Identification of Two Novel Human Acyl-CoA Wax Alcohol Acyltransferases

The esterification of alcohols such as sterols, diacylglycerols, and monoacylglycerols with fatty acids represents the formation of both storage and cytoprotective molecules. Conversely, the overproduction of these molecules is associated with several disease pathologies, including atherosclerosis and obesity. The human acyl-CoA:diacylglycerol acyltransferase (DGAT) 2 gene superfamily comprises seven members, four of which have been previously implicated in the synthesis of di- or triacylglycerol. The remaining 3 members comprise an X-linked locus and have not been characterized. We describe here the expression of DGAT2 and the three X-linked genes in Saccharomyces cerevisiae strains virtually devoid of neutral lipids. All four gene products mediate the synthesis of triacylglycerol; however, two of the X-linked genes act as acyl-CoA wax alcohol acyltransferases (AWAT 1 and 2) that predominantly esterify long chain (wax) alcohols with acyl-CoA-derived fatty acids to produce wax esters. AWAT1 and AWAT2 have very distinct substrate preferences in terms of alcohol chain length and fatty acyl saturation. The enzymes are expressed in many human tissues but predominate in skin. In situ hybridizations demonstrate a differentiation-specific expression pattern within the human sebaceous gland for the two AWAT genes, consistent with a significant role in the composition of sebum.

In eukaryotes, the cytoplasmic storage of fatty acids in the form of triacylglycerol (TAG) 1 serves to provide reservoirs for membrane formation and maintenance, lipoprotein trafficking, detoxification of fatty acid and alcohol substrates, epidermal integrity, and fuel in times of stress or nutrient deprivation (reviewed in Refs. [1][2][3]. By contrast, the subcellular and extracellular accumulation of TAG and other neutral lipids has been linked to several human disease states such as diabetes mellitus (4), obesity (4), atherosclerosis (5,6), and non-alcoholic fatty liver disease (7). Understanding the metabolic pathways of triglyceride (TG) synthesis and the roles of the enzymes involved in these reactions may hasten the development of therapeutic interventions by clarifying the pathophysiological processes of these diseases.
Diacylglycerol is esterified to triglyceride by an acyl-CoA: diacylglycerol acyltransferase (DGAT) reaction. There are at least two independent mammalian enzymes known to catalyze this reaction, DGAT1 (8) and DGAT2 (9,10). DGAT1 is a member of the acyl-CoA:cholesterol acyltransferase (ACAT) gene family with high levels of expression in human small intestine, colon, testis, and skeletal muscle. Mice lacking DGAT1 are surprisingly healthy with normal serum TAG but are resistant to diet-induced obesity, and have impaired sebaceous gland secretion (11)(12)(13)(14). Subsequent to the discovery of DGAT1, DGAT2, the original member of a second human DGAT family, was identified by sequence similarity to lipid droplet proteins purified from Mortierella ramanniana, an oleaginous fungus (10). When expressed in insect cells, DGAT2 produces robust DGAT activity. DGAT2 shares no sequence similarity with DGAT1 and exhibits widespread expression in humans, with particularly high levels in liver and adipose tissue. Recently it has been shown that DGAT2 Ϫ/Ϫ mice are severely depleted of triglycerides in their tissues and plasma, and possess poor skin barrier function, leading to early death (15). DGAT1 was unable to fully compensate for the loss of DGAT2, suggesting different roles for the two enzymes, and that DGAT2 is the enzyme responsible for the majority of TAG synthesis in mice.
In Saccharomyces cerevisiae, DGA1 is the sole member of the DGAT2 gene family and is responsible for a large portion of triglyceride synthesis in this model organism (16 -18). This activity is supplemented by that of LRO1 (an ortholog of mammalian lecithin cholesterol acyltransferase), which also esterifies diacylglycerol but uses phospholipids as the acyl donor (19,20). While their relative contribution varies with culture con-ditions, together these enzymes are responsible for about 98% of triglyceride synthesis in yeast. Deletion of ARE2, a steryl ester synthase, in conjunction with DGA1 and LRO1 eliminates all detectable TAG synthesis in yeast (17). As described here, we have utilized this "triple" deletion strain to study mammalian mediators of triglyceride metabolism such as the DGAT2 gene family.
The human DGAT2 gene family consists of seven members: DGAT2 (10), 3 acyl-CoA monoacylglycerol acyltransferases (MGATs 1, 2, and 3; Refs. [21][22][23], and the three genes characterized by our study. Based on expression patterns, MGAT2 and MGAT3 were proposed as the major mediators of the intestinal MAG pathway. This pathway predominates in the intestine, whereby 2-monoacylglycerol, a product of partial lipolysis of triglyceride, is re-esterified using a fatty acyl-CoA to produce diacylglycerol. Moreover, the substrate preference of MGAT3 (49% sequence identity with human DGAT2) for 2-monoacylglycerol, compared with MGAT1 and MGAT2, which primarily use the sn-1 and sn-3 stereoisomers of MAG, implicates it as the major MGAT involved in intestinal fat absorption. Interestingly; however, there is no mouse ortholog of MGAT3. Murine MGAT2 is 46% identical to DGAT2 and is predominantly expressed in small intestine; however, human MGAT2 is more widespread (22, 24 -26). The MGATs also exhibit some DGAT activity and so the precise physiologic roles of these members of the DGAT2 gene family remain to be determined. Two other DGAT2 family members were originally reported (human DGAT candidate (hDC) genes 3 and 4, Ref. 10), and in this report we describe the final member of the human family, which we temporarily name DGA2. We have isolated and expressed cDNAs for these three genes and defined the substrate specificity, tissue expression patterns, and genealogic relationships of the enzymes, relative to human DGAT2. Markedly, all three enzymes are capable of esterifying DAG. However, the preferred alcohol substrates for two of them are long chain alcohols leading to the production of wax esters. We therefore rename these enzymes acyl-CoA wax alcohol acyltransferase (AWAT) 1 and 2.

