2.1. Cloning and protein purification

The hmo gene (orf22) was amplified from Amycolatopsis orientalis genomic DNA by polymerase chain reaction and was subcloned into the expression vector pET-28a(+) to generate the protein with an N-terminal His₆ tag. The expression plasmid was transformed into Escherichia coli BL21(DE3) cells by electroporation and the cells were grown on LB agar plates containing 50 µg ml⁻¹ kanamycin for 16 h at 37°C. A single colony was grown overnight in 5 ml LB medium containing 50 µg ml⁻¹ kanamycin at 37°C. The cell culture was used to inoculate 1 l LB medium containing 50 µg l⁻¹ kanamycin. For protein expression and purification, the transformed E. coli cells were cultured, induced with 200 µl 1.0 M IPTG at an OD₆₀₀ of 0.6 and grown for a further 24 h at 16°C. The cells were harvested by centrifugation, resuspended in 20 ml binding buffer (20 mM HEPES pH 7.5, 500 mM NaCl, 10 mM imidazole, 10% glycerol) and lysed using a microfluidizer or by sonication. The supernatant of the lysate after centrifugation at 16 000 rev min⁻¹ for 30 min was applied onto an Ni²⁺–NTA resin column. The bound protein was sequentially washed with 10 ml binding buffer (50 mM HEPES pH 8, 200 mM NaCl, 20 mM imidazole) and 10 ml wash buffer (50 mM HEPES pH 8, 200 mM NaCl, 80 mM imidazole) before elution with 10 ml elution buffer (20 mM HEPES pH 8, 500 mM NaCl, 250 mM imidazole). Gel filtration was performed using an ÄKTA FPLC system equipped with a HiLoad 16/60 Superdex 200 column (Amersham Bioscience) under isocratic conditions (20 mM HEPES pH 8, 100 mM NaCl).

2.2. Crystallization and data collection

The purified proteins were crystallized using the hanging-drop vapor-diffusion method. Hmo and its Y128F, Y128C and R163L mutants were concentrated to 7 mg ml⁻¹ in 50 mM HEPES pH 8.0 buffer solution and crystallized using a solution consisting of 35% Tascimate, 0.1 M bis-Tris propane pH 7.0 in a 50:50 volume ratio. Crystals appeared within five days in VDX48 plates (Hampton Research) with sealant at 20°C. For Hmo and its mutants in complex with (S)-mandelate (SMA), benzoylformate (BF), benzoic acid (BA), phenyl­pyruvate (PPY), oxaloacetate (OAA) or mandelamide (MAAD), each crystal was soaked with 10–30 mM of the designated ligand dissolved in the mother solution for between 10 min and 24 h before data collection. All crystals were transferred to a solution containing cryoprotectants and flash-cooled in liquid nitrogen prior to data collection. The cryoprotectant-containing solution for Hmo and the Y128F mutant consisted of 20%(w/v) glycerol and 35% Tascimate. X-ray diffraction data were recorded at an operating temperature of 100 K using an ADSC Quantum 315 or an MX300HE CCD detector on beamlines 13B1, 13C1, 15A1 or 05A at the National Synchrotron Radiation Research Center, Taiwan or beamline 44XU at SPring-8, Japan. All of the crystals belonged to the same space group: I422.

2.3. Structure determination and refinement

Data were indexed and scaled using the HKL-2000 package (Otwinowski & Minor, 1997). The crystal structures were determined by the molecular-replacement (MR) method using Phaser-MR from the CCP4 suite (McCoy et al., 2007; Winn et al., 2011). The crystal structure of hydroxyacid oxidase (PDB entry 3sgz; Chen et al., 2012) was used as the search model for solving the initial phase. The polypeptide structures were built and refined using REFMAC (Murshudov et al., 2011). Subsequent iterative cycles of model building and refinement were performed using Coot (Emsley et al., 2010) and PHENIX (Afonine et al., 2012). All non-H atoms were refined with anisotropic displacement parameters. Both protein structures and electron-density maps were generated using PyMOL (DeLano, 2002). Detailed refinement statistics are presented in Table 1.

