Cloning and Functional Expression of UGT Genes Encoding Sterol Glucosyltransferases from Saccharomyces cerevisiae, Candida albicans, Pichia pastoris, andDictyostelium discoideum *

Sterol glucosides, typical membrane-bound lipids of many eukaryotes, are biosynthesized by a UDP-glucose:sterol glucosyltransferase (EC 2.4.1.173). We cloned genes from three different yeasts and from Dictyostelium discoideum, the deduced amino acid sequences of which all showed similarities with plant sterol glucosyltransferases (Ugt80A1, Ugt80A2). These genes fromSaccharomyces cerevisiae (UGT51 =YLR189C), Pichia pastoris(UGT51B1), Candida albicans(UGT51C1), and Dictyostelium discoideum(ugt52) were expressed in Escherichia coli. In vitro enzyme assays with cell-free extracts of the transgenicE. coli strains showed that the genes encode UDP-glucose:sterol glucosyltransferases which can use different sterols such as cholesterol, sitosterol, and ergosterol as sugar acceptors. AnS. cerevisiae null mutant of UGT51 had lost its ability to synthesize sterol glucoside but exhibited normal growth under various culture conditions. Expression of eitherUGT51 or UGT51B1 in this null mutant under the control of a galactose-induced promoter restored sterol glucoside synthesis in vitro. Lipid extracts of these cells contained a novel glycolipid. This lipid was purified and identified as ergosterol-β-d-glucopyranoside by nuclear magnetic resonance spectroscopy. These data prove that the cloned genes encode sterol-β-d-glucosyltransferases and that sterol glucoside synthesis is an inherent feature of eukaryotic microorganisms.

Sterol glycosides are widespread membrane lipids, occurring in all plants, several algae (1)(2)(3), some fungi (4 -9), slime molds (10 -12), Dictyostelium (13), a few bacteria (14 -19), and even animals (20 -23). The knowledge base for sterol glycosides is rather limited compared with free sterols and sterol esters, where the synthesis, transport, and functions have been studied extensively in animals (24 -29), plants, (30 -35), and yeast (36 -40). The basis for studies on the functions of sterol glycosides is the assumption that free sterols and sterol glycosides differ physiologically. It is obvious that the attachment of a glycosyl moiety to the sterol backbone alters the physical properties of this lipid. As a result, there are changes in the properties of membranes containing different proportions of free sterols and sterol glycosides. Such changes have been studied with artificial membranes in terms of membrane fluidity, permeability, hydration, and phase behavior (41)(42)(43)(44). However, we still do not know how free sterols and sterol glycosides differ physiologically in biological membranes and why many eukaryotic organisms synthesize sterol glycosides. One of the main reasons for our limited knowledge in this field is the lack of a genetic approach. The objective of the present work was the isolation and characterization of sterol glycosyltransferase genes from eukaryotic organisms. We expect that genetic manipulation of these genes will facilitate the elucidation of sterol glycoside functions in these organisms.
The predominating sugar moiety in sterol glycosides is glucose. Besides plants, UDP-glucose:sterol glucosyltransferase activity was determined by in vitro enzyme assays in Saccharomyces cerevisiae (45)(46)(47)(48), Candida bogoriensis (49), other fungi (50), and Physarum polycephalum (51,11). Although the lipid composition of S. cerevisiae was studied in detail during the last decades, there are only rare reports on the actual isolation of sterol glucoside (SG) 1 from this yeast (8,9). The expression level of sterol glucosyltransferase gene(s) seems to be so low that very sensitive radioactive assays are required to detect enzyme activity in vitro. However, expression is too low for accumulation of a significant amount of SG in vivo suitable for lipid analysis. Another possibility is the existence of a sterol glucoside hydrolase that counteracts the sterol glucosyltransferase keeping the amount of SG very low. Such a hydrolase has been detected in plants (52,53) but not in S. cerevisiae.
In contrast to S. cerevisiae, significant amounts of SG were occasionally found in other fungi such as C. bogoriensis (4) and Phytium sylvaticum (5), whereas sterol mannosides were detected in the human pathogen Candida albicans (6,7).
