A hypothalamic pathway for Augmentor α–controlled body weight regulation

To elucidate the physiological roles of Augα and Augβ in mammals, we generated individual knockout mice and double knockout mice of Augα and Augβ. Augα-knockout mice were generated by replacing exons 1 to 4 with a green fluorescent protein (GFP) expression cassette to enable analysis of the expression pattern of endogenous Augα protein in both tissues and cells. Similarly, Augβ-knockout mice were generated by replacing exon 1 with a GFP expression cassette. Double knockout mice were generated by crossing Augα- and Augβ-deficient mice (SI Appendix, Fig. S1 A–C). We also generated another line of knockout mice by replacing Augα with a LacZ cassette (i.e., AugαLacZ/+ and AugαLacZ/LacZ mice; SI Appendix, Fig. S1 D and E).

Using Augα-Gfp/+ and Augβ-Gfp/+ mice, we next analyzed the expression of Augα and Augβ primarily within the brain. Visualization of Augα-Gfp/+ mice by immunofluorescence microscopy revealed strong expression in the hypothalamus (Fig. 1A). To determine whether loss of Augα may cause changes in cytoarchitecture and in the localization of neuronal cell populations expressing Augα, we generated and similarly analyzed Augα-Gfp/Gfp–knockout mice. No obvious defects were observed in the overall brain anatomy and in the region containing cells labeled by Augα–GFP (Fig. 1A). The most robust expression of Augα was detected within the ARC, with weaker expression in the paraventricular nucleus (PVN), dorsomedial nucleus (DMH), and suprachiasmatic nucleus (SCN; Fig. 1 A and B).

Next, we determined the cell types expressing Augα within the ARC nucleus. The ARC nucleus contains two major neuronal cell populations that control energy metabolism, including hunger-promoting AgRP neurons and satiety-promoting proopiomelanocortin (POMC) neurons (14, 15). To determine the specific cells within the ARC that express Augα, we immunolabeled postnatal day 60 (P60) coronal sections of Augα-Gfp/+ mice with anti-neuropeptide Y (NPY) antibody, which also labels AgRP neurons (SI Appendix, Fig. S2A), and with an anti-POMC antibody that labels POMC neurons (SI Appendix, Fig. S2B). This experiment showed increased coexpression of NPY-positive neurons with GFP-labeled neurons, suggesting endogenous expression of Augα in AgRP neurons within the ARC. Moreover, published single-cell RNA-sequencing data (16) revealed high expression of Augα (Fam150B) within AgRP neurons among 18 neuronal cell populations (SI Appendix, Fig. S2 C and D). To further establish colocalization of Augα–Gfp within the ARC neuronal populations, we genetically labeled Augα–Gfp-expressing cells within the ARC nucleus using AgRP-Cre; Ai14-tdT mice that label all AgRP-expressing neurons (in red) and POMC-Cre; Ai14-tdT mice that label all POMC-expressing neurons (in red; Fig. 1 C and D). The Augα–Gfp neurons within the ARC showed strong colocalization with the AgRP-expressing neurons and minimal colocalization with POMC neurons. It is noteworthy that we were not able to detect Augβ expression in the brain, which is consistent with open-label RNA-sequencing data (17), which also did not reveal Augβ mRNA expression in the brain (SI Appendix, Fig. S2E). Taken together these experiments show a particularly high expression of Augα in AgRP neurons in the ARC nucleus of the hypothalamus.

AgRP neurons located within the ARC of the hypothalamus are critical for regulating food intake and energy homeostasis (18). In response to fasting, AgRP neurons show increased expression of various orexigenic molecules, including AgRP peptide, which regulates neurons in the PVN (19). To determine whether Augα expression within the ARC nucleus is involved in control of hunger and satiety, we performed a 16-h food starvation study on the Augα-Gfp/+; AgRP-Cre; Ai14-tdT and Augα-Gfp/Gfp; AgRP-Cre; Ai14-tdT animals. We observed that Augα is metabolically stimulated upon fasting within AgRP neurons, while its expression within PVN remains unchanged (Fig. 1 E–G and SI Appendix, Fig. S3). RNA seq (17) data also showed similarly increased expression of Augα in AgRP neurons upon fasting versus refeeding (SI Appendix, Fig. S2E). However, the increased expression of Augα did not affect AgRP neuron activation, as shown by c-Fos immunostaining in Augα-Gfp/Gfp versus Augα-Gfp/+ animals (Fig. 1 H and I), suggesting that Augα may act downstream from AgRP signaling.