Table 1 Data-collection and refinement statistics for Hmo and its Y128F, Y128C and R163L mutants Values in parentheses are for the highest resolution shell. SMA, (S)-mandelate; BF, benzoylformate; FMN, riboflavin mononucleotide; 5DAFMN, 5-­deazariboflavin mononucleotide; PPY, phenylpyruvate; MA_FMN, malonyl–FMN; MAAD_FMN, mandelamide–FMN.   Hmo Hmo–SMA Hmo–BF Y128F Y128F–SMA Y128F–BF Y128F–5DAFMN Y128F–5DAFMN–BA PDB code 5zzp 5zzr 6a08 6a13 6a0v 6a19 6a1h 6a1l Data collection  Wavelength (Å) 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0  Space group I422 I422 I422 I422 I422 I422 I422 I422  a, b, c (Å) 137.9, 137.9, 112.3 137.7, 137.7, 111.6 137.9, 137.9, 111.8 137.5, 137.5, 112.0 137.8, 137.8, 111.8 137.9, 137.9, 112.3 137.4, 137.4, 112.4 137.7, 137.7, 111.6  α, β, γ (°) 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90  Resolution range (Å) 30–1.39 (1.44–1.39) 30–1.31 (1.36–1.31) 30–1.55 (1.61–1.55) 30–1.70 (1.76–1.70) 30–1.39 (1.44–1.39) 30–1.55 (1.61–1.55) 30–1.36 (1.41–1.36) 30–1.40 (1.45–1.40)  Rmerge† (%) 3.7 (73.0) 3.7 (67.0) 4.0 (79.0) 3.5 (69.2) 4.3 (73.1) 3.6 (75.0) 2.8 (62.0) 4.2 (81.0)  〈I/σ(I)〉 36.1 (2.4) 34.5 (2.3) 43.1 (2.9) 41.3 (3.1) 41.0 (2.9) 32.9 (2.7) 41.2 (3.4) 29.6 (2.4)  Completeness (%) 99.9 (100.0) 99.0 (100.0) 100.0 (100.0) 100.0 (100.0) 100.0 (100.0) 99.9 (100.0) 99.9 (100.0) 100.0 (100.0)  Multiplicity 9.4 (9.4) 9.8 (9.5) 12.8 (11.2) 11.2 (11.1) 12.2 (12.0) 9.6 (9.2) 9.7 (9.4) 10.5 (10.4) Refinement  Resolution range (Å) 30–1.39 (1.44–1.39) 30–1.31 (1.36–1.31) 30–1.55 (1.61–1.55) 30–1.70 (1.76–1.70) 30–1.39 (1.44–1.39) 30–1.55 (1.61–1.55) 30–1.36 (1.41–1.36) 30–1.40 (1.45–1.40)  Rwork‡ (%) 17.0 (28.0) 16.6 (22.9) 16.9 (22.0) 17.8 (24.2) 16.7 (22.9) 15.7 (24.4) 16.8 (23.0) 17.5 (28.0)  Rfree§ (%) 18.3 (30.0) 17.7 (22.3) 17.9 (20.1) 20.0 (26.2) 18.4 (23.1) 17.3 (26.6) 18.1 (26.0) 18.6 (30.1)  R.m.s. deviations   Bond lengths (Å) 0.010 0.009 0.008 0.018 0.008 0.009 0.017 0.006   Bond angles (°) 1.41 1.41 1.26 1.490 1.3361 1.460 1.421 0.982  No. of reflections 100548 112497 75720 58599 98014 74156 114232 104841  No. of atoms   Protein 2785 2622 2684 2510 2636 2703 2548 2574   Ligand/ion 31 53 65 31 53 53 31 55   Water 403 402 329 316 400 400 440 335  B factors (Å2)   Protein 18.2 19.7 19.1 18.6 16.9 21.9 18.4 20.5   Ligand/ion 10.7 19.0 20.6 10.8 17.8 29.4 12.1 17.1   Water 32.4 33.9 31.7 32.1 32.2 37.3 34.9 26.6   Y128F–5DAFMN–BF Y128F–5DAFMN–PPY Y128F–PPY–FMN Y128F–MA_FMN Y128F–BF–FMN Y128C–BF R163L–MAAD_FMN PDB code 6a1m 6a1p 6a1r 6a21 6a23 5zzz 6a3t Data collection  Wavelength (Å) 1.0 1.0 1.0 1.0 1.0 1.0 1.0  Space group I422 I422 I422 I422 I422 I422 I422  a, b, c (Å) 137.9, 137.9, 111.2 138.0, 138.0, 111.1 138.1, 138.1, 112.1 137.6, 137.6, 112.1 137.8, 137.8, 111.8 138.2, 138.2, 112.1 137.4, 137.4, 111.7  α, β, γ (°) 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90  Resolution range (Å) 30–1.55 (1.61–1.55) 30–1.51 (1.56–1.51) 30–1.65 (1.71–1.65) 30–1.50 (1.55–1.50) 30–1.65 (1.71–1.65) 30–1.45 (1.50–1.45) 30–2.51 (2.60–2.51)  Rmerge† (%) 3.9 (57.0) 3.6 (72.1) 4.4 (73.0) 3.2 (64.3) 3.8 (70.1) 4.4 (69.2) 10.1 (66.0)  〈I/σ(I)〉 40.8 (2.4) 30.4 (2.3) 36.8 (3.16) 40.0 (3.2) 28.4 (2.0) 33.5 (2.6) 15.2 (2.4)  Completeness (%) 99.7 (97.2) 99.8 (98.3) 100.0 (100.0) 99.9 (100.0) 99.3 (100.0) 99.9 (100.0) 99.9 (100.0)  Multiplicity 11.4 (9.1) 9.4 (7.7) 12.2 (12.1) 9.8 (9.8) 9.1 (8.8) 9.9 (9.9) 9.5 (9.3) Refinement      Resolution range (Å) 30–1.55 (1.61–1.55) 30–1.51 (1.56–1.51) 30–1.65 (1.71–1.65) 30–1.50 (1.55–1.50) 30–1.65 (1.71–1.65) 30–1.45 (1.50–1.45) 30–2.51 (2.60–2.51)  Rwork‡ (%) 17.8 (25.7) 17.0 (26.1) 16.3 (20.6) 16.7 (22.4) 17.3 (24.1) 17.5 (22.8) 17.8 (21.6)  Rfree§ (%) 20.1 (26.5) 19.7 (26.8) 18.5 (22.5) 19.0 (22.9) 19.4 (24.9) 19.2 (27.2) 22.4 (29.1)  R.m.s. deviations   Bond lengths (Å) 0.018 0.015 0.019 0.018 0.006 0.010 0.008   Bond angles (°) 1.501 1.357 1.529 1.557 1.173 1.300 1.29  No. of reflections 75042 83439 64602 84212 59812 93717 17837  No. of atoms   Protein 2485 2539 2586 2626 2606 2481 2496   Ligand/ion 53 44 41 37 51 42 43   Water 298 331 303 375 301 357 91  B factors (Å2)   Protein 19.3 22.2 17.9 19.0 19.1 18.7 40.29   Ligand/ion 14.8 18.2 11.7 16.7 16.6 20.0 40.89   Water 34.2 35.2 30.2 33.2 30.4 31.6 38.94 †Rmerge = , where 〈I(hkl)〉 is the average intensity value of the equivalent reflections. ‡Rwork = . §Rfree was calculated from 5% of data that were randomly excluded from refinement.

