Biochemical and Structural Studies of 6-Carboxy-5,6,7,8-tetrahydropterin Synthase Reveal the Molecular Basis of Catalytic Promiscuity within the Tunnel-fold Superfamily*

Background: The bacterial homolog, 6-carboxy-5,6,7,8-tetrahydropterin synthase, of the eukaryotic 6-pyruvoyltetrahydropterin synthase enzyme acts on the same substrate but produces different products. Results: Structural and biochemical studies trace the differential reactivity to four residues. Conclusion: Differential reactivity between the enzyme homologs is a result of small changes in the enzyme active site. Significance: This work furthers our understanding of how novel activities may arise from common protein-folds. 6-Pyruvoyltetrahydropterin synthase (PTPS) homologs in both mammals and bacteria catalyze distinct reactions using the same 7,8-dihydroneopterin triphosphate substrate. The mammalian enzyme converts 7,8-dihydroneopterin triphosphate to 6-pyruvoyltetrahydropterin, whereas the bacterial enzyme catalyzes the formation of 6-carboxy-5,6,7,8-tetrahydropterin. To understand the basis for the differential activities we determined the crystal structure of a bacterial PTPS homolog in the presence and absence of various ligands. Comparison to mammalian structures revealed that although the active sites are nearly structurally identical, the bacterial enzyme houses a His/Asp dyad that is absent from the mammalian protein. Steady state and time-resolved kinetic analysis of the reaction catalyzed by the bacterial homolog revealed that these residues are responsible for the catalytic divergence. This study demonstrates how small variations in the active site can lead to the emergence of new functions in existing protein folds.

The rapid sequencing of bacterial genomes has led to an explosion of sequence information for proteins whose functions are not known. It is rarely possible to predict enzymatic function purely on the basis of sequence because proteins with nearly identical three-dimensional structures can catalyze vastly different reactions. To bridge the gap from sequence to function, a molecular level understanding of the amino acid changes that permit evolution of novel enzymatic activity in existing folds is highly desirable.
The tunnel-fold (T-fold) superfamily is comprised of a widely distributed group of enzymes that catalyze transformations leading to the production of purines and pterins (1,2). The mammalian tunnel-fold enzyme 6-pyruvoyltetrahydropterin (PPH 4 ) 4 synthase (mPTPS) is required for the biosynthesis of 6R-L-erythro-5, 6,7, ) in eukaryotes ( Fig. 1) (3). BH 4 is an essential cofactor for many enzymes including nitric-oxide synthase and phenylalanine hydroxylase where it can act as either an electron donor or in oxygen insertion (4). In addition, this cofactor has been implicated in a vast array of physiological roles including hyperphenylalaninaemia, cellular proliferation, vascular dysfunction, and various neurological disorders (4,5). mPTPS catalyzes the conversion of 7,8dihydroneopterin triphosphate (H 2 NTP) to PPH 4 , which is subsequently converted to BH 4 by the NADPH-dependent sepiapterin reductase (6,7). More recently, an alternative pathway has been identified in higher organisms, including humans, which allows for bypassing the enzyme sepiapterin reductase through the combined actions of a carbonyl reductase and an aldose reductase or by members of the aldo-keto reductase family of enzymes (8 -11). BH 4 is typically not produced in bacteria, but is present in glycosylated forms in certain cyanobacteria and Chlorobium tepidum (12)(13)(14)(15)(16).
Although most bacteria do not posses BH 4 , a PTPS homolog is found in virtually all bacterial genomes for production of pterins and deazapurines. Previous studies from our laboratory have established that the Escherichia coli PTPS homolog, 6-carboxy-5,6,7,8-tetrahydropterin synthase, catalyzes the second step in the biosynthesis of pyrrolopyrimidine nucleosides involving the conversion of H 2 NTP to 6-carboxy-5,6,7,8-tetrahydropterin (CPH 4 ) (17). CPH 4 is a precursor to Ͼ30 natural products that range from antibiotic and anticancer agents, produced by various strains of Actinomyces, to the modified tRNA base queuosine, which is found in nearly all kingdoms of life ( Fig. 1) (18). CPH 4 synthase is promiscuous and in addition to H 2 NTP converts sepiapterin and PPH 4 to CPH 4 (19). It is not clear how this bacterial protein, which is closely related at the sequence level to its mammalian homolog involved in the biosynthesis of BH 4 , catalyzes such distinct transformations. Herein we report a structural and functional investigation of CPH 4 synthase from E. coli (QueD, so named for its role in the biosynthesis of queuosine). Comparison of the structure of QueD to its mammalian homolog revealed amino acid substitutions in the active site that may account for the distinct catalytic outcome of the bacterial enzyme. Steady state and rapid kinetic analysis of wild-type CPH 4 synthase and its site-directed variants have led to insights regarding the role of these residues in catalysis. This study highlights how minor changes in this highly conserved active site have led to the emergence of catalytic promiscuity and the evolution of catalytic function to support novel biosynthetic pathways.

