Purification and cloning of the GTP cyclohydrolase I feedback regulatory protein, GFRP.

The activity of GTP cyclohydrolase I, the initial enzyme of the de novo pathway for biosynthesis of tetrahydrobiopterin, the cofactor required for aromatic amino acid hydroxylations and nitric oxide synthesis, is sensitive to end-product feedback inhibition by tetrahydrobiopterin. This inhibition by tetrahydrobiopterin is mediated by the GTP cyclohydrolase I feedback regulatory protein GFRP, previously named p35 (Harada, T., Kagamiyama, H., and Hatakeyama, K. (1993) Science 260, 1507-1510), and L-phenylalanine specifically reverses the tetrahydrobiopterin-dependent inhibition. As a first step in the investigation of the physiological role of this unique mechanism of regulation, a convenient procedure has been developed to co-purify to homogeneity both GTP cyclohydrolase I and GFRP from rat liver. GTP cyclohydrolase I and GFRP exist in a complex which can be bound to a GTP-affinity column from which GTP cyclohydrolase I and GFRP are separately and selectively eluted. GFRP is dissociated from the GTP agarose-bound complex with 0.2 M NaCl, a concentration of salt which also effectively blocks the tetrahydrobiopterin-dependent inhibitory activity of GFRP. GTP cyclohydrolase I is then eluted from the GTP-agarose column with GTP. Both GFRP and GTP cyclohydrolase I were then purified separately to near homogeneity by sequential high performance anion exchange and gel filtration chromatography. GFRP was found to have a native molecular mass of 20 kDa and consist of a homodimer of 9.5-kDa subunits. Based on peptide sequences obtained from purified GFRP, oligonucleotides were synthesized and used to clone a cDNA from a rat liver cDNA library by polymerase chain reaction-based methods. The cDNA contained an open reading frame that encoded a novel protein of 84 amino acids (calculated molecular mass 9665 daltons). This protein when expressed in Escherichia coli as a thioredoxin fusion protein had tetrahydrobiopterin-dependent GTP cyclohydrolase I inhibitory activity. Northern blot analysis indicated the presence of an 0.8-kilobase GFRP mRNA in most rat tissues, the amounts generally correlating with levels of GTP cyclohydrolase I and tetrahydrobiopterin. Thus, mRNA levels were relatively high in liver and kidney and somewhat lower in testis, heart, brain, and lung. These results suggest that GFRP is widely expressed and may play a role in regulating not only phenylalanine metabolism in the liver, but also the production of biogenic amine neurotransmitters as well as nitric oxide synthesis.

expressed and may play a role in regulating not only phenylalanine metabolism in the liver, but also the production of biogenic amine neurotransmitters as well as nitric oxide synthesis.
6R-L-erythro-5,6,7,8-Tetrahydrobiopterin (BH 4 ) 1 is the required cofactor for the members of the family of aromatic amino acid hydroxylases (1) and also for the three different isoforms of nitric-oxide synthase (2). The de novo pathway for biosynthesis has been well characterized (3), yet little is known of the regulation of this pathway in vivo and how this regulation impinges on the activities of the enzymes for which it is required. Studies of acute and chronic regulation of BH 4 levels in vivo have generally focused on the first enzyme of the BH 4 biosynthetic pathway, GTP cyclohydrolase I, since changes in its activity usually mediate or closely correlate with changes in BH 4 levels (3). Genetic deficiencies of GTP cyclohydrolase I activity have been described which result in hyperphenylalaninemia due to the lack of BH 4 (4). Recently, a hereditary form of progressive dystonia was shown to be caused by a mutation in the GTP cyclohydrolase I gene resulting in decreased dopamine synthesis in the central nervous system (5). In the adrenal medulla and cortex, catecholamine synthesis as well as BH 4 levels and GTP cyclohydrolase I activity have been shown to be hormonally regulated (6), probably through the action of cAMP-mediated pathways leading to increased protein synthesis (7). Recently, reserpine, which had previously been found to increase BH 4 levels and GTP cyclohydrolase I activity in the adrenal cortex (8), was shown to increase levels of GTP cyclohydrolase I mRNA in peripheral and central neurons (9). Activation of the immune system results in the production of interferon-␥ which then stimulates an increase in GTP cyclohydrolase I activity in macrophages resulting in increased extracellular dihydroneopterin levels in humans and intracellular BH 4 levels and BH 4 -dependent nitric oxide synthesis in rodents (10). GTP cyclohydrolase I mRNA and BH 4 synthesis are also induced by interferon-␥ in lymphocytes (11), by interleukin-1␤ in smooth muscle cells (12), and by cytokines in endothelial cells where it regulates the activity of the endothelial form of nitric-oxide synthase (13). Administration of bacterial endotoxin (lipopolysaccharide) to rats has been shown to increase GTP cyclohydrolase I activity and BH 4 levels in cerebellum, liver, spleen, and adrenal gland (14).
The mechanism of two other types of acute regulation of BH 4 biosynthesis, phenylalanine-induced increases in BH 4 levels * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) U53710.
