The biosynthesis and storage of triglyceride (TG) is an important cellular process conserved from yeast to man. Most mammalian cells accumulate TG in lipid droplets, most prominent in adipocytes, which are specialized to store large amounts of the TG over long periods. In this study, we followed TG biosynthesis and targeting by fluorescence imaging in living 3T3-L1 adipocytes and COS7 fibroblasts. Key findings were (i) not only TG but also its direct metabolic precursor diacylglycerol, DG, accumulates on lipid droplets; (ii) the essential enzyme diacylglycerol acyltransferase 2 associates specifically with lipid droplets where it catalyzes the conversion of DG to TG and (iii) individual lipid droplets within one cell acquire TG at very different rates, suggesting unequal access to the biosynthetic machinery. We conclude that at least part of TG biosynthesis takes place in the immediate vicinity of lipid droplets. In vitro assays on purified lipid droplets show that this fraction of the biosynthetic TG is directly inserted into the growing droplet.
All organisms store metabolic energy to satisfy needs when nutrient levels decrease. Mammals store most of this energy as triglyceride (TG) in the lipid droplets of adipocytes (1). Adipocytes are specialists with high rates of metabolite uptake and synthesis of TG, which is stored in large lipid droplets (10- to 100-μm diameter) and mobilized upon demand by the co-ordinated action of several lipases (2). Although essentially all cell types contain small lipid droplets, most non-adipocytes store only little TG and do not form large droplets. However, a basal ability to esterify fatty acids and glycerol is important for all cells because even low amounts of free fatty acid are toxic (3).
Little is known about mechanisms by which cells package newly synthesized TG into storage droplets (1). Lipid droplets in animals, plants and yeast consist of a hydrophobic core made from neutral lipids surrounded by a phospholipid monolayer with proteins attached both integrally and peripherally (4,5). Biogenesis of neutral lipids is assumed to occur at the endoplasmic reticulum, ER. In mammals, the pathway starts by acylation of glycerol-3-phosphate with two fatty acids to give phosphatidic acid, PA. Subsequent dephosphorylation yields diacylglycerol (DG), the precursor for synthesis of both phospholipids and, by a third acylation, TG. Two enzymes diacylglycerol acyltransferase (DGAT)1 (6) and DGAT2 (7) catalyze the TG synthesis. Recently, microscopic analysis of plant cells overexpressing tagged DGAT1 and DGAT2 pinpointed their localization to distinct regions of the ER, excluding each other (8). Also in mammalian cells, overexpressed, tagged DGAT2 localized to the ER (9). Analysis of phenotypes of knockout mice point to a complex physiological organization of TG synthesis. Mice lacking the ubiquitously expressed DGAT1 have reduced body fat, a lactation defect and are resistant to diet-induced obesity (10). Interestingly, they show normal TG levels in plasma and adipose tissue. DGAT2 is expressed in many tissues with high levels in liver and white adipose tissue. DGAT2 deficiency in mice is postnatally lethal because of severe disturbance of energy metabolism and skin barrier function (11). DGAT1’s inability to compensate for the loss of DGAT2 supports the notion of two different roles for the two enzymes in TG metabolism.
To address the functional organization of TG biosynthesis, we performed biochemical and microscopical studies of the lipid flow during TG synthesis in COS7 fibroblasts and differentiated 3T3-L1 adipocytes, which are amenable to both live cell fluorescence microscopy and subcellular localization studies by immunofluorescence microscopy. For that, we have recently introduced polyene lipids (12), which are fluorescent analogues of high structural similarity to natural lipids. These analogues show both a very similar metabolic behavior and kinetics of uptake and transport. Preliminary data also showed efficient incorporation into storage lipids and targeting to lipid droplets (12). In the current study, we use polyene fatty acids to visualize TG biogenesis and targeting to lipid droplets in vivo, complemented by subcellular imaging of DGAT2, a key enzyme of TG synthesis.
