AoiQ Catalyzes Geminal Dichlorination of 1,3-Diketone Natural Products

Abstract

Enzymes that can perform halogenation of aliphatic carbons are of significant interests to the synthetic and biocatalysis communities. Here we describe the characterization of AoiQ, a single-component flavin-dependent halogenase (FDH) that catalyzes gem-dichlorination of 1,3-diketone substrates in the biosynthesis of dichlorodiaporthin. AoiQ represents the first biochemically reconstituted FDH that can halogenate an enolizable sp³ hybridized carbon atom.

Graphical Abstract

Halogenation is one of the most important chemical transformations in organic synthesis. Introduction of halogen atoms can profoundly improve the biological activities and properties of small molecules.¹ Synthetically, carbon–halogen bonds can enable nucleophilic substitutions or metal-catalyzed cross-coupling reactions.² Diverse halogenation methodologies have been developed, with a continuing emphasis on improving regioselectivity and reaction conditions.^2,3 In parallel, Nature has evolved a family of halogenases in the biosynthesis of more than 5000 halogenated natural products,^4–6 some of which have been used in chemoenzymatic synthesis.^7–11 Among the characterized halogenases, the most versatile are flavin-dependent halogenases (FDHs) that generate hypohalous acid (HOX) through attack of a halogen anion (X⁻) on a flavin-4a-OOH intermediate. The diffusible HOX is proposed to interact with a lysine side chain either through hydrogen bonding or forming a haloamine (N₆-X-Lys). Either HOX or N₆-X-Lys could be the donor of X⁺ to carbon nucleophiles.^12–15 So far, nearly all FDHs are associated with halogenation of aromatic substrates (Figure S1).^16,17

In contrast, the roles of FDHs in catalyzing halogenation of aliphatic carbon nucleophiles are largely unexplored. Activated sp³ carbons, such as the α-carbon of a 1,3-diketo moiety (pKa ~9), would in theory serve as a suitable nucleophile.¹⁸ Müller and coworkers demonstrated through an in vivo biotransformation study that dichlorination of chlorotonil A could be facilitated by the FDH CtoA (Figure 1A).¹⁹ Aliphatic FDHs CmlS and PloK have been identified to function in the biosynthesis of chloramphenicol²⁰ and 3-chloro-6-hydroxymellein,²¹ respectively. Although the X-ray crystal structure of CmlS is available²⁰, biochemical confirmation of these FDHs is lacking. It was therefore exciting when Chankhamjon et al. reported the discovery of AoiQ, a bifunctional methyltransferase (MT)-FDH fusion protein from Aspergillus oryzae that catalyzes chlorination of orthosporins (1–3) (Figure 1B).²² Under the reported AoiQ assay conditions, chlorinated products derived from orthosporin (4) were observed. In 4, the proposed site of chlorination is α to the hydroxyl group, which is a highly unactivated position (pKa > 40).¹⁸ Generating a carbon nucleophile at C11 is therefore incompatible with conventional electrophilic mechanism of FDHs.^12,13,17 To account for this, AoiQ was proposed to catalyze the chlorination of 4 to 2 via a radical mechanism (Figure 1B).²² Such unusual proposed activity of AoiQ therefore warrants further investigation of the halogenation steps.

Figure 1.

Natural products derived from chlorination of sp³ hybridized carbons are shown in (A). The previously reported enzymatic radical chlorination and methylation catalyzed by AoiQ is shown in (B).

The coumarins 4 and 5 are synthesized by the nonreducing polyketide synthase (NRPKS) AoiG and the MT AoiF encoded in the aoi biosynthetic gene cluster (BGC) (Figure 2A).^22,23 In A. oryzae, aoiQ encoding the FDH-MT is unclustered with aoiG and aoiF, and inactivation of aoiQ abolished the production of 1-3.²² When the genes near aoiQ were reexamined, we found aoiQ lies in the middle of a different BGC (dia), which encodes a NRPKS (diaA), a β-lactamase like enzyme (diaB) involved in product release,²⁴ a short-chain dehydrogenase/reductase (SDR, diaC), and a flavin-dependent monooxygenase (FMO, diaD)(Figure 2A). Searching the genome database showed that homologs of the dia cluster are widely conserved (Figure S2), and all include an ortholog of aoiQ. In the ngv cluster from P. nalgiovense, which is a producer of 3,²⁵ the MT and FDH domains are encoded in separate genes (Figure 2A). Hence, we hypothesize AoiQ has a dedicated role in the dia BGC.

Figure 2.

Biosynthesis of chlorinated orthosporin 1. (A) Comparison of aoi, dia and ngv clusters. In ngv cluster, FDH and MT are encoded by separate genes. (B) LC-MS analysis S. cerevisiae expressing different combinations of the dia genes. For each trace, the gene indicated in red represents addition from the previous construct. (C) Proposed biosynthetic pathway of the aoi and dia clusters.