2.4. Site-directed mutagenesis

The site-directed Hmo point mutants Y128C, Y128F and R163L were generated using QuikChange (Stratagene). The primers for mutant preparation are listed in Supplementary Table S1. The mutations were confirmed by DNA sequencing. Each mutated protein was purified using the same protocol as used for wild-type Hmo.

2.5. Chemoenzymatic synthesis of 5-deazaflavin mononucleotide

To obtain 5-deazaflavin mononucleotide, the commercially available compounds 3,4-dimethylaniline, sodium cyanoboro­hydride and D-ribose were used as starting materials, following the synthetic procedure described in Supplementary Scheme S1. Chemically synthesized deazariboflavin was purified using column chromatography, characterized by mass spectrometry or NMR and converted to catalytically active 5-deaza-FMN by riboflavin kinase. The gene coding for riboflavin kinase was amplified from the genomic DNA of E. coli K12 and was subcloned into a pET-28a vector to carry an N-terminal His₆ tag. The constructed plasmid was then transformed into E. coli BL21(DE3) cells for overexpression. The overexpressed protein was purified using Ni²⁺–NTA affinity chromatography and concentrated to 15 mg ml⁻¹. The reaction was initiated by the addition of 2 mg ml⁻¹ riboflavin kinase to a reaction solution consisting of HEPES pH 7.5 with 5 mM ATP, 5 mM MgSO₄ and 10 mM 5-deazariboflavin overnight. The sample solution was purified by HPLC and confirmed by mass spectrometry.

