Isolation of a gene encoding a 1,2-diacylglycerol-sn-acetyl-CoA acetyltransferase from developing seeds of Euonymus alatus.

1,2-Diacyl-3-acetyl-sn-glycerols (ac-TAG) are unusual triacylglycerols that constitute the major storage lipid in the seeds of Euonymus alatus (Burning Bush). These ac-TAGs have long-chain acyl groups esterified at both the sn-1 and sn-2 positions of glycerol. Cell-free extracts of developing seeds of E. alatus contain both long-chain acyl-CoA and acetyl-CoA sn-1,2-diacylglycerol acyltransferase (DGAT) activity. We have isolated a gene from developing seeds of Euonymus alatus that shows a very high sequence similarity to the members of the DGAT1 gene family (i.e. related to acyl-CoA:cholesterol acyltransferases). This Euonymus DGAT1 gene, when expressed in wild type yeast, results in a 5-fold enhancement of long-chain triacylglycerol (lc-TAG) accumulation, as well as the appearance of low levels of ac-TAG. Hydrogenated ac-TAG molecular species were identified by gas chromatography-mass spectrometry. Microsomes isolated from this transformed yeast show diacylglycerol:acetyl-CoA acetyltransferase activity, which is about 40-fold higher than that measured in microsomes prepared from yeast transformed with the empty vector or with the Arabidopsis thaliana DGAT1 gene. The specific activity of this microsomal acetyltransferase activity is of the same order of magnitude as the microsomal long-chain DGAT activities measured for yeast lines transformed with the empty vector or either the Arabidopsis or Euonymus DGAT1 genes. Despite this, ac-TAG accumulation in yeast transformed with the Euonymus DGAT1 gene was very low (0.26% of lc-TAG), whereas lc-TAG accumulation was enhanced. Possible reasons for this anomaly are discussed. Expression of the Euonymus DGAT1-like gene in yeast lines where endogenous TAG synthesis has been deleted confirmed that the gene product has both long-chain acyl- and acetyltransferase activity.

The occurrence and the structural characterization of the 3-acetyl-1,2-long-chain diacyl-sn-glycerols (1,2-diacyl-3-acetins, or ac-TAG) 1 in seed oils were reported by Kleiman et al. (1). Unlike most long-chain triacylglycerols (lc-TAG), these ac-TAGs exhibit a strong optical activity because of asymmetry at the central carbon atom of the glycerol moiety introduced by the acetyl group. They have been found mainly in plants of the family Celastraceae but also in the families Lardizabalaceae, Ranunculaceae, Rosaceae, and Balsaminaceae. These unusual triacylglycerols are found in varying amounts, but in the genus Euonymus they can represent up to 98% of the total triacylglycerols in the seed oil. In Euonymus the sn-1 and sn-2 positions of the acetyl glycerides are esterified with common longchain fatty acids, predominantly palmitate, oleate, and linoleate. Short-chain fatty acids are also found esterified in the triacylglycerols of milk from ruminants, particularly at the sn-3 position (2), and ac-TAGs have been identified in bovine udder lipids (3). Thus it is likely that ac-TAGs are often a minor component in the human diet. In addition, SALATRIM (short and long acyltriglyceride molecule) triacylglycerols have been developed as a commercial reduced calorie fat (4,5). These are interesterified TAGs containing saturated fatty acids (largely stearate) and short-chain fatty acids (acetate, propionate, and/or butyrate). Lipase-catalyzed interesterification has also been used to produce acetylacylglycerols (6).
Triacylglycerols can be by synthesized by sn-1,2-diacylglycerol:acyl-CoA acyltransferases (EC 2.3.1.20) (DGAT) or by various transacylases. The first DGAT gene to be cloned was from mouse (7) and is a member of the DGAT1 gene family, which is closely related to the cholesterol:acyl-CoA acyltransferase gene family. Homologous DGAT1 genes have been cloned from Arabidopsis and other plants (8 -11). In Arabidopsis there is only one DGAT1 gene, and studies with mutant lines generated by chemical mutagenesis or tDNA tagging indicate that this gene contributes significantly to seed oil content (10,11). A second family of DGAT genes (DGAT2) was first identified in the oleaginous fungus Morteriella ramanniana, based upon purification of a DGAT activity, peptide sequencing, and subsequent isolation of the corresponding genes (12). DGAT2 genes do not share sequence similarity to DGAT1 genes. Several DGAT2 homologues are reported in fungi, plants, and mammals (13,14). Most recently, a gene encoding a bifunctional wax ester synthase/acyl:CoA diacylglycerol acyltransferase was characterized in Acinetobacter calcoaceticus (15). Several related open reading frames have been identified in Arabidopsis. This novel long-chain acyl-CoA acyltransferase is not related to the DGAT1 or DGAT2 gene families described above, and any role in triacylglycerol synthesis in higher plants still has to be established. Transacylases can produce triacylglycerols via acyl-CoA-independent reactions. A phospholipid:1,2-diacylglycerol acyltransferase has been reported to be involved in triacylglycerol synthesis in yeast and higher plants (16). Triacylglycerol synthesis has been described in developing seed extracts from various plant species using medium-chain acyl-CoAs and/or medium-chain diacylglycerols in DGAT as-says (17)(18)(19)(20). Frequently, the substrate specificity and selectivity of these DGATs are fairly broad. Our in vivo labeling of developing Euonymus seeds with exogenous [ 14 C]acetate produced [acetyl- 14 C]1,2-diacyl-3-acetins with negligible kinetic lag phase, suggesting that is no intermediate acetyl lipid involved in biosynthesis, i.e. that a transacylase mechanism of synthesis was unlikely. 2 In this paper we present evidence for a 1,2-diacylglycerol:acetyl-CoA acetyltransferase activity in extracts of developing Euonymus seeds, and we describe the isolation and characterization of a Euonymus DGAT1 gene that encodes an DGAT with both long-chain acyl-and acetyltransferase activity.

