Yeast-derived recombinant DG42 protein of Xenopus can synthesize hyaluronan in vitro.

We demonstrate in this report that the Xenopus DG42 gene product made in the yeast Saccharomyces cerevisiae can synthesize authentic high molecular weight hyaluronan (hyaluronic acid; HA) in vitro. Saccharomyces are eukaryotes that do not naturally produce HA or any other molecules known to contain glucuronic acid. Therefore bakers' yeast is a good model system to determine the enzymatic activity of the DG42 protein, which is the topic of recent debate. Membrane extracts prepared from cells expressing DG42 encoded on a plasmid incorporated [14C]glucuronic acid and N-[3H]acetylglucosamine from exogenously supplied UDP-sugar nucleotides into a high molecular mass (10(6) to 10(7) Da) polymer in the presence of magnesium ions. Both sugar precursors were simultaneously required for elongation. Control extracts prepared from cells with the vector plasmid alone or the DG42 cDNA in the antisense orientation did not display this activity. The double-labeled polysaccharide product synthesized in vitro was deemed to be HA by enzymatic analyses; specific HA lyase could degrade the polymer, but it was unaffected by protease or chitinase treatments. The fragments generated by HA lyase were identical to fragments derived from authentic vertebrate HA as compared by high performance liquid chromatography. We conclude that DG42 is a membrane-associated HA synthase capable of transferring both glucuronic acid and N-acetylglucosamine groups.

Glycoconjugates of the extracellular matrix are essential constituents of multicellular organisms throughout embryogenesis and adult life. One such molecule, HA, 1 is a linear polysaccharide composed of alternating GlcA and GlcNAc residues (reviewed in Ref. 1). A recent controversy in carbohydrate biosynthesis is focused on the identity of the eukaryotic HA synthases (2). The function of DG42 was unknown at the time of its discovery as a major transcript accumulated during Xenopus gastrulation (3). The sequence similarity to some enzymes with a GlcNAc transferase activity, at first noted with Rhizobium NodC and fungal chitin synthases (4,5) followed by streptococcal HA synthase (6,7), yielded a valuable clue to the possible function of DG42.
At least two groups have overexpressed members of an apparent class of candidate HA synthases, Xenopus DG42 (8) or a putative mouse analog HAS (76.7% identity to DG42; Ref. 9), in mammalian cells and observed increases in HA biosynthesis above basal levels. On the other hand, another group has reported that a DG42-fusion protein derived from in vitro translation reactions can produce chitin-like chains of 2 to 6 GlcNAc monomers, but not HA polysaccharide (10). The same group subsequently detected an activity from detergent extracts of zebra fish embryos that could make chitin oligosaccharides (11). The peak of this activity on gel filtration chromatography profiles contained a putative DG42 analog that was detected by cross-reaction with an antibody against Xenopus DG42. The chitin oligosaccharide synthase activity was discrete from an HA-synthesizing activity that eluted earlier on the gel filtration profiles (11).
To avoid the difficulties in interpreting relative increases in HA biosynthesis in mammalian expression systems, we have performed studies employing bakers' yeast as the heterologous host. There are no reports of HA or UDP-GlcA production, capsule formation, or exopolysaccharide synthesis by any Saccharomyces in the Medline data base since 1966. Since the entire genome of Saccharomyces cerevisiae has been sequenced, we searched the Saccharomyces Genome Database (Stanford) for proteins similar to known UDPglucose 6-dehydrogenases (EC 1.1.1.22), the enzymes that catalyze UDP-GlcA precursor production. We used the primary sequences of the enzymes from bovine liver (A54926; Ref. 12) or from Group A Streptococcus (HasB, A46089; Ref. 13) as queries in computer homology analyses (TBLASTN version 1.4.9; Ref. 14). We could not identify a potential yeast analog (smallest sum probability ϭ 0.992-0.82). On the other hand, parallel searches using HasA, the streptococcal HA synthase (L20853; Ref. 6), as the query detected its sequence similarity (smallest sum probability ϭ 10 Ϫ4 -10 Ϫ5 ) to all three of the yeast chitin synthases, CHS1, CHS2, and CHS3, a previously noted homology (6,7). Therefore, several apparent advantages with the yeast system include: (i) no other potential subunits or machinery specific to HA biosynthesis, (ii) no endogenous HA chains on the recombinant HAS enzyme due to the lack of the HA precursor UDP-GlcA, and (iii) the ease of molecular biological manipulation. Here we report that the Xenopus DG42 gene product produced in yeast can synthesize high molecular weight HA in vitro.