EXPERIMENTAL PROCEDURES
General-Molecular biology and yeast procedures were performed according to conventional protocols (27,28). Complete (yeast extract, peptone, dextrose; YPD), synthetic complete (SC), and selective media (with galactose for cDNA expression) were prepared as described (27). DNA-modifying reagents were purchased from New England Biolabs. Oligonucleotide primers were synthesized by Invitrogen. Human cDNA obtained as part of a Quick Screen cDNA panel of human tissues was obtained from Clontech (BD Biosciences). Skin cDNA was obtained from Invitrogen. IMAGE clones were obtained from Invitrogen. The Qiaquick Gel Extraction Kit (Qiagen) and the Wizard PCR Preps DNA Purification System (Promega) were used to purify PCR products. The Prime-It random priming probe synthesis kit was obtained from Stratagene. [ 32 P]dCTP, [9,  Identification, Isolation, and Construction of Expression Plasmids with Human DGA2, hDC3, and hDC4 -Human expressed sequence tags, genomic sequences, and predicted mRNA sequences for DGA2, hDC4, and hDC3 were obtained through a comprehensive BLAST search (29,30) through their homology to human DGAT2 (GenBank TM accession number AF384161). Sequence comparisons were made across all currently available genome data sets at NCBI, Ensembl, UCSC, and Celera. The resultant sequences were then used to predict the coding sequence of the candidate genes, analyzed, and compared using DNAStrider (31) and the ClustalW program in MacVector (32). Protein sequences were analyzed using the PROSITE and SMART analysis programs (33) (34). The full-length coding sequence for DGAT2 and hDC3 (GenBank TM accession number BG743707) were obtained as IMAGE cDNA clones in the pOTB7 (clone ID 4644380) and pCMV-SPORT6 (clone ID 4778300) vectors, respectively, and sequenced. Expression plasmids for the full-length human DGAT2 and hDC3 were engineered by subcloning the cDNA fragments into the EcoRI and XmaI, and EcoRI sites, respectively, of the yeast galactose inducible vector, pRS423-GP, downstream of the GAL 1/10 promoter (35). Using predicted cDNA-flanking regions, oligo primer pairs (Supplementary  Table I) were designed to amplify the full-length coding region of DGA2 and hDC4 from thymus and lung cDNA, respectively. Amplification was performed with nested PCR using 2 ng of human cDNA, 10ϫ PCR buffer, 1.75 mM MgCl 2 , 1.5 units of Taq polymerase, (TaqDNA polymerase, Invitrogen), 1 M oligo primers, and 0.4 units of Pfu DNA polymerase (Stratagene) in the Gene Amp PCR System 2400 (Applied Biosystems). PCR reactions were held at 94°C for 5 min followed by 28 cycles of 94°C for 15 s, 55°C for 30 s, and 72°C for 1.5 min. PCR products were gel-isolated, purified, TA-cloned into the EcoRI and NotI sites of the pCR 2.1-TOPO vector (Invitrogen), and transformed into chemically competent Top10 Escherichia coli (Rapid One Shot Chemical Transformation, Invitrogen). The full-length hDC4 was then subcloned into the EcoRI site of the pRS423-GP vector downstream of the GAL 1/10 promoter. The full-length DGA2 was subcloned into the EcoRI and NotI sites of the pRS423-GP vector, also downstream to the GAL 1/10 promoter. The full-length cDNA of DGA2 and hDC4 was completely sequenced.
Expression of DGA2, hDC3, and hDC4 -RNA was prepared from transformed yeast strains grown in SC-His ϩ 2% galactose ϩ 1% raffinose or SC-His ϩ 2% dextrose (control). 10 g of RNA from each culture was resolved on a 1.2% agarose, formaldehyde gel, transferred to nylon membrane (Hybond-N, Amersham Biosciences), and hybridized in Quik-Hyb buffer (Stratagene) at 65°C for 1 h with a random hexamer-primed, [ 32 P]dCTP-labeled gene specific probe, generated with PCR (Table I). Molecular weight markers were supplied by Clontech.