To determine whether Augα governs any of the metabolic phenotype during fasting, we performed fasting and refeeding experiments on Augα-knockout and wild-type littermate mice in metabolic cages. Single animals were housed for 4 d in the metabolic cage, and at the end of day four, the mice were food deprived for 16 h and then allowed to feed for the next 24 h. During food deprivation, significantly higher energy expenditure and respiratory exchange rate (RER) of Augα-deficient mice were detected due to increased CO₂ production during the dark phases, indicating a preference for carbohydrate consumption over lipids with activity trending higher (SI Appendix, Fig. S4). A significant increase in water intake was also detected in Augα-deficient mice during the dark phase of the fasting period (SI Appendix, Fig. S4E). However, upon refeeding, a small but significant decrease in both food and water intake was observed in Augα-deficient mice as compared to their wild-type littermates (SI Appendix, Fig. S5). Taken together, the results showed that Augα expression increases upon fasting similar to the expression of other orexigenic genes, including AgRP and Npy. These results link Augα to the metabolic circuit of AgRP-controlled energy homeostasis (20).

To determine the effect of Augα deficiency on animal physiology, we assessed the weight and body composition of the Augα-knockout and littermate control mice. The mice were born according to the Mendelian ratio. We observed that both Augα male and female knockout mice, derived from both Gfp- and LacZ-knockin mouse lines, gained less weight with age than wild-type littermate or heterozygous mice on standard chow diet. Differences in body weights were noticed from the age of 4 wk onward. Moreover, from the age of 14 wk onward, Augα-knockout animals (male and female) were significantly leaner than Augα heterozygous or wild-type animals. There was no significant weight difference between heterozygous and wild-type mice (Fig. 2 A, B, D, and E). The Augα and Augβ double knockout mice and Augα single knockout mice showed similar age-dependent weight differences compared to the double heterozygous mice. In contrast, the Augβ-knockout mice, which carry one Augα allele, showed no weight difference compared to double heterozygous littermates. These results demonstrate that Augα deficiency is critical for thinness, while Augβ deficiency does not cause thinness (Fig. 2C). In addition, Augα-knockout mice gained significantly less weight than littermate controls when the mice were fed a high-fat diet (HFD), suggesting that Augα deficiency provides resistance against HFD-induced weight gain (Fig. 2F). Inspection of the gross morphology of the animals revealed a substantial reduction in the circumference of the animals but no change in their length, indicating that the differences seen in the weights of males and females are not due to change in body length (Fig. 2G). We next analyzed the body composition of adult Augα-knockout mice and their littermate wild-type as well as heterozygous controls by magnetic resonance imaging (MRI). This analysis revealed that Augα-knockout mice have significantly less overall body fat and more lean mass than wild-type and heterozygous littermate mice with respect to their overall body weight (Fig. 2 H–J).

Furthermore, to determine the contributions of each of the fat deposits, fat from subcutaneous white adipose tissue (scWAT), retroperitoneal WAT (rWAT), and gonadal WAT (gWAT) as well as brown adipose tissue (BAT) from Augα-knockout and wild-type animals was isolated. The results presented in Fig. 3A revealed significantly reduced weight of scWAT, rWAT, and gWAT in Augα-knockout mice. Overall gross anatomy revealed reduction in the size of the adipose tissue depots in the Augα-knockout mice, which were further analyzed by hematoxylin and eosin (H&E) staining to assess fat composition and adiposity. Overall, lower percentages of large adipocytes were detected in Augα-knockout mice than in their littermate controls; conversely, a higher percentage of small adipocytes was detected in Augα-knockout mice (Fig. 3 B–E). To gain insight into the mechanism of reduced fat deposition in the adipocytes of Augα-knockout mice, qRT-PCR analysis was carried out for genes involved in the control of thermogenesis and fat oxidation. The experiment presented in Fig. 3 F–H reveals increased expression levels of mitochondrial brown fat uncoupling protein 1 (UCP1), peroxisome proliferator–activated receptor coactivator 1α (PGC1α), cell death activator (CIDEA-A), and β3 adrenergic receptor (β3AR) in WAT depots, suggesting some browning of WAT. To determine a potential link between decreased Augα expression in brain and increased browning of WAT, we next analyzed the levels of norepinephrine (NE) produced in adipose tissue and in the sera of Augα-knockout and control mice. This experiment revealed a strong increase of NE levels in scWAT (Fig. 3I). It was previously demonstrated that hypothalamic PVN neurons regulate NE production within adipocytes through a cascade of neuronal connections, resulting in enhancement of thermogenesis and fat oxidation (21). While our experiments reveal changes in adipose tissue depots in Augα-knockout mice, these results may not represent the sole mechanism underlying the leanness of Augα-knockout mice.

To approach the mechanisms contributing to the lean phenotype of the Augα-knockout mice, we compared 3- to 6-mo–old Augα-knockout mice with littermate control mice fed on a standard diet in metabolic cages for 2 d. We observed that Augα-knockout mice were slightly more active during daytime but had a significantly higher activity during the night period with two separate peaks. Coherent with this, the augmented activity was associated with higher energy expenditure and RER contributed by increased O₂ consumption and CO₂ production (Fig. 4 A–E). However, while during the night period, food and water intake tend to be increased in the lean knockout mice, this did not reach statistical significance (Fig. 4 F and G). We also assessed these animals for glucose tolerance and insulin sensitivity. The experiment presented in Fig. 4 H–I illustrates that Augα deficiency leads to improved glucose tolerance and sensitivity to insulin, consistent with increased energy expenditure and physical activity.

