Ferryl Derivatives of Human Indoleamine 2,3-Dioxygenase*

The critical role of the ferryl intermediate in catalyzing the oxygen chemistry of monooxygenases, oxidases, or peroxidases has been known for decades. In contrast, its involvement in heme-based dioxygenases, such as human indoleamine 2,3-dioxygenase (hIDO), was not recognized until recently. In this study, H2O2 was used as a surrogate to generate the ferryl intermediate of hIDO. Spectroscopic data demonstrate that the ferryl species is capable of oxidizing azinobis(3-ethylbenzothiazoline-6-sulfonic acid) but not l-Trp. Kinetic studies reveal that the conversion of the ferric enzyme to the ferryl intermediate facilitates the l-Trp binding rate by >400-fold; conversely, l-Trp binding to the enzyme retards the peroxide reaction rate by ∼9-fold, because of the significant elevation of the entropic barrier. The unfavorable entropic factor for the peroxide reaction highlights the scenario that the structure of hIDO is not optimized for utilizing H2O2 as a co-substrate for oxidizing l-Trp. Titration studies show that the ferryl intermediate possesses two substrate-binding sites with a Kd of 0.3 and 440 μm and that the electronic properties of the ferryl moiety are sensitive to the occupancy of the two substrate-binding sites. The implications of the data are discussed in the context of the structural and functional relationships of the enzyme.

Indoleamine 2,3-dioxygenase (IDO) 2 and tryptophan dioxygenase (TDO) are the only two heme-based dioxygenases in humans that are responsible for the conversion of L-tryptophan (L-Trp) to N-formylkynurenine (NFK) via the kynurenine pathway (1). The majority of our dietary L-Trp is metabolized by TDO in the liver (2,3). TDO is hence critical for the regulation of our systemic L-Trp levels. In contrast to the hepatic TDO, IDO is inducible by IFN-␥ and is ubiquitously distributed in all tissues (1,4). It plays important immunosuppressive roles under various physiological and pathophysiological conditions (5). Recently, IDO has attracted a great deal of attention because of its recognition as a potential target for cancer therapy (6,7).
It is generally accepted that for heme proteins to carry out oxygen chemistry, the relatively inert dioxygen has to be activated first. In monooxygenases, such as P450s, dioxygen is acti-vated by two-electron reduction to peroxide. The subsequent heterolytic O-O bond cleavage of the heme-bound peroxide leads to a ferryl species (Fe 4ϩ ϭO 2Ϫ ), with a -cation radical residing on the porphyrin ring (i.e. the so-called compound I species), and a water molecule (1,9,10). The compound I (Cmpd-I) species thus produced is a strong oxidant that is capable of inserting a single oxygen to organic substrates (RH) to generate oxygenated products (ROH) (11) as shown in Reaction 1.
The overall reaction consumes two electrons and two protons. The requirement of the reductive activation of dioxygen in P450s is confirmed by the observation that the reactions can be bypassed by reacting the ferric enzymes with H 2 O 2 (the so-called "peroxide shunt"). A similar mechanism is believed to be operative in oxidases (12)(13)(14)(15), peroxidases, and catalases (16,17).
In contrast to the P450 reactions, the catalytic cycle of IDO and TDO is initiated by the binding of O 2 and L-Trp to the ferrous enzyme to generate the ternary complex, which turns over to produce the product, NFK, leaving the heme iron in the active ferrous state that is ready for a new turnover (see Reaction 2).

and references therein).
Crystallographic data of human isoforms of IDO, hIDO, show that the proximal heme ligand is a histidine (His-346) (18). His-346 is linked to Arg-343, Asp-274, Thr-395, and a propionate group of the heme via an extended H-bonding network involving two intervening water molecules (18,19), giving the histidine imidazolate character, as indicated by the relatively high frequency of the proximal iron-histidine stretching mode ( Fe-His ), at 236 cm Ϫ1 (20). The imidazolate ligand leads to a relatively low redox potential of the heme iron (Ϫ30 and ϩ16 mV for the substrate-free and L-Trp-bound enzyme, respectively) (21), with respect to that of globins (ϩ50/ϩ160 mV) (22,23). Intriguingly, the redox potentials of hIDO are more comparable with those of peroxidases (Ϫ300 to ϩ100 mV) (24,25), which also possess a proximal imidazolate ligand (20), suggesting that the electron-donating properties of the proximal histidine ligand of hIDO may be critical for modulating the reactivity of the ferric superoxide and stabilizing the high valent iron oxo intermediates during the dioxygenase chemistry (see below).
Structure-based sequence homology data (18,28) suggest that the substrate, L-Trp, binds to the distal heme pocket and that substrate binding introduces the ordering of the (360 -380) loop, which anchors the substrate in a unique regioorientation with respect to the heme-bound O 2 via H-binding interactions, thereby facilitating the dioxygenase chemistry (29,30). In the past few decades, it was widely believed that the two atoms of heme-bound O 2 are simultaneously incorporated into the substrate by IDO, setting it apart from monooxygenases (1,9,10). However, recent data demonstrated that the dioxygenase reaction follows a stepwise oxygen insertion mechanism (29). As illustrated in Scheme 1, the reaction is initiated by radical addition of the heme-bound O 2 , with ferric superoxide character, to the C 2 of the indole ring of L-Trp to generate a compound II (Cmpd-II) type of ferryl species (Fe 4ϩ ϭO 2Ϫ without a -cation radical on the porphyrin) and an indole 2,3-epoxide intermediate via a 2-alkylperoxo transition state. It is followed by the subsequent insertion of the ferryl oxygen to the epoxide to generate the product NFK. The two-step ferryl-based mechanism introduces a paradigm shift in our understanding of the heme-based dioxygenase chemistry. It also reveals the unusual reactivities of ferric superoxide and Cmpd-II type of ferryl species. Computational studies show that the ammonium group of the L-Trp plays a critical role in activating the ferric superoxide for its insertion to the substrate, as well as in protonating the epoxide moiety of the intermediate for its ring opening reaction and subsequent electrophilic addition to the ferryl moiety (29).
