Conserved YjgF Protein Family Deaminates Reactive Enamine/Imine Intermediates of Pyridoxal 5′-Phosphate (PLP)-dependent Enzyme Reactions*

Background: YjgF proteins are conserved across the three domains of life. Results: YjgF proteins deaminate unstable products of PLP-dependent enzyme threonine dehydratase by positioning the enamine/imine close to water in the active site. Conclusion: YjgF is an enamine/imine deaminase and is renamed RidA. Significance: RidA removes reactive metabolites released by PLP-dependent enzymes before they can damage cellular components. The YjgF/YER057c/UK114 family of proteins is conserved in all domains of life, suggesting that the role of these proteins arose early and was maintained throughout evolution. Metabolic consequences of lacking this protein in Salmonella enterica and other organisms have been described, but the biochemical function of YjgF remained unknown. This work provides the first description of a conserved biochemical activity for the YjgF protein family. Our data support the conclusion that YjgF proteins have enamine/imine deaminase activity and accelerate the release of ammonia from reactive enamine/imine intermediates of the pyridoxal 5′-phosphate-dependent threonine dehydratase (IlvA). Results from structure-guided mutagenesis experiments suggest that YjgF lacks a catalytic residue and that it facilitates ammonia release by positioning a critical water molecule in the active site. YjgF is renamed RidA (reactive intermediate/imine deaminase A) to reflect the conserved activity of the protein family described here. This study, combined with previous physiological studies on yjgF mutants, suggests that intermediates of pyridoxal 5′-phosphate-mediated reactions may have metabolic consequences in vivo that were previously unappreciated. The conservation of the RidA/YjgF family suggests that reactive enamine/imine metabolites are of concern to all organisms.

Data presented herein show that IlvA-generated enamine/ imine intermediates serve as substrates for YjgF. Results here show that YjgF family members increased the rate of hydrolysis of the enamine/imine metabolites. This study identified the first conserved biochemical function for the YjgF/YER057c/ UK114 family and led to the hypothesis that the role of this ubiquitous family is to accelerate the release of ammonia from enamine/imine intermediates in the low water environment of the cell.

EXPERIMENTAL PROCEDURES
Chemicals-Potassium ferricyanide and other chemicals used in this study were purchased from Sigma-Aldrich. L-[U- 14 C]Threonine (175 mCi/mmol) was purchased from American Radiolabeled Chemicals Inc.
Bacterial Strains and Plasmids-Pyrococcus furiosus and Bacillus subtilis DNA were generously provided by J. Escalante-Semerena. All constructs for overexpression were in vector pET20bϩ (Novagen) except for the plasmid containing chrD (pET15b (23)) and that containing rutC (pCA24N (24)). Sitedirected mutagenesis of yjgF was performed using the QuikChange II site-directed mutagenesis kit, according to the manufacturer's protocol (Agilent Technologies).
Enzyme Purifications-IlvA and YjgF were purified using C-terminal His 6 tags and nickel ion affinity chromatography, as described previously (13). Variants YjgF Y17F , YjgF R105A , YjgF E120K , and YjgF E120A and protein homologs from B. subtilis (YabJ) and P. furiosus (PF0668) were purified following the same protocol as for YjgF. Briefly, Escherichia coli BL21AI cells carrying the appropriate plasmid were grown to OD 650 ϭ 0.7, expression was induced with 0.2% (w/v) arabinose, and growth continued for 9.5 h at 37°C. Cells were disrupted with a French pressure cell, and extract was clarified by centrifugation (48,000 ϫ g for 45 min). Cell-free extract was applied to nickelnitrilotriacetic superflow resin, and proteins were purified according to manufacturer's protocol (Qiagen) using column chromatography. Purified protein was concentrated in a 3000 molecular weight cut-off centrifugal filtration unit (Millipore) and then dialyzed using a 10,000 molecular weight cut-off Slide-A-Lyzer dialysis cassette (Thermo Scientific) in 10 mM HEPES, pH 8.0, with 10 mM EGTA, then in 10 mM HEPES, pH 8.0, and lastly in 10 mM HEPES, pH 8.0, with 20% (v/v) glycerol. RutC (YcdK) from E. coli was purified from a complete set of E. coli, K-12 ORF Archive (ASKA) collection (JW0995), with an N-terminal His 6 tag (24). Cells were grown to OD 650 ϭ 0.5 and induced with 0.2 mM isopropyl-␤-D-thiogalactoside and then grown for 9 h more at 37°C. ChrD was purified with an N-terminal His 6 tag. Cells were grown to OD 650 ϭ 0.7 and induced with 0.2 mM isopropyl-␤-D-thiogalactoside and 0.2% (w/v) arabinose and then grown for 9.5 h more at 37°C. Protein purification procedures for RutC and ChrD were the same as for other YjgF homologs above. All variants and homologs were stably expressed in S. enterica, as determined by SDS-PAGE with Coomassie Brilliant Blue staining and/or Western blotting.
