Biosynthetic Origin of the Octose Core and Its Mechanism of Assembly during Apramycin Biosynthesis

Abstract

Apramycin is an aminoglycoside antibiotic isolated from Streptoalloteichus tenebrarius and S. hindustanus that has found clinical use in veterinary medicine. The apramycin structure is notable for its atypical eight-carbon bicyclic dialdose (octose) moiety. While the apramycin biosynthetic gene cluster (apr) has been identified and several of the encoded genes functionally characterized, how the octose core itself is assembled has remained elusive. Nevertheless, recent gene deletion studies have hinted at an N-acetyl aminosugar being a key precursor to the octose, and this hypothesis is consistent with the additional feeding experiments described in the present report. Moreover, bioinformatic analysis indicates that AprG may be structurally similar to GlcNAc-2-epimerase and hence recognize GlcNAc or a structurally similar substrate suggesting a potential role in octose formation. AprG with an extended N-terminal sequence was therefore expressed, purified, and assayed in vitro demonstrating that it does indeed catalyze a transaldolation reaction between GlcNAc or GalNAc and 6′-oxo-lividamine to afford 7′-N-acetyldemethylaprosamine with the same 6′-R and 7′-S stereochemistry as those observed in the apramycin product. Biosynthesis of the octose core in apramycin thus proceeds in the [6 + 2] manner with GlcNAc or GalNAc as the two-carbon donor, which has not been previously reported for biological octose formation, as well as novel inverting stereochemistry of the transferred fragment. Consequently, AprG appears to be a new transaldolase that lacks any apparent sequence similarity to the currently known aldolases and catalyzes a transaldolation for which there is no established biological precedent.

Graphical Abstract

INTRODUCTION

Apramycin (1) isolated from Streptoalloteichus tenebrarius (S. tenebrarius) and S. hindustanus^1–3 is an aminoglycoside^4–8 commonly used as a veterinary antibiotic.⁹ Its mode of action involves inhibition of bacterial protein synthesis by blocking the ribosomal translocation step.¹⁰ The structure of apramycin contains an atypical eight carbon bicyclic dialdose (octose) core (see 1), which has glycosidic linkages with 2-deoxystreptamine (2-DOS, 2) at C1′ and a 4-amino-4-deoxyglucose moiety at C8′ (Figure 1).¹¹ Previous studies have shown that apramycin (1) is not susceptible to modification by most aminoglycoside-modifying enzymes with the exception of two aminoglycoside N-acetyltransferases.^12–15 Thus, the unique bicyclic dialdose structure of apramycin may serve as a novel scaffold for the design and development of next-generation aminoglycoside antibiotics against resistant organisms.¹⁶

Figure 1.

Biosynthetic pathway for apramycin (1).^17,21,24,25 Conversion of paromamine (4) to lividamine (5) is catalyzed by AprD4 and AprD3.^17–19 Subsequent AprQ mediated C6′ oxidation of 5 to oxolividamine (6)¹⁷ sets the stage for octose assembly. The final steps include AprI catalyzed N-7′ methylation of 9 to give 10,²⁴ AprH mediated attachment of a 4-aminosugar, and AprZ catalyzed dephosphorylation of 11 to yield 1.²¹ The annotation of apr genes is listed in Table S1.

How apramycin (1) is assembled has attracted much attention over the years.^17–25 The identification of its biosynthetic gene cluster (apr cluster)^1–3 lead to significant progress in understanding the biological formation of apramycin (see Figure 1). Apramycin (1) contains five amino groups at C1, C3, C2′, C7′ and C4″. Two of these (i.e., at C1 and C3) are part of the 2-DOS (2) moiety and incorporated via the aminotransferase AprS (Figure 1 and Table S1).²⁶ Coupling of GlcNAc (3) to C4-OH of 2-DOS (2) and subsequent deacetylation are catalyzed by the glycosyltransferase AprM and the deacetylase AprN, respectively, resulting in the C2′ amino group (Figure 1 and Table S1).^27,28 While the origin of the C4″-amino sugar in 1 has not been studied, it may involve reductive amination of a 4-ketohexose sugar catalyzed by AprL, which is annotated as a transaminase. Attachment of the resulting 4-aminosugar to C8′–OH of the octose core may be catalyzed by AprH, which is the only unassigned glycosyltransferase encoded in the apr gene cluster. However, these reactions bracket the unknown intermediary steps that are critical to the formation of the apramycin octose core as well introduction of the remaining amino group at C7′ (i.e., 6 → 7).

