Reinvestigation of the C-Glycoside Synthase in Alnumycin Biosynthesis Reveals a Conserved Mechanism of C–C Bond Formation

Abstract

C-Nucleosides are defined by their unusual C–C glycosidic linkage between the nucleobase and the monosaccharide moiety, which distinguishes them from common N-nucleosides. Several enzymes have been identified to catalyze this atypical C–C bond formation. For instance, YeiN catalyzes the reversible cleavage of pseudouridine 5′-phosphate, yielding ribose 5-phosphate (R5P) and uracil via a Schiff base intermediate formed between R5P and an active-site lysine residue. In alnumycin biosynthesis, the C–C glycosidic bond between R5P and a naphthoquinone heterocycle, prealnumycin, has been shown to be installed by AlnA. While AlnA shares 41% sequence identity with YeiN, a distinct mechanism involving an ene-diol tautomer of R5P had been proposed based on previous biochemical studies and X-ray crystallography. Herein, the mechanism of AlnA is reevaluated using juglone (5-hydroxy-1,4-naphthalenedione) as a prealnumycin analog. By employing isotopologues and protein mass spectrometry, the involvement of an ene-diol intermediate and an alternative Morita–Baylis–Hillman mechanism in AlnA catalysis can both be ruled out. Further analysis of juglone reactivity showed that it can be reduced either enzymatically when coupled to glucose oxidase or nonenzymatically through autoreduction yielding 1,4,5-naphthalenetriol. This hydroquinone derivative of juglone serves as the true substrate of AlnA such that the C-glycosylation mechanism is no different from that of YeiN. These findings unravel the correct substrate of the C-glycoside synthase AlnA and unify the mechanisms of AlnA, YeiN, and other C-glycoside synthases. These results highlight that accurate substrate identification is essential for mechanistic study of enzyme catalysis and call for a reevaluation of the biosynthetic pathway of alnumycin and other naphthoquinone-derived natural products.

Graphical Abstract

INTRODUCTION

The C-nucleosides represent a small class of natural products that structurally resemble the more common N-nucleo-sides;^1–7 however, unlike N-nucleosides, where the heterocycle is linked to C1 of D-ribofuranose via a carbon–nitrogen glycosidic bond, C-nucleosides instead feature a carbon–carbon glycosidic bond. This key structural feature imparts to C-nucleosides distinct metabolic stability against acid hydrolysis or enzymatic phosphorolysis making them promising candidates for drug development.¹ A notable example is remdesivir, a C-nucleoside prodrug that has gained widespread attention as a treatment for COVID-19.⁸ Remdesivir and many other natural or non-natural C-nucleosides demonstrate antiviral activities through incorporation into RNA in place of a canonical N-nucleotide.^9–13

To date, several enzymes have been identified to catalyze formation of the unusual C–C glycosidic bond in several C-nucleosides.^14–19 Pseudouridine 5′-phosphate C-glycosidase (YeiN) is one such example. Despite being initially discovered in the degradation pathway of pseudouridine (1) (Figure 1A),^14,20 it was later realized that these enzymes not only participate in the catabolism of primary metabolites but also function as C-glycoside synthases in the biosynthesis of C-nucleoside secondary metabolites. Biochemical characterization of SdmA¹⁷ and OzmB¹⁹ from the biosynthetic pathways of showdomycin (7) and oxazinomycin (10), respectively (Figure 1A), has shown that these C-glycoside synthases likely operate through a consensus mechanism similar to that of YeiN,^14,21 involving the formation of a Schiff base (12) between a highly conserved lysine residue and ribose 5-phosphate (R5P, 4) (Figures 1B and S2). Nucleophilic attack by the heterocycle (3, 5 or 8) at the resulting iminium group (12) establishes the C–C glycosidic linkage in 2, 6 or 9. The catalytic lysine is then eliminated with the assistance of a conserved glutamate residue (Figure 1B,C), and the acyclic ribose moiety undergoes recyclization to yield the C-nucleotide product.^14,21

Figure 1.

(A) Reactions catalyzed by YeiN-type C-glycoside synthases. The nucleophilic carbon atoms and the electrophilic carbon atoms involved in C-glycosidic bond formation are highlighted in red and blue, respectively. (B) The general mechanism proposed for YeiN-type enzymes. (C) Sequence alignment of E. coli YeiN (EcYeiN), SdmA, AlnA and OzmB highlighting the conserved lysine and glutamate residues.