Although the dioxygenase reaction of IDO does not consume any electrons, a small portion of the ternary complex can leak out the active cycle by releasing O 2 as superoxide (1, 31). Consequently, a significant amount of ferric enzyme accumulates during multiple turnovers, in particular under electron-deficient conditions. The inactive ferric species may reenter the active cycle by binding superoxide. Accordingly, although controversial, superoxide has been suggested as a physiologically important co-substrate for IDO (32,33). Along these lines, the potential role of H 2 O 2 as a co-substrate for IDO has also been considered (34), despite the fact that H 2 O 2 possesses an extra redox equivalent. The in vivo H 2 O 2 production rate has been estimated to be ϳ0.1 mM min Ϫ1 under physiological conditions and at a much higher rate in adverse environments (35). It has been demonstrated that, with H 2 O 2 as a co-substrate, the human isoforms (hIDO) can oxidize melatonin, serotonin, and tryptamine, but not L-Trp (34), whereas the rabbit isoform can catalyze the demethylation reaction of benzphetamine (36). In this study, we aimed to systematically examine the reactivity of hIDO toward H 2 O 2 , as well as the chemical and structural properties of the ferryl intermediate derived from the H 2 O 2 reaction by using optical absorption and resonance Raman (RR) spectroscopies.

EXPERIMENTAL PROCEDURES
Materials-Horse heart myoglobin, L-tryptophan (L-Trp), 2,2Ј-azinobis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS) in the crystallized ammonium salt form, and hydrogen peroxide (30%) were obtained from Sigma. H 2 18 O 2 (0.2%) was purchased from Icon Isotopes (Summit, NJ). All other chemicals used throughout this study were of analytical reagent grade and were used without further purification. All solutions were prepared with deionized water treated with a Millipore water purification system (Millipore Corp., Bedford, MA). The hIDO protein samples were buffered with 100 mM phosphate at pH 7.4, unless otherwise indicated. The concentration of hIDO was estimated based on the Soret absorption coefficient ⑀ ϭ 172 mM Ϫ1 cm Ϫ1 for the ferric form (21). The concentration of H 2 O 2 was determined by using ⑀ 203 nm ϭ 72.4 M Ϫ1 cm Ϫ1 or ⑀ 254 nm ϭ 19.3 m Ϫ1 cm Ϫ1 (37).
Protein Expression and Purification-The recombinant hIDO protein was prepared based on previously described procedures (20). The catalytic activity of hIDO was confirmed by monitoring the product (NFK) formation rate at 321 nm, with an extinction coefficient of 3.75 mM Ϫ1 cm Ϫ1 (38,39). The protein was flash-frozen in liquid nitrogen and stored at Ϫ80°C prior to usage.
Stopped-flow Measurements-The transient absorption spectra and kinetics associated with the reaction of hIDO with H 2 O 2 in the presence or absence of L-Trp were measured with a *180 stopped-flow instrument from Applied Photophysics Inc. (Leatherhead, UK) equipped with a photodiode array detector as described elsewhere (27). The lifetime of the substrate-free ferryl species thus produced was short (ϳseconds) at high [H 2 O 2 ] (Ͼ2 mM), as heme was decomposed due to secondary reactions at longer time scales. At low [H 2 O 2 ] (Յ0.2 mM), the lifetime of the ferryl species was longer (ϳminutes). L-Trp binding to hIDO significantly increased the lifetime of the ferryl species. The initial formation of the ferryl species in each reaction was analyzed using PRO-K software (Applied Photophysics, Ltd.). All the reactions were carried out under pseudo firstorder conditions, with the concentration of H 2 O 2 at least 5-fold higher than that of hIDO.
The L-Trp affinity toward the ferryl species was measured by mixing hIDO with a mixture of H 2 O 2 (2 mM)/L-Trp (0 -15 mM) in the stopped-flow instrument. The L-Trp binding curve was obtained by plotting ⌬A (defined by ⌬A 418 nm Ϫ ⌬A 450 nm ) as a function of L-Trp. A binding model with two independent binding sites, each described by the following Equation 1, was used to fit the data by using the Origin 6.1 program (Microcal, Inc., Northampton, MA) (31). Here, K d is the substrate dissociation constant; ⌬A max is the maximum ⌬A.