We expressed combinations of dia genes in Saccharomyces cerevisiae BJ5464-NpgA²⁶ and analyzed the metabolites. When DiaA and DiaB were coexpressed, 6 (50 mg/L) with the minor 8 (8 mg/L) and 9 (5 mg/L) were formed (Figure 2B, ii). NMR analysis confirmed that 6 is the 1,3-diketo-containing coumarin.²⁴ 8 and 9 are derived from ketoreduction of 6 at C10 and C12, respectively (Figure 2B, iv; Tables S9, S11–S12 and Figures S41–S45, S48–S57), which are likely catalyzed by endogenous ketoreductases in yeast.²⁷ The 1,3-diketone in 6 implicates that 6 may be the precursor to 1–3 instead of previously proposed 4.²² To test this, AoiQ was coexpressed with DiaAB. However, while trace amounts of 1 can be detected by selected ion monitoring, 6 was completely consumed. Numerous minor products can be observed, most at titers that are too low to be isolated (Figure 2B, iv). Surprisingly, when the SDR DiaC was coexpressed, the yeast strain produced predominantly 1 (Figure 2B, v). When DiaABC were expressed without AoiQ, the yeast strain produced only the C10-reduced 8. No reduction of 6 to 9 can be detected, suggesting DiaC is site-selective towards the C10 ketone. To test whether 6 can be directly converted to 1, we supplemented 6 to yeast expressing either AoiQ or DiaC, or both. Consistent with the reconstitution data, only the coexpression of DiaC and AoiQ can lead to formation of 1 (Figure S11). In addition to 1, 6 was also reduced to 8, and methylated by the MT domain of AoiQ to 10 (Figure S11, Table S13 and Figures S58–S62). In contrast, when AoiG and AoiF were reconstituted with AoiQ in a heterologous host, we were only able to observe the formation of 5 (4 mg/L) (Table S11 and Figures S5, S39–S40) and methylated 7 (3 mg/L) (Figure 2C, Table S10 and Figures S5, S46–S47). No chlorinated product can be observed.

To generalize this finding to homologous clusters of dia, we reconstituted the NRPKS, β-lactamase, FDH and MT orthologs from ngv, ndl (A. nidulans) and hzn (T. harzianum t-22) (Figure S6). In each case, O-6-methylated 3 was detected as the major product (~5 mg/L) (Table S6 and Figures S35–S36). This indicates the MT from these clusters can only catalyze O-6 methylation, while the MT domain of AoiQ can catalyze both O-6 andO-8 methylations to give 1. More importantly, these studies demonstrate the dia BGCs are responsible for the biosynthesis of 1 or 3. While dia and aoi pathways both produce coumarin-containing polyketides, only the 1,3-diketone 6 produced by DiaAB can be halogenated and further transformed into 1, consistent with the known mechanisms of FDHs.

To understand the sequence of reactions converting 6 to 1, we expressed the MT and FDH domains of AoiQ as standalone proteins AoiQ-MT and AoiQ-FDH, respectively. Purified AoiQ-FDH showed yellow color, and was confirmed to copurify with ~85% FAD occupancy (Figure S8). The function of DiaC in reduction of 6 to 8 was first confirmed in vitro (K_M = 196.7 ± 35.4 μM, k_cat = 0.75 ± 0.1 min⁻¹ and k_cat/K_M = 3.8 × 10³ min⁻¹ M⁻¹) (Figure S22). Next in a coupled assay, 6 (0.5 mM) was added to both AoiQ-FDH and DiaC. We detected two new products with m/z of 304 and 270 (Figure 3B, i), which matched those of the doubly chlorinated 2 and singly chlorinated 12, respectively (Figure 3A). The conversion of 6 to both products, especially to 2, increased when the concentration of AoiQ-FDH was increased (Figure 3B). Both 2 (6 mg/L) and 12 (7 mg/L) were isolated from a heterologous host expressing AoiQ-FDH with DiaAB, and were structurally confirmed (Tables S5 and S14; Figures S10, S30–S34, S63–S67). Mutation of AoiQ-FDH Lys75, which interacts with HOCl, to alanine completely abolished halogenase activity (Figure S17).^12,13 AoiQ-FDH catalyzed chlorination of 6 in the absence of any exogenously added flavin reductase, which indicates AoiQ-FDH is a single-component FDH and can catalyze the reduction of FAD to FADH₂.^28,29 Indeed, AoiQ-FDH has moderate flavin reductase activity as monitored through the consumption of NADH or NADPH (NADH: K_M = 43.5 ± 3.6 μM, k_cat = 0.22 ± 1.0 min⁻¹ and k_cat/K_M = 5.1 × 10³ min⁻¹ M⁻¹, NADPH: K_M = 30.5 ± 6.0 μM, k_cat = 0.04 ± 0.02 min⁻¹ and k_cat/K_M = 1.3 × 10³ min⁻¹ M⁻¹) (Figure S18).

Figure 3.

Verification of AoiQ function. (A)Proposed reaction mechanism of AoiQ-FDH, AoiQ-MT (or NgvF), and DiaC in the biosynthesis of 1 and 3. (B) In vitro reaction of 6 with AoiQ-FDH and DiaC. The peak labeled in * is proposed to be a decomposition product. (C) LC-MS profiles of AoiQ-FDH assays for the conversion 12 to 21 and 2 to 19. (D) Proposed aromatic chlorination catalyzed by AoiQ-FDH.

