Selectivity of substrate binding and ionization of 2-methyl-3-hydroxypyridine-5-carboxylic acid oxygenase

Abstract

2-Methyl-3-hydroxypyridine-5-carboxylic acid (MHPC) oxygenase (EC 1.14.12.4) from Pseudomonas sp. MA-1 is a flavin-dependent monooxygenase that catalyzes a hydroxylation and aromatic ring cleavage reaction. The functional roles of two residues, Tyr223 and Tyr82, located ~ 5 Å away from MHPC, were characterized using site-directed mutagenesis, along with ligand binding, product analysis and transient kinetic experiments. Mutation of Tyr223 resulted in enzyme variants that were impaired in their hydroxylation activity and had K_d values for substrate binding 5–10-fold greater than the wild-type enzyme. Because this residue is adjacent to the water molecule that is located next to the 3-hydroxy group of MHPC, the results indicate that the interaction between Tyr223, H₂O and the 3-hydroxyl group of MHPC are important for substrate binding and hydroxylation. By contrast, the K_d for substrate binding of Tyr82His and Tyr82Phe variants were similar to that of the wild-type enzyme. However, only ~ 40–50% of the substrate was hydroxylated in the reactions of both variants, whereas most of the substrate was hydroxylated in the wild-type enzyme reaction. In free solution, MHPC or 5-hydroxynicotinic acid exists in a mixture of monoanionic and tripolar ionic forms, whereas only the tripolar ionic form binds to the wild-type enzyme. The binding of tripolar ionic MHPC would allow efficient hydroxylation through an electrophilic aromatic substitution mechanism. For the Tyr82His and Tyr82Phe variants, both forms of substrates can bind to the enzymes, indicating that the mutation at Tyr82 abolished the selectivity of the enzyme towards the tripolar ionic form. Transient kinetic studies indicated that the hydroxylation rate constants of both Tyr82 variants are approximately two- to 2.5-fold higher than that of the wild-type enzyme. Altogether, our findings suggest that Tyr82 is important for the binding selectivity of MHPC oxygenase towards the tripolar ionic species, whereas the interaction between Tyr223 and the substrate is important for ensuring hydroxylation. These results highlight how the active site of a flavoenzyme is able to deal with the presence of multiple forms of a substrate in solution and ensure efficient hydroxylation.

Abbreviations

3HB

3-hydroxybenzoate

3HB6H

3-hydroxybenzoate 6-hydroxylase

5HN

5-hydroxynicotinic acid

5HP2M

5-hydroxypicolinic acid 2-monooxygenase

5PAO

5-pyridoxic acid oxygenase

6HN3M

6-hydroxynicotinate 3-monooxygenase

DHPH

2,6-dihydroxypyridine 3-hydroxylase

FAMS

α-(N-formylaminomethylene)succinic acid

G6PD

glucose 6-phosphate dehydrogenase

G6P

glucose 6-phosphate

HOD

1(H)-3-hydroxy-4-oxoquinaldine 2,4-dioxygenase

HspB

6-hydroxy-3-succinoyl-pyridine 3-monooxygenase

IPTG

isopropyl thio-β-d-galactoside

mHBH

3-hydroxybenzoate 4-hydroxylase

MHPC

2-methyl-3-hydroxypyridine-5-carboxylic acid

MHPCO

2-methyl-3-hydroxypyridine-5-carboxylic acid oxygenase

MSD

mass selective detector

pHBH

p-hydroxybenzoate hydroxylase

PHHY

phenol hydroxylase

pOHB

p-hydroxybenzoate

QDO

1(H)-3-hydroxy-4-oxoquinoline 2,4-dioxygenase

Introduction

Hydroxylation and oxygenation are important biological reactions in the degradation of aromatic compounds by aerobic microorganisms. These reactions convert stable aromatic compounds derived from lignin, pesticides or herbicides into metabolites that can be assimilated by soil microbes [1]. Generally, the hydroxylation reaction to convert phenolic derivatives to catecholic derivatives precedes the ring-cleavage reaction that generates aliphatic compounds [2]. Although the hydroxylation of phenolic derivatives can be catalyzed by many flavin-dependent enzymes [3-5], the ring-cleavage reactions are usually performed by enzymes with non-heme iron cofactors [6]. Some exceptions to this are the enzymes: 2-methyl-3-hydroxypyridine-5-carboxylic acid (MHPC) oxygenase (MHPCO) (EC 1.14.12.4), 5-pyridoxic acid oxygenase (5PAO) (EC 1.14.12.5) [7], 1(H)-3-hydroxy-4-oxoquinoline 2,4-dioxygenase (QDO) (EC 1.13.11.47) and 1(H)-3-hydroxy-4-oxoquinaldine 2,4-dioxygenase (HOD) (EC 1.13.11.48) [8, 9], in which the aromatic ring-cleavage reaction occurs without involvement of a metal cofactor. HOD catalyzes the conversion of 1(H)-3-hydroxy-4-oxoquinaldine to N-acetyl-anthranilic acid, whereas QDO catalyzes the conversion of 1(H)-3-hydroxy-4-oxoquinoline to N-formyl-anthranilic acid and both enzymes do not have any bound cofactors [10-12]. MHPCO and 5PAO are also metal-free enzymes. They are flavoenzyme hydroxylases that catalyze a ring-cleavage reaction in addition to a flavin-dependent hydroxylation [7, 13, 14].

MHPCO is an FAD-containing enzyme initially discovered as a part of the vitamin B₆ degradation pathway of the soil bacterium, Pseudomomas sp. MA-1 [15, 16], and subsequently also found in the nitrogen fixing bacterium, Mesorhizobium loti MAFF303099 [17, 18]. The enzyme is among a few flavin-dependent monooxygenases that utilize hydroxypyridine compounds as substrates. Other hydroxypyridine–utilizing flavoenzymes include 5PAO, 6-hydroxy-3-succinoyl-pyridine 3-monooxygenases (HspB) (EC 1.14.13.163) [19], 6-hydroxynicotinate 3-monooxygenase (6HN3M) (EC 1.14.13.114) [20], 2,6-dihydroxypyridine 3-monooxygenase (DHPH) (EC 1.14.13.10) [21] and the recently reported 5-hydroxypicolinic acid 2-monooxygenase (5HP2M) [22]. These enzyme were found in the degradation pathways of pyridine derivatives such as nicotine in soil bacteria [23-25].

The overall reaction of MHPCO is the conversion of MHPC into an aliphatic compound, α-(N-acetylaminomethylene)succinic acid (Scheme 1). The enzyme can also use 5-hydroxynicotinic acid (5HN) as a substrate to generate α-(N-formylaminomethylene)succinic acid (FAMS) (Scheme 1). Previous studies indicated that MHPCO forms a reactive C4a-hydroperoxyflavin intermediate that subsequently hydroxylates MHPC or 5HN to yield hydroxylated MHPC or 5HN. The hydroxylated product presumably undergoes hydration and re-arrangement to form the ring-opening product [7, 13, 14, 26-28]. Studies using flavin analogs have shown that the hydroxylation mechanism occurs via an electrophilic aromatic substitution mechanism [29]. The ability of the enzyme to convert 5HN into FAMS instead of 5,6-dihydroxynicotinic acid (5,6-HNA) also suggests that the ring-cleavage reaction is an activity intrinsic to MHPCO because, otherwise, the hydroxylated 5HN would go through a re-aromatization process to give the aromatic product, 5,6-HNA, as occurs in the reactions of other flavoenzyme hydroxylases (Scheme 2) [13]. The mechanism by which the ring-cleavage reaction occurs is not well understood, although a few models have been proposed [7, 18, 27, 28]. MHPCO from Pseudomonas MA-1 and M. loti were cloned and expressed in Escherichia coli and these two enzymes share ~ 95% sequence identity [17]. The enzyme shares ~ 20–30% sequence identity with other enzymes in the Class A flavin-dependent monooxygenases [4, 30]. The X-ray structure of the MHPCO:MHPC complex indicates that the active site contains seven water molecules surrounding MHPC. These water molecules mediate the interactions between MHPC and Tyr223, Tyr82 and Arg211 (Fig. 1), and are suggested to be important for the catalytic reaction [7, 18]. Tyr223 is located close to the 3-O(H) group of MHPC, Tyr82 is close to the MHPC pyridinium nitrogen and Arg211 is close to the carboxylate group of MHPC (Fig. 1). However, the functional roles of these residues have never been investigated.