Peroxidase Activity-The peroxidase reaction of hIDO was determined by adding H 2 O 2 to a mixture of hIDO and ABTS in 100 mM phosphate buffer at pH 7.4 at room temperature (ϳ25°C) in a 1-cm optical cuvette. The reaction was monitored by optical absorption measurements with an Agilent 8453 diode array spectrophotometer (Agilent Technologies, Santa Clara, CA) as a function of time. H 2 O 2 and the ABTS were in large excess with respect to the enzyme (ϳ80 -100 nM, unless otherwise noted) to ensure that their concentrations remain constant during the initial phase of the reactions. The activity of hIDO was determined based on the initial slopes of the formation kinetics of the ABTS radical cation by using ⑀ 735 nm ϭ 15 mM Ϫ1 cm Ϫ1 (40,41). The pH-dependent studies were carried out from pH 4.8 to 10.0. At least three measurements were made and averaged for each condition.
Resonance Raman Measurements-To measure the RR spectra, the 413.1 nm excitation from a krypton ion laser (Spectra Physics, Mountain View, CA) was focused to an ϳ30-m spot on the spinning quartz cell rotating at ϳ6,000 rpm (42). The scattered light, collected at a right angle to the incident laser beam, was focused on the 100-m-wide entrance slit of a 1.25-m Spex spectrometer equipped with a 1200 grooves/mm grating (Horiba Jobin Yvon, Edison, NJ), where it was dispersed and then detected by a liquid nitrogen-cooled CCD detector (Princeton Instruments, Trenton, NJ). A holographic notch filter (Kaiser, Ann Arbor, MI) was used to remove the laser line. The Raman shifts were calibrated by using indene for the 200 -1700 cm Ϫ1 spectral window. Each ferric spectrum was obtained with a laser power of 8 milliwatts and an acquisition time of 1 h. Because of the instability of the ferryl species and its photolability, each ferryl spectrum was obtained by summing 12 individual spectra of fresh samples, each with a 10 -20-s acquisition time and a laser power of ϳ8 milliwatts.

RESULTS AND DISCUSSION
Reaction Kinetics with H 2 O 2 -The reaction of hIDO with H 2 O 2 was initiated in a stopped-flow apparatus and monitored by a photodiode detector. As shown in Fig. 1a, the ferric enzyme has Soret and visible bands at 404 and 499/533 nm, respectively, as well as a charge transfer band at 633 nm, characteristic of a six-coordinate water-bound ferric heme (39). Following the initiation of the reaction with 2 mM H 2 O 2 , the Soret and visible bands gradually shift to 415 and 547/587 nm, respectively. It is concurrent with the reduction in intensity of the 633-nm band. The product is assigned to a compound II (Cmpd-II) type of ferryl species (Fe 4ϩ ϭO 2Ϫ ). As shown in the inset of In the presence of 25 mM L-Trp, the ferric enzyme exhibits the Soret and visible bands at 410 and 540/576 nm, respectively ( Fig. 2a), characteristic for a hydroxide-bound ferric heme (20). The mixing of the ferric L-Trp-bound enzyme with 2 mM H 2 O 2 leads to the shift of the Soret and visible bands to 412 and 542/ 587 nm, indicating the formation of a distinct Cmpd-II type of ferryl species. The distinct maxima of the Soret and visible bands, as well as the significant reduction in their intensities, as compared with those obtained in the absence of substrate, indicate that the electronic properties of the ferryl species are perturbed by substrate binding. As shown in the inset in Fig. 2a, the reaction of the substrate-bound enzyme hIDO, like the substrate-free enzyme, follows single exponential kinetics, with a rate constant of 1.3 s Ϫ1 . However, the observed rate exhibits a hyperbolic relationship, instead of a linear relationship, with [H 2 O 2 ] (Fig. 2b).
Considering the fact that the distal ligand-binding site of the heme is occupied by a hydroxide ion, we hypothesize the following two-step reaction mechanisms.
The reaction was initiated by the dissociation of the hydroxide ion from the ferric heme iron. It was followed by the binding of H 2 O 2 to the heme iron to form the ferric-(hydro)peroxo intermediate (P) that subsequently converts to the Cmpd-II ferryl species. At low [H 2 O 2 ] (Ͻ 0.6 mM), the reaction is rate-limited by the bimolecular binding of H 2 O 2 (Step 2), and hence it exhibits a linear dependence on the [H 2 O 2 ], with a bimolecular rate constant of 9.1 ϫ 10 2 M Ϫ1 s Ϫ1 (see inset in Fig. 2b). At high [H 2 O 2 ], the bimolecular binding of H 2 O 2 to the enzyme is much faster; the reaction is rate-limited by the dissociation of  the heme iron-bound hydroxide (Step 1) with a rate constant of ϳ6 s Ϫ1 (Fig. 2b), accounting for the hyperbolic dependence of the kinetic rate on [H 2 O 2 ]. It is important to note that a similar two-step mechanism is plausibly operative for the reaction of the substrate-free enzyme with H 2 O 2 . However, instead of a hydroxide, the distal heme ligand is a water molecule, which is a considerably weaker ligand; ligand dissociation is hence not rate-limiting under all conditions studied (Fig. 1b).
As summarized in Table 1, L-Trp binding to hIDO leads to ϳ9-fold slower H 2 O 2 -binding rate (9.1 ϫ 10 2 versus 8.4 ϫ 10 3 M Ϫ1 s Ϫ1 ). The substrate-retarded ligand binding is similar to that observed in cyanide-binding reactions, in which prebinding of L-Trp to hIDO leads to ϳ13-fold slower binding rate (31). The data suggest that L-Trp binding induces conformational changes to the protein matrix, thereby restricting ligand entrance and binding. This hypothesis is consistent with recent crystallographic and spectroscopic data showing that the distal heme pocket of the substrate-free hIDO is open and wateraccessible, while that of L-Trp-bound enzyme is more enclosed due to substrate-induced open-to-closed conformational change (18,28).
