Arg2) Is Sufficient to Convey the Therapeutic Metabolic Effects of Fasting Yiming Zhang

Washington University School of Medicine Digital Commons@Becker

Open Access Publications

2019 Hepatic arginase 2 (Arg2) is sufficient to convey the therapeutic metabolic effects of fasting Yiming Zhang

Cassandra B. Higgins

Hannah M. Fortune

Phillip Chen

Alicyn I. Stothard

See next page for additional authors

Follow this and additional works at: https://digitalcommons.wustl.edu/open_access_pubs Authors Yiming Zhang, Cassandra B. Higgins, Hannah M. Fortune, Phillip Chen, Alicyn I. Stothard, Allyson L. Mayer, Benjamin M. Swarts, and Brian J. DeBosch ARTICLE

https://doi.org/10.1038/s41467-019-09642-8 OPEN Hepatic arginase 2 (Arg2) is sufﬁcient to convey the therapeutic metabolic effects of fasting

Yiming Zhang1, Cassandra B. Higgins1, Hannah M. Fortune1, Phillip Chen1, Alicyn I. Stothard2, Allyson L. Mayer 1, Benjamin M. Swarts2 & Brian J. DeBosch1,3

Caloric restriction and intermittent fasting are emerging therapeutic strategies against obesity, insulin resistance and their complications. However, the effectors that drive this

1234567890():,; response are not completely defined. Here we identify arginase 2 (Arg2) as a fasting-induced hepatocyte factor that protects against hepatic and peripheral fat accumulation, hepatic inflammatory responses, and insulin and glucose intolerance in obese murine models. Arg2 is upregulated in fasting conditions and upon treatment with the hepatocyte glucose transporter inhibitor trehalose. Hepatocyte-specific Arg2 overexpression enhances basal thermogenesis, and protects from weight gain, insulin resistance, glucose intolerance, hepatic steatosis and hepatic inflammation in diabetic mouse models. Arg2 suppresses expression of the regulator of G-protein signalling (RGS) 16, and genetic RGS16 reconstitution reverses the effects of Arg2 overexpression. We conclude that hepatocyte Arg2 is a critical effector of the hepatic glucose fasting response and define a therapeutic target to mitigate the complications of obesity and non-alcoholic fatty liver disease.

1 Department of Pediatrics, Washington University School of Medicine, St. Louis, MO 63110, USA. 2 Department of Chemistry & Biochemistry, Central Michigan University, Mt. Pleasant, MI 48859, USA. 3 Department of Cell Biology & Physiology, Washington University School of Medicine, St. Louis, MO 63110, USA. Correspondence and requests for materials should be addressed to B.J.D. (email: [email protected])

NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8

besigenic cardiometabolic disease arises from a con- that the trehalose class of disaccharide GLUT inhibitors activates stellation of environmental, genetic and meta- and epi- this pathway. More broadly, this study identifies a seminal link O 1 genetic factors . Clinically, intensive lifestyle between hepatic Arg2 and G-protein signaling in the pathogenesis management remains the primary means of treating obesity and of metabolic disease. its comorbidities, including insulin resistance, non-alcoholic fatty liver disease (NAFLD), and thermic depression. Intensive lifestyle management encompasses a wide range of interventions, such as Results increased locomotion, selective macronutrient elimination (e.g., Arg2 is induced during fasting and pseudo-fasting. We pre- ketogenic diets), intermittent fasting (IF), and caloric restriction viously demonstrated that glucose transporter inhibition by the (CR)2. Although lifestyle measures effectively treat obesity, hepatic GLUT inhibitor, trehalose, enhances peripheral thermo- insulin resistance and NAFLD2–8, caloric restriction, fasting and genesis, and mitigates hepatic steatosis by inducing hepatocyte extreme dietary alterations are often unsustainable, unpalatable, fasting-like pathways24,26,29. RNA-seq analysis of trehalose- and are in some cases associated with clinical complications9–11. treated primary hepatocytes identified hepatocyte arginase 2 At the nexus of portal and systemic circulations, the hepatocyte (Arg2) to be among the most highly upregulated genes relative to coordinates peripheral responses to changes in nutrient flux. untreated hepatocytes. We confirmed these data by quantitative During physiological fasting, the hepatocyte provides glucose and real-time RT-PCR (qPCR) and immunoblot analysis, demon- ketones as fuel for fasting peripheral organs. In addition, strating rapid, dose-dependent, cell autonomous Arg2 induction hepatocyte-derived endocrine signals, such as fibroblast growth within 24 h post trehalose (1 mM, 10 mM, and 100 mM) treat- factor 21 (FGF21), enhance peripheral insulin sensitivity and ment (Fig. 1a). This Arg2 response was energy deficit-dependent, promote basal thermogenesis. These adaptations ostensibly as indicated by pre-treating cells with pyruvate as an energy maintain thermic equilibrium during prolonged fasting and substrate to bypass the GLUTs targeted by 100 mM trehalose. hibernation, and maximize post-fasting nutrient absorption12–16. Pyruvate significantly reduced the trehalose-induced Arg2 The physiological fasting response is itself induced by several key response to trehalose (Fig. 1b). Because trehalose catabolism by transcriptional regulators that are also activated during fasting: host (brush border, hepatocyte, or renal) or microbiotic trehalases peroxisome proliferator activated receptor α (PPARα), PPARγ can theoretically reduce the bioavailability of orally administered coactivator 1α (PGC1α), SIRT117,18, and transcription factor EB trehalose therapy, we compared the potency of native trehalose (TFEB)19–23. However, surprisingly little is known regarding with that of a trehalase-resistant analog, lactotrehalose32–34,to hepatocyte-intrinsic factors that regulate hepatic and extrahepatic induce Arg2 expression in vitro (Fig. 1c). Lactotrehalose-induced responses to fasting. Arg2 expression indeed was significantly enhanced at 10 mM and We recently demonstrated that hepatic glucose transport 100 mM when compared with native trehalose at equivalent doses blockade is sufficient to confer or augment the adaptive metabolic (Fig. 1c). Trehalose similarly induced Arg2 mRNA and protein in changes associated with generalized caloric restriction and livers from mice fed 3% trehalose water (5 days) ad libitum intermittent fasting24–29. Germline deletion of the hepatic glucose in vivo (Fig. 1d, and Supplementary Fig. 6). The similarities transporter (GLUT) 8 increased hepatocyte fatty acid oxidation between trehalose-induced pseudo-fasting responses in the liver and decreased fructose-induced de novo lipogenesis in isolated and physiological macronutrient withdrawal prompted us to hepatocytes25. In vivo, targeted GLUT8 disruption increased quantify the Arg2 response in livers from mice fasting up to 24 h. basal thermogenesis25,27,28, protected from diet-induced NAFLD qPCR and immunoblot analysis identified Arg2 to be activated and insulin resistance, and augmented canonical hepatic fasting during physiological fasting (Fig. 1e, and Supplementary Fig. 6). responses (e.g., PPARα and FGF2125,27,28). Similarly, pharma- Arg2 upregulation during fasting is cell-intrinsic, as determined cological hepatic GLUT inhibition by the GLUT inhibitor, tre- by subjecting isolated primary murine hepatocytes to serum and halose, activated hepatocyte fasting responses—autophagic flux glucose deprivation (24 h), after which Arg2 mRNA expression and FGF21 secretion—that was dependent on hepatocyte was significantly upregulated (Fig. 1f). TFEB24,30,31. The hepatic response to pharmacological hepatic We demonstrated that hepatocyte glucose transport blockade GLUT inhibition is to activate thermogenesis, and reduce hepatic attenuates insulin resistance and hepatic steatosis in multiple steatosis and insulin resistance24,26,30,32. The full mechanisms obese models24,25,27–29. To test the hypothesis that hepatocyte driving the hepatocyte glucose fasting response and the utility of Arg2 activation downstream of the fasting response is sufficient these responses against cardiometabolic disease, however, are to drive these therapeutic effects, we examined the cell-intrinsic only now being elucidated. consequences of forced Arg2 overexpression on hepatocyte Here, we identify the arginine metabolizing enzyme, arginase 2 triglyceride accumulation, inflammatory cytokine and chemo- (Arg2) as a hepatocyte glucose fasting-induced factor that is kine activation, and insulin signal transduction. BSA-conjugated sufficient to convey the adaptive hepatic and peripheral metabolic fatty acid treatment (FFA) both without and with 1 nM changes induced during CR and IF. Hepatocyte Arg2 mRNA and lipopolysaccharide (LPS) induced TG accumulation and inflam- protein expression is induced during acute and prolonged fasting, matory gene markers interleukin 1β (Il-1β), interleukin 6 (Il-6), and after treatment with the hepatic GLUT inhibitor, trehalose. tumor necrosis factor α (Tnf-α), monocyte chemoattractant Hepatocyte-specific Arg2 overexpression increases basal ther- protein 1 (Mcp1; also known as Ccl2), and CXC chemokine mogenesis, enhances insulin sensitivity, and decreases hepatic ligand 9 (Cxcl9) in wild-type primary hepatocyte cultures steatosis and inflammation in genetic and diet-induced models of (Fig. 1g). In each condition, fat accumulation and cytokine/ diabetes and NAFLD. Unbiased transcriptomic analysis identifies chemokine expression was significantly attenuated by Arg2 the regulator of G-protein-coupled receptor signaling (RGS) 16 to overexpression (Fig. 1g). In parallel, we verified expression of be among the most highly suppressed hepatic genes in Arg2- functional protein by quantifying urea production in vitro, which overexpressing db/db diabetic mice. Moreover, hepatocyte- revealed an eightfold increase in urea production upon Arg2 specific genetic RGS16 reconstitution in the context of Arg2 overexpression (not shown). To define the effect of Arg2 overexpression reverses the hepatic and peripheral therapeutic overexpression on hepatocyte insulin sensitivity in vitro, primary effects of Arg2 overexpression. We conclude that Arg2 represents hepatocytes were transfected with β-galactosidase or Arg2, an effector of the hepatocyte-intrinsic fasting response that serum-deprived overnight, then stimulated with 10 nM recom- modulates hepatic and extrahepatic metabolic homeostasis, and binant insulin. Arg2-overexpressing cultures exhibited enhanced

2 NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8 ARTICLE

abcd **** ** * Control 200 20 10 * 3% Trehalose (5 days) 150 **** 8 15 100 ** 6 5 **** 20 3 Arg2 * Arg2 n.s. Arg2 4 n.s. 4 2 5 1 2 3 Relative expression Relative Relative expression Relative Relative expression Relative 0 0 0 Arg2 2 1 Relative expression Relative Control Control Control Starved 0 Trehalose Trehalose – + 1 mM Trehalose 1 mM 1 mM Trehalose 1 mM 10 mM Trehalose 10 mM 10 mM Trehalose 10 mM ARG2 39 kDa 100 mM Trehalose 100 mM 100 mM Trehalose 100 mM Trehalose/Pyruvate 1 mM LactoTrehalose

10 mM LactoTrehalose β 100 mM LactoTrehalose -Actin 42 kDa efControl Fed Starved (12 h) Starved (24 h) Starved (24 h) **** ** 10 6 ** 8 6 **** Fasting 0 h 12 h 24 h 4 Arg2 4 39 kDa Arg2 ARG2 2 g 2 Control β-Actin 42 kDa Relative expression Relative

Relative expression Relative FFA FFA/LPS 0 0 AdArg2 FFA AdArg2 FFA/LPS h Control Ad-Arg2 **** 4 *** ** 3000 **** Insulin –+++ –+++ 2500 ) ** * ArG2 39 kDa –1 3 2000 1500 p-AKT (Ser473) 60 kDa 2 2 Arg2 1 1

