Genetic Nicotinamide N-Methyltransferase (Nnmt) Deficiency in Male Mice Improves Insulin Sensitivity in Diet-Induced Obesity but Does Not Affect Glucose Tolerance

Nicotinamide N-methyltransferase (NNMT) is expressed in most tissues including skeletal muscle (SKM), fat, and liver (1–3). In humans, NNMT expression is highest in the liver, followed by fat, whereas in mice, adipose expression appears to exceed that of the liver by at least 10-fold (3,4). NNMT is highly expressed in several human cancers (5). It catalyzes methylation of nicotinamide (NAM) and other pyridine derivatives using S-adenosylmethionine (SAM) to yield the products 1-methylnicotinamide (MNAM) and S-adenosylhomocysteine (SAH). NNMT may play a role in the biotransformation of xenobiotics (6,7). NAM is the amide of nicotinic acid, and together these are known as niacin/vitamin B3. After the methylation of NAM to MNAM, it no longer functions as an NAD⁺ precursor and can be excreted via urine (8,9). Increased urinary concentration of MNAM has been shown in humans with type 2 diabetes (T2D) and obesity-prone mouse strains (10). SAM is a methyl donor in numerous reactions catalyzed by methyltransferases including histone methylation, which can affect gene regulation and metabolism (11).

In high-fat diet (HFD)–fed mice, adipose NNMT protein (12), mRNA (13), and enzyme activity (4) were increased. Human adipose NNMT (aNNMT) expression appears to be increased in T2D and to correlate negatively with insulin sensitivity (IS) among individuals with diabetes (2); however, an association with obesity remains unclear (2,14). Investigations of aNnmt and liver (l)Nnmt expression reveal a high interstrain variability in mice (∼20-fold in white adipose tissue [WAT], ∼150-fold in liver) and a similar intersubject variability in humans (2). Suggestive of a causative role of NNMT in the development of obesity and associated metabolic dysregulation, antisense oligonucleotide knockdown (ASO-KD) protected against these phenotypes by increasing cellular energy expenditure (EE) in HFD-fed mice (12). ASO-KD reduced Nnmt in liver and WAT and led to reduced weight gain, relative fat mass, plasma insulin levels, and plasma triglyceride levels and improved glucose tolerance (12).

Here, we aimed to broaden the body of knowledge regarding the role of NNMT in metabolic disease. With a constitutive knockout, we investigate the effects of Nnmt deletion on energy metabolism, glucose homeostasis, and the development of obesity. Furthermore, we examine a human weight reduction (WR) study for NNMT expression, MNAM levels, and changes of metabolic parameters.

NNMT knockout (NNMT^−/−) experiments were conducted according to institutional ethical guidelines and approved by LAGeSo (Landesamt für Gesundheit und Soziales) Berlin (G0276/15) and by personal licenses for Jacob Jelsing (2013-15-2934-00784) issued by the Danish Committee for Animal Research.

VelociGene (Regeneron Pharmaceuticals, Inc.) technology (15) was used to generate conditional NNMT^fl/fl mice on a C57BL/6 background at Regeneron Pharmaceuticals Inc. Details of the targeting strategy are depicted in Fig. 3A. For whole-body Nnmt deletion, mice were crossbred with a ZP3-Cre deleter (C57BL/6-Tg(Zp3-cre)93Knw/J; The Jackson Laboratory).

NNMT^−/− and wild-type littermate controls (WT) were shipped to Sanofi (Frankfurt am Main, Germany) or Charité–Universitätsmedizin Berlin (Berlin, Germany) for experiments.

The 6- to 10-week-old males or females were fed an HFD (fat 60% kcal; carbohydrate 20% kcal) (D12492; Research Diets, Inc.), a Western diet (WD) (fat 47% kcal; carbohydrate 34% kcal) (TD.97366; ssniff special diets GmbH), or a standard diet (STD) (Rat/Mouse-Maintenance; ssniff special diets GmbH) ad libitum (Supplementary Tables 5–7) with free access to water. Mice were maintained in individually ventilated cages (2–4/cage) environmentally controlled with a 12-h light/dark cycle and monitored weekly. Body composition was assessed by ¹H-MRS using a Minispec LF50 Body Composition Analyzer (Bruker BioSpin).

