V. A. Zammit, Division of Metabolic and Vascular Health, Warwick Medical School, University of Warwick, Coventry CV4 7AL, UK Fax: +44 2476 968653 Tel: +44 2476 522798 E-mail: firstname.lastname@example.org
The two diacylglycerol acyltransferases, DGAT1 and DGAT2, are known to have non-redundant functions, in spite of catalysing the same reaction and being present in the same cell types. The basis for this distinctiveness, which is reflected in the very different phenotypes of Dgat1−/− and Dgat2−/− mice, has not been resolved. Using selective inhibitors of human DGAT1 and DGAT2 on HepG2 cells and gene silencing, we show that, although DGAT2 activity accounts for a modest fraction (< 20%) of overall cellular DGAT activity, inhibition of DGAT2 activity specifically inhibits (and is rate-limiting for) the incorporation of de novo synthesized fatty acids and of glycerol into cellular and secreted triglyceride to a much greater extent than it affects the incorporation of exogenously added oleate. By contrast, inhibition of DGAT1 affects equally the incorporation of glycerol and exogenous (preformed) oleate into cellular and secreted triacylglycerol (TAG). These data indicate that DGAT2 acts upstream of DGAT1, largely determines the rate of de novo synthesis of triglyceride, and uses nascent diacylglycerol and de novo synthesized fatty acids as substrates. By contrast, the data suggest that DGAT1 functions in the re-esterification of partial glycerides generated by intracellular lipolysis, using preformed (exogenous) fatty acids. Therefore, we describe distinct but synergistic roles of the two DGATs in an integrated pathway of TAG synthesis and secretion, with DGAT2 acting upstream of DGAT1.
The two diacylglycerol acyltransferases (DGAT1 and DGAT2) catalyse the same reaction, namely, the final step in the synthesis of triacylglyceride (TAG) involving the esterification of diacylglyceride (DAG) by long-chain acyl-CoA esters , although DGAT1 is known also to catalyse the esterification of monoacylgyceride (MAG) . The two DGATs belong to unrelated protein families. A prominent feature of their functions is that they are largely non-redundant, i.e. they do not functionally compensate or substitute for each other’s catalytic function or loss of expression in vivo. This is reflected in the very different phenotypes of mice carrying global tissue disruption of the respective genes. Thus, Dgat1−/− mice have a mild phenotype that is characterized by reductions in TAG content of tissues (e.g. liver, adipose) and mild hypotriglyceridaemia, accompanied by resistance to the development of obesity and insulin resistance . By contrast, Dgat2−/− mice die soon after birth, suffering from severe lipopenia and major skin function disorders . Overexpression of DGAT1 and DGAT2 in McA-RH7777 cells results in the accumulation of lipid droplets of very distinct sizes and morphology and in different locations within the cells [5,6]. Similarly, in the liver in vivo, only suppression of DGAT2, and not DGAT1, reverses diet-induced hepatic steatosis and insulin resistance . Overexpression of DGAT1 in fast and slow muscle fibres in vivo results in improved insulin sensitivity  whereas overexpression of DGAT2 in fast-twitch fibres results in whole-body insulin resistance , even though both models have increased TAG content in the muscle. When overexpressed in plant cells, the two proteins are localized to different discrete compartments within the endoplasmic reticular membrane (ERM) . Furthermore, DGAT2, but not DGAT1, colocalizes with proteins that catalyse reactions essential for de novo lipogenesis (glycerol-3-phosphate acyltransferase in plants  and stearoyl-CoA desaturase 1, SCD1, in liver ). Both DGAT1 and DGAT2 are expressed in the ERM, but DGAT2 has also been reported to be present in the cytosolic lipid-droplet proteome and on structures where the ERM comes into close contact with mitochondria [13,14] and where the synthetic pathway leading to DAG synthesis is initiated [15–17].
A major unresolved question in the synthesis and secretion of TAG in triglyceride-rich lipoprotein secreting cell types (hepatocytes and enterocytes) is how TAG present in, and metabolized from, cytosolic lipid droplets is made available, after hydrolysis, for the formation of lumenal non-apoB-associated lipid droplets within the ERM . Our working model [19,20] has proposed that DAG, by virtue of its rapid transmembrane diffusion, connects DGAT activities expressed on both sides of the ERM. This has been supported experimentally by the detection of DGAT activity on both aspects of the ERM (‘overt’ and ‘latent’ DGAT [19,21,22]) in isolated microsomes and in HepG2 cells . Further support for such a role for DAG comes from the demonstration of intralumenal TAG synthesis from substrates provided on the cytosolic aspect of the membrane [24,25]; the demonstration of TAG accumulation in cytosolic and endoplasmic reticular lumenal droplets in vivo after overexpression of the two DGATs ; and, latterly, topological studies on DGAT2 and DGAT1 in the ERM  in which we showed that whereas (in agreement with ) DGAT2 activity is expressed entirely on the cytosolic aspect of the ERM, DGAT1 activity is expressed on both aspects of the ERM .
