Diabetes Volume 64, September 2015 3069

Basem M. Abdallah,1 Nicholas Ditzel,1 Jorge Laborda,2 Gerard Karsenty,3 and Moustapha Kassem1,4,5

DLK1 Regulates Whole-Body Glucose Metabolism: A Negative Feedback Regulation of the - Loop

Diabetes 2015;64:3069–3080 | DOI: 10.2337/db14-1642

The endocrine role of the skeleton in regulating energy through, in part, the osteocalcin (OCN). OCN metabolism is supported by a feed-forward loop between signals in b-cells through its bona fide receptor G- circulating osteoblast (OB)-derived undercarboxylated –coupled receptor (Gprc6a) to increase b-cell prolif- osteocalcin (Glu-OCN) and pancreatic b-cell insulin; in erationandinsulinsecretionandactsonperipheraltissues turn, insulin favors osteocalcin (OCN) bioactivity. These to increase energy expenditure (1-3). In turn, insulin sig- data suggest the existence of a negative regulation of this naling in osteoblasts (OBs) stimulates the activation of METABOLISM cross talk between OCN and insulin. Recently, we OCN by promoting its decarboxylation (Glu-OCN) through fi identi ed delta like-1 (DLK1) as an endocrine regulator the bone resorption arm of bone remodeling (1,4). The of bone turnover. Because DLK1 is colocalized with insulin physiological relevance of these findings have been sup- in pancreatic b-cells, we examined the role of DLK1 in ported through studies demonstrating skeleton as a site insulin signaling in OBs and energy metabolism. We show of insulin resistance in mice fed a high-fat diet (5). More- that Glu-OCN specifically stimulates Dlk1 expression by 2 2 over, patients with a dominant negative in the pancreas. Conversely, Dlk1-deficient (Dlk1 / )mice exhibited increased circulating Glu-OCN levels and in- Gprc6a show evidence of glucose intolerance (3). In all creased insulin sensitivity, whereas mice overexpressing likelihood, the Glu-OCN-insulin feed-forward loop must Dlk1 in OB displayed reduced insulin secretion and sen- be under a negative regulation to protect from hypoglyce- sitivity due to impaired insulin signaling in OB and low- mia. Soluble factors responsible for this regulation have 2 2 fi ered Glu-OCN serum levels. Furthermore, Dlk1 / mice not yet been identi ed despite the demonstrated ability treated with Glu-OC experienced significantly lower of transcription factors activating transcription factor 4 blood glucose levels than Glu-OCN–treated wild-type (ATF4) and forkhead box protein O1 (FoxO1) to regulate mice. The data suggest that Glu-OCN–controlled produc- glucose metabolism through a negative regulation of OCN tion of DLK1 by pancreatic b-cells acts as a negative feed- bioavailability (6,7). back mechanism to counteract the stimulatory effects of Delta like-1 (DLK1), also known as preadipocyte factor- insulin on OB production of Glu-OCN, a potential mecha- 1 (Pref-1), is a transmembrane protein belonging to the nism preventing OCN-induced hypoglycemia. Notch/serrate/delta family (8,9). The full ectodomain of DLK1 is proteolytically cleaved to generate a soluble active protein named fetal antigen-1 (FA1), which is secreted by A growing body of work indicates that bone is an endocrine cells of pancreas, ovary, Leydig cells of the tes- endocrine organ that regulates glucose metabolism tis, adrenal glands, and pituitary gland (10). DLK1 has

1Molecular Endocrinology Laboratory (KMEB), Department of Endocrinology, Odense Corresponding author: Basem M. Abdallah, [email protected]. University Hospital and University of Southern Denmark, Odense, Denmark Received 29 October 2014 and accepted 21 April 2015. 2Department of Inorganic and Organic Chemistry and Biochemistry, University of This article contains Supplementary Data online at http://diabetes Castilla–La Mancha Medical School, Albacete, Spain .diabetesjournals.org/lookup/suppl/doi:10.2337/db14-1642/-/DC1. 3Department of Genetics and Development, College of Physicians and Surgeons, Columbia University, New York, NY © 2015 by the American Diabetes Association. Readers may use this article as 4DanStem (Danish Stem Cell Center), Panum Institute, University of Copenhagen, long as the work is properly cited, the use is educational and not for profit, and Copenhagen, Denmark the work is not altered. 5Stem Cell Unit, Department of Anatomy, College of Medicine, King Saud University, Riyadh, Saudi Arabia 3070 Dlk1 Regulates Osteocalcin-Insulin Loop Diabetes Volume 64, September 2015 been shown to inhibit both adipogenesis (11,12) and were purified through Histopaque 1100 (120 mL 1119 osteoblastogenesis (13,14). In addition, DLK1 favors Histopaque + 100 mL 1077 Histopaque; Sigma-Aldrich) gra- bone resorption through a nuclear factor-kB–dependent dient centrifugation and cultured overnight in RPMI medium pathway (13). Consistent with these data, serum levels of (Gibco) supplemented with L-glutamine 10% FBS and peni- DLK1 were increased in -deficient postmeno- cillin (100 units/mL)/streptomycin (100 mg/mL) at 37°C. pausal women (15) and inversely correlated with total Mouse recombinant Glu-OCN was provided by G.K. bone mineral density (BMD) in patients with anorexia Conditioned medium (CM) containing sDLK1 protein was nervosa (16) or hypothalamic amenorrhea (17). collected from NIH3T3 mouse fibroblast cells cultured in Several lines of evidence suggest that DLK1 plays a role serum-free medium for 24 h. The expression plasmid in energy metabolism. For instance, mice overexpressing PHD184, containing the full-length human Dlk1 cDNA, soluble DLK1 (sDLK1) exhibit a marked reduction in was used (27). Mouse insulin signaling pathway RT2 Pro- white adipose tissue mass and impaired whole-body filer PCR array (catalog no. PAMM-030Z; QIAGEN) was glucose tolerance and insulin sensitivity (18,19). Further- used with the SYBR Green quantitative PCR method. more, increasing expression of Dlk1 has been shown to be Biochemical Assays associated with insulin resistance in diabetic Goto-Kakizaki ELISA measurements of (Millipore A/S), in- rat (20) and mice (21). Additionally, human studies have sulin (Mercordia), total serum OCN (Immutopics Interna- demonstrated changes in serum levels of FA1 in an ex- treme nutritional state (22) and during weight loss follow- tional), Gla and Glu-OCN (Takara), serum type 1 cross-linked C-telopeptide (CTX) (IDS Nordic, Helrev, ing bariatric surgery (23). Denmark), and sDLK1 (MyBioSource, Inc.) were used. On the basis of the inhibitory effects of DLK1 on bone remodeling and energy metabolism, we hypothesized that OB Differentiation DLK1 regulates glucose homeostasis by negatively regu- Cells were differentiated in a-minimum essential medium lating the OCN-insulin loop. To test this hypothesis, we (Gibco) containing 10% FBS, 100 units/mL penicillin, studied the effect of either loss or gain of Dlk1 function 100 mg/mL streptomycin, 50 mg/mL vitamin C (Sigma- on insulin signaling in OB and whole-glucose metabolism Aldrich), and 10 mmol/L b-glycerol-phosphate (Sigma-Aldrich) in mice. The data identify Dlk1 as a novel negative reg- in the presence or absence of 10 nmol/L insulin. ulator of energy metabolism through controlling OCN bioavailability. Alkaline Phosphatase and Alizarin Red Staining Cells were stained with naphthol AS-TR phosphate solution RESEARCH DESIGN AND METHODS containing Fast Red TR (Sigma-Aldrich) as described pre- viously (13). Alkaline phosphatase activity was measured Animals using p-nitrophenyl phosphate (Fluka Chemie) as substrate All experimental procedures were approved by the Danish 2/2 (28). Cells were stained with 40 mmol/L Alizarin Red S Animal Ethical committee. Dlk1-deficient (Dlk1 ) mice (Sigma-Aldrich), pH 4.2, for 10 min at room temperature were obtained from J. Laborda (University of Castilla–La as previously described (13). Mancha, Ciudad Real, Spain) (24). Osteoblast-specific Dlk1-overexpressing mice (expressing Dlk1 under collagen RNA Extraction and Real-Time PCR Analysis 3.6-kb promoter Col1-Dlk1) with high circulating levels of RNA was extracted using TRIzol (Invitrogen). cDNA was sDLK1 were generated by our group (13). Mice were bred synthesized using a RevertAid H Minus First Strand cDNA and housed under standard conditions (21°C, 55% rela- Synthesis Kit (Fermentas). Quantitative real-time PCR was tive humidity) on a 12-h light, 12-h dark cycle. Ad libitum performed with an Applied Biosystems 7500 Real-Time food (Altromin) and water were provided. PCR System using Fast SYBR Green Master Mix (Applied For the effect of Glu-OCN on glucose metabolism in Biosystems) with specific primers. After normalization to 2/2 vivo, 12-week-old wild-type (WT) and Dlk1 mice (n = b-actin mRNA levels, a relative expression level of each 6/group) were implanted subcutaneously with osmotic target was calculated by a comparative CT method D T pumps (Alzet, Karlslunde, Denmark) containing Glu- [(1 / (2 C ), where DCT is the difference between CT target OCN (0.3 ng/h delivery) or vehicle for a period of 28 days. and CT reference] using Microsoft Excel 2007 software.

