Introduction

Large quantities of wastes containing the pyridine ring are produced from coal and shale oil processing, pharmaceutical, food, dye and tobacco industries and agriculture1,2. They are classified as priority pollutants by USEPA (United States Environmental Protection Agency) because of their toxicity, carcinogenicity and environmental hazards3,4. Their hydrophilic nature contributes to their environmental danger5,6. Many pyridine derivatives such as 2- and 3-hydroxypyridine, nicotinic acid and nicotine are degraded by aerobic microorganisms to generate 5-hydroxy-2-pyridone (2,5-dihydroxypyridine, 2,5-DHP) as an intermediate (Fig. 1)7,8,9,10,11,12,13. 2,5-DHP was shown to cause DNA strand scission and suggested as a potential carcinogen14. 2,5-DHP is subsequently oxidized to form N-formylmaleamic acid (NFM) and then hydrolyzed to maleamate and formate by some Pseudomonas species (shown in the red oval of Fig. 1)7,10,13. This pathway was first shown by Behrman and Stanier in Pseudomonas putida N-9 (formerly known as P. fluorescens)15,16.

Figure 1
figure 1

Pathways that degrade pyridine derivatives through 2,5-DHP7,8,9,10,11,12,13.

(a) 2-Picolinamide degrading pathway through 2,5-DHP to maleamate7,8; (b) 2-Picolinic acid degrading pathway through 2,5-DHP to α-oxoglutaric acid7; (c) 2-Hydroxypyridine degrading pathway through 2,5-DHP to blue pigment or to maleamate7,9; (d) Nicotine degrading pathway through 2,5-DHP to maleamate10,11; (e) 3-Hydroxypyridine degrading pathway through 2,5-DHP to maleamate12. (f) Nicotinic acid degrading pathway through 2,5-DHP to maleamate13; (g) 3-Hydroxy-methylpyridine degrading pathway through 2,5-DHP to maleamate7; (h) 3-Methylpyridine degrading pathway through 2,5-DHP to maleamate7; Compounds in brackets were not detected. Reactions in the red oval were studied in this paper.

Full size image

2,5-DHP oxygenase, termed NicX or Hpo for the enzyme from P. putida strain KT2440 or strain S16, respectively, catalyzes the dioxygenation of 2,5-DHP10,13; however, it has been disputed for about 40 years whether or not 2,5-DHP dioxygenases also catalyze NFM hydrolysis13,17,18,19,20,21. Investigation of the further transformation of NFM has not been straightforward for several reasons. Gauthier and Rittenberg studied the enzymatic hydrolysis of what they thought was NFM and they reported that the purified 2,5-DHP oxygenase had both dioxygenase and deformylase activities18. Behrman showed that the compound synthesized by Gauthier and Rittenberg was the trans isomer, N-formylfumaramic acid (NFF) and that both of these N-formylamides were rapidly hydrolyzed to the corresponding amides non-enzymatically. Maleamic acid is the next intermediate in the bacterial pathway whereas fumaramic acid is not metabolized by the cells. Further work was hampered by the lack of a convenient synthetic source of NFM, as the photochemical isomerization of NFF gives a mixture of geometric isomers19. Jiménez et al. prepared a homogeneous sample by cleavage of 2,5-DHP with a highly purified oxygenase (NicX), demonstrated that NicX did not catalyze the hydrolysis of NFM and identified a separate deformylase (NicD) which did so in strain KT244013. NFM formation and hydrolysis were studied in strain S1610, but NFF was mistakenly used as the deformylase substrate by following the method of Gauthier and Rittenberg18,19. Two syntheses of authentic NFM were reported in 200822.

NicX was characterized and shown to belong to a new non-heme iron(II) ring-cleavage dioxygenase family13. However, it is still unclear whether these properties are the general characteristics of the family or specific to NicX. The similarities and differences between the new family and other dioxygenases are also unknown. A 2-His-1-carboxylate motif is a nearly universal platform for other non-heme iron(II) dioxygenases23, but this architecture has not yet been confirmed in the new family.

Recently, we found a 2,5-DHP dioxygenase gene (hpo) in a nicotine-degrading strain P. putida S16. Some basic catalytic characteristics of C-terminal His6-tagged 2,5-DHP dioxygenase (His-Hpo), such as temperature effects, optimal pH, Km value and kcat value, were characterized10. In this study, two 2,5-DHP dioxygenases (Hpo and NicX) and a NFM deformylase (Nfo) were heterologously expressed and purified. Using a genuine sample of NFM, we now conclude that neither NicX nor Hpo could further hydrolyze NFM to maleamate while the deformylase isolated from the nicotine-degrading organism is capable of catalyzing the hydrolysis of both NFM and NFF, but at different rates. Kinetic studies show that the catalytic efficiency of Nfo for NFM is about 1,400 times greater than that for NFF which is consistent with the semi-quantitative results of Hillenbrand, E. L. (M.S. thesis, the Ohio State University, 1980). Comparing the catalytic characteristics of Hpo and NicX, we found some similarities and differences between them. Using structure prediction and site-specific mutagenesis analysis, we obtained data that support the 2-His-1-carboxylate motif of the new non-heme iron(II) dioxygenases family. The motif HX52HXD is different from the sequence motif of other non-heme iron(II) enzymes23.

