|Home | About | Journals | Submit | Contact Us | Français|
The hypothalamic neuronal circuits that modulate energy homeostasis become mature and functional during early postnatal life. However, the molecular mechanism underlying this developmental process remains largely unknown. Here we use a mouse genetic approach to investigate the role of gamma-protocadherins (Pcdh-γs) in hypothalamic neuronal circuits. First, we show that rat insulin promoter (RIP)-Cre conditional knockout mice lacking Pcdh-γs in a broad subset of hypothalamic neurons are obese and hyperphagic. Second, specific deletion of Pcdh-γs in anorexigenic proopiomelanocortin (POMC) expressing neurons also leads to obesity. Using cell lineage tracing, we show that POMC and RIP-Cre expressing neurons do not overlap but interact with each other in the hypothalamus. Moreover, excitatory synaptic inputs are reduced in Pcdh-γ deficient POMC neurons. Genetic evidence from both knockout models shows that Pcdh-γs can regulate POMC neuronal function autonomously and non-autonomously through cell-cell interaction. Taken together, our data demonstrate that Pcdh-γs regulate the formation and functional integrity of hypothalamic feeding circuitry in mice.
The hypothalamus plays a crucial role in the control of feeding behavior and energy expenditure (Elmquist and Flier, 2004; Morton et al., 2006). Increasing evidence has demonstrated that the neuronal connectivity in the hypothalamus develops postnatally. In rodents the arcuate neurons that include POMC/CART and AgRP/NPY neurons are immature at birth and their projections do not fully innervate the dorsomedial hypothalamus (DMH), paraventricular nuclei (PVH) and lateral hypothalamus (LH) until the second to third postnatal week (Bouret et al., 2004a; Grove and Smith, 2003; Nilsson et al., 2005). During this period, leptin functions to promote the outgrowth and innervation of the arcuate axons (Bouret et al., 2004b). Concomitant with the development of ARH projections, two populations of PVH target neurons - neurosecretory (NS) and preautonomic (PA) cells - dramatically change their responses to melanocortins and NPY respectively (Melnick et al., 2007). Furthermore, the hypothalamic feeding circuits maintains considerable plasticity even in adulthood (Horvath, 2005). Rapid changes of synaptic inputs onto the arcuate neurons have been documented upon administration of leptin or estrogen mimics (Gao et al., 2007; Pinto et al., 2004). While ciliary neurotrophic factor (CNTF) is found to promote neurogenesis in the adult hypothalamus and regulate energy homeostasis (Kokoeva et al., 2005). Taken together, these findings indicate that the postnatal maturation of neuronal circuits is important in the maintenance of functional integrity of the hypothalamus.
Despite the importance of postnatal development of hypothalamic feeding circuitry, less is known about the molecular factors involved in this process. The recently discovered protocadherin genes are attractive candidates for regulating synaptic development (Benson et al., 2001; Takeichi, 2007). The three families of protocadherins, Pcdh-α, -β and -γ, encode nearly 60 isoforms, all of which share sequence homology with classic cadherins in their extracellular domains (Wu and Maniatis, 1999). Individual neurons express multiple Pcdh isoforms through a combination of differential promoter activation, cis-alternative splicing and allelic exclusion (Esumi et al., 2005; Tasic et al., 2002; Wang et al., 2002a). Pcdh proteins are enriched at but not limited to synapses (Kohmura et al., 1998; Phillips et al., 2003; Wang et al., 2002b). Together, these features have led to the hypothesis that Pcdh molecules might play a role in establishing neuronal connectivity in the vertebrate nervous system (Morishita and Yagi, 2007). For instance, it has been shown that olfactory and serotonergic axons fail to mature appropriately in the target areas in mice lacking Pcdh-αs (Hasegawa et al., 2008; Katori et al., 2009). In Pcdh-γ null mice, massive neuronal death and synaptic loss are found in the spinal interneurons during late embryonic development (Wang et al., 2002b). More importantly, the synapse deficit appears to be independent from neuronal apoptosis (Weiner et al., 2005).
Protocadherins are abundantly expressed in the adult hypothalamus. However, the neonatal lethality of Pcdh-γ nullizygous mice precludes investigation of their role in postnatal neural development. In this study, using a conditional knockout approach, we have found that mutant mice deficient for gamma-protocadherins in subsets of hypothalamic neurons are obese and hyperphagic due to abnormal neuronal connectivity in the hypothalamus. Therefore, Pcdh-γ mediated intercellular interactions play a critical role in the postnatal development of functional POMC neuronal circuitry in the hypothalamus.
The transgenic mouse strains including RIP-Cre, R26R-LacZ, R26R-GFP, mT/mG and BAX+/-, were originally obtained from Jackson Laboratory. The Pcdh-γfC3 and POMC-Cre transgenic mice were recently described (Lefebvre et al., 2008; Prasad et al., 2008; Xu et al., 2005b). The Pcdh-γfC3 conditional allele was originally generated in 129S7 ES cells. After germline transmission, the allele was backcrossed with C57BL/6-Tyr c-Brd for 4 generations. All experiments were carried out on 129S7/C57BL/6-Tyr c-Brd F4 hybrids. Both RIP-Cre; Pcdh-γfC3/fC3 and POMC-Cre; Pcdh-γfC3/fC3 conditional mutants were generated from double heterozygous mating. All mice were fed TEKLAD mouse diet 7904 (minimal 17% protein and 11% fat, Harlan TEKLAD, Madison WI) and water ad libitum unless otherwise stated. All procedures were reviewed and approved by the Northwestern University Institutional Animal Care and Use Committee.
