q 2002 Elsevier Science Ltd. All rights reserved

Transcription

1 doi: /s (02) available online at on Bw J. Mol. Biol. (2002) 320, Crystal Structure of the Ternary Complex of the Catalytic Domain of Human Phenylalanine Hydroxylase with Tetrahydrobiopterin and 3-(2-Thienyl)-L-alanine, and its Implications for the Mechanism of Catalysis and Substrate Activation Ole Andreas Andersen 1, Torgeir Flatmark 2 and Edward Hough 1 * 1 Department of Chemistry University of Tromsø, N-9037 Tromso, Norway 2 Department of Biochemistry and Molecular Biology University of Bergen Årstadveien 19, N-5009 Bergen Norway *Corresponding author Phenylalanine hydroxylase catalyzes the stereospecific hydroxylation of L-phenylalanine, the committed step in the degradation of this amino acid. We have solved the crystal structure of the ternary complex (hpheoh Fe(II) BH 4 THA) of the catalytically active Fe(II) form of a truncated form (DN1 102/DC ) of human phenylalanine hydroxylase (hpheoh), using the catalytically active reduced cofactor 6(R)-L-erythro- 5,6,7,8-tetrahydrobiopterin (BH 4 ) and 3-(2-thienyl)-L-alanine (THA) as a substrate analogue. The analogue is bound in the second coordination sphere of the catalytic iron atom with the thiophene ring stacking against the imidazole group of His285 (average interplanar distance 3.8 Å) and with a network of hydrogen bonds and hydrophobic contacts. Binding of the analogue to the binary complex hpheoh Fe(II) BH 4 triggers structural changes throughout the entire molecule, which adopts a slightly more compact structure. The largest change occurs in the loop region comprising residues , where the maximum r.m.s. displacement (9.6 Å) is at Tyr138. This loop is refolded, bringing the hydroxyl oxygen atom of Tyr Å closer to the iron atom and into the active site. The iron geometry is highly distorted square pyramidal, and Glu330 adopts a conformation different from that observed in the hpheoh Fe(II) BH 4 structure, with bidentate iron coordination. BH 4 binds in the second coordination sphere of the catalytic iron atom, and is displaced 2.6 Å in the direction of Glu286 and the iron atom, relative to the hpheoh Fe(II) BH 4 structure, thus changing its hydrogen bonding network. The active-site structure of the ternary complex gives new insight into the substrate specificity of the enzyme, notably the low affinity for L-tyrosine. Furthermore, the structure has implications both for the catalytic mechanism and the molecular basis for the activation of the full-length tetrameric enzyme by its substrate. The large conformational change, moving Tyr138 from a surface position into the active site, may reflect a possible functional role for this residue. q 2002 Elsevier Science Ltd. All rights reserved Keywords: phenylalanine hydroxylase; tetrahydrobiopterin; thienylalanine; conformational change; protein crystallography Abbreviations used: 4a-OH-BH 4, 4a-hydroxy-tetrahydrobiopterin; BH 2, L-erythro-7,8-dihydrobiopterin; BH 4,6(R)-L-erythro-5,6,7,8-tetrahydrobiopterin; HPA, hyperphenylalaninemia; hpheoh, human phenylalanine hydroxylase; htyroh, human tyrosine hydroxylase; L-Tyr, L-tyrosine; L-Phe, L-phenylalanine; PheOH, phenylalanine hydroxylase; PKU, phenylketonuria; rpheoh, rat phenylalanine hydroxylase; rtyroh, rat tyrosine hydroxylase; THA, 3-(2-thienyl)-L-alanine; TyrOH, tyrosine hydroxylase. address of the corresponding author: edward@james.chem.uit.no Introduction The non-heme iron enzyme phenylalanine hydroxylase (PheOH, phenylalanine 4-monooxygenase, EC ) catalyzes the hydroxylation of the essential aromatic amino acid L-phenylalanine (L-Phe) to L-tyrosine (L-Tyr) in the presence of the specific pterin cofactor 6(R)-L-erythro-5,6,7,8- tetrahydrobiopterin (BH 4 ) and dioxygen. The reaction is the rate-limiting step in the degradation of /02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved

2 1096 Ternary Complex of Phenylalanine Hydroxylase Figure 1. Stereo picture of the electron density at (a) the THA-binding site and (b) the BH 4 -binding site in the ternary complex. Blue electron density is from s-weighted 2F o 2 F c maps at 1.2s while red omit electron density is from s-weighted F o 2 F c maps at 2.7s (a) omitting THA and the side-chain of Tyr138 and (b) omitting BH 4, the side-chain of Glu330 and Wat2. The Figure was produced using BOBSCRIPT. 57 L-Phe to carbon dioxide and water. 1 Inborn errors that reduce or destroy the activity of the enzyme are responsible for the human autosomal recessive disease phenylketonuria (PKU)/hyperphenylalaninemia (HPA). The disease causes elevated concentrations of L-Phe in the blood, which can impair the normal development of the brain and cause severe mental retardation. In most of Europe, approximately 1 in 10,000 live births reportedly has the disorder 2 and more than 400 different mutations are associated with PKU/HPA. 3 Most of the mutations are found in the catalytic domain 3 and they demonstrate different clinical, metabolic and enzymatic phenotypes. 4,5 Recent crystallographic studies on human phenylalanine hydroxylase (hpheoh) 6 9 and rat phenylalanine hydroxylase (rpheoh) 10 have made it possible to define the structural phenotypes of the different genotypes. 11 A limitation in the assignment of the structural phenotypes has been that they have been based on the structures of catalytically inactive Fe(III) forms of the enzyme, which also lack structural information on substrate binding. Following our recently solved crystal structure of the catalytically active Fe(II) form of the truncated form DN1 102/DC hPheOH and its binary complex with the reduced pterin cofactor (BH 4 ), 12 we now present the crystal structure of a ternary complex with the substrate analogue 3-(2-thienyl)-L-alanine (THA). THA is a substrate for rpheoh 13 and hpheoh 14 and binds competitively to L-Phe at the active site. 15 Binding of the substrate analogue also triggers a conformational change similar to that observed upon binding of L-Phe. 14,16,17 The structure reveals the binding sites of the pterin cofactor and the substrate under near-turnover conditions, i.e. in the absence of dioxygen, and provides new insights into the substrate specificity and catalytic mechanism of the enzyme. It shows that substrate binding triggers a substantial structural change in the catalytic domain, particularly in the active-site region. This change may