Anhydroexfoliamycin (β-17, also known as alnumycin C1) is a C-nucleoside natural product produced by Streptomyces exfoliatus, and exhibits antioxidant and neuroprotective activities.^22–26 It is also an intermediate produced during the assembly of the polyketide antibiotic alnumycin A (19) in Streptomyces sp. CM020 (Figure 2).²³ The structure of alnumycin A (19) features a naphthoquinone core attached to an unusual 1,3-dioxane moiety which together with its antibacterial and cytostatic activities has led to much interest in its biosynthesis.^27,28 Alnumycin C (C1 or C2) is characterized by a ribofuranose unit attached to the prealnumycin core (15) through a C1′–C8 glycosidic linkage. Construction of this linkage involves a C-glycosylation reaction between R5P (4) and prealnumycin (15) to yield alnumycin P (β-16 and α-16) catalyzed by AlnA, which is a homologue of EcYeiN (41% sequence identity).²⁹ Alnumycin P is the 5′-phosphorylated precursor to alnumycin C, and its dephosphorylation is catalyzed by the HAD family phosphatase AlnB (16 → 17) before subsequent rearrangement of the ribose moiety to a tetrahydrofuran-dioxolane bicycle (17 → 18) catalyzed by the cofactor-independent oxidase Aln6.²³ Finally, the NADPH-dependent oxidoreductase Aln4 is responsible for catalyzing reduction of the hemiacetal and the rearrangement of the resulting diol to give alnumycin A (18 → 19).²³

Figure 2.

Proposed biosynthetic pathway of alnumycin A by Oja et al.²⁹ Percentiles represent the distribution of isomers, either isolated from an enzymatic reaction (e.g., alnumycin P) or produced during fermentation (e.g., alnumycin C and B).

While much information is available regarding the late stages of alnumycin A biosynthesis, several observations suggest an incomplete picture of the overall pathway and enzymes involved. Gene deletion experiments following heterologous expression of the alnumycin biosynthetic gene cluster in Streptomyces albus revealed the intermediacy of 17 and 18 along the biosynthetic pathway.²³ However, both 17 and 18 are isolated as mixtures of C1′ epimers from the deletion mutants Δaln6 and Δaln4, respectively.²³ Furthermore, in vitro assay of AlnA with 4 and 15 also produces a mixture of C1′ epimers of 16.²⁹ In fact, a total of six stereoisomers of alnumycin A (19) were isolated from the native strain S. sp. CM020.³⁰ While these observations contradict general expectations regarding enzyme stereoselectivity and thus raise important questions about the possible involvement of nonenzymatic transformations, they align with the known propensity of C-glycoside to undergo nonenzymatic C1′ epimerization.^17,19 Moreover, the proposed mechanism of C-glycosylation catalyzed by AlnA diverges considerably from the mechanism employed by its homologues YeiN, SdmA and OzmB despite the similarity of their sequences (a minimal of 41% sequence identity to AlnA) and structures.⁶ Specifically, R5P (4) is hypothesized to serve as an electrophile in the reaction catalyzed by YeiN, SdmA and OzmB (see Figure 1B),^14,17,19 but is proposed to act as a nucleophile in the case of AlnA (see Figure 3).²⁹

Figure 3.

Previously proposed mechanism of AlnA.²³

In this study, the late-stage biosynthesis of alnumycin A is reinvestigated with a focus on elucidating the reaction mechanism of AlnA and the stereochemical course of its transformation. The findings presented indicate that a significant revision of the biosynthetic pathway for alnumycin A is necessary. Utilizing the electrophilic quinone juglone as an analog of the proposed substrate prealnumycin, AlnA is shown to instead recognize the reduced and nucleophilic hydroquinone derivative 1,4,5-naphthalenetriol as the actual substrate. This implies a reversal in the polarity of the mechanism from that previously proposed such that R5P serves as the electrophile rather than prealnumycin. AlnA catalysis thus appears to follow the established mechanistic paradigm of other YeiN-type enzymes. Furthermore, the distinct chemical reactivities of hydroquinones versus quinones are highlighted as they can play a pivotal role in modulating the observed outcomes of both enzyme catalyzed and nonenzyme catalyzed reactions alike. These insights not only refine our understanding of alnumycin biosynthesis but also underscore the importance of characterizing the redox state of the substrates as it can be very sensitive to in vitro versus in vivo conditions and thus confound the mechanistic analysis.

RESULTS AND DISCUSSION

The Distinction Between the Reactions Catalyzed by AlnA and YeiN.

The original mechanistic proposal for AlnA is built upon the X-ray crystal structure of the binary complex of AlnA with ribulose 5-phosphate (Ru5P, 20) and the observation that AlnA catalyzes equilibration between R5P (4) and Ru5P (20).^23,29 Consequently, a mechanism is proposed to involve tautomerization of R5P or Ru5P to the ene-diol intermediate 21 (Figure 3), which then undergoes 1,4-Michael addition to the electrophilic quinone moiety of prealnumycin (15) upon deprotonation by an active site glutamate residue (Glu29). Residues Lys159 and Lys86 were speculated to facilitate the formation of the C-glycosidic linkage (21 → 22 → 23) by stabilizing the development of negative charges in 22 and 23 through electrostatic interactions.²⁹ The intermediate 23 then proceeds through sequential tautomerization and hemiketal cyclization steps leading to the fructofuranose 25. Subsequent dehydration and tautomerization results in the production of alnumycin P (16). The fact that the isolated C-nucleotide product from the AlnA reaction is a mixture of β-16 and α-16 isomers suggests that the final cyclization (24 → 25 → 26 → 16) may occur nonenzymatically.²⁹