Results

Characterization of Hpo

2,5-DHP dioxygenase activity was purified by following the ability to decrease the absorption of 2,5-DHP at 320 nm. After three steps of anion-exchange column chromatography, the recombinant Hpo was purified about 16-fold with an overall yield of 24% (Table 1). The molecular mass of the purified bands on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and native polyacrylamide gel electrophoresis (native-PAGE) were approximately 38 kDa and 120 kDa, respectively and size-exclusion chromatography analysis revealed that the mass of the purified protein as it existed in solution was about 119 ± 5 kDa, which implied a trimer (Fig. 2A, 2B, 2C). The UV-visible spectrum (data not shown) showed only a maximum at 280 nm associated with the protein and no evidence for heme, flavin or bound ferric iron, just as was found for the dioxygenase purified by Gauthier and Rittenberg17.

Table 1 Hpo purification from E. coli BL21(DE3) harboring pET28a-hpo
Full size table
Figure 2
figure 2

Characterization of purified Hpo.

(A) SDS-PAGE. M, marker proteins; 1, cell-free extracts of induced E. coli BL21(DE3) without plasmid; 2, cell-free extracts of induced E. coli BL21(DE3) harboring plasmid pET28a-hpo; 3, Q Sepharose XL pool; 4, Source 30Q fraction; 5, Mono Q fraction. (B) Native-PAGE of purified Hpo. M, native protein marker; 1, purified Hpo. (C) Elution profile of size-exclusion chromatography analysis of purified Hpo. Inverted triangles indicate the peak positions of standard proteins; inset is the fitted standard curve of a base-10 logarithm of the MW of the standard proteins versus their Kav values; each error bar represents standard deviation of three independent experiments. (D) O2 consumption and substrate depletion during 2,5-DHP degradation. Each error bar represents standard deviation of three independent experiments; all samples contained 0.025 mM Fe2+ in 1 ml of 20 mM Tris-HCl buffer (pH 7.5) at 28°C and the following additions: 0.2 mM 2,5-DHP (open triangles), 0.2 mM 2,5-DHP and 9.6 μg Hpo (closed triangles). 2,5-DHP depletion during the reactions catalyzed by Hpo was also assayed (circles). (E), (F), (G) and (H) LC-MS analysis of 18O labeling experiments. LC-MS spectrum of (E) 1 mM substrate 2,5-DHP and the product from the reaction occurring in (F) H2O and O2, (G) H218O and O2 and (H) H2O and 18O2. The molecular ion peaks for 2,5-DHP, NFM and 18O labeled NFM are indicated.

Full size image

The purified enzyme was stored at –80°C in 20 mM Tris-HCl buffer (pH 7.5) containing 1 mM dithiothreitol (DTT), under which the enzyme retained more than 90% of its activity for 30 days. Contrastingly, the enzyme lost all of its activity after being stored at 0°C in 20 mM Tris-HCl buffer (pH 7.5) for 1 day. The addition of 1 mM DTT could not regenerate the activity (data not shown).

To compare the steady-state kinetic data between Hpo and NicX, NicX was purified from E. coli BL21(DE3) harboring plasmid pET24a-nicX (Table 2). The calculated Km of Hpo for 2,5-DHP was 0.138 ± 0.013 mM, more than that of NicX (Km = 0.074 ± 0.003 mM) and the values of Vmax and kcat of Hpo were 20.5 ± 1.0 U mg−1 and 12.9 ± 0.6 s−1, higher than the values of NicX (2.00 ± 0.04 U mg−1 and 1.28 ± 0.03 s−1) (Fig. S1).

Table 2 Primers used for amplification and mutagenesis
Full size table

O2 consumption and 18O labeling

During the first 15 s of the enzymatic reaction, the rates of decrease for O2 and 2,5-DHP were approximately linear, 0.88 ± 0.01 μmol L−1 s−1 for O2 and 0.90 ± 0.02 μmol L−1 s−1 for 2,5-DHP (Fig. 2D). Therefore, the ratio of O2 used per 2,5-DHP oxidation was approximately 1:1.