Targeting of neutral lipids to lipid droplets in COS7 cells
When COS7 fibroblasts were incubated for 30 min with fluorescent all-trans-6,8,10,12,14-hexadecapentaenoic acid (t16:5)-fatty acid, lipid droplets became intensely stained (Figure 1). To visualize the flow of lipid into lipid droplets, living COS7 cells were incubated with 50 μm of t16:5-fatty acid and image stacks were acquired over 15 min (Figure 2A). After 2 min, a diffuse perinuclear staining, presumably ER, with some punctate components was observed, together with staining of the nuclear envelope. Already 1 min later, obvious punctate lipid droplet staining appeared and continuously increased in intensity over the next minutes. The fluorescent lipid species present at the various time-points during microscopy were identified by thin-layer chromatography (TLC) analysis (Figure 2B, left panel; Table 1). After initial adsorption of t16:5-fatty acid (0 min), rapid incorporation into phosphatidylcholine, PC and PA (1 min) was observed, followed by incorporation into DG (2 min) and TG (5 min). Labeling of the metabolic end products PC and TG increased constantly, while labeling of the metabolic intermediates PA and DG reached a plateau. For comparison, a corresponding experiment was also performed with radioactive [3H]palmitic acid (Figure 2B, right panel). Both gave very similar results, with the respective lipid metabolites becoming detectable at similar time-points. Differences found included lower and higher amounts of free radioactive fatty acid and PA, respectively. Metabolism downstream of DG, i.e. to PC and TG was virtually identical. Total uptake of both fluorescent and radioactive fatty acid was very similar, at 2–3 nmol/mg of protein in the first 10 min in several experiments.
Table 1. Correlation of fatty acid incorporation into cellular lipids and targeting to lipid dropletsa
PC (% of total)
Free fatty acid (% of total)
DG (% of total)
TG (% of total)
% staining in lipid droplets
Number of droplets (% of 9-min value)
Distribution of fluorescent lipid species and appearance of fluorescence in lipid droplets were quantified from TLC (n = 4, average ± SD) and microscopy time series (n = 3, average ± SD), respectively, such as shown in Figure 2 and specified in Materials and Methods. Results are displayed as per cent of total fluorescence of the TLC lane or the microscopy frame at the respective time-point. Values are corrected for background. For some time-points, not both TLC analysis and imaging are available as indicated by an em-dash. The number of droplets is presented relative to their number at time-point 9 min. After this time-point, individual lipid droplets could not be resolved as they are usually clustered.
13.9 ± 1.6
68.6 ± 2.0
10.4 ± 2.5
1.8 ± 0.6
22.2 ± 1.0
54.1 ± 2.0
14.7 ± 1.4
2.3 ± 0.8
28.8 ± 1.3
40.6 ± 1.8
19.6 ± 1.1
3.4 ± 1.1
14.5 ± 5.1
33.7 ± 1.9
30.7 ± 2.7
21.6 ± 0.9
7.7 ± 2.8
19.7 ± 3.7
35.2 ± 2.2
25.4 ± 1.1
21.1 ± 1.6
10.2 ± 1.6
24.0 ± 6.1
26.5 ± 4.6
39.7 ± 4.2
17.4 ± 3.7
19.1 ± 1.9
17.0 ± 2.6
26.3 ± 5.5
37.1 ± 4.2
15.6 ± 2.2
17.0 ± 3.2
22.3 ± 2.9
34.5 ± 3.4
11.3 ± 3.2
16.0 ± 1.7
27.9 ± 2.5
Individual fluorescent lipid species present at the various time-points were quantified from the TLC plates and correlated to the pixel intensities of lipid droplets in microscopic images (Table 1). Already at early time-points (3–5 min), when very little TG was present, intense staining of lipid droplets was observed. Also at later time-points, the fraction of cellular fluorescent TG was too low to account for the fluorescence intensity of lipid droplets, indicating the presence of at least one more fluorescent lipid on lipid droplets. Free fatty acid, whose band intensity remained nearly constant during the time–course, is unlikely to contribute to a significant extent to the increasing lipid droplet fluorescence. The best correlation between droplet fluorescence and fluorescent lipid species would be achieved, if at any moment, in addition to the TG, about half of the fluorescent DG would localize to lipid droplets. To exclude that the amount of polyene lipids in the droplets is overestimated because of hyperchromic effects in the hydrophobic environment, we measured fluorescence emission intensity of the t16:5 fatty acid in PBS and organic solvents of different polarity. Compared with PBS, the fluorescence intensity in various alcohols was found to be about 15% higher, in hexane about 30% weaker (data not shown). This excludes a large hyperchromic effect of the fluorophor in an apolar environment, consistent with previous measurements by others (13). A direct characterization of fluorescent lipid species present on lipid droplets, i.e. analysis of purified lipid droplets from the labeled COS7 cells, failed. Consistent with observations by others (14), the small lipid droplets of fibroblast-like cells, which are not fed with large amounts of fatty acid for many hours, could not be purified by flotation on a density gradient.