The fundamental distinction between AlnA and YeiN (as well as SdmA and OzmB) lies in the polarity of the addition reaction. In YeiN, R5P forms an electrophilic Schiff base (12) with an active site lysine residue (Figure 1B). In the case of the AlnA reaction, however, R5P instead functions as a nucleophile that adds to the electrophilic Michael acceptor prealnumycin (15) (Figure 3). This distinction also implies two different roles played by the highly conserved and catalytically essential lysine residue in the two enzyme active sites. Specifically, Lys159 in AlnA is believed to provide noncovalent electrostatic stabilization during the addition reaction (Figure 3), whereas Lys166 in YeiN instead forms a covalent adduct with the ribose substrate in the catalytic cycle (see Figure 1B). Another distinguishing aspect is the requirement of glucose, glucose oxidase and catalase to lower the oxygen level in order to reconstitute AlnA activity in vitro;^23,29,31 however, the proposed mechanism of AlnA does not seem to be oxygen-sensitive. Furthermore, the homologous enzyme, YeiN, shows no inhibition by oxygen during catalysis.^14,20 The disparity between the observed and expected behavior of AlnA, especially given its similarity to YeiN, prompted a reinvestigation of the AlnA catalyzed reaction.

Reconstitution of AlnA and AlnB Activity.

Commercially available juglone (27), which structurally resembles prealnumycin (15), has been shown to be a substrate for AlnA.³¹ Therefore, it was used to verify the activities of AlnA and AlnB. The preliminary assays involved the incubation of 0.5 mM juglone (27), 5 mM R5P (4), 2 mM MgCl₂, 50 mM NaCl, 30 μM AlnA and 10 μM AlnB in 100 μL 50 mM Tris-HCl buffer (pH 7.0) under aerobic conditions. Assays were also performed with the additional inclusion of 50 mM glucose, 1 μM glucose oxidase and 1 μM catalase as previously described.^23,29,31 After incubation at room temperature for 3 h, the reaction was quenched by the addition of 200 μL MeOH containing 0.25% (v/v) formic acid. LC-HRMS analysis of the reaction supernatant revealed the production of ribosyljuglone (29) (calcd m/z for C₁₅H₁₃O₇⁻ [M–H]⁻ 305.0667; found: 305.0675). As previously reported, inclusion of the glucose oxidase system appeared to be critical for the AlnA/AlnB reaction, as 29 was not observed in its absence (Figure 4, trace d) unless the reaction was conducted under anaerobic conditions (Figure 4, trace e). When AlnB was not included in the aerobic incubation of AlnA in the presence of the glucose oxidase system, the formation of ribosyljuglone 5′-phosphate (28) was observed instead of 29 (Figure 4, trace b). Sufficient 28 could be acquired under this condition to permit structure characterization by NMR (Figure 5A). ¹H NMR analysis revealed it to be an inseparable mixture of α- and β-ribofuranosyl isomers (α-28 and β-28, respectively) (Figure 5A, spectrum a). These results demonstrated the success of in vitro reconstitution of AlnA and AlnB activity using juglone and reproduced the apparent oxygen sensitivity of the reaction as previously reported for the native substrate prealumycin (15).

Figure 4.

Characterization of AlnA and AlnB with and without the glucose oxidase system (GO/Glc/CAT). (A) Reactions catalyzed by AlnA and AlnB using juglone (27) as substrate; (B) HPLC analysis of the AlnA and AlnB coupled assay. GO: glucose oxidase, Glc: D-glucose, CAT: catalase.

Figure 5.

(A) ¹H NMR analysis of 28 isolated from different incubation conditions. Aerobic incubation of AlnA with juglone (27), glucose oxidase, glucose and catalase along with either (a) natural abundance R5P (4); (b) [1-D]-R5P; or (c) [2-D]-R5P. (d) Anaerobic incubation of AlnA with juglone and R5P without glucose oxidase, glucose and catalase. (e) Anaerobic incubation of AlnA and 1,4,5-napthalenetriol (35) and R5P. (f) β-28 standard. (B) Structures of mechanistic probes for the AlnA catalyzed reaction.

AlnA Catalysis Does Not Involve an Ene-Diol Intermediate.