Two hypotheses have been proposed for the assembly of the octose core in apramycin (1). One possibility is that the C7′ and C8′ atoms are derived from the addition of pyruvate to 6 in a reaction catalyzed by a transketolase/transaldolase followed by transamination of the C7′-keto group (Figure 2, path A).¹ However, only two genes in apr are annotated as encoding a transaminase (i.e., AprS and AprL), and both have already been assigned other functions. The alternative hypothesis is that C7′ and C8′ of 1 are derived from an amino acid such as glycine or a derivative thereof (Figure 2, path B).²⁹ However, early feeding experiments revealed that neither glycine nor serine is a precursor of carbon source to apramycin.²⁹

Figure 2.

Proposed precursors providing the C7′, C8′, and 7′-N centers of the octose moiety in 7.

Recent results, however, have suggested a third possibility. In particular, 7′-N-acetyl-demethylaprosamine (7) accumulates in the fermentation media when aprU is deleted, and AprU has been shown to catalyze C5-phosphorylation of 7 to yield 8 (Figure 1).²¹ It was also shown that 8 could be deacetylated to give 9 by AprP (Figure 1).²¹ These observations implied the precursor to the octose unit has an N-acetyl group making N-acetylglucosamine (GlcNAc, 12) or its derivative a possible candidate (Figure 2, path C). We then analyze the apr cluster searching for genes encoding GlcNAc-related enzymes with a focus on aprG as it has not been assigned a function. However, the BlastP analysis provided little insight regarding the potential function of AprG because its closest homologues are hypothetical proteins except for a glycoside hydrolase family 127 protein (Saccharopolyspora hirsuta). Nevertheless, HHpred analysis revealed that AprG may be structurally similar to GlcNAc-2-epimerase (Nostoc sp. KVJ10, accession no. 6F04_A). Hence, AprG may recognize GlcNAc (12) or an N-acetyl aminosugar as its substrate and act as the coupling enzyme to afford the octose core. Reported herein is a description of the origins of the C7′, C8′, and 7′-N centers in 1 along with a functional and mechanistic characterization of AprG catalysis.

RESULTS AND DISCUSSION

Glucosamine is a Building Block of the Octose Core in Apramycin.

To explore whether GlcNAc (12) is a precursor to octose formation during apramycin biosynthesis, feeding experiments were performed with Streptomyces tenebrarius NRRL B-12390 using [1-¹³C]-glucose, [U–¹³C₆]-glucose and [1-¹³C]-glucosamine. Each ¹³C-labeled isotopologue was separately added to 50 mL of cultures at 0, 24, 48, and 72 h to a final concentration of ca. 2 mM for each addition (see Section S3 in the SI). After incubation for a total of 144 h, each culture was collected, and the production of apramycin (1) was analyzed by liquid chromatography with mass spectrometric detection (LC-MS). Compared to the control with natural abundance glucose, different levels of ¹³C enrichment in 1 were observed in all three ¹³C-labeled trials (compare with Figure S3B–E). In particular, approximately 67% more ¹³C incorporation was detected when [1-¹³C]-glucosamine instead of [1-¹³C]-glucose was used in the incubation. Thus, glucosamine appears to be a more immediate precursor to octose biosynthesis in vivo (Figure S3C,E). The ¹³C-enriched apramycin produced in the above feeding experiments was then derivatized using benzyl chloroformate (CbzCl) and isolated using high-performance liquid chromatography (see Section S3.4 in the SI) for further analysis by ¹³C nuclear magnetic resonance (NMR) spectroscopy (see Section S10.1 in the SI). Significant ¹³C double enrichment was detected at both C1′ and C8′ of 1 upon supplementation with [1-¹³C]-glucosamine (Figure S26, trace F). Taken together, these experiments demonstrate that glucosamine or GlcNAc may serve as an immediate biosynthetic precursor to the octose moiety in 1 with the C8′ and C7′ centers originating in C1 and C2 of glucosamine, respectively. While these results themselves do not necessarily rule out either Path A or B in Figure 2, they do show that Path C is a viable third alternative.

