Hyaluronan synthases; mechanisms, myths, & mysteries of three types of unique bifunctional glycosyltransferases

Abstract

Hyaluronan (HA), the essential [-3-GlcNAc-1-β-4-GlcA-1-β-]_n matrix polysaccharide in vertebrates and molecular camouflage coating in select pathogens, is polymerized by “HA synthase” (HAS) enzymes. The first HAS identified three decades ago opened the window for new insights and biotechnological tools. This review discusses current understanding of HA biosynthesis, its biotechnological utility, and addresses some misconceptions in the literature.

HASs are fascinating enzymes that polymerize two different UDP-activated sugars via different glycosidic linkages. Therefore, these catalysts were the first examples to break the “one enzyme/one sugar transferred” dogma. Three distinct types of these bifunctional glycosyltransferases (GTs) with disparate architectures and reaction modes are known. Based on biochemical and structural work, we present an updated classification system. Class I membrane-integrated HASs employ a processive chain elongation mechanism and secrete HA across the plasma membrane. This complex operation is accomplished by functionally integrating a cytosolic catalytic domain with a channel-forming transmembrane region. Class I enzymes, containing a single GT family-2 (GT-2) module that adds both monosaccharide units to the nascent chain, are further subdivided into two groups that construct the polymer with opposite molecular directionalities: Class I-R and I-NR elongate the HA polysaccharide at either the reducing or the non-reducing end, respectively. In contrast, Class II HASs are membrane-associated peripheral synthases with a non-processive, non-reducing end elongation mechanism using two independent GT-2 modules (one for each type of monosaccharide) and require a separate secretion system for HA export. We discuss recent mechanistic insights into HA biosynthesis that promise biotechnological benefits and exciting engineering approaches.

Introduction

Hyaluronan (hyaluronic acid, hyaluronate, HA) is an essential polysaccharide in the extracellular matrix of mammals and found in all chordates from the Amphioxus lancet to humans (Fig. 1). This ‘wonder goo’ plays many roles in development, health, and disease as covered by multiple reviews (Toole 2000; Tammi et al. 2002; Itano et al. 2008; Simpson et al. 2022) with more physiological functions being reported almost monthly. Interestingly, certain microbial animal pathogens also employ an extracellular HA coating (capsule) as a virulence factor; this ‘self molecule appears to serve as molecular camouflage to dampen host immune defenses (Wessels 2019; Guan et al. 2020). In the last three decades, we have made strides from the identification of various HA biosynthesis enzymes, the HA synthases (HASs), to a better understanding of their structure and function (Weigel and DeAngelis 2007). Recent studies provide the first structural snapshots of a HAS at different stages during HA biosynthesis (Maloney et al. 2022). This review presents an updated classification system for HASs, discusses reaction mechanisms and biotechnological utility, and points out some potential technical misconceptions as well as areas ripe for future investigations.

Fig. 1.

Schematic of HA matrix, polymer structure and synthases. A) Eukaryotic cells have an HA polysaccharide coating (along with various proteins, not shown) on their extracellular surfaces (HA matrix and chain) while select microbial cells have an HA-rich capsule. B) HA polysaccharide has a [-3-GlcNAc-1-β-4-GlcA-1-β-]_n repeating structure where n can range up to 10⁴. C) Depending on the species, one of three types of HA synthase enzymes (see Table 1) polymerizes the monosaccharides from UDP-sugar donors into the HA chain that is then secreted or transported into the extracellular space. Class II HASs synthesize HA in vivo on a glycolipid anchor (indicated by a lipid-linked square) generated by enzymes encoded in the corresponding capsular polysaccharide biosynthesis gene operon.

HA synthase basics

General HAS reaction

HASs belong to glycosyltransferase (GT) family-2 (GT-2), a huge conglomeration of glycan-building enzymes, including cellulose, chitin and glycogen synthases as well as several Golgi localized glycoprotein N-glycan transferases (Lombard et al. 2014). HASs synthesize the [-3-GlcNAc-1-β-4-GlcA-1-β-]_n disaccharide repeats of the HA polysaccharide by repetitive addition of the monosaccharide units from uridine diphosphate-activated (UDP) donors (Fig. 1) (Weigel and DeAngelis 2007). This reaction releases UDP as a second reaction product, which has been shown to competitively inhibit some GT-2s in vitro (Tlapak-Simmons et al. 2004; Omadjela et al. 2013; Purushotham et al. 2016). Both UDP-N-acetylglucosamine (GlcNAc) and UDP-glucuronic acid (GlcA) precursors are present in the cytoplasm; their levels and interplay with the synthetic machinery may define the amount and size of the final HA product formed (Zimmer et al. 2022).

In general, Class I enzymes are integrated into the plasma membrane whereas Class II HAS, so far only been identified in Gram-negative bacteria, associates with the cytosolic side of the inner membrane (Jing and DeAngelis 2000). Class I HASs are found in prokaryotes (Gram-positive Group A and C Streptococci), a Chlorella virus, and vertebrates. This group is further classified based on the chain elongation mechanism, as described later (Fig. 2).