NNMT^−/− and WT females were fed a WD for 18 weeks. From week 13 to 18, animals were treated daily with a vehicle (30% PEG200) by oral gavage to simulate pharmacological intervention with a test compound, analogous to another set of experiments (16), but not in weeks 1–12 of the WD, in which we assessed body composition, body weight (BW) gain, and food intake. In week 18, an oral glucose tolerance test (oGTT) was performed in PEG200-treated mice. In our article, we mainly used data from weeks 1–12 before PEG200 treatment.

C57BL/6N females (Taconic) were randomized into three groups receiving PBS (vehicle), control-ASO, or NNMT-ASO (37.5 mg/kg) (Supplementary Table 1) (taken from Kraus et al. [12]) subcutaneously every other day for 10 weeks on a WD. ASOs were partly 2′-O-(methoxy)-ethyl modified to improve resistance against nucleases, synthetized on a 250-μmol scale by phosphoramidite chemistry, purified using ion-exchange chromatography (Q Sepharose FF), desalted by ultrafiltration, and freeze dried, and purity was determined by reverse-phase high-performance liquid chromatography (>90%). Control- and NNMT-ASO were detected as free acids by their masses of 7215.6 and 7251.9 Da.

GTTs were performed in overnight-fasted (16 h) mice orally applying 2 g (WD) or intraperitoneally applying 1 g/2 g (HFD/SD) glucose/kg BW. Glucose and insulin (oGTT only 15 min value) were measured at 0, 15, 30, 60, and 120 min via the tail.

Hyperinsulinemic-euglycemic (HE) clamps in conscious, restrained mice and activity and indirect calorimetry (TSE LabMaster System) were performed as described previously (17).

Terminally, mice were sacrificed, and samples were analyzed, collected, immediately processed, and/or appropriately stored. Detailed information for mouse phenotyping is summarized in Table 1.

Western blot of epididymal WAT (eWAT) using Total OXPHOS WB-Antibody-Cocktail (ab110413; Abcam), rabbit-anti-mouse NNMT (ARP42281_T100; Aviva Systems Biology), rabbit-anti-mouse α-tubulin (2144; NEB), and horseradish peroxidase–anti-rabbit IgG (7074; Cell Signaling Technology) for detection was conducted as described previously (18,19). Optical density (OD) was quantified by ImageJ (20).

Plasma glucose was measured in duplicate (≤10%) with Contour XT glucometer (Ascensia), nonesterified fatty acids (NEFAs) by NEFA HR (2) kit (FUJIFILM Wako Diagnostics U.S.A.), insulin via Mouse Insulin ELISA Kit (Mercodia), and other metabolic parameters on a Cobas 6000c (501-module; Roche Diagnostics) according to manufacturer instructions using the following: Autokit Total Ketone Bodies (FUJIFILM Wako Diagnostics U.S.A.), Gluco-quant Glucose/HK Kit, cholesterol with CHOD-PAP kit, and Crep2 Creatinine plus V2 (all Roche Diagnostics).

Plasma/tissue MNAM levels were measured by a liquid chromatography-tandem mass spectroscopy method with chromatographic separation and quantification as described previously (2).

Mouse/human RNA was extracted using TRIzol, DNase digested, and transcribed into cDNA via RevertAid Reverse Transcriptase as described previously (17). Quantitative real-time PCR was performed on a Roche LightCycler 96 (Roche Diagnostics). Human/mouse NNMT/Nnmt expression was normalized to cyclophilin (PPIA/Ppia), GAPDH/Gapdh, and/or β-actin (Actb). For the primers used, see Supplementary Table 8.