Lipid Synthesis-Metabolic pulse labeling in exponential phase was performed as previously described (17,19). A dilute inoculation (1:20 -1:30) of a saturated culture grown in SC-His ϩ 2% dextrose, followed by washes and resuspension in SC-His ϩ 2% galactose ϩ 1% raffinose, was grown at 30°C to early logarithmic phase (A 600 ϭ 0.3-0.4). For in vivo DGAT assays, 10 ml of cells were added to 0.01 Ci/ml of [ 3 H] or [ 14 C]oleate or 0.25 Ci/ml of [ 3 H]palmitate and incubated at 30°C for 1 h with shaking, washed twice with 0.5% tergitol and once with dH 2 O, and lyophilized. Lipids were extracted from dried cell pellets by cell wall hydrolysis (lyticase) and organic extraction (37) and were resolved by TLC in petroleum ether/diethyl ether/acetic acid (84:15:1) and the two solvent resolution system as described (38). For in vivo AWAT assays, cells were incubated with 70 nmol of hexadecanol, 30 nmol of [ 14 C]hexadecanol, and 100 nmol of oleate. Lipids were resolved in hexane for 30 min followed by air drying for 15 min and subsequent development in toluene for 30 min. Each lane was cut according to lipid standards and counted by liquid scintillation. Assays were performed on three independent isolates of each genotype. Statistical analysis was performed using Student's t tests.
Lipid Accumulation-Assessment of the steady-state accumulation of lipids was performed by labeling a dilute inoculation (ϳ1:1000 of a saturated culture) of each strain into 10 ml of SC-His ϩ 2% galactose ϩ 1% raffinose with 0.01 Ci/ml [ 3 H] or [ 14 C]oleate or 0.25 Ci/ml [ 3 H]palmitate and overnight growth at 30°C. Cells were washed, and lipids were extracted and analyzed as described above.
In Vitro Assay of DGAT and AWAT Activity and Substrate Specificity in Yeast Microsomes-A dilute inoculation of each strain into 500 ml of SC-His ϩ 2% galactose ϩ 1% raffinose was grown overnight at 30°C into log phase. Cells were washed and lysed, and microsomes were prepared from a 100,000 ϫ g spin, as previously described (39). Protein concentrations were determined (40) and diacylglycerol esterification assays were performed as described previously (17,19,41). In brief, microsomes (100 g) were preincubated for 5 min at 37°in 100 mM Tris, pH 7.5 containing 0.25 M sucrose, 1 mM EDTA, 10 mM MgCl 2 , 20 M fatty acid free bovine serum albumin, 160 M dioleoylglycerol (in 5 l acetone, final volume 200 l). The reaction was initiated by the addition of [ 14 C]oleoyl-CoA (10 -20,000 dpm/nmol, 40 M) and stopped after 10 min by the addition of 4 ml of chloroform/methanol (2:1, v/v). Carrier TAG (10 g) and internal standard ([ 3 H]triolein, 30,000 dpm) were added followed by 0.8 ml of water to separate phases. The lower chlo-roform layer containing the lipids was removed, evaporated to dryness, resuspended in 50 l of chloroform, and individual lipid classes separated on Varian chromatography paper using a hexane/diethyl ether/ acetic acid (170:30:1, v/v) mobile phase. The spot corresponding to TG was cut out and counted.
Similar assays were performed in triplicate for AWAT activity using at least three independent yeast microsome preparations. Initially (Table II), conditions were similar to those used for DGAT activities (i.e. 100 g microsomes, assayed for 20 min in a final volume of 250 l) with the substitution of 200 M cetyl alcohol (unlabeled or 14 C-labeled 10,000 dpm/nmol) for DAG. After separation by TLC, the spot corresponding to wax ester was cut out and counted. The AWAT assays were subsequently optimized to be linear with time and so that the reaction velocity was proportional to the amount of protein (Table III). For AWAT1 (DGA2), microsomes (10 g) were preincubated for 5 min at 30°C in 100 mM Tris, pH 7.5 containing 0.25 M sucrose, 1 mM EDTA, 20 mM MgCl 2 , 20 M fatty acid-free bovine serum albumin, 200 M wax alcohol in 2 l of acetone). The reaction (final volume 250 l) was initiated by the addition of [ 14 C]oleoyl-CoA (40 M) and stopped after 10 min by the addition of chloroform/methanol. Cholesteryl ester (15 g) and [ 3 H]cholesteryl ester (30,000 dpm) were added as carrier and internal standard, respectively and lipids extracted and separated as described above. Cholesteryl ester and wax ester have the same solubility in the mobile phase. The assay for AWAT2 (hDC4) was performed as AWAT1 except there was no preincubation and the assay length was 20 min. In some AWAT assays [ 14 C]cetyl alcohol was used to follow esterification. Kill reactions and vector controls were included in all experiments. To assess substrate specificities, the various wax alcohols and acyl-CoAs were substituted for cetyl alcohol or oleoyl-CoA at the concentrations specified above. Activity is presented as pmol/min/mg protein.
Glycerol acyltransferase activity was assayed according to the procedure of Lee et al. (42) using [ 14 C]oleoyl-CoA and 5% glycerol. Monoacylglycerol acyltransferase (MGAT) was assayed as previously described (23) Table I) designed to amplify a portion of the gene of interest. The initial PCR was held at 94°C for 5 min followed by 28 -30 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 1.5 min. The second PCR was conducted under similar conditions using 10% of the initial PCR product, 1 M internal oligo primers, and a 72°C extension time ranging from 40 s to 1 min for 28 -30 cycles. Tissue expression for MGAT3 was determined using the above components held at 94°C for 5 min followed by 35 cycles of 94°C for 15 s, 53°C for 1 min, and 72°C for 2 min. PCR products were resolved on a 1-2% agarose, 0.5 g/ml ethidium bromide gel. Splice variants for hDC4 and MGAT1 were gel-isolated, purified, and sequenced. First strand cDNA synthesis was performed on mRNA from human adipose tissue (kindly provided by A. Ferrante, Columbia University, NY), with and without reverse transcriptase (SuperScriptII, Invitrogen), using 1 g of RNA. 10% of the product was used as the template in the initial reaction of a nested PCR (as described above) designed to amplify a portion of each of the DGAT2 gene family members. Primers were similar to those utilized in experiments above except for the 2nd reaction using MGAT2 and MGAT3 (see Supplementary Table I). Expression of the DGAT2 gene family in skin was determined via standard and nested PCR as described above, using 2 ng of human skin cDNA and primers specific to each gene (Supplementary Table I).
