Original Article BMP4 gene therapy enhances insulin sensitivity but not adipose tissue browning in obese mice Jenny M. Hoffmann 1, John R. Grünberg 1, Ann Hammarstedt 1, Tobias Kroon 2, Thomas U. Greiner 3, Stefanie Maurer 2, Ivet Elias 4, 5, Vilborg Palsdottir 6, Fatima Bosch 4, 5, Jeremie Boucher 1, 2, 7, Shahram Hedjazifar 1, Ulf Smith 1, * ABSTRACT Objective: Bone morphogenetic protein 4 (BMP4) adeno-associated viral vectors of serotype 8 (AAV8) gene therapy targeting the liver prevents the development of obesity in initially lean mice by browning the large subcutaneous white adipose tissue (WAT) and enhancing energy expenditure. Here, we examine whether this approach could also reduce established obesity. Methods: Dietary-induced obese C57BL6/N mice received AAV8 BMP4 gene therapy at 17e18 weeks of age. They were kept on a high-fat diet and phenotypically characterized for an additional 10e12 weeks. Following termination, the mice underwent additional characterization in vitro. Results: Surprisingly, we observed no effect on body weight, browning of WAT, or energy expenditure in these obese mice, but whole-body insulin sensitivity and glucose tolerance were robustly improved. Insulin signaling and insulin-stimulated glucose uptake were increased in both adipose cells and skeletal muscle. BMP4 also decreased hepatic glucose production and reduced gluconeogenic enzymes in the liver, but not in the kidney, in addition to enhancing insulin action in the liver. Conclusions: Our findings show that BMP4 prevents, but does not reverse, established obesity in adult mice, while it improves insulin sensitivity independent of weight reduction. The BMP antagonist Noggin was increased in WAT in obesity, which may account for the lack of browning. Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Keywords Obesity; Gene therapy; Insulin sensitivity; Adipose tissue; Skeletal muscle; Liver 1. INTRODUCTION secretion. However, BMP4 signaling is antagonized by increased secretion of BMP antagonists, and two of these antagonists are Noggin White adipose tissue (WAT) expansion can occur through enlargement [5,6] and Gremlin 1 [7]. We found Gremlin1 to be increased in human of existing adipocytes (hypertrophic expansion) and/or by recruitment hypertrophic WAT [4], whereas in the mouse, WAT Noggin, but not and differentiation of new adipocytes (hyperplastic expansion) [1]. Gremlin1, is increased [8]. White adipocytes have a mesenchymal stem cell origin [2], and the We and others have described that BMP4 also has a role in enhancing complex molecular events regulating commitment and differentiation beige adipogenesis in human preadipocytes [4], transgenic mice are still unclear. One of the early, and central, events in white adipocyte overexpressing Bmp4 in WAT [9], as well as in lean mature mice commitment is the action of bone morphogenetic protein 4 (BMP4), following BMP4 gene therapy [8]. In the latter study, we treated adult, which results in the activation of peroxisome proliferatoreactivated lean mice with adeno-associated viral vectors of serotype 8 (AAV8) receptor gamma (PPARg) and commitment of preadipocytes [2,3]. carrying the Bmp4 gene and targeting the liver, resulting in increased Interestingly, BMP4 is secreted by mature adipocytes and increased in circulating BMP4 levels, which targeted the SubQ WAT and induced hypertrophic obesity [4], thereby eliciting a positive feedback signal to browning. The mice had increased energy expenditure and were recruit new adipocytes. We have also previously shown that primary protected from diet-induced obesity, despite the finding that BMP4 human preadipocytes, isolated from the subcutaneous (SubQ) WAT actually inhibits BAT activation, as also shown in direct in vitro ex- depot, undergo improved differentiation in vitro when treated with periments [10]. However, these results support a beneficial effect of BMP4 [5], consistent with data in other models [2]. Hypertrophic ad- BMP4 only in preventing obesity. Its potential role in treating obesity ipocytes have increased endogenous BMP4 gene expression and and insulin resistance is still unknown. 1 The Lundberg Laboratory for Diabetes Research, Department of Molecular and Clinical Medicine, the Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden 2Bioscience Metabolism, Research and Early Development, Cardiovascular, Renal and Metabolic (CVRM), BioPharmaceuticals R&D, AstraZeneca, Gothenburg, Sweden 3The Wallenberg Laboratory, Department of Molecular and Clinical Medicine, The Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden 4Center of Animal Biotechnology and Gene Therapy and Department of Biochemistry and Molecular Biology, School of Veterinary Medicine, Universitat Autònoma de Barcelona, 08193, Bellaterra, Spain 5Centro de Investigación Biomédica en Red de Diabetes y Enfermedades Metabólicas Asociadas (CIBERDEM), Spain 6Department of Physiology/Endo- crinology, The Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden 7Wallenberg Centre for Molecular and Translational Medicine, University of Goth- enburg, Sweden *Corresponding author. The Lundberg Laboratory for Diabetes Research, Department of Molecular and Clinical Medicine, Sahlgrenska University Hospital, Blå Stråket 5, SE- 413 45, Gothenburg, Sweden. E-mail: ulf.smith@medic.gu.se (U. Smith). Received October 30, 2019  Revision received November 21, 2019  Accepted November 22, 2019  Available online 17 December 2019 https://doi.org/10.1016/j.molmet.2019.11.016 MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 15 www.molecularmetabolism.com Original Article Therefore, in the present study, we tested whether BMP4 gene therapy BMP4-treated mice did not reduce their body weights nor were they could also be used to treat already established obesity. Our results protected from further weight gain (Figure 1E). Food intake was also show that obesity is not reduced but that BMP4 improves whole-body not changed compared with the control mice (Figure 1F), nor were insulin sensitivity, enhances insulin signaling in all key metabolic tis- there any differences in the weights of Epi or SubQ WAT depots sues, and reduces key gluconeogenic enzymes in the liver despite no (Figure 1G), in SubQ adipocyte size (Figure 1H), in VO2 (Figure 1I), or in weight loss. animal movements (data not shown). Adipocyte size in the Epi depot was not analyzed. 