Posttranscriptional control of the expression and function of diacylglycerol acyltransferase-1 in mouse adipocytes

without affecting triglyceride lipolysis rate, leading to >2-fold increase in intracellular triglyceride accumulation. No change in adipocyte morphology, or in the expression levels of LPL, PPAR- g , and aP2 was evident secondary to DGAT1-overexpression at different stages in 3T3-L1 differentiation. These data suggest that dysregulation of DGAT1 may play a role in the development of obesity, and manipulation of the steady-state level of DGAT1 protein may offer a potential means to treat or prevent obesity.


Summary
Acyl-CoA:diacylglycerol acyltransferase-1 (DGAT1) catalyzes the final step of triglyceride synthesis in mammalian cells. Data obtained from DGAT1-knockout mice have indicated that this enzyme plays an important role in energy homeostasis. We investigated the regulation of the expression and function of DGAT1 in mouse 3T3-L1 cell as a model for mammalian adipocytes. We demonstrated that the DGAT1 protein level increased by ~90-fold following differentiation of 3T3-L1 into mature adipocytes, a change which was accompanied by ~7-fold increase in DGAT1 mRNA. On the other hand, forced overexpression of DGAT1 mRNA by >20-fold via a recombinantadenovirus only resulted in ~2-fold increase in DGAT1 protein in mature adipocytes, and little increase in preadipocytes. These results indicated that gene expression of DGAT1 in adipocytes is subjected to rigorous posttranscriptional regulation, which is modulated significantly by the differentiation status of 3T3-L1 cells. Protein stability is not a significant factor in the control of DGAT1 expression. The steady-state levels of DGAT1 were unaffected by blockage of proteolytic pathways by ALLN. However, translational control was suggested by sequence analysis of the 5'-UTR of hDGAT1 mRNA. We found that the level of DGAT1 activity was predominantly a function of the steady-state level of DGAT1 protein. No significant functional changes were observed when the conserved tyrosine-phosphorylation site in hDGAT1 was mutated by a single base-pair

Introduction
Adiposity, or the amount of triglyceride (TG) stored in adipocytes, is fundamentally a net result of two opposing metabolic processes, i.e., lipogenesis and lipolysis. These are highly regulated processes influenced by various hormones, cytokines, nutrients and other factors (c.f. (1)) that may either serve as substrates or function as signaling molecules for one or more specific metabolic pathways. Acyl-CoA:diacylglycerol acyltransferase (DGAT; EC2.3.1.20) is thought to be a key enzyme in controlling the synthetic rate of TG in adipocytes (2). It catalyzes the last step in the de novo TG synthetic pathway employed by various types of cells, producing TG from its two substrates, diacylglycerol (DAG) and fatty acyl-CoA (for review c.f. (3,4)). Two DGAT genes have been identified. DGAT1 is a member of the ACAT (acyl-CoA cholesterol acyltransferase) gene family, and was cloned from human and mouse tissues through a homology search for ACAT-like sequences (5,6). More recently, DGAT2 was cloned from animals and fungi (7,8); DGAT2 bears no sequence homology with DGAT1.
The function of DGAT1 gene in TG metabolism has been assessed in knockout mice (9,10). Mice deficient in DGAT1 are viable and appear to have histologically normal adipose tissue (9). However, these mice have decreased adiposity, and less TG content in muscle and the liver (10). They also display a significant defect in TG synthesis in the mammary and sebaceous glands (9,11). Reexamination of DGAT1 deficient mice revealed that as much as half normal levels of DGAT activity were present in adipose tissue. This remaining DGAT activity was more readily displayed in an in vitro assay at lower magnesium concentrations, and was thought to be mediated by DGAT2 (7).
Interestingly, DGAT1 deficient mice showed increased insulin and leptin sensitivity (10), and they had significant resistance to diet-induced obesity due to increased energy expenditure (9). These experiments point to a significant role of DGAT1 in systemic energy homeostasis.
Despite much existing data concerning the biology of DGAT, how DGAT's expression and function are regulated in mammalian adipocytes remains poorly understood.
