Saccharomyces cerevisiae GNA1, an Essential Gene Encoding a Novel Acetyltransferase Involved in UDP-N-acetylglucosamine Synthesis*

The Saccharomyces cerevisiae gene,YFL017C, for a putative acetyltransferase was characterized. Disruption of YFL017C was lethal, leading to a morphology similar to those caused by the depletion ofAGM1 or UAP1, the genes encoding phospho-N-acetylglucosamine mutase and UDP-N-acetylglucosamine pyrophosphorylase, respectively. This implies the involvement of YFL017C in UDP-N-acetylglucosamine synthesis. The recombinant protein for YFL017C displayed phosphoglucosamine acetyltransferase activities in vitro and utilized glucosamine 6-phosphate as the substrate. When incubated with Agm1p and Uap1p, the Yfl017c protein produced UDP-N-acetylglucosamine from glucosamine 6-phosphate. These results indicate that YFL017C specifies glucosamine-6-phosphate acetyltransferase; therefore, the gene was designated GNA1(glucosamine-6-phosphateacetyltransferase). In addition, whereas bacterial phosphoglucosamine acetyltransferase and UDP-N-acetylglucosamine pyrophosphorylase activities are intrinsic in a single polypeptide, they are encoded by distinct essential genes in yeast. When the sequence of ScGna1p was compared with those of other acetyltransferases, Ile97, Glu98, Val102, Gly112, Leu115, Ile116, Phe142, Tyr143, and Gly147 were found to be highly conserved. When alanine was substituted for these amino acids, the enzyme activity for the substituted Phe142 or Tyr143 enzymes was severely diminished. Although the activity of Y143A was too low to perform kinetics, F142A displayed a significantly increased K m value for acetyl-CoA, suggesting that the Phe142 and Tyr143 residues are essential for the catalysis.

karyotes and eukaryotes. In bacteria, UDP-GlcNAc is a major cytoplasmic precursor of the cell wall peptide glycan, and it is also used as the disaccharide moiety of lipid A (4 -6). In eukaryotes, it serves as the substrate of chitin synthase and is also utilized for the GlcNAc moiety of N-linked glycosylation as well as the glycosylphosphatidylinositol-anchor of cellular proteins (7,8).
In Saccharomyces cerevisiae, GFA1 and AGM1 are responsible for the synthesis of GlcN-6-P from Fru-6-P and for the interconversion of N-acetylglucosamine 6-phosphate (Glc-NAc-6-P) and GlcNAc-1-P, respectively (18,19). As Agm1p utilizes phospho-N-acetylglucosamine (phospho-GlcNAc) as the substrate, it has been assumed that acetyl transfer precedes the mutase reaction, and, therefore, GlcN-6-P would serve as the recipient of the acetyl moiety (19). In our previous paper, we reported that the yeast UAP1 gene encodes UDP-GlcNAc pryophosphorylase and demonstrated that, unlike the product of the E. coli GlmU gene, phospho-GlcN acetyltransferase activity is not intrinsic in the yeast Uap1 protein (20). The human UDP-GlcNAc pyrophosphorylase was also identified and shown to be highly related to the yeast enzyme. Curiously, the human enzyme turned out to be identical to the antigen of male infertility patients (20 -22). Furthermore, the human enzyme recognizes both UDP-GlcNAc and UDP-GalNAc, but the deletion of 17 amino acids in the C-terminal region, which occurs physiologically by alternative splicing, alters its substrate specificity (22).
In this paper, we report the S. cerevisiae gene for phospho-GlcN acetyltransferase. The enzyme utilizes GlcN-6-P as the substrate, and its deletion is lethal, leading to a morphology similar to those observed in UAP1-and AGM1-deficient cells. The important residues for the catalysis are proposed on the basis of mutation analysis. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM 1 of ScGNA1 (YFL107C) was amplified by polymerase chain reaction with the S. cerevisiae genomic DNA extracted from strain A451 (MAT␣ can1 aro7 can1 leu2 trp1 ura3) as a template. It was then cloned at the XbaI site of pUC18 or pYEUra3 (Toyobo), generating pUC-ScGNA1 and pYEU-ScGNA1, respectively. Primers used for polymerase chain reaction were 5Ј-AAGGATCCAGCT-TACCCGATGGATTTTATATA-3Ј and 5Ј-AAGAATTCCTATTTTCTA-ATTTGCATTTCCAC-3Ј. The Candida albicans homolog of ScGNA1 was cloned by screening a C. albicans genomic DNA library as a probe using the 0.2-kilobase fragment of the C. albicans genomic DNA, whose sequence was available in the C. albicans data base and was found to be related to YFL017C. The filters were hybridized and washed under stringent conditions (20 mM sodium phosphate (pH 7.2), 5 ϫ SSC (1 ϫ SSC contains 150 mM NaCl and 15 mM sodium citrate), 5 ϫ Denhardt's solution, 0.1% SDS, 25% formamide at 42°C for hybridization; 0.1 ϫ SSC and 0.1% SDS at 50°C for washing). After the third screening, DNA was extracted from bacterial cells that were strongly hybridized with the probe DNA, and the insert DNA was cloned between the BamHI and SalI sites of pUC19 for further plasmid construction. Radiolabeling of the probe DNA was performed by the random priming method using [␣-32 P]dCTP (23), and DNA sequencing was carried out as described elsewhere (23). Construction of the C. albicans genomic DNA library was already reported (24).