EXPERIMENTAL PROCEDURES
Materials-All materials were purchased commercially (unless otherwise noted) and were of the highest purity. All assays were carried out in a Coy anaerobic chamber in an atmosphere of 95-97% N 2 , 3-5% H 2 . All buffers and materials were deoxygenated in the chamber several days prior to use.
Expression and Purification of E. coli QueD and Variants, mPTPS, and Sepiapterin Reductase-All proteins were expressed and purified as described previously (19). Selenomethioninesubstituted QueD was obtained by growth in minimal media using a method to suppress methionine biosynthesis (20). Protein concentration was determined by the Bradford method using bovine serum albumin as a standard.
Diffraction data were collected at SSRL on beam lines 9-2 and 11-1, and the data were reduced and scaled using Crystal Clear (unliganded selenomethionine-substituted protein) or XDS (21,22). The structure of the unliganded, selenomethioninesubstituted protein was solved by molecular replacement with PHASER (23) using PDB structure 2OBA as the search model. Liganded structures were solved in the same manner using the unliganded protein structure (PDB code 4NTN) as the search model. The structures were refined using REFMAC5 (24) and rebuilt using COOT (25). TLS parameters were refined (one TLS group per protein chain) for the liganded and unliganded selenomethionine protein structures. Only protein residues were included in the TLS groups (26). No cutoff was used in the refinement. All other calculations were performed using programs from the CCP4 package (27). Coordinates for unliganded selenomethionine-substituted QueD (PDB code 4NTN), selenomethionine-substituted protein soaked with sepiapterin (PDB code 4NTM), and the C27A variant of QueD soaked with sepiapterin (PDB code 4NTK) were deposited in the Protein Data Bank. Structure figures were prepared using PYMOL (28).
Enzymatic Preparation and Purification of H 2 NTP-H 2 NTP was produced as reported previously (29). The lyophilized material was dissolved in deoxygenated water in the anaerobic chamber and aliquots were frozen at Ϫ80°C.
HPLC Assays to Detect Turnover of QueD and mPTPS with H 2 NTP and Sepiapterin-The assays (0.1 ml) contained 20 mM PIPES/NaOH (pH 7.4), 10 mM dithiothreitol, 10 FIGURE 1. QueD homologs catalyze distinct reactions in mammals and bacteria. Bacterial QueD catalyzes the second step in the biosynthesis of 7-deazapurine containing compounds, including toyocamycin and the modified tRNA base queuosine (18). mPTPS catalyzes the second reaction in the biosynthesis of tetrahydrobiopterin. and 50 M substrate and were initiated by the addition of 10 M enzyme. The reactions were quenched after 1 h by addition of 30% (w/v) trichloroacetic acid to a final concentration of 10% (v/v). Precipitated protein was removed by centrifugation at 13,000 ϫ g. The supernatant (75 l) was placed into a glass vial that was sealed inside the anaerobic chamber and analyzed by an ion-pairing HPLC method described previously (19). The eluents were monitored using a diode array detector and the chromatograms were analyzed for appearance of CPH 4 (298 nm) or disappearance of sepiapterin (420 nm) and H 2 NTP (330 nm), as appropriate.
Steady State Kinetics Experiments with QueD-HPLC-based assays to determine steady state kinetic parameters for the conversion of H 2 NTP to CPH 4 were carried out as described above. Velocities were obtained from the linear portion of the reaction at each concentration of H 2 NTP. The enzyme concentration for the assays was in the 0.1-1 M range. Steady state analysis with sepiapterin as substrate was carried out in the same manner as the steady state H 2 NTP assays except that conversion of sepiapterin to CPH 4 was monitored (in a total assay volume of 1 ml) by the change in absorbance at 420 nm (⑀ 420 ϭ 10.4 mM Ϫ1 cm Ϫ1 ) (30).
