The Accessibility of Iron at the Active Site of Recombinant Human Phenylalanine Hydroxylase to Water As Studied by 1H NMR Paramagnetic Relaxation

The high-spin (S = 5/2) Fe(III) ion at the active site of recombinant human phenylalanine hydroxylase (PAH) has a paramagnetic effect on the longitudinal relaxation rate of water protons. This effect is proportional to the concentration of enzyme, with a paramagnetic molar-relaxivity value at 400 MHz and 25 °C of 1.3 (± 0.03) × 103 s−1 m −1. The value of the Arrhenius activation energy (E a ) for the relaxation rate was −14.4 ± 1.1 kJ/mol for the resting enzyme, indicating a fast exchange of water protons in the paramagnetic environment. The frequency dependence of the relaxation rate also supported this hypothesis. Thus, the recombinant human PAH appears to have a more solvent-accessible catalytic iron than the rat enzyme, in which the water coordinated to the metal is slowly exchanging with the solvent. These findings may be related to the level of basal activity before activation for these enzymes, which is higher for human than for rat PAH. In the presence of saturating (5 mm) concentrations of the substratel-Phe, the paramagnetic molar relaxivity for human PAH decreased to 0.72 (± 0.05) × 103 s−1 m −1 with no significant change in theE a . Effective correlation times (τC) of 1.8 (± 0.3) × 10−10 and 1.25 (± 0.2) × 10−10 s−1 were calculated for the enzyme and the enzyme-substrate complex, respectively, and most likely represent the electron spin relaxation rate (τS) for Fe(III) in each case. Together with the paramagnetic molar-relaxivity values, the τC values were used to estimate Fe(III)-water distances. It seems that at least one of the three water molecules coordinated to the iron in the resting rat and human enzymes is displaced from coordination on the binding of l-Phe at the active site.

catalyzes the hydroxylation of L-Phe to L-Tyr, which is the rate-limiting step in the catabolic pathway for L-Phe, taking place mainly in the liver (1,2). Like the other mammalian aromatic amino acid hydroxylases, tyrosine hydroxylase, and tryptophan hydroxylase, PAH is a mononuclear nonheme iron containing enzyme which requires (6R)-L-erythro-tetrahydrobiopterin (BH 4 ) and dioxygen as additional substrates (3). The iron is ferric in the enzyme as isolated and in the catalytic cycle it is reduced to Fe(II) by BH 4 (2,4,5). Deficiency of human PAH activity causes phenylketonuria, which represents the most prevalent inborn error of amino acid metabolism (6).
Our group has recently described the expression and the molecular and kinetic characterization of recombinant human PAH (7)(8)(9)(10). The catalytic and physicochemical properties of this enzyme are essentially the same as those reported for the human liver enzyme (7,11). The catalytic domain of the human PAH has been crystallized and the three-dimensional structure has recently been solved (12,13). This crystal structure provides a frame in the understanding of the effect of mutations in PAH causing phenylketonuria. Moreover, large efforts are being made by several groups in this field to solve the mechanism for catalysis and regulation in light of the three-dimensional structure. The largest body of information about PAH has been previously obtained on the enzyme isolated from rat liver (2,3). The rat and human PAH share a 96% sequence homology (14) and they have many similar molecular and kinetic properties (2,3,7,8,10,15): (a) they have comparable K m -values for L-Phe and tetrahydropterin cofactors; (b) they consist of dimers and tetramers in equilibrium; (c) they are activated by a number of processes, e.g. incubation with lysolecithin, high pH, deletion of the N-terminal domain, phosphorylation at Ser-16 and by the substrate L-Phe, the two latter mechanisms being of physiological importance; (d) the activation by L-Phe, which binds with positive cooperativity, results in conformational changes involving the tertiary and the quaternary structure, shifting the dimer 7 tetramer equilibrium toward the tetrameric form. However, there is a clear difference between the human and the rat PAH with respect to the extent of the response to preincubation with L-Phe and to other modes of activation. Thus, the native and recombinant rat PAH are activated 8-to 30-fold by L-Phe (1,16), whereas the recombinant human PAH is only activated about 2-3-fold (8,10). Even lower values of activation of the human enzyme have been reported by Kowlessur et al. (17). The degree of activation by phosphorylation is also lower for human than for rat PAH (10). Thus, although the activity of the maximally activated enzyme is about 2-fold lower for the human than for the rat PAH (1,8,16), the level of basal activity in the absence of activating treatments is higher for the recombinant human than for the rat enzyme, which * This work was supported by grants from the Research Council of Norway. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Present address: Science Institute, University of Iceland, Dunhaga 3, 107 Reykjavik, Iceland.

