Complete In Vitro Reconstitution of the Apramycin Biosynthetic Pathway Demonstrates the Unusual Incorporation of a β-d-Sugar Nucleotide in the Final Glycosylation Step

Abstract

Apramycin is a widely used aminoglycoside antibiotic with applications in veterinary medicine. It is composed of a 4-amino-4-deoxy-d-glucose moiety and the pseudodisaccharide aprosamine, which is an adduct of 2-deoxystreptamine and an unusual eight-carbon bicyclic dialdose. Despite its extensive study and relevance to medical practice, the biosynthetic pathway of this complex aminoglycoside nevertheless remains incomplete. Herein, the remaining unknown steps of apramycin biosynthesis are reconstituted in vitro, thereby leading to a comprehensive picture of its biological assembly. In particular, phosphomutase AprJ and nucleotide transferase AprK are found to catalyze the conversion of glucose 6-phosphate to NDP-β-d-glucose as a critical biosynthetic intermediate. Moreover, the dehydrogenase AprD5 and transaminase AprL are identified as modifying this intermediate via introduction of an amino group at the 4″ position without requiring prior 6″-deoxygenation as is typically encountered in aminosugar biosynthesis. Finally, the glycoside hydrolase family 65 protein AprO is shown to utilize NDP-β-d-glucose or NDP-4″-amino-4″-deoxy-β-d-glucose to form the 8′,1″-O-glycosidic linkage of saccharocin or apramycin, respectively. As the activated sugar nucleotides in all known natural glycosylation reactions involve either NDP-α-d-hexoses or NDP-β-l-hexoses, the reported chemistry expands the scope of known biological glycosylation reactions to NDP-β-d-hexoses, with important implications for the understanding and repurposing of aminoglycoside biosynthesis.

Graphical Abstract

INTRODUCTION

Apramycin (1) is an aminoglycoside antibiotic^1,2 isolated from Streptoalloteichus tenebrarius, Streptoalloteichus hindustanus, and Saccharopolyspora hirsute^3–6 and is widely used as a veterinary antibiotic and marker of drug resistance in biochemical research. Like other aminoglycosides containing the 2-deoxystreptamine (2-DOS, 5) aglycon (e.g., neomycin, kanamycin, gentamicin, and tobramycin), apramycin targets the bacterial ribosome and inhibits protein synthesis, resulting in antibacterial activity.⁷ The structure of apramycin is characterized by a disaccharide at position-4 of 2-DOS (5) that is composed of an octose (the first sugar) not found in other aminoglycosides and 4-amino-4-deoxy-d-glucose (the second sugar) (see Figure 1). The latter two sugars are linked by an O-glycosidic bond between the C8′ and C1″ anomeric carbons of the octose and the 4-aminosugar, respectively. Aprosamine (3), which is produced via hydrolysis of the 8′, 1″-O-glycosidic linkage of 1, and saccharocin (2), which bears a hydroxyl group instead of an amino group at the 4″-position of the second sugar, both exhibit weaker biological activity than apramycin (1).^5,6,8 Interestingly, aprosamine (3), unlike apramycin (1), can cause aminoglycoside-induced misreading during translation, which has been suggested to be a key factor in aminoglycoside ototoxicity.⁹ Hence, understanding the chemistry that underpins the biosynthesis of the unique structure of apramycin (1) and the functional groups important for its biological activity may provide new insights for the development of next-generation aminoglycoside antibiotics.

Figure 1.

Structure of apramycin and its related compounds.

The apramycin biosynthetic gene cluster (i.e., apr cluster, Figure 2A) was identified in 2004 from S. tenebrarius (GenBank accession: AJ629123.1) and in 2005 from S. hindustanus (GenBank accession: AJ875019.1). The proteins encoded show high sequence homology (over 70% identity) with their counterparts in each cluster. Most of the enzymes encoded by the apr cluster have been characterized, and the biosynthetic pathway to aprosamine 5-phosphate (11) has been established (Figure 2B).^10–20 A key recent discovery is that the formation of the octose unit is catalyzed by AprG in a transaldol reaction using N-acetylglucosamine (GlcNAc) or N-acetylgalactosamine (GalNAc) as the two-carbon donor and 6′-oxo-lividamine (8) as the aldol acceptor (8 → 9).¹⁹ While the periplasm-localized phosphatase AprZ has been demonstrated to catalyze the final hydrolysis of the phosphate group at the 5-position (12 → 1) to give apramycin, a step proposed to facilitate the efflux of apramycin (1) from the cell,²⁰ it remains unclear how the 4-amino-4-deoxy-d-glucose moiety is assembled (11 → 12). As mentioned earlier, the 4-aminosugar moiety plays a crucial role in the biological activity of apramycin (1), thus investigation of the construction of the 4-aminosugar moiety and its attachment to the disaccharide unit is important to fully comprehend the biosynthesis of apramycin.

Figure 2.

(A) Apramycin biosynthetic gene cluster. (B) Biosynthetic pathway for apramycin (1). Annotation of the apr genes is provided in Table S1.