Furthermore, although insect cells have been used for exogenous expression of the mammalian DGAT1 gene, with resultant increases in intracellular TG synthesis (6,12), studies of the effects of overexpressing DGAT1 gene in mammalian adipocytes are lacking. In the present paper, we report our studies of the regulation of DGAT1 expression and function in mouse 3T3-L1 cells in relation to their differentiation status.
We demonstrated that DGAT1 expression is not only regulated transcriptionally as previously shown, but also strictly controlled posttranscriptionally. This posttranscriptional mechanism is heavily modulated by the differentiation status of 3T3-L1 cells. We further showed that DGAT1 activity and its effect on the rate of intracellular TG accumulation depend primarily on the steady-state levels of DGAT1 protein.
Mutation at the putative tyrosine-phosphorylation site in DGAT1 causes no significant changes in its activity or overall rates of TG accumulation. Finally, we showed that overexpression of DGAT1 in 3T3-L1 cells does not alter the course of adipocyte differentiation. florescence protein (GFP) was constructed similarly. Lysates were prepared from the infected monolayer 293 cells by freeze-and-thaw methods, and purification of viral particles was achieved using CsCl gradient centrifugation as previously described (13,14). Viral titers were determined by TCID 50 (15,16).

Experimental
Mutation of hDGAT1 at the conserved tyrosine phosphorylation site. The point mutation converting Tyr 316 to Phe, a structurally related amino acid residue, was made using a PCR-based strategy. A pair of PCR primers was made commercially: the forward 5' primer corresponding to nucleotide sequence 978-997 in hDGAT1 and the reverse 3' primer corresponding to nucleotide sequence 1179-1221 with a base substitution from A to T at position 1190. This substitution converts the codon TAC for Tyr to the codon TTC for Phe. The resultant PCR fragments was digested, in preparation for subsequent ligation, with Eco47III and HaeII, which are located near the 5' and 3' ends of the fragment, respectively. The full-length hDGAT1 in pcDNA3.1 was excised with Eco47III (which cuts hDGAT1 sequence at position 1069) and XbaI (which cuts at the polylinker region of the vector, 3' to the insert sequence). While the larger Eco47III/XbaI fragment (fragment A) from this digestion containing the vector sequence and part of the 5' DGAT sequence was saved for re-ligation, the smaller Eco47III/XbaI fragment was further digested with HaeII (position 1209) to produce a 3' HaeII/XbaI DGAT fragment (fragment B). Finally, a 3-piece re-ligation was carried out between fragments A and B and the Eco47/HaeII PCR fragment to generate pc(T316F). This plasmid, which contains a full-length hDGAT1 with phenylalanine substitution for tyrosine, was used to generate the recombinant adenovirus, Ad-T316F, similarly as described (above).
Cell culture, differentiation and adenoviral infection: 3T3-L1 preadipocytes were maintained in high glucose (25mM) DMEM with 10% bovine calf serum. Medium was changed every other day, and cells were always split and passed before they reached full confluence. To induce differentiation, confluent cells were cultured in maintenance medium for 48 hours before changing to an induction medium of high glucose DMEM containing 10% bovine fetal serum, 1 µM dexamethasone and 5 µg/ml insulin. Cells were routinely cultured in induction medium for 10 days, unless otherwise indicated, to obtain full differentiation. Again, medium was changed every other day. To infect cells with recombinant adenovirus, cells were first incubated with viruses (typically 30 pfu/cell) in serum-free DMEM for 2 hours. In most experiments, polylysine at concentration of 0.75 ug/ml was added during the incubation to facilitate viral transduction (17). An equal volume of DMEM medium containing 4% bovine calf serum (final concentration 2%) was then added, and the cells were incubated for another 12 hours before the medium was changed. Assays were carried out at least 48 hours after viral infection. RT-PCR: RNA was prepared from cultured cells with TriZol reagent using the protocol provided by the manufacturer (Life Technologies). Reverse transcription was carried out with a reverse transcriptase and random hexamer primers. One to two ul of the reaction mixture containing reverse transcripts was used for subsequent PCR reaction using Taq polymerase and sequence-specific oligonuclear primers. To amplify a DNA fragment of less than 500-bp, thirty-five cycles were typically used as follows: denaturing 94 o C/15 sec; annealing 55 o C/15 sec; polymerization 72 o C/30 sec. The relative RNA levels were determined by "quantitative RT-PCR". For this assay, primers for small PCR fragments of about 150 bp were designed. Amplification was pre-titrated within a linear range, using a limited number of cycles and 32 P-dCTP labeling for quantification (18,19). The probes used for hybridization were generated by RT-PCR labeling with 32 P-dCTP using RNAs isolated from mouse adipose tissue as initial templates. Specific oligonucleotide primers used in the PCR labeling were made of about 20 base-pairs based on gene bank database sequences for specific adipocyte differentiation marker proteins described in the text.