3 Ternary Complex of Phenylalanine Hydroxylase 1097 Figure 2. An illustration of the BH 4 and THA-binding sites in the ternary complex. The Figure was produced using Ligplot, 58 and edited using CorelDRAW 9.0. represent the epicenter of the global conformational transition and catalytic activation that occurs in the full-length tetrameric enzyme upon substrate binding. 18 Results Well defined crystals of the binary hpheoh Fe(II) BH 4 complex were treated with the substrate analogue THA by adding solid THA to the crystallization drops. The whole procedure of crystallization, post-crystallization diffusion soaking in THA, flash-cooling in liquid nitrogen and mounting of crystals was carried out anaerobically as described. 12 When observed in the microscope, the crystals appeared to be unaffected by the THA soaking, but the diffraction pattern revealed a mosaicity that was two to three times higher than that of the binary crystals. All three axes of the unit cell were 1 2 Å shorter than for the binary complex, and useable data were obtained to 2.5 Å. It should be noted, however, that the binary crystals of the enzyme were found to deform/disintegrate when exposed to L-Phe. Data collected for cofactor-free crystals soaked in L-Phe revealed the same high level of mosaicity as that observed for hpheoh Fe(II) BH 4 THA but processing of these data was unsuccessful. The reason why we have been unsuccessful using L-Phe as substrate is not clear, but does suggest that possible lattice or structural changes may simply have exceeded that tolerable within the crystals. Co-crystallization of hpheoh substrate/substrate-analogue complexes using previously known crystallization conditions 19 failed, as did experiments to find new crystallization conditions for such a complex. The structure was refined to a final R work and R free of 22.0% and 26.7%, respectively. The mean error of the atomic positions was determined to 0.29 Å using the s A method. 20 The final model contains 307 residues, a ferrous iron, 39 water molecules, one THA molecule and the reduced cofactor (BH 4 ). The electron density for both THA and BH 4 is very good (Figure 1(a) and (b)), as is the electron density for most of the amino acid side-chains with the exception of the loop residues (see below) and a few surface-located side-chains. Binding of the substrate analogue 3-(2-thienyl)- L-alanine (THA) All atoms of the substrate analogue THA are well defined, consistent with the low B-factors (below 22 Å 2 ) estimated during refinement. THA binds in the second coordination sphere of the catalytic iron atom, with the five-membered thiophene ring packing against the imidazole group of the iron ligand His285 (Figures 1(a) and 2) with an average interplanar distance of 3.8 Å. This distance

4 1098 Ternary Complex of Phenylalanine Hydroxylase Figure 3. Stereo picture of the ternary hpheoh Fe(II) BH 4 THA complex (blue/cyan) superimposed on the binary hpheoh Fe(II) BH 4 complex 12 (red/orange) (PDB entry 1J8U). Black ball-and-stick models of BH 4, THA and iron are shown for the ternary structure. BH 4 in the binary structure is omitted for clarity. The highest r.m.s. displacement (see the text) occurs in loop residues coloured cyan in the ternary complex and orange in the binary complex. The r.m.s. maximum is located at Tyr138. The Figure was produced using MOLSCRIPT. 59 is similar to that observed for the phenyl ring of L-Phe in the ternary complex of hpheoh as determined by a combined nuclear magnetic resonance (NMR) and molecular docking analysis. 15 However, the hydrogen bonding pattern of the main-chain THA is somewhat different from that found for L-Phe in that structure. Thus, the amino N forms a water-mediated hydrogen bond to Tyr277 (3.2 and 3.1 Å), and hydrogen bonds to Thr278 O (2.8 Å) and a water molecule (2.8 Å), which in turn is hydrogen bonded to Gly346 O (2.8 Å), Glu353 O 12 (2.7 Å) and possibly Ser350 O g (3.4 Å). THA OT1 is hydrogen bonded to Arg270 N h1 (3.3 Å), Thr278N (3.0 Å) and possibly Thr278 O (3.1 Å) while THA OT2 is hydrogen bonded to Arg270 N h2 (2.9 Å), Ser349 O g (2.4 Å) and possibly Ser349 O. Hydrophobic contacts are formed from Gly346 C a (3.8 Å) and Phe331 C z (3.7 Å) to THA C d, from Phe331 C z (3.8 Å) to THA C 12 and from Ser350C b to THA C (3.5 Å) (Figure 2). Binding of the substrate analogue 3-(2-thienyl)- L-alanine (THA) triggers large-scale structural changes All previous crystal structures of the binary complexes of the double truncated form ( catalytic domain ) of hpheoh 8,9,12 can be superimposed onto the non-liganded structure 6 (PDB entry 1PAH) with r.m.s. deviations for main-chain atoms of between 0.21 and 0.31 Å. By contrast, a superposition of the ternary complex hpheoh Fe(II) BH 4 THA onto the non-liganded structure reveals an r.m.s. deviation for main-chain atoms of 2.2 Å, and superposition onto other hpheoh structures 8,9,12 gives similar r.m.s. values. Thus, the ternary complex structure is significantly different from the substrate-free structures and demonstrates that substrate-analogue binding triggers large-scale structural changes. The ternary enzyme is slightly smaller compared to the binary and ligand-free structures of hpheoh due to a more compact packing. The average atomic distance to the centre of mass is 17.9 Å for the THA-bound structure compared to 18.1 Å for the binary and ligand-free structures. The superposition of the ternary hpheoh Fe(II) BH 4 THA complex onto the binary hpheoh Fe(II) BH 4 complex 12 (PDB entry 1J8U) is shown in Figure 3. In addition to general adjustments throughout the whole structure, part of the chain comprising residues (mostly loop residues) is substantially refolded. The largest displacement is observed for Tyr138 (r.m.s. displacement of 9.6 Å). The hydroxyl in this residue is displaced by 20.7 Å to a partially buried position in the active site (Figure 3), with its O h only 6.5 Å away from the iron atom, 5.7 Å away from BH 4 C4a and 3.7 Å away from THA C 11. The phenol ring is packed between Leu248 (closest distance 3.6 Å) and Val379 (closest distance 3.4 Å) forming a hydrophobic cluster (Figure 4), and Tyr138 O h forms an intramolecular hydrogen bond to a water molecule, which is also hydrogen bonded to O2 0 in BH 4 (Figure 2). In the substrate-free crystal structures of hpheoh, 6 9,12 Tyr138 is located on the surface of the protein with a solvent-exposed side-chain and appears to have no specific importance, except for a possible contribution to protein stability. 21 Tyr277 adopts a conformation slightly different from that in the substrate-free structures of hpheoh and its hydroxyl oxygen atom is displaced 6.3 Å compared with the binary hpheoh Fe(II) BH 4 complex to form a possible water-mediated hydrogen bond with THA N. Displacement of the pterin cofactor upon substrate binding In addition to the large motion of the region, including the reorientation of Tyr138 from a surface position to a location in the active site, a