§ To whom correspondence should be addressed: Laboratory of Cell Biology, 36 Convent Dr., Bldg. 36, Rm. 3A-17, National Institute of Mental Health, Bethesda, MD 20892. Tel: 301-402-4897; Fax: 301-480-9284; E-mail: milstien@codon.nih.gov. and end-product feedback regulation by BH 4 , have recently been clarified. In 1975, Leeming et al. (15) discovered that plasma BH 4 levels in normal children increased rapidly following a phenylalanine challenge but not in children who had a defect in BH 4 biosynthesis, suggesting that phenylalanine might stimulate BH 4 biosynthesis. This possibility was strengthened by the finding that administration of phenylalanine to rats, together with labeled guanosine to label GTP pools, stimulated the incorporation of label into BH 4 (16). However, in the ensuing years, the mechanism of the phenylalanine effect remained enigmatic since phenylalanine does not stimulate GTP cyclohydrolase I activity in vitro nor does it have any effects on the other enzymes of the de novo pathway. 2 In what was considered to be an unrelated type of regulation, it has long been known that GTP cyclohydrolase I activity was negatively regulated by the end product of the pathway, BH 4 , and it was assumed that this was an intrinsic property of the enzyme (17). A link between the phenylalanine stimulatory effect and the BH 4 inhibitory effect was established recently when Harada et al. (18) found that pure, recombinant GTP cyclohydrolase I was not inhibited by BH 4 and that another protein, called p35 because of its apparent molecular mass, was present in rat liver extracts and conferred BH 4 -dependent inhibitory sensitivity to GTP cyclohydrolase. Furthermore, it was also found that the BH 4 -dependent inhibitory activity could be reversed by L-phenylalanine (18). Thus, this putative auxiliary protein, when complexed to GTP cyclohydrolase I, constitutes a unique regulatory system allowing hepatic phenylalanine hydroxylation to increase in response to dietary phenylalanine by increasing synthesis of the rate-limiting cofactor and turning off BH 4 synthesis when it is present in higher than needed concentrations (Fig. 1).
In this report, a convenient method is described to co-purify GTP cyclohydrolase I and its associated feedback regulatory protein to homogeneity which facilitated the molecular cloning of the regulatory protein. Our results suggest that this regulatory protein may play an important role not only in regulating phenylalanine metabolism, but also in the regulation of catecholamine and serotonin synthesis, as well as nitric oxide production. It is suggested that this protein should more properly be named GTP cyclohydrolase I feedback regulatory protein, or GFRP, rather than p35.

Materials
GTP, GTP immobilized on 4% agarose by coupling periodate oxidized GTP to an adipic acid dihydrazide linker, amino acids, and other common biochemicals were obtained from Sigma. Intestinal alkaline phosphatase was from Calbiochem. Pterins were all obtained from B. Schircks Lab (Jona, Switzerland).

Enzyme Assays
GTP Cyclohydrolase I Activity-Two different methods were used to determine GTP cyclohydrolase I activity which gave identical results for purified GTP cyclohydrolase I preparations. For cruder material and in some experiments with purified enzyme, activity was determined by measurement of neopterin produced after oxidation and dephosphorylation of the enzymatic product, dihydroneopterin triphosphate, by modifications of previously described methods (19). In brief, samples to be assayed were buffer-exchanged into buffer A (50 mM KPO 4 (pH 7.4), 1 mM EDTA, 1 mM DTT, 0.1% Tween 20, 10% glycerol) by the spin column method through Sephadex G25-M, and aliquots were incubated at 37°C for various times (usually 15 min) in a total volume of 50 l containing 0.5 mM Tris-HCl (pH 7.4), 1 mM DTT, 50 g of BSA, and 1 mM GTP. Reactions were terminated and the product was oxidized by the addition of 5 l of iodine solution (1% I 2 , 2% KI in 1 M HCl, stored at 4°C in a brown bottle). After 20 min at room temperature, excess iodine was reduced by addition of 5 l of 2% ascorbic acid, and the pH was made alkaline with 8 l of 1 N NaOH. The samples were then mixed, centrifuged briefly, and dephosphorylated by addition of 1 l of intestinal alkaline phosphatase (10 units) and incubation for 20 min at 37°C. Finally, 100 l of 1 M H 3 PO 4 were added, and samples were clarified by centrifuging at 10,000 ϫ g for 2 min in a microcentrifuge. Neopterin was measured by reverse phase HPLC with fluorometric detection as described previously (20).
For more highly purified preparations, GTP cyclohydrolase I activity was determined by measurement of neopterin triphosphate. This method could not be used for less purified samples due to the lower sensitivity of fluorescence detection in the required HPLC solvent systems and to the presence of contaminating phosphatases. In brief, assays were carried out exactly as described above except that reactions were terminated by addition of 100 l of 1 M H 3 PO 4 , and dihydroneopterin triphosphate was oxidized to neopterin triphosphate by adding 10 mg of MnO 2 and incubating for 10 min at room temperature. After centrifugation at 10,000 ϫ g for 2 min, supernatants were stored on ice until analyzed. Preliminary experiments indicated that samples could be kept on ice for up to 2 h without significant dephosphorylation occurring. Neopterin triphosphate was measured by anion exchange HPLC (Whatman SCX, 0.4 M potassium monophosphate, 0.8 M KCl at 2 ml/min) with fluorometric detection as described by Zagalak et al. (21). Neopterin triphosphate was eluted at 12 min under these conditions and was well separated from other unconjugated pterins. Neopterin was used to construct standard curves.