Ϫ1
, outer-sphere contribution to the relaxation rate; C , dipolar correlation time; M , exchange lifetime; S , electron spin relaxation rate; I and S , proton and electron Larmor frequencies, respectively. seems to be in agreement with the properties of PAH from human liver (16,17).
By using 1 H NMR paramagnetic relaxation of the water protons, we have shown earlier that the Fe(III) ion at the active site of resting rat PAH contains coordinated water and that following the binding of both L-Phe and L-noradrenaline at least one water molecule is displaced from coordination (18). In this study we have measured the paramagnetic effect of the ferric site in recombinant human PAH on the water protons in the absence and the presence of the substrate L-Phe. The differences encountered between the accessibility of the solvent to the iron in the rat and the human enzymes are interpreted based on the different degree of activity in the resting enzymes.

EXPERIMENTAL PROCEDURES
Materials-L-Phenylalanine and bathophenanthroline disulfonic acid were from Sigma and 2 H 2 O (99.8%) from Aldrich. Expression of recombinant human PAH in Escherichia coli (TB1) as fusion protein with maltose-binding protein, purification of the fusion proteins by affinity chromatography on amylose resin followed by high-performance size exclusion chromatography, cleavage by factor Xa or enterokinase, and further purification of the hydroxylase was performed as described (7). Rat PAH was isolated from rat liver by the method (procedure II D) of Shiman et al. (19). The enzyme activity was measured by determination of L-Tyr formed by high pressure liquid chromatography and fluorimetric detection as described (7). The concentration of purified PAH (both rat and human) was estimated using an absorption coefficient at 280 nm of 1.0 cm Ϫ1 for 1 mg/ml (7,19).
Metal Analyses-The metal content of the enzyme preparations was determined by a Perkin-Elmer model 402 atomic absorption spectrophotometer equipped with a graphite furnace (type HGA-76B from Perkin-Elmer).
Preparation of Iron-free Apoenzymes-The iron was extracted from the isolated rat and human holoenzymes by a modification of the method of Gottschall et al. (20) using 1 mM bathophenanthroline, 1 mM L-Phe, 0.5 mM BH 4 , and 1 mM dithiothreitol. The formation of the Fe(II)-bathophenanthroline complex was followed by measuring the increase in absorbance at 535 nm. The concentration of the released iron was calculated from the molar extinction coefficient for the Fe(II)bathophenanthroline complex (⑀ 535 ϭ 22,000 M Ϫ1 s Ϫ1 ) (21). The enzyme was separated from the chelating agent and other low molecular weight compounds by gel filtration on a G-25 Sephadex column (1.5 ϫ 20-cm) equilibrated in 20 mM Na-Hepes, 0.2 M NaCl, pH 7.0, and 1 mM L-Phe, included to stabilize the apoenzyme. The catalytically inactive apoenzyme was concentrated in Centriplus 30 concentrators (Amicon, MA).
NMR Measurements-Longitudinal relaxation rates of the residual water signal (HDO) were measured on enzyme samples prepared by 3-4 cycles of 20-fold concentration and dilution in 2 H 2 O, containing 20 mM potassium phosphate of pH 7.2 (pH value determined with an Ingold electrode and representing the uncorrected value in 2 H 2 O) and 0.2 M KCl, using Centricon 30 microconcentrators (Amicon). The NMR spectra were recorded on a Bruker DMX-400 and in experiments with variable field, measurements at 250 and 600 MHz were made on Bruker AM-250 and Bruker DRX-600 spectrometers, respectively. The longitudinal relaxation times (T 1OBS ) were measured at the indicated probe temperatures by using a standard inversion-recovery sequence, with acquisition parameters including 16K data points, four transients per time increment and recycle delay (Ͼ5 ϫ T 1 ). In the titration experiments, the enzyme solution was allowed to equilibrate at the indicated concentrations of L-Phe in the NMR tubes for 5 min at room temperature, before measurements. Titrations were performed by adding microliter amounts of concentrated solutions of L-Phe. Final dilution of the samples was Յ2.5%.
Theoretical Details-The longitudinal paramagnetic relaxation time of the bound water proton (T 1M ) is described by the dipolar term of the Solomon-Bloembergen equation (22,23), i.e.
Ϫ1 ϭ 2͑␥ I g␤͒ 2 S͑S ϩ 1͒ where (␥ I is the nuclear gyromagnetic ratio, g is the electronic g factor (isotropic splitting factor), ␤ is the Bohr magneton, S is the electronic spin at the ground state of the paramagnetic ion (24), r is the metalproton internuclear distance, I and S are the nuclear and electron Larmor precession frequencies, respectively, and C the effective dipolar correlation time, which describes the molecular events which modulate the electron-nuclear dipolar coupling and can be calculated from the frequency dependence of the longitudinal paramagnetic relaxation (25).
The observed longitudinal relaxation rate of water protons in a protein solution (T 1OBS Ϫ1 is the longitudinal relaxation rate of water protons due to the paramagnetic ion and T 1D Ϫ1 is the diamagnetic contribution of the protein due to the effects of protein residues-solvent interactions. T 1D Ϫ1 was estimated using sodium dithionite-treated enzyme, in which the iron is fully reduced (26), or using apoenzyme forms without iron. T 1P Ϫ1 values for water protons, normalized to a subunit concentration (C s ) of 1 M are expressed as paramagnetic molar relaxivity (T 1P ), which is related to T 1M by the following expression (27): where q is the number of water ligands that are coordinated to the paramagnetic ion (i.e. in the first coordination sphere) and T O.S. Ϫ1 is the outer sphere contribution to the relaxation rate.
Ϫ1 is usually small in paramagnetic systems with coordinated water. The effect of varying temperature and frequency on T 1P Ϫ1 was used to determine the predominant contributions (T 1M , M , or T O.S. ) to the observed relaxation rate (25). Further theoretical considerations relevant for this paper have been described (18).