Oil-red-O staining for intracellular TG
Western blot: This analysis was performed according to standard methods (21). The primary polyclonal antibodies against DGAT were generated from rabbits using MAPconjugated synthetic peptides corresponding to the first 20 amino acid residues of mouse DGAT1. Specific protein bands were visualized using ECL immunodetection system (Amersham Life Science) after blotting with horse radish peroxidase-conjugated second antibodies against rabbit IgG. In some experiments, 131 I-labeled mouse anti-rabbit antibodies were used and specific protein bands were visualized after autoradiographic exposure. Lipids were re-dissolved in 20 µl hexane, and then analyzed by thin-layerchromatography (TLC) as described (22) except that plates were developed in hexanediethyl ether-glacial acetic acid 70:30:1 (v/v/v). Chromatographic bands containing TG were cut out after iodine staining, and quantitated by scintillation counting. Total TG mass: Cellular lipids were extracted using hexane/isopropanol (3:2) as described above. TG mass was determined enzymatically using colorimetric kits (Trig/GB from Roche Diagnostics). TG standards were used for quantification.

In vitro
Statistical analysis: Statistica V6.0 was used for all analyses. The significance of differences was determined using student's t test. A two-tailed p value <0.05 was considered to indicate statistical significance. An estimated 90-fold increase in DGAT1 protein level was observed (9-fold densitometric difference between lanes d and d' x 10-fold difference in protein loading).

Results
This marked disparity between the increases in DGAT1 protein and mRNA indicates that the expression of DGAT1 gene was regulated by an additional posttranscriptional mechanism during normal differentiation of 3T3-L1 cells.
To further characterize this posttranscriptional regulation, 3T3-L1 cells were directed to produce high levels of full-length DGAT1 mRNA from an exogenous human DGAT1 (hDGAT1) gene containing the entire 5' and 3' untranslated regions, and the resultant DGAT1 protein levels were followed. Forced high-level expression of hDGAT1 mRNA was achieved by using a recombinant adenovirus carrying a hDGAT1 expression cassette (Ad-hDGAT1) under the control of the cytomegalovirus promoter/enhancer (pCMV).
We first transduced fully-differentiated 3T3-L1 adipocytes (PID 10) with Ad-hDGAT1.  Figure 4, a slight, statistically insignificant decrease in the fractional turnover rate was observed in Ad-hDGAT1 transduced cells, as compared with that in Ad-GFP transduced cells. These results further substantiated our conclusion that the >2-fold increase in TG accumulation in hDGAT1transduced adipocytes ( Figure 3C) was due directly to an elevated TG synthesis rate; there was no significant change in TG lipolysis rate in these cells.
Mutation at the putative tyrosine-phosphorylation site in DGAT1 has no significant effect on DGAT1 activity and overall cellular TG accumulation. Previous biochemical studies suggested that DGAT activity is regulated by phosphorylation of the enzyme (25)(26)(27)(28). Since these data were all obtained before the cloning of the DGAT genes, it is unclear which DGAT was particularly modified in particular experiments. Lau and Rodriguez reported that DGAT isolated from rat adipocytes was bound and inactivated by an ATP-dependent tyrosine kinase; this process could be reversed by incubating the sample with a crude preparation containing protein phosphatase(s) (28). Both human and mouse DGAT1 contain a single conserved sequence for tyrosine-phosphorylation (5,6).
