Increased Very Low Density Lipoprotein Secretion and Gonadal Fat Mass in Mice Overexpressing Liver DGAT1*

Acyl-CoA:diacylglycerol acyltransferases (DGATs) catalyze the last step in triglyceride (TG) synthesis. The genes for two DGAT enzymes, DGAT1 and DGAT2, have been identified. To examine the roles of liver DGAT1 and DGAT2 in TG synthesis and very low density lipoprotein (VLDL) secretion, liver DGAT1- and DGAT2-overexpressing mice were created by adenovirus-mediated gene transfection. DGAT1-overexpressing mice had markedly increased DGAT activity in the presence of the permeabilizing agent alamethicin. This suggests that DGAT1 possesses latent DGAT activity on the lumen of the endoplasmic reticulum. DGAT1-overexpressing mice showed increased VLDL secretion, resulting in increased gonadal (epididymal or parametrial) fat mass but not subcutaneous fat mass. The VLDL-mediated increase in gonadal fat mass might be due to the 4-fold greater expression of the VLDL receptor protein in gonadal fat than in subcutaneous fat. DGAT2-overexpressing mice had increased liver TG content, but VLDL secretion was not affected. These results indicate that DGAT1 but not DGAT2 has a role in VLDL synthesis and that increased plasma VLDL concentrations may promote obesity, whereas increased DGAT2 activity has a role in steatosis.

Triglyceride (TG) 1 is the major energy storage form and is synthesized primarily in three tissues: liver, adipose, and small intestine. In the liver, synthesized TG is either stored in cytoplasmic droplets or secreted as very low density lipoprotein (VLDL) particles. Acyl-CoA:diacylglycerol acyltransferase (DGAT) is a membrane-bound enzyme that catalyzes the last step in the synthesis of TG. Classified by detergent sensitivity, two types of DGATs in microsomes have been proposed: the overt type (on the cytosol) catalyzes the synthesis of TG destined for cytoplasmic droplets, and latent type (on the lumen of the endoplasmic reticulum) catalyzes the synthesis of TG for VLDL formation (1). Because it has been well established that cytosolic droplet TG cannot be incorporated en bloc into VLDL (2), the relative activities of these two functions of DGAT may have a significant impact on the level of triglyceridemia as well as on the development of steatosis.
Regarding the molecular aspects of DGAT, the cDNAs of DGATs, viz. DGAT1 and DGAT2, have been recently cloned and sequenced (3,4). DGAT1 and DGAT2 are unrelated proteins that exhibit DGAT activity. DGAT1 is expressed ubiquitously, with the highest expression levels in the small intestine (3), and the phenotypes of DGAT1-null mice have been extensively examined (5)(6)(7)(8)(9). DGAT1-null mice are viable, can still synthesize TG, and have normal 4-h fasted plasma TG levels (5). These mice have reduced adiposity and are resistant to diet-induced obesity through a mechanism that involves increased energy expenditure (5). The increased energy expenditure is due partly to the increased peripheral leptin sensitivity in DGAT1-deficient mice (6). DGAT2 is expressed ubiquitously, with high expression levels in the liver and white adipose tissue (WAT) (4). In contrast to DGAT1-deficient mice, DGAT2deficient mice are lipopenic and die soon after birth from profound reductions in substrates for energy metabolism and from impaired permeability barrier function in the skin (10). It is evident that DGAT1 and DGAT2 have distinct roles in the whole body, but the roles of DGAT1 and DGAT2 in tissues such as liver, adipose, and small intestine that synthesize TG have not been well studied. In this work, to elucidate the roles of DGAT1 and DGAT2 in the liver, especially in overt and latent DGAT activities, we created liver DGAT1-and DGAT2-overexpressing mice by adenovirus-mediated gene transfection, and we investigated their phenotypes.

EXPERIMENTAL PROCEDURES
Preparation of Recombinant Adenoviruses-The full-length mouse Dgat1 (GenBank TM accession number NM_010046) and Dgat2 (Gen-Bank TM accession number AF384160) coding sequences were amplified by PCR. The primers used for amplifying DGAT1 and DGAT2 cDNAs were tailed with either an XbaI site (forward) or a KpnI site (reverse). PCR products were subcloned into the XbaI/KpnI site of the pShuttle vector provided in the BD Adeno-X expression system (BD Biosciences). All sequences of DGATs were verified by DNA sequencing. An I-CeuI/ PI-SceI restriction fragment from pShuttle containing the cytomegalovirus-IE promoter/enhancer 5Ј to the DGAT cDNAs insert and the polyadenylation signal was ligated into an adenoviral DNA backbone that was also restricted with I-CeuI and PI-SecI. Following amplification and purification of recombinant viral DNA from bacteria, recombinant viral DNA was further amplified by transfecting PacI-linearized recombinant viral DNA into human embryonic kidney 293 cells via the use of Lipofectamine reagent (Invitrogen). The adenoviruses containing DGAT1 cDNA (Ad-DGAT1) and DGAT2 cDNA (Ad-DGAT2) were purified for injection into mice using the BD Adeno-X purification kit (BD Biosciences). BD Adeno-X enhanced green fluorescent protein (Ad-GFP; BD Biosciences) was used as a control.
