Increased Energy Expenditure and Protection From Diet-Induced Obesity in Mice Lacking the cGMP-Specific Phosphodiesterase PDE9

The lack of effective therapies for the obesity epidemic is a serious challenge, as recent studies report that ∼40% of Americans are clinically obese, with the prevalence of obesity rising worldwide (1). Obesity is comorbid with insulin resistance and type 2 diabetes along with other chronic illnesses, such as cardiovascular diseases, nonalcoholic fatty liver disease, asthma, and certain cancers (2). Thus, finding ways to treat or reduce obesity are important for decreasing diabetes and related diseases. At its simplest, obesity can be managed by either reducing the calories consumed or by increasing energy expenditure. Changes in diet and exercise are the most commonly recommended methods to treat obesity, but rarely succeed over the long-term (3). Much effort has been directed toward reducing caloric intake, including pharmacological approaches such as appetite suppressants and blockade of fat absorption (4). These are frequently accompanied by unacceptable side effects or eventual lack of efficacy. Invasive procedures such as bariatric surgery have also been prescribed (4), and while to date, this approach has proven effective at ameliorating insulin resistance and providing sustained weight loss, the risks associated with such radical surgeries should not be underestimated (5). Alternatively, one potential means to increase energy expenditure aside from physical exercise is energy-consuming futile metabolic cycles and uncoupled respiration (6).

Brown adipose tissue (BAT) in mammals is a mechanism for maintaining body temperature. Brown adipocytes can regulate uncoupled respiration via oxidative metabolism that generates heat without ATP synthesis. These adipocytes contain a unique mitochondrial protein called uncoupling protein 1 (UCP1). When activated, UCP1 is a gated pore that allows H⁺ to pass through the inner mitochondrial membrane, thereby uncoupling oxidative phosphorylation from ATP production (7). While UCP1 is the signature protein of BAT, it can also be expressed in cells of white adipose tissue (WAT) in response to stimuli that increase cyclic nucleotides (i.e., cAMP and cGMP). These WAT adipocytes that have a brownlike phenotype are sometimes called “beige” adipocytes (8,9). During the last decade, there has been growing appreciation that adult humans possess significant amounts of brown and beige adipocytes that are rich in mitochondria and UCP1 (10–13). Thus, activation of brown/beige adipocyte thermogenesis to increase net energy expenditure might be an attractive therapeutic target for obesity and metabolic disease.

Increasing the concentration of cyclic nucleotides cAMP and cGMP in adipocytes is associated with an increase in UCP1 expression and thermogenic activity (14–17). For cAMP, this is classically observed in response to cold temperature, leading to increased secretion of norepinephrine from sympathetic nerves that activate the adipocyte’s β-adrenergic receptors (AR) (18). In addition to the important physiological role of the sympathetic nervous system and catecholamines, we and other have shown that the cardiac natriuretic peptides (NP), atrial NP (ANP) and B-type NP (BNP), are also capable of increasing thermogenic activity of adipocytes via increasing intracellular cGMP and thereby exert antiobesity effects (19–25). Levels of cyclic nucleotides are controlled not only by their synthesis, but also their degradation. The phosphodiesterase (PDE) enzymes break the phosphodiester bond in cyclic nucleotides, rendering them inactive (26). Based on our new findings that cold exposure decreases the expression of cGMP-specific PDE9, one of the most selective PDEs to degrade cGMP over cAMP (27), we asked whether decreasing the degradation of cGMP would have similar prothermogenic and antiobesity effects. The safe clinical utility for pharmacological inhibition of several PDEs has been demonstrated for a number of diseases, and the inhibition of multiple PDEs has been associated with increased adipocyte browning (reviewed in Ref. 17). Several small-molecule PDE9 inhibitors have entered clinical trials for conditions such as Alzheimer disease, schizophrenia, and sickle cell disease in which their safety has been established (28–34).