Expression and Purification of the Recombinant Proteins-The coding regions of ScGNA1 and CaGNA1 were cloned between the BamHI and EcoRI sites of pGEX-2T (25), and the resulting plasmids were transfected into E. coli JM109 to express the recombinant yeast proteins as a fusion with glutathione S-transferase (GST). Induction and expression of the recombinant Gna1 proteins was carried out with isopropyl-␤-D-thiogalactopyranoside as described (24,25). At 4 h after the addition of isopropyl-␤-D-thiogalactopyranoside, the bacterial cells were harvested, suspended in a buffer containing 20 mM Tris-HCl (pH 7.5), 0.5 mM EDTA, 10%(v/v) glycerol, 1 mM phenylmethylsulfonyl fluoride, and lysed by sonication. After cell debris was removed by centrifugation at 15,000 ϫ g at 4°C for 30 min, GST-Gna1 fusion proteins were purified by glutathione-Sepharose CL-4B column chromatography, as described (25), and analyzed by sodium dodecyl sulfate (SDS)polyacrylamide gel electrophoresis. The primers used for amplifying the CaGNA1 ORF were 5Ј-AAGGATCCTTACCACAAGGTTATACATTCA-GA-3Ј and 5Ј-AAGAATTCCTAGAATCTACATACCATTTCAAC-3Ј.
Assays for Phospho-GlcN Acetyltransferase-Because CoA reacts with 2-nitrobenzoic acid and releases 4-nitrothiophenolate (26,27), an assay for phospho-GlcN acetyltransferase was performed in 50 l of a reaction mixture containing 50 mM Tris-HCl (pH 7.5), 5 mM MgCl 2 , 200 M GlcN-6-P, 200 M Ac-CoA, 10% (v/v) glycerol, and approximately 0.1 g of the recombinant GST-Gna1p fusion protein. After incubation at 30°C for the indicated time, the reaction was terminated by adding 50 l of a solution containing 50 mM Tris-HCl (pH 7.5) and 6.4 M guanidine hydrochloride and then 50 l of a solution containing 50 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 20 M 2-nitrobenzoic acid. The amounts of CoA produced by the Gna1 protein were estimated from that of 4-nitrothiophenolate by measuring the optimal density at 412 nm. The UDP-GlcNAc synthesis assay was carried out in a 20 l standard reaction mixture containing 50 mM Tris-HCl (pH 8.3), 5 mM MgCl 2 , 200 M GlcN-6-P, 200 M Ac-CoA, 10% (v/v) glycerol, 0.1 M [␣-32 P]UTP (specific activity, 1 ϫ 10 3 cpm/pmol) and approximately 0.1 g each of Gna1, Agm1, and Uap1 proteins at 30°C for 10 min. Two microliters of the each reaction mixture were spotted onto polyethyleneimine-cellulose plates, and nucleotide sugars were separated by thin layer chromatography (TLC) in a solution that was prepared by mixing 6 g of Na 2 B 4 O 7 ⅐10H 2 O, 3 g of H 3 BO 3 , and 25 ml of ethylene glycol in 70 ml of H 2 O (28). The radioactive spots were visualized by autoradiography.
To test the ability of CaGNA to complement ScGNA1, the entire ORF of CaGNA1 was cloned at the BamHI site (downstream of the ADH1 promoter) of pGBT9-T (24). Thus, the transcription of CaGNA1 from this plasmid was under the control of the ADH1 promoter. The resulting plasmid was transfected into gna1⌬ cells. After the trp ϩ cells were selected in the presence of galactose, they were transferred to plates containing glucose and further cultured for 3 days.
