Histone H4 lysine 16 acetylation controls central carbon metabolism and diet-induced obesity in mice

Noncommunicable diseases (NCDs) account for over 70% of deaths world-wide. Previous work has linked NCDs such as type 2 diabetes (T2D) to disruption of chromatin regulators. However, the exact molecular origins of these chronic conditions remain elusive. Here, we identify the H4 lysine 16 acetyltransferase MOF as a critical regulator of central carbon metabolism. High-throughput metabolomics unveil a systemic amino acid and carbohydrate imbalance in Mof deficient mice, manifesting in T2D predisposition. Oral glucose tolerance testing (OGTT) reveals defects in glucose assimilation and insulin secretion in these animals. Furthermore, Mof deficient mice are resistant to diet-induced fat gain due to defects in glucose uptake in adipose tissue. MOF-mediated H4K16ac deposition controls expression of the master regulator of glucose metabolism, Pparg and the entire downstream transcriptional network. Glucose uptake and lipid storage can be reconstituted in MOF-depleted adipocytes in vitro by ectopic Glut4 expression, PPARγ agonist thiazolidinedione (TZD) treatment or SIRT1 inhibition. Hence, chronic imbalance in H4K16ac promotes a destabilisation of metabolism triggering the development of a metabolic disorder, and its maintenance provides an unprecedented regulatory epigenetic mechanism controlling diet-induced obesity. Misregulation of chromatin has been linked to many conditions, including obesity, but the details remain unclear. Here the authors identify the H4 lysine 16 acetyltransferase MOF as a master regulator of glucose metabolism that is required for normal glucose uptake and fat storage.

C hronic diseases, in particular noncommunicable diseases (NCDs), are major health challenges for modern societies and account for over 70% of deaths worldwide. Metabolic disorders such as obesity and hyperglycemia are considered major risk factors for both development and maintenance of NCDs 1,2 .
Epigenetic regulators have been implicated in molecular regulation of metabolic pathways and are thereby responsible for keeping the metabolic state of the organism in check [3][4][5] . Central players of major metabolic reaction modules are transcriptionally 6,7 and post translationally 8 regulated by epigenetic regulators. For example, transcription factors such as PGC1α 9 , C/EBPα 10 , or CRTC2 11 responsible for control of metabolic networks are dynamically regulated by acetylation in response to metabolic stimuli. Consequently, alterations in levels of epigenetic regulators can translate into metabolite imbalances. In fact, both histone methyltransferase and histone deacetylase enzymes have been implicated in the regulation of obesity, including as factors mediating resistance to dietinduced obesity [12][13][14][15] . However, the contribution of a large and important class of chromatin-modifying enzymes, the lysine acetyltransferases (KAT), to these processes has not been systematically described to date.
MOF (also known as KAT8) is the major KAT responsible for acetylation of histone H4 at lysine 16 (H4K16ac), a unique histone mark that directly modulates higher-order chromatin structure 16 and the only acetylation mark known to be intergenerationally maintained 17 . Genome-wide association studies have identified MOF as a risk locus for multiple traits frequently associated with obesity and T2D, including waist circumference 18 , serum triglyceride (TG) levels 19 and body mass index (BMI) [20][21][22][23] . Whether changes in MOF levels are causative of these phenotypes as well as the nature of the underlying pathogenic mechanisms remain enigmatic.
In this study, we systematically interrogate the metabolic roles of MOF across the body using metabolite profiling of six metabolically important organs. Complete MOF knockout is deleterious for cells and tissues [24][25][26][27][28] and causes embryonic lethality in mice 29 . However, Mof haploinsufficiency is tolerated in mice 30 and humans 31 . Thus, we used Mof heterozygous mice (Mof +/− ) to dissect the metabolic consequences of reduced MOF activity in multiple tissues. Metabolomic analyses revealed that reduction of this central KAT results in global remodeling of primary nutrient metabolism, including altered carbohydrate and amino acid metabolism in multiple tissues. At the physiological level, these mice show disrupted glucose assimilation and impaired insulin response, suggestive of a predisposition to developing a metabolic syndrome. Indeed, when challenged with a high-fat diet (HFD), Mof +/− animals exhibit impaired fat storage and resistance to weight gain. We demonstrate that MOF-mediated H4K16ac is required for the transcriptional regulation of PPARγ and the downstream transcriptional network mediating glucose uptake and neutral lipid storage in white adipose tissue.
Taken together, we reveal MOF and its major target H4K16ac as the first KAT involved in diet-induced obesity resistance. Furthermore, we implicate MOF in regulation of systemic glucose and amino acid balance and suggest that reduced MOF levels predispose animals to a metabolic syndrome with features common to T2D.
Given that every organ has a unique set of metabolic requirements, we expected that altered Mof dosage would produce disparate effects on individual tissues. In spite of this, differentially expressed metabolites (DMs) in spleen, heart, and brain (the most strongly affected tissues) belonged predominantly to three major primary nutrient classes: amino acids, lipids, and carbohydrates as well as their derivatives (Fig. 1e, f, Supplementary Fig. 1d). In Mof +/− brain samples, the strongest effect was scored in the central carbon metabolism pathways, including deregulation of glycolysis and pyruvate metabolism (Fig. 1f, Supplementary Fig. 1d, e). The brain and heart, organs which are characterized by a high energetic expenditure, both showed an increase in metabolites related to energy metabolism (Fig. 1f). The heart and spleen showed marked reductions in multiple features that were annotated as proteinogenic amino acids and associated pathways (Fig. 1f). All tissues showed deregulation of multiple amino acid metabolism pathways (valine, leucine/isoleucine, glycine, serine threonine, alanine, aspartate, and glutamate), the non-oxidative branch of the pentose phosphate pathway, TCA, as well as vitamin B5 and B6 metabolism. Despite their inherent physiological differences, multiple organs show recurring patterns of deregulation in similar metabolic pathways following MOF reduction, serving as crucial indicators of systemic metabolite destablization in Mof heterozygous animals.
Next, we focused on DMs in the serum. Its connection to all organ systems make the serum valuable for obtaining a broad overview of circulating metabolites and identifying potential systemic effects. We scored a significant correlation between metabolites found in the serum and those found in the heart, SKM, and spleen ( Supplementary Fig. 1g). Detailed pathway enrichment analysis of serum DMs revealed two important trends. First, most of the significantly deregulated pathways are also deregulated in at least one other tissue set ( Fig. 1f, g). Second, metabolism of alanine, aspartate, glutamate, and pentose phosphates, which are consistently deregulated in the heart, spleen, and brain, also show disruption in the serum (Fig. 1g highlighted in red). Central carbon and amino acid metabolism were the most significantly altered classes of metabolite in the serum of Mof +/− mice (Fig. 1g).
Chronic metabolic rewiring as observed in Mof +/− mice can result in pathologies. For example, an altered serum glutamine to glutamate ratio has emerged as a reliable biomarker for predicting metabolic disorders such as T2D 32 . The circulating serum metabolite profile of Mof +/− mice showed depletion of a number of features matching the mass of glucogenic amino acids (glutamic acid, aspartic acid, and glycine), elevated glutamine and a concomitant increase in hexose (glucose) levels (Fig. 1g, h; Supplementary Data 1 and 2). This is indicative of a possible shift in organ nutrient utilization, probably triggered by impaired glucose consumption. Despite elevated levels of circulating hexose in the serum, we detected reduced hexose and increased lactate in Mof +/− brains, the highest glucose-consuming organ of the body. We also observed increased consumption of alternative fuel sources including acetate, primary amino acids (and derivatives) as well as ketone derivatives in Mof +/− hearts. Ketone bodies are generated in the liver for extrahepatic consumption 33 , providing an important source of nutrition under starvation and particularly in diabetic states 34 . Notably, livers of Mof +/− mice showed an accumulation of ketone derivatives ( Supplementary Fig. 1d-f), suggesting that energy metabolism is disrupted in Mof +/− animals.
Given a potential global deregulation of energy metabolism in Mof heterozygous animals, we decided to complement our analysis by concentrating on the visceral white adipocyte tissue (WAT) as the major energy reservoir of the body. To this end, we performed polar and apolar metabolomics from WAT (Supplementary Fig. 2a). Similar to our observations in brain, heart, and spleen, we detected an impairment of carbohydrate and amino acid metabolism in WAT of Mof +/− mice ( Supplementary  Fig. 2b, c). Remarkably, we found that Mof +/− adipocytes showed significant deregulation of the insulin pathway (Supplementary Fig. 2b), suggesting that Mof +/− animals may be more susceptible to glucose deregulation.
Our systematic metabolomic profiling unveils a role for MOF in controlling the levels of the major nutritional sources in multiple organs. Collectively, the overarching deregulated pathways in the sampled tissues were related to central carbon and amino acid metabolism. Amino acid levels can impact insulin secretion and are frequently associated with metabolic syndromes [35][36][37][38][39] . Thereby, we hypothesize that Mof haploinsufficiency may affect insulin sensing or signaling, leading to aberrant global glucose homeostasis and thereby predisposing Mof +/− animals to the development of metabolic disorders.