60 kDa protein (TG AKT OD500:OD562

0 expression Relative 0 GAPDH 37 kDa

10 **** * *** * i Control db/+ Control Ad-Arg2 p = 0.056 10 mM fructose db/db 8 *** ** n.s. n.s. **** 6 4 4 ** 1.5 ** *** *** 1.5 * **** n.s. n.s. 4 **** ** ** 3 3 ** n.s. 1.0 1.0 2

2 2 expression Relative Arg2 Arg2 0

p-AKT/AKT 0.5 0.5 1 1 Arg2/GAPDH II-6 II-1 Tnf- CcI2 Relative expression Relative

Relative expression Relative CxcI9 Relative expression Relative 0 0 0.0 expression Relative 0.0

Fig. 1 Arg2 is upregulated during fasting and trehalose treatment. a Arg2 gene expression by qPCR analysis of cultured primary hepatocytes treated with 1–100 mM trehalose (n = 6 cultures per group, derived from three mice). b Arg2 gene expression in response to 100 mM trehalose in the presence or absence of energetic reconstitution with pyruvate in primary hepatocytes (n = 6 hepatocyte cultures per group, derived from three mice). c Arg2 gene expression by qPCR analysis of cultured primary hepatocytes treated with trehalose, or the trehalase-resistant analog, lactotrehalose (n = 6 cultures per group, derived from three mice). d Hepatic Arg2 expression by qPCR (top) and immunoblot (bottom) analysis of crude liver lysates from mice subjected to trehalose feeding (3% water ad libitum, 5 days) (n = 5 mice per group). e Arg2 expression by qPCR and immunoblot analysis of crude liver lysates from mice subjected to fasting (n = 5 mice per group). f Arg2 expression by qPCR analysis of primary hepatocytes subjected to low glucose and low serum treatment (24 h). g Left panels: in vitro enzymatic-colorimetric triglyceride accumulation assay and Right: Arg2 gene expression in response to BSA- conjugated fatty acids with or without 1 nM lipopolysaccharide (LPS) and with or without adenoviral Arg2 overexpression in primary hepatocytes. Below: Inflammatory gene marker expression in parallel hepatocyte cultures analyzed above (n = 6 cultures per group, derived from three mice). h Immunoblot analysis of AKT phosphorylation in AML12 hepatocytes treated with or without 10 nM insulin following overnight serum deprivation (n = 6 cultures per group). i Arg2 gene expression in primary hepatocyte cultures following 24 h treatment with 10 mM fructose, or in crude liver lysates from db/+mice or db/db diabetic mice (n = 4 mice per group). For box plots, the midline represents the median, boxes represent the interquartile range and whiskers show the full range of values. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 relative to bracketed control, by two-tailed Student’s t-test insulin-induced AKT phosphorylation (Fig. 1h, and Supplemen- fructose-treated and untreated primary hepatocyte cultures, and tary Fig. 6). Together, these data suggested that Arg2 over- in LepR heterozygous (db/+) mice with LepR-deficient (db/db) expression cell autonomously mitigates hepatocyte fat diabetic mice, and in mice fed with a high-fat, high-sucrose diet accumulation, LPS-stimulated inflammatory responses and (HFHS, 14 weeks). This revealed significantly reduced Arg2 insulin sensitivity through the AKT pathway. expression in db/db mice and in response to fructose treatment To define whether suppression of Arg2 correlates with (Fig. 1i). Therefore, Arg2 is induced during fasting and is reduced metabolic disease, we compared hepatocyte Arg2 expression in in states of nutritional excess.

NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8

c AAV8-Control NCD AAV8-Control HFrD a AAV8-Arg2 NCD AAV8-Arg2 HFrD Experimental outline –1 0 4 Time 100 * ** (weeks) 80

NAFLD Progression 60 Mice obtained AAV8 HFrD Sample (age: 4 weeks old injection + feeding collection 40 * ** body weight: ~17 g) acclimation 20 Total weight (%) Total

b AAV8-Control NCD 0 AAV8-Control HFrD Lean Fat AAV8-Arg2 NCD AAV8-Arg2 HFrD e AAV8-Control NCD AAV8-Control HFrD 30 n.s. 30 n.s. AAV8-Arg2 NCD AAV8-Arg2 HFrD n.s. ** ** 25 n.s. 25 ** 3 n.s. 250 n.s. * * n.s. **** 20 20 200 ) n.s. n.s. ) –1 –1 2 n.s. 150

Body weight (g) 15 15 10 weight (g) Body 10 100

1 dL (mg 0 0 mL (ng

Serum insulin 50 012345 012345 Serum glucose Time (weeks) Time (weeks) 0 0 4 * d AAV8-Control NCD AAV8-Arg2 NCD AAV8-Control HFrD AAV8-Arg2 HFrD 3 * VO2 (light) VCO2 (light) RER (Light) Heat (Light) 2 8000 8000 n.s. **** 1.2 ) 40 )

) **** –1 p = 0.072 –1 1 –1 n.s. **** * **** Relative HOMA2 Insulin resistance 6000 6000 *kg 30

*kg 0.9 *kg –1 –1 0 –1 4000 4000 0.6 20 RER kacl*h (ml*h AAV8-Control NCD (ml*h 2 g 2 2000 2000 0.3 10 AAV8-Arg2 NCD VO VCO 0 0 0.0 Heat ( 0 AAV8-Control HFrD VO2 (Dark) VCO2 (Dark) RER (Dark) Heat (Dark) AAV8-Arg2 HFrD n.s. 8 *** 8000 8000 1.2 ) 40 )

) **** **** –1 * –1 **** **** –1 ** *** n.s. * 6000 6000 0.9 *kg 30 6 *kg n.s. *kg –1

–1 n.s. –1 * * **** 4000 4000 0.6 20 4 * RER kacl*h (ml*h * 2 (ml*h 2 2000 2000 0.3 10 2 VO VCO Relative expression Relative 0 0 0.0 Heat ( 0 0 ck 1 Irs1 bp1 G gat ITT Igf Ass1 f GTT AAV8-Control HFrD Mo ) ) AAV8-Arg2 HFrD –1

–1 200 400 n.s. **** ) )

p = 0.083 * –1 300 –1 30,000 150 12,000 *** * n.s. ** 200 * 100 ** 20,000 n.s. ** 8000 120 min 120 120 min 120 –1 100 –1 50 10,000 4000 0 0 Blood glucose (mg dL (mg glucose Blood Blood glucose (mg dL (mg glucose Blood 0306090 120 0306090 120 0 0

Time (min) AUCdL (mg

Fig. 2 Hepatic Arg2 improves whole-body metabolism and insulin sensitivity. a Experimental design to test the role of Arg2 in protection from HFrD feeding. b, c Body weight (b) and echoMRI analysis of body composition (c) of AAV8-Control or AAV8-Arg2 WT mice fed NCD (n = 6–7 mice per group) or HFrD (n = 8 mice per group). d Indirect calorimetric quantiﬁcation of VO2, VCO2, RER, and energy expenditure during light and dark cycles in AAV8- control and -Arg2 mice fed the indicated diet (n = 6 mice for NCD groups and n = 8 mice for HFrD groups). e Serum insulin (left), serum glucose (middle), and HOMA2 IR (right) in AAV8-control and -Arg2 mice fed NCD or HFrD. f Intraperitoneal glucose tolerance test (GTT) and insulin tolerance test (ITT) (n = 5 mice per group). g Hepatic mRNA expression of insulin-responsive genes. For box plots, the midline represents the median, boxes represent the interquartile range and whiskers show the full range of values. For bar graphs, data represent mean+s.e.m. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001 relative to vehicle treatment, by two-tailed Student’s t-test

Liver Arg2 blocks hepatic steatosis and insulin resistance. or brown adipose tissue (Supplementary Fig. 1a). Importantly, we Forced Arg2 expression in vitro reduced steatosis and enhanced did not observe changes in hepatic Arg1 mRNA expression in insulin signaling. In vivo, hepatocyte-specific Arg2 over- hepatic Arg2-overexpressing mice (Supplementary Fig. 1a). expression was achieved by delivering adeno-associated virus 8 by Hepatic Arg2 overexpression significantly increased lean mass tail-vein injection in mice (hereafter, AAV8-Arg2 mice). We fed and reduced fat mass in NCD and HFrD-fed mice (Fig. 2b, c). these mice either a normal chow diet (NCD) or a 5-week high- Moreover, Arg2 attenuated HFrD-induced weight gain without fructose diet (HFrD, experimental design in Fig. 2a) to test altering food intake (Supplementary Fig. 1b). Accordingly, whether forced Arg2 expression in vivo also enhances insulin indirect calorimetry revealed enhanced basal thermogenesis and signaling and reduces hepatic steatosis. We first demonstrated oxygen and carbon dioxide exchange in AAV8-Arg2 versus restricted hepatic overexpression of the Arg2 construct by qPCR HFrD-fed AAV8-controls (Fig. 2d), without altering light- or and western blot analysis, without upregulation in skeletal muscle dark-cycle movement (Supplementary Fig. 1c).

4 NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8 ARTICLE

The generalized fasting response induces peripheral lipolysis to confirmed by significantly reduced quantitative intrahepatic generate substrate for liver fatty acid oxidation and ketogenesis. triglycerides, cholesterol and low-density lipoprotein cholesterol Analogously, hepatic Arg2 overexpression increased serum-free without changes in hepatic free fatty acids (FFA) (Fig. 3b). fatty acid (FFA) in AAV8-Arg2 HFrD mice with no significant Although we observed no changes in serum ALT, AST or changes in serum triglyceride (TG), total cholesterol, and LDL-C albumin in any group (Supplementary Fig. 1g), hepatic (Supplementary Fig. 1d). AAV8-Arg2 mice exhibited lower serum inflammatory gene expression (e.g., Il-1β, Il-6, Tnfα, Ccl2, Cxcl9) insulin and glucose concentrations in both NCD-fed and HFrD- was significantly lower in the livers of AAV8-Arg2 mice when fed group which resulted in lower HOMA of insulin resistance compared with AAV8-Controls in response to HFrD (Fig. 3c). (HOMA-IR) (Fig. 2e). Glucose tolerance test (GTT) and insulin Analysis of major metabolic pathways yielded no consistent tolerance testing (ITT) confirmed improved glucose tolerance and changes in de novo lipogenic genes, import export (Cd36/FAT, insulin sensitivity secondary to hepatic Arg2 overexpression in Mttp), or fatty acid oxidation (Acox1, Cpt1α, Cpt1β, Ucp2, Pparα, HFrD-fed mice (Fig. 2f), without effects on glucose or insulin Pgc1α, Fgf21) in HFrD-fed AAV8-Arg2 mice relative to littermate tolerance in NCD-fed mice (Supplementary Fig. 1e). With regard AAV8-Controls (Supplementary Fig. 1h–j). to insulin-responsive gene activation, Arg2 significantly increased expression of insulin receptor substrate 1 (Irs1), insulin-like growth factor binding protein 1 (Igfbp1), and argininosuccinate Hepatic Arg2 enhances thermogenesis in db/db diabetic mice. synthase 1 (Ass1). With regard to genes involved in de novo To address the generalizability of the protective effects of hepatic lipogenesis, only monoacylglycerol O-acyltransferase 1 (Mogat1) Arg2, we explored the effects of Arg2 expression in a genetic (Fig. 2g) was reduced in AAV8-Arg2 mice. model of obesity, leptin receptor-deficient (db/db) diabetic mice Hepatic analysis revealed lower liver weight and liver weight- (Experimental outline in Fig. 4a). qPCR and immunoblot analysis to-body-weight ratios in HFrD-fed AAV8-Arg2 mice when confirmed restricted Arg2 overexpression in liver without Arg2 compared with HFrD-fed AAV8-Controls, whereas no changes overexpression in visceral white adipose, brown adipose or ske- were detected between NCD-fed AAV8-Arg2 and AAV8-control letal muscle tissue (Supplementary Fig. 2a, b). Hepatic Arg1 groups (Supplementary Fig. 1f). Histological analysis of livers mRNA was increased by 10% uniquely in this genetically obese from control and AAV8-Arg2 mice by H&E and Oil Red O model (but not in the HFrD model), whereas hepatic Arg2 staining demonstrated that Arg2 attenuated HFrD-induced increased greater than 20-fold in db/db AAV8-Arg2 mice (Sup- hepatic lipid accumulation (Fig. 3a). These findings were plementary Fig. 2a). Hepatic Arg2 overexpression attenuated

a AAV8-Control AAV8-Arg2

NCD HFrD NCD HFrD H&E Oil Red O

b AAV8-Control NCD AAV8-Control HFrD c AAV8-Control NCD AAV8-Control HFrD AAV8-Arg2 NCD AAV8-Arg2 HFrD AAV8-Arg2 NCD AAV8-Arg2 HFrD

8 60 ** 15 n.s. n.s. n.s. * 6 n.s.