Cell Cultures and Reagents Western Blot Assays Clonal insulin-secreting INS-1E cells were cultured as Forty micrograms of protein were separated on 8–12% described previously (25). Primary osteoprogenitors (OBs) NuPAGE Novex Bis-Tris gels (Invitrogen) followed by were isolated from the calvarias of neonatal (3–4-day-old) transfer to a polyvinylidene fluoride membrane (Millipore mice and cultured as described previously (13). Primary A/S). Antibodies against the , total or islets were isolated and cultured from 12-week-old mice Ser-473 phosphorylated AKT (p-AKT), insulin-like growth as previously described (26). In brief, pancreata were in- factor 1 receptor, and p-38 were obtained from Cell Signal- fused with 3–4 mL collagenase P solution (Roche) in ing Technology (Herlev, Denmark). Anti-DLK1 and IRS-1 Hanks’ balanced salt solution (Invitrogen) (13 supple- were from Millipore. Anti-phospho-ERK1/2, anti-ERK2 mented 0.35 g NaHCO3/L, pH 7.4, and 1% BSA). Islets (C-14, sc-154), and anti-b-actin were purchased from Santa diabetes.diabetesjournals.org Abdallah and Associates 3071

Cruz Biotechnology, Inc. (Aarhus, Denmark). Quantification 2 or 20 mmol/L glucose with various dilutions of sDLK1-CM of Western blots was performed with ImageJ software. for 30 min at 37°C with shaking. INS-1E cells were stimu- Metabolic Studies lated in 24-well culture plates, and insulin released into the medium was measured by ELISA and normalized to the pro- Glucose Metabolic Studies tein content measured by Bradford protein assay. Glucose tolerance test (GTT) and insulin tolerance test (ITT) were performed on 10- to 12-week-old mice. For DEXA and Microcomputed Tomography Scanning 2 GTT, overnight-fasted mice were injected with D-glucose Fat mass (g), bone mineral content (g), and BMD (g/cm ), 2g/kgi.p.,andglucoselevelswere measured using an Accu- were measured using DEXA PIXImus2 version 1.44 (Lunar Chek glucometer (Roche Diagnostics Corp., Indianapolis, Corporation, Madison, WI) as described previously (13). IN). For ITT, 5-h–fasted mice were injected with insulin The tibiae of 2-month-old mice were scanned using 0.5 units/kg i.p. (Eli Lilly and Company, Indianapolis, IN). a high-resolution microcomputed tomography (micro- For glucose-stimulated insulin secretion (GSIS), overnight- CT) system (vivaCT 40; SCANCO Medical, Bassersdorf, fasted mice were injected with glucose 2 g/kg i.p., and Switzerland) as described previously (29). serum insulin was measured using mouse ultrasensitive insulin ELISA (ALPCO). Bone Dynamic Histomorphometry Mice were injected with calcein 30 mg/kg (Fluka Chemie) Insulin Secretion Measurements at 9 and 2 days, respectively, before necropsy. ImageJ Cultured mouse islets were washed in Krebs-Ringer bi- version 1.45s analysis software was used to measure carbonate (KRB) buffer (135 mmol/L NaCl, 3.6 mmol/L mineral apposition rate (MAR) (in mm/day), mineralizing KCl, 5 mmol/L NaHCO3, 0.5 mmol/L NaH2PO4,0.5mmol/L surface per bone surface (MS/BS), and bone formation m 2 m MgCl2,1.5mmol/LCaCl2, 10 mmol/L HEPES [pH 7.4], and rate (BFR) (in m / m/day) in the frozen sections of 0.1% BSA). Islets were incubated in KRB buffer containing tibia as previously described (30).