These previous findings raised the question how these various DGAT activities contribute towards the distinctive functions of DGAT1 and DGAT2 in the cellular accumulation and secretion of TAG in hepatocytes; differential inhibition of each of them could potentially be used as part of pharmacological strategies aimed at targeting specific outcomes with respect to hepatic steatosis and/or hypertriglyceridaemia, both associated with (patho)physiological states characterized by insulin resistance. Of particular interest [20,28,29] is the involvement of de novo fatty acid synthesis, which is essential [30–32] (as is desaturation [33–35]) for hepatic TAG synthesis but which, apparently paradoxically, makes only a minor contribution towards the acyl chain composition of secreted TAG [30,31] except under conditions of greatly increased carbohydrate-induced hepatic lipogenesis .
In the present study, the availability of selective inhibitors for endoplasmic reticular overt and latent human (h) DGAT1  and for hDGAT2, combined with the use of suitably labelled substrates, has enabled us to identify largely distinct roles for DGAT1 and DGAT2 in (a) the incorporation of de novo synthesized fatty acids and nascent DAG into TAG, (b) the retention of the glyceride moiety within TAG, and (c) the cellular accumulation and secretion of TAG in a human hepatocyte-derived cell line. We find that the two DGATs act primarily in series, rather than in parallel, with DGAT2 acting upstream of DGAT1 by catalysing the initial formation of TAG from de novo synthesized (‘new’) fatty acids and nascent diglycerides. Our findings further suggest that DGAT1 promotes TAG accumulation by using preformed or ‘old’ fatty acids to catalyse the re-esterification of partial glycerides formed by lipase-mediated hydrolysis of TAG in cytosolic and endoplasmic reticular lumenal lipid droplets [35,37].
In addition to the two DGAT1 inhibitor compounds (iA and iB) described in , the compound N-(4,5-dihydronaphtho[1,2-d]thiazol-2-yl)-2-(3,4-dimethoxy phenyl) acetamide (iC) (Fig. S1) was used as an inhibitor of both hDGAT2, at low concentrations, and hDGAT1, at high concentrations (Fig. 1A). The IC50 values for inhibition by compound iC of hDGAT2 and hDGAT1 expressed heterologously in insect cell membranes were 8.5 ± 0.6 μm and 70.8 ± 4.6 μm respectively (Fig. 1). The effects of iC were also tested on overt DGAT activity in HepG2 cells (Fig. 1B) that had had their plasma membrane permeabilized to small molecules with mild digitonin treatment, but with retention of the intactness of the ERM . This enabled us to quantify the dose–response relationship for the inhibition by iC of overt DGAT activity (i.e. activity on structures that face the cytosol, e.g. the ERM, mitochondria and lipid droplets to which cytosolic acyl-CoA has access ). As expected, the dose–response curve (Fig. 1B) showed a pattern of inhibition that was intermediate between the respective dose–response curves observed for iC inhibition of hDGAT2 and hDGAT1 (broken lines in Fig. 1B) when these were expressed in insect cell membranes (Fig. 1A). Thus, at lower concentrations (≤ 15 μm), inhibition of overt DGAT activity by iC followed the curve expected for inhibition of hDGAT2, whereas at higher concentrations there was an increasing component that had the characteristics of hDGAT1 inhibition. These data were as anticipated from our previous observation that overt DGAT activity in hepatocytes is due to both DGAT2 and DGAT1 . They also suggested that iC could be used as a selective inhibitor of DGAT2 when used at low concentrations. Thus, it could be calculated that at 15 μm iC inhibits DGAT2 by ∼ 80% whereas DGAT1 activity is inhibited by only ∼ 20%. The availability of such a selective inhibitor of the two enzymes, coupled with the availability of the two highly specific inhibitors for DGAT1 (iA, which is selective for latent DGAT1, and iB, which inhibits overt and latent DGAT1 with equal potency ) enabled us to study the relative roles of DGAT1 and DGAT2 in the synthesis and secretion of triglyceride in HepG2 cells.
Contributions of DGAT2 to overall cellular DGAT activity
As shown in Fig. 2A, 15 μm iC inhibited a minor (∼ 20%) proportion of overt DGAT activity (measured in cells treated only with digitonin). As this concentration of iC inhibits 80% of DGAT2 (and 20% of DGAT1 activity), the results suggested that DGAT2 constitutes a relatively small proportion of overt DGAT activity, and an even smaller fraction (∼ 12%) of overall cellular DGAT activity (Fig. 2A, Control 2). At 70 μm iC, a concentration at which > 95% of DGAT2 and ∼ 50% of DGAT1 activity are inhibited (Fig. 1A), there was a further inhibition of overt DGAT activity of HepG2 cells, as expected from previous observations that DGAT1 is partly expressed on the overt aspect of the ERM (cf. 23). When latent DGAT activity was also exposed to oleoyl-CoA substrate (by treating the cells sequentially with digitonin and alamethacin ) to obtain total DGAT activity, 15 μm iC did not inhibit the latent DGAT activity, as expected from the exclusively cytosol-facing topology of the DGAT2 active site in the ERM  and the absence of DGAT2 activity expression on the latent aspect of the ERM . However, at 70 μm iC, a substantial proportion of latent DGAT activity (which is due exclusively to DGAT1 ) was inhibited by iC, as expected from the ability of higher concentrations of iC also to inhibit DGAT1 (Fig. 1A). By comparison, iB (240 nm), a specific inhibitor of hDGAT1 described in , inhibited ∼ 80% of overt DGAT activity and ∼ 90% of total cellular DGAT activity, as expected from its ability to inhibit both overt and latent DGAT1 with equal potency . Routinely, we observed a residual DGAT-type activity (∼ 8% of total cellular DGAT activity) that was overt and resistant to individual or combined inhibitor action; this is considered to be due to the activity of other enzymes with DGAT-like activity (e.g. monoacylglycerol acyltransferase, MGAT1 , which, when expressed in insect cell membranes, was found to be insensitive to all three inhibitors used in the present study; data not shown).
