Pyrazofurin Biosynthesis Involves Nonenzymatic Ring Contraction of a Pyridazine Intermediate Triggered by a Rieske Enzyme-Catalyzed Oxygenation

Abstract

Pyrazofurin and formycin A are two pyrazole-containing C-nucleosides, exhibiting antibacterial, antiviral, and antitumor activities. Despite previous studies elucidating the early biosynthetic steps of pyrazole formation, how the linear hydrazone intermediate is cyclized to a pyrazole ring remains unknown. Herein, substrate analogs are used to show that the amidohydrolase PyfA mediates intramolecular cyclization of 2-hydrazinylidenepentanedioic acid to yield a 4,5-dihydropyridazine ring, which can undergo PyfO-catalyzed dehydrogenation to generate a pyridazine product. More importantly, the Rieske enzyme PyfB is demonstrated to catalyze oxygenation of 4,6-dihydroxypyridazine-3-carboxylic acid. The resulting six-membered pyridazine product is found to undergo uncatalyzed rearrangement to the five-membered pyrazole ring. This work thus highlights a unique pyrazole-forming pathway involving a rare pyridazine intermediate that undergoes oxidation-triggered nonenzymatic ring contraction to yield the pyrazole core.

Graphical Abstract

INTRODUCTION

Pyrazofurin (1) and formycin A (2) are two C-nucleosides isolated from species of Streptomyces.^1-3 Pyrazofurin exhibits both antiviral and antitumor activities.⁴ Once it is 5′-monophosphorylated in vivo, it becomes an inhibitor of orotidine monophosphate decarboxylase, thereby impeding de novo pyrimidine nucleoside biosynthesis.⁵ In addition, formycin A also shows potent antibacterial and antiviral properties by targeting adenosine-utilizing enzymes.⁶ The structures of both 1 and 2 include a pyrazole moiety in their nucleobases (Figure 1A). Pyrazoles are aromatic heterocycles containing two adjacent nitrogen atoms in a five-membered ring. Although many steps in the biosynthesis of 1 and 2 have been elucidated, including the PyfQ/ForT-catalyzed C-glycosidation between phosphoribosyl pyrophosphate (PRPP, 3) and pyrazoles 4 and 5 to give 6 and 7, respectively, along with subsequent modification of the nucleobases to give 1 and 2 (Figure 1A),^7-10 how the pyrazole rings 4 and 5 are constructed remains unsolved. Considering pyrazole is the structural core of several naturally occurring heterocycles, such as l-α-amino-β-(pyrazolyl-N)-propanoic acid, fluviols, and withasomnine (see Figure S2),¹¹ learning how the pyrazole ring is assembled in 1 and 2 will not only resolve the mystery of its formation but may also shed light on the biosynthesis of other pyrazole-containing natural products.

Figure 1.

(A) PyfQ- and ForT-catalyzed C-glycosidation during the biosynthesis of pyrazofurin (1) and formycin A (2), respectively. (B) Established early steps in the biosynthesis of pyrazoles 4 or 5. AA indicates glycine, l-alanine, l-serine, l-threonine, or l-proline.

Previous studies have demonstrated that the initial steps in the biosynthesis of 4 and 5 appear to be the same.^12,13 As shown in Figure 1B, the initial conversion of l-lysine to N⁶-hydroxylated l-lysine (8) is catalyzed by the flavin adenine dinucleotide (FAD)-dependent hydroxylase PyfH/ForK. PyfG/ForJ, which contains cupin and methionyl-tRNA synthetase-like domains, then catalyzes N–N bond formation between 8 and d-glutamic acid to form hydrazine 9. The nicotinamide adenine dinucleotide (NAD)-dependent oxidor-eductase PyfI/ForL catalyzes further oxidation of 9, followed by hydrolysis to release α-hydrazino D-glutamic acid (10) and 11. Next, the ATP-dependent ligase PyfJ/ForM catalyzes aminoacylation of the terminal nitrogen in 10 with a proteinogenic amino acid, such as glycine, l-alanine, l-serine, l-threonine, or l-proline. This is followed by dehydrogenation of the C2–N bond to yield hydrazone 13. However, the remaining steps that result in cyclization of linear hydrazone intermediate 13 to pyrazoles 4 and 5 have yet to be elucidated. This is of particular interest as the pyrazole moiety is not commonly found in nature, and very few pyrazole-forming enzymes have been identified.^14,15

RESULTS AND DISCUSSION

Compared with the pyrazofurin gene cluster (pyf), the formycin A gene cluster (for) harbors two unique genes forF and forI, which are annotated as encoding a dehydrogenase and an aminotransferase, respectively (Figure 2).¹⁶ These two genes are likely associated with the installation of the C4 amino group in pyrazole 5. Therefore, efforts were initially focused on the biosynthesis of pyrazole 4 in the pyrazofurin pathway, which may also contribute to the assembly of 5. The pyf gene cluster contains five gene products that have not been characterized and are potentially involved in pyrazole formation (Figure 2). Among them, putative amidohydrolase PyfA is hypothesized to be linked to cleavage of the N-aminoacyl group in 13 or another intermediate in the pathway. The remaining four genes are annotated as encoding two flavin mononucleotide (FMN)-dependent monooxygenases (PyfM and PyfN), an FAD-dependent oxidoreductase (PyfO), and a Rieske enzyme (PyfB). A combination of these four oxidases was thus hypothesized to catalyze the six-electron oxidation of 13 to form 4.