EXPERIMENTAL PROCEDURES
Plasmid Construction-Molecular biology reagents were from Promega, and, unless noted, all other reagents were from Sigma. The DG42 gene was obtained from the cDNA plasmid pC4202 (3; generously provided by I. Dawid). The coding region of 588 residues was amplified by 15 cycles of the polymerase chain reaction with Taq DNA polymerase and oligonucleotide primers (Ransom Hill Bioscience) designed to add flanking EcoRI restriction sites (15). The cassette was subcloned by standard methods into pYES2 (Invitrogen), an episomal shuttle plasmid possessing the yeast GAL1 promoter for foreign gene expression, using Escherichia coli as the intermediary host (15). The plasmids were characterized with respect to insert size and orientation by restriction mapping; pYES/DGϩ or pYES/DGϪ contain the gene in the protein expressing orientation or the antisense direction, respectively. Plasmids were transformed into S. cerevisiae INSc1 (Invitrogen) using the lithium acetate/polyethylene glycol method (15). Transformants were selected and maintained on uracil-deficient synthetic media (SM-U) with glucose as the carbon source at 30°C.
Expression and Membrane Preparation-For expression experiments, the various constructs were grown to midlogarithmic phase (0.2-0.8 A 600 ) in SM-U with 1.5% raffinose and 5% glycerol instead of glucose. Galactose, the inducer of the GAL1 promoter, was added to 2%, and the cultures were grown for a further 5-16 h. All the following steps were performed at 4°C or on ice. Cells were collected and washed twice with 50 mM sodium phosphate buffer, 1 mM EDTA, pH 7, by repeated centrifugation at 3,000 ϫ g for 5 min. The cell pellet was suspended in half of its volume of a lysis buffer, SB, composed of the above wash buffer supplemented with 5% glycerol, 1 mM dithiothreitol, 1 mM phenylmethanesulfonyl fluoride, 2 g/ml pepstatin A, and 1 mM benzamidine and frozen at Ϫ80°C until needed. Lysis was performed by three cycles of freeze/thawing followed by vigorous mixing (1 min vortex/1 min ice, 6 cycles) with glass beads (50% total volume; 400 -600 m). The lysate was removed after the beads had settled and been subjected to ultrasonication (20 s, setting 3, Heat Systems W-380 with microtip probe). The debris and unbroken cells were removed from the lysate by low speed centrifugation (2,000 ϫ g, 1 min). The supernatant fraction was diluted 10-fold in SB, and the membranes were harvested by ultracentrifugation (100,000 ϫ g, 1 h). The resulting membrane pellet was resuspended by gentle ultrasonication in SB and washed by another round of ultracentrifugation. The final membrane fraction was resuspended in SB and stored frozen in aliquots at Ϫ80°C. Protein was quantitated by the Coomassie binding assay using a bovine serum albumin standard (Pierce; Ref. 16) Polysaccharide Synthase Assays-The incorporation of sugars into high molecular weight HA polysaccharide was monitored by using UDP-[ 14 C]GlcA (291 mCi/mmol; ICN) and/or UDP-[ 3 H]GlcNAc (27.3 Ci/mmol; DuPont NEN) precursors. Membrane preparations were incubated at 30°C in a volume of 100 l in a buffer typically containing: 50 mM Tris, pH 7.5, 20 mM MgCl 2 , 1 mM dithiothreitol, 0 -180 M UDP-GlcA, 0 -300 M UDP-GlcNAc. Reactions were terminated by the addition of SDS to 2% (w/v). Either paper chromatography (17) or gel filtration chromatography on Sephacryl S-500HR (Pharmacia Biotech Inc.) was utilized to separate products from substrates; the radioactive polymers at the origin of the paper chromatogram or in the column fractions were detected by liquid scintillation counting.
