Molecular Directionality of Polysaccharide Polymerization by thePasteurella multocida Hyaluronan Synthase*

Hyaluronan (HA), a long linear polymer composed of alternating glucuronic acid and N-acetylglucosamine residues, is an essential polysaccharide in vertebrates and a putative virulence factor in certain microbes. All known HA synthases utilize UDP-sugar precursors. Previous reports describing the HA synthase enzymes from Streptococcus bacteria and mammals, however, did not agree on the molecular directionality of polymer elongation. We show here that a HA synthase, PmHAS, from Gram-negative P. multocida bacteria polymerizes the HA chain by the addition of sugar units to the nonreducing terminus. Recombinant PmHAS will elongate exogenous HA oligosaccharide acceptors to form long polymersin vitro; thus far no other HA synthase has displayed this capability. The directionality of synthesis was established definitively by testing the ability of PmHAS to elongate defined oligosaccharide derivatives. Analysis of the initial stages of synthesis demonstrated that PmHAS added single monosaccharide units sequentially. Apparently the fidelity of the individual sugar transfer reactions is sufficient to generate the authentic repeating structure of HA. Therefore, simultaneous addition of disaccharide block units is not required as hypothesized in some recent models of polysaccharide biosynthesis. PmHAS appears distinct from other known HA synthases based on differences in sequence, topology in the membrane, and putative reaction mechanism.

Polysaccharides are the most abundant biomaterials on earth, yet many of the molecular details of their biosynthesis and function are not clear. HA 1 is a linear polysaccharide of the glycosaminoglycan class composed of up to thousands of ␤(1,4)GlcUA-␤(1,3)GlcNAc repeats. In vertebrates, HA is a major structural element of the extracellular matrix and plays roles in adhesion and recognition (1). HA has a high negative charge density and numerous hydroxyl groups; therefore, the molecule assumes an extended, hydrated conformation in solution. The viscoelastic properties of cartilage and synovial fluid are in part the result of the physical properties of the HA polysaccharide. HA also interacts with proteins such as CD44, RHAMM, and fibrinogen, thereby influencing many natural processes such as angiogenesis, cancer, cell motility, wound healing, and cell adhesion (2).
HA is also made by certain microbes. Some bacterial pathogens, namely Gram-negative Pasteurella multocida Type A and Gram-positive Streptococcus Group A and C, produce extracellular HA capsules (3,4); this coating protects the microbes from host defenses including complement and phagocytosis (5,6). The Paramecium bursaria chlorella virus (PBCV-1) directs the algal host cells to produce a HA surface coating early in infection, but the biological role of HA in the viral life cycle is not yet known (7).
The various HA synthases, the enzymes that polymerize HA, utilize UDP-GlcUA and UDP-GlcNAc sugar nucleotide precursors in the presence of a divalent Mn 2ϩ or Mg 2ϩ ion (7)(8)(9). The HA synthase activity from all sources is localized in the membrane fraction. The enzymes were all identified by molecular genetic means due to the innate problems of membrane protein purification. In all cases, a single species of polypeptide catalyzes the transfer of two distinct sugars; in contrast, the vast majority of other known glycosyltransferases transfer only one monosaccharide.
HasA (or SpHAS) from Group A Streptococcus pyogenes was the first HA synthase to be described at the molecular level (10). The various vertebrate homologs (Xenopus frog DG42 or XlHAS1; murine and human HAS1, HAS2, and HAS3) and the viral enzyme, A98R, are quite similar at the amino acid level to certain regions of the HasA polypeptide chain (ϳ30% identity overall). At least 7 short motifs (5-9 residues) interspersed throughout these enzymes are identical or quite conserved. The evolutionary relationship among these HA synthases from such dissimilar sources is not clear at present. The enzymes are predicted to have a similar overall topology in the bilayer; membrane-associated regions at the amino and the carboxyl termini flank a large cytoplasmic central domain (ϳ200 amino acids; reviewed in Ref. 8). The amino-terminal region appears to contain two transmembrane segments, whereas the carboxyl-terminal region appears to contain three to five membraneassociated or transmembrane segments depending on the species. Very little of these HAS polypeptide chains are expected to be exposed to the outside of the cell.