Figure 1—Recombinant Glu-OCN stimulates Dlk1 expression by pancreatic islet cells in vitro and in vivo. A: Real-time PCR analysis of Dlk1 expression in the insulinoma INS-1E (i), preadipocyte 3T3-L1 (ii), and fibroblast NIH3T3 (iii) cell lines treated with increasing concentrations of Glu-OCN (0.01–30 ng/mL) for 4 h. B: Stimulatory effect of Glu-OCN on sDLK1 secretion by primary mouse islets. Mouse islets were isolated and cultured as described in RESEARCH DESIGN AND METHODS and treated with vehicle (control) or increasing concentrations of Glu-OCN for 4 h. sDLK1 released in the media was measured by ELISA, and values were normalized to cellular protein content. C: In vivo effect of Glu-OCN on pancreatic Dlk1 expression. Glu-OCN (1 mg/kg) or PBS (vehicle, control) was injected intraperitoneally in 2-month-old female WT mice (n =4–5/group). D: Four hours after Glu-OCN injection, serum sDLK1 was measured by ELISA, and Dlk1 gene expression was quantified in selected tissues by quantitative real-time RT-PCR. Data are mean 6 SEM of three independent experiments. *P < 0.05, **P < 0.005 compared with control noninduced. WAT, white adipose tissue (inguinal fat). 3072 Dlk1 Regulates Osteocalcin-Insulin Loop Diabetes Volume 64, September 2015

Statistical Analysis mRNA expression was not detectable in mouse 3T3-L1 All values are expressed as mean 6 SEM. Comparisons be- preadipocytes or NIH3T3 fibroblasts (Fig. 1A, ii and iii). tween groups were performed using unpaired Student t test To determine the specificity of Glu-OCN action on Dlk1 (two-tailed). P , 0.05 was considered statistically significant. production by b-cells in vivo, we injected WT mice intra- peritoneally with either Glu-OCN (1 mg/kg) or vehicle as RESULTS described previously (31) and measured circulating sDLK1 Glu-OCN Stimulates Dlk1 Expression by b-Cells In levels as well as the expression of Dlk1 mRNA 4 h later. Vitro and In Vivo This experiment showed that Glu-OCN significantly b While looking for expressed in pancreatic -cells by increased serum sDLK1 levels (Fig. 1C) because of its Dlk1 Glu-OCN, we examined the possible regulation of stimulatory effect on Dlk1 expression in pancreas by b expression by Glu-OCN in -cells. Treatment of the .2.5-fold compared with controls and no other endocrine b Dlk1 -cell line INS-1E with Glu-OCN stimulated mRNA organs (Fig. 1D). expression in a dose-dependent manner (Fig. 1A,i).Further- more, Glu-OCN stimulated sDLK1 secretion by cultured DLK1 Inhibits Insulin-Induced OB Differentiation primary mouse pancreatic islets in a dose-dependent man- We then asked whether DLK1 affects insulin signaling ner (Fig. 1B). On the other hand, Glu-OCN–induced Dlk1 in OB. As shown in Fig. 2A, insulin enhanced OB

Figure 2—DLK1 inhibits insulin signaling in OB. A: Real-time PCR analysis of osteogenic markers in WT-OBs treated with OB induction medium in the presence or absence of 10 nmol/L insulin for 7 days. B: Western blot analysis of the expression of insulin-related during long-term differentiation into OB lineage in the presence and absence of insulin. C: Real-time PCR analysis of insulin-induced OB markers Ocn, Alp, Runx2, and Col1a1 in Col1-Dlk1 OBs and Dlk12/2 OB compared with WT-OB in the presence and absence of 10 nmol/L insulin. Alizarin Red staining is shown. D: Western blot analysis of insulin signaling in Col1-Dlk1 OBs and Dlk12/2 OBs compared with WT- OBs. Relative protein levels of p-AKT are represented as fold change to control after normalization to total AKT (T-AKT) protein levels. E: Real-time PCR analysis and Western blot analysis of INSR protein at baseline. F: Annotation analysis of downregulated insulin target by Col1-Dlk1 OBs compared with WT-OBs upon insulin (10 nmol/L) treatment for 12 h in serum-free medium. Genes downregulated by twofold or less in Col1-Dlk1 OBs were annotated according to their gene function and presented as a percent of the total downregulated genes. G: Real-time PCR analysis of Foxo1 expression in Col1-Dlk1 OBs and Dlk12/2 OBs compared with WT-OBs. Expression was normalized to b-actin expression levels and represents percent induction over noninduced control cells. Data are mean 6 SEM of three independent experiments. *P < 0.05, **P < 0.005 vs. WT-OB. diabetes.diabetesjournals.org Abdallah and Associates 3073 differentiation of WT OBs (WT-OB) as assessed by the In addition, we observed that 70% of differentially upregulation of the OB markers Runx2, type I collagen upregulated genes in response to insulin were down- (Col1a1), Ocn, and Alp, and this stimulatory effect of in- regulated in Col1-Dlk1 OBs, including the insulin target sulin was additive to OB induction medium. Additionally, genes Cebpb, Adra1d, and Dusp14 and the insulin signaling insulin treatment together with OB induction medium genes Irs1, Insl3, Ptpn1, and Gsk3b as assessed by insulin over 6 days increased the protein levels of INSR1, IRS1, signaling pathway PCR array analysis (Fig. 2F and Supple- and p-AKT compared with control cells treated with OB mentary Table 1). induction medium alone (Fig. 2B). Insulin-induced OB We observed that DLK1-impaired insulin signaling in differentiation was significantly reduced in OBs isolated Col1-Dlk1 OBs was associated with a significant upregula- from Col1-Dlk1 mice (Col1-Dlk1 OBs) (13), as shown by tion of FoxO1 (Fig. 2G), a downstream target of insulin decreased expression of all tested OB markers and poor that negatively regulates insulin signaling in OBs (6), 2/2 formation of mineralized matrix visualized by Alizarin whereas FoxO1 was downregulated in Dlk1 OBs. 2 2 Red staining compared with WT-OB controls (Fig. 2C). Transient transfection of Dlk1 / OBs with Dlk1 cDNA 2/2 On the other hand, OBs isolated from Dlk1 mice plasmid (Supplementary Fig. 1A) reproduced the data 2/2 (Dlk1 OBs) exhibited a higher expression of Alp, Ocn, obtained from Col1-Dlk1 OBs, including the inhibition and Runx2 than WT-OBs (Fig. 2C). of insulin-induced AKT phosphorylation (Supplementary Fig. 1B) and the impairment of insulin-induced OB differ- DLK1 Inhibits Insulin Signaling in OB entiation (Supplementary Fig. 1C and D). In addition, As shown in Fig. 2D, the insulin-induced phosphorylation treatment of WT-OBs with sDLK1 inhibited insulin- of AKT Ser-473 was impaired in Col1-Dlk1 OBs and en- induced p-AKT in a paracrine fashion (Fig. 3A). Thus, 2 2 hanced in Dlk1 / OBs compared with WT-OBs. On the these data identify DLK1 as an autocrine/paracrine antag- other hand, insulin receptor (Insr) mRNA and protein onist of insulin signaling in OBs. accumulation were not affected by Dlk1 expression in We also examined the effect of sDLK1 on insulin OBs (Fig. 2E), suggesting that DLK1 regulates insulin sig- secretion by isolated mouse islets and the b-cell line INS- naling in OBs downstream of Insr. 1E under low- and high-glucose stimulatory conditions.