HA is an extracellular polysaccharide aiding in extracellular matrix formation (animals) or encapsulation (microbes, virally infected hosts). It is noteworthy that, depending on species and exact tissue, HA chains ranging in length from ~10 kDa to 10 MDa (~25 to 25,000 sugar repeats) are efficiently synthesized and exported during or after biosynthesis.

The known HAS enzymes require divalent metals (e.g. magnesium or manganese) for activity, with the specific cation preference being synthase dependent (DeAngelis et al. 1997; Kumari and Weigel 1997; Pummill et al. 1998; Tlapak-Simmons et al. 2004; Blackburn et al. 2018). As in many types of GTs, the cation complexes the diphosphate group of the UDP-sugar substrates, facilitating transition state formation and/or providing leaving group assistance. A conserved Asp-X-Asp motif contributes to metal coordination. Of note, recent work on the viral HAS revealed a second Mn²⁺-binding site also in proximity to the nucleotide’s diphosphate, and mutagenesis experiments demonstrated its importance for catalytic activity, for hitherto unknown reasons (Maloney et al. 2022).

In all known HASs, it appears that only one sugar unit is transferred to the nascent chain at a time (Fig. 2); this fact has been shown for PmHAS (DeAngelis 1999a,1999b) and viral HAS (Maloney et al. 2022) and agrees well with the chain elongation mechanism employed by cellulose synthase (Morgan et al. 2016). We speculate that this mechanism also applies to the rest of the HAS enzymes. This model is in contrast to some earlier hypotheses on cellulose and HA biosynthesis that assumed that the glycosyl units of the disaccharide units were added simultaneously (Saxena et al. 2001; Carpita 2011; Weigel 2015).

Fig. 2.

Schematic models of HA biosynthesis past and present. A) The first reported model of the formation of HA biosynthesis (adapted from Markovitz et al. 1959; reproduced with permission from J. Biol. Chem.) where alternating attack by incoming UDP-sugars results in the formation of sugar repeats. B) Modern models of the two types of Class I HAS; while both have channels to secrete the identical HA chain, the two types differ in molecular directionality of chain elongation. For I-R synthases, the monosaccharide of the UDP-donor attacks the reducing end of the nascent UDP-HA chain intermediate; these enzymes have been proposed to dimerize to simultaneously bind UDP-linked HA and donor sugars. On the other hand, for I-NR HASs, the terminal monosaccharide at the non-reducing end of the nascent HA chain attacks the UDP-sugar donor.

The overall reaction catalyzed by all known HASs is shown in Equations (1) or (2). The official IUBMB enzyme classification for a HAS is EC 2.4.1.212; however, depending on the particular enzyme involved, there are multiple routes to make the same identical HA polysaccharide.

(1)

(2)

HA is quite unlike the other structurally similar glycosaminoglycans found in vertebrates, heparan sulfate/heparin, chondroitin sulfate, and keratan sulfate (Pedersen et al. 2000). This disparity is because HA is neither made in the Golgi apparatus on a protein core (using different types of peripheral membrane GTs), nor is the chain modified by sulfation or epimerization in post-polymerization processes.

HAS nomenclature

The various HAS proteins are named in a fashion akin to DNA-acting enzymes; the first letters of the genera and species are employed to denote the source organism. Thus, the synthase from Streptococcus pyogenes is SpHAS, and a human enzyme is HsHAS. In the latter case, there are three isozymes named HAS1, 2 and 3, but some extant chordates have four genes (some may be pseudogenes) based on ancestral repeated gene duplication events. One of the initial vertebrate HASs to be identified from the frog Xenopus laevis (DeAngelis and Achyuthan 1996; Meyer and Kreil 1996) was initially named DG42 (for an RNA differentially expressed during gastrulation) (Sargent et al. 1986), but formally would be called XlHAS1.

Enzyme sequence identification

The HA polysaccharide was found to be abundant in certain vertebrate tissues (notably, the seminal discovery in the vitreous humor of mammalian eye by Karl Meyer in the 1930s (Meyer and Palmer 1936)) and the surface coating of Group A and C Streptococci (Kendall et al. 1937). The ‘synthase’ (‘synthetase’ or ‘polymerase’ in some older nomenclatures) activity that generated HA was initially studied with radiolabeled sugar incorporation assays of crude membrane preparations by Albert Dorfman and colleagues starting in the mid-1950s (Markovitz et al. 1959). These pioneering biochemical experiments laid the groundwork for generating Equation (1) and spawned the first mechanistic hypothesis on HA chain elongation in 1959 (Fig. 2A).

The primary sequences of the first HAS enzymes were identified in 1993–1998 by combinations of molecular and genetic methodologies (e.g. transposon insertional mutagenesis, expression cloning, degenerate polymerase chain reaction) because these low abundance membrane proteins were rather recalcitrant to classical analyses (DeAngelis et al. 1993; Itano and Kimata 1996; Spicer et al. 1996; DeAngelis et al. 1998). Indeed, the early reports using biochemical purification, affinity labeling and/or immunoreagents mis-identified (Mian 1986; Prehm 1989; Lansing et al. 1993) or missed pinpointing (van de Rijn and Drake 1992) the authentic HASs.