Here we describe the identification and characterization of sterol glucosyltransferases from three different yeasts as well as from Dictyostelium discoideum. Since the genes cloned in this work all belong to a superfamily of UDP-glycosyltransferases (UGTs) (54,55), we use the gene symbol UGT for each new member. The UGT51 gene of S. cerevisiae was identified by a homology probing approach with the cDNA sequences of UDP-glucose:sterol glucosyltransferases from plants that were recently cloned from oat (ugt80A1, 2 GenBank TM accession number Z83832) and Arabidopsis (ugt80A2, GenBank TM accession number Z83833) (56,57). The UGT51 genes from C. albicans, Pichia pastoris, and UGT52 from D. discoideum were cloned by either homology probing or PCR-based strategies. Moreover, we identified ergosterol as an in vivo substrate for the sterol glucosyltransferases from S. cerevisiae and P. pastoris. Furthermore, a new glycolipid was identified in S. cerevisiae as the product of the in vivo action of the yeast sterol glucosyltransferase from P. pastoris.

Media and Growth of Bacteria and Saccharomyces
E. coli was grown in Luria-Bertani broth (LB, Duchefa, Haarlem, The Netherlands). Ampicillin (100 mg/liter) and 1.5% agar (Difco) was included for solid LB media. Yeast synthetic dextrose medium (SD) was prepared according to Ausubel et al. (62).
Cloning of Sterol Glucosyltransferases from S. cerevisiae, C. albicans, P. pastoris, and D. discoideum S. cerevisiae-UGT51 (YLR189C, GenBank TM accession number U17246) is a gene located on chromosome XII of S. cerevisiae. We isolated a genomic 6359-bp NdeI/SpeI DNA fragment containing UGT51 by restriction digestion of cosmid 9470 DNA that we received from the American Type Culture Collection. This fragment was cloned into pBluescript II KS giving the plasmid pUGT51g. This plasmid was used to clone the open reading frame (ORF) UGT51 into expression vectors for E. coli (pET19b 3 pUGT51x) and S. cerevisiae (pGAL4 3 pUGT51xy). In addition, PCR fragments of UGT51 corresponding to polypeptides with N-terminal or C-terminal deletions of amino acids were cloned into the following plasmids: pN690x, pN722x, pN799x, pC1043x, and pN722xy.
C. albicans-PCR with genomic DNA of C. albicans was performed with the primer DW3 5Ј-GSI WCI VGI GGI GAY GTH CAR CC-3Ј and WA3 5Ј-GTI GTI CCI SHI CCI SCR TGR TG-3Ј. The resulting PCR fragment was used to synthesize a digoxigenin-labeled probe for the screening of a fosmid library of genomic DNA of C. albicans strain 1161. 3 Three positive fosmid clones were isolated: 2H6, 3E6, and 8C12. An 8.2-kb HindIII fragment of 2H6 was identified by agarose gel separation of the digestion fragments and hybridization with the probe. This fragment contained the ORF UGT51C1 and was ligated with pUC18/HindIII (3 pUGT51C1g). A PCR fragment corresponding to the ORF UGT51C1 was cloned into an expression vector for E. coli (pBAD-TOPO 3 pUGT51C1x).
P. pastoris-PCR with genomic DNA of P. pastoris was performed with the primer DW5 5Ј-TTY ACI ATG CCI TGG ACI MSI AC-3Ј and DW30 5Ј-YKI GRI SHI GCI SCI GTI GTN CC-3Ј. The resulting PCR fragment was used to synthesize a digoxigenin-labeled probe for the screening of a genomic DNA library of P. pastoris strain GS115 (63,64). An SphI/EcoRV fragment of a positive clone was isolated according to the method described above for Candida. This fragment contained the ORF UGT51B1 and was ligated with pUC19/SphI/SmaI (3 pUGT51B1g). A PCR fragment corresponding to the ORF UGT51B1 was cloned into expression vectors for E. coli (pBAD-TOPO 3 pUGT51B1x) and S. cerevisiae (pYES2 3 pUGT51B1xy).