Reactions were carried out using O-18 labeled water or oxygen and the products were analyzed by liquid chromatography–mass spectrometry (LC-MS) to assess the source of the added oxygen. The substrate 2,5-DHP was completely transformed in the presence of O2, while it was partially transformed in the presence of 18O2 probably due to an insufficient 18O2 concentration (Fig. 2E, 2F, 2G, 2H). The single product molecular ion peak detected in the presence of H2 18O was at m/z 142.0152, corresponding to NFM lacking 18O (theoretical m/z = 142.0146) (Fig. 2F, 2G). In contrast, the molecular ion peak of NFM with two 18O atoms (theoretical m/z = 146.0231) was found for sample produced in the presence of 18O2 (Fig. 2H). Thus the source of both oxygen atoms in the product is dioxygen, not water.

Dioxygenase substrate specificity

Many pyridine and benzene derivatives, including 2,5-DHP, 2,3-DHP, 2,4-DHP, 2,6-DHP, 2-hydroxypyridine, 3-hydroxypyridine, 4-hydroxypyridine, 2,3-pyridinedicarboxylic acid, 2,4-pyridinedicarboxylic acid, 2,5-pyridinedicarboxylic acid, 2,6-pyridinedicarboxylic acid, 3,4-pyridinedicarboxylic acid, 3,5-pyridinedicarboxylic acid, 2-picolinic acid, 4-picolinic acid, nicotinic acid, 6-hydroxynicotinic acid, pyridoxal hydrochloride, pyridoxamine hydrochloride, catechol, resorcinol, hydroquinone, pyrogallol, protocatechuate, p-hydroxybenzoic acid, gentisate and gallate, were tested as potential substrates of Hpo by monitoring substrates decrease and O2 consumption. Only 2,5-DHP was oxidized, showing that the substrate specificity of Hpo is very strict.

Aminopeptidase activity of 2,5-DHP dioxygenase

The sequence of Hpo, like that of NicX, contains an aminopeptidase (AmpS)-like conserved domain according to blastp analysis24. Many proteins homologous to Hpo are predicted to belong to the leucine aminopeptidase (LAP) or M29 peptidase families (Fig. 3A) and such a sequence could reasonably account for the previously proposed deformylase activity of the protein. To ascertain the LAP activity of the new dioxygenase family, Hpo and NicX were used to incubate with l-Leu-p-nitroaniline under many conditions. No activity was detected for either sample (data not shown).

Figure 3
figure 3

Amino acid sequence alignment analysis of Hpo.

Details of all sequences are included in supplementary information. The sequence alignments were performed with CLUSTAL W. The same residues are highlighted in red and the highly conserved residues are colored in red. (A) Partial sequence alignment of various proteins homologous to Hpo. The probable iron(II) binding sites are indicated by the black circles. The abbreviations of proteins are listed as follows: HYP, hypothetical protein; PLAP, predicted LAP; PPase, predicted M29 family peptidase. (B) Phylogenetic analysis of non-heme iron(II) ring-cleavage dioxygenase. The phylogenetic tree was constructed with the neighbor joining method using MEGA 4.1. Hpo and NicX are indicated by black stars. The abbreviations of dioxygenases are listed as follows: 2,3-CTD, catechol 2,3-dioxygenase; HPCD, homoprotocatechuate 2,3-dioxygenase; DHPAD, 3,4-dihydroxyphenylacetate 2,3-dioxygenase; DHBD, 2,3-dihydroxybiphenyl dioxygenase; BphC, 2,3-dihydroxybiphenyl-1,2-dioxygenase; Ligab, protocatechuate 4,5-dioxygenase; PCD, protocatechuate 4,5-dioxygenase. (C) Secondary structure predictions of Hpo and NicX. The secondary structures were predicted by PSIPRED v3.3 on the PSIPRED server (http://bioinf.cs.ucl.ac.uk/psipred.html)32. The blue and orange shapes indicate the secondary structure of Hpo and NicX, respectively: helix, α-helix; arrows, β-strand.

Full size image

Phylogenetic and secondary structure analyses

Phylogenetic analysis showed that Hpo and NicX do not belong to any known non-heme iron(II) ring-cleavage dioxygenase family (Fig. 3B). Structure prediction indicated that the secondary structures of Hpo and NicX are similar, although the amino acid sequence identity is only 43% (Fig. 3C). The sequence alignment shows that Nfo is highly similar with those of the α/β-hydrolase-fold enzyme superfamily (Fig. S2)13,25. The predicted secondary structure of Nfo shows remarkable similarity to NicD (Fig. S2)13.