RESULTS AND DISCUSSION
The Enzyme Preparations-The specific activities of the isolated recombinant human and rat liver PAH preparations with 1 mM L-Phe and 500 M BH 4 in the presence and the absence of prior incubation (5 min, 25°C) with 1 mM L-Phe are shown in Table I. Although the activity of the rat liver holoenzyme was not stimulated by addition of ferrous ammonium sulfate (100 M) in the assay mixture, the activity of the recombinant human PAH increased by about 10% in the presence of ferrous ions, indicating the presence of some iron-free apoenzyme in the preparations. Thus, as determined by atomic absorption spectroscopy the preparations of rat and human PAH used in this study contained 0.95 Ϯ 0.03 and 0.48 Ϯ 0.05 atom of iron/mol subunit, respectively. By measuring the formation of the Fe(II)-bathophenanthroline complex at 535 nm, it was determined that a maximal amount of about 0.5 atom of iron/mol subunit was extracted from both the rat and human enzymes after a 30 -45-min incubation period with the chelator at reducing conditions (see "Experimental Procedures"). After iron extraction, the resulting proteins were devoid of catalytic activity and were referred to as apoenzymes (Table I). However, as shown by atomic absorption spectroscopy measurements, although no remaining iron was present in the apoenzyme forms of the recombinant human PAH, the apoenzyme of the rat liver enzyme contained 0.4 -0.45 atom/subunit. This population of inactive iron in rat PAH has previously been found not to be reduced by the tetrahydropterin cofactor and not to participate in catalysis (4,5,26). After reconstitution of the holoenzyme from the apoenzyme by incubation with 0.1 mM ferrous ammonium sulfate, full activity was recovered.