DGAT2 localizes to lipid droplets in COS7 cells
The likely presence of DG on lipid droplets pointed to a possible DG-to-TG conversion at this organelle. However, the acyltransferases that catalyze this reaction, DGAT1 and DGAT2, were reported to reside in the ER (8,9). Revisiting the amino acid sequence of DGAT2, we noted an uncharged stretch (Q67-T97), which may allow for the proteins association with the lipid droplets hydrophobic core or the surrounding membrane monolayer in analogy to the caveolins (15–17). We designed DGAT2 fusion proteins featuring an hemagglutinin (HA)-tag at the N- or C-terminus or internally at a position 17. The latter construct was generated because in DGAT2, both the extreme N- and C-termini are highly conserved in all mammalian species (7), suggesting their functional importance.
When COS7 cells transiently expressing HA-tagged DGAT2 were analyzed by immunofluorescence microscopy, all three tagged constructs were detected in the ER (Figure 3A and B, lower rows), in accordance with other reports (8,9). In contrast, when the untagged wild-type DGAT2 was transiently expressed, only a small fraction of the protein was found in the ER but the majority localized to lipid droplets (Figure 3B, top row) where it formed partial, but usually incomplete, rings contacting the lipid droplets (Figure 3A, top row). Noteworthy, in cells strongly overexpressing wild-type DGAT2, much more of the protein was found in the ER (data not shown), pointing to a saturation of targeting mechanisms. The endogenous level of DGAT2 in COS7 cells was too low to allow detection with our antibody (data not shown). The association of DGAT2 with lipid droplets was dependent on fatty acid supply (Figure 4A). In cells that were kept in medium lacking oleate, also the overexpressed untagged wild-type DGAT2 was found in the ER (Figure 4A, bottom row). To analyze the association with lipid droplets at higher resolution, oleate-treated COS7 cells transfected with wild-type DGAT2 were analyzed by immuno-electron microscopy (EM) using the anti-DGAT2 antibody (Figure 4B,C). Consistent with the observations by others, lipid droplet morphology is only partially retained as the lipid core often condenses and is lost during sample preparation (18). In transfected cells, DGAT2 was highly localized to lipid droplets, defined as large, electron-lucent structures lacking a limiting membrane bilayer (Figure 4B). Frequently, DGAT2 labeling on double membranes in close proximity to lipid droplets and occasionally on the lipid cores could be detected. Untransfected control cells and transfected cells not incubated with the primary antibody showed only negligible background labeling (data not shown). To confirm the association of DGAT2 with lipid droplets, oleate-induced lipid droplets were isolated by flotation on a sucrose gradient. This droplet fraction was processed for immuno-EM using the anti-DGAT2 antibody. Droplets from cells transfected with wild-type DGAT2 showed strong staining, both at areas where droplets were covered with a surface layer (18) (Figure 5B, upper panels) and directly on the droplet core itself (Figure 5B, lower panels).
To further corroborate our findings, biochemical activity assays were performed on purified lipid droplets from COS7 cells (Figure 5C). When incubated with DG and radioactive acyl-CoA, lipid droplets of cells transfected with DGAT2 efficiently synthesized radiolabeled TG (Figure 5C, lanes 5–7). Even without addition of DG, a small amount of TG was produced using the endogenous DG on lipid droplets (Figure 5C, lane 9). A small but detectable amount of TG was made also in droplets from non-transfected cells (Figure 5C, lanes 1–3). The bottom fractions of both transfected and non-transfected cells showed robust TG synthesis (Figure 5C, lanes 11 and 12). The presence of DGAT2 on the purified droplets of transfected cells was confirmed by Western blotting (Figure 5D). Noteworthy, acyl-CoA cholesteryl acyltransferase 1 (ACAT1), a lipid-metabolizing membrane protein of the ER, was lacking in the purified droplets (Figure 5D), demonstrating the absence of ER contaminations.
