Hyaluronan Synthases*

In 1934, Meyer and Palmer isolated a novel, high Mr glycosaminoglycan from the vitreous of the eye (1). They showed that this substance contained a hexuronic acid, an amino sugar, and no sulfoesters and proposed the name hyaluronic acid (hyaluronan, HA), from the Greek hyaloid (vitreous) and uronic acid. It took 20 years before Weissmann and Meyer (2) finally established the precise structure of the repeating disaccharide unit of hyaluronic acid (GlcAb(133)GlcNAcb(134)). The number of repeating disaccharides in an HA molecule can exceed 30,000, a Mr .10 . MedLine surveys for reports describing the structure, synthesis, degradation, and biology of HA reveal a steadily increasing interest in this biopolymer during the four decades following the determination of its structure: 790 papers published from 1966 to 1975; 2200 from 1976 to 1985; over 3300 from 1986 to 1996. During this time, HA has been identified in virtually every tissue in vertebrates and has achieved widespread use in various clinical applications, most notably and appropriately as an intra-articular matrix supplement (3) and in eye surgery. This period has also seen a transition from the original perception that HA is primarily a passive structural component in the matrix of a few connective tissues and in the capsule of certain strains of bacteria to a recognition that this ubiquitous macromolecule is dynamically involved in many biological processes: from modulating cell migration and differentiation during embryogenesis (4) to regulation of extracellular matrix organization and metabolism (5) to important roles in the complex processes of metastasis, wound healing, and inflammation (6, 7). Further, it is becoming clear that HA is highly metabolically active and that cells focus much attention on the processes of its synthesis and catabolism. For example, the half-life of HA in tissues ranges from 1 to 3 weeks in cartilage (8) to ,1 day in epidermis (9). In this report, we describe recent advances that provide exciting new insights into the biosynthetic side of these metabolic processes.

In 1934, Meyer and Palmer isolated a novel, high M r glycosaminoglycan from the vitreous of the eye (1). They showed that this substance contained a hexuronic acid, an amino sugar, and no sulfoesters and proposed the name hyaluronic acid (hyaluronan, HA), 1 from the Greek hyaloid (vitreous) and uronic acid. It took 20 years before Weissmann and Meyer (2) finally established the precise structure of the repeating disaccharide unit of hyaluronic acid (GlcA␤(133)GlcNAc␤ (134)). The number of repeating disaccharides in an HA molecule can exceed 30,000, a M r Ͼ10 7 . MedLine surveys for reports describing the structure, synthesis, degradation, and biology of HA reveal a steadily increasing interest in this biopolymer during the four decades following the determination of its structure: 790 papers published from 1966 to 1975; 2200 from 1976 to 1985; over 3300 from 1986 to 1996. During this time, HA has been identified in virtually every tissue in vertebrates and has achieved widespread use in various clinical applications, most notably and appropriately as an intra-articular matrix supplement (3) and in eye surgery. This period has also seen a transition from the original perception that HA is primarily a passive structural component in the matrix of a few connective tissues and in the capsule of certain strains of bacteria to a recognition that this ubiquitous macromolecule is dynamically involved in many biological processes: from modulating cell migration and differentiation during embryogenesis (4) to regulation of extracellular matrix organization and metabolism (5) to important roles in the complex processes of metastasis, wound healing, and inflammation (6,7). Further, it is becoming clear that HA is highly metabolically active and that cells focus much attention on the processes of its synthesis and catabolism. For example, the half-life of HA in tissues ranges from 1 to 3 weeks in cartilage (8) to Ͻ1 day in epidermis (9). In this report, we describe recent advances that provide exciting new insights into the biosynthetic side of these metabolic processes.