By contrast, the cloned DGAT2 contains several putative protein kinase C phosphorylation sites (7). To verify potential tyrosine-phosphorylation regulation of DGAT1, we made a single point mutation at the conserved tyrosine-phosphorylation site, converting tyrosine 316 to phenylalanine ( Figure 5A). Clearly, this mutant hDGAT1 would be unable to serve as a substrate for a tyrosine kinase.
The mutant hDGAT1 (T316F) was introduced into 3T3-L1 adipocytes using the same technique employing recombinant adenovirus. Viral transduction gave rise to a comparable high-degree expression of the mutant mRNA, T316F, compared with the wild type mRNA ( Figure 5B, compare lanes b with c). Similarly, the protein levels of T316F and the wild-type hDGAT1 were increased by 2-and 2.5-fold, respectively, over the control ( Figure 5C, Mb), suggesting that the T316F mutant was subjected to the same posttranscriptional control as the wild-type. Little DGAT1 protein was found in the cytosol in either instance ( Figure 5C, Cyt), suggesting that the mutation did not alter subcellular distribution of hDGAT1 protein either. Furthermore, microsomal membranes isolated from the T316F-transduced cells (T316F) exhibited a 1.7-fold increase in DGAT activity over the control (GFP); in the same set of experiments, the wild-type hDGAT1 caused a 2.1-fold increase in DGAT activity ( Figure 5D, left panel). These results suggest that the mutant hDGAT1 is functionally active. However the "nonphosphorylated" enzyme did not result in higher enzymatic activity more than what might be accounted for by the increased protein expression of the mutant hDATG1. This was in contrast to what might be expected if the "nonphosphorylated" form of this enzyme represents a more active DGAT1, a scenario suggested previously for "overall" DGAT activity being subjected to tyrosine-phosphorylation modification (28). Under the above assay condition (with 150 mM MgCl 2 ), DGAT2 activity is largely suppressed (7).
In order to assess if any compensatory changes in DGAT2 activity occur in DGAT1-20 by guest on March 24, 2020 http://www.jbc.org/ Downloaded from overexpressing cells, experiments were repeated at 10 mM MgCl 2 to optimally display DGAT2 activity (7). Under such a condition, a fuller range of DGAT activity (both DGAT1 and DGAT2) was measured. As shown in Figure 5D, right panel, higher DGAT activities (a difference by the factor of 2-2.5 between the two assay conditions) were observed. However, there was no significant difference in the relative increase among the three specimens. These results suggested that DGAT2 levels were preserved in all these cells, and there was no down-regulation of DGAT2 in cells transduced with Ad-hDGAT1 or Ad-T316F in response to the increased DGAT1 expression and function.
Finally, consistent with the increase in total DGAT activity in wild-type hDGAT1-and mutant T316F-expressing cells adipocytes, a 2.6-and 2.3-fold increase, respectively, in total intracellular TG content were obtained 5 days after viral transduction ( Figure 5E). bypasses DGAT and appears to operate in oilseed maturation may play a limited role in mammalian gastrointestinal system (4,29). Recently, another major TG biosynthetic pathway has been identified in yeast mediated by Lro1, a homolog of human lecithin cholesterol acyltransferase, which esterifies DAG using phosphatidylcholine as the acyl donor, bypassing DGAT (30). It is unclear if this type of TG biosynthesis plays any significant role in mammalian cells. With regard to the glycerol-phosphate pathway, two DGAT genes have been identified, and there may be several members in the DGAT2 family (7,8). DGAT2 activity does not appear to be up-regulated in DGAT1 deficient mice (7). We have no evidence that it is down-regulated in DGAT1 overexpressing adipocytes, since the difference in total DGAT activity resulting from overexpressing DGAT1 is preserved under the assay conditions optimal for DGAT2 (see text). An Unlike a previous report (28) suggesting that DGAT activity is significantly regulated by tyrosine-phosphorylation, our results showed that a site-specific mutation in the conserved tyrosine-phosphorylation site in hDGAT1 did not result in any enhanced DGAT activity, as compared with the wild-type hDGAT1. The "nonphosphorylated DGAT1" is an active form of the enzyme, however, since the mutant hDGAT1, T316F, displayed an appropriate increase in DGAT activity and intracellular TG accumulation in Ad-T316F transduced adipocytes, compatible with an increase in its protein level. Our results further showed that at the whole cell level, TG accumulation was not affected disproportionately by the presence of the overexpressed "constitutively nonphosphorylated" form of DGAT1. Thus, our data suggest strongly that tyrosinephosphorylation does not play a significant role in regulating DGAT1 activity. We cannot rule out the possibility, although unlikely, that the conversion of the tyrosine to the phenylalanine residue in our T361F construct caused a significant alteration in DGAT1 conformation. Such a conformational change may not abolish its enzymatic function, but cause a decrease in activity compared with the native "nonphosphorylated" molecules.