RNA of five liver samples of NNMT^−/− and WT mice (fed for 12 weeks with an HFD) was isolated using RNeasy Plus Universal kit (Qiagen) according to the manufacturer protocol. The 100 ng/μL RNA samples (RNA integrity number >7, OD_260/280 ≥2.0, OD_260/230 ≥1.6) were prepared for next-generation RNA sequencing. RNA library preparation using the Illumina TruSeq RNA Sample Prep Kit V2 including polyA selection and single-end sequencing (50 base pairs, adaptor trimming, demultiplexing) were performed on an Illumina HiSeq 2500 System on High Output Run Mode at ATLAS Biolabs. Summarized results are in Supplementary Table 9.

Study protocols were approved by the Institutional Review Board of Charité–Universitätsmedizin Berlin. It is registered under NCT00850629 (clinicaltrials.gov). All subjects gave written informed consent (21,22). Briefly, obese subjects were anthropometrically and metabolically phenotyped (oGTT, calorimetry, bioimpedance, HE clamps) and subcutaneous WAT (scWAT)/SKM biopsy samples were taken before (T-3) and after (T0) 3 months of WR through lifestyle intervention and analyzed by RNA sequencing to discover differentially expressed genes (23).

All data are presented as the mean ± SD or Tukey box-and-whisker plots. Whiskers represent the range except where data lie >1.5 interquartile ranges above/below the 75th/25th percentile; outliers are depicted as black dots and are included in each analysis.

GraphPad Prism 6 (GraphPad Software, La Jolla, CA) and R version 3.1.1 (24) were used for statistics and graphs.

Statistical significance was considered as *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001. Differences were calculated using a two-tailed unpaired Student t test with Welch correction or Mann-Whitney test, one-way or two-way (repeated-measures) ANOVA adjusted with a Tukey or Bonferroni multiple-comparisons test (mct) for post hoc analysis, or Kruskal-Wallis test by ranks with Dunn mct. Correlations were calculated using Pearson or Spearman correlation coefficient, all depending on whether data passed the normal distribution tests.

Mouse RNA sequencing raw data were analyzed using Array Studio Software (OmicSoft Corporation, Cary, NC). RNA-sequencing reads (fastq files) were mapped to the mouse genome with OmicSoft gene model B38 (20141031) and OSA4 as aligner. The number of fragments per kilobase of transcript per million mapped reads (FPKM) and counts were calculated. Genes with FPKM <1 were filtered out according to the 50% fraction rule to exclude minimally expressed genes. As quality control assessment, principal component analysis was applied. Between WT and NNMT^−/− mice, differentially expressed genes were detected using the DESeq2 package in SPSS with a cutoff of a >1.5 fold change in expression and P < 0.05 adjusted with Benjamini-Hochberg procedure to control false discovery rates. All differentially expressed genes are listed in Supplementary Table 4.

Human data were analyzed using R by two-tailed Wilcoxon matched-pairs signed rank test, paired two-way ANOVA adjusted with Bonferroni mct. Correlations were calculated by Spearman correlation with Bonferroni mct. A linear regression model was used and adjusted for confounders (age, sex, and BMI at T-3). Statistical tests were used appropriately; details are in figure legends.

We investigated the effects of NNMT-ASO-KD (Supplementary Table 1) on BW, body composition, and metabolic parameters in female mice fed a WD for 12 weeks. NNMT-ASO-KD reduced Nnmt expression in 3T3-L1 adipocytes (Fig. 1A) and in vivo in liver (Fig. 1B) and scWAT (Fig. 1C) compared with vehicle and control-ASO and in visceral WAT (vWAT) (Fig. 1D) compared with vehicle. MNAM concentration was reduced in scWAT (Fig. 1E) and vWAT (Fig. 1F) but not in plasma (Fig. 1G). During 10 weeks of WD, NNMT-ASO-KD mice showed no significantly reduced BW at any time point compared with vehicle or control-ASO mice (Fig. 2A); however, their cumulative weight gain was decreased (Fig. 2B). Although lean mass remained similar (Fig. 2C), NNMT-ASO mice gained less fat than vehicle and by trend compared with control-ASO mice (Fig. 2D). In mice fed ad libitum, plasma glucose was decreased throughout (Fig. 2E) and insulin was decreased at the end of WD in NNMT-ASO-KD mice (Fig. 2F). After 10 weeks, the NNMT-ASO–treated mice showed diminished glucose excursions in an oGTT (Fig. 2G) with a significantly lower area under the curve for glucose (Fig. 2H), but no difference in plasma insulin levels at 15 min (Fig. 2I). Having observed the beneficial effects of NNMT-ASO-KD on BW, fat, and glucose tolerance also in WD-fed mice, we next generated an Nnmt knockout.