Three-dimensional structure prediction and directed mutagenesis

Since the PSI-BLAST programs revealed no known 3D structure that was homologous to Hpo, the protein sequence was submitted to the mGenThreader server for fold assignment. Protein structures with significant hits were chosen as templates for further modeling (Table S1). Five models were generated for Hpo and the one with the lowest DOPE score was used for visualization (Table S2, Fig. 4A). The protein is predicted to contain two global domains. One is constituted mainly by residues from the N-terminal region (residues 1–126 and 173–182) and shows a Rossmann fold structure. It contains three parallel β strands (β2-β1-β3) linked to two pairs of α helices (α1–α2 and α3–α4). The other domain is mainly constituted by two layers of β sheets (β4-β5-β8-β11-β12-β16-β17-β18 and β6-β7-β9-β10-β13-β14-β1) that are flanked by α helices (α7–α8) at each side (Fig. 4A).

Figure 4
figure 4

Predicted three-dimensional model of Hpo and directed mutagenesis of supposed iron(II) binding sites.

(A) Predicted three-dimensional structural model of Hpo. (B) Comparison of the iron(II) binding sites. Purple residues, the iron(II) binding sites of BphC from Bhurkholderia xenovorans LB400 (PDB_ID, 1HAN); yellow residues, the predicted iron(II) binding sites of Hpo. (C) SDS-PAGE analysis of purified His-H257R, His-H310K and His-D312E variant proteins; M, marker proteins; 1, cell-free extracts of induced E. coli BL21(DE3) without plasmid; 2, 4 and 6 cell-free extracts of induced E. coli BL21(DE3) harboring plasmid pET28a-hisH257R, pET28a-hisH310K and pET28a-hisD312E; 3, 5 and 7, purified His-H257R, His-H310K and His-D312E variants. (D) Circular dichroism spectra of His-Hpo, His-H257R, His-H310K and His-D312E. (E) Enzyme activity comparison of His-Hpo and the His-H257R, His-H310K and His-D312E variant proteins; each error bar represents standard deviation of three independent experiments.

Full size image

Despite the phylogenetic diversity of non-heme iron(II) ring-cleavage dioxygenases (Fig. 3B), each is thought to contain a 2-His-1-carboxylate facial triad for coordinating the metal ion; significantly, a similar motif was predicted for Hpo (consisting of H257, H310 and D312) (Fig. 4B). To test the significance of this motif, E. coli BL21(DE3) strains harboring pET28a-hisH257R, pET28a-hisH310K or pET28a-hisD312E plasmid were separately constructed (Table 2). The His6-tagged variant proteins of His-H257R, His-H310K and His-D312E were purified (Fig. 4C). In these variants, the amino acid H257, H310 or D312 was replaced by R257, K310 or E312. The far-UV circular dichroism analysis showed that the secondary structures of the mutation variants were not changed (Fig. 4D, Table S3) and these variants lost almost all of their 2,5-DHP dioxygenase activities (Fig. 4E). This result implies that the amino acids H257, H310 and D312 are important or crucial for the activity of Hpo.

The predicted tertiary structure of Nfo shows a β-sheet enwrapped by a series of α helices (Fig. S3). The conserved catalytic triad characterized by a nucleophilic residue at the end of β5 (Ser94), an acidic residue at the end of β6 (Asp118), as well as a histidine located in a loop followed β8 (His238) was predicted (Fig. S3).

Deformylase activity

LC-MS analysis showed that both Hpo and NicX catalyze the formation of NFM from 2,5-DHP (Fig. S4). There was no significant hydrolysis of standard NFM after incubation at pH 7, 25°C, for 15 min (Fig. 5A, 5B). Neither Hpo nor NicX promoted the process (Fig. 5B, 5C, 5D), while Nfo catalyzed the process completely (Fig. 5E, Fig. S5). Similar results were obtained using NFF as a substrate (Fig. 5F, 5G, 5H, 5I, 5J). However, kinetic studies showed that Nfo catalyzed the hydrolysis of NFM much faster (Fig. 6). The calculated Km, Vmax and kcat for NFM are 0.94 ± 0.06 mM, 1,106 ± 19 U mg−1 and 542 ± 9 s−1, while these values for NFF are 5.83 ± 0.46 mM, 4.8 ± 0.3 U mg−1 and 2.4 ± 0.1 s−1, respectively (Fig. 6). According to these results, we can deduce that 2,5-DHP dioxygenases (Hpo and NicX) have no ability to catalyze NFM hydrolysis even if it is formed by 2,5-DHP reaction with O2. NFM hydrolysis is catalyzed by NFM deformylase, which can also catalyze NFF hydrolysis but at a reduced rate.

Figure 5
figure 5

Mass spectra of reaction solutions using NFM or NFF as substrate.

