Tetrahydrobiopterin and its functions

Transcription

1 Tetrahydrobiopterin and its functions B. Thöny Division of Clinical Chemistry and Biochemistry, University Children s ospital, Zurich, Switzerland This article is dedicated to my collaborator, Walter Leimbacher, who died on June 4, 2004 at the age of 58. Introduction Tetrahydrobiopterin (B 4 ) is essential for diverse processes and is ubiquitously present in all tissues of higher organisms. It is well established as essential cofactor for various enzyme activities to activate dioxygen, and for less defined functions at the cellular level. The latter function of B 4 seems to have a dual nature, depending on the concentration of the cofactor and the cell type, between proliferative activity and trigger for apoptosis. Since B 4 (and related tetrahydropterins) can form radicals, it can act as a generator or scavenger of reactive oxygen species in cells. More characteristics are arising based on recent observations, including chaperon function reported at least for the hepatic phenylalanine hydroxylase. The B 4 cofactor is endogenously synthesized and a (genetic) deficiency in the biosynthesis or regeneration leads to neurological abnormalities, including Dopa-responsive dystonia or severe monoamine neurotransmitter depletion. owever, also under normal B 4 concentrations its availability becomes limiting in various pathological situations involving endothelial dysfunction, for instance, in diabetes or coronary heart diseases. From such developments it is expected that research on cofactor function will become of even broader interest in unraveling pathophysiological connections. The following overview is basically an extension of a previous review article on the biosynthesis, regeneration, and functions of the B 4 cofactor [1]. The main advances since the last review involve progress on the GTP cyclohydrolase I feedback regulatory protein (GFRP) with structural and functional aspects, on the postulation of alternative pathway steps for the biosynthesis of B 4 in peripheral organs based on the discovery of patients with sepiapterin reductase deficiency, on the function of B 4 cofactor as a chaperon, and on the generated mouse models for B 4 deficiency. The focus of this overview will again be on the structure and function of the mammalian enzymes responsible for the synthesis and regeneration of B 4. (A comprehensive review covering B 4 biology also in nonmammalian species can be found in [2]). Taken together, advances in the knowledge of cofactor function is not solely an academic exercise for biochemists and geneticists, but should be the basis for understanding the pathophysiology in order to eventually help patients with different defects involving directly or indirectly the B 4 cofactor. 495

2 Biosynthesis of B 4 In the following, the de novo biosynthesis of the B 4 cofactor carried out by the three classical enzymes is described. In addition, a hypothetical alternative way to circumvent the last enzymatic step by involving other reductases is presented (from compound 5 to 11 in scheme 1; see also below). Part of this alternative pathway is also called salvage pathway, which feeds B 4 from dihydrobiopterin via the ADP-dependent dihydrofolate reductase. As will be discussed, and in contrast to our previous view, this alternative pathway seems to compensate for a genetic defect in the last biosynthetic step at least in peripheral tissues. Reaction mechanism of the de novo pathway B 4 biosynthesis proceeds in the de novo pathway in a Mg 2+ -, Zn 2+ -, and ADPdependent reaction from GTP via the two intermediates, 7,8-dihydroneopterin triphosphate ( 2 TP, compound 5) and 6-pyruvoyl-5,6,7,8-tetrahydropterin (PTP, compound 8), which have been isolated although they are rather unstable (Figure 1). The three enzymes GTP cyclohydrolase I (EC ; GTPC), 6-pyruvoyl-tetrahydropterin synthase (EC ; PTPS), and sepiapterin reductase (EC ; SR) are required and sufficient to carry out the proper stereospecific reaction to 6R-L-erythro-5,6,7,8-tetrahydrobiopterin (compound 11). With the crystallographic structures, including the characteristics of the active centers of all three enzymes, the essential information for the interpretation of the reaction mechanism is available. Moreover, MR studies on the reaction mechanisms of all three enzymes revealed the details of the hydrogen transfer process and the stereochemical course of the reactions [3]. The committing step is carried out by GTPC, a homodecamer containing a single zinc ion in each subunit, consisting of a tightly associated dimer of two pentamers [4, 5]. GTPC contains 10 equivalent active centers with 10-Å deep pockets. The interface of 3 subunits, two from one pentamer and one from the other forms an active site. This cavity is structurally stabilized by two loops (residues and in the E. coli enzyme) formed by hydrogen bond interactions and a salt bridge (Glu111 Arg153). An intramolecular disulfide bridge (Cys110 Cys181) is not essential for active site integrity as shown by Cys-mutant structures [6]. The atomic structure of this pocket is not only highly selective for GTP, but also provides residues for complete charge compensation in order to render obsolete Mg 2+ -assisted binding to the protein, as found in other nucleoside triphosphate-binding proteins. 496

3 PPP 2C GTP 11 4a ,6,7,8-Tetrahydrobiopterin +ADP - ADP PPP 2C C 10 2 GTPC C P - P P P P - P P P PTPS SR Pyruvoyl-tetrahydropterin 2 -PPP +ADP - ADP + PPP PPP 6 2 PPP 7,8-Dihydroneopterin triphosphate Scheme 1: Reaction mechanism for the de novo pathway for B 4 biosynthesis 497

4 498

5 Figure 1: Three-dimensional structures of B4 metabolizing enzymes. Ribbon-type respresentations of the main-chain foldings of the proteins and enzymes involved in de novo B4 biosynthesis and regenaration are shown: GTPC, GTP cyclohydrolase I (homodecamer); GFRP, GTP cyclohydrolase feed-back regulatory protein (homopentamer); PTPS, 6-pyruvoyl tetrahydropterin synthase (homohexamer); SR, sepiapterin reductase (homodimer); PCD/DCo, pterin-4a-carbinolamine dehydratase/dimerization cofactor of hepatocyte nuclear factor-1α (homotetramer); DPR, dihydropteridine reductase (homodimer). Substrates are shown in ball-and-stick representation with atoms in standard colors. n the right side the enzymes are shown rotated by 90º around the x-axis. The crystal structure coordinates used are GTPC from E. coli (Protein Data Bank entry code 1GTP), GFRP from rat (1JG5), PTPS from rat liver (1GTQ, 1B66), SR from mouse (1SEP), PCD/DCo from rat (1DC, 1DCP), and DPR from rat liver (1DR). The figure was created using the programs MLSCRIPT (Kraulis, 1991, J Appl Crystallogr 24:945), REDER (Merrit and Murphy, 1994, Acta Crystallog Sct D Biol Chrystallogr D50:869), and/or PyML ( A catalytic mechanism was proposed based primarily on structural analysis obtained from E. coli GTPC co-crystals with the dgtp analog and several active site mutants, and single turn-over experiments using a kinetically competent reaction intermediate [6, 7] (see Scheme 1). The purine hydrolysis reaction may be initiated by protonation of -7 by a specific histidine residue (is179 in the E. coli enzyme). Attack of water at C-8 and suc- 499