2.6. Preparation of 5-deaza-FMN-containing Hmo and its Y128F mutant

The deflavination and reconstitution of Hmo and its Y128F mutant were performed on an Ni²⁺–NTA resin column by circulating a solution of HEPES buffer containing 10 mM deazariboflavin (Hefti et al., 2003). The Ni²⁺–NTA resin column was loaded and saturated with a solution of the target protein at a flow rate of 0.2 ml min⁻¹ using a peristaltic pump, which was installed between the sample reservoir and the column, to ensure maximal binding. Ten volumes of denaturing buffer (50 mM HEPES, 2 M potassium bromide, 2 M urea pH 8.0) were subsequently used to unfold Hmo or its Y128F mutant, allowing removal of the FMN prosthetic cofactor from Hmo or its Y128F mutant. On-column replacement with 5-deaza-FMN was completed by circulating a buffer solution consisting of 50 mM HEPES pH 8.0, 10 mM 5-deaza-FMN at 0.2 ml min⁻¹ overnight. Hmo or its Y128F mutant containing 5-deaza-FMN was eluted using 10 ml elution buffer (20 mM HEPES pH 8, 200 mM NaCl, 250 mM imidazole). An analytical gel-filtration experiment using a Superdex 200 16/20 column confirmed that 5-deaza-FMN-containing Hmo (5-deaza-Hmo) and its Y128F mutant have an unchanged tetrameric state after the refolding process. The purified enzymes were concentrated to 10 mg ml⁻¹ for enzymatic assays and protein crystallization.

2.7. Enzymatic assays using 5-deaza-Hmo and its Y128F mutant

A typical enzymatic reaction (10 µg 5-deaza-Hmo or its Y128F mutant) was carried out in 200 µl buffer solution (20 mM HEPES, 100 mM NaCl pH 7.5, 5 mM S-mandelate, benzoylformate or benzoic acid) at 25°C for 2 h. Reactions were quenched using 6 N HCl and subjected to HPLC-MS analysis (using an Agilent 1260 Infinity Quaternary LC module connected to a Thermo-Finnigan LTQ-XL). Analytes were separated using a reverse-phase C₁₈ column (4.6 × 250 mm, 5 µm, C₁₈ Prodigy, Phenomenex) at a flow rate of 1 ml min⁻¹ in a mobile system programmed as a linear gradient from 2% to 40% solvent B against solvent A over 25 min followed by 98% solvent B for a further 8 min (solvent A, water with 1% formic acid; solvent B, acetonitrile with 1% formic acid). The data were recorded using a UV detector with the wavelength set to 254/280 nm and the mass spectrometer in positive mode.

2.8. UV–Vis absorption measurements of reactions of Hmo and its Y128F mutant

Ultraviolet and visible absorption spectra were recorded using a Beckman spectrophotometer (DU-800). In aqueous solution, freshly prepared Hmo or its Y128F mutant (0.125 mM in 0.05 M HEPES, 0.1 M NaCl pH 7.5) were mixed with substrates (2.5–5 mM) and incubated in a quartz cuvette at ambient temperature for 2 h. The absorption spectra of the reactions were recorded from 300 to 600 nm. For the redissolved crystals/crystalloids, more than 100 crystals of Hmo or its Y128F mutant (after soaking with substrates at 5 mM for 2 h) were picked from hanging-drop crystallization plates and redissolved in the mother liquor in a UV quartz cuvette before spectral scanning.

3.1. Inhibition of α-ketoacids

To investigate the mechanism of the four-electron oxidative decarboxylation reaction catalyzed by the Hmo single mutant Y128F, we first solved crystal structures of the Y128F mutant in complex with different ligands such as (S)-mandelate, (S)-2-phenylpropionate, benzoylformate, benzaldehyde and benzoate. The ternary complexes of Hmo and its Y128F mutant were then compared, whereupon it was observed that the superpositioned complexes showed no apparent structural variations between Hmo and the Y128F mutant (r.m.s.d. of <0.1 Å; Supplementary Fig. S2) in terms of chemical conformations and spatial positions of both the proteins and ligands. When the crystals were soaked with benzoylformate, the formation of an N5-benzoyl-FMN adduct with an N5–C′α linkage was observed [Fig. 2(a)]. MS analysis confirmed that the benzoyl moiety was covalently linked to FMN [Fig. 2(b)]. When the crystals were soaked with benzaldehyde, the formation of a covalent adduct was not observed. This outcome suggests that the α-ketoacid moiety is a prerequisite for the formation of the covalent adduct, in which the decarboxylation of the terminal carboxyl group of the α-ketoacid is likely to take place after the formation of the N5–C′α linkage. Moreover, when the crystals of the Y128F mutant were soaked with (S)-3-phenyllactate, phenylpyruvate or phenylacetate [Fig. 2(a)], we observed that (S)-3-phenyllactate was oxidized to phenylacetate (a four-electron oxidative decarboxylation), phenylpyruvate ended up as an N5-phenylacetyl-FMN adduct and phenylacetate stayed as it was. When the crystals were soaked with oxaloacetate, an N5-malonyl-FMN adduct was found [Fig. 2(a)], leading overall to the conclusion that the formation of the covalent N5 adducts is α-ketoacid-dependent.