(A) Initial sample of 0.1 mM NFM. (B) 0.1 mM NFM incubated at room temperature for 15 min. (C) 0.1 mM NFM incubated with 9.6 mg l−1 Hpo and 0.025 mM Fe2+ at room temperature for 15 min. (D) 0.1 mM NFM incubated with 9.6 mg l−1 NicX and 0.025 mM Fe2+ at room temperature for 15 min. (E) 0.1 mM NFM incubated with 9.6 mg l−1 Nfo at room temperature for 15 min. (F) Initial sample of 0.1 mM NFF. (G) 0.1 mM NFF incubated at room temperature for 15 min. (H) 0.1 mM NFF incubated with 1 g l−1 Hpo and 0.025 mM Fe2+ at room temperature for 15 min. (I) 0.1 mM NFF incubated with 1 g l−1 NicX and 0.025 mM Fe2+ at room temperature for 15 min. (J) 0.1 mM NFF incubated with 1 g l−1 Nfo at room temperature for 15 min.

Full size image
Figure 6
figure 6

Kinetic studies of Nfo for NFM and NFF.

The experiments were performed in 1 ml of 20 mM Na2HPO4-NaH2PO4 buffer (pH 6.7) at 25°C. Red line, fitted Michaelis-Menten curve of Nfo for NFM; blue line, fitted Michaelis-Menten curve of Nfo for NFF; each error bar represents standard deviation of three independent experiments. The inset table shows the Michaelis-Menten equation and the fitted parameters.

Full size image

Discussion

The mechanism of NFM deformylation has been in dispute. Gauthier and Rittenberg deduced that it was catalyzed by 2,5-DHP oxygenase18, while Behrman showed that the spontaneous hydrolysis of NFM was undoubtedly important under the conditions that Gauthier and Rittenberg used18,19. Jiménez and coworkers found a NFM deformylase activity13. In our studies, we found that 2,5-DHP dioxygenases Hpo and NicX catalyze 2,5-DHP ring-cleavage to generate NFM (Fig. S4), but neither enzyme could catalyze NFM deformylation (Fig. 5C, 5D). Although NFM does hydrolyze spontaneously (Fig. 5B)19, it occurs much more slowly than the reaction catalyzed by Nfo (Fig. 5E). These findings are in good agreement with the results of Jiménez et al.13. Interestingly, hydrolysis of the synthetic trans-isomer, NFF, can also be catalyzed by Nfo and much faster than the nonenzymic hydrolysis (Fig. 5F, 5G, 5J, Fig. 6). Kinetic studies suggested that the apparent affinity of Nfo for NFM is about 6 times greater than for NFF affinity and the catalytic efficiency for NFM is about 1,400 times larger than that for NFF (Fig. 6).

Sequence alignment and phylogenetic analysis indicates that Hpo is the second member of a new ring-cleavage dioxygenase family, whose founding member is NicX13. The predicted secondary structures and 3D models of the two members of the new ring-cleavage dioxygenase family are similar. They are both predicted to contain two anti-parallel β sheets (β18-β17-β16-β4-β5-β8-β11-β12 and β6-β7-β13-β14-β15-β9-β10 in Hpo, β21-β20-β19-β5-β6-β11-β14-β15 and β7-β10-β16-β17-β18-β12-β13 in NicX) flanked by α helices (α6 and α7 in Hpo, α7 and α8 in NicX) at each side of the C terminal domain (Fig. 3C, Fig. 4A)13. These similarities are probably the reasons for the similar properties of this new family. Hpo is a trimer in solution (Fig. 2A, 2B, 2C) and NicX is a hexameric assembly composed of 2 cyclic trimers26 suggesting a sort of "trimer-module" is conserved in both NicX and Hpo. Hpo and NicX catalyze 5,6-dioxygenation reactions that convert 1 mole of 2,5-DHP to NFM consuming 1 mole O2; are unstable at 0°C without DTT; are totally inhibited by metal-chelators or oxidants; depend on iron(II) and exhibit strict substrate specificity for 2,5-DHP13. Although AmpS-like conserved domains are present in each protein, neither Hpo nor NicX has LAP activity.

Some differences exist between the two dioxygenases. Identities of the amino acid sequences between the two dioxygenases are only 43%. These may lead to higher efficient binding pockets of Hpo than those of NicX, which may cause their differences in kinetics: the apparent affinity of Hpo for 2,5-DHP is less than that of NicX while the catalytic efficiency of Hpo is 5 times greater than that of NicX (Fig. S1). The predicted 3D structures of NicX and of Hpo were some different. The N-terminal domain of NicX was predicted to contain six β strands (β2-β1-β3-β4-β9-β8) flanked by six α helices (α1-α2-α3-α4-α5-α6)13, while the N-terminal domain of Hpo was predicted to contain only 3 β strands (β2-β1-β3) and 5 α helices, with 3 at one side (α1, α2 and α5) and another 2 (α3 and α4) at the other side (Fig. 4A).