Therefore, the two remaining questions to be addressed in this work are the biosynthetic formation of 4-amino-4-deoxy-d-glucose and the mechanism of its attachment to aprosamine (3). While aminosugars are common structural components of natural products,^21,22 most of them are derivatives of 6-deoxyhexoses, and their formation via 4,6-dehydration catalyzed by a NAD-dependent nucleoside diphosphate (NDP)-sugar 4,6-dehydratase has been well documented.^23,24 However, the aminosugar in apramycin is not a 6-deoxyhexose but a rare 4-amino-4-deoxy-d-glucose which has been found only in 4-trehalosamine²⁵ and analogues of apramycin,⁵ adiposins,²⁶ and sorbistins.^27,28 How this unusual sugar is biosynthesized has never been investigated, but the initial steps clearly do not involve 4,6-dehydration. Second, the coupling of 4-amino-4-deoxy-d-glucose to aprosamine results in an 8′,1″-O-glycosidic linkage between the anomeric carbons of two sugars, either of which can potentially be the sugar donor. Since the glycosyl donor substrate is typically a sugar nucleotide in the glycosylation reaction, it is not apparent whether an NDP-4-aminoglucose or a nucleotidyl derivative at the 8′-position of aprosamine 5-phosphate (11) is the sugar donor that generates the 8′,1″-O-glycosidic linkage in apramycin. Moreover, the enzymes involved in 4″-transamination and 8′-glycosylation have not been identified.

In the work reported herein, all enzymes responsible for 8′,1″-O-glycosidic bond formation and 4″-transamination during apramycin biosynthesis are identified and characterized in vitro using recombinant enzymes. An important observation is that AprO catalyzes glycosylation using NDP-β-d-glucose provided by AprJ and AprK as the glycosyl donor. This is a highly unusual example in which the sugar donor is neither an NDP-α-d-hexose nor an NDP-β-l-hexose but rather an NDP-β-d-hexose. Moreover, bioinformatic analysis also narrowed down the candidate enzymes involved in the introduction of the amino group at the 4″ position, and in vitro experiments with AprD5 and AprL demonstrate that transamination takes place on NDP-β-d-glucose before glycosylation.

RESULTS

Analysis of Unassigned Genes in the Apramycin Biosynthetic Gene Cluster.

There are 10 genes in the apr cluster that have not yet been assigned a biosynthetic function: AprA, D1, D2, D5, F, H, J, K, L, and O (Figure 2). AprA has the DUF371 domain at the N-terminus, which belongs to the archaeal family of unknown function. However, HHpred²⁹ analysis identified the rRNA small subunit methyltransferase (RsmI; PDB ID: 5HW4) as one of the top hits for AprA. Thus, AprA might be involved in RNA methylation. AprF contains a pentapeptide repeat with unknown function;³⁰ however, members of this family have been reported to interact with DNA-binding proteins such as DNA gyrase.³¹ Accordingly, AprA and AprF are unlikely to be directly involved in the apramycin biosynthetic pathway. Transamination of a common sugar typically involves both an NAD(P)-dependent oxidoreductase and a PLP-dependent aminotransferase. Candidates for transamination at the 4″ position thus include AprD1, D2, D5, and L. The remaining enzymes AprH, J, K, and O are likely involved in formation of the 8′,1″-O-glycosidic bond. AprH belongs to glycosyltransferase family 28 (GT28) in the Carbohydrate-Active Enzymes (CAZy) database.³² GT28 enzymes, such as MurG,³³ UgtP (YpfP),³⁴ and monogalactosyldiacylglycerol synthase,³⁵ catalyze glycosylation using a UDP-α-d-sugar as the glycosyl donor with an inversion of stereochemistry at the anomeric carbon during the reaction. AprO belongs to glycoside hydrolase family 65 (GH65) in the CAZy database. GH65 enzymes can catalyze not only the hydrolysis but also the reversible phosphorolysis of glycosidic bonds.³⁶ For example, kojibiose phosphorylase, which shows 27% identity with AprO, synthesizes kojibiose (Glc-α1,2-Glc) using β-d-glucose 1-phosphate (β-G1P, 13; see Figure 3 for the structure) as a glycosyl donor and d-glucose (14) as an acceptor.³⁷ AprJ and AprK are annotated as phosphatase and an adenylyltransferase, respectively. It has been reported that the aprJ gene-deletion mutant of S. tenebrarius produced aprosamine (3) and aprosamine 5-phosphate (11) instead of apramycin (1) and apramycin 5-phosphate (12).¹² Therefore, AprH or AprO may catalyze the glycosyltransfer reaction, and AprJ or AprK is expected to provide the glycosyl donor substrate.

Figure 3.

(A) Potential precursors to the second sugar. (B) Screening of potential sugar donors for O-glycosidic bond formation in the AprHJKO one-pot reaction. Each extracted ion chromatogram (EIC) trace (a–e) is labeled with the compound number of the donor candidate used in the assay. (C) AprHJKO one-pot reaction excluding one of the four enzymes. EIC trace (a) without AprH, (b) without AprJ, (c) without AprK, and (d) without AprO. (D) Assay to determine the reaction sequence of AprJKO. EIC trace (a) AprJK then AprO, (b) AprJO then AprK, and (c) AprKO then AprJ. (E) Scheme of the one-pot reaction with AprJ, AprK, and AprO. EICs at m/z 1041.2100 corresponding to [M – H]⁻ for the tetra-dinitrophenyl (DNP) derivatives of 3 (red trace) and at m/z 1203.2628 corresponding to [M – H]⁻ for the tetra-DNP derivatives of 2 (black trace).

Reconstitution of the Enzyme Reaction for the 8′,1″-O-Glycosidic Bond Formation.