6 ceeding opening of the imidazole ring result in a first intermediate, the 2-amino-5-formylamino-6-ribofuranoside triphosphate (compound 2). Protonation of the bridging atom in the furanose ring and release of C-8 of GTP by another histidine residue (is112, E. coli) as formate yields the Schiff s base intermediate 3. The subsequent Amadori rearrangement catalyzed by involving the γ-phosphate of GTP and a serine residue (Ser135, E. coli), and keto-enol tautomerization results in the intermediate compound 4. The last step is the ring expansion by closing between -7 and C-2, yielding the product 2 TP (compound 5). It is assumed that this last reaction step takes place rather at the protein surface or in solution after dissociation, as the spatial structure of the active site pocket does not favor the reaction to occur inside. The reaction from 2 TP (compound 5) to PTP (compound 8) is catalyzed by PTPS in a Zn 2+ - and Mg 2+ -dependent reaction without consuming an external reducing agent (Figure 1). This conversion involves a stereospecific reduction by an internal redox transfer between atoms -5, C-6, and C-1, oxidation of both side-chain hydroxyl groups, and an unusual triphosphate elimination at the C-2 -C-3 bond in the side chain. Crystallographic analysis revealed that PTPS is composed of a pair of trimers arranged in a head-tohead fashion to form the functional hexamer [8]. The homohexamer contains six active sites that are located on the interface of three monomers, two subunits from one trimer and one subunit from the other trimer. The catalytic center and the reaction mechanism were studied by crystallographic and kinetic analysis of wild-type and mutant PTPS from the rat [9, 10]. In addition, the crystal structure of the inactive mutant Cys42Ala PTPS in complex with its natural substrate 2 TP was determined [10]. Each catalytic center harbors a Zn 2+ -metal binding site in a 12-Å deep cavity. The active site pocket with the specific pterin-anchoring Glu residue for salt-bridging plus two hydrogen-bonding amino acids appears to be similar to the equivalent sites in GTPC, sepiapterin reductase, dihydroneopterin epimerase, and neopterin aldolase. The active site pocket contains in addition two catalytic motifs: a Zn 2+ -binding site and an intersubunit catalytic triad formed by a Cys, an Asp, and a is residue. The tetravalent co-ordination of the transition metal is accomplished via the ε-atoms of three is residues and a fourth ligand provided by the pyruvoyl moiety of the 2 TP substrate. Unfortunately, neither the triphosphate nor the putative Mg 2+ ion moieties could be defined in the electron density map. Zn 2+ plays a crucial role in catalysis as it activates the protons of the substrate and stabilizes the intermediates (compounds 6 and 7). The proposed reaction mechanism is the following: Protonation of -5 and abstraction of a proton from the C-1 side-chain carbonyl atom lead to the -5-C-6 double-bond reduction (compound 6). Stereospecific protonation of C-6 and oxidation of C-1 - to C-1 = leads to compound 7. The last step is the abstraction of a proton from the C-2 carbon of the carbonyl side-chain, followed by triphosphate elimination and tautomerization to yield PTP (compound 8). The Cys residue (Cys42 in the rat enzyme) of the catalytic triad appears to be the general base for stereospecific protein abstraction in both reaction steps. 500

7 The final step is the ADP-dependent reduction of the two side-chain keto groups of PTP (compound 8) by SR. The overall structure of SR is a homodimer stabilized by a common four-helix bundle [11]. Each monomer contributes with two α-helices to the central dimerization domain and forms a separate complex composed of seven parallel β-sheets surrounded by α-helices. The C-terminal end of the β-sheets contains in close vicinity ADP and the pterin binding site, the latter comprising a 15-Å deep pocket. The pterin substrate is anchored by the guanidino moiety of a specific Asp (Asp258 in the mouse SR; Asp258 is also the anchoring residue for -acetyl serotonin inhibitor binding; see below). The C-1 carbon of the pterin side-chain is in direct proximity to ADP and a Tyr -group (Tyr171 in the mouse SR). This Tyr is the central active site residue for optimal proton transfer [12]. Based on kinetic, crystallographic, and MR data, the initial step is the ADP-dependent reduction at the side-chain C-1 -keto function leading to the formation of oxopropyl tetrahydropterin (compound 9) [3]. Internal rearrangement of the keto group via side-chain isomerization leads to the 1 -keto compound 6-lactoyl tetrahydropterin (compound 10). Intermediate compound 10 is then reduced to B 4 (compound 11) in a second ADP-dependent reduction step. While the pterin substrate remains bound to the active site, the redox cofactor has to be renewed after the first reduction. It is thus assumed that ADP is exchanged at the opening located at the opposite side of the pterin-binding and entry pocket. Alternative routes for biosynthesis of B 4 Besides the involvement in the de novo biosynthesis of B 4, SR may also participate in the pterin salvage pathway by catalyzing the conversion of sepiapterin (Scheme 2, compound 17) to 7,8-dihydrobiopterin (compound 16) that is then transformed to B 4 by dihydrofolate reductase (DFR; EC ) [13]. Both reactions consume ADP. Although SR is sufficient to complete the B 4 biosynthesis, a family of alternative ADPdependent aldo-keto reductases, including carbonyl reductases (CR), aldose reductases (AR), and the 3α-hydroxysteroid dehydrogenase type 2 (AKR1C3) may participate in the diketo reduction of the carbonyl side-chain in vivo [14 16]. Moreover, based on the discovery of the autosomal-recessive deficiency for SR, which presents with neurotransmitter deficiency but without hyperphenylalaninemia (see also below), different routes for the final two-step reaction of B 4 biosynthesis involving these alternative reductases were proposed [16, 17] (Scheme 2). CR converts 6-pyruvoyl tetrahydropterin (compound 8) to both, the 2 -oxo-tetrahydropterin (compound 9) and the lactoyl tetrahydropterin (or 1 - oxo-tetrahydropterin; compound 10); however, the rate of production is much more favorable for the 1 -oxo-tetrahydropterin intermediate. n the other hand, the 3α-hydroxysteroid dehydrogenase type 2 efficiently converts compounds 8 to 9. AR can convert 6- pyruvoyl tetrahydropterin to 6-lactoyl tetrahydropterin (compound 10), or to 2 -oxotetrahydropterin (compound 9) to B 4. In summary, alternative routes for B 4 in the absence of SR involve in one pathway option via the 2 -oxo-tetrahydropterin the concerted action of 3α-hydroxysteroid dehydrogenase type 2 and AR, thereby consuming 2 501