D. discoideum-We used the cDNA clone FC-AZ07 (GenBank TM Accession numbers C25752 and C25753 from H. Urushihara, University of Tsukuba) 4 from the cDNA library of D. discoideum strain CAX3 for the PCR synthesis of a digoxigenin-labeled probe. With this probe, we screened a -ZAP cDNA library of D. discoideum strain AX4 (from W. F. Loomis, University of California). In vivo excision of a positive clone resulted in the plasmid pUGT52c that had a 3331-bp cDNA insert within the vector pBluescript I SK. A PCR fragment corresponding to the ORF ugt52 was cloned into an expression vector for E. coli (pBAD-TOPO 3 pUGT52x). All plasmids used in this study are listed in Table I.

Expression of Sterol Glucosyltransferases
E. coli-Cells of BL21(DE3) and TOP 10 were transformed with plasmids derived from pET19b and pBAD TOPO, respectively. Cells were grown at 30°C in 25 ml of LB medium containing 100 mg/liter ampicillin to an absorbance between 0.5 and 0.8 at 600 nm. Induction was performed by adding 0.4 mM isopropyl-1-thio-␤-D-galactopyrano- indicate identical amino acids in four or five sequences. The polypeptides of the three yeasts and D. discoideum contain long N-terminal extensions that do not exist in the plant polypeptides (residues 1-820). Within this N-terminal extension there are only a few sequence similarities between the four species. In contrast to this N-terminal region, all sequences show significant similarity in their C-terminal parts (residues 1030 -1425). Within this region four highly conserved boxes are underlined and numbered. The conserved box 4 corresponds to the signature sequence of a protein superfamily of UDP-glycosyltransferases (54). The region named psbd corresponds to a putative steroid-binding domain found in solanidine glucosyltransferase (67) and steroid UDP-glucuronosyltransferases. Clusters of asparagine, glutamine, or threonine in the sequence from D. discoideum are indicated by dotted lines.

Sterol Glucosyltransferase Assay
The assay mixture contained in a total volume of 100 l the following: 80 l of E. coli or S. cerevisiae cell homogenate or membrane fraction, respectively, either 10 l of a solution of 4 mM cholesterol in ethanol (400 M final concentration) or 8 l of 500,000 dpm [4-14 C]cholesterol in ethanol (final concentration 45 M, specific activity 1.85 GBq/mmol), and either 100,000 dpm UDP-[U-14 C]glucose (final concentration 1.5 M, specific activity 10.8 GBq/mmol) or 360 M unlabeled UDP-glucose. After incubating for 2 h at 30°C, the reaction was terminated by the addition of 0.9 ml of 0.45% NaCl solution and 4 ml of chloroform/methanol, 2:1. The extracted lipids were separated by thin layer chromatography with chloroform/methanol (85:15, v/v). The radioactivity on the silica gel plate was detected by radioscanning with a BAS-1000 BioImaging Analyzer (Raytest, Straubenhardt, Germany).

Lipid Extraction and Analysis
S. cerevisiae cells were grown at 30°C on defined, lipid-free media described above. Harvesting of the cells by centrifugation was followed by the extraction of lipids with chloroform/methanol, 1:2, and chloroform/methanol, 2:1. The extracted lipids were separated by thin layer chromatography on silica gel 60 (Merck, Darmstadt, Germany) with chloroform/methanol, 85:15. Glycolipids were visualized with a spray of ␣-naphthol/sulfuric acid mixture. For nuclear magnetic resonance spectroscopy and mass spectrometry, the novel glycolipid was purified by column chromatography on silica gel by elution with acetone. The lipid was acetylated with acetic anhydride in pyridine and subjected to preparative thin layer chromatography in diethyl ether.

Mass Spectrometry
Mass spectrometric analysis of the peracetylated sterol glycoside was carried out with a Hewlett-Packard model 5989 spectrometer using the direct insert probe mode. Temperature was raised from 80 to 325°C at a rate of 30°C/min. EI-mass spectra were recorded at 70 eV with an ion source temperature of 200°C.