Figure 2 Crystal structures of acyl-FMNred adducts and inhibition mechanism by α-ketoacids. (a) Structures of acyl-FMNred adducts in crystals of the Y128F mutant soaked with benzoylformate (left), phenylpyruvate (center) or oxaloacetate (right). The flavin adducts all are at the si-face of the isoalloxazine ring. (b) LC traces and mass spectra of FMN (i) and phenylacetyl-FMNred (ii). (c, d) Weighted 2Fo − Fc electron-density maps (gray) and unbiased Fo − Fc difference OMIT electron-density maps (blue) for FMN in Hmo (c) and the Y128F mutant (d) without (left) or with a ligand (S-mandelate, center; benzoylformate, right), where the extent of polarization is justified by OMIT electron density (the wild type or Y128F mutant and the absence or presence of a ligand seem to be key factors). The 2Fo − Fc electron-density map is contoured at 2σ; the unbiased Fo − Fc OMIT difference electron-density map is contoured at 4σ. Free ligands, FMN, FMN adducts and active-site residues are colored cyan, yellow, orange and green, respectively. See Supplementary Figs. S3(a), S3(b) and S2(c) for stereoviews and Fo − Fc difference electron-density maps.

In most flavin-dependent oxidoreductases the FMN_ox cofactor acts as an electron sink (electrophile) that accepts electrons or hydrides conveyed from a substrate or NADH/NADPH in the reductive half-reaction. The nitroalkane oxidase from the fungus Fusarium oxysporum and the alkyldihydroxyacetone phosphate synthase in human fibroblasts are two rare cases in which a carbanion is generated at the active site prior to addition to N5 of the flavin cofactor as a covalent adduct (Razeto et al., 2007; Heroux et al., 2009). However, a question arises in the context of how two electrophiles (FMN_ox and α-ketoacid) are covalently associated in the Y128F mutant. One likelihood is that the α-ketoacid undergoes decarboxylation in the first instance to form a localized C′α carbanion that then acts on FMN_ox, but an α-ketoacid that spontaneously undergoes decarboxylation is chemically untenable. The second possibility is that the sp² N5 atom of FMN_ox acts as a nucleophile, but this scenario likewise contradicts the current understanding: FMN_ox is a strong electrophile that accepts electrons. Nevertheless, an extra chunk of electron density at the top of N5 of FMN_ox was observed in unbiased difference electron-density maps, and this electron density was denser in the Y128F mutant than in the wild type [Figs. 2(c) and 2(d)]. This additional electron density suggests that the C4α=N5 double bond in FMN_ox is polarized to a C4α⁺–N5⁻ ylide (a tertiary carbocation and a tetrahedral sp³ amine anion), as manifested by uneven wedge-shaped electron density for the π-bond [Figs. 2(c), 2(d), 3(a) and 3(b)]. The extent of polarization appears to be a function of an active-site perturbation ensemble (e.g. the point mutation), reflecting cooperative interplay of the hydrogen-bond network between water, FMN and active-site residues as well as ligands [Figs. 3(c) and 3(d)]. The formation of the adduct is thereby proposed to take place as follows: the sp³ N5 atom of the polarized FMN_ox attacks the carbonyl C atom of the α-ketoacid to form a covalent C′α–N5 adduct, whereupon decarboxylation takes place, resulting in a localized C′α carbanion. The lone pair of the C′α carbanion subsequently hybridizes with the π orbital of N5 to reinstate the neutrality of C4α. Upon collapse of the C′α oxyanion, N5-acyl-FMN_red results via a series of bond rearrangements. This species has a hydroquinone-like structure, with the acyl moiety protruding out of the plane defined by the isoalloxazine ring of FMN_ox [Fig. 2(a)].

Figure 3 The effect of 4-OH of Tyr128 on polarization, hydrogen-bond networking and reactivity. (a, b) A close-up view of the wedge-shaped electron density on top of C4α=N5 of FMNox shown by unbiased difference electron-density maps (blue, positive electron density) contoured at 4σ in the wild type (a) and the Y128F mutant (b), suggesting that the electrons in the π-orbital of C4α=N5 are polarized from C4α to N5 (to form a C4α+–N5− ylide), with this being more significant in the Y128F mutant than in the wild type. (c, d) The point mutation Y128F disturbs the active-site hydrogen-bonding network between water, FMN and the catalytic dyad [the Y128F mutant loses the hydrogen bond between Tyr128 and H2O (187) but gains a new hydrogen bond between His252 and H2O (257)]. (e) Hydrogen peroxide was modeled at C4α, where the distance between Tyr128 and the terminal O atom of Cα-OOH is 2.7 Å. (f, g) The phenyl ring of benzoylformate, which is bulkier and takes up space, limits the access of dioxygen to the reaction center, while the methyl group of pyruvate, which is smaller and takes up less space, allows the access of dioxygen to the reaction center [the phenyl ring that bulges out at the substrate entrance in (f) prohibits exposure of FMNred to the bulk solvent, as opposed to the methyl group in (g) which allows exposure of FMNred to the bulk solvent]. Therefore, the size of the substrates is another factor in leverage of the oxidation cascade. (h) Aside from the active-site perturbation effect, the absence of the p-­OH group also introduces some space allowing access of O2 to the C4α redox-active center. (i) The sulfhydryl group (SH) of the Y128C mutant has been oxidized to a sulfenyl group (S-OH), as it is vulnerable to ROS generated in the active site. Free ligands, FMN and active-site residues are colored cyan, yellow and green, respectively.