Scheme 1

Catalytic reaction of MHPCO.

Scheme 2

An alternative path for a hydroxylated 5HN acid.

Figure 1

The active site of the wtMHPC:MHPC complex (Protein Data Bank code: 3GMC). Residues surrounding the MHPC (white) binding site, Tyr223, Tyr82 and Arg211, are shown in green. FAD is shown in yellow. Hydrogen bonding between the substrate and the side chain of two tyrosines is indicated by yellow dash lines. The hydrogen-bonding distance is within 3 Å.

MHPCO is also interesting for its ability to selectively bind a tripolar ionic specie of MHPC. The hydroxypyridine structure of MHPC can exist in an equilibrium between a tripolar ionic and monoanionic species at neutral pH (Scheme 3) [31]. The results obtained from spectroscopic studies of the enzyme-bound ligand and studies with the substrate analog, N-methyl-5-hydroxynicotinic acid, indicate that the wild-type enzyme only binds to the tripolar ionic form of MHPC [31]. Binding of the tripolar ionic species allows the -OH group of MHPC to be in the deprotonated form that would facilitate an electrophilic aromatic substitution mechanism [7]. Therefore, the active site residues must contain catalytic features that facilitate the selective binding of the tripolar ionic MHPC.

Scheme 3

Two forms of hydroxypyridine substrate exist at neutral pH.

In the present study, the functional roles of Tyr82 and Tyr223 in the MHPCO reaction were investigated by exploring the catalytic properties of their variants. Multiple characteristics were evaluated for the different variants, including ligand-binding affinity, efficiency of the ring-cleavage reaction, selectivity for the ligand, and the rate constant of hydroxylation. The results obtained show that the interaction of Tyr223 with MHPC is crucial for substrate binding and hydroxylation, whereas Tyr82 is important for the selective binding of the tripolar ionic form of the ligand, which is necessary for efficient hydroxylation.

Results

Expression and isolation of MHPCO variants

MHPCO variants were expressed in E. coli BL21(DE3) using an auto-induction system. The auto-induction media allows cells to reach higher densities, resulting in a greater protein yield than the conventional isopropyl thio-β-d-galactoside (IPTG) induction method [32]. Variants that were soluble and overexpressed well were purified to homogeneity. Typically, growth in the auto-induction medium (4 L) resulted in ~ 50 g of cell paste and 350–370 mg of purified MHPCO. Three out of six variants of Tyr223 (Tyr223Arg, Tyr223Lys and Tyr223Asp) and three out of six variants of Tyr82 (Tyr82Glu, Tyr82Lys and Tyr82Asp) were insoluble at all conditions (auto-induction and IPTG-induction media at 16 and 25 °C). Therefore, only the soluble variants Tyr223Phe, Tyr223Glu, Tyr223His, Tyr223Thr, Tyr82Phe, Tyr82His and Tyr82Arg (see 4) were studied further.

Using the method described in the Experimental procedures, the extinction coefficients of the MHPCO variants were determined. All values are in the same range as the wild-type MHPCO (Table 1), indicating that the variants bind to FAD in a similar fashion as the wild-type enzyme. The results also indicate that all variants have FAD fully bound to their active sites.

Table 1. Summary of the extinction coefficients, dissociation constants of 5HN binding to MHPCO variants, and the effects of 5HN on the reductive half-reaction and flavin conformation during substrate binding

Enzyme	Extinction coefficient (mm^−1·cm^−1)	Kd^5HN (μm) at 25 °C	Specific activity in the absence of 5HN (μmol·min^−1·mg^−1)	Specific activity in the presence of 5HN (μmol·min^−1·mg^−1)	Stimulation fold	Flavin Conformation during substrate binding
WT	11.62 ± 0.31 (455 nm)	24 ± 3	0.047 ± 0.003	2.04 ± 0.09	43.3	Out
Y82F	11.89 ± 0.08 (455 nm)	19 ± 2	0.13 ± 0.03	0.60 ± 0.03	4.5	Out
Y82R	11.29 ± 0.12 (455 nm)	16 ± 3	0.07 ± 0.02	0.099 ± 0.009	1.0	In
Y82H	12.21 ± 0.21 (456 nm)	32 ± 5	0.042 ± 0.003	0.40 ± 0.04	9.3	Out
Y223F	11.45 ± 0.03 (458 nm)	165 ± 23	0.035 ± 0.004	0.37 ± 0.02	9.5	Out
Y223E	12.12 ± 0.09 (455 nm)	310 ± 16	0.31 ± 0.06	0.27 ± 0.05	–	Out
Y223H	11.98 ± 0.02 (455 nm)	86 ± 7	0.133 ± 0.007	0.63 ± 0.03	5.2	Out
Y223T	12.00 ± 0.05 (455 nm)	76 ± 10	0.151 ± 0.005	0.74 ± 0.02	5.1	Out

Binding of 5HN to MHPCO variants

Dissociation constants for 5HN binding to the MHPCO variants were determined by changes in the spectrum of enzyme-bound FAD upon 5HN binding (see 4). 5HN was used in the binding experiment instead of MHPC because 5HN was further used to explore the ability of the MHPCO variants to catalyze the ring-cleavage reaction. The reaction of MHPC and 5HN catalyzed by MHPCO occurs in a similar fashion. However, 5HN has the potential to form an aromatic product after rearomatization as a result of the lack of a 2-CH₃-substituent (Scheme 2) [7, 13, 26]. A summary of the obtained K_d values (Table 1) indicated that the values for all of the Tyr223 variants (75–300 μm) were ~ 5–10-fold larger than that of the wild-type enzyme, suggesting that this residue is important for substrate binding to MHPCO. By contrast, mutations at the Tyr82 position have much less of an effect on the K_d of 5HN binding, giving values in the range 20–30 μm, comparable to that of the wild-type enzyme.

It is known that Class A single-component flavoenzymes exhibit conformational dynamics in which the flavin can alternate between the ‘in’ and ‘out’ conformations, which are classified according to the position of the FAD isoalloxazine moiety. The ‘out’ conformation in the prototype enzyme, p-hydroxybenzoate hydroxylase (PHBH) (EC 1.14.13.2) is defined as the position in which the isoalloxazine moves out of the active site. This position is assumed to be relevant for the substrate-bound form because the isoalloxazine ring is ready for receiving a hydride equivalent from NAD(P)H that binds outside of the substrate pocket [33, 34]. The ‘in’ conformation is the conformation suitable for the reaction of reduced flavin and oxygen during the oxidative half-reaction because formation of C4a-hydroperoxyflavin can be protected from solvent. Previous studies of PHBH have shown a correlation between the conformational dynamics of PHBH and the enzyme difference spectra upon the binding of p-hydroxybenzoate (pOHB) or other substrate analogs [35-37]. Based on this analogy, the results of the binding studies of 5HN to the different MHPCO variants showed that most of the changes in the MHPCO spectra were similar to the changes observed for the flavin ‘out’ conformation of PHBH (Fig. 2). The only exception to this was the Tyr82Arg variant, in which the difference spectra showed a similar pattern to the flavin ‘in’ conformation in PHBH.