Figure 3—DLK1 inhibits insulin-induced OCN production and carboxylation. A: Effect of sDLK1 on insulin-induced AKT phosphorylation. Insulin-induced AKT phosphorylation in WT-OB cells was treated with either control-CM or sDLK1-CM (50% dilution) and visualized by Western blot analysis. B: Effect of sDLK1 on insulin secretion by primary isolated mouse islets. Islets were incubated for 30 min at 37°C in KRB buffer with 2 or 20 mmol/L glucose in the presence of control-CM or various dilutions of sDlk1-CM. Insulin secretion in CM was determined by ELISA and normalized to cellular protein content. C and D: Effect of sDLK1-CM on Ocn gene expression by WT-OBs as measured by real-time PCR analysis as well as on total OCN secreted protein in the culture medium as measured by ELISA. Cells were cultured in OB induction medium and treated with various dilutions of sDLK1-CM for 24 h. E and F: ELISA measurements of total OCN and Gla-OCN and Glu-OCN in serum from 2-month-old Col1-Dlk1 and Dlk12/2 mice and their WT littermate controls (n = 8 mice/group). Data are mean 6 SEM of three independent experiments. *P < 0.05, **P < 0.005. 3074 Dlk1 Regulates Osteocalcin-Insulin Loop Diabetes Volume 64, September 2015 sDLK1-CM at various dilutions did not affect the secre- F). Taken together, DLK1 reduced OCN production by OB, tion of insulin by pancreatic islets (Fig. 3B) or INS-1E cells leaving insulin secretion by b-cells unaffected. A (Supplementary Fig. 2 ). In addition, the expressions of DLK1 Negatively Regulates Glucose Metabolism the insulin genes Ins1 and Ins2 and the cell cycle gene Next, we performed metabolic studies to examine the Cdk2 were not affected in INS-1E cells upon sDLK1-CM biological consequences of impaired OB insulin signaling stimulation (Supplementary Fig. 2B). and reduced serum Glu-OCN in Col1-Dlk1 mice on whole- DLK1 Inhibits OCN Bioactivity body glucose metabolism. Both fasted and fed blood glu- In WT-OBs, sDLK1 inhibited Ocn expression (Fig. 3C)as cose levels were significantly increased in Col1-Dlk1 mice well as the secretion of OCN in the CM (Fig. 3D) in a dose- by 48.2% and 33.8%, respectively, compared with WT dependent fashion. We also studied the role of DLK1 in littermates (Fig. 4A). The hyperglycemia observed in regulating OCN activity in vivo, measuring the total cir- Col1-Dlk1 mice was associated with a 46% reduction in culating OCN as well as the ratio of Glu/Gla OCN in the insulin levels (Fig. 4B). GTT (Fig. 4C) revealed that Col1- 2/2 serum of Col1-Dlk1 and Dlk1 mice. Of note, Col1-Dlk1 Dlk1 mice display impaired glucose tolerance with a higher mice displayed 36.3% and 43.7% reduction in total OCN initial rise in blood glucose and slower glucose clearance and active Glu-OCN serum levels, respectively, compared rate, whereas ITT revealed reduced insulin sensitivity in 2/2 with WT controls, whereas in Dlk1 mice, we observed Col1-Dlk1 mice compared with WT controls (Fig. 4D). a significant increase in the serum levels of total OCN and The impaired insulin sensitivity of Col1-Dlk1 mice was Glu-OCN by 19.8% and 48.1%, respectively (Fig. 3E and associated with a 45.6% reduction in serum levels of

