. 6G). The fact that Bdnf expression was altered only in areas of the brain where there was strong Cre expression and Mecp2 was deleted (PVN), suggests that Bdnf levels are directly affected by the loss of MeCP2.
In addition to BDNF, CRH has also been suggested as a downstream mediator of MC4R signaling (Lu et al., 2003). RNA in situ hybridization revealed a significant decrease in Crh in the PVN of Mecp2 CKO mice (Fig. 7 A–B). Finally, overexpression of SIM1 completely rescues the hyperphagia that occurs in Ay mice where MC4R signaling has been disrupted, suggesting that Sim1 also acts downstream of the Mc4r in PVN neurons to regulate food intake (Kublaoui et al., 2006). QRT-PCR demonstrated that, in contrast to both Bdnf and Crh, the expression of Sim1 is not altered in the Mecp2 CKO hypothalamus (Fig. S3 A). To confirm that the expression of other genes that are highly expressed in the PVN was similarly unaltered in Mecp2 CKO mice, we examined the expression of arginine vasopressin (Avp) and oxytocin (Oxt) by QRT-PCR (Fig. S3 B–C). We found that both Avp and Oxt are expressed at WT levels in the Mecp2 CKO hypothalamus suggesting that the cells in the PVN retain at least some of their normal function.
In this study, we demonstrate that deletion of Mecp2 in Sim1-expressing neurons results in a subset of the phenotypes observed in RTT and related MECP2 disorders. We found that Mecp2 CKO mice recapitulated the increased body weight and abnormal stress response that are observed in mice where Mecp2 is mutated in all neurons. This result is interesting in light of the fact that patients with RTT syndrome have also been found to exhibit an abnormal physiological response to stress (Motil et al., 2006). Importantly, we did not recapitulate all of the phenotypes typically seen upon deletion of Mecp2 from the entire brain such as motor coordination abnormalities and learning and memory deficits, demonstrating that the function of MeCP2 in Sim1-expressiong neurons is likely not important for those behaviors. By removing Mecp2 from Sim1-expressiong neurons we also uncovered a novel role for MeCP2 in the MC4R signaling pathway that regulates food intake and in pathways important for aggression. Although aggression and hyperphagia are not typically seen in patients with classic RTT, patients with atypical RTT owing to favorable XCI or to hypomorphic MECP2 alleles do indeed manifest aggressive behavior (personal communication, P. Huppke, Georg August University; H. Zoghbi, unpublished data) and some are overweight (Couvert et al., 2001; Kleefstra et al., 2002; Zappella et al., 2001). It is possible that these phenotypes are not observed upon MeCP2 dysfunction in all neurons due to secondary changes that may mask certain primary effects of MeCP2 dysfunction. It is noteworthy that deletion of Mecp2 from other selected brain regions such as the Purkinje cells of the cerebellum or tyrosine hydroxylase neurons does not result in similar phenotypes (Neul and Zoghbi, unpublished data), highlighting the specificity of the phenotypes observed in this study and the important role of Mecp2 in the hypothalamus.
Mecp2 CKO mice displayed increased aggressive behavior that was apparent strictly upon changes in social situation such as exposure to either an unfamiliar or an intruder mouse. The fact that aggression is not detected in the home cage but is precipitated by the stress of unfamiliarity is interesting given that patients with autism spectrum disorders typically manifest aggression when stressed or frustrated. When interpreting the resident intruder data it is of note that the Mecp2 CKO mice were heavier than their control littermates. The intruder mice were slightly smaller than all of the mice that were tested although this difference in body weight was more pronounced in the case of the Mecp2 CKO mice because the latter are much larger than their control littermates. While it is possible that the increased body weight of the Mecp2 CKO mice may have contributed to the development of their increased aggression, the fact that the aggressive behavior was detected only in response to a smaller novel mouse and not in response to their smaller regular cage mates suggests that the increased body weight alone is not likely responsible for the aggressive phenotype.
Stress undoubtedly represents an altered physiological state; thus the finding that Mecp2 CKO mice cannot easily adjust to a new state adds to the growing evidence suggesting that MeCP2 is important for the modulation of gene expression in response to alterations in neuronal activity owing to new physiologic states (Chahrour and Zoghbi, 2007; Chen et al., 2003). For example, MeCP2 is important for the neuronal activity-dependent regulation of Bdnf (Zhou et al., 2006). Furthermore, genes that are normally induced in response to stress such as serum glucocorticoid-inducible kinase 1 (Sgk1) and FK506-binding protein 5 (Fkbp5) are misregulated in the Mecp2 null brain (Nuber et al., 2005). Although we do not yet understand the molecular basis for the aggression phenotype seen in the Mecp2 CKO mice, we have demonstrated that MeCP2’s function, specifically in Sim1-expressing neurons, is critical to ensure proper social behavior in response to novel and stressful social situations.