Mof haploinsufficient mice show impaired glucose sensing. To investigate the physiological consequences of the systemic central carbon metabolism imbalance, we metabolically phenotyped young (8-week-old) and old (26-week-old) Mof +/− animals. We detected no significant differences in body mass or composition between Mof +/− and Mof +/+ animals at either age (Fig. 2a). Oral glucose tolerance testing (OGTT) in young Mof +/− mice revealed impaired glucose tolerance, which was further aggravated in older animals (Fig. 2b). Mof +/− mice showed delayed glucose absorption and a significantly higher glycemic peak 15 min after oral glucose administration (Fig. 2c). Furthermore, overnight-fasted Mof +/− mice showed attenuated insulin production upon glucose challenge compared to littermate controls, despite secreting similar levels of insulin at steady state ( Fig. 2de). We conclude that Mof +/− animals show significant defects in glucose sensing and impaired glucose-stimulated insulin secretion.
We could recapitulate these glucose sensing defects in vitro using differentiated endocrine cells derived from the human PANC-1 cell line transfected with MOF siRNAs ( Supplementary  Fig. 3a, b). MOF-depleted endocrine cells show attenuated insulin production and reduced insulin mRNA expression upon glucose challenge ( Supplementary Fig. 3a). Hence, MOF levels likely influence insulin production and/or secretion upon glucose challenge.
Intriguingly, we observed a strong upregulation of insulin receptor substrate 1 (Irs1) mRNA in the pancreas, heart, and SKM of Mof +/− mice ( Supplementary Fig. 3c), which may reflect a compensatory mechanism to cope with impaired glucose sensing. Importantly, no change in the frequency of insulinproducing cells ( Supplementary Fig. 3d), nor in the size, morphology, endocrine organization or inflammation status of the pancreatic islets was observed in Mof +/− mice ( Supplementary Fig. 3e). Collectively, these data indicate that Mof reduction affects glucose-sensing rather than production. MOF is a transcriptional regulator 26,40 . To investigate whether the glucose sensing defect observed after Mof reduction was underpinned by specific gene expression changes, we took advantage of publicly available scRNA-seq datasets from different tissues of healthy wild-type mice 41 (Fig. 2f). These data were used to generate a gene regulatory network (GRN) 42 setting Mof as the regulatory gene node. This analysis allowed us to predict candidate genes that are regulated by Mof and subsequently classify them according to their metabolic pathways. This unbiased analysis identified enrichment of pathways related to insulin signaling and T2D predisposition in multiple organs, thereby predicting a correlation between Mof levels and diabetes predisposition on the gene regulatory level (Fig. 2f). We furthermore correlated the fold changes (FC) of DM detected in the liver and SKM of Mof +/− mice with those detected in the same tissues in prediabetic and diabetic human patients 37,43 . Indeed, we observed a positive correlation between the metabolites upregulated in Mof +/− mice and increasing during the progression from prediabetic to diabetic state in human patients 37 (Fig. 2g), with amino acid metabolism representing the strongest deregulated class in both species ( Supplementary Fig. 3f).
Collectively, these data identify defective glucose assimilation and impaired insulin secretion as the major pathological consequences of Mof haploinsufficiency. We hypothesize that Mof +/− animals are susceptible to development of a metabolic syndrome that has the potential to evolve into T2D.
Mof haploinsufficient mice are resistant to HFD-induced obesity. To test if reduction of MOF can indeed increase the risk of developing a metabolic syndrome, we challenged Mof +/− mice with nutrient stress. A key trigger for metabolic disorders is high caloric food intake. We therefore fed 7-week-old Mof +/− and Mof +/+ mice a HFD ad libitum. Strikingly, after 26 weeks of HFD, Mof +/− mice gained~50% less body weight compared to controls (Fig. 3a, b). This surprising resistance to weight gain in HFD was predominantly caused by the inability of Mof +/− mice to increase fat mass, while lean mass, food intake, total water content, and basal temperature were unaffected ( Supplementary Fig. 4a-c). Resistance to HFD-induced weight gain of Mof +/− mice showed no sex bias, with a similar phenotype observed in both male (Fig. 3a) and female ( Supplementary Fig. 4d, e) animals. We also found that MOF mRNA levels positively correlate with the body mass index (BMI) of healthy human subjects (Fig. 3c), suggesting that MOF-mediated regulation of weight gain may be conserved in mammals.
To infer the immediate response of Mof +/− mice to HFD, we focused on earlier time points. After 1 week of HFD challenge, Mof +/− mice showed an elevated glycemic peak 15 min post oral Fig. 1 Chronic reduction of MOF triggers systemic metabolic destablization. a Experimental overview of metabolite extraction scheme from different tissues of 8-week-old wild-type (+/+, gray, n = 4) and Mof heterozygous (+/−, green, n = 3) mice. See also: Supplementary Fig. 1a-c. b Partial least squares discriminant analysis (PLS-DA) scores plot of identified features across all seven analyzed tissue types. Two components covering maximum feature variance are shown. Mof +/+ -Hollow circles; Mof +/− -Filled circles. c Heatmap representing relative levels of different metabolites across all six tissues and the serum of both genotypes. Unsupervised k-means clustering is used to segregate the data into six major classes, closely mirroring tissuespecific metabolite signatures. See also: Supplementary Fig. 1d. d Stacked-plot showing the differentially regulated (black) and total number of metabolites (gray) found per tissue. Bold black labels represent tissues with significant enrichment p < 0.05 scored by two-sided Fisher's exact test. e Donut-plot showing the distribution of metabolite classes from significantly deregulated features, based on the highest probability annotations. Key metabolite classes are highlighted as bold. See also: Supplementary Fig. 1e-g. f Heatmap of the enriched deregulated pathways found in the brain (yellow), spleen (gray), and heart (blue). Color-scale depicts the number of contributing DMs in the pathway. KEGG pathway names denoted in black are common to all three tissues. g Sankey-plot showing deregulated pathways in serum and contributing metabolites (based on most probable annotation from deregulated feature ionMz). Metabolite colors depict direction of change, up-(red) and down-(blue) regulation (q value < 0.1). h Glutamine and glutamate Log 2 fold change found in serum of Mof +/− animals. Post hoc calculation of q-value was performed using the Storey method 97 and significance set as q value < 0.1 (details in "Methods"). See also: Supplementary Fig. 1 and 2; Supplementary Data 1 and 2. gavage ( Fig. 3d, e, Supplementary Fig. 4f), similar to their response under standard diet (SD) conditions (Fig. 2), as well as reduced insulin secretion (Fig. 3f). After 8 weeks of HFD, control animals showed elevated basal glucose levels and a higher OGTT glycemic peak (Fig. 3d, Supplementary Fig. 4f), accompanied by increased body weight. At the same time, Mof +/− mice showed higher glycaemia, but a sharp resistance to gaining weight or fat mass (Fig. 3b, Supplementary Fig. 4b, d, e). A long-term consequence of persistent elevation in serum glucose levels is the modification of hemoglobin to glycated hemoglobin (HbA1c).
The phenotypes exhibited by Mof +/− mice following the HFD challenge provide strong evidence for a metabolic predisposition of these animals toward a prediabetic state. To substantiate this hypothesis, we specifically searched for further evidence of multiorgan comorbidities that are frequently associated with a prediabetic state [46][47][48] (Supplementary Fig. 4g-m). HFDchallenged Mof +/− mice showed a significant decrease in pancreatic islet area compared to control animals ( Supplementary   Fig. 2 Mof haploinsufficient mice show impaired glucose tolerance and insulin secretion. a Barplots showing the total body weight (left), overall percentage of lean and fat mass, and total water of 8-week-old ("young", middle) or 26-week-old ("old", right) Mof +/+ and Mof +/− animals under steady state (open bars) or after 16 h of starvation (filled bars). Dots in the barplots represent independent animals and error bars represent ±SEM. n = 4-5/ group. b Blood glucose levels before and after oral glucose tolerance test (OGTT); 1 g/kg mouse; n = 5/group in 8-(young, left) and 26-week-old (old, right) n = 5/Mof +/+ and n = 4/Mof +/− male mice. Statistical analysis performed by two side, two-way ANOVA. p values are indicated in the figure. Error bars represent means ± SEM. c Glycemic peaks are represented as a ratio of peak glucose level (15 min after oral administration) over steady-state (fasting) level. Statistical analysis performed by two-sided Mann-Whitney test, *p = 0.05. Error bars indicate ±SEM. n = 4-5/group. Serum insulin levels at (d) steady state (fasting) and (e) at maxima (15 min after oral glucose administration), represented relative to fasting levels. Statistical analysis performed by two-sided Mann-Whitney test, *p = 0.05. Biological replicates: young: Mof +/+ n = 5; Mof +/− n = 4 and old: Mof +/+ n = 5 and Mof +/− n = 4. Error bars indicate ±SEM. f (i) UMAP-plots showing the major populations found in single-cell data from wild-type skeletal muscle (SKM), pancreas, white adipocytes (WAT), liver, kidney, and heart. Data were obtained in the PangloDB dataset (top panel). (ii) scRNA-seq data from indicated tissues was used to infer a gene regulatory network (GRN) using Mof as the regulatory node. Edges represent gene subsets. (iii) KEGG Gene Set Enrichment Analysis results depicted in barplots (colored by tissue type). Benjamini-Hochberg method applied to calculate FDR adjusted p values: ***p = 0.001, and *p = 0.05. See also Supplementary Source file for the original scRNA-seq data 41,114 . g Comparative analysis of the tendencies of identified mouse metabolites that were previously scored in transition from normal to prediabetic (NtoP) and subsequently to diabetic stages (PtoT) in patients; original data from 37 . Color bar indicates Spearman correlation scores (r); *p < 0.05. See also: Supplementary Fig. 3.  Supplementary Fig. 4f. e Glycemic peak, represented as a ratio of peak glucose level (15 min after oral administration) relative to steady-state (fasting) level. Error bars indicate ±SEM. Statistical analysis performed by one-way ANOVA, followed by Tukey test *p = 0.04. n = 5/group. f Peak serum insulin levels (15 min after oral glucose administration) relative to steady state (fasting) levels. Error bars indicate ±SEM. Statistical analysis performed by one-way ANOVA, followed by Tukey test *p = 0.02 n = 5/group. g Percentage of glycated hemoglobin (HbA1c) in SD (Mof +/+ n = 7; Mof +/− n = 5) or HFD (n = 5/group). Error bars indicate ±SEM. Statistical analysis performed by one-way ANOVA followed by Tukey test, *p = 0.02, ***p = 0.0004. h Free water in SD or HFD as a percentage of total body composition (n = 5/group). Statistical analysis performed as in (g). Statistical analysis performed by two-sided one-way ANOVA followed by Tukey test, *p = 0.04, **p = 0.006. Error bars indicate ±SEM. See also: Supplementary Fig. 4. Fig. 4g). HFD treatment resulted in increased inflammation in spleens and livers of both Mof +/+ and Mof +/− mice when compared to their respective SD controls (Supplementary Fig. 4h-j). However, Mof +/− mice showed a more severe inflammatory response in both organs ( Supplementary Fig. 4h-j).