Figure 2

Difference spectra of Tyr82 (A) and Tyr223 variants (B) upon 5HN binding. Absorbance changes upon addition of 5HN to the enzymes indicate the interaction of 5HN with enzyme-bound FAD. Based on analogy with wild-type MHPCO, the absorption characteristics of wild-type, Tyr82Phe, Tyr82His and the other Tyr223 variants are consistent with a change in the Flavin conformation from the ‘in’ to the ‘out’ conformation upon 5HN binding.

Stimulation of the flavin reduction upon substrate binding

One of the unique properties of Class A single-component flavin-dependent monooxygenases is the stimulation of flavin reduction by NADH (reductive half-reaction) upon substrate binding [4, 34, 38]. Previous studies of MHPCO show that both MHPC and 5HN can enhance the enzyme reduction rate by ~ 1600- and 1250-fold, respectively. This substrate stimulation effect can be assessed using steady-state assays in which the rates of NADH oxidation in the presence and absence of ligand are compared [7]. The steady-state assay is much more economical in enzyme consumption than stopped-flow experiments for evaluating the effects of ligand binding on each variant. The results shown in Table 1 indicate that, for most of the Tyr82 and Tyr223 variants under the assay conditions employed, the binding of 5HN can stimulate rate of flavin reduction (NADH oxidation) by ~ 5–10-fold. These observed stimulation effects were all less than the effect observed for the wild-type enzyme.

Interestingly, the presence of 5HN in the reduction assay for the Tyr223Glu variant could not enhance the NADH oxidation activity (Table 1). Although the K_d value for 5HN binding to Tyr223Glu was higher than those of the other variants (Table 1), the conditions employed (200 μm 5HN; 1.2 μm enzyme; K_d value of 300 μm) should still give at least 0.48 μm of the binary complex of MHPCO:5HN. These data indicate that, although 5HN can bind to Tyr223Glu, the protein–ligand interaction necessary for triggering the change in the flavin conformation to the conformation that can enhance the flavin reduction rate is disrupted.

Identification of ring-opening ability by the MHPCO variants

To investigate whether the MHPCO variants can catalyze ring-cleavage reactions, 5HN was used as a substrate for MHPCO. Unlike MHPC, the hydroxylated 5HN that lacks the -CH₃ group at the C2-position should be able to lose the 2-H atom, resulting in the formation of 5,6-dihydroxynicotinic acid (Scheme 2). Therefore, variants that still possess hydroxylation activity but lack ring-cleavage ability may generate 5,6-dihydroxynicotinic acid as a product. In this experiment, multiple turnover reactions of MHPCO with 5HN as a substrate were carried out with a constant supply of NADH produced by a glucose 6-phosphate dehydrogenase (G6PD) and glucose 6-phosphate (G6P) system, such that the spectral changes were not influenced in any way by the change in NADH and NAD⁺ absorption. Difference spectra experiments were set up such that all reagents in the sample and reference cuvettes were the same, except that 5HN was placed in the sample cuvette. The changes observed for the spectra were thus a result of the change of 5HN into the oxygenated product (see 4). Because 5HN has absorption peaks at 260, 285 and 323 nm, whereas the ring-cleaved product, FAMS, has a λ_max at 260 nm, the decrease of absorption at wavelengths > 300 nm and the increase of absorption at 260 nm indicates the conversion of 5HN to the ring-cleaved product [13].

All four variants of Tyr223 (Tyr223Phe, Tyr223Glu, Tyr223His and Tyr223Thr) could not catalyze the ring-cleavage reaction or hydroxylation of 5HN. Spectra of the reactions recorded at the reaction times between 0 min and 180 min were all identical (data not shown). HPLC analysis was also carried out and could not detect either hydroxylated 5HN or the ring-cleavage product. These variants could all bind to 5HN (although with higher K_d values than the wild-type enzyme) (Table 1). Therefore, the disruption of the interaction between Tyr223 and 5HN significantly affects oxygenation of the substrate but not the binding. These results also imply that proper interaction of Tyr223 and 5HN is crucial for hydroxylation to occur, possibly by allowing 5HN or MHPC to bind to a tripolar ionic species so that the substrate can readily undergo electrophilic aromatic substitution.

For Tyr82 variants (Tyr82Phe, Tyr82Arg and Tyr82His), HPLC analysis and UV-visible spectroscopy indicated that only Tyr82Phe and Tyr82His could convert 5HN to FAMS (Fig. 3), whereas Tyr82Arg could not. These results imply that the interaction between the N-atom of 5HN and the -OH of Tyr82 is not specifically required for hydroxylation because replacement of Tyr82 with Phe still resulted in hydroxylation of 5HN. However, Tyr82 may be important for selective binding of the substrate in its tripolar ionic form. Therefore, the Tyr82Phe and Tyr82His variants were investigated further with regard to their ligand-binding selectivity.

Figure 3

Product conversion by Tyr82 variants that can undergo a ring-opening reaction. (A) Spectra of multiple turnover reactions of Tyr82Phe. (B) Spectra of multiple turnover reactions of Tyr82His. Difference spectra in (A) and (B) were recorded from the reactions in which all reagents in the sample and reference cells were the same except that 5HN was only present in the sample cell (see 4). The absorption increase at 260 nm resulted from the formation of the aliphatic FAMS product. The results indicate that only Tyr82Phe and Tyr82His can catalyze ring-cleavage reactions, whereas the other variants cannot catalyze oxygenation reactions because the spectra of all the reactions remained unchanged over time (data not shown).

pK_a determination of 5HN and spectra of various ionic species of 5HN

To identify the 5HN species that binds to MHPCO, spectra of various ionic forms of 5HN were recorded at different pH in buffers with pH in the range 0.5–12.0 (Fig. S1). At a pH below 1.5, 5HN is in the cationic form in which N-1 is protonated. The cationic 5HN has a maximum absorption at 296 nm. When the pH was increased to pH 3, the maximum absorption wavelength was shifted to 292 nm, presumably because of the formation of dipolar 5HN (Scheme 4). At neutral pH, 5HN is in an equilibrium between the tripolar and monoanionic species (Scheme 3). The mixture of these two species has maximum absorption at 285 and 323 nm (Fig. S1). At a pH above 9, all of the 5HN was present as a dianionic species with a maximum absorption at 315 nm. Based on these absorbance changes as a function of pH, three pK_a values, 1.87, 4.59 and 8.28, associated with each ionization state were identified (Fig. S2).

Scheme 4

Different 5HN species exist at different pH ranges. 5HN form A = cationic ion. 5HN form B = dipolar ion. 5HN form C = mixture between tripolar ion and monoanionic ion. 5HN form D = dianionic ion.

Because 5HN at neutral pH exists in equilibrium between the two forms (tripolar ion and the monoanionic species), we decreased the polarity of the system by adding isopropanol (polarity index of 3.9) into the 5HN solution so that all the 5HN was converted into the monoanionic species (Fig. S3). The 5HN monoanionic spectrum obtained was subtracted from the mixture spectra of monoanionic and tripolar species to identify the spectrum of the tripolar 5HN (Fig. 4) (see 4). The tripolar ion spectrum obtained has absorption peaks at 260 and 320 nm, which is similar to the known spectrum of N-methyl-5-hydroxynicotinic acid that mainly exists in the tripolar form at neutral pH [31].

Figure 4

Absorption characteristics of MHPC (A) and 5HN (B) in different media at pH 7. At neutral pH, the substrate exists in equilibrium between both tripolar and monoanionic ion forms (Scheme 2). Isopropanol (final of 80%) was added into an aqueous solution of MHPC and 5HN to push the equilibrium towards the monoanionic species. A spectrum of the tripolar form was derived by subtraction of the spectrum of the monoanionic species that contributed to the spectrum of the mixture.

