Hyaluronan (HA) is an abundant and essential extracellular matrix polysaccharide in vertebrates. It performs a plethora of developmental and physiological functions with critical contributions to fertilization, cardio- and angiogenesis, wound healing, and joint lubrication ^1–3.

HA is a heteropolysaccharide of alternating glucuronic acid (GlcA) and N-acetylglucosamine (GlcNAc) sugars that reaches several megadaltons in molecular weight under normal tissue homeostasis. It is synthesized from UDP-activated monosaccharides by HA synthase (HAS), a membrane-integrated processive glycosyltransferase (GT) of family-2 by the CAZy classification^4,5. HAS performs two functions: it synthesizes HA via a cytosolic catalytic domain and translocates the nascent HA polymer across the plasma membrane through a channel formed by its own membrane-spanning segment. These reactions are functionally coupled by spatially associating the catalytic and transmembrane (TM) domains⁶.

HAS was among the first GT enzymes with confirmed dual substrate specificity and alternating regio selectivity of the glycosyl transfer reaction⁴. It uses a single catalytic domain that adopts a GT-A fold⁷ to bind either UDP-activated GlcA or GlcNAc and transfers the donor sugar to a GlcNAc or GlcA acceptor, respectively. The different glycosyl transfer reactions form a β-(1,3) linkage between GlcA and GlcNAc and a β-(1,4) linkage between GlcNAc and GlcA.

Recent structural and biochemical analyses of Chlorella virus and Xenopus laevis isoform-1 HAS (CvHAS and XlHAS-1, respectively) provided detailed insights into the multitasking of the enzyme^6,8–10. First, to initiate HA biosynthesis, HAS binds and hydrolyzes UDP-GlcNAc, such that the released GlcNAc monosaccharide can prime polymer biosynthesis. Second, the GlcNAc-primed enzyme binds the second substrate, UDP-GlcA, leading to the formation of a ternary complex that facilitates glycosyl transfer and the formation of a β-linked GlcAβ1-3GlcNAc disaccharide. The subsequent binding of a UDP-GlcNAc substrate molecule presumably translocates the HA disaccharide by an unknown mechanism and facilitates the transfer of GlcNAc to the nascent HA polymer, thereby forming a β-1,4 linkage to GlcA. These steps must be repeated thousands of times to synthesize and secrete an HA polysaccharide of great than 10,000 disaccharide units¹¹.

While UDP-GlcNAc is a common metabolite in all reported HA-producing species, the physiological concentration of UDP-GlcA, however, may limit HA biosynthesis under certain conditions¹². Accordingly, the Chlorella virus encodes a UDP-glucose dehydrogenase enzyme that generates UDP-GlcA from UDP-glucose¹³. Further, expressing this enzyme in engineered HA-producing systems increased the overall production levels^14,15.

In this study, we delineate differences in UDP-GlcA coordination by CvHAS using cryo-electron microscopy (cryo-EM). Enzymology and biochemical analysis were utilized to understand how UDP-GlcA interaction strength and turnover efficiency vary in the presence and the absence of a GlcNAc primer. We further show that CvHAS exhibits a degree of promiscuity toward acceptors for glycosyl transfer, enabling the biosynthesis of unnatural complex carbohydrates. Lastly, our structural analysis reveals a dodecyl maltoside-inhibited state of CvHAS in which the detergent molecule occupies the acceptor-binding site, thereby providing insights into HAS substrate promiscuity and potential inhibitor design.

We used cryo-EM analysis to gain structural insights into the interaction of CvHAS with its substrate UDP-GlcA. The catalytically inactive (D302N) CvHAS mutant was reconstituted into MSPE3D1 lipid nanodiscs as described before⁶ and complexed with high affinity nanobodies that recognize cytosolic and extracellular epitopes. This complex was then incubated with UDP-GlcA:Mn²⁺ in the absence of a receiving GlcNAc monosaccharide, prior to cryo grid preparation. The obtained cryo EM dataset was processed as outlined in Fig. S1.