GFRP Activity-To determine BH 4 -dependent GTP cyclohydrolase I inhibitory activity, GTP cyclohydrolase I activity was measured exactly as described above in the presence of 10 M BH 4 . In all assays of samples in which GFRP was completely separated from GTP cyclohydrolase I (see below), enough purified GTP cyclohydrolase I was added to generate 20 pmol of product per min. One unit of inhibitory activity corresponded to a BH 4 -dependent decrease of GTP cyclohydrolase I activity of 1 pmol/min of dihydroneopterin triphosphate. When GTP cyclohydrolase I activity was determined by the conversion of dihydroneopterin triphosphate to neopterin as described above, it was necessary to wash the reverse phase HPLC column for an additional 10 min to elute biopterin prior to injection of the next sample. Addition of biopterin (or other pterins) during the alkaline phosphatase step had no significant effects on the yield of product. Assays were routinely carried out with two different concentrations of the material being assayed to ensure that the inhibitory response was proportional to the amount of GFRP added.  , in a triphosphate elimination-intramolecular reduction reaction catalyzed by 6-pyruvoyltetrahydropterin synthase (6PPH 4 synthase). Tetrahydrobiopterin is then formed by reduction of the side chain carbonyl groups of 6-pyruvoyltetrahydropterin synthase in NADPH-dependent reactions by the enzyme sepiapterin reductase or by the combined action of 6PPH 4 reductase and sepiapterin reductase (not shown). When GTP cyclohydrolase I is complexed with GFRP, GTP cyclohydrolase I activity is inhibited by BH 4 (Ϫ). Phenylalanine, which is hydroxylated to tyrosine in a BH 4 -dependent reaction catalyzed by phenylalanine hydroxylase (PAH), specifically reverses the inhibitory effect of BH 4 (ϩ), while having no direct effects in the absence of BH 4 .

Other Assays
Protein was determined by the Coomassie dye binding assay (Pierce) with BSA as a standard.

Peptide Analysis
An aliquot of purified GFRP (8 g) was dried in vacuo, redissolved in 50 l of 8 M urea in 0.4 M ammonium bicarbonate, and then subjected to reduction, alkylation, and proteolytic digestion with 1.5 g of modified trypsin (Promega) (22). The peptides in the resulting digest were fractionated by reverse-phase HPLC on a Vydac 218TP52 column (Separations Group, 2.1 ϫ 250 mm) at 35°C using the gradient described by Fernandez et al. (23) on a System Gold HPLC equipped with a Model 507 autosampler, Model 126 programmable solvent module, and Model 168 diode array detector (Beckman). Solvent A was 0.1% trifluoroacetic acid in water, and solvent B was 0.1% trifluoroacetic acid in acetonitrile. Column effluent was monitored at 215 and 280 nm, and fractions were collected at 30-s intervals and stored at Ϫ70°C. Fractions (125 l) containing tryptic peptides were applied in 30-l aliquots to a Biobrene (Applied Biosystems Inc.) treated glass fiber filter and dried prior to amino acid sequencing on a Model 477A pulsed-liquid protein Sequencer equipped with a Model 120A PTH analyzer (ABI) using methods and cycles supplied by the manufacturer. Data were collected and analyzed on a Model 610A data analysis system (Applied Biosystems Inc.).

Co-purification of GTP Cyclohydrolase I and GFRP
All procedures were carried out at room temperature unless indicated otherwise.
Extract-Frozen rat livers from male Sprague-Dawley rats (300 g, Harlan Bioproducts for Science) were suspended in 900 ml of extraction buffer B (50 mM KPO 4 , pH 7.4, 1 mM DTT, 1 mM EDTA, 2 g/ml each aprotinin, leupeptin, pepstatin, 0.2 mM phenylmethylsulfonyl fluoride, 0.1% Tween 20; 10% glycerol) in a beaker placed in an ice bath. After 30 min, livers were minced with a scissors and homogenized in a Waring blendor for 2 min at medium speed and 2 min at high speed. The homogenate was centrifuged at 20,000 ϫ g for 60 min, the supernatant was carefully decanted, and the loose pellet was discarded.
Heat Treatment-The extract was heated to 65°C in a boiling water bath in 200-ml portions while stirring with a thermometer. The heattreated extract was rapidly cooled in an ice bath to Ͻ20°C and then centrifuged at 20,000 ϫ g for 30 min. The large pellet was briefly extracted by stirring with 200 ml of buffer B, centrifuged for 15 min at 20,000 ϫ g, and the supernatants from both centrifugations were combined.
Acid Precipitation-The heat-treated supernatant was placed in a beaker in an ice bath with a stirring bar and a pH electrode. The pH was adjusted to 5.0 by dropwise addition of glacial acetic acid. After stirring for an additional 30 min, the sample was centrifuged for 30 min at 20,000 ϫ g, and the supernatant was discarded. The dark brown pellet was rapidly resuspended in 50 ml of cold buffer B by homogenizing in a glass Dounce homogenizer.
Ammonium Sulfate Precipitation-While stirring in an ice bath, finely powdered ammonium sulfate (0.134 g/ml) was added to the redissolved acid precipitate in small portions over 10 min. After stirring for an additional 20 min, the sample was centrifuged at 20,000 ϫ g for 20 min, and the pellet was discarded. An additional portion of ammonium sulfate (0.179 g/ml) was then added to the supernatant as above to precipitate GTP cyclohydrolase I and the associated GFRP activity. After centrifugation for 20 min at 20,000 ϫ g, the pellet was dissolved in 10 ml of buffer B.
Dialysis-The resuspended ammonium sulfate pellet was dialyzed overnight at 4°C against 1 liter of buffer B (Spectra/Por, 6 -8000 MWCO). Dialyzed samples could be rapidly frozen and stored at Ϫ70°C for several weeks without loss of activity.