Identification of the form of 5HN and MHPC bound to the Tyr82 and Tyr223 variants

The specific species of 5HN bound to the MHPCO variants and wild-type enzyme was identified based on the difference spectra (see 4). 5HN (10 μm) was only added into a sample cuvette, whereas an equal concentration of enzyme was added in both sample and reference cuvettes. The enzyme concentration was gradually increased so that the mixture of 5HN in the sample cell shifted towards more being in the enzyme-bound state. With the highest enzyme concentration used (120 μm), the 5HN in the sample cuvette should mainly exist in the enzyme-bound form (82% for the wild-type enzyme, 85% for the Tyr82Phe variant and 78% for the Tyr82His variant).

The results for the wild-type enzyme clearly showed an increase in absorption at 323 nm upon binding, consistent with the change of 5HN from a mixture toward the tripolar species. These data agree with the previous report of binding of MHPC to wild-type MHPCO [31]. However, for the Tyr82Phe and Tyr82His variants, this shift towards the tripolar species upon increasing the enzyme concentration was not as clear as in the wild-type enzyme. Absorbance at 323 nm (Fig. 4) was used to estimate the amount of tripolar 5HN bound to the enzyme. For the binding of 5HN (10 μm) and wild-type MHPCO (25 μm) (Fig. 5A), the concentration of enzyme-bound MHPCO was 4.58 μm. The absorbance value of this binding mixture at 323 nm corresponded to ~ 4.60 μm tripolar ionic 5HN, indicating that all the wild-type MHPCO-bound 5HN is in the tripolar ionic form. For the binding of the Tyr82 variants, the enzyme concentration was increased to 120 μm to obtain 8.54 μm of the Tyr82Phe-bound 5HN and 7.81 μm for the Tyr82His-bound 5HN (Fig. S4). Based on their absorbance characteristics, the concentration of tripolar ionic 5HN bound to the Tyr82Phe and Tyr82His variants were estimated as 1.91 μm (23%) and 4.4 μm (56%), respectively. These data indicate that Tyr82 is important for promoting the binding of 5HN in the tripolar form.

Figure 5

Identification of the specific form of 5HN or MHPC that binds to the wild-type MHPCO and Tyr82Phe variant. Green lines represent free 5HN or MHPC in neutral solution (sodium phosphate pH 7.0, whereas red lines represent enzyme-bound 5HN or MHPC in the presence of 25 μm enzyme: (A) wild-type MHPCO-bound 5HN; (B) wild-type MHPCO-bound MHPC; (C) Tyr82Phe-bound 5HN; and (D) Tyr82Phe-bound MHPC. Increasing concentrations of wild-type enzyme in solution resulted in an increase in absorption at wavelengths > 320 nm, indicating a shift towards the tripolar ionic species. By contrast, increasing the Tyr82Phe concentration resulted in greater absorption at 290 nm, indicating a shift towards the monoanionic species.

Because the absorption peak of the monoanionic form of 5HN is ~ 290 nm, which is in the region where interference from intrinsic protein absorption occurs, similar experiments with MHPC were carried out. Results from the MHPC binding agree well with the 5HN binding results (Fig. 5B,D) in that Tyr82Phe and Tyr82His did not clearly shift the bound MHPC toward tripolar ion characteristics as seen in the wild-type enzyme (Table 2).

Table 2. Summary of the percentage of product formation, percentage of coupling path, percentage of enzyme-bound tripolar ionic MHPC and hydroxylation rate constants of wild-type MHPCO, and the Tyr82Phe and Tyr82His variants

Enzyme	MHPC
	Percentage conversiona	Percentage coupling pathb	Percentage tripolar MHPCc (%)	kOH(s^−1)b
Wild-type MHPCO	99 ± 6	100% (uncoupling was not detected)	100	0.9
Tyr82Phe	49 ± 5	25% (A400)	50	2.5
Tyr82His	57 ± 7	30% (A400)	32	1.9

• ^a Estimated from the amount of substrate consumed.
• ^b Calculated from the amplitude changes of coupling and uncoupling paths of stopped-flow data.
• ^c Estimated from the absorbance at 323 nm of the tripolar ionic MHPC.

Similar experiments with Tyr223 variants were also performed, although meaningful results could not be obtained because the variants have high K_d values for 5HN (Table 1) and only formed minimal levels of the enzyme–5HN complex (data not shown).

Percentage of product formation

Single turnover reactions of the MHPCO variants and wild-type enzymes were performed by mixing the substrate bound reduced enzyme with the oxygenated buffer (see 4). Enzyme is the limiting reagent under these experimental conditions. The results summarized in Table 2 indicate that single turnover reactions of wild-type enzyme with MHPC and 5HN gave 99 ± 6% and 100 ± 11% product formation, respectively (the calculation was based on the substrate consumed), whereas the reactions of Tyr82Phe and Tyr82His gave lower percentages of product formation (Table 2). These data indicate that both variants can convert 5HN and MHPC into aliphatic ring-cleaved products but with lower efficiency than the wild-type enzyme.

Transient kinetics of the reaction of reduced wild-type MHPCO:MHPC complex with oxygen

To evaluate the role of Tyr82 in the oxygenation reaction of MHPCO, reactions of wild-type and Tyr82Phe and Tyr82His variants were investigated and compared. The kinetics of wild-type MHPCO and MHPC was investigated by mixing the reduced enzyme in the presence of MHPC with various oxygen concentrations and changes in A₄₀₅ and A₄₅₆ were monitored (see 4). The results indicated that the kinetic traces consist of two phases (Fig. S5), similar to previously reported data [13, 29]. The first phase is characterized by an increase in absorbance at 405 nm and the observed rate constants for this phase were linearly dependent on the oxygen concentration, consistent with a bimolecular rate constant of 5.1 ± 0.1 × 10⁴ m⁻¹·s⁻¹. The second phase is characterized by a decrease in absorbance at 405 nm and a large increase of absorbance at 456 nm. The observed rate constant of this first phase (0.9 s⁻¹) is independent of the oxygen concentration. These results are interpreted according to the model described in Scheme 5 in which the first phase is the formation of C4a-hydroperoxyflavin and the second phase is the hydroxylation step resulting in C4a-hydroxyflavin. As a result of the fast decay of the C4a-hydroxyflavin, this intermediate could not be detected [13, 29].

Scheme 5

The reaction of MHPC bound reduced wild-type MHPCO with oxygen.

The reaction was also monitored by fluorescence detection using excitation wavelengths of 400 and 456 nm and emission of wavelengths > 495 nm. Traces that resulted from both excitations were biphasic in which the first phase had insignificant changes in fluorescence. The second phase showed a large increase in fluorescence, with a rate constant of 1 s⁻¹. These data are in agreement with the absorption data and can be explained in accordance with a model in which the first phase represents the formation of C4a-hydroperoxyflavin and the second phase is the hydroxylation step to form C4a-hydroxyflavin. Because the dehydration of C4a-hydroxyflavin to yield the oxidized enzyme is very fast, the spectroscopic signal of this intermediate could not be detected and the fluorescence signal was only from the final oxidized enzyme. All of the wild-type enzyme data were very similar to the results previously reported.