With respect to the reaction pathway utilized by this group of enzymes, mixed findings have been reported from indirect experiments. The Group A streptococcal enzyme was reported to add sugars to the nonreducing terminus of the growing chain as determined by selective labeling and degradation studies (11). Using a similar approach, however, two laboratories working with the enzyme preparations from mammalian cells concluded that the new sugars were added to the reducing end of the nascent chain (12,13). In comparing these various studies, the analysis of the enzymatically released sugars from the streptococcal system added more rigorous support for their interpretation (11). In another type of experiment, HA made in mammalian cells was reported to have a covalently attached UDP group as measured by an incorporation of low amounts of radioactivity derived from 32 P-labeled UDP-sugar into an anionic polymer (14). These data implied that the last sugar was transferred to the reducing end of the polymer. Thus it remains unclear if these rather similar HAS polypeptides from vertebrates and streptococci actually utilize different reaction pathways.
We recently reported the identification and molecular cloning of a unique HA synthase, PmHAS, from the fowl cholera pathogen, Type A P. multocida (15). Expression of this single 972-residue protein allowed the Escherichia coli host cells to produce HA capsules in vivo; normally E. coli does not make HA. Overall, the deduced PmHAS sequence is very different from the other known HA synthases. There appear to be only two short potential sequence motifs ((D/N)DGS(S/T); DSD(D/ T)Y) in common between PmHAS and Group A HasA. Instead, a portion of the central region of the new enzyme is more homologous to the amino termini of other bacterial glycosyltransferases that produce different capsular polysaccharides or lipopolysaccharides. Furthermore, even though PmHAS is about twice as long as any other HAS enzyme, it only has two predicted transmembrane-spanning helices separated by ϳ320 residues. Thus at least a third of the polypeptide is predicted not to be in the cytoplasm.
In this report, definitive proof is presented that PmHAS adds sugars to the nonreducing end of the growing polymer chain, in contrast to the reports with mammalian enzymes. Furthermore, it is shown that the correct monosaccharides are added sequentially in a stepwise fashion to the nascent chain. This is the first direct demonstration of HA polysaccharide polymerization in such a fashion.

EXPERIMENTAL PROCEDURES
All reagents were from Sigma or Fisher unless noted otherwise. HA Synthase Isolation and Assays-Membrane preparations containing recombinant PmHAS (rPmHAS) (GenBank™ number AF036004) were isolated from E. coli SURE(pPmHAS) as described (15). Membrane preparations containing native PmHAS were obtained from the P. multocida strain P-1059 (ATCC #15742) as described (9), except that 1 mM ␤-mercaptoethanol was substituted for thioglycollate throughout the procedure. PmHAS was assayed in 50 mM Tris, pH 7.2, 20 mM MnC1 2 , and UDP-sugars (UDP-[ 14 C]GlcUA, 0.3 Ci/mmol (NEN Life Science Products) and UDP-GlcNAc) at 30°C. The reaction products were analyzed by various chromatographic methods as described below. Membrane preparations containing other recombinant HAS enzymes, Group A streptococcal HasA or Xenopus DG42 produced in the yeast Saccharomyces cerevisiae, were prepared as described previously (16).
Acceptor Oligosaccharides-Uronic acid was quantitated by the carbazole method (17). Even-numbered HA oligosaccharides ((GlcNAc-GlcUA) n ) were generated by degradation of HA (from Group C Streptococcus) with either ovine testicular hyaluronidase Type V (n ϭ 2-5) or Streptomyces hyaluroniticus HA lyase (n ϭ 2 or 3) in 30 mM sodium acetate, pH 5.2, at 30°C overnight. The latter enzyme employs an elimination mechanism to cleave the chain, resulting in an unsaturated ⌬GlcUA residue at the nonreducing terminus of each fragment (18). For further purification and desalting, some preparations were subjected to gel filtration with P-2 resin (Bio-Rad) in 0.2 M ammonium formate and lyophilization. Odd-numbered HA oligosaccharides (GlcNAc(GlcUA-GlcNAc) n ) ending in a GlcNAc residue were prepared by mercuric acetate treatment of partial HA digests generated by HA lyase (n ϭ 2-7; gift of Dr. G. Sugumaran;Ref. 19). The masses of the HA oligosaccharides were verified by matrix-assisted laser desorption ionization timeof-flight mass spectrometry. Sugars in water were mixed with an equal volume of 5 mg/ml 6-azo-2-thiothymine in 50% acetonitrile, 0.1% trifluoroacetic acid and rapidly air-dried on the target plate. The negative ions produced by pulsed nitrogen laser irradiation were analyzed in linear mode (20-kV acceleration; Perceptive Voyager®).