The proposed mechanism in Figure 3 suggests that the ene-diol 21 is generated from either R5P (4) or Ru5P (20). Hence, deuterium labeled isotopologues of 4 were prepared and incubated with AlnA to test for solvent exchange (Figure 5B). However, when [1-²H]-R5P (98% ²H-enriched, 1D-4) was employed under the aforementioned assay condition, less than 1% deuterium wash-out of the corresponding product 28 was observed by mass spectrometry (Figure S3). Isolation and structural characterization of 28 indicated that the deuterium was retained at C1′ (Figure 5A, spectrum b). Similarly, when [2-²H]-R5P (93% ²H-enriched, 2D-4) was employed, approximately 4% deuterium wash-out was noted and the deuterium was retained at C2′ (Figures 5A, spectrum c and S3). To corroborate these observations, the reaction was also conducted in buffered D₂O (>90% ²H-enriched) with natural abundance substrates. LC-HRMS analysis of the AlnA product (28) revealed approximately 3% deuterium incorporation into the product from solvent (Figure S3). All these observations indicated that AlnA does not catalyze solvent exchange at either C1′ or C2′ of R5P. Therefore, if enolization of R5P is a component of the AlnA catalytic cycle, the catalytic base(s) responsible for deprotonation at C1′ and C2′ must not be solvent exchangeable during turnover.

Four candidate residues, including Glu29, Lys86, Asp142 and Lys159, that may serve as catalytic acids or bases can be located in the active site pocket of the AlnA crystal structure (Figure S4). Alanine substitution of each of these residues resulted in a greater than 94% reduction in enzymatic activity, highlighting their critical roles in catalysis (Figure S6). However, Lys86 and Lys159 are unlikely to effect deprotonation and protonation without solvent exchange due to rotameric scrambling of hydrogens on the ε-amino group. Despite significantly diminished activity, the E29A mutant can still produce 28 without loss of deuterium from the deuterated substrates (Figures S5 and S6). While the activity of the D142A variant was too low to confirm whether deuterium retention occurs, at least two catalytic bases are necessary for nonsolvent-exchangeable deprotonation and reprotonation at both C1′ and C2′. Thus, these results suggest that deprotonation at either C1′ or C2′ is unlikely a step in the AlnA catalyzed reaction as previously proposed.

Ruling Out an Umpolung Mechanism.

The absence of an ene-diol intermediate (21) implies that C1 of R5P unlikely functions as a nucleophile in the AlnA catalyzed reaction. However, if R5P instead acts as an electrophile, then formation of the C–C glycosidic bond between R5P and juglone (27) or prealnumycin (15) demands transformation of the latter quinone substrates from being electrophilic Michael acceptors to nucleophilic donors. One possible scenario is an umpolung reaction involving a Morita–Baylis–Hillman-type mechanism facilitated by the addition of a nucleophilic active site residue (e.g., Lys159) to reverse the electronic properties of the quinone substrate.^32–36 As shown in Figure 6, the nucleophilic residue may first attack C2 of juglone, and the resulting protein-tethered enolate (30) then undergoes aldol coupling to the linearized R5P (11). Subsequent dehydration of intermediate 31 can lead to the α,β-unsaturated ketone 32, which may be subject to 5-exo-trig attack by the C4′–OH, resulting in ring-closure to give 33. Elimination of the nucleophilic residue would then yield the quinone adduct (28). Furthermore, the juglone-protein adduct 30 may be susceptible to auto-oxidation in air leading to a covalent dead-end shunt product such as 34. Lowering the concentration of oxygen in the activity assay may therefore be important to prevent the formation of 34 and enzyme inactivation.

Figure 6.

Proposed Morita–Baylis–Hillman-type mechanism of AlnA.

To explore this possibility, 50 μM wild-type AlnA was incubated with 0.5 mM juglone (27) in 100 μL 50 mM Tris-HCl buffer (pH 8.0) for 3 h under aerobic conditions. Protein mass analysis revealed that 90% of the AlnA was modified by a single juglone molecule demonstrated by a mass shift of +172 Da with approximately 10% of AlnA modified by two juglone molecules (Figure S7). However, this modification did not occur at the conserved Lys159 or Lys86 residues, as covalent modification was still observed when K159A and K86A variants were employed in the incubation (Figure S7). However, 18% of the AlnA-C251S mutant was found to be modified with a single juglone (Figure S7), which suggested that Cys251 rather than Lys159 or Lys86 is the primary site of covalent modification. However, formation of the Cys251-juglone adduct is unlikely to be catalytically relevant, because Cys251 is 19.7 Å away from C1 of Ru5P in the active site judging from the crystal structure of AlnA-Ru5P binary complex. Furthermore, the C251S AlnA variant produced a similar amount of C-glycoside product 28 compared to wild-type AlnA in an end-point assay (Figure S6). Importantly, ¹H NMR analysis of the reaction product from the anaerobic AlnA reaction with R5P and juglone revealed that the product is still a mixture of α-28 and β-28 epimers (Figure 5A, spectrum d). If the C-glycosylation proceeds as shown in Figure 6, however, only one stereoisomer is expected, because the ring closure step (32 → 33) will necessarily occur within the enzyme active site. Therefore, a Morita-Baylis-Hillman type mechanism mediated by a protein residue is unlikely to be the mode of catalysis for AlnA.

Formation of 1,4,5-Naphthalenetriol.