Each recombinant adenovirus (Ad-DGAT1, Ad-DGAT2, or Ad-GFP) was injected intravenously in a single dose (2 ϫ 10 9 plaque-forming units in 200 l) into a non-fasted mouse. The C57BL/6J mice (8 weeks of age) used in this experiment were purchased from The Jackson Laboratory (Bar Harbor, ME). The animals were fed standard rodent chow (CE2, CLEA, Tokyo). Animals were exposed to 12-h light/12-h dark cycles and maintained at a constant temperature of 22°C.
Northern Blotting-The cDNA clones containing the coding sequences of mouse DGAT1 and DGAT2 were obtained by PCR as described above. cDNAs encoding the rat VLDL receptor (VLDLR) and mouse perilipin were kindly provided by Dr. T. Yamamoto (Tohoku University Gene Research Center) and Dr. C. Londos (National Institutes of Health), respectively. RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. A portion of the RNA (20 g/lane) was denatured with glyoxal and dimethyl sulfoxide and analyzed by electrophoresis on 1% agarose gels. After transfer to nylon membranes (PerkinElmer Life Sciences) and UV cross-linking, RNA blots were stained with methylene blue to locate 28 S and 18 S rRNAs (11). The stained 18 S rRNA is shown in Figs. 1 and 7 to indicate that the amount of loaded RNAs was similar in each group. cDNAs were labeled with [␣-32 P]dCTP (Amersham Biosciences, Buckinghamshire, United Kingdom) using a random primer labeling kit (Rediprime II DNA labeling system, Amersham Biosciences). The amounts of mRNAs were quantitated with a BAS-2000 imaging analyzer (Fujifilm, Tokyo).
Liver Homogenization and Permeabilization of Microsomes for DGAT Assay-To examine the effects of homogenization methods on the recovery and intactness of microsomes, liver tissues (0.2 g) were homogenized in three different types of tissue grinders: Dounce-type (Kimble/ Kontes, Vineland, NJ), type B pestle, 30 strokes; motor-driven Pottertype (Iuchi, Osaka, Japan), 200, 400, 600, 1000, 2000, and 4000 rpm for 3 ϫ 10 s; and Polytron (PT10/35, Kinematica AG, Lucerne, Switzerland), setting 4 for 3 ϫ 10 s. All homogenization steps were carried out in 2 ml of ice-cold medium containing 300 mM sucrose, 1 mM EGTA, and 5 mM Tris-HCl (pH 7.4), followed by centrifugation at 10,000 ϫ g for 15 min at 4°C. The resultant supernatant was recentrifuged at 100,000 ϫ g for 70 min at 4°C to obtain the microsomes. The microsomes were resuspended to a final volume of 1 ml and separated into aliquots, which were snap-frozen in liquid N 2 and stored at Ϫ80°C. The recovery of microsomes was compared by measuring the total mannose-6-phosphatase activity in microsomal fractions. The intactness of microsomes was estimated by measuring the mannose-6-phosphatase activities in microsomal fractions in the presence and absence of permeabilizing agents. Potter-type homogenization at 200 or 400 rpm for 3 ϫ 10 s, a relatively mild method, resulted in a poor recovery of total microsomes (32-48% relative to those obtained by Polytron homogenizer) but a good recovery of intact microsomes (81-84%), whereas Polytron homogenization, a powerful method, resulted in a relatively good recovery of total microsomes but a poor recovery of intact microsomes (54%). Because the intactness of microsomes was more important than the recovery of microsomes in measuring overt and latent DGAT activities of microsomes, we chose Potter-type homogenization at 400 rpm for 3 ϫ 10 s in the following experiment.
To examine the effects of permeabilizing agents on DGAT activity, a portion of the microsomes (2 mg/ml protein) was permeabilized by preincubation on ice for 30 min with either the channel-forming peptide alamethicin (final concentration of 0.06 mg/ml in 0.38% ethanol; Sigma) or deoxycholate (final concentration of 2 mg/ml; Sigma) (1,12,13). The permeabilizing efficiency estimated by mannose-6-phosphatase activity was not different between these two agents, but DGAT activity in the presence of deoxycholate was 76% lower than in the presence of alamethicin (data not shown). It was shown previously that taurocholate inhibits DGAT activity by 30%, but alamethicin does not (1). Thus, in the study, we used alamethicin as a permeabilizing agent of microsomes.