The reaction of reduced Tyr82His:MHPC complex with oxygen in the presence of sodium azide

To better resolve the kinetics of the individual intermediates, sodium azide was added to the reaction to slow down the rate constant of C4a-hydroxyflavin dehydration. Sodium azide is commonly used in many transient kinetic studies of flavoenzyme monooxygenase to slow down the rate of C4a-hydroxyflavin dehydration without causing other interference in the reaction [13, 29, 39]. The kinetics of the reaction of reduced Tyr82His:MHPC was investigated using stopped-flow absorbance (A₄₀₀ and A₄₅₆) and fluorescence (excitation wavelengths of 380 and 470 nm) as described in the Experimental procedures. The kinetics of the absorbance increase at 400 nm is biphasic with the observed rate constants of both phases being linearly dependent on the oxygen concentration. The first phase had a bimolecular rate constant of 3.6 ± 0.3 × 10⁴ m⁻¹·s⁻¹, whereas the second phase was 7.6 ± 0.3 × 10² m⁻¹·s⁻¹. The amplitude of the first phase was ~ 30%, whereas the second phase was ~ 70% of the total absorbance increase at 400 nm. For kinetic traces at 456 nm, the reaction was analyzed according to three phases (data not shown). The first phase was a lag phase without significant absorption change while the second phase showed an absorption increase of ~ 40% of the total amplitude change. The observed rate constant of this phase was constant at around 0.5 s⁻¹. The third phase showed an absorbance increase that comprised ~ 60% of the total amplitude change. The observed rate constant of this phase was dependent on the oxygen concentration with a bimolecular rate constant of 7.6 ± 0.3 × 10² m⁻¹·s⁻¹.

The kinetics of the reaction of the complex of reduced Tyr82His:MHPC with oxygen was detected by fluorescence excitation at 380 nm and found to be triphasic. The first phase was a lag phase without significant fluorescence signal changes. The second phase was a large fluorescence increase with an observed rate constant of 1.9 s⁻¹, whereas the third phase was seen as a fluorescence decrease with an observed rate constant of ~ 0.5 s⁻¹ (Fig. 6). In analogy to the wild-type enzyme [13], and in lieu of the results of the absorbance changes noted above, the first lag phase corresponded to the formation of C4a-hydroperoxyflavin, whereas the second phase corresponded to the hydroxylation step to form C4a-hydroxyflavin, which is highly fluorescent. The third phase was the H₂O elimination from C4a-hydroxyflavin to form the oxidized enzyme. For the kinetic trace from the excitation at 470 nm that specifically monitored the formation of the oxidized enzyme, the data only showed a large fluorescence increase with an observed rate constant of 0.5 s⁻¹ that is independent of oxygen concentration (Fig. 6). This phase was synchronized with the third phase of the reaction excited at 380 nm and the third phase of the absorption change at 456 nm. We also noted a slower phase (0.09 s⁻¹) with a small fluorescence increase. This phase may be the result of a small fraction of less active enzyme or to the dissociation of FAD from the enzyme active site.

Figure 6

Oxidative half-reaction of Tyr82His observed on a stopped-flow spectrophotometer in both absorption and fluorescence modes. Kinetic traces of the reaction from low to high oxygen concentrations are shown from the right to left traces. The experiment was carried out in the presence of 50 mm sodium azide to distinguish the hydroxylation step.

Altogether, the data indicate that the re-oxidation of Tyr82His occurs via two pathways that result from two fractions of enzymes (Scheme 6). The upper pathway (coupling path) forms C4a-hydroperoxyflavin (bimolecular rate constant of 3.6 ± 0.3 × 10⁴ m^−1·s⁻¹), which further results in C4a-hydroxyflavin (hydroxylation rate constant of 1.9 s⁻¹) before formation of the oxidized enzyme with the dehydration rate constant of 0.5 s⁻¹. The other pathway (uncoupling path) is the direct oxidation of reduced enzyme to form oxidized FAD and H₂O₂ with a bimolecular rate constant of 7.6 ± 0.3 × 10² m^−1·s⁻¹. Based changes in A₄₀₀ and A₄₅₆, ~ 30–40% of the enzyme reacts via the coupling path, whereas 60–70% reacts via the uncoupling path.

Scheme 6

The reoxidation of the Ty82His:MHPC complex in the presence of sodium azide.

The reaction of reduced Tyr82Phe:MHPC complex with oxygen in the presence of sodium azide

Similar to the reaction of Tyr82His described above, the kinetics of re-oxidation of reduced Tyr82Phe:MHPC was detected using stopped-flow absorbance (A₄₀₀ and A₄₅₆) and fluorescence (excitation wavelengths of 390 and 470 nm) modes in the presence of sodium azide as described in the Experimental procedures. The sodium azide is added to slow down the dehydration step.

The kinetics of absorbance increase at 400 nm showed four phases with observed rate constants of the first and second phases linearly dependent on the oxygen concentration. The first phase has a bimolecular rate constant of 4.1 ± 0.9 × 10⁴ m⁻¹·s⁻¹, whereas the second phase had a rate constant of 2.1 ± 0.2 × 10³ m⁻¹·s⁻¹. The third phase resulted in an absorbance increase at 400 nm at oxygen concentrations of 0.13 and 0.31 mm, whereas the absorbance decrease at 400 nm was observed for oxygen concentrations of 0.61 and 1.03 mm (with the same observed rate constant of 0.5 s⁻¹). The fourth phase had a small absorption increase or decrease with a rate constant of 0.1 s⁻¹. This observation was likely a result of the reduced enzyme reacting with oxygen via coupling and uncoupling pathways, as described for Tyr82His above. The first phase is the formation of C4a-hydroperoxyflavin because this phase showed very little change in A₄₅₀ (data not shown). At higher oxygen concentrations, the rate of C4a-hydroperoxyflavin formation in the coupling path is faster than the reactions at lower oxygen concentrations, resulting in a greater accumulation of the C4a-hydroperoxyflavin intermediate. Therefore, the amplitude change of kinetic traces observed in the third and fourth phases varied according to the oxygen concentration. Kinetic analysis based on absorption data alone cannot unambiguously assign the hydroxylation or dehydration step.

The kinetics of the reaction of the reduced Tyr82Phe:MHPC complex with oxygen was triphasic as detected by fluorescence excitation at 390 nm. The first phase was a lag phase with a very small fluorescence change. The second phase was a large fluorescence increase with an observed rate constant of 2.5 s⁻¹, whereas the third phase was a fluorescence decrease with an observed rate constant of 0.5 s⁻¹ (Fig. 7). By analogy to the wild-type enzyme [13] and the reaction of Tyr82His described above, the second phase observed by fluorescence detection is the hydroxylation step to form C4a-hydroxyflavin, which is highly fluorescent, whereas the third phase is the dehydration of C4a-hydroxyflavin to form the oxidized enzyme. For the kinetic trace from the excitation at 470 nm that specifically monitored formation of the oxidized enzyme, only a large fluorescence increase with an observed rate constant of 0.5 s⁻¹ was observed (Fig. 7). This phase was synchronized with the third phase of the reaction excited at 390 nm. We also noted a slower phase (0.06 s⁻¹) of small fluorescence increase as in the case of Tyr82His. This phase is considered to be a result of dissociation of a small amount of free FAD from the enzyme active site.

Figure 7

Oxidative half-reaction of Tyr82Phe observed on a stopped-flow spectrophotometer in both absorption and fluorescence modes. Kinetic traces of the reaction from low to high oxygen concentrations are shown from the right to left traces. The experiment was carried out in the presence of 50 mm sodium azide to distinguish the hydroxylation step.

Altogether, the data indicate that the re-oxidation of Tyr82Phe occurs via two pathways (Scheme 7). The upper pathway (coupling path) is the reaction of the enzyme to form C4a-hydroperoxyflavin (bimolecular rate constant of 4.1 ± 0.9 × 10⁴ m⁻¹·s⁻¹) that further results in C4a-hydroxyflavin (hydroxylation rate constant of 2.5 s⁻¹) and the oxidized enzyme (the dehydration step of 0.5 s⁻¹). The other pathway is the uncoupling path in which the reduced enzyme directly oxidizes to form H₂O₂ with a bimolecular rate constant of 2.1 ± 0.2 × 10³ m⁻¹·s⁻¹, without formation of C4a-hydroperoxyflavin. Based on changes in A₄₀₀, ~ 25–30% of the enzyme reacts via the coupling path, whereas 70–75% reacts via the uncoupling path.

Scheme 7

The reoxidation of the Tyr82Phe:MHPC complex in the presence of sodium azide.