We hypothesized that the absence of PDE9 would counteract diet-induced weight gain and improve glucose handling via increasing uncoupled energy expenditure. We used a global Pde9a knockout mouse and show that these Pde9a^−/− mice resist diet-induced weight gain and have concomitant improvements associated with reduced adiposity. Changes in energy expenditure associated with Pde9a knockout at any given time were small; however, over the extended course of the study, these modest changes became significant, leading to a reduced body weight and adiposity. As most people become obese slowly, gaining ∼0.5–1.0 kg/year (35,36), we posit that inhibition of PDE9 may be useful for counteracting this slow weight gain over time.

See Supplementary Table 1 for more information.

Animal studies were conducted at Sanford Burnham Prebys Medical Discovery Institute (SBP), Vanderbilt University Medical Center (VUMC), and Johns Hopkins University and School of Medicine (JHSOM). Pde9a^−/− mice on a C57BL/6J background were previously described (37). Mice were maintained on a 12-h light/dark cycle with two to five animals per cage and ad libitum food and water. At 6 weeks of age, SBP and VUMC mice received nutrient-matched Surwit diets: a control low-fat diet (LFD) (10.5% calories from fat: D12328; Research Diets) or high-fat diet (HFD) (58.0% calories from fat: D12330; Research Diets). JHSOM mice were fed HFD (60.0% calories from fat: D12492; Research Diets). Mice were weighed twice weekly, and body composition was measured using a Minispec Body Composition Analyzer (Bruker) at SBP and Vanderbilt Mouse Metabolic Phenotyping Center (VMMPC) or EchoMRI-100 (Echo Medical Systems) at JHSOM. Mice were fasted for 5-h prior to euthanasia by CO₂ asphyxiation and exsanguination via cardiac puncture. All procedures were approved by the Institutional Animal Care and Use Committees at SBP, VUMC, and JHSOM.

Chow-fed 13–16-week-old male mice were housed at 30°C for 2–4 days prior to habitation at 6°C for the indicated times. Additional mice received intraperitoneal CL-316,243 (1 mg/kg body weight/day) for 1 week as in Liu et al. (38).

IngJ6 and Bat8 immortalized mouse cell lines (8) and human multipotent adipose-derived stem cells (hMADS) (39) were cultured and differentiated as described.

RNA was extracted using TRIzol and the aqueous phase purified on Zymo-Spin IIICG Columns. Instruments used were the Roche LightCycler 480 II or Applied Biosystems QuantStudio 6 Flex System. Three replicates of each sample were averaged, normalized to mRplp0 (36B4), and analyzed using the Pfaffl method. Primer sequences are in Supplementary Table 2.

Western blotting was as described (24). Image acquisition was on Bio-Rad Digital ChemiDoc MP with IR or Typhoon FLA9000 Variable Mode Imager.

Tissues were fixed in 10% formaldehyde (4°C overnight) and stored in 70% ethanol (4°C). Histology was performed by the Vanderbilt Translational Pathology Shared Resource. Slides were imaged at ×20 original magnification with a Leica SCN400 Slide Scanner in the Digital Histology Shared Resource.

Lipids were extracted from frozen liver and quantified by the VMMPC Lipid Core.

For both the glucose tolerance test (GTT) and insulin tolerance test (ITT), mice were fasted for 5-h during the light cycle. Intraperitoneal injections of 1.0 g/kg dextrose in 0.9% saline for GTT or 0.5 units/kg insulin in 0.9% saline for ITT. Glucose was measured from tail-vein blood with a Bayer CONTOUR glucometer and glucose test strips. Clamp studies were done in chronically catheterized conscious mice by the VMMPC.

Prior to euthanasia, mice were fasted for 5 h, and glucose was measured from tail-vein blood. Mice were euthanized, and plasma was obtained from blood via cardiac puncture into tubes containing 8 µL of 0.5 mol/L EDTA. Insulin was measured from plasma stored at −80°C by ELISA. ALT, triglyceride, and cholesterol quantification was performed on plasma stored at −20°C by the Vanderbilt Translational Pathology Shared Resource.