EXPERIMENTAL PROCEDURES
Materials-Developing seeds of Euonymus alatus were harvested during the early fall of 1999 on Michigan State University campus, courtesy of the Grounds Department of Campus Parks, and the endosperm plus embryo tissue were dissected and stored at Ϫ80°C. Arabidopsis thaliana (ecotype Colombia-2) developing seeds were collected from the siliques of 6 -8-week-old plants, grown in the growth chambers (22°C, 16-h light period at 80 -100 microeinsteins of light intensity). Yeast (Saccharomyces cerevisiae) cultures ("wild type" is Invitrogen strain INVSc1; genotype, MATa his3D1 leu2 trp1-289 ura3-52) were grown in either YPD medium or selective synthetic SC medium (according to the protocol of Invitrogen for pYES2CT) supplemented with 2% glucose and were grown at 28°C in a shaker. Induction media for yeast cells containing constructs consisted of SC medium lacking uracil and supplemented with 2% galactose and 1% raffinose. Strain H1266 (MATa are2-⌬::LEU2 dga1-⌬::KanMX4 lro1-⌬::TRP1 ADE2) was obtained from Prof. Sten Stymne, Swedish University of Agricultural Sciences. This line has the three genes encoding almost all of the TAG synthesis capacity of the yeast cell deleted (13). Escherichia coli strain HB101 was grown at 37°C in Luria Broth media (21), supplemented with the appropriate antibiotics for selection of the constructs: 100 g/ml ampicillin (pYES2CT) or 50 g/ml kanamycin (pE1776).
Isolation of Euonymus and Arabidopsis cDNAs Encoding DGAT1-Total RNA was extracted from developing seeds of Euonymus according to the protocol of Chung et al. (22) and from developing seeds of Arabidopsis by the method of Ruuska and Ohlrogge (23). For Euonymus gene isolation, alignment of the deduced amino acid sequence of several plant and mammalian DGAT1 genes showed very well conserved sequences. On the basis of the most conserved sequences between these DGATs, seven sets of degenerated primers were designed and tested in a PCR for accuracy with the Arabidopsis DGAT1 EST AA042298 as template. Two sets of primers (MP1 and MP6) were found to be successful and were used for an RT-PCR with Euonymus total RNA of developing seeds. The cDNA of the total RNA was made with an oligo(dT) primer (Invitrogen). Because the resolution of the products on gel was not clear, the primers were labeled with 32 P, and the products were separated on a polyacrylamide gel. A 330-bp radioactive band was recovered, cloned into the TopoPCR2.1 vector, and sequenced. With gene-specific primers designed on the basis of the obtained 330-bp sequence, 3Ј-and 5Ј-RACE yielded the cDNA ends, using a RACE kit (Invitrogen). Primers used to obtain the 330-bp fragment are MP1 and MP6, and primers for the 3Ј-RACE and primers for the 5Ј-RACE are MP 10, 11, and 16 and MP 30, 15, and 31, respectively.
For a full-length Euonymus cDNA, first strand cDNA was prepared from total RNA using an oligo(dT) primer and reverse transcriptase superscript according to Invitrogen. With the 3Ј primer DAGFEaYes (containing a BamHI site and a yeast translation signal) and 5Ј primer DAGREaYes (containing a XhoI site), a PCR was set up, using high fidelity polymerase Pwo (Roche Applied Science). A product of 1.5 kb, encompassing the coding region and appropriate yeast signals, was obtained, gel-purified (Qiagen kit), and cloned into the BamHI/XhoI site of the yeast vector pYES2CT (Invitrogen). The presence of the insert was verified by whole cell PCR using DGAT1-specific primers, and the correct base composition was verified by sequence analysis, yielding a suitable clone designated pYES2EaPCR5.1. Data base searches were done using the BLAST algorithm (24). DNA sequences and the deduced amino acid sequence were analyzed with the Vector NTI Suite of Infor-Max. The Arabidopsis DGAT1 cDNA was isolated via RT-PCR of total RNA from developing seeds, using the primers DAGFAtYes (containing a BamHI site) and DAGRAtYes (containing an XhoI site). The 1.6-kb PCR fragment was purified with the kit from Qiagen and cloned in the vector pYES2CT. Whole cell PCR of obtained colonies with DGAT1specific primers confirmed the presence of the insert, and the DNA sequence was verified, yielding the clone pYES2At2.2.