Disruption in the balance between pro-inflammatory T helper 17 cells (Th17) and anti-inflammatory regulatory T cells (Treg) is a wellknown hallmark of chronic tissue inflammation [49][50][51] . Chronic metabolic disorders, such as diabetes, promote the expansion and polarization of Th17 cells [52][53][54] , and interference with this process has been shown to mitigate diabetic symptoms in patients 55 . Spleens of both control and Mof +/− animals displayed increased frequencies of Th17 cells and concomitant decreases in the Treg cell population under HFD conditions ( Supplementary Fig. 4k). In stark contrast, livers of HFD-fed Mof +/− animals displayed a significantly higher frequency of pro-inflammatory Th17 cells than control animals ( Supplementary Fig. 4l, m).
Notably, the liver is one of the main organs showing increased frequency of intra-tissue Th17 cells during diabetic onset 56,57 . To validate the tissue-intrinsic contribution of MOF to the aggravated inflammatory response in the liver microenvironment, we generated liver organoids from HFD-challenged Mof +/− mice and performed co-culture experiments with wild-type naïve T helper cells (CD4+ cells) ( Supplementary Fig. 4n, o). We scored increased polarization of activated T cells toward the Th17 profile specifically in cells that were co-cultured with liver organoids derived from HFD-fed Mof +/− mice ( Supplementary Fig. 4o). Our results suggest that the unique liver microenvironment of Mof +/− animals can polarize T cells toward a pro-inflammatory Th17 profile upon HFD exposure.
Taken together, Mof +/− animals show a striking resistance to HFD-induced gains in fat mass. The resistance to diet-induced obesity is accompanied by a marked deregulation of glucose homeostasis, elevated glycated-hemoglobin levels, and multitissue inflammation which combined are strong evidence for a diabetic predisposition caused by Mof haploinsufficiency.
MOF regulates the gene expression profile of adipose tissue. In order to understand the molecular mechanisms underlying the resistance to HFD-induced obesity, we performed mRNA-seq of visceral WAT, which are the main fat reserves in adult mice ( Fig. 4a, Supplementary Data 3). First, we compared the gene expression profiles from SD-fed 8-week-old Mof +/+ and Mof +/− mice to identify the differences at baseline prior to nutrient stress (Fig. 4a, DEG geno ). Although neither genotype displayed gross phenotypic changes under SD conditions, the PCA analysis could already discriminate a clear separation between the gene expression profiles of Mof +/+ and Mof +/− mice ( Supplementary  Fig. 5a). Mof +/− WAT showed significant downregulation of genes associated with inflammation, tissue remodeling, and angiogenesis ( Supplementary Fig. 5b-d)-all pathways associated with the development of obesity.
Next, to take both the impact of genotype and food intake into account, we cross-compared differentially regulated genes (p < 0.05) identified upon SD (Supplementary Fig. 5b) with those identified in mice fed a HFD for 26 weeks ( Supplementary  Fig. 5c). We performed unbiased clustering (based on expression similarity) on the differentially regulated genes in all four datasets combined (4,088 genes). Out of five identified clusters, cluster 1 (593 genes) showed the highest expression in Mof +/− mice independently of their diet and is therefore the most insightful for uncovering MOF-specific functions in WAT (Fig. 4b). Interestingly, these genes are associated with metabolic pathways such as endocrine disorders, glucagon signaling, parathyroid hormone, and insulin resistance (Fig. 4b, Supplementary Fig. 5d and To identify the pathways under direct regulation by MOF, we performed MOF ChIP-seq in WAT isolated from young Mof +/+ and Mof +/− animals fed a SD. We identified 19,753 peaks (mapping to 9,423 genes) in Mof +/+ WAT and 716 peaks (mapping to 444 genes) in Mof +/− WAT (Supplementary Data 3 and 4). In both datasets, MOF peaks were enriched at promoters (±1 kb). Control peaks can be grouped into two main clusters (kmeans = 2): cluster 1 (regions = 8,254 mapping to 4,439 genes) showing a mild decrease; and cluster 2 (regions = 11,486 mapping to 6,570 genes) showing a very clear reduction in the Mof +/− dataset (Fig. 4c, Supplementary  Fig. 5e). Overlapping the MOF-bound genes of both clusters with significantly downregulated genes identified in the Mof +/− WAT (SD) RNA-seq ( Supplementary Fig. 5b) revealed that the vast majority (81.5%) of downregulated genes are direct targets of MOF (Fig. 4d). Notably, pathway enrichment analysis revealed a strong association with central carbon metabolism and inositol phosphate metabolism/signaling (Fig. 4d). In WAT, phosphoinositides are not only important signaling molecules, but are key modulators of energetic balance in response to metabolic stress [58][59][60] . Hence, our genome-wide analysis uncovers a mechanistic link between MOFmediated gene regulation and the energetic demand for fat storage in adipose tissues. Besides the prominent metabolic signature, many direct MOF targets in WAT were also related to tissue inflammation ( Fig. 4d)-an important driver for obesity which appeared to be lost in the adipose tissue of Mof +/− animals.
A large proportion of MOF targets is regulated by deposition of H4K16ac. We cross-compared the list of genes downregulated in Mof +/− WAT (SD) against existing H4K16ac ChIP-seq profiles 61 and found a significant enrichment of H4K16ac domains at these genes ( Supplementary Fig. 5f), suggesting they are likely to be regulated by MOF via H4K16ac. Furthermore, we compared the list of downregulated genes with those scored in obese Trim28 +/ D9 haploinsufficient animals 15 . This analysis revealed a robust anti-correlation ( Supplementary Fig. 5g), implying that the metabolic disorder triggered by Mof haploinsufficiency is distinct from the one found in Trim28 +/D9 obese mice and therefore unique to MOF function.
To gain mechanistic insight into how Mof +/− animals react to HFD stress, we compared the transcriptional changes elicited by HFD treatment between Mof +/+ and Mof +/− animals ( Fig. 4a, Supplementary Fig. 5h, DEG diet ). In wild-type animals, HFD treatment results in activation of pathways associated with the development of obesity, such as carbohydrate metabolism, hormone and cytokine production, adipocyte differentiation, and lipid storage (Fig. 4e). Although Mof +/− mice showed hyperactivation of insulin signaling (Fig. 4b, qPCR validation in Supplementary Fig. 6d) and upregulation of lipid biosynthesis genes, genes involved in lipolysis or fatty acid uptake remained unaffected ( Supplementary Fig. 6d). In agreement with this, our metabolomic analysis revealed deregulation of the insulin pathway in WAT ( Supplementary Fig. 2b). We conclude that impaired lipid storage observed in WAT of HFD-induced obesity-resistant Mof +/− animals is neither caused by a defect in lipogenesis arising from impaired insulin signaling nor as a result of increased lipid breakdown or impaired lipid transport. In remarkable agreement with a role for MOF in regulating genes linked to energy balance and adipose tissue inflammation, these pathways are completely absent in HFD-treated Mof +/− animals (Supplementary Figs. 2 and 6).
To conclude, the transcriptome of Mof +/− WAT shows a downregulation of pathways associated with the development of obesity. ChIP-seq analysis in the same tissue further identifies direct MOF/H4K16ac-mediated gene regulation of metabolic pathways associated with carbon metabolism. Taken together, molecular analyses indicate a critical role for MOF in balancing energy expenditure and storage in adipose tissue.