HA Biosynthesis
It is now clear that a single protein utilizes both sugar substrates to synthesize HA (10). The abbreviation HAS, for the HA synthase, has gained widespread support for designating this class of enzymes and should now be accepted as standard nomenclature. Markovitz et al. (11) successfully character-ized the HAS activity from Streptococcus pyogenes and discovered the enzyme's membrane localization and its requirements for sugar nucleotide precursors and Mg 2ϩ . Prehm (12) found that elongating HA, made by B6 cells, was digested by hyaluronidase added to the medium and proposed that HAS resides at the plasma membrane. Philipson and Schwartz (13) also showed that HAS activity cofractionated with plasma membrane markers in mouse oligodendroglioma cells. HAS assembles high M r HA that is simultaneously extruded through the membrane into the extracellular space (or to make the cell capsule in the case of bacteria) as glycosaminoglycan synthesis proceeds. This mode of biosynthesis is unique among macromolecules since nucleic acids, proteins, and lipids are synthesized in the nucleus, endoplasmic reticulum/Golgi, cytoplasm, or mitochondria. The extrusion of the growing chain into the extracellular space would also allow unconstrained polymer growth, thereby achieving the exceptionally large size of HA, whereas confinement of synthesis within a Golgi or post-Golgi compartment could limit the overall amount or length of the polymers formed. High concentrations of HA within a confined lumen could also create a high viscosity environment that might be deleterious for other organelle functions.
In 1983, Prehm (14) proposed a novel mechanism for HA synthesis that was distinctly different from that for other glycosaminoglycans, such as chondroitin sulfate and heparan sulfate. These latter glycosaminoglycans are elongated on core proteins by transfer of an appropriate sugar from a sugar nucleotide onto the nonreducing terminus of a growing chain (15). However, Prehm (14) proposed that HA synthesis occurs at the reducing terminus of a growing HA chain by a two-site mechanism (Fig. 1). In this mechanism, the reducing end sugar of the growing HA chain (either in the GlcNAc or GlcA site) would remain covalently bound to a terminal UDP, and the next sugar to be added from the second site would be transferred as the UDP-sugar onto the reducing end sugar with displacement of its terminal UDP. The HA chain would then be in the second site. This unusual mode of synthesis does not occur with the eukaryotic heparan sulfate synthase (16) or the bacterial, K5 (17), or K4 (18) polysaccharide synthases, each of which utilizes the same nucleotide sugar substrates, and it remains to be verified using purified recombinant HAS.
Several studies attempted to solubilize, identify, and purify HAS from strains of Streptococci that make a capsular coat of HA as well as from eukaryotic cells (11-13, 19 -21). Although the streptococcal (19,20) and murine oligodendroglioma enzymes (21) were successfully detergent-solubilized and studied, efforts to purify an active HAS for further study or molecular cloning remained unsuccessful for decades. Prehm and Mausolf (19) used periodate-oxidized UDP-GlcA or UDP-GlcNAc to affinity label a protein of ϳ52 kDa in streptococcal membranes that co-purified with HA. This led to a report (22) claiming that the Group C streptococcal HAS had been cloned, which was unfortunately erroneous. This study failed to demonstrate expression of an active synthase and may have actually cloned a peptide transporter. Triscott and van de Rijn (20) used digitonin to solubilize HAS from streptococcal membranes in an active form. van de Rijn and Drake (23) selectively radiolabeled three streptococcal membrane proteins of 42, 33, and 27 kDa with 5-azido-UDP-GlcA and suggested that the 33-kDa protein was HAS. As shown later (24,25), however, HAS actually turned out to be the 42-kDa protein.
Despite these efforts, progress in understanding the regulation and mechanisms of HA synthesis was essentially stalled, since there were no molecular probes for HAS mRNA or HAS protein. A major breakthrough occurred in 1993 when DeAngelis et al. (24,25) reported the molecular cloning and characterization of the Group A streptococcal gene encoding the protein HasA, known to be in an operon required for bacterial HA synthesis (26), although the function of this protein, which we now propose to designate as spHAS (the S. pyogenes HAS), was unknown. spHAS was subsequently proven to be responsible for HA elongation (see below) and was the first glycosaminoglycan synthase identified and cloned and then successfully expressed (10). The S. pyogenes HA synthesis operon encodes two other proteins. HasB is a UDP-glucose dehydrogenase, which is required to convert UDP-glucose to UDP-GlcA, one of the substrates for HA synthesis (26). HasC is a UDP-glucose pyrophosphorylase, which is required to convert glucose 1-phosphate and UTP to UDP-glucose (27). Co-transfection of both hasA and hasB genes into either acapsular Streptococcus strains or Enteroccus faecalis conferred them with the ability to synthesize HA and form a capsule (24,25). This provided the first strong evidence that HasA is an HA synthase.

