All chemicals were purchased as reagent grade and used without further purification. Anhydrous dichloromethane (CH₂Cl₂) was purchased from a commercial source without further distillation. Pulverized MS 4Å (Aldrich) for glycosylation was activated by heating at 350°C for 10 h. Reactions were monitored by analytical thin-layer chromatography (TLC) in EM silica gel 60 F254 plates and visualized under UV (254 nm) and/or by staining with acidic ceric ammonium molybdate or p-anisaldehyde. Flash chromatography was performed on a silica gel (Merck) of 40–63 μm particle size. ¹H NMR spectra were recorded on a Bruker AVANCE 600 (600 MHz) spectrometer at 25°C. Chemical shifts (in ppm) were assigned relative to CDCl₃ (δ = 7.24 ppm) and D₂O (δ = 4.80 ppm). ¹³C NMR spectra were obtained using a Bruker AVANCE 600 spectrometer and were calibrated with CDCl₃ (δ = 77.00 ppm). Coupling constants (J) are reported in hertz (Hz). Splitting patterns are described by using the following abbreviations: s, singlet; brs, broad singlet; d doublet; brd, broad doublet; t, triplet; q, quartet; dt, doublet of triplets; tt, triplet of triplets; qt, quartet of triplets; m, multiplet. High-resolution ESI-TOF was recorded on a Bruker Daltonics spectrometer.

Benzyl α-L-fucopyranoside (100 mg, 0.4 mmol) and dibutyltin oxide (107 mg, 0.432 mmol) were dissolved in 30 ml of methanol and refluxed for 2 h. After evaporation of the solvent, the residue was dried under vacuum and then dissolved in 2.5 ml of toluene. After the solution was cooled to 0°C for 5 min, a solution of acetic anhydride (0.432 mmol, 40.8 μL) in anhydrous DCM (0.1 ml) was added dropwise and then allowed to react at room temperature for 8 h. The resulting mixture was directly purified by flash column chromatography, yielding the acetylated products.

Benzyl α-L-fucopyranoside (1.0 g, 4 mmol) and dibutyltin oxide (1.07 g, 4.32 mmol) were dissolved in 300 ml of methanol and refluxed for 2 h. After evaporation of the solvent, the residue was dried under vacuum and then dissolved in 25 ml of solvent (toluene oracetonitrile). After the solution was cooled to 0°C for 5 min, a solution of benzoyl chloride (4.32 mmol, 502 μL) in anhydrous DCM (1 ml) was added dropwise and then allowed to react at room temperature for 8 h. The resulting mixture was directly purified by flash column chromatography, yielding the benzoylated products.

To a solution of the suitably O-protected benzyl glycoside carrying two unprotected OH groups at C2/C3, C2/C4, or C3/C4 (0.2 g, 0.6 mmol) in DCM (5 ml) was added pyridine (0.269 ml, 3.35 mmol) at -20°C. Trifluoromethanesulfonic anhydride (0.281 ml, 1.67 mmol) in DCM (1 ml) was added dropwise, and the mixture was stirred while allowing to warm from -20 to 10°C for over 2 h. The resulting mixture was subsequently diluted with DCM (50.0 ml) and washed with 1 M HCl (10.0 ml), aqueous NaHCO₃ (10.0 ml), water (10.0 ml), and brine (10.0 ml). The organic phase was dried over Na₂SO₄ and concentrated in vacuum at low temperatures. The residue was used directly in the next step without further purification.

TBAN₃/TBANO₂ (5.0 equiv.) was all added at once to a solution of the protected triflate residue in dry toluene (5 ml) at 25°C. After stirring at room temperature (25°C) for 8 h, the mixture was diluted with EtOAc (50.0 ml) and washed with water (10.0 ml). The organic phase was dried with MgSO₄ and concentrated in vacuum. Purification of the residue by flash column chromatography (10:1, hexanes/ethyl acetate) afforded the inversion products.