We also found that Mecp2 CKO mice are obese based on their significantly increased body weight, close to a 100% increase in body fat content by 20 weeks of age, and an apparent resistance to leptin. Mecp2 CKO mice display normal levels of activity and basal metabolic rates but they consume more chow than control littermates. Once again, deletion of Mecp2 in Sim1-expressing neurons impairs the ability of the mouse to adapt to changing physiology; in this case Mecp2 CKO mice do not stop eating once they have ingested a sufficient number of calories demonstrating that they are unable to properly respond to satiety signals.
Hyperphagia and obesity are characteristic of mice that have disrupted MC4R signaling. MC4Rs function by integrating an agonist satiety signal from alpha-MSH (a cleavage product of POMC), and an antagonist signal provided by AgRP (Oswal and Yeo, 2007). The essential role of MC4Rs in the control of food intake is evident based on the hyperphagia and severe obesity seen in Mc4r −/− mice (Huszar et al., 1997), mice that overexpress the MC4R antagonist (Ay) (Kesterson et al., 1997), and in humans with naturally occurring mutations in MC4R (Lubrano-Berthelier et al., 2006). MC4Rs are densely expressed in the PVN, an area that receives inputs from both POMC and AgRP/NPY neurons, and their function, specifically in the PVN, is essential to maintain proper control of food intake (Balthasar et al., 2005). We demonstrate that Mecp2 CKO mice express normal levels of Mc4r, Pomc1, and Agrp but they have a small decrease in Npy in the ARC and a profound induction of Npy in the DMH. The finding that Npy is decreased in the ARC is not surprising since NPY neurons in the ARC typically respond to increases in leptin by downregulating the expression of Npy (Stephens et al., 1995). Increased Npy expression in the DMH, however, has been noted in a variety of rodent models that exhibit reduced function of melanocortin pathways including the agouti yellow mice (Ay), tubby mice, diet-induced obese mice (DIO), Mc4r −/− mice, and WT lactating rats (Chen et al., 2004; Guan et al., 1998; Kesterson et al., 1997). Importantly, increased expression of Npy in the DMH is not associated with all models of genetic obesity, since leptin-deficient ob/ob mice do not display detectable levels of Npy in the DMH (Kesterson et al., 1997). Thus, the ectopic expression of Npy in the DMH serves as a marker of disrupted MC4R signaling which seems to be occurring in the Mecp2 CKO mice.
Because our data point to a disruption of MC4R signaling in the Mecp2 CKO mice despite normal expression of upstream components of the pathway, we examined some of the downstream mediators of MC4R signaling in the PVN. Recent studies have suggested that both CRH and BDNF function as downstream effectors in the MC4R signaling pathway (Lu et al., 2003; Tsao et al., 2007; Xu et al., 2003). We found decreased expression of Crh in the PVN of the Mecp2 CKO mice. This is interesting because MC4R agonists increase the expression of Crh in the PVN (Lu et al., 2003), whereas Mc4r −/− mice have low levels of Crh in the PVN (see discussion in (Lu et al., 2003)), suggesting that the decreased Crh observed in the Mecp2 CKO mice may represent another marker of disrupted MC4R signaling.
We also found that Bdnf expression was decreased strictly in the PVN of Mecp2 CKO mice. Studies in neuronal cultures suggest that Bdnf is a direct target of MeCP2, and that MeCP2 functions as a transcriptional repressor to inhibit Bdnf transcription specifically from promoter III (Chen et al., 2003; Martinowich et al., 2003). These data conflict, however, with recent in vivo evidence demonstrating that BDNF levels are low rather than high in Mecp2 null mouse brain (Chang et al., 2006). Our data showing that Bdnf is decreased in neurons lacking MeCP2 is consistent with the in vivo data. Because low levels of Bdnf occur only in the cells where Mecp2 has been deleted in the Mecp2 CKO mice, it seems likely that the decreased expression is occurring as a primary result of MeCP2 deficiency. Given recent data showing that MeCP2 binds to the promoter region of many actively expressed genes (Yasui et al., 2007), that MeCP2 can function as both an activator and a repressor of gene expression in the hypothalamus (Chahrour et al., 2008), that Bdnf is significantly down-regulated in Mecp2 null mice, and that MeCP2 overexpression results in increased BDNF III transcript levels in cultured rat neurons and in MECP2 transgenic mice (Chahrour et al., 2008; Chang et al., 2006; Klein et al., 2007), we propose that MeCP2 functions as an activator of Bdnf transcription.
