Whole-body Insulin Resistance in the Absence of Obesity in FVB Mice With Overexpression of Dgat1 in Adipose Tissue

The transgene contains, from the 5′-end to the 3′-end, a 5.4-kb aP2 promoter, a full-length human Dgat1 (hDgat1) cDNA, and a 0.97-kb PCR fragment of human growth hormone genomic DNA containing two introns (four and five), three exons (three, four, and five), and the poly A signal (Fig. 1A). Transgenic mice were derived from fertilized eggs of the CBA/C57BL6 hybrid. Lines of aP2-Dgat1/FVB mice were developed after six backcrosses with the wild-type FVB strain (The Jackson Laboratory, Bar Harbor, ME). Genotyping was carried out by PCR or Southern blot using genomic DNA extracted from mouse tails.

Mice were fed normal chow or Western-type high-fat diets (TD 88137; Harlan Teklad, Madison, WI). The high-fat diet contains 42% calories from anhydrous milk fat. The time and duration of high-fat diet treatments are indicated in the text or relevant figure legends.

Intraperitoneal glucose tolerance tests were carried out after a 12-h fast. Glucose (1 g/kg body wt) was administered by intraperitoneal injection. Blood samples were obtained from tail veins before injection and 30, 60, and 120 min after injection, and blood glucose levels were determined by glucometry.

Blood was drawn from 12-h–fasted mice. Plasma obtained after centrifugation was snap frozen in liquid nitrogen and stored at −80°C until use. Glucose and FFA levels were determined using enzymatic colorimetric assay kits from Sigma and Wako Chemicals (Neuss, Germany), respectively. Insulin, leptin, and adiponectin levels were determined using enzyme-linked immunosorbent assay or radioimmunoassay methods (Linco Research).

DGAT activity was measured in vitro in membrane fractions isolated from adipose tissue or liver. Assays were carried out (using 150 mmol/l MgCl₂ in the incubation), exactly as previously described (10).

Lipids were extracted from liver using chloroform/methanol/HCl as described (11). [³H]triolein (0.25 μCi) was added to homogenates as an internal control for triglyceride recovery. Triglyceride mass in lipid extracts was enzymatically determined with a colorimetric kit (Trig/GB; Roche Diagnostics). Lipid extraction from muscle specimens was performed similarly, but extreme care was exercised during tissue dissection to avoid perimuscular/perivascular fat contamination.

The assay was based on previously described methods (12,13). Adipocyte lipolysis was measured in triplicate in a 96-well microplate in the presence or absence of isoproterenol (10⁻⁷ mol/l). At the end of a 2-h incubation at 37°C, an aliquot of culture medium was carefully transferred into a new 96-well microplate for free glycerol measurement using a glycerol determination kit from Sigma-Aldrich. Glycerol levels were normalized by the cell numbers, which were determined in the original microplate by measuring DNA contents. Cells in culture wells were first lysed in 0.1% Triton X-100 and digested overnight by proteinase K (20 μg/ml) at 50°C, and Hoechst 33342 was added (5 μg/ml). Fluorescence was measured at λ = 360 nm (excitation) and λ = 460 nm (emission), using a fluorescence plate reader (Cytofluor II; PerSeptive Biosystems). A linear standard curve was made using calf thymus DNA.

Muscle 2-deoxyglucose uptake was carried out in isolated intact soleus muscles based on published methods (14,15). Pairs of soleus muscles were isolated from overnight-fasted mice, and one was used for basal and the other for insulin (10 nmol/l)-stimulated uptake of 2-deoxy-d-[1-³H]glucose. [¹⁴C]mannitol was used to control for nonspecific radioactivity trapping in the extracellular spaces of the muscle. After subtracting nonspecific [¹⁴C]mannitol trapping from the total 2-deoxy-d-[1-³H]glucose counts in the muscle, specific 2-deoxyglucose uptakes were expressed as counts per unit wet muscle weight.

Adipose 2-deoxyglucose uptake was measured as described (16,17), with minor modification. Briefly, adipose tissue was cut into ∼10-mg pieces and incubated for 30 min at 37°C under 5% CO₂ in DMEM (25 mmol/l glucose) supplemented with 2% BSA. After three washes with Krebs-Ringer bicarbonate HEPES-based buffer (KRBH) containing 1% BSA (pH 7.4), tissues were incubated for 15 min at 37°C under 5% CO₂ in 0.4 ml KRBH containing 0.1% BSA with (insulin stimulated) or without (basal) 100 nmol/l insulin. The uptake was initiated by adding 0.1 ml of the same buffer containing 10 mmol/l glucose and 2 μCi 2-deoxy-d-[1-³H]glucose. At the end of the 20-min incubation at 37°C, the uptake was terminated by adding 20 ml ice-cold KRBH buffer, followed by three washes with the same ice-cold KRBH buffer. After tissue digestion in 0.5 N NaOH, [³H] was scintillation counted and data processed in the same fashion as described above for muscles.