Correlation between the enzyme-bound substrate form, product formation and the fraction of enzyme forming C4a-hydroperoxyflavin

The data provided in Table 2 clearly indicate that Tyr82His and Tyr82Phe bind both forms of MHPC. Their reactions with oxygen proceed via the uncoupling and coupling paths. The correlation between these data implies that only the enzyme fraction that binds the tripolar ionic MHPC could form C4a-hydroperoxyflavin and catalyze hydroxylation of the substrates. The reduced enzyme fraction that binds the monoanionic form of MHPC was directly oxidized to form H₂O₂. The data also suggest that the functional role of Tyr82 in wild-type MHPCO is a key factor dictating the selective binding of the substrate in the tripolar ionic form.

Discussion

The results of the present study clearly indicate that the residue located ~ 5 Å away from the -OH group of MHPC (Tyr223) is important for the hydroxylation and ring-cleavage activity because mutation of this residue into Phe, His or Glu abolished this activity in the enzyme. Another Tyr residue (Tyr82) that is located ~ 5 Å away from the N-atom of MHPC is important for the selective binding of MHPC in the tripolar ionic form. Unlike the wild-type MHPCO in which the binding is specific for the tripolar ion form of substrate, Tyr82Phe and Tyr82His also bind to the monoanionic forms of MHPC and 5HN (50–75% of the monoanionic form bound).

The interaction between Tyr223 and the -OH of 5HN or MHPC is likely mediated via a water molecule. Based on the X-ray structures of MHPCO: MHPC and MHPCO:5HN complexes, one water molecule is located between the -OH group of MHPC and Tyr223 [18] (J. Kobayashi, H. Yoshida, B. Mikami, H. Hayashi, S. Kamitori & T. Yagi, unpublished results). It is conceivable that this interaction can stabilize the negative charge of the deprotonated hydroxyl group of MHPC or 5HN, which would facilitate the electrophilic aromatic substitution reaction depicted in Scheme 8. This mechanism is considered to be common for Class A flavoprotein monooxygenases [30, 38] that catalyze the incorporation of an oxygen atom from C4a-hydroperoxyflavin into an aromatic substrate. The deprotonation of the substrate -OH group would make the substrate a better nucleophile, subsequently making the oxygen atom transfer from flavin more favorable [7, 38, 40].

Scheme 8

Proposed water-mediated interactions between two tyrosine residues and the MHPC substrate.

The presence of a conserved Tyr close to a hydroxyl group of an aromatic substrate is common among the Class A aromatic hydroxylases. For PHBH, a hydrogen-bond network that is formed by Tyr201, Tyr385, His72 and water molecules [34, 41] controls the deprotonation of pOHB [34, 42]. The Tyr201 in contact with the substrate pOHB is particularly crucial for the deprotonation of pOHB [41]. Based on the structure of m-hydroxybenzoate 4-hydroxylase (mHBH) (EC 1.14.13.2), the deprotonation of m-hydroxybenzoate is assumed to be involved with the neighboring residues Tyr271 and Asp75. For phenol hydroxylase (PHHY) (EC 1.14.13.29), residues similar to those in mHBH, Tyr289 and an Asp54, are located ~ 2.4 A from the hydroxyl group of the substrate. For 3-hydroxybenzoate (3HB) 6-hydroxylase (3HB6H) (EC 1.14.13.24) that catalyzes para-hydroxylation of 3HB, the His213 residue, and not a Tyr, is located close to the hydroxyl group of 3HB [43]. This residue was proposed to be important for the deprotonation process of 3HB [43, 44].

Based on the structures of M. loti MHPCO-MHPC [18] and the recently deposited M. loti MHPCO-5HN (Protein Data Bank code: 4H2Q) (J. Kobayashi, H. Yoshida, B. Mikami, H. Hayashi, S. Kamitori & T. Yagi, unpublished results), MHPCO is different from the aforementioned enzymes in that the enzyme-bound MHPC has no direct contact with any active site residues; all interactions are mediated via water [18]. Mutation of Tyr223 leads to impaired activity of MHPCO and a decrease in ligand-binding affinity (Table 1). The data imply that the hydroxyl group of Tyr223 is important for controlling a hydrogen-bonding interaction between water and substrate. This result is similar to those of other enzymes. The variants of Tyr201Phe in pHBH [45] resulted in the abolishment of the hydroxylation reaction. For PHHY, the Tyr289Phe variant has a greater dissociation constant (~ 15-fold increase, from 1.8 μm to 27 μm) compared to the wild-type enzyme, similar to the results of Tyr223 variants of MHPCO shown in Table 1.

The results shown in Tables 1 and 2 indicate that Tyr82 is crucial for selective binding of MHPC or 5HN (Table S1) in the tripolar ionic form but not critical for ligand-binding affinity or hydroxylation activity. The data provided in Table 1 also indicate that Tyr82 variants have K_d values in a similar range to the wild-type enzyme, implying that this residue is not the key determinant for the ligand-binding affinity. For Tyr82Arg in which its side chain is longer and has a positive charge, the difference spectra indicate that the FAD conformation of this variant is different from the ligand-binding mode of the wild-type enzyme and the other variants (Table 1). Tyr82Phe and Tyr82His variants can still catalyze hydroxylation and ring-cleavage of MHPC with ~ 40–50% efficiency. This value correlated with the tripolar ionic form of MHPC bound on Tyr82Phe and Tyr82His (Figs 5 and S4). The results imply that only the enzyme-bound tripolar ionic form is oxygenated by MHPCO. The binding of the tripolar ionic form of MHPC may also be necessary for the ability of the enzyme to form C4a-hydroperoxyflavin. For Tyr82Phe and Tyr82His, ~ 75% and 70% of enzyme bound ligand could not form C4a-hydroperoxyflavin (Figs 6 and 7). For the enzyme fractions that could form C4a-hydroperoxyflavin, their rate constants for the hydroxylation were 2.5 s⁻¹ and 1.9 s⁻¹ for Tyr82Phe and Tyr82His, respectively, which is ~ 2–2.5-fold faster than that of the wild-type enzyme (0.9 s⁻¹). These results clearly indicate that the hydroxyl group of Tyr82 is not compulsory for the hydroxylation, as is the case for Tyr223. The functional role of Tyr82 is likely to promote the selectivity of substrate binding towards the tripolar ionic form (Fig. 5) [31]. For the Tyr82Phe or Tyr82His variants, the ligand selectivity for the tripolar ionic form is impaired, and both mutants could bind to a mixture of tripolar and monoanionic forms of MHPC and 5HN (Fig. 5).

The functional role of Tyr82 in selection of substrate may also be present in other enzymes that use hydroxypyridine substrates. Besides MHPCO, several enzymes of the Class A flavoenzyme monooxygenases that use hydroxypyridine-substrates are known, such as 5PAO [7, 46], HspB [19], 6HN3M [20], DHPH [21] and 5HP2M [22]. Among these, only the structure of DHPH is available [21]. Because DHP is also a hydroxypyridine derivative, an ionic equilibrium similar to that found for MHPC and 5HN also exists. It will be interesting to determine whether the functional roles of active site residues of DHPH are similar to those of MHPCO.

In conclusion, the present study has clarified the functional roles of Tyr223 and Tyr82 of MHPCO. Tyr223 is crucial for the hydroxylation activity, whereas Tyr82 is needed for the selective binding of the tripolar ionic form of the substrate. The selection of substrate binding in the tripolar ionic form facilitates the oxygen transfer via an electrophilic aromatic substitution mechanism. Our findings highlight the active site features of a flavoenzyme that allow it to overcome the presence of a mixed ionic solution of hydroxypyridine substrates. It would be interesting to determine whether this catalytic mechanism is also prevalent for other hydroxypyridine-utilizing enzymes.