After the generation of conditional NNMT^fl/fl mice (Fig. 3A), we crossbred them with a global ZP3-Cre deleter to generate whole-body Nnmt knockout mice (NNMT^−/−) and compared them to NNMT^fl/flxCre⁻ control littermates (i.e., WT). We validated the Nnmt deletion on DNA, RNA, and protein level (Fig. 3B–D and Supplementary Fig. 1A and B). Consistent with Nnmt deletion, MNAM concentration in NNMT^−/− mice was reduced by >99% in plasma, liver, and vWAT (Fig. 3E–H), whereas the NAM concentration was increased in plasma but not in liver or vWAT (Fig. 3I–K), and plasma SAM and SAH remained unchanged (Fig. 3L).

Kraus et al. (12) aligned Nnmt expression data and mouse strains by their susceptibility for obesity and showed that increasing Nnmt in WAT but not in liver matched increasing susceptibility to develop obesity. Using a second data set from www.BioGPS.org containing additional strains and their aNnmt expression, no such correlation (between expression and propensity for obesity) was observed (Supplementary Fig. 2). We first examined the potential effects of Nnmt deletion in the absence of any dietary challenge. We analyzed male NNMT^−/− mice on STD for BW, body composition, and basal metabolic parameters including glucose and insulin. Furthermore, we investigated them by intraperitoneal glucose tolerance test (ipGTT); characterized their food and water intake, respiratory exchange ratio (RER), EE, locomotor activity, O₂ consumption, and CO₂ production in metabolic cages; and performed HE clamps to estimate IS by glucose infusion rate (GIR), endogenous glucose production (EGP), and rate of glucose disappearance (R_d). However, these investigations did not reveal differences between NNMT^−/− and WT mice (Supplementary Table 2).

We next challenged NNMT^−/− males that were fed an HFD, not a WD, for 10 weeks. No impact of Nnmt deletion on BW or composition was detected (Fig. 4A–C). Challenging the HFD-fed mice with glucose showed nearly identical glucose tolerance and similar insulin concentrations (Fig. 4D and E). In week 11, NNMT^−/− mice were analyzed in metabolic cages. Both groups showed similar food and water intake (Fig. 4F and G). We did not reveal differences in O₂ consumption (data not shown), RER (Fig. 4H), or EE (Fig. 4I). Locomotor activity was low during the daytime and increased rapidly with the onset of nighttime, yet without significant differences (Fig. 4J).

Finally, we investigated the IS of HFD-fed mice by performing HE clamps. Conscious, restrained WT and NNMT^−/− mice were clamped with similar mean steady-state plasma glucose concentrations (130 vs. 127 mg/dL) (Fig. 5A, top panel), whereas mean GIR was significantly elevated in NNMT^−/− mice compared with WT mice (24 vs. 11 mg/kg/min), indicating improved IS in HFD-fed Nnmt-deficient mice (Fig. 5A, bottom panel). Plasma insulin level was comparable in either phase (Fig. 5B). Plasma NEFAs were similar in the basal phase, but insulin-mediated NEFA reduction was elevated in NNMT^−/− mice (56%) compared with WT mice (33%) (Fig. 5C). Although basal EGP was similar (9 mg/kg/min), EGP under hyperinsulinemic conditions was almost completely suppressed in NNMT^−/− mice, and WT mice showed only partial EGP reduction (23%) (Fig. 5D). Accordingly, R_d and glucose clearance were increased in NNMT^−/− mice in the hyperinsulinemic phase, but not in the basal phase (Fig. 5E and F). Calculating the IS index (ISI) of the steady state, NNMT^−/− mice showed a significant increase (0.28 mg/kg/min/[mU/L]) compared with WT (0.10 mg/kg/min/[mU/L], P = 0.017). The organ-specific glucose uptake (GU) in SKM and WAT revealed no significant difference; however, it did reveal a trend in SKM for NNMT^−/− mice (Fig. 5G).