In situ hybridization was performed on freshly cut human scalp frozen sections using antisense probes for DGA2, hDC3, hDC4 and counterpart sense probe controls using standard procedures. Genespecific PCR primer pairs described in Supplementary Table I were used to amplify templates from the pRS423-GP vector harboring each cDNA insert. The resulting PCR products (ϳ1 kb for DGA2, and ϳ0.5 kb for hDC3 and hDC4) were ligated to the pCRII dual promoter vector (T7 and SP6) using the TA-cloning kit (Invitrogen). After sequencing to determine the orientation of the insert, the vector was linearized using either BamHI (T7, sense) or XhoI (SP6, antisense). Subsequently, in vitro transcription and DIG-labeling reactions were carried out following the manufacturer's instructions (Roche Applied Science). Finally, the DIG-labeled RNA probes were ethanol-precipitated, quantified, and tested for alkaline phosphatase activity on nitrocellulose membranes to determine the optimal concentration for both sense and antisense probes. For hybridization, the slides were fixed in 4% paraformalde-hyde/1ϫ phosphate-buffered saline, washed in phosphate-buffered saline, followed by acetylation using acetic anhydride and subsequent washing in phosphate-buffered saline. After prehybridization in hybridization buffer (Roche Applied Science) for 2 h at 58°C, the probes were added and the slides incubated overnight at the same temperature in a humidified chamber. The next day, slides were washed at different stringencies of SSC, treated with RNase A, blocked in blocking buffer, and then incubated in anti-DIG (1:500) overnight at 4°C. On the third day, the slides were washed, and the signal detected using NBT/BCIP substrate. After 24 h, the reaction was stopped, and slides were dehydrated and mounted. Pictures were taken on a regular compound light microscope equipped with a camera.

RESULTS
The Human DGAT2 Gene Family, Conserved Motifs and Evolutionary Relationships-There are six paralogous genes to human DGAT2 whose chromosomal locations based on the human genome databases at NCBI, Ensembl, UCSC, and Celera are listed in Table I. In addition to DGAT2 and MGATs 1-3, we used the previously termed hDC3 and hDC4 (human DGAT candidate genes 3 and 4), and DGA2 (diacyl-glycerol acyltransferases) as acronyms for the uncharacterized family members, upon the assumption that these genes direct the synthesis of triglyceride. The predicted polypeptides of these six genes exhibit 27-34% and 46 -51% sequence identity to the primordial yeast DGA1 and human DGAT2 proteins, respectively, with marked conservation toward the C terminus of the protein (Fig. 1A). All seven human sequences and the yeast Dga1p are predicted transmembrane proteins (Fig. 1A). An iterative algorithm (PSI-BLAST, Ref. 30), suggests these gene products to be derived from an ancestor with lysophosphatidic acid acyltransferase (LPAAT) or glycerol-3-phosphate acyltransferases (GPAT) activity. These enzymes and its relatives mediate the final steps in phosphatidic acid biosynthesis. This conservation includes those residues implicated by site-directed mutagenesis as forming the active sites of bacterial GPAT (43).
The MGAT Arm of the DGAT2 Gene Family-The MGAT branch of the DGAT2 gene family has been extensively characterized (21)(22)(23)(24)(25)(26), primarily in murine tissues. We performed nested PCRs using cDNA from a variety of tissues to confirm and further discern the tissue expression of human MGATs 1-3 ( Fig. 2A). Similar to mouse MGAT1, the human paralog is transcribed in stomach, uterus, kidney, adipose, and liver. However, unlike mice, human brain, lung, thymus, prostate, testes, colon, and notably, small intestine had significant expression. Two splice variants of human MGAT1 exist, and the larger of the two is the predominant transcript found in thymus and testes. Based on sequence analysis, the larger transcript results from the splicing out of the third exon of MGAT1 and an alternatively spliced-in portion of intron 4; this introduces an early stop codon at base pair 312, truncating the protein. The  1. A, sequence alignments of the human DGAT2 family. Amino acid sequences of the DGAT2 family were aligned using the ClustalW multiple sequence alignment program. Dark and light shading indicate identity and similarity of residues, respectively. Conservation of putative active site residues based on mutagenesis of GPATs and LPAATs are indicated by an asterisk (*). Putative transmembrane domains are indicated by underline. B, dendrogram depicting the evolutionary relationship of the human DGAT2 gene family. A phylogenetic tree indicating the relatedness of DGAT2 family sequences was constructed using the ClustalW program. DGA2 (GenBank TM accession number AY947638), hDC3 (GenBank TM accession number AL357752), and hDC4 (GenBank TM accession number AY605053) appear to be derived from hDGAT2 and are more closely related to each other than to the MGAT subfamily. C, genomic structure spanning DGA2, hDC3, and hDC4 on the human X-chromosome. The representation is based on physical data from the NCBI, Ensembl, UCSC, and Celera genome datasets. Shaded boxes depict exons. The locus of three genes spans about 200 kbp with a significant density of predicted genes in the intergenic regions. Extended intergenic regions are indicated by the line breaks. ATG represents the likely translational starts, with the direction of transcription indicated by the arrows. smaller transcript is a product of exon 3 being spliced out, also predicting a truncated protein due to an early stop at base pair 294.