The nucleotide sequences of the sterol glucosyltransferases have been deposited in the GenBank TM data base under accession numbers AF091398 (C. albicans), AF098916 (D. discoideum), and AF091397 (P. pastoris). (66) data base search with the amino acid sequences of plant sterol glucosyltransferases revealed similarities to Ugt51 (Ylr189c, PIR: locus S51434) which was described as a hypothetical protein from S. cerevisiae with unknown function. The first aim of the present work was to elucidate the function of this protein. UGT51 contains an open reading frame (ORF) of 3594 bp encoding a polypeptide of 1198 amino acids with a calculated molecular mass of 136 kDa. We isolated a 6359-bp NdeI/SpeI fragment of the cosmid 9470 (accession number U17246) from the chromosome XII of S. cerevisiae that contained UGT51. This fragment was cloned into pBluescript (3 pUGT51c) and used to subclone the ORF UGT51 into expression vectors for E. coli and S. cerevisiae.

Isolation and Cloning of Putative Sterol Glycosyltransferases-A BLAST
Furthermore, we wanted to isolate and clone homologous genes from other organisms. The identification of novel sterol glycosyltransferase genes will provide the basis for genetic manipulation of these genes and contribute to the elucidation of sterol glycoside functions that may differ in these organisms. We cloned the homologous genes by a PCR-based strategy with degenerated oligonucleotide primers that were derived from conserved regions of Ugt51 and the plant sterol glucosyltransferases. PCR products obtained from C. albicans and P. pastoris were sequenced, and the deduced amino acid sequences of these fragments showed similarities with Ugt51 and the plant sterol glucosyltransferases. These PCR products were labeled and used to isolate the complete genes from genomic libraries of these organisms. An 8.2-kb fragment of C. albicans that contained the incomplete MDL1 gene and an unknown ORF was cloned into pUC18 (3 pUGT51C1c). Both strands of the unknown ORF were sequenced, and these data were published in nucleotide data bases under the accession number AF091398. The ORF of 4548 bp, called UGT51C1, encoded a polypeptide of 1516 amino acids with a calculated molecular mass of 171 kDa.
From P. pastoris we cloned and sequenced a genomic fragment of 6777 bp (accession no. AF091397). It contained an ORF (UGT51B1) of 3633 bp encoding a polypeptide of 1211 amino acids with a calculated molecular mass of 136 kDa.
A BLAST search (66) with Ugt51 revealed sequence similarity with a cDNA clone from D. discoideum CAX3 (clone FC-AZ07, accession numbers C25752 and C25753). This clone was used to label a PCR fragment that was suitable for the isolation of longer fragments from a D. discoideum AX4 cDNA library. The longest clone (pUGT52c) had an insert of 3331 bp that was sequenced on both strands (accession number AF098916). It contained an ORF from bp 244 -3312 (ugt52) that encoded a polypeptide of 1023 amino acids with a calculated molecular mass of 114 kDa. Fig. 1 shows a comparison of the deduced amino acid sequences of the plant sterol glucosyltransferases with those of Ugt51p and the gene products of the homologous genes from Pichia, Candida, and Dictyostelium. The polypeptides showed only a few similarities in their N-terminal parts, but within a C-terminal region of about 360 amino acids (residues 1030 -1425) similarities were significant as follows: 69% identity between the three yeasts, 36 -39% between the yeasts and Dictyostelium, 34 -37% identity between the yeasts and the plants, 86% between the plants, and 33% between the plants and Dictyostelium. Near their C-terminal ends all sequences showed a 29-amino acid region of striking homology (box 4 in Fig. 1). This region was very similar to a "signature sequence" that is characteristic for a superfamily of nucleoside diphosphosugar glycosyltransferases and that was suggested to represent a UDP-sugar binding domain (54). For this reason, all open reading frames mentioned in this work were classified as UGTs (UDP-glycosyltransferases) in agreement with the UGT Nomenclature Committee (54) and the Saccharomyces Genome Data base. 5 Most of the members of this superfamily, but not all, use UDP-sugars as substrates. The conserved boxes 1 and 3 of the sequences shown in Fig. 1 are also present in some prokaryotic members of the UGT superfamily, e.g. a rhamnosyltransferase from Pseudomonas aeruginosa (GenBank TM accession number L28170), but are absent in the eukaryotic members of the superfamily. The region named psbd in Fig. 1 corresponds to a putative steroid-binding domain found in a solanidine glucosyltransferase from potato (see Ref. 67, Gen-Bank TM accession number U82367) and mammalian steroid UDP-glucuronosyltransferases (e.g. UGT2B7, GenBank TM accession number J05428). The UGTs cloned in the present work showed only a few similarities with this putative steroid-binding domain. These data suggest that the novel UGTs cloned in the present work were more related to bacterial glycosyltransferases than to other eukaryotic glycosyltransferases, although some of them use structurally similar substrates. The conserved box 2 in Fig. 1 is not present in other members of the UGT superfamily and seemed to be characteristic for sterol glucosyltransferases.

