Modulation of glycosphingolipid metabolism significantly improves hepatic insulin sensitivity and reverses hepatic steatosis in mice


  • Potential conflict of interest: Nothing to report.


Nonalcoholic fatty liver disease (NAFLD) is associated with obesity, insulin resistance, and type 2 diabetes. The hyperinsulinemia that occurs as a consequence of insulin resistance is thought to be an important contributor to the development of fatty liver. We have shown that the iminosugar N-(5'-adamantane-1'-yl-methoxy)-pentyl-1-deoxynojirimycin (AMP-DNM), an inhibitor of the enzyme glucosylceramide synthase, is a potent enhancer of insulin signaling in rodent models for insulin resistance and type 2 diabetes. The present study was designed to assess the impact of AMP-DNM on insulin levels, liver triglyceride synthesis, and gene expression profile. Treatment of ob/ob mice with AMP-DNM restored insulin signaling in the liver, corrected blood glucose values to levels found in lean mice, and decreased insulin concentration. The expression of sterol regulatory element-binding protein 1c target genes involved in fatty acid synthesis normalized. AMP-DNM treatment significantly reduced liver to body weight ratio and reversed hepatic steatosis, comprising fat as well as inflammatory markers. In addition, AMP-DNM treatment corrected to a large extent the gene expression profile of ob/ob mice livers toward the profile of lean mice. Conclusion: Pharmacological lowering of glycosphingolipids with the iminosugar AMP-DNM is a promising approach to restore insulin signaling and improve glucose homeostasis as well as hepatic steatosis. (HEPATOLOGY 2009.)

The prevalence of the metabolic syndrome is rising dramatically in westernized societies.1 Fundamental features of the metabolic syndrome are obesity, hyperglycemia, high blood pressure, and elevated triglycerides (TGs). These abnormalities are considered to be the driving forces in the development of diabetes and heart disease. However, the precise underlying mechanism is still poorly understood and may involve other factors. Candidates in this respect are low-grade inflammation throughout the body and fatty liver, both of which occur in concurrent obesity.2, 3

The proinflammatory environment that exists in adipose tissue of the majority of obese individuals4 is thought to precede insulin resistance by enhancing deregulation of adipose tissue, leading to adipocyte cell death.5 Subsequently, increased amounts of free fatty acids coming from the inflamed adipose tissue reach the liver, where they decrease insulin sensitivity.6 In the liver, insulin resistance prevents suppression of glucose production, signaling the pancreas to produce more insulin. Despite hepatic insulin resistance for the glucose homeostasis pathway, insulin sensitivity remains in the lipogenic pathway.7 Hyperinsulinemia triggers the activation of sterol regulatory element-binding protein 1c (SREBP1c), the transcription factor that controls genes in the lipogenic pathway such as fatty acid synthase. The combination of increased flux of free fatty acids to the liver from dysfunctional adipose tissue, together with induced fatty acid synthesis in the liver, leads to lipid accumulation in the liver. Steatosis is the first relatively benign stage of nonalcoholic liver disease. In the next stage, nonalcoholic steatohepatitis (NASH) can develop, which may result in cirrhosis and organ failure.8

Improving sensitivity of the liver for insulin is of utmost importance to break the cycle of hyperinsulinemia to hepatic steatosis. Suitable insulin-sensitizing agents that enhance pathways that suppress gluconeogenesis and enhance peripheral glucose uptake are needed. In this connection, the recent realization that insulin sensitivity is modulated by the lipid composition of the membrane harboring the insulin receptor is of interest. Evidence is accumulating that especially glycosphingolipids play an important regulatory role in insulin sensitivity.9–12 Recent findings suggest that excessive levels of glycosphingolipids modify the direct interaction between the insulin receptor and caveolin-1, thereby influencing the membrane localization and downstream signaling capacity of the insulin receptor.13 In leptin-deficient ob/ob mice, insulin resistance could be a direct consequence of altered GM3 levels, because leptin administration was recently shown to affect genes involved in glycosphingolipid synthesis, including GM3 synthase.14 The importance of GM3 in insulin signaling has been confirmed in studies that use inhibitors of glycosphingolipid biosynthesis.15, 16 Two different classes of inhibitors of glucosylceramide synthase were used in these studies: the iminosugar N-(5'-adamantane-1'-yl-methoxy)-pentyl-1-deoxynojirimycin (AMP-DNM) and Genz-123346 ((1R,2R)-nonanoic acid[2-(2',3'-dihydro-benzo [1, 4] dioxin-6'-yl)-2-hydroxy-1-pyrrolidin-1-ylmethyl-ethyl]-amide-l-tartaric acid salt). These compounds exert their function by reducing the synthesis of glucosylceramide, the precursor for more complex glycosphingolipids such as gangliosides. Importantly, pharmacological glycosphingolipid lowering was found to increase hepatic insulin signaling and decrease insulin concentrations, in combination with reductions in blood glucose and improved oral glucose tolerance. An additional beneficial feature of AMP-DNM is its ability to improve adipocyte function and reduce inflammation in adipose tissue of obese mice.17

The present study was designed to assess the hypothesis that increased insulin sensitivity by way of glycosphingolipid lowering results in reduction of fatty liver and restoration of normal liver physiology. We therefore treated lean and ob/ob mice with AMP-DNM and determined the effect on glucose homeostasis, liver fat storage, and gene expression.