2. RESULTS We next examined markers of browning, which were not increased at the transcriptional level in the SubQ WAT (Figure 2A; including UCP1 The mice were fed a high-fat diet (HFD) for 11 weeks prior to the AAV8 protein, data not shown), and OXPHOS activation was similar injections to allow increased body weight. Body weights and blood (Figure 2B). We also examined whether BMP4 gene therapy enhanced glucose levels were used to match the two groups for the AAV8 BMP4 markers of terminal adipocyte differentiation in the SubQ WAT. As and AAV8 control injections for cohort 1 (at study week 0; Fig. S1a and shown in Figure 2C, only the expression of adiponectin and G proteine b) and later also for a second cohort of mice (cohort 2, also injected at coupled receptor 120 (Gpr120) were significantly increased. There study week 0; Fig. S2a and f). Schematic figures of the study designs were no differences in markers of WAT fibrosis or inflammation for cohorts 1 and 2 are shown in Figs. S1c and S2b. Although initial (Figure 2D) nor in circulating levels of SAA3 (Figure 2E). However, AAV8 design and matching of the mice were similar, different phenotyping BMP4 mice had significantly higher serum levels of adiponectin procedures were performed, and cohort 2 was used to examine he- (Figure 2F) and lower levels of free fatty acids (Figure 2G), suggesting patic glucose production through a pyruvate tolerance test and for enhanced insulin sensitivity. labeled tissue glucose uptake. BMP4 has recently been shown to directly target and inhibit BAT cells [10], which is also seen in lean mice following BMP4 gene therapy [8], 2.1. Increased hepatic and serum BMP4 levels following AAV8 leading to increased lipid accumulation and inhibition of its lipolytic and BMP4 injections, but not in peripheral tissues oxidative effects [8,10]. We therefore also examined BAT but saw a Twelve weeks after tail-vein injection of 5  1011 vg/mice of AAV8 Ctrl similar increase in BAT lipid accumulation in both groups of obese and AAV8 BMP4, vector genome copy number was determined in liver mice (Fig. S1e). Furthermore, both Ucp1 mRNA (Fig. S1f) and protein and epididymal fat (Epi) of injected mice from cohort 1. As shown in (Fig. S1g) were similar, and other mitochondrial and white adipose Figure 1A, we found a very high transduction of the liver (vector marker genes, including Pparg and Fabp4, were also similar (Fig. S1f). genome/diploid genome), while the levels were marginal in Epi WAT. Hormone-sensitive lipase (HSL) protein was also not different between This result is consistent with the high tropism for the liver of the AAV8 the two groups (Fig. S1h). Consistent with findings in WAT, we also vectors after intravascular administration. Moreover, when the saw increased endogenous tissue Bmp4 and Noggin gene expression expression levels of the mouse codoneoptimized BMP4 (moBMP4) in BAT in both obese groups compared with lean control mice were measured by quantitative reverse transcriptase polymerase chain (Fig. S1i). Thus, BAT in obese mice seems to have already acquired a reaction (RT-qPCR) in the liver and Epi WAT of the AAV8 BMP4-treated beige phenotype, which was not further enhanced by AAV8 BMP4, mice, very high levels were observed in the liver of these mice, while probably because tissue BMP4 is already increased in obesity. Epi fat again expressed only marginal levels (Figure 1B), which is Taken together, these results show that BMP4 gene therapy in initially consistent with the use of the liver-specific human alpha 1-antitrypsin obese mice does not increase browning of the SubQ WAT or reduce (hAAT) promoter. body weight. In addition, BMP4 gene therapy does not reduce fibrosis Next, we evaluated the effect of the AAV8 BMP4 treatment on serum or inflammation in WAT and also has no impact on markers of BAT and tissue protein BMP4 levels in the same mice. As expected, function in obesity. hepatic BMP4 protein (Figure 1C) was markedly increased, but the vectors did not increase BMP4 protein in the other tissues exam- 2.3. Reduced BMP signaling in subcutaneous adipose tissue but ined, including skeletal muscle (tibialis anterior; TA) or SubQ WAT, not in skeletal muscle or in liver similar to previous reports with these vectors [8]. Endogenous BMP4 Our current findings suggest the induction of a BMP4 resistance in protein was low in the livers from control animals compared with murine WAT cells in obesity. In fact, similar to BAT, we found increased the AAV8 BMP4 mice (Figure 1C), but it was clearly expressed at the mRNA levels of endogenous Bmp4 in WAT in the obese HFD groups gene level (Ct values around 31 in both groups; see also Figure 3E). compared with a group of lean animals previously studied [5] (CD Ctrl) This was further verified in an uncut immunoblot of liver BMP4 as well as increased expression of the BMP4 antagonist Noggin protein, which clearly showed that it was present as an unprocessed (Figure 3A). In line with this, pSMAD1/5/8 levels were reduced in SubQ protein in the control animals (Fig. S1d). Endogenous BMP4 was adipocytes in both obese groups (Figure 3B), and there was no sig- highly expressed in WAT and skeletal muscles (TA), but there was nificant difference between the groups. In contrast, neither Noggin nor no difference between the groups (Figure 1C). Mice treated with the BMP4 mRNA levels were different between the groups in skeletal AAV8 BMP4 vectors had increased circulating BMP4 levels muscle (Figure 3C), and pSMAD1/5/8 was also increased in both measured at termination as expected (Figure 1D). These results groups (Figure 3D). demonstrate that AAV8 BMP4 mice have increased hepatic and Endogenous BMP4 mRNA levels were also similar in the liver circulating BMP4 levels and that the BMP4 expression lasted for the (Figure 3E), while the expression of Noggin was very low in both lean full duration of the study. and obese mice (data not shown) and pSMAD1/5/8 was increased, particularly in the AAV8 BMP4 mice (Figure 3F). Total SMAD protein 2.2. AAV8 BMP4 gene therapy in obese mice does not induce was unchanged in the liver (Figure 3F). weight loss or browning of WAT We also validated that BMP activity in serum from obese untreated HFD In contrast to our previous findings in which initially lean mice fed a mice was lower than that in serum from lean mice by using an in vitro HFD were protected from obesity [8], the current initially obese AAV8 BMP reporter cell assay system, which suggested the presence of 16 MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). www.molecularmetabolism.com Liver EPI A B 0,4 0.4 Ctrl Ctrl moBMP4 expression (AU) 0,3 0.3 BMP4 BMP4 Vector genomes / diploid genome Ctrl BMP4 0.2 0,2 Liver N.D. 6.86 ± 0.52 EPI N.D. 0.04 ± 0.02 0.1 0,1 N.D. N.D. N.D. 0 C Liver TA SubQ WAT -25 kDa -25 kDa -25 kDa BMP4 BMP4 BMP4 -20 kDa -20 kDa -20 kDa β-Tubulin -55 kDa GAPDH -37 kDa ACTIN -37 kDa BMP4 BMP4 BMP4 Ctrl Ctrl D E 60 Ctrl -25 kDa 55 BMP4 Body weight (g) -20 kDa 50 IP serum pooled BMP4 mice Neg. Ctrl Pos. Ctrl IP serum pooled Ctrl mice 45 40 35 Ctrl BMP4 30 0 1 2 3 4 5 6 7 8 9 10 11 12 Weeks following AAV administration F G H 0.08 3 100 SubQ adipocyte diameter WAT depot weight (g) (norm. to body weight) 80 0.06 Food intake 2 60 (um) 0.04 Ctrl BMP4 40 1 0.02 20 0.00 0 0 EPI SubQ I 100 VO2 [ml/(min*kg)] 90 80 70 60 Ctrl BMP4 50 10 12 14 16 18 20 22 00 02 04 06 Time of day (h) Figure 1: The effect of BMP4 gene therapy on body weight gain in obese mice. Vector gene copy number was determined in DNA isolated from liver and Epi WAT by qPCR with primers specific for BMP4. Liver showed high transduction compared with Epi WAT (A). Mouse codon-optimized BMP4 (moBMP4) (described