Thin-layer Chromatography-Reaction mixes contained 50 mM HEPES, pH 8.0, 10 M PLP, 0.9 M IlvA, 50 mM (60 nCi) L-[U- 14 C]threonine, and 3 M YjgF when applicable. Reactions were incubated for 2 h at 37°C, and then radiolabeled compounds were separated by thin-layer chromatography. Spots containing 5 nCi were applied to silica, cellulose, and polyethyleneimine-cellulose plates, and each type of plate was run in each of the following solvent systems: 1:1. The positions of radioactive spots were detected using a Cyclone Storage phosphor screen (Packard Instrument Co.) and a Typhoon FLA 9000 scanner (GE Healthcare). In all solvent/plate combinations, a single 14 C-labeled product with the same retention factor as 2-ketobutyrate was detected.
Mass Spectral Analysis-Reactions containing 50 mM ammonium bicarbonate, IlvA, and 50 mM threonine, in the presence and absence of YjgF, were incubated for 2 h. The proteins were removed, and the sample was submitted to positive (ϩ) and negative (Ϫ) time-of-flight mass spectral analysis, carried out at the University of Wisconsin Biotechnology Center.
Kinetic Assays-2-Ketobutyrate formation was assayed by following absorbance at 230 nm, as described previously (25). Assays contained 50 mM each MES, HEPES, and TAPS (pH 7.5) or TAPS, CHES, and CAPS (pH 9.5) buffer, 10 M PLP, 0.9 M IlvA, and 1.6 M YjgF or YjgF variants/homologs when applicable. (Assays were also performed in the absence of additional PLP with no change in results.) Substrate threonine or serine was added to begin reactions, and concentrations were varied as indicated. Each reaction was performed in a 200-l final volume and assayed continuously in a quartz 96-well plate using a SpectraMax Plus (Molecular Devices) at room temperature (22°C). Initial rates were calculated from the change in A 230 nm over the first 30 s (pH Յ 9.0) or over 30 -60 s (pH Ͼ 9.0). Data IlvA dehydrates threonine to generate enamine/imine tautomer intermediates. The imine is hydrolyzed non-enzymatically to form the final products 2-ketobutyrate and ammonia. The data reported under "Results" showed that a biochemical function of YjgF is to catalyze the hydrolysis of these intermediates, thus removing enamine/imine from solution and increasing the rate of production of 2-ketobutyrate.
pH Titration-pH titrations were performed with mixtures of the following buffers, each at 50 mM: MES, HEPES, and TAPS (pH Յ 8.6) or TAPS, CHES, and CAPS (pH Ն 8.3) (26). In reactions carried out at pH values greater than 9.0 without YjgF, a 30-s burst phase was observed before the reaction velocities became linear. To accommodate the burst, initial velocities for all reactions above pH 9.0 were calculated from 30 to 60 s. Data were plotted using GraphPad Prism 4.0b.