Expression of Euonymus (Ea) DGAT1 and Arabidopsis (At) DGAT1 in Yeast Strains-The Ea cDNA and the At cDNA, cloned into the yeast vector pYES2CT, as well as the empty vector, were transferred to the S. cerevisiae strains INVSc1 and H1266 via transformation. Three yeast colonies for each construct were grown in liquid medium and analyzed for lipid content. For the growth phase-dependent analysis with semiquantitative lipid analysis by TLC, a small 5-ml culture of each colony was started in SC medium with 2% glucose and grown overnight. This culture was diluted 1:100 in a volume of 200 ml and grown overnight in SC medium with 2% glucose until the OD was between 1 and 2. This culture was subsequently centrifuged and washed with sterile water and resuspended in 400 ml of SC medium supplemented with 2% galactose and 1% raffinose, with a starting OD of 0.4. Growth was followed over time, and 40-ml samples were taken at early and midexponential phase and beginning, mid-, and late stationary phase. These samples were washed, pelleted, and stored at Ϫ20°C and analyzed for lipids as described below. For the full and quantitative analysis of lipids, 800 ml of yeast cultures of S. cerevisiae and H1266 carrying either pYES2CT, pYES2EaPCR5.1, or pYES2At2.2 were grown until the start of the stationary phase and were treated as described above.
Lipid Analysis-Yeast pellets were quenched by heating in 5 volumes of hot isopropyl alcohol for 10 min to inactivate lipases. The pellets were then vortexed with glass beads and re-extracted with hot isopropyl alcohol. The pellet was extracted with hexane/isopropyl alcohol (25). The initial hot isopropyl alcohol extracts were combined with 1.5 volumes of hexane and added to the lipid extraction. The lipids recovered from the extractions were evaporated to dryness under nitrogen. For semi-quantitative lipid analysis, the total lipid extract was separated by TLC (0.25-mm thickness K6 Silica Gel 60 A plates, Whatman), developed with hexane/diethyl ether/acetic acid 80:20:1 (v/v/v). Visualization of bands was carried out with iodine staining, and bands were compared with mass standards.
For quantitative lipid analysis and identification, GC standards were prepared. 3-Acetyl-1,2-dipalmitin was prepared by the acetylation of 1,2-dipalmitoylglycerol with acetic anhydride in pyridine and purified by preparative TLC. A mixture of palmitoylstearoyl-3-acetin and distearoyl-3-acetin was prepared by hydrogenation of the purified 1,2diacyl-3-acetin fraction isolated from E. alatus seeds. Dipentadecanoyl-␣-acetin was prepared by the interesterification of triacetin (1 mmol) and tripentadecanoin (2 mmol). These triacylglycerols were heated in a capped tube at 250°C in the presence of sodium methoxide (0.2 mmol) catalyst for 1 h. The resulting dipentadecanoyl-␣-acetin was separated from the scrambled products by preparative TLC. The lipids recovered from the large scale yeast cultures were first weighed, and then triheptadecanoin (lc-TAG) and dipentadecanoyl-␣-acetin (ac-TAG) internal standards were added. An aliquot of the sample was hydrogenated. Hydrogenations were carried out by continuous stirring of lipids (5-20 mg) in hexane (2-3 ml) with 2-3 mg of platinum(IV) oxide catalyst and hydrogen at slightly above atmospheric pressure at room temperature for 2 h. The hydrogenated lipid was then treated with ethereal diazomethane to convert free fatty acids to fatty acid methyl esters, because free fatty acids will co-chromatograph with ac-TAG on silica TLC. An aliquot was then transmethylated by using the method of Ichihara et al. (26) and analyzed by GC (with flame ionization detector (FID)) to determine total fatty acid content (against 17:0 from added triheptadecanoin). The lc-TAG and ac-TAG fractions were isolated by preparative TLC using 1-mm thickness K6 Silica Gel 60 A plates (Whatman) and were developed multiple times with hexane/diethyl ether/acetic acid 80:20:1 (v/v/v). Plates were viewed under UV light after spraying with ethanolic dichlorofluorescein (0.2% w/v) to locate the bands. Lipids were eluted from the silica with hexane/isopropyl alcohol 3:2 (v/v). The hydrogenated triacylglycerols were analyzed by high temperature GC using a DB-5ht capillary column (30 m ϫ 0.25 mm inner diameter, 2 M. Pollard, unpublished data. 0.1-m film thickness) with helium carrier gas at 1.5 ml/min. The column was temperature programmed at 5°C/min to 360°C, with the injector set at 380°C and a 40:1 split ratio, and the FID detector temperature was also 380°C. For GC-MS to identify the hydrogenated ac-TAG fractions, a Hewlett-Packard 5890 gas chromatograph-coupled MSD 5972 mass analyzer was used, with the mass analyzer set in electron impact mode scanning from 40 to 700 atomic mass units. Long-chain TAG and ac-TAG quantification was based on relative FID responses relative to triheptadecanoin or to acetyl-dipentadecanoin internal standards, respectively. Since within ac-TAG or lc-TAG molecular species series observed the response factor variation per the internal standard was small (Ϫ5 to Ϫ10%, or Ϫ5 to ϩ15%, respectively), no response factor correction was included.