In apramycin biosynthesis, the introduction of an amino group at the 4″-position may occur before or after the formation of the 8′,1″-O-glycosidic bond. Since saccharocin (2) has been isolated as a natural product (Figure 1), it should be possible to form an 8′,1″-O-glycosidic bond using d-glucose (14) instead of 4-amino-4-deoxy-d-glucose as the substrate. In order to test which enzyme is responsible for O-glycosidic bond formation and what is the appropriate form of the sugar donor, the purified recombinant enzymes AprH, J, K, and O were incubated with specific glucose derivatives, aprosamine 5-phosphosphate (11), and the appropriate cofactors in one pot prior to assay for saccharocin 5-phosphate (15). Each of the four His₆-tagged proteins being tested from S. tenebrarius was heterologously expressed in E. coli and purified by Ni-affinity column chromatography (Figure S1). In order to prepare aprosamine 5-phosphate (11), aprosamine (3) was obtained by acid hydrolysis of apramycin (1) as previously reported³ (Figures S23–S27). Although AprU is a kinase that catalyzes the phosphorylation of the hydroxyl group at the 5-position of 9 (Figure 2B),²⁰ it may have loose substrate specificity for other apramycin biosynthetic intermediates possessing the octose moiety. Indeed, aprosamine (3) was fully converted by AprU to the phosphorylated compound (Figure S2), and the product was isolated from a large-scale reaction with AprU (Figures S28–S33). An inspection of the ¹H NMR and correlation spectra indicated that the product was a mixture of the two C8′-hemiacetal isomers 11a and 11b. A comparison of the NMR spectra of the AprU-phosphorylated compound and aprosamine (3) showed that the quartet signal corresponding to H5 (δ 3.63–3.73 in 3) was shifted downfield by 0.5 ppm (δ 4.18 in 11) and the C5 signal (δ 75.1 in 3) was shifted downfield by 3.8 ppm (δ 78.9 in 11) (Table S3). All NMR data indicated that the AprU-phosphorylated product is aprosamine 5-phosphate (11).

The resulting aprosamine 5-phosphate (11) was then assayed with purified AprH, AprJ, AprK, and AprO in one pot with ATP, MgCl₂, and a specific glucose derivative. d-Glucose (14), G6P (4), α-d-glucose 1-phosphate (α-G1P, 16), UDP-α-d-glucose (17), and α,α-trehalose (18) were tested as candidate sugar donors (Figure 3A). After overnight incubation, the 5-phosphate group was removed by treatment with AprZ prior to derivatization with 1-fluoro-2,4-dinitrobenzene (DNFB) under basic conditions.¹⁹ Only G6P (4) was almost completely consumed, yielding a product with m/z 1203.2633 in negative mode upon liquid chromatography with mass spectrometric analysis (LC–MS), which was consistent with the pseudomolecular ion of the tetra-dinitrophenyl (DNP) derivative of 2 (calcd. [C₄₅H₄₈N₁₂O₂₈ – H]⁻ m/z 1203.2628) (Figure 3B). To clarify which enzymes are required for product formation, the enzymes AprH, J, K, and O were sequentially removed from the aforementioned one-pot reaction with ATP-Mg²⁺, G6P (4), and aprosamine 5-phosphate (11). The coupling reaction proceeded only when AprJ, K, and O were all present (Figure 3C), suggesting that AprH is not required for the expected glycosylation.

To determine the sequence of reactions catalyzed by AprJ, AprK, and AprO, different combinations of two out of the three enzymes were used in the incubation with ATP-Mg²⁺, G6P (4), and aprosamine 5-phosphate (11). After the removal of the first two enzymes with ultrafiltration after 2 h, the third enzyme was added to the filtrate and the reaction continued for another 2 h. Product was detected only when AprJ and AprK were used in the first incubation and AprO was used in the second reaction (Figure 3D). Based on these results, AprJ and AprK, which are annotated as a phosphatase and an adenylyltransferase, respectively, were proposed to catalyze the modification of the substrates before AprO catalyzes the subsequent coupling to form the glycosidic linkage. Moreover, LC–MS analysis of the reaction mixture without derivatization showed a peak with m/z 590.0898 in the AprJ-AprK reaction, which corresponds to an adenylated glucose-phosphate (calcd. [C₁₆H₂₅N₅O₁₅P₂ + H]⁺ m/z 590.0895) (Figure S3). This AprJ-AprK product was consumed in the presence of AprO (Figure S3, trace d) and not formed unless both AprJ and AprK were present (Figure S4). These findings suggested that the AprJ-AprK product is ADP-glucose, which serves as the sugar donor in the glycosylation reaction catalyzed by AprO (Figure 3E).

Characterization of the Glycosylation Product Produced in the AprJKO One-Pot Reaction.

Structural characterization of the glycosylation product was challenging due to difficulties in preparing sufficient aprosamine 5-phosphate (11) for a large-scale reaction. Therefore, aprosamine (3) was used as a starting substrate and AprU was added to the one-pot reaction to prepare 11 in situ. The large-scale AprJ-AprK-AprO-AprU reaction was conducted, and the product was isolated using a combination of both cationic and anionic exchange resins (Figures S34–S40 and Table S4). In the ¹H NMR spectrum, H1″ and H8′ were observed as doublet signals with coupling constants of 3.9 and 8.6 Hz, respectively, indicating that H1″ is in the equatorial position and H8′ is in the axial position. These signals are similar to those of the anomeric protons of saccharocin (2)⁵ and apramycin 5-phosphate (12).²⁰ The correlation between H4′ and H8′ in the NOESY spectrum also implied that H8′ is fixed in the axial position (Figure S36). In addition, H5 was observed as a quartet signal rather than a triplet due to coupling with the phosphorus atom (J_H5–P = 9.2 Hz), and the C5 signal was shifted downfield compared to the C5 of saccharocin (2). Overall, these NMR data indicated that the glycosylation product from the AprJ-AprK-AprO-AprU reaction is saccharocin 5-phosphate (15). In other words, the 8′,1″-equatorial, axial-O-glycosidic linkage is formed during the glycosylation step, probably mediated by AprO (Figure 3E).

Characterization of the Glycosyl Donor Produced by AprJ and AprK.