5 Ternary Complex of Phenylalanine Hydroxylase 1099 Figure 4. Stereo picture of the packing of Tyr138 in the hydrophobic core at the active site. The green model illustrates the position of residues in the binary hpheoh Fe(II) BH 4 complex 12 when superimposed on the ternary structure based on conserved active-site residues (His285, His290 and Fe). The Figure was produced using MOLSCRIPT. 59 significant displacement was observed for the pterin cofactor BH 4 (Figures 2 and 5). Iron to pterin (C4a, O4 and N5) distances for all crystal and NMR structures of PheOH and tyrosine hydroxylase (TyrOH) are compared in Table 1. The pterin-binding site of the binary hpheoh Fe(II) BH 4 is similar to that found in the crystal structure of the binary complex (hpheoh Fe(III) BH 2 ) of hpheoh with the oxidized L-erythro-7,8-dihydrobiopterin (BH 2 ) cofactor (PDB entries 1LRM and 1DMW). The orientation of the pterin cofactor in these two structures is similar to that of the hpheoh Fe(III) BH 2 - L-Phe NMR and molecular docking structure and the present hpheoh Fe(II) BH 4 THA structure. However, the pterin cofactor is displaced significantly compared to the binary complexes. Superimposing the binary hpheoh Fe(II) BH 4 complex on the ternary hpheoh Fe(II) BH 4 THA complex based on conserved active-site residues (His285, His290 and Fe) revealed a mean displacement of 2.6 Å for BH 4 in the direction of Glu286 and iron upon THA binding (Figure 5). Iron distances are shortened from 5.9, 3.8 and 5.7 Å for C4a, O4 and N5, respectively, in the crystal structure of binary hpheoh Fe(II) BH 4 complex to 4.5, 3.4 and 3.7 Å in the current THA ternary structure (Table 1). However, the pterin cofactor is still not coordinated directly to the iron atom. This contrasts with the combined NMR and molecular modeling studies on hpheoh, in which the distance between BH 2 O4 and the iron atom was estimated to be 2.6 Å, and thus compatible with direct coordination to the iron atom. 15 However, this NMR/ molecular modeling structure of cofactor and substrate bound at the active site was modeled into the rigid structure of the non-liganded double truncated form of hpheoh and thus did not take into account any bound water molecules or Figure 5. Stereo picture of the binding site of BH 4 in the ternary hpheoh Fe(II) BH 4 THA structure. Side-chains for Leu248 and Leu249 are omitted for clarity. All potential hydrogen bonds to the pterin moiety are shown as dotted lines. The green model of BH 4 illustrates its position in the binary hpheoh Fe(II) BH 4 complex 12 when superimposed on the ternary structure using conserved active-site residues (His285, His290 and Fe). The Figure was produced using MOLSCRIPT. 59

6 1100 Ternary Complex of Phenylalanine Hydroxylase Table 1. Comparison of metal to pterin distances (Å) Phenylalanine hydroxylase Tyrosine hydroxylase X-ray hpheoh Fe(II) BH 4 THA X-ray hpheoh 12 Fe(II) BH 4 X-ray hpheoh 8 Fe(III) BH 2 NMR hpheoh Fe(III) BH 2 L-Phe 15 X-ray rtyroh 22 Fe(III) BH 2 NMR htyroh Fe(III) BH 2 L-Phe 60 Fe C4a Fe O ^ Fe N ^ possible conformational changes in the active site associated with substrate binding. The p-stacking interactions of the cofactor with Phe254 are, however, similar for all three crystal structures of hpheoh pterin complexes. The side-chain of Phe254 in hpheoh Fe(II) BH 4 THA is displaced about 1.9 Å in the same direction as the pterin displacement compared with the binary hpheoh Fe(II) BH 4 and hpheoh Fe(III) BH 2 structures, and the average interplanar phenyl-cofactor distance is 3.7 Å. The loop residues form the same pattern of direct hydrogen bonds to the pterin as in the binary hpheoh Fe(II) BH 4 and hpheoh Fe(III) BH 2 complexes (except the pterin O2 0 Ser251 O g bond, which is not present in the published hpheoh Fe(III) BH 2 crystal structure). 8 These residues are thus displaced about 2.6 Å (relative to the active site) in the same direction as the pterin displacement. Glu286 is not displaced significantly and its carbonyl group forms hydrogen bonds directly to N3 and O4 of BH 4 ; these two connections are water-mediated hydrogen bonds in the binary complexes hpheoh Fe(II) BH 4 and hpheoh Fe(III) BH 2. BH 4 C3 0 forms hydrophobic contacts with Leu255 C d1 (3.6 Å), Ser 251 C a (3.8 Å) and Phe254 C d2 (3.8 Å), while BH 4 C8a and C7 forms contacts with Leu248 C d1 (3.7 and 3.8 Å, respectively). The torsion angle between the hydroxyl groups in the dihydroxypropyl side-chain of BH 4 in the ternary complex is similar (2 598) to that in the binary hpheoh Fe(II) BH 4 complex (2658) and the ternary hpheoh Fe(III) BH 2 L-Phe NMR structure (2608), enabling the BH 4 O2 0 to make a hydrogen bond with the side-chain oxygen atom of Ser251. A crystal structure of the binary hpheoh Fe(III) BH 2 complex to 2.1 Å resolution (data not shown) was obtained. The electron density maps showed unambiguous positions for all cofactor atoms, including the dihydroxypropyl side-chain, and revealed that the angle between the hydroxyl groups of the side-chain is Thus, the same hydrogen bond is formed between O2 0 and Ser251 O g (2.7 Å) as for the hpheoh Fe(II) BH 4 complex. The ternary complex has a distorted square pyramidal, five-coordinated iron atom Previously determined crystal structures of amino acid hydroxylases 6 10,12,22,23 have shown that the iron ligands are consistently two histidine residues, a monodentate glutamic acid residue and a varying number of water molecules. The iron atom in the present ternary complex is fivecoordinated by two histidine residues, a bidentate glutamic acid residue, and a single water molecule (Wat2) in a highly distorted square pyramidal geometry with Glu330 O 11 as the axial ligand (Figures 1(b) and 2). The ligand iron distances are 2.3 Å (His285), 2.3 Å (His290), 2.6 Å (Glu330 