Functional Expression of the UGTs in E. coli and in Vitro Determination of Sterol Glycosyltransferase Activity-The
ORFs of the DNA fragments cloned in this work were expressed in E. coli to examine the function of the expression products. This was done by in vitro determination of sterol glycosyltrans- 5 Saccharomyces Genome Data base available at the following on-line address: http://genome-www.stanford.edu/.

FIG. 3. In vitro activity of recombinant S. cerevisiae sterol glucosyltransferase (Ugt51p) expressed in E. coli as full-length or as Nor C-terminally shortened fragments.
Cell-free homogenates from transgenic E. coli were used for in vitro enzyme assays. The assay mixtures contained UDP-glucose and cholesterol with either radiolabeled glucose (A) or radiolabeled cholesterol (B). The labeled lipophilic reaction products were separated by thin layer chromatography with subsequent analysis by a BAS 1000 BioImaging Analyzer. SG was identified by cochromatography with an authentic standard. The transgenic E. coli harbored the control plasmid pET19b (lanes 1, 1Ј), pUGT51x (encoding the complete polypeptide Ugt51p, lanes 2, 2Ј), pN690x (encoding Ugt51p that lacked 690 N-terminal amino acids, lanes 3, 3Ј), pN722x (encoding Ugt51p that lacked 722 N-terminal amino acids, lanes 4, 4Ј), pN799x (encoding Ugt51p that lacked 799 N-terminal amino acids, lanes 5, 5Ј), and pC1043x (encoding Ugt51p that lacked 155 C-terminal amino acids, lanes 6, 6Ј). ferase activity in cell-free extracts using radiolabeled substrates. E. coli is suitable for such experiments since untransformed cells do not contain a sterol glycosyltransferase activity. The ORF UGT51 from S. cerevisiae was cloned into the expression vector pET19b (pUGT51x). The other UGTs from C. albicans, P. pastoris, and D. discoideum were amplified by PCR and subcloned into the expression vector pBAD-TOPO (pUGT51C1x, pUGT51B1x, and pUGT52x). Transformed and induced E. coli cells were disrupted, and the homogenates or crude membrane fractions were used for in vitro enzyme assays. Membrane fractions had the advantage that most of the intracellular UDP-glucose was eliminated. It turned out that more than 50% of the expressed protein was not in the soluble fraction but was found in the sediment of membranes and inclusion bodies (data not shown).
To measure the sterol glycosyltransferase activity, we performed in vitro assays with various radiolabeled sugar donors (NDP-sugars) and acceptors (sterols). After thin layer chromatography (TLC), the radioactivity in the lipophilic reaction products was determined. Table II shows a comparison of the substrate specificities of the recombinant enzymes from S. cerevisiae, C. albicans, P. pastoris, D. discoideum, and Avena sativa. These data should not be considered as a quantitative determination of substrate affinities, since the assays were not performed under linearized conditions. These results rather suggest which substrates were accepted or discriminated against by each of the heterologously expressed proteins. With UDP-[U-14 C]glucose as the donor, all expressed UGT proteins used various sterols as sugar acceptors, whereas there was no background SG synthesis of a control E. coli homogenate. The UGT proteins glucosylated sterols with a planar backbone such as cholesterol, ergosterol, ␤-sitosterol, stigmasterol, and even the steroidal alkaloid tomatidine. However, other lipids, e.g. ceramide and dioleoyl glycerol, were not accepted. Regarding the sugar moiety of the donor, all enzymes exhibited a distinct specificity for glucose (UDP-glucose), whereas UDP-galactose, UDP-glucuronic acid, UDP-mannose, and GDP-mannose were not accepted. This was of particular relevance in the case of the Ugt51C1 from C. albicans in view of the identification of sterol mannoside in this organism (see "Discussion"). UDP-xylose, as a pyranose, mimics UDP-glucose except for the lacking CH 2 OH moiety. It was incorporated into cholesterol xyloside by the Avena, Saccharomyces, and Pichia enzymes at a low rate. UDP served as the best nucleotide moiety of the glucose donors, but CDP-glucose and GDP-glucose were also used, although at significantly lower rates. TDP-glucose was not accepted. These in vitro data gave strong evidence that all UGT genes cloned in this work encode specific sterol glucosyltransferases using UDP-glucose as sugar donor and different planar sterols as sugar acceptors.