Thermodynamic Parameters Associated with the H 2 O 2 Reaction-As listed in Table 1, the reaction rate of hIDO (in either substrate-free or L-Trp-bound form) with H 2 O 2 is significantly faster than Mb (ϳ300 M Ϫ1 s Ϫ1 ) but much slower than peroxidases (ϳ1-5 ϫ 10 7 M Ϫ1 s Ϫ1 ), such as horseradish peroxidase (HRP), lactoperoxidase, or cytochrome c peroxidase. To investigate the energetic parameters associated with the hIDO reaction, the reaction was studied as a function of temperature from ϳ2 to 40°C. As shown in Fig. 3, a and b, the observed rate constants obtained from single exponential fits of the kinetic traces for the substrate-free and L-Trp-bound enzyme correlate linearly with [H 2 O 2 ]. The activation enthalpies (⌬H ‡ ) and entropies (⌬S ‡ ) obtained from Eyring plots of the data (Fig. 3c) are summarized in Table 1. The data show that the enthalpic (⌬H ‡ ) and entropic (ϪT⌬S ‡ ) contribution to the activation free energy barrier for the reaction of the substrate-free enzyme are 10.8 and 1.0 kcal mol Ϫ1 , respectively, at room temperature (T ϭ 298 K), whereas those of the L-Trp-bound enzyme are 6.9 and 6.5 kcal mol Ϫ1 , respectively. The data indicate that L-Trp binding to the enzyme lowers the enthalpic barrier by 3.9 kcal mol Ϫ1 ; however, the gain in the enthalpic stabilization is overcompensated by the elevation of the entropic barrier (from 1.0 to 6.5 kcal mol Ϫ1 ). As a result, the activation free energy barrier (⌬G ‡ ) is raised by 1.6 kcal mol Ϫ1 (from 11.8 and 13.4 kcal mol Ϫ1 ). It is conceivable that the open-to-closed conformational change introduced by L-Trp binding leads to favorable intramolecular interactions in the transition state with respect to the reactant (hence a lower ⌬H ‡ ), but at the same time it locks the transition state in a more rigid conformation (hence a more negative ⌬S ‡ because of lower conformational freedom).
As listed in Table 1, the enthalpic barrier for the Mb reaction (ϳ14 kcal mol Ϫ1 ) is significantly higher as compared with the L-Trp-bound hIDO, but the entropic barrier is much lower (Ϫ0.3 kcal mol Ϫ1 ). However, the enthalpic barrier for the peroxidase reactions is ϳ5-6 kcal mol Ϫ1 , similar to that of the   L-Trp-bound enzyme. Although the entropic barriers for the peroxidase reactions have not been reported, the fact that the peroxidase reaction rates are ϳ4 orders of magnitude faster than that of L-Trp-bound hIDO indicates that the entropic contribution to the activation free energy of the peroxidase reactions are insignificant. These data indicate that although the chemical environment of the active site of the L-Trp-bound hIDO is analogous to that of peroxidases, in particular the positive polar distal environment for the heme iron-bound ligand and the imidazolate character of the proximal heme ligand (20), the conformation of hIDO is optimized to carry out the L-Trp dioxygenase reaction by using O 2 as the co-substrate, and hence a high entropic penalty has to be paid to overcome the energy barrier when H 2 O 2 is used as the co-substrate.
In summary, the aforementioned data demonstrate that the reaction of hIDO with H 2 O 2 leads to Cmpd-II-like ferryl adducts regardless of L-Trp binding. In the presence of L-Trp, no oxidation product of L-Trp was observable (data not shown), indicating that neither the heme iron-bound peroxide nor the ferryl adduct is capable of oxidizing L-Trp. Consistent with this observation, Ferry et al. (34) have reported that, by using H 2 O 2 as a co-substrate, hIDO converts melatonin, serotonin, and tryptamine, but not L-Trp, to oxygenated products. The fact that, when O 2 is used as a co-substrate, hIDO is able to oxidize L-Trp, but not melatonin (34), indicates that the oxygen chemistry carried out by hIDO is modulated by the structural and chemical properties of the substrate.
Peroxidase Activity-To confirm the assignment of the ferryl adduct, we examined its activity toward ABTS. ABTS is an azo dye that is widely used as a substrate for activity assays of peroxidases, as it reacts with Cmpd-I and/or Cmpd-II types of ferryl intermediates generated from peroxidase reactions to produce a stable colored product ABTS . ϩ , readily detectable by absorption spectroscopic measurements (40,41). As shown in Fig. 4a, the reaction of hIDO with H 2 O 2 in the presence of ABTS leads to the generation of ABTS . ϩ , as is evident by the characteristic absorption band at ϳ500 -900 nm. The cation radical was not observable in the control experiments without hIDO and/or H 2 O 2 (Fig. 4a). In addition, the amount of radical generated is linearly correlated with [hIDO] (Fig. 4b), confirming that the radical production is directly linked to hIDO. pHdependent studies show that the optimum condition for ABTS . ϩ production is at ϳpH 8.0, which matches the optimum condition for the production of the Cmpd-II-like ferryl species (Fig.  4c), consistent with the scenario that the active species oxidizing ABTS is the ferryl adduct. It is important to note that the data at pH Ͻ5 or Ͼ10 are unavailable, because the enzyme is not stable enough to allow data collection under these conditions.