Single Turnover Stopped-flow UV-Visible Spectroscopy-All single turnover experiments were carried out using a BioLogic SFM 400 fitted with a stopped-flow head; a stream of nitrogen gas maintained anaerobicity in the syringe compartment. All solutions were prepared in the anaerobic chamber and loaded into syringes before being transferred to the instrument. The solutions were the same composition as described for the above HPLC assays except one syringe contained substrate, whereas the other contained enzyme, and the two were mixed in a 1:1 ratio. To maintain a single turnover regime, the substrate concentration (50 M after mixing) was kept below the enzyme concentration (250 M after mixing) in all experiments. Data obtained with the diode array detector in the 259 -701 nm range were analyzed using KinTek Explorer software (version 3.0) (31,32). In each case, the data were fit to one or more exponentials (as described under "Results") to analyze the spectral changes corresponding to turnover.
Single Turnover Quenched-flow Experiments-These experiments were carried out essentially as described above for the stopped-flow experiments with the exception that the reactions were quenched with 30% (w/v) trichloroacetic acid to a final concentration of 10% (v/v). Samples recovered after the quenches were analyzed by HPLC as described above.
Assays with mPTPS-Reactions were conducted in the same manner as the HPLC-monitored experiments described for turnover of E. coli QueD with H 2 NTP or sepiapterin at concentrations of mPTPS and substrate indicated in the figure legends.

RESULTS
Overall Structure of QueD-A ribbon representation of wildtype QueD is shown in Fig. 2, and refinement statistics are summarized in Table 1. QueD is a homohexamer with each monomer containing 121 amino acids (33). QueD crystallizes with the biological assembly, a homohexamer in the asymmetric unit. As with the mPTPS homolog (1, 34), QueD adopts a tunnel-fold common to proteins that bind purine and pterin sub-strates. The structure consists of four sequential antiparallel ␤-strands with two antiparallel ␣-helices on the concave face of the ␤-sheet between the 2nd and 3rd strands (␤␤␣␣␤␤). The subunits assemble into a hollow barrel with an approximate 3-fold axis down the center of the cavity and approximate 2-fold axes in the equatorial plane (Fig. 2). In the structures we are reporting, the packing of hexamers in the unit cell differs between crystal forms. The C27A variant crystallized such that the 3-fold axis of the barrel is parallel to the crystallographic a axis for all hexamers. For selenomethionine-substituted protein, which crystallized under different conditions, the hexamer 3-fold axis is not parallel to a crystallographic axis and adjacent hexamers are oriented with the 3-fold axes rotated by 90 degrees.
The Active Site Architecture of QueD-The six active sites of QueD are located at the interface of the monomers that comprise the homohexamer (Fig. 2). As expected from the sequence similarity to the mPTPS homolog, and from biochemical studies showing the presence of ϳ1.1 eq of zinc divalent cation per monomer (19), we observe electron density consistent with the cation in each active site. The catalytic zinc ions (shown as green spheres) are positioned near the equator and toward the outside of the assembly, near the confluence of three protein chains. As with mPTPS, the zinc divalent cation is coordinated by the imidazole side chains of three histidine residues (His 16 , His 31 , and His 33 ) from a single subunit (35). A water molecule occupies the fourth coordination position of the zinc cation. As in mPTPS, the essential Cys 27 residue (Cys 42 in mPTPS) in the active site of QueD is in a position to interact with an His 71 -Asp 70 dyad (His 89 -Asp 88 in mPTPS) activating it to catalyze proton abstraction from the substrate to initiate catalysis (29,35). Interestingly, a second unique dyad, His 25 -Asp 54 , is also present in QueD, which we hypothesize is responsible for promoting the novel retroaldol cleavage to form CPH 4 by additional interactions with Cys 27 .
In addition to the structure of the wild-type enzyme, we also determined the structure of QueD with the product CPH 4 in the active site (PDB code 4NTM), and the QueD variant C27A with the substrate sepiapterin in the active site (PDB code 4NTK) ( Table 1). Illustrations of the active site residues and corresponding ligand electron density are shown in Fig. 3, A and B. The electron density that is observed when sepiapterin is soaked into selenomethionine-substituted QueD is consistent with formation of the product CPH 4 that clearly has a shorter side chain than sepiapterin (Fig. 3A), demonstrating that the crystals are catalytically active and convert sepiapterin to product.