Experimental procedures

Chemicals and reagents

The chemicals and reagents used were of analytical grade. 5HN, G6P, G6PD, NAD⁺ and NADH were purchased from Sigma (St Louis, MO, USA). 5HN has an extinction coefficient of 4.19 mm⁻¹·cm⁻¹ in 0.1 m NaOH (at 315 nm). FAD was purchased from Tokyo Chemical Industry Co., Ltd (Tokyo, Japan). SDS used for determination of the extinction coefficient was purchased from Merck (Merck, Whitehouse Station, NJ, USA). Pfu DNA polymerase and 1X Pfu buffer were obtained from Fermentas (Fermentas, Glen Burnie, MD, USA). MHPC used in the binding experiments was synthesized as described previously [26]. MHPC has an extinction coefficient of 4.4 mm⁻¹·cm⁻¹ at 326 nm [47]. DpnI was obtained from New England Biolabs (New England Biolabs, Beverly, MA, USA).

Site-directed mutagenesis

Site-directed mutagenesis at positions Tyr223 and Tyr82 were carried out by performing PCR using a thermocycler (GeneAmp PCR system, model 2004; Applied Biosystems, Foster City, CA, USA; or MyCycler, Thermal Cycler; Bio-Rad, Hercules, CA, USA). The mutagenic forward and reverse primers were custom synthesized by Bio Basic Inc. (Markham, ON, Canada). All of the mutagenic primers are listed in Tables S2 and S3.

The PCR reaction contained dNTPs, the forward and reverse primers, ~ 10 μg of pET11a-MHPCO (the plasmid encoding wild-type enzyme) as the template, and Pfu DNA polymerase in 1X Pfu buffer. Sterile deionized water was then added into the mixture to give a volume of 50 μL. DpnI was added to the PCR products to digest the original DNA template. After digestion, the reaction mixture containing the nicked PCR products was transformed into XL1-blue E. coli competent cells for amplification. Plasmid extraction and purification were in accordance with the manufacturer's instuctions for the FavorPrep Plasmid Extraction Kit (Favorgen Biotech Corp., Changzhi, Taiwan).

To confirm the accuracy of the site-directed mutagenesis, the purified plasmids were sent for DNA sequencing using the forward primer T7 promoter 5′-TAA TAC GAC TCA CTA TAG GG-3′ and the reverse primer T7 terminator 5′-GCT AGT TAT TGC TCA GCG G-3′. DNA sequencing was carried out by Macrogen Inc. (Seoul, Korea).

Expression of MHPCO variants

An auto-induction medium [32] was used for the overexpression of MHPCO variants. pET11a expression vectors containing MHPCO mutant genes were transformed into BL21(DE3) competent E. coli cells. Bacteria were grown overnight at 37 °C on an LB agar plate containing 100 μg·mL⁻¹ ampicillin. A single colony was then inoculated into 100 mL of ZYM-5052 starter medium containing 50 μg·mL⁻¹ ampicillin. The starter culture was incubated in a shaking incubator at 37 °C overnight. The starter (1%) was inoculated into 650 mL of ZYP-5052 medium and the culture was incubated until D₆₀₀ reached 1.0 (after ~ 2.5–3.0 h). Subsequently, the temperature for incubation was switched to 25 °C, and the cells were incubated for another 14 h before being harvested. The resulting bacterial cell paste was kept at −80 °C until use.

Purification of MHPCO variants

MHPCO variants were overexpressed and purified in accordance with the protocol used for wild-type MHPCO purification [26] with some modifications: the bacterial cell paste was thawed, and lysed by sonication, whereas the temperature of the suspension during sonication was maintained below 15 °C in an ice-bath. Cell debris was separated by centrifugation. Supernatant was collected and defined as the crude extract. Excess FAD (approximate final concentration 0.5 mm) was added into the crude extract before performing ammonium sulfate precipitation to obtain a pellet with 40–80% ammonium sulfate saturation. The resulting precipitated MHPCO was then dialyzed against the buffer before loading onto a Q-sepharose column (pH 7.5) (GE Healthcare, Milwaukee, WI, USA). Protein elution was performed using 25 mm Tris-H₂SO₄ buffer (pH 7.5) containing a gradient of 5 mm to 250 mm NaCl. Eluted fractions were collected and pooled according to A₂₈₀ and A₄₅₀ values. The pooled fractions were concentrated using an Amicon stirred-cell (8050 and 8200; Amicon, Amicon Corp., Danvers, MA, USA) equipped with a YM-10 membrane (Millipore, Millipore, Billerica, MA, USA) (10-kDa cut-off). The concentrated enzyme was passed through a G-25 Sephadex column (GE Healthcare) for buffer exchange, and the stock enzyme solution was aliquoted in 500-μL aliquots and stored at −80 °C until use. DTT (1 mm) and EDTA (0.5 mm) were added to all buffers during purification to maintain the enzyme in active form [26]. Approximately 300 mg of enzyme was obtained from 4 L of bacterial culture (50 g of cell paste).

MHPCO variants from each step of the purification were analyzed for protein content using Bradford Reagent (Bio-Rad) and the purity was checked using SDS/PAGE. BSA was used as a standard protein for constructing a protein calibration curve.

MHPCO activity assay

MHPCO catalyzes the conversion of MHPC or 5HN to yield ring-opening products. Because NADH is used as an external reductant in the reductive half-reaction of MHPCO, the decrease in absorption at 370 nm as a result of NADH consumption can be used as a correlate to enzyme activity. NADH consumption was monitored at 370 nm instead of at its λ_max at 340 nm to avoid interference from 5HN absorption. The extinction coefficient of NADH at 370 nm is 2470 m⁻¹·cm⁻¹. Specific activity values of MHPCO variants can be calculated based on the initial rates obtained. Activity assay data were also used to evaluate whether the presence of 5HN at the enzyme active site can enhance the reductive half-reaction (Table 1). The stimulation of NADH reduction indicates that the variants can bind to 5HN.

Determination of extinction coefficients of MHPCO variants

The purified MHPCO variants were diluted in 100 mm sodium phosphate (pH 7.0) buffer until A₄₅₀ of 0.2 or 0.3 was reached (final volume after dilution equals to 900 μL). A 2% SDS solution (100 μL) was then added, resulting in denaturation of protein and release of free FAD into the solution. Concentrations of the released flavin cofactor were calculated using the known molar extinction coefficient of free FAD of 11.3 mm⁻¹·cm⁻¹ at 450 nm.

Characterization of the ring-opening ability of MHPCO variants using the G6PD system to continuously supply NADH

In this experiment, a G6PD system was used to constantly generate NADH for the MHPCO reaction. A double-beam UV-visible spectrophotometer model UV-2501PC was used to monitor the difference absorption spectra. The reaction mixture containing MHPCO variants (5–10 μm), NAD⁺, G6P and G6PD enzyme was placed in both sample and reference cells. Approximately 62 μm 5HN was added into the sample cell to initiate the reaction. The same volume of buffer was added into the reference cuvette to correct for the dilution factor. If the variant can perform the ring-opening reaction, 5HN will be converted to the corresponding FAMS. Because the spectrum of the product FAMS has an absorption λ_max at 260 nm, whereas 5,6-DHNA should have λ_max at a longer wavelength, the coupled reaction was monitored by a spectrophotometer to monitor the product conversion process.