Rate-limiting role of DGAT2 for de novo TAG synthesis
Measurement of the absolute rate of the net synthesis of triglycerides is quantified by the incorporation of 3H-glycerol, being the rate of incorporation of the glyceroyl backbone into DAG used for esterification to TAG (by DGAT activity) net of the 3H-label lost upon hydrolysis of TAG without subsequent re-esterification. By contrast, incorporation of exogenously added 14C-labelled fatty acids can occur both during initial formation of TAG and during re-esterification of partial glycerides generated from TAG lipolysis. In particular, incorporation of fatty acids into TAG can also occur independently of de novo glyceride synthesis when partial glyceride molecules (DAG, MAG) generated from the hydrolysis of cellular TAG (including that already present in the cell prior to the addition of radiolabelled substrates) are re-esterified. Consequently, differential labelling of TAG with 3H-glycerol and 14C-fatty acid (preformed or synthesized de novo) is indicative of the source of the fatty acids and diglyceride used by the two DGATs. Therefore, in the present study we have used dual labelling, namely with 3H-glycerol and either 14C-oleate or 14C-acetate, so that we could measure concurrently the rates of incorporation of the fatty acid (preformed or de novo synthesized, respectively) and glyceroyl moieties into TAG.
In spite of the small proportion of total cellular DGAT activity accounted for by DGAT2 (see black bars in Fig. 2B), its inhibition had the highest observed effect on the incorporation of 3H-glycerol into the combined cellular and secreted TAG fractions (Fig. 2B). As at this concentration of iC 80% of DGAT2 is inhibited, this indicates that DGAT2 exerts a highly rate-limiting effect on 3H-glycerol incorporation into TAG even though it accounts for a small fraction of overall cellular DGAT activity. The observations in Fig. 2B suggest (a) that DGAT2 catalyses preferentially the incorporation of newly synthesized (nascent) DAG (with newly incorporated 3H-glycerol) into TAG, i.e. that it mediates the initial incorporation of 3H-glycerol into triglycerides rather than the re-esterification of partial glycerides formed by TAG lipolysis; and (b) that DGAT2 does not use exogenously added preformed fatty acid (14C-oleate) as substrate.
By contrast, the rates of incorporation of 3H-glycerol and 14C-oleate were inhibited equally and stoichiometrically (in a ratio of ∼ 1 glycerol : 3 oleate molecules; see legend to Fig. 4, later) by iA (at 0.15 μm), which inhibits latent DGAT1, or by iB (at 240 nm) which inhibits both overt and latent DGAT1 . Importantly, both these effects were quantitatively proportional to the degree of inhibition of cellular DGAT activity by these inhibitors (Fig. 2B). These data suggest that loss of DGAT1 activity results in the simultaneous loss of 3H-glycerol and 14C-oleate retention in the TAG fraction, as would be expected if DGAT1 inhibition prevented the re-esterification of DAG and/or MAG resulting from TAG hydrolysis (Fig. 2B).
Effects of sequential DGAT2 and DGAT1 inhibition on TAG synthesis and secretion
In order to test the above inferences, we performed experiments in which the rates of incorporation of 3H-glycerol and 14C-oleate into cellular and secreted TAG were measured simultaneously at increasing concentrations of iC, to titrate the activities of DGAT2 and DGAT1 sequentially (see Fig. 1A). This allowed us to obtain IC50 values for the inhibition of each of four cellular functions simultaneously, namely (a) the incorporation of 3H-glycerol into cellular and (b) secreted TAG; and (c) the incorporation of 14C-oleate into cellular and (d) secreted TAG. The data in Fig. 3 show that each of these four functions showed distinctive IC50 values for inhibition by iC, such that between them they spanned the entire range of iC action (see Fig. 1A) from an IC50 corresponding to exclusive DGAT2 involvement (IC50 = 3.1 ± 0.4 μm for incorporation of 3H-glycerol into cellular TAG) to an IC50 indicative of exclusive DGAT1 action (82 ± 5 μm iC for incorporation of 14C-oleate into cellular TAG) (Fig. 3A). It is noteworthy that the two extremes (differing by almost two orders of magnitude) were observed for the incorporation of 3H-glycerol and 14C-oleate into the same fraction, namely cellular TAG which constituted > 98% of overall TAG labelling (see the legend to Fig. 3A). These data indicated that de novo synthesis of TAG (which is measured by 3H-glycerol incorporation) is primarily a function of DGAT2 (lowest IC50) and that it is largely independent of the incorporation of preformed fatty acid (14C-oleate) into cellular TAG (Fig. 3A). As the latter was inhibited with an IC50 of 82.5 μm this was indicative of mediation through the inhibition of DGAT1.