Efforts were subsequently focused on understanding why AlnA catalysis is inhibited in the presence of molecular oxygen and why this inhibition is prevented in the presence of glucose oxidase and catalase. It was noticed that solutions of juglone (27) prepared in Tris-HCl buffer underwent a gradual color change from bright yellow to wine red upon standing under anaerobic conditions. Time-course UV–vis analysis of an anaerobic solution of 0.5 mM juglone in 50 mM Tris-HCl buffer (pH 8.0) at 25 °C revealed the formation of a species with a broad absorption band between 300 and 350 nm (Figure 7B). The product spectrum resembled that reported for 1,4,5-naphthalenetriol (35), which is the reduced form of juglone.³⁷ A chemical standard of 1,4,5-naphthalenetriol was subsequently synthesized (see Supporting Information) and its UV–vis spectrum matched that observed following prolonged anaerobic incubation of juglone. This observation suggested that juglone undergoes disproportionation via the intermediacy of a semiquinone radical under anaerobic conditions similar to what has been seen for benzoquinones (see Figure S8 for proposed mechanism).³⁸ Specifically, this apparent autoreduction may be initiated with deprotonation of juglone and the resulting anion 36 can reduce a second molecule of juglone to its semiquinone form 38 (Figure 7A). This semiquinone radical can be further disproportionated to afford juglone (27) and 1,4,5-napthalenetriol (35). This observation suggested that 35 may be the true substrate of AlnA leading to apparent inhibition in the presence of molecular oxygen, which can both scavenge the semiquinone radical 38 and promote auto-oxidation of 35.

Figure 7.

Formation of 1,4,5-napthalenetriol (35) from juglone (27) under two different assay conditions. (A and B) Autoreduction of juglone. (C and D) Glucose oxidase-mediated reduction of juglone. (E) HPLC analysis of the AlnA reaction in the presence of various reducing agents.

A similar color change from yellow to red was also noted for AlnA/AlnB activity assays conducted aerobically in the presence of the glucose oxidase system. Glucose oxidase utilizes flavin adenosine dinucleotide (FAD) as a cofactor to accept electrons from glucose. FAD then undergoes reduction to FADH₂,³⁹ which can further reduce various electron acceptors including oxygen, benzoquinones, naphthoquinones and anthraquinones.⁴⁰ Therefore, in addition to reducing O₂ concentrations,⁴¹ the glucose oxidase system may also participate in the AlnA catalytic cycle by catalyzing the net reduction of juglone (27) by glucose providing the putative substrate 35 (Figure 7C). Indeed, when 20 μM glucose oxidase was incubated with 50 mM glucose and 0.5 mM juglone in 50 mM Tris-HCl buffer (pH 8.0) under anaerobic conditions, the characteristic absorption spectrum of 1,4,5-naphthalenetriol (35) became evident after just 10 min (Figure 7D). This rate of formation is significantly faster compared to that via autoreduction, where the formation of 35 was less apparent before 30 min. LC-HRMS analysis identified the expected mass of 35 (calcd m/z for C₁₀H₇O₃⁻ [M–H]⁻ 175.0401; found: 175.0409) consistent with glucose oxidase catalyzed reduction of juglone by glucose (Figure S9).

1,4,5-Naphthalenetriol is the True Substrate of AlnA.

Unlike juglone, 1,4,5-naphthalenetriol is a potent nucleophile, which would allow the C-glycosylation between R5P (4) and 1,4,5-naphthalenetriol (35) to proceed through the same mechanism as YeiN. Besides the glucose oxidase system, other FADH₂-generating enzymes such as Escherichia coli flavodoxin reductase (EcFdr)⁴² and putidaredoxin reductase (CamA)⁴³ were both found to facilitate the AlnA catalyzed C-glycosylation reaction under aerobic conditions (Figure 7E). Moreover, chemical reductants including Na₂S₂O₄, NADH, or NADH with phenazine methosulfate (PMS), which are capable of reducing quinones to hydroquinones, were also found to support the AlnA catalyzed reaction even though NADH is an inefficient oxygen scavenger. These findings indicate that the apparent inhibition of AlnA by O₂ is not due to a direct effect on the catalytic mechanism but rather depletion of the true substrate 35 of AlnA catalysis.

These observations prompted a further comparison of the chemistry shared by AlnA and other homologous C-glycoside synthases such as YeiN and OzmB. While no ΨMP (2) formation was detected when uracil (3) and R5P (4) were incubated with AlnA (Figure S10), AlnA was able to catalyze C-glycosidic bond formation between the OzmB substrate 8 and R5P (4) under anaerobic conditions. The formation of the nascent C-glycoside product 9 was indirectly confirmed through OzmC-mediated dephosphorylation and auto-oxidation, yielding detectable indochrome (Figure S11).¹⁹ Importantly, a Schiff base adduct (e.g., 12) between the active site lysine residue of AlnA and R5P (4) was detected by protein mass spectrometry after treatment with NaBH₄ (Figure S12). Tryptic proteolysis revealed that the covalent modification occurred at the highly conserved Lys159 residue in AlnA (Figure S12). These observations suggest that AlnA may activate R5P (4) as an electrophile during catalysis of the C-glycosylation reaction through formation of the iminium intermediate 12 between the key catalytic residue Lys159 and R5P. As in the case of YeiN,^14,21 formation of the iminium and subsequent elimination of Lys159 may be mediated by the conserved Glu29, thereby explaining its importance to AlnA catalysis.