Thermodynamic binding of a substrate analog, 5HN, to MHPCO variants

5HN stock solutions at four different concentrations (in the range 2.05–15.31 mm) were prepared. Purified MHPCO variants were titrated with 5HN, starting from lower to higher concentrations, and a spectrum was recorded after each titration. Absorbance changes at wavelengths > 320 nm were used to calculate the binding affinity between variants and ligand. The calculations were done by plotting the maximum change of absorbance (ΔA/ΔA_max) versus the concentration of the ligand.

pK_a determination of 5HN

The absorption spectra of the 5HN solutions in buffers from pH 0.5 to pH 12.0 (at increments of ~ 0.25) were recorded. Buffers used in this experiment were: potassium chloride/hydrochloric acid (pH 0.50–2.50); sodium formate (pH 2.75–4.00); sodium acetate (pH 4.25–5.50); sodium phosphate (pH 5.75–8.25); and sodium carbonate (pH 8.50–12.00). The pK_a of 5HN can be calculated based on the spectrophotometric data. A₃₀₈ and A₃₃₀ values were plotted against pH. The plot was fitted with the equation below to obtain pK_a values [48, 49]:

(1)

where L_HA is absorbance value for HA and L_A⁻ is the absorbance value for A⁻.

Identification of a specific form of 5HN that binds to MHPCO variants

In this experiment, the ionization status of the 5HN bound to the MHPCO variants was investigated. Previously, the natural substrate, MHPC which has a similar structure to 5HN, was shown to bind to the wild-type enzyme in a tripolar ionic form [31]. To obtain the spectrum of each ionization species of 5HN (Fig. S1), the spectra of 5HN was recorded at four different pH ranges; pH < 2, pH in the range 2–4, pH between 4.8 and 8, and pH > 8. Using the Handerson–Hasselbach equation, we calculated the spectra associated with all 5HN species.

At neutral pH, 5HN exists in an ionic equilibrium of two forms: tripolar 5HN and monoanionic 5HN. Using a water-miscible solvent that can increase hydrophobicity of the system, the spectroscopic characteristics of the pure form of the monoanionic species could be obtained. The pure monoanionic spectrum of 5HN was obtained by adding 80% isopropanol into the solution. The pure monoanionic spectrum was used to subtract from the mixture spectrum to obtain the tripolar 5HN shown in Fig. 4. The derived tripolar ionic spectrum is similar to the spectrum of N-methyl-5HN, which is a 5HN analog with pure tripolar ionic characteristics.

The specific form of 5HN that binds to the MHPCO variant was identified by subtracting the enzyme spectrum from the spectra of the enzyme and 5HN mixture. In this experiment (Fig. 5), the enzyme solution with same final concentration was added to both sample and reference cuvettes, whereas 5HN (final concentration of 10 μm) was added only to the sample cuvette. Therefore, the difference spectrum obtained was mostly the spectrum of 5HN bound on the enzymes plus some amount of free 5HN. The tripolar ion species has distinct characteristics with a λ_max at 323 nm, whereas the anion has a λ_max at 285 nm. Similar experiments with the natural substrate, MHPC, were also carried out for the binding of Tyr82Phe, Tyr82His and Tyr82Arg.

Pre-steady-state kinetics: oxidative half-reaction of MHPCO variants

Variants with ring-opening ability were investigated further for their oxidative half-reaction using absorption and fluorescence detection. Pre-steady-state kinetic studies were performed using a model SF-61DX or model SHU-61SX2 stopped-flow spectrophotometer (TgK Scientific Ltd, Bradford-on-Avon, UK) in single-mixing mode. The stopped-flow flow unit was made anaerobic by flushing with a solution of sodium dithionite. An anaerobic buffer prepared by bubbling the buffer with oxygen-free nitrogen gas for 8 min was used to rinse the flow unit before the experiment. A glass tonometer used for holding the enzyme solution was initially made anaerobic by placing it inside the anaerobic glove box (Belle Technology, Wemouth, UK) overnight. An anaerobic solution of the MHPC-bound Tyr82-MHPCO variants was reduced stoichiometrically with a solution of sodium dithionite in the anaerobic glove box. Dithionite was used to reduce the enzyme because the reduction can be carried out with a stoichiometric amount and is much more rapid than by pyridine nucleotide. The MHPC-bound reduced enzyme solution was placed in the glass tonometer and the solution was loaded into syringe A of the stopped-flow machine. Buffer was equilibrated with four concentrations of oxygen before being loaded into syringe B of the machine. The reaction was conducted in 100 mm sodium phosphate buffer (pH 7.0) at 4 °C. Final concentrations after mixing were 16 μm enzyme, 300 μm MHPC, 50 mm sodium azide and 0.13, 0.31, 0.61 and 1.03 mm O_2. The dissolved oxygen concentration in buffer solution was calculated as described previously [50]. For absorption detection, stopped-flow experiments were monitored for changes in A₄₀₀ and A₄₅₆ for at least 100 s. For fluorescence detection, the reactions were monitored by excitation at 380 nm (Tyr82His) or 390 nm (Tyr82Phe) with emission at wavelengths > 455 nm and also by excitation at 470 nm with emission at wavelengths > 495 nm. The stopped-flow data were analyzed, and the k_obs was calculated using the exponential fits in program a (developed by C.-J. Chiu. R. Chung, J. Diverno and D. P. Ballou at the University of Michigan, Ann Arbor, MI, USA).

Product conversion of Tyr82 variants having ring-opening ability

A substrate-bound enzyme solution was reduced stoichiometrically using a sodium dithionite solution inside the anaerobic glovebox. Air-saturated sodium phosphate buffer (pH 7.0) containing substrate (MHPC or 5HN) was placed in a 1.5-mL microcentrifuge tube, wrapped with parafilm and transferred into the glovebox. The solution of substrate-bound reduced enzyme was immediately mixed with the oxygenated solution containing MHPC or 5HN. Final concentrations of reagents after mixing were 25 μm enzyme, 200 μm 5HN or 300 μm MHPC and 0.13 mm oxygen. The reaction was left for 15 min before being quenched by 0.1% formic acid. Denatured protein resulting from the acid quenching was separated from the supernatant using a Microcon filtration device (Millipore). Product generated and substrate left in the filtrate were analyzed using an HPLC (1100 Series or 1260 Infinity; Agilent Technologies Inc., Santa Clara, CA, USA) equipped with a UV-visible diode array detector and an ESI mass spectrometer (mass selective detector; MSD) (6120 Quadrupole; Agilent Technologies). Zorbax SB-Aq, inner diameter 4.6 mm × 150 mm (5 μm), an alkyl reversed-phase bonded phase column, was used as the stationary phase. HPLC samples were prepared in 50 mm sodium phosphate buffer containing 0.1% formic acid before each sample (50 μL) was injected for analysis. Gradient elution was performed using a mixture of 50 mm ammonium formate (pH 3.0) and acetonitrile as the mobile phase. The HPLC was operated at a flow rate of 0.5 mL·min⁻¹. The diode array detector detector was set to monitor absorption at 260, 285, 320, 340 and 450 nm. The MSD was set up in positive mode. SCAN mode of the MSD was set to look at all m/z in the range 100–500. The selected-ion monitoring mode of the MSD was set to detect m/z of 140 and 174, which belong to 5HN substrate and the ring-opening product, respectively. Percentage of product conversion or the coupling ratio could be calculated based on MSD or diode array detector data. Because FAMS is not commercially available, the amount of 5HN left in the filtrate was used to calculate the percentage of product conversion. Because the limiting reaction in this reaction is the enzyme concentration (25 μm), a 100% conversion should result in the production of 175 μm 5HN or 275 μm MHPC, whereas 0% conversion should result in 200 μm 5HN or 300 μm MHPC.

Acknowledgements

This work was supported by the grants from the Thailand Research Fund RTA5680001 and Mahidol University to P.C. and from the Thailand Research Fund RSA5580050 to J.S. T.L. received partial support from Faculty of Graduate Studies, Mahidol University. We are grateful to David P. Ballou (University of Michigan, Ann Arbor, MI, USA) and Jirundon Yuvaniyama (Mahidol University, Bangkok, Thailand) for useful discussions.

Author contributions

TL performed the experiments, analyzed the data and wrote the paper. JS performed the experiments and analyzed the data. PC designed the research, analyzed the data and wrote the paper.