Oligosaccharide Mapping-Carbohydrate digests were analyzed by liquid chromatography utilizing the method of Gherezghiher et al. (19). Polysaccharide samples were exchanged into 50 mM ammonium acetate buffer, pH 5.5, and digested with HA lyase at 400 units/ml at 37°C for 6.5 h. The oligosaccharides for standards were prepared by digestion of authentic HA from rooster comb. The digests were lyophilized twice to remove the volatile salts before resuspension and injection onto a Varian Micropak AX-5 column (4 mm ϫ 30 cm). The column was eluted isocratically with 4% methanol, 0.3 M ammonium formate, pH 5.5, at 1 ml/min. The standard oligosaccharides were monitored by UV absorbance; the action of the enzyme creates a 4,5-double bond that was detected at 232 nm (18). The radiolabeled sugars were measured by liquid scintillation counting of the fractions (0.5 min).

RESULTS
Sugar Incorporation by DG42 in Vitro-Membrane preparations derived from cells with the plasmid pYES/DGϩ encoding DG42 in the correct orientation behind the GAL promoter incorporated the sugars from UDP-GlcNAc and UDP-GlcA into high molecular weight product (Table I). Both precursors were simultaneously required for polymerization. Mg 2ϩ ion is also required for catalysis; membranes incubated with EDTA instead of metal ion did not incorporate radioactivity from the precursors into polymer. Specific activities of ϳ5 pmol of GlcA transferred (g of protein) Ϫ1 h Ϫ1 were typically obtained. Preparations derived from cells with either pYES2 vector or the antisense plasmid, pYES/DGϪ, did not display any similar activity.
Characterization of the DG42 Protein-Membranes derived from cells with pYES/DGϩ were compared before and after the high speed washing step described under "Experimental Procedures." The second membrane pellet contained on average ϳ86% of the activity found in the first pellet even though the total protein concentration was reduced by ϳ40%. No activity was found in the washes. This result suggests that DG42 is associated with the membranes. Total yeast cell lysates possessed only about 10% of the activity observed with the membranes, but this decrease is most likely due to the detrimental effects of cytosolic enzymes on the sugar precursors. By Western blot analysis using antiserum to a DG42-fusion protein supplied by I. Dawid (3), only membranes derived from yeast with pYES/DGϩ possessed a unique immunoreactive band migrating at ϳ67 kDa by SDS-polyacrylamide gel electrophoresis (data not shown). This estimate is very similar to the predicted size from the deduced sequence (3) and the experimental values of radiolabeled embryo extracts (3) or lysates from cells infected with recombinant DG42-vaccinia virus (8).
Analysis of the Polysaccharide Product-Membranes derived from the pYES/DG plasmids were incubated with both UDP-[ 14 C]GlcA and UDP-[ 3 H]GlcNAc, and the reaction mixtures were analyzed by gel filtration chromatography on Sephacryl S-500HR (Fig. 1). A double-labeled polymer produced by membranes of cells with pYES/DGϩ eluted with an apparent molecular mass ranging from 10 6 to 10 7 Da; this value is based on the elution position of the blue dextran standard (average 2 ϫ10 6 Da; Pharmacia) and the resin exclusion limit (Ն2 ϫ 10 7 Da). No such polymer was detected in parallel runs with the antisense plasmid control (data not shown).