Several studies have documented the importance of BDNF in the VMH and DMH for feeding regulation (Tsao et al., 2007; Unger et al., 2007; Xu et al., 2003). Ay obese mice that have disrupted MC4R signaling have a 30–40% decrease in Bdnf expression in the VMH but no changes were observed in the PVN, DMH or lateral hypothalamus (LH) (Xu et al., 2003). Our data suggest a disruption in MC4R signaling, although we see decreased Bdnf expression in the PVN, and not the VMH. This result is not surprising since we have only deleted Mecp2 from the PVN, not the VMH, and we believe that the misregulation of Bdnf is occurring as a direct result of the absence of MeCP2. Future studies should examine Bdnf expression in the VMH and the PVN of Mc4r −/− mice to determine whether decreased expression occurs in one or both nuclei. Data supporting the fact that BDNF is important specifically in the PVN has come from a recent study that demonstrated that BDNF administration into the PVN was sufficient to significantly decrease food intake and body weight in mice (Wang et al., 2007). MC4Rs are densely expressed in the PVN and activation of these receptors leads to an acute release of BDNF in the hypothalamus which is necessary for MC4R-induced effects on appetite (Nicholson et al., 2007). If MeCP2 is important for the activation of Bdnf transcription, then the required release of BDNF in response to MC4R activation would likely be impaired, at least in the PVN of Mecp2 CKO mice.
We have demonstrated that MeCP2 plays an important role in the hypothalamus, an area of the brain where neurons are constantly responding to a variety of physiological stimuli. Specifically, we have demonstrated that the loss of MeCP2 seems to disrupt the typical homeostatic responses to food intake and to novel environment and stress consistent with a role for MeCP2 in the modulation of neuronal responses to changing stimuli (Chahrour and Zoghbi, 2007; Chen et al., 2003; Zhou et al., 2006). It is interesting that we found decreased expression of Bdnf in the PVN of Mecp2 CKO mice since conditional deletion of Bdnf from post-mitotic neurons results in enhanced aggression, and obesity (Koizumi et al., 2006; Lyons et al., 1999; Rios et al., 2001). Therefore, our data support a model where MeCP2 is required for the proper expression of Bdnf in the PVN, and we propose that misregulation of Bdnf expression may lead to disrupted social behavior in the form of aggression, and to hyperphagia and obesity owing to disrupted MC4R signaling.
Importantly, by using a conditional knockout strategy we were able to reproduce a subset of RTT or Mecp2 null phenotypes and to uncover novel functions of MeCP2. The many phenotypes that are seen in patients with RTT stem from MeCP2 dysfunction in neurons throughout the brain. Our study suggests that conditional deletion of Mecp2 in different cell populations or different regions of the brain will enable us to map the neuroanatomic origins of the complex behaviors and phenotypes seen in RTT and MECP2 disorders. Lastly, this study revealed that the more restrictive we are in choosing the neurons in which to delete Mecp2, the more we will learn about the function of MeCP2 in specific neurons and the neurobiological basis of certain behaviors.
Mice were maintained on a 12h light, 12h dark cycle with regular mouse chow and water ad libitum. Mecp2flox mice were a gift from Adrian Bird and Sim1-cre mice were from Brad Lowell and Joel Elmquist. All of the mice used in these experiments were generated by crossing heterozygous female Mecp2flox/+ mice that had been backcrossed to 129 Sv/Ev for 7 generations to male Sim1-cre mice on a pure FVB background. Sixty-four F1 129/FVB male littermates representing all 4 possible genotypes were group-housed immediately after weaning so that each cage contained one mouse of each genotype. All research and animal care procedures were approved by the Baylor College of Medicine Animal Care and Use Committee.