Three distinct types of HA synthases

A generalized overview of HASs

All known, experimentally confirmed HA synthases are bifunctional GTs that add both the GlcA and GlcNAc components to the nascent HA chain. These findings broke the longstanding glycobiology dogma that ‘one enzyme transfers one type of sugar and forms one type of linkage’ (DeAngelis 1999a, 1999b; Weigel and DeAngelis 2007); it was assumed that a pair of GT enzymes (i.e. a separate polypeptide for each monosaccharide unit) would be necessary to form the disaccharide repeats of HA. Following their discoveries, it became apparent that the HASs from different organisms did not always appear or behave equivalently. Accordingly, biochemical and structural features of these unique enzymes were employed to assign the various HASs into three main types (Weigel and DeAngelis 2007).

Features of the streptococcal HASs

The original prototype HAS from the Gram-positive Streptococcus pyogenes, SpHAS (encoded by the hasA gene), was shown to be a single polypeptide species with one identifiable GT domain that alternates between adding GlcA or GlcNAc units (DeAngelis et al. 1993; Blackburn et al. 2018), but the details of modulating the UDP-sugar binding and elongation reaction are still unclear. The streptococcal HAS is a membrane-integrated protein (predicted to have 4 transmembrane (TM) segments) that also secretes HA across the plasma membrane (Heldermon et al. 2001; Hubbard et al. 2012).

The use of disparate, native enzyme sources in the past gave conflicting answers regarding the general elongation mechanism for the HASs. Streptococcus HAS elongated HA on its reducing end (Class I-R) based on pulse-chase labeling reactions with an Escherichia coli-derived recombinant enzyme as dissected by exoglycosidase digestion time courses (Tlapak-Simmons et al. 2005). This method was further refined by the use of recombinant SpHAS (as well as XlHAS1 for a parallel comparison) expressed in Saccharomyces cerevisiae, a heterologous expression host lacking HA as well as the UDP-GlcA substrate (Bodevin-Authelet et al. 2005). Here, HA biosynthesis could be cleanly initiated in vitro without any background. Other support is derived from mass spectral analyses of early reaction products during synthesis in vitro with purified recombinant SeHAS (Streptococcus equisimilis) enzyme (Weigel et al. 2015). Further, in contrast to non-reducing end polymerization, HA elongation at the reducing end requires its activation by covalent coupling to a nucleotide (Equation (2)). Indeed, using isotope labeling, the SeHAS-produced HA polymer was shown to be covalently attached to UDP, thereby supporting the reducing-end elongation mechanism (Tlapak-Simmons et al. 2005; Blackburn et al. 2018).

Chitin oligosaccharides ([-4-GlcNAc-1-β-]_n where n = ~2–6) fused onto HA chains as a non-reducing end cap were also reported (Weigel et al. 2015). HA derived from native Streptococcus that was digested with hyaluronidase also appeared to have chitin caps at the non-reducing termini. These findings imply that (i) UDP-GlcNAc is the initial acceptor for SeHAS, (ii) this group is extended by sequential addition of several GlcNAc units in the early stages, and then (iii) followed by disaccharide repeat addition to make the very large HA chain. The role(s) of the theorized chitin cap in vivo remains unknown and so is the molecular cue that switches from chitin to HA biosynthesis. Several functions of a chitin cap have been discussed, including serving as a plug in the membrane to prevent ion leakage during HA synthase maturation or participating in interactions with other proteins (Weigel 2015). Some controversy also exists on whether vertebrate HASs can make chitin as well (Varki 1996). At this point, it is unclear whether the proposed chitin cap is indeed necessary for in vivo HA biosynthesis or may be the result of particular experimental conditions.

Additional insights into the Class I-R reaction mechanism came from biochemical and biophysical analyses of SeHAS (Blackburn et al. 2018). Initially, streptococcal enzymes were hypothesized to function as monomers based on radiation inactivation analysis (Tlapak-Simmons et al. 1998); the actual target size reported was ~50% larger than the predicted polypeptide (but less than the postulated dimer), therefore, a complex of the protein monomer with ~16 ‘essential lipid molecules’ was invoked in their model. Later, single particle photobleaching and biochemical experiments showed that the SeHAS enzyme functions as a homodimer (Blackburn et al. 2018). This finding was supported by the purification of a mixed complex of wildtype and inactive mutant SeHAS, which lacked catalytic activity, thus suggesting that an SeHAS homodimer is the biolofically functional unit. According to this updated model, it is predicted that the GT domains of a dimer would bind the UDP-attached HA polymer as well as the UDP-sugar substrate (Fig. 2B).

Although initially proposed under the assumption of a monomeric functional polypeptide unit, the “pendulum hypothesis” (Weigel 2015) agrees well with the above-described dimeric functional unit of Class I-R enzymes. One protomer would interact and position the UDP-attached reducing end of HA (see Equation (2)), while the other binds the incoming UDP-donor substrate. Glycosyl transfer would than attach the nascent HA chain to the substrate molecule, such that the HA-attached UDP originates from the last incorporated substrate. According to this reaction scheme, the nascent HA chain functions as the donor and the substrate molecule as the acceptor of the glycosyl transfer reaction (Fig. 2B). Further information derived from structural biology will be required to fully elucidate the molecular details of the Class I-R HASs.