Body weights were measured weekly. Food intake was determined in single-housed mice by weighing remaining diet. Data was discarded in the event of food spillage in the cage. Fat composition and lean mass were measured using an MRI whole body composition analyzer (Echo medical Systems, Houston, TX). The compensatory refeeding experiments were performed as previously described (Xu et al., 2005a).
Oxygen consumption was measured by an indirect calorimeter chamber (AccuScan Instruments Inc., Columbus, OH) at the University of Cincinnati Mouse Metabolic Phenotyping Center. Male mice, 10-12 weeks old, were placed in the chambers 3 hours before the dark cycle and energy expenditure measurements were collected for 24 hours.
Blood samples were collected from cardiac puncture on deeply anesthetized mice between 2 pm and 4 pm. Serum samples were collected by centrifugation and stored at -80°C. Glucose, triglycerides, and cholesterol were assayed by using commercial kits from Analox Instruments LTD (GMRD-002A, GMRD-195, GMRD-084 respectively). Insulin was measured by Mouse Insulin ELISA (Crystal Chem Inc), and leptin was assayed using mouse endocrine mutiplex kit from Linco (Kohsaka et al., 2007). LH was measured by radioimmunoassay. T3, T4 and corticosterone levels were assayed by the Hormone Assay & Analytical Services Core at Vanderbilt University.
Glucose tolerance tests were performed between 10 am to 11 am after a 17-hour fast. Blood samples were harvested from tail vein bleed. Glucose levels were determined using a Precision Xtra glucose monitor (Abbott Laboratories) at 0, 5, 15, 30, 60 and 120 minutes after intraperitoneal injection of glucose (2 g/kg body weight). Islet glucose-responsive insulin secretion was examined in static incubation studies. Briefly, pancreatic islets were isolated via collagenase digestion and left to recover overnight in RPMI-based media. Islets were then incubated for 1h in Krebs Ringer Buffer with 2 mM glucose to reach basal levels of insulin secretion. Groups of five size-matched islets were hand-picked and statically incubated in 2 mM, 12 mM or 20 mM of glucose for 1h. Supernatant was collected, and islets were then sonicated in acid-ethanol solution to obtain total insulin content. Amounts of secreted insulin and insulin content were determined by ELISA as described above.
The hypothalamus was dissected out using the optic chiasm as the anterior boundary, the mammillary bodies as the posterior border, and anterior commissure as the horizontal limit. Tissues were snap frozen in liquid nitrogen and stored at −80°C for subsequent analysis. Total RNAs were purified using RNeasy mini kits (Qiagen). Reverse transcription and quantitative PCR were performed with the ABI Prism 7700/7900 PCR instrument (Applied Biosystems) using SYBR green PCR Master Mix reagent kit (Applied Biosystems) as previously described (Kohsaka et al., 2007).
Mice were anesthetized with i.p. injection of Avertin (2.5% tribromoethanol) and perfused with PBS followed by 4% PFA. The brain and pancreas were post-fixed with 4% PFA for 4 hours at 4°C, cryoprotected with 30% sucrose and then frozen in Tissue-Tek O.C.T. 12 μm- or 20 μm- sections were collected for various antibody staining. The following primary antibodies were used in this studies: guinea pig anti-insulin (1:100, DAKO), monoclonal rat anti-BrdU (1:1500, AbD Serotec); mouse monoclonal anti-class III β-tubulin (TuJ1, 1:1000, Covance); mouse monoclonal anti-GFAP (1:500, Chemicon); rabbit polyclonal anti-pSTAT3 (1:200, Cell Signaling Technology), rabbit anti-POMC (1:400, Phoenix pharmaceuticals), rabbit anti-AgRP (1:2000, Phoenix pharmaceuticals), sheep anti-a MSH (1:4000, Millipore), rat-anti-GFP (1:5000, MBL), guinea pig anti-VGLUT2 (1:3000, Millipore), rabbit anti-GFP (1:10,000, Invitrogen), rabbit anti-LacZ (a gift from Dr. Holmgren). Single and double staining were visualized with FITC-, DyLight488, DyLight 594- and Cy3-conjugated secondary antibodies (1: 500 and 1: 1000, Jackson ImmunoReseach) and counter stained with DAPI (Sigma-Aldrich). Images were collected with Zeiss Axiovision 4.6 using a Zeiss Axiovert 200M with Apotome module.
For BrdU labeling, pregnant mice at E12.5 or E14.5 were injected intraperitoneally with 50mg/kg body weight BrdU solution (BD Biosciences), and sacrificed one hour after injection. To evaluate leptin-induced STAT3 phosphorylation, 15 days old sex matched mice were intraperitoneally injected with leptin (National Hormone and Peptide Program; 10.0 mg/kg body weight) or a similar volume of saline at 10 a.m. 45 minutes after leptin injection, the mice were anesthetized and perfused with 4% PFA in PBS. For pSTAT3 single or double labeling, sections were sequentially processed with 1% NaOH, 0.3% glycine and 0.03% Sodium dodecyl sulfate, followed by antibody staining as described above. To colocalize POMC expressing neurons, mice were pre-treated with i.c.v. injection of colchicine (Sigma) 24 h before euthanasia to block axonal transport. In brief, deeply anesthetized mice received 1 μl colchicine solution (20μg/μl in 0.9% sterile saline, 0.75 μg/g body weight) at 1 mm lateral and 0.3 mm posterior to bregma and 2 mm ventral to the surface of the skull using standard sterile surgical procedures. Brains were harvested 24 hours later for immunohistochemistry as described above.