Mice used for energy balance studies were maintained at VUMC and JHSOM and fed HFD for 15 and 26 weeks, respectively. Indirect calorimetry and measurements of food intake and physical activity were performed on mice in the Promethion System (Sable Systems International) by the VMMPC or the Comprehensive Lab Animal Monitoring System (CLAMS) by the Rodent Metabolism Core at JHSOM. Light and dark cycles were analyzed independently.

High-resolution respirometry was performed using Oxygraph-2k (Oroboros Instruments) in the VMMPC. Interscapular BAT (iBAT) from age-matched chow-fed mice was dissected into 1.6–2.5-mg pieces and permeabilized by sequential incubation in biopsy preservation solution (BIOPS), BIOPS plus saponin, and MiR05, each for 20 min at 4°C. Assays were performed in MiR05 at 37°C, and substrates were added in the order indicated. Data were normalized to wet tissue weight.

Data are mean ± SEM using Prism version 9.1.0 for Windows 64-bit (GraphPad Software). Unless otherwise stated, analyses were performed using two-way ANOVA. Post hoc analyses of ANOVA used Sidak multiple comparisons test for Pde9a genotype only and indicated on figures with the symbols: *Pde9a^+/+ versus Pde9a^−/−, †Pde9a^+/+ versus Pde9a^+/−, and ‡Pde9a^+/− versus Pde9a^−/−; and one symbol, P < 0.05; two symbols, P < 0.01; or three symbols, P < 0.001. For analyses of body mass and blood glucose over time, two-way ANOVA with repeated measures was performed comparing the factors Pde9a genotype and time, analyzing diet groups independently. Average energy expenditure, VO₂, and respiratory quotient for light and dark cycles were analyzed using the least squares method of multiple linear regression with model including the continuous effect of body mass and the categorical effects of diet, genotype, and instrument (i.e., Promethion at VUMC or CLAMS at JHSOM).

Data/resources are available upon request from the corresponding author.

We and others previously established that the cGMP–cGMP-dependent protein kinase (PKG) signaling axis in adipocytes can increase brown and beige adipocyte activity, including mitochondrial biogenesis, UCP1 expression, uncoupled respiration, and energy expenditure (19–25,40–42). In addition, the dynamic control of the ratio between the guanylyl cyclase-containing NP receptor, NPRA, to the clearance receptor, NPRC, is an important regulator of NP signaling (reviewed in Ref. 17), and we previously showed that cold temperature exposure reduces expression of NPRC (20). Given that PDE5 and PDE9 are selective for degrading cGMP, we sought to determine if they also played a role in the brown fat thermogenic program. Pde5a and Pde9a expression were measured in iBAT following both short-term and long-term cold exposure. We found that Pde9a, but not Pde5a, gene expression was significantly suppressed by cold exposure at both time points (Fig. 1A–D). A similar suppression of Pde9a, but not Pde5a, expression in iBAT was found in mice treated with the β₃-AR agonist CL-316,243 (Fig. 1E and F). In an immortalized brown adipocyte cell line (Bat8), Pde9a expression was reduced in response to the β-AR agonist isoproterenol in a comparable manner (Fig. 1G and H). These findings indicate that cold-induced β-AR signaling suppresses Pde9a, but not Pde5a, gene expression.

As these findings suggest that a reduction in Pde9a may be important for cold-mediated induction of the brown fat thermogenic program, we asked whether global loss of PDE9 would affect AT browning. Male Pde9a^+/+ and Pde9a^−/− mice were exposed to cold temperature for 3 days to cause a mild induction of the thermogenic program. Pde9a^−/− mice demonstrated a significant increase in the amount of UCP1⁺ adipocytes in their inguinal WAT (iWAT), while both iBAT and epididymal WAT (eWAT) appeared to have a reduction in adipocyte size (Fig. 1I–K). Together, these findings suggest that PDE9 is a negative regulator of the cold-mediated BAT thermogenic program.