Human MGAT2 transcripts were identified in liver, prostate, small intestine, colon, and adipose (Fig. 2, A and C), confirming previous findings (22,26). In contrast to mMGAT2 (24,25) and a previous report on hMGAT2 (22), no expression was found in kidney tissue. Consistent with previous reports (22,26), splice variants were also exhibited by MGAT2, especially in adipose tissue.
Extensive searching of the available murine databases indicates that the MGAT3 gene is not present in mice. It has been previously reported that hMGAT3 expression is restricted to the gastrointestinal tract and the liver (23). In our studies ( Fig.  2A) the full-length hMGAT3 is found in liver, small intestine, and colon. However, splice variants of hMGAT3 are ubiquitously found in many human tissues. The larger of these splice variants results from the splicing out of exon 5, leading to a truncated protein because of the creation of an early stop codon at base pair 774.
The functional significance of the human MGAT1, MGAT2, and MGAT3 splice variants remains to be determined. MGAT activity has been considered a reaction that predominates in the intestine but also in liver and fat, at least in rodents (44). The expression patterns of this branch of the human DGAT2 family are inconsistent with this concept. However, if the aforementioned conservation with the LPAAT enzymes is significant, then the majority of these splice variants would be anticipated to be inactive because of premature truncation. This may implicate these proteins as having alternate activities or regulatory roles in neutral lipid metabolism. For example, why is the larger MGAT1 message the sole transcript found in thymus and testes? Interestingly, in humans, all of the MGATs, including the full-length transcripts, are expressed in adipose tissue (Fig. 2C).
DGAT2 and the Sex-linked Arm of the DGAT2 Gene Family-Human and murine DGAT2 have been characterized extensively (9, 10). By contrast the X-linked hDC4-hDC3-DGA2 cluster (Xq13.1) of this gene family has not been investigated previously with regard to function or expression. We therefore focused our efforts on defining the role of this subfamily in neutral lipid biosynthesis. All three genes, DGA2 (amino acids 34 -328), hDC3 (amino acids 38 -337), and hDC4 (amino acids 38 -333), possess DAGAT (diacylglycerol acyltransferase) domains that essentially are regions of amino acid sequence similarity common to the DGAT2 family. DGA2 is previously undescribed and comprises seven exons covering 6.1 kb on the direct strand, ϳ190-kb downstream from hDC4 and 29-kb downstream from hDC3. It encodes a 329-amino acid protein with 51% amino acid identity to hDGAT2 and a predicted molecular mass of 37.9 kDa. The initial 44 amino acids of DGA2 are predicted to be a signal peptide. DGA2 has one potential N-glycosylation site and one potential protein kinase C phosphorylation site. Interestingly, DGA2 possesses a region with remote similarity to the IIGF (insulin-like and insulin growth factor) domain, which belongs to a family of proteins that include insulin-related growth factors.
hDC4 (10), encodes a 334-amino acid protein with extensive similarity to the phosphate acyltransferase domains (PlsC, COG0204) of GPAT (amino acids 37-226), and LPAAT (amino acids 103-227). hDC4 has a predicted molecular mass of 38.2 kDa, and is 48% identical to hDGAT2 and 51% identical to DGA2 (Table I). The initial 35 amino acids of hDC4 are predicted to be a signal peptide and the remaining sequence predicts two transmembrane regions (amino acids 39 -58 and 127-149) as well as one potential N-glycosylation site, one potential tyrosine sulfation site, and two potential protein kinase C phosphorylation sites. hDC4 covers 8.5 kb on the reverse strand of the X chromosome and comprises seven exons. hDC3 (10) encodes a 338-amino acid protein also with a phosphate acyltransferase domain similar to the GPAT (amino acids 45-233) and LPAAT (amino acids 115-227) family of acyltransferases. hDC3 is 50% identical to hDGAT2, 51% identical to hDC4, and 51% identical to DGA2. It has a predicted molecular mass of 38.7 kDa and possesses one transmembrane region (amino acids 21-43) in addition to one potential Nglycosylation site and three potential protein kinase C phosphorylation sites. hDC4 is comprised of seven exons covering 28.22 kb on the direct strand of the X chromosome, ϳ130-kb downstream from the DGA3 gene.
To examine the expression of DGA2, hDC3, and hDC4 in humans, nested PCR was performed using cDNA generated from a variety of tissues (Fig. 2B). DGA2 is expressed in all tissues except the spleen; interestingly, the strongest bands were found in thymus, prostate, and testes. hDC4 is expressed in all tissues surveyed except the placenta whereas hDC3 is  Table I). Splice variants for MGAT1 were gel-isolated, purified, and sequenced. B, analysis of human DGA2, hDC3, and hDC4 tissue expression was performed via nested PCR using the same cDNA tissue panel and method as above. The splice variant of hDC4 was isolated, purified, and sequenced. C, nested RT-PCR was performed on mRNA using primers specific for each gene (Supplementary Table I) (Fig. 2C).