Features of the vertebrate and viral HASs

Within several years of the SpHAS discovery, the vertebrate (HAS1,2,3 isozymes (Weigel et al. 1997)) and viral (CvHAS of Paramecium bursaria Chlorella virus 1, PBCV-1 (DeAngelis et al. 1997)) HASs were identified. The vertebrate HASs produce HA of the glycocalyx and extracellular matrix. HA is essential in mammals. Identifying the viral enzyme as an authentic HAS was the first GT found to be encoded by a virus and the first instance of this ‘animal’ polymer in a plant-like organism.

These enzymes display similarities in overall TM architecture, with two TM helices on the N-terminal and four on the C-terminal side of the cytosolic GT domain (Figs 2B and 3). The animal and viral HASs share the features of a single GT-2 catalytic domain that operates as in Equation (1), but their reaction mechanism is inverted compared to streptococcal HAS (Equation (2)). Here, new sugar units are added to the non-reducing (NR) end of the nascent chain (Fig. 2B) (Hoshi et al. 2004; Blackburn et al. 2018; Maloney et al. 2022), thus these catalysts are termed Class I-NR HASs.

Fig. 3.

Structural models of the various HASs. The experimentally determined structure of CvHAS (PDB: 7SPA; Class I-NR) is shown as a cartoon with TM helices colored in gray, interface helices in blue, and the catalytic GT-2 domain in magenta and cyan for beta-strands and helices, respectively. The priming GlcNAc monosaccharide is shown with orange spheres for carbon atoms. AlphaFold2 (AF) models are shown for HAS from Homo sapiens (Hs; Class I-NR), Streptococcus equisimilis (Se; Class I-R), and Pasteurella multocida (Pm; Class II). The putative HA translocation pathways are shown as a blue dashed line. SeHAS is likely to form an HA translocation channel at a dimer interface (the second HAS polypeptide is shown in light blue to the left 3D structure), but further work is needed to confirm. Aspartate residues (the first Asp in D-X-D motif) implicated in substrate binding are shown and labeled for all species. The indicated Trp (W) residues mark the acceptor position above the Class I catalytic pockets.

This NR end reaction mode is typical for the vast majority of known GT enzymes and contrasts the reducing-end elongation mechanism employed by streptococcal HASs described above. The major difference is that in NR end elongation, the nascent HA chain serves as the acceptor of the transfer reaction by mediating a nucleophilic attack on the substrate’s donor sugar.

Recent cryo electron microscopy (EM) and functional analyses of viral HAS provided new insights into the initiation stage of HA biosynthesis (Fig. 3) (Maloney et al. 2022). First, it was shown that the enzyme’s GT domain tightly interacts with a channel-forming TM region to couple HA synthesis with secretion. Second, biochemical and structural data revealed that HA biosynthesis starts from a GlcNAc monosaccharide primer hydrolytically released from the substrate UDP-GlcNAc. GlcA, however, does not initiate biosynthesis. Third, because the enzyme’s catalytic pocket and TM channel are juxtaposed, HAS can ‘hold on’ to the nascent polysaccharide between elongation steps, thereby enabling processive chain elongation.

Another important finding from these cryo EM studies is that the viral enzyme functions as a monomer. Despite using large amounts of purified enzyme for these studies, no evidence of homo-oligomerization was observed, in contrast to trimerization of plant cellulose synthases (Purushotham et al. 2020). This model is in agreement with the previous radiation inactivation data on XlHAS1 (Pummill et al. 2001); in these studies, the comparison of native and fusion protein constructs allowed internal calibration of the functional target size as a monomer. However, recombinantly over-expressed human HASs were reported to form homo- (HAS2-HAS2) and heterooligomers (HAS1-HAS2 and HAS3-HAS1) by FRET imaging and immunoprecipitation assays (Bart et al. 2015). Therefore, more studies are needed to identify the presence and physiological relevance of HAS oligomers during HA biosynthesis in vivo in various species.

Class I HASs form an HA secretion channel

Based on the predicted transmembrane segments of SpHAS and the observation that a single gene conferred HA production, it was hypothesized that the enzyme could transport HA through a membrane pore (Heldermon et al. 2001). In vitro HA biosynthesis assays with proteoliposome reconstituted enzymes demonstrated that SeHAS and CvHAS are both necessary and sufficient to translocate HA across the vesicle membrane during synthesis (Hubbard et al. 2012; Blackburn et al. 2018). Structural analyses of CvHAS revealed that the enzyme creates a continuous TM channel upon priming with a GlcNAc monosaccharide (Maloney et al. 2022). The channel is electropositive and lined with several polar residues that could contact HA during secretion. In an extended conformation, about ten glycosyl units, i.e. five HA disaccharide repeat units, would span the TM channel (Fig. 3).

HA likely adopts multiple conformations in an aqueous environment such as the extracellular matrix. Spectroscopic analyses, modeling, and high-resolution structural insights showed when bound to some proteins, the acetamido and carboxylate groups of an HA repeat unit are positioned on the same side of the disaccharide (Almond et al. 2006; Banerji et al. 2007). How the polymer chain migrates through HAS’s TM channel is unknown to date. However, the CvHAS channel is sufficiently wide to allow conformational flexibility during translocation. Further high-resolution structural insights into HA translocation intermediates are necessary to determine the precise translocation mechanism.