For secreted TAG (which constituted < 2% of total TAG labelling), the IC50 for the inhibitory effects of iC on 3H-glycerol incorporation was several-fold lower than that for 14C-oleate incorporation (Fig. 3B), but both values were intermediate between the extreme values described above for cellular TAG labelling. These observations suggested that DGAT1 (inhibited at higher iC concentrations) is heavily involved in the use of 14C-oleate during the extensive cycling between TAG and partial glycerides generated by lipolysis which occurs prior to TAG secretion [32,39].
This inference was tested by studying the dose–response curves for inhibition of either latent DGAT1 only (≤ 0.15 μm iA) or both overt and latent DGAT1 (≤ 240 nm iB)  in the four functions listed above (3H-glycerol and 14C-oleate into cellular or secreted TAG) which were quantified simultaneously. The data are shown in Fig. 4. It is apparent that (a) both incorporation of 3H-glycerol and 14C-oleate (preformed fatty acid) into cellular or secreted TAG had identical IC50 values for either of the two DGAT1 inhibitors; (b) that these IC50 values were identical to those observed for the inhibition of hDGAT1 and hDGAT2 overexpressed in insect cells ; and (c) that the maximal degree of inhibition was only 60–70%. These data indicated (a) that incorporation of preformed fatty acid (14C-oleate) into TAG by DGAT1 activity (overt and latent forms) is strictly coupled to the retention of 3H-glycerol within the TAG fraction (i.e. through the re-esterification of partial glycerides formed by TAG-lipase action), and (b) that only ∼ 70% of the pool of labelled TAG is subject to hydrolysis/re-esterification.
Effects of DGAT1 and DGAT2 inhibition on the incorporation of de novo synthesized fatty acids into TAG
In order to test the above inference that DGAT2 uses preferentially de novo synthesized fatty acids as substrates, we performed experiments in which 3H-glycerol incorporation into TAG was quantified in the presence of 0.5 mm14C-acetate (and unlabelled oleate). This allowed us to compare the effects of the inhibition of DGAT2 or DGAT1 on the incorporation of de novo synthesized fatty acids into cellular and secreted TAG when the availability of substrate for de novo fatty acid synthesis was increased. The inhibitor concentrations used were those known to be optimally effective in inhibiting DGAT2 selectively (15 μm iC), latent DGAT1 specifically (1.5 μm iA) or both overt and latent DGAT1 (240 nm iB). The data are presented in Fig. 5.
Inhibition of DGAT1 (either latent or both latent and overt) resulted in equal inhibition of incorporation of 14C-oleate and 3H-glycerol into cellular TAG (Fig. 5A); by contrast, when 0.5 mm acetate was present (Fig. 5C), the degree of inhibition of 3H-glycerol into cellular TAG by these inhibitors of DGAT1 was approximately halved (compare Fig. 5A,C). In control experiments, addition of unlabelled acetate (0.5 mm) to incubations in which oleate was the labelled substrate did not alter the degree of inhibition of the incorporation of 14C-oleate into cellular TAG (not shown) but lowered the degree of inhibition of 3H-glycerol incorporation by iA and iB (cf. Fig. 5C). These observations showed that, when denovo fatty acid synthesis was increased (by addition of acetate), the ability of DGAT1 inhibition to affect 3H-glycerol incorporation/retention within the TAG fraction was markedly reduced, i.e. that 3H-DAG containing de novo synthesized fatty acids is a poorer substrate for DGAT1 than that containing exogenously added oleate.
Selective DGAT2 inhibition (using 15 μm iC) resulted in very different degrees of inhibition of 14C-oleate and 3H-glycerol incorporation into cellular TAG (Fig. 5A), with a much greater (threefold) inhibition of 3H-glycerol incorporation compared with that of 14C-oleate incorporation into cellular TAG (Fig. 5A), in accordance with the dose–response relationships in Fig. 3A. When 14C-acetate was the labelled substrate (Fig. 5C), selective inhibition of DGAT2 (with 15 μm iC) inhibited completely the incorporations of both 14C-acetate and 3H-glycerol into cellular TAG (Fig. 5C), as a result of a more than threefold greater inhibition of 14C-acetate than that of 14C-oleate by 15 μm iC (compare Fig. 5A,C). Consequently, there was a parallel effect of the inhibition of DGAT2 on the incorporation of 14C-acetate and 3H-glycerol into cellular TAG (Fig. 5C) in contrast to the markedly different responses to DGAT2 inhibition of the incorporation of 3H-glycerol and 14C-oleate (Fig. 5A), indicating that DGAT2 is primarily involved in the de novo incorporation of both glycerol and acetate-derived fatty acids into TAG.