in the Supplemental Methods section) expression was analyzed by RT-qPCR in liver and Epi WAT. ND, nondetectable; AU, arbitrary units; Epi WAT, epididymal WAT (B). At termination, the AAV8 BMP4 mice had increased BMP4 protein in liver lysates but not in lysates from tibialis anterior (TA) skeletal muscle or subcutaneous (SubQ) WAT (C). AAV8 BMP4 mice also had increased circulating BMP4 levels (D). Beads and antibody were used as negative control in (D). There was no difference in weight gain (E) or food intake (F) between the groups or in weight of WAT depots (G), SubQ adipocyte size (H), or oxygen consumption (I) between the AAV8 BMP4 mice and control group. All graphs display the means  SEM. A-I: Material from cohort 1. A-B: n ¼ 3 þ 3. E-H: Ctrl n ¼ 15, BMP4 n ¼ 14. (I): Ctrl n ¼ 11, BMP4 n ¼ 12. Statistics were calculated using Student’s t-test. MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 17 www.molecularmetabolism.com Original Article A B 1.5 * -55 kDa SubQ WAT/18S Relative mRNA CV-ATP5A 1.0 -48 kDa CIII-UQCRC2 CIV-MTCO1 -40 kDa 0.5 -30 kDa CII-SDHB Ctrl BMP4 CI-NDUFB8 -20 kDa 0 Pgc1 Ucp1 Tmem26 Cox2 Ctrl BMP4 C D Relative mRNA SubQ WAT/18S Relative mRNA SubQ WAT / 18S 3,0 3.0 * 2.0 2.5 2,5 2.5 ** 1.6 2,0 2.0 1.2 1.5 1,5 1.5 0.8 1.0 1,0 1.0 Ctrl 0.5 0,5 Ctrl 0.4 0.5 BMP4 BMP4 0,00 0 GLUT4 Adiponec n Fabp4 PPARg Znf423 C/EBPa GPR120 E F G 1.0 10,000 * 1,200 * Free fatty acids (uM) Adiponectin (ng/ml) 1,000 0.8 8,000 SAA3 (ug/ml) 800 0.6 6,000 600 0.4 4,000 400 0.2 2,000 200 0 0 0 Figure 2: BMP4 gene therapy does not enhance browning of subcutaneous (SubQ) WAT in obesity. There was no increased expression of beige/brown adipocyte genes in subcutaneous (SubQ) WAT in BMP4 mice (A) or SubQ WAT mitochondrial OXPHOS proteins (B) compared with the control group. Increased markers of adipose cell differentiation in AAV8 BMP4 SubQ WAT but only adiponectin and Gpr120 mRNA levels were significantly increased (C). There was no difference in expression of inflammatory and fibrotic genes in SubQ WAT (D) and no difference in circulating SAA3 (E). AAV8 BMP4 mice had increased levels of circulating adiponectin (F) and reduced free fatty acids (G) in fasting serum samples. A-G show material from cohort 1. Graphs display means  SEM. A: PGC1a, TMEM26, COX2 Ctrl n ¼ 15, BMP4 n ¼ 14. UCP1 Ctrl n ¼ 14, BMP4 n ¼ 12. B: n ¼ 4. C: GLUT4, adiponectin, FABP4, PPARg Ctrl n ¼ 15, BMP4 n ¼ 14. ZFP423 Ctrl n ¼ 13, BMP4 n ¼ 11. C/EBPa Ctrl n ¼ 12, BMP4 n ¼ 8. GPR120 Ctrl n ¼ 12, BMP4 n ¼ 10. D: CTGF, IL6, F4/80, SAA3 Ctrl n ¼ 15, BMP4 n ¼ 14. TNFa Ctrl n ¼ 14, BMP4 n ¼ 13. E: Ctrl n ¼ 12, BMP4 n ¼ 11. F: n ¼ 14. G: Ctrl n ¼ 13, BMP4 n ¼ 14. Statistics in (G) were calculated using the ManneWhitney nonparametric U-test. Statistics were otherwise calculated using Student’s t-test. increased circulating BMP antagonists in obesity (Figure 3G). Taken was associated with altered glucose tolerance and insulin sensitivity, together, increased Noggin and reduced pSMAD 1/5/8 specifically in the mice underwent an intraperitoneal (i.p.) glucose tolerance test WAT suggest that the reason for the lack of response to BMP4 therapy (GTT) and insulin tolerance test (ITT) 10e11 weeks after vector in obesity, measured as browning of WAT and body weight reduction, administration. The AAV8 BMP4 mice had significantly better glucose is due to BMP4 resistance in the adipose tissue. tolerance compared with controls (Figure 4B), with lower insulin levels at baseline after 4 h of fasting and later during the GTT (Figure 4C). 2.4. Improved insulin sensitivity and glucose homeostasis in obese Consistent with this, the AAV8 BMP4 mice also had significantly better AAV8 BMP4 mice whole-body insulin sensitivity (Figure 4D, left) and glucose area under We next assessed the potential effects of AAV8 BMP4 gene therapy on the curve (Figure 4D, right). We also performed immunostaining and glucose homeostasis. Ten weeks after AAV8 injections and on quantification of pancreatic islet size, which showed no difference continued HFD, the control mice in cohort 1 tended to have further between the AAV8 BMP4 and AAV8 control mice (Figure 4E and increased fasting insulin levels, while the AAV8 BMP4 mice had quantified diagram right). Taken together, these data show that BMP4 significantly lower fasting insulin levels compared with both samples enhances insulin sensitivity and that the lower insulin levels during the taken prior to (at lower body weights) as well as samples taken 10 GTT are a consequence of the improved insulin sensitivity and inde- weeks after vector administration (Figure 4A). To assess whether this pendent of any effects on body weight. 18 MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). www.molecularmetabolism.com A * B 6 ** Relative mRNA / 18S 5 ** SubQ WAT 4 -60 kDa phos SMAD1/5/8 3 * ACTIN -37 kDa 2 Lean CD Ctrl Endo Bmp4 Pooled HFD Pooled HFD 1 Noggin ctrl mice BMP4 mice 0 CD HFD HFD CD Ctrl Ctrl HFD BMP4 HFD Ctrl Ctrl BMP4 C 1.5 1,5 D d phos SMAD1/5/8 -60 kDa Relative mRNA / 18S 1.2 1,2 Gastrocnemius GAPDH -37 kDa 0.9 0,9 CD Ctrl HFD Ctrl HFD BMP4 0.6 0,6 2,5 2.5 * phos SMAD / GAPDH 0.3 0,3 Endo Bmp4 2,0 2.0 Noggin 1,5 1.5 0.0 0,0 CDCD Ctrl HFD HFD Ctrl HFDHFD BMP4 1.0 1,0 Ctrl Ctrl BMP4 0.5 0,5 0,0 0 HFD Ctrl HFD B4 Ctrl BMP4 E F 3,0 Endogenous Bmp4 phos SMAD1/5/8 -60 kDa 3.0 2,5 2.5 β-Tubulin -55 kDa Relative mRNA / 18S 2,0 2.0 HFD Ctrl HFD BMP4 Liver 1.5 1,5 1,0 1.0 -60 kDa Total SMAD1/5/8 0.5 0,5 0 0,0 β-Tubulin -55 kDa CD HFD HFD Ctrl BMP4 Ctrl BMP4 hepG2 ctrl HFD Ctrl HFD BMP4 G *** phos SMAD / β-Tubulin 3 50 * 2 40 BRE activity 1 30 20 0 HFD HFD 10 Ctrl BMP4 0 Lean Obese (CD) (HFD) Figure 3: Reduced BMP signaling in subcutaneous adipose tissue but not in skeletal muscle or in liver. Increased mRNA levels of endogenous Bmp4 and the BMP antagonist Noggin in subcutaneous (SubQ) WAT in HFD-fed compared with lean (CD Ctrl) mice (A). Low pSMAD1/5/8 activation in lysates of SubQ adipocytes in both obese groups (B); BMP4-stimulated lean WT adipocytes were used as positive control in (B). Gene expression of endogenous BMP4 and of the antagonist Noggin in gastrocnemius biopsies from lean control mice, compared with obese control and BMP4 mice (C). pSMAD1/5/8 is significantly increased in gastrocnemius skeletal muscle from BMP4 obese compared with control obese mice (D) (all samples were run on a single, large gel). Gene expression of endogenous BMP4 in liver tissue from lean control and obese AAV8 BMP4 and control mice (E). pSMAD1/5/8 is significantly increased in liver tissue in obese AAV8 BMP4 compared with obese control mice; however, the expression of total SMAD 1/5/8 is unchanged (F). Reduced BMP-response element activity in C2C12-BRE cells treated with serum from obese (HFD-fed) mice compared with serum from lean (CD-fed) mice (G). A-G show material from cohort 1. Graphs display means  SEM. A: endo BMP4, Noggin CD Ctrl n ¼ 10, HFD Ctrl n ¼ 13, HFD BMP4 n ¼ 8. C: Endo BMP4/Noggin CD Ctrl n ¼ 9, HFD Ctrl/BMP4 n ¼ 8. E: CD Ctrl n ¼ 8, HFD/BMP4 n ¼ 8. G: lean n ¼ 9, obese n ¼ 11. Statistics were calculated using Student’s t-test. MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 19 www.molecularmetabolism.com Original Article A B C Insulin during glucose tolerance test (ng/ml) Circulating fasting insulin 10 25 14 Blood glucose (mmol/l) # 8 * 20 * 12 * * * # 10 * 6 15 8 (ng/ml) 4 10 6 Ctrl 4 2 5 Ctrl BMP4 2 0 BMP4 0 0 Before AAV After AAV 0 15 30 60 90 120 0 Fasted, 15´15GTT 30´30 GTT injection injection Minutes after ip glucose Minutes after AAV after after AAVip glucose after AAV (GTT 0´) D E 15000 12 1500 * 15,000 Mean Islet Area (um2) 10 * AUC (increase gluose x min/L) Blood glucose (mmol/l) 10000 10,000 8 ** * 1000 6 5000 5,000 4 500 2 Ctrl 00 BMP4 0 0 0 15 30 60 90 120 Ctrl BMP4 Minutes after ip insulin F SubQ adipocytes G TA skeletal muscle Ins Ins Ins Ins -150 kDa -60 kDa IRS1 -130 kDa phos AKT IR b -95 kDa AKT -60 kDa AKT -60 kDa GAPDH -37 kDa β-Tubulin -55 kDa Ctrl BMP4 1.0 1 * phos AKT / total AKT (TA) phos IRb -95 kDa 0,8 0.8 phos AKT -60 kDa β-Tubulin -55 kDa 0,6 0.6 Ctrl BMP4 Ctrl BMP4 0,4 0.4 0,2 0.2 00 Ctrl BMP4 H I # 30 # # * * 1,000 1000 Ctrl * * Plasma glucose (mM) 25 BMP4 Rg'(nmol/g/min) 800 20 600 15 400 10 5 Ctrl 200 BMP4 0 0 0 15 20 25 35 45 60 Minutes after oral glucose gavage Figure 4: BMP4-treated obese mice have improved insulin sensitivity. Serum insulin levels were similar in both groups before AAV8 injections (A). Ten weeks following AAV8 administration and on continued HFD, control mice showed increased insulin levels, while levels were decreased in AAV8 BMP4 mice (A). AAV8 BMP4 mice had enhanced glucose tolerance (B), lower insulin levels at baseline and during the glucose tolerance test (C), and improved insulin sensitivity (D). There was no difference in pancreas islet size between the groups (E). AAV8 BMP4 mice had significantly increased Insulin Receptor Substrate (IRS) 1 protein and a trend for increased IRb protein in isolated SubQ adipocytes. Insulin- stimulated IRb and AKT phosphorylation were increased in SubQ adipocytes from AAV8 BMP4 compared with control cells, and basal AKT phosphorylation was also increased (F). Increased basal AKT phosphorylation in biopsies of tibialis anterior (TA) skeletal muscle from AAV8 BMP4 mice compared with control mice (G). Blood glucose levels were decreased during oral administration of glucose (H), and the AAV8 BMP4 mice showed tissue-specific increased glucose uptake in soleus skeletal muscle and BAT (I). A-G show material from cohort 1. H-I show material from cohort 2. All graphs show means  SEM. A-D: n ¼ 9, E: n ¼ 3, H-I: n ¼ 5. Statistics were calculated using Student’s t-test. 20 MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). www.molecularmetabolism.com Next, we examined insulin signaling in two target tissues, isolated significantly reduced in the livers of the AAV8 BMP4 mice compared SubQ adipocytes and skeletal muscle, from the equally obese AAV8 with the equally obese control mice. This effect was not seen in the BMP4-treated and -untreated mice. As shown in Figure 4F, insulin kidney, which is another gluconeogenic tissue, of the AAV8 BMP4- receptor substrate 1 (IRS1) protein was increased in BMP4 mice, and treated mice (Figure 5D) [11]. Moreover, endogenous BMP4 expres- IR-b also tended to be increased, whereas there was no difference in sion was similar (Figure 5D). AKT protein. Furthermore, insulin-stimulated IR-b phosphorylation To elucidate whether the effect of BMP4 on gluconeogenesis is (Tyr1146) and downstream pS 473AKT phosphorylation were signifi- mediated by direct suppressive effects on key markers of gluconeo- cantly increased (Figure 4F). We also found significantly increased AKT genesis or by enhancing insulin action as in adipose tissue and skeletal phosphorylation in excised TA skeletal muscles (Figure 4G and graph muscles, we treated human liver cells (IHH) with recombinant BMP4 below), supporting enhanced insulin sensitivity in skeletal muscles, alone or with insulin. Interestingly, recombinant BMP4 alone was which is likely an important reason for the increased whole-body in- sufficient to suppress these key markers of gluconeogenesis sulin effect (Figure 4D). (Figure 5E). However, BMP4 also enhanced the insulin-induced sup- pression of the gluconeogenic enzymes (Figure 5E). As expected, 2.5. Tissue glucose uptake Gremlin1, which is a key BMP4 antagonist in human cells, prevented We then characterized the phenotype of the improved glucose ho- BMP4-induced suppression of the gluconeogenic enzymes and pre- meostasis in cohort 2 mice, which were treated identically to the mice vented the positive effect of BMP4 with insulin (Figure 5E). in cohort 1. We confirmed their similar body weights during the study The important role of PGC1-a in the regulation of hepatic gluconeo- (Fig. S2c) and the enhanced insulin sensitivity through an ITT genesis is well established. Acetylation of PGC1-a has been shown to (Fig. S2d). To further characterize these mice, we administered oral modulate its activity in several contexts, including gluconeogenesis glucose via gavage and injected 14C-2deoxy-glucose to conscious [12]. We also found that acetylation of hepatic PGC1-a in AAV8 BMP4 AAV8 BMP4 and control mice. Similar to cohort 1, BMP4-treated mice mice was significantly increased compared with that of the control from cohort 2 had reduced fasting plasma insulin at baseline (Fig. S2e) mice (Figure 5F). and markedly improved whole-body glycemic control following oral Collectively, our results show that BMP4 exerts insulin-like effects on glucose (Figure 4H), despite similarly reduced insulin levels as seen in the suppression of hepatic gluconeogenic enzymes in both murine and cohort 1 throughout the experiment, thus confirming the increased human cells and that this effect could be exerted through modulation of insulin sensitivity. Tissue-specific glucose uptake (Rg’) was signifi- PGC1-a acetylation. The suppressive effect of BMP4 on hepatic cantly increased in the skeletal muscle and BAT of AAV8 BMP4 mice gluconeogenesis is also consistent with the markedly improved PTT (Figure 4I). seen in AAV8 BMP4 mice. Taken together, these findings in obese mice are consistent with our previous results in lean, AAV8 BMP4-treated mice on a chow diet [5] 3. DISCUSSION and show that AAV8 BMP4 has a robust, positive effect on glucose tolerance and insulin sensitivity in both lean and obese mice. Tissue The current study was designed to determine whether BMP4 gene glucose uptake following oral glucose was increased in BAT and therapy, shown to prevent the development of obesity [8], could also skeletal muscle, consistent with the enhanced insulin signaling in be a potential treatment for established obesity. However, in contrast to skeletal muscle and also seen in white adipocytes. lean mice, BMP4 gene therapy did not reduce body weight or fat mass. Consistent with this, there was no browning of WAT or reduced 2.6. BMP4 also enhances hepatic insulin signaling and reduces inflammation or fibrosis, as