Mof depletion results in impaired lipid storage and glucose uptake in adipocytes. Our genome-wide analyses prompted us to examine the morphology of WAT after HFD treatment. Consistent with the transcriptional deregulation of lipid storage and other obesity-related pathways (Fig. 4e), we found the WAT of HFD-treated Mof +/− animals to be devoid of neutral lipid droplets ( Fig. 5a, b, Supplementary Fig. 6a). Neither the overall number of mature adipocytes (Fig. 5c) nor the number of adipocyte progenitor cells ( Supplementary Fig. 6b) were significantly decreased, suggesting that impaired lipid storage and lipogenesis, rather than defective adipogenesis, are the predominant causes for obesity resistance in Mof +/− mice.
To narrow down the molecular mechanism for lipid storage defects observed in Mof +/− animals, we queried whether the genes in five selected gene ontology (GO) terms broadly related to adipocyte lipid metabolism were deregulated in WAT of Mof +/− animals under SD conditions. Among the selected GO terms, the only pathway significantly enriched for deregulated genes in Mof +/− WAT was glucose homeostasis (Fig. 5d and Supplementary Fig. 6c). This result was of particular significance as it closely supports our data showing MOF's involvement in regulating glucose homeostasis in vivo (Fig. 2a, b) and central carbon metabolism at the cellular level (Fig. 4d).
SKM and adipose tissue are the two major tissues responsible for absorbing glucose upon food intake, thereby regulating glycemia. In WAT, dietary sugars are converted into fatty acids and subsequently stored as TGs. The insulin-sensitive glucose transporter GLUT4 is responsible for 70% of glucose uptake in adipocytes and SKM 62,63 . Interestingly, we detected a significant decrease of Slc2a4 (also known as Glut4) levels in WAT and SKM of HFD-fed Mof +/− animals (Fig. 5e, Supplementary Fig. 6d), suggesting possible perturbations of glucose import in these organs.
Alterations in GLUT4 levels may have dramatic effects on adipocytes' ability to store fat. In fact, transgenic mice overexpressing GLUT4 showed adipocyte hyperplasia due to increased lipogenesis [64][65][66] . Conversely, Glut4-null mice show strongly reduced adipose tissue deposits 67 . Accordingly, we observe a significant increase in levels of Rbp4 RNA (Supplementary Fig. 6c), encoding a transporter protein known to be upregulated in Glut4-null animals and contributing to insulin resistance in obese and T2D patients [68][69][70] .
Since impaired lipid storage in WAT can be a direct outcome of defective adipocyte differentiation, we sorted mesenchymal stromal cells (MSCs) from Mof +/+ and Mof +/− animals to monitor their differentiation in vitro ( Fig. 5f, g). While Mof +/+ MSCs showed normal differentiation toward the adipocyte lineage, they failed to increase their lipid reserves after glucose and insulin challenge (Fig. 5g). Thus, we could confirm that Mof +/− mice are able to maintain normal adipocyte differentiation, but the adipocytes fail to respond to glucose.
To investigate tissue-intrinsic effects of Mof deletion on glucose flux in adipocytes, we took advantage of an in vitro conditional Mof knockout system based on Mof flox/flox CreERT2 tg (Cag-Cre/ Mof-iKO) mice 26 . We isolated mouse embryonic fibroblasts (MEFs) from these mice and differentiated them into adipocytelike cells (iAdipo) 71 (Fig. 5h, Supplementary Fig. 6e, f, see "Methods"). Upon 4-hydroxy-tamoxifen (4-OHT)-induced Mof deletion ( Supplementary Fig. 6g), Mof-iKO adipocytes showed decreased Glut4 mRNA and GLUT4 protein levels ( Fig. 5i and Supplementary Fig. 6h) and hyper-activation of insulin signaling, fully recapitulating our observations from adipocytes in vivo (Fig. 4b, Supplementary Fig. 6i). Notably, along with low Glut4/ GLUT4 levels Mof-iKO adipocytes showed impaired insulinmediated glucose uptake (Fig. 5j). In addition, Mof-iKO adipocytes showed glucose-fueled elevation in extracellular acidification rate (ECAR), while glycolytic flux (Fig. 5k) and oxygen consumption rate ( Supplementary Fig. 6j) were significantly reduced (Fig. 5k) in Mof-iKO iAdipo. This difference in glycolytic rates was even more pronounced after insulinstimulated glucose uptake in these cells. While control cells showed insulin sensitivity and a concomitant elevation in glycolytic rate, this response was abrogated in Mof-iKO adipocytes (Fig. 5k). Our in vitro analyses therefore support the hypothesis that Mof reduction results in intracellular glucose deprivation in adipocytes.
Mof +/− animals show dramatic defects in lipid storage (Figs. 3a and 5a). Whilst lipid biosynthesis and transport genes are unaffected, we detect deregulation of glucose homeostasis genes. Moreover, Mof +/− animals show a downregulation of Glut4 in WAT, resulting in impaired glucose uptake.
MOF regulates fat storage by regulating the Glut4 transcription network. Glucose uptake is a requirement for neutral lipid storage in adipose tissue. To understand the interplay between defects in lipid storage and glucose uptake in Mof +/− animals, we challenged in vitro differentiated adipocytes (generated as in Fig. 5h) with glucose and insulin and performed neutral lipid staining. Similar to our observations in vivo, Mof-iKO adipocytes failed to increase their lipid storage upon combined glucose and insulin challenge (Fig. 6a, Supplementary Fig. 7a). Treating wildtype cells with MG149, an inhibitor with efficacy against MYST family HATs, including MOF 72 , mimics the fat storage (  Fig. 6g). We therefore treated Mof-iKO cells with Ex-527, an inhibitor of the HDAC SIRT1 responsible for deacetylation of H4K16ac [73][74][75] . Remarkably, Ex-527 treatment not only rescued the H4K16ac levels ( Supplementary Fig. 7e), but also restored lipid storage in Mof-iKO cells (Fig. 6b, Supplementary Fig. 7a), indicating that chromatin regulation of H4K16ac is indeed essential. Notably, Ex-527 treatment significantly boosted lipid storage even in control cells. Moreover, we found that control iAdipo showed a significant increase in H4K16ac levels after glucose and insulin treatment, while Mof +/− cells failed to do so (Fig. 6c).
Lipid storage defects upon deletion of MOF could be rescued by ectopic expression of either Mof or Glut4, while the expression of a Mof catalytic mutant (E350Q) was unable to restore H4K16ac levels nor lipid storage defects (Fig. 6d, e, Supplementary Fig. 7a, f-h). These results imply that the impaired capacity of Mofdeficient adipocytes to store fat is, at least in part, due to defective glucose uptake by GLUT4. Treating Mof-iKO adipocytes with chloroquine-a drug that enhances GLUT4 translocation and fusion with the plasma membrane in response to insulin 76 -failed to restore fat storage in adipocytes (Fig. 6f, Supplementary  Fig. 7a). This finding establishes that a decrease in GLUT4 levels, rather than defects in its intracellular transport, is the primary cause for reduced storage of neutral lipids in Mof-iKO adipocytes.
Given that Mof depletion results in downregulation of Glut4 (Figs. 5e, 6g, Supplementary Fig. 6d), we next investigated whether the transcriptional network regulating Glut4 expression was affected in Mof +/− animals. We observed significant downregulation of the Glut4 transcriptional activators Pparg, Ppargc1a (also known as: Pgc1α) and Mef2c 77,78 , as well as of Glut4 itself in WAT and SKM of SD-fed Mof +/− mice (Fig. 6g,  Supplementary Fig. 7i-k). This pointed toward a perturbation of   the whole transcriptional network regulating glucose import. ChIP-qPCR in WAT detected no enrichment of MOF at the promoters of Glut4, Pgc1α or Mef2c (Fig. 6h), suggesting that expression of these genes is not directly regulated by MOF. Conversely, the gene coding for PPARγ (Pparg), the major regulator of adipocyte biology and glucose homeostasis 79 , was downregulated in both Mof +/− WAT and glucose/insulinchallenged Mof-iKO cells ( Fig. 6 and Supplementary Fig. 7i, j). In contrast to Glut4, Pgc1α, and Mef2c, MOF was bound to the promoter of the main isoform of Pparg and this binding was significantly reduced in Mof +/− WAT (Fig. 6h). Moreover, using the Mof-iKO model, we confirmed that MOF and H4K16ac at the Pparg promoter increased upon glucose and insulin treatment (Fig. 6i). We conclude that expression of Pparg via MOFmediated deposition of H4K16ac regulates Glut4 expression in response to insulin signaling upstream of the transcriptional activators Pgc1α and Mef2c (Fig. 6g). Intriguingly, we observe a specific effect on the glucose homeostasis transcriptional network rather than a global deregulation of all PPARγ target genes in Mof +/− WAT ( Supplementary Fig. 7l, m).
Finally, treatment of Mof-iKO adipocytes with thiazolidinedione (TZD; a potent PPARγ agonist widely used to treat T2D patients 80 ) upon insulin/glucose challenge partially rescued lipid storage defects in Mof-depleted cells (Fig. 7a) and restored the expression of Pgc1α and Glut4 (Fig. 7b, c). Of note, TZD treatment had no effect on bulk H4K16ac levels in these cells (Fig. 7d). siRNA-mediated knockdown of Glut4 in wild-type cells phenocopied lipid storage defects observed in Mof-iKO cells, as expected ( Fig. 7e-h). In contrast to Mof-iKO cells, however, TZD treatment was not able to rescue lipid storage in Glut4 knockdown cells (Fig. 7e-h).