By contrast, electron density after soaking sepiapterin into crystals of the catalytically inactive C27A variant shows the side chain to be long enough to be uncleaved (Fig. 3B). Therefore, we initially built sepiapterin into the electron density in each of the active sites. However, after refinement, the terminal methyl group of the side chain was out of the electron density in all six active sites. Moreover, a negative difference density peak appeared centered on the methyl carbon in several, but not all of the active sites. The lack of electron density for the methyl group and presence of difference peaks persisted through several refinement cycles. Although kinetic results (discussed below) led us to believe that no reaction should occur in crystals of the C27A variant, we replaced sepiapterin with CPH 4 in the model. Two positive difference peaks appeared, one 2.3 Å from the zinc atom, corresponding to the 2Ј-oxygen of the sepiapterin side chain, and the second in the plane of the O-C-C-O of the side chain corresponding to the methyl group. This placement of the methyl carbon requires the second carbon to be sp 2 hybridized. Subsequent refinement modeling the ligand as the enol form of sepiapterin resulted in the loss of the difference density peaks and led to electron density that covered the methyl group, which is planar with the OCCO of the side chain. The electron density for the methyl carbon is slightly weaker than that of the surrounding atoms, which may imply some structural heterogeneity in the bound ligand. Therefore, it appears that whereas the C27A variant is catalytically inactive, the predominant form of bound sepiapterin is the deprotonated enolate form with both oxygen atoms coordinated, asymmetrically, to the zinc cation. Stopped-flow evidence supporting this tautomerization of sepiapterin will be presented below.
The bound ligands have an extensive array of hydrogen bonds that satisfy each of the allowed hydrogen bond donor/ acceptor sites of the pyrimidine ring (see Fig. 4). All six active sites appear to be fully occupied with ligand in both soaked structures. No large protein conformational changes occur upon sepiapterin binding the protein. The core root mean square deviation is 0.42 Å when unliganded and liganded selenomethionine structures are overlaid using the SSM algorithm as implemented in COOT (36).
Although ligand binding induces no overall changes, there are active site adjustments when ligands bind. The largest movements are of the side chains of Glu 54 and Phe 55 , which move ϳ2 Å into the active site. The side chain of Glu 54 forms a hydrogen bond with N1 of the pterin ring and Phe 55 stacks with the pterin ring of the ligand. Crystal contacts for the selenomethionine protein⅐CPH 4 complex are different from those of the C27A protein⅐sepiapterin complex, so these small conformational changes that are the same in both structures are presumably not an artifact of crystal contacts. Structures of unliganded and sepiapterin-soaked native QueD have been deposited in the Protein Data Bank (PDB codes 3QN9, 3QN0, and 3QNA). These structures show the same small adjustments in active site conformation upon ligand binding. It is important to note that sepiapterin is not the natural substrate of the enzyme and is used as a model for the second step in the reaction catalyzed by QueD, as described in this article. Therefore, we cannot exclude the possibility that there may be more significant conforma- tional changes that occur on native substrate binding during the first half-reaction. The other significant difference in the active site is rotation of the Cys 27 sulfur toward the position of His 25 causing its rotation away from the active site. Both Cys 27 and His 25 occupy alternate conformations in the CPH 4 -containing crystal (Fig. 3A); in conformation A, the sulfur of Cys 27 and the nitrogen of His 25 are separated by 3.4 Å. In contrast, the distance between these atoms is 4.1 Å in the unliganded structure.
The structural data highlight the substantial similarities of mPTPS and QueD, as well as the conspicuous His 25 -Asp 54 dyad in the vicinity of the substrate, which may be driving the functional differences. In the remainder of this article we focus on biochemical studies of residues Cys 27 , His 25 , Asp 54 , Asp 70 , and His 71 to better understand their role in catalysis.