LacZ staining was performed as described before (Wang et al., 2002a). For in situ hybridization, a GFP riboprobe was generated by in vitro transcription with T3 polymerases and the sample processing, hybridization and development of fluorescent signals were performed as previously described (Wang et al., 2002a).
For quantification of active caspase 3 positive cells, immunohistochemistry staining was processed as described above. Pictures were taken every 3rd sections using a 10× objective with ARC positioned in the midline, starting from where ARH and VMH were first visible in a caudal to rostral manner throughout the entire ARC. Active caspase 3-positive cells were found in the ARC, the DMH, the VMH, the LH and PVH, and were counted using Zeiss Axiovision 4.6 software by an observer blinded to the genotype. Cell counts were performed every 3rd section in 3-5 animals per group (12 sections/animal) and represent the average total number of cells counted.
Anesthetized animals were first perfused transcardially with saline, followed by a fixative containing 4% PFA and 0.1% Glutaraldehyde in 0.1M phosphate buffer (PB, pH 7.2-7.4). After an overnight fixation with 4% PFA, the brain tissues that included the hypothalamus were cut into 50 μm coronal sections using a vibrotome (Leica VT1000S). The vibrotome sections were then extensively washed with PB, blocked with 50 mM Glycine, and permeabilized with 0.05% Triton X-100/PB for 30 minutes. After blocking with 5% bovine serum albumin/2% goat serum / PBS for 30 minutes, the sections were incubated with 1% goat serum/0.1M PBS containing rabbit anti-LacZ antibody (1:200) overnight at 4°C. After multiple washes, the samples were incubated with NanaGold-conjugated Fab' anti-rabbit secondary antibody (1:250, Nanoprobes, NY)) at room temperature for 1 hour and at 4°C for overnight. The brain slices were subsequently washed with 0.1M PBS and fixed with 2% glutaraldehyde in PBS for one and half hours. After washes with PBS and distilled water, silver enhancement (HQ Silver enhancement kit, Nanoprobes, NY) was performed in the dark for 12 min. Finally, the samples were fixed in 0.2% OsO4 for 30 min, block stained with uranyl acetate, dehydrated in ethanol and flat embedded in LX112. Semi-thin sections were cut at 1 μm and stained with 1% Toluidine Blue to evaluate the quality of preservation. Ultra-thin sections (50 nm) were cut, mounted on formvar-coated single-slot copper grids and stained with uranyl acetate and lead citrate by standard methods. Stained grids were examined under Philips CM-12 electron microscope (FEI; Eindhoven, The Netherlands) and photographed with a Gatan (4k ×2.7k) digital camera (Gatan, Inc., Pleasanton, CA) at different magnifications. After obtaining digital montages with Adobe Photoshop, symmetric and asymmetric synapses were counted on all lacZ-positive cells only if the pre- and/or postsynaptic membrane specializations were clearly seen. Synapses without clearly symmetric or asymmetric membrane specializations were excluded from the assessment.
All data are presented as means with standard error of the means. Detailed information on sample sizes, genotype and p values are marked on the individual figures or described in the corresponded legends. Comparisons between two groups are analyzed using two-tailed unpaired student's t-test or two-way ANOVA. For two-way ANOVA, if the overall F value was significant, Bonferroni post-hoc test were performed using GraphPad.
To evaluate the function of Pcdh-γs in a tissue-specific fashion, we used a recently characterized conditional Pcdh-γ allele (Lefebvre et al., 2008; Prasad et al., 2008), Pcdh-γfC3. In this allele, the shared third constant exon (C3) is fused in-frame with GFP and flanked by loxP sites (Figure 1A). Therefore, Pcdh-γ–GFP provides an efficient way to monitor expression. The Cre-mediated excision of Pcdh-γfC3 leads to a complete deletion of constant exon 3 including the spicing acceptor site as well as the polyA adenylation site for all Pcdh-γ mRNAs. Consequently, western blot analysis failed to detect any Pcdh-γ proteins in Actin-Cre; Pcdh-γfC3/fC3 mice, indicating that this allele when excised is effectively a null (Lefebvre et al., 2008; Prasad et al., 2008).