Together our data show that MOF orchestrates glucose uptake and subsequent lipid storage by transcriptional regulation of Pparg. In the absence of a MOF-elicited PPARγ response, the activation of the downstream transcriptional network mediating glucose uptake cannot be triggered (Fig. 7i). The lack of glucose influx, in turn, results in reduced neutral lipid storage in adipocytes. We conclude that failure in glucose uptake in adipose tissue is a major consequence of Mof haploinsufficiency, resulting in resistance to HFD-induced obesity in these animals (Fig. 8).

Discussion
Our work identifies a fundamental role for MOF in regulating systemic metabolic homeostasis. Using organ-specific metabolomics, we identify global destablization of carbohydrate and amino acid metabolism in the serum and multiple tissues of Mof haploinsufficient mice. These metabolite disruptions predispose Mof +/− mice to developing metabolic disorders, as confirmed by physiological indicators including impaired glucose assimilation, reduced insulin secretion and elevated glycated-hemoglobin (Hb1ac) levels in these animals. When challenged with prolonged HFD stress, Mof +/− mice showed surprising resistance to fat mass gain which prevented them from developing obesity. Adipose tissue of HFD-challenged Mof +/− mice is devoid of neutral lipid storage as a consequence of impaired glucose uptake in this tissue. Molecularly, we established that MOF-mediated H4K16ac is a master regulator of Pparg expression and is thereby responsible for sustaining glucose uptake and lipid storage in adipocytes. Therefore, we provide the first experimental evidence of a direct causal relationship between a KAT, glucose homeostasis, lipid deposition and obesity.
Chronic Mof depletion promotes a pathological insulin-glucose balance. The resistance of Mof +/− mice to gaining weight upon HFD stress represents a double-edged sword. Despite the apparent benefit in ameliorating obesity, this phenotype occurs at the expense of normal central carbon and amino acid metabolism. These metabolic alterations are, in turn, associated with defective glucose sensing and failed glycaemic homeostasis.
Insulin-dependent glucose homeostasis depends on the interplay between multiple tissues. In a healthy individual, spikes in glucose levels are rapidly sensed by pancreatic β cells, promoting insulin secretion. As a consequence, insulin reaches tissues such Fig. 5 Mof depletion impairs glucose uptake and fat storage in adipocytes. a Immunofluorescence images of WAT after 26 weeks on HFD (SD controls are shown in Supplementary Fig. 6a). DAPI (blue) and neutral lipid staining by BODIPY-493 (green). Highlighted regions show recidual fat storage. Scale bar: 50 µm, enlarged view: 5 µm. b Violin-plot showing the MFI quantification of (a). Error bars indicate ±SEM. Statistical analysis was performed by oneway ANOVA followed by Kruskal-Wallis test, *p = 0.04, ****p = 0.0001. n = 3 biological replicates. Dotted lines inside the violin plots show the quartiles and dashed lines depict the medians. Dotted horizontal line marks the zero. c Violin-plot showing total quantification of adipocytes per section (n = 3, 2-3 sections per animal). Normality was scored as in (b). Statistical analysis using Mann-Whitney test showed no significance. Dotted lines show the quartiles and dashed lines depict the median. d Barplot showing enrichment test for target GO terms and DEG in WAT. The black box represents differentially regulated genes in WAT of Mof +/− animals on SD. Statistical test was performed using the absolute numbers of genes and p values were scored by the two-sided Fisher's exact test. The red bar highlights the only significantly enriched pathway. See also: Supplementary Fig. 6b. as WAT or SKM, triggering expression and translocation of GLUT4 to boost glucose uptake and reduce extracellular glucose levels. Insulin release is transient and ceases immediately once basal glucose levels are reinstated. We propose a model in which insulin secretion in Mof +/− animals is not completely blocked, but strongly impaired upon glucose challenge, arguing that glucose sensing is the major defect in these animals (Fig. 8). If glucose is not appropriately sensed in β cells, serum glucose levels will remain high. Stimulating glucose uptake by tissues via insulin signaling represents a potential coping mechanism against this persistent hyperglycemia. Mof +/− animals show a hyper-insulin response in the pancreas, WAT, SKM, and heart. However, the ability of intracellular insulin signaling to produce the desired attenuating effect on serum glucose levels is severely hampered by inadequate levels of the GLUT4 transporter in the membranes of these cells.
Recently, pancreas-specific Mof depletion was reported to promote expansion of β cells at the expense of α-cells, ultimately leading to reduced glycemia upon glucose challenge 81 . Surprisingly though, insulin tolerance tests from the same study indicated significant resistance to regulating blood glucose levels upon insulin treatment. This impaired response agrees with our observations that chronic MOF reduction causes impaired glucose sensing due to downregulation of glucose transporters. An independent study has shown that Mof deletion in the mediobasal hypothalamus reduces hypothalamic polysialic acid (PSA), resulting in overfeeding and exacerbated diet-induced obesity 82 . In contrast, Mof +/− mice show normal feeding behavior and lean mass gain equal to littermate controls. We find that the resistance of Mof +/− mice to diet-induced obesity is a direct result of the failure of MOF-mediated glucose uptake in adipose tissue. Furthermore, given the tissue-specific functions of MOF and H4K16ac it is possible that while altered PSA levels are the primary defect upon complete knockout in the hypothalamus, the haploinsufficient model reveals more systemic effects of Mof reduction.
MOF regulates glucose uptake in adipocytes via the insulindependent glucose transporter network. We established that MOF orchestrates glucose uptake and subsequent neutral lipid storage by regulating the transcriptional network upstream of Glut4 in adipocytes, in particular by binding and promoting expression of Pparg. Remarkably, both Mof +/− and Pparg +/− mice are protected against fat mass gain when fed a HFD 83 . The phenotypic similarities between the Pparg and Mof haploinsufficient mouse models strongly support a functional interaction between MOF and the downstream PPARγ transcriptional axis in adipose tissue.
Treatment with the PPARγ inhibitor TZD partially restores the expression of Pgc1α, Mef2c, and Glut4 as well as lipid storage in Mof-depleted adipocytes. Given that TZD rescue is incomplete, it remains unclear whether pathways such as increased thermogenesis or browning of adipose tissue may additionally be contributing to obesity resistance in Mof +/− mice. MOF ChIPseq revealed that Mof +/− WAT retained~5% of mapped gene targets found in Mof +/+ WAT, indicating that MOF loss likely affects multiple other pathways beyond glucose sensing.
Another intriguing question is whether MOF depletion alters the microbiome. Gut microbiota are known to exert a critical role in metabolic diseases including obesity, partially by eliciting local immune reactions 84 . Considering that the hepatic microenvironment of Mof +/− mice promotes Th17 cell polarization in the liver, it would be important to test whether a similar immune response occurs in the gut.
Altered MOF levels may be implicated in noncommunicable diseases. Our systematic metabolomic analysis showed differential expression of biomarkers for multiple NCDs following Mof depletion, indicating that diabetic predisposition is only one potential angle of interrogating this data. For instance, Mof +/− hearts show deregulation of glyoxylate/dicarboxylate, aminoacyl-tRNA biosynthesis and vitamin B6 metabolism, which have all been associated with coronary heart disease [85][86][87] . Consistent with this finding, mice with conditional Mof knockout in the heart succumb to heart failure and premature death 25 . Our analysis also revealed deregulation of vitamin B6 metabolism in the brains of Mof +/− mice. Vitamin B6 serves as an important cofactor for many metabolic processes such as protein folding, amino acid biosynthesis, and glycogen degradation 88 . Perturbations in vitamin B6 are associated with several pathologies including autoimmune inflammatory disorders 89 and stroke 90 . These heart-and brain-specific metabolic disturbances suggest that MOF depletion might also facilitate the onset of other, more tissue-localized NCDs.
MOF in human metabolism. Taken together, this work identifies the lysine acetyltransferase MOF as a key regulator of glucose and Dashed horizontal lines mark the mean area of controls. a Control, Mof-iAdipo or Mg149-treated (50 nM) control cells at baseline ("-", upper panel) or after 15 min of insulin/glucose challenge ("+", lower panel). Statistical analysis was performed using the raw data and one-way ANOVA followed by Kruskal-Wallis comparison test, ****p = 10 −16 . b Control and Mof-iAdipo cells treated with Ex-527 (200 nM) after insulin/glucose challenge. Statistical analysis was performed using the raw data and one-way ANOVA followed by Kruskal-Wallis comparison test, ***p = 0.006. See Supplementary Fig. 7a for baseline images. c Scatter-plot depicting the H4K16ac MFI in control and Mof-iKO iAdipo with (open circles) or without (filled circles) insulin and glucose challenge. Each dot represents a single iAdipo. Statistical analysis was performed using the raw data and two-sided one-way ANOVA followed by Kruskal-Wallis comparison test, **p = 0.014, ****p = 10 −16 . Number of biological replicates n = 4. d Control and Mof-iAdipo cells at baseline, following overexpression of Glut4, or following ectopic expression of wild-type Mof (wt-Mof) or Mof catalytic mutant (Mof-E350Q) challenged with insulin/glucose (+Ins/Gluc). Statistical analysis was performed using the raw data and two-sided one-way ANOVA followed by Dunn's multiple comparison test, ***p = 0.001. Note that the data for untreated wild type (n = 1299) and Mof-iKO (n = 4644) is identical to the data depicted in (a). e Heatmap showing the H4K16ac levels from wild type and Mof-iKO upon ectopic expression of Glut4, wt-Mof or Mof-E350Q with or without glucose and insulin treatment. Statistical analysis was performed using the raw data and two-sided two-way ANOVA followed by Sidak multiple comparison post test, ****p = 10 −16 . f Control and Mof-iAdipo cells treated with chloroquine at baseline or upon insulin/glucose challenge. Statistical analysis was performed using the raw data and two-sided one-way ANOVA followed by Dunn's multiple comparison test, ***p = 0.0001, ****p = 10 −16 . amino acid metabolism and adipose tissue homeostasis. Our findings have important clinical implications for understanding the contribution of epigenetic regulators to metabolic disorders such as obesity, cardiovascular disorders, and T2D as well as cancer. Importantly, alterations in MOF levels can serve as an important biomarker to predict future risk of developing metabolic syndromes. We envision that advances in cell-type-specific delivery of small molecule inhibitors aimed at modulating acetylation levels could be an exciting avenue for future research in the battle against obesity and its comorbidities.