Figure 4—DLK1 expression in OB inhibits whole-body glucose metabolism. A: Blood glucose levels at fed and fasted conditions in Col1- Dlk1 and WT mice. B: Serum insulin levels in Col1-Dlk1 and WT mice. C and D: GTTs and ITTs in 2-month-old Col1-Dlk1 and WT mice. E: Adiponectin serum levels in Col1-Dlk1 and WT mice. F: GSIS test measuring serum insulin stimulation after glucose injection in Col1-Dlk1 and WT mice. G: Real-time PCR analysis of Ins1 and Ins2 expression in pancreas from Col1-Dlk1 and WT mice. H: Histological analysis of Col1-Dlk1 and WT islets showing hematoxylin-eosin (H&E) staining and double immunostaining for insulin/Ki67. Scale bar = 100 mm. I and J: Percentage of b-cell area and Ki67-proliferating b-cells in Col1-Dlk1 mice. K: Real-time PCR analysis of insulin target genes in white fat from Col1-Dlk1 and WT mice. Data are mean 6 SEM (n =5–7 mice/group). *P < 0.05, **P < 0.005 vs. WT mice. +ve, positive. diabetes.diabetesjournals.org Abdallah and Associates 3075 adiponectin, a hormone that also regulates bone remodel- mice were reduced by 28.4% and 36.4%, respectively, com- ing (Fig. 4E) (32,33). We show a significant reduction in pared with WT mice. Fed insulin serum level was increased 2/2 insulin levels after glucose injection, indicating that insulin by 48.3% in Dlk1 mice compared with WT controls (Fig. secretion was impaired in mice overexpressing Dlk1 in OB 5B). GTT revealed a significant reduction in blood glucose (Fig. 4F). Accordingly, Col1-Dlk1 mouse islets exhibited re- levels compared with WT controls (Fig. 5C), and ITT 2 2 duced insulin expression (Ins1 and Ins2 genes) (Fig. 4G) showed that insulin sensitivity increased in Dlk1 / mice and a significant reduction in b-cell area and proliferation (Fig. 5D). A 34% increase in adiponectin serum levels com- (by 37% and 65.3%, respectively) compared with WT islets pared with WT mice was noted (Fig. 5E). In contrast to (Fig. 4H–J). Finally, expression of the insulin target genes Col1-Dlk1 mice, a GSIS test showed a significant increase 2/2 Cebpa, Pparg2, aP2,andFas was significantly reduced in in insulin stimulation by glucose in Dlk1 mice (Fig. 5F). white fat of Col1-Dlk1 mice compared with WT controls In addition, Dlk1 deficiency resulted in a significant upre- (Fig. 4K). These data demonstrate that DLK1, through its gulation of Ins1 and Ins2 gene expression and a significant expression in OB, negatively regulates insulin sensitivity increase in size and proliferation of pancreatic b-cells (by and secretion in mice. 46.4% and 71.7%, respectively) compared with WT controls 2/2 (Fig. 5G–J). Accordingly, the expression of insulin target Increased Insulin Secretion and Sensitivity in Dlk1 fi Dlk12/2 K Mice genes was signi cantly increased in fat (Fig. 5 ). Because Dlk1 is not an OB-specific gene (9), we investi- DLK1 Is a Negative Regulator of OCN-Insulin Feed- gated the effect of DLK1 loss of function on energy me- Forward Loop 2/2 tabolism using general Dlk1 mice (24). As shown in To examine whether DLK1 modulates the effect of OCN 2 2 Fig. 5A, fasted and fed blood glucose levels in adult Dlk1 / on glucose metabolism, we studied the effect of Glu-OCN

Figure 5—Loss of function of DLK1 improves glucose sensitivity and secretion. A: Blood glucose levels in Dlk12/2 and WT newborn pups before milk ingestion and 2-month-old mice at fed and fasted conditions. B: Serum insulin levels in Dlk12/2 and WT mice. C and D: GTT and ITT in 2-month-old Dlk12/2 and WT mice. E: Adiponectin serum levels in Dlk12/2 and WT mice. F: GSIS test. G: Real-time PCR analysis of Ins1 and Ins2 expression in pancreas from Dlk12/2 or WT mice. H: Hematoxylin-eosin (H&E) staining and double immunostaining for insulin/ Ki67 on Dlk12/2 or WT pancreatic islet sections. Scale bar = 100 mm. I and J: b-Cell area and Ki67-positive b-cells in islets from Dlk12/2 mice vs. WT controls. K: Real-time PCR analysis of insulin target genes in Dlk12/2 and WT white fat. Data are mean 6 SEM (n =6–8 mice/ group). *P < 0.05, **P < 0.005 vs. WT mice. +ve, positive. 3076 Dlk1 Regulates Osteocalcin-Insulin Loop Diabetes Volume 64, September 2015 on glucose metabolism in mice lacking Dlk1. For that (35). Total BMD (Fig. 7B) and micro-CT analysis of both 2/2 purpose, we implanted WT and Dlk1 mice with os- trabecular and cortical bone of the proximal tibia did not 2 2 motic pumps delivering either Glu-OCN (0.3 ng/h) or reveal significant differences between Dlk1 / and WT vehicle for 28 days. As reported previously (34), the mice (Fig. 7C–E). No histological changes were observed 2 2 data show that WT mice infused with Glu-OCN were hy- in the tibial growth plate between Dlk1 / mice and their poglycemic due to increased insulin sensitivity and secre- WT controls (Fig. 7F) and both MAR and BFR were com- 2/2 tion compared with WT mice infused with vehicle (Fig. parable between Dlk1 and WT mice (Fig. 7G). In addi- 2/2 6A–E). Of note, Dlk1 mice implanted with pumps de- tion, the osteoclastic bone resorption was not affected as livering Glu-OCN displayed significantly lower blood glu- evidenced by absence of significant changes in serum CTX cose levels and increased insulin levels, glucose clearance compared with WT controls (Fig. 7H). rate (GTT), insulin sensitivity (ITT), and GSIS compared 2/2 with either WT mice infused with Glu-OCN or Dlk1 DISCUSSION A–E mice infused with vehicle (Fig. 6 ). Thus, these data In this study, we show that DLK1 acts as a negative regula- – demonstrate that DLK1 protects against Glu-OCN induced tor of the OCN-insulin feed-forward loop, thus revealing hypoglycemia. a new control mechanism protecting from OCN-induced Loss of Dlk1 Function Does Not Affect Bone hypoglycemia. In this negative feedback loop, OB-secreted Remodeling in Mice Glu-OCN stimulates the production of DLK1 by b-cells, To determine whether the metabolic effects of DLK1 are which inhibits insulin signaling in OB and, consequently, caused secondary to changes in skeletal turnover, we regulates the bioavailability of active Glu-OCN (Fig. 8). 2 2 studied the skeletal phenotype of Dlk1 / mice. As DLK1 has already been implicated in many aspects of 2/2 reported previously (24) and shown in Fig. 7A, Dlk1 energy metabolism, starting with its role as an inhibitor of embryos were smaller during development and postnatally adipogenesis in vitro and in vivo (35) and its association

Figure 6—DLK1 antagonizes Glu-OCN–induced hypoglycemia. WT and Dlk12/2 mice were implanted with osmotic pumps delivering vehicle or Glu-OCN (0.3 ng/h) over a period of 28 days. Glucose metabolic studies were performed at day 21. A and B: Blood glucose and serum insulin levels for the fed condition. C:GTT.D:ITT.E: GSIS. Data are mean 6 SEM (n = 6 mice/group). *P < 0.05, **P < 0.005 (Dlk12/2-Glu-OCN vs. Dlk12/2-vehicle); #P < 0.05, ##P < 0.005 (Dlk12/2-Glu-OCN vs. WT-Glu-OCN). diabetes.diabetesjournals.org Abdallah and Associates 3077