Affinity Chromatography-In preliminary experiments, it was found that in contrast to previous claims that eukaryotic GTP cyclohydrolase I could not be purified by affinity chromatography on a GTP-affinity matrix (24), rat liver GTP cyclohydrolase I activity purified 200-fold through the dialysis step described above was found to strongly bind to GTP-agarose affinity media and could be eluted specifically with GTP. However, when this was done, almost all of the BH 4 -dependent inhibitory activity that was present in the dialyzed sample and bound to the GTP-agarose column was eluted by GTP together with GTP cyclohydrolase I activity (data not shown). Further attempts to separate the inhibitory activity from the GTP cyclohydrolase I activity by anion exchange chromatography or by gel filtration were not successful. However, it was discovered that it was possible to dissociate GFRP from the GTP cyclohydrolase I-GFRP complex by washing the affinity column with 0.2 M NaCl prior to specific elution of GTP cyclohydrolase I with GTP (see "Results"). In this protocol, the dialyzed sample was applied to a column of GTP-agarose (1 ml) pre-equilibrated in buffer B. After washing with 50 ml of buffer B, GFRP was eluted with 10 ml of buffer B containing 0.2 M NaCl. After a further wash with 1 ml of buffer B alone, the GTP cyclohydrolase activity I was eluted with 10 ml of buffer B containing 5 mM GTP. The salt eluate and the GTP eluate were separately concentrated to small volumes in centrifugal concentrators (Centricon 10, Amicon). The GTP-affinity column was regenerated by washing with 50 ml of water, 1 ml of 10% sodium dodecyl sulfate, 50 ml of water, and finally with 5 ml of 50% glycerol before storage at Ϫ20°C. It should be noted that the same column was reused more than 10 times without noticeable loss of GTP cyclohydrolase I binding capacity.
Further Purification of GTP Cyclohydrolase I-Affinity-purified GTP cyclohydrolase I was purified to homogeneity in high yield by a two-step procedure utilizing high performance anion exchange chromatography followed by high performance gel filtration. The concentrated, affinitypurified GTP cyclohydrolase I (Ͻ1 ml) was diluted with 4 ml of buffer B and injected onto a Mono Q column (Pharmacia, HR 5/5) equilibrated with buffer B at a flow rate of 1 ml/min. The chromatography system consisted of an all titanium Gilson HPLC system with a Rainin Macintosh controller. The absorbance was monitored at 280 nm with a Gilson 121 UV detector. All runs were made at room temperature. After washing with buffer B until the absorbance reached base line, a linear gradient to 0.4 M NaCl in buffer B was run over 80 min, and 1-min fractions were collected. Groups of 4 -5 fractions were collected and transferred to an ice bath until assayed. Fractions with the highest activity were pooled and concentrated to 200 l with Centricon 30 concentrators (Amicon). The concentrated GTP cyclohydrolase I fractions were then purified by gel filtration on a Superose 6 (Pharmacia) column using the HPLC apparatus described above and eluted with buffer C (same as B, without protease inhibitors) containing 0.1 M NaCl at a flow rate of 0.3 ml/min. Fractions (0.3 ml) with the greatest activity were pooled, concentrated as above, and stored in aliquots at Ϫ70°C.
Further Purification of GFRP-GFRP eluted from the GTP-affinity column by 0.2 M NaCl was purified to homogeneity by a similar two-step HPLC procedure carried out in the reverse order, gel filtration followed by Mono Q chromatography. The NaCl eluate from the GTP-agarose column was concentrated to 200 l with Centricon 10 concentrators and injected onto the HPLC gel filtration columns at a flow rate of 0.4 ml/min with buffer C (containing 0.1 M NaCl). A Zorbax GF-250 precolumn and column (DuPont NEN) were coupled in series with a Superose 6 column to increase the separation capacity. Fractions from the tandem gel filtration columns with the highest BH 4 -dependent GTP cyclohydrolase I inhibiting activity were pooled, concentrated with Centricon 10 concentrators, and buffer-exchanged into buffer C with a PD-10 column (Pharmacia). Mono Q chromatography was then carried out as described above for GTP cyclohydrolase I. Fractions with GFRP activity were concentrated with a Centricon 10 concentrator and stored at Ϫ70°C.

Cloning and Expression of Rat Liver GFRP cDNA
Degenerate primers Sm1 (coding strand, ATGACNGGNGTNAC-NCARACT) and Sm5 (noncoding strand, GTYTGNCCNACNCCNGT-CAT) were used on DNA prepared from a rat liver cDNA library (provided by M. Brownstein) in the pCD SP6/T7 vector. PCR amplification (96°C, 1 min; 55°C, 1 min; 72°C, 4 min; 30 cycles) gave a 0.19-kilobase product whose predicted amino acid sequence exactly matched the target peptide sequence. To amplify the 5Ј end of the cDNA, a pair of nested vector primers, up3 (AACTGCTCCTCAGTGGATGTTGC) and up2 (CTAGGCCTGTACGGAAGTGTTAC), were used with Sm5 and a nested primer based on the cDNA sequence, Sm7 (AAGCCCCTGCAT-TCCAGCTTG). To amplify the 3Ј end of the cDNA, primers I and Sm6 (GACGATGGTGGGTGATGAGCA) were used in conjunction with vector primer dn2 (ATTCAGTTGTGGTTTGTCCAAAC). PCR products (96°C, 1 min; 60°C, 1 min; 72°C, 4 min; 30 cycles) of 250-1300 bases resulting from primers I and dn2 or up3 and Sm5 were size-selected on a 2% NuSieve gel and reamplified with primers dn2 and Sm6 or up2 and Sm7, respectively. The resulting products of about 650 bases (3Ј end) and 230 -330 bases (5Ј end) were subcloned into pUC18 using a BamHI site in the cDNA sequence and either XhoI or Pst sites in the pCD-derived portion of the product and sequenced. To obtain a fragment containing the complete coding sequence, the PCR reactions were repeated using primers derived from the 5Ј and 3Ј ends of the cDNA and containing a built-in KpnI or XbaI site (GFR3, CCGGTACCCATGC-CCTACCTGCTCATCAGCACTCA; GFR4, GCCTCTAGATTGGCTTG-GATTAGCTGTGTAGG). Digestion of the resulting product with KpnI and XbaI allowed it to be inserted into the KpnI and XbaI sites of the pTrxFus expression vector (Invitrogen) with the initiating methionine in-frame with the thioredoxin gene. Cleavage of the fusion protein with enterokinase should produce GFRP protein with two additional amino acids at the amino-terminal, valine and proline. Proper insertion of the fragment and the absence of PCR-generated mutations was confirmed by sequencing the expression vector construct. A GenBank TM search did not reveal any known proteins with significant homology to GFRP.