Microsomal Assays for DGAT Activity-Cultures of S. cerevisiae strains INVSc1 and H1266 carrying pYES2CT, pYES2EaPCR5.1, or pYES2At2.2 were grown in SC medium lacking uracil and supplemented with 1% raffinose and 2% galactose. At the beginning stationary phase, 100-ml cultures were centrifuged (ϳ0.5 g of yeast pellet), and microsomes were isolated as described by Dahlqvist et al. (16) with some minor modifications. The yeast pellet was resuspended in 4 ml of ice-cold buffer (20 mM Tris-HCl, pH 7.9, 10 mM MgCl 2 , 1 mM EDTA, 5% (v/v) glycerol, 1 mM dithiothreitol, and 0.3 M ammonium sulfate) and vortexed vigorously with 2-ml glass beads for 5 min. The resulting suspension was centrifuged at 1,500 ϫ g for 15 min at 6°C. The supernatant was subsequently centrifuged at 100,000 ϫ g for 1.5 h at 6°C. The pellet was resuspended in cold 100 mM potassium phosphate, pH 7.2, to 1.5-3.5 mg of protein/ml, and aliquots were stored at Ϫ80°C. Protein concentrations of the yeast microsomes were determined according to the dye-binding method of Bradford (27) (20 -50 M), and 50 mM potassium phosphate buffer in a total volume of 100 l. The reaction was carried out at 30°C for 15 min. The reaction mixture was immediately quenched in hot isopropyl alcohol, and lipids were extracted with hexane/isopropyl alcohol (25). Total lipids were analyzed on K6 Silica TLC plates developed with hexane/diethyl ether/acetic acid, 80:20:1 (v/v/v). Labeled bands were located and quantified using an Instant Imager. To identify the putative [ 14 C]ac-TAG fraction, it was recovered from the silica, and reanalyzed by C18 reverse phase TLC and by argentation TLC. For reverse phase TLC, KC18 plates (Whatman) were developed three times in methanol. Argentation TLC plates were prepared by impregnation of K6 silica plates with 15% (w/v) silver nitrate in acetonitrile, activation at 110°C for 1 h, and triacylglycerols separated by multiple developments in toluene.
DGAT Assays with Euonymus Seed Extracts-Frozen embryo and endosperm tissue was homogenized in 2 volumes of chilled buffer containing 0.3 M sucrose, 10 mM NaF, 5 mM MgCl 2 , 2 mM dithiothreitol, 1 mM EDTA, and 40 mM HEPES-NaOH, pH 7.4, and filtered through two layers of Miracloth. The residue was re-homogenized in 2 more volumes of buffer and filtered. The filtrates were combined and constitute the cell-free homogenate. The cell-free homogenate was frozen and stored at Ϫ70°C until used for enzyme assays and contained 12-17 mg of protein/ml. The acetyltransferase assay contained [1-14 C]acetyl-CoA (100 M, 200,000 dpm) plus 140 l of homogenization buffer in a total volume of 200 l. 1,2-Dihexanoyl-sn-glycerol was added as 1 l of ethanol solution. To initiate the assay, 20 l of cell-free homogenate was added. The reaction was run at 25°C for 15 min and then terminated, and lipids were extracted as described above. Long-chain acyltransferase assays contained [1-14 C]palmitoyl-CoA and 20 -40 l of cell-free homogenate and were run for 30 min. Normal phase TLC of the labeled lipid products was performed as described above. For subsequent reversed phase TLC analysis, the neutral lipids were eluted from the silica with hexane/isopropyl alcohol. The recovered lipids were analyzed on KC18F reversed phase TLC plates with multiple development in methanol/isopropyl alcohol, 4:1 (v/v). A 3-acetyl-dihexanoin standard was prepared by reacting 1,2-dihexanoyl-sn-glycerol in pyridine:acetic anhydride (1:1, v/v) overnight at room temperature.