The HAS Family
The elusive HA synthase gene was finally cloned by a transposon mutagenesis approach (24), in which an acapsular mutant Group A strain was created containing a transposon interruption of the HA synthesis operon. Known sequences of the transposon allowed the region of the junction with streptococcal DNA to be identified and then cloned from wild-type cells. The encoded spHAS (25) was 5-10% identical to a family of yeast chitin synthases and 30% identical to the Xenopus laevis protein DG42 (developmentally expressed during gastrulation (28)), whose function was unknown at the time. DeAngelis and Weigel (10) expressed the active recombinant spHAS in Escherichia coli and showed that this single purified gene product synthesizes high M r HA when incubated in vitro with UDP-GlcA and UDP-GlcNAc, thereby showing that both glycosyltransferase activities required for HA synthesis are catalyzed by the same protein, as first proposed in 1959 (11). This set the stage for the almost simultaneous identification of eukaryotic HAS cDNAs in 1996 by four laboratories revealing that HAS is a multigene family encoding distinct isozymes. Two genes (HAS1 and HAS2) were quickly discovered in mammals (29 -34), and a third gene has now been found (35). Fig. 2 compares the predicted amino acid sequences of spHAS, human HAS1 and HAS2, mouse HAS1, HAS2, and HAS3, and frog HAS.
Further, preliminary studies 2 have also identified the authentic HAS gene from Group C Streptococcus equisimilis (seHAS); the seHAS protein has a high level of identity to the spHAS enzyme. Membranes prepared from E. coli expressing recombinant seHAS synthesize HA when both substrates are provided. These results confirm that the earlier report of Lansing et al. (22) claiming to have cloned the Group C HAS was wrong. Unfortunately, several studies have employed antibody to this uncharacterized 52-kDa streptococcal protein to investigate what was believed to be eukaryotic HAS (36 -41). In view of subsequent developments, it must be considered that these reports reached incorrect conclusions and should be reexamined, since they were not in fact studying HAS.
Itano and Kimata (29) used expression cloning in a mutant mouse mammary carcinoma cell line, unable to synthesize HA, to clone the first putative mammalian HAS cDNA (mmHAS1). Subclones defective in HA synthesis fell into three separate classes that were complementary for HA synthesis in somatic cell fusion experiments, suggesting that at least three proteins are required. Two of these classes maintained some HA synthetic activity, whereas one showed none. The latter cell line was used in transient transfection experiments with cDNA prepared from the parental cells to identify a single protein that restored HA synthetic activity. Sequence analyses revealed a deduced primary structure for a protein of ϳ65 kDa with a predicted membrane topology similar to that of spHAS (25). mmHAS1 is 30% identical to spHAS and 55% identical to DG42. The same month this report appeared, three other groups submitted papers describing cDNAs encoding what was initially thought to be the same mouse and human enzyme. However, through an extraordinary circumstance, each of the four laboratories had discovered a separate HAS isozyme in both species. Using a similar functional cloning approach to that of Itano and Kimata (29), Shyjan et al. (30) identified the human homolog of HAS1. A mesenteric lymph node cDNA library was used to transfect murine mucosal T lymphocytes that were then screened for their ability to adhere in a rosette assay. Adhesion of one transfectant was inhibited by antisera to CD44, a known cell surface HA-binding protein, and was abrogated directly by pretreatment with hyaluronidase. Thus, rosetting by this transfectant required synthesis of HA. Cloning and sequencing of the responsible cDNA identified hs-HAS1. Itano and Kimata (31) also reported a human HAS1 cDNA isolated from a fetal brain library. The hsHAS1 cDNAs reported by the two groups, however, differ in length; they encode a 578 (30) or a 543 (31) amino acid protein. HAS activity has only been demonstrated for the longer form.