Steady State Kinetic Characterization of QueD and Site-directed Variants-To understand how the difference in reactivity between QueD and mPTPS may relate to the structural differences, we examined the effect of mutations to either the conserved or unique dyad on the activity of QueD. In initial end point assays to assess activity, QueD was incubated with sepiapterin or H 2 NTP and the reaction mixtures were examined by where F o is the observed structure-factor amplitude and F c is the calculated structure-factor amplitude. e As calculated using Molprobity (46). HPLC for CPH 4 production. The data (Fig. 5) show that, as expected, the C27A variant of QueD cannot convert either substrate to CPH 4 confirming the essential role of this residue in catalysis and the lack of turnover in the crystal (see above). Mutation of the conserved dyad (D70N/H71A) greatly diminished CPH 4 production from H 2 NTP, whereas mutation of the unique dyad (H25A/D54N) completely abolished this activity. Interestingly, both dyad mutants were able to produce CPH 4 from sepiapterin. The quadruple variant (H25A/D54N/D70N/ H71A), where both dyads are deleted, is completely inactive under all conditions tested. Steady state kinetic analyses of QueD and variants were carried out with H 2 NTP and sepiapterin to quantify the roles of the conserved residues. Unfortunately, we were not able to obtain reliable rates at very low concentrations of substrate, hampering determination of accurate K m values. Therefore, for the foregoing discussion of the stopped-flow data, we only use the data in Fig. 6 to estimate the lower limit of the turnover number of wild-type enzyme with H 2 NTP, and of wild-type H25A/ D54N and D70N/H71A variants with sepiapterin. The C27A variant is catalytically inactive under all conditions examined.
Pre-steady State Kinetic Characterization of Wild-type QueD and Site-directed Variants-To gain additional insights into the role(s) of the conserved residues, single turnover stopped-flow experiments were undertaken with H 2 NTP and sepiapterin as substrates. In experiments where QueD is mixed with H 2 NTP, an intermediate with a max at ϳ440 nm builds in an interval of 4.5 s after mixing and disappears in the next ϳ50 s (Fig. 7A). The data are consistent with a model that included steps for formation and disappearance of an intermediate. The data are best fit with three exponentials, the first two corresponding to the formation (1.5 and 0.6 s Ϫ1 ) and the third to the disappearance (0.1 s Ϫ1 ) of the 440 nm transient (Fig. 7B). Because the turnover number of the enzyme with H 2 NTP is 0.013 s Ϫ1 the two phases are fast enough to correspond to formation and disappearance of a kinetically competent intermediate.
Quenched The C27A variant is clearly inactive in overall turnover (Fig.  5). Nevertheless, we carried out a stopped-flow analysis of this variant with H 2 NTP and sepiapterin. We observe no detectable change in the UV-visible spectrum with H 2 NTP. However, the UV-visible spectrum of sepiapterin undergoes a red shift upon mixing with the enzyme leading to a 440-nm species, which is indistinguishable from that formed during turnover with H 2 NTP (Fig. 8A). The spectral change at 440 nm is fit to a single exponential yielding a rate constant of ϳ0.4 s Ϫ1 (Fig. 8B). Therefore, whereas the variant is inactive with respect to overall turnover, it is capable of binding sepiapterin and catalyzing formation of the transient that we observed with H 2 NTP and the catalytically active QueD variants. The slower rate constant for formation of sepiapterin observed in the rapid quench assays with H 2 NTP as substrate may reflect the rate of conversion of this transient enolate species to sepiapterin in solution upon being released from the active site after the quench.
The two dyad variants are differentially affected in turnover with H 2 NTP. The unique dyad H25A/D54N variant clearly forms the same 440-nm intermediate, but this intermediate is not turned over further (Fig. 7B). Analysis of the data at 440 nm reveals two phases with rate constants of 0.23 and 0.05 s Ϫ1 . Therefore, the absence of production of CPH 4 with the unique dyad (Fig. 5)   iant was assayed with H 2 NTP as substrate in our stopped-flow analysis (Fig. 7B).
The stopped-flow data with H 2 NTP as substrate support the hypothesis that a sepiapterin-like molecule is an intermediate of the reaction catalyzed by QueD. When the wild-type enzyme is mixed with sepiapterin, the 420-nm peak corresponding to the substrate is lost with concomitant build-up of a species with a max at ϳ340 nm (Fig. 9A). This 340-nm intermediate then disappears along with the spectral features near 420 nm over ϳ60 s. The fits of the data at 420 and 340 nm yield rate constants of 0.6 and 0.004 s Ϫ1 corresponding to conversion of sepiapterin to the 340-nm intermediate and its subsequent conversion to product (Fig. 9B). As with turnover of the wild-type enzyme with H 2 NTP, these rates are of the same order of magnitude as the turnover number of the enzyme with sepiapterin as substrate (0.01 s Ϫ1 ). Interestingly, the conserved dyad D70N/ H71A variant exhibits nearly overlapping kinetic profiles that are also fit by similar rate constants (0.4 and 0.1 s Ϫ1 , respectively) (Fig. 9B). However, whereas a disappearance of the substrate (at 420 nm) is observed with the H25A/D54N variant, consistent with the fact that this variant turns over with sepiapterin, no intermediates build up at 340 nm during the process (Fig. 9B).