To characterize the chemical structure of the AprJ-AprK product, the AprJ-AprK reaction was conducted with ATP, MgCl₂, and G6P (4) in the presence of inorganic pyrophosphatase (IPP). IPP was added to degrade the pyrophosphate byproduct of adenylation in order to increase the yield of the AprJ-AprK product (Figure S5). After the degradation of excess adenine-nucleotides with shrimp alkaline phosphatase (rSAP), the AprJ-AprK product was purified by using anion-exchange resin. The excess inorganic salt was then removed by gel filtration prior to analysis by NMR (Figurse S41–S45 and Table S5). In the ¹H NMR spectrum, H1″ (δ 5.15) of the AprJ-AprK product was observed as a triplet with a coupling constant of 8.0 Hz, indicating that the coupling between H1″ and H2″ is as large as that between H1″ and the phosphorus atom. On the other hand, in the ¹H NMR spectrum of ADP-α-d-glucose, H1″ has been reported to be a doublet of doublets, with coupling constants of 7.3 Hz between H1″ and the phosphorus atom and 3.6 Hz between H1″ and H2″.³⁸ Therefore, NMR analysis of the AprJ-AprK product strongly suggested that H1″ is in the axial position, and the AprJ-AprK product is ADP-β-d-glucose (19a), which has not been previously found in nature (Figure 3E).

AprJ Catalyzes Intramolecular Phosphate Transfer Interconverting G6P and β-G1P.

AprJ is a HAD family enzyme (Table S1). HAD enzymes typically possess an active site composed of signature motifs and a bound magnesium ion (Figure S6B).^39,40,41 In reactions catalyzed by HAD phosphatases, the nucleophilic Asp residue attacks the phosphate group of the substrate, resulting in phosphorylation of the enzyme (Figure S6C).⁴¹ Another Asp residue (Asp+2) in the catalytic motif acts as a base to facilitate the hydrolysis of the aspartyl phosphate produced above, thereby completing the dephosphorylation reaction. HAD dehalogenase is also known for utilizing a similar attack by the nucleophilic Asp residue via a covalent intermediate (Figure S6C).⁴² The HAD family of enzymes also includes β-phosphoglucomutases (βPGMs). βPGM accepts a phosphate group from its activator, α-d-glucose 1,6-bisphosphate (α-G1,6P₂, 20).⁴³ This aspartylphosphate group can be transferred from the enzyme to β-G1P (13) or G6P (4) to generate β-G1,6P₂ (21) (Figure S6C).^43–45 The catalytic cycle is then completed by taking back the other phosphate group from β-G1,6P₂ (21). HAD βPGM is widely found in bacteria,^46–50 and βPGM from Lactococcus lactis has been extensively studied.^{43–45,51–56} AprJ is expected to catalyze the conversion of G6P (4) to β-G1P (13) following the same mechanism (Figure 4A), but there is a low degree of sequence identity between AprJ and the reported HAD βPGMs (20% or less, Figure S6A).

Figure 4.

(A) Proposed mechanism of intramolecular phosphate transfer mediated by AprJ. (B) LC–MS analysis of acetyl derivatives of the AprJ-catalyzed reaction. EIC trace: (a) G6P (4) was reacted as a starting substrate. (b) EDTA was added to reaction (a). (c) β-G1P (13) was reacted as a starting substrate. (d) EDTA was added to reaction (c). (e) Acetyl derivatives of G6P (4), β-G1P (13), and α-G1P (16) as standards. EICs at m/z 427.0647 corresponding to [M – H]⁻ for the tetra-acetyl derivatives of 4, 13, and 16. (C) Deconvoluted mass spectra of AprJ: (a) as-isolated, (b) incubated with α-G1,6P₂ (20), (c) incubated with α-G1P (16), (d) incubated with β-G1P (13), and (e) incubated with G6P (4). Each spectrum (b–e) is labeled with the compound number of the glucose phosphate used as an activator. The calculated mass value of His₆-tagged AprJ is 27562 Da.

The βPGM activity of AprJ was assayed by adding α-G1,6P₂ (20) as an activator in the presence of MgCl₂. Although β-G1,6P₂ (21) is expected to be a natural intermediate, we decided to use the commercially available α-G1,6P₂ (20) for the characterization of AprJ according to the reported HAD βPGMs.^46,47,50 After the reaction, the assay mixture was treated with pyridine and acetic anhydride to acetylate the glucose phosphates. LC–MS analysis of the acetyl derivatives showed that G6P (4) was converted to β-G1P (13) in the presence of AprJ despite a very low yield (Figure 4B, trace a). When β-G1P (13) was reacted with AprJ, G6P (4) was clearly formed (Figure 4B, trace c), indicating that the equilibrium of the AprJ-mediated reaction is largely biased toward G6P (4). The equilibrium constant (K_eq = [β-G1P]/[G6P]) was determined to be 0.031 ± 0.001 at pH 8.0 by the ratio of the products calculated from an integration of the resonances in the ³¹P NMR spectrum (Figure S7). K_eq is similar to those obtained for HAD βPGMs from Neisseria perflava (0.041 at pH 6.5),⁴⁸ from Euglena gracilis (0.035 at pH 7.0),⁴⁹ from Lactobacillus brevis (0.054 at pH 6.7),⁵⁷ and from E. coli (0.021 at pH 8.0).⁵⁰ AprJ activity was completely abolished by the addition of ethylenediamine tetraacetate (EDTA) (Figure 4B, traces b and d), suggesting that AprJ requires a metal ion for activity.