The polysaccharide product synthesized by membranes from pYES/DGϩ cells and purified on the S-500HR column was deemed to be HA by its complete sensitivity to the specific Streptomyces HA lyase. Untreated material remained at the origin of the paper chromatogram, but at least 99% of the 3 H and 14 C label migrated away from the origin after digestion. These labeled fragments also eluted in the totally included Aliquots of membranes (50 g of protein) were incubated with the radiolabeled sugar donor UDP-[ 14 C]GlcA (60 M, 0.05 Ci) and either 0 or 300 M unlabeled UDP-GlcNAc, the other precursor of HA, in 50 mM Tris, 1 mM dithiothreitol, pH 7.5. Additionally, either 20 mM MgCl 2 (ϩ) or 5 mM EDTA (Ϫ) were included in the reactions as noted in the "Mg 2ϩ " column. The reactions were terminated after 1 h at 30°C, and the polymer product was quantitated by the paper chromatography method. Only membranes derived from cells expressing the DG42 gene product, and not the vector or antisense plasmid, incorporated radiolabel into polysaccharide. Substitution of Mg 2ϩ with EDTA resulted in the complete loss of the activity observed in incubations of both UDP-sugars with pYES/DGϩ. Mg 2ϩ and both precursors are simultaneously required for HA synthesis in vitro. Yeast-derived DG42 Can Synthesize HA 23658 fractions of the S-500HR column upon rechromatography (data not shown). Portions of the intact polysaccharide were also treated with chitinase or protease in parallel experiments, and all of the radiolabel derived from both sugar precursors remained at the origin of the paper chromatogram (data not shown). Furthermore, we did not detect the release of any smaller radiolabeled sugar fragments by chitinase utilizing gel filtration analysis of the digest on Sephadex G-25 (detection limit Ͻ0.3% of starting material).

Plasmid
The fragments derived from HA lyase digestion of the radiolabeled polysaccharide were further characterized by high performance liquid chromatography on a Micropak AX-5 amino column (Fig. 2). The radiolabeled fragments eluted with the same retention times as the major tetrasaccharide and hexasaccharide products derived from parallel digests of authentic vertebrate HA; these oligosaccharides are the limit digest products made by the Streptomyces HA lyase (18). DISCUSSION As noted in the introduction and a commentary elsewhere (2), a definitive answer on the nature of the enzymatic activity of the DG42 polypeptide was not previously available. Due to the innate ability of mammalian cells to produce HA, the earlier experiments in the other model cell systems could not rule out the possibility that components, such as catalytic or regulatory species, distinct from the DG42 polypeptide were also required for HA biosynthesis. It is quite plausible that the HAS protein of Itano and Kimata (9) actually complemented the defect of their particular cell line, and perhaps the HAS molecule alone was not sufficient for HA production. In the case of Meyer and Kreil (8), overexpression of DG42 could have potentially elevated the concentration of a limiting biosynthetic component in a cell line with a low endogenous HA production capability, rather than the DG42 polypeptide be an actual HA synthase. These scenarios seem quite possible in light of information reported by Semino et al. (11) that some irrelevant molecules, such as an integrin subunit and p34, could increase HA production when overexpressed in mammalian cells.
Bakers' yeast does not possess any of the required components specific to HA elongation and, therefore, in some respects is a better model system to dissect HA biosynthesis than the vertebrate cell systems with the endogenous ability to produce HA. The data in this report clearly show that the recombinant DG42 gene product copolymerizes GlcA and GlcNAc groups to form the repeating, linear HA polysaccharide. The size of the polymer produced by recombinant yeast membranes was comparable to high quality, purified vertebrate HA polysaccharide (1). Therefore, Xenopus DG42, and, very probably, mouse HAS (9) are genuine HA synthases which require no other catalytic subunits for HA elongation.
The attempt by Semino and Robbins (10) to detect HA polymerization with recombinant DG42 in an in vitro system did not succeed. We believe that their recombinant DG42 molecule could not make HA for perhaps several reasons: (i) the protein was produced with a potentially disruptive T7-peptide tag at the amino terminus, (ii) no lipids or membranes were available during translation in the in vitro system, or (iii) other proteinfolding or post-translational modification machinery is required. Any one or all of these reasons could cause misfolding, and the DG42 polypeptide would not be able to elongate HA. For example, we have found evidence that the streptococcal HasA amino terminus is important for activity; deletion of the first 24 residues results in an inactive product. 2 Furthermore, we could detect neither significant protein production nor HA synthase activity of recombinant DG42 produced by several foreign gene expression systems employing E. coli as the host (data not shown). The DG42 protein prepared in the yeast system, however, was active with respect to HA elongation in our very first experiment.