The GFRP-thioredoxin fusion protein was expressed in Escherichia coli GI724 cells (Invitrogen) after transformation of the cells with the pTrxFus vector by induction with 100 M L-tryptophan overnight at 30°C. Cells were harvested by centrifugation, freeze-thawed twice, and then sonicated on ice in 20 mM Tris-HCl (pH 8), 2.5 mM EDTA. The extract was clarified by centrifugation, and the fusion protein was purified by affinity chromatography on a ThioBond column (Invitrogen) according to the manufacturer's directions. The fusion protein was then purified by chromatography on a Mono Q column (Pharmacia) eluted with a linear NaCl gradient in buffer B. Aliquots of fractions were analyzed by SDS-PAGE on 16% Tricine gels (Novex), and fractions containing 23-kDa fusion protein as well as GFRP activity were pooled and concentrated with Centricon 10 microconcentrators.

Northern Blot
A rat multiple tissue Northern blot containing approximately equal amounts of polyadenylated mRNA from various rat tissues (Clontech) was prehybridized for 4 h in 5 ϫ SSPE, 5 ϫ Denhardt's, 50% formamide, 2% SDS, 100 g/ml denatured salmon DNA at 42°C. The GFR3/4 PCR fragment was labeled with [ 32 P]dCTP by Klenow DNA polymerase with the random priming method and then hybridized to the blot for 16 h in the same solution used for prehybridization followed by three washes in 0.2 ϫ SSPE, 0.1% SDS at 60°C. Radioactive bands were visualized on a Fuji phosphoimager.

RESULTS
The results of a typical purification of both GTP cyclohydrolase I and GFRP from rat liver are given in Table I. Similar results have been obtained in more than 5 preparations.
Purification of GFRP-GFRP activity separated from GTP cyclohydrolase I activity by salt elution of the GTP-agarose column, as described under "Experimental Procedures," was eluted from the tandem gel filtration columns as a small ultraviolet absorbing peak at 75 min ( Fig. 2A). This elution time corresponded to a molecular mass of 20 kDa. No inhibitory activity was detected in fractions corresponding to a molecular mass of 35 kDa. Concentration of the fractions containing GFRP activity followed by anion exchange chromatography on Mono Q yielded a major 280 nm absorbing peak which contained all of the GFRP activity (Fig. 2B). Analysis of this highly purified preparation of GFRP by SDS-polyacrylamide gel electrophoresis demonstrated the presence of a major Coomassiestained band with a molecular mass of 9.5 kDa (Fig. 2C). Thus, GFRP apparently consists of a homodimer of 9.5-kDa subunits.
Purification and Properties of GTP Cyclohydrolase I-The further purification of affinity-purified GTP cyclohydrolase I by anion exchange chromatography on a Mono Q column is shown in Fig. 3A. A major peak of protein and activity was detected eluting at approximately 0.2 M NaCl. However, it should be noted that the GTP cyclohydrolase I peak was found to contain significant amounts of dihydroneopterin triphosphate which was produced during the GTP elution step from the GTPaffinity column and subsequent concentration of the GTP cyclohydrolase I activity. The bound dihydroneopterin triphosphate contributes to the absorbance profile, and it is necessary to correct activity assays for the presence of preformed dihydroneopterin phosphate. Peak fractions of GTP cyclohydrolase I activity were then pooled, concentrated, and separated by high performance gel filtration (Fig. 3B). The GTP cyclohydrolase I activity eluted at 46 min, corresponding to a molecular mass of 300 kDa, in agreement with previous estimations (25).
Purified GTP cyclohydrolase I showed a single band at 30 kDa by SDS-polyacrylamide gel electrophoresis (Fig. 3C), in agreement with that previously reported (25). As described previously for recombinant GTP cyclohydrolase I (26), purified rat liver GTP cyclohydrolase I showed sigmoidal kinetics with varying GTP concentration with a Hill coefficient of 2 and an apparent K m for GTP of 31.1 Ϯ 8.9 M (data not shown).