To a solution of compound 31 (100 mg, 0.3 mmol) in DCM (5.0 ml) was added pydine (79.5 μL, 0.987 mmol) at 0°C. After the solution was cooled to 0°C for 5 min, a solution of benzoyl chloride (0.493 mmol, 57.3 μL) in anhydrous DCM (0.1 ml) was added dropwise and then allowed to react at room temperature for 8 h. The resulting mixture was subsequently diluted with DCM (50.0 ml) and washed with 1 M HCl (10.0 ml), aqueous NaHCO₃ (10.0 ml), water (10.0 ml), and brine (10.0 ml). The organic phase was dried over Na₂SO₄ and concentrated in vacuum at low temperatures. The residue was used directly in the next step without further purification. The benzoylated residue was dissolved in pyridine (1.5 ml), and thioacetic acid (3.0 ml) was added.(Khedri et al., 2014). The reaction was allowed to stir at room temperature for 12 h and monitored by TLC. The reaction mixture was concentrated in high vacuum to remove the pyridine and thioacetic acid. The residue was run through a short silica gel column to remove the lower polarity impurities (EA/Hexane, 1/4 to 1/1). The residue was concentrated and directly dissolved in MeOH (10 ml), and then NaOMe (20 mg) was added. The reaction was allowed to stir at room temperature for 6 h. After the benzoyl group had been removed completely, the reaction was neutralized by IR120, filtered, and concentrated in vacuum. Purification of the residue by flash column chromatography afforded 84 mg of the di-acetamido product S2 with 76% yield. To the solution of S2 in MeOH (10 ml) was added 10% Pd-C (84 mg). The resulting mixture was hydrogenated on a hydrogenation apparatus for 12 h under a hydrogen atmosphere (2-3 psi). The resulting suspension was filtered and concentrated. Silica gel column purification (DCM/MeOH = 10:1, by column) produced the desired product 40 (59 mg, 97% yield).

The reaction was carried out in 2 ml Tris–HCl buffer (100 mM, pH 7.5) containing compound 40 (10 mg), sodium pyruvate (5.0 equiv.), and PmNanA (P. multocida sialic acid synthase) (Li et al., 2008) (0.2 mg) by incubating at 37°C for 12 h with agitation (140 rpm). The product formation was monitored by thin-layer chromatography (eluent: butanol: AcOH: H₂O = 5 : 3: 2) (Supplementary Figure S1). When no additional product formation was observed, the reaction was quenched by adding an equal volume of cold 95% EtOH and incubating on ice for 30 min to precipitate the protein followed by centrifugation to remove the precipitates. The supernatant was concentrated by rotary evaporation, and the product was purified using a C-18 column.

The reaction was carried out in 2 ml Tris–HCl buffer (150 mM, pH 7.5) containing compound 40 (10 mg), phosphoenolpyruvate (5.0 equiv.), MgCl₂ (20 mM), and NeuB3 (pseudaminic acid synthase) (Chou et al., 2005) (0.2 mg) by incubating at 37°C for 15.5 h with agitation (140 rpm). The product formation was monitored by thin-layer chromatography (eluent: butanol: AcOH: H₂O = 5: 3: 2) (Supplementary Figure S1). When no additional product formation was observed, the reaction was quenched by adding an equal volume of cold 95% EtOH and incubating on ice for 30 min to precipitate the protein followed by centrifugation to remove the precipitates. The supernatant was concentrated by rotary evaporation, and the product was purified using a C-18 column.

Previous studies on inversion reactions in the synthesis of 2,4-di-azido-2,4-di-deoxy-α-L-Rha recommended harsh conditions (110°C in DMF/HMPA); however, the yield was not satisfactory (58%/69%), and no detailed explanations were given (Sanapala and Kulkarni, 2016a; Flack et al., 2020). In our study, using 3-O-acetyl-α-L-Fuc 1 as reactant and TBAN₃ (5.0 equiv.×2) in the double inversion process resulted in a mixture of 2,4-di-azido-3-O-acetyl-2,4,6-trideoxy-α-L-Glu 2 and 2,4-di-azido-3-O-acetyl-2,4-di-deoxy-α-L-Rha 3 (Table 1, Entry 1, NMR ratio 51/49, 2: 3). To our surprise, a better result was obtained when 3-O-benzoyl-α-L-Fuc 4 was used instead as the reactant. In this case, 2,4-di-azido-3-O-benzoyl-2,4-di-deoxy-α-L-Rha 6 was the major product of the process (Table 1, Entry 3, NMR ratio 38/62, 5: 6). We think the main reason for this behavior is most likely due to the formation of competing acetoxonium ion intermediate 9 in the reaction pathway (Scheme 1). Specifically, when the triflates of compound 8 was shifted far toward the ⁴ C ₁ chair form, an anti-di-axial relationship is established between the ester and the triflate group at the 2-position, and the undesirable 9 forms quickly to compete with the attacking azide anion in a nonpolar solvent. For the ¹ C ₄ chair form of compound 8, the acetoxonium ion intermediate 9 also could be obtained slowly. In this case, nucleophilic attack on the acetyloxonium ion from the equatorial/axial 2-position is preferred, opening the five-member-ring to produce L-Glu 2. Attack at the equatorial/axial 3-position appeared to be unfavored because no 2-O-acetyl-3,4-di-azido-3,4,6-tri-deoxy-α-L-Alt 10 was identified in the reaction. Moreover, the anti-di-equatorial configuration between the triflate and ester is less likely to produce an intermediate benzoxonium ion leading to the double inversion (DI) as the major instead of neighboring group participation (NGP). For substrate 7, reaction with 4-OTf does not have any complicating side reaction. Furthermore, acetyloxonium ion intermediate is unlikely to form because of the 3,4-cis configuration.