DGAT Activity in Cell-free Extracts of Developing Euonymus
Seeds-Developing seeds of E. alatus deposit large amounts of ac-TAG as their principal storage lipid (1). There is no report on the enzyme activity responsible for this sn-3 acetylation reaction. Labeled lipids from incubation of cell-free homogenates prepared from developing E. alatus seed tissues with [ 14 C]acetyl-CoA were analyzed by TLC as shown in Fig. 1. In the absence of exogenous diacylglycerol (DAG), normal phase TLC showed a major labeled band (Fig. 1, lane 2) that co-eluted with endogenous ac-TAG. When this labeled band was recovered and analyzed by C18 reversed phase TLC, the radioactivity migrated with the mass bands corresponding to the major ac-TAG molecular species (Fig.  1, lane 4). To confirm further the reaction as a 1,2-diacyl-snglcyerol:acyl-CoA acyltransferase reaction, a unique exogenous DAG, 1,2-dihexanoyl-sn-glycerol, was added to the assays, resulting in a novel band that co-chromatographed with the synthetic 3-acetyl-1,2-dihexanoin standard in both normal (Fig. 1, lanes 1  and 3) and reverse phase TLC systems (lanes 5 and 6). The acetyltransferase assay gave a linear initial rate of incorporation into ac-TAG up to 15 min, and activity was proportional to protein up to 20 l of extract (0.3 mg of protein) (data not shown). Extract boiled for 5 min was inactive. There was no measurable lag phase before the label appears in the ac-TAG product, indicating that no detectable [ 14 C]acetyl-lipid intermediate was formed. Thus the activity was defined as a DGAT by the incorporation of the labeled acetyl group from acetyl-CoA, by the acetyl incorporation kinetics, and by the use of a distinctive acyl acceptor. When the homogenate was assayed for long-chain TAG synthesis at 5-60 M palmitoyl-CoA concentrations, the lc-DGAT activity was about half of the acetyltransferase-specific activity measured at the same acetyl-CoA concentration.
Identification of a DGAT1 Gene in Euonymus-Total RNA was isolated from developing E. alata seeds at early to midmaturation, a period during which the accumulation of acetyl glycerides reaches a maximum, 2 and thus mRNA levels for the responsible biosynthetic enzyme(s) would be high. An RT-PCR gene-cloning strategy was used for the isolation of putative TAG synthesis genes. Several classes of TAG-synthesizing genes/proteins have been described, with the DGAT1 family being the first to have been identified. Because nulls for the single DGAT1 gene present in A. thaliana result in the reduction of seed oil content by 35-45% (10,11), the DGAT1 gene product contributes substantially to seed oil content. As other TAG-synthesizing enzymes might compensate for the loss of the DGAT1 activity, it is likely that in Arabidopsis the DGAT1 gene could contribute much more than half of the total oil accumulation in the wild type. Thus our strategy was to isolate a putative DGAT1 gene from E. alatus and investigate whether it would be involved in 1,2-diacyl-3-acetin production.  left-hand panel (lanes 1-3). The ac-TAG bands, C18/ C18, or C16/C18 from the endogenous diacylglycerols or C6/C6 from dihexanoin were isolated and re-analyzed by C18 reversed phase TLC, as shown on the right-hand panel (lanes 4 -6). In lanes 4 and 6, the We searched for a DGAT1-like gene in Euonymus using degenerate primers that were made on the basis of well conserved regions of several DGAT1 proteins. These primers were used for RT-PCR of total RNA from Euonymus developing seeds. A small fragment of 330 bp was obtained, showing a high degree of similarity with the DGAT1 genes. 3Ј-and 5Ј-RACE, using gene-specific primers, yielded a 1.5-kb fragment that very likely encompassed the full open reading frame. The deduced amino acid sequence is highly similar to all DGAT1 proteins described so far in plants (Fig. 2), with 50.7% identity and 91% similarity to the Arabidopsis DGAT1 protein. The only region of the putative Euonymus DGAT1 protein that is greatly different from the other DGAT1 proteins is the Nterminal end (93 amino acids). DGAT1 enzymes are membrane-associated proteins. The predicted transmembrane regions and putative substrate-binding site and active site residues described by others are present in the Euonymus DGAT1 gene product (8 -11, 28).
Lipid Analysis of Wild Type Yeast Transformed with Euonymus or Arabidopsis DGAT1 Genes-In order to analyze the function of the DGAT1 gene of Euonymus, the cDNA was cloned into the yeast vector pYES2CT and expressed in wild type S. cerevisiae cells (strain INVSc1) under inducible conditions. As controls, the yeast was also transformed with the empty pYES2CT vector and the pYES2CT vector containing the A. thaliana DGAT1 gene. Oil extracted from mature Arabidopsis seeds contains only 0.01 Ϯ 0.003% ac-TAG (data not shown), and thus any triacylglycerol-synthesizing gene from Arabidopsis would not be expected to have acetyltransferase activity. Total lipids extracted from yeast cells were analyzed semi-quantitatively by TLC with iodine staining. Triacylglycerols, including 1,2-diacyl-3-acetins, were analyzed quantitatively by the addition of odd-chain internal standards, namely acetyl-1,2-dipentadecanoin (for ac-TAG) and triheptadecanoin (for lc-TAG), followed by hydrogenation, TLC separation, and GC. Cultures of all the transformed yeast lines were sampled from early logarithmic to late stationary phase, and TLC analysis showed optimum accumulation of TAG at early stationary phase. This is in agreement with the observations on wild type yeast (13,29). There was no significant difference in the growth rates of the yeast lines. Thus the full TAG analyses were conducted at 48 h after culture inoculation. We also tested if the supplementation of acetate would increase the lc-TAG or ac-TAG production. Acetate (5 mM) was found to marginally increase TAG synthesis and was included in the medium because it might cause higher acetyl-CoA pools in the yeast cells.