Despite similar body composition and glucose tolerance, HE clamp results strongly suggest substantially improved IS in HFD-fed NNMT^−/− males with regard not only to insulin-stimulated GU but also to the inhibition of lipolysis and EGP.

Given reports that adipose SIRT1 (silent mating type information regulation 2 homolog 1) depletion may reduce IS (25–27), and that NNMT reduction may increase SIRT1 activity (12), we wondered whether increased SIRT1 may mediate the improved IS in NNMT^−/− mice. However, adipose Sirt1 expression, SIRT1 protein expression, and expression of target genes did not differ from those of controls (Fig. 5H–J). Hepatic SIRT1, Sirt1, and G6pc expression were not different from those of controls, whereas Pck1 expression was increased in NNMT^−/− mice (Fig. 5H–J). Analyzing representative respiratory chain complex genes/proteins as a proxy for cellular oxygen consumption in adipose tissue of NNMT^−/− mice, we observed no significant differences (Fig. 5K and L and Supplementary Fig. 3).

To elucidate whether this enhanced IS is reflected in gene regulation, we performed RNA sequencing with liver samples of NNMT^−/− and WT mice. However, between genotypes no obvious separation of both groups in the principal component analysis was observed, suggesting no fundamental regulatory difference (Supplementary Fig. 4). In detailed analysis, we found no highly differentially regulated genes aside from Nnmt, which was virtually absent in NNMT^−/− (Supplementary Fig. 1B). All of the other 23 differentially expressed genes that exceeded cutoff were deemed unlikely to explain the observed HE clamp results (Supplementary Table 3).

Finally, to corroborate our ASO-KD results, we analyzed NNMT^−/− females during a 12-week WD, not an HFD, and observed decreased BW, with NNMT^−/− mice finally weighing 6 g less than WT mice, whereas the initial BW and food intake did not differ (Fig. 6A–C). This was accounted for mainly by a difference in fat rather than lean mass (Fig. 6D and E). During a WD, fed glucose levels showed only a minor increase and remained comparable while insulin rose significantly less (37%) in NNMT^−/− mice (Fig. 6F–H). However, Nnmt deletion did not reveal any benefit on glucose tolerance (Fig. 6I) and 15-min insulin measurement showed no significant difference investigated by oGTT (Fig. 6J). Likewise, other metabolic parameters such as NEFAs, triglycerides, cholesterol, HDL, and LDL analyzed in plasma did not differ between WD-fed WT and NNMT^−/− females (data not shown). Fed with an HFD, female NNMT^−/− mice did not show changes in BW and composition (data not shown).

To explore the role of NNMT and its metabolites in humans and to check which findings from the mouse are transferrable, we analyzed data from a human WR study (23). Study participants were extensively characterized before (T-3) and after (T0) a 3-month WR intervention study. SKM and scWAT biopsy samples were taken and RNA sequenced (21,22).

