Biosynthesis of Pteridines

GTP cyclohydrolase I catalyzes a ring expansion affording dihydroneopterin triphosphate from GTP. [1′,2′,3′,4′,5′-13C5,2′-2H1]GTP was prepared enzymatically from [U-13C6]glucose for use as enzyme substrate. Multinuclear NMR experiments showed that the reaction catalyzed by GTP cyclohydrolase I involves the release of a proton from C-2′ of GTP that is exchanged with the bulk solvent. Subsequently, a proton is reintroduced stereospecifically from the bulk solvent. This is in line with an Amadori rearrangement mechanism. The proton introduced from solvent occupies the pro-7R position in the enzyme product. The data also confirm that the reaction catalyzed by pyruvoyltetrahydropterin synthase results in the incorporation of solvent protons into positions C-6 and C-3′ of the enzyme product. On the other hand, the reaction catalyzed by sepiapterin reductase does not involve any detectable incorporation of solvent protons into tetrahydrobiopterin.

Multinuclear NMR experiments showed that the reaction catalyzed by GTP cyclohydrolase I involves the release of a proton from C-2 of GTP that is exchanged with the bulk solvent. Subsequently, a proton is reintroduced stereospecifically from the bulk solvent. This is in line with an Amadori rearrangement mechanism. The proton introduced from solvent occupies the pro-7R position in the enzyme product. The data also confirm that the reaction catalyzed by pyruvoyltetrahydropterin synthase results in the incorporation of solvent protons into positions C-6 and C-3 of the enzyme product. On the other hand, the reaction catalyzed by sepiapterin reductase does not involve any detectable incorporation of solvent protons into tetrahydrobiopterin.
Pteridines serve as cofactors for a variety of enzyme-catalyzed reactions. Specifically, tetrahydrofolate (in bacteria and eukaryotic organisms) and tetrahydromethanopterin (in archaea) mediate the transfer of one-carbon fragments, tetrahydrobiopterin (BH 4 ) 1 is implicated in the hydroxylation of aromatic amino acids and the formation of nitric oxide in animals (1,2), and molybdopterin is required as cofactor by a variety of redox enzymes, e.g. xanthine dehydrogenase. The metabolic roles of these cofactors have been reviewed repeatedly (3)(4)(5)(6).
The formation of pterins by ring expansion of guanosine, including an Amadori rearrangement of the ribose moiety, was first suggested by Weygand et al. (7) on basis of in vivo studies using 14 C-labeled precursors. Subsequent studies by Brown and Burg (8) and by Shiota et al. (9) showed that the first committed step in the biosynthesis of tetrahydrofolate and BH 4 is catalyzed by the enzyme GTP cyclohydrolase I. More specifically, C-8 of GTP (Fig. 1, compound 1) is released as formate, carbon atoms 1Ј and 2Ј of the ribose moiety are utilized for the formation of the dihydropyrazine ring, and carbon atoms 3Ј-5Ј of GTP afford the position 6 side chain of dihydroneopterin triphosphate (NH 2 TP) (Fig. 1, compound 2) (for a review, see Ref. 3).
The product of GTP cyclohydrolase I, NH 2 TP, is converted to BH 4 (compound 4) by the consecutive action of pyruvoyltetrahydropterin synthase (PPH 4 synthase) and sepiapterin reductase (10 -12). PPH 4 synthase catalyzes the elimination of triphosphate from NH 2 TP as well as a series of tautomerization reactions that are conducive to the formation of a tetrahydropterin from the dihydropterin substrate. Both carbonyl groups of the resulting pyruvoyltetrahydropterin (PPH 4 , compound 3) are subsequently reduced by the action of sepiapterin reductase.
The three-dimensional structures of GTP cyclohydrolase I from Escherichia coli (13,14), PPH 4 synthase from rat (15,16), and sepiapterin reductase (17) from mouse have been determined by x-ray crystallography. The folding patterns of GTP cyclohydrolase I and PPH 4 synthase are surprisingly similar in the absence of significant sequence similarity (13).
