Engineer P. multocida Heparosan Synthase 2 (PmHS2) for Size-Controlled Synthesis of Longer Heparosan Oligosaccharides

Abstract

Pasteurella multocida heparosan synthase 2 (PmHS2) is a dual-function polysaccharide synthase having both α1–4-N-acetylglucosaminyltransferase (α1–4-GlcNAcT) and β1–4-glucuronyltransferase (β1–4-GlcAT) activities located in two separate catalytic domains. We found that removing PmHS2 N-terminal 80-amino acid residues improved enzyme stability and expression level while retaining its substrate promiscuity. We also identified the reverse glycosylation activities of PmHS2 which complicated its application in size-controlled synthesis of oligosaccharides longer than hexasaccharide. Engineered Δ80PmHS2 single-function-glycosyltransferase mutants Δ80PmHS2_D291N (α1–4-GlcNAcT lacking both forward and reverse β1–4-GlcAT activities) and Δ80PmHS2_D569N (β1–4-GlcAT lacking both forward and reverse α1–4-GlcNAcT activities) were designed and showed to minimize side product formation. They were successfully used in a sequential one-pot multienzyme (OPME) platform for size-controlled high-yield production of oligosaccharides up to decasaccharide. The study draws attention to the consideration of reverse glycosylation activities of glycosyltransferases, including polysaccharide synthases, when applying them in the synthesis of oligosaccharides and polysaccharides. The mutagenesis strategy has the potential to be extended to other multifunctional polysaccharide synthases with reverse glycosylation activities to generate catalysts with improved synthetic efficiency.

Graphical Abstract

Heparin (HP) and heparan sulfate (HS) glycosaminoglycans (GAGs) are structurally complex, sulfated linear polysaccharides known to bind to over 300 different human proteins.¹ They are biosynthesized from heparosan, which is a polysaccharide backbone containing a −4GlcNAcα1–4GlcAβ1- disaccharide repeat, followed by orchestrated functions of a group of carbohydrate post-glycosylational modification (PGM) enzymes to produce diverse sulfation and epimerization patterns.^2–3 HP is a clinically used anticoagulant while both HP and HS may have broader potential applications including cancer treatment and protection against virus infections.^4–5 Due to their structure complexity, detailed structure-activity relationships of HP and HS remain unclear.^6–7 Structurally defined synthetic compounds will provide invaluable standards and probes to better understand their important roles and accelerate their applications.

Despite much progress,^8–12 the process for chemical synthesis of HS and HP oligosaccharides is still tedious and challenging especially for longer structure targets. Chemoenzymatic methods are powerful alternatives.^13–18 Glycosyltransferases and polysaccharide synthases involved in the production of heparosan-based capsular polysaccharides have been identified from pathogenic bacteria Escherichia coli K5,¹⁹ Pasteurella multocida Type D (PmHS1),²⁰ and Avibacterium paragallinarum genotype II.²¹ Pasteurella multocida heparosan synthase 2 (PmHS2) which was not part of bacterial capsule biosynthesis loci has been discovered in Pasteurella multocida type A, D, and F strains.²² Both PmHS1²⁰ and PmHS2²² are dual function polysaccharide synthases having both α1–4-N-acetylglucosaminyltransferase (α1–4-GlcNAcT) and β1–4-glucuronyltransferase (β1–4-GlcAT) activities contributed by two separate catalytic domains¹⁴ similar to the case in PmHS1.^{14, 23} Compared to PmHS1, PmHS2 has higher expression levels in E. coli, no requirement for an oligosaccharide primer, and a higher substrate promiscuity.²⁴ PmHS2 has been used broadly in enzymatic synthesis of heparosan polysaccharides,^14–15 and in chemoenzymatic synthesis of HP and HS oligosaccharides and their derivatives with^16–17 or without¹⁸ Escherichia coli K5 α1–4-GlcNAcT (EcKfiA).