Other oligosaccharides that are structurally similar to HA were also tested in HAS assays. The structure of heparosan pentamer derived from the E. coli Various oligosaccharides were radiolabeled by reduction with 4 to 6 equivalents of sodium borotritide (20 mM, NEN Life Science Products; 0.2 Ci/mmol) in 15 mM NaOH at 30°C for 2 h (21). [ 3 H]Oligosaccharides were desalted on a P-2 column in 0.2 M ammonium formate to remove unincorporated tritium and lyophilized. Some labeled oligosaccharides were further purified preparatively by paper chromatography with Whatman 1 developed in pyridine/ethyl acetate/acetic acid/H 2 O (5:5: 1:3) before use as an acceptor.
Chromatographic Analyses of HA Synthase Reaction Products-Paper chromatography with Whatman No. 3M developed in ethanol 1 M ammonium acetate, pH 5.5 (65:35), was used to separate high molecular weight HA product (which remains at the origin) from UDP-sugars and small acceptor oligosaccharides (22). In the conventional HAS assay, radioactive UDP-sugars are polymerized into HA. To obtain the size distribution of the HA polymerization products, some samples were also separated by gel filtration chromatography with Sephacryl S-200 (Amersham Pharmacia Biotech) columns in 0.2 M NaCl, 5 mM Tris, pH 8. Columns were calibrated with dextran standards. The identity of the polymer products was assessed by sensitivity to specific HA lyase and the requirement for the simultaneous presence of both UDP-sugar precursors during the reaction. Thin layer chromatography (TLC) on high performance silica plates with application zones (Whatman) utilizing butanol/acetic acid/water (1.5:1:1 or 1.25:1:1) development solvent separated 3 H-labeled oligosaccharides in reaction mixes. Radioactive molecules were visualized after impregnation with EnHance spray (NEN Life Science Products) and fluorography at Ϫ80°C.

RESULTS
HA Oligosaccharides Serve as Acceptors for rPmHAS-In our previous report, membrane preparations from recombinant E. coli containing rPmHAS protein had HA synthase activity as judged by incorporation of radiolabel from UDP-[ 14 C]GlcUA into polymer when co-incubated with both UDP-GlcNAc and Mn 2ϩ ion (15). Based on the similarity at the amino acid level of PmHAS to some lipopolysaccharide transferases, it seemed possible that HA oligosaccharides would serve as acceptors for GlcUA and GlcNAc transfer. The addition of unlabeled evennumbered HA tetramer (from testicular hyaluronidase digests) to reaction mixtures with rPmHAS stimulated incorporation of radiolabel from UDP-[ 14 C]GlcUA into HA polymer by ϳ20to 60-fold in comparison to reactions without oligosaccharides (Fig. 1). On the other hand, structurally similar sugars, hep- arosan pentamer or chitotetraose, did not stimulate incorporation of the radiolabel. The free monosaccharides GlcUA and GlcNAc, either singly or in combination at concentrations of up to 100 mM, did not serve as acceptors; likewise, the ␤-methyl glycosides of these sugars did not stimulate HAS activity (not shown).
The activity of rPmHAS was dependent on the simultaneous incubation with both UDP-sugar precursors and Mn 2ϩ ion. The level of incorporation was dependent on protein concentration, on HA oligosaccharide concentration, and on incubation time (Fig. 2). HA synthesized in the presence or the absence of HA oligosaccharides was sensitive to HA lyase (Ͼ95% destroyed) and had a molecular weight of Ն1-5 ϫ 10 4 Da (ϳ50 -250 monosaccharides). No requirement for a lipid-linked intermediate was observed as neither bacitracin (0.5 mg/ml) nor tunicamycin (0.2 mg/ml) altered the level of incorporation in comparison to parallel reactions with no inhibitor.