Endogenous DGAT2 localizes to the vicinity of lipid droplets in adipocytes
Adipocytes are an important system for studies of lipid uptake and neutral lipid biogenesis. Their big lipid droplets are easily detectable by light microscopy. Higher levels of DGAT2 expression are found in 3T3-L1 adipocytes, which allowed visualization of the endogenous protein. DGAT2 was found to associate with lipid droplets (Figure 6), confirming our results obtained in COS7 fibroblasts. Upon careful analysis, the DGAT2-positive structures appeared to be in contact with parts of the lipid droplet surface, not fully surrounding them. The DGAT2-positive structures were continuous, but not overlapping, with the protein disulfide isomerase (PDI)-positive ER (Figure 6A, inset, two color merge), suggesting that DGAT2 resides either on the droplet surface that is not covered by ER or in an ER subregion that excludes PDI. Attempts to colocalize DGAT2 with other ER markers (KDEL/BiP, calnexin and calreticulin) did not clarify this issue because they gave very weak and poorly defined stainings; only PDI in combination with formaldehyde + glutaraldehyde fixation resulted in a reliable ER staining in the differentiated cells. However, even this advanced fixation protocol was not able to fully preserve the morphology of the big lipid droplets found in 3T3-L1 adipocytes, as illustrated by apparently shrunken and ruptured 4,4-difluoro-1,3,5,7,8-pentamethyl-4-bora-3a,4a-diaza-s-indacene (BODIPY 493/503)-positive lipid cores (Figure 6).
Targeting of neutral lipids to lipid droplets in adipocytes
Lipid flow into lipid droplets was also studied in adipocytes. Living 3T3-L1 adipocytes were incubated with 50 μm of t16:5-fatty acid, and image stacks were acquired over 30 min (Figure 7A). After 1–2 min of incubation, a diffuse reticular staining, presumably ER, appeared followed by a rapid accumulation (∼2 min) of fluorescence in small, round droplets of about 1 μm size found at the bottom of the cells and near big lipid droplets, which initially appeared as characteristic unstained, round structures. With time (2–6 min), the intensity of the diffuse ER staining increased. In larger lipid droplets with diameters of about 5 μm, fluorescence was first detected after 3 min, followed by a continuous increase in labeling intensity. After 15–30 min, these droplets reached staining intensities comparable to those of small droplets after 3–5 min. Often, some of the larger droplets stained very weakly, even if neighboring droplets showed intense staining at much earlier time-points. Higher-magnification images allowed for a more detailed analysis of this inhomogeneity (Figure 7B,C). It became clear that filling of individual lipid droplets with fluorescent lipid occurred with very different kinetics, even for those of identical size (Figure 7C). Frequently, we observed the appearance of brightly stained partial rings around weakly stained larger droplets (Figure 7B), reminiscent of EM pictures of ER in contact with lipid droplets (18). Like COS7 cells, 3T3-L1 adipocytes rapidly metabolize the t16:5 polyene fatty acid to various intermediates and the final products, phospholipids and TG, as revealed by TLC analysis (Figure 8A). After initial adsorption of t16:5-fatty acid (0 min), rapid incorporation into PA (1 min) was followed by conversion to DG and PC (3 min) and finally by incorporation into TG (5 min). With increasing time, labeling of the metabolic end products PC and TG increased constantly, while labeling of the metabolic intermediate DG reached a plateau. Upon a 10- to 60-min chase, free fatty acid and PA levels were strongly reduced, while DG concentration decreased by ∼60%. When the images (Figure 7A) were correlated to the fluorescent lipids present at each time-point, the first visualization of small droplets coincided with the synthesis of DG, PA and PC (2–3 min). Again, positive labeling of big lipid droplets, starting at 3 min, was seen before metabolic appearance of TG (5–10 min). This finding pointed to the possibility that also in adipocytes, not only TG but also some of its metabolic precursors would accumulate on lipid droplets. To address this question, we repeated both imaging and TLC analysis of t16:5 metabolism in the presence of 1.2 mm 2-bromo-octanoic acid, an inhibitor of TG biosynthesis (19), which does not influence the fluorescence of the polyene fluorophor (data not shown). When adipocytes were incubated in the presence of 2-bromo-octanoic acid (Figure 8B), almost no TG was synthesized and both its immediate precursor DG and free fatty acid strongly accumulated. Production of PA and PC, as well as fatty acid and PA clearance during chase, was slightly delayed and reduced, indicating a small inhibitory effect on other acyltransferase reactions in lipid metabolism.
In the presence of the inhibitor, lipid traffic was changed (Figure 9). The diffuse reticular staining was more intense, indicating an ER accumulation of lipid. The number of small fluorescent lipid droplets was strongly reduced, and they could only be detected at later time-points (5–6 min). After 5 min, a delayed acquisition of fluorescent lipid by large lipid droplets started, after metabolic accumulation of DG (Figure 8B). Remarkably, the intensity of lipid droplets increased continuously throughout the course of the experiment (Figure 9), although only very little TG was produced (Figure 8B).