40 tissue) 10 * * *** * * * * tissue)

–1 4 –1 **** 20 5 2 g mg Hepatic FFA μ ( 0 Relative expression 0 (nmol mg 0 Hepatic triglycerides Il-6 Il-1 Ccl2 80 n.s. ** 4 n.s. * Tnf- Cxcl9 60 3 tissue) tissue)

–1 40 –1 2 20 1 g mg g mg μ μ Hepatic LDL-C ( (

Hepatic cholesterol 0 0

Fig. 3 Hepatic Arg2 protects against hepatic steatosis and inflammation. a Representative hematoxylin and eosin (H&E)-stained (top) and Oil Red O- stained (bottom) liver sections from AAV8-Control (left) and AAV8-Arg2 (right) mice that were fed NCD or HFrD for 5 weeks. Scale bar, 100 µm. b Triglyceride, cholesterol, non-esterified fatty acid, and LDL-C contents in liver of AAV8-Control or AAV8-Arg2 injected WT mice that fed NCD (n = 6to 7 mice per group) or HFrD (n = 8 mice per group). c Hepatic expression of genes involved in cytokine and chemokine inflammatory responses. Gene expression was normalized to 36B4 mRNA levels. For box plots, the midline represents the median, boxes represent the interquartile range, and whiskers show the full range of values. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001 relative to vehicle treatment, by two-tailed Student’s t-test

NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8

a Experimental outline 510 Time (weeks)

db/db mice obtained AAV8 Sample (age: 5 weeks old injectiom + collection body weight: ~30 g acclimation

b c db/db AAV8-Control db/db AAV8-Control db/db AAV8-Arg2 db/db AAV8-Arg2 d db/db AAV8-Control db/db AAV8-Arg2 55 **** 70 50 **** p = 0.060 VCO2 (Light) VO2 (Light) RER (Light) Heat (Light) 45 *** 60 * * 6000 6000 1.2 25 40 ** ) ) ****

* )

**** –1 50 –1 **** 35 n.s. –1 20

0.9 *kg *kg *kg 30 4000 4000 –1 –1

40 –1 Body weight weight (g) Body 15

Total weight (g) Total weight 0.6

0 RER 10 30 (ml*h 2 2000 (ml*h 2000 0 5 6 7 8 9 10 11 2 0.3 Lean Fat 5

Age (weeks) VO VCO 0 0 0.0 Heat (kacl*h 0 e db/db AAV8-Control db/db AAV8-Arg2 VCO2 (Dark) VO2 (Dark) RER (Dark) Heat (Dark) ** p = 0.10 30 **** 15 500 n.s. 2.5 6000 6000 **** 1.2 ) * ) **** –1 ) –1 400 2.0 –1 0.9 *kg ) *kg ) –1 10 4000 *kg 4000 20 –1 –1 –1

300 1.5 –1 0.6

200 1.0 RER (ml*h 10 2 2000 (ml*h

(ng mL (ng 5 2000 (mg dL (mg 2 0.3 Serum insulin 100 0.5 Serum glucose Relative HOMA2 VO insulin resistance Heat (kacl*h VCO 0 0 0.0 0 0 0.0 0

f db/db AAV8-Control GTT ITT )

) db/db AAV8-Arg2

–1 500 –1 800 * ) n.s. 400 ) –1 600 80,000 * –1 40,000 ** * 300 * n.s. n.s. 400 60,000 30,000 200 120 min 120 120 min 120 –1 200 40,000 100 –1 20,000 20,000 0

0 dL (mg glucose Blood 10,000 Blood glucose (mg dL (mg glucose Blood 0306090120 0306090120 0 Time (min) 0 AUCdL (mg

Time (min) AUCdL (mg

g db/db AAV8-Control db/db AAV8-Arg2 h db/db db/db db/db AAV8-Control 10 * AAV8-Control AAV8-Arg2 db/db AAV8-Arg2 8 5 p-AKT (Ser473) 60 kDa * 6 ** ** * * 4 4 AKT 60 kDa 3 2 2 Relative expression

β-Actin 42 kDa p-AKT/AKT 0 1

Irs1 (Relative expression) Ass1 Gck 0 Igfbp1 Mogat1

Fig. 4 Hepatic Arg2 increases thermogenesis and insulin sensitivity in db/db mice. a Experiment schematic used to test the role of AAV8-mediated mouse Arg2 overexpression in db/db mice. b, c Body weight (b) and body composition (c) of AAV8-Control or AAV8-Arg2 injected db/db mice (n = 8 mice per group). d VO2, VCO2, energy expenditure, and RER during light and dark cycles in AAV8-injected db/db mice (n = 8 mice per group). e Serum insulin (left), glucose (middle), and HOMA2 IR (right) (n = 8 per group). f Intraperitoneal glucose tolerance test (GTT) (above) and insulin tolerance test (ITT) (below) (n = 5 per group). g Hepatic mRNA expression of insulin-responsive genes in liver samples from db/db AAV8-Control or AAV8-Arg2 mice (n = 8 mice per group). Gene expression was normalized to 36B4 mRNA levels. h Western blot analysis of total and phosphorylated AKT in liver samples from db/db AAV8-Control and AAV8-Arg2 mice (n = 4 mice per group). β-actin was probed as a loading control. For box plots, the midline represents the median, boxes represent the interquartile range and whiskers show the full range of values. For bar graphs, data represent mean+s.e.m. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001; n.s., not significant; relative to control treatment, by two-tailed Student’s t-test body weight gain in db/db AAV8-Arg2 mice compared with the TG, decreased total cholesterol and LDL-C, but it did not change db/db AAV8-Control mice (Fig. 4b) without changes in total food FFA (Supplementary Fig. 2e). Analysis of insulin and glucose intake or locomotion (Supplementary Fig. 2c, d). Accordingly, db/ homeostasis revealed that serum insulin and homeostatic model db Arg2 mice had greater lean mass and lower fat mass (Fig. 4c), assessment of insulin resistance (HOMA2-IR) were significantly fi which corresponded with signi cantly higher VO2, VCO2, and decreased in db/db AAV8-Arg2 mice (Fig. 4e). Similarly, AAV8- energy expenditure in db/db AAV8-Arg2 mice versus db/db Arg2-treated db/db mice exhibited significantly improved insulin AAV8-Arg2 control mice. We did not observe changes in tolerance testing (ITT) and trends toward improved glucose tol- food consumption or movement (Fig. 4d, and Supplementary erance when compared with AAV8-Control injected db/db lit- Fig. 2c, d). Hepatic Arg2 overexpression also increased circulating termates (Fig. 4f). Consistent with these physiological findings,

6 NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8 ARTICLE gene expression of insulin-responsive genes, Irs1, Igfbp1, and AAV8-Arg2 mice (Fig. 5a). Histological analysis of H&E and Oil Ass1, were elevated in livers of db/db AAV8-Arg2 mice versus Red O staining confirmed reduced lipid accumulation in db/db AA8-Controls (Fig. 4g). db/db AAV8-Arg2 mice exhibited sig- AAV8-Arg2 mice when compared with AAV8-control littermates nificantly enhanced hepatic AKT phosphorylation relative to db/ (Fig. 5b). Biochemically, we observed lower serum ALT and db AAV8-Control littermates (Fig. 4h, and Supplementary Fig. 7). trends toward lower AST with elevated serum albumin con- Similarly, analysis of iBAT, skeletal muscle and visceral adipose centration in db/db AAV8-Arg2 mice relative to those db/db tissues each revealed trends toward increased AKT phosphor- AAV8-Control mice (Fig. 5c). Attenuated steatosis in db/db ylation in db/db AAV8-Arg2 mice relative to db/db AAV8- AAV8-Arg2 mice was paralleled by decreased fatty acid synthetic Control littermates (Supplementary Fig. 2f). genes (Fig. 5d), and fatty acid uptake and export genes (Fig. 5e). However, no consistent changes in gluconeogenic or fatty acid oxidation gene programs were observed (Supplementary Fig. 2g). Hepatic Arg2 mitigates hepatic steatosis in db/db mice.We Arg2 overexpression also attenuated the hepatic inflammatory examined the role of Arg2 in mediating the major hallmarks of gene expression response, as assessed by qPCR analysis of Il-1β, hepatic steatosis and inflammation. Similar to the observations Il-6, Tnf-α, Ccl2, and Cxcl9 in db/db mice as compared with their made in HFrD-fed AAV8-Arg2 mice, hepatic Arg2 over- littermate controls at 10 weeks of age (Fig. 5f). Moreover, in expression decreased hepatic TG, FFA, and LDL-C content, hepatic tissue extracts and in serum, only hepatic alanine was without significantly altering total hepatic cholesterol in db/db significantly altered in db/db AAV8-Arg2 mice versus db/db

a db/db AAV8-Control db/db AAV8-Arg2 b db/db AAV8-Control db/db AAV8-Arg2

n.s. 150 ** 8

6 100 tissue) tissue) H&E –1

–1 4 50 g mg

g mg 2 μ μ ( ( Hepatic cholesterol Hepatic

Hepatic triglycerides 0 0 ** 15 * 4

3 10 tissue) tissue) –1 Oil Red Oil Red O

–1 2 5

g mg 1 Hepatic FFA Hepatic μ Hepatic LDL-C Hepatic ( (nmol mg (nmol 0 0 c d db/db AAV8-Control db/db AAV8-Arg2 db/db AAV8-Control db/db AAV8-Arg2

150 200 n.s. 6 2.5 * * * ) ** * ** ** *** * ) –1 2.0 ) –1 150 5 100 –1 1.5 100 4 1.0

ALT (U L ALT 50 3 ALT (U L ALT 50 0.5 Albumin (g dL 2 0 0 0 Relative expression 0.0 Fasn Scd1 Acc1 Gpam Ppar e f Chrebp rebp-1c db/db AAV8-Control db/db AAV8-Control S db/db AAV8-Arg2 db/db AAV8-Arg2

2.0 4 ****n.s. ** 1.5 3

1.0 2

0.5 1 Relative expression 0.0 Relative expression 0 Mttp l-1 IL-6 Cd36 I Tnf- Ccl2 Cxcl9