The activity of the native PmHAS from P. multocida membranes, however, was not stimulated by the addition of HA oligosaccharides under similar conditions (not shown). It is likely that native PmHAS enzyme has an attached or bound nascent HA chain that was initiated in the bacterium before membrane isolation. The recombinant enzyme, on the other hand, is expected to lack a nascent HA chain as the E. coli host does not produce the UDP-GlcUA precursor needed to make HA polysaccharide (10). Therefore, the exogenous HA-derived oligosaccharide has access to the active site of rPmHAS and can be elongated.
Sugar Transfer by rPmHAS Occurs at the Nonreducing End-The tetramer from ovine testicular hyaluronidase digests of HA terminates at the nonreducing end with a GlcUA residue; this molecule served as an acceptor for HA elongation by rPmHAS. On the other hand, the ⌬tetramer and ⌬hexamer oligosaccharides produced by the action of Streptomyces HA lyase did not stimulate HA polymerization (Fig. 1, unsHA4/6). As a result of the lyase eliminative cleavage, the terminal unsaturated sugar is missing the C4 hydroxyl of GlcUA (18), which would normally be extended by the HA synthase. The lack of subsequent polymerization onto this terminal unsaturated sugar is analogous to the case of dideoxynucleotides causing chain termination if present during DNA synthesis. A closed pyranose ring at the reducing terminus was not required by PmHAS, since reduction with borohydride did not effect the ability of the HA tetramer to serve as an acceptor; this finding also allowed the use of borotritide labeling to monitor the fate of oligosaccharides.
Other Recombinant HASs Do Not Utilize HA Oligosaccharide Acceptors-Neither recombinant Group A HasA nor recombinant DG42 produced in yeast elongated HA-derived oligosaccharides into larger polymers. First, the addition of HA tetramer or a series of longer oligosaccharides neither stimulated nor inhibited the incorporation of radiolabeled UDP-sugar precursors into HA by these enzymes significantly (ՆϮ5% control value). In parallel experiments, the HAS activity of HasA or DG42 were not affected by the addition of chitin-derived oligosaccharides (data not shown). Second, the recombinant enzymes did not elongate radiolabeled HA tetramer in the presence of UDP-sugars (Table I). These same preparations of enzymes, however, were highly active in the conventional HAS assay in which radiolabeled UDP-sugars were polymerized into HA.
rPmHAS Elongates HA via Stepwise Addition of Single Sugars-TLC was utilized to monitor the PmHAS-catalyzed elongation reactions containing 3 H-labeled oligosaccharides and various combinations of UDP-sugar nucleotides. Fig. 4A clearly shows that rPmHAS elongated HA-derived tetramer by a single sugar unit if the next appropriate UDP-sugar precursor was available in the reaction mixture. GlcNAc derived from UDP-GlcNAc was added onto the GlcUA residue at the nonreducing terminus of the tetramer acceptor to form a pentamer. On the other hand, inclusion of only UDP-GlcUA did not alter the mobility of the oligosaccharide. If both HA precursors were supplied, then various longer products were made. In parallel reactions, control membranes prepared from host cells with vector plasmid did not alter the mobility of the radiolabeled HA tetramer under any circumstances (not shown). In similar analyses monitored by TLC, rPmHAS did not utilize labeled chitopentaose as an acceptor (Fig. 4B).
HA-derived oligosaccharides with either GlcUA or GlcNAc at the nonreducing terminus served as acceptors for rPmHAS (Fig. 5). Within 2 min, 2 to 6 sugar units were added, and after 20 min, 9 to Ն15 units were added. In the experiments with HA tetramer and both sugars at the 20-min time point, a ladder of even-and odd-numbered products is produced. Therefore, in combination with the results of the single UDP-sugar experiments, it appears that PmHAS transfers individual monosaccharides sequentially during the polymerization reaction.