CvHAS binds UDP-GlcA in two different binding poses – Focused refinement of particles harboring a UDP-GlcA molecule at the active site resolved two substrate binding poses, interpreted as proofreading and inserted states. The inserted pose resembles the previously reported UDP-GlcA conformation in the presence of a priming GlcNAc sugar¹⁰. In this state, the nucleotide’s uracyl moiety is surrounded by Tyr91, His174, and Lys177, its diphosphate group, together with CvHAS’s DxD motif (Asp201 and Asp203), coordinates a divalent cation, most likely Mn²⁺, and the glucuronic acid donor sugar resides in a pocket underneath the acceptor binding site (Fig 1B). GlcA’s ring oxygen resides within hydrogen bonding distance to the Nε of Trp342, and its C6 carboxylate group, although poorly resolved, is positioned near the guanidinium group of Arg341 (Fig 1C and D). The side chain’s Nε is within 3.5 Å of GlcA’s carboxyl group, potentially accounting for a weak electrostatic interaction. An alternative side chain conformation for Arg341 was well-supported by cryo-EM density and thus modeled. In this position, the residue’s guanidinium group is about 4.6 Å from GlcA’s C6 carboxylate (Fig 1D) and likely interacts with it via a mediating water molecule. In addition, the conserved Asp201 forms a salt bridge with Lys177 at the back of the nucleotide binding pocket (Fig. 1C).

Previous analysis identified a second divalent cation binding site in CvHAS’s catalytic pocket⁶. This site is created by Glu93 together with the C-terminal Asp of the DxD motif (Asp203) and appears to be metal occupied in nucleotide-bound and apo states. This is also the case in both of our new cryo EM maps, suggesting that the site is sufficient to coordinate a hydrated cation (Fig S2A-D). Further, this coordination site is required for catalytic activity as replacing Glu93 with Ala renders CvHAS inactive.

In the second ‘proofreading’ UDP-GlcA binding pose resolved in our dataset, the ligand’s nucleotide and donor sugar moieties are less deeply inserted into the catalytic pocket (Fig. 1E-H). The uracil moiety is shifted by about 1.5 Å towards the entrance of the catalytic pocket, resulting in the rotation of His174 away from Tyr91. Further, the nucleotide’s ribose moiety is tilted away from the DxD motif, which places the attached diphosphate group also closer to the binding cleft’s entrance. In this position, the diphosphate is suitably positioned to contribute to cation coordination at the second metal binding site, together with Asp203 and Glu93 (Fig 1E). Indeed, we observe strong cryo EM density at this site, consistent with an octahedrally coordinated manganese cation (Fig S2C and S2D).

The partial insertion of UDP-GlcA in the proofreading pose further allows Asp201 of the DxD motif to move away from its binding partner Lys177 to interact with the ribose’s C2 and C3 hydroxyl groups (Fig 1H). The ammonium group of Lys177, instead, moves towards the donor sugar and, together with the guanidinium group of Arg341, sandwiches GlcA’s carboxylate group (Fig 1F). Both side chains are well resolved in the cryo EM map and about 3.5 Å away from the carboxylate group, which is also resolved at a slightly lower contour level (Fig 1F). Compared to the inserted UDP-GlcA state described above, these interactions create a basic pocket that recognizes GlcA’s carboxylate group.

Next to the interactions with Lys177 and Arg341, GlcA’s carboxylate is further framed by the C-terminal segment of the priming loop that leads into the conserved GDD motif (residues 300-302). The backbone conformation of the priming loop differs slightly in both UDP-GlcA bound conformations. Most notably, while the entire backbone is well resolved in the inserted UDP-GlcA bound pose (Fig S2E), the density of the conserved Gly300 is essentially absent in the proofreading state (Fig S2F), suggesting that this residue and parts of the preceding priming loop are flexible until UDP-GlcA is fully inserted into the catalytic pocket. The conserved Arg348 that belongs to an amphipathic helix at the cytosolic water-lipid interface, interacts with the backbone carbonyl oxygen of Gly300 in the inserted state.