To complete the enzymatic synthesis of 1, we added either AoiQ-MT or the homolog NgvF to the assay. In a one-pot enzymatic reaction with AoiQ-FDH and DiaC, addition of AoiQ-MT or NgvF and SAM led to the successful conversion of 6 to either 1 or 3, respectively (Figures 3A and S17). To investigate the timing of chlorination vs methylation reaction, we performed a preparative scale reaction with 6 and NgvF, which led to the isolation and characterization of the O-6-methyoxylated 13 (Table S15 and Figures S17A, S68–S72). When 13 was added to AoiQ-FDH and DiaC, only the C10-ketoreduced 14 was observed, and no chlorinated products can be detected. In contrast, adding 2 to AoiQ-MT led to methylation and formation of 1. Therefore, 6 must undergo chlorination prior to methylation. (Figure 3A).

The conversion of 6 to 2 in the above assay points to a sequence of reactions in which AoiQ can gem-dichlorinate the 1,3-ketone, followed by deacetylation, and C10 ketoreduction by DiaC. The α-chlorinated, 1,3-diketone intermediates such as 15 and 17 (Figure 3A) were not detected from assays, indicating deacetylation can take place readily under aqueous conditions. 15 was chemically prepared from 6 using ZnCl₂/Pb(OAc)₄(Table S17 and Figures S75–S79). When placed in buffer for 1 hr, approximately 50% of 15 was converted to the α-haloketone 16 (Table S18 and Figures S80–S84), confirming deacetylation can take place nonenzymatically (Figure S20). Longer incubation resulted in complete decomposition of 16 into a multitude of minor compounds. This is consistent with the lack of products when DiaC was not included in the yeast reconstitution (Figure 2B). Indeed, addition of DiaC to 15 led to formation of 12; while addition of both AoiQ-FDH and DiaC led to conversion of 15 to 2 (Figure S20). The reduction of 16 to 12 by DiaC was kinetically characterized (K_M = 28.5 ± 9.7 μM, k_cat = 7.0 ± 0.8 min⁻¹ and k_cat/K_M = 2.5 × 10⁵ min⁻¹ M⁻¹) to be > 60-times more efficient compared to the reduction of 6 to 8, indicating deacetylated compounds such as 16 preferred by DiaC. α-β scission of a 1,3-diketone after halogenation as seen in 15 to 16, was similarly observed when dibrominated 3-oxo-hexanoyl homoserine lactones (DBOHHL) readily underwent cleavage to α,α-dibromoethanoyl homoserine lactone (DBEHL).³⁰ Such scission can take place with water attacking the β-carbonyl followed by the elimination of the α-halogenated carbonylmethyl group, which is an enhanced leaving group due to stabilization of the negative charge by the halogen substitution (Figure 3A). This is likely the route that forms the terminal dichloromethyl group in chloramphenicol.²⁰

In the heterologous expression of DiaAB and AoiQ-FDH to obtain 2 and 12, we observed an accompanying product 19 (5 mg/L) that is trichlorinated based on MWT (Figures S10). A characteristic bathochromic shift of λ_max caused by aromatic chlorination was also observed for 19 (λ_max = 339 nm) when compared to that of 2 (λ_max = 326 nm) (Figure S23).³¹ 19 was confirmed to be 5-chloro-demethyldichlorodiaporthin (Figure 3A, Table S19 and Figures S85–89). This surprising result indicates that AoiQ-FDH can also catalyze aromatic chlorination. This was demonstrated in an in vitro assay in which 2 is converted to 19 in the presence of AoiQ-FDH (Figure 3C, iii). Similarly, 4 and 12 can be converted to 20 and 21, respectively (Figure 3 and Figure S24, Tables S20–S21 and Figures S90–S99). The C10-reduced 8 was also converted to C5-chlorinated 22 (major) and C7-chlorinated 23 (minor) (Figure 3D, Tables S22–S23, Figures S26, S100–S110). A likely explanation for the ability of AoiQ to halogenate nucleophiles separated by six carbons is that the active site of AoiQ-FDH could accommodate multiple binding modes of the substrates, which enable halogenating species to position close to both nucleophilic carbons.³² While the C11 α-carbon is the preferred nucleophile in 6, the absence of such nucleophile in C10-reduced substrates led to the chlorination of alternative nucleophiles, namely the aromatic C5 through the deprotonation of C6 phenol in the active site. Aromatic chlorination of 4 that yields 20, which is nearly indistinguishable to 12 by retention time and HRMS analysis, could have led to the interpretation of AoiQ being able to chlorinate unactivated sp³ carbons in the initial report.²²

In summary, we have characterized the single component FDH AoiQ that can perform gem-α,α-dichlorination of an enolizable 1,3-diketone substrate, which leads to an unusual polyketide shortening reaction. The ability of AoiQ to halogenate nonaromatic substrates can be explored for use in biocatalysis.