To test whether AprJ is phosphorylated like the reported HAD βPGMs, AprJ was reacted with glucose phosphate (20, 16, 13, or 4), followed by protein mass analysis. When α-G1,6P₂ (20) was used as an activator, a signal with an increase of 80 Da was detected, which is consistent with phosphorylation of AprJ (Figure 4C, trace b). The +80 Da signal was not detected when α-G1P (16), β-G1P (13), or G6P (4) was reacted. These results suggested that AprJ utilizes an aspartyl phosphate residue activated by α-G1,6P₂ (20) during the catalytic cycle. It has been proposed that the Asp residue in motif I (xDxDx) is phosphorylated as a nucleophilic residue, and the neighboring Asp residue (Asp+2) acts as a base during the catalytic cycle (Figure 4A, S6C).^41,43 According to this hypothesis, Asp15 was predicted to be the nucleophilic Asp in AprJ, so the phosphorylation of AprJ should also be expected with the D17N mutant. However, the +80 Da signal was not observed with either the D15N or D17N AprJ mutant (Figure S8), which likewise demonstrated no other observable activity (Figure S9). Consequently, both Asp15 and Asp17 are important residues for AprJ catalysis.

The results of the experiments described above demonstrated that AprJ catalyzes an intramolecular phosphate transfer reaction of G6P (4) using α-G1,6P₂ (20) as an activator, which generates β-G1P (13). However, in the AprJ-AprK-AprO and AprJ-AprK reactions described in the previous sections, the conversion of G6P (4) to β-G1P (13) catalyzed by AprJ could proceed without the addition of α-G1,6P₂ (20). In the study of L. lactis βPGM, it had been proposed that β-G1P (13) itself may be an activator for βPGM.⁴³ Although the phosphorylation of AprJ was not detected when G6P (4) or β-G1P (13) was used as an activator (Figure 4C), it is possible that G6P (4) and β-G1P (13) are less efficient activators than α-G1,6P₂ (20) and therefore did not accumulate enough phosphorylated AprJ for detection. Indeed, when G6P (4) or β-G1P (13) was incubated with AprJ in the presence of MgCl₂, the production of free d-glucose (14) was detected (Figure S10), suggesting that AprJ is able to dephosphorylate G6P (4) and β-G1P (13). The phosphate group of G6P (4) might be transferred to the catalytic residue to activate AprJ under conditions without an appropriate activator.

AprK Is Responsible for the Conversion of β-G1P to NDP-β-d-glucose.

The results of the one-pot reaction using AprJ and AprK and the characterization of AprJ suggested that AprK may catalyze the transfer of nucleotides to β-G1P (13). To test this hypothesis, AprK was reacted with glucose phosphate (13, 16, or 4) in the presence of ATP and MgCl₂. When β-G1P (13) was used as the substrate, ADP-β-d-glucose (19a) was formed (Figure 5A). When α-G1P (16) or G6P (4) was used as a substrate, the adenylylation reaction did not proceed, indicating that AprK selectively recognizes β-G1P (13) as a substrate to form ADP-β-d-glucose (19a). In order to investigate the substrate specificity, a nucleotide triphosphate (NTP) and β-G1P (13) were reacted with AprK in the presence of MgCl₂, and the resulting pyrophosphate was degraded to monophosphate by inorganic pyrophosphatase (IPP) and quantified by a colorimetric assay using malachite green and ammonium molybdate.⁵⁸ As a result, the best activity was observed with CTP, which produced 2.5 times more pyrophosphate than ATP under the tested conditions (Figure 5B). No activity was detected when TTP was used. These results suggested that AprK prefers smaller nucleobases to adenine, such as cytosine and uridine. CDP-β-d-glucose (19c) and UDP-β-d-glucose (19u) were isolated from the AprK assay mixture using the same method as that for ADP-β-d-glucose (19a), and their structures were determined by NMR analysis (Figures S46–S55 and Table S5).

Figure 5.

(A) LC–MS analysis of the AprK-catalyzed reaction using ATP as a nucleotide donor. EIC traces following incubation with (a) β-G1P (13), (b) α-G1P (16), (c) G6P (4), and (d) ADP-β-d-glucose (19a) as a standard. EICs at m/z 590.0895 correspond to [M + H]⁺ for ADP-glucose. (B) Screening of nucleotide donors for AprK. The generated pyrophosphate was quantified by a colorimetric assay using malachite green and ammonium molybdate after degradation by IPP. The reaction without enzyme was performed separately, and its value was subtracted as background. Each data bar is the average of three replicates, and error bars denote standard errors above and below the mean. ND, not detected. Reaction conditions: 0.5 mM β-G1P (13), 0.5 mM NTP, 1 mM MgCl₂, 0.5 μM IPP, 0.4 μM AprK, 50 mM HEPES-Na⁺ (pH 8.5), 28 °C, and 15 min.

AprO is a Unique GH65 Enzyme That Utilizes NDP-β-d-Glucose as a Glycosyl Donor Substrate.

The above results suggested that NDP-β-d-glucose (19) produced in the presence of AprK is likely the sugar donor substrate during glycosidic bond formation mediated by AprO. Such a proposed function of AprO based on the results of the AprJ-AprK-AprO reaction is unprecedented because glucose 1-phosphate is expected to be the sugar donor for a phosphorylase of the GH65 subfamily containing AprO. To investigate the substrate specificity of AprO, aprosamine 5-phosphate (11) and donor candidates (16, 13, 17, and 19u) were reacted with AprO. LC–MS analysis of the assay mixture treated with AprZ and DNFB after reaction showed no product formation if either anomer of glucose 1-phosphate was used as the sugar donor. In contrast, when using UDP-β-d-glucose (19u), the acceptor substrate was completely consumed, resulting in the formation of the desired product (Figure 6). AprO thus apparently prefers NDP-glucose over glucose 1-phosphate. Moreover, when using UDP-α-d-glucose (17), which is a common sugar donor for glycosylation in nature, no reaction is observed (Figure 6, trace c). Therefore, the anomeric position of the NDP-sugar donor in the AprO reaction must have the β-configuration. The time-course assay indicated that UDP-β-d-glucose (19u) was the best sugar donor for AprO among the three tested nucleotides (Figure S12). The AprO-catalyzed reaction demonstrated 3.5 and 5.4 times more conversion when using UDP-β-d-glucose than when using CDP-β-d-glucose (19c) and ADP-β-d-glucose (19a), respectively, under the assay conditions. This result suggested that AprO recognizes and distinguishes the nucleobase moiety of the sugar donor.