AprG is Responsible for Octose Core Formation in Apramycin.

To identify the enzyme responsible for octose assembly, all genes in the apr cluster were analyzed using BlastP³⁰ and HHpred³¹ (Table S1). While no putative transaldolase genes could be located in the apr cluster, as aforementioned, HHpred analysis of the AprG sequence led to N-acylglucosamine 2-epimerase (Nostoc sp. KVJ10, accession no. 6F04_A) as one of the top hits. Because N-acylglucosamine 2-epimerase catalyzes the interconversion of GlcNAc (12) and N-acetyl-mannosamine (ManNAc, 17),³² AprG may also recognize an N-acetyl amino sugar as its substrate. It was thus hypothesized that AprG may be the enzyme responsible for the delivery of a C₂ fragment donated from GlcNAc (12) or a derivative thereof to the acceptor 6′-oxolividamine (6), which is the AprQ product,¹⁷ to form the octose core.

To test this hypothesis, AprG was overexpressed and purified from Escherichia coli as a soluble N-His₆-tagged construct after recognizing that the correct gene sequence begins 129 residues from the 5′-end of the originally published¹ aprG gene from S. tenebrarius (Figure S1). Lividamine (5) was prepared as previously published;¹⁸ however, efforts to prepare recombinant AprQ (see Figure 1) for in situ production of 6 from 5 were unsuccessful. Therefore, the homologous enzymes NeoQ and BtrQ, which catalyze the same reaction in the biosynthesis of neomycin and butirosin, respectively,^26,33–35 were instead isolated in soluble form and confirmed to catalyze the oxidization of 5 to 6 (Figures S5 and S6). BtrQ was found to provide a greater extent of conversion on overnight incubation with lividamine (5) compared to NeoQ and was therefore chosen for the preparation of 6 (Figure S5B). However, overoxidation to yield 18 was also observed as previously reported for AprQ (Figure S6).^17,23

GlcNAc (12) along with several other commercially available GlcNAc derivatives including GlcNAc-1-P (13), GlcNAc-6-P (14), UDP-GlcNAc (15), GalNAc (16), and ManNAc (17) were each incubated with AprG, BtrQ, and lividamine (5) for 14 h. The reaction mixture was then treated with 1-fluoro-2,4-dinitrobenzene (DNFB) under alkaline conditions, and the resulting dinitrophenyl (DNP) derivatives were extracted with ethyl acetate followed by LC-MS analysis (see Supporting Information, Section S5.1). Product formation was observed only in the incubations with GlcNAc (12) and GalNAc (16), where three peaks with the same m/z as that expected for a DNP derivative of 7 were identified (calcd. C₃₄H₃₆N₁₀O₂₀⁻ [M – H]⁻ m/z 903.2035, obsd. 903.2015) (Figure 3, trace a–f). The peaks detected by LC-MS analysis were thus consistent with the formation of compound 7 or isomers thereof from either compound 12 or compound 16 during the incubation period. Due to the scarcity of the samples and the tedious purification procedures, the identities of these peaks were not further characterized. When the assays with GlcNAc (12) and GalNAc (16) were treated with 2,4-dinitrophenylhydrazine (DNPH) after 14 h of incubation, production of 22 and 23 was observed consistent with the respective formation of the four-carbon sugars d-erythrose (20) and d-threose (21) as well (Figures 4 and S8). Moreover, time courses based on UV-absorbance at 350 nm of the DNPH-derivatives of d-threose and d-erythrose produced from GalNAc (16) and GlcNAc (12), respectively, demonstrated more rapid accumulation of the former implying that GalNAc is the more efficient substrate under the assayed conditions (see Figure S12). These results support a model in which AprG catalyzes a transaldolation between 12/16 and 6 to yield 7 and 20/21 thereby generating the octose core.