When the priming loop is flexible in the proofreading state, however, Arg348 bends away by about 2 Å to adopt a cation-π stacking interaction with Tyr299 (Fig. 1E). Arginine 348 is conserved among bacterial and eukaryotic HASs, and Tyr299 is conservatively substituted with Phe in vertebrate HASs, suggesting this interaction is preserved across species.

We next thought to validate the contribution of conformational changes in the uracil binding groove to catalytic activity of CvHAS. To this end, we tested the functional relevance of Tyr91 and His174 through site-directed mutagenesis. CvHAS’ activity can be quantified in vitro by measuring the accumulation of tritiated HA by scintillation counting, as previously described¹⁶. Accordingly, replacing Tyr91 with Ala abolishes catalytic activity of CvHAS, while substituting the residue with Phe maintains about 80% activity, relative to the wild-type enzyme (Fig 1I). Similarly drastic effects are observed when replacing His174 with either Ala or Trp, its corresponding substitution in vertebrate HASs. The Ala mutant is essentially inactive while the H174W mutant retains about 25% of wild-type activity (Fig 1I). All mutants purify similar to the wild-type enzyme, suggesting that the substitutions do not affect the overall protein (Fig. S3).

CvHAS binds UDP-GlcA with low micromolar affinity – We performed isothermal titration calorimetry to determine the apparent affinity of UDP-GlcA binding to CvHAS. Titrating UDP-GlcA into a cell containing the catalytically inactive D302N CvHAS mutant at 45 µM revealed saturable exothermic heat responses that could be fit to a single binding curve. By this method, we obtained dissociation constants (K_d) for UDP-GlcA of 69 and 24 µM in the absence and the presence of a GlcNAc primer, respectively (Fig 2A and 2B). Under similar conditions, no binding was observed for UDP-GlcNAc, UDP alone, as well as UDP-glucose (Fig S4). While the lack of a detectable heat signal for UDP-GlcNAc binding is unknown, the apparent lack of UDP and UDP-glucose binding suggests the contribution of UDP-GlcA’s carboxylate to binding.

We next thought to compare substrate turnover efficiency for UDP-GlcA and UDP-GlcNAc under initiation and synthesis conditions. To this end, an enzyme-coupled assay was utilized that links the release of UDP upon glycosyl transfer to the activities of pyruvate kinase and lactate dehydrogenase, as previously described^16,17. First, measurements of UDP-GlcA and UDP-GlcNAc turnover in the absence of an acceptor substrate were performed. We observed an incremental increase in reaction rate in response to UDP-GlcNAc titration, allowing us to derive an apparent K_m of 744 μM. (Fig 2C and S5A).

Second, when titrating UDP-GlcA, no apparent change in substrate turnover rate above background was observed (Fig 2D), as also previously described⁶. This suggests that CvHAS is unable to hydrolyze UDP-GlcA.

Third, a kinetic experiment reflecting the first glycosyl transfer reaction performed by CvHAS was setup. Here, excess GlcNAc was supplemented to generate a primed state and UDP-GlcA was titrated. Under those conditions, a steep kinetic response to substrate titration was observed, with an apparent K_m for UDP-GlcA of 147 μM (Fig 2D and S5B).

Fourth, to compare kinetics of substrate turnover during initiation and processive HA synthesis, a separate set of experiments was performed. First, UDP-GlcNAc was titrated in the presence of excess UDP-GlcA resulting in a response similar to the acceptor-free condition (Fig 2C). However, the maximum reaction velocity at 20 mM UDP-GlcNAc was approximately 25% lower than that measured in the presence of UDP-GlcNAc only (Fig. 2C). Second, titration of UDP-GlcA at an excess of UDP-GlcNAc revealed a constant or even slightly declining substrate turnover rate (Fig 2D). This finding suggests that the UDP-GlcNAc turnover rate remains the same, regardless of whether GlcA or presumably water forms the acceptor of the glycosyl transfer reaction.