Regulation of GTP Cyclohydrolase I Activity by GFRP-In preliminary attempts to purify the GTP cyclohydrolase I-GFRP complex after elution from the GTP-affinity column by GTP alone, most of the BH 4 -dependent inhibitory activity was found to be present in trailing fractions of the GTP cyclohydrolase I peak during a subsequent gel filtration chromatography step. Although no detectable BH 4 -dependent GTP cyclohydrolase I inhibitory activity was found in the 35-kDa region of the gel filtration chromatogram as expected based on the report by Harada et al. (18), some inhibitory activity was detected in TABLE I Purification of rat liver GTP cyclohydrolase I and GFRP Because Ͼ75% of the total GFRP activity is associated with GTP cyclohydrolase I in extracts of rat liver, the purification factors and yields for GTP cyclohydrolase I and GFRP are essentially identical through the affinity chromatography step (GTP-Ag).
Step a ND, not done. b BH 4 -dependent inhibitory activity in these fractions was determined with added, purified GTP cyclohydrolase I. The BH 4 -dependent inhibitory activity in the previous steps was measured by determining the effect of BH 4 on endogenous GTP cyclohydrolase I activity (see "Experimental Procedures"). One unit of GTP cyclohydrolase I activity is defined as the formation of 1 pmol/min of dihydroneopterin triphosphate and 1 unit of GFRP inhibitory activity as the BH 4 -dependent decrease of 1 pmol/min in the rate of formation of dihydroneopterin triphosphate, measured as described under "Experimental Procedures." fractions corresponding to much lower molecular mass (data not shown). This observation raised the possibilities that the higher ionic strength buffer used in the gel filtration step might be causing dissociation of bound GFRP from GTP cyclohydrolase I and also that the actual molecular mass of GFRP might be less than the expected 35 kDa. The first possibility was investigated by examining the effect of increasing salt concentration on the BH 4 -dependent inhibition of affinity-purified GTP cyclohydrolase I-GFRP complex. As shown in Fig. 4, high salt concentration alone strongly inhibited GTP cyclohydrolase I activity, an effect which has previously been noted with human GTP cyclohydrolase I and was ascribed to a decrease in the affinity of the enzyme for GTP (27). However, intermediate salt concentrations, from 100 -300 mM, had little or no direct effect on GTP cyclohydrolase I activity determined in the absence of BH 4 but effectively relieved the BH 4 -dependent inhibition. In confirmation of the proposal that the reversal of the BH 4 -dependent inhibition of GTP cyclohydrolase I by salt was caused by dissociation of GFRP, 0.2 M NaCl effectively released the GFRP activity from the GTP-affinity column where it was tightly associated with GTP cyclohydrolase I. The bound GTP cyclohydrolase I was not eluted from the column by this salt concentration. Interestingly, L-phenylalanine which reverses the BH 4 -dependent inhibitory effect (see below) did not cause release of GFRP or GTP cyclohydrolase I from the affinity column. Another pertinent finding was that GFRP eluted at approximately 0.2 M NaCl from a Mono Q anion exchange column, which is the same salt concentration that elutes GTP cyclohydrolase I activity.
The specificity of the pterin-dependent inhibition of GTP cyclohydrolase I by GFRP was examined using the purified enzyme and regulatory protein. The natural 6R-BH 4 isomer had the most potent inhibitory effect and was severalfold more potent than 7-BH 4 . The concentration of 6R-BH 4 which gave 50% inhibition of GTP cyclohydrolase I was about 3 M which is comparable to the half-maximally effective concentration of 2 M previously found (18). In contrast to previous studies (18), however, all of the dihydropterins examined, including 7,8dihydrobiopterin, 7,8-dihydroneopterin, and sepiapterin, were nearly as potent as 6R-BH 4 when assayed with purified GFRP (Table II). Furthermore, xanthopterin, which exists in solution as a hydrated dihydropterin, was also found to have potent inhibitory power. In contrast, other fully oxidized pterins that were tested, with the exception of pterin-6-carboxylic acid, had no GFRP inhibitory activity.
In the feedback inhibitory scheme for the action of GFRP in the regulation of GTP cyclohydrolase I activity (Fig. 1), Lphenylalanine abrogates the BH 4 -dependent inhibitory effect. As shown in Fig. 5, L-phenylalanine specifically and dose-dependently reversed the BH 4 -dependent inhibition of GTP cyclohydrolase I activity in the presence of purified GFRP. 0.4 mM L-phenylalanine restored nearly one-half of the GTP cyclohydrolase I activity inhibited by 10 M BH 4 and GFRP. Other amino acids, including D-phenylalanine, L-tyrosine, and L-tryptophan were ineffective, in agreement with previous studies using a crude GFRP preparation (18).
Cloning and Expression of GFRP-Purified GFRP was digested with trypsin, and the resulting peptides were isolated by reverse-phase HPLC. Three relatively long peptides were sequenced by automated Edman degradation, the sequences of which are indicated in Fig. 6. Oligonucleotide primers based on the peptide sequences were used for PCR screening of a rat liver cDNA library. After several rounds of PCR screening, a full-length cDNA was obtained which contained an open reading frame coding for an 84-amino acid protein with a calculated molecular mass of 9665 daltons (Fig. 6). The deduced peptide sequence contained the complete sequences of the three isolated peptides. The GFRP cDNA was ligated into the pTrxFus vector and expressed as a 23-kDa thioredoxin fusion protein.