DG is found on lipid droplets
To characterize the molecular species of polyene lipids present on lipid droplets in adipocytes, droplets were purified by flotation on a sucrose gradient. Cells were incubated with 50 μm t16:5-fatty acid in the absence or presence of 1.2 mm 2-bromo-octanoate inhibitor, followed by homogenization and gradient analysis. Lipids in gradient fractions were analyzed by TLC (Figure 10). The top fractions containing the lipid droplets contained fluorescent TG as well as DG (Figure 10A, left). The TG-to-DG ratio in droplets from cells with inhibited TG synthesis (Figure 10A, right) changed considerably, as more fluorescent DG and also some fluorescent fatty acid, but substantially less TG could be detected. Fluorescent PC was hardly detectable on lipid droplets. Compared with earlier experiments (Figure 8B), where metabolism was stopped immediately, inhibition of TG biosynthesis appeared less efficient when lipid droplets were purified by flotation on sucrose gradients (Figure 10). We speculate that the long ultracentrifugation work up (about 40 min) allows for some TG biosynthesis as harvested cells not immediately subjected to lipid extraction also synthesize labeled TG (Figure 10B). To exclude that the presence of DG on lipid droplets is a peculiarity of polyene lipids, we repeated the analysis for endogenous lipids after labeling with [3H]palmitic acid. Lipid droplets contained a significant amount of DG (Figure 10B, left) already under normal conditions. Upon inhibition of TG biosynthesis, major amounts of DG accumulated at the droplets after 30 min of labeling (Figure 10B, right).
The major component of the hydrophobic core of adipocyte lipid droplets is the storage lipid TG. In this study, we show that these organelles also contain considerable amounts of DG. Although smaller amounts of DG on lipid droplets are produced during lipolysis (20), our experimental setup confirms its presence during lipogenesis. Upon acylation, DG becomes converted to TG, a reaction catalyzed by the DGAT proteins. We find DGAT2 associated with many lipid droplets in both fibroblasts and adipocytes. These findings are in contrast to two recent publications (8,9), which localized DGAT2 to the ER by immunofluorescence microscopy without any indication of lipid droplet staining. This discrepancy may be explained by the use of epitope-tagged constructs combined with overexpression in those studies. Also in our hands, overexpression of three different tagged DGAT2 fusion proteins resulted in their localization to the bulk ER. Only the untagged protein localized to lipid droplets. Interestingly, DGAT2 is found to undergo a transition between lipid droplet association and a localization in the bulk ER, depending on the lipid state of COS7 cells. Activation of TG synthesis by oleate supply is necessary for association of DGAT2 with lipid droplets. Because many lipid droplets are in direct vicinity of ER cisternae (18), droplet-associated DGAT2 can in principle localize to the lipid droplet surface itself and/or to the directly opposing ER membrane. The hairpin-like topology of the DGAT2 membrane domain (8,9) would, by analogy with caveolins (15–17), allow for both localizations. By immunofluorescence analysis (Figure 5A,B), wild-type DGAT2 clearly separates from the bulk of the sec61-positive ER and surrounds lipid droplets in ring-like structures. Yet, by close inspection, a fraction of these rings still contains a low level of ER marker, suggesting that part of them actually are specific ER structures. This is supported by the EM analysis of DGAT2 localization in Figures 4 and 5, which localized DGAT2 to both the droplet surface itself and surrounding or adherent membrane structures. Consistent with this notion, a fraction of the expressed DGAT2 purifies with lipid droplets, separating from the ER membrane marker ACAT1 (Figure 5D). These data actually argue that a relative, small fraction of the DGAT2 resides directly on the droplet surface itself, but it should be kept in mind that the (i) biochemical droplet purification is usually not complete and (ii) in the biochemical analysis, those cells that very strongly overexpress the protein (and therefore contain it in the ER) tend to dominate the picture. Interestingly, the small fraction of DGAT2 on the droplets seems to have disproportional high enzymatic activity (compare Figure 5D, LD versus bottom, with Figure 5C, lanes 5–7 versus 12, in which the same fractions were analyzed), suggesting that the droplet DGAT2 is much more active than the pool localized to the ER. In adipocytes, a large fraction, but not all, of both the big and the small lipid droplets showed association with DGAT2, which presumably accounts for the observed differences in the rate by which individual lipid droplets acquire