previously seen in lean mice [5]. This lack hepatic gluconeogenic enzymes of response in WAT was also associated with increased Noggin mRNA We next examined whether BMP4 also could regulate hepatic glucose and reduced pSMAD1/5/8, supporting that the adipose tissues become production. Similar to our findings in WAT, we found markers of BMP4 resistant in obesity in contrast to skeletal muscles and liver. The increased insulin sensitivity also in the liver in the BMP4 mice from BMP reporter system data also supported the concept that obesity is a cohort 1 with increased IRb- and IRS1 protein (Fig. S3a blots and condition of increased BMP antagonism in the circulation, and this may quantifications). emanate from the expanded adipose tissue. The AAV8-mediated in- To test this in vivo, we performed a pyruvate tolerance test (PTT) to crease in circulating levels of BMP4 were similar to the levels we evaluate hepatic gluconeogenesis, which showed a significantly reported in initially lean AAV8 BMP4 mice [8] but still could not reduced hepatic glucose production in the AAV8 BMP4 mice overcome the WAT BMP4 resistance in established obesity. An inter- (Figure 5A). Both fed glucose and insulin levels were significantly esting alternative approach of overcoming adipose tissue BMP4 decreased in AAV8 BMP4 mice compared with those in controls resistance would thus be to inhibit Noggin in vivo in adult obese mice, (Figs. S2f and g), supporting that BMP4 also has positive effects on and such studies are in development. both hepatic insulin resistance and hepatic gluconeogenesis. We also show for the first time that obese AAV8 BMP4 mice exhibited To further validate this, we injected mice in cohort 2 with insulin in the clear improvement in insulin sensitivity in all key metabolic targets for vena cava 5 min prior to sacrifice and evaluated effects on insulin insulin and also in glucose homoeostasis compared with control mice, signaling in the liver. Both pS473AKT and pYIRb were increased despite similar body weight and fat mass. We also found improved compared with control mice, showing positive effects of AAV8 BMP4 whole-body insulin sensitivity in our previous study in lean mice [5], also on hepatic insulin sensitivity (Figure 5B). but this could have been secondary to the increased browning of WAT. As insulin exerts inhibitory effects on hepatic gluconeogenesis, we However, the current study clearly shows that this is not the case. asked whether AAV8 BMP4 also suppresses key markers of gluco- Peripheral tissue glucose uptake following an oral glucose gavage with neogenesis. As shown in Figure 5C, gene expression of key rate- tracer also showed that the skeletal muscles and BAT had significantly limiting enzymes of gluconeogenesis including glucose 6- increased glucose uptake, which likely contributes to the improved phosphatase (G6pc), phosphoenolpyruvate carboxykinase 1 (Pck1), insulin sensitivity. BAT is a metabolically important tissue in mice for fructose-1, 6-bisphosphatase 1 (Fbp1), and glucokinase (Gck) were both glucose and lipid metabolism [13]. The increase in insulin- MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 21 www.molecularmetabolism.com Original Article A 20 ** * * B # -95kDa # IRb 18 PTT glucose (mmol/L) 16 phos AKT -60kDa 14 β-Tubulin -55kDa 12 10 8 phos IRb -95kDa 6 4 AKT -60kDa 2 β-Tubulin -55kDa 0 Ctrl + ins BMP4 + ins 0 15 30 60 90 120 Minutes after ip pyruvate C D 2.0 2,0 7 Ctrl * Ctrl Relative mRNA kindney cortex / 18S BMP4 Relative mRNA liver tissue / 18s BMP4 6 1.5 1,5 5 4 * 1,0 1.0 3 2 * 0.5 0,5 * 1 0 0,0 0 G6pc Pck1 Fbp1 Gck endo Bmp4 G6pc Pck1 Fbp1 E F Basal recBMP4 Acetyl- -110 kDa recBMP4+recGremlin1 PGC1a -80 kDa -110 kDa Relative mRNA IHH liver cells / 18s 1.4 * PGC1a -80 kDa BMP4 BMP4 Neg. Ctrl Ctrl Ctrl BMP4 Ctrl BMP4 Ctrl Ctrl BMP4 ** * 1.2 1.0 *** *** ** ** *** ** 0.8 ** 0.6 * 1.4 * *** * * Acetyl-PGC1a / Total PGC1a 0.4 1.2 0.2 ** 1.0 0 - ** Insulin Insulin Insulin Insulin 0.8 - - - 0.6 G6PC PCK1 FBP1 GCKR 0.4 0.2 0 Figure 5: BMP4 treatment reduces hepatic gluconeogenesis and increases PGC1-a acetylation in liver. AAV8 BMP4 mice had improved pyruvate tolerance compared with controls (A). Insulin-stimulated hepatic AKT and IRb phosphorylation was increased in obese AAV8 BMP4 mice compared with controls (B). Reduced expression of gluconeogenic markers in liver tissue from obese AAV8 BMP4 mice compared with controls (C). However, these markers were unaltered in the renal cortex from control diet-fed AAV8 BMP4 mice compared with controls, and endogenous BMP4 levels were also not different (D). Recombinant BMP4 treatment of IHH cells also reduced the expression of gluconeogenic markers (E). This effect was blunted by recombinant Gremlin1. Acetylated PGC1-a was increased in liver tissue from AAV8 BMP4 mice compared with controls (F). Beads and antibody were used as negative control in (F). A and B show material from cohort 2. C, D, and F show material from cohort 1. All graphs show means  SEM. A: n ¼ 5, B: n ¼ 2, C: n ¼ 8, D: n ¼ 7 þ 8, E: n ¼ 4e5 (n ¼ 5 for all conditions not stimulated with insulin and n ¼ 4 for all insulin-stimulated conditions), F: n ¼ 5. Statistics were calculated using Student’s t- test. 22 MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). www.molecularmetabolism.com stimulated glucose uptake in BAT indicates that insulin sensitivity is HFD (45 kcal% fat; D12451, Research Diets, New Brunswick, NJ, USA). also increased. A first cohort of mice was initially thoroughly characterized; these mice Further detailed studies also showed that all metabolic target tissues are referred to as cohort 1 throughout the manuscript. Later on, we for insulin, including WAT, had improved cellular insulin signaling, added a second cohort of mice, referred to as cohort 2, to shed more clearly suggesting that the canonical pSMAD1/5/8 is not a mediator of light on findings from cohort 1. The mice were started on an HFD at 6 this effect of BMP4, in contrast to that of browning of the white adipose weeks of age and kept on the diet until vector administration (study cells. An additional positive effect of BMP4 was the reduced expression week 0). Upon vector administration, the mice in cohort 1 and 2 were of key hepatic gluconeogenic enzymes in AAV8 BMP4 mice, which 17 and 18 weeks of age, respectively. The mice were kept on an HFD may, at least partly, be accounted for by the enhanced hepatic insulin until the endpoint (12 and 10 weeks after vector administration). For effect. However, a similar direct effect of BMP4 was seen in human cohort 1, results from the initial GTT and body weights at study liver cells in vitro, and this was additive to that of insulin. Thus, BMP4 week 1 (Figs. S1a and b) were used to match the two groups for exerts direct insulin-like effects in the liver and also enhances the subsequent vector administration (described below) at study week 0. effect of insulin. We could confirm that this is a direct effect of BMP4 Food intake was measured over 1 week. For cohort 