Only a C-terminal Conserved Region of UGT51 Was Essential for Enzyme Activity-An alignment of the different sterol glucosyltransferases ( Fig. 1 and Fig. 2) showed that the sequences shared significant similarities only at their C-terminal parts and that the polypeptides from the three yeasts and from Dictyostelium had large N-terminal extensions that did not exist in the plant polypeptides. This phenomenon raised the question which parts of the polypeptides were necessary for their function. We selected one of the proteins, Ugt51p from S.  cerevisiae, to investigate this phenomenon. Therefore, various fragments of Ugt51p, obtained by deleting different N-or Cterminal sequences, were expressed in E. coli (Fig. 2). The cell-free homogenates of the transgenic E. coli were used for in vitro enzyme assays performed with cholesterol and UDP-glucose. Ugt51p fragments with N-terminal deletions of either 690 or 722 amino acids synthesized SG (Fig. 3). These fragments contained the conserved region indicated in Fig. 2. In contrast, elimination of parts of the conserved region by deletion of either 799 N-terminal or 155 C-terminal amino acids resulted in complete loss of enzyme activity. These data indicate that at least 722 N-terminal amino acids of Ugt51p are not required for in vitro enzyme function. Ugt51p fragments containing the complete conserved region of the sterol glucosyltransferases are sufficient for in vitro SG synthesis.
Deletion and Overexpression of UGT51 in S. cerevisiae-To address the question whether the in vitro synthesis of SG reflects the in vivo function of Ugt51p, we deleted UGT51 in S. cerevisiae strain UTL7A. The null mutant was still viable and grew like the parental strain on complex and minimal media, at low and elevated temperatures, under different conditions of osmotic stress and in the presence of nystatin. Fig. 4 shows that cell-free homogenates of the parental strain UTL7A synthesized low amounts of SG in vitro (1700 dpm of radiolabeled product), whereas the enzyme activity was completely lost in the null mutant. The labeled compound near the front probably was sterol ester. Then we transformed the null mutant with a plasmid that led to the expression of complete UGT51 under the control of a galactose-induced promoter. The in vitro SG synthesis was restored in these cells and was more than 10-fold higher (22,000 dpm of labeled product) than with UTL7A cells (Fig. 4). These data prove that the expression of UGT51 in S. cerevisiae resulted in the biosynthesis of an enzymatically active sterol glucosyltransferase. The expression of the N722 fragment in the null mutant also resulted in a detectable enzyme activity (1500 dpm of labeled product). However, the activity resulting from the expression of the truncated enzyme was much lower than that observed for the expression of the wild-type enzyme.
Isolation of a Novel Glycolipid from S. cerevisiae and Its Identification as Ergosterol-␤-D-Glucopyranoside-The lipids of S. cerevisiae were extracted and separated by thin layer chromatography to address the question whether SG was accumulated in vivo (Fig. 5A). The total lipid extracts of the S. cerevisiae wild-type cells and the null mutant did not contain detectable amounts of SG. Also, expression of the N722 fragment in null mutant cells did not result in the biosynthesis of significant amounts of SG, whereas cells expressing the complete UGT51 contained a new glycolipid that co-chromatographed with an authentic SG standard.
We found that a heterologous expression of UGT51B1 from P. pastoris in the S. cerevisiae knock-out mutant led to the accumulation of larger amounts of the novel glycolipid (Fig. 5B) as compared with the homologous expression of UGT51. We made use of this large accumulation of the glycolipid for its purification from lipid extracts of S. cerevisiae expressing UGT51B1 from P. pastoris. This was done by column chromatography on silica gel, acetylation of the isolated glycolipid, and subsequent thin layer chromatography. The peracetylated glycolipid was subjected to structural analysis by mass spectrometry and nuclear magnetic resonance (NMR) spectroscopy.