Methods
Cell culture Primary mouse embryonic fibroblast culture. Primary MEF cells were isolated from embryos of wild-type and Cre-ERT2tg Mof flox/flox (Mof-iKO) mice at 13.5-15 days after mating. Embryos were removed from the uterus and shortly rinsed with 70% EtOH and ice-cold PBS. The chorion, placenta, heart, liver, yolk sac, and head were dissected out from the embryo. The remaining body was transferred onto a new dish and minced finely. Afterwards, 1 ml Trypsin/1 mM EDTA was added and incubated for 30 min at 37°C. Then, Dulbecco's modified Eagle's medium (DMEM, Life Technologies, 31966047) supplemented with 10% FCS and 1% penicillin-streptomycin (DMEM-10%) was added and the sample was centrifuged for 5 min at 1000 × g. The pellet was resuspended in DMEM, plated in 10-cm petri- Fig. 8 Mof Haploinsufficiency Protects against Diet-induced Obesity. Schematic overview describing the proposed mechanism by which chronic reduction of MOF results in global remodeling of metabolism toward diabetic predisposition. In wild-type cells, insulin binds to the insulin receptor, which will trigger a cascade of phosphorylation events culminating in the phosphorylation of GLUT4 inside storage vesicles. These will be transported to the cytoplasmic membrane to regulate glucose import. In parallel, insulin triggers the transcription of Glut4 predominantly through the transcriptional network downstream of PPARγ and PGC1α. In healthy adipose tissue and skeletal muscle (SKM), MOF mediates glucose uptake by positively regulating gene expression of the nuclear hormone receptor Pparg, thereby triggering the downstream transcriptional network resulting in increased levels of the major insulin-dependent glucose transporter GLUT4. On the other hand, Mof haploinsufficient animals show impaired insulin secretion upon glucose challenge, leading to severe destablization of glucose homeostasis. Whilst insulin signaling and GLUT4 vesicle transport in Mof-depleted cells are functional, the transcriptional network regulating Glut4 expression is impaired, resulting in reduced glucose import. Limitation of intracellular glucose is responsible for impaired neutral lipid storage in Mof +/− adipocytes. This study identifies the epigenetic regulator MOF as the first histone acetyltransferase to regulate the onset of diet-induced obesity. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-26277-w ARTICLE dishes and cultured at 37°C with 5% CO2 in a controlled incubator. Confluence was reached after 2-3 days in culture and part of the cells were frozen as passage 1, while the rest were used for MEF expansion and genotype confirmation.
In vitro differentiated adipocyte culture. Primary MEFs were differentiated to adipocytes using the protocol of Fei et al. 71 with minor modifications. In brief, 1 × 10 6 MEFs (passage 3 or 4) were plated in six-well plates in DMEM-10% supplemented with 0.5 mM methylisobutylxanthine (Sigma, I7018), 1 M dexamethasone (DEX), 10 g/ml insulin and 10 M troglitazone (Tocris Bioscience, 3114/10). This media was kept for 3 days, and then replaced with DMEM-10% for another 2 days. After 5 days in culture 4-OHT was added to the culture twice for Mof deletion. For downstream analysis, cells were harvested before differentiation (MEFs) and after differentiation (iAdipo) followed or not by Mof deletion.
PANC-1 cell culture. Human pancreatic cancer cell line PANC-1 was obtained from the cell line depository of the BIOSS excellence cluster (Freiburg, Germany) and cultured in DMEM containing 10% FCS (DMEM-10%), at 37°C with 5% CO 2 in a controlled incubator.
In vitro differentiated pancreatic islet culture. PANC-1 cells were differentiated into ductal and islets-like cells following published protocol 91,92 . In brief, PANC-1 were seeded at 1 × 10 5 cells/ml density in 12-well plates with DMEM-10% supplemented with 1% BSA, transferrin (Sigma, 5.5 μg/ml) and insulin-like growth factor 1 (10 ng/ml, PeproTech) for a period of 7 days. Media was changed every other day and differentiation confirmed by qRT-PCR assays.
Liver organoid formation and maintenance. Livers from 26-week-old HFD-treated Mof +/+ and Mof +/− animals were harvested and placed in cold PBS. Ductal liver organoids were generated according to published protocol 93 with modifications. In brief, livers were minced thoroughly and washed three times in cold wash buffer (GlutaMAX, Gibco #35050-038, 1% FCS, 100 U/ml penicillin, and 100 μg/ml streptomycin (Gibco #15140-122)) to remove fat. Tissue was digested in prewarmed digestion media (wash buffer, 0.25 mg/ml Collagenase (Sigma Aldrich #C2674), 0.125 mg/ml (Life Technologies #17105-041)) at 37°C on a rotating wheel for 1 h. Tissue was pelleted, supernatant discarded and 1 h digestion was repeated using fresh digestion buffer. Ductal structures were pelleted and washed three times in cold wash buffer. Ductal fragments were handpicked and seeded into Matrigel ® (Corning, #356231) domes (50 fragments per 50 μl of Matrigel ® ). Upon solidification of the basement matrix, domes were overlaid with isolation medium and incubated for 3 days under standard cell culturing conditions (37°C with 5% CO 2 ) after which media was exchanged to expansion media 93 . Ductal organoids were maintained and a 14 day hepatocyte differentiation protocol was performed as described in ref. 93 .
Liver organoid and CD4+ T cell co-culture Fluorescence-activated cell sorting. For co-culture, wild-type thymi were isolated and thymocytes retrieved by mechanical filtering on a 70-μm cell strainer. The cells were resuspended in PBS, centrifuged at 500 × g for 5 min and the pellet was resuspended in ACK lysing buffer for erythrocyte cleaning. After 10 min, PBS was added to the cellular suspension followed by centrifugation. The remaining leukocytes were resuspended in FACS buffer containing zombie-dye for dead cells, fluorophore-conjugated primary antibodies against CD3 (PERCP-Cy5.5), CD4 (A700), and CD8a (PE) and kept on ice for 30 min. Afterwards, cells were resuspended in a 1x v/v FACS buffer and centrifuged at 3000 g for 30 s. The pellet was resuspended in FACS buffer and naive CD8a+ cells sorted using the BD Aria-FACS Fusion II using a nozzle of 70 μm and maximum flow of 15,000 events/s rate. A 95% pure population was sorted and used for co-culture.
Organoids and naive Th cell co-culture. Sorted purified naive T cells were then coculture with differentiated hepatic organoids at a ratio of 2:1 (liver organoids: T cells) in the presence of IL-2 (10 ng/mL, Peprotech 212-12). T cells were further separated into two groups (a) stimulated with CD3:CD28 dynabeads (Thermo Fisher, 11456D) in a 1:1 ratio or (b) no stimuli. Five days prior to the co-culture experiment differentiated hepatic organoids were cultured in AIM V™ media (Thermo Fisher) supplemented with 3 μM DEX. After 5 days of co-culture, cells were harvested and prepared for flow cytometry analysis.
For cytokine flow cytometry evaluation we added 1X Brefeldin-A for at least 10 h prior to analysis. The cells were then transferred to 1.5 ml tubes and centrifuged at 3000 × g for 30 s. The remaining pellet was washed 2× with FACS buffer and fluorophore-conjugated primary antibodies against CD4 (PE), and viability dye (zombie-dye aqua or green) was added and incubated for 30 min incubation on ice. Then the sample was centrifuged and pellet washed twice with 1× wash buffer (Foxp3/Transcription Factor Staining Buffer Set, Affymetrix, eBioscience, USA, 00-5523-00), followed by 40 min incubation on ice with the "Perm/Fix" solution (1:4 dilution) (Foxp3/Transcription Factor Staining Buffer Set, Affy-metrix, eBioscience, USA, 00-5523-00). Samples were washed once with 1× wash buffer and twice with FACS buffer followed by incubation with fluorophoreconjugated primary antibodies against IFN-gamma (FITC), IL-17A (A647), RORgammat (PERCP-Cy5.5), FoxP3 (AF700), and/or IL-4 (BV421), IL-10(BV421) as indicated in the figures. The cells were incubated for~16 h, washed twice with FACS buffer and immediately acquired using the BD Fortessa II or Fortessa I cytometer. Data were next analyzed using the FlowJo v.10 software.