Ramström’s group concluded that the neighboring group’s participation would compete with inversion when ester protecting groups were used (Pei et al., 2005). The formation of acetoxonium ion intermediate 9 is slower in a nonpolar solvent (toluene) than in a polar solvent (DMF), favoring an intermolecular S _N 2 reaction (NGP), which is dependent on the nucleophile concentration. As we expected, the yield of 3 was dramatically increased when the concentration of TBAN₃ was increased close to the saturation limit (20.0 equiv. × 2), resulting in near quantitative formation of 3 (Table 1, Entry 2, 95% isolated yield). Furthermore, in contrast to the high nucleophile concentration for the double inversion of α-L-Fuc 1 in toluene, the α-L-Rha 6 was formed almost quantitatively for α-L-Fuc 4 in TBAN₃ (8.0 equiv. × 2, Table 1, Entry 4, which was less than half of the saturation limit, 20.0 × 2 equiv.). Even when the gram-scale reaction of 3-O-benzoyl α-L-Fuc 4 (1g/3g/5 g) was performed, this double inversion strategy still managed a high yield (95%/93%/90%) of α-L-Rha 6. In the 5 g scale, the intermediate product 2-O-triflate-3-O-benzoyl-4-azido-2,4,6-tri-deoxy α-L-Glu 11 was isolated in 5% yield, agreeing with what we proposed in Scheme 1. As a result, the neighboring group participation (NGP) was avoided, and the overall yield increased.

To further analyze these findings and explore how reactivity is affected by configurations of triflate and the neighboring ester group, we designed other inversion systems (Table 1). We prepared a range of compounds by protecting one of the hydroxyl groups in the 2-, 3-, or 4-position with a benzoyl group and subsequently tested in the azide-based double inversion reactions. Our experimental results (Table 1) indicated that a range of configurations provided high yields of diazide products. Exceptions can be found in entries 7–9 wherein the precursor vicinal diol bears the trans-di-equatorial configuration. This may be the result of initial S_N2 displacement followed by conformational shift and antiperiplanar elimination. Scheme 2 depicts plausible pathways to elimination products. In addition, Entry 11 provided poor results.

In Scheme 2, we discussed completely the mechanisms of entries 7–9. All of these involve vicinal di-equatorial triflates with a neighboring benzoyl group. The Vicinal Triflate Effect (Hale et al., 2014) had been used to explain the reactivity of the inversion reaction, which suggested an α-trans-relationship between C (2/3/4-carbon)-triflates and a vicinal axially oriented strong electron-withdrawing group, indicating poor S _N 2 displacements. Based on the Vicinal Triflate Effect, when the conformation shifted, the α-trans-relationship between C (3)-OTf and the benzoyl group was formatted, which led to antiperiplanar elimination (Schemes 2A,B). For Entry 9, when the conformation shifted, the intermediate II could not format the α-trans-relationship between C (4)-OTf and the neighboring group (Scheme 2C). Meanwhile, the C (3)-proton in the α-trans-axial relationship with C (4)-triflate led to an antiperiplanar elimination reaction, which competed with the triflate S _N 2 displacement, and resulted in product 24.

More interestingly, we thought compound 25 was likely to form a α-trans-axial relationship between the C (2)-triflate and the C (3)-benzoyl group, which would perform poorly in S _N 2 displacement. However, the double inversion product 26 was obtained in good yield (79%), suggesting that the ⁴ C ₁ chair conformation was exclusively populated for the intermediate of 25 triflate. Such conformation usually engages in S _N 2 displacement readily.

As we discussed previously, the double inversion reaction proceeded smoothly when the ester group was in the α-cis or α-trans-equatorial relationship with the triflate groups (compounds 4 and 12). However, for α-L-Rha 27 and β-L-Rha 28, only the β-isomer 28 provided double inversion product 29 in excellent yield (85%), whereas the α-isomer proved inefficient. Despite the fact, we are still unclear about how anomeric configuration influences the S _N 2 transition state.

To prepare di-NAc sugars to obtain nonulosonic acid precursors, we need to synthesize 6-deoxy-L-AltdiNA 40 (Scheme 4). Fascione’s group used the Lattrell–Dax reaction (Guo and O'Doherty, 2007; 2008) with β-Rha 34 as the substrate and found that the epimerization product 35 was obtained in moderate yield in the TBANO₂/acetonitrile system (46%) (Flack et al., 2020). They also found α-Rha 32 was inefficient for the reaction. Considering these unsatisfactory results, we tested α-Rha 30 and β-Rha 36 in the TBANO₂/toluene system. As we expected, β-Rha 36 gave β-Alt 37 in excellent yield (80%) over two steps, and α-Rha 30 provided the epimerization product α-Alt 31 in moderate yield (25%) (Scheme 3).