fluorescent lipid. If TG synthesis was inhibited, fluorescent lipid accumulated in the ER, but considerable amounts of fluorescent DG still entered the large lipid droplets by an unknown mechanism. Interestingly, under those conditions, endogenous DGAT2 levels in 3T3-L1 adipocytes fell below detection limit of our antibody (data not shown), pointing to a destabilization or downregulation of the protein. Our observation that the presence of DG on droplets precedes appearance of TG is in accordance with earlier radiolabeling experiments in adipose tissue (21) and supports the proposed model (Figure 11) as this confirms the availability of both substrates for TG biosynthesis, DG and acyl-CoA. Recent analyses of the proteome of lipid droplets (22–25) consistently found fatty acyl-CoA synthetases on the droplet surface itself, providing the activated substrate for the DGAT2-catalyzed reaction. Very recently, acyl-CoA synthetase activity as well as DGAT activity were also found on a purified droplet fraction from HuH7 hepatoma cells (26). Taken together, our findings point to a dedicated role of DGAT2 in long-term lipid storage, while the ER-resident DGAT1 may be more important for synthesis of TG destined for secretion. This would be consistent with the tissue expression pattern of the two enzymes as DGAT2 is highly expressed in adipose tissue, whereas DGAT1 has its highest expression in liver and gut. Furthermore, overexpression of DGAT1 in a hepatoma cell line led to peripheral accumulation of neutral lipid, while DGAT2 overexpression resulted in formation of large lipid droplets (11).
Our finding that TG is synthesized close to the lipid droplet surface demonstrates that existing lipid droplets can grow by influx of TG. It has, however, no direct impact on the question of how nascent lipid droplets are formed in the first place. The widely accepted current model of TG biosynthesis links TG production to lipid droplet formation (27) as biosynthesized TG is thought to accumulate between the leaflets of the ER membrane before budding off and forming a lipid droplet. Those small lipid droplets are believed to grow by homotypic fusion (28,29). Our findings support the notion that lipid droplets have additional ways of growing in size besides droplet to droplet fusion.
Materials and Methods
Fluorescent t16:5 was synthesized as described (12). [9,10-3H]palmitic acid was obtained from Hartmann. BODIPY 493/503 was from Molecular Probes. [3H]oleoyl-CoA and [3H]palmitoyl-CoA were synthesized as described (30). Delipidated BSA was from Applichem; cell culture reagents and Lipofectamine 2000 from Invitrogen. Hanks’ balanced salt solution (HBSS), common lab chemicals and solvents were from Sigma. Silica TLC plates were obtained from Merck; no. 1.5 glass bottom dishes from Mattek.
Cell culture and lipid delivery
COS7 and 3T3-L1 cells were maintained as described previously (12). Confluent 3T3-L1 cells were differentiated to adipocytes with 500 μm 3-isobutyl-1-methylxanthine (IBMX), 10 μm dexamethasone (DEX) and 5 μg/mL insulin. After 2 days, IBMX and DEX were omitted. For the next 5 days, fresh medium was applied every day until at least 90% of cells was differentiated. After a total of 7 days of differentiation, adipocytes were used. Sixteen hours prior to live cell microscopy or lipid analysis, adipocytes were transferred to low-glucose microscopy medium lacking phenol red and supplemented with delipidated serum (12). For live cell microscopy, cells were grown in glass bottom dishes, for immunofluorescence microscopy on coverslips, and for lipid analysis in 10-cm plastic dishes. Fatty acids were delivered as BSA complexes in HBSS as described (12).
Purification of lipid droplets
3T3-L1 adipocytes in a 15-cm dish were incubated with 50 μm t16:5-fatty acid or [3H]palmitic acid in the absence or presence of 1.2 mm 2-bromo-octanoic acid. For the latter, cells had been incubated with 1.2 mm 2-bromo-octanoic acid for 2 h before. After 30 min, cells were washed and scraped in ice-cold disruption buffer [100 mm KCl, 25 mm Tris, pH 7.4, 5 mm EGTA and 1 mm ethylenediaminetetraacetic acid (EDTA)] before homogenization in a cooled European Molecular Biology Laboratory (EMBL) cell cracker (HGM) with five strokes using a maximum clearance of 18 μm. Homogenization was confirmed by inspection under a microscope before adjusting the lysate to 540 mm sucrose. This solution was overlaid with ice-cold 270 mm and 135 mm sucrose in disruption buffer and top buffer (25 mm Tris, pH 7.4, 1 mm EGTA and 1 mm EDTA). If applying, all buffers were supplemented with 1.2 mm 2-bromo-octanoic acid. The gradients were centrifuged in a swing-out rotor at 200 000 × g at 4°C for 30 min. Seven fractions were collected from the top. Lipids were extracted and analyzed by TLC as described (12).