Figure 7—Dlk12/2 mice display a normal bone mass. A: Whole-mount staining for bone and cartilage in E17.5 Dlk12/2 and WT embryos. Dlk12/2 embryos showed reduced size during development. B: Total body weight, as measured gravimetrically, and bone mineral content (BMC), total BMD, and fat and lean mass measured using PIXImus2 in 2-month-old Dlk12/2 mice and their WT littermates. C: Three-dimensional micro-CT image reconstruction with median values of distal femur and proximal tibia from Dlk12/2 mice and WT controls. Micro-CT analysis of trabecular (D) and cortical (E) bone parameters in the proximal tibia of 2-month-old Dlk12/2 and WT mice. Trabecular bone parameters are bone volume/total volume (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), and connectivity (CD). Cortical bone parameters are BV/TV, cortical thickness (Ct.Th), bone surface/bone volume (BS/BV), and density bone volume (Density.BV). F: Histological sections of tibia bone from WT and Dlk12/2 mice stained with Alcian Blue showing the thickness of the growth plate. G:Dynamic histomorphometrics of proximal tibial metaphysis after fluorescent imaging microscopy. MAR, MS/BS, and BFR were comparable between Dlk12/2 and WT mice. H: Serum CTX levels. Data are mean 6 SEM (n =6/group).**P < 0.005 vs. WT mice. d, day; gm, gram.

with insulin resistance in both rodents and humans receptor to inhibit insulin-stimulated AKT phosphoryla- (19,20,36). The current study uncovers a new mechanism tion of FoxO1. Regulation of the AKT-FoxO1 signal by by which DLK1 links bone and energy metabolism. DLK1 is supported further by the increased phosphory- We previously demonstrated that Dlk1 expression and lation of AKT and the reduced expression of FoxO1 in 2 2 secretion in b-cells are stimulated in vitro and in vivo by Dlk1 / OBs. Suppression of AKT activation appears to recombinant Glu-OCN, an inducer of insulin expression be a common mechanism used by DLK1 to inhibit insulin by b-cells (34). Considering that DLK1 is colocalized with signaling in other biological processes. Indeed, it has been insulin in adult b-cells (37,38) and that the secretion of demonstrated in the inhibition of insulin-induced chon- DLK1 and insulin have been reported to be stimulated by drogenesis in the mouse cell line ATDC5 (40) and in re- the same (including and pro- ducing insulin-stimulated glucose uptake in skeletal lactin) (37,39), it is plausible that OCN uses a similar muscles in vivo in Dlk1-overexpressing mice (19). On mechanism to stimulate the secretion of sDLK1 and in- the other hand, the biological activity of OCN is nega- sulin. In this context, it is important to note that OCN tively regulated by the OB-expressed gene Esp (embryonic favors the receptor Gprc6a in insulin secretion in b-cells stem cell phosphatase), encoding for a protein tyrosine (31). Thus, it is plausible that the effect of OCN on DLK1 phosphatase (OST-PTP) that decreases OCN bioactivity by secretion by b-cells is also mediated through Gprc6a, but inhibiting insulin signaling in OBs (2). Despite that the more experiments are needed to prove this point. two recently identified negative regulators of Glu-OCN The data reveal that DLK1 negatively regulates OCN production ATF4 and FoxO1 were reported to function bioactivity by acting downstream from the insulin through an Esp-dependent regulatory mechanism (6,7), 3078 Dlk1 Regulates Osteocalcin-Insulin Loop Diabetes Volume 64, September 2015

sensitivity in an OB-dependent manner. Increased circu- lating levels of sDLK1 in transgenic mice overexpressing DLK1 in fat under an aP2 promoter (aP2-Dlk1) was pre- viously demonstrated to induce insulin resistance due to impaired insulin signaling and reduced insulin-induced glucose uptake in muscle and adipose tissue without af- fecting insulin secretion by b-cells (18,19). We therefore do not exclude the possibility of a contribution by other insulin target tissues, including fat and muscle, in the development of insulin resistance in Col1-Dlk1 mice. How- ever, the reported reduced insulin secretion by b-cells in the current Col1-Dlk1 mice but not in aP2-Dlk1 mice, de- spite high serum levels of sDLK1, supports the specific action of DLK1 on insulin secretion by b-cells through its function in OB to regulate Glu-OCN. We show that sDLK1 did not exert a regulatory effect on insulin production by b-cells, thus excluding the pos- sible endocrine function of sDLK1 in controlling insulin production by islets in the current Col1-Dlk1 mice. In addition, we show that OCN-induced hypoglycemia was 2/2 significantly pronounced in Dlk1 mice infused with Glu-OCN compared with WT controls. To our knowledge, this report is the first to demonstrate that general Dlk1- Figure 8—Proposed model of DLK1 action in regulating the OCN- null mice display increased insulin secretion by b-cells and insulin feed-forward loop. OB-secreted Glu-OCN stimulates DLK1 enhanced insulin sensitivity through an OCN-dependent production by islet b-cells. DLK1 exerts a negative feedback mech- anism that impairs insulin signaling–induced OCN production by mechanism. Thus, DLK1 affects insulin secretion by OB, thus antagonizing Glu-OCN–induced hypoglycemia. P, phos- b-cells through an OB-dependent mechanism, whereas it phorylation. regulates insulin sensitivity in an endocrine fashion. The current findings provide a mechanistic explanation for the observed association between increased levels of DLK1 and impaired insulin sensitivity in diabetic mice (21) and rats (20), in obese patients (46), and in patients with type 2 dia- we could not detect any changes in Esp expression in OBs betes (36). Although our studies are conducted in murine 2/2 or bone tissue derived from Col1-Dlk1 or Dlk1 mice models, these findings may be relevant to normal human (data not shown). Thus, the current data suggest that physiology. Despite that in many physiological situations find- DLK1 is a novel class of OCN regulator acting through ings in mice predict normal human physiology, some of the an Esp-independent mechanism. human data related to the role of OCN in glucose homeostasis A growing body of evidence supports the function of seem to be at variance with the murine data. For example, DLK1 as a noncanonical NOTCH receptor ligand that reduced levels of Glu-OCN by antiresorptive therapies in regulates Notch signaling (41–43). In this regard, it is humans do not cause significant changes in glucose metabo- worth mentioning that Notch signaling has been shown lism. Reduced Gla and Glu forms of OCN by bisphosphonate to be involved in insulin sensitivity. Genetic or pharma- treatment for osteoporosis are not associated with insulin cologic inhibition of hepatic Notch signaling in obese mice secretion or resistance (47). Additionally, the antiresorptive simultaneously improves glucose tolerance and reduces therapy did not affect the risk for developing diabetes in three hepatic triglyceride content (44). In addition, Notch sig- randomized placebo-controlled trials in postmenopausal naling appears to be involved in the development of in- women (48). On the other hand, the association between sulin resistance through a FoxO1-dependent mechanism serum increased sDLK1 and the insulin resistance phenotype (45). The current data indicate that FoxO1 expression is has been reported in rodent (20,21) and human studies (46) modulated by DLK1, thus linking its activity to a potential as well as in patients with type 2 diabetes (36,49). Thus, modulation of NOTCH signaling in OBs. Nevertheless, follow-up studies are needed to corroborate the relevance of more studies are needed to test this possibility. changes in serum sDLK1 to Glu-OCN regulation of glucose To study the involvement of DLK1 in regulating the metabolism in humans. endocrine function of bone in vivo, we compared the glucose metabolism phenotype of Col1-Dlk1 mice and 2/2 Dlk1 mice. Our metabolic studies in Col1-Dlk1 mice Acknowledgments. The authors thank Bianca Jørgensen and Lone revealed the role of DLK1 in regulating glucose metabo- Christiansen (Odense University Hospital, Odense, Denmark) for excellent tech- lism by controlling both insulin secretion and insulin nical assistance. diabetes.diabetesjournals.org Abdallah and Associates 3079