8 ADP. In the other pathway option via the 1 -oxo-tetrahydropterin intermediate, AR, CR, and dihydrofolate reductase are required. This second route involves consecutively the compounds 8, 10, 17, 16 to 11, and consumes 3 ADP due to a non-enzymatic reduction step to sepiapterin (from compound 10 to 17). It was proposed that due to low expression or activity of dihydrofolate reductase and 3α-hydroxysteroid dehydrogenase type 2 in human brain, the biosynthesis from 6-pyruvoyl tetrahydropterin to B 4 in the absence of SR cannot be completed. This leads to central B 4 deficiency with accumulation of the unstable 1 -oxo-tetrahydropterin (compound 10) and its degradation products, detected as abnormal pterin metabolites, i.e. compound ADP +ADP - ADP + 6-Pyruvoyl-tetrahydropterin - ADP + AR/CR SR/AKR1C3/CR 17 2 Sepiapterin SR/CR +ADP - ADP ,8-Dihydrobiopterin non-enzymatic Lactoyl tetrahydropterin DFR +ADP - ADP + SR SR +ADP +ADP -ADP + - ADP a ,6,7,8-Tetrahydrobiopterin ydroxy-2 -oxopropyl tetrahydropterin AR Scheme 2: Reaction mechanism for alternative routes for B 4 biosynthesis Enzyme structures GTPC and GFRP Structural crystallographic data are available for the recombinant GTPC from the human and the E. coli enzymes [4, 5,18,19], for the rat GFRP [20], and for the GTPC-GFRP cocomplex from rat [19]. The homodecameric GTPC is composed of two dimers of pentamers (Figure 1). Regarding the E. coli enzyme, each subunit contains 221 amino acids and folds into an α+β structure with a predominantly helical -terminus. The -terminal antiparallel helix pair α 2 /α 3 is remote from the rest of the molecule. The compact C-terminal body of the monomer (residues ) is formed by a central four-stranded antiparallel β-sheet (β-1 to β-4) that is flanked on both sides by α-helices (α 4, α 5, and α 6 ) (Figure 2A). The -terminal helix α 1 lies on top of helices α 4 and α 5. The association of 502

9 two GTPC monomers to dimers is driven by the formation of a four-helix bundle by helices α 2 and α 3. Five monomers interact along their β-sheets to form a 20-stranded antiparallel β-barrel with a diameter of 35 Å. The decamer is formed by a face-to-face association of two pentamers whereby the antiparallel helix pair α 2 /α 3 of one monomer is intertwined with that of another monomer. The decamer has a toroidal shape with an approximate height of 65 Å and a diameter of 100 Å. It encloses a cavity of 30 Å x 30 Å x 15 Å that is accessible through the pores formed by the five helix bundles in the center of the pentamers, but has no opening at the decamer equator. The active site is located at the interface of three subunits, two from one pentamer and one from the other. There are thus ten equivalent active sites per functional unit. The GTPC contains a zinc ion bound to the active site, i.e. bound to Cys-110, is-113, and Cys-181 in the bacterial enzyme (complexed by Cys-141, is-143, and Cys-212 in the human GTPC). Sequence information of a number of GTPCs from diverse organisms is available (Figure 2A). igh overall sequence similarity is found among the mammalian enzymes (approximately 90% identity). A comparison of all sequences reveals that mainly the C- terminal sequence is evolutionarily conserved; i.e. the C-terminal 120 residues between the human and the E. coli enzymes are 60% identical. Moreover, almost all residues participating in pterin binding and/or catalysis appear to be conserved. This high sequence homology suggests that the tertiary and quaternary structures of GTPC are most probably very similar, an assumption that was confirmed by solving the structure of the rat [19] and human enzymes, the latter with a deletion of amino acid residues 2 41 [5]. The high diversity of the -termini between the mammalian and the non-vertebrate enzyme, most notably the -terminal extension in the Drosophila sequence, may be due to different regulating functions such as, for instance, docking sites for the GFRP (see also below). In the 360 kda GTPC-GFRP complex, the decameric GTPC is sandwiched by two GFRP homopentamers. The model for the rat complex contains 10 GTPC units (residues ) and 10 GFRP subunits (residues 1 84), 10 phenylalanine molecules, and 10 zinc ions (Figure 1). Each monomer is folded into an α/β-structure with a sixstranded antiparallel β-sheet and two α-helices, arranged onto a ββαββαββ topology. The monomers form a propeller-like pentameric-disc with an overall dimension of 50 Å x 50 Å x 20 Å for this complex. The contacts between GFRP and GTPC are, besides salt bridging, primarily due to van der Waals interactions, as the concave side of GFRP pentamer exhibits neutral to negative surface potential, whereas the contact surface of the GTPC is mainly positively charged. The phenylalanine binding sites are located at the interface between GFRP and GTPC, and phenylalanine binding enhances the association of these proteins. The complex structure suggests that phenylalanine-induced GTPC- GFRP complex formation enhances GTPC activity by locking the enzyme in the active site. 503