Electron impact (EI) mass spectrometry analysis of the peracetylated glycolipid was performed using the direct insert probe mode. Two characteristic fragments were obtained, one (m/z ϭ 331) from a terminal single tetra-O-acetylated hexosyl residue and the other (m/z ϭ 378) from the steroid (Fig. 6). The molecular ion [M Ϫ H] Ϫ (m/z ϭ 726) was found to be in agreement with the calculated mass for tetra-O-acetylhexosylergosterol (C 42 H 62 O 10 ; M r ϭ 726.94).
The structure of the purified tetra-O-acetylhexosyl-ergosterol was further investigated by 1 H NMR spectroscopy at 600 MHz. The 1 H NMR spectrum revealed signals for four O-acetyl groups (␦ 2.011, 1.984, 1.954, and 1.937), indicating one terminal hexose and no additional OAc group derived from a putative OH group of the steroid. The anomeric signal of the hexose (H-1, ␦ 4.551) expressed a coupling constant J 1, 2 of 8 Hz (Table  III), thus showing ␤-configuration. All other hexose ring protons showed high coupling constant values (J 2, 3 9.5, J 3, 4 9.6, and J 4, 5 9.8) characteristic for the gluco configuration in a pyranoside that was further supported by the chemical shift data (Table III). These data are in full agreement with a terminal tetra-O-acetyl ␤-glucopyranose (68).
Characteristic signals for the steroid were found at ␦ 5.501 and 5.312 and were diagnostic for the conjugated diene system in ring B (H-6 and H-7) of ergosterol (69). Signals for H-22 and H-23 at the isolated double bond between C-22 and C-23 (␦ 5.156 and 5.105, respectively) expressed a high coupling constant value of J 22, 23 15.2 Hz, thus indicating E configuration (Fig. 7). Signals for H-3, H-4␣, and H-4␤ were at ␦ 3.531, 2.367, and 2.173, respectively. All other signals could be assigned and part of the coupling constant values determined based on relay COSY and 1 H, 13 C HMQC experiments ( Fig. 7 and Table III).

DISCUSSION
This work reports the cloning, identification, and characterization of novel UDP-glucose:sterol glucosyltransferases from S. cerevisiae, C. albicans, P. pastoris, and D. discoideum. We also isolated a sterol glucoside from transgenic S. cerevisiae expressing UGT51B1 from P. pastoris. This is the first report on the structural analysis of a glycosylated sterol from S. cerevisiae which we identified as ergosterol-␤-D-glucoside. These data prove that UGT51B1 encodes a UDP-glucose:sterol ␤-D-glucosyltransferase. Combined with in vitro data on the substrate specificities of the identified glucosyltransferases from different organisms, we assume that all the recombinant enzymes that we characterized in this work are sterol ␤-Dglucosyltransferases. The enzymes show a narrow specificity for the sugar donor using predominantly UDP-glucose. In contrast, the in vitro specificity for the sterol acceptor is broad, pointing to the biosynthesis of glucosyl sterols with different sterol moieties depending on the individual sterol composition of each organism.
The isolation of novel sterol glucosyltransferases represents one step in our approach to elucidate the functions of sterol glucosides in these organisms. Amino acid sequence comparisons showed that these genes represent a gene family with members in plants, fungi, and other eukaryotic organisms (Fig.  1). This gene family is part of a superfamily of UDP-glycosyltransferases (UGTs) (54), which is characterized by a C-terminal signature amino acid sequence (Fig. 1).
Glucose is the main sugar moiety in the sterol glycosides of many organisms, but other sugars have also been found (1). This raises the question whether there are physiological differences between sterol glycosides having different hydrophilic head groups. Since C. albicans was reported to contain mainly cholesterol mannoside (6,7), it was our intention to isolate the gene encoding the corresponding mannosyltransferase. Our data on the substrate specificity of Ugt51C1p from C. albicans showed that it is not a mannosyltransferase as expected but a glucosyltransferase. The discrepancy between the glucose preference of Ugt51C1p and the presence of mainly sterol mannoside in C. albicans points to the possibility that this organism may contain two sterol glycosyltransferases with different sugar specificities. The ongoing C. albicans genome sequencing project should answer the question whether there is a gene homologous to UGT51C1 which may encode a mannosyltransferase.