The synthetic hydroquinone 35 is stable as a solid if kept under anaerobic conditions but readily oxidizes to juglone in aqueous buffer when exposed to air. Therefore, assays of AlnA with 35 were conducted under anaerobic conditions to prevent auto-oxidation of 35. Accordingly, the freshly prepared 35 (0.5 mM) was assayed with 5 mM R5P (4), 2 mM MgCl₂ and 30 μM AlnA in 100 μL 50 mM Tris-HCl buffer (pH 7.0) for 3 h followed by removal of proteins via ultrafiltration under anaerobic conditions. The correct exact mass for both 41 (calcd m/z for C₁₅H₁₈O₁₀P⁺ [M + H]⁺ 389.0632; found: 389.0617) and its oxidized counterparts 28 (calcd m/z for C₁₅H₁₆O₁₀P⁺ [M + H]⁺ 387.0476; found: 387.0485) were detected by LC-HRMS (Figure 8B, trace b). Moreover, exposure of the reaction filtrate to air for 1 h led to the complete conversion of 41 to 28 (Figure 8B, trace d), indicating that the immediate hydroquinone product 41 is susceptible to oxygen and underwent rapid auto-oxidation toward 28 during LC-HRMS analysis. On the other hand, when 10 μM AlnB was included in the activity assay, the dephosphorylated products 43 and 29 were detected (Figure 8B, traces c and e). Among these, the oxidized products 28 and 29 were readily isolated for comprehensive NMR characterization, which revealed that 28 is isolated as a mixture of two isomers β-28 (74%) and α-28 (26%) (Figure 5A, spectrum e) and 29 is isolated as a mixture of β-29 (79%) and α-29 (21%) (see Supporting Information). These results established that hydroquinone 35 is a substrate of AlnA.

Figure 8.

(A) Characterization of AlnA and AlnB using 1,4,5-napthalenetriol (35) as substrate. HPLC analysis of (B) AlnA and AlnB coupled reaction; (C) time-course activity assay of AlnA; (D) isomerization of β-28.

While β-28 and α-28 are poorly separated by reverse-phase chromatography, these two diastereomers can be resolved using anion exchange HPLC. As shown in Figure 8C, a time-course analysis of the above AlnA activity assay revealed the appearance of just β-28 after a 2 min incubation, with the α-epimer only becoming detectable after 10 min. Furthermore, α-28 continued to gradually accumulate from 20 to 60 min while the formation of β-28 reached a plateau during that period (Figure 8C), and isolated β-28 did not undergo observable C1′ epimerization in 100 mM KPi buffer (pH 7.0) even in the presence of 30 μM AlnA (Figure 8D). However, β-28 did undergo epimerization in the presence of NADH/PMS (Figure 8D), which implies a nonenzymatic process initiated by reduction of the quinone to the hydroquinone (β-28 → β-41), followed by epimerization (β-41 → 42 → α-41) and reoxidation (α-41 → α-28). These observations suggest that AlnA catalyzes stereoselective C-glycosylation with 1,4,5-naphthalenetriol (35) as the substrate. Therefore, the mechanism of AlnA catalysis follows the characteristic behavior of C-nucleoside synthases.^14,19

Apart from the major products, several other minor enzyme-dependent species with the characteristic juglone chromophore were also detected in the AlnA/AlnB coupled reaction (Figure 8B, traces c and e). One species which exhibits the same exact mass as 29 was identified as its β-pyranosyl isomer (β-pyran-29). Two other species eluting at 7.8 and 8.1 min (P7.8 and P8.1, respectively) had exact mass consistent with juglone modified with two ribose moieties. To determine the sites of ribosylation, the AlnA/AlnB enzyme assay was scaled up such that P7.8 and P8.1 could be isolated for further analysis (Figure S13). The latter also appeared to be a mixture of multiple diribosylated juglone isomers, and only the major species was in sufficient quantity for structure determination. ¹H NMR analysis revealed C-glycosylation at both C2 and C7 of juglone in both P7.8 and the major species in P8.1, and both products are consistent with the structure of 46 (see Tables S6 and S7). These observations suggest that AlnA has a flexible active site and can accept large substrates like 43 for a second round of C-ribosylation.

AlnB is a Stereoselective Phosphatase.