WR showed no differential regulation in SKM; however, the WR led to increased aNNMT (Fig. 7A and B), especially in 18- to 39- and 61- to 90-year-old subjects (Supplementary Fig. 5A and B). Before WR, aNNMT correlated negatively with IS, given by the ISI, or positively with HOMA-insulin resistance (IR), but these correlations disappeared after WR (Fig. 7C and D). Even after adjustment in a linear regression model with confounders (age, sex, BMI at T-3), the baseline correlation remained significant (ISI at T-3: R = −0.299, P = 0.001; HOMA-IR at T-3: R = 0.324, P = 0.011). No association between aNNMT and BMI, as a parameter for obesity, plasma NAM, or MNAM was present (Supplementary Fig. 5C–E). Both plasma NAM and MNAM were lower after WR (Fig. 7E and F). Neither NAM nor MNAM correlated with IS or BMI before WR. Surprisingly, despite the overall reduction, MNAM correlated positively with IS after WR (Fig. 7G and H), suggesting that those with the greatest reductions had lesser improvements in IS. In contrast, NAM correlated negatively with IS after WR (Supplementary Fig. 5F and G). A summary of findings from our WR study is found in Supplementary Table 4.

In female NNMT-ASO-KD mice fed a WD, we report reduced BW, fat, plasma glucose, and insulin and improved glucose tolerance. This is highly consistent with findings in male NNMT-ASO-KD mice reported by Kraus et al. (12). Building on this, we generated an Nnmt-deficient constitutive knockout mouse and demonstrated the deletion of Nnmt DNA, drastic reductions in RNA, and virtually no protein or MNAM; the faint protein band in vWAT may represent nonspecific antibody binding rather than residual amounts of NNMT protein. In WD-fed NNMT^−/− females, we observed reduced BW, fat, and plasma insulin but no changes in glucose or glucose tolerance, whereas in HFD-fed NNMT^−/− males we report increased IS under hyperinsulinemic conditions without changes in BW, fat, fasting glucose, fasting insulin, or glucose tolerance. This suggests a BW-independent modulation of IS in this group.

Our investigations add further evidence that a reduction of NNMT inhibits weight gain and protects against associated negative metabolic effects in models of diet-induced obesity. This may represent a viable pharmacological target for future treatment of such conditions (16,28,29). Consistent effects of NNMT-specific ASOs, small-molecule inhibition (12,16,28,29), and the genetic model described here provide strong evidence for the specificity of NNMT modulation.

An important, if somewhat surprising, finding is that complete genetic Nnmt deletion did not consistently meet all aspects of the obesity-resistant phenotype observed in NNMT-ASO-KD. NNMT^−/− mice showed different responses depending on sex and detailed diet composition.

We confirmed that NNMT is the primary enzyme catalyzing the methylation of NAM and that this is not directly compensated by any other gene. It is, however, possible that epigenetic changes, mediated through altered metabolite concentrations during early development, are able to compensate for the NNMT loss on a system level and thus lead to metabolic normalization. In studies with ASOs or inhibitor, adult mice are used and, as such, this developmental window of opportunity may be missed (30). Alternatively, residual NNMT activity (10–40%) (12,16) may impede compensatory mechanisms that mediate adaptation to a null situation. Furthermore, it is possible that NNMT concentration within a certain range (in adipose or other tissues) is required to mediate the observed positive effects, whereas complete deletion negates these. Investigations of tissue-specific NNMT^−/− mice could help to elucidate this and are currently underway.

An additional possibility is that some compensatory mechanisms are triggered upstream of the protein function (i.e., it may be the genetic mutation itself or the complete lack of transcript RNA rather than the loss of protein function that triggers these). In knockdown models, the DNA remains intact, and so such a compensatory mechanism would not be triggered (31). Similar discrepancies between ASO-KD and genetic knockout have previously been described (17,31–35).

We analyzed SIRT1 and several target genes in liver and eWAT of HFD-fed NNMT^−/− males and compared the results with those previously published (9,12). We observed increased hepatic expression of Pck1 in NNMT^−/− mice, but no change in Sirt1, SIRT1 (protein), or G6pc, whereas Hong et al. (9) reported reduced SIRT1 and G6pc in NNMT-KD mice but unchanged Sirt1 and Pck1, and Kraus et al. (12) reported reduced Sirt1 but unchanged Pck1 and G6pc. In WAT, Kraus et al. (12) reported increased expression of SIRT1 target genes Cd36, Cat, and Sdhb. We also measured these along with Sirt1 and SIRT1 but did not reveal any differences between NNMT^−/− and WT mice. Our data, therefore, do not directly support the hypothesis that NNMT, through its product MNAM, increases the stability of SIRT1 protein as in the study by Hong et al. (9). The discrepancies among these three works highlight the need for further investigations.