Fig. 5 Hepatic Arg2 attenuates hepatic steatosis and inflammation in db/db mice. a Representative hematoxylin and eosin (H&E)-stained (top) and Oil Red O-stained (bottom) liver sections from AAV8-Control (left) and AAV8-Arg2 (right) db/db mice at 10 weeks of age. Scale bar, 100 μm. b Triglyceride, cholesterol, non-esterified fatty acid, and LDL-C contents in liver of AAV8-Control and AAV8-Arg2 injected db/db mice (n = 8 mice per group). c Serum ALT (left), AST (middle), and albumin (right) in db/db AAV8-Control and AAV8-Arg2 mice (n = 7 mice per group). d, e Hepatic mRNA expression of the indicated genes involved in de novo lipogenesis (d), and fatty acid uptake and export (e)indb/db AAV8-Control and AAV8-Arg2 mice (n = 8 mice per group). Gene expression was normalized to 36B4 mRNA levels. f Inflammatory mediator mRNA expression by qPCR in liver from db/db AAV8-Control and AAV8-Arg2 mice (n = 8 mice per group). For box plots, the midline represents the median, boxes represent the interquartile range and whiskers show the full range of values. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001; n.s., not significant; relative to control treatment, by two-tailed Student’s t-test

NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8

a c db/db Control ASO Experimental outline 56 10 Time db/db Arg2 ASO (weeks)

db/db mice obtained ASO ASO Sample 40 ** * (age: 5 weeks old injection + injection collection body weight: ~25 g acclimation every week 35 **** 30

25 20 0 Rectal temperature(°C) b db/db Control ASO db/db Control ASO cle db/db Arg2 ASO asted db/db Arg2 ASO F 45 Light cycle Dark cy ** * n.s. 60 n.s. 40 * d p = 0.099 55 db/db Control ASO db/db Arg2 ASO 35 n.s. 30 n.s. 50 20 **** 600 **** 8 25 **** 45

Body weight (g) weight Body 15 6 )

20 )

–1 400 –1 0 Total weight (%) 30 5678910 0 10 4 Lean Fat

(ng mL 200 Age (weeks) (mg dL 5 2 Serum insulin Serum glucose Relative HOMA2 Insulin resistance 0 0 0

e db/db Control ASO GTT ITT db/db Arg2 ASO ) ) –1 –1 500 n.s. 500 n.s. * ) 400 n.s. n.s. 400 ) –1 n.s. n.s. 50,000 –1 40,000 n.s. *** n.s. * 300 300 40,000 n.s. 30,000 200 200 120 min 120 30,000 min 120 –1 100 100 –1 20,000 20,000 0 0 10,000 Blood glucose (mg dL Blood glucose (mg dL 0 30 60 90 120 10,000 0 30 60 90 120 Time (min) 0 Time (min) 0 AUCdL (mg AUCdL (mg Fig. 6 Hepatic Arg2 deﬁciency exacerbates hyperinsulinemia in db/db mice. a Experiment schematic used to test the role of hepatic Arg2 knockdown in db/db mice. b Body weight and body composition of Control ASO or Arg2 ASO-treated db/db mice (n = 8 mice per group). c Rectal temperature in db/db Control ASO or Arg2 ASO mice (n = 8 mice per group). d Serum insulin (left), glucose (middle), and HOMA2 IR (right) in Control ASO or Arg2 ASO-treated db/db mice (n = 8 mice per group). e Intraperitoneal glucose tolerance test (GTT) and insulin tolerance test (ITT) (n = 5 mice per group). For box plots, the midline represents the median, boxes represent the interquartile range and whiskers show the full range of values. For bar graphs, data represent mean+s.e.m. *P <0.05,**P < 0.01, ***P < 0.005, ****P < 0.0001; n.s., not signiﬁcant; relative to control treatment, by two-tailed Student’s t-test

AAV8-Control mice after multiple comparison correction, as lower in db/db Arg2 ASO-treated mice when compared with quantified by targeted metabolomic analysis (Supplementary db/db Control ASO-treated mice (Fig. 6c). Table 1). Surprisingly, no significant steady-state changes in urea Quantifying serum parameters in db/db Arg2 ASO-treated cycle intermediaries or in arginine itself were defined in serum or mice revealed lower serum triglycerides and glucose, but elevated hepatic extracts from db/db AAV8-Control and db/db AAV8- total cholesterol and LDL cholesterol as well as elevated insulin Arg2 mice. and calculated HOMA2 in db/db Arg2 ASO-treated mice when compared with db/db Control ASO-treated mice (Fig. 6d, and Supplementary Fig. 3c), although fasting glucose was lower in Higher cholesterol and insulin in Arg2ASO db/db mice. Arg2 ASO mice, in concordance with lower gluconeogenic gene Hepatic Arg2 overexpression protected mice from diet- and (G6pc and Fbp1) expression (Fig. 6d, and Supplementary Fig. 3d). genetically-induced insulin resistance, adiposity and hepatic Total areas under the insulin- and glucose tolerance testing curves steatosis and inflammation. To examine the consequences of were unchanged in db/db Control versus Arg2 ASO-treated mice liver-selective Arg2 deficiency, we treated db/db mice with either (Fig. 6e). scrambled or Arg2-directed antisense oligonucleotides, followed by metabolic assessment (Fig. 6a). This resulted in significant hepatic Arg2 mRNA knockdown without knockdown in iBAT, Liver steatosis and inflammation after liver Arg2 knockdown. SKM or WAT (Supplementary Fig. 3a) or in kidney (not shown). We next examined hepatic effects of liver Arg2 deficiency in There were no hepatic changes in Arg1 mRNA expression db/db mice. Total liver weight and liver weight-to-body weight (Supplementary Fig. 3a). Endpoint total body weight, total fat ratios were elevated in liver Arg2-deficient mice (Fig. 7a). Con- mass, and fat percentage did not significantly change (Fig. 6b), comitantly, liver Arg2-deficient db/db mice exhibited increased although food consumption was slightly lower in liver Arg2- neutral lipid accumulation, as determined by routine histological deficient mice (Supplementary Fig. 3b), and core temperature was and Oil Red O analyses (Fig. 7b). Accordingly, hepatic

8 NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8 ARTICLE

a db/db db/db Control ASO Arg2 ASO

b db/db Control ASO db/db Arg2 ASO

*** ****

H&E 3 8

6 2 4 1 Liver/body 2 Liver weight (g) Weight ratio (%) 0 0 Oil Red O

c db/db Control ASO db/db Arg2 ASO 80 **** 6 ** 4 n.s. 8 **** 60 3 6 4 tissue) tissue) tissue) tissue) –1 –1 –1 40 –1 2 4 2 20 1 g mg g mg g mg 2 Hepatic FFA Hepatic μ μ μ Hepatic LDL-C ( ( ( (nmol mg Hepatic cholesterol Hepatic triglycerides 0 0 0 0 e d db/db Control ASO db/db Control ASO db/db Arg2 ASO db/db Arg2 ASO * n.s. 800 ** 1000 5 ) 40 *** –1 ) ) 600 800 –1 –1 30 600 4 *** 400 400 20 * 3 * ALT (U L ALT (U L 200 p = 0.067 200 10

Albumin (g dL 2 0 0 0 Relative expression 0 Il-6 Il-1 Tnf- Ccl2 Cxcl9

Fig. 7 Hepatic Arg2 deficiency exacerbates hepatic steatosis and inflammation in db/db mice. a Representative hematoxylin and eosin (H&E)-stained (top) and Oil Red O-stained (bottom) liver sections from Control ASO and Arg2 ASO db/db mice at 10 weeks of age. Scale bar, 100 μm. b Liver weight and liver weight-to-body weight ratio in Control ASO and Arg2 ASO db/db mice. c Triglyceride, cholesterol, non-esterified fatty acid, and LDL-C contents in liver of AAV8-Control and AAV8-Arg2 injected db/db mice (n = 8 mice per group). d Serum ALT (left), AST (middle), and albumin (right) in db/db AAV8-Control and AAV8-Arg2 mice (n = 7 mice per group). e Hepatic mRNA expression of the indicated genes involved in hepatic inflammation in db/db Control and Arg2 ASO mice (n = 8 mice per group). For box plots, the midline represents the median, boxes represent the interquartile range and whiskers show the full range of values. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001; n.s., not significant; relative to control treatment, by two-tailed Student’s t-test triglycerides, total cholesterol and LDL-C and serum ALT and component analyses revealed distinct gene clusters between db/db AST were all increased (Fig. 7c), without alterations in hepatic and db/db AAV8-Arg2 groups (Fig. 8a). This was corroborated by synthetic function in liver Arg2-deficient mice versus db/db unsupervised gene-level clustering revealing distinctive patterns controls (Fig. 7d). Hepatic inflammatory marker gene expression of significantly upregulated and downregulated genes when of Il-6, Tnf-α, Ccl2, and Cxcl9 was increased in hepatic Arg2- comparing db/db AAV8-Arg2 and db/db AAV8-Control mice deficient db/db mice versus db/db controls (Fig. 7e). Whereas (Fig. 8b). Kyoto Encyclopedia of Genes and Genomes (KEGG) gluconeogenic gene expression in Arg2-deficient db/db livers pathway enrichment analysis of differentially expressed genes (Supplementary Fig. 3d), genes involved in de novo lipogenesis: revealed enhanced phosphatidylinositol signaling, mTOR and Scd1, Gpam, and Pparγ were elevated in Arg2-deficient db/db FOXO signaling, consistent with enhanced hepatic insulin sig- livers when compared with db/db controls. naling. Conversely, consistent with fasting-state cellular physiology, significant downregulation of fructose, mannose, amino acid and nucleotide sugar metabolism, fatty acid elongation and bio- Transcriptomics reveal inverse Arg2-Rgs16 regulation.To synthesis, and carbon/pyruvate metabolism (Fig. 8c). Volcano explore potential mechanisms by which hepatic Arg2 over- plot analysis confirmed upregulation of Arg2 gene expression in expression improves hepatic lipid homeostasis, inflammation and db/db AAV8-Arg2 mice, and identified the hepatocyte regulator insulin sensitivity, we examined unbiased transcriptional profiling of G-protein signaling 16 (RGS16) as among the most highly by RNA-seq to analyze transcriptomic changes in the livers of suppressed genes in db/db AAV8-Arg2 mice (Fig. 8d). In light of db/db AAV8-Arg2 mice and AAV8-Control mice. Principal the described pathogenic role of RGS16 in hepatic steatosis35,we

NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8

a 50 b AAV8-Control AAV8-Arg2 c mmu04550 Signaling pathways regulating pluripotency of stem cells mmu04310 Wnt signaling pathway mmu04068 FoxO signaling pathway 40 mmu04360 Axon guidance mmu04722 Neurotrophin signaling pathway 30 2 mmu04261 Adrenergic signaling in cardiomyocytes mmu04010 MAPK signaling pathway 20 mmu04330 Notch signaling pathway mmu04150 mTOR signaling pathway 10 mmu04660 T cell receptor signaling pathway mmu04024 cAMP signaling pathway mmu04916 Melanogenesis –80 –60 –40 –20 20 40 60 80 mmu04012 ErbB signaling pathway PC2 (25.69%) –10 mmu04919 Thyroid hormone signaling pathway mmu04520 Adherens junction –20 0 mmu04022 cGMP-PKG signaling pathway mmu04020 Calcium signaling pathway mmu04390 Hippo signaling pathway –30 mmu04120 Ubiquitin mediated proteolysis d PC1 (57.32%) mmu04070 Phosphatidylinositol signaling system mmu00051 Fructose and mannose me tabolism mmu00640 Propanoate metabolism mmu00520 Amino sugar and nucleotide sugar metabolism 10 Arg2 mmu03060 Protein export mmu00062 Fatty acid elongation 8 –2 mmu00592 alpha-Linolenic acid metabolism mmu01040 Biosynthesis of unsaturated fatty acids mmu00620 Pyruvate metabolism -Value) 6 mmu04146 Peroxisome P Rgs16 mmu04723 Retrograde endocannabinoid signaling 4 mmu00280 Valine, leucine and isoleucine degradation mmu01200 Carbon metabolism 2 mmu04145 Phagosome p = 0.05 mmu00480 Glutathione metabolism –Log10( mmu03050 Proteasome 0 –4 mmu03320 PPAR signaling pathway –4 –2 0 2 4 6 mmu04612 Antigen processing and presentation mmu01212 Fatty acid metabolism Log2 (Fold change) mmu00190 Oxidative phosphorylation mmu03010 Ribosome

e r 2 = 0.5268 3 P = 0.0415 15 4 r 2 = 0.2158 r 2 = 0.1085 3 P = 0.0043 2 10 P = 0.0935 2 Rgs16 Rgs16 Rgs16 1 5 1 Relative expression Relative expression Relative expression 0 0 0 0.0 0.5 1.0 1.5 0204060 01020304050 Arg2 Arg2 Arg2 Relative expression Relative expression Relative expression

f AAV8-Control NCD AAV8-Control HFrD g db/db AAV8-Control AAV8-Arg2 NCD AAV8-Arg2 HFrD db/db AAV8-Arg2 *** 14 4 ** 2.0 **** 2.0 12 * 10 3 1.5 1.5 8 2 P = 0.064 6 ** 1.0 1.0 Rgs16 4 Rgs16 1 0.5 0.5

2 RGS16/GAPDH RGS16/GAPDH Relative expression (Relative expression) 0 0 Relative expression

0.0 (Relative expression) 0.0 AAV8-Control AAV8-Arg2 db/db db/db AAV8-Control AAV8-Arg2 NCDHFrD NCD HFrD

RGS16 28 kDa RGS16 28 kDa GAPDH 37 kDa GAPDH 37 kDa

Fig. 8 Inverse Arg2-RGS16 regulatory relationship is identified by unbiased transcriptomics. a Principal component analysis of transcriptomic data from db/db AAV8-Control and AAV8-Arg2 livers. b Heat map demonstrating gene-level regulation in db/db AAV8-Control and AAV8-Arg2 livers. c KEGG pathway analysis demonstrating top up- and downregulated gene pathways in db/db AAV8-Arg2 livers. d Volcano plot demonstrating significantly altered genes by expression level and P-value. RGS16 and Arg2 are highlighted as highly regulated genes identified within the dataset. e Linear regression analysis demonstrating inverse Arg2 and RGS16 correlation in (left to right:) db/+ versus db/db livers, in HFrD-fed AAV8-control and AAV8-Arg2 mice, and in db/db AAV8-control and AAV8-Arg2 mice. f, g RGS16 gene expression and protein abundance in HFrD-fed (f) and db/db (g) AAV8-control and AAV8- Arg2 mouse livers. For box plots, the midline represents the median, boxes represent the interquartile range and whiskers show the full range of values. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001; n.s., not significant; relative to control treatment, by two-tailed Student’s t-test further interrogated whether Arg2 and RGS16 anti-correlate in deficient mice exhibited upregulated hepatic Rgs16 mRNA multiple models of metabolic disease. In db/+ mice, in wild-type (Supplementary Fig. 4). Together, the data identify an inverse mice overexpressing Arg2 and in db/db AAV8-Arg2 mice, we relationship between hepatic Rgs16 and Arg2. observed an inverse relationship between Arg2 and Rgs16 expression (Fig. 8e). Moreover, hepatic Rgs16 mRNA and protein levels were significantly suppressed by direct Arg2 expression in Hepatic Rgs16 reconstitution reverses hepatic Arg2 effects.We the livers of chow and HFrD-fed and db/db AAV8-Arg2 mice next examined the hypothesis that genetic reconstitution of (Fig. 8f, g, and Supplementary Fig. 8), whereas hepatic Arg2- hepatic RGS16 in the context of Arg2 overexpression would

10 NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8 ARTICLE

a b db/db AAV8-Control Experimental outline 5 610 Time db/db AAV8-Arg2 (weeks) db/db AAV8-Arg2 Ad-Rgs16 c db/db mice obtained AAV8 Ad Sample 45 * 70 (age: 5 weeks old injection + injection collection n.s. ** body weight: ~30 g) acclimation every week 40 n.s. * 60 n.s. ** 35 50 n.s. d db/db AAV8-Control db/db AAV8-Arg2 db/db AAV8-Arg2 Ad-Rgs16 30 40 VO2 (Light) VCO2 (Light) RER (Light) Heat (Light) 25 Total weight (g) Total weight

Body weight (g) weight Body 30 0 0

**** ) 5000 ) 5000 1.2 25 **** ) *** **** **** **** 5678910 Lean Fat –1 **** –1 **** –1 4000 4000 20 Age (Weeks) *kg *kg 0.9 *kg –1 –1

–1 3000 3000 15 0.6 e db/db AAV8-Control db/db AAV8-Arg2

2000 2000 RER 10 kcal*h (ml*h (ml*h 2 db/db AAV8-Arg2 Ad-Rgs16 2 1000 1000 0.3 5 *** * VO 10 **** 300

0 0 0.0 Heat ( 0 8 VO2 (Dark) VCO2 (Dark) RER (Dark) Heat (Dark) ) ) * –1 –1 200 n.s. 6 ) 5000 ) VCO **** ) **** **** 5000 **** 1.2 25 **** **** –1 –1 *** 4 –1 100 (ng mL (ng

4000 4000 20 mL (mg *kg *kg 0.9

*kg 2 Serum insulin Serum –1 Serum glucose –1

–1 3000 3000 15 0.6 0 0

2000 2000 RER 10 (ml*h 3 **** (ml*h 2 2 1000 1000 0.3 5 * VO VCO 0 0 0.0 Heat (kcal*h 0 2

f db/db AAV8-Control 1 db/db AAV8-Arg2 Relative HOMA2 Relative GTT ITT insulin resistance 0 db/db AAV8-Arg2 800 * Ad-Rgs16 600 n.s. n.s. * ) * ) * ) * * 600 80,000 * ) 50,000 –1 * * –1 –1

n.s. –1 400 P = 0.13 400 60,000 40,000 *

(mg dL (mg 200 30,000 (mg dL (mg 120 min 120 min

200 40,000 Blood glucose Blood AUC AUC Blood glucose Blood

–1 –1 20,000 0 20,000 0 10,000 0306090120 0 306090120 (mg dL (mg Time (min) 0 Time (min) dL (mg 0

g AAV8-Arg2 AAV8-Control AAV8-Arg2 Ad-Rgs16 db/db AAV8-Control db/db AAV8-Arg2 db/db AAV8-Arg2 Ad-Rgs16 ** n.s. 3 * ** 6 1.5 * * p-AKT (Ser473) 60 kDa

AKT 60 kDa 2 4 1.0 -Actin -Actin β ARG2 39 kDa β 1 2 0.5 RGS16 28 kDa p-AKT/AKT ARG2/ β RGS16/ -Actin 42 kDa 0

0 expression) (Relative 0.0 (Relative expression) (Relative (Relative expression) (Relative

Fig. 9 Hepatic RGS16 reconstitution reverses Arg2-mediated improvements in peripheral fat accumulation and glucose homeostasis. a Experimental schematic used to test the role of AAV8-mediated mouse Arg2 and Ad-RGS16 overexpression in db/db mice. b Body weight and c body composition of

AAV8-Control or AAV8-Arg2 db/db mice expressing Ad-Control or Ad-RGS16 (n = 8 per group). d VO2, VCO2, energy expenditure, and RER during light and dark cycles in AAV8-Control or AAV8-Arg2 db/db mice expressing Ad-Control or Ad-RGS16 (n = 8 per group). e Serum insulin (left), glucose (middle), and HOMA2 IR (right) in AAV8-Control or AAV8-Arg2 db/db mice expressing Ad-Control or Ad-RGS16 (n = 8 per group). f Intraperitoneal GTT and ITT (n = 5 mice per group). g Western blot analysis of total and phosphorylated AKT in liver samples from db/db AAV8-Control and AAV8-Arg2 mice (n = 4 mice per group). For box plots, the midline represents the median, boxes represent the interquartile range and whiskers show the full range of values. For bar graphs, data represent mean+s.e.m. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001; n.s., not significant; relative to control treatment, by two-tailed Student’s t-test reverse the improvements in hepatic steatosis, inflammation and adenovirus-treated cohorts (Fig. 9b) than in the control cohorts insulin and glucose intolerance observed in Arg2-treated db/db (Fig. 4b). Nevertheless, these changes were reversed by RGS16 mice. To that end, we treated db/db AAV8-Control and AAV8- reconstitution in db/db AAV8-Arg2 mice when compared with Arg2 mice with adenovirus encoding GFP (Ad-Control) or RGS16 Ad-Control-treated mice (Fig. 9b, c). db/db AAV8-Arg2 mice (Ad-Rgs16, Experimental outline, Fig. 9a). We confirmed Arg2 again exhibited enhanced energy expenditure and oxygen con- overexpression, Rgs16 suppression and mild Rgs16 reconstitution sumption by indirect calorimetry when compared with db/db in response to our AAV8 and adenoviral manipulation (Supple- AAV8-control mice without altering locomotion or food con- mentary Fig. 5a, b). In db/db mice, Arg2 overexpression atte- sumption (Fig. 9d, and Supplementary Fig. 5c, d). However, nuated weight gain, increased fat mass and decreased lean mass, RGS16 reconstitution reversed the energetic enhancements in although we noted the weight bifurcation occurred later in the db/db AAV8-Arg2 mice (Fig. 9d). Accordingly, Rgs16

NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8

a AAV8-Arg2 b db/db AAV8-Control db/db AAV8-Arg2 AAV8-Control AAV8-Arg2 Ad-Rgs16 db/db AAV8-Arg2 Ad-Rgs16 ** 80 2.5 *** ** 60 2.0 H&E tissue) tissue) 1.5

–1 40 –1 1.0 20 0.5 (ug mg (ug 0 mg (ug Hepatic cholesterol 0.0 Hepatic triglycerides **** 4 *** 4 ** ** 3 3 Oil Red O tissue) tissue) –1

2 –1 2 1 1 Hepatic FFA Hepatic Hepatic LDL-C (ug mg 0 (nmol mg 0

c db/db AAV8-Arg2 db/db AAV8-Arg2 Ad-Rgs16 d db/db AAV8-Control e db/db AAV8-Control db/db AAV8-Arg2 db/db AAV8-Arg2

120 200 ) 5 n.s. ** db/db AAV8-Arg2 Ad-Rgs16 db/db AAV8-Arg2 Ad-Rgs16 –1 ) ) –1 –1 150 80 4 15 **** 6 *** 100 n.s. n.s. n.s. 40 3 10 * * 4 *** 50 ** n.s.

ALT (U L 2 AST (U L * * *** n.s. * n.s. n.s. * 0 0 Albumin (g dL 0 5 2 n.s.