Assay of DGAT Activity-DGAT activity was determined by measuring the incorporation of the [ 14 C]oleoyl moiety into trioleoylglycerol with [ 14 C]oleoyl-CoA (acyl donor) and sn-1,2-dioleoylglycerol (acyl acceptor) as described previously (14,15). The acyl acceptors were introduced into the reaction mixture by liposomes prepared with phosphatidylcholine (molar ratio of acyl acceptor to phosphatidylcholine of 1:5). The reaction mixture (200 l) contained 100 mM Tris-HCl (pH 7.4), 10 mM MgCl 2 , 1.25 mg/ml fatty acid-free bovine serum albumin (Serologicals Corp., Kankakee, IL), 200 mM sucrose, 25 M [ 14 C]oleoyl-CoA (PerkinElmer Life Sciences), 200 M acyl acceptors, and 10 g of either permeabilized or non-permeabilized liver microsomal protein. After a 10-min incubation at 37°C, lipids were extracted with 2:1(v/v) chloroform/methanol. After centrifugation to remove debris, aliquots of the organic phase-containing lipids were dried and separated on Silica Gel 60 TLC plates (Merck, Darmstadt, Germany) with 80:20:1 (v/v/v) hexane/ethyl ether/acetic acid. Individual lipid moieties were identified by standards with exposure to I 2 vapor. The TLC plates were exposed to an Fujifilm imaging plate to assess the formation of 14 C-labeled lipid products. Imaging signals were visualized and quantitated with a BAS-2000 imaging analyzer.
To estimate overt and latent DGAT activities, mannose-6-phosphatase activity, a marker of latent enzyme in microsomes, was used for correction of membrane damage and inside-out sealing of microsome in overt DGAT activity (1). Thus, the following equations were used to calculate overt and latent DGAT activities: , where D 0 and D t represent DGAT activity before and after treatment with alamethicin, respectively, and M 0 and M t represent mannose-6-phosphatase activity before and after treatment with alamethicin, respectively (1,16).
Electron Microscopy-Diced mouse liver samples were fixed with 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.3) for 24 h. After washing with phosphate buffer, glutaraldehyde-fixed blocks were placed in 2% osmium tetroxide in phosphate buffer for 2 h at 4°C, dehydrated, infiltrated with Epon 812, and polymerized for 3 days. Sections were cut out with Ultratome III (LKB Bromma, Stockholm, Sweden) and stained with saturated uranyl acetate and lead acetate. Thin sections were examined with a Hitachi H-300 electron microscope.
Serum Lipoprotein Analysis-Serum obtained from the retro-orbital plexus of an individual mouse was subjected to gel filtration HPLC on two tandemly connected TSK-Gel Lipopropak XL columns (300 ϫ 7.8 mm; Tosoh Corp., Tokyo) as described (17). TG and cholesterol were measured simultaneously using an on-line dual detection system. The particle size of lipoproteins was estimated by the elution time of the column and appropriate size markers.
Liver and Microsomal Lipid Analysis-Lipids in the liver were extracted quantitatively with ice-cold 2:1 (v/v) chloroform/methanol by the method of Folch et al. (18). Total cholesterol and TG concentrations in the liver homogenates and microsomal fractions were measured by enzymatic colorimetric methods using the cholesterol and TG E tests (Wako Pure Chemicals, Osaka), respectively.
Measurement of in Vivo VLDL Secretion Rates-Hepatic production of VLDL TG was measured in Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2injected mice. On day 12 after adenovirus injection, Triton WR-1339 (Sigma) was intravenously administered after a 4-h fast (19). Mice were bled prior to injection and at 1, 2, 3, and 4 h post-injection. Increases in serum TG concentrations over 4 h in Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2-injected mice after intravenous injection of Triton WR-1339 were measured by enzymatic colorimetric methods using the TG E test.
Immunoblotting-Pooled membrane fractions of either isolated epididymal or subcutaneous adipocytes from three mice of each group were prepared by collagenase digestion and differential centrifugation as described by Kono et al. (20). Ten g of membrane protein obtained by centrifugation at 175,000 ϫ g for 60 min at 4°C was applied to 12.5% SDS-polyacrylamide gel and transferred to Clear Blot Membrane-P (Atto Corp., Tokyo). Immunoblot analysis was performed using the enhanced chemiluminescence Western blotting detection system kit (Amersham Biosciences). Membrane sheets were first incubated with antibody against the VLDLR (sc-18824, Santa Cruz Biotechnology, Inc.) for 1 h at 22°C, washed three times, and incubated with horseradish peroxidase-conjugated anti-mouse IgG according to the protocol recommended by the manufacturer. The amount of VLDLR was quantitated using NIH Image Version 1.6.3.