Mice were anesthetized (avertin) and perfused with 10% formalin for 8 minutes. Brains were dissected and post-fixed in 10% formalin overnight at 4°C. The samples were cryoprotected in 30% sucrose/1 X PBS and 50 μm sections were cut from hypothalamus blocks and suspended in 24 well tissue culture plates containing 1 X PBS. Sections were blocked in 2% normal goat sera with 0.3% triton X-100 for 1 hour at 4°C, incubated for 48-hours in a 1:100 dilution of anti-MeCP2 (Upstate cat# 07–013) and then incubated for 48 hours in 1:500 Cy3 labeled goat-anti rabbit (Jackson ImmunoResearch Labs). 1:2000 TOTO-3 (Invitrogen cat#T3604) was added to the last wash. Sections were mounted with ProLong Gold antifade mounting medium (Invitrogen cat# P-36931). Images were collected from optical sections that were obtained using a Zeiss 510 confocal microscope and processed using ImageJ (http://rsb.info.nih.gov/ij/).
Mice were left undisturbed for 12 hr and then retro-orbitally bled between 07:00–09:00. Before the first bleed, half of the animals were restrained in 50mL conical tubes for 30 min (stress condition). The remaining animals were kept in their home cages (basal condition). Three weeks later the mice were bled again, this time with the conditions reversed. Blood was collected in pre-chilled tubes containing lithium heparin, centrifuged at 0.8× g for 10 min and serum was collected and frozen at −80°C until it was analyzed. Serum corticosterone levels were measured using an enzyme-linked immunosorbent assay (ELISA) (IDS Inc., Fountain Hills, AZ). The data were analyzed by one-way ANOVA with Tukey’s post hoc analysis.
The experiment was performed as previously described (Spencer et al., 2006) with a few modifications. The open field apparatus consisted of a clear, open-topped Plexiglas box (40 × 40 × 30 cm) with photo beams to record the movement of the mouse within the box. Overhead lighting provided 200 lux of illumination and a white noise generator (Lafayette Instruments, Lafayette, IN) maintained background noise at 60 dB. Fifteen week old mice were placed in the center of the box and their activity was quantified over a 30-min period. Data were collected by a computer-operated digiscan optical animal activity system (Acuscan Electronics, Columbus, OH). Data were analyzed by one-way ANOVA with Tukey’s post hoc analysis.
The assay was performed as described (Spencer et al., 2006) with a few modifications. The light-dark box consisted of a clear Plexiglas chamber (36 × 20 × 26cm) with an open top separated from a covered black chamber (15.5 × 20 × 26cm) by a black partition with a small opening. The open chamber was illuminated to 700 lux. Fourteen week old mice were placed into the illuminated side and allowed to explore freely for 10 min. Mice were scored for the number and latency of entries and time spent in each compartment using a hand-held computer (Psion Workabout mx, Psion Teklogix) with the Observer program (Noldus Information Technologies). An entry was scored when the mouse placed all four feet into either the light or dark zone. White noise was present at 60 dB in the test chamber. Data were analyzed by two-way ANOVA for a Mecp2flox allele X Sim1-cre allele interaction and by one-way ANOVA with Tukey’s post hoc analysis.
This test was performed as previously described (Spencer et al., 2005) with a few modifications. Test mice at 20 weeks of age were individually housed in standard housing cages for 4 days. Each cage was separated into 2 compartments by a perforated barrier. On day 5 of individual housing, age- and gender-matched C57/BL6 partner mice were placed into the compartment opposite the test mice. Paired mice were co-housed in the separate halves of the partitioned cage for at least 18 hours. Following this period of induced familiarity, the time that test mice displayed directed interest in their partner mice was recorded during three different paradigms: test subject versus familiar partner, test subject versus unfamiliar partner and repeated test subject versus familiar partner. Each behavioral paradigm was assessed during three 5-minute intervals and was performed in sequential order. The presence or absence of aggressive behavior (direct attacks through the partition holes and tail rattling) was also scored during the three test intervals. Data were analyzed using a repeated measures ANOVA (genotype × time) and Tukey’s post hoc analysis.
Forty-two week old mice were individually housed for 2 weeks prior to the test to establish dominance in their home cage. Each singly housed test male was confronted with a lighter, group-housed wild type opponent (C57BL/6) in its home cage for 10 min. The occurrence of mounting, biting, tail rattling and attacking was recorded. To avoid injuries, the experiment was stopped if intense fighting occurred. Data were analyzed by a non-parametric Kruskal-Wallis test.