Expression of DGAT2 in Yeast Cells Deficient in TG Synthesis-To examine the expression and biochemical activity of DGAT2, we expressed its cDNA in a diacylglycerol esterification-deficient yeast strain (SCY2056, (17) (Fig. 4, A and B) and to a lesser but significant (p Ͻ 0.05) extent, diacylglycerol synthesis (not shown) (21). It has been previously shown that yeast deficient in triglyceride and steryl ester synthesis display minimal cytoplasmic lipid droplets when stained with the vital stain Nile Red and followed by fluorescence microscopy (17). Upon expression of human DGAT2, these strains exhibit a marked accumulation of cytoplasmic neutral lipid droplets (not shown). In addition, microsomes from null TAG background strains of yeast, transformed with an expression vector harboring no insert or the cDNA insert for DGAT2 were assayed in vitro for the incorporation of [ 14 C]oleate and DAG into triglyceride. As shown in Table II, in an in vitro microsomal assay, DGAT2 forms triglyceride at a rate of 65.4 pmol/min/mg protein.
Expression of the Sex-linked Arm of the DGAT2 Gene Family in Yeast-The previous experiments with human DGAT2 demonstrate the utility of this heterologous system in the analysis of mammalian triglyceride synthesis as was demonstrated previously for sterol esterification by members of the ACAT gene family (35,45). We thus performed similar investigations expressing the cDNA of DGA2, hDC3, and hDC4 in the null TAG background strains of yeast. Northern blot hybridization analysis confirmed the high level expression of DGA2, hDC3, and hDC4 transcripts of the predicted size (Fig. 3).
During pulse and steady state [ 3 H]/[ 14 C]oleate and [ 3 H]palmitate metabolic labeling, DGA2, hDC3, and hDC4 exhibit significant triglyceride production, compared with vectortransformed control (VC) strains (Fig. 4, A and B). Notably, hDC3 demonstrated a marked incorporation of [ 3 H]]oleate into TAG. Microsomes from null TAG background strains of yeast, transformed with an expression vector containing DGA2, hDC3, and hDC4 were assayed in vitro for the incorporation of [ 14 C]oleate and DAG into triglyceride. In each case significant synthesis of TAG was detected (Table II).
Substrate Specificity of Sex-linked Arm of the DGAT2 Gene Family-Although DGA2, hDC3, and hDC4 were capable of synthesizing TG in transformed yeast (Fig. 4, A and B) and demonstrate in vitro DGAT activity (Table II), the low specific activity, particularly compared with DGAT2, suggests that DAG is not their primary substrate. Under the conditions used here we demonstrated a significant accumulation of diacylglycerol in strains expressing DGAT2, consistent with the known MGAT activity of this enzyme. However, there was no detectable production of diacylglycerol, phospholipid or cholesteryl ester by DGA2, hDC3, or hDC4 above background strains (not shown). Similarly in vitro assays of microsomes (23,42) prepared from these strains, did not indicate any increase in esterification of MAG or glycerol above vector control (not shown). This prompted us to search for other alcohols as substrates. Recently a novel bifunctional wax ester synthase/acyl-CoA:diacylglycerol acyltransferase that shares no detectable similarity with the DGAT2 gene family was discovered in Acinetobacter calcoaceticus (46). This, in combination with the fact that a mammalian AWAT sequence was unknown at the time of this study, led us to assay DGA2, hDC3 and hDC4 strains for this activity. Initial experiments were performed using cetyl alcohol (C16) and oleoyl-CoA as substrates in reaction conditions identical to those used for DGAT assays.
Both DGA2 (31.7 pmol/min/mg) and hDC4 (1066 pmol/min/ mg) but not DGAT2 or hDC3 demonstrated significant AWAT activity using [ 14 C]oleoyl-CoA as a monitor for wax ester formation (Table II). Wax esters co-migrate with sterol esters on thin layer chromatography. Although it is unlikely that the activity observed is due to sterol ester formation (yeast sterols are not preferred substrates for mammalian enzymes, Refs. 35,47), a similar in vitro assay was performed using [ 14 C]cetyl alcohol rather than radiolabeled oleoyl-CoA. Again, significant AWAT activity was observed with DGA2 (113 pmol/min/mg) and hDC4 (1232 pmol/min/mg) demonstrating that wax ester, not sterol ester, is the enzymatic product and leading us to rename these proteins AWAT1 and AWAT2, respectively.