Statistical Methods-Statistical comparisons of groups were made by one-way analysis of variance, and the VLDL secretion curve in Fig. 4 was compared by repeated measure analysis (StatView Version 5.0, Abacus Concepts, Inc., Berkeley, CA). When statistically significant, each group was compared with the others by Fisher's protected least significant difference test.

RESULTS
Overexpression of DGAT1 or DGAT2 in Mouse Liver-To gain insight into the physiological roles of DGAT1 and DGAT2 in the liver, either DGAT1 or DGAT2 was overexpressed in the liver by adenovirus-mediated gene transfection. We injected 2 ϫ 10 9 plaque-forming units of recombinant virus containing DGAT1 cDNA, DGAT2 cDNA, or GFP cDNA (control) into mice. This dose has been found to cause expression of the foreign gene in the majority of hepatocytes (21). On day 12 after administration of the recombinant viruses, the expression lev-els of DGAT1 and DGAT2 were measured by Northern blotting. Because the molecular sizes of DGAT1 mRNAs derived from the endogenous Dgat1 gene and the Ad-DGAT1 construct were almost the same, they could not be distinguished by Northern blotting, whereas because the molecular size of DGAT2 mRNA derived from the Ad-DGAT2 construct was smaller than that of DGAT2 mRNA derived from the endogenous Dgat2 gene, they could be distinguished (Fig. 1).
The DGAT1 mRNA levels in Ad-DGAT1-injected mice were increased by 19-fold in the liver and by 1.7-fold in the gastrocnemius compared with endogenous DGAT1 mRNA levels in control Ad-GFP-injected mice, but no changes were seen in brown adipose tissue (BAT) and epididymal WAT. DGAT2 expression in Ad-DGAT1-injected mice was not affected in all of these tissues. The DGAT2 mRNA levels in Ad-DGAT2-injected mice were increased by 4-fold in the liver compared with endogenous DGAT2 mRNA levels in control Ad-GFP-injected mice, but were not altered in BAT, WAT, and the gastrocnemius. The DGAT1 expression levels in Ad-DGAT2-injected mice were not affected in any of these tissues. Because DGAT1 expression in the gastrocnemius was increased slightly in both Ad-DGAT1-injected (significantly) and Ad-DGAT2-injected (insignificantly) mice, an increase in gastrocnemius DGAT1 expression might be due to metabolic changes in the whole body, secondary to liver DGAT overexpression. Because, as estimated by Northern blotting, the expression level of DGAT1 in normal liver was very low compared with that of DGAT2, when expressed as the -fold increase relative to endogenous DGAT1 mRNA levels, the increased DGAT1 mRNA levels in Ad-DGAT1-injected mice were much higher than the increased DGAT2 mRNA levels in Ad-DGAT2-injected mice. Plasma aspartate aminotransferase and alanine aminotransferase levels were not elevated after gene transfection, and liver damage did not occur (data not shown).
DGAT1 Possesses Latent DGAT Activity-To examine whether overexpressed liver DGAT1 and DGAT2 are functional, the overt and latent DGAT activities in liver microsomes from Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2-injected mice were estimated (Fig. 2). DGAT activity was assessed by the formation of 14 C-labeled TG from [ 14 C]oleoyl-CoA ( Fig. 2A). Overt DGAT activity in microsomes is defined as enzyme activity on the cytosol involved in the synthesis of cytosolic droplet TG, whereas latent DGAT activity is defined as enzyme activity on the endoplasmic reticulum (ER) lumen involved in VLDL synthesis (1). Overt and latent DGAT activities were estimated by measuring DGAT and mannose-6-phosphatase activities in the absence and presence of alamethicin as described under "Experimental Procedures." DGAT activity in the absence of alamethicin was significantly increased in both DGAT1-and DGAT2-overexpressing mice compared with control Ad-GFP-injected mice, whereas DGAT activity in the presence of alamethicin (total DGAT activity in microsomes) was markedly increased by 3.2-fold in DGAT1-overexpressing mice, but did not differ in DGAT2-overexpressing mice (Fig. 2B). When latent and overt DGAT activities were estimated by measuring the mannose-6-phosphatase activity (22), latent DGAT activity was increased by 4.2-fold in DGAT1-overexpressing mice, and overt DGAT activity was increased by 2.5-fold in DGAT2-overexpressing mice (Fig. 2C).