O 11 ), 2.4 Å (Glu330 O 12 ) and 2.4 Å (Wat2). Both histidine residues and the glutamic acid residue have good electron densities with low B-factors (below 32 Å 2 ). The water ligand has a slightly higher B-factor (44 Å 2 ) and is displaced about 0.9 Å compared to Wat2 in the binary hpheoh Fe(II) BH 4 complex. No density appeared for either Wat1 or Wat3, which is in conformity with the 1 H NMR studies on full-length hpheoh, suggesting that at least one of the coordinating water molecules is displaced from coordination upon the binding of L-Phe at the active site. 24 Magnetic circular dichroism (MCD) studies on rpheoh further supports a five-coordinate square pyramidal Fe(II) site upon addition of pterin in the presence of L-Phe. 25 Discussion The present crystal structure of the ternary complex hpheoh Fe(II) BH 4 THA has given valuable new information related to the question of the substrate-binding site, the substrate specificity and the conformational transition (hysteresis) that occurs in the enzyme upon substrate binding. Furthermore, the structure has important implications for the catalytic mechanism and defines clearly the amino acid residues of the active-site crevice structure that are involved in the binding of pterin cofactor and substrate under near-turnover conditions (in the absence of dioxygen) as well as providing a structural explanation for the diseaseassociated PKU/HPA mutations related to these binding sites. The substrate specificity and substratebinding site The substrate specificity of PheOH has been studied extensively by Kaufman. 16 For maximum activity, a substrate including an unmodified alanine residue must be attached to an aromatic ring that may contain a number of subsitutions and still be hydroxylated as long as the alanine part is intact. Of particular interest was the finding that

7 Ternary Complex of Phenylalanine Hydroxylase 1101 Figure 6. Stereo picture of the substrate-binding site with (a) THA (green), the modelled L-Phe (red phenyl group) and L-mTyr (red phenyl group and blue hydroxyl oxygen atom) and (b) L-Tyr (blue; see the text). The Figure was produced using MOLSCRIPT (2-thienyl)-L-alanine (THA) is hydroxylated by PheOH 14,16 and that this analogue induces a comprehensive global conformational transition (and activation of the enzyme) similar to L-Phe, 14,16,17 although it has a slightly lower affinity of binding than L-Phe. 14 Interestingly, recent NMR studies on the double truncated form DN1 102/ DC hPheOH have demonstrated that L-Phe bound at the active site is displaced by THA. 15 Since THA binds competitively to L-Phe, 16,17 the present structure allows us to model the physiological substrate L-Phe into the active site (Figure 6(a)) assuming the position of the main chain of the substrates and the orientation of the ring structure (x 1 and x 2 angles) to be conserved. In this model, the phenyl group is positioned appropriately in the hydrophobic cluster, 3.6 Å from the phenyl group of Phe331, 3.7 Å from the side-chain of Trp326 as well as 3.7 Å from the C a atoms of Gly346 and Pro281. The side-chains of Phe331 and Trp326 are both displaced (about 2 Å and 3 Å, respectively) upon binding of substrate, resulting in hydrophobic contacts with its phenyl group. Thus, the modeled L-Phe was found to interact with four main residues at the active site, i.e. Arg270, Thr278, His285 and Ser349, as well as with Tyr277, Pro281, Trp326, Phe331, Gly346, Ser350 and Glu353, and interestingly, human missense single-point mutations related to PKU/HPA have been reported for eight of these residues (Table 2). It should be noted that one of the interactions with Arg270, the interaction with the side-chain of Ser349 and the interactions with Pro281 and His285 are present also in the NMR/molecular docking structure, 15 and the displacement of Trp326 to accommodate hydrophobic contacts with L-Phe was predicted in that study, whereas the other observed interactions are unique for the crystal structure. Furthermore, studies on chimeric forms of pterin-dependent aromatic amino acid hydroxylases have revealed that their substrate specificity is determined by the catalytic domain and that none of the chimeric enzymes, containing the catalytic domain of PheOH, were able to hydroxylate L-Tyr. 26 L-meta-tyrosine (L-mTyr), which is a substrate for PheOH, 16 was modeled into the binding site in the same manner as L-Phe (see above) (Figure 6(a)). Its hydroxyl oxygen atom could form a hydrogen bond (3.0 Å) with Tyr138 O h and its nearest carbon atom is Tyr138C 11 (3.2 Å). On the other hand, when L-Tyr (L-pTyr) was similarly modeled into the active site, its oxygen atom was only 2.5 Å away from the nearest carbon atom in the side-chain of Trp326. This steric hindrance could be minimized slightly by rotating the sidechain x-values (while conserving the position of the main-chain). The best manual fit was found by rotating x 1, 98, bringing the hydroxyl oxygen

8 1102 Ternary Complex of Phenylalanine Hydroxylase Table 2. Amino acid residues in the active-site crevice structure involved in the binding of pterin cofactor and substrate and missense single-point mutations that have been reported in these residues in human hyperphenylalaninemias Residue Mutations MutNo a Comment Reference BH 4 Tyr138 No Gly247 G247V 229 4% r.a. b 44 Leu248 L248R 230 L248P 231 Leu249 L249F 232 L249H 233 Ser251 No Phe254 F254I 237 Leu255 L255V % r.a. 44, 45 L255S 239 1% r.a. 45 His264 c H264L 250 Glu286 No Ala322 c A322T 314 A322G % r.a. 46 Tyr325 c Y325C 324 PKU d Glu330 c E330D 328 PKU e L-Phe Arg270 R270K 257 R270S 258 2% r.a. 45 Tyr277 Y277D 268 Y277C 269 Thr278 T278A 270 T278N 271 T278I 272 Pro281 P281L 275,1% r.a. 47 His285 No Trp326 No Phe331 F331L 329 F331C 330 Gly346 G346R 350 Ser349 S349P 355,1% r.a. 48, 49 S349L 356 Ser350 S350T 358 Glu353 No a Mutation number in the data base ( ca/pahdb_new/ b The abbreviation % r.a. represents the residual activity of the recombinant enzyme as a percentage of the wild-type expressed enzyme. c Binds BH 4 only in the binary complex via a water bridge. d Classic PKU, with genotype Y325C/L348V. e Classic PKU, with genotype E330D/R408W. atom 3.0 Å away from carbon atoms in both Trp326 and Glu330 (Figure 6(b)). Whereas L-Phe and, to a certain degree, L-mTyr, are well accommodated in the present active-site structure, and with specific interactions similar to that of THA, the hydroxyl group of L-Tyr is not accepted, due to steric hindrance. The resulting strained binding site for L-Tyr may well explain why this amino acid is not a substrate, but is an appropriate leaving product. Thus, the crystal structure is in complete agreement with the substrate specificity observed in steady-state kinetics. 