Although incorporation of labels into secreted TAG constituted ≤ 2% of overall TAG labelling, the reciprocal relationships between the effects of the DGAT1 and DGAT2 inhibition on the incorporation of labelled preformed (oleate) and de novo synthesized fatty acid (from acetate) were also apparent when incorporation into this fraction was measured (Fig. 5B,D). Thus, inhibition of DGAT1 affected 14C-oleate incorporation into secreted TAG to a twofold greater extent than that of 14C-acetate (compare Fig. 5B,D), whereas DGAT2 inhibition showed the opposite pattern.
Gene-silencing experiments confirm the rate-limiting role of DGAT2 on TAG synthesis
Because iC is a selective and not a specific inhibitor of DGAT2 at low concentrations, it was considered necessary to test the above conclusions, and particularly the rate-limiting role of DGAT2 in net TAG synthesis (3H-glycerol incorporation), by comparing the effects of the inhibitors with those of gene silencing of the two enzymes. The results are shown in Fig. 6, in which it can be seen that, as in the case of inhibition by iC, the small loss in cellular DGAT induced by the silencing of DGAT2 is accompanied by an inordinately large inhibition of TAG synthesis; the inhibition of 3H-glycerol incorporation was seven to eightfold higher, in each case, than the loss of cellular DGAT activity (Fig. 6, right). By contrast, the much larger degree of DGAT activity achieved by silencing of DGAT1 (or by iA or iB treatment) was accompanied by strictly proportionate inhibition of 3H-glycerol incorporation (cf. Figs 2 and 4).
The respective roles of DGAT1 and DGAT2 in triglyceride synthesis, and particularly the rationale for their evident non-redundancy in vivo, in spite of catalysing the identical reaction, have not been elucidated. The present study provides evidence that the two enzymes act primarily in series rather than in parallel in hepatocytes; i.e. that DGAT2 acts upstream of DGAT1 and is primarily responsible for the initial synthesis of TAG using de novo synthesized fatty acids and nascent DAG containing these fatty acids. By contrast, the data indicate that DGAT1 functions to retain the glyceride moiety within the TAG fraction by catalysing the re-esterification (with preformed, exogenously added oleate) of DAG and MAG formed after lipase-mediated hydrolysis of TAG. The central evidence for these conclusions is (a) that, although DGAT2 activity constitutes a relatively minor proportion of overall cellular DGAT activity (< 20%), its inhibition results in a near-total inhibition of 3H-glycerol incorporation into TAG; (b) that, in dose–response experiments with iC (a preferential inhibitor of DGAT2), incorporation of the glyceroyl moiety into TAG is inhibited in parallel with DGAT2 inhibition, with DGAT1 playing a very minor role in this process; and (c) that, whereas inhibition of DGAT2 has a minor effect on the incorporation of preformed fatty acid (exogenous oleate) into TAG, it is rate-limiting for the incorporation of both glycerol and de novo synthesized fatty acid into TAG.
The large effect of DGAT2 inhibition (disproportionate compared with the contribution of its activity towards overall cellular DGAT activity) on both 14C-acetate and 3H-glycerol incorporation into TAG shows that this enzyme uses primarily nascent diglyceride and de novo synthesized fatty acids as substrates, and that these pools of substrates are not available to DGAT1. The observation that inhibition of the incorporation of preformed fatty acid (14C-oleate) into cellular TAG showed characteristics corresponding solely to those of the inhibition of DGAT1 (Fig. 3A) suggests that DGAT1, but not DGAT2, is primarily responsible for the downstream re-esterification of the partial glycerides (DAG, MAG) formed after lipase-mediated hydrolysis of TAG (Fig. 7). This conclusion was supported by experiments in which small interfering RNA (siRNA) mediated knockdown of DGAT1 (versus that of DGAT2) also showed equal inhibitions of cellular DGAT activity and 3H-glycerol incorporation into TAG. This would imply that, in hepatocytes, DGAT1 acts primarily to retain the glyceroyl moiety within TAG by re-esterifying DAG and MAG with preformed (‘old’) fatty acid.
The data in Fig. 5 also show that in the case of secreted TAG there is a much lower degree of dissociation between the labelling of the glyceride backbone and the de novo synthesized acyl moieties esterified to it, in agreement with previous observations that de novo synthesized fatty acids are enriched in TAG that is preferentially secreted without prior hydrolysis/re-esterificaton cycling and that this is more pronounced under conditions of increased de novo fatty acid synthesis (e.g. after fructose feeding in vivo) [31,32,40].