Because DGAT protein concentrations in microsomal fractions were not measured because of the lack of appropriate antibodies to DGAT1 and DGAT2, comparison of DGAT activity on a protein basis was not made between DGAT1-and DGAT2-overexpressing mice. However, if we assume that alamethicin did not affect the DGAT1 and DGAT2 molecules, this indicates that the latent/overt DGAT activity was much larger in DGAT1-overexpressing mice than in DGAT2-overexpressing mice; the latent/overt DGAT activity in Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2-injected mice was 2.4, 14.1, and 0.35, respectively. Thus, these data indicate that DGAT1 but not DGAT2 possesses marked latent DGAT activity, whereas DGAT2 might possess overt DGAT activity. Total RNA was isolated from the liver, BAT, epididymal WAT, and the gastrocnemius of non-fasted Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2-injected male mice on day 12 after injection of each adenovirus construct, and 20 g of RNA was subjected to electrophoresis, transferred to a nylon membrane, and hybridized with the indicated 32 P-labeled cDNA probes. Each lane represents a sample from an individual mouse. The molecular sizes of DGAT1, endogenous DGAT2, and Ad-derived DGAT2 mRNAs are 1.8, 2.5, and 1.7 kb, respectively. 18 S rRNA was used as a loading control. Quantification of DGAT1 and DGAT2 mRNAs is expressed relative to Ad-GFP-injected mice. **, p Ͻ 0.01 versus Ad-GFP-injected mice; ***, p Ͻ 0.001 versus Ad-GFP-injected mice.
ity was increased in DGAT1-overexpressing mice, we expected that blood VLDL concentrations might be increased in DGAT1overexpressing mice. Thus, we measured serum lipoprotein profiles in DGAT1-and DGAT2-overexpressing mice by HPLC (Fig. 3). As expected, compared with the control Ad-GFP-injected mice, VLDL TG concentrations were increased by 1.6fold in DGAT1-overexpressing mice, but were not changed in DGAT2-overexpressing mice (Fig. 3A and Table I). High den-sity lipoprotein (HDL) TG concentrations were very low in each group of mice. VLDL cholesterol concentrations were also increased in DGAT1-overexpressing mice ( Fig. 3B and Table I). However, low density lipoprotein (LDL) and HDL cholesterol concentrations did not differ among the different groups of mice. The particle size of lipoproteins was estimated by the elution time of TSK-Gel Lipopropak XL columns. VLDL particle size was significantly enlarged in DGAT1-overexpressing 1 and 4), Ad-DGAT1-injected (lanes 2 and 5), and Ad-DGAT2-injected (lanes 3 and 6) male mouse liver microsomes prepared using a Potter-type homogenizer in the absence and presence of alamethicin. Each lane represents a sample from an individual mouse. Mice were killed on day 12 after adenovirus administration. The amount of TG corresponds to DGAT activity. CE, cholesteryl ester; FFA, free fatty acid. Additional bands between TG and FFA and between FFA and oleoyl-CoA were observed when both microsomes and alamethicin were added to the incubation mixture, but their molecules have not been identified. B, quantification of DGAT activities in A. C, overt and latent DGAT activities in liver microsomes isolated from Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2-injected mice estimated by the activity of mannose-6-phosphatase, a marker of latent enzyme in microsomes, as described under "Experimental Procedures." The mannose-6-phosphatase activities in microsomal preparations before and after alamethicin treatment were found to be 23.8 and 167 nmol of P i formed per min/mg of protein, respectively. Data represent means Ϯ S.E. (n ϭ 4). **, p Ͻ 0.01 versus Ad-GFP-injected mice; ***, p Ͻ 0.001 versus Ad-GFP-injected mice.

TABLE I Serum lipoprotein profiles in Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2-injected mice
Serum lipoproteins in male mice were assayed on day 12 after adenovirus injection as described in the legend to Fig. 3 3. Increased serum VLDL concentration and particle size in DGAT1-overexpressing mice. Serum from each group of mice on day 12 after injection of Ad-GFP, Ad-DGAT1, or Ad-DGAT2 was subjected to gel filtration HPLC, and the TG and cholesterol content of each fraction was measured as described under "Experimental Procedures." Representative data from individual mice are shown in A (TG) and B (cholesterol). The mean of VLDL particle size is shown in C. Data represent means Ϯ S.E. (n ϭ 10 -12). IDL, intermediate density lipoprotein. **, p Ͻ 0.01 versus Ad-GFPinjected mice. mice (Fig. 3C), but there were no changes in HDL particle size among the different groups of mice (data not shown).