RNA isolation, reverse transcription using random hexamer primers, and subsequent RT-PCR (MJR Cycler, Reno, NV) were all carried out as previously described (18). Sequence-specific primers were shown in Table 1. Levels of gene expression were expressed as ratios relative to β-actin.

Western blot was performed as described (10) using an enhanced chemiluminescence immunodetection system (Amersham Life Science). Immunoblotting was carried out using specific antibodies to sterol regulatory element–binding protein-1c (SREBP1c) (gift from Jay D. Horton, University of Texas Southwestern Medical Center) and to both phospho–protein kinase B (Ser⁴⁷³) and phospho-insulin receptor (Tyr^1,131)/(Tyr^1,146) (Cell Signaling Technology, Beverly, MA). β-Actin was used as a loading control.

Adipose tissue was fixed in Bouin’s fixative or 4% paraformaldehyde in 1 × PBS (pH 7.4) for 2 days. Specimens were washed with 75% ethanol three times and then soaked in 75% ethanol for >12 h. After dehydration in graded ethanol solutions and paraffin embedding, the tissue was cut in 7-μm sections and stained with hematoxylin/eosin before microscopic examination.

Data were expressed as means ± SD. Differences between study groups were analyzed by Student’s t test using Statistica V6.0. A two-tailed P value of <0.05 was used to indicate statistical significance.

We used the 5.4-kb aP2 promoter, which is widely used for adipose transgene expression (1,5,6,19), to direct the expression of a full-length hDgat1 cDNA (10,20). hDgat1 shares 80% identity in DNA sequence with mouse Dgat1. Overexpression of hDgat1 in Hep G2 and McRH7777, human and rodent cell lines, respectively, resulted in similar increases in intracellular triglyceride levels (data not shown), suggesting that hDgat1 is equally functional in humans and rodents. As shown in Fig. 1A, the ap2-Dgat1 transgene also contains a 3′-human growth hormone genomic sequence, serving as a RNA processing substrate. We obtained four aP2-hDgat1 transgenic lines. Southern blot (Fig. 1B) showed high copy numbers of the transgene in positive founder mice (compare the band intensity of the transgene with that of the endogenous Dgat1), and the same digest pattern in different lines was consistent with a tandem repeat arrangement of the transgene in each line. Initially derived as B6CBA/F1 hybrids, transgenic mice were backcrossed onto the FVB genetic background (six consecutive backcrosses), and only male mice were used in this study. Adipose expression of the transgene was shown by RT-PCR (Fig. 1C) using a pair of primers specific for the transgene (Fig. 1, legend). Real-time PCR using a pair of primers common to both endogenous and the transgene Dgat1 showed an approximately threefold increase (P < 0.01) in overall Dgat1 mRNA levels in both epididymal and inguinal white adipose tissue of the transgenic mice line II (Fig. 1D). Similar quantitative results were obtained from mouse line IV (data not shown). Functional assays with isolated microsomal membranes showed a more than twofold increase (P < 0.01) in DGAT activity in transgenic white adipose tissue compared with wild-type white adipose tissue (Fig. 1E).

No phenotypic difference was observed between aP2-Dgat1/FVB and the wild-type littermates when animals were maintained on a chow diet as long as 44 weeks. After challenge with a 42% high-fat diet (Harlan Teklad) treatment for 4–8 weeks, both the transgenic and wild-type mice gained weight on the high-fat diet, and aP2-Dgat1 transgenic mice were only 5–8% heavier than the age-, sex-, and diet-matched wild-type controls; line II (Dgat1-II) was 7% heavier (P = 0.052) and line IV (Dgat1-IV) 5% heavier (P = 0.044) than the wild-type controls (Table 2). To better characterize aP2-Dgat1/FVB mice, detailed metabolic studies were carried out in Dgat1-II. Figure 2A and B show growth curves of the transgenic and wild-type mice up to 45 weeks. No significant difference in body weight was detected between the transgenic and wild-type mice on chow diets (Fig. 2A). On a 42% high-fat diet started after weaning, both aP2-Dgat1 and wild-type mice were heavier. The weight gain in the transgenic mice was only marginally more than the wild-type mice on this high-fat diet, with the biggest difference of ∼8% at week 28 (Fig. 2B). This growth curve is in agreement with the weight data presented in Table 2. Consistent with the slight weight gain, the transgenic mice consumed ∼8% (P < 0.05) more food than the wild-type mice during this entire period of 10 weeks on high-fat diet (Fig. 2C).