Our conclusions explain the well-established observation that very low density lipoprotein (VLDL)-TAG formation is strictly linked to the rate of de novo fatty acid synthesis in the liver [30,31,40]. Thus, although in humans and animal models fed a normal carbohydrate diet the absolute contribution of de novo synthesized fatty acids to VLDL-TAG is low (≤ 5%) [41,42], inhibition of de novo fatty acid synthesis results in an almost complete cessation of TAG synthesis and VLDL-TAG secretion [30,31,40]. In addition, because fatty acid synthesis results in the formation of a saturated fatty acid, this process is closely coupled to desaturation, catalysed by SCD1. Inhibition of SCD1 or global disruption of the Scd1 gene similarly has an almost complete inhibitory effect on TAG synthesis in vivo  which cannot be rescued by dietary triolein feeding , suggesting that the DGAT enzyme that utilizes denovo synthesized (and desaturated) fatty acids exerts a rate-limiting effect on overall TAG synthesis and secretion, as would be predicted from the upstream function of DGAT2, with respect to DGAT1, suggested by our data.
It is noteworthy that even maximal inhibition of total DGAT1 activity (with 240 nm iB) resulted in ≤ 70% inhibition of TAG labelling with either 14C-oleate or 3H-glycerol. This suggests that a proportion (∼ 30%) of newly synthesized TAG exists in a pool that bypasses the hydrolysis/re-esterification cycling and is incorporated directly into cellular lipid droplets or used without prior hydrolysis for secretion. It is well established that a sizable minority of TAG is secreted without prior hydrolysis/re-esterification [44,45]. In this respect, Steiner and colleagues [32,39] showed that only about 70% of newly synthesized hepatic TAG is racemized in vivo before secretion and that ∼ 30% of newly synthesized TAG is incorporated into cellular or secreted TAG without entering the pool that undergoes hydrolysis and re-esterification [32,39]. Our observations are also in agreement with those on lipid droplet formation which have suggested that small nascent and ‘old’ large lipid droplets in the cytosolic compartment of McArdle cells have different dynamics and that, whereas there is transfer of TAG from small nascent droplets to larger lipid droplets, this does not happen in the reverse direction . In this context, it is of interest that different lipid droplets may have differential access to lipogenic enzymes  and that DGAT2, but not DGAT1, is associated with lipid droplets .
The suggested primary role for DGAT1 in re-esterifying partial glycerides after their formation through TAG hydrolysis implies that this enzyme is very important in the accumulation of cellular TAG as it rescues partial glycerides from further hydrolysis or diversion to phospholipids. This would explain why overexpression of DGAT1 in mouse liver in vivo increases hepatic TAG content . DGAT1 has considerable MGAT activity in its own right  and may therefore be responsible for the re-esterification to TAG of both DAG and MAG, obviating the requirement for substantial MGAT activity in adult liver . DGAT1 activity may need to be substantially higher compared with that of DGAT2 in hepatocytes (as in Fig. 2A), as its catalytic activity would need to accommodate the high rate of cycling between TAG and partial glycerides [44,45].
Because of the rapidity of the hydrolysis/re-esterification cycling, the de novo synthesized fatty acids incorporated into TAG during initial (DGAT2-mediated) synthesis may be rapidly replaced by preformed fatty acids, thus resulting in the low absolute contribution of de novo synthesized fatty acids to secreted TAG unless the rate of their synthesis is increased by for example a high carbohydrate diet [30,31].
Our conclusions may also explain the recent observation, obtained with liver-specific DGAT1 knockout mice, that DGAT1 is required for the development of steatosis due to excess exogenous (preformed) fatty acid supply to the liver, whereas steatosis associated with increased hepatic de novo fatty acid synthesis is not DGAT1 dependent . These data are fully consistent with our conclusion that DGAT2 utilizes substrates that are, or incorporate, de novo synthesized fatty acids, whereas DGAT1 is primarily involved in esterifying DAG with ‘old’, preformed fatty acids. More generally, our conclusions also explain the extreme lipopenic phenotype of global Dgat2 gene disruption in mice .
Other observations suggest that, as indicated by our present data, DGAT2 and DGAT1 may have access to different pools of fatty acids and diglycerides in animal and plant cells. Thus, DGAT1 and DGAT2 have been shown to occupy distinct micro-domains within the endoplasmic reticulum , whereas DGAT2 is also localized in lipid droplets  and in structures where the endoplasmic reticulum comes in close proximity to lipid droplets or mitochondria . Overexpression of DGAT1 and DGAT2 in McArdle cells results in lipid droplets of different morphology and intracellular distribution, leading to the suggestion that they have access to different pools of substrates . Importantly, DGAT2, but not DGAT1, has been shown to colocalize with SCD1  in the liver and with glycerol-3-phosphate acyltransferase(s) in plant cells . Our conclusions that DGAT1 and DGAT2 may use different pools of DAG may also explain the apparently paradoxical observations  that, when DGAT2 is overexpressed in the liver of transgenic mice, there is a greater increase in the hepatic levels of DAG than of TAG [52,53]. In these livers, DAG would be expected to accumulate through the action of lipase(s) on the increased amounts of TAG synthesized by overexpressed DGAT2 under conditions where the enzyme for which it is the substrate (DGAT1) is expressed only at normal levels.