A hypothetical mechanism for GTP cyclohydrolase I proposed by Brown and Burg (8), Wolf and Brown (18), and Shiota et al. (9,19) implicates the hydrolytic opening of the imidazole ring of GTP followed by an Amadori rearrangement of the ribose moiety. Whereas the details of this complex reaction sequence are incompletely understood, it is immediately obvious that the reaction formally requires the removal of hydrogen from C-2Ј of GTP followed by the reintroduction of hydrogen, which becomes ultimately located at C-7 of the product. The experiments reported in this paper were designed in order to determine the details of the hydrogen transfer process and the stereochemical course of the reaction. 13 C 6 ]Glucose (99% enrichment) and D 2 O (99.9% enrichment) were purchased from Isotec (Miamisburg, OH) and Euriso-Top (Gif-sur-Yvette, France), respectively. D-Neopterin, L-sepiapterin, L-biopterin, and (6R)-5,6,7,8-tetrahydro-L-biopterin dihydrochloride were obtained from Schircks Laboratories (Jona, Switzerland). Alkaline phosphatase and lysozyme were from Boehringer Mannheim. Other enzymes were purchased from Sigma. All other reagents were of the highest available purity.

Materials-[U-
Preparation of [U- 13  . At intervals, aliquots of 100 mM guanine in 0.5 M NaOD (total volume, 740 l) and of [U-13 C 5 ]ribulose 5-phosphate (see above) (total volume, 1200 l) in D 2 O were added. Phosphoenolpyruvate was replenished, and the pH was adjusted by addition of concentrated HCl. The reaction was monitored by ion pair high pressure liquid chromatography using a column of Nucleosil RP18 (5 m, Macherey & Nagel, Dü ren, Germany; 4 ϫ 250 mm). The eluent contained 1% triethylammonium phosphate, pH 7.0, and 0.8% isopropanol (20). The effluent was monitored by a Hewlett-Packard HP1504A diode array detector. The reaction was terminated after 17 h. The pH of the solution was adjusted to 4.0 by the addition of concentrated HCl. Precipitate was removed by centrifugation. The supernatant was placed on a column of DEAE-cellulose DE 52 (20 ϫ 145 mm) that had been equilibrated with 30 mM LiCl, pH 3.5. The column was developed with a linear gradient of 30 -195 mM LiCl at a temperature of 4°C. The effluent was monitored photometrically at 280 nm. Fractions containing GTP were combined. The pH was adjusted to 6.5 by the addition of ammonium bicarbonate. The solution was lyophilized. The residue was extracted with dry methanol. The lithium salt of GTP was collected by centrifugation. The residue was dried under reduced pressure. The yield of GTP was 12.7 mg (23 mol, 33% based on glucose) as estimated photometrically (⑀ 252 nm 1 cm ϭ 13,700 M Ϫ1 ) (21). 4 Synthase-The open reading frame specifying human PPH 4 synthase was amplified by PCR using ATCC clone 79818 (22) as a template. The oligonucleotides PPH4 -1F (5Ј-GAGGAGAAATTAACTATGAGCACGGAAGGTGGTGG-CCGTCG-3Ј) and PPH4S-2Pst (5Ј-ATTATTCTGCAGTGCTAACCCCA-ATAGCTATTCTCC-3Ј) were used as primers. The resulting DNA fragment was reamplified using the oligonucleotides UniEcoRI (5Ј-CAAT-TTGAATTCATTAAAGAGGAGAAATTAACTATG-3Ј) and PPH4 -2Pst as primers. After restriction with EcoRI and PstI, the resulting PCR fragment was ligated into the vector pMAL-c2 (New England Biolabs, Schwalbach, Germany) yielding the plasmid pMAL-PPH4S, which specifies a fusion protein comprising the maltose-binding protein of E. coli and human PPH 4 synthase. The plasmid was transformed into E. coli strain XL-1 blue, yielding the recombinant strain E. coli XL-1 blue pMAL-PPH4S.

Construction of an Expression Plasmid for Sepiapterin Reductase-
Plasmid pNCO-SR (17) was digested with EcoRI and HindIII. The resulting 0.8-kilobase pair fragment was ligated into the vector pMAL-c2, which had been prepared with the same enzymes. The resulting plasmid pMAL-SR specifies a fusion protein comprising maltose-binding protein of E. coli and murine sepiapterin reductase. The plasmid was transformed into E. coli strain XL-1 blue by electroporation yielding the recombinant strain E. coli XL-1 blue pMAL-SR.