Pcdh-γs are abundantly expressed in the neuroendocrine tissues including pancreatic β cells and the hypothalamus (Figure 1). To selectively delete Pcdh-γs in these tissues, we chose a previously characterized rat insulin promoter driven Cre (RIP-Cre) transgenic line in which Cre is expressed in the pancreatic β cells as well as hypothalamic neurons (Gannon et al., 2000). By crossing RIP-Cre with a reporter line Gt(Rosa)26Sortm1sor(R26R-LacZ), we confirmed the expression pattern of RIP-Cre (Figure 1B). In the hypothalamus, Cre activity is strong in the ARH, VMH and DMH and weak in the LH and PVH as visualized by the expression of LacZ (Figure 1B, right panel and data not shown). We next intercrossed RIP-Cre with Pcdh-γfC3 to generate the RIP-Cre; Pcdh-γfC3/fC3 mice (designated as Pcdh-γ KORIP-Cre). Pcdh-γ KORIP-Cre mice were born at the expected Mendelian frequency, viable and fertile. To verify that Pcdh-γs are deleted in the Cre-expressing cells, we used immunostaining for Pcdh-γ-GFP to monitor expression. Figure 1C shows that Pcdh-γs were deleted in the majority of pancreatic β cells. For the hypothalamic samples, it is however difficult to distinguish a truly GFP-negative neuron from its surrounding cells because Pcdh-γ proteins are localized in both soma and neuronal processes. To overcome this technical limitation, we performed in-situ hybridization to detect Pcdh-γ mRNAs that are mostly localized in cell bodies (Wang et al., 2002a). This result clearly showed that Pcdh-γ mRNAs were absent in subsets of hypothalamic neurons (Figure 1D). Therefore, we successfully inactivated Pcdh-γ genes in the majority of β-cells and a subset of hypothalamic neurons.
To investigate the physiological consequences of Pcdh-γ ablation in neuroendocrine tissues, we compared the body weights of RIP-Cre; Pcdh-γfC3/fC3 (Pcdh-γ KORIP-Cre) with those of their RIP-Cre; Pcdh-γfC3/+ littermates. We chose the RIP-Cre positive littermates as the control because recent data suggested that RIP-Cre expression may independently effect β cell function (Lee et al., 2006). We found no difference in body weight or body composition between Pcdh-γfC3/fC3 mice and Pcdh-γfC3/+ or Pcdh-γ+/+ littermate controls (Figure S1), indicating that Pcdh-γfC3/fC3 alone has no effect on the regulation of body weight. Both male and female Pcdh-γ KORIP-Cre mice weighed significantly more than the littermate controls, starting from 7 weeks after birth (Figure 2A and 2B). There was no difference in linear growth between the genotypes (Figure S2). MRI measurement showed that the increased body weight of Pcdh-γ KORIP-Cre mice was mainly due to increased fat composition at 12 weeks of age (Figure 2B).
We next monitored daily food consumption of Pcdh-γ KORIP-Cre mice during a 9-week interval (Figure 2C). In comparison with their littermate controls, both male and female Pcdh-γ KORIP-Cre mice consumed significantly more food starting at 5-6 weeks even prior to the onset of obesity in these mice (Figure 2C). Interestingly, we found that Pcdh-γ KORIP-Cre mice maintained a normal compensatory refeeding response after a 48-hour fast (Figure 2D). To fully examine energy balance in Pcdh-γ KORIP-Cre mice, we subjected ad lib fed mice to indirect calorimetry and measured VO2, VCO2, respiratory quotient and heat (Figure S3). While there were no significant changes in metabolic rate (VO2) in adult Pcdh-γ KORIP-Cre mice, the respiration quotient (RQ) was significantly reduced, indicative of increased fat oxidation (Nogueiras et al., 2007). The increased lipid oxidation could be an intrinsic metabolic phenotype for Pcdh-γ KORIP-Cre mice as reported in other animal models (Nogueiras et al., 2007; Tong et al., 2008; Wortley et al., 2005). Alternatively, it might be secondary to obesity. Our data currently do not distinguish these two alternatives. Nevertheless, our analyses on food intake and energy expenditure of Pcdh-γ KORIP-Cre mice show that chronic hyperphagia is a contributing cause of obesity in Pcdh-γ KORIP-Cre mice.
Central administration of insulin decreases food intake and body adiposity, thus we tested whether insulin deficiency may contribute to obesity and hyperphagia observed in Pcdh-γ KORIP-Cre mice (Bruning et al., 2000; Choudhury et al., 2005; Niswender et al., 2004). Pcdh-γs are abundantly expressed in pancreatic β cells (Figure 1C, left). To examine whether the inactivation of Pcdh-γs might affect β-cell insulin secretion, we analyzed the glucose regulation in both 4-week and 12-week old Pcdh-γ KORIP-Cre mice. In young and pre-obese (4 weeks old) mice, glucose and insulin levels under both fasting and fed conditions were normal (Figure S4). Since neither hyperglycemia nor hypoinsulinemia were present in the pre-obese young mice, it is unlikely that hyperphagia in these animals can be attributed to hypoinsulinemia. In contrast, at a latter age (12 week old) following the onset of obesity, Pcdh-γ KORIP-Cre mice developed elevated levels of insulin yet they still had normal glucose levels (Supplemental Table 1), consistent with the onset of obesity-related insulin resistance. Furthermore, we observed no significant difference between Pcdh-γ KORIP-Cre and control mice in glucose tolerance upon intraperitoneal (i.p.) glucose challenge (Figure S5A). These results are consistent with β -cell compensation for insulin resistance. To further assess β-cell function in Pcdh-γ KORIP-Cre mice, we isolated pancreatic islets from these obese mice and performed in vitro glucose challenge studies. We found that islets from Pcdh-γ KORIP-Cre and control mice released equivalent amounts of insulin following glucose stimulation (Figure S5B). Collectively, our data demonstrate that Pcdh-γ KORIP-Cre mice have normal insulin secretion and that Pcdh-γs are not required for normal β-cell glucose responsiveness.