Figure 2.

Biosynthetic gene clusters of pyrazofurin (pyf) and formycin A ( for). Colored genes are responsible for the formation of pyrazoles 4 and 5. Genes highlighted in gray in the table have been characterized, while those in blue have not.

Sequence analysis of PyfB shows that it shares 34% sequence identity with the phthalate dioxygenase (PDO) from C o m a m o n a s t e s t o s t e r o n i K F 1 ( N C B I I D : WP_003054408.1).¹⁷ The function of PDO is to catalyze the dihydroxylation of phthalate (14) to form 15 (Figure 3A).¹⁷ This suggests that PyfB may also accept an aromatic compound as its substrate. One possible substrate for PyfB is 6-hydroxypyridazine-3-carboxylic acid (16), which could undergo dihydroxylation to generate 17, similar to the reaction catalyzed by PDO (Figure 3B). Subsequently, 17 could be dehydrogenated to pyridazine 18 that may undergo ring contraction to yield 4, analogous to a similar reaction reported for the conversion of 19 to 20 (Figure 3C).¹⁸

Figure 3.

(A) Reaction catalyzed by PDO, (B) proposed reaction of PyfB with 16 and subsequent formation of 4, and (C) ring contraction of a pyridazinone (19) to yield a pyrazole (20).

To investigate the function of Rieske enzyme PyfB, N-His₆-tagged PyfB was expressed and purified from Escherichia coli. Rieske enzymes are known to bind a mononuclear nonheme iron in addition to a [Fe₂S₂] cluster.^19,20 Nonetheless, the iron content analysis showed that the as-isolated PyfB contained only 0.85 ± 0.02 Fe per protein monomer, suggesting that the [Fe₂S₂] cluster and the nonheme mononuclear iron center in as-isolated PyfB were not fully occupied. Therefore, reconstitution of PyfB with 6-fold FeCl₃ and Na₂S was performed under anaerobic conditions, which increased the iron content of reconstituted PyfB to 2.78 ± 0.11 Fe per monomer. Furthermore, the UV–vis absorption spectrum of reconstituted PyfB exhibited three broad bands around 330, 460, and 560 nm (Figure S3), which are characteristic of a Rieske [Fe₂S₂] cluster.^21,22 To test the activity of PyfB, 0.4 mM 16 (commercially purchased) was incubated with 20 μM PyfB, 25 μM reductase VanB from Pseudomonas aeruginosa,²³ and an NADH regeneration system containing 0.2 mM NAD, 5 mM sodium formate, and 1.4 mU/μL formate dehydrogenase from Candida boidinii in a 50 mM 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid (HEPES) buffer (pH 7.5) aerobically at room temperature overnight. However, no reaction was observed, suggesting that 16 is not a substrate of PyfB.

As another substrate candidate, commercially available 4,6-dihydroxypyridazine-3-carboxylic acid (21) was then tested with PyfB under the above conditions. This led to near complete consumption of 21 after a 1 h incubation with concomitant formation of a new species at 5.0 min with λ_max at 287 nm (Figure 4B). Formation of this species was dependent on the presence of both PyfB and O₂ (Figure 4B). Indeed, the ESI-HRMS analysis of this new species indicated a monooxygenated product with an m/z of 171.0045 (caled for C₅H₃N₂O₅⁻ [M─H]⁻: 171.0047), which is 16 Da greater than that of 21. To confirm that this product is derived from 21, [¹³C₅]-21 was synthesized and used as the substrate. The resulting product showed a 5 Da increase in exact mass (Figure S4), indicating that it is indeed derived from 21.

Figure 4.

(A) Schematic and (B) HPLC analysis of the PyfB reaction using 21 as the substrate. The peak labeled with an asterisk represents an unknown species derived from NAD. (C) ¹³C NMR spectra of [¹³C₅]-23 under C–H coupling or decoupling mode.