In contrast to the aP2-Dgat1/C57BL6 mice (4), there was no adipocyte hypertrophy in aP2-Dgat1/FVB mice. Adipocyte cell sizes were similar between transgenic and wild-type mice in both retroperitoneal and inguinal fat pads (Fig. 2D). Whole-body fat composition measured by dual-energy X-ray absorptiometry also showed no statistically significant difference between aP2-Dgat1 and the wild-type control mice (Fig. 2E). Since there was no evidence of fat expansion despite increased Dgat1 expression and DGAT activity in the transgenic adipose tissue, we measured rates of lipolysis in adipocytes isolated from mice fed high-fat diets. Compared with the wild-type adipocytes, basal lipolysis rates in aP2-Dgat1 adipocytes were >25-fold (P < 0.05) higher. Isoproterenol increased lipolysis rates in both the transgenic and the wild-type adipocytes, but the aP2-Dgat1 adipocytes maintained an eightfold (P < 0.05) higher rate compared with the wild type (Fig. 2F). This elevated adipose lipolysis was further supported by the finding that the 12-h fasting plasma FFA levels in aP2-Dgat1 mice were ∼60% (P < 0.05) higher than in the wild type (Fig. 2G).

Despite the lack of significant obesity, aP2-Dgat1 mice developed marked fasting hyperinsulinemia (more than fivefold increase in fasting plasma insulin) and hyperglycemia (fasting plasma glucose >200 mg/dl) after 4–8 weeks of high-fat diet treatment in 34-week-old male mice (Table 2). To further characterize the insulin-resistant phenotype, male Dgat1-II and control mice in several different age-groups were challenged with a high-fat diet for 6 weeks, and intraperitoneal glucose tolerance tests were performed. Although significant differences were not detected between groups of young transgenic and wild-type mice who were switched to high-fat diets at <20 weeks (data not shown), marked glucose intolerance was observed in older aP2-Dgat1 mice who started high-fat diets at either 24 or 29 weeks (Fig. 3A and B). Fasting blood glucose reached 200 mg/dl in 35-week-old Dgat1-II mice in comparison with 125 mg/dl in the wild-type control mice (P < 0.05) (Fig. 3B).

Fasting plasma levels of insulin, adiponectin, and leptin were measured in 35-week-old male mice following a 6-week high-fat diet challenge. Insulin levels in the transgenic mice were more than fivefold (P < 0.01) higher than those in the wild-type mice (Fig. 3C), indicating insulin resistance. There was no difference in adiponectin levels (Fig. 3D). However, plasma levels of leptin in aP2-Dgat1 mice were about twofold (P < 0.05) higher than the wild type (Fig. 3E). Since food intake was not suppressed in aP2-Dgat1 mice but actually increased (Fig. 2C), the elevated leptin levels in these mice suggest leptin resistance.

Because of increased lipolytic activity in aP2-Dgat1 adipose tissue, with resultant elevation of plasma FFA levels, we investigated whether muscle and liver had alteration in lipid contents and insulin sensitivity. Triglyceride in soleus muscle was not increased in the transgenic mice (Fig. 4A). Furthermore, 2-deoxyglucose uptake in isolated soleus muscle strips (Fig. 4B) indicated no difference in insulin sensitivity (Fig. 4C) between the transgenic and the wild-type mice. In contrast, fatty liver was observed in aP2-Dgat1 mice (Fig. 4D). Livers in transgenic mice were enlarged by 20% (P < 0.05) by weight (Fig. 4E), and quantification of liver triglycerides revealed nearly threefold (P < 0.05) greater triglycerides per unit liver mass in the transgenic livers (Fig. 4F). The development of fatty liver was accompanied by markedly elevated intracellular protein levels of SREBP1c, a master regulator of hepatic lipogenesis (Fig. 4G, top). To investigate whether hepatic insulin signaling was altered, we measured phosphorylation levels of the insulin receptor and the downstream protein kinase B (Akt) at two well-recognized phosphorylation sites, Tyr^1,146 (insulin receptor) and Ser⁴⁷³ (Akt), respectively. The levels of phosphorylated insulin receptor and Akt were both decreased markedly in aP2-Dgat1 liver (Fig. 4G, middle), indicating that the aP2-Dgat1 fatty liver was insulin resistant.