0 0 Relative expression Relative Relative expression Relative -6 f- l9 -1 L Irs1 Il I Tn Ccl2 Ass1 Gck Cxc Igfbp1 Mogat1

Fig. 10 Hepatic RGS16 reconstitution reverses Arg2-mediated improvements in hepatic fat accumulation. a H&E and Oil Red O-stained liver sections from db/db AAV8-Control and AAV8-Arg2 mice treated with Ad-GFP or Ad-RGS16. Scale bar, 100 µm. b Hepatic triglycerides, cholesterol, non-esterified fatty acids and LDL-C in livers from db/db AAV8-Control and AAV8-Arg2 mice treated with Ad-GFP or Ad-RGS16. c Serum ALT, AST and albumin in db/db AAV8-Arg2 mice expressing Ad-GFP or Ad-RGS16. d Hepatic inflammatory gene marker mRNA by qPCR. e mRNA expression of insulin-responsive genes by qPCR (n = 8 mice per group). For box plots, the midline represents the median, boxes represent the interquartile range and whiskers show the full range of values. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001; n.s., not significant; relative to control treatment, by two-tailed Student’s t-test reconstitution substantially exacerbated insulin resistance in db/db and ornithine36–43. More recently, these enzymes have been mice as defined by serum insulin, glucose, and HOMA2-IR implicated in the biosynthesis of polyamines, and in proline, (Fig. 9e). This was further confirmed by glucose and insulin tol- glutamate, creatine, and agmatine biosynthesis as well38,40,42,44. erance testing (Fig. 9f). Hepatic insulin sensitivity, as quantified by The two vertebrate arginases, type I and type II evolved separately insulin-stimulated AKT phosphorylation, was impaired by Rgs16 from an early gene duplication event36, which ultimately man- expression in the liver (Fig. 9g, and Supplementary Fig. 9) when ifests as both distinct and overlapping enzymatic functions, which compared with db/db AAV8-Arg2 mice treated with Ad-Control. yet remain to be fully appreciated42. We demonstrate here that In the liver, Arg2 decreased hepatic TG, cholesterol, FFA, and Arg2 is more profoundly induced during fasting in liver versus LDL-C in db/db Arg2 mice, and these effects were significantly Arg1. This opens the possibility now that the arginases regulate reversed upon Rgs16 reconstitution in db/db AAV8-Arg2 mice fasting biology in the liver. Although the total abundance of Arg1 (Fig. 10a) with no significant changes detected in serum TG, is much greater than that of Arg2 in the basal setting42, the data cholesterol, FFA and LDL-C (Supplementary Fig. 5e). Histologi- here suggest that Arg1 and Arg2 are in fact differentially regulated cal analysis by H&E and Oil Red O demonstrated that Arg2 during fasting, and may thus serve divergent functions as part of overexpression reduced hepatic lipid accumulation, which was the physiological fasting response. The fundamental biological reversed in db/db mice also expressing AdRGS16 (Fig. 10b). purpose of hepatocyte Arg2 regulation during fed and fasting Similarly, serum ALT, AST as well as hepatic inflammatory states, and during health and disease remains an intriguing area marker expression in db/db mice were each significantly elevated of further exploration. in db/db AAV8-Arg2 mice expressing Ad-Rgs16 as compared The fasting response enhances hepatic and extrahepatic energy with db/db AAV8-Arg2 Ad-Control littermates (Fig. 10c, d). homeostasis, and yet the intermediaries that dictate these effects Rgs16 overexpression reversed Arg2 effects on Igfbp1 expression, remain poorly understood. Here, we demonstrate that hepatocyte and RGS16 overexpression generally increased genes involved in arginase 2 is upregulated during fasting, and after trehalose and de novo lipogenesis in liver, including Chrebp, Acc1, Gpam, and trehalose analog treatment. Furthermore, Arg2 overexpression Pparγ (Fig. 10e). However, the expression of mRNAs encoding per se attenuates insulin resistance, hepatic steatosis and beta oxidation and fatty acid uptake and export were largely inflammation in high-fructose-fed and genetically obese models. unchanged by hepatic Arg2 overexpression or Rgs16 reconstitu- Arg2 is therefore a tractable target that can be used to leverage the tion (Supplementary Fig. 5f). therapeutic hepatic fasting response. Although it must be noted that some of the individual metabolic improvements observed Discussion upon AAV8-Arg2 expression were modest. For example, we Arginases are ancient, highly conserved enzymes36,37, the purpose observed 20–25% improvements in glucose and insulin tolerance of which is long considered to be conversion of L-arginine to urea in HFrD-fed AAV8-Arg2 and db/db x AAV8-Arg2 mice.

12 NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8 ARTICLE

However, it should also be noted that the magnitude of changes negative feedback on Gi/Gq-mediated fat oxidation35,54. This in glucose and insulin tolerance here mirrors that observed in negative feedback appropriately moderates the prolonged fasting humans subjected to short-term intermittent fasting or caloric response (e.g., during hibernation). In overfed contexts, in which restriction, which results in 11–30% improvements in HOMA-IR, excess caloric and fat dissipation is an appropriate full-time and 11–40% improvement in fasting plasma insulin levels7. compensation, RGS16 function becomes maladaptive. The Indeed even high-intensity, supervised diet and exercise training maladaptive nature of this regulation is supported by comparing regimens over a 4-month period produced even more modest Arg2 expression in db/db vs db/+ mice, HFrD-fed vs Chow-fed improvements in glucose tolerance, and only for the highest- mice and in primary hepatocyte cultures treated with or without intensity intervention groups45. 10 mM fructose. In each of these overfed contexts, Arg2 expres- The observation that hepatic Arg2 activation is therapeutic sion was suppressed. Our data demonstrating that Arg2 acts originated from prior observations that hepatic fasting signals upstream of Rgs16 to suppress its induction may reflect a mole- (e.g., TFEB, PPARα) mediated the therapeutic effects of genetic or cular bypass in this negative feedback loop. Clinically, this Arg2- pharmacological glucose transport blockade27,29. However, the regulated bypass route may be exploited to treat NAFLD and observation that hepatic glucose uptake is impaired in diabetic insulin resistance. patients46 may raise an apparent paradox, in light of the intrinsic In summary, this report identifies hepatic Arg2 suppression of therapeutic effect of hepatic glucose transport blockade. At least Rgs16 as a metabolic regulatory axis during fasting. Moreover, we two key differences between the fasted liver and the diabetic liver identify pharmacologic fasting mimetics—trehalose and lacto- resolve this apparent paradox. First, insulin stimulates hepatic trehalose—which activate this axis analogous to physiological glucose uptake by twofold over the fasting state47. In diabetic fasting. Therefore, the hepatic intermediaries that mediate the patients, insulin-stimulated glucose uptake is impaired (by 17.4% generalized macronutrient fasting response can be directly in the Iozzo study46) but this degree of blunted fed-state glucose leveraged to treat cardiometabolic disease. uptake would not be expected to mimic a true hepatocyte fasting response. Second, basal hepatic glucose uptake should remain Methods elevated through hepatic facilitative transporters GLUT1, GLUT2, Cell lines. AML12 cells (CRL-2254) were purchased directly from the American and GLUT8. This is because basal hyperglycemia in the diabetic Type Culture Collection (ATCC) and propagated and maintained precisely per patient provides a glucose concentration gradient that drives manufacturer specification. these facilitative transporters even in the absence of insulin- stimulated glucose transport. This rise in basal hepatic glucose Primary hepatocyte cell isolation, culture and treatment. Primary murine hepatocytes obtained from WT mice were isolated24,25,31 and cultured and uptake would be expected to further impede the hepatocyte maintained in regular DMEM growth media (Sigma, #D5796) containing 10% FBS. fasting response. Thus, although insulin-stimulated glucose For in vitro starvation experiments, “starved” media contained 1 g/L glucose and uptake is relatively impaired in the insulin-resistant state, our 0.5% FBS was used. Cultures were lysed in Trizol and subjected to downstream data suggest that pharmacological hepatic GLUT blockade or analysis. In vitro genetic knockdown was achieved via siRNA transfection using targeted hepatic Arg2 activation can drive the therapeutic hepatic Lipofectamine 3000 from Invitrogen (L3000015). Trehalose was obtained from Sigma Aldrich (St. Louis, MO) and was >97% purity by HPLC. Three percent glucose response in diet-induced and genetic models of obesity trehalose water (w/v) fed ad libitum was used in all in vivo experiments. and insulin resistance. fi Taken further in context, Arg2 appears to serve tissue-speci c, In vitro insulin signaling. AML12 cells were seeded in 6-well plates. Cells were and context-dependent functions in energy homeostasis. These infected with adenoviruses, Ad-GFP, or Ad-Arg2, for 24 h in regular DMEM context-dependent functions underscore the importance of HAM/F12 growth media and starved in serum-free DMEM HAM/F12 media for leveraging hepatocyte fasting responses over less targeted 16 h before being treated with either serum-free media or serum-free media with approaches to metabolic therapy. Endothelial Arg2 over- insulin (10 nM) for 10 min. expression surprisingly impaired endothelial cell autophagy and AMPK signaling48, and macrophage Arg2 appears to have pro- Animal studies. All animal protocols were approved by the Washington Uni- fl 49–51 versity School of Medicine Animal Studies Committee. Male C57B/6J mice and in ammatory functions , whereas germline Arg2 deletion db/db mice were purchased directly from the Jackson Laboratory (Bar Harbor, ME) effects are not yet clear49,52. The clinical implications of these and housed a 12-h alternating light-dark, temperature-controlled, specific data should thus be underscored, especially in view of recent pathogen-free barrier facility prior to and throughout experimentation. momentum to inhibit arginases in the setting of cardiovascular All animals received humane care and procedures were performed in 53 fi accordance with the approved guidelines by the Animal Studies Committee at and other disease . Our liver-de cient Arg2 model suggests that Washington University School of Medicine. All animal studies were performed in blocking Arg2 without sufficient specificity may have negative accordance with the criteria and ethical regulations outlined by the Institutional metabolic consequences, particularly if hepatic Arg2 is inad- Animal Care and Use Committee (IACUC). vertently targeted. Here, we utilized a targeted, hepatocyte- specific overexpression approach to demonstrate further adaptive Quantitative real-time RT-PCR (qRT-PCR). Total RNA was prepared by metabolic effects of hepatic Arg2 overexpression. Mechanistically, homogenizing snap-frozen livers or cultured hepatocytes in Trizol reagent (Invi- trogen #15596026) according to the manufacturer’s protocol. cDNA was prepared our genetic complementation data indicate that Arg2 cell using Qiagen Quantitect reverse transcriptase kit (Qiagen #205310). Real-time autonomously enhances hepatic insulin signaling in vitro and qPCR was performed with Step-One Plus Real-Time PCR System (Applied Bio- in vivo and reduces hepatic fat accumulation via systems) using SYBR Green master Mix Reagent (Applied Biosystems) and specific Rgs16 suppression. The data give rise to the hypothesis that primer pairs. Relative gene expression was calculated by a comparative method Rgs16 suppresses hepatocyte-protective G-protein-coupled using values normalized to the expression of an internal control gene. Primers are shown in Supplementary Table 2. receptor signaling events, the origin and nature of which remain to be addressed. fi Immunoblotting. Tissues were homogenized in RIPA buffer supplemented with RGS16 was previously identi ed as the only one of 21 RGS protease and phosphatase inhibitors (Thermo Scientific). After homogenization, genes that was regulated by circadian rhythms in the murine lysate was centrifuged at 18,000 × g for 15 min at 4 °C, and the supernatant was suprachiasmatic nucleus and liver54,55. Hepatic Rgs16 was sub- recovered. Protein concentration was determined by BCA Assay Kit (Thermo sequently determined to suppress hepatic fat oxidation by upre- Scientific) and was adjusted to 2 mg/mL. Samples for western blotting were pre- 35 pared by adding Laemmli buffer at a ratio of 1:1 and heating at 95 °C for 5 min. gulating the carbohydrate response element binding protein . The prepared samples were subjected to 10% or 13% SDS-PAGE, followed by Because the net physiological response to fasting in the liver is to electrical transfer onto a nitrocellulose membrane using the Trans-Blot Turbo upregulate fat oxidation, the proposed Rgs16 function is to exert system (Bio-Rad). After blocking the membrane with 5% milk in TBST, the

NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8 membrane was incubated in primary antibody at 4 °C overnight. The blot was RNA-seq. RNA-seq was performed by the Washington University Genome developed after secondary antibody incubation using Pierce ECL Western Blotting Technology Access Center (GTAC). Library preparation was performed with 10 µg Substrate (Thermo Scientific). Blots were developed according to the manu- of total RNA with a Bioanalyzer RIN score greater than 8.0. Ribosomal RNA was facturer’s instructions. Protein expression levels were quantified with Image Lab removed by poly-A selection using Oligo-dT beads (mRNA Direct kit, Life software and normalized to the levels of β-actin or GAPDH. Technologies). mRNA was then fragmented in buffer containing 40 mM Tris acetate pH 8.2, 100 mM potassium acetate and 30 mM magnesium acetate and heating to 94 °C for 150 s. mRNA was reverse transcribed to yield cNDA using Antibodies. Antibodies against phosphor-IRS1 (Ser636/639) (no. 2388), IRS1 (no. SuperScript III RT enzyme (Life Technologies, per manufacturer’s instructions) 2382), phosphor-AKT (Ser473) (no. 9271), AKT (no. 9272), GAPDH (no. 5174), β and random hexamers. A second strand reaction was performed to yield ds-cDNA. and -actin (no. 3700) were purchased from Cell Signaling Technology (CST) cDNA was blunt ended, had an A base added to the 3′ ends, and then had Illumina (Beverly, MA, USA). Anti-Arg2 antibody was purchased from Santa Cruz Bio- sequencing adapters ligated to the ends. Ligated fragments were then amplified for technology (no. Ab154422). Antibody against RGS16 (no. NBP2-01584) was 12 cycles using primers incorporating unique index tags. Fragments were purchased from Novus Biologicals, LLC (no. Ab203071) (Littleton, CO, USA). The sequenced on an Illumina HiSeq-3000 using single reads extending 50 bases. dilution ratio for all primary antibodies was 1:1000. The secondary antibodies used RNA-seq reads were aligned to the Ensembl release 76 top-level assembly with in this study were peroxidase-conjugated anti-rabbit IgG (A7074) and anti-mouse STAR version 2.0.4b. Gene counts were derived from the number of uniquely IgG (A7076) purchased from CST, in which were used at a 1:5000 dilution. aligned unambiguous reads by Subread:featureCount version 1.4.5. Transcript counts were produced by Sailfish version 0.6.3. Sequencing performance was Histological analysis. Formalin-fixed paraffin-embedded liver sections were assessed for total number of aligned reads, total number of uniquely aligned reads, stained by H&E via the Washington University Digestive Diseases Research Core genes and transcripts detected, ribosomal fraction known junction saturation and Center. OCT-embedded frozen liver sections were stained by Oil Red O according read distribution over known gene models with RSeQC version 2.3. to standard protocols flowered by microscopic examination. Three liver sections To enhance the biological interpretation of the large set of transcripts, grouping of were examined and evaluated for each animal. For Oil red O staining, ice-cold genes/transcripts based on functional similarity was achieved using the R/Bioconductor methanol-fixed frozen sections from mice were stained according to described packages GAGE and Pathview. GAGE and Pathview were also used to generate protocols 24,25,31. pathway maps on known signaling and metabolism pathways curated by KEGG. Lactotrehalose synthesis and purification was carried out using chemoenzymatic synthesis and purification methods we described56. Reactions Targeted metabolomics. We performed targeted metabolomics as reported with were performed in 50 mM HEPES buffer (pH 7.4) containing 10 mM galactose, minor modifications57. Briefly, the liver samples were homogenized in water μ fi 40 mM UDP-glucose, 20 mM MgCl2, and 9.8 M TreT at 70 °C. We con rmed (4 mL/g liver). The amino acids in 20 µL of mouse serum or liver homogenate were 98–99% purity by 1H-NMR (not shown). We confirmed 98–99% purity by 1H- extracted with protein precipitation in the presence of internal standards NMR prior to experiments. (13C6,15N-Ile, d3-Leu, d8-Lys, d8-Phe, d8-Trp, d4-Tyr, d8-Val, d7-Pro, 13C4-Thr, d3-Met, d2-Gly, d3,15N2-Asn, d4-Cit, d3-Asp, 13C5-Gln, 13C6-His, d3-Glu, d4- Ala, d3-Ser, 13C5-Orn, and 13C6-Arg). Quality control (QC) samples for livers AAV8- and adenovirus-mediated overexpression. Serotype 8 AAV (AAV8) was and sera were prepared from pooled partial study samples and injected every five administered via tail vein as we previously reported29. All viral vectors (RGS16, study samples to monitor intra-batch precision. Only the lipid species with CV% < Arg2, shTFEB) were obtained directly from Vector Biolabs Inc (Philadelphia, PA). 15% for QC injections are reported. The Ile, Leu, Lys, Phe, Trp, Tyr, Val, Pro, Thr, Met, Gly, Asn, Cit, Asp, Gln, His, Glu, Ala, Ser, Orn, and Arg were analyzed on Antisense oligonucleotides. We obtained validated Arg2-specific antisense oli- 4000 QTRAP mass spectrometer coupled with a Prominence LC-20AD HPLC gonucleotides from IONIS Pharmaceuticals (Carlsbad, CA). Mice were treated system. Data processing was conducted with Analyst 1.5.1 (Applied Biosystems). precisely as we previously reported27,32, prior to in vivo and post-mortem assays. Statistical analyses. Data were analyzed using GraphPad Prism version 6.0 (RRID: fi fi Insulin and glucose tolerance testing. For insulin tolerance tests (ITT), mice SCR_015807). p < 0.05 was de ned as statistically signi cant. Data shown are as were injected with 0.75 IU per kg body weight of insulin (Humalog, Eli Lilly) mean ± SEM. In box-whisker plots: middle bar represents the dataset median; boxes intraperitoneally after 4 h of fasting on Aspen bedding. For glucose tolerance tests represent 25 and 75 percentile lines; whiskers represent maximum and minimum (GTT), mice were injected with 2 g per kg body weight of glucose intraperitoneally values in the dataset. In dot plots: the line represents the data mean. Unpaired two- tailed homoscedastic T-tests with Bonferroni post hoc correction for multiple after fasting for 4 h on aspen bedding. db/db mice were injected with 1 g per kg fi body weight of glucose intraperitoneally after fasting for 16 h on aspen bedding. comparisons were used for all analyses unless otherwise noted in the gure legends. Blood samples were measured at different time points with a glucometer (Arkray Two-way ANOVA was also used for analyses with two independent variables. USA, Inc., Minneapolis, MN, USA). Reporting summary. Further information on experimental design is available in the Nature Research Reporting Summary linked to this article. Clinical chemistry measurements and hepatic lipid analyses. For all other serum analyses, submandibular blood collection was performed immediately prior to sacrifice and serum was separated. Insulin ELISA (Millipore #EZRMI-13K), Data availability triglycerides (Thermo Fisher Scientific #TR22421), cholesterol (Thermo Fisher There are no restrictions on data or material availability. The data that support the Scientific #TR13421), and free fatty acids (Wako Diagnostics #999-34691, #995- findings of this study are available from the authors on reasonable request. Raw RNA 34791, #991-34891, #993-35191) quantification were performed using commer- sequencing data files have been deposited into the Gene Expression Omnibus database cially available reagents according to manufacturer’s directions. Albumin levels (www.ncbi.nlm.nih.gov/geo/) with accession number GSE126134. were quantified using an AMS LIASYS Chemistry Analyzer. Hepatic lipids were extracted from ~100 mg hepatic tissue homogenized in 2:1 chloroform:methanol. In total, 0.25–0.5% of each extract was evaporated overnight Received: 20 May 2018 Accepted: 22 March 2019 prior to biochemical quantification of triglycerides, LDL-C, cholesterol, and free fatty acids using reagents described above, precisely according to manufacturer’s directions.

Body composition analysis. Body composition analysis was carried out in una- nesthetized mice using an EchoMRI 3-1 device (Echo Medical Systems) via the References Washington University Diabetic Mouse Models Phenotyping Core Facility. 1. González-Muniesa, P. et al. Obesity. Nat. Rev. Dis. Primers 3, 17034 (2017). 2. Longo, V. D. & Mattson, M. P. Fasting: molecular mechanisms and clinical – Indirect calorimetry and food intake measurement. All measurements were applications. Cell Metab. 19, 181 192 (2014). performed in a PhenoMaster System (TSE systems) via the Washington University 3. Martin, B., Mattson, M. P. & Maudsley, S. Caloric restriction and intermittent Diabetic Mouse Models Phenotyping Core Facility, which allowed metabolic per- fasting: two potential diets for successful brain aging. Ageing Res. Rev. 5, – formance measurement and activity monitoring by an infrared light = beam frame. 332 353 (2006). Mice were placed at room temperature (22–24 °C) in separate chambers of the 4. Ahmet, I., Wan, R., Mattson, M. P., Lakatta, E. G. & Talan, M. Cardioprotection PhenoMaster open-circuit calorimetry. Mice were allowed to acclimatize in the by intermittent fasting in rats. Circulation 112, 3115–3121 (2005). chambers for 4 h. Food and water were provided ad libitum in the appropriate 5. MATTSON, M. & Wan, R. Beneﬁcial effects of intermittent fasting and caloric devices. The parameters of indirect calorimetry (VO2, VCO2, respiratory exchange restriction on the cardiovascular and cerebrovascular systems. J. Nutr. ratio (RER), heat and movement) were measured for at least 24 h for a minimum of Biochem. 16, 129–137 (2005). one light cycle (6:01 am to 6:00 pm) and one dark cycle (6:01 pm to 6:00 am). 6. Patterson, R. E. et al. Intermittent fasting and human metabolic health. J. Presented data are average values obtained in these recordings. Acad. Nutr. Diet. 115, 1203–1212 (2018).