Another possibility is that a functional tyrosine-phosphorylation site is present in a yetunidentified DGAT gene. Phosphorylation modification of DGAT activity by cAMPdependent or Ca2+/calmodulin-dependent protein kinases was also reported (26,27). But these phosphorylations resulted in enhanced, rather than decreased, DGAT activity. Such modification is likely present in DGAT2 since the cloned DGAT2 gene contains several putative protein kinase C phosphorylation sites (7). Nevertheless, the phosphorylation of the serine/threonine residues needs to be verified at specific sites in DGAT2 sequence using site-specific mutation, an approach we have taken to determine the functionality of the tyrosine-phosphorylation site of DGAT1. Consistent with the lack of the tyrosinephosphorylation regulation in DGAT1 activity is the fact that the DGAT activity and the cellular TG incorporation rate correlate closely only with the steady-state protein level of DGAT1. This is true in both Ad-hDGAT1 and Ad-T316F transduced adipocytes. It is clear that the DGAT1 activity is determined primarily by the steady-state level of Posttranscriptional control of gene expression appears to be an important mechanism involved in many aspects of lipid metabolism in response to various physiological stimuli. Gene expression of several major apolipoproteins is controlled at the posttranscriptional level (34)(35)(36)(37)(38). The secretion rates of apolipoprotein B (apoB), for example, depend largely on the lipid availability in the ER, which permits continuous synthesis of this protein to be coupled with regulated lipoprotein secretion. In the absence of lipid ligands, either translation of apoB is retarded, or the newly synthesized polypeptides are quickly targeted for proteosomal degradation (c.f. (39,40)). HMG-CoA reductase, the rate limiting enzyme for cholesterol synthesis, is another example of posttranscriptional regulation (41)(42)(43)(44)(45)(46). In addition to transcriptional suppression by sterols, this enzyme is further regulated by mevalonate through a posttranscriptional mechanism, primarily by accelerating the degradation of this protein (41,44,45). We do not know at this time how posttranscriptional regulation of the expression of DGAT1 protein is achieved. But a need for a more rapid regulation (as opposed to transcriptional regulation) in response to metabolic stimuli (hormones or lipid intermediates) is conceivable. As demonstrated in enterocytes, DGAT is a component of the "TG synthetase" complex that also contains acyl CoA ligase, acyl CoA acyltransferase and monoacylglycerol acyltransferase (47). If DGAT1 is also part of a protein complex in adipocytes, it is possible that the other components of the complex serve to stabilize DGAT1 protein when it is associated with the complex. Therefore, one could speculate that when DGAT1 protein is produced in large quantities, it would be in free form and, therefore, targeted for rapid protein degradation. This hypothesis could be further tested by examining and comparing DGAT1 protein levels in the presence and absence of specific protease inhibitors that may block the proteolytic pathway responsible for the targeted DGAT1 degradation. We have assessed DGAT1 protein levels in 3T3-L1 adipocytes in the presence and absence of ALLN, a nonspecific proteosome inhibitor.