A revised HA biosynthesis model for Class I-NR enzymes

The structural and functional analyses of CvHAS suggest a revised working model for HA biosynthesis (Fig. 4). First, the enzyme initiates HA biosynthesis by binding and hydrolyzing a UDP-GlcNAc substrate molecule. The released GlcNAc monosaccharide is stabilized at the acceptor binding site near the entrance to a TM channel. The primer’s acetamido group ensures the selectivity of the catalytic pocket for UDP-GlcA, leading to the formation of a GlcNAc-GlcA disaccharide. Prior or concomitant to binding of the next substrate molecule (UDP-GlcNAc), the disaccharide unit has to translocate into HAS’s TM channel to position GlcA at the acceptor site. In this position, GlcA’s carboxylate group likely prevents binding of a UDP-GlcA substrate molecule. The process of stepwise elongation and translocation continues until HA’s final size has been reached. By an unknown mechanism, HA biosynthesis terminates and the polymer can be released into the extracellular space. It is also possible that HAS remains associated with the completed HA chain for some time, thus acting as an anchor on the cell surface. However, this behavior would block further production of new HA chains by the same enzyme molecule. Thus, a more likely scenario is that matrix proteins tie up the HA polysaccharide in the extracellular space to limit diffusion (Day and Prestwich 2002).

Fig. 4.

Model of HA biosynthesis by Class I-NR enzymes. The postulated steps of polysaccharide initiation and production in order from left to right. HAS generates a monosaccharide primer by hydrolyzing a UDP-GlcNAc substrate. Substrate selectivity is determined by the nature of the accepting glycosyl unit. GlcNAc’s acetamido group prevents binding of a UDP-GlcNAc substrate and a GlcA acceptor is incompatible with binding of a UDP-GlcA substrate. The control of HA chain elongation and termination is not yet understood.

Features of the Pasteurella HASs

In another HAS surprise, a totally different enzyme was identified in another bacterial pathogen in 1998, the Gram-negative Pasteurella multocida, PmHAS (encoded by the hyaD gene) (DeAngelis et al. 1998). PmHAS has a primary sequence dissimilar from streptococcal, vertebrate, or viral HASs. Based on sequence analysis and mutagenesis, PmHAS possesses two GT-2 domains, one for each sugar unit that is transferred (Jing and DeAngelis 2000; Jing and DeAngelis 2003) in contrast to the single GT-2 of the Class I enzymes. The GlcNAc-Tase and GlcA-Tase appeared to act independently, resulting in non-processive polymerization. While no experimental 3D structures are available for PmHAS, a crystal structure for the similar E. coli chondroitin synthase, KfoC, was reported (Osawa et al. 2009), revealing two separate domains, each with a UDP-sugar binding site, in agreement with earlier biochemical and mutagenesis studies of PmHAS (Jing and DeAngelis 2003). Radiation inactivation studies for PmHAS indicate that it functions as a monomer (Pummill et al. 2001).

Although the last HAS to be discovered, PmHAS was the first to be absolutely verified to operate by non-reducing end elongation (DeAngelis 1999a, 1999b). In this case, the enzyme’s propensity to use artificial acceptors allowed clear proof that one GT domain of PmHAS added a sugar unit to the nascent chain on the non-reducing terminus. Then a second sugar is added to the new terminus via the other GT domain, and so on.

The core catalytic region of the PmHAS enzyme (i.e. residues 1–703) is not membrane-integrated and truncated recombinant versions are water soluble (Fig. 3) (Jing and DeAngelis 2000). Instead, the PmHAS’ carboxyl terminus (residues ~704–972) likely anchors the enzyme peripherally to the plasma membrane. It is important to note that PmHAS functions in vivo in concert with other membrane-associated enzymes to produce and secrete a lipid-linked capsular HA polymer (Whitfield et al. 2020). During this process, the nascent HA polymer is attached to a membrane-bound glycolipid (Fig. 1C). The corresponding secretion machinery (called HexA, -B, -C, -D) consists of an ABC transporter (HexAB) integrated into the inner bacterial membrane, a periplasmic but inner membrane-anchored subunit (HexC), as well as an outer membrane component (HexD) with a large periplasmic domain. All of these genes are encoded in the same capsule biosynthesis locus as the HAS gene (reviewed in (Caffalette et al. 2020).

It is conceivable that PmHAS interacts with the transport machinery to spatio-temporally couple HA biosynthesis with export. While the directionality of HA secretion from the interior of the cell to the extracellular capsule by HexABCD has not been established, it is possible that the transporter engages with the lipid anchor while the nascent HA chain is extended by PmHAS from the non-reducing terminus. This lipid-linked sugar, however, is not required for HA polymerization in vitro. From a biotechnological standpoint, the non-processive nature of PmHAS allows useful chemoenzymatic syntheses for the preparation of either defined oligosaccharides or quasi-monodisperse polysaccharides (reviewed in DeAngelis et al 2013) that are beyond the capabilities of any other currently known HAS.

An updated classification system

Based on biochemical and structural work outlined above, we present an updated classification system (Table 1) derived from a previous HAS review in 2007 (Weigel and DeAngelis 2007) that includes three distinct types of HASs:

Table 1.

Updated HA synthase classification system.