Purification of Recombinant GTP Cyclohydrolase I of E. coli-The recombinant E. coli strain M15 (pREP4, pECHI) (14) was grown aerobically as described in LB medium containing 150 mg of ampicillin and 22 mg of kanamycin per liter. At an A 600 of 0.7, isopropylthiogalactoside was added to a concentration of 1 mM. Cells were harvested by centrifugation after 18 h of incubation. Wet cell mass (8 g) was suspended in 100 ml of a solution containing 20 mM Tris hydrochloride, pH 7.5, 10 mM DTT, 1 mM phenylmethylsulfonyl fluoride, and 14 mg of lysozyme. The suspension was incubated at 37°C for 40 min and was then subjected to ultrasonic treatment and centrifuged. The supernatant was applied to a column of DEAE-cellulose (DE 52, Whatman, 40 ϫ 160 mm) that had been equilibrated with 20 mM Tris hydrochloride, pH 7.5. The column was washed with 300 ml of 20 mM Tris hydrochloride, pH 7.5, and was subsequently developed with a gradient of 0 -0.25 M NaCl in 20 mM Tris hydrochloride, pH 7.5. Fractions were combined. Alkaline phosphatase (0.14 mg/ml) was added, and the solution was incubated for 2 h at 37°C in order to hydrolyze phosphorylated pteridines. The solution was loaded on a column of DEAE-cellulose (20 ϫ 100 mm) that was developed as described above. Fractions were collected and concentrated by ultrafiltration. The protein showed a single band in SDS-polyacrylamide gel electrophoresis. The specific activity was 96 nmol min Ϫ1 mg Ϫ1 . Fractions were combined and concentrated. The concentration of GTP cyclohydrolase I was monitored photometrically using an absorbance coefficient of ⑀ 280 nm 1 cm ϭ 0.33 ml mg Ϫ1 . The enzyme was stored at 4°C in 10 mM potassium phosphate, pH 7.5, containing 1 mM DTT.
Purification of Fusion Proteins-The recombinant fusion proteins consisting of maltose-binding protein and PPH 4 synthase or sepiapterin reductase were purified by the following procedure. The respective E. coli strains were grown in LB medium containing ampicillin (170 mg per liter). At an A 600 of 0.6, isopropylthiogalactoside was added to a concentration of 1 mM. After 5 h, the cells were harvested by centrifugation. Cells were homogenized by ultrasonic treatment in 50 mM Tris hydrochloride, pH 8.0. The suspension was centrifuged. The crude extract was applied to an amylose resin column (New England Biolabs; 20 ϫ 30 mm). The column was washed with 50 mM Tris hydrochloride, pH 8.0, and was then developed with 50 mM Tris hydrochloride, pH 8.0, containing 100 mM maltose.
Enzyme Assays-Assay mixtures for the determination of GTP cyclohydrolase I activity contained 100 mM Tris hydrochloride, pH 8. an excess of sepiapterin reductase (2 mg per ml), and protein in a total volume of 300 l. The solutions were incubated at 37°C. The reaction was terminated by the addition of 50 l of 1 M hydrochloric acid containing 1% iodine and 2% potassium iodide. Biopterin was determined by reversed phase high pressure liquid chromatography using a column of Nucleosil 10-C18 (Macherey & Nagel; 4 ϫ 250 mm). The eluent contained 3% methanol, 2% acetonitrile, 0.4% phosphoric acid, and water in a total volume of 1 liter. The flow rate was 1 ml min Ϫ1 . The effluent was monitored fluorometrically (excitation, 350 nm; emission, 450 nm). The retention time of biopterin was 8 min (24).
Assay mixtures for the determination of sepiapterin reductase activity contained 50 mM Tris hydrochloride, pH 8.0, 100 M NADPH, 40 M sepiapterin, and protein in a total volume of 300 l. The reaction mixture was incubated at 37°C. The reaction was terminated as described above. Biopterin was determined by high pressure liquid chromatography (see above). One unit of each of the enzymes under study catalyzes the formation of 1 nmol of product/min at 37°C.