Nevertheless, the expression level (17–20 mg/L culture) and the stability of PmHS2¹⁸ can be improved. Additionally, PmHS2 was efficient in synthesizing short oligosaccharides up to hexasaccharides²⁵ but longer glycans were obtained in lower yields (e.g. 35% yield for a decasaccharide).¹⁷ On the other hand, longer HS oligosaccharides (e.g. octasaccharides–dodecasaccharide) were required for strong interaction with some HP-binding proteins.^26–29 For example, decasaccharide was found to be minimum for efficient binding of bone morphogenetic protein-2 (BMP-2).³⁰ We planned to engineer PmHS2 to improve its application in size-controlled chemoenzymatic synthesis of heparosan, HS, HP oligosaccharides and derivatives.

Analyzing PmHS2 protein sequence using XtalPredRF³¹ (Figure S1A) and BLAST (Figure S1B) predicted that its N-terminal 80 amino acid residues might be nonessential to its glycosyltransferase activities and unfavorable for its crystallizability and stability. Indeed, removing the N-terminal 80 amino acid residues of PmHS2 (Figure S2) (17–20 mg/L culture) resulted in Δ80PmHS2 (Figure S3) (60–80 mg/L) with a 3–4-fold improved expression level (Figure S4) and an improved thermal stability (Figure S5). In contrast to PmHS2 which precipitated easily during dialysis, Δ80PmHS2 remained soluble. Δ80PmHS2 could also survive lyophilization without loss of activity (Figure S6). On the other hand, Δ80PmHS2 (pI 6.61) and PmHS2³² (pI 6.83) share similar pH profiles (Figure S7) and donor substrate promiscuities (Figure S8). A previously synthesized stable GlcAβ2AA¹⁸ was used as a fluorescent-labeled acceptor substrate in these assays to allow easy product detection by high-performance liquid chromatography (HPLC) with a diode array or UV/Vis detector.

To facilitate reaction monitoring, product purification, and allow easy removal from the products for downstream conjugation with proteins or other molecules, a fluorophore tag was introduced to two possible monosaccharide substrates. GlcAβProNHFmoc (O1) and GlcNAcαProNHFmoc were synthesized from the corresponding glycosylpropylazides GlcAβProN₃¹⁸ and GlcNAcαProN₃³³ by catalytic hydrogenation followed by coupling with N-(9-fluorenylmethoxycarbonyloxy)succinimide (Fmoc-Suc) (see ESI).

Activity comparison (Table S1) showed that GlcAβProNHFmoc (O1) was a more efficient substrate than GlcNAcαProNHFmoc for both PmHS2 (157-fold) and Δ80PmHS2 (130-fold) in disaccharide production. Apparent kinetics studies with GlcAβ2AA (Table S2) showed that PmHS2 (3.4±0.2 s⁻¹) and Δ80PmHS2 (3.5±0.2 s⁻¹) have similar k_cat values while PmHS2 (4.0±0.6 mM) has a lower K_M value than Δ80PmHS2 (5.3±0.7 mM), resulting in a slightly higher catalytic efficiency for PmHS2 (0.9 s⁻¹ mM⁻¹) than Δ80PmHS2 (0.7 s⁻¹ mM⁻¹).

Starting from GlcAβProNHFmoc (O1), Δ80PmHS2 with improved expression and stability permitted size-controlled gram-scale synthesis of heparosan oligosaccharides ranging from disaccharide (O2) to hexasaccharide (O6) in excellent yields using a sequential one-pot multienzyme (OPME) platform. In this platform (Scheme 1), GlcNAc-activation/transfer (OPME1) and GlcA-activation/transfer (OPME2) systems (each contains Δ80PmHS2, a kinase, a nucleotidyltransferase, and an inorganic pyrophosphatase) were used alternately to extend the acceptor substrate chain one monosaccharide at a time. Each OPME reaction was carried out for 1–2 days, the product was purified and used as the acceptor substrate for the next OPME reaction.

Scheme 1.

Δ80PmHS2-dependent sequential one-pot multienzyme (OPME) synthesis of heparosan oligosaccharides O2–O6.