Figure 6.

LC–MS analysis of DNP derivatives following the AprO-catalyzed reaction. AprO and aprosamine 5-phosphate (11 were incubated with (a) α-G1P (16), (b) β-G1P (13), (c) UDP-α-d-glucose (17), or (d) UDP-β-d-glucose (19u). Trace (e) DNP derivatives of saccharocin (2) and aprosamine (3) as standards. EICs at m/z 1041.2100 corresponding to [M – H]⁻ for the tetra-DNP derivatives of 3 (red trace) and at m/z 1203.2628 corresponding to [M – H]⁻ for the tetra-DNP derivatives of 2 (black trace).

To investigate the recognition mechanism of NDP-β-d-glucose (19) by AprO, a model structure of AprO was constructed using AlphaFold2 (Figure S14).⁵⁹ Three crystal structures of GH65 glycoside phosphorylases have been reported: maltose phosphorylase from Levilactobacillus brevis (LbMP, PDB: 1H54),⁶⁰ kojibiose phosphorylase from Caldicellulosiruptor saccharolyticus (CsKP, PDB: 3WIR),⁶¹ and 2-O-α-glucosylglycerol phosphorylase (GGP, PDB: 4KTP).⁶² The phosphate-binding site of GH65 is composed of Tyr, Lys, and two Ser residues that are highly conserved among the GH65 phosphorylases. In AprO, Lys604, Ser639, and Ser640 are present at the corresponding positions, and the Tyr residue is replaced by His348 (Figures S13 and S14). The model structure of AprO reveals a relatively large space around the phosphate-binding site (Figure S14A), which suggests that this distinctive spatial feature of AprO may be involved in the recognition of sugar nucleotides. Furthermore, the sequence similarity network (SSN) constructed with 11,066 sequences of GH65 enzymes registered in the CAZy database revealed that the network in which AprO belongs is not part of the largest network encompassing MP, KP, GGP, and trehalose phosphorylase (TreP)⁶³ but forms its own group (Figure S15). The SSN results emphasize the uniqueness of AprO within the GH65 enzyme family.

Timing of the Introduction of the Amino Group at the 4″ Position in Apramycin.

Saccharocin (2), which has a hydroxyl group instead of a primary amine at the 4″ position, is produced by Saccharopolyspora hirsuta, which also produces apramycin.^5,6 While the complete genome of S. hirsuta has not yet been determined, the results of shotgun sequencing were registered at NCBI in 2019 (GenBank assembly: GCA_008630535.1) in which Contig 4 contains a gene cluster (i.e., sacc cluster) very similar to the apr cluster of S. tenebrarius (Figure S17). Interestingly, the apr cluster of S. tenebrarius encodes a putative aminotransferase AprL as well as three putative NAD(P)-dependent oxidoreductases (i.e., AprD1, D2, and D5) (Figure 2A), but no homologues of aprD5 or aprL genes are found in the sacc cluster. A search for aprD5 and aprL equivalents led to the identification of two homologous genes present next to each other in contig 5, suggesting that the sacc cluster and the gene set including the putative aprD5 and aprL homologues are not colocalized in the genome of S. hirsuta. This finding implies that the reason that S. hirsuta can produce both saccharocin (2) and apramycin (1) may be a consequence of the differential regulation of the sacc cluster versus the distant gene set including the aprD5 and aprL homologues. These results also suggested that AprD5 and AprL and their homologues are likely the enzymes responsible for introducing the amino group at the 4″ position in apramycin (1).

To test this hypothesis, the His-tagged AprD5 and AprL from S. tenebrarius were expressed in Streptomyces lividans TK24 and purified by Ni-affinity column chromatography (Figure S1). To examine whether transamination occurs after glycosidic bond formation, saccharocin 5-phosphate (15) was reacted with AprD5 and AprL in the presence of NAD, l-glutamate, and PLP. NAD and l-glutamate were added as cofactors for AprD5 and amino donors for AprL, respectively. LC–MS analysis of the reaction mixture after derivatization showed the corresponding peak for the DNP derivatives of the unreacted substrate but no peaks corresponding to apramycin (1) (Figure S18). Therefore, saccharocin 5-phosphate (15) is not a substrate for AprD5 and AprL.