Enzymatic Formation of Isotope-labeled Pteridines-For experiments to be performed in D 2 O, all reaction components were lyophilized repeatedly and subsequently dissolved in 99.9% D 2 O. Enzyme reactions were performed as follows. The reaction mixtures were subjected to NMR analysis without further processing. Pyruvoyl-5,6,7,8-tetrahydropterin-To a solution of NH 2 TP prepared as described above, a solution (100 l) containing 60 mM MgCl 2 , 10 mM potassium phosphate, pH 7.5, 1 mM DTT, and 27 units of PPH 4 synthase (fusion protein) was added anaerobically by injection through the rubber septum. The reaction mixture was incubated for 1 h at 37°C.
NMR Spectroscopy-NMR measurements were performed at 27°C using a Bruker DRX 500 spectrometer operating at 500.13 and 125.7 MHz for 1 H and 13 C NMR experiments, respectively. The spectrometer was equipped with a lock-switch unit for 2 H-decoupling experiments using the lock channel. D 2 O (100 l) was added anaerobically to reaction mixtures containing H 2 O as solvent. The NMR tubes were flushed with argon for 10 min. The solution was injected into the NMR tube, and the tube was sealed with an Omni-Fit sample tube valve (Wilmad, Buena, NY).
One-dimensional 1 H and 13 C NMR experiments and two-dimensional double-quantum-filtered COSY, NOESY, HMQC, and HMQC-TOCSY experiments were performed with standard Bruker software (XWIN-NMR). 13 C NMR spectra were obtained using a dual 13 C/ 1 H probehead, and proton-detection experiments were obtained using a 1 H/ 13 C/ 15 N inverse triple resonance probehead. Simultaneous 1 H and 2 H decoupling of 13 C was achieved with a WALTZ16 pulse sequence during relaxation ( 1 H) and acquisition ( 1 H, 2 H). 2 H decoupling was via the lock channel using a lock-switch unit (Bruker). Prior to Fourier transformation, the free induction decay was multiplied with a Gaussian function. NMR spectra were simulated using the NMR-SIM software package from Bruker.

RESULTS
Experiments were designed to study the transfer of deuterium in the enzyme reaction. At the outset of the experiments, we assumed tentatively that the putative Amadori rearrangement might proceed by reincorporation of a proton released from C-2Ј into the enzyme product. Based on this working hypothesis (which turned out to be incorrect), we decided to conduct the experiments with GTP labeled with deuterium in the 2Ј position. In order to increase the selectivity and sensitivity of 13 C NMR observations, it also appeared advantageous to label the ribose side chain of substrate GTP with 13 C. Consequently, [1Ј,2Ј,3Ј,4Ј,5Ј-13 C 5 ,2Ј-2 H 1 ]GTP (Fig. 2, compound 1a) was prepared from [U-13 C 6 ]glucose as described under "Experimental Procedures." Briefly, [U-13 C 6 ]glucose was converted to [U-13 C 5 ]ribulose 5-phosphate in a one-pot reaction using hexokinase, glucose-6-phosphate dehydrogenase, and 6-phosphogluconate dehydrogenase. NADPH generated by the reaction sequence was dehydrogenated using glutamate dehydrogen-    ase, with ␣-ketoglutarate serving as hydride acceptor (25).
[U-13 C 5 ]Ribulose phosphate was converted to GTP in a one-pot reaction performed in D 2 O as solvent. The reaction mixtures contained all enzymes required for generation of GTP. Pentose phosphate isomerase was added in a large excess (26). The phosphate donor was phosphoenolpyruvate. Under these experimental conditions, deuterium from solvent was incorporated into the 2Ј position of the ribose moiety with an enrichment above 90%. The labeling pattern of the multiply labeled GTP sample was confirmed by multinuclear NMR experiments ( Table I). The presence of 2 H at the 2Ј position was confirmed by the large upfield 13 C shift of C-2Ј and the smaller upfield shifts of C-1Ј and C-3Ј. All carbon atoms of the ribose moiety showed the expected 13 C 13 C coupling patterns (data not shown). The labeled GTP was converted to NH 2 TP by the catalytic action of GTP cyclohydrolase I in H 2 O or D 2 O solution. In the experiment using H 2 O as solvent, the 13 C signal of C-7 of NH 2 TP appears as a doublet of doublets due to 13 C 13 C coupling via one bond (to C-6) and via two bonds (to C-1Ј) (Fig. 3A). This signal pattern was not affected by 2 H decoupling. It follows that the enzyme product had the structure 2a (Fig. 2).