2, body weight and since it was antagonized by Gremlin1. Taken together, our results raise fed glucose measurements at study week 1 (Figs. S2a and f) were the intriguing question of whether BMP4 gene therapy targeting the used to match the two groups for subsequent vector administration at liver can be used to treat hyperglycemia in animal models of diabetes study week 0. At the endpoint, the mice were euthanized using iso- regardless of obesity. This potential therapeutic approach is currently flurane (Baxter Medical AB, Kista, Sweden), and blood and tissues were under investigation. treated as described below for subsequent analyses. Figs. S1c and S2b Our finding of increased PGC1-a acetylation in liver in BMP4-treated show the study designs for cohorts 1 and 2. Exact information on which mice is also interesting as PGC1-a is a well-known regulator of both assay was performed in which cohort of mice can be found in the figure hepatic gluconeogenesis and lipid metabolism [12]. Recently, selective legends. All animal experiments were approved by the Research Animal inhibition of PGC1-a, by increasing its acetylation, was shown to reduce Ethics Committee at the University of Gothenburg, Sweden. hepatic gluconeogenesis and improve insulin sensitivity in T2D mice without increasing liver triglycerides or cholesterol levels [14]. In our 5.2. Recombinant adeno-associated viral vectors study, there were also no differences in liver lipid accumulation be- Briefly, recombinant AAV8 encoding a codon-optimized murine Bmp4 tween AAV8 BMP4 and control mice, as examined by histological cDNA sequence under control of the hAAT promoter was produced by staining and triglyceride quantification (Figs. S3b and c). The lower triple transfection of HEK293 cells followed by an optimized cesium insulin levels as a consequence of the improved insulin sensitivity may chloride gradient-based purification that renders vector batches of high also contribute to this. It is well known that chronic hyperinsulinemia in purity [18]. A noncoding plasmid carrying the hAAT promoter was used obesity is closely associated with increased hepatic de novo lipogenesis to produce empty vectors for control mice [19]. The vectors (5  1011 and the associated progression of nonalcoholic fatty liver disease [15]. viral particles/200 mL saline/mouse; 9 mg/mL saline; Fresenius Kabi, The detailed molecular mechanisms by which BMP4 gene therapy Bad Homburg, Germany) were administered via intravenous tail-vein improves insulin sensitivity in all key metabolic tissues need to be injection at study week 0 (when the mice were 17 weeks old) [8]. further explored. However, it should also be emphasized that the More details of the vectors and biodistribution can be found in the observed elevated serum adiponectin levels and enhanced transcrip- Supplemental methods section. tional activation of Gpr120 in WAT may also contribute to the increased whole-body insulin sensitivity. Both adiponectin and GPR120 activation 5.3. GTT, ITT, and PTT have previously been shown to improve whole-body insulin sensitivity After 4 h of fasting, the mice received i.p. injections containing glucose and enhance cellular insulin signaling [16,17]. (1 g/kg body weight [b.w.], Fresenius Kabi), insulin (human recombi- nant insulin, Actrapid Penfill, Novo Nordisk, Bagsvaerd, Denmark), or 4. CONCLUSIONS pyruvate (2.0 g/kg b.w. pyruvate, SigmaeAldrich, St. Louis, MO, USA) dissolved in NaCl. Following pilot experiments, the appropriate insulin Our results show that obesity is a state of BMP4 resistance in WAT, a concentrations were determined to 1.0 U/kg b.w. for cohort 1 and likely consequence of increased Noggin, but this is not seen in skeletal 1.2 U/kg b.w. for cohort 2. Blood samples were collected from the tail muscles or in liver. This resistance of local tissue to BMP4 inhibits the vein at 0, 15, 30, 60, 90, and 120 min after injections for subsequent positive effects of increased BMP4 on SubQ WAT browning and whole- serum analyses. Blood glucose concentrations were determined at body energy expenditure, which is seen in lean animals [8]. However, these time points using an Accu-Check glucometer (Roche Di- BMP4 still has impressive positive effects on whole-body insulin agnostics, Basel, Switzerland). Saline (9 mg/mL), insulin (1.0 U/kg sensitivity in obese mice with increased insulin signaling in WAT, b.w.), or glucose solution (1 g/kg b.w.) was given as fluid replacement skeletal muscles, and liver; increased glucose uptake in skeletal or to compensate for high or low blood glucose levels. muscles and BAT; and also suppressive effects on key markers of hepatic gluconeogenesis. Thus, BMP4 and its antagonists are potential 5.4. Oxygen consumption and activity targets in obesity, insulin resistance, and type 2 diabetes. Oxygen (VO2) consumption, carbon dioxide production, and activity were recorded as previously described [8,20] by indirect calorimetry in 5. METHODS an INCA Metabolic system (Somedic, Hörby, Sweden) and by use of telemetry devices (MiniMitter, Bend, OR, USA). The measurements 5.1. Animals were performed 8e9 weeks after vector administration. Data were Male C57BL6/N mice (Taconic, Hudson, NY, USA) were group caged, recorded for 22 h at 20  C during the light and dark phase. Animals maintained on a 12-h lightedark cycle in a temperature- (þ21  C) and had ad libitum access to food and water. VO2 was normalized to body humidity-controlled room, and underwent weekly weighing. Food and weights. Data for the first hour were discarded to account for animal drinking water were administered ad libitum, and the mice were fed an acclimatization. MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 23 www.molecularmetabolism.com Original Article 5.5. Glucose uptake 5.6. In vivo insulin stimulation In vivo insulin stimulation was performed as previously described [21]. 5.5.1. In vivo protocol In brief, animals were fasted for 3 h and anesthetized with an i.p. Animals were fasted for 3 h and orally gavaged with glucose (1 g/kg). injection of pentobarbital (120 mg/kg). Insulin was injected (5 mU/g) Fifteen minutes after oral glucose load, w2.5 mCi of 2-deoxy-D-[14C]- through the inferior vena cava. After 5 min, tissues were harvested. glucose (14C-2DG, Perkin Elmer, Sweden) was administered via i.p. Tissues were homogenized and lysed at 4  C, and lysates were injection (t ¼ 0). Blood samples (w20 mL) were collected from the tip subjected to Western blotting to assess insulin-stimulated phosphor- of the tail at 16, 1, 5, 10, 20, 30, and 45 min, with plasma glucose ylation cascades. measured directly at each time point (Accu-Chek, Roche Diagnostics). Blood samples were immediately centrifuged and stored at 80  C 5.7. Serum chemistry pending analysis of plasma insulin (Crystal Chem, catalog #90080) and Insulin (Ultrasensitive Mouse Insulin ELISA kit #90080, Chrystal Chem plasma 14C-2DG activity. After 45 min, animals were killed by cervical Inc., Downers Grove, IL, USA), adiponectin (Quantikine ELISA, MRP300 