Compared with the study by Kraus et al. (12), we additionally included a vehicle-control to control-ASO and NNMT-ASO and found an effect of control-ASO compared with vehicle (Fig. 1). This could indicate a general (energy-consuming) RNA interference effect responding to the exogenous scrambled siRNA, but not to vehicle-only (36–39), alongside the NNMT metabolic effects that are missed without vehicle.

Sex and age differences in DNA methylation starting at 4 weeks and a gradual establishment of sex-dependent hypomethylation were reported for the lNnmt gene among others (40). This may also occur in other tissues and could explain the sex differences in response to diet and also differences between ASO-KD and knockout mice. lNnmt expression showed sex differences in mice with an altered methylation pattern in which lNnmt levels were lower in males (40). Because NNMT is also involved in histone methylation via SAM (30) and histone methylation can also be sex specific, the deletion of Nnmt could also lead to sex differences in the methylation pattern and, therefore, produce different responses between NNMT^−/− males and females per se, and in particular in response to different diets. Beyond that, a limitation of our investigation is the use of different diets for males and females. Male mice were fed an HFD, whereas females were fed a WD, resulting in more calories originating from carbohydrates (Supplementary Tables 6 and 7). Simply, this could underlie the differences observed between sexes (41). Further research on NNMT^−/− mice is warranted to fully elucidate whether sex and/or the dietary nutrient content underlie the observed differences.

Despite unchanged glucose tolerance, we observed markedly improved IS in male NNMT^−/− mice being fed an HFD investigated in an HE clamp. Both increased R_d and sustained insulin-mediated suppression of EGP contributed to the higher GIR in NNMT^−/− mice. Additionally, in the distinct SKM and adipose tissues tested, a trend for but no significant difference in GU was observed. It is likely that a summation effect and possibly increased hepatic GU may be responsible for the significantly elevated R_d (Fig. 5). Because improved IS could not be explained by differences in hepatic gene regulation (Supplementary Table 7), extrahepatic processes or post-translational and/or flux regulation mediated by MNAM may contribute to the improved IS without affecting steady-state gene expression.

Several plausible mechanisms have been proposed through which increased NNMT activity could mediate either the development of, or protection against, obesity and loss of IS. For instance, increased plasma MNAM may induce inflammation and insulin resistance (2,8). Otherwise, studies have shown MNAM to have antithrombotic and anti-inflammatory effects, to lower plasma triglycerides, and to protect against endothelial dysfunction (42), and it is implicated in mediating extended life span in Caenorhabditis elegans (43). Other mechanisms through which NNMT activity might regulate energy metabolism are the modulation of NAD⁺ synthesis affecting NAD(P)⁺ concentration and NAD(P)⁺/NAD(P)H ratio, changes in cellular methylation potential (SAM/SAH ratio), or reduction in polyamine flux.

A distinct β-cell phenotype of NNMT^−/− mice could explain the discrepancy of improved IS without enhanced glucose tolerance due to an altered modulation of insulin secretion. An important difference of the whole-body Nnmt deletion is that it also affects pancreatic islets, whereas the NNMT-ASO-KD does not. However, this was not observed during the glucose challenge of NNMT^−/− mice, and merely an impaired insulin secretion could be causative (Fig. 6G and H). Additionally, in PANC-1 cells, a pancreatic cancer cell line, NNMT-ASO-KD reduced cell proliferation and survival under metabolic stress (44).