The ¹³C NMR spectrum of the product formed from [¹³C₅]-21 is shown in Figures 4C and S5, demonstrating a total of five resonances, including a doublet of doublet of doublets at 195.16 ppm indicative of a carbonyl. The product was thus assigned as 23 (Figure 4A), with the aforementioned resonance corresponding to C4 based on comparison of its three coupling constants (¹J_C4–C3 = 69.6 Hz, ¹J_C4–C5 = 45.1 Hz, ²J_C4-C7 = 5.9 Hz) with those for C3, C5, and C7, respectively. Furthermore, the resonance at 167.79 ppm (dd, ¹J_C7–C3 = 63.8 Hz, ²J_C7–C4 = 5.0 Hz) corresponds to the C7 carboxylate, while the peak at 135.11 ppm (ddd, ¹J_C3–C4 = 68.4 Hz, ¹J_C3–C7 = 63.8 Hz,²J_C3–C5 = 10.8 Hz) is the C3 imine carbon. The latter assignment is consistent with sp² hybridization of C3 in 23 and thus the absence of a C3–H coupling interaction in the ¹³C NMR, as well as the similar chemical shift (149 ppm) measured for the hydrazone carbon in 13.¹³ The doublet at 175.57 ppm (¹J_C6–C5 = 52.8 Hz) is consistent with C6, which is tentatively assigned as an iminol; however, the tautomeric amide remains possible. The remaining C5 signal is a doublet of doublet of doublets at 75.12 ppm (¹J_C5–C6 = 54.1 Hz, ¹J_C5–C4 = 44.9 Hz, ²J_C5–C3 = 11.4 Hz), consistent with the CHOH moiety. To verify that there is one hydrogen atom attached to C5, the ¹³C NMR experiment was repeated in the C–H coupling mode. As anticipated, the C5 carbon shows an additional ¹J_C–H coupling of 152.1 Hz (Figure 4C). Hence, PyfB is proposed to catalyze the dihydroxylation of 21 to form 22, analogous to the reaction catalyzed by PDO, followed by dehydration of the C4 gem-diol to yield 23 (Figures 3A and 4A). Alternatively, dihydroxylation at C5 and C6 could also lead to 23 (Figure S6).

During the characterization of 23, it was found that 23 is prone to slow degradation with a half-life on the order of hours. Specifically, 23 degraded to two new species with high-performance liquid chromatography (HPLC) retention times of 4.8 and 5.1 min after being kept at room temperature overnight in the absence of any enzymes (Figure 5B). Furthermore, HPLC analysis showed that the species at 5.1 min had both the same retention time and ESI-HRMS profile as pyrazole 4 (calcd for C₅H₃N₂O₅⁻ [M–H]⁻: 171.0047, obsd 171.0049) (Figure 5B). Likewise, incubation of this product with PRPP and the C-glycosidase PyfQ, followed by treatment with calf intestinal alkaline phosphatase, generated the C-nucleoside 27 (calcd for C₉H₁₁N₂O₇⁻ [M–H]⁻: 259.0572, found: 259.0577; Figure 5C). Therefore, pyrazole 4 can be generated upon nonenzymatic degradation of the PyfB product 23. The species eluting at 4.8 min had an m/z at 159.0041, which increased by 4 Da when [¹³C₅]-23 was used as the starting material (Figure S7), indicating the loss of one carbon atom from 23 and the chemical formula C₄H₄N₂O₅ (calcd for [M–H]⁻: 159.0047). While the structure of this product remains to be fully characterized due to difficulties in its isolation, it is unlikely to be related to pyrazole formation. Nevertheless, the above finding is significant, as it led to the proposal that 23 may undergo hydrolysis to 24, followed by tautomerization to 25 before cyclization via intramolecular imine condensation to afford pyrazole 4 (Figure 5A). While 23 can nonenzymatically degrade to 4, it cannot be excluded that an enzyme is involved in this transformation. Indeed, the involvement of an enzyme, which has not yet been identified, could help prevent formation of the unidentified side product and improve overall flux through the biosynthetic pathway.

Figure 5.

(A) Scheme of nonenzymatic ring contraction of 23 to pyrazole 4 and subsequent PyfQ-catalyzed C-glycosidation. (B) HPLC analysis of 23 in the absence of enzymes: (i) as-purified 23, (ii) overnight incubation of 23 in 50 mM HEPES (pH7.5) buffer, and (iii) standard of 4. The asterisk-labeled peak eluting at 4.8 min in trace ii is another degradation product of 23 but is not related to pyrazole formation. (C) LC-MS analysis of the PyfQ reaction with (i) the species eluting at 5.1 min (i.e., 4) in trace ii of Figure 5B and (ii) the standard of 4. In both cases, the product obtained is 27.