AMP-DNM, (N-(5'-adamantane-1'-yl-methoxy)-pentyl-1-deoxynojirimycin; bw, body weight; Ig, immunoglobulin; mRNA, messenger RNA; NASH, nonalcoholic steatohepatitis; NK, natural killer; RIPA lysis buffer; 25mM Tris, pH 7.4, 150mM NaCl, 1mM EDTA 1% Nonidet P-40 0.5% sodium deoxycholate, and 0.01% SDS; SREBP1c, sterol regulatory element-binding protein 1c; TG, triglyceride.

Materials and Methods

AMP-DNM was synthesized as described.18 All solvents and reagents used were of analytical grade.


Male wild-type (≈23 g) and ob/ob (≈31 g) C57BL/6J mice (7 weeks old; Harlan Laboratories, Horst, The Netherlands) were maintained in controlled conditions (20–22°C, 50%–60% humidity, 12-hour light/dark cycle) with food and water ad libitum. Mice received the rodent AM-II diet (Hope Farms, Woerden, The Netherlands) with 0 or 100 mg AMP-DNM/kg body weight (bw)/day mixed in the food. Each treatment group consisted of five animals. The study followed the Dutch guidelines for the use of experimental animals and was approved by the Academic Medical Center Animal Experiments Committee.

Plasma and Tissue Sampling.

Blood samples were collected from the tail vein or through retro-orbital plexus puncture. A large blood sample was collected by way of cardiac puncture, and plasma was stored at −20°C. The liver was quickly excised and weighed, and parts were either used directly, snap-frozen in liquid nitrogen and stored at −80°C, or fixed in 10% buffered formalin and embedded in paraffin for further analysis.

Analytical Procedures.

Blood glucose levels were determined in plasma of nonfasted animals using a hand-held Glucometer (Ascensia Elite; Bayer A.G., Leverkusen, Germany). Cholesterol and TGs in liver samples were determined using colorimetric enzymatic kit from Biolabo (Maizy, France). To correct the obtained lipid values for the amount of tissue, the protein content of the liver was measured using the bicinchoninic acid method (Pierce; Perbio Science Nederland BV, Etten-Leur, The Netherlands). Hemoglobin A1c levels were measured using a single measurement A1C now device (Metrika, Sunnyvale, CA). Insulin levels were determined by enzyme-linked immunosorbent assay (Crystal Chem Inc., Downers Grove, IL). Oral glucose tolerance tests were performed as described.15 For the analysis of ceramide, glucosylceramide, and gangliosides, the lipids from 50 μL plasma were extracted with 2 mL chloroform/methanol (1:1, vol/vol) followed by deacylation in 500 μL 0.1 mol/L NaOH in methanol using a microwave oven (CEM microwave Solids/Moisture System SAM-155). The deacylated lipids were derivatized for 30 minutes with the addition of 25 μL O-phtaldehyde reagent to 50 μL lipid mixture and separated uisng high-performance liquid chromatography. Gangliosides were detected as recently described19 by analysis of the acidic glycolipid fraction obtained by the Folch extraction. In short, the upper-phase was desalted on a C18 Sep-Pak (Bakerbond) column and the eluted gangliosides were digested with ceramide glycanase. The released oligosaccharides were labeled at their reducing end with the fluorescent compound anthranilic acid (2-aminobenzoic acid), prior to analysis using normal-phase high-performance liquid chromatography.


Paraffin sections were dewaxed and stained with hematoxylin-eosin for general histology, with periodic acid-Schiff's reagent to visualize glycogen, or with 0.2% picro-sirius red to detect fibrillar collagen deposits. To detect neutral lipids, cryostat sections were stained with 0.3% Oil Red O. For immunohistology, dewaxed paraffin sections were incubated with 0.3% H2O2 in methanol to inactivate endogenous peroxidase. Upon antigen retrieval using either sodium citrate (0.01 M [pH 6.0] for 10 minutes at 100°C) or pepsin (0.25% in 0.01 M HCl for 10 minutes at 37°C), and blocking with 5% (vol/vol) normal goat serum, sections were incubated with rabbit polyclonal immunoglobulin (Ig) G anti–collagen I (GeneTex, Irvine, CA), anti–collagen III (BioLogo, Kronshagen, Germany), anti–collagen IV (Euro-Diagnostica, Malmö, Sweden), anti–cleaved caspase 3 (Cell Signaling Technology, Beverly, MA), mouse monoclonal IgG1 anti–Ki-67 (MIB-1; DAKO, Glostrup, Denmark), or mouse monoclonal IgG2a anti–smooth muscle actin (1A4; DAKO). Cryostat sections were fixed in cold acetone, and after inactivation of endogenous peroxidase using 0.1% (vol/vol) NaN3 and 0.3% H2O2 in phosphate-buffered saline and blocking in 5% (vol/vol) normal goat serum, sections were incubated with rat monoclonal IgG2a anti-mouse LY-49G2 (4D11; BD Pharmingen, Breda, The Netherlands). LY-49G2 is expressed by approximately 50% of splenic natural killer (NK) cells.20 After washing, bound primary antibodies were detected using appropriate secondary horseradish peroxidase–conjugated polyclonal goat IgG antibodies, directed against rabbit IgG (PowerVision; Immunologic, Duiven, The Netherlands), mouse IgG1, mouse IgG2a, or rat IgG2a (all from Southern Biotech, Birmingham, AL). Bound peroxidase activity was visualized using H2O2 and 3,3'-diaminobenzidine as chromogen. Sections were counterstained with either hematoxylin or methylgreen and mounted using Pertex.