The 2-His-1-carboxylate metal-binding motif, consisting of two His residues and an Asp/Glu residue, that exists in a wide array of non-heme iron(II) enzymes23 is also predicted to be present in Hpo (composed of H257, H310, D312). Site-directed mutations led to variants at these positions, suggesting that they were all necessary for their activity (Fig. 4D, 4E, Table S3). Different from other known non-heme iron(II) ring-cleavage dioxygenases, the carboxylate donor of the 2-His-1-carboxylate motif in the new family might be an Asp residue rather than Glu. The D312 residue cannot be replaced by E312 residue in Hpo (Fig. 4E). The sequence of 2-His-1-carboxylate motif in the new family (HX52HXD) is distinct from other non-heme iron(II) enzymes23.

Sequence alignment and structure predictions reveal that Nfo is a member of α/β-hydrolase-fold superfamily which NicD belongs to (Fig. S2, Fig. S3)13,25. The residues forming the catalytic triads of Nfo (S94, D118 and H238) and NicD (S101, D125 and H245) were proposed to be the same (Fig. S2, Fig. S3)13. Although Jiménez et al. proved the indispensability of S101, D125 and H245 for the activity of NicD with site-directed mutagenesis13, many things about structures and catalytic mechanisms of Nfo and NicD are still unknown. For example, it should be clarified why Nfo and NicD have deformylase activity while other α/β-hydrolases do not when even the catalytic triads are conserved13,25. An x-ray structure of Nfo should answer these questions. This work has provided basic knowledge that can be used for the analysis of nicotine metabolism in bacteria.

In summary, Hpo is the second member of a new iron(II)-dependent ring-cleavage dioxygenase family, which works with an independent N-formylamide deformylase Nfo to catalyze the transformation of 2,5-DHP to maleamate. This study provides a better understanding of the catabolism of 2,5-DHP.

Methods

Materials

l-(-)-Nicotine (≥ 99% purity) was obtained from Fluka Chemie GmbH (Buchs Corp., Buchs, Switzerland). LAP and maleamic acid were from Sigma–Aldrich (Sigma–Aldrich Chemie GmbH, Buchs, SG, Switzerland). Fumaramic acid was made by enzymatic hydrolysis of NFF and characterized by its UV spectrum19,22. 2,5-DHP was purchased from SynChem OHG (Kassel Corp., Kassel, Germany). 18O2 and H218O were obtained from the Shanghai Research Institute of Chemical Industry. NFM used as standard was synthesized by the second method described by Behrman et al.22. NFF was made as previously reported but note that these papers refer to the product as NFM10,18. Behrman described further information on this synthesis in 197619. The MutantBest Kit was purchased from TaKaRa Biotechnology (China). Q Sepharose XL, Source 30Q, Mono Q 5/50 GL, 5 ml HisTrap FF and 5 ml HiTrap desalting columns were from GE Healthcare (Uppsala, Sweden).

Strains, media and culture conditions

E. coli DH5α was grown at 37°C and 220 rpm in lysogenic broth (LB) medium containing the appropriate antibiotics. E. coli BL21(DE3) was cultured at appropriate induction expressing conditions in LB medium with 50 μg ml−1 kanamycin. P. putida S16 was cultured at 30°C in nicotine medium as previously described2.

Gene isolation, site-directed mutagenesis and plasmid constructions

DNA was isolated, manipulated and transformed according to standard protocols27. All primers are listed in Table 2. The hpo gene was PCR amplified from the genome DNA of strain S16 using primers hishpo-f/hishpo-r or hpo-f/hpo-r. The variants of hpoH257R, hpoH310K and hpoD312E were acquired by TaKaRa MutantBest Kit according to the manufacturer's instructions. All the products were digested with NcoI and XhoI and cloned into pET28a, then transformed into E. coli DH5α or E. coli BL21(DE3). The nicX gene was PCR amplified from the genome DNA of P. putida KT2440 using primers nicX-f/nicX-r, digested with NdeI and HindIII, cloned into pET24a, then transformed into E. coli DH5α and E. coli BL21(DE3).

Activity analyses of proteins

2,5-DHP dioxygenase activity was assayed by detecting the absorption of 2,5-DHP at 320 nm (ε320 = 5,200 cm−1 M−1) with a UV-visible 2550 spectrophotometer (Shimadzu, Japan) at 20°C17. The reaction system used 1 ml of 20 mM Tris-HCl buffer (pH 7.5) containing 0.2 mM 2,5-DHP, 0.025 mM Fe2+ and 9.6 mg l−1 protein. For potential substrates, the oxygen consumption of the reaction was calculated by using a micro-respiration system (Unisense, Denmark) and the substrates decrease was monitored by HPLC. At 20°C, one unit of Hpo in 20 mM Tris-HCl buffer (pH 7.5) catalyzes the oxidation of 1 μmol of 2,5-DHP in 1 min.