	Class I		Class II
	I-R	I-NR
	SpHAS, SeHAS (& allies; CPS1, BcHAS)	CvHAS (XlHAS, MmHAS, HsHAS isozymes)	PmHAS (& allies)
Number of GT-2 modules (UDP-sugar binding sites)	One	One	Two
Predicted membrane topology	Integral	Integral	Peripheral (soluble)
Intrinsic transmembrane channel	Yes	Yes	No
HA chain growth directionality	Reducing	Non-reducing	Non-reducing
Intrinsic polymerization mode	Processive	Processive	Non-processive

The various known HASs have distinct biochemical and structural characteristics classified into three distinct types. The parentheses indicate HASs that have not been definitively confirmed, but are quite likely to be grouped in these classes. Class I: (i) the same GT domain (CAZY GT-2) adds two different monosaccharide units, and (ii) possess a transmembrane channel for HA secretion. This class is subdivided based on chain elongation directionality into either: I-R: addition of incoming sugar to reducing end, or I-NR: addition of incoming sugar to non-reducing end. Class II: two fused, but independent GT-2 domains; no intrinsic transport functionality. (Sp, Streptococcus pyogenes; Se, S. equisimilis; Cv, Chlorella virus PBCV-1; Xl, Xenopus laevis, frog; Mm, Mus musculus, mouse; Hs, Homo sapiens, human; Pm, Pasteurella multocida; CPS1, HAS candidate from Cryptococcus neoformans; BcHAS, hasA candidate from Bacillus cereus).

• (a) Class I transmembrane synthases employing processive chain elongation that is subdivided into two groups, Class I-R (reducing end addition; Group A and C Streptococci) and Class I-NR (non-reducing end addition; vertebrates and PBCV-1 virus), and

• (b) Class II peripheral membrane synthases exhibiting intrinsic non-processive elongation (non-reducing end addition; Pasteurella multocida Type A and close allies in the Avibacterium genus).

The major updates from the 2007 review include: vertebrate HAS are now combined with the viral HAS based on (i) the directionality of chain elongation and (ii) the use of a single GT-2 domain to polymerize two different monosaccharides. Both Class I catalysts couple HA synthesis with membrane translocation, but Class II does not. Sequence similarity among the three HAS types is likely due to functional convergent evolution using the GT-2 domain structure as discussed in more detail later as well as in previous reviews (DeAngelis 2002; Weigel and DeAngelis 2007).

Myths: equivalence, transport and inhibitor

We have made great strides understanding HASs, but some theoretical misconceptions and a few details in the past literature will be very problematic if used as guidance, especially by novices to the field.

Myth 1: all HAS enzymes are the same

By definition, all HAS enzymes produce the same polysaccharide repeat unit. However, as explained above, the details of the initiation event (priming) and the directionality of extension (and thereby the reaction mechanism) differ between the three known HAS types.

Considering the available structural and biochemical data on bacterial and viral HASs, we can state that the two Class I types, I-R and I-NR are truly different. Analyses of viral HAS undoubtedly show that the enzyme functions as a monomer and elongates the non-reducing end of the HA polymer. AlphaFold2 predicted models of human HAS reveal a striking structural similarity with CvHAS, hence a similar reaction mechanism most likely applies to the vertebrate enzymes as well. In contrast, a wealth of biochemical data confirmed that streptococcal HASs elongate HA on the reducing end and may function as a homo-oligomer, thereby representing mechanistically distinct enzymes. Therefore, researchers need to keep in mind that insights gained from the bacterial enzymes cannot always be applied directly to the viral or vertebrate synthases and vice versa.

In summary, various GTs were most likely independently evolved multiple times, rather than being passed between various organisms. Therefore, the three HAS synthase types need to be considered as non-identical and caution is paramount during extrapolation of findings and paradigms.

Myth 2: the Class I enzymes require an ABC transporter to secrete HA to make a capsule or matrix

The hypothesis that an ABC transporter is required for Class I-enzyme mediated HA secretion (Ouskova et al. 2004; Prehm and Schumacher 2004) seems to have been put to rest. The discoverers of the streptococcal enzyme initially proposed that an intrinsic pore would form in the membrane based on the assumptions that (i) only a single hasA gene encoding a protein with transmembrane segments was needed to confer HA biosynthesis to hosts that lacked the HA synthases (Heldermon et al. 2001) and (ii) the HAS enzyme activity (either native or recombinant derived enzyme) in vitro was dependent on lipids (Triscott and van de Rijn 1986).

The ultimate proof of a coupled synthesis and translocation reaction came from in vitro studies on purified SeHAS (Hubbard et al. 2012). It was demonstrated that SeHAS, when reconstituted into proteoliposomes and exposed to substrates added to the ‘outside’ of the vesicles, produced HA that accumulated inside the vesicles. In the absence of any other proteinaceous components, these experiments demonstrated (a) that SeHAS is necessary and sufficient for HA synthesis and translocation, and (b) no external energy source (such as ATP or the proton motive force) is necessary for HA secretion.