The substrate specificities of AWAT1 and AWAT2 were then investigated in detail in assays that were optimized so that the reaction velocity was proportional to the amount of microsomal protein. As shown (Table III, A) the alcohol substrate specificities of the two enzymes are strikingly dissimilar. Using oleoyl-CoA as the acyl donor, AWAT1 (DGA2) has a definite preference for decyl alcohol (C10), with less activity using C16 and C18 unsaturated alcohols. AWAT1 utilizes arachidyl alcohol about 20% as well as decyl alcohol, demonstrating its relatively poor activity using saturated long chain alcohols (C16, C18,

TABLE II
In vitro acyltransferase assays Microsomes from TG-null yeast strains transformed with human DGAT2, DGA2, hDC3, or hDC4 were incubated in the presence of dioleoylglycerol or cetyl alcohol and ͓ 14 C͔oleoyl-CoA and assayed for esterification as described in "Experimental Procedures." Data are expressed as pmol of neutral lipid (TAG or Wax ester) formed per min/mg protein as a mean Ϯ S.D. DGAT activities are from three replicate experiments, and AWAT activities are from a representative experiment performed in triplicate. Background levels to the assay from strains transformed with vector control have been subtracted. DGAT activity values for all strains and AWAT activity from DGA2 and hDC4 were significantly elevated relative to the vector control (not shown).  3. Northern blot analysis of human DGA2, hDC3, and hDC4 expression in TG-null strains of yeast. 10 g of RNA from cultures of transformed TG-null yeast strains grown in either dextrose (D) or galactose (G) was resolved on a 1.2% agarose, formaldehyde gel, transferred to nylon membrane, and hybridized in Quik-Hyb buffer at 65°C for 1 h with a random hexamer-primed, [ 32 P]dCTP-labeled genespecific probe. After washing in 0.1 ϫ SSC ϩ 0.1% SDS at 65°C for 30 min, the membrane was exposed to X-Ray film. Molecular weight markers were supplied by Clontech. The predicted length of DGA2 is 987 bp, hDC3-1014 bp, and hDC4 -1002 bp. and C20). In contrast, AWAT2 exhibited no activity using decyl alcohol and significantly preferred the C16 and C18 alcohols. AWAT1 and AWAT2 also showed a difference in acyl-CoA preference ( Table 3, B). Using cetyl alcohol as the acyl acceptor, AWAT1 shows a strong preference toward the saturated acyl group; it uses oleoyl-CoA (C18:1) only 40% as well as stearoyl-CoA (C18:0). In contrast, AWAT2 demonstrates significant activity using all four acyl-CoAs and utilizes unsaturated acyl-CoAs twice as well as saturated acyl-CoAs under the conditions employed.

Expression of AWAT1, AWAT2, and hDC3 in Human
Skin-In addition to several other neutral lipids such as TAG, wax esters are major components of sebum, a production of the sebaceous gland. We therefore tested human skin cDNA preparations for the expression of these genes and discovered that transcripts from all seven members of the DGAT2 family were detectable, with particularly high levels of expression of the X-linked subfamily (Fig. 5A). To identify the cell types involved we performed in situ hybridizations of human skin sections with sense and antisense probes for AWAT1, AWAT2, and hDC3 (Fig. 5B). Control sense probes did not result in section staining. Expression of AWAT1 and AWAT2 was clearly limited to the sebaceous gland, with AWAT2 primarily restricted to the cytoplasm of undifferentiated peripheral sebocytes. AWAT1 was expressed in more mature, centrally located cells just before their rupture and sebum release. hDC3 transcripts were also localized to the cells of the sebaceous gland, however the predominant staining corresponded to nuclei, raising the possibility that the gene is transcribed but not translated in sebocytes.

DISCUSSION
Neutral lipids such as TAG and SE are commonly sequestered in the core of cytoplasmic lipid droplets until such time when hydrolysis allows their return to the metabolic fray. This efficient process of energy storage and release is widely used in nature to facilitate such diverse processes as hibernation in animals, embryo development in eggs and plant seeds, and perhaps sporulation in yeast. Neutral lipid biosynthesis therefore represents an advantageous response to overabundance and thus potential toxicity of membrane lipids and energy. Given this pivotal role in cellular metabolism, it is perhaps not surprising to discover so many distinct pathways (PDAT-, DGAT1-and DGAT2-mediated, to date) and genes for the production of TG. The complexity of the large gene family associated with the DGAT2 reaction is perhaps more surprising and likely reflects alternate substrate specificities. For example, although MGATs 1-3 appear to be genealogically closely re-lated to DGAT2, they differ markedly in their substrate specificity and expression patterns. Similarly, we demonstrate here that the previously uncharacterized X-linked members of this family have only modest DGAT activity in vivo and in vitro. Instead, we demonstrate that two members of this subdivision of the DGAT2 gene family are responsible for the generation of acyl esters of long chain alcohols (fatty alcohols) in the synthesis of wax esters. Interestingly, plant orthologs of DGAT2 have been reported to mediate the synthesis of both TG and wax esters (49). Recently, the murine ortholog of AWAT2 has been demonstrated to also perform this reaction (50). Based on sequence conservation, the three X-linked members of this 7-component family are closer to DGAT2 than to the MGAT subfamily (Fig. 1B). Thus it is possible that the final member of this gene family, hDC3 is a bona fide DGAT enzyme. The enzyme was expressed in yeast and clearly conferred the ability to synthesize significant levels of triacylglycerol, particularly with oleate as a substrate. Alternatively, hDC3 may be an orphan waiting for an activity and substrate to be ascribed. Interestingly, the hDC3 mRNA appears to remain within the nucleus in in situ hybridizations of skin sections.