The UV–Vis spectrum of the Y128F mutant protein solution exhibits a typical FMN_ox absorbance profile [two absorbances at 370 and 450 nm; Fig. 4(a), i]. FMN_ox turned colorless when phenylpyruvate was added to the solution, suggesting the formation of an acyl-FMN species [Fig. 4(a), ii]. The spectrum is dissimilar to that observed when phenyllactate was added [in which the two typical absorbances at 370 and 450 nm disappeared, suggesting the reduction of FMN_ox to FMN_red; Fig. 4(a), iv] by a small hump at 340 nm [Fig. 4(a), ii]. The redissolved solution of Y128F mutant crystals/crystalloids that had been pre-soaked with phenylpyruvate also turned colorless, with a profile similar to that in solution [with a smaller hump at 340 nm; Fig. 4(a), iii] (Sucharitakul et al., 2007; Thotsaporn et al., 2011). While O₂ is a small hydrophobic molecule that freely diffuses through spaces and tunnels in protein matrices (Baron, McCammon et al., 2009; Baron, Riley et al., 2009), Hmo may have evolved a discrete channel or pockets that temporarily limit the access of O₂ to C4α of N5-acylated isoalloxazine. The metastable N5-alkyl-FMN_red in aqueous solution is thus attributable to dysfunction of the charge-transfer cage in the absence of Hmo or its Y128F mutant. The inhibited Hmo and mutants identified here differ from the conventional inactivation of FMN_ox, which requires chemically activated agents (Walsh, 1980, 1984).

Figure 4 Spectra, synthesis, reaction and structure of 5-deaza-FMN. (a) UV–Vis spectra of FMN in Hmo and its Y128F mutant: (i) FMNox in Hmo (370/450 nm, green line), (ii) acyl-FMNred in the Y128F mutant [340 nm, blue line; addition of phenylpyruvate (PPY) for 2 h], (iii) acylated FMN in redissolved Y128F mutant crystals/crystalloids soaked with PPY (acyl-FMNred, 340 nm, dotted blue line), (iv) FMNred in Hmo [a shoulder at 330 nm, red; addition of phenyllactate (PLA)]. (b) (i) Chemical synthesis and LC purification of 5-deazariboflavin (the inset shows the mass spectrum of 5-deazariboflavin), (ii) enzymatic synthesis and LC purification of 5-deaza-FMN (the inset shows the mass spectrum of 5-deaza-FMN). (c) Enzymatic reactions of Hmo and its Y128F mutant harboring 5-deaza-FMN: (i) enzymatic reaction with Hmo harboring 5-deaza-FMN in the presence of S-mandelate, (ii) enzymatic reaction with the Y128F mutant harboring 5-deaza-FMN in the presence of S-mandelate, (iii) control reaction of wild-type Hmo in the presence of S-mandelate, (iv) control reaction of the Y128F mutant in the presence of S-mandelate. (d, e) Crystal structures of the Y128F mutant harboring 5-deaza-FMN soaked with S-mandelate (d) or phenyllactate (e), which have been transformed into benzoylformate or phenylpyruvate, respectively. Unlike FMN in the wild type or the Y128F mutant, no electron density emerges at the top of C5 or between C′α and C5. (f) The structure of an α-mandelamide–N5-FMNred adduct in the crystal of the R163L mutant soaked with nondecarboxylable α-mandelamide (the chemical structure is shown). The flavin adduct is on the si-face of the isoalloxazine ring. The 2Fo − Fc electron-density map is contoured at 2σ. The unbiased Fo − Fc difference electron-density map is contoured at 4σ in positive electron density. Free ligands, FMN, FMN adducts and active-site residues are colored cyan, yellow, orange and green, respectively. See Supplementary Figs. S2(j)–S2(m) for stereoviews and 2Fo − Fc difference electron-density maps.