Myth 3: 4-methylumbelliferone (4-MU) is a “specific HAS inhibitor”

The major mechanism of action for 4-MU is the depletion of the cytoplasmic pools of UDP-sugar donors of HAS. The exogenously added 4-MU is glucuronidated instead of the monosaccharide being attached to UDP (Kakizaki et al. 2004), thereby depleting one of HAS’ substrates (‘no precursors, no polymer’). A direct effect on the action of the HAS has not been shown. Therefore, alternate terminology for 4-MU may be “HA biosynthesis inhibitor” or “cytoplasmic UDP-GlcA metabolic suppressor.” The heparan sulfate/heparin and chondroitin sulfate GAGs are made in the Golgi, with specific nucleotide-sugar transporters ensuring substrate availability inside the organelle. Accordingly, their biosynthesis levels have not been reported to be appreciably affected by 4-MU. Furthermore, while 4-MU is an approved drug in some countries for biliary disease, its use as a HA suppressing agent for other diseases such as cancer must recognize that the reagent will have pleiotropic, off-target effects as well; recent reports indicate various perturbations that can occur (Karousou et al. 2023; Tsitrina et al. 2023). At this point, no selective potent HAS inhibitors have been reported, but perhaps future in silico studies or library screens with HASs will prove fruitful for leads toward a useful laboratory reagent or novel therapeutic.

Biotechnology: in vivo recombinant HA production systems

The various HA synthases have been harnessed to yield two types of recombinant polymer production systems based on either: a) in vivo growth of engineered cells with HAS genes, or b) in vitro chemoenzymatic syntheses with purified HAS enzymes (not discussed here, but reviewed in DeAngelis et al. 2013). In general, in vivo systems are used to produce large amounts of HA in safer, more efficient hosts.

For most microbial hosts, in addition to HAS, only one other cytosolic enzyme, UDP-glucose-6-dehydrogenase (converts UDP-glucose to UDP-GlcA) is required for HA polymerization. All eubacteria make the other needed substrate, UDP-GlcNAc, for their cell walls, thus it is typically available in sufficient quantities. Gram-positive microbes are often preferred as they do not make endotoxin, a problematic contaminant for medical devices and medicines. For metabolic engineering beyond direct HAS gene manipulation, two major strategies to increase HA production levels are (a) to increase levels of the required donors by altering the promoters and the catalysts of the UDP-sugar synthesis pathways, and (b) to inactivate other non-HA metabolic pathways to avoid funneling off the donor sugars to non-target products. However, empirical testing is needed as the bioengineering results are not always predictable.

The streptococcal HAS was first employed in an industrial host, Bacillus subtilis, a Gram-positive microbe with GRAS (generally regarded as safe) status in chemically defined media (Widner et al. 2005). Numerous other heterologous host species, including a range of bacteria, yeast and plants (Yoshimura et al. 2015; Cheng et al. 2016; de Oliveira et al. 2016; Chahuki et al. 2019; Nazeri et al. 2021) have since been employed. Yields were increased by manipulating the streptococcal HAS expression system (e.g. promoters, codon optimization, mutants) as well as the host’s metabolic pathways (DeAngelis 2012; Cerminati et al. 2021) (Yu and Stephanopoulos 2008) and membranes (Westbrook et al. 2018). A study focused on the SeHAS enzyme’s carboxyl terminus (predicted to be located inside the cytoplasm) altered HA product size (Yang et al. 2017); it was hypothesized that this region could be involved in chain binding/retention, thus altering processivity. This result potentially supports a model for size control based on relative affinities of the synthase for HA where weaker binding yields lower molecular weight (MW) chains and stronger interactions yield higher MWs (Weigel and DeAngelis 2007).

The recombinant streptococcal HAS in bacterial membranes, while not truly in vivo production, has also been used to generate ‘HA brushes’ on surfaces (e.g. glass, silica) with megaDalton polymer chains (Wei et al. 2019) that are controlled by UDP-sugar or metal ion availability and stabilized by chemical cross-linking. Such brushes can be regenerated several times as well as be micro-patterned with lasers. With further refinements, such systems may be used in future biomaterials or nanoparticles, but one obstacle is that any bacterial enzyme will be immunogenic if introduced into patients.

Over the years, production systems with other HASs including vertebrate HAS isozymes and PmHAS have been reported that generate HA in various chain sizes and yields (Mao et al. 2009; Jia et al. 2013; Jeong et al. 2014; de Oliveira et al. 2016; Gomes et al. 2019; Nazeri et al. 2021). However, to date, the absolute yields generally are not as high as for the streptococcal systems.

Smaller MW HA chains have been reported to exhibit different and often antagonistic biological effects compared to larger chains, establishing the need for good sources of such materials. One approach is the use of a host co-expressing hyaluronidases to produce shorter HA fragments; the initially large chains are subsequently digested by the glycosidase action (Jin et al. 2016). This strategy must be calibrated to obtain the desired target size, otherwise mostly di- or tetrasaccharides (the limit digest size, depending on the hyaluronidase species) are obtained.

New Frontiers and remaining questions

While the HAS field has made significant progress, much is still unknown and there are vistas worth exploring. Some of the “new areas” posed fifteen years ago (Weigel and DeAngelis 2007) remain unsolved, so hopefully in the following decades these questions will not remain.