Effect of Recombinant Human Phenylalanine Hydroxylase on the Water Proton Relaxation Rates-T 1OBS
Ϫ1 values were measured at 400 MHz on the bulk residual water signal (HDO)   (Fig. 1). In this concentration range we found no significant diamagnetic contribution of the protein to the relaxation rate, measured either with the iron-free human apoenzyme or with the dithionite reduced enzyme (Fig. 1). The large effect of the enzyme as isolated on T 1OBS Ϫ1 indicates that the high-spin Fe(III) (S ϭ 5/2) site in the recombinant human PAH is accessible to exchangeable water molecules, as was previously found for the rat liver enzyme (18). Accordingly, the x-ray structure of the catalytic domain of human PAH has shown that the ferric iron is six-coordinated to His-285, His-290, Glu-330 and to three water molecules, referred to as Wat (1), Wat (2), and Wat (3) (12). As shown by magnetic circular dichroism and x-ray absorption spectroscopy, the iron sites for both the resting ferric (inactive) and ferrous (active) forms in the rat enzyme also seem to be six-coordinate distorted octahedral and substrate binding results in geometric and electronic structural changes at the iron center (28).
Although the paramagnetic molar relaxivity (T 1P Ϫ1 ⅐C s Ϫ1 ) at 298 K is higher for the rat than for the human enzymes, the values become similar when they are normalized to molar concentration of iron (T 1P

⅐C Fe
Ϫ1 ) ( Table II). The paramagnetic molar relaxivity of human PAH was found to decrease about 2-fold when 5 mM L-Phe was added (Table II), indicating either occlusion or displacement of the coordinated water molecules. The effect of L-Phe binding to the human enzyme on 1/T 1P was studied in more detail as a function of ligand con-centration (Fig. 2). The titration curve was found to be hyperbolic, unlike the curves obtained on titration of rat liver PAH either with L-Phe or L-noradrenaline, which are three-phasic and nonhyperbolic (18). The three-phasic curves have been interpreted as being the result of the change from a system in which water is coordinated to Fe(III) at the active site and slowly exchanging with the bulk water, to a system in which water is fast exchanging at a site close to the iron, but not coordinated (18). In the case of the recombinant human PAH the hyperbolic titration curve (Fig. 2) seems to indicate the absence of exchange limitations in the enzyme as isolated, as well as in the L-Phe-enzyme complex.
Temperature and Field Dependence of Water Proton Relaxation Rates-The effect of temperature (16 -40°C) on the paramagnetic molar relaxivity of water protons is shown in Fig. 3. The Arrhenius activation energy (E a ) for the paramagnetic contribution to the relaxation (T 1P Ϫ1 ) was only slightly affected by the binding of L-Phe and showed just a small decrease from Ϫ14.4 Ϯ 1.1 kJ/mol in the resting enzyme to Ϫ17.9 Ϯ 0.4 kJ/mol in the presence of 5 mM L-Phe (enzyme-substrate complex) (Table II). Typical E a for T 1P Ϫ1 in slow exchange processes are Ն9 kJ/mol because, in general, the exchange-lifetime, M , has positive temperature coefficients (25,29). Our results thus indicate that both in the resting human enzyme and in its complex with L-Phe, the relaxation of water protons is not exchange-limited. This is different from the situation in the resting nonactivated rat liver PAH, in which E a was found to be 11.3 Ϯ 0.8 kJ/mol (18) consistent with exchange limitations for the water molecules coordinated to the iron, indicating that the

), Arrhenius activation energy (E a ) for the paramagnetic contribution to the relaxation, effective correlation times ( C ), and estimated water-iron distances (r) for the rat and the recombinant human PAH in the absence and the presence of L-Phe
Human PAH Rat PAH a protein may impose hindrances to the free exchange of the coordinated water. This limitation was abolished in the L-Pheactivated enzyme (E a ϭ Ϫ1.5 Ϯ 0.2 kJ/mol) ( Table II).
The paramagnetic contribution to the relaxation time (T 1P ) for solutions of human PAH was measured at 295 K and three different Larmor frequencies, i.e. 250, 400, and 600 MHz, and found to be frequency-dependent both in the presence and the absence of 5 mM L-Phe (data not shown), also consistent with a fast exchange condition of water protons in the paramagnetic environment (25). A linear fit of the T 1P ⅐C s values at the three frequencies was used to calculate the effective dipolar correlation time ( C , Eq. 1) under the assumption that it is constant in this frequency range (25,30) (Table II). The C values thus obtained for both the human enzyme (1.8 (Ϯ0.3) ϫ 10 Ϫ10 s) and its enzyme-substrate complex (1.25 (Ϯ0.2) ϫ 10 Ϫ10 s) seem to be dominated by S , the electron-spin relaxation time for the high-spin Fe(III) center (18,25,27). For the rat PAH a fre-quency dependence of the relaxation rates was only observed for the enzyme-substrate complex, yielding a C value of 4.2 (Ϯ0.5)⅐10 Ϫ10 s (18). These C values should be considered as an approximation, because they were calculated from measurements at high field (Ն100 MHz), where dispersion of magnetization may occur (24).