Consistent with a study involving obese adults (2), we observed no correlations between BW/fat and MNAM in our human study. Our negative correlation between aNNMT and IS is in agreement with the study by Kannt et al. (2) and the idea that increased aNNMT expression/activity is associated with insulin resistance. The increased aNNMT as a result of WR is somewhat surprising but is not in direct contradiction with previously published data, which showed a decrease only in subjects with T2D or impaired glucose tolerance after a 12-week exercise program (2). Another study (45) found increased NNMT in SKM and elevated plasma MNAM after a short and intense WR intervention study. Indeed, after WR and contrary to expectations (2,46), MNAM correlated positively with ISI. The fact that a correlation between NNMT and IS disappeared after WR and that correlations between the metabolites NAM, MNAM, and IS appeared only after WR is surprising and points to a complex regulation of these metabolites. This remains to be elucidated.

Acknowledgments. The authors thank Diana Woellner (Charité–Universitätsmedizin Berlin, Berlin, Germany, and DZHK [German Centre for Cardiovascular Research], partner site Berlin) and Marie-Christin Gaerz (Charité–Universitätsmedizin Berlin) for assistance in (mouse) experiments; Nadine Huckauf (Charité–Universitätsmedizin Berlin) and Candy Kalischke (Charité–Universitätsmedizin Berlin) for assistance with biochemical experiments and quantitative PCR; and Hui Tang (Charité–Universitätsmedizin Berlin) for assistance during HE clamps. The authors also thank Beate Greiner (Sanofi Research and Development, Frankfurt am Main, Germany) for large-scale synthesis of the ASOs; Mostafa Kabiri (Sanofi Research and Development) for help with design, generation, and breeding of the conditional NNMT^−/− mouse strain; Gitte Hansen (Gubra ApS, Hørsholm, Denmark) for conducting the ASO study; and Claire Kammermeier (Sanofi Research and Development), Martin Stephan (Sanofi Research and Development), Pierre Wenski (Sanofi Research and Development), Jörn Wandschneider (Sanofi Research and Development), Uwe Butty (Sanofi Research and Development), Mandy Grohmann (Sanofi Research and Development), and Kerstin Lentz (Sanofi Research and Development) for technical assistance. The Genotype-Tissue Expression (GTEx) Project was supported by the Common Fund of the Office of the Director of the National Institutes of Health and by the National Cancer Institute; National Human Genome Research Institute; National Heart, Lung, and Blood Institute; National Institute on Drug Abuse; National Institute of Mental Health; and National Institute of Neurological Disorders and Stroke. The expression data for human NNMT described in the Introduction were obtained from the GTEx Portal on 15 January 2018.

Funding. This work was supported by a joint laboratory between Charité and Sanofi, Deutsches Zentrum für Herz-Kreislauf-Forschung (DZHK/BMBF), the ICEMED Consortium, and the Berlin Institute of Health.

Duality of Interest. S.B., K.M., and J.S. received research support in relation to the joint laboratory between Charité–Universitätsmedizin Berlin and Sanofi. M.B. is an employee of Treamid Therapeutics GmbH. K.J.-H., R.E., A.P., F.B., D.M., and A.K. are employees of Sanofi. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.B. designed the studies; performed experiments; contributed to discussion of the results; supervised the project; and wrote, critically reviewed, and edited the manuscript. J.P. performed experiements; contributed to discussion of the results; and wrote, critically reviewed, and edited the manuscript. M.B. performed experiments, analyzed the clinical study data, contributed to discussion of the results, and critically reviewed the manuscript. K.J.-H. designed and synthetized the ASOs. R.E., A.P., and F.B. performed experiments. D.M. analyzed mouse next-generation sequencing data. K.M. performed experiments, assessed and analyzed the clinical study data, contributed to discussion of the results, and critically reviewed the manuscript. J.S. designed the studies, supervised the project, assessed the clinical study, contributed to discussion of the results, and critically reviewed and edited the manuscript. A.K. designed the studies, supervised the project, performed experiments, contributed to discussion of the results, and critically reviewed and edited the manuscript. S.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018.