3.2. 5-Deaza-FMN in oxidation

A photoreduction mechanism has been proposed for the LMO-mediated oxidative decarboxylation reaction (Ghisla & Massey, 1977, 1989; Ghisla et al., 1979), in which the reactants need to be activated by photosensitization. In the present case (Hmo and its Y128F mutant), the C4α⁺−N5⁻ ylide of FMN_ox appears to be the key factor. To validate this proposition, we chemoenzymatically synthesized 1 g of the FMN analog 3,10-dimethyl-5-deaza-isoalloxazine ribitol phosphate (5-deaza-FMN_ox) following previously reported methods [Fig. 4(b); the modified method is described in the supporting information] (Carlson & Kiessling, 2004; Kittleman et al., 2007; Mansurova et al., 2008; Osborne et al., 2000). Biochemically, 5-deaza-FMN-containing Hmo or its Y128F mutant is able to oxidize (S)-mandelate to benzoylformate but not to benzoate, confirming that the enzyme was refolded successfully as the wild type and supporting N5 as the pivotal factor in the oxidative decarboxylation reaction [Fig. 4(c)]. Similarly, benzoylformate but not benzoate was found in 5-deaza-FMN_ox-containing crystals of Hmo or its Y128F mutant soaked with (S)-mandelate [Figs. 4(d) and 4(e)]. No N5-acyl adducts can be found in 5-deaza-FMN-containing crystals of Hmo or its Y128F mutant soaked with benzoylformate or phenylpyruvate at various concentrations at different time intervals. Furthermore, the lack of visible electron density at the top of C5 or between C5 and C′α of benzoylformate (4.0 Å) suggests that the C4α=C5 double bond in 5-deaza-FMN is less polarizable. Our structural interrogation supports the decarboxylation of the α-ketoacid taking place after or in concert with the formation of the C′α–N5 bond. The R163L mutant (a low-activity mutant) was further examined using the nondecarboxylable substrates α-(S)-mandelamide [2-(S)-hydroxy-2-phenylethylamide] or benzoylamide (2-keto-2-phenylethylamide), in which the former is oxidized to form the latter. When the R163L mutant crystals were soaked with benzoylamide, an α-hydroxyamide-FMN adduct was formed [Fig. 4(f)], leading to the unequivocal conclusion that N5 of FMN_ox has a nucleophilic propensity and that the C′α—N5 bond is formed prior to α-ketoacid decarboxylation.

3.3. Four-electron oxidation to benzoate

We propose that C4α-OOH-N5-acyl-FMN is the key intermediate in the four-electron oxidation of an α-hydroxyacid mediated by Hmo and its Y128F mutant on the basis of the following facts: (i) Hmo and its Y128F mutant catalyze the four-electron oxidation of an α-hydroxyacid via an α-ketoacid to an acid with one O atom from O₂ incorporated into the terminal carboxylic group, (ii) H₂O₂ is not able to oxidize the α-ketoacid in the absence of Hmo or its Y128F mutant, (iii) the pro-R α-ketoacid is covalently linked to FMN_ox in the Y128F mutant, forming an N5-acyl-FMN_red adduct, (iv) the oxidation cascade stalls at the α-ketoacid using Hmo or its Y128F mutant with 5-deaza-FMN_ox in lieu of FMN_ox and (v) the sp³ N5 in FMN_red is highly reactive, as exemplified in UbiX, a flavin prenyltransferase involved in bacterial ubiquinone biosynthesis (White et al., 2015). The formation of the intermediate is somewhat similar to the mechanism proposed for EncM, which catalyses an oxidative Favorskii-type re­arrangement reaction (Teufel et al., 2015). The major discrepancy, however, is that O₂ in Hmo and its Y128F mutant mediates the transient formation of C4α-OOH-N5-acyl-FMN prior to its release as H₂O₂.