Total RNA Isolation, Complementary RNA and Complementary DNA Synthesis, and Messenger RNA Expression.

Total RNA was extracted from approximately 30 mg frozen tissues with TRIzol reagent (Invitrogen, Breda, The Netherlands) followed by LiCl precipitation, and further purified using the RNeasy Mini Kit following the manufacturer's RNA Cleanup protocol (Qiagen, Valencia, CA). Purified RNA concentration was measured on a NanoDrop ND-1000 Spectrophotometer (Thermo Fisher Scientific, Wilmington, DE) and the quality was assessed using the RNA 6000 Nano LabChip Kit in an Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, TX).

For complementary DNA synthesis, equal amounts of RNA were treated with RQ1 RNase-free DNase (1 unit/2 μg total RNA; Promega, Leiden, The Netherlands) and reverse-transcribed with SuperScript II Reverse Transcriptase and random hexamers (Invitrogen, Breda, The Netherlands) according to the manufacturers' protocols. Gene expression analysis was performed on a Bio-Rad MyiQ Single-Color Real-Time PCR Detection System using the Bio-Rad iQ SYBR Green Supermix (Bio-Rad Laboratories Inc., Hercules, CA). Polymerase chain reaction primers were designed on the basis of Primer Express 1.7 software using the manufacturer's default settings (Applied Biosystems, Nieuwerkerk aan den IJssel, The Netherlands) and validated for identical efficiencies using hypoxanthine-guanine phosphoribosyl transferase, cyclophilin, and acidic ribosomal phosphoprotein P0 (36B4) as standard housekeeping genes.

For microarray analysis, the starting amount of total RNA used for complementary RNA synthesis was 500 ng. The first and second strand synthesis, as well as in vitro transcription and complementary RNA purification, were performed using Illumina TotalPrep RNA Amplification Kit (Ambion Inc., Austin, TX) according to the manufacturer's protocol. The concentration and quality of complementary RNA were assessed using the same method as for total RNA.

Gene Expression Profiling and Data Analysis.

Liver genome-wide gene expression profile was obtained using the Sentrix Mouse-6 Expression Beadchip (Illumina Inc., San Diego, CA), containing >47,000 50-mer oligonucleotide probes.21 Five microarrays were used for each of the strains (lean and obese) and treatments (AMP-DNM and control diet) studied, summing up to 20. BeadStudio version 2.2.22 (Illumina Inc., San Diego, CA) was used for the scanned image processing, outlier removal, local background correction, and summarization of multiple bead intensities per probe. The data were further background corrected, log transformed, and normalized using quantile normalization. To detect the differential gene expression, a moderated F test was performed (to detect changes in more than one comparison), followed by a moderated t test applied to each comparison separately. Genes were considered significant if the P values, adjusted for multiple testing by using Benjamini and Hochberg's method, were less than 0.05. The false discovery rate was thereby controlled to be less than 5%. Pathway, network, and gene-set enrichment analyses were applied, using the MetaCore suit (version 4.7, build 12996; GeneGo, Inc., St. Joseph, MI).22

Western Blot Analysis.

For western blot analysis, liver sections (10-15 mg) were pulverized under liquid nitrogen and the frozen powder was added to RIPA lysis buffer with protease inhibitors (Roche Molecular Biochemicals, Almere, Netherlands) and left to stand on ice for 30 minutes. Lysates were clarified by way of centrifugation (16,000g for 10 minutes), and supernatants were collected and stored at −80°C. Equal amounts of protein (40 μg) were subjected to electrophoresis on 10% sodium dodecyl sulfate–polyacrylamide gels and then transferred to nitrocellulose membranes (Protran BA, Whatman, Germany) using an electroblotting apparatus (Bio-Rad Laboratories, Hercules, CA). The blots were blocked in Starting Block buffer (Pierce, Rockford, IL) and incubated with indicated antibodies diluted in block buffer containing 0.01% (vol/vol) Tween-20 overnight at 4°C. Blots were washed for 30 minutes in trishydroxymethylaminomethane-buffered saline (10 mM Tris-HCl [pH 8.0], 150 mM NaCl) containing 0.05% (vol/vol) Tween-20. Proteins were detected with an infrared fluorescence detection system and secondary antibodies (Odyssey and IRDye 680 antibody, LI-COR Biosciences, Germany). The densitometric analysis of bands was performed with Image Quant software. To confirm equal protein loading and blotting, total protein on the blots was detected with a MemCode reversible protein stain kit (Pierce, Rockford, IL).