In a similar experiment with D 2 O as solvent, the { 1 H}-decoupled 13 C signal of C-7 appeared as a broad hump (Fig. 3B). Simultaneous { 1 H, 2 H}-decoupling revealed a double doublet pattern for C-7 that differed from that in Fig. 3A by an upfield shift of 0.30 ppm. The signal of C-6 was also offset to high field by 0.06 ppm. The difference between the spectra obtained in the two experiments indicates that, contrary to our initial hypothesis, the reaction in water resulted in the virtually complete loss of the deuterium bound at C-2Ј of the GTP substrate, whereas the product obtained in D 2 O as solvent contained deuterium in the position 7 methylene group of NH 2 TP. In summary, the structure of the enzyme product obtained in D 2 O is described by the structure 2b (Fig. 2).
It follows that the formation of NH 2 TP (compound 2) involves the efficacious exchange of a proton released from C-2Ј of substrate GTP with solvent water. For the analysis of the stereochemical course of the H addition step, it was therefore in order to conduct experiments in D 2 O, irrespective of the presence or absence of deuterium in the 2Ј position of substrate GTP.
The chemical shifts of the diastereotopic H-atoms at C-7 of NH 2 TP (compound 2) are degenerate, because the effect of the chirality in C-1Ј and C-2Ј does not extend far enough to cause any significant chiral induction at the C-7 methylene group (data not shown). In order to determine the configuration of the CHD group in position 7 of the pteridine obtained in D 2 O as solvent, it was therefore necessary to generate a more closely adjacent chiral center. This was achieved by enzymatic conversion of NH 2 TP to pyruvoyltetrahydropterin (PPH 4 , compound 3) or BH 4 (compound 4). PPH 4 was obtained from [1Ј,2Ј,3Ј,4Ј,5Ј-13 C 5 , 2Ј-2 H 1 ]GTP in D 2 O as solvent by the consecutive action of GTP cyclohydrolase and PPH 4 synthase in D 2 O as solvent. 13 C NMR signals of the product are shown in Fig. 4. In the { 1 H, 2 H}-decoupled spectrum (Fig. 4A), all carbon atoms appear as multiplets due to 13 C 13 C coupling. On basis of chemical shift and coupling arguments, all signals can be unequivocally assigned. Notably, the signal of C-1Ј appears at 96.3 ppm, thus suggesting that the carbonyl group is hydrated. On the other hand, C-2Ј resonates at 210.5 ppm and is apparently not hydrated to an appreciable degree. In the absence of 2 H-decoupling (Fig. 4B), the signal of C-7 appears as a doublet of triplets indicating 13 C 2 H coupling of 22 Hz. Simultaneous { 1 H, 2 H}-decoupling simplifies this multiplet to a doublet (Fig. 4A). These findings again confirm the presence of exactly one deuterium atom at C-7.
The 3Ј methyl carbon appears as a quartet in the { 1 H, 2 H}decoupled spectrum and as a pattern of 10 lines in the { 1 H}decoupled spectrum. Closer analysis shows that the latter pattern is a quartet of triplets arising by 2 H coupling. It follows that the methyl group contains exactly one deuterium atom. The signals of the 1Ј and 2Ј carbon atoms are not affected by { 2 H} decoupling, as they do not carry directly bound deuterium. The signal of C-6 is only resolved by simultaneous { 1 H, 2 H}- decoupling, indicating the presence of a directly bound deuterium atom. In summary, the molecular species can be described by structure 3a (Fig. 2).