Our results did not shown any increase in DGAT1 protein in the presence of this protease inhibitor, and do not, therefore, support the notion that posttranslational protein In the absence of evidence for protein stability as an explanation for the posttranscriptional control of DGAT1 protein, we speculate that DGAT1 expression is controlled at the translational level. Translational regulation of protein expression has been described for various types of mammalian proteins (48)(49)(50)(51) including ApoB (52,53). Although the efficiency of protein translation may be controlled both at the initiation and elongation, translation initiation has been more extensively studied. In this regard, the primary and secondary structures of the 5'-untranslated region (5'-UTR) of the mRNA may constitute important cis-acting regulatory elements. The cap structure at the 5' terminus (54) and the internal ribosome-entry sites (IRESes) (55), the oligopyrimidine tracts at the extreme 5' terminus (56), the Kozak sequence about the initiation codon (57), the presence or absence of secondary structures (58) or upstream AUGs (uAUGs) (51,59) have all been shown to influence the efficiency of translation initiation. hDGAT1 has a 244 bp 5'-UTR with a potential to form extensive secondary structures because of its high GC content (74% GC overall, and above 80% in several stretches of sequences; c.f. (5)); it also contain an uAUG. These features make it possible that the initiation of hDGAT1 translation is inefficient at baseline, and may, thus, be subjected to significant translational regulation. Of note, a recent study of hDGAT1 gene in Turkish women revealed sites of polymorphism in the 5'-UTR that appear to link to We do not know if there are any elements in the 3'-UTR of hDGAT1 that may function to influence its translation efficiency, albeit no similar AU-rich sequence was found. The hypothesis that translational initiation control of hDGAT1 plays a significant role in protein expression will have to be formally tested in future experiments, in which both the 5'-UTR and the 3'-UTR may be systematically investigated under controlled experimental conditions; specific cis-acting elements may be identified and mapped with the use of a report gene (e.g., the CAT or luciferase gene) linked immediately downstream or upstream of the hDGAT1 5'-UTR or 3'-UTR, respectively.
Although DGAT1-deficient mice have less fat mass, they are not lipoatrophic (9) because of the redundancy in TG biosynthesis. Thus, even when one enzyme is permanently absent, adequate TG biosynthesis can still be sustained by another enzyme(s). TG homeostasis may, therefore, be regulated by controlling the expression and functions of the involved enzyme(s). In addition to transcriptional regulation, we have demonstrated a rigorous posttranscriptional mechanism that appears to control the protein level of DGAT1. This control is not absolute, and a significant amount of fluctuations in protein expression can occur. Importantly, any changes in the level of DGAT1 protein appear to be effectively linked to metabolic changes. This has been demonstrated here with a >2-fold increase in TG accumulation secondary to ~2-fold increase in DGAT1 protein in 3T3-L1 adipocytes. This degree of increase in TG stores is certainly significant with regard to energy homeostasis. If such an increase were to occur in transgenic mice, the weight gain due to excessive fat accumulation would be substantial. Compared with the downward deviation in TG homeostasis, in which only about one third of normal TG stores is affected due to DGAT1 deficiency in the knockout mice (10), it appears that the upward deviation in adipocyte TG mass is relatively easier when DGAT1 protein level is increased. These phenomena may be explained by the evolutionary pressure on selection for mechanisms that ensure adequate TG store for survival and successful breeding, but less pressure on selection against over-storage of fat. Similar ideas have been suggested when the role of leptin in energy homeostasis is examined (cf. (63)).
The fact that increased protein levels of DGAT1 can result in substantial increases in TG mass in adipocytes may point to a potential role of this enzyme in the development of obesity. Further identification of mechanism(s) or potential molecular mediator(s) for posttranscriptional control of DGAT1 expression may allow us to develop therapeutic interventions to correct defects related to the activity of this enzyme. A transgenic mouse with adipose tissue-specific DGAT1 overexpression, which is being created in our      Total cellular TG mass. Cells as in B, except that assays were done 5 days after viral transduction. Asterisk, p<0.01; double asterisks, p<0.05.