Total lipids were extracted and hydrogenated, and lc-TAG and ac-TAG fractions were isolated by preparative TLC. Chromatograms of the acetyl glyceride fraction from the empty vector and Euonymus DGAT1-transformed yeast are shown in Fig. 3. The major peak is the ␣-acetyl-␣,␤-dipentadecanoin internal standard (C32, not counting the glycerol carbon atoms). Three additional peaks are observed only for the Euonymus DGAT1-transformed yeast and correspond to C34, C36, and C38 ac-TAG, having retention times identical to synthetic standards. The sample was then analyzed by GC-MS. Fig. 4 shows the mass spectra for the C32 (internal standard) and C36 peaks from Fig. 3. The mass spectrum of the C32 internal standard shows diagnostic ions at m/z ϭ 225 and 341 atomic mass units, corresponding to CH 3 (CH 2 ) 13 CO ϩ and (M-CH 3 (CH 2 ) 13 COO) ϩ ions. The mass spectrum of the C36 peak shows diagnostic ions at m/z ϭ 239, 267, 355, and 383 atomic mass units, which are assigned as CH 3 (CH 2 ) 14 CO ϩ , CH 3 (CH 2 ) 16 CO ϩ , (M-CH 3 (CH 2 ) 16 COO) ϩ , and (M-CH 3 (CH 2 ) 18 COO) ϩ ions, respectively. The molecular weight is given by (RCO ϩ (M-RCOO) ϩ 16) atomic mass units. Thus the molecular weight is given by (239 ϩ 383 ϩ 16) or (267 ϩ 355 ϩ 16) ϭ 638. In addition there is a small peak at m/z ϭ 578 corresponding to (M-HOAc) ϩ , and the "fingerprint" regions below m/z 200 atomic mass units match very closely for the standard and the C36 peak. Thus the structure of this particular peak is unambiguously acetyl-palmitoyl-stearoyl-glycerol. Likewise, the C34 and C38 peaks shown in Fig. 3, panel 2, are acetyldipalmitin and acetyldistearin. The acyl distribution of the hydrogenated ac-TAG shown in Fig. 3 is a reflection of the fatty acid composition of yeast, which is predominantly a mixture of 16:0, 16:1, 18:0, and 18:1 and which will therefore be expected to produce C34, C36, and C38 acetylglycerides. Integration of the GC peaks gave 49, 43, and 8%, respectively, of ac-TAG species.
The results of the expression of the Arabidopsis and Euonymus DGAT1 genes on lipid accumulation are shown in Table I. The Arabidopsis and Euonymus DGAT1 genes provided 2.7and 6.4-fold increases in total fatty acids, respectively, relative to the empty vector control, and increases in lc-TAG of 5.1-and 19.8-fold, respectively. Thus the expression of DGAT1 genes is not simply diverting acyl groups into TAG but is stimulating net fatty acid accumulation. Most or all of the increase in fatty acid content can be ascribed to the increase in TAG content. The amount of lc-TAG as a fraction of the total lipids, measured as fatty acid content, is 87% in the Arabidopsis DGAT1-transformed yeast line. This is greater than the TAG content reported for wild type S. cerevisiae (10 -50%) (30), and for yeast transformed with the LRO1 gene (66%) (16). The expression of

FIG. 4. Electron impact mass spectra for acetyl glycerides.
A is the acetyl-1,2-dipentadecanoin (C32) internal standard; B is the putative C36 acetyl glyceride peak (retention time ϭ 24.4. min) from GC analysis of the hydrogenated acetyl glyceride fraction isolated from yeast transformed with vector pYES2CT containing the Euonymus DGAT1 gene (see Fig. 3).
the Arabidopsis DGAT1 gene in yeast also resulted in the appearance of very low levels of acetyl glycerides. However, the acetyl glyceride content to total triacylglycerols was 3-fold higher with the Euonymus gene, being 0.26%, whereas with the Arabidopsis gene it was 0.09%. The addition of acetate into the yeast culture had only a small increasing effect on the synthesis of the ac-TAG (data not shown).
DGAT Activity in Wild Type Yeast Expressing the Euonymus or Arabidopsis DGAT1 Genes-Microsomal fractions prepared from wild type yeasts expressing the Euonymus DGAT1 gene, the Arabidopsis DGAT1 gene, or the empty vector were assayed for DGAT activity using [ 14 C]acetyl-CoA or [ 14 C]oleoyl-CoA. Product analysis by TLC is shown in Figs. 5 and 6, and quantification of the relative specific activities is shown in Fig. 7.