Assays for Ferricyanide Reduction-Reactions contained 50 mM each TAPS, CHES, and CAPS buffer, pH 9.5, 10 M PLP, 0.9 M IlvA, and 1.6 M YjgF when applicable. Ferricyanide was added to a final concentration of 1, 2, or 5 mM as indicated. Reduction of ferricyanide was monitored continuously at 420 nm in a 96-well quartz plate, as described previously (20). Because ferricyanide also absorbs at 230 nm, 2-ketobutyrate concentrations in these reactions were determined by 2,4-dinitrophenylhydrazine derivatization, as described (27).

YjgF Increases Rate of 2-Ketobutyrate Formation from
Threonine-Results from thin-layer chromatography experiments showed the formation of a single product in reaction mixtures containing IlvA and threonine. The presence of YjgF did not yield an additional product. The single product was verified as 2-ketobutyrate using time-of-flight mass spectrometry (expected m/z, 101.0245; observed m/z, 101.0248).
The initial rates of 2-ketobutyrate formation by IlvA were determined with and without YjgF by following absorbance at 230 nm in a continuous assay (25). The presence of YjgF in the reaction mix significantly increased the rate of 2-ketobutyrate formation ( Fig. 2A). At pH 7.5, with 15 mM threonine as substrate, the initial rate of 2-ketobutyrate formation was nearly 2-fold higher in the presence of YjgF when compared with IlvA alone, whereas at pH 9.5, YjgF stimulated the initial rate 4 -5fold (Fig. 2B). YjgF and IlvA proteins were present at monomeric concentrations of 1.6 and 0.9 M, respectively. Because YjgF is a trimer (1, 2) and IlvA is a tetramer (28), the YjgF:IlvA oligomeric ratio in these experiments was ϳ2:1. Results from the spectrophotometric assay were corroborated by two independent methods that quantified 2-ketobutyrate using a discontinuous assay. Derivatization of 2-ketobutyrate by o-phenylenediamine (29) or 2,4-dinitrophenylhydrazine (27) detected an increased rate of 2-ketobutyrate formation when YjgF was present in the reaction mixtures comparable with that observed by following A 230 nm .
Activity of YjgF Is Greater at High pH-The pH of the reaction was titrated from 7.0 to 10.0, using MES/HEPES/TAPS buffer for pH Յ8.6 and TAPS/CHES/CAPS buffer for pH Ն8.3 and 15 mM threonine as substrate. The effect of YjgF on the initial rate of 2-ketobutyrate formation was significantly higher with increasing pH (Figs. 2B and Fig. 3). The rate increase with increased pH could be attributed in part to a greater stability of the enamine/imine intermediate at high pH (20,30). Notably, in reactions with IlvA alone at pH values greater than 9.0, a 30-s burst phase was observed before the reaction velocities became linear. To accommodate the burst, initial velocities for all reac-tions above pH 9.0 were determined from 30 to 60 s instead of from the first 30 s. This burst was not observed in the presence of YjgF. The presence of a burst in the initial velocity of the reaction was consistent with a reactive product inhibiting IlvA.
YjgF Is an Enamine/Imine Deaminase-The availability of enamine/imine metabolites in the IlvA reaction mixes was determined with ferricyanide. Ferricyanide reacts with the enamine/imine metabolites of threonine released from IlvA to produce a decarboxylated non-ketobutyrate product and reduced ferricyanide in a 1:1 ratio (20, 31). Ferricyanide reduction was monitored as a decrease in absorbance at 420 nm. In a  reaction mix with IlvA, 10 mM threonine, and 5 mM ferricyanide, ferricyanide was rapidly reduced in the first 15 min, with Ͼ98% reduced within an hour. In contrast, in the presence of YjgF, ferricyanide reduction occurred slowly, and a plateau was reached after ϳ10 min, with only 24% of the ferricyanide reduced after 2 h (Fig. 4A). These results suggested that the YjgF protein removed the enamine/imine reaction intermediates from solution. A series of reactions was set up that contained IlvA, 10 mM threonine, and different concentrations of ferricyanide in the presence or absence of YjgF. After 2 h, the amounts of 2-ketobutyrate produced and ferricyanide reduced were determined, and the data are shown in Fig. 4B. At all concentrations of ferricyanide, YjgF dramatically decreased the amount that was reduced. In the absence of YjgF, the ferricyanide was almost completely reduced, thereby eliminating an equimolar amount of the enamine/imine metabolites and leaving the remainder to be converted to 2-ketobutyrate by hydrolysis in solution. For example, in the presence of 5 mM ferricyanide and IlvA alone, 50% (5 mM) of the reaction intermediates from threonine reduced ferricyanide, whereas the other 50% proceeded to form 2-ketobutyrate. In contrast, when YjgF was present, Ͼ90% of the substrate threonine was converted to 2-ketobutyrate, and only ϳ10% of the reaction intermediates were available to reduce ferricyanide (Fig. 4B, right two bars). In each reaction condition, the sum of reduced ferricyanide and 2-ketobutyrate was ϳ10 mM, indicating that all of the threonine provided was consumed by IlvA and accounted for in the two products. Control experiments measuring product formation confirmed that ferricyanide did not affect the initial rate of 2-ketobutyrate formation with or without YjgF. These data support the conclusion that YjgF eliminated enamine/imine intermediates from solution before ferricyanide could react with them.