The expression of different fragments of UGT51 in E. coli and S. cerevisiae showed that at least 722 N-terminal amino acids of the polypeptide are not required for the in vitro glucosyltransferase activity of the protein (Fig. 3 and Fig. 4). No significant difference in enzyme activity was observed between the wild-type and the truncated protein upon expression in E. coli. Interestingly, upon expression in S. cerevisiae, a significantly lower sterol glycosyltransferase activity was observed for the truncated form (Fig. 4). In line with this observation, expression in S. cerevisiae of the wild-type form but not of the truncated form led to the biosynthesis of detectable amounts of FIG. 7. Two-dimensional 1 H, 13 C HMQC spectrum of 3␤-(2,3,4,6-tetra-O-acetyl-␤-D-glucopyranosyloxy)ergosta-5,7,22E-trien which was isolated in its nonacetylated form from S. cerevisiae expressing UGT51B1 from P. pastoris. The spectrum was measured in a solution of CDCl 3 at 300 K. Assignments of protons to the glycosyl and ergosterol part (in italics) are indicated. The corresponding 1 H NMR spectrum (600 MHz) is displayed along the horizontal (F2) axis and the 13 C NMR spectrum (90.6 MHz) along the vertical (F1) axis. SG (Fig. 5). These observations might suggest that the Nterminal region contains yeast-specific regulatory elements for the activity, expression, or stability of the protein. However, it remains to be investigated what function this large N-terminal extension of about 700 amino acids may have. There are only low sequence similarities between these N-terminal extensions of the three yeasts and D. discoideum and they do not show significant similarity to any sequence from the data bases. They do not contain known membrane targeting motifs nor membrane spanning domains. Interestingly, this N-terminal extension is not present in the plant enzymes (Fig. 1).
By deletion of UGT51 in S. cerevisiae we showed for the first time that SG is non-essential for growth of S. cerevisiae. In addition, no differences in growth were observed under various stress conditions. This is consistent with three genome-wide expression studies of S. cerevisiae which showed no alterations in UGT51 expression during the sporulation process (71), shift from anaerobic to aerobic conditions (72), and when grown in rich or minimal media (73). The expression level of UGT51 seems to be very low, and we do not know whether a low amount of SG is sufficient to fulfill its physiological function or whether larger amounts need to be biosynthesized, e.g. under special growing conditions, that are sufficient for altering membrane properties. Lipid analyses of other yeasts and fungi showed that some of them contain significantly larger amounts of sterol glycosides than Saccharomyces and that an increase in its synthesis was induced under stress conditions. 6 In this context, it should be mentioned that insertional mutagenesis experiments with the rice blast fungus Magnaporthe grisea and a subsequent screening procedure revealed mutants with reduced pathogenicity (74). One of these mutants identifies the gene PTH8 (GenBank TM accession number AF027983), the partial sequence of which shows striking similarity to the sterol glucosyltransferases from yeasts. These data suggest that SG may be involved in the pathogenicity of phytopathogenic fungi.
Another interesting observation concerns some regions of the cDNA of the D. discoideum sterol glucosyltransferase which are rich in AAC codons pointing to clusters of asparagine, glutamine, or threonine (see Fig. 1). Such clusters were reported to be characteristic for developmentally regulated genes (75)(76)(77). A hypothetical change of SG synthesis during development would be consistent with the observations that SG synthesis alters during development of D. discoideum (13) and that SG accumulation was induced by the process of differentiation from amoeboid stage to plasmodium in P. polycephalum (78,51).
In summary, we have shown that SG synthesis is not restricted to plants but that it is also a common feature of yeasts, fungi, and other eukaryotes. We have identified and characterized four novel sterol glucosyltransferases from yeasts and D. discoideum. Our work provides the basis for new approaches to elucidate the functions of SG in these organisms.