As shown in Scheme 1, disaccharide GlcNAcα1–4GlcAβProNHFmoc (O2) was enzymatically synthesized from O1 using a one-pot four-enzyme GlcNAc-activation and transfer system (OPME1) containing Bifidobacterium longum N-acetylhexosamine-1-kinase (BLNahK),³⁴ Pasteurella multocida N-acetylglucosamine-1-phosphate uridylyltransferase (PmGlmU),³⁵ Pasteurella multocida inorganic pyrophosphatase (PmPpA),³³ and Δ80PmHS2. The reaction went to completion and the product was readily purified by passing the reaction mixture through a C18 cartridge and eluting with a gradient solution of CH₃CN in water. Purified disaccharide O2 (2.12 g) was obtained in an excellent 96% yield. Trisaccharide GlcAβ1–4GlcNAcα1–4GlcAβProNHFmoc (O3) (1.94 g) was synthesized from purified disaccharide O2 using a one-pot four-enzyme GlcA-activation and transfer system (OPME2) containing Arabidopsis thaliana glucuronokinase (AtGlcAK),³⁶ Bifidobacterium longum UDP-sugar pyrophosphorylase (BLUSP),³⁷ PmPpA, and Δ80PmHS2 with a purified yield of 99%. Repeating the alternate OPME1 and OPME2 reactions with C18-cartridge-based product purification after each OPME reaction led to the formation of tetrasaccharide O4 (1.87 g, 98%), pentasaccharide O5 (1.52 g, 87%), and hexasaccharide O6 (1.60 g, 99%).

OPME reactions of GlcNAc activation and transfer to glucuronides O1, O3, and O5 (OPME1) were highly efficient and N-acetylglucosaminides O2, O4, and O6 were obtained in nearly quantitative yields (96–99%). In the case of OPME reactions of GlcA activation and transfer to N-acetylglucosaminides O2 and O4 (OPME2) for the formation of glucuronides O3 and O5, the reaction with O2 went well and O3 was obtained in an excellent 99% yield. However, when tetrasaccharide O4 (an N-acetylglucosaminide longer than O2) was used as an acceptor for the formation of O5, the presence of O3 byproduct was observed, indicating that the terminal GlcNAc in the acceptor O4 was removed in the reaction. The side product formation was minimized by monitoring the reaction progress carefully and stopping the reaction promptly to obtain O5 in 87% yield.

Overall, the sequential OPME platform containing Δ80PmHS2 was efficient in gram-scale synthesis of heparosan oligosaccharides up to hexasaccharide O6 from monosaccharide O1. When hexasaccharide O6 (an N-acetylglucosaminide even longer than O4) was used as the acceptor substrate for the β1–4-GlcAT activity of Δ80PmHS2, the formation of both longer and shorter oligosaccharide byproducts (Figure S9) was observed, which complicated product purification and lowered synthetic yields.

We hypothesized that the formation of longer and shorter oligosaccharide byproducts in the OPME reaction for the synthesis of O7 from O6 was due to the reverse glycosylation activity of Δ80PmHS2. Such activity was reported for some glycosyltransferases involved in natural product glycosylation,^38–39 mammalian sialyltransferases,⁴⁰ and bacterial sialyltransferases Pasteurella multocida sialyltransferase 1 (PmST1) and Photobacterium damselae α2–6-sialyltransferase (Pd2,6ST),^41–42 but was not observed for others.⁴⁰ Such activity, however, has never been shown for polysaccharide synthases. To test the hypothesis, hexasaccharide O6 was incubated with Δ80PmHS2 in the absence or the presence of different concentrations of uridine 5’-diphosphate (UDP). As shown in Table S3, the amounts of Δ80PmHS2 reverse glycosylation products (oligosaccharides of various sizes ranging from mono- to dodecasaccharide) increased and the concentration of O6 decreased significantly with the increase of UDP concentration. The same effects were observed for PmHS2 (Table S3), demonstrating that the reverse glycosylation property was not introduced by the N-terminal protein sequence truncation in Δ80PmHS2. The chain reactions caused by the forward and reverse glycosylation reactions of Δ80PmHS2 in the presence of O6 and UDP are illustrated in Scheme 2. Starting from O6, the reverse α1–4-GlcNAcT activity of Δ80PmHS2 produces O5 and UDP-GlcNAc. The resulting O5 is used by the reverse β1–4-GlcAT activity of Δ80PmHS2 to produce O4 and UDP-GlcA. On the other hand, the UDP-GlcA obtained is used together with O6 by the β1–4-GlcAT activity of Δ80PmHS2 to produce O7 and UDP. The α1–4-GlcNAcT activity of Δ80PmHS2 uses O7 and UDP-GlcNAc to produce O8 and UDP. Similarly, the newly formed longer and shorter oligosaccharides in the reaction mixture are used as the substrates in combined Δ80PmHS2-catalyzed forward and reverse glycosylation reactions for the formation of additional oligosaccharides of longer and shorter lengths.