Increased VLDL Secretion in DGAT1-overexpressing Mice-The increase in the VLDL concentration in DGAT1-overexpressing mice may be due to an increase in VLDL secretion or/and inhibition of VLDL clearance. To determine the cause, the rate of hepatic VLDL secretion was estimated by monitoring serum TG concentrations in the presence of Triton WR-1339 (Fig. 4), which blocks lipolysis of VLDL in peripheral tissues. The TG concentration significantly increased in DGAT1-overexpressing mice compared with control mice following Triton WR-1339 administration (by repeated analysis of variance, p Ͻ 0.01). An increase in the TG concentration in DGAT2-overexpressing mice was not observed. Thus, the increase in the blood TG concentration in DGAT1-overexpressing mice was due, at least in part, to increased VLDL secretion.
Electron Microscopy Study of the ER in Livers from DGAToverexpressing Mice-Next, to examine whether the increased rate of VLDL secretion observed in DGAT1-overexpressing mice is due to increased VLDL synthesis in the liver ER, an electron microscopy study was conducted (Fig. 5). The VLDL secretion pathway in the liver has been elucidated (23)(24)(25). TG is synthesized in the smooth ER, and this lipid particle then moves to the smooth surface ends of rough ER cisternae. At the junction between the smooth ER and the rough ER, apolipoprotein B is bound to the lipid particle to form VLDL. Alternatively, in the rough ER, TG and apolipoprotein B are synthesized and assembled into VLDL with the microsomal TG transfer protein MTP (26). Electron microscopy examination of the liver revealed that the rough ER from DGAT1-overexpressing mice was markedly dilated and contained small particles, possibly synthesized TG particles in the lumen of the rough ER, but these changes were not observed in control and DGAT2overexpressing mice. The diameter of these small particles was 46 Ϯ 2 nm (means Ϯ S.E., n ϭ 21), similar to the size of VLDL in blood (Fig. 3C). Indeed, the TG concentration in the microsomal fractions from DGAT1-overexpressing mice was 2.1-fold larger than in control mice (microsomal TG concentrations in Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2-injected mice of 95 Ϯ 15, 206 Ϯ 28, and 117 Ϯ 21 g/mg of protein, respectively (n ϭ 4); p Ͻ 0.01). Even when expressed as wet liver weight, the microsomal TG amount in DGAT1-overexpressing mice was 2.2-fold larger than in control mice (microsomal TG amounts in Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2-injected mice of 3.6 Ϯ 0.5, 7.8 Ϯ 1.1, and 5.3 Ϯ 1.2 mg/g of liver, respectively (n ϭ 4); p Ͻ 0.05). This finding supports the hypothesis that increased latent DGAT activity, which occurs in the lumen of the rough ER, leads to increased VLDL secretion in DGAT1-overexpressing mice.
DGAT2 Overexpression Increases Liver TG Concentrations-As hepatic overexpression of DGAT1 and DGAT2 increased DGAT activity (Fig. 2), we examined whether the liver TG concentration was increased in these DGAT-overexpressing mice on day 12 after administration of the recombinant viruses (Fig. 6A). Accumulation of TG in DGAT1-overexpressing mice was not expected because of lower overt DGAT activity and/or an increase in latent DGAT activity, which promotes the excretion of TG from the liver to the blood circulation as VLDL. However, either DGAT1-or DGAT2-overexpressing mice increased liver TG concentrations; DGAT1-and DGAT2-overexpressing mice showed 1.9-and 3.1-fold increases in liver TG concentrations, respectively, compared with control mice. Similar increases in TG concentrations were noted on day 6 after administration of the recombinant viruses (data not shown). Although an ϳ2-fold increase in the TG concentration in microsomes was observed in DGAT1-overexpressing mice, the contribution of the TG amount in microsomes to that in the total liver was Ͻ10% (estimated by the average amount of TG in microsomal fractions and homogenates); the 1.9-fold increase in the TG amount in the total liver in DGAT1-overexpressing mice was not due to the TG increase in the microsomes. Thus, both DGAT1 and DGAT2 contributed to cytosolic triglyceride levels. The larger increase in TG accumulation in DGAT2-overexpressing mice indicated that DGAT2 possessed potent activity to synthesize TG in the liver cytosol. The liver cholesterol concentration did not differ among the three groups of mice (Fig. 6B).