To confirm that HA was synthesized under the titration conditions, we visualized the HA product by gel electrophoresis and ‘Stains-All’ staining¹⁰. To this end, synthesis reactions were quenched after 1.5 hours and electrophoresed through an agarose gel, as previously described¹⁰. Previous work showed that HA made by CvHAS appears as a polydisperse distribution in a molecular weight range between 30-300 kDa. Our HA electrophoresis experiment confirmed formation of a similar product, with HA signal initially observed at a UDP-GlcNAc concentration of 625 μM (Fig 2F), and a UDP-GlcA concentration of 156 μM (Fig 2G).

GlcNAc primer affinity – Turnover of UDP-GlcA by CvHAS requires a GlcNAc acceptor, either in the form of a primer or the non-reducing end terminal HA moiety. The enzyme-coupled assay described above allowed us to determine the apparent affinity of CvHAS for the GlcNAc primer. While only minimal UDP-GlcA hydrolysis by CvHAS is observed in the absence of a GlcNAc primer⁶, CvHAS shows a drastic increase in UDP-GlcA consumption in the presence of saturating GlcNAc. This is likely due to the supplemented monosaccharide serving as the acceptor. Accordingly, titrating GlcNAc into a reaction of CvHAS at a constant UDP-GlcA concentration of 2 mM revealed a saturable increase in UDP-GlcA turnover with an apparent K_m of 4.0 mM (Fig 2E and S5C). We note that the apparent maximum catalytic rate observed in the presence of saturating GlcNAc is about twice the rate observed in the presence of saturating UDP-GlcA and UDP-GlcNAc (Fig. 2C and E). While the reaction in the presence of the GlcNAc primer (Fig. 2E) likely generates disaccharides that diffuse away from the enzymes, the formation of a proper HA polymer in the presence of both UDP-activated substrates (Fig. 2C) appears to reduce CvHAS’ overall catalytic rate.

Detergent interactions at the acceptor site – Structural and functional analyses of CvHAS are routinely performed in lipid nanodiscs or the detergent glyco-diosgenin (GDN)⁶. Both conditions support the catalytic activity of the enzyme. However, the protein is initially extracted from membranes in the detergent dodecyl-β-D-maltopyranoside (DDM) that is later removed during the purification procedure (see Methods). In a DDM-solubilized state, CvHAS and XlHAS-1 are catalytically inactive (Fig S6), yet regain activity after nanodisc reconstitution or exchange into GDN detergent.

To our surprise, careful sorting of the UDP-GlcA supplemented cryo EM dataset revealed a CvHAS subpopulation that was not bound to the substrate, but, instead, a DDM molecule near the active site (Fig 3A and S7). This was observed for >50% of the analyzed particles. The first glucosyl unit of the maltoside moiety stacks against Trp342 at the acceptor site. The second non-reducing glucosyl unit protrudes from there into the nucleotide binding cleft, thereby overlapping with the donor sugar binding site (Fig 3B). DDM’s interaction with Trp342 is intriguing because this residue also stabilizes the GlcNAc monosaccharide in the primed state⁶ or the terminal GlcNAc unit in an HA associated state¹⁰. Indeed, the maltose moiety occupies a position similar to the previously described HA disaccharide binding pose obtained after extending a GlcNAc primer with GlcA¹⁰ (Fig. 3C).

Relative to the HA disaccharide-bound state, the reducing end glucosyl unit of DDM is shifted toward the TM channel by about 2 Å (Fig 3C). In this position, its C6 hydroxyl group is in hydrogen bonding distance to Ser345, and the ring oxygen interacts with the guanidinium group of Arg303, which is about 3.1 Å away (Fig 3B). Additionally, a previously described ‘switch’ loop (residues 267-271) at the back of the substrate binding groove⁶, moves approximately 3.6 Å away from the maltose moiety relative to the HA disaccharide-bound structure. This conformational change prevents direct interactions of the switch loop with DDM’s head group (Fig 3C and S8).