The fusion protein was purified by affinity chromatography on phenylarsine oxide-modified agarose which specifically binds the vicinal dithiols of thioredoxin. The fusion protein demonstrated BH 4 -dependent GFRP activity which was specifically reversed or blocked by L-phenylalanine (Fig. 7). Pure, recombinant thioredoxin (Sigma) had no effect on either GTP cyclohydrolase I or GFRP activities. Enterokinase treatment cleaved the fusion protein into its constitutive components. However, GFRP activity of the fusion protein was lost after treatment with enterokinase although GFRP isolated from rat liver was not digested by enterokinase nor did it lose activity when treated in the same manner as the fusion protein. Cleavage of the fusion protein would yield recombinant GFRP with two additional amino acids at the amino-terminal which may interfere with proper folding or association with GTP cyclohydrolase I. FIG. 4. The effect of increasing salt concentration on tetrahydrobiopterin-dependent inhibition of GTP cyclohydrolase I mediated by GFRP. The GTP cyclohydrolase I-GFRP complex was purified though the ammonium sulfate step according to the purification procedure described under "Experimental Procedures." The complex was then isolated by affinity chromatography on a GTP-agarose column which was washed with buffer and eluted with 5 mM GTP in the same buffer (without any added salt). This material was concentrated with a centrifugal concentrator, and GTP cyclohydrolase I activity was determined as described under "Experimental Procedures" in the absence or presence of 2 M BH 4 and the indicated concentrations of NaCl.

TABLE II
Effect of various pterins on GFRP activity GTP cyclohydrolase activity I was determined in the presence of GFRP and the absence or presence of the indicated concentration of pterin as described under "Experimental Procedures." Samples were incubated for 15 min at 37°C and the formation of dihydroneopterin triphosphate was measured. Reaction mixtures contained 0.2 g (0.7 pmol) of purified GTP cyclohydrolase I and 0.18 g (10 pmol) of purified GFRP. FIG. 5. L-Phenylalanine specifically reverses tetrahydrobiopterin-dependent inhibition of GTP cyclohydrolase I mediated by GFRP. BH 4 (2 M) was added to reaction mixtures containing purified GTP cyclohydrolase I and GFRP in the absence or presence of the indicated amino acids. GTP cyclohydrolase I activity was determined by measuring the amount of dihydroneopterin triphosphate formed after 15 min at 37°C as described under "Experimental Procedures." C, control GTP cyclohydrolase I activity determined in the absence of BH 4 .
Northern Blot Analysis-Because it appears as though there is a stoichiometic complex between GTP cyclohydrolase I and GFRP in the liver, it was important to determine whether GFRP is also associated with GTP cyclohydrolase I in other tissues. To this end, Northern blot analysis of GFRP mRNA levels was performed on several rat tissues. The GFRP probe detected a major mRNA species of 800 base pairs (Fig. 8). The minor bands at 1.4 and 6 kilobases presumably represent incompletely spliced RNAs. Testis, heart, brain, and lung had lower levels of GFRP mRNA, while levels in spleen and muscle were much lower. DISCUSSION GTP cyclohydrolase I activity is regulated in a unique feedback regulatory loop by tetrahydrobiopterin in conjunction with a regulatory protein. Our observation that GTP cyclohydrolase I and the small feedback regulatory protein GFRP exist in a tight complex has facilitated the purification of GFRP to homogeneity from rat liver. In the purification procedure described in this report, 200-fold purified GTP cyclohydrolase I-GFRP complex was bound to a GTP-affinity column from which the GFRP could be eluted selectively. GFRP was then purified to homogeneity in high yield, with a final purification factor of nearly 70,000-fold from rat liver. This was approximately the same extent of purification as was found for GTP cyclohydrolase I specifically eluted from the GTP-affinity column with GTP and then purified by anion exchange chroma-tography and gel filtration. The native molecular mass of GFRP was found to be 20 kDa by gel filtration, consisting of a homodimer of 9.5-kDa subunits as determined by SDS-polyacrylamide gel electrophoresis. It is likely that the higher molecular mass of 35 kDa previously found by gel filtration of crude rat liver extracts (18) was caused by slow dissociation of GFRP from GTP cyclohydrolase I during the chromatography, a phenomenon noted during this study. GFRP, similar to GTP cyclohydrolase I, is a relatively stable protein. It was not highly sensitive to proteolysis, losing only about 30% of its inhibitory activity after digestion with trypsin for 2 h; it was also heatstable since no activity was lost during the purification when preparations were heated to 65°C, although most of the inhibitory activity was destroyed after 5 min in a boiling water bath.
In the affinity purification scheme developed in this study, GTP cyclohydrolase I was purified from rat liver to an apparent extent of 91,000-fold and a recovery of 65%. However, as shown in Table I, the total GTP cyclohydrolase I activity increased by about 3.5-fold during the purification when the activity present in liver extracts was purified through the ammonium sulfate step as described under "Experimental Procedures." This increase in activity probably results both from the removal of inhibitory proteins (28), as well as dissociation of bound endogenous BH 4 from the GTP cyclohydrolase I-GFRP complex. Thus, when purified GTP cyclohydrolase I is added to rat liver extracts, the activity is strongly inhibited, giving Ͻ50% recovery of the added activity in assays containing total amounts of protein normally used for GTP cyclohydrolase I activity measurements. 2 The phenomenon of increased GTP cyclohydrolase I activity after heat treatment, which has been described previously (29), suggests that the actual overall recovery of stable GTP cyclohydrolase I activity purified by the procedure described here should be considered to be closer to 25% rather than 65%, still a substantial improvement over a previously described procedure (25). Several factors contribute to this improvement. GTP cyclohydrolase I was found to be sensitive to sulfhydryl reagents and, in the absence of added DTT, rapidly loses activity. Thus, all buffers used in the purification contained DTT plus EDTA added to diminish metal-catalyzed oxidations. Furthermore, both glycerol and the nonionic detergent Tween 20 were present throughout since when the purification procedure was carried out in their absence, the proteins in the concentrated affinity-purified GTP cyclohydrolase I fraction noticeably aggregated and became intractably insoluble. Most importantly, although several laboratories have pre- viously failed to demonstrate binding of eukaryotic GTP cyclohydrolase I to a GTP-affinity column (24), it now seems likely that this failure was due both to the relative high abundance of other GTP-binding proteins in unpurified tissue or cell extracts and the presence of bound GTP on GTP cyclohydrolase I. As described here, when GTP cyclohydrolase I was first purified about 200-fold before being applied to GTP-agarose affinity columns, it bound very tightly and could then be eluted specifically by GTP resulting in a further large degree of purification. It should be noted that in contrast to a previous report (25), purified GTP cyclohydrolase I was quite stable when stored in the presence of DTT, Tween 20, and glycerol, losing less than 20% of its activity after 2 months at Ϫ70°C.