16 Site-directed mutagenesis has been performed on TyrOH to identify residues responsible for substrate binding. 27 Arg316 (Arg270 in PheOH) was shown to be critical for substrate binding with a 400-fold higher K m value for the Arg316Lys mutation, whereas the Asp328Ser (Asp282 in PheOH) mutation showed a 26-fold higher K m value. In the present crystal structure, the Arg270 side-chain forms a salt-bridge with the carboxyl group of THA (and L-Phe), explaining the critical role of this residue in substrate binding. Asp282 does not bind to the substrate, but its carboxyl group forms a salt-bridge with Arg270 and is thus important for substrate binding by providing stability and correct positioning of Arg270. Conformational changes at the active site upon pterin cofactor and substrate binding A comparison of the non-liganded and the binary complexes of the double truncated form DN1 102/DC hPheOH with oxidized (BH 2 ) and reduced (BH 4 ) cofactor have revealed some important conformational changes of the active-site structure upon cofactor binding. Thus, in the hpheoh Fe(III) BH 2 complex, 8 the loop between residues 245 and 250 shows the largest displacement (the C a atom of Gly247 moves,1.3 Å toward the pterin ring), in the direction of the iron atom, and thus is able to form several hydrogen bonds to the pterin ring. Furthermore, the Leu248 side-chain changes its conformation as compared to the non-liganded structure, and now faces the active site. Leu255 also shifts its conformation to accommodate the dihydropropyl sidechain of the pterin molecule. In the hpheoh Fe(II) BH 4 structure, 12 the overall fold is very similar to that reported for hpheoh Fe(III) BH 2. However, superposition of the two structures has revealed that the reduced cofactor is displaced about 0.5 Å away from Ser251, and that the pterin ring is rotated about 108 (along the C4a C8a bond) with the pyrimidine ring rotated towards Phe254. The angle between the hydroxyl groups in the dihydroxypropyl group is 2658, which enables the BH 4 O2 0 to make a strong hydrogen bond (2.4 Å) with the side-chain oxygen atom of Ser251, while O1 0 forms water-mediated hydrogen bonds between residues Ala322 and Glu330. In addition, Glu330 adopts a completely different conformation in the hpheoh Fe(II) BH 4 structure. On the basis of these crystallographic data, it is evident that in both oxidation states of the enzyme and cofactor the pterin cofactor binds to the enzyme active site by an induced-fit mechanism involving a conformational change in the active-site crevice structure of the protein. In the present study, it is further demonstrated that the binding of the substrate analogue THA to the binary complex hpheoh Fe(II) BH 4 triggers large additional conformational changes at the active-site crevice structure. The largest deviation occurs in the region comprising the residues with a maximum main-chain r.m.s. displacement (9.6 Å) at Tyr138. The hydroxyl oxygen atom of this residue is indeed displaced 18.5 Å closer to the catalytic iron atom. Furthermore, Glu330 adopts a bidentate coordination and conformation not observed previously in any crystal structure of

9 Ternary Complex of Phenylalanine Hydroxylase 1103 PheOH. Thus, the binding of both the pterin cofactor and the substrate induces conformational changes at the active site that have not been detected by any alternative biophysical method. Functional implications The PheOH-catalyzed hydroxylation of L-Phe is a three-substrate reaction with specific binding sites for L-Phe, BH 4 and dioxygen, and there is general agreement that the Fe(II) centre participates directly in oxygen incorporation. 21,28,29 In the present structure of the unproductive (anaerobic) ternary complex, the binding sites for L-Phe and BH 4 and the iron coordination are clearly defined. The structure has revealed a five-coordinated iron atom with a distorted square-pyramidal coordination, as proposed previously on the basis of MCD spectral analysis of rpheoh. 25 This finding is consistent with the ordered reaction mechanism proposed for the enzyme, wherein cofactor and substrate must be present before any product is released. 28,30 The position of dioxygen binding to the iron is not established unequivocally, and attempts to bind NO and CO (by diffusion into crystals of the ternary complex) have not been successful so far. Dioxygen may bind either into the position occupied by Wat2 (as shown in Figure 7(b)) or into the open coordination position, as favoured in a recent model. 30 In the catalytic reaction, dioxygen is cleaved, incorporating one of the oxygen atoms into BH 4 to form 4a-hydroxy-tetrahydrobiopterin (4a-OH-BH 4 ), 31,32 while the other oxygen atom is incorporated into the substrate to generate L-Tyr. L-Phe/L-Tyr were modeled manually into the THA-binding site by conserving the main-chain and the x 1 and x 2 angles (Figure 7(a)) and 4a-OH-BH 4 (calculated using the perturbative Becke Perdew model (pbp/dn pp ) 33,34 of the PC SPARTAN PRO programme package) was superimposed on BH 4 (Figure 7(a)) to simulate the positions of the oxygen atoms of dioxygen after the reaction but prior to product and cofactor release. On the basis of their relative positions, dioxygen was modeled manually into the active site, positioning one of the atoms (proximal oxygen) at Wat2 and the other atom (distal oxygen) in the direction of the hydroxyl oxygen atom in 4a-OH-BH 4 (Figure 7(b)). The proximal oxygen atom is then 3.3 Å from L-Tyr O h and 3.0 Å from the closest carbon atom in THA/L-Phe, while the distal oxygen atom is 0.9 Å from the 4a-OH-BH 4 hydroxyl oxygen atom and 2.2 Å from BH 4 C4a. However, modeling dioxygen in the same manner in the open coordination position (at the position of Wat3 in the binary hpheoh Fe(II) BH 4 complex) 12 leaves the proximal oxygen atom 5.6 Å from L-Tyr O h and 5.1 Å from the closest carbon atom in THA/L-Phe, while the distal oxygen atom is 1.8 Å from the 4a-OH-BH 4 hydroxyl oxygen atom and 1.9 Å from BH 4 C4a. Binding of dioxygen at the