Biochemical Characterization of the Promiscuity of mPTPS-Additional experiments were undertaken to confirm that the only product of the mPTPS reaction with H 2 NTP was PPH 4 . Within 1 min of mixing 0.1 mM mPTPS with 0.1 mM H 2 NTP, we observe a peak by HPLC for PPH 4 at ϳ5.8 min. PPH 4 is also a substrate for QueD and we do not observe evidence for a similar peak in any of our studies with QueD. Surprisingly, however, in reactions with mPTPS we observe that in addition to PPH 4 , CPH 4 also forms within the first 5 min and continues to build until all the PPH 4 is exhausted (Fig. 10A). To confirm that PPH 4 is the source of the resulting CPH 4 , the reaction was coupled to sepiapterin reductase. Sepiapterin reductase catalyzes the conversion of PPH 4 to tetrahydrobiopterin. Under these conditions, CPH 4 formation is inhibited and only tetrahydrobiopterin forms. Moreover, to confirm that the CPH 4 -forming activity is enzymatic and does not result from degradative breakdown of PPH 4 , we carried out the same reaction in the presence of 10-fold less mPTPS. Although PPH 4 formation was complete within the same time frame in this experiment, CPH 4 formed at a 10-fold reduced rate (data not shown). Finally, mPTPS assayed with sepiapterin also exhibited the capacity to form CPH 4 (Fig. 10B). The mPTPS used here is His 6 -tagged and was purified by affinity chromatography, whereas QueD used in all the experiments in this article was not and was purified by anion exchange and hydrophobic interaction chromatography steps. One can envision that mPTPS expressed in E. coli may be contaminated by a small amount of endogenous QueD. However, the unique subunit arrangement that produces the active site, which is contributed by residues from 3 adjacent subunits, require that the well expressed recombinant protein incorporate at minimum two subunits of endogenous QueD in the proper orientation. Finally, mPTPS has been shown to complement E. coli QueD (37). This observation could not be explained previously but it is compatible with the biochemical data shown here. Our observations support the notion that the capacity for synthesis of CPH 4 is a promiscuous activity that is present in mPTPS and that the unique dyad simply accelerates the production of this transformation in QueD.

DISCUSSION
mPTPS and QueD both utilize H 2 NTP as substrate, but despite a strikingly similar overall-fold and active site architec- ture they produce distinctly different major products. The active sites of both proteins are composed of a constellation of residues that are contributed by adjacent subunits in the biological assembly (34). The substrate in each active site is bound to a conserved zinc divalent cation via the C1Ј and C2Ј hydroxyl groups and the proteins retain similar binding interactions with the substrate (35). Each active site houses an essential cysteine residue whose activation with a conserved Asp/His dyad is proposed to initiate catalysis (29). The most conspicuous structural difference between the two active sites is the presence of an additional His/Asp dyad in the bacterial enzyme on the opposite face of the Cys 27 residue relative to the conserved dyad. Therefore, we investigated site-directed variants of QueD to determine the contribution(s) of each dyad to the functional differences between QueD and mPTPS.
Although the biological role of QueD is to convert H 2 NTP to CPH 4 , we have shown previously that the enzyme also utilizes sepiapterin as substrate converting it to CPH 4 (19). Our working hypothesis was that the reaction likely involves intermediates that resemble this alternate substrate. Indeed time-resolved studies of QueD reveal the build up and disappearance of a kinetically competent intermediate with a max of 440 nm. The identity of the intermediate is difficult to establish unambiguously; however, rapid quench studies reveal transient for-mation of sepiapterin in the same time frame as that of the 440-nm transient. The differences between the 440-nm transient observed by stopped-flow and the solution spectrum of sepiapterin may represent tautomerization to the enolate form, which is also observed in the crystal structure of the C27A variant, despite the inability of this variant to catalyze the overall reaction.