When microsomes were incubated with [ 14 C]acetyl-CoA, only the yeast line transformed with the Euonymus DGAT1 gene showed significant accumulation of a labeled product that comigrates with the authentic 1,2-diacyl-3-acetin standard on silica TLC (Fig. 5). When this band was recovered and reanalyzed on two totally different TLC systems, its chromatographic behavior was completely consistent with its identification as [ 14 C]ac-TAG. C18 reverse-phase TLC (Fig. 6A) shows three labeled bands and the assignment of the probable molecular species. The lower band co-migrates with 16:0/18:1 and 18:1/18:1 acetyl glyceride standards, and the spacing of the bands is consistent with two carbon or one double bond differences in acyl composition. On silver nitrate TLC, the putative [ 14 C]ac-TAG sample eluted with monoenoic and dienoic AcTAG standards, as expected (Fig. 6B). This product analysis shows that the cloned Euonymus DGAT1 gene, expressed in yeast, has acetyltransferase activity that results in the expected ac-TAG products. When microsomal assays were conducted with [ 14 C]oleoyl-CoA as substrate, long-chain TAG was a major labeled product, along with oleic acid (it has a similar R f value to ac-TAG). Inspection of Fig. 5 suggests that there are not large differences in long-chain DGAT activity between the three yeast lines.
The DGAT-specific activities from yeast microsomes in the above experiments are compared in Fig. 7. The results for the acetyltransferase activity are clear; only the Euonymus DGAT1 gene produced significant activity and with an order of magnitude comparable with long-chain DGAT activities. The yeast transformed with the Euonymus DGAT1 gene showed about a  40-fold increase in acetyltransferase activity versus the yeast transformed with the Arabidopsis DGAT1 gene. The long-chain DGAT activity we report in the control yeast ( Fig. 7; 180 pmol/ min/mg microsomal protein) is similar to that reported for wild type yeast (340 Ϯ 40 pmol/min/mg microsomal protein) by Oelkers et al. (31). However, the relative specific activities for long-chain DGAT (Fig. 7) for the three yeast lines (Euonymus DGAT1 Ͼ control Ͼ Arabidopsis DGAT1) bear no relationship to the TAG accumulation observed (Table I; Arabidopsis DGAT1 Ͼ Ͼ Euonymus DGAT1 Ͼ Ͼ control). The control DGAT activity comes largely from the DGA1 gene in yeast, which is a member of the DGAT2 gene family and is the dominant TAG synthesizing gene at the onset of stationary phase (31).

TAG Content and DGAT Activity Analyses of TAG-null Yeasts Transformed with Euonymus or Arabidopsis DGAT1
Genes-A triple mutant yeast line, H1266, deleted in the DGA1, LRO1, and ARE2 genes (13), was also transformed with the plant genes or empty vector control. The DGA1 gene encodes a DGAT2 protein that is largely responsible for TAG synthesis during stationary phase, whereas the LRO1 gene encodes a phospholipid:diacylglycerol acyltransferase that contributes significantly to TAG synthesis during logarithmic phase growth. The acyl-CoA:sterol acyltransferase encoded by ARE2 has residual DGAT activity. When all three genes were inactivated, the yeast cell cannot synthesize TAG nor do microsomes exhibit any DGAT activity. TAG accumulation and microsomal DGAT activities with acetyl-CoA and palmitoyl-CoA substrates were determined (Table II) in transformed H1266 lines. With the much reduced endogenous background in H1266, the long-chain DGAT activity of both plant genes becomes clearly apparent. In this background the acetyltransferase activity of the Euonymus DGAT1 gene is still much higher than the Arabidopsis gene, but the long-chain DGAT activities are severalfold higher, primarily because the substrate in these assays was palmitoyl-CoA. Again, the Euonymus DGAT1 gene gives a (2-fold) greater microsomal long-chain DGAT activity, but lc-TAG accumulation is much higher (12-fold) in the yeast line transformed with the Arabidopsis gene. DISCUSSION Our enzymology experiments with cell-free extracts of developing E. alatus seeds demonstrate an acetyl-CoA-utilizing DGAT activity (Fig. 1). The DGAT1 gene we identified from developing E. alatus seeds clearly has substantial acetyl-CoA transferase activity, as demonstrated by the microsomal assays in transformed yeast and in contrast to yeast transformed with the Arabidopsis DGAT1 gene as a control, where the activity is almost undetectable. The variant acyltransferase we have identified is striking because it involves the complete removal of the hydrophobic domain of the substrate molecule. To our knowledge, this is the first report of a gene from any source encoding an acyltransferase that retains its long-chain acyl-CoA acyltransferase activity and yet can also function with an acyl substrate with no hydrophobic domain.
Euonymus seed oil contains about 95% 1,2-diacyl-3-acetins, whereas Arabidopsis seed oil contains essentially none. The DGAT1 gene with acetyl-CoA transferase activity we describe is likely involved in acetyl glyceride accumulation in developing Euonymus seeds. However, the fact that Euonymus is a genetically uncharacterized plant, even to its ploidy level, would allow for other DGAT1 isoforms, and thus the DGAT1 gene may not be the only gene required for expression of the ac-TAG phenotype. By measuring bulk tissue concentrations of acetyl-CoA in various seeds, we have ruled out that the primary cause of the ac-TAG phenotype may be a very high acetyl-CoA pool concentration (32). However, a channeled supply of acetyl-CoA, perhaps from an association of ATP-citrate lyase with the DGAT1 protein, might not result in a detectable change in the bulk acetyl-CoA concentration.