14 NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8 ARTICLE

7. Barnosky, A. R., Hoddy, K. K., Unterman, T. G. & Varady, K. A. Intermittent 37. Morris, S. M., Bhamidipati, D. & Kepka-Lenhart, D. Human type II arginase: fasting vs daily calorie restriction for type 2 diabetes prevention: a review of sequence analysis and tissue-specific expression. Gene 193, 157–161 (1997). human findings. Transl. Res. 164, 302–311 (2014). 38. Caldwell, R. B., Toque, H. A., Narayanan, S. P. & Caldwell, R. W. Arginase: an 8. Chung, H. et al. Time-restricted feeding improves insulin resistance and old enzyme with new tricks. Trends Pharmacol. Sci. 36, 395–405 (2015). hepatic steatosis in a mouse model of postmenopausal obesity. Metabolism 65, 39. Caldwell, R. W., Rodriguez, P. C., Toque, H. A., Narayanan, S. P. & Caldwell, 1743–1754 (2018). R. B. Arginase: a multifaceted enzyme important in health and disease. 9. Kosinski, C. & Jornayvaz, F. R. Effects of ketogenic diets on cardiovascular risk Physiol. Rev. 98, 641–665 (2018). factors: evidence from animal and human studies. Nutrients 9,1–16 (2017). 40. Morris, S. M. Jr Enzymes of arginine metabolism. J. Nutr. 134, 2743S–2747S 10. Kang, H. C., Chung, D. E., Kim, D. W. & Kim, H. D. Early- and late-onset (2004). complications of the ketogenic diet for intractable epilepsy. Epilepsia 45, 41. Morris, S. M. Jr Arginine metabolism revisited. J. Nutr. 146, 2579S–2586S (2016). 1116–1123 (2004). 42. Cederbaum, S. D. et al. Arginases I and II: do their functions overlap? Mol. 11. Steriade, C., Andrade, D. M., Faghfoury, H., Tarnopolsky, M. A. & Tai, P. Genet. Metab. 81,38–44 (2004). Mitochondrial encephalopathy with lactic acidosis and stroke-like episodes 43. Crombez, E. A. & Cederbaum, S. D. Hyperargininemia due to liver arginase (MELAS) may respond to adjunctive ketogenic diet. Pediatr. Neurol. 50, deficiency. Mol. Genet. Metab. 84, 243–251 (2005). 498–502 (2014). 44. Morris, S. M. Arginine metabolism: boundaries of our knowledge. J. Nutr. 137, 12. Potthoff, M. J., Kliewer, S. A. & Mangelsdorf, D. J. Endocrine fibroblast growth 1602S–1609S (2007). factors 15/19 and 21: From feast to famine. Genes Dev. 26, 312–324 (2012). 45. McAuley, K. A. et al. Intensive lifestyle changes are necessary to improve 13. Potthoff, M. J. FGF21 and metabolic disease in 2016: a new frontier in FGF21 insulin sensitivity. Diabetes. Care 25, 445 LP–445452 (2002). biology. Nat. Rev. Endocrinol. 13,74–76 (2016). 46. Iozzo, P. et al. Insulin-mediated hepatic glucose uptake is impaired in type 2 14. Potthoff, M. J. & Finck, B. N. Head over hepatocytes for FGF21. Diabetes 63, Diabetes: evidence for a relationship with glycemic control. J. Clin. Endocrinol. 4013–4015 (2014). Metab. 88, 2055–2060 (2003). 15. Alemán, G. et al. Increase in FGF21 stimulates browning markers in white 47. Honka, M.-J. et al. Insulin-stimulated glucose uptake in skeletal muscle, adipose tissue in rats fed a low protein high carbohydrate diet during acute adipose tissue and liver: a positron emission tomography study. Eur. J. cold exposure. FASEB. J. 31(652), 12–652.12 (2017). Endocrinol. 178, 523–531 (2018). 16. Gaich, G. et al. The effects of LY2405319, an FGF21 Analog, in obese human 48. Xiong, Y. et al. regulation of MTOR and PRKAA / AMPK signaling in subjects with type 2 diabetes. Cell Metab. 18, 333–340 (2013). advance datherosclerosis ARG2 impairs endothelial autophagy through 17. Colak, Y. et al. SIRT1 as a potential therapeutic target for treatment of regulation of MTOR and PRKAA/AMPK signaling in advanced nonalcoholic fatty liver disease. Med. Sci. Monit. 17, HY5–Y9 (2011). atherosclerosis, 8627 https://doi.org/10.4161/15548627.2014.981789 (2014). 18. Bordone, L. et al. SIRT1 transgenic mice show phenotypes resembling calorie 49. Liu, C. et al. Targeting arginase-II protects mice from high-fat-diet-induced restriction. Aging Cell 6, 759–767 (2007). hepatic steatosis through suppression of macrophage inflammation. Sci. Rep. 19. Badman, M. K. et al. Hepatic fibroblast growth factor 21 is regulated by 6, 20405 (2016). PPARa and is a key mediator of hepatic lipid metabolism in ketotic states. Cell 50. Ming, X.-F. et al. Arginase II promotes macrophage inflammatory responses Metab. 5, 426–437 (2007). through mitochondrial reactive oxygen species, contributing to insulin 20. Inagaki, T. et al. Endocrine regulation of the fasting response by PPARα- resistance and atherogenesis. Am. Heart J. 1, e000992–e000992 (2012). mediated induction of fibroblast growth factor 21. Cell Metab. 5, 415–425 51. Yang, Z. & Ming, X. F. Functions of arginase isoforms in macrophage (2007). inflammatory responses: Impact on cardiovascular diseases and metabolic 21. Nakamura, M. T., Yudell, B. E. & Loor, J. J. Regulation of energy metabolism disorders. Front. Immunol. 5,1–10 (2014). by long-chain fatty acids. Prog. Lipid. Res. 53, 124–144 (2014). 52. Navarro, L. A. et al. Arginase 2 deficiency results in spontaneous 22. Pastore, N. et al. TFE3 regulates whole‐body energy metabolism in steatohepatitis: a novel link between innate immune activation and hepatic de cooperation with TFEB. EMBO Mol. Med. 9, 605–621 (2017). novo lipogenesis. J. Hepatol. 62, 412–420 (2015). 23. Markan, K. R. et al. Circulating FGF21 is liver derived and enhances glucose 53. Pernow, J. & Jung, C. Arginase as a potential target in the treatment of uptake during refeeding and overfeeding. Diabetes 63, 4057–4063 (2014). cardiovascular disease: Reversal of arginine steal? Cardiovasc. Res. 98, 334–343 24. DeBosch, B. J. et al. Trehalose inhibits solute carrier 2A (SLC2A) proteins to (2013). induce autophagy and prevent hepatic steatosis. Sci. Signal. 9, ra21–ra21 54. Huang, J. et al. Feeding and fasting controls liver expression of a regulator of (2016). G protein signaling (Rgs16) in periportal hepatocytes. Comp. Hepatol. 5,1–11 25. DeBosch, B. J., Chen, Z., Saben, J. L., Finck, B. N. & Moley, K. H. Glucose (2006). transporter 8 (GLUT8) mediates fructose-induced de Novo lipogenesis and 55. Masri, S. et al. Partitioning circadian transcription by SIRT6 leads to macrosteatosis. J. Biol. Chem. 289, 10989–10998 (2014). segregated control of cellular metabolism. Cell 158, 659–672 (2014). 26. Mardones, P., Rubinsztein, D. C. & Hetz, C. Mystery solved: trehalose 56. Bassily, R. W., El-Sokkary, R. I., Silwanis, B. A., Nematalla, A. S. & Nashed, M. kickstarts autophagy by blocking glucose transport. Sci. Signal. 9, fs2 LP–fs2 A. An improved synthesis of 4-azido-4-deoxy- and 4-amino-4-deoxy-α, α- L2 (2016). trehalose and their epimers. Carbohydr. Res. 239, 197–207 (1993). 27. Mayer, A. L. et al. Enhanced hepatic PPARα activity links GLUT8 deficiency 57. Prinsen, H. C. M. T. et al. Rapid quantification of underivatized amino acids to augmented peripheral fasting responses in male mice. Endocrinology 159, in plasma by hydrophilic interaction liquid chromatography (HILIC) coupled 2110–2126 (2018). with tandem mass-spectrometry. J. Inherit. Metab. Dis. 39, 651–660 (2016). 28. DeBosch, B. J., Chen, Z., Finck, B. N., Chi, M. & Moley, K. H. Glucose transporter-8 (GLUT8) mediates glucose intolerance and dyslipidemia in high-fructose diet-fed male mice. Mol. Endocrinol. 27, 1887–1896 (2013). Acknowledgements 29. Zhang, Y. et al. Transcription factor EB (TFEB)-dependent induction of We thank Mark Graham (IONIS Pharmaceuticals, Carlsbad, CA) for providing Arg2- thermogenesis by the hepatocyte solute carrier 2A (SLC2A) inhibitor, specific antisense oligonucleotide. This work was supported by the Office of the Assistant trehalose. Autophagy. 0, 1 (2018). Secretary of Defense for Health Affairs, through the Peer Reviewed Medical Research 30. Zhang, Y. et al. TFEB-dependent induction of thermogenesis by the Program under Award No. W81XWH-17-1-0133. This work was also supported by hepatocyte SLC2A inhibitor trehalose. Autophagy. 14, 1959–1975 (2018). grants from the NIH/National Center for Advancing Translational Sciences (NCATS) 31. Mayer, A. L. et al. SLC2A8 (GLUT8) is a mammalian trehalose transporter grant #UL1TR002345, the NIH R56 (DK115764), the Children’s Discovery Institute (MI- required for trehalose-induced autophagy. Sci. Rep. 6, 38586 (2016). FR-2014-426 and MI-II2017-593), AGA-Gilead Sciences Research Scholar Award in 32. Higgins, C. B. et al. Hepatocyte ALOXE3 is induced during adaptive fasting Liver Disease, the Washington University Digestive Disease Research Core Center (BJD) and enhances insulin sensitivity by activating hepatic PPARγ. JCI Insight 3 (P30DK52574), the Washington University Diabetes Research Center (P30DK020579), https://doi.org/10.1172/jci.insight.120794 (2018). Nutrition & Obesity Research Center (P30DK056341), and the Robert Wood Johnson 33. Walmagh, M., Zhao, R. & Desmet, T. Trehalose analogues: latest insights in Foundation. B.M.S. was supported by National Institutes of Health (R15 AI117670). A.L. properties and biocatalytic production. Int. J. Mol. Sci. 16, 13729–13745 M. was supported by the Washington University Spencer T. Olin Fellowship, the (2015). Washington University NIGMS Institutional Training Grant in Cell and Molecular 34. O’Neill, M. K., Piligian, B. F., Olson, C. D., Woodruff, P. J. & Swarts, B. M. Biosciences (T32GM007067), and National Science Foundation Graduate Student Fel- Tailoring trehalose for biomedical and biotechnological applications. Pure lowship (DGE-1143954). Appl. Chem. 89, 1223–1249 (2017). 35. Pashkov, V. et al. Regulator of G protein signaling (Rgs16) inhibits hepatic fatty acid oxidation in a carbohydrate response element-binding protein Author contributions (ChREBP)-dependent manner. J. Biol. Chem. 286, 15116–15125 (2011). B.J.D. conceived and coordinated the study and wrote the paper. Y.Z., C.B.H., H.F., and 36. Perozich, J., Hempel, J., Morris Jr, S. M. Roles of conserved residues in the B.J.D. designed, performed and analyzed the experiments. A.L.M. analyzed the experi- arginase family. Biochim. Biophys. Acta. 1382,23–37 (1998). ments. B.M.S. and A.I.S. designed and analyzed the lactotrehalose experiments and

NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications 15 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09642-8

designed and optimized lactotrehalose synthesis. All authors reviewed the results and Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in approved the ﬁnal version of the manuscript. published maps and institutional afﬁliations.

Additional information Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- Open Access This article is licensed under a Creative Commons 019-09642-8. Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give Competing interests: A provisional patent application, “AAV-mediated hepatocyte appropriate credit to the original author(s) and the source, provide a link to the Creative Arginase II overexpression as a treatment against metabolic disease” is pending, on Commons license, and indicate if changes were made. The images or other third party which B.J.D. is the lead inventor. The data in Figs. 1–5 and 9–10 and Supplementary material in this article are included in the article’s Creative Commons license, unless Figs. 1, 2, and 5 are pertinent to the ﬁled application. The authors declare no competing indicated otherwise in a credit line to the material. If material is not included in the interests. article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from Reprints and permission information is available online at http://npg.nature.com/ the copyright holder. To view a copy of this license, visit http://creativecommons.org/ reprintsandpermissions/ licenses/by/4.0/.

Journal peer review information: Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports © The Author(s) 2019 are available.

16 NATURE COMMUNICATIONS | (2019) 10:1587 | https://doi.org/10.1038/s41467-019-09642-8 | www.nature.com/naturecommunications