As a consequence of CH acidity, the protons at C-3Ј of PPH 4 (compound 3) are subjected to further hydrogen exchange with solvent at prolonged (10 h at 30°C) incubation. Samples that were kept for extended time periods after completion of the enzyme reaction showed spectroscopic evidence of a molecular species carrying two deuterium atoms at the 3Ј methyl group (0.23 ppm upfield shifted, but otherwise unchanged doublet of doublets in a { 1 H, 2 H}-decoupled 13 C NMR spectrum obtained after 10 h at 30°C; data not shown).
In a subsequent experiment, BH 4 (compound 4) was obtained from [1Ј,2Ј,3Ј,4Ј,5Ј-13 C 5 ,2Ј-2 H 1 ]GTP in D 2 O as solvent by the consecutive action of GTP cyclohydrolase I, PPH 4 synthase, and sepiapterin reductase (Fig. 4, C and D). { 1 H, 2 H}-Decoupling (Fig. 4C) did not affect the signals of C-1Ј and C-2Ј, thus indicating that these carbon atoms did not carry directly bound deuterium. On the other hand, it was obvious from comparison of the { 1 H}-decoupled spectrum (Fig. 4D) with the { 1 H, 2 H}decoupled spectrum (Fig. 4C) that each of the carbon atoms 6, 7, and 3Ј carried one directly bound deuterium atom. The biopterin species obtained in this experiment is represented by structure 4a (Fig. 2).
Based on these results, we can now address the configuration of the C-7 methylene groups in the isotopomers 3b and 4a. Two different strategies were used for the analysis, (i) one-dimensional 1 H NMR analysis of PPH 4 obtained from unlabeled GTP (i.e. Fig. 2, compound 3b), and (ii) 1 H 13 C correlation spectroscopy with BH 4 obtained from [1Ј,2Ј,3Ј,4Ј,5Ј-13 C 5 ,2Ј-2 H 1 ]GTP (i.e. Fig. 2, compound 4a). 1 H NMR signals of PPH 4 formed from unlabeled GTP in H 2 O as solvent by the consecutive action of GTP cyclohydrolase and PPH 4 synthase are shown in Fig. 5C. Signals could be assigned by analysis of the coupling pattern in the one-dimensional spectrum and by two-dimensional DQF-COSY (not shown) and NOESY experiments (see Fig. 7A). The prochiral C-7 protons of PPH 4 devoid of isotope labels afford two double doublets centered at approximately 3.22 and 3.45 ppm (Fig. 5C). The double doublet at 3.22 ppm is contaminated by three lines (marked by asterisks), which were contributed by an unidentified component of the reaction mixture. The hydrogen at C-6 appears as a double doublet arising by coupling to both C-7 protons. A simulation of the nonlinear three-spin coupling pattern involving the geminal protons at C-7 and the proton at C-6 of compound 3 is shown in Fig. 5D.
The { 2 H}-decoupled 1 H NMR spectrum of PPH 4 obtained in D 2 O from unlabeled GTP (i.e. isotopomer 3b) is shown in Fig.  5A. The downfield C-7 proton appears as a singlet at 3.43 ppm. The signals of the proton at C-6 and the upfield geminal proton at C-7 are both absent, thus indicating the incorporation of 2 H into the respective positions (the signals marked by asterisks in Fig. 5A represent the unknown contaminant mentioned above). The upfield shift of the singlet at 3.43 ppm relative to the double doublet pattern in Fig. 5C is caused by the ␣ and ␤ shift contributions of the deuterium atoms present at C-6 and C-7 of PPH 4 in isotopomer 3b (see Fig. 4, A and B, for analysis of 2 H substitution). The data indicate that deuterium incorporated from solvent D 2 O by GTP cyclohydrolase I is specifically contributed to the position 7 methylene proton resonating at higher field (3.22 ppm) in the enzyme product, PPH 4 . As shown below, this signifies that deuterium is incorporated into the pro-7R position.