Statistical Analysis.

All measurements were performed in duplicate. Statistical significance of differences between groups was evaluated by way of analysis of variance and the Student t test, and significance was set at P < 0.05.


AMP-DNM Improves Glucose Homeostasis in ob/ob Mice.

C57BL/6J ob/ob mice were fed for 5 weeks with either standard AM-II chow or AM-II chow supplemented with 100 mg/kg bw/day AMP-DNM. Lean C57BL/6J mice were treated the same way and served as control groups. The compound was well tolerated and caused no overt side effects. As described,17 AMP-DNM markedly reduced blood glucose, hemoglobin A1c, and insulin levels and homeostasis model assessment score index in ob/ob mice (Table 1). In lean mice blood glucose somewhat decreased with AMP-DNM treatment but the effect was not significant and treatment did not result in hypoglycemia. Treatment with 100 mg/kg bw/day AMP-DNM reduced total body weight gain by 16% and 5% in ob/ob and lean mice respectively. It is worth noting that a lower dose of 25 mg/kg bw/day AMP-DNM reduces plasma glucose and restores glucose tolerance in the absence of reduction of body weight gain.15 AMP-DNM reduced plasma glycosphingolipid levels in both ob/ob and lean mice (GlcCer in ob/ob control 12.9 ± 2.8 nmol/mL; ob/ob AMP-DNM 3.7 ± 1.0; lean control 6.2 ± 0.6; lean AMP-DNM 2.9 ± 0.6) as well as liver glycosphingolipid levels (GlcCer in control ob/ob 55.8 ± 11.0 nmol/g wet weight; AMP-DNM 42.2 ± 16.4). This was accompanied by increased insulin signaling in the liver at the level of Akt and mTOR in AMP-DNM treated ob/ob mice (Fig. 1).

Table 1. Plasma Metabolic Parameters and Lipids of Lean and ob/ob Mice Treated with 0 or 100 mg/kg AMP-DNM
VehicleAMP-DNM (100 mg/kg)VehicleAMP-DNM (100 mg/kg)
  • Data are expressed as the mean ± standard deviation. Lean or ob/ob mice were treated with 0 (vehicle) or 100 mg/kg/day AMP-DNM mixed into the diet. Measurements were conducted after 5 weeks of treatment.

  • Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; AUC GTT, area under the glucose tolerance curve; HbA1c, hemoglobin A1c; HOMA, homeostasis model assessment score; NM, not measured.

  • *

    P < 0.05,

  • **

    P < 0.01,

  • ***

    P < 0.001 AMP-DNM versus vehicle-treated mice.

General parameters    
 Body weight (g) at treatment start23.2 ± 0.924.1 ± 1.432.9 ± 1.933.6 ± 1.7
 Body weight (g) at treatment end28.3 ± 2.527.4 ± 2.149.0 ± 2.342.7 ± 2.3***
 Liver weight (g)1.4 ± 0.11.5 ± 0.14.5 ± 0.52.5 ± 0.4***
 Liver weight/body weight (%)5.0 ± 0.15.6 ± 0.29.1 ± 0.36.0 ± 0.3***
Metabolic parameters    
 Plasma insulin (ng/mL)0.8 ± 0.2NM22.8 ± 13.16.3 ± 2.7**
 Blood glucose fasted (mmol/L)7.4 ± 1.3NM10.1 ± 1.87.3 ± 0.7**
 Blood glucose nonfasted (mmol/L)10.0 ± 0.48.8 ± 1.116.3 ± 4.011.6 ± 1.6*
 HbA1C (%)4.9 ± 0.74.8 ± 0.59.6 ± 1.25.1 ± 1.1***
 HOMA index6.0 ± 2.0NM259 ± 17450 ± 25**
 AUC GTT1,236 23921737
 AST70.2 ± 13.2 138.5 ± 20.4137.0 ± 20.2
 ALT29.3 ± 3.5 97.0 ± 15.0106.3 ± 15.0
Plasma lipids (nmol/mL)    
 Ceramide6.8 ± 1.46.3 ± 0.815.5 ± 2.912.9 ± 3.0
 Glccer6.2 ± 0.62.9 ± 0.6***12.9 ± 2.83.7 ± 1.0**
 GM30.25 ± 0.020.13 ± 0.02***0.26 ± 0.040.13 ± 0.01**
 GM20.63 ± 0.010.53 ± 0.441.8 ± 0.41.2 ± 1.2
 GM2-GL18.4 ± 1.64.0 ± 0.9***30.0 ± 6.66.7 ± 0.2**
Figure 1.