Figure 3.

Screening of potential C₂ donor substrates for the AprG-catalyzed transaldolation reaction. (A) Structures of possible C₂ donors. (B) Extracted ion chromatograms (EICs) at m/z 903.2035 corresponding to [M – H]⁻ for the tridinitrophenyl (DNP)-derivatives of 7, which was obtained after workup of the BtrQ/AprG reactions and then treated with 1-fluoro-2,4-dinitrobenzene (DNFB). Each EIC trace (a–f) is labeled with the compound number of the C₂ donor candidate used in the assay.

Figure 4.

BtrQ/AprG/AprU reaction with lividamine (5) and GalNAc (16). (A) Scheme of the one-pot reaction with BtrQ, AprG, and AprU. (B) Electrospray ionization (ESI) mass spectra of the one-pot reaction with (a) BtrQ, (b) BtrQ + AprG, (c) BtrQ + AprG + AprU, (d) BtrQ + AprU, and (e) lividamine (5) standard. All spectra were acquired in positive ion mode. Compound 5: [M+1]⁺ m/z 308.1816; compound 7: [M+1]⁺ m/z 407.2136; compound 8: [M+1]⁺ m/z 487.1800; and compound 18: [M+1]⁺ m/z 322.1609.

A significant amount of the carboxylic acid 18 was also generated with GalNAc (16) as the C₂-donor in the one-pot reaction (Figure 4B, trace b, Figure S7) indicating that 6′-oxolividamine (6) was not fully consumed by AprG but was instead further oxidized by BtrQ to 18. Therefore, AprU, which catalyzes 5-phosphorylation of 7,²¹ and ATP were included in the BtrQ/AprG reaction in order to drive the AprG-catalyzed reaction toward the phosphorylated octose product 8. LC-MS analysis showed that the signal corresponding to 18 (calcd C₁₂H₂₃N₃O₇ [M + H]⁺ 322.1609) was no longer discernible being replaced by a signal at m/z 487.1789 consistent with accumulation of 8 (calcd C₁₆H₃₁N₄O₁₁P [M + H]⁺ 487.1800, Figure 4B, trace c). Furthermore, control experiments with AprU and either 7 or 18 indicated that AprU only catalyzes the phosphorylation of the former²¹ and not the latter (see Figure 4B, trace c,d). Hence, the detection of m/z 487.1789 strongly indicated that the formation of the octose core is catalyzed by AprG and AprU and helps to drive the AprG reaction (6 →7) to completion.

Characterization of the AprG Product.

While LC-MS analysis of the BtrQ/AprG/AprU reaction product described above was consistent with the formation of 7 and thus AprG-catalyzed octose formation, the structure of this product remained fully established. Therefore, AprZ (fused with maltose-binding protein at N-terminal (MBP-AprZ)), which is an alkaline phosphatase that catalyzes C5-dephosphorylation of apramycin-5-phosphate (11) and its biosynthetic intermediates,²¹ was used to convert the isolated BtrQ/AprG/AprU reaction product (8) back to 7. As shown in Figure S9, dephosphorylation of the product catalyzed by AprZ was indeed observed, indicating that the inclusion of AprU in the initial reaction followed by workup with AprZ helped to improve production of 7. Subsequent purification was then achieved through a combination of both cationic and anionic exchange resins (Section S5.5) prior to analysis by NMR.