Without { 2 H}-decoupling, the 1 H signal of the 3Ј methyl group appears as a triplet with 2.2 Hz spacing in Fig. 5B. { 2 H}-Decoupling converts this triplet to a singlet with an up-field shift of 17 ppb as compared with PPH 4 obtained in H 2 O as solvent (Fig. 5C). In agreement with the 13 C observations described above, this indicates the presence of exactly one 2 H atom in the methyl group of PPH 4 obtained in D 2 O as solvent. However, it should be noted that prolonged incubation in D 2 O has been shown to result in additional wash-in of deuterium from solvent (see above).
Thus, C-7 shows correlations to the directly attached diastereotopic H atoms as well as to H-6. C-6 shows strong correlation signals to the protons at C-1Ј and C-7 and a weak correlation signal extending to the proton at C-2Ј. C-1Ј shows correlation to the directly attached proton, the proton at C-6, and the proton on C-2Ј.
In the corresponding spectra of BH 4 obtained from [1Ј,2Ј,3Ј,4Ј,5Ј-13 C 5 ,2Ј-2 H 1 ]GTP (Compound 1a) in D 2 O solution (i.e. the BH 4 isotopomer 4a, Fig. 2), several 1 H 13 C correlation signals are absent. Specifically, all correlation signals to the upfield C-7 methylene proton and the C-6 proton are absent. This signifies that the upfield position 7 methylene proton and the proton at C-6 have been replaced by deuterium. This finding is well in line with the observations, obtained by a different experimental technique, in the case of PPH 4 (see above). It should be noted that Fig. 6, B and D, shows cross-peaks that do not match either the 13 C or 1 H signal positions of BH 4 . The occurrence of these signals is not surprising because the enzymatic reaction mixtures were subjected to NMR analysis without isolation and purification of the chemically labile BH 4 and contain all of the numerous ingredients specified under "Experimental Procedures." It is a specific advantage that these signals are clearly separated from the BH 4 signals in the twodimensional matrix, in contrast to the signal overlap (Fig. 5, asterisks) occurring in the one-dimensional spectra.
For the interpretation of data in stereochemical terms, the diastereomeric assignment of the geminal C-7 protons remains to be established. Expansion of the relevant portions from NOESY experiments (mixing time, 1 s) with unlabeled PPH 4 and BH 4 are shown in Fig. 7. As expected, the strongest NOE transfer was observed between the proximal geminal protons at C-7. In both spectra in Fig. 7, considerably stronger NOE interactions were found between H-6 and the respective downfield proton at C-7 as compared with the NOE between H-6 and the upfield methylene proton. This indicates that in PPH 4 , as well as in BH 4 , the distances between H-6 and the downfield proton at C-7 are substantially shorter than the distance between H-6 and the upfield proton.
Irrespective of the conformational equilibrium of (6R)-5,6,7,8-tetrahydropterins with a side chain in position 6, as a consequence of ring strain, the pro-S proton at C-7 will be closer to H-6 that the pro-R proton. This can be easily demonstrated with Newman projections of this model compound (Fig.  8). Therefore, the downfield protons at C-7 of PPH 4 and BH 4 can be assigned as pro-7S (natural PPH 4 must be the 6R enantiomer, because in both compounds, the proton introduced from solvent by GTP cyclohydrolase I occupies the upfield position).
The average conformations of PPH 4 and BH 4 can be estimated from vicinal coupling constants between the methylene group protons and H-6. In both compounds, small coupling constants of 3.1 and 3.2 Hz, respectively, were found between the protons at positions pro-7S and 6, indicating an estimated dihedral angle of about 55° (27). Similar values were found by Armarego et al. (28) for BH 4 hydrochloride. The corresponding vicinal coupling constants for the protons at positions pro-7R and 6 were 6.6 and 7.5 Hz, respectively. This suggests a conformational equilibrium of pseudo-equatorial and pseudo-axial conformation of the side chain in position 6 ( Fig. 8), in agreement with a molecular dynamics simulation for BH 4 (29,30). DISCUSSION Simon et al. (31) and Brown and co-workers (8,18) had proposed an Amadori rearrangement for the remodeling of the ribose moiety by GTP cyclohydrolase, but no direct evidence had been obtained. The experiments reported in this paper show that the removal of hydrogen from C-2Ј of GTP catalyzed by GTP cyclohydrolase and the introduction of hydrogen conducive to the formation of the position 7 methylene group of NH 2 TP occurs by exchange with the bulk solvent as shown by deuterium washout from [1Ј,2Ј,3Ј,4Ј,5Ј-13 C 5 ,2Ј-2 H]GTP occurring in H 2 O as solvent, and by deuterium incorporation with unlabeled GTP as substrate (Fig. 9)  vent. It follows that the exchanged species is a proton as opposed to a hydride ion.