Animal husbandry. Animals were maintained on a 14/10-h light-dark cycle under controlled humidity and temperature. For SD experiments the animals were provided with a normal chow diet 15% fat (Ssniff GmbH), while obese animals were fed with HFD (60% kcal% fat, Research Diet). Both groups were fed ad libitum, had free access to water and refinement methods as nesting material were employed. The gain of total body mass was recorded every second week. Every mouse strain in this study was backcrossed with C57BL/6J mice. All animal procedures are in agreement with and approved by the responsible Animal Welfare Committees (Regierungspräsidium Freiburg, Karlsruhe/ Germany, license, G-17/150), conforming to the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health (publication No. 85-23, revised 1996).

Mouse phenotypic assays
Oral glucose tolerance test and plasma insulin measurements. For the OGTT, mice were fasted overnight (16 h), before starvation the total body weight, fat, and lean mass was recorded as well as free and total water contribution using the EchoMRI™ scan. Then, after 16 h the animals were again scanned using the EchoMRI™ scan and basal blood glucose was measured by Accu-check stripes and the OneTouch Vita blood glucose meter. Mice were given glucose bolus (1 g/kg body weight) by oral gavage. Blood glucose levels were measured using a OneTouch Vita blood glucose meter at −5 (to avoid stress bias), 0, 15, 30, 45, and 60 min after glucose. Blood drawn using heparinized capillary tubes was centrifuged at 2000 × g for 15 min at 4°C. Plasma obtained was used for insulin level measurement by ELISA (Mercodia Ultrasensitive Mouse Insulin Kit).
Glycated-hemoglobin measurement. For glycated-hemoglobin (HbA1c) animals were euthanized by asphyxiation and blood was taken using EDTA-coated syringes. Whole blood was used and HbA1c levels measured by colorimetric assay using hemoglobin A1c human kit (Crystal Cheme, 80099).
Immunofluorescence stainings and image acquisition. WAT, liver, and spleen, whole organs were dissected, weighed, and embedded into OCT (Tissue-Tek, SA62550-01) and frozen gradually: first the blocks were left on a dry-ice board for 15 min followed by liquid nitrogen submersion. OCT blocks were sectioned with a cryotome (Leica) at −20°C into 7-10-μm slides.
For BODIPY staining, WAT specimens were blocked in 2% calf serum PBS-T for 15 min at RT before incubating with the BODIPY probe, BP (Invitrogen). Each slide was rinsed two times with PBS-T, for 5 min each. Specimens were incubated with BP diluted in PBS (1:300) for 30 min at 37°C in the dark. After rinsing with PBS-T, VectorShield with DAPI and coverslip were mounted. While for GLUT4 staining WAT sections were embedded in paraffin as describe for H&E staining, sections were deparaffinized and rehydrated stepwise using an EtOH gradient (2 × 100% xylene 5 min each, 2 × 100% ethanol 5 min each, 2 × 95% ethanol 5 min each, 70% ethanol for 5 min) and rinsed in dH 2 O for 5 min.
For liver and spleen images, cryosections were fixed in cold methanol for 20 min on ice followed by 3x PBS-T wash, 10 min each. Samples were then blocked with a 2% BSA-PBS-T solution for 2 h at RT. After, specimens were washed as above and liver sections were incubated with the Phalloidin-A647 probe for 30 min at RT. Then liver samples were washed 3× in PBS-T and mounted as above. While spleen sections were incubated with fluorophore-conjugated primary antibodies against IgD (FITC, eBioscience, 115993), F4/80 (BV421,BD 565411), CD11c (APC, BD 550261), and propidium iodide for 12 h at −4°C. Specimens were subsequently washed 3× with PBS-T and samples were mounted with pure VectorShield. Images were acquired using either the LSM780 or Airyscan (Zeiss) and analyzed using the ImageJ software. Images were taken using the ×40 magnification.
Oil red O and BODIPY staining in adipocytes. Oil red O staining was performed 5 days after the adipocytes induction. In which cells were washed twice with cold PBS and fixed with 1% methanol-free paraformaldehyde in PBS for 30 min at RT. Then the well was washed 2× with PBS, 3× with water, one time with 60% isopropyl alcohol, and stained with oil red O (six parts of 0.6% oil red O dye in isopropyl alcohol and four parts water) for 2 h. The excess stain was removed by washing with ice-cold 60% isopropanol and 2× with water. The plate was then observed under an inverted-brightfield microscope. BODIPY staining followed the same protocol as the one performed for WAT.
Image quantifications. For all morphometric analysis of adipocytes, islets, hepatocytes, and spleen follicles 3-5 animals of each genotype and food group were analyzed. Briefly, individual channels were converted to 8-bit grayscale and measurement scale was converted from pixels to mm. An identical threshold was applied to all images from the same channel to exclude background signals and watershed was applied as a mask. The neutral lipid storage was measured using the ImageJ macro "Lipid droplets Tool". Images were acquired using the 40× objectives and a threshold of 1 µM settled for droplets identification.
Sample collection. Mice were sacrificed by cervical dislocation and all tissues were harvested in the same chronological order for every animal. Excess blood was washed off by a quick rinse in freshly prepared ammonium carbonate wash solution (75 mM, pH 7.4). Each tissue was rapidly dissected into 2-4 slices, on ice. WAT samples were collected in 0.5 mL PBS on ice. Samples were flash frozen and stored in −80°C until metabolite extraction.
Polar metabolite extraction from brain, heart, liver, kidney, and spleen. Approximately 50-mg tissue slice was used per extraction. Tissue homogenization was performed in Milli-Q water pre-heated to 80°C (1 ml/ 50 mg tissue) with metal bead-based homogenizer-Retsch mixer mill MM 400. Lysates were incubated at 80°C for 3 min in a thermomixer, before being spun down at 7000 g for 3 min at 22°C. The pellets were used for protein estimation. Supernatant containing metabolites was transferred to a fresh tube and stored −80°C, till FIA-MS analysis (U.Sauer Lab, ETH Zurich).
Polar and apolar metabolite extraction from WAT. Thirty milligrams tissue was homogenized by mechanical disruption as mentioned above. Polar and apolar metabolites were extracted by adding 400 µl of cold methanol and 400 µl cold MiliQ. Samples were then vortex for 30 s. Next, 800 µl of ice-cold chloroform was added and vortex for 3 min, followed by centrifugation at 20,000 × g for 2 min at 4°C. The polar (top) and the organic phase were separated and transferred to fresh ice-cold 1.5 mL eppendorfs. Universal internal standard mix (ISTD) controls were added for absolute quantification before samples were vacuum dried and stored at −80°C, till LC-MS/MS analysis (Metabolomics Core, MPIIE Freiburg).
Serum metabolite extraction. Minimum 50 μl of blood was retrieved from the mouse heart and placed in a 1.5 ml tube. Then 4× volume of cold ultrapure methanol (MS-grade) was added to the whole-blood specimens and vortexed thrice 30 secs each before incubation at −20°C for 1 h. Afterward, the samples were centrifuged at 7000 × g for 3 min at RT. The supernatants were transferred to a new 1.5 ml tube and stored at −80°C till FIA-MS analysis.
Mouse metabolomics data analysis Spectral data processing and ion annotation. Brain, liver, spleen, kidney, skeleton muscle, and serum mass spectrometry data were processed and analyzed with Matlab (The Mathworks, Natick). Negatively charged ions were tentatively annotated to the Human Metabolome Database (HMDB) 94 based on accurate mass using 0.001 Da tolerance. Importantly, accurate mass does not allow distinguishing between compounds with identical molecular formulas and, hence, ions can match multiple compounds. The WAT dataset was analyzed using the TASQ software using the MPIIE automated pipeline followed by iPath3 of the polar Mof +/+ versus Mof +/− comparison.
Fold change and q value calculation. Mouse tissue metabolomics data were analyzed through two orthogonal pipelines, which were cross checked to verify robust identification of similar metabolite FC: (a) MetaboAnalyst software 4.0 95 based analysis was performed based on published protocol from Chong et al. 96  Metabolite class annotation and tissue enrichment. Metabolic class was generated by intersection of inchikey from uniquely identified metabolite names with the HMDB metabolite attributes list. The tissue enrichment was scored by constructing a contingency table querying the total number of metabolites, total number of deregulated metabolites and their related class. Metabolic class enrichment was calculated by Fisher.
Cross-tissue correlation. To identify significant cross-tissue correlations we subject the log 2 fold change of differential metabolites from each tissue set and construct a matrix containing their correlation coefficient and significance by Pearson correlation. Metabolites with a correlation p value < 0.05 in 95% were considered as significant. Calculations were done using R.
Other data analyses. Metabolite class annotation was performed using ClassyFire 100 followed by manual curation. Deregulated metabolites Z-Score were plotted as heatmap using the R pheatmap package.
Data analysis and visualization. Identified and quantified peak intensities were analyzed using MSstats 101 . Peptide intensities from evidence.txt were logtransformed and normalized by the "equalizeMedians" option. Statistical comparison was based on a built-in linear mixed-effect model 102  Bulk RNA-seq bioinformatic analysis. Reads for mouse bulk RNA-seq datasets were mapped following the default SnakePipes parameters of bulk RNA-seq pipeline 105 .