AMP-DNM increases insulin signaling. We treated ob/ob mice and lean controls (n = 5 per group) with 0 or 100 mg/kg bw AMP-DNM mixed into the diet. After 5 weeks of treatment, animals were fasted for 4 hours and infused with vehicle (0.9% physiological salt solution) or insulin solution for 30 minutes in the bloodstream through the hepatic portal vein. Livers were harvested and the levels of Akt mTOR were detected using western blot analysis.

AMP-DNM Decreases Hepatic Fat Content.

We next examined if the beneficial effects of AMP-DNM on glucose homeostasis influenced liver morphology and fat content. Histology of ob/ob liver sections, stained with hematoxylin-eosin (Fig. 2A) and Oil-Red-O (Fig. 2B), revealed extensive steatosis in the centrilobular area with hepatocytes containing micro- and macro-lipid droplets. Interestingly, the liver weights of AMP-DNM–treated ob/ob mice were significantly lower (Table 1), and histological analysis revealed a marked reduction in macrovesicular steatosis (Fig. 2B). Normal liver histology was not noticeably altered in lean mice that had received AMP-DNM (Fig. 2A). Biochemical analysis confirmed robust reductions of hepatic TGs in both ob/ob and lean mice treated with AMP-DNM (Fig. 2D). There was no effect on liver cholesterol content after AMP-DNM treatment in lean or ob/ob mice (Fig. 2E). Serum levels of alanine aminotransferase and aspartate aminotransferase were only modestly increased in ob/ob mice compared with lean mice. Drug treatment had no statistically significant effect on the circulating levels of the liver enzymes (Table 1).

Figure 2.

AMP-DNM decreases hepatic fat content. We treated ob/ob mice and lean controls (n = 5 per group) with 0 or 100 mg/kg bw AMP-DNM mixed into the diet. After 5 weeks, livers were dissected and processed for histology applying staining with (A) hematoxylin-eosin, (B) Oil-Red-O, or (C) periodic acid-Schiff. (D,E) Biochemical measurements of total liver TG (D) and total cholesterol levels (E). Data are expressed as the mean ± standard error of the mean (n = 5 per group). *Significant difference in lipid parameters between treatment groups and the control group (P < 0.05).

AMP-DNM Inhibits Hepatic Lipogenic and Glucose Production Pathways.

We studied hepatic lipid metabolism to understand the molecular mechanism behind AMP-DNM–induced reduction of liver TG content. The transcription factor SREBP1c is activated by insulin and enhances the transcription of genes required for fatty acid and TG biosynthesis in the liver. In ob/ob mice, nuclear SREBP1c levels are extremely high and fatty acid synthesis is accelerated. Using real-time polymerase chain reaction, we indeed found that expression of Srebp1c and its target genes were up-regulated in the untreated ob/ob mice compared with untreated lean mice (Fig 3A). Treatment with AMP-DNM had no effect on SREBP1c at the messenger RNA (mRNA) level. However, various lipogenic genes were down-regulated after AMP-DNM treatment suggesting that SREBP1c activity was decreased (Fig. 3A). The change in mRNA level of Fas was also confirmed at the level of protein expression (Fig. 3B). The expression of Pparγ was increased in ob/ob mice compared with lean mice and was not corrected by AMP-DNM treatment. The expression of genes involved in fatty acid β-oxidation is known to be increased in ob/ob mice compared with lean mice.23 Treatment with AMP-DNM did not further increase the expression of genes involved in β-oxidation, such as Pdk4, Ccpt1a, and Lcad (Fig. 3C). AMP-DNM treatment significantly reduced the mRNA expression levels of glucose-6-phosphatase, confirming that glucose output from the liver is reduced (Fig. 3D). Glycogen content in livers from AMP-DNM–treated ob/ob mice was increased compared with livers from untreated animals (Fig. 2C). Both reduced glucose-6-phosphatase expression and increased glycogen content are indicative of increased insulin signaling in the liver.

Figure 3.

AMP-DNM inhibits hepatic lipogenic and glucose production pathways. Total protein and RNA were extracted from frozen liver samples of ob/ob mice and lean controls (n = 5 per group) treated with 0 or 100 mg/kg bw AMP-DNM for 5 weeks. (A) AMP-DNM reduced mRNA encoding enzymes involved in hepatic lipogenesis without changing SREBP1c mRNA. ATP CL, ATP citrate lyase; FAS, fatty acid synthase; PPARγ, peroxisome proliferator-activated receptor gamma; SCD1, stearoyl–coenzyme A desaturase 1. (B) AMP-DNM reduced protein expression level of fatty acid synthase. (C) AMP-DNM did not change the expression of genes involved in β-oxidation. CPT1a, carnitine palmitoyl transferase 1; LCAD, long chain–specific acyl–coenzyme A dehydrogenase; PDK4, pyruvate dehydrogenase kinase, isoenzyme 4. (D) AMP-DNM reduced glucose-6-phosphatase mRNA. Data are expressed as the mean ± standard error of the mean (n = 5 per group). *Significant difference in mRNA expression between untreated ob/ob mice compared with either lean controls or AMP-DNM–treated ob/ob mice (P < 0.05).