10 504

11 Figure 2: Amino acid sequence alignments of B 4 synthesizing enzymes. Alignments were generated using the Clustal W program [216]. Identical residues and residues with functional similarities (V/L/I, R/K, M/L, D/E) are highlighted with shaded boxes. Residues involved in catalysis and/or substrate binding are marked with a triangle. The identified phosphorylation site in PTPS (Serine 19 in the human sequence) is marked with a star. In case of SR, residues involved in cofactor binding are additionally marked. The accession numbers for the amino acid sequences for (A) GTPC are P_ (human), AAA37757 (mouse), AAA41299 (rat), P_ (Drosophila), CAA87397 (yeast), Q94465 (Dictyostelium), CAA45365 (E. coli); for (B) PTPS are Q03393 (human), AAD15827 (mouse), P27213 (rat), P48611 (Drosophila); and for (C) SR are P35270 (human), Q64105 (mouse), P18297 (rat), AAC60297 (Fugu), and AAD12760 (Drosophila). (D) GFRP are P_ (human), P_ (mouse), P_ (rat). 505

12 Sequence information is available for the human, rat, and mouse GFRP, which share 94% sequence homology with conservation of the secondary structural elements (Figure 2D; GTPC is not regulated by GFRP in bacteria which do therefore not contain the corresponding protein). PTPS nly the recombinant rat liver enzyme could be crystallized, which yielded interpretable diffraction data (residues 7 144) [9]. Each subunit folds into a compact, single-domain α+β structure (Figure 1). The monomer consists of a sequential, four-stranded, antiparallel β-sheet (β-1 to β-4) with a 25-residue, helix-containing (α A ) insertion between β-1 and β-2 at the bottom of the molecule (Figure 2B). Strands 3 and 4 are connected via an α- helical turn. A segment between strands 2 and 3 forms a pair of antiparallel helices, α B and α C, layered on one side of the four β-sheets. Three monomers assemble into a trimer, forming a 12-stranded antiparallel β-barrel structure surrounded by a ring of α-helices. Crystallographic and experimental data revealed that the mammalian PTPS is homohexameric and dissociates into trimers. Two trimers arrange in a head-to-head fashion to form a barrel, the functional PTPS hexamer with an overall shape of 60 Å x 60 Å x 60 Å. Due to the relatively tilted order of the β-sheet, the pore in the trimer is conically shaped with a 6 12 Å diameter. In the hexameric enzyme, the pore has a smaller opening towards the trimer interface, and opens up also equatorially to the hexamer surroundings. The inside of the barrel accumulates a cluster of basic and aromatic residues stretching radially into the pore. Each subunit of the homohexamer contains one putative active site that is located at the interface of three monomers, two subunits from one trimer and one subunit from the other trimer. Each active site harbors three histidine residues (is23, is48, and is50 in the rat enzyme), with the e-atoms coordinating the binding of the Zn 2+ -metal. In the unliganded state a water molecule is bound as the fourth ligand. In a substrate complex structure the C1 and C2 side-chain hydrolysis of the dihydroneopterin ligand displaces the bound water, yielding a pentavalent co-ordination of the transition metal [10]. Complete PTPS sequences from several mammals and from Drosophila are available, while partial amino acid sequences from salmon are also known (Figure 2B). With the exception of the C-terminal extension in the fly enzyme, the overall amino acid identity is around 80%. All residues involved in substrate binding and catalysis are conserved among these sequences. The high degree of subunit conservation between the species implies that all these PTPS form a homohexameric structure. SR The 1.25 Å-crystal structure of the mouse SR in complex with ADP has been solved [11]. The 261 amino acids of the monomer fold into a single domain α/β-structure. A sev- 506

13 en-stranded parallel β-sheet, β A β G, in the center of the molecule is sandwiched by two arrays of three a-helices (α C, α B, α G and α D, α E, α F ) (Figure 2C). The association of two monomers to the active homodimeric SR leads to the formation of a four-helix bundle (Figure 1; helices α E and α F of each monomer). Due to the two-folded crystallographic symmetry of the homodimeric molecule, the parallel β-sheets in monomer A is in an antiparallel orientation relative to the β-sheet of monomer B enclosing an angle of 90. The overall dimensions of the SR dimer are 40 Å x 50 Å x 80 Å. The two substrate pockets bind sepiapterin (or 6-pyruvoyl-tetrahydropterin; compound 8 in Figure 1), and the cofactor ADP/ADP from opposite sides to the enzyme. Amino acid sequence comparison for the available SRs revealed a high degree of similarity among the mammalian enzymes (around 90%), as well as a high degree of conservation compared to the fish and fly enzymes (Figure 2C). It is thus assumed that the tertiary and quaternary structures of SR are conserved throughout these species. From both amino acid sequence and three-dimensional structure, SR can be assigned to the family of short-chain dehydrogenases/reductases. They all use AD() or ADP() as the cofactor and contain a strictly conserved Tyr-X-X-X-Lys sequence motif (Tyr171 and Lys175 in the mouse SR sequence). Comparison of the three biosynthetic enzymes There is no sequence identity among the three B 4 -biosynthetic enzymes. owever, a comparison of polypeptide secondary structure revealed that the C-terminal domain of GTPC shows a similar subunit fold with PTPS, and also with neopterin aldolase, neopterin epimerase, and uroate oxidase, whereas the overall structure of SR is entirely different [21, 22]. The tertiary structure of GTPC and PTPS revealed for both enzymes a central pore for which the function is still unknown; at least it seems not to be directly involved in catalysis and substrate binding. A speculative but attractive role for these cavities might be channeling of the unstable pterin intermediates from one enzyme to the other (pterin compounds 5 and 8). A potential substrate channeling would imply that the biosynthetic enzymes must interact and thus form some type of a super-complex. Such physical interaction has not yet been observed. In terms of complex formation with their natural substrates, all three enzymes show the same active site architecture. Furthermore, this pterin-binding motif unit is quite similar to the GTP-binding motif in small GTP-binding proteins [11]. 507