We asked whether blocking PDE9 activity in adipocytes in vitro using small-molecule inhibitors would increase PKG signaling and the adipocyte thermogenic program. We used mouse cell lines representative of white (IngJ6) and brown (Bat8) adipocytes (8), hMADS (43), and two different PDE9 inhibitors: BAY73–6691 (BAY) and PF-04447943 (PF). As shown in Fig. 2A and B, phosphorylation of vasodilator-stimulated phosphoprotein (VASP) at serine 239, a marker of PKG signaling (44), was increased in both IngJ6 and Bat8 cells treated with BAY above the level seen with ANP alone. BAY also slightly increased VASP phosphorylation in IngJ6 independent of ANP (Fig. 2A). The PF compound was also tested in Bat8 cells, where it increased ANP-evoked VASP^(S239) phosphorylation in a dose-dependent manner (Fig. 2C). The hMADS cell line can express thermogenic markers and increase mitochondrial biogenesis after differentiation in response to catecholamines as well as cardiac NPs (20,39,43). We stimulated cGMP-PKG signaling by including the NPs, ANP and BNP, and/or PF during the last 7 days of differentiation. As shown in Fig. 2D, both the NPs and PF were able to increase UCP1 expression. Also shown is that the cell-permeable cGMP analog pCPT-cGMP raised UCP1 expression (Fig. 2D).

To test our hypothesis that PDE9 ablation would improve diet-induced obesity and glucose handling, male Pde9a^+/+, Pde9a^+/−, and Pde9a^−/− mice were fed nutrient-matched control LFD or HFD for 16 weeks beginning at 6 weeks of age. Overall, Pde9a genotype affected body weight and body weight gain throughout the feeding study (Fig. 3A–D and Supplementary Fig. 1). The reduced body weight of HFD-fed Pde9a^−/− was attributed to a reduction in fat mass (Fig. 3E and G). Of note, Pde9a^+/− mice exhibited an intermediate obesity phenotype when challenged with HFD. These data demonstrate that even haploinsufficiency of Pde9a confers resistance to HFD. Female Pde9a^−/− mice also gained less weight than Pde9a^+/+ throughout the 16 weeks of HFD feeding (Supplementary Fig. 2). Compared with males, the differences in body weight were not as pronounced, but were also a result of reduced fat mass.

Intraperitoneal GTT (IP-GTT) and IP-ITT were used to assess glucose handling and insulin sensitivity in male mice. In mice fed LFD, Pde9a genotype did not affect glucose handling. For HFD-fed mice, glucose tolerance was essentially identical between Pde9a^+/+ and Pde9a^+/− mice, while Pde9a^−/− mice exhibited a modest improvement (Fig. 4A and B and Supplementary Fig. 3A). ITT revealed very little difference between the Pde9a genotypes (Fig. 4C and Supplementary Fig. 3B and C). Fasting glucose and insulin levels were not significantly different (Fig. 4D and E). These findings are consistent with other observations in this study that modest increases in “adipose browning” and energy expenditure are sufficient to mitigate the more overt features of insulin resistance. In female mice on HFD, absence of Pde9a was also associated with a modest improvement in IP-GTT and IP-ITT (Supplementary Fig. 4A–F). Fasting glucose and insulin levels were not different (Supplementary Fig. 4G and H). Hyperinsulinemic-euglycemic clamps were performed in chronically catheterized conscious mice from another cohort of Pde9a^+/+ and Pde9a^−/− male mice that were fed HFD for 16 weeks. The body weights of these mice were closely matched, thus controlling for the previous weight difference. In these mice, all measurements were largely the same between the Pde9a^+/+ and Pde9a^−/− mice (Supplementary Fig. 3D–F). Pde9a^−/− mice showed an increase in fasting glucose turnover rate; however, the rates of glucose uptake and endogenous glucose production during the hyperinsulinemic clamp were similar (Supplementary Fig. 3G).