Based on the molecular identification of spHAS as an authentic HA synthase and regions of near identity among DG42, spHAS, and NodC (a ␤-GlcNAc transferase nodulation factor in Rhizobium), Spicer et al. (32) used a degenerate RT-PCR approach to clone a mouse embryo cDNA encoding a second distinct enzyme, which is designated mmHAS2. Transfection of mmHAS2 cDNA into COS cells directed de novo production of an HA cell coat detected by a particle exclusion assay, thereby providing strong evidence that the HAS2 protein can synthesize HA. Using a similar approach, Watanabe and Yamaguchi (33) screened a human fetal brain cDNA library to identify hsHAS2. Fulop et al. (34) independently used a similar strategy to identify mmHAS2 in RNA isolated from ovarian cumulus cells actively synthesizing HA, a critical process for normal cumulus oophorus expansion in the pre-ovulatory follicle. Cumulus cell-oocyte complexes were isolated from mice immediately after initiating an ovulatory cycle, before HA synthesis begins, and at later times when HA synthesis is just beginning (3 h) or already apparent (4 h). RT-PCR showed that HAS2 mRNA was absent initially but expressed at high levels 3-4 h later suggesting that transcription of HAS2 regulates HA synthesis in this process. Both hsHAS2 and mmHAS2 are 552 2 K. Kumari and P. Weigel, manuscript in preparation. FIG. 1. Proposed mechanism of HA synthesis. The repeating disaccharide (shown in brackets) is synthesized by extension of the polymer at the reducing end via a two-site mechanism (14), as described in the text. Definitive evidence for this unusual mode of saccharide synthesis using purified, recombinant HAS has not yet been obtained. amino acids in length and are 98% identical. mmHAS1 is 583 amino acids long and 95% identical to hsHAS1, which is 578 amino acids long.
Most recently Spicer et al. (35) used a PCR approach to identify a third HAS gene in mammals. The mmHAS3 protein is 554 amino acids long (Fig. 2) and 57, 71, 56, and 28% identical, respectively, to mmHAS1, mmHAS2, DG42, and spHAS. Spicer et al. (42) have also localized the three human and mouse genes to three different chromosomes (HAS1 to hsChr 19/mmChr 17; HAS2 to hsChr 8/mmChr 15; HAS3 to hsChr 16/mmChr 8). Localization of the three HAS genes on different chromosomes and the appearance of HA throughout the vertebrate class suggest that this gene family is ancient and that isozymes appeared by duplication early in the evolution of vertebrates. The high identity (ϳ30%) between the bacterial and eukaryotic HASs also indicates a common ancestoral gene. Perhaps primitive bacteria usurped the HAS gene from an early vertebrate ancestor before the eukaryotic gene products became larger and more complex. Alternatively, the bacteria could have obtained a larger vertebrate HAS gene and deleted regulatory sequences nonessential for enzyme activity.
The discovery of X. laevis DG42 by Dawid and co-workers (28) played a significant role in these recent developments, even though this protein was not known to be an HA synthase. Nonetheless, that DG42 and spHAS were 30% identical was critical for designing oligonucleotides that allowed identification of mammalian HAS2 (32,34). Ironically, definitive evidence that DG42 is a bona fide HA synthase was reported only after the discoveries of the mammalian isozymes, when DeAngelis and Achyuthan (43) expressed the recombinant protein in yeast (an organism that cannot synthesize HA) and showed that it synthesizes HA when isolated membranes are provided with the two substrates. Meyer and Kreil (44) also showed that lysates from cells transfected with cDNA for DG42 synthesize elevated levels of HA. Now that its function is known, DG42 can, therefore, be designated xlHAS (Fig. 2).  (30,31,33), mouse (29,32,34,35), bacteria (25), and frog (28) are shown. Asterisks denote residues conserved in all family members. The shaded and boldface areas are amino acids identical to the smallest member of the family shown, spHAS. These presumably include any specific residues necessary to create an active HAS. Solid dots above the sequences indicate those amino acids conserved in all members of the broader ␤-glycosyltransferase family including NodC, chitin synthases, and cellulose synthase (48). The diamond denotes the Cys residue (Cys-225 in spHAS) conserved within the family. The approximate midpoints predicted for MD1-MD7 are indicated above the sequences. Three or more possible whole or partial HA-binding domains (B-X 7 -B, where B is a basic residue (46)) and many potential phosphorylation sites are present in all family members (not shown). The carboxyl region of the central domain, Ala-145-Pro-317, in spHAS is particularly conserved, with few gaps, in the family. hs, Homo sapiens; mm, Mus musculis; sp, S. pyogenes; xl, X. laevis.