Site-directed Mutagenesis-A series of the ScGNA1 mutants harboring an alanine substitution for Ile 97 , Glu 98 , Val 102 , Gln 107 , Gly 112 , Leu 115 , Ile 116 , Leu 119 , Phe 142 , Tyr 143 , or Gly 147 were generated by the oligonucleotide-directed dual amber method as described (29) with Mutan-Express Km TM (Takara). The entire ORF of the ScGNA1 gene was cloned between the BamHI and EcoRI sites of pKF18k (Takara) using EcoRI linker and hybridized with oligonucleotides containing the indicated mutations. The resulting mutant ScGNA1 genes were excised from the vector and ligated with pGEX-2T, which had been digested with BamHI and EcoRI. All the mutations were confirmed by sequencing the DNA.

Identification of the Yeast GlcN-6-P Acetyltransferase Gene-
Previously, we demonstrated that unlike the bacterial enzyme, the eukaryotic UDP-GlcNAc pyrophosphorylase is a monofunctional protein with no phospho-GlcN acetyltransferase activity (20). When searching the S. cerevisiae data base, we found YFL017C on chromosome VI encodes a protein with homology to known acetyltransferases. To examine the physiological function of YFL017C, we created an S. cerevisiae strain in which the endogenous YFL017C gene was disrupted, but where episomal copies of YFL017C, whose transcription was under the control of the GAL1 promoter, were maintained. The resulting strain grew on galactose plates, but died on glucose plates, confirming that YFL017C is an essential gene. When cultured on glucose plates, most of the YFL017C-deficient cells swelled and often then lysed. This morphology was quite similar to that caused by the depletion of AGM1 or UAP1 (Fig. 1), suggesting that YFL017C is involved in the synthesis of UDP-GlcNAc. Although the expected product of YFL017C showed only a limited sequence similarity to the C-terminal phospho-GlcN acetyltransferase domain of the bacterial GlmU gene product (15% identity, see below), YFL017C was called GNA1 (glucosamine synthesis acetyltransferase 1) in this study.
If GNA1 is really a gene required for an essential metabolic pathway, such as UDP-GlcNAc synthesis, it should be preserved in other organisms. A search of the C. albicans data base revealed a partial nucleotide sequence that shared sequence similarity with S. cerevisiae GNA1 (ScGNA1). By screening the C. albicans genomic DNA library with the DNA fragment of the above nucleotide sequence as a probe, we were able to clone and sequence the C. albicans homolog of GNA1 (CaGNA1). The predicted products of ScGNA1 and CaGNA1 are closely related to each other (Fig. 2) (44% sequence identity over the entire protein). In addition, when expressed under the control of the ADH1 promoter, CaGNA1 rescued an S. cerevisiae gna1⌬ null mutant (data not shown), demonstrating that the CaGNA1 functionally complements ScGNA1. Furthermore, a search of the Schizosaccharomyces pombe and Caenorhabditis elegans data bases revealed that S. pombe SPAC16E8 and C. elegans B0024-12 encode proteins which share a high degree  of sequence identity with ScGna1p over the entire protein (25.2% identity between ScGna1p and the SPAC16E8 product and 28.9% identity between ScGna1p and the B0024-12 product) (Fig. 2). Therefore, we referred to these genes as S. pombe GNA1 (SpGNA1) and C. elegans GNA1 (CeGNA1), respectively. Although the functionality of SpGNA1 and CeGNA1 remains to be established, all the above results strongly support the idea that GNA1 encodes ubiquitous acetyltransferase necessary for UDP-GlcNAc synthesis.
In UDP-GlcNAc synthesis, there is a process involving acetyltransferase that is the acetylation of phospho-GlcN. Accordingly, we asked whether GNA1 specifies phospho-GlcN acetyltransferase. Both ScGna1 and CaGna1 proteins were expressed in E. coli as a fusion with GST and then purified by affinity column chromatography (Fig. 3A). When incubated with GlcN-6-P and Ac-CoA, GST-ScGna1p and GST-CaGna1p produced CoA, whereas GST alone did not (Fig. 3B). The amounts of CoA released from Ac-CoA reached a plateau within 5 min (Fig. 3B), and GlcN-6-P was preferably utilized as the substrate (Fig. 3C). Although small amounts of CoA were produced when higher concentrations of GlcN-1-P were present (at concentrations higher than 100 M), the K m value for GlcN-1-P was about 25 times higher than that for GlcN-6-P (3.0 mM for GlcN-1-P versus 124 M for GlcN-6-P) (Fig. 3C, Table I). No CoA production was observed when galactosamine 1-phosphate (GalN-1-P) was used as the substrate (Fig. 3C).