Food and water intake measurements were performed similarly to what was described in (Wade et al., 2008) with a few modifications. Mice were individually housed for 8 days in cages (45 × 24 × 17 cm) with feeders and water bottles mounted at one end. Intake of food and water were determined daily. The cages were placed on activity-monitoring platforms with two load beams at the front of the platform and a central pivot that allowed estimation of the position of the animal’s center of gravity (DiLog Instruments, Tallahassee, FL). Data were collected from the activity platform into daily event files using two personal computers (DiLog Instruments, Tallahassee, FL). A movement event was defined as the detection of a change in the animal’s center of gravity beyond a radius of 1 cm (calculated online from the animal’s body weight and the forces on two load beams after filtering with a 500 millisecond moving average window). The movement event files recorded the onset of movement events sampled every 20 milliseconds as well as the distance moved in x and y. The amount of food consumed was measured by weighing the chow placed in the cage each morning and subtracting the amount of chow left after a 24-hour period. Mean food intake and movement for the last 4 days of data collection was used for group comparisons to allow 4 days of acclimation to the housing conditions. Data were analyzed by one-way ANOVA with Tukey’s post hoc analysis.
Indirect calorimetry was performed to determine oxygen consumption. Mice were placed in a 4-chamber indirect open circuit calorimeter system (Oxymax; Columbus Instruments, Columbus, OH) maintained at 22°C. Food was removed at 1200 h. The mice were placed in the calorimetry chamber at 1300 h and were removed from the system at 1800 h. Each chamber (dimensions: 20 × 10 × 12 cm) received an airflow rate of 600 ml/min, and samples were collected at 15-min intervals. The calorimetry chambers were placed within photobeam activity monitors (San Diego Instruments, San Diego, CA) in order to measure physical activity simultaneously. Estimates of resting oxygen consumption or basal metabolic rate were made by averaging VO2 measurements during periods when the mice broke ten beams or less. Data were analyzed by one-way ANOVA with Tukey’s post hoc analysis.
Weight gain was measured each week beginning after weaning until 18 weeks of age. Nose-to-anus distance was measured after lightly anesthetizing each mouse in order to permit similar body extension. Intra-abdominal fat pads were dissected and weighed after the mice were anesthetized with isofluorane and then cervically dislocated. A Lunar PIXImus Densitometer (GE Medical Systems) was used to perform DEXA scans and Lunar PIXImus 2.10 software was used to analyze the results in order to estimate % body fat. The average of 2 scans was used for each mouse. Before the scan, mice were anesthetized using avertin at a dosage of 0.017 – 0.020 cc/g. Body weight data were analyzed using repeated measures ANOVA with Tukey’s post hoc, body length and fat pad weights were analyzed by one-way ANOVA with Tukey’s post hoc, and percent body fat was analyzed by Student’s t-test.
Blood was collected between 1300 and 1600 by retro-orbital bleed. Serum leptin levels were measured using a commercially prepared mouse leptin enzyme-linked immunoassay kit (American Laboratory Products Company, Windham, NH; cat.# 22-LEP-E06, lot# 090107). Data were analyzed by one-way ANOVA with Tukey’s post hoc.
Coronal sections from 2–4 mice of each genotype were analyzed for each gene. Tissue preparation and automated ISH were performed as previously described (Carson et al., 2002; Visel et al., 2004; Yaylaoglu et al., 2005) and as described online at http://www.genepaint.org/RNA.htm. Mc4r, Pomc1, Agrp, Npy, Bdnf, and Crh anti sense probes were generated from cDNA clones. The cDNA templates were amplified by PCR and used for in vitro transcription of digoxygenin labeled riboprobe. We performed quantitative analysis of the Bdnf ISH signal on sections spanning the entire PVN and VMH using the Celldetekt protocol to determine cellular gene expression strengths (Carson et al., 2005). Data were analyzed using a two-tailed Student’s t-test.
We are grateful to B. Lowell and J. Elmquist for the gift of Sim1-cre mice; to A. Bird for Mecp2flox mice; to the Baylor College of Medicine MRDDRC confocal core; to Christina Thaller and the Baylor College of Medicine in situ core; the Emory University Biomarkers Core Lab for technical assistance; to the CNRC Body Composition Lab for DEXA scans; and to M. Ramocki and other Zoghbi lab members for helpful comments on the manuscript. This work was funded by National Institutes of Health/National Institute of Neurological Disorders and Stroke grant NS057819 (HZ), National Institute of Child Health and Human Development Mental Retardation and Developmental Disabilities Research Center HD024064 (HZ), the International Rett Syndrome Foundation, the Simons Foundation and Autism Speaks (RS). H. Zoghbi is a Howard Hughes Medical Institute investigator.