DGAT1-overexpressing Mice Have Increased Epididymal WAT Weight-To examine the effects of increased VLDL secretion in DGAT1-overexpressing mice on the whole body, the body and tissue weights of Ad-GFP-, Ad-DGAT1-, and Ad-DGAT2-injected mice were measured on day 12 after administration of the recombinant viruses (Table II). There were no changes in total body weights among these three groups of mice. However, epididymal WAT weight was significantly increased by 1.2-fold in DGAT1-overexpressing mice, but abdominal subcutaneous WAT weight was not altered. No increase in epididymal WAT weight was observed in DGAT2-overexpressing mice. In female mice, DGAT1 overexpression also caused a 1.5-fold increase in parametrial WAT weight (p Ͻ 0.01), but did not alter subcutaneous WAT weight (data not shown). These results suggest that TG transfer from the liver to WAT might lead to increased gonadal WAT weight. Because the observation periods were short (12 days), increases in WAT weight might not cause apparent obesity. Although its mechanism was not clear, liver weight was slightly decreased in DGAT1-overexpressing mice irrespective of the increased lipid accumulation (Fig. 6A). However, we speculated that increased DGAT1 expression over a longer time period might lead to increased liver weight with a gradual accumulation of TG. The glucose and insulin tolerance curves on day 12 were not significantly different among these groups (data not shown).
Gonadal WAT Expresses a Large Amount of VLDLR-To elucidate the mechanism for the higher sensitivity of VLDL in gonadal WAT, the expression levels of VLDLR in subcutaneous WAT and epididymal WAT were examined in normal mice (Fig.  7). The expression level of VLDLR mRNA was 6-fold larger in epididymal WAT than in abdominal subcutaneous WAT (Fig. 7,  A and B). Expression of perilipin (a fat droplet marker) was also increased by 2-fold in epididymal WAT, but its increase was much lower than the VLDLR mRNA increase. In female mice, a 4-fold increase in VLDLR mRNA was also observed in parametrial WAT (p Ͻ 0.01) (data not shown). Immunoblot analysis of the VLDLR also confirmed that the VLDLR protein was increased by ϳ4-fold in adipocytes from epididymal WAT compared with those from abdominal subcutaneous WAT (Fig.  7C). Furthermore, overexpression of liver DGAT1 or DGAT2 did not regulate the expression levels of VLDLR mRNA in epididymal WAT (Ad-GFP-injected mice, 100 Ϯ 16%; Ad-DGAT1-injected mice, 116 Ϯ 7%; and Ad-DGAT2-injected mice, 119 Ϯ 13%; n ϭ 3) and subcutaneous WAT (Ad-GFP-injected mice, 100 Ϯ 6%; Ad-DGAT1-injected mice, 81 Ϯ 5%; and Ad-DGAT2-injected mice, 89 Ϯ 5%; n ϭ 3). These data indicate that the preferential increase in gonadal WAT mass observed in DGAT1-overexpressing mice might be due to high VLDLR expression levels in gonadal WAT and increased VLDL secretion.

DISCUSSION
In this study, to examine the roles of liver DGAT1 and DGAT2, we investigated the effects of overexpression of liver DGAT1 and DGAT2 by adenovirus-mediated gene transfection on DGAT activity, TG synthesis, and VLDL secretion. DGAT1overexpressing mice had increased latent DGAT activity and a dilated ER, whereas these changes were not observed in DGAT2-overexpressing mice (Figs. 2 and 5). As expected from the increased latent DGAT activity in DGAT1-overexpressing mice, VLDL secretion and particle size were increased (Figs. 3 and 4), resulting in increased gonadal fat mass expressing a large amount of VLDLR ( Fig. 7 and Table II).
Our in vivo evidence that DGAT1 is located in the lumen of the ER and promotes VLDL secretion is in a good agreement with recent in vitro studies. Overexpression of human DGAT1 in McArdle rat hepatoma cells (RH7777) results in increased TG-rich VLDL secretion (27). When DGAT1 and DGAT2 are overexpressed in RH7777 cells, small lipid droplets around the cell periphery are observed in DGAT1-expressing cells, whereas numerous large cytosolic lipid droplets are observed in DGAT2-expressing cells (Fig. 1 from Ref. 10).
Judging from the abundance of DGAT2 mRNA, it was speculated that DGAT2 rather than DGAT1 is the major DGAT in the liver, although its protein levels and subcellular localization have not been determined (28,29). In a previous study, DGAT1 and DGAT2 activities were distinguished by sensitivity to MgCl 2 ; DGAT2 but not DGAT1 activity is inhibited at higher concentrations of MgCl 2 (100 mM) (4). The recent observation that niacin inhibits both MgCl 2 (high concentration)-sensitive DGAT2 activity and overt DGAT activity in HepG2 cells also suggests that DGAT2 has potent overt DGAT activity (13). These observations confirm that DGATs in the liver have a distinct role; DGAT2 plays a major role in cytosolic lipid accumulation, whereas DGAT1 plays a role in VLDL assembly. However, contrary to this hypothesis, DGAT1-null mice fail to decrease 4-h fasted serum TG concentrations (5). We speculated that a 4-h fasting might not be long enough to eliminate the contribution of ingested dietary fat to serum TG levels, and the actual VLDL TG levels might be decreased in a 12-h fasted state. It is also conceivable that ablation of DGAT1 in the whole body might up-regulate adaptive mechanism(s) that enhance VLDL secretion. Regulation of VLDL assembly and secretion is a complex process that requires a coordinated function of many enzymes such as acyl-CoA:cholesterol acyltransferases and MTP and those related to apolipoprotein B synthesis and transport of acylcarnitine from the cytosol to the ER, including DGATs (30 -32). Thus, these other players might be involved in the adaptive mechanism(s).