Possible Active Sites of ScGna1p-A number of acetyltransferases have been identified in various organisms. When the amino acid sequence of ScGna1p was compared with known and putative acetyltransferases, we found that although the bacterial GlmU protein is not highly related to ScGna1p, some of the amino acids in the two short regions of ScGna1p are highly conserved among several acetyltransferases (Fig. 6). These regions, designated domain I and domain II, encompass the amino acid positions from 97 to 117 and from 142 to 147, respectively. To examine the importance of these domains for catalysis, we mutated the highly conserved amino acids in these regions Ile 97 , Glu 98 , Val 102 , Gly 112 , Leu 115 , Ile 116 , Phe 142 , Tyr 143 , and Gly 147 to alanine. As a control, a nonconserved Leu 119 was also converted to alanine. After confirming the nucleotide sequences, all the mutant enzymes were expressed as a fusion with GST and purified by affinity column chromatography (Fig. 7A). As shown in Fig. 7B, alanine substitutions for Val 102 , Gly 112 , Leu 115 , Ile 116 , Leu 119 , and Gly 147 only weakly impaired the activity. In contrast, those for Phe 142 and Tyr 143 severely diminished the activity. Furthermore, although the activity of Y143A was too low to perform kinetics, alanine substitution for the neighboring amino acid Phe 142 increased the K m values for Ac-CoA and GlcN-6-P by about 10-and 5-fold, respectively (Table I). The activities of I97A and E98A were also affected to a lesser extent (Fig. 7B). Interestingly, these two mutants displayed different features of the binding affinities to the cofactor and the substrate. Mutation of Ile 97 increased the K m value for GlcN-6-P by about 3-fold, whereas that of Glu 98 raised the K m value for Ac-CoA by about 2.5-fold (Table I).
In human spermine/spermidine N 1 -acetyltransferase (HsS-SAT), it was demonstrated that the amino acid sequence motif RGFGIGS beginning at the position of 101 is required for the Ac-CoA binding and that Arg 101 and the proximal glycine loops in this motif are essential for the enzyme activity (30). Moreover, the double mutant R101A/E152K acts as a dominant negative (31). In our sequence alignment, Arg 101 and Gly 106 of HsSSAT correspond to Gln 107 and Gly 112 of ScGna1p, respectively (Fig. 6). As indicated above, G112A restored the significant activity, and other amino acids in the RGFGIGS motif are not highly conserved in Gna1p-related acetyltransferases listed in Table I and Fig. 6. Accordingly, we mutated Gln 107 of ScGna1p to alanine. Though the effect of the mutation of Gln 107 was not as severe as those of Phe 142 and Tyr 143 , the activity and the kcat values of Q107A were significantly lower than that of the wild type, and the K m value for Ac-CoA was about five times higher than that of the wild type enzyme (Fig. 7B, Table I). From these results, it was speculated that Phe 142 and Tyr 143 of ScGna1p are essential residues for the catalysis, and that Ile 97 , Glu 98 , and Gln 107 also locate in the active pocket and support the binding of the cofactor and the substrate. DISCUSSION In this paper, we have identified the yeast GNA1 as the gene for phospho-GlcN acetyltransferase, and demonstrated that unlike the case in bacteria, yeast UDP-GlcNAc pyrophosphorylase and phospho-GlcN acetyltransferase are encoded by distinct genes. Because UAP1, AGM1, GNA1, and GFA1 are all essential genes (18 -20), every step of the UDP-GlcNAc synthesis is essential for the viability of yeast cells. In addition, yeast has only one type of acetylated amino sugar, that is GlcNAc; neither GalNAc nor N-acetylmannosamine is present in yeast cells. Thus, Gna1p is expected to be the sole amino sugar acetyltransferase present in yeast cells.
ScGna1p showed a relatively higher K m value for Ac-CoA (228 M) compared with those of other acetyltransferases (e.g. 3 M for HsSSAT and 3.8 M for choline acetyltransferase) (30 -32). This is also true for the E. coli phospho-GlcN acetyltransferase (320 M) (26). Probably, low affinity to Ac-CoA is unique feature of phospho-GlcN acetyltransferases. The K m value of ScGna1p for GlcN-6-P was also comparatively high (124 M). The enzyme displayed a reasonable substrate specificity to GlcN-6-P, however, because the K m value for GlcN-1-P was about 25 times greater than that for GlcN-6-P. Nevertheless, ScGna1p produced UDP-GlcNAc from GlcN-1-P in the presence of Agm1p, Uap1p, and UTP. Because Uap1p does not utilize GlcNAc-6-P as the substrate (20), the results obtained in this study indicate that Agm1p has a broad substrate specificity and recognizes phospho-GlcN as well. In fact, not only phospho-GlcNAc but also phosphoglucose can be the substrate of Agm1p (19).