YjgF Active Site Configuration Is Suggested by Activity of Variant Proteins-The data above suggested that YjgF hydrolyzed an enamine/imine intermediate in a reaction that released ammonia and formed the stable product 2-ketobutyrate. Structural studies and bioinformatic analyses have suggested residues that could comprise an active site of YjgF (Fig. 5A). The structure of E. coli homolog TdcF (73% identical to S. enterica YjgF) with serine, threonine, and the 2-ketobutyrate enol bound implicated residues Arg-105 and Glu-120 in direct binding of substrate (5). Modeling the enamine/imine substrate into this putative active site suggested that Arg-105 formed a bidentate salt bridge with the carboxylic acid of the substrate, as proposed by Burman et al. (5). However, based on the activity of YjgF and the apo structure of TdcF containing an ordered water (5), we predicted that Glu-120 did not bind the substrate, but rather formed hydrogen bonds with a water molecule in the active site (Fig. 5B). In this scenario, the hydroxyl group of Tyr-17 would stabilize the imine nitrogen. This conformation would place the water Ͻ2 Å from C2 of the substrate, and the hydrogen bond between the Tyr-17 and the imine nitrogen would draw electron density away from C2, making the attack by water more favorable (Fig. 5B).
Variants of S. enterica YjgF were generated to probe the active site mechanism. Four stable YjgF variants (YjgF Y17F , YjgF R105A , YjgF E120K , and YjgF E120A ) were purified and assayed. The activity of each variant was determined as a ratio of the initial rate of 2-ketobutyrate formation with/without the YjgF protein. The data in Fig. 6 showed that all four variants were defective, with YjgF R105A having the lowest activity. None of the mutations completely eliminated activity. Additionally, YjgF R105A was the only variant that still exhibited burst kinetics. These results are consistent with the model that the role of the active site is to bring the imine and water together to facilitate the attack by water and release of ammonia. This interpretation is consistent with the dominant role of Arg-105 as this residue would hold the imine in place. An alternative model suggesting that Glu-120 had an acid/base role in activating the water molecule was eliminated by the behavior of variants YjgF E120K and YjgF E120A . The activity remaining in both variants indicated that Glu-120 was not an essential catalytic residue. These data further suggested that the Glu-120 residue was critical for the increased activity of the YjgF protein at higher pH. Each of the four variants, with the exception of YjgF R105A , complemented a yjgF mutant defect in vivo. This result emphasized the importance of the Arg-105 residue and suggested that the decrease in activity of this variant in vitro was physiologically relevant.
YjgF Increased Rate of Pyruvate Formation from IlvA and Serine-IlvA catalyzes the dehydration of serine to generate pyruvate via the corresponding enamine and imine intermediates (15,32). Pyruvate formation from IlvA and serine was measured at pH 9.5 by following A 230 nm in the presence and absence of YjgF. Like 2-ketobutyrate formation, the rate of pyruvate formation was 2.2-fold faster (at 30 mM serine) in the presence of YjgF (Fig. 7). This result identified a second substrate for YjgF and was consistent with the active site configuration proposed in Fig. 5, which did not implicate residues to bind C3 or C4 present in the enamine/imine derived from threonine.