Estimation of the Fe(III)-Water Proton Internuclear Distances in Recombinant Human Phenylalanine Hydroxylase and the Enzyme-Substrate
Complex-For nonexchange-limited processes, the paramagnetic relaxation rate of the water proton resonance is dependent on the distance between the exchangeable water molecules and the ferric iron. Distances can be estimated from the paramagnetic molar relaxivity values (Table II) using the value of C (see Refs. 24 and 25, and Equation 1). Assuming that the paramagnetic contribution to the relaxation rate of the bulk water is mainly because of the exchange of one of the water molecules coordinated to the paramagnetic ion (q ϭ 1, Eq. 2), T 1M is calculated to be 6.6 ϫ 10 Ϫ6 s and the estimated distance (r) between the iron and the water protons (averaged distance for the two protons) is 3.0 Ϯ 0.3 Å (Eq. 1). Assuming that two (q ϭ 2) or three (q ϭ 3) of the water molecules coordinated to the Fe(III) (12) contribute equally to the relaxation of the bulk water r increases to 3.4 Ϯ 0.4 Å and 3.7 Ϯ 0.4 Å, respectively. Thus, our data fit best with the enhancement of solvent bulk water proton relaxation rates being due to one or two of the coordinated water molecules transferring the paramagnetic effects to the bulk water through exchange. These water molecules are most likely Wat (1) and Wat (2), the most mobile of the three coordinated water molecules to the Fe(III) in the crystal structure of the catalytic domain (12). Although Wat (1) has a temperature factor that is slightly higher than that of Wat (2) (48.7 Å 2 versus 35.8 Å 2 ), it is hydrogen bonded (2.8 Å) to the hydroxyl group in the phenolic ring of Tyr-325, whereas Wat (2) is not stabilized by any additional interaction with the protein. Moreover, Wat (3) is hydrogen bonded (2.7 Å) to Glu-286 and has a low temperature factor (18.7 A 2 ) and probably contributes little to the transfer of paramagnetic effect to the bulk water.
The estimated water protons-Fe(III) distances increased by 13% on incubation of the enzyme with L-Phe, regardless of the value of q, indicating a displacement of, at least, one of the coordinated water molecules, as earlier found for rat PAH in the presence of L-Phe (18) and for other systems in which bound water is known to be displaced upon binding of ligands (27,31,32). However, to date there is no crystal structure of complexes of the enzyme with L-Phe, and a possible displacement of water on binding of the substrate has not been proved. The binding of L-Phe to either the rat or the human PAH does not affect the spin state of the Fe(III) (26) 2 and the decrease in paramagnetic molar relaxivity upon binding cannot be because of high-spin to low-spin transition of the iron.
Most groups working with PAH interpret the activation of PAH by L-Phe as the result of its cooperative binding at an allosteric site, which is physically different to the binding site of L-Phe at the active site (1, 2). We have, however, interpreted the activation as the result of the cooperative binding of L-Phe at the active site (18,33) because: L-noradrenaline, a ligand binding at the active site of the enzyme by coordination to the iron, binds with positive cooperativity to rat PAH and induces conformational changes similar to L-Phe (18,26,33). Thus, displacement of water from the active site Fe could be because of the rearrangement of the coordination geometry of the metal upon substrate binding and enzyme activation. Although the local effects related to water displacement seem to be similar for rat and human PAH, larger conformational effects seem to be induced on the rat enzyme by activation with L-Phe increasing the limited accessibility of the active site, a limitation that is not observed in the human enzyme.
In conclusion, the results presented here support that the observed limitation to the exchange of the coordinated water in PAH isolated from the rat liver is not detected for the human enzyme. This may be related to the state of activation, which is higher in human than in rat PAH (1). These results have important implications for understanding the structural and regulatory differences between the hydroxylases from both sources, as well as the phenylalanine homeostasis in man (1). Moreover, we have further shown that the binding of L-Phe and activation of both human and rat enzyme is accompanied by the displacement of at least one water molecule from coordination to the iron.