dislocation and tissues were collected, weighed, placed in a 2-mL tube R&D Systems, Minneapolis, MN, USA), free fatty acids (kit #ab65341, (Sarstedt, catalog #72.694.007) containing 6 ceramic beads (Retsch, Abcam, Cambridge, UK), and serum amyloid A isoform 3 (SAA3) catalog # 5.368.0090), and snap frozen in liquid N2. (#EZMSAA3-12K, Merck Millipore, Darmstadt, Germany) were measured in serum according to the protocols provided by the 5.5.2. Measurement of plasma and tissue 14C-2DG activity manufacturers. A 5-mL plasma aliquot was added to 10 mL of scintillation fluid (Optiphase HiSafe 3, PerkinElmer, Sweden, catalog #1200.437) for 5.8. Liver triglyceride content determination of plasma concentration time course of 14C-2DG. Tissue Triglyceride content (Kit #10010303, Cayman Chemicals, Ann Arbor, levels of phosphorylated 14C-2DG (14C-2DG-6-P) were determined MI, USA) was measured in liver biopsies according to the protocol following H2O extraction. Ice-cold Milli-Q H2O (1.0 mL) was added to provided by the manufacturer. each tube, homogenized for 2  20 s at 5000 rpm (Precellys 24, Bertin, France) and an additional 5 min at 25 Hz (Mixer Mill 301, 5.9. Insulin stimulation Retsch, Germany), and centrifuged (10 min, 10,000 g at 4  C). A 350- mL aliquot of the supernatant was added to 10 mL of scintillation fluid 5.9.1. Isolated adipocytes for determination of total tissue 14C-2DG activity (14C-2DG-6-P þ 14C- Isolation was performed essentially as described [8,22]. Briefly, bi- 2DG). To isolate 14C-2DG-6-P activity, an additional 350 mL aliquot of opsies of SubQ WAT were washed, cut, and incubated with collage- the supernatant was loaded on to a solid-phase extraction (SPE) col- nase (1 mg/mL; SigmaeAldrich). The isolated adipocytes were filtered umn (Isolute 200 mg PE-AX, Biotage, catalog #503-0020-B). Prior to through a 250-mm nylon mesh and washed. Adipocyte diameter was placing the supernatant on the columns, each SPE column was acti- measured as previously described [5,22]. Adipocytes were stimulated vated (2 mL, methanol) and equilibrated (2 mL, Milli-Q H2O). SPE with 1 nM human recombinant insulin (Actrapid Penfill) for 15 min. columns were washed with 2 mL Milli-Q H2O (flow through discarded), and phosphates were eluted, collect with 2 mL 5% NaCl (aq), and 5.9.2. Human hepatocyte cell line (IHH) added to 10 mL of scintillation fluid. Plasma and tissue 14C activities Cells were seeded on gelatin-coated plates and cultured in Williams E were measured using liquid scintillation spectrometry (LS6500, medium (Life Technologies, MA, USA), supplemented with 10% fetal Beckman Coulter USA). bovine serum. For insulin signaling analysis, the cells were starved for 16 h with 0.1% serum and incubated with recombinant BMP4 (50 ng/ 5.5.3. Calculation of tissue-specific glucose clearance (Kg’) and mL), recombinant Gremlin1 (200 ng/mL), and/or insulin (10 nM), fol- utilization (Rg’) indices lowed by protein and RNA extractions. The clearance rate of 14C-2DG-6-P (Kg’), represents an index of the ability of a particular tissue to USE (metabolize) glucose, calculated 5.10. Immunoprecipitation using the relation: mDG 6P 5.10.1. Immunoprecipitation of BMP4 K 0g ¼ Pooled serum from control and BMP4 mice were precleaned using A/G ZT cDG ðt Þdt PLUS-Agarose beads (Santa Cruz Biotechnology, Dallas, TX, USA), immunoprecipitated with a BMP4 antibody (sc-73536, Santa Cruz 0 Biotechnology, Dallas, TX, USA), and analyzed by subsequent immu- where mDG-6-P is the activity of 14C-2DG-6-P per unit tissue mass noblotting. Beads and antibody were used as a negative control. Re- (dpm/g) at the end of the experiment, cDG(t) is the plasma concen- combinant BMP4 protein was used as positive control. tration time course of 14C-2DG and T is the total 14C-2DG exposure time (i.e., experiment length, 45 min). The glucose metabolic index or 5.10.2. Immunoprecipitation and acetylation analysis of PGC1-a estimate of glucose utilization (Rg’) is calculated from Kg’ by the Liver tissue lysates obtained from AAV8 control and AAV8 BMP4 mice equation: were precleaned with A/G PLUS-Agarose beads before immunopre- cipitation with a PGC1-a antibody (sc-13067, Santa Cruz Biotech- R 0g ¼ CG $K 0g nology, Dallas, TX, USA). Samples were then immunoblotted with same PGC1-a antibody or acetylated-lysine antibody (#9681, Cell where CG is the average plasma glucose concentration during the 14C- Signaling). Beads and antibody were used as a negative control. 2DG exposure time. 24 MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). www.molecularmetabolism.com 5.11. BMP reporter assay U.S., S.H., and J.M.H. wrote the final article, and all authors approved BMP-reporter assay experiments were performed essentially as pre- and contributed. viously described [23]. Briefly, stably transfected C2CL2-BRE cells, containing the pGL3-luciferase reporter construct fused to BMP- FUNDING specific response elements (BRE) from the Id1-promoter, were allowed to attach overnight before starvation in DMEM, 0.1% fetal This work was funded by the Swedish Research Council, the Swedish bovine serum (FBS) for 6 h. After starvation, cells were incubated with Diabetes Association, the Novo Nordisk Foundation, the Torsten DMEM, 0.1% FBS supplemented with 2% mouse serum for 20 h, after Söderberg Foundation, the Edgar Sjölund Foundation, the EFSD/Lilly which luciferase activity was measured. The quantified luciferase European Diabetes Research Programme, and the West Sweden ALF signal was related to total protein concentration. program. ACKNOWLEDGEMENTS 5.12. Immunoblotting Immunoblotting was performed essentially as described [3], with the We thank our colleagues at the Lundberg Laboratory for Diabetes Research for fruitful following primary antibodies: total OXPHOS Antibody Cocktail discussions. (ab110413, Abcam); p IR/IGF beta (3021), p AKT (9271), AKT (9272), p38 MAPK (9212), p p38 MAPK (9211), HSL (4107), p SMAD1/5/8 CONFLICT OF INTEREST (13820), and B tubulin (2128) (all from Cell Signaling Technology, Danvers, MA, USA); IRS1 (06e248, Upstate/Merck Millipore); IR beta None declared. (sc-711), p JNK (sc-6254), BMP4 (sc-73536), SMAD 1/5/8 (N-18)-R (sc 6031-R), GAPDH (sc-47724), and actin (sc-8432) (all from Santa APPENDIX A. SUPPLEMENTARY DATA Cruz Biotechnology, Dallas, TX, USA); and UCP1 (MAB6158, R&D Systems). Recombinant mouse BMP4 (R&D Systems) was used as a Supplementary data to this article can be found online at https://doi.org/10.1016/j. positive control where indicated. Quantifications were performed with molmet.2019.11.016. normalization against loading controls. 