To examine whether sugar nucleotides are the substrates for transamination, NDP-β-d-glucose (19) was incubated with AprD5 and AprL in the presence of NAD, l-glutamate, and PLP. Anion-exchange high-performance liquid chromatography (HPLC) analysis showed the conversion of UDP-β-d-glucose (19u) to a new product (Figure 7A). The product peak was isolated and analyzed by electrospray ionization mass spectrometry (ESI–MS) and showed a mass spectrum consistent with the expected amination product (Figure 7B, obsd 564.0668, calcd [C₁₅H₂₅N₃O₁₆P₂ – H]⁻ m/z 564.0637). CDP-β-d-glucose (19c) and ADP-β-d-glucose (19a) could also be converted to the expected amination products, respectively (Figure S19). Unfortunately, we could not isolate sufficient amination product for NMR characterization due to low conversion. To confirm that the desired UDP-β-d-aminosugar is formed by AprD5 and AprL, the amination product was reacted with AprO and aprosamine 5-phosphate (11). LC–MS analysis of the derivatized product showed that apramycin (1) was formed by the reaction of the AprD5-AprL product with AprO (Figure 7C), indicating that UDP-4″-amino-4″-deoxy-β-d-glucose (22u) was formed in the presence of AprD5 and AprL. These results thus verified that AprD5 and AprL catalyze the modification of NDP-β-d-glucose (19) to provide NDP-4″-amino-4″-deoxy-β-d-glucose (22), which is used as a donor substrate by AprO, leading to the assembly of the apramycin skeleton (Figure 8).

Figure 7.

(A) HPLC analysis of the AprD5-AprL reaction with UDP-β-d-glucose (19u). A solution of NAD, 19u, l-Glu, and PLP was reacted with (a) AprD5 and AprL, (b) AprD5, or (c) AprL; trace (d) 19u was used as a standard. The product peak marked with a star was isolated and analyzed by ESI–MS. (B) Mass spectrum of the isolated product in the negative scan. (C) LC–MS analysis of the DNP derivatives of the AprO-catalyzed reaction with 11 and the product provided by AprD5 and AprL. NAD, 19u, l-Glu, and PLP were reacted (a) with AprD5 and AprL, (b) with AprD5, (c) with AprL, or (d) without AprD5 and AprL, and then after the enzymes were removed by ultrafiltration, AprO and 11 were added. Trace (e) DNP derivatives of saccharocin (2) and apramycin (1) were used as standards. EICs at m/z 1203.2628 corresponding to [M – H]⁻ for the tetra-DNP derivatives of 2 (blue trace) and at m/z 1368.2803 corresponding to [M – H]⁻ for the penta-DNP derivatives of 1 (black trace).

Figure 8.

Final phase of apramycin (1) and saccharocin (2) biosynthesis.

DISCUSSION

In this study, the functions of the remaining uncharacterized enzymes in the apr cluster were elucidated, thereby establishing the complete biosynthetic pathway for apramycin (see Figures 2 and 8). The 4-amino-4-deoxy-d-glucose moiety is derived from G6P (4), which is consistent with previous results indicating the incorporation of intact glucose into the second sugar moiety of apramycin.¹⁹ Equilibration of G6P (4) with β-G1P (13) is catalyzed by AprJ and biased toward G6P (K_eq = 0.031). AprK then catalyzes a nucleotidylyl transfer reaction to produce NDP-β-d-glucose (19) from β-G1P (13). Although AprK has a relatively broad substrate tolerance for nucleotide donors, it tends to prefer cytosine and uridine bases. Subsequently, AprD5 catalyzes the oxidation of the 4″ position of 19, and AprL catalyzes the subsequent transamination to afford NDP-4″-amino-4″-deoxy-β-d-glucose (22). AprO, which is responsible for the glycosylation to aprosamine 5-phosphate (11), utilizes 22 as a sugar donor to produce apramycin 5-phosphate (12). When AprO is supplied with 19 instead of 22, saccharocin 5-phosphate (15) will be produced. The investigations for the substrate specificity of AprO showed that AprO clearly recognizes the stereochemistry at the 1″ position of the NDP-sugar donor substrates as well as the nucleobase moiety. Aprosamine 5-phosphate (11) exists as a mixture of 8′-hemiacetal isomers in aqueous solution, but AprO probably recognizes only the isomer with 8′-OH in the equatorial position as the sugar acceptor, resulting in the formation of the 8′,1″-equatorial, axial-O-glycosidic bond characteristic of apramycin (1) and saccharocin (2).

Anomeric carbons of two sugars linked by O-glycosidic bonds are found in sucrose (23) and trehalose (18) as well as the antibiotics tunicamycin (24)^64,65 and BE-7585A (25)⁶⁶ (Figure 9). The glycosidic linkages of sucrose (23) and trehalose (18) can be synthesized by appropriate enzymes such as sucrose synthase (EC 2.4.1.13), sucrose 6-phosphate synthase (EC 2.4.1.14), trehalose synthetase (EC 2.4.1.245), and trehalose 6-phosphate synthase (EC 2.4.1.15) (Figure S20A,B). Sucrose (23) and trehalose (18) can also be biosynthesized via glycosylation by sucrose phosphorylase (EC 2.4.1.7) and trehalose phosphorylase (EC 2.4.1.64), respectively. In the cases involving glycosyltransferases, UDP-α-d-glucose (17) is used as the donor substrate. Likewise, the 1′,1″-axial, axial-O-glycosidic bond in BE-7585A (25) is constructed by trehalose 6-phosphate synthase-like glycosyltransferase BexG2 also using UDP-α-d-glucose (17) as a sugar donor (Figure S20C).⁶⁷ How the 11′,1″-equatorial, axial-O-glycosidic bond in tunicamycin (24) is formed has not yet been established. When sugar nucleotides are used as sugar donors in the construction of O-glycosidic bonds linking two anomeric carbons, the stereochemistry of the anomeric carbon of the sugar donor is generally shown to be retained before and after the glycosylation reaction. In contrast, the AprO-catalyzed glycosylation proceeds with stereochemical inversion at the anomeric carbon of the sugar donor (Figure 8). It should be noted that both AprO and trehalose phosphorylase belong to the GH65 family, and all known GH65 enzymes use an inverting catalytic mechanism; however, as previously mentioned, AprO is a unique enzyme in that its sugar donor is a sugar nucleotide instead of β-G1P (13).