14 Genes encoding the biosynthetic enzymes The corresponding genes for all four proteins, including GFRP, are known from human and mouse, in some instances also from other organisms. An overview of gene names, chromosomal location, and exon content for the human genes is presented in Table 1. Enzyme (EC number) Gene Chromosomal o. of o. of amino Size location exons acid residues kda x no. of subunits GTPC ( ) GC1 14q x 10 GFRP GCFR 15q x 5 x 2 PTPS ( ) PTS 11q x 6 SR ( ) SPR 2p x 2 PCD/DCo ( ) PCBD 10q x 4 DPR ( ) QDPR 4p x 2 Table 1: uman B 4 -metabolizing enzymes. For references see text or the CBI database ( ote that PCD/DCo is homotetrameric (α 4 ) only as a carbinolamine dehydratase, whereas in complex with F-1α it is heteroterameric (α 2 β 2 ). Furthermore, a second copy for PCD/DCo, DCo2, was identified (for details see text). GFRP forms two pentameric complexes with the decameric GTPC. GC1 uman and mouse GTPC is encoded by a single copy gene, GC1, and is composed of 6 exons spanning about 30 kb [23]. The location of the 5 introns in mouse and human genes is at identical sites. In the fruit fly Drosophila, GTPC is encoded by the Punch locus. There, three out of four introns have identical positions (introns 2 4 [24]), whereas the single intron present in the Dictyostelium discoideum gene is inserted at a different site [25]. The yeast gene has no intronic sequences [26]. Although complete cdas for the GTPC were isolated from many diverse organisms (see for instance [27]), information on gene organization is mostly missing. Alternative splicing was observed for human exons 5 and 6, and Drosophila exon 1. Up to four types of GTPC-cDAs with different 3 ends were isolated from human liver or cell lines [28, 29]. Type 1 cda with 250 codons has the longest coding region and the greatest similarity to that reported from rat and mouse [30, 31]. Furthermore, the full-length type 1 GTPC-cDA was isolated from a pheochromocytoma cda library [32]. Alternative usage of the splice acceptor site within exon 6 generates the shorter type 2 mra [23, 33]. Type 3 mra contains, besides exon 1 5, an extension of exon 5 due to non-usage of the splice sites from intron 5. The type 4 mra variant is spliced within exon 6 without affecting the reading frame. Putative proteins derived from type 2 and 3 mras miss the C-terminal 40 amino acids- 508

15 containing residues involved in pterin binding and catalysis, and are highly conserved compared to the E. coli enzyme (see Figure 2A). In agreement with this is the observation that individual expression of recombinant proteins from these cdas in bacterial cells yielded GTPC activity only with the type 1 cda. It was further demonstrated that the type 2 mra exerts a dominant-negative effect on the wild-type cda, similar to the effect of some GC mutants [34], and may thus contribute to the regulation of B 4 production in specific cells [33]. Drosophila contains two 5 splice variants for the GTPC mra. These alternative exons confer distinct -terminal domains to each predicted protein and cannot be aligned with the -termini of any other GTPC. It was speculated that different GTPC isoforms in Drosophila and in higher eukaryotes might have specific subcellular and/or tissue distribution, and that alternative isoforms could associate and respond to different regulatory or modifying proteins. So far there is no experimental evidence for such regulatory diversity. owever, it was observed that two different complex forms of GTPC activities exist in human liver upon separation by molecular mass (400 kda and 600 kda). In this case both forms contained the same protein subunit as component with the identical mass as judged by SDS-PAGE analysis [35]. The transcription start sites and analysis for the upstream promoter sequences from the human, rat, and mouse GC1 genes are available. Functional cis-acting elements for basal, camp, and RF4/F-Y dependent expression, including a CRE-Sp1-CCAAT-box were described within the first 150 bp upstream of the cap sites in these promoters [36 41]. owever, the sequences neither revealed typical sites known to be involved in for instance interferon-γ signal transduction, nor was any response observed upon interferon-γ treatment of transfected human cells with corresponding reporter constructs. Since it is well established that GTPC gene expression can be induced by interferon-γ (and other compounds; see below) in various rodent and human cells (see below), further studies are necessary to characterize the mammalian GC1 promoter. PTS rganization of the human and mouse PTS genes is highly conserved. They are both spanning a region of 6 7 kb of genomic DA and contain six exons with their introns at identical positions [42 44]. riginally, a splicing polymorphism that occurred in at least some human cell types leading to skipping of the 23 bp exon 3, and thus aberrant protein expression was detected in normal controls (and in patients). onetheless, all cells and tissues investigated appeared to have expressed the same human PTPS-mRA. owever, a more detailed study revealed that the amount of exon-3-containing mra was correlated closely to PTPS activity, concluding that exon 3 skipping during transcript splicing is a major cause or regulator of the low PTPS protein expression in different human cell types (see also below) [45]. The human but not the mouse genome contains in addition a retropseudogene, PTS-P1, located on chromosome 9q13. In this retropseudogene, the 5 25 codons and an internal fragment of 23 bp corresponding to exon 509