For these results, usually, a simple explanation is that axially disposed OBn and developing 1,3-diaxial interactions retard the approach of nucleophile with α-Rha 30 and 32 but not with β-Rha 34 and 36, especially a significant increase in productivity based on solvent effection. Thus, compounds 34 and 36, both with β-anomer, yielded compound 35 (46%) in acetonitrile and compound 37 (80%) in toluene. For compound 30, both inversed nitro-product and nitrite-ester were formed by N/O-attack of nitrite to achieve epimerization of the substrates without a neighboring equatorial ester group. In this case (Figure 1), the nitro-product was unstable and subsequently underwent the antiperiplanar elimination reaction to form double-bond product 39 (in 70% yield).

Based on our discussions on neighboring group participation and solvent effect, we provided a new chemical synthetic strategy to prepare 6-deoxy-L-AltdiNAc 40 with 49% overall yield (Scheme 4). Compared with the 11% overall yield reported by Fascione’s group, our route provides a more efficient chemical synthesis strategy to obtain the important pseudaminic acid precursor 40 for enzymatic synthesized Pse.

The pseudaminic acid was expressed by Pseudomonas species, which exhibited the same connectivity as legionaminic acid (Leg) but in a different stereochemical configuration. Like Neu5Ac, Pse and Leg are well-characterized microbial nonulosonic acids (NulO). Enzymatic synthesis of NulOs generally proceeds by the aldol reaction of a 6-carbon monosaccharide with 3-carbon pyruvate, activation to a CMP-sugar, and incorporation into glycans. Previous studies showed the condensation of phosphoenolpyuvate with N-acetyl-D-mannosamine (ManNAc) was catalyzed by sialic acid synthase to form sialic acid and phosphate (Blacklow and Warren, 1962; Masson and Holbein, 1983; Angata and Varki, 2002). In particular, PmNanA, a sialic acid synthase, exhibited good tolerance toward mannose derivatives in ManNAc modifications (Li et al., 2008; Khedri et al., 2014). In addition, a study also characterized enzymes of pseudaminic acid enzymatic synthesis (Schoenhofen et al., 2006). However, not many studies have been carried out on the conversion of non-natural modified AltdiNAc derivatives by a pseudaminic acid aldolase (PseI)-catalyzed reaction.

Based on the experience from the enzymatic synthesis of the non-natural C-5-, C-7-, and C-8-modified pseudaminic acid derivatives, we can synthesize new, desired isomers of pseudaminic acid for further research. To date, none of the enzymes for enzymatic synthesis of nonulosonic acid precursors have been studied in the aspect of configurations. In order to explore the sialidase substrate specificity for various configurations and to synthesize new nonulosonic acid derivatives enzymatically for further research, we chose compounds 6, 15, 26, and 29 for deprotection to obtain 41, 42, 43, and 44. Individually, the deprotection products were combined with compound 40 and catalyzed by PseI and PmNanA.

For PmNanA, all of diacetamido pyranoside derivatives tested were completely incompatible. Although 2,4-diNAc Lyx 43 and ManNAc share similar configurations, we did not observe effective catalysis of PmNanA. We attributed the ineffectiveness of 2,4-diNAc Lyx 43 to the loss of hydrogen bonding to PmNanA. For PseI, the results (Table 2) indicated the importance of 2- and 4-NHAc group configuration in the reaction. Clearly, Pse (Pse5Ac7Ac) 45 was obtained in excellent yield from 6-deoxy-L-AltdiNAc 40 (in 80% yield). Similar to precursor 40, the only difference was that compound 41 has opposite configuration of C3-OH. Using substrate 41, a new nonulosonic acid derivative 46 was obtained in a 61% isolated yield as the isomer of Pse. Different from precursor 40, compound 44 contains two acetamido groups (on C2 and C4) and one hydroxyl group (on C3) with the opposite configuration of C2, C3, and C4. Similar to compound 44, 43 (like 40) only contains C3-OH. For compounds 43 and 44, although 2- and 4-NHAcs are present, their configuration proved to be inefficient for PseI. For compound 42, the change of the functional group order will further increase the structural complexity to lose binding interactions with PseI. In sum, we found that both the configuration of di-acetamido groups and the placement of acetamido groups were important for PseI substrate specificity. Under this guideline, including the diverse substitution on the two amino groups, diverse pseudaminic acid derivatives could be accessed to elucidate substrate specificities of bacterial, human, and viral pseudaminic acid synthase.