We hypothesize that the role of the conserved dyad is to initiate catalysis with H 2 NTP as substrate, as mutation of these residues leads to ϳ80-fold decrease in overall turnover with H 2 NTP without effecting turnover parameters regarding catalysis with sepiapterin. By contrast, mutation of the unique dyad leads to a ϳ160-fold drop in activity of the enzyme with sepiapterin and buildup of the 440-nm intermediate when assayed with H 2 NTP, which is not carried through further to CPH 4 . As described above, the time-resolved data clearly show that a sepiapterin analog is an intermediate in the reaction. This observation, in the context of the fact that sepiapterin is a sub-strate, suggests that the conversion of H 2 NTP to CPH 4 is a multistep process. Although we cannot assign specific roles to these residues, the dyads clearly have differential contributions at distinct stages in the catalytic cycle. A mechanistic paradigm for QueD is shown in Fig. 11. The proposed mechanism has many features in common with those

Catalytic Promiscuity in the Tunnel-fold
proposed for mPTPS. Specifically, both enzymes have an essential active site cysteine residue, a divalent zinc ion, and a conserved dyad within hydrogen bonding distance to the conserved Cys. The first step in the reaction is the binding of the substrate to the active site via hydrogen bonding and coordination of the 1Ј-and 2Ј-hydroxyls to the active site zinc divalent cation. In the structure of the C27A variant of QueD complexed with sepiapterin, the hydroxyl oxygen atoms are bound asymmetrically and at distances of ϳ2 and ϳ 2.3 Å from the zinc ion.
In the catalytic cycle the required Cys residue, which is activated by interaction with the conserved Asp 70 -His 71 dyad, abstracts a proton from C2Ј eliminating the triphosphate and is subsequently tautomerized to a sepiapterin-like intermediate. The unique dyad, either directly or in concert with the catalytically essential Cys 27 , activates a water molecule for the next half-reaction, which entails the elimination of the C2Ј-C3Ј as acetaldehyde to form the CPH 4 product. We favor a role for Cys 27 in this half-reaction as well because the C27A variant is completely inactive in conversion of sepiapterin to CPH 4 . Again, the C27A variant appears to catalyze a change in the conjugation of sepiapterin upon binding leading to a 440-nm species, corroborating the planar geometry for sepiapterin observed in the x-ray crystal structure of the C27A variant. This mechanism accounts for all our observations on QueD as well as the literature on the mechanism of the mPTPS enzyme (38 -42).
Recent studies have shown that the mPTPS enzyme can complement production of queuosine in a ⌬queD strain of E. coli (37). This has been somewhat puzzling, as mPTPS had never been shown to catalyze the conversion of H 2 NTP to CPH 4 . Our biochemical data with purified mPTPS provide an experimental basis of the complementation. As shown in Fig.  10, in addition to converting H 2 NTP to PPH 4 , mPTPS also catalyzes conversion of PPH 4 to CPH 4 , albeit at vastly slower rates. The extent of modification of tRNA to queuosine under normal growth conditions is not known, but the overexpression of the mammalian protein is likely to produce the necessary pool of CPH 4 to support the tRNA modification. As is clear in the proposed mechanism, the active site zinc cation should be able to promote a number of tautomerizations and one can readily propose mechanisms by which zinc-bound PPH 4 would be converted to CPH 4 . We note that whereas the unique dyad of QueD appears to be necessary for conversion of sepiapterin to CPH 4 , it is not essential. The active site of mPTPS appears to have been set up to permit zinc-mediated tautomerizations and small variations in active site environment, such as the introduction of the unique dyad, are sufficient to amplify rates of the reactions that produce CPH 4 .
The differences in reactivity between the highly similar mPTPS and bacterial QueD homologs provide an interesting case of evolution of new catalytic activities in existing enzyme folds. The conversion of H 2 NTP to CPH 4 , albeit at very low levels, is clearly a promiscuous activity as the biosynthesis of queuosine is only carried out in prokaryotes. The concept of enzyme promiscuity leading to novel metabolic pathways was first proposed by Jensen (43) (for a review, see Ref. 44). Following a gene duplication event, enzymes that possess some lowlevel secondary activity will have a selective advantage toward evolution of a new activity as it will theoretically require fewer advantageous mutations to elevate this ability to a prominent component of a metabolic pathway. An enzyme possessing this low-level catalytic promiscuity will therefore have an evolutionary "head start." Recently, a model regarding the evolution of new metabolic pathways, the innovation-amplification-divergence model, describes quite exquisitely how these new activities could be borne of previous weak activities in ancestral enzymes (45). QueD and mPTPS may represent naturally occurring examples of the evolutionary models that posit promiscuous activity is required for emergence of new biological function(s).