Steady-state kinetic measurements show that the ABTS . ϩ production rate follows typical Michaelis-Menten behavior as a function of [ABTS], with K m and k cat of 0.061 mM and 61 s Ϫ1 , respectively (Fig. 5a). As summarized in Table 2, the K m value is comparable with that reported for Mb (0.077 mM) but is much lower than that of HRP (0.64 mM). However, the k cat is ϳ140fold higher than that of Mb (0.43 s Ϫ1 ) but is comparable with that of HRP (45.5 s Ϫ1 ). Based on these data, the efficiency of hIDO (k cat /K m ϭ 1.0 ϫ 10 6 M Ϫ1 s Ϫ1 ) is ϳ180and 14-fold higher than that of Mb and HRP, respectively, at neutral pH. Steadystate kinetic measurements of the activity as a function of [H 2 O 2 ] show that the K m and k cat values are 4.0 mM and 85 s Ϫ1 , respectively (Fig. 5b). The k cat is similar to that determined by the data shown in Fig. 5a, confirming the reliability of the measurements. The K m is 4-and 120-fold higher than Mb (0.98 mM) and HRP (0.032 mM), respectively, manifesting the relatively lower affinity of hIDO toward H 2 O 2 . Competition experiments show that L-Trp efficiently inhibits the ABTS . ϩ production (Fig. 5c), with an inhibition constant (K i ) of 0.2 M based on the fitting of the data with a competitive inhibition model described in Equation 2 (43).
Here, V max and K m are the Michaelis-Menten constants associated with the data shown in Fig. 5a. The data indicate that L-Trp competes with ABTS for the active site, thereby inhibiting the reaction. The K i value is significantly lower than the dissociation constant (K d ) of L-Trp from the ferric hIDO (ϳ900 M), suggesting that L-Trp preferentially binds to the ferryl species, instead of the ferric enzyme. To evaluate this hypothesis, the affinity of the ferryl species toward L-Trp was determined (see below).
Affinity of the Ferryl Species toward L-Trp-To determine the L-Trp affinity, the ferryl species was generated by mixing hIDO with a mixture of H 2 O 2 (2 mM)/L-Trp (0 -15 mM) in a stoppedflow system and probed by a photodiode detector. As shown in the inset of Fig. 6b, a peak and trough at 450 and 418 nm, respectively, were observed in the L-Trp-bound minus substrate-free difference spectrum of the ferryl species. The L-Trp binding curve was hence obtained by plotting ⌬A 418 nm Ϫ ⌬A 450 nm as a function of [L-Trp]. The data showed two transitions with K d 1 and K d 2 of 0.3 and 440 M (Fig. 6c), indicating the presence of two L-Trp-binding sites in the ferryl species.
It is well accepted that the steady-state kinetics of hIDO exhibits substrate-inhibition behavior, indicating the presence of two substrate-binding sites, an active site and an inhibitory site (39). The inhibitory site has been identified in the CObound ferrous complex at low temperature (44) and the CNbound ferric adduct at room temperature (31) but not in ligandfree ferric and ferrous states (31,39). The current data demonstrate that the ferryl species also possesses two substrate-binding sites, with K d of 0.3 and 440 M. The K d values are significantly lower than those of the CN-bound ferric enzyme (18 M and 26 mM) (31), suggesting that the highly charged ferryl moiety (i.e. Fe 4ϩ ϭO 2Ϫ ) in the active site of the enzyme promotes L-Trp binding. In any case, the data highlight the fact that L-Trp affinities toward the two binding sites are intricately modulated by the redox state of the heme iron, as

Ferryl Derivatives of hIDO
well as the nature of ligand bound to it. It is noteworthy that the K d 1 value (0.3 M) is in good agreement with the K i (L-Trp), 0.2 M, for the ABTS reaction (Fig. 5c), confirming that L-Trp competes with ABTS for the active substrate-binding site of the enzyme.
It is important to note that the H 2 O 2 binding rate constant of the substrate-free ferric enzyme is 8.4 ϫ 10 3 M Ϫ1 s Ϫ1 (Table 1), whereas the L-Trp binding rate constant of the ligand-free ferric enzyme is 5.5 ϫ 10 3 M Ϫ1 s Ϫ1 (31). Under the conditions employed in this study (i.e. [H 2 O 2 ] ϭ 2 mM), the H 2 O 2 binding rate is ϳ16 s Ϫ1 . In the low [L-Trp] window (Ͻ40 M), the L-Trp binding rate is Ͻ0.22 s Ϫ1 . Hence, H 2 O 2 binding is expected to occur prior to L-Trp binding. This scenario is consistent with the kinetic data shown in Fig. 6b. Specifically, in the presence of 36 M L-Trp, the kinetic trace can be fitted with a single exponential function with a rate of 13.3 s Ϫ1 , which is similar to that of the comparable reaction carried out in the absence of L-Trp (data not shown), confirming that H 2 O 2 binding precedes L-Trp binding (note: prebinding of L-Trp is expected to retard H 2 O 2 binding by ϳ9-fold, see Table 1). On the basis of the data, we hypothesize the following sequential mechanism shown in Reaction 3. REACTION 3 Here Fe 4ϩ ϭO 2Ϫ (1ϫTrp) denotes the ferryl species with one L-Trp bound to it. As the substrate-free ferryl intermediate was not observed during the reaction; the O-O bond cleavage reaction (that leads to the ferryl species) and the subsequent binding of L-Trp to the ferryl species must be much faster than H 2 O 2 binding. On the basis of this analysis, L-Trp binding to the ferryl species is expected to be at least 5-fold faster than 13.3 s Ϫ1 (i.e. Ͼ2 ϫ 10 6 M Ϫ1 s Ϫ1 ), highlighting the fact that L-Trp binding to the ferryl species is at least 400-fold faster than that of the ligand-free ferric enzyme (Ͼ2 ϫ 10 6 versus 5.5 ϫ 10 3 M Ϫ1 s Ϫ1 ).