To assess insulin signaling in AT, phosphorylation of AKT at Ser⁴⁷³ was measured postmortem in eWAT, iWAT, and iBAT (Fig. 4F). Three representative mice were chosen from each genotype × diet treatment group based upon their closeness to the mean terminal body weight of their group. Loss of Pde9a was associated with increased AKT phosphorylation and therefore insulin signaling in AT regardless of dietary treatment (Fig. 4F).

The reduction in body weight and adiposity in Pde9a^−/− male mice was associated with reduced liver mass. Consistent with the observation that overall fat mass was lower in Pde9a^−/− mice on HFD, the weights of individual adipose depots that were collected exhibited a similar trend (Fig. 5A–C). This also corresponded to a reduction in adipocyte size (Fig. 5D–F). On HFD, the iBAT of Pde9a^+/+ mice had large unilocular adipocytes, whereas Pde9a^−/− had smaller multilocular adipocytes (Fig. 5F). In female mice on HFD, tissue weights exhibited a similar trend toward reduced mass in the Pde9a^−/−, which was particularly evident for iWAT and iBAT (Supplementary Fig. 5). In the liver, absence of Pde9a led to a greater protection during HFD feeding. Hematoxylin and eosin (H-E) staining of the livers revealed prominent steatosis in the Pde9a^+/+ HFD-fed mice, which was significantly reduced in the HFD fed Pde9a^−/− livers (Fig. 5G). This corresponded to a significant reduction in liver weight in Pde9a^−/− mice (Fig. 5H). Once again, Pde9a^+/− mice showed an intermediate reduction in liver weight. As expected, we observed a strong trend toward reduced hepatic triglyceride levels in the HFD-fed Pde9a^−/− mice (Fig. 5I), with minor differences in fatty acid composition between genotypes (Supplementary Table 3). Hepatic steatosis frequently leads to liver damage. Circulating levels of plasma ALT, a biomarker of hepatocyte membrane damage, were measured. In the Pde9a^−/− mice, ALT levels were significantly reduced (Fig. 5J). Plasma triglyceride levels were unchanged and cholesterol showed a trend toward being reduced in HFD-fed Pde9a^−/− mice (Supplementary Fig. 6). Absence of Pde9a was also associated with a significant reduction in heart and spleen weight and a trend toward reduced kidney weight with HFD feeding (Fig. 5K–L).

To determine whether there were differences in energy expenditure that would account for the differences in body weights and fat mass in HFD-fed Pde9a^−/− mice, indirect calorimetry was measured on male mice at the end of the diet period. As shown in Supplementary Fig. 7, food intake and physical activity cannot account for the decreased weight and fat mass in the Pde9a^−/− mice. Importantly, we found significantly increased energy expenditure (P = 0.047) and a trend toward increased oxygen consumption (P = 0.057) in Pde9a^−/− mice during the dark cycle (Fig. 6A and B). Respiratory quotient was not changed (Fig. 6C).

We hypothesized that the increased energy expenditure in Pde9a^−/− mice was due to increased PKG signaling and thermogenesis in the AT. Significant changes in resting VASP^(S239) phosphorylation were not observed in iBAT or iWAT (Supplementary Fig. 8). It is likely that differences in unstimulated VASP^(S239) phosphorylation are difficult to observe in vivo due to the continuous interplay of basal stimuli and dephosphorylation events. However, as shown in Fig. 7, a signature group of genes involved in energy expenditure was increased in the ATs of male Pde9a^−/− mice. Compared with Pde9a^+/+ mice, in Pde9a^−/− mice fed LFD, these genes were increased in iWAT, while in HFD-fed mice, these genes were increased in iBAT. Once again, in the Pde9a^+/− mice, expression of each of these genes was intermediate between the Pde9a^+/+ and Pde9a^−/−, which is consistent with a gene dosage effect of Pde9a on body weight and adiposity (Fig. 7F and L). Pde9a expression in Pde9a^+/+ mice was compared across dietary treatments, and HFD did not affect its expression in either iBAT or iWAT (Fig. 7F and L and Supplementary Fig. 9F and L). In female iBAT, there were no significant differences in this panel of thermogenic genes between genotypes (Supplementary Fig. 9).