Taken together, the metabolic parameters suggest that the age-dependent thinness of Augα-knockout mice is the likely result of increased physical activity linked to higher energy expenditure and superior glucose clearance. However, longer and more detailed metabolic analysis may reveal the full contribution of altered food intake, possibly with behavioral changes, to the Augα-knockout mice lean phenotype.

To identify the mechanism of neuronal control of Augα-knockout mice thinness, we took a cue from our previous studies where we showed that Augα induces neurite elongation in neuroblastoma cell lines (10). To determine whether Augα can also mediate a similar response in mouse neuronal cultures, we isolated neural stem cells from a wild-type animal and stimulated them in the presence of recombinant mouse Augα protein. This experiment showed a robust increase in neurite formation, as shown in SI Appendix, Fig. S6A by staining with microtubule-associated protein-2. We next examined the possibility of whether Augα deletion may affect the arborization of AgRP neuronal projections to the PVN, a critical region that regulates energy homeostasis downstream of ARC. In this experiment serial sections of the PVN isolated from adult and P15 brains of Augα-Gfp/+; AgRP-Cre; Ai14-tdT and Augα-Gfp/Gfp; AgRP-Cre; Ai14-tdT mice were analyzed for AgRP axon arborization into the PVN. We observed that axonal projections of AgRP neurons onto PVN neurons were significantly reduced in numbers, indicating a disruption in metabolic circuitry (Fig. 5). These results suggest a potential role of Augα in the development of the metabolic circuit.

It was previously demonstrated that Alk is expressed in a subset of PVN neurons and that knockout of Alk leads to thinness (22). Additionally, in situ hybridization data of Alk from Allen Brain Atlas and experiments performed in our laboratory confirmed Alk localization in the PVN (SI Appendix, Fig. S6 B and C). Mechanistically, Augα binding stimulates tyrosine autophosphorylation and Alk activation, leading to stimulation of Alk-dependent intracellular signaling pathways (10). To determine if Augα expression in ARC and/or the PVN influences Alk activation within the PVN, we performed immunolabeling of Alk using antibodies that bind to the Alk extracellular domain as a measure for Alk level and phospho-Alk (pAlk)–selective antibodies as a measure of Alk activation in the brains of Augα-Gfp/+; AgRP-Cre; Ai14-tdT and Augα-Gfp/Gfp; AgRP-Cre; Ai14-tdT mice. This experiment showed that tdTomato-labeled AgRP projections are in close proximity to pAlk-expressing cell bodies, suggesting that Augα expressed in AgRP neurons may mediate activation of Alk within the PVN. Moreover, GFP-labeled Augα neurons are in the vicinity of the pAlk-expressing neurons in the PVN, revealing close proximity of endogenous Augα with Alk, consistent with Augα-induced Alk activation in these neurons. Further analysis of the Augα-Gfp/Gfp; AgRP-Cre; Ai14-tdT brains revealed that the level of pAlk is reduced in Augα-deficient mice (Fig. 5 C and D and SI Appendix, Fig. S6F), while the level of total Alk remains unaffected in the PVN of Augα-deficient mice (SI Appendix, Fig. S6 D and E). These results provide a mechanistic link to an earlier finding demonstrating that deletion of Alk from the PVN results in thinning of mice (22). We propose that Augα expressed in AgRP and/or the PVN stimulates Alk activation in the PVN, an area within the hypothalamus implicated in control of energy metabolism (22).

Experiments presented in this manuscript show that the Augα-activated Alk signaling pathway may operate within the hypothalamus, ARC, and PVN neurons, a region in which signaling via MC4 receptors (MC4Rs) controls an important metabolic process that regulates food intake. It was shown that binding of AgRP or melanocyte-stimulating hormone (α-MSH) to MC4R in the PVN promotes hunger or satiety, respectively, and that aberrant activation of this pathway results in severe cases of obesity (23–25). To determine whether the Augα–Alk signaling pathway is linked to or part of the α-MSH–MC4R signaling pathway, we analyzed single-cell RNA-sequencing data from the PVN to identify neurons that express Alk. Interestingly, the data presented in Fig. 6 A–D show that Alk is not expressed in MC4R-expressing neurons, which otherwise coexpress thyrotropin-releasing hormone (TRH). Alk, on the other hand, is expressed in corticosterone-releasing hormone (CRH)–positive neurons. This result was confirmed by immunostaining analysis (Fig. 6E). These experiments show that MC4R and Alk are expressed in discrete populations of cells within the PVN, demonstrating that two distinct mechanisms that are also responsible for pigmentation operate within the PVN to control body weight.