Scheme 2.

Schematic illustration of the chain reactions caused by forward and reverse glycosylation activities of Δ80PmHS2 in the presence of hexasaccharide O6 and UDP.

Time course studies using GlcNAc-terminated O6 as the substrate for Δ80PmHS2 (Figure S10) showed that longer incubation times led to a continuous decrease of O6 concentration and increased dispersity of oligosaccharide products with a preference toward the accumulation of GlcNAc-terminated oligosaccharides (O2, O4, O8, and O10). The production of oligosaccharides of different sizes indicated that both α1–4-GlcNAcT and β1–4-GlcAT activities of Δ80PmHS2 have the corresponding reverse glycosylation activities. Indeed, incubating Δ80PmHS2 with GlcA-terminated pentasaccharide O5 in the presence of UDP (Figure S11) also showed a time-dependent increase of the product dispersity although GlcA-terminated oligosaccharide products (O3, O7, and O9) dominated. Increased product dispersity with the increase of time was also observed previously in PmHS2-catalyzed polymerization reactions,^{15, 43} although reverse glycosylation was not realized as a likely major contributor.

The formation of UDP-GlcNAc by reverse glycosylation of O6 using Δ80PmHS2 in the presence of UDP was further confirmed by a coupled enzyme assay (Scheme 3) inspired by Mehr et al. for indirectly detecting CMP-sialic acid formed in a reverse-sialyltransferase reaction.⁴¹ To do this, an additional glycosyltransferase Neisseria meningitidis β1–3-N-acetylglucosaminyltransferase (NmLgtA)⁴⁴ and its acceptor 4-methylumberliferyl β-lactoside (LacβMU)⁴⁵ were added to the reaction of Δ80PmHS2 in the presence of O6 and UDP. As expected (Table S4), GlcNAcβ1–3LacβMU formed by NmLgtA-catalyzed reaction was observed only in the reaction containing all components including NmLgtA, Δ80PmHS2, and UDP (Reaction 1 in Table S4) but not in the one lacking NmLgtA (Reaction 2 in Table S4), Δ80PmHS2 (Reaction 3 in Table S4), or UDP (Reaction 4 in Table S4).

Scheme 3.

Schematic illustration for indirectly detecting UDP-GlcNAc formed in the reverse glycosylation reaction of Δ80PmHS2 in the presence of UDP and hexasaccharide O6 using a coupled enzyme assay with NmLgtA and LacβMU.

Since reverse α1–4-GlcNAcT and β1–4-GlcAT activities of Δ80PmHS2 cause challenges in size-controlled enzymatic synthesis and purification of longer oligosaccharide products, mutagenesis studies of Δ80PmHS2 were planned to generate single-function glycosyltransferases by mutating the key catalytic base residues of the other glycosyltransferase domains. The previously reported strategy of generating single-functional glycosyltransferases of PmHS2¹⁴ and PmHS1²³ by mutating the glycosyltransferase DXD motifs was not adopted due to the significant decrease of the stability of the PmHS2 mutants¹⁴ which would adversely affect their application in synthesis. In the absence of known PmHS2 crystal structures, a Δ80PmHS2 I-TASSER^46–47 model was generated. The β1–4-GlcAT domain in Δ80PmHS2 I-TASSER model aligned well with the GalNAcT and GlcAT domains of Escherichia coli K4 chondroitin polymerase (EcKfoC) (PDB ID: 2Z86 and 2Z87)⁴⁸ (Figure S12), identifying D291 as a possible catalytic base. Similarly, a potential key catalytic residue (D569) in the Δ80PmHS2 α1–4-GlcNAcT domain was identified by aligning its I-TASSER model with the structures of Neisseria meningitidis lipopolysaccharyl-α1,4-galactosyltransferase (NmLgtC)^49–51 and bovine α1–3-galactosyltransferase (α3GalT)⁵² (Figure S13).