16 3 (codons 54 61) are entirely absent. The overall similarity to the 3 portion of the PTPScDA is 74%. The Drosophila purple gene produces two PTPS-mRAs: a head specific one with three introns expressed from a proximal promoter and a constitutive mra with an additional intron in the 5 non-translated region expressed from a distal promoter [46]. The reading frame encoding PTPS from both mras is the same. When compared to the human and mouse intron-exon organization, the first two introns in the Drosophila genecoding sequence have identical positions, and the third Drosophila intron is located two codons upstream of the corresponding human intron 5 position. o information is thus far available on transcriptional start sites and corresponding promoter analyses for the mammalian PTS genes. SPR Genomic organizations of the mouse and human SPR genes encoding SR are very similar [47, 48]. They both span a region of 4 5 kb and the reading frames are split into 3 exons. o alternative splice variants have been observed. nly the mouse harbors a genomic pseudogene (Sprp) containing exons 1 and 2 plus the intervening and partial flanking sequences for these two exons with an overall similarity to the functional SPR gene of 82%. The single gene for SR in Drosophila contains no introns and encodes a single transcript of 1.4 kb that is present in heads and bodies of adults [49]. In all three species, the transcriptional start sites (+1) were determined for the SPR genes [47, 48]. o TATA-like sequences and no CAAT-box motifs were found in the upstream vicinities of the +1 sites. Furthermore, for the mouse SPR fusion studies with a reporter gene were conducted revealing that the promoter contains a sequence between 83 and 51 upstream of the transcriptional start site that is essential for expression. GCFR The cdas for the rat and human GFRP proteins were described [50, 51], but cdas and/or genes structures from other organisms are also available on the CBI database (see also Figure 2D). The corresponding human GCFR gene, which is located on chromosome 15q15, contains 3 exons and spans a region of roughly 4 kb (accession number U78190). Although promoter elements have not yet been described, transcription analyses in murine or human tissues by orthern blots (or RT-PCR) revealed the presence of up to three different GCFR-specific transcripts in all samples investigated, including brain [51, 52]. 510

17 CpG abundance An increased abundance of the dinucleotide CpG in a DA sequence of at least 200 bp spanning exon 1 and the transcriptional start site in comparison with the residual DA sequence of a gene are known to be typical for housekeeping genes (C+G content >50%, frequency of observed vs. expected CpG =0.6; [53]). Such a CpG content analysis with the human and mouse B 4 -biosynthetic genes revealed, as expected, that GC1 is a regulated gene, whereas PTS and SPR are predicted to be constitutively expressed (not known for GCFR; conducted with the CpGPlot program by Gardiner-Garden and Frommer; [1, 54]). owever, more detailed analyses are required to better define the promoters for the mammalian B 4 -biosynthetic genes. Regulation of enzyme expression and activity Regulation of B 4 biosynthesis appears to be complex, and an integrated picture of the signal transduction and control pathways does not yet exist. Depending on the cell or tissue type, all enzymes are thought to be constitutively expressed, such as in the liver or in some brain regions. In other tissues, enzyme expression can be induced or is completely absent. Expression of GTPC at least is inducible, and PTPS activity can be elevated to some extent. Moreover, post-translational modification(s) of all three biosynthetic enzymes and regulation of GTPC by the GFRP may modulate enzymatic activities. A detailed review on the regulation of B 4 -biosynthetic enzymes can also be found in [2]. GTPC The committing step for B 4 biosynthesis is the major controlling point for cofactor biosynthesis. GTPC activity can be regulated at the transcriptional and post-translational levels and by the GTPC-interacting protein GFRP that modulates enzyme activity. In tissues like liver and many brain regions that synthesize monoamine neurotransmitters, GTPC is constitutively present [55], whereas expression is inducible in virtually all other cell types or organs [2]. The regulation on the transcriptional level by various immune stimuli such as cytokines, phytohemagglutinin, and endotoxin (lipopolysaccharide) in a cell- and tissue-specific mode is probably predominant. Cytokines such as interferon-γ, tumor necrosis factor-α, stem-cell factor (or kit ligand), interleukin-1β, glial cell linederived neurotrophic factor (GDF), or a specific combination of these, induce GTPC gene expression in vitro and/or in vivo in various human and rodent cells, including T-lymphocytes, macrophages, monocytes, fibroblasts, endothelial cells, smooth-muscle cells, bone marrow-derived mast cells, mesangial cells, glioma cells, rat lungs, and locus coeruoleus in mouse brain (for review see [2]; [56 59]). Some of these stimulatory agents, for instance lectin, may act in an indirect way by first triggering cytokine release from T- lymphocytes, which then stimulates GTPC activities in other blood cells. In humans, a 511

18 biochemical consequence of immunostimulation is the excretion of both neopterin and 7,8-dihydroneopterin by activated macrophages and consequently accumulation in plasma and urine. A physiological function for these compounds has not been established. Lipopolysaccharide-treated rats show de novo expression and increased enzyme activity of GTPC in brain, liver, spleen, and adrenal gland. In cultured dopamine neurons of the hypothalamus and mesencephalon, or in the human neuroblastoma cell line SK--BE, but also in other cell types, camp and depolarization of membrane potential were found to stimulate GTPC-mRA expression [39, 60]. Increased levels of GTPC mra were also observed in peripheral and central neurons upon treatment with the catecholaminedepleting drug reserpine [61], and in brain catecholaminergic regions with addition of estrogen [62], or by GDF in primary dopaminergic neurons [59]. Furthermore, stimulation of GTPC-mRA expression was also reported upon phenylalanine treatment of UVEC cells [52], and arginine or hydrogen peroxide ( 2 2 ) in endothelial cells [63, 64]. n the other hand, anti-inflammatory cytokines, for instance IL-4, IL-10 and TGF-β, down-regulate GTPC expression [2, 63]. Glucocorticoids or cyclosporin A were reported to down-regulate GTPC gene expression, but also to have the opposite effect, i.e. to increase mra expression and enzyme activity (thereby decreasing the is) [57, 65]. pposing effects of glucocorticoids (and potentially other compounds) may depend on co-treatment with cytokines, and/or cell types [66] (for more detailed overiews see [63, 67]). Post-translational processing of GTPC involves cleavage of the -terminal 11 amino acids for at least the rat liver enzyme and protein phosphorylation [30, 68 70]. Whereas a regulatory effect for the -terminal processing is not known, phosphorylation modulates enzyme activity: (1) Protein kinase C stimulators like phorbol ester, platelet-derived growth factor, and angiotensin II or specific inhibitors caused an increase or reduction, respectively, of GTPC phosphorylation. (2) Concomitantly, phosphorylation coincides with elevated enzyme activity and an increase of cellular B 4 levels. (3) In vitro phosphorylation with purified enzymes demonstrated that GTPC is modified by casein kinase II and/or protein kinase C. The primary amino acid sequence of GTPC reveals several conserved sites for potential phosphorylation by casein kinase II. owever, only one serine residue (Ser167 in the rat and mouse sequences), which is conserved between the human, rat, mouse and Drosophila enzymes, was proposed to be a potential target site for protein kinase C. Further effectors for GTPC enzymatic regulation are its substrate GTP, the pathway-end product B 4 (and other reduced pteridines), phenylalanine, and calcium ions. B 4 as feed-back inhibitor and phenylalanine as feed-forward stimulator modulate enzymatic activity, with different affinities of these effectors, via two homopentameric GFRP that form a sandwich with the homodecameric GTPC [50, 71 74]. A physiological consequence of GFRP action is the high plasma B 4 concentrations observed in patients with hyperphenylalaninemia caused by PA deficiency [71]. Each GFRP has a cavity at the interface with GTPC that binds phenylalanine and enhances the association of the GTPC-GRFP 512