Inspection of the ¹H NMR and correlation spectroscopy spectra indicated that the product was purified as a mixture of the two C8′-hemiacetal isomers 7a and 7b (Figure 5A). Thus, the corresponding H-8′_axial (7a) and H-8′_equatorial (7b) signals are, respectively, at 4.99 (J_{8′axial-7′}= 8.5 Hz) and 5.33 ppm (J_{8′equatorial-7′} = 2.0 Hz) (Figures S27 and S28). Furthermore, the stereochemistry at C6′ was confirmed to be 6′-R due to the absence of a nuclear Overhauser enhancement spectroscopy (NOESY) correlation between H-6′ (4.14–4.11 ppm) and H-8′_axial (4.99 ppm) (Figures S29 and S30). In contrast, the stereochemistry at C7′ was determined to be 7′-S based on the small J_7′-6′ coupling constant (3.0 Hz), which is also present for H-7′ (J_7′-6′ (3.0 Hz) in apramycin (1).¹¹ More importantly, a correlation between H-7′ and H-5′ was observed in the NOESY scans (Figures S27 and S31). These findings collectively established the 6′-R and 7′-S stereochemistry of the isolated product and its assignment as 7 (see Figure 5A), which is consistent with the 6′-R and 7′-S configurations of apramycin (1).¹¹ This assignment is also in agreement with the ΔaprU fermentation results previously published in 2021;²¹ however, in that report the stereochemistry at C6′ and C7′ was not assigned nor was the occurrence of two epimers at C8′. The R-stereochemistry at C6′ has been shown to be critical for the binding of 1 to the bacterial ribosome.³⁶

Figure 5.

(A) Possible mechanisms of octose formation catalyzed by AprG. (B) Proposed stereochemical course of the C–C bond formation is catalyzed by AprG. (C) ESI mass spectra of the octose product (8) formed in the BtrQ/AprG/AprU reaction using (a) unlabeled 12 or (b) [2-²H]-12.

Mechanistic Study of the AprG-Catalyzed Reaction.

The aforementioned stereochemical analysis indicates that AprG catalyzes a transaldol reaction in which the si face of the transferred two-carbon N-acetyl-acetaldehyde equivalent (24) nucleophilically attacks the re face of acceptor aldehyde 6 during octose formation (see Figure 5B). However, the four-carbon byproduct (20/21) is instead eliminated from the re face of the two-carbon equivalent (24) in 12/16. Consequently, a stereochemical inversion takes place such that the 2-R center in the donor substrate (12/16) becomes the 7′-S center in the aldol product (7). This differs from other known transaldolases where elimination and addition are typically at the same face of the transferred fragment leading to retention of stereochemistry.^37–39 In terms of the sequence of C–C bond-forming events, AprG-catalyzed transaldolation may proceed via two possible mechanisms. Mechanism A (Figure 5A) involves cleavage of the C2–C3 bond in 12/16, followed by the C–C bond formation between the resulting C₂-donor (24) and C6′ of 6′-oxolividamine (6) to yield 26, which after intramolecular cyclization gives 7. Alternatively, C6′–C7′ bond formation to yield 28 may occur prior to the elimination of the four-carbon aldose (20/21) in a retro-aldol process to give 29, which will then cyclize to afford 7 via 26 (mechanism B).

Mechanisms A and B in Figure 5A are distinguished by the fate of the C2–H bond in the C₂ donor. In mechanism A, this bond remains intact throughout the catalytic cycle, whereas in mechanism B it undergoes cleavage, allowing for potential incorporation of a solvent Hydron at C7′ in 7. Therefore, [2-²H]-12 was synthesized (Supporting Information, Section S9.1) and used as the substrate in the coupled BtrQ/AprG/AprU assay as an initial test between these two mechanisms. LC-MS analysis of the resulting product 8 demonstrated an increase of one mass unit (Figure 5C) indicating the retention of the deuterium. The complementary experiment was also performed by conducting the BtrQ/AprG/AprU assay in buffered D₂O. In this case, the BtrQ product (6, calcd [M + H]⁺ m/z 322.1609) acquired a single deuterium indicated by minor and major peaks at m/z 322.1604 and 323.1676, respectively, consistent with the partial exchange of the acidic C5′ Hydron in the acceptor substrate 6 with solvent (Figure S18). However, incorporation of a second deuterium in the AprG/AprU product 8 (unlabeled calcd. [M + H]⁺ m/z 487.1800) was not observed with minor and major peaks, respectively, at m/z 487.1790 and 488.1862 alone (Figure S19). These results imply that the C2–H bond remains intact throughout the transaldolation and thus supports the elimination of the four-carbon sugar taking place before the C–C bond formation (i.e., mechanism A).