Enamine/Imine Deaminase Activity Is Encoded in YjgF
Homologs from Diverse Organisms-Homologs of YjgF from B. subtilis (YabJ; 50% identity), P. furiosus (PF0668; 56% identity), Cucumis sativus (ChrD; 42% identity), Homo sapiens (UK114; 45% identity), and E. coli (RutC/YcdK; 29% identity) were assayed for 2-ketobutyrate formation activity (see Fig. 5A for sequence alignment). Each of the five homologs significantly increased the rate of 2-ketobutyrate formation from threonine over that of IlvA alone (Fig. 8). Low identity homolog RutC had FIGURE 5. Alignment of family members and proposed mechanism for YjgF. A, amino acid sequences of YjgF family members assayed in this study were aligned using CLUSTALW (45). The numbering of residues is based on the S. enterica YjgF sequence. YjgF is from S. enterica; YabJ is from B. subtilis; PF0668 is from P. furiosus; ChrD is from C. sativus; UK114 is from H. sapiens; and RutC is from E. coli K12. Boxed residues in the S. enterica sequence indicate amino acids mutated in this study. Asterisks indicate residues that are invariant among high identity family members (9) and have been predicted to be involved in active site structure or catalysis (5). B, a reaction scheme for the deamination of iminobutyrate by YjgF. The backbone of Cys-107 and the side chain of Glu-120 hydrogen-bond with water, whereas the side chain of Arg-105 forms a salt bridge with the carboxyl group of the substrate and the Tyr-17 hydroxyl group hydrogen bonds with the imine nitrogen. This setup facilitates the attack of water on C2 of the imine. Then a rearrangement occurs, releasing the products ammonia and 2-ketobutyrate.  less activity in this assay than the higher identity homologs. This result was consistent with the reduced activity of the YjgF Y17F variant because RutC also has a phenylalanine at the residue corresponding to Tyr-17 in the wild-type protein (Fig.  5A).

DISCUSSION
The YjgF/YER057c/UK114 family of proteins is broadly conserved and found in all domains of life. Past work, primarily in S. enterica, suggested that the diverse phenotypes of yjgF mutants reflected a global metabolic role for this protein family (11)(12)(13)33). The results reported herein define an enamine/ imine deaminase activity associated with the YjgF protein and homologs from all domains of life, including humans. This is the first report of a conserved biochemical activity for this ubiquitous family. The activity described here for YjgF proteins suggests that the role of these proteins in vivo is to accelerate the release of ammonia from enamine/imine intermediates generated by PLP-dependent dehydratases. The conservation of YjgF proteins may be an indication that the low free-water environment of the cell extends the half-life of these metabolites. These findings have implications for reactions that use solvent water to complete product formation and may indicate a need for accessory proteins that were not anticipated from past in vitro work performed in aqueous solutions.
Proposed Active Site Configuration for YjgF Function-Previous structural studies (5,9) and the biochemical analyses of YjgF variants performed here support a mechanism for YjgF catalytic activity (Fig. 5B). We propose that Arg-105 forms a bidentate salt bridge with the carboxylic acid of the imine/ enamine substrate, consistent with a previous structure containing the 2-ketobutyrate enol in the active site (5). In this scenario, the hydroxyl group of Tyr-17 stabilizes the nitrogen of the substrate, drawing electron density and activating the substrate for nucleophilic attack by a water molecule stabilized by Glu-120 and the backbone of Cys-107. Evidence for this placement of water was found in the apo structure of TdcF, where an ordered water was within an adequate distance to hydrogen-bond to Glu-120 and Cys-107 (5). After the nucleophilic attack, a rearrangement could displace ammonia and generate 2ketobutyrate. The decreased activities of variants YjgF Y17F , YjgF R105A , YjgF E120A , and YjgF E120K are consistent with this general reaction scheme. Arg-105 is a critical residue, responsible for binding the substrate and required to complement a yjgF mutant defect in vivo. Significantly, the YjgF R105A variant had the biggest effect on in vitro activity and was the only variant that failed to eliminate the burst kinetics observed in high pH reactions with IlvA alone. Tyr-17 played a significant role in stabilizing substrate, as indicated by ϳ50% reduction in activity at both pH levels when this residue was substituted. The behavior of the YjgF E120A and YjgF E120K variants indicates that Glu-120 has a more significant role at higher pH, where enamine/ imines are known to be more stable (20,30). This result could reflect a need to optimally orient the water molecule in the active site for the nucleophilic attack. Although altering the active site residues led to decreased activities, no mutation completely abolished function. This result suggests that YjgF acts to place water and reactive enamine/imine substrates in close proximity to facilitate a more rapid release of ammonia and does so without a catalytic residue. In the cellular environment, where available water may be limiting, the impact of YjgF proteins is expected to be significant in facilitating stable product formation.