HA is produced during the entire lifetime of higher animals. One of the most obvious previous arguments against DG42 being an eukaryotic HA synthase was the apparent temporal 2 P. L. DeAngelis and P. H. Weigel, unpublished data. Membranes from cells with pYES/DGϩ (100 g of protein) were incubated with both radiolabeled precursors diluted to roughly equivalent specific activity (60 M, 0.2 Ci each) under typical buffer conditions with Mg 2ϩ . After 30 min, additional unlabeled precursors were added to 180 M each, and the incubation was continued for 90 min. SDS was added to 2% (w/v), and the reaction was heated at 95°C for 1 min. The mixture was clarified by centrifugation and injected onto a Sephacryl S-500HR column (1 ϫ 50 cm) equilibrated in 0.2 M NaCl, 5 mM Tris, pH 8. The radioactivity in a portion of each fraction (0.2 ml of 1 ml) was measured ( 3 H, circles; 14 C, squares). BD marks the elution position of the peak of blue dextran (average ϳ2 ϫ 10 6 Da; Pharmacia). The excluded volume is at 14 ml. High molecular weight product composed of both radiolabeled sugars was produced only by samples from cells with pYES/DGϩ and not by cells with vector or pYES/DGϪ. The remainder of the material in fractions 14 -24 was pooled and aliquots were treated with degradative enzymes to assess the nature of the polysaccharide (see "Results" and Fig. 2).

FIG. 2.
Mapping of HA lyase digests. High molecular weight, double-labeled material from recombinant pDG42ϩ yeast was purified on the S-500HR column (fractions 14 -24; depicted in Fig. 1) and treated with HA lyase. The radiolabeled fragments (1/4 of pool) were analyzed by chromatography on a Micropak AX-5 amino column as described in the text. The arrowhead marks the elution time for unretained molecules. The column was standardized with an HA lyase digest of authentic HA (5 g). Two UV-absorbing peaks were observed in the standard that eluted at 5.9 min and 8.5 min (not shown); these species correspond to the unsaturated tetrasaccharide and hexasaccharide, respectively, and are the final products formed by the Streptomyces enzyme (18,19). The radiolabeled peaks ( 14 C, squares; 3 H, circles) eluted with the same retention times as the oligosaccharides derived from pure authentic HA. Yeast-derived recombinant DG42 can make HA.
Yeast-derived DG42 Can Synthesize HA 23659 restriction of DG42 expression to the period of gastrulation (3). Our observation that DG42 can indeed produce HA suggests that several genes may exist in the vertebrate genome that encode different developmental or tissue-specific isozymes of HA synthase. It is quite likely that a family of eukaryotic HA synthases will be found upon further examination of sequenced genomes as well as the result of directed explorations. DG42 alone can transfer both GlcA and GlcNAc groups to form HA. In almost all known cases of carbohydrate biosynthesis, one glycosyltransferase transfers only one type of sugar subunit to form a certain, specific linkage. There is precedent, however, for one glycosyltransferase having the ability to transfer two distinct sugars. Based on genetic and biochemical evidence, the streptococcal HA synthase, HasA, has been demonstrated to transfer both GlcA and GlcNAc to the growing HA chain. Only one gene encoding one protein on a recombinant plasmid is required for HA synthesis in both Gram-positive and Gram-negative bacteria, as long as the UDP-GlcA precursor is present or supplied (6,20). 2 Immunoaffinity-purified HasA can also synthesize HA chains (17). There is also some evidence that another enzyme, the synthase involved in producing the repeating disaccharides of heparin, is a single polypeptide that can similarly execute both functions (21). It is quite interesting that these enzymes which synthesize structurally related, linear glycosaminoglycans appear to share this characteristic.
In conclusion, the Xenopus DG42 protein is an HA synthase, a transferase that can form HA chains by alternating addition of both GlcA and GlcNAc groups; no other specific catalytic subunits appear to be required. It is extremely likely that most of the newly discovered vertebrate DG42 homologs (9) are also bona fide HA synthases. The elucidation of the details of the enzymatic mechanism should be an active field of research now that a prototype eukaryotic HA synthase has been identified and experimental model systems are available. Basic information on HA biosynthesis will surely aid our understanding of numerous phenomena including development, intercellular adhesion and recognition, cell motility, and cancer.