Hepatic DGAT activity also increased by ∼30% (P < 0.01) (Fig. 4H, top), consistent with a general upregulation of hepatic lipogenesis. However, the increased DGAT activity was not associated with an increase in DGAT1 mRNA levels (Fig. 4H, bottom), consistent with our previous finding of posttranscriptional regulation of DGAT activity (10). Additionally, the lack of increase in DGAT1 mRNA in the liver of aP2-Dgat1 mice indicates that significant expression of the transgene in the liver is unlikely. We also measured genes involved in gluconeogenesis (the catalytic subunit of glucose-6-phosphatase and phosphoenolpyruvate carboxykinase) and genes involved in the glycolytic pathway (liver pyruvate kinase and glucokinase). Importantly, glucose-6-phosphate was increased fivefold (P < 0.01), suggesting upregulation of hepatic glucose production, which was in agreement with hepatic steatosis and hepatic insulin resistance.

To obtain insights into the mechanism by which overexpression of Dgat1 in adipose tissue causes elevations in plasma FFAs and hepatic lipotoxicity, we further examined adipose tissue. We measured basal and insulin-stimulated 2-deoxyglucose uptake (Fig. 5A) and found that the aP2-Dgat1 adipose tissue was insulin resistant compared with the wild-type adipose tissue (insulin sensitivity index: 0.6 vs. 1.25, P < 0.05; Fig. 5B). Then, we compared mRNA levels of several genes involved in glucose and lipid metabolic pathways between aP2-Dgat1 and wild-type adipose tissue (Fig. 5C). We first examined genes in the insulin signaling pathway. As shown in Fig. 5C, Akt2 was downregulated by 50% (P < 0.05) in transgenic mice, which may in part contribute to the decrease of insulin-stimulated 2-deoxyglucose uptake. Phosphatidylinositol 3-kinase tended to be reduced as well (P = 0.07). Next, we looked at genes involved in fatty acid and triglyceride synthesis. DGAT1 mRNA was increased by twofold (P < 0.01) due to exogenous Dgat1 expression. ACC1 (acetyl CoA carboxylase-1) and ACS1 (acyl-CoA synthetase-1), two enzymes involved in de novo fatty acyl-CoA synthesis, were also upregulated by 1-fold (P < 0.01) and 0.7-fold (P < 0.01), respectively. However, genes related to FFA uptake were downregulated; mRNA levels of CD36/FAT (fatty acid translocase) and A-FABP (adipose fatty acid binding protein) were 1.2-fold (P < 0.05) and 0.8-fold (P < 0.05) lower than the wild-type, respectively. Finally, we examined genes related to triglyceride storage/lipolysis. mRNA levels of the newly described adipocyte triglyceride lipase, ATGL (21) (also independently termed as desnutrin [22] and iPLA₂ζ [23]), were 2.9-fold (P < 0.01) higher than the wild type, which supports the observed high lipolysis rates in the aP2-Dgat1 adipose tissue (Fig. 2F).

The present study demonstrated an obesity-independent whole-body insulin-resistant phenotype in FVB transgenic mice with adipose overexpression of Dgat1. Our data support the model depicted in Fig. 6. In this model, overexpression of the Dgat1 gene in adipose tissue of FVB mice led to increased triglyceride synthesis. However, this increase in triglyceride synthesis did not result in triglyceride storage or fat mass expansion within the adipose tissue due to a concomitant increase in lipolysis, most likely mediated by increased expression of the recently identified novel adipose tissue lipase, ATGL (21). There was also evidence of decreased expression of CD36 and A-FABP genes in the adipose tissue, raising the possibility that adipose tissue uptake of FFAs was reduced. A combination of increased lipolysis and decreased FFA uptake could be the basis for the lack of adipocyte hypertrophy in aP2-Dgat1 mice. Under the condition of high-fat feeding, the inability of adipose tissue to store triglycerides resulted in increased levels of circulating FFAs. This, in turn, resulted in excessive FFA delivery to the liver, causing hepatosteatosis, hepatic insulin resistance, and whole-body glucose intolerance, despite sparing of muscle, which retained normal triglyceride content and insulin sensitivity.