D291N and D569N mutants of Δ80PmHS2 were generated (Figure S14). Their expression levels (50–60 and 60–70 mg/L culture, respectively) were comparable to that of Δ80PmHS2 (60–80 mg/L culture). The D291N mutant retained the α1–4-GlcNAcT activity and lost the β1–4-GlcAT activity of Δ80PmHS2. In contrast, the D569N mutant retained Δ80PmHS2 β1–4-GlcAT activity while its α1–4-GlcNAcT activity decreased 548-fold (Figure S15). The corresponding reverse glycosylation activities of Δ80PmHS2 also decreased significantly in the mutants. As shown in Figure S16, under conditions mimicking synthetic reactions with 30 mM UDP (Figures S16A–B), no reverse glycosylation oligosaccharide products were observed for either mutants, demonstrating the efficiency of the mutants in blocking the cascade chain reactions shown in Scheme 2 and avoiding the generation of multiple oligosaccharide byproducts.

Δ80PmHS2_D569N mutant was used as a single function β1–4-GlcAT (lacking both forward and reverse α1–4-GlcNAcT activities) in high-yield OPME synthesis (OPME2) (Scheme 4) of longer GlcA-terminated heparosan oligosaccharides including heptasaccharide O7 (566 mg, 98%) and nonasaccharide O9 (445 mg, 96%), respectively, from the corresponding GlcNAc-terminated Δ80PmHS2_D569N oligosaccharide acceptors O6 and O8. Δ80PmHS2_D291N mutant was used as a single function α1–4-GlcNAcT (lacking both forward and reverse β1–4-GlcAT activities) for high-yield OPME (OPME1) synthesis of longer GlcNAc-terminated oligosaccharides including octasaccharide O8 (504 mg, 99%) and decasaccharide O10 (430 mg, 98%), respectively, from the corresponding GlcA-terminated Δ80PmHS2_D291N oligosaccharide acceptors O7 and O9. In these preparative-scale OPME reactions, unwanted oligosaccharide byproducts were not observed.

Scheme 4.

Sequential OPME synthesis of heparosan oligosaccharides O7–O10.

Various NMR experiments for O1–O10 including ¹H and ¹³C NMR, HSQC, and HSQC-TOCSY (90 ms and 10 ms) enabled signal assignments and the observation of key correlations. HSQC spectra provided C−H coupling information, and HSQC-TOCSY with 90 ms and 10 ms mixing times indicated independent coupling correlations of terminal and internal GlcA or GlcNAc residues. In the example of O3 containing two GlcA residues, the chemical shifts of the internal GlcA are more downfield for H3 (differ by 0.30 ppm), H4 (differ by 0.17 ppm), H5 (differ by 0.07 ppm), C1 (differ by 0.11 ppm), C2 (differ by 0.73 ppm), C3 (differ by 0.34 ppm), C4 (differ by 4.95 ppm), and C5 (differ by 3.18 ppm), but more upfield for H-1 (differ by 0.18 ppm) than that of the terminal GlcA.

In conclusion, N-terminal truncated Δ80PmHS2 with improved expression level and stability was shown to be an efficient catalyst for gram-scale sequential OPME synthesis of heparosan oligosaccharides up to hexasaccharide O6. Reverse glycosylation activities of Δ80PmHS2 were characterized and shown to be responsible for poor yields and complications in Δ80PmHS2-involved OPME synthesis of longer oligosaccharides. Key catalytic base residues for the β1–4-GlcAT and the α1–4-GlcNAcT activities of Δ80PmHS2 were identified. Δ80PmHS2_D569N and Δ80PmHS2_D291N mutants were generated as single functional β1–4-GlcAT and α1–4-GlcNAcT with significantly decreased reverse α1–4-GlcNAcT and reverse β1–4-GlcAT glycosyltransferase activities, respectively. They have been used as efficient catalysts for sequential OPME synthesis of longer length heparosan oligosaccharides (O7–O10). The study draws attention to the consideration of reverse glycosylation activities of glycosyltransferases including polysaccharide synthases when applying them in the synthesis of oligosaccharides and polysaccharides. The mutagenesis strategy has the potential to be extended to other multifunctional polysaccharide synthases with reverse glycosylation activities, especially those use sugar nucleotides containing the same nucleotide component, to generate catalysts with improved synthetic efficiency.