open coordination position seems unlikely, since BH 4 O4 is close (,1.0 Å). However, the proposed dioxygen-binding site at the Wat2 position is in a hydrophobic microenvironment consisting of His285, BH 4, THA/L-Phe and Pro281 with the side-chain of Pro Å from the distal oxygen atom. It is important to note here that the P281L mutation is associated with severe PKU. 35,47 A displacement of the BH 4 molecule upon substrate binding has been proposed to accommodate a possible Fe(II)-peroxo-BH 4 intermediate, 12,36 and is indeed confirmed in the present structure (Figure 5). The crystallographic data are entirely consistent with the occurrence of a large-scale conformational transition that brings the two substrates into the close proximity required for reaction. Thus, the pterin C4a Fe(II) distance of 4.5 Å in the ternary complex is far more appropriate for the formation of a bridging dioxygen molecule between the iron atom and BH 4 (as a putative Fe(II) O O BH 4 intermediate) than the C4a Fe(II) distance of 5.9 Å in the binary BH 4 structure, implying that the bridging dioxygen intermediate can be formed only after substrate binding. On this basis, the following ordered reaction mechanism can be proposed (Figure 8), wherein cofactor and substrate are bound before any product is released. Although little is known about the mechanism of reduction, 28 prereduction of the active-site iron (step 1) is an obligate event prior to catalysis. 37 Significant changes occur in the active site upon reduction, including a reduced affinity for two of the coordinated water molecules (Wat1 and Wat2), a displacement of Glu330 and possible disorder of its side-chain. 12 The reversible binding of BH 4 (step 2) changes the overall coordination geometry and causes the Glu330 ligand to change its coordination to the iron atom. 12 The reversible binding of L-Phe (step 3) triggers a further conformational change altering the iron coordination to a highly distorted square pyramidal geometry where Glu330 adopts yet another conformation with bidentate iron coordination (Figure 2). Furthermore, the position of BH 4 in the active site is altered (Figure 5), favouring dioxygen binding at the position occupied by Wat2 (Figure 7(b)) and the formation of a putative Fe(II) O O BH 4 (4a-peroxy-BH 4 ) intermediate 25,38 (step 4). Although the molecular mechanism of dioxygen activation is still an unsolved question, a heterolytic cleavage of the oxygen oxygen bond 30,38,39 produces a molecule of 4a-OH-BH 4 and an oxidizing species, the so-called activated oxygen intermediate (most likely an oxyferryl species) (step 5), leading finally to release of the products (step 6). Experimental evidence has been presented that the formation of the hydroxylating intermediate is the rate-limiting step in the tyrosine hydroxylasecatalyzed reaction. 39,40,63 In the non-heme iron enzyme extradiol dioxygenase, a tyrosine residue at the active site has been proposed to stabilize a radical intermediate in the catalytic cycle. 41,42 In hpheoh, Tyr325 has been considered to have a similar function, 6 but site-directed mutagenesis revealed that it plays no

10 1104 Ternary Complex of Phenylalanine Hydroxylase Figure 7. (a) Stereo picture of the active site of the ternary structure. THA is replaced by the modelled L-Phe and L-Tyr (green), and BH 4 is replaced by the superpositioned hydroxylated cofactor (4a-OH-BH 4 ; blue). (b) Stereo picture of the active site of the ternary structure with a proposed binding site of dioxygen (red). One of the oxygen atoms is positioned at Wat2, while the other atom is positioned in the direction of the C4a attached hydroxyl oxygen atom in 4a-OH-BH 4 when superimposed on BH 4. The Figure was produced using MOLSCRIPT. 59 direct role in the catalytic reaction. 8 It remains to be seen if the substrate-induced reorientation of Tyr138 gives this residue a similar function. Alternatively, Tyr138 may contribute to determine the substrate specificity, since this is a phenylalanine residue (Phe214) in TyrOH. It is interesting to note that a tyrosine residue is present at the equivalent position of tryptophan hydroxylase. Another possibility is that Tyr138 plays an important role in the regulation of the enzyme. The relation between the substrate-induced conformational change in the catalytic domain and that observed in the full-length tetrameric wild-type enzyme Although there is general agreement about the importance of L-Phe (and some substrate analogues) for the activation of the tetrameric fulllength wild-type enzyme, 4,18 determination of the mechanism by which the substrate induces the related global conformational change has remained elusive. 4,18 In the present study, it has been shown that binding of the substrate at the active site triggers large-scale structural changes in the catalytic domain, including the active-site crevice structure, which are likely to represent the epicenter of the global conformational change observed in the full-- length tetrameric enzyme, and thus delineate a molecular mechanism for substrate-activation of the tetramer. The conformational changes allow the cofactor and substrate to access the active site during enzyme turnover more freely. 10,43 It is notable that, so far, hpheoh has not been crystallized in the full-length form, and that further studies on this enzyme form are required to demonstrate how the conformational change in the catalytic domain may be transmitted to other parts of the enzyme. Structural insight into the effect of single-point mutations of active-site residues Mutations in the human gene encoding the PheOH enzyme result in the autosomal recessively inherited disease PKU/HPA, and more than 400 mutations have so far been identified. The