AMP-DNM Corrects the Gene Expression Profile in the ob/ob Liver.

To obtain a comprehensive picture of the total impact of AMP-DNM treatment, we compared liver gene expression profiles using microarrays of the lean and ob/ob mice fed by control or AMP-DNM–supplemented diet. A large number of genes (4,013) was found to be significantly differentially regulated in ob/ob mice compared with lean animals on a control diet (adjusted P value < 0.05, fold change ≥1.4). Pathway analysis in a MetaCore suit revealed that these genes fell into numerous differentially expressed pathways, among which extracellular matrix remodeling, glucose, lipid, and glutathione metabolism (Table 2). (For a complete list of differentially regulated genes, see Supporting Information.) As expected, numerous genes corresponded to those previously reported to be differentially regulated in ob/ob mice compared with lean mice.24, 25

Table 2. Most Affected Differentially Regulated Pathways in Livers of ob/ob Mice in Comparison with Lean Controls
Process and PathwayChanged Genes/Total Genes in Pathway
  1. The pathways are significantly differentially regulated (P < 0.001); the number of regulated genes and total number of genes per pathway are indicated.

  2. Abbreviations: dATP, deoxyadenosine triphosphate; dITP, deoxyinosine triphosphate; PPAR, peroxisome proliferator-activated receptor.

Regulation of lipid metabolism 
 PPAR regulation of lipid metabolism10/20
 Regulation of fatty acid synthase activity in hepatocytes8/16
 G-alpha(q) regulation of lipid metabolism10/23
Cell adhesion and cytoskeleton remodeling 
 Cell adhesion and ECM remodeling21/51
 Role of tetraspanins in the integrin-mediated cell adhesion15/37
 Endothelial cell contacts by nonjunctional mechanisms11/24
Immune response 
 Lipoxin inhibitory action on neutrophil migration15/35
 Lectin Induced complement pathway15/38
 Classic complement pathway15/40
Carbohydrates metabolism 
 Glycolysis and gluconeogenesis7/13
Vitamin and cofactor metabolism 
 Glutathione metabolism15/39
 Vitamin K metabolism4/4
 dATP/dITP metabolism21/54
 (L)-selenoaminoacids incorporation in proteins during translation16/38
 Regulation of cell cycle14/32

Treatment of ob/ob mice with AMP-DNM had a remarkable effect on total gene expression. We found that 81% of the 4,013 differentially regulated genes were to a certain extent corrected toward the lean phenotype. Moreover, 45% (1,459) of these showed a complete correction in expression (Fig. 4). The pathways that were most corrected in gene expression are listed in Supporting Table 2. Notably, they indicate changes in lipid and carbohydrate metabolism, matching our biochemical and real-time polymerase chain reaction observations (Figs. 2 and 3).

Figure 4.

Gene expression changes induced by AMP-DNM. Diagram shows the number of genes differentially expressed in the livers of obese mice compared with their lean counterparts. Genes that are not corrected by AMP-DNM treatment (ob/ob) are represented by the light gray box; genes corrected to a certain extent toward lean expression profile by AMP-DNM treatment (ob/ob + AMP-DNM) are represented by the dark gray box. The white box represents genes completely corrected toward lean expression profile after treatment with AMP-DNM (P < 0.05; fold change ≥1.4).

Pathways regulated in lean animals treated with AMP-DNM were, not surprisingly, glutathione, cholesterol, and sphingolipid metabolism (data not shown). Expression changes in these pathways indicated enhanced detoxification, induced glycosphingolipid synthesis, and, as reported earlier,26 increased cholesterol synthesis.

AMP-DNM Reduces Hepatic Markers Associated with Matrix Remodeling and Inflammation.

Matrix remodeling and immune responses belonged to the pathways most influenced by AMP-DNM treatment (Supporting Table 2). This prompted us to look in detail at changes in expression of specific genes.

Interestingly, the expression level of almost all the genes involved in extracellular matrix homeostasis corrected towards the lean phenotype by AMP-DNM treatment (apart from procollagen type XXVII alpha 1, which was even more down-regulated) (Table 3). The expression level of 9 out of 23 corrected genes did not differ from that in the lean animals. Of special interest was correction of collagens 1, 3, and 4, tissue inhibitor of metalloproteinase 1 (Timp) and vinculin (Vcl), well-known markers of fibrosis.27, 28 Picro-sirius red staining indeed showed slightly increased deposition of fibrillar collagens in the centrilobular area of ob/ob livers. Treatment of ob/ob mice with AMP-DNM largely normalized this pattern to that of control lean mice (data not shown).

Table 3. Differentially Expressed Genes Involved in Extracellular Matrix Remodeling
SymbolGene NameFold Change in Gene Expression
ob/ob Control/Lean Controlob/ob AMP-DNM/Lean Control
  • Data represent gene expression levels in ob/ob mice versus lean and AMP-DNM–treated ob/ob mice versus lean measured by microarray analysis.