Duality of Interest. No potential conflicts of interest relevant to this article factor-1 (Pref-1): a new model of partial lipodystrophy. Diabetes 2008;57:3258– were reported. 3266 Author Contributions. B.M.A. contributed to designing and performing 20. Barbu A, Hedlund GP, Lind J, Carlsson C. Pref-1 and adipokine expression the experiments and wrote the manuscript. N.D. contributed to performing the in adipose tissues of GK and Zucker rats. Mol Cell Endocrinol 2009;299:163–171 experiments. J.L., G.K., and M.K. contributed to designing the experiments and to 21. Shimomura I, Shimano H, Korn BS, Bashmakov Y, Horton JD. Nuclear sterol the discussion, review, and editing of the manuscript. B.M.A. is the guarantor of regulatory element-binding proteins activate genes responsible for the entire this work and, as such, had full access to all the data in the study and takes program of unsaturated fatty acid biosynthesis in transgenic mouse liver. J Biol responsibility for the integrity of the data and the accuracy of the data analysis. Chem 1998;273:35299–35306 22. Andries A, Niemeier A, Støving RK, et al. Serum levels of fetal antigen 1 in References extreme nutritional states. ISRN Endocrinol 2012;2012:592648 1. Ferron M, Wei J, Yoshizawa T, et al. Insulin signaling in osteoblasts in- 23. O’Connell J, Lynch L, Hogan A, Cawood TJ, O’Shea D. Preadipocyte factor-1 tegrates bone remodeling and energy metabolism. Cell 2010;142:296–308 is associated with metabolic profile in severe obesity. J Clin Endocrinol Metab 2. Lee NK, Sowa H, Hinoi E, et al. Endocrine regulation of energy metabolism 2011;96:E680–E684 by the skeleton. Cell 2007;130:456–469 24. Raghunandan R, Ruiz-Hidalgo M, Jia Y, et al. Dlk1 influences differentiation 3. Oury F, Ferron M, Huizhen W, et al. Osteocalcin regulates murine and human and function of B lymphocytes. Stem Cells Dev 2008;17:495–507 fertility through a pancreas-bone-testis axis. J Clin Invest 2013;123:2421–2433 25. Merglen A, Theander S, Rubi B, Chaffard G, Wollheim CB, Maechler P. 4. Fulzele K, Riddle RC, DiGirolamo DJ, et al. Insulin receptor signaling in Glucose sensitivity and metabolism-secretion coupling studied during two-year osteoblasts regulates postnatal bone acquisition and body composition. Cell continuous culture in INS-1E insulinoma cells. Endocrinology 2004;145:667–678 2010;142:309–319 26. Carter JD, Dula SB, Corbin KL, Wu R, Nunemaker CS. A practical guide to 5. Wei J, Ferron M, Clarke CJ, et al. Bone-specific insulin resistance disrupts rodent islet isolation and assessment. Biol Proced Online 2009;11:3–31 whole-body glucose homeostasis via decreased osteocalcin activation. J Clin 27. Abdallah BM, Ding M, Jensen CH, et al. Dlk1/FA1 is a novel endocrine Invest 2014;124:1–13 regulator of bone and fat mass and its serum level is modulated by growth 6. Rached M-T, Kode A, Silva BC, et al. FoxO1 expression in osteoblasts hormone. Endocrinology 2007;148:3111–3121 regulates glucose homeostasis through regulation of osteocalcin in mice. J Clin 28. Qiu W, Hu Y, Andersen TE, et al. Tumor necrosis factor receptor superfamily Invest 2010;120:357–368 member 19 (TNFRSF19) regulates differentiation fate of human mesenchymal 7. Yoshizawa T, Hinoi E, Jung DY, et al. The transcription factor ATF4 regulates (stromal) stem cells through canonical Wnt signaling and C/EBP. J Biol Chem glucose metabolism in mice through its expression in osteoblasts. J Clin Invest 2010;285:14438–14449 2009;119:2807–2817 29. Ding M, Danielsen CC, Hvid I. Age-related three-dimensional micro- 8. Abdallah BM, Kassem M. New factors controlling the balance between architectural adaptations of subchondral bone tissues in guinea pig primary osteoblastogenesis and adipogenesis. Bone 2012;50:540–545 osteoarthrosis. Calcif Tissue Int 2006;78:113–122 9. Laborda J. The role of the epidermal -like protein dlk in cell 30. Dempster DW, Compston JE, Drezner MK, et al. Standardized nomencla- differentiation. Histol Histopathol 2000;15:119–129 ture, symbols, and units for bone histomorphometry: a 2012 update of the report 10. Jensen CH, Krogh TN, Højrup P, et al. Protein structure of fetal antigen 1 of the ASBMR Histomorphometry Nomenclature Committee. J Bone Miner Res (FA1). A novel circulating human epidermal-growth-factor-like protein expressed 2013;28:2–17 in neuroendocrine tumors and its relation to the gene products of dlk and pG2. 31. Pi M, Wu Y, Quarles LD. GPRC6A mediates responses to osteocalcin in Eur J Biochem 1994;225:83–92 b-cells in vitro and pancreas in vivo. J Bone Miner Res 2011;26:1680–1683 11. Smas CM, Sul HS. Pref-1, a protein containing EGF-like repeats, inhibits 32. Kajimura D, Lee HW, Riley KJ, et al. Adiponectin regulates bone mass via adipocyte differentiation. Cell 1993;73:725–734 opposite central and peripheral mechanisms through FoxO1 [published correction 12. Sul HS, Smas C, Mei B, Zhou L. Function of pref-1 as an inhibitor of adi- appears in Cell Metab 2014;19:891]. Cell Metab 2013;17:901–915 pocyte differentiation. Int J Obes Relat Metab Disord 2000;24(Suppl. 4):S15–S19 33. Yamauchi T, Kamon J, Waki H, et al. The fat-derived hormone adiponectin 13. Abdallah BM, Ditzel N, Mahmood A, et al. DLK1 is a novel regulator of bone reverses insulin resistance associated with both lipoatrophy and obesity. Nat Med mass that mediates estrogen deficiency-induced bone loss in mice. J Bone Miner 2001;7:941–946 Res 2011;26:1457–1471 34. Ferron M, Hinoi E, Karsenty G, Ducy P. Osteocalcin differentially regulates 14. Abdallah BM, Jensen CH, Gutierrez G, Leslie RG, Jensen TG, Kassem M. beta cell and adipocyte gene expression and affects the development of meta- Regulation of human skeletal stem cells differentiation by Dlk1/Pref-1. J Bone bolic diseases in wild-type mice. Proc Natl Acad Sci U S A 2008;105:5266–5270 Miner Res 2004;19:841–852 35. Moon YS, Smas CM, Lee K, et al. Mice lacking paternally expressed Pref-1/ 15. Abdallah BM, Bay-Jensen A-C, Srinivasan B, et al. Estrogen inhibits Dlk1/ Dlk1 display growth retardation and accelerated adiposity. Mol Cell Biol 2002;22: FA1 production: a potential mechanism for estrogen effects on bone turnover. J 5585–5592 Bone Miner Res 2011;26:2548–2551 36. Kavalkova P, Touskova V, Roubicek T, et al. Serum preadipocyte factor-1 16. Fazeli PK, Bredella MA, Misra M, et al. Preadipocyte factor-1 is associated concentrations in females with obesity and type 2 diabetes mellitus: the influence with marrow adiposity and bone mineral density in women with anorexia ner- of very low calorie diet, acute hyperinsulinemia, and fenofibrate treatment. Horm vosa. J Clin Endocrinol Metab 2010;95:407–413 Metab Res 2013;45:820–826 17. Aronis KN, Kilim H, Chamberland JP, Breggia A, Rosen C, Mantzoros CS. 37. Carlsson C, Tornehave D, Lindberg K, et al. Growth hormone and Preadipocyte factor-1 levels are higher in women with hypothalamic amenorrhea stimulate the expression of rat preadipocyte factor-1/delta-like protein in pan- and are associated with bone mineral content and bone mineral density through creatic islets: molecular cloning and expression pattern during development a mechanism independent of . J Clin Endocrinol Metab 2011;96:E1634– and growth of the endocrine pancreas. Endocrinology 1997;138:3940–3948 E1639 38. Tornehave D, Jansen P, Teisner B, Rasmussen HB, Chemnitz J, Moscoso G. 18. Lee K, Villena JA, Moon YS, et al. Inhibition of adipogenesis and de- Fetal antigen 1 (FA1) in the human pancreas: cell type expression, topological velopment of glucose intolerance by soluble preadipocyte factor-1 (Pref-1). J Clin and quantitative variations during development. Anat Embryol (Berl) 1993;187: Invest 2003;111:453–461 335–341 19. Villena JA, Choi CS, Wang Y, et al. Resistance to high-fat diet-induced 39. Friedrichsen BN, Carlsson C, Møldrup A, et al. Expression, biosynthesis obesity but exacerbated insulin resistance in mice overexpressing preadipocyte and release of preadipocyte factor-1/ delta-like protein/fetal antigen-1 in 3080 Dlk1 Regulates Osteocalcin-Insulin Loop Diabetes Volume 64, September 2015