In the presence of the highest [L-Trp] employed (15 mM), L-Trp binding rate is ϳ83 s Ϫ1 , which is much faster than H 2 O 2 binding (16 s Ϫ1 ), indicating L-Trp binding precedes H 2 O 2 binding as illustrated in Reaction 4. REACTION 4 Here, Fe 3ϩ (1ϫTrp) and P(1ϫTrp) denote the ferric and P species with one L-Trp bound to them, whereas Fe 4ϩ ϭO 2Ϫ (2ϫTrp) stands for the ferryl species with two L-Trp bound to it. Consistent with this scenario, our data show transient population of the 1ϫTrp-bound ferric species, Fe 3ϩ (1x Trp), during the reaction in the presence 15 mM L-Trp (data not shown). As no additional intermediate was observed during the subsequent conversion of the 1ϫTrp-bound ferric species to the 2ϫTrp-bound ferryl species, the data suggest that the O-O bond cleavage reaction and the binding of the second L-Trp to the 1ϫTrp-bound ferryl species are much faster than the preceding reaction step.
The reactions occurring in the medium [L-Trp] window, however, were complicated by the competition between L-Trp and H 2 O 2 binding to the ferric enzyme. Nonetheless, for the determination of the equilibrium K d values, all the spectra of the ferryl species were obtained at the end of the reaction; hence, it is independent of the binding sequence. More importantly, the fact that the spectrum of the 2ϫTrp-bound ferryl species obtained by mixing hIDO with the mixture of H 2 O 2 (2 mM)/L-Trp (15 mM) (Fig. 6a) is fully consistent with that obtained by mixing the mixture of hIDO/L-Trp (25 mM) with H 2 O 2 (2 mM) (Fig. 2a) confirms the reliability of our data.
As shown in Fig. 6a, the optical absorption spectrum of the ferryl derivative of the substrate-free enzyme has Soret and visible bands at 415 and 547/587 nm, respectively. Binding of one molecule of L-Trp to the active binding site causes the shift of these bands to 412 and 542/587 nm, whereas the binding of an additional molecule of L-Trp to the second binding site leads to the reduction of the Soret intensity and changes in the relative intensity of the 542/587 nm bands. The data indicate that the electronic properties of the ferryl species are significantly perturbed by the occupancy of the two substrate-binding sites. Previous kinetic studies of hIDO show that a 412 nm species was populated at 20 s following the initiation of the dioxygenase reaction in the presence of 100 M L-Trp. This intermediate was originally assigned to the L-Trp-bound ferric enzyme (29); however, subsequent studies revealed that the K d of L-Trp for the ferric enzyme is ϳ900 M (31, 39), which is much higher than the 100 M employed for the studies, indicating that the 412 nm species is not the L-Trp-bound ferric enzyme. The similarity of the spectral feature of the reported 412 nm species to the 1ϫTrp-bound ferryl adduct shown in Fig. 6a prompts us to hypothesize that they are the same species, although additional studies are required to fully understand why a stable 1ϫTrpbound ferryl species is populated at the end of the dioxygenase reaction and what the fate of this ferryl species is. Nonetheless, our current data demonstrate that the ferryl species exhibits Ͼ3000-fold higher affinity toward L-Trp as compared with the ferric enzyme (0.3 versus 900 M), supporting the assignment of this 412 nm species as the 1ϫTrp-bound ferryl intermediate.
Steady-state kinetic studies of the dioxygenase reaction of hIDO (31,39) show that the K m (L-Trp) is 15 M, which is considerably lower than the K d (Trp) for the ligand-free ferric (900 M) or ferrous enzyme (400 M). Accordingly, we have proposed that O 2 binding occurs prior to L-Trp binding during the multiple turnover of the dioxygenase reaction under physiological conditions (31,39). Our current finding further supports this hypothesis by demonstrating that, with respect to the ligand-free ferric enzyme, the ferryl species shows Ͼ400-fold faster binding rate (Ͼ2 ϫ 10 6 M Ϫ1 s Ϫ1 versus 5.5 ϫ 10 3 M Ϫ1 s Ϫ1 ) and 3,000-fold higher affinity (K d of 0.3 versus 900 M) toward L-Trp. Similarly, cyanide-bound ferric enzyme has been shown to exhibit an ϳ20-fold faster binding rate and ϳ50-fold higher affinity toward L-Trp with respect to the ligand-free ferric enzyme (31,39), manifesting the fact that ligand binding to the distal heme pocket introduces structural changes to the protein matrix, thereby facilitating substrate binding.