  • *

    P < 0.05.

Col1a1Procollagen, type I, alpha 19.35*3.05*
Col1a2Procollagen, type I, alpha 22.80*1.34
Col3a1Procollagen, type III, alpha 16.39*1.49
Col4a1Procollagen, type IV, alpha 12.31*1.43
Col4a2Procollagen, type IV, alpha 21.93*1.24
Col4a5Procollagen, type IV, alpha 52.13*1.92
Col5a1Procollagen, type V, alpha 12.16*1.06
Col6a1Procollagen, type VI, alpha 11.81*1.11
Col6a1Procollagen, type VI, alpha 12.01*1.17
Col12a1Procollagen, type XII, alpha 13.30*2.45*
Col15a1Procollagen, type XV2.72*1.79
Col16a1Procollagen, type XVI, alpha 11.69*−1.07
Col20a1Collagen, type XX, alpha 1−1.99*−1.80*
Col27a1Procollagen, type XXVII, alpha 1−2.15*−2.81*
Mmp12Matrix metalloproteinase 1257.73*31.90*
Mmp13Matrix metalloproteinase 1320.53*9.12*
Mmp15Matrix metalloproteinase 15−1.63*−1.51*
Mmp23Matrix metalloproteinase 231.89*1.24
Mmp24Matrix metalloproteinase 243.20*1.81
Mmp24Matrix metalloproteinase 242.36*1.69*
Mmp24Matrix metalloproteinase 24−1.60*−1.21
PlatPlasminogen activator, tissue7.47*2.54*
Timp1Tissue inhibitor of metalloproteinase 14.70*2.17*

Although the liver of ob/ob mice does not show a marked inflammatory phenotype,29 a significant increase in mRNA expression of several chemokines (monocyte-chemo attractant protein-1; macrophage inflammatory protein-1α; macrophage inflammatory protein-1β) and their receptors (chemokine receptors 1 and 2) was observed in ob/ob mice livers compared with lean mice (Table 4). We also found increased expression of several macrophage-associated markers although the total number of macrophages did not increase as indicated by the unchanged expression of F4/80, a macrophage-specific marker (Table 4). This implies that the resident liver cells are the source of these cytokines. Treatment with AMP-DNM normalized the expression of the above-mentioned chemokines and macrophage-associated markers such as CD11c, osteopontin, and YM1 (Table 4).

Table 4. Differentially Expressed Genes Involved in Inflammation
SymbolGene NameLean Controlob/ob Controlob/ob AMP-DNM
  • Data represent gene expression levels relative to housekeeping genes measured by real-time polymerase chain reaction shown as the mean ± standard error of the mean (n = 5 per group).

  • *

    Significant difference in mRNA expression between untreated ob/ob mice compared with lean controls and AMP-DNM–treated ob/ob mice compared with untreated ob/ob mice (P < 0.05).

F4/80Cell surface glycoprotein F4/807.85 ± 3.847.38 ± 2.475.26 ± 1.11
Mcp-1Monocyte-chemoattractant protein-10.83 ± 0.432.71 ± 0.91*1.53 ± 0.76
MIP1αMacrophage inflammatory protein-1α0.57 ± 0.282.15 ± 1.06*0.64 ± 0.27*
MIP1βMacrophage inflammatory protein-1β2.01 ± 0.305.53 ± 2.19*2.78 ± 0.85
Ccr1Chemokine-receptor 10.25 ± 0.101.20 ± 0.73*0.54 ± 0.23
Ccr2Chemokine-receptor 20.76 ± 0.191.91 ± 0.96*0.56 ± 0.33*
CD11cIntegrin alpha-X1.99 ± 0.798.50 ± 2.95*2.95 ± 0.93*
YM1Chitinase-3-like protein 30.39 ± 0.199.89 ± 7.56*3.12 ± 2.18*
OPNOsteopontin218 ± 791166 ± 491*355 ± 110*


Leptin-deficient ob/ob mice rapidly develop insulin resistance and concomitant hyperinsulinemia. These animals also develop a prominent hepatic steatosis that is attributed to excessive de novo lipogenesis driven by the chronic high levels of insulin.30

We15 and others16 have shown that inhibition of glycosphingolipid synthesis has a beneficial effect on hepatic and peripheral insulin sensitivity and hyperinsulinemia in various rodent models of insulin resistance and type 2 diabetes. The present study shows that AMP-DNM treatment prevents hepatomegaly in ob/ob mice and results in improvements in liver insulin signaling. At the same time, fatty acid synthesis in the liver is decreased and the development of abundant steatosis in the centrilobular area is markedly prevented concurrent with normalization of glycogen storage in livers of treated animals.