The molecular mechanism by which increased DGAT1 activity caused increased lipolysis, upregulation of ATGL, and downregulation of CD36 and A-FABP in the adipose tissue remains speculative at the present time. Nevertheless, these data indicate that adipose overexpression of Dgat1, a gene involved in lipid esterification and triglyceride synthesis, is the primary cause of inappropriate energy redistribution and whole-body insulin resistance in FVB mice under diet-permissive conditions. The possibility that hepatosteatosis and hepatic insulin resistance was caused by overexpression of the aP2-Dgat1 transgene in the liver has been effectively ruled out by the lack of increase in hepatic DGAT1 mRNA levels in aP2-Dgat1 mice. Although hepatic DGAT activity in aP2-Dgat1 mice was 30% increased, this was likely due to posttranscriptional upregulation, as was previously demonstrated for this gene (10). The increased hepatic FFA delivery in aP2-Dgat1 mice was associated with upregulation of SREBP1c and hepatic steatosis, and the posttranscriptional upregulation of DGAT is likely modulated by the physiological response to the increased FFA uptake and/or the physiological status of the hepatocytes (10). In fact, increased DGAT activity (primarily due to posttranscriptional regulation) was also demonstrated in a hyperlipidemic, insulin-resistant hamster model (24).

Our data demonstrating development of systemic insulin resistance due to adipose Dgat1 overexpression are consistent with prior studies demonstrating that Dgat1 deficiency resulted in an obesity-resistant and insulin-sensitive phenotype (8). Similarly, leptin resistance in our aP2-Dgat1 mice is consistent with increased leptin sensitivity in Dgat1-deficient mice (25), although it remains unclear how alterations in Dgat1 expression change leptin sensitivity. Our results, however, are in contrast to that of Chen et al. (4) with the aP2-Dgat1/C57BL6 model. They reported ∼20% greater weight gain in their transgenic mice compared with wild type on a high-fat diet. Although there was a similar two- to threefold increase in Dgat1 expression and activity, adipocyte hypertrophy was an easily recognizable feature in their mouse model. However, insulin resistance was not observed in aP2-Dgat1/C57BL6 mice (4). A likely reason for persistence of normal insulin sensitivity in aP2-Dgat1/C57BL6 mice is that the adipose tissue in those mice maintained the ability to store lipids effectively, thereby avoiding lipotoxicity in other organs. Differences between the two studies may be best explained by the differences in the genetic backgrounds of the mice. Compared with C56Bl6, FVB is more resistant to diet-induced obesity (9). However, when rendered genetically obese by crossing with ob/ob mice (26) or lipoatrophic by crossing with the A-ZIP/F-1 mice (27), FVB exhibited higher plasma FFAs and more severe hyperglycemia than C57BL6. It is likely that the elevated plasma FFA levels in aP2-Dgat1/FVB mice results from genetic interactions between Dgat1 and the FVB background, leading to increased adipose lipolysis, decreased adipose FFA uptake, and inappropriate energy redistribution. The fact that the systemic manifestations of overexpression of adipose Dgat1 differ from one mouse strain to another is consistent with a general scheme of oligo- or polygenic determination of overall metabolic regulation. Discrepancies in phenotypic outcome have also been reported for several other genes in different genetic backgrounds (28–32).

Recent studies have indicated that adipokines can play important roles in insulin resistance. We did not detect any difference in adipokine production between aP2-Dgat1/FVB mice and wild-type littermates. Plasma levels of adiponectin (Fig. 3D), resistin, interleukin-6, and tumor necorsis factor-α were all unchanged (data not shown). It is possible that the gross plasma levels of certain adipokines may not reflect true functional changes; adiponectin is a clear example because of specific signaling properties associated with its various oligomerization states (33,34). In addition, some as yet undiscovered adipokine(s) may actually play a role in mediating the phenotype we have observed in the aP2-Dgat1 FVB mice. In fact, in mice with targeted GLUT4 knockout in adipose tissue, systemic insulin resistance was demonstrated in the absence of measurable changes in known adipokines or lipid mediators (35). Thus, the existence of some novel adipokine(s) is conceivable.

With the caveat that we have not eliminated the possibility that an adipose-secreted adipokine is playing a role, the present study in aP2-Dgat1/FVB mice strongly support a lipotoxicity model in which increased circulating FFAs are causally linked to hepatic steatosis and insulin resistance (36–40). Moreover, our data stress that metabolic impairment is not necessarily correlated with fat mass expansion (1–4). Rather, it results from a relatively low uptake and storage capacity of adipose tissue per se. When lipid availability exceeds the uptake and storage capacity of adipose tissue, fat will be deposited in nonadipose tissues, causing lipotoxicity and insulin resistance (41). This is well demonstrated by our aP2-Dgat1/FVB mouse, which is characterized by increased lipolysis, lack of fat mass expansion, increased circulating FFAs, and both insulin and leptin resistance.

This work was supported by grants from the National Institutes of Health (DK60530 to Y.-H.Y. and HL55638 to H.N.G.)

We thank Dr. Ira J. Goldberg for critical review of the manuscript.