In brief, reads were mapped using STAR software alignment against mouse genome version GRCm38. The total number of sequenced reads averaged 15 million pairs of which mean alignment of 65% of the reads were uniquely mapped. Reads were counted with featureCounts (subread-1.5.0-p1). Differential expression analysis was performed with DESEQ2 (v1.26). Multi-FASTA QC statistics indicated data were of high quality and sequencing depth was sufficient to test for differential expression between conditions. Differential genes were called with an FDR threshold of 0.05. For the Trim28 +/D9 dataset the differentially expressed genes matrix was used as given 15 . Correlations between datasets and graphical output were conducted in an R environment.
Single-cell RNA-seq and gene regulatory network construction. Mouse single-cell count matrix from SKM, adipocyte, kidney, pancreas, liver, and heart single-cell analysis was obtained from the PanglaoDB database 41 . These data were then further processed using the Seurat v3 algorithm, in which standard preprocessing was followed. In brief, we filtered cells that have unique feature counts over 2500 or less than 200 and cells that showed >5% mitochondrial counts. After QC filtering, the data were normalized by employing a global-scaling normalization method that normalizes the feature expression measurements for each cell by the total expression, multiples the results by 10000 and log-transforms the product. Highly variable transcripts were identified using the Seurat3::FindVaribleFeatures 106 function. The data were scaled to provide an equal weight in downstream analysis and buffer the noise of highly-expressed genes. Then, linear dimensional reduction was applied on the scaled data. PCA and "Elbow plot" were used to define the dimensionality of each dataset. Finally, to explore feature expression similarities and defining cell populations we generated the UMAP for each tissue dataset and extracted their differentially expressed genes. After cluster identification and validation that most of the expected organ cell populations were present in the dataset, we used the tissue gene expression matrix to build a GRN using the GENIE3 algorithm. To predict the candidate genes we restrict the candidate regulators to Mof and use the "Random Forest" machine learning algorithm. The resulting weight matrix was used to explore gene set enrichment pathways related to Mof. Those analyses were performed in a R environment.
Kyoto Encyclopedia of Genes and Genomes enrichment analysis. For KEGG enrichment analysis, the list of candidate genes was subjected to the clusterProfiler() software. Wherein the function clusterProfiler::enrichKEGG 107 with a 0.01 p value cutoff was used. Enrichment analyses were conducted in the R environment.
Other data analyses. Heatmaps and scatter plots were generated in R with pheatmap software and qplots default settings. For heatmap representation the data were transformed in Z-score prior representation.

Chromatin immunoprecipitation (ChIP)
Chromatin preparation and immunoprecipitation. ChIP was performed according to published protocol 108 with the following modifications: WAT of 8-week-old SDfed animals was minced and transferred into a dounce homogenizer (loose pestle), covered with 1% formaldehyde in PBS, and dissociated, followed by 15 min incubation at room temperature (RT). During incubation, the tissue suspension was filtered using a nylon 70 µM cell strainer (Falcon, 352350). Excess formaldehyde was quenched by incubation with 125 mM glycine at RT for 5 min and the suspension was pelleted at 500 g for 5 min. Cell pellets were washed twice in ice-cold PBS and pelleted for 5 min at 500 g. For NEXSON-based nuclei isolation cell pellets were resuspended in 1 ml of lysis buffer (10 mM Tris-HCl pH 8, 10 mM NaCl, 0.2% Igepal, protease inhibitor cocktail). Cell suspension was transferred into Covaris MilliTube (cat. No. 520130) and sonicated in Covaris instrument (E220) for 30 s at peak power 75 W, duty factor 10% and 200 cycles/burst. Nuclei were pelleted at 1000 g at 20°C for 5 min. Pellets were resuspended in 0.5% SDS and incubated at RT for 10 min. Sample was diluted 1:5 in the digestion buffer (1.1% Triton, 1x CutSmart buffer, H2Odest). Up to 1 × 10 6 nuclei were digested with Cvik1 (5U/100,000 cells) at 25°C for 17 h, shaking at 800 g. Nuclei were pelleted for 5 min at 1000 g and washed in a nuclei wash solution (10 mM Tris-HCl pH 8, 0.25% Triton, 0.2 mg/ml BSA) until residual fat was removed. Nuclei were pelleted at 5000 g for 10 min, resuspended in the shearing buffer (10 mM Tris-HCl pH 8, 0.1% SDS, 1 mM EDTA), transferred into Covaris MicroTube (cat. No. 520052) and sonicated for 6 min at peak power 105 W, duty factor 2% and 200 cycles/burst. The sample was diluted 1:6 in 1× iC1b buffer (iDeal ChIP-seq, Diagenode) and incubated with 2 µg of antibody o/n at 4°C. The next day protein G Dynabeads were added and incubated for 3 h at 4°C. Beads were washed and DNA was eluted according to manufacturer's instructions (iDeal ChIP-seq, Diagenode). DNA was purified using MinElute PCR purification kit (Qiagen) for further analysis.
Library preparation. Libraries were prepared using the NEBNext Ultra II DNA Library Prep Kit for Illumina according to the manufacturer's instructions (New England Biolabs, E7645). ChIP libraries were sequenced with 2 × 50 bp paired-end reads on an Illumina NovaSeq6000 sequencer.
Processing of ChIP-seq datasets. ChIP-seq datasets were mapped to GRCm38/Mm10 with default paired-end parameters from snakePipes version 2.1.2 105 . Peaks were called with MACS2 (Galaxy version 2.1.1.20160309.3), and bandwidth was set to 200, lower mfold bound to 5, upper mfold bound to 500, and the q-value to 0.1. Input was used as control to call peaks. Peaks from each cell type were merged using cat, BEDtools sort (Galaxy version 2.27.0.0) and piped to BEDtools merge with a distance of 1000 bp. BamCompare (Galaxy version 2.5.1.0.0) with a bin size of 50 bp. The data were normalized as Log 2 fold change over H3 immunoprecipitation. Heatmaps were generated using deepTools2 compute matrix and plotHeatmap function from Deep-Tools2 version 3.5.0 109 . Boxplots were generated using DeepTools2 multi-BigwigSummary scores. The Bioconductor package ChIP-Seeker 110 was used to retrieve the nearest genes around the peak, annotate the genomic region of the peak, and retrieve peak features annotation. GO term analysis was performed using the Metascape 111 platform and the ShinyGo v0.61 database 112 .
Differentiated adipocyte (iAdipo) glycolytic capacity analysis. After Mof deletion iAdipo cells were plated in 96-well XF-plates (Agillent) and resuspended in glucosephenol free media with 2 mM glutamine and incubated for 40 min at 37°C under deprived CO 2 conditions. The plates were assayed on the XFe 96 analyzer, to calculate glycolytic rate by measuring the ECAR over time, after sequential addition of respiratory and glycolytic inhibitors. First, basal ECAR was measured, in the absence of glucose or pyruvate. Then, glucose (10 mM) was added to enhance glycolysis, before inhibiting mitochondrial ATP synthesis (with 1 μM Oligomycin) to estimate maximum glycolytic capacity. Finally, the glucose analog 2-deoxy-glucose (2-DG, 50 mM) was added to inhibit glycolysis. Assay data were exported as a graph pad file and plotting/statistical analysis was performed using the Prism software.
Knockdown, ectopic expression and chemical treatments Mof knockdown in in vitro differentiated pancreatic islets and insulin response. For siRNA knockdowns, 5 nM of Silencer Select (Ambion) control siRNA 4390846 or siRNA s38569 for MOF were performed on in vitro differentiated pancreatic islets for 3 days in triplicates using RNAiMAX (Thermo Fisher Scientific). Before knockdown, cells were seeded on 12-well plates in which in the bottom of the well we insert a coverslip. After, silencing cells were then crosslinked with 4% methanolfree formaldehyde in PBS at RT for 10 min and permeabilized with 0.1% Triton-X and 1% BSA in PBS for 30 min at RT. Primary antibody against insulin were diluted (1:100) in FACS buffer and incubated at 4°C for~16 h. Wells were then washed 2× with PBS-T and secondary fluorescently labeled antibodies were used to reveal target proteins. DAPI was used to stain DNA. Imaging was performed with the LSM780 confocal microscope.
Statistical analysis. All bar plots depicted in this manuscript show the mean value ± standard error of the mean (±SEM) and the solid line depicts the median. All violin plots use dotted lines to show the quartiles and dashed or solid lines to indicate the medians. Boxplots were generated using the boxplot() function in R with the upper whisker = min(max(x)), Q_3 + 1.5 * interquartile range (IQR) and the lower whisker = max(min(x), Q_1 -1.5 * IQR) where IQR = Q_3 -Q_1, the box length. Notches in boxplots display a confidence interval around the median which is normally based on the median ± 1.58*IQR/sqrt(n). Data were tested statistically using two-tailed or one-tailed tests as indicated in the figure. Normality was scored by the Shapiro-Wilk and D'Agostino Pearson omnibus test. P values above 0.05 weeres considered to be statistically significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All sequencing data from this study have been uploaded to the NCBI GEO database. Raw data pertaining to RNA-seq and ChIP-seq experiments are deposited under the accession number GSE156463. Source data are provided with this paper.