Because GTP cyclohydrolase I appears to be a ubiquitous enzyme, expressed either constitutively or inducibly in almost every cell type, a major question is whether GFRP is also present in other tissues besides liver and regulates GTP cyclohydrolase I activity in those tissues. The distribution of GTP cyclohydrolase I activity and BH 4 content have been determined in various tissues and body fluids of humans, monkeys, dogs, and mice (19). As expected, there is a good correlation between them with highest levels of both generally found in liver, pineal, bone marrow, and adrenal. Both GTP cyclohydrolase I activity (14) and mRNA (30) as well as BH 4 levels are induced in brain, liver, spleen, and adrenal gland of lipopolysaccharide-treated rats (14). By Northern blot analysis, GFRP mRNA levels in the rat parallel the GTP cyclohydrolase I expression and BH 4 content, being greatest in liver and kidney, although significant amounts were also found in testis, brain, heart, and lung (Fig. 8). It should be noted that by direct enzymatic assays, no free GFRP activity could be detected in any tissue except liver where the majority of the BH 4 -dependent GTP cyclohydrolase I inhibitory activity occurs in a complex with GTP cyclohydrolase I and suggests that GFRP may also exist in such a complex in other tissues. Further studies are in progress on determining the distribution of GFRP in various rat tissues using a specific antibody to detect GFRP.
The interaction of GFRP with GTP cyclohydrolase I creates a binding site for pterins/amino acids which has similar binding specificity as the analogous binding sites of phenylalanine hydroxylase. As with the hydroxylase, the natural cofactor, 6Rtetrahydrobiopterin, is a better inhibitor of GTP cyclohydrolase activity than the 7 isomer or the 6-methyltetrahydropterin analog, and dihydropterins are also effective inhibitors (Table  II). Furthermore, L-phenylalanine is the only amino acid which has been found to reverse the BH 4 inhibitory effect and neither D-phenylalanine nor L-tyrosine have this ability (Fig. 5) or are substrates or inhibitors of phenylalanine hydroxylase (1). The tetrahydrobiopterin-GFRP-dependent inhibition of GTP cyclo-hydrolase I activity results from a decrease in catalytic activity rather than an alteration in affinity for GTP. The mechanism whereby association of GFRP with GTP cyclohydrolase I confers sensitivity to feedback inhibition by BH 4 has not yet been elucidated. A plausible scheme is presented in Fig. 9. GFRP and GTP cyclohydrolase I are considered to be normally tightly associated because only a small fraction of GFRP in liver extracts appears to be unbound, and mixing equimolar amounts of GFRP and GTP cyclohydrolase I is sufficient to generate a complex which is BH 4 -sensitive. There is little evidence for binding of BH 4 to the GTP cyclohydrolase I-GFRP complex. However, when crude rat liver extracts were fractionated by gel filtration, fractions that contained BH 4 -sensitive GTP cyclohydrolase I activity also contained BH 4 and there was no BH 4 associated with affinity-purified GFRP. 2 The GTP cyclohydrolase I in the tripartite complex likely undergoes a conformational change which reduces the enzyme activity. It is proposed that L-phenylalanine specifically reverses the inhibitory activity by displacing BH 4 from the complex. L-Phenylalanine is shown bound to the GFRP-GTP cyclohydrolase I complex because, as discussed above, it does not cause release of GFRP from the complex when it is bound to GTP-agarose, and, in addition, it has been shown that L-phenylalanine alters GTP cyclohydrolase I activity, converting the kinetics with GTP from sigmoidal to hyperbolic (18). Finally, raising the ionic strength causes dissociation of the inhibitory complex and removes sensitivity to BH 4 as well as restoring full GTP cyclohydrolase I activity. The number of molecules of GFRP bound to decameric GTP cyclohydrolase I has not yet been established. Further characterization of the GFRP regulatory mechanism by physical and kinetic methods should be facilitated by the cloning and overexpression of larger amounts of GFRP. Studies are underway to determine the physiological role of GFRP in the regulation of other BH 4 -dependent processes, such as biogenic amine neurotransmitter sythesis and nitric oxide production. FIG. 9. Proposed mechanism for the regulation of GTP cyclohydrolase I activity by GFRP and BH 4 . Fully active GTP cyclohydrolase I is indicated by the large circle. The binding of BH 4 to the GTP cyclohydrolase I-GFRP complex induces a conformational change which decreases activity. Phenylalanine displaces the bound BH 4 and restores activity. In the presence of high salt, GFRP dissociates from GTP cyclohydrolase I.