The DGAT1 gene identified from developing E. alatus seeds has substantial acetyltransferase activity in the transformed yeast (Fig. 7) and also results in accumulation of low but measurable amounts of the expected acetyl glyceride product (Table I). However, the expression of this gene does not result in as strong an acetyl glyceride phenotype as might be expected from consideration of enzyme activity alone. This result is similar to a number of other studies where expression of genes involved in the synthesis of unusual fatty acids has yielded only low levels of the expected product (33)(34)(35). In this study the limitation is clearly not the result of low expression of the transgene. By comparing the expression of Euonymus and Arabidopsis DGAT1 genes, which have 51% identity and 91% similarity, it is clear that the yeast long-chain microsomal DGAT activity does not predict the level of lc-TAG accumulation. This is the case whether the genes are expressed in wild type (Table  I and Fig. 7) or in a TAG-synthesizing null line of yeast (Table  II). In both cell lines, expression of the Euonymus DGAT resulted in higher microsomal DGAT activity, whereas expression of the Arabidopsis DGAT gave 4-and 12-fold higher TAG accumulations for wild type and H1266 backgrounds, respectively. Euonymus seed extracts also contain similar acetyl-CoA and long-chain acyl-CoA DGAT activity, but in the seed the latter activity is barely represented in terms of end-product accumulation (1). Thus, TAG accumulation in both yeast and Euonymus reflects a higher degree of complexity than indicated by in vitro enzyme activity assays. Although catalyzing a simple acyl transfer reaction, DGAT1 is a very large integral membrane protein, having 60-kDa subunits with 5-7 transmembrane domains and likely existing as a tetramer (36,37), and by inference it may have some sophisticated regulatory mechanisms associated with its structure.
With the demonstrated lack of correlation between longchain DGAT activity and lc-TAG accumulation in yeast, we should hardly be surprised if the acetyltransferase activity of the Euonymus DGAT1 gene product is not reflected in a large accumulation of ac-TAG product. One simple interpretation is that the gene we have cloned produces an enzyme with both acetyl and long-chain acyltransferase activity and can switch between either specificity depending on conditions. Such conditions might include metabolite concentrations or association with other proteins. This interpretation would fit the observation that expression of the Euonymus DGAT1 protein in yeast causes large increases in lc-TAG. It would be assumed that the conditions in vivo favor the long-chain DGAT activity. In developing Euonymus seeds the broad specificity DGAT1 may function as an acetyltransferase because of localized conditions in the membrane or perhaps because of an as yet unidentified additional gene product. It needs to be stressed that, as is usually the case with assay of microsomal acyl-CoA-dependent enzymes of lipid synthesis, even at fairly low acyl-CoA assay concentrations, most of the acyl-CoA will sequester into the membranes, giving a higher molar ratio of acyl-CoA to membrane lipid than is expected in vivo. In vivo, most of the endogenous acyl-CoA pool will be bound to acyl-CoA-binding protein (38). In our long-chain DGAT1 assay at an oleoyl-CoA concentration of 50 M and containing on average 10 g of microsomal protein, there is an ϳ1:2 acyl-CoA to membrane lipid ratio in the assay. This compares with a bulk cellular ratio estimated at about 1:30, using a value for the long-chain acyl-CoA concentration in yeast cells of 1.5-2 nmol/10 9 cells (39), and our estimate of total phospholipids at 40 -60 nmol/10 9 cells. With such microsomal assays, the cooperativity of substrate utilization might be hidden, acyl-CoA substrate inhibition might occur, or factors that regulate activity in vivo might be lost during extract preparation. DGAT enzyme activities with distinct specificities and/or selectivities have been described in assays with microsomes isolated from various developing seeds where the seed oils have unique compositions (18 -20), but none of the genes corresponding to such activities has yet been cloned, expressed in a heterologous system, and the enzyme activity characterized. The acetyltransferase activity of the Euonymus DGAT1 demonstrates that a fatty acyl chain is not an obligatory requirement for catalysis, although the enzyme has high long-chain acyltransferase activity. Whether long-chain acyl-CoA and acetyl-CoA compete for the same binding site on the enzyme is unknown. The site(s) in the Euonymus DGAT1 protein that is responsible for the acetyltransferase specificity is unknown. There are two regions where the Euonymus and Arabidopsis DGAT1 amino acid sequences diverge significantly, at the N terminus (residues 1-82) and at residues 243-268 in the Euonymus gene. All the plant DGAT1 proteins are divergent in these two regions (Fig. 1), so speculation about the cause of the specificity change is not possible. We note that there are 15 single amino acid changes unique to the Euonymus sequence alone in the conserved region and that the residues 97-114 have been identified previously as a putative acyl-CoA-binding site. Understanding the origin of the unusual specificity in the Euonymus DGAT1 gene is one of our future research goals. Also, the ability to use a hydrophilic substrate in the ongoing kinetic characterization may be a useful asset in unraveling the regulatory complexity of the DGAT1 proteins. Plant expression studies are also underway to determine whether the Euonymus DGAT1 gene expressed in another oilseed can substantially alter the TAG composition.