The fact that PyfB catalyzes the oxidation of 21 to 23, which undergoes nonenzymatic conversion to biosynthetic intermediate 4, suggests that 21 is likely an intermediate of the pyrazofurin biosynthetic pathway. To form 21 from 13, the N-aminoacyl group of the PyfK product 13 must be removed, which may be catalyzed by the amidohydrolase PyfA. To test whether PyfA catalyzes hydrolysis of the amide bond in 13 to release free hydrazone 28, 13 (0.9 mM) was prepared upon incubation of PyfJ and PyfK with 10 and different amino acids in a 50 mM HEPES buffer (pH 7.5) before filtration to deproteinize the solution. This was followed by the addition of PyfA to 6 μM and further incubation at room temperature for 2 h (Supporting Information). The expected hydrazone 28, however, was not detected by the LC-MS analysis of the reaction mixture, and substrate 13 remained essentially unconsumed, regardless of the aminoacyl moiety (Figure S8). These results showed that PyfA does not recognize 13 as a substrate (Figure 6) and thus implied that the backbone of 13 must first be modified prior to the hydrolysis of the Naminoacyl group.

Figure 6.

Proposed biosynthesis of pyrazole 4.

Hence, it was postulated that 13 first undergoes hydroxylation at C3 before being accepted by PyfA and hydrolysis of the N-aminoacyl group (13 → 29 → 30, Figure 6). The hydroxylated hydrazone 30 may then undergo intramolecular cyclization to the six-membered 4,5-dihydropyridazine 31 prior to desaturation to yield PyfB substrate 21 (Figure 6). Among the remaining four uncharacterized gene products related to pyrazole formation, either or both of the two FMN-dependent monooxygenases PyfM and PyfN may be responsible for the proposed C3 hydroxylation of 13, while PyfA could mediate the intramolecular condensation of 30 in addition to the hydrolysis of 29. Lastly, the desaturation from 31 to 21 would then be catalyzed by the FAD-dependent oxidoreductase PyfO.

However, attempts to test the functions of PyfM and PyfN were hampered by difficulties with heterologous expression in E. coli, which led to inclusion bodies in both cases. Moreover, the preparation of the proposed intermediates 29, 30, and 31 also turned out to be difficult. Thus, it was not able to test PyfA and PyfO directly with these hydroxylated compounds. Despite these setbacks, PyfA did exhibit cyclization activity with the nonhydroxylated hydrazone 28 to yield 34, which was generated in situ from hydrazine (32) and α-ketoglutaric acid (33) (Figure 7A,B). Furthermore, incubation of 34 with PyfO aerobically also led to the formation of 16 based on LC-MS analysis and comparison with an authentic standard (Figure 7A,C). These results provide compelling support for the proposed activities of PyfA and PyfO and thus the pathway proposed in Figure 6.

Figure 7.

(A) Reactions catalyzed by PyfA and PyfO using nonhydroxylated substrates 28 and 34, respectively. HPLC analyses of (B) the PyfA reaction with 28 and (C) the PyfO reaction with 34.

CONCLUSIONS

In summary, the biosynthesis of the pyrazole ring in pyrazofurin is proposed to follow the pathway shown in Figure 6. Either or both of the two FMN monooxygenases PyfM and PyfN are proposed to be involved in C3 hydroxylation of 13. The resulting product 29 could then serve as the substrate for amidohydrolase PyfA that may cleave the N-aminoacyl group and further mediate intramolecular cyclization to yield dihydropyridazine 31. The FAD-dependent oxidoreductase PyfO could catalyze the oxidation of 31 to pyridazine 21. While direct evidence supporting the proposed conversion of 13 to 21 remains to be acquired, the cyclase and desaturase activities observed for PyfA and PyfO with deoxysubstrate analogs 28 and 34, respectively, are consistent with this hypothesis. Moreover, definitive evidence is now provided for the oxygenation of 21 to yield the pyridazine 23 catalyzed by the Rieske enzyme PyfB. Nonenzymatic rearrangement of 23 via a ring contraction produces the pyrazole 4, which is the substrate for the C-glycosidation reaction.^7,8 Formation of the pyrazole during the biosynthesis of formycin A is thus proposed to involve a similar process with the addition of an extra amination reaction (Figure S9).

Taken together, this study provides critical insight into the biosynthesis of pyrazofurin and formycin A. In particular, the results presented here reveal a unique pathway to construct the pyrazole core in nature, which entails the formation of a six-membered pyridazine ring prior to its rearrangement to a fivemembered pyrazole ring. While the details of pyridazine 21 formation remain to be fully characterized, the proposed overall biosynthetic pathway is based on a sound foundation since the activities of the enzymes involved in the pathway have been unambiguously demonstrated. Thus, this work sets the stage for further investigation of the biosynthesis of pyrazole-containing natural products, and the outcomes hold promise for engineering methods to prepare pyrazole derivatives that may be of value in clinical applications.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c15879.

• Additional experimental details, materials, and methods, including chemical synthesis, protein expression and purification, in vitro enzymatic assays, supplementary tables and figures, and NMR spectra of synthetic compounds (PDF)