For EM analysis of lipid droplets from COS7 fibroblast, cells were incubated with 100 μm oleic acid for 16 h (31). Cells were washed and scraped into ice-cold dissociation buffer (100 mm KCl, 25 mm HEPES, pH 7.4, 5 mm EGTA and 1 mm EDTA) containing protease inhibitors (Roche) and lysed by sonication for 10 seconds. The lysate was adjusted to 540 mm sucrose and overlaid with ice-cold 270 mm and 135 mm sucrose in dissociation buffer and top buffer (25 mm HEPES, pH 7.4, 1 mm EDTA and 1 mm EGTA). The gradients were centrifuged in a swing-out rotor at 166 000 × g at 4°C for 45 min.
For DGAT2 activity assays, COS7 fibroblast was incubated with 50 μm oleic acid for 36 h (31). Cells were washed and scraped into ice-cold homogenization buffer (20 mm HEPES, pH 7.4, 200 mm sucrose and 1 mm DTT) before homogenization with five strokes in a 0.7 × 40-mm needle and in a cooled EMBL cell cracker with 13 strokes using a 18-μm clearance. The homogenate was centrifuged at 1000 × g at 4°C for 10 min, and the supernatant was mixed with an equal amount of buffer (20 mm HEPES, pH 7.4, 2 m sucrose and 1 mm DTT) and overlaid with homogenization buffer. The gradients were centrifuged in a swing-out rotor at 200 000 × g at 4°C for 2.5 h. A top fraction (3 mL) containing lipid droplets and an intermediate and bottom fractions (5 mL each) were collected and aliquots thereof analyzed by Western blotting and used for DGAT2 activity assays.
DGAT2 activity assay
Aliquots of the lipid droplet fraction (200 μL) or bottom fraction (120 μL) were mixed with 100 or 180 μL, respectively, of assay buffer (200 mm Tris, pH 7.4, 10 mm MgCl2, 2 mg/mL delipidated BSA, 20 μm oleoyl-CoA, 20 μm palmitoyl-CoA, 0 or 200 μm dioleoylglycerol and 0.35 μCi each of [9,10-3H]oleoyl-CoA, [9,10-3H]pamitoyl-CoA and [9,10-3H]myristoyl-CoA) and incubated at 30°C for 1 h. A control sample replacing the lipid droplet fraction with homogenization buffer was carried along. Lipids were extracted, separated by TLC in iso-hexane/diethyl ether/acetic acid 80/20/1 and visualized by fluorography.
pN-HA-DGAT2 and pC-HA-DGAT2 code for human DGAT2 tagged with a triple HA-tag at the N- or C-terminus, respectively. p-int-HA-DGAT2 codes for human DGAT2 with an internal triple HA-tag. This construct translates as follows: the first 17 aa of DGAT2 are followed by a 2-aa spacer (SR), the triple HA-tag and a 2-aa spacer (SR) and the remaining 369 aa of DGAT2. p-wt-DGAT2 codes for untagged, wild-type, human DGAT2. The DGAT2 sequence was amplified from IMAGE EST clone 3903313. pADRP-mRFP1 is coding for a fusion protein of human adipose differentiation related protein (ADRP) followed by a linker (five aa; SPVAT) and monomeric red fluorescent protein 1 (mRFP1). mRFP1 was obtained from R. E. Tsien; the ADRP sequence was amplified from I.M.A.G.E. Consortium EST clone 4801004. All constructs were verified by sequencing. p-sec61-mRFP1 is described elsewhere as pmRFP1-ER (12).