LAP activity was assayed by determining the generation of p-nitroaniline at 405 nm (ε405 = 9,920 cm−1 M−1) utilizing l-Leu-p-nitroaniline as substrate28. One unit of enzyme catalyzes the production of 1 μmol of p- nitroaniline in 1 min. The reaction mixture was 5 mM l-Leu-p-nitroaniline, 0.025 mM metal ion and 9.6 mg l−1 enzyme in 1 ml of 20 mM Tris-HCl buffer (pH 7.5). LAP was used as the positive control.

Deformylase activity was measured directly by following the disappearance of NFM or NFF by mass spectrometry or, more easily, by monitoring the decrease in absorbance at 290 nm attendant upon formation of the corresponding amide19. ε290 is 185 cm−1 M−1 for NFM, 475 cm−1 M−1 for NFF, 48 cm−1 M−1 for maleamic acid and 121 cm−1 M−1 for fumaramic acid. For the mass spectrometric assay, 1 ml of 20 mM Tris-HCl buffer (pH 7) containing 0.1 mM substrate, 0.025 mM Fe2+ and appropriate concentrations of enzymes (9.6 mg l−1 for NFM and 1 g l−1 for NFF) were incubated at room temperature for 15 min. The UV assay used 1 ml of 20 mM Na2HPO4-NaH2PO4 buffer (pH 6.7) containing 1 mM substrate and appropriate concentrations of Nfo (2 mg l−1 for NFM and 0.1 g l−1 for NFF) that were incubated at 20°C. Because there is about 1% (w/w) sodium iodate in the NFM sample, more sodium iodate (1%, w/w) was added to the NFM catalyzed system to see if iodate affects the activity of Nfo; it does not. All these experiments were performed in a dark room. At 25°C, one unit of Nfo in 20 mM Na2HPO4-NaH2PO4 buffer (pH 6.7) can catalyze the hydrolysis of 1 μmol of substrate in 1 min.

Purification of proteins

E. coli BL21(DE3) strains carrying the recombinant plasmids were cultured to OD600 0.6 ~ 0.8 at 37°C, 220 rpm. The cells were induced by addition of isopropyl thiogalactoside (IPTG) to a final concentration of 0.1 mM, followed by growth at 30°C for 10 h. All the steps of chromatography were performed by an ÄKTA purifier (GE Healthcare Bio-Sciences, Sweden). Ni binding buffer contained 50 mM NaH2PO4, 300 mM NaCl and 10 mM imidazole (pH 8.0), while Ni elution buffer contained 50 mM NaH2PO4, 300 mM NaCl and 250 mM imidazole (pH 8.0). The low salt buffer was 20 mM Tris-HCl (pH 7.5) and the high salt buffer was 20 mM Tris-HCl plus 1 M NaCl (pH 7.5). One mM DTT was added to all the buffers before use.

The cells expressing His6-tagged proteins were resuspended in Ni binding buffer (OD600 50) after centrifugation (5,000 × g, 10 min) and sonicated (150 W, on ice for 10 min). After centrifugation (12,000 × g, 20 min), the supernatant was loaded onto a 5 ml HisTrap FF column pre-equilibrated with Ni binding buffer and the column was washed with 100 mM imidazole. His6-tagged proteins were eluted by 150 mM imidazole and 250 mM imidazole was applied to clear the column. Concentration of the proteins was performed by ultrafiltration (Millipore Corporation, USA) and the concentrates were desalted through a 5 ml HiTrap desalting columns to change the buffer to low salt buffer. The purified proteins were stored at –80°C.

The non-His6-tagged protein (Hpo) was purified as follows. The cell-free supernatant obtained after sonication was applied to a 1.6 × 10-cm Q Sepharose XL column pre-equilibrated with low salt buffer and eluted with a step gradient of 0, 0.6 M and 1 M NaCl in 20 mM Tris-HCl buffer (pH 7.5). Hpo eluting with 0.6 M NaCl was pooled, concentrated and desalted. The Hpo solution from the last step was loaded onto a Source 30Q column (1.6 × 10-cm) equilibrated with low salt buffer and the column was washed with a step gradient of 0.35 M, 0.45 M and 1 M NaCl. The active fractions containing 0.45 M NaCl were concentrated and desalted. Finally, the enzyme solution was applied to a 0.5 × 5-cm Mono Q column pre-equilibrated with 20 mM Tris-HCl buffer (pH 7.5). The column was washed with 0.45 M NaCl and Hpo was eluted with 0.55 M NaCl. The purified Hpo solution was desalted to change the buffer to 20 mM Tris-HCl (pH 7.5) and stored at –80°C until further use. All proteins were quantified by the Bradford method using bovine serum albumin as standard29.

Electrophoresis

Electrophoresis was performed using a 1% agarose gel30. SDS-PAGE and native-PAGE were carried out as previously described31.