11 Ternary Complex of Phenylalanine Hydroxylase 1105 Figure 8. Reaction pathway of PheOH. The catalytic cycle consists of a reduction of the iron centre [1] which converts the six-coordinate, high-spin Fe(III) to a four-coordinate, high-spin Fe(II); reversible binding of BH 4, which results in a six-coordinate Fe(II) and a change in coordination of Glu330 [2]; reversible binding of L-Phe to give a highly distorted square pyramidal five-coordinated Fe(II), with a bidentate coordination of Glu330 [3]; reversible binding of dioxygen to give a five-coordinated Fe(II)-O 2 intermediate followed by the formation of the putative Fe(II) O O BH 4 intermediate [4]; heterolytic cleavage of the oxygen oxygen bond and the formation of a 4a-OH-BH 4 and an oxyferryl species [5], and products are finally released [6]. resulting metabolic and clinical phenotypes range in severity from mild forms of non-pku HPA to the severe forms of PKU. The structural basis of some of the metabolic and enzymatic phenotypes has been discussed recently 11 in the light of the four crystal structures of inactive Fe(III) forms of the enzyme, but without information on the mode of substrate binding and associated conformational changes in the catalytic domain. Table 2 presents a summary of the current information on the reported missense single-point mutations of active-site residues, involved in the binding of the reduced pterin cofactor and the substrate. Of these mutations, 13 are found in nine residues involved in cofactor binding, and 14 mutations in are found in eight residues involved in substrate binding. Only a few of these mutations have been expressed as recombinant enzymes. 5,44 49 As expected, mutations of residues at the pterin cofactor and substrate-binding sites result in reduced affinities for BH 4 and L-Phe, respectively, and thus reduced catalytic efficiency in steady-state kinetic analyses. In addition, some of these mutations affect the overall structure and stability of the enzyme. 5,45 Materials and Methods Crystallization and data collection Expression and purification of the double truncated mutant (DN1 102/DC ) of hpheoh were carried out as described. 50,51 Anaerobic co-crystallization of hpheoh Fe(II) in complex with BH 4 was undertaken essentially as described 12 but with some modifications. The drops contained initial concentrations of BH 4 of 5 mm, and 15 mm sodium dithionite was used as the reducing agent. After four days of growth, solid THA was added in excess to the drop (anaerobic) and left for 24 hours to allow diffusion. A crystal of approximate size 0.6 mm 0.2 mm 0.05 mm was flash-frozen in liquid nitrogen and data were collected at 100 K on the Swiss Norwegian Beamline (BM01) at the European Synchrotron Radiation Facility (ESRF) in Grenoble (France). A wavelength of Å and a MAR345 Research imaging plate system were used. Processing of

12 1106 Ternary Complex of Phenylalanine Hydroxylase Table 3. Summary of data-collection and refinement statistics for the hpheoh Fe(II) BH 4 THA complex Resolution (Å) ( ) a Cell dimensions a, b, c (Å) 65.17, , Space group C222 1 No. observations 59,237 (8578) a No. unique reflections 14,178 (2071) a Multiplicity 4.2 (4.1) a R merge (%) 9.2 (37.7) a I/s(I) 7.1 (2.1) a Completeness (%) 94.4 (95.2) a Wilson B-factor (Å 2 ) 41.3 No. atoms in refinement 2586 No. solvent molecules 39 R work (%) 22.0 R free (%) 55 (10% of data) 26.7 Average B-factor all atoms (Å 2 ) 34.3 Average THA B-factor (Å 2 ) 18.6 Average BH 4 B-factor (Å 2 ) 31.7 r.m.s. deviations Bond lengths (Å) Bond angles (deg.) 1.12 Dihedral angles (deg.) 21.9 Improper angles (deg.) 0.87 Luzatti r.m.s. coordinate error (Å) s A r.m.s. coordinate error (Å) f and c angles in most-favoured regions (%) b 87.3 f and c angles in additional allowed regions (%) b 12.7 a Statistics for highest-resolution shell are given in parentheses. b Statistics from Ramachandran plots 61 are calculated by PROCHECK. 62 the data was done using DENZO, 52 while scaling and merging were carried out using SCALA in the CCP4 program suite. 53 Model building and refinement The refinement was initiated by a rigid-body refinement of the catalytic domain of hpheoh Fe(III) 6 (PDB entry 1PAH) using CNS version 1.0: 54 10% of the reflections were excluded in the refinement for crossvalidation. 55 The initial R-factor was 50.7%, suggesting large deviations from the starting model, but three rounds of simulated annealing (CNS) dropped the R-factor to 33.6%. At this point, 88 residues ( , and ) were outside electron density and subsequently removed from the model. The omitted residues were gradually re-added, in between rounds of energy minimization and refinement, in their correct positions at the ends of the remaining four peptide chains. The model building was done using the graphical programme O 56 and s A -weighted 20 2F o 2 F c and F o 2 F c maps calculated in CNS. The electron densities for both BH 4 and THA were good even after one round of simulated annealing, but they were added to the model after all three rounds of simulated annealing. Water oxygen atoms were added throughout the proceeding refinement as clear 2F o 2 F c and F o 2 F c densities appeared in positions compatible with hydrogen bonds to the protein or other water molecules. Luzatti 64 and s A coordinate errors and r.m.s. deviations concerning bond lengths, bond angles, dihedral angles and improper angles were calculated using CNS. Statistics of the processing and of the final model is presented in Table 3. A crystal structure to 2.1 Å of the binary hpheoh Fe(III) BH 2 complex was obtained (data not shown) by using known crystallization conditions, 8 modified with a high concentration (5 mm) of BH 2. This structure (PDB entry 1LRM) was refined by CNS to a final R work and R free of 21.1% and 23.8%, respectively. Protein Data Bank accession numbers Atomic coordinates and structure factors have been deposited at the Protein Data Bank (PDB), Research Collaboratory for Structural Bioinformatics (RCSB), with accession number 1KW0. The coordinates will remain privileged until 28th January Acknowledgments This work has been supported by grants from the Norwegian Research Council (NFR), the Norwegian Council on Cardiovascular Diseases, Rebergs legat, L. Meltzer Høyskolefond, the Novo Nordisk Foundation and the European Commission. We thank Ali Sepulveda Muñoz for expert technical assistance in preparing the bacterial extracts and fusion protein, and the staff of the Swiss Norwegian Beamlines in Grenoble (France). References 1. Kaufman, S. (1993). The phenylalanine hydroxylating system. Advan. Enzymol. Relat. Areas. Mol. Biol. 67, Bickel, H., Bachmann, C., Beckers, R., Brandt, N. J., Clayton, B. E., Corrado, G. et al. (1981). Neonatal mass screening for metabolic disorders. Eur. J. Pediatr. 137, Scriver, C. R., Waters, P. J., Sarkissian, C., Ryan, S., Prevost, L., Côté, D. et al. (2000). PAHdb: A locusspecific knowledgebase. Hum. Mutat. 15, Flatmark, T. & Stevens, R. C. (1999). Structural insight into the aromatic amino acid hydroxylases and their disease-related mutant forms. Chem. Rev. 99,