It will be of interest to confirm whether these specialized functions and pathway relationships of DGAT2 and DGAT1 are specific to hepatocytes or whether they are also present in other cell types, e.g. in muscle, in which overexpression of either of the two enzymes (in one or both fibre types, respectively) has opposite effects on whole-body insulin sensitivity in spite of similar increases in tissue triglyceride levels [8,9].
Previous reports have purported to show that, in isolated cells, DGAT1 can compensate for the absence of DGAT2 when the incorporation of substrates is studied, although not quantified [4,54]. This conclusion obviously cannot be extrapolated to the situation in vivo where the respective transgenic mice have very different phenotypes. In this respect, it is important to note that the incorporation of 14C-oleate into TAG by fibroblasts from Dgat2−/− mice  does not necessarily indicate net TAG synthesis, owing to the propensity of DGAT1 to use exogenous preformed fatty acids for re-esterification of partial glycerides (see above).
Role of DAG in TAG secretion
We have previously shown that DGAT1 catalytic activity is expressed approximately equally on the cytosolic and luminal aspects of the ERM in HepG2 cells . The current data may explain why this is necessary. In hepatocytes, TAG lipases occur in association with cytosolic and endoplasmic reticular lumen lipid droplets . We suggest that the intra-endoplasmic reticular lipase(s)  hydrolyse TAG in endoplasmic reticular lumenal lipid droplets (non-apoB containing ) and that latent DGAT1 activity re-esterifies the DAG and MAG generated, in order to maintain the intra-luminal endoplasmic reticular pool of TAG that is used for the ‘second step’ lipidation of nascent VLDL . We have previously suggested that DAG equilibrates rapidly through the ERM  to provide the substrate for latent DGAT1-mediated synthesis of luminal lipid droplets. Therefore, overt and latent DGAT1 activities could both be involved in retaining the glyceride backbone within the TAG present within the cytosolic or endoplasmic reticular luminal droplets (Fig. 7). Our present data show that DGAT2 does not perform this partial glyceride re-esterification function on the cytosolic aspect of the ERM because, possibly owing to its localization within the cell and its close association with other enzymes, it appears to be specialized for the incorporation of de novo synthesized fatty acids and nascent DAG into newly synthesized TAG. Therefore, re-esterification of DAG and MAG necessitates the expression of DGAT1 activity also on the cytosolic aspect of the ERM, as we have shown experimentally . This explains our present observations that there were only minor differences in the labelling of cellular or secreted TAG when either only latent DGAT1 was inhibited (with iA) or both overt and latent DGAT1 activities were inhibited (with iB), as both types of activity would be expected to contribute towards determining the DAG available for re-esterification reactions on either aspect of the membrane owing to the rapid equilibration of DAG across the membrane.
In conclusion, DGAT2 is suggested to act upstream of DGAT1 in HepG2 cells and to be involved in the incorporation of de novo synthesized fatty acids and nascent glyceroyl moieties into TAG, whereas DGAT1 incorporates preformed (‘old’) fatty acid, remodelling TAG acyl composition in the process. Inhibition or knockdown of DGAT2 largely abolishes de novo synthesis of TAG. This explains the severely lipopenic phenotype of Dgat2−/− mice  but not of Dgat1−/− animals . The different cellular distribution, micro-compartmentation, substrate pool utilization and association of DGAT2 with specific partner-proteins (SCD1  and glycerol-3-phosphate acyltransferase ) are reminiscent of the distinction by PPARα between ‘new’ and ‘old’ fat as a source of its activating ligands . It rationalizes the involvement of DGAT2 in the interactions between carbohydrate-induced lipogenesis and hypertriglyceridaemia in conditions that are characterized by insulin resistance  and the central importance of de novo lipogenesis in high carbohydrate diet induced hepatic accumulation of triglycerides and development of hypertriglyceridaemia .
Materials and methods
RPMI-1640 medium (2000 mg·L−1 glucose and without sodium pyruvate), l-glutamine, fetal bovine serum and trypsin/EDTA were obtained from Invitrogen (Paisley, UK). DMEM (1000 mg·L−1 glucose) was from Sigma (Poole, UK). The hepatoblastoma cell line HepG2 was obtained from the European Collection of Cell Cultures (ECACC). Radiolabelled [1-14C]oleoyl-CoA (specific activity 50 μCi·μmol−1) was obtained from Amersham (GE Healthcare UK Ltd). [1-14C]oleic acid (specific activity 58.2 mCi·mmol−1) and [2-14C]acetic acid (acetic acid sodium salt, specific activity 50 mCi·mmol−1) were from Perkin Elmer. [1,2,3-3H]glycerol (specific activity 40 Ci·mmol−1) was purchased from American Radiolabelled Chemical Inc. Alamethicin, 1,2-dioleoyl-sn-glycerol, oleoyl coenzyme A lithium salt, sodium oleate, essentially fatty-acid-free BSA and the lipid standards glyceryl trioleate and oleic acid were purchased from Sigma. Digitonin was purchased from Fischer Scientific UK. LK6D 19-channel thin-layer chromatography plates were purchased from VWR-Jencons, UK. All other chemicals used were of analytical grade. siRNA transfection reagents, Smartpool On-Target Plus siRNA and scrambled siRNA preparations specific for DGAT1 and DGAT2 were from Dharmacon Inc. (Epsom, UK) and were used according to the supplier’s instructions.