Strikingly, DDM’s dodecyl alkyl chain extends into a hydrophobic tunnel that is formed by CvHAS’ transmembrane region. The tunnel is formed by Interface Helices two and three as well as TM helix 3 (Fig 3A). It is open to the lipid bilayer environment which likely allows lipid acyl chains to partially enter the enzyme in a biological membrane. The tunnel is lined by Ser343, Trp346, Cys347, Gln374, Ile375, Phe378, Phe379, Arg442 and Met446 (Fig 3B).

Acceptor promiscuity – The intriguing binding pose of DDM’s maltoside headgroup at the acceptor site prompted us to investigate whether CvHAS accepts other carbohydrates as glycosyl transfer substrates. To this end, we monitored the increase of UDP-GlcA turnover by CvHAS in the presence of selected mono- and disaccharides. As discussed, addition of GlcNAc to a reaction of CvHAS and UDP-GlcA dramatically increases UDP-GlcA turnover, which is monitored using the enzyme-coupled reaction described above.

Under similar conditions, we tested the suitability of the monosaccharides D-galactose, D-glucose, D-mannose, L-rhamnose and L-arabinose to prime CvHAS. None of the monosaccharides tested were able to increase UDP-GlcA turnover above the unsubstituted condition (Fig S9). Next, the disaccharides cellobiose, chitobiose, maltose, sucrose, and xylobiose were tested. Of those, only cellobiose and chitobiose increased UDP-GlcA turnover at elevated concentrations between 10 and 50 mM (Fig 4A).

Interestingly, cellotriose and cellotetraose failed to elicit a similar effect on UDP-GlcA turnover (Fig S9).

Compared to the GlcNAc monosaccharide, the stimulatory effects of chito- and cellobiose are reduced to about 15 and 50%, respectively, at the highest concentrations tested (50 mM). Consistent with previous observations, no increase in UDP-GlcNAc turnover was observed in the presence of added mono- and disaccharides (Fig 4B). Instead, cello- and chitobiose as well as GlcNAc slightly reduced its turnover, compared to the unsubstituted reaction.

Increasing UDP-GlcA turnover in the presence of cellobiose or chitobiose suggests that the disaccharides may serve as acceptors of the GlcA transfer reaction. We employed thin layer chromatography (TLC) to separate the reactants and products and visualized them with thymol stain (see Methods). As shown in Fig. S10A, UDP-GlcA, GlcA, and cellobiose are appreciably separated on Silica Gel 60 plates using a 5:3:2 volume ration of butanol/ethanol/water as solvent. UDP-GlcA shows the least mobility and remains close to the origin position, clearly separated from GlcA. In the presence of cellobiose and UDP-GlcA, a CvHAS product species can be detected that migrates slightly above the UDP-GlcA position (Fig. S10A).

Repeating the TLC analysis with diphenylamine staining instead of thymol provides a different colorimetric readout for cello- and chitobiose adducts (dark blue and burgundy bands). For both disaccharides, we detect putative monosaccharide adducts under reaction conditions that migrate below the cello- and chitobiose species (Fig 4C). The appearance of these products depends on the presence of CvHAS, UDP-GlcA, and the disaccharide, suggesting that the bands indeed represent novel reaction products of CvHAS. Conversely, replacing UDP-GlcA with UDP-GlcNAc as the substrate did not result in the formation of detectable new products with cello- or chitobiose, suggesting that these acceptors can only be extended with GlcA (Fig 4D).