13 Ternary Complex of Phenylalanine Hydroxylase Waters, P. J., Parniak, M. A., Nowacki, P. & Scriver, C. R. (1998). In vitro expression analysis of mutations in phenylalanine hydroxylase: linking genotype to phenotype and structure to function. Hum. Mutat. 11, Erlandsen, H., Fusetti, F., Martínez, A., Hough, E., Flatmark, T. & Stevens, R. C. (1997). Crystal structure of the catalytic domain of human phenylalanine hydroxylase reveals the structural basis for phenylketonuria. Nature Struct. Biol. 4, Fusetti, F., Erlandsen, H., Flatmark, T. & Stevens, R. C. (1998). Structure of tetrameric human phenylalanine hydroxylase and its implications for phenylketonuria. J. Biol. Chem. 273, Erlandsen, H., Bjørgo, E., Flatmark, T. & Stevens, R. C. (2000). Crystal structure and site-specific mutagenesis of pterin-bound human phenylalanine hydroxylase. Biochemisty, 39, Erlandsen, H., Flatmark, T., Stevens, R. C. & Hough, E. (1998). Crystallographic analysis of the human phenylalanine hydroxylase catalytic domain with bound catechol inhibitors at 2.0 Å resolution. Biochemistry, 37, Kobe, B., Jennings, I. G., House, C. M., Michell, B. J., Goodwill, K. E., Santarsiero, B. D., Stevens, R. C., Cotton, R. G. H. & Kemp, B. E. (1999). Structural basis of autoregulation of phenylalanine hydroxylase. Nature Struct. Biol. 6, Erlandsen, H. & Stevens, R. C. (1999). The structural basis of phenylketonuria. Mol. Genet. Metab. 68, Andersen, O. A., Flatmark, T. & Hough, E. (2001). High resolution crystal structures of the catalytic domain of human phenylalanine hydroxylase in its calalytically active Fe(II) form and binary complex with tetrahydrobiopterin. J. Mol. Biol. 314, Kaufman, S. & Mason, K. (1982). Specificity of amino acids as activators and substrates for phenylalanine hydroxylase. J. Biol. Chem. 257, Stokka, A. J. & Flatmark, T. (2002). 3-(2-Thienyl)-Lalanine as a competitive substrate analogue and activator of human phenylalanine hydroxylase. In Chemistry and Biology of Pteridines and Folates (Milstien, S., Kapatos, G., Levine, R. A. & Shane, B., eds.), pp , Kluwer Academic Publishers. 15. Teigen, K., Frøystein, N. A. & Martínez, A. (1999). The structural basis of the recognition of phenylalanine and pterin cofactors by phenylalanine hydroxylase: implications for the catalytic mechanism. J. Mol. Biol. 294, Kaufman, S. (1987). Phenylalanine 4-monooxygenase from rat liver. Methods Enzymol. 142, Døskeland, A. P., Døskeland, S. O., Øgreid, D. & Flatmark, T. (1984). The effect of ligands of phenylalanine 4-monooxygenase on the camp-dependent phosphorylation of the enzyme. J. Biol. Chem. 259, Flatmark, T., Stokka, A. J. & Berge, S. V. (2001). Use of surface plasmon resonance for real-time measurements of the global conformational transition in human phenylalanine hydroxylase in response to substrate binding and catalytic activation. Anal. Biochem. 294, Erlandsen, H., Martínez, A., Knappskog, P. M., Haavik, J., Hough, E. & Flatmark, T. (1997). Crystallization and preliminary diffraction analysis of a truncated homodimer of human phenylalanine hydroxylase. FEBS Letters, 406, Read, R. J. (1986). Improved Fourier coefficients for maps using phases from partial structures with errors. Acta Crystallog. sect. A, 42, Pace, C. N., Horn, G., Hebert, E. J., Bechert, J., Shaw, K., Urbanikova, L. et al. (2001). Tyrosine hydrogen bonds make a large contribution to protein stability. J. Mol. Biol. 312, Goodwill, K. E., Sabatier, C. & Stevens, R. C. (1998). Crystal structure of tyrosine hydroxylase with bound cofactor analogue and iron at 2.3 Å resolution: self-hydroxylation of Phe300 and the pterin-binding site. Biochemistry, 37, Goodwill, K. E., Sabatier, C., Marks, C., Raag, R., Fitzpatrick, P. F. & Stevens, R. C. (1997). Crystal structure of tyrosine hydroxylase at 2.3 Å resolution and its implications for inherited neurodegenerative diseases. Nature Struct. Biol. 4, Olafsdottir, S. & Martínez, A. (1999). The accessibility of iron at the active site of recombinant human phenylalanine hydroxylase to water as studied by 1 H NMR paramagnetic relaxation. Effect of L-Phe and comparison with the rat enzyme. J. Biol. Chem. 274, Kemsley, J. N., Mitic, N., Zaleski, K. L., Caradonna, J. P. & Solomon, E. I. (1999). Circular dichroism and magnetic circular dichroism spectroscopy of the catalytically competent ferrous active site of phenylalanine hydroxylase and its interaction with pterin cofactor. J. Am. Chem. Soc. 121, Daubner, S. C., Hillas, P. J. & Fitzpatrick, P. F. (1997). Characterization of chimeric pterin-dependent hydroxylases: contributions of the regulatory domains of tyrosine and phenylalanine hydroxylase to substrate specificity. Biochemistry, 36, Daubner, S. C. & Fitzpatrick, P. F. (1999). Site-directed mutants of charged residues in the active site of tyrosine hydroxylase. Biochemistry, 38, Kappock, T. J. & Caradonna, J. P. (1996). Pterindependent amino acid hydroxylases. Chem. Rev. 96, Chen, D. & Frey, P. A. (1998). Phenylalanine hydroxylase from Chromobacterium violaceum. Uncoupled oxidation of tetrahydropterin and the role of iron in hydroxylation. J. Biol. Chem. 273, Solomon, E. I., Brunold, T. C., Davis, M. I., Kemsley, J. N., Lee, S.-K., Lehnert, N. et al. (2000). Geometric and electronic structure/function correlations in non-heme iron enzymes. Chem. Rev. 100, Dix, T. A. & Benkovic, S. J. (1988). Mechanism of oxygen activation by pteridine-dependent monooxygenases. Accts Chem. Res. 21, Haavik, J. & Flatmark, T. (1987). Isolation and characterization of tetrahydropterin oxidation products generated in the tyrosine 3-monooxygenase (tyrosine hydroxylase) reaction. Eur. J. Biochem. 168, Becke, A. D. (1988). Density-functional exchangeenergy approximation with correct asymptotic behaviour. Phys. Rev. A, 38, Perdew, J. P. (1986). Density-functional approximation for the correlation energy of the inhomogeneous electron gas. Phys. Rev. ser. B, 33, Dworniczak, B., Grudda, K., Stumper, J., Bartholome, K., Aulehla-Scholz, C. & Horst, J. (1991). Phenylalanine hydroxylase gene: Novel missense mutation in exon 7 causing severe phenylketonuria. Genomics, 9, Loeb, K. E., Westre, T. E., Kappock, T. J., Mitic, N., Glasfeld, E., Caradonna, J. P. et al. (1997). Spectroscopic characterization of the catalytically competent