Structural Characterization of the Ferryl Species-The structural properties of the ferryl derivative of hIDO was investigated by RR spectroscopy in the absence or presence of 75 M or 32 mM L-Trp (for the substrate-free, 1ϫTrp-bound or 2ϫTrp-bound state, respectively) and compared with the ligand-free ferric species under comparable conditions.
The high frequency RR data show that substrate-free ferric enzyme (Fig. 7a) has a mixed 6-coordinate high and low spin configuration, as indicated by the 2 / 3 , modes at 1562/1484 and 1581/1511 cm Ϫ1 , respectively, consistent with a waterbound ferric heme. The presence of 75 M L-Trp, which is much lower than the K d for L-Trp (900 M) (39), does not affect the spectrum, whereas the presence 32 mM L-Trp leads to spin transition to a six-coordinate low spin state, as evident by the 2 / 3 modes at 1578/1500 cm Ϫ1 . As reported previously (20), it reflects the deprotonation of the distal water ligand to a hydroxide due to its close proximity to the bound substrate. The low frequency RR spectrum of the substrate-free enzyme (Fig. 8a) exhibits 7 / 8 modes at 676/336 cm Ϫ1 , as well as propionate bending (␦ propionate ) and vinyl bending (␦ vinyl ) modes at 386 and 417 cm Ϫ1 , respectively. L-Trp binding leads to the disappearance of the ␦ propionate mode, consistent with the establishment of the H-bonding interaction between L-Trp and one of the two propionates (18,20,28). It also causes the splitting of the ␦ vinyl mode into two bands at 422 and 435 cm Ϫ1 , suggesting that the H-bonding interaction indirectly leads to conformational changes to the two vinyl groups.
However, the substrate-free ferryl species exhibits 4 and 2 / 3 at 1376 and 1584/1507 cm Ϫ1 , respectively (Fig. 7b), consistent with that reported for the high valent iron oxo heme complex (45)(46)(47)(48). In the low frequency window (Fig. 8b), the 7 / 8 modes at 676/337 cm Ϫ1 are similar to those of the ferric species; however, the ␦ vinyl mode shifts to 419 cm Ϫ1 and the intensity of the ␦ propionate mode diminishes, reflecting the unique conformations of the peripheral groups of the ferryl heme. The binding of one molecule of L-Trp to the enzyme significantly modulates the spectral features. In particular, the 2 / 3 modes shift to 1579/1503 cm Ϫ1 (with the intensity of the 3 mode considerably enhanced), the 8 mode shifts to 332 cm Ϫ1 , and the ␦ vinyl mode splits into two bands at 423 and 433 cm Ϫ1 . In addition, various in-plane asymmetric modes and outof-plane modes are activated, indicating that L-Trp binding to the ferryl species causes the reduction of the in-plane symmetry, thereby introducing out-of-plane distortion of the porphyrin macrocycle of the heme (29,49). Binding of the second L-Trp to the enzyme leads to minor changes in the 770 -820 cm Ϫ1 region.
To identify the Fe 4ϩ ϭO 2Ϫ stretching mode ( FeϭO ) of the ferryl species, H 2 16 O 2 -H 2 18 O 2 isotope substitution experiments were carried out. In general, in the isotope difference spectrum, all the heme modes are cancelled out, and the remaining peak and trough are attributed to 16 O-and 18 Orelated modes, respectively. Intriguingly, the substrate-free spectrum shows two positive peaks at 786 and 810 cm Ϫ1 , in conjunction with a single negative peak at 752 cm Ϫ1 (Fig. 9b). The single negative peak at 752 cm Ϫ1 is assigned to the FeϭO(18) mode of the ferryl species based on its similarity to those of the ferryl derivatives of other heme proteins (45). The FeϭO(16) mode is predicted to be at ϳ787 cm Ϫ1 based on an  isotope shift of 35 cm Ϫ1 estimated for an FeϭO harmonic oscillator. The splitting of this mode into two peaks at 786 and 810 cm Ϫ1 is attributed to Fermi resonance coupling interaction with an intrinsic heme mode. Likewise, the FeϭO(18) mode of the 1ϫTrp-bound species is present at 764 cm Ϫ1 , and based on this frequency, the FeϭO(16) mode is predicted to be at 799 cm Ϫ1 . Again, the FeϭO(16) mode splits into two peaks at 797 and 808 cm Ϫ1 , because of Fermi resonance coupling interaction. It is important to note that the spectral feature of the 1ϫTrp-bound ferryl species is similar to that populated during the turnover of the dioxygenase reaction in the presence of 100 M L-Trp (29), despite the fact that the reported time-resolved spectrum did not resolve the two components of the FeϭO(16) mode because of its significantly lower signal-to-noise ratio. The frequencies of the FeϭO mode of the Cmpd-II type of ferryl derivative of HRP has been found to be at 787 cm Ϫ1 at basic pH, which shifts to 774 cm Ϫ1 at neutral pH (50). The downshift of the frequency is believed to be a result of the weakening of the FeϭO bond due to an H-bond donated to it from its surroundings. Likewise, the FeϭO mode of the ferryl derivative of Mb was reported to be 804 and 790 at pH 8.5 and 4.5, respectively (51). Accordingly, the ϳ12 cm Ϫ1 lower FeϭO mode of the substrate-free hIDO with respect to that of the 1ϫTrp-bound enzyme (787 versus 799 cm Ϫ1 ) suggests that the H-bonding interactions between the ferryl moiety and its environment is relatively stronger in the substrate-free enzyme (hence a weaker FeϭO bond).