pancreatic beta-cells: possible physiological implications. J Endocrinol 2003; 45. Pajvani UB, Shawber CJ, Samuel VT, et al. Inhibition of Notch signaling 176:257–266 ameliorates insulin resistance in a FoxO1-dependent manner. Nat Med 2011;17: 40. Chen L, Qanie D, Jafari A, et al. Delta-like 1/fetal antigen-1 (Dlk1/FA1) is 961–967 a novel regulator of chondrogenic cell differentiation via inhibition of the Akt 46. Chacón MR, Miranda M, Jensen CH, et al. Human serum levels of fetal kinase-dependent pathway. J Biol Chem 2011;286:32140–32149 antigen 1 (FA1/Dlk1) increase with obesity, are negatively associated with 41. Baladrón V, Ruiz-Hidalgo MJ, Nueda ML, et al. Dlk acts as a negative insulin sensitivity and modulate inflammation in vitro. Int J Obes (Lond) regulator of Notch1 activation through interactions with specific EGF-like repeats. 2008;32:1122–1129 Exp Cell Res 2005;303:343–359 47. Hong SH, Koo JW, Hwang JK, et al. Changes in serum osteocalcin are not 42. Bray SJ, Takada S, Harrison E, Shen SC, Ferguson-Smith AC. The atypical associated with changes in glucose or insulin for osteoporotic patients treated mammalian ligand Delta-like homologue 1 (Dlk1) can regulate Notch signalling in with bisphosphonate. J Bone Metab 2013;20:37–41 Drosophila. BMC Dev Biol 2008;8:11 48. Schwartz AV, Schafer AL, Grey A, et al. Effects of antiresorptive therapies on 43. Müller D, Cherukuri P, Henningfeld K, et al. Dlk1 promotes a fast motor glucose metabolism: results from the FIT, HORIZON-PFT, and FREEDOM trials. J neuron biophysical signature required for peak force execution. Science 2014; Bone Miner Res 2013;28:1348–1354 343:1264–1266 49. Flehmig G, Scholz M, Klöting N, et al. Identification of adipokine clusters 44. Valenti L, Mendoza RM, Rametta R, et al. Hepatic notch signaling correlates with related to parameters of fat mass, insulin sensitivity and inflammation. PLoS One insulin resistance and nonalcoholic fatty liver disease. Diabetes 2013;62:4052–4062 2014;9:e99785