The metabolic changes by which AMP-DNM exerts its beneficial effect on the fatty liver of ob/ob mice warrant discussion. We hypothesized that restoring insulin sensitivity in ob/ob mice by AMP-DNM treatment could break the vicious cycle of insulin resistance and increased lipogenesis in the livers of ob/ob mice. The hepatic steatosis found in ob/ob mice is mostly attributed to excessive de novo lipogenesis driven by high levels of insulin.30 Reducing glycosphingolipids with AMP-DNM restored insulin signaling in various organs of the ob/ob mice, including the liver. Circulating insulin was markedly lower in AMP-DNM–treated ob/ob mice, reflecting their overall improved insulin sensitivity. The observed reduction in expression of genes involved in fatty acid synthesis is likely a direct consequence of reduced insulin concentration in AMP-DNM–treated mice.

The storage of TGs in liver is not toxic per se and protects the organ from free fatty acid overload. Steatosis progresses to NASH when the adaptive mechanisms that normally protect hepatocytes from fatty acid-mediated lipotoxicity become exhausted. This process triggers responses that involve the activation of hepatic stellate cells and the production of chemokines by myofibroblasts to attract various kinds of inflammatory cells to the liver. However, the ob/ob mice liver is not an ideal model to study NASH; due to the lack of leptin in these mice, full activation of stellate cells does not occur, thereby preventing development of steatohepatitis and frank fibrosis.29, 31, 32 In accordance, α smooth muscle actin–positive stellate cells were not observed in livers from ob/ob mice (data not shown).

We noted slightly increased picro-sirius red staining in the centrilobular area of livers of ob/ob mice that was absent in livers of ob/ob mice treated with AMP-DNM. Immunohistological analysis for collagen type I confirmed this observation (data not shown). Gene expression analysis using a microarray approach showed that pathways involved in inflammation and matrix remodeling were only slightly activated in the livers of ob/ob mice compared with lean mice. We could not detect increased numbers of total macrophages in the ob/ob mice livers by looking at F4/80 expression, suggesting that there was no recruitment of inflammatory macrophages toward the liver. Interestingly, markers generally associated with activated macrophages (such as YM1) were elevated in ob/ob mouse livers, and most likely originated from endogenous sources such as Kupffer cells. In addition, it was found that CD11c, one of the defining markers of inflammatory macrophages in adipose tissue, was highly expressed in ob/ob mouse livers. Treatment with AMP-DNM normalized the expression of these inflammatory markers. This is in line with the earlier reported anti-inflammatory capacity of AMP-DNM.17 This finding is of particular interest, because inflammation is considered to be a contributing factor to the development of insulin resistance and obesity-related liver injury.31

NK T cells are suggested to play a protective role in the inflammatory response in hepatic steatosis, and the reduction in NK T cell numbers in humans and animal models are implicated in disease severity.33 NK T cell numbers have been found to be reduced in ob/ob mouse livers. We therefore tested the possibility that NK T cell numbers increased after reduction of steatosis with AMP-DNM in our mouse model. We were able to detect a few NK T cells by way of immunohistochemical analysis in untreated ob/ob mice livers, but we were unable to detect significant changes in their amount upon treatment (Supporting Fig. 1).

Next to metabolic corrections in the liver and local tempering of inflammation, other factors might contribute to the improvement in the steatotic livers of AMP-DNM–treated ob/ob mice. We recently reported that AMP-DNM treatment enhances adipocyte function and reduces inflammation in adipose tissue of ob/ob mice.17 As a result of the improved adipose tissue function imposed by AMP-DNM, less inflammatory mediators such as Ccl2 and osteopontin as well as free fatty acids are released in the circulation. Other organs, such as the liver, might profit from this. In the same study, AMP-DNM treatment was found to increase production of adiponectin by adipose tissue. This adipokine, reported to be reduced in patients with obesity, insulin resistance, and cardiovascular disease,34 is known to exert a beneficial, insulin-sensitizing effect on various tissues. Furthermore, hypoadiponectinemia is a feature of NASH independent of insulin resistance.35 Importantly, recent studies show that adiponectin also protects against alcoholic liver disease by enhancing fat oxidation, reducing lipid synthesis, and preventing hepatic steatosis.36

The role of glycosphingolipids in diabetes development has been firmly established by us and others.37 In obesity-induced insulin resistance the excess amounts of glycosphingolipids disturb insulin signaling through interference with the insulin receptor. Another link between glycosphingolipids and insulin resistance has been reported by Ilan and coworkers,38 who showed that the intraperitoneal injection of glycosphingolipid emulsions protect beta-cell function in the pancreas through modulation of NK T cell activity in a cytokine-mediated model of insulin resistance type 1 (Cohen diabetic rat). Whether the beneficial effects of AMP-DNM are partly attributed to modulation of NK T cell activity remains to be established.

Our present study and earlier investigations in rodent models shed light on the therapeutic potential of AMP-DNM for type 2 diabetes and hepatic steatosis. It is well-documented that in diabetic subjects, de novo lipogenesis is elevated despite insulin resistance in the gluconeogenesis pathway; however, the precise cause of this selective hepatic insulin resistance is unknown.7 The noted ability of AMP-DNM to increase insulin sensitivity and concomitantly reduce lipogenesis in ob/ob mice is therefore of particular interest.


We thank Peter Dubbelhuis and Sijmen Kuiper for outstanding technical expertise.