The H 2

16
O 2 -H 2 18 O 2 isotope difference spectrum of the 2ϫTrp-bound ferryl species, however, exhibits a single peak and trough at 800 and 761 cm Ϫ1 , respectively, which are assigned to the FeϭO (16) and FeϭO(18) modes, respectively. The isotopic shift of 39 cm Ϫ1 is close to the theoretical value of 35 cm Ϫ1 . The frequency of the FeϭO mode is similar to that of the 1ϫTrp-bound species, but its spectral width is 5 cm Ϫ1 broader (11 versus 16 cm Ϫ1 ), indicating that the binding of an additional L-Trp to the second substrate-binding site increases the conformational freedom of the protein matrix surrounding it. The fact that the Fermi coupling interaction is absent suggests that the second L-Trp binding perturbs the symmetry of the Fe 4ϩ ϭO 2Ϫ moiety with respect to the heme. It is noteworthy that the FeϭO mode of the 2ϫTrp bound ferryl species is analogous to that reported by Yanagisawa et al. (49), in the presence of 10 mM L-Trp.

CONCLUSIONS
This study demonstrates that the reaction of the ferric hIDO with H 2 O 2 leads to the formation of the Cmpd-II type of ferryl adducts. With respect to the ligand-free ferric enzyme, the ferryl intermediate exhibits ϳ400-fold faster binding rate (Ͼ2 ϫ 10 6 M Ϫ1 s Ϫ1 versus 5.5 ϫ 10 3 M Ϫ1 s Ϫ1 ) and 3000-fold higher affinity (K d of 0.3 versus 900 M) toward L-Trp, manifesting the fact that ligand binding in hIDO introduces conformational changes to the protein matrix, thereby promoting L-Trp binding. On the contrary, L-Trp binding to hIDO reduces the H 2 O 2 reaction rate from 8.4 ϫ 10 3 to 9.1 ϫ 10 2 M Ϫ1 s Ϫ1 , indicating that substrate binding also introduces structural changes to the enzyme, thereby restricting ligand access and binding to the active site. Temperature-dependent studies reveal that the free energy barrier for the H 2 O 2 reaction at room temperature is 1.6 kcal mol Ϫ1 higher for the L-Trp-bound enzyme with respect to the substrate-free enzyme (13.4 versus 11.8 kcal mol Ϫ1 ), due to unfavorable entropic factors, highlighting the fact that the structure of hIDO is not optimized for utilizing H 2 O 2 as the co-substrate for the oxygenation reaction of L-Trp. Consistently, activity studies demonstrate that the ferryl species is capable of oxidizing ABTS, but not L-Trp; moreover, L-Trp effectively inhibits the ABTS activity via competitive binding to the active site. L-Trp concentration-dependent studies show that the ferryl species exhibits two substrate-binding sites with K d of 0.3 and 440 M, in good agreement with earlier data showing that hIDO possesses two substrate-binding sites (39). The fact that the second substrate-binding site is detectable in the ferryl species, the CN-bound ferric adduct (31), and CO-bound ferrous complex (at low temperature) (44) but not in the ligandfree ferric and ferrous states (31,39) suggests that ligand binding to the heme iron in the active site induces allosteric structural changes to the inhibitory site, thereby affecting its affinity toward L-Trp. RR data show that the electronic properties of the ferryl species are significantly perturbed by the occupancy of the two substrate-binding sites.
Taken together the data provide the first detailed characterization of the ferryl derivatives of hIDO, with or without the substrate-binding site(s) occupied by L-Trp. They also highlight the intricate interplay between ligand and substrate binding in hIDO, which is of potential physiological significance. The    (56). However, the K m (L-Trp) is 15 M (39), 25-fold lower than the K d (L-Trp) of the ligand-free ferrous enzyme (400 M) (31). These data support the scenario that hIDO binds O 2 prior to L-Trp binding during the multiple turnovers under physiological conditions. It also provides a rational explanation for the finding that the enzyme turns over efficiently in the presence of Ͻ50 M L-Trp (29,56), despite the fact that under this condition only Ͻ5% ferrous enzyme is in the L-Trp-bound form (based on the K d (L-Trp) of 400 M) (31). The preferential binding of O 2 prior to substrate binding is unique as compared with the other heme dioxygenase, TDO, and monooxygenase type of enzymes (such as P450s), in which substrate-binding precedes O 2 binding (1,11). Previous studies showed that, in TDO, substrate binding promotes ligand binding (1), and it is probably beneficial for TDO to bind substrate prior to O 2 . However, in P450s, the sequential binding of substrate and O 2 is believed to be part of a cellular protection mechanism to prevent the release of cytotoxic superoxide under substrate-deficient conditions. For hIDO, the preferential binding of O 2 prior to substrate binding is plausibly important for ensuring the fast turnover of the enzyme, especially under low [L-Trp] conditions.