The identification of 43 from the coupled AlnA/AlnB assay using hydroquinone 35 as the substrate implies that AlnB catalyzes dephosphorylation of the hydroquinone 41 (Figure 8). To investigate whether AlnB can also catalyze hydrolysis of the quinone 28, 0.2 mM purified α-28 or β-28 was assayed in the presence of 2 mM MgCl₂ and 10 μM AlnB in 50 mM HEPES buffer (pH 7.0) under aerobic conditions. LC-MS analysis of the reaction mixtures revealed hydrolysis of only the β isomer and no apparent reaction with the α isomer (Figure 9A). To compare the relative rates of dephosphorylation between hydroquinone and quinone substrates, a solution of ca. 0.2 mM 41 was enzymatically prepared from the AlnA assay under anaerobic conditions. The reaction mixture was deproteinized and directly subjected to the AlnB reaction without further purification using the same assay conditions as described above but under anaerobic conditions. Within an hour, more than 98% of 41, detected in the form of 28, was consumed (Figure 9B) while only 10% of the β-28 was consumed during the same period. Taken together, these findings suggest that AlnB is a phosphatase that is stereoselective for the β-ribofuranosyl isomer. However, AlnB is more tolerant of structural variation in the aglycone moiety and is able to accept both the hydroquinone and quinone as the substrate. The greater extent of turnover of the hydroquinone 41 versus the quinone 28 under equivalent conditions suggests that 41 rather than 28 is likely the native substrate of this enzyme.

Figure 9.

HPLC analysis of the AlnB activity assay using (A) quinone substrates α-28 and β-28, and (B) hydroquinone substrate 41. Calf intestinal alkaline phosphatase (CIP, shown in panel A) was used as a positive control to dephosphorylate α-28.

Ribose 5-Phosphate Isomerase Activity.

Previous research on AlnA has shown that it can also accept Ru5P (20) as an alternative substrate to produce the same ribofuranosyl product 17 (see Figure 3).²³ Indeed, incubation of AlnA with 1,4,5-naphthalenetriol (35) and Ru5P (20) under the same assay conditions as with R5P (4) demonstrated formation of 28 as the major product (β-28:α-28 = 3:1) upon LC-HRMS and anion-exchange HPLC analysis following exposure to air (Figure 10A). Unlike the reaction of 35 with R5P, however, two additional minor product peaks were also observed in the reaction with Ru5P. When the reaction was repeated with D-arabinose 5-phosphate (Ara5P, 47), which is the C2 epimer of R5P, there was a significant reduction in overall product formation with three product peaks formed in comparable amounts (Figure 10A). The three products identified in the assay with 35 and Ara5P (47) had the same retention time as 28 and the two minor products observed with 35 and Ru5P (20). Subsequent isolation and structural characterization led to their assignment as the arabinosyl derivatives α-48 and β-48 (i.e., the C2′ epimers of α-28 and β-28) (see Table S8 and Figure S14). This result suggested tautomerization of Ru5P (20) to both R5P (4) and Ara5P (47) with an estimated 97% stereoselective preference for the former based on the relative HPLC peak integrations and thus enzyme involvement.

Figure 10.

HPLC analysis of the (A) AlnA and (C) YeiN catalyzed reactions using different pentose phosphate substrates. Inset: extracted ion chromatogram (EIC) traces corresponding to the [M–H]⁻ signal of 28 or 48. ³¹P NMR analysis of the ribose 5-phosphate isomerase activity of (B) AlnA and (D) YeiN. (E) Proposed reaction pathway of pentose phosphate isomerization and the formation of arabinosyl isomers.

To corroborate the detection of arabinosyl isomers α-48 and β-48, 100 μM AlnA was incubated with 4 mM R5P (4) together with 2 mM MgCl₂ in the absence of 35 and the reaction progress was monitored by ³¹P NMR. Following a 30 min incubation, formation of Ru5P (20) was detected which is consistent with the previously reported R5P isomerase activity of AlnA (Figure 10B).²⁹ Moreover, a third resonance consistent with the ³¹P chemical shift of Ara5P became discernible after 24 h based on comparison with a standard. Similar changes were not observed in the control NMR time courses of R5P and MgCl₂ without AlnA. These observations are consistent with AlnA catalyzed interconversion of R5P (4) to Ara5P (47) via Ru5P (20) with only R5P and Ara5P being substrates for C-glycosylation (Figure 10E).

Whether R5P isomerase activity is shared by other YeiN-type C-glycoside synthases was also examined by incubating 4 mM R5P (4) with 100 μM YeiN, SdmA or OzmB and monitoring the reaction by ³¹P NMR. In all cases, the ³¹P signal corresponding to Ru5P (20) was observed within 4 h of initiating the incubation but remained absent in the control without enzyme (Figures 10D and S15). Furthermore, an additional ³¹P signal corresponding to the formation of Ara5P (47) could be detected within 24 h in the SdmA and YeiN assays as opposed to the control. Furthermore, incubation of 10 μM YeiN with 0.5 mM Ru5P (20) and 5 mM uracil (3) resulted in the detection of two products by HPLC with the major product having the same retention time as ΨMP (2) (Figure 10C). The minor species had the same retention time as β-arabinosyl ΨMP (Ara-ΨMP, 49), which could also be prepared by incubating YeiN, Ara5P (47) and uracil (3). These observations indicated that YeiN is capable of catalyzing the isomerization of Ru5P (20) to R5P and Ara5P prior to C-glycosylation.