The chromosomal location of these genes may implicate them as candidates for several human syndromes ranging from obesity to skin disorders. For example, the hDC3, hDC4, and DGA2 genes comprise a cluster of 3 physically linked genes  1 and 2 In vitro AWAT assays were performed using microsomal preparations from TG-null yeast strains transformed with human AWAT1 (DGA2), AWAT2 (hDC4), or a vector control harboring no insert. Microsomes were assayed using the optimized reaction conditions described under "Experimental Procedures." Activity is in pmol of wax ester formed per min/mg protein, expressed as a mean Ϯ S.D. from a representative experiment performed in triplicate. Vector control background has been removed. In vitro determination of alcohol and acyl-CoA substrate specificities were performed with ͓ 14 C͔oleoyl-CoA and the indicated alcohols or ͓ 14 C͔cetyl alcohol and the indicated acyl-CoA donors (Table III, 5. Expression of DGA2, hDC3, and hDC4 in human skin. A, analysis of tissue expression of the human DGAT2 gene family in skin was performed via standard PCR using skin cDNA and primers specific for each gene (see Supplementary Table I) as described in "Experimental Procedures." B, in situ hybridization was performed on freshly cut human scalp frozen sections using antisense probes for DGA2, hDC3, and hDC4 and counterpart sense probe controls as described in "Experimental Procedures." Gene-specific PCR primer pairs (Supplementary  Table I) were used to amplify templates from the pRS423-GP vector harboring each cDNA insert.
localizing to the X chromosome, in a contig of ϳ200 kbp, in mice and humans (Fig. 1C). Interestingly there are several undefined sex-linked obesity (e.g. Borjeson-Forssman-Lehmann and Wilson-Turner syndromes, see Ref. 51 for review) and dermatological syndromes (52). Similarly, it is interesting to note that chromosome 11 (specifically 11q13) containing DGAT2, MGAT2, uncoupling protein UCP2, and UCP3 (53,54), and BBS1, the gene most commonly involved in Bardet-Biedl syndrome (55), has been linked to obesity and hyperinsulinemia. Similarly, one genome scan also implicated the same region of 11q13 in childhood and adolescent obesity (56). Moreover, several linkage studies suggest that 7q22.1, a region flanking the leptin gene but also MGAT3, is closely linked to obesity and body mass index in humans (57,58). Variation at these loci is clearly worthy of investigation as causative agents of several disease syndromes. In addition to their roles as safe harbors for toxic fatty acids or alternative form of energy storage, molecules such as wax esters and TAG function as a hydrophobic permeability barrier to limit dehydration from tissue surfaces (59,60). As such, both molecules are significant components of plant cuticle (61), insect exoskeleton coating (62), and mammalian sebum. Indeed, murine DGAT1 and DGAT2 utilize long chain alcohols, in addition to DAG, as substrates (50), and are critical for epidermal integrity (12,15). Wax esters are also the main component of spermaceti that allows for sperm whales to regulate their buoyancy. Whereas the many commercial applications of whale oil prompted sperm whale hunting, purification of wax esters from the jojoba plant currently meets market demands for wax esters. The AWAT reaction thus has both commercial and biological relevance.
Until recently (Ref. 50 and this study), the AWAT reaction was uncharacterized at the molecular level. A novel bifunctional wax ester synthase/acyl-CoA:diacylglycerol acyltransferase that shares no detectable similarity with the DGAT2 gene family and has no human counterpart was discovered in Acinetobacter calcoaceticus (46). Thus, human AWAT1, murine wax ester synthase (50), and its human ortholog, AWAT2, likely represent the most significant contributors to the pool of wax esters in sebum in mammals. Indeed, the tissue and cell type expression pattern for these genes is consistent with a major role for these enzymes in the secretion of sebum by the sebocyte. The most common saturated fatty acids found in sebum wax esters are C16 Ͼ C14 Ͼ C15 whereas palmitoleoyl (C16:1) acid is the most abundant mono-unsaturated fatty acid, which is present at five times the amount of oleic acid (63). The murine wax ester synthase was shown to prefer mono-and polyunsaturated fatty alcohols less than 20 carbons in length (when palmitoyl-CoA is used as the acyl-CoA), and short chain mono-and polyunsaturated fatty acyl-CoAs (when hexadecanol is used as the fatty alcohol), but not monoacylglycerol or diacylglycerol as substrates (50). We found similar fatty acyl-CoA and fatty alcohol preferences for its human ortholog, AWAT2, except that AWAT2 is able to use C16:0 as well as C16:1 fatty alcohols (using oleyl-CoA as the acyl-CoA). We also found that AWAT2 does possess modest but significant DGAT activity but no MGAT activity. Interestingly AWAT1 prefers saturated acyl-CoAs while AWAT2 and its murine ortholog (50) prefer unsaturated fatty acids. The diet appears to have little effect on wax ester composition with linoleic acid, an exclusively dietary fatty acid accounting for Ͻ1% of wax ester fatty acids (64). This would explain why disruption of stearoyl-CoA desaturase (responsible for synthesis of palmitoleic and oleic acids) causes atrophy of sebaceous gland (48).
Strikingly, AWAT1 and 2 are expressed at different stages in the differentiation of sebocytes (Fig. 5B), suggesting a temporal reason for quality control of the wax esters. It is possible that as the sebocyte matures, a cycle of wax ester hydrolysis and reesterification occurs, such that the wax esters are remodeled prior to their secretion. The relevance of this is not clear, however it may serve to homogenize the sebum in terms of fatty acid and alcohol saturation or chain length such that it is optimally hydrophobic and effective as a permeability barrier.