Size-exclusion chromatography analysis

Size-exclusion chromatography analysis was performed as described by Jiménez et al.26. The Superdex 200 10/300 GL Tricorn column (GE Healthcare) was equilibrated with 20 mM Tris-HCl (pH 7.5) containing 1 mM DTT. Thyroglobulin (669 kDa), apoferritin (443 kDa), aldolase (158 kDa), conalbumin (75 kDa) and ovalbumin (44 kDa) in the same buffer were used to calibrate the column. The Kav values of the proteins were calculated according to the formula Kav = (Ve – V0)/(Vt – V0) (Ve: elution volume, V0: void volume, Vt: total volume of the column). The size of Hpo was determined by its Kav value three times according to a fitted standard curve of a base-10 logarithm of the molecular weight (MW) of the standard proteins versus their Kav values.

18O labeling

H218O (≥ 98% purity) and 18O2 (≥ 98% purity) were used to determine the source of the oxygen atoms in NFM. The H218O labeling reaction was 1 ml of H218O containing 1 mM 2,5-DHP, 0.025 mM Fe2+ and 10 μg ml−1 Hpo, incubated at 25°C for 5 min. The 18O2 labeling reaction was performed in a rubber sealed bottle attached to an anaerobic workstation (model AW200SG; Electrotek Ltd, UK) filled with a gas mixture of 10% H2, 10% CO2 and 80% N2. One ml of H2O containing 1 mM 2,5-DHP and 0.025 mM Fe2+ was injected into the bottle, air in the bottle was replaced by the continuous gas mixture, 18O2 was injected and 10 μg Hpo was added to the bottle, with the mixture incubated at 25°C for 5 min. Another reaction containing 1 ml H2O, 1 mM 2,5-DHP, 0.025 mM Fe2+ and 10 μg ml−1 Hpo in air was used as a control. The supernatants were directly analyzed by LC-MS after ethanol precipitation and centrifugation (12,000 × g, 2 min).

Analytical techniques

2,5-DHP, 2,5-DHP analogues, NFM and NFF were monitored at 254 nm by HPLC using an Agilent eclipse XDB C-18 (5 μm, 0.46 × 15 cm). The mobile phase was a mixture (90:10, vol/vol) of H2O containing 0.1% HCOOH and methanol containing 0.1% HCOOH, the flow rate was 0.2 ml min−1 and the column temperature was 30°C. Mass spectra of 2,5-DHP, NFM and maleamic acid were acquired using an Agilent 1290 (ultra performance liquid chromatography, UPLC) coupling with an Agilent 6230 electrospray ionization time-of-flight mass spectrometry (ESI-TOF-MS). The reaction mixtures were precipitated with ethanol and centrifuged (12,000 × g, 2 min). The supernatants were analyzed by direct-insertion mass spectra in the negative-ion turbo ion spray ionization mode. The theoretical m/z values of 2,5-DHP, NFM, NFM with two 18O atoms and maleamic acid were 110.0248, 142.0146, 146.0231 and 114.0197, respectively.

Structure predicting

The secondary structures of the proteins were predicted by PSIPRED v3.3 on PSIPRED server (http://bioinf.cs.ucl.ac.uk/psipred.html)32. The method of three-dimensional modeling was as previously described31; templates for further structure prediction of Hpo are listed in Table S1. The iron(II) binding sites of BphC (PDB_ID, 1HAN), which is a extradiol dioxygenase from Bhurkholderia xenovorans LB400, was used to compare with the predicted iron(II) binding sites of Hpo33. The comparison was illustrated by Discovery Studio 3.5 (Accelrys, San Diego, CA). The sequence of the Nfo protein was searched in the protein data bank by the PSI-BLAST program on the NCBI web site. Among all the similar sequences revealed by BLAST program, the epoxide hydrolase from Agrobacterium radiobacter AD1 (PDB_ID: 1EHY) shows the highest sequence identity of 36% and a very low E value as well34. Thus, its crystal structure was used as the template to construct the homology model of Nfo. Ten homology models were constructed by the MODELLER package. The structure with the best score is displayed (Fig. S3).

Secondary structure analysis by circular dichroism spectra

The circular dichroism spectra of His-Hpo and the variant proteins were obtained by JASCO J-815 at 25°C. The samples were 0.1 mg ml−1 His-Hpo, His-H257R, His-H310K and His-D312E in 20 mM NaH2PO4-Na2HPO4 buffer (pH 7.5), respectively. The measurement range was 190 nm – 250 nm; the cell length was 1.0 mm; 10 spectra were accumulated for each run. According to the circular dichroism spectra, the secondary structure compositions of the proteins were calculated by using the software CDPro (CONTINLL, reference protein set: SMP50).