Expression of human DGAT1 and DGAT2 in insect cells
The cDNAs for hDGAT1 and hDGAT2 were prepared and transfected into insect cells using Baculovirus as described previously .
HepG2 cell culture
HepG2 cells were cultured routinely in low glucose RPMI-1640 medium (5 mm glucose) with 2 mm l-glutamine, supplemented with 10% fetal bovine serum. Cells were grown in an incubator at 37 °C with a water-saturated 5% CO2 atmosphere. The cell medium was changed every 3–4 days. Cells were subcultured every 5–6 days, just prior to reaching about 80–90% confluency, into new vented culture flasks or into six-well plates for experimental work. HepG2 cells were used after 20–25 subcultures. Unless specified otherwise, cells were passaged and harvested after 0.25% trypsin/EDTA treatment.
Measurement of overt and latent DGAT activity
The method was as described in . Assays were carried out using detached cells. Confluent cells were trypsinized and incubated on ice in CSK medium  containing 30 μg·mL−1 digitonin for 5 min, to expose the overt DGAT activity. To detect latent DGAT activity, after digitonin treatment the cells were washed using ice-cold CSK buffer and resuspended in CSK buffer; they were then further incubated with 20 μg·mL−1 alamethacin on ice for 30 min.
Overt DGAT activity in HepG2 cells and microsomes was measured by carrying out the DGAT assay using digitonin-only treated HepG2 cells . Latent DGAT activity was measured in HepG2 cells after permeabilization of their endoplasmic reticulum with alamethacin and quantified as the difference between total DGAT activity and overt DGAT activity. Inhibitors were added in dimethylsulfoxide; the solvent was added to the same final concentration (0.01%) in all assays.
Assay of cellular TAG accumulation and secretion
Cells were plated at 1.5 × 106 cells·well−1 in 10 cm2 six-multiwell dishes in 2 mL of growth medium on day 1. On day 3, cells were washed in phosphate-buffered saline, the medium was replaced with serum-free medium and cells were pre-incubated for 30 min with increasing concentrations of the inhibitors dissolved in dimethylsulfoxide. The control reactions had only dimethylsulfoxide added to them (final concentration 0.001% dimethylsulfoxide in all incubations). At the end of the incubation period, the substrates were added to each well and cells were incubated at 37 °C for 2 h. This time period was determined in preliminary experiments to give linear rates of incorporation. Final substrate concentrations were 0.7 mm [1-14C]oleate (2.0 × 103 dpm·nmol−1), 0.25 mm glycerol (10.0 × 103 dpm·nmol−1). All incubations were performed in duplicate, and experiments were performed for the number of times indicated in the figure legends. When the effect of inhibitors on incorporation of de novo synthesized fatty acid into cellular and secreted TAG was to be studied 5 mm [2-14C]acetate (1 μCi·mmol−1) was added.
Treatment of cells with siRNA specific for DGAT1 or DGAT2
Cells were plated at a density of 3.3 × 105 cells·well−1 in six-well plates. Transfections were carried out 24 h after plating, using Smartpool siRNA (Dharmocon) designed by the manufacturers for DGAT1 and DGAT2, and transfection reagent DharmaFect 4 (Dharmocon) according to the manufacturer’s instructions. To the transfection volume (400 μL) of serum-free and antibiotic-free RPMI medium were added different amounts of a 5-μm stock of siRNA-DGAT1 or siRNA-DGAT2 or siRNA-control and 4–10 μL of transfection reagent to optimize knockdown of the respective proteins. The mixture was incubated for 20 min at room temperature and then diluted with 1.6 mL of RPMI medium containing 10% fetal bovine serum and added to the cells. Control cells were treated with scrambled siRNA (Dharmacon). Untreated cells were incubated with the transfection reagent only. Optimization of the transfection conditions showed that maximal effects were obtained at 70 nm siRNA for both DGAT1 and DGAT2, with 6 μL of transfection reagent. Cells were cultured for up to 96 h with siRNA-containing medium which was supplemented with fresh siRNA every 24 h. Treatment for 72 h was found to yield optimal effects. At the time of harvesting, cells were trypsinized, washed, sedimented and used for DGAT assays or for quantification of the incorporation of 3H-glycerol into TAG.
Extraction and analysis of lipids
Lipids were extracted from the cells or the media using a chloroform–methanol mixture (2 : 1 v/v) and the lipids were saponified. Extraction, chromatography and quantification of TAG labelling were performed as described in reference 23.
The data presented are means ± SEM for the number of independent experiments indicated. Statistical significance was calculated by using Student’s t-test, and P < 0.05 was considered significant.
HRW was supported by a UK BBSRC-CASE studentship in association with AstraZeneca Ltd.