19 complex. Furthermore, this phenylalanine locks the enzyme in the active state [19, 74]. GFRP-mediated end-product inhibition by B 4 can be reverted to an active form by phenylalanine; however, the binding site for B 4 and its mechanism for GFRP-mediated feed-back inhibition are not clear. A functional high-affinity, EF-hand-like calcium-binding site was reported specifically for the eukaryotic GTPC. It was speculated that mutant GTPC with loss of calcium ion binding capacity might be a primary cause for doparesponsive dystonia (see below), however, the role for calcium binding remains unclear [75]. GFRP-mRA studies by orthern blot analysis and in situ hybridization revealed that the expression pattern in rat tissues correlates with that of GTPC, i.e. GFRP is expressed in peripheral organs like liver and heart, and also in the brain [50, 76]. Proinflammatory stimuli such as bacterial lipopolysaccharide or interferon-γ stimulate expression and activity of GTPC, while down-regulating expression of GFRP, as shown in human cell lines or in vivo with rats. Moreover, down-regulation of GFRP-mRA occurred also in presence of phenylalanine, thus rendering B 4 biosynthesis independent of metabolic control by phenylalanine [51, 52]. Furthermore, inhibition of GTPC activity by 2,4-diamino-6- hydroxypyrimidine (DAP) was shown using recombinant proteins to be mediated by GFRP binding to its cognate GTPC enzyme. Thus 2,4-diamino-6-hydroxypyrimidine is not competing with GTP at the substrate binding site of GTPC [77]. PTPS The PTPS-mRA is considered to be constitutively expressed but the protein or its activity is not ubiquitously present, at least not in higher animals; moreover, PTPS expression is strictly regulated in some cell types by an unknown mechanism at the transcriptional level (see below). Following immunostimulatory induction of GTPC by cytokines, PTPS becomes the rate-limiting step in B 4 biosynthesis, at least in man, where PTPS activity is much lower than in rodents [78, 79]. Whereas GTPC activity can be stimulated up to 100-fold in cytokine treated cells, PTPS activity remained unaffected in some experiments but was stimulated in others. In any case, PTPS activity was reported to be maximally elevated by a factor of 2 to 4 [45, 80 84]. Elevation of PTPS-mRA (and GTPC-mRA; see above) by a factor of 3-4 was observed in rat adrenal glands following reserpine treatment [85]. A regulatory mechanism for PTPS expression at the mra (or pre-mra) level was observed for various human cell types, including blood-derived cells, fibroblasts, endothelial cells, and other cell lines [45]. Although normal levels of PTPS mra could be detected in all cells, PTPS protein presence and enzymatic activity were barely detectable in some cells. A detailed study of PTPS-mRA revealed the presence of two types of PTPS transcripts: the normal functional mra and a species lacking the 23 bp-exon 3, which results in a premature stop codon. Furthermore, the ratio of exon-3-containing to exon-3-lacking PTPS-mRA closely correlated to PTPS activity, and from this it was concluded that exon 3 skipping is a major cause of the low PTPS protein expression in these 513

20 cells. Furthermore, mra species lacking exon 3 were detected also in human brain cda libraries and in a PTPS-deficient patient [42, 43, 86]. The significance of such a post-transcriptional event as potential regulatory mechanism controlling the expression level of PTPS also in other cells remains to be clarified ,6,7,8-Tetrahydrobiopterin + 2 PA 12 2 Phe, Tyr,Trp +AD - AD + DPR PA Tyr, L-Dopa, 5--Trp 14 2 q-dihydrobiopterin - 2 PCD/DCo 13 2 Tetrahydrobiopterin- 4a-carbinolamine non-enzymatic ,8-Dihydrobiopterin Primapterin (7-dihydrobiopterin) Scheme 3: Reaction mechanism for the B 4 -regenerative pathway With regard to post-translational modifications, PTPS isolated from rat liver was found to be -terminally processed by having the first four amino acids removed [87]. Furthermore, human PTPS was also shown to be subject to regulatory phosphorylation at Ser19, whereby the cyclic-gmp-dependent protein kinase type II seemed to be responsible for 514