Although PAT1, which denotes putative acetyltransferase, is not listed in the yeast data base, a search of the literature revealed that the amino acid sequence of ScGna1p is identical to that of previously reported Pat1p (33). PAT1 was isolated as the gene that suppressed the histone H1-dependent growth of FIG. 7. Effects on ScGna1p activity of alanine substitution for the conserved amino acids. The wild type and the mutant ScGna1 proteins bearing an alanine substitution for each of the amino acids that are highly conserved in several acetyltransferases listed in Fig. 6 were expressed as a fusion with GST and purified with glutathione-Sepharose beads. A, approximately 1-g amounts of the wild type (WT) and the indicated mutant proteins were separated on a 12.5% SDSpolyacrylamide gel and stained with Coomassie Brilliant Blue. The positions of the protein size markers and GST-ScUap1p are indicated in kilodaltons and by the arrowhead, respectively. B, approximately 0.1 g of the purified wild type (WT) and the indicated mutant proteins were incubated with 200 M each of GlcN-6-P and Ac-CoA at 30°C for 10 min, and the amounts of the released CoA that represent the enzyme activities were determined as described in Fig. 3. The activities of the mutant enzymes are indicated as a percent of that of the wild type. the S. cerevisiae YRL4 cells (33). Depletion of Pat1p function results in multiple phenotypical changes. The PAT1-deficient cells displayed enlarged cell volume, increased sensitivity to the microtuble inhibitor, benomyl, aberrant spindle structure, and incompletion of cytokinesis. Furthermore, by fluorescenceactivated cell sorter analysis, it was demonstrated that PAT1 is required for multiple steps in the cell cycle, such as exit from G 0 , progression of DNA synthesis, and mitosis (33). All these facts suggest a physiological link between UDP-GlcNAc synthesis and cell cycle progression. As UDP-GlcNAc is mainly used as the cell wall precursor in yeast, we wonder whether Gna1p has additional substrates, such as core histones, whose acetylation is essential for gene expression and cell cycle progression; however, no histone acetyltransferase activity was detected in the recombinant ScGna1p.
Lin et al. (33) cloned the PAT1/GNA1 homolog from S. pombe by the functional complementation of an S. cerevisiae pat1⌬ null mutation. Although the amino acid sequence of the C-terminal region of S. pombe Pat1p (SpPat1p) is identical to that of SpGna1p, the previously identified SpPat1p has 245 additional amino acids at the N terminus (33). Because we could not find the DNA sequence corresponding to SpPat1p in the S. pombe data base, the presence of the extra N-terminal amino acids in SpPat1p may be unique to a certain S. pombe strain.
Sequence comparisons of the various acetyltransferases revealed that ScGna1p contains two short domains, where several amino acids are highly conserved among several known and putative acetyltransferases. By the mutation study, we found that Phe 142 and Tyr 143 of ScGna1p are crucial amino acids for the catalytic reaction and that Ile 97 , Glu 98 , and Gln 107 also play important roles in the binding of the cofactor and the substrate. As mentioned before, arginine at the position of 101 and the proximal glycine loops of HsSSAT have been implicated to be the active site of the enzyme (30,31). Although Arg 101 of HsSSAT is converted to glutamine in the yeast Gna1 proteins, our results indicate that Gln 107 of ScGna1p is also involved in the Ac-CoA binding. In contrast, alanine substitution for the highly conserved Gly 112 , which corresponds to the third glycine in the RGFGIGS motif, had little effect on the activity of ScGna1p, and other glycines in the RGFGIGS motif are not highly conserved in ScGna1p and its homologs. Recently, the three-dimensional structures of the yeast histone acetyltransferase (Hat1p) and Serratia marcescens aminoglycoside 3-N-acetyltransferase, both of which belong to the GCN5-related superfamily, were solved, and the R/Q-X-X-G-X-G motif, where X denotes some variations, was demonstrated to form an important part of the pyrophosphate-binding loop (34,35). Nevertheless, our results suggest that Phe 142 and Tyr 143 in domain II of Gna1p play more important roles in the Ac-CoA binding. Because neither Phe 142 nor Tyr 143 is conserved in most of GCN5-related N-acetyltransferases, it is still difficult to predict the exact roles of these two amino acids in catalysis. Possibly, Phe 142 and Tyr 143 in domain II are involved in the interaction with the adenine base of Ac-CoA.