Because Pcdh-γ deficiency does not affect β-cell function in Pcdh-γ KORIP-Cre mice, we focused on analyzing the role of Pcdh-γs in the hypothalamus. To examine whether the endocrine function of the hypothalamus is affected in Pcdh-γ KORIP-Cre mice, we measured levels of thyroid hormones (T3 and T4), luteinizing hormone (LH) and corticosterone and found no significant difference between genotypes (Supplemental Table 1). Thus, deletion of Pcdh-γ in subsets of hypothalamic neurons appears to have no overt impact on markers of hypothalamic neuroendocrine homeostasis. Histological analysis showed that all distinct hypothalamic nuclei are present in the hypothalamus of adult Pcdh-γ KORIP-Cre mice (Figure 3A). Neurogenesis of hypothalamic neurons occurs between E11.5-E16.5 during embryonic development (Markakis, 2002). We used BrdU to pulse-label hypothalamic progenitors at both E12.5 and E14.5. We observed similar numbers of BrdU positive cells in Pcdh-γ KORIP-Cre and control embryos (Figure 3B). Thus, it appears that Pcdh-γs do not play a role in neurogenesis of hypothalamic neurons.
Pcdh-γs are known to regulate both neuronal survival and synaptic development by affecting intercellular interactions (Fernandez-Monreal et al., 2008; Lefebvre et al., 2008; Prasad et al., 2008; Wang et al., 2002b; Weiner et al., 2005). Neuronal apoptosis is one of the known defects in Pcdh-γ null mice (Lefebvre et al., 2008; Prasad et al., 2008; Wang et al., 2002b), therefore we examined whether Pcdh-γ deficient hypothalamic neurons underwent increased apoptosis in Pcdh-γ KORIP-Cre mice. Using active-caspase-3 staining, we surveyed cell death in the hypothalamus at different developmental stages including E17, E18, P0, P2, P3, P5 and P10. We found that in both Pcdh-γ KORIP-Cre and littermate controls, a significant amount of cell death occurred at both P0 and P5. Quantitative analysis showed that the number of apoptotic cells significantly increased in the hypothalamic region of Pcdh-γ KORIP-Cre mice at P0 but not at P5 (Figure 3C, 3D and 3E). We further confirmed that the apoptotic cells were differentiated neurons and not astrocytes (Figure 3F). Consistent with the findings in the spinal cord and retina, Pcdh-γs also regulate the survival of hypothalamic neurons. Moreover, loss of hypothalamic neurons could be a contributing factor for the obese phenotype seen in Pcdh-γ KORIP-Cre mice.
The identities of RIP-Cre expressing neurons in the hypothalamus are unknown, making it difficult to explain how Pcdh-γ deficiency in RIP-Cre expressing cells might impact the development of neuronal circuitry in the hypothalamus. Therefore, we first used Cre-mediated cell labeling techniques to examine the intercellular connectivity of RIP-Cre expressing cells. Two Cre reporter transgenic mouse lines ROSA26-GFP (Gt(ROSA)26Sortm2Sho/J) and mT/mG (Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J) are used in our analysis. In the ROSA26-GFP line, GFP primarily labels the cell bodies of Cre-expressing neurons (Figure 4B), whereas membrane-bound GFP in the mT/mG line robustly labels all neuronal processes and membrane organelles in the soma. For example, to label neuronal projections of POMC-Cre expressing neurons, we generated POMC-Cre transgenic mice with the mT/mG reporter allele. In these mice, membrane bound GFP labeled both the soma of POMC neurons in the ARH (Figure 4C) and the α-MSH-positive axonal projections in the DMH (Figure 4D). Previous studies have shown that RIP-Cre expressing cells do not overlap with POMC-positive or AgRP-positive neurons in the ARH (Choudhury et al., 2005). The double immune staining of the hypothalamic sections from RIP-Cre; R26R-GFP mice confirmed this observation and we found that the cell bodies of RIP-Cre expressing neurons and POMC neurons are often adjacent to each other in the ARH (Figure 4B and data not shown), These observations suggest that RIP-Cre and POMC expressing cells might be in contact through their neuronal processes. RIP-Cre positive cells are also negative for GFAP, excluding the possibility that RIP-Cre cells are astrocytes (Figure S6). To better visualize neuronal projections of RIP-Cre expressing cells, we labeled their neuronal processes with the mT/mG reporter allele. We found that the POMC-positive soma are clearly surrounded by membrane GFP-labeled neuronal processes originated from RIP-Cre positive cells (Figure 4E). These neuronal processes on POMC soma might originate either proximally or distally from RIP-Cre expressing cells. Reciprocally, α-MSH positive axons project to the vicinity of the cell bodies of RIP-Cre expressing neurons in the ARH (Figure 4F, left). In the target areas to which POMC axons project, α-MSH positive axonal projections interconnect with the soma and neuronal processes of RIP-Cre-expressing neurons in the DMH and LH (Figure 4F and 4G). Due to the technical limitation of partial labeling, we were unable to visualize extensive synaptic contacts between the two cell types. Therefore, it remains to be determined whether the interactions between two cell types are synaptic or extrasynaptic. In addition, we found that RIP-Cre expressing cells in the ARH and DMH express VGlut2 (Figure 4H and DMH data not shown), a vesicle transporter for the excitatory neurotransmitter glutamate, suggesting that these neurons are glutamatergic. Taken together, our analysis provided the evidence that RIP-Cre expressing neurons interact with POMC neurons.