Molecular details on the Class I HAS single domain substrate recognition/usage

A pair of glucose-based sugars, GlcNAc and GlcA, is transferred by the same HAS active site to two different ring positions on the nascent HA chain. How does this catalytic center make the repeating HA chain with the observed high fidelity? Some crosstalk between HA’s terminal glycosyl unit and the incoming substrate must exist to ensure (a) alternating polymerization of GlcNAc and GlcA and (b) regio- and stereo-selective glycosyl transfer. Once this is better understood and if the catalyst is malleable, perhaps the creation of novel designer polysaccharides beyond HA with artificial sugar units and/or linkages will be possible.

Explanation of Class II efficiency

PmHAS is a non-processive, yet speedy elongation catalyst based on observations that the wild-type bifunctional enzyme had the same efficiency in vitro as a pair of mutant catalysts with either single-action GlcNAc-Tase or GlcA-Tase activity (Jing and DeAngelis 2000). Is this simply kinetic optimization with the substrates or do guide elements (e.g. molecular steering, surface channels) assist the non-reducing terminus transiting between the two GT sites? We know that the optimal acceptors in vitro possess 3 to 4 sugar units (Williams et al. 2006) and assume that the binding site size is comparable, but beyond that, little is known.

Polymer size control

While initiation and elongation has been partly explained, the ultimate control of chain length by the intrinsic properties of the Class I HAS enzymes (not by the simple control of substrate UDP-GlcA levels) observed in vitro as well as in the living cell is still unsolved. Several hypotheses were presented in 2007 (Weigel and DeAngelis 2007) and none have been either substantially strengthened or diminished since then in a coherent, comprehensive fashion. Likewise, the nature of HA size control for the Class II PmHAS in vivo is not known yet; is the HAS itself or the transport system or both components responsible?

Post-translational modification and trafficking of Class I-NR HASs

Several reports of mammalian HAS modification (e.g. phosphorylation, glycosylation) have been made (Vigetti and Passi 2014; de Mera et al. 2019; Kasai et al. 2020; Caon et al. 2021). However, the functional consequences of these modifications in modulating catalysis, turnover, and/or intracellular transport are not yet completely defined. The modifications appear to be important based on site-specific mutagenesis experiments. Perhaps the combined use of more defined in vitro systems, selective modification inhibitors, and/or fluorophore-tagged HAS probes in vivo will yield better mechanistic insights in the next decade.

How is HA biosynthesis controlled during vesicular trafficking? HA does not appear to significantly accumulate in the ER or Golgi vesicles; if it did so, then the bulky polymer most likely would impair the organelles’ functions. Hence, how is HA biosynthesis confined to the plasma membrane? Perhaps some post-translational modification(s) transiently silence HAS during vesicular trafficking.

Confirmation of other HAS candidates

Based on sequence homologies to the rigorously verified HASs discussed above, other enzyme candidates have been reported. A protein found in some strains of the fungal pathogen Cryptococcus neoformans, CPS1, resembles the S. pneumoniae Cps3S capsular synthase (~26% identity at protein sequence level) and to a slightly lesser degree, the streptococcal HASs. The cryptococcal CPS1 is quite likely to be an authentic HAS based on the reported biochemical tests (e.g. binding by versican, FACE disaccharide gels) (Jong et al. 2007), but further analysis of the polysaccharide product (especially using the HA-specific Streptomyces hyaluronidase and/or NMR) is needed to truly confirm the polysaccharide structure.

In another case, a gene on the pXO1 plasmid in certain virulent Bacillus anthracis strains (e.g. Sterne) was noted for potentially encoding a HAS based on its close similarity to the streptococcal HASs (~52% identity at protein sequence level) (Okinaka et al. 1999). This causative agent of anthrax employs toxins (also encoded on the pXO1 plasmid) and a poly-glutamic acid capsule (encoded on another plasmid) as virulence factors. This SpHAS-like gene has not been authenticated since its initial report, but this is probably due in part to a predicted inactivating frameshift mutation. Later, some very pathogenic strains of B. cereus, a species related to the anthrax pathogen, were found to possess a similar plasmid to pXO1 that also encoded a very similar HAS candidate (~99% identical to the B. anthracis gene) located` in an operon with UDP-glucose dehydrogenase and UDP-glucose pyrophosphorylase like the Group A and C Streptococci, but with a different gene order (Oh et al. 2011). The large capsule in these B. cereus strains was removed by treatment with testicular hyaluronidase or by knockout of the HAS-like gene. Again, more specific tools and/or biophysical analyses will be needed to confirm this polysaccharide’s structure as HA.

In closing, the various HA synthases promise to be exciting subjects for both basic and applied sciences if we remain cognizant of their enzymological similarities and differences as well as stay observant for any new features or behaviors that are sure to emerge.

Author contributions

Paul DeAngelis (Conceptualization [equal], Writing—original draft [lead], Writing—review & editing [equal]), and Jochen Zimmer (Conceptualization [equal], Writing—original draft [supporting], Writing—review & editing [equal])

Funding

Work by PD was supported by federal (NIH, NSF) and state agencies (Oklahoma Center for Advancement of Science and Technology) and the Presbyterian Health Foundation. JZ is an Investigator of the Howard Hughes Medical Institute and acknowledges NIH funding for HA research.

 

Conflict of interest statement: None declared.

Data availability

Alphafold HAS structures available upon request.