The apparent GlcUA transfer rate of PmHAS is faster than the GlcNAc transfer rate, as indicated by buildup of products terminating in GlcUA at the reducing end in experiments with either acceptor at the 2-min time point. This finding may be the result of the higher relative affinity of PmHAS enzyme for the UDP-GlcUA substrate than UDP-GlcNAc, as measured by the apparent Michaelis constant (K m ) in previous kinetic analyses (20 M versus 75 M, respectively; Ref. 9).

DISCUSSION
Potential Polymer Retention Mechanisms-An intrinsic and essential feature of polysaccharide synthesis is the repetitive addition of sugar monomer units to the growing polymer. The glycosyltransferase is expected to remain in association with the nascent chain. This feature is particularly relevant for HA biosynthesis, as the HA polysaccharide product in all known cases is transported out of the cell; if the polymer was released, then the HAS would not have another chance to elongate that particular molecule. Three possible mechanisms for maintaining the growing polymer chain at the active site of the enzyme are immediately obvious. First, the enzyme possesses a carbohydrate polymer binding pocket or cleft. Second, the nascent chain is covalently attached to the enzyme during its synthesis. Third, the enzyme binds to the nucleotide base or the lipid moiety of the precursor while the nascent polymer chain is still covalently attached. Thus far, the molecular details of the vast majority of polysaccharide synthases are lacking.
PmHAS and Acceptor Oligosaccharides-The HAS activity of the native PmHAS enzyme found in P. multocida membrane preparations was not stimulated by addition of HA oligosaccharides; theoretically, the endogenous nascent HA chain initiated in vivo renders the exogenously supplied acceptor unnecessary. However, recombinant PmHAS produced in an E. coli strain that lacks the UDP-GlcUA precursor and, thus, lacks a nascent HA chain, is able to bind and to elongate exogenous HA oligosaccharides. As mentioned above, there are three likely means for a nascent HA chain to be held at or near the active site. In the case of PmHAS, it appears that a HAbinding site exists near or at the sugar transferase catalytic site.
Defined oligosaccharides that vary in size and composition may be utilized to discern the nature of the interaction between PmHAS and the sugar chain. For example, it appears that the putative HA polymer-binding pocket of PmHAS will bind and elongate at least an intact HA trisaccharide (reduced tetramer). The monosaccharides GlcUA or GlcNAc, however, even in combination at high concentration, are not effective acceptors. Oligosaccharide binding to PmHAS appears to be quite selective, because the heparosan pentamer, which only differs in the glycosidic linkages from HA-derived oligosaccharides, does not serve as an acceptor. Future studies will further examine the structural requirements for the acceptor molecule as well as the identity of the oligosaccharide-binding site on the PmHAS polypeptide.
The recombinant PmHAS enzyme, however, will produce HA chains without the addition of exogenous HA-derived oligosaccharide, albeit at a lower rate. Perhaps chain initiation is the rate-limiting step in HA biosynthesis. Thus the stimulation of sugar incorporation into HA chains observed in the presence of  The various recombinant enzymes were tested for their ability to convert HA tetramer into higher molecular weight products. The reactions contained radiolabeled HA tetramer (5-8 ϫ 10 5 dpm), 750 M UDP-GlcNAc, 360 M UDP-GlcUA, 20 mM XCI 2 , 50 mM Tris, pH 7-7.6 (the respective X cation and pH values for used for each enzyme were: PmHAS, Mn 2ϩ /7.2; Xenopus DG42, Mg 2ϩ /7.6; Group A streptococcal HasA, Mg 2ϩ /7.0), and enzyme (units/reaction listed). As a control, parallel reactions in which the metal ion was chelated (22 mM EDTA final; EDTA column, rows with ϩ) were tested; without free metal ion, the HAS enzymes do not catalyze polymerization. After a 1-h incubation, the reactions were terminated and subjected to descending paper chromatography. Only rPmHAS could elongate HA tetramer even though all three membrane preparations were very active in the conventional HAS assay (incorporation of [ 14 C]GlcUA from UDP-GlcUA into polymer when supplied UDP-GlcNAc). r-, recombinant. HA-derived acceptors is likely to be due to the circumvention of the initial kinetic obstacle.