The evidence that RIP-Cre expressing neurons interact with POMC neurons prompted us to analyze the function of the first order melanocortinergic signaling neurons in Pcdh-γ KORIP-Cre mice. Using quantitative RT-PCR, we determined the expression of several hypothalamic neuropeptides that are known to regulate energy homeostasis (Kohsaka et al., 2007). We found that anorexigenic POMC expression decreased significantly in Pcdh-γ KORIP-Cre hypothalamus in both genders (Figure 5A). The expression levels of AgRP and NPY in Pcdh-γ KORIP-Cre mice were downregulated at 14 weeks of age, but recovered to a normal level at 30 weeks (Figure 5A and Figure S7). Although our data did not argue that the phenotype of Pcdh-γ KORIP-Cre mice was solely caused by defects in POMC neurons, we decided to focus our subsequent work on analyzing the role of Pcdh-γs in POMC neurons because Pcdh-γ deficiency exhibited a much stronger effect on the regulation of POMC expression (Figure 5A) and markers and genetic tools for POMC neurons made it a tractable system for more detailed studies.
Using both immunostaining and RT-qPCR analysis, we showed that the onset of POMC downregulation in Pcdh-γ KORIP-Cre mice occurred during early postnatal development, and that this effect became obvious from P10 throughout adulthood (Figure 5B-C and Figure S8). Decreased POMC mRNA levels could be due to either transcriptional downregulation of the POMC gene, or loss of POMC expressing cells. Previous studies showed that either defect can cause obesity in mice (Xu et al., 2005a; Yaswen et al., 1999). To distinguish between these two alternatives, we first examined whether POMC positive neurons undergo apoptosis. We found no POMC neurons that were positive for active Caspase-3 in Pcdh-γ KORIP-Cre mice at different developmental stages (data not shown). However, we could not rule out the possibility that a small set of POMC neurons, which were missed in our assay, did undergo apoptosis. Consistent with the observation that no POMC neurons were found to undergo apoptosis, immunostaining and RT-qPCR analysis also demonstrated that the POMC signals were not changed in Pcdh-γ KORIP-Cre mice during the neonatal period (P0 in Figure 5C and Figure S8A).
Two additional lines of evidence further suggested that POMC neurons were not lost in Pcdh-γ KORIP-Cre mice. First, POMC and AgRP expressing neurons are two major groups of leptin target cells in the ARH (Elias et al., 1999). Upon intraperitoneal leptin injection, we found no significant difference between the number of phospho-Stat3 (pStat3) positive neurons in the ARH of Pcdh-γ KORIP-Cre mice and that of their control littermates (Figure 5D), suggesting no significant loss of leptin-responsive neurons in Pcdh-γ KORIP-Cre mice. Second, Pcdh-γ KORIP-Cre mice exhibited feeding behavior different from that of mice in which POMC neurons were specifically ablated. For example, Pcdh-γ KORIP-Cre mice were capable of increasing feeding in response to starvation (Figure 2D), whereas POMC neuron-ablated mice failed to do so (Xu et al., 2005a).
It has been previously shown that Pcdh-γs can regulate neuronal connectivity without affecting neuronal survival (Weiner et al., 2005). To further examine whether POMC downregulation is independent from apoptosis, we generated double knock-out mice for Pcdh-γ and the proapoptotic gene Bax (RIP-Cre; Pcdh-γfC3/fC3; Bax-/-), in which neuronal apoptosis is blocked (White et al., 1998). Active caspase-3 staining confirmed that no apoptotic cells were found in the hypothalamus of these double knock-out mice at P0 (Figure 5E). More importantly, POMC expression was still significantly reduced in RIP-Cre; Pcdh-γfC3/fC3; Bax-/- mice, without any detectable cell death (Figure 5F). These results indicated that the reduction of POMC signal in Pcdh-γ KORIP-Cre mice was due to transcriptional downregulation rather than a loss of POMC expressing cells.
RIP-Cre; Pcdh-γfC3/fC3; Bax-/- mice were generated at a ratio of 1 out of 32 from an intercross of RIP-Cre; Pcdh-γ+/fC3; Bax+/- mice. Because Bax-/- mice are sterile, it is extremely difficult to generate a sufficient number of gender and genotype matched littermate animals for longitudinal metabolic studies. Furthermore, the infertility and gender related developmental defects for Bax-/- mice would add more complexity for phenotypic analysis (Forger et al., 2004; Russell et al., 2002). Therefore, we used Bax mutation only for the molecular analysis of POMC expression in Pcdh-γ KORIP-Cre mice. Our data on RIP-Cre; Pcdh-γfC3/fC3; Bax-/- mice demonstrate that defects in Pcdh-γ mediated cell-cell interaction could affect the development of POMC neurons without a loss of Pcdh-γ deficient neurons.
In summary, we have demonstrated that Pcdh-γ KORIP-Cre mice are obese and hyperpagic. Two non-overlapping populations of hypothalamic neurons, RIP-Cre expressing and POMC-positive neurons, interact with each other. Deletion of Pcdh-γs within the RIP-Cre expressing neuronal population, affects the postnatal maturation of POMC neurons and downregulates POMC expression in a non-autonomous fashion. Thus, Pcdh-γ mediated cell-cell interactions between RIP-Cre positive neurons and POMC neurons appear to be critical for the functional integrity of POMC neuronal circuits.