5.13. Measurement of islet size REFERENCES Paraffin-embedded paraformaldehyde-fixed pancreases were sectioned in 200-mm increments and stained with polyclonal guinea [1] Wang, Q.A., Tao, C., Gupta, R.K., Scherer, P.E., 2013. Tracking adipogenesis pig anti-insulin antibody (1:1000, Dako, Jena, Germany) and mouse during white adipose tissue development, expansion and regeneration. Nature monoclonal anti-glucagon antibody (1:2000, SigmaeAldrich). Sec- Medicine 19:1338e1344. ondary antibodies used were Alexa FluorÔ 488 anti-guinea pig and [2] Bowers, R.R., Lane, M.D., 2007. A role for bone morphogenetic protein-4 in Alexa FluorÔ 594 anti-mouse (Thermo Fisher Scientific, Waltham, MA, adipocyte development. Cell Cycle 6:385e389. USA), essentially as described [24]. Islet size was determined by [3] Hammarstedt, A., Hedjazifar, S., Jenndahl, L., Gogg, S., Grünberg, J.R., measuring the insulin and glucagon area of 50 islets per mouse. Gustafson, B., et al., 2013. WISP2 regulates preadipocyte commitment and PPARgamma activation by BMP4. Proceedings of the National Academy of 5.14. Gene expression analysis Sciences of the United States of America 110:2563e2568. Real-time RT-PCR gene expression analysis was performed as [4] Gustafson, B., Hammarstedt, A., Hedjazifar, S., Hoffmann, J.M., described [8] using the TaqMan System and normalized to 18S RNA. Svensson, P.A., Grimsby, J., et al., 2015. BMP4 and BMP antagonists regulate Gene-specific primers and probes were designed using the software human white and beige adipogenesis. Diabetes 64:1670e1681. Primer Express or purchased on demand (all from Applied Biosystems, [5] Gustafson, B., Smith, U., 2012. The WNT inhibitor Dickkopf 1 and bone Carlsbad, CA, USA). Sequences and assay IDs are available upon morphogenetic protein 4 rescue adipogenesis in hypertrophic obesity in request. humans. Diabetes 61:1217e1224. [6] Wang, G., Zhang, H., Zhao, Y., Li, J., Cai, J., Wang, P., et al., 2005. Noggin and 5.15. Statistical calculations and figures bFGF cooperate to maintain the pluripotency of human embryonic stem cells in Statistics were performed using Excel (Microsoft Office, Redmond, WA, the absence of feeder layers. Biochemical and Biophysical Research Com- USA) or IBM SPSS Statistics version 24 (IBM, Armonk, NY, USA). munications 330:934e942. Comparisons were performed using two-tailed Student’s t-test or [7] Topol, L.Z., Bardot, B., Zhang, Q., Resau, J., Huillard, E., Marx, M., et al., 2000. ManneWhitney nonparametric U-test (stated in figure legends). Re- Biosynthesis, post-translation modification, and functional characterization of sults are reported as means  SEM. Individual data points are included Drm/Gremlin. Journal of Biological Chemistry 275:8785e8793. in figures where n  10. Significances are indicated in figures ac- [8] Hoffmann, J.M., Grünberg, J.R., Church, C., Elias, I., Palsdottir, V., cording to the following: #p < 0.1, *p < 0.05, **p < 0.01, Jansson, J.O., et al., 2017. BMP4 gene therapy in mature mice reduces BAT ***p < 0.001. activation but protects from obesity by browning subcutaneous adipose tissue. Cell Reports 20:1038e1049. AUTHOR CONTRIBUTIONS [9] Qian, S.W., Tang, Y., Li, X., Liu, Y., Zhang, Y.Y., Huang, H.Y., et al., 2013. BMP4-mediated brown fat-like changes in white adipose tissue alter glucose U.S., S.H., J.M.H., and A.H. designed the experiments; J.M.H., J.G., and energy homeostasis. Proceedings of the National Academy of Sciences of A.H., J.B., T.K., and S.M. performed animal husbandry and mouse the United States of America 110:E798eE807. in vivo experiments; and J.M.H., S.H., and T.G. performed the in vitro [10] Modica, S., Straub, L.G., Balaz, M., Sun, W., Varga, L., Stefanicka, P., et al., work. F.B. and I.E. generated and characterized the viral vectors, and 2016. Bmp4 promotes a Brown to white-like adipocyte shift. Cell Reports 16: V.P. contributed to the energy expenditure and activity measurements. 2243e2258. MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 25 www.molecularmetabolism.com Original Article [11] Triplitt, C.L., 2012. Understanding the kidneys’ role in blood glucose regula- [19] Vilà, L., Elias, I., Roca, C., Ribera, A., Ferré, T., Casellas, A., et al., 2014. AAV8- tion. American Journal of Managed Care 18:S11eS16. mediated Sirt1 gene transfer to the liver prevents high carbohydrate diet- [12] Finck, B.N., Kelly, D.P., 2006. PGC-1 coactivators: inducible regulators of induced nonalcoholic fatty liver disease. Molecular Therapy Methods and energy metabolism in health and disease. Journal of Clinical Investigation 116: Clinical Development 1. 615e622. [20] Wernstedt, I., Edgley, A., Berndtsson, A., Fäldt, J., Bergström, G., [13] Stanford, K.I., Middelbeek, R.J., Townsend, K.L., An, D., Nygaard, E.B., Wallenius, V., et al., 2006. Reduced stress- and cold-induced increase in Hitchcox, K.M., et al., 2013. Brown adipose tissue regulates glucose ho- energy expenditure in interleukin-6-deficient mice. American Journal of meostasis and insulin sensitivity. Journal of Clinical Investigation 123:215e Physiology e Regulatory, Integrative and Comparative Physiology 291:R551e 223. R557. [14] Sharabi, K., Lin, H., Tavares, C.D.J., Dominy, J.E., Camporez, J.P., Perry, R.J., [21] Jing, E., Emanuelli, B., Hirschey, M.D., Boucher, J., Lee, K.Y., Lombard, D., et al., 2017. Selective chemical inhibition of PGC-1a gluconeogenic activity et al., 2011. Sirtuin-3 (Sirt3) regulates skeletal muscle metabolism and insulin ameliorates type 2 diabetes. Cell 169:148e160. signaling via altered mitochondrial oxidation and reactive oxygen species [15] Tilg, H., Moschen, A.R., Roden, M., 2017. NAFLD and diabetes mellitus. Nature production. Proceedings of the National Academy of Sciences of the United Reviews Gastroenterology & Hepatology 14:32e42. States of America 108:14608e14613. [16] Awazawa, M., Ueki, K., Inabe, K., Yamauchi, T., Kubota, N., Kaneko, K., et al., [22] Rotter, V., Nagaev, I., Smith, U., 2003. Interleukin-6 (IL-6) induces insulin 2011. Adiponectin enhances insulin sensitivity by increasing hepatic IRS-2 resistance in 3T3-L1 adipocytes and is, like IL-8 and tumor necrosis factor- expression via a macrophage-derived IL-6-dependent pathway. Cell Meta- alpha, overexpressed in human fat cells from insulin-resistant subjects. bolism 13:401e412. Journal of Biological Chemistry 278:45777e45784. [17] Oh, D.Y., Walenta, E., Akiyama, T.E., Lagakos, W.S., Lackey, D., [23] Herrera, B., Inman, G.J., 2009. A rapid and sensitive bioassay for the Pessentheiner, A.R., et al., 2014. A Gpr120 selective agonist improves insulin simultaneous measurement of multiple bone morphogenetic proteins. Identi- resistance and chronic inflammation in obese mice. Nature Medicine 20:942e fication and quantification of BMP4, BMP6 and BMP9 in bovine and human 947. serum. BMC Cell Biology 10. [18] Ayuso, E., Mingozzi, F., Montane, J., Leon, X., Anguela, X.M., Haurigot, V., [24] Lof-Ohlin, Z.M., Nyeng, P., Bechard, M.E., Hess, K., Bankaitis, E., Greiner, T.U., et al., 2010. High AAV vector purity results in serotype- and tissue- et al., 2017. EGFR signalling controls cellular fate and pancreatic organo- independent enhancement of transduction efficiency. Gene Therapy 17: genesis by regulating apicobasal polarity. Nature Cell Biology 19:1313e1325. 503e510. 26 MOLECULAR METABOLISM 32 (2020) 15e26 Ó 2019 The Author(s). Published by Elsevier GmbH. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). www.molecularmetabolism.com