Plasma TG levels are reduced by 70 -90% in DGAT2-null mice at birth relative to littermate controls (10). This result can be explained as follows. Plasma free fatty acids are markedly reduced by 70 -90% in DGAT2-null mice, with a marked reduction in fatty acid supply from fat mass (10). Because it is well known that plasma free fatty acids promote VLDL secretion (33), a decrease in plasma free fatty acids might lead to a decrease in fasting plasma TG levels in DGAT2-null mice.
The idea that a similar enzyme but in a different orientation of the ER plays a different role is not new. Acyl-CoA:cholesterol acyltransferases catalyze the formation of cholesteryl ester from cholesterol and fatty acyl-CoA (34), whereas DGATs catalyze the formation of TG from diacylglycerol and fatty acyl-CoA. Like DGATs, two genes that encode ACAT1 and ACAT2 have been identified with distinct roles (35)(36)(37). Although the membrane topology of ACAT2 is still controversial (38,39), it secretes more VLDL cholesterol compared with ACAT1 when overexpressed in RH7777 cells (27). These data suggest that DGAT1 and ACAT2 might coordinately participate in VLDL formation within the ER lumen.
The other important finding of this study is that VLDL increased gonadal fat mass weight. The common causes of increased VLDL synthesis include increased free fatty acid flux into hepatocytes, as seen in obesity, and increased hepatic de novo lipogenesis, as seen in excessive intake of alcohol or carbohydrate (40). VLDL transports fatty acids from the liver to adipose and other peripheral tissues. Because adipose tissues receive fat from VLDL and chylomicrons, an increase in blood VLDL concentrations might be one of the causes of obesity. However, it has not been elucidated whether the observed increased VLDL secretion is a cause or a result of obesity. Our results that gonadal adipose tissues expressed a large amount of VLDLR (Fig. 7) and that liver DGAT1-overexpressing mice showed enhanced VLDL secretion and increased gonadal adipose tissue weight, whereas DGAT2-overexpressing mice did not show enhanced VLDL secretion or increased fat mass weights (Table II), support the hypothesis that VLDL causes some types of obesity, viz. abdominal obesity. Ablation of the Srebp-1 (sterol regulatory element-einding protein-1) gene results in marked reductions in hepatic lipogenesis, but does not decrease VLDL secretion or the amount of WAT in hybrids between C57BL/6J and 129Sv/Ev mouse strains (41) and in  7. Increased VLDLR expression in epididymal WAT. A and B, total RNA was isolated from abdominal subcutaneous WAT and epididymal WAT, and 20 g of RNA was subjected to electrophoresis, transferred to a nylon membrane, and hybridized with the 32 P-labeled VLDLR and perilipin cDNA probes. Perilipin mRNA was measured on the same membrane sheet as a control. Each lane represents a sample from an individual mouse. The molecular sizes of VLDLR and perilipin mRNAs are 3.4 and 2.7 kb, respectively (A). Quantification of each mRNA is expressed relative to subcutaneous WAT (B). Data represent means Ϯ S.E. (n ϭ 3). **, p Ͻ 0.01 (significantly different); ***, p Ͻ 0.001 (significantly different). C, VLDLR proteins in pooled membrane fractions of isolated adipocytes from either abdominal subcutaneous WAT or epididymal WAT from three mice were measured by immunoblotting. The molecular size of the VLDLR protein is 135 kDa. ob/ob mice (42). Mice lacking the VLDLR exhibit normal lipid levels and a modest decrease in epididymal adipose tissue weights and are protected from diet-induced obesity (43,44). These knockout mouse studies and our DGAT study support the hypothesis that an increased plasma VLDL concentration is an important determinant of intra-abdominal obesity, possibly mediated by the VLDLR.
In conclusion, DGAT1 but not DGAT2 has a role in VLDL synthesis, and increased blood VLDL concentrations may promote obesity, whereas increased DGAT2 activity has a role in steatosis. In addition, the finding that gonadal fat expresses a large amount of VLDLR highlights the importance of the VLDLR in the etiology of intra-abdominal obesity. Further studies using liver-specific DGAT knockout mice are needed to examine whether a decrease in liver DGAT activity is safe and effective in preventing fatty liver and VLDL oversecretion.