Acceptors and Other HA and Glycosaminoglycan Synthases-Previously no HA synthase had been shown to utilize an exogenous acceptor or primer sugar. In an early study of a cell-free HA synthesis system, preparations of native Group A streptococcal HAS were neither inhibited nor stimulated by the addition of various HA oligosaccharides, including the HA tetramer derived from testicular hyaluronidase digests (11). These membrane preparations were isolated from cultures that were producing copious amounts of HA polysaccharide. The cells were hyaluronidase-treated to facilitate handling. Therefore, it is quite likely that the native streptococcal enzyme was isolated with a small nascent HA chain attached to or bound to the protein much as suspected in the case of native PmHAS. Theoretically, the existing nascent chain formed in vivo would block the entry and subsequent utilization of an exogenous acceptor by the isolated enzyme in vitro. With the advent of molecularly cloned HAS genes, it is possible to prepare virgin enzymes lacking a nascent HA chain if the proper host is utilized for expression. The yeast S. cerevisiae, an organism whose genome has been totally sequenced, does not possess the UDP-glucose dehydrogenase that is required for UDP-GlcUA precursor synthesis. Nonetheless, the virgin yeast-derived recombinant streptococcal or vertebrate enzymes did not utilize HA acceptor oligosaccharides in our experiments in vitro. Possible explanations for this finding are that the enzymes lack an accessible binding site for the HA-derived acceptor chains tested, or the enzymes utilize a different polymer retention mechanism.
Recently, it has been postulated that certain vertebrate HAS enzymes, Xenopus DG42 and the Brachydanio zebrafish homolog in particular, can produce chitin oligosaccharides under certain conditions (23,24). Another possibility forwarded was that chitin oligosaccharide primers are used to initiate HA chains, and polymerization would occur at the nonreducing terminus (24). More defined enzyme systems will be needed to address this difficult issue in the vertebrate system. With respect to PmHAS, however, chitotetraose and chitopentaose neither stimulated HA production nor served as acceptors in our experiments.
In the case of the biosynthesis of the other glycosaminogly-can polysaccharides, heparin and chondroitin, some details are available (reviewed in Refs. 25 and 26). Both heparin and chondroitin are synthesized by the addition of sugar units to the nonreducing end of the polymer chain. In vivo, the glycosyltransferases initiate chain elongation on primer tetrasaccharides (xylose-galactose-galactose-GlcUA) that are attached to serine residues of proteoglycan core molecules. In vitro, enzyme extracts transfer a single sugar to exogenously added heparin or chondroitin oligosaccharides (26 -29); unfortunately, the subsequent sugar of the disaccharide unit is usually not added, and processive elongation to longer polymers does not occur. Therefore it is likely that some component is altered or missing in the in vitro system. In the case of heparin biosynthesis, it is postulated that a single enzyme transfers both GlcUA and GlcNAc sugars to the glycosaminoglycan chain based on co-purification or expression studies (27,28). Recent work with the E. coli K5 KfiC enzyme, which polymerizes heparosan, indicates that a single protein can transfer both sugars to the nonreducing end of acceptor molecules in vitro (30). Processive elongation, however, was not demonstrated in these experiments; crude cell lysates transferred a single sugar to defined even-or odd-numbered oligosaccharides. However, their initial mutagenesis experiments suggest that at least two independent sites were involved in the transfer of the two monosaccharides (30).
Overview of Previous Models of Polysaccharide Synthesis-Several models describing various facets of the biosynthesis of HA and other polysaccharides have been proposed in the literature, but the lack of purified, stable enzymes and/or the difficulty of monitoring early stages of the reaction have prevented the rigorous testing of these hypotheses. The first theoretical model of HA biosynthesis based on direct logic proposed that two sites transferred the sugars individually from precursors to the nonreducing terminus of the nascent chain in an alternating fashion (31). Subsequent work by this laboratory on the Group A streptococcal HAS enzyme gave corroborating data (11). On the other hand, two other groups concluded that HA was extended by the addition of sugars to the reducing end in experiments with native mammalian HAS preparations (12)(13)(14). It is obvious that direct experiments with defined systems utilizing purified enzyme will be necessary to address these issues.