The Pcdh-γ mediated intercellular interactions and neuronal connectivity provide an explanation for the non-autonomous function of Pcdh-γ in the regulation of POMC neurons during postnatal development. Defective intercellular interactions between RIP-Cre neurons and POMC neurons are most likely responsible for the downregulation of POMC expression seen in the hypothalamus of Pcdh-γ KORIP-Cre mice. If so, given the homophilic nature of Pcdh-γ interaction (Fernandez-Monreal et al., 2008), we predicted that Pcdh-γ deletion within POMC neurons themselves would give rise to similar defects in regulation of POMC expression and control of body weight and appetite.
To test the hypothesis that Pcdh-γ deficiency in POMC neurons causes a defect in the hypothalamic regulation of energy homeostasis, we generated POMC-Cre; Pcdh-γfC3/fC3 (designated as Pcdh-γ KOPOMC-Cre) mice. As predicted, Pcdh-γ KOPOMC-Cre mice exhibited a significant increase in body weight in comparison with littermate control mice (POMC-Cre; Pcdh-γ+/fC3) (Figure 6A). Daily food intake experiments showed that Pcdh-γ KOPOMC-Cre mice were hyperphagic (Figure 6B). Body composition analysis further demonstrated that increased adiposity accounted for most of the weight gain in these mutant mice (Figure 6C). Like Pcdh-γ KORIP-Cre mice, blood levels of thyroid hormones, luteinizing hormone (LH) and corticosterone were normal in Pcdh-γ KOPOMC-Cre mice (Supplementary Table 2). Thus, the similarity of the obesity phenotypes in Pcdh-γ KOPOMC-Cre and Pcdh-γ KORIP-Cre mice supports the notion that Pcdh-γ mediated intercellular interactions play a critical role during the postnatal development of hypothalamic neuronal circuitry.
To further examine the role of Pcdh-γs in POMC neurons, we asked whether Pcdh-γ depletion in POMC neurons would affect POMC expression the same way it does in RIP-Cre neurons. We failed to detect any increase in apoptosis in the hypothalamus of Pcdh-γ KOPOMC-Cre mice at P0 and P5 and found that no POMC neurons were positive for active-caspase-3 (Figure S9). Using both immunohistochemistry and qPCR analyses to detect POMC expression, we showed that similar to Pcdh-γ KORIP-Cre mice, the postnatal expression of POMC was dramatically downregulated in Pcdh-γ KOPOMC-Cre mice (Figure 7A-B and Figure S10). It has previously been shown that leptin induces c-fos expression in POMC-positive ARH neurons, but not in AgRP/NPY neurons (Elias et al., 1999). We identified c-fos positive neurons in the ARH of both Pcdh-γ KOPOMC-Cre and control littermate mice injected with leptin (Figure 7C-D), suggesting that leptin-responsive neurons remained in Pcdh-γ KOPOMC-Cre mice. Therefore, in Pcdh-γ KOPOMC-Cre mice, Pcdh-γs also autonomously regulate POMC transcription during the critical period of postnatal development. Taken together, our genetic studies on Pcdh-γ function clearly show that the absence of Pcdh-γs in either POMC neurons themselves or cells that interconnect with POMC neurons (RIPCre neurons) causes a downregulation of postnatal POMC expression. This downregulation serves as an indication that abnormal development and function of POMC neuron circuitry exist in both Pcdh-γ KOPOMC-Cre and Pcdh-γ KORIP-Cre mice. The simplest model derived from these complementary data is that Pcdh-γ mediated intercellular interactions are critical for the connectivity and proper function of POMC neurons.
Pcdh-γs have an important function in maintaining synaptic connectivity for spinal cord interneurons, independently from their role in regulating neuronal survival (Weiner et al., 2005). Therefore, we examined the possibility that Pcdh-γ deficiency could lead to loss of synapses. In Pcdh-γfC3 mice, all neurons are labeled with Pcdh-γ-GFP, thus preventing us from using an established POMC-GFP transgene to label live POMC positive neurons for electrophysiological recording. To overcome this technical limitation, we generated compound Pcdh-γ KOPOMC-Cre mice with the R26R-LacZ reporter allele to label POMC neurons by LacZ. To study the synaptic inputs onto the POMC perikaryal membrane, we performed immuno-electron microscopy analysis on LacZ-positive neurons using a NanoGold and silver enhancement method (Figure S11). After identifying POMC (LacZ-positive) cells (Figure 7E), we quantitatively analyzed the numbers of symmetric (inhibitory) and asymmetric (excitatory) synapses formed on randomly picked POMC neurons from mutant and control mice. This analysis showed that asymmetric synapses are significantly reduced in Pcdh-γ deficient POMC neurons (Figure 7F), thus providing evidence that Pcdh-γs are important for synapse formation in POMC neurons. We speculate that the deficit of excitatory synaptic inputs to POMC neurons might cause a decrease in neural activity and might consequently downregulate POMC expression in Pcdh-γ deleted POMC neurons. Thus, abnormal synaptic connectivity is an underlying defect in Pcdh-γ deficient hypothalamic neurons.
In this study, we show that Pcdh-γs, a large family of cell surface molecules, play an important role in the formation of functional hypothalamic feeding circuitry during early postnatal development. Using Cre-mediated cell tracing, we show that RIP-Cre expressing neurons interact with POMC neurons in the hypothalamus. Using both RIP-Cre and POMC-Cre to delete Pcdh-γs in subsets of hypothalamic neurons, we demonstrate that intercellular interactions mediated by Pcdh-γs are critical for the maturation of peptidergic POMC neurons in the ARH. With the loss of Pcdh-γs in the hypothalamic melanocortinergic circuitry, abnormal neuronal connectivity in Pcdh-γ conditional KO mice, as manifested by a developmental downregulation of POMC expression and a reduction of excitatory synaptic inputs on POMC neurons, contributes to hyperphagia and obesity.