Structural Features of the HAS Proteins
(63% of the total protein) and 307-328 residues long in the eukaryotic HAS members (54 -56% of the total protein). The exact number and orientation of membrane domains and the topological organization of extracellular and intracellular loops have not yet been experimentally determined for any HAS. spHAS is the only HAS family member to date that has been purified and partially characterized (10). Initial studies using spHAS/alkaline phosphatase fusion proteins indicate that the N terminus, C terminus, and the large central domain of spHAS are, in fact, inside the cell (45). spHAS has 6 cysteines, whereas HAS1, HAS2, and HAS3 have 13, 14, and 14 Cys residues, respectively. Two of the 6 Cys residues in spHAS are conserved and identical in HAS1 and HAS2. Only one conserved Cys residue is found at the same position (Cys-225 in spHAS) in the HAS family members shown in Fig. 2. This may be an essential Cys whose modification by sulfhydryl poisons partially inhibits enzyme activity. 2 The possible presence of disulfide bonds or the identification of critical Cys residues needed for any of the multiple HAS functions noted below has not yet been elucidated for any members of the HAS family.
In addition to the proposed unique mode of synthesis at the plasma membrane, the HAS enzyme family is highly unusual in the large number of functions required for the overall polymerization of HA. At least six discrete activities could be present within the HAS enzyme: binding sites for each of the two different sugar nucleotide precursors (UDP-GlcNAc and UDP-GlcA), two different glycosyltransferase activities, one or more binding sites that anchor the growing HA polymer to the enzyme (perhaps related to a B-X 7 -B motif (46)), and a ratchetlike transfer reaction that moves the growing polymer one sugar at a time. This later activity is likely coincident with the stepwise advance of the polymer through the membrane. All of these functions, and perhaps others as yet unknown, are present in a relatively small protein ranging in size from 419 (spHAS) to 588 (xlHAS) amino acids.
Although all the available evidence supports the conclusion that only the spHAS protein is required for HA biosynthesis in bacteria or in vitro, it is possible that the larger eukaryotic HAS family members are part of multicomponent complexes (47). Since the eukaryotic HAS proteins are ϳ40% larger than spHAS, their additional protein domains could be involved in more elaborate functions such as intracellular trafficking and localization, regulation of enzyme activity, and mediating interactions with other cellular components.
The unexpected finding that there are multiple vertebrate HAS genes encoding different synthases strongly supports the emerging consensus that HA is an important regulator of cell behavior and not simply a structural component in tissues.
Thus, in less than 6 months, the field moved from one known, cloned HAS (spHAS) to recognition of a multigene family that promises rapid, numerous, and exciting future advances in our understanding of the synthesis and biology of HA.
FIG. 3. Proposed membrane topology for the HAS family. Very similar hydropathy plots and primary structure (28 -71% identity) among all the HAS isozymes suggest that they are similarly organized within the membrane. The scheme depicts the N and C termini and the large central domain, between MD2 and MD3, inside the cell. The larger eukaryotic HASs (thick line) have additional amino acids in all regions (see Fig. 2) compared with the bacterial HASs (thin line), except for the highly conserved carboxyl 178 residues of the central domain and MD1-MD5. In particular, the carboxyl ϳ25% of the eukaryotic HASs has two additional predicted membrane domains (MD6 and MD7), missing in the bacterial proteins. The conserved Cys is indicated by the circled C. MD5 can be modeled as an amphipathic helix, which would orient the C terminus of all HAS members inside.