To determine if these changes in thermogenic gene expression altered the function of the iBAT, we performed high-resolution respirometry of iBAT. Pde9a^−/− iBAT from both male and female mice demonstrated greater oxygen consumption than Pde9a^+/+ when the respiratory substrates were added (Fig. 8A and B). This indicates that loss of PDE9 increases the respiratory capacity of the iBAT, even though changes in the expression levels of the components of thermogenesis were not vastly different between genotypes.

Analogous to sympathetic nervous system–induced adipocyte browning using the cAMP–cAMP-dependent protein kinase pathway, NP-evoked browning uses cGMP-PKG–mediated signaling (15–17). Stimulation of this pathway has been shown to improve metabolic dysfunction by increasing BAT thermogenesis and browning of WAT (19–25,40–42). Additionally, we have shown that removal of the NP clearance receptor, NPRC, in adipocytes augments cGMP-PKG signaling, leading to increased thermogenic energy expenditure, which consequently reduces obesity and improves glucose handling (20,45). In the present studies, we observed that cold exposure or β₃-AR agonist treatment led to significant decreases in the expression of PDE9, which is highly selective for cGMP (27,28), in brown adipocytes. This observation is like our earlier finding that cold exposure also suppresses the expression of NPRC in AT (20). Therefore, we asked whether genetic deletion of PDE9 would have a similar effect to deletion of NPRC. Interestingly, PDE9 has been suggested to preferentially degrade NP-evoked cGMP (27,37,46). PDE enzymes are already known to play an important role in adipocytes, and nonselective PDE inhibitors have long been known to increase adipocyte lipolysis and thermogenesis (reviewed in Ref. 17). Furthermore, sildenafil, which inhibits the other cGMP-specific PDE expressed in adipocytes, PDE5, has been shown to improve glucose uptake in skeletal muscle and increase energy expenditure (47). While sildenafil did not affect BAT UCP1 expression (47), in a later study, it was reported to induce browning of iWAT (48). Though PDE9 has been detected in human AT (49), its role in adipose browning and energy balance had not been previously studied.

In this study, we found that inhibition of PDE9 increased PKG signaling and UCP1 expression in adipocytes (Fig. 8C). Global gene deletion of Pde9a resulted in the mice being resistant to HFD-induced obesity due to increased iBAT thermogenic gene expression and an associated increase in energy expenditure. There were no differences in physical activity or food intake (Pde9a^−/− mice actually tended to eat slightly more food than Pde9a^+/+). This reduction in adiposity ameliorated metabolic dysfunction in several ways. Most noticeably, HFD-induced hepatic steatosis and liver damage was greatly reduced in the Pde9a^−/− mice. The larger caloric deficit due to energy expenditure in Pde9a^−/− mice primarily reduced the hepatic lipid content, with WAT depots somewhat less affected. Thus, in Pde9a^−/− mice, lipids were appropriately stored in the adipose instead of elsewhere ectopically. Additionally, glucose handling was modestly improved in the Pde9a^−/− compared with Pde9a^+/+ mice when challenged with HFD. Upon close evaluation of this glucose handling phenotype, we found that these improvements in Pde9a^−/− mice were not due to increased glucose uptake or elevated insulin secretion, but rather were due to the reduced adiposity. Nevertheless, as obesity and increased adiposity are closely associated with impaired insulin sensitivity and poorer glucose handling, PDE9 inhibition has potential as an anti–type 2 diabetes therapeutic. While our findings seem most consistent with a major role for PDE9 in AT, only an adipocyte-specific deletion of Pde9a will unequivocally establish its role in this cell type. In contrast, since small-molecule PDE9 inhibitors that are under development and in clinical trials will affect the enzyme wherever it is expressed, the whole-body Pde9a^−/− mouse serves as preclinical evidence that these inhibitors could have beneficial consequences for metabolic disease.