Diversity among YjgF Homologs-YjgF homologs from five diverse organisms were found to have the same activity as the S. enterica protein when measured in assays described here and previously (13). The diversity of homologs that have activity in the assays minimizes the possibility that a protein-protein interaction is required for YjgF activity. Consistent with the activity demonstrated here, some enzymes that participate in other enamine deamination reactions can be defined as low identity members of the YjgF family (25-30% identical to S. enterica). In Pseudomonas pseudoalcaligenes, a YjgF-like protein was shown to deaminate the short-lived enamine 2-aminomuconate to 4-oxalocrotonic acid during the degradation of nitrobenzene (34), and the same reaction has been reported for homologs NbzE in nitrobenzene-degrading Pseudomonas putida HS12 (35) and AmnD in Pseudomonas sp. AP-3 (36). In Burkholderia cepacia, a YjgF homolog increased the rate of 2-aminomuconate deamination more than 70-fold when compared with non-enzymatic hydrolysis in a tryptophan degradation pathway (37). It is likely that this enhanced catalysis was due to an enamine deamination similar to that described here. Finally, in Bordetella sp. strain 10d, AhdB deaminates 2amino-5-carboxymuconic 6-semialdehyde, an unstable enamine intermediate (38,39). Interestingly, rutC in E. coli is included in a pyrimidine degradation operon, encoding a pathway that proceeds via the unstable enamine intermediate 3-aminoacrylate (40). Although the biochemical role of RutC in this pathway was not defined, our data support the prediction that RutC acts on the unstable enamine intermediate.
In addition to RutC, other YjgF family members are encoded within operons dedicated to a specific metabolic pathway. Often these operons contain a dedicated threonine and/or serine dehydratase. For example, the anaerobically expressed tdc operon in E. coli (41) and the alanylclavam biosynthetic operon in Streptomyces clavuligerus (42) have genes encoding both threonine/serine dehydratases and YjgF homologs. These genomic organizations are consistent with the model that the YjgF proteins act on short-lived reactive metabolites generated during specific metabolic processes.
Change to Genome Annotation and Significance of Rid Proteins-We have renamed yjgF as ridA (reactive intermediate/imine deaminase A) to reflect the biochemical activity of the gene product and to meet the standards of gene nomenclature. This study determined that RidA acts on reaction intermediates from a PLP-dependent enzyme, threonine dehydratase, to increase the rate of product formation above the solvent-mediated reaction. This activity has two anticipated consequences in vivo: 1) increasing the hydration rates in low water environment to improve product generation and 2) removing reactive metabolites. We suggest that the diverse phenotypes described for ridA (yjgF) mutants in S. enterica (10,11,13,33) result from the accumulation of enamine/imine intermediates and that the nucleophilic nature of these metabolites allows them to react with and modify cellular targets, altering their function. The plausibility of this scenario is supported by literature reports that the aminoacrylate enamine derived from serine can inactivate a number of enzymes in vitro (43,44). Thus, it is reasonable to suggest that Rid proteins evolved to facilitate the release of ammonia from enamine/ imine intermediates generated during normal metabolism, and as a result, reduce the impact of these reactive metabolites on cellular components.