21 the phosphoserine modification [88]. The molecular basis for the at least three-fold stimulation of the phosphorylated PTPS versus the non-modified protein when tested in CS- 1 cells is not yet understood. evertheless, phosphoserine modification appears to be essential as a phosphorylation-deficient mutant of PTPS was identified from a patient with a defect in B 4 biosynthesis [89]. It was speculated that in cultured dopamine neurons, where B 4 biosynthesis can be stimulated by camp, the observed short-term increase in B 4 levels may be attributed to camp-dependent phosphorylation of PTPS [85]. SR SR seems to be ubiquitously expressed, as it was found in all mammalian tissues, and its expression remains unaffected by cytokine treatment. A recent examination of the expression pattern in brain revealed that SR is expressed in all regions, in contrast to GTPC or PTPS which are basically localized to monoamine neurons [55, 90] (A. Résibois and Thöny B, in preparation). Several alternative aldose-ketose reductases can substitute for SR (see below). Alternatively, SR may also have other functions as an aldose-ketose reductase besides being involved in B 4 biosynthesis (for instance detoxification [91]). Biochemical evidence for alternative reductases comes from the AKR family of enzymes that commit the last biosynthetic steps from 6-pyruvoyl-tetrahydrobiopterin to B 4 (compounds 8 to 11 in Scheme 2) [16]. The discovery of patients with defective SR who present with monoamine neurotransmitter deficiency but without hyperphenylalaninemia is probably the most compelling proof for the in vivo potential of alternative aldose-ketose reductases replacing SR in the B 4 biosynthesis [17, 92]. From -terminal sequence analysis of SR purified from rat erythrocytes it is known that the protein begins with an -acetyl-methionyl residue [93]. Furthermore, SR was reported to be phosphorylated in vitro by calmodulin-dependent protein kinase II and protein kinase C at one or several sites [12, 94]. Phosphorylation did not modify the kinetic properties of the purified rat and human enzymes, but it became more sensitive to calciumactivated proteases, a situation reminiscent of tyrosine hydroxylase [95]. Localization of biosynthetic enzymes As mentioned above, immunohistochemical staining with anti-gtpc and anti-ptps antibodies in rat brain revealed that the expression pattern is highly specific and co-localization was generally found with aromatic amino acid hydroxylases [55] (and A. Résibois and Thöny B, in preparation). In contrast, SR was detected in all brain regions examined, pointing towards additional functions as reductase besides being involved in B 4 biosynthesis [90]. In addition, studies with all 3 B 4 -biosynthetic enzymes, GTPC, PTPS, and SR, revealed some prominent nuclear localization in specific but various cell types, including neurons [96]. 515

22 Regeneration of B 4 Regeneration of B 4 is an essential part of the phenylalanine hydroxylating system (see also Cofactor functions ). During the catalytic event of aromatic amino acid hydroxylases, molecular oxygen is transferred to the corresponding amino acid and B 4 is oxidized to B 4-4a-carbinolamine (Scheme 3) [97, 98]. Two enzymes are involved in its subsequent dehydratation and reduction to B 4 : pterin-4a-carbinolamine dehydratase (EC ; PCD) and dihydropteridine reductase (EC ; DPR). Enzymatic recycling of B 4 is essential for phenylalanine metabolism (1) to ensure a continuous supply of reduced cofactor, and (2) to prevent accumulation of harmful metabolites produced by rearrangement of B 4-4a-carbinolamine. The primary structure of PCD is identical with a protein of the cell nucleus that has transcription function, named dimerization cofactor (DCo) of hepatocyte nuclear factor 1α (F-1α), recently reported to have general transcriptional function [99 101]. In the following, PCD will be designated as the dual-function protein PCD/DCo. Furthermore, a functional homologue, DCo2, was identified, which partially complements PCD/DCo mutants in mice and man [102,103]. Reaction mechanism of the regenerative pathway The dehydratation of B 4-4a-carbinolamine (compound 13), the first product of the reaction of aromatic amino acid hydroxylases (Scheme 3), is catalyzed by the enzyme PCD/DCo. The human cytoplasmic PCD/DCo, whose sequence is identical to that of the rat protein, is a homotetramer with a molecular mass of 11.9 kda per subunit (Figure 3A) [100,101]. Using chemically synthesized pterin-4a-carbinolamine it has been demonstrated that the enzyme shows little sensitivity to the structure or configuration of the 6- substituent of its substrate, and to the 4a(R)-and 4a(S)-hydroxy stereoisomers [104]. bviously, the binding pocket has a relatively high degree of flexibility and might not be designed to recognize only B 4-4a-carbinolamine. X-ray crystal structures of the tetrameric enzyme complexed with the product analogue 7,8-dihydrobiopterin (compound 16) revealed four active sites harboring three essential and conserved histidines (is61, is62, is79 in the human or rat enzyme; see Figure 3A; [105]). Detailed enzymatic studies on the stereospecificity and catalytic function revealed a dehydratation mechanism in which the three histidines in PCD/DCo are crucial for activity [ ]. A is61 Ala/is62 Ala double mutant was fully inactivated and showed a significantly increased dissociation constant of quinonoid 6,6-dimethyl-7,8-dihydropterin. Moreover, is61 and is79 act as general acid catalysts for the stereospecific elimination of the 4a(R)- and 4a(S)-hydroxyl groups, respectively (see Figure 3A). The role of is62 is primarily to bind substrate, with an additional component of base catalysis [107]. The quinonoid dihydrobiopterin (compound 14) product is a strong inhibitor of PCD/DCo with a K I -value of about one half of its respective K m -value, and no inhibition was observed with 7,8-dihydrobiopterin (compound 16) [104]. Furthermore, PA is not inhibited by its cofactor product, B 4-4a-carbinolamine, but by primapterin (compound 15). In the 516

23 absence of PCD/DCo, dehydratation of B 4-4a-carbinolamine occurs also non-enzymatically, but at a rate that is, at least in the liver, insufficient to maintain B 4 in the reduced state [109]. As a consequence, liver PCD/DCo deficiency in man causes B 4-4a-carbinolamine to be rearranged via a spiro structure intermediate to dihydroprimapterin (7-substituted dihydrobiopterin; compound 15) that is excreted in the urine [110,111]. Figure 3: Amino acid sequence alignments of B 4 regenerating enzymes. Alignments were generated using the Clustal W program [216]. Identical residues and residues with functional similarities (V/L/I, R/K, M/L, D/E) are highlighted with shaded boxes. Residues in PCD/DCo involved in catalysis and/or substrate binding are marked with a triangle. For DPR, the potential active site pocket residues are marked with a triangle. The accession numbers for the amino acid sequences for (A) PCD/DCo are P80095 (human), A47189 (rat), P61458 (mouse), AA06395 (chicken), AAC25196 (Drosophila); BAA17842 (cyanobacteria), AAC06420 (Aquifex), P43335 (Pseudomonas aeruginosa), DCo2 Q905 (human) for (B) DPR are P09417 (human), P11348 (rat). 517