Immunolabelling and antibodies
COS7 cells were fixed with 3.7% formaldehyde. For some experiments, 1.25% glutaraldehyde was used in addition. Wild-type adipocytes were fixed with 3.7% formaldehyde and 0.25% glutaraldehyde, followed by background fluorescence quenching using sodium borohydride (50 mg/mL in PBS, 30 min). All cells were permeabilized with 0.05% saponin in PBS. If applying, BODIPY 493/503 was added during incubation with secondary antibody. Rabbit anti-HA polyclonal antibody (Y-11) and mouse monoclonal anti-HA (F-7) were from Santa Cruz, mouse monoclonal anti-PDI from Stressgen, mouse monoclonal anti-GAPDH from Novus Biological and rabbit polyclonal anti-acyl-CoA cholesteryl acyltransferase 1, anti-ACAT1 [DM10 (32)], was generously provided by Dr Ta-Yuan Chang. A polyclonal antibody against 6His-tagged mouse DGAT2 lacking aa 69–112 was raised (Eurogentec) and used after affinity purification (Figure S1). Secondary antibodies were from Molecular Probes.
Imaging of polyene lipids in living cells at 37°C employing a two-photon microscope was performed using a Biorad Radiance 2100 MP setup attached to a Nikon Eclipse TE300 inverted microscope (Biorad Hercules) equipped with a Planapo 60× (1.2 NA). Pulsed excitation (pulse duration 200 femtoseconds, repetition rate 76 MHz and average power 600 mW) at 725 nm was provided by a 5-W Verdi/Mira laser (Coherent). Fluorescent images were acquired using the non-descanned detectors. Long and short wavelengths were split using a 560DCXR beamsplitter (Chroma), and the longer were filtered with a D630/50m bandpass filter (Chroma). Images were acquired using LaserSharp 2000 software (Biorad). All live cell imaging was performed at 37°C; the temperature of the specimen was controlled with a Tempcontrol-37-2-digital unit (Bachofer) that of the objective by an objective heater (Bioptechs). Phase contrast and epifluorescence microscopy of fixed COS7 cells was performed using an Olympus IX70 microscope (Olympus) equipped with a U-Plan FL 60× (1.25 NA). A Polychrome II xenon light source was used (TillPhotonics). For polyenes, a FT395-BP475/65 (Carl Zeiss) and for mRFP1, a FT595-BP630/60 (Chroma) filter combination was used. Digital images were acquired with a NTE/CCD-512-EBFT camera (Roper Scientific). Confocal fluorescence scanning microscopy of fixed cells was performed using a Zeiss LSM 510 Meta microscope (Carl Zeiss) equipped with Plan apochromat differential interference contrast (DIC) 100× (1.4 NA). Differential interference contrast microscopy was performed using a Zeiss Axiovert 200M microscope equipped with a Plan apochromat DIC 63× (1.4 NA). Images were processed employing Adobe Photoshop 6.0 (Adobe). Three-dimensional reconstruction was performed using Imaris x64 4.5.2 software (Bitplane).
Quantification of signal intensities
To quantify fluorescence in microscopy images, different horizontal layers (z-stack) were collapsed onto each other by adding corresponding pixel intensities. Fluorescence in lipid droplets was determined as percentage of total cellular fluorescence after background corrections employing ImageGauge V3.3 software (Fuji). This software was also used to quantify fluorescence signal intensities on TLC plates. Intensities of individual bands were calculated as percentage of total fluorescence per lane after background correction.
For immuno-EM, COS7 cells were incubated overnight in the presence of 100 μm oleate and fixed in 2% paraformaldehyde and 0.2% glutaraldehyde in 0.1 m PHEM buffer [60 mm piperazine-1,4-bis(2-ethanesulfonic acid) (PIPES), 25 mm HEPES, 2 mm MgCl2 and 10 mm EGTA], pH 6.9. Cells were embedded in 10% gelatin, cryoprotected using polyvinylpyrrolidone (PVP)-sucrose and snap frozen onto specimen holders in liquid N2. Thin sections (100 nm) were picked up with a 1:1 mixture of 2.3 m sucrose and 2% methyl cellulose. Alternatively, isolated lipid droplets were applied to Formvar/carbon-coated copper grids and fixed in 2% paraformaldehyde and 0.2% glutaraldehyde. Immunolabeling was performed using the anti-DGAT2 antibody and 10-nm gold-conjugated protein A. Grids were viewed using a Jeol 1011 transmission electron microscope.
This work was financially supported by the Deutsche Forschungsgemeinschaft (DFG) (SFB-TR13 Project D2) as well as a postdoctoral fellowship to L. K. We would like to thank Dr Rob Parton for providing the possibility to perform EM analyses in his laboratory. We thank the MPI-CBG Light Microscopy Facility for continuous support.