The prevailing model of Pcdh function in vertebrate neural development derives from the analysis of two Pcdh gene clusters, Pcdh-γ and Pcdh-α (Emond and Jontes, 2008; Hasegawa et al., 2008; Lefebvre et al., 2008; Prasad et al., 2008; Wang et al., 2002b; Weiner et al., 2005). Two seemingly independent Pcdh roles emerge from these studies. First, Pcdhs are required for the survival of specific neuronal subpopulations. For example, Pcdh-γ deficiency causes significant increase in apoptosis of spinal interneurons and retina ganglion cells (Lefebvre et al., 2008; Wang et al., 2002b). Knock-down of Pcdh-α by morphonilos in zebrafish also induces excessive neuronal death during neurogenesis (Emond and Jontes, 2008). It is worth noting that apoptosis has only been seen in certain subsets of neurons while some neuronal populations such as motor neurons and sensory neurons have been spared in both species. Second, Pcdhs are required for the establishment of synaptic connectivity. Pcdh-γs appear to be required for the maturation and maintenance of synapses in the spinal cord, as synaptic loss persists even in the absence of neuronal death in Pcdh-γdel/del, Bax-/- double knockout mice (Weiner et al., 2005). Pcdh-αs have also been shown to affect the axonal coalescence of olfactory sensory neurons (Hasegawa et al., 2008). The findings of the present study are consistent with these two roles of protocadherins during neural development. First, we have shown that deletion of Pcdh-γs increases the incidence of programmed cell death in the developing hypothalamus. Second, we have found that excitatory synaptic inputs to POMC soma decrease significantly in the absence of Pcdh-γs. Both cellular defects likely contributed to the metabolic and behavioral phenotypes seen in the Pcdh-γ conditional KO mice. In Pcdh-γ KORIP-Cre mice, both neuronal death and abnormal connectivity could contribute to the non-autonomous effect of Pcdh-γ deficiency on POMC neurons. In Pcdh-γ KOPOMC-Cre mice, the observed synaptic defects in POMC neurons provide direct evidence for a synaptic function of Pcdh-γs in the feeding circuitry. In rodents, from birth to 3 weeks of age, axons from the arcuate POMC neurons innervate the PVH, DMH and LH (Simerly, 2008), forming synaptic connections with target neurons in these areas. At the same time, POMC neurons establish inhibitory synapses with neighboring AgRP/NPY neurons, and receive excitatory inputs including those from the VMH (Cowley et al., 2001; Dhillon et al., 2006; Horvath et al., 1997; Horvath et al., 1992; Sternson et al., 2005; Tong et al., 2008) and likely from neighboring RIP-Cre expressing neurons (this study). Loss of Pcdh-γs in either RIP-Cre positive neurons or POMC neurons themselves affects the postnatal development of POMC neurons and leads to a downregulation of POMC expression. Given the nature of protocadherin molecules and their actions in other neuronal circuits, our data strongly support the notion that Pcdh-γs regulate the formation of interneuronal connections of the melanocortinergic circuitry. During normal development of hypothalamic feeding circuitry, Pcdh-γs may function as an adhesive force to stabilize appropriate synapses within hypothalamic neuronal circuits. Our recent proteomic analysis reveals that Pcdh-γs form large protein complexes with postsynaptic components of excitatory synapses (Han et al., 2009), providing further molecular evidence for a critical role of Pcdh-γ in synaptic development.
On a broader scale, our data add to the emerging concept that altering synaptic connectivity of the hypothalamic feeding circuitry during a critical developmental period has a profound effect on the regulation of energy homeostasis later in life (Horvath, 2005; Simerly, 2008). It has also been shown that environmental cues (i.e., nutrition status and energy availability) are capable of influencing the formation of the hypothalamic feeding circuitry (Bouret et al., 2008; Enriori et al., 2007; Simerly, 2008; Yura et al., 2005). Cell surface heparan sulfate proteoglycans such as syndecan-3 has been shown to modulate AgRP and regulate feeding behavior in mice (Reizes et al., 2001). Pcdhs may be among the molecular components that regulate this developmental process. With the rapid increase of juvenile obesity in the world population, it is becoming increasingly important to understand the molecular basis of how metabolic and nutritional status can imprint hypothalamic development during early postnatal life. Intervening with this early developmental process may influence body weight and energy homeostasis in adulthood (Friedman, 2003).
We thank Dr. Jon Levine for technical advice and Drs. Joshua Sanes, Kelly Mayo, Robert Holmgren and Dan Tombly for comments on the manuscript. We thank Dr. Robert Holmgren for rabbit anti-β-galactosidase antibody, Dr. Jianhua Cang for the help with stereotaxic injection and Dr. Jon Levine for measuring serum LH level. We thank The University of Cincinnati Mouse Metabolic Phenotyping Center (DK5963) and Hormone Assay & Analytical Services Core at Vanderbilt University (DK20593) for measuring metabolic parameters and serum hormone levels, respectively. This work was supported by 5R01NS051253 from NIH and a grant from the Whitehall Foundation to XW.
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.