Similar results were obtained with SdmA and OzmB as well. Thus, incubation of 0.8 mM Ru5P and 80 μM native SdmA substrate 5 with 10 μM SdmA for 1 h yielded two distinct C-glycoside products (Figure S16). The major product was identified as the C-ribotide product 6, as its retention time matched that of the product formed using R5P. The minor product, which eluted at the same retention time as the product formed using Ara5P, was assigned to the arabinosyl isomer (i.e., C2′ epimer of 6) even though NMR characterization of this species is unavailable due to the instability of the enzymatic product. Likewise, when 10 μM OzmB was incubated with 5 mM Ru5P and 1 mM native OzmD substrate 8 for 16 h, the C-glycoside product 9 was detected after workup with phosphatase OzmC and auto-oxidation to yield stable indochrome (Figure S17). Therefore, R5P isomerase activity is a common feature of all known YeiN-family C-glycoside synthases; yet, this activity does not appear to be immediately relevant to the formation of C–C glycosidic bonds at least from a mechanistic perspective since Ru5P was not the direct substrate in the reaction. However, it is possible that both processes share the same active site residues to facilitate catalysis.

CONCLUSIONS

AlnA is a YeiN-type C-glycoside synthase in the biosynthesis of alnumycin A that can catalyze C-glycosylation of R5P (4) with 1,4,5-naphthalenetriol (35), which is the reduced form of juglone (27). In contrast, juglone itself is not a substrate for AlnA; however, it can undergo autoreduction to yield 1,4,5-napthalenetrial under anaerobic conditions or be reduced by coupling with the glucose oxidase system. These findings can account for the previously reported requirement of the glucose oxidase system in order to observe AlnA activity under aerobic conditions.^23,31 This discovery is significant since it also uncovers that AlnA catalysis proceeds via the same mechanism as other YeiN-type C-glycoside synthases involving the formation of a key Schiff base (12) between a highly conserved lysine residue and ribose 5-phosphate (4) as shown in Figure 1B.¹⁴ Furthermore, all currently known YeiN-type enzymes exhibit a moonlighting ribose 5-phosphate isomerase activity; however, this function appears to be unrelated to their roles in the respective C-glycosylation reactions. While AlnA catalyzes stereoselective C-glycosylation, the resulting product is susceptible to nonenzymatic rearrangement of the ribose ring leading to the formation of C1′ epimeric pairs. However, subsequent dephosphorylation is carried out by the stereoselective phosphatase AlnB, which exclusively recognizes the β-isomer thereby facilitating efficient processing of correctly configured biosynthetic intermediates.

The observed activity of AlnA with 1,4,5-naphthalenetriol (35) instead of juglone (27) also implies a revision of the alnumycin A biosynthetic pathway (Figure 11). In the revised pathway, the hydroquinone derivative (50) of prealnumycin rather than prealnumycin (15) itself is proposed as the native substrate for AlnA in vivo. Isolation of prealnumycin (15) from the ΔalnA deletion mutant²⁷ is thus likely the consequence of nonenzymatic auto-oxidation of the hydroquinone intermediate 50 in the biosynthetic pathway. Consequently, intermediate 50 undergoes stereoselective C-glycosylation and dephosphorylation catalyzed by AlnA and AlnB to generate 52, which is also prone to auto-oxidation resulting in the isolation of 17. While the previous characterization of Aln6 indicated that quinone 17 can be accepted by Aln6 and no reducing agent is required for this transformation,²³ it remains to be thoroughly examined whether hydroquinone 52 or quinone 17 serves as the true substrate of Aln6. This question, along with the substrate specificity of Aln4, which is the next enzyme in the pathway, will be explored in future studies.

Figure 11.

Revised biosynthetic pathway of alnumycin A.

It remains unclear whether the hydroquinone or quinone form of alnumycin A is the biologically active species, as current studies on its biological activity have focused exclusively on the quinone form. The quinone and hydroquinone forms of alnumycin A likely exhibit distinct mechanisms of action due to their opposing electronic properties. The observed alkylating activity of juglone implies that the quinone form of alnumycin A may exert its biological effect through covalent modification of enzyme thiol residues, similar to the mechanism of action of showdomycin (7).⁴⁴ Although these aspects require further clarification in future studies, the biosynthetic investigation of alnumycin A presented here demonstrates that hydroquinones are the genuine biosynthetic intermediates. Quinone- and hydroquinone-containing natural products, typically biosynthesized via polyketide synthases, are widely distributed in nature and play various biological roles.^45,46 The findings of this study highlight the importance of determining the redox states of these compounds, not only as a prerequisite for investigating enzyme mechanisms but also as a critical factor in elucidating their biological functions given the significantly different chemical properties of quinones and hydroquinones.

Supplementary Material

ASSOCIATED CONTENT

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c03469.

Description of experimental procedures, full characterization of all chemical compounds, supplementary tables and figures, and copies of additional spectroscopic data (PDF)