To confirm that the extended disaccharide species indeed contain GlcA, we repeated the synthesis reaction in the presence of ¹⁴C-labeled UDP-GlcA. TLC analysis followed by autoradiography revealed species migrating slightly above and below the GlcA monosaccharide band for chito- and cellobiose, respectively (Fig S10C). While the mobility of the individual bands between TLC experiments shows some variability, our data suggests that cellobiose and chitobiose can indeed serve as GlcA acceptors, albeit at high concentrations.

HA biosynthesis is a multi-step process. It involves the sequential binding of UDP-activated GlcA and GlcNAc to HAS’s catalytic site, the transfer of the donor sugar to an acceptor, and the translocation of the nascent HA polymer across the plasma membrane⁴. Important differences exist in how CvHAS interacts with its substrate. The more abundant UDP-GlcNAc substrate is readily bound and hydrolyzed by apo CvHAS to initiate HA biosynthesis. Turnover of the second substrate, UDP-GlcA, however, is inefficient in the absence of an accepting carbohydrate moiety and does not prime the synthesis reaction.

The resolved proofreading and inserted binding poses of UDP-GlcA suggest that recruitment of this substrate occurs in multiple steps. The proofreading pose, in which the carboxylate is recognized between positively charged residues, may serve to distinguish UDP-GlcA from UDP-Glc, which is the more abundant metabolite. Upon its proper insertion into the catalytic cleft, the subtle reorganization of the surrounding priming loop likely positions and stabilizes UDP-GlcA for glycosyl transfer.

Observing two distinct UDP-sugar binding poses is not entirely unprecedented for processive GTs. A comparison of cryo-EM structures for P. sojae and C. albicans chitin synthase (CHS) in complex with UDP-GlcNAc revealed that CHSs likely utilize an analogous mechanism for substrate insertion^18,19. In one case, the GlcNAc moiety points almost 180 degree away from the active site and must therefore undergo large-scale spatial rearrangements to adopt its proper position for glycosyl transfer (Fig S11).

It is currently unclear why GlcA cannot prime HA biosynthesis. The turnover of UDP-GlcA is substantially increased in the presence of an accepting glycosyl unit, likely due to the positioning of a suitable nucleophile to mediate the attack on the donor sugar. Because the apparent dissociation constants for UDP-GlcA in the absence and the presence of GlcNAc are similar, a GlcNAc primer has only modest effects on substrate binding but significantly affects the stability of the substrate at the active site.

HA biosynthesis requires regio- and stereoselective transfer reactions. Under non-physiological in vitro conditions, however, the enzyme can be coaxed into utilizing alternative acceptors, such as cello- and chitobiose. These disaccharides are unlikely to be physiological substrates as they are absent from the cytosol. Their ability to serve as acceptors is likely due to the conservation of the acceptor binding site. Similar architectures are found in cellulose and chitin synthases that bind β-linked glucose and GlcNAc units, respectively, via the Trp of the conserved QxxRW motif and surrounding residues^18–20. However, some crosstalk between the bound acceptor and UDP-sugar substrate must exist because cello- and chitobiose can only be extended with GlcA, and not with GlcNAc. The observation that non-physiological acceptors can be elongated by HAS may be exploited in the future for the design of novel biomaterials.

CvHAS accommodates a DDM molecule at its acceptor site and an adjacent hydrophobic tunnel. DDM’s dodecyl tail, which likely mimics a phospholipid acyl chain, seals off a lateral connection to the active site. The coordination of DDM’s maltoside moiety, an α-linked glucose disaccharide, is consistent with priming by cellobiose and chitobiose.

However, maltose alone fails to stimulate UDP-GlcA turnover, suggesting that either an α-linked disaccharide alone does not bind to the acceptor site or cannot serve as an acceptor. In the resolved position, DDM blocks a site critical for HAS function. It is unlikely that DDM can be replaced by the HA biosynthesis reaction alone, which may contribute to CvHAS’s catalytic inactivity in a DDM-solubilized state. In the absence of any HAS-specific binder, features contributing to DDM coordination may be exploited in the future for the design of HA biosynthesis inhibitors.