Palm and Simon (32, 33) and Simon and Kraus (34) studied the Amadori rearrangement of a glycoside formed from [2-3 H]glucose and p-toluidine. The tritium label was washed out during the course of the acid-catalyzed reaction. The same experiment with natural abundance glucose and tritiated water yielded Amadori product with tritium in position 1. Our findings are consistent with an Amadori rearrangement of the ribose moiety as an essential part of the GTP cyclohydrolasecatalyzed reaction. Specifically, the hypothetical intermediate 5 obtained by the hydrolytic release of formate from the imidazole ring of GTP could form an iminium ion by opening of the furanose ring (Fig. 10). A proton could then be abstracted from the activated position 2Ј methine group under formation of an enol-type intermediate. Reprotonation of the enol in the position 1Ј could then yield the 2Ј-keto-3Ј,4Ј,5Ј-trihydroxypentyltype intermediate 6. Formation of a cyclic imine could occur as the final reaction step. López et al. (35) reported a nonenzymatical Amadori rearrangement of a 6-amino-5-ribosylaminopyrimidine under formation of a pteridine ring system.
The formation of the position 7 methylene group occurs by stereospecific protonation of the carbon atom equivalent to C-1Ј of the enzyme substrate, GTP. The proton introduced from the solvent occupies the pro-7R position in the enzyme product, NH 2 TP (compound 2).
As a consequence of the utilization of PPH 4 synthase as an auxiliary enzyme, our data also show that protons are introduced from solvent into positions 6 and 3Ј of PPH 4 by the action of this enzyme. This is well in line with earlier observations. Thus, Ghisla et al. (36,37) had observed 2 H incorporation from D 2 O in positions C-6 and C-3Ј of PPH 4 by 1 H NMR experiments. On the other hand, Le Van et al. (38) found that 3 H is not transferred to BH 4 from GTP labeled with 3 H in position 3Ј or 4Ј, thus indicating that the 3Ј and 4Ј hydrogen atoms of GTP had been washed out by the enzyme-catalyzed conversion of GTP to BH 4 . Both studies indicated that the hydrogen atoms introduced at C-6 and C-3Ј had to be introduced from solvent.
The present study also shows that the position 3Ј methyl group of PPH 4 undergoes slow, spontaneous (i.e. nonenzymatic) proton exchange with the solvent. This is not surprising in light of the methylketone structure of PPH 4 .
The position 1Ј carbonyl group of PPH 4 is hydrated to more than 90%. On the other hand, the position 2Ј carbonyl group shows no evidence of hydration. Hydration is a common finding in 1,2-diketones, e.g. butanedione. However, it is not immediately obvious why the hydration appears to involve exclusively the 1Ј carbonyl group of PPH 4 .
The experiments also show that the reaction catalyzed by sepiapterin reductase does not involve deuterium incorporation from the solvent into the C-1Ј and C-2Ј positions of BH 4 under nonequilibrium conditions. Sepiapterin reductase can catalyze the reduction of both carbonyl groups of PPH 4 (39). On basis of biochemical and crystallographic data (17,40,41), it has been proposed that the position 1Ј carbonyl group is reduced first, and that the enzyme subsequently catalyzes an isomerization via an enediol intermediate conducive to shifting of the carbonyl group from the 2Ј position in 6-(1Ј-hydroxy-2Ј-oxopropyl)tetrahydropterin (Fig. 11, compound 7) to the 1Ј position in 6-lactoyltetrahydropterin. This rearrangement has been proposed to preceed the reduction of the second carbonyl group. If the rearrangement reaction indeed plays a role in the overall reaction mechanism, it must occur without significant hydrogen exchange with the bulk solvent, because no deuterium incorporation from D 2 O as solvent has been observed.