Recently, general mechanistic models have been proposed for ␤-glycosyltransferases that synthesize polysaccharides (32,33). The hypotheses were based loosely on the putative mechanism of carbohydrate hydrolases that cleave polymer chains, but in the case of synthesis, the reaction would run in the reverse direction. It was also proposed in both models that three binding sites for sugar-nucleotides and/or sugars are utilized to synthesize the polymer in a processive fashion. The directionality of chain synthesis was undecided until electron microscopy and x-ray diffraction data from cellulose fibrils protruding from Acetobacter bacteria implied that new sugars were added to the nonreducing terminus (33). In these models, the two monosaccharides of the disaccharide repeat are simultaneously added onto the polymer chain bound to the enzyme. This reaction pathway allows the formation of ␤-linked bonds from ␣-linked UDP-sugars by an inversion mechanism and removes the need for the polymer chain or the protein to rotate during the elongation reaction. This latter feature was invoked in part to eliminate topological problems during the formation of insoluble cellulose fibrils (32). The models were then extended further to other ␤-linked polysaccharide synthases due to the similarities of certain putative domains and/or motifs at the protein sequence level (32,33).
Formation of the Disaccharide Repeat Structure of HA by FIG. 5. TLC analysis of the early stages of HA elongation. Radiolabeled HA pentamer (5ϫ10 3 dpm 3 H (HA5)) or HA tetramer (25ϫ10 3 dpm 3 H (HA4)) was incubated with rPmHAS and various combinations of UDP-sugars (as in Fig. 4) for 2 or 20 min. Portions (1.5 l) of the supernatant were spotted onto the TLC plate and developed in 1.5:1:1 solvent. This autoradiogram (1-month exposure) shows the single addition of a sugar onto an acceptor when only the appropriate precursor is supplied (HA4, N lane and HA5, A lane). If both UDPsugars are supplied (ϩAN lanes), then a ladder of products with final sizes of 6, 8, and 10 sugars is formed from either HA4 or HA5 in 2 min. After 20 min, a range of odd-and even-numbered product sugars are observed in reactions with HA4 and both UDP-sugars. In the 20-min reaction with HA5 and both UDP-sugars, the HA products are so large that they do not migrate from the application zone.
PmHAS-We have found that recombinant PmHAS adds single monosaccharides in a sequential fashion to the nonreducing termini of the nascent HA chain. Elongation of HA polymers containing hundreds of sugars was demonstrated in vitro. The simultaneous formation of the disaccharide repeat unit is not necessary for generating the alternating structure of the HA molecule. The intrinsic specificity and fidelity of each halfreaction (e.g. GlcUA added to a GlcNAc residue or vice versa) apparently is sufficient to synthesize authentic HA chains.
A great technical benefit resulting from the alternating disaccharide structure of HA is that the reaction may be dissected by controlling the availability of UDP-sugar nucleotides. By omitting or supplying precursors in a reaction mixture, the glycosyltransferase may be stopped and started at different stages of synthesis of the heteropolysaccharide. In contrast, there is no facile way to control in a stepwise fashion the glycosyltransferase enzymes that produce important homopolysaccharides such as chitin, cellulose, starch, and glycogen. The lessons learned with PmHAS in the future, however, may be applicable to the study of other enzyme systems.
Two Classes of HA Synthases-It has been established that one polypeptide species transfers both GlcUA and GlcNAc during HA biosynthesis in all known cases (7,8,10,15,16,22), but at least two evolutionary paths may have led to the creation of HA synthases. I propose that two distinct classes of enzyme exist based on differences in the amino acid sequences, the predicted polypeptide topology in the membrane bilayer, and the putative reaction pathway. Class I HASs would include the previously described streptococcal, viral, and vertebrate enzymes. At present, P. multocida PmHAS is the only known member of the Class II HA synthase. In the near future, it will be interesting to examine and to compare the reaction mechanisms of the glycosaminoglycan synthases and other ␤-glycosyltransferases in more detail.