Figure 9.

Structures of natural products containing an O-glycosidic bond between the anomeric centers of two sugars. Anomeric carbons derived from the donor and acceptor substrates are highlighted with red and blue circles, respectively.

In the biosynthesis of natural products, monosaccharides used for glycosylation are normally activated with nucleotide monophosphate (NMP) or nucleotide diphosphate (NDP). While it is known that only 9 sugar nucleotides (eight NDP-sugars along with CMP-N-acetyl-neuraminic acid) are employed to construct eukaryotic glycoproteins and glycolipids, >100 different sugar nucleotides are used in the biosynthesis of prokaryotic polysaccharides and glycosylated natural products.²¹ The naturally available NDP-sugars include NDP-α-d-hexose and NDP-β-l-hexose, whereas NDP-β-d-hexose and NDP-α-l-hexose are considered to be non-natural analogs. G6P (4) and d-fructose 6-phosphate (F6P) serve as the precursors to most sugar nucleotides. F6P can undergo further transformation to d-mannose 6-phosphate, d-glucosamine 6-phosphate, or d-N-acetylglucosamine 6-phosphate. In all cases, a d-hexose 6-phosphate is converted to the corresponding α-d-hexose 1-phosphate by distinct but related phosphohexose mutases and then activated by the appropriate nucleotidyl transferase to become an NDP-α-d-hexose. An NDP-β-l-hexose can be generated by inverting the C5″ stereochemistry of the corresponding NDP-α-d-hexose during its biosynthesis. Therefore, the stereochemistry at the C1″ position of an NDP-hexose results entirely from the activity of phosphohexose mutase. The present experiments reveal that AprJ converts G6P (4) to β-G1P (13) rather than α-G1P (16), which is activated by AprK, specific for β-G1P (13), to produce NDP-β-d-glucose (19) in apramycin biosynthesis. This finding indicates that NDP-hexoses used in natural product biosynthesis are not limited to NDP-α-d-hexose and NDP-β-l-hexose but should now include NDP-β-d-glucose.

Among naturally occurring NDP-hexoses, NDP-β-d-(amino)glucose used in apramycin biosynthesis is the first example of this configuration. The anomeric configuration of ADP-l-glycero-β-d-manno-heptose (ADP-LDmanHep),⁶⁸ which is a building block in the biosynthesis of lipopolysaccharide, is the same. Thus, the discovery of NDP-β-d-(amino)glucose in apramycin biosynthesis further extends the story from high-carbon sugars (i.e., those with more than seven carbons in the backbone) to hexoses. ADP-LDmanHep is biosynthesized by four enzymes, including an isomerase GmhA, a bifunctional kinase/adenylyltransferase HldE, a phosphatase GmhB, and an epimerase HldD using sedoheptulose 7-phosphate as the starting substrate. It is the phosphorylation step catalyzed by HldE that determines the stereochemistry at C1 of ADP-LDmanHep (Figure S21).⁶⁹ Lipopolysaccharide biosynthesis in Gram-negative bacteria has been well studied using E. coli as a model system,⁷⁰ in which three LDmanHeps are incorporated into the core oligosaccharide chain by WaaC,^71,72 WaaF,⁷² and WaaQ.⁷³ Interestingly, all three enzymes belong to the GT9 CAZy family and catalyze the glycosylation using ADP-LDmanHep as a sugar donor substrate with an inverting mechanism.

An S_N2-like mechanism has been proposed for glycosylation reactions that proceed with stereochemical inversion of the anomeric carbon of the sugar donor (Figure S22).⁷⁴ In this mechanism, the interaction between the noncovalent electron pair of the ring oxygen and the σ* orbital of the C1″–O bond can facilitate the formation of the oxocarbenium cationic transition state (or intermediate). However, since the C1″ substituent is in the equatorial position in the most stable conformation of NDP-β-d-glucose (19), the nonbonding electron pair of the ring oxygen cannot interact with the σ* orbital of the C1″–O bond. The glycosyltransferase VinC has been shown to accept the non-natural sugar donor TDP-β-d-hexose in vitro and catalyze an inverting glycosylation.⁷⁵ In this case, it has been proposed that the conformational change in the hexose moiety moves the TDP group to the pseudoaxial position, thus satisfying the stereoelectronic requirement and promoting the formation of the glycosidic bond. A similar conformational change likely occurs prior to nucleophilic attack by the acceptor substrate in the AprO-catalyzed reaction as well (Figure S22B). Structural and computational studies on GH94 cellobiose phosphorylase, which is structurally similar to GH65 enzymes and degrades cellobiose (d-G1c-β1,4-d-G1c) to α-G1P (16) and d-glucose (14) by the same inversion mechanism as GH65, supported a conformational change of the glycon moiety of cellobiose from a chair to a boat conformation prior to nucleophilic attack by phosphate from the axial orientation.⁷⁶ AprO also has a Glu residue that is highly conserved among the GH65 and GH94 enzymes. The Glu495 may act as a base that activates 8′_eq-OH of aprosamine 5-phosphate (11) (Figure S14).

In summary, there are two significant findings of this work. First, all enzymes involved in the assembly of the 4-amino-4-deoxy-d-glucose moiety of apramycin have been identified and fully characterized, leading to a comprehensive understanding of apramycin biosynthesis. Second, AprO is a unique glycosyltransferase, as its utilization of UDP-β-d-(amino)-glucose as the sugar donor is unprecedented. In addition, AprO is the first known glycosyltransferase in the GH65 CAZy family which now contains hydrolases, phosphorylases, and glycosyltransferases. These results may open the door to the chemoenzymatic preparation of sugar nucleotides that are otherwise not found in nature.