An unexpected finding of these studies was that Pde9a^+/− mice displayed an intermediate phenotype. This would suggest a gene dosage effect of Pde9a, and thus, even partial inhibition of PDE9 may be able to augment energy expenditure and improve metabolic disease. Furthermore, this supports the robustness of PDE9 inhibition, as even partial removal of Pde9a results in a noticeable improvement in susceptibility to HFD.

PDE9 inhibitors have potential to be useful therapeutics for reducing weight gain and thereby ameliorating associated comorbidities, such as type 2 diabetes. Most importantly, current studies indicate they are safe and well tolerated in humans (28–34). Moreover, analogous to what we show in this study with Pde9a^−/− mice, our collaborators have found that the PF PDE9 inhibitor reduces body weight by increasing energy expenditure and thereby improves glucose handling in a model of cardiometabolic disease (50). Comparable to our findings, the results of Mishra et al. (50) show that PDE9 inhibition is associated with increased respiration and expression of thermogenic genes including Ucp1 (50). Together, these observations suggest that PDE9 inhibitors can increase thermogenic gene expression in adipocytes via increasing PKG signaling, and these therapeutic implications may be translatable to humans.

Our studies show that loss of Pde9a increases iBAT respiratory capacity, modestly increased global energy expenditure, and resulted in an associated decrease in weight gain. At the end of the HFD, this culminated in a large reduction in weight gain and significant metabolic improvements. As most people become obese slowly, a therapeutic approach that blunts weight gain over extended time may be preferable to one that causes rapid reductions in body weight, especially if that approach loses effectiveness over time. Together, these studies suggest that PDE9 inhibition may be a useful approach for augmenting adipose thermogenesis to combat weight gain and improve metabolic health.

Acknowledgments. The authors thank Bruce Spiegelman (Dana-Farber Cancer Institute, Harvard Medical School) for the IngJ6 and Bat8 cells and Ez-Zoubir Amri (Institut de Biologie Valrose, Université Côte d’Azur) for the hMADS.

Funding. This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases grant R01 DK103056 (to S.C.), National Heart, Lung, and Blood Institute R35-HL135827 (to D.A.K.), and American Heart Association grant 16SFRN28620000 (to D.A.K. and S.C.). R.P.C. was supported by National Institute of Diabetes and Digestive and Kidney Diseases grant F32 DK116520. F.S. was supported by American Diabetes Association grant 1-18-PDF-110. The Vanderbilt Mouse Metabolic Phenotyping Center is supported in part by National Institute of Diabetes and Digestive and Kidney Diseases grant DK059637 and National Institutes of Health grants S10RR028101 and S10OD025199. The Vanderbilt Translational Pathology Shared Resource is supported by National Cancer Institute/NIH Cancer Center support grants 2P30 CA068485-14, 5P30 CA68485-19, 5U24DK059637-13, and S10 OD023475-01A1. The Lipid Core is supported by National Institute of Diabetes and Digestive and Kidney Diseases grant DK020593.

Duality of Interest. D.A.K. is a co-inventor on a patent filed by Johns Hopkins University regarding uses of PDE9 inhibitors for the treatment of cardiometabolic disorders and obesity. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. R.P.C., D.A.K., and S.C. conceived and designed the study. R.P.C., D.L., F.S., M.K.C., and S.M. performed experiments. R.P.C. analyzed the data and wrote the manuscript. D.A.K. and S.C. reviewed and edited the manuscript. S.C. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. This study was presented at EMBO Workshop: Organ crosstalk in energy balance and metabolic disease, Cádiz, Spain, 8–11 April 2019.