If the retro-aldol bond cleavage between C2–C3 in 12/16 is the first step as proposed in mechanism A, then AprG may also catalyze degradation of 12/16 even in the absence of an acceptor. Indeed, upon incubation of AprG with 16 alone, d-threose (21) and acetamidoacetaldehyde (25) were both detected by LC-MS and coelution with standards (Figure S11 and Section S9.2 in the SI). Furthermore, the incubation of lividamine (5) with BtrQ/AprG/AprU in the presence of 25, which is expected to be in equilibrium with 24 in free solution, also led to the production of phosphorylated octose 8 based on LC-MS analysis (Figure S16B(a)). These results support the intermediacy of 24 in mechanism A for the AprG-catalyzed reaction (Figure 5A). As described above, however, the C2-Hydron in 24 does not appear to exchange with the solvent during the catalytic cycle of AprG, which suggests that 24 once generated in the active site with 6 undergoes more rapid C–C bond formation versus dissociation to bulk solvent.

Previous studies have shown that the catalytic cycles of aldolases typically proceed via one of two mechanisms based on how the transferred fragment corresponding to the enolate 24 is activated.^38,40 Class I aldolases possess an active site lysine residue that forms a Schiff base with the aldol-donor (e.g., 30) to facilitate catalysis (Figure 6A, route A). In contrast, class II aldolases instead utilize a divalent metal ion, such as zinc, to stabilize a discrete enolate intermediate (e.g., 32, Figure 6A, route B). However, when the AprG assay was conducted in the presence of ethylenediaminetetraacetic acid, there was no significant reduction in the level of product formation (Figure S13). Instead, AprG contains two lysine residues located in the putative active site of a structural model built using Alphafold2 (Lys42 and Lys257, Figure S20). Moreover, the sequence alignment of AprG with several hypothetical proteins having sequence similarity to AprG (ranging from 88 to 36% sequence identity) showed that these two lysine residues are conserved among them (Figure S2). The importance of these two residues was investigated by preparing K42A and K257A mutants (Section S6). While the K42A mutant was able to catalyze the formation of 21 and 25 from 16, albeit with poor yield, no such reaction was observed in the presence of the K257A mutant (Figures S15 and S16A(b,c)). When the putative two-carbon fragment 25 was synthesized and provided as the substrate instead of 16, only the K42A mutant showed activity to afford 8 (Figure S16B(b,c)). Therefore, whereas the K42A mutation appears to primarily affect catalysis of the initial retro-aldol cleavage of 16 to yield transferred two-carbon fragment 25, it has much less of an impact on the subsequent aldol addition of 25 with 6 to yield 7, which can then be phosphorylated to 8. In contrast, the K257A mutant no longer appears to catalyze either retro-aldol cleavage of 16 to 21 and 25 or the subsequent aldol coupling of 25 with 6.

Figure 6.

Proposed mechanism of retro-aldol bond cleavage catalyzed by AprG to form the C₂-donor. In the absence of an acceptor compound (e.g., 6), 25 along with 21 were detected as reaction products.

These observations otherwise suggest that Lys257 is critical for the complete catalytic cycle, whereas Lys42 is required for only the initial retro-aldol elimination to generate the transferred two-carbon fragment. One possible mechanism similar to that of the class I aldolases involves the formation of a Schiff base (30) between Lys257 and the donor substrate with Lys42 potentially serving as a general base (see Figure 6A, route A). Efforts to trap the putative imine intermediate (30) by reduction with NaBH₄ while incubating the wild-type AprG or K257A mutant with either 16 or 25 were unsuccessful; however, a mass increase of 85 corresponding to adduct formation with 25 was observed with the K42A mutant (see Figure S17). These observations are consistent with a mechanism in which the donor substrate is activated through the formation of a Schiff base with a catalytic lysine residue such as K257 (Figure 6A, route A), despite the apparent lack of homology between AprG and other class I aldolases or transaldolases characterized to date.