Superposition of the benzoylformate-liganded ternary complex of the Y128F mutant with that of the wild type shows no apparent discrepancies (r.m.s.d. of 0.06 Å) except for the p-OH group of Tyr128 (Supplementary Fig. S4). Given that the oxidative decarboxylation of an α-hydroxyacid is cata­lytically executed by the Y128F mutant, the p-OH group ought to play a crucial role in leverage of the oxidation cascade. This effect is commensurate with a recent report that a single mutation, C65D, of phenylacetone monooxygenase converts a monooxygenase to an oxidase (in contrast to this report) by facilitating the discharge of H₂O₂ (Brondani et al., 2014). C4α-OOH-FMN_red, which is a reactive intermediate in a typical monooxygenase/oxidase-catalyzed reaction, was modeled and optimized in the structure of Hmo, in which the p-OH group is within hydrogen-bonding distance (2.7 Å) of C4α-OOH [Fig. 3(e)]. On the basis of this model, the p-OH group is in a position to protonate C4α-OO⁻-FMN_red to form C4α-OOH-FMN_red, thereby neutralizing or facilitating the discharge of H₂O₂ from C4α-OOH. In contrast, C4α-OO⁻-FMN_red may diverge in the absence of the p-OH group. One implication is that the Y128F mutant works like a monooxygenase, whereby Baeyer–Villiger-type reactions result. That is, the C4α-OO⁻ anion attacks the α-carbon (C′α) of pro-R benzoylformate to form a tetrahedral oxyanion species. Upon the collapse of the α-oxyanion the terminal carboxyl group migrates to the distal O atom of C4α-OO⁻ to form a mixed-anhydride species; subsequent hydrolysis would give rise to benzoate and formate (Torres Pazmiño et al., 2010). This type of reaction, however, was ruled out because no benzoate was detected in the reactions catalyzed by the 5-deaza-FMN-containing Y128F mutant. This fact, however, underscores the importance of the sp³ N5 of C4α-OO⁻-FMN_red in the four-electron oxidation reaction, where the reduced or polarized sp³ N5 actually has a better Bürgi–Dunitz angle and is at a short distance from C′α of pro-R benzoylformate, favoring the formation of C4α-OO⁻-N5-alkyl-FMN_red before the release of H₂O₂.

3.4. Proposed catalytic mechanism of oxidative decarboxylation

In a general four-electron oxidation reaction, one molecule of (S)-mandelate should theoretically yield two equivalents of H₂O₂. The molar ratio of H₂O₂ versus benzoate, however, did not follow this stoichiometry (it was much less than unity; Yeh et al., 2019). This fact, in contrast, is consistent with a disproportionation reaction of FMN peroxide, in which one O atom goes to benzoate and the other ends up as water. To search for clues, we re-examined the active-site geometry of Hmo and its Y128F mutant liganded with substrates (mandelate or lactate) or products (benzoylformate or pyruvate). The redox-active center C4α=N5 of isoalloxazine is surrounded by a constellation of active-site residues [Val78 and Ala79 at the bottom and Tyr(Phe)128 and His252 at the top], where it is accessible only from the upper front side. The reaction center is sealed to form a narrow and low-dielectric milieu suitable for hydride transfer/electron tunneling when a substrate and redox-active center C4α=N5 approach each other. Interestingly, the substrate pair mandelate/benzoyl­formate fits better than the alternative pair lactate/pyruvate because of the bulky phenyl group in the former [Figs. 3(f) and 3(g)]. The redox chamber in the Y128F mutant, on the other hand, is not as tight as that in the wild type owing to the lack of the p-OH group [Fig. 3(h)]. This flaw is exacerbated when lactate (with a smaller methyl group) is used [Figs. 3(f) and 3(g)]. In this context, O₂ is relatively accessible to FMN_red via a temporal space/tunnel to form C4α-OO⁻-FMN before the release of the α-ketoacid (the non-ping-pong mechanism). Meanwhile, the pro-R α-ketoacid is accessible by sp³ N5 to form C4α-COOH-N5-oxyalkylate-FMN_red. Upon decarboxyl­ation, a C4α-COOH-N5-aloxyl-FMN_red C′α carb­anion results. The C′α carbanion that intramolecularly attacks the distal O atom of C4α-OOH then leads to heterolytic cleavage of the peroxide scissile bond. Upon return of the C′α oxyanion, benzoate and H₂O are formed in concert with the regeneration of FMN_ox (Fig. 5).

Figure 5 The proposed mechanisms of oxidative decarboxylation catalyzed by the Y128F mutant.

The peroxide anion radical:FMN semiquinone caged pair is likely to proceed through a single-electron transfer from FMN_red to O₂ in a given flavoenzyme, where the reactivity depends on the active-site polarity ensemble of factors including bound water, charge distribution, hydrogen bonds, van der Waal forces etc. (Fagan & Palfey, 2010). The Y128C mutant (in which the bulky phenyl group is replaced by a sulfhydryl group) that can transform (S)-mandelate to benzoate was used to assess the extent of active-site perturbation. The structure of the Y128C mutant crystallized and soaked with (S)-mandelate revealed that the sulfhydryl (SH) group of the Y128C mutant has been oxidized to a sulfenyl group (S-OH), in contrast to the other sulfhydryl groups, which are not changed [Fig. 3(i)]. This result indicates that the sulfhydryl group of the Y128C mutant is relatively accessible and sensitive to the local unregulated reactive oxygen species (ROS; Chaiyen et al., 2012).