CONCLUSIONS

This report explains the biosynthetic origin of the octose moiety found in apramycin (1). In particular, AprG is shown to operate as a transaldolase that catalyzes the transfer of a two-carbon N-acetyl-acetaldehyde equivalent (24) from either GlcNAc (12) or GalNAc (16) to the six-carbon acceptor 6′-oxolividamine (6). This results in the formation of a new C6′–C7′ bond which upon cyclization via intramolecular hemiacetal formation generates the octose core in 7. Moreover, AprQ catalyzed oxidation of lividamine (5) yields the aldehyde acceptor substrate (6) for AprG; however, the flavin adenine dinucleotide-dependent dehydrogenase BtrQ was instead used in the coupled assay because AprQ could not be expressed in soluble form in our hand. As previously reported for AprQ,^17,23 BtrQ can also catalyze further oxidation of 6 resulting in the dead-end carboxylic acid 18 and thereby poison apramycin biosynthesis under aerobic conditions. This suggests a potential biosynthetic role for the phosphorylase AprU, which can shift the 6 + 12/16 ⇌ 7 + 20/21 equilibrium to remove 6 and effectively trap the initial octose intermediate as the 5′-phosphate 8 thereby avoiding dead-end oxidation.

While AprG functions as a transaldolase, its sequence shows no significant resemblance to any known aldolases. Nevertheless, initial mechanistic characterization suggests that it may utilize the active site lysine residue K257 to form a Schiff base with the two-carbon donor (GlcNAc or GalNAc) and thereby facilitate the transfer of the C₂ fragment. Moreover, the reaction appears to proceed via retro-aldol elimination of the C₂ fragment from the donor preceding the subsequent addition of aldol to the C₆ acceptor rather than vice versa. This is consistent with the classic mechanism of the class I aldolases despite the lack of any apparent sequence similarity to other known enzymes in this class.

There are two essential paradigms known regarding the biosynthesis of octose-containing secondary metabolites. The first involves [5 + 3]-coupling between a five-carbon sugar (e.g., adenosine-5-phosphate or uridine-5-phosphate) and a three-carbon donor, which is exemplified by the biosyntheses of 2-keto-3-deoxy-d-manno-octulosonic acid (Kdo, 34)^41,42 as well as various octosyl acid (35) containing natural products such as ezomycin, malayamycin, nikkomycin, and polyoxin (Figure 7A(a)).^43–45 While phosphoenolpyruvate (PEP, 33) is the most common three-carbon donor, d-fructose 6-phosphate (36) or d-sedoheptulose 7-phosphate (37) can also serve in this role as observed during the biosynthesis of lincosamide (38, Figure 7A(b)),³⁹ which is a key precursor to lincomycin. The second known principal mechanism of octose formation involves a [6 + 2]-coupling between a six-carbon sugar (e.g., an NDP-hexose) and a two-carbon donor, which has been either demonstrated or proposed to be operant during the biosynthesis of yersiniose (40),⁴⁶ hibarimicins,⁴⁷ angelmicin,⁴⁸ pillaromycin,⁴⁹ namenamicin,⁵⁰ and aldgamycins.⁵¹ In this case, the two-carbon building block is believed to originate in pyruvate (39) or hydroxypyruvate (41) with a thiamine pyrophosphate (TPP) dependent decarboxylation reaction facilitating the transfer as verified in the production of yersiniose catalyzed by TPP-dependent YerE (Figure 7B(a)).⁴⁶ Therefore, while the biosynthesis of apramycin belongs to the [6 + 2] category, the immediate two-carbon donor is not pyruvate as is usually the case but rather an aminosugar such as GlcNAc or GalNAc (Figure 7B(b)). Consequently, the identification and characterization of AprG not only represent the discovery of a new transaldolase that catalyzes a unique inversion of stereochemistry in the transferred moiety and lacks any apparent sequence similarity with the currently known aldolases but also add a new dimension to our understanding of biological octose construction.

Figure 7.

(A) Octose formation involving [5 + 3] coupling: (a) three-carbon donor is PEP (33) in the reaction catalyzed by KDO synthase; (b) three-carbon donor is derived from F6P (36) or S7P (37) via the retro-aldol reaction catalyzed by LmbR. (B) Octose formation involving [6 + 2] coupling: (a) two-carbon donor is pyruvate (39) or hydroxypyruvate (41); (b) GlcNAc (12) or GalNAc (16) serves as the two-carbon donor in the AprG-catalyzed reaction.